netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Sasha Levin <sashal@kernel.org>
To: linux-kernel@vger.kernel.org, stable@vger.kernel.org
Cc: Chuck Lever <chuck.lever@oracle.com>,
	Eli Dorfman <eli@vastdata.com>,
	Anna Schumaker <Anna.Schumaker@Netapp.com>,
	Sasha Levin <sashal@kernel.org>,
	linux-nfs@vger.kernel.org, netdev@vger.kernel.org
Subject: [PATCH AUTOSEL 5.2 05/63] xprtrdma: Send Queue size grows after a reconnect
Date: Tue,  1 Oct 2019 12:40:27 -0400	[thread overview]
Message-ID: <20191001164125.15398-5-sashal@kernel.org> (raw)
In-Reply-To: <20191001164125.15398-1-sashal@kernel.org>

From: Chuck Lever <chuck.lever@oracle.com>

[ Upstream commit 98ef77d1aaa7a2f4e1b2a721faa084222021fda7 ]

Eli Dorfman reports that after a series of idle disconnects, an
RPC/RDMA transport becomes unusable (rdma_create_qp returns
-ENOMEM). Problem was tracked down to increasing Send Queue size
after each reconnect.

The rdma_create_qp() API does not promise to leave its @qp_init_attr
parameter unaltered. In fact, some drivers do modify one or more of
its fields. Thus our calls to rdma_create_qp must use a fresh copy
of ib_qp_init_attr each time.

This fix is appropriate for kernels dating back to late 2007, though
it will have to be adapted, as the connect code has changed over the
years.

Reported-by: Eli Dorfman <eli@vastdata.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
 net/sunrpc/xprtrdma/verbs.c | 26 ++++++++++++++------------
 1 file changed, 14 insertions(+), 12 deletions(-)

diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 84bb379245406..3f67c395845df 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -607,10 +607,10 @@ void rpcrdma_ep_destroy(struct rpcrdma_xprt *r_xprt)
  * Unlike a normal reconnection, a fresh PD and a new set
  * of MRs and buffers is needed.
  */
-static int
-rpcrdma_ep_recreate_xprt(struct rpcrdma_xprt *r_xprt,
-			 struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
+static int rpcrdma_ep_recreate_xprt(struct rpcrdma_xprt *r_xprt,
+				    struct ib_qp_init_attr *qp_init_attr)
 {
+	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 	int rc, err;
 
 	trace_xprtrdma_reinsert(r_xprt);
@@ -627,7 +627,7 @@ rpcrdma_ep_recreate_xprt(struct rpcrdma_xprt *r_xprt,
 	}
 
 	rc = -ENETUNREACH;
-	err = rdma_create_qp(ia->ri_id, ia->ri_pd, &ep->rep_attr);
+	err = rdma_create_qp(ia->ri_id, ia->ri_pd, qp_init_attr);
 	if (err) {
 		pr_err("rpcrdma: rdma_create_qp returned %d\n", err);
 		goto out3;
@@ -644,16 +644,16 @@ rpcrdma_ep_recreate_xprt(struct rpcrdma_xprt *r_xprt,
 	return rc;
 }
 
-static int
-rpcrdma_ep_reconnect(struct rpcrdma_xprt *r_xprt, struct rpcrdma_ep *ep,
-		     struct rpcrdma_ia *ia)
+static int rpcrdma_ep_reconnect(struct rpcrdma_xprt *r_xprt,
+				struct ib_qp_init_attr *qp_init_attr)
 {
+	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 	struct rdma_cm_id *id, *old;
 	int err, rc;
 
 	trace_xprtrdma_reconnect(r_xprt);
 
-	rpcrdma_ep_disconnect(ep, ia);
+	rpcrdma_ep_disconnect(&r_xprt->rx_ep, ia);
 
 	rc = -EHOSTUNREACH;
 	id = rpcrdma_create_id(r_xprt, ia);
@@ -675,7 +675,7 @@ rpcrdma_ep_reconnect(struct rpcrdma_xprt *r_xprt, struct rpcrdma_ep *ep,
 		goto out_destroy;
 	}
 
-	err = rdma_create_qp(id, ia->ri_pd, &ep->rep_attr);
+	err = rdma_create_qp(id, ia->ri_pd, qp_init_attr);
 	if (err)
 		goto out_destroy;
 
