1 /* 2 * Sharing QEMU devices via vhost-user protocol 3 * 4 * Copyright (c) Coiby Xu <coiby.xu@gmail.com>. 5 * Copyright (c) 2020 Red Hat, Inc. 6 * 7 * This work is licensed under the terms of the GNU GPL, version 2 or 8 * later. See the COPYING file in the top-level directory. 9 */ 10 #include "qemu/osdep.h" 11 #include "qemu/main-loop.h" 12 #include "qemu/vhost-user-server.h" 13 #include "block/aio-wait.h" 14 15 /* 16 * Theory of operation: 17 * 18 * VuServer is started and stopped by vhost_user_server_start() and 19 * vhost_user_server_stop() from the main loop thread. Starting the server 20 * opens a vhost-user UNIX domain socket and listens for incoming connections. 21 * Only one connection is allowed at a time. 22 * 23 * The connection is handled by the vu_client_trip() coroutine in the 24 * VuServer->ctx AioContext. The coroutine consists of a vu_dispatch() loop 25 * where libvhost-user calls vu_message_read() to receive the next vhost-user 26 * protocol messages over the UNIX domain socket. 27 * 28 * When virtqueues are set up libvhost-user calls set_watch() to monitor kick 29 * fds. These fds are also handled in the VuServer->ctx AioContext. 30 * 31 * Both vu_client_trip() and kick fd monitoring can be stopped by shutting down 32 * the socket connection. Shutting down the socket connection causes 33 * vu_message_read() to fail since no more data can be received from the socket. 34 * After vu_dispatch() fails, vu_client_trip() calls vu_deinit() to stop 35 * libvhost-user before terminating the coroutine. vu_deinit() calls 36 * remove_watch() to stop monitoring kick fds and this stops virtqueue 37 * processing. 38 * 39 * When vu_client_trip() has finished cleaning up it schedules a BH in the main 40 * loop thread to accept the next client connection. 41 * 42 * When libvhost-user detects an error it calls panic_cb() and sets the 43 * dev->broken flag. Both vu_client_trip() and kick fd processing stop when 44 * the dev->broken flag is set. 45 * 46 * It is possible to switch AioContexts using 47 * vhost_user_server_detach_aio_context() and 48 * vhost_user_server_attach_aio_context(). They stop monitoring fds in the old 49 * AioContext and resume monitoring in the new AioContext. The vu_client_trip() 50 * coroutine remains in a yielded state during the switch. This is made 51 * possible by QIOChannel's support for spurious coroutine re-entry in 52 * qio_channel_yield(). The coroutine will restart I/O when re-entered from the 53 * new AioContext. 54 */ 55 56 static void vmsg_close_fds(VhostUserMsg *vmsg) 57 { 58 int i; 59 for (i = 0; i < vmsg->fd_num; i++) { 60 close(vmsg->fds[i]); 61 } 62 } 63 64 static void vmsg_unblock_fds(VhostUserMsg *vmsg) 65 { 66 int i; 67 for (i = 0; i < vmsg->fd_num; i++) { 68 qemu_set_nonblock(vmsg->fds[i]); 69 } 70 } 71 72 static void panic_cb(VuDev *vu_dev, const char *buf) 73 { 74 error_report("vu_panic: %s", buf); 75 } 76 77 void vhost_user_server_ref(VuServer *server) 78 { 79 assert(!server->wait_idle); 80 server->refcount++; 81 } 82 83 void vhost_user_server_unref(VuServer *server) 84 { 85 server->refcount--; 86 if (server->wait_idle && !server->refcount) { 87 aio_co_wake(server->co_trip); 88 } 89 } 90 91 static bool coroutine_fn 92 vu_message_read(VuDev *vu_dev, int conn_fd, VhostUserMsg *vmsg) 93 { 94 struct iovec iov = { 95 .iov_base = (char *)vmsg, 96 .iov_len = VHOST_USER_HDR_SIZE, 97 }; 98 int rc, read_bytes = 0; 99 Error *local_err = NULL; 100 const size_t max_fds = G_N_ELEMENTS(vmsg->fds); 101 VuServer *server = container_of(vu_dev, VuServer, vu_dev); 102 QIOChannel *ioc = server->ioc; 103 104 vmsg->fd_num = 0; 105 if (!ioc) { 106 error_report_err(local_err); 107 goto fail; 108 } 109 110 assert(qemu_in_coroutine()); 111 do { 112 size_t nfds = 0; 113 int *fds = NULL; 114 115 /* 116 * qio_channel_readv_full may have short reads, keeping calling it 117 * until getting VHOST_USER_HDR_SIZE or 0 bytes in total 118 */ 119 rc = qio_channel_readv_full(ioc, &iov, 1, &fds, &nfds, &local_err); 120 if (rc < 0) { 121 if (rc == QIO_CHANNEL_ERR_BLOCK) { 122 assert(local_err == NULL); 123 qio_channel_yield(ioc, G_IO_IN); 124 continue; 125 } else { 126 error_report_err(local_err); 127 goto fail; 128 } 129 } 130 131 if (nfds > 0) { 132 if (vmsg->fd_num + nfds > max_fds) { 133 error_report("A maximum of %zu fds are allowed, " 134 "however got %zu fds now", 135 max_fds, vmsg->fd_num + nfds); 136 g_free(fds); 137 goto fail; 138 } 139 memcpy(vmsg->fds + vmsg->fd_num, fds, nfds * sizeof(vmsg->fds[0])); 140 vmsg->fd_num += nfds; 141 g_free(fds); 142 } 143 144 if (rc == 0) { /* socket closed */ 145 goto fail; 146 } 147 148 iov.iov_base += rc; 149 iov.iov_len -= rc; 150 read_bytes += rc; 151 } while (read_bytes != VHOST_USER_HDR_SIZE); 152 153 /* qio_channel_readv_full will make socket fds blocking, unblock them */ 154 vmsg_unblock_fds(vmsg); 155 if (vmsg->size > sizeof(vmsg->payload)) { 156 error_report("Error: too big message request: %d, " 157 "size: vmsg->size: %u, " 158 "while sizeof(vmsg->payload) = %zu", 159 vmsg->request, vmsg->size, sizeof(vmsg->payload)); 160 goto fail; 161 } 162 163 struct iovec iov_payload = { 164 .iov_base = (char *)&vmsg->payload, 165 .iov_len = vmsg->size, 166 }; 167 if (vmsg->size) { 168 rc = qio_channel_readv_all_eof(ioc, &iov_payload, 1, &local_err); 169 if (rc != 1) { 170 if (local_err) { 171 error_report_err(local_err); 172 } 173 goto fail; 174 } 175 } 176 177 return true; 178 179 fail: 180 vmsg_close_fds(vmsg); 181 182 return false; 183 } 184 185 static coroutine_fn void vu_client_trip(void *opaque) 186 { 187 VuServer *server = opaque; 188 VuDev *vu_dev = &server->vu_dev; 189 190 while (!vu_dev->broken && vu_dispatch(vu_dev)) { 191 /* Keep running */ 192 } 193 194 if (server->refcount) { 195 /* Wait for requests to complete before we can unmap the memory */ 196 server->wait_idle = true; 197 qemu_coroutine_yield(); 198 server->wait_idle = false; 199 } 200 assert(server->refcount == 0); 201 202 vu_deinit(vu_dev); 203 204 /* vu_deinit() should have called remove_watch() */ 205 assert(QTAILQ_EMPTY(&server->vu_fd_watches)); 206 207 object_unref(OBJECT(server->sioc)); 208 server->sioc = NULL; 209 210 object_unref(OBJECT(server->ioc)); 211 server->ioc = NULL; 212 213 server->co_trip = NULL; 214 if (server->restart_listener_bh) { 215 qemu_bh_schedule(server->restart_listener_bh); 216 } 217 aio_wait_kick(); 218 } 219 220 /* 221 * a wrapper for vu_kick_cb 222 * 223 * since aio_dispatch can only pass one user data pointer to the 224 * callback function, pack VuDev and pvt into a struct. Then unpack