Skip to contentSkip to conversation

btrfs: block group cleanups

PATCH Superseded Josef Bacik 3 participants 11 messages Jul 14, 2022 Btrfs file system

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Josef Bacik

author
#0 lore raw

cover btrfs: block group cleanups

I'm reworking our relocation and delete unused block group workqueues which require some cleanups of how we deal with flags on the block group. We've had a bit field for various flags on the block group for a while, but there's a subtle gotcha with this bitfield in that you have to protect every modification with bg->lock in order to not mess with the values, and there were a few places that we weren't holding the lock.

Rework these to be normal flags, and then go behind this conversion and cleanup some of the usage of the different flags. Additionally there's a cleanup around when to break out of the background workers. Thanks,

Josef

Josef Bacik (5):
btrfs: use btrfs_fs_closing for background bg work
btrfs: convert block group bit field to use bit helpers
btrfs: remove block_group->lock protection for TO_COPY
btrfs: simplify btrfs_put_block_group_cache
btrfs: remove BLOCK_GROUP_FLAG_HAS_CACHING_CTL

 fs/btrfs/block-group.c      | 95 ++++++++++++++++---------------------
 fs/btrfs/block-group.h      | 19 ++++----
 fs/btrfs/dev-replace.c      | 11 ++---
 fs/btrfs/extent-tree.c      |  7 ++-
 fs/btrfs/free-space-cache.c | 18 +++----
 fs/btrfs/scrub.c            | 16 +++----
 fs/btrfs/volumes.c          | 13 +++--
 fs/btrfs/zoned.c            | 30 ++++++++----
 8 files changed, 103 insertions(+), 106 deletions(-)
Signature · 1 line
2.26.3
author
#1 lore raw

PATCH 1/5 btrfs: use btrfs_fs_closing for background bg work

For both the deletion of unused block groups and the async balance work we'll only skip running if we don't have BTRFS_FS_OPEN set, however during teardown we park the cleaner kthread and cancel the reclaim work right away, effectively stopping the work from running.

Fix the condition to bail if btrfs_fs_closing() is true instead of if !BTRFS_FS_OPEN, which is cleared much further in the process.

Signed-off-by Josef Bacik <[email protected]>
---
 fs/btrfs/block-group.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
fs/btrfs/block-group.c +2 -2
index c3aecfb0a71d..1e61d967d8be 100644
@@ -1318,7 +1318,7 @@ void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info)
 	const bool async_trim_enabled = btrfs_test_opt(fs_info, DISCARD_ASYNC);
 	int ret = 0;
 
-	if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags))
+	if (btrfs_fs_closing(fs_info))
 		return;
 
 	/*

@@ -1557,7 +1557,7 @@ void btrfs_reclaim_bgs_work(struct work_struct *work)
 	struct btrfs_block_group *bg;
 	struct btrfs_space_info *space_info;
 
-	if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags))
+	if (btrfs_fs_closing(fs_info))
 		return;
 
 	if (!btrfs_should_reclaim(fs_info))
Signature · 1 line
2.26.3
author
#3 lore raw

PATCH 2/5 btrfs: convert block group bit field to use bit helpers

In reply to Josef Bacik

We use a bit field in the btrfs_block_group for different flags, however this is awkward because we have to hold the block_group->lock for any modification of any of these fields, and makes the code clunky for a few of these flags. Convert these to a properly flags setup so we can utilize the bit helpers.

Signed-off-by Josef Bacik <[email protected]>
---
 fs/btrfs/block-group.c      | 21 +++++++++++++--------
 fs/btrfs/block-group.h      | 20 ++++++++++++--------
 fs/btrfs/dev-replace.c      |  6 +++---
 fs/btrfs/extent-tree.c      |  7 +++++--
 fs/btrfs/free-space-cache.c | 18 +++++++++---------
 fs/btrfs/scrub.c            | 13 +++++++------
 fs/btrfs/volumes.c          | 11 ++++++-----
 fs/btrfs/zoned.c            | 30 ++++++++++++++++++++----------
 8 files changed, 75 insertions(+), 51 deletions(-)
fs/btrfs/block-group.c +13 -8
index 1e61d967d8be..e27098571011 100644
@@ -789,7 +789,7 @@ int btrfs_cache_block_group(struct btrfs_block_group *cache, int load_cache_only
 		cache->cached = BTRFS_CACHE_FAST;
 	else
 		cache->cached = BTRFS_CACHE_STARTED;
-	cache->has_caching_ctl = 1;
+	set_bit(BLOCK_GROUP_FLAG_HAS_CACHING_CTL, &cache->runtime_flags);
 	spin_unlock(&cache->lock);
 
 	write_lock(&fs_info->block_group_cache_lock);

@@ -1005,11 +1005,13 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 		kobject_put(kobj);
 	}
 
-	if (block_group->has_caching_ctl)
+	if (test_bit(BLOCK_GROUP_FLAG_HAS_CACHING_CTL,
+		     &block_group->runtime_flags))
 		caching_ctl = btrfs_get_caching_control(block_group);
 	if (block_group->cached == BTRFS_CACHE_STARTED)
 		btrfs_wait_block_group_cache_done(block_group);
-	if (block_group->has_caching_ctl) {
+	if (test_bit(BLOCK_GROUP_FLAG_HAS_CACHING_CTL,
+		     &block_group->runtime_flags)) {
 		write_lock(&fs_info->block_group_cache_lock);
 		if (!caching_ctl) {
 			struct btrfs_caching_control *ctl;

@@ -1086,7 +1088,8 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 		goto out;
 
 	spin_lock(&block_group->lock);
-	block_group->removed = 1;
+	set_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags);
+
 	/*
 	 * At this point trimming or scrub can't start on this block group,
 	 * because we removed the block group from the rbtree

@@ -2440,7 +2443,8 @@ void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans)
 		ret = insert_block_group_item(trans, block_group);
 		if (ret)
 			btrfs_abort_transaction(trans, ret);
-		if (!block_group->chunk_item_inserted) {
+		if (!test_bit(BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED,
+			      &block_group->runtime_flags)) {
 			mutex_lock(&fs_info->chunk_mutex);
 			ret = btrfs_chunk_alloc_add_chunk_item(trans, block_group);
 			mutex_unlock(&fs_info->chunk_mutex);

@@ -3989,7 +3993,8 @@ void btrfs_put_block_group_cache(struct btrfs_fs_info *info)
 		while (block_group) {
 			btrfs_wait_block_group_cache_done(block_group);
 			spin_lock(&block_group->lock);
-			if (block_group->iref)
+			if (test_bit(BLOCK_GROUP_FLAG_IREF,
+				     &block_group->runtime_flags))
 				break;
 			spin_unlock(&block_group->lock);
 			block_group = btrfs_next_block_group(block_group);

@@ -4002,7 +4007,7 @@ void btrfs_put_block_group_cache(struct btrfs_fs_info *info)
 		}
 
 		inode = block_group->inode;
-		block_group->iref = 0;
+		clear_bit(BLOCK_GROUP_FLAG_IREF, &block_group->runtime_flags);
 		block_group->inode = NULL;
 		spin_unlock(&block_group->lock);
 		ASSERT(block_group->io_ctl.inode == NULL);

@@ -4144,7 +4149,7 @@ void btrfs_unfreeze_block_group(struct btrfs_block_group *block_group)
 
 	spin_lock(&block_group->lock);
 	cleanup = (atomic_dec_and_test(&block_group->frozen) &&
-		   block_group->removed);
+		   test_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags));
 	spin_unlock(&block_group->lock);
 
 	if (cleanup) {
fs/btrfs/block-group.h +12 -8
index 35e0e860cc0b..8008a391ed8c 100644
@@ -46,6 +46,17 @@ enum btrfs_chunk_alloc_enum {
 	CHUNK_ALLOC_FORCE_FOR_EXTENT,
 };
 
