Merge tag 'nfs-for-5.13-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client bugfixes from Trond Myklebust:
"Stable fixes:
   - Fix v4.0/v4.1 SEEK_DATA return -ENOTSUPP when set NFS_V4_2 config
   - Fix Oops in xs_tcp_send_request() when transport is disconnected
   - Fix a NULL pointer dereference in pnfs_mark_matching_lsegs_return()

  Bugfixes:
   - Fix instances where signal_pending() should be fatal_signal_pending()
   - fix an incorrect limit in filelayout_decode_layout()
   - Fixes for the SUNRPC backlogged RPC queue
   - Don't corrupt the value of pg_bytes_written in nfs_do_recoalesce()
   - Revert commit 586a0787ce35 ("Clean up rpcrdma_prepare_readch()")"

* tag 'nfs-for-5.13-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  nfs: Remove trailing semicolon in macros
  xprtrdma: Revert 586a0787ce35
  NFSv4: Fix v4.0/v4.1 SEEK_DATA return -ENOTSUPP when set NFS_V4_2 config
  NFS: Clean up reset of the mirror accounting variables
  NFS: Don't corrupt the value of pg_bytes_written in nfs_do_recoalesce()
  NFS: Fix an Oopsable condition in __nfs_pageio_add_request()
  SUNRPC: More fixes for backlog congestion
  SUNRPC: Fix Oops in xs_tcp_send_request() when transport is disconnected
  NFSv4: Fix a NULL pointer dereference in pnfs_mark_matching_lsegs_return()
  SUNRPC in case of backlog, hand free slots directly to waiting task
  pNFS/NFSv4: Remove redundant initialization of 'rd_size'
  NFS: fix an incorrect limit in filelayout_decode_layout()
  fs/nfs: Use fatal_signal_pending instead of signal_pending

7 files changed, 74 insertions, 33 deletions

diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index f555d335e910..42623d6b8f0e 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -1677,13 +1677,6 @@ call_reserveresult(struct rpc_task *task)
 		return;
 	}
 
-	/*
-	 * Even though there was an error, we may have acquired
-	 * a request slot somehow.  Make sure not to leak it.
-	 */
-	if (task->tk_rqstp)
-		xprt_release(task);
-
 	switch (status) {
 	case -ENOMEM:
 		rpc_delay(task, HZ >> 2);
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index e5b5a960a69b..3509a7f139b9 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -70,6 +70,7 @@
 static void	 xprt_init(struct rpc_xprt *xprt, struct net *net);
 static __be32	xprt_alloc_xid(struct rpc_xprt *xprt);
 static void	 xprt_destroy(struct rpc_xprt *xprt);
+static void	 xprt_request_init(struct rpc_task *task);
 
 static DEFINE_SPINLOCK(xprt_list_lock);
 static LIST_HEAD(xprt_list);
@@ -1606,17 +1607,40 @@ xprt_transmit(struct rpc_task *task)
 	spin_unlock(&xprt->queue_lock);
 }
 
-static void xprt_add_backlog(struct rpc_xprt *xprt, struct rpc_task *task)
+static void xprt_complete_request_init(struct rpc_task *task)
+{
+	if (task->tk_rqstp)
+		xprt_request_init(task);
+}
+
+void xprt_add_backlog(struct rpc_xprt *xprt, struct rpc_task *task)
 {
 	set_bit(XPRT_CONGESTED, &xprt->state);
-	rpc_sleep_on(&xprt->backlog, task, NULL);
+	rpc_sleep_on(&xprt->backlog, task, xprt_complete_request_init);
+}
+EXPORT_SYMBOL_GPL(xprt_add_backlog);
+
+static bool __xprt_set_rq(struct rpc_task *task, void *data)
+{
+	struct rpc_rqst *req = data;
+
+	if (task->tk_rqstp == NULL) {
+		memset(req, 0, sizeof(*req));	/* mark unused */
+		task->tk_rqstp = req;
+		return true;
+	}
+	return false;
 }
 
