From a8768a134518e406d41799a3594aeb74e0889cf7 Mon Sep 17 00:00:00 2001 From: Li Nan Date: Wed, 8 May 2024 17:20:53 +0800 Subject: [PATCH 001/209] md: do not delete safemode_timer in mddev_suspend The deletion of safemode_timer in mddev_suspend() is redundant and potentially harmful now. If timer is about to be woken up but gets deleted, 'in_sync' will remain 0 until the next write, causing array to stay in the 'active' state instead of transitioning to 'clean'. Commit 0d9f4f135eb6 ("MD: Add del_timer_sync to mddev_suspend (fix nasty panic))" introduced this deletion for dm, because if timer fired after dm is destroyed, the resource which the timer depends on might have been freed. However, commit 0dd84b319352 ("md: call __md_stop_writes in md_stop") added __md_stop_writes() to md_stop(), which is called before freeing resource. Timer is deleted in __md_stop_writes(), and the origin issue is resolved. Therefore, delete safemode_timer can be removed safely now. Signed-off-by: Li Nan Reviewed-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240508092053.1447930-1-linan666@huaweicloud.com --- drivers/md/md.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index aff9118ff69750..09c55d9a2c5424 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -479,7 +479,6 @@ int mddev_suspend(struct mddev *mddev, bool interruptible) */ WRITE_ONCE(mddev->suspended, mddev->suspended + 1); - del_timer_sync(&mddev->safemode_timer); /* restrict memory reclaim I/O during raid array is suspend */ mddev->noio_flag = memalloc_noio_save(); From 03e792eaf18ec2e93e2c623f9f1a4bdb97fe4126 Mon Sep 17 00:00:00 2001 From: Li Nan Date: Sun, 26 May 2024 02:52:56 +0800 Subject: [PATCH 002/209] md: change the return value type of md_write_start to void Commit cc27b0c78c79 ("md: fix deadlock between mddev_suspend() and md_write_start()") aborted md_write_start() with false when mddev is suspended, which fixed a deadlock if calling mddev_suspend() with holding reconfig_mutex(). Since mddev_suspend() now includes lockdep_assert_not_held(), it no longer holds the reconfig_mutex. This makes previous abort unnecessary. Now, remove unnecessary abort and change function return value to void. Signed-off-by: Li Nan Reviewed-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240525185257.3896201-2-linan666@huaweicloud.com --- drivers/md/md.c | 14 ++++---------- drivers/md/md.h | 2 +- drivers/md/raid1.c | 3 +-- drivers/md/raid10.c | 3 +-- drivers/md/raid5.c | 3 +-- 5 files changed, 8 insertions(+), 17 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index 09c55d9a2c5424..6bac20e82ff02f 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -8640,12 +8640,12 @@ EXPORT_SYMBOL(md_done_sync); * A return value of 'false' means that the write wasn't recorded * and cannot proceed as the array is being suspend. */ -bool md_write_start(struct mddev *mddev, struct bio *bi) +void md_write_start(struct mddev *mddev, struct bio *bi) { int did_change = 0; if (bio_data_dir(bi) != WRITE) - return true; + return; BUG_ON(mddev->ro == MD_RDONLY); if (mddev->ro == MD_AUTO_READ) { @@ -8678,15 +8678,9 @@ bool md_write_start(struct mddev *mddev, struct bio *bi) if (did_change) sysfs_notify_dirent_safe(mddev->sysfs_state); if (!mddev->has_superblocks) - return true; + return; wait_event(mddev->sb_wait, - !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags) || - is_md_suspended(mddev)); - if (test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)) { - percpu_ref_put(&mddev->writes_pending); - return false; - } - return true; + !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)); } EXPORT_SYMBOL(md_write_start); diff --git a/drivers/md/md.h b/drivers/md/md.h index ca085ecad50449..487582058f7417 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -785,7 +785,7 @@ extern void md_unregister_thread(struct mddev *mddev, struct md_thread __rcu **t extern void md_wakeup_thread(struct md_thread __rcu *thread); extern void md_check_recovery(struct mddev *mddev); extern void md_reap_sync_thread(struct mddev *mddev); -extern bool md_write_start(struct mddev *mddev, struct bio *bi); +extern void md_write_start(struct mddev *mddev, struct bio *bi); extern void md_write_inc(struct mddev *mddev, struct bio *bi); extern void md_write_end(struct mddev *mddev); extern void md_done_sync(struct mddev *mddev, int blocks, int ok); diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 7b8a71ca66dde0..0d80ff471c73dd 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -1687,8 +1687,7 @@ static bool raid1_make_request(struct mddev *mddev, struct bio *bio) if (bio_data_dir(bio) == READ) raid1_read_request(mddev, bio, sectors, NULL); else { - if (!md_write_start(mddev,bio)) - return false; + md_write_start(mddev,bio); raid1_write_request(mddev, bio, sectors); } return true; diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index a4556d2e46bf95..f8d7c02c6ed561 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -1836,8 +1836,7 @@ static bool raid10_make_request(struct mddev *mddev, struct bio *bio) && md_flush_request(mddev, bio)) return true; - if (!md_write_start(mddev, bio)) - return false; + md_write_start(mddev, bio); if (unlikely(bio_op(bio) == REQ_OP_DISCARD)) if (!raid10_handle_discard(mddev, bio)) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 2bd1ce9b39226a..a84389311dd1ea 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -6078,8 +6078,7 @@ static bool raid5_make_request(struct mddev *mddev, struct bio * bi) ctx.do_flush = bi->bi_opf & REQ_PREFLUSH; } - if (!md_write_start(mddev, bi)) - return false; + md_write_start(mddev, bi); /* * If array is degraded, better not do chunk aligned read because * later we might have to read it again in order to reconstruct From 611d5cbc0b35a752e657a83eebadf40d814d006b Mon Sep 17 00:00:00 2001 From: Li Nan Date: Sun, 26 May 2024 02:52:57 +0800 Subject: [PATCH 003/209] md: fix deadlock between mddev_suspend and flush bio Deadlock occurs when mddev is being suspended while some flush bio is in progress. It is a complex issue. T1. the first flush is at the ending stage, it clears 'mddev->flush_bio' and tries to submit data, but is blocked because mddev is suspended by T4. T2. the second flush sets 'mddev->flush_bio', and attempts to queue md_submit_flush_data(), which is already running (T1) and won't execute again if on the same CPU as T1. T3. the third flush inc active_io and tries to flush, but is blocked because 'mddev->flush_bio' is not NULL (set by T2). T4. mddev_suspend() is called and waits for active_io dec to 0 which is inc by T3. T1 T2 T3 T4 (flush 1) (flush 2) (third 3) (suspend) md_submit_flush_data mddev->flush_bio = NULL; . . md_flush_request . mddev->flush_bio = bio . queue submit_flushes . . . . md_handle_request . . active_io + 1 . . md_flush_request . . wait !mddev->flush_bio . . . . mddev_suspend . . wait !active_io . . . submit_flushes . queue_work md_submit_flush_data . //md_submit_flush_data is already running (T1) . md_handle_request wait resume The root issue is non-atomic inc/dec of active_io during flush process. active_io is dec before md_submit_flush_data is queued, and inc soon after md_submit_flush_data() run. md_flush_request active_io + 1 submit_flushes active_io - 1 md_submit_flush_data md_handle_request active_io + 1 make_request active_io - 1 If active_io is dec after md_handle_request() instead of within submit_flushes(), make_request() can be called directly intead of md_handle_request() in md_submit_flush_data(), and active_io will only inc and dec once in the whole flush process. Deadlock will be fixed. Additionally, the only difference between fixing the issue and before is that there is no return error handling of make_request(). But after previous patch cleaned md_write_start(), make_requst() only return error in raid5_make_request() by dm-raid, see commit 41425f96d7aa ("dm-raid456, md/raid456: fix a deadlock for dm-raid456 while io concurrent with reshape)". Since dm always splits data and flush operation into two separate io, io size of flush submitted by dm always is 0, make_request() will not be called in md_submit_flush_data(). To prevent future modifications from introducing issues, add WARN_ON to ensure make_request() no error is returned in this context. Fixes: fa2bbff7b0b4 ("md: synchronize flush io with array reconfiguration") Signed-off-by: Li Nan Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240525185257.3896201-3-linan666@huaweicloud.com --- drivers/md/md.c | 26 +++++++++++++++----------- 1 file changed, 15 insertions(+), 11 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index 6bac20e82ff02f..e204e36e317081 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -549,13 +549,9 @@ static void md_end_flush(struct bio *bio) rdev_dec_pending(rdev, mddev); - if (atomic_dec_and_test(&mddev->flush_pending)) { - /* The pair is percpu_ref_get() from md_flush_request() */ - percpu_ref_put(&mddev->active_io); - + if (atomic_dec_and_test(&mddev->flush_pending)) /* The pre-request flush has finished */ queue_work(md_wq, &mddev->flush_work); - } } static void md_submit_flush_data(struct work_struct *ws); @@ -586,12 +582,8 @@ static void submit_flushes(struct work_struct *ws) rcu_read_lock(); } rcu_read_unlock(); - if (atomic_dec_and_test(&mddev->flush_pending)) { - /* The pair is percpu_ref_get() from md_flush_request() */ - percpu_ref_put(&mddev->active_io); - + if (atomic_dec_and_test(&mddev->flush_pending)) queue_work(md_wq, &mddev->flush_work); - } } static void md_submit_flush_data(struct work_struct *ws) @@ -616,8 +608,20 @@ static void md_submit_flush_data(struct work_struct *ws) bio_endio(bio); } else { bio->bi_opf &= ~REQ_PREFLUSH; - md_handle_request(mddev, bio); + + /* + * make_requst() will never return error here, it only + * returns error in raid5_make_request() by dm-raid. + * Since dm always splits data and flush operation into + * two separate io, io size of flush submitted by dm + * always is 0, make_request() will not be called here. + */ + if (WARN_ON_ONCE(!mddev->pers->make_request(mddev, bio))) + bio_io_error(bio);; } + + /* The pair is percpu_ref_get() from md_flush_request() */ + percpu_ref_put(&mddev->active_io); } /* From acc6680af28696a037ede62867e731841d4454c2 Mon Sep 17 00:00:00 2001 From: Li Nan Date: Wed, 29 May 2024 04:31:49 +0800 Subject: [PATCH 004/209] md: make md_flush_request() more readable Setting bio to NULL and checking 'if(!bio)' is redundant and looks strange, just consolidate them into one condition. There are no functional changes. Suggested-by: Christoph Hellwig Signed-off-by: Li Nan Reviewed-by: Christoph Hellwig Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240528203149.2383260-1-linan666@huaweicloud.com --- drivers/md/md.c | 28 +++++++++++++--------------- 1 file changed, 13 insertions(+), 15 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index e204e36e317081..b9b15aa79496fb 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -657,24 +657,22 @@ bool md_flush_request(struct mddev *mddev, struct bio *bio) WARN_ON(percpu_ref_is_zero(&mddev->active_io)); percpu_ref_get(&mddev->active_io); mddev->flush_bio = bio; - bio = NULL; - } - spin_unlock_irq(&mddev->lock); - - if (!bio) { + spin_unlock_irq(&mddev->lock); INIT_WORK(&mddev->flush_work, submit_flushes); queue_work(md_wq, &mddev->flush_work); - } else { - /* flush was performed for some other bio while we waited. */ - if (bio->bi_iter.bi_size == 0) - /* an empty barrier - all done */ - bio_endio(bio); - else { - bio->bi_opf &= ~REQ_PREFLUSH; - return false; - } + return true; } - return true; + + /* flush was performed for some other bio while we waited. */ + spin_unlock_irq(&mddev->lock); + if (bio->bi_iter.bi_size == 0) { + /* pure flush without data - all done */ + bio_endio(bio); + return true; + } + + bio->bi_opf &= ~REQ_PREFLUSH; + return false; } EXPORT_SYMBOL(md_flush_request); From 35f20acaa3585f25f8356da0ee6bc143e0256522 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 4 Jun 2024 19:25:28 +0200 Subject: [PATCH 005/209] md/raid0: don't free conf on raid0_run failure The core md code calls the ->free method which already frees conf. Fixes: 0c031fd37f69 ("md: Move alloc/free acct bioset in to personality") Signed-off-by: Christoph Hellwig Reviewed-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240604172607.3185916-2-hch@lst.de --- drivers/md/raid0.c | 21 +++++---------------- 1 file changed, 5 insertions(+), 16 deletions(-) diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c index c5d4aeb68404c9..81c01347cd24e6 100644 --- a/drivers/md/raid0.c +++ b/drivers/md/raid0.c @@ -365,18 +365,13 @@ static sector_t raid0_size(struct mddev *mddev, sector_t sectors, int raid_disks return array_sectors; } -static void free_conf(struct mddev *mddev, struct r0conf *conf) -{ - kfree(conf->strip_zone); - kfree(conf->devlist); - kfree(conf); -} - static void raid0_free(struct mddev *mddev, void *priv) { struct r0conf *conf = priv; - free_conf(mddev, conf); + kfree(conf->strip_zone); + kfree(conf->devlist); + kfree(conf); } static int raid0_set_limits(struct mddev *mddev) @@ -415,7 +410,7 @@ static int raid0_run(struct mddev *mddev) if (!mddev_is_dm(mddev)) { ret = raid0_set_limits(mddev); if (ret) - goto out_free_conf; + return ret; } /* calculate array device size */ @@ -427,13 +422,7 @@ static int raid0_run(struct mddev *mddev) dump_zones(mddev); - ret = md_integrity_register(mddev); - if (ret) - goto out_free_conf; - return 0; -out_free_conf: - free_conf(mddev, conf); - return ret; + return md_integrity_register(mddev); } /* From 17f91ac0843b50462a9c9c8f18df962338bd3db2 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 4 Jun 2024 19:25:29 +0200 Subject: [PATCH 006/209] md/raid1: don't free conf on raid0_run failure The core md code calls the ->free method which already frees conf. Fixes: 07f1a6850c5d ("md/raid1: fail run raid1 array when active disk less than one") Signed-off-by: Christoph Hellwig Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240604172607.3185916-3-hch@lst.de --- drivers/md/raid1.c | 14 +++----------- 1 file changed, 3 insertions(+), 11 deletions(-) diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 0d80ff471c73dd..3d54f30112a0e8 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -3203,7 +3203,6 @@ static int raid1_set_limits(struct mddev *mddev) return queue_limits_set(mddev->gendisk->queue, &lim); } -static void raid1_free(struct mddev *mddev, void *priv); static int raid1_run(struct mddev *mddev) { struct r1conf *conf; @@ -3237,7 +3236,7 @@ static int raid1_run(struct mddev *mddev) if (!mddev_is_dm(mddev)) { ret = raid1_set_limits(mddev); if (ret) - goto abort; + return ret; } mddev->degraded = 0; @@ -3251,8 +3250,7 @@ static int raid1_run(struct mddev *mddev) */ if (conf->raid_disks - mddev->degraded < 1) { md_unregister_thread(mddev, &conf->thread); - ret = -EINVAL; - goto abort; + return -EINVAL; } if (conf->raid_disks - mddev->degraded == 1) @@ -3276,14 +3274,8 @@ static int raid1_run(struct mddev *mddev) md_set_array_sectors(mddev, raid1_size(mddev, 0, 0)); ret = md_integrity_register(mddev); - if (ret) { + if (ret) md_unregister_thread(mddev, &mddev->thread); - goto abort; - } - return 0; - -abort: - raid1_free(mddev, conf); return ret; } From ab99a87542f194f28e2364a42afbf9fb48b1c724 Mon Sep 17 00:00:00 2001 From: Ofir Gal Date: Fri, 7 Jun 2024 10:27:44 +0300 Subject: [PATCH 007/209] md/md-bitmap: fix writing non bitmap pages __write_sb_page() rounds up the io size to the optimal io size if it doesn't exceed the data offset, but it doesn't check the final size exceeds the bitmap length. For example: page count - 1 page size - 4K data offset - 1M optimal io size - 256K The final io size would be 256K (64 pages) but md_bitmap_storage_alloc() allocated 1 page, the IO would write 1 valid page and 63 pages that happens to be allocated afterwards. This leaks memory to the raid device superblock. This issue caused a data transfer failure in nvme-tcp. The network drivers checks the first page of an IO with sendpage_ok(), it returns true if the page isn't a slabpage and refcount >= 1. If the page !sendpage_ok() the network driver disables MSG_SPLICE_PAGES. As of now the network layer assumes all the pages of the IO are sendpage_ok() when MSG_SPLICE_PAGES is on. The bitmap pages aren't slab pages, the first page of the IO is sendpage_ok(), but the additional pages that happens to be allocated after the bitmap pages might be !sendpage_ok(). That cause skb_splice_from_iter() to stop the data transfer, in the case below it hangs 'mdadm --create'. The bug is reproducible, in order to reproduce we need nvme-over-tcp controllers with optimal IO size bigger than PAGE_SIZE. Creating a raid with bitmap over those devices reproduces the bug. In order to simulate large optimal IO size you can use dm-stripe with a single device. Script to reproduce the issue on top of brd devices using dm-stripe is attached below (will be added to blktest). I have added some logs to test the theory: ... md: created bitmap (1 pages) for device md127 __write_sb_page before md_super_write offset: 16, size: 262144. pfn: 0x53ee === __write_sb_page before md_super_write. logging pages === pfn: 0x53ee, slab: 0 <-- the only page that allocated for the bitmap pfn: 0x53ef, slab: 1 pfn: 0x53f0, slab: 0 pfn: 0x53f1, slab: 0 pfn: 0x53f2, slab: 0 pfn: 0x53f3, slab: 1 ... nvme_tcp: sendpage_ok - pfn: 0x53ee, len: 262144, offset: 0 skbuff: before sendpage_ok() - pfn: 0x53ee skbuff: before sendpage_ok() - pfn: 0x53ef WARNING at net/core/skbuff.c:6848 skb_splice_from_iter+0x142/0x450 skbuff: !sendpage_ok - pfn: 0x53ef. is_slab: 1, page_count: 1 ... Cc: stable@vger.kernel.org Reviewed-by: Christoph Hellwig Signed-off-by: Ofir Gal Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240607072748.3182199-1-ofir.gal@volumez.com --- drivers/md/md-bitmap.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/md/md-bitmap.c b/drivers/md/md-bitmap.c index 0a2d37eb38ef92..08232d8dc815ee 100644 --- a/drivers/md/md-bitmap.c +++ b/drivers/md/md-bitmap.c @@ -227,6 +227,8 @@ static int __write_sb_page(struct md_rdev *rdev, struct bitmap *bitmap, struct block_device *bdev; struct mddev *mddev = bitmap->mddev; struct bitmap_storage *store = &bitmap->storage; + unsigned int bitmap_limit = (bitmap->storage.file_pages - pg_index) << + PAGE_SHIFT; loff_t sboff, offset = mddev->bitmap_info.offset; sector_t ps = pg_index * PAGE_SIZE / SECTOR_SIZE; unsigned int size = PAGE_SIZE; @@ -269,11 +271,9 @@ static int __write_sb_page(struct md_rdev *rdev, struct bitmap *bitmap, if (size == 0) /* bitmap runs in to data */ return -EINVAL; - } else { - /* DATA METADATA BITMAP - no problems */ } - md_super_write(mddev, rdev, sboff + ps, (int) size, page); + md_super_write(mddev, rdev, sboff + ps, (int)min(size, bitmap_limit), page); return 0; } From 0476d09c36a845757dd9fd1c80fbbf45b0faeb3c Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:40 +0800 Subject: [PATCH 008/209] md: rearrange recovery_flags Currently there are lots of flags with the same confusing prefix "MD_REOCVERY_", and there are two main types of flags, sync thread runnng status, I prefer prefix "SYNC_THREAD_", and sync thread action, I perfer prefix "SYNC_ACTION_". For now, rearrange and update comment to improve code readability, there are no functional changes. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-2-yukuai1@huaweicloud.com --- drivers/md/md.h | 52 ++++++++++++++++++++++++++++++++++++------------- 1 file changed, 38 insertions(+), 14 deletions(-) diff --git a/drivers/md/md.h b/drivers/md/md.h index 487582058f7417..1ee129c6f98ff5 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -551,22 +551,46 @@ struct mddev { }; enum recovery_flags { + /* flags for sync thread running status */ + + /* + * set when one of sync action is set and new sync thread need to be + * registered, or just add/remove spares from conf. + */ + MD_RECOVERY_NEEDED, + /* sync thread is running, or about to be started */ + MD_RECOVERY_RUNNING, + /* sync thread needs to be aborted for some reason */ + MD_RECOVERY_INTR, + /* sync thread is done and is waiting to be unregistered */ + MD_RECOVERY_DONE, + /* running sync thread must abort immediately, and not restart */ + MD_RECOVERY_FROZEN, + /* waiting for pers->start() to finish */ + MD_RECOVERY_WAIT, + /* interrupted because io-error */ + MD_RECOVERY_ERROR, + + /* flags determines sync action */ + + /* if just this flag is set, action is resync. */ + MD_RECOVERY_SYNC, + /* + * paired with MD_RECOVERY_SYNC, if MD_RECOVERY_CHECK is not set, + * action is repair, means user requested resync. + */ + MD_RECOVERY_REQUESTED, /* - * If neither SYNC or RESHAPE are set, then it is a recovery. + * paired with MD_RECOVERY_SYNC and MD_RECOVERY_REQUESTED, action is + * check. */ - MD_RECOVERY_RUNNING, /* a thread is running, or about to be started */ - MD_RECOVERY_SYNC, /* actually doing a resync, not a recovery */ - MD_RECOVERY_RECOVER, /* doing recovery, or need to try it. */ - MD_RECOVERY_INTR, /* resync needs to be aborted for some reason */ - MD_RECOVERY_DONE, /* thread is done and is waiting to be reaped */ - MD_RECOVERY_NEEDED, /* we might need to start a resync/recover */ - MD_RECOVERY_REQUESTED, /* user-space has requested a sync (used with SYNC) */ - MD_RECOVERY_CHECK, /* user-space request for check-only, no repair */ - MD_RECOVERY_RESHAPE, /* A reshape is happening */ - MD_RECOVERY_FROZEN, /* User request to abort, and not restart, any action */ - MD_RECOVERY_ERROR, /* sync-action interrupted because io-error */ - MD_RECOVERY_WAIT, /* waiting for pers->start() to finish */ - MD_RESYNCING_REMOTE, /* remote node is running resync thread */ + MD_RECOVERY_CHECK, + /* recovery, or need to try it */ + MD_RECOVERY_RECOVER, + /* reshape */ + MD_RECOVERY_RESHAPE, + /* remote node is running resync thread */ + MD_RESYNCING_REMOTE, }; enum md_ro_state { From a85aa09da2f2773c685310666ef09f935ff68a45 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:41 +0800 Subject: [PATCH 009/209] md: add a new enum type sync_action In order to make code related to sync_thread cleaner in following patches, also add detail comment about each sync action. And also prepare to remove the related recovery_flags in the fulture. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-3-yukuai1@huaweicloud.com --- drivers/md/md.h | 57 ++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 56 insertions(+), 1 deletion(-) diff --git a/drivers/md/md.h b/drivers/md/md.h index 1ee129c6f98ff5..e5001d39c82d4b 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -34,6 +34,61 @@ */ #define MD_FAILFAST (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT) +/* Status of sync thread. */ +enum sync_action { + /* + * Represent by MD_RECOVERY_SYNC, start when: + * 1) after assemble, sync data from first rdev to other copies, this + * must be done first before other sync actions and will only execute + * once; + * 2) resize the array(notice that this is not reshape), sync data for + * the new range; + */ + ACTION_RESYNC, + /* + * Represent by MD_RECOVERY_RECOVER, start when: + * 1) for new replacement, sync data based on the replace rdev or + * available copies from other rdev; + * 2) for new member disk while the array is degraded, sync data from + * other rdev; + * 3) reassemble after power failure or re-add a hot removed rdev, sync + * data from first rdev to other copies based on bitmap; + */ + ACTION_RECOVER, + /* + * Represent by MD_RECOVERY_SYNC | MD_RECOVERY_REQUESTED | + * MD_RECOVERY_CHECK, start when user echo "check" to sysfs api + * sync_action, used to check if data copies from differenct rdev are + * the same. The number of mismatch sectors will be exported to user + * by sysfs api mismatch_cnt; + */ + ACTION_CHECK, + /* + * Represent by MD_RECOVERY_SYNC | MD_RECOVERY_REQUESTED, start when + * user echo "repair" to sysfs api sync_action, usually paired with + * ACTION_CHECK, used to force syncing data once user found that there + * are inconsistent data, + */ + ACTION_REPAIR, + /* + * Represent by MD_RECOVERY_RESHAPE, start when new member disk is added + * to the conf, notice that this is different from spares or + * replacement; + */ + ACTION_RESHAPE, + /* + * Represent by MD_RECOVERY_FROZEN, can be set by sysfs api sync_action + * or internal usage like setting the array read-only, will forbid above + * actions. + */ + ACTION_FROZEN, + /* + * All above actions don't match. + */ + ACTION_IDLE, + NR_SYNC_ACTIONS, +}; + /* * The struct embedded in rdev is used to serialize IO. */ @@ -571,7 +626,7 @@ enum recovery_flags { /* interrupted because io-error */ MD_RECOVERY_ERROR, - /* flags determines sync action */ + /* flags determines sync action, see details in enum sync_action */ /* if just this flag is set, action is resync. */ MD_RECOVERY_SYNC, From e792a4c2156a392d8126bf0496f74407a21a8824 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:42 +0800 Subject: [PATCH 010/209] md: add new helpers for sync_action The new helpers will get current sync_action of the array, will be used in later patches to make code cleaner. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-4-yukuai1@huaweicloud.com --- drivers/md/md.c | 79 +++++++++++++++++++++++++++++++++++++++++++++++++ drivers/md/md.h | 3 ++ 2 files changed, 82 insertions(+) diff --git a/drivers/md/md.c b/drivers/md/md.c index b9b15aa79496fb..4ce8d164cde97d 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -69,6 +69,16 @@ #include "md-bitmap.h" #include "md-cluster.h" +static const char *action_name[NR_SYNC_ACTIONS] = { + [ACTION_RESYNC] = "resync", + [ACTION_RECOVER] = "recover", + [ACTION_CHECK] = "check", + [ACTION_REPAIR] = "repair", + [ACTION_RESHAPE] = "reshape", + [ACTION_FROZEN] = "frozen", + [ACTION_IDLE] = "idle", +}; + /* pers_list is a list of registered personalities protected by pers_lock. */ static LIST_HEAD(pers_list); static DEFINE_SPINLOCK(pers_lock); @@ -4868,6 +4878,75 @@ metadata_store(struct mddev *mddev, const char *buf, size_t len) static struct md_sysfs_entry md_metadata = __ATTR_PREALLOC(metadata_version, S_IRUGO|S_IWUSR, metadata_show, metadata_store); +enum sync_action md_sync_action(struct mddev *mddev) +{ + unsigned long recovery = mddev->recovery; + + /* + * frozen has the highest priority, means running sync_thread will be + * stopped immediately, and no new sync_thread can start. + */ + if (test_bit(MD_RECOVERY_FROZEN, &recovery)) + return ACTION_FROZEN; + + /* + * read-only array can't register sync_thread, and it can only + * add/remove spares. + */ + if (!md_is_rdwr(mddev)) + return ACTION_IDLE; + + /* + * idle means no sync_thread is running, and no new sync_thread is + * requested. + */ + if (!test_bit(MD_RECOVERY_RUNNING, &recovery) && + !test_bit(MD_RECOVERY_NEEDED, &recovery)) + return ACTION_IDLE; + + if (test_bit(MD_RECOVERY_RESHAPE, &recovery) || + mddev->reshape_position != MaxSector) + return ACTION_RESHAPE; + + if (test_bit(MD_RECOVERY_RECOVER, &recovery)) + return ACTION_RECOVER; + + if (test_bit(MD_RECOVERY_SYNC, &recovery)) { + /* + * MD_RECOVERY_CHECK must be paired with + * MD_RECOVERY_REQUESTED. + */ + if (test_bit(MD_RECOVERY_CHECK, &recovery)) + return ACTION_CHECK; + if (test_bit(MD_RECOVERY_REQUESTED, &recovery)) + return ACTION_REPAIR; + return ACTION_RESYNC; + } + + /* + * MD_RECOVERY_NEEDED or MD_RECOVERY_RUNNING is set, however, no + * sync_action is specified. + */ + return ACTION_IDLE; +} + +enum sync_action md_sync_action_by_name(const char *page) +{ + enum sync_action action; + + for (action = 0; action < NR_SYNC_ACTIONS; ++action) { + if (cmd_match(page, action_name[action])) + return action; + } + + return NR_SYNC_ACTIONS; +} + +const char *md_sync_action_name(enum sync_action action) +{ + return action_name[action]; +} + static ssize_t action_show(struct mddev *mddev, char *page) { diff --git a/drivers/md/md.h b/drivers/md/md.h index e5001d39c82d4b..88add162b08ea0 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -864,6 +864,9 @@ extern void md_unregister_thread(struct mddev *mddev, struct md_thread __rcu **t extern void md_wakeup_thread(struct md_thread __rcu *thread); extern void md_check_recovery(struct mddev *mddev); extern void md_reap_sync_thread(struct mddev *mddev); +extern enum sync_action md_sync_action(struct mddev *mddev); +extern enum sync_action md_sync_action_by_name(const char *page); +extern const char *md_sync_action_name(enum sync_action action); extern void md_write_start(struct mddev *mddev, struct bio *bi); extern void md_write_inc(struct mddev *mddev, struct bio *bi); extern void md_write_end(struct mddev *mddev); From 207c5656c33d56b3759d0876b68fa56cb56e5c51 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:43 +0800 Subject: [PATCH 011/209] md: factor out helper to start reshape from action_store() There are no functional changes, just to make code cleaner and prepare for following refactor. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-5-yukuai1@huaweicloud.com --- drivers/md/md.c | 65 +++++++++++++++++++++++++++++++------------------ 1 file changed, 41 insertions(+), 24 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index 4ce8d164cde97d..b34ae9fbd24671 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -5071,6 +5071,45 @@ static void frozen_sync_thread(struct mddev *mddev) mutex_unlock(&mddev->sync_mutex); } +static int mddev_start_reshape(struct mddev *mddev) +{ + int ret; + + if (mddev->pers->start_reshape == NULL) + return -EINVAL; + + ret = mddev_lock(mddev); + if (ret) + return ret; + + if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) { + mddev_unlock(mddev); + return -EBUSY; + } + + if (mddev->reshape_position == MaxSector || + mddev->pers->check_reshape == NULL || + mddev->pers->check_reshape(mddev)) { + clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); + ret = mddev->pers->start_reshape(mddev); + if (ret) { + mddev_unlock(mddev); + return ret; + } + } else { + /* + * If reshape is still in progress, and md_check_recovery() can + * continue to reshape, don't restart reshape because data can + * be corrupted for raid456. + */ + clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); + } + + mddev_unlock(mddev); + sysfs_notify_dirent_safe(mddev->sysfs_degraded); + return 0; +} + static ssize_t action_store(struct mddev *mddev, const char *page, size_t len) { @@ -5090,32 +5129,10 @@ action_store(struct mddev *mddev, const char *page, size_t len) clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); set_bit(MD_RECOVERY_RECOVER, &mddev->recovery); } else if (cmd_match(page, "reshape")) { - int err; - if (mddev->pers->start_reshape == NULL) - return -EINVAL; - err = mddev_lock(mddev); - if (!err) { - if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) { - err = -EBUSY; - } else if (mddev->reshape_position == MaxSector || - mddev->pers->check_reshape == NULL || - mddev->pers->check_reshape(mddev)) { - clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - err = mddev->pers->start_reshape(mddev); - } else { - /* - * If reshape is still in progress, and - * md_check_recovery() can continue to reshape, - * don't restart reshape because data can be - * corrupted for raid456. - */ - clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - } - mddev_unlock(mddev); - } + int err = mddev_start_reshape(mddev); + if (err) return err; - sysfs_notify_dirent_safe(mddev->sysfs_degraded); } else { if (cmd_match(page, "check")) set_bit(MD_RECOVERY_CHECK, &mddev->recovery); From c8ecfe680c371db2a6d125de5d6bc2398950e9cf Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:44 +0800 Subject: [PATCH 012/209] md: replace sysfs api sync_action with new helpers To get rid of extrem long if else if usage, and make code cleaner. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-6-yukuai1@huaweicloud.com --- drivers/md/md.c | 94 +++++++++++++++++++++++++++---------------------- 1 file changed, 52 insertions(+), 42 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index b34ae9fbd24671..d035cd52e49a2d 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -4950,27 +4950,9 @@ const char *md_sync_action_name(enum sync_action action) static ssize_t action_show(struct mddev *mddev, char *page) { - char *type = "idle"; - unsigned long recovery = mddev->recovery; - if (test_bit(MD_RECOVERY_FROZEN, &recovery)) - type = "frozen"; - else if (test_bit(MD_RECOVERY_RUNNING, &recovery) || - (md_is_rdwr(mddev) && test_bit(MD_RECOVERY_NEEDED, &recovery))) { - if (test_bit(MD_RECOVERY_RESHAPE, &recovery)) - type = "reshape"; - else if (test_bit(MD_RECOVERY_SYNC, &recovery)) { - if (!test_bit(MD_RECOVERY_REQUESTED, &recovery)) - type = "resync"; - else if (test_bit(MD_RECOVERY_CHECK, &recovery)) - type = "check"; - else - type = "repair"; - } else if (test_bit(MD_RECOVERY_RECOVER, &recovery)) - type = "recover"; - else if (mddev->reshape_position != MaxSector) - type = "reshape"; - } - return sprintf(page, "%s\n", type); + enum sync_action action = md_sync_action(mddev); + + return sprintf(page, "%s\n", md_sync_action_name(action)); } /** @@ -5113,35 +5095,63 @@ static int mddev_start_reshape(struct mddev *mddev) static ssize_t action_store(struct mddev *mddev, const char *page, size_t len) { + int ret; + enum sync_action action; + if (!mddev->pers || !mddev->pers->sync_request) return -EINVAL; + action = md_sync_action_by_name(page); - if (cmd_match(page, "idle")) - idle_sync_thread(mddev); - else if (cmd_match(page, "frozen")) - frozen_sync_thread(mddev); - else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) - return -EBUSY; - else if (cmd_match(page, "resync")) - clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - else if (cmd_match(page, "recover")) { - clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - set_bit(MD_RECOVERY_RECOVER, &mddev->recovery); - } else if (cmd_match(page, "reshape")) { - int err = mddev_start_reshape(mddev); - - if (err) - return err; + /* TODO: mdadm rely on "idle" to start sync_thread. */ + if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) { + switch (action) { + case ACTION_FROZEN: + frozen_sync_thread(mddev); + return len; + case ACTION_IDLE: + idle_sync_thread(mddev); + break; + case ACTION_RESHAPE: + case ACTION_RECOVER: + case ACTION_CHECK: + case ACTION_REPAIR: + case ACTION_RESYNC: + return -EBUSY; + default: + return -EINVAL; + } } else { - if (cmd_match(page, "check")) + switch (action) { + case ACTION_FROZEN: + set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); + return len; + case ACTION_RESHAPE: + clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); + ret = mddev_start_reshape(mddev); + if (ret) + return ret; + break; + case ACTION_RECOVER: + clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); + set_bit(MD_RECOVERY_RECOVER, &mddev->recovery); + break; + case ACTION_CHECK: set_bit(MD_RECOVERY_CHECK, &mddev->recovery); - else if (!cmd_match(page, "repair")) + fallthrough; + case ACTION_REPAIR: + set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); + set_bit(MD_RECOVERY_SYNC, &mddev->recovery); + fallthrough; + case ACTION_RESYNC: + case ACTION_IDLE: + clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); + break; + default: return -EINVAL; - clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); - set_bit(MD_RECOVERY_SYNC, &mddev->recovery); + } } + if (mddev->ro == MD_AUTO_READ) { /* A write to sync_action is enough to justify * canceling read-auto mode From df79234bdc3f441bec99dfc8199b6f2c673203ed Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:45 +0800 Subject: [PATCH 013/209] md: remove parameter check_seq for stop_sync_thread() Caller will always set MD_RECOVERY_FROZEN if check_seq is true, and always clear MD_RECOVERY_FROZEN if check_seq is false, hence replace the parameter with test_bit() to make code cleaner. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-7-yukuai1@huaweicloud.com --- drivers/md/md.c | 26 +++++++++++--------------- 1 file changed, 11 insertions(+), 15 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index d035cd52e49a2d..44cb18ec1c52d9 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -4961,15 +4961,10 @@ action_show(struct mddev *mddev, char *page) * @locked: if set, reconfig_mutex will still be held after this function * return; if not set, reconfig_mutex will be released after this * function return. - * @check_seq: if set, only wait for curent running sync_thread to stop, noted - * that new sync_thread can still start. */ -static void stop_sync_thread(struct mddev *mddev, bool locked, bool check_seq) +static void stop_sync_thread(struct mddev *mddev, bool locked) { - int sync_seq; - - if (check_seq) - sync_seq = atomic_read(&mddev->sync_seq); + int sync_seq = atomic_read(&mddev->sync_seq); if (!test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) { if (!locked) @@ -4990,7 +4985,8 @@ static void stop_sync_thread(struct mddev *mddev, bool locked, bool check_seq) wait_event(resync_wait, !test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) || - (check_seq && sync_seq != atomic_read(&mddev->sync_seq))); + (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery) && + sync_seq != atomic_read(&mddev->sync_seq))); if (locked) mddev_lock_nointr(mddev); @@ -5001,7 +4997,7 @@ void md_idle_sync_thread(struct mddev *mddev) lockdep_assert_held(&mddev->reconfig_mutex); clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - stop_sync_thread(mddev, true, true); + stop_sync_thread(mddev, true); } EXPORT_SYMBOL_GPL(md_idle_sync_thread); @@ -5010,7 +5006,7 @@ void md_frozen_sync_thread(struct mddev *mddev) lockdep_assert_held(&mddev->reconfig_mutex); set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - stop_sync_thread(mddev, true, false); + stop_sync_thread(mddev, true); } EXPORT_SYMBOL_GPL(md_frozen_sync_thread); @@ -5035,7 +5031,7 @@ static void idle_sync_thread(struct mddev *mddev) return; } - stop_sync_thread(mddev, false, true); + stop_sync_thread(mddev, false); mutex_unlock(&mddev->sync_mutex); } @@ -5049,7 +5045,7 @@ static void frozen_sync_thread(struct mddev *mddev) return; } - stop_sync_thread(mddev, false, false); + stop_sync_thread(mddev, false); mutex_unlock(&mddev->sync_mutex); } @@ -6544,7 +6540,7 @@ void md_stop_writes(struct mddev *mddev) { mddev_lock_nointr(mddev); set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - stop_sync_thread(mddev, true, false); + stop_sync_thread(mddev, true); __md_stop_writes(mddev); mddev_unlock(mddev); } @@ -6612,7 +6608,7 @@ static int md_set_readonly(struct mddev *mddev) set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); } - stop_sync_thread(mddev, false, false); + stop_sync_thread(mddev, false); wait_event(mddev->sb_wait, !test_bit(MD_SB_CHANGE_PENDING, &mddev->sb_flags)); mddev_lock_nointr(mddev); @@ -6658,7 +6654,7 @@ static int do_md_stop(struct mddev *mddev, int mode) set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); } - stop_sync_thread(mddev, true, false); + stop_sync_thread(mddev, true); if (mddev->sysfs_active || test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) { From 5ce10a38590c77f20d0dc706944f79e7d56a7400 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:46 +0800 Subject: [PATCH 014/209] md: don't fail action_store() if sync_thread is not registered MD_RECOVERY_RUNNING will always be set when trying to register a new sync_thread, however, if md_start_sync() turns out to do nothing, MD_RECOVERY_RUNNING will be cleared in this case. And during the race window, action_store() will return -EBUSY, which will cause some mdadm tests to fail. For example: The test 07reshape5intr will add a new disk to array, then start reshape: mdadm /dev/md0 --add /dev/xxx mdadm --grow /dev/md0 -n 3 And add_bound_rdev() from mdadm --add will set MD_RECOVERY_NEEDED, then during the race windown, mdadm --grow will fail. Fix the problem by waiting in action_store() during the race window, fail only if sync_thread is registered. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-8-yukuai1@huaweicloud.com --- drivers/md/md.c | 85 +++++++++++++++++++------------------------------ drivers/md/md.h | 2 -- 2 files changed, 33 insertions(+), 54 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index 44cb18ec1c52d9..86abd0fe0681ac 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -753,7 +753,6 @@ int mddev_init(struct mddev *mddev) mutex_init(&mddev->open_mutex); mutex_init(&mddev->reconfig_mutex); - mutex_init(&mddev->sync_mutex); mutex_init(&mddev->suspend_mutex); mutex_init(&mddev->bitmap_info.mutex); INIT_LIST_HEAD(&mddev->disks); @@ -5021,34 +5020,6 @@ void md_unfrozen_sync_thread(struct mddev *mddev) } EXPORT_SYMBOL_GPL(md_unfrozen_sync_thread); -static void idle_sync_thread(struct mddev *mddev) -{ - mutex_lock(&mddev->sync_mutex); - clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - - if (mddev_lock(mddev)) { - mutex_unlock(&mddev->sync_mutex); - return; - } - - stop_sync_thread(mddev, false); - mutex_unlock(&mddev->sync_mutex); -} - -static void frozen_sync_thread(struct mddev *mddev) -{ - mutex_lock(&mddev->sync_mutex); - set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - - if (mddev_lock(mddev)) { - mutex_unlock(&mddev->sync_mutex); - return; - } - - stop_sync_thread(mddev, false); - mutex_unlock(&mddev->sync_mutex); -} - static int mddev_start_reshape(struct mddev *mddev) { int ret; @@ -5056,24 +5027,13 @@ static int mddev_start_reshape(struct mddev *mddev) if (mddev->pers->start_reshape == NULL) return -EINVAL; - ret = mddev_lock(mddev); - if (ret) - return ret; - - if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) { - mddev_unlock(mddev); - return -EBUSY; - } - if (mddev->reshape_position == MaxSector || mddev->pers->check_reshape == NULL || mddev->pers->check_reshape(mddev)) { clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); ret = mddev->pers->start_reshape(mddev); - if (ret) { - mddev_unlock(mddev); + if (ret) return ret; - } } else { /* * If reshape is still in progress, and md_check_recovery() can @@ -5083,7 +5043,6 @@ static int mddev_start_reshape(struct mddev *mddev) clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); } - mddev_unlock(mddev); sysfs_notify_dirent_safe(mddev->sysfs_degraded); return 0; } @@ -5097,36 +5056,53 @@ action_store(struct mddev *mddev, const char *page, size_t len) if (!mddev->pers || !mddev->pers->sync_request) return -EINVAL; +retry: + if (work_busy(&mddev->sync_work)) + flush_work(&mddev->sync_work); + + ret = mddev_lock(mddev); + if (ret) + return ret; + + if (work_busy(&mddev->sync_work)) { + mddev_unlock(mddev); + goto retry; + } + action = md_sync_action_by_name(page); /* TODO: mdadm rely on "idle" to start sync_thread. */ if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery)) { switch (action) { case ACTION_FROZEN: - frozen_sync_thread(mddev); - return len; + md_frozen_sync_thread(mddev); + ret = len; + goto out; case ACTION_IDLE: - idle_sync_thread(mddev); + md_idle_sync_thread(mddev); break; case ACTION_RESHAPE: case ACTION_RECOVER: case ACTION_CHECK: case ACTION_REPAIR: case ACTION_RESYNC: - return -EBUSY; + ret = -EBUSY; + goto out; default: - return -EINVAL; + ret = -EINVAL; + goto out; } } else { switch (action) { case ACTION_FROZEN: set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); - return len; + ret = len; + goto out; case ACTION_RESHAPE: clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); ret = mddev_start_reshape(mddev); if (ret) - return ret; + goto out; break; case ACTION_RECOVER: clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); @@ -5144,7 +5120,8 @@ action_store(struct mddev *mddev, const char *page, size_t len) clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); break; default: - return -EINVAL; + ret = -EINVAL; + goto out; } } @@ -5152,14 +5129,18 @@ action_store(struct mddev *mddev, const char *page, size_t len) /* A write to sync_action is enough to justify * canceling read-auto mode */ - flush_work(&mddev->sync_work); mddev->ro = MD_RDWR; md_wakeup_thread(mddev->sync_thread); } + set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); md_wakeup_thread(mddev->thread); sysfs_notify_dirent_safe(mddev->sysfs_action); - return len; + ret = len; + +out: + mddev_unlock(mddev); + return ret; } static struct md_sysfs_entry md_scan_mode = diff --git a/drivers/md/md.h b/drivers/md/md.h index 88add162b08ea0..732053b905b229 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -595,8 +595,6 @@ struct mddev { */ struct list_head deleting; - /* Used to synchronize idle and frozen for action_store() */ - struct mutex sync_mutex; /* The sequence number for sync thread */ atomic_t sync_seq; From 7d9f107a4e946bb52b7502eed9ed8f316700397e Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:47 +0800 Subject: [PATCH 015/209] md: use new helpers in md_do_sync() Make code cleaner. and also use the action_name directly in kernel log: - "check" instead of "data-check" - "repair" instead of "requested-resync" Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-9-yukuai1@huaweicloud.com --- drivers/md/md.c | 21 +++++---------------- drivers/md/md.h | 2 +- 2 files changed, 6 insertions(+), 17 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index 86abd0fe0681ac..5fa7b5f4bc6de5 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -8929,7 +8929,8 @@ void md_do_sync(struct md_thread *thread) sector_t last_check; int skipped = 0; struct md_rdev *rdev; - char *desc, *action = NULL; + enum sync_action action; + const char *desc; struct blk_plug plug; int ret; @@ -8960,21 +8961,9 @@ void md_do_sync(struct md_thread *thread) goto skip; } - if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) { - if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) { - desc = "data-check"; - action = "check"; - } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) { - desc = "requested-resync"; - action = "repair"; - } else - desc = "resync"; - } else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) - desc = "reshape"; - else - desc = "recovery"; - - mddev->last_sync_action = action ?: desc; + action = md_sync_action(mddev); + desc = md_sync_action_name(action); + mddev->last_sync_action = desc; /* * Before starting a resync we must have set curr_resync to diff --git a/drivers/md/md.h b/drivers/md/md.h index 732053b905b229..ee06cb076f8c7d 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -432,7 +432,7 @@ struct mddev { * when the sync thread is "frozen" (interrupted) or "idle" (stopped * or finished). It is overwritten when a new sync operation is begun. */ - char *last_sync_action; + const char *last_sync_action; sector_t curr_resync; /* last block scheduled */ /* As resync requests can complete out of order, we cannot easily track * how much resync has been completed. So we occasionally pause until From d249e541887a966df37544f7c4d301cdee0f0e27 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:48 +0800 Subject: [PATCH 016/209] md: replace last_sync_action with new enum type The only difference is that "none" is removed and initial last_sync_action will be idle. On the one hand, this value is introduced by commit c4a395514516 ("MD: Remember the last sync operation that was performed"), and the usage described in commit message is not affected. On the other hand, last_sync_action is not used in mdadm or mdmon, and none of the tests that I can find. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-10-yukuai1@huaweicloud.com --- drivers/md/dm-raid.c | 2 +- drivers/md/md.c | 7 ++++--- drivers/md/md.h | 9 ++++----- 3 files changed, 9 insertions(+), 9 deletions(-) diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c index abe88d1e673582..052c00c1eb1542 100644 --- a/drivers/md/dm-raid.c +++ b/drivers/md/dm-raid.c @@ -3542,7 +3542,7 @@ static void raid_status(struct dm_target *ti, status_type_t type, recovery = rs->md.recovery; state = decipher_sync_action(mddev, recovery); progress = rs_get_progress(rs, recovery, state, resync_max_sectors); - resync_mismatches = (mddev->last_sync_action && !strcasecmp(mddev->last_sync_action, "check")) ? + resync_mismatches = mddev->last_sync_action == ACTION_CHECK ? atomic64_read(&mddev->resync_mismatches) : 0; /* HM FIXME: do we want another state char for raid0? It shows 'D'/'A'/'-' now */ diff --git a/drivers/md/md.c b/drivers/md/md.c index 5fa7b5f4bc6de5..ab492e88586729 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -768,7 +768,7 @@ int mddev_init(struct mddev *mddev) init_waitqueue_head(&mddev->recovery_wait); mddev->reshape_position = MaxSector; mddev->reshape_backwards = 0; - mddev->last_sync_action = "none"; + mddev->last_sync_action = ACTION_IDLE; mddev->resync_min = 0; mddev->resync_max = MaxSector; mddev->level = LEVEL_NONE; @@ -5149,7 +5149,8 @@ __ATTR_PREALLOC(sync_action, S_IRUGO|S_IWUSR, action_show, action_store); static ssize_t last_sync_action_show(struct mddev *mddev, char *page) { - return sprintf(page, "%s\n", mddev->last_sync_action); + return sprintf(page, "%s\n", + md_sync_action_name(mddev->last_sync_action)); } static struct md_sysfs_entry md_last_scan_mode = __ATTR_RO(last_sync_action); @@ -8963,7 +8964,7 @@ void md_do_sync(struct md_thread *thread) action = md_sync_action(mddev); desc = md_sync_action_name(action); - mddev->last_sync_action = desc; + mddev->last_sync_action = action; /* * Before starting a resync we must have set curr_resync to diff --git a/drivers/md/md.h b/drivers/md/md.h index ee06cb076f8c7d..41781e41d8ffc2 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -426,13 +426,12 @@ struct mddev { struct md_thread __rcu *thread; /* management thread */ struct md_thread __rcu *sync_thread; /* doing resync or reconstruct */ - /* 'last_sync_action' is initialized to "none". It is set when a - * sync operation (i.e "data-check", "requested-resync", "resync", - * "recovery", or "reshape") is started. It holds this value even + /* + * Set when a sync operation is started. It holds this value even * when the sync thread is "frozen" (interrupted) or "idle" (stopped - * or finished). It is overwritten when a new sync operation is begun. + * or finished). It is overwritten when a new sync operation is begun. */ - const char *last_sync_action; + enum sync_action last_sync_action; sector_t curr_resync; /* last block scheduled */ /* As resync requests can complete out of order, we cannot easily track * how much resync has been completed. So we occasionally pause until From bbf2076277b137f03624259da0a0369af88f3a68 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:49 +0800 Subject: [PATCH 017/209] md: factor out helpers for different sync_action in md_do_sync() Make code cleaner by replacing if else if with switch, and it's more obvious now what is doing for each sync_action. There are no functional changes. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-11-yukuai1@huaweicloud.com --- drivers/md/md.c | 123 ++++++++++++++++++++++++++++-------------------- 1 file changed, 73 insertions(+), 50 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index ab492e88586729..ec2ef4dd42cf56 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -8914,6 +8914,77 @@ void md_allow_write(struct mddev *mddev) } EXPORT_SYMBOL_GPL(md_allow_write); +static sector_t md_sync_max_sectors(struct mddev *mddev, + enum sync_action action) +{ + switch (action) { + case ACTION_RESYNC: + case ACTION_CHECK: + case ACTION_REPAIR: + atomic64_set(&mddev->resync_mismatches, 0); + fallthrough; + case ACTION_RESHAPE: + return mddev->resync_max_sectors; + case ACTION_RECOVER: + return mddev->dev_sectors; + default: + return 0; + } +} + +static sector_t md_sync_position(struct mddev *mddev, enum sync_action action) +{ + sector_t start = 0; + struct md_rdev *rdev; + + switch (action) { + case ACTION_CHECK: + case ACTION_REPAIR: + return mddev->resync_min; + case ACTION_RESYNC: + if (!mddev->bitmap) + return mddev->recovery_cp; + return 0; + case ACTION_RESHAPE: + /* + * If the original node aborts reshaping then we continue the + * reshaping, so set again to avoid restart reshape from the + * first beginning + */ + if (mddev_is_clustered(mddev) && + mddev->reshape_position != MaxSector) + return mddev->reshape_position; + return 0; + case ACTION_RECOVER: + start = MaxSector; + rcu_read_lock(); + rdev_for_each_rcu(rdev, mddev) + if (rdev->raid_disk >= 0 && + !test_bit(Journal, &rdev->flags) && + !test_bit(Faulty, &rdev->flags) && + !test_bit(In_sync, &rdev->flags) && + rdev->recovery_offset < start) + start = rdev->recovery_offset; + rcu_read_unlock(); + + /* If there is a bitmap, we need to make sure all + * writes that started before we added a spare + * complete before we start doing a recovery. + * Otherwise the write might complete and (via + * bitmap_endwrite) set a bit in the bitmap after the + * recovery has checked that bit and skipped that + * region. + */ + if (mddev->bitmap) { + mddev->pers->quiesce(mddev, 1); + mddev->pers->quiesce(mddev, 0); + } + return start; + default: + return MaxSector; + } +} + #define SYNC_MARKS 10 #define SYNC_MARK_STEP (3*HZ) #define UPDATE_FREQUENCY (5*60*HZ) @@ -9032,56 +9103,8 @@ void md_do_sync(struct md_thread *thread) spin_unlock(&all_mddevs_lock); } while (mddev->curr_resync < MD_RESYNC_DELAYED); - j = 0; - if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) { - /* resync follows the size requested by the personality, - * which defaults to physical size, but can be virtual size - */ - max_sectors = mddev->resync_max_sectors; - atomic64_set(&mddev->resync_mismatches, 0); - /* we don't use the checkpoint if there's a bitmap */ - if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) - j = mddev->resync_min; - else if (!mddev->bitmap) - j = mddev->recovery_cp; - - } else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) { - max_sectors = mddev->resync_max_sectors; - /* - * If the original node aborts reshaping then we continue the - * reshaping, so set j again to avoid restart reshape from the - * first beginning - */ - if (mddev_is_clustered(mddev) && - mddev->reshape_position != MaxSector) - j = mddev->reshape_position; - } else { - /* recovery follows the physical size of devices */ - max_sectors = mddev->dev_sectors; - j = MaxSector; - rcu_read_lock(); - rdev_for_each_rcu(rdev, mddev) - if (rdev->raid_disk >= 0 && - !test_bit(Journal, &rdev->flags) && - !test_bit(Faulty, &rdev->flags) && - !test_bit(In_sync, &rdev->flags) && - rdev->recovery_offset < j) - j = rdev->recovery_offset; - rcu_read_unlock(); - - /* If there is a bitmap, we need to make sure all - * writes that started before we added a spare - * complete before we start doing a recovery. - * Otherwise the write might complete and (via - * bitmap_endwrite) set a bit in the bitmap after the - * recovery has checked that bit and skipped that - * region. - */ - if (mddev->bitmap) { - mddev->pers->quiesce(mddev, 1); - mddev->pers->quiesce(mddev, 0); - } - } + max_sectors = md_sync_max_sectors(mddev, action); + j = md_sync_position(mddev, action); pr_info("md: %s of RAID array %s\n", desc, mdname(mddev)); pr_debug("md: minimum _guaranteed_ speed: %d KB/sec/disk.\n", speed_min(mddev)); From bc49694a9e8fd1b36bca47d9a54ec8da8e39012f Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:50 +0800 Subject: [PATCH 018/209] md: pass in max_sectors for pers->sync_request() For different sync_action, sync_thread will use different max_sectors, see details in md_sync_max_sectors(), currently both md_do_sync() and pers->sync_request() in eatch iteration have to get the same max_sectors. Hence pass in max_sectors for pers->sync_request() to prevent redundant code. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-12-yukuai1@huaweicloud.com --- drivers/md/md.c | 5 +++-- drivers/md/md.h | 3 ++- drivers/md/raid1.c | 5 ++--- drivers/md/raid10.c | 8 ++------ drivers/md/raid5.c | 3 +-- 5 files changed, 10 insertions(+), 14 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index ec2ef4dd42cf56..c0426a6d2fd197 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -9186,7 +9186,8 @@ void md_do_sync(struct md_thread *thread) if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) break; - sectors = mddev->pers->sync_request(mddev, j, &skipped); + sectors = mddev->pers->sync_request(mddev, j, max_sectors, + &skipped); if (sectors == 0) { set_bit(MD_RECOVERY_INTR, &mddev->recovery); break; @@ -9276,7 +9277,7 @@ void md_do_sync(struct md_thread *thread) mddev->curr_resync_completed = mddev->curr_resync; sysfs_notify_dirent_safe(mddev->sysfs_completed); } - mddev->pers->sync_request(mddev, max_sectors, &skipped); + mddev->pers->sync_request(mddev, max_sectors, max_sectors, &skipped); if (!test_bit(MD_RECOVERY_CHECK, &mddev->recovery) && mddev->curr_resync > MD_RESYNC_ACTIVE) { diff --git a/drivers/md/md.h b/drivers/md/md.h index 41781e41d8ffc2..2dc52edec3fe63 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -729,7 +729,8 @@ struct md_personality int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev); int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev); int (*spare_active) (struct mddev *mddev); - sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped); + sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, + sector_t max_sector, int *skipped); int (*resize) (struct mddev *mddev, sector_t sectors); sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks); int (*check_reshape) (struct mddev *mddev); diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 3d54f30112a0e8..2bbfb4e682b2ff 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -2756,12 +2756,12 @@ static struct r1bio *raid1_alloc_init_r1buf(struct r1conf *conf) */ static sector_t raid1_sync_request(struct mddev *mddev, sector_t sector_nr, - int *skipped) + sector_t max_sector, int *skipped) { struct r1conf *conf = mddev->private; struct r1bio *r1_bio; struct bio *bio; - sector_t max_sector, nr_sectors; + sector_t nr_sectors; int disk = -1; int i; int wonly = -1; @@ -2777,7 +2777,6 @@ static sector_t raid1_sync_request(struct mddev *mddev, sector_t sector_nr, if (init_resync(conf)) return 0; - max_sector = mddev->dev_sectors; if (sector_nr >= max_sector) { /* If we aborted, we need to abort the * sync on the 'current' bitmap chunk (there will diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index f8d7c02c6ed561..4e804602d1e53a 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -3139,12 +3139,12 @@ static void raid10_set_cluster_sync_high(struct r10conf *conf) */ static sector_t raid10_sync_request(struct mddev *mddev, sector_t sector_nr, - int *skipped) + sector_t max_sector, int *skipped) { struct r10conf *conf = mddev->private; struct r10bio *r10_bio; struct bio *biolist = NULL, *bio; - sector_t max_sector, nr_sectors; + sector_t nr_sectors; int i; int max_sync; sector_t sync_blocks; @@ -3174,10 +3174,6 @@ static sector_t raid10_sync_request(struct mddev *mddev, sector_t sector_nr, return 0; skipped: - max_sector = mddev->dev_sectors; - if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) || - test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) - max_sector = mddev->resync_max_sectors; if (sector_nr >= max_sector) { conf->cluster_sync_low = 0; conf->cluster_sync_high = 0; diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index a84389311dd1ea..013adc5ba0e124 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -6457,11 +6457,10 @@ static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *sk } static inline sector_t raid5_sync_request(struct mddev *mddev, sector_t sector_nr, - int *skipped) + sector_t max_sector, int *skipped) { struct r5conf *conf = mddev->private; struct stripe_head *sh; - sector_t max_sector = mddev->dev_sectors; sector_t sync_blocks; int still_degraded = 0; int i; From 305a5170dc5cf3d395bb4c4e9239bca6d0b54b49 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 11 Jun 2024 21:22:51 +0800 Subject: [PATCH 019/209] md/raid5: avoid BUG_ON() while continue reshape after reassembling Currently, mdadm support --revert-reshape to abort the reshape while reassembling, as the test 07revert-grow. However, following BUG_ON() can be triggerred by the test: kernel BUG at drivers/md/raid5.c:6278! invalid opcode: 0000 [#1] PREEMPT SMP PTI irq event stamp: 158985 CPU: 6 PID: 891 Comm: md0_reshape Not tainted 6.9.0-03335-g7592a0b0049a #94 RIP: 0010:reshape_request+0x3f1/0xe60 Call Trace: raid5_sync_request+0x43d/0x550 md_do_sync+0xb7a/0x2110 md_thread+0x294/0x2b0 kthread+0x147/0x1c0 ret_from_fork+0x59/0x70 ret_from_fork_asm+0x1a/0x30 Root cause is that --revert-reshape update the raid_disks from 5 to 4, while reshape position is still set, and after reassembling the array, reshape position will be read from super block, then during reshape the checking of 'writepos' that is caculated by old reshape position will fail. Fix this panic the easy way first, by converting the BUG_ON() to WARN_ON(), and stop the reshape if checkings fail. Noted that mdadm must fix --revert-shape as well, and probably md/raid should enhance metadata validation as well, however this means reassemble will fail and there must be user tools to fix the wrong metadata. Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240611132251.1967786-13-yukuai1@huaweicloud.com --- drivers/md/raid5.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 013adc5ba0e124..547fd15115cdc5 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -6254,7 +6254,9 @@ static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *sk safepos = conf->reshape_safe; sector_div(safepos, data_disks); if (mddev->reshape_backwards) { - BUG_ON(writepos < reshape_sectors); + if (WARN_ON(writepos < reshape_sectors)) + return MaxSector; + writepos -= reshape_sectors; readpos += reshape_sectors; safepos += reshape_sectors; @@ -6272,14 +6274,18 @@ static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *sk * to set 'stripe_addr' which is where we will write to. */ if (mddev->reshape_backwards) { - BUG_ON(conf->reshape_progress == 0); + if (WARN_ON(conf->reshape_progress == 0)) + return MaxSector; + stripe_addr = writepos; - BUG_ON((mddev->dev_sectors & - ~((sector_t)reshape_sectors - 1)) - - reshape_sectors - stripe_addr - != sector_nr); + if (WARN_ON((mddev->dev_sectors & + ~((sector_t)reshape_sectors - 1)) - + reshape_sectors - stripe_addr != sector_nr)) + return MaxSector; } else { - BUG_ON(writepos != sector_nr + reshape_sectors); + if (WARN_ON(writepos != sector_nr + reshape_sectors)) + return MaxSector; + stripe_addr = sector_nr; } From 3253aba3408aa4eb2e4e09365eede3e63ef7536b Mon Sep 17 00:00:00 2001 From: Andreas Hindborg Date: Tue, 11 Jun 2024 13:45:49 +0200 Subject: [PATCH 020/209] rust: block: introduce `kernel::block::mq` module Add initial abstractions for working with blk-mq. This patch is a maintained, refactored subset of code originally published by Wedson Almeida Filho [1]. [1] https://github.com/wedsonaf/linux/tree/f2cfd2fe0e2ca4e90994f96afe268bbd4382a891/rust/kernel/blk/mq.rs Cc: Wedson Almeida Filho Signed-off-by: Andreas Hindborg Reviewed-by: Benno Lossin Link: https://lore.kernel.org/r/20240611114551.228679-2-nmi@metaspace.dk Signed-off-by: Jens Axboe --- rust/bindings/bindings_helper.h | 3 + rust/helpers.c | 16 ++ rust/kernel/block.rs | 5 + rust/kernel/block/mq.rs | 98 +++++++++++ rust/kernel/block/mq/gen_disk.rs | 215 ++++++++++++++++++++++++ rust/kernel/block/mq/operations.rs | 245 ++++++++++++++++++++++++++++ rust/kernel/block/mq/raw_writer.rs | 55 +++++++ rust/kernel/block/mq/request.rs | 253 +++++++++++++++++++++++++++++ rust/kernel/block/mq/tag_set.rs | 86 ++++++++++ rust/kernel/error.rs | 6 + rust/kernel/lib.rs | 2 + 11 files changed, 984 insertions(+) create mode 100644 rust/kernel/block.rs create mode 100644 rust/kernel/block/mq.rs create mode 100644 rust/kernel/block/mq/gen_disk.rs create mode 100644 rust/kernel/block/mq/operations.rs create mode 100644 rust/kernel/block/mq/raw_writer.rs create mode 100644 rust/kernel/block/mq/request.rs create mode 100644 rust/kernel/block/mq/tag_set.rs diff --git a/rust/bindings/bindings_helper.h b/rust/bindings/bindings_helper.h index ddb5644d4fd900..84f601d7068e37 100644 --- a/rust/bindings/bindings_helper.h +++ b/rust/bindings/bindings_helper.h @@ -7,6 +7,8 @@ */ #include +#include +#include #include #include #include @@ -20,6 +22,7 @@ /* `bindgen` gets confused at certain things. */ const size_t RUST_CONST_HELPER_ARCH_SLAB_MINALIGN = ARCH_SLAB_MINALIGN; +const size_t RUST_CONST_HELPER_PAGE_SIZE = PAGE_SIZE; const gfp_t RUST_CONST_HELPER_GFP_ATOMIC = GFP_ATOMIC; const gfp_t RUST_CONST_HELPER_GFP_KERNEL = GFP_KERNEL; const gfp_t RUST_CONST_HELPER_GFP_KERNEL_ACCOUNT = GFP_KERNEL_ACCOUNT; diff --git a/rust/helpers.c b/rust/helpers.c index 2c37a0f5d7a84f..3df5217fb2fff8 100644 --- a/rust/helpers.c +++ b/rust/helpers.c @@ -186,3 +186,19 @@ static_assert( __alignof__(size_t) == __alignof__(uintptr_t), "Rust code expects C `size_t` to match Rust `usize`" ); + +// This will soon be moved to a separate file, so no need to merge with above. +#include +#include + +void *rust_helper_blk_mq_rq_to_pdu(struct request *rq) +{ + return blk_mq_rq_to_pdu(rq); +} +EXPORT_SYMBOL_GPL(rust_helper_blk_mq_rq_to_pdu); + +struct request *rust_helper_blk_mq_rq_from_pdu(void *pdu) +{ + return blk_mq_rq_from_pdu(pdu); +} +EXPORT_SYMBOL_GPL(rust_helper_blk_mq_rq_from_pdu); diff --git a/rust/kernel/block.rs b/rust/kernel/block.rs new file mode 100644 index 00000000000000..150f710efe5b46 --- /dev/null +++ b/rust/kernel/block.rs @@ -0,0 +1,5 @@ +// SPDX-License-Identifier: GPL-2.0 + +//! Types for working with the block layer. + +pub mod mq; diff --git a/rust/kernel/block/mq.rs b/rust/kernel/block/mq.rs new file mode 100644 index 00000000000000..fb0f393c1cea6c --- /dev/null +++ b/rust/kernel/block/mq.rs @@ -0,0 +1,98 @@ +// SPDX-License-Identifier: GPL-2.0 + +//! This module provides types for implementing block drivers that interface the +//! blk-mq subsystem. +//! +//! To implement a block device driver, a Rust module must do the following: +//! +//! - Implement [`Operations`] for a type `T`. +//! - Create a [`TagSet`]. +//! - Create a [`GenDisk`], via the [`GenDiskBuilder`]. +//! - Add the disk to the system by calling [`GenDiskBuilder::build`] passing in +//! the `TagSet` reference. +//! +//! The types available in this module that have direct C counterparts are: +//! +//! - The [`TagSet`] type that abstracts the C type `struct tag_set`. +//! - The [`GenDisk`] type that abstracts the C type `struct gendisk`. +//! - The [`Request`] type that abstracts the C type `struct request`. +//! +//! The kernel will interface with the block device driver by calling the method +//! implementations of the `Operations` trait. +//! +//! IO requests are passed to the driver as [`kernel::types::ARef`] +//! instances. The `Request` type is a wrapper around the C `struct request`. +//! The driver must mark end of processing by calling one of the +//! `Request::end`, methods. Failure to do so can lead to deadlock or timeout +//! errors. Please note that the C function `blk_mq_start_request` is implicitly +//! called when the request is queued with the driver. +//! +//! The `TagSet` is responsible for creating and maintaining a mapping between +//! `Request`s and integer ids as well as carrying a pointer to the vtable +//! generated by `Operations`. This mapping is useful for associating +//! completions from hardware with the correct `Request` instance. The `TagSet` +//! determines the maximum queue depth by setting the number of `Request` +//! instances available to the driver, and it determines the number of queues to +//! instantiate for the driver. If possible, a driver should allocate one queue +//! per core, to keep queue data local to a core. +//! +//! One `TagSet` instance can be shared between multiple `GenDisk` instances. +//! This can be useful when implementing drivers where one piece of hardware +//! with one set of IO resources are represented to the user as multiple disks. +//! +//! One significant difference between block device drivers implemented with +//! these Rust abstractions and drivers implemented in C, is that the Rust +//! drivers have to own a reference count on the `Request` type when the IO is +//! in flight. This is to ensure that the C `struct request` instances backing +//! the Rust `Request` instances are live while the Rust driver holds a +//! reference to the `Request`. In addition, the conversion of an integer tag to +//! a `Request` via the `TagSet` would not be sound without this bookkeeping. +//! +//! [`GenDisk`]: gen_disk::GenDisk +//! [`GenDisk`]: gen_disk::GenDisk +//! [`GenDiskBuilder`]: gen_disk::GenDiskBuilder +//! [`GenDiskBuilder::build`]: gen_disk::GenDiskBuilder::build +//! +//! # Example +//! +//! ```rust +//! use kernel::{ +//! alloc::flags, +//! block::mq::*, +//! new_mutex, +//! prelude::*, +//! sync::{Arc, Mutex}, +//! types::{ARef, ForeignOwnable}, +//! }; +//! +//! struct MyBlkDevice; +//! +//! #[vtable] +//! impl Operations for MyBlkDevice { +//! +//! fn queue_rq(rq: ARef>, _is_last: bool) -> Result { +//! Request::end_ok(rq); +//! Ok(()) +//! } +//! +//! fn commit_rqs() {} +//! } +//! +//! let tagset: Arc> = +//! Arc::pin_init(TagSet::new(1, 256, 1), flags::GFP_KERNEL)?; +//! let mut disk = gen_disk::GenDiskBuilder::new() +//! .capacity_sectors(4096) +//! .build(format_args!("myblk"), tagset)?; +//! +//! # Ok::<(), kernel::error::Error>(()) +//! ``` + +pub mod gen_disk; +mod operations; +mod raw_writer; +mod request; +mod tag_set; + +pub use operations::Operations; +pub use request::Request; +pub use tag_set::TagSet; diff --git a/rust/kernel/block/mq/gen_disk.rs b/rust/kernel/block/mq/gen_disk.rs new file mode 100644 index 00000000000000..3b9edb96c8ff77 --- /dev/null +++ b/rust/kernel/block/mq/gen_disk.rs @@ -0,0 +1,215 @@ +// SPDX-License-Identifier: GPL-2.0 + +//! Generic disk abstraction. +//! +//! C header: [`include/linux/blkdev.h`](srctree/include/linux/blkdev.h) +//! C header: [`include/linux/blk_mq.h`](srctree/include/linux/blk_mq.h) + +use crate::block::mq::{raw_writer::RawWriter, Operations, TagSet}; +use crate::error; +use crate::{bindings, error::from_err_ptr, error::Result, sync::Arc}; +use core::fmt::{self, Write}; + +/// A builder for [`GenDisk`]. +/// +/// Use this struct to configure and add new [`GenDisk`] to the VFS. +pub struct GenDiskBuilder { + rotational: bool, + logical_block_size: u32, + physical_block_size: u32, + capacity_sectors: u64, +} + +impl Default for GenDiskBuilder { + fn default() -> Self { + Self { + rotational: false, + logical_block_size: bindings::PAGE_SIZE as u32, + physical_block_size: bindings::PAGE_SIZE as u32, + capacity_sectors: 0, + } + } +} + +impl GenDiskBuilder { + /// Create a new instance. + pub fn new() -> Self { + Self::default() + } + + /// Set the rotational media attribute for the device to be built. + pub fn rotational(mut self, rotational: bool) -> Self { + self.rotational = rotational; + self + } + + /// Validate block size by verifying that it is between 512 and `PAGE_SIZE`, + /// and that it is a power of two. + fn validate_block_size(size: u32) -> Result<()> { + if !(512..=bindings::PAGE_SIZE as u32).contains(&size) || !size.is_power_of_two() { + Err(error::code::EINVAL) + } else { + Ok(()) + } + } + + /// Set the logical block size of the device to be built. + /// + /// This method will check that block size is a power of two and between 512 + /// and 4096. If not, an error is returned and the block size is not set. + /// + /// This is the smallest unit the storage device can address. It is + /// typically 4096 bytes. + pub fn logical_block_size(mut self, block_size: u32) -> Result { + Self::validate_block_size(block_size)?; + self.logical_block_size = block_size; + Ok(self) + } + + /// Set the physical block size of the device to be built. + /// + /// This method will check that block size is a power of two and between 512 + /// and 4096. If not, an error is returned and the block size is not set. + /// + /// This is the smallest unit a physical storage device can write + /// atomically. It is usually the same as the logical block size but may be + /// bigger. One example is SATA drives with 4096 byte physical block size + /// that expose a 512 byte logical block size to the operating system. + pub fn physical_block_size(mut self, block_size: u32) -> Result { + Self::validate_block_size(block_size)?; + self.physical_block_size = block_size; + Ok(self) + } + + /// Set the capacity of the device to be built, in sectors (512 bytes). + pub fn capacity_sectors(mut self, capacity: u64) -> Self { + self.capacity_sectors = capacity; + self + } + + /// Build a new `GenDisk` and add it to the VFS. + pub fn build( + self, + name: fmt::Arguments<'_>, + tagset: Arc>, + ) -> Result> { + let lock_class_key = crate::sync::LockClassKey::new(); + + // SAFETY: `tagset.raw_tag_set()` points to a valid and initialized tag set + let gendisk = from_err_ptr(unsafe { + bindings::__blk_mq_alloc_disk( + tagset.raw_tag_set(), + core::ptr::null_mut(), // TODO: We can pass queue limits right here + core::ptr::null_mut(), + lock_class_key.as_ptr(), + ) + })?; + + const TABLE: bindings::block_device_operations = bindings::block_device_operations { + submit_bio: None, + open: None, + release: None, + ioctl: None, + compat_ioctl: None, + check_events: None, + unlock_native_capacity: None, + getgeo: None, + set_read_only: None, + swap_slot_free_notify: None, + report_zones: None, + devnode: None, + alternative_gpt_sector: None, + get_unique_id: None, + // TODO: Set to THIS_MODULE. Waiting for const_refs_to_static feature to + // be merged (unstable in rustc 1.78 which is staged for linux 6.10) + // https://github.com/rust-lang/rust/issues/119618 + owner: core::ptr::null_mut(), + pr_ops: core::ptr::null_mut(), + free_disk: None, + poll_bio: None, + }; + + // SAFETY: `gendisk` is a valid pointer as we initialized it above + unsafe { (*gendisk).fops = &TABLE }; + + let mut raw_writer = RawWriter::from_array( + // SAFETY: `gendisk` points to a valid and initialized instance. We + // have exclusive access, since the disk is not added to the VFS + // yet. + unsafe { &mut (*gendisk).disk_name }, + )?; + raw_writer.write_fmt(name)?; + raw_writer.write_char('\0')?; + + // SAFETY: `gendisk` points to a valid and initialized instance of + // `struct gendisk`. We have exclusive access, so we cannot race. + unsafe { + bindings::blk_queue_logical_block_size((*gendisk).queue, self.logical_block_size) + }; + + // SAFETY: `gendisk` points to a valid and initialized instance of + // `struct gendisk`. We have exclusive access, so we cannot race. + unsafe { + bindings::blk_queue_physical_block_size((*gendisk).queue, self.physical_block_size) + }; + + // SAFETY: `gendisk` points to a valid and initialized instance of + // `struct gendisk`. `set_capacity` takes a lock to synchronize this + // operation, so we will not race. + unsafe { bindings::set_capacity(gendisk, self.capacity_sectors) }; + + if !self.rotational { + // SAFETY: `gendisk` points to a valid and initialized instance of + // `struct gendisk`. This operation uses a relaxed atomic bit flip + // operation, so there is no race on this field. + unsafe { bindings::blk_queue_flag_set(bindings::QUEUE_FLAG_NONROT, (*gendisk).queue) }; + } else { + // SAFETY: `gendisk` points to a valid and initialized instance of + // `struct gendisk`. This operation uses a relaxed atomic bit flip + // operation, so there is no race on this field. + unsafe { + bindings::blk_queue_flag_clear(bindings::QUEUE_FLAG_NONROT, (*gendisk).queue) + }; + } + + crate::error::to_result( + // SAFETY: `gendisk` points to a valid and initialized instance of + // `struct gendisk`. + unsafe { + bindings::device_add_disk(core::ptr::null_mut(), gendisk, core::ptr::null_mut()) + }, + )?; + + // INVARIANT: `gendisk` was initialized above. + // INVARIANT: `gendisk` was added to the VFS via `device_add_disk` above. + Ok(GenDisk { + _tagset: tagset, + gendisk, + }) + } +} + +/// A generic block device. +/// +/// # Invariants +/// +/// - `gendisk` must always point to an initialized and valid `struct gendisk`. +/// - `gendisk` was added to the VFS through a call to +/// `bindings::device_add_disk`. +pub struct GenDisk { + _tagset: Arc>, + gendisk: *mut bindings::gendisk, +} + +// SAFETY: `GenDisk` is an owned pointer to a `struct gendisk` and an `Arc` to a +// `TagSet` It is safe to send this to other threads as long as T is Send. +unsafe impl Send for GenDisk {} + +impl Drop for GenDisk { + fn drop(&mut self) { + // SAFETY: By type invariant, `self.gendisk` points to a valid and + // initialized instance of `struct gendisk`, and it was previously added + // to the VFS. + unsafe { bindings::del_gendisk(self.gendisk) }; + } +} diff --git a/rust/kernel/block/mq/operations.rs b/rust/kernel/block/mq/operations.rs new file mode 100644 index 00000000000000..9ba7fdfeb4b22c --- /dev/null +++ b/rust/kernel/block/mq/operations.rs @@ -0,0 +1,245 @@ +// SPDX-License-Identifier: GPL-2.0 + +//! This module provides an interface for blk-mq drivers to implement. +//! +//! C header: [`include/linux/blk-mq.h`](srctree/include/linux/blk-mq.h) + +use crate::{ + bindings, + block::mq::request::RequestDataWrapper, + block::mq::Request, + error::{from_result, Result}, + types::ARef, +}; +use core::{marker::PhantomData, sync::atomic::AtomicU64, sync::atomic::Ordering}; + +/// Implement this trait to interface blk-mq as block devices. +/// +/// To implement a block device driver, implement this trait as described in the +/// [module level documentation]. The kernel will use the implementation of the +/// functions defined in this trait to interface a block device driver. Note: +/// There is no need for an exit_request() implementation, because the `drop` +/// implementation of the [`Request`] type will be invoked by automatically by +/// the C/Rust glue logic. +/// +/// [module level documentation]: kernel::block::mq +#[macros::vtable] +pub trait Operations: Sized { + /// Called by the kernel to queue a request with the driver. If `is_last` is + /// `false`, the driver is allowed to defer committing the request. + fn queue_rq(rq: ARef>, is_last: bool) -> Result; + + /// Called by the kernel to indicate that queued requests should be submitted. + fn commit_rqs(); + + /// Called by the kernel to poll the device for completed requests. Only + /// used for poll queues. + fn poll() -> bool { + crate::build_error(crate::error::VTABLE_DEFAULT_ERROR) + } +} + +/// A vtable for blk-mq to interact with a block device driver. +/// +/// A `bindings::blk_mq_ops` vtable is constructed from pointers to the `extern +/// "C"` functions of this struct, exposed through the `OperationsVTable::VTABLE`. +/// +/// For general documentation of these methods, see the kernel source +/// documentation related to `struct blk_mq_operations` in +/// [`include/linux/blk-mq.h`]. +/// +/// [`include/linux/blk-mq.h`]: srctree/include/linux/blk-mq.h +pub(crate) struct OperationsVTable(PhantomData); + +impl OperationsVTable { + /// This function is called by the C kernel. A pointer to this function is + /// installed in the `blk_mq_ops` vtable for the driver. + /// + /// # Safety + /// + /// - The caller of this function must ensure that the pointee of `bd` is + /// valid for reads for the duration of this function. + /// - This function must be called for an initialized and live `hctx`. That + /// is, `Self::init_hctx_callback` was called and + /// `Self::exit_hctx_callback()` was not yet called. + /// - `(*bd).rq` must point to an initialized and live `bindings:request`. + /// That is, `Self::init_request_callback` was called but + /// `Self::exit_request_callback` was not yet called for the request. + /// - `(*bd).rq` must be owned by the driver. That is, the block layer must + /// promise to not access the request until the driver calls + /// `bindings::blk_mq_end_request` for the request. + unsafe extern "C" fn queue_rq_callback( + _hctx: *mut bindings::blk_mq_hw_ctx, + bd: *const bindings::blk_mq_queue_data, + ) -> bindings::blk_status_t { + // SAFETY: `bd.rq` is valid as required by the safety requirement for + // this function. + let request = unsafe { &*(*bd).rq.cast::>() }; + + // One refcount for the ARef, one for being in flight + request.wrapper_ref().refcount().store(2, Ordering::Relaxed); + + // SAFETY: + // - We own a refcount that we took above. We pass that to `ARef`. + // - By the safety requirements of this function, `request` is a valid + // `struct request` and the private data is properly initialized. + // - `rq` will be alive until `blk_mq_end_request` is called and is + // reference counted by `ARef` until then. + let rq = unsafe { Request::aref_from_raw((*bd).rq) }; + + // SAFETY: We have exclusive access and we just set the refcount above. + unsafe { Request::start_unchecked(&rq) }; + + let ret = T::queue_rq( + rq, + // SAFETY: `bd` is valid as required by the safety requirement for + // this function. + unsafe { (*bd).last }, + ); + + if let Err(e) = ret { + e.to_blk_status() + } else { + bindings::BLK_STS_OK as _ + } + } + + /// This function is called by the C kernel. A pointer to this function is + /// installed in the `blk_mq_ops` vtable for the driver. + /// + /// # Safety + /// + /// This function may only be called by blk-mq C infrastructure. + unsafe extern "C" fn commit_rqs_callback(_hctx: *mut bindings::blk_mq_hw_ctx) { + T::commit_rqs() + } + + /// This function is called by the C kernel. It is not currently + /// implemented, and there is no way to exercise this code path. + /// + /// # Safety + /// + /// This function may only be called by blk-mq C infrastructure. + unsafe extern "C" fn complete_callback(_rq: *mut bindings::request) {} + + /// This function is called by the C kernel. A pointer to this function is + /// installed in the `blk_mq_ops` vtable for the driver. + /// + /// # Safety + /// + /// This function may only be called by blk-mq C infrastructure. + unsafe extern "C" fn poll_callback( + _hctx: *mut bindings::blk_mq_hw_ctx, + _iob: *mut bindings::io_comp_batch, + ) -> core::ffi::c_int { + T::poll().into() + } + + /// This function is called by the C kernel. A pointer to this function is + /// installed in the `blk_mq_ops` vtable for the driver. + /// + /// # Safety + /// + /// This function may only be called by blk-mq C infrastructure. This + /// function may only be called once before `exit_hctx_callback` is called + /// for the same context. + unsafe extern "C" fn init_hctx_callback( + _hctx: *mut bindings::blk_mq_hw_ctx, + _tagset_data: *mut core::ffi::c_void, + _hctx_idx: core::ffi::c_uint, + ) -> core::ffi::c_int { + from_result(|| Ok(0)) + } + + /// This function is called by the C kernel. A pointer to this function is + /// installed in the `blk_mq_ops` vtable for the driver. + /// + /// # Safety + /// + /// This function may only be called by blk-mq C infrastructure. + unsafe extern "C" fn exit_hctx_callback( + _hctx: *mut bindings::blk_mq_hw_ctx, + _hctx_idx: core::ffi::c_uint, + ) { + } + + /// This function is called by the C kernel. A pointer to this function is + /// installed in the `blk_mq_ops` vtable for the driver. + /// + /// # Safety + /// + /// - This function may only be called by blk-mq C infrastructure. + /// - `_set` must point to an initialized `TagSet`. + /// - `rq` must point to an initialized `bindings::request`. + /// - The allocation pointed to by `rq` must be at the size of `Request` + /// plus the size of `RequestDataWrapper`. + unsafe extern "C" fn init_request_callback( + _set: *mut bindings::blk_mq_tag_set, + rq: *mut bindings::request, + _hctx_idx: core::ffi::c_uint, + _numa_node: core::ffi::c_uint, + ) -> core::ffi::c_int { + from_result(|| { + // SAFETY: By the safety requirements of this function, `rq` points + // to a valid allocation. + let pdu = unsafe { Request::wrapper_ptr(rq.cast::>()) }; + + // SAFETY: The refcount field is allocated but not initialized, so + // it is valid for writes. + unsafe { RequestDataWrapper::refcount_ptr(pdu.as_ptr()).write(AtomicU64::new(0)) }; + + Ok(0) + }) + } + + /// This function is called by the C kernel. A pointer to this function is + /// installed in the `blk_mq_ops` vtable for the driver. + /// + /// # Safety + /// + /// - This function may only be called by blk-mq C infrastructure. + /// - `_set` must point to an initialized `TagSet`. + /// - `rq` must point to an initialized and valid `Request`. + unsafe extern "C" fn exit_request_callback( + _set: *mut bindings::blk_mq_tag_set, + rq: *mut bindings::request, + _hctx_idx: core::ffi::c_uint, + ) { + // SAFETY: The tagset invariants guarantee that all requests are allocated with extra memory + // for the request data. + let pdu = unsafe { bindings::blk_mq_rq_to_pdu(rq) }.cast::(); + + // SAFETY: `pdu` is valid for read and write and is properly initialised. + unsafe { core::ptr::drop_in_place(pdu) }; + } + + const VTABLE: bindings::blk_mq_ops = bindings::blk_mq_ops { + queue_rq: Some(Self::queue_rq_callback), + queue_rqs: None, + commit_rqs: Some(Self::commit_rqs_callback), + get_budget: None, + put_budget: None, + set_rq_budget_token: None, + get_rq_budget_token: None, + timeout: None, + poll: if T::HAS_POLL { + Some(Self::poll_callback) + } else { + None + }, + complete: Some(Self::complete_callback), + init_hctx: Some(Self::init_hctx_callback), + exit_hctx: Some(Self::exit_hctx_callback), + init_request: Some(Self::init_request_callback), + exit_request: Some(Self::exit_request_callback), + cleanup_rq: None, + busy: None, + map_queues: None, + #[cfg(CONFIG_BLK_DEBUG_FS)] + show_rq: None, + }; + + pub(crate) const fn build() -> &'static bindings::blk_mq_ops { + &Self::VTABLE + } +} diff --git a/rust/kernel/block/mq/raw_writer.rs b/rust/kernel/block/mq/raw_writer.rs new file mode 100644 index 00000000000000..9222465d670bfe --- /dev/null +++ b/rust/kernel/block/mq/raw_writer.rs @@ -0,0 +1,55 @@ +// SPDX-License-Identifier: GPL-2.0 + +use core::fmt::{self, Write}; + +use crate::error::Result; +use crate::prelude::EINVAL; + +/// A mutable reference to a byte buffer where a string can be written into. +/// +/// # Invariants +/// +/// `buffer` is always null terminated. +pub(crate) struct RawWriter<'a> { + buffer: &'a mut [u8], + pos: usize, +} + +impl<'a> RawWriter<'a> { + /// Create a new `RawWriter` instance. + fn new(buffer: &'a mut [u8]) -> Result> { + *(buffer.last_mut().ok_or(EINVAL)?) = 0; + + // INVARIANT: We null terminated the buffer above. + Ok(Self { buffer, pos: 0 }) + } + + pub(crate) fn from_array( + a: &'a mut [core::ffi::c_char; N], + ) -> Result> { + Self::new( + // SAFETY: the buffer of `a` is valid for read and write as `u8` for + // at least `N` bytes. + unsafe { core::slice::from_raw_parts_mut(a.as_mut_ptr().cast::(), N) }, + ) + } +} + +impl Write for RawWriter<'_> { + fn write_str(&mut self, s: &str) -> fmt::Result { + let bytes = s.as_bytes(); + let len = bytes.len(); + + // We do not want to overwrite our null terminator + if self.pos + len > self.buffer.len() - 1 { + return Err(fmt::Error); + } + + // INVARIANT: We are not overwriting the last byte + self.buffer[self.pos..self.pos + len].copy_from_slice(bytes); + + self.pos += len; + + Ok(()) + } +} diff --git a/rust/kernel/block/mq/request.rs b/rust/kernel/block/mq/request.rs new file mode 100644 index 00000000000000..a0e22827f3f4ec --- /dev/null +++ b/rust/kernel/block/mq/request.rs @@ -0,0 +1,253 @@ +// SPDX-License-Identifier: GPL-2.0 + +//! This module provides a wrapper for the C `struct request` type. +//! +//! C header: [`include/linux/blk-mq.h`](srctree/include/linux/blk-mq.h) + +use crate::{ + bindings, + block::mq::Operations, + error::Result, + types::{ARef, AlwaysRefCounted, Opaque}, +}; +use core::{ + marker::PhantomData, + ptr::{addr_of_mut, NonNull}, + sync::atomic::{AtomicU64, Ordering}, +}; + +/// A wrapper around a blk-mq `struct request`. This represents an IO request. +/// +/// # Implementation details +/// +/// There are four states for a request that the Rust bindings care about: +/// +/// A) Request is owned by block layer (refcount 0) +/// B) Request is owned by driver but with zero `ARef`s in existence +/// (refcount 1) +/// C) Request is owned by driver with exactly one `ARef` in existence +/// (refcount 2) +/// D) Request is owned by driver with more than one `ARef` in existence +/// (refcount > 2) +/// +/// +/// We need to track A and B to ensure we fail tag to request conversions for +/// requests that are not owned by the driver. +/// +/// We need to track C and D to ensure that it is safe to end the request and hand +/// back ownership to the block layer. +/// +/// The states are tracked through the private `refcount` field of +/// `RequestDataWrapper`. This structure lives in the private data area of the C +/// `struct request`. +/// +/// # Invariants +/// +/// * `self.0` is a valid `struct request` created by the C portion of the kernel. +/// * The private data area associated with this request must be an initialized +/// and valid `RequestDataWrapper`. +/// * `self` is reference counted by atomic modification of +/// self.wrapper_ref().refcount(). +/// +#[repr(transparent)] +pub struct Request(Opaque, PhantomData); + +impl Request { + /// Create an `ARef` from a `struct request` pointer. + /// + /// # Safety + /// + /// * The caller must own a refcount on `ptr` that is transferred to the + /// returned `ARef`. + /// * The type invariants for `Request` must hold for the pointee of `ptr`. + pub(crate) unsafe fn aref_from_raw(ptr: *mut bindings::request) -> ARef { + // INVARIANT: By the safety requirements of this function, invariants are upheld. + // SAFETY: By the safety requirement of this function, we own a + // reference count that we can pass to `ARef`. + unsafe { ARef::from_raw(NonNull::new_unchecked(ptr as *const Self as *mut Self)) } + } + + /// Notify the block layer that a request is going to be processed now. + /// + /// The block layer uses this hook to do proper initializations such as + /// starting the timeout timer. It is a requirement that block device + /// drivers call this function when starting to process a request. + /// + /// # Safety + /// + /// The caller must have exclusive ownership of `self`, that is + /// `self.wrapper_ref().refcount() == 2`. + pub(crate) unsafe fn start_unchecked(this: &ARef) { + // SAFETY: By type invariant, `self.0` is a valid `struct request` and + // we have exclusive access. + unsafe { bindings::blk_mq_start_request(this.0.get()) }; + } + + /// Try to take exclusive ownership of `this` by dropping the refcount to 0. + /// This fails if `this` is not the only `ARef` pointing to the underlying + /// `Request`. + /// + /// If the operation is successful, `Ok` is returned with a pointer to the + /// C `struct request`. If the operation fails, `this` is returned in the + /// `Err` variant. + fn try_set_end(this: ARef) -> Result<*mut bindings::request, ARef> { + // We can race with `TagSet::tag_to_rq` + if let Err(_old) = this.wrapper_ref().refcount().compare_exchange( + 2, + 0, + Ordering::Relaxed, + Ordering::Relaxed, + ) { + return Err(this); + } + + let request_ptr = this.0.get(); + core::mem::forget(this); + + Ok(request_ptr) + } + + /// Notify the block layer that the request has been completed without errors. + /// + /// This function will return `Err` if `this` is not the only `ARef` + /// referencing the request. + pub fn end_ok(this: ARef) -> Result<(), ARef> { + let request_ptr = Self::try_set_end(this)?; + + // SAFETY: By type invariant, `this.0` was a valid `struct request`. The + // success of the call to `try_set_end` guarantees that there are no + // `ARef`s pointing to this request. Therefore it is safe to hand it + // back to the block layer. + unsafe { bindings::blk_mq_end_request(request_ptr, bindings::BLK_STS_OK as _) }; + + Ok(()) + } + + /// Return a pointer to the `RequestDataWrapper` stored in the private area + /// of the request structure. + /// + /// # Safety + /// + /// - `this` must point to a valid allocation of size at least size of + /// `Self` plus size of `RequestDataWrapper`. + pub(crate) unsafe fn wrapper_ptr(this: *mut Self) -> NonNull { + let request_ptr = this.cast::(); + // SAFETY: By safety requirements for this function, `this` is a + // valid allocation. + let wrapper_ptr = + unsafe { bindings::blk_mq_rq_to_pdu(request_ptr).cast::() }; + // SAFETY: By C API contract, wrapper_ptr points to a valid allocation + // and is not null. + unsafe { NonNull::new_unchecked(wrapper_ptr) } + } + + /// Return a reference to the `RequestDataWrapper` stored in the private + /// area of the request structure. + pub(crate) fn wrapper_ref(&self) -> &RequestDataWrapper { + // SAFETY: By type invariant, `self.0` is a valid allocation. Further, + // the private data associated with this request is initialized and + // valid. The existence of `&self` guarantees that the private data is + // valid as a shared reference. + unsafe { Self::wrapper_ptr(self as *const Self as *mut Self).as_ref() } + } +} + +/// A wrapper around data stored in the private area of the C `struct request`. +pub(crate) struct RequestDataWrapper { + /// The Rust request refcount has the following states: + /// + /// - 0: The request is owned by C block layer. + /// - 1: The request is owned by Rust abstractions but there are no ARef references to it. + /// - 2+: There are `ARef` references to the request. + refcount: AtomicU64, +} + +impl RequestDataWrapper { + /// Return a reference to the refcount of the request that is embedding + /// `self`. + pub(crate) fn refcount(&self) -> &AtomicU64 { + &self.refcount + } + + /// Return a pointer to the refcount of the request that is embedding the + /// pointee of `this`. + /// + /// # Safety + /// + /// - `this` must point to a live allocation of at least the size of `Self`. + pub(crate) unsafe fn refcount_ptr(this: *mut Self) -> *mut AtomicU64 { + // SAFETY: Because of the safety requirements of this function, the + // field projection is safe. + unsafe { addr_of_mut!((*this).refcount) } + } +} + +// SAFETY: Exclusive access is thread-safe for `Request`. `Request` has no `&mut +// self` methods and `&self` methods that mutate `self` are internally +// synchronized. +unsafe impl Send for Request {} + +// SAFETY: Shared access is thread-safe for `Request`. `&self` methods that +// mutate `self` are internally synchronized` +unsafe impl Sync for Request {} + +/// Store the result of `op(target.load())` in target, returning new value of +/// target. +fn atomic_relaxed_op_return(target: &AtomicU64, op: impl Fn(u64) -> u64) -> u64 { + let old = target.fetch_update(Ordering::Relaxed, Ordering::Relaxed, |x| Some(op(x))); + + // SAFETY: Because the operation passed to `fetch_update` above always + // return `Some`, `old` will always be `Ok`. + let old = unsafe { old.unwrap_unchecked() }; + + op(old) +} + +/// Store the result of `op(target.load)` in `target` if `target.load() != +/// pred`, returning true if the target was updated. +fn atomic_relaxed_op_unless(target: &AtomicU64, op: impl Fn(u64) -> u64, pred: u64) -> bool { + target + .fetch_update(Ordering::Relaxed, Ordering::Relaxed, |x| { + if x == pred { + None + } else { + Some(op(x)) + } + }) + .is_ok() +} + +// SAFETY: All instances of `Request` are reference counted. This +// implementation of `AlwaysRefCounted` ensure that increments to the ref count +// keeps the object alive in memory at least until a matching reference count +// decrement is executed. +unsafe impl AlwaysRefCounted for Request { + fn inc_ref(&self) { + let refcount = &self.wrapper_ref().refcount(); + + #[cfg_attr(not(CONFIG_DEBUG_MISC), allow(unused_variables))] + let updated = atomic_relaxed_op_unless(refcount, |x| x + 1, 0); + + #[cfg(CONFIG_DEBUG_MISC)] + if !updated { + panic!("Request refcount zero on clone") + } + } + + unsafe fn dec_ref(obj: core::ptr::NonNull) { + // SAFETY: The type invariants of `ARef` guarantee that `obj` is valid + // for read. + let wrapper_ptr = unsafe { Self::wrapper_ptr(obj.as_ptr()).as_ptr() }; + // SAFETY: The type invariant of `Request` guarantees that the private + // data area is initialized and valid. + let refcount = unsafe { &*RequestDataWrapper::refcount_ptr(wrapper_ptr) }; + + #[cfg_attr(not(CONFIG_DEBUG_MISC), allow(unused_variables))] + let new_refcount = atomic_relaxed_op_return(refcount, |x| x - 1); + + #[cfg(CONFIG_DEBUG_MISC)] + if new_refcount == 0 { + panic!("Request reached refcount zero in Rust abstractions"); + } + } +} diff --git a/rust/kernel/block/mq/tag_set.rs b/rust/kernel/block/mq/tag_set.rs new file mode 100644 index 00000000000000..f9a1ca655a35be --- /dev/null +++ b/rust/kernel/block/mq/tag_set.rs @@ -0,0 +1,86 @@ +// SPDX-License-Identifier: GPL-2.0 + +//! This module provides the `TagSet` struct to wrap the C `struct blk_mq_tag_set`. +//! +//! C header: [`include/linux/blk-mq.h`](srctree/include/linux/blk-mq.h) + +use core::pin::Pin; + +use crate::{ + bindings, + block::mq::{operations::OperationsVTable, request::RequestDataWrapper, Operations}, + error, + prelude::PinInit, + try_pin_init, + types::Opaque, +}; +use core::{convert::TryInto, marker::PhantomData}; +use macros::{pin_data, pinned_drop}; + +/// A wrapper for the C `struct blk_mq_tag_set`. +/// +/// `struct blk_mq_tag_set` contains a `struct list_head` and so must be pinned. +/// +/// # Invariants +/// +/// - `inner` is initialized and valid. +#[pin_data(PinnedDrop)] +#[repr(transparent)] +pub struct TagSet { + #[pin] + inner: Opaque, + _p: PhantomData, +} + +impl TagSet { + /// Try to create a new tag set + pub fn new( + nr_hw_queues: u32, + num_tags: u32, + num_maps: u32, + ) -> impl PinInit { + // SAFETY: `blk_mq_tag_set` only contains integers and pointers, which + // all are allowed to be 0. + let tag_set: bindings::blk_mq_tag_set = unsafe { core::mem::zeroed() }; + let tag_set = core::mem::size_of::() + .try_into() + .map(|cmd_size| { + bindings::blk_mq_tag_set { + ops: OperationsVTable::::build(), + nr_hw_queues, + timeout: 0, // 0 means default which is 30Hz in C + numa_node: bindings::NUMA_NO_NODE, + queue_depth: num_tags, + cmd_size, + flags: bindings::BLK_MQ_F_SHOULD_MERGE, + driver_data: core::ptr::null_mut::(), + nr_maps: num_maps, + ..tag_set + } + }); + + try_pin_init!(TagSet { + inner <- PinInit::<_, error::Error>::pin_chain(Opaque::new(tag_set?), |tag_set| { + // SAFETY: we do not move out of `tag_set`. + let tag_set = unsafe { Pin::get_unchecked_mut(tag_set) }; + // SAFETY: `tag_set` is a reference to an initialized `blk_mq_tag_set`. + error::to_result( unsafe { bindings::blk_mq_alloc_tag_set(tag_set.get())}) + }), + _p: PhantomData, + }) + } + + /// Return the pointer to the wrapped `struct blk_mq_tag_set` + pub(crate) fn raw_tag_set(&self) -> *mut bindings::blk_mq_tag_set { + self.inner.get() + } +} + +#[pinned_drop] +impl PinnedDrop for TagSet { + fn drop(self: Pin<&mut Self>) { + // SAFETY: By type invariant `inner` is valid and has been properly + // initialized during construction. + unsafe { bindings::blk_mq_free_tag_set(self.inner.get()) }; + } +} diff --git a/rust/kernel/error.rs b/rust/kernel/error.rs index 55280ae9fe40e1..145f5c3970096f 100644 --- a/rust/kernel/error.rs +++ b/rust/kernel/error.rs @@ -126,6 +126,12 @@ impl Error { self.0 } + #[cfg(CONFIG_BLOCK)] + pub(crate) fn to_blk_status(self) -> bindings::blk_status_t { + // SAFETY: `self.0` is a valid error due to its invariant. + unsafe { bindings::errno_to_blk_status(self.0) } + } + /// Returns the error encoded as a pointer. #[allow(dead_code)] pub(crate) fn to_ptr(self) -> *mut T { diff --git a/rust/kernel/lib.rs b/rust/kernel/lib.rs index fbd91a48ff8bc5..2cf7c6b6f66b9d 100644 --- a/rust/kernel/lib.rs +++ b/rust/kernel/lib.rs @@ -27,6 +27,8 @@ compile_error!("Missing kernel configuration for conditional compilation"); extern crate self as kernel; pub mod alloc; +#[cfg(CONFIG_BLOCK)] +pub mod block; mod build_assert; pub mod error; pub mod init; From bc5b533b91ef0b8a09fe507e23d1c6c43c1fb0f5 Mon Sep 17 00:00:00 2001 From: Andreas Hindborg Date: Tue, 11 Jun 2024 13:45:50 +0200 Subject: [PATCH 021/209] rust: block: add rnull, Rust null_blk implementation This patch adds an initial version of the Rust null block driver. Signed-off-by: Andreas Hindborg Reviewed-by: Benno Lossin Link: https://lore.kernel.org/r/20240611114551.228679-3-nmi@metaspace.dk Signed-off-by: Jens Axboe --- drivers/block/Kconfig | 9 ++++++ drivers/block/Makefile | 3 ++ drivers/block/rnull.rs | 73 ++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 85 insertions(+) create mode 100644 drivers/block/rnull.rs diff --git a/drivers/block/Kconfig b/drivers/block/Kconfig index 5b9d4aaebb81d4..ed209f4f279839 100644 --- a/drivers/block/Kconfig +++ b/drivers/block/Kconfig @@ -354,6 +354,15 @@ config VIRTIO_BLK This is the virtual block driver for virtio. It can be used with QEMU based VMMs (like KVM or Xen). Say Y or M. +config BLK_DEV_RUST_NULL + tristate "Rust null block driver (Experimental)" + depends on RUST + help + This is the Rust implementation of the null block driver. For now it + is only a minimal stub. + + If unsure, say N. + config BLK_DEV_RBD tristate "Rados block device (RBD)" depends on INET && BLOCK diff --git a/drivers/block/Makefile b/drivers/block/Makefile index 101612cba303a3..1105a2d4fdcb00 100644 --- a/drivers/block/Makefile +++ b/drivers/block/Makefile @@ -9,6 +9,9 @@ # needed for trace events ccflags-y += -I$(src) +obj-$(CONFIG_BLK_DEV_RUST_NULL) += rnull_mod.o +rnull_mod-y := rnull.o + obj-$(CONFIG_MAC_FLOPPY) += swim3.o obj-$(CONFIG_BLK_DEV_SWIM) += swim_mod.o obj-$(CONFIG_BLK_DEV_FD) += floppy.o diff --git a/drivers/block/rnull.rs b/drivers/block/rnull.rs new file mode 100644 index 00000000000000..b0227cf9ddd387 --- /dev/null +++ b/drivers/block/rnull.rs @@ -0,0 +1,73 @@ +// SPDX-License-Identifier: GPL-2.0 + +//! This is a Rust implementation of the C null block driver. +//! +//! Supported features: +//! +//! - blk-mq interface +//! - direct completion +//! - block size 4k +//! +//! The driver is not configurable. + +use kernel::{ + alloc::flags, + block::mq::{ + self, + gen_disk::{self, GenDisk}, + Operations, TagSet, + }, + error::Result, + new_mutex, pr_info, + prelude::*, + sync::{Arc, Mutex}, + types::ARef, +}; + +module! { + type: NullBlkModule, + name: "rnull_mod", + author: "Andreas Hindborg", + license: "GPL v2", +} + +struct NullBlkModule { + _disk: Pin>>>, +} + +impl kernel::Module for NullBlkModule { + fn init(_module: &'static ThisModule) -> Result { + pr_info!("Rust null_blk loaded\n"); + let tagset = Arc::pin_init(TagSet::new(1, 256, 1), flags::GFP_KERNEL)?; + + let disk = gen_disk::GenDiskBuilder::new() + .capacity_sectors(4096 << 11) + .logical_block_size(4096)? + .physical_block_size(4096)? + .rotational(false) + .build(format_args!("rnullb{}", 0), tagset)?; + + let disk = Box::pin_init(new_mutex!(disk, "nullb:disk"), flags::GFP_KERNEL)?; + + Ok(Self { _disk: disk }) + } +} + +struct NullBlkDevice; + +#[vtable] +impl Operations for NullBlkDevice { + #[inline(always)] + fn queue_rq(rq: ARef>, _is_last: bool) -> Result { + mq::Request::end_ok(rq) + .map_err(|_e| kernel::error::code::EIO) + // We take no refcounts on the request, so we expect to be able to + // end the request. The request reference must be unique at this + // point, and so `end_ok` cannot fail. + .expect("Fatal error - expected to be able to end request"); + + Ok(()) + } + + fn commit_rqs() {} +} From d37a9ab8331cfc0fc2eac0480f0af624c0144a92 Mon Sep 17 00:00:00 2001 From: Andreas Hindborg Date: Tue, 11 Jun 2024 13:45:51 +0200 Subject: [PATCH 022/209] MAINTAINERS: add entry for Rust block device driver API Add an entry for the Rust block device driver abstractions. Signed-off-by: Andreas Hindborg Link: https://lore.kernel.org/r/20240611114551.228679-4-nmi@metaspace.dk Signed-off-by: Jens Axboe --- MAINTAINERS | 14 ++++++++++++++ 1 file changed, 14 insertions(+) diff --git a/MAINTAINERS b/MAINTAINERS index aacccb376c28a1..aa1321fdc30010 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -3781,6 +3781,20 @@ F: include/linux/blk* F: kernel/trace/blktrace.c F: lib/sbitmap.c +BLOCK LAYER DEVICE DRIVER API [RUST] +M: Andreas Hindborg +R: Boqun Feng +L: linux-block@vger.kernel.org +L: rust-for-linux@vger.kernel.org +S: Supported +W: https://rust-for-linux.com +B: https://github.com/Rust-for-Linux/linux/issues +C: https://rust-for-linux.zulipchat.com/#narrow/stream/Block +T: git https://github.com/Rust-for-Linux/linux.git rust-block-next +F: drivers/block/rnull.rs +F: rust/kernel/block.rs +F: rust/kernel/block/ + BLOCK2MTD DRIVER M: Joern Engel L: linux-mtd@lists.infradead.org From 5db755fbb1a0de4a4cfd5d5edfaa19853b9c56e6 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:47:56 +0200 Subject: [PATCH 023/209] ubd: refactor the interrupt handler Instead of a separate handler function that leaves no work in the interrupt hanler itself, split out a per-request end I/O helper and clean up the coding style and variable naming while we're at it. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Acked-By: Anton Ivanov Link: https://lore.kernel.org/r/20240531074837.1648501-2-hch@lst.de Signed-off-by: Jens Axboe --- arch/um/drivers/ubd_kern.c | 49 ++++++++++++++------------------------ 1 file changed, 18 insertions(+), 31 deletions(-) diff --git a/arch/um/drivers/ubd_kern.c b/arch/um/drivers/ubd_kern.c index ef805eaa9e013d..0c9542d58c01b7 100644 --- a/arch/um/drivers/ubd_kern.c +++ b/arch/um/drivers/ubd_kern.c @@ -447,43 +447,30 @@ static int bulk_req_safe_read( return n; } -/* Called without dev->lock held, and only in interrupt context. */ -static void ubd_handler(void) +static void ubd_end_request(struct io_thread_req *io_req) { - int n; - int count; - - while(1){ - n = bulk_req_safe_read( - thread_fd, - irq_req_buffer, - &irq_remainder, - &irq_remainder_size, - UBD_REQ_BUFFER_SIZE - ); - if (n < 0) { - if(n == -EAGAIN) - break; - printk(KERN_ERR "spurious interrupt in ubd_handler, " - "err = %d\n", -n); - return; - } - for (count = 0; count < n/sizeof(struct io_thread_req *); count++) { - struct io_thread_req *io_req = (*irq_req_buffer)[count]; - - if ((io_req->error == BLK_STS_NOTSUPP) && (req_op(io_req->req) == REQ_OP_DISCARD)) { - blk_queue_max_discard_sectors(io_req->req->q, 0); - blk_queue_max_write_zeroes_sectors(io_req->req->q, 0); - } - blk_mq_end_request(io_req->req, io_req->error); - kfree(io_req); - } + if (io_req->error == BLK_STS_NOTSUPP && + req_op(io_req->req) == REQ_OP_DISCARD) { + blk_queue_max_discard_sectors(io_req->req->q, 0); + blk_queue_max_write_zeroes_sectors(io_req->req->q, 0); } + blk_mq_end_request(io_req->req, io_req->error); + kfree(io_req); } static irqreturn_t ubd_intr(int irq, void *dev) { - ubd_handler(); + int len, i; + + while ((len = bulk_req_safe_read(thread_fd, irq_req_buffer, + &irq_remainder, &irq_remainder_size, + UBD_REQ_BUFFER_SIZE)) >= 0) { + for (i = 0; i < len / sizeof(struct io_thread_req *); i++) + ubd_end_request((*irq_req_buffer)[i]); + } + + if (len < 0 && len != -EAGAIN) + pr_err("spurious interrupt in %s, err = %d\n", __func__, len); return IRQ_HANDLED; } From 31ade7d4fdcf382beb8cb229a1f5d77e0f239672 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:47:57 +0200 Subject: [PATCH 024/209] ubd: untagle discard vs write zeroes not support handling Discard and Write Zeroes are different operation and implemented by different fallocate opcodes for ubd. If one fails the other one can work and vice versa. Split the code to disable the operations in ubd_handler to only disable the operation that actually failed. Fixes: 50109b5a03b4 ("um: Add support for DISCARD in the UBD Driver") Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: Martin K. Petersen Acked-By: Anton Ivanov Link: https://lore.kernel.org/r/20240531074837.1648501-3-hch@lst.de Signed-off-by: Jens Axboe --- arch/um/drivers/ubd_kern.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/arch/um/drivers/ubd_kern.c b/arch/um/drivers/ubd_kern.c index 0c9542d58c01b7..093c87879d08ba 100644 --- a/arch/um/drivers/ubd_kern.c +++ b/arch/um/drivers/ubd_kern.c @@ -449,10 +449,11 @@ static int bulk_req_safe_read( static void ubd_end_request(struct io_thread_req *io_req) { - if (io_req->error == BLK_STS_NOTSUPP && - req_op(io_req->req) == REQ_OP_DISCARD) { - blk_queue_max_discard_sectors(io_req->req->q, 0); - blk_queue_max_write_zeroes_sectors(io_req->req->q, 0); + if (io_req->error == BLK_STS_NOTSUPP) { + if (req_op(io_req->req) == REQ_OP_DISCARD) + blk_queue_max_discard_sectors(io_req->req->q, 0); + else if (req_op(io_req->req) == REQ_OP_WRITE_ZEROES) + blk_queue_max_write_zeroes_sectors(io_req->req->q, 0); } blk_mq_end_request(io_req->req, io_req->error); kfree(io_req); From a00d4bfce7c6d7fa4712b8133ec195c9bd142ae6 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:47:58 +0200 Subject: [PATCH 025/209] rbd: increase io_opt again Commit 16d80c54ad42 ("rbd: set io_min, io_opt and discard_granularity to alloc_size") lowered the io_opt size for rbd from objset_bytes which is 4MB for typical setup to alloc_size which is typically 64KB. The commit mostly talks about discard behavior and does mention io_min in passing. Reducing io_opt means reducing the readahead size, which seems counter-intuitive given that rbd currently abuses the user max_sectors setting to actually increase the I/O size. Switch back to the old setting to allow larger reads (the readahead size despite it's name actually limits the size of any buffered read) and to prepare for using io_opt in the max_sectors calculation and getting drivers out of the business of overriding the max_user_sectors value. Signed-off-by: Christoph Hellwig Acked-by: Ilya Dryomov Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-4-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/rbd.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c index 26ff5cd2bf0abc..46dc487ccc17eb 100644 --- a/drivers/block/rbd.c +++ b/drivers/block/rbd.c @@ -4955,8 +4955,8 @@ static int rbd_init_disk(struct rbd_device *rbd_dev) struct queue_limits lim = { .max_hw_sectors = objset_bytes >> SECTOR_SHIFT, .max_user_sectors = objset_bytes >> SECTOR_SHIFT, + .io_opt = objset_bytes, .io_min = rbd_dev->opts->alloc_size, - .io_opt = rbd_dev->opts->alloc_size, .max_segments = USHRT_MAX, .max_segment_size = UINT_MAX, }; From a23634644afc2f7c1bac98776440a1f3b161819e Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:47:59 +0200 Subject: [PATCH 026/209] block: take io_opt and io_min into account for max_sectors The soft max_sectors limit is normally capped by the hardware limits and an arbitrary upper limit enforced by the kernel, but can be modified by the user. A few drivers want to increase this limit (nbd, rbd) or adjust it up or down based on hardware capabilities (sd). Change blk_validate_limits to default max_sectors to the optimal I/O size, or upgrade it to the preferred minimal I/O size if that is larger than the kernel default if no optimal I/O size is provided based on the logic in the SD driver. This keeps the existing kernel default for drivers that do not provide an io_opt or very big io_min value, but picks a much more useful default for those who provide these hints, and allows to remove the hacks to set the user max_sectors limit in nbd, rbd and sd. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Acked-by: Ilya Dryomov Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-5-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 7 +++++++ drivers/block/nbd.c | 2 +- drivers/block/rbd.c | 1 - drivers/scsi/sd.c | 29 +++++------------------------ 4 files changed, 13 insertions(+), 26 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index effeb9a639bb45..a49abdb3554834 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -153,6 +153,13 @@ static int blk_validate_limits(struct queue_limits *lim) if (lim->max_user_sectors < PAGE_SIZE / SECTOR_SIZE) return -EINVAL; lim->max_sectors = min(max_hw_sectors, lim->max_user_sectors); + } else if (lim->io_opt) { + lim->max_sectors = + min(max_hw_sectors, lim->io_opt >> SECTOR_SHIFT); + } else if (lim->io_min && + lim->io_min > (BLK_DEF_MAX_SECTORS_CAP << SECTOR_SHIFT)) { + lim->max_sectors = + min(max_hw_sectors, lim->io_min >> SECTOR_SHIFT); } else { lim->max_sectors = min(max_hw_sectors, BLK_DEF_MAX_SECTORS_CAP); } diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c index 22a79a62cc4eab..ad887d614d5b3f 100644 --- a/drivers/block/nbd.c +++ b/drivers/block/nbd.c @@ -1808,7 +1808,7 @@ static struct nbd_device *nbd_dev_add(int index, unsigned int refs) { struct queue_limits lim = { .max_hw_sectors = 65536, - .max_user_sectors = 256, + .io_opt = 256 << SECTOR_SHIFT, .max_segments = USHRT_MAX, .max_segment_size = UINT_MAX, }; diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c index 46dc487ccc17eb..22ad704f81d8b9 100644 --- a/drivers/block/rbd.c +++ b/drivers/block/rbd.c @@ -4954,7 +4954,6 @@ static int rbd_init_disk(struct rbd_device *rbd_dev) rbd_dev->layout.object_size * rbd_dev->layout.stripe_count; struct queue_limits lim = { .max_hw_sectors = objset_bytes >> SECTOR_SHIFT, - .max_user_sectors = objset_bytes >> SECTOR_SHIFT, .io_opt = objset_bytes, .io_min = rbd_dev->opts->alloc_size, .max_segments = USHRT_MAX, diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index f6c822c9cbd2d3..3dff9150ce11e2 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -3593,7 +3593,7 @@ static int sd_revalidate_disk(struct gendisk *disk) struct request_queue *q = sdkp->disk->queue; sector_t old_capacity = sdkp->capacity; unsigned char *buffer; - unsigned int dev_max, rw_max; + unsigned int dev_max; SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_revalidate_disk\n")); @@ -3675,34 +3675,15 @@ static int sd_revalidate_disk(struct gendisk *disk) else blk_queue_io_min(sdkp->disk->queue, 0); - if (sd_validate_opt_xfer_size(sdkp, dev_max)) { - q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks); - rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks); - } else { - q->limits.io_opt = 0; - rw_max = min_not_zero(logical_to_sectors(sdp, dev_max), - (sector_t)BLK_DEF_MAX_SECTORS_CAP); - } - /* * Limit default to SCSI host optimal sector limit if set. There may be * an impact on performance for when the size of a request exceeds this * host limit. */ - rw_max = min_not_zero(rw_max, sdp->host->opt_sectors); - - /* Do not exceed controller limit */ - rw_max = min(rw_max, queue_max_hw_sectors(q)); - - /* - * Only update max_sectors if previously unset or if the current value - * exceeds the capabilities of the hardware. - */ - if (sdkp->first_scan || - q->limits.max_sectors > q->limits.max_dev_sectors || - q->limits.max_sectors > q->limits.max_hw_sectors) { - q->limits.max_sectors = rw_max; - q->limits.max_user_sectors = rw_max; + q->limits.io_opt = sdp->host->opt_sectors << SECTOR_SHIFT; + if (sd_validate_opt_xfer_size(sdkp, dev_max)) { + q->limits.io_opt = min_not_zero(q->limits.io_opt, + logical_to_bytes(sdp, sdkp->opt_xfer_blocks)); } sdkp->first_scan = 0; From b3491b0db165c0cbe25874da66d97652c03db654 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:48:00 +0200 Subject: [PATCH 027/209] sd: simplify the ZBC case in provisioning_mode_store Don't reset the discard settings to no-op over and over when a user writes to the provisioning attribute as that is already the default mode for ZBC devices. In hindsight we should have made writing to the attribute fail for ZBC devices, but the code has probably been around for far too long to change this now. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-6-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/sd.