SUNRPC: Ensure we release the socket write lock if the rpc_task exits early
authorTrond Myklebust <Trond.Myklebust@netapp.com>
Mon, 7 Jan 2013 19:30:46 +0000 (14:30 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 17 Jan 2013 16:50:46 +0000 (08:50 -0800)
commit11d607dd94b7e1e4190f20581643555e620e3366
tree43d6ee6738bb7046f74073d995f0724067a39126
parentd5b319960ab9e0770b6ecae94045b512cd3029b7
SUNRPC: Ensure we release the socket write lock if the rpc_task exits early

commit 87ed50036b866db2ec2ba16b2a7aec4a2b0b7c39 upstream.

If the rpc_task exits while holding the socket write lock before it has
allocated an rpc slot, then the usual mechanism for releasing the write
lock in xprt_release() is defeated.

The problem occurs if the call to xprt_lock_write() initially fails, so
that the rpc_task is put on the xprt->sending wait queue. If the task
exits after being assigned the lock by __xprt_lock_write_func, but
before it has retried the call to xprt_lock_and_alloc_slot(), then
it calls xprt_release() while holding the write lock, but will
immediately exit due to the test for task->tk_rqstp != NULL.

Reported-by: Chris Perl <chris.perl@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/sunrpc/sched.c
net/sunrpc/xprt.c