master
c 529 lines 15.6 KB
Raw
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/error-report.h"
12 #include "qemu/main-loop.h"
13 #include "qemu/vhost-user-server.h"
14 #include "qemu/aio-wait.h"
15
16 /*
17 * Theory of operation:
18 *
19 * VuServer is started and stopped by vhost_user_server_start() and
20 * vhost_user_server_stop() from the main loop thread. Starting the server
21 * opens a vhost-user UNIX domain socket and listens for incoming connections.
22 * Only one connection is allowed at a time.
23 *
24 * The connection is handled by the vu_client_trip() coroutine in the
25 * VuServer->ctx AioContext. The coroutine consists of a vu_dispatch() loop
26 * where libvhost-user calls vu_message_read() to receive the next vhost-user
27 * protocol messages over the UNIX domain socket.
28 *
29 * When virtqueues are set up libvhost-user calls set_watch() to monitor kick
30 * fds. These fds are also handled in the VuServer->ctx AioContext.
31 *
32 * Both vu_client_trip() and kick fd monitoring can be stopped by shutting down
33 * the socket connection. Shutting down the socket connection causes
34 * vu_message_read() to fail since no more data can be received from the socket.
35 * After vu_dispatch() fails, vu_client_trip() calls vu_deinit() to stop
36 * libvhost-user before terminating the coroutine. vu_deinit() calls
37 * remove_watch() to stop monitoring kick fds and this stops virtqueue
38 * processing.
39 *
40 * When vu_client_trip() has finished cleaning up it schedules a BH in the main
41 * loop thread to accept the next client connection.
42 *
43 * When libvhost-user detects an error it calls panic_cb() and sets the
44 * dev->broken flag. Both vu_client_trip() and kick fd processing stop when
45 * the dev->broken flag is set.
46 *
47 * It is possible to switch AioContexts using
48 * vhost_user_server_detach_aio_context() and
49 * vhost_user_server_attach_aio_context(). They stop monitoring fds in the old
50 * AioContext and resume monitoring in the new AioContext. The vu_client_trip()
51 * coroutine remains in a yielded state during the switch. This is made
52 * possible by QIOChannel's support for spurious coroutine re-entry in
53 * qio_channel_yield(). The coroutine will restart I/O when re-entered from the
54 * new AioContext.
55 */
56
57 static void vmsg_close_fds(VhostUserMsg *vmsg)
58 {
59 int i;
60 for (i = 0; i < vmsg->fd_num; i++) {
61 close(vmsg->fds[i]);
62 }
63 }
64
65 static bool vmsg_unblock_fds(VhostUserMsg *vmsg, Error **errp)
66 {
67 int i;
68
69 /*
70 * These messages carry fd used to map memory, not to send/receive messages,
71 * so this operation is useless. In addition, in some systems this
72 * operation can fail (e.g. in macOS setting an fd returned by shm_open()
73 * non-blocking fails with errno = ENOTTY)
74 */
75 if (vmsg->request == VHOST_USER_ADD_MEM_REG ||
76 vmsg->request == VHOST_USER_SET_MEM_TABLE) {
77 return true;
78 }
79
80 for (i = 0; i < vmsg->fd_num; i++) {
81 if (!qemu_set_blocking(vmsg->fds[i], false, errp)) {
82 return false;
83 }
84 }
85
86 return true;
87 }
88
89 static void panic_cb(VuDev *vu_dev, const char *buf)
90 {
91 error_report("vu_panic: %s", buf);
92 }
93
94 void vhost_user_server_inc_in_flight(VuServer *server)
95 {
96 assert(!server->wait_idle);
97 qatomic_inc(&server->in_flight);
98 }
99
