Merge branch 'for-2.6.38' of git://linux-nfs.org/~bfields/linux

* 'for-2.6.38' of git://linux-nfs.org/~bfields/linux: (62 commits)
  nfsd4: fix callback restarting
  nfsd: break lease on unlink, link, and rename
  nfsd4: break lease on nfsd setattr
  nfsd: don't support msnfs export option
  nfsd4: initialize cb_per_client
  nfsd4: allow restarting callbacks
  nfsd4: simplify nfsd4_cb_prepare
  nfsd4: give out delegations more quickly in 4.1 case
  nfsd4: add helper function to run callbacks
  nfsd4: make sure sequence flags are set after destroy_session
  nfsd4: re-probe callback on connection loss
  nfsd4: set sequence flag when backchannel is down
  nfsd4: keep finer-grained callback status
  rpc: allow xprt_class->setup to return a preexisting xprt
  rpc: keep backchannel xprt as long as server connection
  rpc: move sk_bc_xprt to svc_xprt
  nfsd4: allow backchannel recovery
  nfsd4: support BIND_CONN_TO_SESSION
  nfsd4: modify session list under cl_lock
  Documentation: fl_mylease no longer exists
  ...

Fix up conflicts in fs/nfsd/vfs.c with the vfs-scale work.  The
vfs-scale work touched some msnfs cases, and this merge removes support
for that entirely, so the conflict was trivial to resolve.

10 files changed, 141 insertions, 94 deletions

diff --git a/net/sunrpc/auth_gss/gss_krb5_crypto.c b/net/sunrpc/auth_gss/gss_krb5_crypto.c
index 75ee993ea057..9576f35ab701 100644
--- a/net/sunrpc/auth_gss/gss_krb5_crypto.c
+++ b/net/sunrpc/auth_gss/gss_krb5_crypto.c
@@ -137,7 +137,7 @@ arcfour_hmac_md5_usage_to_salt(unsigned int usage, u8 salt[4])
 		ms_usage = 13;
 		break;
 	default:
-		return EINVAL;;
+		return -EINVAL;
 	}
 	salt[0] = (ms_usage >> 0) & 0xff;
 	salt[1] = (ms_usage >> 8) & 0xff;
diff --git a/net/sunrpc/auth_gss/svcauth_gss.c b/net/sunrpc/auth_gss/svcauth_gss.c
index dec2a6fc7c12..bcdae78fdfc6 100644
--- a/net/sunrpc/auth_gss/svcauth_gss.c
+++ b/net/sunrpc/auth_gss/svcauth_gss.c
@@ -67,7 +67,6 @@ static int netobj_equal(struct xdr_netobj *a, struct xdr_netobj *b)
 
 #define	RSI_HASHBITS	6
 #define	RSI_HASHMAX	(1<<RSI_HASHBITS)
-#define	RSI_HASHMASK	(RSI_HASHMAX-1)
 
 struct rsi {
 	struct cache_head	h;
@@ -319,7 +318,6 @@ static struct rsi *rsi_update(struct rsi *new, struct rsi *old)
 
 #define	RSC_HASHBITS	10
 #define	RSC_HASHMAX	(1<<RSC_HASHBITS)
-#define	RSC_HASHMASK	(RSC_HASHMAX-1)
 
 #define GSS_SEQ_WIN	128
 
diff --git a/net/sunrpc/cache.c b/net/sunrpc/cache.c
index e433e7580e27..72ad836e4fe0 100644
--- a/net/sunrpc/cache.c
+++ b/net/sunrpc/cache.c
@@ -37,7 +37,7 @@
 
 #define	 RPCDBG_FACILITY RPCDBG_CACHE
 
-static void cache_defer_req(struct cache_req *req, struct cache_head *item);
+static bool cache_defer_req(struct cache_req *req, struct cache_head *item);
 static void cache_revisit_request(struct cache_head *item);
 
 static void cache_init(struct cache_head *h)
@@ -128,6 +128,7 @@ static void cache_fresh_locked(struct cache_head *head, time_t expiry)
 {
 	head->expiry_time = expiry;
 	head->last_refresh = seconds_since_boot();
+	smp_wmb(); /* paired with smp_rmb() in cache_is_valid() */
 	set_bit(CACHE_VALID, &head->flags);
 }
 
