From 62b2417e84ba0734d3f4f95a17e6f5f0be54d75a Mon Sep 17 00:00:00 2001
From: NeilBrown <neilb@suse.com>
Date: Mon, 10 Apr 2017 12:19:40 +1000
Subject: sunrpc: don't check for failure from mempool_alloc()

When mempool_alloc() is allowed to sleep (GFP_NOIO allows
sleeping) it cannot fail.
So rpc_alloc_task() cannot fail, so rpc_new_task doesn't need
to test for failure.
Consequently rpc_new_task() cannot fail, so the callers
don't need to test.

Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
---
 net/sunrpc/clnt.c  | 8 --------
 net/sunrpc/sched.c | 5 -----
 2 files changed, 13 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 52da3ce54bb5..b5cb921775a0 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -1042,8 +1042,6 @@ struct rpc_task *rpc_run_task(const struct rpc_task_setup *task_setup_data)
 	struct rpc_task *task;
 
 	task = rpc_new_task(task_setup_data);
-	if (IS_ERR(task))
-		goto out;
 
 	rpc_task_set_client(task, task_setup_data->rpc_client);
 	rpc_task_set_rpc_message(task, task_setup_data->rpc_message);
@@ -1053,7 +1051,6 @@ struct rpc_task *rpc_run_task(const struct rpc_task_setup *task_setup_data)
 
 	atomic_inc(&task->tk_count);
 	rpc_execute(task);
-out:
 	return task;
 }
 EXPORT_SYMBOL_GPL(rpc_run_task);
@@ -1140,10 +1137,6 @@ struct rpc_task *rpc_run_bc_task(struct rpc_rqst *req)
 	 * Create an rpc_task to send the data
 	 */
 	task = rpc_new_task(&task_setup_data);
-	if (IS_ERR(task)) {
-		xprt_free_bc_request(req);
-		goto out;
-	}
 	task->tk_rqstp = req;
 
 	/*
@@ -1158,7 +1151,6 @@ struct rpc_task *rpc_run_bc_task(struct rpc_rqst *req)
 	WARN_ON_ONCE(atomic_read(&task->tk_count) != 2);
 	rpc_execute(task);
 
-out:
 	dprintk("RPC: rpc_run_bc_task: task= %p\n", task);
 	return task;
 }
diff --git a/net/sunrpc/sched.c b/net/sunrpc/sched.c
index 5db68b371db2..0cc83839c13c 100644
--- a/net/sunrpc/sched.c
+++ b/net/sunrpc/sched.c
@@ -965,11 +965,6 @@ struct rpc_task *rpc_new_task(const struct rpc_task_setup *setup_data)
 
 	if (task == NULL) {
 		task = rpc_alloc_task();
-		if (task == NULL) {
-			rpc_release_calldata(setup_data->callback_ops,
-					setup_data->callback_data);
-			return ERR_PTR(-ENOMEM);
-		}
 		flags = RPC_TASK_DYNAMIC;
 	}
 
-- 
cgit v1.2.3


From 9378b274e1eb6925db315e345f48850d2d5d9789 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:22:29 -0400
Subject: xprtrdma: Cancel refresh worker during buffer shutdown

Trying to create MRs while the transport is being torn down can
cause a crash.

Fixes: e2ac236c0b65 ("xprtrdma: Allocate MRs on demand")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/verbs.c | 1 +
 1 file changed, 1 insertion(+)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 3b332b395045..2e4fc1e46947 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -1037,6 +1037,7 @@ void
 rpcrdma_buffer_destroy(struct rpcrdma_buffer *buf)
 {
 	cancel_delayed_work_sync(&buf->rb_recovery_worker);
+	cancel_delayed_work_sync(&buf->rb_refresh_worker);
 
 	while (!list_empty(&buf->rb_recv_bufs)) {
 		struct rpcrdma_rep *rep;
-- 
cgit v1.2.3


From e2a4f4fbefc5e5b7b4435f73711b7be94f780584 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:22:38 -0400
Subject: sunrpc: Export xprt_force_disconnect()

xprt_force_disconnect() is already invoked from the socket
transport. I want to invoke xprt_force_disconnect() from the
RPC-over-RDMA transport, which is a separate module from sunrpc.ko.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprt.c | 1 +
 1 file changed, 1 insertion(+)

(limited to 'net')

diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index b530a2852ba8..3e63c5e97ebe 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -651,6 +651,7 @@ void xprt_force_disconnect(struct rpc_xprt *xprt)
 	xprt_wake_pending_tasks(xprt, -EAGAIN);
 	spin_unlock_bh(&xprt->transport_lock);
 }
+EXPORT_SYMBOL_GPL(xprt_force_disconnect);
 
 /**
  * xprt_conditional_disconnect - force a transport to disconnect
-- 
cgit v1.2.3


From 33849792cbcdae2b04819cfb09fe3dca0a84a11e Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:22:46 -0400
Subject: xprtrdma: Detect unreachable NFS/RDMA servers more reliably

Current NFS clients rely on connection loss to determine when to
retransmit. In particular, for protocols like NFSv4, clients no
longer rely on RPC timeouts to drive retransmission: NFSv4 servers
are required to terminate a connection when they need a client to
retransmit pending RPCs.

When a server is no longer reachable, either because it has crashed
or because the network path has broken, the server cannot actively
terminate a connection. Thus NFS clients depend on transport-level
keepalive to determine when a connection must be replaced and
pending RPCs retransmitted.

However, RDMA RC connections do not have a native keepalive
mechanism. If an NFS/RDMA server crashes after a client has sent
RPCs successfully (an RC ACK has been received for all OTW RDMA
requests), there is no way for the client to know the connection is
moribund.

