linux-nfs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Chuck Lever <chuck.lever@oracle.com>
To: anna.schumaker@netapp.com
Cc: linux-rdma@vger.kernel.org, linux-nfs@vger.kernel.org
Subject: [PATCH v3 08/11] xprtrdma: Serialize credit accounting again
Date: Fri, 04 Mar 2016 11:28:27 -0500	[thread overview]
Message-ID: <20160304162827.13590.44906.stgit@oracle120-ib.cthon.org> (raw)
In-Reply-To: <20160304162447.13590.9524.stgit@oracle120-ib.cthon.org>

Commit fe97b47cd623 ("xprtrdma: Use workqueue to process RPC/RDMA
replies") replaced the reply tasklet with a workqueue that allows
RPC replies to be processed in parallel. Thus the credit values in
RPC-over-RDMA replies can be applied in a different order than in
which the server sent them.

To fix this, revert commit eba8ff660b2d ("xprtrdma: Move credit
update to RPC reply handler"). Reverting is done by hand to
accommodate code changes that have occurred since then.

Fixes: fe97b47cd623 ("xprtrdma: Use workqueue to process . . .")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/rpc_rdma.c  |    9 +--------
 net/sunrpc/xprtrdma/verbs.c     |   27 ++++++++++++++++++++++++++-
 net/sunrpc/xprtrdma/xprt_rdma.h |    1 +
 3 files changed, 28 insertions(+), 9 deletions(-)

diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index 35f8108..888823b 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -797,7 +797,6 @@ rpcrdma_reply_handler(struct rpcrdma_rep *rep)
 	__be32 *iptr;
 	int rdmalen, status, rmerr;
 	unsigned long cwnd;
-	u32 credits;
 
 	dprintk("RPC:       %s: incoming rep %p\n", __func__, rep);
 
@@ -928,15 +927,9 @@ out:
 	if (req->rl_nchunks)
 		r_xprt->rx_ia.ri_ops->ro_unmap_sync(r_xprt, req);
 
-	credits = be32_to_cpu(headerp->rm_credit);
-	if (credits == 0)
-		credits = 1;	/* don't deadlock */
-	else if (credits > r_xprt->rx_buf.rb_max_requests)
-		credits = r_xprt->rx_buf.rb_max_requests;
-
 	spin_lock_bh(&xprt->transport_lock);
 	cwnd = xprt->cwnd;
-	xprt->cwnd = credits << RPC_CWNDSHIFT;
+	xprt->cwnd = atomic_read(&r_xprt->rx_buf.rb_credits) << RPC_CWNDSHIFT;
 	if (xprt->cwnd > cwnd)
 		xprt_release_rqst_cong(rqst->rq_task);
 
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 878f1bf..fc1ef5f 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -190,6 +190,28 @@ rpcrdma_receive_worker(struct work_struct *work)
 	rpcrdma_reply_handler(rep);
 }
 
+/* Perform basic sanity checking to avoid using garbage
+ * to update the credit grant value.
+ */
+static void
+rpcrdma_update_granted_credits(struct rpcrdma_rep *rep)
+{
+	struct rpcrdma_msg *rmsgp = rdmab_to_msg(rep->rr_rdmabuf);
+	struct rpcrdma_buffer *buffer = &rep->rr_rxprt->rx_buf;
+	u32 credits;
+
+	if (rep->rr_len < RPCRDMA_HDRLEN_ERR)
+		return;
+
+	credits = be32_to_cpu(rmsgp->rm_credit);
+	if (credits == 0)
+		credits = 1;	/* don't deadlock */
+	else if (credits > buffer->rb_max_requests)
+		credits = buffer->rb_max_requests;
+
+	atomic_set(&buffer->rb_credits, credits);
+}
+
 static void
 rpcrdma_recvcq_process_wc(struct ib_wc *wc)
 {
@@ -211,7 +233,8 @@ rpcrdma_recvcq_process_wc(struct ib_wc *wc)
 	ib_dma_sync_single_for_cpu(rep->rr_device,
 				   rdmab_addr(rep->rr_rdmabuf),
 				   rep->rr_len, DMA_FROM_DEVICE);
-	prefetch(rdmab_to_msg(rep->rr_rdmabuf));
+
+	rpcrdma_update_granted_credits(rep);
 
