vhost: add vhost-user client mode

Add a new paramter (flags) to rte_vhost_driver_register(). DPDK
vhost-user acts as client mode when RTE_VHOST_USER_CLIENT flag
is set.  The flags would also allow future extensions without
breaking the API (again).

The rest is straingfoward then: allocate a unix socket, and
bind/listen for server, connect for client.

This extension is for vhost-user only, therefore we simply quit
and report error when any flags are given for vhost-cuse.

Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>

doc/guides/rel_notes/release_16_07.rst

@@ -51,6 +51,18 @@ New Features
 
   The ioports are mapped in memory when using Linux UIO.
 
+* **Added vhost-user client mode.**
+
+  DPDK vhost-user could be the server as well as the client. It supports
+  server mode only before, now it also supports client mode. Client mode
+  is enabled when ``RTE_VHOST_USER_CLIENT`` flag is set while calling
+  ``rte_vhost_driver_register``.
+
+  When DPDK vhost-user restarts from normal or abnormal quit (say crash),
+  the client mode would allow DPDK to establish the connect again.  Note
+  that a brand new QEMU version (v2.7 or above) is needed, otherwise, the
+  reconnect won't work.
+
 * **Added AES-CTR support to AESNI MB PMD.**
 
   Now AESNI MB PMD supports 128/192/256-bit counter mode AES encryption and

drivers/net/vhost/rte_eth_vhost.c

@@ -467,7 +467,7 @@ eth_dev_start(struct rte_eth_dev *dev)
 	int ret = 0;
 
 	if (rte_atomic16_cmpset(&internal->once, 0, 1)) {
-		ret = rte_vhost_driver_register(internal->iface_name);
+		ret = rte_vhost_driver_register(internal->iface_name, 0);
 		if (ret)
 			return ret;
 	}

examples/tep_termination/main.c

@@ -1254,7 +1254,7 @@ main(int argc, char *argv[])
 	rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
 
 	/* Register CUSE device to handle IOCTLs. */
-	ret = rte_vhost_driver_register((char *)&dev_basename);
+	ret = rte_vhost_driver_register((char *)&dev_basename, 0);
 	if (ret != 0)
 		rte_exit(EXIT_FAILURE, "CUSE device setup failure.\n");
 

examples/vhost/main.c

@@ -1488,7 +1488,7 @@ main(int argc, char *argv[])
 		rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
 
 	/* Register vhost(cuse or user) driver to handle vhost messages. */
-	ret = rte_vhost_driver_register((char *)&dev_basename);
+	ret = rte_vhost_driver_register(dev_basename, 0);
 	if (ret != 0)
 		rte_exit(EXIT_FAILURE, "vhost driver register failure.\n");
 

lib/librte_vhost/rte_virtio_net.h

@@ -51,6 +51,8 @@
 #include <rte_mempool.h>
 #include <rte_ether.h>
 
+#define RTE_VHOST_USER_CLIENT		(1ULL << 0)
+
 /* Enum for virtqueue management. */
 enum {VIRTIO_RXQ, VIRTIO_TXQ, VIRTIO_QNUM};
 
@@ -85,11 +87,14 @@ uint64_t rte_vhost_feature_get(void);
 
 int rte_vhost_enable_guest_notification(int vid, uint16_t queue_id, int enable);
 
-/* Register vhost driver. dev_name could be different for multiple instance support. */
+/**
-int rte_vhost_driver_register(const char *dev_name);
+ * Register vhost driver. path could be different for multiple
+ * instance support.
+ */
+int rte_vhost_driver_register(const char *path, uint64_t flags);
 
 /* Unregister vhost driver. This is only meaningful to vhost user. */
-int rte_vhost_driver_unregister(const char *dev_name);
+int rte_vhost_driver_unregister(const char *path);
 
