Merge tag 'nfs-for-4.2-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client updates from Trond Myklebust:
 "Highlights include:

  Stable patches:
   - Fix a crash in the NFSv4 file locking code.
   - Fix an fsync() regression, where we were failing to retry I/O in
     some circumstances.
   - Fix an infinite loop in NFSv4.0 OPEN stateid recovery
   - Fix a memory leak when an attempted pnfs fails.
   - Fix a memory leak in the backchannel code
   - Large hostnames were not supported correctly in NFSv4.1
   - Fix a pNFS/flexfiles bug that was impeding error reporting on I/O.
   - Fix a couple of credential issues in pNFS/flexfiles

  Bugfixes + cleanups:
   - Open flag sanity checks in the NFSv4 atomic open codepath
   - More NFSv4 delegation related bugfixes
   - Various NFSv4.1 backchannel bugfixes and cleanups
   - Fix the NFS swap socket code
   - Various cleanups of the NFSv4 SETCLIENTID and EXCHANGE_ID code
   - Fix a UDP transport deadlock issue

  Features:
   - More RDMA client transport improvements
   - NFSv4.2 LAYOUTSTATS functionality for pnfs flexfiles"

* tag 'nfs-for-4.2-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (87 commits)
  nfs: Remove invalid tk_pid from debug message
  nfs: Remove invalid NFS_ATTR_FATTR_V4_REFERRAL checking in nfs4_get_rootfh
  nfs: Drop bad comment in nfs41_walk_client_list()
  nfs: Remove unneeded micro checking of CONFIG_PROC_FS
  nfs: Don't setting FILE_CREATED flags always
  nfs: Use remove_proc_subtree() instead remove_proc_entry()
  nfs: Remove unused argument in nfs_server_set_fsinfo()
  nfs: Fix a memory leak when meeting an unsupported state protect
  nfs: take extra reference to fl->fl_file when running a LOCKU operation
  NFSv4: When returning a delegation, don't reclaim an incompatible open mode.
  NFSv4.2: LAYOUTSTATS is optional to implement
  NFSv4.2: Fix up a decoding error in layoutstats
  pNFS/flexfiles: Fix the reset of struct pgio_header when resending
  pNFS/flexfiles: Turn off layoutcommit for servers that don't need it
  pnfs/flexfiles: protect ktime manipulation with mirror lock
  nfs: provide pnfs_report_layoutstat when NFS42 is disabled
  nfs: verify open flags before allowing open
  nfs: always update creds in mirror, even when we have an already connected ds
  nfs: fix potential credential leak in ff_layout_update_mirror_cred
  pnfs/flexfiles: report layoutstat regularly
  ...

15 files changed, 751 insertions, 532 deletions

diff --git a/net/sunrpc/Makefile b/net/sunrpc/Makefile
index 936ad0a15371..b512fbd9d79a 100644
--- a/net/sunrpc/Makefile
+++ b/net/sunrpc/Makefile
@@ -14,6 +14,6 @@ sunrpc-y := clnt.o xprt.o socklib.o xprtsock.o sched.o \
 	    sunrpc_syms.o cache.o rpc_pipe.o \
 	    svc_xprt.o
 sunrpc-$(CONFIG_SUNRPC_DEBUG) += debugfs.o
-sunrpc-$(CONFIG_SUNRPC_BACKCHANNEL) += backchannel_rqst.o bc_svc.o
+sunrpc-$(CONFIG_SUNRPC_BACKCHANNEL) += backchannel_rqst.o
 sunrpc-$(CONFIG_PROC_FS) += stats.o
 sunrpc-$(CONFIG_SYSCTL) += sysctl.o
diff --git a/net/sunrpc/backchannel_rqst.c b/net/sunrpc/backchannel_rqst.c
index 9dd0ea8db463..9825ff0f91d6 100644
--- a/net/sunrpc/backchannel_rqst.c
+++ b/net/sunrpc/backchannel_rqst.c
@@ -37,16 +37,18 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 static inline int xprt_need_to_requeue(struct rpc_xprt *xprt)
 {
-	return xprt->bc_alloc_count > 0;
+	return xprt->bc_alloc_count < atomic_read(&xprt->bc_free_slots);
 }
 
 static inline void xprt_inc_alloc_count(struct rpc_xprt *xprt, unsigned int n)
 {
+	atomic_add(n, &xprt->bc_free_slots);
 	xprt->bc_alloc_count += n;
 }
 
 static inline int xprt_dec_alloc_count(struct rpc_xprt *xprt, unsigned int n)
 {
+	atomic_sub(n, &xprt->bc_free_slots);
 	return xprt->bc_alloc_count -= n;
 }
 
@@ -60,13 +62,62 @@ static void xprt_free_allocation(struct rpc_rqst *req)
 
 	dprintk("RPC:        free allocations for req= %p\n", req);
 	WARN_ON_ONCE(test_bit(RPC_BC_PA_IN_USE, &req->rq_bc_pa_state));
-	xbufp = &req->rq_private_buf;
+	xbufp = &req->rq_rcv_buf;
 	free_page((unsigned long)xbufp->head[0].iov_base);
 	xbufp = &req->rq_snd_buf;
 	free_page((unsigned long)xbufp->head[0].iov_base);
 	kfree(req);
 }
 