In addition, new RDMA requests are subject to the RPC-over-RDMA
credit limit. If the client has consumed all granted credits with
NFS traffic, it is not allowed to send another RDMA request until
the server replies. Thus it has no way to send a true keepalive when
the workload has already consumed all credits with pending RPCs.

To address this, forcibly disconnect a transport when an RPC times
out. This prevents moribund connections from stopping the
detection of failover or other configuration changes on the server.

Note that even if the connection is still good, retransmitting
any RPC will trigger a disconnect thanks to this logic in
xprt_rdma_send_request:

	/* Must suppress retransmit to maintain credits */
	if (req->rl_connect_cookie == xprt->connect_cookie)
		goto drop_connection;
	req->rl_connect_cookie = xprt->connect_cookie;

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/transport.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index c717f5410776..acf5d81f4d6e 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -484,6 +484,27 @@ xprt_rdma_set_port(struct rpc_xprt *xprt, u16 port)
 	dprintk("RPC:       %s: %u\n", __func__, port);
 }
 
+/**
+ * xprt_rdma_timer - invoked when an RPC times out
+ * @xprt: controlling RPC transport
+ * @task: RPC task that timed out
+ *
+ * Invoked when the transport is still connected, but an RPC
+ * retransmit timeout occurs.
+ *
+ * Since RDMA connections don't have a keep-alive, forcibly
+ * disconnect and retry to connect. This drives full
+ * detection of the network path, and retransmissions of
+ * all pending RPCs.
+ */
+static void
+xprt_rdma_timer(struct rpc_xprt *xprt, struct rpc_task *task)
+{
+	dprintk("RPC: %5u %s: xprt = %p\n", task->tk_pid, __func__, xprt);
+
+	xprt_force_disconnect(xprt);
+}
+
 static void
 xprt_rdma_connect(struct rpc_xprt *xprt, struct rpc_task *task)
 {
@@ -776,6 +797,7 @@ static struct rpc_xprt_ops xprt_rdma_procs = {
 	.alloc_slot		= xprt_alloc_slot,
 	.release_request	= xprt_release_rqst_cong,       /* ditto */
 	.set_retrans_timeout	= xprt_set_retrans_timeout_def, /* ditto */
+	.timer			= xprt_rdma_timer,
 	.rpcbind		= rpcb_getport_async,	/* sunrpc/rpcb_clnt.c */
 	.set_port		= xprt_rdma_set_port,
 	.connect		= xprt_rdma_connect,
-- 
cgit v1.2.3


From fff09594edf5e9b8595a2cefdc07e54b70f81729 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:22:54 -0400
Subject: xprtrdma: Refactor rpcrdma_ia_open()

In order to unload a device driver and reload it, xprtrdma will need
to close a transport's interface adapter, and then call
rpcrdma_ia_open again, possibly finding a different interface
adapter.

Make rpcrdma_ia_open safe to call on the same transport multiple
times.

This is a refactoring change only.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/transport.c |  6 +++---
 net/sunrpc/xprtrdma/verbs.c     | 46 ++++++++++++++++++++---------------------
 net/sunrpc/xprtrdma/xprt_rdma.h |  7 ++++++-
 3 files changed, 32 insertions(+), 27 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index acf5d81f4d6e..83e219d7aba4 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -66,8 +66,8 @@ static unsigned int xprt_rdma_slot_table_entries = RPCRDMA_DEF_SLOT_TABLE;
 unsigned int xprt_rdma_max_inline_read = RPCRDMA_DEF_INLINE;
 static unsigned int xprt_rdma_max_inline_write = RPCRDMA_DEF_INLINE;
 static unsigned int xprt_rdma_inline_write_padding;
-static unsigned int xprt_rdma_memreg_strategy = RPCRDMA_FRMR;
-		int xprt_rdma_pad_optimize = 0;
+unsigned int xprt_rdma_memreg_strategy		= RPCRDMA_FRMR;
+int xprt_rdma_pad_optimize;
 
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
 
@@ -396,7 +396,7 @@ xprt_setup_rdma(struct xprt_create *args)
 
 	new_xprt = rpcx_to_rdmax(xprt);
 
-	rc = rpcrdma_ia_open(new_xprt, sap, xprt_rdma_memreg_strategy);
+	rc = rpcrdma_ia_open(new_xprt, sap);
 	if (rc)
 		goto out1;
 
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 2e4fc1e46947..d4880d50d7be 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -413,13 +413,16 @@ out:
  * Exported functions.
  */
 
-/*
- * Open and initialize an Interface Adapter.
- *  o initializes fields of struct rpcrdma_ia, including
- *    interface and provider attributes and protection zone.
+/**
+ * rpcrdma_ia_open - Open and initialize an Interface Adapter.
+ * @xprt: controlling transport
+ * @addr: IP address of remote peer
+ *
+ * Returns 0 on success, negative errno if an appropriate
+ * Interface Adapter could not be found and opened.
  */
 int
-rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg)
+rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr)
 {
 	struct rpcrdma_ia *ia = &xprt->rx_ia;
 	int rc;
@@ -427,7 +430,7 @@ rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg)
 	ia->ri_id = rpcrdma_create_id(xprt, ia, addr);
 	if (IS_ERR(ia->ri_id)) {
 		rc = PTR_ERR(ia->ri_id);
-		goto out1;
+		goto out_err;
 	}
 	ia->ri_device = ia->ri_id->device;
 
@@ -435,10 +438,10 @@ rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg)
 	if (IS_ERR(ia->ri_pd)) {
 		rc = PTR_ERR(ia->ri_pd);
 		pr_err("rpcrdma: ib_alloc_pd() returned %d\n", rc);
-		goto out2;
+		goto out_err;
 	}
 