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 3dff9150ce11e2..83aa17fea39d39 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -461,14 +461,13 @@ provisioning_mode_store(struct device *dev, struct device_attribute *attr, if (!capable(CAP_SYS_ADMIN)) return -EACCES; - if (sd_is_zoned(sdkp)) { - sd_config_discard(sdkp, SD_LBP_DISABLE); - return count; - } - if (sdp->type != TYPE_DISK) return -EINVAL; + /* ignore the provisioning mode for ZBC devices */ + if (sd_is_zoned(sdkp)) + return count; + mode = sysfs_match_string(lbp_mode, buf); if (mode < 0) return -EINVAL; From b0dadb86a90bd5a7b723f9d3a6cf69da9b596496 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:48:01 +0200 Subject: [PATCH 028/209] sd: add a sd_disable_discard helper Add helper to disable discard when it is not supported and use it instead of sd_config_discard in the I/O completion handler. This avoids touching more fields than required in the I/O completion handler and prepares for converting sd to use the atomic queue limits API. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-7-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/sd.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 83aa17fea39d39..f07d90474e682b 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -821,6 +821,12 @@ static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd, return protect; } +static void sd_disable_discard(struct scsi_disk *sdkp) +{ + sdkp->provisioning_mode = SD_LBP_DISABLE; + blk_queue_max_discard_sectors(sdkp->disk->queue, 0); +} + static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode) { struct request_queue *q = sdkp->disk->queue; @@ -2245,12 +2251,12 @@ static int sd_done(struct scsi_cmnd *SCpnt) case 0x24: /* INVALID FIELD IN CDB */ switch (SCpnt->cmnd[0]) { case UNMAP: - sd_config_discard(sdkp, SD_LBP_DISABLE); + sd_disable_discard(sdkp); break; case WRITE_SAME_16: case WRITE_SAME: if (SCpnt->cmnd[1] & 8) { /* UNMAP */ - sd_config_discard(sdkp, SD_LBP_DISABLE); + sd_disable_discard(sdkp); } else { sdkp->device->no_write_same = 1; sd_config_write_same(sdkp); From 9972b8ce0d4ba373901bdd1e15e4de58fcd7f662 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:48:02 +0200 Subject: [PATCH 029/209] sd: add a sd_disable_write_same helper Add helper to disable WRITE SAME when it is not supported and use it instead of sd_config_write_same in the I/O completion handler. This avoids touching more fields than required in the I/O completion handler and prepares for converting sd to use the atomic queue limits API. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-8-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/sd.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index f07d90474e682b..70211d0b187652 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -1004,6 +1004,13 @@ static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd) return sd_setup_write_same10_cmnd(cmd, false); } +static void sd_disable_write_same(struct scsi_disk *sdkp) +{ + sdkp->device->no_write_same = 1; + sdkp->max_ws_blocks = 0; + blk_queue_max_write_zeroes_sectors(sdkp->disk->queue, 0); +} + static void sd_config_write_same(struct scsi_disk *sdkp) { struct request_queue *q = sdkp->disk->queue; @@ -2258,8 +2265,7 @@ static int sd_done(struct scsi_cmnd *SCpnt) if (SCpnt->cmnd[1] & 8) { /* UNMAP */ sd_disable_discard(sdkp); } else { - sdkp->device->no_write_same = 1; - sd_config_write_same(sdkp); + sd_disable_write_same(sdkp); req->rq_flags |= RQF_QUIET; } break; From d15b9bd42cd3b2077812d4bf32f532a9bd5c4914 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:48:03 +0200 Subject: [PATCH 030/209] sd: simplify the disable case in sd_config_discard Fall through to the main call to blk_queue_max_discard_sectors given that max_blocks has been initialized to zero above instead of duplicating the call. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-9-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/sd.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 70211d0b187652..0dbc6eb7a7cac3 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -844,8 +844,7 @@ static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode) case SD_LBP_FULL: case SD_LBP_DISABLE: - blk_queue_max_discard_sectors(q, 0); - return; + break; case SD_LBP_UNMAP: max_blocks = min_not_zero(sdkp->max_unmap_blocks, From f1e8185fc12c699c3abf4f39b1ff5d7793da3a95 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:48:04 +0200 Subject: [PATCH 031/209] sd: factor out a sd_discard_mode helper Split the logic to pick the right discard mode into a little helper to prepare for further changes. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-10-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/sd.c | 37 ++++++++++++++++++++----------------- 1 file changed, 20 insertions(+), 17 deletions(-) diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 0dbc6eb7a7cac3..39eddfac09ef8f 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -3201,6 +3201,25 @@ static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer) return; } +static unsigned int sd_discard_mode(struct scsi_disk *sdkp) +{ + if (!sdkp->lbpvpd) { + /* LBP VPD page not provided */ + if (sdkp->max_unmap_blocks) + return SD_LBP_UNMAP; + return SD_LBP_WS16; + } + + /* LBP VPD page tells us what to use */ + if (sdkp->lbpu && sdkp->max_unmap_blocks) + return SD_LBP_UNMAP; + if (sdkp->lbpws) + return SD_LBP_WS16; + if (sdkp->lbpws10) + return SD_LBP_WS10; + return SD_LBP_DISABLE; +} + /** * sd_read_block_limits - Query disk device for preferred I/O sizes. * @sdkp: disk to query @@ -3239,23 +3258,7 @@ static void sd_read_block_limits(struct scsi_disk *sdkp) sdkp->unmap_alignment = get_unaligned_be32(&vpd->data[32]) & ~(1 << 31); - if (!sdkp->lbpvpd) { /* LBP VPD page not provided */ - - if (sdkp->max_unmap_blocks) - sd_config_discard(sdkp, SD_LBP_UNMAP); - else - sd_config_discard(sdkp, SD_LBP_WS16); - - } else { /* LBP VPD page tells us what to use */ - if (sdkp->lbpu && sdkp->max_unmap_blocks) - sd_config_discard(sdkp, SD_LBP_UNMAP); - else if (sdkp->lbpws) - sd_config_discard(sdkp, SD_LBP_WS16); - else if (sdkp->lbpws10) - sd_config_discard(sdkp, SD_LBP_WS10); - else - sd_config_discard(sdkp, SD_LBP_DISABLE); - } + sd_config_discard(sdkp, sd_discard_mode(sdkp)); } out: From 9c1d339a1bf45f4d3a2e77bbf24b0ec51f02551c Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:48:05 +0200 Subject: [PATCH 032/209] sd: cleanup zoned queue limits initialization Consolidate setting zone-related queue limits in sd_zbc_read_zones instead of splitting them between sd_zbc_revalidate_zones and sd_zbc_read_zones, and move the early_zone_information initialization in sd_zbc_read_zones above setting up the queue limits. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-11-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/sd_zbc.c | 18 ++++++++---------- 1 file changed, 8 insertions(+), 10 deletions(-) diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c index 806036e48abeda..1c24c844e8d178 100644 --- a/drivers/scsi/sd_zbc.c +++ b/drivers/scsi/sd_zbc.c @@ -565,12 +565,6 @@ int sd_zbc_revalidate_zones(struct scsi_disk *sdkp) sdkp->zone_info.zone_blocks = zone_blocks; sdkp->zone_info.nr_zones = nr_zones; - blk_queue_chunk_sectors(q, - logical_to_sectors(sdkp->device, zone_blocks)); - - /* Enable block layer zone append emulation */ - blk_queue_max_zone_append_sectors(q, 0); - flags = memalloc_noio_save(); ret = blk_revalidate_disk_zones(disk); memalloc_noio_restore(flags); @@ -625,6 +619,10 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, u8 buf[SD_BUF_SIZE]) if (ret != 0) goto err; + nr_zones = round_up(sdkp->capacity, zone_blocks) >> ilog2(zone_blocks); + sdkp->early_zone_info.nr_zones = nr_zones; + sdkp->early_zone_info.zone_blocks = zone_blocks; + /* The drive satisfies the kernel restrictions: set it up */ blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q); if (sdkp->zones_max_open == U32_MAX) @@ -632,10 +630,10 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, u8 buf[SD_BUF_SIZE]) else disk_set_max_open_zones(disk, sdkp->zones_max_open); disk_set_max_active_zones(disk, 0); - nr_zones = round_up(sdkp->capacity, zone_blocks) >> ilog2(zone_blocks); - - sdkp->early_zone_info.nr_zones = nr_zones; - sdkp->early_zone_info.zone_blocks = zone_blocks; + blk_queue_chunk_sectors(q, + logical_to_sectors(sdkp->device, zone_blocks)); + /* Enable block layer zone append emulation */ + blk_queue_max_zone_append_sectors(q, 0); return 0; From 804e498e0496d889090f32f812b5ce1a4f2aa63e Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:48:06 +0200 Subject: [PATCH 033/209] sd: convert to the atomic queue limits API Assign all queue limits through a local queue_limits variable and queue_limits_commit_update so that we can't race updating them from multiple places, and freeze the queue when updating them so that in-progress I/O submissions don't see half-updated limits. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: John Garry Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-12-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/sd.c | 130 ++++++++++++++++++++++++------------------ drivers/scsi/sd.h | 6 +- drivers/scsi/sd_zbc.c | 15 ++--- 3 files changed, 85 insertions(+), 66 deletions(-) diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 39eddfac09ef8f..049071b5681989 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -101,12 +101,13 @@ MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC); #define SD_MINORS 16 -static void sd_config_discard(struct scsi_disk *, unsigned int); -static void sd_config_write_same(struct scsi_disk *); +static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim, + unsigned int mode); +static void sd_config_write_same(struct scsi_disk *sdkp, + struct queue_limits *lim); static int sd_revalidate_disk(struct gendisk *); static void sd_unlock_native_capacity(struct gendisk *disk); static void sd_shutdown(struct device *); -static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer); static void scsi_disk_release(struct device *cdev); static DEFINE_IDA(sd_index_ida); @@ -456,7 +457,8 @@ provisioning_mode_store(struct device *dev, struct device_attribute *attr, { struct scsi_disk *sdkp = to_scsi_disk(dev); struct scsi_device *sdp = sdkp->device; - int mode; + struct queue_limits lim; + int mode, err; if (!capable(CAP_SYS_ADMIN)) return -EACCES; @@ -472,8 +474,13 @@ provisioning_mode_store(struct device *dev, struct device_attribute *attr, if (mode < 0) return -EINVAL; - sd_config_discard(sdkp, mode); - + lim = queue_limits_start_update(sdkp->disk->queue); + sd_config_discard(sdkp, &lim, mode); + blk_mq_freeze_queue(sdkp->disk->queue); + err = queue_limits_commit_update(sdkp->disk->queue, &lim); + blk_mq_unfreeze_queue(sdkp->disk->queue); + if (err) + return err; return count; } static DEVICE_ATTR_RW(provisioning_mode); @@ -556,6 +563,7 @@ max_write_same_blocks_store(struct device *dev, struct device_attribute *attr, { struct scsi_disk *sdkp = to_scsi_disk(dev); struct scsi_device *sdp = sdkp->device; + struct queue_limits lim; unsigned long max; int err; @@ -577,8 +585,13 @@ max_write_same_blocks_store(struct device *dev, struct device_attribute *attr, sdkp->max_ws_blocks = max; } - sd_config_write_same(sdkp); - + lim = queue_limits_start_update(sdkp->disk->queue); + sd_config_write_same(sdkp, &lim); + blk_mq_freeze_queue(sdkp->disk->queue); + err = queue_limits_commit_update(sdkp->disk->queue, &lim); + blk_mq_unfreeze_queue(sdkp->disk->queue); + if (err) + return err; return count; } static DEVICE_ATTR_RW(max_write_same_blocks); @@ -827,17 +840,15 @@ static void sd_disable_discard(struct scsi_disk *sdkp) blk_queue_max_discard_sectors(sdkp->disk->queue, 0); } -static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode) +static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim, + unsigned int mode) { - struct request_queue *q = sdkp->disk->queue; unsigned int logical_block_size = sdkp->device->sector_size; unsigned int max_blocks = 0; - q->limits.discard_alignment = - sdkp->unmap_alignment * logical_block_size; - q->limits.discard_granularity = - max(sdkp->physical_block_size, - sdkp->unmap_granularity * logical_block_size); + lim->discard_alignment = sdkp->unmap_alignment * logical_block_size; + lim->discard_granularity = max(sdkp->physical_block_size, + sdkp->unmap_granularity * logical_block_size); sdkp->provisioning_mode = mode; switch (mode) { @@ -875,7 +886,8 @@ static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode) break; } - blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9)); + lim->max_hw_discard_sectors = max_blocks * + (logical_block_size >> SECTOR_SHIFT); } static void *sd_set_special_bvec(struct request *rq, unsigned int data_len) @@ -1010,9 +1022,9 @@ static void sd_disable_write_same(struct scsi_disk *sdkp) blk_queue_max_write_zeroes_sectors(sdkp->disk->queue, 0); } -static void sd_config_write_same(struct scsi_disk *sdkp) +static void sd_config_write_same(struct scsi_disk *sdkp, + struct queue_limits *lim) { - struct request_queue *q = sdkp->disk->queue; unsigned int logical_block_size = sdkp->device->sector_size; if (sdkp->device->no_write_same) { @@ -1066,8 +1078,8 @@ static void sd_config_write_same(struct scsi_disk *sdkp) } out: - blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks * - (logical_block_size >> 9)); + lim->max_write_zeroes_sectors = + sdkp->max_ws_blocks * (logical_block_size >> SECTOR_SHIFT); } static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd) @@ -2523,7 +2535,7 @@ static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp, #define READ_CAPACITY_RETRIES_ON_RESET 10 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp, - unsigned char *buffer) + struct queue_limits *lim, unsigned char *buffer) { unsigned char cmd[16]; struct scsi_sense_hdr sshdr; @@ -2597,7 +2609,7 @@ static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp, /* Lowest aligned logical block */ alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size; - blk_queue_alignment_offset(sdp->request_queue, alignment); + lim->alignment_offset = alignment; if (alignment && sdkp->first_scan) sd_printk(KERN_NOTICE, sdkp, "physical block alignment offset: %u\n", alignment); @@ -2608,7 +2620,7 @@ static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp, if (buffer[14] & 0x40) /* LBPRZ */ sdkp->lbprz = 1; - sd_config_discard(sdkp, SD_LBP_WS16); + sd_config_discard(sdkp, lim, SD_LBP_WS16); } sdkp->capacity = lba + 1; @@ -2711,13 +2723,14 @@ static int sd_try_rc16_first(struct scsi_device *sdp) * read disk capacity */ static void -sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer) +sd_read_capacity(struct scsi_disk *sdkp, struct queue_limits *lim, + unsigned char *buffer) { int sector_size; struct scsi_device *sdp = sdkp->device; if (sd_try_rc16_first(sdp)) { - sector_size = read_capacity_16(sdkp, sdp, buffer); + sector_size = read_capacity_16(sdkp, sdp, lim, buffer); if (sector_size == -EOVERFLOW) goto got_data; if (sector_size == -ENODEV) @@ -2737,7 +2750,7 @@ sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer) int old_sector_size = sector_size; sd_printk(KERN_NOTICE, sdkp, "Very big device. " "Trying to use READ CAPACITY(16).\n"); - sector_size = read_capacity_16(sdkp, sdp, buffer); + sector_size = read_capacity_16(sdkp, sdp, lim, buffer); if (sector_size < 0) { sd_printk(KERN_NOTICE, sdkp, "Using 0xffffffff as device size\n"); @@ -2796,9 +2809,8 @@ sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer) */ sector_size = 512; } - blk_queue_logical_block_size(sdp->request_queue, sector_size); - blk_queue_physical_block_size(sdp->request_queue, - sdkp->physical_block_size); + lim->logical_block_size = sector_size; + lim->physical_block_size = sdkp->physical_block_size; sdkp->device->sector_size = sector_size; if (sdkp->capacity > 0xffffffff) @@ -3220,11 +3232,11 @@ static unsigned int sd_discard_mode(struct scsi_disk *sdkp) return SD_LBP_DISABLE; } -/** - * sd_read_block_limits - Query disk device for preferred I/O sizes. - * @sdkp: disk to query +/* + * Query disk device for preferred I/O sizes. */ -static void sd_read_block_limits(struct scsi_disk *sdkp) +static void sd_read_block_limits(struct scsi_disk *sdkp, + struct queue_limits *lim) { struct scsi_vpd *vpd; @@ -3258,7 +3270,7 @@ static void sd_read_block_limits(struct scsi_disk *sdkp) sdkp->unmap_alignment = get_unaligned_be32(&vpd->data[32]) & ~(1 << 31); - sd_config_discard(sdkp, sd_discard_mode(sdkp)); + sd_config_discard(sdkp, lim, sd_discard_mode(sdkp)); } out: @@ -3277,11 +3289,9 @@ static void sd_read_block_limits_ext(struct scsi_disk *sdkp) rcu_read_unlock(); } -/** - * sd_read_block_characteristics - Query block dev. characteristics - * @sdkp: disk to query - */ -static void sd_read_block_characteristics(struct scsi_disk *sdkp) +/* Query block device characteristics */ +static void sd_read_block_characteristics(struct scsi_disk *sdkp, + struct queue_limits *lim) { struct request_queue *q = sdkp->disk->queue; struct scsi_vpd *vpd; @@ -3307,29 +3317,26 @@ static void sd_read_block_characteristics(struct scsi_disk *sdkp) #ifdef CONFIG_BLK_DEV_ZONED /* sd_probe rejects ZBD devices early otherwise */ if (sdkp->device->type == TYPE_ZBC) { - /* - * Host-managed. - */ - disk_set_zoned(sdkp->disk); + lim->zoned = true; /* * Per ZBC and ZAC specifications, writes in sequential write * required zones of host-managed devices must be aligned to * the device physical block size. */ - blk_queue_zone_write_granularity(q, sdkp->physical_block_size); + lim->zone_write_granularity = sdkp->physical_block_size; } else { /* * Host-aware devices are treated as conventional. */ - WARN_ON_ONCE(blk_queue_is_zoned(q)); + lim->zoned = false; } #endif /* CONFIG_BLK_DEV_ZONED */ if (!sdkp->first_scan) return; - if (blk_queue_is_zoned(q)) + if (lim->zoned) sd_printk(KERN_NOTICE, sdkp, "Host-managed zoned block device\n"); else if (sdkp->zoned == 1) sd_printk(KERN_NOTICE, sdkp, "Host-aware SMR disk used as regular disk\n"); @@ -3605,8 +3612,10 @@ static int sd_revalidate_disk(struct gendisk *disk) struct scsi_device *sdp = sdkp->device; struct request_queue *q = sdkp->disk->queue; sector_t old_capacity = sdkp->capacity; + struct queue_limits lim; unsigned char *buffer; unsigned int dev_max; + int err; SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_revalidate_disk\n")); @@ -3627,12 +3636,14 @@ static int sd_revalidate_disk(struct gendisk *disk) sd_spinup_disk(sdkp); + lim = queue_limits_start_update(sdkp->disk->queue); + /* * Without media there is no reason to ask; moreover, some devices * react badly if we do. */ if (sdkp->media_present) { - sd_read_capacity(sdkp, buffer); + sd_read_capacity(sdkp, &lim, buffer); /* * Some USB/UAS devices return generic values for mode pages * until the media has been accessed. Trigger a READ operation @@ -3651,10 +3662,10 @@ static int sd_revalidate_disk(struct gendisk *disk) if (scsi_device_supports_vpd(sdp)) { sd_read_block_provisioning(sdkp); - sd_read_block_limits(sdkp); + sd_read_block_limits(sdkp, &lim); sd_read_block_limits_ext(sdkp); - sd_read_block_characteristics(sdkp); - sd_zbc_read_zones(sdkp, buffer); + sd_read_block_characteristics(sdkp, &lim); + sd_zbc_read_zones(sdkp, &lim, buffer); sd_read_cpr(sdkp); } @@ -3680,31 +3691,36 @@ static int sd_revalidate_disk(struct gendisk *disk) /* Some devices report a maximum block count for READ/WRITE requests. */ dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks); - q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max); + lim.max_dev_sectors = logical_to_sectors(sdp, dev_max); if (sd_validate_min_xfer_size(sdkp)) - blk_queue_io_min(sdkp->disk->queue, - logical_to_bytes(sdp, sdkp->min_xfer_blocks)); + lim.io_min = logical_to_bytes(sdp, sdkp->min_xfer_blocks); else - blk_queue_io_min(sdkp->disk->queue, 0); + lim.io_min = 0; /* * Limit default to SCSI host optimal sector limit if set. There may be * an impact on performance for when the size of a request exceeds this * host limit. */ - q->limits.io_opt = sdp->host->opt_sectors << SECTOR_SHIFT; + lim.io_opt = sdp->host->opt_sectors << SECTOR_SHIFT; if (sd_validate_opt_xfer_size(sdkp, dev_max)) { - q->limits.io_opt = min_not_zero(q->limits.io_opt, + lim.io_opt = min_not_zero(lim.io_opt, logical_to_bytes(sdp, sdkp->opt_xfer_blocks)); } sdkp->first_scan = 0; set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity)); - sd_config_write_same(sdkp); + sd_config_write_same(sdkp, &lim); kfree(buffer); + blk_mq_freeze_queue(sdkp->disk->queue); + err = queue_limits_commit_update(sdkp->disk->queue, &lim); + blk_mq_unfreeze_queue(sdkp->disk->queue); + if (err) + return err; + /* * For a zoned drive, revalidating the zones can be done only once * the gendisk capacity is set. So if this fails, set back the gendisk diff --git a/drivers/scsi/sd.h b/drivers/scsi/sd.h index 49dd600bfa4825..b4170b17bad47a 100644 --- a/drivers/scsi/sd.h +++ b/drivers/scsi/sd.h @@ -239,7 +239,8 @@ static inline int sd_is_zoned(struct scsi_disk *sdkp) #ifdef CONFIG_BLK_DEV_ZONED -int sd_zbc_read_zones(struct scsi_disk *sdkp, u8 buf[SD_BUF_SIZE]); +int sd_zbc_read_zones(struct scsi_disk *sdkp, struct queue_limits *lim, + u8 buf[SD_BUF_SIZE]); int sd_zbc_revalidate_zones(struct scsi_disk *sdkp); blk_status_t sd_zbc_setup_zone_mgmt_cmnd(struct scsi_cmnd *cmd, unsigned char op, bool all); @@ -250,7 +251,8 @@ int sd_zbc_report_zones(struct gendisk *disk, sector_t sector, #else /* CONFIG_BLK_DEV_ZONED */ -static inline int sd_zbc_read_zones(struct scsi_disk *sdkp, u8 buf[SD_BUF_SIZE]) +static inline int sd_zbc_read_zones(struct scsi_disk *sdkp, + struct queue_limits *lim, u8 buf[SD_BUF_SIZE]) { return 0; } diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c index 1c24c844e8d178..f685838d9ed214 100644 --- a/drivers/scsi/sd_zbc.c +++ b/drivers/scsi/sd_zbc.c @@ -582,13 +582,15 @@ int sd_zbc_revalidate_zones(struct scsi_disk *sdkp) /** * sd_zbc_read_zones - Read zone information and update the request queue * @sdkp: SCSI disk pointer. + * @lim: queue limits to read into * @buf: 512 byte buffer used for storing SCSI command output. * * Read zone information and update the request queue zone characteristics and * also the zoned device information in *sdkp. Called by sd_revalidate_disk() * before the gendisk capacity has been set. */ -int sd_zbc_read_zones(struct scsi_disk *sdkp, u8 buf[SD_BUF_SIZE]) +int sd_zbc_read_zones(struct scsi_disk *sdkp, struct queue_limits *lim, + u8 buf[SD_BUF_SIZE]) { struct gendisk *disk = sdkp->disk; struct request_queue *q = disk->queue; @@ -626,14 +628,13 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, u8 buf[SD_BUF_SIZE]) /* The drive satisfies the kernel restrictions: set it up */ blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q); if (sdkp->zones_max_open == U32_MAX) - disk_set_max_open_zones(disk, 0); + lim->max_open_zones = 0; else - disk_set_max_open_zones(disk, sdkp->zones_max_open); - disk_set_max_active_zones(disk, 0); - blk_queue_chunk_sectors(q, - logical_to_sectors(sdkp->device, zone_blocks)); + lim->max_open_zones = sdkp->zones_max_open; + lim->max_active_zones = 0; + lim->chunk_sectors = logical_to_sectors(sdkp->device, zone_blocks); /* Enable block layer zone append emulation */ - blk_queue_max_zone_append_sectors(q, 0); + lim->max_zone_append_sectors = 0; return 0; From 969f17e10f5b732c05186ee0126c8a08166d0cda Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:48:07 +0200 Subject: [PATCH 034/209] sr: convert to the atomic queue limits API Assign all queue limits through a local queue_limits variable and queue_limits_commit_update so that we can't race updating them from multiple places, and free the queue when updating them so that in-progress I/O submissions don't see half-updated limits. Also use the chance to clean up variable names to standard ones. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-13-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/sr.c | 42 +++++++++++++++++++++++++----------------- 1 file changed, 25 insertions(+), 17 deletions(-) diff --git a/drivers/scsi/sr.c b/drivers/scsi/sr.c index 7ab000942b97fc..3f491019103e0c 100644 --- a/drivers/scsi/sr.c +++ b/drivers/scsi/sr.c @@ -111,7 +111,7 @@ static struct lock_class_key sr_bio_compl_lkclass; static int sr_open(struct cdrom_device_info *, int); static void sr_release(struct cdrom_device_info *); -static void get_sectorsize(struct scsi_cd *); +static int get_sectorsize(struct scsi_cd *); static int get_capabilities(struct scsi_cd *); static unsigned int sr_check_events(struct cdrom_device_info *cdi, @@ -473,15 +473,15 @@ static blk_status_t sr_init_command(struct scsi_cmnd *SCpnt) return BLK_STS_IOERR; } -static void sr_revalidate_disk(struct scsi_cd *cd) +static int sr_revalidate_disk(struct scsi_cd *cd) { struct scsi_sense_hdr sshdr; /* if the unit is not ready, nothing more to do */ if (scsi_test_unit_ready(cd->device, SR_TIMEOUT, MAX_RETRIES, &sshdr)) - return; + return 0; sr_cd_check(&cd->cdi); - get_sectorsize(cd); + return get_sectorsize(cd); } static int sr_block_open(struct gendisk *disk, blk_mode_t mode) @@ -494,13 +494,16 @@ static int sr_block_open(struct gendisk *disk, blk_mode_t mode) return -ENXIO; scsi_autopm_get_device(sdev); - if (disk_check_media_change(disk)) - sr_revalidate_disk(cd); + if (disk_check_media_change(disk)) { + ret = sr_revalidate_disk(cd); + if (ret) + goto out; + } mutex_lock(&cd->lock); ret = cdrom_open(&cd->cdi, mode); mutex_unlock(&cd->lock); - +out: scsi_autopm_put_device(sdev); if (ret) scsi_device_put(cd->device); @@ -685,7 +688,9 @@ static int sr_probe(struct device *dev) blk_pm_runtime_init(sdev->request_queue, dev); dev_set_drvdata(dev, cd); - sr_revalidate_disk(cd); + error = sr_revalidate_disk(cd); + if (error) + goto unregister_cdrom; error = device_add_disk(&sdev->sdev_gendev, disk, NULL); if (error) @@ -714,13 +719,14 @@ static int sr_probe(struct device *dev) } -static void get_sectorsize(struct scsi_cd *cd) +static int get_sectorsize(struct scsi_cd *cd) { + struct request_queue *q = cd->device->request_queue; static const u8 cmd[10] = { READ_CAPACITY }; unsigned char buffer[8] = { }; - int the_result; + struct queue_limits lim; + int err; int sector_size; - struct request_queue *queue; struct scsi_failure failure_defs[] = { { .result = SCMD_FAILURE_RESULT_ANY, @@ -736,10 +742,10 @@ static void get_sectorsize(struct scsi_cd *cd) }; /* Do the command and wait.. */ - the_result = scsi_execute_cmd(cd->device, cmd, REQ_OP_DRV_IN, buffer, + err = scsi_execute_cmd(cd->device, cmd, REQ_OP_DRV_IN, buffer, sizeof(buffer), SR_TIMEOUT, MAX_RETRIES, &exec_args); - if (the_result) { + if (err) { cd->capacity = 0x1fffff; sector_size = 2048; /* A guess, just in case */ } else { @@ -789,10 +795,12 @@ static void get_sectorsize(struct scsi_cd *cd) set_capacity(cd->disk, cd->capacity); } - queue = cd->device->request_queue; - blk_queue_logical_block_size(queue, sector_size); - - return; + lim = queue_limits_start_update(q); + lim.logical_block_size = sector_size; + blk_mq_freeze_queue(q); + err = queue_limits_commit_update(q, &lim); + blk_mq_unfreeze_queue(q); + return err; } static int get_capabilities(struct scsi_cd *cd) From 1652b0bafeaa8281ca9a805d81e13d7647bd2f44 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:48:08 +0200 Subject: [PATCH 035/209] block: remove unused queue limits API Remove all APIs that are unused now that sd and sr have been converted to the atomic queue limits API. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: John Garry Reviewed-by: Nitesh Shetty Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-14-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 190 ----------------------------------------- include/linux/blkdev.h | 24 ------ 2 files changed, 214 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index a49abdb3554834..0b038729608f4b 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -293,24 +293,6 @@ int queue_limits_set(struct request_queue *q, struct queue_limits *lim) } EXPORT_SYMBOL_GPL(queue_limits_set); -/** - * blk_queue_chunk_sectors - set size of the chunk for this queue - * @q: the request queue for the device - * @chunk_sectors: chunk sectors in the usual 512b unit - * - * Description: - * If a driver doesn't want IOs to cross a given chunk size, it can set - * this limit and prevent merging across chunks. Note that the block layer - * must accept a page worth of data at any offset. So if the crossing of - * chunks is a hard limitation in the driver, it must still be prepared - * to split single page bios. - **/ -void blk_queue_chunk_sectors(struct request_queue *q, unsigned int chunk_sectors) -{ - q->limits.chunk_sectors = chunk_sectors; -} -EXPORT_SYMBOL(blk_queue_chunk_sectors); - /** * blk_queue_max_discard_sectors - set max sectors for a single discard * @q: the request queue for the device @@ -352,139 +334,6 @@ void blk_queue_max_write_zeroes_sectors(struct request_queue *q, } EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors); -/** - * blk_queue_max_zone_append_sectors - set max sectors for a single zone append - * @q: the request queue for the device - * @max_zone_append_sectors: maximum number of sectors to write per command - * - * Sets the maximum number of sectors allowed for zone append commands. If - * Specifying 0 for @max_zone_append_sectors indicates that the queue does - * not natively support zone append operations and that the block layer must - * emulate these operations using regular writes. - **/ -void blk_queue_max_zone_append_sectors(struct request_queue *q, - unsigned int max_zone_append_sectors) -{ - unsigned int max_sectors = 0; - - if (WARN_ON(!blk_queue_is_zoned(q))) - return; - - if (max_zone_append_sectors) { - max_sectors = min(q->limits.max_hw_sectors, - max_zone_append_sectors); - max_sectors = min(q->limits.chunk_sectors, max_sectors); - - /* - * Signal eventual driver bugs resulting in the max_zone_append - * sectors limit being 0 due to the chunk_sectors limit (zone - * size) not set or the max_hw_sectors limit not set. - */ - WARN_ON_ONCE(!max_sectors); - } - - q->limits.max_zone_append_sectors = max_sectors; -} -EXPORT_SYMBOL_GPL(blk_queue_max_zone_append_sectors); - -/** - * blk_queue_logical_block_size - set logical block size for the queue - * @q: the request queue for the device - * @size: the logical block size, in bytes - * - * Description: - * This should be set to the lowest possible block size that the - * storage device can address. The default of 512 covers most - * hardware. - **/ -void blk_queue_logical_block_size(struct request_queue *q, unsigned int size) -{ - struct queue_limits *limits = &q->limits; - - limits->logical_block_size = size; - - if (limits->discard_granularity < limits->logical_block_size) - limits->discard_granularity = limits->logical_block_size; - - if (limits->physical_block_size < size) - limits->physical_block_size = size; - - if (limits->io_min < limits->physical_block_size) - limits->io_min = limits->physical_block_size; - - limits->max_hw_sectors = - round_down(limits->max_hw_sectors, size >> SECTOR_SHIFT); - limits->max_sectors = - round_down(limits->max_sectors, size >> SECTOR_SHIFT); -} -EXPORT_SYMBOL(blk_queue_logical_block_size); - -/** - * blk_queue_physical_block_size - set physical block size for the queue - * @q: the request queue for the device - * @size: the physical block size, in bytes - * - * Description: - * This should be set to the lowest possible sector size that the - * hardware can operate on without reverting to read-modify-write - * operations. - */ -void blk_queue_physical_block_size(struct request_queue *q, unsigned int size) -{ - q->limits.physical_block_size = size; - - if (q->limits.physical_block_size < q->limits.logical_block_size) - q->limits.physical_block_size = q->limits.logical_block_size; - - if (q->limits.discard_granularity < q->limits.physical_block_size) - q->limits.discard_granularity = q->limits.physical_block_size; - - if (q->limits.io_min < q->limits.physical_block_size) - q->limits.io_min = q->limits.physical_block_size; -} -EXPORT_SYMBOL(blk_queue_physical_block_size); - -/** - * blk_queue_zone_write_granularity - set zone write granularity for the queue - * @q: the request queue for the zoned device - * @size: the zone write granularity size, in bytes - * - * Description: - * This should be set to the lowest possible size allowing to write in - * sequential zones of a zoned block device. - */ -void blk_queue_zone_write_granularity(struct request_queue *q, - unsigned int size) -{ - if (WARN_ON_ONCE(!blk_queue_is_zoned(q))) - return; - - q->limits.zone_write_granularity = size; - - if (q->limits.zone_write_granularity < q->limits.logical_block_size) - q->limits.zone_write_granularity = q->limits.logical_block_size; -} -EXPORT_SYMBOL_GPL(blk_queue_zone_write_granularity); - -/** - * blk_queue_alignment_offset - set physical block alignment offset - * @q: the request queue for the device - * @offset: alignment offset in bytes - * - * Description: - * Some devices are naturally misaligned to compensate for things like - * the legacy DOS partition table 63-sector offset. Low-level drivers - * should call this function for devices whose first sector is not - * naturally aligned. - */ -void blk_queue_alignment_offset(struct request_queue *q, unsigned int offset) -{ - q->limits.alignment_offset = - offset & (q->limits.physical_block_size - 1); - q->limits.misaligned = 0; -} -EXPORT_SYMBOL(blk_queue_alignment_offset); - void disk_update_readahead(struct gendisk *disk) { blk_apply_bdi_limits(disk->bdi, &disk->queue->limits); @@ -514,26 +363,6 @@ void blk_limits_io_min(struct queue_limits *limits, unsigned int min) } EXPORT_SYMBOL(blk_limits_io_min); -/** - * blk_queue_io_min - set minimum request size for the queue - * @q: the request queue for the device - * @min: smallest I/O size in bytes - * - * Description: - * Storage devices may report a granularity or preferred minimum I/O - * size which is the smallest request the device can perform without - * incurring a performance penalty. For disk drives this is often the - * physical block size. For RAID arrays it is often the stripe chunk - * size. A properly aligned multiple of minimum_io_size is the - * preferred request size for workloads where a high number of I/O - * operations is desired. - */ -void blk_queue_io_min(struct request_queue *q, unsigned int min) -{ - blk_limits_io_min(&q->limits, min); -} -EXPORT_SYMBOL(blk_queue_io_min); - /** * blk_limits_io_opt - set optimal request size for a device * @limits: the queue limits @@ -841,25 +670,6 @@ void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua) } EXPORT_SYMBOL_GPL(blk_queue_write_cache); -/** - * disk_set_zoned - inidicate a zoned device - * @disk: gendisk to configure - */ -void disk_set_zoned(struct gendisk *disk) -{ - struct request_queue *q = disk->queue; - - WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED)); - - /* - * Set the zone write granularity to the device logical block - * size by default. The driver can change this value if needed. - */ - q->limits.zoned = true; - blk_queue_zone_write_granularity(q, queue_logical_block_size(q)); -} -EXPORT_SYMBOL_GPL(disk_set_zoned); - int bdev_alignment_offset(struct block_device *bdev) { struct request_queue *q = bdev_get_queue(bdev); diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 24c36929920b76..ad995e7a769811 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -332,8 +332,6 @@ struct queue_limits { typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx, void *data); -void disk_set_zoned(struct gendisk *disk); - #define BLK_ALL_ZONES ((unsigned int)-1) int blkdev_report_zones(struct block_device *bdev, sector_t sector, unsigned int nr_zones, report_zones_cb cb, void *data); @@ -638,18 +636,6 @@ static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) return sector >> ilog2(disk->queue->limits.chunk_sectors); } -static inline void disk_set_max_open_zones(struct gendisk *disk, - unsigned int max_open_zones) -{ - disk->queue->limits.max_open_zones = max_open_zones; -} - -static inline void disk_set_max_active_zones(struct gendisk *disk, - unsigned int max_active_zones) -{ - disk->queue->limits.max_active_zones = max_active_zones; -} - static inline unsigned int bdev_max_open_zones(struct block_device *bdev) { return bdev->bd_disk->queue->limits.max_open_zones; @@ -929,24 +915,14 @@ static inline void queue_limits_cancel_update(struct request_queue *q) /* * Access functions for manipulating queue properties */ -extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int); void blk_queue_max_secure_erase_sectors(struct request_queue *q, unsigned int max_sectors); extern void blk_queue_max_discard_sectors(struct request_queue *q, unsigned int max_discard_sectors); extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q, unsigned int max_write_same_sectors); -extern void blk_queue_logical_block_size(struct request_queue *, unsigned int); -extern void blk_queue_max_zone_append_sectors(struct request_queue *q, - unsigned int max_zone_append_sectors); -extern void blk_queue_physical_block_size(struct request_queue *, unsigned int); -void blk_queue_zone_write_granularity(struct request_queue *q, - unsigned int size); -extern void blk_queue_alignment_offset(struct request_queue *q, - unsigned int alignment); void disk_update_readahead(struct gendisk *disk); extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); -extern void blk_queue_io_min(struct request_queue *q, unsigned int min); extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); extern void blk_set_stacking_limits(struct queue_limits *lim); From 73e3715ed14844067c5c598e72777641004a7f60 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 31 May 2024 09:48:09 +0200 Subject: [PATCH 036/209] block: add special APIs for run-time disabling of discard and friends A few drivers optimistically try to support discard, write zeroes and secure erase and disable the features from the I/O completion handler if the hardware can't support them. This disable can't be done using the atomic queue limits API because the I/O completion handlers can't take sleeping locks or freeze the queue. Keep the existing clearing of the relevant field to zero, but replace the old blk_queue_max_* APIs with new disable APIs that force the value to 0. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: John Garry Reviewed-by: Nitesh Shetty Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240531074837.1648501-15-hch@lst.de Signed-off-by: Jens Axboe --- arch/um/drivers/ubd_kern.c | 4 ++-- block/blk-settings.c | 41 ------------------------------------ drivers/block/xen-blkfront.c | 4 ++-- drivers/scsi/sd.c | 4 ++-- include/linux/blkdev.h | 28 ++++++++++++++++++------ 5 files changed, 28 insertions(+), 53 deletions(-) diff --git a/arch/um/drivers/ubd_kern.c b/arch/um/drivers/ubd_kern.c index 093c87879d08ba..cdcb75a68989dd 100644 --- a/arch/um/drivers/ubd_kern.c +++ b/arch/um/drivers/ubd_kern.c @@ -451,9 +451,9 @@ static void ubd_end_request(struct io_thread_req *io_req) { if (io_req->error == BLK_STS_NOTSUPP) { if (req_op(io_req->req) == REQ_OP_DISCARD) - blk_queue_max_discard_sectors(io_req->req->q, 0); + blk_queue_disable_discard(io_req->req->q); else if (req_op(io_req->req) == REQ_OP_WRITE_ZEROES) - blk_queue_max_write_zeroes_sectors(io_req->req->q, 0); + blk_queue_disable_write_zeroes(io_req->req->q); } blk_mq_end_request(io_req->req, io_req->error); kfree(io_req); diff --git a/block/blk-settings.c b/block/blk-settings.c index 0b038729608f4b..996f247fc98e80 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -293,47 +293,6 @@ int queue_limits_set(struct request_queue *q, struct queue_limits *lim) } EXPORT_SYMBOL_GPL(queue_limits_set); -/** - * blk_queue_max_discard_sectors - set max sectors for a single discard - * @q: the request queue for the device - * @max_discard_sectors: maximum number of sectors to discard - **/ -void blk_queue_max_discard_sectors(struct request_queue *q, - unsigned int max_discard_sectors) -{ - struct queue_limits *lim = &q->limits; - - lim->max_hw_discard_sectors = max_discard_sectors; - lim->max_discard_sectors = - min(max_discard_sectors, lim->max_user_discard_sectors); -} -EXPORT_SYMBOL(blk_queue_max_discard_sectors); - -/** - * blk_queue_max_secure_erase_sectors - set max sectors for a secure erase - * @q: the request queue for the device - * @max_sectors: maximum number of sectors to secure_erase - **/ -void blk_queue_max_secure_erase_sectors(struct request_queue *q, - unsigned int max_sectors) -{ - q->limits.max_secure_erase_sectors = max_sectors; -} -EXPORT_SYMBOL(blk_queue_max_secure_erase_sectors); - -/** - * blk_queue_max_write_zeroes_sectors - set max sectors for a single - * write zeroes - * @q: the request queue for the device - * @max_write_zeroes_sectors: maximum number of sectors to write per command - **/ -void blk_queue_max_write_zeroes_sectors(struct request_queue *q, - unsigned int max_write_zeroes_sectors) -{ - q->limits.max_write_zeroes_sectors = max_write_zeroes_sectors; -} -EXPORT_SYMBOL(blk_queue_max_write_zeroes_sectors); - void disk_update_readahead(struct gendisk *disk) { blk_apply_bdi_limits(disk->bdi, &disk->queue->limits); diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c index fd7c0ff2139cee..9b4ec3e4908cce 100644 --- a/drivers/block/xen-blkfront.c +++ b/drivers/block/xen-blkfront.c @@ -1605,8 +1605,8 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id) blkif_req(req)->error = BLK_STS_NOTSUPP; info->feature_discard = 0; info->feature_secdiscard = 0; - blk_queue_max_discard_sectors(rq, 0); - blk_queue_max_secure_erase_sectors(rq, 0); + blk_queue_disable_discard(rq); + blk_queue_disable_secure_erase(rq); } break; case BLKIF_OP_FLUSH_DISKCACHE: diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 049071b5681989..d957e29b17a98a 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -837,7 +837,7 @@ static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd, static void sd_disable_discard(struct scsi_disk *sdkp) { sdkp->provisioning_mode = SD_LBP_DISABLE; - blk_queue_max_discard_sectors(sdkp->disk->queue, 0); + blk_queue_disable_discard(sdkp->disk->queue); } static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim, @@ -1019,7 +1019,7 @@ static void sd_disable_write_same(struct scsi_disk *sdkp) { sdkp->device->no_write_same = 1; sdkp->max_ws_blocks = 0; - blk_queue_max_write_zeroes_sectors(sdkp->disk->queue, 0); + blk_queue_disable_write_zeroes(sdkp->disk->queue); } static void sd_config_write_same(struct scsi_disk *sdkp, diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index ad995e7a769811..ac8e0cb2353a0e 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -912,15 +912,31 @@ static inline void queue_limits_cancel_update(struct request_queue *q) mutex_unlock(&q->limits_lock); } +/* + * These helpers are for drivers that have sloppy feature negotiation and might + * have to disable DISCARD, WRITE_ZEROES or SECURE_DISCARD from the I/O + * completion handler when the device returned an indicator that the respective + * feature is not actually supported. They are racy and the driver needs to + * cope with that. Try to avoid this scheme if you can. + */ +static inline void blk_queue_disable_discard(struct request_queue *q) +{ + q->limits.max_discard_sectors = 0; +} + +static inline void blk_queue_disable_secure_erase(struct request_queue *q) +{ + q->limits.max_secure_erase_sectors = 0; +} + +static inline void blk_queue_disable_write_zeroes(struct request_queue *q) +{ + q->limits.max_write_zeroes_sectors = 0; +} + /* * Access functions for manipulating queue properties */ -void blk_queue_max_secure_erase_sectors(struct request_queue *q, - unsigned int max_sectors); -extern void blk_queue_max_discard_sectors(struct request_queue *q, - unsigned int max_discard_sectors); -extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q, - unsigned int max_write_same_sectors); void disk_update_readahead(struct gendisk *disk); extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); From 899ee2c3829c5ac14bfc7d3c4a5846c0b709b78f Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:11 +0200 Subject: [PATCH 037/209] block: initialize integrity buffer to zero before writing it to media Metadata added by bio_integrity_prep is using plain kmalloc, which leads to random kernel memory being written media. For PI metadata this is limited to the app tag that isn't used by kernel generated metadata, but for non-PI metadata the entire buffer leaks kernel memory. Fix this by adding the __GFP_ZERO flag to allocations for writes. Fixes: 7ba1ba12eeef ("block: Block layer data integrity support") Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Kanchan Joshi Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240613084839.1044015-2-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index 2e3e8e04961eae..af7f71d16114de 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -432,6 +432,7 @@ bool bio_integrity_prep(struct bio *bio) unsigned long start, end; unsigned int len, nr_pages; unsigned int bytes, offset, i; + gfp_t gfp = GFP_NOIO; if (!bi) return true; @@ -454,11 +455,19 @@ bool bio_integrity_prep(struct bio *bio) if (!bi->profile->generate_fn || !(bi->flags & BLK_INTEGRITY_GENERATE)) return true; + + /* + * Zero the memory allocated to not leak uninitialized kernel + * memory to disk. For PI this only affects the app tag, but + * for non-integrity metadata it affects the entire metadata + * buffer. + */ + gfp |= __GFP_ZERO; } /* Allocate kernel buffer for protection data */ len = bio_integrity_bytes(bi, bio_sectors(bio)); - buf = kmalloc(len, GFP_NOIO); + buf = kmalloc(len, gfp); if (unlikely(buf == NULL)) { printk(KERN_ERR "could not allocate integrity buffer\n"); goto err_end_io; From d11854ed05635e4a73fa61a988ffdd0978c9e202 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:12 +0200 Subject: [PATCH 038/209] md/raid0: don't free conf on raid0_run failure The core md code calls the ->free method which already frees conf. Fixes: 0c031fd37f69 ("md: Move alloc/free acct bioset in to personality") Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Yu Kuai Link: https://lore.kernel.org/r/20240613084839.1044015-3-hch@lst.de Signed-off-by: Jens Axboe --- drivers/md/raid0.c | 21 +++++---------------- 1 file changed, 5 insertions(+), 16 deletions(-) diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c index c5d4aeb68404c9..81c01347cd24e6 100644 --- a/drivers/md/raid0.c +++ b/drivers/md/raid0.c @@ -365,18 +365,13 @@ static sector_t raid0_size(struct mddev *mddev, sector_t sectors, int raid_disks return array_sectors; } -static void free_conf(struct mddev *mddev, struct r0conf *conf) -{ - kfree(conf->strip_zone); - kfree(conf->devlist); - kfree(conf); -} - static void raid0_free(struct mddev *mddev, void *priv) { struct r0conf *conf = priv; - free_conf(mddev, conf); + kfree(conf->strip_zone); + kfree(conf->devlist); + kfree(conf); } static int raid0_set_limits(struct mddev *mddev) @@ -415,7 +410,7 @@ static int raid0_run(struct mddev *mddev) if (!mddev_is_dm(mddev)) { ret = raid0_set_limits(mddev); if (ret) - goto out_free_conf; + return ret; } /* calculate array device size */ @@ -427,13 +422,7 @@ static int raid0_run(struct mddev *mddev) dump_zones(mddev); - ret = md_integrity_register(mddev); - if (ret) - goto out_free_conf; - return 0; -out_free_conf: - free_conf(mddev, conf); - return ret; + return md_integrity_register(mddev); } /* From 799af947ed132956d6de6d77a5bc053817ccb06b Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:13 +0200 Subject: [PATCH 039/209] md/raid1: don't free conf on raid0_run failure The core md code calls the ->free method which already frees conf. Fixes: 07f1a6850c5d ("md/raid1: fail run raid1 array when active disk less than one") Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240613084839.1044015-4-hch@lst.de Signed-off-by: Jens Axboe --- drivers/md/raid1.c | 14 +++----------- 1 file changed, 3 insertions(+), 11 deletions(-) diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 7b8a71ca66dde0..1f321826ef02ba 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -3204,7 +3204,6 @@ static int raid1_set_limits(struct mddev *mddev) return queue_limits_set(mddev->gendisk->queue, &lim); } -static void raid1_free(struct mddev *mddev, void *priv); static int raid1_run(struct mddev *mddev) { struct r1conf *conf; @@ -3238,7 +3237,7 @@ static int raid1_run(struct mddev *mddev) if (!mddev_is_dm(mddev)) { ret = raid1_set_limits(mddev); if (ret) - goto abort; + return ret; } mddev->degraded = 0; @@ -3252,8 +3251,7 @@ static int raid1_run(struct mddev *mddev) */ if (conf->raid_disks - mddev->degraded < 1) { md_unregister_thread(mddev, &conf->thread); - ret = -EINVAL; - goto abort; + return -EINVAL; } if (conf->raid_disks - mddev->degraded == 1) @@ -3277,14 +3275,8 @@ static int raid1_run(struct mddev *mddev) md_set_array_sectors(mddev, raid1_size(mddev, 0, 0)); ret = md_integrity_register(mddev); - if (ret) { + if (ret) md_unregister_thread(mddev, &mddev->thread); - goto abort; - } - return 0; - -abort: - raid1_free(mddev, conf); return ret; } From 63e649594ab19cc3122a2d0fc2c94b19932f0b19 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:14 +0200 Subject: [PATCH 040/209] dm-integrity: use the nop integrity profile Use the block layer built-in nop profile instead of duplicating it. Tested by: $ dd if=/dev/urandom of=key.bin bs=512 count=1 $ cryptsetup luksFormat -q --type luks2 --integrity hmac-sha256 \ --integrity-no-wipe /dev/nvme0n1 key.bin $ cryptsetup luksOpen /dev/nvme0n1 luks-integrity --key-file key.bin and then doing mkfs.xfs and simple I/O on the mount file system. Signed-off-by: Christoph Hellwig Reviewed-by: Milan Broz Reviewed-by: Chaitanya Kulkarni Reviewed-by: Hannes Reinecke Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240613084839.1044015-5-hch@lst.de Signed-off-by: Jens Axboe --- drivers/md/dm-crypt.c | 4 ++-- drivers/md/dm-integrity.c | 20 -------------------- 2 files changed, 2 insertions(+), 22 deletions(-) diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c index 1b7a97cc377943..1dfc462f29cd6f 100644 --- a/drivers/md/dm-crypt.c +++ b/drivers/md/dm-crypt.c @@ -1176,8 +1176,8 @@ static int crypt_integrity_ctr(struct crypt_config *cc, struct dm_target *ti) struct blk_integrity *bi = blk_get_integrity(cc->dev->bdev->bd_disk); struct mapped_device *md = dm_table_get_md(ti->table); - /* From now we require underlying device with our integrity profile */ - if (!bi || strcasecmp(bi->profile->name, "DM-DIF-EXT-TAG")) { + /* We require an underlying device with non-PI metadata */ + if (!bi || strcmp(bi->profile->name, "nop")) { ti->error = "Integrity profile not supported."; return -EINVAL; } diff --git a/drivers/md/dm-integrity.c b/drivers/md/dm-integrity.c index 417fddebe367a2..c1cc27541673c7 100644 --- a/drivers/md/dm-integrity.c +++ b/drivers/md/dm-integrity.c @@ -350,25 +350,6 @@ static struct kmem_cache *journal_io_cache; #define DEBUG_bytes(bytes, len, msg, ...) do { } while (0) #endif -static void dm_integrity_prepare(struct request *rq) -{ -} - -static void dm_integrity_complete(struct request *rq, unsigned int nr_bytes) -{ -} - -/* - * DM Integrity profile, protection is performed layer above (dm-crypt) - */ -static const struct blk_integrity_profile dm_integrity_profile = { - .name = "DM-DIF-EXT-TAG", - .generate_fn = NULL, - .verify_fn = NULL, - .prepare_fn = dm_integrity_prepare, - .complete_fn = dm_integrity_complete, -}; - static void dm_integrity_map_continue(struct dm_integrity_io *dio, bool from_map); static void integrity_bio_wait(struct work_struct *w); static void dm_integrity_dtr(struct dm_target *ti); @@ -3656,7 +3637,6 @@ static void dm_integrity_set(struct dm_target *ti, struct dm_integrity_c *ic) struct blk_integrity bi; memset(&bi, 0, sizeof(bi)); - bi.profile = &dm_integrity_profile; bi.tuple_size = ic->tag_size; bi.tag_size = bi.tuple_size; bi.interval_exp = ic->sb->log2_sectors_per_block + SECTOR_SHIFT; From e9f5f44ad3725335d9c559c3c22cd3726152a7b1 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:15 +0200 Subject: [PATCH 041/209] block: remove the blk_integrity_profile structure Block layer integrity configuration is a bit complex right now, as it indirects through operation vectors for a simple two-dimensional configuration: a) the checksum type of none, ip checksum, crc, crc64 b) the presence or absence of a reference tag Remove the integrity profile, and instead add a separate csum_type flag which replaces the existing ip-checksum field and a new flag that indicates the presence of the reference tag. This removes up to two layers of indirect calls, remove the need to offload the no-op verification of non-PI metadata to a workqueue and generally simplifies the code. The downside is that block/t10-pi.c now has to be built into the kernel when CONFIG_BLK_DEV_INTEGRITY is supported. Given that both nvme and SCSI require t10-pi.ko, it is loaded for all usual configurations that enabled CONFIG_BLK_DEV_INTEGRITY already, though. Signed-off-by: Christoph Hellwig Reviewed-by: Kanchan Joshi Reviewed-by: Hannes Reinecke Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240613084839.1044015-6-hch@lst.de Signed-off-by: Jens Axboe --- block/Kconfig | 8 +- block/Makefile | 3 +- block/bio-integrity.c | 32 ++-- block/blk-integrity.c | 66 ++++---- block/blk-mq.c | 13 +- block/blk.h | 8 + block/t10-pi.c | 241 ++++++++++------------------ drivers/md/dm-crypt.c | 2 +- drivers/nvme/host/Kconfig | 1 - drivers/nvme/host/core.c | 17 +- drivers/nvme/target/Kconfig | 1 - drivers/nvme/target/io-cmd-bdev.c | 16 +- drivers/scsi/Kconfig | 1 - drivers/scsi/sd_dif.c | 19 +-- drivers/target/target_core_iblock.c | 49 +++--- include/linux/blk-integrity.h | 35 ++-- include/linux/blkdev.h | 9 +- include/linux/t10-pi.h | 8 - 18 files changed, 215 insertions(+), 314 deletions(-) diff --git a/block/Kconfig b/block/Kconfig index dc12af58dbaeca..5b623b876d3b4a 100644 --- a/block/Kconfig +++ b/block/Kconfig @@ -62,6 +62,8 @@ config BLK_DEV_BSGLIB config BLK_DEV_INTEGRITY bool "Block layer data integrity support" + select CRC_T10DIF + select CRC64_ROCKSOFT help Some storage devices allow extra information to be stored/retrieved to help protect the data. The block layer @@ -72,12 +74,6 @@ config BLK_DEV_INTEGRITY T10/SCSI Data Integrity Field or the T13/ATA External Path Protection. If in doubt, say N. -config BLK_DEV_INTEGRITY_T10 - tristate - depends on BLK_DEV_INTEGRITY - select CRC_T10DIF - select CRC64_ROCKSOFT - config BLK_DEV_WRITE_MOUNTED bool "Allow writing to mounted block devices" default y diff --git a/block/Makefile b/block/Makefile index 168150b9c51025..ddfd21c1a9ffc9 100644 --- a/block/Makefile +++ b/block/Makefile @@ -26,8 +26,7 @@ obj-$(CONFIG_MQ_IOSCHED_KYBER) += kyber-iosched.o bfq-y := bfq-iosched.o bfq-wf2q.o bfq-cgroup.o obj-$(CONFIG_IOSCHED_BFQ) += bfq.o -obj-$(CONFIG_BLK_DEV_INTEGRITY) += bio-integrity.o blk-integrity.o -obj-$(CONFIG_BLK_DEV_INTEGRITY_T10) += t10-pi.o +obj-$(CONFIG_BLK_DEV_INTEGRITY) += bio-integrity.o blk-integrity.o t10-pi.o obj-$(CONFIG_BLK_MQ_PCI) += blk-mq-pci.o obj-$(CONFIG_BLK_MQ_VIRTIO) += blk-mq-virtio.o obj-$(CONFIG_BLK_DEV_ZONED) += blk-zoned.o diff --git a/block/bio-integrity.c b/block/bio-integrity.c index af7f71d16114de..31dbc2853f92e3 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -378,10 +378,9 @@ EXPORT_SYMBOL_GPL(bio_integrity_map_user); * bio_integrity_process - Process integrity metadata for a bio * @bio: bio to generate/verify integrity metadata for * @proc_iter: iterator to process - * @proc_fn: Pointer to the relevant processing function */ static blk_status_t bio_integrity_process(struct bio *bio, - struct bvec_iter *proc_iter, integrity_processing_fn *proc_fn) + struct bvec_iter *proc_iter) { struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk); struct blk_integrity_iter iter; @@ -392,17 +391,18 @@ static blk_status_t bio_integrity_process(struct bio *bio, iter.disk_name = bio->bi_bdev->bd_disk->disk_name; iter.interval = 1 << bi->interval_exp; - iter.tuple_size = bi->tuple_size; iter.seed = proc_iter->bi_sector; iter.prot_buf = bvec_virt(bip->bip_vec); - iter.pi_offset = bi->pi_offset; __bio_for_each_segment(bv, bio, bviter, *proc_iter) { void *kaddr = bvec_kmap_local(&bv); iter.data_buf = kaddr; iter.data_size = bv.bv_len; - ret = proc_fn(&iter); + if (bio_data_dir(bio) == WRITE) + blk_integrity_generate(&iter, bi); + else + ret = blk_integrity_verify(&iter, bi); kunmap_local(kaddr); if (ret) @@ -448,12 +448,10 @@ bool bio_integrity_prep(struct bio *bio) return true; if (bio_data_dir(bio) == READ) { - if (!bi->profile->verify_fn || - !(bi->flags & BLK_INTEGRITY_VERIFY)) + if (!(bi->flags & BLK_INTEGRITY_VERIFY)) return true; } else { - if (!bi->profile->generate_fn || - !(bi->flags & BLK_INTEGRITY_GENERATE)) + if (!(bi->flags & BLK_INTEGRITY_GENERATE)) return true; /* @@ -488,7 +486,7 @@ bool bio_integrity_prep(struct bio *bio) bip->bip_flags |= BIP_BLOCK_INTEGRITY; bip_set_seed(bip, bio->bi_iter.bi_sector); - if (bi->flags & BLK_INTEGRITY_IP_CHECKSUM) + if (bi->csum_type == BLK_INTEGRITY_CSUM_IP) bip->bip_flags |= BIP_IP_CHECKSUM; /* Map it */ @@ -511,12 +509,10 @@ bool bio_integrity_prep(struct bio *bio) } /* Auto-generate integrity metadata if this is a write */ - if (bio_data_dir(bio) == WRITE) { - bio_integrity_process(bio, &bio->bi_iter, - bi->profile->generate_fn); - } else { + if (bio_data_dir(bio) == WRITE) + bio_integrity_process(bio, &bio->bi_iter); + else bip->bio_iter = bio->bi_iter; - } return true; err_end_io: @@ -539,15 +535,13 @@ static void bio_integrity_verify_fn(struct work_struct *work) struct bio_integrity_payload *bip = container_of(work, struct bio_integrity_payload, bip_work); struct bio *bio = bip->bip_bio; - struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk); /* * At the moment verify is called bio's iterator was advanced * during split and completion, we need to rewind iterator to * it's original position. */ - bio->bi_status = bio_integrity_process(bio, &bip->bio_iter, - bi->profile->verify_fn); + bio->bi_status = bio_integrity_process(bio, &bip->bio_iter); bio_integrity_free(bio); bio_endio(bio); } @@ -569,7 +563,7 @@ bool __bio_integrity_endio(struct bio *bio) struct bio_integrity_payload *bip = bio_integrity(bio); if (bio_op(bio) == REQ_OP_READ && !bio->bi_status && - (bip->bip_flags & BIP_BLOCK_INTEGRITY) && bi->profile->verify_fn) { + (bip->bip_flags & BIP_BLOCK_INTEGRITY) && bi->csum_type) { INIT_WORK(&bip->bip_work, bio_integrity_verify_fn); queue_work(kintegrityd_wq, &bip->bip_work); return false; diff --git a/block/blk-integrity.c b/block/blk-integrity.c index ccbeb6dfa87a4d..17d37badfbb8bc 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -123,10 +123,10 @@ int blk_integrity_compare(struct gendisk *gd1, struct gendisk *gd2) struct blk_integrity *b1 = &gd1->queue->integrity; struct blk_integrity *b2 = &gd2->queue->integrity; - if (!b1->profile && !b2->profile) + if (!b1->tuple_size && !b2->tuple_size) return 0; - if (!b1->profile || !b2->profile) + if (!b1->tuple_size || !b2->tuple_size) return -1; if (b1->interval_exp != b2->interval_exp) { @@ -150,10 +150,13 @@ int blk_integrity_compare(struct gendisk *gd1, struct gendisk *gd2) return -1; } - if (b1->profile != b2->profile) { + if (b1->csum_type != b2->csum_type || + (b1->flags & BLK_INTEGRITY_REF_TAG) != + (b2->flags & BLK_INTEGRITY_REF_TAG)) { pr_err("%s: %s/%s type %s != %s\n", __func__, gd1->disk_name, gd2->disk_name, - b1->profile->name, b2->profile->name); + blk_integrity_profile_name(b1), + blk_integrity_profile_name(b2)); return -1; } @@ -217,14 +220,37 @@ static inline struct blk_integrity *dev_to_bi(struct device *dev) return &dev_to_disk(dev)->queue->integrity; } +const char *blk_integrity_profile_name(struct blk_integrity *bi) +{ + switch (bi->csum_type) { + case BLK_INTEGRITY_CSUM_IP: + if (bi->flags & BLK_INTEGRITY_REF_TAG) + return "T10-DIF-TYPE1-IP"; + return "T10-DIF-TYPE3-IP"; + case BLK_INTEGRITY_CSUM_CRC: + if (bi->flags & BLK_INTEGRITY_REF_TAG) + return "T10-DIF-TYPE1-CRC"; + return "T10-DIF-TYPE3-CRC"; + case BLK_INTEGRITY_CSUM_CRC64: + if (bi->flags & BLK_INTEGRITY_REF_TAG) + return "EXT-DIF-TYPE1-CRC64"; + return "EXT-DIF-TYPE3-CRC64"; + case BLK_INTEGRITY_CSUM_NONE: + break; + } + + return "nop"; +} +EXPORT_SYMBOL_GPL(blk_integrity_profile_name); + static ssize_t format_show(struct device *dev, struct device_attribute *attr, char *page) { struct blk_integrity *bi = dev_to_bi(dev); - if (bi->profile && bi->profile->name) - return sysfs_emit(page, "%s\n", bi->profile->name); - return sysfs_emit(page, "none\n"); + if (!bi->tuple_size) + return sysfs_emit(page, "none\n"); + return sysfs_emit(page, "%s\n", blk_integrity_profile_name(bi)); } static ssize_t tag_size_show(struct device *dev, struct device_attribute *attr, @@ -326,28 +352,6 @@ const struct attribute_group blk_integrity_attr_group = { .attrs = integrity_attrs, }; -static blk_status_t blk_integrity_nop_fn(struct blk_integrity_iter *iter) -{ - return BLK_STS_OK; -} - -static void blk_integrity_nop_prepare(struct request *rq) -{ -} - -static void blk_integrity_nop_complete(struct request *rq, - unsigned int nr_bytes) -{ -} - -static const struct blk_integrity_profile nop_profile = { - .name = "nop", - .generate_fn = blk_integrity_nop_fn, - .verify_fn = blk_integrity_nop_fn, - .prepare_fn = blk_integrity_nop_prepare, - .complete_fn = blk_integrity_nop_complete, -}; - /** * blk_integrity_register - Register a gendisk as being integrity-capable * @disk: struct gendisk pointer to make integrity-aware @@ -363,11 +367,11 @@ void blk_integrity_register(struct gendisk *disk, struct blk_integrity *template { struct blk_integrity *bi = &disk->queue->integrity; + bi->csum_type = template->csum_type; bi->flags = BLK_INTEGRITY_VERIFY | BLK_INTEGRITY_GENERATE | template->flags; bi->interval_exp = template->interval_exp ? : ilog2(queue_logical_block_size(disk->queue)); - bi->profile = template->profile ? template->profile : &nop_profile; bi->tuple_size = template->tuple_size; bi->tag_size = template->tag_size; bi->pi_offset = template->pi_offset; @@ -394,7 +398,7 @@ void blk_integrity_unregister(struct gendisk *disk) { struct blk_integrity *bi = &disk->queue->integrity; - if (!bi->profile) + if (!bi->tuple_size) return; /* ensure all bios are off the integrity workqueue */ diff --git a/block/blk-mq.c b/block/blk-mq.c index 3b4df8e5ac9e5f..0d4cd39c3d25da 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -804,10 +804,8 @@ static void blk_complete_request(struct request *req) if (!bio) return; -#ifdef CONFIG_BLK_DEV_INTEGRITY if (blk_integrity_rq(req) && req_op(req) == REQ_OP_READ) - req->q->integrity.profile->complete_fn(req, total_bytes); -#endif + blk_integrity_complete(req, total_bytes); /* * Upper layers may call blk_crypto_evict_key() anytime after the last @@ -875,11 +873,9 @@ bool blk_update_request(struct request *req, blk_status_t error, if (!req->bio) return false; -#ifdef CONFIG_BLK_DEV_INTEGRITY if (blk_integrity_rq(req) && req_op(req) == REQ_OP_READ && error == BLK_STS_OK) - req->q->integrity.profile->complete_fn(req, nr_bytes); -#endif + blk_integrity_complete(req, nr_bytes); /* * Upper layers may call blk_crypto_evict_key() anytime after the last @@ -1264,10 +1260,9 @@ void blk_mq_start_request(struct request *rq) WRITE_ONCE(rq->state, MQ_RQ_IN_FLIGHT); rq->mq_hctx->tags->rqs[rq->tag] = rq; -#ifdef CONFIG_BLK_DEV_INTEGRITY if (blk_integrity_rq(rq) && req_op(rq) == REQ_OP_WRITE) - q->integrity.profile->prepare_fn(rq); -#endif + blk_integrity_prepare(rq); + if (rq->bio && rq->bio->bi_opf & REQ_POLLED) WRITE_ONCE(rq->bio->bi_cookie, rq->mq_hctx->queue_num); } diff --git a/block/blk.h b/block/blk.h index 189bc25beb502a..79e8d5d4fe0caf 100644 --- a/block/blk.h +++ b/block/blk.h @@ -9,6 +9,7 @@ #include #include "blk-crypto-internal.h" +struct blk_integrity_iter; struct elevator_type; /* Max future timer expiry for timeouts */ @@ -673,4 +674,11 @@ int bdev_open(struct block_device *bdev, blk_mode_t mode, void *holder, const struct blk_holder_ops *hops, struct file *bdev_file); int bdev_permission(dev_t dev, blk_mode_t mode, void *holder); +void blk_integrity_generate(struct blk_integrity_iter *iter, + struct blk_integrity *bi); +blk_status_t blk_integrity_verify(struct blk_integrity_iter *iter, + struct blk_integrity *bi); +void blk_integrity_prepare(struct request *rq); +void blk_integrity_complete(struct request *rq, unsigned int nr_bytes); + #endif /* BLK_INTERNAL_H */ diff --git a/block/t10-pi.c b/block/t10-pi.c index f4cc91156da1f2..dadecf621497bb 100644 --- a/block/t10-pi.c +++ b/block/t10-pi.c @@ -11,17 +11,14 @@ #include #include #include +#include "blk.h" -typedef __be16 (csum_fn) (__be16, void *, unsigned int); - -static __be16 t10_pi_crc_fn(__be16 crc, void *data, unsigned int len) -{ - return cpu_to_be16(crc_t10dif_update(be16_to_cpu(crc), data, len)); -} - -static __be16 t10_pi_ip_fn(__be16 csum, void *data, unsigned int len) +static __be16 t10_pi_csum(__be16 csum, void *data, unsigned int len, + unsigned char csum_type) { - return (__force __be16)ip_compute_csum(data, len); + if (csum_type == BLK_INTEGRITY_CSUM_IP) + return (__force __be16)ip_compute_csum(data, len); + return cpu_to_be16(crc_t10dif_update(be16_to_cpu(csum), data, len)); } /* @@ -29,48 +26,44 @@ static __be16 t10_pi_ip_fn(__be16 csum, void *data, unsigned int len) * 16 bit app tag, 32 bit reference tag. Type 3 does not define the ref * tag. */ -static blk_status_t t10_pi_generate(struct blk_integrity_iter *iter, - csum_fn *fn, enum t10_dif_type type) +static void t10_pi_generate(struct blk_integrity_iter *iter, + struct blk_integrity *bi) { - u8 offset = iter->pi_offset; + u8 offset = bi->pi_offset; unsigned int i; for (i = 0 ; i < iter->data_size ; i += iter->interval) { struct t10_pi_tuple *pi = iter->prot_buf + offset; - pi->guard_tag = fn(0, iter->data_buf, iter->interval); + pi->guard_tag = t10_pi_csum(0, iter->data_buf, iter->interval, + bi->csum_type); if (offset) - pi->guard_tag = fn(pi->guard_tag, iter->prot_buf, - offset); + pi->guard_tag = t10_pi_csum(pi->guard_tag, + iter->prot_buf, offset, bi->csum_type); pi->app_tag = 0; - if (type == T10_PI_TYPE1_PROTECTION) + if (bi->flags & BLK_INTEGRITY_REF_TAG) pi->ref_tag = cpu_to_be32(lower_32_bits(iter->seed)); else pi->ref_tag = 0; iter->data_buf += iter->interval; - iter->prot_buf += iter->tuple_size; + iter->prot_buf += bi->tuple_size; iter->seed++; } - - return BLK_STS_OK; } static blk_status_t t10_pi_verify(struct blk_integrity_iter *iter, - csum_fn *fn, enum t10_dif_type type) + struct blk_integrity *bi) { - u8 offset = iter->pi_offset; + u8 offset = bi->pi_offset; unsigned int i; - BUG_ON(type == T10_PI_TYPE0_PROTECTION); - for (i = 0 ; i < iter->data_size ; i += iter->interval) { struct t10_pi_tuple *pi = iter->prot_buf + offset; __be16 csum; - if (type == T10_PI_TYPE1_PROTECTION || - type == T10_PI_TYPE2_PROTECTION) { + if (bi->flags & BLK_INTEGRITY_REF_TAG) { if (pi->app_tag == T10_PI_APP_ESCAPE) goto next; @@ -82,15 +75,17 @@ static blk_status_t t10_pi_verify(struct blk_integrity_iter *iter, iter->seed, be32_to_cpu(pi->ref_tag)); return BLK_STS_PROTECTION; } - } else if (type == T10_PI_TYPE3_PROTECTION) { + } else { if (pi->app_tag == T10_PI_APP_ESCAPE && pi->ref_tag == T10_PI_REF_ESCAPE) goto next; } - csum = fn(0, iter->data_buf, iter->interval); + csum = t10_pi_csum(0, iter->data_buf, iter->interval, + bi->csum_type); if (offset) - csum = fn(csum, iter->prot_buf, offset); + csum = t10_pi_csum(csum, iter->prot_buf, offset, + bi->csum_type); if (pi->guard_tag != csum) { pr_err("%s: guard tag error at sector %llu " \ @@ -102,33 +97,13 @@ static blk_status_t t10_pi_verify(struct blk_integrity_iter *iter, next: iter->data_buf += iter->interval; - iter->prot_buf += iter->tuple_size; + iter->prot_buf += bi->tuple_size; iter->seed++; } return BLK_STS_OK; } -static blk_status_t t10_pi_type1_generate_crc(struct blk_integrity_iter *iter) -{ - return t10_pi_generate(iter, t10_pi_crc_fn, T10_PI_TYPE1_PROTECTION); -} - -static blk_status_t t10_pi_type1_generate_ip(struct blk_integrity_iter *iter) -{ - return t10_pi_generate(iter, t10_pi_ip_fn, T10_PI_TYPE1_PROTECTION); -} - -static blk_status_t t10_pi_type1_verify_crc(struct blk_integrity_iter *iter) -{ - return t10_pi_verify(iter, t10_pi_crc_fn, T10_PI_TYPE1_PROTECTION); -} - -static blk_status_t t10_pi_type1_verify_ip(struct blk_integrity_iter *iter) -{ - return t10_pi_verify(iter, t10_pi_ip_fn, T10_PI_TYPE1_PROTECTION); -} - /** * t10_pi_type1_prepare - prepare PI prior submitting request to device * @rq: request with PI that should be prepared @@ -225,81 +200,15 @@ static void t10_pi_type1_complete(struct request *rq, unsigned int nr_bytes) } } -static blk_status_t t10_pi_type3_generate_crc(struct blk_integrity_iter *iter) -{ - return t10_pi_generate(iter, t10_pi_crc_fn, T10_PI_TYPE3_PROTECTION); -} - -static blk_status_t t10_pi_type3_generate_ip(struct blk_integrity_iter *iter) -{ - return t10_pi_generate(iter, t10_pi_ip_fn, T10_PI_TYPE3_PROTECTION); -} - -static blk_status_t t10_pi_type3_verify_crc(struct blk_integrity_iter *iter) -{ - return t10_pi_verify(iter, t10_pi_crc_fn, T10_PI_TYPE3_PROTECTION); -} - -static blk_status_t t10_pi_type3_verify_ip(struct blk_integrity_iter *iter) -{ - return t10_pi_verify(iter, t10_pi_ip_fn, T10_PI_TYPE3_PROTECTION); -} - -/* Type 3 does not have a reference tag so no remapping is required. */ -static void t10_pi_type3_prepare(struct request *rq) -{ -} - -/* Type 3 does not have a reference tag so no remapping is required. */ -static void t10_pi_type3_complete(struct request *rq, unsigned int nr_bytes) -{ -} - -const struct blk_integrity_profile t10_pi_type1_crc = { - .name = "T10-DIF-TYPE1-CRC", - .generate_fn = t10_pi_type1_generate_crc, - .verify_fn = t10_pi_type1_verify_crc, - .prepare_fn = t10_pi_type1_prepare, - .complete_fn = t10_pi_type1_complete, -}; -EXPORT_SYMBOL(t10_pi_type1_crc); - -const struct blk_integrity_profile t10_pi_type1_ip = { - .name = "T10-DIF-TYPE1-IP", - .generate_fn = t10_pi_type1_generate_ip, - .verify_fn = t10_pi_type1_verify_ip, - .prepare_fn = t10_pi_type1_prepare, - .complete_fn = t10_pi_type1_complete, -}; -EXPORT_SYMBOL(t10_pi_type1_ip); - -const struct blk_integrity_profile t10_pi_type3_crc = { - .name = "T10-DIF-TYPE3-CRC", - .generate_fn = t10_pi_type3_generate_crc, - .verify_fn = t10_pi_type3_verify_crc, - .prepare_fn = t10_pi_type3_prepare, - .complete_fn = t10_pi_type3_complete, -}; -EXPORT_SYMBOL(t10_pi_type3_crc); - -const struct blk_integrity_profile t10_pi_type3_ip = { - .name = "T10-DIF-TYPE3-IP", - .generate_fn = t10_pi_type3_generate_ip, - .verify_fn = t10_pi_type3_verify_ip, - .prepare_fn = t10_pi_type3_prepare, - .complete_fn = t10_pi_type3_complete, -}; -EXPORT_SYMBOL(t10_pi_type3_ip); - static __be64 ext_pi_crc64(u64 crc, void *data, unsigned int len) { return cpu_to_be64(crc64_rocksoft_update(crc, data, len)); } -static blk_status_t ext_pi_crc64_generate(struct blk_integrity_iter *iter, - enum t10_dif_type type) +static void ext_pi_crc64_generate(struct blk_integrity_iter *iter, + struct blk_integrity *bi) { - u8 offset = iter->pi_offset; + u8 offset = bi->pi_offset; unsigned int i; for (i = 0 ; i < iter->data_size ; i += iter->interval) { @@ -311,17 +220,15 @@ static blk_status_t ext_pi_crc64_generate(struct blk_integrity_iter *iter, iter->prot_buf, offset); pi->app_tag = 0; - if (type == T10_PI_TYPE1_PROTECTION) + if (bi->flags & BLK_INTEGRITY_REF_TAG) put_unaligned_be48(iter->seed, pi->ref_tag); else put_unaligned_be48(0ULL, pi->ref_tag); iter->data_buf += iter->interval; - iter->prot_buf += iter->tuple_size; + iter->prot_buf += bi->tuple_size; iter->seed++; } - - return BLK_STS_OK; } static bool ext_pi_ref_escape(u8 *ref_tag) @@ -332,9 +239,9 @@ static bool ext_pi_ref_escape(u8 *ref_tag) } static blk_status_t ext_pi_crc64_verify(struct blk_integrity_iter *iter, - enum t10_dif_type type) + struct blk_integrity *bi) { - u8 offset = iter->pi_offset; + u8 offset = bi->pi_offset; unsigned int i; for (i = 0; i < iter->data_size; i += iter->interval) { @@ -342,7 +249,7 @@ static blk_status_t ext_pi_crc64_verify(struct blk_integrity_iter *iter, u64 ref, seed; __be64 csum; - if (type == T10_PI_TYPE1_PROTECTION) { + if (bi->flags & BLK_INTEGRITY_REF_TAG) { if (pi->app_tag == T10_PI_APP_ESCAPE) goto next; @@ -353,7 +260,7 @@ static blk_status_t ext_pi_crc64_verify(struct blk_integrity_iter *iter, iter->disk_name, seed, ref); return BLK_STS_PROTECTION; } - } else if (type == T10_PI_TYPE3_PROTECTION) { + } else { if (pi->app_tag == T10_PI_APP_ESCAPE && ext_pi_ref_escape(pi->ref_tag)) goto next; @@ -374,23 +281,13 @@ static blk_status_t ext_pi_crc64_verify(struct blk_integrity_iter *iter, next: iter->data_buf += iter->interval; - iter->prot_buf += iter->tuple_size; + iter->prot_buf += bi->tuple_size; iter->seed++; } return BLK_STS_OK; } -static blk_status_t ext_pi_type1_verify_crc64(struct blk_integrity_iter *iter) -{ - return ext_pi_crc64_verify(iter, T10_PI_TYPE1_PROTECTION); -} - -static blk_status_t ext_pi_type1_generate_crc64(struct blk_integrity_iter *iter) -{ - return ext_pi_crc64_generate(iter, T10_PI_TYPE1_PROTECTION); -} - static void ext_pi_type1_prepare(struct request *rq) { struct blk_integrity *bi = &rq->q->integrity; @@ -467,33 +364,61 @@ static void ext_pi_type1_complete(struct request *rq, unsigned int nr_bytes) } } -static blk_status_t ext_pi_type3_verify_crc64(struct blk_integrity_iter *iter) +void blk_integrity_generate(struct blk_integrity_iter *iter, + struct blk_integrity *bi) { - return ext_pi_crc64_verify(iter, T10_PI_TYPE3_PROTECTION); + switch (bi->csum_type) { + case BLK_INTEGRITY_CSUM_CRC64: + ext_pi_crc64_generate(iter, bi); + break; + case BLK_INTEGRITY_CSUM_CRC: + case BLK_INTEGRITY_CSUM_IP: + t10_pi_generate(iter, bi); + break; + default: + break; + } } -static blk_status_t ext_pi_type3_generate_crc64(struct blk_integrity_iter *iter) +blk_status_t blk_integrity_verify(struct blk_integrity_iter *iter, + struct blk_integrity *bi) { - return ext_pi_crc64_generate(iter, T10_PI_TYPE3_PROTECTION); + switch (bi->csum_type) { + case BLK_INTEGRITY_CSUM_CRC64: + return ext_pi_crc64_verify(iter, bi); + case BLK_INTEGRITY_CSUM_CRC: + case BLK_INTEGRITY_CSUM_IP: + return t10_pi_verify(iter, bi); + default: + return BLK_STS_OK; + } } -const struct blk_integrity_profile ext_pi_type1_crc64 = { - .name = "EXT-DIF-TYPE1-CRC64", - .generate_fn = ext_pi_type1_generate_crc64, - .verify_fn = ext_pi_type1_verify_crc64, - .prepare_fn = ext_pi_type1_prepare, - .complete_fn = ext_pi_type1_complete, -}; -EXPORT_SYMBOL_GPL(ext_pi_type1_crc64); - -const struct blk_integrity_profile ext_pi_type3_crc64 = { - .name = "EXT-DIF-TYPE3-CRC64", - .generate_fn = ext_pi_type3_generate_crc64, - .verify_fn = ext_pi_type3_verify_crc64, - .prepare_fn = t10_pi_type3_prepare, - .complete_fn = t10_pi_type3_complete, -}; -EXPORT_SYMBOL_GPL(ext_pi_type3_crc64); +void blk_integrity_prepare(struct request *rq) +{ + struct blk_integrity *bi = &rq->q->integrity; + + if (!(bi->flags & BLK_INTEGRITY_REF_TAG)) + return; + + if (bi->csum_type == BLK_INTEGRITY_CSUM_CRC64) + ext_pi_type1_prepare(rq); + else + t10_pi_type1_prepare(rq); +} + +void blk_integrity_complete(struct request *rq, unsigned int nr_bytes) +{ + struct blk_integrity *bi = &rq->q->integrity; + + if (!(bi->flags & BLK_INTEGRITY_REF_TAG)) + return; + + if (bi->csum_type == BLK_INTEGRITY_CSUM_CRC64) + ext_pi_type1_complete(rq, nr_bytes); + else + t10_pi_type1_complete(rq, nr_bytes); +} MODULE_DESCRIPTION("T10 Protection Information module"); MODULE_LICENSE("GPL"); diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c index 1dfc462f29cd6f..6c013ceb0e5f1d 100644 --- a/drivers/md/dm-crypt.c +++ b/drivers/md/dm-crypt.c @@ -1177,7 +1177,7 @@ static int crypt_integrity_ctr(struct crypt_config *cc, struct dm_target *ti) struct mapped_device *md = dm_table_get_md(ti->table); /* We require an underlying device with non-PI metadata */ - if (!bi || strcmp(bi->profile->name, "nop")) { + if (!bi || bi->csum_type != BLK_INTEGRITY_CSUM_NONE) { ti->error = "Integrity profile not supported."; return -EINVAL; } diff --git a/drivers/nvme/host/Kconfig b/drivers/nvme/host/Kconfig index b309c8be720f47..a3caef75aa0a83 100644 --- a/drivers/nvme/host/Kconfig +++ b/drivers/nvme/host/Kconfig @@ -1,7 +1,6 @@ # SPDX-License-Identifier: GPL-2.0-only config NVME_CORE tristate - select BLK_DEV_INTEGRITY_T10 if BLK_DEV_INTEGRITY config BLK_DEV_NVME tristate "NVM Express block device" diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index f5d150c62955d8..14bac248cde4ca 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -1744,17 +1744,16 @@ static bool nvme_init_integrity(struct gendisk *disk, struct nvme_ns_head *head) case NVME_NS_DPS_PI_TYPE3: switch (head->guard_type) { case NVME_NVM_NS_16B_GUARD: - integrity.profile = &t10_pi_type3_crc; + integrity.csum_type = BLK_INTEGRITY_CSUM_CRC; integrity.tag_size = sizeof(u16) + sizeof(u32); integrity.flags |= BLK_INTEGRITY_DEVICE_CAPABLE; break; case NVME_NVM_NS_64B_GUARD: - integrity.profile = &ext_pi_type3_crc64; + integrity.csum_type = BLK_INTEGRITY_CSUM_CRC64; integrity.tag_size = sizeof(u16) + 6; integrity.flags |= BLK_INTEGRITY_DEVICE_CAPABLE; break; default: - integrity.profile = NULL; break; } break; @@ -1762,22 +1761,22 @@ static bool nvme_init_integrity(struct gendisk *disk, struct nvme_ns_head *head) case NVME_NS_DPS_PI_TYPE2: switch (head->guard_type) { case NVME_NVM_NS_16B_GUARD: - integrity.profile = &t10_pi_type1_crc; + integrity.csum_type = BLK_INTEGRITY_CSUM_CRC; integrity.tag_size = sizeof(u16); - integrity.flags |= BLK_INTEGRITY_DEVICE_CAPABLE; + integrity.flags |= BLK_INTEGRITY_DEVICE_CAPABLE | + BLK_INTEGRITY_REF_TAG; break; case NVME_NVM_NS_64B_GUARD: - integrity.profile = &ext_pi_type1_crc64; + integrity.csum_type = BLK_INTEGRITY_CSUM_CRC64; integrity.tag_size = sizeof(u16); - integrity.flags |= BLK_INTEGRITY_DEVICE_CAPABLE; + integrity.flags |= BLK_INTEGRITY_DEVICE_CAPABLE | + BLK_INTEGRITY_REF_TAG; break; default: - integrity.profile = NULL; break; } break; default: - integrity.profile = NULL; break; } diff --git a/drivers/nvme/target/Kconfig b/drivers/nvme/target/Kconfig index 872dd1a0acd804..c42aec41cc7b1f 100644 --- a/drivers/nvme/target/Kconfig +++ b/drivers/nvme/target/Kconfig @@ -6,7 +6,6 @@ config NVME_TARGET depends on CONFIGFS_FS select NVME_KEYRING if NVME_TARGET_TCP_TLS select KEYS if NVME_TARGET_TCP_TLS - select BLK_DEV_INTEGRITY_T10 if BLK_DEV_INTEGRITY select SGL_ALLOC help This enabled target side support for the NVMe protocol, that is diff --git a/drivers/nvme/target/io-cmd-bdev.c b/drivers/nvme/target/io-cmd-bdev.c index 6426aac2634aeb..b628bc5ee99847 100644 --- a/drivers/nvme/target/io-cmd-bdev.c +++ b/drivers/nvme/target/io-cmd-bdev.c @@ -61,15 +61,17 @@ static void nvmet_bdev_ns_enable_integrity(struct nvmet_ns *ns) { struct blk_integrity *bi = bdev_get_integrity(ns->bdev); - if (bi) { + if (!bi) + return; + + if (bi->csum_type == BLK_INTEGRITY_CSUM_CRC) { ns->metadata_size = bi->tuple_size; - if (bi->profile == &t10_pi_type1_crc) + if (bi->flags & BLK_INTEGRITY_REF_TAG) ns->pi_type = NVME_NS_DPS_PI_TYPE1; - else if (bi->profile == &t10_pi_type3_crc) - ns->pi_type = NVME_NS_DPS_PI_TYPE3; else - /* Unsupported metadata type */ - ns->metadata_size = 0; + ns->pi_type = NVME_NS_DPS_PI_TYPE3; + } else { + ns->metadata_size = 0; } } @@ -102,7 +104,7 @@ int nvmet_bdev_ns_enable(struct nvmet_ns *ns) ns->pi_type = 0; ns->metadata_size = 0; - if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY_T10)) + if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)) nvmet_bdev_ns_enable_integrity(ns); if (bdev_is_zoned(ns->bdev)) { diff --git a/drivers/scsi/Kconfig b/drivers/scsi/Kconfig index 065db86d602164..37c24ffea65cc0 100644 --- a/drivers/scsi/Kconfig +++ b/drivers/scsi/Kconfig @@ -82,7 +82,6 @@ comment "SCSI support type (disk, tape, CD-ROM)" config BLK_DEV_SD tristate "SCSI disk support" depends on SCSI - select BLK_DEV_INTEGRITY_T10 if BLK_DEV_INTEGRITY help If you want to use SCSI hard disks, Fibre Channel disks, Serial ATA (SATA) or Parallel ATA (PATA) hard disks, diff --git a/drivers/scsi/sd_dif.c b/drivers/scsi/sd_dif.c index 1df847b5f74764..6f0921c7db787b 100644 --- a/drivers/scsi/sd_dif.c +++ b/drivers/scsi/sd_dif.c @@ -47,18 +47,13 @@ void sd_dif_config_host(struct scsi_disk *sdkp) memset(&bi, 0, sizeof(bi)); /* Enable DMA of protection information */ - if (scsi_host_get_guard(sdkp->device->host) & SHOST_DIX_GUARD_IP) { - if (type == T10_PI_TYPE3_PROTECTION) - bi.profile = &t10_pi_type3_ip; - else - bi.profile = &t10_pi_type1_ip; + if (scsi_host_get_guard(sdkp->device->host) & SHOST_DIX_GUARD_IP) + bi.csum_type = BLK_INTEGRITY_CSUM_IP; + else + bi.csum_type = BLK_INTEGRITY_CSUM_CRC; - bi.flags |= BLK_INTEGRITY_IP_CHECKSUM; - } else - if (type == T10_PI_TYPE3_PROTECTION) - bi.profile = &t10_pi_type3_crc; - else - bi.profile = &t10_pi_type1_crc; + if (type != T10_PI_TYPE3_PROTECTION) + bi.flags |= BLK_INTEGRITY_REF_TAG; bi.tuple_size = sizeof(struct t10_pi_tuple); @@ -76,7 +71,7 @@ void sd_dif_config_host(struct scsi_disk *sdkp) sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIX %s, application tag size %u bytes\n", - bi.profile->name, bi.tag_size); + blk_integrity_profile_name(&bi), bi.tag_size); out: blk_integrity_register(disk, &bi); } diff --git a/drivers/target/target_core_iblock.c b/drivers/target/target_core_iblock.c index 7f6ca81778453b..a3e09adc4e767c 100644 --- a/drivers/target/target_core_iblock.c +++ b/drivers/target/target_core_iblock.c @@ -148,35 +148,38 @@ static int iblock_configure_device(struct se_device *dev) dev->dev_attrib.is_nonrot = 1; bi = bdev_get_integrity(bd); - if (bi) { - struct bio_set *bs = &ib_dev->ibd_bio_set; - - if (!strcmp(bi->profile->name, "T10-DIF-TYPE3-IP") || - !strcmp(bi->profile->name, "T10-DIF-TYPE1-IP")) { - pr_err("IBLOCK export of blk_integrity: %s not" - " supported\n", bi->profile->name); - ret = -ENOSYS; - goto out_blkdev_put; - } + if (!bi) + return 0; - if (!strcmp(bi->profile->name, "T10-DIF-TYPE3-CRC")) { - dev->dev_attrib.pi_prot_type = TARGET_DIF_TYPE3_PROT; - } else if (!strcmp(bi->profile->name, "T10-DIF-TYPE1-CRC")) { + switch (bi->csum_type) { + case BLK_INTEGRITY_CSUM_IP: + pr_err("IBLOCK export of blk_integrity: %s not supported\n", + blk_integrity_profile_name(bi)); + ret = -ENOSYS; + goto out_blkdev_put; + case BLK_INTEGRITY_CSUM_CRC: + if (bi->flags & BLK_INTEGRITY_REF_TAG) dev->dev_attrib.pi_prot_type = TARGET_DIF_TYPE1_PROT; - } + else + dev->dev_attrib.pi_prot_type = TARGET_DIF_TYPE3_PROT; + break; + default: + break; + } - if (dev->dev_attrib.pi_prot_type) { - if (bioset_integrity_create(bs, IBLOCK_BIO_POOL_SIZE) < 0) { - pr_err("Unable to allocate bioset for PI\n"); - ret = -ENOMEM; - goto out_blkdev_put; - } - pr_debug("IBLOCK setup BIP bs->bio_integrity_pool: %p\n", - &bs->bio_integrity_pool); + if (dev->dev_attrib.pi_prot_type) { + struct bio_set *bs = &ib_dev->ibd_bio_set; + + if (bioset_integrity_create(bs, IBLOCK_BIO_POOL_SIZE) < 0) { + pr_err("Unable to allocate bioset for PI\n"); + ret = -ENOMEM; + goto out_blkdev_put; } - dev->dev_attrib.hw_pi_prot_type = dev->dev_attrib.pi_prot_type; + pr_debug("IBLOCK setup BIP bs->bio_integrity_pool: %p\n", + &bs->bio_integrity_pool); } + dev->dev_attrib.hw_pi_prot_type = dev->dev_attrib.pi_prot_type; return 0; out_blkdev_put: diff --git a/include/linux/blk-integrity.h b/include/linux/blk-integrity.h index 7428cb43952da0..56ce1ae355805d 100644 --- a/include/linux/blk-integrity.h +++ b/include/linux/blk-integrity.h @@ -10,7 +10,7 @@ enum blk_integrity_flags { BLK_INTEGRITY_VERIFY = 1 << 0, BLK_INTEGRITY_GENERATE = 1 << 1, BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, - BLK_INTEGRITY_IP_CHECKSUM = 1 << 3, + BLK_INTEGRITY_REF_TAG = 1 << 3, }; struct blk_integrity_iter { @@ -19,22 +19,10 @@ struct blk_integrity_iter { sector_t seed; unsigned int data_size; unsigned short interval; - unsigned char tuple_size; - unsigned char pi_offset; const char *disk_name; }; -typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *); -typedef void (integrity_prepare_fn) (struct request *); -typedef void (integrity_complete_fn) (struct request *, unsigned int); - -struct blk_integrity_profile { - integrity_processing_fn *generate_fn; - integrity_processing_fn *verify_fn; - integrity_prepare_fn *prepare_fn; - integrity_complete_fn *complete_fn; - const char *name; -}; +const char *blk_integrity_profile_name(struct blk_integrity *bi); #ifdef CONFIG_BLK_DEV_INTEGRITY void blk_integrity_register(struct gendisk *, struct blk_integrity *); @@ -44,14 +32,17 @@ int blk_rq_map_integrity_sg(struct request_queue *, struct bio *, struct scatterlist *); int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); -static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) +static inline bool +blk_integrity_queue_supports_integrity(struct request_queue *q) { - struct blk_integrity *bi = &disk->queue->integrity; + return q->integrity.tuple_size; +} - if (!bi->profile) +static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) +{ + if (!blk_integrity_queue_supports_integrity(disk->queue)) return NULL; - - return bi; + return &disk->queue->integrity; } static inline struct blk_integrity * @@ -60,12 +51,6 @@ bdev_get_integrity(struct block_device *bdev) return blk_get_integrity(bdev->bd_disk); } -static inline bool -blk_integrity_queue_supports_integrity(struct request_queue *q) -{ - return q->integrity.profile; -} - static inline unsigned short queue_max_integrity_segments(const struct request_queue *q) { diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index ac8e0cb2353a0e..bdd33388e1ced8 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -105,9 +105,16 @@ enum { struct disk_events; struct badblocks; +enum blk_integrity_checksum { + BLK_INTEGRITY_CSUM_NONE = 0, + BLK_INTEGRITY_CSUM_IP = 1, + BLK_INTEGRITY_CSUM_CRC = 2, + BLK_INTEGRITY_CSUM_CRC64 = 3, +} __packed ; + struct blk_integrity { - const struct blk_integrity_profile *profile; unsigned char flags; + enum blk_integrity_checksum csum_type; unsigned char tuple_size; unsigned char pi_offset; unsigned char interval_exp; diff --git a/include/linux/t10-pi.h b/include/linux/t10-pi.h index 248f4ac9564258..d2bafb76badfb9 100644 --- a/include/linux/t10-pi.h +++ b/include/linux/t10-pi.h @@ -48,11 +48,6 @@ static inline u32 t10_pi_ref_tag(struct request *rq) return blk_rq_pos(rq) >> (shift - SECTOR_SHIFT) & 0xffffffff; } -extern const struct blk_integrity_profile t10_pi_type1_crc; -extern const struct blk_integrity_profile t10_pi_type1_ip; -extern const struct blk_integrity_profile t10_pi_type3_crc; -extern const struct blk_integrity_profile t10_pi_type3_ip; - struct crc64_pi_tuple { __be64 guard_tag; __be16 app_tag; @@ -79,7 +74,4 @@ static inline u64 ext_pi_ref_tag(struct request *rq) return lower_48_bits(blk_rq_pos(rq) >> (shift - SECTOR_SHIFT)); } -extern const struct blk_integrity_profile ext_pi_type1_crc64; -extern const struct blk_integrity_profile ext_pi_type3_crc64; - #endif From e8bc14d116aeac8f0f133ec8d249acf4e0658da7 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:16 +0200 Subject: [PATCH 042/209] block: remove the blk_flush_integrity call in blk_integrity_unregister Now that there are no indirect calls for PI processing there is no way to dereference a NULL pointer here. Additionally drivers now always freeze the queue (or in case of stacking drivers use their internal equivalent) around changing the integrity profile. This is effectively a revert of commit 3df49967f6f1 ("block: flush the integrity workqueue in blk_integrity_unregister"). Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240613084839.1044015-7-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-integrity.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/block/blk-integrity.c b/block/blk-integrity.c index 17d37badfbb8bc..24f04575096d39 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -401,8 +401,6 @@ void blk_integrity_unregister(struct gendisk *disk) if (!bi->tuple_size) return; - /* ensure all bios are off the integrity workqueue */ - blk_flush_integrity(); blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, disk->queue); memset(bi, 0, sizeof(*bi)); } From 1366251a794b149a132ef8423c8946b6e565a923 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:17 +0200 Subject: [PATCH 043/209] block: factor out flag_{store,show} helper for integrity Factor the duplicate code for the generate and verify attributes into common helpers. Signed-off-by: Christoph Hellwig Reviewed-by: Keith Busch Reviewed-by: Chaitanya Kulkarni Reviewed-by: Martin K. Petersen Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240613084839.1044015-8-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-integrity.c | 53 +++++++++++++++++++++---------------------- 1 file changed, 26 insertions(+), 27 deletions(-) diff --git a/block/blk-integrity.c b/block/blk-integrity.c index 24f04575096d39..24671d9f90a124 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -243,6 +243,28 @@ const char *blk_integrity_profile_name(struct blk_integrity *bi) } EXPORT_SYMBOL_GPL(blk_integrity_profile_name); +static ssize_t flag_store(struct device *dev, struct device_attribute *attr, + const char *page, size_t count, unsigned char flag) +{ + struct blk_integrity *bi = dev_to_bi(dev); + char *p = (char *) page; + unsigned long val = simple_strtoul(p, &p, 10); + + if (val) + bi->flags |= flag; + else + bi->flags &= ~flag; + return count; +} + +static ssize_t flag_show(struct device *dev, struct device_attribute *attr, + char *page, unsigned char flag) +{ + struct blk_integrity *bi = dev_to_bi(dev); + + return sysfs_emit(page, "%d\n", !!(bi->flags & flag)); +} + static ssize_t format_show(struct device *dev, struct device_attribute *attr, char *page) { @@ -275,49 +297,26 @@ static ssize_t read_verify_store(struct device *dev, struct device_attribute *attr, const char *page, size_t count) { - struct blk_integrity *bi = dev_to_bi(dev); - char *p = (char *) page; - unsigned long val = simple_strtoul(p, &p, 10); - - if (val) - bi->flags |= BLK_INTEGRITY_VERIFY; - else - bi->flags &= ~BLK_INTEGRITY_VERIFY; - - return count; + return flag_store(dev, attr, page, count, BLK_INTEGRITY_VERIFY); } static ssize_t read_verify_show(struct device *dev, struct device_attribute *attr, char *page) { - struct blk_integrity *bi = dev_to_bi(dev); - - return sysfs_emit(page, "%d\n", !!(bi->flags & BLK_INTEGRITY_VERIFY)); + return flag_show(dev, attr, page, BLK_INTEGRITY_VERIFY); } static ssize_t write_generate_store(struct device *dev, struct device_attribute *attr, const char *page, size_t count) { - struct blk_integrity *bi = dev_to_bi(dev); - - char *p = (char *) page; - unsigned long val = simple_strtoul(p, &p, 10); - - if (val) - bi->flags |= BLK_INTEGRITY_GENERATE; - else - bi->flags &= ~BLK_INTEGRITY_GENERATE; - - return count; + return flag_store(dev, attr, page, count, BLK_INTEGRITY_GENERATE); } static ssize_t write_generate_show(struct device *dev, struct device_attribute *attr, char *page) { - struct blk_integrity *bi = dev_to_bi(dev); - - return sysfs_emit(page, "%d\n", !!(bi->flags & BLK_INTEGRITY_GENERATE)); + return flag_show(dev, attr, page, BLK_INTEGRITY_GENERATE); } static ssize_t device_is_integrity_capable_show(struct device *dev, From 1d59857ed2ec4d506e346859713c4325b5053da3 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:18 +0200 Subject: [PATCH 044/209] block: use kstrtoul in flag_store Use the text to integer helper that has error handling and doesn't modify the input pointer. Signed-off-by: Christoph Hellwig Reviewed-by: Keith Busch Reviewed-by: Chaitanya Kulkarni Reviewed-by: Martin K. Petersen Reviewed-by: Hannes Reinecke Reviewed-by: Kanchan Joshi Link: https://lore.kernel.org/r/20240613084839.1044015-9-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-integrity.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/block/blk-integrity.c b/block/blk-integrity.c index 24671d9f90a124..58760a6d6b2209 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -247,8 +247,12 @@ static ssize_t flag_store(struct device *dev, struct device_attribute *attr, const char *page, size_t count, unsigned char flag) { struct blk_integrity *bi = dev_to_bi(dev); - char *p = (char *) page; - unsigned long val = simple_strtoul(p, &p, 10); + unsigned long val; + int err; + + err = kstrtoul(page, 10, &val); + if (err) + return err; if (val) bi->flags |= flag; From 43c5dbe98a3953e07f4fbf89aa137b9207d52378 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:19 +0200 Subject: [PATCH 045/209] block: don't require stable pages for non-PI metadata Non-PI metadata doesn't contain checksums and thus doesn't require stable pages. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240613084839.1044015-10-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-integrity.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/block/blk-integrity.c b/block/blk-integrity.c index 58760a6d6b2209..1d2d371cd632d3 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -379,7 +379,8 @@ void blk_integrity_register(struct gendisk *disk, struct blk_integrity *template bi->tag_size = template->tag_size; bi->pi_offset = template->pi_offset; - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, disk->queue); + if (bi->csum_type != BLK_INTEGRITY_CSUM_NONE) + blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, disk->queue); #ifdef CONFIG_BLK_INLINE_ENCRYPTION if (disk->queue->crypto_profile) { @@ -404,7 +405,8 @@ void blk_integrity_unregister(struct gendisk *disk) if (!bi->tuple_size) return; - blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, disk->queue); + if (bi->csum_type != BLK_INTEGRITY_CSUM_NONE) + blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, disk->queue); memset(bi, 0, sizeof(*bi)); } EXPORT_SYMBOL(blk_integrity_unregister); From 3c3e85ddffae93eba1a257eb6939bf5dc1e93b9e Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:20 +0200 Subject: [PATCH 046/209] block: bypass the STABLE_WRITES flag for protection information Currently registering a checksum-enabled (aka PI) integrity profile sets the QUEUE_FLAG_STABLE_WRITE flag, and unregistering it clears the flag. This can incorrectly clear the flag when the driver requires stable writes even without PI, e.g. in case of iSCSI or NVMe/TCP with data digest enabled. Fix this by looking at the csum_type directly in bdev_stable_writes and not setting the queue flag. Also remove the blk_queue_stable_writes helper as the only user in nvme wants to only look at the actual QUEUE_FLAG_STABLE_WRITE flag as it inherits the integrity configuration by other means. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240613084839.1044015-11-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-integrity.c | 6 ------ drivers/nvme/host/multipath.c | 3 ++- include/linux/blkdev.h | 12 ++++++++---- 3 files changed, 10 insertions(+), 11 deletions(-) diff --git a/block/blk-integrity.c b/block/blk-integrity.c index 1d2d371cd632d3..bec0d1df387ce9 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -379,9 +379,6 @@ void blk_integrity_register(struct gendisk *disk, struct blk_integrity *template bi->tag_size = template->tag_size; bi->pi_offset = template->pi_offset; - if (bi->csum_type != BLK_INTEGRITY_CSUM_NONE) - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, disk->queue); - #ifdef CONFIG_BLK_INLINE_ENCRYPTION if (disk->queue->crypto_profile) { pr_warn("blk-integrity: Integrity and hardware inline encryption are not supported together. Disabling hardware inline encryption.\n"); @@ -404,9 +401,6 @@ void blk_integrity_unregister(struct gendisk *disk) if (!bi->tuple_size) return; - - if (bi->csum_type != BLK_INTEGRITY_CSUM_NONE) - blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, disk->queue); memset(bi, 0, sizeof(*bi)); } EXPORT_SYMBOL(blk_integrity_unregister); diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c index d8b6b4648eaff9..12c59db02539e5 100644 --- a/drivers/nvme/host/multipath.c +++ b/drivers/nvme/host/multipath.c @@ -875,7 +875,8 @@ void nvme_mpath_add_disk(struct nvme_ns *ns, __le32 anagrpid) nvme_mpath_set_live(ns); } - if (blk_queue_stable_writes(ns->queue) && ns->head->disk) + if (test_bit(QUEUE_FLAG_STABLE_WRITES, &ns->queue->queue_flags) && + ns->head->disk) blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, ns->head->disk->queue); #ifdef CONFIG_BLK_DEV_ZONED diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index bdd33388e1ced8..f9089750919c6b 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -571,8 +571,6 @@ bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); #define blk_queue_noxmerges(q) \ test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) -#define blk_queue_stable_writes(q) \ - test_bit(QUEUE_FLAG_STABLE_WRITES, &(q)->queue_flags) #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) #define blk_queue_zone_resetall(q) \ @@ -1300,8 +1298,14 @@ static inline bool bdev_synchronous(struct block_device *bdev) static inline bool bdev_stable_writes(struct block_device *bdev) { - return test_bit(QUEUE_FLAG_STABLE_WRITES, - &bdev_get_queue(bdev)->queue_flags); + struct request_queue *q = bdev_get_queue(bdev); + +#ifdef CONFIG_BLK_DEV_INTEGRITY + /* BLK_INTEGRITY_CSUM_NONE is not available in blkdev.h */ + if (q->integrity.csum_type != 0) + return true; +#endif + return test_bit(QUEUE_FLAG_STABLE_WRITES, &q->queue_flags); } static inline bool bdev_write_cache(struct block_device *bdev) From 9f4aa46f2a7401025d8561495cf8740f773310fc Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:21 +0200 Subject: [PATCH 047/209] block: invert the BLK_INTEGRITY_{GENERATE,VERIFY} flags Invert the flags so that user set values will be able to persist revalidating the integrity information once we switch the integrity information to queue_limits. Signed-off-by: Christoph Hellwig Reviewed-by: Chaitanya Kulkarni Reviewed-by: Hannes Reinecke Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240613084839.1044015-12-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 4 ++-- block/blk-integrity.c | 18 +++++++++--------- include/linux/blk-integrity.h | 4 ++-- 3 files changed, 13 insertions(+), 13 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index 31dbc2853f92e3..173ffd4d623788 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -448,10 +448,10 @@ bool bio_integrity_prep(struct bio *bio) return true; if (bio_data_dir(bio) == READ) { - if (!(bi->flags & BLK_INTEGRITY_VERIFY)) + if (bi->flags & BLK_INTEGRITY_NOVERIFY) return true; } else { - if (!(bi->flags & BLK_INTEGRITY_GENERATE)) + if (bi->flags & BLK_INTEGRITY_NOGENERATE) return true; /* diff --git a/block/blk-integrity.c b/block/blk-integrity.c index bec0d1df387ce9..b37b8855eed147 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -254,10 +254,11 @@ static ssize_t flag_store(struct device *dev, struct device_attribute *attr, if (err) return err; + /* the flags are inverted vs the values in the sysfs files */ if (val) - bi->flags |= flag; - else bi->flags &= ~flag; + else + bi->flags |= flag; return count; } @@ -266,7 +267,7 @@ static ssize_t flag_show(struct device *dev, struct device_attribute *attr, { struct blk_integrity *bi = dev_to_bi(dev); - return sysfs_emit(page, "%d\n", !!(bi->flags & flag)); + return sysfs_emit(page, "%d\n", !(bi->flags & flag)); } static ssize_t format_show(struct device *dev, struct device_attribute *attr, @@ -301,26 +302,26 @@ static ssize_t read_verify_store(struct device *dev, struct device_attribute *attr, const char *page, size_t count) { - return flag_store(dev, attr, page, count, BLK_INTEGRITY_VERIFY); + return flag_store(dev, attr, page, count, BLK_INTEGRITY_NOVERIFY); } static ssize_t read_verify_show(struct device *dev, struct device_attribute *attr, char *page) { - return flag_show(dev, attr, page, BLK_INTEGRITY_VERIFY); + return flag_show(dev, attr, page, BLK_INTEGRITY_NOVERIFY); } static ssize_t write_generate_store(struct device *dev, struct device_attribute *attr, const char *page, size_t count) { - return flag_store(dev, attr, page, count, BLK_INTEGRITY_GENERATE); + return flag_store(dev, attr, page, count, BLK_INTEGRITY_NOGENERATE); } static ssize_t write_generate_show(struct device *dev, struct device_attribute *attr, char *page) { - return flag_show(dev, attr, page, BLK_INTEGRITY_GENERATE); + return flag_show(dev, attr, page, BLK_INTEGRITY_NOGENERATE); } static ssize_t device_is_integrity_capable_show(struct device *dev, @@ -371,8 +372,7 @@ void blk_integrity_register(struct gendisk *disk, struct blk_integrity *template struct blk_integrity *bi = &disk->queue->integrity; bi->csum_type = template->csum_type; - bi->flags = BLK_INTEGRITY_VERIFY | BLK_INTEGRITY_GENERATE | - template->flags; + bi->flags = template->flags; bi->interval_exp = template->interval_exp ? : ilog2(queue_logical_block_size(disk->queue)); bi->tuple_size = template->tuple_size; diff --git a/include/linux/blk-integrity.h b/include/linux/blk-integrity.h index 56ce1ae355805d..bafa01d4e7f95b 100644 --- a/include/linux/blk-integrity.h +++ b/include/linux/blk-integrity.h @@ -7,8 +7,8 @@ struct request; enum blk_integrity_flags { - BLK_INTEGRITY_VERIFY = 1 << 0, - BLK_INTEGRITY_GENERATE = 1 << 1, + BLK_INTEGRITY_NOVERIFY = 1 << 0, + BLK_INTEGRITY_NOGENERATE = 1 << 1, BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, BLK_INTEGRITY_REF_TAG = 1 << 3, }; From c6e56cf6b2e79a463af21286ba951714ed20828c Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 13 Jun 2024 10:48:22 +0200 Subject: [PATCH 048/209] block: move integrity information into queue_limits Move the integrity information into the queue limits so that it can be set atomically with other queue limits, and that the sysfs changes to the read_verify and write_generate flags are properly synchronized. This also allows to provide a more useful helper to stack the integrity fields, although it still is separate from the main stacking function as not all stackable devices want to inherit the integrity settings. Even with that it greatly simplifies the code in md and dm. Note that the integrity field is moved as-is into the queue limits. While there are good arguments for removing the separate blk_integrity structure, this would cause a lot of churn and might better be done at a later time if desired. However the integrity field in the queue_limits structure is now unconditional so that various ifdefs can be avoided or replaced with IS_ENABLED(). Given that tiny size of it that seems like a worthwhile trade off. Signed-off-by: Christoph Hellwig Reviewed-by: Hannes Reinecke Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240613084839.1044015-13-hch@lst.de Signed-off-by: Jens Axboe --- Documentation/block/data-integrity.rst | 49 +------- block/blk-integrity.c | 124 ++----------------- block/blk-settings.c | 118 +++++++++++++++++- block/t10-pi.c | 12 +- drivers/md/dm-core.h | 1 - drivers/md/dm-integrity.c | 27 ++--- drivers/md/dm-table.c | 161 +++++-------------------- drivers/md/md.c | 72 +++-------- drivers/md/md.h | 5 +- drivers/md/raid0.c | 7 +- drivers/md/raid1.c | 10 +- drivers/md/raid10.c | 10 +- drivers/md/raid5.c | 2 +- drivers/nvdimm/btt.c | 13 +- drivers/nvme/host/core.c | 70 +++++------ drivers/scsi/sd.c | 8 +- drivers/scsi/sd.h | 12 +- drivers/scsi/sd_dif.c | 34 +++--- include/linux/blk-integrity.h | 27 ++--- include/linux/blkdev.h | 12 +- include/linux/t10-pi.h | 12 +- 21 files changed, 289 insertions(+), 497 deletions(-) diff --git a/Documentation/block/data-integrity.rst b/Documentation/block/data-integrity.rst index 6a760c0eb1924e..99905e880a0e56 100644 --- a/Documentation/block/data-integrity.rst +++ b/Documentation/block/data-integrity.rst @@ -153,18 +153,11 @@ bio_free() will automatically free the bip. 4.2 Block Device ---------------- -Because the format of the protection data is tied to the physical -disk, each block device has been extended with a block integrity -profile (struct blk_integrity). This optional profile is registered -with the block layer using blk_integrity_register(). - -The profile contains callback functions for generating and verifying -the protection data, as well as getting and setting application tags. -The profile also contains a few constants to aid in completing, -merging and splitting the integrity metadata. +Block devices can set up the integrity information in the integrity +sub-struture of the queue_limits structure. Layered block devices will need to pick a profile that's appropriate -for all subdevices. blk_integrity_compare() can help with that. DM +for all subdevices. queue_limits_stack_integrity() can help with that. DM and MD linear, RAID0 and RAID1 are currently supported. RAID4/5/6 will require extra work due to the application tag. @@ -250,42 +243,6 @@ will require extra work due to the application tag. integrity upon completion. -5.4 Registering A Block Device As Capable Of Exchanging Integrity Metadata --------------------------------------------------------------------------- - - To enable integrity exchange on a block device the gendisk must be - registered as capable: - - `int blk_integrity_register(gendisk, blk_integrity);` - - The blk_integrity struct is a template and should contain the - following:: - - static struct blk_integrity my_profile = { - .name = "STANDARDSBODY-TYPE-VARIANT-CSUM", - .generate_fn = my_generate_fn, - .verify_fn = my_verify_fn, - .tuple_size = sizeof(struct my_tuple_size), - .tag_size = , - }; - - 'name' is a text string which will be visible in sysfs. This is - part of the userland API so chose it carefully and never change - it. The format is standards body-type-variant. - E.g. T10-DIF-TYPE1-IP or T13-EPP-0-CRC. - - 'generate_fn' generates appropriate integrity metadata (for WRITE). - - 'verify_fn' verifies that the data buffer matches the integrity - metadata. - - 'tuple_size' must be set to match the size of the integrity - metadata per sector. I.e. 8 for DIF and EPP. - - 'tag_size' must be set to identify how many bytes of tag space - are available per hardware sector. For DIF this is either 2 or - 0 depending on the value of the Control Mode Page ATO bit. - ---------------------------------------------------------------------- 2007-12-24 Martin K. Petersen diff --git a/block/blk-integrity.c b/block/blk-integrity.c index b37b8855eed147..05a48689a424b2 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -107,63 +107,6 @@ int blk_rq_map_integrity_sg(struct request_queue *q, struct bio *bio, } EXPORT_SYMBOL(blk_rq_map_integrity_sg); -/** - * blk_integrity_compare - Compare integrity profile of two disks - * @gd1: Disk to compare - * @gd2: Disk to compare - * - * Description: Meta-devices like DM and MD need to verify that all - * sub-devices use the same integrity format before advertising to - * upper layers that they can send/receive integrity metadata. This - * function can be used to check whether two gendisk devices have - * compatible integrity formats. - */ -int blk_integrity_compare(struct gendisk *gd1, struct gendisk *gd2) -{ - struct blk_integrity *b1 = &gd1->queue->integrity; - struct blk_integrity *b2 = &gd2->queue->integrity; - - if (!b1->tuple_size && !b2->tuple_size) - return 0; - - if (!b1->tuple_size || !b2->tuple_size) - return -1; - - if (b1->interval_exp != b2->interval_exp) { - pr_err("%s: %s/%s protection interval %u != %u\n", - __func__, gd1->disk_name, gd2->disk_name, - 1 << b1->interval_exp, 1 << b2->interval_exp); - return -1; - } - - if (b1->tuple_size != b2->tuple_size) { - pr_err("%s: %s/%s tuple sz %u != %u\n", __func__, - gd1->disk_name, gd2->disk_name, - b1->tuple_size, b2->tuple_size); - return -1; - } - - if (b1->tag_size && b2->tag_size && (b1->tag_size != b2->tag_size)) { - pr_err("%s: %s/%s tag sz %u != %u\n", __func__, - gd1->disk_name, gd2->disk_name, - b1->tag_size, b2->tag_size); - return -1; - } - - if (b1->csum_type != b2->csum_type || - (b1->flags & BLK_INTEGRITY_REF_TAG) != - (b2->flags & BLK_INTEGRITY_REF_TAG)) { - pr_err("%s: %s/%s type %s != %s\n", __func__, - gd1->disk_name, gd2->disk_name, - blk_integrity_profile_name(b1), - blk_integrity_profile_name(b2)); - return -1; - } - - return 0; -} -EXPORT_SYMBOL(blk_integrity_compare); - bool blk_integrity_merge_rq(struct request_queue *q, struct request *req, struct request *next) { @@ -217,7 +160,7 @@ bool blk_integrity_merge_bio(struct request_queue *q, struct request *req, static inline struct blk_integrity *dev_to_bi(struct device *dev) { - return &dev_to_disk(dev)->queue->integrity; + return &dev_to_disk(dev)->queue->limits.integrity; } const char *blk_integrity_profile_name(struct blk_integrity *bi) @@ -246,7 +189,8 @@ EXPORT_SYMBOL_GPL(blk_integrity_profile_name); static ssize_t flag_store(struct device *dev, struct device_attribute *attr, const char *page, size_t count, unsigned char flag) { - struct blk_integrity *bi = dev_to_bi(dev); + struct request_queue *q = dev_to_disk(dev)->queue; + struct queue_limits lim; unsigned long val; int err; @@ -254,11 +198,18 @@ static ssize_t flag_store(struct device *dev, struct device_attribute *attr, if (err) return err; - /* the flags are inverted vs the values in the sysfs files */ + /* note that the flags are inverted vs the values in the sysfs files */ + lim = queue_limits_start_update(q); if (val) - bi->flags &= ~flag; + lim.integrity.flags &= ~flag; else - bi->flags |= flag; + lim.integrity.flags |= flag; + + blk_mq_freeze_queue(q); + err = queue_limits_commit_update(q, &lim); + blk_mq_unfreeze_queue(q); + if (err) + return err; return count; } @@ -355,52 +306,3 @@ const struct attribute_group blk_integrity_attr_group = { .name = "integrity", .attrs = integrity_attrs, }; - -/** - * blk_integrity_register - Register a gendisk as being integrity-capable - * @disk: struct gendisk pointer to make integrity-aware - * @template: block integrity profile to register - * - * Description: When a device needs to advertise itself as being able to - * send/receive integrity metadata it must use this function to register - * the capability with the block layer. The template is a blk_integrity - * struct with values appropriate for the underlying hardware. See - * Documentation/block/data-integrity.rst. - */ -void blk_integrity_register(struct gendisk *disk, struct blk_integrity *template) -{ - struct blk_integrity *bi = &disk->queue->integrity; - - bi->csum_type = template->csum_type; - bi->flags = template->flags; - bi->interval_exp = template->interval_exp ? : - ilog2(queue_logical_block_size(disk->queue)); - bi->tuple_size = template->tuple_size; - bi->tag_size = template->tag_size; - bi->pi_offset = template->pi_offset; - -#ifdef CONFIG_BLK_INLINE_ENCRYPTION - if (disk->queue->crypto_profile) { - pr_warn("blk-integrity: Integrity and hardware inline encryption are not supported together. Disabling hardware inline encryption.\n"); - disk->queue->crypto_profile = NULL; - } -#endif -} -EXPORT_SYMBOL(blk_integrity_register); - -/** - * blk_integrity_unregister - Unregister block integrity profile - * @disk: disk whose integrity profile to unregister - * - * Description: This function unregisters the integrity capability from - * a block device. - */ -void blk_integrity_unregister(struct gendisk *disk) -{ - struct blk_integrity *bi = &disk->queue->integrity; - - if (!bi->tuple_size) - return; - memset(bi, 0, sizeof(*bi)); -} -EXPORT_SYMBOL(blk_integrity_unregister); diff --git a/block/blk-settings.c b/block/blk-settings.c index 996f247fc98e80..f11c8676eb4c67 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -6,7 +6,7 @@ #include #include #include -#include +#include #include #include #include @@ -97,6 +97,36 @@ static int blk_validate_zoned_limits(struct queue_limits *lim) return 0; } +static int blk_validate_integrity_limits(struct queue_limits *lim) +{ + struct blk_integrity *bi = &lim->integrity; + + if (!bi->tuple_size) { + if (bi->csum_type != BLK_INTEGRITY_CSUM_NONE || + bi->tag_size || ((bi->flags & BLK_INTEGRITY_REF_TAG))) { + pr_warn("invalid PI settings.\n"); + return -EINVAL; + } + return 0; + } + + if (!IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)) { + pr_warn("integrity support disabled.\n"); + return -EINVAL; + } + + if (bi->csum_type == BLK_INTEGRITY_CSUM_NONE && + (bi->flags & BLK_INTEGRITY_REF_TAG)) { + pr_warn("ref tag not support without checksum.\n"); + return -EINVAL; + } + + if (!bi->interval_exp) + bi->interval_exp = ilog2(lim->logical_block_size); + + return 0; +} + /* * Check that the limits in lim are valid, initialize defaults for unset * values, and cap values based on others where needed. @@ -105,6 +135,7 @@ static int blk_validate_limits(struct queue_limits *lim) { unsigned int max_hw_sectors; unsigned int logical_block_sectors; + int err; /* * Unless otherwise specified, default to 512 byte logical blocks and a @@ -230,6 +261,9 @@ static int blk_validate_limits(struct queue_limits *lim) lim->misaligned = 0; } + err = blk_validate_integrity_limits(lim); + if (err) + return err; return blk_validate_zoned_limits(lim); } @@ -263,13 +297,24 @@ int queue_limits_commit_update(struct request_queue *q, struct queue_limits *lim) __releases(q->limits_lock) { - int error = blk_validate_limits(lim); + int error; - if (!error) { - q->limits = *lim; - if (q->disk) - blk_apply_bdi_limits(q->disk->bdi, lim); + error = blk_validate_limits(lim); + if (error) + goto out_unlock; + +#ifdef CONFIG_BLK_INLINE_ENCRYPTION + if (q->crypto_profile && lim->integrity.tag_size) { + pr_warn("blk-integrity: Integrity and hardware inline encryption are not supported together.\n"); + error = -EINVAL; + goto out_unlock; } +#endif + + q->limits = *lim; + if (q->disk) + blk_apply_bdi_limits(q->disk->bdi, lim); +out_unlock: mutex_unlock(&q->limits_lock); return error; } @@ -575,6 +620,67 @@ void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev, } EXPORT_SYMBOL_GPL(queue_limits_stack_bdev); +/** + * queue_limits_stack_integrity - stack integrity profile + * @t: target queue limits + * @b: base queue limits + * + * Check if the integrity profile in the @b can be stacked into the + * target @t. Stacking is possible if either: + * + * a) does not have any integrity information stacked into it yet + * b) the integrity profile in @b is identical to the one in @t + * + * If @b can be stacked into @t, return %true. Else return %false and clear the + * integrity information in @t. + */ +bool queue_limits_stack_integrity(struct queue_limits *t, + struct queue_limits *b) +{ + struct blk_integrity *ti = &t->integrity; + struct blk_integrity *bi = &b->integrity; + + if (!IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)) + return true; + + if (!ti->tuple_size) { + /* inherit the settings from the first underlying device */ + if (!