it 225 * and pass them to vu_kick_cb 226 */ 227 static void kick_handler(void *opaque) 228 { 229 VuFdWatch *vu_fd_watch = opaque; 230 VuDev *vu_dev = vu_fd_watch->vu_dev; 231 232 vu_fd_watch->cb(vu_dev, 0, vu_fd_watch->pvt); 233 234 /* Stop vu_client_trip() if an error occurred in vu_fd_watch->cb() */ 235 if (vu_dev->broken) { 236 VuServer *server = container_of(vu_dev, VuServer, vu_dev); 237 238 qio_channel_shutdown(server->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL); 239 } 240 } 241 242 static VuFdWatch *find_vu_fd_watch(VuServer *server, int fd) 243 { 244 245 VuFdWatch *vu_fd_watch, *next; 246 QTAILQ_FOREACH_SAFE(vu_fd_watch, &server->vu_fd_watches, next, next) { 247 if (vu_fd_watch->fd == fd) { 248 return vu_fd_watch; 249 } 250 } 251 return NULL; 252 } 253 254 static void 255 set_watch(VuDev *vu_dev, int fd, int vu_evt, 256 vu_watch_cb cb, void *pvt) 257 { 258 259 VuServer *server = container_of(vu_dev, VuServer, vu_dev); 260 g_assert(vu_dev); 261 g_assert(fd >= 0); 262 g_assert(cb); 263 264 VuFdWatch *vu_fd_watch = find_vu_fd_watch(server, fd); 265 266 if (!vu_fd_watch) { 267 VuFdWatch *vu_fd_watch = g_new0(VuFdWatch, 1); 268 269 QTAILQ_INSERT_TAIL(&server->vu_fd_watches, vu_fd_watch, next); 270 271 vu_fd_watch->fd = fd; 272 vu_fd_watch->cb = cb; 273 qemu_set_nonblock(fd); 274 aio_set_fd_handler(server->ioc->ctx, fd, true, kick_handler, 275 NULL, NULL, NULL, vu_fd_watch); 276 vu_fd_watch->vu_dev = vu_dev; 277 vu_fd_watch->pvt = pvt; 278 } 279 } 280 281 282 static void remove_watch(VuDev *vu_dev, int fd) 283 { 284 VuServer *server; 285 g_assert(vu_dev); 286 g_assert(fd >= 0); 287 288 server = container_of(vu_dev, VuServer, vu_dev); 289 290 VuFdWatch *vu_fd_watch = find_vu_fd_watch(server, fd); 291 292 if (!vu_fd_watch) { 293 return; 294 } 295 aio_set_fd_handler(server->ioc->ctx, fd, true, 296 NULL, NULL, NULL, NULL, NULL); 297 298 QTAILQ_REMOVE(&server->vu_fd_watches, vu_fd_watch, next); 299 g_free(vu_fd_watch); 300 } 301 302 303 static void vu_accept(QIONetListener *listener, QIOChannelSocket *sioc, 304 gpointer opaque) 305 { 306 VuServer *server = opaque; 307 308 if (server->sioc) { 309 warn_report("Only one vhost-user client is allowed to " 310 "connect the server one time"); 311 return; 312 } 313 314 if (!vu_init(&server->vu_dev, server->max_queues, sioc->fd, panic_cb, 315 vu_message_read, set_watch, remove_watch, server->vu_iface)) { 316 error_report("Failed to initialize libvhost-user"); 317 return; 318 } 319 320 /* 321 * Unset the callback function for network listener to make another 322 * vhost-user client keeping waiting until this client disconnects 323 */ 324 qio_net_listener_set_client_func(server->listener, 325 NULL, 326 NULL, 327 NULL); 328 server->sioc = sioc; 329 /* 330 * Increase the object reference, so sioc will not freed by 331 * qio_net_listener_channel_func which will call object_unref(OBJECT(sioc)) 332 */ 333 object_ref(OBJECT(server->sioc)); 334 qio_channel_set_name(QIO_CHANNEL(sioc), "vhost-user client"); 335 server->ioc = QIO_CHANNEL(sioc); 336 object_ref(OBJECT(server->ioc)); 337 338 /* TODO vu_message_write() spins if non-blocking! */ 339 qio_channel_set_blocking(server->ioc, false, NULL); 340 341 server->co_trip = qemu_coroutine_create(vu_client_trip, server); 342 343 