Update bcachefs sources to 5b8c4a1366 bcachefs: bcachefs_metadata_ver…

…sion_deleted_inodes

.bcachefs_revision

@@ -1 +1 @@
-33a60d9b05f523be93973b25e0df1ab2d65fa4fc
+5b8c4a1366df20bc043404cb882230ce86296590

libbcachefs/bcachefs.h

@@ -208,6 +208,7 @@
 #include "fifo.h"
 #include "nocow_locking_types.h"
 #include "opts.h"
+#include "recovery_types.h"
 #include "seqmutex.h"
 #include "util.h"
 
@@ -452,6 +453,7 @@ enum gc_phase {
 	GC_PHASE_BTREE_backpointers,
 	GC_PHASE_BTREE_bucket_gens,
 	GC_PHASE_BTREE_snapshot_trees,
+	GC_PHASE_BTREE_deleted_inodes,
 
 	GC_PHASE_PENDING_DELETE,
 };
@@ -655,48 +657,6 @@ enum bch_write_ref {
 	BCH_WRITE_REF_NR,
 };
 
-#define PASS_SILENT		BIT(0)
-#define PASS_FSCK		BIT(1)
-#define PASS_UNCLEAN		BIT(2)
-#define PASS_ALWAYS		BIT(3)
-
-#define BCH_RECOVERY_PASSES()									\
-	x(alloc_read,			PASS_ALWAYS)						\
-	x(stripes_read,			PASS_ALWAYS)						\
-	x(initialize_subvolumes,	0)							\
-	x(snapshots_read,		PASS_ALWAYS)						\
-	x(check_topology,		0)							\
-	x(check_allocations,		PASS_FSCK)						\
-	x(set_may_go_rw,		PASS_ALWAYS|PASS_SILENT)				\
-	x(journal_replay,		PASS_ALWAYS)						\
-	x(check_alloc_info,		PASS_FSCK)						\
-	x(check_lrus,			PASS_FSCK)						\
-	x(check_btree_backpointers,	PASS_FSCK)						\
-	x(check_backpointers_to_extents,PASS_FSCK)						\
-	x(check_extents_to_backpointers,PASS_FSCK)						\
-	x(check_alloc_to_lru_refs,	PASS_FSCK)						\
-	x(fs_freespace_init,		PASS_ALWAYS|PASS_SILENT)				\
-	x(bucket_gens_init,		0)							\
-	x(check_snapshot_trees,		PASS_FSCK)						\
-	x(check_snapshots,		PASS_FSCK)						\
-	x(check_subvols,		PASS_FSCK)						\
-	x(delete_dead_snapshots,	PASS_FSCK|PASS_UNCLEAN)					\
-	x(fs_upgrade_for_subvolumes,	0)							\
-	x(check_inodes,			PASS_FSCK|PASS_UNCLEAN)					\
-	x(check_extents,		PASS_FSCK)						\
-	x(check_dirents,		PASS_FSCK)						\
-	x(check_xattrs,			PASS_FSCK)						\
-	x(check_root,			PASS_FSCK)						\
-	x(check_directory_structure,	PASS_FSCK)						\
-	x(check_nlinks,			PASS_FSCK)						\
-	x(fix_reflink_p,		0)							\
-
-enum bch_recovery_pass {
-#define x(n, when)	BCH_RECOVERY_PASS_##n,
-	BCH_RECOVERY_PASSES()
-#undef x
-};
-
 struct bch_fs {
 	struct closure		cl;
 

libbcachefs/bcachefs_format.h

@@ -1629,7 +1629,9 @@ struct bch_sb_field_journal_seq_blacklist {
 	x(major_minor,			BCH_VERSION(1,  0),		\
 	  0)								\
 	x(snapshot_skiplists,		BCH_VERSION(1,  1),		\
-	  BIT_ULL(BCH_RECOVERY_PASS_check_snapshots))
+	  BIT_ULL(BCH_RECOVERY_PASS_check_snapshots))			\
+	x(deleted_inodes,		BCH_VERSION(1,  2),		\
+	  BIT_ULL(BCH_RECOVERY_PASS_check_inodes))
 
 enum bcachefs_metadata_version {
 	bcachefs_metadata_version_min = 9,
@@ -2251,7 +2253,9 @@ enum btree_id_flags {
 	x(bucket_gens,		14,	0,					\
 	  BIT_ULL(KEY_TYPE_bucket_gens))					\
 	x(snapshot_trees,	15,	0,					\
-	  BIT_ULL(KEY_TYPE_snapshot_tree))
+	  BIT_ULL(KEY_TYPE_snapshot_tree))					\
+	x(deleted_inodes,	16,	BTREE_ID_SNAPSHOTS,			\
+	  BIT_ULL(KEY_TYPE_set))
 
 enum btree_id {
 #define x(name, nr, ...) BTREE_ID_##name = nr,

libbcachefs/fs-io.c

@@ -35,7 +35,7 @@
 
 #include <trace/events/writeback.h>
 
-static void bch2_clamp_data_hole(struct inode *, u64 *, u64 *, unsigned);
+static int bch2_clamp_data_hole(struct inode *, u64 *, u64 *, unsigned, bool);
 
 struct folio_vec {
 	struct folio	*fv_folio;
@@ -3410,11 +3410,15 @@ static int __bchfs_fallocate(struct bch_inode_info *inode, int mode,
 		}
 
 		if (!(mode & FALLOC_FL_ZERO_RANGE)) {
-			ret = drop_locks_do(&trans,
-				(bch2_clamp_data_hole(&inode->v,
-						      &hole_start,
-						      &hole_end,
-						      opts.data_replicas), 0));
+			if (bch2_clamp_data_hole(&inode->v,
+						 &hole_start,
+						 &hole_end,
+						 opts.data_replicas, true))
+				ret = drop_locks_do(&trans,
+					(bch2_clamp_data_hole(&inode->v,
+							      &hole_start,
+							      &hole_end,
+							      opts.data_replicas, false), 0));
 			bch2_btree_iter_set_pos(&iter, POS(iter.pos.inode, hole_start));
 
