Merge branch 'sunrpc'

* sunrpc:
  SUNRPC: Use MSG_SENDPAGE_NOTLAST in xs_send_pagedata()
  SUNRPC: Move AF_LOCAL receive data path into a workqueue context
  SUNRPC: Move UDP receive data path into a workqueue context
  SUNRPC: Move TCP receive data path into a workqueue context
  SUNRPC: Refactor TCP receive

include/linux/sunrpc/xprtsock.h

@@ -44,6 +44,8 @@ struct sock_xprt {
 	 */
 	unsigned long		sock_state;
 	struct delayed_work	connect_worker;
+	struct work_struct	recv_worker;
+	struct mutex		recv_mutex;
 	struct sockaddr_storage	srcaddr;
 	unsigned short		srcport;
 

net/sunrpc/xprtsock.c

@@ -360,8 +360,10 @@ static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned i
 		int flags = XS_SENDMSG_FLAGS;
 
 		remainder -= len;
-		if (remainder != 0 || more)
+		if (more)
 			flags |= MSG_MORE;
+		if (remainder != 0)
+			flags |= MSG_SENDPAGE_NOTLAST | MSG_MORE;
 		err = do_sendpage(sock, *ppage, base, len, flags);
 		if (remainder == 0 || err != len)
 			break;
@@ -823,6 +825,7 @@ static void xs_reset_transport(struct sock_xprt *transport)
 
 	kernel_sock_shutdown(sock, SHUT_RDWR);
 
+	mutex_lock(&transport->recv_mutex);
 	write_lock_bh(&sk->sk_callback_lock);
 	transport->inet = NULL;
 	transport->sock = NULL;
@@ -833,6 +836,7 @@ static void xs_reset_transport(struct sock_xprt *transport)
 	xprt_clear_connected(xprt);
 	write_unlock_bh(&sk->sk_callback_lock);
 	xs_sock_reset_connection_flags(xprt);
+	mutex_unlock(&transport->recv_mutex);
 
 	trace_rpc_socket_close(xprt, sock);
 	sock_release(sock);
@@ -886,6 +890,7 @@ static void xs_destroy(struct rpc_xprt *xprt)
 
 	cancel_delayed_work_sync(&transport->connect_worker);
 	xs_close(xprt);
+	cancel_work_sync(&transport->recv_worker);
 	xs_xprt_free(xprt);
 	module_put(THIS_MODULE);
 }
@@ -906,44 +911,36 @@ static int xs_local_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
 }
 
 /**
- * xs_local_data_ready - "data ready" callback for AF_LOCAL sockets
- * @sk: socket with data to read
+ * xs_local_data_read_skb
+ * @xprt: transport
+ * @sk: socket
+ * @skb: skbuff
  *
  * Currently this assumes we can read the whole reply in a single gulp.
  */
-static void xs_local_data_ready(struct sock *sk)
+static void xs_local_data_read_skb(struct rpc_xprt *xprt,
+		struct sock *sk,
+		struct sk_buff *skb)
 {
 	struct rpc_task *task;
-	struct rpc_xprt *xprt;
 	struct rpc_rqst *rovr;
-	struct sk_buff *skb;
-	int err, repsize, copied;
+	int repsize, copied;
 	u32 _xid;
 	__be32 *xp;
 
-	read_lock_bh(&sk->sk_callback_lock);
-	dprintk("RPC:       %s...\n", __func__);
-	xprt = xprt_from_sock(sk);
-	if (xprt == NULL)
-		goto out;
-
-	skb = skb_recv_datagram(sk, 0, 1, &err);
-	if (skb == NULL)
-		goto out;
-
 	repsize = skb->len - sizeof(rpc_fraghdr);
 	if (repsize < 4) {
 		dprintk("RPC:       impossible RPC reply size %d\n", repsize);
-		goto dropit;
+		return;
 	}
 
 	/* Copy the XID from the skb... */
 	xp = skb_header_pointer(skb, sizeof(rpc_fraghdr), sizeof(_xid), &_xid);
 	if (xp == NULL)
-		goto dropit;
+		return;
 