aio_context_acquire(server->ctx); 344 vhost_user_server_attach_aio_context(server, server->ctx); 345 aio_context_release(server->ctx); 346 } 347 348 void vhost_user_server_stop(VuServer *server) 349 { 350 aio_context_acquire(server->ctx); 351 352 qemu_bh_delete(server->restart_listener_bh); 353 server->restart_listener_bh = NULL; 354 355 if (server->sioc) { 356 VuFdWatch *vu_fd_watch; 357 358 QTAILQ_FOREACH(vu_fd_watch, &server->vu_fd_watches, next) { 359 aio_set_fd_handler(server->ctx, vu_fd_watch->fd, true, 360 NULL, NULL, NULL, NULL, vu_fd_watch); 361 } 362 363 qio_channel_shutdown(server->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL); 364 365 AIO_WAIT_WHILE(server->ctx, server->co_trip); 366 } 367 368 aio_context_release(server->ctx); 369 370 if (server->listener) { 371 qio_net_listener_disconnect(server->listener); 372 object_unref(OBJECT(server->listener)); 373 } 374 } 375 376 /* 377 * Allow the next client to connect to the server. Called from a BH in the main 378 * loop. 379 */ 380 static void restart_listener_bh(void *opaque) 381 { 382 VuServer *server = opaque; 383 384 qio_net_listener_set_client_func(server->listener, vu_accept, server, 385 NULL); 386 } 387 388 /* Called with ctx acquired */ 389 void vhost_user_server_attach_aio_context(VuServer *server, AioContext *ctx) 390 { 391 VuFdWatch *vu_fd_watch; 392 393 server->ctx = ctx; 394 395 if (!server->sioc) { 396 return; 397 } 398 399 qio_channel_attach_aio_context(server->ioc, ctx); 400 401 QTAILQ_FOREACH(vu_fd_watch, &server->vu_fd_watches, next) { 402 aio_set_fd_handler(ctx, vu_fd_watch->fd, true, kick_handler, NULL, 403 NULL, NULL, vu_fd_watch); 404 } 405 406 aio_co_schedule(ctx, server->co_trip); 407 } 408 409 /* Called with server->ctx acquired */ 410 void vhost_user_server_detach_aio_context(VuServer *server) 411 { 412 if (server->sioc) { 413 VuFdWatch *vu_fd_watch; 414 415 QTAILQ_FOREACH(vu_fd_watch, &server->vu_fd_watches, next) { 416 aio_set_fd_handler(server->ctx, vu_fd_watch->fd, true, 417 NULL, NULL, NULL, NULL, vu_fd_watch); 418 } 419 420 qio_channel_detach_aio_context(server->ioc); 421 } 422 423 server->ctx = NULL; 424 } 425 426 bool vhost_user_server_start(VuServer *server, 427 SocketAddress *socket_addr, 428 AioContext *ctx, 429 uint16_t max_queues, 430 const VuDevIface *vu_iface, 431 Error **errp) 432 { 433 QEMUBH *bh; 434 QIONetListener *listener; 435 436 if (socket_addr->type != SOCKET_ADDRESS_TYPE_UNIX && 437 socket_addr->type != SOCKET_ADDRESS_TYPE_FD) { 438 error_setg(errp, "Only socket address types 'unix' and 'fd' are supported"); 439 return false; 440 } 441 442 listener = qio_net_listener_new(); 443 if (qio_net_listener_open_sync(listener, socket_addr, 1, 444 errp) < 0) { 445 object_unref(OBJECT(listener)); 446 return false; 447 } 448 449 bh = qemu_bh_new(restart_listener_bh, server); 450 451 /* zero out unspecified fields */ 452 *server = (VuServer) { 453 .listener = listener, 454 .restart_listener_bh = bh, 455 .vu_iface = vu_iface, 456 .max_queues = max_queues, 457 .ctx = ctx, 458 }; 459 460 qio_net_listener_set_name(server->listener, "vhost-user-backend-listener"); 461 462 qio_net_listener_set_client_func(server->listener, 463 vu_accept, 464 server, 465 NULL); 466 467 QTAILQ_INIT(&server->vu_fd_watches); 468 return true; 469 } 470