-	switch (memreg) {
+	switch (xprt_rdma_memreg_strategy) {
 	case RPCRDMA_FRMR:
 		if (frwr_is_supported(ia)) {
 			ia->ri_ops = &rpcrdma_frwr_memreg_ops;
@@ -452,28 +455,23 @@ rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr, int memreg)
 		}
 		/*FALLTHROUGH*/
 	default:
-		pr_err("rpcrdma: Unsupported memory registration mode: %d\n",
-		       memreg);
+		pr_err("rpcrdma: Device %s does not support memreg mode %d\n",
+		       ia->ri_device->name, xprt_rdma_memreg_strategy);
 		rc = -EINVAL;
-		goto out3;
+		goto out_err;
 	}
 
 	return 0;
 
-out3:
-	ib_dealloc_pd(ia->ri_pd);
-	ia->ri_pd = NULL;
-out2:
-	rpcrdma_destroy_id(ia->ri_id);
-	ia->ri_id = NULL;
-out1:
+out_err:
+	rpcrdma_ia_close(ia);
 	return rc;
 }
 
-/*
- * Clean up/close an IA.
- *   o if event handles and PD have been initialized, free them.
- *   o close the IA
+/**
+ * rpcrdma_ia_close - Clean up/close an IA.
+ * @ia: interface adapter to close
+ *
  */
 void
 rpcrdma_ia_close(struct rpcrdma_ia *ia)
@@ -483,12 +481,14 @@ rpcrdma_ia_close(struct rpcrdma_ia *ia)
 		if (ia->ri_id->qp)
 			rdma_destroy_qp(ia->ri_id);
 		rpcrdma_destroy_id(ia->ri_id);
-		ia->ri_id = NULL;
 	}
+	ia->ri_id = NULL;
+	ia->ri_device = NULL;
 
 	/* If the pd is still busy, xprtrdma missed freeing a resource */
 	if (ia->ri_pd && !IS_ERR(ia->ri_pd))
 		ib_dealloc_pd(ia->ri_pd);
+	ia->ri_pd = NULL;
 }
 
 /*
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index 171a35116de9..af844fc30bd4 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -497,10 +497,15 @@ struct rpcrdma_xprt {
  * Default is 0, see sysctl entry and rpc_rdma.c rpcrdma_convert_iovs() */
 extern int xprt_rdma_pad_optimize;
 
+/* This setting controls the hunt for a supported memory
+ * registration strategy.
+ */
+extern unsigned int xprt_rdma_memreg_strategy;
+
 /*
  * Interface Adapter calls - xprtrdma/verbs.c
  */
-int rpcrdma_ia_open(struct rpcrdma_xprt *, struct sockaddr *, int);
+int rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr);
 void rpcrdma_ia_close(struct rpcrdma_ia *);
 bool frwr_is_supported(struct rpcrdma_ia *);
 bool fmr_is_supported(struct rpcrdma_ia *);
-- 
cgit v1.2.3


From 91a10c52975a8c89e146a4f740e64cd147ba8e8a Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:23:02 -0400
Subject: xprtrdma: Use same device when mapping or syncing DMA buffers

When the underlying device driver is reloaded, ia->ri_device will be
replaced. All cached copies of that device pointer have to be
updated as well.

Commit 54cbd6b0c6b9 ("xprtrdma: Delay DMA mapping Send and Receive
buffers") added the rg_device field to each regbuf. As part of
handling a device removal, rpcrdma_dma_unmap_regbuf is invoked on
all regbufs for a transport.

Simply calling rpcrdma_dma_map_regbuf for each Receive buffer after
the driver has been reloaded should reinitialize rg_device correctly
for every case except rpcrdma_wc_receive, which still uses
rpcrdma_rep::rr_device.

Ensure the same device that was used to map a Receive buffer is also
used to sync it in rpcrdma_wc_receive by using rg_device there
instead of rr_device.

This is the only use of rr_device, so it can be removed.

The use of regbufs in the send path is also updated, for
completeness.

Fixes: 54cbd6b0c6b9 ("xprtrdma: Delay DMA mapping Send and ... ")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/rpc_rdma.c  |  4 ++--
 net/sunrpc/xprtrdma/verbs.c     | 12 ++++++------
 net/sunrpc/xprtrdma/xprt_rdma.h |  7 ++++++-
 3 files changed, 14 insertions(+), 9 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index a044be2d6ad7..e68131bed3ce 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -494,7 +494,7 @@ rpcrdma_prepare_hdr_sge(struct rpcrdma_ia *ia, struct rpcrdma_req *req,
 	}
 	sge->length = len;
 
-	ib_dma_sync_single_for_device(ia->ri_device, sge->addr,
+	ib_dma_sync_single_for_device(rdmab_device(rb), sge->addr,
 				      sge->length, DMA_TO_DEVICE);
 	req->rl_send_wr.num_sge++;
 	return true;
@@ -523,7 +523,7 @@ rpcrdma_prepare_msg_sges(struct rpcrdma_ia *ia, struct rpcrdma_req *req,
 	sge[sge_no].addr = rdmab_addr(rb);
 	sge[sge_no].length = xdr->head[0].iov_len;
 	sge[sge_no].lkey = rdmab_lkey(rb);
-	ib_dma_sync_single_for_device(device, sge[sge_no].addr,
+	ib_dma_sync_single_for_device(rdmab_device(rb), sge[sge_no].addr,
 				      sge[sge_no].length, DMA_TO_DEVICE);
 
 	/* If there is a Read chunk, the page list is being handled
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index d4880d50d7be..c8813fb2163f 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -180,7 +180,7 @@ rpcrdma_wc_receive(struct ib_cq *cq, struct ib_wc *wc)
 	rep->rr_wc_flags = wc->wc_flags;
 	rep->rr_inv_rkey = wc->ex.invalidate_rkey;
 