100 void vhost_user_server_dec_in_flight(VuServer *server)
101 {
102 if (qatomic_fetch_dec(&server->in_flight) == 1) {
103 if (server->wait_idle) {
104 aio_co_wake(server->co_trip);
105 }
106 }
107 }
108
109 bool vhost_user_server_has_in_flight(VuServer *server)
110 {
111 return qatomic_load_acquire(&server->in_flight) > 0;
112 }
113
114 static bool coroutine_fn
115 vu_message_read(VuDev *vu_dev, int conn_fd, VhostUserMsg *vmsg)
116 {
117 struct iovec iov = {
118 .iov_base = (char *)vmsg,
119 .iov_len = VHOST_USER_HDR_SIZE,
120 };
121 int rc, read_bytes = 0;
122 Error *local_err = NULL;
123 const size_t max_fds = G_N_ELEMENTS(vmsg->fds);
124 VuServer *server = container_of(vu_dev, VuServer, vu_dev);
125 QIOChannel *ioc = server->ioc;
126
127 vmsg->fd_num = 0;
128 if (!ioc) {
129 goto fail;
130 }
131
132 assert(qemu_in_coroutine());
133 do {
134 size_t nfds = 0;
135 int *fds = NULL;
136
137 /*
138 * qio_channel_readv_full may have short reads, keeping calling it
139 * until getting VHOST_USER_HDR_SIZE or 0 bytes in total
140 */
141 rc = qio_channel_readv_full(ioc, &iov, 1, &fds, &nfds, 0, &local_err);
142 if (rc < 0) {
143 if (rc == QIO_CHANNEL_ERR_BLOCK) {
144 assert(local_err == NULL);
145 if (server->ctx) {
146 server->in_qio_channel_yield = true;
147 qio_channel_yield(ioc, G_IO_IN);
148 server->in_qio_channel_yield = false;
149 } else {
150 return false;
151 }
152 continue;
153 } else {
154 error_report_err(local_err);
155 goto fail;
156 }
157 }
158
159 if (nfds > 0) {
160 if (vmsg->fd_num + nfds > max_fds) {
161 error_report("A maximum of %zu fds are allowed, "
162 "however got %zu fds now",
163 max_fds, vmsg->fd_num + nfds);
164 g_free(fds);
165 goto fail;
166 }
167 memcpy(vmsg->fds + vmsg->fd_num, fds, nfds * sizeof(vmsg->fds[0]));
168 vmsg->fd_num += nfds;
169 g_free(fds);
170 }
171
172 if (rc == 0) { /* socket closed */
173 goto fail;
174 }
175
176 iov.iov_base += rc;
177 iov.iov_len -= rc;
178 read_bytes += rc;
179 } while (read_bytes != VHOST_USER_HDR_SIZE);
180
181 /* qio_channel_readv_full will make socket fds blocking, unblock them */
182 if (!vmsg_unblock_fds(vmsg, &local_err)) {
183 error_report_err(local_err);
184 goto fail;
185 }
186 if (vmsg->size > sizeof(vmsg->payload)) {
187 error_report("Error: too big message request: %d, "
188 "size: vmsg->size: %u, "
189 "while sizeof(vmsg->payload) = %zu",
190 vmsg->request, vmsg->size, sizeof(vmsg->payload));
191 goto fail;
192 }
193
194 struct iovec iov_payload = {
195 .iov_base = (char *)&vmsg->payload,
196 .iov_len = vmsg->size,
197 };
198 if (vmsg->size) {
199 rc = qio_channel_readv_all_eof(ioc, &iov_payload, 1, &local_err);
200 if (rc != 1) {
201 if (local_err) {
202 error_report_err(local_err);
203 }
204 goto fail;
205 }
206 }
207
208 return true;
209
210 fail:
211 vmsg_close_fds(vmsg);
212
213 return false;
214 }
215
216 static coroutine_fn void vu_client_trip(void *opaque)
217 {
218 VuServer *server = opaque;
219 VuDev *vu_dev = &server->vu_dev;
220
221 while (!vu_dev->broken) {
222 if (server->quiescing) {
223 server->co_trip = NULL;
224 aio_wait_kick();
225 return;
226 }
227 /* vu_dispatch() returns false if server->ctx went away */
228 if (!vu_dispatch(vu_dev) && server->ctx) {
229 break;
230 }
231 }
232
233 if (vhost_user_server_has_in_flight(server)) {
234 /* Wait for requests to complete before we can unmap the memory */
