From: Chuck Lever <chuck.lever@oracle.com>
To: linux-rdma@vger.kernel.org, linux-nfs@vger.kernel.org
Subject: [PATCH v2 03/10] xprtrdma: unmap all FMRs during transport disconnect
Date: Sat, 08 Nov 2014 20:14:29 -0500 [thread overview]
Message-ID: <20141109011429.8806.34346.stgit@manet.1015granger.net> (raw)
In-Reply-To: <20141109010328.8806.5861.stgit@manet.1015granger.net>
When using RPCRDMA_MTHCAFMR memory registration, after a few
transport disconnect / reconnect cycles, ib_map_phys_fmr() starts to
return EINVAL because the provider has exhausted its map pool.
Make sure that all FMRs are unmapped during transport disconnect,
and that ->send_request remarshals them during an RPC retransmit.
This resets the transport's MRs to ensure that none are leaked
during a disconnect.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
---
net/sunrpc/xprtrdma/transport.c | 2 +-
net/sunrpc/xprtrdma/verbs.c | 42 ++++++++++++++++++++++++++++++++++++++-
2 files changed, 42 insertions(+), 2 deletions(-)
diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index 6a4615d..cfe9a81 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -599,7 +599,7 @@ xprt_rdma_send_request(struct rpc_task *task)
if (req->rl_niovs == 0)
rc = rpcrdma_marshal_req(rqst);
- else if (r_xprt->rx_ia.ri_memreg_strategy == RPCRDMA_FRMR)
+ else if (r_xprt->rx_ia.ri_memreg_strategy != RPCRDMA_ALLPHYSICAL)
rc = rpcrdma_marshal_chunks(rqst, 0);
if (rc < 0)
goto failed_marshal;
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index af45cf3..3c88276 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -62,6 +62,7 @@
#endif
static void rpcrdma_reset_frmrs(struct rpcrdma_ia *);
+static void rpcrdma_reset_fmrs(struct rpcrdma_ia *);
/*
* internal functions
@@ -868,8 +869,19 @@ retry:
rpcrdma_ep_disconnect(ep, ia);
rpcrdma_flush_cqs(ep);
- if (ia->ri_memreg_strategy == RPCRDMA_FRMR)
+ switch (ia->ri_memreg_strategy) {
+ case RPCRDMA_FRMR:
rpcrdma_reset_frmrs(ia);
+ break;
+ case RPCRDMA_MTHCAFMR:
+ rpcrdma_reset_fmrs(ia);
+ break;
+ case RPCRDMA_ALLPHYSICAL:
+ break;
+ default:
+ rc = -EIO;
+ goto out;
+ }
xprt = container_of(ia, struct rpcrdma_xprt, rx_ia);
id = rpcrdma_create_id(xprt, ia,
@@ -1289,6 +1301,34 @@ rpcrdma_buffer_destroy(struct rpcrdma_buffer *buf)
kfree(buf->rb_pool);
}
+/* After a disconnect, unmap all FMRs.
+ *
+ * This is invoked only in the transport connect worker in order
+ * to serialize with rpcrdma_register_fmr_external().
+ */
+static void
+rpcrdma_reset_fmrs(struct rpcrdma_ia *ia)
+{
+ struct rpcrdma_xprt *r_xprt =
+ container_of(ia, struct rpcrdma_xprt, rx_ia);
+ struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
+ struct list_head *pos;
+ struct rpcrdma_mw *r;
+ LIST_HEAD(l);
+ int rc;
+
+ list_for_each(pos, &buf->rb_all) {
+ r = list_entry(pos, struct rpcrdma_mw, mw_all);
+
+ INIT_LIST_HEAD(&l);
+ list_add(&r->r.fmr->list, &l);
+ rc = ib_unmap_fmr(&l);
+ if (rc)
+ dprintk("RPC: %s: ib_unmap_fmr failed %i\n",
+ __func__, rc);
+ }
+}
+
/* After a disconnect, a flushed FAST_REG_MR can leave an FRMR in
* an unusable state. Find FRMRs in this state and dereg / reg
* each. FRMRs that are VALID and attached to an rpcrdma_req are
next prev parent reply other threads:[~2014-11-09 1:14 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-11-09 1:14 [PATCH v2 00/10] NFS/RDMA patches for 3.19 Chuck Lever
2014-11-09 1:14 ` [PATCH v2 01/10] xprtrdma: Return an errno from rpcrdma_register_external() Chuck Lever
2014-11-09 1:14 ` [PATCH v2 02/10] xprtrdma: Cap req_cqinit Chuck Lever
2014-11-09 10:13 ` Sagi Grimberg
2014-11-09 21:43 ` Chuck Lever
2014-11-09 1:14 ` Chuck Lever [this message]
2014-11-09 1:14 ` [PATCH v2 04/10] xprtrdma: Refactor tasklet scheduling Chuck Lever
2014-11-09 1:14 ` [PATCH v2 05/10] xprtrdma: Re-write rpcrdma_flush_cqs() Chuck Lever
2014-11-09 1:14 ` [PATCH v2 06/10] xprtrdma: Enable pad optimization Chuck Lever
2014-11-10 14:36 ` Anna Schumaker
2014-11-10 14:54 ` Chuck Lever
2014-11-10 15:05 ` J. Bruce Fields
2014-11-09 1:15 ` [PATCH v2 07/10] xprtrdma: Display async errors Chuck Lever
2014-11-11 14:30 ` Sagi Grimberg
2014-11-11 16:52 ` Chuck Lever
2014-11-11 18:49 ` Sagi Grimberg
2014-11-11 20:30 ` Or Gerlitz
2014-11-09 1:15 ` [PATCH v2 08/10] SUNRPC: serialize iostats updates Chuck Lever
2014-11-09 1:15 ` [PATCH v2 09/10] NFS: SETCLIENTID XDR buffer sizes are incorrect Chuck Lever
2014-11-10 15:22 ` Anna Schumaker
2014-11-10 16:21 ` Trond Myklebust
2014-11-09 1:15 ` [PATCH v2 10/10] NFS: Clean up nfs4_init_callback() Chuck Lever
2014-11-26 16:25 ` [PATCH v2 00/10] NFS/RDMA patches for 3.19 Anna Schumaker
2014-11-26 16:36 ` 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=20141109011429.8806.34346.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