-	ib_dma_sync_single_for_cpu(rep->rr_device,
+	ib_dma_sync_single_for_cpu(rdmab_device(rep->rr_rdmabuf),
 				   rdmab_addr(rep->rr_rdmabuf),
 				   rep->rr_len, DMA_FROM_DEVICE);
 
@@ -878,7 +878,6 @@ struct rpcrdma_rep *
 rpcrdma_create_rep(struct rpcrdma_xprt *r_xprt)
 {
 	struct rpcrdma_create_data_internal *cdata = &r_xprt->rx_data;
-	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 	struct rpcrdma_rep *rep;
 	int rc;
 
@@ -894,7 +893,6 @@ rpcrdma_create_rep(struct rpcrdma_xprt *r_xprt)
 		goto out_free;
 	}
 
-	rep->rr_device = ia->ri_device;
 	rep->rr_cqe.done = rpcrdma_wc_receive;
 	rep->rr_rxprt = r_xprt;
 	INIT_WORK(&rep->rr_work, rpcrdma_reply_handler);
@@ -1232,17 +1230,19 @@ rpcrdma_alloc_regbuf(size_t size, enum dma_data_direction direction,
 bool
 __rpcrdma_dma_map_regbuf(struct rpcrdma_ia *ia, struct rpcrdma_regbuf *rb)
 {
+	struct ib_device *device = ia->ri_device;
+
 	if (rb->rg_direction == DMA_NONE)
 		return false;
 
-	rb->rg_iov.addr = ib_dma_map_single(ia->ri_device,
+	rb->rg_iov.addr = ib_dma_map_single(device,
 					    (void *)rb->rg_base,
 					    rdmab_length(rb),
 					    rb->rg_direction);
-	if (ib_dma_mapping_error(ia->ri_device, rdmab_addr(rb)))
+	if (ib_dma_mapping_error(device, rdmab_addr(rb)))
 		return false;
 
-	rb->rg_device = ia->ri_device;
+	rb->rg_device = device;
 	rb->rg_iov.lkey = ia->ri_pd->local_dma_lkey;
 	return true;
 }
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index af844fc30bd4..9d58260533fc 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -164,6 +164,12 @@ rdmab_to_msg(struct rpcrdma_regbuf *rb)
 	return (struct rpcrdma_msg *)rb->rg_base;
 }
 
+static inline struct ib_device *
+rdmab_device(struct rpcrdma_regbuf *rb)
+{
+	return rb->rg_device;
+}
+
 #define RPCRDMA_DEF_GFP		(GFP_NOIO | __GFP_NOWARN)
 
 /* To ensure a transport can always make forward progress,
@@ -209,7 +215,6 @@ struct rpcrdma_rep {
 	unsigned int		rr_len;
 	int			rr_wc_flags;
 	u32			rr_inv_rkey;
-	struct ib_device	*rr_device;
 	struct rpcrdma_xprt	*rr_rxprt;
 	struct work_struct	rr_work;
 	struct list_head	rr_list;
-- 
cgit v1.2.3


From bebd031866caa404c522e91bb6fd0c69be04c707 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:23:10 -0400
Subject: xprtrdma: Support unplugging an HCA from under an NFS mount

The device driver for the underlying physical device associated
with an RPC-over-RDMA transport can be removed while RPC-over-RDMA
transports are still in use (ie, while NFS filesystems are still
mounted and active). The IB core performs a connection event upcall
to request that consumers free all RDMA resources associated with
a transport.

There may be pending RPCs when this occurs. Care must be taken to
release associated resources without leaving references that can
trigger a subsequent crash if a signal or soft timeout occurs. We
rely on the caller of the transport's ->close method to ensure that
the previous RPC task has invoked xprt_release but the transport
remains write-locked.

A DEVICE_REMOVE upcall forces a disconnect then sleeps. When ->close
is invoked, it destroys the transport's H/W resources, then wakes
the upcall, which completes and allows the core driver unload to
continue.

BugLink: https://bugzilla.linux-nfs.org/show_bug.cgi?id=266
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/transport.c | 29 +++++++++++++---
 net/sunrpc/xprtrdma/verbs.c     | 74 ++++++++++++++++++++++++++++++++++++++---
 net/sunrpc/xprtrdma/xprt_rdma.h |  7 ++++
 3 files changed, 101 insertions(+), 9 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index 83e219d7aba4..62ecbccd9748 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -457,19 +457,33 @@ out1:
 	return ERR_PTR(rc);
 }
 
-/*
- * Close a connection, during shutdown or timeout/reconnect
+/**
+ * xprt_rdma_close - Close down RDMA connection
+ * @xprt: generic transport to be closed
+ *
+ * Called during transport shutdown reconnect, or device
+ * removal. Caller holds the transport's write lock.
  */
 static void
 xprt_rdma_close(struct rpc_xprt *xprt)
 {
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
+	struct rpcrdma_ep *ep = &r_xprt->rx_ep;
+	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
+
+	dprintk("RPC:       %s: closing xprt %p\n", __func__, xprt);
 