+enum btrfs_block_group_flags {
+	BLOCK_GROUP_FLAG_IREF,
+	BLOCK_GROUP_FLAG_HAS_CACHING_CTL,
+	BLOCK_GROUP_FLAG_REMOVED,
+	BLOCK_GROUP_FLAG_TO_COPY,
+	BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
+	BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED,
+	BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
+	BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
+};
+
 struct btrfs_caching_control {
 	struct list_head list;
 	struct mutex mutex;

@@ -95,16 +106,9 @@ struct btrfs_block_group {
 
 	/* For raid56, this is a full stripe, without parity */
 	unsigned long full_stripe_len;
+	unsigned long runtime_flags;
 
 	unsigned int ro;
-	unsigned int iref:1;
-	unsigned int has_caching_ctl:1;
-	unsigned int removed:1;
-	unsigned int to_copy:1;
-	unsigned int relocating_repair:1;
-	unsigned int chunk_item_inserted:1;
-	unsigned int zone_is_active:1;
-	unsigned int zoned_data_reloc_ongoing:1;
 
 	int disk_cache_state;
fs/btrfs/dev-replace.c +3 -3
index f43196a893ca..f85bbd99230b 100644
@@ -546,7 +546,7 @@ static int mark_block_group_to_copy(struct btrfs_fs_info *fs_info,
 			continue;
 
 		spin_lock(&cache->lock);
-		cache->to_copy = 1;
+		set_bit(BLOCK_GROUP_FLAG_TO_COPY, &cache->runtime_flags);
 		spin_unlock(&cache->lock);
 
 		btrfs_put_block_group(cache);

@@ -577,7 +577,7 @@ bool btrfs_finish_block_group_to_copy(struct btrfs_device *srcdev,
 		return true;
 
 	spin_lock(&cache->lock);
-	if (cache->removed) {
+	if (test_bit(BLOCK_GROUP_FLAG_REMOVED, &cache->runtime_flags)) {
 		spin_unlock(&cache->lock);
 		return true;
 	}

@@ -611,7 +611,7 @@ bool btrfs_finish_block_group_to_copy(struct btrfs_device *srcdev,
 
 	/* Last stripe on this device */
 	spin_lock(&cache->lock);
-	cache->to_copy = 0;
+	clear_bit(BLOCK_GROUP_FLAG_TO_COPY, &cache->runtime_flags);
 	spin_unlock(&cache->lock);
 
 	return true;
fs/btrfs/extent-tree.c +5 -2
index c6a9237b31b9..151f254d5c11 100644
@@ -3816,7 +3816,9 @@ static int do_allocation_zoned(struct btrfs_block_group *block_group,
 	       block_group->start == fs_info->data_reloc_bg ||
 	       fs_info->data_reloc_bg == 0);
 
-	if (block_group->ro || block_group->zoned_data_reloc_ongoing) {
+	if (block_group->ro ||
+	    test_bit(BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
+		     &block_group->runtime_flags)) {
 		ret = 1;
 		goto out;
 	}

@@ -3893,7 +3895,8 @@ static int do_allocation_zoned(struct btrfs_block_group *block_group,
 		 * regular extents) at the same time to the same zone, which
 		 * easily break the write pointer.
 		 */
-		block_group->zoned_data_reloc_ongoing = 1;
+		set_bit(BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
+			&block_group->runtime_flags);
 		fs_info->data_reloc_bg = 0;
 	}
 	spin_unlock(&fs_info->relocation_bg_lock);
fs/btrfs/free-space-cache.c +9 -9
index 996da650ecdc..fd73327134ac 100644
@@ -126,10 +126,9 @@ struct inode *lookup_free_space_inode(struct btrfs_block_group *block_group,
 		block_group->disk_cache_state = BTRFS_DC_CLEAR;
 	}
 
-	if (!block_group->iref) {
+	if (!test_and_set_bit(BLOCK_GROUP_FLAG_IREF,
+			      &block_group->runtime_flags))
 		block_group->inode = igrab(inode);
-		block_group->iref = 1;
-	}
 	spin_unlock(&block_group->lock);
 
 	return inode;

@@ -241,8 +240,8 @@ int btrfs_remove_free_space_inode(struct btrfs_trans_handle *trans,
 	clear_nlink(inode);
 	/* One for the block groups ref */
 	spin_lock(&block_group->lock);
-	if (block_group->iref) {
-		block_group->iref = 0;
+	if (test_and_clear_bit(BLOCK_GROUP_FLAG_IREF,
+			       &block_group->runtime_flags)) {
 		block_group->inode = NULL;
 		spin_unlock(&block_group->lock);
 		iput(inode);

@@ -2860,7 +2859,8 @@ void btrfs_dump_free_space(struct btrfs_block_group *block_group,
 	if (btrfs_is_zoned(fs_info)) {
 		btrfs_info(fs_info, "free space %llu active %d",
 			   block_group->zone_capacity - block_group->alloc_offset,
-			   block_group->zone_is_active);
+			   test_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
+				    &block_group->runtime_flags));
 		return;
 	}
 

@@ -3992,7 +3992,7 @@ int btrfs_trim_block_group(struct btrfs_block_group *block_group,
 	*trimmed = 0;
 
 	spin_lock(&block_group->lock);
-	if (block_group->removed) {
+	if (test_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags)) {
 		spin_unlock(&block_group->lock);
 		return 0;
 	}

@@ -4022,7 +4022,7 @@ int btrfs_trim_block_group_extents(struct btrfs_block_group *block_group,
 	*trimmed = 0;
 
 	spin_lock(&block_group->lock);
-	if (block_group->removed) {
+	if (test_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags)) {
 		spin_unlock(&block_group->lock);
 		return 0;
 	}

@@ -4044,7 +4044,7 @@ int btrfs_trim_block_group_bitmaps(struct btrfs_block_group *block_group,
 	*trimmed = 0;
 
 	spin_lock(&block_group->lock);
-	if (block_group->removed) {
+	if (test_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags)) {
 		spin_unlock(&block_group->lock);
 		return 0;
 	}
fs/btrfs/scrub.c +7 -6
index 3afe5fa50a63..b7be62f1cd8e 100644
@@ -3266,7 +3266,7 @@ static int scrub_simple_mirror(struct scrub_ctx *sctx,
 		}
 		/* Block group removed? */
 		spin_lock(&bg->lock);
-		if (bg->removed) {
+		if (test_bit(BLOCK_GROUP_FLAG_REMOVED, &bg->runtime_flags)) {
 			spin_unlock(&bg->lock);
 			ret = 0;
 			break;

@@ -3606,7 +3606,7 @@ static noinline_for_stack int scrub_chunk(struct scrub_ctx *sctx,
 		 * kthread or relocation.
 		 */
 		spin_lock(&bg->lock);
-		if (!bg->removed)
+		if (!test_bit(BLOCK_GROUP_FLAG_REMOVED, &bg->runtime_flags))
 			ret = -EINVAL;
 		spin_unlock(&bg->lock);
 

@@ -3765,7 +3765,8 @@ int scrub_enumerate_chunks(struct scrub_ctx *sctx,
 
 		if (sctx->is_dev_replace && btrfs_is_zoned(fs_info)) {
 			spin_lock(&cache->lock);
-			if (!cache->to_copy) {
+			if (!test_bit(BLOCK_GROUP_FLAG_TO_COPY,
+				      &cache->runtime_flags)) {
 				spin_unlock(&cache->lock);
 				btrfs_put_block_group(cache);
 				goto skip;