@@ -208,11 +209,36 @@ static inline int cache_is_valid(struct cache_detail *detail, struct cache_head
 		/* entry is valid */
 		if (test_bit(CACHE_NEGATIVE, &h->flags))
 			return -ENOENT;
-		else
+		else {
+			/*
+			 * In combination with write barrier in
+			 * sunrpc_cache_update, ensures that anyone
+			 * using the cache entry after this sees the
+			 * updated contents:
+			 */
+			smp_rmb();
 			return 0;
+		}
 	}
 }
 
+static int try_to_negate_entry(struct cache_detail *detail, struct cache_head *h)
+{
+	int rv;
+
+	write_lock(&detail->hash_lock);
+	rv = cache_is_valid(detail, h);
+	if (rv != -EAGAIN) {
+		write_unlock(&detail->hash_lock);
+		return rv;
+	}
+	set_bit(CACHE_NEGATIVE, &h->flags);
+	cache_fresh_locked(h, seconds_since_boot()+CACHE_NEW_EXPIRY);
+	write_unlock(&detail->hash_lock);
+	cache_fresh_unlocked(h, detail);
+	return -ENOENT;
+}
+
 /*
  * This is the generic cache management routine for all
  * the authentication caches.
@@ -251,14 +277,8 @@ int cache_check(struct cache_detail *detail,
 			case -EINVAL:
 				clear_bit(CACHE_PENDING, &h->flags);
 				cache_revisit_request(h);
-				if (rv == -EAGAIN) {
-					set_bit(CACHE_NEGATIVE, &h->flags);
-					cache_fresh_locked(h, seconds_since_boot()+CACHE_NEW_EXPIRY);
-					cache_fresh_unlocked(h, detail);
-					rv = -ENOENT;
-				}
+				rv = try_to_negate_entry(detail, h);
 				break;
-
 			case -EAGAIN:
 				clear_bit(CACHE_PENDING, &h->flags);
 				cache_revisit_request(h);
@@ -268,9 +288,11 @@ int cache_check(struct cache_detail *detail,
 	}
 
 	if (rv == -EAGAIN) {
-		cache_defer_req(rqstp, h);
-		if (!test_bit(CACHE_PENDING, &h->flags)) {
-			/* Request is not deferred */
+		if (!cache_defer_req(rqstp, h)) {
+			/*
+			 * Request was not deferred; handle it as best
+			 * we can ourselves:
+			 */
 			rv = cache_is_valid(detail, h);
 			if (rv == -EAGAIN)
 				rv = -ETIMEDOUT;
@@ -618,18 +640,19 @@ static void cache_limit_defers(void)
 		discard->revisit(discard, 1);
 }
 
-static void cache_defer_req(struct cache_req *req, struct cache_head *item)
+/* Return true if and only if a deferred request is queued. */
+static bool cache_defer_req(struct cache_req *req, struct cache_head *item)
 {
 	struct cache_deferred_req *dreq;
 
 	if (req->thread_wait) {
 		cache_wait_req(req, item);
 		if (!test_bit(CACHE_PENDING, &item->flags))
-			return;
+			return false;
 	}
 	dreq = req->defer(req);
 	if (dreq == NULL)
-		return;
+		return false;
 	setup_deferral(dreq, item, 1);
 	if (!test_bit(CACHE_PENDING, &item->flags))
 		/* Bit could have been cleared before we managed to
@@ -638,6 +661,7 @@ static void cache_defer_req(struct cache_req *req, struct cache_head *item)
 		cache_revisit_request(item);
 
 	cache_limit_defers();
+	return true;
 }
 
 static void cache_revisit_request(struct cache_head *item)
diff --git a/net/sunrpc/svc.c b/net/sunrpc/svc.c
index 0e659c665a8d..08e05a8ce025 100644
--- a/net/sunrpc/svc.c
+++ b/net/sunrpc/svc.c
@@ -1001,6 +1001,7 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 	rqstp->rq_splice_ok = 1;
 	/* Will be turned off only when NFSv4 Sessions are used */
 	rqstp->rq_usedeferral = 1;
+	rqstp->rq_dropme = false;
 
