Merge tag 'nfs-for-3.13-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

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

   - Changes to the RPC socket code to allow NFSv4 to turn off
     timeout+retry:
      * Detect TCP connection breakage through the "keepalive" mechanism
   - Add client side support for NFSv4.x migration (Chuck Lever)
   - Add support for multiple security flavour arguments to the "sec="
     mount option (Dros Adamson)
   - fs-cache bugfixes from David Howells:
     * Fix an issue whereby caching can be enabled on a file that is
       open for writing
   - More NFSv4 open code stable bugfixes
   - Various Labeled NFS (selinux) bugfixes, including one stable fix
   - Fix buffer overflow checking in the RPCSEC_GSS upcall encoding"

* tag 'nfs-for-3.13-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (68 commits)
  NFSv4.2: Remove redundant checks in nfs_setsecurity+nfs4_label_init_security
  NFSv4: Sanity check the server reply in _nfs4_server_capabilities
  NFSv4.2: encode_readdir - only ask for labels when doing readdirplus
  nfs: set security label when revalidating inode
  NFSv4.2: Fix a mismatch between Linux labeled NFS and the NFSv4.2 spec
  NFS: Fix a missing initialisation when reading the SELinux label
  nfs: fix oops when trying to set SELinux label
  nfs: fix inverted test for delegation in nfs4_reclaim_open_state
  SUNRPC: Cleanup xs_destroy()
  SUNRPC: close a rare race in xs_tcp_setup_socket.
  SUNRPC: remove duplicated include from clnt.c
  nfs: use IS_ROOT not DCACHE_DISCONNECTED
  SUNRPC: Fix buffer overflow checking in gss_encode_v0_msg/gss_encode_v1_msg
  SUNRPC: gss_alloc_msg - choose _either_ a v0 message or a v1 message
  SUNRPC: remove an unnecessary if statement
  nfs: Use PTR_ERR_OR_ZERO in 'nfs/nfs4super.c'
  nfs: Use PTR_ERR_OR_ZERO in 'nfs41_callback_up' function
  nfs: Remove useless 'error' assignment
  sunrpc: comment typo fix
  SUNRPC: Add correct rcu_dereference annotation in rpc_clnt_set_transport
  ...

4 files changed, 234 insertions, 81 deletions

diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c
index 084656671d6e..97912b40c254 100644
--- a/net/sunrpc/auth_gss/auth_gss.c
+++ b/net/sunrpc/auth_gss/auth_gss.c
@@ -420,41 +420,53 @@ static void gss_encode_v0_msg(struct gss_upcall_msg *gss_msg)
 	memcpy(gss_msg->databuf, &uid, sizeof(uid));
 	gss_msg->msg.data = gss_msg->databuf;
 	gss_msg->msg.len = sizeof(uid);
-	BUG_ON(sizeof(uid) > UPCALL_BUF_LEN);
+
+	BUILD_BUG_ON(sizeof(uid) > sizeof(gss_msg->databuf));
 }
 
-static void gss_encode_v1_msg(struct gss_upcall_msg *gss_msg,
+static int gss_encode_v1_msg(struct gss_upcall_msg *gss_msg,
 				const char *service_name,
 				const char *target_name)
 {
 	struct gss_api_mech *mech = gss_msg->auth->mech;
 	char *p = gss_msg->databuf;
-	int len = 0;
-
-	gss_msg->msg.len = sprintf(gss_msg->databuf, "mech=%s uid=%d ",
-				   mech->gm_name,
-				   from_kuid(&init_user_ns, gss_msg->uid));
-	p += gss_msg->msg.len;
+	size_t buflen = sizeof(gss_msg->databuf);
+	int len;
+
+	len = scnprintf(p, buflen, "mech=%s uid=%d ", mech->gm_name,
+			from_kuid(&init_user_ns, gss_msg->uid));
+	buflen -= len;
+	p += len;
+	gss_msg->msg.len = len;
 	if (target_name) {
-		len = sprintf(p, "target=%s ", target_name);
+		len = scnprintf(p, buflen, "target=%s ", target_name);
+		buflen -= len;
 		p += len;
 		gss_msg->msg.len += len;
 	}
 	if (service_name != NULL) {
-		len = sprintf(p, "service=%s ", service_name);
+		len = scnprintf(p, buflen, "service=%s ", service_name);
+		buflen -= len;
 		p += len;
 		gss_msg->msg.len += len;
 	}
 	if (mech->gm_upcall_enctypes) {
-		len = sprintf(p, "enctypes=%s ", mech->gm_upcall_enctypes);
+		len = scnprintf(p, buflen, "enctypes=%s ",
+				mech->gm_upcall_enctypes);
+		buflen -= len;
 		p += len;
 		gss_msg->msg.len += len;
 	}
-	len = sprintf(p, "\n");
+	len = scnprintf(p, buflen, "\n");
+	if (len == 0)
+		goto out_overflow;
 	gss_msg->msg.len += len;
 
