This repository has been archived by the owner on Jun 18, 2024. It is now read-only.
[PATCH] SCX: scx_ops_bypass() can't grab a mutex #198
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This fixes a bug introduced by 34c6075 ("scx: Fix scx_ops_bypass_depth going out of sync"). It added mutex_lock(scx_ops_enable_mutex) to scx_ops_bypass(), which can obviously lead to problem if scx_ops_bypass() is invoked from the disable path while someone is holding the mutex and scheduled out. [ 246.906433] INFO: task sched_ext_ops_h:1531 blocked for more than 122 seconds. [ 246.907862] Not tainted 6.9.0-rc5-work-02067-g7f2a15fc2eb2-dirty #34 [ 246.909213] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 246.910973] task:sched_ext_ops_h state:D stack:0 pid:1531 tgid:1531 ppid:2 flags:0x00004000 [ 246.913004] Sched_ext: qmap (enabled+all) [ 246.913007] Call Trace: [ 246.914486] <TASK> [ 246.914956] __schedule+0x750/0xcd0 [ 246.915770] schedule+0x41/0xb0 [ 246.916503] schedule_preempt_disabled+0x15/0x20 [ 246.917542] __mutex_lock+0x511/0xb00 [ 246.918379] ? scx_ops_bypass+0x54/0x1a0 [ 246.919232] mutex_lock_nested+0x1b/0x20 [ 246.920110] scx_ops_bypass+0x54/0x1a0 [ 246.920961] ? irqentry_exit+0x59/0x80 [ 246.921814] ? sysvec_irq_work+0x41/0x80 [ 246.922705] scx_ops_disable_workfn+0xd1/0x920 [ 246.923714] ? lock_acquire+0xcb/0x230 [ 246.924567] ? kthread_worker_fn+0x80/0x2a0 [ 246.925504] ? lock_release+0xd8/0x2d0 [ 246.926355] ? kthread_worker_fn+0xbe/0x2a0 [ 246.927262] kthread_worker_fn+0x10a/0x2a0 [ 246.928182] ? scx_dump_task+0x2f0/0x2f0 [ 246.929080] kthread+0xec/0x110 [ 246.929814] ? __kthread_init_worker+0x100/0x100 [ 246.930855] ? kthread_blkcg+0x40/0x40 [ 246.931718] ret_from_fork+0x36/0x40 [ 246.932545] ? kthread_blkcg+0x40/0x40 [ 246.933411] ret_from_fork_asm+0x11/0x20 [ 246.934311] </TASK> [ 246.934798] [ 246.934798] Showing all locks held in the system: [ 246.936169] 1 lock held by khungtaskd/60: [ 246.937081] #0: ffffffff8335b128 (rcu_read_lock){....}-{1:2}, at: debug_show_all_locks+0x30/0x190 [ 246.939051] 1 lock held by scx_qmap/1529: [ 246.939963] 1 lock held by sched_ext_ops_h/1531: [ 246.940999] #0: ffffffff83272e28 (scx_ops_enable_mutex){+.+.}-{3:3}, at: scx_ops_bypass+0x54/0x1a0 There's no reason to grab the mutex there. We can just test scx_enabled() without holding the mutex. Drop the broken locking.
htejun
force-pushed
the
htejun/fix-stop-hang
branch
from
0dd603c
to
a95e0bf
Compare
Byte-Lab
approved these changes
May 5, 2024
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This fixes a bug introduced by 34c6075 ("scx: Fix scx_ops_bypass_depth going out of sync"). It added mutex_lock(scx_ops_enable_mutex) to scx_ops_bypass(), which can obviously lead to problem if scx_ops_bypass() is invoked from the disable path while someone is holding the mutex and scheduled out.
[ 246.906433] INFO: task sched_ext_ops_h:1531 blocked for more than 122 seconds.
[ 246.907862] Not tainted 6.9.0-rc5-work-02067-g7f2a15fc2eb2-dirty #34
[ 246.909213] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 246.910973] task:sched_ext_ops_h state:D stack:0 pid:1531 tgid:1531 ppid:2 flags:0x00004000
[ 246.913004] Sched_ext: qmap (enabled+all)
[ 246.913007] Call Trace:
[ 246.914486]
[ 246.914956] __schedule+0x750/0xcd0
[ 246.915770] schedule+0x41/0xb0
[ 246.916503] schedule_preempt_disabled+0x15/0x20
[ 246.917542] __mutex_lock+0x511/0xb00
[ 246.918379] ? scx_ops_bypass+0x54/0x1a0
[ 246.919232] mutex_lock_nested+0x1b/0x20
[ 246.920110] scx_ops_bypass+0x54/0x1a0
[ 246.920961] ? irqentry_exit+0x59/0x80
[ 246.921814] ? sysvec_irq_work+0x41/0x80
[ 246.922705] scx_ops_disable_workfn+0xd1/0x920
[ 246.923714] ? lock_acquire+0xcb/0x230
[ 246.924567] ? kthread_worker_fn+0x80/0x2a0
[ 246.925504] ? lock_release+0xd8/0x2d0
[ 246.926355] ? kthread_worker_fn+0xbe/0x2a0
[ 246.927262] kthread_worker_fn+0x10a/0x2a0
[ 246.928182] ? scx_dump_task+0x2f0/0x2f0
[ 246.929080] kthread+0xec/0x110
[ 246.929814] ? __kthread_init_worker+0x100/0x100
[ 246.930855] ? kthread_blkcg+0x40/0x40
[ 246.931718] ret_from_fork+0x36/0x40
[ 246.932545] ? kthread_blkcg+0x40/0x40
[ 246.933411] ret_from_fork_asm+0x11/0x20
[ 246.934311]
[ 246.934798]
[ 246.934798] Showing all locks held in the system:
[ 246.936169] 1 lock held by khungtaskd/60:
[ 246.937081] #0: ffffffff8335b128 (rcu_read_lock){....}-{1:2}, at: debug_show_all_locks+0x30/0x190
[ 246.939051] 1 lock held by scx_qmap/1529:
[ 246.939963] 1 lock held by sched_ext_ops_h/1531:
[ 246.940999] #0: ffffffff83272e28 (scx_ops_enable_mutex){+.+.}-{3:3}, at: scx_ops_bypass+0x54/0x1a0
There's no reason to grab the mutex there. We can just test scx_enabled() without holding the mutex. Drop the broken locking.