(ti->flags & BLK_INTEGRITY_STACKED)) { + ti->flags = BLK_INTEGRITY_DEVICE_CAPABLE | + (bi->flags & BLK_INTEGRITY_REF_TAG); + ti->csum_type = bi->csum_type; + ti->tuple_size = bi->tuple_size; + ti->pi_offset = bi->pi_offset; + ti->interval_exp = bi->interval_exp; + ti->tag_size = bi->tag_size; + goto done; + } + if (!bi->tuple_size) + goto done; + } + + if (ti->tuple_size != bi->tuple_size) + goto incompatible; + if (ti->interval_exp != bi->interval_exp) + goto incompatible; + if (ti->tag_size != bi->tag_size) + goto incompatible; + if (ti->csum_type != bi->csum_type) + goto incompatible; + if ((ti->flags & BLK_INTEGRITY_REF_TAG) != + (bi->flags & BLK_INTEGRITY_REF_TAG)) + goto incompatible; + +done: + ti->flags |= BLK_INTEGRITY_STACKED; + return true; + +incompatible: + memset(ti, 0, sizeof(*ti)); + return false; +} +EXPORT_SYMBOL_GPL(queue_limits_stack_integrity); + /** * blk_queue_update_dma_pad - update pad mask * @q: the request queue for the device diff --git a/block/t10-pi.c b/block/t10-pi.c index dadecf621497bb..cd7fa60d63ff21 100644 --- a/block/t10-pi.c +++ b/block/t10-pi.c @@ -116,7 +116,7 @@ static blk_status_t t10_pi_verify(struct blk_integrity_iter *iter, */ static void t10_pi_type1_prepare(struct request *rq) { - struct blk_integrity *bi = &rq->q->integrity; + struct blk_integrity *bi = &rq->q->limits.integrity; const int tuple_sz = bi->tuple_size; u32 ref_tag = t10_pi_ref_tag(rq); u8 offset = bi->pi_offset; @@ -167,7 +167,7 @@ static void t10_pi_type1_prepare(struct request *rq) */ static void t10_pi_type1_complete(struct request *rq, unsigned int nr_bytes) { - struct blk_integrity *bi = &rq->q->integrity; + struct blk_integrity *bi = &rq->q->limits.integrity; unsigned intervals = nr_bytes >> bi->interval_exp; const int tuple_sz = bi->tuple_size; u32 ref_tag = t10_pi_ref_tag(rq); @@ -290,7 +290,7 @@ static blk_status_t ext_pi_crc64_verify(struct blk_integrity_iter *iter, static void ext_pi_type1_prepare(struct request *rq) { - struct blk_integrity *bi = &rq->q->integrity; + struct blk_integrity *bi = &rq->q->limits.integrity; const int tuple_sz = bi->tuple_size; u64 ref_tag = ext_pi_ref_tag(rq); u8 offset = bi->pi_offset; @@ -330,7 +330,7 @@ static void ext_pi_type1_prepare(struct request *rq) static void ext_pi_type1_complete(struct request *rq, unsigned int nr_bytes) { - struct blk_integrity *bi = &rq->q->integrity; + struct blk_integrity *bi = &rq->q->limits.integrity; unsigned intervals = nr_bytes >> bi->interval_exp; const int tuple_sz = bi->tuple_size; u64 ref_tag = ext_pi_ref_tag(rq); @@ -396,7 +396,7 @@ blk_status_t blk_integrity_verify(struct blk_integrity_iter *iter, void blk_integrity_prepare(struct request *rq) { - struct blk_integrity *bi = &rq->q->integrity; + struct blk_integrity *bi = &rq->q->limits.integrity; if (!(bi->flags & BLK_INTEGRITY_REF_TAG)) return; @@ -409,7 +409,7 @@ void blk_integrity_prepare(struct request *rq) void blk_integrity_complete(struct request *rq, unsigned int nr_bytes) { - struct blk_integrity *bi = &rq->q->integrity; + struct blk_integrity *bi = &rq->q->limits.integrity; if (!(bi->flags & BLK_INTEGRITY_REF_TAG)) return; diff --git a/drivers/md/dm-core.h b/drivers/md/dm-core.h index 08700bfc3e2343..14a44c0f82868b 100644 --- a/drivers/md/dm-core.h +++ b/drivers/md/dm-core.h @@ -206,7 +206,6 @@ struct dm_table { bool integrity_supported:1; bool singleton:1; - unsigned integrity_added:1; /* * Indicates the rw permissions for the new logical device. This diff --git a/drivers/md/dm-integrity.c b/drivers/md/dm-integrity.c index c1cc27541673c7..2a89f8eb4713c9 100644 --- a/drivers/md/dm-integrity.c +++ b/drivers/md/dm-integrity.c @@ -3475,6 +3475,17 @@ static void dm_integrity_io_hints(struct dm_target *ti, struct queue_limits *lim limits->dma_alignment = limits->logical_block_size - 1; limits->discard_granularity = ic->sectors_per_block << SECTOR_SHIFT; } + + if (!ic->internal_hash) { + struct blk_integrity *bi = &limits->integrity; + + memset(bi, 0, sizeof(*bi)); + bi->tuple_size = ic->tag_size; + bi->tag_size = bi->tuple_size; + bi->interval_exp = + ic->sb->log2_sectors_per_block + SECTOR_SHIFT; + } + limits->max_integrity_segments = USHRT_MAX; } @@ -3631,19 +3642,6 @@ static int initialize_superblock(struct dm_integrity_c *ic, return 0; } -static void dm_integrity_set(struct dm_target *ti, struct dm_integrity_c *ic) -{ - struct gendisk *disk = dm_disk(dm_table_get_md(ti->table)); - struct blk_integrity bi; - - memset(&bi, 0, sizeof(bi)); - bi.tuple_size = ic->tag_size; - bi.tag_size = bi.tuple_size; - bi.interval_exp = ic->sb->log2_sectors_per_block + SECTOR_SHIFT; - - blk_integrity_register(disk, &bi); -} - static void dm_integrity_free_page_list(struct page_list *pl) { unsigned int i; @@ -4629,9 +4627,6 @@ static int dm_integrity_ctr(struct dm_target *ti, unsigned int argc, char **argv } } - if (!ic->internal_hash) - dm_integrity_set(ti, ic); - ti->num_flush_bios = 1; ti->flush_supported = true; if (ic->discard) diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index b2d5246cff2102..fd789eeb62d943 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -425,6 +425,13 @@ static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev, q->limits.logical_block_size, q->limits.alignment_offset, (unsigned long long) start << SECTOR_SHIFT); + + /* + * Only stack the integrity profile if the target doesn't have native + * integrity support. + */ + if (!dm_target_has_integrity(ti->type)) + queue_limits_stack_integrity_bdev(limits, bdev); return 0; } @@ -702,9 +709,6 @@ int dm_table_add_target(struct dm_table *t, const char *type, t->immutable_target_type = ti->type; } - if (dm_target_has_integrity(ti->type)) - t->integrity_added = 1; - ti->table = t; ti->begin = start; ti->len = len; @@ -1119,99 +1123,6 @@ static int dm_table_build_index(struct dm_table *t) return r; } -static bool integrity_profile_exists(struct gendisk *disk) -{ - return !!blk_get_integrity(disk); -} - -/* - * Get a disk whose integrity profile reflects the table's profile. - * Returns NULL if integrity support was inconsistent or unavailable. - */ -static struct gendisk *dm_table_get_integrity_disk(struct dm_table *t) -{ - struct list_head *devices = dm_table_get_devices(t); - struct dm_dev_internal *dd = NULL; - struct gendisk *prev_disk = NULL, *template_disk = NULL; - - for (unsigned int i = 0; i < t->num_targets; i++) { - struct dm_target *ti = dm_table_get_target(t, i); - - if (!dm_target_passes_integrity(ti->type)) - goto no_integrity; - } - - list_for_each_entry(dd, devices, list) { - template_disk = dd->dm_dev->bdev->bd_disk; - if (!integrity_profile_exists(template_disk)) - goto no_integrity; - else if (prev_disk && - blk_integrity_compare(prev_disk, template_disk) < 0) - goto no_integrity; - prev_disk = template_disk; - } - - return template_disk; - -no_integrity: - if (prev_disk) - DMWARN("%s: integrity not set: %s and %s profile mismatch", - dm_device_name(t->md), - prev_disk->disk_name, - template_disk->disk_name); - return NULL; -} - -/* - * Register the mapped device for blk_integrity support if the - * underlying devices have an integrity profile. But all devices may - * not have matching profiles (checking all devices isn't reliable - * during table load because this table may use other DM device(s) which - * must be resumed before they will have an initialized integity - * profile). Consequently, stacked DM devices force a 2 stage integrity - * profile validation: First pass during table load, final pass during - * resume. - */ -static int dm_table_register_integrity(struct dm_table *t) -{ - struct mapped_device *md = t->md; - struct gendisk *template_disk = NULL; - - /* If target handles integrity itself do not register it here. */ - if (t->integrity_added) - return 0; - - template_disk = dm_table_get_integrity_disk(t); - if (!template_disk) - return 0; - - if (!integrity_profile_exists(dm_disk(md))) { - t->integrity_supported = true; - /* - * Register integrity profile during table load; we can do - * this because the final profile must match during resume. - */ - blk_integrity_register(dm_disk(md), - blk_get_integrity(template_disk)); - return 0; - } - - /* - * If DM device already has an initialized integrity - * profile the new profile should not conflict. - */ - if (blk_integrity_compare(dm_disk(md), template_disk) < 0) { - DMERR("%s: conflict with existing integrity profile: %s profile mismatch", - dm_device_name(t->md), - template_disk->disk_name); - return 1; - } - - /* Preserve existing integrity profile */ - t->integrity_supported = true; - return 0; -} - #ifdef CONFIG_BLK_INLINE_ENCRYPTION struct dm_crypto_profile { @@ -1423,12 +1334,6 @@ int dm_table_complete(struct dm_table *t) return r; } - r = dm_table_register_integrity(t); - if (r) { - DMERR("could not register integrity profile."); - return r; - } - r = dm_table_construct_crypto_profile(t); if (r) { DMERR("could not construct crypto profile."); @@ -1688,6 +1593,14 @@ int dm_calculate_queue_limits(struct dm_table *t, blk_set_stacking_limits(limits); + t->integrity_supported = true; + for (unsigned int i = 0; i < t->num_targets; i++) { + struct dm_target *ti = dm_table_get_target(t, i); + + if (!dm_target_passes_integrity(ti->type)) + t->integrity_supported = false; + } + for (unsigned int i = 0; i < t->num_targets; i++) { struct dm_target *ti = dm_table_get_target(t, i); @@ -1738,6 +1651,18 @@ int dm_calculate_queue_limits(struct dm_table *t, dm_device_name(t->md), (unsigned long long) ti->begin, (unsigned long long) ti->len); + + if (t->integrity_supported || + dm_target_has_integrity(ti->type)) { + if (!queue_limits_stack_integrity(limits, &ti_limits)) { + DMWARN("%s: adding target device (start sect %llu len %llu) " + "disabled integrity support due to incompatibility", + dm_device_name(t->md), + (unsigned long long) ti->begin, + (unsigned long long) ti->len); + t->integrity_supported = false; + } + } } /* @@ -1761,36 +1686,6 @@ int dm_calculate_queue_limits(struct dm_table *t, return validate_hardware_logical_block_alignment(t, limits); } -/* - * Verify that all devices have an integrity profile that matches the - * DM device's registered integrity profile. If the profiles don't - * match then unregister the DM device's integrity profile. - */ -static void dm_table_verify_integrity(struct dm_table *t) -{ - struct gendisk *template_disk = NULL; - - if (t->integrity_added) - return; - - if (t->integrity_supported) { - /* - * Verify that the original integrity profile - * matches all the devices in this table. - */ - template_disk = dm_table_get_integrity_disk(t); - if (template_disk && - blk_integrity_compare(dm_disk(t->md), template_disk) >= 0) - return; - } - - if (integrity_profile_exists(dm_disk(t->md))) { - DMWARN("%s: unable to establish an integrity profile", - dm_device_name(t->md)); - blk_integrity_unregister(dm_disk(t->md)); - } -} - static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev, sector_t start, sector_t len, void *data) { @@ -2004,8 +1899,6 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, else blk_queue_flag_set(QUEUE_FLAG_NONROT, q); - dm_table_verify_integrity(t); - /* * Some devices don't use blk_integrity but still want stable pages * because they do their own checksumming. diff --git a/drivers/md/md.c b/drivers/md/md.c index aff9118ff69750..67ece2cd725f50 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -2410,36 +2410,10 @@ static LIST_HEAD(pending_raid_disks); */ int md_integrity_register(struct mddev *mddev) { - struct md_rdev *rdev, *reference = NULL; - if (list_empty(&mddev->disks)) return 0; /* nothing to do */ - if (mddev_is_dm(mddev) || blk_get_integrity(mddev->gendisk)) - return 0; /* shouldn't register, or already is */ - rdev_for_each(rdev, mddev) { - /* skip spares and non-functional disks */ - if (test_bit(Faulty, &rdev->flags)) - continue; - if (rdev->raid_disk < 0) - continue; - if (!reference) { - /* Use the first rdev as the reference */ - reference = rdev; - continue; - } - /* does this rdev's profile match the reference profile? */ - if (blk_integrity_compare(reference->bdev->bd_disk, - rdev->bdev->bd_disk) < 0) - return -EINVAL; - } - if (!reference || !bdev_get_integrity(reference->bdev)) - return 0; - /* - * All component devices are integrity capable and have matching - * profiles, register the common profile for the md device. - */ - blk_integrity_register(mddev->gendisk, - bdev_get_integrity(reference->bdev)); + if (mddev_is_dm(mddev) || !blk_get_integrity(mddev->gendisk)) + return 0; /* shouldn't register */ pr_debug("md: data integrity enabled on %s\n", mdname(mddev)); if (bioset_integrity_create(&mddev->bio_set, BIO_POOL_SIZE) || @@ -2459,32 +2433,6 @@ int md_integrity_register(struct mddev *mddev) } EXPORT_SYMBOL(md_integrity_register); -/* - * Attempt to add an rdev, but only if it is consistent with the current - * integrity profile - */ -int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev) -{ - struct blk_integrity *bi_mddev; - - if (mddev_is_dm(mddev)) - return 0; - - bi_mddev = blk_get_integrity(mddev->gendisk); - - if (!bi_mddev) /* nothing to do */ - return 0; - - if (blk_integrity_compare(mddev->gendisk, rdev->bdev->bd_disk) != 0) { - pr_err("%s: incompatible integrity profile for %pg\n", - mdname(mddev), rdev->bdev); - return -ENXIO; - } - - return 0; -} -EXPORT_SYMBOL(md_integrity_add_rdev); - static bool rdev_read_only(struct md_rdev *rdev) { return bdev_read_only(rdev->bdev) || @@ -5755,14 +5703,20 @@ static const struct kobj_type md_ktype = { int mdp_major = 0; /* stack the limit for all rdevs into lim */ -void mddev_stack_rdev_limits(struct mddev *mddev, struct queue_limits *lim) +int mddev_stack_rdev_limits(struct mddev *mddev, struct queue_limits *lim, + unsigned int flags) { struct md_rdev *rdev; rdev_for_each(rdev, mddev) { queue_limits_stack_bdev(lim, rdev->bdev, rdev->data_offset, mddev->gendisk->disk_name); + if ((flags & MDDEV_STACK_INTEGRITY) && + !queue_limits_stack_integrity_bdev(lim, rdev->bdev)) + return -EINVAL; } + + return 0; } EXPORT_SYMBOL_GPL(mddev_stack_rdev_limits); @@ -5777,6 +5731,14 @@ int mddev_stack_new_rdev(struct mddev *mddev, struct md_rdev *rdev) lim = queue_limits_start_update(mddev->gendisk->queue); queue_limits_stack_bdev(&lim, rdev->bdev, rdev->data_offset, mddev->gendisk->disk_name); + + if (!queue_limits_stack_integrity_bdev(&lim, rdev->bdev)) { + pr_err("%s: incompatible integrity profile for %pg\n", + mdname(mddev), rdev->bdev); + queue_limits_cancel_update(mddev->gendisk->queue); + return -ENXIO; + } + return queue_limits_commit_update(mddev->gendisk->queue, &lim); } EXPORT_SYMBOL_GPL(mddev_stack_new_rdev); diff --git a/drivers/md/md.h b/drivers/md/md.h index ca085ecad50449..6733b0b0abf999 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -809,7 +809,6 @@ extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev); extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors); extern int md_check_no_bitmap(struct mddev *mddev); extern int md_integrity_register(struct mddev *mddev); -extern int md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev); extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); extern int mddev_init(struct mddev *mddev); @@ -908,7 +907,9 @@ void md_autostart_arrays(int part); int md_set_array_info(struct mddev *mddev, struct mdu_array_info_s *info); int md_add_new_disk(struct mddev *mddev, struct mdu_disk_info_s *info); int do_md_run(struct mddev *mddev); -void mddev_stack_rdev_limits(struct mddev *mddev, struct queue_limits *lim); +#define MDDEV_STACK_INTEGRITY (1u << 0) +int mddev_stack_rdev_limits(struct mddev *mddev, struct queue_limits *lim, + unsigned int flags); int mddev_stack_new_rdev(struct mddev *mddev, struct md_rdev *rdev); void mddev_update_io_opt(struct mddev *mddev, unsigned int nr_stripes); diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c index 81c01347cd24e6..62634e2a33bd0f 100644 --- a/drivers/md/raid0.c +++ b/drivers/md/raid0.c @@ -377,13 +377,18 @@ static void raid0_free(struct mddev *mddev, void *priv) static int raid0_set_limits(struct mddev *mddev) { struct queue_limits lim; + int err; blk_set_stacking_limits(&lim); lim.max_hw_sectors = mddev->chunk_sectors; lim.max_write_zeroes_sectors = mddev->chunk_sectors; lim.io_min = mddev->chunk_sectors << 9; lim.io_opt = lim.io_min * mddev->raid_disks; - mddev_stack_rdev_limits(mddev, &lim); + err = mddev_stack_rdev_limits(mddev, &lim, MDDEV_STACK_INTEGRITY); + if (err) { + queue_limits_cancel_update(mddev->gendisk->queue); + return err; + } return queue_limits_set(mddev->gendisk->queue, &lim); } diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 1f321826ef02ba..779cad62f6f8c0 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -1907,9 +1907,6 @@ static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev) if (mddev->recovery_disabled == conf->recovery_disabled) return -EBUSY; - if (md_integrity_add_rdev(rdev, mddev)) - return -ENXIO; - if (rdev->raid_disk >= 0) first = last = rdev->raid_disk; @@ -3197,10 +3194,15 @@ static struct r1conf *setup_conf(struct mddev *mddev) static int raid1_set_limits(struct mddev *mddev) { struct queue_limits lim; + int err; blk_set_stacking_limits(&lim); lim.max_write_zeroes_sectors = 0; - mddev_stack_rdev_limits(mddev, &lim); + err = mddev_stack_rdev_limits(mddev, &lim, MDDEV_STACK_INTEGRITY); + if (err) { + queue_limits_cancel_update(mddev->gendisk->queue); + return err; + } return queue_limits_set(mddev->gendisk->queue, &lim); } diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index a4556d2e46bf95..5f6885b53b691a 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -2083,9 +2083,6 @@ static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev) if (rdev->saved_raid_disk < 0 && !_enough(conf, 1, -1)) return -EINVAL; - if (md_integrity_add_rdev(rdev, mddev)) - return -ENXIO; - if (rdev->raid_disk >= 0) first = last = rdev->raid_disk; @@ -3980,12 +3977,17 @@ static int raid10_set_queue_limits(struct mddev *mddev) { struct r10conf *conf = mddev->private; struct queue_limits lim; + int err; blk_set_stacking_limits(&lim); lim.max_write_zeroes_sectors = 0; lim.io_min = mddev->chunk_sectors << 9; lim.io_opt = lim.io_min * raid10_nr_stripes(conf); - mddev_stack_rdev_limits(mddev, &lim); + err = mddev_stack_rdev_limits(mddev, &lim, MDDEV_STACK_INTEGRITY); + if (err) { + queue_limits_cancel_update(mddev->gendisk->queue); + return err; + } return queue_limits_set(mddev->gendisk->queue, &lim); } diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 2bd1ce9b39226a..675c68fa6c6403 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -7708,7 +7708,7 @@ static int raid5_set_limits(struct mddev *mddev) lim.raid_partial_stripes_expensive = 1; lim.discard_granularity = stripe; lim.max_write_zeroes_sectors = 0; - mddev_stack_rdev_limits(mddev, &lim); + mddev_stack_rdev_limits(mddev, &lim, 0); rdev_for_each(rdev, mddev) queue_limits_stack_bdev(&lim, rdev->bdev, rdev->new_data_offset, mddev->gendisk->disk_name); diff --git a/drivers/nvdimm/btt.c b/drivers/nvdimm/btt.c index 1e5aedaf8c7bd9..c5f8451b494d6c 100644 --- a/drivers/nvdimm/btt.c +++ b/drivers/nvdimm/btt.c @@ -1504,6 +1504,11 @@ static int btt_blk_init(struct btt *btt) }; int rc; + if (btt_meta_size(btt) && IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)) { + lim.integrity.tuple_size = btt_meta_size(btt); + lim.integrity.tag_size = btt_meta_size(btt); + } + btt->btt_disk = blk_alloc_disk(&lim, NUMA_NO_NODE); if (IS_ERR(btt->btt_disk)) return PTR_ERR(btt->btt_disk); @@ -1516,14 +1521,6 @@ static int btt_blk_init(struct btt *btt) blk_queue_flag_set(QUEUE_FLAG_NONROT, btt->btt_disk->queue); blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, btt->btt_disk->queue); - if (btt_meta_size(btt) && IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)) { - struct blk_integrity bi = { - .tuple_size = btt_meta_size(btt), - .tag_size = btt_meta_size(btt), - }; - blk_integrity_register(btt->btt_disk, &bi); - } - set_capacity(btt->btt_disk, btt->nlba * btt->sector_size >> 9); rc = device_add_disk(&btt->nd_btt->dev, btt->btt_disk, NULL); if (rc) diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 14bac248cde4ca..5a673fa5cb2612 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -1723,11 +1723,12 @@ int nvme_getgeo(struct block_device *bdev, struct hd_geometry *geo) return 0; } -static bool nvme_init_integrity(struct gendisk *disk, struct nvme_ns_head *head) +static bool nvme_init_integrity(struct gendisk *disk, struct nvme_ns_head *head, + struct queue_limits *lim) { - struct blk_integrity integrity = { }; + struct blk_integrity *bi = &lim->integrity; - blk_integrity_unregister(disk); + memset(bi, 0, sizeof(*bi)); if (!head->ms) return true; @@ -1744,14 +1745,14 @@ static bool nvme_init_integrity(struct gendisk *disk, struct nvme_ns_head *head) case NVME_NS_DPS_PI_TYPE3: switch (head->guard_type) { case NVME_NVM_NS_16B_GUARD: - integrity.csum_type = BLK_INTEGRITY_CSUM_CRC; - integrity.tag_size = sizeof(u16) + sizeof(u32); - integrity.flags |= BLK_INTEGRITY_DEVICE_CAPABLE; + bi->csum_type = BLK_INTEGRITY_CSUM_CRC; + bi->tag_size = sizeof(u16) + sizeof(u32); + bi->flags |= BLK_INTEGRITY_DEVICE_CAPABLE; break; case NVME_NVM_NS_64B_GUARD: - integrity.csum_type = BLK_INTEGRITY_CSUM_CRC64; - integrity.tag_size = sizeof(u16) + 6; - integrity.flags |= BLK_INTEGRITY_DEVICE_CAPABLE; + bi->csum_type = BLK_INTEGRITY_CSUM_CRC64; + bi->tag_size = sizeof(u16) + 6; + bi->flags |= BLK_INTEGRITY_DEVICE_CAPABLE; break; default: break; @@ -1761,16 +1762,16 @@ static bool nvme_init_integrity(struct gendisk *disk, struct nvme_ns_head *head) case NVME_NS_DPS_PI_TYPE2: switch (head->guard_type) { case NVME_NVM_NS_16B_GUARD: - integrity.csum_type = BLK_INTEGRITY_CSUM_CRC; - integrity.tag_size = sizeof(u16); - integrity.flags |= BLK_INTEGRITY_DEVICE_CAPABLE | - BLK_INTEGRITY_REF_TAG; + bi->csum_type = BLK_INTEGRITY_CSUM_CRC; + bi->tag_size = sizeof(u16); + bi->flags |= BLK_INTEGRITY_DEVICE_CAPABLE | + BLK_INTEGRITY_REF_TAG; break; case NVME_NVM_NS_64B_GUARD: - integrity.csum_type = BLK_INTEGRITY_CSUM_CRC64; - integrity.tag_size = sizeof(u16); - integrity.flags |= BLK_INTEGRITY_DEVICE_CAPABLE | - BLK_INTEGRITY_REF_TAG; + bi->csum_type = BLK_INTEGRITY_CSUM_CRC64; + bi->tag_size = sizeof(u16); + bi->flags |= BLK_INTEGRITY_DEVICE_CAPABLE | + BLK_INTEGRITY_REF_TAG; break; default: break; @@ -1780,9 +1781,8 @@ static bool nvme_init_integrity(struct gendisk *disk, struct nvme_ns_head *head) break; } - integrity.tuple_size = head->ms; - integrity.pi_offset = head->pi_offset; - blk_integrity_register(disk, &integrity); + bi->tuple_size = head->ms; + bi->pi_offset = head->pi_offset; return true; } @@ -2105,11 +2105,6 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns, if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) && ns->head->ids.csi == NVME_CSI_ZNS) nvme_update_zone_info(ns, &lim, &zi); - ret = queue_limits_commit_update(ns->disk->queue, &lim); - if (ret) { - blk_mq_unfreeze_queue(ns->disk->queue); - goto out; - } /* * Register a metadata profile for PI, or the plain non-integrity NVMe @@ -2117,9 +2112,15 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns, * I/O to namespaces with metadata except when the namespace supports * PI, as it can strip/insert in that case. */ - if (!nvme_init_integrity(ns->disk, ns->head)) + if (!nvme_init_integrity(ns->disk, ns->head, &lim)) capacity = 0; + ret = queue_limits_commit_update(ns->disk->queue, &lim); + if (ret) { + blk_mq_unfreeze_queue(ns->disk->queue); + goto out; + } + set_capacity_and_notify(ns->disk, capacity); /* @@ -2191,14 +2192,6 @@ static int nvme_update_ns_info(struct nvme_ns *ns, struct nvme_ns_info *info) struct queue_limits lim; blk_mq_freeze_queue(ns->head->disk->queue); - if (unsupported) - ns->head->disk->flags |= GENHD_FL_HIDDEN; - else - nvme_init_integrity(ns->head->disk, ns->head); - set_capacity_and_notify(ns->head->disk, get_capacity(ns->disk)); - set_disk_ro(ns->head->disk, nvme_ns_is_readonly(ns, info)); - nvme_mpath_revalidate_paths(ns); - /* * queue_limits mixes values that are the hardware limitations * for bio splitting with what is the device configuration. @@ -2221,7 +2214,16 @@ static int nvme_update_ns_info(struct nvme_ns *ns, struct nvme_ns_info *info) lim.io_opt = ns_lim->io_opt; queue_limits_stack_bdev(&lim, ns->disk->part0, 0, ns->head->disk->disk_name); + if (unsupported) + ns->head->disk->flags |= GENHD_FL_HIDDEN; + else + nvme_init_integrity(ns->head->disk, ns->head, &lim); ret = queue_limits_commit_update(ns->head->disk->queue, &lim); + + set_capacity_and_notify(ns->head->disk, get_capacity(ns->disk)); + set_disk_ro(ns->head->disk, nvme_ns_is_readonly(ns, info)); + nvme_mpath_revalidate_paths(ns); + blk_mq_unfreeze_queue(ns->head->disk->queue); } diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index d957e29b17a98a..e01393ed42076b 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -2482,11 +2482,13 @@ static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer return 0; } -static void sd_config_protection(struct scsi_disk *sdkp) +static void sd_config_protection(struct scsi_disk *sdkp, + struct queue_limits *lim) { struct scsi_device *sdp = sdkp->device; - sd_dif_config_host(sdkp); + if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY)) + sd_dif_config_host(sdkp, lim); if (!sdkp->protection_type) return; @@ -3677,7 +3679,7 @@ static int sd_revalidate_disk(struct gendisk *disk) sd_read_app_tag_own(sdkp, buffer); sd_read_write_same(sdkp, buffer); sd_read_security(sdkp, buffer); - sd_config_protection(sdkp); + sd_config_protection(sdkp, &lim); } /* diff --git a/drivers/scsi/sd.h b/drivers/scsi/sd.h index b4170b17bad47a..726f1613f6cb56 100644 --- a/drivers/scsi/sd.h +++ b/drivers/scsi/sd.h @@ -220,17 +220,7 @@ static inline sector_t sectors_to_logical(struct scsi_device *sdev, sector_t sec return sector >> (ilog2(sdev->sector_size) - 9); } -#ifdef CONFIG_BLK_DEV_INTEGRITY - -extern void sd_dif_config_host(struct scsi_disk *); - -#else /* CONFIG_BLK_DEV_INTEGRITY */ - -static inline void sd_dif_config_host(struct scsi_disk *disk) -{ -} - -#endif /* CONFIG_BLK_DEV_INTEGRITY */ +void sd_dif_config_host(struct scsi_disk *sdkp, struct queue_limits *lim); static inline int sd_is_zoned(struct scsi_disk *sdkp) { diff --git a/drivers/scsi/sd_dif.c b/drivers/scsi/sd_dif.c index 6f0921c7db787b..ae6ce6f5d622d9 100644 --- a/drivers/scsi/sd_dif.c +++ b/drivers/scsi/sd_dif.c @@ -24,14 +24,15 @@ /* * Configure exchange of protection information between OS and HBA. */ -void sd_dif_config_host(struct scsi_disk *sdkp) +void sd_dif_config_host(struct scsi_disk *sdkp, struct queue_limits *lim) { struct scsi_device *sdp = sdkp->device; - struct gendisk *disk = sdkp->disk; u8 type = sdkp->protection_type; - struct blk_integrity bi; + struct blk_integrity *bi = &lim->integrity; int dif, dix; + memset(bi, 0, sizeof(*bi)); + dif = scsi_host_dif_capable(sdp->host, type); dix = scsi_host_dix_capable(sdp->host, type); @@ -39,40 +40,33 @@ void sd_dif_config_host(struct scsi_disk *sdkp) dif = 0; dix = 1; } - if (!dix) { - blk_integrity_unregister(disk); + if (!dix) return; - } - - memset(&bi, 0, sizeof(bi)); /* Enable DMA of protection information */ if (scsi_host_get_guard(sdkp->device->host) & SHOST_DIX_GUARD_IP) - bi.csum_type = BLK_INTEGRITY_CSUM_IP; + bi->csum_type = BLK_INTEGRITY_CSUM_IP; else - bi.csum_type = BLK_INTEGRITY_CSUM_CRC; + bi->csum_type = BLK_INTEGRITY_CSUM_CRC; if (type != T10_PI_TYPE3_PROTECTION) - bi.flags |= BLK_INTEGRITY_REF_TAG; + bi->flags |= BLK_INTEGRITY_REF_TAG; - bi.tuple_size = sizeof(struct t10_pi_tuple); + bi->tuple_size = sizeof(struct t10_pi_tuple); if (dif && type) { - bi.flags |= BLK_INTEGRITY_DEVICE_CAPABLE; + bi->flags |= BLK_INTEGRITY_DEVICE_CAPABLE; if (!sdkp->ATO) - goto out; + return; if (type == T10_PI_TYPE3_PROTECTION) - bi.tag_size = sizeof(u16) + sizeof(u32); + bi->tag_size = sizeof(u16) + sizeof(u32); else - bi.tag_size = sizeof(u16); + bi->tag_size = sizeof(u16); } sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIX %s, application tag size %u bytes\n", - blk_integrity_profile_name(&bi), bi.tag_size); -out: - blk_integrity_register(disk, &bi); + blk_integrity_profile_name(bi), bi->tag_size); } - diff --git a/include/linux/blk-integrity.h b/include/linux/blk-integrity.h index bafa01d4e7f95b..d201140d77a336 100644 --- a/include/linux/blk-integrity.h +++ b/include/linux/blk-integrity.h @@ -11,6 +11,7 @@ enum blk_integrity_flags { BLK_INTEGRITY_NOGENERATE = 1 << 1, BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, BLK_INTEGRITY_REF_TAG = 1 << 3, + BLK_INTEGRITY_STACKED = 1 << 4, }; struct blk_integrity_iter { @@ -23,11 +24,15 @@ struct blk_integrity_iter { }; const char *blk_integrity_profile_name(struct blk_integrity *bi); +bool queue_limits_stack_integrity(struct queue_limits *t, + struct queue_limits *b); +static inline bool queue_limits_stack_integrity_bdev(struct queue_limits *t, + struct block_device *bdev) +{ + return queue_limits_stack_integrity(t, &bdev->bd_disk->queue->limits); +} #ifdef CONFIG_BLK_DEV_INTEGRITY -void blk_integrity_register(struct gendisk *, struct blk_integrity *); -void blk_integrity_unregister(struct gendisk *); -int blk_integrity_compare(struct gendisk *, struct gendisk *); int blk_rq_map_integrity_sg(struct request_queue *, struct bio *, struct scatterlist *); int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); @@ -35,14 +40,14 @@ int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); static inline bool blk_integrity_queue_supports_integrity(struct request_queue *q) { - return q->integrity.tuple_size; + return q->limits.integrity.tuple_size; } static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) { if (!blk_integrity_queue_supports_integrity(disk->queue)) return NULL; - return &disk->queue->integrity; + return &disk->queue->limits.integrity; } static inline struct blk_integrity * @@ -119,17 +124,6 @@ blk_integrity_queue_supports_integrity(struct request_queue *q) { return false; } -static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b) -{ - return 0; -} -static inline void blk_integrity_register(struct gendisk *d, - struct blk_integrity *b) -{ -} -static inline void blk_integrity_unregister(struct gendisk *d) -{ -} static inline unsigned short queue_max_integrity_segments(const struct request_queue *q) { @@ -157,4 +151,5 @@ static inline struct bio_vec *rq_integrity_vec(struct request *rq) return NULL; } #endif /* CONFIG_BLK_DEV_INTEGRITY */ + #endif /* _LINUX_BLK_INTEGRITY_H */ diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index f9089750919c6b..0c247a71688561 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -334,6 +334,8 @@ struct queue_limits { * due to possible offsets. */ unsigned int dma_alignment; + + struct blk_integrity integrity; }; typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx, @@ -419,10 +421,6 @@ struct request_queue { struct queue_limits limits; -#ifdef CONFIG_BLK_DEV_INTEGRITY - struct blk_integrity integrity; -#endif /* CONFIG_BLK_DEV_INTEGRITY */ - #ifdef CONFIG_PM struct device *dev; enum rpm_status rpm_status; @@ -1300,11 +1298,9 @@ static inline bool bdev_stable_writes(struct block_device *bdev) { struct request_queue *q = bdev_get_queue(bdev); -#ifdef CONFIG_BLK_DEV_INTEGRITY - /* BLK_INTEGRITY_CSUM_NONE is not available in blkdev.h */ - if (q->integrity.csum_type != 0) + if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && + q->limits.integrity.csum_type != BLK_INTEGRITY_CSUM_NONE) return true; -#endif return test_bit(QUEUE_FLAG_STABLE_WRITES, &q->queue_flags); } diff --git a/include/linux/t10-pi.h b/include/linux/t10-pi.h index d2bafb76badfb9..1773610010ebaf 100644 --- a/include/linux/t10-pi.h +++ b/include/linux/t10-pi.h @@ -39,12 +39,8 @@ struct t10_pi_tuple { static inline u32 t10_pi_ref_tag(struct request *rq) { - unsigned int shift = ilog2(queue_logical_block_size(rq->q)); + unsigned int shift = rq->q->limits.integrity.interval_exp; -#ifdef CONFIG_BLK_DEV_INTEGRITY - if (rq->q->integrity.interval_exp) - shift = rq->q->integrity.interval_exp; -#endif return blk_rq_pos(rq) >> (shift - SECTOR_SHIFT) & 0xffffffff; } @@ -65,12 +61,8 @@ static inline u64 lower_48_bits(u64 n) static inline u64 ext_pi_ref_tag(struct request *rq) { - unsigned int shift = ilog2(queue_logical_block_size(rq->q)); + unsigned int shift = rq->q->limits.integrity.interval_exp; -#ifdef CONFIG_BLK_DEV_INTEGRITY - if (rq->q->integrity.interval_exp) - shift = rq->q->integrity.interval_exp; -#endif return lower_48_bits(blk_rq_pos(rq) >> (shift - SECTOR_SHIFT)); } From 52734206b2273aa0b3448980a58ec920a8e571a5 Mon Sep 17 00:00:00 2001 From: Gabriel Krisman Bertazi Date: Thu, 23 May 2024 17:45:17 -0400 Subject: [PATCH 049/209] io_uring: Drop per-ctx dummy_ubuf Commit 19a63c402170 ("io_uring/rsrc: keep one global dummy_ubuf") replaced it with a global static object but this stayed behind. Fixes: 19a63c402170 ("io_uring/rsrc: keep one global dummy_ubuf") Signed-off-by: Gabriel Krisman Bertazi Link: https://lore.kernel.org/r/20240523214517.31803-1-krisman@suse.de Signed-off-by: Jens Axboe --- include/linux/io_uring_types.h | 1 - 1 file changed, 1 deletion(-) diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h index 7a6b190c7da74a..91224bbcfa73ff 100644 --- a/include/linux/io_uring_types.h +++ b/include/linux/io_uring_types.h @@ -373,7 +373,6 @@ struct io_ring_ctx { struct io_restriction restrictions; /* slow path rsrc auxilary data, used by update/register */ - struct io_mapped_ubuf *dummy_ubuf; struct io_rsrc_data *file_data; struct io_rsrc_data *buf_data; From 4faf204386db68d8457b5ecd6cbcef56fcca328a Mon Sep 17 00:00:00 2001 From: Gabriel Krisman Bertazi Date: Thu, 23 May 2024 17:45:35 -0400 Subject: [PATCH 050/209] io_uring/rsrc: Drop io_copy_iov in favor of iovec API Instead of open coding an io_uring function to copy iovs from userspace, rely on the existing iovec_from_user function. While there, avoid repeatedly zeroing the iov in the !arg case for io_sqe_buffer_register. tested with liburing testsuite. Signed-off-by: Gabriel Krisman Bertazi Link: https://lore.kernel.org/r/20240523214535.31890-1-krisman@suse.de Signed-off-by: Jens Axboe --- io_uring/rsrc.c | 60 +++++++++++++++++-------------------------------- 1 file changed, 21 insertions(+), 39 deletions(-) diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c index 65417c9553b1d8..338e771bcafb3b 100644 --- a/io_uring/rsrc.c +++ b/io_uring/rsrc.c @@ -85,31 +85,6 @@ static int io_account_mem(struct io_ring_ctx *ctx, unsigned long nr_pages) return 0; } -static int io_copy_iov(struct io_ring_ctx *ctx, struct iovec *dst, - void __user *arg, unsigned index) -{ - struct iovec __user *src; - -#ifdef CONFIG_COMPAT - if (ctx->compat) { - struct compat_iovec __user *ciovs; - struct compat_iovec ciov; - - ciovs = (struct compat_iovec __user *) arg; - if (copy_from_user(&ciov, &ciovs[index], sizeof(ciov))) - return -EFAULT; - - dst->iov_base = u64_to_user_ptr((u64)ciov.iov_base); - dst->iov_len = ciov.iov_len; - return 0; - } -#endif - src = (struct iovec __user *) arg; - if (copy_from_user(dst, &src[index], sizeof(*dst))) - return -EFAULT; - return 0; -} - static int io_buffer_validate(struct iovec *iov) { unsigned long tmp, acct_len = iov->iov_len + (PAGE_SIZE - 1); @@ -419,8 +394,9 @@ static int __io_sqe_buffers_update(struct io_ring_ctx *ctx, struct io_uring_rsrc_update2 *up, unsigned int nr_args) { + struct iovec __user *uvec = u64_to_user_ptr(up->data); u64 __user *tags = u64_to_user_ptr(up->tags); - struct iovec iov, __user *iovs = u64_to_user_ptr(up->data); + struct iovec fast_iov, *iov; struct page *last_hpage = NULL; __u32 done; int i, err; @@ -434,21 +410,23 @@ static int __io_sqe_buffers_update(struct io_ring_ctx *ctx, struct io_mapped_ubuf *imu; u64 tag = 0; - err = io_copy_iov(ctx, &iov, iovs, done); - if (err) + iov = iovec_from_user(&uvec[done], 1, 1, &fast_iov, ctx->compat); + if (IS_ERR(iov)) { + err = PTR_ERR(iov); break; + } if (tags && copy_from_user(&tag, &tags[done], sizeof(tag))) { err = -EFAULT; break; } - err = io_buffer_validate(&iov); + err = io_buffer_validate(iov); if (err) break; - if (!iov.iov_base && tag) { + if (!iov->iov_base && tag) { err = -EINVAL; break; } - err = io_sqe_buffer_register(ctx, &iov, &imu, &last_hpage); + err = io_sqe_buffer_register(ctx, iov, &imu, &last_hpage); if (err) break; @@ -970,8 +948,9 @@ int io_sqe_buffers_register(struct io_ring_ctx *ctx, void __user *arg, { struct page *last_hpage = NULL; struct io_rsrc_data *data; + struct iovec fast_iov, *iov = &fast_iov; + const struct iovec __user *uvec = (struct iovec * __user) arg; int i, ret; - struct iovec iov; BUILD_BUG_ON(IORING_MAX_REG_BUFFERS >= (1u << 16)); @@ -988,24 +967,27 @@ int io_sqe_buffers_register(struct io_ring_ctx *ctx, void __user *arg, return ret; } + if (!arg) + memset(iov, 0, sizeof(*iov)); + for (i = 0; i < nr_args; i++, ctx->nr_user_bufs++) { if (arg) { - ret = io_copy_iov(ctx, &iov, arg, i); - if (ret) + iov = iovec_from_user(&uvec[i], 1, 1, &fast_iov, ctx->compat); + if (IS_ERR(iov)) { + ret = PTR_ERR(iov); break; - ret = io_buffer_validate(&iov); + } + ret = io_buffer_validate(iov); if (ret) break; - } else { - memset(&iov, 0, sizeof(iov)); } - if (!iov.iov_base && *io_get_tag_slot(data, i)) { + if (!iov->iov_base && *io_get_tag_slot(data, i)) { ret = -EINVAL; break; } - ret = io_sqe_buffer_register(ctx, &iov, &ctx->user_bufs[i], + ret = io_sqe_buffer_register(ctx, iov, &ctx->user_bufs[i], &last_hpage); if (ret) break; From 4313c8d79a6d8b6213fa3271f7b758bb64f87a24 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 3 Jun 2024 11:19:10 -0600 Subject: [PATCH 051/209] io_uring/eventfd: move to more idiomatic RCU free usage In some ways, it just "happens to work" currently with using the ops field for both the free and signaling bit. But it depends on ordering of operations in terms of freeing and signaling. Clean it up and use the usual refs == 0 under RCU read side lock to determine if the ev_fd is still valid, and use the reference to gate the freeing as well. Fixes: 21a091b970cd ("io_uring: signal registered eventfd to process deferred task work") Signed-off-by: Jens Axboe --- io_uring/io_uring.c | 49 ++++++++++++++++++++++++--------------------- io_uring/io_uring.h | 4 ++-- io_uring/register.c | 6 +++--- 3 files changed, 31 insertions(+), 28 deletions(-) diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 816e93e7f94907..b874836ee49db5 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -541,29 +541,33 @@ static __cold void io_queue_deferred(struct io_ring_ctx *ctx) } } -void io_eventfd_ops(struct rcu_head *rcu) +void io_eventfd_free(struct rcu_head *rcu) { struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); - int ops = atomic_xchg(&ev_fd->ops, 0); - if (ops & BIT(IO_EVENTFD_OP_SIGNAL_BIT)) - eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); + eventfd_ctx_put(ev_fd->cq_ev_fd); + kfree(ev_fd); +} - /* IO_EVENTFD_OP_FREE_BIT may not be set here depending on callback - * ordering in a race but if references are 0 we know we have to free - * it regardless. - */ - if (atomic_dec_and_test(&ev_fd->refs)) { - eventfd_ctx_put(ev_fd->cq_ev_fd); - kfree(ev_fd); - } +void io_eventfd_do_signal(struct rcu_head *rcu) +{ + struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); + + eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); + + if (atomic_dec_and_test(&ev_fd->refs)) + io_eventfd_free(rcu); } static void io_eventfd_signal(struct io_ring_ctx *ctx) { struct io_ev_fd *ev_fd = NULL; - rcu_read_lock(); + if (READ_ONCE(ctx->rings->cq_flags) & IORING_CQ_EVENTFD_DISABLED) + return; + + guard(rcu)(); + /* * rcu_dereference ctx->io_ev_fd once and use it for both for checking * and eventfd_signal @@ -576,24 +580,23 @@ static void io_eventfd_signal(struct io_ring_ctx *ctx) * the function and rcu_read_lock. */ if (unlikely(!ev_fd)) - goto out; - if (READ_ONCE(ctx->rings->cq_flags) & IORING_CQ_EVENTFD_DISABLED) - goto out; + return; + if (!atomic_inc_not_zero(&ev_fd->refs)) + return; if (ev_fd->eventfd_async && !io_wq_current_is_worker()) goto out; if (likely(eventfd_signal_allowed())) { eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); } else { - atomic_inc(&ev_fd->refs); - if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_SIGNAL_BIT), &ev_fd->ops)) - call_rcu_hurry(&ev_fd->rcu, io_eventfd_ops); - else - atomic_dec(&ev_fd->refs); + if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_SIGNAL_BIT), &ev_fd->ops)) { + call_rcu_hurry(&ev_fd->rcu, io_eventfd_do_signal); + return; + } } - out: - rcu_read_unlock(); + if (atomic_dec_and_test(&ev_fd->refs)) + call_rcu(&ev_fd->rcu, io_eventfd_free); } static void io_eventfd_flush_signal(struct io_ring_ctx *ctx) diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h index 726e6367af4d37..2b08b402b716d8 100644 --- a/io_uring/io_uring.h +++ b/io_uring/io_uring.h @@ -106,10 +106,10 @@ bool io_match_task_safe(struct io_kiocb *head, struct task_struct *task, enum { IO_EVENTFD_OP_SIGNAL_BIT, - IO_EVENTFD_OP_FREE_BIT, }; -void io_eventfd_ops(struct rcu_head *rcu); +void io_eventfd_do_signal(struct rcu_head *rcu); +void io_eventfd_free(struct rcu_head *rcu); void io_activate_pollwq(struct io_ring_ctx *ctx); static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx) diff --git a/io_uring/register.c b/io_uring/register.c index c0010a66a6f2c2..212711e9bc8a67 100644 --- a/io_uring/register.c +++ b/io_uring/register.c @@ -63,9 +63,9 @@ static int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg, ev_fd->eventfd_async = eventfd_async; ctx->has_evfd = true; - rcu_assign_pointer(ctx->io_ev_fd, ev_fd); atomic_set(&ev_fd->refs, 1); atomic_set(&ev_fd->ops, 0); + rcu_assign_pointer(ctx->io_ev_fd, ev_fd); return 0; } @@ -78,8 +78,8 @@ int io_eventfd_unregister(struct io_ring_ctx *ctx) if (ev_fd) { ctx->has_evfd = false; rcu_assign_pointer(ctx->io_ev_fd, NULL); - if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_FREE_BIT), &ev_fd->ops)) - call_rcu(&ev_fd->rcu, io_eventfd_ops); + if (atomic_dec_and_test(&ev_fd->refs)) + call_rcu(&ev_fd->rcu, io_eventfd_free); return 0; } From 31718d0ba5777340c23604708b471e02fb9e0c2f Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 3 Jun 2024 11:51:19 -0600 Subject: [PATCH 052/209] io_uring/eventfd: move eventfd handling to separate file This is pretty nicely abstracted already, but let's move it to a separate file rather than have it in the main io_uring file. With that, we can also move the io_ev_fd struct and enum out of global scope. Signed-off-by: Jens Axboe --- include/linux/io_uring_types.h | 8 -- io_uring/Makefile | 6 +- io_uring/eventfd.c | 160 +++++++++++++++++++++++++++++++++ io_uring/eventfd.h | 8 ++ io_uring/io_uring.c | 82 +---------------- io_uring/io_uring.h | 6 -- io_uring/register.c | 56 +----------- 7 files changed, 173 insertions(+), 153 deletions(-) create mode 100644 io_uring/eventfd.c create mode 100644 io_uring/eventfd.h diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h index 91224bbcfa73ff..a2227ab7fd1666 100644 --- a/include/linux/io_uring_types.h +++ b/include/linux/io_uring_types.h @@ -211,14 +211,6 @@ struct io_submit_state { struct blk_plug plug; }; -struct io_ev_fd { - struct eventfd_ctx *cq_ev_fd; - unsigned int eventfd_async: 1; - struct rcu_head rcu; - atomic_t refs; - atomic_t ops; -}; - struct io_alloc_cache { void **entries; unsigned int nr_cached; diff --git a/io_uring/Makefile b/io_uring/Makefile index fc1b23c524e83b..61923e11c76772 100644 --- a/io_uring/Makefile +++ b/io_uring/Makefile @@ -4,9 +4,9 @@ obj-$(CONFIG_IO_URING) += io_uring.o opdef.o kbuf.o rsrc.o notif.o \ tctx.o filetable.o rw.o net.o poll.o \ - uring_cmd.o openclose.o sqpoll.o \ - xattr.o nop.o fs.o splice.o sync.o \ - msg_ring.o advise.o openclose.o \ + eventfd.o uring_cmd.o openclose.o \ + sqpoll.o xattr.o nop.o fs.o splice.o \ + sync.o msg_ring.o advise.o openclose.o \ epoll.o statx.o timeout.o fdinfo.o \ cancel.o waitid.o register.o \ truncate.o memmap.o diff --git a/io_uring/eventfd.c b/io_uring/eventfd.c new file mode 100644 index 00000000000000..b9384503a2b752 --- /dev/null +++ b/io_uring/eventfd.c @@ -0,0 +1,160 @@ +// SPDX-License-Identifier: GPL-2.0 +#include +#include +#include +#include +#include +#include +#include +#include + +#include "io-wq.h" +#include "eventfd.h" + +struct io_ev_fd { + struct eventfd_ctx *cq_ev_fd; + unsigned int eventfd_async: 1; + struct rcu_head rcu; + atomic_t refs; + atomic_t ops; +}; + +enum { + IO_EVENTFD_OP_SIGNAL_BIT, +}; + +static void io_eventfd_free(struct rcu_head *rcu) +{ + struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); + + eventfd_ctx_put(ev_fd->cq_ev_fd); + kfree(ev_fd); +} + +static void io_eventfd_do_signal(struct rcu_head *rcu) +{ + struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); + + eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); + + if (atomic_dec_and_test(&ev_fd->refs)) + io_eventfd_free(rcu); +} + +void io_eventfd_signal(struct io_ring_ctx *ctx) +{ + struct io_ev_fd *ev_fd = NULL; + + if (READ_ONCE(ctx->rings->cq_flags) & IORING_CQ_EVENTFD_DISABLED) + return; + + guard(rcu)(); + + /* + * rcu_dereference ctx->io_ev_fd once and use it for both for checking + * and eventfd_signal + */ + ev_fd = rcu_dereference(ctx->io_ev_fd); + + /* + * Check again if ev_fd exists incase an io_eventfd_unregister call + * completed between the NULL check of ctx->io_ev_fd at the start of + * the function and rcu_read_lock. + */ + if (unlikely(!ev_fd)) + return; + if (!atomic_inc_not_zero(&ev_fd->refs)) + return; + if (ev_fd->eventfd_async && !io_wq_current_is_worker()) + goto out; + + if (likely(eventfd_signal_allowed())) { + eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); + } else { + if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_SIGNAL_BIT), &ev_fd->ops)) { + call_rcu_hurry(&ev_fd->rcu, io_eventfd_do_signal); + return; + } + } +out: + if (atomic_dec_and_test(&ev_fd->refs)) + call_rcu(&ev_fd->rcu, io_eventfd_free); +} + +void io_eventfd_flush_signal(struct io_ring_ctx *ctx) +{ + bool skip; + + spin_lock(&ctx->completion_lock); + + /* + * Eventfd should only get triggered when at least one event has been + * posted. Some applications rely on the eventfd notification count + * only changing IFF a new CQE has been added to the CQ ring. There's + * no depedency on 1:1 relationship between how many times this + * function is called (and hence the eventfd count) and number of CQEs + * posted to the CQ ring. + */ + skip = ctx->cached_cq_tail == ctx->evfd_last_cq_tail; + ctx->evfd_last_cq_tail = ctx->cached_cq_tail; + spin_unlock(&ctx->completion_lock); + if (skip) + return; + + io_eventfd_signal(ctx); +} + +int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg, + unsigned int eventfd_async) +{ + struct io_ev_fd *ev_fd; + __s32 __user *fds = arg; + int fd; + + ev_fd = rcu_dereference_protected(ctx->io_ev_fd, + lockdep_is_held(&ctx->uring_lock)); + if (ev_fd) + return -EBUSY; + + if (copy_from_user(&fd, fds, sizeof(*fds))) + return -EFAULT; + + ev_fd = kmalloc(sizeof(*ev_fd), GFP_KERNEL); + if (!ev_fd) + return -ENOMEM; + + ev_fd->cq_ev_fd = eventfd_ctx_fdget(fd); + if (IS_ERR(ev_fd->cq_ev_fd)) { + int ret = PTR_ERR(ev_fd->cq_ev_fd); + kfree(ev_fd); + return ret; + } + + spin_lock(&ctx->completion_lock); + ctx->evfd_last_cq_tail = ctx->cached_cq_tail; + spin_unlock(&ctx->completion_lock); + + ev_fd->eventfd_async = eventfd_async; + ctx->has_evfd = true; + atomic_set(&ev_fd->refs, 1); + atomic_set(&ev_fd->ops, 0); + rcu_assign_pointer(ctx->io_ev_fd, ev_fd); + return 0; +} + +int io_eventfd_unregister(struct io_ring_ctx *ctx) +{ + struct io_ev_fd *ev_fd; + + ev_fd = rcu_dereference_protected(ctx->io_ev_fd, + lockdep_is_held(&ctx->uring_lock)); + if (ev_fd) { + ctx->has_evfd = false; + rcu_assign_pointer(ctx->io_ev_fd, NULL); + if (atomic_dec_and_test(&ev_fd->refs)) + call_rcu(&ev_fd->rcu, io_eventfd_free); + return 0; + } + + return -ENXIO; +} diff --git a/io_uring/eventfd.h b/io_uring/eventfd.h new file mode 100644 index 00000000000000..d394f49c632105 --- /dev/null +++ b/io_uring/eventfd.h @@ -0,0 +1,8 @@ + +struct io_ring_ctx; +int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg, + unsigned int eventfd_async); +int io_eventfd_unregister(struct io_ring_ctx *ctx); + +void io_eventfd_flush_signal(struct io_ring_ctx *ctx); +void io_eventfd_signal(struct io_ring_ctx *ctx); diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index b874836ee49db5..96f6da0bf5cd98 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -101,6 +101,7 @@ #include "poll.h" #include "rw.h" #include "alloc_cache.h" +#include "eventfd.h" #define IORING_MAX_ENTRIES 32768 #define IORING_MAX_CQ_ENTRIES (2 * IORING_MAX_ENTRIES) @@ -541,87 +542,6 @@ static __cold void io_queue_deferred(struct io_ring_ctx *ctx) } } -void io_eventfd_free(struct rcu_head *rcu) -{ - struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); - - eventfd_ctx_put(ev_fd->cq_ev_fd); - kfree(ev_fd); -} - -void io_eventfd_do_signal(struct rcu_head *rcu) -{ - struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); - - eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); - - if (atomic_dec_and_test(&ev_fd->refs)) - io_eventfd_free(rcu); -} - -static void io_eventfd_signal(struct io_ring_ctx *ctx) -{ - struct io_ev_fd *ev_fd = NULL; - - if (READ_ONCE(ctx->rings->cq_flags) & IORING_CQ_EVENTFD_DISABLED) - return; - - guard(rcu)(); - - /* - * rcu_dereference ctx->io_ev_fd once and use it for both for checking - * and eventfd_signal - */ - ev_fd = rcu_dereference(ctx->io_ev_fd); - - /* - * Check again if ev_fd exists incase an io_eventfd_unregister call - * completed between the NULL check of ctx->io_ev_fd at the start of - * the function and rcu_read_lock. - */ - if (unlikely(!ev_fd)) - return; - if (!atomic_inc_not_zero(&ev_fd->refs)) - return; - if (ev_fd->eventfd_async && !io_wq_current_is_worker()) - goto out; - - if (likely(eventfd_signal_allowed())) { - eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); - } else { - if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_SIGNAL_BIT), &ev_fd->ops)) { - call_rcu_hurry(&ev_fd->rcu, io_eventfd_do_signal); - return; - } - } -out: - if (atomic_dec_and_test(&ev_fd->refs)) - call_rcu(&ev_fd->rcu, io_eventfd_free); -} - -static void io_eventfd_flush_signal(struct io_ring_ctx *ctx) -{ - bool skip; - - spin_lock(&ctx->completion_lock); - - /* - * Eventfd should only get triggered when at least one event has been - * posted. Some applications rely on the eventfd notification count - * only changing IFF a new CQE has been added to the CQ ring. There's - * no depedency on 1:1 relationship between how many times this - * function is called (and hence the eventfd count) and number of CQEs - * posted to the CQ ring. - */ - skip = ctx->cached_cq_tail == ctx->evfd_last_cq_tail; - ctx->evfd_last_cq_tail = ctx->cached_cq_tail; - spin_unlock(&ctx->completion_lock); - if (skip) - return; - - io_eventfd_signal(ctx); -} - void __io_commit_cqring_flush(struct io_ring_ctx *ctx) { if (ctx->poll_activated) diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h index 2b08b402b716d8..cd43924eed04e1 100644 --- a/io_uring/io_uring.h +++ b/io_uring/io_uring.h @@ -104,12 +104,6 @@ bool __io_alloc_req_refill(struct io_ring_ctx *ctx); bool io_match_task_safe(struct io_kiocb *head, struct task_struct *task, bool cancel_all); -enum { - IO_EVENTFD_OP_SIGNAL_BIT, -}; - -void io_eventfd_do_signal(struct rcu_head *rcu); -void io_eventfd_free(struct rcu_head *rcu); void io_activate_pollwq(struct io_ring_ctx *ctx); static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx) diff --git a/io_uring/register.c b/io_uring/register.c index 212711e9bc8a67..f121e02f5e10e6 100644 --- a/io_uring/register.c +++ b/io_uring/register.c @@ -27,65 +27,11 @@ #include "cancel.h" #include "kbuf.h" #include "napi.h" +#include "eventfd.h" #define IORING_MAX_RESTRICTIONS (IORING_RESTRICTION_LAST + \ IORING_REGISTER_LAST + IORING_OP_LAST) -static int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg, - unsigned int eventfd_async) -{ - struct io_ev_fd *ev_fd; - __s32 __user *fds = arg; - int fd; - - ev_fd = rcu_dereference_protected(ctx->io_ev_fd, - lockdep_is_held(&ctx->uring_lock)); - if (ev_fd) - return -EBUSY; - - if (copy_from_user(&fd, fds, sizeof(*fds))) - return -EFAULT; - - ev_fd = kmalloc(sizeof(*ev_fd), GFP_KERNEL); - if (!ev_fd) - return -ENOMEM; - - ev_fd->cq_ev_fd = eventfd_ctx_fdget(fd); - if (IS_ERR(ev_fd->cq_ev_fd)) { - int ret = PTR_ERR(ev_fd->cq_ev_fd); - kfree(ev_fd); - return ret; - } - - spin_lock(&ctx->completion_lock); - ctx->evfd_last_cq_tail = ctx->cached_cq_tail; - spin_unlock(&ctx->completion_lock); - - ev_fd->eventfd_async = eventfd_async; - ctx->has_evfd = true; - atomic_set(&ev_fd->refs, 1); - atomic_set(&ev_fd->ops, 0); - rcu_assign_pointer(ctx->io_ev_fd, ev_fd); - return 0; -} - -int io_eventfd_unregister(struct io_ring_ctx *ctx) -{ - struct io_ev_fd *ev_fd; - - ev_fd = rcu_dereference_protected(ctx->io_ev_fd, - lockdep_is_held(&ctx->uring_lock)); - if (ev_fd) { - ctx->has_evfd = false; - rcu_assign_pointer(ctx->io_ev_fd, NULL); - if (atomic_dec_and_test(&ev_fd->refs)) - call_rcu(&ev_fd->rcu, io_eventfd_free); - return 0; - } - - return -ENXIO; -} - static __cold int io_probe(struct io_ring_ctx *ctx, void __user *arg, unsigned nr_args) { From 7cc167a8abaa1f4b904bd85a2927b44c5602ad7c Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Fri, 14 Jun 2024 10:57:03 -0600 Subject: [PATCH 053/209] io_uring: use 'state' consistently __io_submit_flush_completions() assigns ctx->submit_state to a local variable and uses it in all but one spot, switch that forgotten statement to using 'state' as well. Signed-off-by: Jens Axboe --- io_uring/io_uring.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 96f6da0bf5cd98..0c86f504fc6659 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -1390,7 +1390,7 @@ void __io_submit_flush_completions(struct io_ring_ctx *ctx) } __io_cq_unlock_post(ctx); - if (!wq_list_empty(&ctx->submit_state.compl_reqs)) { + if (!wq_list_empty(&state->compl_reqs)) { io_free_batch_list(ctx, state->compl_reqs.first); INIT_WQ_LIST(&state->compl_reqs); } From 940b8c41dd4ecbc3e334dfe46b3694ba7a91406e Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 13 Jun 2024 19:28:27 +0000 Subject: [PATCH 054/209] io_uring/io-wq: make io_wq_work flags atomic The work flags can be set/accessed from different tasks, both the originator of the request, and the io-wq workers. While modifications aren't concurrent, it still makes KMSAN unhappy. There's no real downside to just making the flag reading/manipulation use proper atomics here. Signed-off-by: Jens Axboe --- include/linux/io_uring_types.h | 2 +- io_uring/io-wq.c | 19 ++++++++++--------- io_uring/io-wq.h | 2 +- io_uring/io_uring.c | 12 ++++++------ 4 files changed, 18 insertions(+), 17 deletions(-) diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h index a2227ab7fd1666..d3f88b5e11955b 100644 --- a/include/linux/io_uring_types.h +++ b/include/linux/io_uring_types.h @@ -50,7 +50,7 @@ struct io_wq_work_list { struct io_wq_work { struct io_wq_work_node list; - unsigned flags; + atomic_t flags; /* place it here instead of io_kiocb as it fills padding and saves 4B */ int cancel_seq; }; diff --git a/io_uring/io-wq.c b/io_uring/io-wq.c index 7d3316fe9bfc46..913c92249522e8 100644 --- a/io_uring/io-wq.c +++ b/io_uring/io-wq.c @@ -159,7 +159,7 @@ static inline struct io_wq_acct *io_get_acct(struct io_wq *wq, bool bound) static inline struct io_wq_acct *io_work_get_acct(struct io_wq *wq, struct io_wq_work *work) { - return io_get_acct(wq, !(work->flags & IO_WQ_WORK_UNBOUND)); + return io_get_acct(wq, !(atomic_read(&work->flags) & IO_WQ_WORK_UNBOUND)); } static inline struct io_wq_acct *io_wq_get_acct(struct io_worker *worker) @@ -451,7 +451,7 @@ static void __io_worker_idle(struct io_wq *wq, struct io_worker *worker) static inline unsigned int io_get_work_hash(struct io_wq_work *work) { - return work->flags >> IO_WQ_HASH_SHIFT; + return atomic_read(&work->flags) >> IO_WQ_HASH_SHIFT; } static bool io_wait_on_hash(struct io_wq *wq, unsigned int hash) @@ -592,8 +592,9 @@ static void io_worker_handle_work(struct io_wq_acct *acct, next_hashed = wq_next_work(work); - if (unlikely(do_kill) && (work->flags & IO_WQ_WORK_UNBOUND)) - work->flags |= IO_WQ_WORK_CANCEL; + if (do_kill && + (atomic_read(&work->flags) & IO_WQ_WORK_UNBOUND)) + atomic_or(IO_WQ_WORK_CANCEL, &work->flags); wq->do_work(work); io_assign_current_work(worker, NULL); @@ -891,7 +892,7 @@ static bool io_wq_worker_wake(struct io_worker *worker, void *data) static void io_run_cancel(struct io_wq_work *work, struct io_wq *wq) { do { - work->flags |= IO_WQ_WORK_CANCEL; + atomic_or(IO_WQ_WORK_CANCEL, &work->flags); wq->do_work(work); work = wq->free_work(work); } while (work); @@ -926,7 +927,7 @@ static bool io_wq_work_match_item(struct io_wq_work *work, void *data) void io_wq_enqueue(struct io_wq *wq, struct io_wq_work *work) { struct io_wq_acct *acct = io_work_get_acct(wq, work); - unsigned long work_flags = work->flags; + unsigned int work_flags = atomic_read(&work->flags); struct io_cb_cancel_data match = { .fn = io_wq_work_match_item, .data = work, @@ -939,7 +940,7 @@ void io_wq_enqueue(struct io_wq *wq, struct io_wq_work *work) * been marked as one that should not get executed, cancel it here. */ if (test_bit(IO_WQ_BIT_EXIT, &wq->state) || - (work->flags & IO_WQ_WORK_CANCEL)) { + (work_flags & IO_WQ_WORK_CANCEL)) { io_run_cancel(work, wq); return; } @@ -982,7 +983,7 @@ void io_wq_hash_work(struct io_wq_work *work, void *val) unsigned int bit; bit = hash_ptr(val, IO_WQ_HASH_ORDER); - work->flags |= (IO_WQ_WORK_HASHED | (bit << IO_WQ_HASH_SHIFT)); + atomic_or(IO_WQ_WORK_HASHED | (bit << IO_WQ_HASH_SHIFT), &work->flags); } static bool __io_wq_worker_cancel(struct io_worker *worker, @@ -990,7 +991,7 @@ static bool __io_wq_worker_cancel(struct io_worker *worker, struct io_wq_work *work) { if (work && match->fn(work, match->data)) { - work->flags |= IO_WQ_WORK_CANCEL; + atomic_or(IO_WQ_WORK_CANCEL, &work->flags); __set_notify_signal(worker->task); return true; } diff --git a/io_uring/io-wq.h b/io_uring/io-wq.h index 2b2a6406dd8ee8..b3b004a7b62528 100644 --- a/io_uring/io-wq.h +++ b/io_uring/io-wq.h @@ -56,7 +56,7 @@ bool io_wq_worker_stopped(void); static inline bool io_wq_is_hashed(struct io_wq_work *work) { - return work->flags & IO_WQ_WORK_HASHED; + return atomic_read(&work->flags) & IO_WQ_WORK_HASHED; } typedef bool (work_cancel_fn)(struct io_wq_work *, void *); diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 0c86f504fc6659..872477412726eb 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -462,9 +462,9 @@ static void io_prep_async_work(struct io_kiocb *req) } req->work.list.next = NULL; - req->work.flags = 0; + atomic_set(&req->work.flags, 0); if (req->flags & REQ_F_FORCE_ASYNC) - req->work.flags |= IO_WQ_WORK_CONCURRENT; + atomic_or(IO_WQ_WORK_CONCURRENT, &req->work.flags); if (req->file && !(req->flags & REQ_F_FIXED_FILE)) req->flags |= io_file_get_flags(req->file); @@ -480,7 +480,7 @@ static void io_prep_async_work(struct io_kiocb *req) io_wq_hash_work(&req->work, file_inode(req->file)); } else if (!req->file || !S_ISBLK(file_inode(req->file)->i_mode)) { if (def->unbound_nonreg_file) - req->work.flags |= IO_WQ_WORK_UNBOUND; + atomic_or(IO_WQ_WORK_UNBOUND, &req->work.flags); } } @@ -520,7 +520,7 @@ static void io_queue_iowq(struct io_kiocb *req) * worker for it). */ if (WARN_ON_ONCE(!same_thread_group(req->task, current))) - req->work.flags |= IO_WQ_WORK_CANCEL; + atomic_or(IO_WQ_WORK_CANCEL, &req->work.flags); trace_io_uring_queue_async_work(req, io_wq_is_hashed(&req->work)); io_wq_enqueue(tctx->io_wq, &req->work); @@ -1736,14 +1736,14 @@ void io_wq_submit_work(struct io_wq_work *work) io_arm_ltimeout(req); /* either cancelled or io-wq is dying, so don't touch tctx->iowq */ - if (work->flags & IO_WQ_WORK_CANCEL) { + if (atomic_read(&work->flags) & IO_WQ_WORK_CANCEL) { fail: io_req_task_queue_fail(req, err); return; } if (!io_assign_file(req, def, issue_flags)) { err = -EBADF; - work->flags |= IO_WQ_WORK_CANCEL; + atomic_or(IO_WQ_WORK_CANCEL, &work->flags); goto fail; } From 5e3b7009f116f684ac6b93d8924506154f3b1f6d Mon Sep 17 00:00:00 2001 From: Andreas Hindborg Date: Sat, 15 Jun 2024 01:53:50 +0200 Subject: [PATCH 055/209] rust: block: do not use removed queue limit API The Rust block layer API was using the old queue limit API, which was just removed. Use the new API instead. Reported-by: Boqun Feng Fixes: 3253aba3408a ("rust: block: introduce `kernel::block::mq` module") Signed-off-by: Andreas Hindborg Link: https://lore.kernel.org/r/20240614235350.621121-1-nmi@metaspace.dk Signed-off-by: Jens Axboe --- rust/kernel/block/mq/gen_disk.rs | 20 +++++++------------- 1 file changed, 7 insertions(+), 13 deletions(-) diff --git a/rust/kernel/block/mq/gen_disk.rs b/rust/kernel/block/mq/gen_disk.rs index 3b9edb96c8ff77..e06044b549e0be 100644 --- a/rust/kernel/block/mq/gen_disk.rs +++ b/rust/kernel/block/mq/gen_disk.rs @@ -95,11 +95,17 @@ impl GenDiskBuilder { ) -> Result> { let lock_class_key = crate::sync::LockClassKey::new(); + // SAFETY: `bindings::queue_limits` contain only fields that are valid when zeroed. + let mut lim: bindings::queue_limits = unsafe { core::mem::zeroed() }; + + lim.logical_block_size = self.logical_block_size; + lim.physical_block_size = self.physical_block_size; + // SAFETY: `tagset.raw_tag_set()` points to a valid and initialized tag set let gendisk = from_err_ptr(unsafe { bindings::__blk_mq_alloc_disk( tagset.raw_tag_set(), - core::ptr::null_mut(), // TODO: We can pass queue limits right here + &mut lim, core::ptr::null_mut(), lock_class_key.as_ptr(), ) @@ -141,18 +147,6 @@ impl GenDiskBuilder { raw_writer.write_fmt(name)?; raw_writer.write_char('\0')?; - // SAFETY: `gendisk` points to a valid and initialized instance of - // `struct gendisk`. We have exclusive access, so we cannot race. - unsafe { - bindings::blk_queue_logical_block_size((*gendisk).queue, self.logical_block_size) - }; - - // SAFETY: `gendisk` points to a valid and initialized instance of - // `struct gendisk`. We have exclusive access, so we cannot race. - unsafe { - bindings::blk_queue_physical_block_size((*gendisk).queue, self.physical_block_size) - }; - // SAFETY: `gendisk` points to a valid and initialized instance of // `struct gendisk`. `set_capacity` takes a lock to synchronize this // operation, so we will not race. From e21d12c7cd5cef8a6c5367f96aaab01249216ded Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Tue, 11 Jun 2024 11:36:36 +0900 Subject: [PATCH 056/209] block: Improve checks on zone resource limits Make sure that the zone resource limits of a zoned block device are correct by checking that: (a) If the device has a max active zones limit, make sure that the max open zones limit is lower than the max active zones limit. (b) If the device has zone resource limits, check that the limits values are lower than the number of sequential zones of the device. If it is not, assume that the zoned device has no limits by setting the limits to 0. For (a), a check is added to blk_validate_zoned_limits() and an error returned if the max open zones limit exceeds the value of the max active zone limit (if there is one). For (b), given that we need the number of sequential zones of the zoned device, this check is added to disk_update_zone_resources(). This is safe to do as that function is executed with the disk queue frozen and the check executed after queue_limits_start_update() which takes the queue limits lock. Of note is that the early return in this function for zoned devices that do not use zone write plugging (e.g. DM devices using native zone append) is moved to after the new check and adjustment of the zone resource limits so that the check applies to any zoned device. Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Reviewed-by: Niklas Cassel Reviewed-by: Benjamin Marzinski Link: https://lore.kernel.org/r/20240611023639.89277-2-dlemoal@kernel.org Signed-off-by: Jens Axboe --- block/blk-settings.c | 8 ++++++++ block/blk-zoned.c | 20 ++++++++++++++++---- 2 files changed, 24 insertions(+), 4 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index f11c8676eb4c67..f574181105f8a5 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -80,6 +80,14 @@ static int blk_validate_zoned_limits(struct queue_limits *lim) if (WARN_ON_ONCE(!IS_ENABLED(CONFIG_BLK_DEV_ZONED))) return -EINVAL; + /* + * Given that active zones include open zones, the maximum number of + * open zones cannot be larger than the maximum number of active zones. + */ + if (lim->max_active_zones && + lim->max_open_zones > lim->max_active_zones) + return -EINVAL; + if (lim->zone_write_granularity < lim->logical_block_size) lim->zone_write_granularity = lim->logical_block_size; diff --git a/block/blk-zoned.c b/block/blk-zoned.c index 52abebf56027a8..8f89705f5e1c50 100644 --- a/block/blk-zoned.c +++ b/block/blk-zoned.c @@ -1647,8 +1647,22 @@ static int disk_update_zone_resources(struct gendisk *disk, return -ENODEV; } + lim = queue_limits_start_update(q); + + /* + * Some devices can advertize zone resource limits that are larger than + * the number of sequential zones of the zoned block device, e.g. a + * small ZNS namespace. For such case, assume that the zoned device has + * no zone resource limits. + */ + nr_seq_zones = disk->nr_zones - nr_conv_zones; + if (lim.max_open_zones >= nr_seq_zones) + lim.max_open_zones = 0; + if (lim.max_active_zones >= nr_seq_zones) + lim.max_active_zones = 0; + if (!disk->zone_wplugs_pool) - return 0; + goto commit; /* * If the device has no limit on the maximum number of open and active @@ -1657,9 +1671,6 @@ static int disk_update_zone_resources(struct gendisk *disk, * dynamic zone write plug allocation when simultaneously writing to * more zones than the size of the mempool. */ - lim = queue_limits_start_update(q); - - nr_seq_zones = disk->nr_zones - nr_conv_zones; pool_size = max(lim.max_open_zones, lim.max_active_zones); if (!pool_size) pool_size = min(BLK_ZONE_WPLUG_DEFAULT_POOL_SIZE, nr_seq_zones); @@ -1673,6 +1684,7 @@ static int disk_update_zone_resources(struct gendisk *disk, lim.max_open_zones = 0; } +commit: return queue_limits_commit_update(q, &lim); } From 7f91ccd8a608dbe39b97a6e43d635378d493f77e Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Tue, 11 Jun 2024 11:36:37 +0900 Subject: [PATCH 057/209] dm: Call dm_revalidate_zones() after setting the queue limits dm_revalidate_zones() is called from dm_set_zone_restrictions() when the mapped device queue limits are not yet set. However, dm_revalidate_zones() calls blk_revalidate_disk_zones() and this function consults and modifies the mapped device queue limits. Thus, currently, blk_revalidate_disk_zones() operates on limits that are not yet initialized. Fix this by moving the call to dm_revalidate_zones() out of dm_set_zone_restrictions() and into dm_table_set_restrictions() after executing queue_limits_set(). To further cleanup dm_set_zones_restrictions(), the message about the type of zone append (native or emulated) is also moved inside dm_revalidate_zones(). Fixes: 1c0e720228ad ("dm: use queue_limits_set") Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Benjamin Marzinski Reviewed-by: Niklas Cassel Link: https://lore.kernel.org/r/20240611023639.89277-3-dlemoal@kernel.org Signed-off-by: Jens Axboe --- drivers/md/dm-table.c | 15 +++++++++++---- drivers/md/dm-zone.c | 25 ++++++++++--------------- drivers/md/dm.h | 1 + 3 files changed, 22 insertions(+), 19 deletions(-) diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index fd789eeb62d943..2805cd2b7c8bf2 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -1921,10 +1921,7 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, dm_table_any_dev_attr(t, device_is_not_random, NULL)) blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q); - /* - * For a zoned target, setup the zones related queue attributes - * and resources necessary for zone append emulation if necessary. - */ + /* For a zoned table, setup the zone related queue attributes. */ if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) && limits->zoned) { r = dm_set_zones_restrictions(t, q, limits); if (r) @@ -1935,6 +1932,16 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, if (r) return r; + /* + * Now that the limits are set, check the zones mapped by the table + * and setup the resources for zone append emulation if necessary. + */ + if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) && limits->zoned) { + r = dm_revalidate_zones(t, q); + if (r) + return r; + } + dm_update_crypto_profile(q, t); /* diff --git a/drivers/md/dm-zone.c b/drivers/md/dm-zone.c index 5d66d916730efa..75d0019a0649d4 100644 --- a/drivers/md/dm-zone.c +++ b/drivers/md/dm-zone.c @@ -166,14 +166,22 @@ static int dm_check_zoned_cb(struct blk_zone *zone, unsigned int idx, * blk_revalidate_disk_zones() function here as the mapped device is suspended * (this is called from __bind() context). */ -static int dm_revalidate_zones(struct mapped_device *md, struct dm_table *t) +int dm_revalidate_zones(struct dm_table *t, struct request_queue *q) { + struct mapped_device *md = t->md; struct gendisk *disk = md->disk; int ret; + if (!get_capacity(disk)) + return 0; + /* Revalidate only if something changed. */ - if (!disk->nr_zones || disk->nr_zones != md->nr_zones) + if (!disk->nr_zones || disk->nr_zones != md->nr_zones) { + DMINFO("%s using %s zone append", + disk->disk_name, + queue_emulates_zone_append(q) ? "emulated" : "native"); md->nr_zones = 0; + } if (md->nr_zones) return 0; @@ -240,9 +248,6 @@ int dm_set_zones_restrictions(struct dm_table *t, struct request_queue *q, lim->max_zone_append_sectors = 0; } - if (!get_capacity(md->disk)) - return 0; - /* * Count conventional zones to check that the mapped device will indeed * have sequential write required zones. @@ -269,16 +274,6 @@ int dm_set_zones_restrictions(struct dm_table *t, struct request_queue *q, return 0; } - if (!md->disk->nr_zones) { - DMINFO("%s using %s zone append", - md->disk->disk_name, - queue_emulates_zone_append(q) ? "emulated" : "native"); - } - - ret = dm_revalidate_zones(md, t); - if (ret < 0) - return ret; - if (!static_key_enabled(&zoned_enabled.key)) static_branch_enable(&zoned_enabled); return 0; diff --git a/drivers/md/dm.h b/drivers/md/dm.h index 53ef8207fe2c15..c984ecb64b1e89 100644 --- a/drivers/md/dm.h +++ b/drivers/md/dm.h @@ -103,6 +103,7 @@ int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t); */ int dm_set_zones_restrictions(struct dm_table *t, struct request_queue *q, struct queue_limits *lim); +int dm_revalidate_zones(struct dm_table *t, struct request_queue *q); void dm_zone_endio(struct dm_io *io, struct bio *clone); #ifdef CONFIG_BLK_DEV_ZONED int dm_blk_report_zones(struct gendisk *disk, sector_t sector, From 73a74af0c72b7cfd843cbd93e088fc5c51471a84 Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Tue, 11 Jun 2024 11:36:38 +0900 Subject: [PATCH 058/209] dm: Improve zone resource limits handling The generic stacking of limits implemented in the block layer cannot correctly handle stacking of zone resource limits (max open zones and max active zones) because these limits are for an entire device but the stacking may be for a portion of that device (e.g. a dm-linear target that does not cover an entire block device). As a result, when DM devices are created on top of zoned block devices, the DM device never has any zone resource limits advertized, which is only correct if all underlying target devices also have no zone resource limits. If at least one target device has resource limits, the user may see either performance issues (if the max open zone limit of the device is exceeded) or write I/O errors if the max active zone limit of one of the underlying target devices is exceeded. While it is very difficult to correctly and reliably stack zone resource limits in general, cases where targets are not sharing zone resources of the same device can be dealt with relatively easily. Such situation happens when a target maps all sequential zones of a zoned block device: for such mapping, other targets mapping other parts of the same zoned block device can only contain conventional zones and thus will not require any zone resource to correctly handle write operations. For a mapped device constructed with such targets, which includes mapped devices constructed with targets mapping entire zoned block devices, the zone resource limits can be reliably determined using the non-zero minimum of the zone resource limits of all targets. For mapped devices that include targets partially mapping the set of sequential write required zones of zoned block devices, instead of advertizing no zone resource limits, it is also better to set the mapped device limits to the non-zero minimum of the limits of all targets. In this case the limits for a target depend on the number of sequential zones being mapped: if this number of zone is larger than the limits, then the limits of the device apply and can be used. If on the other hand the target maps a number of zones smaller than the limits, then no limits is needed and we can assume that the target has no limits (limits set to 0). This commit improves zone resource limits handling as described above by modifying dm_set_zones_restrictions() to iterate the targets of a mapped device to evaluate the max open and max active zone limits. This relies on an internal "stacking" of the limits of the target devices combined with a direct counting of the number of sequential zones mapped by the targets. 1) For a target mapping an entire zoned block device, the limits for the target are set to the limits of the device. 2) For a target partially mapping a zoned block device, the number of mapped sequential zones is used to determine the limits: if the target maps more sequential write required zones than the device limits, then the limits of the device are used as-is. If the number of mapped sequential zones is lower than the limits, then we assume that the target has no limits (limits set to 0). As this evaluation is done for each target, the zone resource limits for the mapped device are evaluated as the non-zero minimum of the limits of all the targets. For configurations resulting in unreliable limits, i.e. a table containing a target partially mapping a zoned device, a warning message is issued. The counting of mapped sequential zones for the target is done using the new function dm_device_count_zones() which performs a report zones on the entire block device with the callback dm_device_count_zones_cb(). This count of mapped sequential zones is also used to determine if the mapped device contains only conventional zones. This allows simplifying dm_set_zones_restrictions() to not do a report zones just for this. For mapped devices mapping only conventional zones, as before, the mapped device is changed to a regular device by setting its zoned limit to false and clearing all its zone related limits. Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Benjamin Marzinski Reviewed-by: Niklas Cassel Link: https://lore.kernel.org/r/20240611023639.89277-4-dlemoal@kernel.org Signed-off-by: Jens Axboe --- drivers/md/dm-zone.c | 180 +++++++++++++++++++++++++++++++++++-------- 1 file changed, 150 insertions(+), 30 deletions(-) diff --git a/drivers/md/dm-zone.c b/drivers/md/dm-zone.c index 75d0019a0649d4..ac9f1f82108bf4 100644 --- a/drivers/md/dm-zone.c +++ b/drivers/md/dm-zone.c @@ -145,21 +145,6 @@ bool dm_is_zone_write(struct mapped_device *md, struct bio *bio) } } -/* - * Count conventional zones of a mapped zoned device. If the device - * only has conventional zones, do not expose it as zoned. - */ -static int dm_check_zoned_cb(struct blk_zone *zone, unsigned int idx, - void *data) -{ - unsigned int *nr_conv_zones = data; - - if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) - (*nr_conv_zones)++; - - return 0; -} - /* * Revalidate the zones of a mapped device to initialize resource necessary * for zone append emulation. Note that we cannot simply use the block layer @@ -228,13 +213,127 @@ static bool dm_table_supports_zone_append(struct dm_table *t) return true; } +struct dm_device_zone_count { + sector_t start; + sector_t len; + unsigned int total_nr_seq_zones; + unsigned int target_nr_seq_zones; +}; + +/* + * Count the total number of and the number of mapped sequential zones of a + * target zoned device. + */ +static int dm_device_count_zones_cb(struct blk_zone *zone, + unsigned int idx, void *data) +{ + struct dm_device_zone_count *zc = data; + + if (zone->type != BLK_ZONE_TYPE_CONVENTIONAL) { + zc->total_nr_seq_zones++; + if (zone->start >= zc->start && + zone->start < zc->start + zc->len) + zc->target_nr_seq_zones++; + } + + return 0; +} + +static int dm_device_count_zones(struct dm_dev *dev, + struct dm_device_zone_count *zc) +{ + int ret; + + ret = blkdev_report_zones(dev->bdev, 0, BLK_ALL_ZONES, + dm_device_count_zones_cb, zc); + if (ret < 0) + return ret; + if (!ret) + return -EIO; + return 0; +} + +struct dm_zone_resource_limits { + unsigned int mapped_nr_seq_zones; + struct queue_limits *lim; + bool reliable_limits; +}; + +static int device_get_zone_resource_limits(struct dm_target *ti, + struct dm_dev *dev, sector_t start, + sector_t len, void *data) +{ + struct dm_zone_resource_limits *zlim = data; + struct gendisk *disk = dev->bdev->bd_disk; + unsigned int max_open_zones, max_active_zones; + int ret; + struct dm_device_zone_count zc = { + .start = start, + .len = len, + }; + + /* + * If the target is not the whole device, the device zone resources may + * be shared between different targets. Check this by counting the + * number of mapped sequential zones: if this number is smaller than the + * total number of sequential zones of the target device, then resource + * sharing may happen and the zone limits will not be reliable. + */ + ret = dm_device_count_zones(dev, &zc); + if (ret) { + DMERR("Count %s zones failed %d", disk->disk_name, ret); + return ret; + } + + /* + * If the target does not map any sequential zones, then we do not need + * any zone resource limits. + */ + if (!zc.target_nr_seq_zones) + return 0; + + /* + * If the target does not map all sequential zones, the limits + * will not be reliable. + */ + if (zc.target_nr_seq_zones < zc.total_nr_seq_zones) + zlim->reliable_limits = false; + + /* + * If the target maps less sequential zones than the limit values, then + * we do not have limits for this target. + */ + max_active_zones = disk->queue->limits.max_active_zones; + if (max_active_zones >= zc.target_nr_seq_zones) + max_active_zones = 0; + zlim->lim->max_active_zones = + min_not_zero(max_active_zones, zlim->lim->max_active_zones); + + max_open_zones = disk->queue->limits.max_open_zones; + if (max_open_zones >= zc.target_nr_seq_zones) + max_open_zones = 0; + zlim->lim->max_open_zones = + min_not_zero(max_open_zones, zlim->lim->max_open_zones); + + /* + * Also count the total number of sequential zones for the mapped + * device so that when we are done inspecting all its targets, we are + * able to check if the mapped device actually has any sequential zones. + */ + zlim->mapped_nr_seq_zones += zc.target_nr_seq_zones; + + return 0; +} + int dm_set_zones_restrictions(struct dm_table *t, struct request_queue *q, struct queue_limits *lim) { struct mapped_device *md = t->md; struct gendisk *disk = md->disk; - unsigned int nr_conv_zones = 0; - int ret; + struct dm_zone_resource_limits zlim = { + .reliable_limits = true, + .lim = lim, + }; /* * Check if zone append is natively supported, and if not, set the @@ -249,32 +348,53 @@ int dm_set_zones_restrictions(struct dm_table *t, struct request_queue *q, } /* - * Count conventional zones to check that the mapped device will indeed - * have sequential write required zones. + * Determine the max open and max active zone limits for the mapped + * device by inspecting the zone resource limits and the zones mapped + * by each target. */ - md->zone_revalidate_map = t; - ret = dm_blk_report_zones(disk, 0, UINT_MAX, - dm_check_zoned_cb, &nr_conv_zones); - md->zone_revalidate_map = NULL; - if (ret < 0) { - DMERR("Check zoned failed %d", ret); - return ret; + for (unsigned int i = 0; i < t->num_targets; i++) { + struct dm_target *ti = dm_table_get_target(t, i); + + if (!ti->type->iterate_devices || + ti->type->iterate_devices(ti, + device_get_zone_resource_limits, &zlim)) { + DMERR("Could not determine %s zone resource limits", + disk->disk_name); + return -ENODEV; + } } /* - * If we only have conventional zones, expose the mapped device as - * a regular device. + * If we only have conventional zones mapped, expose the mapped device + + as a regular device. */ - if (nr_conv_zones >= ret) { + if (!zlim.mapped_nr_seq_zones) { lim->max_open_zones = 0; lim->max_active_zones = 0; + lim->max_zone_append_sectors = 0; + lim->zone_write_granularity = 0; + lim->chunk_sectors = 0; lim->zoned = false; clear_bit(DMF_EMULATE_ZONE_APPEND, &md->flags); + md->nr_zones = 0; disk->nr_zones = 0; return 0; } - if (!static_key_enabled(&zoned_enabled.key)) + /* + * Warn once (when the capacity is not yet set) if the mapped device is + * partially using zone resources of the target devices as that leads to + * unreliable limits, i.e. if another mapped device uses the same + * underlying devices, we cannot enforce zone limits to guarantee that + * writing will not lead to errors. Note that we really should return + * an error for such case but there is no easy way to find out if + * another mapped device uses the same underlying zoned devices. + */ + if (!get_capacity(disk) && !zlim.reliable_limits) + DMWARN("%s zone resource limits may be unreliable", + disk->disk_name); + + if (lim->zoned && !static_key_enabled(&zoned_enabled.key)) static_branch_enable(&zoned_enabled); return 0; } From eaa3706fedc6a4142c251b2d4005d850caeabe50 Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Tue, 11 Jun 2024 11:36:39 +0900 Subject: [PATCH 059/209] dm: Remove unused macro DM_ZONE_INVALID_WP_OFST With the switch to using the zone append emulation of the block layer zone write plugging, the macro DM_ZONE_INVALID_WP_OFST is no longer used in dm-zone.c. Remove its definition. Fixes: f211268ed1f9 ("dm: Use the block layer zone append emulation") Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Reviewed-by: Benjamin Marzinski Reviewed-by: Niklas Cassel Link: https://lore.kernel.org/r/20240611023639.89277-5-dlemoal@kernel.org Signed-off-by: Jens Axboe --- drivers/md/dm-zone.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/drivers/md/dm-zone.c b/drivers/md/dm-zone.c index ac9f1f82108bf4..70719bf32a2e5e 100644 --- a/drivers/md/dm-zone.c +++ b/drivers/md/dm-zone.c @@ -13,8 +13,6 @@ #define DM_MSG_PREFIX "zone" -#define DM_ZONE_INVALID_WP_OFST UINT_MAX - /* * For internal zone reports bypassing the top BIO submission path. */ From 81cc927d9c5eefd4a1b08e16b0ab2263f36d03f7 Mon Sep 17 00:00:00 2001 From: Gabriel Krisman Bertazi Date: Thu, 23 May 2024 17:45:17 -0400 Subject: [PATCH 060/209] io_uring: Drop per-ctx dummy_ubuf Commit 19a63c402170 ("io_uring/rsrc: keep one global dummy_ubuf") replaced it with a global static object but this stayed behind. Fixes: 19a63c402170 ("io_uring/rsrc: keep one global dummy_ubuf") Signed-off-by: Gabriel Krisman Bertazi Link: https://lore.kernel.org/r/20240523214517.31803-1-krisman@suse.de Signed-off-by: Jens Axboe --- include/linux/io_uring_types.h | 1 - 1 file changed, 1 deletion(-) diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h index b48570eaa4497f..93c9044ec3fef3 100644 --- a/include/linux/io_uring_types.h +++ b/include/linux/io_uring_types.h @@ -373,7 +373,6 @@ struct io_ring_ctx { struct io_restriction restrictions; /* slow path rsrc auxilary data, used by update/register */ - struct io_mapped_ubuf *dummy_ubuf; struct io_rsrc_data *file_data; struct io_rsrc_data *buf_data; From f4eaf8eda89e1ae5d8274297094687245293deff Mon Sep 17 00:00:00 2001 From: Gabriel Krisman Bertazi Date: Thu, 23 May 2024 17:45:35 -0400 Subject: [PATCH 061/209] io_uring/rsrc: Drop io_copy_iov in favor of iovec API Instead of open coding an io_uring function to copy iovs from userspace, rely on the existing iovec_from_user function. While there, avoid repeatedly zeroing the iov in the !arg case for io_sqe_buffer_register. tested with liburing testsuite. Signed-off-by: Gabriel Krisman Bertazi Link: https://lore.kernel.org/r/20240523214535.31890-1-krisman@suse.de Signed-off-by: Jens Axboe --- io_uring/rsrc.c | 60 +++++++++++++++++-------------------------------- 1 file changed, 21 insertions(+), 39 deletions(-) diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c index edb9c5baf2e290..e89c5e2326a257 100644 --- a/io_uring/rsrc.c +++ b/io_uring/rsrc.c @@ -85,31 +85,6 @@ static int io_account_mem(struct io_ring_ctx *ctx, unsigned long nr_pages) return 0; } -static int io_copy_iov(struct io_ring_ctx *ctx, struct iovec *dst, - void __user *arg, unsigned index) -{ - struct iovec __user *src; - -#ifdef CONFIG_COMPAT - if (ctx->compat) { - struct compat_iovec __user *ciovs; - struct compat_iovec ciov; - - ciovs = (struct compat_iovec __user *) arg; - if (copy_from_user(&ciov, &ciovs[index], sizeof(ciov))) - return -EFAULT; - - dst->iov_base = u64_to_user_ptr((u64)ciov.iov_base); - dst->iov_len = ciov.iov_len; - return 0; - } -#endif - src = (struct iovec __user *) arg; - if (copy_from_user(dst, &src[index], sizeof(*dst))) - return -EFAULT; - return 0; -} - static int io_buffer_validate(struct iovec *iov) { unsigned long tmp, acct_len = iov->iov_len + (PAGE_SIZE - 1); @@ -420,8 +395,9 @@ static int __io_sqe_buffers_update(struct io_ring_ctx *ctx, struct io_uring_rsrc_update2 *up, unsigned int nr_args) { + struct iovec __user *uvec = u64_to_user_ptr(up->data); u64 __user *tags = u64_to_user_ptr(up->tags); - struct iovec iov, __user *iovs = u64_to_user_ptr(up->data); + struct iovec fast_iov, *iov; struct page *last_hpage = NULL; __u32 done; int i, err; @@ -435,21 +411,23 @@ static int __io_sqe_buffers_update(struct io_ring_ctx *ctx, struct io_mapped_ubuf *imu; u64 tag = 0; - err = io_copy_iov(ctx, &iov, iovs, done); - if (err) + iov = iovec_from_user(&uvec[done], 1, 1, &fast_iov, ctx->compat); + if (IS_ERR(iov)) { + err = PTR_ERR(iov); break; + } if (tags && copy_from_user(&tag, &tags[done], sizeof(tag))) { err = -EFAULT; break; } - err = io_buffer_validate(&iov); + err = io_buffer_validate(iov); if (err) break; - if (!iov.iov_base && tag) { + if (!iov->iov_base && tag) { err = -EINVAL; break; } - err = io_sqe_buffer_register(ctx, &iov, &imu, &last_hpage); + err = io_sqe_buffer_register(ctx, iov, &imu, &last_hpage); if (err) break; @@ -971,8 +949,9 @@ int io_sqe_buffers_register(struct io_ring_ctx *ctx, void __user *arg, { struct page *last_hpage = NULL; struct io_rsrc_data *data; + struct iovec fast_iov, *iov = &fast_iov; + const struct iovec __user *uvec = (struct iovec * __user) arg; int i, ret; - struct iovec iov; BUILD_BUG_ON(IORING_MAX_REG_BUFFERS >= (1u << 16)); @@ -989,24 +968,27 @@ int io_sqe_buffers_register(struct io_ring_ctx *ctx, void __user *arg, return ret; } + if (!arg) + memset(iov, 0, sizeof(*iov)); + for (i = 0; i < nr_args; i++, ctx->nr_user_bufs++) { if (arg) { - ret = io_copy_iov(ctx, &iov, arg, i); - if (ret) + iov = iovec_from_user(&uvec[i], 1, 1, &fast_iov, ctx->compat); + if (IS_ERR(iov)) { + ret = PTR_ERR(iov); break; - ret = io_buffer_validate(&iov); + } + ret = io_buffer_validate(iov); if (ret) break; - } else { - memset(&iov, 0, sizeof(iov)); } - if (!iov.iov_base && *io_get_tag_slot(data, i)) { + if (!iov->iov_base && *io_get_tag_slot(data, i)) { ret = -EINVAL; break; } - ret = io_sqe_buffer_register(ctx, &iov, &ctx->user_bufs[i], + ret = io_sqe_buffer_register(ctx, iov, &ctx->user_bufs[i], &last_hpage); if (ret) break; From 60b6c075e8eb8bd23c106e2ab13370a146a94a5b Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 3 Jun 2024 11:19:10 -0600 Subject: [PATCH 062/209] io_uring/eventfd: move to more idiomatic RCU free usage In some ways, it just "happens to work" currently with using the ops field for both the free and signaling bit. But it depends on ordering of operations in terms of freeing and signaling. Clean it up and use the usual refs == 0 under RCU read side lock to determine if the ev_fd is still valid, and use the reference to gate the freeing as well. Fixes: 21a091b970cd ("io_uring: signal registered eventfd to process deferred task work") Signed-off-by: Jens Axboe --- io_uring/io_uring.c | 49 ++++++++++++++++++++++++--------------------- io_uring/io_uring.h | 4 ++-- io_uring/register.c | 6 +++--- 3 files changed, 31 insertions(+), 28 deletions(-) diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 154b25b8a613b1..0a24feec27f7b9 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -541,29 +541,33 @@ static __cold void io_queue_deferred(struct io_ring_ctx *ctx) } } -void io_eventfd_ops(struct rcu_head *rcu) +void io_eventfd_free(struct rcu_head *rcu) { struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); - int ops = atomic_xchg(&ev_fd->ops, 0); - if (ops & BIT(IO_EVENTFD_OP_SIGNAL_BIT)) - eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); + eventfd_ctx_put(ev_fd->cq_ev_fd); + kfree(ev_fd); +} - /* IO_EVENTFD_OP_FREE_BIT may not be set here depending on callback - * ordering in a race but if references are 0 we know we have to free - * it regardless. - */ - if (atomic_dec_and_test(&ev_fd->refs)) { - eventfd_ctx_put(ev_fd->cq_ev_fd); - kfree(ev_fd); - } +void io_eventfd_do_signal(struct rcu_head *rcu) +{ + struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); + + eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); + + if (atomic_dec_and_test(&ev_fd->refs)) + io_eventfd_free(rcu); } static void io_eventfd_signal(struct io_ring_ctx *ctx) { struct io_ev_fd *ev_fd = NULL; - rcu_read_lock(); + if (READ_ONCE(ctx->rings->cq_flags) & IORING_CQ_EVENTFD_DISABLED) + return; + + guard(rcu)(); + /* * rcu_dereference ctx->io_ev_fd once and use it for both for checking * and eventfd_signal @@ -576,24 +580,23 @@ static void io_eventfd_signal(struct io_ring_ctx *ctx) * the function and rcu_read_lock. */ if (unlikely(!ev_fd)) - goto out; - if (READ_ONCE(ctx->rings->cq_flags) & IORING_CQ_EVENTFD_DISABLED) - goto out; + return; + if (!atomic_inc_not_zero(&ev_fd->refs)) + return; if (ev_fd->eventfd_async && !io_wq_current_is_worker()) goto out; if (likely(eventfd_signal_allowed())) { eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); } else { - atomic_inc(&ev_fd->refs); - if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_SIGNAL_BIT), &ev_fd->ops)) - call_rcu_hurry(&ev_fd->rcu, io_eventfd_ops); - else - atomic_dec(&ev_fd->refs); + if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_SIGNAL_BIT), &ev_fd->ops)) { + call_rcu_hurry(&ev_fd->rcu, io_eventfd_do_signal); + return; + } } - out: - rcu_read_unlock(); + if (atomic_dec_and_test(&ev_fd->refs)) + call_rcu(&ev_fd->rcu, io_eventfd_free); } static void io_eventfd_flush_signal(struct io_ring_ctx *ctx) diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h index 726e6367af4d37..2b08b402b716d8 100644 --- a/io_uring/io_uring.h +++ b/io_uring/io_uring.h @@ -106,10 +106,10 @@ bool io_match_task_safe(struct io_kiocb *head, struct task_struct *task, enum { IO_EVENTFD_OP_SIGNAL_BIT, - IO_EVENTFD_OP_FREE_BIT, }; -void io_eventfd_ops(struct rcu_head *rcu); +void io_eventfd_do_signal(struct rcu_head *rcu); +void io_eventfd_free(struct rcu_head *rcu); void io_activate_pollwq(struct io_ring_ctx *ctx); static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx) diff --git a/io_uring/register.c b/io_uring/register.c index c0010a66a6f2c2..212711e9bc8a67 100644 --- a/io_uring/register.c +++ b/io_uring/register.c @@ -63,9 +63,9 @@ static int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg, ev_fd->eventfd_async = eventfd_async; ctx->has_evfd = true; - rcu_assign_pointer(ctx->io_ev_fd, ev_fd); atomic_set(&ev_fd->refs, 1); atomic_set(&ev_fd->ops, 0); + rcu_assign_pointer(ctx->io_ev_fd, ev_fd); return 0; } @@ -78,8 +78,8 @@ int io_eventfd_unregister(struct io_ring_ctx *ctx) if (ev_fd) { ctx->has_evfd = false; rcu_assign_pointer(ctx->io_ev_fd, NULL); - if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_FREE_BIT), &ev_fd->ops)) - call_rcu(&ev_fd->rcu, io_eventfd_ops); + if (atomic_dec_and_test(&ev_fd->refs)) + call_rcu(&ev_fd->rcu, io_eventfd_free); return 0; } From 200f3abd14db55f9aadcb74f4e7a678f1c469ba1 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 3 Jun 2024 11:51:19 -0600 Subject: [PATCH 063/209] io_uring/eventfd: move eventfd handling to separate file This is pretty nicely abstracted already, but let's move it to a separate file rather than have it in the main io_uring file. With that, we can also move the io_ev_fd struct and enum out of global scope. Signed-off-by: Jens Axboe --- include/linux/io_uring_types.h | 8 -- io_uring/Makefile | 6 +- io_uring/eventfd.c | 160 +++++++++++++++++++++++++++++++++ io_uring/eventfd.h | 8 ++ io_uring/io_uring.c | 82 +---------------- io_uring/io_uring.h | 6 -- io_uring/register.c | 56 +----------- 7 files changed, 173 insertions(+), 153 deletions(-) create mode 100644 io_uring/eventfd.c create mode 100644 io_uring/eventfd.h diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h index 93c9044ec3fef3..850e30be932256 100644 --- a/include/linux/io_uring_types.h +++ b/include/linux/io_uring_types.h @@ -211,14 +211,6 @@ struct io_submit_state { struct blk_plug plug; }; -struct io_ev_fd { - struct eventfd_ctx *cq_ev_fd; - unsigned int eventfd_async: 1; - struct rcu_head rcu; - atomic_t refs; - atomic_t ops; -}; - struct io_alloc_cache { void **entries; unsigned int nr_cached; diff --git a/io_uring/Makefile b/io_uring/Makefile index fc1b23c524e83b..61923e11c76772 100644 --- a/io_uring/Makefile +++ b/io_uring/Makefile @@ -4,9 +4,9 @@ obj-$(CONFIG_IO_URING) += io_uring.o opdef.o kbuf.o rsrc.o notif.o \ tctx.o filetable.o rw.o net.o poll.o \ - uring_cmd.o openclose.o sqpoll.o \ - xattr.o nop.o fs.o splice.o sync.o \ - msg_ring.o advise.o openclose.o \ + eventfd.o uring_cmd.o openclose.o \ + sqpoll.o xattr.o nop.o fs.o splice.o \ + sync.o msg_ring.o advise.o openclose.o \ epoll.o statx.o timeout.o fdinfo.o \ cancel.o waitid.o register.o \ truncate.o memmap.o diff --git a/io_uring/eventfd.c b/io_uring/eventfd.c new file mode 100644 index 00000000000000..b9384503a2b752 --- /dev/null +++ b/io_uring/eventfd.c @@ -0,0 +1,160 @@ +// SPDX-License-Identifier: GPL-2.0 +#include +#include +#include +#include +#include +#include +#include +#include + +#include "io-wq.h" +#include "eventfd.h" + +struct io_ev_fd { + struct eventfd_ctx *cq_ev_fd; + unsigned int eventfd_async: 1; + struct rcu_head rcu; + atomic_t refs; + atomic_t ops; +}; + +enum { + IO_EVENTFD_OP_SIGNAL_BIT, +}; + +static void io_eventfd_free(struct rcu_head *rcu) +{ + struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); + + eventfd_ctx_put(ev_fd->cq_ev_fd); + kfree(ev_fd); +} + +static void io_eventfd_do_signal(struct rcu_head *rcu) +{ + struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); + + eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); + + if (atomic_dec_and_test(&ev_fd->refs)) + io_eventfd_free(rcu); +} + +void io_eventfd_signal(struct io_ring_ctx *ctx) +{ + struct io_ev_fd *ev_fd = NULL; + + if (READ_ONCE(ctx->rings->cq_flags) & IORING_CQ_EVENTFD_DISABLED) + return; + + guard(rcu)(); + + /* + * rcu_dereference ctx->io_ev_fd once and use it for both for checking + * and eventfd_signal + */ + ev_fd = rcu_dereference(ctx->io_ev_fd); + + /* + * Check again if ev_fd exists incase an io_eventfd_unregister call + * completed between the NULL check of ctx->io_ev_fd at the start of + * the function and rcu_read_lock. + */ + if (unlikely(!ev_fd)) + return; + if (!atomic_inc_not_zero(&ev_fd->refs)) + return; + if (ev_fd->eventfd_async && !io_wq_current_is_worker()) + goto out; + + if (likely(eventfd_signal_allowed())) { + eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); + } else { + if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_SIGNAL_BIT), &ev_fd->ops)) { + call_rcu_hurry(&ev_fd->rcu, io_eventfd_do_signal); + return; + } + } +out: + if (atomic_dec_and_test(&ev_fd->refs)) + call_rcu(&ev_fd->rcu, io_eventfd_free); +} + +void io_eventfd_flush_signal(struct io_ring_ctx *ctx) +{ + bool skip; + + spin_lock(&ctx->completion_lock); + + /* + * Eventfd should only get triggered when at least one event has been + * posted. Some applications rely on the eventfd notification count + * only changing IFF a new CQE has been added to the CQ ring. There's + * no depedency on 1:1 relationship between how many times this + * function is called (and hence the eventfd count) and number of CQEs + * posted to the CQ ring. + */ + skip = ctx->cached_cq_tail == ctx->evfd_last_cq_tail; + ctx->evfd_last_cq_tail = ctx->cached_cq_tail; + spin_unlock(&ctx->completion_lock); + if (skip) + return; + + io_eventfd_signal(ctx); +} + +int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg, + unsigned int eventfd_async) +{ + struct io_ev_fd *ev_fd; + __s32 __user *fds = arg; + int fd; + + ev_fd = rcu_dereference_protected(ctx->io_ev_fd, + lockdep_is_held(&ctx->uring_lock)); + if (ev_fd) + return -EBUSY; + + if (copy_from_user(&fd, fds, sizeof(*fds))) + return -EFAULT; + + ev_fd = kmalloc(sizeof(*ev_fd), GFP_KERNEL); + if (!ev_fd) + return -ENOMEM; + + ev_fd->cq_ev_fd = eventfd_ctx_fdget(fd); + if (IS_ERR(ev_fd->cq_ev_fd)) { + int ret = PTR_ERR(ev_fd->cq_ev_fd); + kfree(ev_fd); + return ret; + } + + spin_lock(&ctx->completion_lock); + ctx->evfd_last_cq_tail = ctx->cached_cq_tail; + spin_unlock(&ctx->completion_lock); + + ev_fd->eventfd_async = eventfd_async; + ctx->has_evfd = true; + atomic_set(&ev_fd->refs, 1); + atomic_set(&ev_fd->ops, 0); + rcu_assign_pointer(ctx->io_ev_fd, ev_fd); + return 0; +} + +int io_eventfd_unregister(struct io_ring_ctx *ctx) +{ + struct io_ev_fd *ev_fd; + + ev_fd = rcu_dereference_protected(ctx->io_ev_fd, + lockdep_is_held(&ctx->uring_lock)); + if (ev_fd) { + ctx->has_evfd = false; + rcu_assign_pointer(ctx->io_ev_fd, NULL); + if (atomic_dec_and_test(&ev_fd->refs)) + call_rcu(&ev_fd->rcu, io_eventfd_free); + return 0; + } + + return -ENXIO; +} diff --git a/io_uring/eventfd.h b/io_uring/eventfd.h new file mode 100644 index 00000000000000..d394f49c632105 --- /dev/null +++ b/io_uring/eventfd.h @@ -0,0 +1,8 @@ + +struct io_ring_ctx; +int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg, + unsigned int eventfd_async); +int io_eventfd_unregister(struct io_ring_ctx *ctx); + +void io_eventfd_flush_signal(struct io_ring_ctx *ctx); +void io_eventfd_signal(struct io_ring_ctx *ctx); diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 0a24feec27f7b9..d10678b9d519cf 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -101,6 +101,7 @@ #include "poll.h" #include "rw.h" #include "alloc_cache.h" +#include "eventfd.h" #define IORING_MAX_ENTRIES 32768 #define IORING_MAX_CQ_ENTRIES (2 * IORING_MAX_ENTRIES) @@ -541,87 +542,6 @@ static __cold void io_queue_deferred(struct io_ring_ctx *ctx) } } -void io_eventfd_free(struct rcu_head *rcu) -{ - struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); - - eventfd_ctx_put(ev_fd->cq_ev_fd); - kfree(ev_fd); -} - -void io_eventfd_do_signal(struct rcu_head *rcu) -{ - struct io_ev_fd *ev_fd = container_of(rcu, struct io_ev_fd, rcu); - - eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); - - if (atomic_dec_and_test(&ev_fd->refs)) - io_eventfd_free(rcu); -} - -static void io_eventfd_signal(struct io_ring_ctx *ctx) -{ - struct io_ev_fd *ev_fd = NULL; - - if (READ_ONCE(ctx->rings->cq_flags) & IORING_CQ_EVENTFD_DISABLED) - return; - - guard(rcu)(); - - /* - * rcu_dereference ctx->io_ev_fd once and use it for both for checking - * and eventfd_signal - */ - ev_fd = rcu_dereference(ctx->io_ev_fd); - - /* - * Check again if ev_fd exists incase an io_eventfd_unregister call - * completed between the NULL check of ctx->io_ev_fd at the start of - * the function and rcu_read_lock. - */ - if (unlikely(!ev_fd)) - return; - if (!atomic_inc_not_zero(&ev_fd->refs)) - return; - if (ev_fd->eventfd_async && !io_wq_current_is_worker()) - goto out; - - if (likely(eventfd_signal_allowed())) { - eventfd_signal_mask(ev_fd->cq_ev_fd, EPOLL_URING_WAKE); - } else { - if (!atomic_fetch_or(BIT(IO_EVENTFD_OP_SIGNAL_BIT), &ev_fd->ops)) { - call_rcu_hurry(&ev_fd->rcu, io_eventfd_do_signal); - return; - } - } -out: - if (atomic_dec_and_test(&ev_fd->refs)) - call_rcu(&ev_fd->rcu, io_eventfd_free); -} - -static void io_eventfd_flush_signal(struct io_ring_ctx *ctx) -{ - bool skip; - - spin_lock(&ctx->completion_lock); - - /* - * Eventfd should only get triggered when at least one event has been - * posted. Some applications rely on the eventfd notification count - * only changing IFF a new CQE has been added to the CQ ring. There's - * no depedency on 1:1 relationship between how many times this - * function is called (and hence the eventfd count) and number of CQEs - * posted to the CQ ring. - */ - skip = ctx->cached_cq_tail == ctx->evfd_last_cq_tail; - ctx->evfd_last_cq_tail = ctx->cached_cq_tail; - spin_unlock(&ctx->completion_lock); - if (skip) - return; - - io_eventfd_signal(ctx); -} - void __io_commit_cqring_flush(struct io_ring_ctx *ctx) { if (ctx->poll_activated) diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h index 2b08b402b716d8..cd43924eed04e1 100644 --- a/io_uring/io_uring.h +++ b/io_uring/io_uring.h @@ -104,12 +104,6 @@ bool __io_alloc_req_refill(struct io_ring_ctx *ctx); bool io_match_task_safe(struct io_kiocb *head, struct task_struct *task, bool cancel_all); -enum { - IO_EVENTFD_OP_SIGNAL_BIT, -}; - -void io_eventfd_do_signal(struct rcu_head *rcu); -void io_eventfd_free(struct rcu_head *rcu); void io_activate_pollwq(struct io_ring_ctx *ctx); static inline void io_lockdep_assert_cq_locked(struct io_ring_ctx *ctx) diff --git a/io_uring/register.c b/io_uring/register.c index 212711e9bc8a67..f121e02f5e10e6 100644 --- a/io_uring/register.c +++ b/io_uring/register.c @@ -27,65 +27,11 @@ #include "cancel.h" #include "kbuf.h" #include "napi.h" +#include "eventfd.h" #define IORING_MAX_RESTRICTIONS (IORING_RESTRICTION_LAST + \ IORING_REGISTER_LAST + IORING_OP_LAST) -static int io_eventfd_register(struct io_ring_ctx *ctx, void __user *arg, - unsigned int eventfd_async) -{ - struct io_ev_fd *ev_fd; - __s32 __user *fds = arg; - int fd; - - ev_fd = rcu_dereference_protected(ctx->io_ev_fd, - lockdep_is_held(&ctx->uring_lock)); - if (ev_fd) - return -EBUSY; - - if (copy_from_user(&fd, fds, sizeof(*fds))) - return -EFAULT; - - ev_fd = kmalloc(sizeof(*ev_fd), GFP_KERNEL); - if (!ev_fd) - return -ENOMEM; - - ev_fd->cq_ev_fd = eventfd_ctx_fdget(fd); - if (IS_ERR(ev_fd->cq_ev_fd)) { - int ret = PTR_ERR(ev_fd->cq_ev_fd); - kfree(ev_fd); - return ret; - } - - spin_lock(&ctx->completion_lock); - ctx->evfd_last_cq_tail = ctx->cached_cq_tail; - spin_unlock(&ctx->completion_lock); - - ev_fd->eventfd_async = eventfd_async; - ctx->has_evfd = true; - atomic_set(&ev_fd->refs, 1); - atomic_set(&ev_fd->ops, 0); - rcu_assign_pointer(ctx->io_ev_fd, ev_fd); - return 0; -} - -int io_eventfd_unregister(struct io_ring_ctx *ctx) -{ - struct io_ev_fd *ev_fd; - - ev_fd = rcu_dereference_protected(ctx->io_ev_fd, - lockdep_is_held(&ctx->uring_lock)); - if (ev_fd) { - ctx->has_evfd = false; - rcu_assign_pointer(ctx->io_ev_fd, NULL); - if (atomic_dec_and_test(&ev_fd->refs)) - call_rcu(&ev_fd->rcu, io_eventfd_free); - return 0; - } - - return -ENXIO; -} - static __cold int io_probe(struct io_ring_ctx *ctx, void __user *arg, unsigned nr_args) { From f2a93294edce87c909d61e18b506404127394891 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Fri, 14 Jun 2024 10:57:03 -0600 Subject: [PATCH 064/209] io_uring: use 'state' consistently __io_submit_flush_completions() assigns ctx->submit_state to a local variable and uses it in all but one spot, switch that forgotten statement to using 'state' as well. Signed-off-by: Jens Axboe --- io_uring/io_uring.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index d10678b9d519cf..57382e523b3367 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -1390,7 +1390,7 @@ void __io_submit_flush_completions(struct io_ring_ctx *ctx) } __io_cq_unlock_post(ctx); - if (!wq_list_empty(&ctx->submit_state.compl_reqs)) { + if (!wq_list_empty(&state->compl_reqs)) { io_free_batch_list(ctx, state->compl_reqs.first); INIT_WQ_LIST(&state->compl_reqs); } From 3474d1b93f897ab33ce160e759afd47d5f412de4 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 13 Jun 2024 19:28:27 +0000 Subject: [PATCH 065/209] io_uring/io-wq: make io_wq_work flags atomic The work flags can be set/accessed from different tasks, both the originator of the request, and the io-wq workers. While modifications aren't concurrent, it still makes KMSAN unhappy. There's no real downside to just making the flag reading/manipulation use proper atomics here. Signed-off-by: Jens Axboe --- include/linux/io_uring_types.h | 2 +- io_uring/io-wq.c | 19 ++++++++++--------- io_uring/io-wq.h | 2 +- io_uring/io_uring.c | 12 ++++++------ 4 files changed, 18 insertions(+), 17 deletions(-) diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h index 850e30be932256..1052a68fd68df4 100644 --- a/include/linux/io_uring_types.h +++ b/include/linux/io_uring_types.h @@ -50,7 +50,7 @@ struct io_wq_work_list { struct io_wq_work { struct io_wq_work_node list; - unsigned flags; + atomic_t flags; /* place it here instead of io_kiocb as it fills padding and saves 4B */ int cancel_seq; }; diff --git a/io_uring/io-wq.c b/io_uring/io-wq.c index 7d3316fe9bfc46..913c92249522e8 100644 --- a/io_uring/io-wq.c +++ b/io_uring/io-wq.c @@ -159,7 +159,7 @@ static inline struct io_wq_acct *io_get_acct(struct io_wq *wq, bool bound) static inline struct io_wq_acct *io_work_get_acct(struct io_wq *wq, struct io_wq_work *work) { - return io_get_acct(wq, !(work->flags & IO_WQ_WORK_UNBOUND)); + return io_get_acct(wq, !(atomic_read(&work->flags) & IO_WQ_WORK_UNBOUND)); } static inline struct io_wq_acct *io_wq_get_acct(struct io_worker *worker) @@ -451,7 +451,7 @@ static void __io_worker_idle(struct io_wq *wq, struct io_worker *worker) static inline unsigned int io_get_work_hash(struct io_wq_work *work) { - return work->flags >> IO_WQ_HASH_SHIFT; + return atomic_read(&work->flags) >> IO_WQ_HASH_SHIFT; } static bool io_wait_on_hash(struct io_wq *wq, unsigned int hash) @@ -592,8 +592,9 @@ static void io_worker_handle_work(struct io_wq_acct *acct, next_hashed = wq_next_work(work); - if (unlikely(do_kill) && (work->flags & IO_WQ_WORK_UNBOUND)) - work->flags |= IO_WQ_WORK_CANCEL; + if (do_kill && + (atomic_read(&work->flags) & IO_WQ_WORK_UNBOUND)) + atomic_or(IO_WQ_WORK_CANCEL, &work->flags); wq->do_work(work); io_assign_current_work(worker, NULL); @@ -891,7 +892,7 @@ static bool io_wq_worker_wake(struct io_worker *worker, void *data) static void io_run_cancel(struct io_wq_work *work, struct io_wq *wq) { do { - work->flags |= IO_WQ_WORK_CANCEL; + atomic_or(IO_WQ_WORK_CANCEL, &work->flags); wq->do_work(work); work = wq->free_work(work); } while (work); @@ -926,7 +927,7 @@ static bool io_wq_work_match_item(struct io_wq_work *work, void *data) void io_wq_enqueue(struct io_wq *wq, struct io_wq_work *work) { struct io_wq_acct *acct = io_work_get_acct(wq, work); - unsigned long work_flags = work->flags; + unsigned int work_flags = atomic_read(&work->flags); struct io_cb_cancel_data match = { .fn = io_wq_work_match_item, .data = work, @@ -939,7 +940,7 @@ void io_wq_enqueue(struct io_wq *wq, struct io_wq_work *work) * been marked as one that should not get executed, cancel it here. */ if (test_bit(IO_WQ_BIT_EXIT, &wq->state) || - (work->flags & IO_WQ_WORK_CANCEL)) { + (work_flags & IO_WQ_WORK_CANCEL)) { io_run_cancel(work, wq); return; } @@ -982,7 +983,7 @@ void io_wq_hash_work(struct io_wq_work *work, void *val) unsigned int bit; bit = hash_ptr(val, IO_WQ_HASH_ORDER); - work->flags |= (IO_WQ_WORK_HASHED | (bit << IO_WQ_HASH_SHIFT)); + atomic_or(IO_WQ_WORK_HASHED | (bit << IO_WQ_HASH_SHIFT), &work->flags); } static bool __io_wq_worker_cancel(struct io_worker *worker, @@ -990,7 +991,7 @@ static bool __io_wq_worker_cancel(struct io_worker *worker, struct io_wq_work *work) { if (work && match->fn(work, match->data)) { - work->flags |= IO_WQ_WORK_CANCEL; + atomic_or(IO_WQ_WORK_CANCEL, &work->flags); __set_notify_signal(worker->task); return true; } diff --git a/io_uring/io-wq.h b/io_uring/io-wq.h index 2b2a6406dd8ee8..b3b004a7b62528 100644 --- a/io_uring/io-wq.h +++ b/io_uring/io-wq.h @@ -56,7 +56,7 @@ bool io_wq_worker_stopped(void); static inline bool io_wq_is_hashed(struct io_wq_work *work) { - return work->flags & IO_WQ_WORK_HASHED; + return atomic_read(&work->flags) & IO_WQ_WORK_HASHED; } typedef bool (work_cancel_fn)(struct io_wq_work *, void *); diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 57382e523b3367..438c44ca3abd63 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -462,9 +462,9 @@ static void io_prep_async_work(struct io_kiocb *req) } req->work.list.next = NULL; - req->work.flags = 0; + atomic_set(&req->work.flags, 0); if (req->flags & REQ_F_FORCE_ASYNC) - req->work.flags |= IO_WQ_WORK_CONCURRENT; + atomic_or(IO_WQ_WORK_CONCURRENT, &req->work.flags); if (req->file && !(req->flags & REQ_F_FIXED_FILE)) req->flags |= io_file_get_flags(req->file); @@ -480,7 +480,7 @@ static void io_prep_async_work(struct io_kiocb *req) io_wq_hash_work(&req->work, file_inode(req->file)); } else if (!req->file || !S_ISBLK(file_inode(req->file)->i_mode)) { if (def->unbound_nonreg_file) - req->work.flags |= IO_WQ_WORK_UNBOUND; + atomic_or(IO_WQ_WORK_UNBOUND, &req->work.flags); } } @@ -520,7 +520,7 @@ static void io_queue_iowq(struct io_kiocb *req) * worker for it). */ if (WARN_ON_ONCE(!same_thread_group(req->task, current))) - req->work.flags |= IO_WQ_WORK_CANCEL; + atomic_or(IO_WQ_WORK_CANCEL, &req->work.flags); trace_io_uring_queue_async_work(req, io_wq_is_hashed(&req->work)); io_wq_enqueue(tctx->io_wq, &req->work); @@ -1736,14 +1736,14 @@ void io_wq_submit_work(struct io_wq_work *work) io_arm_ltimeout(req); /* either cancelled or io-wq is dying, so don't touch tctx->iowq */ - if (work->flags & IO_WQ_WORK_CANCEL) { + if (atomic_read(&work->flags) & IO_WQ_WORK_CANCEL) { fail: io_req_task_queue_fail(req, err); return; } if (!io_assign_file(req, def, issue_flags)) { err = -EBADF; - work->flags |= IO_WQ_WORK_CANCEL; + atomic_or(IO_WQ_WORK_CANCEL, &work->flags); goto fail; } From 11d194669271642a5d1bfff6c8011478309e7849 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Fri, 14 Jun 2024 18:34:50 -0600 Subject: [PATCH 066/209] io_uring/rsrc: remove redundant __set_current_state() post schedule() We're guaranteed to be in a TASK_RUNNING state post schedule, so we never need to set the state after that. While in there, remove the other __set_current_state() as well, and just call finish_wait() when we now we're going to break anyway. This is easier to grok than manual __set_current_state() calls. Reported-by: Linus Torvalds Signed-off-by: Jens Axboe --- io_uring/rsrc.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c index e89c5e2326a257..60c00144471a61 100644 --- a/io_uring/rsrc.c +++ b/io_uring/rsrc.c @@ -224,7 +224,7 @@ __cold static int io_rsrc_ref_quiesce(struct io_rsrc_data *data, ret = io_run_task_work_sig(ctx); if (ret < 0) { - __set_current_state(TASK_RUNNING); + finish_wait(&ctx->rsrc_quiesce_wq, &we); mutex_lock(&ctx->uring_lock); if (list_empty(&ctx->rsrc_ref_list)) ret = 0; @@ -232,7 +232,6 @@ __cold static int io_rsrc_ref_quiesce(struct io_rsrc_data *data, } schedule(); - __set_current_state(TASK_RUNNING); mutex_lock(&ctx->uring_lock); ret = 0; } while (!list_empty(&ctx->rsrc_ref_list)); From 3b87184f7eff27fef7d7ee18b65f173152e1bb81 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Sat, 15 Jun 2024 20:47:14 -0600 Subject: [PATCH 067/209] io_uring/advise: support 64-bit lengths The existing fadvise/madvise support only supports 32-bit lengths. Add support for 64-bit lengths, enabled by the application setting sqe->off rather than sqe->len for the length. If sqe->len is set, then that is used as the 32-bit length. If sqe->len is zero, then sqe->off is read for full 64-bit support. Older kernels will return -EINVAL if 64-bit support isn't available. Fixes: 4840e418c2fc ("io_uring: add IORING_OP_FADVISE") Fixes: c1ca757bd6f4 ("io_uring: add IORING_OP_MADVISE") Reported-by: Stefan Signed-off-by: Jens Axboe --- io_uring/advise.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) diff --git a/io_uring/advise.c b/io_uring/advise.c index 7085804c513c09..cb7b881665e57f 100644 --- a/io_uring/advise.c +++ b/io_uring/advise.c @@ -17,14 +17,14 @@ struct io_fadvise { struct file *file; u64 offset; - u32 len; + u64 len; u32 advice; }; struct io_madvise { struct file *file; u64 addr; - u32 len; + u64 len; u32 advice; }; @@ -33,11 +33,13 @@ int io_madvise_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) #if defined(CONFIG_ADVISE_SYSCALLS) && defined(CONFIG_MMU) struct io_madvise *ma = io_kiocb_to_cmd(req, struct io_madvise); - if (sqe->buf_index || sqe->off || sqe->splice_fd_in) + if (sqe->buf_index || sqe->splice_fd_in) return -EINVAL; ma->addr = READ_ONCE(sqe->addr); - ma->len = READ_ONCE(sqe->len); + ma->len = READ_ONCE(sqe->off); + if (!ma->len) + ma->len = READ_ONCE(sqe->len); ma->advice = READ_ONCE(sqe->fadvise_advice); req->flags |= REQ_F_FORCE_ASYNC; return 0; @@ -78,11 +80,13 @@ int io_fadvise_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) { struct io_fadvise *fa = io_kiocb_to_cmd(req, struct io_fadvise); - if (sqe->buf_index || sqe->addr || sqe->splice_fd_in) + if (sqe->buf_index || sqe->splice_fd_in) return -EINVAL; fa->offset = READ_ONCE(sqe->off); - fa->len = READ_ONCE(sqe->len); + fa->len = READ_ONCE(sqe->addr); + if (!fa->len) + fa->len = READ_ONCE(sqe->len); fa->advice = READ_ONCE(sqe->fadvise_advice); if (io_fadvise_force_async(fa)) req->flags |= REQ_F_FORCE_ASYNC; From d9c2332199d073c5edd7163d64fbdee6224d8c08 Mon Sep 17 00:00:00 2001 From: Jiapeng Chong Date: Fri, 14 Jun 2024 09:03:43 +0000 Subject: [PATCH 068/209] bdev: make blockdev_mnt static The blockdev_mnt are not used outside the file bdev.c, so the modification is defined as static. block/bdev.c:377:17: warning: symbol 'blockdev_mnt' was not declared. Should it be static? Reported-by: Abaci Robot jpg: Remove closes bugzilla link Reviewed-by: Christoph Hellwig Signed-off-by: Jiapeng Chong Signed-off-by: John Garry Reviewed-by: Bart Van Assche Fixes: 8f3a608827d1 ("bdev: open block device as files") Tested-by: John Garry Link: https://lore.kernel.org/r/20240614090345.655716-2-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- block/bdev.