-static void xprt_wake_up_backlog(struct rpc_xprt *xprt)
+bool xprt_wake_up_backlog(struct rpc_xprt *xprt, struct rpc_rqst *req)
 {
-	if (rpc_wake_up_next(&xprt->backlog) == NULL)
+	if (rpc_wake_up_first(&xprt->backlog, __xprt_set_rq, req) == NULL) {
 		clear_bit(XPRT_CONGESTED, &xprt->state);
+		return false;
+	}
+	return true;
 }
+EXPORT_SYMBOL_GPL(xprt_wake_up_backlog);
 
 static bool xprt_throttle_congested(struct rpc_xprt *xprt, struct rpc_task *task)
 {
@@ -1626,7 +1650,7 @@ static bool xprt_throttle_congested(struct rpc_xprt *xprt, struct rpc_task *task
 		goto out;
 	spin_lock(&xprt->reserve_lock);
 	if (test_bit(XPRT_CONGESTED, &xprt->state)) {
-		rpc_sleep_on(&xprt->backlog, task, NULL);
+		xprt_add_backlog(xprt, task);
 		ret = true;
 	}
 	spin_unlock(&xprt->reserve_lock);
@@ -1703,11 +1727,11 @@ EXPORT_SYMBOL_GPL(xprt_alloc_slot);
 void xprt_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *req)
 {
 	spin_lock(&xprt->reserve_lock);
-	if (!xprt_dynamic_free_slot(xprt, req)) {
+	if (!xprt_wake_up_backlog(xprt, req) &&
+	    !xprt_dynamic_free_slot(xprt, req)) {
 		memset(req, 0, sizeof(*req));	/* mark unused */
 		list_add(&req->rq_list, &xprt->free);
 	}
-	xprt_wake_up_backlog(xprt);
 	spin_unlock(&xprt->reserve_lock);
 }
 EXPORT_SYMBOL_GPL(xprt_free_slot);
@@ -1894,10 +1918,10 @@ void xprt_release(struct rpc_task *task)
 	xdr_free_bvec(&req->rq_snd_buf);
 	if (req->rq_cred != NULL)
 		put_rpccred(req->rq_cred);
-	task->tk_rqstp = NULL;
 	if (req->rq_release_snd_buf)
 		req->rq_release_snd_buf(req);
 
+	task->tk_rqstp = NULL;
 	if (likely(!bc_prealloc(req)))
 		xprt->ops->free_slot(xprt, req);
 	else
diff --git a/net/sunrpc/xprtrdma/rpc_rdma.c b/net/sunrpc/xprtrdma/rpc_rdma.c
index 649f7d8b9733..c335c1361564 100644
--- a/net/sunrpc/xprtrdma/rpc_rdma.c
+++ b/net/sunrpc/xprtrdma/rpc_rdma.c
@@ -628,8 +628,9 @@ out_mapping_err:
 	return false;
 }
 
-/* The tail iovec might not reside in the same page as the
- * head iovec.
+/* The tail iovec may include an XDR pad for the page list,
+ * as well as additional content, and may not reside in the
+ * same page as the head iovec.
  */
 static bool rpcrdma_prepare_tail_iov(struct rpcrdma_req *req,
 				     struct xdr_buf *xdr,
@@ -747,19 +748,27 @@ static bool rpcrdma_prepare_readch(struct rpcrdma_xprt *r_xprt,
 				   struct rpcrdma_req *req,
 				   struct xdr_buf *xdr)
 {
-	struct kvec *tail = &xdr->tail[0];
-
 	if (!rpcrdma_prepare_head_iov(r_xprt, req, xdr->head[0].iov_len))
 		return false;
 
-	/* If there is a Read chunk, the page list is handled
+	/* If there is a Read chunk, the page list is being handled
 	 * via explicit RDMA, and thus is skipped here.
 	 */
 
-	if (tail->iov_len) {
-		if (!rpcrdma_prepare_tail_iov(req, xdr,
-					      offset_in_page(tail->iov_base),
-					      tail->iov_len))
+	/* Do not include the tail if it is only an XDR pad */
+	if (xdr->tail[0].iov_len > 3) {
+		unsigned int page_base, len;
+
+		/* If the content in the page list is an odd length,
+		 * xdr_write_pages() adds a pad at the beginning of
+		 * the tail iovec. Force the tail's non-pad content to
+		 * land at the next XDR position in the Send message.
+		 */
+		page_base = offset_in_page(xdr->tail[0].iov_base);
+		len = xdr->tail[0].iov_len;
+		page_base += len & 3;
+		len -= len & 3;
+		if (!rpcrdma_prepare_tail_iov(req, xdr, page_base, len))
 			return false;
 		kref_get(&req->rl_kref);
 	}
diff --git a/net/sunrpc/xprtrdma/transport.c b/net/sunrpc/xprtrdma/transport.c
index 09953597d055..19a49d26b1e4 100644
--- a/net/sunrpc/xprtrdma/transport.c
+++ b/net/sunrpc/xprtrdma/transport.c
@@ -520,9 +520,8 @@ xprt_rdma_alloc_slot(struct rpc_xprt *xprt, struct rpc_task *task)
 	return;
 
 out_sleep:
-	set_bit(XPRT_CONGESTED, &xprt->state);
-	rpc_sleep_on(&xprt->backlog, task, NULL);
 	task->tk_status = -EAGAIN;
+	xprt_add_backlog(xprt, task);
 }
 
