btrfs: zoned: fix critical section of relocation inode writeback
authorNaohiro Aota <naohiro.aota@wdc.com>
Tue, 7 Jun 2022 07:08:30 +0000 (16:08 +0900)
committerDavid Sterba <dsterba@suse.com>
Tue, 21 Jun 2022 12:46:30 +0000 (14:46 +0200)
commit19ab78ca86981e0e1e73036fb73a508731a7c078
tree533944319310feff0c84f82b14de79ae654165c8
parent343d8a30851c48a4ef0f5ef61d5e9fbd847a6883
btrfs: zoned: fix critical section of relocation inode writeback

We use btrfs_zoned_data_reloc_{lock,unlock} to allow only one process to
write out to the relocation inode. That critical section must include all
the IO submission for the inode. However, flush_write_bio() in
extent_writepages() is out of the critical section, causing an IO
submission outside of the lock. This leads to an out of the order IO
submission and fail the relocation process.

Fix it by extending the critical section.

Fixes: 35156d852762 ("btrfs: zoned: only allow one process to add pages to a relocation inode")
CC: stable@vger.kernel.org # 5.16+
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
fs/btrfs/extent_io.c