+static int xprt_alloc_xdr_buf(struct xdr_buf *buf, gfp_t gfp_flags)
+{
+	struct page *page;
+	/* Preallocate one XDR receive buffer */
+	page = alloc_page(gfp_flags);
+	if (page == NULL)
+		return -ENOMEM;
+	buf->head[0].iov_base = page_address(page);
+	buf->head[0].iov_len = PAGE_SIZE;
+	buf->tail[0].iov_base = NULL;
+	buf->tail[0].iov_len = 0;
+	buf->page_len = 0;
+	buf->len = 0;
+	buf->buflen = PAGE_SIZE;
+	return 0;
+}
+
+static
+struct rpc_rqst *xprt_alloc_bc_req(struct rpc_xprt *xprt, gfp_t gfp_flags)
+{
+	struct rpc_rqst *req;
+
+	/* Pre-allocate one backchannel rpc_rqst */
+	req = kzalloc(sizeof(*req), gfp_flags);
+	if (req == NULL)
+		return NULL;
+
+	req->rq_xprt = xprt;
+	INIT_LIST_HEAD(&req->rq_list);
+	INIT_LIST_HEAD(&req->rq_bc_list);
+
+	/* Preallocate one XDR receive buffer */
+	if (xprt_alloc_xdr_buf(&req->rq_rcv_buf, gfp_flags) < 0) {
+		printk(KERN_ERR "Failed to create bc receive xbuf\n");
+		goto out_free;
+	}
+	req->rq_rcv_buf.len = PAGE_SIZE;
+
+	/* Preallocate one XDR send buffer */
+	if (xprt_alloc_xdr_buf(&req->rq_snd_buf, gfp_flags) < 0) {
+		printk(KERN_ERR "Failed to create bc snd xbuf\n");
+		goto out_free;
+	}
+	return req;
+out_free:
+	xprt_free_allocation(req);
+	return NULL;
+}
+
 /*
  * Preallocate up to min_reqs structures and related buffers for use
  * by the backchannel.  This function can be called multiple times
@@ -87,9 +138,7 @@ static void xprt_free_allocation(struct rpc_rqst *req)
  */
 int xprt_setup_backchannel(struct rpc_xprt *xprt, unsigned int min_reqs)
 {
-	struct page *page_rcv = NULL, *page_snd = NULL;
-	struct xdr_buf *xbufp = NULL;
-	struct rpc_rqst *req, *tmp;
+	struct rpc_rqst *req;
 	struct list_head tmp_list;
 	int i;
 
@@ -106,7 +155,7 @@ int xprt_setup_backchannel(struct rpc_xprt *xprt, unsigned int min_reqs)
 	INIT_LIST_HEAD(&tmp_list);
 	for (i = 0; i < min_reqs; i++) {
 		/* Pre-allocate one backchannel rpc_rqst */
-		req = kzalloc(sizeof(struct rpc_rqst), GFP_KERNEL);
+		req = xprt_alloc_bc_req(xprt, GFP_KERNEL);
 		if (req == NULL) {
 			printk(KERN_ERR "Failed to create bc rpc_rqst\n");
 			goto out_free;
@@ -115,41 +164,6 @@ int xprt_setup_backchannel(struct rpc_xprt *xprt, unsigned int min_reqs)
 		/* Add the allocated buffer to the tmp list */
 		dprintk("RPC:       adding req= %p\n", req);
 		list_add(&req->rq_bc_pa_list, &tmp_list);
-
-		req->rq_xprt = xprt;
-		INIT_LIST_HEAD(&req->rq_list);
-		INIT_LIST_HEAD(&req->rq_bc_list);
-
-		/* Preallocate one XDR receive buffer */
-		page_rcv = alloc_page(GFP_KERNEL);
-		if (page_rcv == NULL) {
-			printk(KERN_ERR "Failed to create bc receive xbuf\n");
-			goto out_free;
-		}
-		xbufp = &req->rq_rcv_buf;
-		xbufp->head[0].iov_base = page_address(page_rcv);
-		xbufp->head[0].iov_len = PAGE_SIZE;
-		xbufp->tail[0].iov_base = NULL;
-		xbufp->tail[0].iov_len = 0;
-		xbufp->page_len = 0;
-		xbufp->len = PAGE_SIZE;
-		xbufp->buflen = PAGE_SIZE;
-
-		/* Preallocate one XDR send buffer */
-		page_snd = alloc_page(GFP_KERNEL);
-		if (page_snd == NULL) {
-			printk(KERN_ERR "Failed to create bc snd xbuf\n");
-			goto out_free;
-		}
-
-		xbufp = &req->rq_snd_buf;
-		xbufp->head[0].iov_base = page_address(page_snd);
-		xbufp->head[0].iov_len = 0;
-		xbufp->tail[0].iov_base = NULL;
-		xbufp->tail[0].iov_len = 0;
-		xbufp->page_len = 0;
-		xbufp->len = 0;
-		xbufp->buflen = PAGE_SIZE;
 	}
 