 	/* Look up and lock the request corresponding to the given XID */
-	spin_lock(&xprt->transport_lock);
+	spin_lock_bh(&xprt->transport_lock);
 	rovr = xprt_lookup_rqst(xprt, *xp);
 	if (!rovr)
 		goto out_unlock;
@@ -961,50 +958,68 @@ static void xs_local_data_ready(struct sock *sk)
 	xprt_complete_rqst(task, copied);
 
  out_unlock:
-	spin_unlock(&xprt->transport_lock);
- dropit:
-	skb_free_datagram(sk, skb);
- out:
-	read_unlock_bh(&sk->sk_callback_lock);
+	spin_unlock_bh(&xprt->transport_lock);
+}
+
+static void xs_local_data_receive(struct sock_xprt *transport)
+{
+	struct sk_buff *skb;
+	struct sock *sk;
+	int err;
+
+	mutex_lock(&transport->recv_mutex);
+	sk = transport->inet;
+	if (sk == NULL)
+		goto out;
+	for (;;) {
+		skb = skb_recv_datagram(sk, 0, 1, &err);
+		if (skb == NULL)
+			break;
+		xs_local_data_read_skb(&transport->xprt, sk, skb);
+		skb_free_datagram(sk, skb);
+	}
+out:
+	mutex_unlock(&transport->recv_mutex);
+}
+
+static void xs_local_data_receive_workfn(struct work_struct *work)
+{
+	struct sock_xprt *transport =
+		container_of(work, struct sock_xprt, recv_worker);
+	xs_local_data_receive(transport);
 }
 
 /**
- * xs_udp_data_ready - "data ready" callback for UDP sockets
- * @sk: socket with data to read
+ * xs_udp_data_read_skb - receive callback for UDP sockets
+ * @xprt: transport
+ * @sk: socket
+ * @skb: skbuff
  *
  */
-static void xs_udp_data_ready(struct sock *sk)
+static void xs_udp_data_read_skb(struct rpc_xprt *xprt,
+		struct sock *sk,
+		struct sk_buff *skb)
 {
 	struct rpc_task *task;
-	struct rpc_xprt *xprt;
 	struct rpc_rqst *rovr;
-	struct sk_buff *skb;
-	int err, repsize, copied;
+	int repsize, copied;
 	u32 _xid;
 	__be32 *xp;
 
-	read_lock_bh(&sk->sk_callback_lock);
-	dprintk("RPC:       xs_udp_data_ready...\n");
-	if (!(xprt = xprt_from_sock(sk)))
-		goto out;
-
-	if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
-		goto out;
-
 	repsize = skb->len - sizeof(struct udphdr);
 	if (repsize < 4) {
 		dprintk("RPC:       impossible RPC reply size %d!\n", repsize);
-		goto dropit;
+		return;
 	}
 
 	/* Copy the XID from the skb... */
 	xp = skb_header_pointer(skb, sizeof(struct udphdr),
 				sizeof(_xid), &_xid);
 	if (xp == NULL)
-		goto dropit;
+		return;
 
 	/* Look up and lock the request corresponding to the given XID */
-	spin_lock(&xprt->transport_lock);
+	spin_lock_bh(&xprt->transport_lock);
 	rovr = xprt_lookup_rqst(xprt, *xp);
 	if (!rovr)
 		goto out_unlock;
@@ -1025,10 +1040,54 @@ static void xs_udp_data_ready(struct sock *sk)
 	xprt_complete_rqst(task, copied);
 
  out_unlock:
-	spin_unlock(&xprt->transport_lock);
- dropit:
-	skb_free_datagram(sk, skb);
- out:
+	spin_unlock_bh(&xprt->transport_lock);
+}
+
+static void xs_udp_data_receive(struct sock_xprt *transport)
+{
+	struct sk_buff *skb;
+	struct sock *sk;
+	int err;
+
+	mutex_lock(&transport->recv_mutex);
+	sk = transport->inet;
+	if (sk == NULL)
+		goto out;
+	for (;;) {
+		skb = skb_recv_datagram(sk, 0, 1, &err);
+		if (skb == NULL)
+			break;
+		xs_udp_data_read_skb(&transport->xprt, sk, skb);
+		skb_free_datagram(sk, skb);
+	}
+out:
+	mutex_unlock(&transport->recv_mutex);
+}
+
+static void xs_udp_data_receive_workfn(struct work_struct *work)
+{
+	struct sock_xprt *transport =
+		container_of(work, struct sock_xprt, recv_worker);
+	xs_udp_data_receive(transport);
+}
+
+/**
+ * xs_data_ready - "data ready" callback for UDP sockets
+ * @sk: socket with data to read
+ *
+ */
+static void xs_data_ready(struct sock *sk)
+{
+	struct rpc_xprt *xprt;
+
+	read_lock_bh(&sk->sk_callback_lock);
+	dprintk("RPC:       xs_data_ready...\n");
+	xprt = xprt_from_sock(sk);
+	if (xprt != NULL) {
+		struct sock_xprt *transport = container_of(xprt,
+				struct sock_xprt, xprt);
+		queue_work(rpciod_workqueue, &transport->recv_worker);
+	}
 	read_unlock_bh(&sk->sk_callback_lock);
 }
 
