All of lore.kernel.org
 help / color / mirror / Atom feed
From: Chuck Lever <chuck.lever@oracle.com>
To: linux-rdma@vger.kernel.org, linux-nfs@vger.kernel.org
Subject: [PATCH v1 08/11] xprtrdma: Disconnect on flushed completion
Date: Fri, 21 Feb 2020 17:00:49 -0500	[thread overview]
Message-ID: <20200221220049.2072.12038.stgit@manet.1015granger.net> (raw)
In-Reply-To: <20200221214906.2072.32572.stgit@manet.1015granger.net>

Completion errors after a disconnect often occur much sooner than a
CM_DISCONNECT event. Use this to try to detect connection loss more
quickly.

Note that other kernel ULPs do take care to disconnect explicitly
when a WR is flushed.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 include/trace/events/rpcrdma.h  |    3 ++-
 net/sunrpc/xprtrdma/frwr_ops.c  |   24 ++++++++++++++++--------
 net/sunrpc/xprtrdma/verbs.c     |   37 ++++++++++++++++++++++++++++---------
 net/sunrpc/xprtrdma/xprt_rdma.h |    1 +
 4 files changed, 47 insertions(+), 18 deletions(-)

diff --git a/include/trace/events/rpcrdma.h b/include/trace/events/rpcrdma.h
index ce2126a90806..18369943da61 100644
--- a/include/trace/events/rpcrdma.h
+++ b/include/trace/events/rpcrdma.h
@@ -109,7 +109,7 @@
 		__assign_str(port, rpcrdma_portstr(r_xprt));
 	),
 
-	TP_printk("peer=[%s]:%s r_xprt=%p: rc=%d connect status=%d",
+	TP_printk("peer=[%s]:%s r_xprt=%p: rc=%d connection status=%d",
 		__get_str(addr), __get_str(port), __entry->r_xprt,
 		__entry->rc, __entry->connect_status
 	)
@@ -409,6 +409,7 @@
 
 DEFINE_CONN_EVENT(connect);
 DEFINE_CONN_EVENT(disconnect);
+DEFINE_CONN_EVENT(flush_dct);
 
 DEFINE_RXPRT_EVENT(xprtrdma_create);
 DEFINE_RXPRT_EVENT(xprtrdma_op_destroy);
diff --git a/net/sunrpc/xprtrdma/frwr_ops.c b/net/sunrpc/xprtrdma/frwr_ops.c
index 1f34aa49679c..69d5910f04a0 100644
--- a/net/sunrpc/xprtrdma/frwr_ops.c
+++ b/net/sunrpc/xprtrdma/frwr_ops.c
@@ -358,8 +358,8 @@ struct rpcrdma_mr_seg *frwr_map(struct rpcrdma_xprt *r_xprt,
 
 /**
  * frwr_wc_fastreg - Invoked by RDMA provider for a flushed FastReg WC
- * @cq:	completion queue (ignored)
- * @wc:	completed WR
+ * @cq: completion queue
+ * @wc: WCE for a completed FastReg WR
  *
  */
 static void frwr_wc_fastreg(struct ib_cq *cq, struct ib_wc *wc)
@@ -371,6 +371,8 @@ static void frwr_wc_fastreg(struct ib_cq *cq, struct ib_wc *wc)
 	/* WARNING: Only wr_cqe and status are reliable at this point */
 	trace_xprtrdma_wc_fastreg(wc, frwr);
 	/* The MR will get recycled when the associated req is retransmitted */
+
+	rpcrdma_flush_disconnect(cq, wc);
 }
 
 /**
@@ -441,8 +443,8 @@ static void __frwr_release_mr(struct ib_wc *wc, struct rpcrdma_mr *mr)
 
 /**
  * frwr_wc_localinv - Invoked by RDMA provider for a LOCAL_INV WC
- * @cq:	completion queue (ignored)
- * @wc:	completed WR
+ * @cq: completion queue
+ * @wc: WCE for a completed LocalInv WR
  *
  */
 static void frwr_wc_localinv(struct ib_cq *cq, struct ib_wc *wc)
@@ -455,12 +457,14 @@ static void frwr_wc_localinv(struct ib_cq *cq, struct ib_wc *wc)
 	/* WARNING: Only wr_cqe and status are reliable at this point */
 	trace_xprtrdma_wc_li(wc, frwr);
 	__frwr_release_mr(wc, mr);
+
+	rpcrdma_flush_disconnect(cq, wc);
 }
 