 	/*
@@ -167,7 +181,10 @@ out_free:
 	/*
 	 * Memory allocation failed, free the temporary list
 	 */
-	list_for_each_entry_safe(req, tmp, &tmp_list, rq_bc_pa_list) {
+	while (!list_empty(&tmp_list)) {
+		req = list_first_entry(&tmp_list,
+				struct rpc_rqst,
+				rq_bc_pa_list);
 		list_del(&req->rq_bc_pa_list);
 		xprt_free_allocation(req);
 	}
@@ -217,9 +234,15 @@ static struct rpc_rqst *xprt_alloc_bc_request(struct rpc_xprt *xprt, __be32 xid)
 	struct rpc_rqst *req = NULL;
 
 	dprintk("RPC:       allocate a backchannel request\n");
-	if (list_empty(&xprt->bc_pa_list))
+	if (atomic_read(&xprt->bc_free_slots) <= 0)
 		goto not_found;
-
+	if (list_empty(&xprt->bc_pa_list)) {
+		req = xprt_alloc_bc_req(xprt, GFP_ATOMIC);
+		if (!req)
+			goto not_found;
+		/* Note: this 'free' request adds it to xprt->bc_pa_list */
+		xprt_free_bc_request(req);
+	}
 	req = list_first_entry(&xprt->bc_pa_list, struct rpc_rqst,
 				rq_bc_pa_list);
 	req->rq_reply_bytes_recvd = 0;
@@ -245,11 +268,21 @@ void xprt_free_bc_request(struct rpc_rqst *req)
 
 	req->rq_connect_cookie = xprt->connect_cookie - 1;
 	smp_mb__before_atomic();
-	WARN_ON_ONCE(!test_bit(RPC_BC_PA_IN_USE, &req->rq_bc_pa_state));
 	clear_bit(RPC_BC_PA_IN_USE, &req->rq_bc_pa_state);
 	smp_mb__after_atomic();
 
-	if (!xprt_need_to_requeue(xprt)) {
+	/*
+	 * Return it to the list of preallocations so that it
+	 * may be reused by a new callback request.
+	 */
+	spin_lock_bh(&xprt->bc_pa_lock);
+	if (xprt_need_to_requeue(xprt)) {
+		list_add_tail(&req->rq_bc_pa_list, &xprt->bc_pa_list);
+		xprt->bc_alloc_count++;
+		req = NULL;
+	}
+	spin_unlock_bh(&xprt->bc_pa_lock);
+	if (req != NULL) {
 		/*
 		 * The last remaining session was destroyed while this
 		 * entry was in use.  Free the entry and don't attempt
@@ -260,14 +293,6 @@ void xprt_free_bc_request(struct rpc_rqst *req)
 		xprt_free_allocation(req);
 		return;
 	}
-
-	/*
-	 * Return it to the list of preallocations so that it
-	 * may be reused by a new callback request.
-	 */
-	spin_lock_bh(&xprt->bc_pa_lock);
-	list_add_tail(&req->rq_bc_pa_list, &xprt->bc_pa_list);
-	spin_unlock_bh(&xprt->bc_pa_lock);
 }
 
 /*
@@ -311,6 +336,7 @@ void xprt_complete_bc_request(struct rpc_rqst *req, uint32_t copied)
 
 	spin_lock(&xprt->bc_pa_lock);
 	list_del(&req->rq_bc_pa_list);
+	xprt->bc_alloc_count--;
 	spin_unlock(&xprt->bc_pa_lock);
 
