1 files changed, 49 insertions, 353 deletions

diff --git a/lib/librte_eal/linuxapp/eal/eal_vfio_mp_sync.c b/lib/librte_eal/linuxapp/eal/eal_vfio_mp_sync.c
index 7cc3c152..680a24aa 100644
--- a/lib/librte_eal/linuxapp/eal/eal_vfio_mp_sync.c
+++ b/lib/librte_eal/linuxapp/eal/eal_vfio_mp_sync.c
@@ -1,32 +1,16 @@
 /* SPDX-License-Identifier: BSD-3-Clause
- * Copyright(c) 2010-2014 Intel Corporation
+ * Copyright(c) 2010-2018 Intel Corporation
  */
 
+#include <unistd.h>
 #include <string.h>
-#include <fcntl.h>
-#include <sys/socket.h>
-#include <pthread.h>
-
-/* sys/un.h with __USE_MISC uses strlen, which is unsafe */
-#ifdef __USE_MISC
-#define REMOVED_USE_MISC
-#undef __USE_MISC
-#endif
-#include <sys/un.h>
-/* make sure we redefine __USE_MISC only if it was previously undefined */
-#ifdef REMOVED_USE_MISC
-#define __USE_MISC
-#undef REMOVED_USE_MISC
-#endif
 
+#include <rte_compat.h>
 #include <rte_log.h>
-#include <rte_eal_memconfig.h>
-#include <rte_malloc.h>
 #include <rte_vfio.h>
+#include <rte_eal.h>
 
-#include "eal_filesystem.h"
 #include "eal_vfio.h"
-#include "eal_thread.h"
 
 /**
  * @file
@@ -37,358 +21,70 @@
 
 #ifdef VFIO_PRESENT
 
-#define SOCKET_PATH_FMT "%s/.%s_mp_socket"
-#define CMSGLEN (CMSG_LEN(sizeof(int)))
-#define FD_TO_CMSGHDR(fd, chdr) \
-		do {\
-			(chdr).cmsg_len = CMSGLEN;\
-			(chdr).cmsg_level = SOL_SOCKET;\
-			(chdr).cmsg_type = SCM_RIGHTS;\
-			memcpy((chdr).__cmsg_data, &(fd), sizeof(fd));\
-		} while (0)
-#define CMSGHDR_TO_FD(chdr, fd) \
-			memcpy(&(fd), (chdr).__cmsg_data, sizeof(fd))
-
-static pthread_t socket_thread;
-static int mp_socket_fd;
-
-
-/* get socket path (/var/run if root, $HOME otherwise) */
-static void
-get_socket_path(char *buffer, int bufsz)
-{
-	const char *dir = "/var/run";
-	const char *home_dir = getenv("HOME");
-
-	if (getuid() != 0 && home_dir != NULL)
-		dir = home_dir;
-
-	/* use current prefix as file path */
-	snprintf(buffer, bufsz, SOCKET_PATH_FMT, dir,
-			internal_config.hugefile_prefix);
-}
-
-
-
-/*
- * data flow for socket comm protocol:
- * 1. client sends SOCKET_REQ_CONTAINER or SOCKET_REQ_GROUP
- * 1a. in case of SOCKET_REQ_GROUP, client also then sends group number
- * 2. server receives message
- * 2a. in case of invalid group, SOCKET_ERR is sent back to client
- * 2b. in case of unbound group, SOCKET_NO_FD is sent back to client
- * 2c. in case of valid group, SOCKET_OK is sent and immediately followed by fd
- *
- * in case of any error, socket is closed.
- */
-
-/* send a request, return -1 on error */
-int
-vfio_mp_sync_send_request(int socket, int req)
-{
-	struct msghdr hdr;
-	struct iovec iov;
-	int buf;
-	int ret;
-
-	memset(&hdr, 0, sizeof(hdr));
-
-	buf = req;
-
-	hdr.msg_iov = &iov;
-	hdr.msg_iovlen = 1;
-	iov.iov_base = (char *) &buf;
-	iov.iov_len = sizeof(buf);
-
-	ret = sendmsg(socket, &hdr, 0);
-	if (ret < 0)
-		return -1;
-	return 0;
-}
-
-/* receive a request and return it */
-int
-vfio_mp_sync_receive_request(int socket)
-{
-	int buf;
-	struct msghdr hdr;
-	struct iovec iov;
-	int ret, req;
-
-	memset(&hdr, 0, sizeof(hdr));
-
-	buf = SOCKET_ERR;
-
-	hdr.msg_iov = &iov;
-	hdr.msg_iovlen = 1;
-	iov.iov_base = (char *) &buf;
-	iov.iov_len = sizeof(buf);
-
-	ret = recvmsg(socket, &hdr, 0);
-	if (ret < 0)
-		return -1;
-
-	req = buf;
-
-	return req;
-}
-
-/* send OK in message, fd in control message */
-int
-vfio_mp_sync_send_fd(int socket, int fd)
+static int
+vfio_mp_primary(const struct rte_mp_msg *msg, const void *peer)
 {
-	int buf;
-	struct msghdr hdr;
-	struct cmsghdr *chdr;
-	char chdr_buf[CMSGLEN];
-	struct iovec iov;
+	int fd = -1;
 	int ret;
+	struct rte_mp_msg reply;
+	struct vfio_mp_param *r = (struct vfio_mp_param *)reply.param;
+	const struct vfio_mp_param *m =
+		(const struct vfio_mp_param *)msg->param;
 