 	/* Setup reply header */
 	rqstp->rq_xprt->xpt_ops->xpo_prep_reply_hdr(rqstp);
@@ -1102,7 +1103,7 @@ svc_process_common(struct svc_rqst *rqstp, struct kvec *argv, struct kvec *resv)
 		*statp = procp->pc_func(rqstp, rqstp->rq_argp, rqstp->rq_resp);
 
 		/* Encode reply */
-		if (*statp == rpc_drop_reply) {
+		if (rqstp->rq_dropme) {
 			if (procp->pc_release)
 				procp->pc_release(rqstp, NULL, rqstp->rq_resp);
 			goto dropit;
diff --git a/net/sunrpc/svc_xprt.c b/net/sunrpc/svc_xprt.c
index 3f2c5559ca1a..ab86b7927f84 100644
--- a/net/sunrpc/svc_xprt.c
+++ b/net/sunrpc/svc_xprt.c
@@ -13,6 +13,7 @@
 #include <linux/sunrpc/stats.h>
 #include <linux/sunrpc/svc_xprt.h>
 #include <linux/sunrpc/svcsock.h>
+#include <linux/sunrpc/xprt.h>
 
 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
 
@@ -128,6 +129,9 @@ static void svc_xprt_free(struct kref *kref)
 	if (test_bit(XPT_CACHE_AUTH, &xprt->xpt_flags))
 		svcauth_unix_info_release(xprt);
 	put_net(xprt->xpt_net);
+	/* See comment on corresponding get in xs_setup_bc_tcp(): */
+	if (xprt->xpt_bc_xprt)
+		xprt_put(xprt->xpt_bc_xprt);
 	xprt->xpt_ops->xpo_free(xprt);
 	module_put(owner);
 }
@@ -303,6 +307,15 @@ static void svc_thread_dequeue(struct svc_pool *pool, struct svc_rqst *rqstp)
 	list_del(&rqstp->rq_list);
 }
 
+static bool svc_xprt_has_something_to_do(struct svc_xprt *xprt)
+{
+	if (xprt->xpt_flags & ((1<<XPT_CONN)|(1<<XPT_CLOSE)))
+		return true;
+	if (xprt->xpt_flags & ((1<<XPT_DATA)|(1<<XPT_DEFERRED)))
+		return xprt->xpt_ops->xpo_has_wspace(xprt);
+	return false;
+}
+
 /*
  * Queue up a transport with data pending. If there are idle nfsd
  * processes, wake 'em up.
@@ -315,8 +328,7 @@ void svc_xprt_enqueue(struct svc_xprt *xprt)
 	struct svc_rqst	*rqstp;
 	int cpu;
 
-	if (!(xprt->xpt_flags &
-	      ((1<<XPT_CONN)|(1<<XPT_DATA)|(1<<XPT_CLOSE)|(1<<XPT_DEFERRED))))
+	if (!svc_xprt_has_something_to_do(xprt))
 		return;
 
 	cpu = get_cpu();
@@ -343,28 +355,7 @@ void svc_xprt_enqueue(struct svc_xprt *xprt)
 		dprintk("svc: transport %p busy, not enqueued\n", xprt);
 		goto out_unlock;
 	}
-	BUG_ON(xprt->xpt_pool != NULL);
-	xprt->xpt_pool = pool;
-
-	/* Handle pending connection */
-	if (test_bit(XPT_CONN, &xprt->xpt_flags))
-		goto process;
-
-	/* Handle close in-progress */
-	if (test_bit(XPT_CLOSE, &xprt->xpt_flags))
-		goto process;
-
-	/* Check if we have space to reply to a request */
-	if (!xprt->xpt_ops->xpo_has_wspace(xprt)) {
-		/* Don't enqueue while not enough space for reply */
-		dprintk("svc: no write space, transport %p  not enqueued\n",
-			xprt);
-		xprt->xpt_pool = NULL;
-		clear_bit(XPT_BUSY, &xprt->xpt_flags);
-		goto out_unlock;
-	}
 
