4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 #include <sys/socket.h>
38 /* sys/un.h with __USE_MISC uses strlen, which is unsafe */
40 #define REMOVED_USE_MISC
44 /* make sure we redefine __USE_MISC only if it was previously undefined */
45 #ifdef REMOVED_USE_MISC
47 #undef REMOVED_USE_MISC
52 #include <rte_tailq.h>
53 #include <rte_eal_memconfig.h>
54 #include <rte_malloc.h>
56 #include "eal_filesystem.h"
57 #include "eal_pci_init.h"
61 * VFIO socket for communication between primary and secondary processes.
63 * This file is only compiled if CONFIG_RTE_EAL_VFIO is set to "y".
68 #define SOCKET_PATH_FMT "%s/.%s_mp_socket"
69 #define CMSGLEN (CMSG_LEN(sizeof(int)))
70 #define FD_TO_CMSGHDR(fd, chdr) \
72 (chdr).cmsg_len = CMSGLEN;\
73 (chdr).cmsg_level = SOL_SOCKET;\
74 (chdr).cmsg_type = SCM_RIGHTS;\
75 memcpy((chdr).__cmsg_data, &(fd), sizeof(fd));\
77 #define CMSGHDR_TO_FD(chdr, fd) \
78 memcpy(&(fd), (chdr).__cmsg_data, sizeof(fd))
80 static pthread_t socket_thread;
81 static int mp_socket_fd;
84 /* get socket path (/var/run if root, $HOME otherwise) */
86 get_socket_path(char *buffer, int bufsz)
88 const char *dir = "/var/run";
89 const char *home_dir = getenv("HOME");
91 if (getuid() != 0 && home_dir != NULL)
94 /* use current prefix as file path */
95 snprintf(buffer, bufsz, SOCKET_PATH_FMT, dir,
96 internal_config.hugefile_prefix);
102 * data flow for socket comm protocol:
103 * 1. client sends SOCKET_REQ_CONTAINER or SOCKET_REQ_GROUP
104 * 1a. in case of SOCKET_REQ_GROUP, client also then sends group number
105 * 2. server receives message
106 * 2a. in case of invalid group, SOCKET_ERR is sent back to client
107 * 2b. in case of unbound group, SOCKET_NO_FD is sent back to client
108 * 2c. in case of valid group, SOCKET_OK is sent and immediately followed by fd
110 * in case of any error, socket is closed.
113 /* send a request, return -1 on error */
115 vfio_mp_sync_send_request(int socket, int req)
122 memset(&hdr, 0, sizeof(hdr));
128 iov.iov_base = (char *) &buf;
129 iov.iov_len = sizeof(buf);
131 ret = sendmsg(socket, &hdr, 0);
137 /* receive a request and return it */
139 vfio_mp_sync_receive_request(int socket)
146 memset(&hdr, 0, sizeof(hdr));
152 iov.iov_base = (char *) &buf;
153 iov.iov_len = sizeof(buf);
155 ret = recvmsg(socket, &hdr, 0);
164 /* send OK in message, fd in control message */
166 vfio_mp_sync_send_fd(int socket, int fd)
170 struct cmsghdr *chdr;
171 char chdr_buf[CMSGLEN];
175 chdr = (struct cmsghdr *) chdr_buf;
176 memset(chdr, 0, sizeof(chdr_buf));
177 memset(&hdr, 0, sizeof(hdr));
181 iov.iov_base = (char *) &buf;
182 iov.iov_len = sizeof(buf);
183 hdr.msg_control = chdr;
184 hdr.msg_controllen = CMSGLEN;
187 FD_TO_CMSGHDR(fd, *chdr);
189 ret = sendmsg(socket, &hdr, 0);
195 /* receive OK in message, fd in control message */
197 vfio_mp_sync_receive_fd(int socket)
201 struct cmsghdr *chdr;
202 char chdr_buf[CMSGLEN];
208 chdr = (struct cmsghdr *) chdr_buf;
209 memset(chdr, 0, sizeof(chdr_buf));
210 memset(&hdr, 0, sizeof(hdr));
214 iov.iov_base = (char *) &buf;
215 iov.iov_len = sizeof(buf);
216 hdr.msg_control = chdr;