235 server->wait_idle = true;
236 qemu_coroutine_yield();
237 server->wait_idle = false;
238 }
239 assert(!vhost_user_server_has_in_flight(server));
240
241 vu_deinit(vu_dev);
242
243 /* vu_deinit() should have called remove_watch() */
244 assert(QTAILQ_EMPTY(&server->vu_fd_watches));
245
246 object_unref(OBJECT(server->sioc));
247 server->sioc = NULL;
248
249 object_unref(OBJECT(server->ioc));
250 server->ioc = NULL;
251
252 server->co_trip = NULL;
253 if (server->restart_listener_bh) {
254 qemu_bh_schedule(server->restart_listener_bh);
255 }
256 aio_wait_kick();
257 }
258
259 /*
260 * a wrapper for vu_kick_cb
261 *
262 * since aio_dispatch can only pass one user data pointer to the
263 * callback function, pack VuDev and pvt into a struct. Then unpack it
264 * and pass them to vu_kick_cb
265 */
266 static void kick_handler(void *opaque)
267 {
268 VuFdWatch *vu_fd_watch = opaque;
269 VuDev *vu_dev = vu_fd_watch->vu_dev;
270
271 vu_fd_watch->cb(vu_dev, 0, vu_fd_watch->pvt);
272
273 /* Stop vu_client_trip() if an error occurred in vu_fd_watch->cb() */
274 if (vu_dev->broken) {
275 VuServer *server = container_of(vu_dev, VuServer, vu_dev);
276
277 qio_channel_shutdown(server->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL);
278 }
279 }
280
281 static VuFdWatch *find_vu_fd_watch(VuServer *server, int fd)
282 {
283
284 VuFdWatch *vu_fd_watch, *next;
285 QTAILQ_FOREACH_SAFE(vu_fd_watch, &server->vu_fd_watches, next, next) {
286 if (vu_fd_watch->fd == fd) {
287 return vu_fd_watch;
288 }
289 }
290 return NULL;
291 }
292
293 static void
294 set_watch(VuDev *vu_dev, int fd, int vu_evt,
295 vu_watch_cb cb, void *pvt)
296 {
297
298 VuServer *server = container_of(vu_dev, VuServer, vu_dev);
299 g_assert(vu_dev);
300 g_assert(fd >= 0);
301 g_assert(cb);
302
303 VuFdWatch *vu_fd_watch = find_vu_fd_watch(server, fd);
304
305 if (!vu_fd_watch) {
306 vu_fd_watch = g_new0(VuFdWatch, 1);
307
308 QTAILQ_INSERT_TAIL(&server->vu_fd_watches, vu_fd_watch, next);
309
310 vu_fd_watch->fd = fd;
311 vu_fd_watch->cb = cb;
312 /* TODO: handle error more gracefully than aborting */
313 qemu_set_blocking(fd, false, &error_abort);
314 aio_set_fd_handler(server->ctx, fd, kick_handler,
315 NULL, NULL, NULL, vu_fd_watch);
316 vu_fd_watch->vu_dev = vu_dev;
317 vu_fd_watch->pvt = pvt;
318 }
319 }
320
321
322 static void remove_watch(VuDev *vu_dev, int fd)
323 {
324 VuServer *server;
325 g_assert(vu_dev);
326 g_assert(fd >= 0);
327
328 server = container_of(vu_dev, VuServer, vu_dev);
329
330 VuFdWatch *vu_fd_watch = find_vu_fd_watch(server, fd);
331
332 if (!vu_fd_watch) {
333 return;
334 }
335 aio_set_fd_handler(server->ctx, fd, NULL, NULL, NULL, NULL, NULL);
336
337 QTAILQ_REMOVE(&server->vu_fd_watches, vu_fd_watch, next);
338 g_free(vu_fd_watch);
339 }
340
341
342 static void vu_accept(QIONetListener *listener, QIOChannelSocket *sioc,
343 gpointer opaque)
344 {
345 VuServer *server = opaque;
346 Error *local_err = NULL;
347
348 if (server->sioc) {
349 warn_report("Only one vhost-user client is allowed to "
350 "connect the server one time");
351 return;
352 }
353
354 if (!vu_init(&server->vu_dev, server->max_queues, sioc->fd, panic_cb,
355 vu_message_read, set_watch, remove_watch, server->vu_iface)) {
356 error_report("Failed to initialize libvhost-user");