@@ -1243,12 +1302,12 @@ static inline int xs_tcp_read_reply(struct rpc_xprt *xprt,
 	dprintk("RPC:       read reply XID %08x\n", ntohl(transport->tcp_xid));
 
 	/* Find and lock the request corresponding to this xid */
-	spin_lock(&xprt->transport_lock);
+	spin_lock_bh(&xprt->transport_lock);
 	req = xprt_lookup_rqst(xprt, transport->tcp_xid);
 	if (!req) {
 		dprintk("RPC:       XID %08x request not found!\n",
 				ntohl(transport->tcp_xid));
-		spin_unlock(&xprt->transport_lock);
+		spin_unlock_bh(&xprt->transport_lock);
 		return -1;
 	}
 
@@ -1257,7 +1316,7 @@ static inline int xs_tcp_read_reply(struct rpc_xprt *xprt,
 	if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
 		xprt_complete_rqst(req->rq_task, transport->tcp_copied);
 
-	spin_unlock(&xprt->transport_lock);
+	spin_unlock_bh(&xprt->transport_lock);
 	return 0;
 }
 
@@ -1277,10 +1336,10 @@ static int xs_tcp_read_callback(struct rpc_xprt *xprt,
 	struct rpc_rqst *req;
 
 	/* Look up and lock the request corresponding to the given XID */
-	spin_lock(&xprt->transport_lock);
+	spin_lock_bh(&xprt->transport_lock);
 	req = xprt_lookup_bc_request(xprt, transport->tcp_xid);
 	if (req == NULL) {
-		spin_unlock(&xprt->transport_lock);
+		spin_unlock_bh(&xprt->transport_lock);
 		printk(KERN_WARNING "Callback slot table overflowed\n");
 		xprt_force_disconnect(xprt);
 		return -1;
@@ -1291,7 +1350,7 @@ static int xs_tcp_read_callback(struct rpc_xprt *xprt,
 
 	if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
 		xprt_complete_bc_request(req, transport->tcp_copied);
-	spin_unlock(&xprt->transport_lock);
+	spin_unlock_bh(&xprt->transport_lock);
 
 	return 0;
 }
@@ -1391,6 +1450,44 @@ static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, uns
 	return len - desc.count;
 }
 
+static void xs_tcp_data_receive(struct sock_xprt *transport)
+{
+	struct rpc_xprt *xprt = &transport->xprt;
+	struct sock *sk;
+	read_descriptor_t rd_desc = {
+		.count = 2*1024*1024,
+		.arg.data = xprt,
+	};
+	unsigned long total = 0;
+	int read = 0;
+
+	mutex_lock(&transport->recv_mutex);
+	sk = transport->inet;
+	if (sk == NULL)
+		goto out;
+
+	/* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
+	for (;;) {
+		lock_sock(sk);
+		read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
+		release_sock(sk);
+		if (read <= 0)
+			break;
+		total += read;
+		rd_desc.count = 65536;
+	}
+out:
+	mutex_unlock(&transport->recv_mutex);
+	trace_xs_tcp_data_ready(xprt, read, total);
+}
+
+static void xs_tcp_data_receive_workfn(struct work_struct *work)
+{
+	struct sock_xprt *transport =
+		container_of(work, struct sock_xprt, recv_worker);
+	xs_tcp_data_receive(transport);
+}
+
 /**
  * xs_tcp_data_ready - "data ready" callback for TCP sockets
  * @sk: socket with data to read
@@ -1398,34 +1495,24 @@ static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, uns
  */
 static void xs_tcp_data_ready(struct sock *sk)
 {
+	struct sock_xprt *transport;
 	struct rpc_xprt *xprt;
-	read_descriptor_t rd_desc;
-	int read;
-	unsigned long total = 0;
 
 	dprintk("RPC:       xs_tcp_data_ready...\n");
 
 	read_lock_bh(&sk->sk_callback_lock);
-	if (!(xprt = xprt_from_sock(sk))) {
-		read = 0;
+	if (!(xprt = xprt_from_sock(sk)))
 		goto out;
-	}
+	transport = container_of(xprt, struct sock_xprt, xprt);
+
 	/* Any data means we had a useful conversation, so
 	 * the we don't need to delay the next reconnect
 	 */
 	if (xprt->reestablish_timeout)
 		xprt->reestablish_timeout = 0;
+	queue_work(rpciod_workqueue, &transport->recv_worker);
 