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/block/bdev.c b/block/bdev.c index 353677ac49b3b9..ced4ac990ec88f 100644 --- a/block/bdev.c +++ b/block/bdev.c @@ -385,7 +385,7 @@ static struct file_system_type bd_type = { }; struct super_block *blockdev_superblock __ro_after_init; -struct vfsmount *blockdev_mnt __ro_after_init; +static struct vfsmount *blockdev_mnt __ro_after_init; EXPORT_SYMBOL_GPL(blockdev_superblock); void __init bdev_cache_init(void) From c3042a5403ef2be622023fcc3b11fc1aa08ba7fa Mon Sep 17 00:00:00 2001 From: John Garry Date: Fri, 14 Jun 2024 09:03:44 +0000 Subject: [PATCH 069/209] block: Drop locking annotation for limits_lock Currently compiling block/blk-settings.c with C=1 gives the following warning: block/blk-settings.c:262:9: warning: context imbalance in 'queue_limits_commit_update' - wrong count at exit request_queue.limits_lock is a mutex. Sparse locking annotation for mutexes are currently not supported - see [0] - so drop that locking annotation. [0] https://lore.kernel.org/lkml/cover.1579893447.git.jbi.octave@gmail.com/T/#mbb8bda6c0a7ca7ce19f46df976a8e3b489745488 Fixes: d690cb8ae14bd ("block: add an API to atomically update queue limits") Signed-off-by: John Garry Reviewed-by: Bart Van Assche Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240614090345.655716-3-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- block/blk-settings.c | 1 - include/linux/blkdev.h | 1 - 2 files changed, 2 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index f574181105f8a5..377a86fe0fcc41 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -303,7 +303,6 @@ int blk_set_default_limits(struct queue_limits *lim) */ int queue_limits_commit_update(struct request_queue *q, struct queue_limits *lim) - __releases(q->limits_lock) { int error; diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 0c247a71688561..83af6ac5aa4819 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -893,7 +893,6 @@ static inline unsigned int blk_chunk_sectors_left(sector_t offset, */ static inline struct queue_limits queue_limits_start_update(struct request_queue *q) - __acquires(q->limits_lock) { mutex_lock(&q->limits_lock); return q->limits; From 66088084fdabb6e5075cd19e8ffe15b8bc7e3708 Mon Sep 17 00:00:00 2001 From: John Garry Date: Fri, 14 Jun 2024 09:03:45 +0000 Subject: [PATCH 070/209] block: BFQ: Refactor bfq_exit_icq() to silence sparse warning Currently building for C=1 generates the following warning: block/bfq-iosched.c:5498:9: warning: context imbalance in 'bfq_exit_icq' - different lock contexts for basic block Refactor bfq_exit_icq() into a core part which loops for the actuators, and only lock calling this routine when necessary. Signed-off-by: John Garry Reviewed-by: Bart Van Assche Link: https://lore.kernel.org/r/20240614090345.655716-4-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- block/bfq-iosched.c | 38 ++++++++++++++++++++------------------ 1 file changed, 20 insertions(+), 18 deletions(-) diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c index 4b88a54a9b76cb..36a4998c4b378b 100644 --- a/block/bfq-iosched.c +++ b/block/bfq-iosched.c @@ -5463,40 +5463,42 @@ static void bfq_exit_icq_bfqq(struct bfq_io_cq *bic, bool is_sync, } } +static void _bfq_exit_icq(struct bfq_io_cq *bic, unsigned int num_actuators) +{ + struct bfq_iocq_bfqq_data *bfqq_data = bic->bfqq_data; + unsigned int act_idx; + + for (act_idx = 0; act_idx < num_actuators; act_idx++) { + if (bfqq_data[act_idx].stable_merge_bfqq) + bfq_put_stable_ref(bfqq_data[act_idx].stable_merge_bfqq); + + bfq_exit_icq_bfqq(bic, true, act_idx); + bfq_exit_icq_bfqq(bic, false, act_idx); + } +} + static void bfq_exit_icq(struct io_cq *icq) { struct bfq_io_cq *bic = icq_to_bic(icq); struct bfq_data *bfqd = bic_to_bfqd(bic); unsigned long flags; - unsigned int act_idx; + /* * If bfqd and thus bfqd->num_actuators is not available any * longer, then cycle over all possible per-actuator bfqqs in * next loop. We rely on bic being zeroed on creation, and * therefore on its unused per-actuator fields being NULL. - */ - unsigned int num_actuators = BFQ_MAX_ACTUATORS; - struct bfq_iocq_bfqq_data *bfqq_data = bic->bfqq_data; - - /* + * * bfqd is NULL if scheduler already exited, and in that case * this is the last time these queues are accessed. */ if (bfqd) { spin_lock_irqsave(&bfqd->lock, flags); - num_actuators = bfqd->num_actuators; - } - - for (act_idx = 0; act_idx < num_actuators; act_idx++) { - if (bfqq_data[act_idx].stable_merge_bfqq) - bfq_put_stable_ref(bfqq_data[act_idx].stable_merge_bfqq); - - bfq_exit_icq_bfqq(bic, true, act_idx); - bfq_exit_icq_bfqq(bic, false, act_idx); - } - - if (bfqd) + _bfq_exit_icq(bic, bfqd->num_actuators); spin_unlock_irqrestore(&bfqd->lock, flags); + } else { + _bfq_exit_icq(bic, BFQ_MAX_ACTUATORS); + } } /* From b83bd486b43d2b7f10595a9d7a52d41023eaa9c1 Mon Sep 17 00:00:00 2001 From: Kanchan Joshi Date: Mon, 17 Jun 2024 10:19:18 +0530 Subject: [PATCH 071/209] block: cleanup flag_{show,store} Remove a superfluous argument that flag_show and flag_store currently take. Signed-off-by: Kanchan Joshi Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240617044918.374608-1-joshi.k@samsung.com Signed-off-by: Jens Axboe --- block/blk-integrity.c | 15 +++++++-------- 1 file changed, 7 insertions(+), 8 deletions(-) diff --git a/block/blk-integrity.c b/block/blk-integrity.c index 05a48689a424b2..010decc892eaa0 100644 --- a/block/blk-integrity.c +++ b/block/blk-integrity.c @@ -186,8 +186,8 @@ const char *blk_integrity_profile_name(struct blk_integrity *bi) } EXPORT_SYMBOL_GPL(blk_integrity_profile_name); -static ssize_t flag_store(struct device *dev, struct device_attribute *attr, - const char *page, size_t count, unsigned char flag) +static ssize_t flag_store(struct device *dev, const char *page, size_t count, + unsigned char flag) { struct request_queue *q = dev_to_disk(dev)->queue; struct queue_limits lim; @@ -213,8 +213,7 @@ static ssize_t flag_store(struct device *dev, struct device_attribute *attr, return count; } -static ssize_t flag_show(struct device *dev, struct device_attribute *attr, - char *page, unsigned char flag) +static ssize_t flag_show(struct device *dev, char *page, unsigned char flag) { struct blk_integrity *bi = dev_to_bi(dev); @@ -253,26 +252,26 @@ static ssize_t read_verify_store(struct device *dev, struct device_attribute *attr, const char *page, size_t count) { - return flag_store(dev, attr, page, count, BLK_INTEGRITY_NOVERIFY); + return flag_store(dev, page, count, BLK_INTEGRITY_NOVERIFY); } static ssize_t read_verify_show(struct device *dev, struct device_attribute *attr, char *page) { - return flag_show(dev, attr, page, BLK_INTEGRITY_NOVERIFY); + return flag_show(dev, page, BLK_INTEGRITY_NOVERIFY); } static ssize_t write_generate_store(struct device *dev, struct device_attribute *attr, const char *page, size_t count) { - return flag_store(dev, attr, page, count, BLK_INTEGRITY_NOGENERATE); + return flag_store(dev, page, count, BLK_INTEGRITY_NOGENERATE); } static ssize_t write_generate_show(struct device *dev, struct device_attribute *attr, char *page) { - return flag_show(dev, attr, page, BLK_INTEGRITY_NOGENERATE); + return flag_show(dev, page, BLK_INTEGRITY_NOGENERATE); } static ssize_t device_is_integrity_capable_show(struct device *dev, From bb7e5a193d8becf3920e3848287f1b23c5fc9b24 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 18 Jun 2024 11:27:53 +0800 Subject: [PATCH 072/209] block, bfq: remove blkg_path() After commit 35fe6d763229 ("block: use standard blktrace API to output cgroup info for debug notes"), the field 'bfqg->blkg_path' is not used and hence can be removed, and therefor blkg_path() is not used anymore and can be removed. Signed-off-by: Yu Kuai Reviewed-by: Jan Kara Link: https://lore.kernel.org/r/20240618032753.3502528-1-yukuai1@huaweicloud.com Signed-off-by: Jens Axboe --- block/bfq-cgroup.c | 51 --------------------------------------------- block/bfq-iosched.h | 3 --- block/blk-cgroup.h | 13 ------------ 3 files changed, 67 deletions(-) diff --git a/block/bfq-cgroup.c b/block/bfq-cgroup.c index d442ee358fc257..b758693697c093 100644 --- a/block/bfq-cgroup.c +++ b/block/bfq-cgroup.c @@ -797,57 +797,6 @@ void bfq_bic_update_cgroup(struct bfq_io_cq *bic, struct bio *bio) */ bfq_link_bfqg(bfqd, bfqg); __bfq_bic_change_cgroup(bfqd, bic, bfqg); - /* - * Update blkg_path for bfq_log_* functions. We cache this - * path, and update it here, for the following - * reasons. Operations on blkg objects in blk-cgroup are - * protected with the request_queue lock, and not with the - * lock that protects the instances of this scheduler - * (bfqd->lock). This exposes BFQ to the following sort of - * race. - * - * The blkg_lookup performed in bfq_get_queue, protected - * through rcu, may happen to return the address of a copy of - * the original blkg. If this is the case, then the - * bfqg_and_blkg_get performed in bfq_get_queue, to pin down - * the blkg, is useless: it does not prevent blk-cgroup code - * from destroying both the original blkg and all objects - * directly or indirectly referred by the copy of the - * blkg. - * - * On the bright side, destroy operations on a blkg invoke, as - * a first step, hooks of the scheduler associated with the - * blkg. And these hooks are executed with bfqd->lock held for - * BFQ. As a consequence, for any blkg associated with the - * request queue this instance of the scheduler is attached - * to, we are guaranteed that such a blkg is not destroyed, and - * that all the pointers it contains are consistent, while we - * are holding bfqd->lock. A blkg_lookup performed with - * bfqd->lock held then returns a fully consistent blkg, which - * remains consistent until this lock is held. - * - * Thanks to the last fact, and to the fact that: (1) bfqg has - * been obtained through a blkg_lookup in the above - * assignment, and (2) bfqd->lock is being held, here we can - * safely use the policy data for the involved blkg (i.e., the - * field bfqg->pd) to get to the blkg associated with bfqg, - * and then we can safely use any field of blkg. After we - * release bfqd->lock, even just getting blkg through this - * bfqg may cause dangling references to be traversed, as - * bfqg->pd may not exist any more. - * - * In view of the above facts, here we cache, in the bfqg, any - * blkg data we may need for this bic, and for its associated - * bfq_queue. As of now, we need to cache only the path of the - * blkg, which is used in the bfq_log_* functions. - * - * Finally, note that bfqg itself needs to be protected from - * destruction on the blkg_free of the original blkg (which - * invokes bfq_pd_free). We use an additional private - * refcounter for bfqg, to let it disappear only after no - * bfq_queue refers to it any longer. - */ - blkg_path(bfqg_to_blkg(bfqg), bfqg->blkg_path, sizeof(bfqg->blkg_path)); bic->blkcg_serial_nr = serial_nr; } diff --git a/block/bfq-iosched.h b/block/bfq-iosched.h index 467e8cfc41a249..08ddf2cfae5b1c 100644 --- a/block/bfq-iosched.h +++ b/block/bfq-iosched.h @@ -1003,9 +1003,6 @@ struct bfq_group { /* must be the first member */ struct blkg_policy_data pd; - /* cached path for this blkg (see comments in bfq_bic_update_cgroup) */ - char blkg_path[128]; - /* reference counter (see comments in bfq_bic_update_cgroup) */ refcount_t ref; diff --git a/block/blk-cgroup.h b/block/blk-cgroup.h index 90b3959d88cfa4..bd472a30bc61f6 100644 --- a/block/blk-cgroup.h +++ b/block/blk-cgroup.h @@ -300,19 +300,6 @@ static inline struct blkcg *cpd_to_blkcg(struct blkcg_policy_data *cpd) return cpd ? cpd->blkcg : NULL; } -/** - * blkg_path - format cgroup path of blkg - * @blkg: blkg of interest - * @buf: target buffer - * @buflen: target buffer length - * - * Format the path of the cgroup of @blkg into @buf. - */ -static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen) -{ - return cgroup_path(blkg->blkcg->css.cgroup, buf, buflen); -} - /** * blkg_get - get a blkg reference * @blkg: blkg to get From 28d8c13830cc530996157e22ecf22def90cb7f35 Mon Sep 17 00:00:00 2001 From: Jeff Johnson Date: Mon, 17 Jun 2024 18:13:32 -0700 Subject: [PATCH 073/209] amiflop: add missing MODULE_DESCRIPTION() macro With ARCH=m68k, make allmodconfig && make W=1 C=1 reports: WARNING: modpost: missing MODULE_DESCRIPTION() in drivers/block/amiflop.o Add the missing invocation of the MODULE_DESCRIPTION() macro. Signed-off-by: Jeff Johnson Reviewed-by: Geert Uytterhoeven Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240617-md-m68k-drivers-block-v1-1-b200599a315e@quicinc.com Signed-off-by: Jens Axboe --- drivers/block/amiflop.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/block/amiflop.c b/drivers/block/amiflop.c index a25414228e4741..310254c994aa88 100644 --- a/drivers/block/amiflop.c +++ b/drivers/block/amiflop.c @@ -232,6 +232,7 @@ static DEFINE_MUTEX(amiflop_mutex); static unsigned long int fd_def_df0 = FD_DD_3; /* default for df0 if it doesn't identify */ module_param(fd_def_df0, ulong, 0); +MODULE_DESCRIPTION("Amiga floppy driver"); MODULE_LICENSE("GPL"); /* From ba8df22e25e7e906254f4f490d7bcfbe637152aa Mon Sep 17 00:00:00 2001 From: Jeff Johnson Date: Mon, 17 Jun 2024 18:13:33 -0700 Subject: [PATCH 074/209] ataflop: add missing MODULE_DESCRIPTION() macro With ARCH=m68k, make allmodconfig && make W=1 C=1 reports: WARNING: modpost: missing MODULE_DESCRIPTION() in drivers/block/ataflop.o Add the missing invocation of the MODULE_DESCRIPTION() macro. Signed-off-by: Jeff Johnson Reviewed-by: Geert Uytterhoeven Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240617-md-m68k-drivers-block-v1-2-b200599a315e@quicinc.com Signed-off-by: Jens Axboe --- drivers/block/ataflop.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/block/ataflop.c b/drivers/block/ataflop.c index cacc4ba942a814..b19884da43760c 100644 --- a/drivers/block/ataflop.c +++ b/drivers/block/ataflop.c @@ -2197,4 +2197,5 @@ static void __exit atari_floppy_exit(void) module_init(atari_floppy_init) module_exit(atari_floppy_exit) +MODULE_DESCRIPTION("Atari floppy driver"); MODULE_LICENSE("GPL"); From 465478bb00168a7620788990b1679c5067d421f2 Mon Sep 17 00:00:00 2001 From: Jeff Johnson Date: Mon, 17 Jun 2024 18:13:34 -0700 Subject: [PATCH 075/209] z2ram: add missing MODULE_DESCRIPTION() macro With ARCH=m68k, make allmodconfig && make W=1 C=1 reports: WARNING: modpost: missing MODULE_DESCRIPTION() in drivers/block/z2ram.o Add the missing invocation of the MODULE_DESCRIPTION() macro. Signed-off-by: Jeff Johnson Reviewed-by: Geert Uytterhoeven Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240617-md-m68k-drivers-block-v1-3-b200599a315e@quicinc.com Signed-off-by: Jens Axboe --- drivers/block/z2ram.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/block/z2ram.c b/drivers/block/z2ram.c index 7c5f4e4d9b5037..4b7219be1bb860 100644 --- a/drivers/block/z2ram.c +++ b/drivers/block/z2ram.c @@ -409,4 +409,5 @@ static void __exit z2_exit(void) module_init(z2_init); module_exit(z2_exit); +MODULE_DESCRIPTION("Amiga Zorro II ramdisk driver"); MODULE_LICENSE("GPL"); From dc2e77979412d289df9049d8c693761db8602867 Mon Sep 17 00:00:00 2001 From: Gabriel Krisman Bertazi Date: Fri, 14 Jun 2024 12:30:44 -0400 Subject: [PATCH 076/209] net: Split a __sys_bind helper for io_uring io_uring holds a reference to the file and maintains a sockaddr_storage address. Similarly to what was done to __sys_connect_file, split an internal helper for __sys_bind in preparation to supporting an io_uring bind command. Reviewed-by: Jens Axboe Signed-off-by: Gabriel Krisman Bertazi Reviewed-by: Kuniyuki Iwashima Acked-by: Jakub Kicinski Link: https://lore.kernel.org/r/20240614163047.31581-1-krisman@suse.de Signed-off-by: Jens Axboe --- include/linux/socket.h | 2 ++ net/socket.c | 25 ++++++++++++++++--------- 2 files changed, 18 insertions(+), 9 deletions(-) diff --git a/include/linux/socket.h b/include/linux/socket.h index 89d16b90370bd4..b3000f49e9f59f 100644 --- a/include/linux/socket.h +++ b/include/linux/socket.h @@ -442,6 +442,8 @@ extern int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr, extern int __sys_socket(int family, int type, int protocol); extern struct file *__sys_socket_file(int family, int type, int protocol); extern int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen); +extern int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address, + int addrlen); extern int __sys_connect_file(struct file *file, struct sockaddr_storage *addr, int addrlen, int file_flags); extern int __sys_connect(int fd, struct sockaddr __user *uservaddr, diff --git a/net/socket.c b/net/socket.c index e416920e9399ec..fd0714e10cedd1 100644 --- a/net/socket.c +++ b/net/socket.c @@ -1822,6 +1822,20 @@ SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol, return __sys_socketpair(family, type, protocol, usockvec); } +int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address, + int addrlen) +{ + int err; + + err = security_socket_bind(sock, (struct sockaddr *)address, + addrlen); + if (!err) + err = READ_ONCE(sock->ops)->bind(sock, + (struct sockaddr *)address, + addrlen); + return err; +} + /* * Bind a name to a socket. Nothing much to do here since it's * the protocol's responsibility to handle the local address. @@ -1839,15 +1853,8 @@ int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen) sock = sockfd_lookup_light(fd, &err, &fput_needed); if (sock) { err = move_addr_to_kernel(umyaddr, addrlen, &address); - if (!err) { - err = security_socket_bind(sock, - (struct sockaddr *)&address, - addrlen); - if (!err) - err = READ_ONCE(sock->ops)->bind(sock, - (struct sockaddr *) - &address, addrlen); - } + if (!err) + err = __sys_bind_socket(sock, &address, addrlen); fput_light(sock->file, fput_needed); } return err; From bb6aaf736680f0f3c2e6281735c47c64e2042819 Mon Sep 17 00:00:00 2001 From: Gabriel Krisman Bertazi Date: Fri, 14 Jun 2024 12:30:45 -0400 Subject: [PATCH 077/209] net: Split a __sys_listen helper for io_uring io_uring holds a reference to the file and maintains a sockaddr_storage address. Similarly to what was done to __sys_connect_file, split an internal helper for __sys_listen in preparation to support an io_uring listen command. Reviewed-by: Jens Axboe Signed-off-by: Gabriel Krisman Bertazi Reviewed-by: Kuniyuki Iwashima Acked-by: Jakub Kicinski Link: https://lore.kernel.org/r/20240614163047.31581-2-krisman@suse.de Signed-off-by: Jens Axboe --- include/linux/socket.h | 1 + net/socket.c | 23 ++++++++++++++--------- 2 files changed, 15 insertions(+), 9 deletions(-) diff --git a/include/linux/socket.h b/include/linux/socket.h index b3000f49e9f59f..c1f16cdab67762 100644 --- a/include/linux/socket.h +++ b/include/linux/socket.h @@ -449,6 +449,7 @@ extern int __sys_connect_file(struct file *file, struct sockaddr_storage *addr, extern int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen); extern int __sys_listen(int fd, int backlog); +extern int __sys_listen_socket(struct socket *sock, int backlog); extern int __sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len); extern int __sys_getpeername(int fd, struct sockaddr __user *usockaddr, diff --git a/net/socket.c b/net/socket.c index fd0714e10cedd1..fcbdd5bc47ac2f 100644 --- a/net/socket.c +++ b/net/socket.c @@ -1870,23 +1870,28 @@ SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen) * necessary for a listen, and if that works, we mark the socket as * ready for listening. */ +int __sys_listen_socket(struct socket *sock, int backlog) +{ + int somaxconn, err; + + somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn); + if ((unsigned int)backlog > somaxconn) + backlog = somaxconn; + + err = security_socket_listen(sock, backlog); + if (!err) + err = READ_ONCE(sock->ops)->listen(sock, backlog); + return err; +} int __sys_listen(int fd, int backlog) { struct socket *sock; int err, fput_needed; - int somaxconn; sock = sockfd_lookup_light(fd, &err, &fput_needed); if (sock) { - somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn); - if ((unsigned int)backlog > somaxconn) - backlog = somaxconn; - - err = security_socket_listen(sock, backlog); - if (!err) - err = READ_ONCE(sock->ops)->listen(sock, backlog); - + err = __sys_listen_socket(sock, backlog); fput_light(sock->file, fput_needed); } return err; From 7481fd93fa0a851740e26026485f56a1305454ce Mon Sep 17 00:00:00 2001 From: Gabriel Krisman Bertazi Date: Fri, 14 Jun 2024 12:30:46 -0400 Subject: [PATCH 078/209] io_uring: Introduce IORING_OP_BIND IORING_OP_BIND provides the semantic of bind(2) via io_uring. While this is an essentially synchronous system call, the main point is to enable a network path to execute fully with io_uring registered and descriptorless files. Signed-off-by: Gabriel Krisman Bertazi Link: https://lore.kernel.org/r/20240614163047.31581-3-krisman@suse.de Signed-off-by: Jens Axboe --- include/uapi/linux/io_uring.h | 1 + io_uring/net.c | 36 +++++++++++++++++++++++++++++++++++ io_uring/net.h | 3 +++ io_uring/opdef.c | 13 +++++++++++++ 4 files changed, 53 insertions(+) diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h index 994bf7af0efe20..4ef153d95c875e 100644 --- a/include/uapi/linux/io_uring.h +++ b/include/uapi/linux/io_uring.h @@ -257,6 +257,7 @@ enum io_uring_op { IORING_OP_FUTEX_WAITV, IORING_OP_FIXED_FD_INSTALL, IORING_OP_FTRUNCATE, + IORING_OP_BIND, /* this goes last, obviously */ IORING_OP_LAST, diff --git a/io_uring/net.c b/io_uring/net.c index 7c98c4d5094633..8cc4cfc2fef312 100644 --- a/io_uring/net.c +++ b/io_uring/net.c @@ -51,6 +51,11 @@ struct io_connect { bool seen_econnaborted; }; +struct io_bind { + struct file *file; + int addr_len; +}; + struct io_sr_msg { struct file *file; union { @@ -1715,6 +1720,37 @@ int io_connect(struct io_kiocb *req, unsigned int issue_flags) return IOU_OK; } +int io_bind_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) +{ + struct io_bind *bind = io_kiocb_to_cmd(req, struct io_bind); + struct sockaddr __user *uaddr; + struct io_async_msghdr *io; + + if (sqe->len || sqe->buf_index || sqe->rw_flags || sqe->splice_fd_in) + return -EINVAL; + + uaddr = u64_to_user_ptr(READ_ONCE(sqe->addr)); + bind->addr_len = READ_ONCE(sqe->addr2); + + io = io_msg_alloc_async(req); + if (unlikely(!io)) + return -ENOMEM; + return move_addr_to_kernel(uaddr, bind->addr_len, &io->addr); +} + +int io_bind(struct io_kiocb *req, unsigned int issue_flags) +{ + struct io_bind *bind = io_kiocb_to_cmd(req, struct io_bind); + struct io_async_msghdr *io = req->async_data; + int ret; + + ret = __sys_bind_socket(sock_from_file(req->file), &io->addr, bind->addr_len); + if (ret < 0) + req_set_fail(req); + io_req_set_res(req, ret, 0); + return 0; +} + void io_netmsg_cache_free(const void *entry) { struct io_async_msghdr *kmsg = (struct io_async_msghdr *) entry; diff --git a/io_uring/net.h b/io_uring/net.h index 0eb1c1920fc9a9..49f9a7bc111390 100644 --- a/io_uring/net.h +++ b/io_uring/net.h @@ -49,6 +49,9 @@ int io_sendmsg_zc(struct io_kiocb *req, unsigned int issue_flags); int io_send_zc_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe); void io_send_zc_cleanup(struct io_kiocb *req); +int io_bind_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe); +int io_bind(struct io_kiocb *req, unsigned int issue_flags); + void io_netmsg_cache_free(const void *entry); #else static inline void io_netmsg_cache_free(const void *entry) diff --git a/io_uring/opdef.c b/io_uring/opdef.c index 2e3b7b16effb34..1424c2d9e9ef2d 100644 --- a/io_uring/opdef.c +++ b/io_uring/opdef.c @@ -495,6 +495,16 @@ const struct io_issue_def io_issue_defs[] = { .prep = io_ftruncate_prep, .issue = io_ftruncate, }, + [IORING_OP_BIND] = { +#if defined(CONFIG_NET) + .needs_file = 1, + .prep = io_bind_prep, + .issue = io_bind, + .async_size = sizeof(struct io_async_msghdr), +#else + .prep = io_eopnotsupp_prep, +#endif + }, }; const struct io_cold_def io_cold_defs[] = { @@ -716,6 +726,9 @@ const struct io_cold_def io_cold_defs[] = { [IORING_OP_FTRUNCATE] = { .name = "FTRUNCATE", }, + [IORING_OP_BIND] = { + .name = "BIND", + }, }; const char *io_uring_get_opcode(u8 opcode) From ff140cc8628abfb1755691d16cfa8788d8820ef7 Mon Sep 17 00:00:00 2001 From: Gabriel Krisman Bertazi Date: Fri, 14 Jun 2024 12:30:47 -0400 Subject: [PATCH 079/209] io_uring: Introduce IORING_OP_LISTEN IORING_OP_LISTEN provides the semantic of listen(2) via io_uring. While this is an essentially synchronous system call, the main point is to enable a network path to execute fully with io_uring registered and descriptorless files. Signed-off-by: Gabriel Krisman Bertazi Link: https://lore.kernel.org/r/20240614163047.31581-4-krisman@suse.de Signed-off-by: Jens Axboe --- include/uapi/linux/io_uring.h | 1 + io_uring/net.c | 28 ++++++++++++++++++++++++++++ io_uring/net.h | 3 +++ io_uring/opdef.c | 13 +++++++++++++ 4 files changed, 45 insertions(+) diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h index 4ef153d95c875e..2aaf7ee256ac42 100644 --- a/include/uapi/linux/io_uring.h +++ b/include/uapi/linux/io_uring.h @@ -258,6 +258,7 @@ enum io_uring_op { IORING_OP_FIXED_FD_INSTALL, IORING_OP_FTRUNCATE, IORING_OP_BIND, + IORING_OP_LISTEN, /* this goes last, obviously */ IORING_OP_LAST, diff --git a/io_uring/net.c b/io_uring/net.c index 8cc4cfc2fef312..db4a4a03ce3ab6 100644 --- a/io_uring/net.c +++ b/io_uring/net.c @@ -56,6 +56,11 @@ struct io_bind { int addr_len; }; +struct io_listen { + struct file *file; + int backlog; +}; + struct io_sr_msg { struct file *file; union { @@ -1751,6 +1756,29 @@ int io_bind(struct io_kiocb *req, unsigned int issue_flags) return 0; } +int io_listen_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe) +{ + struct io_listen *listen = io_kiocb_to_cmd(req, struct io_listen); + + if (sqe->addr || sqe->buf_index || sqe->rw_flags || sqe->splice_fd_in || sqe->addr2) + return -EINVAL; + + listen->backlog = READ_ONCE(sqe->len); + return 0; +} + +int io_listen(struct io_kiocb *req, unsigned int issue_flags) +{ + struct io_listen *listen = io_kiocb_to_cmd(req, struct io_listen); + int ret; + + ret = __sys_listen_socket(sock_from_file(req->file), listen->backlog); + if (ret < 0) + req_set_fail(req); + io_req_set_res(req, ret, 0); + return 0; +} + void io_netmsg_cache_free(const void *entry) { struct io_async_msghdr *kmsg = (struct io_async_msghdr *) entry; diff --git a/io_uring/net.h b/io_uring/net.h index 49f9a7bc111390..52bfee05f06a14 100644 --- a/io_uring/net.h +++ b/io_uring/net.h @@ -52,6 +52,9 @@ void io_send_zc_cleanup(struct io_kiocb *req); int io_bind_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe); int io_bind(struct io_kiocb *req, unsigned int issue_flags); +int io_listen_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe); +int io_listen(struct io_kiocb *req, unsigned int issue_flags); + void io_netmsg_cache_free(const void *entry); #else static inline void io_netmsg_cache_free(const void *entry) diff --git a/io_uring/opdef.c b/io_uring/opdef.c index 1424c2d9e9ef2d..2dd49cf22f642d 100644 --- a/io_uring/opdef.c +++ b/io_uring/opdef.c @@ -503,6 +503,16 @@ const struct io_issue_def io_issue_defs[] = { .async_size = sizeof(struct io_async_msghdr), #else .prep = io_eopnotsupp_prep, +#endif + }, + [IORING_OP_LISTEN] = { +#if defined(CONFIG_NET) + .needs_file = 1, + .prep = io_listen_prep, + .issue = io_listen, + .async_size = sizeof(struct io_async_msghdr), +#else + .prep = io_eopnotsupp_prep, #endif }, }; @@ -729,6 +739,9 @@ const struct io_cold_def io_cold_defs[] = { [IORING_OP_BIND] = { .name = "BIND", }, + [IORING_OP_LISTEN] = { + .name = "LISTEN", + }, }; const char *io_uring_get_opcode(u8 opcode) From dd9300e9eaeeb212f77ffeb72d1d8756107f1f1f Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:28 +0200 Subject: [PATCH 080/209] xen-blkfront: don't disable cache flushes when they fail MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit blkfront always had a robust negotiation protocol for detecting a write cache. Stop simply disabling cache flushes in the block layer as the flags handling is moving to the atomic queue limits API that needs user context to freeze the queue for that. Instead handle the case of the feature flags cleared inside of blkfront. This removes old debug code to check for such a mismatch which was previously impossible to hit, including the check for passthrough requests that blkfront never used to start with. Signed-off-by: Christoph Hellwig Acked-by: Roger Pau Monné Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-2-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/xen-blkfront.c | 44 +++++++++++++++++++----------------- 1 file changed, 23 insertions(+), 21 deletions(-) diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c index 9b4ec3e4908cce..851b03844edd13 100644 --- a/drivers/block/xen-blkfront.c +++ b/drivers/block/xen-blkfront.c @@ -788,6 +788,11 @@ static int blkif_queue_rw_req(struct request *req, struct blkfront_ring_info *ri * A barrier request a superset of FUA, so we can * implement it the same way. (It's also a FLUSH+FUA, * since it is guaranteed ordered WRT previous writes.) + * + * Note that can end up here with a FUA write and the + * flags cleared. This happens when the flag was + * run-time disabled after a failing I/O, and we'll + * simplify submit it as a normal write. */ if (info->feature_flush && info->feature_fua) ring_req->operation = @@ -795,8 +800,6 @@ static int blkif_queue_rw_req(struct request *req, struct blkfront_ring_info *ri else if (info->feature_flush) ring_req->operation = BLKIF_OP_FLUSH_DISKCACHE; - else - ring_req->operation = 0; } ring_req->u.rw.nr_segments = num_grant; if (unlikely(require_extra_req)) { @@ -887,16 +890,6 @@ static inline void flush_requests(struct blkfront_ring_info *rinfo) notify_remote_via_irq(rinfo->irq); } -static inline bool blkif_request_flush_invalid(struct request *req, - struct blkfront_info *info) -{ - return (blk_rq_is_passthrough(req) || - ((req_op(req) == REQ_OP_FLUSH) && - !info->feature_flush) || - ((req->cmd_flags & REQ_FUA) && - !info->feature_fua)); -} - static blk_status_t blkif_queue_rq(struct blk_mq_hw_ctx *hctx, const struct blk_mq_queue_data *qd) { @@ -908,12 +901,22 @@ static blk_status_t blkif_queue_rq(struct blk_mq_hw_ctx *hctx, rinfo = get_rinfo(info, qid); blk_mq_start_request(qd->rq); spin_lock_irqsave(&rinfo->ring_lock, flags); - if (RING_FULL(&rinfo->ring)) - goto out_busy; - if (blkif_request_flush_invalid(qd->rq, rinfo->dev_info)) - goto out_err; + /* + * Check if the backend actually supports flushes. + * + * While the block layer won't send us flushes if we don't claim to + * support them, the Xen protocol allows the backend to revoke support + * at any time. That is of course a really bad idea and dangerous, but + * has been allowed for 10+ years. In that case we simply clear the + * flags, and directly return here for an empty flush and ignore the + * FUA flag later on. + */ + if (unlikely(req_op(qd->rq) == REQ_OP_FLUSH && !info->feature_flush)) + goto complete; + if (RING_FULL(&rinfo->ring)) + goto out_busy; if (blkif_queue_request(qd->rq, rinfo)) goto out_busy; @@ -921,14 +924,14 @@ static blk_status_t blkif_queue_rq(struct blk_mq_hw_ctx *hctx, spin_unlock_irqrestore(&rinfo->ring_lock, flags); return BLK_STS_OK; -out_err: - spin_unlock_irqrestore(&rinfo->ring_lock, flags); - return BLK_STS_IOERR; - out_busy: blk_mq_stop_hw_queue(hctx); spin_unlock_irqrestore(&rinfo->ring_lock, flags); return BLK_STS_DEV_RESOURCE; +complete: + spin_unlock_irqrestore(&rinfo->ring_lock, flags); + blk_mq_end_request(qd->rq, BLK_STS_OK); + return BLK_STS_OK; } static void blkif_complete_rq(struct request *rq) @@ -1627,7 +1630,6 @@ static irqreturn_t blkif_interrupt(int irq, void *dev_id) blkif_req(req)->error = BLK_STS_OK; info->feature_fua = 0; info->feature_flush = 0; - xlvbd_flush(info); } fallthrough; case BLKIF_OP_READ: From be60e7700e6df1e16a2f60f45bece08e6140a46d Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:29 +0200 Subject: [PATCH 081/209] sd: remove sd_is_zoned Since commit 7437bb73f087 ("block: remove support for the host aware zone model"), only ZBC devices expose a zoned access model. sd_is_zoned is used to check for that and thus return false for host aware devices. Replace the helper with the simple open coded TYPE_ZBC check to fix this. Fixes: 7437bb73f087 ("block: remove support for the host aware zone model") Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Reviewed-by: Johannes Thumshirn Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240617060532.127975-3-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/sd.c | 6 +----- drivers/scsi/sd.h | 5 ----- drivers/scsi/sd_zbc.c | 13 ++++--------- 3 files changed, 5 insertions(+), 19 deletions(-) diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index e01393ed42076b..664523048ce819 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -466,10 +466,6 @@ provisioning_mode_store(struct device *dev, struct device_attribute *attr, if (sdp->type != TYPE_DISK) return -EINVAL; - /* ignore the provisioning mode for ZBC devices */ - if (sd_is_zoned(sdkp)) - return count; - mode = sysfs_match_string(lbp_mode, buf); if (mode < 0) return -EINVAL; @@ -2288,7 +2284,7 @@ static int sd_done(struct scsi_cmnd *SCpnt) } out: - if (sd_is_zoned(sdkp)) + if (sdkp->device->type == TYPE_ZBC) good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr); SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt, diff --git a/drivers/scsi/sd.h b/drivers/scsi/sd.h index 726f1613f6cb56..7603b3c67b233f 100644 --- a/drivers/scsi/sd.h +++ b/drivers/scsi/sd.h @@ -222,11 +222,6 @@ static inline sector_t sectors_to_logical(struct scsi_device *sdev, sector_t sec void sd_dif_config_host(struct scsi_disk *sdkp, struct queue_limits *lim); -static inline int sd_is_zoned(struct scsi_disk *sdkp) -{ - return sdkp->zoned == 1 || sdkp->device->type == TYPE_ZBC; -} - #ifdef CONFIG_BLK_DEV_ZONED int sd_zbc_read_zones(struct scsi_disk *sdkp, struct queue_limits *lim, diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c index f685838d9ed214..8cc9c025017961 100644 --- a/drivers/scsi/sd_zbc.c +++ b/drivers/scsi/sd_zbc.c @@ -232,7 +232,7 @@ int sd_zbc_report_zones(struct gendisk *disk, sector_t sector, int zone_idx = 0; int ret; - if (!sd_is_zoned(sdkp)) + if (sdkp->device->type != TYPE_ZBC) /* Not a zoned device */ return -EOPNOTSUPP; @@ -300,7 +300,7 @@ static blk_status_t sd_zbc_cmnd_checks(struct scsi_cmnd *cmd) struct scsi_disk *sdkp = scsi_disk(rq->q->disk); sector_t sector = blk_rq_pos(rq); - if (!sd_is_zoned(sdkp)) + if (sdkp->device->type != TYPE_ZBC) /* Not a zoned device */ return BLK_STS_IOERR; @@ -521,7 +521,7 @@ static int sd_zbc_check_capacity(struct scsi_disk *sdkp, unsigned char *buf, static void sd_zbc_print_zones(struct scsi_disk *sdkp) { - if (!sd_is_zoned(sdkp) || !sdkp->capacity) + if (sdkp->device->type != TYPE_ZBC || !sdkp->capacity) return; if (sdkp->capacity & (sdkp->zone_info.zone_blocks - 1)) @@ -598,13 +598,8 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, struct queue_limits *lim, u32 zone_blocks = 0; int ret; - if (!sd_is_zoned(sdkp)) { - /* - * Device managed or normal SCSI disk, no special handling - * required. - */ + if (sdkp->device->type != TYPE_ZBC) return 0; - } /* READ16/WRITE16/SYNC16 is mandatory for ZBC devices */ sdkp->device->use_16_for_rw = 1; From 308ad58af49d6c4c3b7a36b98972cc9db4d7b36a Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:30 +0200 Subject: [PATCH 082/209] sd: move zone limits setup out of sd_read_block_characteristics Move a bit of code that sets up the zone flag and the write granularity into sd_zbc_read_zones to be with the rest of the zoned limits. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-4-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/sd.c | 21 +-------------------- drivers/scsi/sd_zbc.c | 9 +++++++++ 2 files changed, 10 insertions(+), 20 deletions(-) diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 664523048ce819..66f7d1e3429c86 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -3312,29 +3312,10 @@ static void sd_read_block_characteristics(struct scsi_disk *sdkp, blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q); } - -#ifdef CONFIG_BLK_DEV_ZONED /* sd_probe rejects ZBD devices early otherwise */ - if (sdkp->device->type == TYPE_ZBC) { - lim->zoned = true; - - /* - * Per ZBC and ZAC specifications, writes in sequential write - * required zones of host-managed devices must be aligned to - * the device physical block size. - */ - lim->zone_write_granularity = sdkp->physical_block_size; - } else { - /* - * Host-aware devices are treated as conventional. - */ - lim->zoned = false; - } -#endif /* CONFIG_BLK_DEV_ZONED */ - if (!sdkp->first_scan) return; - if (lim->zoned) + if (sdkp->device->type == TYPE_ZBC) sd_printk(KERN_NOTICE, sdkp, "Host-managed zoned block device\n"); else if (sdkp->zoned == 1) sd_printk(KERN_NOTICE, sdkp, "Host-aware SMR disk used as regular disk\n"); diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c index 8cc9c025017961..360ec980499529 100644 --- a/drivers/scsi/sd_zbc.c +++ b/drivers/scsi/sd_zbc.c @@ -601,6 +601,15 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, struct queue_limits *lim, if (sdkp->device->type != TYPE_ZBC) return 0; + lim->zoned = true; + + /* + * Per ZBC and ZAC specifications, writes in sequential write required + * zones of host-managed devices must be aligned to the device physical + * block size. + */ + lim->zone_write_granularity = sdkp->physical_block_size; + /* READ16/WRITE16/SYNC16 is mandatory for ZBC devices */ sdkp->device->use_16_for_rw = 1; sdkp->device->use_10_for_rw = 0; From c9055b44abe60da69aa4ee4fdcb78ee7fe733335 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:31 +0200 Subject: [PATCH 083/209] loop: stop using loop_reconfigure_limits in __loop_clr_fd __loop_clr_fd wants to clear all settings on the device. Prepare for moving more settings into the block limits by open coding loop_reconfigure_limits. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Reviewed-by: Bart Van Assche Link: https://lore.kernel.org/r/20240617060532.127975-5-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/loop.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 93780f41646b75..fd671028fa8554 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -1133,6 +1133,7 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, static void __loop_clr_fd(struct loop_device *lo, bool release) { + struct queue_limits lim; struct file *filp; gfp_t gfp = lo->old_gfp_mask; @@ -1156,7 +1157,14 @@ static void __loop_clr_fd(struct loop_device *lo, bool release) lo->lo_offset = 0; lo->lo_sizelimit = 0; memset(lo->lo_file_name, 0, LO_NAME_SIZE); - loop_reconfigure_limits(lo, 512, false); + + /* reset the block size to the default */ + lim = queue_limits_start_update(lo->lo_queue); + lim.logical_block_size = SECTOR_SIZE; + lim.physical_block_size = SECTOR_SIZE; + lim.io_min = SECTOR_SIZE; + queue_limits_commit_update(lo->lo_queue, &lim); + invalidate_disk(lo->lo_disk); loop_sysfs_exit(lo); /* let user-space know about this change */ From ae0d40ff49642651f969883ef9fc79d69c1632d7 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:32 +0200 Subject: [PATCH 084/209] loop: always update discard settings in loop_reconfigure_limits Simplify loop_reconfigure_limits by always updating the discard limits. This adds a little more work to loop_set_block_size, but doesn't change the outcome as the discard flag won't change. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Reviewed-by: Bart Van Assche Link: https://lore.kernel.org/r/20240617060532.127975-6-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/loop.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index fd671028fa8554..ce197cbea5f434 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -975,8 +975,7 @@ loop_set_status_from_info(struct loop_device *lo, return 0; } -static int loop_reconfigure_limits(struct loop_device *lo, unsigned short bsize, - bool update_discard_settings) +static int loop_reconfigure_limits(struct loop_device *lo, unsigned short bsize) { struct queue_limits lim; @@ -984,8 +983,7 @@ static int loop_reconfigure_limits(struct loop_device *lo, unsigned short bsize, lim.logical_block_size = bsize; lim.physical_block_size = bsize; lim.io_min = bsize; - if (update_discard_settings) - loop_config_discard(lo, &lim); + loop_config_discard(lo, &lim); return queue_limits_commit_update(lo->lo_queue, &lim); } @@ -1086,7 +1084,7 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, else bsize = 512; - error = loop_reconfigure_limits(lo, bsize, true); + error = loop_reconfigure_limits(lo, bsize); if (WARN_ON_ONCE(error)) goto out_unlock; @@ -1496,7 +1494,7 @@ static int loop_set_block_size(struct loop_device *lo, unsigned long arg) invalidate_bdev(lo->lo_device); blk_mq_freeze_queue(lo->lo_queue); - err = loop_reconfigure_limits(lo, arg, false); + err = loop_reconfigure_limits(lo, arg); loop_update_dio(lo); blk_mq_unfreeze_queue(lo->lo_queue); From a17ece76bcfe7b86327b19cae1652d7c62068a30 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:33 +0200 Subject: [PATCH 085/209] loop: regularize upgrading the block size for direct I/O The LOOP_CONFIGURE path automatically upgrades the block size to that of the underlying file for O_DIRECT file descriptors, but the LOOP_SET_BLOCK_SIZE path does not. Fix this by lifting the code to pick the block size into common code. Signed-off-by: Christoph Hellwig Reviewed-by: Hannes Reinecke Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Link: https://lore.kernel.org/r/20240617060532.127975-7-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/loop.c | 25 +++++++++++++++---------- 1 file changed, 15 insertions(+), 10 deletions(-) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index ce197cbea5f434..eea3e4919e356e 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -975,10 +975,24 @@ loop_set_status_from_info(struct loop_device *lo, return 0; } +static unsigned short loop_default_blocksize(struct loop_device *lo, + struct block_device *backing_bdev) +{ + /* In case of direct I/O, match underlying block size */ + if ((lo->lo_backing_file->f_flags & O_DIRECT) && backing_bdev) + return bdev_logical_block_size(backing_bdev); + return SECTOR_SIZE; +} + static int loop_reconfigure_limits(struct loop_device *lo, unsigned short bsize) { + struct file *file = lo->lo_backing_file; + struct inode *inode = file->f_mapping->host; struct queue_limits lim; + if (!bsize) + bsize = loop_default_blocksize(lo, inode->i_sb->s_bdev); + lim = queue_limits_start_update(lo->lo_queue); lim.logical_block_size = bsize; lim.physical_block_size = bsize; @@ -997,7 +1011,6 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, int error; loff_t size; bool partscan; - unsigned short bsize; bool is_loop; if (!file) @@ -1076,15 +1089,7 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, if (!(lo->lo_flags & LO_FLAGS_READ_ONLY) && file->f_op->fsync) blk_queue_write_cache(lo->lo_queue, true, false); - if (config->block_size) - bsize = config->block_size; - else if ((lo->lo_backing_file->f_flags & O_DIRECT) && inode->i_sb->s_bdev) - /* In case of direct I/O, match underlying block size */ - bsize = bdev_logical_block_size(inode->i_sb->s_bdev); - else - bsize = 512; - - error = loop_reconfigure_limits(lo, bsize); + error = loop_reconfigure_limits(lo, config->block_size); if (WARN_ON_ONCE(error)) goto out_unlock; From 4ce37fe0938b02b7b947029c40b72d76a22a3882 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:34 +0200 Subject: [PATCH 086/209] loop: also use the default block size from an underlying block device Fix the code in loop_reconfigure_limits to pick a default block size for O_DIRECT file descriptors to also work when the loop device sits on top of a block device and not just on a regular file on a block device based file system. Signed-off-by: Christoph Hellwig Reviewed-by: Hannes Reinecke Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Link: https://lore.kernel.org/r/20240617060532.127975-8-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/loop.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index eea3e4919e356e..6a4826708a3acf 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -988,10 +988,16 @@ static int loop_reconfigure_limits(struct loop_device *lo, unsigned short bsize) { struct file *file = lo->lo_backing_file; struct inode *inode = file->f_mapping->host; + struct block_device *backing_bdev = NULL; struct queue_limits lim; + if (S_ISBLK(inode->i_mode)) + backing_bdev = I_BDEV(inode); + else if (inode->i_sb->s_bdev) + backing_bdev = inode->i_sb->s_bdev; + if (!bsize) - bsize = loop_default_blocksize(lo, inode->i_sb->s_bdev); + bsize = loop_default_blocksize(lo, backing_bdev); lim = queue_limits_start_update(lo->lo_queue); lim.logical_block_size = bsize; From 97dd4a43d69b74a114be466d6887e257971adfe9 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:35 +0200 Subject: [PATCH 087/209] loop: fold loop_update_rotational into loop_reconfigure_limits This prepares for moving the rotational flag into the queue_limits and also fixes it for the case where the loop device is backed by a block device. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Reviewed-by: Bart Van Assche Link: https://lore.kernel.org/r/20240617060532.127975-9-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/loop.c | 23 ++++------------------- 1 file changed, 4 insertions(+), 19 deletions(-) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 6a4826708a3acf..8991de8fb1bb0b 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -916,24 +916,6 @@ static void loop_free_idle_workers_timer(struct timer_list *timer) return loop_free_idle_workers(lo, false); } -static void loop_update_rotational(struct loop_device *lo) -{ - struct file *file = lo->lo_backing_file; - struct inode *file_inode = file->f_mapping->host; - struct block_device *file_bdev = file_inode->i_sb->s_bdev; - struct request_queue *q = lo->lo_queue; - bool nonrot = true; - - /* not all filesystems (e.g. tmpfs) have a sb->s_bdev */ - if (file_bdev) - nonrot = bdev_nonrot(file_bdev); - - if (nonrot) - blk_queue_flag_set(QUEUE_FLAG_NONROT, q); - else - blk_queue_flag_clear(QUEUE_FLAG_NONROT, q); -} - /** * loop_set_status_from_info - configure device from loop_info * @lo: struct loop_device to configure @@ -1003,6 +985,10 @@ static int loop_reconfigure_limits(struct loop_device *lo, unsigned short bsize) lim.logical_block_size = bsize; lim.physical_block_size = bsize; lim.io_min = bsize; + if (!backing_bdev || bdev_nonrot(backing_bdev)) + blk_queue_flag_set(QUEUE_FLAG_NONROT, lo->lo_queue); + else + blk_queue_flag_clear(QUEUE_FLAG_NONROT, lo->lo_queue); loop_config_discard(lo, &lim); return queue_limits_commit_update(lo->lo_queue, &lim); } @@ -1099,7 +1085,6 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, if (WARN_ON_ONCE(error)) goto out_unlock; - loop_update_rotational(lo); loop_update_dio(lo); loop_sysfs_init(lo); From bbe5c84122b35c37f2706872fe34da66f0854b56 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:36 +0200 Subject: [PATCH 088/209] virtio_blk: remove virtblk_update_cache_mode virtblk_update_cache_mode boils down to a single call to blk_queue_write_cache. Remove it in preparation for moving the cache control flags into the queue_limits. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Stefan Hajnoczi Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Reviewed-by: Johannes Thumshirn Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240617060532.127975-10-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/virtio_blk.c | 13 +++---------- 1 file changed, 3 insertions(+), 10 deletions(-) diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index 2351f411fa4680..378b241911ca87 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -1089,14 +1089,6 @@ static int virtblk_get_cache_mode(struct virtio_device *vdev) return writeback; } -static void virtblk_update_cache_mode(struct virtio_device *vdev) -{ - u8 writeback = virtblk_get_cache_mode(vdev); - struct virtio_blk *vblk = vdev->priv; - - blk_queue_write_cache(vblk->disk->queue, writeback, false); -} - static const char *const virtblk_cache_types[] = { "write through", "write back" }; @@ -1116,7 +1108,7 @@ cache_type_store(struct device *dev, struct device_attribute *attr, return i; virtio_cwrite8(vdev, offsetof(struct virtio_blk_config, wce), i); - virtblk_update_cache_mode(vdev); + blk_queue_write_cache(disk->queue, virtblk_get_cache_mode(vdev), false); return count; } @@ -1528,7 +1520,8 @@ static int virtblk_probe(struct virtio_device *vdev) vblk->index = index; /* configure queue flush support */ - virtblk_update_cache_mode(vdev); + blk_queue_write_cache(vblk->disk->queue, virtblk_get_cache_mode(vdev), + false); /* If disk is read-only in the host, the guest should obey */ if (virtio_has_feature(vdev, VIRTIO_BLK_F_RO)) From 6b377787a306253111404325aee98005b361e59a Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:37 +0200 Subject: [PATCH 089/209] nbd: move setting the cache control flags to __nbd_set_size Move setting the cache control flags in nbd in preparation for moving these flags into the queue_limits structure. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Josef Bacik Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-11-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/nbd.c | 17 +++++++---------- 1 file changed, 7 insertions(+), 10 deletions(-) diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c index ad887d614d5b3f..44b8c671921e5c 100644 --- a/drivers/block/nbd.c +++ b/drivers/block/nbd.c @@ -342,6 +342,12 @@ static int __nbd_set_size(struct nbd_device *nbd, loff_t bytesize, lim.max_hw_discard_sectors = UINT_MAX; else lim.max_hw_discard_sectors = 0; + if (!(nbd->config->flags & NBD_FLAG_SEND_FLUSH)) + blk_queue_write_cache(nbd->disk->queue, false, false); + else if (nbd->config->flags & NBD_FLAG_SEND_FUA) + blk_queue_write_cache(nbd->disk->queue, true, true); + else + blk_queue_write_cache(nbd->disk->queue, true, false); lim.logical_block_size = blksize; lim.physical_block_size = blksize; error = queue_limits_commit_update(nbd->disk->queue, &lim); @@ -1286,19 +1292,10 @@ static void nbd_bdev_reset(struct nbd_device *nbd) static void nbd_parse_flags(struct nbd_device *nbd) { - struct nbd_config *config = nbd->config; - if (config->flags & NBD_FLAG_READ_ONLY) + if (nbd->config->flags & NBD_FLAG_READ_ONLY) set_disk_ro(nbd->disk, true); else set_disk_ro(nbd->disk, false); - if (config->flags & NBD_FLAG_SEND_FLUSH) { - if (config->flags & NBD_FLAG_SEND_FUA) - blk_queue_write_cache(nbd->disk->queue, true, true); - else - blk_queue_write_cache(nbd->disk->queue, true, false); - } - else - blk_queue_write_cache(nbd->disk->queue, false, false); } static void send_disconnects(struct nbd_device *nbd) From af2814149883e2c1851866ea2afcd8eadc040f79 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:38 +0200 Subject: [PATCH 090/209] block: freeze the queue in queue_attr_store queue_attr_store updates attributes used to control generating I/O, and can cause malformed bios if changed with I/O in flight. Freeze the queue in common code instead of adding it to almost every attribute. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240617060532.127975-12-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-mq.c | 5 +++-- block/blk-sysfs.c | 9 ++------- 2 files changed, 5 insertions(+), 9 deletions(-) diff --git a/block/blk-mq.c b/block/blk-mq.c index 0d4cd39c3d25da..58b0d6c7cc34d6 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -4631,13 +4631,15 @@ int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr) int ret; unsigned long i; + if (WARN_ON_ONCE(!q->mq_freeze_depth)) + return -EINVAL; + if (!set) return -EINVAL; if (q->nr_requests == nr) return 0; - blk_mq_freeze_queue(q); blk_mq_quiesce_queue(q); ret = 0; @@ -4671,7 +4673,6 @@ int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr) } blk_mq_unquiesce_queue(q); - blk_mq_unfreeze_queue(q); return ret; } diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index f0f9314ab65c61..5c787965b7d09e 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -189,12 +189,9 @@ static ssize_t queue_discard_max_store(struct request_queue *q, if ((max_discard_bytes >> SECTOR_SHIFT) > UINT_MAX) return -EINVAL; - blk_mq_freeze_queue(q); lim = queue_limits_start_update(q); lim.max_user_discard_sectors = max_discard_bytes >> SECTOR_SHIFT; err = queue_limits_commit_update(q, &lim); - blk_mq_unfreeze_queue(q); - if (err) return err; return ret; @@ -241,11 +238,9 @@ queue_max_sectors_store(struct request_queue *q, const char *page, size_t count) if (ret < 0) return ret; - blk_mq_freeze_queue(q); lim = queue_limits_start_update(q); lim.max_user_sectors = max_sectors_kb << 1; err = queue_limits_commit_update(q, &lim); - blk_mq_unfreeze_queue(q); if (err) return err; return ret; @@ -585,13 +580,11 @@ static ssize_t queue_wb_lat_store(struct request_queue *q, const char *page, * ends up either enabling or disabling wbt completely. We can't * have IO inflight if that happens. */ - blk_mq_freeze_queue(q); blk_mq_quiesce_queue(q); wbt_set_min_lat(q, val); blk_mq_unquiesce_queue(q); - blk_mq_unfreeze_queue(q); return count; } @@ -722,9 +715,11 @@ queue_attr_store(struct kobject *kobj, struct attribute *attr, if (!entry->store) return -EIO; + blk_mq_freeze_queue(q); mutex_lock(&q->sysfs_lock); res = entry->store(q, page, length); mutex_unlock(&q->sysfs_lock); + blk_mq_unfreeze_queue(q); return res; } From 70905f8706b62113ae32c8df721384ff6ffb6c6a Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:39 +0200 Subject: [PATCH 091/209] block: remove blk_flush_policy Fold blk_flush_policy into the only caller to prepare for pending changes to it. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240617060532.127975-13-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-flush.c | 33 +++++++++++++++------------------ 1 file changed, 15 insertions(+), 18 deletions(-) diff --git a/block/blk-flush.c b/block/blk-flush.c index c17cf8ed8113db..2234f8b3fc05f2 100644 --- a/block/blk-flush.c +++ b/block/blk-flush.c @@ -100,23 +100,6 @@ blk_get_flush_queue(struct request_queue *q, struct blk_mq_ctx *ctx) return blk_mq_map_queue(q, REQ_OP_FLUSH, ctx)->fq; } -static unsigned int blk_flush_policy(unsigned long fflags, struct request *rq) -{ - unsigned int policy = 0; - - if (blk_rq_sectors(rq)) - policy |= REQ_FSEQ_DATA; - - if (fflags & (1UL << QUEUE_FLAG_WC)) { - if (rq->cmd_flags & REQ_PREFLUSH) - policy |= REQ_FSEQ_PREFLUSH; - if (!(fflags & (1UL << QUEUE_FLAG_FUA)) && - (rq->cmd_flags & REQ_FUA)) - policy |= REQ_FSEQ_POSTFLUSH; - } - return policy; -} - static unsigned int blk_flush_cur_seq(struct request *rq) { return 1 << ffz(rq->flush.seq); @@ -399,12 +382,26 @@ bool blk_insert_flush(struct request *rq) { struct request_queue *q = rq->q; unsigned long fflags = q->queue_flags; /* may change, cache */ - unsigned int policy = blk_flush_policy(fflags, rq); struct blk_flush_queue *fq = blk_get_flush_queue(q, rq->mq_ctx); + unsigned int policy = 0; /* FLUSH/FUA request must never be merged */ WARN_ON_ONCE(rq->bio != rq->biotail); + if (blk_rq_sectors(rq)) + policy |= REQ_FSEQ_DATA; + + /* + * Check which flushes we need to sequence for this operation. + */ + if (fflags & (1UL << QUEUE_FLAG_WC)) { + if (rq->cmd_flags & REQ_PREFLUSH) + policy |= REQ_FSEQ_PREFLUSH; + if (!(fflags & (1UL << QUEUE_FLAG_FUA)) && + (rq->cmd_flags & REQ_FUA)) + policy |= REQ_FSEQ_POSTFLUSH; + } + /* * @policy now records what operations need to be done. Adjust * REQ_PREFLUSH and FUA for the driver. From 1122c0c1cc71f740fa4d5f14f239194e06a1d5e7 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:40 +0200 Subject: [PATCH 092/209] block: move cache control settings out of queue->flags Move the cache control settings into the queue_limits so that the flags can be set atomically with the device queue frozen. Add new features and flags field for the driver set flags, and internal (usually sysfs-controlled) flags in the block layer. Note that we'll eventually remove enough field from queue_limits to bring it back to the previous size. The disable flag is inverted compared to the previous meaning, which means it now survives a rescan, similar to the max_sectors and max_discard_sectors user limits. The FLUSH and FUA flags are now inherited by blk_stack_limits, which simplified the code in dm a lot, but also causes a slight behavior change in that dm-switch and dm-unstripe now advertise a write cache despite setting num_flush_bios to 0. The I/O path will handle this gracefully, but as far as I can tell the lack of num_flush_bios and thus flush support is a pre-existing data integrity bug in those targets that really needs fixing, after which a non-zero num_flush_bios should be required in dm for targets that map to underlying devices. Signed-off-by: Christoph Hellwig Acked-by: Ulf Hansson Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-14-hch@lst.de Signed-off-by: Jens Axboe --- .../block/writeback_cache_control.rst | 67 +++++++++++-------- arch/um/drivers/ubd_kern.c | 2 +- block/blk-core.c | 2 +- block/blk-flush.c | 9 ++- block/blk-mq-debugfs.c | 2 - block/blk-settings.c | 29 ++------ block/blk-sysfs.c | 29 +++++--- block/blk-wbt.c | 4 +- drivers/block/drbd/drbd_main.c | 2 +- drivers/block/loop.c | 9 +-- drivers/block/nbd.c | 14 ++-- drivers/block/null_blk/main.c | 12 ++-- drivers/block/ps3disk.c | 7 +- drivers/block/rnbd/rnbd-clt.c | 10 +-- drivers/block/ublk_drv.c | 8 ++- drivers/block/virtio_blk.c | 20 ++++-- drivers/block/xen-blkfront.c | 8 ++- drivers/md/bcache/super.c | 7 +- drivers/md/dm-table.c | 39 +++-------- drivers/md/md.c | 8 ++- drivers/mmc/core/block.c | 42 ++++++------ drivers/mmc/core/queue.c | 12 ++-- drivers/mmc/core/queue.h | 3 +- drivers/mtd/mtd_blkdevs.c | 5 +- drivers/nvdimm/pmem.c | 4 +- drivers/nvme/host/core.c | 7 +- drivers/nvme/host/multipath.c | 6 -- drivers/scsi/sd.c | 28 +++++--- include/linux/blkdev.h | 38 +++++++++-- 29 files changed, 227 insertions(+), 206 deletions(-) diff --git a/Documentation/block/writeback_cache_control.rst b/Documentation/block/writeback_cache_control.rst index b208488d0aae85..c575e08beda8e3 100644 --- a/Documentation/block/writeback_cache_control.rst +++ b/Documentation/block/writeback_cache_control.rst @@ -46,41 +46,50 @@ worry if the underlying devices need any explicit cache flushing and how the Forced Unit Access is implemented. The REQ_PREFLUSH and REQ_FUA flags may both be set on a single bio. +Feature settings for block drivers +---------------------------------- -Implementation details for bio based block drivers --------------------------------------------------------------- +For devices that do not support volatile write caches there is no driver +support required, the block layer completes empty REQ_PREFLUSH requests before +entering the driver and strips off the REQ_PREFLUSH and REQ_FUA bits from +requests that have a payload. -These drivers will always see the REQ_PREFLUSH and REQ_FUA bits as they sit -directly below the submit_bio interface. For remapping drivers the REQ_FUA -bits need to be propagated to underlying devices, and a global flush needs -to be implemented for bios with the REQ_PREFLUSH bit set. For real device -drivers that do not have a volatile cache the REQ_PREFLUSH and REQ_FUA bits -on non-empty bios can simply be ignored, and REQ_PREFLUSH requests without -data can be completed successfully without doing any work. Drivers for -devices with volatile caches need to implement the support for these -flags themselves without any help from the block layer. +For devices with volatile write caches the driver needs to tell the block layer +that it supports flushing caches by setting the + BLK_FEAT_WRITE_CACHE -Implementation details for request_fn based block drivers ---------------------------------------------------------- +flag in the queue_limits feature field. For devices that also support the FUA +bit the block layer needs to be told to pass on the REQ_FUA bit by also setting +the -For devices that do not support volatile write caches there is no driver -support required, the block layer completes empty REQ_PREFLUSH requests before -entering the driver and strips off the REQ_PREFLUSH and REQ_FUA bits from -requests that have a payload. For devices with volatile write caches the -driver needs to tell the block layer that it supports flushing caches by -doing:: + BLK_FEAT_FUA + +flag in the features field of the queue_limits structure. + +Implementation details for bio based block drivers +-------------------------------------------------- + +For bio based drivers the REQ_PREFLUSH and REQ_FUA bit are simplify passed on +to the driver if the drivers sets the BLK_FEAT_WRITE_CACHE flag and the drivers +needs to handle them. + +*NOTE*: The REQ_FUA bit also gets passed on when the BLK_FEAT_FUA flags is +_not_ set. Any bio based driver that sets BLK_FEAT_WRITE_CACHE also needs to +handle REQ_FUA. - blk_queue_write_cache(sdkp->disk->queue, true, false); +For remapping drivers the REQ_FUA bits need to be propagated to underlying +devices, and a global flush needs to be implemented for bios with the +REQ_PREFLUSH bit set. -and handle empty REQ_OP_FLUSH requests in its prep_fn/request_fn. Note that -REQ_PREFLUSH requests with a payload are automatically turned into a sequence -of an empty REQ_OP_FLUSH request followed by the actual write by the block -layer. For devices that also support the FUA bit the block layer needs -to be told to pass through the REQ_FUA bit using:: +Implementation details for blk-mq drivers +----------------------------------------- - blk_queue_write_cache(sdkp->disk->queue, true, true); +When the BLK_FEAT_WRITE_CACHE flag is set, REQ_OP_WRITE | REQ_PREFLUSH requests +with a payload are automatically turned into a sequence of a REQ_OP_FLUSH +request followed by the actual write by the block layer. -and the driver must handle write requests that have the REQ_FUA bit set -in prep_fn/request_fn. If the FUA bit is not natively supported the block -layer turns it into an empty REQ_OP_FLUSH request after the actual write. +When the BLK_FEAT_FUA flags is set, the REQ_FUA bit simplify passed on for the +REQ_OP_WRITE request, else a REQ_OP_FLUSH request is sent by the block layer +after the completion of the write request for bio submissions with the REQ_FUA +bit set. diff --git a/arch/um/drivers/ubd_kern.c b/arch/um/drivers/ubd_kern.c index cdcb75a68989dd..19e01691ea0ea7 100644 --- a/arch/um/drivers/ubd_kern.c +++ b/arch/um/drivers/ubd_kern.c @@ -835,6 +835,7 @@ static int ubd_add(int n, char **error_out) struct queue_limits lim = { .max_segments = MAX_SG, .seg_boundary_mask = PAGE_SIZE - 1, + .features = BLK_FEAT_WRITE_CACHE, }; struct gendisk *disk; int err = 0; @@ -882,7 +883,6 @@ static int ubd_add(int n, char **error_out) } blk_queue_flag_set(QUEUE_FLAG_NONROT, disk->queue); - blk_queue_write_cache(disk->queue, true, false); disk->major = UBD_MAJOR; disk->first_minor = n << UBD_SHIFT; disk->minors = 1 << UBD_SHIFT; diff --git a/block/blk-core.c b/block/blk-core.c index 82c3ae22d76d88..2b45a4df9a1aa1 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -782,7 +782,7 @@ void submit_bio_noacct(struct bio *bio) if (WARN_ON_ONCE(bio_op(bio) != REQ_OP_WRITE && bio_op(bio) != REQ_OP_ZONE_APPEND)) goto end_io; - if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) { + if (!bdev_write_cache(bdev)) { bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA); if (!bio_sectors(bio)) { status = BLK_STS_OK; diff --git a/block/blk-flush.c b/block/blk-flush.c index 2234f8b3fc05f2..30b9d5033a2b85 100644 --- a/block/blk-flush.c +++ b/block/blk-flush.c @@ -381,8 +381,8 @@ static void blk_rq_init_flush(struct request *rq) bool blk_insert_flush(struct request *rq) { struct request_queue *q = rq->q; - unsigned long fflags = q->queue_flags; /* may change, cache */ struct blk_flush_queue *fq = blk_get_flush_queue(q, rq->mq_ctx); + bool supports_fua = q->limits.features & BLK_FEAT_FUA; unsigned int policy = 0; /* FLUSH/FUA request must never be merged */ @@ -394,11 +394,10 @@ bool blk_insert_flush(struct request *rq) /* * Check which flushes we need to sequence for this operation. */ - if (fflags & (1UL << QUEUE_FLAG_WC)) { + if (blk_queue_write_cache(q)) { if (rq->cmd_flags & REQ_PREFLUSH) policy |= REQ_FSEQ_PREFLUSH; - if (!(fflags & (1UL << QUEUE_FLAG_FUA)) && - (rq->cmd_flags & REQ_FUA)) + if ((rq->cmd_flags & REQ_FUA) && !supports_fua) policy |= REQ_FSEQ_POSTFLUSH; } @@ -407,7 +406,7 @@ bool blk_insert_flush(struct request *rq) * REQ_PREFLUSH and FUA for the driver. */ rq->cmd_flags &= ~REQ_PREFLUSH; - if (!(fflags & (1UL << QUEUE_FLAG_FUA))) + if (!supports_fua) rq->cmd_flags &= ~REQ_FUA; /* diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index 770c0c2b72faaa..e8b9db7c30c455 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -93,8 +93,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(INIT_DONE), QUEUE_FLAG_NAME(STABLE_WRITES), QUEUE_FLAG_NAME(POLL), - QUEUE_FLAG_NAME(WC), - QUEUE_FLAG_NAME(FUA), QUEUE_FLAG_NAME(DAX), QUEUE_FLAG_NAME(STATS), QUEUE_FLAG_NAME(REGISTERED), diff --git a/block/blk-settings.c b/block/blk-settings.c index f11c8676eb4c67..536ee202fcdccb 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -261,6 +261,9 @@ static int blk_validate_limits(struct queue_limits *lim) lim->misaligned = 0; } + if (!(lim->features & BLK_FEAT_WRITE_CACHE)) + lim->features &= ~BLK_FEAT_FUA; + err = blk_validate_integrity_limits(lim); if (err) return err; @@ -454,6 +457,8 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, { unsigned int top, bottom, alignment, ret = 0; + t->features |= (b->features & BLK_FEAT_INHERIT_MASK); + t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); t->max_user_sectors = min_not_zero(t->max_user_sectors, b->max_user_sectors); @@ -711,30 +716,6 @@ void blk_set_queue_depth(struct request_queue *q, unsigned int depth) } EXPORT_SYMBOL(blk_set_queue_depth); -/** - * blk_queue_write_cache - configure queue's write cache - * @q: the request queue for the device - * @wc: write back cache on or off - * @fua: device supports FUA writes, if true - * - * Tell the block layer about the write cache of @q. - */ -void blk_queue_write_cache(struct request_queue *q, bool wc, bool fua) -{ - if (wc) { - blk_queue_flag_set(QUEUE_FLAG_HW_WC, q); - blk_queue_flag_set(QUEUE_FLAG_WC, q); - } else { - blk_queue_flag_clear(QUEUE_FLAG_HW_WC, q); - blk_queue_flag_clear(QUEUE_FLAG_WC, q); - } - if (fua) - blk_queue_flag_set(QUEUE_FLAG_FUA, q); - else - blk_queue_flag_clear(QUEUE_FLAG_FUA, q); -} -EXPORT_SYMBOL_GPL(blk_queue_write_cache); - int bdev_alignment_offset(struct block_device *bdev) { struct request_queue *q = bdev_get_queue(bdev); diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 5c787965b7d09e..4f524c1d5e08bd 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -423,32 +423,41 @@ static ssize_t queue_io_timeout_store(struct request_queue *q, const char *page, static ssize_t queue_wc_show(struct request_queue *q, char *page) { - if (test_bit(QUEUE_FLAG_WC, &q->queue_flags)) - return sprintf(page, "write back\n"); - - return sprintf(page, "write through\n"); + if (q->limits.features & BLK_FLAGS_WRITE_CACHE_DISABLED) + return sprintf(page, "write through\n"); + return sprintf(page, "write back\n"); } static ssize_t queue_wc_store(struct request_queue *q, const char *page, size_t count) { + struct queue_limits lim; + bool disable; + int err; + if (!strncmp(page, "write back", 10)) { - if (!test_bit(QUEUE_FLAG_HW_WC, &q->queue_flags)) - return -EINVAL; - blk_queue_flag_set(QUEUE_FLAG_WC, q); + disable = false; } else if (!strncmp(page, "write through", 13) || - !strncmp(page, "none", 4)) { - blk_queue_flag_clear(QUEUE_FLAG_WC, q); + !strncmp(page, "none", 4)) { + disable = true; } else { return -EINVAL; } + lim = queue_limits_start_update(q); + if (disable) + lim.flags |= BLK_FLAGS_WRITE_CACHE_DISABLED; + else + lim.flags &= ~BLK_FLAGS_WRITE_CACHE_DISABLED; + err = queue_limits_commit_update(q, &lim); + if (err) + return err; return count; } static ssize_t queue_fua_show(struct request_queue *q, char *page) { - return sprintf(page, "%u\n", test_bit(QUEUE_FLAG_FUA, &q->queue_flags)); + return sprintf(page, "%u\n", !!(q->limits.features & BLK_FEAT_FUA)); } static ssize_t queue_dax_show(struct request_queue *q, char *page) diff --git a/block/blk-wbt.c b/block/blk-wbt.c index 64472134dd26df..1a5e4b049ecd1d 100644 --- a/block/blk-wbt.c +++ b/block/blk-wbt.c @@ -206,8 +206,8 @@ static void wbt_rqw_done(struct rq_wb *rwb, struct rq_wait *rqw, */ if (wb_acct & WBT_DISCARD) limit = rwb->wb_background; - else if (test_bit(QUEUE_FLAG_WC, &rwb->rqos.disk->queue->queue_flags) && - !wb_recent_wait(rwb)) + else if (blk_queue_write_cache(rwb->rqos.disk->queue) && + !wb_recent_wait(rwb)) limit = 0; else limit = rwb->wb_normal; diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 113b441d4d3670..bf42a46781fa21 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2697,6 +2697,7 @@ enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsig * connect. */ .max_hw_sectors = DRBD_MAX_BIO_SIZE_SAFE >> 8, + .features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA, }; device = minor_to_device(minor); @@ -2736,7 +2737,6 @@ enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsig disk->private_data = device; blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, disk->queue); - blk_queue_write_cache(disk->queue, true, true); device->md_io.page = alloc_page(GFP_KERNEL); if (!device->md_io.page) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 8991de8fb1bb0b..08d0fc7f17b701 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -985,6 +985,9 @@ static int loop_reconfigure_limits(struct loop_device *lo, unsigned short bsize) lim.logical_block_size = bsize; lim.physical_block_size = bsize; lim.io_min = bsize; + lim.features &= ~BLK_FEAT_WRITE_CACHE; + if (file->f_op->fsync && !(lo->lo_flags & LO_FLAGS_READ_ONLY)) + lim.features |= BLK_FEAT_WRITE_CACHE; if (!backing_bdev || bdev_nonrot(backing_bdev)) blk_queue_flag_set(QUEUE_FLAG_NONROT, lo->lo_queue); else @@ -1078,9 +1081,6 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, lo->old_gfp_mask = mapping_gfp_mask(mapping); mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS)); - if (!(lo->lo_flags & LO_FLAGS_READ_ONLY) && file->f_op->fsync) - blk_queue_write_cache(lo->lo_queue, true, false); - error = loop_reconfigure_limits(lo, config->block_size); if (WARN_ON_ONCE(error)) goto out_unlock; @@ -1131,9 +1131,6 @@ static void __loop_clr_fd(struct loop_device *lo, bool release) struct file *filp; gfp_t gfp = lo->old_gfp_mask; - if (test_bit(QUEUE_FLAG_WC, &lo->lo_queue->queue_flags)) - blk_queue_write_cache(lo->lo_queue, false, false); - /* * Freeze the request queue when unbinding on a live file descriptor and * thus an open device. When called from ->release we are guaranteed diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c index 44b8c671921e5c..cb1c86a6a3fb9d 100644 --- a/drivers/block/nbd.c +++ b/drivers/block/nbd.c @@ -342,12 +342,14 @@ static int __nbd_set_size(struct nbd_device *nbd, loff_t bytesize, lim.max_hw_discard_sectors = UINT_MAX; else lim.max_hw_discard_sectors = 0; - if (!(nbd->config->flags & NBD_FLAG_SEND_FLUSH)) - blk_queue_write_cache(nbd->disk->queue, false, false); - else if (nbd->config->flags & NBD_FLAG_SEND_FUA) - blk_queue_write_cache(nbd->disk->queue, true, true); - else - blk_queue_write_cache(nbd->disk->queue, true, false); + if (!(nbd->config->flags & NBD_FLAG_SEND_FLUSH)) { + lim.features &= ~(BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA); + } else if (nbd->config->flags & NBD_FLAG_SEND_FUA) { + lim.features |= BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA; + } else { + lim.features |= BLK_FEAT_WRITE_CACHE; + lim.features &= ~BLK_FEAT_FUA; + } lim.logical_block_size = blksize; lim.physical_block_size = blksize; error = queue_limits_commit_update(nbd->disk->queue, &lim); diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c index 75f189e42f885d..21f9d256e88402 100644 --- a/drivers/block/null_blk/main.c +++ b/drivers/block/null_blk/main.c @@ -1928,6 +1928,13 @@ static int null_add_dev(struct nullb_device *dev) goto out_cleanup_tags; } + if (dev->cache_size > 0) { + set_bit(NULLB_DEV_FL_CACHE, &nullb->dev->flags); + lim.features |= BLK_FEAT_WRITE_CACHE; + if (dev->fua) + lim.features |= BLK_FEAT_FUA; + } + nullb->disk = blk_mq_alloc_disk(nullb->tag_set, &lim, nullb); if (IS_ERR(nullb->disk)) { rv = PTR_ERR(nullb->disk); @@ -1940,11 +1947,6 @@ static int null_add_dev(struct nullb_device *dev) nullb_setup_bwtimer(nullb); } - if (dev->cache_size > 0) { - set_bit(NULLB_DEV_FL_CACHE, &nullb->dev->flags); - blk_queue_write_cache(nullb->q, true, dev->fua); - } - nullb->q->queuedata = nullb; blk_queue_flag_set(QUEUE_FLAG_NONROT, nullb->q); diff --git a/drivers/block/ps3disk.c b/drivers/block/ps3disk.c index b810ac0a5c4b97..8b73cf459b5937 100644 --- a/drivers/block/ps3disk.c +++ b/drivers/block/ps3disk.c @@ -388,9 +388,8 @@ static int ps3disk_probe(struct ps3_system_bus_device *_dev) .max_segments = -1, .max_segment_size = dev->bounce_size, .dma_alignment = dev->blk_size - 1, + .features = BLK_FEAT_WRITE_CACHE, }; - - struct request_queue *queue; struct gendisk *gendisk; if (dev->blk_size < 512) { @@ -447,10 +446,6 @@ static int ps3disk_probe(struct ps3_system_bus_device *_dev) goto fail_free_tag_set; } - queue = gendisk->queue; - - blk_queue_write_cache(queue, true, false); - priv->gendisk = gendisk; gendisk->major = ps3disk_major; gendisk->first_minor = devidx * PS3DISK_MINORS; diff --git a/drivers/block/rnbd/rnbd-clt.c b/drivers/block/rnbd/rnbd-clt.c index b7ffe03c61606d..02c4b173182719 100644 --- a/drivers/block/rnbd/rnbd-clt.c +++ b/drivers/block/rnbd/rnbd-clt.c @@ -1389,6 +1389,12 @@ static int rnbd_client_setup_device(struct rnbd_clt_dev *dev, le32_to_cpu(rsp->max_discard_sectors); } + if (rsp->cache_policy & RNBD_WRITEBACK) { + lim.features |= BLK_FEAT_WRITE_CACHE; + if (rsp->cache_policy & RNBD_FUA) + lim.features |= BLK_FEAT_FUA; + } + dev->gd = blk_mq_alloc_disk(&dev->sess->tag_set, &lim, dev); if (IS_ERR(dev->gd)) return PTR_ERR(dev->gd); @@ -1397,10 +1403,6 @@ static int rnbd_client_setup_device(struct rnbd_clt_dev *dev, blk_queue_flag_set(QUEUE_FLAG_SAME_COMP, dev->queue); blk_queue_flag_set(QUEUE_FLAG_SAME_FORCE, dev->queue); - blk_queue_write_cache(dev->queue, - !!(rsp->cache_policy & RNBD_WRITEBACK), - !!(rsp->cache_policy & RNBD_FUA)); - return rnbd_clt_setup_gen_disk(dev, rsp, idx); } diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index 4e159948c912c2..e45c65c1848d31 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -487,8 +487,6 @@ static void ublk_dev_param_basic_apply(struct ublk_device *ub) struct request_queue *q = ub->ub_disk->queue; const struct ublk_param_basic *p = &ub->params.basic; - blk_queue_write_cache(q, p->attrs & UBLK_ATTR_VOLATILE_CACHE, - p->attrs & UBLK_ATTR_FUA); if (p->attrs & UBLK_ATTR_ROTATIONAL) blk_queue_flag_clear(QUEUE_FLAG_NONROT, q); else @@ -2210,6 +2208,12 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub, struct io_uring_cmd *cmd) lim.max_zone_append_sectors = p->max_zone_append_sectors; } + if (ub->params.basic.attrs & UBLK_ATTR_VOLATILE_CACHE) { + lim.features |= BLK_FEAT_WRITE_CACHE; + if (ub->params.basic.attrs & UBLK_ATTR_FUA) + lim.features |= BLK_FEAT_FUA; + } + if (wait_for_completion_interruptible(&ub->completion) != 0) return -EINTR; diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index 378b241911ca87..b1a3c293528519 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -1100,6 +1100,7 @@ cache_type_store(struct device *dev, struct device_attribute *attr, struct gendisk *disk = dev_to_disk(dev); struct virtio_blk *vblk = disk->private_data; struct virtio_device *vdev = vblk->vdev; + struct queue_limits lim; int i; BUG_ON(!virtio_has_feature(vblk->vdev, VIRTIO_BLK_F_CONFIG_WCE)); @@ -1108,7 +1109,17 @@ cache_type_store(struct device *dev, struct device_attribute *attr, return i; virtio_cwrite8(vdev, offsetof(struct virtio_blk_config, wce), i); - blk_queue_write_cache(disk->queue, virtblk_get_cache_mode(vdev), false); + + lim = queue_limits_start_update(disk->queue); + if (virtblk_get_cache_mode(vdev)) + lim.features |= BLK_FEAT_WRITE_CACHE; + else + lim.features &= ~BLK_FEAT_WRITE_CACHE; + blk_mq_freeze_queue(disk->queue); + i = queue_limits_commit_update(disk->queue, &lim); + blk_mq_unfreeze_queue(disk->queue); + if (i) + return i; return count; } @@ -1504,6 +1515,9 @@ static int virtblk_probe(struct virtio_device *vdev) if (err) goto out_free_tags; + if (virtblk_get_cache_mode(vdev)) + lim.features |= BLK_FEAT_WRITE_CACHE; + vblk->disk = blk_mq_alloc_disk(&vblk->tag_set, &lim, vblk); if (IS_ERR(vblk->disk)) { err = PTR_ERR(vblk->disk); @@ -1519,10 +1533,6 @@ static int virtblk_probe(struct virtio_device *vdev) vblk->disk->fops = &virtblk_fops; vblk->index = index; - /* configure queue flush support */ - blk_queue_write_cache(vblk->disk->queue, virtblk_get_cache_mode(vdev), - false); - /* If disk is read-only in the host, the guest should obey */ if (virtio_has_feature(vdev, VIRTIO_BLK_F_RO)) set_disk_ro(vblk->disk, 1); diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c index 851b03844edd13..9aafce3e5987bf 100644 --- a/drivers/block/xen-blkfront.c +++ b/drivers/block/xen-blkfront.c @@ -959,6 +959,12 @@ static void blkif_set_queue_limits(const struct blkfront_info *info, lim->max_secure_erase_sectors = UINT_MAX; } + if (info->feature_flush) { + lim->features |= BLK_FEAT_WRITE_CACHE; + if (info->feature_fua) + lim->features |= BLK_FEAT_FUA; + } + /* Hard sector size and max sectors impersonate the equiv. hardware. */ lim->logical_block_size = info->sector_size; lim->physical_block_size = info->physical_sector_size; @@ -987,8 +993,6 @@ static const char *flush_info(struct blkfront_info *info) static void xlvbd_flush(struct blkfront_info *info) { - blk_queue_write_cache(info->rq, info->feature_flush ? true : false, - info->feature_fua ? true : false); pr_info("blkfront: %s: %s %s %s %s %s %s %s\n", info->gd->disk_name, flush_info(info), "persistent grants:", info->feature_persistent ? diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c index 4d11fc664cb0b8..cb6595c8b5514e 100644 --- a/drivers/md/bcache/super.c +++ b/drivers/md/bcache/super.c @@ -897,7 +897,6 @@ static int bcache_device_init(struct bcache_device *d, unsigned int block_size, sector_t sectors, struct block_device *cached_bdev, const struct block_device_operations *ops) { - struct request_queue *q; const size_t max_stripes = min_t(size_t, INT_MAX, SIZE_MAX / sizeof(atomic_t)); struct queue_limits lim = { @@ -909,6 +908,7 @@ static int bcache_device_init(struct bcache_device *d, unsigned int block_size, .io_min = block_size, .logical_block_size = block_size, .physical_block_size = block_size, + .features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA, }; uint64_t n; int idx; @@ -975,12 +975,7 @@ static int bcache_device_init(struct bcache_device *d, unsigned int block_size, d->disk->fops = ops; d->disk->private_data = d; - q = d->disk->queue; - blk_queue_flag_set(QUEUE_FLAG_NONROT, d->disk->queue); - - blk_queue_write_cache(q, true, true); - return 0; out_bioset_exit: diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index fd789eeb62d943..03abdae646829c 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -1686,34 +1686,16 @@ int dm_calculate_queue_limits(struct dm_table *t, return validate_hardware_logical_block_alignment(t, limits); } -static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev, - sector_t start, sector_t len, void *data) -{ - unsigned long flush = (unsigned long) data; - struct request_queue *q = bdev_get_queue(dev->bdev); - - return (q->queue_flags & flush); -} - -static bool dm_table_supports_flush(struct dm_table *t, unsigned long flush) +/* + * Check if a target requires flush support even if none of the underlying + * devices need it (e.g. to persist target-specific metadata). + */ +static bool dm_table_supports_flush(struct dm_table *t) { - /* - * Require at least one underlying device to support flushes. - * t->devices includes internal dm devices such as mirror logs - * so we need to use iterate_devices here, which targets - * supporting flushes must provide. - */ for (unsigned int i = 0; i < t->num_targets; i++) { struct dm_target *ti = dm_table_get_target(t, i); - if (!ti->num_flush_bios) - continue; - - if (ti->flush_supported) - return true; - - if (ti->type->iterate_devices && - ti->type->iterate_devices(ti, device_flush_capable, (void *) flush)) + if (ti->num_flush_bios && ti->flush_supported) return true; } @@ -1855,7 +1837,6 @@ static int device_requires_stable_pages(struct dm_target *ti, int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, struct queue_limits *limits) { - bool wc = false, fua = false; int r; if (dm_table_supports_nowait(t)) @@ -1876,12 +1857,8 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, if (!dm_table_supports_secure_erase(t)) limits->max_secure_erase_sectors = 0; - if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) { - wc = true; - if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_FUA))) - fua = true; - } - blk_queue_write_cache(q, wc, fua); + if (dm_table_supports_flush(t)) + limits->features |= BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA; if (dm_table_supports_dax(t, device_not_dax_capable)) { blk_queue_flag_set(QUEUE_FLAG_DAX, q); diff --git a/drivers/md/md.c b/drivers/md/md.c index 67ece2cd725f50..2f4c5d1755d857 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -5785,7 +5785,10 @@ struct mddev *md_alloc(dev_t dev, char *name) int partitioned; int shift; int unit; - int error ; + int error; + struct queue_limits lim = { + .features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA, + }; /* * Wait for any previous instance of this device to be completely @@ -5825,7 +5828,7 @@ struct mddev *md_alloc(dev_t dev, char *name) */ mddev->hold_active = UNTIL_STOP; - disk = blk_alloc_disk(NULL, NUMA_NO_NODE); + disk = blk_alloc_disk(&lim, NUMA_NO_NODE); if (IS_ERR(disk)) { error = PTR_ERR(disk); goto out_free_mddev; @@ -5843,7 +5846,6 @@ struct mddev *md_alloc(dev_t dev, char *name) disk->fops = &md_fops; disk->private_data = mddev; - blk_queue_write_cache(disk->queue, true, true); disk->events |= DISK_EVENT_MEDIA_CHANGE; mddev->gendisk = disk; error = add_disk(disk); diff --git a/drivers/mmc/core/block.c b/drivers/mmc/core/block.c index 367509b5b6466c..2c9963248fcbd6 100644 --- a/drivers/mmc/core/block.c +++ b/drivers/mmc/core/block.c @@ -2466,8 +2466,7 @@ static struct mmc_blk_data *mmc_blk_alloc_req(struct mmc_card *card, struct mmc_blk_data *md; int devidx, ret; char cap_str[10]; - bool cache_enabled = false; - bool fua_enabled = false; + unsigned int features = 0; devidx = ida_alloc_max(&mmc_blk_ida, max_devices - 1, GFP_KERNEL); if (devidx < 0) { @@ -2499,7 +2498,24 @@ static struct mmc_blk_data *mmc_blk_alloc_req(struct mmc_card *card, */ md->read_only = mmc_blk_readonly(card); - md->disk = mmc_init_queue(&md->queue, card); + if (mmc_host_cmd23(card->host)) { + if ((mmc_card_mmc(card) && + card->csd.mmca_vsn >= CSD_SPEC_VER_3) || + (mmc_card_sd(card) && + card->scr.cmds & SD_SCR_CMD23_SUPPORT)) + md->flags |= MMC_BLK_CMD23; + } + + if (md->flags & MMC_BLK_CMD23 && + ((card->ext_csd.rel_param & EXT_CSD_WR_REL_PARAM_EN) || + card->ext_csd.rel_sectors)) { + md->flags |= MMC_BLK_REL_WR; + features |= (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA); + } else if (mmc_cache_enabled(card->host)) { + features |= BLK_FEAT_WRITE_CACHE; + } + + md->disk = mmc_init_queue(&md->queue, card, features); if (IS_ERR(md->disk)) { ret = PTR_ERR(md->disk); goto err_kfree; @@ -2539,26 +2555,6 @@ static struct mmc_blk_data *mmc_blk_alloc_req(struct mmc_card *card, set_capacity(md->disk, size); - if (mmc_host_cmd23(card->host)) { - if ((mmc_card_mmc(card) && - card->csd.mmca_vsn >= CSD_SPEC_VER_3) || - (mmc_card_sd(card) && - card->scr.cmds & SD_SCR_CMD23_SUPPORT)) - md->flags |= MMC_BLK_CMD23; - } - - if (md->flags & MMC_BLK_CMD23 && - ((card->ext_csd.rel_param & EXT_CSD_WR_REL_PARAM_EN) || - card->ext_csd.rel_sectors)) { - md->flags |= MMC_BLK_REL_WR; - fua_enabled = true; - cache_enabled = true; - } - if (mmc_cache_enabled(card->host)) - cache_enabled = true; - - blk_queue_write_cache(md->queue.queue, cache_enabled, fua_enabled); - string_get_size((u64)size, 512, STRING_UNITS_2, cap_str, sizeof(cap_str)); pr_info("%s: %s %s %s%s\n", diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c index 241cdc2b2a2a3b..97ff993d31570c 100644 --- a/drivers/mmc/core/queue.c +++ b/drivers/mmc/core/queue.c @@ -344,10 +344,12 @@ static const struct blk_mq_ops mmc_mq_ops = { }; static struct gendisk *mmc_alloc_disk(struct mmc_queue *mq, - struct mmc_card *card) + struct mmc_card *card, unsigned int features) { struct mmc_host *host = card->host; - struct queue_limits lim = { }; + struct queue_limits lim = { + .features = features, + }; struct gendisk *disk; if (mmc_can_erase(card)) @@ -413,10 +415,12 @@ static inline bool mmc_merge_capable(struct mmc_host *host) * mmc_init_queue - initialise a queue structure. * @mq: mmc queue * @card: mmc card to attach this queue + * @features: block layer features (BLK_FEAT_*) * * Initialise a MMC card request queue. */ -struct gendisk *mmc_init_queue(struct mmc_queue *mq, struct mmc_card *card) +struct gendisk *mmc_init_queue(struct mmc_queue *mq, struct mmc_card *card, + unsigned int features) { struct mmc_host *host = card->host; struct gendisk *disk; @@ -460,7 +464,7 @@ struct gendisk *mmc_init_queue(struct mmc_queue *mq, struct mmc_card *card) return ERR_PTR(ret); - disk = mmc_alloc_disk(mq, card); + disk = mmc_alloc_disk(mq, card, features); if (IS_ERR(disk)) blk_mq_free_tag_set(&mq->tag_set); return disk; diff --git a/drivers/mmc/core/queue.h b/drivers/mmc/core/queue.h index 9ade3bcbb714e4..1498840a4ea008 100644 --- a/drivers/mmc/core/queue.h +++ b/drivers/mmc/core/queue.h @@ -94,7 +94,8 @@ struct mmc_queue { struct work_struct complete_work; }; -struct gendisk *mmc_init_queue(struct mmc_queue *mq, struct mmc_card *card); +struct gendisk *mmc_init_queue(struct mmc_queue *mq, struct mmc_card *card, + unsigned int features); extern void mmc_cleanup_queue(struct mmc_queue *); extern void mmc_queue_suspend(struct mmc_queue *); extern void mmc_queue_resume(struct mmc_queue *); diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c index 3caa0717d46c01..1b9f57f231e8be 100644 --- a/drivers/mtd/mtd_blkdevs.c +++ b/drivers/mtd/mtd_blkdevs.c @@ -336,6 +336,8 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new) lim.logical_block_size = tr->blksize; if (tr->discard) lim.max_hw_discard_sectors = UINT_MAX; + if (tr->flush) + lim.features |= BLK_FEAT_WRITE_CACHE; /* Create gendisk */ gd = blk_mq_alloc_disk(new->tag_set, &lim, new); @@ -373,9 +375,6 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new) spin_lock_init(&new->queue_lock); INIT_LIST_HEAD(&new->rq_list); - if (tr->flush) - blk_queue_write_cache(new->rq, true, false); - blk_queue_flag_set(QUEUE_FLAG_NONROT, new->rq); blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, new->rq); diff --git a/drivers/nvdimm/pmem.c b/drivers/nvdimm/pmem.c index 598fe2e89bda45..aff818469c114c 100644 --- a/drivers/nvdimm/pmem.c +++ b/drivers/nvdimm/pmem.c @@ -455,6 +455,7 @@ static int pmem_attach_disk(struct device *dev, .logical_block_size = pmem_sector_size(ndns), .physical_block_size = PAGE_SIZE, .max_hw_sectors = UINT_MAX, + .features = BLK_FEAT_WRITE_CACHE, }; int nid = dev_to_node(dev), fua; struct resource *res = &nsio->res; @@ -495,6 +496,8 @@ static int pmem_attach_disk(struct device *dev, dev_warn(dev, "unable to guarantee persistence of writes\n"); fua = 0; } + if (fua) + lim.features |= BLK_FEAT_FUA; if (!devm_request_mem_region(dev, res->start, resource_size(res), dev_name(&ndns->dev))) { @@ -543,7 +546,6 @@ static int pmem_attach_disk(struct device *dev, } pmem->virt_addr = addr; - blk_queue_write_cache(q, true, fua); blk_queue_flag_set(QUEUE_FLAG_NONROT, q); blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, q); if (pmem->pfn_flags & PFN_MAP) diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 5a673fa5cb2612..9fc5e36fe2e55e 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -2056,7 +2056,6 @@ static int nvme_update_ns_info_generic(struct nvme_ns *ns, static int nvme_update_ns_info_block(struct nvme_ns *ns, struct nvme_ns_info *info) { - bool vwc = ns->ctrl->vwc & NVME_CTRL_VWC_PRESENT; struct queue_limits lim; struct nvme_id_ns_nvm *nvm = NULL; struct nvme_zone_info zi = {}; @@ -2106,6 +2105,11 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns, ns->head->ids.csi == NVME_CSI_ZNS) nvme_update_zone_info(ns, &lim, &zi); + if (ns->ctrl->vwc & NVME_CTRL_VWC_PRESENT) + lim.features |= BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA; + else + lim.features &= ~(BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA); + /* * Register a metadata profile for PI, or the plain non-integrity NVMe * metadata masquerading as Type 0 if supported, otherwise reject block @@ -2132,7 +2136,6 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns, if ((id->dlfeat & 0x7) == 0x1 && (id->dlfeat & (1 << 3))) ns->head->features |= NVME_NS_DEAC; set_disk_ro(ns->disk, nvme_ns_is_readonly(ns, info)); - blk_queue_write_cache(ns->disk->queue, vwc, vwc); set_bit(NVME_NS_READY, &ns->flags); blk_mq_unfreeze_queue(ns->disk->queue); diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c index 12c59db02539e5..3d0e23a0a4ddd8 100644 --- a/drivers/nvme/host/multipath.c +++ b/drivers/nvme/host/multipath.c @@ -521,7 +521,6 @@ static void nvme_requeue_work(struct work_struct *work) int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head) { struct queue_limits lim; - bool vwc = false; mutex_init(&head->lock); bio_list_init(&head->requeue_list); @@ -562,11 +561,6 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head) if (ctrl->tagset->nr_maps > HCTX_TYPE_POLL && ctrl->tagset->map[HCTX_TYPE_POLL].nr_queues) blk_queue_flag_set(QUEUE_FLAG_POLL, head->disk->queue); - - /* we need to propagate up the VMC settings */ - if (ctrl->vwc & NVME_CTRL_VWC_PRESENT) - vwc = true; - blk_queue_write_cache(head->disk->queue, vwc, vwc); return 0; } diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 66f7d1e3429c86..d8ee4a4d4a6283 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -120,17 +120,18 @@ static const char *sd_cache_types[] = { "write back, no read (daft)" }; -static void sd_set_flush_flag(struct scsi_disk *sdkp) +static void sd_set_flush_flag(struct scsi_disk *sdkp, + struct queue_limits *lim) { - bool wc = false, fua = false; - if (sdkp->WCE) { - wc = true; + lim->features |= BLK_FEAT_WRITE_CACHE; if (sdkp->DPOFUA) - fua = true; + lim->features |= BLK_FEAT_FUA; + else + lim->features &= ~BLK_FEAT_FUA; + } else { + lim->features &= ~(BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA); } - - blk_queue_write_cache(sdkp->disk->queue, wc, fua); } static ssize_t @@ -168,9 +169,18 @@ cache_type_store(struct device *dev, struct device_attribute *attr, wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0; if (sdkp->cache_override) { + struct queue_limits lim; + sdkp->WCE = wce; sdkp->RCD = rcd; - sd_set_flush_flag(sdkp); + + lim = queue_limits_start_update(sdkp->disk->queue); + sd_set_flush_flag(sdkp, &lim); + blk_mq_freeze_queue(sdkp->disk->queue); + ret = queue_limits_commit_update(sdkp->disk->queue, &lim); + blk_mq_unfreeze_queue(sdkp->disk->queue); + if (ret) + return ret; return count; } @@ -3663,7 +3673,7 @@ static int sd_revalidate_disk(struct gendisk *disk) * We now have all cache related info, determine how we deal * with flush requests. */ - sd_set_flush_flag(sdkp); + sd_set_flush_flag(sdkp, &lim); /* Initial block count limit based on CDB TRANSFER LENGTH field size. */ dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS; diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 0c247a71688561..acdfe5122faa44 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -282,6 +282,28 @@ static inline bool blk_op_is_passthrough(blk_opf_t op) return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; } +/* flags set by the driver in queue_limits.features */ +enum { + /* supports a volatile write cache */ + BLK_FEAT_WRITE_CACHE = (1u << 0), + + /* supports passing on the FUA bit */ + BLK_FEAT_FUA = (1u << 1), +}; + +/* + * Flags automatically inherited when stacking limits. + */ +#define BLK_FEAT_INHERIT_MASK \ + (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA) + + +/* internal flags in queue_limits.flags */ +enum { + /* do not send FLUSH or FUA command despite advertised write cache */ + BLK_FLAGS_WRITE_CACHE_DISABLED = (1u << 31), +}; + /* * BLK_BOUNCE_NONE: never bounce (default) * BLK_BOUNCE_HIGH: bounce all highmem pages @@ -292,6 +314,8 @@ enum blk_bounce { }; struct queue_limits { + unsigned int features; + unsigned int flags; enum blk_bounce bounce; unsigned long seg_boundary_mask; unsigned long virt_boundary_mask; @@ -536,12 +560,9 @@ struct request_queue { #define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */ #define QUEUE_FLAG_SYNCHRONOUS 11 /* always completes in submit context */ #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ -#define QUEUE_FLAG_HW_WC 13 /* Write back caching supported */ #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ #define QUEUE_FLAG_STABLE_WRITES 15 /* don't modify blks until WB is done */ #define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */ -#define QUEUE_FLAG_WC 17 /* Write back caching */ -#define QUEUE_FLAG_FUA 18 /* device supports FUA writes */ #define QUEUE_FLAG_DAX 19 /* device supports DAX */ #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */ #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */ @@ -951,7 +972,6 @@ void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev, sector_t offset, const char *pfx); extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); -extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua); struct blk_independent_access_ranges * disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges); @@ -1304,14 +1324,20 @@ static inline bool bdev_stable_writes(struct block_device *bdev) return test_bit(QUEUE_FLAG_STABLE_WRITES, &q->queue_flags); } +static inline bool blk_queue_write_cache(struct request_queue *q) +{ + return (q->limits.features & BLK_FEAT_WRITE_CACHE) && + !(q->limits.flags & BLK_FLAGS_WRITE_CACHE_DISABLED); +} + static inline bool bdev_write_cache(struct block_device *bdev) { - return test_bit(QUEUE_FLAG_WC, &bdev_get_queue(bdev)->queue_flags); + return blk_queue_write_cache(bdev_get_queue(bdev)); } static inline bool bdev_fua(struct block_device *bdev) { - return test_bit(QUEUE_FLAG_FUA, &bdev_get_queue(bdev)->queue_flags); + return bdev_get_queue(bdev)->limits.features & BLK_FEAT_FUA; } static inline bool bdev_nowait(struct block_device *bdev) From bd4a633b6f7c3c6b6ebc1a07317643270e751a94 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:41 +0200 Subject: [PATCH 093/209] block: move the nonrot flag to queue_limits Move the nonrot flag into the queue_limits feature field so that it can be set atomically with the queue frozen. Use the chance to switch to defaulting to non-rotational and require the driver to opt into rotational, which matches the polarity of the sysfs interface. For the z2ram, ps3vram, 2x memstick, ubiblock and dcssblk the new rotational flag is not set as they clearly are not rotational despite this being a behavior change. There are some other drivers that unconditionally set the rotational flag to keep the existing behavior as they arguably can be used on rotational devices even if that is probably not their main use today (e.g. virtio_blk and drbd). The flag is automatically inherited in blk_stack_limits matching the existing behavior in dm and md. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-15-hch@lst.de Signed-off-by: Jens Axboe --- arch/m68k/emu/nfblock.c | 1 + arch/um/drivers/ubd_kern.c | 1 - arch/xtensa/platforms/iss/simdisk.c | 5 +++- block/blk-mq-debugfs.c | 1 - block/blk-sysfs.c | 39 ++++++++++++++++++++++++++--- drivers/block/amiflop.c | 5 +++- drivers/block/aoe/aoeblk.c | 1 + drivers/block/ataflop.c | 5 +++- drivers/block/brd.c | 2 -- drivers/block/drbd/drbd_main.c | 3 ++- drivers/block/floppy.c | 3 ++- drivers/block/loop.c | 8 +++--- drivers/block/mtip32xx/mtip32xx.c | 1 - drivers/block/n64cart.c | 2 -- drivers/block/nbd.c | 5 ---- drivers/block/null_blk/main.c | 1 - drivers/block/pktcdvd.c | 1 + drivers/block/ps3disk.c | 3 ++- drivers/block/rbd.c | 3 --- drivers/block/rnbd/rnbd-clt.c | 4 --- drivers/block/sunvdc.c | 1 + drivers/block/swim.c | 5 +++- drivers/block/swim3.c | 5 +++- drivers/block/ublk_drv.c | 9 +++---- drivers/block/virtio_blk.c | 4 ++- drivers/block/xen-blkfront.c | 1 - drivers/block/zram/zram_drv.c | 2 -- drivers/cdrom/gdrom.c | 1 + drivers/md/bcache/super.c | 2 -- drivers/md/dm-table.c | 12 --------- drivers/md/md.c | 13 ---------- drivers/mmc/core/queue.c | 1 - drivers/mtd/mtd_blkdevs.c | 1 - drivers/nvdimm/btt.c | 1 - drivers/nvdimm/pmem.c | 1 - drivers/nvme/host/core.c | 1 - drivers/nvme/host/multipath.c | 1 - drivers/s390/block/dasd_genhd.c | 1 - drivers/s390/block/scm_blk.c | 1 - drivers/scsi/sd.c | 4 +-- include/linux/blkdev.h | 10 ++++---- 41 files changed, 83 insertions(+), 88 deletions(-) diff --git a/arch/m68k/emu/nfblock.c b/arch/m68k/emu/nfblock.c index 642fb80c5c4e31..8eea7ef9115146 100644 --- a/arch/m68k/emu/nfblock.c +++ b/arch/m68k/emu/nfblock.c @@ -98,6 +98,7 @@ static int __init nfhd_init_one(int id, u32 blocks, u32 bsize) { struct queue_limits lim = { .logical_block_size = bsize, + .features = BLK_FEAT_ROTATIONAL, }; struct nfhd_device *dev; int dev_id = id - NFHD_DEV_OFFSET; diff --git a/arch/um/drivers/ubd_kern.c b/arch/um/drivers/ubd_kern.c index 19e01691ea0ea7..9f1e76ddda5a26 100644 --- a/arch/um/drivers/ubd_kern.c +++ b/arch/um/drivers/ubd_kern.c @@ -882,7 +882,6 @@ static int ubd_add(int n, char **error_out) goto out_cleanup_tags; } - blk_queue_flag_set(QUEUE_FLAG_NONROT, disk->queue); disk->major = UBD_MAJOR; disk->first_minor = n << UBD_SHIFT; disk->minors = 1 << UBD_SHIFT; diff --git a/arch/xtensa/platforms/iss/simdisk.c b/arch/xtensa/platforms/iss/simdisk.c index defc67909a9c74..d6d2b533a5744d 100644 --- a/arch/xtensa/platforms/iss/simdisk.c +++ b/arch/xtensa/platforms/iss/simdisk.c @@ -263,6 +263,9 @@ static const struct proc_ops simdisk_proc_ops = { static int __init simdisk_setup(struct simdisk *dev, int which, struct proc_dir_entry *procdir) { + struct queue_limits lim = { + .features = BLK_FEAT_ROTATIONAL, + }; char tmp[2] = { '0' + which, 0 }; int err; @@ -271,7 +274,7 @@ static int __init simdisk_setup(struct simdisk *dev, int which, spin_lock_init(&dev->lock); dev->users = 0; - dev->gd = blk_alloc_disk(NULL, NUMA_NO_NODE); + dev->gd = blk_alloc_disk(&lim, NUMA_NO_NODE); if (IS_ERR(dev->gd)) { err = PTR_ERR(dev->gd); goto out; diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index e8b9db7c30c455..4d0e62ec88f033 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -84,7 +84,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(NOMERGES), QUEUE_FLAG_NAME(SAME_COMP), QUEUE_FLAG_NAME(FAIL_IO), - QUEUE_FLAG_NAME(NONROT), QUEUE_FLAG_NAME(IO_STAT), QUEUE_FLAG_NAME(NOXMERGES), QUEUE_FLAG_NAME(ADD_RANDOM), diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 4f524c1d5e08bd..637ed3bbbfb46f 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -263,6 +263,39 @@ static ssize_t queue_dma_alignment_show(struct request_queue *q, char *page) return queue_var_show(queue_dma_alignment(q), page); } +static ssize_t queue_feature_store(struct request_queue *q, const char *page, + size_t count, unsigned int feature) +{ + struct queue_limits lim; + unsigned long val; + ssize_t ret; + + ret = queue_var_store(&val, page, count); + if (ret < 0) + return ret; + + lim = queue_limits_start_update(q); + if (val) + lim.features |= feature; + else + lim.features &= ~feature; + ret = queue_limits_commit_update(q, &lim); + if (ret) + return ret; + return count; +} + +#define QUEUE_SYSFS_FEATURE(_name, _feature) \ +static ssize_t queue_##_name##_show(struct request_queue *q, char *page) \ +{ \ + return sprintf(page, "%u\n", !!(q->limits.features & _feature)); \ +} \ +static ssize_t queue_##_name##_store(struct request_queue *q, \ + const char *page, size_t count) \ +{ \ + return queue_feature_store(q, page, count, _feature); \ +} + #define QUEUE_SYSFS_BIT_FNS(name, flag, neg) \ static ssize_t \ queue_##name##_show(struct request_queue *q, char *page) \ @@ -289,7 +322,7 @@ queue_##name##_store(struct request_queue *q, const char *page, size_t count) \ return ret; \ } -QUEUE_SYSFS_BIT_FNS(nonrot, NONROT, 1); +QUEUE_SYSFS_FEATURE(rotational, BLK_FEAT_ROTATIONAL) QUEUE_SYSFS_BIT_FNS(random, ADD_RANDOM, 0); QUEUE_SYSFS_BIT_FNS(iostats, IO_STAT, 0); QUEUE_SYSFS_BIT_FNS(stable_writes, STABLE_WRITES, 0); @@ -526,7 +559,7 @@ static struct queue_sysfs_entry queue_hw_sector_size_entry = { .show = queue_logical_block_size_show, }; -QUEUE_RW_ENTRY(queue_nonrot, "rotational"); +QUEUE_RW_ENTRY(queue_rotational, "rotational"); QUEUE_RW_ENTRY(queue_iostats, "iostats"); QUEUE_RW_ENTRY(queue_random, "add_random"); QUEUE_RW_ENTRY(queue_stable_writes, "stable_writes"); @@ -624,7 +657,7 @@ static struct attribute *queue_attrs[] = { &queue_write_zeroes_max_entry.attr, &queue_zone_append_max_entry.attr, &queue_zone_write_granularity_entry.attr, - &queue_nonrot_entry.attr, + &queue_rotational_entry.attr, &queue_zoned_entry.attr, &queue_nr_zones_entry.attr, &queue_max_open_zones_entry.attr, diff --git a/drivers/block/amiflop.c b/drivers/block/amiflop.c index a25414228e4741..ff45701f7a5e31 100644 --- a/drivers/block/amiflop.c +++ b/drivers/block/amiflop.c @@ -1776,10 +1776,13 @@ static const struct blk_mq_ops amiflop_mq_ops = { static int fd_alloc_disk(int drive, int system) { + struct queue_limits lim = { + .features = BLK_FEAT_ROTATIONAL, + }; struct gendisk *disk; int err; - disk = blk_mq_alloc_disk(&unit[drive].tag_set, NULL, NULL); + disk = blk_mq_alloc_disk(&unit[drive].tag_set, &lim, NULL); if (IS_ERR(disk)) return PTR_ERR(disk); diff --git a/drivers/block/aoe/aoeblk.c b/drivers/block/aoe/aoeblk.c index b6dac8cee70fe1..2028795ec61cbb 100644 --- a/drivers/block/aoe/aoeblk.c +++ b/drivers/block/aoe/aoeblk.c @@ -337,6 +337,7 @@ aoeblk_gdalloc(void *vp) struct queue_limits lim = { .max_hw_sectors = aoe_maxsectors, .io_opt = SZ_2M, + .features = BLK_FEAT_ROTATIONAL, }; ulong flags; int late = 0; diff --git a/drivers/block/ataflop.c b/drivers/block/ataflop.c index cacc4ba942a814..4ee10a742bdb93 100644 --- a/drivers/block/ataflop.c +++ b/drivers/block/ataflop.c @@ -1992,9 +1992,12 @@ static const struct blk_mq_ops ataflop_mq_ops = { static int ataflop_alloc_disk(unsigned int drive, unsigned int type) { + struct queue_limits lim = { + .features = BLK_FEAT_ROTATIONAL, + }; struct gendisk *disk; - disk = blk_mq_alloc_disk(&unit[drive].tag_set, NULL, NULL); + disk = blk_mq_alloc_disk(&unit[drive].tag_set, &lim, NULL); if (IS_ERR(disk)) return PTR_ERR(disk); diff --git a/drivers/block/brd.c b/drivers/block/brd.c index 558d8e67056608..b25dc463b5e3a6 100644 --- a/drivers/block/brd.c +++ b/drivers/block/brd.c @@ -366,8 +366,6 @@ static int brd_alloc(int i) strscpy(disk->disk_name, buf, DISK_NAME_LEN); set_capacity(disk, rd_size * 2); - /* Tell the block layer that this is not a rotational device */ - blk_queue_flag_set(QUEUE_FLAG_NONROT, disk->queue); blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, disk->queue); blk_queue_flag_set(QUEUE_FLAG_NOWAIT, disk->queue); err = add_disk(disk); diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index bf42a46781fa21..2ef29a47807550 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2697,7 +2697,8 @@ enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsig * connect. */ .max_hw_sectors = DRBD_MAX_BIO_SIZE_SAFE >> 8, - .features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA, + .features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | + BLK_FEAT_ROTATIONAL, }; device = minor_to_device(minor); diff --git a/drivers/block/floppy.c b/drivers/block/floppy.c index 25c9d85667f1a2..6d7f7df97c3a6c 100644 --- a/drivers/block/floppy.c +++ b/drivers/block/floppy.c @@ -4516,7 +4516,8 @@ static bool floppy_available(int drive) static int floppy_alloc_disk(unsigned int drive, unsigned int type) { struct queue_limits lim = { - .max_hw_sectors = 64, + .max_hw_sectors = 64, + .features = BLK_FEAT_ROTATIONAL, }; struct gendisk *disk; diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 08d0fc7f17b701..86b5d956dc4e02 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -985,13 +985,11 @@ static int loop_reconfigure_limits(struct loop_device *lo, unsigned short bsize) lim.logical_block_size = bsize; lim.physical_block_size = bsize; lim.io_min = bsize; - lim.features &= ~BLK_FEAT_WRITE_CACHE; + lim.features &= ~(BLK_FEAT_WRITE_CACHE | BLK_FEAT_ROTATIONAL); if (file->f_op->fsync && !(lo->lo_flags & LO_FLAGS_READ_ONLY)) lim.features |= BLK_FEAT_WRITE_CACHE; - if (!backing_bdev || bdev_nonrot(backing_bdev)) - blk_queue_flag_set(QUEUE_FLAG_NONROT, lo->lo_queue); - else - blk_queue_flag_clear(QUEUE_FLAG_NONROT, lo->lo_queue); + if (backing_bdev && !bdev_nonrot(backing_bdev)) + lim.features |= BLK_FEAT_ROTATIONAL; loop_config_discard(lo, &lim); return queue_limits_commit_update(lo->lo_queue, &lim); } diff --git a/drivers/block/mtip32xx/mtip32xx.c b/drivers/block/mtip32xx/mtip32xx.c index 43a187609ef794..1dbbf72659d549 100644 --- a/drivers/block/mtip32xx/mtip32xx.c +++ b/drivers/block/mtip32xx/mtip32xx.c @@ -3485,7 +3485,6 @@ static int mtip_block_initialize(struct driver_data *dd) goto start_service_thread; /* Set device limits. */ - blk_queue_flag_set(QUEUE_FLAG_NONROT, dd->queue); blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, dd->queue); dma_set_max_seg_size(&dd->pdev->dev, 0x400000); diff --git a/drivers/block/n64cart.c b/drivers/block/n64cart.c index 27b2187e7a6d55..b9fdeff31cafdf 100644 --- a/drivers/block/n64cart.c +++ b/drivers/block/n64cart.c @@ -150,8 +150,6 @@ static int __init n64cart_probe(struct platform_device *pdev) set_capacity(disk, size >> SECTOR_SHIFT); set_disk_ro(disk, 1); - blk_queue_flag_set(QUEUE_FLAG_NONROT, disk->queue); - err = add_disk(disk); if (err) goto out_cleanup_disk; diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c index cb1c86a6a3fb9d..6cddf5baffe02a 100644 --- a/drivers/block/nbd.c +++ b/drivers/block/nbd.c @@ -1867,11 +1867,6 @@ static struct nbd_device *nbd_dev_add(int index, unsigned int refs) goto out_err_disk; } - /* - * Tell the block layer that we are not a rotational device - */ - blk_queue_flag_set(QUEUE_FLAG_NONROT, disk->queue); - mutex_init(&nbd->config_lock); refcount_set(&nbd->config_refs, 0); /* diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c index 21f9d256e88402..83a4ebe4763ae5 100644 --- a/drivers/block/null_blk/main.c +++ b/drivers/block/null_blk/main.c @@ -1948,7 +1948,6 @@ static int null_add_dev(struct nullb_device *dev) } nullb->q->queuedata = nullb; - blk_queue_flag_set(QUEUE_FLAG_NONROT, nullb->q); rv = ida_alloc(&nullb_indexes, GFP_KERNEL); if (rv < 0) diff --git a/drivers/block/pktcdvd.c b/drivers/block/pktcdvd.c index 8a2ce80700109d..7cece5884b9c67 100644 --- a/drivers/block/pktcdvd.c +++ b/drivers/block/pktcdvd.c @@ -2622,6 +2622,7 @@ static int pkt_setup_dev(dev_t dev, dev_t* pkt_dev) struct queue_limits lim = { .max_hw_sectors = PACKET_MAX_SECTORS, .logical_block_size = CD_FRAMESIZE, + .features = BLK_FEAT_ROTATIONAL, }; int idx; int ret = -ENOMEM; diff --git a/drivers/block/ps3disk.c b/drivers/block/ps3disk.c index 8b73cf459b5937..ff45ed76646957 100644 --- a/drivers/block/ps3disk.c +++ b/drivers/block/ps3disk.c @@ -388,7 +388,8 @@ static int ps3disk_probe(struct ps3_system_bus_device *_dev) .max_segments = -1, .max_segment_size = dev->bounce_size, .dma_alignment = dev->blk_size - 1, - .features = BLK_FEAT_WRITE_CACHE, + .features = BLK_FEAT_WRITE_CACHE | + BLK_FEAT_ROTATIONAL, }; struct gendisk *gendisk; diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c index 22ad704f81d8b9..ec1f1c7d4275cd 100644 --- a/drivers/block/rbd.c +++ b/drivers/block/rbd.c @@ -4997,9 +4997,6 @@ static int rbd_init_disk(struct rbd_device *rbd_dev) disk->fops = &rbd_bd_ops; disk->private_data = rbd_dev; - blk_queue_flag_set(QUEUE_FLAG_NONROT, q); - /* QUEUE_FLAG_ADD_RANDOM is off by default for blk-mq */ - if (!ceph_test_opt(rbd_dev->rbd_client->client, NOCRC)) blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, q); diff --git a/drivers/block/rnbd/rnbd-clt.c b/drivers/block/rnbd/rnbd-clt.c index 02c4b173182719..4918b0f68b46cd 100644 --- a/drivers/block/rnbd/rnbd-clt.c +++ b/drivers/block/rnbd/rnbd-clt.c @@ -1352,10 +1352,6 @@ static int rnbd_clt_setup_gen_disk(struct rnbd_clt_dev *dev, if (dev->access_mode == RNBD_ACCESS_RO) set_disk_ro(dev->gd, true); - /* - * Network device does not need rotational - */ - blk_queue_flag_set(QUEUE_FLAG_NONROT, dev->queue); err = add_disk(dev->gd); if (err) put_disk(dev->gd); diff --git a/drivers/block/sunvdc.c b/drivers/block/sunvdc.c index 5286cb8e0824d1..2d38331ee66793 100644 --- a/drivers/block/sunvdc.c +++ b/drivers/block/sunvdc.c @@ -791,6 +791,7 @@ static int probe_disk(struct vdc_port *port) .seg_boundary_mask = PAGE_SIZE - 1, .max_segment_size = PAGE_SIZE, .max_segments = port->ring_cookies, + .features = BLK_FEAT_ROTATIONAL, }; struct request_queue *q; struct gendisk *g; diff --git a/drivers/block/swim.c b/drivers/block/swim.c index 6731678f3a41db..126f151c4f2cf0 100644 --- a/drivers/block/swim.c +++ b/drivers/block/swim.c @@ -787,6 +787,9 @@ static void swim_cleanup_floppy_disk(struct floppy_state *fs) static int swim_floppy_init(struct swim_priv *swd) { + struct queue_limits lim = { + .features = BLK_FEAT_ROTATIONAL, + }; int err; int drive; struct swim __iomem *base = swd->base; @@ -820,7 +823,7 @@ static int swim_floppy_init(struct swim_priv *swd) goto exit_put_disks; swd->unit[drive].disk = - blk_mq_alloc_disk(&swd->unit[drive].tag_set, NULL, + blk_mq_alloc_disk(&swd->unit[drive].tag_set, &lim, &swd->unit[drive]); if (IS_ERR(swd->unit[drive].disk)) { blk_mq_free_tag_set(&swd->unit[drive].tag_set); diff --git a/drivers/block/swim3.c b/drivers/block/swim3.c index a04756ac778ee8..90be1017f7bfcd 100644 --- a/drivers/block/swim3.c +++ b/drivers/block/swim3.c @@ -1189,6 +1189,9 @@ static int swim3_add_device(struct macio_dev *mdev, int index) static int swim3_attach(struct macio_dev *mdev, const struct of_device_id *match) { + struct queue_limits lim = { + .features = BLK_FEAT_ROTATIONAL, + }; struct floppy_state *fs; struct gendisk *disk; int rc; @@ -1210,7 +1213,7 @@ static int swim3_attach(struct macio_dev *mdev, if (rc) goto out_unregister; - disk = blk_mq_alloc_disk(&fs->tag_set, NULL, fs); + disk = blk_mq_alloc_disk(&fs->tag_set, &lim, fs); if (IS_ERR(disk)) { rc = PTR_ERR(disk); goto out_free_tag_set; diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index e45c65c1848d31..4fcde099935868 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -484,14 +484,8 @@ static inline unsigned ublk_pos_to_tag(loff_t pos) static void ublk_dev_param_basic_apply(struct ublk_device *ub) { - struct request_queue *q = ub->ub_disk->queue; const struct ublk_param_basic *p = &ub->params.basic; - if (p->attrs & UBLK_ATTR_ROTATIONAL) - blk_queue_flag_clear(QUEUE_FLAG_NONROT, q); - else - blk_queue_flag_set(QUEUE_FLAG_NONROT, q); - if (p->attrs & UBLK_ATTR_READ_ONLY) set_disk_ro(ub->ub_disk, true); @@ -2214,6 +2208,9 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub, struct io_uring_cmd *cmd) lim.features |= BLK_FEAT_FUA; } + if (ub->params.basic.attrs & UBLK_ATTR_ROTATIONAL) + lim.features |= BLK_FEAT_ROTATIONAL; + if (wait_for_completion_interruptible(&ub->completion) != 0) return -EINTR; diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index b1a3c293528519..13a2f24f176628 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -1451,7 +1451,9 @@ static int virtblk_read_limits(struct virtio_blk *vblk, static int virtblk_probe(struct virtio_device *vdev) { struct virtio_blk *vblk; - struct queue_limits lim = { }; + struct queue_limits lim = { + .features = BLK_FEAT_ROTATIONAL, + }; int err, index; unsigned int queue_depth; diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c index 9aafce3e5987bf..fa3a2ba525458b 100644 --- a/drivers/block/xen-blkfront.c +++ b/drivers/block/xen-blkfront.c @@ -1146,7 +1146,6 @@ static int xlvbd_alloc_gendisk(blkif_sector_t capacity, err = PTR_ERR(gd); goto out_free_tag_set; } - blk_queue_flag_set(QUEUE_FLAG_VIRT, gd->queue); strcpy(gd->disk_name, DEV_NAME); ptr = encode_disk_name(gd->disk_name + sizeof(DEV_NAME) - 1, offset); diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c index 3acd7006ad2ccd..aad840fc7e18e3 100644 --- a/drivers/block/zram/zram_drv.c +++ b/drivers/block/zram/zram_drv.c @@ -2245,8 +2245,6 @@ static int zram_add(void) /* Actual capacity set using sysfs (/sys/block/zram/disksize */ set_capacity(zram->disk, 0); - /* zram devices sort of resembles non-rotational disks */ - blk_queue_flag_set(QUEUE_FLAG_NONROT, zram->disk->queue); blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, zram->disk->queue); blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, zram->disk->queue); ret = device_add_disk(NULL, zram->disk, zram_disk_groups); diff --git a/drivers/cdrom/gdrom.c b/drivers/cdrom/gdrom.c index eefdd422ad8e9f..71cfe7a85913c4 100644 --- a/drivers/cdrom/gdrom.c +++ b/drivers/cdrom/gdrom.c @@ -744,6 +744,7 @@ static int probe_gdrom(struct platform_device *devptr) .max_segments = 1, /* set a large max size to get most from DMA */ .max_segment_size = 0x40000, + .features = BLK_FEAT_ROTATIONAL, }; int err; diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c index cb6595c8b5514e..baa364eedd0051 100644 --- a/drivers/md/bcache/super.c +++ b/drivers/md/bcache/super.c @@ -974,8 +974,6 @@ static int bcache_device_init(struct bcache_device *d, unsigned int block_size, d->disk->minors = BCACHE_MINORS; d->disk->fops = ops; d->disk->private_data = d; - - blk_queue_flag_set(QUEUE_FLAG_NONROT, d->disk->queue); return 0; out_bioset_exit: diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index 03abdae646829c..c062af32970934 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -1716,12 +1716,6 @@ static int device_dax_write_cache_enabled(struct dm_target *ti, return false; } -static int device_is_rotational(struct dm_target *ti, struct dm_dev *dev, - sector_t start, sector_t len, void *data) -{ - return !bdev_nonrot(dev->bdev); -} - static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev, sector_t start, sector_t len, void *data) { @@ -1870,12 +1864,6 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, if (dm_table_any_dev_attr(t, device_dax_write_cache_enabled, NULL)) dax_write_cache(t->md->dax_dev, true); - /* Ensure that all underlying devices are non-rotational. */ - if (dm_table_any_dev_attr(t, device_is_rotational, NULL)) - blk_queue_flag_clear(QUEUE_FLAG_NONROT, q); - else - blk_queue_flag_set(QUEUE_FLAG_NONROT, q); - /* * Some devices don't use blk_integrity but still want stable pages * because they do their own checksumming. diff --git a/drivers/md/md.c b/drivers/md/md.c index 2f4c5d1755d857..c23423c51fb7c2 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -6151,20 +6151,7 @@ int md_run(struct mddev *mddev) if (!mddev_is_dm(mddev)) { struct request_queue *q = mddev->gendisk->queue; - bool nonrot = true; - rdev_for_each(rdev, mddev) { - if (rdev->raid_disk >= 0 && !bdev_nonrot(rdev->bdev)) { - nonrot = false; - break; - } - } - if (mddev->degraded) - nonrot = false; - if (nonrot) - blk_queue_flag_set(QUEUE_FLAG_NONROT, q); - else - blk_queue_flag_clear(QUEUE_FLAG_NONROT, q); blk_queue_flag_set(QUEUE_FLAG_IO_STAT, q); /* Set the NOWAIT flags if all underlying devices support it */ diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c index 97ff993d31570c..b4f62fa845864c 100644 --- a/drivers/mmc/core/queue.c +++ b/drivers/mmc/core/queue.c @@ -387,7 +387,6 @@ static struct gendisk *mmc_alloc_disk(struct mmc_queue *mq, blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, mq->queue); blk_queue_rq_timeout(mq->queue, 60 * HZ); - blk_queue_flag_set(QUEUE_FLAG_NONROT, mq->queue); blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, mq->queue); dma_set_max_seg_size(mmc_dev(host), queue_max_segment_size(mq->queue)); diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c index 1b9f57f231e8be..bf8369ce7ddf1d 100644 --- a/drivers/mtd/mtd_blkdevs.c +++ b/drivers/mtd/mtd_blkdevs.c @@ -375,7 +375,6 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new) spin_lock_init(&new->queue_lock); INIT_LIST_HEAD(&new->rq_list); - blk_queue_flag_set(QUEUE_FLAG_NONROT, new->rq); blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, new->rq); gd->queue = new->rq; diff --git a/drivers/nvdimm/btt.c b/drivers/nvdimm/btt.c index c5f8451b494d6c..e474afa8e9f68d 100644 --- a/drivers/nvdimm/btt.c +++ b/drivers/nvdimm/btt.c @@ -1518,7 +1518,6 @@ static int btt_blk_init(struct btt *btt) btt->btt_disk->fops = &btt_fops; btt->btt_disk->private_data = btt; - blk_queue_flag_set(QUEUE_FLAG_NONROT, btt->btt_disk->queue); blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, btt->btt_disk->queue); set_capacity(btt->btt_disk, btt->nlba * btt->sector_size >> 9); diff --git a/drivers/nvdimm/pmem.c b/drivers/nvdimm/pmem.c index aff818469c114c..501cf226df0187 100644 --- a/drivers/nvdimm/pmem.c +++ b/drivers/nvdimm/pmem.c @@ -546,7 +546,6 @@ static int pmem_attach_disk(struct device *dev, } pmem->virt_addr = addr; - blk_queue_flag_set(QUEUE_FLAG_NONROT, q); blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, q); if (pmem->pfn_flags & PFN_MAP) blk_queue_flag_set(QUEUE_FLAG_DAX, q); diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 9fc5e36fe2e55e..0d753fe71f35b0 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -3744,7 +3744,6 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, struct nvme_ns_info *info) if (ctrl->opts && ctrl->opts->data_digest) blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, ns->queue); - blk_queue_flag_set(QUEUE_FLAG_NONROT, ns->queue); if (ctrl->ops->supports_pci_p2pdma && ctrl->ops->supports_pci_p2pdma(ctrl)) blk_queue_flag_set(QUEUE_FLAG_PCI_P2PDMA, ns->queue); diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c index 3d0e23a0a4ddd8..58c13304e558e0 100644 --- a/drivers/nvme/host/multipath.c +++ b/drivers/nvme/host/multipath.c @@ -549,7 +549,6 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head) sprintf(head->disk->disk_name, "nvme%dn%d", ctrl->subsys->instance, head->instance); - blk_queue_flag_set(QUEUE_FLAG_NONROT, head->disk->queue); blk_queue_flag_set(QUEUE_FLAG_NOWAIT, head->disk->queue); blk_queue_flag_set(QUEUE_FLAG_IO_STAT, head->disk->queue); /* diff --git a/drivers/s390/block/dasd_genhd.c b/drivers/s390/block/dasd_genhd.c index 4533dd055ca8e3..1aa426b1deddc7 100644 --- a/drivers/s390/block/dasd_genhd.c +++ b/drivers/s390/block/dasd_genhd.c @@ -68,7 +68,6 @@ int dasd_gendisk_alloc(struct dasd_block *block) blk_mq_free_tag_set(&block->tag_set); return PTR_ERR(gdp); } - blk_queue_flag_set(QUEUE_FLAG_NONROT, gdp->queue); /* Initialize gendisk structure. */ gdp->major = DASD_MAJOR; diff --git a/drivers/s390/block/scm_blk.c b/drivers/s390/block/scm_blk.c index 1d456a5a3bfb8e..2e2309fa9a0b34 100644 --- a/drivers/s390/block/scm_blk.c +++ b/drivers/s390/block/scm_blk.c @@ -475,7 +475,6 @@ int scm_blk_dev_setup(struct scm_blk_dev *bdev, struct scm_device *scmdev) goto out_tag; } rq = bdev->rq = bdev->gendisk->queue; - blk_queue_flag_set(QUEUE_FLAG_NONROT, rq); blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, rq); bdev->gendisk->private_data = scmdev; diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index d8ee4a4d4a6283..a42c3c45e86830 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -3318,7 +3318,7 @@ static void sd_read_block_characteristics(struct scsi_disk *sdkp, rcu_read_unlock(); if (rot == 1) { - blk_queue_flag_set(QUEUE_FLAG_NONROT, q); + lim->features &= ~BLK_FEAT_ROTATIONAL; blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q); } @@ -3646,7 +3646,7 @@ static int sd_revalidate_disk(struct gendisk *disk) * cause this to be updated correctly and any device which * doesn't support it should be treated as rotational. */ - blk_queue_flag_clear(QUEUE_FLAG_NONROT, q); + lim.features |= BLK_FEAT_ROTATIONAL; blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q); if (scsi_device_supports_vpd(sdp)) { diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index acdfe5122faa44..988e3248cffeb7 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -289,14 +289,16 @@ enum { /* supports passing on the FUA bit */ BLK_FEAT_FUA = (1u << 1), + + /* rotational device (hard drive or floppy) */ + BLK_FEAT_ROTATIONAL = (1u << 2), }; /* * Flags automatically inherited when stacking limits. */ #define BLK_FEAT_INHERIT_MASK \ - (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA) - + (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL) /* internal flags in queue_limits.flags */ enum { @@ -553,8 +555,6 @@ struct request_queue { #define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */ #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */ #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */ -#define QUEUE_FLAG_NONROT 6 /* non-rotational device (SSD) */ -#define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */ #define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */ #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */ #define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */ @@ -589,7 +589,7 @@ bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) #define blk_queue_noxmerges(q) \ test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) -#define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) +#define blk_queue_nonrot(q) ((q)->limits.features & BLK_FEAT_ROTATIONAL) #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) #define blk_queue_zone_resetall(q) \ From 39a9f1c334f9f27b3b3e6d0005c10ed667268346 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:42 +0200 Subject: [PATCH 094/209] block: move the add_random flag to queue_limits Move the add_random flag into the queue_limits feature field so that it can be set atomically with the queue frozen. Note that this also removes code from dm to clear the flag based on the underlying devices, which can't be reached as dm devices will always start out without the flag set. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-16-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-mq-debugfs.c | 1 - block/blk-sysfs.c | 6 +++--- drivers/block/mtip32xx/mtip32xx.c | 1 - drivers/md/dm-table.c | 18 ------------------ drivers/mmc/core/queue.c | 2 -- drivers/mtd/mtd_blkdevs.c | 3 --- drivers/s390/block/scm_blk.c | 4 ---- drivers/scsi/scsi_lib.c | 3 +-- drivers/scsi/sd.c | 11 +++-------- include/linux/blkdev.h | 5 +++-- 10 files changed, 10 insertions(+), 44 deletions(-) diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index 4d0e62ec88f033..6b7edb50bfd3fa 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -86,7 +86,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(FAIL_IO), QUEUE_FLAG_NAME(IO_STAT), QUEUE_FLAG_NAME(NOXMERGES), - QUEUE_FLAG_NAME(ADD_RANDOM), QUEUE_FLAG_NAME(SYNCHRONOUS), QUEUE_FLAG_NAME(SAME_FORCE), QUEUE_FLAG_NAME(INIT_DONE), diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 637ed3bbbfb46f..9174aca3b85526 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -323,7 +323,7 @@ queue_##name##_store(struct request_queue *q, const char *page, size_t count) \ } QUEUE_SYSFS_FEATURE(rotational, BLK_FEAT_ROTATIONAL) -QUEUE_SYSFS_BIT_FNS(random, ADD_RANDOM, 0); +QUEUE_SYSFS_FEATURE(add_random, BLK_FEAT_ADD_RANDOM) QUEUE_SYSFS_BIT_FNS(iostats, IO_STAT, 0); QUEUE_SYSFS_BIT_FNS(stable_writes, STABLE_WRITES, 0); #undef QUEUE_SYSFS_BIT_FNS @@ -561,7 +561,7 @@ static struct queue_sysfs_entry queue_hw_sector_size_entry = { QUEUE_RW_ENTRY(queue_rotational, "rotational"); QUEUE_RW_ENTRY(queue_iostats, "iostats"); -QUEUE_RW_ENTRY(queue_random, "add_random"); +QUEUE_RW_ENTRY(queue_add_random, "add_random"); QUEUE_RW_ENTRY(queue_stable_writes, "stable_writes"); #ifdef CONFIG_BLK_WBT @@ -665,7 +665,7 @@ static struct attribute *queue_attrs[] = { &queue_nomerges_entry.attr, &queue_iostats_entry.attr, &queue_stable_writes_entry.attr, - &queue_random_entry.attr, + &queue_add_random_entry.attr, &queue_poll_entry.attr, &queue_wc_entry.attr, &queue_fua_entry.attr, diff --git a/drivers/block/mtip32xx/mtip32xx.c b/drivers/block/mtip32xx/mtip32xx.c index 1dbbf72659d549..c6ef0546ffc9d2 100644 --- a/drivers/block/mtip32xx/mtip32xx.c +++ b/drivers/block/mtip32xx/mtip32xx.c @@ -3485,7 +3485,6 @@ static int mtip_block_initialize(struct driver_data *dd) goto start_service_thread; /* Set device limits. */ - blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, dd->queue); dma_set_max_seg_size(&dd->pdev->dev, 0x400000); /* Set the capacity of the device in 512 byte sectors. */ diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index c062af32970934..0a3838e45affd4 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -1716,14 +1716,6 @@ static int device_dax_write_cache_enabled(struct dm_target *ti, return false; } -static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev, - sector_t start, sector_t len, void *data) -{ - struct request_queue *q = bdev_get_queue(dev->bdev); - - return !blk_queue_add_random(q); -} - static int device_not_write_zeroes_capable(struct dm_target *ti, struct dm_dev *dev, sector_t start, sector_t len, void *data) { @@ -1876,16 +1868,6 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, else blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, q); - /* - * Determine whether or not this queue's I/O timings contribute - * to the entropy pool, Only request-based targets use this. - * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not - * have it set. - */ - if (blk_queue_add_random(q) && - dm_table_any_dev_attr(t, device_is_not_random, NULL)) - blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q); - /* * For a zoned target, setup the zones related queue attributes * and resources necessary for zone append emulation if necessary. diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c index b4f62fa845864c..da00904d4a3c7e 100644 --- a/drivers/mmc/core/queue.c +++ b/drivers/mmc/core/queue.c @@ -387,8 +387,6 @@ static struct gendisk *mmc_alloc_disk(struct mmc_queue *mq, blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, mq->queue); blk_queue_rq_timeout(mq->queue, 60 * HZ); - blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, mq->queue); - dma_set_max_seg_size(mmc_dev(host), queue_max_segment_size(mq->queue)); INIT_WORK(&mq->recovery_work, mmc_mq_recovery_handler); diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c index bf8369ce7ddf1d..47ead84407cdcf 100644 --- a/drivers/mtd/mtd_blkdevs.c +++ b/drivers/mtd/mtd_blkdevs.c @@ -374,9 +374,6 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new) /* Create the request queue */ spin_lock_init(&new->queue_lock); INIT_LIST_HEAD(&new->rq_list); - - blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, new->rq); - gd->queue = new->rq; if (new->readonly) diff --git a/drivers/s390/block/scm_blk.c b/drivers/s390/block/scm_blk.c index 2e2309fa9a0b34..3fcfe029db1b3a 100644 --- a/drivers/s390/block/scm_blk.c +++ b/drivers/s390/block/scm_blk.c @@ -439,7 +439,6 @@ int scm_blk_dev_setup(struct scm_blk_dev *bdev, struct scm_device *scmdev) .logical_block_size = 1 << 12, }; unsigned int devindex; - struct request_queue *rq; int len, ret; lim.max_segments = min(scmdev->nr_max_block, @@ -474,9 +473,6 @@ int scm_blk_dev_setup(struct scm_blk_dev *bdev, struct scm_device *scmdev) ret = PTR_ERR(bdev->gendisk); goto out_tag; } - rq = bdev->rq = bdev->gendisk->queue; - blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, rq); - bdev->gendisk->private_data = scmdev; bdev->gendisk->fops = &scm_blk_devops; bdev->gendisk->major = scm_major; diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c index ec39acc986d6ec..54f771ec8cfb5e 100644 --- a/drivers/scsi/scsi_lib.c +++ b/drivers/scsi/scsi_lib.c @@ -631,8 +631,7 @@ static bool scsi_end_request(struct request *req, blk_status_t error, if (blk_update_request(req, error, bytes)) return true; - // XXX: - if (blk_queue_add_random(q)) + if (q->limits.features & BLK_FEAT_ADD_RANDOM) add_disk_randomness(req->q->disk); WARN_ON_ONCE(!blk_rq_is_passthrough(req) && diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index a42c3c45e86830..a27f1c7f1b61d5 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -3301,7 +3301,6 @@ static void sd_read_block_limits_ext(struct scsi_disk *sdkp) static void sd_read_block_characteristics(struct scsi_disk *sdkp, struct queue_limits *lim) { - struct request_queue *q = sdkp->disk->queue; struct scsi_vpd *vpd; u16 rot; @@ -3317,10 +3316,8 @@ static void sd_read_block_characteristics(struct scsi_disk *sdkp, sdkp->zoned = (vpd->data[8] >> 4) & 3; rcu_read_unlock(); - if (rot == 1) { - lim->features &= ~BLK_FEAT_ROTATIONAL; - blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q); - } + if (rot == 1) + lim->features &= ~(BLK_FEAT_ROTATIONAL | BLK_FEAT_ADD_RANDOM); if (!sdkp->first_scan) return; @@ -3599,7 +3596,6 @@ static int sd_revalidate_disk(struct gendisk *disk) { struct scsi_disk *sdkp = scsi_disk(disk); struct scsi_device *sdp = sdkp->device; - struct request_queue *q = sdkp->disk->queue; sector_t old_capacity = sdkp->capacity; struct queue_limits lim; unsigned char *buffer; @@ -3646,8 +3642,7 @@ static int sd_revalidate_disk(struct gendisk *disk) * cause this to be updated correctly and any device which * doesn't support it should be treated as rotational. */ - lim.features |= BLK_FEAT_ROTATIONAL; - blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q); + lim.features |= (BLK_FEAT_ROTATIONAL | BLK_FEAT_ADD_RANDOM); if (scsi_device_supports_vpd(sdp)) { sd_read_block_provisioning(sdkp); diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 988e3248cffeb7..cf1bbf566b2bcd 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -292,6 +292,9 @@ enum { /* rotational device (hard drive or floppy) */ BLK_FEAT_ROTATIONAL = (1u << 2), + + /* contributes to the random number pool */ + BLK_FEAT_ADD_RANDOM = (1u << 3), }; /* @@ -557,7 +560,6 @@ struct request_queue { #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */ #define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */ #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */ -#define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */ #define QUEUE_FLAG_SYNCHRONOUS 11 /* always completes in submit context */ #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ @@ -591,7 +593,6 @@ bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) #define blk_queue_nonrot(q) ((q)->limits.features & BLK_FEAT_ROTATIONAL) #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) -#define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) #define blk_queue_zone_resetall(q) \ test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags) #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags) From cdb2497918cc2929691408bac87b58433b45b6d3 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:43 +0200 Subject: [PATCH 095/209] block: move the io_stat flag setting to queue_limits Move the io_stat flag into the queue_limits feature field so that it can be set atomically with the queue frozen. Simplify md and dm to set the flag unconditionally instead of avoiding setting a simple flag for cases where it already is set by other means, which is a bit pointless. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-17-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-mq-debugfs.c | 1 - block/blk-mq.c | 6 +++++- block/blk-sysfs.c | 2 +- drivers/md/dm-table.c | 12 +++++++++--- drivers/md/dm.c | 13 +++---------- drivers/md/md.c | 5 ++--- drivers/nvme/host/multipath.c | 2 +- include/linux/blkdev.h | 9 +++++---- 8 files changed, 26 insertions(+), 24 deletions(-) diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index 6b7edb50bfd3fa..cbe99444ed1a54 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -84,7 +84,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(NOMERGES), QUEUE_FLAG_NAME(SAME_COMP), QUEUE_FLAG_NAME(FAIL_IO), - QUEUE_FLAG_NAME(IO_STAT), QUEUE_FLAG_NAME(NOXMERGES), QUEUE_FLAG_NAME(SYNCHRONOUS), QUEUE_FLAG_NAME(SAME_FORCE), diff --git a/block/blk-mq.c b/block/blk-mq.c index 58b0d6c7cc34d6..cf67dc13f7dd4c 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -4116,7 +4116,11 @@ struct request_queue *blk_mq_alloc_queue(struct blk_mq_tag_set *set, struct request_queue *q; int ret; - q = blk_alloc_queue(lim ? lim : &default_lim, set->numa_node); + if (!lim) + lim = &default_lim; + lim->features |= BLK_FEAT_IO_STAT; + + q = blk_alloc_queue(lim, set->numa_node); if (IS_ERR(q)) return q; q->queuedata = queuedata; diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 9174aca3b85526..6f58530fb3c08e 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -324,7 +324,7 @@ queue_##name##_store(struct request_queue *q, const char *page, size_t count) \ QUEUE_SYSFS_FEATURE(rotational, BLK_FEAT_ROTATIONAL) QUEUE_SYSFS_FEATURE(add_random, BLK_FEAT_ADD_RANDOM) -QUEUE_SYSFS_BIT_FNS(iostats, IO_STAT, 0); +QUEUE_SYSFS_FEATURE(iostats, BLK_FEAT_IO_STAT) QUEUE_SYSFS_BIT_FNS(stable_writes, STABLE_WRITES, 0); #undef QUEUE_SYSFS_BIT_FNS diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index 0a3838e45affd4..5d5431e531aea9 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -579,6 +579,12 @@ int dm_split_args(int *argc, char ***argvp, char *input) return 0; } +static void dm_set_stacking_limits(struct queue_limits *limits) +{ + blk_set_stacking_limits(limits); + limits->features |= BLK_FEAT_IO_STAT; +} + /* * Impose necessary and sufficient conditions on a devices's table such * that any incoming bio which respects its logical_block_size can be @@ -617,7 +623,7 @@ static int validate_hardware_logical_block_alignment(struct dm_table *t, for (i = 0; i < t->num_targets; i++) { ti = dm_table_get_target(t, i); - blk_set_stacking_limits(&ti_limits); + dm_set_stacking_limits(&ti_limits); /* combine all target devices' limits */ if (ti->type->iterate_devices) @@ -1591,7 +1597,7 @@ int dm_calculate_queue_limits(struct dm_table *t, unsigned int zone_sectors = 0; bool zoned = false; - blk_set_stacking_limits(limits); + dm_set_stacking_limits(limits); t->integrity_supported = true; for (unsigned int i = 0; i < t->num_targets; i++) { @@ -1604,7 +1610,7 @@ int dm_calculate_queue_limits(struct dm_table *t, for (unsigned int i = 0; i < t->num_targets; i++) { struct dm_target *ti = dm_table_get_target(t, i); - blk_set_stacking_limits(&ti_limits); + dm_set_stacking_limits(&ti_limits); if (!ti->type->iterate_devices) { /* Set I/O hints portion of queue limits */ diff --git a/drivers/md/dm.c b/drivers/md/dm.c index 13037d6a6f62a2..8a976cee448bed 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -2386,22 +2386,15 @@ int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t) struct table_device *td; int r; - switch (type) { - case DM_TYPE_REQUEST_BASED: + WARN_ON_ONCE(type == DM_TYPE_NONE); + + if (type == DM_TYPE_REQUEST_BASED) { md->disk->fops = &dm_rq_blk_dops; r = dm_mq_init_request_queue(md, t); if (r) { DMERR("Cannot initialize queue for request-based dm mapped device"); return r; } - break; - case DM_TYPE_BIO_BASED: - case DM_TYPE_DAX_BIO_BASED: - blk_queue_flag_set(QUEUE_FLAG_IO_STAT, md->queue); - break; - case DM_TYPE_NONE: - WARN_ON_ONCE(true); - break; } r = dm_calculate_queue_limits(t, &limits); diff --git a/drivers/md/md.c b/drivers/md/md.c index c23423c51fb7c2..8db0db8d5a27ac 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -5787,7 +5787,8 @@ struct mddev *md_alloc(dev_t dev, char *name) int unit; int error; struct queue_limits lim = { - .features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA, + .features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | + BLK_FEAT_IO_STAT, }; /* @@ -6152,8 +6153,6 @@ int md_run(struct mddev *mddev) if (!mddev_is_dm(mddev)) { struct request_queue *q = mddev->gendisk->queue; - blk_queue_flag_set(QUEUE_FLAG_IO_STAT, q); - /* Set the NOWAIT flags if all underlying devices support it */ if (nowait) blk_queue_flag_set(QUEUE_FLAG_NOWAIT, q); diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c index 58c13304e558e0..eea727cfa9e67d 100644 --- a/drivers/nvme/host/multipath.c +++ b/drivers/nvme/host/multipath.c @@ -538,6 +538,7 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head) blk_set_stacking_limits(&lim); lim.dma_alignment = 3; + lim.features |= BLK_FEAT_IO_STAT; if (head->ids.csi != NVME_CSI_ZNS) lim.max_zone_append_sectors = 0; @@ -550,7 +551,6 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head) ctrl->subsys->instance, head->instance); blk_queue_flag_set(QUEUE_FLAG_NOWAIT, head->disk->queue); - blk_queue_flag_set(QUEUE_FLAG_IO_STAT, head->disk->queue); /* * This assumes all controllers that refer to a namespace either * support poll queues or not. That is not a strict guarantee, diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index cf1bbf566b2bcd..5fafb2f95fd1a3 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -295,6 +295,9 @@ enum { /* contributes to the random number pool */ BLK_FEAT_ADD_RANDOM = (1u << 3), + + /* do disk/partitions IO accounting */ + BLK_FEAT_IO_STAT = (1u << 4), }; /* @@ -558,7 +561,6 @@ struct request_queue { #define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */ #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */ #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */ -#define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */ #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */ #define QUEUE_FLAG_SYNCHRONOUS 11 /* always completes in submit context */ #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ @@ -577,8 +579,7 @@ struct request_queue { #define QUEUE_FLAG_SQ_SCHED 30 /* single queue style io dispatch */ #define QUEUE_FLAG_SKIP_TAGSET_QUIESCE 31 /* quiesce_tagset skip the queue*/ -#define QUEUE_FLAG_MQ_DEFAULT ((1UL << QUEUE_FLAG_IO_STAT) | \ - (1UL << QUEUE_FLAG_SAME_COMP) | \ +#define QUEUE_FLAG_MQ_DEFAULT ((1UL << QUEUE_FLAG_SAME_COMP) | \ (1UL << QUEUE_FLAG_NOWAIT)) void blk_queue_flag_set(unsigned int flag, struct request_queue *q); @@ -592,7 +593,7 @@ bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); #define blk_queue_noxmerges(q) \ test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) #define blk_queue_nonrot(q) ((q)->limits.features & BLK_FEAT_ROTATIONAL) -#define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) +#define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) #define blk_queue_zone_resetall(q) \ test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags) #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags) From 1a02f3a73f8c670eddeb44bf52a75ae7f67cfc11 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:44 +0200 Subject: [PATCH 096/209] block: move the stable_writes flag to queue_limits Move the stable_writes flag into the queue_limits feature field so that it can be set atomically with the queue frozen. The flag is now inherited by blk_stack_limits, which greatly simplifies the code in dm, and fixed md which previously did not pass on the flag set on lower devices. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-18-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-mq-debugfs.c | 1 - block/blk-sysfs.c | 29 +---------------------------- drivers/block/drbd/drbd_main.c | 5 ++--- drivers/block/rbd.c | 9 +++------ drivers/block/zram/zram_drv.c | 2 +- drivers/md/dm-table.c | 19 ------------------- drivers/md/raid5.c | 6 ++++-- drivers/mmc/core/queue.c | 5 +++-- drivers/nvme/host/core.c | 9 +++++---- drivers/nvme/host/multipath.c | 4 ---- drivers/scsi/iscsi_tcp.c | 8 ++++---- include/linux/blkdev.h | 9 ++++++--- 12 files changed, 29 insertions(+), 77 deletions(-) diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index cbe99444ed1a54..eb73f1d348e5a9 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -88,7 +88,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(SYNCHRONOUS), QUEUE_FLAG_NAME(SAME_FORCE), QUEUE_FLAG_NAME(INIT_DONE), - QUEUE_FLAG_NAME(STABLE_WRITES), QUEUE_FLAG_NAME(POLL), QUEUE_FLAG_NAME(DAX), QUEUE_FLAG_NAME(STATS), diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 6f58530fb3c08e..cde525724831ef 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -296,37 +296,10 @@ static ssize_t queue_##_name##_store(struct request_queue *q, \ return queue_feature_store(q, page, count, _feature); \ } -#define QUEUE_SYSFS_BIT_FNS(name, flag, neg) \ -static ssize_t \ -queue_##name##_show(struct request_queue *q, char *page) \ -{ \ - int bit; \ - bit = test_bit(QUEUE_FLAG_##flag, &q->queue_flags); \ - return queue_var_show(neg ? !bit : bit, page); \ -} \ -static ssize_t \ -queue_##name##_store(struct request_queue *q, const char *page, size_t count) \ -{ \ - unsigned long val; \ - ssize_t ret; \ - ret = queue_var_store(&val, page, count); \ - if (ret < 0) \ - return ret; \ - if (neg) \ - val = !val; \ - \ - if (val) \ - blk_queue_flag_set(QUEUE_FLAG_##flag, q); \ - else \ - blk_queue_flag_clear(QUEUE_FLAG_##flag, q); \ - return ret; \ -} - QUEUE_SYSFS_FEATURE(rotational, BLK_FEAT_ROTATIONAL) QUEUE_SYSFS_FEATURE(add_random, BLK_FEAT_ADD_RANDOM) QUEUE_SYSFS_FEATURE(iostats, BLK_FEAT_IO_STAT) -QUEUE_SYSFS_BIT_FNS(stable_writes, STABLE_WRITES, 0); -#undef QUEUE_SYSFS_BIT_FNS +QUEUE_SYSFS_FEATURE(stable_writes, BLK_FEAT_STABLE_WRITES); static ssize_t queue_zoned_show(struct request_queue *q, char *page) { diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 2ef29a47807550..f92673f05c7abc 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2698,7 +2698,8 @@ enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsig */ .max_hw_sectors = DRBD_MAX_BIO_SIZE_SAFE >> 8, .features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | - BLK_FEAT_ROTATIONAL, + BLK_FEAT_ROTATIONAL | + BLK_FEAT_STABLE_WRITES, }; device = minor_to_device(minor); @@ -2737,8 +2738,6 @@ enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsig sprintf(disk->disk_name, "drbd%d", minor); disk->private_data = device; - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, disk->queue); - device->md_io.page = alloc_page(GFP_KERNEL); if (!device->md_io.page) goto out_no_io_page; diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c index ec1f1c7d4275cd..008e850555f41a 100644 --- a/drivers/block/rbd.c +++ b/drivers/block/rbd.c @@ -4949,7 +4949,6 @@ static const struct blk_mq_ops rbd_mq_ops = { static int rbd_init_disk(struct rbd_device *rbd_dev) { struct gendisk *disk; - struct request_queue *q; unsigned int objset_bytes = rbd_dev->layout.object_size * rbd_dev->layout.stripe_count; struct queue_limits lim = { @@ -4979,12 +4978,14 @@ static int rbd_init_disk(struct rbd_device *rbd_dev) lim.max_write_zeroes_sectors = objset_bytes >> SECTOR_SHIFT; } + if (!ceph_test_opt(rbd_dev->rbd_client->client, NOCRC)) + lim.features |= BLK_FEAT_STABLE_WRITES; + disk = blk_mq_alloc_disk(&rbd_dev->tag_set, &lim, rbd_dev); if (IS_ERR(disk)) { err = PTR_ERR(disk); goto out_tag_set; } - q = disk->queue; snprintf(disk->disk_name, sizeof(disk->disk_name), RBD_DRV_NAME "%d", rbd_dev->dev_id); @@ -4996,10 +4997,6 @@ static int rbd_init_disk(struct rbd_device *rbd_dev) disk->minors = RBD_MINORS_PER_MAJOR; disk->fops = &rbd_bd_ops; disk->private_data = rbd_dev; - - if (!ceph_test_opt(rbd_dev->rbd_client->client, NOCRC)) - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, q); - rbd_dev->disk = disk; return 0; diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c index aad840fc7e18e3..f8f1b5b54795ac 100644 --- a/drivers/block/zram/zram_drv.c +++ b/drivers/block/zram/zram_drv.c @@ -2208,6 +2208,7 @@ static int zram_add(void) #if ZRAM_LOGICAL_BLOCK_SIZE == PAGE_SIZE .max_write_zeroes_sectors = UINT_MAX, #endif + .features = BLK_FEAT_STABLE_WRITES, }; struct zram *zram; int ret, device_id; @@ -2246,7 +2247,6 @@ static int zram_add(void) /* Actual capacity set using sysfs (/sys/block/zram/disksize */ set_capacity(zram->disk, 0); blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, zram->disk->queue); - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, zram->disk->queue); ret = device_add_disk(NULL, zram->disk, zram_disk_groups); if (ret) goto out_cleanup_disk; diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index 5d5431e531aea9..aaf379cb15d91f 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -1819,13 +1819,6 @@ static bool dm_table_supports_secure_erase(struct dm_table *t) return true; } -static int device_requires_stable_pages(struct dm_target *ti, - struct dm_dev *dev, sector_t start, - sector_t len, void *data) -{ - return bdev_stable_writes(dev->bdev); -} - int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, struct queue_limits *limits) { @@ -1862,18 +1855,6 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, if (dm_table_any_dev_attr(t, device_dax_write_cache_enabled, NULL)) dax_write_cache(t->md->dax_dev, true); - /* - * Some devices don't use blk_integrity but still want stable pages - * because they do their own checksumming. - * If any underlying device requires stable pages, a table must require - * them as well. Only targets that support iterate_devices are considered: - * don't want error, zero, etc to require stable pages. - */ - if (dm_table_any_dev_attr(t, device_requires_stable_pages, NULL)) - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, q); - else - blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, q); - /* * For a zoned target, setup the zones related queue attributes * and resources necessary for zone append emulation if necessary. diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 675c68fa6c6403..e875763d69917d 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -7082,12 +7082,14 @@ raid5_store_skip_copy(struct mddev *mddev, const char *page, size_t len) err = -ENODEV; else if (new != conf->skip_copy) { struct request_queue *q = mddev->gendisk->queue; + struct queue_limits lim = queue_limits_start_update(q); conf->skip_copy = new; if (new) - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, q); + lim.features |= BLK_FEAT_STABLE_WRITES; else - blk_queue_flag_clear(QUEUE_FLAG_STABLE_WRITES, q); + lim.features &= ~BLK_FEAT_STABLE_WRITES; + err = queue_limits_commit_update(q, &lim); } mddev_unlock_and_resume(mddev); return err ?: len; diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c index da00904d4a3c7e..d0b3ca8a11f071 100644 --- a/drivers/mmc/core/queue.c +++ b/drivers/mmc/core/queue.c @@ -378,13 +378,14 @@ static struct gendisk *mmc_alloc_disk(struct mmc_queue *mq, lim.max_segments = host->max_segs; } + if (mmc_host_is_spi(host) && host->use_spi_crc) + lim.features |= BLK_FEAT_STABLE_WRITES; + disk = blk_mq_alloc_disk(&mq->tag_set, &lim, mq); if (IS_ERR(disk)) return disk; mq->queue = disk->queue; - if (mmc_host_is_spi(host) && host->use_spi_crc) - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, mq->queue); blk_queue_rq_timeout(mq->queue, 60 * HZ); dma_set_max_seg_size(mmc_dev(host), queue_max_segment_size(mq->queue)); diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 0d753fe71f35b0..5ecf762d7c8837 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -3724,6 +3724,7 @@ static void nvme_ns_add_to_ctrl_list(struct nvme_ns *ns) static void nvme_alloc_ns(struct nvme_ctrl *ctrl, struct nvme_ns_info *info) { + struct queue_limits lim = { }; struct nvme_ns *ns; struct gendisk *disk; int node = ctrl->numa_node; @@ -3732,7 +3733,10 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, struct nvme_ns_info *info) if (!ns) return; - disk = blk_mq_alloc_disk(ctrl->tagset, NULL, ns); + if (ctrl->opts && ctrl->opts->data_digest) + lim.features |= BLK_FEAT_STABLE_WRITES; + + disk = blk_mq_alloc_disk(ctrl->tagset, &lim, ns); if (IS_ERR(disk)) goto out_free_ns; disk->fops = &nvme_bdev_ops; @@ -3741,9 +3745,6 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, struct nvme_ns_info *info) ns->disk = disk; ns->queue = disk->queue; - if (ctrl->opts && ctrl->opts->data_digest) - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, ns->queue); - if (ctrl->ops->supports_pci_p2pdma && ctrl->ops->supports_pci_p2pdma(ctrl)) blk_queue_flag_set(QUEUE_FLAG_PCI_P2PDMA, ns->queue); diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c index eea727cfa9e67d..173796f2ddea9f 100644 --- a/drivers/nvme/host/multipath.c +++ b/drivers/nvme/host/multipath.c @@ -868,10 +868,6 @@ void nvme_mpath_add_disk(struct nvme_ns *ns, __le32 anagrpid) nvme_mpath_set_live(ns); } - if (test_bit(QUEUE_FLAG_STABLE_WRITES, &ns->queue->queue_flags) && - ns->head->disk) - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, - ns->head->disk->queue); #ifdef CONFIG_BLK_DEV_ZONED if (blk_queue_is_zoned(ns->queue) && ns->head->disk) ns->head->disk->nr_zones = ns->disk->nr_zones; diff --git a/drivers/scsi/iscsi_tcp.c b/drivers/scsi/iscsi_tcp.c index 60688f18fac6f7..c708e105963833 100644 --- a/drivers/scsi/iscsi_tcp.c +++ b/drivers/scsi/iscsi_tcp.c @@ -1057,15 +1057,15 @@ static umode_t iscsi_sw_tcp_attr_is_visible(int param_type, int param) return 0; } -static int iscsi_sw_tcp_slave_configure(struct scsi_device *sdev) +static int iscsi_sw_tcp_device_configure(struct scsi_device *sdev, + struct queue_limits *lim) { struct iscsi_sw_tcp_host *tcp_sw_host = iscsi_host_priv(sdev->host); struct iscsi_session *session = tcp_sw_host->session; struct iscsi_conn *conn = session->leadconn; if (conn->datadgst_en) - blk_queue_flag_set(QUEUE_FLAG_STABLE_WRITES, - sdev->request_queue); + lim->features |= BLK_FEAT_STABLE_WRITES; return 0; } @@ -1083,7 +1083,7 @@ static const struct scsi_host_template iscsi_sw_tcp_sht = { .eh_device_reset_handler= iscsi_eh_device_reset, .eh_target_reset_handler = iscsi_eh_recover_target, .dma_boundary = PAGE_SIZE - 1, - .slave_configure = iscsi_sw_tcp_slave_configure, + .device_configure = iscsi_sw_tcp_device_configure, .proc_name = "iscsi_tcp", .this_id = -1, .track_queue_depth = 1, diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 5fafb2f95fd1a3..8936eb6ba60956 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -298,13 +298,17 @@ enum { /* do disk/partitions IO accounting */ BLK_FEAT_IO_STAT = (1u << 4), + + /* don't modify data until writeback is done */ + BLK_FEAT_STABLE_WRITES = (1u << 5), }; /* * Flags automatically inherited when stacking limits. */ #define BLK_FEAT_INHERIT_MASK \ - (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL) + (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \ + BLK_FEAT_STABLE_WRITES) /* internal flags in queue_limits.flags */ enum { @@ -565,7 +569,6 @@ struct request_queue { #define QUEUE_FLAG_SYNCHRONOUS 11 /* always completes in submit context */ #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ -#define QUEUE_FLAG_STABLE_WRITES 15 /* don't modify blks until WB is done */ #define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */ #define QUEUE_FLAG_DAX 19 /* device supports DAX */ #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */ @@ -1323,7 +1326,7 @@ static inline bool bdev_stable_writes(struct block_device *bdev) if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && q->limits.integrity.csum_type != BLK_INTEGRITY_CSUM_NONE) return true; - return test_bit(QUEUE_FLAG_STABLE_WRITES, &q->queue_flags); + return q->limits.features & BLK_FEAT_STABLE_WRITES; } static inline bool blk_queue_write_cache(struct request_queue *q) From aadd5c59c910427c0464c217d5ed588ff14e2502 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:45 +0200 Subject: [PATCH 097/209] block: move the synchronous flag to queue_limits Move the synchronous flag into the queue_limits feature field so that it can be set atomically with the queue frozen. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-19-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-mq-debugfs.c | 1 - drivers/block/brd.c | 2 +- drivers/block/zram/zram_drv.c | 4 ++-- drivers/nvdimm/btt.c | 3 +-- drivers/nvdimm/pmem.c | 4 ++-- include/linux/blkdev.h | 7 ++++--- 6 files changed, 10 insertions(+), 11 deletions(-) diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index eb73f1d348e5a9..957774e40b1d0c 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -85,7 +85,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(SAME_COMP), QUEUE_FLAG_NAME(FAIL_IO), QUEUE_FLAG_NAME(NOXMERGES), - QUEUE_FLAG_NAME(SYNCHRONOUS), QUEUE_FLAG_NAME(SAME_FORCE), QUEUE_FLAG_NAME(INIT_DONE), QUEUE_FLAG_NAME(POLL), diff --git a/drivers/block/brd.c b/drivers/block/brd.c index b25dc463b5e3a6..d77deb571dbd06 100644 --- a/drivers/block/brd.c +++ b/drivers/block/brd.c @@ -335,6 +335,7 @@ static int brd_alloc(int i) .max_hw_discard_sectors = UINT_MAX, .max_discard_segments = 1, .discard_granularity = PAGE_SIZE, + .features = BLK_FEAT_SYNCHRONOUS, }; list_for_each_entry(brd, &brd_devices, brd_list) @@ -366,7 +367,6 @@ static int brd_alloc(int i) strscpy(disk->disk_name, buf, DISK_NAME_LEN); set_capacity(disk, rd_size * 2); - blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, disk->queue); blk_queue_flag_set(QUEUE_FLAG_NOWAIT, disk->queue); err = add_disk(disk); if (err) diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c index f8f1b5b54795ac..efcb8d9d274c31 100644 --- a/drivers/block/zram/zram_drv.c +++ b/drivers/block/zram/zram_drv.c @@ -2208,7 +2208,8 @@ static int zram_add(void) #if ZRAM_LOGICAL_BLOCK_SIZE == PAGE_SIZE .max_write_zeroes_sectors = UINT_MAX, #endif - .features = BLK_FEAT_STABLE_WRITES, + .features = BLK_FEAT_STABLE_WRITES | + BLK_FEAT_SYNCHRONOUS, }; struct zram *zram; int ret, device_id; @@ -2246,7 +2247,6 @@ static int zram_add(void) /* Actual capacity set using sysfs (/sys/block/zram/disksize */ set_capacity(zram->disk, 0); - blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, zram->disk->queue); ret = device_add_disk(NULL, zram->disk, zram_disk_groups); if (ret) goto out_cleanup_disk; diff --git a/drivers/nvdimm/btt.c b/drivers/nvdimm/btt.c index e474afa8e9f68d..e79c06d65bb77b 100644 --- a/drivers/nvdimm/btt.c +++ b/drivers/nvdimm/btt.c @@ -1501,6 +1501,7 @@ static int btt_blk_init(struct btt *btt) .logical_block_size = btt->sector_size, .max_hw_sectors = UINT_MAX, .max_integrity_segments = 1, + .features = BLK_FEAT_SYNCHRONOUS, }; int rc; @@ -1518,8 +1519,6 @@ static int btt_blk_init(struct btt *btt) btt->btt_disk->fops = &btt_fops; btt->btt_disk->private_data = btt; - blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, btt->btt_disk->queue); - set_capacity(btt->btt_disk, btt->nlba * btt->sector_size >> 9); rc = device_add_disk(&btt->nd_btt->dev, btt->btt_disk, NULL); if (rc) diff --git a/drivers/nvdimm/pmem.c b/drivers/nvdimm/pmem.c index 501cf226df0187..b821dcf018f6ae 100644 --- a/drivers/nvdimm/pmem.c +++ b/drivers/nvdimm/pmem.c @@ -455,7 +455,8 @@ static int pmem_attach_disk(struct device *dev, .logical_block_size = pmem_sector_size(ndns), .physical_block_size = PAGE_SIZE, .max_hw_sectors = UINT_MAX, - .features = BLK_FEAT_WRITE_CACHE, + .features = BLK_FEAT_WRITE_CACHE | + BLK_FEAT_SYNCHRONOUS, }; int nid = dev_to_node(dev), fua; struct resource *res = &nsio->res; @@ -546,7 +547,6 @@ static int pmem_attach_disk(struct device *dev, } pmem->virt_addr = addr; - blk_queue_flag_set(QUEUE_FLAG_SYNCHRONOUS, q); if (pmem->pfn_flags & PFN_MAP) blk_queue_flag_set(QUEUE_FLAG_DAX, q); diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 8936eb6ba60956..cee7b44a142513 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -301,6 +301,9 @@ enum { /* don't modify data until writeback is done */ BLK_FEAT_STABLE_WRITES = (1u << 5), + + /* always completes in submit context */ + BLK_FEAT_SYNCHRONOUS = (1u << 6), }; /* @@ -566,7 +569,6 @@ struct request_queue { #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */ #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */ #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */ -#define QUEUE_FLAG_SYNCHRONOUS 11 /* always completes in submit context */ #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ #define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */ @@ -1315,8 +1317,7 @@ static inline bool bdev_nonrot(struct block_device *bdev) static inline bool bdev_synchronous(struct block_device *bdev) { - return test_bit(QUEUE_FLAG_SYNCHRONOUS, - &bdev_get_queue(bdev)->queue_flags); + return bdev->bd_disk->queue->limits.features & BLK_FEAT_SYNCHRONOUS; } static inline bool bdev_stable_writes(struct block_device *bdev) From f76af42f8bf13d2620084f305f01691de9238fc7 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:46 +0200 Subject: [PATCH 098/209] block: move the nowait flag to queue_limits Move the nowait flag into the queue_limits feature field so that it can be set atomically with the queue frozen. Stacking drivers are simplified in that they now can simply set the flag, and blk_stack_limits will clear it when the features is not supported by any of the underlying devices. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-20-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-mq-debugfs.c | 1 - block/blk-mq.c | 2 +- block/blk-settings.c | 9 +++++++++ drivers/block/brd.c | 4 ++-- drivers/md/dm-table.c | 18 +++--------------- drivers/md/md.c | 18 +----------------- drivers/nvme/host/multipath.c | 3 +-- include/linux/blkdev.h | 9 +++++---- 8 files changed, 22 insertions(+), 42 deletions(-) diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index 957774e40b1d0c..62b132e9a9ce3b 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -96,7 +96,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(ZONE_RESETALL), QUEUE_FLAG_NAME(RQ_ALLOC_TIME), QUEUE_FLAG_NAME(HCTX_ACTIVE), - QUEUE_FLAG_NAME(NOWAIT), QUEUE_FLAG_NAME(SQ_SCHED), QUEUE_FLAG_NAME(SKIP_TAGSET_QUIESCE), }; diff --git a/block/blk-mq.c b/block/blk-mq.c index cf67dc13f7dd4c..43235acc87505f 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -4118,7 +4118,7 @@ struct request_queue *blk_mq_alloc_queue(struct blk_mq_tag_set *set, if (!lim) lim = &default_lim; - lim->features |= BLK_FEAT_IO_STAT; + lim->features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT; q = blk_alloc_queue(lim, set->numa_node); if (IS_ERR(q)) diff --git a/block/blk-settings.c b/block/blk-settings.c index 536ee202fcdccb..bf4622c19b5c09 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -459,6 +459,15 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, t->features |= (b->features & BLK_FEAT_INHERIT_MASK); + /* + * BLK_FEAT_NOWAIT needs to be supported both by the stacking driver + * and all underlying devices. The stacking driver sets the flag + * before stacking the limits, and this will clear the flag if any + * of the underlying devices does not support it. + */ + if (!(b->features & BLK_FEAT_NOWAIT)) + t->features &= ~BLK_FEAT_NOWAIT; + t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); t->max_user_sectors = min_not_zero(t->max_user_sectors, b->max_user_sectors); diff --git a/drivers/block/brd.c b/drivers/block/brd.c index d77deb571dbd06..a300645cd9d4a5 100644 --- a/drivers/block/brd.c +++ b/drivers/block/brd.c @@ -335,7 +335,8 @@ static int brd_alloc(int i) .max_hw_discard_sectors = UINT_MAX, .max_discard_segments = 1, .discard_granularity = PAGE_SIZE, - .features = BLK_FEAT_SYNCHRONOUS, + .features = BLK_FEAT_SYNCHRONOUS | + BLK_FEAT_NOWAIT, }; list_for_each_entry(brd, &brd_devices, brd_list) @@ -367,7 +368,6 @@ static int brd_alloc(int i) strscpy(disk->disk_name, buf, DISK_NAME_LEN); set_capacity(disk, rd_size * 2); - blk_queue_flag_set(QUEUE_FLAG_NOWAIT, disk->queue); err = add_disk(disk); if (err) goto out_cleanup_disk; diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index aaf379cb15d91f..84d636712c7284 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -582,7 +582,7 @@ int dm_split_args(int *argc, char ***argvp, char *input) static void dm_set_stacking_limits(struct queue_limits *limits) { blk_set_stacking_limits(limits); - limits->features |= BLK_FEAT_IO_STAT; + limits->features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT; } /* @@ -1746,12 +1746,6 @@ static bool dm_table_supports_write_zeroes(struct dm_table *t) return true; } -static int device_not_nowait_capable(struct dm_target *ti, struct dm_dev *dev, - sector_t start, sector_t len, void *data) -{ - return !bdev_nowait(dev->bdev); -} - static bool dm_table_supports_nowait(struct dm_table *t) { for (unsigned int i = 0; i < t->num_targets; i++) { @@ -1759,10 +1753,6 @@ static bool dm_table_supports_nowait(struct dm_table *t) if (!dm_target_supports_nowait(ti->type)) return false; - - if (!ti->type->iterate_devices || - ti->type->iterate_devices(ti, device_not_nowait_capable, NULL)) - return false; } return true; @@ -1824,10 +1814,8 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, { int r; - if (dm_table_supports_nowait(t)) - blk_queue_flag_set(QUEUE_FLAG_NOWAIT, q); - else - blk_queue_flag_clear(QUEUE_FLAG_NOWAIT, q); + if (!dm_table_supports_nowait(t)) + limits->features &= ~BLK_FEAT_NOWAIT; if (!dm_table_supports_discards(t)) { limits->max_hw_discard_sectors = 0; diff --git a/drivers/md/md.c b/drivers/md/md.c index 8db0db8d5a27ac..f1c7d4f281c521 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -5788,7 +5788,7 @@ struct mddev *md_alloc(dev_t dev, char *name) int error; struct queue_limits lim = { .features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | - BLK_FEAT_IO_STAT, + BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT, }; /* @@ -6150,13 +6150,6 @@ int md_run(struct mddev *mddev) } } - if (!mddev_is_dm(mddev)) { - struct request_queue *q = mddev->gendisk->queue; - - /* Set the NOWAIT flags if all underlying devices support it */ - if (nowait) - blk_queue_flag_set(QUEUE_FLAG_NOWAIT, q); - } if (pers->sync_request) { if (mddev->kobj.sd && sysfs_create_group(&mddev->kobj, &md_redundancy_group)) @@ -7115,15 +7108,6 @@ static int hot_add_disk(struct mddev *mddev, dev_t dev) set_bit(MD_SB_CHANGE_DEVS, &mddev->sb_flags); if (!mddev->thread) md_update_sb(mddev, 1); - /* - * If the new disk does not support REQ_NOWAIT, - * disable on the whole MD. - */ - if (!bdev_nowait(rdev->bdev)) { - pr_info("%s: Disabling nowait because %pg does not support nowait\n", - mdname(mddev), rdev->bdev); - blk_queue_flag_clear(QUEUE_FLAG_NOWAIT, mddev->gendisk->queue); - } /* * Kick recovery, maybe this spare has to be added to the * array immediately. diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c index 173796f2ddea9f..61a162c9cf4e6c 100644 --- a/drivers/nvme/host/multipath.c +++ b/drivers/nvme/host/multipath.c @@ -538,7 +538,7 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head) blk_set_stacking_limits(&lim); lim.dma_alignment = 3; - lim.features |= BLK_FEAT_IO_STAT; + lim.features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT; if (head->ids.csi != NVME_CSI_ZNS) lim.max_zone_append_sectors = 0; @@ -550,7 +550,6 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head) sprintf(head->disk->disk_name, "nvme%dn%d", ctrl->subsys->instance, head->instance); - blk_queue_flag_set(QUEUE_FLAG_NOWAIT, head->disk->queue); /* * This assumes all controllers that refer to a namespace either * support poll queues or not. That is not a strict guarantee, diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index cee7b44a142513..f3d4519d609d95 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -304,6 +304,9 @@ enum { /* always completes in submit context */ BLK_FEAT_SYNCHRONOUS = (1u << 6), + + /* supports REQ_NOWAIT */ + BLK_FEAT_NOWAIT = (1u << 7), }; /* @@ -580,12 +583,10 @@ struct request_queue { #define QUEUE_FLAG_ZONE_RESETALL 26 /* supports Zone Reset All */ #define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */ #define QUEUE_FLAG_HCTX_ACTIVE 28 /* at least one blk-mq hctx is active */ -#define QUEUE_FLAG_NOWAIT 29 /* device supports NOWAIT */ #define QUEUE_FLAG_SQ_SCHED 30 /* single queue style io dispatch */ #define QUEUE_FLAG_SKIP_TAGSET_QUIESCE 31 /* quiesce_tagset skip the queue*/ -#define QUEUE_FLAG_MQ_DEFAULT ((1UL << QUEUE_FLAG_SAME_COMP) | \ - (1UL << QUEUE_FLAG_NOWAIT)) +#define QUEUE_FLAG_MQ_DEFAULT (1UL << QUEUE_FLAG_SAME_COMP) void blk_queue_flag_set(unsigned int flag, struct request_queue *q); void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); @@ -1348,7 +1349,7 @@ static inline bool bdev_fua(struct block_device *bdev) static inline bool bdev_nowait(struct block_device *bdev) { - return test_bit(QUEUE_FLAG_NOWAIT, &bdev_get_queue(bdev)->queue_flags); + return bdev->bd_disk->queue->limits.features & BLK_FEAT_NOWAIT; } static inline bool bdev_is_zoned(struct block_device *bdev) From f467fee48da4500786e145489787b37adae317c3 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:47 +0200 Subject: [PATCH 099/209] block: move the dax flag to queue_limits Move the dax flag into the queue_limits feature field so that it can be set atomically with the queue frozen. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-21-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-mq-debugfs.c | 1 - drivers/md/dm-table.c | 4 ++-- drivers/nvdimm/pmem.c | 7 ++----- drivers/s390/block/dcssblk.c | 2 +- include/linux/blkdev.h | 6 ++++-- 5 files changed, 9 insertions(+), 11 deletions(-) diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index 62b132e9a9ce3b..f4fa820251ce83 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -88,7 +88,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(SAME_FORCE), QUEUE_FLAG_NAME(INIT_DONE), QUEUE_FLAG_NAME(POLL), - QUEUE_FLAG_NAME(DAX), QUEUE_FLAG_NAME(STATS), QUEUE_FLAG_NAME(REGISTERED), QUEUE_FLAG_NAME(QUIESCED), diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index 84d636712c7284..e44697037e86f4 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -1834,11 +1834,11 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, limits->features |= BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA; if (dm_table_supports_dax(t, device_not_dax_capable)) { - blk_queue_flag_set(QUEUE_FLAG_DAX, q); + limits->features |= BLK_FEAT_DAX; if (dm_table_supports_dax(t, device_not_dax_synchronous_capable)) set_dax_synchronous(t->md->dax_dev); } else - blk_queue_flag_clear(QUEUE_FLAG_DAX, q); + limits->features &= ~BLK_FEAT_DAX; if (dm_table_any_dev_attr(t, device_dax_write_cache_enabled, NULL)) dax_write_cache(t->md->dax_dev, true); diff --git a/drivers/nvdimm/pmem.c b/drivers/nvdimm/pmem.c index b821dcf018f6ae..1dd74c969d5a09 100644 --- a/drivers/nvdimm/pmem.c +++ b/drivers/nvdimm/pmem.c @@ -465,7 +465,6 @@ static int pmem_attach_disk(struct device *dev, struct dax_device *dax_dev; struct nd_pfn_sb *pfn_sb; struct pmem_device *pmem; - struct request_queue *q; struct gendisk *disk; void *addr; int rc; @@ -499,6 +498,8 @@ static int pmem_attach_disk(struct device *dev, } if (fua) lim.features |= BLK_FEAT_FUA; + if (is_nd_pfn(dev)) + lim.features |= BLK_FEAT_DAX; if (!devm_request_mem_region(dev, res->start, resource_size(res), dev_name(&ndns->dev))) { @@ -509,7 +510,6 @@ static int pmem_attach_disk(struct device *dev, disk = blk_alloc_disk(&lim, nid); if (IS_ERR(disk)) return PTR_ERR(disk); - q = disk->queue; pmem->disk = disk; pmem->pgmap.owner = pmem; @@ -547,9 +547,6 @@ static int pmem_attach_disk(struct device *dev, } pmem->virt_addr = addr; - if (pmem->pfn_flags & PFN_MAP) - blk_queue_flag_set(QUEUE_FLAG_DAX, q); - disk->fops = &pmem_fops; disk->private_data = pmem; nvdimm_namespace_disk_name(ndns, disk->disk_name); diff --git a/drivers/s390/block/dcssblk.c b/drivers/s390/block/dcssblk.c index 6d1689a2717e5f..d5a5d11ae0dcdf 100644 --- a/drivers/s390/block/dcssblk.c +++ b/drivers/s390/block/dcssblk.c @@ -548,6 +548,7 @@ dcssblk_add_store(struct device *dev, struct device_attribute *attr, const char { struct queue_limits lim = { .logical_block_size = 4096, + .features = BLK_FEAT_DAX, }; int rc, i, j, num_of_segments; struct dcssblk_dev_info *dev_info; @@ -643,7 +644,6 @@ dcssblk_add_store(struct device *dev, struct device_attribute *attr, const char dev_info->gd->fops = &dcssblk_devops; dev_info->gd->private_data = dev_info; dev_info->gd->flags |= GENHD_FL_NO_PART; - blk_queue_flag_set(QUEUE_FLAG_DAX, dev_info->gd->queue); seg_byte_size = (dev_info->end - dev_info->start + 1); set_capacity(dev_info->gd, seg_byte_size >> 9); // size in sectors diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index f3d4519d609d95..7022e06a3dd9a3 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -307,6 +307,9 @@ enum { /* supports REQ_NOWAIT */ BLK_FEAT_NOWAIT = (1u << 7), + + /* supports DAX */ + BLK_FEAT_DAX = (1u << 8), }; /* @@ -575,7 +578,6 @@ struct request_queue { #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ #define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */ -#define QUEUE_FLAG_DAX 19 /* device supports DAX */ #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */ #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */ #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */ @@ -602,7 +604,7 @@ bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) #define blk_queue_zone_resetall(q) \ test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags) -#define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags) +#define blk_queue_dax(q) ((q)->limits.features & BLK_FEAT_DAX) #define blk_queue_pci_p2pdma(q) \ test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags) #ifdef CONFIG_BLK_RQ_ALLOC_TIME From 8023e144f9d6e35f8786937e2f0c2fea0aba6dbc Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:48 +0200 Subject: [PATCH 100/209] block: move the poll flag to queue_limits Move the poll flag into the queue_limits feature field so that it can be set atomically with the queue frozen. Stacking drivers are simplified in that they now can simply set the flag, and blk_stack_limits will clear it when the features is not supported by any of the underlying devices. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-22-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-core.c | 5 ++-- block/blk-mq-debugfs.c | 1 - block/blk-mq.c | 31 +++++++++++--------- block/blk-settings.c | 10 ++++--- block/blk-sysfs.c | 4 +-- drivers/md/dm-table.c | 54 +++++++++-------------------------- drivers/nvme/host/multipath.c | 12 +------- include/linux/blkdev.h | 4 ++- 8 files changed, 45 insertions(+), 76 deletions(-) diff --git a/block/blk-core.c b/block/blk-core.c index 2b45a4df9a1aa1..8d9fbd353fc7fc 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -791,7 +791,7 @@ void submit_bio_noacct(struct bio *bio) } } - if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags)) + if (!(q->limits.features & BLK_FEAT_POLL)) bio_clear_polled(bio); switch (bio_op(bio)) { @@ -915,8 +915,7 @@ int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags) return 0; q = bdev_get_queue(bdev); - if (cookie == BLK_QC_T_NONE || - !test_bit(QUEUE_FLAG_POLL, &q->queue_flags)) + if (cookie == BLK_QC_T_NONE || !(q->limits.features & BLK_FEAT_POLL)) return 0; blk_flush_plug(current->plug, false); diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index f4fa820251ce83..3a21527913840d 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -87,7 +87,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(NOXMERGES), QUEUE_FLAG_NAME(SAME_FORCE), QUEUE_FLAG_NAME(INIT_DONE), - QUEUE_FLAG_NAME(POLL), QUEUE_FLAG_NAME(STATS), QUEUE_FLAG_NAME(REGISTERED), QUEUE_FLAG_NAME(QUIESCED), diff --git a/block/blk-mq.c b/block/blk-mq.c index 43235acc87505f..e2b9710ddc5ad1 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -4109,6 +4109,12 @@ void blk_mq_release(struct request_queue *q) blk_mq_sysfs_deinit(q); } +static bool blk_mq_can_poll(struct blk_mq_tag_set *set) +{ + return set->nr_maps > HCTX_TYPE_POLL && + set->map[HCTX_TYPE_POLL].nr_queues; +} + struct request_queue *blk_mq_alloc_queue(struct blk_mq_tag_set *set, struct queue_limits *lim, void *queuedata) { @@ -4119,6 +4125,8 @@ struct request_queue *blk_mq_alloc_queue(struct blk_mq_tag_set *set, if (!lim) lim = &default_lim; lim->features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT; + if (blk_mq_can_poll(set)) + lim->features |= BLK_FEAT_POLL; q = blk_alloc_queue(lim, set->numa_node); if (IS_ERR(q)) @@ -4273,17 +4281,6 @@ static void blk_mq_realloc_hw_ctxs(struct blk_mq_tag_set *set, mutex_unlock(&q->sysfs_lock); } -static void blk_mq_update_poll_flag(struct request_queue *q) -{ - struct blk_mq_tag_set *set = q->tag_set; - - if (set->nr_maps > HCTX_TYPE_POLL && - set->map[HCTX_TYPE_POLL].nr_queues) - blk_queue_flag_set(QUEUE_FLAG_POLL, q); - else - blk_queue_flag_clear(QUEUE_FLAG_POLL, q); -} - int blk_mq_init_allocated_queue(struct blk_mq_tag_set *set, struct request_queue *q) { @@ -4311,7 +4308,6 @@ int blk_mq_init_allocated_queue(struct blk_mq_tag_set *set, q->tag_set = set; q->queue_flags |= QUEUE_FLAG_MQ_DEFAULT; - blk_mq_update_poll_flag(q); INIT_DELAYED_WORK(&q->requeue_work, blk_mq_requeue_work); INIT_LIST_HEAD(&q->flush_list); @@ -4798,8 +4794,10 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, fallback: blk_mq_update_queue_map(set); list_for_each_entry(q, &set->tag_list, tag_set_list) { + struct queue_limits lim; + blk_mq_realloc_hw_ctxs(set, q); - blk_mq_update_poll_flag(q); + if (q->nr_hw_queues != set->nr_hw_queues) { int i = prev_nr_hw_queues; @@ -4811,6 +4809,13 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set, set->nr_hw_queues = prev_nr_hw_queues; goto fallback; } + lim = queue_limits_start_update(q); + if (blk_mq_can_poll(set)) + lim.features |= BLK_FEAT_POLL; + else + lim.features &= ~BLK_FEAT_POLL; + if (queue_limits_commit_update(q, &lim) < 0) + pr_warn("updating the poll flag failed\n"); blk_mq_map_swqueue(q); } diff --git a/block/blk-settings.c b/block/blk-settings.c index bf4622c19b5c09..026ba68d829856 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -460,13 +460,15 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, t->features |= (b->features & BLK_FEAT_INHERIT_MASK); /* - * BLK_FEAT_NOWAIT needs to be supported both by the stacking driver - * and all underlying devices. The stacking driver sets the flag - * before stacking the limits, and this will clear the flag if any - * of the underlying devices does not support it. + * BLK_FEAT_NOWAIT and BLK_FEAT_POLL need to be supported both by the + * stacking driver and all underlying devices. The stacking driver sets + * the flags before stacking the limits, and this will clear the flags + * if any of the underlying devices does not support it. */ if (!(b->features & BLK_FEAT_NOWAIT)) t->features &= ~BLK_FEAT_NOWAIT; + if (!(b->features & BLK_FEAT_POLL)) + t->features &= ~BLK_FEAT_POLL; t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); t->max_user_sectors = min_not_zero(t->max_user_sectors, diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index cde525724831ef..da4e96d686f91e 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -394,13 +394,13 @@ static ssize_t queue_poll_delay_store(struct request_queue *q, const char *page, static ssize_t queue_poll_show(struct request_queue *q, char *page) { - return queue_var_show(test_bit(QUEUE_FLAG_POLL, &q->queue_flags), page); + return queue_var_show(q->limits.features & BLK_FEAT_POLL, page); } static ssize_t queue_poll_store(struct request_queue *q, const char *page, size_t count) { - if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags)) + if (!(q->limits.features & BLK_FEAT_POLL)) return -EINVAL; pr_info_ratelimited("writes to the poll attribute are ignored.\n"); pr_info_ratelimited("please use driver specific parameters instead.\n"); diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index e44697037e86f4..ca1f136575cff4 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -582,7 +582,7 @@ int dm_split_args(int *argc, char ***argvp, char *input) static void dm_set_stacking_limits(struct queue_limits *limits) { blk_set_stacking_limits(limits); - limits->features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT; + limits->features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT | BLK_FEAT_POLL; } /* @@ -1024,14 +1024,13 @@ bool dm_table_request_based(struct dm_table *t) return __table_type_request_based(dm_table_get_type(t)); } -static bool dm_table_supports_poll(struct dm_table *t); - static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md) { enum dm_queue_mode type = dm_table_get_type(t); unsigned int per_io_data_size = 0, front_pad, io_front_pad; unsigned int min_pool_size = 0, pool_size; struct dm_md_mempools *pools; + unsigned int bioset_flags = 0; if (unlikely(type == DM_TYPE_NONE)) { DMERR("no table type is set, can't allocate mempools"); @@ -1048,6 +1047,9 @@ static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device * goto init_bs; } + if (md->queue->limits.features & BLK_FEAT_POLL) + bioset_flags |= BIOSET_PERCPU_CACHE; + for (unsigned int i = 0; i < t->num_targets; i++) { struct dm_target *ti = dm_table_get_target(t, i); @@ -1060,8 +1062,7 @@ static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device * io_front_pad = roundup(per_io_data_size, __alignof__(struct dm_io)) + DM_IO_BIO_OFFSET; - if (bioset_init(&pools->io_bs, pool_size, io_front_pad, - dm_table_supports_poll(t) ? BIOSET_PERCPU_CACHE : 0)) + if (bioset_init(&pools->io_bs, pool_size, io_front_pad, bioset_flags)) goto out_free_pools; if (t->integrity_supported && bioset_integrity_create(&pools->io_bs, pool_size)) @@ -1404,14 +1405,6 @@ struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector) return &t->targets[(KEYS_PER_NODE * n) + k]; } -static int device_not_poll_capable(struct dm_target *ti, struct dm_dev *dev, - sector_t start, sector_t len, void *data) -{ - struct request_queue *q = bdev_get_queue(dev->bdev); - - return !test_bit(QUEUE_FLAG_POLL, &q->queue_flags); -} - /* * type->iterate_devices() should be called when the sanity check needs to * iterate and check all underlying data devices. iterate_devices() will @@ -1459,19 +1452,6 @@ static int count_device(struct dm_target *ti, struct dm_dev *dev, return 0; } -static bool dm_table_supports_poll(struct dm_table *t) -{ - for (unsigned int i = 0; i < t->num_targets; i++) { - struct dm_target *ti = dm_table_get_target(t, i); - - if (!ti->type->iterate_devices || - ti->type->iterate_devices(ti, device_not_poll_capable, NULL)) - return false; - } - - return true; -} - /* * Check whether a table has no data devices attached using each * target's iterate_devices method. @@ -1817,6 +1797,13 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, if (!dm_table_supports_nowait(t)) limits->features &= ~BLK_FEAT_NOWAIT; + /* + * The current polling impementation does not support request based + * stacking. + */ + if (!__table_type_bio_based(t->type)) + limits->features &= ~BLK_FEAT_POLL; + if (!dm_table_supports_discards(t)) { limits->max_hw_discard_sectors = 0; limits->discard_granularity = 0; @@ -1858,21 +1845,6 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, return r; dm_update_crypto_profile(q, t); - - /* - * Check for request-based device is left to - * dm_mq_init_request_queue()->blk_mq_init_allocated_queue(). - * - * For bio-based device, only set QUEUE_FLAG_POLL when all - * underlying devices supporting polling. - */ - if (__table_type_bio_based(t->type)) { - if (dm_table_supports_poll(t)) - blk_queue_flag_set(QUEUE_FLAG_POLL, q); - else - blk_queue_flag_clear(QUEUE_FLAG_POLL, q); - } - return 0; } diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c index 61a162c9cf4e6c..4933194d00e592 100644 --- a/drivers/nvme/host/multipath.c +++ b/drivers/nvme/host/multipath.c @@ -538,7 +538,7 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head) blk_set_stacking_limits(&lim); lim.dma_alignment = 3; - lim.features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT; + lim.features |= BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT | BLK_FEAT_POLL; if (head->ids.csi != NVME_CSI_ZNS) lim.max_zone_append_sectors = 0; @@ -549,16 +549,6 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head) head->disk->private_data = head; sprintf(head->disk->disk_name, "nvme%dn%d", ctrl->subsys->instance, head->instance); - - /* - * This assumes all controllers that refer to a namespace either - * support poll queues or not. That is not a strict guarantee, - * but if the assumption is wrong the effect is only suboptimal - * performance but not correctness problem. - */ - if (ctrl->tagset->nr_maps > HCTX_TYPE_POLL && - ctrl->tagset->map[HCTX_TYPE_POLL].nr_queues) - blk_queue_flag_set(QUEUE_FLAG_POLL, head->disk->queue); return 0; } diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 7022e06a3dd9a3..cd27b66cbacc00 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -310,6 +310,9 @@ enum { /* supports DAX */ BLK_FEAT_DAX = (1u << 8), + + /* supports I/O polling */ + BLK_FEAT_POLL = (1u << 9), }; /* @@ -577,7 +580,6 @@ struct request_queue { #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */ #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ -#define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */ #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */ #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */ #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */ From b1fc937a55f5735b98d9dceae5bb6ba262501f56 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:49 +0200 Subject: [PATCH 101/209] block: move the zoned flag into the features field Move the zoned flags into the features field to reclaim a little bit of space. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-23-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 5 ++--- drivers/block/null_blk/zoned.c | 2 +- drivers/block/ublk_drv.c | 2 +- drivers/block/virtio_blk.c | 5 +++-- drivers/md/dm-table.c | 11 ++++++----- drivers/md/dm-zone.c | 2 +- drivers/md/dm-zoned-target.c | 2 +- drivers/nvme/host/zns.c | 2 +- drivers/scsi/sd_zbc.c | 2 +- include/linux/blkdev.h | 9 ++++++--- 10 files changed, 23 insertions(+), 19 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index 026ba68d829856..96e07f24bd9aa1 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -68,7 +68,7 @@ static void blk_apply_bdi_limits(struct backing_dev_info *bdi, static int blk_validate_zoned_limits(struct queue_limits *lim) { - if (!lim->zoned) { + if (!(lim->features & BLK_FEAT_ZONED)) { if (WARN_ON_ONCE(lim->max_open_zones) || WARN_ON_ONCE(lim->max_active_zones) || WARN_ON_ONCE(lim->zone_write_granularity) || @@ -602,8 +602,7 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, b->max_secure_erase_sectors); t->zone_write_granularity = max(t->zone_write_granularity, b->zone_write_granularity); - t->zoned = max(t->zoned, b->zoned); - if (!t->zoned) { + if (!(t->features & BLK_FEAT_ZONED)) { t->zone_write_granularity = 0; t->max_zone_append_sectors = 0; } diff --git a/drivers/block/null_blk/zoned.c b/drivers/block/null_blk/zoned.c index f118d304f31080..ca8e739e76b981 100644 --- a/drivers/block/null_blk/zoned.c +++ b/drivers/block/null_blk/zoned.c @@ -158,7 +158,7 @@ int null_init_zoned_dev(struct nullb_device *dev, sector += dev->zone_size_sects; } - lim->zoned = true; + lim->features |= BLK_FEAT_ZONED; lim->chunk_sectors = dev->zone_size_sects; lim->max_zone_append_sectors = dev->zone_append_max_sectors; lim->max_open_zones = dev->zone_max_open; diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index 4fcde099935868..69c16018cbb19a 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -2196,7 +2196,7 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub, struct io_uring_cmd *cmd) if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED)) return -EOPNOTSUPP; - lim.zoned = true; + lim.features |= BLK_FEAT_ZONED; lim.max_active_zones = p->max_active_zones; lim.max_open_zones = p->max_open_zones; lim.max_zone_append_sectors = p->max_zone_append_sectors; diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index 13a2f24f176628..cea45b296f8bec 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -728,7 +728,7 @@ static int virtblk_read_zoned_limits(struct virtio_blk *vblk, dev_dbg(&vdev->dev, "probing host-managed zoned device\n"); - lim->zoned = true; + lim->features |= BLK_FEAT_ZONED; virtio_cread(vdev, struct virtio_blk_config, zoned.max_open_zones, &v); @@ -1546,7 +1546,8 @@ static int virtblk_probe(struct virtio_device *vdev) * All steps that follow use the VQs therefore they need to be * placed after the virtio_device_ready() call above. */ - if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) && lim.zoned) { + if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) && + (lim.features & BLK_FEAT_ZONED)) { blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, vblk->disk->queue); err = blk_revalidate_disk_zones(vblk->disk); if (err) diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index ca1f136575cff4..df6313c3fe6ba4 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -1605,12 +1605,12 @@ int dm_calculate_queue_limits(struct dm_table *t, ti->type->iterate_devices(ti, dm_set_device_limits, &ti_limits); - if (!zoned && ti_limits.zoned) { + if (!zoned && (ti_limits.features & BLK_FEAT_ZONED)) { /* * After stacking all limits, validate all devices * in table support this zoned model and zone sectors. */ - zoned = ti_limits.zoned; + zoned = (ti_limits.features & BLK_FEAT_ZONED); zone_sectors = ti_limits.chunk_sectors; } @@ -1658,12 +1658,12 @@ int dm_calculate_queue_limits(struct dm_table *t, * zoned model on host-managed zoned block devices. * BUT... */ - if (limits->zoned) { + if (limits->features & BLK_FEAT_ZONED) { /* * ...IF the above limits stacking determined a zoned model * validate that all of the table's devices conform to it. */ - zoned = limits->zoned; + zoned = limits->features & BLK_FEAT_ZONED; zone_sectors = limits->chunk_sectors; } if (validate_hardware_zoned(t, zoned, zone_sectors)) @@ -1834,7 +1834,8 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, * For a zoned target, setup the zones related queue attributes * and resources necessary for zone append emulation if necessary. */ - if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) && limits->zoned) { + if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) && + (limits->features & limits->features & BLK_FEAT_ZONED)) { r = dm_set_zones_restrictions(t, q, limits); if (r) return r; diff --git a/drivers/md/dm-zone.c b/drivers/md/dm-zone.c index 5d66d916730efa..88d313229b43ff 100644 --- a/drivers/md/dm-zone.c +++ b/drivers/md/dm-zone.c @@ -263,7 +263,7 @@ int dm_set_zones_restrictions(struct dm_table *t, struct request_queue *q, if (nr_conv_zones >= ret) { lim->max_open_zones = 0; lim->max_active_zones = 0; - lim->zoned = false; + lim->features &= ~BLK_FEAT_ZONED; clear_bit(DMF_EMULATE_ZONE_APPEND, &md->flags); disk->nr_zones = 0; return 0; diff --git a/drivers/md/dm-zoned-target.c b/drivers/md/dm-zoned-target.c index 12236e6f46f39c..cd0ee144973f9f 100644 --- a/drivers/md/dm-zoned-target.c +++ b/drivers/md/dm-zoned-target.c @@ -1009,7 +1009,7 @@ static void dmz_io_hints(struct dm_target *ti, struct queue_limits *limits) limits->max_sectors = chunk_sectors; /* We are exposing a drive-managed zoned block device */ - limits->zoned = false; + limits->features &= ~BLK_FEAT_ZONED; } /* diff --git a/drivers/nvme/host/zns.c b/drivers/nvme/host/zns.c index 77aa0f440a6d2a..06f2417aa50de7 100644 --- a/drivers/nvme/host/zns.c +++ b/drivers/nvme/host/zns.c @@ -108,7 +108,7 @@ int nvme_query_zone_info(struct nvme_ns *ns, unsigned lbaf, void nvme_update_zone_info(struct nvme_ns *ns, struct queue_limits *lim, struct nvme_zone_info *zi) { - lim->zoned = 1; + lim->features |= BLK_FEAT_ZONED; lim->max_open_zones = zi->max_open_zones; lim->max_active_zones = zi->max_active_zones; lim->max_zone_append_sectors = ns->ctrl->max_zone_append; diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c index 360ec980499529..d3f84665946ec4 100644 --- a/drivers/scsi/sd_zbc.c +++ b/drivers/scsi/sd_zbc.c @@ -601,7 +601,7 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, struct queue_limits *lim, if (sdkp->device->type != TYPE_ZBC) return 0; - lim->zoned = true; + lim->features |= BLK_FEAT_ZONED; /* * Per ZBC and ZAC specifications, writes in sequential write required diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index cd27b66cbacc00..bdc30c1fb1b57b 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -313,6 +313,9 @@ enum { /* supports I/O polling */ BLK_FEAT_POLL = (1u << 9), + + /* is a zoned device */ + BLK_FEAT_ZONED = (1u << 10), }; /* @@ -320,7 +323,7 @@ enum { */ #define BLK_FEAT_INHERIT_MASK \ (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \ - BLK_FEAT_STABLE_WRITES) + BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED) /* internal flags in queue_limits.flags */ enum { @@ -372,7 +375,6 @@ struct queue_limits { unsigned char misaligned; unsigned char discard_misaligned; unsigned char raid_partial_stripes_expensive; - bool zoned; unsigned int max_open_zones; unsigned int max_active_zones; @@ -654,7 +656,8 @@ static inline enum rpm_status queue_rpm_status(struct request_queue *q) static inline bool blk_queue_is_zoned(struct request_queue *q) { - return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && q->limits.zoned; + return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && + (q->limits.features & BLK_FEAT_ZONED); } #ifdef CONFIG_BLK_DEV_ZONED From a52758a39768f441e468a41da6c15a59d6d6011a Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:50 +0200 Subject: [PATCH 102/209] block: move the zone_resetall flag to queue_limits Move the zone_resetall flag into the queue_limits feature field so that it can be set atomically with the queue frozen. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-24-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-mq-debugfs.c | 1 - drivers/block/null_blk/zoned.c | 3 +-- drivers/block/ublk_drv.c | 4 +--- drivers/block/virtio_blk.c | 3 +-- drivers/nvme/host/zns.c | 3 +-- drivers/scsi/sd_zbc.c | 5 +---- include/linux/blkdev.h | 6 ++++-- 7 files changed, 9 insertions(+), 16 deletions(-) diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index 3a21527913840d..f2fd72f4414ae8 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -91,7 +91,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(REGISTERED), QUEUE_FLAG_NAME(QUIESCED), QUEUE_FLAG_NAME(PCI_P2PDMA), - QUEUE_FLAG_NAME(ZONE_RESETALL), QUEUE_FLAG_NAME(RQ_ALLOC_TIME), QUEUE_FLAG_NAME(HCTX_ACTIVE), QUEUE_FLAG_NAME(SQ_SCHED), diff --git a/drivers/block/null_blk/zoned.c b/drivers/block/null_blk/zoned.c index ca8e739e76b981..b42c00f1313254 100644 --- a/drivers/block/null_blk/zoned.c +++ b/drivers/block/null_blk/zoned.c @@ -158,7 +158,7 @@ int null_init_zoned_dev(struct nullb_device *dev, sector += dev->zone_size_sects; } - lim->features |= BLK_FEAT_ZONED; + lim->features |= BLK_FEAT_ZONED | BLK_FEAT_ZONE_RESETALL; lim->chunk_sectors = dev->zone_size_sects; lim->max_zone_append_sectors = dev->zone_append_max_sectors; lim->max_open_zones = dev->zone_max_open; @@ -171,7 +171,6 @@ int null_register_zoned_dev(struct nullb *nullb) struct request_queue *q = nullb->q; struct gendisk *disk = nullb->disk; - blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q); disk->nr_zones = bdev_nr_zones(disk->part0); pr_info("%s: using %s zone append\n", diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index 69c16018cbb19a..4fdff13fc23b8a 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -248,8 +248,6 @@ static int ublk_dev_param_zoned_validate(const struct ublk_device *ub) static void ublk_dev_param_zoned_apply(struct ublk_device *ub) { - blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, ub->ub_disk->queue); - ub->ub_disk->nr_zones = ublk_get_nr_zones(ub); } @@ -2196,7 +2194,7 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub, struct io_uring_cmd *cmd) if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED)) return -EOPNOTSUPP; - lim.features |= BLK_FEAT_ZONED; + lim.features |= BLK_FEAT_ZONED | BLK_FEAT_ZONE_RESETALL; lim.max_active_zones = p->max_active_zones; lim.max_open_zones = p->max_open_zones; lim.max_zone_append_sectors = p->max_zone_append_sectors; diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index cea45b296f8bec..6c64a67ab9c901 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -728,7 +728,7 @@ static int virtblk_read_zoned_limits(struct virtio_blk *vblk, dev_dbg(&vdev->dev, "probing host-managed zoned device\n"); - lim->features |= BLK_FEAT_ZONED; + lim->features |= BLK_FEAT_ZONED | BLK_FEAT_ZONE_RESETALL; virtio_cread(vdev, struct virtio_blk_config, zoned.max_open_zones, &v); @@ -1548,7 +1548,6 @@ static int virtblk_probe(struct virtio_device *vdev) */ if (IS_ENABLED(CONFIG_BLK_DEV_ZONED) && (lim.features & BLK_FEAT_ZONED)) { - blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, vblk->disk->queue); err = blk_revalidate_disk_zones(vblk->disk); if (err) goto out_cleanup_disk; diff --git a/drivers/nvme/host/zns.c b/drivers/nvme/host/zns.c index 06f2417aa50de7..99bb89c2495ae3 100644 --- a/drivers/nvme/host/zns.c +++ b/drivers/nvme/host/zns.c @@ -108,13 +108,12 @@ int nvme_query_zone_info(struct nvme_ns *ns, unsigned lbaf, void nvme_update_zone_info(struct nvme_ns *ns, struct queue_limits *lim, struct nvme_zone_info *zi) { - lim->features |= BLK_FEAT_ZONED; + lim->features |= BLK_FEAT_ZONED | BLK_FEAT_ZONE_RESETALL; lim->max_open_zones = zi->max_open_zones; lim->max_active_zones = zi->max_active_zones; lim->max_zone_append_sectors = ns->ctrl->max_zone_append; lim->chunk_sectors = ns->head->zsze = nvme_lba_to_sect(ns->head, zi->zone_size); - blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, ns->queue); } static void *nvme_zns_alloc_report_buffer(struct nvme_ns *ns, diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c index d3f84665946ec4..f7067afac79c14 100644 --- a/drivers/scsi/sd_zbc.c +++ b/drivers/scsi/sd_zbc.c @@ -592,8 +592,6 @@ int sd_zbc_revalidate_zones(struct scsi_disk *sdkp) int sd_zbc_read_zones(struct scsi_disk *sdkp, struct queue_limits *lim, u8 buf[SD_BUF_SIZE]) { - struct gendisk *disk = sdkp->disk; - struct request_queue *q = disk->queue; unsigned int nr_zones; u32 zone_blocks = 0; int ret; @@ -601,7 +599,7 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, struct queue_limits *lim, if (sdkp->device->type != TYPE_ZBC) return 0; - lim->features |= BLK_FEAT_ZONED; + lim->features |= BLK_FEAT_ZONED | BLK_FEAT_ZONE_RESETALL; /* * Per ZBC and ZAC specifications, writes in sequential write required @@ -630,7 +628,6 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, struct queue_limits *lim, sdkp->early_zone_info.zone_blocks = zone_blocks; /* The drive satisfies the kernel restrictions: set it up */ - blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q); if (sdkp->zones_max_open == U32_MAX) lim->max_open_zones = 0; else diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index bdc30c1fb1b57b..1077cb8d8fd808 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -316,6 +316,9 @@ enum { /* is a zoned device */ BLK_FEAT_ZONED = (1u << 10), + + /* supports Zone Reset All */ + BLK_FEAT_ZONE_RESETALL = (1u << 11), }; /* @@ -586,7 +589,6 @@ struct request_queue { #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */ #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */ #define QUEUE_FLAG_PCI_P2PDMA 25 /* device supports PCI p2p requests */ -#define QUEUE_FLAG_ZONE_RESETALL 26 /* supports Zone Reset All */ #define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */ #define QUEUE_FLAG_HCTX_ACTIVE 28 /* at least one blk-mq hctx is active */ #define QUEUE_FLAG_SQ_SCHED 30 /* single queue style io dispatch */ @@ -607,7 +609,7 @@ bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); #define blk_queue_nonrot(q) ((q)->limits.features & BLK_FEAT_ROTATIONAL) #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) #define blk_queue_zone_resetall(q) \ - test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags) + ((q)->limits.features & BLK_FEAT_ZONE_RESETALL) #define blk_queue_dax(q) ((q)->limits.features & BLK_FEAT_DAX) #define blk_queue_pci_p2pdma(q) \ test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags) From 9c1e42e3c876c66796eda23e79836a4d92613a61 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:51 +0200 Subject: [PATCH 103/209] block: move the pci_p2pdma flag to queue_limits Move the pci_p2pdma flag into the queue_limits feature field so that it can be set atomically with the queue frozen. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-25-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-mq-debugfs.c | 1 - drivers/nvme/host/core.c | 8 +++----- include/linux/blkdev.h | 7 ++++--- 3 files changed, 7 insertions(+), 9 deletions(-) diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index f2fd72f4414ae8..8b5a68861c119b 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -90,7 +90,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(STATS), QUEUE_FLAG_NAME(REGISTERED), QUEUE_FLAG_NAME(QUIESCED), - QUEUE_FLAG_NAME(PCI_P2PDMA), QUEUE_FLAG_NAME(RQ_ALLOC_TIME), QUEUE_FLAG_NAME(HCTX_ACTIVE), QUEUE_FLAG_NAME(SQ_SCHED), diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 5ecf762d7c8837..31e752e8d632cd 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -3735,6 +3735,9 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, struct nvme_ns_info *info) if (ctrl->opts && ctrl->opts->data_digest) lim.features |= BLK_FEAT_STABLE_WRITES; + if (ctrl->ops->supports_pci_p2pdma && + ctrl->ops->supports_pci_p2pdma(ctrl)) + lim.features |= BLK_FEAT_PCI_P2PDMA; disk = blk_mq_alloc_disk(ctrl->tagset, &lim, ns); if (IS_ERR(disk)) @@ -3744,11 +3747,6 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, struct nvme_ns_info *info) ns->disk = disk; ns->queue = disk->queue; - - if (ctrl->ops->supports_pci_p2pdma && - ctrl->ops->supports_pci_p2pdma(ctrl)) - blk_queue_flag_set(QUEUE_FLAG_PCI_P2PDMA, ns->queue); - ns->ctrl = ctrl; kref_init(&ns->kref); diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 1077cb8d8fd808..ab0f7dfba556eb 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -319,6 +319,9 @@ enum { /* supports Zone Reset All */ BLK_FEAT_ZONE_RESETALL = (1u << 11), + + /* supports PCI(e) p2p requests */ + BLK_FEAT_PCI_P2PDMA = (1u << 12), }; /* @@ -588,7 +591,6 @@ struct request_queue { #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */ #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */ #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */ -#define QUEUE_FLAG_PCI_P2PDMA 25 /* device supports PCI p2p requests */ #define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */ #define QUEUE_FLAG_HCTX_ACTIVE 28 /* at least one blk-mq hctx is active */ #define QUEUE_FLAG_SQ_SCHED 30 /* single queue style io dispatch */ @@ -611,8 +613,7 @@ bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); #define blk_queue_zone_resetall(q) \ ((q)->limits.features & BLK_FEAT_ZONE_RESETALL) #define blk_queue_dax(q) ((q)->limits.features & BLK_FEAT_DAX) -#define blk_queue_pci_p2pdma(q) \ - test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags) +#define blk_queue_pci_p2pdma(q) ((q)->limits.features & BLK_FEAT_PCI_P2PDMA) #ifdef CONFIG_BLK_RQ_ALLOC_TIME #define blk_queue_rq_alloc_time(q) \ test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags) From 8c8f5c85b20d0a7dc0ab9b2a17318130d69ceb5a Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:52 +0200 Subject: [PATCH 104/209] block: move the skip_tagset_quiesce flag to queue_limits Move the skip_tagset_quiesce flag into the queue_limits feature field so that it can be set atomically with the queue frozen. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-26-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-mq-debugfs.c | 1 - drivers/nvme/host/core.c | 8 +++++--- include/linux/blkdev.h | 6 ++++-- 3 files changed, 9 insertions(+), 6 deletions(-) diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index 8b5a68861c119b..344f9e503bdb32 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -93,7 +93,6 @@ static const char *const blk_queue_flag_name[] = { QUEUE_FLAG_NAME(RQ_ALLOC_TIME), QUEUE_FLAG_NAME(HCTX_ACTIVE), QUEUE_FLAG_NAME(SQ_SCHED), - QUEUE_FLAG_NAME(SKIP_TAGSET_QUIESCE), }; #undef QUEUE_FLAG_NAME diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 31e752e8d632cd..bf410d10b12006 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -4489,13 +4489,15 @@ int nvme_alloc_io_tag_set(struct nvme_ctrl *ctrl, struct blk_mq_tag_set *set, return ret; if (ctrl->ops->flags & NVME_F_FABRICS) { - ctrl->connect_q = blk_mq_alloc_queue(set, NULL, NULL); + struct queue_limits lim = { + .features = BLK_FEAT_SKIP_TAGSET_QUIESCE, + }; + + ctrl->connect_q = blk_mq_alloc_queue(set, &lim, NULL); if (IS_ERR(ctrl->connect_q)) { ret = PTR_ERR(ctrl->connect_q); goto out_free_tag_set; } - blk_queue_flag_set(QUEUE_FLAG_SKIP_TAGSET_QUIESCE, - ctrl->connect_q); } ctrl->tagset = set; diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index ab0f7dfba556eb..2c433ebf6f2030 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -322,6 +322,9 @@ enum { /* supports PCI(e) p2p requests */ BLK_FEAT_PCI_P2PDMA = (1u << 12), + + /* skip this queue in blk_mq_(un)quiesce_tagset */ + BLK_FEAT_SKIP_TAGSET_QUIESCE = (1u << 13), }; /* @@ -594,7 +597,6 @@ struct request_queue { #define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */ #define QUEUE_FLAG_HCTX_ACTIVE 28 /* at least one blk-mq hctx is active */ #define QUEUE_FLAG_SQ_SCHED 30 /* single queue style io dispatch */ -#define QUEUE_FLAG_SKIP_TAGSET_QUIESCE 31 /* quiesce_tagset skip the queue*/ #define QUEUE_FLAG_MQ_DEFAULT (1UL << QUEUE_FLAG_SAME_COMP) @@ -629,7 +631,7 @@ bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags) #define blk_queue_sq_sched(q) test_bit(QUEUE_FLAG_SQ_SCHED, &(q)->queue_flags) #define blk_queue_skip_tagset_quiesce(q) \ - test_bit(QUEUE_FLAG_SKIP_TAGSET_QUIESCE, &(q)->queue_flags) + ((q)->limits.features & BLK_FEAT_SKIP_TAGSET_QUIESCE) extern void blk_set_pm_only(struct request_queue *q); extern void blk_clear_pm_only(struct request_queue *q); From 339d3948c07b4aa2940aeb874294a7d6782cec16 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 17 Jun 2024 08:04:53 +0200 Subject: [PATCH 105/209] block: move the bounce flag into the features field Move the bounce flag into the features field to reclaim a little bit of space. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Hannes Reinecke Link: https://lore.kernel.org/r/20240617060532.127975-27-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 1 - block/blk.h | 2 +- drivers/scsi/scsi_lib.c | 2 +- include/linux/blkdev.h | 6 ++++-- 4 files changed, 6 insertions(+), 5 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index 96e07f24bd9aa1..d0e9096f93ca8a 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -479,7 +479,6 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, b->max_write_zeroes_sectors); t->max_zone_append_sectors = min(queue_limits_max_zone_append_sectors(t), queue_limits_max_zone_append_sectors(b)); - t->bounce = max(t->bounce, b->bounce); t->seg_boundary_mask = min_not_zero(t->seg_boundary_mask, b->seg_boundary_mask); diff --git a/block/blk.h b/block/blk.h index 79e8d5d4fe0caf..fa32f7fad5d7e6 100644 --- a/block/blk.h +++ b/block/blk.h @@ -394,7 +394,7 @@ struct bio *__blk_queue_bounce(struct bio *bio, struct request_queue *q); static inline bool blk_queue_may_bounce(struct request_queue *q) { return IS_ENABLED(CONFIG_BOUNCE) && - q->limits.bounce == BLK_BOUNCE_HIGH && + (q->limits.features & BLK_FEAT_BOUNCE_HIGH) && max_low_pfn >= max_pfn; } diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c index 54f771ec8cfb5e..e2f7bfb2b9e450 100644 --- a/drivers/scsi/scsi_lib.c +++ b/drivers/scsi/scsi_lib.c @@ -1986,7 +1986,7 @@ void scsi_init_limits(struct Scsi_Host *shost, struct queue_limits *lim) shost->dma_alignment, dma_get_cache_alignment() - 1); if (shost->no_highmem) - lim->bounce = BLK_BOUNCE_HIGH; + lim->features |= BLK_FEAT_BOUNCE_HIGH; dma_set_seg_boundary(dev, shost->dma_boundary); dma_set_max_seg_size(dev, shost->max_segment_size); diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 2c433ebf6f2030..e96ba7b97288d2 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -325,6 +325,9 @@ enum { /* skip this queue in blk_mq_(un)quiesce_tagset */ BLK_FEAT_SKIP_TAGSET_QUIESCE = (1u << 13), + + /* bounce all highmem pages */ + BLK_FEAT_BOUNCE_HIGH = (1u << 14), }; /* @@ -332,7 +335,7 @@ enum { */ #define BLK_FEAT_INHERIT_MASK \ (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \ - BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED) + BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | BLK_FEAT_BOUNCE_HIGH) /* internal flags in queue_limits.flags */ enum { @@ -352,7 +355,6 @@ enum blk_bounce { struct queue_limits { unsigned int features; unsigned int flags; - enum blk_bounce bounce; unsigned long seg_boundary_mask; unsigned long virt_boundary_mask; From 3e05b222382ec67dce7358d50b6006e91d028d8b Mon Sep 17 00:00:00 2001 From: Gabriel Krisman Bertazi Date: Tue, 18 Jun 2024 22:06:18 -0400 Subject: [PATCH 106/209] io_uring: Fix probe of disabled operations io_probe checks io_issue_def->not_supported, but we never really set that field, as we mark non-supported functions through a specific ->prep handler. This means we end up returning IO_URING_OP_SUPPORTED, even for disabled operations. Fix it by just checking the prep handler itself. Fixes: 66f4af93da57 ("io_uring: add support for probing opcodes") Signed-off-by: Gabriel Krisman Bertazi Link: https://lore.kernel.org/r/20240619020620.5301-2-krisman@suse.de Signed-off-by: Jens Axboe --- io_uring/opdef.c | 8 ++++++++ io_uring/opdef.h | 4 ++-- io_uring/register.c | 2 +- 3 files changed, 11 insertions(+), 3 deletions(-) diff --git a/io_uring/opdef.c b/io_uring/opdef.c index 2dd49cf22f642d..a2be3bbca5ffa1 100644 --- a/io_uring/opdef.c +++ b/io_uring/opdef.c @@ -751,6 +751,14 @@ const char *io_uring_get_opcode(u8 opcode) return "INVALID"; } +bool io_uring_op_supported(u8 opcode) +{ + if (opcode < IORING_OP_LAST && + io_issue_defs[opcode].prep != io_eopnotsupp_prep) + return true; + return false; +} + void __init io_uring_optable_init(void) { int i; diff --git a/io_uring/opdef.h b/io_uring/opdef.h index 7ee6f5aa90aa3c..14456436ff74ac 100644 --- a/io_uring/opdef.h +++ b/io_uring/opdef.h @@ -17,8 +17,6 @@ struct io_issue_def { unsigned poll_exclusive : 1; /* op supports buffer selection */ unsigned buffer_select : 1; - /* opcode is not supported by this kernel */ - unsigned not_supported : 1; /* skip auditing */ unsigned audit_skip : 1; /* supports ioprio */ @@ -47,5 +45,7 @@ struct io_cold_def { extern const struct io_issue_def io_issue_defs[]; extern const struct io_cold_def io_cold_defs[]; +bool io_uring_op_supported(u8 opcode); + void io_uring_optable_init(void); #endif diff --git a/io_uring/register.c b/io_uring/register.c index f121e02f5e10e6..e28cc226217cdb 100644 --- a/io_uring/register.c +++ b/io_uring/register.c @@ -59,7 +59,7 @@ static __cold int io_probe(struct io_ring_ctx *ctx, void __user *arg, for (i = 0; i < nr_args; i++) { p->ops[i].op = i; - if (!io_issue_defs[i].not_supported) + if (io_uring_op_supported(i)) p->ops[i].flags = IO_URING_OP_SUPPORTED; } p->ops_len = i; From 6bc9199d0c84f5cd72922223231c7708698059a2 Mon Sep 17 00:00:00 2001 From: Gabriel Krisman Bertazi Date: Tue, 18 Jun 2024 22:06:19 -0400 Subject: [PATCH 107/209] io_uring: Allocate only necessary memory in io_probe We write at most IORING_OP_LAST entries in the probe buffer, so we don't need to allocate temporary space for more than that. As a side effect, we no longer can overflow "size". Signed-off-by: Gabriel Krisman Bertazi Link: https://lore.kernel.org/r/20240619020620.5301-3-krisman@suse.de Signed-off-by: Jens Axboe --- io_uring/register.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) diff --git a/io_uring/register.c b/io_uring/register.c index e28cc226217cdb..e3c20be5a19803 100644 --- a/io_uring/register.c +++ b/io_uring/register.c @@ -39,9 +39,10 @@ static __cold int io_probe(struct io_ring_ctx *ctx, void __user *arg, size_t size; int i, ret; + if (nr_args > IORING_OP_LAST) + nr_args = IORING_OP_LAST; + size = struct_size(p, ops, nr_args); - if (size == SIZE_MAX) - return -EOVERFLOW; p = kzalloc(size, GFP_KERNEL); if (!p) return -ENOMEM; @@ -54,8 +55,6 @@ static __cold int io_probe(struct io_ring_ctx *ctx, void __user *arg, goto out; p->last_op = IORING_OP_LAST - 1; - if (nr_args > IORING_OP_LAST) - nr_args = IORING_OP_LAST; for (i = 0; i < nr_args; i++) { p->ops[i].op = i; From f6860b6069b92559f5cdb65f48e2d82051eaebca Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 19 Jun 2024 17:45:33 +0200 Subject: [PATCH 108/209] block: remove the unused blk_bounce enum The enum has been replaced with the BLK_FEAT_BOUNCE_HIGH flag. Reported-by: Keith Busch Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240619154623.450048-2-hch@lst.de Signed-off-by: Jens Axboe --- include/linux/blkdev.h | 9 --------- 1 file changed, 9 deletions(-) diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index e96ba7b97288d2..f7d275e3fb2c1e 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -343,15 +343,6 @@ enum { BLK_FLAGS_WRITE_CACHE_DISABLED = (1u << 31), }; -/* - * BLK_BOUNCE_NONE: never bounce (default) - * BLK_BOUNCE_HIGH: bounce all highmem pages - */ -enum blk_bounce { - BLK_BOUNCE_NONE, - BLK_BOUNCE_HIGH, -}; - struct queue_limits { unsigned int features; unsigned int flags; From 4e54ea72edd68d074be2403f3efc67ff0541e298 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 19 Jun 2024 17:45:34 +0200 Subject: [PATCH 109/209] block: fix spelling and grammar for in writeback_cache_control.rst Suggested-by: Damien Le Moal Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: Martin K. Petersen Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240619154623.450048-3-hch@lst.de Signed-off-by: Jens Axboe --- Documentation/block/writeback_cache_control.rst | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/Documentation/block/writeback_cache_control.rst b/Documentation/block/writeback_cache_control.rst index c575e08beda8e3..c3707d07178045 100644 --- a/Documentation/block/writeback_cache_control.rst +++ b/Documentation/block/writeback_cache_control.rst @@ -70,8 +70,8 @@ flag in the features field of the queue_limits structure. Implementation details for bio based block drivers -------------------------------------------------- -For bio based drivers the REQ_PREFLUSH and REQ_FUA bit are simplify passed on -to the driver if the drivers sets the BLK_FEAT_WRITE_CACHE flag and the drivers +For bio based drivers the REQ_PREFLUSH and REQ_FUA bit are simply passed on to +the driver if the driver sets the BLK_FEAT_WRITE_CACHE flag and the driver needs to handle them. *NOTE*: The REQ_FUA bit also gets passed on when the BLK_FEAT_FUA flags is @@ -89,7 +89,7 @@ When the BLK_FEAT_WRITE_CACHE flag is set, REQ_OP_WRITE | REQ_PREFLUSH requests with a payload are automatically turned into a sequence of a REQ_OP_FLUSH request followed by the actual write by the block layer. -When the BLK_FEAT_FUA flags is set, the REQ_FUA bit simplify passed on for the +When the BLK_FEAT_FUA flags is set, the REQ_FUA bit is simply passed on for the REQ_OP_WRITE request, else a REQ_OP_FLUSH request is sent by the block layer after the completion of the write request for bio submissions with the REQ_FUA bit set. From bae1c74316b86c67c95658c3a0cd312cec9aad77 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 19 Jun 2024 17:45:35 +0200 Subject: [PATCH 110/209] block: renumber and rename the cache disabled flag Start with the first bit, and drop the plural-S from the name. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240619154623.450048-4-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-sysfs.c | 6 +++--- include/linux/blkdev.h | 6 +++--- 2 files changed, 6 insertions(+), 6 deletions(-) diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index da4e96d686f91e..59e6d111ed059a 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -429,7 +429,7 @@ static ssize_t queue_io_timeout_store(struct request_queue *q, const char *page, static ssize_t queue_wc_show(struct request_queue *q, char *page) { - if (q->limits.features & BLK_FLAGS_WRITE_CACHE_DISABLED) + if (q->limits.features & BLK_FLAG_WRITE_CACHE_DISABLED) return sprintf(page, "write through\n"); return sprintf(page, "write back\n"); } @@ -452,9 +452,9 @@ static ssize_t queue_wc_store(struct request_queue *q, const char *page, lim = queue_limits_start_update(q); if (disable) - lim.flags |= BLK_FLAGS_WRITE_CACHE_DISABLED; + lim.flags |= BLK_FLAG_WRITE_CACHE_DISABLED; else - lim.flags &= ~BLK_FLAGS_WRITE_CACHE_DISABLED; + lim.flags &= ~BLK_FLAG_WRITE_CACHE_DISABLED; err = queue_limits_commit_update(q, &lim); if (err) return err; diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index f7d275e3fb2c1e..713a98b6dbba08 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -339,8 +339,8 @@ enum { /* internal flags in queue_limits.flags */ enum { - /* do not send FLUSH or FUA command despite advertised write cache */ - BLK_FLAGS_WRITE_CACHE_DISABLED = (1u << 31), + /* do not send FLUSH/FUA commands despite advertising a write cache */ + BLK_FLAG_WRITE_CACHE_DISABLED = (1u << 0), }; struct queue_limits { @@ -1339,7 +1339,7 @@ static inline bool bdev_stable_writes(struct block_device *bdev) static inline bool blk_queue_write_cache(struct request_queue *q) { return (q->limits.features & BLK_FEAT_WRITE_CACHE) && - !(q->limits.flags & BLK_FLAGS_WRITE_CACHE_DISABLED); + !(q->limits.flags & BLK_FLAG_WRITE_CACHE_DISABLED); } static inline bool bdev_write_cache(struct block_device *bdev) From 5543217be468268dfedf504f4969771b9a377353 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 19 Jun 2024 17:45:36 +0200 Subject: [PATCH 111/209] block: move the misaligned flag into the features field Move the misaligned flags into the features field to reclaim a little bit of space. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240619154623.450048-5-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 20 ++++++++++---------- include/linux/blkdev.h | 4 +++- 2 files changed, 13 insertions(+), 11 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index d0e9096f93ca8a..a1b10404e500bc 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -258,7 +258,7 @@ static int blk_validate_limits(struct queue_limits *lim) if (lim->alignment_offset) { lim->alignment_offset &= (lim->physical_block_size - 1); - lim->misaligned = 0; + lim->features &= ~BLK_FEAT_MISALIGNED; } if (!(lim->features & BLK_FEAT_WRITE_CACHE)) @@ -470,6 +470,8 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, if (!(b->features & BLK_FEAT_POLL)) t->features &= ~BLK_FEAT_POLL; + t->flags |= (b->flags & BLK_FEAT_MISALIGNED); + t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); t->max_user_sectors = min_not_zero(t->max_user_sectors, b->max_user_sectors); @@ -494,8 +496,6 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, t->max_segment_size = min_not_zero(t->max_segment_size, b->max_segment_size); - t->misaligned |= b->misaligned; - alignment = queue_limit_alignment_offset(b, start); /* Bottom device has different alignment. Check that it is @@ -509,7 +509,7 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, /* Verify that top and bottom intervals line up */ if (max(top, bottom) % min(top, bottom)) { - t->misaligned = 1; + t->flags |= BLK_FEAT_MISALIGNED; ret = -1; } } @@ -531,28 +531,28 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, /* Physical block size a multiple of the logical block size? */ if (t->physical_block_size & (t->logical_block_size - 1)) { t->physical_block_size = t->logical_block_size; - t->misaligned = 1; + t->flags |= BLK_FEAT_MISALIGNED; ret = -1; } /* Minimum I/O a multiple of the physical block size? */ if (t->io_min & (t->physical_block_size - 1)) { t->io_min = t->physical_block_size; - t->misaligned = 1; + t->flags |= BLK_FEAT_MISALIGNED; ret = -1; } /* Optimal I/O a multiple of the physical block size? */ if (t->io_opt & (t->physical_block_size - 1)) { t->io_opt = 0; - t->misaligned = 1; + t->flags |= BLK_FEAT_MISALIGNED; ret = -1; } /* chunk_sectors a multiple of the physical block size? */ if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) { t->chunk_sectors = 0; - t->misaligned = 1; + t->flags |= BLK_FEAT_MISALIGNED; ret = -1; } @@ -566,7 +566,7 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, /* Verify that new alignment_offset is on a logical block boundary */ if (t->alignment_offset & (t->logical_block_size - 1)) { - t->misaligned = 1; + t->flags |= BLK_FEAT_MISALIGNED; ret = -1; } @@ -729,7 +729,7 @@ int bdev_alignment_offset(struct block_device *bdev) { struct request_queue *q = bdev_get_queue(bdev); - if (q->limits.misaligned) + if (q->limits.flags & BLK_FEAT_MISALIGNED) return -1; if (bdev_is_partition(bdev)) return queue_limit_alignment_offset(&q->limits, diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 713a98b6dbba08..7ad2b1240fc0bf 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -341,6 +341,9 @@ enum { enum { /* do not send FLUSH/FUA commands despite advertising a write cache */ BLK_FLAG_WRITE_CACHE_DISABLED = (1u << 0), + + /* I/O topology is misaligned */ + BLK_FEAT_MISALIGNED = (1u << 1), }; struct queue_limits { @@ -374,7 +377,6 @@ struct queue_limits { unsigned short max_integrity_segments; unsigned short max_discard_segments; - unsigned char misaligned; unsigned char discard_misaligned; unsigned char raid_partial_stripes_expensive; unsigned int max_open_zones; From 4cac3d3a712b5c76d462b29b73b9e58c0b6d9946 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 19 Jun 2024 17:45:37 +0200 Subject: [PATCH 112/209] block: remove the discard_alignment flag queue_limits.discard_alignment is never read except in the places where it is stacked into another limit. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240619154623.450048-6-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 10 ---------- drivers/md/dm-cache-target.c | 1 - drivers/md/dm-clone-target.c | 1 - drivers/md/dm-table.c | 1 - include/linux/blkdev.h | 1 - 5 files changed, 14 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index a1b10404e500bc..62588d9a38e39a 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -578,16 +578,6 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, if (b->discard_granularity) { alignment = queue_limit_discard_alignment(b, start); - if (t->discard_granularity != 0 && - t->discard_alignment != alignment) { - top = t->discard_granularity + t->discard_alignment; - bottom = b->discard_granularity + alignment; - - /* Verify that top and bottom intervals line up */ - if ((max(top, bottom) % min(top, bottom)) != 0) - t->discard_misaligned = 1; - } - t->max_discard_sectors = min_not_zero(t->max_discard_sectors, b->max_discard_sectors); t->max_hw_discard_sectors = min_not_zero(t->max_hw_discard_sectors, diff --git a/drivers/md/dm-cache-target.c b/drivers/md/dm-cache-target.c index 16884b5850532f..2d8dd9283ff4cf 100644 --- a/drivers/md/dm-cache-target.c +++ b/drivers/md/dm-cache-target.c @@ -3403,7 +3403,6 @@ static void set_discard_limits(struct cache *cache, struct queue_limits *limits) limits->max_hw_discard_sectors = origin_limits->max_hw_discard_sectors; limits->discard_granularity = origin_limits->discard_granularity; limits->discard_alignment = origin_limits->discard_alignment; - limits->discard_misaligned = origin_limits->discard_misaligned; } static void cache_io_hints(struct dm_target *ti, struct queue_limits *limits) diff --git a/drivers/md/dm-clone-target.c b/drivers/md/dm-clone-target.c index ad79b52ffc1434..b4384a8b13e360 100644 --- a/drivers/md/dm-clone-target.c +++ b/drivers/md/dm-clone-target.c @@ -2059,7 +2059,6 @@ static void set_discard_limits(struct clone *clone, struct queue_limits *limits) limits->max_hw_discard_sectors = dest_limits->max_hw_discard_sectors; limits->discard_granularity = dest_limits->discard_granularity; limits->discard_alignment = dest_limits->discard_alignment; - limits->discard_misaligned = dest_limits->discard_misaligned; limits->max_discard_segments = dest_limits->max_discard_segments; } diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index df6313c3fe6ba4..502ebc78d490f6 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -1808,7 +1808,6 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q, limits->max_hw_discard_sectors = 0; limits->discard_granularity = 0; limits->discard_alignment = 0; - limits->discard_misaligned = 0; } if (!dm_table_supports_write_zeroes(t)) diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 7ad2b1240fc0bf..86410ce41bf60e 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -377,7 +377,6 @@ struct queue_limits { unsigned short max_integrity_segments; unsigned short max_discard_segments; - unsigned char discard_misaligned; unsigned char raid_partial_stripes_expensive; unsigned int max_open_zones; unsigned int max_active_zones; From 7d4dec525f5fd555037486af4d02dd3682655ba1 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 19 Jun 2024 17:45:38 +0200 Subject: [PATCH 113/209] block: move the raid_partial_stripes_expensive flag into the features field Move the raid_partial_stripes_expensive flags into the features field to reclaim a little bit of space. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240619154623.450048-7-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 4 ---- drivers/md/bcache/super.c | 4 ++-- drivers/md/raid5.c | 2 +- include/linux/blkdev.h | 7 +++++-- 4 files changed, 8 insertions(+), 9 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index 62588d9a38e39a..008fed84edb432 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -556,10 +556,6 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, ret = -1; } - t->raid_partial_stripes_expensive = - max(t->raid_partial_stripes_expensive, - b->raid_partial_stripes_expensive); - /* Find lowest common alignment_offset */ t->alignment_offset = lcm_not_zero(t->alignment_offset, alignment) % max(t->physical_block_size, t->io_min); diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c index baa364eedd0051..283b2511c6d21f 100644 --- a/drivers/md/bcache/super.c +++ b/drivers/md/bcache/super.c @@ -1416,8 +1416,8 @@ static int cached_dev_init(struct cached_dev *dc, unsigned int block_size) } if (bdev_io_opt(dc->bdev)) - dc->partial_stripes_expensive = - q->limits.raid_partial_stripes_expensive; + dc->partial_stripes_expensive = q->limits.features & + BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE; ret = bcache_device_init(&dc->disk, block_size, bdev_nr_sectors(dc->bdev) - dc->sb.data_offset, diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index e875763d69917d..72f91eaa3201c4 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -7707,7 +7707,7 @@ static int raid5_set_limits(struct mddev *mddev) blk_set_stacking_limits(&lim); lim.io_min = mddev->chunk_sectors << 9; lim.io_opt = lim.io_min * (conf->raid_disks - conf->max_degraded); - lim.raid_partial_stripes_expensive = 1; + lim.features |= BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE; lim.discard_granularity = stripe; lim.max_write_zeroes_sectors = 0; mddev_stack_rdev_limits(mddev, &lim, 0); diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 86410ce41bf60e..1fa2b148c20696 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -328,6 +328,9 @@ enum { /* bounce all highmem pages */ BLK_FEAT_BOUNCE_HIGH = (1u << 14), + + /* undocumented magic for bcache */ + BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE = (1u << 15), }; /* @@ -335,7 +338,8 @@ enum { */ #define BLK_FEAT_INHERIT_MASK \ (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \ - BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | BLK_FEAT_BOUNCE_HIGH) + BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | BLK_FEAT_BOUNCE_HIGH | \ + BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE) /* internal flags in queue_limits.flags */ enum { @@ -377,7 +381,6 @@ struct queue_limits { unsigned short max_integrity_segments; unsigned short max_discard_segments; - unsigned char raid_partial_stripes_expensive; unsigned int max_open_zones; unsigned int max_active_zones; From 5ddb88f22eb97218d9295e69c39e0ff7cc64e09c Mon Sep 17 00:00:00 2001 From: Andreas Hindborg Date: Thu, 20 Jun 2024 10:57:21 +0200 Subject: [PATCH 114/209] rust: block: do not use removed queue flag API `blk_queue_flag_set` and `blk_queue_flag_clear` was removed in favor of a new API. This caused a build error for Rust block device abstractions. Thus, use the new feature passing API instead of the old removed API. Fixes: bd4a633b6f7c ("block: move the nonrot flag to queue_limits") Signed-off-by: Andreas Hindborg Link: https://lore.kernel.org/r/20240620085721.1218296-1-nmi@metaspace.dk Signed-off-by: Jens Axboe --- rust/kernel/block/mq/gen_disk.rs | 17 +++-------------- 1 file changed, 3 insertions(+), 14 deletions(-) diff --git a/rust/kernel/block/mq/gen_disk.rs b/rust/kernel/block/mq/gen_disk.rs index e06044b549e0be..f548a619984778 100644 --- a/rust/kernel/block/mq/gen_disk.rs +++ b/rust/kernel/block/mq/gen_disk.rs @@ -100,6 +100,9 @@ impl GenDiskBuilder { lim.logical_block_size = self.logical_block_size; lim.physical_block_size = self.physical_block_size; + if self.rotational { + lim.features = bindings::BLK_FEAT_ROTATIONAL; + } // SAFETY: `tagset.raw_tag_set()` points to a valid and initialized tag set let gendisk = from_err_ptr(unsafe { @@ -152,20 +155,6 @@ impl GenDiskBuilder { // operation, so we will not race. unsafe { bindings::set_capacity(gendisk, self.capacity_sectors) }; - if !self.rotational { - // SAFETY: `gendisk` points to a valid and initialized instance of - // `struct gendisk`. This operation uses a relaxed atomic bit flip - // operation, so there is no race on this field. - unsafe { bindings::blk_queue_flag_set(bindings::QUEUE_FLAG_NONROT, (*gendisk).queue) }; - } else { - // SAFETY: `gendisk` points to a valid and initialized instance of - // `struct gendisk`. This operation uses a relaxed atomic bit flip - // operation, so there is no race on this field. - unsafe { - bindings::blk_queue_flag_clear(bindings::QUEUE_FLAG_NONROT, (*gendisk).queue) - }; - } - crate::error::to_result( // SAFETY: `gendisk` points to a valid and initialized instance of // `struct gendisk`. From 8d1dfd51c84e202df05a999ce82cb27554f7d152 Mon Sep 17 00:00:00 2001 From: John Garry Date: Thu, 20 Jun 2024 12:53:50 +0000 Subject: [PATCH 115/209] block: Pass blk_queue_get_max_sectors() a request pointer Currently blk_queue_get_max_sectors() is passed a enum req_op. In future the value returned from blk_queue_get_max_sectors() may depend on certain request flags, so pass a request pointer. Reviewed-by: Christoph Hellwig Reviewed-by: Keith Busch Reviewed-by: Luis Chamberlain Reviewed-by: Martin K. Petersen Signed-off-by: John Garry Reviewed-by: Hannes Reinecke Acked-by: Darrick J. Wong Reviewed-by: Darrick J. Wong Link: https://lore.kernel.org/r/20240620125359.2684798-2-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- block/blk-merge.c | 3 ++- block/blk-mq.c | 2 +- block/blk.h | 6 ++++-- 3 files changed, 7 insertions(+), 4 deletions(-) diff --git a/block/blk-merge.c b/block/blk-merge.c index 8534c35e04976a..8957e08e020c2d 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -593,7 +593,8 @@ static inline unsigned int blk_rq_get_max_sectors(struct request *rq, if (blk_rq_is_passthrough(rq)) return q->limits.max_hw_sectors; - max_sectors = blk_queue_get_max_sectors(q, req_op(rq)); + max_sectors = blk_queue_get_max_sectors(rq); + if (!q->limits.chunk_sectors || req_op(rq) == REQ_OP_DISCARD || req_op(rq) == REQ_OP_SECURE_ERASE) diff --git a/block/blk-mq.c b/block/blk-mq.c index e2b9710ddc5ad1..47fe9d19b8f109 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -3036,7 +3036,7 @@ void blk_mq_submit_bio(struct bio *bio) blk_status_t blk_insert_cloned_request(struct request *rq) { struct request_queue *q = rq->q; - unsigned int max_sectors = blk_queue_get_max_sectors(q, req_op(rq)); + unsigned int max_sectors = blk_queue_get_max_sectors(rq); unsigned int max_segments = blk_rq_get_max_segments(rq); blk_status_t ret; diff --git a/block/blk.h b/block/blk.h index fa32f7fad5d7e6..20c5718815e2e7 100644 --- a/block/blk.h +++ b/block/blk.h @@ -182,9 +182,11 @@ static inline unsigned int blk_rq_get_max_segments(struct request *rq) return queue_max_segments(rq->q); } -static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q, - enum req_op op) +static inline unsigned int blk_queue_get_max_sectors(struct request *rq) { + struct request_queue *q = rq->q; + enum req_op op = req_op(rq); + if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE)) return min(q->limits.max_discard_sectors, UINT_MAX >> SECTOR_SHIFT); From f70167a7a6e7e8a6911f3a216dc044cbfe7c1983 Mon Sep 17 00:00:00 2001 From: John Garry Date: Thu, 20 Jun 2024 12:53:51 +0000 Subject: [PATCH 116/209] block: Generalize chunk_sectors support as boundary support The purpose of the chunk_sectors limit is to ensure that a mergeble request fits within the boundary of the chunck_sector value. Such a feature will be useful for other request_queue boundary limits, so generalize the chunk_sectors merge code. This idea was proposed by Hannes Reinecke. Reviewed-by: Martin K. Petersen Signed-off-by: John Garry Reviewed-by: Hannes Reinecke Acked-by: Darrick J. Wong Reviewed-by: Darrick J. Wong Link: https://lore.kernel.org/r/20240620125359.2684798-3-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- block/blk-merge.c | 20 ++++++++++++++------ drivers/md/dm.c | 2 +- include/linux/blkdev.h | 13 +++++++------ 3 files changed, 22 insertions(+), 13 deletions(-) diff --git a/block/blk-merge.c b/block/blk-merge.c index 8957e08e020c2d..68969e27c83193 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -154,6 +154,11 @@ static struct bio *bio_split_write_zeroes(struct bio *bio, return bio_split(bio, lim->max_write_zeroes_sectors, GFP_NOIO, bs); } +static inline unsigned int blk_boundary_sectors(const struct queue_limits *lim) +{ + return lim->chunk_sectors; +} + /* * Return the maximum number of sectors from the start of a bio that may be * submitted as a single request to a block device. If enough sectors remain, @@ -167,12 +172,13 @@ static inline unsigned get_max_io_size(struct bio *bio, { unsigned pbs = lim->physical_block_size >> SECTOR_SHIFT; unsigned lbs = lim->logical_block_size >> SECTOR_SHIFT; + unsigned boundary_sectors = blk_boundary_sectors(lim); unsigned max_sectors = lim->max_sectors, start, end; - if (lim->chunk_sectors) { + if (boundary_sectors) { max_sectors = min(max_sectors, - blk_chunk_sectors_left(bio->bi_iter.bi_sector, - lim->chunk_sectors)); + blk_boundary_sectors_left(bio->bi_iter.bi_sector, + boundary_sectors)); } start = bio->bi_iter.bi_sector & (pbs - 1); @@ -588,19 +594,21 @@ static inline unsigned int blk_rq_get_max_sectors(struct request *rq, sector_t offset) { struct request_queue *q = rq->q; - unsigned int max_sectors; + struct queue_limits *lim = &q->limits; + unsigned int max_sectors, boundary_sectors; if (blk_rq_is_passthrough(rq)) return q->limits.max_hw_sectors; + boundary_sectors = blk_boundary_sectors(lim); max_sectors = blk_queue_get_max_sectors(rq); - if (!q->limits.chunk_sectors || + if (!boundary_sectors || req_op(rq) == REQ_OP_DISCARD || req_op(rq) == REQ_OP_SECURE_ERASE) return max_sectors; return min(max_sectors, - blk_chunk_sectors_left(offset, q->limits.chunk_sectors)); + blk_boundary_sectors_left(offset, boundary_sectors)); } static inline int ll_new_hw_segment(struct request *req, struct bio *bio, diff --git a/drivers/md/dm.c b/drivers/md/dm.c index 8a976cee448bed..7d107ae06e1ae1 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -1188,7 +1188,7 @@ static sector_t __max_io_len(struct dm_target *ti, sector_t sector, return len; return min_t(sector_t, len, min(max_sectors ? : queue_max_sectors(ti->table->md->queue), - blk_chunk_sectors_left(target_offset, max_granularity))); + blk_boundary_sectors_left(target_offset, max_granularity))); } static inline sector_t max_io_len(struct dm_target *ti, sector_t sector) diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 0e8253c1507a51..fb7d4c21bba87f 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -907,14 +907,15 @@ static inline bool bio_straddles_zones(struct bio *bio) } /* - * Return how much of the chunk is left to be used for I/O at a given offset. + * Return how much within the boundary is left to be used for I/O at a given + * offset. */ -static inline unsigned int blk_chunk_sectors_left(sector_t offset, - unsigned int chunk_sectors) +static inline unsigned int blk_boundary_sectors_left(sector_t offset, + unsigned int boundary_sectors) { - if (unlikely(!is_power_of_2(chunk_sectors))) - return chunk_sectors - sector_div(offset, chunk_sectors); - return chunk_sectors - (offset & (chunk_sectors - 1)); + if (unlikely(!is_power_of_2(boundary_sectors))) + return boundary_sectors - sector_div(offset, boundary_sectors); + return boundary_sectors - (offset & (boundary_sectors - 1)); } /** From c34fc6f26ab86d03a2d47446f42b6cd492dfdc56 Mon Sep 17 00:00:00 2001 From: Prasad Singamsetty Date: Thu, 20 Jun 2024 12:53:52 +0000 Subject: [PATCH 117/209] fs: Initial atomic write support An atomic write is a write issued with torn-write protection, meaning that for a power failure or any other hardware failure, all or none of the data from the write will be stored, but never a mix of old and new data. Userspace may add flag RWF_ATOMIC to pwritev2() to indicate that the write is to be issued with torn-write prevention, according to special alignment and length rules. For any syscall interface utilizing struct iocb, add IOCB_ATOMIC for iocb->ki_flags field to indicate the same. A call to statx will give the relevant atomic write info for a file: - atomic_write_unit_min - atomic_write_unit_max - atomic_write_segments_max Both min and max values must be a power-of-2. Applications can avail of atomic write feature by ensuring that the total length of a write is a power-of-2 in size and also sized between atomic_write_unit_min and atomic_write_unit_max, inclusive. Applications must ensure that the write is at a naturally-aligned offset in the file wrt the total write length. The value in atomic_write_segments_max indicates the upper limit for IOV_ITER iovcnt. Add file mode flag FMODE_CAN_ATOMIC_WRITE, so files which do not have the flag set will have RWF_ATOMIC rejected and not just ignored. Add a type argument to kiocb_set_rw_flags() to allows reads which have RWF_ATOMIC set to be rejected. Helper function generic_atomic_write_valid() can be used by FSes to verify compliant writes. There we check for iov_iter type is for ubuf, which implies iovcnt==1 for pwritev2(), which is an initial restriction for atomic_write_segments_max. Initially the only user will be bdev file operations write handler. We will rely on the block BIO submission path to ensure write sizes are compliant for the bdev, so we don't need to check atomic writes sizes yet. Signed-off-by: Prasad Singamsetty jpg: merge into single patch and much rewrite Acked-by: Darrick J. Wong Reviewed-by: Martin K. Petersen Signed-off-by: John Garry Reviewed-by: Darrick J. Wong Link: https://lore.kernel.org/r/20240620125359.2684798-4-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- fs/aio.c | 8 ++++---- fs/btrfs/ioctl.c | 2 +- fs/read_write.c | 18 +++++++++++++++++- include/linux/fs.h | 17 +++++++++++++++-- include/uapi/linux/fs.h | 5 ++++- io_uring/rw.c | 9 ++++----- 6 files changed, 45 insertions(+), 14 deletions(-) diff --git a/fs/aio.c b/fs/aio.c index 57c9f7c077e601..93ef59d358b389 100644 --- a/fs/aio.c +++ b/fs/aio.c @@ -1516,7 +1516,7 @@ static void aio_complete_rw(struct kiocb *kiocb, long res) iocb_put(iocb); } -static int aio_prep_rw(struct kiocb *req, const struct iocb *iocb) +static int aio_prep_rw(struct kiocb *req, const struct iocb *iocb, int rw_type) { int ret; @@ -1542,7 +1542,7 @@ static int aio_prep_rw(struct kiocb *req, const struct iocb *iocb) } else req->ki_ioprio = get_current_ioprio(); - ret = kiocb_set_rw_flags(req, iocb->aio_rw_flags); + ret = kiocb_set_rw_flags(req, iocb->aio_rw_flags, rw_type); if (unlikely(ret)) return ret; @@ -1594,7 +1594,7 @@ static int aio_read(struct kiocb *req, const struct iocb *iocb, struct file *file; int ret; - ret = aio_prep_rw(req, iocb); + ret = aio_prep_rw(req, iocb, READ); if (ret) return ret; file = req->ki_filp; @@ -1621,7 +1621,7 @@ static int aio_write(struct kiocb *req, const struct iocb *iocb, struct file *file; int ret; - ret = aio_prep_rw(req, iocb); + ret = aio_prep_rw(req, iocb, WRITE); if (ret) return ret; file = req->ki_filp; diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index efd5d6e9589e09..6ad524b894fc85 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -4627,7 +4627,7 @@ static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool goto out_iov; init_sync_kiocb(&kiocb, file); - ret = kiocb_set_rw_flags(&kiocb, 0); + ret = kiocb_set_rw_flags(&kiocb, 0, WRITE); if (ret) goto out_iov; kiocb.ki_pos = pos; diff --git a/fs/read_write.c b/fs/read_write.c index ef6339391351c2..90e283b31ca181 100644 --- a/fs/read_write.c +++ b/fs/read_write.c @@ -730,7 +730,7 @@ static ssize_t do_iter_readv_writev(struct file *filp, struct iov_iter *iter, ssize_t ret; init_sync_kiocb(&kiocb, filp); - ret = kiocb_set_rw_flags(&kiocb, flags); + ret = kiocb_set_rw_flags(&kiocb, flags, type); if (ret) return ret; kiocb.ki_pos = (ppos ? *ppos : 0); @@ -1736,3 +1736,19 @@ int generic_file_rw_checks(struct file *file_in, struct file *file_out) return 0; } + +bool generic_atomic_write_valid(struct iov_iter *iter, loff_t pos) +{ + size_t len = iov_iter_count(iter); + + if (!iter_is_ubuf(iter)) + return false; + + if (!is_power_of_2(len)) + return false; + + if (!IS_ALIGNED(pos, len)) + return false; + + return true; +} diff --git a/include/linux/fs.h b/include/linux/fs.h index 0283cf366c2a93..e049414bef7da6 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -125,8 +125,10 @@ typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset, #define FMODE_EXEC ((__force fmode_t)(1 << 5)) /* File writes are restricted (block device specific) */ #define FMODE_WRITE_RESTRICTED ((__force fmode_t)(1 << 6)) +/* File supports atomic writes */ +#define FMODE_CAN_ATOMIC_WRITE ((__force fmode_t)(1 << 7)) -/* FMODE_* bits 7 to 8 */ +/* FMODE_* bit 8 */ /* 32bit hashes as llseek() offset (for directories) */ #define FMODE_32BITHASH ((__force fmode_t)(1 << 9)) @@ -317,6 +319,7 @@ struct readahead_control; #define IOCB_SYNC (__force int) RWF_SYNC #define IOCB_NOWAIT (__force int) RWF_NOWAIT #define IOCB_APPEND (__force int) RWF_APPEND +#define IOCB_ATOMIC (__force int) RWF_ATOMIC /* non-RWF related bits - start at 16 */ #define IOCB_EVENTFD (1 << 16) @@ -351,6 +354,7 @@ struct readahead_control; { IOCB_SYNC, "SYNC" }, \ { IOCB_NOWAIT, "NOWAIT" }, \ { IOCB_APPEND, "APPEND" }, \ + { IOCB_ATOMIC, "ATOMIC"}, \ { IOCB_EVENTFD, "EVENTFD"}, \ { IOCB_DIRECT, "DIRECT" }, \ { IOCB_WRITE, "WRITE" }, \ @@ -3403,7 +3407,8 @@ static inline int iocb_flags(struct file *file) return res; } -static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags) +static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags, + int rw_type) { int kiocb_flags = 0; @@ -3422,6 +3427,12 @@ static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags) return -EOPNOTSUPP; kiocb_flags |= IOCB_NOIO; } + if (flags & RWF_ATOMIC) { + if (rw_type != WRITE) + return -EOPNOTSUPP; + if (!