 	req->rq_private_buf.len = copied;
diff --git a/net/sunrpc/bc_svc.c b/net/sunrpc/bc_svc.c
deleted file mode 100644
index 15c7a8a1c24f..000000000000
--- a/net/sunrpc/bc_svc.c
+++ /dev/null
@@ -1,63 +0,0 @@
-/******************************************************************************
-
-(c) 2007 Network Appliance, Inc.  All Rights Reserved.
-(c) 2009 NetApp.  All Rights Reserved.
-
-NetApp provides this source code under the GPL v2 License.
-The GPL v2 license is available at
-http://opensource.org/licenses/gpl-license.php.
-
-THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
-CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
-EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
-PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
-PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
-LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
-NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
-SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-******************************************************************************/
-
-/*
- * The NFSv4.1 callback service helper routines.
- * They implement the transport level processing required to send the
- * reply over an existing open connection previously established by the client.
- */
-
-#include <linux/module.h>
-
-#include <linux/sunrpc/xprt.h>
-#include <linux/sunrpc/sched.h>
-#include <linux/sunrpc/bc_xprt.h>
-
-#define RPCDBG_FACILITY	RPCDBG_SVCDSP
-
-/* Empty callback ops */
-static const struct rpc_call_ops nfs41_callback_ops = {
-};
-
-
-/*
- * Send the callback reply
- */
-int bc_send(struct rpc_rqst *req)
-{
-	struct rpc_task *task;
-	int ret;
-
-	dprintk("RPC:       bc_send req= %p\n", req);
-	task = rpc_run_bc_task(req, &nfs41_callback_ops);
-	if (IS_ERR(task))
-		ret = PTR_ERR(task);
-	else {
-		WARN_ON_ONCE(atomic_read(&task->tk_count) != 1);
-		ret = task->tk_status;
-		rpc_put_task(task);
-	}
-	dprintk("RPC:       bc_send ret= %d\n", ret);
-	return ret;
-}
-
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index e6ce1517367f..cbc6af923dd1 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -891,15 +891,8 @@ void rpc_task_set_client(struct rpc_task *task, struct rpc_clnt *clnt)
 			task->tk_flags |= RPC_TASK_SOFT;
 		if (clnt->cl_noretranstimeo)
 			task->tk_flags |= RPC_TASK_NO_RETRANS_TIMEOUT;
-		if (sk_memalloc_socks()) {
-			struct rpc_xprt *xprt;
-
-			rcu_read_lock();
-			xprt = rcu_dereference(clnt->cl_xprt);
-			if (xprt->swapper)
-				task->tk_flags |= RPC_TASK_SWAPPER;
-			rcu_read_unlock();
-		}
+		if (atomic_read(&clnt->cl_swapper))
+			task->tk_flags |= RPC_TASK_SWAPPER;
 		/* Add to the client's list of all tasks */
 		spin_lock(&clnt->cl_lock);
 		list_add_tail(&task->tk_task, &clnt->cl_tasks);
@@ -1031,15 +1024,14 @@ EXPORT_SYMBOL_GPL(rpc_call_async);
  * rpc_run_bc_task - Allocate a new RPC task for backchannel use, then run
  * rpc_execute against it
  * @req: RPC request
- * @tk_ops: RPC call ops
  */
-struct rpc_task *rpc_run_bc_task(struct rpc_rqst *req,
-				const struct rpc_call_ops *tk_ops)
+struct rpc_task *rpc_run_bc_task(struct rpc_rqst *req)
 {
 	struct rpc_task *task;
 	struct xdr_buf *xbufp = &req->rq_snd_buf;
 	struct rpc_task_setup task_setup_data = {
-		.callback_ops = tk_ops,
+		.callback_ops = &rpc_default_ops,
+		.flags = RPC_TASK_SOFTCONN,
 	};
 
 	dprintk("RPC: rpc_run_bc_task req= %p\n", req);
@@ -1614,6 +1606,7 @@ call_allocate(struct rpc_task *task)
 					req->rq_callsize + req->rq_rcvsize);
 	if (req->rq_buffer != NULL)
 		return;
+	xprt_inject_disconnect(xprt);
 
 	dprintk("RPC: %5u rpc_buffer allocation failed\n", task->tk_pid);
 
@@ -1951,33 +1944,36 @@ call_bc_transmit(struct rpc_task *task)
 {
 	struct rpc_rqst *req = task->tk_rqstp;
 
-	if (!xprt_prepare_transmit(task)) {
-		/*
-		 * Could not reserve the transport. Try again after the
-		 * transport is released.
-		 */
-		task->tk_status = 0;
-		task->tk_action = call_bc_transmit;
-		return;
-	}
+	if (!xprt_prepare_transmit(task))
+		goto out_retry;
 
-	task->tk_action = rpc_exit_task;
 	if (task->tk_status < 0) {
 		printk(KERN_NOTICE "RPC: Could not send backchannel reply "
 			"error: %d\n", task->tk_status);
-		return;
+		goto out_done;
 	}
+	if (req->rq_connect_cookie != req->rq_xprt->connect_cookie)
+		req->rq_bytes_sent = 0;
 
 	xprt_transmit(task);
+
+	if (task->tk_status == -EAGAIN)
+		goto out_nospace;
+
 	xprt_end_transmit(task);
 	dprint_status(task);
 	switch (task->tk_status) {
 	case 0:
 		/* Success */
-		break;
 	case -EHOSTDOWN:
 	case -EHOSTUNREACH:
 	case -ENETUNREACH:
+	case -ECONNRESET:
+	case -ECONNREFUSED:
+	case -EADDRINUSE:
+	case -ENOTCONN:
+	case -EPIPE:
+		break;
 	case -ETIMEDOUT:
 		/*
 		 * Problem reaching the server.  Disconnect and let the
@@ -2002,6 +1998,13 @@ call_bc_transmit(struct rpc_task *task)
 		break;
 	}
 	rpc_wake_up_queued_task(&req->rq_xprt->pending, task);
+out_done:
+	task->tk_action = rpc_exit_task;
+	return;
+out_nospace:
+	req->rq_connect_cookie = req->rq_xprt->connect_cookie;
+out_retry:
+	task->tk_status = 0;
 }
 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
 