@@ -3782,7 +3783,7 @@ int scrub_enumerate_chunks(struct scrub_ctx *sctx,
 		 * repair extents.
 		 */
 		spin_lock(&cache->lock);
-		if (cache->removed) {
+		if (test_bit(BLOCK_GROUP_FLAG_REMOVED, &cache->runtime_flags)) {
 			spin_unlock(&cache->lock);
 			btrfs_put_block_group(cache);
 			goto skip;

@@ -3942,8 +3943,8 @@ int scrub_enumerate_chunks(struct scrub_ctx *sctx,
 		 * balance is triggered or it becomes used and unused again.
 		 */
 		spin_lock(&cache->lock);
-		if (!cache->removed && !cache->ro && cache->reserved == 0 &&
-		    cache->used == 0) {
+		if (!test_bit(BLOCK_GROUP_FLAG_REMOVED, &cache->runtime_flags) &&
+		    !cache->ro && cache->reserved == 0 && cache->used == 0) {
 			spin_unlock(&cache->lock);
 			if (btrfs_test_opt(fs_info, DISCARD_ASYNC))
 				btrfs_discard_queue_work(&fs_info->discard_ctl,
fs/btrfs/volumes.c +6 -5
index bf4e140f6bfc..7637eae1a699 100644
@@ -5590,7 +5590,7 @@ int btrfs_chunk_alloc_add_chunk_item(struct btrfs_trans_handle *trans,
 	if (ret)
 		goto out;
 
-	bg->chunk_item_inserted = 1;
+	set_bit(BLOCK_GROUP_FLAG_CHUNK_ITEM_INSERTED, &bg->runtime_flags);
 
 	if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) {
 		ret = btrfs_add_system_chunk(fs_info, &key, chunk, item_size);

@@ -6149,7 +6149,7 @@ static bool is_block_group_to_copy(struct btrfs_fs_info *fs_info, u64 logical)
 	cache = btrfs_lookup_block_group(fs_info, logical);
 
 	spin_lock(&cache->lock);
-	ret = cache->to_copy;
+	ret = test_bit(BLOCK_GROUP_FLAG_TO_COPY, &cache->runtime_flags);
 	spin_unlock(&cache->lock);
 
 	btrfs_put_block_group(cache);

@@ -8243,7 +8243,8 @@ static int relocating_repair_kthread(void *data)
 	if (!cache)
 		goto out;
 
-	if (!cache->relocating_repair)
+	if (!test_bit(BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
+		      &cache->runtime_flags))
 		goto out;
 
 	ret = btrfs_may_alloc_data_chunk(fs_info, target);

@@ -8281,12 +8282,12 @@ bool btrfs_repair_one_zone(struct btrfs_fs_info *fs_info, u64 logical)
 		return true;
 
 	spin_lock(&cache->lock);
-	if (cache->relocating_repair) {
+	if (test_and_set_bit(BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
+			     &cache->runtime_flags)) {
 		spin_unlock(&cache->lock);
 		btrfs_put_block_group(cache);
 		return true;
 	}
-	cache->relocating_repair = 1;
 	spin_unlock(&cache->lock);
 
 	kthread_run(relocating_repair_kthread, cache,
fs/btrfs/zoned.c +20 -10
index 96836dd5de78..02e26e8f8ab2 100644
@@ -1442,7 +1442,9 @@ int btrfs_load_block_group_zone_info(struct btrfs_block_group *cache, bool new)
 		}
 		cache->alloc_offset = alloc_offsets[0];
 		cache->zone_capacity = caps[0];
-		cache->zone_is_active = test_bit(0, active);
+		if (test_bit(0, active))
+			set_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
+				&cache->runtime_flags);
 		break;
 	case BTRFS_BLOCK_GROUP_DUP:
 		if (map->type & BTRFS_BLOCK_GROUP_DATA) {

@@ -1476,7 +1478,9 @@ int btrfs_load_block_group_zone_info(struct btrfs_block_group *cache, bool new)
 				goto out;
 			}
 		} else {
-			cache->zone_is_active = test_bit(0, active);
+			if (test_bit(0, active))
+				set_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
+					&cache->runtime_flags);
 		}
 		cache->alloc_offset = alloc_offsets[0];
 		cache->zone_capacity = min(caps[0], caps[1]);

@@ -1494,7 +1498,7 @@ int btrfs_load_block_group_zone_info(struct btrfs_block_group *cache, bool new)
 		goto out;
 	}
 
-	if (cache->zone_is_active) {
+	if (test_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE, &cache->runtime_flags)) {
 		btrfs_get_block_group(cache);
 		spin_lock(&fs_info->zone_active_bgs_lock);
 		list_add_tail(&cache->active_bg_list, &fs_info->zone_active_bgs);

@@ -1862,7 +1866,8 @@ bool btrfs_zone_activate(struct btrfs_block_group *block_group)
 
 	spin_lock(&space_info->lock);
 	spin_lock(&block_group->lock);
-	if (block_group->zone_is_active) {
+	if (test_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
+		     &block_group->runtime_flags)) {
 		ret = true;
 		goto out_unlock;
 	}

@@ -1888,7 +1893,7 @@ bool btrfs_zone_activate(struct btrfs_block_group *block_group)
 	}
 
 	/* Successfully activated all the zones */
-	block_group->zone_is_active = 1;
+	set_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE, &block_group->runtime_flags);
 	space_info->active_total_bytes += block_group->length;
 	spin_unlock(&block_group->lock);
 	btrfs_try_granting_tickets(fs_info, space_info);

@@ -1917,7 +1922,8 @@ static int do_zone_finish(struct btrfs_block_group *block_group, bool fully_writ
 	int i;
 
 	spin_lock(&block_group->lock);
-	if (!block_group->zone_is_active) {
+	if (!test_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
+		      &block_group->runtime_flags)) {
 		spin_unlock(&block_group->lock);
 		return 0;
 	}

@@ -1956,7 +1962,8 @@ static int do_zone_finish(struct btrfs_block_group *block_group, bool fully_writ
 		 * Bail out if someone already deactivated the block group, or
 		 * allocated space is left in the block group.
 		 */
-		if (!block_group->zone_is_active) {
+		if (!test_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE,
+			      &block_group->runtime_flags)) {
 			spin_unlock(&block_group->lock);
 			btrfs_dec_block_group_ro(block_group);
 			return 0;

@@ -1969,7 +1976,8 @@ static int do_zone_finish(struct btrfs_block_group *block_group, bool fully_writ
 		}
 	}
 
-	block_group->zone_is_active = 0;
+	clear_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE, &block_group->runtime_flags);
+
 	block_group->alloc_offset = block_group->zone_capacity;
 	block_group->free_space_ctl->free_space = 0;
 	btrfs_clear_treelog_bg(block_group);

@@ -2178,13 +2186,15 @@ void btrfs_zoned_release_data_reloc_bg(struct btrfs_fs_info *fs_info, u64 logica
 	ASSERT(block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA));
 
 	spin_lock(&block_group->lock);
-	if (!block_group->zoned_data_reloc_ongoing)
+	if (!test_bit(BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
+		      &block_group->runtime_flags))
 		goto out;
 
 	/* All relocation extents are written. */
 	if (block_group->start + block_group->alloc_offset == logical + length) {
 		/* Now, release this block group for further allocations. */
-		block_group->zoned_data_reloc_ongoing = 0;
+		clear_bit(BLOCK_GROUP_FLAG_ZONED_DATA_RELOC,
+			  &block_group->runtime_flags);
 	}
 
 out:
Signature · 1 line
2.26.3
author
#6 lore raw

PATCH 3/5 btrfs: remove block_group->lock protection for TO_COPY

In reply to Josef Bacik

We use this during device replace for zoned devices, we were simply taking the lock because it was in a bit field and we needed the lock to be safe with other modifications in the bitfield. With the bit helpers we no longer require that locking.