(ki->ki_filp->f_mode & FMODE_CAN_ATOMIC_WRITE)) + return -EOPNOTSUPP; + } kiocb_flags |= (__force int) (flags & RWF_SUPPORTED); if (flags & RWF_SYNC) kiocb_flags |= IOCB_DSYNC; @@ -3613,4 +3624,6 @@ extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len, extern int generic_fadvise(struct file *file, loff_t offset, loff_t len, int advice); +bool generic_atomic_write_valid(struct iov_iter *iter, loff_t pos); + #endif /* _LINUX_FS_H */ diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h index 45e4e64fd6643c..191a7e88a8ab7d 100644 --- a/include/uapi/linux/fs.h +++ b/include/uapi/linux/fs.h @@ -329,9 +329,12 @@ typedef int __bitwise __kernel_rwf_t; /* per-IO negation of O_APPEND */ #define RWF_NOAPPEND ((__force __kernel_rwf_t)0x00000020) +/* Atomic Write */ +#define RWF_ATOMIC ((__force __kernel_rwf_t)0x00000040) + /* mask of flags supported by the kernel */ #define RWF_SUPPORTED (RWF_HIPRI | RWF_DSYNC | RWF_SYNC | RWF_NOWAIT |\ - RWF_APPEND | RWF_NOAPPEND) + RWF_APPEND | RWF_NOAPPEND | RWF_ATOMIC) /* Pagemap ioctl */ #define PAGEMAP_SCAN _IOWR('f', 16, struct pm_scan_arg) diff --git a/io_uring/rw.c b/io_uring/rw.c index 1a2128459cb4c8..c004d21e2f12e3 100644 --- a/io_uring/rw.c +++ b/io_uring/rw.c @@ -772,7 +772,7 @@ static bool need_complete_io(struct io_kiocb *req) S_ISBLK(file_inode(req->file)->i_mode); } -static int io_rw_init_file(struct io_kiocb *req, fmode_t mode) +static int io_rw_init_file(struct io_kiocb *req, fmode_t mode, int rw_type) { struct io_rw *rw = io_kiocb_to_cmd(req, struct io_rw); struct kiocb *kiocb = &rw->kiocb; @@ -787,7 +787,7 @@ static int io_rw_init_file(struct io_kiocb *req, fmode_t mode) req->flags |= io_file_get_flags(file); kiocb->ki_flags = file->f_iocb_flags; - ret = kiocb_set_rw_flags(kiocb, rw->flags); + ret = kiocb_set_rw_flags(kiocb, rw->flags, rw_type); if (unlikely(ret)) return ret; kiocb->ki_flags |= IOCB_ALLOC_CACHE; @@ -832,8 +832,7 @@ static int __io_read(struct io_kiocb *req, unsigned int issue_flags) if (unlikely(ret < 0)) return ret; } - - ret = io_rw_init_file(req, FMODE_READ); + ret = io_rw_init_file(req, FMODE_READ, READ); if (unlikely(ret)) return ret; req->cqe.res = iov_iter_count(&io->iter); @@ -1013,7 +1012,7 @@ int io_write(struct io_kiocb *req, unsigned int issue_flags) ssize_t ret, ret2; loff_t *ppos; - ret = io_rw_init_file(req, FMODE_WRITE); + ret = io_rw_init_file(req, FMODE_WRITE, WRITE); if (unlikely(ret)) return ret; req->cqe.res = iov_iter_count(&io->iter); From 0f9ca80fa4f9670ba09721e4e36b8baf086a500c Mon Sep 17 00:00:00 2001 From: Prasad Singamsetty Date: Thu, 20 Jun 2024 12:53:53 +0000 Subject: [PATCH 118/209] fs: Add initial atomic write support info to statx Extend statx system call to return additional info for atomic write support support for a file. Helper function generic_fill_statx_atomic_writes() can be used by FSes to fill in the relevant statx fields. For now atomic_write_segments_max will always be 1, otherwise some rules would need to be imposed on iovec length and alignment, which we don't want now. Signed-off-by: Prasad Singamsetty jpg: relocate bdev support to another patch Reviewed-by: Darrick J. Wong Reviewed-by: Martin K. Petersen Signed-off-by: John Garry Acked-by: Darrick J. Wong Link: https://lore.kernel.org/r/20240620125359.2684798-5-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- fs/stat.c | 34 ++++++++++++++++++++++++++++++++++ include/linux/fs.h | 3 +++ include/linux/stat.h | 3 +++ include/uapi/linux/stat.h | 12 ++++++++++-- 4 files changed, 50 insertions(+), 2 deletions(-) diff --git a/fs/stat.c b/fs/stat.c index 70bd3e888cfa30..72d0e6357b91f1 100644 --- a/fs/stat.c +++ b/fs/stat.c @@ -89,6 +89,37 @@ void generic_fill_statx_attr(struct inode *inode, struct kstat *stat) } EXPORT_SYMBOL(generic_fill_statx_attr); +/** + * generic_fill_statx_atomic_writes - Fill in atomic writes statx attributes + * @stat: Where to fill in the attribute flags + * @unit_min: Minimum supported atomic write length in bytes + * @unit_max: Maximum supported atomic write length in bytes + * + * Fill in the STATX{_ATTR}_WRITE_ATOMIC flags in the kstat structure from + * atomic write unit_min and unit_max values. + */ +void generic_fill_statx_atomic_writes(struct kstat *stat, + unsigned int unit_min, + unsigned int unit_max) +{ + /* Confirm that the request type is known */ + stat->result_mask |= STATX_WRITE_ATOMIC; + + /* Confirm that the file attribute type is known */ + stat->attributes_mask |= STATX_ATTR_WRITE_ATOMIC; + + if (unit_min) { + stat->atomic_write_unit_min = unit_min; + stat->atomic_write_unit_max = unit_max; + /* Initially only allow 1x segment */ + stat->atomic_write_segments_max = 1; + + /* Confirm atomic writes are actually supported */ + stat->attributes |= STATX_ATTR_WRITE_ATOMIC; + } +} +EXPORT_SYMBOL_GPL(generic_fill_statx_atomic_writes); + /** * vfs_getattr_nosec - getattr without security checks * @path: file to get attributes from @@ -659,6 +690,9 @@ cp_statx(const struct kstat *stat, struct statx __user *buffer) tmp.stx_dio_mem_align = stat->dio_mem_align; tmp.stx_dio_offset_align = stat->dio_offset_align; tmp.stx_subvol = stat->subvol; + tmp.stx_atomic_write_unit_min = stat->atomic_write_unit_min; + tmp.stx_atomic_write_unit_max = stat->atomic_write_unit_max; + tmp.stx_atomic_write_segments_max = stat->atomic_write_segments_max; return copy_to_user(buffer, &tmp, sizeof(tmp)) ? -EFAULT : 0; } diff --git a/include/linux/fs.h b/include/linux/fs.h index e049414bef7da6..db26b4a70c628b 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -3235,6 +3235,9 @@ extern const struct inode_operations page_symlink_inode_operations; extern void kfree_link(void *); void generic_fillattr(struct mnt_idmap *, u32, struct inode *, struct kstat *); void generic_fill_statx_attr(struct inode *inode, struct kstat *stat); +void generic_fill_statx_atomic_writes(struct kstat *stat, + unsigned int unit_min, + unsigned int unit_max); extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int); extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int); void __inode_add_bytes(struct inode *inode, loff_t bytes); diff --git a/include/linux/stat.h b/include/linux/stat.h index bf92441dbad284..3d900c86981c5b 100644 --- a/include/linux/stat.h +++ b/include/linux/stat.h @@ -54,6 +54,9 @@ struct kstat { u32 dio_offset_align; u64 change_cookie; u64 subvol; + u32 atomic_write_unit_min; + u32 atomic_write_unit_max; + u32 atomic_write_segments_max; }; /* These definitions are internal to the kernel for now. Mainly used by nfsd. */ diff --git a/include/uapi/linux/stat.h b/include/uapi/linux/stat.h index 67626d53531664..887a2528644168 100644 --- a/include/uapi/linux/stat.h +++ b/include/uapi/linux/stat.h @@ -126,9 +126,15 @@ struct statx { __u64 stx_mnt_id; __u32 stx_dio_mem_align; /* Memory buffer alignment for direct I/O */ __u32 stx_dio_offset_align; /* File offset alignment for direct I/O */ - __u64 stx_subvol; /* Subvolume identifier */ /* 0xa0 */ - __u64 __spare3[11]; /* Spare space for future expansion */ + __u64 stx_subvol; /* Subvolume identifier */ + __u32 stx_atomic_write_unit_min; /* Min atomic write unit in bytes */ + __u32 stx_atomic_write_unit_max; /* Max atomic write unit in bytes */ + /* 0xb0 */ + __u32 stx_atomic_write_segments_max; /* Max atomic write segment count */ + __u32 __spare1[1]; + /* 0xb8 */ + __u64 __spare3[9]; /* Spare space for future expansion */ /* 0x100 */ }; @@ -157,6 +163,7 @@ struct statx { #define STATX_DIOALIGN 0x00002000U /* Want/got direct I/O alignment info */ #define STATX_MNT_ID_UNIQUE 0x00004000U /* Want/got extended stx_mount_id */ #define STATX_SUBVOL 0x00008000U /* Want/got stx_subvol */ +#define STATX_WRITE_ATOMIC 0x00010000U /* Want/got atomic_write_* fields */ #define STATX__RESERVED 0x80000000U /* Reserved for future struct statx expansion */ @@ -192,6 +199,7 @@ struct statx { #define STATX_ATTR_MOUNT_ROOT 0x00002000 /* Root of a mount */ #define STATX_ATTR_VERITY 0x00100000 /* [I] Verity protected file */ #define STATX_ATTR_DAX 0x00200000 /* File is currently in DAX state */ +#define STATX_ATTR_WRITE_ATOMIC 0x00400000 /* File supports atomic write operations */ #endif /* _UAPI_LINUX_STAT_H */ From 9da3d1e912f3953196e66991d75208cde3e845e1 Mon Sep 17 00:00:00 2001 From: John Garry Date: Thu, 20 Jun 2024 12:53:54 +0000 Subject: [PATCH 119/209] block: Add core atomic write support Add atomic write support, as follows: - add helper functions to get request_queue atomic write limits - report request_queue atomic write support limits to sysfs and update Doc - support to safely merge atomic writes - deal with splitting atomic writes - misc helper functions - add a per-request atomic write flag New request_queue limits are added, as follows: - atomic_write_hw_max is set by the block driver and is the maximum length of an atomic write which the device may support. It is not necessarily a power-of-2. - atomic_write_max_sectors is derived from atomic_write_hw_max_sectors and max_hw_sectors. It is always a power-of-2. Atomic writes may be merged, and atomic_write_max_sectors would be the limit on a merged atomic write request size. This value is not capped at max_sectors, as the value in max_sectors can be controlled from userspace, and it would only cause trouble if userspace could limit atomic_write_unit_max_bytes and the other atomic write limits. - atomic_write_hw_unit_{min,max} are set by the block driver and are the min/max length of an atomic write unit which the device may support. They both must be a power-of-2. Typically atomic_write_hw_unit_max will hold the same value as atomic_write_hw_max. - atomic_write_unit_{min,max} are derived from atomic_write_hw_unit_{min,max}, max_hw_sectors, and block core limits. Both min and max values must be a power-of-2. - atomic_write_hw_boundary is set by the block driver. If non-zero, it indicates an LBA space boundary at which an atomic write straddles no longer is atomically executed by the disk. The value must be a power-of-2. Note that it would be acceptable to enforce a rule that atomic_write_hw_boundary_sectors is a multiple of atomic_write_hw_unit_max, but the resultant code would be more complicated. All atomic writes limits are by default set 0 to indicate no atomic write support. Even though it is assumed by Linux that a logical block can always be atomically written, we ignore this as it is not of particular interest. Stacked devices are just not supported either for now. An atomic write must always be submitted to the block driver as part of a single request. As such, only a single BIO must be submitted to the block layer for an atomic write. When a single atomic write BIO is submitted, it cannot be split. As such, atomic_write_unit_{max, min}_bytes are limited by the maximum guaranteed BIO size which will not be required to be split. This max size is calculated by request_queue max segments and the number of bvecs a BIO can fit, BIO_MAX_VECS. Currently we rely on userspace issuing a write with iovcnt=1 for pwritev2() - as such, we can rely on each segment containing PAGE_SIZE of data, apart from the first+last, which each can fit logical block size of data. The first+last will be LBS length/aligned as we rely on direct IO alignment rules also. New sysfs files are added to report the following atomic write limits: - atomic_write_unit_max_bytes - same as atomic_write_unit_max_sectors in bytes - atomic_write_unit_min_bytes - same as atomic_write_unit_min_sectors in bytes - atomic_write_boundary_bytes - same as atomic_write_hw_boundary_sectors in bytes - atomic_write_max_bytes - same as atomic_write_max_sectors in bytes Atomic writes may only be merged with other atomic writes and only under the following conditions: - total resultant request length <= atomic_write_max_bytes - the merged write does not straddle a boundary Helper function bdev_can_atomic_write() is added to indicate whether atomic writes may be issued to a bdev. If a bdev is a partition, the partition start must be aligned with both atomic_write_unit_min_sectors and atomic_write_hw_boundary_sectors. FSes will rely on the block layer to validate that an atomic write BIO submitted will be of valid size, so add blk_validate_atomic_write_op_size() for this purpose. Userspace expects an atomic write which is of invalid size to be rejected with -EINVAL, so add BLK_STS_INVAL for this. Also use BLK_STS_INVAL for when a BIO needs to be split, as this should mean an invalid size BIO. Flag REQ_ATOMIC is used for indicating an atomic write. Co-developed-by: Himanshu Madhani Signed-off-by: Himanshu Madhani Reviewed-by: Martin K. Petersen Signed-off-by: John Garry Reviewed-by: Keith Busch Link: https://lore.kernel.org/r/20240620125359.2684798-6-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- Documentation/ABI/stable/sysfs-block | 53 +++++++++++++++++ block/blk-core.c | 19 ++++++ block/blk-merge.c | 50 ++++++++++++++-- block/blk-settings.c | 88 ++++++++++++++++++++++++++++ block/blk-sysfs.c | 33 +++++++++++ block/blk.h | 3 + include/linux/blk_types.h | 8 ++- include/linux/blkdev.h | 55 +++++++++++++++++ 8 files changed, 304 insertions(+), 5 deletions(-) diff --git a/Documentation/ABI/stable/sysfs-block b/Documentation/ABI/stable/sysfs-block index 831f19a32e080c..cea8856f798dde 100644 --- a/Documentation/ABI/stable/sysfs-block +++ b/Documentation/ABI/stable/sysfs-block @@ -21,6 +21,59 @@ Description: device is offset from the internal allocation unit's natural alignment. +What: /sys/block//atomic_write_max_bytes +Date: February 2024 +Contact: Himanshu Madhani +Description: + [RO] This parameter specifies the maximum atomic write + size reported by the device. This parameter is relevant + for merging of writes, where a merged atomic write + operation must not exceed this number of bytes. + This parameter may be greater than the value in + atomic_write_unit_max_bytes as + atomic_write_unit_max_bytes will be rounded down to a + power-of-two and atomic_write_unit_max_bytes may also be + limited by some other queue limits, such as max_segments. + This parameter - along with atomic_write_unit_min_bytes + and atomic_write_unit_max_bytes - will not be larger than + max_hw_sectors_kb, but may be larger than max_sectors_kb. + + +What: /sys/block//atomic_write_unit_min_bytes +Date: February 2024 +Contact: Himanshu Madhani +Description: + [RO] This parameter specifies the smallest block which can + be written atomically with an atomic write operation. All + atomic write operations must begin at a + atomic_write_unit_min boundary and must be multiples of + atomic_write_unit_min. This value must be a power-of-two. + + +What: /sys/block//atomic_write_unit_max_bytes +Date: February 2024 +Contact: Himanshu Madhani +Description: + [RO] This parameter defines the largest block which can be + written atomically with an atomic write operation. This + value must be a multiple of atomic_write_unit_min and must + be a power-of-two. This value will not be larger than + atomic_write_max_bytes. + + +What: /sys/block//atomic_write_boundary_bytes +Date: February 2024 +Contact: Himanshu Madhani +Description: + [RO] A device may need to internally split an atomic write I/O + which straddles a given logical block address boundary. This + parameter specifies the size in bytes of the atomic boundary if + one is reported by the device. This value must be a + power-of-two and at least the size as in + atomic_write_unit_max_bytes. + Any attempt to merge atomic write I/Os must not result in a + merged I/O which crosses this boundary (if any). + What: /sys/block//diskseq Date: February 2021 diff --git a/block/blk-core.c b/block/blk-core.c index 8d9fbd353fc7fc..6fc1a5a1980db3 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -174,6 +174,8 @@ static const struct { /* Command duration limit device-side timeout */ [BLK_STS_DURATION_LIMIT] = { -ETIME, "duration limit exceeded" }, + [BLK_STS_INVAL] = { -EINVAL, "invalid" }, + /* everything else not covered above: */ [BLK_STS_IOERR] = { -EIO, "I/O" }, }; @@ -739,6 +741,18 @@ void submit_bio_noacct_nocheck(struct bio *bio) __submit_bio_noacct(bio); } +static blk_status_t blk_validate_atomic_write_op_size(struct request_queue *q, + struct bio *bio) +{ + if (bio->bi_iter.bi_size > queue_atomic_write_unit_max_bytes(q)) + return BLK_STS_INVAL; + + if (bio->bi_iter.bi_size % queue_atomic_write_unit_min_bytes(q)) + return BLK_STS_INVAL; + + return BLK_STS_OK; +} + /** * submit_bio_noacct - re-submit a bio to the block device layer for I/O * @bio: The bio describing the location in memory and on the device. @@ -797,6 +811,11 @@ void submit_bio_noacct(struct bio *bio) switch (bio_op(bio)) { case REQ_OP_READ: case REQ_OP_WRITE: + if (bio->bi_opf & REQ_ATOMIC) { + status = blk_validate_atomic_write_op_size(q, bio); + if (status != BLK_STS_OK) + goto end_io; + } break; case REQ_OP_FLUSH: /* diff --git a/block/blk-merge.c b/block/blk-merge.c index 68969e27c83193..cff20bcc0252a7 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -154,8 +154,16 @@ static struct bio *bio_split_write_zeroes(struct bio *bio, return bio_split(bio, lim->max_write_zeroes_sectors, GFP_NOIO, bs); } -static inline unsigned int blk_boundary_sectors(const struct queue_limits *lim) +static inline unsigned int blk_boundary_sectors(const struct queue_limits *lim, + bool is_atomic) { + /* + * chunk_sectors must be a multiple of atomic_write_boundary_sectors if + * both non-zero. + */ + if (is_atomic && lim->atomic_write_boundary_sectors) + return lim->atomic_write_boundary_sectors; + return lim->chunk_sectors; } @@ -172,8 +180,18 @@ static inline unsigned get_max_io_size(struct bio *bio, { unsigned pbs = lim->physical_block_size >> SECTOR_SHIFT; unsigned lbs = lim->logical_block_size >> SECTOR_SHIFT; - unsigned boundary_sectors = blk_boundary_sectors(lim); - unsigned max_sectors = lim->max_sectors, start, end; + bool is_atomic = bio->bi_opf & REQ_ATOMIC; + unsigned boundary_sectors = blk_boundary_sectors(lim, is_atomic); + unsigned max_sectors, start, end; + + /* + * We ignore lim->max_sectors for atomic writes because it may less + * than the actual bio size, which we cannot tolerate. + */ + if (is_atomic) + max_sectors = lim->atomic_write_max_sectors; + else + max_sectors = lim->max_sectors; if (boundary_sectors) { max_sectors = min(max_sectors, @@ -311,6 +329,11 @@ struct bio *bio_split_rw(struct bio *bio, const struct queue_limits *lim, *segs = nsegs; return NULL; split: + if (bio->bi_opf & REQ_ATOMIC) { + bio->bi_status = BLK_STS_INVAL; + bio_endio(bio); + return ERR_PTR(-EINVAL); + } /* * We can't sanely support splitting for a REQ_NOWAIT bio. End it * with EAGAIN if splitting is required and return an error pointer. @@ -596,11 +619,12 @@ static inline unsigned int blk_rq_get_max_sectors(struct request *rq, struct request_queue *q = rq->q; struct queue_limits *lim = &q->limits; unsigned int max_sectors, boundary_sectors; + bool is_atomic = rq->cmd_flags & REQ_ATOMIC; if (blk_rq_is_passthrough(rq)) return q->limits.max_hw_sectors; - boundary_sectors = blk_boundary_sectors(lim); + boundary_sectors = blk_boundary_sectors(lim, is_atomic); max_sectors = blk_queue_get_max_sectors(rq); if (!boundary_sectors || @@ -806,6 +830,18 @@ static enum elv_merge blk_try_req_merge(struct request *req, return ELEVATOR_NO_MERGE; } +static bool blk_atomic_write_mergeable_rq_bio(struct request *rq, + struct bio *bio) +{ + return (rq->cmd_flags & REQ_ATOMIC) == (bio->bi_opf & REQ_ATOMIC); +} + +static bool blk_atomic_write_mergeable_rqs(struct request *rq, + struct request *next) +{ + return (rq->cmd_flags & REQ_ATOMIC) == (next->cmd_flags & REQ_ATOMIC); +} + /* * For non-mq, this has to be called with the request spinlock acquired. * For mq with scheduling, the appropriate queue wide lock should be held. @@ -829,6 +865,9 @@ static struct request *attempt_merge(struct request_queue *q, if (req->ioprio != next->ioprio) return NULL; + if (!blk_atomic_write_mergeable_rqs(req, next)) + return NULL; + /* * If we are allowed to merge, then append bio list * from next to rq and release next. merge_requests_fn @@ -960,6 +999,9 @@ bool blk_rq_merge_ok(struct request *rq, struct bio *bio) if (rq->ioprio != bio_prio(bio)) return false; + if (blk_atomic_write_mergeable_rq_bio(rq, bio) == false) + return false; + return true; } diff --git a/block/blk-settings.c b/block/blk-settings.c index b193068040569c..37fe4c8f6b6ba1 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -135,6 +135,92 @@ static int blk_validate_integrity_limits(struct queue_limits *lim) return 0; } +/* + * Returns max guaranteed bytes which we can fit in a bio. + * + * We request that an atomic_write is ITER_UBUF iov_iter (so a single vector), + * so we assume that we can fit in at least PAGE_SIZE in a segment, apart from + * the first and last segments. + */ +static +unsigned int blk_queue_max_guaranteed_bio(struct queue_limits *lim) +{ + unsigned int max_segments = min(BIO_MAX_VECS, lim->max_segments); + unsigned int length; + + length = min(max_segments, 2) * lim->logical_block_size; + if (max_segments > 2) + length += (max_segments - 2) * PAGE_SIZE; + + return length; +} + +static void blk_atomic_writes_update_limits(struct queue_limits *lim) +{ + unsigned int unit_limit = min(lim->max_hw_sectors << SECTOR_SHIFT, + blk_queue_max_guaranteed_bio(lim)); + + unit_limit = rounddown_pow_of_two(unit_limit); + + lim->atomic_write_max_sectors = + min(lim->atomic_write_hw_max >> SECTOR_SHIFT, + lim->max_hw_sectors); + lim->atomic_write_unit_min = + min(lim->atomic_write_hw_unit_min, unit_limit); + lim->atomic_write_unit_max = + min(lim->atomic_write_hw_unit_max, unit_limit); + lim->atomic_write_boundary_sectors = + lim->atomic_write_hw_boundary >> SECTOR_SHIFT; +} + +static void blk_validate_atomic_write_limits(struct queue_limits *lim) +{ + unsigned int chunk_sectors = lim->chunk_sectors; + unsigned int boundary_sectors; + + if (!lim->atomic_write_hw_max) + goto unsupported; + + boundary_sectors = lim->atomic_write_hw_boundary >> SECTOR_SHIFT; + + if (boundary_sectors) { + /* + * A feature of boundary support is that it disallows bios to + * be merged which would result in a merged request which + * crosses either a chunk sector or atomic write HW boundary, + * even though chunk sectors may be just set for performance. + * For simplicity, disallow atomic writes for a chunk sector + * which is non-zero and smaller than atomic write HW boundary. + * Furthermore, chunk sectors must be a multiple of atomic + * write HW boundary. Otherwise boundary support becomes + * complicated. + * Devices which do not conform to these rules can be dealt + * with if and when they show up. + */ + if (WARN_ON_ONCE(do_div(chunk_sectors, boundary_sectors))) + goto unsupported; + + /* + * The boundary size just needs to be a multiple of unit_max + * (and not necessarily a power-of-2), so this following check + * could be relaxed in future. + * Furthermore, if needed, unit_max could even be reduced so + * that it is compliant with a !power-of-2 boundary. + */ + if (!is_power_of_2(boundary_sectors)) + goto unsupported; + } + + blk_atomic_writes_update_limits(lim); + return; + +unsupported: + lim->atomic_write_max_sectors = 0; + lim->atomic_write_boundary_sectors = 0; + lim->atomic_write_unit_min = 0; + lim->atomic_write_unit_max = 0; +} + /* * Check that the limits in lim are valid, initialize defaults for unset * values, and cap values based on others where needed. @@ -272,6 +358,8 @@ static int blk_validate_limits(struct queue_limits *lim) if (!(lim->features & BLK_FEAT_WRITE_CACHE)) lim->features &= ~BLK_FEAT_FUA; + blk_validate_atomic_write_limits(lim); + err = blk_validate_integrity_limits(lim); if (err) return err; diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 59e6d111ed059a..1a984179f3acc5 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -118,6 +118,30 @@ static ssize_t queue_max_discard_segments_show(struct request_queue *q, return queue_var_show(queue_max_discard_segments(q), page); } +static ssize_t queue_atomic_write_max_bytes_show(struct request_queue *q, + char *page) +{ + return queue_var_show(queue_atomic_write_max_bytes(q), page); +} + +static ssize_t queue_atomic_write_boundary_show(struct request_queue *q, + char *page) +{ + return queue_var_show(queue_atomic_write_boundary_bytes(q), page); +} + +static ssize_t queue_atomic_write_unit_min_show(struct request_queue *q, + char *page) +{ + return queue_var_show(queue_atomic_write_unit_min_bytes(q), page); +} + +static ssize_t queue_atomic_write_unit_max_show(struct request_queue *q, + char *page) +{ + return queue_var_show(queue_atomic_write_unit_max_bytes(q), page); +} + static ssize_t queue_max_integrity_segments_show(struct request_queue *q, char *page) { return queue_var_show(q->limits.max_integrity_segments, page); @@ -505,6 +529,11 @@ QUEUE_RO_ENTRY(queue_discard_max_hw, "discard_max_hw_bytes"); QUEUE_RW_ENTRY(queue_discard_max, "discard_max_bytes"); QUEUE_RO_ENTRY(queue_discard_zeroes_data, "discard_zeroes_data"); +QUEUE_RO_ENTRY(queue_atomic_write_max_bytes, "atomic_write_max_bytes"); +QUEUE_RO_ENTRY(queue_atomic_write_boundary, "atomic_write_boundary_bytes"); +QUEUE_RO_ENTRY(queue_atomic_write_unit_max, "atomic_write_unit_max_bytes"); +QUEUE_RO_ENTRY(queue_atomic_write_unit_min, "atomic_write_unit_min_bytes"); + QUEUE_RO_ENTRY(queue_write_same_max, "write_same_max_bytes"); QUEUE_RO_ENTRY(queue_write_zeroes_max, "write_zeroes_max_bytes"); QUEUE_RO_ENTRY(queue_zone_append_max, "zone_append_max_bytes"); @@ -626,6 +655,10 @@ static struct attribute *queue_attrs[] = { &queue_discard_max_entry.attr, &queue_discard_max_hw_entry.attr, &queue_discard_zeroes_data_entry.attr, + &queue_atomic_write_max_bytes_entry.attr, + &queue_atomic_write_boundary_entry.attr, + &queue_atomic_write_unit_min_entry.attr, + &queue_atomic_write_unit_max_entry.attr, &queue_write_same_max_entry.attr, &queue_write_zeroes_max_entry.attr, &queue_zone_append_max_entry.attr, diff --git a/block/blk.h b/block/blk.h index 20c5718815e2e7..d0a986d8ee507e 100644 --- a/block/blk.h +++ b/block/blk.h @@ -194,6 +194,9 @@ static inline unsigned int blk_queue_get_max_sectors(struct request *rq) if (unlikely(op == REQ_OP_WRITE_ZEROES)) return q->limits.max_write_zeroes_sectors; + if (rq->cmd_flags & REQ_ATOMIC) + return q->limits.atomic_write_max_sectors; + return q->limits.max_sectors; } diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h index 781c4500491bde..632edd71f8c63c 100644 --- a/include/linux/blk_types.h +++ b/include/linux/blk_types.h @@ -162,6 +162,11 @@ typedef u16 blk_short_t; */ #define BLK_STS_DURATION_LIMIT ((__force blk_status_t)17) +/* + * Invalid size or alignment. + */ +#define BLK_STS_INVAL ((__force blk_status_t)19) + /** * blk_path_error - returns true if error may be path related * @error: status the request was completed with @@ -370,7 +375,7 @@ enum req_flag_bits { __REQ_SWAP, /* swap I/O */ __REQ_DRV, /* for driver use */ __REQ_FS_PRIVATE, /* for file system (submitter) use */ - + __REQ_ATOMIC, /* for atomic write operations */ /* * Command specific flags, keep last: */ @@ -402,6 +407,7 @@ enum req_flag_bits { #define REQ_SWAP (__force blk_opf_t)(1ULL << __REQ_SWAP) #define REQ_DRV (__force blk_opf_t)(1ULL << __REQ_DRV) #define REQ_FS_PRIVATE (__force blk_opf_t)(1ULL << __REQ_FS_PRIVATE) +#define REQ_ATOMIC (__force blk_opf_t)(1ULL << __REQ_ATOMIC) #define REQ_NOUNMAP (__force blk_opf_t)(1ULL << __REQ_NOUNMAP) diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index fb7d4c21bba87f..4816f3b1d52892 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -377,6 +377,16 @@ struct queue_limits { unsigned int discard_alignment; unsigned int zone_write_granularity; + /* atomic write limits */ + unsigned int atomic_write_hw_max; + unsigned int atomic_write_max_sectors; + unsigned int atomic_write_hw_boundary; + unsigned int atomic_write_boundary_sectors; + unsigned int atomic_write_hw_unit_min; + unsigned int atomic_write_unit_min; + unsigned int atomic_write_hw_unit_max; + unsigned int atomic_write_unit_max; + unsigned short max_segments; unsigned short max_integrity_segments; unsigned short max_discard_segments; @@ -1403,6 +1413,30 @@ static inline int queue_dma_alignment(const struct request_queue *q) return q ? q->limits.dma_alignment : 511; } +static inline unsigned int +queue_atomic_write_unit_max_bytes(const struct request_queue *q) +{ + return q->limits.atomic_write_unit_max; +} + +static inline unsigned int +queue_atomic_write_unit_min_bytes(const struct request_queue *q) +{ + return q->limits.atomic_write_unit_min; +} + +static inline unsigned int +queue_atomic_write_boundary_bytes(const struct request_queue *q) +{ + return q->limits.atomic_write_boundary_sectors << SECTOR_SHIFT; +} + +static inline unsigned int +queue_atomic_write_max_bytes(const struct request_queue *q) +{ + return q->limits.atomic_write_max_sectors << SECTOR_SHIFT; +} + static inline unsigned int bdev_dma_alignment(struct block_device *bdev) { return queue_dma_alignment(bdev_get_queue(bdev)); @@ -1644,6 +1678,27 @@ struct io_comp_batch { void (*complete)(struct io_comp_batch *); }; +static inline bool bdev_can_atomic_write(struct block_device *bdev) +{ + struct request_queue *bd_queue = bdev->bd_queue; + struct queue_limits *limits = &bd_queue->limits; + + if (!limits->atomic_write_unit_min) + return false; + + if (bdev_is_partition(bdev)) { + sector_t bd_start_sect = bdev->bd_start_sect; + unsigned int alignment = + max(limits->atomic_write_unit_min, + limits->atomic_write_hw_boundary); + + if (!IS_ALIGNED(bd_start_sect, alignment >> SECTOR_SHIFT)) + return false; + } + + return true; +} + #define DEFINE_IO_COMP_BATCH(name) struct io_comp_batch name = { } #endif /* _LINUX_BLKDEV_H */ From 9abcfbd235f59fb5b6379e5bc0231dad831ebace Mon Sep 17 00:00:00 2001 From: Prasad Singamsetty Date: Thu, 20 Jun 2024 12:53:55 +0000 Subject: [PATCH 120/209] block: Add atomic write support for statx Extend statx system call to return additional info for atomic write support support if the specified file is a block device. Reviewed-by: Martin K. Petersen Signed-off-by: Prasad Singamsetty Signed-off-by: John Garry Reviewed-by: Keith Busch Acked-by: Darrick J. Wong Reviewed-by: Darrick J. Wong Reviewed-by: Luis Chamberlain Link: https://lore.kernel.org/r/20240620125359.2684798-7-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- block/bdev.c | 36 ++++++++++++++++++++++++++---------- fs/stat.c | 16 +++++++++------- include/linux/blkdev.h | 6 ++++-- 3 files changed, 39 insertions(+), 19 deletions(-) diff --git a/block/bdev.c b/block/bdev.c index ced4ac990ec88f..1b4af2cc3b1e6c 100644 --- a/block/bdev.c +++ b/block/bdev.c @@ -1260,23 +1260,39 @@ void sync_bdevs(bool wait) } /* - * Handle STATX_DIOALIGN for block devices. - * - * Note that the inode passed to this is the inode of a block device node file, - * not the block device's internal inode. Therefore it is *not* valid to use - * I_BDEV() here; the block device has to be looked up by i_rdev instead. + * Handle STATX_{DIOALIGN, WRITE_ATOMIC} for block devices. */ -void bdev_statx_dioalign(struct inode *inode, struct kstat *stat) +void bdev_statx(struct inode *backing_inode, struct kstat *stat, + u32 request_mask) { struct block_device *bdev; - bdev = blkdev_get_no_open(inode->i_rdev); + if (!(request_mask & (STATX_DIOALIGN | STATX_WRITE_ATOMIC))) + return; + + /* + * Note that backing_inode is the inode of a block device node file, + * not the block device's internal inode. Therefore it is *not* valid + * to use I_BDEV() here; the block device has to be looked up by i_rdev + * instead. + */ + bdev = blkdev_get_no_open(backing_inode->i_rdev); if (!bdev) return; - stat->dio_mem_align = bdev_dma_alignment(bdev) + 1; - stat->dio_offset_align = bdev_logical_block_size(bdev); - stat->result_mask |= STATX_DIOALIGN; + if (request_mask & STATX_DIOALIGN) { + stat->dio_mem_align = bdev_dma_alignment(bdev) + 1; + stat->dio_offset_align = bdev_logical_block_size(bdev); + stat->result_mask |= STATX_DIOALIGN; + } + + if (request_mask & STATX_WRITE_ATOMIC && bdev_can_atomic_write(bdev)) { + struct request_queue *bd_queue = bdev->bd_queue; + + generic_fill_statx_atomic_writes(stat, + queue_atomic_write_unit_min_bytes(bd_queue), + queue_atomic_write_unit_max_bytes(bd_queue)); + } blkdev_put_no_open(bdev); } diff --git a/fs/stat.c b/fs/stat.c index 72d0e6357b91f1..bd0698dfd7b36e 100644 --- a/fs/stat.c +++ b/fs/stat.c @@ -265,6 +265,7 @@ static int vfs_statx(int dfd, struct filename *filename, int flags, { struct path path; unsigned int lookup_flags = getname_statx_lookup_flags(flags); + struct inode *backing_inode; int error; if (flags & ~(AT_SYMLINK_NOFOLLOW | AT_NO_AUTOMOUNT | AT_EMPTY_PATH | @@ -290,13 +291,14 @@ static int vfs_statx(int dfd, struct filename *filename, int flags, stat->attributes |= STATX_ATTR_MOUNT_ROOT; stat->attributes_mask |= STATX_ATTR_MOUNT_ROOT; - /* Handle STATX_DIOALIGN for block devices. */ - if (request_mask & STATX_DIOALIGN) { - struct inode *inode = d_backing_inode(path.dentry); - - if (S_ISBLK(inode->i_mode)) - bdev_statx_dioalign(inode, stat); - } + /* + * If this is a block device inode, override the filesystem + * attributes with the block device specific parameters that need to be + * obtained from the bdev backing inode. + */ + backing_inode = d_backing_inode(path.dentry); + if (S_ISBLK(backing_inode->i_mode)) + bdev_statx(backing_inode, stat, request_mask); path_put(&path); if (retry_estale(error, lookup_flags)) { diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 4816f3b1d52892..2800231046a98b 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -1638,7 +1638,8 @@ int sync_blockdev(struct block_device *bdev); int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend); int sync_blockdev_nowait(struct block_device *bdev); void sync_bdevs(bool wait); -void bdev_statx_dioalign(struct inode *inode, struct kstat *stat); +void bdev_statx(struct inode *backing_inode, struct kstat *stat, + u32 request_mask); void printk_all_partitions(void); int __init early_lookup_bdev(const char *pathname, dev_t *dev); #else @@ -1656,7 +1657,8 @@ static inline int sync_blockdev_nowait(struct block_device *bdev) static inline void sync_bdevs(bool wait) { } -static inline void bdev_statx_dioalign(struct inode *inode, struct kstat *stat) +static inline void bdev_statx(struct inode *backing_inode, struct kstat *stat, + u32 request_mask) { } static inline void printk_all_partitions(void) From caf336f81b3a3ca744e335972e86ec7244512d4a Mon Sep 17 00:00:00 2001 From: John Garry Date: Thu, 20 Jun 2024 12:53:56 +0000 Subject: [PATCH 121/209] block: Add fops atomic write support Support atomic writes by submitting a single BIO with the REQ_ATOMIC set. It must be ensured that the atomic write adheres to its rules, like naturally aligned offset, so call blkdev_dio_invalid() -> blkdev_atomic_write_valid() [with renaming blkdev_dio_unaligned() to blkdev_dio_invalid()] for this purpose. The BIO submission path currently checks for atomic writes which are too large, so no need to check here. In blkdev_direct_IO(), if the nr_pages exceeds BIO_MAX_VECS, then we cannot produce a single BIO, so error in this case. Finally set FMODE_CAN_ATOMIC_WRITE when the bdev can support atomic writes and the associated file flag is for O_DIRECT. Reviewed-by: Martin K. Petersen Signed-off-by: John Garry Reviewed-by: Keith Busch Acked-by: Darrick J. Wong Reviewed-by: Darrick J. Wong Link: https://lore.kernel.org/r/20240620125359.2684798-8-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- block/fops.c | 20 +++++++++++++++++--- 1 file changed, 17 insertions(+), 3 deletions(-) diff --git a/block/fops.c b/block/fops.c index 376265935714fd..be36c9fbd500bb 100644 --- a/block/fops.c +++ b/block/fops.c @@ -34,9 +34,12 @@ static blk_opf_t dio_bio_write_op(struct kiocb *iocb) return opf; } -static bool blkdev_dio_unaligned(struct block_device *bdev, loff_t pos, - struct iov_iter *iter) +static bool blkdev_dio_invalid(struct block_device *bdev, loff_t pos, + struct iov_iter *iter, bool is_atomic) { + if (is_atomic && !generic_atomic_write_valid(iter, pos)) + return true; + return pos & (bdev_logical_block_size(bdev) - 1) || !bdev_iter_is_aligned(bdev, iter); } @@ -72,6 +75,8 @@ static ssize_t __blkdev_direct_IO_simple(struct kiocb *iocb, bio.bi_iter.bi_sector = pos >> SECTOR_SHIFT; bio.bi_write_hint = file_inode(iocb->ki_filp)->i_write_hint; bio.bi_ioprio = iocb->ki_ioprio; + if (iocb->ki_flags & IOCB_ATOMIC) + bio.bi_opf |= REQ_ATOMIC; ret = bio_iov_iter_get_pages(&bio, iter); if (unlikely(ret)) @@ -343,6 +348,9 @@ static ssize_t __blkdev_direct_IO_async(struct kiocb *iocb, task_io_account_write(bio->bi_iter.bi_size); } + if (iocb->ki_flags & IOCB_ATOMIC) + bio->bi_opf |= REQ_ATOMIC; + if (iocb->ki_flags & IOCB_NOWAIT) bio->bi_opf |= REQ_NOWAIT; @@ -359,12 +367,13 @@ static ssize_t __blkdev_direct_IO_async(struct kiocb *iocb, static ssize_t blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter) { struct block_device *bdev = I_BDEV(iocb->ki_filp->f_mapping->host); + bool is_atomic = iocb->ki_flags & IOCB_ATOMIC; unsigned int nr_pages; if (!iov_iter_count(iter)) return 0; - if (blkdev_dio_unaligned(bdev, iocb->ki_pos, iter)) + if (blkdev_dio_invalid(bdev, iocb->ki_pos, iter, is_atomic)) return -EINVAL; nr_pages = bio_iov_vecs_to_alloc(iter, BIO_MAX_VECS + 1); @@ -373,6 +382,8 @@ static ssize_t blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter) return __blkdev_direct_IO_simple(iocb, iter, bdev, nr_pages); return __blkdev_direct_IO_async(iocb, iter, bdev, nr_pages); + } else if (is_atomic) { + return -EINVAL; } return __blkdev_direct_IO(iocb, iter, bdev, bio_max_segs(nr_pages)); } @@ -612,6 +623,9 @@ static int blkdev_open(struct inode *inode, struct file *filp) if (!bdev) return -ENXIO; + if (bdev_can_atomic_write(bdev) && filp->f_flags & O_DIRECT) + filp->f_mode |= FMODE_CAN_ATOMIC_WRITE; + ret = bdev_open(bdev, mode, filp->private_data, NULL, filp); if (ret) blkdev_put_no_open(bdev); From bf4ae8f2e6407a779c0368eb0f3e047a8333be17 Mon Sep 17 00:00:00 2001 From: John Garry Date: Thu, 20 Jun 2024 12:53:57 +0000 Subject: [PATCH 122/209] scsi: sd: Atomic write support Support is divided into two main areas: - reading VPD pages and setting sdev request_queue limits - support WRITE ATOMIC (16) command and tracing The relevant block limits VPD page need to be read to allow the block layer request_queue atomic write limits to be set. These VPD page limits are described in sbc4r22 section 6.6.4 - Block limits VPD page. There are five limits of interest: - MAXIMUM ATOMIC TRANSFER LENGTH - ATOMIC ALIGNMENT - ATOMIC TRANSFER LENGTH GRANULARITY - MAXIMUM ATOMIC TRANSFER LENGTH WITH BOUNDARY - MAXIMUM ATOMIC BOUNDARY SIZE MAXIMUM ATOMIC TRANSFER LENGTH is the maximum length for a WRITE ATOMIC (16) command. It will not be greater than the device MAXIMUM TRANSFER LENGTH. ATOMIC ALIGNMENT and ATOMIC TRANSFER LENGTH GRANULARITY are the minimum alignment and length values for an atomic write in terms of logical blocks. Unlike NVMe, SCSI does not specify an LBA space boundary, but does specify a per-IO boundary granularity. The maximum boundary size is specified in MAXIMUM ATOMIC BOUNDARY SIZE. When used, this boundary value is set in the WRITE ATOMIC (16) ATOMIC BOUNDARY field - layout for the WRITE_ATOMIC_16 command can be found in sbc4r22 section 5.48. This boundary value is the granularity size at which the device may atomically write the data. A value of zero in WRITE ATOMIC (16) ATOMIC BOUNDARY field means that all data must be atomically written together. MAXIMUM ATOMIC TRANSFER LENGTH WITH BOUNDARY is the maximum atomic write length if a non-zero boundary value is set. For atomic write support, the WRITE ATOMIC (16) boundary is not of much interest, as the block layer expects each request submitted to be executed atomically. However, the SCSI spec does leave itself open to a quirky scenario where MAXIMUM ATOMIC TRANSFER LENGTH is zero, yet MAXIMUM ATOMIC TRANSFER LENGTH WITH BOUNDARY and MAXIMUM ATOMIC BOUNDARY SIZE are both non-zero. This case will be supported. To set the block layer request_queue atomic write capabilities, sanitize the VPD page limits and set limits as follows: - atomic_write_unit_min is derived from granularity and alignment values. If no granularity value is not set, use physical block size - atomic_write_unit_max is derived from MAXIMUM ATOMIC TRANSFER LENGTH. In the scenario where MAXIMUM ATOMIC TRANSFER LENGTH is zero and boundary limits are non-zero, use MAXIMUM ATOMIC BOUNDARY SIZE for atomic_write_unit_max. New flag scsi_disk.use_atomic_write_boundary is set for this scenario. - atomic_write_boundary_bytes is set to zero always SCSI also supports a WRITE ATOMIC (32) command, which is for type 2 protection enabled. This is not going to be supported now, so check for T10_PI_TYPE2_PROTECTION when setting any request_queue limits. To handle an atomic write request, add support for WRITE ATOMIC (16) command in handler sd_setup_atomic_cmnd(). Flag use_atomic_write_boundary is checked here for encoding ATOMIC BOUNDARY field. Trace info is also added for WRITE_ATOMIC_16 command. Reviewed-by: Martin K. Petersen Signed-off-by: John Garry Acked-by: Darrick J. Wong Reviewed-by: Darrick J. Wong Link: https://lore.kernel.org/r/20240620125359.2684798-9-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- drivers/scsi/scsi_trace.c | 22 +++++++++ drivers/scsi/sd.c | 93 ++++++++++++++++++++++++++++++++++++- drivers/scsi/sd.h | 8 ++++ include/scsi/scsi_proto.h | 1 + include/trace/events/scsi.h | 1 + 5 files changed, 124 insertions(+), 1 deletion(-) diff --git a/drivers/scsi/scsi_trace.c b/drivers/scsi/scsi_trace.c index 41a9500759130f..3e47c4472a80e7 100644 --- a/drivers/scsi/scsi_trace.c +++ b/drivers/scsi/scsi_trace.c @@ -325,6 +325,26 @@ scsi_trace_zbc_out(struct trace_seq *p, unsigned char *cdb, int len) return ret; } +static const char * +scsi_trace_atomic_write16_out(struct trace_seq *p, unsigned char *cdb, int len) +{ + const char *ret = trace_seq_buffer_ptr(p); + unsigned int boundary_size; + unsigned int nr_blocks; + sector_t lba; + + lba = get_unaligned_be64(&cdb[2]); + boundary_size = get_unaligned_be16(&cdb[10]); + nr_blocks = get_unaligned_be16(&cdb[12]); + + trace_seq_printf(p, "lba=%llu txlen=%u boundary_size=%u", + lba, nr_blocks, boundary_size); + + trace_seq_putc(p, 0); + + return ret; +} + static const char * scsi_trace_varlen(struct trace_seq *p, unsigned char *cdb, int len) { @@ -385,6 +405,8 @@ scsi_trace_parse_cdb(struct trace_seq *p, unsigned char *cdb, int len) return scsi_trace_zbc_in(p, cdb, len); case ZBC_OUT: return scsi_trace_zbc_out(p, cdb, len); + case WRITE_ATOMIC_16: + return scsi_trace_atomic_write16_out(p, cdb, len); default: return scsi_trace_misc(p, cdb, len); } diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index a27f1c7f1b61d5..525f48c97f5e15 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -939,6 +939,64 @@ static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd) return scsi_alloc_sgtables(cmd); } +static void sd_config_atomic(struct scsi_disk *sdkp, struct queue_limits *lim) +{ + unsigned int logical_block_size = sdkp->device->sector_size, + physical_block_size_sectors, max_atomic, unit_min, unit_max; + + if ((!sdkp->max_atomic && !sdkp->max_atomic_with_boundary) || + sdkp->protection_type == T10_PI_TYPE2_PROTECTION) + return; + + physical_block_size_sectors = sdkp->physical_block_size / + sdkp->device->sector_size; + + unit_min = rounddown_pow_of_two(sdkp->atomic_granularity ? + sdkp->atomic_granularity : + physical_block_size_sectors); + + /* + * Only use atomic boundary when we have the odd scenario of + * sdkp->max_atomic == 0, which the spec does permit. + */ + if (sdkp->max_atomic) { + max_atomic = sdkp->max_atomic; + unit_max = rounddown_pow_of_two(sdkp->max_atomic); + sdkp->use_atomic_write_boundary = 0; + } else { + max_atomic = sdkp->max_atomic_with_boundary; + unit_max = rounddown_pow_of_two(sdkp->max_atomic_boundary); + sdkp->use_atomic_write_boundary = 1; + } + + /* + * Ensure compliance with granularity and alignment. For now, keep it + * simple and just don't support atomic writes for values mismatched + * with max_{boundary}atomic, physical block size, and + * atomic_granularity itself. + * + * We're really being distrustful by checking unit_max also... + */ + if (sdkp->atomic_granularity > 1) { + if (unit_min > 1 && unit_min % sdkp->atomic_granularity) + return; + if (unit_max > 1 && unit_max % sdkp->atomic_granularity) + return; + } + + if (sdkp->atomic_alignment > 1) { + if (unit_min > 1 && unit_min % sdkp->atomic_alignment) + return; + if (unit_max > 1 && unit_max % sdkp->atomic_alignment) + return; + } + + lim->atomic_write_hw_max = max_atomic * logical_block_size; + lim->atomic_write_hw_boundary = 0; + lim->atomic_write_hw_unit_min = unit_min * logical_block_size; + lim->atomic_write_hw_unit_max = unit_max * logical_block_size; +} + static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd, bool unmap) { @@ -1237,6 +1295,26 @@ static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd) return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1; } +static blk_status_t sd_setup_atomic_cmnd(struct scsi_cmnd *cmd, + sector_t lba, unsigned int nr_blocks, + bool boundary, unsigned char flags) +{ + cmd->cmd_len = 16; + cmd->cmnd[0] = WRITE_ATOMIC_16; + cmd->cmnd[1] = flags; + put_unaligned_be64(lba, &cmd->cmnd[2]); + put_unaligned_be16(nr_blocks, &cmd->cmnd[12]); + if (boundary) + put_unaligned_be16(nr_blocks, &cmd->cmnd[10]); + else + put_unaligned_be16(0, &cmd->cmnd[10]); + put_unaligned_be16(nr_blocks, &cmd->cmnd[12]); + cmd->cmnd[14] = 0; + cmd->cmnd[15] = 0; + + return BLK_STS_OK; +} + static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd) { struct request *rq = scsi_cmd_to_rq(cmd); @@ -1302,6 +1380,10 @@ static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd) if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) { ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks, protect | fua, dld); + } else if (rq->cmd_flags & REQ_ATOMIC && write) { + ret = sd_setup_atomic_cmnd(cmd, lba, nr_blocks, + sdkp->use_atomic_write_boundary, + protect | fua); } else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) { ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks, protect | fua, dld); @@ -3264,7 +3346,7 @@ static void sd_read_block_limits(struct scsi_disk *sdkp, sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]); if (!sdkp->lbpme) - goto out; + goto config_atomic; lba_count = get_unaligned_be32(&vpd->data[20]); desc_count = get_unaligned_be32(&vpd->data[24]); @@ -3279,6 +3361,15 @@ static void sd_read_block_limits(struct scsi_disk *sdkp, get_unaligned_be32(&vpd->data[32]) & ~(1 << 31); sd_config_discard(sdkp, lim, sd_discard_mode(sdkp)); + +config_atomic: + sdkp->max_atomic = get_unaligned_be32(&vpd->data[44]); + sdkp->atomic_alignment = get_unaligned_be32(&vpd->data[48]); + sdkp->atomic_granularity = get_unaligned_be32(&vpd->data[52]); + sdkp->max_atomic_with_boundary = get_unaligned_be32(&vpd->data[56]); + sdkp->max_atomic_boundary = get_unaligned_be32(&vpd->data[60]); + + sd_config_atomic(sdkp, lim); } out: diff --git a/drivers/scsi/sd.h b/drivers/scsi/sd.h index 7603b3c67b233f..36382eca941ce2 100644 --- a/drivers/scsi/sd.h +++ b/drivers/scsi/sd.h @@ -115,6 +115,13 @@ struct scsi_disk { u32 max_unmap_blocks; u32 unmap_granularity; u32 unmap_alignment; + + u32 max_atomic; + u32 atomic_alignment; + u32 atomic_granularity; + u32 max_atomic_with_boundary; + u32 max_atomic_boundary; + u32 index; unsigned int physical_block_size; unsigned int max_medium_access_timeouts; @@ -148,6 +155,7 @@ struct scsi_disk { unsigned security : 1; unsigned ignore_medium_access_errors : 1; unsigned rscs : 1; /* reduced stream control support */ + unsigned use_atomic_write_boundary : 1; }; #define to_scsi_disk(obj) container_of(obj, struct scsi_disk, disk_dev) diff --git a/include/scsi/scsi_proto.h b/include/scsi/scsi_proto.h index 843106e1109f43..70e1262b2e202e 100644 --- a/include/scsi/scsi_proto.h +++ b/include/scsi/scsi_proto.h @@ -120,6 +120,7 @@ #define WRITE_SAME_16 0x93 #define ZBC_OUT 0x94 #define ZBC_IN 0x95 +#define WRITE_ATOMIC_16 0x9c #define SERVICE_ACTION_BIDIRECTIONAL 0x9d #define SERVICE_ACTION_IN_16 0x9e #define SERVICE_ACTION_OUT_16 0x9f diff --git a/include/trace/events/scsi.h b/include/trace/events/scsi.h index 8e2d9b1b0e77ab..05f1945ed204ec 100644 --- a/include/trace/events/scsi.h +++ b/include/trace/events/scsi.h @@ -102,6 +102,7 @@ scsi_opcode_name(WRITE_32), \ scsi_opcode_name(WRITE_SAME_32), \ scsi_opcode_name(ATA_16), \ + scsi_opcode_name(WRITE_ATOMIC_16), \ scsi_opcode_name(ATA_12)) #define scsi_hostbyte_name(result) { result, #result } From 84f3a3c01d70efba736bc42155cf32722067b327 Mon Sep 17 00:00:00 2001 From: John Garry Date: Thu, 20 Jun 2024 12:53:58 +0000 Subject: [PATCH 123/209] scsi: scsi_debug: Atomic write support Add initial support for atomic writes. As is standard method, feed device properties via modules param, those being: - atomic_max_size_blks - atomic_alignment_blks - atomic_granularity_blks - atomic_max_size_with_boundary_blks - atomic_max_boundary_blks These just match sbc4r22 section 6.6.4 - Block limits VPD page. We just support ATOMIC WRITE (16). The major change in the driver is how we lock the device for RW accesses. Currently the driver uses a per-device lock for accessing device metadata and "media" data (calls to do_device_access()) atomically for the duration of the whole read/write command. This should not suit verifying atomic writes. Reason being that currently all reads/writes are atomic, so using atomic writes does not prove anything. Change device access model to basis that regular writes only atomic on a per-sector basis, while reads and atomic writes are fully atomic. As mentioned, since accessing metadata and device media is atomic, continue to have regular writes involving metadata - like discard or PI - as atomic. We can improve this later. Currently we only support model where overlapping going reads or writes wait for current access to complete before commencing an atomic write. This is described in 4.29.3.2 section of the SBC. However, we simplify, things and wait for all accesses to complete (when issuing an atomic write). Reviewed-by: Martin K. Petersen Signed-off-by: John Garry Acked-by: Darrick J. Wong Reviewed-by: Darrick J. Wong Reviewed-by: Luis Chamberlain Link: https://lore.kernel.org/r/20240620125359.2684798-10-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- drivers/scsi/scsi_debug.c | 588 +++++++++++++++++++++++++++++--------- 1 file changed, 454 insertions(+), 134 deletions(-) diff --git a/drivers/scsi/scsi_debug.c b/drivers/scsi/scsi_debug.c index acf0592d63dae4..6b8397b3a55f59 100644 --- a/drivers/scsi/scsi_debug.c +++ b/drivers/scsi/scsi_debug.c @@ -69,6 +69,8 @@ static const char *sdebug_version_date = "20210520"; /* Additional Sense Code (ASC) */ #define NO_ADDITIONAL_SENSE 0x0 +#define OVERLAP_ATOMIC_COMMAND_ASC 0x0 +#define OVERLAP_ATOMIC_COMMAND_ASCQ 0x23 #define LOGICAL_UNIT_NOT_READY 0x4 #define LOGICAL_UNIT_COMMUNICATION_FAILURE 0x8 #define UNRECOVERED_READ_ERR 0x11 @@ -103,6 +105,7 @@ static const char *sdebug_version_date = "20210520"; #define READ_BOUNDARY_ASCQ 0x7 #define ATTEMPT_ACCESS_GAP 0x9 #define INSUFF_ZONE_ASCQ 0xe +/* see drivers/scsi/sense_codes.h */ /* Additional Sense Code Qualifier (ASCQ) */ #define ACK_NAK_TO 0x3 @@ -152,6 +155,12 @@ static const char *sdebug_version_date = "20210520"; #define DEF_VIRTUAL_GB 0 #define DEF_VPD_USE_HOSTNO 1 #define DEF_WRITESAME_LENGTH 0xFFFF +#define DEF_ATOMIC_WR 0 +#define DEF_ATOMIC_WR_MAX_LENGTH 8192 +#define DEF_ATOMIC_WR_ALIGN 2 +#define DEF_ATOMIC_WR_GRAN 2 +#define DEF_ATOMIC_WR_MAX_LENGTH_BNDRY (DEF_ATOMIC_WR_MAX_LENGTH) +#define DEF_ATOMIC_WR_MAX_BNDRY 128 #define DEF_STRICT 0 #define DEF_STATISTICS false #define DEF_SUBMIT_QUEUES 1 @@ -374,7 +383,9 @@ struct sdebug_host_info { /* There is an xarray of pointers to this struct's objects, one per host */ struct sdeb_store_info { - rwlock_t macc_lck; /* for atomic media access on this store */ + rwlock_t macc_data_lck; /* for media data access on this store */ + rwlock_t macc_meta_lck; /* for atomic media meta access on this store */ + rwlock_t macc_sector_lck; /* per-sector media data access on this store */ u8 *storep; /* user data storage (ram) */ struct t10_pi_tuple *dif_storep; /* protection info */ void *map_storep; /* provisioning map */ @@ -398,12 +409,20 @@ struct sdebug_defer { enum sdeb_defer_type defer_t; }; +struct sdebug_device_access_info { + bool atomic_write; + u64 lba; + u32 num; + struct scsi_cmnd *self; +}; + struct sdebug_queued_cmd { /* corresponding bit set in in_use_bm[] in owning struct sdebug_queue * instance indicates this slot is in use. */ struct sdebug_defer sd_dp; struct scsi_cmnd *scmd; + struct sdebug_device_access_info *i; }; struct sdebug_scsi_cmd { @@ -463,7 +482,8 @@ enum sdeb_opcode_index { SDEB_I_PRE_FETCH = 29, /* 10, 16 */ SDEB_I_ZONE_OUT = 30, /* 0x94+SA; includes no data xfer */ SDEB_I_ZONE_IN = 31, /* 0x95+SA; all have data-in */ - SDEB_I_LAST_ELEM_P1 = 32, /* keep this last (previous + 1) */ + SDEB_I_ATOMIC_WRITE_16 = 32, + SDEB_I_LAST_ELEM_P1 = 33, /* keep this last (previous + 1) */ }; @@ -497,7 +517,8 @@ static const unsigned char opcode_ind_arr[256] = { 0, 0, 0, SDEB_I_VERIFY, SDEB_I_PRE_FETCH, SDEB_I_SYNC_CACHE, 0, SDEB_I_WRITE_SAME, SDEB_I_ZONE_OUT, SDEB_I_ZONE_IN, 0, 0, - 0, 0, 0, 0, 0, 0, SDEB_I_SERV_ACT_IN_16, SDEB_I_SERV_ACT_OUT_16, + 0, 0, 0, 0, + SDEB_I_ATOMIC_WRITE_16, 0, SDEB_I_SERV_ACT_IN_16, SDEB_I_SERV_ACT_OUT_16, /* 0xa0; 0xa0->0xbf: 12 byte cdbs */ SDEB_I_REPORT_LUNS, SDEB_I_ATA_PT, 0, SDEB_I_MAINT_IN, SDEB_I_MAINT_OUT, 0, 0, 0, @@ -547,6 +568,7 @@ static int resp_write_buffer(struct scsi_cmnd *, struct sdebug_dev_info *); static int resp_sync_cache(struct scsi_cmnd *, struct sdebug_dev_info *); static int resp_pre_fetch(struct scsi_cmnd *, struct sdebug_dev_info *); static int resp_report_zones(struct scsi_cmnd *, struct sdebug_dev_info *); +static int resp_atomic_write(struct scsi_cmnd *, struct sdebug_dev_info *); static int resp_open_zone(struct scsi_cmnd *, struct sdebug_dev_info *); static int resp_close_zone(struct scsi_cmnd *, struct sdebug_dev_info *); static int resp_finish_zone(struct scsi_cmnd *, struct sdebug_dev_info *); @@ -788,6 +810,11 @@ static const struct opcode_info_t opcode_info_arr[SDEB_I_LAST_ELEM_P1 + 1] = { resp_report_zones, zone_in_iarr, /* ZONE_IN(16), REPORT ZONES) */ {16, 0x0 /* SA */, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xbf, 0xc7} }, +/* 31 */ + {0, 0x0, 0x0, F_D_OUT | FF_MEDIA_IO, + resp_atomic_write, NULL, /* ATOMIC WRITE 16 */ + {16, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff} }, /* sentinel */ {0xff, 0, 0, 0, NULL, NULL, /* terminating element */ {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0} }, @@ -835,6 +862,13 @@ static unsigned int sdebug_unmap_granularity = DEF_UNMAP_GRANULARITY; static unsigned int sdebug_unmap_max_blocks = DEF_UNMAP_MAX_BLOCKS; static unsigned int sdebug_unmap_max_desc = DEF_UNMAP_MAX_DESC; static unsigned int sdebug_write_same_length = DEF_WRITESAME_LENGTH; +static unsigned int sdebug_atomic_wr = DEF_ATOMIC_WR; +static unsigned int sdebug_atomic_wr_max_length = DEF_ATOMIC_WR_MAX_LENGTH; +static unsigned int sdebug_atomic_wr_align = DEF_ATOMIC_WR_ALIGN; +static unsigned int sdebug_atomic_wr_gran = DEF_ATOMIC_WR_GRAN; +static unsigned int sdebug_atomic_wr_max_length_bndry = + DEF_ATOMIC_WR_MAX_LENGTH_BNDRY; +static unsigned int sdebug_atomic_wr_max_bndry = DEF_ATOMIC_WR_MAX_BNDRY; static int sdebug_uuid_ctl = DEF_UUID_CTL; static bool sdebug_random = DEF_RANDOM; static bool sdebug_per_host_store = DEF_PER_HOST_STORE; @@ -1188,6 +1222,11 @@ static inline bool scsi_debug_lbp(void) (sdebug_lbpu || sdebug_lbpws || sdebug_lbpws10); } +static inline bool scsi_debug_atomic_write(void) +{ + return sdebug_fake_rw == 0 && sdebug_atomic_wr; +} + static void *lba2fake_store(struct sdeb_store_info *sip, unsigned long long lba) { @@ -1815,6 +1854,14 @@ static int inquiry_vpd_b0(unsigned char *arr) /* Maximum WRITE SAME Length */ put_unaligned_be64(sdebug_write_same_length, &arr[32]); + if (sdebug_atomic_wr) { + put_unaligned_be32(sdebug_atomic_wr_max_length, &arr[40]); + put_unaligned_be32(sdebug_atomic_wr_align, &arr[44]); + put_unaligned_be32(sdebug_atomic_wr_gran, &arr[48]); + put_unaligned_be32(sdebug_atomic_wr_max_length_bndry, &arr[52]); + put_unaligned_be32(sdebug_atomic_wr_max_bndry, &arr[56]); + } + return 0x3c; /* Mandatory page length for Logical Block Provisioning */ } @@ -3377,16 +3424,238 @@ static inline struct sdeb_store_info *devip2sip(struct sdebug_dev_info *devip, return xa_load(per_store_ap, devip->sdbg_host->si_idx); } +static inline void +sdeb_read_lock(rwlock_t *lock) +{ + if (sdebug_no_rwlock) + __acquire(lock); + else + read_lock(lock); +} + +static inline void +sdeb_read_unlock(rwlock_t *lock) +{ + if (sdebug_no_rwlock) + __release(lock); + else + read_unlock(lock); +} + +static inline void +sdeb_write_lock(rwlock_t *lock) +{ + if (sdebug_no_rwlock) + __acquire(lock); + else + write_lock(lock); +} + +static inline void +sdeb_write_unlock(rwlock_t *lock) +{ + if (sdebug_no_rwlock) + __release(lock); + else + write_unlock(lock); +} + +static inline void +sdeb_data_read_lock(struct sdeb_store_info *sip) +{ + BUG_ON(!sip); + + sdeb_read_lock(&sip->macc_data_lck); +} + +static inline void +sdeb_data_read_unlock(struct sdeb_store_info *sip) +{ + BUG_ON(!sip); + + sdeb_read_unlock(&sip->macc_data_lck); +} + +static inline void +sdeb_data_write_lock(struct sdeb_store_info *sip) +{ + BUG_ON(!sip); + + sdeb_write_lock(&sip->macc_data_lck); +} + +static inline void +sdeb_data_write_unlock(struct sdeb_store_info *sip) +{ + BUG_ON(!sip); + + sdeb_write_unlock(&sip->macc_data_lck); +} + +static inline void +sdeb_data_sector_read_lock(struct sdeb_store_info *sip) +{ + BUG_ON(!sip); + + sdeb_read_lock(&sip->macc_sector_lck); +} + +static inline void +sdeb_data_sector_read_unlock(struct sdeb_store_info *sip) +{ + BUG_ON(!sip); + + sdeb_read_unlock(&sip->macc_sector_lck); +} + +static inline void +sdeb_data_sector_write_lock(struct sdeb_store_info *sip) +{ + BUG_ON(!sip); + + sdeb_write_lock(&sip->macc_sector_lck); +} + +static inline void +sdeb_data_sector_write_unlock(struct sdeb_store_info *sip) +{ + BUG_ON(!sip); + + sdeb_write_unlock(&sip->macc_sector_lck); +} + +/* + * Atomic locking: + * We simplify the atomic model to allow only 1x atomic write and many non- + * atomic reads or writes for all LBAs. + + * A RW lock has a similar bahaviour: + * Only 1x writer and many readers. + + * So use a RW lock for per-device read and write locking: + * An atomic access grabs the lock as a writer and non-atomic grabs the lock + * as a reader. + */ + +static inline void +sdeb_data_lock(struct sdeb_store_info *sip, bool atomic) +{ + if (atomic) + sdeb_data_write_lock(sip); + else + sdeb_data_read_lock(sip); +} + +static inline void +sdeb_data_unlock(struct sdeb_store_info *sip, bool atomic) +{ + if (atomic) + sdeb_data_write_unlock(sip); + else + sdeb_data_read_unlock(sip); +} + +/* Allow many reads but only 1x write per sector */ +static inline void +sdeb_data_sector_lock(struct sdeb_store_info *sip, bool do_write) +{ + if (do_write) + sdeb_data_sector_write_lock(sip); + else + sdeb_data_sector_read_lock(sip); +} + +static inline void +sdeb_data_sector_unlock(struct sdeb_store_info *sip, bool do_write) +{ + if (do_write) + sdeb_data_sector_write_unlock(sip); + else + sdeb_data_sector_read_unlock(sip); +} + +static inline void +sdeb_meta_read_lock(struct sdeb_store_info *sip) +{ + if (sdebug_no_rwlock) { + if (sip) + __acquire(&sip->macc_meta_lck); + else + __acquire(&sdeb_fake_rw_lck); + } else { + if (sip) + read_lock(&sip->macc_meta_lck); + else + read_lock(&sdeb_fake_rw_lck); + } +} + +static inline void +sdeb_meta_read_unlock(struct sdeb_store_info *sip) +{ + if (sdebug_no_rwlock) { + if (sip) + __release(&sip->macc_meta_lck); + else + __release(&sdeb_fake_rw_lck); + } else { + if (sip) + read_unlock(&sip->macc_meta_lck); + else + read_unlock(&sdeb_fake_rw_lck); + } +} + +static inline void +sdeb_meta_write_lock(struct sdeb_store_info *sip) +{ + if (sdebug_no_rwlock) { + if (sip) + __acquire(&sip->macc_meta_lck); + else + __acquire(&sdeb_fake_rw_lck); + } else { + if (sip) + write_lock(&sip->macc_meta_lck); + else + write_lock(&sdeb_fake_rw_lck); + } +} + +static inline void +sdeb_meta_write_unlock(struct sdeb_store_info *sip) +{ + if (sdebug_no_rwlock) { + if (sip) + __release(&sip->macc_meta_lck); + else + __release(&sdeb_fake_rw_lck); + } else { + if (sip) + write_unlock(&sip->macc_meta_lck); + else + write_unlock(&sdeb_fake_rw_lck); + } +} + /* Returns number of bytes copied or -1 if error. */ static int do_device_access(struct sdeb_store_info *sip, struct scsi_cmnd *scp, - u32 sg_skip, u64 lba, u32 num, bool do_write, - u8 group_number) + u32 sg_skip, u64 lba, u32 num, u8 group_number, + bool do_write, bool atomic) { int ret; - u64 block, rest = 0; + u64 block; enum dma_data_direction dir; struct scsi_data_buffer *sdb = &scp->sdb; u8 *fsp; + int i; + + /* + * Even though reads are inherently atomic (in this driver), we expect + * the atomic flag only for writes. + */ + if (!do_write && atomic) + return -1; if (do_write) { dir = DMA_TO_DEVICE; @@ -3406,21 +3675,26 @@ static int do_device_access(struct sdeb_store_info *sip, struct scsi_cmnd *scp, fsp = sip->storep; block = do_div(lba, sdebug_store_sectors); - if (block + num > sdebug_store_sectors) - rest = block + num - sdebug_store_sectors; - ret = sg_copy_buffer(sdb->table.sgl, sdb->table.nents, + /* Only allow 1x atomic write or multiple non-atomic writes at any given time */ + sdeb_data_lock(sip, atomic); + for (i = 0; i < num; i++) { + /* We shouldn't need to lock for atomic writes, but do it anyway */ + sdeb_data_sector_lock(sip, do_write); + ret = sg_copy_buffer(sdb->table.sgl, sdb->table.nents, fsp + (block * sdebug_sector_size), - (num - rest) * sdebug_sector_size, sg_skip, do_write); - if (ret != (num - rest) * sdebug_sector_size) - return ret; - - if (rest) { - ret += sg_copy_buffer(sdb->table.sgl, sdb->table.nents, - fsp, rest * sdebug_sector_size, - sg_skip + ((num - rest) * sdebug_sector_size), - do_write); + sdebug_sector_size, sg_skip, do_write); + sdeb_data_sector_unlock(sip, do_write); + if (ret != sdebug_sector_size) { + ret += (i * sdebug_sector_size); + break; + } + sg_skip += sdebug_sector_size; + if (++block >= sdebug_store_sectors) + block = 0; } + ret = num * sdebug_sector_size; + sdeb_data_unlock(sip, atomic); return ret; } @@ -3596,70 +3870,6 @@ static int prot_verify_read(struct scsi_cmnd *scp, sector_t start_sec, return ret; } -static inline void -sdeb_read_lock(struct sdeb_store_info *sip) -{ - if (sdebug_no_rwlock) { - if (sip) - __acquire(&sip->macc_lck); - else - __acquire(&sdeb_fake_rw_lck); - } else { - if (sip) - read_lock(&sip->macc_lck); - else - read_lock(&sdeb_fake_rw_lck); - } -} - -static inline void -sdeb_read_unlock(struct sdeb_store_info *sip) -{ - if (sdebug_no_rwlock) { - if (sip) - __release(&sip->macc_lck); - else - __release(&sdeb_fake_rw_lck); - } else { - if (sip) - read_unlock(&sip->macc_lck); - else - read_unlock(&sdeb_fake_rw_lck); - } -} - -static inline void -sdeb_write_lock(struct sdeb_store_info *sip) -{ - if (sdebug_no_rwlock) { - if (sip) - __acquire(&sip->macc_lck); - else - __acquire(&sdeb_fake_rw_lck); - } else { - if (sip) - write_lock(&sip->macc_lck); - else - write_lock(&sdeb_fake_rw_lck); - } -} - -static inline void -sdeb_write_unlock(struct sdeb_store_info *sip) -{ - if (sdebug_no_rwlock) { - if (sip) - __release(&sip->macc_lck); - else - __release(&sdeb_fake_rw_lck); - } else { - if (sip) - write_unlock(&sip->macc_lck); - else - write_unlock(&sdeb_fake_rw_lck); - } -} - static int resp_read_dt0(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) { bool check_prot; @@ -3669,6 +3879,7 @@ static int resp_read_dt0(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) u64 lba; struct sdeb_store_info *sip = devip2sip(devip, true); u8 *cmd = scp->cmnd; + bool meta_data_locked = false; switch (cmd[0]) { case READ_16: @@ -3727,6 +3938,10 @@ static int resp_read_dt0(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) atomic_set(&sdeb_inject_pending, 0); } + /* + * When checking device access params, for reads we only check data + * versus what is set at init time, so no need to lock. + */ ret = check_device_access_params(scp, lba, num, false); if (ret) return ret; @@ -3746,29 +3961,33 @@ static int resp_read_dt0(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) return check_condition_result; } - sdeb_read_lock(sip); + if (sdebug_dev_is_zoned(devip) || + (sdebug_dix && scsi_prot_sg_count(scp))) { + sdeb_meta_read_lock(sip); + meta_data_locked = true; + } /* DIX + T10 DIF */ if (unlikely(sdebug_dix && scsi_prot_sg_count(scp))) { switch (prot_verify_read(scp, lba, num, ei_lba)) { case 1: /* Guard tag error */ if (cmd[1] >> 5 != 3) { /* RDPROTECT != 3 */ - sdeb_read_unlock(sip); + sdeb_meta_read_unlock(sip); mk_sense_buffer(scp, ABORTED_COMMAND, 0x10, 1); return check_condition_result; } else if (scp->prot_flags & SCSI_PROT_GUARD_CHECK) { - sdeb_read_unlock(sip); + sdeb_meta_read_unlock(sip); mk_sense_buffer(scp, ILLEGAL_REQUEST, 0x10, 1); return illegal_condition_result; } break; case 3: /* Reference tag error */ if (cmd[1] >> 5 != 3) { /* RDPROTECT != 3 */ - sdeb_read_unlock(sip); + sdeb_meta_read_unlock(sip); mk_sense_buffer(scp, ABORTED_COMMAND, 0x10, 3); return check_condition_result; } else if (scp->prot_flags & SCSI_PROT_REF_CHECK) { - sdeb_read_unlock(sip); + sdeb_meta_read_unlock(sip); mk_sense_buffer(scp, ILLEGAL_REQUEST, 0x10, 3); return illegal_condition_result; } @@ -3776,8 +3995,9 @@ static int resp_read_dt0(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) } } - ret = do_device_access(sip, scp, 0, lba, num, false, 0); - sdeb_read_unlock(sip); + ret = do_device_access(sip, scp, 0, lba, num, 0, false, false); + if (meta_data_locked) + sdeb_meta_read_unlock(sip); if (unlikely(ret == -1)) return DID_ERROR << 16; @@ -3967,6 +4187,7 @@ static int resp_write_dt0(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) u64 lba; struct sdeb_store_info *sip = devip2sip(devip, true); u8 *cmd = scp->cmnd; + bool meta_data_locked = false; switch (cmd[0]) { case WRITE_16: @@ -4025,10 +4246,17 @@ static int resp_write_dt0(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) "to DIF device\n"); } - sdeb_write_lock(sip); + if (sdebug_dev_is_zoned(devip) || + (sdebug_dix && scsi_prot_sg_count(scp)) || + scsi_debug_lbp()) { + sdeb_meta_write_lock(sip); + meta_data_locked = true; + } + ret = check_device_access_params(scp, lba, num, true); if (ret) { - sdeb_write_unlock(sip); + if (meta_data_locked) + sdeb_meta_write_unlock(sip); return ret; } @@ -4037,22 +4265,22 @@ static int resp_write_dt0(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) switch (prot_verify_write(scp, lba, num, ei_lba)) { case 1: /* Guard tag error */ if (scp->prot_flags & SCSI_PROT_GUARD_CHECK) { - sdeb_write_unlock(sip); + sdeb_meta_write_unlock(sip); mk_sense_buffer(scp, ILLEGAL_REQUEST, 0x10, 1); return illegal_condition_result; } else if (scp->cmnd[1] >> 5 != 3) { /* WRPROTECT != 3 */ - sdeb_write_unlock(sip); + sdeb_meta_write_unlock(sip); mk_sense_buffer(scp, ABORTED_COMMAND, 0x10, 1); return check_condition_result; } break; case 3: /* Reference tag error */ if (scp->prot_flags & SCSI_PROT_REF_CHECK) { - sdeb_write_unlock(sip); + sdeb_meta_write_unlock(sip); mk_sense_buffer(scp, ILLEGAL_REQUEST, 0x10, 3); return illegal_condition_result; } else if (scp->cmnd[1] >> 5 != 3) { /* WRPROTECT != 3 */ - sdeb_write_unlock(sip); + sdeb_meta_write_unlock(sip); mk_sense_buffer(scp, ABORTED_COMMAND, 0x10, 3); return check_condition_result; } @@ -4060,13 +4288,16 @@ static int resp_write_dt0(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) } } - ret = do_device_access(sip, scp, 0, lba, num, true, group); + ret = do_device_access(sip, scp, 0, lba, num, group, true, false); if (unlikely(scsi_debug_lbp())) map_region(sip, lba, num); + /* If ZBC zone then bump its write pointer */ if (sdebug_dev_is_zoned(devip)) zbc_inc_wp(devip, lba, num); - sdeb_write_unlock(sip); + if (meta_data_locked) + sdeb_meta_write_unlock(sip); + if (unlikely(-1 == ret)) return DID_ERROR << 16; else if (unlikely(sdebug_verbose && @@ -4176,7 +4407,8 @@ static int resp_write_scat(struct scsi_cmnd *scp, goto err_out; } - sdeb_write_lock(sip); + /* Just keep it simple and always lock for now */ + sdeb_meta_write_lock(sip); sg_off = lbdof_blen; /* Spec says Buffer xfer Length field in number of LBs in dout */ cum_lb = 0; @@ -4219,7 +4451,11 @@ static int resp_write_scat(struct scsi_cmnd *scp, } } - ret = do_device_access(sip, scp, sg_off, lba, num, true, group); + /* + * Write ranges atomically to keep as close to pre-atomic + * writes behaviour as possible. + */ + ret = do_device_access(sip, scp, sg_off, lba, num, group, true, true); /* If ZBC zone then bump its write pointer */ if (sdebug_dev_is_zoned(devip)) zbc_inc_wp(devip, lba, num); @@ -4258,7 +4494,7 @@ static int resp_write_scat(struct scsi_cmnd *scp, } ret = 0; err_out_unlock: - sdeb_write_unlock(sip); + sdeb_meta_write_unlock(sip); err_out: kfree(lrdp); return ret; @@ -4277,14 +4513,16 @@ static int resp_write_same(struct scsi_cmnd *scp, u64 lba, u32 num, scp->device->hostdata, true); u8 *fs1p; u8 *fsp; + bool meta_data_locked = false; - sdeb_write_lock(sip); + if (sdebug_dev_is_zoned(devip) || scsi_debug_lbp()) { + sdeb_meta_write_lock(sip); + meta_data_locked = true; + } ret = check_device_access_params(scp, lba, num, true); - if (ret) { - sdeb_write_unlock(sip); - return ret; - } + if (ret) + goto out; if (unmap && scsi_debug_lbp()) { unmap_region(sip, lba, num); @@ -4295,6 +4533,7 @@ static int resp_write_same(struct scsi_cmnd *scp, u64 lba, u32 num, /* if ndob then zero 1 logical block, else fetch 1 logical block */ fsp = sip->storep; fs1p = fsp + (block * lb_size); + sdeb_data_write_lock(sip); if (ndob) { memset(fs1p, 0, lb_size); ret = 0; @@ -4302,8 +4541,8 @@ static int resp_write_same(struct scsi_cmnd *scp, u64 lba, u32 num, ret = fetch_to_dev_buffer(scp, fs1p, lb_size); if (-1 == ret) { - sdeb_write_unlock(sip); - return DID_ERROR << 16; + ret = DID_ERROR << 16; + goto out; } else if (sdebug_verbose && !ndob && (ret < lb_size)) sdev_printk(KERN_INFO, scp->device, "%s: %s: lb size=%u, IO sent=%d bytes\n", @@ -4320,10 +4559,12 @@ static int resp_write_same(struct scsi_cmnd *scp, u64 lba, u32 num, /* If ZBC zone then bump its write pointer */ if (sdebug_dev_is_zoned(devip)) zbc_inc_wp(devip, lba, num); + sdeb_data_write_unlock(sip); + ret = 0; out: - sdeb_write_unlock(sip); - - return 0; + if (meta_data_locked) + sdeb_meta_write_unlock(sip); + return ret; } static int resp_write_same_10(struct scsi_cmnd *scp, @@ -4466,25 +4707,30 @@ static int resp_comp_write(struct scsi_cmnd *scp, return check_condition_result; } - sdeb_write_lock(sip); - ret = do_dout_fetch(scp, dnum, arr); if (ret == -1) { retval = DID_ERROR << 16; - goto cleanup; + goto cleanup_free; } else if (sdebug_verbose && (ret < (dnum * lb_size))) sdev_printk(KERN_INFO, scp->device, "%s: compare_write: cdb " "indicated=%u, IO sent=%d bytes\n", my_name, dnum * lb_size, ret); + + sdeb_data_write_lock(sip); + sdeb_meta_write_lock(sip); if (!comp_write_worker(sip, lba, num, arr, false)) { mk_sense_buffer(scp, MISCOMPARE, MISCOMPARE_VERIFY_ASC, 0); retval = check_condition_result; - goto cleanup; + goto cleanup_unlock; } + + /* Cover sip->map_storep (which map_region()) sets with data lock */ if (scsi_debug_lbp()) map_region(sip, lba, num); -cleanup: - sdeb_write_unlock(sip); +cleanup_unlock: + sdeb_meta_write_unlock(sip); + sdeb_data_write_unlock(sip); +cleanup_free: kfree(arr); return retval; } @@ -4528,7 +4774,7 @@ static int resp_unmap(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) desc = (void *)&buf[8]; - sdeb_write_lock(sip); + sdeb_meta_write_lock(sip); for (i = 0 ; i < descriptors ; i++) { unsigned long long lba = get_unaligned_be64(&desc[i].lba); @@ -4544,7 +4790,7 @@ static int resp_unmap(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) ret = 0; out: - sdeb_write_unlock(sip); + sdeb_meta_write_unlock(sip); kfree(buf); return ret; @@ -4702,12 +4948,13 @@ static int resp_pre_fetch(struct scsi_cmnd *scp, rest = block + nblks - sdebug_store_sectors; /* Try to bring the PRE-FETCH range into CPU's cache */ - sdeb_read_lock(sip); + sdeb_data_read_lock(sip); prefetch_range(fsp + (sdebug_sector_size * block), (nblks - rest) * sdebug_sector_size); if (rest) prefetch_range(fsp, rest * sdebug_sector_size); - sdeb_read_unlock(sip); + + sdeb_data_read_unlock(sip); fini: if (cmd[1] & 0x2) res = SDEG_RES_IMMED_MASK; @@ -4866,7 +5113,7 @@ static int resp_verify(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) return check_condition_result; } /* Not changing store, so only need read access */ - sdeb_read_lock(sip); + sdeb_data_read_lock(sip); ret = do_dout_fetch(scp, a_num, arr); if (ret == -1) { @@ -4888,7 +5135,7 @@ static int resp_verify(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) goto cleanup; } cleanup: - sdeb_read_unlock(sip); + sdeb_data_read_unlock(sip); kfree(arr); return ret; } @@ -4934,7 +5181,7 @@ static int resp_report_zones(struct scsi_cmnd *scp, return check_condition_result; } - sdeb_read_lock(sip); + sdeb_meta_read_lock(sip); desc = arr + 64; for (lba = zs_lba; lba < sdebug_capacity; @@ -5032,11 +5279,70 @@ static int resp_report_zones(struct scsi_cmnd *scp, ret = fill_from_dev_buffer(scp, arr, min_t(u32, alloc_len, rep_len)); fini: - sdeb_read_unlock(sip); + sdeb_meta_read_unlock(sip); kfree(arr); return ret; } +static int resp_atomic_write(struct scsi_cmnd *scp, + struct sdebug_dev_info *devip) +{ + struct sdeb_store_info *sip; + u8 *cmd = scp->cmnd; + u16 boundary, len; + u64 lba, lba_tmp; + int ret; + + if (!scsi_debug_atomic_write()) { + mk_sense_invalid_opcode(scp); + return check_condition_result; + } + + sip = devip2sip(devip, true); + + lba = get_unaligned_be64(cmd + 2); + boundary = get_unaligned_be16(cmd + 10); + len = get_unaligned_be16(cmd + 12); + + lba_tmp = lba; + if (sdebug_atomic_wr_align && + do_div(lba_tmp, sdebug_atomic_wr_align)) { + /* Does not meet alignment requirement */ + mk_sense_buffer(scp, ILLEGAL_REQUEST, INVALID_FIELD_IN_CDB, 0); + return check_condition_result; + } + + if (sdebug_atomic_wr_gran && len % sdebug_atomic_wr_gran) { + /* Does not meet alignment requirement */ + mk_sense_buffer(scp, ILLEGAL_REQUEST, INVALID_FIELD_IN_CDB, 0); + return check_condition_result; + } + + if (boundary > 0) { + if (boundary > sdebug_atomic_wr_max_bndry) { + mk_sense_invalid_fld(scp, SDEB_IN_CDB, 12, -1); + return check_condition_result; + } + + if (len > sdebug_atomic_wr_max_length_bndry) { + mk_sense_invalid_fld(scp, SDEB_IN_CDB, 12, -1); + return check_condition_result; + } + } else { + if (len > sdebug_atomic_wr_max_length) { + mk_sense_invalid_fld(scp, SDEB_IN_CDB, 12, -1); + return check_condition_result; + } + } + + ret = do_device_access(sip, scp, 0, lba, len, 0, true, true); + if (unlikely(ret == -1)) + return DID_ERROR << 16; + if (unlikely(ret != len * sdebug_sector_size)) + return DID_ERROR << 16; + return 0; +} + /* Logic transplanted from tcmu-runner, file_zbc.c */ static void zbc_open_all(struct sdebug_dev_info *devip) { @@ -5063,8 +5369,7 @@ static int resp_open_zone(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) mk_sense_invalid_opcode(scp); return check_condition_result; } - - sdeb_write_lock(sip); + sdeb_meta_write_lock(sip); if (all) { /* Check if all closed zones can be open */ @@ -5113,7 +5418,7 @@ static int resp_open_zone(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) zbc_open_zone(devip, zsp, true); fini: - sdeb_write_unlock(sip); + sdeb_meta_write_unlock(sip); return res; } @@ -5140,7 +5445,7 @@ static int resp_close_zone(struct scsi_cmnd *scp, return check_condition_result; } - sdeb_write_lock(sip); + sdeb_meta_write_lock(sip); if (all) { zbc_close_all(devip); @@ -5169,7 +5474,7 @@ static int resp_close_zone(struct scsi_cmnd *scp, zbc_close_zone(devip, zsp); fini: - sdeb_write_unlock(sip); + sdeb_meta_write_unlock(sip); return res; } @@ -5212,7 +5517,7 @@ static int resp_finish_zone(struct scsi_cmnd *scp, return check_condition_result; } - sdeb_write_lock(sip); + sdeb_meta_write_lock(sip); if (all) { zbc_finish_all(devip); @@ -5241,7 +5546,7 @@ static int resp_finish_zone(struct scsi_cmnd *scp, zbc_finish_zone(devip, zsp, true); fini: - sdeb_write_unlock(sip); + sdeb_meta_write_unlock(sip); return res; } @@ -5292,7 +5597,7 @@ static int resp_rwp_zone(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) return check_condition_result; } - sdeb_write_lock(sip); + sdeb_meta_write_lock(sip); if (all) { zbc_rwp_all(devip); @@ -5320,7 +5625,7 @@ static int resp_rwp_zone(struct scsi_cmnd *scp, struct sdebug_dev_info *devip) zbc_rwp_zone(devip, zsp); fini: - sdeb_write_unlock(sip); + sdeb_meta_write_unlock(sip); return res; } @@ -6284,6 +6589,7 @@ module_param_named(lbprz, sdebug_lbprz, int, S_IRUGO); module_param_named(lbpu, sdebug_lbpu, int, S_IRUGO); module_param_named(lbpws, sdebug_lbpws, int, S_IRUGO); module_param_named(lbpws10, sdebug_lbpws10, int, S_IRUGO); +module_param_named(atomic_wr, sdebug_atomic_wr, int, S_IRUGO); module_param_named(lowest_aligned, sdebug_lowest_aligned, int, S_IRUGO); module_param_named(lun_format, sdebug_lun_am_i, int, S_IRUGO | S_IWUSR); module_param_named(max_luns, sdebug_max_luns, int, S_IRUGO | S_IWUSR); @@ -6318,6 +6624,11 @@ module_param_named(unmap_alignment, sdebug_unmap_alignment, int, S_IRUGO); module_param_named(unmap_granularity, sdebug_unmap_granularity, int, S_IRUGO); module_param_named(unmap_max_blocks, sdebug_unmap_max_blocks, int, S_IRUGO); module_param_named(unmap_max_desc, sdebug_unmap_max_desc, int, S_IRUGO); +module_param_named(atomic_wr_max_length, sdebug_atomic_wr_max_length, int, S_IRUGO); +module_param_named(atomic_wr_align, sdebug_atomic_wr_align, int, S_IRUGO); +module_param_named(atomic_wr_gran, sdebug_atomic_wr_gran, int, S_IRUGO); +module_param_named(atomic_wr_max_length_bndry, sdebug_atomic_wr_max_length_bndry, int, S_IRUGO); +module_param_named(atomic_wr_max_bndry, sdebug_atomic_wr_max_bndry, int, S_IRUGO); module_param_named(uuid_ctl, sdebug_uuid_ctl, int, S_IRUGO); module_param_named(virtual_gb, sdebug_virtual_gb, int, S_IRUGO | S_IWUSR); module_param_named(vpd_use_hostno, sdebug_vpd_use_hostno, int, @@ -6361,6 +6672,7 @@ MODULE_PARM_DESC(lbprz, MODULE_PARM_DESC(lbpu, "enable LBP, support UNMAP command (def=0)"); MODULE_PARM_DESC(lbpws, "enable LBP, support WRITE SAME(16) with UNMAP bit (def=0)"); MODULE_PARM_DESC(lbpws10, "enable LBP, support WRITE SAME(10) with UNMAP bit (def=0)"); +MODULE_PARM_DESC(atomic_write, "enable ATOMIC WRITE support, support WRITE ATOMIC(16) (def=0)"); MODULE_PARM_DESC(lowest_aligned, "lowest aligned lba (def=0)"); MODULE_PARM_DESC(lun_format, "LUN format: 0->peripheral (def); 1 --> flat address method"); MODULE_PARM_DESC(max_luns, "number of LUNs per target to simulate(def=1)"); @@ -6392,6 +6704,11 @@ MODULE_PARM_DESC(unmap_alignment, "lowest aligned thin provisioning lba (def=0)" MODULE_PARM_DESC(unmap_granularity, "thin provisioning granularity in blocks (def=1)"); MODULE_PARM_DESC(unmap_max_blocks, "max # of blocks can be unmapped in one cmd (def=0xffffffff)"); MODULE_PARM_DESC(unmap_max_desc, "max # of ranges that can be unmapped in one cmd (def=256)"); +MODULE_PARM_DESC(atomic_wr_max_length, "max # of blocks can be atomically written in one cmd (def=8192)"); +MODULE_PARM_DESC(atomic_wr_align, "minimum alignment of atomic write in blocks (def=2)"); +MODULE_PARM_DESC(atomic_wr_gran, "minimum granularity of atomic write in blocks (def=2)"); +MODULE_PARM_DESC(atomic_wr_max_length_bndry, "max # of blocks can be atomically written in one cmd with boundary set (def=8192)"); +MODULE_PARM_DESC(atomic_wr_max_bndry, "max # boundaries per atomic write (def=128)"); MODULE_PARM_DESC(uuid_ctl, "1->use uuid for lu name, 0->don't, 2->all use same (def=0)"); MODULE_PARM_DESC(virtual_gb, "virtual gigabyte (GiB) size (def=0 -> use dev_size_mb)"); @@ -7563,6 +7880,7 @@ static int __init scsi_debug_init(void) return -EINVAL; } } + xa_init_flags(per_store_ap, XA_FLAGS_ALLOC | XA_FLAGS_LOCK_IRQ); if (want_store) { idx = sdebug_add_store(); @@ -7770,7 +8088,9 @@ static int sdebug_add_store(void) map_region(sip, 0, 2); } - rwlock_init(&sip->macc_lck); + rwlock_init(&sip->macc_data_lck); + rwlock_init(&sip->macc_meta_lck); + rwlock_init(&sip->macc_sector_lck); return (int)n_idx; err: sdebug_erase_store((int)n_idx, sip); From 5f9bbea02f06110ec5cf95a3327019b3194b2d80 Mon Sep 17 00:00:00 2001 From: Alan Adamson Date: Thu, 20 Jun 2024 12:53:59 +0000 Subject: [PATCH 124/209] nvme: Atomic write support MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add support to set block layer request_queue atomic write limits. The limits will be derived from either the namespace or controller atomic parameters. NVMe atomic-related parameters are grouped into "normal" and "power-fail" (or PF) class of parameter. For atomic write support, only PF parameters are of interest. The "normal" parameters are concerned with racing reads and writes (which also applies to PF). See NVM Command Set Specification Revision 1.0d section 2.1.4 for reference. Whether to use per namespace or controller atomic parameters is decided by NSFEAT bit 1 - see Figure 97: Identify – Identify Namespace Data Structure, NVM Command Set. NVMe namespaces may define an atomic boundary, whereby no atomic guarantees are provided for a write which straddles this per-lba space boundary. The block layer merging policy is such that no merges may occur in which the resultant request would straddle such a boundary. Unlike SCSI, NVMe specifies no granularity or alignment rules, apart from atomic boundary rule. In addition, again unlike SCSI, there is no dedicated atomic write command - a write which adheres to the atomic size limit and boundary is implicitly atomic. If NSFEAT bit 1 is set, the following parameters are of interest: - NAWUPF (Namespace Atomic Write Unit Power Fail) - NABSPF (Namespace Atomic Boundary Size Power Fail) - NABO (Namespace Atomic Boundary Offset) and we set request_queue limits as follows: - atomic_write_unit_max = rounddown_pow_of_two(NAWUPF) - atomic_write_max_bytes = NAWUPF - atomic_write_boundary = NABSPF If in the unlikely scenario that NABO is non-zero, then atomic writes will not be supported at all as dealing with this adds extra complexity. This policy may change in future. In all cases, atomic_write_unit_min is set to the logical block size. If NSFEAT bit 1 is unset, the following parameter is of interest: - AWUPF (Atomic Write Unit Power Fail) and we set request_queue limits as follows: - atomic_write_unit_max = rounddown_pow_of_two(AWUPF) - atomic_write_max_bytes = AWUPF - atomic_write_boundary = 0 A new function, nvme_valid_atomic_write(), is also called from submission path to verify that a request has been submitted to the driver will actually be executed atomically. As mentioned, there is no dedicated NVMe atomic write command (which may error for a command which exceeds the controller atomic write limits). Note on NABSPF: There seems to be some vagueness in the spec as to whether NABSPF applies for NSFEAT bit 1 being unset. Figure 97 does not explicitly mention NABSPF and how it is affected by bit 1. However Figure 4 does tell to check Figure 97 for info about per-namespace parameters, which NABSPF is, so it is implied. However currently nvme_update_disk_info() does check namespace parameter NABO regardless of this bit. Signed-off-by: Alan Adamson Reviewed-by: Keith Busch Reviewed-by: Martin K. Petersen jpg: total rewrite Signed-off-by: John Garry Link: https://lore.kernel.org/r/20240620125359.2684798-11-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- drivers/nvme/host/core.c | 52 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 52 insertions(+) diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index bf410d10b12006..89ebfa89613ee6 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -927,6 +927,36 @@ static inline blk_status_t nvme_setup_write_zeroes(struct nvme_ns *ns, return BLK_STS_OK; } +/* + * NVMe does not support a dedicated command to issue an atomic write. A write + * which does adhere to the device atomic limits will silently be executed + * non-atomically. The request issuer should ensure that the write is within + * the queue atomic writes limits, but just validate this in case it is not. + */ +static bool nvme_valid_atomic_write(struct request *req) +{ + struct request_queue *q = req->q; + u32 boundary_bytes = queue_atomic_write_boundary_bytes(q); + + if (blk_rq_bytes(req) > queue_atomic_write_unit_max_bytes(q)) + return false; + + if (boundary_bytes) { + u64 mask = boundary_bytes - 1, imask = ~mask; + u64 start = blk_rq_pos(req) << SECTOR_SHIFT; + u64 end = start + blk_rq_bytes(req) - 1; + + /* If greater then must be crossing a boundary */ + if (blk_rq_bytes(req) > boundary_bytes) + return false; + + if ((start & imask) != (end & imask)) + return false; + } + + return true; +} + static inline blk_status_t nvme_setup_rw(struct nvme_ns *ns, struct request *req, struct nvme_command *cmnd, enum nvme_opcode op) @@ -942,6 +972,9 @@ static inline blk_status_t nvme_setup_rw(struct nvme_ns *ns, if (req->cmd_flags & REQ_RAHEAD) dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH; + if (req->cmd_flags & REQ_ATOMIC && !nvme_valid_atomic_write(req)) + return BLK_STS_INVAL; + cmnd->rw.opcode = op; cmnd->rw.flags = 0; cmnd->rw.nsid = cpu_to_le32(ns->head->ns_id); @@ -1920,6 +1953,23 @@ static void nvme_configure_metadata(struct nvme_ctrl *ctrl, } } + +static void nvme_update_atomic_write_disk_info(struct nvme_ns *ns, + struct nvme_id_ns *id, struct queue_limits *lim, + u32 bs, u32 atomic_bs) +{ + unsigned int boundary = 0; + + if (id->nsfeat & NVME_NS_FEAT_ATOMICS && id->nawupf) { + if (le16_to_cpu(id->nabspf)) + boundary = (le16_to_cpu(id->nabspf) + 1) * bs; + } + lim->atomic_write_hw_max = atomic_bs; + lim->atomic_write_hw_boundary = boundary; + lim->atomic_write_hw_unit_min = bs; + lim->atomic_write_hw_unit_max = rounddown_pow_of_two(atomic_bs); +} + static u32 nvme_max_drv_segments(struct nvme_ctrl *ctrl) { return ctrl->max_hw_sectors / (NVME_CTRL_PAGE_SIZE >> SECTOR_SHIFT) + 1; @@ -1966,6 +2016,8 @@ static bool nvme_update_disk_info(struct nvme_ns *ns, struct nvme_id_ns *id, atomic_bs = (1 + le16_to_cpu(id->nawupf)) * bs; else atomic_bs = (1 + ns->ctrl->subsys->awupf) * bs; + + nvme_update_atomic_write_disk_info(ns, id, lim, bs, atomic_bs); } if (id->nsfeat & NVME_NS_FEAT_IO_OPT) { From 4ac9056e4bd787f1ba2001167c5acc2b5a75ddf9 Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Fri, 21 Jun 2024 12:15:04 +0900 Subject: [PATCH 125/209] null_blk: Do not set disk->nr_zones In null_register_zoned_dev(), there is no need to set disk->nr_zones as the now uncoditional call to blk_revalidate_disk_zones() will do that. So remove the assignment using bdev_nr_zones(). Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240621031506.759397-2-dlemoal@kernel.org Signed-off-by: Jens Axboe --- drivers/block/null_blk/zoned.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/drivers/block/null_blk/zoned.c b/drivers/block/null_blk/zoned.c index b42c00f1313254..9f7151ad93cfc4 100644 --- a/drivers/block/null_blk/zoned.c +++ b/drivers/block/null_blk/zoned.c @@ -171,8 +171,6 @@ int null_register_zoned_dev(struct nullb *nullb) struct request_queue *q = nullb->q; struct gendisk *disk = nullb->disk; - disk->nr_zones = bdev_nr_zones(disk->part0); - pr_info("%s: using %s zone append\n", disk->disk_name, queue_emulates_zone_append(q) ? "emulated" : "native"); From b6cfe2287df6e26d685af8a8a96ed1bf87bdde28 Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Fri, 21 Jun 2024 12:15:05 +0900 Subject: [PATCH 126/209] block: Define bdev_nr_zones() as an inline function There is no need for bdev_nr_zones() to be an exported function calculating the number of zones of a block device. Instead, given that all callers use this helper with a fully initialized block device that has a gendisk, we can redefine this function as an inline helper in blkdev.h. Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240621031506.759397-3-dlemoal@kernel.org Signed-off-by: Jens Axboe --- block/blk-zoned.c | 18 ------------------ include/linux/blkdev.h | 6 +++++- 2 files changed, 5 insertions(+), 19 deletions(-) diff --git a/block/blk-zoned.c b/block/blk-zoned.c index 8f89705f5e1c50..601c21a224c93e 100644 --- a/block/blk-zoned.c +++ b/block/blk-zoned.c @@ -115,24 +115,6 @@ const char *blk_zone_cond_str(enum blk_zone_cond zone_cond) } EXPORT_SYMBOL_GPL(blk_zone_cond_str); -/** - * bdev_nr_zones - Get number of zones - * @bdev: Target device - * - * Return the total number of zones of a zoned block device. For a block - * device without zone capabilities, the number of zones is always 0. - */ -unsigned int bdev_nr_zones(struct block_device *bdev) -{ - sector_t zone_sectors = bdev_zone_sectors(bdev); - - if (!bdev_is_zoned(bdev)) - return 0; - return (bdev_nr_sectors(bdev) + zone_sectors - 1) >> - ilog2(zone_sectors); -} -EXPORT_SYMBOL_GPL(bdev_nr_zones); - /** * blkdev_report_zones - Get zones information * @bdev: Target block device diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 2800231046a98b..1078a7d5129541 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -673,7 +673,6 @@ static inline bool blk_queue_is_zoned(struct request_queue *q) } #ifdef CONFIG_BLK_DEV_ZONED -unsigned int bdev_nr_zones(struct block_device *bdev); static inline unsigned int disk_nr_zones(struct gendisk *disk) { @@ -687,6 +686,11 @@ static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) return sector >> ilog2(disk->queue->limits.chunk_sectors); } +static inline unsigned int bdev_nr_zones(struct block_device *bdev) +{ + return disk_nr_zones(bdev->bd_disk); +} + static inline unsigned int bdev_max_open_zones(struct block_device *bdev) { return bdev->bd_disk->queue->limits.max_open_zones; From caaf7101c01a91a882d3da2f566579dda692367d Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Fri, 21 Jun 2024 12:15:06 +0900 Subject: [PATCH 127/209] block: Cleanup block device zone helpers There is no need to conditionally define on CONFIG_BLK_DEV_ZONED the inline helper functions bdev_nr_zones(), bdev_max_open_zones(), bdev_max_active_zones() and disk_zone_no() as these function will return the correct valu in all cases (zoned device or not, including when CONFIG_BLK_DEV_ZONED is not set). Furthermore, disk_nr_zones() definition can be simplified as disk->nr_zones is always 0 for regular block devices. Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240621031506.759397-4-dlemoal@kernel.org Signed-off-by: Jens Axboe --- include/linux/blkdev.h | 44 ++++++++++++------------------------------ 1 file changed, 12 insertions(+), 32 deletions(-) diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 1078a7d5129541..e89003360c17c7 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -673,11 +673,21 @@ static inline bool blk_queue_is_zoned(struct request_queue *q) } #ifdef CONFIG_BLK_DEV_ZONED - static inline unsigned int disk_nr_zones(struct gendisk *disk) { - return blk_queue_is_zoned(disk->queue) ? disk->nr_zones : 0; + return disk->nr_zones; +} +bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs); +#else /* CONFIG_BLK_DEV_ZONED */ +static inline unsigned int disk_nr_zones(struct gendisk *disk) +{ + return 0; +} +static inline bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs) +{ + return false; } +#endif /* CONFIG_BLK_DEV_ZONED */ static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) { @@ -701,36 +711,6 @@ static inline unsigned int bdev_max_active_zones(struct block_device *bdev) return bdev->bd_disk->queue->limits.max_active_zones; } -bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs); -#else /* CONFIG_BLK_DEV_ZONED */ -static inline unsigned int bdev_nr_zones(struct block_device *bdev) -{ - return 0; -} - -static inline unsigned int disk_nr_zones(struct gendisk *disk) -{ - return 0; -} -static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) -{ - return 0; -} -static inline unsigned int bdev_max_open_zones(struct block_device *bdev) -{ - return 0; -} - -static inline unsigned int bdev_max_active_zones(struct block_device *bdev) -{ - return 0; -} -static inline bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs) -{ - return false; -} -#endif /* CONFIG_BLK_DEV_ZONED */ - static inline unsigned int blk_queue_depth(struct request_queue *q) { if (q->queue_depth) From 8324bb755a80d463ff53379e5d64991656512069 Mon Sep 17 00:00:00 2001 From: John Garry Date: Fri, 21 Jun 2024 18:30:16 +0000 Subject: [PATCH 128/209] block: Fix blk_validate_atomic_write_limits() build for arm32 For arm32, we get the following build warning: In file included from /tmp/next/build/include/linux/printk.h:10, from /tmp/next/build/include/linux/kernel.h:31, from /tmp/next/build/block/blk-settings.c:5: /tmp/next/build/block/blk-settings.c: In function 'blk_validate_atomic_write_limits': /tmp/next/build/include/asm-generic/div64.h:222:35: warning: comparison of distinct pointer types lacks a cast 222 | (void)(((typeof((n)) *)0) == ((uint64_t *)0)); \ | ^~ The divident for do_div() should be 64b, which it is not. Since we want to check 2x unsigned ints, just use % operator. This allows us to drop the chunk_sectors variable. Fixes: 9da3d1e912f3 ("block: Add core atomic write support") Reported-by: Mark Brown Closes: https://lore.kernel.org/linux-next/b765d200-4e0f-48b1-a962-7dfa1c4aef9c@kernel.dk/T/#mbf067b1edd89c7f9d7dac6e258c516199953a108 Signed-off-by: John Garry Link: https://lore.kernel.org/r/20240621183016.3092518-1-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- block/blk-settings.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index 37fe4c8f6b6ba1..ec7dbe93d5c324 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -175,7 +175,6 @@ static void blk_atomic_writes_update_limits(struct queue_limits *lim) static void blk_validate_atomic_write_limits(struct queue_limits *lim) { - unsigned int chunk_sectors = lim->chunk_sectors; unsigned int boundary_sectors; if (!lim->atomic_write_hw_max) @@ -197,7 +196,7 @@ static void blk_validate_atomic_write_limits(struct queue_limits *lim) * Devices which do not conform to these rules can be dealt * with if and when they show up. */ - if (WARN_ON_ONCE(do_div(chunk_sectors, boundary_sectors))) + if (WARN_ON_ONCE(lim->chunk_sectors % boundary_sectors)) goto unsupported; /* From 85f86c5ede7697162c54744258908e657e456f57 Mon Sep 17 00:00:00 2001 From: Jeff Johnson Date: Sat, 1 Jun 2024 23:18:16 +0100 Subject: [PATCH 129/209] cdrom: Add missing MODULE_DESCRIPTION() make allmodconfig && make W=1 C=1 reports: WARNING: modpost: missing MODULE_DESCRIPTION() in drivers/cdrom/cdrom.o Add the missing MODULE_DESCRIPTION() macro invocation. Signed-off-by: Jeff Johnson Link: https://lore.kernel.org/lkml/20240530-cdrom-v1-1-51579c5c240a@quicinc.com Reviewed-by: Phillip Potter Link: https://lore.kernel.org/lkml/ZluYQbvrJkRlhnJC@KernelVM Signed-off-by: Phillip Potter Link: https://lore.kernel.org/r/20240601221816.136977-2-phil@philpotter.co.uk Signed-off-by: Jens Axboe --- drivers/cdrom/cdrom.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/cdrom/cdrom.c b/drivers/cdrom/cdrom.c index 20c90ebb3a3f61..49e4829b726479 100644 --- a/drivers/cdrom/cdrom.c +++ b/drivers/cdrom/cdrom.c @@ -3708,4 +3708,5 @@ static void __exit cdrom_exit(void) module_init(cdrom_init); module_exit(cdrom_exit); +MODULE_DESCRIPTION("Uniform CD-ROM driver"); MODULE_LICENSE("GPL"); From 876835b128976e2e9a7d18daab58b4cba7742787 Mon Sep 17 00:00:00 2001 From: Jeff Johnson Date: Sun, 2 Jun 2024 16:46:25 -0700 Subject: [PATCH 130/209] brd: add missing MODULE_DESCRIPTION() macro make allmodconfig && make W=1 C=1 reports: modpost: missing MODULE_DESCRIPTION() in drivers/block/brd.o Add the missing invocation of the MODULE_DESCRIPTION() macro. Signed-off-by: Jeff Johnson Link: https://lore.kernel.org/r/20240602-md-block-brd-v1-1-e71338e131b6@quicinc.com Signed-off-by: Jens Axboe --- drivers/block/brd.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/block/brd.c b/drivers/block/brd.c index a300645cd9d4a5..2fd1ed1017481b 100644 --- a/drivers/block/brd.c +++ b/drivers/block/brd.c @@ -296,6 +296,7 @@ static int max_part = 1; module_param(max_part, int, 0444); MODULE_PARM_DESC(max_part, "Num Minors to reserve between devices"); +MODULE_DESCRIPTION("Ram backed block device driver"); MODULE_LICENSE("GPL"); MODULE_ALIAS_BLOCKDEV_MAJOR(RAMDISK_MAJOR); MODULE_ALIAS("rd"); From d57afd8bb7f2c4f0d86e9e9b276f7c3a7fedfc6d Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Tue, 28 May 2024 08:40:12 -0600 Subject: [PATCH 131/209] io_uring/msg_ring: tighten requirement for remote posting Currently this is gated on whether or not the target ring needs a local completion - and if so, whether or not we're running on the right task. The use case for same thread cross posting is probably a lot less relevant than remote posting. And since we're going to improve this situation anyway, just gate it on local posting and ignore what task we're currently running on. Signed-off-by: Jens Axboe --- io_uring/msg_ring.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/io_uring/msg_ring.c b/io_uring/msg_ring.c index 81c4a9d437296c..9fdb0cc19bfdd8 100644 --- a/io_uring/msg_ring.c +++ b/io_uring/msg_ring.c @@ -68,9 +68,7 @@ void io_msg_ring_cleanup(struct io_kiocb *req) static inline bool io_msg_need_remote(struct io_ring_ctx *target_ctx) { - if (!target_ctx->task_complete) - return false; - return current != target_ctx->submitter_task; + return target_ctx->task_complete; } static int io_msg_exec_remote(struct io_kiocb *req, task_work_func_t func) From c3ac76f9ca7a621428851149bc56bfca0aacaef4 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 28 Mar 2024 12:38:44 -0600 Subject: [PATCH 132/209] io_uring: add remote task_work execution helper All our task_work handling is targeted at the state in the io_kiocb itself, which is what it is being used for. However, MSG_RING rolls its own task_work handling, ignoring how that is usually done. In preparation for switching MSG_RING to be able to use the normal task_work handling, add io_req_task_work_add_remote() which allows the caller to pass in the target io_ring_ctx. Signed-off-by: Jens Axboe --- io_uring/io_uring.c | 24 ++++++++++++++++-------- io_uring/io_uring.h | 2 ++ 2 files changed, 18 insertions(+), 8 deletions(-) diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 438c44ca3abd63..85b2ce54328c3e 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -1098,9 +1098,10 @@ void tctx_task_work(struct callback_head *cb) WARN_ON_ONCE(ret); } -static inline void io_req_local_work_add(struct io_kiocb *req, unsigned flags) +static inline void io_req_local_work_add(struct io_kiocb *req, + struct io_ring_ctx *ctx, + unsigned flags) { - struct io_ring_ctx *ctx = req->ctx; unsigned nr_wait, nr_tw, nr_tw_prev; struct llist_node *head; @@ -1114,6 +1115,8 @@ static inline void io_req_local_work_add(struct io_kiocb *req, unsigned flags) if (req->flags & (REQ_F_LINK | REQ_F_HARDLINK)) flags &= ~IOU_F_TWQ_LAZY_WAKE; + guard(rcu)(); + head = READ_ONCE(ctx->work_llist.first); do { nr_tw_prev = 0; @@ -1195,13 +1198,18 @@ static void io_req_normal_work_add(struct io_kiocb *req) void __io_req_task_work_add(struct io_kiocb *req, unsigned flags) { - if (req->ctx->flags & IORING_SETUP_DEFER_TASKRUN) { - rcu_read_lock(); - io_req_local_work_add(req, flags); - rcu_read_unlock(); - } else { + if (req->ctx->flags & IORING_SETUP_DEFER_TASKRUN) + io_req_local_work_add(req, req->ctx, flags); + else io_req_normal_work_add(req); - } +} + +void io_req_task_work_add_remote(struct io_kiocb *req, struct io_ring_ctx *ctx, + unsigned flags) +{ + if (WARN_ON_ONCE(!(ctx->flags & IORING_SETUP_DEFER_TASKRUN))) + return; + io_req_local_work_add(req, ctx, flags); } static void __cold io_move_task_work_from_local(struct io_ring_ctx *ctx) diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h index cd43924eed04e1..7a8641214509ff 100644 --- a/io_uring/io_uring.h +++ b/io_uring/io_uring.h @@ -73,6 +73,8 @@ struct file *io_file_get_fixed(struct io_kiocb *req, int fd, unsigned issue_flags); void __io_req_task_work_add(struct io_kiocb *req, unsigned flags); +void io_req_task_work_add_remote(struct io_kiocb *req, struct io_ring_ctx *ctx, + unsigned flags); bool io_alloc_async_data(struct io_kiocb *req); void io_req_task_queue(struct io_kiocb *req); void io_req_task_complete(struct io_kiocb *req, struct io_tw_state *ts); From f33096a3c99c0149be49fe1e107244a7ed860ecb Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 6 Jun 2024 10:28:26 -0600 Subject: [PATCH 133/209] io_uring: add io_add_aux_cqe() helper This helper will post a CQE, and can be called from task_work where we now that the ctx is already properly locked and that deferred completions will get flushed later on. Signed-off-by: Jens Axboe --- io_uring/io_uring.c | 23 +++++++++++++++++++++-- io_uring/io_uring.h | 1 + 2 files changed, 22 insertions(+), 2 deletions(-) diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 85b2ce54328c3e..cdeb94d2a26b61 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -801,19 +801,38 @@ static bool io_fill_cqe_aux(struct io_ring_ctx *ctx, u64 user_data, s32 res, return false; } -bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags) +static bool __io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, + u32 cflags) { bool filled; - io_cq_lock(ctx); filled = io_fill_cqe_aux(ctx, user_data, res, cflags); if (!filled) filled = io_cqring_event_overflow(ctx, user_data, res, cflags, 0, 0); + return filled; +} + +bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags) +{ + bool filled; + + io_cq_lock(ctx); + filled = __io_post_aux_cqe(ctx, user_data, res, cflags); io_cq_unlock_post(ctx); return filled; } +/* + * Must be called from inline task_work so we now a flush will happen later, + * and obviously with ctx->uring_lock held (tw always has that). + */ +void io_add_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags) +{ + __io_post_aux_cqe(ctx, user_data, res, cflags); + ctx->submit_state.cq_flush = true; +} + /* * A helper for multishot requests posting additional CQEs. * Should only be used from a task_work including IO_URING_F_MULTISHOT. diff --git a/io_uring/io_uring.h b/io_uring/io_uring.h index 7a8641214509ff..e1ce908f067992 100644 --- a/io_uring/io_uring.h +++ b/io_uring/io_uring.h @@ -65,6 +65,7 @@ bool io_cqe_cache_refill(struct io_ring_ctx *ctx, bool overflow); int io_run_task_work_sig(struct io_ring_ctx *ctx); void io_req_defer_failed(struct io_kiocb *req, s32 res); bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags); +void io_add_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags); bool io_req_post_cqe(struct io_kiocb *req, s32 res, u32 cflags); void __io_commit_cqring_flush(struct io_ring_ctx *ctx); From 0617bb500bfabf8447062f1e1edde92ed2b638f1 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 28 Mar 2024 11:00:21 -0600 Subject: [PATCH 134/209] io_uring/msg_ring: improve handling of target CQE posting Use the exported helper for queueing task_work for message passing, rather than rolling our own. Note that this is only done for strict data messages for now, file descriptor passing messages still rely on the kernel task_work. It could get converted at some point if it's performance critical. This improves peak performance of message passing by about 5x in some basic testing, with 2 threads just sending messages to each other. Before this change, it was capped at around 700K/sec, with the change it's at over 4M/sec. Signed-off-by: Jens Axboe --- io_uring/msg_ring.c | 90 +++++++++++++++++++++++---------------------- 1 file changed, 47 insertions(+), 43 deletions(-) diff --git a/io_uring/msg_ring.c b/io_uring/msg_ring.c index 9fdb0cc19bfdd8..ad7d67d444612b 100644 --- a/io_uring/msg_ring.c +++ b/io_uring/msg_ring.c @@ -13,7 +13,6 @@ #include "filetable.h" #include "msg_ring.h" - /* All valid masks for MSG_RING */ #define IORING_MSG_RING_MASK (IORING_MSG_RING_CQE_SKIP | \ IORING_MSG_RING_FLAGS_PASS) @@ -71,54 +70,43 @@ static inline bool io_msg_need_remote(struct io_ring_ctx *target_ctx) return target_ctx->task_complete; } -static int io_msg_exec_remote(struct io_kiocb *req, task_work_func_t func) +static void io_msg_tw_complete(struct io_kiocb *req, struct io_tw_state *ts) { - struct io_ring_ctx *ctx = req->file->private_data; - struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg); - struct task_struct *task = READ_ONCE(ctx->submitter_task); - - if (unlikely(!task)) - return -EOWNERDEAD; + struct io_ring_ctx *ctx = req->ctx; - init_task_work(&msg->tw, func); - if (task_work_add(task, &msg->tw, TWA_SIGNAL)) - return -EOWNERDEAD; + io_add_aux_cqe(ctx, req->cqe.user_data, req->cqe.res, req->cqe.flags); + kmem_cache_free(req_cachep, req); + percpu_ref_put(&ctx->refs); +} - return IOU_ISSUE_SKIP_COMPLETE; +static void io_msg_remote_post(struct io_ring_ctx *ctx, struct io_kiocb *req, + int res, u32 cflags, u64 user_data) +{ + req->cqe.user_data = user_data; + io_req_set_res(req, res, cflags); + percpu_ref_get(&ctx->refs); + req->ctx = ctx; + req->task = READ_ONCE(ctx->submitter_task); + req->io_task_work.func = io_msg_tw_complete; + io_req_task_work_add_remote(req, ctx, IOU_F_TWQ_LAZY_WAKE); } -static void io_msg_tw_complete(struct callback_head *head) +static int io_msg_data_remote(struct io_kiocb *req) { - struct io_msg *msg = container_of(head, struct io_msg, tw); - struct io_kiocb *req = cmd_to_io_kiocb(msg); struct io_ring_ctx *target_ctx = req->file->private_data; - int ret = 0; - - if (current->flags & PF_EXITING) { - ret = -EOWNERDEAD; - } else { - u32 flags = 0; - - if (msg->flags & IORING_MSG_RING_FLAGS_PASS) - flags = msg->cqe_flags; - - /* - * If the target ring is using IOPOLL mode, then we need to be - * holding the uring_lock for posting completions. Other ring - * types rely on the regular completion locking, which is - * handled while posting. - */ - if (target_ctx->flags & IORING_SETUP_IOPOLL) - mutex_lock(&target_ctx->uring_lock); - if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags)) - ret = -EOVERFLOW; - if (target_ctx->flags & IORING_SETUP_IOPOLL) - mutex_unlock(&target_ctx->uring_lock); - } + struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg); + struct io_kiocb *target; + u32 flags = 0; - if (ret < 0) - req_set_fail(req); - io_req_queue_tw_complete(req, ret); + target = kmem_cache_alloc(req_cachep, GFP_KERNEL); + if (unlikely(!target)) + return -ENOMEM; + + if (msg->flags & IORING_MSG_RING_FLAGS_PASS) + flags = msg->cqe_flags; + + io_msg_remote_post(target_ctx, target, msg->len, flags, msg->user_data); + return 0; } static int io_msg_ring_data(struct io_kiocb *req, unsigned int issue_flags) @@ -136,7 +124,7 @@ static int io_msg_ring_data(struct io_kiocb *req, unsigned int issue_flags) return -EBADFD; if (io_msg_need_remote(target_ctx)) - return io_msg_exec_remote(req, io_msg_tw_complete); + return io_msg_data_remote(req); if (msg->flags & IORING_MSG_RING_FLAGS_PASS) flags = msg->cqe_flags; @@ -216,6 +204,22 @@ static void io_msg_tw_fd_complete(struct callback_head *head) io_req_queue_tw_complete(req, ret); } +static int io_msg_fd_remote(struct io_kiocb *req) +{ + struct io_ring_ctx *ctx = req->file->private_data; + struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg); + struct task_struct *task = READ_ONCE(ctx->submitter_task); + + if (unlikely(!task)) + return -EOWNERDEAD; + + init_task_work(&msg->tw, io_msg_tw_fd_complete); + if (task_work_add(task, &msg->tw, TWA_SIGNAL)) + return -EOWNERDEAD; + + return IOU_ISSUE_SKIP_COMPLETE; +} + static int io_msg_send_fd(struct io_kiocb *req, unsigned int issue_flags) { struct io_ring_ctx *target_ctx = req->file->private_data; @@ -238,7 +242,7 @@ static int io_msg_send_fd(struct io_kiocb *req, unsigned int issue_flags) } if (io_msg_need_remote(target_ctx)) - return io_msg_exec_remote(req, io_msg_tw_fd_complete); + return io_msg_fd_remote(req); return io_msg_install_complete(req, issue_flags); } From 50cf5f3842af3135b88b041890e7e12a74425fcb Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 6 Jun 2024 12:25:01 -0600 Subject: [PATCH 135/209] io_uring/msg_ring: add an alloc cache for io_kiocb entries With slab accounting, allocating and freeing memory has considerable overhead. Add a basic alloc cache for the io_kiocb allocations that msg_ring needs to do. Unlike other caches, this one is used by the sender, grabbing it from the remote ring. When the remote ring gets the posted completion, it'll free it locally. Hence it is separately locked, using ctx->msg_lock. Signed-off-by: Jens Axboe --- include/linux/io_uring_types.h | 3 +++ io_uring/io_uring.c | 6 ++++++ io_uring/msg_ring.c | 31 +++++++++++++++++++++++++++++-- io_uring/msg_ring.h | 1 + 4 files changed, 39 insertions(+), 2 deletions(-) diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h index 1052a68fd68df4..ede42dce15067c 100644 --- a/include/linux/io_uring_types.h +++ b/include/linux/io_uring_types.h @@ -397,6 +397,9 @@ struct io_ring_ctx { struct callback_head poll_wq_task_work; struct list_head defer_list; + struct io_alloc_cache msg_cache; + spinlock_t msg_lock; + #ifdef CONFIG_NET_RX_BUSY_POLL struct list_head napi_list; /* track busy poll napi_id */ spinlock_t napi_lock; /* napi_list lock */ diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index cdeb94d2a26b61..7ed1e009aaecb0 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -95,6 +95,7 @@ #include "futex.h" #include "napi.h" #include "uring_cmd.h" +#include "msg_ring.h" #include "memmap.h" #include "timeout.h" @@ -315,6 +316,9 @@ static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p) sizeof(struct io_async_rw)); ret |= io_alloc_cache_init(&ctx->uring_cache, IO_ALLOC_CACHE_MAX, sizeof(struct uring_cache)); + spin_lock_init(&ctx->msg_lock); + ret |= io_alloc_cache_init(&ctx->msg_cache, IO_ALLOC_CACHE_MAX, + sizeof(struct io_kiocb)); ret |= io_futex_cache_init(ctx); if (ret) goto err; @@ -351,6 +355,7 @@ static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p) io_alloc_cache_free(&ctx->netmsg_cache, io_netmsg_cache_free); io_alloc_cache_free(&ctx->rw_cache, io_rw_cache_free); io_alloc_cache_free(&ctx->uring_cache, kfree); + io_alloc_cache_free(&ctx->msg_cache, io_msg_cache_free); io_futex_cache_free(ctx); kfree(ctx->cancel_table.hbs); kfree(ctx->cancel_table_locked.hbs); @@ -2599,6 +2604,7 @@ static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx) io_alloc_cache_free(&ctx->netmsg_cache, io_netmsg_cache_free); io_alloc_cache_free(&ctx->rw_cache, io_rw_cache_free); io_alloc_cache_free(&ctx->uring_cache, kfree); + io_alloc_cache_free(&ctx->msg_cache, io_msg_cache_free); io_futex_cache_free(ctx); io_destroy_buffers(ctx); mutex_unlock(&ctx->uring_lock); diff --git a/io_uring/msg_ring.c b/io_uring/msg_ring.c index ad7d67d444612b..47a754e83b49a9 100644 --- a/io_uring/msg_ring.c +++ b/io_uring/msg_ring.c @@ -11,6 +11,7 @@ #include "io_uring.h" #include "rsrc.h" #include "filetable.h" +#include "alloc_cache.h" #include "msg_ring.h" /* All valid masks for MSG_RING */ @@ -75,7 +76,13 @@ static void io_msg_tw_complete(struct io_kiocb *req, struct io_tw_state *ts) struct io_ring_ctx *ctx = req->ctx; io_add_aux_cqe(ctx, req->cqe.user_data, req->cqe.res, req->cqe.flags); - kmem_cache_free(req_cachep, req); + if (spin_trylock(&ctx->msg_lock)) { + if (io_alloc_cache_put(&ctx->msg_cache, req)) + req = NULL; + spin_unlock(&ctx->msg_lock); + } + if (req) + kfree(req); percpu_ref_put(&ctx->refs); } @@ -91,6 +98,19 @@ static void io_msg_remote_post(struct io_ring_ctx *ctx, struct io_kiocb *req, io_req_task_work_add_remote(req, ctx, IOU_F_TWQ_LAZY_WAKE); } +static struct io_kiocb *io_msg_get_kiocb(struct io_ring_ctx *ctx) +{ + struct io_kiocb *req = NULL; + + if (spin_trylock(&ctx->msg_lock)) { + req = io_alloc_cache_get(&ctx->msg_cache); + spin_unlock(&ctx->msg_lock); + } + if (req) + return req; + return kmem_cache_alloc(req_cachep, GFP_KERNEL | __GFP_NOWARN); +} + static int io_msg_data_remote(struct io_kiocb *req) { struct io_ring_ctx *target_ctx = req->file->private_data; @@ -98,7 +118,7 @@ static int io_msg_data_remote(struct io_kiocb *req) struct io_kiocb *target; u32 flags = 0; - target = kmem_cache_alloc(req_cachep, GFP_KERNEL); + target = io_msg_get_kiocb(req->ctx); if (unlikely(!target)) return -ENOMEM; @@ -296,3 +316,10 @@ int io_msg_ring(struct io_kiocb *req, unsigned int issue_flags) io_req_set_res(req, ret, 0); return IOU_OK; } + +void io_msg_cache_free(const void *entry) +{ + struct io_kiocb *req = (struct io_kiocb *) entry; + + kmem_cache_free(req_cachep, req); +} diff --git a/io_uring/msg_ring.h b/io_uring/msg_ring.h index 3987ee6c0e5f1d..3030f3942f0f5e 100644 --- a/io_uring/msg_ring.h +++ b/io_uring/msg_ring.h @@ -3,3 +3,4 @@ int io_msg_ring_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe); int io_msg_ring(struct io_kiocb *req, unsigned int issue_flags); void io_msg_ring_cleanup(struct io_kiocb *req); +void io_msg_cache_free(const void *entry); From 44348870de4b8f292f97b84583a298d66fbaf738 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 24 Jun 2024 19:38:35 +0200 Subject: [PATCH 136/209] block: fix the blk_queue_nonrot polarity Take care of the inverse polarity of the BLK_FEAT_ROTATIONAL flag vs the old nonrot helper. Fixes: bd4a633b6f7c ("block: move the nonrot flag to queue_limits") Reported-by: kernel test robot Reported-by: Keith Busch Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Link: https://lore.kernel.org/r/20240624173835.76753-1-hch@lst.de Signed-off-by: Jens Axboe --- include/linux/blkdev.