-	chdr = (struct cmsghdr *) chdr_buf;
-	memset(chdr, 0, sizeof(chdr_buf));
-	memset(&hdr, 0, sizeof(hdr));
-
-	hdr.msg_iov = &iov;
-	hdr.msg_iovlen = 1;
-	iov.iov_base = (char *) &buf;
-	iov.iov_len = sizeof(buf);
-	hdr.msg_control = chdr;
-	hdr.msg_controllen = CMSGLEN;
-
-	buf = SOCKET_OK;
-	FD_TO_CMSGHDR(fd, *chdr);
-
-	ret = sendmsg(socket, &hdr, 0);
-	if (ret < 0)
-		return -1;
-	return 0;
-}
-
-/* receive OK in message, fd in control message */
-int
-vfio_mp_sync_receive_fd(int socket)
-{
-	int buf;
-	struct msghdr hdr;
-	struct cmsghdr *chdr;
-	char chdr_buf[CMSGLEN];
-	struct iovec iov;
-	int ret, req, fd;
-
-	buf = SOCKET_ERR;
-
-	chdr = (struct cmsghdr *) chdr_buf;
-	memset(chdr, 0, sizeof(chdr_buf));
-	memset(&hdr, 0, sizeof(hdr));
-
-	hdr.msg_iov = &iov;
-	hdr.msg_iovlen = 1;
-	iov.iov_base = (char *) &buf;
-	iov.iov_len = sizeof(buf);
-	hdr.msg_control = chdr;
-	hdr.msg_controllen = CMSGLEN;
-
-	ret = recvmsg(socket, &hdr, 0);
-	if (ret < 0)
-		return -1;
-
-	req = buf;
-
-	if (req != SOCKET_OK)
-		return -1;
-
-	CMSGHDR_TO_FD(*chdr, fd);
-
-	return fd;
-}
-
-/* connect socket_fd in secondary process to the primary process's socket */
-int
-vfio_mp_sync_connect_to_primary(void)
-{
-	struct sockaddr_un addr;
-	socklen_t sockaddr_len;
-	int socket_fd;
-
-	/* set up a socket */
-	socket_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
-	if (socket_fd < 0) {
-		RTE_LOG(ERR, EAL, "Failed to create socket!\n");
+	if (msg->len_param != sizeof(*m)) {
+		RTE_LOG(ERR, EAL, "vfio received invalid message!\n");
 		return -1;
 	}
 
-	get_socket_path(addr.sun_path, sizeof(addr.sun_path));
-	addr.sun_family = AF_UNIX;
-
-	sockaddr_len = sizeof(struct sockaddr_un);
-
-	if (connect(socket_fd, (struct sockaddr *) &addr, sockaddr_len) == 0)
-		return socket_fd;
-
-	/* if connect failed */
-	close(socket_fd);
-	return -1;
-}
-
+	memset(&reply, 0, sizeof(reply));
 