 /* Register callbacks. */
 int rte_vhost_driver_callback_register(struct virtio_net_device_ops const * const);

lib/librte_vhost/vhost_cuse/vhost-net-cdev.c

@@ -352,7 +352,7 @@ static const struct cuse_lowlevel_ops vhost_net_ops = {
  * vhost_net_device_ops are also passed when the device is registered in app.
  */
 int
-rte_vhost_driver_register(const char *dev_name)
+rte_vhost_driver_register(const char *dev_name, uint64_t flags)
 {
 	struct cuse_info cuse_info;
 	char device_name[PATH_MAX] = "";
@@ -364,6 +364,12 @@ rte_vhost_driver_register(const char *dev_name)
 	char fuse_opt_nomulti[] = FUSE_OPT_NOMULTI;
 	char *fuse_argv[] = {fuse_opt_dummy, fuse_opt_fore, fuse_opt_nomulti};
 
+	if (flags) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"vhost-cuse does not support any flags so far\n");
+		return -1;
+	}
+
 	if (access(cuse_device_name, R_OK | W_OK) < 0) {
 		RTE_LOG(ERR, VHOST_CONFIG,
 			"char device %s can't be accessed, maybe not exist\n",

lib/librte_vhost/vhost_user/vhost-net-user.c

@@ -33,6 +33,7 @@
 
 #include <stdint.h>
 #include <stdio.h>
+#include <stdbool.h>
 #include <limits.h>
 #include <stdlib.h>
 #include <unistd.h>
@@ -58,6 +59,7 @@
 struct vhost_user_socket {
 	char *path;
 	int listenfd;
+	bool is_server;
 };
 
 struct vhost_user_connection {
@@ -75,7 +77,7 @@ struct vhost_user {
 
 #define MAX_VIRTIO_BACKLOG 128
 
-static void vhost_user_new_connection(int fd, void *data, int *remove);
+static void vhost_user_server_new_connection(int fd, void *data, int *remove);
 static void vhost_user_msg_handler(int fd, void *dat, int *remove);
 
 static struct vhost_user vhost_user = {
@@ -111,48 +113,6 @@ static const char *vhost_message_str[VHOST_USER_MAX] = {
 	[VHOST_USER_SEND_RARP]  = "VHOST_USER_SEND_RARP",
 };
 
-/**
- * Create a unix domain socket, bind to path and listen for connection.
- * @return
- *  socket fd or -1 on failure
- */
-static int
-uds_socket(const char *path)
-{
-	struct sockaddr_un un;
-	int sockfd;
-	int ret;
-
-	if (path == NULL)
-		return -1;
-
-	sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
-	if (sockfd < 0)
-		return -1;
-	RTE_LOG(INFO, VHOST_CONFIG, "socket created, fd:%d\n", sockfd);
-
-	memset(&un, 0, sizeof(un));
-	un.sun_family = AF_UNIX;
-	snprintf(un.sun_path, sizeof(un.sun_path), "%s", path);
-	ret = bind(sockfd, (struct sockaddr *)&un, sizeof(un));
-	if (ret == -1) {
-		RTE_LOG(ERR, VHOST_CONFIG, "fail to bind fd:%d, remove file:%s and try again.\n",
-			sockfd, path);
-		goto err;
-	}
-	RTE_LOG(INFO, VHOST_CONFIG, "bind to %s\n", path);
-
-	ret = listen(sockfd, MAX_VIRTIO_BACKLOG);
-	if (ret == -1)
-		goto err;
-
-	return sockfd;
-
-err:
-	close(sockfd);
-	return -1;
-}
-
 /* return bytes# of read on success or negative val on failure. */
 static int
 read_fd_message(int sockfd, char *buf, int buflen, int *fds, int fd_num)
@@ -287,32 +247,24 @@ send_vhost_message(int sockfd, struct VhostUserMsg *msg)
 	return ret;
 }
 