-	dprintk("RPC:       %s: closing\n", __func__);
-	if (r_xprt->rx_ep.rep_connected > 0)
+	if (test_and_clear_bit(RPCRDMA_IAF_REMOVING, &ia->ri_flags)) {
+		xprt_clear_connected(xprt);
+		rpcrdma_ia_remove(ia);
+		return;
+	}
+	if (ep->rep_connected == -ENODEV)
+		return;
+	if (ep->rep_connected > 0)
 		xprt->reestablish_timeout = 0;
 	xprt_disconnect_done(xprt);
-	rpcrdma_ep_disconnect(&r_xprt->rx_ep, &r_xprt->rx_ia);
+	rpcrdma_ep_disconnect(ep, ia);
 }
 
 static void
@@ -680,6 +694,8 @@ xprt_rdma_free(struct rpc_task *task)
  * xprt_rdma_send_request - marshal and send an RPC request
  * @task: RPC task with an RPC message in rq_snd_buf
  *
+ * Caller holds the transport's write lock.
+ *
  * Return values:
  *        0:	The request has been sent
  * ENOTCONN:	Caller needs to invoke connect logic then call again
@@ -706,6 +722,9 @@ xprt_rdma_send_request(struct rpc_task *task)
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 	int rc = 0;
 
+	if (!xprt_connected(xprt))
+		goto drop_connection;
+
 	/* On retransmit, remove any previously registered chunks */
 	if (unlikely(!list_empty(&req->rl_registered)))
 		r_xprt->rx_ia.ri_ops->ro_unmap_safe(r_xprt, req, false);
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index c8813fb2163f..938fd9e6f308 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -69,6 +69,8 @@
 /*
  * internal functions
  */
+static void rpcrdma_destroy_mrs(struct rpcrdma_buffer *buf);
+static void rpcrdma_dma_unmap_regbuf(struct rpcrdma_regbuf *rb);
 
 static struct workqueue_struct *rpcrdma_receive_wq;
 
@@ -262,6 +264,21 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 			__func__, ep);
 		complete(&ia->ri_done);
 		break;
+	case RDMA_CM_EVENT_DEVICE_REMOVAL:
+#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
+		pr_info("rpcrdma: removing device for %pIS:%u\n",
+			sap, rpc_get_port(sap));
+#endif
+		set_bit(RPCRDMA_IAF_REMOVING, &ia->ri_flags);
+		ep->rep_connected = -ENODEV;
+		xprt_force_disconnect(&xprt->rx_xprt);
+		wait_for_completion(&ia->ri_remove_done);
+
+		ia->ri_id = NULL;
+		ia->ri_pd = NULL;
+		ia->ri_device = NULL;
+		/* Return 1 to ensure the core destroys the id. */
+		return 1;
 	case RDMA_CM_EVENT_ESTABLISHED:
 		connstate = 1;
 		ib_query_qp(ia->ri_id->qp, attr,
@@ -291,9 +308,6 @@ rpcrdma_conn_upcall(struct rdma_cm_id *id, struct rdma_cm_event *event)
 		goto connected;
 	case RDMA_CM_EVENT_DISCONNECTED:
 		connstate = -ECONNABORTED;
-		goto connected;
-	case RDMA_CM_EVENT_DEVICE_REMOVAL:
-		connstate = -ENODEV;
 connected:
 		dprintk("RPC:       %s: %sconnected\n",
 					__func__, connstate > 0 ? "" : "dis");
@@ -346,6 +360,7 @@ rpcrdma_create_id(struct rpcrdma_xprt *xprt,
 	int rc;
 
 	init_completion(&ia->ri_done);
+	init_completion(&ia->ri_remove_done);
 
 	id = rdma_create_id(&init_net, rpcrdma_conn_upcall, xprt, RDMA_PS_TCP,
 			    IB_QPT_RC);
@@ -468,6 +483,56 @@ out_err:
 	return rc;
 }
 
+/**
+ * rpcrdma_ia_remove - Handle device driver unload
+ * @ia: interface adapter being removed
+ *
+ * Divest transport H/W resources associated with this adapter,
+ * but allow it to be restored later.
+ */
+void
+rpcrdma_ia_remove(struct rpcrdma_ia *ia)
+{
+	struct rpcrdma_xprt *r_xprt = container_of(ia, struct rpcrdma_xprt,
+						   rx_ia);
+	struct rpcrdma_ep *ep = &r_xprt->rx_ep;
+	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
+	struct rpcrdma_req *req;
+	struct rpcrdma_rep *rep;
+
+	cancel_delayed_work_sync(&buf->rb_refresh_worker);
+
+	/* This is similar to rpcrdma_ep_destroy, but:
+	 * - Don't cancel the connect worker.
+	 * - Don't call rpcrdma_ep_disconnect, which waits
+	 *   for another conn upcall, which will deadlock.
+	 * - rdma_disconnect is unneeded, the underlying
+	 *   connection is already gone.
+	 */
+	if (ia->ri_id->qp) {
+		ib_drain_qp(ia->ri_id->qp);
+		rdma_destroy_qp(ia->ri_id);
+		ia->ri_id->qp = NULL;
+	}
+	ib_free_cq(ep->rep_attr.recv_cq);
+	ib_free_cq(ep->rep_attr.send_cq);
+
+	/* The ULP is responsible for ensuring all DMA
+	 * mappings and MRs are gone.
+	 */
+	list_for_each_entry(rep, &buf->rb_recv_bufs, rr_list)
+		rpcrdma_dma_unmap_regbuf(rep->rr_rdmabuf);
+	list_for_each_entry(req, &buf->rb_allreqs, rl_all) {
+		rpcrdma_dma_unmap_regbuf(req->rl_rdmabuf);
+		rpcrdma_dma_unmap_regbuf(req->rl_sendbuf);
+		rpcrdma_dma_unmap_regbuf(req->rl_recvbuf);
+	}
+	rpcrdma_destroy_mrs(buf);
+
+	/* Allow waiters to continue */
+	complete(&ia->ri_remove_done);
+}
+
 /**
  * rpcrdma_ia_close - Clean up/close an IA.
  * @ia: interface adapter to close
@@ -1080,7 +1145,8 @@ rpcrdma_get_mw(struct rpcrdma_xprt *r_xprt)
 
 out_nomws:
 	dprintk("RPC:       %s: no MWs available\n", __func__);
-	schedule_delayed_work(&buf->rb_refresh_worker, 0);
+	if (r_xprt->rx_ep.rep_connected != -ENODEV)
+		schedule_delayed_work(&buf->rb_refresh_worker, 0);
 
