linux-nfs.vger.kernel.org archive mirror
 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 6/6] xprtrdma: Avoid Send Queue wrapping
Date: Mon, 05 Apr 2021 12:35:24 -0400	[thread overview]
Message-ID: <161764052395.29855.5602569735350576645.stgit@manet.1015granger.net> (raw)
In-Reply-To: <161764034907.29855.614994107807503843.stgit@manet.1015granger.net>

Send WRs can be signalled or unsignalled. A signalled Send WR
always has a matching Send completion, while a unsignalled Send
has a completion only if the Send WR fails.

xprtrdma has a Send account mechanism that is designed to reduce
the number of signalled Send WRs. This in turn mitigates the
interrupt rate of the underlying device.

RDMA consumers can't leave all Sends unsignaled, however, because
providers rely on Send completions to maintain their Send Queue head
and tail pointers. xprtrdma counts the number of unsignaled Send WRs
that have been posted to ensure that Sends are signalled often
enough to prevent the Send Queue from wrapping.

This mechanism neglected to account for FastReg WRs, which are
posted on the Send Queue but never signalled. As a result, the
Send Queue wrapped on occasion, resulting in duplication completions
of FastReg and LocalInv WRs.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/frwr_ops.c |   17 +++++++++++++++--
 net/sunrpc/xprtrdma/verbs.c    |   16 +---------------
 2 files changed, 16 insertions(+), 17 deletions(-)

diff --git a/net/sunrpc/xprtrdma/frwr_ops.c b/net/sunrpc/xprtrdma/frwr_ops.c
index 951ae20485f3..43a412ea337a 100644
--- a/net/sunrpc/xprtrdma/frwr_ops.c
+++ b/net/sunrpc/xprtrdma/frwr_ops.c
@@ -390,11 +390,13 @@ static void frwr_cid_init(struct rpcrdma_ep *ep,
  */
 int frwr_send(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 {
+	struct ib_send_wr *post_wr, *send_wr = &req->rl_wr;
 	struct rpcrdma_ep *ep = r_xprt->rx_ep;
-	struct ib_send_wr *post_wr;
 	struct rpcrdma_mr *mr;
+	unsigned int num_wrs;
 
-	post_wr = &req->rl_wr;
+	num_wrs = 1;
+	post_wr = send_wr;
 	list_for_each_entry(mr, &req->rl_registered, mr_list) {
 		struct rpcrdma_frwr *frwr;
 
@@ -409,8 +411,19 @@ int frwr_send(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 		frwr->fr_regwr.wr.send_flags = 0;
 
 		post_wr = &frwr->fr_regwr.wr;
+		++num_wrs;
 	}
 
+	if ((kref_read(&req->rl_kref) > 1) || num_wrs > ep->re_send_count) {
+		send_wr->send_flags |= IB_SEND_SIGNALED;
+		ep->re_send_count = min_t(unsigned int, ep->re_send_batch,
+					  num_wrs - ep->re_send_count);
+	} else {
+		send_wr->send_flags &= ~IB_SEND_SIGNALED;
+		ep->re_send_count -= num_wrs;
+	}
+
+	trace_xprtrdma_post_send(req);
 	return ib_post_send(ep->re_id->qp, post_wr, NULL);
 }
 
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index d4e573eef416..55c45cad2c8a 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -1365,21 +1365,7 @@ static void rpcrdma_regbuf_free(struct rpcrdma_regbuf *rb)
  */
 int rpcrdma_post_sends(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 {
-	struct ib_send_wr *send_wr = &req->rl_wr;
-	struct rpcrdma_ep *ep = r_xprt->rx_ep;
-	int rc;
-
-	if (!ep->re_send_count || kref_read(&req->rl_kref) > 1) {
-		send_wr->send_flags |= IB_SEND_SIGNALED;
-		ep->re_send_count = ep->re_send_batch;
-	} else {
-		send_wr->send_flags &= ~IB_SEND_SIGNALED;
-		--ep->re_send_count;
-	}
-
-	trace_xprtrdma_post_send(req);
-	rc = frwr_send(r_xprt, req);
-	if (rc)
+	if (frwr_send(r_xprt, req))
 		return -ENOTCONN;
 	return 0;
 }



      parent reply	other threads:[~2021-04-05 16:35 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-04-05 16:34 [PATCH v1 0/6] More xprtrdma fixes Chuck Lever
2021-04-05 16:34 ` [PATCH v1 1/6] xprtrdma: rpcrdma_mr_pop() already does list_del_init() Chuck Lever
2021-04-05 16:34 ` [PATCH v1 2/6] xprtrdma: Rename frwr_release_mr() Chuck Lever
2021-04-05 16:35 ` [PATCH v1 3/6] xprtrdma: Clarify use of barrier in frwr_wc_localinv_done() Chuck Lever
2021-04-05 16:35 ` [PATCH v1 4/6] xprtrdma: Do not recycle MR after FastReg/LocalInv flushes Chuck Lever
2021-04-05 16:35 ` [PATCH v1 5/6] xprtrdma: Do not wake RPC consumer on a failed LocalInv Chuck Lever
2021-04-05 16:35 ` Chuck Lever [this message]

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=161764052395.29855.5602569735350576645.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 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).