 /**
@@ -537,10 +536,11 @@ xprt_rdma_free_slot(struct rpc_xprt *xprt, struct rpc_rqst *rqst)
 	struct rpcrdma_xprt *r_xprt =
 		container_of(xprt, struct rpcrdma_xprt, rx_xprt);
 
-	memset(rqst, 0, sizeof(*rqst));
-	rpcrdma_buffer_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst));
-	if (unlikely(!rpc_wake_up_next(&xprt->backlog)))
-		clear_bit(XPRT_CONGESTED, &xprt->state);
+	rpcrdma_reply_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst));
+	if (!xprt_wake_up_backlog(xprt, rqst)) {
+		memset(rqst, 0, sizeof(*rqst));
+		rpcrdma_buffer_put(&r_xprt->rx_buf, rpcr_to_rdmar(rqst));
+	}
 }
 
 static bool rpcrdma_check_regbuf(struct rpcrdma_xprt *r_xprt,
diff --git a/net/sunrpc/xprtrdma/verbs.c b/net/sunrpc/xprtrdma/verbs.c
index 1e965a380896..649c23518ec0 100644
--- a/net/sunrpc/xprtrdma/verbs.c
+++ b/net/sunrpc/xprtrdma/verbs.c
@@ -1201,6 +1201,20 @@ rpcrdma_mr_get(struct rpcrdma_xprt *r_xprt)
 }
 
 /**
+ * rpcrdma_reply_put - Put reply buffers back into pool
+ * @buffers: buffer pool
+ * @req: object to return
+ *
+ */
+void rpcrdma_reply_put(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req)
+{
+	if (req->rl_reply) {
+		rpcrdma_rep_put(buffers, req->rl_reply);
+		req->rl_reply = NULL;
+	}
+}
+
+/**
  * rpcrdma_buffer_get - Get a request buffer
  * @buffers: Buffer pool from which to obtain a buffer
  *
@@ -1228,9 +1242,7 @@ rpcrdma_buffer_get(struct rpcrdma_buffer *buffers)
  */
 void rpcrdma_buffer_put(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req)
 {
-	if (req->rl_reply)
-		rpcrdma_rep_put(buffers, req->rl_reply);
-	req->rl_reply = NULL;
+	rpcrdma_reply_put(buffers, req);
 
 	spin_lock(&buffers->rb_lock);
 	list_add(&req->rl_list, &buffers->rb_send_bufs);
diff --git a/net/sunrpc/xprtrdma/xprt_rdma.h b/net/sunrpc/xprtrdma/xprt_rdma.h
index 436ad7312614..5d231d94e944 100644
--- a/net/sunrpc/xprtrdma/xprt_rdma.h
+++ b/net/sunrpc/xprtrdma/xprt_rdma.h
@@ -479,6 +479,7 @@ struct rpcrdma_req *rpcrdma_buffer_get(struct rpcrdma_buffer *);
 void rpcrdma_buffer_put(struct rpcrdma_buffer *buffers,
 			struct rpcrdma_req *req);
 void rpcrdma_rep_put(struct rpcrdma_buffer *buf, struct rpcrdma_rep *rep);
+void rpcrdma_reply_put(struct rpcrdma_buffer *buffers, struct rpcrdma_req *req);
 
 bool rpcrdma_regbuf_realloc(struct rpcrdma_regbuf *rb, size_t size,
 			    gfp_t flags);
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
index 47aa47a2b07c..316d04945587 100644
--- a/net/sunrpc/xprtsock.c
+++ b/net/sunrpc/xprtsock.c
@@ -1010,6 +1010,8 @@ static int xs_tcp_send_request(struct rpc_rqst *req)
 			kernel_sock_shutdown(transport->sock, SHUT_RDWR);
 		return -ENOTCONN;
 	}
+	if (!transport->inet)
+		return -ENOTCONN;
 
 	xs_pktdump("packet data:",
 				req->rq_svec->iov_base,