 	/* Allow the reply handler and refresh worker to run */
 	cond_resched();
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index 9d58260533fc..1c5de1af195b 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -69,6 +69,7 @@ struct rpcrdma_ia {
 	struct rdma_cm_id 	*ri_id;
 	struct ib_pd		*ri_pd;
 	struct completion	ri_done;
+	struct completion	ri_remove_done;
 	int			ri_async_rc;
 	unsigned int		ri_max_segs;
 	unsigned int		ri_max_frmr_depth;
@@ -78,10 +79,15 @@ struct rpcrdma_ia {
 	bool			ri_reminv_expected;
 	bool			ri_implicit_roundup;
 	enum ib_mr_type		ri_mrtype;
+	unsigned long		ri_flags;
 	struct ib_qp_attr	ri_qp_attr;
 	struct ib_qp_init_attr	ri_qp_init_attr;
 };
 
+enum {
+	RPCRDMA_IAF_REMOVING = 0,
+};
+
 /*
  * RDMA Endpoint -- one per transport instance
  */
@@ -511,6 +517,7 @@ extern unsigned int xprt_rdma_memreg_strategy;
  * Interface Adapter calls - xprtrdma/verbs.c
  */
 int rpcrdma_ia_open(struct rpcrdma_xprt *xprt, struct sockaddr *addr);
+void rpcrdma_ia_remove(struct rpcrdma_ia *ia);
 void rpcrdma_ia_close(struct rpcrdma_ia *);
 bool frwr_is_supported(struct rpcrdma_ia *);
 bool fmr_is_supported(struct rpcrdma_ia *);
-- 
cgit v1.2.3


From 1890896b4e01e3858db71fa55a8edf85e15f9946 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:23:18 -0400
Subject: xprtrdma: Refactor rpcrdma_ep_connect

I'm about to add another arm to

    if (ep->rep_connected != 0)

It will be cleaner to use a switch statement here. We'll be looking
for a couple of specific errnos, or "anything else," basically to
sort out the difference between a normal reconnect and recovery from
device removal.

This is a refactoring change only.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/verbs.c | 109 +++++++++++++++++++++++++-------------------
 1 file changed, 63 insertions(+), 46 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 938fd9e6f308..6479ad3fe69d 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -711,6 +711,57 @@ rpcrdma_ep_destroy(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
 	ib_free_cq(ep->rep_attr.send_cq);
 }
 
+static int
+rpcrdma_ep_reconnect(struct rpcrdma_xprt *r_xprt, struct rpcrdma_ep *ep,
+		     struct rpcrdma_ia *ia)
+{
+	struct sockaddr *sap = (struct sockaddr *)&r_xprt->rx_data.addr;
+	struct rdma_cm_id *id, *old;
+	int err, rc;
+
+	dprintk("RPC:       %s: reconnecting...\n", __func__);
+
+	rpcrdma_ep_disconnect(ep, ia);
+
+	rc = -EHOSTUNREACH;
+	id = rpcrdma_create_id(r_xprt, ia, sap);
+	if (IS_ERR(id))
+		goto out;
+
+	/* As long as the new ID points to the same device as the
+	 * old ID, we can reuse the transport's existing PD and all
+	 * previously allocated MRs. Also, the same device means
+	 * the transport's previous DMA mappings are still valid.
+	 *
+	 * This is a sanity check only. There should be no way these
+	 * point to two different devices here.
+	 */
+	old = id;
+	rc = -ENETUNREACH;
+	if (ia->ri_device != id->device) {
+		pr_err("rpcrdma: can't reconnect on different device!\n");
+		goto out_destroy;
+	}
+
+	err = rdma_create_qp(id, ia->ri_pd, &ep->rep_attr);
+	if (err) {
+		dprintk("RPC:       %s: rdma_create_qp returned %d\n",
+			__func__, err);
+		goto out_destroy;
+	}
+
+	/* Atomically replace the transport's ID and QP. */
+	rc = 0;
+	old = ia->ri_id;
+	ia->ri_id = id;
+	rdma_destroy_qp(old);
+
+out_destroy:
+	rpcrdma_destroy_id(old);
+out:
+	return rc;
+}
+
 /*
  * Connect unconnected endpoint.
  */
@@ -719,61 +770,25 @@ 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 rdma_cm_id *id, *old;
-	struct sockaddr *sap;
 	unsigned int extras;
-	int rc = 0;
+	int rc;
 
-	if (ep->rep_connected != 0) {
 retry:
-		dprintk("RPC:       %s: reconnecting...\n", __func__);
-
-		rpcrdma_ep_disconnect(ep, ia);
-
-		sap = (struct sockaddr *)&r_xprt->rx_data.addr;
-		id = rpcrdma_create_id(r_xprt, ia, sap);
-		if (IS_ERR(id)) {
-			rc = -EHOSTUNREACH;
-			goto out;
-		}
-		/* TEMP TEMP TEMP - fail if new device:
-		 * Deregister/remarshal *all* requests!
-		 * Close and recreate adapter, pd, etc!
-		 * Re-determine all attributes still sane!
-		 * More stuff I haven't thought of!
-		 * Rrrgh!
-		 */
-		if (ia->ri_device != id->device) {
-			printk("RPC:       %s: can't reconnect on "
-				"different device!\n", __func__);
-			rpcrdma_destroy_id(id);
-			rc = -ENETUNREACH;
-			goto out;
-		}
-		/* END TEMP */
-		rc = rdma_create_qp(id, ia->ri_pd, &ep->rep_attr);
-		if (rc) {
-			dprintk("RPC:       %s: rdma_create_qp failed %i\n",
-				__func__, rc);
-			rpcrdma_destroy_id(id);
-			rc = -ENETUNREACH;
-			goto out;
-		}
-
-		old = ia->ri_id;
-		ia->ri_id = id;
-
-		rdma_destroy_qp(old);
-		rpcrdma_destroy_id(old);
-	} else {
+	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);
 		if (rc) {
 			dprintk("RPC:       %s: rdma_create_qp failed %i\n",
 				__func__, rc);
-			/* do not update ep->rep_connected */
-			return -ENETUNREACH;
+			rc = -ENETUNREACH;
+			goto out_noupdate;
 		}
+		break;
+	default:
+		rc = rpcrdma_ep_reconnect(r_xprt, ep, ia);
+		if (rc)
+			goto out;
 	}
 