 /**
  * frwr_wc_localinv_wake - Invoked by RDMA provider for a LOCAL_INV WC
- * @cq:	completion queue (ignored)
- * @wc:	completed WR
+ * @cq: completion queue
+ * @wc: WCE for a completed LocalInv WR
  *
  * Awaken anyone waiting for an MR to finish being fenced.
  */
@@ -475,6 +479,8 @@ static void frwr_wc_localinv_wake(struct ib_cq *cq, struct ib_wc *wc)
 	trace_xprtrdma_wc_li_wake(wc, frwr);
 	__frwr_release_mr(wc, mr);
 	complete(&frwr->fr_linv_done);
+
+	rpcrdma_flush_disconnect(cq, wc);
 }
 
 /**
@@ -562,8 +568,8 @@ void frwr_unmap_sync(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 
 /**
  * frwr_wc_localinv_done - Invoked by RDMA provider for a signaled LOCAL_INV WC
- * @cq:	completion queue (ignored)
- * @wc:	completed WR
+ * @cq:	completion queue
+ * @wc:	WCE for a completed LocalInv WR
  *
  */
 static void frwr_wc_localinv_done(struct ib_cq *cq, struct ib_wc *wc)
@@ -581,6 +587,8 @@ static void frwr_wc_localinv_done(struct ib_cq *cq, struct ib_wc *wc)
 	/* Ensure @rep is generated before __frwr_release_mr */
 	smp_rmb();
 	rpcrdma_complete_rqst(rep);
+
+	rpcrdma_flush_disconnect(cq, wc);
 }
 
 /**
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index a7f46bbbf017..dfe680e3234a 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -129,13 +129,31 @@ static void rpcrdma_xprt_drain(struct rpcrdma_xprt *r_xprt)
 }
 
 /**
+ * rpcrdma_flush_disconnect - Disconnect on flushed completion
+ * @cq: completion queue
+ * @wc: work completion entry
+ *
+ * Must be called in process context.
+ */
+void rpcrdma_flush_disconnect(struct ib_cq *cq, struct ib_wc *wc)
+{
+	struct rpcrdma_xprt *r_xprt = cq->cq_context;
+	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
+
+	if (wc->status != IB_WC_SUCCESS && r_xprt->rx_ep.rep_connected == 1) {
+		r_xprt->rx_ep.rep_connected = -ECONNABORTED;
+		trace_xprtrdma_flush_dct(r_xprt, wc->status);
+		xprt_force_disconnect(xprt);
+	}
+}
+
+/**
  * rpcrdma_wc_send - Invoked by RDMA provider for each polled Send WC
  * @cq:	completion queue
- * @wc:	completed WR
+ * @wc:	WCE for a completed Send WR
  *
  */
-static void
-rpcrdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
+static void rpcrdma_wc_send(struct ib_cq *cq, struct ib_wc *wc)
 {
 	struct ib_cqe *cqe = wc->wr_cqe;
 	struct rpcrdma_sendctx *sc =
@@ -144,21 +162,21 @@ static void rpcrdma_xprt_drain(struct rpcrdma_xprt *r_xprt)
 	/* WARNING: Only wr_cqe and status are reliable at this point */
 	trace_xprtrdma_wc_send(sc, wc);
 	rpcrdma_sendctx_put_locked((struct rpcrdma_xprt *)cq->cq_context, sc);
+	rpcrdma_flush_disconnect(cq, wc);
 }
 
 /**
  * rpcrdma_wc_receive - Invoked by RDMA provider for each polled Receive WC
- * @cq:	completion queue (ignored)
- * @wc:	completed WR
+ * @cq:	completion queue
+ * @wc:	WCE for a completed Receive WR
  *
  */
-static void
-rpcrdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
+static void rpcrdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
 {
 	struct ib_cqe *cqe = wc->wr_cqe;
 	struct rpcrdma_rep *rep = container_of(cqe, struct rpcrdma_rep,
 					       rr_cqe);
-	struct rpcrdma_xprt *r_xprt = rep->rr_rxprt;
+	struct rpcrdma_xprt *r_xprt = cq->cq_context;
 