 out_schedule:
 	queue_work(rpcrdma_receive_wq, &rep->rr_work);
@@ -330,6 +353,7 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 connected:
 		dprintk("RPC:       %s: %sconnected\n",
 					__func__, connstate > 0 ? "" : "dis");
+		atomic_set(&xprt->rx_buf.rb_credits, 1);
 		ep->rep_connected = connstate;
 		rpcrdma_conn_func(ep);
 		wake_up_all(&ep->rep_connect_wait);
@@ -943,6 +967,7 @@ rpcrdma_buffer_create(struct rpcrdma_xprt *r_xprt)
 	buf->rb_max_requests = r_xprt->rx_data.max_requests;
 	buf->rb_bc_srv_max_requests = 0;
 	spin_lock_init(&buf->rb_lock);
+	atomic_set(&buf->rb_credits, 1);
 
 	rc = ia->ri_ops->ro_init(r_xprt);
 	if (rc)
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index 38fe11b..7bf6f43 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -311,6 +311,7 @@ struct rpcrdma_buffer {
 	struct list_head	rb_send_bufs;
 	struct list_head	rb_recv_bufs;
 	u32			rb_max_requests;
+	atomic_t		rb_credits;	/* most recent credit grant */
 
 	u32			rb_bc_srv_max_requests;
 	spinlock_t		rb_reqslock;	/* protect rb_allreqs */


  parent reply	other threads:[~2016-03-04 16:28 UTC|newest]

Thread overview: 35+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-03-04 16:27 [PATCH v3 00/11] NFS/RDMA client patches for v4.6 Chuck Lever
2016-03-04 16:27 ` [PATCH v3 01/11] xprtrdma: Clean up unused RPCRDMA_INLINE_PAD_THRESH macro Chuck Lever
2016-03-08 17:48   ` Sagi Grimberg
2016-03-04 16:27 ` [PATCH v3 02/11] xprtrdma: Clean up physical_op_map() Chuck Lever
2016-03-08 17:48   ` Sagi Grimberg
2016-03-04 16:27 ` [PATCH v3 03/11] xprtrdma: Clean up dprintk format string containing a newline Chuck Lever
2016-03-08 17:48   ` Sagi Grimberg
2016-03-04 16:27 ` [PATCH v3 04/11] xprtrdma: Segment head and tail XDR buffers on page boundaries Chuck Lever
2016-03-04 16:28 ` [PATCH v3 05/11] xprtrdma: Do not wait if ib_post_send() fails Chuck Lever
2016-03-08 17:53   ` Sagi Grimberg
2016-03-08 18:03     ` Chuck Lever
2016-03-09 11:09       ` Sagi Grimberg
2016-03-09 20:47         ` Chuck Lever
2016-03-09 21:40           ` Anna Schumaker
2016-03-10 10:25           ` Sagi Grimberg
2016-03-10 15:04             ` Steve Wise
2016-03-10 15:05               ` Chuck Lever
2016-03-10 15:31                 ` Steve Wise
2016-03-10 15:35                   ` Chuck Lever
2016-03-10 15:54                     ` Steve Wise
2016-03-10 15:58                       ` Chuck Lever
2016-03-10 16:10                         ` Steve Wise
2016-03-10 16:14                           ` Chuck Lever
2016-03-10 16:21                             ` Steve Wise
2016-03-10 16:40             ` Chuck Lever
2016-03-10 17:01               ` Anna Schumaker
2016-03-04 16:28 ` [PATCH v3 06/11] rpcrdma: Add RPCRDMA_HDRLEN_ERR Chuck Lever
2016-03-08 17:53   ` Sagi Grimberg
2016-03-04 16:28 ` [PATCH v3 07/11] xprtrdma: Properly handle RDMA_ERROR replies Chuck Lever
2016-03-04 16:28 ` Chuck Lever [this message]
2016-03-04 16:28 ` [PATCH v3 09/11] xprtrdma: Use new CQ API for RPC-over-RDMA client receive CQs Chuck Lever
2016-03-08 17:55   ` Sagi Grimberg
2016-03-04 16:28 ` [PATCH v3 10/11] xprtrdma: Use an anonymous union in struct rpcrdma_mw Chuck Lever
2016-03-08 17:55   ` Sagi Grimberg
2016-03-04 16:28 ` [PATCH v3 11/11] xprtrdma: Use new CQ API for RPC-over-RDMA client send CQs Chuck Lever

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=20160304162827.13590.44906.stgit@oracle120-ib.cthon.org \
    --to=chuck.lever@oracle.com \
    --cc=anna.schumaker@netapp.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).