-	/* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
-	rd_desc.arg.data = xprt;
-	do {
-		rd_desc.count = 65536;
-		read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
-		if (read > 0)
-			total += read;
-	} while (read > 0);
 out:
-	trace_xs_tcp_data_ready(xprt, read, total);
 	read_unlock_bh(&sk->sk_callback_lock);
 }
 
@@ -1873,7 +1960,7 @@ static int xs_local_finish_connecting(struct rpc_xprt *xprt,
 		xs_save_old_callbacks(transport, sk);
 
 		sk->sk_user_data = xprt;
-		sk->sk_data_ready = xs_local_data_ready;
+		sk->sk_data_ready = xs_data_ready;
 		sk->sk_write_space = xs_udp_write_space;
 		sk->sk_error_report = xs_error_report;
 		sk->sk_allocation = GFP_NOIO;
@@ -2059,7 +2146,7 @@ static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
 		xs_save_old_callbacks(transport, sk);
 
 		sk->sk_user_data = xprt;
-		sk->sk_data_ready = xs_udp_data_ready;
+		sk->sk_data_ready = xs_data_ready;
 		sk->sk_write_space = xs_udp_write_space;
 		sk->sk_allocation = GFP_NOIO;
 
@@ -2650,6 +2737,7 @@ static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
 	}
 
 	new = container_of(xprt, struct sock_xprt, xprt);
+	mutex_init(&new->recv_mutex);
 	memcpy(&xprt->addr, args->dstaddr, args->addrlen);
 	xprt->addrlen = args->addrlen;
 	if (args->srcaddr)
@@ -2703,6 +2791,7 @@ static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
 	xprt->ops = &xs_local_ops;
 	xprt->timeout = &xs_local_default_timeout;
 
+	INIT_WORK(&transport->recv_worker, xs_local_data_receive_workfn);
 	INIT_DELAYED_WORK(&transport->connect_worker,
 			xs_dummy_setup_socket);
 
@@ -2774,21 +2863,20 @@ static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
 
 	xprt->timeout = &xs_udp_default_timeout;
 
+	INIT_WORK(&transport->recv_worker, xs_udp_data_receive_workfn);
+	INIT_DELAYED_WORK(&transport->connect_worker, xs_udp_setup_socket);
+
 	switch (addr->sa_family) {
 	case AF_INET:
 		if (((struct sockaddr_in *)addr)->sin_port != htons(0))
 			xprt_set_bound(xprt);
 
-		INIT_DELAYED_WORK(&transport->connect_worker,
-					xs_udp_setup_socket);
 		xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
 		break;
 	case AF_INET6:
 		if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
 			xprt_set_bound(xprt);
 
-		INIT_DELAYED_WORK(&transport->connect_worker,
-					xs_udp_setup_socket);
 		xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
 		break;
 	default:
@@ -2853,21 +2941,20 @@ static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
 	xprt->ops = &xs_tcp_ops;
 	xprt->timeout = &xs_tcp_default_timeout;
 
+	INIT_WORK(&transport->recv_worker, xs_tcp_data_receive_workfn);
+	INIT_DELAYED_WORK(&transport->connect_worker, xs_tcp_setup_socket);
+
 	switch (addr->sa_family) {
 	case AF_INET:
 		if (((struct sockaddr_in *)addr)->sin_port != htons(0))
 			xprt_set_bound(xprt);
 
-		INIT_DELAYED_WORK(&transport->connect_worker,
-					xs_tcp_setup_socket);
 		xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
 		break;
 	case AF_INET6:
 		if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
 			xprt_set_bound(xprt);
 
-		INIT_DELAYED_WORK(&transport->connect_worker,
-					xs_tcp_setup_socket);
 		xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
 		break;
 	default: