Merge tag 'nfs-for-3.9-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client bugfixes from Trond Myklebust:

 - Fix an Oops in the pNFS layoutget code

 - Fix a number of NFSv4 and v4.1 state recovery deadlocks and hangs due
   to the interaction of the session drain lock and state management
   locks.

 - Remove task->tk_xprt, which was hiding a lot of RCU dereferencing
   bugs

 - Fix a long standing NFSv3 posix lock recovery bug.

 - Revert commit 324d003b0cd8 ("NFS: add nfs_sb_deactive_async to avoid
   deadlock").  It turned out that the root cause of the deadlock was
   due to interactions with the workqueues that have now been resolved.

* tag 'nfs-for-3.9-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (22 commits)
  NLM: Ensure that we resend all pending blocking locks after a reclaim
  umount oops when remove blocklayoutdriver first
  sunrpc: silence build warning in gss_fill_context
  nfs: remove kfree() redundant null checks
  NFSv4.1: Don't decode skipped layoutgets
  NFSv4.1: Fix bulk recall and destroy of layouts
  NFSv4.1: Fix an ABBA locking issue with session and state serialisation
  NFSv4: Fix a reboot recovery race when opening a file
  NFSv4: Ensure delegation recall and byte range lock removal don't conflict
  NFSv4: Fix up the return values of nfs4_open_delegation_recall
  NFSv4.1: Don't lose locks when a server reboots during delegation return
  NFSv4.1: Prevent deadlocks between state recovery and file locking
  NFSv4: Allow the state manager to mark an open_owner as being recovered
  SUNRPC: Add missing static declaration to _gss_mech_get_by_name
  Revert "NFS: add nfs_sb_deactive_async to avoid deadlock"
  SUNRPC: Nuke the tk_xprt macro
  SUNRPC: Avoid RCU dereferences in the transport bind and connect code
  SUNRPC: Fix an RCU dereference in xprt_reserve
  SUNRPC: Pass pointers to struct rpc_xprt to the congestion window
  SUNRPC: Fix an RCU dereference in xs_local_rpcbind
  ...

8 files changed, 42 insertions, 37 deletions

diff --git a/net/sunrpc/auth_gss/auth_gss.c b/net/sunrpc/auth_gss/auth_gss.c
index 6e5c824b040b..911ef008b701 100644
--- a/net/sunrpc/auth_gss/auth_gss.c
+++ b/net/sunrpc/auth_gss/auth_gss.c
@@ -247,8 +247,7 @@ gss_fill_context(const void *p, const void *end, struct gss_cl_ctx *ctx, struct
 		__func__, ctx->gc_expiry, now, timeout);
 	return q;
 err:
-	dprintk("RPC:       %s returns %ld gc_expiry %lu now %lu timeout %u\n",
-		__func__, -PTR_ERR(p), ctx->gc_expiry, now, timeout);
+	dprintk("RPC:       %s returns error %ld\n", __func__, -PTR_ERR(p));
 	return p;
 }
 
@@ -1154,7 +1153,7 @@ gss_marshal(struct rpc_task *task, __be32 *p)
 
 	/* We compute the checksum for the verifier over the xdr-encoded bytes
 	 * starting with the xid and ending at the end of the credential: */
-	iov.iov_base = xprt_skip_transport_header(task->tk_xprt,
+	iov.iov_base = xprt_skip_transport_header(req->rq_xprt,
 					req->rq_snd_buf.head[0].iov_base);
 	iov.iov_len = (u8 *)p - (u8 *)iov.iov_base;
 	xdr_buf_from_iov(&iov, &verf_buf);
diff --git a/net/sunrpc/auth_gss/gss_mech_switch.c b/net/sunrpc/auth_gss/gss_mech_switch.c
index b174fcd9ff4c..f0f4eee63a35 100644
--- a/net/sunrpc/auth_gss/gss_mech_switch.c
+++ b/net/sunrpc/auth_gss/gss_mech_switch.c
@@ -140,7 +140,7 @@ gss_mech_get(struct gss_api_mech *gm)
 
 EXPORT_SYMBOL_GPL(gss_mech_get);
 
-struct gss_api_mech *
+static struct gss_api_mech *
 _gss_mech_get_by_name(const char *name)
 {
 	struct gss_api_mech	*pos, *gm = NULL;
@@ -205,7 +205,7 @@ mech_supports_pseudoflavor(struct gss_api_mech *gm, u32 pseudoflavor)
 	return 0;
 }
 