@@ -700,25 +700,27 @@ rpcrdma_ep_connect(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
 	struct rpcrdma_xprt *r_xprt = container_of(ia, struct rpcrdma_xprt,
 						   rx_ia);
 	struct rpc_xprt *xprt = &r_xprt->rx_xprt;
+	struct ib_qp_init_attr qp_init_attr;
 	int rc;
 
 retry:
+	memcpy(&qp_init_attr, &ep->rep_attr, sizeof(qp_init_attr));
 	switch (ep->rep_connected) {
 	case 0:
 		dprintk("RPC:       %s: connecting...\n", __func__);
-		rc = rdma_create_qp(ia->ri_id, ia->ri_pd, &ep->rep_attr);
+		rc = rdma_create_qp(ia->ri_id, ia->ri_pd, &qp_init_attr);
 		if (rc) {
 			rc = -ENETUNREACH;
 			goto out_noupdate;
 		}
 		break;
 	case -ENODEV:
-		rc = rpcrdma_ep_recreate_xprt(r_xprt, ep, ia);
+		rc = rpcrdma_ep_recreate_xprt(r_xprt, &qp_init_attr);
 		if (rc)
 			goto out_noupdate;
 		break;
 	default:
-		rc = rpcrdma_ep_reconnect(r_xprt, ep, ia);
+		rc = rpcrdma_ep_reconnect(r_xprt, &qp_init_attr);
 		if (rc)
 			goto out;
 	}
-- 
2.20.1


       reply	other threads:[~2019-10-01 16:58 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <20191001164125.15398-1-sashal@kernel.org>
2019-10-01 16:40 ` Sasha Levin [this message]
2019-10-01 16:40 ` [PATCH AUTOSEL 5.2 06/63] 9p: Transport error uninitialized Sasha Levin
2019-10-01 16:40 ` [PATCH AUTOSEL 5.2 15/63] SUNRPC: RPC level errors should always set task->tk_rpc_status Sasha Levin
2019-10-01 16:40 ` [PATCH AUTOSEL 5.2 17/63] netfilter: nf_tables: allow lookups in dynamic sets Sasha Levin
2019-10-01 16:40 ` [PATCH AUTOSEL 5.2 21/63] SUNRPC: Don't try to parse incomplete RPC messages Sasha Levin
2019-10-01 16:40 ` [PATCH AUTOSEL 5.2 22/63] net/sched: act_sample: don't push mac header on ip6gre ingress Sasha Levin
2019-10-01 16:40 ` [PATCH AUTOSEL 5.2 24/63] cdc_ncm: fix divide-by-zero caused by invalid wMaxPacketSize Sasha Levin
2019-10-01 16:40 ` [PATCH AUTOSEL 5.2 25/63] usbnet: ignore endpoints with " Sasha Levin
2019-10-01 16:40 ` [PATCH AUTOSEL 5.2 26/63] net/phy: fix DP83865 10 Mbps HDX loopback disable function Sasha Levin
2019-10-01 16:40 ` [PATCH AUTOSEL 5.2 27/63] net_sched: add max len check for TCA_KIND Sasha Levin
2019-10-01 16:40 ` [PATCH AUTOSEL 5.2 34/63] net/mlx5e: Fix traffic duplication in ethtool steering Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 38/63] arcnet: provide a buffer big enough to actually receive packets Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 40/63] ppp: Fix memory leak in ppp_write Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 49/63] bpf: Fix bpf_event_output re-entry issue Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 50/63] macsec: drop skb sk before calling gro_cells_receive Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 53/63] nfp: flower: fix memory leak in nfp_flower_spawn_vnic_reprs Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 54/63] net: phy: micrel: add Asym Pause workaround for KSZ9021 Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 55/63] mlxsw: spectrum_flower: Fail in case user specifies multiple mirror actions Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 56/63] vrf: Do not attempt to create IPv6 mcast rule if IPv6 is disabled Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 57/63] nfp: flower: prevent memory leak in nfp_flower_spawn_phy_reprs Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 58/63] nfp: abm: fix memory leak in nfp_abm_u32_knode_replace Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 61/63] usbnet: sanity checking of packet sizes and device mtu Sasha Levin
2019-10-01 16:41 ` [PATCH AUTOSEL 5.2 62/63] sch_netem: fix a divide by zero in tabledist() Sasha Levin

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=20191001164125.15398-5-sashal@kernel.org \
    --to=sashal@kernel.org \
    --cc=Anna.Schumaker@Netapp.com \
    --cc=chuck.lever@oracle.com \
    --cc=eli@vastdata.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-nfs@vger.kernel.org \
    --cc=netdev@vger.kernel.org \
    --cc=stable@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).