]> git.openfabrics.org - ~shefty/rdma-dev.git/commit
Btrfs: fix possible deadlock when opening a seed device
authorLi Zefan <lizf@cn.fujitsu.com>
Wed, 7 Dec 2011 03:38:24 +0000 (11:38 +0800)
committerLi Zefan <lizf@cn.fujitsu.com>
Wed, 11 Jan 2012 02:26:54 +0000 (10:26 +0800)
commitb367e47fb3a70f5d24ebd6faf7d42436d485fb2d
tree5acc05bae14070f2dad3433a1482fe3a7bab0744
parentc7c144db531fda414e532adac56e965ce332e2a5
Btrfs: fix possible deadlock when opening a seed device

The correct lock order is uuid_mutex -> volume_mutex -> chunk_mutex,
but when we mount a filesystem which has backing seed devices, we have
this lock chain:

    open_ctree()
        lock(chunk_mutex);
        read_chunk_tree();
            read_one_dev();
                open_seed_devices();
                    lock(uuid_mutex);

and then we hit a lockdep splat.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
fs/btrfs/disk-io.c
fs/btrfs/volumes.c