[PATCH v1 1/3] svcrdma: Fix another Receive buffer leak

[PATCH v1 1/3] svcrdma: Fix another Receive buffer leak

[Chuck Lever]

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>
---
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c |    2 ++
 1 file changed, 2 insertions(+)

diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index d5ec85cb652c..5bb97b5f4606 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -275,6 +275,8 @@ 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;

[PATCH v1 2/3] svcrdma: Remove transport reference counting

[Chuck Lever]

Jason tells me that a ULP cannot rely on getting an ESTABLISHED
and DISCONNECTED event pair for each connection, so transport
reference counting in the CM event handler will never be reliable.

Now that we have ib_drain_qp(), svcrdma should no longer need to
hold transport references while Sends and Receives are posted. So
remove the get/put call sites in the CM event handlers.

This eliminates a significant source of locked memory bus traffic.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_recvfrom.c  |    6 +-----
 net/sunrpc/xprtrdma/svc_rdma_rw.c        |    2 --
 net/sunrpc/xprtrdma/svc_rdma_sendto.c    |    4 ----
 net/sunrpc/xprtrdma/svc_rdma_transport.c |   17 +----------------
 4 files changed, 2 insertions(+), 27 deletions(-)

diff --git a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
index 5bb97b5f4606..c6ea2903c21a 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_recvfrom.c
@@ -257,7 +257,6 @@ static int __svc_rdma_post_recv(struct svcxprt_rdma *rdma,
 {
 	int ret;
 
-	svc_xprt_get(&rdma->sc_xprt);
 	trace_svcrdma_post_recv(ctxt);
 	ret = ib_post_recv(rdma->sc_qp, &ctxt->rc_recv_wr, NULL);
 	if (ret)
@@ -267,7 +266,6 @@ static int __svc_rdma_post_recv(struct svcxprt_rdma *rdma,
 err_post:
 	trace_svcrdma_rq_post_err(rdma, ret);
 	svc_rdma_recv_ctxt_put(rdma, ctxt);
-	svc_xprt_put(&rdma->sc_xprt);
 	return ret;
 }
 
@@ -344,15 +342,13 @@ static void svc_rdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
 	spin_unlock(&rdma->sc_rq_dto_lock);
 	if (!test_bit(RDMAXPRT_CONN_PENDING, &rdma->sc_flags))
 		svc_xprt_enqueue(&rdma->sc_xprt);
-	goto out;
+	return;
 
 flushed:
 post_err:
 	svc_rdma_recv_ctxt_put(rdma, ctxt);
 	set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
 	svc_xprt_enqueue(&rdma->sc_xprt);
-out:
-	svc_xprt_put(&rdma->sc_xprt);
 }
 
 /**
diff --git a/net/sunrpc/xprtrdma/svc_rdma_rw.c b/net/sunrpc/xprtrdma/svc_rdma_rw.c
index c16d10601d65..fe54cbe97a46 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_rw.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_rw.c
@@ -164,7 +164,6 @@ static void svc_rdma_cc_init(struct svcxprt_rdma *rdma,
 {
 	svc_rdma_cc_cid_init(rdma, &cc->cc_cid);
 	cc->cc_rdma = rdma;
-	svc_xprt_get(&rdma->sc_xprt);
 
 	INIT_LIST_HEAD(&cc->cc_rwctxts);
 	cc->cc_sqecount = 0;
@@ -184,7 +183,6 @@ static void svc_rdma_cc_release(struct svc_rdma_chunk_ctxt *cc,
 				    ctxt->rw_nents, dir);
 		svc_rdma_put_rw_ctxt(rdma, ctxt);
 	}
-	svc_xprt_put(&rdma->sc_xprt);
 }
 
 /* State for sending a Write or Reply chunk.
diff --git a/net/sunrpc/xprtrdma/svc_rdma_sendto.c b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
index 73d46e8cdc16..7b94d971feb3 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_sendto.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_sendto.c
@@ -291,8 +291,6 @@ static void svc_rdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
 		set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
 		svc_xprt_enqueue(&rdma->sc_xprt);
 	}
-
-	svc_xprt_put(&rdma->sc_xprt);
 }
 
 /**
@@ -330,7 +328,6 @@ int svc_rdma_send(struct svcxprt_rdma *rdma, struct svc_rdma_send_ctxt *ctxt)
 			continue;
 		}
 
