ocfs2: check if the ocfs2 lock resource has been initialized before calling ocfs2_dlm...
authoralex chen <alex.chen@huawei.com>
Tue, 14 Apr 2015 22:43:49 +0000 (15:43 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 14 Apr 2015 23:48:58 +0000 (16:48 -0700)
commit2f2eca20a09dac0e9d62bf57ce6a0c6ef6cf91e6
tree481e731e8379647344f431c7b72c7ac15f46b984
parent1543306e75ee40662b2c6d37c43c8659f3d6f880
ocfs2: check if the ocfs2 lock resource has been initialized before calling ocfs2_dlm_lock

If ocfs2 lockres has not been initialized before calling ocfs2_dlm_lock,
the lock won't be dropped and then will lead umount hung.  The case is
described below:

ocfs2_mknod
    ocfs2_mknod_locked
        __ocfs2_mknod_locked
            ocfs2_journal_access_di
            Failed because of -ENOMEM or other reasons, the inode lockres
            has not been initialized yet.

    iput(inode)
        ocfs2_evict_inode
            ocfs2_delete_inode
                ocfs2_inode_lock
                    ocfs2_inode_lock_full_nested
                        __ocfs2_cluster_lock
                        Succeeds and allocates a new dlm lockres.
            ocfs2_clear_inode
                ocfs2_open_unlock
                    ocfs2_drop_inode_locks
                        ocfs2_drop_lock
                        Since lockres has not been initialized, the lock
                        can't be dropped and the lockres can't be
                        migrated, thus umount will hang forever.

Signed-off-by: Alex Chen <alex.chen@huawei.com>
Reviewed-by: Joseph Qi <joseph.qi@huawei.com>
Reviewed-by: joyce.xue <xuejiufei@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/ocfs2/dlmglue.c