- process:
 	if (!list_empty(&pool->sp_threads)) {
 		rqstp = list_entry(pool->sp_threads.next,
 				   struct svc_rqst,
@@ -381,13 +372,11 @@ void svc_xprt_enqueue(struct svc_xprt *xprt)
 		rqstp->rq_reserved = serv->sv_max_mesg;
 		atomic_add(rqstp->rq_reserved, &xprt->xpt_reserved);
 		pool->sp_stats.threads_woken++;
-		BUG_ON(xprt->xpt_pool != pool);
 		wake_up(&rqstp->rq_wait);
 	} else {
 		dprintk("svc: transport %p put into queue\n", xprt);
 		list_add_tail(&xprt->xpt_ready, &pool->sp_sockets);
 		pool->sp_stats.sockets_queued++;
-		BUG_ON(xprt->xpt_pool != pool);
 	}
 
 out_unlock:
@@ -426,7 +415,6 @@ static struct svc_xprt *svc_xprt_dequeue(struct svc_pool *pool)
 void svc_xprt_received(struct svc_xprt *xprt)
 {
 	BUG_ON(!test_bit(XPT_BUSY, &xprt->xpt_flags));
-	xprt->xpt_pool = NULL;
 	/* As soon as we clear busy, the xprt could be closed and
 	 * 'put', so we need a reference to call svc_xprt_enqueue with:
 	 */
@@ -722,7 +710,10 @@ int svc_recv(struct svc_rqst *rqstp, long timeout)
 	if (test_bit(XPT_CLOSE, &xprt->xpt_flags)) {
 		dprintk("svc_recv: found XPT_CLOSE\n");
 		svc_delete_xprt(xprt);
-	} else if (test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
+		/* Leave XPT_BUSY set on the dead xprt: */
+		goto out;
+	}
+	if (test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
 		struct svc_xprt *newxpt;
 		newxpt = xprt->xpt_ops->xpo_accept(xprt);
 		if (newxpt) {
@@ -747,28 +738,23 @@ int svc_recv(struct svc_rqst *rqstp, long timeout)
 			spin_unlock_bh(&serv->sv_lock);
 			svc_xprt_received(newxpt);
 		}
-		svc_xprt_received(xprt);
-	} else {
+	} else if (xprt->xpt_ops->xpo_has_wspace(xprt)) {
 		dprintk("svc: server %p, pool %u, transport %p, inuse=%d\n",
 			rqstp, pool->sp_id, xprt,
 			atomic_read(&xprt->xpt_ref.refcount));
 		rqstp->rq_deferred = svc_deferred_dequeue(xprt);
-		if (rqstp->rq_deferred) {
-			svc_xprt_received(xprt);
+		if (rqstp->rq_deferred)
 			len = svc_deferred_recv(rqstp);
-		} else {
+		else
 			len = xprt->xpt_ops->xpo_recvfrom(rqstp);
-			svc_xprt_received(xprt);
-		}
 		dprintk("svc: got len=%d\n", len);
 	}
+	svc_xprt_received(xprt);
 
 	/* No data, incomplete (TCP) read, or accept() */
-	if (len == 0 || len == -EAGAIN) {
-		rqstp->rq_res.len = 0;
-		svc_xprt_release(rqstp);
-		return -EAGAIN;
-	}
+	if (len == 0 || len == -EAGAIN)
+		goto out;
+
 	clear_bit(XPT_OLD, &xprt->xpt_flags);
 
 	rqstp->rq_secure = svc_port_is_privileged(svc_addr(rqstp));
@@ -777,6 +763,10 @@ int svc_recv(struct svc_rqst *rqstp, long timeout)
 	if (serv->sv_stats)
 		serv->sv_stats->netcnt++;
 	return len;
+out:
+	rqstp->rq_res.len = 0;
+	svc_xprt_release(rqstp);
+	return -EAGAIN;
 }
 EXPORT_SYMBOL_GPL(svc_recv);
 
@@ -935,7 +925,12 @@ void svc_close_xprt(struct svc_xprt *xprt)
 	if (test_and_set_bit(XPT_BUSY, &xprt->xpt_flags))
 		/* someone else will have to effect the close */
 		return;
-
+	/*
+	 * We expect svc_close_xprt() to work even when no threads are
+	 * running (e.g., while configuring the server before starting
+	 * any threads), so if the transport isn't busy, we delete
+	 * it ourself:
+	 */
 	svc_delete_xprt(xprt);
 }
 EXPORT_SYMBOL_GPL(svc_close_xprt);
@@ -945,16 +940,16 @@ void svc_close_all(struct list_head *xprt_list)
 	struct svc_xprt *xprt;
 	struct svc_xprt *tmp;
 