Signed-off-by Josef Bacik <[email protected]>
---
 fs/btrfs/dev-replace.c | 5 -----
 fs/btrfs/scrub.c       | 3 ---
 fs/btrfs/volumes.c     | 2 --
 3 files changed, 10 deletions(-)
fs/btrfs/dev-replace.c +0 -5
index f85bbd99230b..488f2105c5d0 100644
@@ -545,10 +545,7 @@ static int mark_block_group_to_copy(struct btrfs_fs_info *fs_info,
 		if (!cache)
 			continue;
 
-		spin_lock(&cache->lock);
 		set_bit(BLOCK_GROUP_FLAG_TO_COPY, &cache->runtime_flags);
-		spin_unlock(&cache->lock);
-
 		btrfs_put_block_group(cache);
 	}
 	if (iter_ret < 0)

@@ -610,9 +607,7 @@ bool btrfs_finish_block_group_to_copy(struct btrfs_device *srcdev,
 	}
 
 	/* Last stripe on this device */
-	spin_lock(&cache->lock);
 	clear_bit(BLOCK_GROUP_FLAG_TO_COPY, &cache->runtime_flags);
-	spin_unlock(&cache->lock);
 
 	return true;
 }
fs/btrfs/scrub.c +0 -3
index b7be62f1cd8e..14af085fe868 100644
@@ -3764,14 +3764,11 @@ int scrub_enumerate_chunks(struct scrub_ctx *sctx,
 		}
 
 		if (sctx->is_dev_replace && btrfs_is_zoned(fs_info)) {
-			spin_lock(&cache->lock);
 			if (!test_bit(BLOCK_GROUP_FLAG_TO_COPY,
 				      &cache->runtime_flags)) {
-				spin_unlock(&cache->lock);
 				btrfs_put_block_group(cache);
 				goto skip;
 			}
-			spin_unlock(&cache->lock);
 		}
 
 		/*
fs/btrfs/volumes.c +0 -2
index 7637eae1a699..83c9bae144c7 100644
@@ -6148,9 +6148,7 @@ static bool is_block_group_to_copy(struct btrfs_fs_info *fs_info, u64 logical)
 
 	cache = btrfs_lookup_block_group(fs_info, logical);
 
-	spin_lock(&cache->lock);
 	ret = test_bit(BLOCK_GROUP_FLAG_TO_COPY, &cache->runtime_flags);
-	spin_unlock(&cache->lock);
 
 	btrfs_put_block_group(cache);
 	return ret;
Signature · 1 line
2.26.3
author
#8 lore raw

PATCH 4/5 btrfs: simplify btrfs_put_block_group_cache

In reply to Josef Bacik

We're breaking out and re-searching for the next block group while evicting any of the block group cache inodes. This is not needed, the block groups aren't disappearing here, we can simply loop through the block groups like normal and iput any inode that we find.

Signed-off-by Josef Bacik <[email protected]>
---
 fs/btrfs/block-group.c | 42 +++++++++++++++---------------------------
 1 file changed, 15 insertions(+), 27 deletions(-)
fs/btrfs/block-group.c +15 -27
index e27098571011..51096f88b8d3 100644
@@ -3984,36 +3984,24 @@ void btrfs_reserve_chunk_metadata(struct btrfs_trans_handle *trans,
 void btrfs_put_block_group_cache(struct btrfs_fs_info *info)
 {
 	struct btrfs_block_group *block_group;
-	u64 last = 0;
 
-	while (1) {
-		struct inode *inode;
-
-		block_group = btrfs_lookup_first_block_group(info, last);
-		while (block_group) {
-			btrfs_wait_block_group_cache_done(block_group);
-			spin_lock(&block_group->lock);
-			if (test_bit(BLOCK_GROUP_FLAG_IREF,
-				     &block_group->runtime_flags))
-				break;
+	block_group = btrfs_lookup_first_block_group(info, 0);
+	while (block_group) {
+		btrfs_wait_block_group_cache_done(block_group);
+		spin_lock(&block_group->lock);
+		if (test_and_clear_bit(BLOCK_GROUP_FLAG_IREF,
+				       &block_group->runtime_flags)) {
+			struct inode *inode = block_group->inode;
+
+			block_group->inode = NULL;
 			spin_unlock(&block_group->lock);
-			block_group = btrfs_next_block_group(block_group);
-		}
-		if (!block_group) {
-			if (last == 0)
-				break;
-			last = 0;
-			continue;
-		}
 
-		inode = block_group->inode;
-		clear_bit(BLOCK_GROUP_FLAG_IREF, &block_group->runtime_flags);
-		block_group->inode = NULL;
-		spin_unlock(&block_group->lock);
-		ASSERT(block_group->io_ctl.inode == NULL);
-		iput(inode);
-		last = block_group->start + block_group->length;
-		btrfs_put_block_group(block_group);
+			ASSERT(block_group->io_ctl.inode == NULL);
+			iput(inode);
+		} else {
+			spin_unlock(&block_group->lock);
+		}
+		block_group = btrfs_next_block_group(block_group);
 	}
 }
Signature · 1 line
2.26.3
author
#10 lore raw

PATCH 5/5 btrfs: remove BLOCK_GROUP_FLAG_HAS_CACHING_CTL

In reply to Josef Bacik

This is used mostly to determine if we need to look at the caching ctl list and clean up any references to this block group. However we never clear this flag, specifically because we need to know if we have to remove a caching ctl we have for this block group still. This is in the remove block group path which isn't a fast path, so the optimization doesn't really matter, simplify this logic and remove the flag.

Signed-off-by Josef Bacik <[email protected]>
---
 fs/btrfs/block-group.c | 46 +++++++++++++++++++-----------------------
 fs/btrfs/block-group.h |  1 -
 2 files changed, 21 insertions(+), 26 deletions(-)
fs/btrfs/block-group.c +21 -25
index 51096f88b8d3..26202918bc9f 100644
@@ -789,7 +789,6 @@ int btrfs_cache_block_group(struct btrfs_block_group *cache, int load_cache_only
 		cache->cached = BTRFS_CACHE_FAST;
 	else
 		cache->cached = BTRFS_CACHE_STARTED;
-	set_bit(BLOCK_GROUP_FLAG_HAS_CACHING_CTL, &cache->runtime_flags);
 	spin_unlock(&cache->lock);
 
 	write_lock(&fs_info->block_group_cache_lock);

@@ -1005,34 +1004,31 @@ int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
 		kobject_put(kobj);
 	}
 
-	if (test_bit(BLOCK_GROUP_FLAG_HAS_CACHING_CTL,
-		     &block_group->runtime_flags))
-		caching_ctl = btrfs_get_caching_control(block_group);
 	if (block_group->cached == BTRFS_CACHE_STARTED)
 		btrfs_wait_block_group_cache_done(block_group);
-	if (test_bit(BLOCK_GROUP_FLAG_HAS_CACHING_CTL,
-		     &block_group->runtime_flags)) {
-		write_lock(&fs_info->block_group_cache_lock);
-		if (!caching_ctl) {
-			struct btrfs_caching_control *ctl;
-
-			list_for_each_entry(ctl,
-				    &fs_info->caching_block_groups, list)
-				if (ctl->block_group == block_group) {
-					caching_ctl = ctl;
-					refcount_inc(&caching_ctl->count);
-					break;
-				}
-		}
-		if (caching_ctl)
-			list_del_init(&caching_ctl->list);
-		write_unlock(&fs_info->block_group_cache_lock);
-		if (caching_ctl) {
-			/* Once for the caching bgs list and once for us. */
-			btrfs_put_caching_control(caching_ctl);
-			btrfs_put_caching_control(caching_ctl);
+
+	write_lock(&fs_info->block_group_cache_lock);
+	caching_ctl = btrfs_get_caching_control(block_group);
+	if (!caching_ctl) {
+		struct btrfs_caching_control *ctl;
+
+		list_for_each_entry(ctl, &fs_info->caching_block_groups, list) {
+			if (ctl->block_group == block_group) {
+				caching_ctl = ctl;
+				refcount_inc(&caching_ctl->count);
+				break;
+			}
 		}
 	}
+	if (caching_ctl)
+		list_del_init(&caching_ctl->list);
+	write_unlock(&fs_info->block_group_cache_lock);
+
+	if (caching_ctl) {
+		/* Once for the caching bgs list and once for us. */
+		btrfs_put_caching_control(caching_ctl);
+		btrfs_put_caching_control(caching_ctl);
+	}
 