-		svc_xprt_get(&rdma->sc_xprt);
 		trace_svcrdma_post_send(ctxt);
 		ret = ib_post_send(rdma->sc_qp, wr, NULL);
 		if (ret)
@@ -340,7 +337,6 @@ int svc_rdma_send(struct svcxprt_rdma *rdma, struct svc_rdma_send_ctxt *ctxt)
 
 	trace_svcrdma_sq_post_err(rdma, ret);
 	set_bit(XPT_CLOSE, &rdma->sc_xprt.xpt_flags);
-	svc_xprt_put(&rdma->sc_xprt);
 	wake_up(&rdma->sc_send_wait);
 	return ret;
 }
diff --git a/net/sunrpc/xprtrdma/svc_rdma_transport.c b/net/sunrpc/xprtrdma/svc_rdma_transport.c
index 3da7901a49e6..aa60f75c8c1d 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_transport.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_transport.c
@@ -271,7 +271,6 @@ static int rdma_cma_handler(struct rdma_cm_id *cma_id,
 	switch (event->event) {
 	case RDMA_CM_EVENT_ESTABLISHED:
 		/* Accept complete */
-		svc_xprt_get(xprt);
 		dprintk("svcrdma: Connection completed on DTO xprt=%p, "
 			"cm_id=%p\n", xprt, cma_id);
 		clear_bit(RDMAXPRT_CONN_PENDING, &rdma->sc_flags);
@@ -282,7 +281,6 @@ static int rdma_cma_handler(struct rdma_cm_id *cma_id,
 			xprt, cma_id);
 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
 		svc_xprt_enqueue(xprt);
-		svc_xprt_put(xprt);
 		break;
 	case RDMA_CM_EVENT_DEVICE_REMOVAL:
 		dprintk("svcrdma: Device removal cma_id=%p, xprt = %p, "
@@ -290,7 +288,6 @@ static int rdma_cma_handler(struct rdma_cm_id *cma_id,
 			rdma_event_msg(event->event), event->event);
 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
 		svc_xprt_enqueue(xprt);
-		svc_xprt_put(xprt);
 		break;
 	default:
 		dprintk("svcrdma: Unexpected event on DTO endpoint %p, "
@@ -539,24 +536,11 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 	return NULL;
 }
 
-/*
- * When connected, an svc_xprt has at least two references:
- *
- * - A reference held by the cm_id between the ESTABLISHED and
- *   DISCONNECTED events. If the remote peer disconnected first, this
- *   reference could be gone.
- *
- * - A reference held by the svc_recv code that called this function
- *   as part of close processing.
- *
- * At a minimum one references should still be held.
- */
 static void svc_rdma_detach(struct svc_xprt *xprt)
 {
 	struct svcxprt_rdma *rdma =
 		container_of(xprt, struct svcxprt_rdma, sc_xprt);
 
-	/* Disconnect and flush posted WQE */
 	rdma_disconnect(rdma->sc_cm_id);
 }
 
@@ -566,6 +550,7 @@ static void __svc_rdma_free(struct work_struct *work)
 		container_of(work, struct svcxprt_rdma, sc_work);
 	struct svc_xprt *xprt = &rdma->sc_xprt;
 
+	/* This blocks until the Completion Queues are empty */
 	if (rdma->sc_qp && !IS_ERR(rdma->sc_qp))
 		ib_drain_qp(rdma->sc_qp);
 

[PATCH v1 3/3] svcrdma: CM event handler clean up

[Chuck Lever]

Now that there's a core tracepoint that reports these events, there's
no need to maintain dprintk() call sites in each arm of the switch
statements.

We also refresh the documenting comments.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/svc_rdma_transport.c |   56 +++++++++++++-----------------
 1 file changed, 25 insertions(+), 31 deletions(-)

diff --git a/net/sunrpc/xprtrdma/svc_rdma_transport.c b/net/sunrpc/xprtrdma/svc_rdma_transport.c
index aa60f75c8c1d..fb044792b571 100644
--- a/net/sunrpc/xprtrdma/svc_rdma_transport.c
+++ b/net/sunrpc/xprtrdma/svc_rdma_transport.c
@@ -237,62 +237,56 @@ static void handle_connect_req(struct rdma_cm_id *new_cma_id,
 	svc_xprt_enqueue(&listen_xprt->sc_xprt);
 }
 