+	/*
+	 * The server is shutting down, and no more threads are running.
+	 * svc_xprt_enqueue() might still be running, but at worst it
+	 * will re-add the xprt to sp_sockets, which will soon get
+	 * freed.  So we don't bother with any more locking, and don't
+	 * leave the close to the (nonexistent) server threads:
+	 */
 	list_for_each_entry_safe(xprt, tmp, xprt_list, xpt_list) {
 		set_bit(XPT_CLOSE, &xprt->xpt_flags);
-		if (test_bit(XPT_BUSY, &xprt->xpt_flags)) {
-			/* Waiting to be processed, but no threads left,
-			 * So just remove it from the waiting list
-			 */
-			list_del_init(&xprt->xpt_ready);
-			clear_bit(XPT_BUSY, &xprt->xpt_flags);
-		}
-		svc_close_xprt(xprt);
+		svc_delete_xprt(xprt);
 	}
 }
 
@@ -1028,6 +1023,7 @@ static struct cache_deferred_req *svc_defer(struct cache_req *req)
 	}
 	svc_xprt_get(rqstp->rq_xprt);
 	dr->xprt = rqstp->rq_xprt;
+	rqstp->rq_dropme = true;
 
 	dr->handle.revisit = svc_revisit;
 	return &dr->handle;
@@ -1065,14 +1061,13 @@ static struct svc_deferred_req *svc_deferred_dequeue(struct svc_xprt *xprt)
 	if (!test_bit(XPT_DEFERRED, &xprt->xpt_flags))
 		return NULL;
 	spin_lock(&xprt->xpt_lock);
-	clear_bit(XPT_DEFERRED, &xprt->xpt_flags);
 	if (!list_empty(&xprt->xpt_deferred)) {
 		dr = list_entry(xprt->xpt_deferred.next,
 				struct svc_deferred_req,
 				handle.recent);
 		list_del_init(&dr->handle.recent);
-		set_bit(XPT_DEFERRED, &xprt->xpt_flags);
-	}
+	} else
+		clear_bit(XPT_DEFERRED, &xprt->xpt_flags);
 	spin_unlock(&xprt->xpt_lock);
 	return dr;
 }
diff --git a/net/sunrpc/svcauth.c b/net/sunrpc/svcauth.c
index 4e9393c24687..7963569fc04f 100644
--- a/net/sunrpc/svcauth.c
+++ b/net/sunrpc/svcauth.c
@@ -118,7 +118,6 @@ EXPORT_SYMBOL_GPL(svc_auth_unregister);
 
 #define	DN_HASHBITS	6
 #define	DN_HASHMAX	(1<<DN_HASHBITS)
-#define	DN_HASHMASK	(DN_HASHMAX-1)
 
 static struct hlist_head	auth_domain_table[DN_HASHMAX];
 static spinlock_t	auth_domain_lock =
diff --git a/net/sunrpc/svcauth_unix.c b/net/sunrpc/svcauth_unix.c
index 560677d187f1..30916b06c12b 100644
--- a/net/sunrpc/svcauth_unix.c
+++ b/net/sunrpc/svcauth_unix.c
@@ -30,7 +30,9 @@
 
 struct unix_domain {
 	struct auth_domain	h;
+#ifdef CONFIG_NFSD_DEPRECATED
 	int	addr_changes;
+#endif /* CONFIG_NFSD_DEPRECATED */
 	/* other stuff later */
 };
 
@@ -64,7 +66,9 @@ struct auth_domain *unix_domain_find(char *name)
 			return NULL;
 		}
 		new->h.flavour = &svcauth_unix;
+#ifdef CONFIG_NFSD_DEPRECATED
 		new->addr_changes = 0;
+#endif /* CONFIG_NFSD_DEPRECATED */
 		rv = auth_domain_lookup(name, &new->h);
 	}
 }
@@ -85,14 +89,15 @@ static void svcauth_unix_domain_release(struct auth_domain *dom)
  */
 #define	IP_HASHBITS	8
 #define	IP_HASHMAX	(1<<IP_HASHBITS)
-#define	IP_HASHMASK	(IP_HASHMAX-1)
 