 	gss_msg->msg.data = gss_msg->databuf;
-	BUG_ON(gss_msg->msg.len > UPCALL_BUF_LEN);
+	return 0;
+out_overflow:
+	WARN_ON_ONCE(1);
+	return -ENOMEM;
 }
 
 static struct gss_upcall_msg *
@@ -463,15 +475,15 @@ gss_alloc_msg(struct gss_auth *gss_auth,
 {
 	struct gss_upcall_msg *gss_msg;
 	int vers;
+	int err = -ENOMEM;
 
 	gss_msg = kzalloc(sizeof(*gss_msg), GFP_NOFS);
 	if (gss_msg == NULL)
-		return ERR_PTR(-ENOMEM);
+		goto err;
 	vers = get_pipe_version(gss_auth->net);
-	if (vers < 0) {
-		kfree(gss_msg);
-		return ERR_PTR(vers);
-	}
+	err = vers;
+	if (err < 0)
+		goto err_free_msg;
 	gss_msg->pipe = gss_auth->gss_pipe[vers]->pipe;
 	INIT_LIST_HEAD(&gss_msg->list);
 	rpc_init_wait_queue(&gss_msg->rpc_waitqueue, "RPCSEC_GSS upcall waitq");
@@ -482,10 +494,17 @@ gss_alloc_msg(struct gss_auth *gss_auth,
 	switch (vers) {
 	case 0:
 		gss_encode_v0_msg(gss_msg);
+		break;
 	default:
-		gss_encode_v1_msg(gss_msg, service_name, gss_auth->target_name);
+		err = gss_encode_v1_msg(gss_msg, service_name, gss_auth->target_name);
+		if (err)
+			goto err_free_msg;
 	};
 	return gss_msg;
+err_free_msg:
+	kfree(gss_msg);
+err:
+	return ERR_PTR(err);
 }
 
 static struct gss_upcall_msg *
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 77479606a971..dab09dac8fc7 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -25,12 +25,12 @@
 #include <linux/namei.h>
 #include <linux/mount.h>
 #include <linux/slab.h>
+#include <linux/rcupdate.h>
 #include <linux/utsname.h>
 #include <linux/workqueue.h>
 #include <linux/in.h>
 #include <linux/in6.h>
 #include <linux/un.h>
-#include <linux/rcupdate.h>
 
 #include <linux/sunrpc/clnt.h>
 #include <linux/sunrpc/addr.h>
@@ -264,6 +264,26 @@ void rpc_clients_notifier_unregister(void)
 	return rpc_pipefs_notifier_unregister(&rpc_clients_block);
 }
 
+static struct rpc_xprt *rpc_clnt_set_transport(struct rpc_clnt *clnt,
+		struct rpc_xprt *xprt,
+		const struct rpc_timeout *timeout)
+{
+	struct rpc_xprt *old;
+
+	spin_lock(&clnt->cl_lock);
+	old = rcu_dereference_protected(clnt->cl_xprt,
+			lockdep_is_held(&clnt->cl_lock));
+
+	if (!xprt_bound(xprt))
+		clnt->cl_autobind = 1;
+
+	clnt->cl_timeout = timeout;
+	rcu_assign_pointer(clnt->cl_xprt, xprt);
+	spin_unlock(&clnt->cl_lock);
+
+	return old;
+}
+
 static void rpc_clnt_set_nodename(struct rpc_clnt *clnt, const char *nodename)
 {
 	clnt->cl_nodelen = strlen(nodename);
@@ -272,12 +292,13 @@ static void rpc_clnt_set_nodename(struct rpc_clnt *clnt, const char *nodename)
 	memcpy(clnt->cl_nodename, nodename, clnt->cl_nodelen);
 }
 