@@ -2476,3 +2479,59 @@ void rpc_show_tasks(struct net *net)
 	spin_unlock(&sn->rpc_client_lock);
 }
 #endif
+
+#if IS_ENABLED(CONFIG_SUNRPC_SWAP)
+int
+rpc_clnt_swap_activate(struct rpc_clnt *clnt)
+{
+	int ret = 0;
+	struct rpc_xprt	*xprt;
+
+	if (atomic_inc_return(&clnt->cl_swapper) == 1) {
+retry:
+		rcu_read_lock();
+		xprt = xprt_get(rcu_dereference(clnt->cl_xprt));
+		rcu_read_unlock();
+		if (!xprt) {
+			/*
+			 * If we didn't get a reference, then we likely are
+			 * racing with a migration event. Wait for a grace
+			 * period and try again.
+			 */
+			synchronize_rcu();
+			goto retry;
+		}
+
+		ret = xprt_enable_swap(xprt);
+		xprt_put(xprt);
+	}
+	return ret;
+}
+EXPORT_SYMBOL_GPL(rpc_clnt_swap_activate);
+
+void
+rpc_clnt_swap_deactivate(struct rpc_clnt *clnt)
+{
+	struct rpc_xprt	*xprt;
+
+	if (atomic_dec_if_positive(&clnt->cl_swapper) == 0) {
+retry:
+		rcu_read_lock();
+		xprt = xprt_get(rcu_dereference(clnt->cl_xprt));
+		rcu_read_unlock();
+		if (!xprt) {
+			/*
+			 * If we didn't get a reference, then we likely are
+			 * racing with a migration event. Wait for a grace
+			 * period and try again.
+			 */
+			synchronize_rcu();
+			goto retry;
+		}
+
+		xprt_disable_swap(xprt);
+		xprt_put(xprt);
+	}
+}
+EXPORT_SYMBOL_GPL(rpc_clnt_swap_deactivate);
+#endif /* CONFIG_SUNRPC_SWAP */
diff --git a/net/sunrpc/debugfs.c b/net/sunrpc/debugfs.c
index 82962f7e6e88..e7b4d93566df 100644
--- a/net/sunrpc/debugfs.c
+++ b/net/sunrpc/debugfs.c
@@ -10,9 +10,12 @@
 #include "netns.h"
 
 static struct dentry *topdir;
+static struct dentry *rpc_fault_dir;
 static struct dentry *rpc_clnt_dir;
 static struct dentry *rpc_xprt_dir;
 
+unsigned int rpc_inject_disconnect;
+
 struct rpc_clnt_iter {
 	struct rpc_clnt	*clnt;
 	loff_t		pos;
@@ -257,6 +260,8 @@ rpc_xprt_debugfs_register(struct rpc_xprt *xprt)
 		debugfs_remove_recursive(xprt->debugfs);
 		xprt->debugfs = NULL;
 	}
+
+	atomic_set(&xprt->inject_disconnect, rpc_inject_disconnect);
 }
 
 void
@@ -266,11 +271,79 @@ rpc_xprt_debugfs_unregister(struct rpc_xprt *xprt)
 	xprt->debugfs = NULL;
 }
 
+static int
+fault_open(struct inode *inode, struct file *filp)
+{
+	filp->private_data = kmalloc(128, GFP_KERNEL);
+	if (!filp->private_data)
+		return -ENOMEM;
+	return 0;
+}
+
+static int
+fault_release(struct inode *inode, struct file *filp)
+{
+	kfree(filp->private_data);
+	return 0;
+}
+
+static ssize_t
+fault_disconnect_read(struct file *filp, char __user *user_buf,
+		      size_t len, loff_t *offset)
+{
+	char *buffer = (char *)filp->private_data;
+	size_t size;
+
+	size = sprintf(buffer, "%u\n", rpc_inject_disconnect);
+	return simple_read_from_buffer(user_buf, len, offset, buffer, size);
+}
+
+static ssize_t
+fault_disconnect_write(struct file *filp, const char __user *user_buf,
+		       size_t len, loff_t *offset)
+{
+	char buffer[16];
+
+	if (len >= sizeof(buffer))
+		len = sizeof(buffer) - 1;
+	if (copy_from_user(buffer, user_buf, len))
+		return -EFAULT;
+	buffer[len] = '\0';
+	if (kstrtouint(buffer, 10, &rpc_inject_disconnect))
+		return -EINVAL;
+	return len;
+}
+
+static const struct file_operations fault_disconnect_fops = {
+	.owner		= THIS_MODULE,
+	.open		= fault_open,
+	.read		= fault_disconnect_read,
+	.write		= fault_disconnect_write,
+	.release	= fault_release,
+};
+
+static struct dentry *
+inject_fault_dir(struct dentry *topdir)
+{
+	struct dentry *faultdir;
+
+	faultdir = debugfs_create_dir("inject_fault", topdir);
+	if (!faultdir)
+		return NULL;
+
+	if (!debugfs_create_file("disconnect", S_IFREG | S_IRUSR, faultdir,
+				 NULL, &fault_disconnect_fops))
+		return NULL;
+
+	return faultdir;
+}
+
 void __exit
 sunrpc_debugfs_exit(void)
 {
 	debugfs_remove_recursive(topdir);
 	topdir = NULL;
+	rpc_fault_dir = NULL;
 	rpc_clnt_dir = NULL;
 	rpc_xprt_dir = NULL;
 }
@@ -282,6 +355,10 @@ sunrpc_debugfs_init(void)
 	if (!topdir)
 		return;
 
