On Wed, Aug 28, 2024 at 07:14:11PM +0300, Fedor Pchelkin wrote:
The local extent changeset is passed to clear_record_extent_bits() where it may have some additional memory dynamically allocated for ulist. When qgroup is disabled, the memory is leaked because in this case the changeset is not released upon __btrfs_qgroup_release_data() return.
Since the recorded contents of the changeset are not used thereafter, just don't pass it.
Found by Linux Verification Center (linuxtesting.org) with Syzkaller.
Reported-by: syzbot+81670362c283f3dd889c@syzkaller.appspotmail.com Closes: https://lore.kernel.org/lkml/000000000000aa8c0c060ade165e@google.com Fixes: af0e2aab3b70 ("btrfs: qgroup: flush reservations during quota disable") Cc: stable@vger.kernel.org # 6.10+ Signed-off-by: Fedor Pchelkin pchelkin@ispras.ru
This version looks even better, to me. Thanks for the catch and fix!
Reviewed-by: Boris Burkov boris@bur.io
v2: rework the fix as Qu Wenruo suggested - just don't pass unneeded changeset. Update the commit title and description accordingly.
fs/btrfs/qgroup.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c index 5d57a285d59b..f6118c5f3c9f 100644 --- a/fs/btrfs/qgroup.c +++ b/fs/btrfs/qgroup.c @@ -4344,10 +4344,9 @@ static int __btrfs_qgroup_release_data(struct btrfs_inode *inode, int ret; if (btrfs_qgroup_mode(inode->root->fs_info) == BTRFS_QGROUP_MODE_DISABLED) {
extent_changeset_init(&changeset);EXTENT_QGROUP_RESERVED, &changeset);
EXTENT_QGROUP_RESERVED, NULL);/* In release case, we shouldn't have @reserved */ -- 2.39.2