-static int rpc_client_register(const struct rpc_create_args *args,
-			       struct rpc_clnt *clnt)
+static int rpc_client_register(struct rpc_clnt *clnt,
+			       rpc_authflavor_t pseudoflavor,
+			       const char *client_name)
 {
 	struct rpc_auth_create_args auth_args = {
-		.pseudoflavor = args->authflavor,
-		.target_name = args->client_name,
+		.pseudoflavor = pseudoflavor,
+		.target_name = client_name,
 	};
 	struct rpc_auth *auth;
 	struct net *net = rpc_net_ns(clnt);
@@ -298,7 +319,7 @@ static int rpc_client_register(const struct rpc_create_args *args,
 	auth = rpcauth_create(&auth_args, clnt);
 	if (IS_ERR(auth)) {
 		dprintk("RPC:       Couldn't create auth handle (flavor %u)\n",
-				args->authflavor);
+				pseudoflavor);
 		err = PTR_ERR(auth);
 		goto err_auth;
 	}
@@ -337,7 +358,8 @@ static struct rpc_clnt * rpc_new_client(const struct rpc_create_args *args,
 {
 	const struct rpc_program *program = args->program;
 	const struct rpc_version *version;
-	struct rpc_clnt		*clnt = NULL;
+	struct rpc_clnt *clnt = NULL;
+	const struct rpc_timeout *timeout;
 	int err;
 
 	/* sanity check the name before trying to print it */
@@ -365,7 +387,6 @@ static struct rpc_clnt * rpc_new_client(const struct rpc_create_args *args,
 	if (err)
 		goto out_no_clid;
 
-	rcu_assign_pointer(clnt->cl_xprt, xprt);
 	clnt->cl_procinfo = version->procs;
 	clnt->cl_maxproc  = version->nrprocs;
 	clnt->cl_prog     = args->prognumber ? : program->number;
@@ -380,16 +401,15 @@ static struct rpc_clnt * rpc_new_client(const struct rpc_create_args *args,
 	INIT_LIST_HEAD(&clnt->cl_tasks);
 	spin_lock_init(&clnt->cl_lock);
 
-	if (!xprt_bound(xprt))
-		clnt->cl_autobind = 1;
-
-	clnt->cl_timeout = xprt->timeout;
+	timeout = xprt->timeout;
 	if (args->timeout != NULL) {
 		memcpy(&clnt->cl_timeout_default, args->timeout,
 				sizeof(clnt->cl_timeout_default));
-		clnt->cl_timeout = &clnt->cl_timeout_default;
+		timeout = &clnt->cl_timeout_default;
 	}
 
+	rpc_clnt_set_transport(clnt, xprt, timeout);
+
 	clnt->cl_rtt = &clnt->cl_rtt_default;
 	rpc_init_rtt(&clnt->cl_rtt_default, clnt->cl_timeout->to_initval);
 
@@ -398,7 +418,7 @@ static struct rpc_clnt * rpc_new_client(const struct rpc_create_args *args,
 	/* save the nodename */
 	rpc_clnt_set_nodename(clnt, utsname()->nodename);
 
-	err = rpc_client_register(args, clnt);
+	err = rpc_client_register(clnt, args->authflavor, args->client_name);
 	if (err)
 		goto out_no_path;
 	if (parent)
@@ -600,6 +620,80 @@ rpc_clone_client_set_auth(struct rpc_clnt *clnt, rpc_authflavor_t flavor)
 }
 EXPORT_SYMBOL_GPL(rpc_clone_client_set_auth);
 