 	ep->rep_connected = 0;
@@ -801,6 +816,8 @@ retry:
 out:
 	if (rc)
 		ep->rep_connected = rc;
+
+out_noupdate:
 	return rc;
 }
 
-- 
cgit v1.2.3


From a9b0e381caa965bc9b00dad1e593c502f48f8c3d Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:23:26 -0400
Subject: xprtrdma: Restore transport after device removal

After a device removal, enable the transport connect worker to
restore normal operation if there is another device with
connectivity to the server.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/verbs.c | 48 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 48 insertions(+)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 6479ad3fe69d..b69daa30874f 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -69,6 +69,7 @@
 /*
  * internal functions
  */
+static void rpcrdma_create_mrs(struct rpcrdma_xprt *r_xprt);
 static void rpcrdma_destroy_mrs(struct rpcrdma_buffer *buf);
 static void rpcrdma_dma_unmap_regbuf(struct rpcrdma_regbuf *rb);
 
@@ -711,6 +712,48 @@ rpcrdma_ep_destroy(struct rpcrdma_ep *ep, struct rpcrdma_ia *ia)
 	ib_free_cq(ep->rep_attr.send_cq);
 }
 
+/* Re-establish a connection after a device removal event.
+ * 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)
+{
+	struct sockaddr *sap = (struct sockaddr *)&r_xprt->rx_data.addr;
+	int rc, err;
+
+	pr_info("%s: r_xprt = %p\n", __func__, r_xprt);
+
+	rc = -EHOSTUNREACH;
+	if (rpcrdma_ia_open(r_xprt, sap))
+		goto out1;
+
+	rc = -ENOMEM;
+	err = rpcrdma_ep_create(ep, ia, &r_xprt->rx_data);
+	if (err) {
+		pr_err("rpcrdma: rpcrdma_ep_create returned %d\n", err);
+		goto out2;
+	}
+
+	rc = -ENETUNREACH;
+	err = rdma_create_qp(ia->ri_id, ia->ri_pd, &ep->rep_attr);
+	if (err) {
+		pr_err("rpcrdma: rdma_create_qp returned %d\n", err);
+		goto out3;
+	}
+
+	rpcrdma_create_mrs(r_xprt);
+	return 0;
+
+out3:
+	rpcrdma_ep_destroy(ep, ia);
+out2:
+	rpcrdma_ia_close(ia);
+out1:
+	return rc;
+}
+
 static int
 rpcrdma_ep_reconnect(struct rpcrdma_xprt *r_xprt, struct rpcrdma_ep *ep,
 		     struct rpcrdma_ia *ia)
@@ -785,6 +828,11 @@ retry:
 			goto out_noupdate;
 		}
 		break;
+	case -ENODEV:
+		rc = rpcrdma_ep_recreate_xprt(r_xprt, ep, ia);
+		if (rc)
+			goto out_noupdate;
+		break;
 	default:
 		rc = rpcrdma_ep_reconnect(r_xprt, ep, ia);
 		if (rc)
-- 
cgit v1.2.3


From 56a6bd154ddd60f7d01d6c0a5c272c1224d16b46 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:23:34 -0400
Subject: xprtrdma: Revert commit d0f36c46deea

Device removal is now adequately supported. Pinning the underlying
device driver to prevent removal while an NFS mount is active is no
longer necessary.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/verbs.c | 33 +++++++--------------------------
 1 file changed, 7 insertions(+), 26 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index b69daa30874f..d7c16005491e 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -53,7 +53,7 @@
 #include <linux/sunrpc/addr.h>
 #include <linux/sunrpc/svc_rdma.h>
 #include <asm/bitops.h>
-#include <linux/module.h> /* try_module_get()/module_put() */
+
 #include <rdma/ib_cm.h>
 
 #include "xprt_rdma.h"
@@ -344,14 +344,6 @@ connected:
 	return 0;
 }
 
-static void rpcrdma_destroy_id(struct rdma_cm_id *id)
-{
-	if (id) {
-		module_put(id->device->owner);
-		rdma_destroy_id(id);
-	}
-}
-
 static struct rdma_cm_id *
 rpcrdma_create_id(struct rpcrdma_xprt *xprt,
 			struct rpcrdma_ia *ia, struct sockaddr *addr)
@@ -386,16 +378,6 @@ rpcrdma_create_id(struct rpcrdma_xprt *xprt,
 		goto out;
 	}
 
