[PATCH v3 0/5] btrfs: zoned: fix deadlock and space reporting issues for zoned filesystems

[PATCH v3 0/5] btrfs: zoned: fix deadlock and space reporting issues for zoned filesystems

[Johannes Thumshirn]

This series fixes premature ENOSPC errors and deadlock issues on zoned BTRFS
filesystems when running xfstest generic/747, which tests garbage collection
on zoned block devices using direct and buffered I/O.

There where two issues revealed in the investigation:

Async reclaim deadlock: On zoned filesystems, the flush state machine
only executes RECLAIM_ZONES and RESET_ZONES in later flush states
(FLUSH_DELALLOC and beyond). By the time these states are reached,
ticket waiters can deadlock waiting for space that can only be freed by
zone reset. The fix adds RECLAIM_ZONES and RESET_ZONES to the first
async reclaim loop (FLUSH_ALLOC) specifically for zoned filesystems,
ensuring zone reset happens early enough to free space for pending
allocation tickets.

Additionally, the series fixes a bug in data relocation block group selection
where the first block group was incorrectly skipped, and adds a new flush
state (BTRFS_RESERVE_FLUSH_DATA_RELOCATION) to use priority reclaim for
zoned data relocation operations.

Using this, generic/747 passes when lowering the fill_percent variable
to 80% (from the 95%) that XFS uses. I'm still examining if this is due
to XFS not having metadata on zoned or other statfs space estimation
bugs or something else.

Changes to v2:
- Re-do the data-relocation block-group selection algorithm

Changes to v1:
- Removed the zone_unusable statfs patch
- Removed the heavy handed flushing patch
- Added fixes tags where needed
- Added more comments

Johannes Thumshirn (5):
  btrfs: zoned: document RECLAIM_ZONES flush state
  btrfs: zoned: decode 'RECLAIM_ZONES' state in tracepoints
  btrfs: zoned: always set data_relocation_bg
  btrfs: zoned: don't account data relocation space-info in statfs free
    space
  btrfs: zoned: fix deadlock waiting for ticket during data relocation

 fs/btrfs/delalloc-space.c    |  2 ++
 fs/btrfs/space-info.c        |  9 +++++++++
 fs/btrfs/space-info.h        | 11 +++++++++++
 fs/btrfs/super.c             |  3 ++-
 fs/btrfs/zoned.c             |  8 +-------
 include/trace/events/btrfs.h |  1 +
 6 files changed, 26 insertions(+), 8 deletions(-)

-- 
2.54.0

[PATCH v3 1/5] btrfs: zoned: document RECLAIM_ZONES flush state

[Johannes Thumshirn]

Document the purpose of the RECLAIM_ZONES flush state.

Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
---
 fs/btrfs/space-info.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/fs/btrfs/space-info.c b/fs/btrfs/space-info.c
index f0436eea1544..739984462677 100644
--- a/fs/btrfs/space-info.c
+++ b/fs/btrfs/space-info.c
@@ -1411,6 +1411,13 @@ static void btrfs_preempt_reclaim_metadata_space(struct work_struct *work)
  *   This is where we reclaim all of the pinned space generated by running the
  *   iputs
  *
+ * RECLAIM_ZONES
+ *   This state only works for the zoned mode. We scan the block groups in the
+ *   reclaim_bgs_list and check if we can relocate them. If yes perform the
+ *   relocation to garbage collect the zone. On each of these runs
+ *   BTRFS_ZONED_SYNC_RECLAIM_BATCH (5) block-groups will be reclaimed, after all
+ *   unused block-groups have been deleted.
+ *
  * RESET_ZONES
  *   This state works only for the zoned mode. We scan the unused block group
  *   list and reset the zones and reuse the block group.
-- 
2.54.0

[PATCH v3 2/5] btrfs: zoned: decode 'RECLAIM_ZONES' state in tracepoints

[Johannes Thumshirn]

Decode the 'RECLAIM_ZONES' state in tracepoints, as of now only the
numerical state is shown in the tracepoints.

Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
---
 include/trace/events/btrfs.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/include/trace/events/btrfs.h b/include/trace/events/btrfs.h
index ec1df8b94517..ed272a100fa8 100644
--- a/include/trace/events/btrfs.h
+++ b/include/trace/events/btrfs.h
@@ -101,6 +101,7 @@ struct find_free_extent_ctl;
 	EM( ALLOC_CHUNK_FORCE,		"ALLOC_CHUNK_FORCE")		\
 	EM( RUN_DELAYED_IPUTS,		"RUN_DELAYED_IPUTS")		\
 	EM( COMMIT_TRANS,		"COMMIT_TRANS")			\
+	EM( RECLAIM_ZONES,		"RECLAIM_ZONES")		\
 	EMe(RESET_ZONES,		"RESET_ZONES")
 
 /*
-- 
2.54.0

[PATCH v3 3/5] btrfs: zoned: always set data_relocation_bg

[Johannes Thumshirn]

When searching for a data relocation block-group on mount,
btrfs_zoned_reserve_data_reloc_bg() is looking for the first empty DATA
block-group. But it first checks if the block-group is empty and if yes
continues the search, and then checks if it is the first DATA block-group.

There is actually no point in looking for the second empty DATA block
group as new DATA allocations will just allocate a new chunk for it. Pick
the first DATA block-group without any allocations done and set it as
relocation block-group.

Singed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
---
 fs/btrfs/zoned.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/fs/btrfs/zoned.c b/fs/btrfs/zoned.c
index 16dd87aa06f2..a4d2fb774f72 100644
--- a/fs/btrfs/zoned.c
+++ b/fs/btrfs/zoned.c
@@ -2763,7 +2763,6 @@ void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info)
 	struct btrfs_block_group *bg;
 	struct list_head *bg_list;
 	u64 alloc_flags;
-	bool first = true;
 	bool did_chunk_alloc = false;
 	int index;
 	int ret;
@@ -2784,13 +2783,9 @@ void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info)
 again:
 	bg_list = &space_info->block_groups[index];
 	list_for_each_entry(bg, bg_list, list) {
-		if (bg->alloc_offset != 0)
-			continue;
 
-		if (first) {
-			first = false;
+		if (bg->alloc_offset != 0)
 			continue;
-		}
 
 		if (space_info == data_sinfo) {
 			/* Migrate the block group to the data relocation space_info. */
@@ -2849,7 +2844,6 @@ void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info)
 		 * We allocated a new block group in the data relocation space_info. We
 		 * can take that one.
 		 */
-		first = false;
 		did_chunk_alloc = true;
 		goto again;
 	}
-- 
2.54.0

[PATCH v3 4/5] btrfs: zoned: don't account data relocation space-info in statfs free space

[Johannes Thumshirn]

Don't account the free space in a data relocation space-info sub-group as
usable free space in statfs.

This is misleading as no user allocations can be made in this space-info
sub-group. It is only a target for relocation.

Fixes: f92ee31e031c ("btrfs: introduce btrfs_space_info sub-group")
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
---
 fs/btrfs/super.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index fb15decb0861..a0dbc0d2213f 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -1739,7 +1739,8 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	int mixed = 0;
 
 	list_for_each_entry(found, &fs_info->space_info, list) {
-		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
+		if (found->flags & BTRFS_BLOCK_GROUP_DATA &&
+		    found->subgroup_id != BTRFS_SUB_GROUP_DATA_RELOC) {
 			int i;
 
 			total_free_data += found->disk_total - found->disk_used;
-- 
2.54.0

[PATCH v3 5/5] btrfs: zoned: fix deadlock waiting for ticket during data relocation

[Johannes Thumshirn]

When performing data relocation on a zoned filesystem, BTRFS can deadlock
in handle_reserve_tickets(). The relocation process is waiting on a space
reservation ticket that can never be fulfilled, because the relocation
itself is the operation responsible for freeing up that space.

Fix this by introducing a new flush state,
BTRFS_RESERVE_FLUSH_ZONED_RELOCATION, specifically for data chunk
allocation during zoned relocation. Like
BTRFS_RESERVE_FLUSH_FREE_SPACE_INODE, this state uses
priority_reclaim_data_space() instead of the normal flushing path, which
avoids re-entering the relocation code and breaking the deadlock cycle.

In btrfs_alloc_data_chunk_ondemand(), select this new flush state when the
inode belongs to a data relocation root on a zoned filesystem.