 struct ip_map {
 	struct cache_head	h;
 	char			m_class[8]; /* e.g. "nfsd" */
 	struct in6_addr		m_addr;
 	struct unix_domain	*m_client;
+#ifdef CONFIG_NFSD_DEPRECATED
 	int			m_add_change;
+#endif /* CONFIG_NFSD_DEPRECATED */
 };
 
 static void ip_map_put(struct kref *kref)
@@ -146,7 +151,9 @@ static void update(struct cache_head *cnew, struct cache_head *citem)
 
 	kref_get(&item->m_client->h.ref);
 	new->m_client = item->m_client;
+#ifdef CONFIG_NFSD_DEPRECATED
 	new->m_add_change = item->m_add_change;
+#endif /* CONFIG_NFSD_DEPRECATED */
 }
 static struct cache_head *ip_map_alloc(void)
 {
@@ -331,6 +338,7 @@ static int __ip_map_update(struct cache_detail *cd, struct ip_map *ipm,
 	ip.h.flags = 0;
 	if (!udom)
 		set_bit(CACHE_NEGATIVE, &ip.h.flags);
+#ifdef CONFIG_NFSD_DEPRECATED
 	else {
 		ip.m_add_change = udom->addr_changes;
 		/* if this is from the legacy set_client system call,
@@ -339,6 +347,7 @@ static int __ip_map_update(struct cache_detail *cd, struct ip_map *ipm,
 		if (expiry == NEVER)
 			ip.m_add_change++;
 	}
+#endif /* CONFIG_NFSD_DEPRECATED */
 	ip.h.expiry_time = expiry;
 	ch = sunrpc_cache_update(cd, &ip.h, &ipm->h,
 				 hash_str(ipm->m_class, IP_HASHBITS) ^
@@ -358,6 +367,7 @@ static inline int ip_map_update(struct net *net, struct ip_map *ipm,
 	return __ip_map_update(sn->ip_map_cache, ipm, udom, expiry);
 }
 
+#ifdef CONFIG_NFSD_DEPRECATED
 int auth_unix_add_addr(struct net *net, struct in6_addr *addr, struct auth_domain *dom)
 {
 	struct unix_domain *udom;
@@ -402,8 +412,7 @@ struct auth_domain *auth_unix_lookup(struct net *net, struct in6_addr *addr)
 		return NULL;
 
 	if ((ipm->m_client->addr_changes - ipm->m_add_change) >0) {
-		if (test_and_set_bit(CACHE_NEGATIVE, &ipm->h.flags) == 0)
-			auth_domain_put(&ipm->m_client->h);
+		sunrpc_invalidate(&ipm->h, sn->ip_map_cache);
 		rv = NULL;
 	} else {
 		rv = &ipm->m_client->h;
@@ -413,6 +422,7 @@ struct auth_domain *auth_unix_lookup(struct net *net, struct in6_addr *addr)
 	return rv;
 }
 EXPORT_SYMBOL_GPL(auth_unix_lookup);
+#endif /* CONFIG_NFSD_DEPRECATED */
 
 void svcauth_unix_purge(void)
 {
@@ -497,7 +507,6 @@ svcauth_unix_info_release(struct svc_xprt *xpt)
  */
 #define	GID_HASHBITS	8
 #define	GID_HASHMAX	(1<<GID_HASHBITS)
-#define	GID_HASHMASK	(GID_HASHMAX - 1)
 
 struct unix_gid {
 	struct cache_head	h;
diff --git a/net/sunrpc/svcsock.c b/net/sunrpc/svcsock.c
index d265aa700bb3..7bd3bbba4710 100644
--- a/net/sunrpc/svcsock.c
+++ b/net/sunrpc/svcsock.c
@@ -331,19 +331,21 @@ int svc_sock_names(struct svc_serv *serv, char *buf, const size_t buflen,
 			len = onelen;
 			break;
 		}
-		if (toclose && strcmp(toclose, buf + len) == 0)
+		if (toclose && strcmp(toclose, buf + len) == 0) {
 			closesk = svsk;
-		else
+			svc_xprt_get(&closesk->sk_xprt);
+		} else
 			len += onelen;
 	}
 	spin_unlock_bh(&serv->sv_lock);
 