 			if (ret)
@@ -3714,7 +3718,8 @@ static int folio_data_offset(struct folio *folio, loff_t pos,
 static loff_t bch2_seek_pagecache_data(struct inode *vinode,
 				       loff_t start_offset,
 				       loff_t end_offset,
-				       unsigned min_replicas)
+				       unsigned min_replicas,
+				       bool nonblock)
 {
 	struct folio_batch fbatch;
 	pgoff_t start_index	= start_offset >> PAGE_SHIFT;
@@ -3731,7 +3736,13 @@ static loff_t bch2_seek_pagecache_data(struct inode *vinode,
 		for (i = 0; i < folio_batch_count(&fbatch); i++) {
 			struct folio *folio = fbatch.folios[i];
 
-			folio_lock(folio);
+			if (!nonblock) {
+				folio_lock(folio);
+			} else if (!folio_trylock(folio)) {
+				folio_batch_release(&fbatch);
+				return -EAGAIN;
+			}
+
 			offset = folio_data_offset(folio,
 					max(folio_pos(folio), start_offset),
 					min_replicas);
@@ -3796,26 +3807,32 @@ static loff_t bch2_seek_data(struct file *file, u64 offset)
 
 	if (next_data > offset)
 		next_data = bch2_seek_pagecache_data(&inode->v,
-						     offset, next_data, 0);
+					offset, next_data, 0, false);
 
 	if (next_data >= isize)
 		return -ENXIO;
 
 	return vfs_setpos(file, next_data, MAX_LFS_FILESIZE);
 }
 
-static bool folio_hole_offset(struct address_space *mapping, loff_t *offset,
-			      unsigned min_replicas)
+static int folio_hole_offset(struct address_space *mapping, loff_t *offset,
+			      unsigned min_replicas, bool nonblock)
 {
 	struct folio *folio;
 	struct bch_folio *s;
 	unsigned i, sectors;
 	bool ret = true;
 
-	folio = filemap_lock_folio(mapping, *offset >> PAGE_SHIFT);
+	folio = __filemap_get_folio(mapping, *offset >> PAGE_SHIFT,
+				    !nonblock ? FGP_LOCK : 0, 0);
 	if (IS_ERR_OR_NULL(folio))
 		return true;
 
+	if (nonblock && !folio_trylock(folio)) {
+		folio_put(folio);
+		return -EAGAIN;
+	}
+
 	s = bch2_folio(folio);
 	if (!s)
 		goto unlock;
@@ -3833,37 +3850,51 @@ static bool folio_hole_offset(struct address_space *mapping, loff_t *offset,
 	ret = false;
 unlock:
 	folio_unlock(folio);
+	folio_put(folio);
 	return ret;
 }
 
 static loff_t bch2_seek_pagecache_hole(struct inode *vinode,
 				       loff_t start_offset,
 				       loff_t end_offset,
-				       unsigned min_replicas)
+				       unsigned min_replicas,
+				       bool nonblock)
 {
 	struct address_space *mapping = vinode->i_mapping;
 	loff_t offset = start_offset;
 
 	while (offset < end_offset &&
-	       !folio_hole_offset(mapping, &offset, min_replicas))
+	       !folio_hole_offset(mapping, &offset, min_replicas, nonblock))
 		;
 
 	return min(offset, end_offset);
 }
 
-static void bch2_clamp_data_hole(struct inode *inode,
-				 u64 *hole_start,
-				 u64 *hole_end,
-				 unsigned min_replicas)
+static int bch2_clamp_data_hole(struct inode *inode,
+				u64 *hole_start,
+				u64 *hole_end,
+				unsigned min_replicas,
+				bool nonblock)
 {
-	*hole_start = bch2_seek_pagecache_hole(inode,
-		*hole_start << 9, *hole_end << 9, min_replicas) >> 9;
+	loff_t ret;
+
+	ret = bch2_seek_pagecache_hole(inode,
+		*hole_start << 9, *hole_end << 9, min_replicas, nonblock) >> 9;
+	if (ret < 0)
+		return ret;
+
+	*hole_start = ret;
 
 	if (*hole_start == *hole_end)
-		return;
+		return 0;
 
-	*hole_end = bch2_seek_pagecache_data(inode,
-		*hole_start << 9, *hole_end << 9, min_replicas) >> 9;
+	ret = bch2_seek_pagecache_data(inode,
+		*hole_start << 9, *hole_end << 9, min_replicas, nonblock) >> 9;
+	if (ret < 0)
+		return ret;
+
+	*hole_end = ret;
+	return 0;
 }
 
 static loff_t bch2_seek_hole(struct file *file, u64 offset)
@@ -3895,12 +3926,12 @@ static loff_t bch2_seek_hole(struct file *file, u64 offset)
 			   BTREE_ITER_SLOTS, k, ret) {
 		if (k.k->p.inode != inode->v.i_ino) {
 			next_hole = bch2_seek_pagecache_hole(&inode->v,
-					offset, MAX_LFS_FILESIZE, 0);
+					offset, MAX_LFS_FILESIZE, 0, false);
 			break;
 		} else if (!bkey_extent_is_data(k.k)) {
 			next_hole = bch2_seek_pagecache_hole(&inode->v,
 					max(offset, bkey_start_offset(k.k) << 9),
-					k.k->p.offset << 9, 0);
+					k.k->p.offset << 9, 0, false);
 
 			if (next_hole < k.k->p.offset << 9)
 				break;