All of lore.kernel.org
 help / color / mirror / Atom feed
From: Chuck Lever <chuck.lever@oracle.com>
To: linux-nfs@vger.kernel.org, linux-rdma@vger.kernel.org
Subject: [PATCH v2 7/8] xprtrdma: Fix cwnd update ordering
Date: Fri, 02 Apr 2021 11:30:55 -0400	[thread overview]
Message-ID: <161737745582.2012531.6232768982739827661.stgit@manet.1015granger.net> (raw)
In-Reply-To: <161737723751.2012531.17740046594972271239.stgit@manet.1015granger.net>

After a reconnect, the reply handler is opening the cwnd (and thus
enabling more RPC Calls to be sent) /before/ rpcrdma_post_recvs()
can post enough Receive WRs to receive their replies. This causes an
RNR and the new connection is lost immediately.

The race is most clearly exposed when KASAN and disconnect injection
are enabled. This slows down rpcrdma_rep_create() enough to allow
the send side to post a bunch of RPC Calls before the Receive
completion handler can invoke ib_post_recv().

Fixes: 2ae50ad68cd7 ("xprtrdma: Close window between waking RPC senders and posting Receives")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
 net/sunrpc/xprtrdma/rpc_rdma.c  |    3 ++-
 net/sunrpc/xprtrdma/verbs.c     |   10 +++++-----
 net/sunrpc/xprtrdma/xprt_rdma.h |    2 +-
 3 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index 292f066d006e..21ddd78a8c35 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -1430,9 +1430,10 @@ void rpcrdma_reply_handler(struct rpcrdma_rep *rep)
 		credits = 1;	/* don't deadlock */
 	else if (credits > r_xprt->rx_ep->re_max_requests)
 		credits = r_xprt->rx_ep->re_max_requests;
+	rpcrdma_post_recvs(r_xprt, credits + (buf->rb_bc_srv_max_requests << 1),
+			   false);
 	if (buf->rb_credits != credits)
 		rpcrdma_update_cwnd(r_xprt, credits);
-	rpcrdma_post_recvs(r_xprt, false);
 
 	req = rpcr_to_rdmar(rqst);
 	if (unlikely(req->rl_reply))
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 0ade69501061..5a2871c4561f 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -544,7 +544,7 @@ int rpcrdma_xprt_connect(struct rpcrdma_xprt *r_xprt)
 	 * outstanding Receives.
 	 */
 	rpcrdma_ep_get(ep);
-	rpcrdma_post_recvs(r_xprt, true);
+	rpcrdma_post_recvs(r_xprt, 1, true);
 
 	rc = rdma_connect(ep->re_id, &ep->re_remote_cma);
 	if (rc)
@@ -1395,21 +1395,21 @@ int rpcrdma_post_sends(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req)
 /**
  * rpcrdma_post_recvs - Refill the Receive Queue
  * @r_xprt: controlling transport instance
- * @temp: mark Receive buffers to be deleted after use
+ * @needed: current credit grant
+ * @temp: mark Receive buffers to be deleted after one use
  *
  */
-void rpcrdma_post_recvs(struct rpcrdma_xprt *r_xprt, bool temp)
+void rpcrdma_post_recvs(struct rpcrdma_xprt *r_xprt, int needed, bool temp)
 {
 	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
 	struct rpcrdma_ep *ep = r_xprt->rx_ep;
 	struct ib_recv_wr *wr, *bad_wr;
 	struct rpcrdma_rep *rep;
-	int needed, count, rc;
+	int count, rc;
 
 	rc = 0;
 	count = 0;
 
-	needed = buf->rb_credits + (buf->rb_bc_srv_max_requests << 1);
 	if (likely(ep->re_receive_count > needed))
 		goto out;
 	needed -= ep->re_receive_count;
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index 31404326f29f..2504f67af63e 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -462,7 +462,7 @@ int rpcrdma_xprt_connect(struct rpcrdma_xprt *r_xprt);
 void rpcrdma_xprt_disconnect(struct rpcrdma_xprt *r_xprt);
 
 int rpcrdma_post_sends(struct rpcrdma_xprt *r_xprt, struct rpcrdma_req *req);
-void rpcrdma_post_recvs(struct rpcrdma_xprt *r_xprt, bool temp);
+void rpcrdma_post_recvs(struct rpcrdma_xprt *r_xprt, int needed, bool temp);
 
 /*
  * Buffer calls - xprtrdma/verbs.c



  parent reply	other threads:[~2021-04-02 15:30 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-04-02 15:30 [PATCH v2 0/8] xprtrdma Receive Queue fixes Chuck Lever
2021-04-02 15:30 ` [PATCH v2 1/8] xprtrdma: Avoid Receive Queue wrapping Chuck Lever
2021-04-02 15:30 ` [PATCH v2 2/8] xprtrdma: Do not refresh Receive Queue while it is draining Chuck Lever
2021-04-02 15:30 ` [PATCH v2 3/8] xprtrdma: Put flushed Receives on free list instead of destroying them Chuck Lever
2021-04-02 15:30 ` [PATCH v2 4/8] xprtrdma: Improve locking around rpcrdma_rep destruction Chuck Lever
2021-04-02 15:30 ` [PATCH v2 5/8] xprtrdma: Improve commentary around rpcrdma_reps_unmap() Chuck Lever
2021-04-02 15:30 ` [PATCH v2 6/8] xprtrdma: Improve locking around rpcrdma_rep creation Chuck Lever
2021-04-02 15:30 ` Chuck Lever [this message]
2021-04-02 15:31 ` [PATCH v2 8/8] xprtrdma: Delete rpcrdma_recv_buffer_put() 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=161737745582.2012531.6232768982739827661.stgit@manet.1015granger.net \
    --to=chuck.lever@oracle.com \
    --cc=linux-nfs@vger.kernel.org \
    --cc=linux-rdma@vger.kernel.org \
    --subject='Re: [PATCH v2 7/8] xprtrdma: Fix cwnd update ordering' \
    /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

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.