-struct gss_api_mech *_gss_mech_get_by_pseudoflavor(u32 pseudoflavor)
+static struct gss_api_mech *_gss_mech_get_by_pseudoflavor(u32 pseudoflavor)
 {
 	struct gss_api_mech *gm = NULL, *pos;
 
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 507b5e84fbdb..a9f7906c1a6a 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -1400,7 +1400,7 @@ call_allocate(struct rpc_task *task)
 {
 	unsigned int slack = task->tk_rqstp->rq_cred->cr_auth->au_cslack;
 	struct rpc_rqst *req = task->tk_rqstp;
-	struct rpc_xprt *xprt = task->tk_xprt;
+	struct rpc_xprt *xprt = req->rq_xprt;
 	struct rpc_procinfo *proc = task->tk_msg.rpc_proc;
 
 	dprint_status(task);
@@ -1508,7 +1508,7 @@ rpc_xdr_encode(struct rpc_task *task)
 static void
 call_bind(struct rpc_task *task)
 {
-	struct rpc_xprt *xprt = task->tk_xprt;
+	struct rpc_xprt *xprt = task->tk_rqstp->rq_xprt;
 
 	dprint_status(task);
 
@@ -1602,7 +1602,7 @@ retry_timeout:
 static void
 call_connect(struct rpc_task *task)
 {
-	struct rpc_xprt *xprt = task->tk_xprt;
+	struct rpc_xprt *xprt = task->tk_rqstp->rq_xprt;
 
 	dprintk("RPC: %5u call_connect xprt %p %s connected\n",
 			task->tk_pid, xprt,
@@ -1685,7 +1685,7 @@ call_transmit(struct rpc_task *task)
 	if (rpc_reply_expected(task))
 		return;
 	task->tk_action = rpc_exit_task;
-	rpc_wake_up_queued_task(&task->tk_xprt->pending, task);
+	rpc_wake_up_queued_task(&task->tk_rqstp->rq_xprt->pending, task);
 }
 
 /*
@@ -1784,7 +1784,7 @@ call_bc_transmit(struct rpc_task *task)
 		 */
 		printk(KERN_NOTICE "RPC: Could not send backchannel reply "
 			"error: %d\n", task->tk_status);
-		xprt_conditional_disconnect(task->tk_xprt,
+		xprt_conditional_disconnect(req->rq_xprt,
 			req->rq_connect_cookie);
 		break;
 	default:
@@ -1836,7 +1836,7 @@ call_status(struct rpc_task *task)
 	case -ETIMEDOUT:
 		task->tk_action = call_timeout;
 		if (task->tk_client->cl_discrtry)
-			xprt_conditional_disconnect(task->tk_xprt,
+			xprt_conditional_disconnect(req->rq_xprt,
 					req->rq_connect_cookie);
 		break;
 	case -ECONNRESET:
@@ -1991,7 +1991,7 @@ out_retry:
 	if (task->tk_rqstp == req) {
 		req->rq_reply_bytes_recvd = req->rq_rcv_buf.len = 0;
 		if (task->tk_client->cl_discrtry)
-			xprt_conditional_disconnect(task->tk_xprt,
+			xprt_conditional_disconnect(req->rq_xprt,
 					req->rq_connect_cookie);
 	}
 }
@@ -2005,7 +2005,7 @@ rpc_encode_header(struct rpc_task *task)
 
 	/* FIXME: check buffer size? */
 
-	p = xprt_skip_transport_header(task->tk_xprt, p);
+	p = xprt_skip_transport_header(req->rq_xprt, p);
 	*p++ = req->rq_xid;		/* XID */
 	*p++ = htonl(RPC_CALL);		/* CALL */
 	*p++ = htonl(RPC_VERSION);	/* RPC version */
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 33811db8788a..846c34fdee9f 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -430,21 +430,23 @@ __xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
  */
 void xprt_release_rqst_cong(struct rpc_task *task)
 {
-	__xprt_put_cong(task->tk_xprt, task->tk_rqstp);
+	struct rpc_rqst *req = task->tk_rqstp;
+
+	__xprt_put_cong(req->rq_xprt, req);
 }
 EXPORT_SYMBOL_GPL(xprt_release_rqst_cong);
 
 /**
  * xprt_adjust_cwnd - adjust transport congestion window
+ * @xprt: pointer to xprt
  * @task: recently completed RPC request used to adjust window
  * @result: result code of completed RPC request
  *
  * We use a time-smoothed congestion estimator to avoid heavy oscillation.
  */
-void xprt_adjust_cwnd(struct rpc_task *task, int result)
+void xprt_adjust_cwnd(struct rpc_xprt *xprt, struct rpc_task *task, int result)
 {
 	struct rpc_rqst *req = task->tk_rqstp;
-	struct rpc_xprt *xprt = task->tk_xprt;
 	unsigned long cwnd = xprt->cwnd;
 
 	if (result >= 0 && cwnd <= xprt->cong) {
@@ -695,7 +697,7 @@ out_abort:
  */
 void xprt_connect(struct rpc_task *task)
 {
-	struct rpc_xprt	*xprt = task->tk_xprt;
+	struct rpc_xprt	*xprt = task->tk_rqstp->rq_xprt;
 
 	dprintk("RPC: %5u xprt_connect xprt %p %s connected\n", task->tk_pid,
 			xprt, (xprt_connected(xprt) ? "is" : "is not"));
@@ -722,13 +724,13 @@ void xprt_connect(struct rpc_task *task)
 		if (xprt_test_and_set_connecting(xprt))
 			return;
 		xprt->stat.connect_start = jiffies;
-		xprt->ops->connect(task);
+		xprt->ops->connect(xprt, task);
 	}
 }
 
 static void xprt_connect_status(struct rpc_task *task)
 {
-	struct rpc_xprt	*xprt = task->tk_xprt;
+	struct rpc_xprt	*xprt = task->tk_rqstp->rq_xprt;
 
 	if (task->tk_status == 0) {
 		xprt->stat.connect_count++;
@@ -832,7 +834,7 @@ static void xprt_timer(struct rpc_task *task)
 	spin_lock_bh(&xprt->transport_lock);
 	if (!req->rq_reply_bytes_recvd) {
 		if (xprt->ops->timer)
-			xprt->ops->timer(task);
+			xprt->ops->timer(xprt, task);
 	} else
 		task->tk_status = 0;
 	spin_unlock_bh(&xprt->transport_lock);
@@ -1091,7 +1093,7 @@ EXPORT_SYMBOL_GPL(xprt_free);
  */
 void xprt_reserve(struct rpc_task *task)
 {
-	struct rpc_xprt	*xprt = task->tk_xprt;
+	struct rpc_xprt	*xprt;
 
 	task->tk_status = 0;
 	if (task->tk_rqstp != NULL)
@@ -1099,7 +1101,10 @@ void xprt_reserve(struct rpc_task *task)
 
 	task->tk_timeout = 0;
 	task->tk_status = -EAGAIN;
+	rcu_read_lock();
+	xprt = rcu_dereference(task->tk_client->cl_xprt);
 	xprt->ops->alloc_slot(xprt, task);
+	rcu_read_unlock();
 }
 
 static inline __be32 xprt_alloc_xid(struct rpc_xprt *xprt)
diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index 558fbab574f0..e03725bfe2b8 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -171,7 +171,7 @@ rpcrdma_create_chunks(struct rpc_rqst *rqst, struct xdr_buf *target,
 		struct rpcrdma_msg *headerp, enum rpcrdma_chunktype type)
 {
 	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
-	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_task->tk_xprt);
+	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(rqst->rq_xprt);
 	int nsegs, nchunks = 0;
 	unsigned int pos;
 	struct rpcrdma_mr_seg *seg = req->rl_segments;
@@ -366,7 +366,7 @@ rpcrdma_inline_pullup(struct rpc_rqst *rqst, int pad)
 int
 rpcrdma_marshal_req(struct rpc_rqst *rqst)
 {
-	struct rpc_xprt *xprt = rqst->rq_task->tk_xprt;
+	struct rpc_xprt *xprt = rqst->rq_xprt;
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
 	char *base;
diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index c9aa7a35f3bf..d0074289708e 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -426,9 +426,8 @@ xprt_rdma_set_port(struct rpc_xprt *xprt, u16 port)
 }
 
 static void
-xprt_rdma_connect(struct rpc_task *task)
+xprt_rdma_connect(struct rpc_xprt *xprt, struct rpc_task *task)
 {
-	struct rpc_xprt *xprt = (struct rpc_xprt *)task->tk_xprt;
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 
 	if (r_xprt->rx_ep.rep_connected != 0) {
@@ -475,7 +474,7 @@ xprt_rdma_reserve_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
 static void *
 xprt_rdma_allocate(struct rpc_task *task, size_t size)
 {
-	struct rpc_xprt *xprt = task->tk_xprt;
+	struct rpc_xprt *xprt = task->tk_rqstp->rq_xprt;
 	struct rpcrdma_req *req, *nreq;
 
