]> git.baikalelectronics.ru Git - kernel.git/commit
btrfs: zoned: use dedicated lock for data relocation
authorNaohiro Aota <naohiro.aota@wdc.com>
Mon, 18 Apr 2022 07:15:03 +0000 (16:15 +0900)
committerDavid Sterba <dsterba@suse.com>
Thu, 21 Apr 2022 14:06:24 +0000 (16:06 +0200)
commit128f0ff6b18a13286766d99445c967b6c1ce4f1c
tree905a717629401721cac29243b43bd23f324d122a
parent2641111f5dc6dd21a6c8fd6f8cc6bc7656b60100
btrfs: zoned: use dedicated lock for data relocation

Currently, we use btrfs_inode_{lock,unlock}() to grant an exclusive
writeback of the relocation data inode in
btrfs_zoned_data_reloc_{lock,unlock}(). However, that can cause a deadlock
in the following path.

Thread A takes btrfs_inode_lock() and waits for metadata reservation by
e.g, waiting for writeback:

prealloc_file_extent_cluster()
  - btrfs_inode_lock(&inode->vfs_inode, 0);
  - btrfs_prealloc_file_range()
  ...
    - btrfs_replace_file_extents()
      - btrfs_start_transaction
      ...
        - btrfs_reserve_metadata_bytes()

Thread B (e.g, doing a writeback work) needs to wait for the inode lock to
continue writeback process:

do_writepages
  - btrfs_writepages
    - extent_writpages
      - btrfs_zoned_data_reloc_lock(BTRFS_I(inode));
        - btrfs_inode_lock()

The deadlock is caused by relying on the vfs_inode's lock. By using it, we
introduced unnecessary exclusion of writeback and
btrfs_prealloc_file_range(). Also, the lock at this point is useless as we
don't have any dirty pages in the inode yet.

Introduce fs_info->zoned_data_reloc_io_lock and use it for the exclusive
writeback.

Fixes: 5ec7c05c9358 ("btrfs: zoned: only allow one process to add pages to a relocation inode")
CC: stable@vger.kernel.org # 5.16.x: fbe485f95555: btrfs: zoned: encapsulate inode locking for zoned relocation
CC: stable@vger.kernel.org # 5.16.x
CC: stable@vger.kernel.org # 5.17
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
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/ctree.h
fs/btrfs/disk-io.c
fs/btrfs/zoned.h