-/* call back when there is new vhost-user connection.  */
+
 static void
-vhost_user_new_connection(int fd, void *dat, int *remove __rte_unused)
+vhost_user_add_connection(int fd, struct vhost_user_socket *vsocket)
 {
-	struct vhost_user_socket *vsocket = dat;
-	int conn_fd;
-	struct vhost_user_connection *conn;
 	int vid;
-	unsigned int size;
+	size_t size;
+	struct vhost_user_connection *conn;
 
-	conn_fd = accept(fd, NULL, NULL);
+	conn = malloc(sizeof(*conn));
-	RTE_LOG(INFO, VHOST_CONFIG,
-		"new virtio connection is %d\n", conn_fd);
-	if (conn_fd < 0)
-		return;
-
-	conn = calloc(1, sizeof(*conn));
 	if (conn == NULL) {
-		close(conn_fd);
+		close(fd);
 		return;
 	}
 
 	vid = vhost_new_device();
 	if (vid == -1) {
+		close(fd);
 		free(conn);
-		close(conn_fd);
 		return;
 	}
 
@@ -323,8 +275,21 @@ vhost_user_new_connection(int fd, void *dat, int *remove __rte_unused)
 
 	conn->vsocket = vsocket;
 	conn->vid = vid;
-	fdset_add(&vhost_user.fdset,
+	fdset_add(&vhost_user.fdset, fd, vhost_user_msg_handler, NULL, conn);
-		conn_fd, vhost_user_msg_handler, NULL, conn);
+}
+
+/* call back when there is new vhost-user connection from client  */
+static void
+vhost_user_server_new_connection(int fd, void *dat, int *remove __rte_unused)
+{
+	struct vhost_user_socket *vsocket = dat;
+
+	fd = accept(fd, NULL, NULL);
+	if (fd < 0)
+		return;
+
+	RTE_LOG(INFO, VHOST_CONFIG, "new vhost user connection is %d\n", fd);
+	vhost_user_add_connection(fd, vsocket);
 }
 
 /* callback when there is message on the connfd */
@@ -452,50 +417,135 @@ vhost_user_msg_handler(int connfd, void *dat, int *remove)
 	}
 }
 
-/**
+static int
- * Creates and initialise the vhost server.
+create_unix_socket(const char *path, struct sockaddr_un *un, bool is_server)
+{
+	int fd;
+
+	fd = socket(AF_UNIX, SOCK_STREAM, 0);
+	if (fd < 0)
+		return -1;
+	RTE_LOG(INFO, VHOST_CONFIG, "vhost-user %s: socket created, fd: %d\n",
+		is_server ? "server" : "client", fd);
+
+	memset(un, 0, sizeof(*un));
+	un->sun_family = AF_UNIX;
+	strncpy(un->sun_path, path, sizeof(un->sun_path));
+
+	return fd;
+}
+
+static int
+vhost_user_create_server(struct vhost_user_socket *vsocket)
+{
+	int fd;
+	int ret;
+	struct sockaddr_un un;
+	const char *path = vsocket->path;
+
+	fd = create_unix_socket(path, &un, vsocket->is_server);
+	if (fd < 0)
+		return -1;
+
+	ret = bind(fd, (struct sockaddr *)&un, sizeof(un));
+	if (ret < 0) {
+		RTE_LOG(ERR, VHOST_CONFIG,
+			"failed to bind to %s: %s; remove it and try again\n",
+			path, strerror(errno));
+		goto err;
+	}
+	RTE_LOG(INFO, VHOST_CONFIG, "bind to %s\n", path);
+
+	ret = listen(fd, MAX_VIRTIO_BACKLOG);
+	if (ret < 0)
+		goto err;
+
+	vsocket->listenfd = fd;
+	fdset_add(&vhost_user.fdset, fd, vhost_user_server_new_connection,
+		  NULL, vsocket);
+
+	return 0;
+
+err:
+	close(fd);
+	return -1;
+}
+
+static int
+vhost_user_create_client(struct vhost_user_socket *vsocket)
+{
+	int fd;
+	int ret;
+	struct sockaddr_un un;
+	const char *path = vsocket->path;
+
+	fd = create_unix_socket(path, &un, vsocket->is_server);
+	if (fd < 0)
+		return -1;
+
+	ret = connect(fd, (struct sockaddr *)&un, sizeof(un));
+	if (ret < 0) {
+		RTE_LOG(ERR, VHOST_CONFIG, "failed to connect to %s: %s\n",
+			path, strerror(errno));
+		close(fd);
+		return -1;
+	}
+
+	vhost_user_add_connection(fd, vsocket);
+
+	return 0;
+}
+
+/*
+ * Register a new vhost-user socket; here we could act as server
+ * (the default case), or client (when RTE_VHOST_USER_CLIENT) flag
+ * is set.
  */
 int
-rte_vhost_driver_register(const char *path)
+rte_vhost_driver_register(const char *path, uint64_t flags)
 {
+	int ret = -1;
 	struct vhost_user_socket *vsocket;
 