 	req = rpcrdma_buffer_get(&rpcx_to_rdmax(xprt)->rx_buf);
@@ -627,7 +626,7 @@ static int
 xprt_rdma_send_request(struct rpc_task *task)
 {
 	struct rpc_rqst *rqst = task->tk_rqstp;
-	struct rpc_xprt *xprt = task->tk_xprt;
+	struct rpc_xprt *xprt = rqst->rq_xprt;
 	struct rpcrdma_req *req = rpcr_to_rdmar(rqst);
 	struct rpcrdma_xprt *r_xprt = rpcx_to_rdmax(xprt);
 
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index 9a66c95b5837..cc1445dc1d1a 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -235,13 +235,13 @@ struct rpcrdma_create_data_internal {
 };
 
 #define RPCRDMA_INLINE_READ_THRESHOLD(rq) \
-	(rpcx_to_rdmad(rq->rq_task->tk_xprt).inline_rsize)
+	(rpcx_to_rdmad(rq->rq_xprt).inline_rsize)
 
 #define RPCRDMA_INLINE_WRITE_THRESHOLD(rq)\
-	(rpcx_to_rdmad(rq->rq_task->tk_xprt).inline_wsize)
+	(rpcx_to_rdmad(rq->rq_xprt).inline_wsize)
 
 #define RPCRDMA_INLINE_PAD_VALUE(rq)\
-	rpcx_to_rdmad(rq->rq_task->tk_xprt).padding
+	rpcx_to_rdmad(rq->rq_xprt).padding
 
 /*
  * Statistics for RPCRDMA
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 68b0a81c31d5..37cbda63f45c 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -770,7 +770,7 @@ static void xs_tcp_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
 		goto out_release;
 	if (req->rq_bytes_sent == req->rq_snd_buf.len)
 		goto out_release;
-	set_bit(XPRT_CLOSE_WAIT, &task->tk_xprt->state);
+	set_bit(XPRT_CLOSE_WAIT, &xprt->state);
 out_release:
 	xprt_release_xprt(xprt, task);
 }
@@ -1005,7 +1005,7 @@ static void xs_udp_data_ready(struct sock *sk, int len)
 
 	UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS);
 
-	xprt_adjust_cwnd(task, copied);
+	xprt_adjust_cwnd(xprt, task, copied);
 	xprt_complete_rqst(task, copied);
 
  out_unlock:
@@ -1646,9 +1646,9 @@ static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t
  *
  * Adjust the congestion window after a retransmit timeout has occurred.
  */
-static void xs_udp_timer(struct rpc_task *task)
+static void xs_udp_timer(struct rpc_xprt *xprt, struct rpc_task *task)
 {
-	xprt_adjust_cwnd(task, -ETIMEDOUT);
+	xprt_adjust_cwnd(xprt, task, -ETIMEDOUT);
 }
 
 static unsigned short xs_get_random_port(void)
@@ -1731,7 +1731,9 @@ static int xs_bind(struct sock_xprt *transport, struct socket *sock)
  */
 static void xs_local_rpcbind(struct rpc_task *task)
 {
-	xprt_set_bound(task->tk_xprt);
+	rcu_read_lock();
+	xprt_set_bound(rcu_dereference(task->tk_client->cl_xprt));
+	rcu_read_unlock();
 }
 
 static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port)
@@ -2205,6 +2207,7 @@ out:
 
 /**
  * xs_connect - connect a socket to a remote endpoint
+ * @xprt: pointer to transport structure
  * @task: address of RPC task that manages state of connect request
  *
  * TCP: If the remote end dropped the connection, delay reconnecting.
@@ -2216,9 +2219,8 @@ out:
  * If a UDP socket connect fails, the delay behavior here prevents
  * retry floods (hard mounts).
  */
-static void xs_connect(struct rpc_task *task)
+static void xs_connect(struct rpc_xprt *xprt, struct rpc_task *task)
 {
-	struct rpc_xprt *xprt = task->tk_xprt;
 	struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
 
 	if (transport->sock != NULL && !RPC_IS_SOFTCONN(task)) {