+/**
+ * rpc_switch_client_transport: switch the RPC transport on the fly
+ * @clnt: pointer to a struct rpc_clnt
+ * @args: pointer to the new transport arguments
+ * @timeout: pointer to the new timeout parameters
+ *
+ * This function allows the caller to switch the RPC transport for the
+ * rpc_clnt structure 'clnt' to allow it to connect to a mirrored NFS
+ * server, for instance.  It assumes that the caller has ensured that
+ * there are no active RPC tasks by using some form of locking.
+ *
+ * Returns zero if "clnt" is now using the new xprt.  Otherwise a
+ * negative errno is returned, and "clnt" continues to use the old
+ * xprt.
+ */
+int rpc_switch_client_transport(struct rpc_clnt *clnt,
+		struct xprt_create *args,
+		const struct rpc_timeout *timeout)
+{
+	const struct rpc_timeout *old_timeo;
+	rpc_authflavor_t pseudoflavor;
+	struct rpc_xprt *xprt, *old;
+	struct rpc_clnt *parent;
+	int err;
+
+	xprt = xprt_create_transport(args);
+	if (IS_ERR(xprt)) {
+		dprintk("RPC:       failed to create new xprt for clnt %p\n",
+			clnt);
+		return PTR_ERR(xprt);
+	}
+
+	pseudoflavor = clnt->cl_auth->au_flavor;
+
+	old_timeo = clnt->cl_timeout;
+	old = rpc_clnt_set_transport(clnt, xprt, timeout);
+
+	rpc_unregister_client(clnt);
+	__rpc_clnt_remove_pipedir(clnt);
+
+	/*
+	 * A new transport was created.  "clnt" therefore
+	 * becomes the root of a new cl_parent tree.  clnt's
+	 * children, if it has any, still point to the old xprt.
+	 */
+	parent = clnt->cl_parent;
+	clnt->cl_parent = clnt;
+
+	/*
+	 * The old rpc_auth cache cannot be re-used.  GSS
+	 * contexts in particular are between a single
+	 * client and server.
+	 */
+	err = rpc_client_register(clnt, pseudoflavor, NULL);
+	if (err)
+		goto out_revert;
+
+	synchronize_rcu();
+	if (parent != clnt)
+		rpc_release_client(parent);
+	xprt_put(old);
+	dprintk("RPC:       replaced xprt for clnt %p\n", clnt);
+	return 0;
+
+out_revert:
+	rpc_clnt_set_transport(clnt, old, old_timeo);
+	clnt->cl_parent = parent;
+	rpc_client_register(clnt, pseudoflavor, NULL);
+	xprt_put(xprt);
+	dprintk("RPC:       failed to switch xprt for clnt %p\n", clnt);
+	return err;
+}
+EXPORT_SYMBOL_GPL(rpc_switch_client_transport);
+
 /*
  * Kill all tasks for the given client.
  * XXX: kill their descendants as well?
@@ -772,6 +866,8 @@ void rpc_task_set_client(struct rpc_task *task, struct rpc_clnt *clnt)
 		atomic_inc(&clnt->cl_count);
 		if (clnt->cl_softrtry)
 			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;
 
@@ -1690,6 +1786,7 @@ call_connect_status(struct rpc_task *task)
 	dprint_status(task);
 
 	trace_rpc_connect_status(task, status);
+	task->tk_status = 0;
 	switch (status) {
 		/* if soft mounted, test if we've timed out */
 	case -ETIMEDOUT:
@@ -1698,12 +1795,14 @@ call_connect_status(struct rpc_task *task)
 	case -ECONNREFUSED:
 	case -ECONNRESET:
 	case -ENETUNREACH:
+		/* retry with existing socket, after a delay */
+		rpc_delay(task, 3*HZ);
 		if (RPC_IS_SOFTCONN(task))
 			break;
-		/* retry with existing socket, after a delay */
-	case 0:
 	case -EAGAIN:
-		task->tk_status = 0;
+		task->tk_action = call_bind;
+		return;
+	case 0:
 		clnt->cl_stats->netreconn++;
 		task->tk_action = call_transmit;
 		return;
@@ -1717,13 +1816,14 @@ call_connect_status(struct rpc_task *task)
 static void
 call_transmit(struct rpc_task *task)
 {
+	int is_retrans = RPC_WAS_SENT(task);
+
 	dprint_status(task);
 
 	task->tk_action = call_status;
 	if (task->tk_status < 0)
 		return;
-	task->tk_status = xprt_prepare_transmit(task);
-	if (task->tk_status != 0)
+	if (!xprt_prepare_transmit(task))
 		return;
 	task->tk_action = call_transmit_status;
 	/* Encode here so that rpcsec_gss can use correct sequence number. */
@@ -1742,6 +1842,8 @@ call_transmit(struct rpc_task *task)
 	xprt_transmit(task);
 	if (task->tk_status < 0)
 		return;
+	if (is_retrans)
+		task->tk_client->cl_stats->rpcretrans++;
 	/*
 	 * On success, ensure that we call xprt_end_transmit() before sleeping
 	 * in order to allow access to the socket to other RPC requests.
@@ -1811,8 +1913,7 @@ call_bc_transmit(struct rpc_task *task)
 {
 	struct rpc_rqst *req = task->tk_rqstp;
 