+	rpc_fault_dir = inject_fault_dir(topdir);
+	if (!rpc_fault_dir)
+		goto out_remove;
+
 	rpc_clnt_dir = debugfs_create_dir("rpc_clnt", topdir);
 	if (!rpc_clnt_dir)
 		goto out_remove;
@@ -294,5 +371,6 @@ sunrpc_debugfs_init(void)
 out_remove:
 	debugfs_remove_recursive(topdir);
 	topdir = NULL;
+	rpc_fault_dir = NULL;
 	rpc_clnt_dir = NULL;
 }
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index 852ae606b02a..5a16d8d8c831 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -1350,6 +1350,11 @@ bc_svc_process(struct svc_serv *serv, struct rpc_rqst *req,
 {
 	struct kvec	*argv = &rqstp->rq_arg.head[0];
 	struct kvec	*resv = &rqstp->rq_res.head[0];
+	struct rpc_task *task;
+	int proc_error;
+	int error;
+
+	dprintk("svc: %s(%p)\n", __func__, req);
 
 	/* Build the svc_rqst used by the common processing routine */
 	rqstp->rq_xprt = serv->sv_bc_xprt;
@@ -1372,21 +1377,36 @@ bc_svc_process(struct svc_serv *serv, struct rpc_rqst *req,
 
 	/*
 	 * Skip the next two words because they've already been
-	 * processed in the trasport
+	 * processed in the transport
 	 */
 	svc_getu32(argv);	/* XID */
 	svc_getnl(argv);	/* CALLDIR */
 
-	/* Returns 1 for send, 0 for drop */
-	if (svc_process_common(rqstp, argv, resv)) {
-		memcpy(&req->rq_snd_buf, &rqstp->rq_res,
-						sizeof(req->rq_snd_buf));
-		return bc_send(req);
-	} else {
-		/* drop request */
+	/* Parse and execute the bc call */
+	proc_error = svc_process_common(rqstp, argv, resv);
+
+	atomic_inc(&req->rq_xprt->bc_free_slots);
+	if (!proc_error) {
+		/* Processing error: drop the request */
 		xprt_free_bc_request(req);
 		return 0;
 	}
+
+	/* Finally, send the reply synchronously */
+	memcpy(&req->rq_snd_buf, &rqstp->rq_res, sizeof(req->rq_snd_buf));
+	task = rpc_run_bc_task(req);
+	if (IS_ERR(task)) {
+		error = PTR_ERR(task);
+		goto out;
+	}
+
+	WARN_ON_ONCE(atomic_read(&task->tk_count) != 1);
+	error = task->tk_status;
+	rpc_put_task(task);
+
+out:
+	dprintk("svc: %s(), error=%d\n", __func__, error);
+	return error;
 }
 EXPORT_SYMBOL_GPL(bc_svc_process);
 #endif /* CONFIG_SUNRPC_BACKCHANNEL */
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 1d4fe24af06a..ab5dd621ae0c 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -68,6 +68,7 @@ static void	 xprt_init(struct rpc_xprt *xprt, struct net *net);
 static void	xprt_request_init(struct rpc_task *, struct rpc_xprt *);
 static void	xprt_connect_status(struct rpc_task *task);
 static int      __xprt_get_cong(struct rpc_xprt *, struct rpc_task *);
+static void     __xprt_put_cong(struct rpc_xprt *, struct rpc_rqst *);
 static void	 xprt_destroy(struct rpc_xprt *xprt);
 
 static DEFINE_SPINLOCK(xprt_list_lock);
@@ -250,6 +251,8 @@ int xprt_reserve_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 	}
 	xprt_clear_locked(xprt);
 out_sleep:
+	if (req)
+		__xprt_put_cong(xprt, req);
 	dprintk("RPC: %5u failed to lock transport %p\n", task->tk_pid, xprt);
 	task->tk_timeout = 0;
 	task->tk_status = -EAGAIN;
@@ -608,8 +611,8 @@ static void xprt_autoclose(struct work_struct *work)
 	struct rpc_xprt *xprt =
 		container_of(work, struct rpc_xprt, task_cleanup);
 
-	xprt->ops->close(xprt);
 	clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
+	xprt->ops->close(xprt);
 	xprt_release_write(xprt, NULL);
 }
 