Fixes: e2a7fd22378f ("btrfs: zoned: add zone reclaim flush state for DATA space_info")
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
---
 fs/btrfs/delalloc-space.c |  2 ++
 fs/btrfs/space-info.c     |  2 ++
 fs/btrfs/space-info.h     | 11 +++++++++++
 3 files changed, 15 insertions(+)

diff --git a/fs/btrfs/delalloc-space.c b/fs/btrfs/delalloc-space.c
index 0970799d0aa4..4293a6383433 100644
--- a/fs/btrfs/delalloc-space.c
+++ b/fs/btrfs/delalloc-space.c
@@ -134,6 +134,8 @@ int btrfs_alloc_data_chunk_ondemand(const struct btrfs_inode *inode, u64 bytes)
 
 	if (btrfs_is_free_space_inode(inode))
 		flush = BTRFS_RESERVE_FLUSH_FREE_SPACE_INODE;
+	else if (btrfs_is_zoned(fs_info) && btrfs_is_data_reloc_root(root))
+		flush = BTRFS_RESERVE_FLUSH_ZONED_RELOCATION;
 
 	return btrfs_reserve_data_bytes(data_sinfo_for_inode(inode), bytes, flush);
 }
diff --git a/fs/btrfs/space-info.c b/fs/btrfs/space-info.c
index 739984462677..e6641597b321 100644
--- a/fs/btrfs/space-info.c
+++ b/fs/btrfs/space-info.c
@@ -1705,6 +1705,7 @@ static int handle_reserve_ticket(struct btrfs_space_info *space_info,
 						ARRAY_SIZE(evict_flush_states));
 		break;
 	case BTRFS_RESERVE_FLUSH_FREE_SPACE_INODE:
+	case BTRFS_RESERVE_FLUSH_ZONED_RELOCATION:
 		priority_reclaim_data_space(space_info, ticket);
 		break;
 	default:
@@ -1968,6 +1969,7 @@ int btrfs_reserve_data_bytes(struct btrfs_space_info *space_info, u64 bytes,
 
 	ASSERT(flush == BTRFS_RESERVE_FLUSH_DATA ||
 	       flush == BTRFS_RESERVE_FLUSH_FREE_SPACE_INODE ||
+	       flush == BTRFS_RESERVE_FLUSH_ZONED_RELOCATION ||
 	       flush == BTRFS_RESERVE_NO_FLUSH, "flush=%d", flush);
 	ASSERT(!current->journal_info || flush != BTRFS_RESERVE_FLUSH_DATA,
 	       "current->journal_info=0x%lx flush=%d",
diff --git a/fs/btrfs/space-info.h b/fs/btrfs/space-info.h
index 24f45072ca4b..aa836e8a9d4a 100644
--- a/fs/btrfs/space-info.h
+++ b/fs/btrfs/space-info.h
@@ -77,6 +77,17 @@ enum btrfs_reserve_flush_enum {
 	 */
 	BTRFS_RESERVE_FLUSH_ALL_STEAL,
 
+	/*
+	 * This is for relocation on zoned filesystems only. We need to use
+	 * priority flushing for this, because otherwise we can deadlock on
+	 * waiting for a ticket, that cannot be granted, because we cannot do
+	 * any allocations.
+	 *
+	 * Apart from being specific to zoned relocation, it is equal to
+	 * BTRFS_FLUSH_FREE_SPACE_INODE.
+	 */
+	BTRFS_RESERVE_FLUSH_ZONED_RELOCATION,
+
 	/*
 	 * This is for btrfs_use_block_rsv only.  We have exhausted our block
 	 * rsv and our global block rsv.  This can happen for things like
-- 
2.54.0

Re: [PATCH v3 3/5] btrfs: zoned: always set data_relocation_bg

[Boris Burkov]

On Fri, May 22, 2026 at 11:02:45AM +0200, Johannes Thumshirn wrote:
> When searching for a data relocation block-group on mount,
> btrfs_zoned_reserve_data_reloc_bg() is looking for the first empty DATA
> block-group. But it first checks if the block-group is empty and if yes
> continues the search, and then checks if it is the first DATA block-group.
> 
> There is actually no point in looking for the second empty DATA block
> group as new DATA allocations will just allocate a new chunk for it. Pick
> the first DATA block-group without any allocations done and set it as
> relocation block-group.
> 