 	spin_lock(&trans->transaction->dirty_bgs_lock);
 	WARN_ON(!list_empty(&block_group->dirty_list));
fs/btrfs/block-group.h +0 -1
index 8008a391ed8c..fffcc7789fa7 100644
@@ -48,7 +48,6 @@ enum btrfs_chunk_alloc_enum {
 
 enum btrfs_block_group_flags {
 	BLOCK_GROUP_FLAG_IREF,
-	BLOCK_GROUP_FLAG_HAS_CACHING_CTL,
 	BLOCK_GROUP_FLAG_REMOVED,
 	BLOCK_GROUP_FLAG_TO_COPY,
 	BLOCK_GROUP_FLAG_RELOCATING_REPAIR,
Signature · 1 line
2.26.3
build bot
#5 lore raw

Re: [PATCH 2/5] btrfs: convert block group bit field to use bit helpers

In reply to Josef Bacik
I love your patch! Yet something to improve:

Hi Josef,

I love your patch! Yet something to improve:

[auto build test ERROR on kdave/for-next]
[also build test ERROR on next-20220714]
[cannot apply to linus/master v5.19-rc6]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Josef-Bacik/btrfs-block-group-cleanups/20220714-083606
base:   https://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux.git for-next
config: hexagon-randconfig-r045-20220714 (https://download.01.org/0day-ci/archive/20220715/[email protected]/config)
compiler: clang version 15.0.0 (https://github.com/llvm/llvm-project 5e61b9c556267086ef9b743a0b57df302eef831b)

reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# https://github.com/intel-lab-lkp/linux/commit/d82c8630686065ad13f9a0a3b73284f9aa3beb2d
git remote add linux-review https://github.com/intel-lab-lkp/linux
git fetch --no-tags linux-review Josef-Bacik/btrfs-block-group-cleanups/20220714-083606
git checkout d82c8630686065ad13f9a0a3b73284f9aa3beb2d
# save the config file
mkdir build_dir && cp config build_dir/.config
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 O=build_dir ARCH=hexagon SHELL=/bin/bash fs/btrfs/

If you fix the issue, kindly add following tag where applicable

Reported-by kernel test robot <[email protected]>

All errors (new ones prefixed by >>):

>> fs/btrfs/block-group.c:1056:24: error: no member named 'zone_is_active' in 'struct btrfs_block_group'
                   WARN_ON(block_group->zone_is_active &&
                           ~~~~~~~~~~~  ^
   include/asm-generic/bug.h:122:25: note: expanded from macro 'WARN_ON'
           int __ret_warn_on = !!(condition);                              \
                                  ^~~~~~~~~
   fs/btrfs/block-group.c:1061:19: error: no member named 'zone_is_active' in 'struct btrfs_block_group'
           if (block_group->zone_is_active)
               ~~~~~~~~~~~  ^
   fs/btrfs/block-group.c:2118:34: error: no member named 'zone_is_active' in 'struct btrfs_block_group'
                                   cache->zone_unusable, cache->zone_is_active,
                                                         ~~~~~  ^
   fs/btrfs/block-group.c:2571:12: error: no member named 'zone_is_active' in 'struct btrfs_block_group'
                                   cache->zone_is_active, &cache->space_info);
                                   ~~~~~  ^
   4 errors generated.
--
>> fs/btrfs/zoned.c:2269:43: error: no member named 'zone_is_active' in 'struct btrfs_block_group'
                                   if (btrfs_zoned_bg_is_full(bg) || bg->zone_is_active) {
                                                                     ~~  ^

1 error generated.

