svcrdma: Fix another Receive buffer leak

[ Upstream commit 64d2642 ]

During a connection tear down, the Receive queue is flushed before
the device resources are freed. Typically, all the Receives flush
with IB_WR_FLUSH_ERR.

However, any pending successful Receives flush with IB_WR_SUCCESS,
and the server automatically posts a fresh Receive to replace the
completing one. This happens even after the connection has closed
and the RQ is drained. Receives that are posted after the RQ is
drained appear never to complete, causing a Receive resource leak.
The leaked Receive buffer is left DMA-mapped.

To prevent these late-posted recv_ctxt's from leaking, block new
Receive posting after XPT_CLOSE is set.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sunrpc/xprtrdma/svc_rdma_recvfrom.c

@@ -265,6 +265,8 @@ static int svc_rdma_post_recv(struct svcxprt_rdma *rdma)
 {
 	struct svc_rdma_recv_ctxt *ctxt;
 
+	if (test_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags))
+		return 0;
 	ctxt = svc_rdma_recv_ctxt_get(rdma);
 	if (!ctxt)
 		return -ENOMEM;