Looks good, thanks.
Reviewed-by: Boris Burkov <boris@bur.io>

> Singed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
> ---
>  fs/btrfs/zoned.c | 8 +-------
>  1 file changed, 1 insertion(+), 7 deletions(-)
> 
> diff --git a/fs/btrfs/zoned.c b/fs/btrfs/zoned.c
> index 16dd87aa06f2..a4d2fb774f72 100644
> --- a/fs/btrfs/zoned.c
> +++ b/fs/btrfs/zoned.c
> @@ -2763,7 +2763,6 @@ void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info)
>  	struct btrfs_block_group *bg;
>  	struct list_head *bg_list;
>  	u64 alloc_flags;
> -	bool first = true;
>  	bool did_chunk_alloc = false;
>  	int index;
>  	int ret;
> @@ -2784,13 +2783,9 @@ void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info)
>  again:
>  	bg_list = &space_info->block_groups[index];
>  	list_for_each_entry(bg, bg_list, list) {
> -		if (bg->alloc_offset != 0)
> -			continue;
>  
> -		if (first) {
> -			first = false;
> +		if (bg->alloc_offset != 0)
>  			continue;
> -		}
>  
>  		if (space_info == data_sinfo) {
>  			/* Migrate the block group to the data relocation space_info. */
> @@ -2849,7 +2844,6 @@ void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info)
>  		 * We allocated a new block group in the data relocation space_info. We
>  		 * can take that one.
>  		 */
> -		first = false;
>  		did_chunk_alloc = true;
>  		goto again;
>  	}
> -- 
> 2.54.0
> 

Re: [PATCH v3 3/5] btrfs: zoned: always set data_relocation_bg

[Naohiro Aota]

On Fri May 22, 2026 at 6:02 PM JST, Johannes Thumshirn wrote:
> When searching for a data relocation block-group on mount,
> btrfs_zoned_reserve_data_reloc_bg() is looking for the first empty DATA
> block-group. But it first checks if the block-group is empty and if yes
> continues the search, and then checks if it is the first DATA block-group.
>
> There is actually no point in looking for the second empty DATA block
> group as new DATA allocations will just allocate a new chunk for it. Pick
> the first DATA block-group without any allocations done and set it as
> relocation block-group.

I think we can add context here.

At first, the commit 694ce5e143d6 ("btrfs: zoned: reserve data_reloc
block group on mount") introduced the functionality. At that time, we
took second unused (used == 0) block group, as the first one might be
a block group used for normal data.

Later, commit daa0fde32235 ("btrfs: zoned: fix data relocation block
group reservation") switched to look for an empty block group
(alloc_offset == 0). At this point, there is no reason taking the second
one anymore. So, this commit is fixing an issue in commit daa0fde32235.

>
> Singed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
> ---
>  fs/btrfs/zoned.c | 8 +-------
>  1 file changed, 1 insertion(+), 7 deletions(-)
>
> diff --git a/fs/btrfs/zoned.c b/fs/btrfs/zoned.c
> index 16dd87aa06f2..a4d2fb774f72 100644
> --- a/fs/btrfs/zoned.c
> +++ b/fs/btrfs/zoned.c
> @@ -2763,7 +2763,6 @@ void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info)
>  	struct btrfs_block_group *bg;
>  	struct list_head *bg_list;
>  	u64 alloc_flags;
> -	bool first = true;
>  	bool did_chunk_alloc = false;
>  	int index;
>  	int ret;
> @@ -2784,13 +2783,9 @@ void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info)

Could you please fix the comment here?