vim +1056 fs/btrfs/block-group.c

7357623a7f4beb4 Qu Wenruo          2020-05-05   888  
e3e0520b32bc3db Josef Bacik        2019-06-20   889  int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
e3e0520b32bc3db Josef Bacik        2019-06-20   890  			     u64 group_start, struct extent_map *em)
e3e0520b32bc3db Josef Bacik        2019-06-20   891  {
e3e0520b32bc3db Josef Bacik        2019-06-20   892  	struct btrfs_fs_info *fs_info = trans->fs_info;
e3e0520b32bc3db Josef Bacik        2019-06-20   893  	struct btrfs_path *path;
32da5386d9a4fd5 David Sterba       2019-10-29   894  	struct btrfs_block_group *block_group;
e3e0520b32bc3db Josef Bacik        2019-06-20   895  	struct btrfs_free_cluster *cluster;
e3e0520b32bc3db Josef Bacik        2019-06-20   896  	struct inode *inode;
e3e0520b32bc3db Josef Bacik        2019-06-20   897  	struct kobject *kobj = NULL;
e3e0520b32bc3db Josef Bacik        2019-06-20   898  	int ret;
e3e0520b32bc3db Josef Bacik        2019-06-20   899  	int index;
e3e0520b32bc3db Josef Bacik        2019-06-20   900  	int factor;
e3e0520b32bc3db Josef Bacik        2019-06-20   901  	struct btrfs_caching_control *caching_ctl = NULL;
e3e0520b32bc3db Josef Bacik        2019-06-20   902  	bool remove_em;
e3e0520b32bc3db Josef Bacik        2019-06-20   903  	bool remove_rsv = false;
e3e0520b32bc3db Josef Bacik        2019-06-20   904  
e3e0520b32bc3db Josef Bacik        2019-06-20   905  	block_group = btrfs_lookup_block_group(fs_info, group_start);
e3e0520b32bc3db Josef Bacik        2019-06-20   906  	BUG_ON(!block_group);
e3e0520b32bc3db Josef Bacik        2019-06-20   907  	BUG_ON(!block_group->ro);
e3e0520b32bc3db Josef Bacik        2019-06-20   908  
e3e0520b32bc3db Josef Bacik        2019-06-20   909  	trace_btrfs_remove_block_group(block_group);
e3e0520b32bc3db Josef Bacik        2019-06-20   910  	/*
e3e0520b32bc3db Josef Bacik        2019-06-20   911  	 * Free the reserved super bytes from this block group before
e3e0520b32bc3db Josef Bacik        2019-06-20   912  	 * remove it.
e3e0520b32bc3db Josef Bacik        2019-06-20   913  	 */
e3e0520b32bc3db Josef Bacik        2019-06-20   914  	btrfs_free_excluded_extents(block_group);
b3470b5dbe1300d David Sterba       2019-10-23   915  	btrfs_free_ref_tree_range(fs_info, block_group->start,
b3470b5dbe1300d David Sterba       2019-10-23   916  				  block_group->length);
e3e0520b32bc3db Josef Bacik        2019-06-20   917  
e3e0520b32bc3db Josef Bacik        2019-06-20   918  	index = btrfs_bg_flags_to_raid_index(block_group->flags);
e3e0520b32bc3db Josef Bacik        2019-06-20   919  	factor = btrfs_bg_type_to_factor(block_group->flags);
e3e0520b32bc3db Josef Bacik        2019-06-20   920  
e3e0520b32bc3db Josef Bacik        2019-06-20   921  	/* make sure this block group isn't part of an allocation cluster */
e3e0520b32bc3db Josef Bacik        2019-06-20   922  	cluster = &fs_info->data_alloc_cluster;
e3e0520b32bc3db Josef Bacik        2019-06-20   923  	spin_lock(&cluster->refill_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20   924  	btrfs_return_cluster_to_free_space(block_group, cluster);
e3e0520b32bc3db Josef Bacik        2019-06-20   925  	spin_unlock(&cluster->refill_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20   926  
e3e0520b32bc3db Josef Bacik        2019-06-20   927  	/*
e3e0520b32bc3db Josef Bacik        2019-06-20   928  	 * make sure this block group isn't part of a metadata
e3e0520b32bc3db Josef Bacik        2019-06-20   929  	 * allocation cluster
e3e0520b32bc3db Josef Bacik        2019-06-20   930  	 */
e3e0520b32bc3db Josef Bacik        2019-06-20   931  	cluster = &fs_info->meta_alloc_cluster;
e3e0520b32bc3db Josef Bacik        2019-06-20   932  	spin_lock(&cluster->refill_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20   933  	btrfs_return_cluster_to_free_space(block_group, cluster);
e3e0520b32bc3db Josef Bacik        2019-06-20   934  	spin_unlock(&cluster->refill_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20   935  
40ab3be102f0a61 Naohiro Aota       2021-02-04   936  	btrfs_clear_treelog_bg(block_group);
c2707a255623435 Johannes Thumshirn 2021-09-09   937  	btrfs_clear_data_reloc_bg(block_group);
40ab3be102f0a61 Naohiro Aota       2021-02-04   938  
e3e0520b32bc3db Josef Bacik        2019-06-20   939  	path = btrfs_alloc_path();
e3e0520b32bc3db Josef Bacik        2019-06-20   940  	if (!path) {
e3e0520b32bc3db Josef Bacik        2019-06-20   941  		ret = -ENOMEM;
9fecd13202f520f Filipe Manana      2020-06-01   942  		goto out;
e3e0520b32bc3db Josef Bacik        2019-06-20   943  	}
e3e0520b32bc3db Josef Bacik        2019-06-20   944  
e3e0520b32bc3db Josef Bacik        2019-06-20   945  	/*
e3e0520b32bc3db Josef Bacik        2019-06-20   946  	 * get the inode first so any iput calls done for the io_list
e3e0520b32bc3db Josef Bacik        2019-06-20   947  	 * aren't the final iput (no unlinks allowed now)
e3e0520b32bc3db Josef Bacik        2019-06-20   948  	 */
e3e0520b32bc3db Josef Bacik        2019-06-20   949  	inode = lookup_free_space_inode(block_group, path);
e3e0520b32bc3db Josef Bacik        2019-06-20   950  
e3e0520b32bc3db Josef Bacik        2019-06-20   951  	mutex_lock(&trans->transaction->cache_write_mutex);
e3e0520b32bc3db Josef Bacik        2019-06-20   952  	/*
e3e0520b32bc3db Josef Bacik        2019-06-20   953  	 * Make sure our free space cache IO is done before removing the
e3e0520b32bc3db Josef Bacik        2019-06-20   954  	 * free space inode
e3e0520b32bc3db Josef Bacik        2019-06-20   955  	 */
e3e0520b32bc3db Josef Bacik        2019-06-20   956  	spin_lock(&trans->transaction->dirty_bgs_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20   957  	if (!list_empty(&block_group->io_list)) {
e3e0520b32bc3db Josef Bacik        2019-06-20   958  		list_del_init(&block_group->io_list);
e3e0520b32bc3db Josef Bacik        2019-06-20   959  
e3e0520b32bc3db Josef Bacik        2019-06-20   960  		WARN_ON(!IS_ERR(inode) && inode != block_group->io_ctl.inode);
e3e0520b32bc3db Josef Bacik        2019-06-20   961  
e3e0520b32bc3db Josef Bacik        2019-06-20   962  		spin_unlock(&trans->transaction->dirty_bgs_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20   963  		btrfs_wait_cache_io(trans, block_group, path);
e3e0520b32bc3db Josef Bacik        2019-06-20   964  		btrfs_put_block_group(block_group);
e3e0520b32bc3db Josef Bacik        2019-06-20   965  		spin_lock(&trans->transaction->dirty_bgs_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20   966  	}
e3e0520b32bc3db Josef Bacik        2019-06-20   967  
e3e0520b32bc3db Josef Bacik        2019-06-20   968  	if (!list_empty(&block_group->dirty_list)) {
e3e0520b32bc3db Josef Bacik        2019-06-20   969  		list_del_init(&block_group->dirty_list);
e3e0520b32bc3db Josef Bacik        2019-06-20   970  		remove_rsv = true;
e3e0520b32bc3db Josef Bacik        2019-06-20   971  		btrfs_put_block_group(block_group);
e3e0520b32bc3db Josef Bacik        2019-06-20   972  	}
e3e0520b32bc3db Josef Bacik        2019-06-20   973  	spin_unlock(&trans->transaction->dirty_bgs_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20   974  	mutex_unlock(&trans->transaction->cache_write_mutex);
e3e0520b32bc3db Josef Bacik        2019-06-20   975  
36b216c85eb9d7f Boris Burkov       2020-11-18   976  	ret = btrfs_remove_free_space_inode(trans, inode, block_group);
