]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
xfs: release the dquot buf outside of qli_lock
authorDarrick J. Wong <djwong@kernel.org>
Tue, 17 Dec 2024 23:00:49 +0000 (15:00 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 8 Feb 2025 09:02:21 +0000 (10:02 +0100)
commiteccc2dfbf7f31dca73ace0a1ab3f6d1a170e7104
treef8b2d3f5d998b3ecbdbb3b8154f2150b6cc5d464
parent2aec9673172fc90bfeb1f3dda9cb5ce3a1f13e6e
xfs: release the dquot buf outside of qli_lock

commit 1aacd3fac248902ea1f7607f2d12b93929a4833b upstream.

Lai Yi reported a lockdep complaint about circular locking:

 Chain exists of:
   &lp->qli_lock --> &bch->bc_lock --> &l->lock

  Possible unsafe locking scenario:

        CPU0                    CPU1
        ----                    ----
   lock(&l->lock);
                                lock(&bch->bc_lock);
                                lock(&l->lock);
   lock(&lp->qli_lock);

I /think/ the problem here is that xfs_dquot_attach_buf during
quotacheck will release the buffer while it's holding the qli_lock.
Because this is a cached buffer, xfs_buf_rele_cached takes b_lock before
decrementing b_hold.  Other threads have taught lockdep that a locking
dependency chain is bp->b_lock -> bch->bc_lock -> l(ru)->lock; and that
another chain is l(ru)->lock -> lp->qli_lock.  Hence we do not want to
take b_lock while holding qli_lock.

Reported-by: syzbot+3126ab3db03db42e7a31@syzkaller.appspotmail.com
Cc: <stable@vger.kernel.org> # v6.13-rc3
Fixes: ca378189fdfa89 ("xfs: convert quotacheck to attach dquot buffers")
Tested-by: syzbot+3126ab3db03db42e7a31@syzkaller.appspotmail.com
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/xfs/xfs_dquot.c