	/* Scan the data space_info to find empty block groups. Take the second one. */

Other than that,

Reviewed-by: Naohiro Aota <naohiro.aota@wdc.com>

>  again:
>  	bg_list = &space_info->block_groups[index];
>  	list_for_each_entry(bg, bg_list, list) {
> -		if (bg->alloc_offset != 0)
> -			continue;
>  
> -		if (first) {
> -			first = false;
> +		if (bg->alloc_offset != 0)
>  			continue;
> -		}
>  
>  		if (space_info == data_sinfo) {
>  			/* Migrate the block group to the data relocation space_info. */
> @@ -2849,7 +2844,6 @@ void btrfs_zoned_reserve_data_reloc_bg(struct btrfs_fs_info *fs_info)
>  		 * We allocated a new block group in the data relocation space_info. We
>  		 * can take that one.
>  		 */
> -		first = false;
>  		did_chunk_alloc = true;
>  		goto again;
>  	}

Re: [PATCH v3 0/5] btrfs: zoned: fix deadlock and space reporting issues for zoned filesystems

[Naohiro Aota]

On Fri May 22, 2026 at 6:02 PM JST, Johannes Thumshirn wrote:
> This series fixes premature ENOSPC errors and deadlock issues on zoned BTRFS
> filesystems when running xfstest generic/747, which tests garbage collection
> on zoned block devices using direct and buffered I/O.
>
> There where two issues revealed in the investigation:
>
> Async reclaim deadlock: On zoned filesystems, the flush state machine
> only executes RECLAIM_ZONES and RESET_ZONES in later flush states
> (FLUSH_DELALLOC and beyond). By the time these states are reached,
> ticket waiters can deadlock waiting for space that can only be freed by
> zone reset. The fix adds RECLAIM_ZONES and RESET_ZONES to the first
> async reclaim loop (FLUSH_ALLOC) specifically for zoned filesystems,
> ensuring zone reset happens early enough to free space for pending
> allocation tickets.
>
> Additionally, the series fixes a bug in data relocation block group selection
> where the first block group was incorrectly skipped, and adds a new flush
> state (BTRFS_RESERVE_FLUSH_DATA_RELOCATION) to use priority reclaim for
> zoned data relocation operations.
>
> Using this, generic/747 passes when lowering the fill_percent variable
> to 80% (from the 95%) that XFS uses. I'm still examining if this is due
> to XFS not having metadata on zoned or other statfs space estimation
> bugs or something else.

For the series,

Reviewed-by: Naohiro Aota <naohiro.aota@wdc.com>

Re: [PATCH v3 0/5] btrfs: zoned: fix deadlock and space reporting issues for zoned filesystems

[David Sterba]

On Fri, May 22, 2026 at 11:02:42AM +0200, Johannes Thumshirn wrote:
> This series fixes premature ENOSPC errors and deadlock issues on zoned BTRFS
> filesystems when running xfstest generic/747, which tests garbage collection
> on zoned block devices using direct and buffered I/O.
> 
> There where two issues revealed in the investigation:
> 
> Async reclaim deadlock: On zoned filesystems, the flush state machine
> only executes RECLAIM_ZONES and RESET_ZONES in later flush states
> (FLUSH_DELALLOC and beyond). By the time these states are reached,
> ticket waiters can deadlock waiting for space that can only be freed by
> zone reset. The fix adds RECLAIM_ZONES and RESET_ZONES to the first
> async reclaim loop (FLUSH_ALLOC) specifically for zoned filesystems,
> ensuring zone reset happens early enough to free space for pending
> allocation tickets.
> 
> Additionally, the series fixes a bug in data relocation block group selection
> where the first block group was incorrectly skipped, and adds a new flush
> state (BTRFS_RESERVE_FLUSH_DATA_RELOCATION) to use priority reclaim for
> zoned data relocation operations.
> 
> Using this, generic/747 passes when lowering the fill_percent variable
> to 80% (from the 95%) that XFS uses. I'm still examining if this is due
> to XFS not having metadata on zoned or other statfs space estimation
> bugs or something else.
> 
> Changes to v2:
> - Re-do the data-relocation block-group selection algorithm
> 
> Changes to v1:
> - Removed the zone_unusable statfs patch
> - Removed the heavy handed flushing patch
> - Added fixes tags where needed
> - Added more comments
> 
> Johannes Thumshirn (5):
>   btrfs: zoned: document RECLAIM_ZONES flush state
>   btrfs: zoned: decode 'RECLAIM_ZONES' state in tracepoints
>   btrfs: zoned: always set data_relocation_bg
>   btrfs: zoned: don't account data relocation space-info in statfs free
>     space
>   btrfs: zoned: fix deadlock waiting for ticket during data relocation

Added to for-next, with updated changelog in patch 3 and deleted
comment. Thanks.