]> git.ipfire.org Git - thirdparty/kernel/linux.git/commit
ext4: guard against EA inode refcount underflow in xattr update
authorAhmet Eray Karadag <eraykrdg1@gmail.com>
Sat, 20 Sep 2025 02:13:43 +0000 (05:13 +0300)
committerTheodore Ts'o <tytso@mit.edu>
Fri, 26 Sep 2025 12:36:08 +0000 (08:36 -0400)
commit57295e835408d8d425bef58da5253465db3d6888
treed57ce0943e8b847df312980fd70343f7d7159b04
parent04a91570ac67760301e5458d65eaf1342ecca314
ext4: guard against EA inode refcount underflow in xattr update

syzkaller found a path where ext4_xattr_inode_update_ref() reads an EA
inode refcount that is already <= 0 and then applies ref_change (often
-1). That lets the refcount underflow and we proceed with a bogus value,
triggering errors like:

  EXT4-fs error: EA inode <n> ref underflow: ref_count=-1 ref_change=-1
  EXT4-fs warning: ea_inode dec ref err=-117

Make the invariant explicit: if the current refcount is non-positive,
treat this as on-disk corruption, emit ext4_error_inode(), and fail the
operation with -EFSCORRUPTED instead of updating the refcount. Delete the
WARN_ONCE() as negative refcounts are now impossible; keep error reporting
in ext4_error_inode().

This prevents the underflow and the follow-on orphan/cleanup churn.

Reported-by: syzbot+0be4f339a8218d2a5bb1@syzkaller.appspotmail.com
Fixes: https://syzbot.org/bug?extid=0be4f339a8218d2a5bb1
Cc: stable@kernel.org
Co-developed-by: Albin Babu Varghese <albinbabuvarghese20@gmail.com>
Signed-off-by: Albin Babu Varghese <albinbabuvarghese20@gmail.com>
Signed-off-by: Ahmet Eray Karadag <eraykrdg1@gmail.com>
Message-ID: <20250920021342.45575-1-eraykrdg1@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
fs/ext4/xattr.c