aboutsummaryrefslogtreecommitdiffstatshomepage
path: root/rust/pin-init/internal/src
diff options
context:
space:
mode:
authorFilipe Manana <fdmanana@suse.com>2025-08-01 16:39:49 +0100
committerDavid Sterba <dsterba@suse.com>2025-08-13 14:08:44 +0200
commitf022499f24e520706b9a8238746e1cacc37eb4e0 (patch)
tree30ac2ab23f28103b7b169b14212bac43d5524f67 /rust/pin-init/internal/src
parentbtrfs: clear block dirty if btrfs_writepage_cow_fixup() failed (diff)
downloadwireguard-linux-f022499f24e520706b9a8238746e1cacc37eb4e0.tar.xz
wireguard-linux-f022499f24e520706b9a8238746e1cacc37eb4e0.zip
btrfs: do not set mtime/ctime to current time when unlinking for log replay
If we are doing an unlink for log replay, we are updating the directory's mtime and ctime to the current time, and this is incorrect since it should stay with the mtime and ctime that were set when the directory was logged. This is the same as when adding a link to an inode during log replay (with btrfs_add_link()), where we want the mtime and ctime to be the values that were in place when the inode was logged. This was found with generic/547 using LOAD_FACTOR=20 and TIME_FACTOR=20, where due to large log trees we have longer log replay times and fssum could detect a mismatch of the mtime and ctime of a directory. Fix this by skipping the mtime and ctime update at __btrfs_unlink_inode() if we are in log replay context (just like btrfs_add_link()). Reviewed-by: Boris Burkov <boris@bur.io> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com>
Diffstat (limited to 'rust/pin-init/internal/src')
0 files changed, 0 insertions, 0 deletions