-	/* FIXME:
-	 * Until xprtrdma supports DEVICE_REMOVAL, the provider must
-	 * be pinned while there are active NFS/RDMA mounts to prevent
-	 * hangs and crashes at umount time.
-	 */
-	if (!ia->ri_async_rc && !try_module_get(id->device->owner)) {
-		dprintk("RPC:       %s: Failed to get device module\n",
-			__func__);
-		ia->ri_async_rc = -ENODEV;
-	}
 	rc = ia->ri_async_rc;
 	if (rc)
 		goto out;
@@ -405,21 +387,20 @@ rpcrdma_create_id(struct rpcrdma_xprt *xprt,
 	if (rc) {
 		dprintk("RPC:       %s: rdma_resolve_route() failed %i\n",
 			__func__, rc);
-		goto put;
+		goto out;
 	}
 	rc = wait_for_completion_interruptible_timeout(&ia->ri_done, wtimeout);
 	if (rc < 0) {
 		dprintk("RPC:       %s: wait() exited: %i\n",
 			__func__, rc);
-		goto put;
+		goto out;
 	}
 	rc = ia->ri_async_rc;
 	if (rc)
-		goto put;
+		goto out;
 
 	return id;
-put:
-	module_put(id->device->owner);
+
 out:
 	rdma_destroy_id(id);
 	return ERR_PTR(rc);
@@ -546,7 +527,7 @@ rpcrdma_ia_close(struct rpcrdma_ia *ia)
 	if (ia->ri_id != NULL && !IS_ERR(ia->ri_id)) {
 		if (ia->ri_id->qp)
 			rdma_destroy_qp(ia->ri_id);
-		rpcrdma_destroy_id(ia->ri_id);
+		rdma_destroy_id(ia->ri_id);
 	}
 	ia->ri_id = NULL;
 	ia->ri_device = NULL;
@@ -800,7 +781,7 @@ rpcrdma_ep_reconnect(struct rpcrdma_xprt *r_xprt, struct rpcrdma_ep *ep,
 	rdma_destroy_qp(old);
 
 out_destroy:
-	rpcrdma_destroy_id(old);
+	rdma_destroy_id(old);
 out:
 	return rc;
 }
-- 
cgit v1.2.3


From 7d7fa9b5509649ee1732fde0c643bea53f8db364 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:23:43 -0400
Subject: xprtrdma: Annotate receive workqueue

Micro-optimize the receive workqueue by marking it's anchor "read-
mostly."

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/verbs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index d7c16005491e..3dbce9ac4327 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -73,7 +73,7 @@ static void rpcrdma_create_mrs(struct rpcrdma_xprt *r_xprt);
 static void rpcrdma_destroy_mrs(struct rpcrdma_buffer *buf);
 static void rpcrdma_dma_unmap_regbuf(struct rpcrdma_regbuf *rb);
 
-static struct workqueue_struct *rpcrdma_receive_wq;
+static struct workqueue_struct *rpcrdma_receive_wq __read_mostly;
 
 int
 rpcrdma_alloc_wq(void)
-- 
cgit v1.2.3


From 0031e47c76ec5ce5b4f64f2750a535a36040c82b Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:23:51 -0400
Subject: xprtrdma: Squelch ENOBUFS warnings

When ro_map is out of buffers, that's not a permanent error, so
don't report a problem.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/rpc_rdma.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index e68131bed3ce..694e9b13ecf0 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -781,9 +781,11 @@ rpcrdma_marshal_req(struct rpc_rqst *rqst)
 	return 0;
 
 out_err:
-	pr_err("rpcrdma: rpcrdma_marshal_req failed, status %ld\n",
-	       PTR_ERR(iptr));
-	r_xprt->rx_stats.failed_marshal_count++;
+	if (PTR_ERR(iptr) != -ENOBUFS) {
+		pr_err("rpcrdma: rpcrdma_marshal_req failed, status %ld\n",
+		       PTR_ERR(iptr));
+		r_xprt->rx_stats.failed_marshal_count++;
+	}
 	return PTR_ERR(iptr);
 }
 
-- 
cgit v1.2.3


From 7ecce75fc3c52590950bc5e68eac0f00e50e8ce4 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:23:59 -0400
Subject: sunrpc: Fix xdr_init_decode_pages() documenting comment

Clean up.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xdr.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'net')

diff --git a/net/sunrpc/xdr.c b/net/sunrpc/xdr.c
index 1f7082144e01..e34f4ee7f2b6 100644
--- a/net/sunrpc/xdr.c
+++ b/net/sunrpc/xdr.c
@@ -807,7 +807,7 @@ void xdr_init_decode(struct xdr_stream *xdr, struct xdr_buf *buf, __be32 *p)
 EXPORT_SYMBOL_GPL(xdr_init_decode);
 
 /**
- * xdr_init_decode - Initialize an xdr_stream for decoding data.
+ * xdr_init_decode_pages - Initialize an xdr_stream for decoding into pages
  * @xdr: pointer to xdr_stream struct
  * @buf: pointer to XDR buffer from which to decode data
  * @pages: list of pages to decode into
-- 
cgit v1.2.3


From 2be1fce95e5b017dd7d23ca039d58cbefd0221e6 Mon Sep 17 00:00:00 2001
From: Chuck Lever <chuck.lever@oracle.com>
Date: Tue, 11 Apr 2017 13:24:07 -0400
Subject: xprtrdma: Remove rpcrdma_buffer::rb_pool

Since commit 1e465fd4ff47 ("xprtrdma: Replace send and receive
arrays"), this field is no longer used.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
---
 net/sunrpc/xprtrdma/xprt_rdma.h | 1 -
 1 file changed, 1 deletion(-)

(limited to 'net')

diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index 1c5de1af195b..1d66acf1a723 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -391,7 +391,6 @@ struct rpcrdma_buffer {
 	spinlock_t		rb_mwlock;	/* protect rb_mws list */
 	struct list_head	rb_mws;
 	struct list_head	rb_all;
-	char			*rb_pool;
 
 	spinlock_t		rb_lock;	/* protect buf lists */
 	int			rb_send_count, rb_recv_count;
-- 
cgit v1.2.3