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index e89003360c17c7..b2f1362c46814f 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -617,7 +617,7 @@ bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) #define blk_queue_noxmerges(q) \ test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) -#define blk_queue_nonrot(q) ((q)->limits.features & BLK_FEAT_ROTATIONAL) +#define blk_queue_nonrot(q) (!((q)->limits.features & BLK_FEAT_ROTATIONAL)) #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) #define blk_queue_zone_resetall(q) \ ((q)->limits.features & BLK_FEAT_ZONE_RESETALL) From cf546dd289e0f6d2594c25e2fb4e19ee67c6d988 Mon Sep 17 00:00:00 2001 From: Mikulas Patocka Date: Mon, 27 May 2024 17:40:10 +0200 Subject: [PATCH 137/209] block: change rq_integrity_vec to respect the iterator If we allocate a bio that is larger than NVMe maximum request size, attach integrity metadata to it and send it to the NVMe subsystem, the integrity metadata will be corrupted. Splitting the bio works correctly. The function bio_split will clone the bio, trim the iterator of the first bio and advance the iterator of the second bio. However, the function rq_integrity_vec has a bug - it returns the first vector of the bio's metadata and completely disregards the metadata iterator that was advanced when the bio was split. Thus, the second bio uses the same metadata as the first bio and this leads to metadata corruption. This commit changes rq_integrity_vec, so that it calls mp_bvec_iter_bvec instead of returning the first vector. mp_bvec_iter_bvec reads the iterator and uses it to build a bvec for the current position in the iterator. The "queue_max_integrity_segments(rq->q) > 1" check was removed, because the updated rq_integrity_vec function works correctly with multiple segments. Signed-off-by: Mikulas Patocka Reviewed-by: Anuj Gupta Reviewed-by: Kanchan Joshi Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/49d1afaa-f934-6ed2-a678-e0d428c63a65@redhat.com Signed-off-by: Jens Axboe --- drivers/nvme/host/pci.c | 6 +++--- include/linux/blk-integrity.h | 14 +++++++------- 2 files changed, 10 insertions(+), 10 deletions(-) diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c index 102a9fb0c65fff..5d8035218de9bd 100644 --- a/drivers/nvme/host/pci.c +++ b/drivers/nvme/host/pci.c @@ -826,9 +826,9 @@ static blk_status_t nvme_map_metadata(struct nvme_dev *dev, struct request *req, struct nvme_command *cmnd) { struct nvme_iod *iod = blk_mq_rq_to_pdu(req); + struct bio_vec bv = rq_integrity_vec(req); - iod->meta_dma = dma_map_bvec(dev->dev, rq_integrity_vec(req), - rq_dma_dir(req), 0); + iod->meta_dma = dma_map_bvec(dev->dev, &bv, rq_dma_dir(req), 0); if (dma_mapping_error(dev->dev, iod->meta_dma)) return BLK_STS_IOERR; cmnd->rw.metadata = cpu_to_le64(iod->meta_dma); @@ -967,7 +967,7 @@ static __always_inline void nvme_pci_unmap_rq(struct request *req) struct nvme_iod *iod = blk_mq_rq_to_pdu(req); dma_unmap_page(dev->dev, iod->meta_dma, - rq_integrity_vec(req)->bv_len, rq_dma_dir(req)); + rq_integrity_vec(req).bv_len, rq_dma_dir(req)); } if (blk_rq_nr_phys_segments(req)) diff --git a/include/linux/blk-integrity.h b/include/linux/blk-integrity.h index d201140d77a336..0fdd62e6d4b02e 100644 --- a/include/linux/blk-integrity.h +++ b/include/linux/blk-integrity.h @@ -90,14 +90,13 @@ static inline bool blk_integrity_rq(struct request *rq) } /* - * Return the first bvec that contains integrity data. Only drivers that are - * limited to a single integrity segment should use this helper. + * Return the current bvec that contains the integrity data. bip_iter may be + * advanced to iterate over the integrity data. */ -static inline struct bio_vec *rq_integrity_vec(struct request *rq) +static inline struct bio_vec rq_integrity_vec(struct request *rq) { - if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1)) - return NULL; - return rq->bio->bi_integrity->bip_vec; + return mp_bvec_iter_bvec(rq->bio->bi_integrity->bip_vec, + rq->bio->bi_integrity->bip_iter); } #else /* CONFIG_BLK_DEV_INTEGRITY */ static inline int blk_rq_count_integrity_sg(struct request_queue *q, @@ -148,7 +147,8 @@ static inline int blk_integrity_rq(struct request *rq) static inline struct bio_vec *rq_integrity_vec(struct request *rq) { - return NULL; + /* the optimizer will remove all calls to this function */ + return (struct bio_vec){ }; } #endif /* CONFIG_BLK_DEV_INTEGRITY */ From 573d5abf3df00c879fbd25774e4cf3e22c9cabd0 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 16:26:22 +0200 Subject: [PATCH 138/209] md: set md-specific flags for all queue limits The md driver wants to enforce a number of flags for all devices, even when not inheriting them from the underlying devices. To make sure these flags survive the queue_limits_set calls that md uses to update the queue limits without deriving them form the previous limits add a new md_init_stacking_limits helper that calls blk_set_stacking_limits and sets these flags. Fixes: 1122c0c1cc71 ("block: move cache control settings out of queue->flags") Reported-by: kernel test robot Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Link: https://lore.kernel.org/r/20240626142637.300624-2-hch@lst.de Signed-off-by: Jens Axboe --- drivers/md/md.c | 14 +++++++++----- drivers/md/md.h | 1 + drivers/md/raid0.c | 2 +- drivers/md/raid1.c | 2 +- drivers/md/raid10.c | 2 +- drivers/md/raid5.c | 2 +- 6 files changed, 14 insertions(+), 9 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index 69ea54aedd99a1..0ff26a547f1afc 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -5853,6 +5853,14 @@ static void mddev_delayed_delete(struct work_struct *ws) kobject_put(&mddev->kobj); } +void md_init_stacking_limits(struct queue_limits *lim) +{ + blk_set_stacking_limits(lim); + lim->features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | + BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT; +} +EXPORT_SYMBOL_GPL(md_init_stacking_limits); + struct mddev *md_alloc(dev_t dev, char *name) { /* @@ -5871,10 +5879,6 @@ struct mddev *md_alloc(dev_t dev, char *name) int shift; int unit; int error; - struct queue_limits lim = { - .features = BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | - BLK_FEAT_IO_STAT | BLK_FEAT_NOWAIT, - }; /* * Wait for any previous instance of this device to be completely @@ -5914,7 +5918,7 @@ struct mddev *md_alloc(dev_t dev, char *name) */ mddev->hold_active = UNTIL_STOP; - disk = blk_alloc_disk(&lim, NUMA_NO_NODE); + disk = blk_alloc_disk(NULL, NUMA_NO_NODE); if (IS_ERR(disk)) { error = PTR_ERR(disk); goto out_free_mddev; diff --git a/drivers/md/md.h b/drivers/md/md.h index c4d7ebf9587d07..28cb4b0b6c1740 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -893,6 +893,7 @@ extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale); extern int mddev_init(struct mddev *mddev); extern void mddev_destroy(struct mddev *mddev); +void md_init_stacking_limits(struct queue_limits *lim); struct mddev *md_alloc(dev_t dev, char *name); void mddev_put(struct mddev *mddev); extern int md_run(struct mddev *mddev); diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c index 62634e2a33bd0f..32d58752477847 100644 --- a/drivers/md/raid0.c +++ b/drivers/md/raid0.c @@ -379,7 +379,7 @@ static int raid0_set_limits(struct mddev *mddev) struct queue_limits lim; int err; - blk_set_stacking_limits(&lim); + md_init_stacking_limits(&lim); lim.max_hw_sectors = mddev->chunk_sectors; lim.max_write_zeroes_sectors = mddev->chunk_sectors; lim.io_min = mddev->chunk_sectors << 9; diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 1a0eba65b8a92b..04a0c2ca173245 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -3194,7 +3194,7 @@ static int raid1_set_limits(struct mddev *mddev) struct queue_limits lim; int err; - blk_set_stacking_limits(&lim); + md_init_stacking_limits(&lim); lim.max_write_zeroes_sectors = 0; err = mddev_stack_rdev_limits(mddev, &lim, MDDEV_STACK_INTEGRITY); if (err) { diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 3334aa803c8380..2a9c4ee982e023 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -3974,7 +3974,7 @@ static int raid10_set_queue_limits(struct mddev *mddev) struct queue_limits lim; int err; - blk_set_stacking_limits(&lim); + md_init_stacking_limits(&lim); lim.max_write_zeroes_sectors = 0; lim.io_min = mddev->chunk_sectors << 9; lim.io_opt = lim.io_min * raid10_nr_stripes(conf); diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 0192a6323f09ba..10219205160bbf 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -7708,7 +7708,7 @@ static int raid5_set_limits(struct mddev *mddev) */ stripe = roundup_pow_of_two(data_disks * (mddev->chunk_sectors << 9)); - blk_set_stacking_limits(&lim); + md_init_stacking_limits(&lim); lim.io_min = mddev->chunk_sectors << 9; lim.io_opt = lim.io_min * (conf->raid_disks - conf->max_degraded); lim.features |= BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE; From 78887d004fb2bb03233122a048eaf46e850dabf4 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 16:26:23 +0200 Subject: [PATCH 139/209] block: correctly report cache type Check the features flag and the override flag using the blk_queue_write_cache, helper otherwise we're going to always report "write through". Fixes: 1122c0c1cc71 ("block: move cache control settings out of queue->flags") Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Reviewed-by: John Garry Link: https://lore.kernel.org/r/20240626142637.300624-3-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-sysfs.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 1a984179f3acc5..3a167abecdceae 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -453,9 +453,9 @@ static ssize_t queue_io_timeout_store(struct request_queue *q, const char *page, static ssize_t queue_wc_show(struct request_queue *q, char *page) { - if (q->limits.features & BLK_FLAG_WRITE_CACHE_DISABLED) - return sprintf(page, "write through\n"); - return sprintf(page, "write back\n"); + if (blk_queue_write_cache(q)) + return sprintf(page, "write back\n"); + return sprintf(page, "write through\n"); } static ssize_t queue_wc_store(struct request_queue *q, const char *page, From ec9b1cf0b0ebfb52274971a8a0e74e0a133f64fb Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 16:26:24 +0200 Subject: [PATCH 140/209] block: rename BLK_FEAT_MISALIGNED This is a flag for ->flags and not a feature for ->features. And fix the one place that actually incorrectly cleared it from ->features. Signed-off-by: Christoph Hellwig Reviewed-by: John Garry Reviewed-by: Damien Le Moal Link: https://lore.kernel.org/r/20240626142637.300624-4-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 18 +++++++++--------- include/linux/blkdev.h | 2 +- 2 files changed, 10 insertions(+), 10 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index ec7dbe93d5c324..ed39a55c5bae7c 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -351,7 +351,7 @@ static int blk_validate_limits(struct queue_limits *lim) if (lim->alignment_offset) { lim->alignment_offset &= (lim->physical_block_size - 1); - lim->features &= ~BLK_FEAT_MISALIGNED; + lim->flags &= ~BLK_FLAG_MISALIGNED; } if (!(lim->features & BLK_FEAT_WRITE_CACHE)) @@ -564,7 +564,7 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, if (!(b->features & BLK_FEAT_POLL)) t->features &= ~BLK_FEAT_POLL; - t->flags |= (b->flags & BLK_FEAT_MISALIGNED); + t->flags |= (b->flags & BLK_FLAG_MISALIGNED); t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors); t->max_user_sectors = min_not_zero(t->max_user_sectors, @@ -603,7 +603,7 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, /* Verify that top and bottom intervals line up */ if (max(top, bottom) % min(top, bottom)) { - t->flags |= BLK_FEAT_MISALIGNED; + t->flags |= BLK_FLAG_MISALIGNED; ret = -1; } } @@ -625,28 +625,28 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, /* Physical block size a multiple of the logical block size? */ if (t->physical_block_size & (t->logical_block_size - 1)) { t->physical_block_size = t->logical_block_size; - t->flags |= BLK_FEAT_MISALIGNED; + t->flags |= BLK_FLAG_MISALIGNED; ret = -1; } /* Minimum I/O a multiple of the physical block size? */ if (t->io_min & (t->physical_block_size - 1)) { t->io_min = t->physical_block_size; - t->flags |= BLK_FEAT_MISALIGNED; + t->flags |= BLK_FLAG_MISALIGNED; ret = -1; } /* Optimal I/O a multiple of the physical block size? */ if (t->io_opt & (t->physical_block_size - 1)) { t->io_opt = 0; - t->flags |= BLK_FEAT_MISALIGNED; + t->flags |= BLK_FLAG_MISALIGNED; ret = -1; } /* chunk_sectors a multiple of the physical block size? */ if ((t->chunk_sectors << 9) & (t->physical_block_size - 1)) { t->chunk_sectors = 0; - t->flags |= BLK_FEAT_MISALIGNED; + t->flags |= BLK_FLAG_MISALIGNED; ret = -1; } @@ -656,7 +656,7 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, /* Verify that new alignment_offset is on a logical block boundary */ if (t->alignment_offset & (t->logical_block_size - 1)) { - t->flags |= BLK_FEAT_MISALIGNED; + t->flags |= BLK_FLAG_MISALIGNED; ret = -1; } @@ -809,7 +809,7 @@ int bdev_alignment_offset(struct block_device *bdev) { struct request_queue *q = bdev_get_queue(bdev); - if (q->limits.flags & BLK_FEAT_MISALIGNED) + if (q->limits.flags & BLK_FLAG_MISALIGNED) return -1; if (bdev_is_partition(bdev)) return queue_limit_alignment_offset(&q->limits, diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index b2f1362c46814f..1a7e9d9c16d78b 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -347,7 +347,7 @@ enum { BLK_FLAG_WRITE_CACHE_DISABLED = (1u << 0), /* I/O topology is misaligned */ - BLK_FEAT_MISALIGNED = (1u << 1), + BLK_FLAG_MISALIGNED = (1u << 1), }; struct queue_limits { From fcf865e357f80285af12c0c9a49f89d71acb7f4b Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 16:26:25 +0200 Subject: [PATCH 141/209] block: convert features and flags to __bitwise types ... and let sparse help us catch mismatches or abuses. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Link: https://lore.kernel.org/r/20240626142637.300624-5-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-sysfs.c | 6 +-- include/linux/blkdev.h | 85 +++++++++++++++++++++--------------------- 2 files changed, 46 insertions(+), 45 deletions(-) diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 3a167abecdceae..2e6d9b918127fe 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -288,7 +288,7 @@ static ssize_t queue_dma_alignment_show(struct request_queue *q, char *page) } static ssize_t queue_feature_store(struct request_queue *q, const char *page, - size_t count, unsigned int feature) + size_t count, blk_features_t feature) { struct queue_limits lim; unsigned long val; @@ -418,7 +418,7 @@ static ssize_t queue_poll_delay_store(struct request_queue *q, const char *page, static ssize_t queue_poll_show(struct request_queue *q, char *page) { - return queue_var_show(q->limits.features & BLK_FEAT_POLL, page); + return queue_var_show(!!(q->limits.features & BLK_FEAT_POLL), page); } static ssize_t queue_poll_store(struct request_queue *q, const char *page, @@ -492,7 +492,7 @@ static ssize_t queue_fua_show(struct request_queue *q, char *page) static ssize_t queue_dax_show(struct request_queue *q, char *page) { - return queue_var_show(blk_queue_dax(q), page); + return queue_var_show(!!blk_queue_dax(q), page); } #define QUEUE_RO_ENTRY(_prefix, _name) \ diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 1a7e9d9c16d78b..b37826b350a2e3 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -283,55 +283,56 @@ static inline bool blk_op_is_passthrough(blk_opf_t op) } /* flags set by the driver in queue_limits.features */ -enum { - /* supports a volatile write cache */ - BLK_FEAT_WRITE_CACHE = (1u << 0), +typedef unsigned int __bitwise blk_features_t; - /* supports passing on the FUA bit */ - BLK_FEAT_FUA = (1u << 1), +/* supports a volatile write cache */ +#define BLK_FEAT_WRITE_CACHE ((__force blk_features_t)(1u << 0)) - /* rotational device (hard drive or floppy) */ - BLK_FEAT_ROTATIONAL = (1u << 2), +/* supports passing on the FUA bit */ +#define BLK_FEAT_FUA ((__force blk_features_t)(1u << 1)) - /* contributes to the random number pool */ - BLK_FEAT_ADD_RANDOM = (1u << 3), +/* rotational device (hard drive or floppy) */ +#define BLK_FEAT_ROTATIONAL ((__force blk_features_t)(1u << 2)) - /* do disk/partitions IO accounting */ - BLK_FEAT_IO_STAT = (1u << 4), +/* contributes to the random number pool */ +#define BLK_FEAT_ADD_RANDOM ((__force blk_features_t)(1u << 3)) - /* don't modify data until writeback is done */ - BLK_FEAT_STABLE_WRITES = (1u << 5), +/* do disk/partitions IO accounting */ +#define BLK_FEAT_IO_STAT ((__force blk_features_t)(1u << 4)) - /* always completes in submit context */ - BLK_FEAT_SYNCHRONOUS = (1u << 6), +/* don't modify data until writeback is done */ +#define BLK_FEAT_STABLE_WRITES ((__force blk_features_t)(1u << 5)) - /* supports REQ_NOWAIT */ - BLK_FEAT_NOWAIT = (1u << 7), +/* always completes in submit context */ +#define BLK_FEAT_SYNCHRONOUS ((__force blk_features_t)(1u << 6)) - /* supports DAX */ - BLK_FEAT_DAX = (1u << 8), +/* supports REQ_NOWAIT */ +#define BLK_FEAT_NOWAIT ((__force blk_features_t)(1u << 7)) - /* supports I/O polling */ - BLK_FEAT_POLL = (1u << 9), +/* supports DAX */ +#define BLK_FEAT_DAX ((__force blk_features_t)(1u << 8)) - /* is a zoned device */ - BLK_FEAT_ZONED = (1u << 10), +/* supports I/O polling */ +#define BLK_FEAT_POLL ((__force blk_features_t)(1u << 9)) - /* supports Zone Reset All */ - BLK_FEAT_ZONE_RESETALL = (1u << 11), +/* is a zoned device */ +#define BLK_FEAT_ZONED ((__force blk_features_t)(1u << 10)) - /* supports PCI(e) p2p requests */ - BLK_FEAT_PCI_P2PDMA = (1u << 12), +/* supports Zone Reset All */ +#define BLK_FEAT_ZONE_RESETALL ((__force blk_features_t)(1u << 11)) - /* skip this queue in blk_mq_(un)quiesce_tagset */ - BLK_FEAT_SKIP_TAGSET_QUIESCE = (1u << 13), +/* supports PCI(e) p2p requests */ +#define BLK_FEAT_PCI_P2PDMA ((__force blk_features_t)(1u << 12)) - /* bounce all highmem pages */ - BLK_FEAT_BOUNCE_HIGH = (1u << 14), +/* skip this queue in blk_mq_(un)quiesce_tagset */ +#define BLK_FEAT_SKIP_TAGSET_QUIESCE ((__force blk_features_t)(1u << 13)) - /* undocumented magic for bcache */ - BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE = (1u << 15), -}; +/* bounce all highmem pages */ +#define BLK_FEAT_BOUNCE_HIGH ((__force blk_features_t)(1u << 14)) + +/* undocumented magic for bcache */ +#define BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE \ + ((__force blk_features_t)(1u << 15)) /* * Flags automatically inherited when stacking limits. @@ -342,17 +343,17 @@ enum { BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE) /* internal flags in queue_limits.flags */ -enum { - /* do not send FLUSH/FUA commands despite advertising a write cache */ - BLK_FLAG_WRITE_CACHE_DISABLED = (1u << 0), +typedef unsigned int __bitwise blk_flags_t; - /* I/O topology is misaligned */ - BLK_FLAG_MISALIGNED = (1u << 1), -}; +/* do not send FLUSH/FUA commands despite advertising a write cache */ +#define BLK_FLAG_WRITE_CACHE_DISABLED ((__force blk_flags_t)(1u << 0)) + +/* I/O topology is misaligned */ +#define BLK_FLAG_MISALIGNED ((__force blk_flags_t)(1u << 1)) struct queue_limits { - unsigned int features; - unsigned int flags; + blk_features_t features; + blk_flags_t flags; unsigned long seg_boundary_mask; unsigned long virt_boundary_mask; From 3302f6f09052274945f877beeb83f74641de2418 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 16:26:26 +0200 Subject: [PATCH 142/209] block: conding style fixup for blk_queue_max_guaranteed_bio "static" never goes on a line of its own. Signed-off-by: Christoph Hellwig Reviewed-by: John Garry Reviewed-by: Damien Le Moal Link: https://lore.kernel.org/r/20240626142637.300624-6-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index ed39a55c5bae7c..c2221b7406d46a 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -142,8 +142,7 @@ static int blk_validate_integrity_limits(struct queue_limits *lim) * so we assume that we can fit in at least PAGE_SIZE in a segment, apart from * the first and last segments. */ -static -unsigned int blk_queue_max_guaranteed_bio(struct queue_limits *lim) +static unsigned int blk_queue_max_guaranteed_bio(struct queue_limits *lim) { unsigned int max_segments = min(BIO_MAX_VECS, lim->max_segments); unsigned int length; From 73781b3b81e76583708a652c853d54d03dce031d Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 16:26:27 +0200 Subject: [PATCH 143/209] block: remove disk_update_readahead Mark blk_apply_bdi_limits non-static and open code disk_update_readahead in the only caller. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Link: https://lore.kernel.org/r/20240626142637.300624-7-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 8 +------- block/blk.h | 2 ++ block/genhd.c | 2 +- include/linux/blkdev.h | 1 - 4 files changed, 4 insertions(+), 9 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index c2221b7406d46a..c692e80bb4f890 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -55,7 +55,7 @@ void blk_set_stacking_limits(struct queue_limits *lim) } EXPORT_SYMBOL(blk_set_stacking_limits); -static void blk_apply_bdi_limits(struct backing_dev_info *bdi, +void blk_apply_bdi_limits(struct backing_dev_info *bdi, struct queue_limits *lim) { /* @@ -434,12 +434,6 @@ int queue_limits_set(struct request_queue *q, struct queue_limits *lim) } EXPORT_SYMBOL_GPL(queue_limits_set); -void disk_update_readahead(struct gendisk *disk) -{ - blk_apply_bdi_limits(disk->bdi, &disk->queue->limits); -} -EXPORT_SYMBOL_GPL(disk_update_readahead); - /** * blk_limits_io_min - set minimum request size for a device * @limits: the queue limits diff --git a/block/blk.h b/block/blk.h index d0a986d8ee507e..95e5a4f81693c4 100644 --- a/block/blk.h +++ b/block/blk.h @@ -358,6 +358,8 @@ bool blk_rq_merge_ok(struct request *rq, struct bio *bio); enum elv_merge blk_try_merge(struct request *rq, struct bio *bio); int blk_set_default_limits(struct queue_limits *lim); +void blk_apply_bdi_limits(struct backing_dev_info *bdi, + struct queue_limits *lim); int blk_dev_init(void); /* diff --git a/block/genhd.c b/block/genhd.c index 8f1f3c6b4d6729..4dc95a46350532 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -524,7 +524,7 @@ int __must_check device_add_disk(struct device *parent, struct gendisk *disk, disk->part0->bd_dev = MKDEV(disk->major, disk->first_minor); } - disk_update_readahead(disk); + blk_apply_bdi_limits(disk->bdi, &disk->queue->limits); disk_add_events(disk); set_bit(GD_ADDED, &disk->state); return 0; diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index b37826b350a2e3..6b88382012e958 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -973,7 +973,6 @@ static inline void blk_queue_disable_write_zeroes(struct request_queue *q) /* * Access functions for manipulating queue properties */ -void disk_update_readahead(struct gendisk *disk); extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); From abfc9d810926dfbf5645c7755c8d5ab96273f27d Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 16:26:28 +0200 Subject: [PATCH 144/209] block: remove the fallback case in queue_dma_alignment Now that all updates go through blk_validate_limits the default of 511 is set at initialization time. Also remove the unused NULL check as calling this helper on a NULL queue can't happen (and doesn't make much sense to start with). Signed-off-by: Christoph Hellwig Reviewed-by: John Garry Reviewed-by: Damien Le Moal Link: https://lore.kernel.org/r/20240626142637.300624-8-hch@lst.de Signed-off-by: Jens Axboe --- include/linux/blkdev.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 6b88382012e958..94fcbc91231208 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -1394,7 +1394,7 @@ static inline bool bdev_is_zone_start(struct block_device *bdev, static inline int queue_dma_alignment(const struct request_queue *q) { - return q ? q->limits.dma_alignment : 511; + return q->limits.dma_alignment; } static inline unsigned int From e94b45d08b5d1c230c0f59c3eed758d28658851e Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 16:26:29 +0200 Subject: [PATCH 145/209] block: move dma_pad_mask into queue_limits dma_pad_mask is a queue_limits by all ways of looking at it, so move it there and set it through the atomic queue limits APIs. Add a little helper that takes the alignment and pad into account to simplify the code that is touched a bit. Note that there never was any need for the > check in blk_queue_update_dma_pad, this probably was just copy and paste from dma_update_dma_alignment. Signed-off-by: Christoph Hellwig Reviewed-by: Damien Le Moal Link: https://lore.kernel.org/r/20240626142637.300624-9-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 2 +- block/blk-map.c | 2 +- block/blk-settings.c | 17 ----------------- drivers/ata/libata-scsi.c | 3 +-- drivers/ata/pata_macio.c | 4 ++-- drivers/scsi/scsi_lib.c | 4 ++-- drivers/ufs/core/ufshcd.c | 10 ++++++---- include/linux/blkdev.h | 12 ++++++++---- 8 files changed, 21 insertions(+), 33 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index 173ffd4d623788..356ca0d3d62f5a 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -312,7 +312,7 @@ int bio_integrity_map_user(struct bio *bio, void __user *ubuf, ssize_t bytes, u32 seed) { struct request_queue *q = bdev_get_queue(bio->bi_bdev); - unsigned int align = q->dma_pad_mask | queue_dma_alignment(q); + unsigned int align = blk_lim_dma_alignment_and_pad(&q->limits); struct page *stack_pages[UIO_FASTIOV], **pages = stack_pages; struct bio_vec stack_vec[UIO_FASTIOV], *bvec = stack_vec; unsigned int direction, nr_bvecs; diff --git a/block/blk-map.c b/block/blk-map.c index 71210cdb34426d..bce144091128f6 100644 --- a/block/blk-map.c +++ b/block/blk-map.c @@ -634,7 +634,7 @@ int blk_rq_map_user_iov(struct request_queue *q, struct request *rq, const struct iov_iter *iter, gfp_t gfp_mask) { bool copy = false, map_bvec = false; - unsigned long align = q->dma_pad_mask | queue_dma_alignment(q); + unsigned long align = blk_lim_dma_alignment_and_pad(&q->limits); struct bio *bio = NULL; struct iov_iter i; int ret = -EINVAL; diff --git a/block/blk-settings.c b/block/blk-settings.c index c692e80bb4f890..2e559cf97cc834 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -768,23 +768,6 @@ bool queue_limits_stack_integrity(struct queue_limits *t, } EXPORT_SYMBOL_GPL(queue_limits_stack_integrity); -/** - * blk_queue_update_dma_pad - update pad mask - * @q: the request queue for the device - * @mask: pad mask - * - * Update dma pad mask. - * - * Appending pad buffer to a request modifies the last entry of a - * scatter list such that it includes the pad buffer. - **/ -void blk_queue_update_dma_pad(struct request_queue *q, unsigned int mask) -{ - if (mask > q->dma_pad_mask) - q->dma_pad_mask = mask; -} -EXPORT_SYMBOL(blk_queue_update_dma_pad); - /** * blk_set_queue_depth - tell the block layer about the device queue depth * @q: the request queue for the device diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c index cdf29b178ddc1e..682971c4cbe418 100644 --- a/drivers/ata/libata-scsi.c +++ b/drivers/ata/libata-scsi.c @@ -1024,7 +1024,6 @@ EXPORT_SYMBOL_GPL(ata_scsi_dma_need_drain); int ata_scsi_dev_config(struct scsi_device *sdev, struct queue_limits *lim, struct ata_device *dev) { - struct request_queue *q = sdev->request_queue; int depth = 1; if (!ata_id_has_unload(dev->id)) @@ -1038,7 +1037,7 @@ int ata_scsi_dev_config(struct scsi_device *sdev, struct queue_limits *lim, sdev->sector_size = ATA_SECT_SIZE; /* set DMA padding */ - blk_queue_update_dma_pad(q, ATA_DMA_PAD_SZ - 1); + lim->dma_pad_mask = ATA_DMA_PAD_SZ - 1; /* make room for appending the drain */ lim->max_segments--; diff --git a/drivers/ata/pata_macio.c b/drivers/ata/pata_macio.c index 3cb455a32d9266..1b85e8bf4ef91b 100644 --- a/drivers/ata/pata_macio.c +++ b/drivers/ata/pata_macio.c @@ -816,7 +816,7 @@ static int pata_macio_device_configure(struct scsi_device *sdev, /* OHare has issues with non cache aligned DMA on some chipsets */ if (priv->kind == controller_ohare) { lim->dma_alignment = 31; - blk_queue_update_dma_pad(sdev->request_queue, 31); + lim->dma_pad_mask = 31; /* Tell the world about it */ ata_dev_info(dev, "OHare alignment limits applied\n"); @@ -831,7 +831,7 @@ static int pata_macio_device_configure(struct scsi_device *sdev, if (priv->kind == controller_sh_ata6 || priv->kind == controller_k2_ata6) { /* Allright these are bad, apply restrictions */ lim->dma_alignment = 15; - blk_queue_update_dma_pad(sdev->request_queue, 15); + lim->dma_pad_mask = 15; /* We enable MWI and hack cache line size directly here, this * is specific to this chipset and not normal values, we happen diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c index e2f7bfb2b9e450..3958a6d14bf457 100644 --- a/drivers/scsi/scsi_lib.c +++ b/drivers/scsi/scsi_lib.c @@ -1139,9 +1139,9 @@ blk_status_t scsi_alloc_sgtables(struct scsi_cmnd *cmd) */ count = __blk_rq_map_sg(rq->q, rq, cmd->sdb.table.sgl, &last_sg); - if (blk_rq_bytes(rq) & rq->q->dma_pad_mask) { + if (blk_rq_bytes(rq) & rq->q->limits.dma_pad_mask) { unsigned int pad_len = - (rq->q->dma_pad_mask & ~blk_rq_bytes(rq)) + 1; + (rq->q->limits.dma_pad_mask & ~blk_rq_bytes(rq)) + 1; last_sg->length += pad_len; cmd->extra_len += pad_len; diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c index 0cf07194bbe89d..b7957a431589dd 100644 --- a/drivers/ufs/core/ufshcd.c +++ b/drivers/ufs/core/ufshcd.c @@ -5193,17 +5193,19 @@ static int ufshcd_change_queue_depth(struct scsi_device *sdev, int depth) } /** - * ufshcd_slave_configure - adjust SCSI device configurations + * ufshcd_device_configure - adjust SCSI device configurations * @sdev: pointer to SCSI device + * @lim: queue limits * * Return: 0 (success). */ -static int ufshcd_slave_configure(struct scsi_device *sdev) +static int ufshcd_device_configure(struct scsi_device *sdev, + struct queue_limits *lim) { struct ufs_hba *hba = shost_priv(sdev->host); struct request_queue *q = sdev->request_queue; - blk_queue_update_dma_pad(q, PRDT_DATA_BYTE_COUNT_PAD - 1); + lim->dma_pad_mask = PRDT_DATA_BYTE_COUNT_PAD - 1; /* * Block runtime-pm until all consumers are added. @@ -8907,7 +8909,7 @@ static const struct scsi_host_template ufshcd_driver_template = { .queuecommand = ufshcd_queuecommand, .mq_poll = ufshcd_poll, .slave_alloc = ufshcd_slave_alloc, - .slave_configure = ufshcd_slave_configure, + .device_configure = ufshcd_device_configure, .slave_destroy = ufshcd_slave_destroy, .change_queue_depth = ufshcd_change_queue_depth, .eh_abort_handler = ufshcd_abort, diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 94fcbc91231208..a53e3434e1a28c 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -401,6 +401,7 @@ struct queue_limits { * due to possible offsets. */ unsigned int dma_alignment; + unsigned int dma_pad_mask; struct blk_integrity integrity; }; @@ -509,8 +510,6 @@ struct request_queue { */ int id; - unsigned int dma_pad_mask; - /* * queue settings */ @@ -981,7 +980,6 @@ extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, sector_t offset); void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev, sector_t offset, const char *pfx); -extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); struct blk_independent_access_ranges * @@ -1433,10 +1431,16 @@ static inline bool bdev_iter_is_aligned(struct block_device *bdev, bdev_logical_block_size(bdev) - 1); } +static inline int blk_lim_dma_alignment_and_pad(struct queue_limits *lim) +{ + return lim->dma_alignment | lim->dma_pad_mask; +} + static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, unsigned int len) { - unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask; + unsigned int alignment = blk_lim_dma_alignment_and_pad(&q->limits); + return !(addr & alignment) && !(len & alignment); } From 69b6517687a4b1fb250bd8c9c193a0a304c8ba17 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Wed, 26 Jun 2024 19:01:58 -0600 Subject: [PATCH 146/209] block: use the right type for stub rq_integrity_vec() For !CONFIG_BLK_DEV_INTEGRITY, rq_integrity_vec() wasn't updated properly. Fix it up. Fixes: cf546dd289e0 ("block: change rq_integrity_vec to respect the iterator") Signed-off-by: Jens Axboe --- include/linux/blk-integrity.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/linux/blk-integrity.h b/include/linux/blk-integrity.h index 0fdd62e6d4b02e..c58634924782d0 100644 --- a/include/linux/blk-integrity.h +++ b/include/linux/blk-integrity.h @@ -145,7 +145,7 @@ static inline int blk_integrity_rq(struct request *rq) return 0; } -static inline struct bio_vec *rq_integrity_vec(struct request *rq) +static inline struct bio_vec rq_integrity_vec(struct request *rq) { /* the optimizer will remove all calls to this function */ return (struct bio_vec){ }; From e269537e491da6336776b5548a3c73f62273aa15 Mon Sep 17 00:00:00 2001 From: Li Nan Date: Tue, 25 Jun 2024 19:55:17 +0800 Subject: [PATCH 147/209] block: clean up the check in blkdev_iomap_begin() It is odd to check the offset amidst a series of assignments. Moving this check to the beginning of the function makes the code look better. Signed-off-by: Li Nan Reviewed-by: Chaitanya Kulkarni Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240625115517.1472120-1-linan666@huaweicloud.com Signed-off-by: Jens Axboe --- block/fops.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/block/fops.c b/block/fops.c index be36c9fbd500bb..9825c1713a49a9 100644 --- a/block/fops.c +++ b/block/fops.c @@ -394,10 +394,11 @@ static int blkdev_iomap_begin(struct inode *inode, loff_t offset, loff_t length, struct block_device *bdev = I_BDEV(inode); loff_t isize = i_size_read(inode); - iomap->bdev = bdev; - iomap->offset = ALIGN_DOWN(offset, bdev_logical_block_size(bdev)); if (offset >= isize) return -EIO; + + iomap->bdev = bdev; + iomap->offset = ALIGN_DOWN(offset, bdev_logical_block_size(bdev)); iomap->type = IOMAP_MAPPED; iomap->addr = iomap->offset; iomap->length = isize - iomap->offset; From 63db4a1f795a19e4e12f036a12a5f61c48b03e5c Mon Sep 17 00:00:00 2001 From: John Garry Date: Thu, 27 Jun 2024 16:07:35 +0000 Subject: [PATCH 148/209] block: Delete blk_queue_flag_test_and_set() Since commit 70200574cc22 ("block: remove QUEUE_FLAG_DISCARD"), blk_queue_flag_test_and_set() has not been used, so delete it. Signed-off-by: John Garry Link: https://lore.kernel.org/r/20240627160735.842189-1-john.g.garry@oracle.com Signed-off-by: Jens Axboe --- block/blk-core.c | 14 -------------- include/linux/blkdev.h | 1 - 2 files changed, 15 deletions(-) diff --git a/block/blk-core.c b/block/blk-core.c index 6fc1a5a1980db3..71b7622c523a30 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -94,20 +94,6 @@ void blk_queue_flag_clear(unsigned int flag, struct request_queue *q) } EXPORT_SYMBOL(blk_queue_flag_clear); -/** - * blk_queue_flag_test_and_set - atomically test and set a queue flag - * @flag: flag to be set - * @q: request queue - * - * Returns the previous value of @flag - 0 if the flag was not set and 1 if - * the flag was already set. - */ -bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q) -{ - return test_and_set_bit(flag, &q->queue_flags); -} -EXPORT_SYMBOL_GPL(blk_queue_flag_test_and_set); - #define REQ_OP_NAME(name) [REQ_OP_##name] = #name static const char *const blk_op_name[] = { REQ_OP_NAME(READ), diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index a53e3434e1a28c..53c41ef4222c3d 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -609,7 +609,6 @@ struct request_queue { void blk_queue_flag_set(unsigned int flag, struct request_queue *q); void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); -bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) From 18048c1af7836b8e31739d9eaefebc2bf76261f7 Mon Sep 17 00:00:00 2001 From: Gulam Mohamed Date: Tue, 18 Jun 2024 16:40:42 +0000 Subject: [PATCH 149/209] loop: Fix a race between loop detach and loop open 1. Userspace sends the command "losetup -d" which uses the open() call to open the device 2. Kernel receives the ioctl command "LOOP_CLR_FD" which calls the function loop_clr_fd() 3. If LOOP_CLR_FD is the first command received at the time, then the AUTOCLEAR flag is not set and deletion of the loop device proceeds ahead and scans the partitions (drop/add partitions) if (disk_openers(lo->lo_disk) > 1) { lo->lo_flags |= LO_FLAGS_AUTOCLEAR; loop_global_unlock(lo, true); return 0; } 4. Before scanning partitions, it will check to see if any partition of the loop device is currently opened 5. If any partition is opened, then it will return EBUSY: if (disk->open_partitions) return -EBUSY; 6. So, after receiving the "LOOP_CLR_FD" command and just before the above check for open_partitions, if any other command (like blkid) opens any partition of the loop device, then the partition scan will not proceed and EBUSY is returned as shown in above code 7. But in "__loop_clr_fd()", this EBUSY error is not propagated 8. We have noticed that this is causing the partitions of the loop to remain stale even after the loop device is detached resulting in the IO errors on the partitions Fix: Defer the detach of loop device to release function, which is called when the last close happens, by setting the lo_flags to LO_FLAGS_AUTOCLEAR at the time of detach i.e in loop_clr_fd() function. Test case involves the following two scripts: script1.sh: while [ 1 ]; do losetup -P -f /home/opt/looptest/test10.img blkid /dev/loop0p1 done script2.sh: while [ 1 ]; do losetup -d /dev/loop0 done Without fix, the following IO errors have been observed: kernel: __loop_clr_fd: partition scan of loop0 failed (rc=-16) kernel: I/O error, dev loop0, sector 20971392 op 0x0:(READ) flags 0x80700 phys_seg 1 prio class 0 kernel: I/O error, dev loop0, sector 108868 op 0x0:(READ) flags 0x0 phys_seg 1 prio class 0 kernel: Buffer I/O error on dev loop0p1, logical block 27201, async page read Signed-off-by: Gulam Mohamed Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240618164042.343777-1-gulam.mohamed@oracle.com Signed-off-by: Jens Axboe --- drivers/block/loop.c | 75 +++++++++++++++++++++----------------------- 1 file changed, 36 insertions(+), 39 deletions(-) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 86b5d956dc4e02..cbda971d22b40b 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -1123,20 +1123,12 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, return error; } -static void __loop_clr_fd(struct loop_device *lo, bool release) +static void __loop_clr_fd(struct loop_device *lo) { struct queue_limits lim; struct file *filp; gfp_t gfp = lo->old_gfp_mask; - /* - * Freeze the request queue when unbinding on a live file descriptor and - * thus an open device. When called from ->release we are guaranteed - * that there is no I/O in progress already. - */ - if (!release) - blk_mq_freeze_queue(lo->lo_queue); - spin_lock_irq(&lo->lo_lock); filp = lo->lo_backing_file; lo->lo_backing_file = NULL; @@ -1161,8 +1153,6 @@ static void __loop_clr_fd(struct loop_device *lo, bool release) mapping_set_gfp_mask(filp->f_mapping, gfp); /* This is safe: open() is still holding a reference. */ module_put(THIS_MODULE); - if (!release) - blk_mq_unfreeze_queue(lo->lo_queue); disk_force_media_change(lo->lo_disk); @@ -1177,11 +1167,7 @@ static void __loop_clr_fd(struct loop_device *lo, bool release) * must be at least one and it can only become zero when the * current holder is released. */ - if (!release) - mutex_lock(&lo->lo_disk->open_mutex); err = bdev_disk_changed(lo->lo_disk, false); - if (!release) - mutex_unlock(&lo->lo_disk->open_mutex); if (err) pr_warn("%s: partition scan of loop%d failed (rc=%d)\n", __func__, lo->lo_number, err); @@ -1230,24 +1216,16 @@ static int loop_clr_fd(struct loop_device *lo) return -ENXIO; } /* - * If we've explicitly asked to tear down the loop device, - * and it has an elevated reference count, set it for auto-teardown when - * the last reference goes away. This stops $!~#$@ udev from - * preventing teardown because it decided that it needs to run blkid on - * the loopback device whenever they appear. xfstests is notorious for - * failing tests because blkid via udev races with a losetup - * /do something like mkfs/losetup -d causing the losetup -d - * command to fail with EBUSY. + * Mark the device for removing the backing device on last close. + * If we are the only opener, also switch the state to roundown here to + * prevent new openers from coming in. */ - if (disk_openers(lo->lo_disk) > 1) { - lo->lo_flags |= LO_FLAGS_AUTOCLEAR; - loop_global_unlock(lo, true); - return 0; - } - lo->lo_state = Lo_rundown; + + lo->lo_flags |= LO_FLAGS_AUTOCLEAR; + if (disk_openers(lo->lo_disk) == 1) + lo->lo_state = Lo_rundown; loop_global_unlock(lo, true); - __loop_clr_fd(lo, false); return 0; } @@ -1714,25 +1692,43 @@ static int lo_compat_ioctl(struct block_device *bdev, blk_mode_t mode, } #endif +static int lo_open(struct gendisk *disk, blk_mode_t mode) +{ + struct loop_device *lo = disk->private_data; + int err; + + err = mutex_lock_killable(&lo->lo_mutex); + if (err) + return err; + + if (lo->lo_state == Lo_deleting || lo->lo_state == Lo_rundown) + err = -ENXIO; + mutex_unlock(&lo->lo_mutex); + return err; +} + static void lo_release(struct gendisk *disk) { struct loop_device *lo = disk->private_data; + bool need_clear = false; if (disk_openers(disk) > 0) return; + /* + * Clear the backing device information if this is the last close of + * a device that's been marked for auto clear, or on which LOOP_CLR_FD + * has been called. + */ mutex_lock(&lo->lo_mutex); - if (lo->lo_state == Lo_bound && (lo->lo_flags & LO_FLAGS_AUTOCLEAR)) { + if (lo->lo_state == Lo_bound && (lo->lo_flags & LO_FLAGS_AUTOCLEAR)) lo->lo_state = Lo_rundown; - mutex_unlock(&lo->lo_mutex); - /* - * In autoclear mode, stop the loop thread - * and remove configuration after last close. - */ - __loop_clr_fd(lo, true); - return; - } + + need_clear = (lo->lo_state == Lo_rundown); mutex_unlock(&lo->lo_mutex); + + if (need_clear) + __loop_clr_fd(lo); } static void lo_free_disk(struct gendisk *disk) @@ -1749,6 +1745,7 @@ static void lo_free_disk(struct gendisk *disk) static const struct block_device_operations lo_fops = { .owner = THIS_MODULE, + .open = lo_open, .release = lo_release, .ioctl = lo_ioctl, #ifdef CONFIG_COMPAT From f1e46758e8b2b04c725ac706b5f455c0de0486a4 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 28 Jun 2024 15:16:48 +0200 Subject: [PATCH 150/209] bcache: work around a __bitwise to bool conversion sparse warning Sparse is a bit dumb about bitwise operation on __bitwise types used in boolean contexts. Add a !! to explicitly propagate to boolean without a warning. Fixes: fcf865e357f8 ("block: convert features and flags to __bitwise types") Reported-by: kernel test robot Signed-off-by: Christoph Hellwig Acked-by: Kent Overstreet Link: https://lore.kernel.org/r/20240628131657.667797-1-hch@lst.de Signed-off-by: Jens Axboe --- drivers/md/bcache/super.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c index 283b2511c6d21f..b5d6ef430b86fc 100644 --- a/drivers/md/bcache/super.c +++ b/drivers/md/bcache/super.c @@ -1416,8 +1416,8 @@ static int cached_dev_init(struct cached_dev *dc, unsigned int block_size) } if (bdev_io_opt(dc->bdev)) - dc->partial_stripes_expensive = q->limits.features & - BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE; + dc->partial_stripes_expensive = !!(q->limits.features & + BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE); ret = bcache_device_init(&dc->disk, block_size, bdev_nr_sectors(dc->bdev) - dc->sb.data_offset, From c546d6f438338017480d105ab597292da67f6f6a Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 06:59:34 +0200 Subject: [PATCH 151/209] block: only zero non-PI metadata tuples in bio_integrity_prep The PI generation helpers already zero the app tag, so relax the zeroing to non-PI metadata. Signed-off-by: Christoph Hellwig Reviewed-by: Kanchan Joshi Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240626045950.189758-2-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index 356ca0d3d62f5a..4da003b86a1bbf 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -456,11 +456,11 @@ bool bio_integrity_prep(struct bio *bio) /* * Zero the memory allocated to not leak uninitialized kernel - * memory to disk. For PI this only affects the app tag, but - * for non-integrity metadata it affects the entire metadata - * buffer. + * memory to disk for non-integrity metadata where nothing else + * initializes the memory. */ - gfp |= __GFP_ZERO; + if (bi->csum_type == BLK_INTEGRITY_CSUM_NONE) + gfp |= __GFP_ZERO; } /* Allocate kernel buffer for protection data */ From c096df908393b0b3445f4335dd9bbd9d98252951 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 06:59:35 +0200 Subject: [PATCH 152/209] block: simplify adding the payload in bio_integrity_prep bio_integrity_add_page can add physically contiguous regions of any size, so don't bother chunking up the kmalloced buffer. Signed-off-by: Christoph Hellwig Reviewed-by: Kanchan Joshi Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240626045950.189758-3-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 32 ++++++-------------------------- 1 file changed, 6 insertions(+), 26 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index 4da003b86a1bbf..259acb60c56a3a 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -428,10 +428,8 @@ bool bio_integrity_prep(struct bio *bio) { struct bio_integrity_payload *bip; struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk); + unsigned int len; void *buf; - unsigned long start, end; - unsigned int len, nr_pages; - unsigned int bytes, offset, i; gfp_t gfp = GFP_NOIO; if (!bi) @@ -471,12 +469,7 @@ bool bio_integrity_prep(struct bio *bio) goto err_end_io; } - end = (((unsigned long) buf) + len + PAGE_SIZE - 1) >> PAGE_SHIFT; - start = ((unsigned long) buf) >> PAGE_SHIFT; - nr_pages = end - start; - - /* Allocate bio integrity payload and integrity vectors */ - bip = bio_integrity_alloc(bio, GFP_NOIO, nr_pages); + bip = bio_integrity_alloc(bio, GFP_NOIO, 1); if (IS_ERR(bip)) { printk(KERN_ERR "could not allocate data integrity bioset\n"); kfree(buf); @@ -489,23 +482,10 @@ bool bio_integrity_prep(struct bio *bio) if (bi->csum_type == BLK_INTEGRITY_CSUM_IP) bip->bip_flags |= BIP_IP_CHECKSUM; - /* Map it */ - offset = offset_in_page(buf); - for (i = 0; i < nr_pages && len > 0; i++) { - bytes = PAGE_SIZE - offset; - - if (bytes > len) - bytes = len; - - if (bio_integrity_add_page(bio, virt_to_page(buf), - bytes, offset) < bytes) { - printk(KERN_ERR "could not attach integrity payload\n"); - goto err_end_io; - } - - buf += bytes; - len -= bytes; - offset = 0; + if (bio_integrity_add_page(bio, virt_to_page(buf), len, + offset_in_page(buf)) < len) { + printk(KERN_ERR "could not attach integrity payload\n"); + goto err_end_io; } /* Auto-generate integrity metadata if this is a write */ From dac18fabba59149acec42621b9b603654e9459b2 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 06:59:36 +0200 Subject: [PATCH 153/209] block: remove allocation failure warnings in bio_integrity_prep Allocation failures already leave a stack trace, so don't spew another warning. Signed-off-by: Christoph Hellwig Reviewed-by: Kanchan Joshi Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240626045950.189758-4-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index 259acb60c56a3a..2da7fb222a3582 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -465,13 +465,11 @@ bool bio_integrity_prep(struct bio *bio) len = bio_integrity_bytes(bi, bio_sectors(bio)); buf = kmalloc(len, gfp); if (unlikely(buf == NULL)) { - printk(KERN_ERR "could not allocate integrity buffer\n"); goto err_end_io; } bip = bio_integrity_alloc(bio, GFP_NOIO, 1); if (IS_ERR(bip)) { - printk(KERN_ERR "could not allocate data integrity bioset\n"); kfree(buf); goto err_end_io; } From df3c485e0e60e8ad87f168092f1513a3d621fa4b Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 06:59:37 +0200 Subject: [PATCH 154/209] block: switch on bio operation in bio_integrity_prep Use a single switch to perform read and write specific checks and exit early for other operations instead of having two checks using different predicates. Signed-off-by: Christoph Hellwig Reviewed-by: Kanchan Joshi Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240626045950.189758-5-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index 2da7fb222a3582..2f8a4617003473 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -435,9 +435,6 @@ bool bio_integrity_prep(struct bio *bio) if (!bi) return true; - if (bio_op(bio) != REQ_OP_READ && bio_op(bio) != REQ_OP_WRITE) - return true; - if (!bio_sectors(bio)) return true; @@ -445,10 +442,12 @@ bool bio_integrity_prep(struct bio *bio) if (bio_integrity(bio)) return true; - if (bio_data_dir(bio) == READ) { + switch (bio_op(bio)) { + case REQ_OP_READ: if (bi->flags & BLK_INTEGRITY_NOVERIFY) return true; - } else { + break; + case REQ_OP_WRITE: if (bi->flags & BLK_INTEGRITY_NOGENERATE) return true; @@ -459,6 +458,9 @@ bool bio_integrity_prep(struct bio *bio) */ if (bi->csum_type == BLK_INTEGRITY_CSUM_NONE) gfp |= __GFP_ZERO; + break; + default: + return true; } /* Allocate kernel buffer for protection data */ From d19b46340b3c0ea66bef0f6c58876cc085813ba8 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 26 Jun 2024 06:59:38 +0200 Subject: [PATCH 155/209] block: remove bio_integrity_process Move the bvec interation into the generate/verify helpers to avoid a bit of argument passing churn. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240626045950.189758-6-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 47 +---------------- block/blk.h | 7 +-- block/t10-pi.c | 97 +++++++++++++++++++++++++++-------- include/linux/blk-integrity.h | 9 ---- 4 files changed, 79 insertions(+), 81 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index 2f8a4617003473..ad296849aa2a9a 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -374,44 +374,6 @@ int bio_integrity_map_user(struct bio *bio, void __user *ubuf, ssize_t bytes, } EXPORT_SYMBOL_GPL(bio_integrity_map_user); -/** - * bio_integrity_process - Process integrity metadata for a bio - * @bio: bio to generate/verify integrity metadata for - * @proc_iter: iterator to process - */ -static blk_status_t bio_integrity_process(struct bio *bio, - struct bvec_iter *proc_iter) -{ - struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk); - struct blk_integrity_iter iter; - struct bvec_iter bviter; - struct bio_vec bv; - struct bio_integrity_payload *bip = bio_integrity(bio); - blk_status_t ret = BLK_STS_OK; - - iter.disk_name = bio->bi_bdev->bd_disk->disk_name; - iter.interval = 1 << bi->interval_exp; - iter.seed = proc_iter->bi_sector; - iter.prot_buf = bvec_virt(bip->bip_vec); - - __bio_for_each_segment(bv, bio, bviter, *proc_iter) { - void *kaddr = bvec_kmap_local(&bv); - - iter.data_buf = kaddr; - iter.data_size = bv.bv_len; - if (bio_data_dir(bio) == WRITE) - blk_integrity_generate(&iter, bi); - else - ret = blk_integrity_verify(&iter, bi); - kunmap_local(kaddr); - - if (ret) - break; - - } - return ret; -} - /** * bio_integrity_prep - Prepare bio for integrity I/O * @bio: bio to prepare @@ -490,7 +452,7 @@ bool bio_integrity_prep(struct bio *bio) /* Auto-generate integrity metadata if this is a write */ if (bio_data_dir(bio) == WRITE) - bio_integrity_process(bio, &bio->bi_iter); + blk_integrity_generate(bio); else bip->bio_iter = bio->bi_iter; return true; @@ -516,12 +478,7 @@ static void bio_integrity_verify_fn(struct work_struct *work) container_of(work, struct bio_integrity_payload, bip_work); struct bio *bio = bip->bip_bio; - /* - * At the moment verify is called bio's iterator was advanced - * during split and completion, we need to rewind iterator to - * it's original position. - */ - bio->bi_status = bio_integrity_process(bio, &bip->bio_iter); + blk_integrity_verify(bio); bio_integrity_free(bio); bio_endio(bio); } diff --git a/block/blk.h b/block/blk.h index 95e5a4f81693c4..47dadd2439b1ca 100644 --- a/block/blk.h +++ b/block/blk.h @@ -9,7 +9,6 @@ #include #include "blk-crypto-internal.h" -struct blk_integrity_iter; struct elevator_type; /* Max future timer expiry for timeouts */ @@ -681,10 +680,8 @@ int bdev_open(struct block_device *bdev, blk_mode_t mode, void *holder, const struct blk_holder_ops *hops, struct file *bdev_file); int bdev_permission(dev_t dev, blk_mode_t mode, void *holder); -void blk_integrity_generate(struct blk_integrity_iter *iter, - struct blk_integrity *bi); -blk_status_t blk_integrity_verify(struct blk_integrity_iter *iter, - struct blk_integrity *bi); +void blk_integrity_generate(struct bio *bio); +void blk_integrity_verify(struct bio *bio); void blk_integrity_prepare(struct request *rq); void blk_integrity_complete(struct request *rq, unsigned int nr_bytes); diff --git a/block/t10-pi.c b/block/t10-pi.c index cd7fa60d63ff21..425e2836f3e1d8 100644 --- a/block/t10-pi.c +++ b/block/t10-pi.c @@ -13,6 +13,15 @@ #include #include "blk.h" +struct blk_integrity_iter { + void *prot_buf; + void *data_buf; + sector_t seed; + unsigned int data_size; + unsigned short interval; + const char *disk_name; +}; + static __be16 t10_pi_csum(__be16 csum, void *data, unsigned int len, unsigned char csum_type) { @@ -364,33 +373,77 @@ static void ext_pi_type1_complete(struct request *rq, unsigned int nr_bytes) } } -void blk_integrity_generate(struct blk_integrity_iter *iter, - struct blk_integrity *bi) +void blk_integrity_generate(struct bio *bio) { - switch (bi->csum_type) { - case BLK_INTEGRITY_CSUM_CRC64: - ext_pi_crc64_generate(iter, bi); - break; - case BLK_INTEGRITY_CSUM_CRC: - case BLK_INTEGRITY_CSUM_IP: - t10_pi_generate(iter, bi); - break; - default: - break; + struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk); + struct bio_integrity_payload *bip = bio_integrity(bio); + struct blk_integrity_iter iter; + struct bvec_iter bviter; + struct bio_vec bv; + + iter.disk_name = bio->bi_bdev->bd_disk->disk_name; + iter.interval = 1 << bi->interval_exp; + iter.seed = bio->bi_iter.bi_sector; + iter.prot_buf = bvec_virt(bip->bip_vec); + bio_for_each_segment(bv, bio, bviter) { + void *kaddr = bvec_kmap_local(&bv); + + iter.data_buf = kaddr; + iter.data_size = bv.bv_len; + switch (bi->csum_type) { + case BLK_INTEGRITY_CSUM_CRC64: + ext_pi_crc64_generate(&iter, bi); + break; + case BLK_INTEGRITY_CSUM_CRC: + case BLK_INTEGRITY_CSUM_IP: + t10_pi_generate(&iter, bi); + break; + default: + break; + } + kunmap_local(kaddr); } } -blk_status_t blk_integrity_verify(struct blk_integrity_iter *iter, - struct blk_integrity *bi) +void blk_integrity_verify(struct bio *bio) { - switch (bi->csum_type) { - case BLK_INTEGRITY_CSUM_CRC64: - return ext_pi_crc64_verify(iter, bi); - case BLK_INTEGRITY_CSUM_CRC: - case BLK_INTEGRITY_CSUM_IP: - return t10_pi_verify(iter, bi); - default: - return BLK_STS_OK; + struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk); + struct bio_integrity_payload *bip = bio_integrity(bio); + struct blk_integrity_iter iter; + struct bvec_iter bviter; + struct bio_vec bv; + + /* + * At the moment verify is called bi_iter has been advanced during split + * and completion, so use the copy created during submission here. + */ + iter.disk_name = bio->bi_bdev->bd_disk->disk_name; + iter.interval = 1 << bi->interval_exp; + iter.seed = bip->bio_iter.bi_sector; + iter.prot_buf = bvec_virt(bip->bip_vec); + __bio_for_each_segment(bv, bio, bviter, bip->bio_iter) { + void *kaddr = bvec_kmap_local(&bv); + blk_status_t ret = BLK_STS_OK; + + iter.data_buf = kaddr; + iter.data_size = bv.bv_len; + switch (bi->csum_type) { + case BLK_INTEGRITY_CSUM_CRC64: + ret = ext_pi_crc64_verify(&iter, bi); + break; + case BLK_INTEGRITY_CSUM_CRC: + case BLK_INTEGRITY_CSUM_IP: + ret = t10_pi_verify(&iter, bi); + break; + default: + break; + } + kunmap_local(kaddr); + + if (ret) { + bio->bi_status = ret; + return; + } } } diff --git a/include/linux/blk-integrity.h b/include/linux/blk-integrity.h index c58634924782d0..804f856ed3e571 100644 --- a/include/linux/blk-integrity.h +++ b/include/linux/blk-integrity.h @@ -14,15 +14,6 @@ enum blk_integrity_flags { BLK_INTEGRITY_STACKED = 1 << 4, }; -struct blk_integrity_iter { - void *prot_buf; - void *data_buf; - sector_t seed; - unsigned int data_size; - unsigned short interval; - const char *disk_name; -}; - const char *blk_integrity_profile_name(struct blk_integrity *bi); bool queue_limits_stack_integrity(struct queue_limits *t, struct queue_limits *b); From aa6ff4eb7c10d9a6532db3ea9e78124bf14e70ae Mon Sep 17 00:00:00 2001 From: Dongliang Cui Date: Fri, 14 Jun 2024 15:49:36 +0800 Subject: [PATCH 156/209] block: Add ioprio to block_rq tracepoint MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Sometimes we need to track the processing order of requests with ioprio set. So the ioprio of request can be useful information. Example： block_rq_insert: 8,0 RA 16384 () 6500840 + 32 be,0,6 [binder:815_3] block_rq_issue: 8,0 RA 16384 () 6500840 + 32 be,0,6 [binder:815_3] block_rq_complete: 8,0 RA () 6500840 + 32 be,0,6 [0] Signed-off-by: Dongliang Cui Reviewed-by: Bart Van Assche Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240614074936.113659-1-dongliang.cui@unisoc.com Signed-off-by: Jens Axboe --- include/trace/events/block.h | 41 ++++++++++++++++++++++++++++-------- 1 file changed, 32 insertions(+), 9 deletions(-) diff --git a/include/trace/events/block.h b/include/trace/events/block.h index 0e128ad5146015..1527d5d45e01a4 100644 --- a/include/trace/events/block.h +++ b/include/trace/events/block.h @@ -9,9 +9,17 @@ #include #include #include +#include #define RWBS_LEN 8 +#define IOPRIO_CLASS_STRINGS \ + { IOPRIO_CLASS_NONE, "none" }, \ + { IOPRIO_CLASS_RT, "rt" }, \ + { IOPRIO_CLASS_BE, "be" }, \ + { IOPRIO_CLASS_IDLE, "idle" }, \ + { IOPRIO_CLASS_INVALID, "invalid"} + #ifdef CONFIG_BUFFER_HEAD DECLARE_EVENT_CLASS(block_buffer, @@ -82,6 +90,7 @@ TRACE_EVENT(block_rq_requeue, __field( dev_t, dev ) __field( sector_t, sector ) __field( unsigned int, nr_sector ) + __field( unsigned short, ioprio ) __array( char, rwbs, RWBS_LEN ) __dynamic_array( char, cmd, 1 ) ), @@ -90,16 +99,20 @@ TRACE_EVENT(block_rq_requeue, __entry->dev = rq->q->disk ? disk_devt(rq->q->disk) : 0; __entry->sector = blk_rq_trace_sector(rq); __entry->nr_sector = blk_rq_trace_nr_sectors(rq); + __entry->ioprio = rq->ioprio; blk_fill_rwbs(__entry->rwbs, rq->cmd_flags); __get_str(cmd)[0] = '\0'; ), - TP_printk("%d,%d %s (%s) %llu + %u [%d]", + TP_printk("%d,%d %s (%s) %llu + %u %s,%u,%u [%d]", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->rwbs, __get_str(cmd), - (unsigned long long)__entry->sector, - __entry->nr_sector, 0) + (unsigned long long)__entry->sector, __entry->nr_sector, + __print_symbolic(IOPRIO_PRIO_CLASS(__entry->ioprio), + IOPRIO_CLASS_STRINGS), + IOPRIO_PRIO_HINT(__entry->ioprio), + IOPRIO_PRIO_LEVEL(__entry->ioprio), 0) ); DECLARE_EVENT_CLASS(block_rq_completion, @@ -113,6 +126,7 @@ DECLARE_EVENT_CLASS(block_rq_completion, __field( sector_t, sector ) __field( unsigned int, nr_sector ) __field( int , error ) + __field( unsigned short, ioprio ) __array( char, rwbs, RWBS_LEN ) __dynamic_array( char, cmd, 1 ) ), @@ -122,16 +136,20 @@ DECLARE_EVENT_CLASS(block_rq_completion, __entry->sector = blk_rq_pos(rq); __entry->nr_sector = nr_bytes >> 9; __entry->error = blk_status_to_errno(error); + __entry->ioprio = rq->ioprio; blk_fill_rwbs(__entry->rwbs, rq->cmd_flags); __get_str(cmd)[0] = '\0'; ), - TP_printk("%d,%d %s (%s) %llu + %u [%d]", + TP_printk("%d,%d %s (%s) %llu + %u %s,%u,%u [%d]", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->rwbs, __get_str(cmd), - (unsigned long long)__entry->sector, - __entry->nr_sector, __entry->error) + (unsigned long long)__entry->sector, __entry->nr_sector, + __print_symbolic(IOPRIO_PRIO_CLASS(__entry->ioprio), + IOPRIO_CLASS_STRINGS), + IOPRIO_PRIO_HINT(__entry->ioprio), + IOPRIO_PRIO_LEVEL(__entry->ioprio), __entry->error) ); /** @@ -180,6 +198,7 @@ DECLARE_EVENT_CLASS(block_rq, __field( sector_t, sector ) __field( unsigned int, nr_sector ) __field( unsigned int, bytes ) + __field( unsigned short, ioprio ) __array( char, rwbs, RWBS_LEN ) __array( char, comm, TASK_COMM_LEN ) __dynamic_array( char, cmd, 1 ) @@ -190,17 +209,21 @@ DECLARE_EVENT_CLASS(block_rq, __entry->sector = blk_rq_trace_sector(rq); __entry->nr_sector = blk_rq_trace_nr_sectors(rq); __entry->bytes = blk_rq_bytes(rq); + __entry->ioprio = rq->ioprio; blk_fill_rwbs(__entry->rwbs, rq->cmd_flags); __get_str(cmd)[0] = '\0'; memcpy(__entry->comm, current->comm, TASK_COMM_LEN); ), - TP_printk("%d,%d %s %u (%s) %llu + %u [%s]", + TP_printk("%d,%d %s %u (%s) %llu + %u %s,%u,%u [%s]", MAJOR(__entry->dev), MINOR(__entry->dev), __entry->rwbs, __entry->bytes, __get_str(cmd), - (unsigned long long)__entry->sector, - __entry->nr_sector, __entry->comm) + (unsigned long long)__entry->sector, __entry->nr_sector, + __print_symbolic(IOPRIO_PRIO_CLASS(__entry->ioprio), + IOPRIO_CLASS_STRINGS), + IOPRIO_PRIO_HINT(__entry->ioprio), + IOPRIO_PRIO_LEVEL(__entry->ioprio), __entry->comm) ); /** From 0676c434a99be42f3bacca4adfd27df65edbf903 Mon Sep 17 00:00:00 2001 From: Ming Lei Date: Thu, 20 Jun 2024 11:06:31 +0800 Subject: [PATCH 157/209] block: check bio alignment in blk_mq_submit_bio IO logical block size is one fundamental queue limit, and every IO has to be aligned with logical block size because our bio split can't deal with unaligned bio. The check has to be done with queue usage counter grabbed because device reconfiguration may change logical block size, and we can prevent the reconfiguration from happening by holding queue usage counter. logical_block_size stays in the 1st cache line of queue_limits, and this cache line is always fetched in fast path via bio_may_exceed_limits(), so IO perf won't be affected by this check. Cc: Yi Zhang Cc: Christoph Hellwig Cc: Ye Bin Cc: stable@vger.kernel.org Signed-off-by: Ming Lei Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240620030631.3114026-1-ming.lei@redhat.com Signed-off-by: Jens Axboe --- block/blk-mq.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) diff --git a/block/blk-mq.c b/block/blk-mq.c index 47fe9d19b8f109..fec2dea5c6e885 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -2909,6 +2909,17 @@ static void blk_mq_use_cached_rq(struct request *rq, struct blk_plug *plug, INIT_LIST_HEAD(&rq->queuelist); } +static bool bio_unaligned(const struct bio *bio, struct request_queue *q) +{ + unsigned int bs_mask = queue_logical_block_size(q) - 1; + + /* .bi_sector of any zero sized bio need to be initialized */ + if ((bio->bi_iter.bi_size & bs_mask) || + ((bio->bi_iter.bi_sector << SECTOR_SHIFT) & bs_mask)) + return true; + return false; +} + /** * blk_mq_submit_bio - Create and send a request to block device. * @bio: Bio pointer. @@ -2961,6 +2972,15 @@ void blk_mq_submit_bio(struct bio *bio) return; } + /* + * Device reconfiguration may change logical block size, so alignment + * check has to be done with queue usage counter held + */ + if (unlikely(bio_unaligned(bio, q))) { + bio_io_error(bio); + goto queue_exit; + } + if (unlikely(bio_may_exceed_limits(bio, &q->limits))) { bio = __bio_split_to_limits(bio, &q->limits, &nr_segs); if (!bio) From 667ea36378cf7f669044b27871c496e1559c872a Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 27 Jun 2024 14:49:11 +0200 Subject: [PATCH 158/209] loop: don't set QUEUE_FLAG_NOMERGES QUEUE_FLAG_NOMERGES isn't really a driver interface, but a user tunable. There also isn't any good reason to set it in the loop driver. The original commit adding it (5b5e20f421c0b6d "block: loop: set QUEUE_FLAG_NOMERGES for request queue of loop") claims that "It doesn't make sense to enable merge because the I/O submitted to backing file is handled page by page." which of course isn't true for multi-page bvec now, and it never has been for direct I/O, for which commit 40326d8a33d ("block/loop: allow request merge for directio mode") alredy disabled the nomerges flag. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Link: https://lore.kernel.org/r/20240627124926.512662-2-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/loop.c | 15 ++------------- 1 file changed, 2 insertions(+), 13 deletions(-) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index cbda971d22b40b..4f0d96876b1f5d 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -211,13 +211,10 @@ static void __loop_update_dio(struct loop_device *lo, bool dio) if (lo->lo_state == Lo_bound) blk_mq_freeze_queue(lo->lo_queue); lo->use_dio = use_dio; - if (use_dio) { - blk_queue_flag_clear(QUEUE_FLAG_NOMERGES, lo->lo_queue); + if (use_dio) lo->lo_flags |= LO_FLAGS_DIRECT_IO; - } else { - blk_queue_flag_set(QUEUE_FLAG_NOMERGES, lo->lo_queue); + else lo->lo_flags &= ~LO_FLAGS_DIRECT_IO; - } if (lo->lo_state == Lo_bound) blk_mq_unfreeze_queue(lo->lo_queue); } @@ -2030,14 +2027,6 @@ static int loop_add(int i) } lo->lo_queue = lo->lo_disk->queue; - /* - * By default, we do buffer IO, so it doesn't make sense to enable - * merge because the I/O submitted to backing file is handled page by - * page. For directio mode, merge does help to dispatch bigger request - * to underlayer disk. We will enable merge once directio is enabled. - */ - blk_queue_flag_set(QUEUE_FLAG_NOMERGES, lo->lo_queue); - /* * Disable partition scanning by default. The in-kernel partition * scanning can be requested individually per-device during its From aa57abe6a7f91fafe53fb98d0f1e74db951bce24 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 27 Jun 2024 14:49:12 +0200 Subject: [PATCH 159/209] megaraid_sas: don't set QUEUE_FLAG_NOMERGES Setting QUEUE_FLAG_NOMERGES was added in commit 15dd03811d99dcf ("scsi: megaraid_sas: NVME Interface detection and prop settings") without any explanation. Drivers should second guess the block layer merge decisions, so remove the flag. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Link: https://lore.kernel.org/r/20240627124926.512662-3-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/megaraid/megaraid_sas_base.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c index 88acefbf9aeaba..6c79c350a4d5ba 100644 --- a/drivers/scsi/megaraid/megaraid_sas_base.c +++ b/drivers/scsi/megaraid/megaraid_sas_base.c @@ -1981,8 +1981,6 @@ megasas_set_nvme_device_properties(struct scsi_device *sdev, lim->max_hw_sectors = max_io_size / 512; lim->virt_boundary_mask = mr_nvme_pg_size - 1; - - blk_queue_flag_set(QUEUE_FLAG_NOMERGES, sdev->request_queue); } /* From 8b77f23fadcbb030a898f168bebe74f465e5d5a2 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 27 Jun 2024 14:49:13 +0200 Subject: [PATCH 160/209] mpt3sas_scsih: don't set QUEUE_FLAG_NOMERGES Setting QUEUE_FLAG_NOMERGES was added in commit d1b01d14b7ba ("scsi: mpt3sas: Set NVMe device queue depth as 128") without any explanation. Drivers should second guess the block layer merge decisions, so remove the flag. Signed-off-by: Christoph Hellwig Reviewed-by: Bart Van Assche Link: https://lore.kernel.org/r/20240627124926.512662-4-hch@lst.de Signed-off-by: Jens Axboe --- drivers/scsi/mpt3sas/mpt3sas_scsih.c | 6 ------ 1 file changed, 6 deletions(-) diff --git a/drivers/scsi/mpt3sas/mpt3sas_scsih.c b/drivers/scsi/mpt3sas/mpt3sas_scsih.c index 12d08d8ba5382d..b050aedc9d4334 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_scsih.c +++ b/drivers/scsi/mpt3sas/mpt3sas_scsih.c @@ -2680,12 +2680,6 @@ scsih_device_configure(struct scsi_device *sdev, struct queue_limits *lim) pcie_device_put(pcie_device); spin_unlock_irqrestore(&ioc->pcie_device_lock, flags); mpt3sas_scsih_change_queue_depth(sdev, qdepth); - /* Enable QUEUE_FLAG_NOMERGES flag, so that IOs won't be - ** merged and can eliminate holes created during merging - ** operation. - **/ - blk_queue_flag_set(QUEUE_FLAG_NOMERGES, - sdev->request_queue); lim->virt_boundary_mask = ioc->page_size - 1; return 0; } From 40988f15907baee227d3b83bd4d8f8fdfeb95dd3 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 27 Jun 2024 14:49:14 +0200 Subject: [PATCH 161/209] rnbd: don't set QUEUE_FLAG_SAME_COMP QUEUE_FLAG_SAME_COMP is already set by default. Signed-off-by: Christoph Hellwig Acked-by: Jack Wang Reviewed-by: Bart Van Assche Link: https://lore.kernel.org/r/20240627124926.512662-5-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/rnbd/rnbd-clt.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/block/rnbd/rnbd-clt.c b/drivers/block/rnbd/rnbd-clt.c index 4918b0f68b46cd..0e3773fe479706 100644 --- a/drivers/block/rnbd/rnbd-clt.c +++ b/drivers/block/rnbd/rnbd-clt.c @@ -1397,7 +1397,6 @@ static int rnbd_client_setup_device(struct rnbd_clt_dev *dev, dev->queue = dev->gd->queue; rnbd_init_mq_hw_queues(dev); - blk_queue_flag_set(QUEUE_FLAG_SAME_COMP, dev->queue); blk_queue_flag_set(QUEUE_FLAG_SAME_FORCE, dev->queue); return rnbd_clt_setup_gen_disk(dev, rsp, idx); } From caffa7cdce47718a0c2e3195c9a1bcf786d655a4 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 27 Jun 2024 14:49:15 +0200 Subject: [PATCH 162/209] rnbd-cnt: don't set QUEUE_FLAG_SAME_FORCE QUEUE_FLAG_SAME_FORCE has been set by rnbd-cnt since the initial merge. There is no good reason for a driver to force exact core delivery, which is tunable for very specific workloads and not a driver setting. Signed-off-by: Christoph Hellwig Acked-by: Jack Wang Reviewed-by: Bart Van Assche Link: https://lore.kernel.org/r/20240627124926.512662-6-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/rnbd/rnbd-clt.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/block/rnbd/rnbd-clt.c b/drivers/block/rnbd/rnbd-clt.c index 0e3773fe479706..c34695d2eea7fe 100644 --- a/drivers/block/rnbd/rnbd-clt.c +++ b/drivers/block/rnbd/rnbd-clt.c @@ -1397,7 +1397,6 @@ static int rnbd_client_setup_device(struct rnbd_clt_dev *dev, dev->queue = dev->gd->queue; rnbd_init_mq_hw_queues(dev); - blk_queue_flag_set(QUEUE_FLAG_SAME_FORCE, dev->queue); return rnbd_clt_setup_gen_disk(dev, rsp, idx); } From 5b026e34120766408e76ba19a0e33a9dc996f9f0 Mon Sep 17 00:00:00 2001 From: Andreas Hindborg Date: Fri, 28 Jun 2024 11:11:52 +0200 Subject: [PATCH 163/209] rust: block: fix generated bindings after refactoring of features Block device features and flags were refactored from `enum` to `#define`. This broke Rust binding generation. This patch fixes the binding generation. Fixes: fcf865e357f8 ("block: convert features and flags to __bitwise types") Signed-off-by: Andreas Hindborg Acked-by: Miguel Ojeda Link: https://lore.kernel.org/r/20240628091152.2185241-1-nmi@metaspace.dk Signed-off-by: Jens Axboe --- rust/bindings/bindings_helper.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/rust/bindings/bindings_helper.h b/rust/bindings/bindings_helper.h index 84f601d7068e37..6deee85a29c896 100644 --- a/rust/bindings/bindings_helper.h +++ b/rust/bindings/bindings_helper.h @@ -9,6 +9,7 @@ #include #include #include +#include #include #include #include @@ -28,3 +29,4 @@ const gfp_t RUST_CONST_HELPER_GFP_KERNEL = GFP_KERNEL; const gfp_t RUST_CONST_HELPER_GFP_KERNEL_ACCOUNT = GFP_KERNEL_ACCOUNT; const gfp_t RUST_CONST_HELPER_GFP_NOWAIT = GFP_NOWAIT; const gfp_t RUST_CONST_HELPER___GFP_ZERO = __GFP_ZERO; +const blk_features_t RUST_CONST_HELPER_BLK_FEAT_ROTATIONAL = BLK_FEAT_ROTATIONAL; From 3991657ae7074c3c497bf095093178bed37ea1b4 Mon Sep 17 00:00:00 2001 From: Anuj Gupta Date: Wed, 26 Jun 2024 15:36:52 +0530 Subject: [PATCH 164/209] block: set bip_vcnt correctly Set the bip_vcnt correctly in bio_integrity_init_user and bio_integrity_copy_user. If the bio gets split at a later point, this value is required to set the right bip_vcnt in the cloned bio. Signed-off-by: Anuj Gupta Signed-off-by: Kanchan Joshi Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240626100700.3629-3-anuj20.g@samsung.com Signed-off-by: Jens Axboe --- block/bio-integrity.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index ad296849aa2a9a..eb3d7bbe1fe83a 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -254,6 +254,7 @@ static int bio_integrity_copy_user(struct bio *bio, struct bio_vec *bvec, bip->bip_flags |= BIP_INTEGRITY_USER | BIP_COPY_USER; bip->bip_iter.bi_sector = seed; + bip->bip_vcnt = nr_vecs; return 0; free_bip: bio_integrity_free(bio); @@ -275,6 +276,7 @@ static int bio_integrity_init_user(struct bio *bio, struct bio_vec *bvec, bip->bip_flags |= BIP_INTEGRITY_USER; bip->bip_iter.bi_sector = seed; bip->bip_iter.bi_size = len; + bip->bip_vcnt = nr_vecs; return 0; } From 1beabab88ecee0698ecee7b54afa9cce7046ef96 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Tue, 18 Jun 2024 14:21:08 +0800 Subject: [PATCH 165/209] blk-throttle: fix lower control under super low iops limit User will configure allowed iops limit in 1s, and calculate_io_allowed() will calculate allowed iops in the slice by: limit * HZ / throtl_slice However, if limit is quite low, the result can be 0, then allowed IO in the slice is 0, this will cause missing dispatch and control will be lower than limit. For example, set iops_limit to 5 with HD disk, and test will found that iops will be 3. This is usually not a big deal, because user will unlikely to configure such low iops limit, however, this is still a problem in the extreme scene. Fix the problem by making sure the wait time calculated by tg_within_iops_limit() should allow at least one IO to be dispatched. Signed-off-by: Yu Kuai Acked-by: Tejun Heo Link: https://lore.kernel.org/r/20240618062108.3680835-1-yukuai1@huaweicloud.com Signed-off-by: Jens Axboe --- block/blk-throttle.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/block/blk-throttle.c b/block/blk-throttle.c index c1bf73f8c75d95..dc6140fa3de061 100644 --- a/block/blk-throttle.c +++ b/block/blk-throttle.c @@ -704,6 +704,9 @@ static unsigned long tg_within_iops_limit(struct throtl_grp *tg, struct bio *bio /* Calc approx time to dispatch */ jiffy_wait = jiffy_elapsed_rnd - jiffy_elapsed; + + /* make sure at least one io can be dispatched after waiting */ + jiffy_wait = max(jiffy_wait, HZ / iops_limit + 1); return jiffy_wait; } From 5476394aa9f27d670dd2bac426fdb6ac12b12cb3 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 27 Jun 2024 13:14:01 +0200 Subject: [PATCH 166/209] block: simplify queue_logical_block_size queue_logical_block_size is never called with a 0 queue, and the logical_block_size field in queue_limits is always initialized for a live queue. Signed-off-by: Christoph Hellwig Reviewed-by: John Garry Link: https://lore.kernel.org/r/20240627111407.476276-2-hch@lst.de Signed-off-by: Jens Axboe --- include/linux/blkdev.h | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 53c41ef4222c3d..4d0d4b83bc740f 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -1227,12 +1227,7 @@ static inline unsigned int bdev_max_segments(struct block_device *bdev) static inline unsigned queue_logical_block_size(const struct request_queue *q) { - int retval = 512; - - if (q && q->limits.logical_block_size) - retval = q->limits.logical_block_size; - - return retval; + return q->limits.logical_block_size; } static inline unsigned int bdev_logical_block_size(struct block_device *bdev) From 319e8cfdf3caf41b98f50ef13542a35acd897bb6 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 27 Jun 2024 13:14:02 +0200 Subject: [PATCH 167/209] block: add helper macros to de-duplicate the queue sysfs attributes A lof the code to implement the queue sysfs attributes is repetitive. Add a few macros to generate the common cases. Signed-off-by: Christoph Hellwig Reviewed-by: John Garry Link: https://lore.kernel.org/r/20240627111407.476276-3-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-sysfs.c | 255 +++++++++++++++------------------------------- 1 file changed, 82 insertions(+), 173 deletions(-) diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index 2e6d9b918127fe..a769fb441b58da 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -100,103 +100,65 @@ queue_ra_store(struct request_queue *q, const char *page, size_t count) return ret; } -static ssize_t queue_max_sectors_show(struct request_queue *q, char *page) -{ - int max_sectors_kb = queue_max_sectors(q) >> 1; - - return queue_var_show(max_sectors_kb, page); -} - -static ssize_t queue_max_segments_show(struct request_queue *q, char *page) -{ - return queue_var_show(queue_max_segments(q), page); -} - -static ssize_t queue_max_discard_segments_show(struct request_queue *q, - char *page) -{ - return queue_var_show(queue_max_discard_segments(q), page); -} - -static ssize_t queue_atomic_write_max_bytes_show(struct request_queue *q, - char *page) -{ - return queue_var_show(queue_atomic_write_max_bytes(q), page); -} - -static ssize_t queue_atomic_write_boundary_show(struct request_queue *q, - char *page) -{ - return queue_var_show(queue_atomic_write_boundary_bytes(q), page); -} - -static ssize_t queue_atomic_write_unit_min_show(struct request_queue *q, - char *page) -{ - return queue_var_show(queue_atomic_write_unit_min_bytes(q), page); -} - -static ssize_t queue_atomic_write_unit_max_show(struct request_queue *q, - char *page) -{ - return queue_var_show(queue_atomic_write_unit_max_bytes(q), page); -} - -static ssize_t queue_max_integrity_segments_show(struct request_queue *q, char *page) -{ - return queue_var_show(q->limits.max_integrity_segments, page); -} - -static ssize_t queue_max_segment_size_show(struct request_queue *q, char *page) -{ - return queue_var_show(queue_max_segment_size(q), page); -} - -static ssize_t queue_logical_block_size_show(struct request_queue *q, char *page) -{ - return queue_var_show(queue_logical_block_size(q), page); -} - -static ssize_t queue_physical_block_size_show(struct request_queue *q, char *page) -{ - return queue_var_show(queue_physical_block_size(q), page); -} - -static ssize_t queue_chunk_sectors_show(struct request_queue *q, char *page) -{ - return queue_var_show(q->limits.chunk_sectors, page); -} - -static ssize_t queue_io_min_show(struct request_queue *q, char *page) -{ - return queue_var_show(queue_io_min(q), page); -} - -static ssize_t queue_io_opt_show(struct request_queue *q, char *page) -{ - return queue_var_show(queue_io_opt(q), page); -} - -static ssize_t queue_discard_granularity_show(struct request_queue *q, char *page) -{ - return queue_var_show(q->limits.discard_granularity, page); +#define QUEUE_SYSFS_LIMIT_SHOW(_field) \ +static ssize_t queue_##_field##_show(struct request_queue *q, char *page) \ +{ \ + return queue_var_show(q->limits._field, page); \ +} + +QUEUE_SYSFS_LIMIT_SHOW(max_segments) +QUEUE_SYSFS_LIMIT_SHOW(max_discard_segments) +QUEUE_SYSFS_LIMIT_SHOW(max_integrity_segments) +QUEUE_SYSFS_LIMIT_SHOW(max_segment_size) +QUEUE_SYSFS_LIMIT_SHOW(logical_block_size) +QUEUE_SYSFS_LIMIT_SHOW(physical_block_size) +QUEUE_SYSFS_LIMIT_SHOW(chunk_sectors) +QUEUE_SYSFS_LIMIT_SHOW(io_min) +QUEUE_SYSFS_LIMIT_SHOW(io_opt) +QUEUE_SYSFS_LIMIT_SHOW(discard_granularity) +QUEUE_SYSFS_LIMIT_SHOW(zone_write_granularity) +QUEUE_SYSFS_LIMIT_SHOW(virt_boundary_mask) +QUEUE_SYSFS_LIMIT_SHOW(dma_alignment) +QUEUE_SYSFS_LIMIT_SHOW(max_open_zones) +QUEUE_SYSFS_LIMIT_SHOW(max_active_zones) +QUEUE_SYSFS_LIMIT_SHOW(atomic_write_unit_min) +QUEUE_SYSFS_LIMIT_SHOW(atomic_write_unit_max) + +#define QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(_field) \ +static ssize_t queue_##_field##_show(struct request_queue *q, char *page) \ +{ \ + return sprintf(page, "%llu\n", \ + (unsigned long long)q->limits._field << SECTOR_SHIFT); \ +} + +QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(max_discard_sectors) +QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(max_hw_discard_sectors) +QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(max_write_zeroes_sectors) +QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(atomic_write_max_sectors) +QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(atomic_write_boundary_sectors) + +#define QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_KB(_field) \ +static ssize_t queue_##_field##_show(struct request_queue *q, char *page) \ +{ \ + return queue_var_show(q->limits._field >> 1, page); \ +} + +QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_KB(max_sectors) +QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_KB(max_hw_sectors) + +#define QUEUE_SYSFS_SHOW_CONST(_name, _val) \ +static ssize_t queue_##_name##_show(struct request_queue *q, char *page) \ +{ \ + return sprintf(page, "%d\n", _val); \ } -static ssize_t queue_discard_max_hw_show(struct request_queue *q, char *page) -{ - - return sprintf(page, "%llu\n", - (unsigned long long)q->limits.max_hw_discard_sectors << 9); -} +/* deprecated fields */ +QUEUE_SYSFS_SHOW_CONST(discard_zeroes_data, 0) +QUEUE_SYSFS_SHOW_CONST(write_same_max, 0) +QUEUE_SYSFS_SHOW_CONST(poll_delay, -1) -static ssize_t queue_discard_max_show(struct request_queue *q, char *page) -{ - return sprintf(page, "%llu\n", - (unsigned long long)q->limits.max_discard_sectors << 9); -} - -static ssize_t queue_discard_max_store(struct request_queue *q, - const char *page, size_t count) +static ssize_t queue_max_discard_sectors_store(struct request_queue *q, + const char *page, size_t count) { unsigned long max_discard_bytes; struct queue_limits lim; @@ -221,28 +183,11 @@ static ssize_t queue_discard_max_store(struct request_queue *q, return ret; } -static ssize_t queue_discard_zeroes_data_show(struct request_queue *q, char *page) -{ - return queue_var_show(0, page); -} - -static ssize_t queue_write_same_max_show(struct request_queue *q, char *page) -{ - return queue_var_show(0, page); -} - -static ssize_t queue_write_zeroes_max_show(struct request_queue *q, char *page) -{ - return sprintf(page, "%llu\n", - (unsigned long long)q->limits.max_write_zeroes_sectors << 9); -} - -static ssize_t queue_zone_write_granularity_show(struct request_queue *q, - char *page) -{ - return queue_var_show(queue_zone_write_granularity(q), page); -} - +/* + * For zone append queue_max_zone_append_sectors does not just return the + * underlying queue limits, but actually contains a calculation. Because of + * that we can't simply use QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES here. + */ static ssize_t queue_zone_append_max_show(struct request_queue *q, char *page) { unsigned long long max_sectors = queue_max_zone_append_sectors(q); @@ -270,23 +215,6 @@ queue_max_sectors_store(struct request_queue *q, const char *page, size_t count) return ret; } -static ssize_t queue_max_hw_sectors_show(struct request_queue *q, char *page) -{ - int max_hw_sectors_kb = queue_max_hw_sectors(q) >> 1; - - return queue_var_show(max_hw_sectors_kb, page); -} - -static ssize_t queue_virt_boundary_mask_show(struct request_queue *q, char *page) -{ - return queue_var_show(q->limits.virt_boundary_mask, page); -} - -static ssize_t queue_dma_alignment_show(struct request_queue *q, char *page) -{ - return queue_var_show(queue_dma_alignment(q), page); -} - static ssize_t queue_feature_store(struct request_queue *q, const char *page, size_t count, blk_features_t feature) { @@ -325,6 +253,16 @@ QUEUE_SYSFS_FEATURE(add_random, BLK_FEAT_ADD_RANDOM) QUEUE_SYSFS_FEATURE(iostats, BLK_FEAT_IO_STAT) QUEUE_SYSFS_FEATURE(stable_writes, BLK_FEAT_STABLE_WRITES); +#define QUEUE_SYSFS_FEATURE_SHOW(_name, _feature) \ +static ssize_t queue_##_name##_show(struct request_queue *q, char *page) \ +{ \ + return sprintf(page, "%u\n", !!(q->limits.features & _feature)); \ +} + +QUEUE_SYSFS_FEATURE_SHOW(poll, BLK_FEAT_POLL); +QUEUE_SYSFS_FEATURE_SHOW(fua, BLK_FEAT_FUA); +QUEUE_SYSFS_FEATURE_SHOW(dax, BLK_FEAT_DAX); + static ssize_t queue_zoned_show(struct request_queue *q, char *page) { if (blk_queue_is_zoned(q)) @@ -337,16 +275,6 @@ static ssize_t queue_nr_zones_show(struct request_queue *q, char *page) return queue_var_show(disk_nr_zones(q->disk), page); } -static ssize_t queue_max_open_zones_show(struct request_queue *q, char *page) -{ - return queue_var_show(bdev_max_open_zones(q->disk->part0), page); -} - -static ssize_t queue_max_active_zones_show(struct request_queue *q, char *page) -{ - return queue_var_show(bdev_max_active_zones(q->disk->part0), page); -} - static ssize_t queue_nomerges_show(struct request_queue *q, char *page) { return queue_var_show((blk_queue_nomerges(q) << 1) | @@ -405,22 +333,12 @@ queue_rq_affinity_store(struct request_queue *q, const char *page, size_t count) return ret; } -static ssize_t queue_poll_delay_show(struct request_queue *q, char *page) -{ - return sprintf(page, "%d\n", -1); -} - static ssize_t queue_poll_delay_store(struct request_queue *q, const char *page, size_t count) { return count; } -static ssize_t queue_poll_show(struct request_queue *q, char *page) -{ - return queue_var_show(!!(q->limits.features & BLK_FEAT_POLL), page); -} - static ssize_t queue_poll_store(struct request_queue *q, const char *page, size_t count) { @@ -485,16 +403,6 @@ static ssize_t queue_wc_store(struct request_queue *q, const char *page, return count; } -static ssize_t queue_fua_show(struct request_queue *q, char *page) -{ - return sprintf(page, "%u\n", !!(q->limits.features & BLK_FEAT_FUA)); -} - -static ssize_t queue_dax_show(struct request_queue *q, char *page) -{ - return queue_var_show(!!blk_queue_dax(q), page); -} - #define QUEUE_RO_ENTRY(_prefix, _name) \ static struct queue_sysfs_entry _prefix##_entry = { \ .attr = { .name = _name, .mode = 0444 }, \ @@ -525,17 +433,18 @@ QUEUE_RO_ENTRY(queue_io_opt, "optimal_io_size"); QUEUE_RO_ENTRY(queue_max_discard_segments, "max_discard_segments"); QUEUE_RO_ENTRY(queue_discard_granularity, "discard_granularity"); -QUEUE_RO_ENTRY(queue_discard_max_hw, "discard_max_hw_bytes"); -QUEUE_RW_ENTRY(queue_discard_max, "discard_max_bytes"); +QUEUE_RO_ENTRY(queue_max_hw_discard_sectors, "discard_max_hw_bytes"); +QUEUE_RW_ENTRY(queue_max_discard_sectors, "discard_max_bytes"); QUEUE_RO_ENTRY(queue_discard_zeroes_data, "discard_zeroes_data"); -QUEUE_RO_ENTRY(queue_atomic_write_max_bytes, "atomic_write_max_bytes"); -QUEUE_RO_ENTRY(queue_atomic_write_boundary, "atomic_write_boundary_bytes"); +QUEUE_RO_ENTRY(queue_atomic_write_max_sectors, "atomic_write_max_bytes"); +QUEUE_RO_ENTRY(queue_atomic_write_boundary_sectors, + "atomic_write_boundary_bytes"); QUEUE_RO_ENTRY(queue_atomic_write_unit_max, "atomic_write_unit_max_bytes"); QUEUE_RO_ENTRY(queue_atomic_write_unit_min, "atomic_write_unit_min_bytes"); QUEUE_RO_ENTRY(queue_write_same_max, "write_same_max_bytes"); -QUEUE_RO_ENTRY(queue_write_zeroes_max, "write_zeroes_max_bytes"); +QUEUE_RO_ENTRY(queue_max_write_zeroes_sectors, "write_zeroes_max_bytes"); QUEUE_RO_ENTRY(queue_zone_append_max, "zone_append_max_bytes"); QUEUE_RO_ENTRY(queue_zone_write_granularity, "zone_write_granularity"); @@ -652,15 +561,15 @@ static struct attribute *queue_attrs[] = { &queue_io_min_entry.attr, &queue_io_opt_entry.attr, &queue_discard_granularity_entry.attr, - &queue_discard_max_entry.attr, - &queue_discard_max_hw_entry.attr, + &queue_max_discard_sectors_entry.attr, + &queue_max_hw_discard_sectors_entry.attr, &queue_discard_zeroes_data_entry.attr, - &queue_atomic_write_max_bytes_entry.attr, - &queue_atomic_write_boundary_entry.attr, + &queue_atomic_write_max_sectors_entry.attr, + &queue_atomic_write_boundary_sectors_entry.attr, &queue_atomic_write_unit_min_entry.attr, &queue_atomic_write_unit_max_entry.attr, &queue_write_same_max_entry.attr, - &queue_write_zeroes_max_entry.attr, + &queue_max_write_zeroes_sectors_entry.attr, &queue_zone_append_max_entry.attr, &queue_zone_write_granularity_entry.attr, &queue_rotational_entry.attr, From 62e35f942231e372f8e465d8484de66a60221226 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 27 Jun 2024 13:14:03 +0200 Subject: [PATCH 168/209] block: pass a gendisk to the queue_sysfs_entry methods The kobject for the queue entries is embedded into a struct gendisk. Pass it to the sysfs methods instead of the request_queue derived from it. Signed-off-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240627111407.476276-4-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-sysfs.c | 180 +++++++++++++++++++++++----------------------- block/elevator.c | 9 +-- block/elevator.h | 4 +- 3 files changed, 96 insertions(+), 97 deletions(-) diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c index a769fb441b58da..60116d13cb8043 100644 --- a/block/blk-sysfs.c +++ b/block/blk-sysfs.c @@ -22,8 +22,8 @@ struct queue_sysfs_entry { struct attribute attr; - ssize_t (*show)(struct request_queue *, char *); - ssize_t (*store)(struct request_queue *, const char *, size_t); + ssize_t (*show)(struct gendisk *disk, char *page); + ssize_t (*store)(struct gendisk *disk, const char *page, size_t count); }; static ssize_t @@ -47,18 +47,18 @@ queue_var_store(unsigned long *var, const char *page, size_t count) return count; } -static ssize_t queue_requests_show(struct request_queue *q, char *page) +static ssize_t queue_requests_show(struct gendisk *disk, char *page) { - return queue_var_show(q->nr_requests, page); + return queue_var_show(disk->queue->nr_requests, page); } static ssize_t -queue_requests_store(struct request_queue *q, const char *page, size_t count) +queue_requests_store(struct gendisk *disk, const char *page, size_t count) { unsigned long nr; int ret, err; - if (!queue_is_mq(q)) + if (!queue_is_mq(disk->queue)) return -EINVAL; ret = queue_var_store(&nr, page, count); @@ -68,42 +68,35 @@ queue_requests_store(struct request_queue *q, const char *page, size_t count) if (nr < BLKDEV_MIN_RQ) nr = BLKDEV_MIN_RQ; - err = blk_mq_update_nr_requests(q, nr); + err = blk_mq_update_nr_requests(disk->queue, nr); if (err) return err; return ret; } -static ssize_t queue_ra_show(struct request_queue *q, char *page) +static ssize_t queue_ra_show(struct gendisk *disk, char *page) { - unsigned long ra_kb; - - if (!q->disk) - return -EINVAL; - ra_kb = q->disk->bdi->ra_pages << (PAGE_SHIFT - 10); - return queue_var_show(ra_kb, page); + return queue_var_show(disk->bdi->ra_pages << (PAGE_SHIFT - 10), page); } static ssize_t -queue_ra_store(struct request_queue *q, const char *page, size_t count) +queue_ra_store(struct gendisk *disk, const char *page, size_t count) { unsigned long ra_kb; ssize_t ret; - if (!q->disk) - return -EINVAL; ret = queue_var_store(&ra_kb, page, count); if (ret < 0) return ret; - q->disk->bdi->ra_pages = ra_kb >> (PAGE_SHIFT - 10); + disk->bdi->ra_pages = ra_kb >> (PAGE_SHIFT - 10); return ret; } #define QUEUE_SYSFS_LIMIT_SHOW(_field) \ -static ssize_t queue_##_field##_show(struct request_queue *q, char *page) \ +static ssize_t queue_##_field##_show(struct gendisk *disk, char *page) \ { \ - return queue_var_show(q->limits._field, page); \ + return queue_var_show(disk->queue->limits._field, page); \ } QUEUE_SYSFS_LIMIT_SHOW(max_segments) @@ -125,10 +118,11 @@ QUEUE_SYSFS_LIMIT_SHOW(atomic_write_unit_min) QUEUE_SYSFS_LIMIT_SHOW(atomic_write_unit_max) #define QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(_field) \ -static ssize_t queue_##_field##_show(struct request_queue *q, char *page) \ +static ssize_t queue_##_field##_show(struct gendisk *disk, char *page) \ { \ return sprintf(page, "%llu\n", \ - (unsigned long long)q->limits._field << SECTOR_SHIFT); \ + (unsigned long long)disk->queue->limits._field << \ + SECTOR_SHIFT); \ } QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(max_discard_sectors) @@ -138,16 +132,16 @@ QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(atomic_write_max_sectors) QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES(atomic_write_boundary_sectors) #define QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_KB(_field) \ -static ssize_t queue_##_field##_show(struct request_queue *q, char *page) \ +static ssize_t queue_##_field##_show(struct gendisk *disk, char *page) \ { \ - return queue_var_show(q->limits._field >> 1, page); \ + return queue_var_show(disk->queue->limits._field >> 1, page); \ } QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_KB(max_sectors) QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_KB(max_hw_sectors) #define QUEUE_SYSFS_SHOW_CONST(_name, _val) \ -static ssize_t queue_##_name##_show(struct request_queue *q, char *page) \ +static ssize_t queue_##_name##_show(struct gendisk *disk, char *page) \ { \ return sprintf(page, "%d\n", _val); \ } @@ -157,7 +151,7 @@ QUEUE_SYSFS_SHOW_CONST(discard_zeroes_data, 0) QUEUE_SYSFS_SHOW_CONST(write_same_max, 0) QUEUE_SYSFS_SHOW_CONST(poll_delay, -1) -static ssize_t queue_max_discard_sectors_store(struct request_queue *q, +static ssize_t queue_max_discard_sectors_store(struct gendisk *disk, const char *page, size_t count) { unsigned long max_discard_bytes; @@ -169,15 +163,15 @@ static ssize_t queue_max_discard_sectors_store(struct request_queue *q, if (ret < 0) return ret; - if (max_discard_bytes & (q->limits.discard_granularity - 1)) + if (max_discard_bytes & (disk->queue->limits.discard_granularity - 1)) return -EINVAL; if ((max_discard_bytes >> SECTOR_SHIFT) > UINT_MAX) return -EINVAL; - lim = queue_limits_start_update(q); + lim = queue_limits_start_update(disk->queue); lim.max_user_discard_sectors = max_discard_bytes >> SECTOR_SHIFT; - err = queue_limits_commit_update(q, &lim); + err = queue_limits_commit_update(disk->queue, &lim); if (err) return err; return ret; @@ -188,15 +182,15 @@ static ssize_t queue_max_discard_sectors_store(struct request_queue *q, * underlying queue limits, but actually contains a calculation. Because of * that we can't simply use QUEUE_SYSFS_LIMIT_SHOW_SECTORS_TO_BYTES here. */ -static ssize_t queue_zone_append_max_show(struct request_queue *q, char *page) +static ssize_t queue_zone_append_max_show(struct gendisk *disk, char *page) { - unsigned long long max_sectors = queue_max_zone_append_sectors(q); - - return sprintf(page, "%llu\n", max_sectors << SECTOR_SHIFT); + return sprintf(page, "%llu\n", + (u64)queue_max_zone_append_sectors(disk->queue) << + SECTOR_SHIFT); } static ssize_t -queue_max_sectors_store(struct request_queue *q, const char *page, size_t count) +queue_max_sectors_store(struct gendisk *disk, const char *page, size_t count) { unsigned long max_sectors_kb; struct queue_limits lim; @@ -207,15 +201,15 @@ queue_max_sectors_store(struct request_queue *q, const char *page, size_t count) if (ret < 0) return ret; - lim = queue_limits_start_update(q); + lim = queue_limits_start_update(disk->queue); lim.max_user_sectors = max_sectors_kb << 1; - err = queue_limits_commit_update(q, &lim); + err = queue_limits_commit_update(disk->queue, &lim); if (err) return err; return ret; } -static ssize_t queue_feature_store(struct request_queue *q, const char *page, +static ssize_t queue_feature_store(struct gendisk *disk, const char *page, size_t count, blk_features_t feature) { struct queue_limits lim; @@ -226,26 +220,27 @@ static ssize_t queue_feature_store(struct request_queue *q, const char *page, if (ret < 0) return ret; - lim = queue_limits_start_update(q); + lim = queue_limits_start_update(disk->queue); if (val) lim.features |= feature; else lim.features &= ~feature; - ret = queue_limits_commit_update(q, &lim); + ret = queue_limits_commit_update(disk->queue, &lim); if (ret) return ret; return count; } -#define QUEUE_SYSFS_FEATURE(_name, _feature) \ -static ssize_t queue_##_name##_show(struct request_queue *q, char *page) \ -{ \ - return sprintf(page, "%u\n", !!(q->limits.features & _feature)); \ -} \ -static ssize_t queue_##_name##_store(struct request_queue *q, \ - const char *page, size_t count) \ -{ \ - return queue_feature_store(q, page, count, _feature); \ +#define QUEUE_SYSFS_FEATURE(_name, _feature) \ +static ssize_t queue_##_name##_show(struct gendisk *disk, char *page) \ +{ \ + return sprintf(page, "%u\n", \ + !!(disk->queue->limits.features & _feature)); \ +} \ +static ssize_t queue_##_name##_store(struct gendisk *disk, \ + const char *page, size_t count) \ +{ \ + return queue_feature_store(disk, page, count, _feature); \ } QUEUE_SYSFS_FEATURE(rotational, BLK_FEAT_ROTATIONAL) @@ -253,35 +248,36 @@ QUEUE_SYSFS_FEATURE(add_random, BLK_FEAT_ADD_RANDOM) QUEUE_SYSFS_FEATURE(iostats, BLK_FEAT_IO_STAT) QUEUE_SYSFS_FEATURE(stable_writes, BLK_FEAT_STABLE_WRITES); -#define QUEUE_SYSFS_FEATURE_SHOW(_name, _feature) \ -static ssize_t queue_##_name##_show(struct request_queue *q, char *page) \ -{ \ - return sprintf(page, "%u\n", !!(q->limits.features & _feature)); \ +#define QUEUE_SYSFS_FEATURE_SHOW(_name, _feature) \ +static ssize_t queue_##_name##_show(struct gendisk *disk, char *page) \ +{ \ + return sprintf(page, "%u\n", \ + !!(disk->queue->limits.features & _feature)); \ } QUEUE_SYSFS_FEATURE_SHOW(poll, BLK_FEAT_POLL); QUEUE_SYSFS_FEATURE_SHOW(fua, BLK_FEAT_FUA); QUEUE_SYSFS_FEATURE_SHOW(dax, BLK_FEAT_DAX); -static ssize_t queue_zoned_show(struct request_queue *q, char *page) +static ssize_t queue_zoned_show(struct gendisk *disk, char *page) { - if (blk_queue_is_zoned(q)) + if (blk_queue_is_zoned(disk->queue)) return sprintf(page, "host-managed\n"); return sprintf(page, "none\n"); } -static ssize_t queue_nr_zones_show(struct request_queue *q, char *page) +static ssize_t queue_nr_zones_show(struct gendisk *disk, char *page) { - return queue_var_show(disk_nr_zones(q->disk), page); + return queue_var_show(disk_nr_zones(disk), page); } -static ssize_t queue_nomerges_show(struct request_queue *q, char *page) +static ssize_t queue_nomerges_show(struct gendisk *disk, char *page) { - return queue_var_show((blk_queue_nomerges(q) << 1) | - blk_queue_noxmerges(q), page); + return queue_var_show((blk_queue_nomerges(disk->queue) << 1) | + blk_queue_noxmerges(disk->queue), page); } -static ssize_t queue_nomerges_store(struct request_queue *q, const char *page, +static ssize_t queue_nomerges_store(struct gendisk *disk, const char *page, size_t count) { unsigned long nm; @@ -290,29 +286,30 @@ static ssize_t queue_nomerges_store(struct request_queue *q, const char *page, if (ret < 0) return ret; - blk_queue_flag_clear(QUEUE_FLAG_NOMERGES, q); - blk_queue_flag_clear(QUEUE_FLAG_NOXMERGES, q); + blk_queue_flag_clear(QUEUE_FLAG_NOMERGES, disk->queue); + blk_queue_flag_clear(QUEUE_FLAG_NOXMERGES, disk->queue); if (nm == 2) - blk_queue_flag_set(QUEUE_FLAG_NOMERGES, q); + blk_queue_flag_set(QUEUE_FLAG_NOMERGES, disk->queue); else if (nm) - blk_queue_flag_set(QUEUE_FLAG_NOXMERGES, q); + blk_queue_flag_set(QUEUE_FLAG_NOXMERGES, disk->queue); return ret; } -static ssize_t queue_rq_affinity_show(struct request_queue *q, char *page) +static ssize_t queue_rq_affinity_show(struct gendisk *disk, char *page) { - bool set = test_bit(QUEUE_FLAG_SAME_COMP, &q->queue_flags); - bool force = test_bit(QUEUE_FLAG_SAME_FORCE, &q->queue_flags); + bool set = test_bit(QUEUE_FLAG_SAME_COMP, &disk->queue->queue_flags); + bool force = test_bit(QUEUE_FLAG_SAME_FORCE, &disk->queue->queue_flags); return queue_var_show(set << force, page); } static ssize_t -queue_rq_affinity_store(struct request_queue *q, const char *page, size_t count) +queue_rq_affinity_store(struct gendisk *disk, const char *page, size_t count) { ssize_t ret = -EINVAL; #ifdef CONFIG_SMP + struct request_queue *q = disk->queue; unsigned long val; ret = queue_var_store(&val, page, count); @@ -333,28 +330,28 @@ queue_rq_affinity_store(struct request_queue *q, const char *page, size_t count) return ret; } -static ssize_t queue_poll_delay_store(struct request_queue *q, const char *page, +static ssize_t queue_poll_delay_store(struct gendisk *disk, const char *page, size_t count) { return count; } -static ssize_t queue_poll_store(struct request_queue *q, const char *page, +static ssize_t queue_poll_store(struct gendisk *disk, const char *page, size_t count) { - if (!(q->limits.features & BLK_FEAT_POLL)) + if (!(disk->queue->limits.features & BLK_FEAT_POLL)) return -EINVAL; pr_info_ratelimited("writes to the poll attribute are ignored.\n"); pr_info_ratelimited("please use driver specific parameters instead.\n"); return count; } -static ssize_t queue_io_timeout_show(struct request_queue *q, char *page) +static ssize_t queue_io_timeout_show(struct gendisk *disk, char *page) { - return sprintf(page, "%u\n", jiffies_to_msecs(q->rq_timeout)); + return sprintf(page, "%u\n", jiffies_to_msecs(disk->queue->rq_timeout)); } -static ssize_t queue_io_timeout_store(struct request_queue *q, const char *page, +static ssize_t queue_io_timeout_store(struct gendisk *disk, const char *page, size_t count) { unsigned int val; @@ -364,19 +361,19 @@ static ssize_t queue_io_timeout_store(struct request_queue *q, const char *page, if (err || val == 0) return -EINVAL; - blk_queue_rq_timeout(q, msecs_to_jiffies(val)); + blk_queue_rq_timeout(disk->queue, msecs_to_jiffies(val)); return count; } -static ssize_t queue_wc_show(struct request_queue *q, char *page) +static ssize_t queue_wc_show(struct gendisk *disk, char *page) { - if (blk_queue_write_cache(q)) + if (blk_queue_write_cache(disk->queue)) return sprintf(page, "write back\n"); return sprintf(page, "write through\n"); } -static ssize_t queue_wc_store(struct request_queue *q, const char *page, +static ssize_t queue_wc_store(struct gendisk *disk, const char *page, size_t count) { struct queue_limits lim; @@ -392,12 +389,12 @@ static ssize_t queue_wc_store(struct request_queue *q, const char *page, return -EINVAL; } - lim = queue_limits_start_update(q); + lim = queue_limits_start_update(disk->queue); if (disable) lim.flags |= BLK_FLAG_WRITE_CACHE_DISABLED; else lim.flags &= ~BLK_FLAG_WRITE_CACHE_DISABLED; - err = queue_limits_commit_update(q, &lim); + err = queue_limits_commit_update(disk->queue, &lim); if (err) return err; return count; @@ -489,20 +486,22 @@ static ssize_t queue_var_store64(s64 *var, const char *page) return 0; } -static ssize_t queue_wb_lat_show(struct request_queue *q, char *page) +static ssize_t queue_wb_lat_show(struct gendisk *disk, char *page) { - if (!wbt_rq_qos(q)) + if (!wbt_rq_qos(disk->queue)) return -EINVAL; - if (wbt_disabled(q)) + if (wbt_disabled(disk->queue)) return sprintf(page, "0\n"); - return sprintf(page, "%llu\n", div_u64(wbt_get_min_lat(q), 1000)); + return sprintf(page, "%llu\n", + div_u64(wbt_get_min_lat(disk->queue), 1000)); } -static ssize_t queue_wb_lat_store(struct request_queue *q, const char *page, +static ssize_t queue_wb_lat_store(struct gendisk *disk, const char *page, size_t count) { + struct request_queue *q = disk->queue; struct rq_qos *rqos; ssize_t ret; s64 val; @@ -515,7 +514,7 @@ static ssize_t queue_wb_lat_store(struct request_queue *q, const char *page, rqos = wbt_rq_qos(q); if (!rqos) { - ret = wbt_init(q->disk); + ret = wbt_init(disk); if (ret) return ret; } @@ -649,14 +648,13 @@ queue_attr_show(struct kobject *kobj, struct attribute *attr, char *page) { struct queue_sysfs_entry *entry = to_queue(attr); struct gendisk *disk = container_of(kobj, struct gendisk, queue_kobj); - struct request_queue *q = disk->queue; ssize_t res; if (!entry->show) return -EIO; - mutex_lock(&q->sysfs_lock); - res = entry->show(q, page); - mutex_unlock(&q->sysfs_lock); + mutex_lock(&disk->queue->sysfs_lock); + res = entry->show(disk, page); + mutex_unlock(&disk->queue->sysfs_lock); return res; } @@ -674,7 +672,7 @@ queue_attr_store(struct kobject *kobj, struct attribute *attr, blk_mq_freeze_queue(q); mutex_lock(&q->sysfs_lock); - res = entry->store(q, page, length); + res = entry->store(disk, page, length); mutex_unlock(&q->sysfs_lock); blk_mq_unfreeze_queue(q); return res; diff --git a/block/elevator.c b/block/elevator.c index f64ebd726e588a..f13d552a32c8b8 100644 --- a/block/elevator.c +++ b/block/elevator.c @@ -709,24 +709,25 @@ static int elevator_change(struct request_queue *q, const char *elevator_name) return ret; } -ssize_t elv_iosched_store(struct request_queue *q, const char *buf, +ssize_t elv_iosched_store(struct gendisk *disk, const char *buf, size_t count) { char elevator_name[ELV_NAME_MAX]; int ret; - if (!elv_support_iosched(q)) + if (!elv_support_iosched(disk->queue)) return count; strscpy(elevator_name, buf, sizeof(elevator_name)); - ret = elevator_change(q, strstrip(elevator_name)); + ret = elevator_change(disk->queue, strstrip(elevator_name)); if (!ret) return count; return ret; } -ssize_t elv_iosched_show(struct request_queue *q, char *name) +ssize_t elv_iosched_show(struct gendisk *disk, char *name) { + struct request_queue *q = disk->queue; struct elevator_queue *eq = q->elevator; struct elevator_type *cur = NULL, *e; int len = 0; diff --git a/block/elevator.h b/block/elevator.h index e9a050a96e5305..3fe18e1a869275 100644 --- a/block/elevator.h +++ b/block/elevator.h @@ -147,8 +147,8 @@ extern void elv_unregister(struct elevator_type *); /* * io scheduler sysfs switching */ -extern ssize_t elv_iosched_show(struct request_queue *, char *); -extern ssize_t elv_iosched_store(struct request_queue *, const char *, size_t); +ssize_t elv_iosched_show(struct gendisk *disk, char *page); +ssize_t elv_iosched_store(struct gendisk *disk, const char *page, size_t count); extern bool elv_bio_merge_ok(struct request *, struct bio *); extern struct elevator_queue *elevator_alloc(struct request_queue *, From 4e63aeb5d0101ddada36a2f64f048e2f9d2202fc Mon Sep 17 00:00:00 2001 From: Baokun Li Date: Tue, 4 Jun 2024 11:05:22 +0800 Subject: [PATCH 169/209] blk-wbt: don't throttle swap writes in direct reclaim Now we avoid throttling swap writes by determining whether the current process is kswapd (aka current_is_kswapd()), but swap writes can come from either kswapd or direct reclaim, so the swap writes from direct reclaim will still be throttled. When a process holds a lock to allocate a free page, and enters direct reclaim because there is no free memory, then it might trigger a hung due to the wbt throttling that causes other processes to fail to get the lock. Both kswapd and direct reclaim set the REQ_SWAP flag, so use REQ_SWAP instead of current_is_kswapd() to avoid throttling swap writes. Also renamed WBT_KSWAPD to WBT_SWAP and WBT_RWQ_KSWAPD to WBT_RWQ_SWAP. Signed-off-by: Baokun Li Reviewed-by: Yu Kuai Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240604030522.3686177-1-libaokun@huaweicloud.com Signed-off-by: Jens Axboe --- block/blk-wbt.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) diff --git a/block/blk-wbt.c b/block/blk-wbt.c index 1a5e4b049ecd1d..6dfc659d22e2b7 100644 --- a/block/blk-wbt.c +++ b/block/blk-wbt.c @@ -37,7 +37,7 @@ enum wbt_flags { WBT_TRACKED = 1, /* write, tracked for throttling */ WBT_READ = 2, /* read */ - WBT_KSWAPD = 4, /* write, from kswapd */ + WBT_SWAP = 4, /* write, from swap_writepage() */ WBT_DISCARD = 8, /* discard */ WBT_NR_BITS = 4, /* number of bits */ @@ -45,7 +45,7 @@ enum wbt_flags { enum { WBT_RWQ_BG = 0, - WBT_RWQ_KSWAPD, + WBT_RWQ_SWAP, WBT_RWQ_DISCARD, WBT_NUM_RWQ, }; @@ -172,8 +172,8 @@ static bool wb_recent_wait(struct rq_wb *rwb) static inline struct rq_wait *get_rq_wait(struct rq_wb *rwb, enum wbt_flags wb_acct) { - if (wb_acct & WBT_KSWAPD) - return &rwb->rq_wait[WBT_RWQ_KSWAPD]; + if (wb_acct & WBT_SWAP) + return &rwb->rq_wait[WBT_RWQ_SWAP]; else if (wb_acct & WBT_DISCARD) return &rwb->rq_wait[WBT_RWQ_DISCARD]; @@ -528,7 +528,7 @@ static bool close_io(struct rq_wb *rwb) time_before(now, rwb->last_comp + HZ / 10); } -#define REQ_HIPRIO (REQ_SYNC | REQ_META | REQ_PRIO) +#define REQ_HIPRIO (REQ_SYNC | REQ_META | REQ_PRIO | REQ_SWAP) static inline unsigned int get_limit(struct rq_wb *rwb, blk_opf_t opf) { @@ -539,13 +539,13 @@ static inline unsigned int get_limit(struct rq_wb *rwb, blk_opf_t opf) /* * At this point we know it's a buffered write. If this is - * kswapd trying to free memory, or REQ_SYNC is set, then + * swap trying to free memory, or REQ_SYNC is set, then * it's WB_SYNC_ALL writeback, and we'll use the max limit for * that. If the write is marked as a background write, then use * the idle limit, or go to normal if we haven't had competing * IO for a bit. */ - if ((opf & REQ_HIPRIO) || wb_recent_wait(rwb) || current_is_kswapd()) + if ((opf & REQ_HIPRIO) || wb_recent_wait(rwb)) limit = rwb->rq_depth.max_depth; else if ((opf & REQ_BACKGROUND) || close_io(rwb)) { /* @@ -622,8 +622,8 @@ static enum wbt_flags bio_to_wbt_flags(struct rq_wb *rwb, struct bio *bio) if (bio_op(bio) == REQ_OP_READ) { flags = WBT_READ; } else if (wbt_should_throttle(bio)) { - if (current_is_kswapd()) - flags |= WBT_KSWAPD; + if (bio->bi_opf & REQ_SWAP) + flags |= WBT_SWAP; if (bio_op(bio) == REQ_OP_DISCARD) flags |= WBT_DISCARD; flags |= WBT_TRACKED; From f62e8edc0a9fda84fe5bf32d5f5874b489d6c301 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 1 Jul 2024 07:17:50 +0200 Subject: [PATCH 170/209] block: remove a duplicate io_min check in blk_validate_limits If io_min is larger than the cap, it must by definition be non-zero. Signed-off-by: Christoph Hellwig Reviewed-by: Chaitanya Kulkarni Reviewed-by: Sagi Grimberg Reviewed-by: Nitesh Shetty Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240701051800.1245240-2-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index 2e559cf97cc834..ff8bbc101fedaa 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -279,8 +279,7 @@ static int blk_validate_limits(struct queue_limits *lim) } else if (lim->io_opt) { lim->max_sectors = min(max_hw_sectors, lim->io_opt >> SECTOR_SHIFT); - } else if (lim->io_min && - lim->io_min > (BLK_DEF_MAX_SECTORS_CAP << SECTOR_SHIFT)) { + } else if (lim->io_min > (BLK_DEF_MAX_SECTORS_CAP << SECTOR_SHIFT)) { lim->max_sectors = min(max_hw_sectors, lim->io_min >> SECTOR_SHIFT); } else { From 37105615f73125cb0466c09796f277a4c46d9295 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 1 Jul 2024 07:17:51 +0200 Subject: [PATCH 171/209] block: don't reduce max_sectors based on io_opt Don't reduce the max_sectors value below the normal cap when the driver advertsizes a very low io_opt. This restores the behavior we had before the recent changes to the max_sectors calculation. Signed-off-by: Christoph Hellwig Reviewed-by: Chaitanya Kulkarni Reviewed-by: Sagi Grimberg Reviewed-by: Nitesh Shetty Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240701051800.1245240-3-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-settings.