+	if (!path)
+		return -1;
+
 	pthread_mutex_lock(&vhost_user.mutex);
 
 	if (vhost_user.vsocket_cnt == MAX_VHOST_SOCKET) {
 		RTE_LOG(ERR, VHOST_CONFIG,
-			"error: the number of servers reaches maximum\n");
+			"error: the number of vhost sockets reaches maximum\n");
-		pthread_mutex_unlock(&vhost_user.mutex);
+		goto out;
-		return -1;
-	}
-
-	vsocket = calloc(sizeof(struct vhost_user_socket), 1);
-	if (vsocket == NULL) {
-		pthread_mutex_unlock(&vhost_user.mutex);
-		return -1;
-	}
-
-	vsocket->listenfd = uds_socket(path);
-	if (vsocket->listenfd < 0) {
-		free(vsocket);
-		pthread_mutex_unlock(&vhost_user.mutex);
-		return -1;
 	}
 
+	vsocket = malloc(sizeof(struct vhost_user_socket));
+	if (!vsocket)
+		goto out;
+	memset(vsocket, 0, sizeof(struct vhost_user_socket));
 	vsocket->path = strdup(path);
 
-	fdset_add(&vhost_user.fdset, vsocket->listenfd,
+	if ((flags & RTE_VHOST_USER_CLIENT) != 0) {
-		vhost_user_new_connection, NULL, vsocket);
+		ret = vhost_user_create_client(vsocket);
+	} else {
+		vsocket->is_server = true;
+		ret = vhost_user_create_server(vsocket);
+	}
+	if (ret < 0) {
+		free(vsocket->path);
+		free(vsocket);
+		goto out;
+	}
 
 	vhost_user.vsockets[vhost_user.vsocket_cnt++] = vsocket;
+
+out:
 	pthread_mutex_unlock(&vhost_user.mutex);
 
-	return 0;
+	return ret;
 }
 
-
 /**
- * Unregister the specified vhost server
+ * Unregister the specified vhost socket
  */
 int
 rte_vhost_driver_unregister(const char *path)
@@ -507,15 +557,16 @@ rte_vhost_driver_unregister(const char *path)
 
 	for (i = 0; i < vhost_user.vsocket_cnt; i++) {
 		if (!strcmp(vhost_user.vsockets[i]->path, path)) {
-			fdset_del(&vhost_user.fdset,
+			if (vhost_user.vsockets[i]->is_server) {
-				vhost_user.vsockets[i]->listenfd);
+				fdset_del(&vhost_user.fdset,
+					vhost_user.vsockets[i]->listenfd);
+				close(vhost_user.vsockets[i]->listenfd);
+				unlink(path);
+			}
 
-			close(vhost_user.vsockets[i]->listenfd);
 			free(vhost_user.vsockets[i]->path);
 			free(vhost_user.vsockets[i]);
 
-			unlink(path);
-
 			count = --vhost_user.vsocket_cnt;
 			vhost_user.vsockets[i] = vhost_user.vsockets[count];
 			vhost_user.vsockets[count] = NULL;