-	task->tk_status = xprt_prepare_transmit(task);
-	if (task->tk_status == -EAGAIN) {
+	if (!xprt_prepare_transmit(task)) {
 		/*
 		 * Could not reserve the transport. Try again after the
 		 * transport is released.
@@ -1900,7 +2001,8 @@ call_status(struct rpc_task *task)
 		rpc_delay(task, 3*HZ);
 	case -ETIMEDOUT:
 		task->tk_action = call_timeout;
-		if (task->tk_client->cl_discrtry)
+		if (!(task->tk_flags & RPC_TASK_NO_RETRANS_TIMEOUT)
+		    && task->tk_client->cl_discrtry)
 			xprt_conditional_disconnect(req->rq_xprt,
 					req->rq_connect_cookie);
 		break;
@@ -1982,7 +2084,6 @@ call_timeout(struct rpc_task *task)
 	rpcauth_invalcred(task);
 
 retry:
-	clnt->cl_stats->rpcretrans++;
 	task->tk_action = call_bind;
 	task->tk_status = 0;
 }
@@ -2025,7 +2126,6 @@ call_decode(struct rpc_task *task)
 	if (req->rq_rcv_buf.len < 12) {
 		if (!RPC_IS_SOFT(task)) {
 			task->tk_action = call_bind;
-			clnt->cl_stats->rpcretrans++;
 			goto out_retry;
 		}
 		dprintk("RPC:       %s: too small RPC reply size (%d bytes)\n",
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 095363eee764..04199bc8416f 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -205,10 +205,8 @@ int xprt_reserve_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
 		goto out_sleep;
 	}
 	xprt->snd_task = task;
-	if (req != NULL) {
-		req->rq_bytes_sent = 0;
+	if (req != NULL)
 		req->rq_ntrans++;
-	}
 
 	return 1;
 
@@ -263,7 +261,6 @@ int xprt_reserve_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 	}
 	if (__xprt_get_cong(xprt, task)) {
 		xprt->snd_task = task;
-		req->rq_bytes_sent = 0;
 		req->rq_ntrans++;
 		return 1;
 	}
@@ -300,10 +297,8 @@ static bool __xprt_lock_write_func(struct rpc_task *task, void *data)
 
 	req = task->tk_rqstp;
 	xprt->snd_task = task;
-	if (req) {
-		req->rq_bytes_sent = 0;
+	if (req)
 		req->rq_ntrans++;
-	}
 	return true;
 }
 
@@ -329,7 +324,6 @@ static bool __xprt_lock_write_cong_func(struct rpc_task *task, void *data)
 	}
 	if (__xprt_get_cong(xprt, task)) {
 		xprt->snd_task = task;
-		req->rq_bytes_sent = 0;
 		req->rq_ntrans++;
 		return true;
 	}
@@ -358,6 +352,11 @@ out_unlock:
 void xprt_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
 {
 	if (xprt->snd_task == task) {
+		if (task != NULL) {
+			struct rpc_rqst *req = task->tk_rqstp;
+			if (req != NULL)
+				req->rq_bytes_sent = 0;
+		}
 		xprt_clear_locked(xprt);
 		__xprt_lock_write_next(xprt);
 	}
@@ -375,6 +374,11 @@ EXPORT_SYMBOL_GPL(xprt_release_xprt);
 void xprt_release_xprt_cong(struct rpc_xprt *xprt, struct rpc_task *task)
 {
 	if (xprt->snd_task == task) {
+		if (task != NULL) {
+			struct rpc_rqst *req = task->tk_rqstp;
+			if (req != NULL)
+				req->rq_bytes_sent = 0;
+		}
 		xprt_clear_locked(xprt);
 		__xprt_lock_write_next_cong(xprt);
 	}
@@ -854,24 +858,36 @@ static inline int xprt_has_timer(struct rpc_xprt *xprt)
  * @task: RPC task about to send a request
  *
  */
-int xprt_prepare_transmit(struct rpc_task *task)
+bool xprt_prepare_transmit(struct rpc_task *task)
 {
 	struct rpc_rqst	*req = task->tk_rqstp;
 	struct rpc_xprt	*xprt = req->rq_xprt;
-	int err = 0;
+	bool ret = false;
 