-
-/*
- * socket listening thread for primary process
- */
-static __attribute__((noreturn)) void *
-vfio_mp_sync_thread(void __rte_unused * arg)
-{
-	int ret, fd, vfio_data;
-
-	/* wait for requests on the socket */
-	for (;;) {
-		int conn_sock;
-		struct sockaddr_un addr;
-		socklen_t sockaddr_len = sizeof(addr);
-
-		/* this is a blocking call */
-		conn_sock = accept(mp_socket_fd, (struct sockaddr *) &addr,
-				&sockaddr_len);
-
-		/* just restart on error */
-		if (conn_sock == -1)
-			continue;
-
-		/* set socket to linger after close */
-		struct linger l;
-		l.l_onoff = 1;
-		l.l_linger = 60;
-
-		if (setsockopt(conn_sock, SOL_SOCKET, SO_LINGER, &l, sizeof(l)) < 0)
-			RTE_LOG(WARNING, EAL, "Cannot set SO_LINGER option "
-					"on listen socket (%s)\n", strerror(errno));
-
-		ret = vfio_mp_sync_receive_request(conn_sock);
-
-		switch (ret) {
-		case SOCKET_REQ_CONTAINER:
-			fd = vfio_get_container_fd();
-			if (fd < 0)
-				vfio_mp_sync_send_request(conn_sock, SOCKET_ERR);
-			else
-				vfio_mp_sync_send_fd(conn_sock, fd);
-			if (fd >= 0)
-				close(fd);
-			break;
-		case SOCKET_REQ_GROUP:
-			/* wait for group number */
-			vfio_data = vfio_mp_sync_receive_request(conn_sock);
-			if (vfio_data < 0) {
-				close(conn_sock);
-				continue;
-			}
-
-			fd = vfio_get_group_fd(vfio_data);
-
-			if (fd < 0)
-				vfio_mp_sync_send_request(conn_sock, SOCKET_ERR);
+	switch (m->req) {
+	case SOCKET_REQ_GROUP:
+		r->req = SOCKET_REQ_GROUP;
+		r->group_num = m->group_num;
+		fd = rte_vfio_get_group_fd(m->group_num);
+		if (fd < 0)
+			r->result = SOCKET_ERR;
+		else if (fd == 0)
 			/* if VFIO group exists but isn't bound to VFIO driver */
-			else if (fd == 0)
-				vfio_mp_sync_send_request(conn_sock, SOCKET_NO_FD);
+			r->result = SOCKET_NO_FD;
+		else {
 			/* if group exists and is bound to VFIO driver */
-			else {
-				vfio_mp_sync_send_request(conn_sock, SOCKET_OK);
-				vfio_mp_sync_send_fd(conn_sock, fd);
-			}
-			break;
-		case SOCKET_CLR_GROUP:
-			/* wait for group fd */
-			vfio_data = vfio_mp_sync_receive_request(conn_sock);
-			if (vfio_data < 0) {
-				close(conn_sock);
-				continue;
-			}
-
-			ret = rte_vfio_clear_group(vfio_data);
-
-			if (ret < 0)
-				vfio_mp_sync_send_request(conn_sock, SOCKET_NO_FD);
-			else
-				vfio_mp_sync_send_request(conn_sock, SOCKET_OK);
-			break;
-		default:
-			vfio_mp_sync_send_request(conn_sock, SOCKET_ERR);
-			break;
+			r->result = SOCKET_OK;
+			reply.num_fds = 1;
+			reply.fds[0] = fd;
 		}
-		close(conn_sock);
-	}
-}
-
-static int
-vfio_mp_sync_socket_setup(void)
-{
-	int ret, socket_fd;
-	struct sockaddr_un addr;
-	socklen_t sockaddr_len;
-
-	/* set up a socket */
-	socket_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
-	if (socket_fd < 0) {
-		RTE_LOG(ERR, EAL, "Failed to create socket!\n");
-		return -1;
-	}
-
-	get_socket_path(addr.sun_path, sizeof(addr.sun_path));
-	addr.sun_family = AF_UNIX;
-
-	sockaddr_len = sizeof(struct sockaddr_un);
-
-	unlink(addr.sun_path);
-
-	ret = bind(socket_fd, (struct sockaddr *) &addr, sockaddr_len);
-	if (ret) {
-		RTE_LOG(ERR, EAL, "Failed to bind socket: %s!\n", strerror(errno));
-		close(socket_fd);
-		return -1;
-	}
-
-	ret = listen(socket_fd, 50);
-	if (ret) {
-		RTE_LOG(ERR, EAL, "Failed to listen: %s!\n", strerror(errno));
-		close(socket_fd);
+		break;
+	case SOCKET_REQ_CONTAINER:
+		r->req = SOCKET_REQ_CONTAINER;
+		fd = rte_vfio_get_container_fd();
+		if (fd < 0)
+			r->result = SOCKET_ERR;
+		else {
+			r->result = SOCKET_OK;
+			reply.num_fds = 1;
+			reply.fds[0] = fd;
+		}
+		break;
+	default:
+		RTE_LOG(ERR, EAL, "vfio received invalid message!\n");
 		return -1;
 	}
 
-	/* save the socket in local configuration */
-	mp_socket_fd = socket_fd;
+	strcpy(reply.name, EAL_VFIO_MP);
+	reply.len_param = sizeof(*r);
 
-	return 0;
+	ret = rte_mp_reply(&reply, peer);
+	if (m->req == SOCKET_REQ_CONTAINER && fd >= 0)
+		close(fd);
+	return ret;
 }
 
-/*
- * set up a local socket and tell it to listen for incoming connections
- */
 int
 vfio_mp_sync_setup(void)
 {
-	int ret;
-	char thread_name[RTE_MAX_THREAD_NAME_LEN];
-
-	if (vfio_mp_sync_socket_setup() < 0) {
-		RTE_LOG(ERR, EAL, "Failed to set up local socket!\n");
-		return -1;
-	}
-
-	ret = pthread_create(&socket_thread, NULL,
-			vfio_mp_sync_thread, NULL);
-	if (ret) {
-		RTE_LOG(ERR, EAL,
-			"Failed to create thread for communication with secondary processes!\n");
-		close(mp_socket_fd);
-		return -1;
-	}
-
-	/* Set thread_name for aid in debugging. */
-	snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN, "vfio-sync");
-	ret = rte_thread_setname(socket_thread, thread_name);
-	if (ret)
-		RTE_LOG(DEBUG, EAL,
-			"Failed to set thread name for secondary processes!\n");
+	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
+		return rte_mp_action_register(EAL_VFIO_MP, vfio_mp_primary);
 
 	return 0;
 }