-/*
- * Handles events generated on the listening endpoint. These events will be
- * either be incoming connect requests or adapter removal  events.
+/**
+ * svc_rdma_listen_handler - Handle CM events generated on a listening endpoint
+ * @cma_id: the server's listener rdma_cm_id
+ * @event: details of the event
+ *
+ * Return values:
+ *     %0: Do not destroy @cma_id
+ *     %1: Destroy @cma_id (never returned here)
+ *
+ * NB: There is never a DEVICE_REMOVAL event for INADDR_ANY listeners.
  */
-static int rdma_listen_handler(struct rdma_cm_id *cma_id,
-			       struct rdma_cm_event *event)
+static int svc_rdma_listen_handler(struct rdma_cm_id *cma_id,
+				   struct rdma_cm_event *event)
 {
 	switch (event->event) {
 	case RDMA_CM_EVENT_CONNECT_REQUEST:
-		dprintk("svcrdma: Connect request on cma_id=%p, xprt = %p, "
-			"event = %s (%d)\n", cma_id, cma_id->context,
-			rdma_event_msg(event->event), event->event);
 		handle_connect_req(cma_id, &event->param.conn);
 		break;
 	default:
-		/* NB: No device removal upcall for INADDR_ANY listeners */
-		dprintk("svcrdma: Unexpected event on listening endpoint %p, "
-			"event = %s (%d)\n", cma_id,
-			rdma_event_msg(event->event), event->event);
 		break;
 	}
-
 	return 0;
 }
 
-static int rdma_cma_handler(struct rdma_cm_id *cma_id,
-			    struct rdma_cm_event *event)
+/**
+ * svc_rdma_cma_handler - Handle CM events on client connections
+ * @cma_id: the server's listener rdma_cm_id
+ * @event: details of the event
+ *
+ * Return values:
+ *     %0: Do not destroy @cma_id
+ *     %1: Destroy @cma_id (never returned here)
+ */
+static int svc_rdma_cma_handler(struct rdma_cm_id *cma_id,
+				struct rdma_cm_event *event)
 {
 	struct svcxprt_rdma *rdma = cma_id->context;
 	struct svc_xprt *xprt = &rdma->sc_xprt;
 
 	switch (event->event) {
 	case RDMA_CM_EVENT_ESTABLISHED:
-		/* Accept complete */
-		dprintk("svcrdma: Connection completed on DTO xprt=%p, "
-			"cm_id=%p\n", xprt, cma_id);
 		clear_bit(RDMAXPRT_CONN_PENDING, &rdma->sc_flags);
 		svc_xprt_enqueue(xprt);
 		break;
 	case RDMA_CM_EVENT_DISCONNECTED:
-		dprintk("svcrdma: Disconnect on DTO xprt=%p, cm_id=%p\n",
-			xprt, cma_id);
-		set_bit(XPT_CLOSE, &xprt->xpt_flags);
-		svc_xprt_enqueue(xprt);
-		break;
 	case RDMA_CM_EVENT_DEVICE_REMOVAL:
-		dprintk("svcrdma: Device removal cma_id=%p, xprt = %p, "
-			"event = %s (%d)\n", cma_id, xprt,
-			rdma_event_msg(event->event), event->event);
 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
 		svc_xprt_enqueue(xprt);
 		break;
 	default:
-		dprintk("svcrdma: Unexpected event on DTO endpoint %p, "
-			"event = %s (%d)\n", cma_id,
-			rdma_event_msg(event->event), event->event);
 		break;
 	}
 	return 0;
@@ -318,7 +312,7 @@ static struct svc_xprt *svc_rdma_create(struct svc_serv *serv,
 	set_bit(XPT_LISTENER, &cma_xprt->sc_xprt.xpt_flags);
 	strcpy(cma_xprt->sc_xprt.xpt_remotebuf, "listener");
 
-	listen_id = rdma_create_id(net, rdma_listen_handler, cma_xprt,
+	listen_id = rdma_create_id(net, svc_rdma_listen_handler, cma_xprt,
 				   RDMA_PS_TCP, IB_QPT_RC);
 	if (IS_ERR(listen_id)) {
 		ret = PTR_ERR(listen_id);
@@ -482,7 +476,7 @@ static struct svc_xprt *svc_rdma_accept(struct svc_xprt *xprt)
 		goto errout;
 
 	/* Swap out the handler */
-	newxprt->sc_cm_id->event_handler = rdma_cma_handler;
+	newxprt->sc_cm_id->event_handler = svc_rdma_cma_handler;
 
 	/* Construct RDMA-CM private message */
 	pmsg.cp_magic = rpcrdma_cmp_magic;