 	dprintk("RPC: %5u xprt_prepare_transmit\n", task->tk_pid);
 
 	spin_lock_bh(&xprt->transport_lock);
-	if (req->rq_reply_bytes_recvd && !req->rq_bytes_sent) {
-		err = req->rq_reply_bytes_recvd;
+	if (!req->rq_bytes_sent) {
+		if (req->rq_reply_bytes_recvd) {
+			task->tk_status = req->rq_reply_bytes_recvd;
+			goto out_unlock;
+		}
+		if ((task->tk_flags & RPC_TASK_NO_RETRANS_TIMEOUT)
+		    && xprt_connected(xprt)
+		    && req->rq_connect_cookie == xprt->connect_cookie) {
+			xprt->ops->set_retrans_timeout(task);
+			rpc_sleep_on(&xprt->pending, task, xprt_timer);
+			goto out_unlock;
+		}
+	}
+	if (!xprt->ops->reserve_xprt(xprt, task)) {
+		task->tk_status = -EAGAIN;
 		goto out_unlock;
 	}
-	if (!xprt->ops->reserve_xprt(xprt, task))
-		err = -EAGAIN;
+	ret = true;
 out_unlock:
 	spin_unlock_bh(&xprt->transport_lock);
-	return err;
+	return ret;
 }
 
 void xprt_end_transmit(struct rpc_task *task)
@@ -912,7 +928,6 @@ void xprt_transmit(struct rpc_task *task)
 	} else if (!req->rq_bytes_sent)
 		return;
 
-	req->rq_connect_cookie = xprt->connect_cookie;
 	req->rq_xtime = ktime_get();
 	status = xprt->ops->send_request(task);
 	if (status != 0) {
@@ -938,12 +953,14 @@ void xprt_transmit(struct rpc_task *task)
 	/* Don't race with disconnect */
 	if (!xprt_connected(xprt))
 		task->tk_status = -ENOTCONN;
-	else if (!req->rq_reply_bytes_recvd && rpc_reply_expected(task)) {
+	else {
 		/*
 		 * Sleep on the pending queue since
 		 * we're expecting a reply.
 		 */
-		rpc_sleep_on(&xprt->pending, task, xprt_timer);
+		if (!req->rq_reply_bytes_recvd && rpc_reply_expected(task))
+			rpc_sleep_on(&xprt->pending, task, xprt_timer);
+		req->rq_connect_cookie = xprt->connect_cookie;
 	}
 	spin_unlock_bh(&xprt->transport_lock);
 }
@@ -1087,11 +1104,9 @@ struct rpc_xprt *xprt_alloc(struct net *net, size_t size,
 	for (i = 0; i < num_prealloc; i++) {
 		req = kzalloc(sizeof(struct rpc_rqst), GFP_KERNEL);
 		if (!req)
-			break;
+			goto out_free;
 		list_add(&req->rq_list, &xprt->free);
 	}
-	if (i < num_prealloc)
-		goto out_free;
 	if (max_alloc > num_prealloc)
 		xprt->max_reqs = max_alloc;
 	else
@@ -1186,6 +1201,12 @@ static void xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
 	req->rq_xprt    = xprt;
 	req->rq_buffer  = NULL;
 	req->rq_xid     = xprt_alloc_xid(xprt);
+	req->rq_connect_cookie = xprt->connect_cookie - 1;
+	req->rq_bytes_sent = 0;
+	req->rq_snd_buf.len = 0;
+	req->rq_snd_buf.buflen = 0;
+	req->rq_rcv_buf.len = 0;
+	req->rq_rcv_buf.buflen = 0;
 	req->rq_release_snd_buf = NULL;
 	xprt_reset_majortimeo(req);
 	dprintk("RPC: %5u reserved req %p xid %08x\n", task->tk_pid,
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index ee03d35677d9..17c88928b7db 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -835,6 +835,8 @@ static void xs_close(struct rpc_xprt *xprt)
 
 	dprintk("RPC:       xs_close xprt %p\n", xprt);
 