e3e0520b32bc3db Josef Bacik        2019-06-20   977  	if (ret)
9fecd13202f520f Filipe Manana      2020-06-01   978  		goto out;
e3e0520b32bc3db Josef Bacik        2019-06-20   979  
16b0c2581e3a81e Filipe Manana      2022-04-13   980  	write_lock(&fs_info->block_group_cache_lock);
08dddb2951c96b5 Filipe Manana      2022-04-13   981  	rb_erase_cached(&block_group->cache_node,
e3e0520b32bc3db Josef Bacik        2019-06-20   982  			&fs_info->block_group_cache_tree);
e3e0520b32bc3db Josef Bacik        2019-06-20   983  	RB_CLEAR_NODE(&block_group->cache_node);
e3e0520b32bc3db Josef Bacik        2019-06-20   984  
9fecd13202f520f Filipe Manana      2020-06-01   985  	/* Once for the block groups rbtree */
9fecd13202f520f Filipe Manana      2020-06-01   986  	btrfs_put_block_group(block_group);
9fecd13202f520f Filipe Manana      2020-06-01   987  
16b0c2581e3a81e Filipe Manana      2022-04-13   988  	write_unlock(&fs_info->block_group_cache_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20   989  
e3e0520b32bc3db Josef Bacik        2019-06-20   990  	down_write(&block_group->space_info->groups_sem);
e3e0520b32bc3db Josef Bacik        2019-06-20   991  	/*
e3e0520b32bc3db Josef Bacik        2019-06-20   992  	 * we must use list_del_init so people can check to see if they
e3e0520b32bc3db Josef Bacik        2019-06-20   993  	 * are still on the list after taking the semaphore
e3e0520b32bc3db Josef Bacik        2019-06-20   994  	 */
e3e0520b32bc3db Josef Bacik        2019-06-20   995  	list_del_init(&block_group->list);
e3e0520b32bc3db Josef Bacik        2019-06-20   996  	if (list_empty(&block_group->space_info->block_groups[index])) {
e3e0520b32bc3db Josef Bacik        2019-06-20   997  		kobj = block_group->space_info->block_group_kobjs[index];
e3e0520b32bc3db Josef Bacik        2019-06-20   998  		block_group->space_info->block_group_kobjs[index] = NULL;
e3e0520b32bc3db Josef Bacik        2019-06-20   999  		clear_avail_alloc_bits(fs_info, block_group->flags);
e3e0520b32bc3db Josef Bacik        2019-06-20  1000  	}
e3e0520b32bc3db Josef Bacik        2019-06-20  1001  	up_write(&block_group->space_info->groups_sem);
e3e0520b32bc3db Josef Bacik        2019-06-20  1002  	clear_incompat_bg_bits(fs_info, block_group->flags);
e3e0520b32bc3db Josef Bacik        2019-06-20  1003  	if (kobj) {
e3e0520b32bc3db Josef Bacik        2019-06-20  1004  		kobject_del(kobj);
e3e0520b32bc3db Josef Bacik        2019-06-20  1005  		kobject_put(kobj);
e3e0520b32bc3db Josef Bacik        2019-06-20  1006  	}
e3e0520b32bc3db Josef Bacik        2019-06-20  1007  
d82c8630686065a Josef Bacik        2022-07-13  1008  	if (test_bit(BLOCK_GROUP_FLAG_HAS_CACHING_CTL,
d82c8630686065a Josef Bacik        2022-07-13  1009  		     &block_group->runtime_flags))
e3e0520b32bc3db Josef Bacik        2019-06-20  1010  		caching_ctl = btrfs_get_caching_control(block_group);
e3e0520b32bc3db Josef Bacik        2019-06-20  1011  	if (block_group->cached == BTRFS_CACHE_STARTED)
e3e0520b32bc3db Josef Bacik        2019-06-20  1012  		btrfs_wait_block_group_cache_done(block_group);
d82c8630686065a Josef Bacik        2022-07-13  1013  	if (test_bit(BLOCK_GROUP_FLAG_HAS_CACHING_CTL,
d82c8630686065a Josef Bacik        2022-07-13  1014  		     &block_group->runtime_flags)) {
16b0c2581e3a81e Filipe Manana      2022-04-13  1015  		write_lock(&fs_info->block_group_cache_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20  1016  		if (!caching_ctl) {
e3e0520b32bc3db Josef Bacik        2019-06-20  1017  			struct btrfs_caching_control *ctl;
e3e0520b32bc3db Josef Bacik        2019-06-20  1018  
e3e0520b32bc3db Josef Bacik        2019-06-20  1019  			list_for_each_entry(ctl,
e3e0520b32bc3db Josef Bacik        2019-06-20  1020  				    &fs_info->caching_block_groups, list)
e3e0520b32bc3db Josef Bacik        2019-06-20  1021  				if (ctl->block_group == block_group) {
e3e0520b32bc3db Josef Bacik        2019-06-20  1022  					caching_ctl = ctl;
e3e0520b32bc3db Josef Bacik        2019-06-20  1023  					refcount_inc(&caching_ctl->count);
e3e0520b32bc3db Josef Bacik        2019-06-20  1024  					break;
e3e0520b32bc3db Josef Bacik        2019-06-20  1025  				}
e3e0520b32bc3db Josef Bacik        2019-06-20  1026  		}
e3e0520b32bc3db Josef Bacik        2019-06-20  1027  		if (caching_ctl)
e3e0520b32bc3db Josef Bacik        2019-06-20  1028  			list_del_init(&caching_ctl->list);
16b0c2581e3a81e Filipe Manana      2022-04-13  1029  		write_unlock(&fs_info->block_group_cache_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20  1030  		if (caching_ctl) {
e3e0520b32bc3db Josef Bacik        2019-06-20  1031  			/* Once for the caching bgs list and once for us. */
e3e0520b32bc3db Josef Bacik        2019-06-20  1032  			btrfs_put_caching_control(caching_ctl);
e3e0520b32bc3db Josef Bacik        2019-06-20  1033  			btrfs_put_caching_control(caching_ctl);
e3e0520b32bc3db Josef Bacik        2019-06-20  1034  		}
e3e0520b32bc3db Josef Bacik        2019-06-20  1035  	}
e3e0520b32bc3db Josef Bacik        2019-06-20  1036  
e3e0520b32bc3db Josef Bacik        2019-06-20  1037  	spin_lock(&trans->transaction->dirty_bgs_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20  1038  	WARN_ON(!list_empty(&block_group->dirty_list));
e3e0520b32bc3db Josef Bacik        2019-06-20  1039  	WARN_ON(!list_empty(&block_group->io_list));
e3e0520b32bc3db Josef Bacik        2019-06-20  1040  	spin_unlock(&trans->transaction->dirty_bgs_lock);
e3e0520b32bc3db Josef Bacik        2019-06-20  1041  
e3e0520b32bc3db Josef Bacik        2019-06-20  1042  	btrfs_remove_free_space_cache(block_group);
e3e0520b32bc3db Josef Bacik        2019-06-20  1043  
e3e0520b32bc3db Josef Bacik        2019-06-20  1044  	spin_lock(&block_group->space_info->lock);
e3e0520b32bc3db Josef Bacik        2019-06-20  1045  	list_del_init(&block_group->ro_list);
e3e0520b32bc3db Josef Bacik        2019-06-20  1046  
e3e0520b32bc3db Josef Bacik        2019-06-20  1047  	if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) {
e3e0520b32bc3db Josef Bacik        2019-06-20  1048  		WARN_ON(block_group->space_info->total_bytes
b3470b5dbe1300d David Sterba       2019-10-23  1049  			< block_group->length);
e3e0520b32bc3db Josef Bacik        2019-06-20  1050  		WARN_ON(block_group->space_info->bytes_readonly
169e0da91a21a57 Naohiro Aota       2021-02-04  1051  			< block_group->length - block_group->zone_unusable);
169e0da91a21a57 Naohiro Aota       2021-02-04  1052  		WARN_ON(block_group->space_info->bytes_zone_unusable
169e0da91a21a57 Naohiro Aota       2021-02-04  1053  			< block_group->zone_unusable);
e3e0520b32bc3db Josef Bacik        2019-06-20  1054  		WARN_ON(block_group->space_info->disk_total
b3470b5dbe1300d David Sterba       2019-10-23  1055  			< block_group->length * factor);
5c21202df2c7336 Naohiro Aota       2022-07-09 @1056  		WARN_ON(block_group->zone_is_active &&
5c21202df2c7336 Naohiro Aota       2022-07-09  1057  			block_group->space_info->active_total_bytes
5c21202df2c7336 Naohiro Aota       2022-07-09  1058  			< block_group->length);
e3e0520b32bc3db Josef Bacik        2019-06-20  1059  	}
b3470b5dbe1300d David Sterba       2019-10-23  1060  	block_group->space_info->total_bytes -= block_group->length;
5c21202df2c7336 Naohiro Aota       2022-07-09  1061  	if (block_group->zone_is_active)