-	if (closesk)
+	if (closesk) {
 		/* Should unregister with portmap, but you cannot
 		 * unregister just one protocol...
 		 */
 		svc_close_xprt(&closesk->sk_xprt);
-	else if (toclose)
+		svc_xprt_put(&closesk->sk_xprt);
+	} else if (toclose)
 		return -ENOENT;
 	return len;
 }
@@ -992,15 +994,17 @@ static int svc_process_calldir(struct svc_sock *svsk, struct svc_rqst *rqstp,
 		vec[0] = rqstp->rq_arg.head[0];
 	} else {
 		/* REPLY */
-		if (svsk->sk_bc_xprt)
-			req = xprt_lookup_rqst(svsk->sk_bc_xprt, xid);
+		struct rpc_xprt *bc_xprt = svsk->sk_xprt.xpt_bc_xprt;
+
+		if (bc_xprt)
+			req = xprt_lookup_rqst(bc_xprt, xid);
 
 		if (!req) {
 			printk(KERN_NOTICE
 				"%s: Got unrecognized reply: "
-				"calldir 0x%x sk_bc_xprt %p xid %08x\n",
+				"calldir 0x%x xpt_bc_xprt %p xid %08x\n",
 				__func__, ntohl(calldir),
-				svsk->sk_bc_xprt, xid);
+				bc_xprt, xid);
 			vec[0] = rqstp->rq_arg.head[0];
 			goto out;
 		}
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 4c8f18aff7c3..856274d7e85c 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -965,6 +965,7 @@ struct rpc_xprt *xprt_alloc(struct net *net, int size, int max_req)
 	xprt = kzalloc(size, GFP_KERNEL);
 	if (xprt == NULL)
 		goto out;
+	kref_init(&xprt->kref);
 
 	xprt->max_reqs = max_req;
 	xprt->slot = kcalloc(max_req, sizeof(struct rpc_rqst), GFP_KERNEL);
@@ -1101,8 +1102,10 @@ found:
 				-PTR_ERR(xprt));
 		return xprt;
 	}
+	if (test_and_set_bit(XPRT_INITIALIZED, &xprt->state))
+		/* ->setup returned a pre-initialized xprt: */
+		return xprt;
 
-	kref_init(&xprt->kref);
 	spin_lock_init(&xprt->transport_lock);
 	spin_lock_init(&xprt->reserve_lock);
 
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 96549df836ee..c431f5a57960 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -2359,6 +2359,15 @@ static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
 	struct svc_sock *bc_sock;
 	struct rpc_xprt *ret;
 
+	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,
+		 * reuse the already-existing one:
+		 */
+		 return args->bc_xprt->xpt_bc_xprt;
+	}
 	xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries);
 	if (IS_ERR(xprt))
 		return xprt;
@@ -2375,16 +2384,6 @@ static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
 	xprt->reestablish_timeout = 0;
 	xprt->idle_timeout = 0;
 
-	/*
-	 * The backchannel uses the same socket connection as the
-	 * forechannel
-	 */
-	xprt->bc_xprt = args->bc_xprt;
-	bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
-	bc_sock->sk_bc_xprt = xprt;
-	transport->sock = bc_sock->sk_sock;
-	transport->inet = bc_sock->sk_sk;
-
 	xprt->ops = &bc_tcp_ops;
 
 	switch (addr->sa_family) {
@@ -2407,6 +2406,20 @@ static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
 			xprt->address_strings[RPC_DISPLAY_PROTO]);
 
 	/*
+	 * Once we've associated a backchannel xprt with a connection,
+	 * we want to keep it around as long as long as the connection
+	 * lasts, in case we need to start using it for a backchannel
+	 * again; this reference won't be dropped until bc_xprt is
+	 * destroyed.
+	 */
+	xprt_get(xprt);
+	args->bc_xprt->xpt_bc_xprt = xprt;
+	xprt->bc_xprt = args->bc_xprt;
+	bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
+	transport->sock = bc_sock->sk_sock;
+	transport->inet = bc_sock->sk_sk;
+
+	/*
 	 * Since we don't want connections for the backchannel, we set
 	 * the xprt status to connected
 	 */
@@ -2415,6 +2428,7 @@ static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
 
 	if (try_module_get(THIS_MODULE))
 		return xprt;
+	xprt_put(xprt);
 	ret = ERR_PTR(-EINVAL);
 out_err:
 	xprt_free(xprt);