+	cancel_delayed_work_sync(&transport->connect_worker);
+
 	xs_reset_transport(transport);
 	xprt->reestablish_timeout = 0;
 
@@ -854,14 +856,6 @@ static void xs_tcp_close(struct rpc_xprt *xprt)
 		xs_tcp_shutdown(xprt);
 }
 
-static void xs_local_destroy(struct rpc_xprt *xprt)
-{
-	xs_close(xprt);
-	xs_free_peer_addresses(xprt);
-	xprt_free(xprt);
-	module_put(THIS_MODULE);
-}
-
 /**
  * xs_destroy - prepare to shutdown a transport
  * @xprt: doomed transport
@@ -869,13 +863,12 @@ static void xs_local_destroy(struct rpc_xprt *xprt)
  */
 static void xs_destroy(struct rpc_xprt *xprt)
 {
-	struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
-
 	dprintk("RPC:       xs_destroy xprt %p\n", xprt);
 
-	cancel_delayed_work_sync(&transport->connect_worker);
-
-	xs_local_destroy(xprt);
+	xs_close(xprt);
+	xs_free_peer_addresses(xprt);
+	xprt_free(xprt);
+	module_put(THIS_MODULE);
 }
 
 static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
@@ -1511,6 +1504,7 @@ static void xs_tcp_state_change(struct sock *sk)
 			transport->tcp_copied = 0;
 			transport->tcp_flags =
 				TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID;
+			xprt->connect_cookie++;
 
 			xprt_wake_pending_tasks(xprt, -EAGAIN);
 		}
@@ -1816,6 +1810,10 @@ static inline void xs_reclassify_socket(int family, struct socket *sock)
 }
 #endif
 
+static void xs_dummy_setup_socket(struct work_struct *work)
+{
+}
+
 static struct socket *xs_create_sock(struct rpc_xprt *xprt,
 		struct sock_xprt *transport, int family, int type, int protocol)
 {
@@ -2112,6 +2110,19 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
 
 	if (!transport->inet) {
 		struct sock *sk = sock->sk;
+		unsigned int keepidle = xprt->timeout->to_initval / HZ;
+		unsigned int keepcnt = xprt->timeout->to_retries + 1;
+		unsigned int opt_on = 1;
+
+		/* TCP Keepalive options */
+		kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
+				(char *)&opt_on, sizeof(opt_on));
+		kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
+				(char *)&keepidle, sizeof(keepidle));
+		kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
+				(char *)&keepidle, sizeof(keepidle));
+		kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
+				(char *)&keepcnt, sizeof(keepcnt));
 
 		write_lock_bh(&sk->sk_callback_lock);
 
@@ -2151,7 +2162,6 @@ static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
 	case 0:
 	case -EINPROGRESS:
 		/* SYN_SENT! */
-		xprt->connect_cookie++;
 		if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
 			xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
 	}
@@ -2498,7 +2508,7 @@ static struct rpc_xprt_ops xs_local_ops = {
 	.send_request		= xs_local_send_request,
 	.set_retrans_timeout	= xprt_set_retrans_timeout_def,
 	.close			= xs_close,
-	.destroy		= xs_local_destroy,
+	.destroy		= xs_destroy,
 	.print_stats		= xs_local_print_stats,
 };
 
@@ -2655,6 +2665,9 @@ static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
 	xprt->ops = &xs_local_ops;
 	xprt->timeout = &xs_local_default_timeout;
 
+	INIT_DELAYED_WORK(&transport->connect_worker,
+			xs_dummy_setup_socket);
+
 	switch (sun->sun_family) {
 	case AF_LOCAL:
 		if (sun->sun_path[0] != '/') {
@@ -2859,8 +2872,8 @@ static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
 	if (args->bc_xprt->xpt_bc_xprt) {
 		/*
 		 * This server connection already has a backchannel
-		 * export; we can't create a new one, as we wouldn't be
-		 * able to match replies based on xid any more.  So,
+		 * transport; we can't create a new one, as we wouldn't
+		 * be able to match replies based on xid any more.  So,
 		 * reuse the already-existing one:
 		 */
 		 return args->bc_xprt->xpt_bc_xprt;