@@ -967,6 +970,7 @@ void xprt_transmit(struct rpc_task *task)
 		task->tk_status = status;
 		return;
 	}
+	xprt_inject_disconnect(xprt);
 
 	dprintk("RPC: %5u xmit complete\n", task->tk_pid);
 	task->tk_flags |= RPC_TASK_SENT;
@@ -1285,6 +1289,7 @@ void xprt_release(struct rpc_task *task)
 	spin_unlock_bh(&xprt->transport_lock);
 	if (req->rq_buffer)
 		xprt->ops->buf_free(req->rq_buffer);
+	xprt_inject_disconnect(xprt);
 	if (req->rq_cred != NULL)
 		put_rpccred(req->rq_cred);
 	task->tk_rqstp = NULL;
diff --git a/net/sunrpc/xprtrdma/fmr_ops.c b/net/sunrpc/xprtrdma/fmr_ops.c
index 302d4ebf6fbf..f1e8dafbd507 100644
--- a/net/sunrpc/xprtrdma/fmr_ops.c
+++ b/net/sunrpc/xprtrdma/fmr_ops.c
@@ -11,6 +11,21 @@
  * can take tens of usecs to complete.
  */
 
+/* Normal operation
+ *
+ * A Memory Region is prepared for RDMA READ or WRITE using the
+ * ib_map_phys_fmr verb (fmr_op_map). When the RDMA operation is
+ * finished, the Memory Region is unmapped using the ib_unmap_fmr
+ * verb (fmr_op_unmap).
+ */
+
+/* Transport recovery
+ *
+ * After a transport reconnect, fmr_op_map re-uses the MR already
+ * allocated for the RPC, but generates a fresh rkey then maps the
+ * MR again. This process is synchronous.
+ */
+
 #include "xprt_rdma.h"
 
 #if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
@@ -50,19 +65,28 @@ fmr_op_init(struct rpcrdma_xprt *r_xprt)
 	struct rpcrdma_mw *r;
 	int i, rc;
 
+	spin_lock_init(&buf->rb_mwlock);
 	INIT_LIST_HEAD(&buf->rb_mws);
 	INIT_LIST_HEAD(&buf->rb_all);
 
-	i = (buf->rb_max_requests + 1) * RPCRDMA_MAX_SEGS;
-	dprintk("RPC:       %s: initializing %d FMRs\n", __func__, i);
+	i = max_t(int, RPCRDMA_MAX_DATA_SEGS / RPCRDMA_MAX_FMR_SGES, 1);
+	i += 2;				/* head + tail */
+	i *= buf->rb_max_requests;	/* one set for each RPC slot */
+	dprintk("RPC:       %s: initalizing %d FMRs\n", __func__, i);
 
+	rc = -ENOMEM;
 	while (i--) {
 		r = kzalloc(sizeof(*r), GFP_KERNEL);
 		if (!r)
-			return -ENOMEM;
+			goto out;
 
-		r->r.fmr = ib_alloc_fmr(pd, mr_access_flags, &fmr_attr);
-		if (IS_ERR(r->r.fmr))
+		r->r.fmr.physaddrs = kmalloc(RPCRDMA_MAX_FMR_SGES *
+					     sizeof(u64), GFP_KERNEL);
+		if (!r->r.fmr.physaddrs)
+			goto out_free;
+
+		r->r.fmr.fmr = ib_alloc_fmr(pd, mr_access_flags, &fmr_attr);
+		if (IS_ERR(r->r.fmr.fmr))
 			goto out_fmr_err;
 
 		list_add(&r->mw_list, &buf->rb_mws);
@@ -71,12 +95,24 @@ fmr_op_init(struct rpcrdma_xprt *r_xprt)
 	return 0;
 
 out_fmr_err:
-	rc = PTR_ERR(r->r.fmr);
+	rc = PTR_ERR(r->r.fmr.fmr);
 	dprintk("RPC:       %s: ib_alloc_fmr status %i\n", __func__, rc);
+	kfree(r->r.fmr.physaddrs);
+out_free:
 	kfree(r);
+out:
 	return rc;
 }
 
+static int
+__fmr_unmap(struct rpcrdma_mw *r)
+{
+	LIST_HEAD(l);
+
+	list_add(&r->r.fmr.fmr->list, &l);
+	return ib_unmap_fmr(&l);
+}
+
 /* Use the ib_map_phys_fmr() verb to register a memory region
  * for remote access via RDMA READ or RDMA WRITE.
  */
@@ -85,12 +121,24 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 	   int nsegs, bool writing)
 {
 	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
-	struct ib_device *device = ia->ri_id->device;
+	struct ib_device *device = ia->ri_device;
 	enum dma_data_direction direction = rpcrdma_data_dir(writing);
 	struct rpcrdma_mr_seg *seg1 = seg;
-	struct rpcrdma_mw *mw = seg1->rl_mw;
-	u64 physaddrs[RPCRDMA_MAX_DATA_SEGS];
 	int len, pageoff, i, rc;
+	struct rpcrdma_mw *mw;
+
+	mw = seg1->rl_mw;
+	seg1->rl_mw = NULL;
+	if (!mw) {
+		mw = rpcrdma_get_mw(r_xprt);
+		if (!mw)
+			return -ENOMEM;
+	} else {
+		/* this is a retransmit; generate a fresh rkey */
+		rc = __fmr_unmap(mw);
+		if (rc)
+			return rc;
+	}
 
 	pageoff = offset_in_page(seg1->mr_offset);
 	seg1->mr_offset -= pageoff;	/* start of page */
@@ -100,7 +148,7 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 		nsegs = RPCRDMA_MAX_FMR_SGES;
 	for (i = 0; i < nsegs;) {
 		rpcrdma_map_one(device, seg, direction);
-		physaddrs[i] = seg->mr_dma;
+		mw->r.fmr.physaddrs[i] = seg->mr_dma;
 		len += seg->mr_len;
 		++seg;
 		++i;
@@ -110,11 +158,13 @@ fmr_op_map(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg,
 			break;
 	}
 
