[PATCH 0/3] btrfs: fixes for races when checking if an inode was logged before

[fdmanana@kernel.org]

From: Filipe Manana <fdmanana@suse.com>

The first patch is actually a new version of a patch sent out before [1],
because holding the inode's log_mutex at inode_logged() can make lockdep
unhappy in situations like when logging conflicting inodes, where we are
already holding the log_mutex of some other inode and could potentially
result in a ABBA deadlock.

The second patch splits part of what the initial version of patch 1 fixed,
but with a different solution that makes the management of an inode's
last_dir_index_offset field simpler.

Patch 3 is completely new.

Details in the change logs.

[1] - https://lore.kernel.org/linux-btrfs/7585d15c0e9c163d5cdf32307014a4e792e62541.1753807163.git.fdmanana@suse.com/

Filipe Manana (3):
  btrfs: fix race between logging inode and checking if it was logged before
  btrfs: fix race between setting last_dir_index_offset and inode logging
  btrfs: avoid load/store tearing races when checking if an inode was logged

 fs/btrfs/btrfs_inode.h |  2 +-
 fs/btrfs/inode.c       |  1 +
 fs/btrfs/tree-log.c    | 78 ++++++++++++++++++++++++++++--------------
 3 files changed, 55 insertions(+), 26 deletions(-)

-- 
2.47.2