 	/* WARNING: Only wr_cqe and status are reliable at this point */
 	trace_xprtrdma_wc_receive(wc);
@@ -179,6 +197,7 @@ static void rpcrdma_xprt_drain(struct rpcrdma_xprt *r_xprt)
 	return;
 
 out_flushed:
+	rpcrdma_flush_disconnect(cq, wc);
 	rpcrdma_rep_destroy(rep);
 }
 
@@ -395,7 +414,7 @@ static int rpcrdma_ep_create(struct rpcrdma_xprt *r_xprt)
 		goto out_destroy;
 	}
 
-	ep->rep_attr.recv_cq = ib_alloc_cq_any(id->device, NULL,
+	ep->rep_attr.recv_cq = ib_alloc_cq_any(id->device, r_xprt,
 					       ep->rep_attr.cap.max_recv_wr,
 					       IB_POLL_WORKQUEUE);
 	if (IS_ERR(ep->rep_attr.recv_cq)) {
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index d2a0f125f7a8..8a3ac9d7ee81 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -452,6 +452,7 @@ struct rpcrdma_xprt {
 /*
  * Endpoint calls - xprtrdma/verbs.c
  */
+void rpcrdma_flush_disconnect(struct ib_cq *cq, struct ib_wc *wc);
 int rpcrdma_xprt_connect(struct rpcrdma_xprt *r_xprt);
 void rpcrdma_xprt_disconnect(struct rpcrdma_xprt *r_xprt);
 


  parent reply	other threads:[~2020-02-21 22:00 UTC|newest]

Thread overview: 22+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-02-21 22:00 [PATCH v1 00/11] NFS/RDMA client side connection overhaul Chuck Lever
2020-02-21 22:00 ` [PATCH v1 01/11] xprtrdma: Invoke rpcrdma_ep_create() in the connect worker Chuck Lever
2020-02-21 22:00 ` [PATCH v1 02/11] xprtrdma: Refactor frwr_init_mr() Chuck Lever
2020-02-21 22:00 ` [PATCH v1 03/11] xprtrdma: Clean up the post_send path Chuck Lever
2020-02-21 22:00 ` [PATCH v1 04/11] xprtrdma: Refactor rpcrdma_ep_connect() and rpcrdma_ep_disconnect() Chuck Lever
2020-02-21 22:00 ` [PATCH v1 05/11] xprtrdma: Allocate Protection Domain in rpcrdma_ep_create() Chuck Lever
2020-03-01 18:11   ` Tom Talpey
2020-03-01 18:29     ` Chuck Lever
2020-03-01 18:38       ` Tom Talpey
2020-02-21 22:00 ` [PATCH v1 06/11] xprtrdma: Invoke rpcrdma_ia_open in the connect worker Chuck Lever
2020-02-21 22:00 ` [PATCH v1 07/11] xprtrdma: Remove rpcrdma_ia::ri_flags Chuck Lever
2020-02-21 22:00 ` Chuck Lever [this message]
2020-02-21 22:00 ` [PATCH v1 09/11] xprtrdma: Merge struct rpcrdma_ia into struct rpcrdma_ep Chuck Lever
2020-02-21 22:01 ` [PATCH v1 10/11] xprtrdma: Extract sockaddr from struct rdma_cm_id Chuck Lever
2020-02-24 16:15   ` Anna Schumaker
2020-02-24 16:18     ` Chuck Lever
2020-02-24 16:23       ` Anna Schumaker
2020-02-21 22:01 ` [PATCH v1 11/11] xprtrdma: kmalloc rpcrdma_ep separate from rpcrdma_xprt Chuck Lever
2020-03-01 18:09 ` [PATCH v1 00/11] NFS/RDMA client side connection overhaul Tom Talpey
2020-03-01 18:12   ` Chuck Lever
2020-03-11 15:27 ` Chuck Lever
2020-03-11 17:16   ` Schumaker, Anna

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200221220049.2072.12038.stgit@manet.1015granger.net \
    --to=chuck.lever@oracle.com \
    --cc=linux-nfs@vger.kernel.org \
    --cc=linux-rdma@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.