-	rc = ib_map_phys_fmr(mw->r.fmr, physaddrs, i, seg1->mr_dma);
+	rc = ib_map_phys_fmr(mw->r.fmr.fmr, mw->r.fmr.physaddrs,
+			     i, seg1->mr_dma);
 	if (rc)
 		goto out_maperr;
 
-	seg1->mr_rkey = mw->r.fmr->rkey;
+	seg1->rl_mw = mw;
+	seg1->mr_rkey = mw->r.fmr.fmr->rkey;
 	seg1->mr_base = seg1->mr_dma + pageoff;
 	seg1->mr_nsegs = i;
 	seg1->mr_len = len;
@@ -137,48 +187,28 @@ fmr_op_unmap(struct rpcrdma_xprt *r_xprt, struct rpcrdma_mr_seg *seg)
 {
 	struct rpcrdma_ia *ia = &r_xprt->rx_ia;
 	struct rpcrdma_mr_seg *seg1 = seg;
-	struct ib_device *device;
+	struct rpcrdma_mw *mw = seg1->rl_mw;
 	int rc, nsegs = seg->mr_nsegs;
-	LIST_HEAD(l);
 
-	list_add(&seg1->rl_mw->r.fmr->list, &l);
-	rc = ib_unmap_fmr(&l);
-	read_lock(&ia->ri_qplock);
-	device = ia->ri_id->device;
+	dprintk("RPC:       %s: FMR %p\n", __func__, mw);
+
+	seg1->rl_mw = NULL;
 	while (seg1->mr_nsegs--)
-		rpcrdma_unmap_one(device, seg++);
-	read_unlock(&ia->ri_qplock);
+		rpcrdma_unmap_one(ia->ri_device, seg++);
+	rc = __fmr_unmap(mw);
 	if (rc)
 		goto out_err;
+	rpcrdma_put_mw(r_xprt, mw);
 	return nsegs;
 
 out_err:
+	/* The FMR is abandoned, but remains in rb_all. fmr_op_destroy
+	 * will attempt to release it when the transport is destroyed.
+	 */
 	dprintk("RPC:       %s: ib_unmap_fmr status %i\n", __func__, rc);
 	return nsegs;
 }
 
-/* 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
-fmr_op_reset(struct rpcrdma_xprt *r_xprt)
-{
-	struct rpcrdma_buffer *buf = &r_xprt->rx_buf;
-	struct rpcrdma_mw *r;
-	LIST_HEAD(list);
-	int rc;
-
-	list_for_each_entry(r, &buf->rb_all, mw_all)
-		list_add(&r->r.fmr->list, &list);
-
-	rc = ib_unmap_fmr(&list);
-	if (rc)
-		dprintk("RPC:       %s: ib_unmap_fmr failed %i\n",
-			__func__, rc);
-}
-
 static void
 fmr_op_destroy(struct rpcrdma_buffer *buf)
 {
@@ -188,10 +218,13 @@ fmr_op_destroy(struct rpcrdma_buffer *buf)
 	while (!list_empty(&buf->rb_all)) {
 		r = list_entry(buf->rb_all.next, struct rpcrdma_mw, mw_all);
 		list_del(&r->mw_all);
-		rc = ib_dealloc_fmr(r->r.fmr);
+		kfree(r->r.fmr.physaddrs);
+
+		rc = ib_dealloc_fmr(r->r.fmr.fmr);
 		if (rc)
 			dprintk("RPC:       %s: ib_dealloc_fmr failed %i\n",
 				__func__, rc);
+
 		kfree(r);
 	}
 }
@@ -202,7 +235,6 @@ const struct rpcrdma_memreg_ops rpcrdma_fmr_memreg_ops = {
 	.ro_open			= fmr_op_open,
 	.ro_maxpages			= fmr_op_maxpages,
 	.ro_init			= fmr_op_init,
-	.ro_reset			= fmr_op_reset,
 	.ro_destroy			= fmr_op_destroy,
 	.ro_displayname			= "fmr",
 };
diff --git a/net/sunrpc/xprtrdma/frwr_ops.c b/net/sunrpc/xprtrdma/frwr_ops.c
index d234521320a4..04ea914201b2 100644
--- a/net/sunrpc/xprtrdma/frwr_ops.c
+++ b/net/sunrpc/xprtrdma/frwr_ops.c
@@ -11,12 +11,136 @@
  * but most complex memory registration mode.
  */