5c21202df2c7336 Naohiro Aota       2022-07-09  1062  		block_group->space_info->active_total_bytes -= block_group->length;
169e0da91a21a57 Naohiro Aota       2021-02-04  1063  	block_group->space_info->bytes_readonly -=
169e0da91a21a57 Naohiro Aota       2021-02-04  1064  		(block_group->length - block_group->zone_unusable);
169e0da91a21a57 Naohiro Aota       2021-02-04  1065  	block_group->space_info->bytes_zone_unusable -=
169e0da91a21a57 Naohiro Aota       2021-02-04  1066  		block_group->zone_unusable;
b3470b5dbe1300d David Sterba       2019-10-23  1067  	block_group->space_info->disk_total -= block_group->length * factor;
e3e0520b32bc3db Josef Bacik        2019-06-20  1068  
e3e0520b32bc3db Josef Bacik        2019-06-20  1069  	spin_unlock(&block_group->space_info->lock);
e3e0520b32bc3db Josef Bacik        2019-06-20  1070  
ffcb9d44572afba Filipe Manana      2020-06-01  1071  	/*
ffcb9d44572afba Filipe Manana      2020-06-01  1072  	 * Remove the free space for the block group from the free space tree
ffcb9d44572afba Filipe Manana      2020-06-01  1073  	 * and the block group's item from the extent tree before marking the
ffcb9d44572afba Filipe Manana      2020-06-01  1074  	 * block group as removed. This is to prevent races with tasks that
ffcb9d44572afba Filipe Manana      2020-06-01  1075  	 * freeze and unfreeze a block group, this task and another task
ffcb9d44572afba Filipe Manana      2020-06-01  1076  	 * allocating a new block group - the unfreeze task ends up removing
ffcb9d44572afba Filipe Manana      2020-06-01  1077  	 * the block group's extent map before the task calling this function
ffcb9d44572afba Filipe Manana      2020-06-01  1078  	 * deletes the block group item from the extent tree, allowing for
ffcb9d44572afba Filipe Manana      2020-06-01  1079  	 * another task to attempt to create another block group with the same
ffcb9d44572afba Filipe Manana      2020-06-01  1080  	 * item key (and failing with -EEXIST and a transaction abort).
ffcb9d44572afba Filipe Manana      2020-06-01  1081  	 */
ffcb9d44572afba Filipe Manana      2020-06-01  1082  	ret = remove_block_group_free_space(trans, block_group);
ffcb9d44572afba Filipe Manana      2020-06-01  1083  	if (ret)
ffcb9d44572afba Filipe Manana      2020-06-01  1084  		goto out;
ffcb9d44572afba Filipe Manana      2020-06-01  1085  
ffcb9d44572afba Filipe Manana      2020-06-01  1086  	ret = remove_block_group_item(trans, path, block_group);
ffcb9d44572afba Filipe Manana      2020-06-01  1087  	if (ret < 0)
ffcb9d44572afba Filipe Manana      2020-06-01  1088  		goto out;
ffcb9d44572afba Filipe Manana      2020-06-01  1089  
e3e0520b32bc3db Josef Bacik        2019-06-20  1090  	spin_lock(&block_group->lock);
d82c8630686065a Josef Bacik        2022-07-13  1091  	set_bit(BLOCK_GROUP_FLAG_REMOVED, &block_group->runtime_flags);
d82c8630686065a Josef Bacik        2022-07-13  1092  
e3e0520b32bc3db Josef Bacik        2019-06-20  1093  	/*
6b7304af62d02d7 Filipe Manana      2020-05-08  1094  	 * At this point trimming or scrub can't start on this block group,
6b7304af62d02d7 Filipe Manana      2020-05-08  1095  	 * because we removed the block group from the rbtree
6b7304af62d02d7 Filipe Manana      2020-05-08  1096  	 * fs_info->block_group_cache_tree so no one can't find it anymore and
6b7304af62d02d7 Filipe Manana      2020-05-08  1097  	 * even if someone already got this block group before we removed it
6b7304af62d02d7 Filipe Manana      2020-05-08  1098  	 * from the rbtree, they have already incremented block_group->frozen -
6b7304af62d02d7 Filipe Manana      2020-05-08  1099  	 * if they didn't, for the trimming case they won't find any free space
6b7304af62d02d7 Filipe Manana      2020-05-08  1100  	 * entries because we already removed them all when we called
6b7304af62d02d7 Filipe Manana      2020-05-08  1101  	 * btrfs_remove_free_space_cache().
e3e0520b32bc3db Josef Bacik        2019-06-20  1102  	 *
e3e0520b32bc3db Josef Bacik        2019-06-20  1103  	 * And we must not remove the extent map from the fs_info->mapping_tree
e3e0520b32bc3db Josef Bacik        2019-06-20  1104  	 * to prevent the same logical address range and physical device space
6b7304af62d02d7 Filipe Manana      2020-05-08  1105  	 * ranges from being reused for a new block group. This is needed to
6b7304af62d02d7 Filipe Manana      2020-05-08  1106  	 * avoid races with trimming and scrub.
6b7304af62d02d7 Filipe Manana      2020-05-08  1107  	 *
6b7304af62d02d7 Filipe Manana      2020-05-08  1108  	 * An fs trim operation (btrfs_trim_fs() / btrfs_ioctl_fitrim()) is
e3e0520b32bc3db Josef Bacik        2019-06-20  1109  	 * completely transactionless, so while it is trimming a range the
e3e0520b32bc3db Josef Bacik        2019-06-20  1110  	 * currently running transaction might finish and a new one start,
e3e0520b32bc3db Josef Bacik        2019-06-20  1111  	 * allowing for new block groups to be created that can reuse the same
e3e0520b32bc3db Josef Bacik        2019-06-20  1112  	 * physical device locations unless we take this special care.
e3e0520b32bc3db Josef Bacik        2019-06-20  1113  	 *
e3e0520b32bc3db Josef Bacik        2019-06-20  1114  	 * There may also be an implicit trim operation if the file system
e3e0520b32bc3db Josef Bacik        2019-06-20  1115  	 * is mounted with -odiscard. The same protections must remain
e3e0520b32bc3db Josef Bacik        2019-06-20  1116  	 * in place until the extents have been discarded completely when
e3e0520b32bc3db Josef Bacik        2019-06-20  1117  	 * the transaction commit has completed.
e3e0520b32bc3db Josef Bacik        2019-06-20  1118  	 */
6b7304af62d02d7 Filipe Manana      2020-05-08  1119  	remove_em = (atomic_read(&block_group->frozen) == 0);
e3e0520b32bc3db Josef Bacik        2019-06-20  1120  	spin_unlock(&block_group->lock);
e3e0520b32bc3db Josef Bacik        2019-06-20  1121  
e3e0520b32bc3db Josef Bacik        2019-06-20  1122  	if (remove_em) {
e3e0520b32bc3db Josef Bacik        2019-06-20  1123  		struct extent_map_tree *em_tree;
e3e0520b32bc3db Josef Bacik        2019-06-20  1124  
e3e0520b32bc3db Josef Bacik        2019-06-20  1125  		em_tree = &fs_info->mapping_tree;
e3e0520b32bc3db Josef Bacik        2019-06-20  1126  		write_lock(&em_tree->lock);
e3e0520b32bc3db Josef Bacik        2019-06-20  1127  		remove_extent_mapping(em_tree, em);
e3e0520b32bc3db Josef Bacik        2019-06-20  1128  		write_unlock(&em_tree->lock);
e3e0520b32bc3db Josef Bacik        2019-06-20  1129  		/* once for the tree */
e3e0520b32bc3db Josef Bacik        2019-06-20  1130  		free_extent_map(em);
e3e0520b32bc3db Josef Bacik        2019-06-20  1131  	}
f6033c5e333238f Xiyu Yang          2020-04-21  1132  
9fecd13202f520f Filipe Manana      2020-06-01  1133  out:
f6033c5e333238f Xiyu Yang          2020-04-21  1134  	/* Once for the lookup reference */
f6033c5e333238f Xiyu Yang          2020-04-21  1135  	btrfs_put_block_group(block_group);
e3e0520b32bc3db Josef Bacik        2019-06-20  1136  	if (remove_rsv)
e3e0520b32bc3db Josef Bacik        2019-06-20  1137  		btrfs_delayed_refs_rsv_release(fs_info, 1);
e3e0520b32bc3db Josef Bacik        2019-06-20  1138  	btrfs_free_path(path);
e3e0520b32bc3db Josef Bacik        2019-06-20  1139  	return ret;
e3e0520b32bc3db Josef Bacik        2019-06-20  1140  }
e3e0520b32bc3db Josef Bacik        2019-06-20  1141
Signature · 2 lines
0-DAY CI Kernel Test Service
https://01.org/lkp