| Message ID | 6e209698b6956fb5cc9da480ac194a6d79426148.1703220899.git.naohiro.aota@wdc.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | btrfs: zoned: fix lock ordering in btrfs_zone_activate() | expand |
On Fri, Dec 22, 2023 at 01:56:34PM +0900, Naohiro Aota wrote: > The btrfs CI reported a lockdep warning as follows by running generic > generic/129. > > ====================================================== > WARNING: possible circular locking dependency detected > 6.7.0-rc5+ #1 Not tainted > ------------------------------------------------------ > kworker/u5:5/793427 is trying to acquire lock: > ffff88813256d028 (&cache->lock){+.+.}-{2:2}, at: btrfs_zone_finish_one_bg+0x5e/0x130 > but task is already holding lock: > ffff88810a23a318 (&fs_info->zone_active_bgs_lock){+.+.}-{2:2}, at: btrfs_zone_finish_one_bg+0x34/0x130 > which lock already depends on the new lock. > > the existing dependency chain (in reverse order) is: > -> #1 (&fs_info->zone_active_bgs_lock){+.+.}-{2:2}: > ... > -> #0 (&cache->lock){+.+.}-{2:2}: > ... > > This is because we take fs_info->zone_active_bgs_lock after a block_group's > lock in btrfs_zone_activate() while doing the opposite other place. > > Fix the issue by expanding the fs_info->zone_active_bgs_lock's critical > section and taking it before a block_group's lock. > > CC: stable@vger.kernel.org # 6.6 > Fixes: a7e1ac7bdc5a ("btrfs: zoned: reserve zones for an active metadata/system block group") > Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com> Added to misc-next, thanks.
diff --git a/fs/btrfs/zoned.c b/fs/btrfs/zoned.c index 12066afc235c..ac9bbe0c4ffe 100644 --- a/fs/btrfs/zoned.c +++ b/fs/btrfs/zoned.c @@ -2072,6 +2072,7 @@ bool btrfs_zone_activate(struct btrfs_block_group *block_group) map = block_group->physical_map; + spin_lock(&fs_info->zone_active_bgs_lock); spin_lock(&block_group->lock); if (test_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE, &block_group->runtime_flags)) { ret = true; @@ -2084,7 +2085,6 @@ bool btrfs_zone_activate(struct btrfs_block_group *block_group) goto out_unlock; } - spin_lock(&fs_info->zone_active_bgs_lock); for (i = 0; i < map->num_stripes; i++) { struct btrfs_zoned_device_info *zinfo; int reserved = 0; @@ -2104,20 +2104,17 @@ bool btrfs_zone_activate(struct btrfs_block_group *block_group) */ if (atomic_read(&zinfo->active_zones_left) <= reserved) { ret = false; - spin_unlock(&fs_info->zone_active_bgs_lock); goto out_unlock; } if (!btrfs_dev_set_active_zone(device, physical)) { /* Cannot activate the zone */ ret = false; - spin_unlock(&fs_info->zone_active_bgs_lock); goto out_unlock; } if (!is_data) zinfo->reserved_active_zones--; } - spin_unlock(&fs_info->zone_active_bgs_lock); /* Successfully activated all the zones */ set_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE, &block_group->runtime_flags); @@ -2125,8 +2122,6 @@ bool btrfs_zone_activate(struct btrfs_block_group *block_group) /* For the active block group list */ btrfs_get_block_group(block_group); - - spin_lock(&fs_info->zone_active_bgs_lock); list_add_tail(&block_group->active_bg_list, &fs_info->zone_active_bgs); spin_unlock(&fs_info->zone_active_bgs_lock); @@ -2134,6 +2129,7 @@ bool btrfs_zone_activate(struct btrfs_block_group *block_group) out_unlock: spin_unlock(&block_group->lock); + spin_unlock(&fs_info->zone_active_bgs_lock); return ret; }
The btrfs CI reported a lockdep warning as follows by running generic generic/129. ====================================================== WARNING: possible circular locking dependency detected 6.7.0-rc5+ #1 Not tainted ------------------------------------------------------ kworker/u5:5/793427 is trying to acquire lock: ffff88813256d028 (&cache->lock){+.+.}-{2:2}, at: btrfs_zone_finish_one_bg+0x5e/0x130 but task is already holding lock: ffff88810a23a318 (&fs_info->zone_active_bgs_lock){+.+.}-{2:2}, at: btrfs_zone_finish_one_bg+0x34/0x130 which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (&fs_info->zone_active_bgs_lock){+.+.}-{2:2}: ... -> #0 (&cache->lock){+.+.}-{2:2}: ... This is because we take fs_info->zone_active_bgs_lock after a block_group's lock in btrfs_zone_activate() while doing the opposite other place. Fix the issue by expanding the fs_info->zone_active_bgs_lock's critical section and taking it before a block_group's lock. CC: stable@vger.kernel.org # 6.6 Fixes: a7e1ac7bdc5a ("btrfs: zoned: reserve zones for an active metadata/system block group") Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com> --- fs/btrfs/zoned.c | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-)