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/block/blk-settings.c b/block/blk-settings.c index ff8bbc101fedaa..9fa4eed4df06b0 100644 --- a/block/blk-settings.c +++ b/block/blk-settings.c @@ -276,7 +276,7 @@ static int blk_validate_limits(struct queue_limits *lim) if (lim->max_user_sectors < PAGE_SIZE / SECTOR_SIZE) return -EINVAL; lim->max_sectors = min(max_hw_sectors, lim->max_user_sectors); - } else if (lim->io_opt) { + } else if (lim->io_opt > (BLK_DEF_MAX_SECTORS_CAP << SECTOR_SHIFT)) { lim->max_sectors = min(max_hw_sectors, lim->io_opt >> SECTOR_SHIFT); } else if (lim->io_min > (BLK_DEF_MAX_SECTORS_CAP << SECTOR_SHIFT)) { From f3bf25d5135539603f24e377c6dec3016fbd9786 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 1 Jul 2024 07:17:52 +0200 Subject: [PATCH 172/209] nvme: don't set io_opt if NOWS is zero NOWS is one of the annoying "0's based values" in NVMe, where 0 means one and we thus can't detect if it isn't set. Thus a NOWS value of 0 means that the Namespace Optimal Write Size is a single LBA, which is clearly bogus. Ignore the value in that case and don't propagate an io_opt value to the block layer. Signed-off-by: Christoph Hellwig Reviewed-by: Chaitanya Kulkarni Reviewed-by: Sagi Grimberg Reviewed-by: Nitesh Shetty Reviewed-by: Johannes Thumshirn Link: https://lore.kernel.org/r/20240701051800.1245240-4-hch@lst.de Signed-off-by: Jens Axboe --- drivers/nvme/host/core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 89ebfa89613ee6..ef00bfc6b8408a 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -2024,7 +2024,8 @@ static bool nvme_update_disk_info(struct nvme_ns *ns, struct nvme_id_ns *id, /* NPWG = Namespace Preferred Write Granularity */ phys_bs = bs * (1 + le16_to_cpu(id->npwg)); /* NOWS = Namespace Optimal Write Size */ - io_opt = bs * (1 + le16_to_cpu(id->nows)); + if (id->nows) + io_opt = bs * (1 + le16_to_cpu(id->nows)); } /* From b0727b1243cd084260e47c51c7950020bfddb636 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 1 Jul 2024 08:40:29 -0600 Subject: [PATCH 173/209] io_uring/msg_ring: check for dead submitter task The change for improving the handling of the target CQE posting inadvertently dropped the NULL check for the submitter task on the target ring, reinstate that. Fixes: 0617bb500bfa ("io_uring/msg_ring: improve handling of target CQE posting") Reported-by: Pavel Begunkov Signed-off-by: Jens Axboe --- io_uring/msg_ring.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) diff --git a/io_uring/msg_ring.c b/io_uring/msg_ring.c index 47a754e83b49a9..c2171495098bfe 100644 --- a/io_uring/msg_ring.c +++ b/io_uring/msg_ring.c @@ -86,16 +86,21 @@ static void io_msg_tw_complete(struct io_kiocb *req, struct io_tw_state *ts) percpu_ref_put(&ctx->refs); } -static void io_msg_remote_post(struct io_ring_ctx *ctx, struct io_kiocb *req, - int res, u32 cflags, u64 user_data) +static int io_msg_remote_post(struct io_ring_ctx *ctx, struct io_kiocb *req, + int res, u32 cflags, u64 user_data) { + req->task = READ_ONCE(ctx->submitter_task); + if (!req->task) { + kmem_cache_free(req_cachep, req); + return -EOWNERDEAD; + } req->cqe.user_data = user_data; io_req_set_res(req, res, cflags); percpu_ref_get(&ctx->refs); req->ctx = ctx; - req->task = READ_ONCE(ctx->submitter_task); req->io_task_work.func = io_msg_tw_complete; io_req_task_work_add_remote(req, ctx, IOU_F_TWQ_LAZY_WAKE); + return 0; } static struct io_kiocb *io_msg_get_kiocb(struct io_ring_ctx *ctx) @@ -125,8 +130,8 @@ static int io_msg_data_remote(struct io_kiocb *req) if (msg->flags & IORING_MSG_RING_FLAGS_PASS) flags = msg->cqe_flags; - io_msg_remote_post(target_ctx, target, msg->len, flags, msg->user_data); - return 0; + return io_msg_remote_post(target_ctx, target, msg->len, flags, + msg->user_data); } static int io_msg_ring_data(struct io_kiocb *req, unsigned int issue_flags) From be4f5d9c992ba1d89ce63ad9e40a99f120882038 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 1 Jul 2024 08:46:25 -0600 Subject: [PATCH 174/209] io_uring/msg_ring: use kmem_cache_free() to free request The change adding caching around the request allocated and freed for data messages changed a kmem_cache_free() to a kfree(), which isn't correct as the request came from slab in the first place. Fix that up and use the right freeing function if the cache is already at its limit. Note that the current mixing of kmem_cache_alloc and kfree is fine, but consistent alloc/free functions should be used as it's otherwise somewhat confusing. Fixes: 50cf5f3842af ("io_uring/msg_ring: add an alloc cache for io_kiocb entries") Signed-off-by: Jens Axboe --- io_uring/msg_ring.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/io_uring/msg_ring.c b/io_uring/msg_ring.c index c2171495098bfe..29fa9285a33d9e 100644 --- a/io_uring/msg_ring.c +++ b/io_uring/msg_ring.c @@ -82,7 +82,7 @@ static void io_msg_tw_complete(struct io_kiocb *req, struct io_tw_state *ts) spin_unlock(&ctx->msg_lock); } if (req) - kfree(req); + kmem_cache_free(req_cachep, req); percpu_ref_put(&ctx->refs); } From 8515f1661ca1f9ad63850a5e1e86599399420d2e Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Mon, 1 Jul 2024 09:23:26 -0600 Subject: [PATCH 175/209] MAINTAINERS: change Pavel Begunkov from io_uring reviewer to maintainer This more accurately describes Pavel's role for the project, so let's make the change to reflect that. Signed-off-by: Jens Axboe --- MAINTAINERS | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/MAINTAINERS b/MAINTAINERS index cf9c9221c38830..ad96b9bd68ac61 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -11570,7 +11570,7 @@ F: include/linux/iosys-map.h IO_URING M: Jens Axboe -R: Pavel Begunkov +M: Pavel Begunkov L: io-uring@vger.kernel.org S: Maintained T: git git://git.kernel.dk/linux-block From 6259151c04d4e0085e00d2dcb471ebdd1778e72e Mon Sep 17 00:00:00 2001 From: Bart Van Assche Date: Thu, 9 May 2024 10:01:48 -0700 Subject: [PATCH 176/209] block: Call .limit_depth() after .hctx has been set Call .limit_depth() after data->hctx has been set such that data->hctx can be used in .limit_depth() implementations. Cc: Christoph Hellwig Cc: Damien Le Moal Cc: Zhiguo Niu Fixes: 07757588e507 ("block/mq-deadline: Reserve 25% of scheduler tags for synchronous requests") Signed-off-by: Bart Van Assche Tested-by: Zhiguo Niu Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240509170149.7639-2-bvanassche@acm.org Signed-off-by: Jens Axboe --- block/blk-mq.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/block/blk-mq.c b/block/blk-mq.c index fec2dea5c6e885..e3c3c0c21b5536 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -448,6 +448,10 @@ static struct request *__blk_mq_alloc_requests(struct blk_mq_alloc_data *data) if (data->cmd_flags & REQ_NOWAIT) data->flags |= BLK_MQ_REQ_NOWAIT; +retry: + data->ctx = blk_mq_get_ctx(q); + data->hctx = blk_mq_map_queue(q, data->cmd_flags, data->ctx); + if (q->elevator) { /* * All requests use scheduler tags when an I/O scheduler is @@ -469,13 +473,9 @@ static struct request *__blk_mq_alloc_requests(struct blk_mq_alloc_data *data) if (ops->limit_depth) ops->limit_depth(data->cmd_flags, data); } - } - -retry: - data->ctx = blk_mq_get_ctx(q); - data->hctx = blk_mq_map_queue(q, data->cmd_flags, data->ctx); - if (!(data->rq_flags & RQF_SCHED_TAGS)) + } else { blk_mq_tag_busy(data->hctx); + } if (data->flags & BLK_MQ_REQ_RESERVED) data->rq_flags |= RQF_RESV; From 39823b47bbd40502632ffba90ebb34fff7c8b5e8 Mon Sep 17 00:00:00 2001 From: Bart Van Assche Date: Thu, 9 May 2024 10:01:49 -0700 Subject: [PATCH 177/209] block/mq-deadline: Fix the tag reservation code The current tag reservation code is based on a misunderstanding of the meaning of data->shallow_depth. Fix the tag reservation code as follows: * By default, do not reserve any tags for synchronous requests because for certain use cases reserving tags reduces performance. See also Harshit Mogalapalli, [bug-report] Performance regression with fio sequential-write on a multipath setup, 2024-03-07 (https://lore.kernel.org/linux-block/5ce2ae5d-61e2-4ede-ad55-551112602401@oracle.com/) * Reduce min_shallow_depth to one because min_shallow_depth must be less than or equal any shallow_depth value. * Scale dd->async_depth from the range [1, nr_requests] to [1, bits_per_sbitmap_word]. Cc: Christoph Hellwig Cc: Damien Le Moal Cc: Zhiguo Niu Fixes: 07757588e507 ("block/mq-deadline: Reserve 25% of scheduler tags for synchronous requests") Signed-off-by: Bart Van Assche Reviewed-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240509170149.7639-3-bvanassche@acm.org Signed-off-by: Jens Axboe --- block/mq-deadline.c | 20 +++++++++++++++++--- 1 file changed, 17 insertions(+), 3 deletions(-) diff --git a/block/mq-deadline.c b/block/mq-deadline.c index 94eede4fb9ebed..acdc28756d9d77 100644 --- a/block/mq-deadline.c +++ b/block/mq-deadline.c @@ -487,6 +487,20 @@ static struct request *dd_dispatch_request(struct blk_mq_hw_ctx *hctx) return rq; } +/* + * 'depth' is a number in the range 1..INT_MAX representing a number of + * requests. Scale it with a factor (1 << bt->sb.shift) / q->nr_requests since + * 1..(1 << bt->sb.shift) is the range expected by sbitmap_get_shallow(). + * Values larger than q->nr_requests have the same effect as q->nr_requests. + */ +static int dd_to_word_depth(struct blk_mq_hw_ctx *hctx, unsigned int qdepth) +{ + struct sbitmap_queue *bt = &hctx->sched_tags->bitmap_tags; + const unsigned int nrr = hctx->queue->nr_requests; + + return ((qdepth << bt->sb.shift) + nrr - 1) / nrr; +} + /* * Called by __blk_mq_alloc_request(). The shallow_depth value set by this * function is used by __blk_mq_get_tag(). @@ -503,7 +517,7 @@ static void dd_limit_depth(blk_opf_t opf, struct blk_mq_alloc_data *data) * Throttle asynchronous requests and writes such that these requests * do not block the allocation of synchronous requests. */ - data->shallow_depth = dd->async_depth; + data->shallow_depth = dd_to_word_depth(data->hctx, dd->async_depth); } /* Called by blk_mq_update_nr_requests(). */ @@ -513,9 +527,9 @@ static void dd_depth_updated(struct blk_mq_hw_ctx *hctx) struct deadline_data *dd = q->elevator->elevator_data; struct blk_mq_tags *tags = hctx->sched_tags; - dd->async_depth = max(1UL, 3 * q->nr_requests / 4); + dd->async_depth = q->nr_requests; - sbitmap_queue_min_shallow_depth(&tags->bitmap_tags, dd->async_depth); + sbitmap_queue_min_shallow_depth(&tags->bitmap_tags, 1); } /* Called by blk_mq_init_hctx() and blk_mq_init_sched(). */ From 3b7c16be30e35ec035b2efcc0f7d7b368789c443 Mon Sep 17 00:00:00 2001 From: Pavel Begunkov Date: Tue, 2 Jul 2024 14:38:11 +0100 Subject: [PATCH 178/209] io_uring/msg_ring: fix overflow posting The caller of io_cqring_event_overflow() should be holding the completion_lock, which is violated by io_msg_tw_complete. There is only one caller of io_add_aux_cqe(), so just add locking there for now. WARNING: CPU: 0 PID: 5145 at io_uring/io_uring.c:703 io_cqring_event_overflow+0x442/0x660 io_uring/io_uring.c:703 RIP: 0010:io_cqring_event_overflow+0x442/0x660 io_uring/io_uring.c:703 __io_post_aux_cqe io_uring/io_uring.c:816 [inline] io_add_aux_cqe+0x27c/0x320 io_uring/io_uring.c:837 io_msg_tw_complete+0x9d/0x4d0 io_uring/msg_ring.c:78 io_fallback_req_func+0xce/0x1c0 io_uring/io_uring.c:256 process_one_work kernel/workqueue.c:3224 [inline] process_scheduled_works+0xa2c/0x1830 kernel/workqueue.c:3305 worker_thread+0x86d/0xd40 kernel/workqueue.c:3383 kthread+0x2f0/0x390 kernel/kthread.c:389 ret_from_fork+0x4b/0x80 arch/x86/kernel/process.c:144 ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:244 Fixes: f33096a3c99c0 ("io_uring: add io_add_aux_cqe() helper") Reported-by: syzbot+f7f9c893345c5c615d34@syzkaller.appspotmail.com Signed-off-by: Pavel Begunkov Link: https://lore.kernel.org/r/c7350d07fefe8cce32b50f57665edbb6355ea8c1.1719927398.git.asml.silence@gmail.com Signed-off-by: Jens Axboe --- io_uring/io_uring.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c index 7ed1e009aaecb0..42139bb85fff86 100644 --- a/io_uring/io_uring.c +++ b/io_uring/io_uring.c @@ -834,7 +834,11 @@ bool io_post_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags */ void io_add_aux_cqe(struct io_ring_ctx *ctx, u64 user_data, s32 res, u32 cflags) { - __io_post_aux_cqe(ctx, user_data, res, cflags); + if (!io_fill_cqe_aux(ctx, user_data, res, cflags)) { + spin_lock(&ctx->completion_lock); + io_cqring_event_overflow(ctx, user_data, res, cflags, 0, 0); + spin_unlock(&ctx->completion_lock); + } ctx->submit_state.cq_flush = true; } From 1c0b3fca381bf879e2168b362692f83808677f95 Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Tue, 2 Jul 2024 16:32:34 +0900 Subject: [PATCH 179/209] null_blk: Fix description of the fua parameter The description of the fua module parameter is defined using MODULE_PARM_DESC() with the first argument passed being "zoned". That is the wrong name, obviously. Fix that by using the correct "fua" parameter name so that "modinfo null_blk" displays correct information. Fixes: f4f84586c8b9 ("null_blk: Introduce fua attribute") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240702073234.206458-1-dlemoal@kernel.org Signed-off-by: Jens Axboe --- drivers/block/null_blk/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c index 83a4ebe4763ae5..5de9ca4eceb4b4 100644 --- a/drivers/block/null_blk/main.c +++ b/drivers/block/null_blk/main.c @@ -227,7 +227,7 @@ MODULE_PARM_DESC(mbps, "Cache size in MiB for memory-backed device. Default: 0 ( static bool g_fua = true; module_param_named(fua, g_fua, bool, 0444); -MODULE_PARM_DESC(zoned, "Enable/disable FUA support when cache_size is used. Default: true"); +MODULE_PARM_DESC(fua, "Enable/disable FUA support when cache_size is used. Default: true"); static unsigned int g_mbps; module_param_named(mbps, g_mbps, uint, 0444); From 98d34c087249d39838874b83e17671e7d5eb1ca7 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 25 Jun 2024 07:52:38 +0200 Subject: [PATCH 180/209] xen-blkfront: fix sector_size propagation to the block layer MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Ensure that info->sector_size and info->physical_sector_size are set before the call to blkif_set_queue_limits by doing away with the local variables and arguments that propagate them. Thanks to Marek Marczykowski-Górecki and Jürgen Groß for root causing the issue. Fixes: ba3f67c11638 ("xen-blkfront: atomically update queue limits") Reported-by: Rusty Bird Signed-off-by: Christoph Hellwig Reviewed-by: Roger Pau Monné Link: https://lore.kernel.org/r/20240625055238.7934-1-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/xen-blkfront.c | 16 +++++----------- 1 file changed, 5 insertions(+), 11 deletions(-) diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c index fa3a2ba525458b..59ce113b882a0e 100644 --- a/drivers/block/xen-blkfront.c +++ b/drivers/block/xen-blkfront.c @@ -1070,8 +1070,7 @@ static char *encode_disk_name(char *ptr, unsigned int n) } static int xlvbd_alloc_gendisk(blkif_sector_t capacity, - struct blkfront_info *info, u16 sector_size, - unsigned int physical_sector_size) + struct blkfront_info *info) { struct queue_limits lim = {}; struct gendisk *gd; @@ -1165,8 +1164,6 @@ static int xlvbd_alloc_gendisk(blkif_sector_t capacity, info->rq = gd->queue; info->gd = gd; - info->sector_size = sector_size; - info->physical_sector_size = physical_sector_size; xlvbd_flush(info); @@ -2320,8 +2317,6 @@ static void blkfront_gather_backend_features(struct blkfront_info *info) static void blkfront_connect(struct blkfront_info *info) { unsigned long long sectors; - unsigned long sector_size; - unsigned int physical_sector_size; int err, i; struct blkfront_ring_info *rinfo; @@ -2360,7 +2355,7 @@ static void blkfront_connect(struct blkfront_info *info) err = xenbus_gather(XBT_NIL, info->xbdev->otherend, "sectors", "%llu", §ors, "info", "%u", &info->vdisk_info, - "sector-size", "%lu", §or_size, + "sector-size", "%lu", &info->sector_size, NULL); if (err) { xenbus_dev_fatal(info->xbdev, err, @@ -2374,9 +2369,9 @@ static void blkfront_connect(struct blkfront_info *info) * provide this. Assume physical sector size to be the same as * sector_size in that case. */ - physical_sector_size = xenbus_read_unsigned(info->xbdev->otherend, + info->physical_sector_size = xenbus_read_unsigned(info->xbdev->otherend, "physical-sector-size", - sector_size); + info->sector_size); blkfront_gather_backend_features(info); for_each_rinfo(info, rinfo, i) { err = blkfront_setup_indirect(rinfo); @@ -2388,8 +2383,7 @@ static void blkfront_connect(struct blkfront_info *info) } } - err = xlvbd_alloc_gendisk(sectors, info, sector_size, - physical_sector_size); + err = xlvbd_alloc_gendisk(sectors, info); if (err) { xenbus_dev_fatal(info->xbdev, err, "xlvbd_add at %s", info->xbdev->otherend); From 93d8032f4143c8d2ac3e10c6504385c26acc511f Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Tue, 2 Jul 2024 09:16:46 -0600 Subject: [PATCH 181/209] io_uring/net: cleanup io_recv_finish() bundle handling Combine the two cases that check for whether or not this is a bundle, rather than having them as separate checks. This is easier to reduce, and it reduces the text associated with it as well. Signed-off-by: Jens Axboe --- io_uring/net.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/io_uring/net.c b/io_uring/net.c index db4a4a03ce3ab6..25223e11958f8c 100644 --- a/io_uring/net.c +++ b/io_uring/net.c @@ -827,20 +827,20 @@ static inline bool io_recv_finish(struct io_kiocb *req, int *ret, bool mshot_finished, unsigned issue_flags) { struct io_sr_msg *sr = io_kiocb_to_cmd(req, struct io_sr_msg); - unsigned int cflags; - - if (sr->flags & IORING_RECVSEND_BUNDLE) - cflags = io_put_kbufs(req, io_bundle_nbufs(kmsg, *ret), - issue_flags); - else - cflags = io_put_kbuf(req, issue_flags); + unsigned int cflags = 0; if (kmsg->msg.msg_inq > 0) cflags |= IORING_CQE_F_SOCK_NONEMPTY; - /* bundle with no more immediate buffers, we're done */ - if (sr->flags & IORING_RECVSEND_BUNDLE && req->flags & REQ_F_BL_EMPTY) - goto finish; + if (sr->flags & IORING_RECVSEND_BUNDLE) { + cflags |= io_put_kbufs(req, io_bundle_nbufs(kmsg, *ret), + issue_flags); + /* bundle with no more immediate buffers, we're done */ + if (req->flags & REQ_F_BL_EMPTY) + goto finish; + } else { + cflags |= io_put_kbuf(req, issue_flags); + } /* * Fill CQE for this receive and see if we should keep trying to From da042a3655151157c06e71a583e883ab2d86d1ff Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 2 Jul 2024 17:10:19 +0200 Subject: [PATCH 182/209] block: split integrity support out of bio.h Split struct bio_integrity_payload and the related prototypes out of bio.h into a separate bio-integrity.h header so that it is only pulled in by the few places that need it. Signed-off-by: Christoph Hellwig Reviewed-by: Anuj Gupta Reviewed-by: Kanchan Joshi Reviewed-by: Johannes Thumshirn Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240702151047.1746127-2-hch@lst.de Signed-off-by: Jens Axboe --- block/bio.c | 2 +- block/blk.h | 1 + block/bounce.c | 2 +- drivers/md/dm.c | 1 + drivers/nvme/host/ioctl.c | 1 + drivers/scsi/sd.c | 3 +- include/linux/bio-integrity.h | 153 +++++++++++++++++++++++++++++++++ include/linux/bio.h | 156 ---------------------------------- include/linux/blk-integrity.h | 1 + 9 files changed, 161 insertions(+), 159 deletions(-) create mode 100644 include/linux/bio-integrity.h diff --git a/block/bio.c b/block/bio.c index e9e809a63c5975..4ca3f31ce45fb5 100644 --- a/block/bio.c +++ b/block/bio.c @@ -4,7 +4,7 @@ */ #include #include -#include +#include #include #include #include diff --git a/block/blk.h b/block/blk.h index 47dadd2439b1ca..401e604f35d2cf 100644 --- a/block/blk.h +++ b/block/blk.h @@ -2,6 +2,7 @@ #ifndef BLK_INTERNAL_H #define BLK_INTERNAL_H +#include #include #include /* for max_pfn/max_low_pfn */ #include diff --git a/block/bounce.c b/block/bounce.c index d6a5219f29dd53..0d898cd5ec497f 100644 --- a/block/bounce.c +++ b/block/bounce.c @@ -10,7 +10,7 @@ #include #include #include -#include +#include #include #include #include diff --git a/drivers/md/dm.c b/drivers/md/dm.c index 7d107ae06e1ae1..92d6eeb0a59327 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -11,6 +11,7 @@ #include "dm-uevent.h" #include "dm-ima.h" +#include #include #include #include diff --git a/drivers/nvme/host/ioctl.c b/drivers/nvme/host/ioctl.c index 8b69427a44762a..fb46f55f8b2894 100644 --- a/drivers/nvme/host/ioctl.c +++ b/drivers/nvme/host/ioctl.c @@ -3,6 +3,7 @@ * Copyright (c) 2011-2014, Intel Corporation. * Copyright (c) 2017-2021 Christoph Hellwig. */ +#include #include /* for force_successful_syscall_return */ #include #include diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 979795dad62b11..8bb3a361185152 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -33,11 +33,12 @@ * than the level indicated above to trigger output. */ +#include #include #include #include #include -#include +#include #include #include #include diff --git a/include/linux/bio-integrity.h b/include/linux/bio-integrity.h new file mode 100644 index 00000000000000..70ef19a0dc7e8b --- /dev/null +++ b/include/linux/bio-integrity.h @@ -0,0 +1,153 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef _LINUX_BIO_INTEGRITY_H +#define _LINUX_BIO_INTEGRITY_H + +#include + +enum bip_flags { + BIP_BLOCK_INTEGRITY = 1 << 0, /* block layer owns integrity data */ + BIP_MAPPED_INTEGRITY = 1 << 1, /* ref tag has been remapped */ + BIP_CTRL_NOCHECK = 1 << 2, /* disable HBA integrity checking */ + BIP_DISK_NOCHECK = 1 << 3, /* disable disk integrity checking */ + BIP_IP_CHECKSUM = 1 << 4, /* IP checksum */ + BIP_INTEGRITY_USER = 1 << 5, /* Integrity payload is user address */ + BIP_COPY_USER = 1 << 6, /* Kernel bounce buffer in use */ +}; + +struct bio_integrity_payload { + struct bio *bip_bio; /* parent bio */ + + struct bvec_iter bip_iter; + + unsigned short bip_vcnt; /* # of integrity bio_vecs */ + unsigned short bip_max_vcnt; /* integrity bio_vec slots */ + unsigned short bip_flags; /* control flags */ + + struct bvec_iter bio_iter; /* for rewinding parent bio */ + + struct work_struct bip_work; /* I/O completion */ + + struct bio_vec *bip_vec; + struct bio_vec bip_inline_vecs[];/* embedded bvec array */ +}; + +#ifdef CONFIG_BLK_DEV_INTEGRITY + +#define bip_for_each_vec(bvl, bip, iter) \ + for_each_bvec(bvl, (bip)->bip_vec, iter, (bip)->bip_iter) + +#define bio_for_each_integrity_vec(_bvl, _bio, _iter) \ + for_each_bio(_bio) \ + bip_for_each_vec(_bvl, _bio->bi_integrity, _iter) + +static inline struct bio_integrity_payload *bio_integrity(struct bio *bio) +{ + if (bio->bi_opf & REQ_INTEGRITY) + return bio->bi_integrity; + + return NULL; +} + +static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag) +{ + struct bio_integrity_payload *bip = bio_integrity(bio); + + if (bip) + return bip->bip_flags & flag; + + return false; +} + +static inline sector_t bip_get_seed(struct bio_integrity_payload *bip) +{ + return bip->bip_iter.bi_sector; +} + +static inline void bip_set_seed(struct bio_integrity_payload *bip, + sector_t seed) +{ + bip->bip_iter.bi_sector = seed; +} + +struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio, gfp_t gfp, + unsigned int nr); +int bio_integrity_add_page(struct bio *bio, struct page *page, unsigned int len, + unsigned int offset); +int bio_integrity_map_user(struct bio *bio, void __user *ubuf, ssize_t len, u32 seed); +void bio_integrity_unmap_free_user(struct bio *bio); +bool bio_integrity_prep(struct bio *bio); +void bio_integrity_advance(struct bio *bio, unsigned int bytes_done); +void bio_integrity_trim(struct bio *bio); +int bio_integrity_clone(struct bio *bio, struct bio *bio_src, gfp_t gfp_mask); +int bioset_integrity_create(struct bio_set *bs, int pool_size); +void bioset_integrity_free(struct bio_set *bs); +void bio_integrity_init(void); + +#else /* CONFIG_BLK_DEV_INTEGRITY */ + +static inline void *bio_integrity(struct bio *bio) +{ + return NULL; +} + +static inline int bioset_integrity_create(struct bio_set *bs, int pool_size) +{ + return 0; +} + +static inline void bioset_integrity_free(struct bio_set *bs) +{ +} + +static inline int bio_integrity_map_user(struct bio *bio, void __user *ubuf, + ssize_t len, u32 seed) +{ + return -EINVAL; +} + +static inline void bio_integrity_unmap_free_user(struct bio *bio) +{ +} + +static inline bool bio_integrity_prep(struct bio *bio) +{ + return true; +} + +static inline int bio_integrity_clone(struct bio *bio, struct bio *bio_src, + gfp_t gfp_mask) +{ + return 0; +} + +static inline void bio_integrity_advance(struct bio *bio, + unsigned int bytes_done) +{ +} + +static inline void bio_integrity_trim(struct bio *bio) +{ +} + +static inline void bio_integrity_init(void) +{ +} + +static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag) +{ + return false; +} + +static inline void *bio_integrity_alloc(struct bio *bio, gfp_t gfp, + unsigned int nr) +{ + return ERR_PTR(-EINVAL); +} + +static inline int bio_integrity_add_page(struct bio *bio, struct page *page, + unsigned int len, unsigned int offset) +{ + return 0; +} +#endif /* CONFIG_BLK_DEV_INTEGRITY */ +#endif /* _LINUX_BIO_INTEGRITY_H */ diff --git a/include/linux/bio.h b/include/linux/bio.h index 818e9361294781..a46e2047bea4d2 100644 --- a/include/linux/bio.h +++ b/include/linux/bio.h @@ -321,69 +321,6 @@ static inline void bio_next_folio(struct folio_iter *fi, struct bio *bio) #define bio_for_each_folio_all(fi, bio) \ for (bio_first_folio(&fi, bio, 0); fi.folio; bio_next_folio(&fi, bio)) -enum bip_flags { - BIP_BLOCK_INTEGRITY = 1 << 0, /* block layer owns integrity data */ - BIP_MAPPED_INTEGRITY = 1 << 1, /* ref tag has been remapped */ - BIP_CTRL_NOCHECK = 1 << 2, /* disable HBA integrity checking */ - BIP_DISK_NOCHECK = 1 << 3, /* disable disk integrity checking */ - BIP_IP_CHECKSUM = 1 << 4, /* IP checksum */ - BIP_INTEGRITY_USER = 1 << 5, /* Integrity payload is user address */ - BIP_COPY_USER = 1 << 6, /* Kernel bounce buffer in use */ -}; - -/* - * bio integrity payload - */ -struct bio_integrity_payload { - struct bio *bip_bio; /* parent bio */ - - struct bvec_iter bip_iter; - - unsigned short bip_vcnt; /* # of integrity bio_vecs */ - unsigned short bip_max_vcnt; /* integrity bio_vec slots */ - unsigned short bip_flags; /* control flags */ - - struct bvec_iter bio_iter; /* for rewinding parent bio */ - - struct work_struct bip_work; /* I/O completion */ - - struct bio_vec *bip_vec; - struct bio_vec bip_inline_vecs[];/* embedded bvec array */ -}; - -#if defined(CONFIG_BLK_DEV_INTEGRITY) - -static inline struct bio_integrity_payload *bio_integrity(struct bio *bio) -{ - if (bio->bi_opf & REQ_INTEGRITY) - return bio->bi_integrity; - - return NULL; -} - -static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag) -{ - struct bio_integrity_payload *bip = bio_integrity(bio); - - if (bip) - return bip->bip_flags & flag; - - return false; -} - -static inline sector_t bip_get_seed(struct bio_integrity_payload *bip) -{ - return bip->bip_iter.bi_sector; -} - -static inline void bip_set_seed(struct bio_integrity_payload *bip, - sector_t seed) -{ - bip->bip_iter.bi_sector = seed; -} - -#endif /* CONFIG_BLK_DEV_INTEGRITY */ - void bio_trim(struct bio *bio, sector_t offset, sector_t size); extern struct bio *bio_split(struct bio *bio, int sectors, gfp_t gfp, struct bio_set *bs); @@ -721,99 +658,6 @@ static inline bool bioset_initialized(struct bio_set *bs) return bs->bio_slab != NULL; } -#if defined(CONFIG_BLK_DEV_INTEGRITY) - -#define bip_for_each_vec(bvl, bip, iter) \ - for_each_bvec(bvl, (bip)->bip_vec, iter, (bip)->bip_iter) - -#define bio_for_each_integrity_vec(_bvl, _bio, _iter) \ - for_each_bio(_bio) \ - bip_for_each_vec(_bvl, _bio->bi_integrity, _iter) - -int bio_integrity_map_user(struct bio *bio, void __user *ubuf, ssize_t len, u32 seed); -void bio_integrity_unmap_free_user(struct bio *bio); -extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int); -extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int); -extern bool bio_integrity_prep(struct bio *); -extern void bio_integrity_advance(struct bio *, unsigned int); -extern void bio_integrity_trim(struct bio *); -extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t); -extern int bioset_integrity_create(struct bio_set *, int); -extern void bioset_integrity_free(struct bio_set *); -extern void bio_integrity_init(void); - -#else /* CONFIG_BLK_DEV_INTEGRITY */ - -static inline void *bio_integrity(struct bio *bio) -{ - return NULL; -} - -static inline int bioset_integrity_create(struct bio_set *bs, int pool_size) -{ - return 0; -} - -static inline void bioset_integrity_free (struct bio_set *bs) -{ - return; -} - -static inline bool bio_integrity_prep(struct bio *bio) -{ - return true; -} - -static inline int bio_integrity_clone(struct bio *bio, struct bio *bio_src, - gfp_t gfp_mask) -{ - return 0; -} - -static inline void bio_integrity_advance(struct bio *bio, - unsigned int bytes_done) -{ - return; -} - -static inline void bio_integrity_trim(struct bio *bio) -{ - return; -} - -static inline void bio_integrity_init(void) -{ - return; -} - -static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag) -{ - return false; -} - -static inline void *bio_integrity_alloc(struct bio * bio, gfp_t gfp, - unsigned int nr) -{ - return ERR_PTR(-EINVAL); -} - -static inline int bio_integrity_add_page(struct bio *bio, struct page *page, - unsigned int len, unsigned int offset) -{ - return 0; -} - -static inline int bio_integrity_map_user(struct bio *bio, void __user *ubuf, - ssize_t len, u32 seed) -{ - return -EINVAL; -} -static inline void bio_integrity_unmap_free_user(struct bio *bio) -{ -} - -#endif /* CONFIG_BLK_DEV_INTEGRITY */ - /* * Mark a bio as polled. Note that for async polled IO, the caller must * expect -EWOULDBLOCK if we cannot allocate a request (or other resources). diff --git a/include/linux/blk-integrity.h b/include/linux/blk-integrity.h index 804f856ed3e571..de98049b7ded91 100644 --- a/include/linux/blk-integrity.h +++ b/include/linux/blk-integrity.h @@ -3,6 +3,7 @@ #define _LINUX_BLK_INTEGRITY_H #include +#include struct request; From 21671a1ed1ff22e158ebe9d619943f926f03f5cd Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 2 Jul 2024 17:10:20 +0200 Subject: [PATCH 183/209] block: also return bio_integrity_payload * from stubs struct bio_integrity_payload is defined unconditionally. No need to return void * from bio_integrity() and bio_integrity_alloc(). Signed-off-by: Christoph Hellwig Reviewed-by: Kanchan Joshi Reviewed-by: Anuj Gupta Reviewed-by: Johannes Thumshirn Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240702151047.1746127-3-hch@lst.de Signed-off-by: Jens Axboe --- include/linux/bio-integrity.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/include/linux/bio-integrity.h b/include/linux/bio-integrity.h index 70ef19a0dc7e8b..cac24dac06fff0 100644 --- a/include/linux/bio-integrity.h +++ b/include/linux/bio-integrity.h @@ -85,7 +85,7 @@ void bio_integrity_init(void); #else /* CONFIG_BLK_DEV_INTEGRITY */ -static inline void *bio_integrity(struct bio *bio) +static inline struct bio_integrity_payload *bio_integrity(struct bio *bio) { return NULL; } @@ -138,8 +138,8 @@ static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag) return false; } -static inline void *bio_integrity_alloc(struct bio *bio, gfp_t gfp, - unsigned int nr) +static inline struct bio_integrity_payload * +bio_integrity_alloc(struct bio *bio, gfp_t gfp, unsigned int nr) { return ERR_PTR(-EINVAL); } From bf4c89fc8797f5c0964a0c3d561fbe7e8483b62f Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 2 Jul 2024 17:10:21 +0200 Subject: [PATCH 184/209] block: don't call bio_uninit from bio_endio Commit b222dd2fdd53 ("block: call bio_uninit in bio_endio") added a call to bio_uninit in bio_endio to work around callers that use bio_init but fail to call bio_uninit after they are done to release the resources. While this is an abuse of the bio_init API we still have quite a few of those left. But this early uninit causes a problem for integrity data, as at least some users need the bio_integrity_payload. Right now the only one is the NVMe passthrough which archives this by adding a special case to skip the freeing if the BIP_INTEGRITY_USER flag is set. Sort this out by only putting bi_blkg in bio_endio as that is the cause of the actual leaks - the few users of the crypto context and integrity data all properly call bio_uninit, usually through bio_put for dynamically allocated bios. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240702151047.1746127-4-hch@lst.de Signed-off-by: Jens Axboe --- block/bio.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) diff --git a/block/bio.c b/block/bio.c index 4ca3f31ce45fb5..68ce75fd9b7c89 100644 --- a/block/bio.c +++ b/block/bio.c @@ -1630,8 +1630,18 @@ void bio_endio(struct bio *bio) goto again; } - /* release cgroup info */ - bio_uninit(bio); +#ifdef CONFIG_BLK_CGROUP + /* + * Release cgroup info. We shouldn't have to do this here, but quite + * a few callers of bio_init fail to call bio_uninit, so we cover up + * for that here at least for now. + */ + if (bio->bi_blkg) { + blkg_put(bio->bi_blkg); + bio->bi_blkg = NULL; + } +#endif + if (bio->bi_end_io) bio->bi_end_io(bio); } From f8924374fd37a8b41d554acd8b7407af7d354c0d Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 2 Jul 2024 17:10:22 +0200 Subject: [PATCH 185/209] block: call bio_integrity_unmap_free_user from blk_rq_unmap_user blk_rq_unmap_user always unmaps user space pass-through request. If such a request has integrity data attached it must come from a user mapping as well. Call bio_integrity_unmap_free_user from blk_rq_unmap_user and remove the nvme_unmap_bio wrapper in the nvme driver. Signed-off-by: Christoph Hellwig Reviewed-by: Kanchan Joshi Reviewed-by: Anuj Gupta Reviewed-by: Johannes Thumshirn Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240702151047.1746127-5-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 1 - block/blk-map.c | 3 +++ drivers/nvme/host/ioctl.c | 15 ++++----------- 3 files changed, 7 insertions(+), 12 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index c4aed1dfa497a3..c8757d47e0ef62 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -174,7 +174,6 @@ void bio_integrity_unmap_free_user(struct bio *bio) bio->bi_integrity = NULL; bio->bi_opf &= ~REQ_INTEGRITY; } -EXPORT_SYMBOL(bio_integrity_unmap_free_user); /** * bio_integrity_add_page - Attach integrity metadata diff --git a/block/blk-map.c b/block/blk-map.c index bce144091128f6..df5f82d114720f 100644 --- a/block/blk-map.c +++ b/block/blk-map.c @@ -757,6 +757,9 @@ int blk_rq_unmap_user(struct bio *bio) bio_release_pages(bio, bio_data_dir(bio) == READ); } + if (bio_integrity(bio)) + bio_integrity_unmap_free_user(bio); + next_bio = bio; bio = bio->bi_next; blk_mq_map_bio_put(next_bio); diff --git a/drivers/nvme/host/ioctl.c b/drivers/nvme/host/ioctl.c index fb46f55f8b2894..f1d58e70933f54 100644 --- a/drivers/nvme/host/ioctl.c +++ b/drivers/nvme/host/ioctl.c @@ -112,13 +112,6 @@ static struct request *nvme_alloc_user_request(struct request_queue *q, return req; } -static void nvme_unmap_bio(struct bio *bio) -{ - if (bio_integrity(bio)) - bio_integrity_unmap_free_user(bio); - blk_rq_unmap_user(bio); -} - static int nvme_map_user_request(struct request *req, u64 ubuffer, unsigned bufflen, void __user *meta_buffer, unsigned meta_len, u32 meta_seed, struct io_uring_cmd *ioucmd, unsigned int flags) @@ -165,7 +158,7 @@ static int nvme_map_user_request(struct request *req, u64 ubuffer, out_unmap: if (bio) - nvme_unmap_bio(bio); + blk_rq_unmap_user(bio); out: blk_mq_free_request(req); return ret; @@ -203,7 +196,7 @@ static int nvme_submit_user_cmd(struct request_queue *q, if (result) *result = le64_to_cpu(nvme_req(req)->result.u64); if (bio) - nvme_unmap_bio(bio); + blk_rq_unmap_user(bio); blk_mq_free_request(req); if (effects) @@ -414,7 +407,7 @@ static void nvme_uring_task_cb(struct io_uring_cmd *ioucmd, struct nvme_uring_cmd_pdu *pdu = nvme_uring_cmd_pdu(ioucmd); if (pdu->bio) - nvme_unmap_bio(pdu->bio); + blk_rq_unmap_user(pdu->bio); io_uring_cmd_done(ioucmd, pdu->status, pdu->result, issue_flags); } @@ -440,7 +433,7 @@ static enum rq_end_io_ret nvme_uring_cmd_end_io(struct request *req, */ if (blk_rq_is_poll(req)) { if (pdu->bio) - nvme_unmap_bio(pdu->bio); + blk_rq_unmap_user(pdu->bio); io_uring_cmd_iopoll_done(ioucmd, pdu->result, pdu->status); } else { io_uring_cmd_do_in_task_lazy(ioucmd, nvme_uring_task_cb); From 85253bac4d02b1f95d6109c221aeccd7a262ec4d Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 2 Jul 2024 17:10:23 +0200 Subject: [PATCH 186/209] block: don't free submitter owned integrity payload on I/O completion Currently __bio_integrity_endio frees the integrity payload unless it is explicitly marked as user-mapped. This means in-kernel callers that allocate their own integrity payload never get to see it on I/O completion. The current two users don't need it as they just pre-mapped PI tuples received over the network, but this limits uses of integrity data lot. Change bio_integrity_endio to call __bio_integrity_endio for block layer generated integrity data only, and leave freeing of submitter allocated integrity data to bio_uninit which also gets called from the final bio_put. This requires that unmapping user mapped or copied integrity data is now always done by the caller, and the special BIP_INTEGRITY_USER flag can go away. Signed-off-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Reviewed-by: Kanchan Joshi Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240702151047.1746127-6-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 57 ++++++++++++++--------------------- block/blk.h | 13 ++++++-- include/linux/bio-integrity.h | 3 +- 3 files changed, 34 insertions(+), 39 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index c8757d47e0ef62..4aa836d603fb23 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -22,9 +22,17 @@ void blk_flush_integrity(void) flush_workqueue(kintegrityd_wq); } -static void __bio_integrity_free(struct bio_set *bs, - struct bio_integrity_payload *bip) +/** + * bio_integrity_free - Free bio integrity payload + * @bio: bio containing bip to be freed + * + * Description: Free the integrity portion of a bio. + */ +void bio_integrity_free(struct bio *bio) { + struct bio_integrity_payload *bip = bio_integrity(bio); + struct bio_set *bs = bio->bi_pool; + if (bs && mempool_initialized(&bs->bio_integrity_pool)) { if (bip->bip_vec) bvec_free(&bs->bvec_integrity_pool, bip->bip_vec, @@ -33,6 +41,8 @@ static void __bio_integrity_free(struct bio_set *bs, } else { kfree(bip); } + bio->bi_integrity = NULL; + bio->bi_opf &= ~REQ_INTEGRITY; } /** @@ -86,7 +96,10 @@ struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio, return bip; err: - __bio_integrity_free(bs, bip); + if (bs && mempool_initialized(&bs->bio_integrity_pool)) + mempool_free(bip, &bs->bio_integrity_pool); + else + kfree(bip); return ERR_PTR(-ENOMEM); } EXPORT_SYMBOL(bio_integrity_alloc); @@ -132,28 +145,6 @@ static void bio_integrity_unmap_user(struct bio_integrity_payload *bip) bio_integrity_unpin_bvec(bip->bip_vec, bip->bip_max_vcnt, dirty); } -/** - * bio_integrity_free - Free bio integrity payload - * @bio: bio containing bip to be freed - * - * Description: Used to free the integrity portion of a bio. Usually - * called from bio_free(). - */ -void bio_integrity_free(struct bio *bio) -{ - struct bio_integrity_payload *bip = bio_integrity(bio); - struct bio_set *bs = bio->bi_pool; - - if (bip->bip_flags & BIP_INTEGRITY_USER) - return; - if (bip->bip_flags & BIP_BLOCK_INTEGRITY) - kfree(bvec_virt(bip->bip_vec)); - - __bio_integrity_free(bs, bip); - bio->bi_integrity = NULL; - bio->bi_opf &= ~REQ_INTEGRITY; -} - /** * bio_integrity_unmap_free_user - Unmap and free bio user integrity payload * @bio: bio containing bip to be unmapped and freed @@ -165,14 +156,9 @@ void bio_integrity_free(struct bio *bio) void bio_integrity_unmap_free_user(struct bio *bio) { struct bio_integrity_payload *bip = bio_integrity(bio); - struct bio_set *bs = bio->bi_pool; - if (WARN_ON_ONCE(!(bip->bip_flags & BIP_INTEGRITY_USER))) - return; bio_integrity_unmap_user(bip); - __bio_integrity_free(bs, bip); - bio->bi_integrity = NULL; - bio->bi_opf &= ~REQ_INTEGRITY; + bio_integrity_free(bio); } /** @@ -273,7 +259,7 @@ static int bio_integrity_copy_user(struct bio *bio, struct bio_vec *bvec, goto free_bip; } - bip->bip_flags |= BIP_INTEGRITY_USER | BIP_COPY_USER; + bip->bip_flags |= BIP_COPY_USER; bip->bip_iter.bi_sector = seed; bip->bip_vcnt = nr_vecs; return 0; @@ -294,7 +280,6 @@ static int bio_integrity_init_user(struct bio *bio, struct bio_vec *bvec, return PTR_ERR(bip); memcpy(bip->bip_vec, bvec, nr_vecs * sizeof(*bvec)); - bip->bip_flags |= BIP_INTEGRITY_USER; bip->bip_iter.bi_sector = seed; bip->bip_iter.bi_size = len; bip->bip_vcnt = nr_vecs; @@ -502,6 +487,8 @@ static void bio_integrity_verify_fn(struct work_struct *work) struct bio *bio = bip->bip_bio; blk_integrity_verify(bio); + + kfree(bvec_virt(bip->bip_vec)); bio_integrity_free(bio); bio_endio(bio); } @@ -522,13 +509,13 @@ bool __bio_integrity_endio(struct bio *bio) struct blk_integrity *bi = blk_get_integrity(bio->bi_bdev->bd_disk); struct bio_integrity_payload *bip = bio_integrity(bio); - if (bio_op(bio) == REQ_OP_READ && !bio->bi_status && - (bip->bip_flags & BIP_BLOCK_INTEGRITY) && bi->csum_type) { + if (bio_op(bio) == REQ_OP_READ && !bio->bi_status && bi->csum_type) { INIT_WORK(&bip->bip_work, bio_integrity_verify_fn); queue_work(kintegrityd_wq, &bip->bip_work); return false; } + kfree(bvec_virt(bip->bip_vec)); bio_integrity_free(bio); return true; } diff --git a/block/blk.h b/block/blk.h index 401e604f35d2cf..2233dc8d36b82a 100644 --- a/block/blk.h +++ b/block/blk.h @@ -202,11 +202,20 @@ static inline unsigned int blk_queue_get_max_sectors(struct request *rq) #ifdef CONFIG_BLK_DEV_INTEGRITY void blk_flush_integrity(void); -bool __bio_integrity_endio(struct bio *); void bio_integrity_free(struct bio *bio); + +/* + * Integrity payloads can either be owned by the submitter, in which case + * bio_uninit will free them, or owned and generated by the block layer, + * in which case we'll verify them here (for reads) and free them before + * the bio is handed back to the submitted. + */ +bool __bio_integrity_endio(struct bio *bio); static inline bool bio_integrity_endio(struct bio *bio) { - if (bio_integrity(bio)) + struct bio_integrity_payload *bip = bio_integrity(bio); + + if (bip && (bip->bip_flags & BIP_BLOCK_INTEGRITY)) return __bio_integrity_endio(bio); return true; } diff --git a/include/linux/bio-integrity.h b/include/linux/bio-integrity.h index cac24dac06fff0..3823d9be0d0790 100644 --- a/include/linux/bio-integrity.h +++ b/include/linux/bio-integrity.h @@ -10,8 +10,7 @@ enum bip_flags { BIP_CTRL_NOCHECK = 1 << 2, /* disable HBA integrity checking */ BIP_DISK_NOCHECK = 1 << 3, /* disable disk integrity checking */ BIP_IP_CHECKSUM = 1 << 4, /* IP checksum */ - BIP_INTEGRITY_USER = 1 << 5, /* Integrity payload is user address */ - BIP_COPY_USER = 1 << 6, /* Kernel bounce buffer in use */ + BIP_COPY_USER = 1 << 5, /* Kernel bounce buffer in use */ }; struct bio_integrity_payload { From 74cc150282e41c6c0704cd305c9a4392dc64ef4d Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 2 Jul 2024 17:10:24 +0200 Subject: [PATCH 187/209] block: don't free the integrity payload in bio_integrity_unmap_free_user Now that the integrity payload is always freed in bio_uninit, don't bother freeing it a little earlier in bio_integrity_unmap_free_user. With that the separate bio_integrity_unmap_free_user can go away by just passing the bio to bio_integrity_unmap_user. Signed-off-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Reviewed-by: Kanchan Joshi Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240702151047.1746127-7-hch@lst.de Signed-off-by: Jens Axboe --- block/bio-integrity.c | 31 +++++++++++-------------------- block/blk-map.c | 2 +- include/linux/bio-integrity.h | 4 ++-- 3 files changed, 14 insertions(+), 23 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index 4aa836d603fb23..4b5c604585561e 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -131,34 +131,25 @@ static void bio_integrity_uncopy_user(struct bio_integrity_payload *bip) bio_integrity_unpin_bvec(copy, nr_vecs, true); } -static void bio_integrity_unmap_user(struct bio_integrity_payload *bip) +/** + * bio_integrity_unmap_user - Unmap user integrity payload + * @bio: bio containing bip to be unmapped + * + * Unmap the user mapped integrity portion of a bio. + */ +void bio_integrity_unmap_user(struct bio *bio) { - bool dirty = bio_data_dir(bip->bip_bio) == READ; + struct bio_integrity_payload *bip = bio_integrity(bio); if (bip->bip_flags & BIP_COPY_USER) { - if (dirty) + if (bio_data_dir(bio) == READ) bio_integrity_uncopy_user(bip); kfree(bvec_virt(bip->bip_vec)); return; } - bio_integrity_unpin_bvec(bip->bip_vec, bip->bip_max_vcnt, dirty); -} - -/** - * bio_integrity_unmap_free_user - Unmap and free bio user integrity payload - * @bio: bio containing bip to be unmapped and freed - * - * Description: Used to unmap and free the user mapped integrity portion of a - * bio. Submitter attaching the user integrity buffer is responsible for - * unmapping and freeing it during completion. - */ -void bio_integrity_unmap_free_user(struct bio *bio) -{ - struct bio_integrity_payload *bip = bio_integrity(bio); - - bio_integrity_unmap_user(bip); - bio_integrity_free(bio); + bio_integrity_unpin_bvec(bip->bip_vec, bip->bip_max_vcnt, + bio_data_dir(bio) == READ); } /** diff --git a/block/blk-map.c b/block/blk-map.c index df5f82d114720f..0e1167b239342f 100644 --- a/block/blk-map.c +++ b/block/blk-map.c @@ -758,7 +758,7 @@ int blk_rq_unmap_user(struct bio *bio) } if (bio_integrity(bio)) - bio_integrity_unmap_free_user(bio); + bio_integrity_unmap_user(bio); next_bio = bio; bio = bio->bi_next; diff --git a/include/linux/bio-integrity.h b/include/linux/bio-integrity.h index 3823d9be0d0790..dd831c269e9948 100644 --- a/include/linux/bio-integrity.h +++ b/include/linux/bio-integrity.h @@ -73,7 +73,7 @@ struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio, gfp_t gfp, int bio_integrity_add_page(struct bio *bio, struct page *page, unsigned int len, unsigned int offset); int bio_integrity_map_user(struct bio *bio, void __user *ubuf, ssize_t len, u32 seed); -void bio_integrity_unmap_free_user(struct bio *bio); +void bio_integrity_unmap_user(struct bio *bio); bool bio_integrity_prep(struct bio *bio); void bio_integrity_advance(struct bio *bio, unsigned int bytes_done); void bio_integrity_trim(struct bio *bio); @@ -104,7 +104,7 @@ static inline int bio_integrity_map_user(struct bio *bio, void __user *ubuf, return -EINVAL; } -static inline void bio_integrity_unmap_free_user(struct bio *bio) +static inline void bio_integrity_unmap_user(struct bio *bio) { } From 2314c2e3a70521f055dd011245dccf6fd97c7ee0 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Sat, 15 Jun 2024 16:51:43 +0800 Subject: [PATCH 188/209] md/raid5: fix spares errors about rcu usage As commit ad8606702f26 ("md/raid5: remove rcu protection to access rdev from conf") explains, rcu protection can be removed, however, there are three places left, there won't be any real problems. drivers/md/raid5.c:8071:24: error: incompatible types in comparison expression (different address spaces): drivers/md/raid5.c:8071:24: struct md_rdev [noderef] __rcu * drivers/md/raid5.c:8071:24: struct md_rdev * drivers/md/raid5.c:7569:25: error: incompatible types in comparison expression (different address spaces): drivers/md/raid5.c:7569:25: struct md_rdev [noderef] __rcu * drivers/md/raid5.c:7569:25: struct md_rdev * drivers/md/raid5.c:7573:25: error: incompatible types in comparison expression (different address spaces): drivers/md/raid5.c:7573:25: struct md_rdev [noderef] __rcu * drivers/md/raid5.c:7573:25: struct md_rdev * Fixes: ad8606702f26 ("md/raid5: remove rcu protection to access rdev from conf") Cc: stable@vger.kernel.org Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240615085143.1648223-1-yukuai1@huaweicloud.com --- drivers/md/raid5.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 10219205160bbf..825fa629db33a3 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -155,7 +155,7 @@ static int raid6_idx_to_slot(int idx, struct stripe_head *sh, return slot; } -static void print_raid5_conf (struct r5conf *conf); +static void print_raid5_conf(struct r5conf *conf); static int stripe_operations_active(struct stripe_head *sh) { @@ -7568,11 +7568,11 @@ static struct r5conf *setup_conf(struct mddev *mddev) if (test_bit(Replacement, &rdev->flags)) { if (disk->replacement) goto abort; - RCU_INIT_POINTER(disk->replacement, rdev); + disk->replacement = rdev; } else { if (disk->rdev) goto abort; - RCU_INIT_POINTER(disk->rdev, rdev); + disk->rdev = rdev; } if (test_bit(In_sync, &rdev->flags)) { @@ -8054,7 +8054,7 @@ static void raid5_status(struct seq_file *seq, struct mddev *mddev) seq_printf (seq, "]"); } -static void print_raid5_conf (struct r5conf *conf) +static void print_raid5_conf(struct r5conf *conf) { struct md_rdev *rdev; int i; @@ -8068,15 +8068,13 @@ static void print_raid5_conf (struct r5conf *conf) conf->raid_disks, conf->raid_disks - conf->mddev->degraded); - rcu_read_lock(); for (i = 0; i < conf->raid_disks; i++) { - rdev = rcu_dereference(conf->disks[i].rdev); + rdev = conf->disks[i].rdev; if (rdev) pr_debug(" disk %d, o:%d, dev:%pg\n", i, !test_bit(Faulty, &rdev->flags), rdev->bdev); } - rcu_read_unlock(); } static int raid5_spare_active(struct mddev *mddev) From ae720670b9fc5ef3588efd5b95e6a0f59a36dec0 Mon Sep 17 00:00:00 2001 From: Yang Li Date: Tue, 18 Jun 2024 09:07:59 +0800 Subject: [PATCH 189/209] md: Remove unneeded semicolon ./drivers/md/md.c:630:21-22: Unneeded semicolon Reported-by: Abaci Robot Closes: https://bugzilla.openanolis.cn/show_bug.cgi?id=9344 Signed-off-by: Yang Li Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240618010759.85416-1-yang.lee@linux.alibaba.com --- drivers/md/md.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index 0ff26a547f1afc..149be072546a59 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -627,7 +627,7 @@ static void md_submit_flush_data(struct work_struct *ws) * always is 0, make_request() will not be called here. */ if (WARN_ON_ONCE(!mddev->pers->make_request(mddev, bio))) - bio_io_error(bio);; + bio_io_error(bio); } /* The pair is percpu_ref_get() from md_flush_request() */ From 1f4a72ff00cafa74b43b0c8a37573c78f86ed1a8 Mon Sep 17 00:00:00 2001 From: Christophe JAILLET Date: Sun, 23 Jun 2024 22:17:54 +0200 Subject: [PATCH 190/209] md-cluster: Constify struct md_cluster_operations 'struct md_cluster_operations' is not modified in this driver. Constifying this structure moves some data to a read-only section, so increase overall security. On a x86_64, with allmodconfig, as an example: Before: ====== text data bss dec hex filename 51941 1442 80 53463 d0d7 drivers/md/md-cluster.o After: ===== text data bss dec hex filename 52133 1246 80 53459 d0d3 drivers/md/md-cluster.o Signed-off-by: Christophe JAILLET Signed-off-by: Song Liu Link: https://lore.kernel.org/r/3727f3ce9693cae4e62ae6778ea13971df805479.1719173852.git.christophe.jaillet@wanadoo.fr --- drivers/md/md-cluster.c | 2 +- drivers/md/md.c | 4 ++-- drivers/md/md.h | 4 ++-- 3 files changed, 5 insertions(+), 5 deletions(-) diff --git a/drivers/md/md-cluster.c b/drivers/md/md-cluster.c index 8e36a0feec0989..139fe2019c1dd6 100644 --- a/drivers/md/md-cluster.c +++ b/drivers/md/md-cluster.c @@ -1570,7 +1570,7 @@ static int gather_bitmaps(struct md_rdev *rdev) return err; } -static struct md_cluster_operations cluster_ops = { +static const struct md_cluster_operations cluster_ops = { .join = join, .leave = leave, .slot_number = slot_number, diff --git a/drivers/md/md.c b/drivers/md/md.c index 149be072546a59..b50721ddf5d37f 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -85,7 +85,7 @@ static DEFINE_SPINLOCK(pers_lock); static const struct kobj_type md_ktype; -struct md_cluster_operations *md_cluster_ops; +const struct md_cluster_operations *md_cluster_ops; EXPORT_SYMBOL(md_cluster_ops); static struct module *md_cluster_mod; @@ -8543,7 +8543,7 @@ int unregister_md_personality(struct md_personality *p) } EXPORT_SYMBOL(unregister_md_personality); -int register_md_cluster_operations(struct md_cluster_operations *ops, +int register_md_cluster_operations(const struct md_cluster_operations *ops, struct module *module) { int ret = 0; diff --git a/drivers/md/md.h b/drivers/md/md.h index 28cb4b0b6c1740..a0d6827dced9b7 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -849,7 +849,7 @@ static inline void safe_put_page(struct page *p) extern int register_md_personality(struct md_personality *p); extern int unregister_md_personality(struct md_personality *p); -extern int register_md_cluster_operations(struct md_cluster_operations *ops, +extern int register_md_cluster_operations(const struct md_cluster_operations *ops, struct module *module); extern int unregister_md_cluster_operations(void); extern int md_setup_cluster(struct mddev *mddev, int nodes); @@ -932,7 +932,7 @@ static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev) } } -extern struct md_cluster_operations *md_cluster_ops; +extern const struct md_cluster_operations *md_cluster_ops; static inline int mddev_is_clustered(struct mddev *mddev) { return mddev->cluster_info && mddev->bitmap_info.nodes > 1; From a1fd37f97808db4fa1bf55da0275790c42521e45 Mon Sep 17 00:00:00 2001 From: Yu Kuai Date: Thu, 27 Jun 2024 19:23:21 +0800 Subject: [PATCH 191/209] md: Don't wait for MD_RECOVERY_NEEDED for HOT_REMOVE_DISK ioctl Commit 90f5f7ad4f38 ("md: Wait for md_check_recovery before attempting device removal.") explained in the commit message that failed device must be reomoved from the personality first by md_check_recovery(), before it can be removed from the array. That's the reason the commit add the code to wait for MD_RECOVERY_NEEDED. However, this is not the case now, because remove_and_add_spares() is called directly from hot_remove_disk() from ioctl path, hence failed device(marked faulty) can be removed from the personality by ioctl. On the other hand, the commit introduced a performance problem that if MD_RECOVERY_NEEDED is set and the array is not running, ioctl will wait for 5s before it can return failure to user. Since the waiting is not needed now, fix the problem by removing the waiting. Fixes: 90f5f7ad4f38 ("md: Wait for md_check_recovery before attempting device removal.") Reported-by: Mateusz Kusiak Closes: https://lore.kernel.org/all/814ff6ee-47a2-4ba0-963e-cf256ee4ecfa@linux.intel.com/ Signed-off-by: Yu Kuai Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240627112321.3044744-1-yukuai1@huaweicloud.com --- drivers/md/md.c | 6 ------ 1 file changed, 6 deletions(-) diff --git a/drivers/md/md.c b/drivers/md/md.c index b50721ddf5d37f..64693913ed186e 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -7765,12 +7765,6 @@ static int md_ioctl(struct block_device *bdev, blk_mode_t mode, return get_bitmap_file(mddev, argp); } - if (cmd == HOT_REMOVE_DISK) - /* need to ensure recovery thread has run */ - wait_event_interruptible_timeout(mddev->sb_wait, - !test_bit(MD_RECOVERY_NEEDED, - &mddev->recovery), - msecs_to_jiffies(5000)); if (cmd == STOP_ARRAY || cmd == STOP_ARRAY_RO) { /* Need to flush page cache, and ensure no-one else opens * and writes From 25b3a8237a03ec0b67b965b52d74862e77ef7115 Mon Sep 17 00:00:00 2001 From: Benjamin Marzinski Date: Tue, 2 Jul 2024 11:18:02 -0400 Subject: [PATCH 192/209] md/raid5: recheck if reshape has finished with device_lock held When handling an IO request, MD checks if a reshape is currently happening, and if so, where the IO sector is in relation to the reshape progress. MD uses conf->reshape_progress for both of these tasks. When the reshape finishes, conf->reshape_progress is set to MaxSector. If this occurs after MD checks if the reshape is currently happening but before it calls ahead_of_reshape(), then ahead_of_reshape() will end up comparing the IO sector against MaxSector. During a backwards reshape, this will make MD think the IO sector is in the area not yet reshaped, causing it to use the previous configuration, and map the IO to the sector where that data was before the reshape. This bug can be triggered by running the lvm2 lvconvert-raid-reshape-linear_to_raid6-single-type.sh test in a loop, although it's very hard to reproduce. Fix this by factoring the code that checks where the IO sector is in relation to the reshape out to a helper called get_reshape_loc(), which reads reshape_progress and reshape_safe while holding the device_lock, and then rechecks if the reshape has finished before calling ahead_of_reshape with the saved values. Also use the helper during the REQ_NOWAIT check to see if the location is inside of the reshape region. Fixes: fef9c61fdfabf ("md/raid5: change reshape-progress measurement to cope with reshaping backwards.") Signed-off-by: Benjamin Marzinski Signed-off-by: Song Liu Link: https://lore.kernel.org/r/20240702151802.1632010-1-bmarzins@redhat.com --- drivers/md/raid5.c | 64 +++++++++++++++++++++++++++++----------------- 1 file changed, 41 insertions(+), 23 deletions(-) diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index 825fa629db33a3..c14cf2410365dd 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -5899,6 +5899,39 @@ static int add_all_stripe_bios(struct r5conf *conf, return ret; } +enum reshape_loc { + LOC_NO_RESHAPE, + LOC_AHEAD_OF_RESHAPE, + LOC_INSIDE_RESHAPE, + LOC_BEHIND_RESHAPE, +}; + +static enum reshape_loc get_reshape_loc(struct mddev *mddev, + struct r5conf *conf, sector_t logical_sector) +{ + sector_t reshape_progress, reshape_safe; + /* + * Spinlock is needed as reshape_progress may be + * 64bit on a 32bit platform, and so it might be + * possible to see a half-updated value + * Of course reshape_progress could change after + * the lock is dropped, so once we get a reference + * to the stripe that we think it is, we will have + * to check again. + */ + spin_lock_irq(&conf->device_lock); + reshape_progress = conf->reshape_progress; + reshape_safe = conf->reshape_safe; + spin_unlock_irq(&conf->device_lock); + if (reshape_progress == MaxSector) + return LOC_NO_RESHAPE; + if (ahead_of_reshape(mddev, logical_sector, reshape_progress)) + return LOC_AHEAD_OF_RESHAPE; + if (ahead_of_reshape(mddev, logical_sector, reshape_safe)) + return LOC_INSIDE_RESHAPE; + return LOC_BEHIND_RESHAPE; +} + static enum stripe_result make_stripe_request(struct mddev *mddev, struct r5conf *conf, struct stripe_request_ctx *ctx, sector_t logical_sector, struct bio *bi) @@ -5913,28 +5946,14 @@ static enum stripe_result make_stripe_request(struct mddev *mddev, seq = read_seqcount_begin(&conf->gen_lock); if (unlikely(conf->reshape_progress != MaxSector)) { - /* - * Spinlock is needed as reshape_progress may be - * 64bit on a 32bit platform, and so it might be - * possible to see a half-updated value - * Of course reshape_progress could change after - * the lock is dropped, so once we get a reference - * to the stripe that we think it is, we will have - * to check again. - */ - spin_lock_irq(&conf->device_lock); - if (ahead_of_reshape(mddev, logical_sector, - conf->reshape_progress)) { - previous = 1; - } else { - if (ahead_of_reshape(mddev, logical_sector, - conf->reshape_safe)) { - spin_unlock_irq(&conf->device_lock); - ret = STRIPE_SCHEDULE_AND_RETRY; - goto out; - } + enum reshape_loc loc = get_reshape_loc(mddev, conf, + logical_sector); + if (loc == LOC_INSIDE_RESHAPE) { + ret = STRIPE_SCHEDULE_AND_RETRY; + goto out; } - spin_unlock_irq(&conf->device_lock); + if (loc == LOC_AHEAD_OF_RESHAPE) + previous = 1; } new_sector = raid5_compute_sector(conf, logical_sector, previous, @@ -6112,8 +6131,7 @@ static bool raid5_make_request(struct mddev *mddev, struct bio * bi) /* Bail out if conflicts with reshape and REQ_NOWAIT is set */ if ((bi->bi_opf & REQ_NOWAIT) && (conf->reshape_progress != MaxSector) && - !ahead_of_reshape(mddev, logical_sector, conf->reshape_progress) && - ahead_of_reshape(mddev, logical_sector, conf->reshape_safe)) { + get_reshape_loc(mddev, conf, logical_sector) == LOC_INSIDE_RESHAPE) { bio_wouldblock_error(bi); if (rw == WRITE) md_write_end(mddev); From 162e06871e6dcde861ef608e0c00a8b6a2d35d43 Mon Sep 17 00:00:00 2001 From: Anuj Gupta Date: Thu, 4 Jul 2024 11:45:15 +0530 Subject: [PATCH 193/209] block: t10-pi: Return correct ref tag when queue has no integrity profile Commit c6e56cf6b2e7 ("block: move integrity information into queue_limits") changed the ref tag calculation logic. It would break if there is no integrity profile. This in turn causes read/write failures for such cases. Fixes: c6e56cf6b2e7 ("block: move integrity information into queue_limits") Signed-off-by: Anuj Gupta Link: https://lore.kernel.org/r/20240704061515.282343-1-joshi.k@samsung.com Signed-off-by: Jens Axboe --- include/linux/t10-pi.h | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/include/linux/t10-pi.h b/include/linux/t10-pi.h index 1773610010ebaf..2c59fe3efcd424 100644 --- a/include/linux/t10-pi.h +++ b/include/linux/t10-pi.h @@ -39,8 +39,11 @@ struct t10_pi_tuple { static inline u32 t10_pi_ref_tag(struct request *rq) { - unsigned int shift = rq->q->limits.integrity.interval_exp; + unsigned int shift = ilog2(queue_logical_block_size(rq->q)); + if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && + rq->q->limits.integrity.interval_exp) + shift = rq->q->limits.integrity.interval_exp; return blk_rq_pos(rq) >> (shift - SECTOR_SHIFT) & 0xffffffff; } @@ -61,8 +64,11 @@ static inline u64 lower_48_bits(u64 n) static inline u64 ext_pi_ref_tag(struct request *rq) { - unsigned int shift = rq->q->limits.integrity.interval_exp; + unsigned int shift = ilog2(queue_logical_block_size(rq->q)); + if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && + rq->q->limits.integrity.interval_exp) + shift = rq->q->limits.integrity.interval_exp; return lower_48_bits(blk_rq_pos(rq) >> (shift - SECTOR_SHIFT)); } From a18df07b7d3dbfa7ae54962cc59569002eaafd6d Mon Sep 17 00:00:00 2001 From: Zhu Yanjun Date: Thu, 4 Jul 2024 03:06:38 +0200 Subject: [PATCH 194/209] null_blk: don't initialize static 'g_virt_boundary' to false No functional changes intended. Signed-off-by: Zhu Yanjun Link: https://lore.kernel.org/r/20240704010638.324349-1-yanjun.zhu@linux.dev Signed-off-by: Jens Axboe --- drivers/block/null_blk/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c index 5de9ca4eceb4b4..cbf4b29b16876d 100644 --- a/drivers/block/null_blk/main.c +++ b/drivers/block/null_blk/main.c @@ -77,7 +77,7 @@ enum { NULL_IRQ_TIMER = 2, }; -static bool g_virt_boundary = false; +static bool g_virt_boundary; module_param_named(virt_boundary, g_virt_boundary, bool, 0444); MODULE_PARM_DESC(virt_boundary, "Require a virtual boundary for the device. Default: False"); From ba942238056584efd3adc278a76592258d500918 Mon Sep 17 00:00:00 2001 From: Anuj Gupta Date: Tue, 2 Jul 2024 15:37:53 +0530 Subject: [PATCH 195/209] block: reuse original bio_vec array for integrity during clone Modify bio_integrity_clone to reuse the original bvec array instead of allocating and copying it, similar to how bio data path is cloned. Suggested-by: Christoph Hellwig Signed-off-by: Anuj Gupta Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Reviewed-by: Ming Lei Link: https://lore.kernel.org/r/20240702100753.2168-1-anuj20.g@samsung.com Signed-off-by: Jens Axboe --- block/bio-integrity.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) diff --git a/block/bio-integrity.c b/block/bio-integrity.c index eb3d7bbe1fe83a..0dc22985fed606 100644 --- a/block/bio-integrity.c +++ b/block/bio-integrity.c @@ -76,7 +76,7 @@ struct bio_integrity_payload *bio_integrity_alloc(struct bio *bio, &bip->bip_max_vcnt, gfp_mask); if (!bip->bip_vec) goto err; - } else { + } else if (nr_vecs) { bip->bip_vec = bip->bip_inline_vecs; } @@ -562,14 +562,11 @@ int bio_integrity_clone(struct bio *bio, struct bio *bio_src, BUG_ON(bip_src == NULL); - bip = bio_integrity_alloc(bio, gfp_mask, bip_src->bip_vcnt); + bip = bio_integrity_alloc(bio, gfp_mask, 0); if (IS_ERR(bip)) return PTR_ERR(bip); - memcpy(bip->bip_vec, bip_src->bip_vec, - bip_src->bip_vcnt * sizeof(struct bio_vec)); - - bip->bip_vcnt = bip_src->bip_vcnt; + bip->bip_vec = bip_src->bip_vec; bip->bip_iter = bip_src->bip_iter; bip->bip_flags = bip_src->bip_flags & ~BIP_BLOCK_INTEGRITY; From dd54fd4e1780b349043f2d74784b8b702dbd84e9 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 5 Jul 2024 07:31:14 +0200 Subject: [PATCH 196/209] loop: remove the unused inode variable in loop_configure Remove the inode variable now that the last user is gone. Fixes: a17ece76bcfe ("loop: regularize upgrading the block size for direct I/O") Reported-by: kernel test robot Signed-off-by: Christoph Hellwig Link: https://lore.kernel.org/r/20240705053114.2042976-1-hch@lst.de Signed-off-by: Jens Axboe --- drivers/block/loop.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 4f0d96876b1f5d..a6f2d63678988b 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -996,7 +996,6 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, const struct loop_config *config) { struct file *file = fget(config->fd); - struct inode *inode; struct address_space *mapping; int error; loff_t size; @@ -1033,7 +1032,6 @@ static int loop_configure(struct loop_device *lo, blk_mode_t mode, goto out_unlock; mapping = file->f_mapping; - inode = mapping->host; if ((config->info.lo_flags & ~LOOP_CONFIGURE_SETTABLE_FLAGS) != 0) { error = -EINVAL; From f4d5dc33c823ef1d7ccbbd2d1e40b871fad0012b Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Thu, 4 Jul 2024 14:28:12 +0900 Subject: [PATCH 197/209] null_blk: Introduce the zone_full parameter Allow creating a zoned null_blk device with the initial state of its sequential write required zones to be FULL. This is convenient to avoid having to first write these zones to perform read performance evaluation or test zone management operations such as zone reset (and zone reset all). Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240704052816.623865-2-dlemoal@kernel.org Signed-off-by: Jens Axboe --- drivers/block/null_blk/main.c | 9 ++++++++- drivers/block/null_blk/null_blk.h | 1 + drivers/block/null_blk/zoned.c | 10 ++++++++-- 3 files changed, 17 insertions(+), 3 deletions(-) diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c index cbf4b29b16876d..9d0f6da77601b4 100644 --- a/drivers/block/null_blk/main.c +++ b/drivers/block/null_blk/main.c @@ -262,6 +262,10 @@ module_param_named(zone_append_max_sectors, g_zone_append_max_sectors, int, 0444 MODULE_PARM_DESC(zone_append_max_sectors, "Maximum size of a zone append command (in 512B sectors). Specify 0 for zone append emulation"); +static bool g_zone_full; +module_param_named(zone_full, g_zone_full, bool, S_IRUGO); +MODULE_PARM_DESC(zone_full, "Initialize the sequential write required zones of a zoned device to be full. Default: false"); + static struct nullb_device *null_alloc_dev(void); static void null_free_dev(struct nullb_device *dev); static void null_del_dev(struct nullb *nullb); @@ -458,6 +462,7 @@ NULLB_DEVICE_ATTR(zone_nr_conv, uint, NULL); NULLB_DEVICE_ATTR(zone_max_open, uint, NULL); NULLB_DEVICE_ATTR(zone_max_active, uint, NULL); NULLB_DEVICE_ATTR(zone_append_max_sectors, uint, NULL); +NULLB_DEVICE_ATTR(zone_full, bool, NULL); NULLB_DEVICE_ATTR(virt_boundary, bool, NULL); NULLB_DEVICE_ATTR(no_sched, bool, NULL); NULLB_DEVICE_ATTR(shared_tags, bool, NULL); @@ -610,6 +615,7 @@ static struct configfs_attribute *nullb_device_attrs[] = { &nullb_device_attr_zone_append_max_sectors, &nullb_device_attr_zone_readonly, &nullb_device_attr_zone_offline, + &nullb_device_attr_zone_full, &nullb_device_attr_virt_boundary, &nullb_device_attr_no_sched, &nullb_device_attr_shared_tags, @@ -700,7 +706,7 @@ static ssize_t memb_group_features_show(struct config_item *item, char *page) "shared_tags,size,submit_queues,use_per_node_hctx," "virt_boundary,zoned,zone_capacity,zone_max_active," "zone_max_open,zone_nr_conv,zone_offline,zone_readonly," - "zone_size,zone_append_max_sectors\n"); + "zone_size,zone_append_max_sectors,zone_full\n"); } CONFIGFS_ATTR_RO(memb_group_, features); @@ -781,6 +787,7 @@ static struct nullb_device *null_alloc_dev(void) dev->zone_max_open = g_zone_max_open; dev->zone_max_active = g_zone_max_active; dev->zone_append_max_sectors = g_zone_append_max_sectors; + dev->zone_full = g_zone_full; dev->virt_boundary = g_virt_boundary; dev->no_sched = g_no_sched; dev->shared_tags = g_shared_tags; diff --git a/drivers/block/null_blk/null_blk.h b/drivers/block/null_blk/null_blk.h index 3234e6c85eed8c..a7bb32f73ec36a 100644 --- a/drivers/block/null_blk/null_blk.h +++ b/drivers/block/null_blk/null_blk.h @@ -101,6 +101,7 @@ struct nullb_device { bool memory_backed; /* if data is stored in memory */ bool discard; /* if support discard */ bool zoned; /* if device is zoned */ + bool zone_full; /* Initialize zones to be full */ bool virt_boundary; /* virtual boundary on/off for the device */ bool no_sched; /* no IO scheduler for the device */ bool shared_tags; /* share tag set between devices for blk-mq */ diff --git a/drivers/block/null_blk/zoned.c b/drivers/block/null_blk/zoned.c index 9f7151ad93cfc4..7996e2e7dce28a 100644 --- a/drivers/block/null_blk/zoned.c +++ b/drivers/block/null_blk/zoned.c @@ -145,7 +145,7 @@ int null_init_zoned_dev(struct nullb_device *dev, zone = &dev->zones[i]; null_init_zone_lock(dev, zone); - zone->start = zone->wp = sector; + zone->start = sector; if (zone->start + dev->zone_size_sects > dev_capacity_sects) zone->len = dev_capacity_sects - zone->start; else @@ -153,7 +153,13 @@ int null_init_zoned_dev(struct nullb_device *dev, zone->capacity = min_t(sector_t, zone->len, zone_capacity_sects); zone->type = BLK_ZONE_TYPE_SEQWRITE_REQ; - zone->cond = BLK_ZONE_COND_EMPTY; + if (dev->zone_full) { + zone->cond = BLK_ZONE_COND_FULL; + zone->wp = zone->start + zone->capacity; + } else{ + zone->cond = BLK_ZONE_COND_EMPTY; + zone->wp = zone->start; + } sector += dev->zone_size_sects; } From ae7e965b36e3132238d16b4ccd223f65162397b5 Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Thu, 4 Jul 2024 14:28:13 +0900 Subject: [PATCH 198/209] dm: Refactor is_abnormal_io() Use a single switch-case to simplify is_abnormal_io() and make this function more readable and easier to modify. Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240704052816.623865-3-dlemoal@kernel.org Signed-off-by: Jens Axboe --- drivers/md/dm.c | 24 +++++++++++------------- 1 file changed, 11 insertions(+), 13 deletions(-) diff --git a/drivers/md/dm.c b/drivers/md/dm.c index 7d107ae06e1ae1..0d80caccbd9eec 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -1598,20 +1598,18 @@ static void __send_abnormal_io(struct clone_info *ci, struct dm_target *ti, static bool is_abnormal_io(struct bio *bio) { - enum req_op op = bio_op(bio); - - if (op != REQ_OP_READ && op != REQ_OP_WRITE && op != REQ_OP_FLUSH) { - switch (op) { - case REQ_OP_DISCARD: - case REQ_OP_SECURE_ERASE: - case REQ_OP_WRITE_ZEROES: - return true; - default: - break; - } + switch (bio_op(bio)) { + case REQ_OP_READ: + case REQ_OP_WRITE: + case REQ_OP_FLUSH: + return false; + case REQ_OP_DISCARD: + case REQ_OP_SECURE_ERASE: + case REQ_OP_WRITE_ZEROES: + return true; + default: + return false; } - - return false; } static blk_status_t __process_abnormal_io(struct clone_info *ci, From 81e7706345f06e1e97a092f59697b7e20a0ee868 Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Thu, 4 Jul 2024 14:28:14 +0900 Subject: [PATCH 199/209] dm: handle REQ_OP_ZONE_RESET_ALL This commit implements processing of the REQ_OP_ZONE_RESET_ALL operation for zoned mapped devices. Given that this operation always has a BIO sector of 0 and a 0 size, processing through the regular BIO __split_and_process_bio() function does not work because this function would always select the first target. Instead, handling of this operation is implemented using the function __send_zone_reset_all(). Similarly to the __send_empty_flush() function, the new __send_zone_reset_all() function manually goes through all targets of a mapped device table doing the following: 1) If the target can natively support REQ_OP_ZONE_RESET_ALL, __send_duplicate_bios() is used to forward the reset all operation to the target. This case is handled with the __send_zone_reset_all_native() function. 2) For other targets, the function __send_zone_reset_all_emulated() is executed to emulate the execution of REQ_OP_ZONE_RESET_ALL using regular REQ_OP_ZONE_RESET operations. Targets that can natively support REQ_OP_ZONE_RESET_ALL are identified using the new target field zone_reset_all_supported. This boolean is set to true in for targets that have reliable zone limits, that is, targets that map all sequential write required zones of their zoned device(s). Setting this field is handled in dm_set_zones_restrictions() and device_get_zone_resource_limits(). For targets with unreliable zone limits, REQ_OP_ZONE_RESET_ALL must be emulated (case 2 above). This is implemented with __send_zone_reset_all_emulated() and is similar to the block layer function blkdev_zone_reset_all_emulated(): first a report zones is done for the zones of the target to identify zones that need reset, that is, any sequential write required zone that is not already empty. This is done using a bitmap and the function dm_zone_get_reset_bitmap() which sets to 1 the bit corresponding to a zone that needs reset. Next, this zone bitmap is inspected and a clone BIO modified to use the REQ_OP_ZONE_RESET operation issued for any zone with its bit set in the zone bitmap. This implementation is more efficient than what the block layer does with blkdev_zone_reset_all_emulated(), which is always used for DM zoned devices currently: as we can natively use REQ_OP_ZONE_RESET_ALL on targets mapping all sequential write required zones, resetting all zones of a zoned mapped device can be much faster compared to always emulating this operation using regular per-zone reset. In the worst case, this implementation is as-efficient as the block layer emulation. This reduction in the time it takes to reset all zones of a zoned mapped device depends directly on the mapped device targets mapping (reliable zone limits or not). Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240704052816.623865-4-dlemoal@kernel.org Signed-off-by: Jens Axboe --- drivers/md/dm-zone.c | 50 ++++++++++++- drivers/md/dm.c | 135 +++++++++++++++++++++++++++++++++- drivers/md/dm.h | 3 + include/linux/device-mapper.h | 7 ++ 4 files changed, 190 insertions(+), 5 deletions(-) diff --git a/drivers/md/dm-zone.c b/drivers/md/dm-zone.c index 4d37e53b50ee64..c0d41c36e06ebf 100644 --- a/drivers/md/dm-zone.c +++ b/drivers/md/dm-zone.c @@ -292,10 +292,12 @@ static int device_get_zone_resource_limits(struct dm_target *ti, /* * If the target does not map all sequential zones, the limits - * will not be reliable. + * will not be reliable and we cannot use REQ_OP_ZONE_RESET_ALL. */ - if (zc.target_nr_seq_zones < zc.total_nr_seq_zones) + if (zc.target_nr_seq_zones < zc.total_nr_seq_zones) { zlim->reliable_limits = false; + ti->zone_reset_all_supported = false; + } /* * If the target maps less sequential zones than the limit values, then @@ -353,6 +355,14 @@ int dm_set_zones_restrictions(struct dm_table *t, struct request_queue *q, for (unsigned int i = 0; i < t->num_targets; i++) { struct dm_target *ti = dm_table_get_target(t, i); + /* + * Assume that the target can accept REQ_OP_ZONE_RESET_ALL. + * device_get_zone_resource_limits() may adjust this if one of + * the device used by the target does not have all its + * sequential write required zones mapped. + */ + ti->zone_reset_all_supported = true; + if (!ti->type->iterate_devices || ti->type->iterate_devices(ti, device_get_zone_resource_limits, &zlim)) { @@ -420,3 +430,39 @@ void dm_zone_endio(struct dm_io *io, struct bio *clone) return; } + +static int dm_zone_need_reset_cb(struct blk_zone *zone, unsigned int idx, + void *data) +{ + /* + * For an all-zones reset, ignore conventional, empty, read-only + * and offline zones. + */ + switch (zone->cond) { + case BLK_ZONE_COND_NOT_WP: + case BLK_ZONE_COND_EMPTY: + case BLK_ZONE_COND_READONLY: + case BLK_ZONE_COND_OFFLINE: + return 0; + default: + set_bit(idx, (unsigned long *)data); + return 0; + } +} + +int dm_zone_get_reset_bitmap(struct mapped_device *md, struct dm_table *t, + sector_t sector, unsigned int nr_zones, + unsigned long *need_reset) +{ + int ret; + + ret = dm_blk_do_report_zones(md, t, sector, nr_zones, + dm_zone_need_reset_cb, need_reset); + if (ret != nr_zones) { + DMERR("Get %s zone reset bitmap failed\n", + md->disk->disk_name); + return -EIO; + } + + return 0; +} diff --git a/drivers/md/dm.c b/drivers/md/dm.c index 0d80caccbd9eec..4b1b69e576a550 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -1606,6 +1606,7 @@ static bool is_abnormal_io(struct bio *bio) case REQ_OP_DISCARD: case REQ_OP_SECURE_ERASE: case REQ_OP_WRITE_ZEROES: + case REQ_OP_ZONE_RESET_ALL: return true; default: return false; @@ -1774,6 +1775,119 @@ static inline bool dm_zone_plug_bio(struct mapped_device *md, struct bio *bio) { return dm_emulate_zone_append(md) && blk_zone_plug_bio(bio, 0); } + +static blk_status_t __send_zone_reset_all_emulated(struct clone_info *ci, + struct dm_target *ti) +{ + struct bio_list blist = BIO_EMPTY_LIST; + struct mapped_device *md = ci->io->md; + unsigned int zone_sectors = md->disk->queue->limits.chunk_sectors; + unsigned long *need_reset; + unsigned int i, nr_zones, nr_reset; + unsigned int num_bios = 0; + blk_status_t sts = BLK_STS_OK; + sector_t sector = ti->begin; + struct bio *clone; + int ret; + + nr_zones = ti->len >> ilog2(zone_sectors); + need_reset = bitmap_zalloc(nr_zones, GFP_NOIO); + if (!need_reset) + return BLK_STS_RESOURCE; + + ret = dm_zone_get_reset_bitmap(md, ci->map, ti->begin, + nr_zones, need_reset); + if (ret) { + sts = BLK_STS_IOERR; + goto free_bitmap; + } + + /* If we have no zone to reset, we are done. */ + nr_reset = bitmap_weight(need_reset, nr_zones); + if (!nr_reset) + goto free_bitmap; + + atomic_add(nr_zones, &ci->io->io_count); + + for (i = 0; i < nr_zones; i++) { + + if (!test_bit(i, need_reset)) { + sector += zone_sectors; + continue; + } + + if (bio_list_empty(&blist)) { + /* This may take a while, so be nice to others */ + if (num_bios) + cond_resched(); + + /* + * We may need to reset thousands of zones, so let's + * not go crazy with the clone allocation. + */ + alloc_multiple_bios(&blist, ci, ti, min(nr_reset, 32), + NULL, GFP_NOIO); + } + + /* Get a clone and change it to a regular reset operation. */ + clone = bio_list_pop(&blist); + clone->bi_opf &= ~REQ_OP_MASK; + clone->bi_opf |= REQ_OP_ZONE_RESET | REQ_SYNC; + clone->bi_iter.bi_sector = sector; + clone->bi_iter.bi_size = 0; + __map_bio(clone); + + sector += zone_sectors; + num_bios++; + nr_reset--; + } + + WARN_ON_ONCE(!bio_list_empty(&blist)); + atomic_sub(nr_zones - num_bios, &ci->io->io_count); + ci->sector_count = 0; + +free_bitmap: + bitmap_free(need_reset); + + return sts; +} + +static void __send_zone_reset_all_native(struct clone_info *ci, + struct dm_target *ti) +{ + unsigned int bios; + + atomic_add(1, &ci->io->io_count); + bios = __send_duplicate_bios(ci, ti, 1, NULL, GFP_NOIO); + atomic_sub(1 - bios, &ci->io->io_count); + + ci->sector_count = 0; +} + +static blk_status_t __send_zone_reset_all(struct clone_info *ci) +{ + struct dm_table *t = ci->map; + blk_status_t sts = BLK_STS_OK; + + for (unsigned int i = 0; i < t->num_targets; i++) { + struct dm_target *ti = dm_table_get_target(t, i); + + if (ti->zone_reset_all_supported) { + __send_zone_reset_all_native(ci, ti); + continue; + } + + sts = __send_zone_reset_all_emulated(ci, ti); + if (sts != BLK_STS_OK) + break; + } + + /* Release the reference that alloc_io() took for submission. */ + atomic_sub(1, &ci->io->io_count); + + return sts; +} + #else static inline bool dm_zone_bio_needs_split(struct mapped_device *md, struct bio *bio) @@ -1784,6 +1898,10 @@ static inline bool dm_zone_plug_bio(struct mapped_device *md, struct bio *bio) { return false; } +static blk_status_t __send_zone_reset_all(struct clone_info *ci) +{ + return BLK_STS_NOTSUPP; +} #endif /* @@ -1797,9 +1915,14 @@ static void dm_split_and_process_bio(struct mapped_device *md, blk_status_t error = BLK_STS_OK; bool is_abnormal, need_split; - need_split = is_abnormal = is_abnormal_io(bio); - if (static_branch_unlikely(&zoned_enabled)) - need_split = is_abnormal || dm_zone_bio_needs_split(md, bio); + is_abnormal = is_abnormal_io(bio); + if (static_branch_unlikely(&zoned_enabled)) { + /* Special case REQ_OP_ZONE_RESET_ALL as it cannot be split. */ + need_split = (bio_op(bio) != REQ_OP_ZONE_RESET_ALL) && + (is_abnormal || dm_zone_bio_needs_split(md, bio)); + } else { + need_split = is_abnormal; + } if (unlikely(need_split)) { /* @@ -1840,6 +1963,12 @@ static void dm_split_and_process_bio(struct mapped_device *md, goto out; } + if (static_branch_unlikely(&zoned_enabled) && + (bio_op(bio) == REQ_OP_ZONE_RESET_ALL)) { + error = __send_zone_reset_all(&ci); + goto out; + } + error = __split_and_process_bio(&ci); if (error || !ci.sector_count) goto out; diff --git a/drivers/md/dm.h b/drivers/md/dm.h index c984ecb64b1e89..cc466ad5cb1df2 100644 --- a/drivers/md/dm.h +++ b/drivers/md/dm.h @@ -110,6 +110,9 @@ int dm_blk_report_zones(struct gendisk *disk, sector_t sector, unsigned int nr_zones, report_zones_cb cb, void *data); bool dm_is_zone_write(struct mapped_device *md, struct bio *bio); int dm_zone_map_bio(struct dm_target_io *io); +int dm_zone_get_reset_bitmap(struct mapped_device *md, struct dm_table *t, + sector_t sector, unsigned int nr_zones, + unsigned long *need_reset); #else #define dm_blk_report_zones NULL static inline bool dm_is_zone_write(struct mapped_device *md, struct bio *bio) diff --git a/include/linux/device-mapper.h b/include/linux/device-mapper.h index 82b2195efaca78..15d28164bbbdb8 100644 --- a/include/linux/device-mapper.h +++ b/include/linux/device-mapper.h @@ -357,6 +357,13 @@ struct dm_target { */ bool discards_supported:1; + /* + * Automatically set by dm-core if this target supports + * REQ_OP_ZONE_RESET_ALL. Otherwise, this operation will be emulated + * using REQ_OP_ZONE_RESET. Target drivers must not set this manually. + */ + bool zone_reset_all_supported:1; + /* * Set if this target requires that discards be split on * 'max_discard_sectors' boundaries. From f2a7bea23710fceb99dac6da4ef82c3cc8932f7f Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Thu, 4 Jul 2024 14:28:15 +0900 Subject: [PATCH 200/209] block: Remove REQ_OP_ZONE_RESET_ALL emulation Now that device mapper can handle resetting all zones of a mapped zoned device using REQ_OP_ZONE_RESET_ALL, all zoned block device drivers support this operation. With this, the request queue feature BLK_FEAT_ZONE_RESETALL is not necessary and the emulation code in blk-zone.c can be removed. Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240704052816.623865-5-dlemoal@kernel.org Signed-off-by: Jens Axboe --- block/blk-core.c | 5 +-- block/blk-zoned.c | 76 ++-------------------------------- drivers/block/null_blk/zoned.c | 2 +- drivers/block/ublk_drv.c | 2 +- drivers/block/virtio_blk.c | 2 +- drivers/nvme/host/zns.c | 2 +- drivers/scsi/sd_zbc.c | 2 +- include/linux/blkdev.h | 5 --- 8 files changed, 9 insertions(+), 87 deletions(-) diff --git a/block/blk-core.c b/block/blk-core.c index 71b7622c523a30..02bceeb36f2c48 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -830,11 +830,8 @@ void submit_bio_noacct(struct bio *bio) case REQ_OP_ZONE_OPEN: case REQ_OP_ZONE_CLOSE: case REQ_OP_ZONE_FINISH: - if (!bdev_is_zoned(bio->bi_bdev)) - goto not_supported; - break; case REQ_OP_ZONE_RESET_ALL: - if (!bdev_is_zoned(bio->bi_bdev) || !blk_queue_zone_resetall(q)) + if (!bdev_is_zoned(bio->bi_bdev)) goto not_supported; break; case REQ_OP_DRV_IN: diff --git a/block/blk-zoned.c b/block/blk-zoned.c index 601c21a224c93e..0007ef9cd5cacd 100644 --- a/block/blk-zoned.c +++ b/block/blk-zoned.c @@ -157,70 +157,6 @@ static inline unsigned long *blk_alloc_zone_bitmap(int node, GFP_NOIO, node); } -static int blk_zone_need_reset_cb(struct blk_zone *zone, unsigned int idx, - void *data) -{ - /* - * For an all-zones reset, ignore conventional, empty, read-only - * and offline zones. - */ - switch (zone->cond) { - case BLK_ZONE_COND_NOT_WP: - case BLK_ZONE_COND_EMPTY: - case BLK_ZONE_COND_READONLY: - case BLK_ZONE_COND_OFFLINE: - return 0; - default: - set_bit(idx, (unsigned long *)data); - return 0; - } -} - -static int blkdev_zone_reset_all_emulated(struct block_device *bdev) -{ - struct gendisk *disk = bdev->bd_disk; - sector_t capacity = bdev_nr_sectors(bdev); - sector_t zone_sectors = bdev_zone_sectors(bdev); - unsigned long *need_reset; - struct bio *bio = NULL; - sector_t sector = 0; - int ret; - - need_reset = blk_alloc_zone_bitmap(disk->queue->node, disk->nr_zones); - if (!need_reset) - return -ENOMEM; - - ret = disk->fops->report_zones(disk, 0, disk->nr_zones, - blk_zone_need_reset_cb, need_reset); - if (ret < 0) - goto out_free_need_reset; - - ret = 0; - while (sector < capacity) { - if (!test_bit(disk_zone_no(disk, sector), need_reset)) { - sector += zone_sectors; - continue; - } - - bio = blk_next_bio(bio, bdev, 0, REQ_OP_ZONE_RESET | REQ_SYNC, - GFP_KERNEL); - bio->bi_iter.bi_sector = sector; - sector += zone_sectors; - - /* This may take a while, so be nice to others */ - cond_resched(); - } - - if (bio) { - ret = submit_bio_wait(bio); - bio_put(bio); - } - -out_free_need_reset: - kfree(need_reset); - return ret; -} - static int blkdev_zone_reset_all(struct block_device *bdev) { struct bio bio; @@ -247,7 +183,6 @@ static int blkdev_zone_reset_all(struct block_device *bdev) int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op, sector_t sector, sector_t nr_sectors) { - struct request_queue *q = bdev_get_queue(bdev); sector_t zone_sectors = bdev_zone_sectors(bdev); sector_t capacity = bdev_nr_sectors(bdev); sector_t end_sector = sector + nr_sectors; @@ -275,16 +210,11 @@ int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op, return -EINVAL; /* - * In the case of a zone reset operation over all zones, - * REQ_OP_ZONE_RESET_ALL can be used with devices supporting this - * command. For other devices, we emulate this command behavior by - * identifying the zones needing a reset. + * In the case of a zone reset operation over all zones, use + * REQ_OP_ZONE_RESET_ALL. */ - if (op == REQ_OP_ZONE_RESET && sector == 0 && nr_sectors == capacity) { - if (!blk_queue_zone_resetall(q)) - return blkdev_zone_reset_all_emulated(bdev); + if (op == REQ_OP_ZONE_RESET && sector == 0 && nr_sectors == capacity) return blkdev_zone_reset_all(bdev); - } while (sector < end_sector) { bio = blk_next_bio(bio, bdev, 0, op | REQ_SYNC, GFP_KERNEL); diff --git a/drivers/block/null_blk/zoned.c b/drivers/block/null_blk/zoned.c index 7996e2e7dce28a..9bc768b2ca56b7 100644 --- a/drivers/block/null_blk/zoned.c +++ b/drivers/block/null_blk/zoned.c @@ -164,7 +164,7 @@ int null_init_zoned_dev(struct nullb_device *dev, sector += dev->zone_size_sects; } - lim->features |= BLK_FEAT_ZONED | BLK_FEAT_ZONE_RESETALL; + lim->features |= BLK_FEAT_ZONED; lim->chunk_sectors = dev->zone_size_sects; lim->max_zone_append_sectors = dev->zone_append_max_sectors; lim->max_open_zones = dev->zone_max_open; diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index 4fdff13fc23b8a..d10a2ea072921d 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -2194,7 +2194,7 @@ static int ublk_ctrl_start_dev(struct ublk_device *ub, struct io_uring_cmd *cmd) if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED)) return -EOPNOTSUPP; - lim.features |= BLK_FEAT_ZONED | BLK_FEAT_ZONE_RESETALL; + lim.features |= BLK_FEAT_ZONED; lim.max_active_zones = p->max_active_zones; lim.max_open_zones = p->max_open_zones; lim.max_zone_append_sectors = p->max_zone_append_sectors; diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index 6c64a67ab9c901..84c3efd0c6117f 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -728,7 +728,7 @@ static int virtblk_read_zoned_limits(struct virtio_blk *vblk, dev_dbg(&vdev->dev, "probing host-managed zoned device\n"); - lim->features |= BLK_FEAT_ZONED | BLK_FEAT_ZONE_RESETALL; + lim->features |= BLK_FEAT_ZONED; virtio_cread(vdev, struct virtio_blk_config, zoned.max_open_zones, &v); diff --git a/drivers/nvme/host/zns.c b/drivers/nvme/host/zns.c index 99bb89c2495ae3..9a06f9d98cd68c 100644 --- a/drivers/nvme/host/zns.c +++ b/drivers/nvme/host/zns.c @@ -108,7 +108,7 @@ int nvme_query_zone_info(struct nvme_ns *ns, unsigned lbaf, void nvme_update_zone_info(struct nvme_ns *ns, struct queue_limits *lim, struct nvme_zone_info *zi) { - lim->features |= BLK_FEAT_ZONED | BLK_FEAT_ZONE_RESETALL; + lim->features |= BLK_FEAT_ZONED; lim->max_open_zones = zi->max_open_zones; lim->max_active_zones = zi->max_active_zones; lim->max_zone_append_sectors = ns->ctrl->max_zone_append; diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c index f7067afac79c14..c8b9654d30f0c3 100644 --- a/drivers/scsi/sd_zbc.c +++ b/drivers/scsi/sd_zbc.c @@ -599,7 +599,7 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, struct queue_limits *lim, if (sdkp->device->type != TYPE_ZBC) return 0; - lim->features |= BLK_FEAT_ZONED | BLK_FEAT_ZONE_RESETALL; + lim->features |= BLK_FEAT_ZONED; /* * Per ZBC and ZAC specifications, writes in sequential write required diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 4d0d4b83bc740f..dc250d8070d215 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -318,9 +318,6 @@ typedef unsigned int __bitwise blk_features_t; /* is a zoned device */ #define BLK_FEAT_ZONED ((__force blk_features_t)(1u << 10)) -/* supports Zone Reset All */ -#define BLK_FEAT_ZONE_RESETALL ((__force blk_features_t)(1u << 11)) - /* supports PCI(e) p2p requests */ #define BLK_FEAT_PCI_P2PDMA ((__force blk_features_t)(1u << 12)) @@ -618,8 +615,6 @@ void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) #define blk_queue_nonrot(q) (!((q)->limits.features & BLK_FEAT_ROTATIONAL)) #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) -#define blk_queue_zone_resetall(q) \ - ((q)->limits.features & BLK_FEAT_ZONE_RESETALL) #define blk_queue_dax(q) ((q)->limits.features & BLK_FEAT_DAX) #define blk_queue_pci_p2pdma(q) ((q)->limits.features & BLK_FEAT_PCI_P2PDMA) #ifdef CONFIG_BLK_RQ_ALLOC_TIME From 2f20872ed43185780a5f30581472599342c86d4a Mon Sep 17 00:00:00 2001 From: Damien Le Moal Date: Thu, 4 Jul 2024 14:28:16 +0900 Subject: [PATCH 201/209] block: Remove blk_alloc_zone_bitmap() Remove the helper function blk_alloc_zone_bitmap() and replace its single call site with a call to bitmap_alloc(). To be consistent with this change, use bitmap_free() to free a disk convnetional zone bitmap. Signed-off-by: Damien Le Moal Reviewed-by: Christoph Hellwig Reviewed-by: Johannes Thumshirn Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240704052816.623865-6-dlemoal@kernel.org Signed-off-by: Jens Axboe --- block/blk-zoned.c | 12 ++---------- 1 file changed, 2 insertions(+), 10 deletions(-) diff --git a/block/blk-zoned.c b/block/blk-zoned.c index 0007ef9cd5cacd..7ace890136a333 100644 --- a/block/blk-zoned.c +++ b/block/blk-zoned.c @@ -150,13 +150,6 @@ int blkdev_report_zones(struct block_device *bdev, sector_t sector, } EXPORT_SYMBOL_GPL(blkdev_report_zones); -static inline unsigned long *blk_alloc_zone_bitmap(int node, - unsigned int nr_zones) -{ - return kcalloc_node(BITS_TO_LONGS(nr_zones), sizeof(unsigned long), - GFP_NOIO, node); -} - static int blkdev_zone_reset_all(struct block_device *bdev) { struct bio bio; @@ -1482,7 +1475,7 @@ void disk_free_zone_resources(struct gendisk *disk) mempool_destroy(disk->zone_wplugs_pool); disk->zone_wplugs_pool = NULL; - kfree(disk->conv_zones_bitmap); + bitmap_free(disk->conv_zones_bitmap); disk->conv_zones_bitmap = NULL; disk->zone_capacity = 0; disk->last_zone_capacity = 0; @@ -1604,7 +1597,6 @@ static int blk_revalidate_conv_zone(struct blk_zone *zone, unsigned int idx, struct blk_revalidate_zone_args *args) { struct gendisk *disk = args->disk; - struct request_queue *q = disk->queue; if (zone->capacity != zone->len) { pr_warn("%s: Invalid conventional zone capacity\n", @@ -1620,7 +1612,7 @@ static int blk_revalidate_conv_zone(struct blk_zone *zone, unsigned int idx, if (!args->conv_zones_bitmap) { args->conv_zones_bitmap = - blk_alloc_zone_bitmap(q->node, args->nr_zones); + bitmap_zalloc(args->nr_zones, GFP_NOIO); if (!args->conv_zones_bitmap) return -ENOMEM; } From 73a768d5f95533574bb8ace34eb683a88c40509e Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 1 Jul 2024 18:51:15 +0200 Subject: [PATCH 202/209] block: factor out a blk_write_zeroes_limit helper Contrary to the comment in __blkdev_issue_write_zeroes, nothing here checks for a potential bi_size overflow. Add a helper mirroring the secure erase code for the check. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240701165219.1571322-6-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-lib.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) diff --git a/block/blk-lib.c b/block/blk-lib.c index 442da9dad04213..297bcf6896930f 100644 --- a/block/blk-lib.c +++ b/block/blk-lib.c @@ -103,24 +103,28 @@ int blkdev_issue_discard(struct block_device *bdev, sector_t sector, } EXPORT_SYMBOL(blkdev_issue_discard); +static sector_t bio_write_zeroes_limit(struct block_device *bdev) +{ + sector_t bs_mask = (bdev_logical_block_size(bdev) >> 9) - 1; + + return min(bdev_write_zeroes_sectors(bdev), + (UINT_MAX >> SECTOR_SHIFT) & ~bs_mask); +} + static int __blkdev_issue_write_zeroes(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, unsigned flags) { struct bio *bio = *biop; - unsigned int max_sectors; if (bdev_read_only(bdev)) return -EPERM; - - /* Ensure that max_sectors doesn't overflow bi_size */ - max_sectors = bdev_write_zeroes_sectors(bdev); - - if (max_sectors == 0) + if (!bdev_write_zeroes_sectors(bdev)) return -EOPNOTSUPP; while (nr_sects) { - unsigned int len = min_t(sector_t, nr_sects, max_sectors); + unsigned int len = min_t(sector_t, nr_sects, + bio_write_zeroes_limit(bdev)); bio = blk_next_bio(bio, bdev, 0, REQ_OP_WRITE_ZEROES, gfp_mask); bio->bi_iter.bi_sector = sector; From ff760a8f0d09f4ba7574ae2ca8be987854f5246d Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 1 Jul 2024 18:51:16 +0200 Subject: [PATCH 203/209] block: remove the LBA alignment check in __blkdev_issue_zeroout __blkdev_issue_zeroout is a purely kernel internal API and thus can rely on the block layer sector alignment checks. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240701165219.1571322-7-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-lib.c | 5 ----- 1 file changed, 5 deletions(-) diff --git a/block/blk-lib.c b/block/blk-lib.c index 297bcf6896930f..601a122af8587c 100644 --- a/block/blk-lib.c +++ b/block/blk-lib.c @@ -209,11 +209,6 @@ int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, unsigned flags) { int ret; - sector_t bs_mask; - - bs_mask = (bdev_logical_block_size(bdev) >> 9) - 1; - if ((sector | nr_sects) & bs_mask) - return -EINVAL; ret = __blkdev_issue_write_zeroes(bdev, sector, nr_sects, gfp_mask, biop, flags); From f6eacb26541ad1eabc40d7e9f5cd86bae7dc0b46 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 1 Jul 2024 18:51:17 +0200 Subject: [PATCH 204/209] block: move read-only and supported checks into (__)blkdev_issue_zeroout Move these checks out of the lower level helpers and into the higher level ones to prepare for refactoring. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240701165219.1571322-8-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-lib.c | 51 ++++++++++++++++++++++--------------------------- 1 file changed, 23 insertions(+), 28 deletions(-) diff --git a/block/blk-lib.c b/block/blk-lib.c index 601a122af8587c..02f69c683a9839 100644 --- a/block/blk-lib.c +++ b/block/blk-lib.c @@ -111,17 +111,12 @@ static sector_t bio_write_zeroes_limit(struct block_device *bdev) (UINT_MAX >> SECTOR_SHIFT) & ~bs_mask); } -static int __blkdev_issue_write_zeroes(struct block_device *bdev, +static void __blkdev_issue_write_zeroes(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, unsigned flags) { struct bio *bio = *biop; - if (bdev_read_only(bdev)) - return -EPERM; - if (!bdev_write_zeroes_sectors(bdev)) - return -EOPNOTSUPP; - while (nr_sects) { unsigned int len = min_t(sector_t, nr_sects, bio_write_zeroes_limit(bdev)); @@ -138,7 +133,6 @@ static int __blkdev_issue_write_zeroes(struct block_device *bdev, } *biop = bio; - return 0; } /* @@ -154,7 +148,7 @@ static unsigned int __blkdev_sectors_to_bio_pages(sector_t nr_sects) return min(pages, (sector_t)BIO_MAX_VECS); } -static int __blkdev_issue_zero_pages(struct block_device *bdev, +static void __blkdev_issue_zero_pages(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, struct bio **biop) { @@ -162,9 +156,6 @@ static int __blkdev_issue_zero_pages(struct block_device *bdev, int bi_size = 0; unsigned int sz; - if (bdev_read_only(bdev)) - return -EPERM; - while (nr_sects != 0) { bio = blk_next_bio(bio, bdev, __blkdev_sectors_to_bio_pages(nr_sects), REQ_OP_WRITE, gfp_mask); @@ -182,7 +173,6 @@ static int __blkdev_issue_zero_pages(struct block_device *bdev, } *biop = bio; - return 0; } /** @@ -208,15 +198,19 @@ int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, unsigned flags) { - int ret; - - ret = __blkdev_issue_write_zeroes(bdev, sector, nr_sects, gfp_mask, - biop, flags); - if (ret != -EOPNOTSUPP || (flags & BLKDEV_ZERO_NOFALLBACK)) - return ret; + if (bdev_read_only(bdev)) + return -EPERM; - return __blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp_mask, - biop); + if (bdev_write_zeroes_sectors(bdev)) { + __blkdev_issue_write_zeroes(bdev, sector, nr_sects, + gfp_mask, biop, flags); + } else { + if (flags & BLKDEV_ZERO_NOFALLBACK) + return -EOPNOTSUPP; + __blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp_mask, + biop); + } + return 0; } EXPORT_SYMBOL(__blkdev_issue_zeroout); @@ -245,21 +239,22 @@ int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, bs_mask = (bdev_logical_block_size(bdev) >> 9) - 1; if ((sector | nr_sects) & bs_mask) return -EINVAL; + if (bdev_read_only(bdev)) + return -EPERM; + if ((flags & BLKDEV_ZERO_NOFALLBACK) && !try_write_zeroes) + return -EOPNOTSUPP; retry: bio = NULL; blk_start_plug(&plug); if (try_write_zeroes) { - ret = __blkdev_issue_write_zeroes(bdev, sector, nr_sects, - gfp_mask, &bio, flags); - } else if (!(flags & BLKDEV_ZERO_NOFALLBACK)) { - ret = __blkdev_issue_zero_pages(bdev, sector, nr_sects, - gfp_mask, &bio); + __blkdev_issue_write_zeroes(bdev, sector, nr_sects, gfp_mask, + &bio, flags); } else { - /* No zeroing offload support */ - ret = -EOPNOTSUPP; + __blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp_mask, + &bio); } - if (ret == 0 && bio) { + if (bio) { ret = submit_bio_wait(bio); bio_put(bio); } From 99800ced26b9d87a918aa9824881bdb90a3c1b03 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 1 Jul 2024 18:51:18 +0200 Subject: [PATCH 205/209] block: refacto blkdev_issue_zeroout Split out two well-defined helpers for hardware supported Write Zeroes and manually writing zeroes using the Write command. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240701165219.1571322-9-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-lib.c | 94 +++++++++++++++++++++++++++++-------------------- 1 file changed, 55 insertions(+), 39 deletions(-) diff --git a/block/blk-lib.c b/block/blk-lib.c index 02f69c683a9839..9dbab472c63f12 100644 --- a/block/blk-lib.c +++ b/block/blk-lib.c @@ -135,6 +135,32 @@ static void __blkdev_issue_write_zeroes(struct block_device *bdev, *biop = bio; } +static int blkdev_issue_write_zeroes(struct block_device *bdev, sector_t sector, + sector_t nr_sects, gfp_t gfp, unsigned flags) +{ + struct bio *bio = NULL; + struct blk_plug plug; + int ret = 0; + + blk_start_plug(&plug); + __blkdev_issue_write_zeroes(bdev, sector, nr_sects, gfp, &bio, flags); + if (bio) { + ret = submit_bio_wait(bio); + bio_put(bio); + } + blk_finish_plug(&plug); + + /* + * For some devices there is no non-destructive way to verify whether + * WRITE ZEROES is actually supported. These will clear the capability + * on an I/O error, in which case we'll turn any error into + * "not supported" here. + */ + if (ret && !bdev_write_zeroes_sectors(bdev)) + return -EOPNOTSUPP; + return ret; +} + /* * Convert a number of 512B sectors to a number of pages. * The result is limited to a number of pages that can fit into a BIO. @@ -175,6 +201,27 @@ static void __blkdev_issue_zero_pages(struct block_device *bdev, *biop = bio; } +static int blkdev_issue_zero_pages(struct block_device *bdev, sector_t sector, + sector_t nr_sects, gfp_t gfp, unsigned flags) +{ + struct bio *bio = NULL; + struct blk_plug plug; + int ret = 0; + + if (flags & BLKDEV_ZERO_NOFALLBACK) + return -EOPNOTSUPP; + + blk_start_plug(&plug); + __blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp, &bio); + if (bio) { + ret = submit_bio_wait(bio); + bio_put(bio); + } + blk_finish_plug(&plug); + + return ret; +} + /** * __blkdev_issue_zeroout - generate number of zero filed write bios * @bdev: blockdev to issue @@ -230,52 +277,21 @@ EXPORT_SYMBOL(__blkdev_issue_zeroout); int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, unsigned flags) { - int ret = 0; - sector_t bs_mask; - struct bio *bio; - struct blk_plug plug; - bool try_write_zeroes = !!bdev_write_zeroes_sectors(bdev); + int ret; - bs_mask = (bdev_logical_block_size(bdev) >> 9) - 1; - if ((sector | nr_sects) & bs_mask) + if ((sector | nr_sects) & ((bdev_logical_block_size(bdev) >> 9) - 1)) return -EINVAL; if (bdev_read_only(bdev)) return -EPERM; - if ((flags & BLKDEV_ZERO_NOFALLBACK) && !try_write_zeroes) - return -EOPNOTSUPP; -retry: - bio = NULL; - blk_start_plug(&plug); - if (try_write_zeroes) { - __blkdev_issue_write_zeroes(bdev, sector, nr_sects, gfp_mask, - &bio, flags); - } else { - __blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp_mask, - &bio); - } - if (bio) { - ret = submit_bio_wait(bio); - bio_put(bio); - } - blk_finish_plug(&plug); - if (ret && try_write_zeroes) { - if (!(flags & BLKDEV_ZERO_NOFALLBACK)) { - try_write_zeroes = false; - goto retry; - } - if (!bdev_write_zeroes_sectors(bdev)) { - /* - * Zeroing offload support was indicated, but the - * device reported ILLEGAL REQUEST (for some devices - * there is no non-destructive way to verify whether - * WRITE ZEROES is actually supported). - */ - ret = -EOPNOTSUPP; - } + if (bdev_write_zeroes_sectors(bdev)) { + ret = blkdev_issue_write_zeroes(bdev, sector, nr_sects, + gfp_mask, flags); + if (!ret) + return ret; } - return ret; + return blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp_mask, flags); } EXPORT_SYMBOL(blkdev_issue_zeroout); From 39722a2f2bcd82bdecc226711412d88b54fcb05b Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 1 Jul 2024 18:51:19 +0200 Subject: [PATCH 206/209] block: limit the Write Zeroes to manually writing zeroes fallback Only fall back from hardware Write Zeroes failures when blkdev_issue_write_zeroes returns -EOPNOTSUPP; Note that blkdev_issue_write_zeroes turns any failure into -EOPNOTSUPP when the write zeroes queue limit has been cleared to 0, so this still catches all I/O errors where the driver detected missing support for the hardware acceleration. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240701165219.1571322-10-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-lib.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/block/blk-lib.c b/block/blk-lib.c index 9dbab472c63f12..2384acd913d21f 100644 --- a/block/blk-lib.c +++ b/block/blk-lib.c @@ -287,7 +287,7 @@ int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, if (bdev_write_zeroes_sectors(bdev)) { ret = blkdev_issue_write_zeroes(bdev, sector, nr_sects, gfp_mask, flags); - if (!ret) + if (ret != -EOPNOTSUPP) return ret; } From bf86bcdb40123ee99669ee91b67e023669433a1a Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 1 Jul 2024 18:51:20 +0200 Subject: [PATCH 207/209] blk-lib: check for kill signal in ioctl BLKZEROOUT Zeroout can access a significant capacity and take longer than the user expected. A user may change their mind about wanting to run that command and attempt to kill the process and do something else with their device. But since the task is uninterruptable, they have to wait for it to finish, which could be many hours. Add a new BLKDEV_ZERO_KILLABLE flag for blkdev_issue_zeroout that checks for a fatal signal at each iteration so the user doesn't have to wait for their regretted operation to complete naturally. Heavily based on an earlier patch from Keith Busch. Signed-off-by: Christoph Hellwig Reviewed-by: Martin K. Petersen Link: https://lore.kernel.org/r/20240701165219.1571322-11-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-lib.c | 66 +++++++++++++++++++++++++++--------------- block/ioctl.c | 2 +- include/linux/blkdev.h | 1 + 3 files changed, 45 insertions(+), 24 deletions(-) diff --git a/block/blk-lib.c b/block/blk-lib.c index 2384acd913d21f..9f735efa6c9459 100644 --- a/block/blk-lib.c +++ b/block/blk-lib.c @@ -115,24 +115,27 @@ static void __blkdev_issue_write_zeroes(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, unsigned flags) { - struct bio *bio = *biop; - while (nr_sects) { unsigned int len = min_t(sector_t, nr_sects, bio_write_zeroes_limit(bdev)); + struct bio *bio; + + if ((flags & BLKDEV_ZERO_KILLABLE) && + fatal_signal_pending(current)) + break; - bio = blk_next_bio(bio, bdev, 0, REQ_OP_WRITE_ZEROES, gfp_mask); + bio = bio_alloc(bdev, 0, REQ_OP_WRITE_ZEROES, gfp_mask); bio->bi_iter.bi_sector = sector; if (flags & BLKDEV_ZERO_NOUNMAP) bio->bi_opf |= REQ_NOUNMAP; bio->bi_iter.bi_size = len << SECTOR_SHIFT; + *biop = bio_chain_and_submit(*biop, bio); + nr_sects -= len; sector += len; cond_resched(); } - - *biop = bio; } static int blkdev_issue_write_zeroes(struct block_device *bdev, sector_t sector, @@ -145,6 +148,12 @@ static int blkdev_issue_write_zeroes(struct block_device *bdev, sector_t sector, blk_start_plug(&plug); __blkdev_issue_write_zeroes(bdev, sector, nr_sects, gfp, &bio, flags); if (bio) { + if ((flags & BLKDEV_ZERO_KILLABLE) && + fatal_signal_pending(current)) { + bio_await_chain(bio); + blk_finish_plug(&plug); + return -EINTR; + } ret = submit_bio_wait(bio); bio_put(bio); } @@ -176,29 +185,34 @@ static unsigned int __blkdev_sectors_to_bio_pages(sector_t nr_sects) static void __blkdev_issue_zero_pages(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, - struct bio **biop) + struct bio **biop, unsigned int flags) { - struct bio *bio = *biop; - int bi_size = 0; - unsigned int sz; + while (nr_sects) { + unsigned int nr_vecs = __blkdev_sectors_to_bio_pages(nr_sects); + struct bio *bio; - while (nr_sects != 0) { - bio = blk_next_bio(bio, bdev, __blkdev_sectors_to_bio_pages(nr_sects), - REQ_OP_WRITE, gfp_mask); + bio = bio_alloc(bdev, nr_vecs, REQ_OP_WRITE, gfp_mask); bio->bi_iter.bi_sector = sector; - while (nr_sects != 0) { - sz = min((sector_t) PAGE_SIZE, nr_sects << 9); - bi_size = bio_add_page(bio, ZERO_PAGE(0), sz, 0); - nr_sects -= bi_size >> 9; - sector += bi_size >> 9; - if (bi_size < sz) + if ((flags & BLKDEV_ZERO_KILLABLE) && + fatal_signal_pending(current)) + break; + + do { + unsigned int len, added; + + len = min_t(sector_t, + PAGE_SIZE, nr_sects << SECTOR_SHIFT); + added = bio_add_page(bio, ZERO_PAGE(0), len, 0); + if (added < len) break; - } + nr_sects -= added >> SECTOR_SHIFT; + sector += added >> SECTOR_SHIFT; + } while (nr_sects); + + *biop = bio_chain_and_submit(*biop, bio); cond_resched(); } - - *biop = bio; } static int blkdev_issue_zero_pages(struct block_device *bdev, sector_t sector, @@ -212,8 +226,14 @@ static int blkdev_issue_zero_pages(struct block_device *bdev, sector_t sector, return -EOPNOTSUPP; blk_start_plug(&plug); - __blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp, &bio); + __blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp, &bio, flags); if (bio) { + if ((flags & BLKDEV_ZERO_KILLABLE) && + fatal_signal_pending(current)) { + bio_await_chain(bio); + blk_finish_plug(&plug); + return -EINTR; + } ret = submit_bio_wait(bio); bio_put(bio); } @@ -255,7 +275,7 @@ int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, if (flags & BLKDEV_ZERO_NOFALLBACK) return -EOPNOTSUPP; __blkdev_issue_zero_pages(bdev, sector, nr_sects, gfp_mask, - biop); + biop, flags); } return 0; } diff --git a/block/ioctl.c b/block/ioctl.c index d570e16958961e..e8e4a4190f183a 100644 --- a/block/ioctl.c +++ b/block/ioctl.c @@ -224,7 +224,7 @@ static int blk_ioctl_zeroout(struct block_device *bdev, blk_mode_t mode, goto fail; err = blkdev_issue_zeroout(bdev, start >> 9, len >> 9, GFP_KERNEL, - BLKDEV_ZERO_NOUNMAP); + BLKDEV_ZERO_NOUNMAP | BLKDEV_ZERO_KILLABLE); fail: filemap_invalidate_unlock(bdev->bd_mapping); diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index dc250d8070d215..02e04df272828e 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -1095,6 +1095,7 @@ int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector, #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ +#define BLKDEV_ZERO_KILLABLE (1 << 2) /* interruptible by fatal signals */ extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, From 25f76c3db2f08428b5acd082a52787164001eb6e Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Sat, 6 Jul 2024 09:52:17 +0200 Subject: [PATCH 208/209] block: add a bvec_phys helper Get callers out of poking into bvec internals a bit more. Not a huge win right now, but with the proposed new DMA mapping API we might end up with a lot more of this otherwise. Signed-off-by: Christoph Hellwig Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240706075228.2350978-2-hch@lst.de Signed-off-by: Jens Axboe --- arch/m68k/emu/nfblock.c | 2 +- block/bio.c | 2 +- block/blk.h | 4 ++-- include/linux/bvec.h | 14 ++++++++++++++ 4 files changed, 18 insertions(+), 4 deletions(-) diff --git a/arch/m68k/emu/nfblock.c b/arch/m68k/emu/nfblock.c index 8eea7ef9115146..874fe958877388 100644 --- a/arch/m68k/emu/nfblock.c +++ b/arch/m68k/emu/nfblock.c @@ -71,7 +71,7 @@ static void nfhd_submit_bio(struct bio *bio) len = bvec.bv_len; len >>= 9; nfhd_read_write(dev->id, 0, dir, sec >> shift, len >> shift, - page_to_phys(bvec.bv_page) + bvec.bv_offset); + bvec_phys(&bvec)); sec += len; } bio_endio(bio); diff --git a/block/bio.c b/block/bio.c index e9e809a63c5975..a3b1b2266c50be 100644 --- a/block/bio.c +++ b/block/bio.c @@ -953,7 +953,7 @@ bool bvec_try_merge_hw_page(struct request_queue *q, struct bio_vec *bv, bool *same_page) { unsigned long mask = queue_segment_boundary(q); - phys_addr_t addr1 = page_to_phys(bv->bv_page) + bv->bv_offset; + phys_addr_t addr1 = bvec_phys(bv); phys_addr_t addr2 = page_to_phys(page) + offset + len - 1; if ((addr1 | mask) != (addr2 | mask)) diff --git a/block/blk.h b/block/blk.h index 47dadd2439b1ca..8e8936e97307c6 100644 --- a/block/blk.h +++ b/block/blk.h @@ -98,8 +98,8 @@ static inline bool biovec_phys_mergeable(struct request_queue *q, struct bio_vec *vec1, struct bio_vec *vec2) { unsigned long mask = queue_segment_boundary(q); - phys_addr_t addr1 = page_to_phys(vec1->bv_page) + vec1->bv_offset; - phys_addr_t addr2 = page_to_phys(vec2->bv_page) + vec2->bv_offset; + phys_addr_t addr1 = bvec_phys(vec1); + phys_addr_t addr2 = bvec_phys(vec2); /* * Merging adjacent physical pages may not work correctly under KMSAN diff --git a/include/linux/bvec.h b/include/linux/bvec.h index bd1e361b351c5a..f41c7f0ef91ed5 100644 --- a/include/linux/bvec.h +++ b/include/linux/bvec.h @@ -280,4 +280,18 @@ static inline void *bvec_virt(struct bio_vec *bvec) return page_address(bvec->bv_page) + bvec->bv_offset; } +/** + * bvec_phys - return the physical address for a bvec + * @bvec: bvec to return the physical address for + */ +static inline phys_addr_t bvec_phys(const struct bio_vec *bvec) +{ + /* + * Note this open codes page_to_phys because page_to_phys is defined in + * , which we don't want to pull in here. If it ever moves to + * a sensible place we should start using it. + */ + return PFN_PHYS(page_to_pfn(bvec->bv_page)) + bvec->bv_offset; +} + #endif /* __LINUX_BVEC_H */ From 09595e0c9d654743483197b2f21dd4ec37c90a27 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Sat, 6 Jul 2024 09:52:18 +0200 Subject: [PATCH 209/209] block: pass a phys_addr_t to get_max_segment_size Work on a single address to simplify the logic, and prepare the callers from using better helpers. Signed-off-by: Christoph Hellwig Reviewed-by: Chaitanya Kulkarni Link: https://lore.kernel.org/r/20240706075228.2350978-3-hch@lst.de Signed-off-by: Jens Axboe --- block/blk-merge.c | 25 +++++++++++-------------- 1 file changed, 11 insertions(+), 14 deletions(-) diff --git a/block/blk-merge.c b/block/blk-merge.c index cff20bcc0252a7..e41ea331809936 100644 --- a/block/blk-merge.c +++ b/block/blk-merge.c @@ -209,23 +209,22 @@ static inline unsigned get_max_io_size(struct bio *bio, /** * get_max_segment_size() - maximum number of bytes to add as a single segment * @lim: Request queue limits. - * @start_page: See below. - * @offset: Offset from @start_page where to add a segment. + * @paddr: address of the range to add + * @max_len: maximum length available to add at @paddr * - * Returns the maximum number of bytes that can be added as a single segment. + * Returns the maximum number of bytes of the range starting at @paddr that can + * be added to a single segment. */ static inline unsigned get_max_segment_size(const struct queue_limits *lim, - struct page *start_page, unsigned long offset) + phys_addr_t paddr, unsigned int len) { - unsigned long mask = lim->seg_boundary_mask; - - offset = mask & (page_to_phys(start_page) + offset); - /* * Prevent an overflow if mask = ULONG_MAX and offset = 0 by adding 1 * after having calculated the minimum. */ - return min(mask - offset, (unsigned long)lim->max_segment_size - 1) + 1; + return min_t(unsigned long, len, + min(lim->seg_boundary_mask - (lim->seg_boundary_mask & paddr), + (unsigned long)lim->max_segment_size - 1) + 1); } /** @@ -258,9 +257,7 @@ static bool bvec_split_segs(const struct queue_limits *lim, unsigned seg_size = 0; while (len && *nsegs < max_segs) { - seg_size = get_max_segment_size(lim, bv->bv_page, - bv->bv_offset + total_len); - seg_size = min(seg_size, len); + seg_size = get_max_segment_size(lim, bvec_phys(bv) + total_len, len); (*nsegs)++; total_len += seg_size; @@ -494,8 +491,8 @@ static unsigned blk_bvec_map_sg(struct request_queue *q, while (nbytes > 0) { unsigned offset = bvec->bv_offset + total; - unsigned len = min(get_max_segment_size(&q->limits, - bvec->bv_page, offset), nbytes); + unsigned len = get_max_segment_size(&q->limits, bvec_phys(bvec), + nbytes); struct page *page = bvec->bv_page; /*