217 hdr.msg_controllen = CMSGLEN;
219 ret = recvmsg(socket, &hdr, 0);
225 if (req != SOCKET_OK)
228 CMSGHDR_TO_FD(*chdr, fd);
233 /* connect socket_fd in secondary process to the primary process's socket */
235 vfio_mp_sync_connect_to_primary(void)
237 struct sockaddr_un addr;
238 socklen_t sockaddr_len;
241 /* set up a socket */
242 socket_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
244 RTE_LOG(ERR, EAL, "Failed to create socket!\n");
248 get_socket_path(addr.sun_path, sizeof(addr.sun_path));
249 addr.sun_family = AF_UNIX;
251 sockaddr_len = sizeof(struct sockaddr_un);
253 if (connect(socket_fd, (struct sockaddr *) &addr, sockaddr_len) == 0)
256 /* if connect failed */
264 * socket listening thread for primary process
266 static __attribute__((noreturn)) void *
267 pci_vfio_mp_sync_thread(void __rte_unused * arg)
269 int ret, fd, vfio_group_no;
271 /* wait for requests on the socket */
274 struct sockaddr_un addr;
275 socklen_t sockaddr_len = sizeof(addr);
277 /* this is a blocking call */
278 conn_sock = accept(mp_socket_fd, (struct sockaddr *) &addr,
281 /* just restart on error */
285 /* set socket to linger after close */
289 setsockopt(conn_sock, SOL_SOCKET, SO_LINGER, &l, sizeof(l));
291 ret = vfio_mp_sync_receive_request(conn_sock);
294 case SOCKET_REQ_CONTAINER:
295 fd = pci_vfio_get_container_fd();
297 vfio_mp_sync_send_request(conn_sock, SOCKET_ERR);
299 vfio_mp_sync_send_fd(conn_sock, fd);
301 case SOCKET_REQ_GROUP:
302 /* wait for group number */
303 vfio_group_no = vfio_mp_sync_receive_request(conn_sock);
304 if (vfio_group_no < 0) {
309 fd = pci_vfio_get_group_fd(vfio_group_no);
312 vfio_mp_sync_send_request(conn_sock, SOCKET_ERR);
313 /* if VFIO group exists but isn't bound to VFIO driver */
315 vfio_mp_sync_send_request(conn_sock, SOCKET_NO_FD);
316 /* if group exists and is bound to VFIO driver */
318 vfio_mp_sync_send_request(conn_sock, SOCKET_OK);
319 vfio_mp_sync_send_fd(conn_sock, fd);
323 vfio_mp_sync_send_request(conn_sock, SOCKET_ERR);
331 vfio_mp_sync_socket_setup(void)
334 struct sockaddr_un addr;
335 socklen_t sockaddr_len;
337 /* set up a socket */
338 socket_fd = socket(AF_UNIX, SOCK_SEQPACKET, 0);
340 RTE_LOG(ERR, EAL, "Failed to create socket!\n");
344 get_socket_path(addr.sun_path, sizeof(addr.sun_path));
345 addr.sun_family = AF_UNIX;
347 sockaddr_len = sizeof(struct sockaddr_un);
349 unlink(addr.sun_path);
351 ret = bind(socket_fd, (struct sockaddr *) &addr, sockaddr_len);
353 RTE_LOG(ERR, EAL, "Failed to bind socket: %s!\n", strerror(errno));
358 ret = listen(socket_fd, 50);
360 RTE_LOG(ERR, EAL, "Failed to listen: %s!\n", strerror(errno));
365 /* save the socket in local configuration */
366 mp_socket_fd = socket_fd;
372 * set up a local socket and tell it to listen for incoming connections
375 pci_vfio_mp_sync_setup(void)
379 if (vfio_mp_sync_socket_setup() < 0) {
380 RTE_LOG(ERR, EAL, "Failed to set up local socket!\n");
384 ret = pthread_create(&socket_thread, NULL,
385 pci_vfio_mp_sync_thread, NULL);
387 RTE_LOG(ERR, EAL, "Failed to create thread for communication with "
388 "secondary processes!\n");