357 return;
358 }
359
360 /*
361 * Unset the callback function for network listener to make another
362 * vhost-user client keeping waiting until this client disconnects
363 */
364 qio_net_listener_set_client_func(server->listener,
365 NULL,
366 NULL,
367 NULL);
368 server->sioc = sioc;
369 /*
370 * Increase the object reference, so sioc will not freed by
371 * qio_net_listener_channel_func which will call object_unref(OBJECT(sioc))
372 */
373 object_ref(OBJECT(server->sioc));
374 qio_channel_set_name(QIO_CHANNEL(sioc), "vhost-user client");
375 server->ioc = QIO_CHANNEL(sioc);
376 object_ref(OBJECT(server->ioc));
377
378 /* TODO vu_message_write() spins if non-blocking! */
379 if (!qio_channel_set_blocking(server->ioc, false, &local_err)) {
380 error_report_err(local_err);
381 vu_deinit(&server->vu_dev);
382 return;
383 }
384
385 qio_channel_set_follow_coroutine_ctx(server->ioc, true);
386
387 vhost_user_server_attach_aio_context(server, server->ctx);
388 }
389
390 /* server->ctx acquired by caller */
391 void vhost_user_server_stop(VuServer *server)
392 {
393 qemu_bh_delete(server->restart_listener_bh);
394 server->restart_listener_bh = NULL;
395
396 if (server->sioc) {
397 VuFdWatch *vu_fd_watch;
398
399 QTAILQ_FOREACH(vu_fd_watch, &server->vu_fd_watches, next) {
400 aio_set_fd_handler(server->ctx, vu_fd_watch->fd,
401 NULL, NULL, NULL, NULL, vu_fd_watch);
402 }
403
404 qio_channel_shutdown(server->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL);
405
406 AIO_WAIT_WHILE(server->ctx, server->co_trip);
407 }
408
409 if (server->listener) {
410 qio_net_listener_disconnect(server->listener);
411 object_unref(OBJECT(server->listener));
412 }
413 }
414
415 /*
416 * Allow the next client to connect to the server. Called from a BH in the main
417 * loop.
418 */
419 static void restart_listener_bh(void *opaque)
420 {
421 VuServer *server = opaque;
422
423 qio_net_listener_set_client_func(server->listener, vu_accept, server,
424 NULL);
425 }
426
427 /* Called with ctx acquired */
428 void vhost_user_server_attach_aio_context(VuServer *server, AioContext *ctx)
429 {
430 VuFdWatch *vu_fd_watch;
431
432 server->ctx = ctx;
433
434 if (!server->sioc) {
435 return;
436 }
437
438 QTAILQ_FOREACH(vu_fd_watch, &server->vu_fd_watches, next) {
439 aio_set_fd_handler(ctx, vu_fd_watch->fd, kick_handler, NULL,
440 NULL, NULL, vu_fd_watch);
441 }
442
443 if (server->co_trip) {
444 /*
445 * The caller didn't fully shut down co_trip (this can happen on
446 * non-polling drains like in bdrv_graph_wrlock()). This is okay as long
447 * as it no longer tries to shut it down and we're guaranteed to still
448 * be in the same AioContext as before.
449 *
450 * co_ctx can still be NULL if we get multiple calls and only just
451 * scheduled a new coroutine in the else branch.
452 */
453 AioContext *co_ctx = qemu_coroutine_get_aio_context(server->co_trip);
454
455 assert(!server->quiescing);
456 assert(!co_ctx || co_ctx == ctx);
457 } else {
458 server->co_trip = qemu_coroutine_create(vu_client_trip, server);
459 assert(!server->in_qio_channel_yield);
460 aio_co_schedule(ctx, server->co_trip);
461 }
462 }
463
464 /* Called with server->ctx acquired */
465 void vhost_user_server_detach_aio_context(VuServer *server)
466 {
467 if (server->sioc) {
468 VuFdWatch *vu_fd_watch;
469
470 QTAILQ_FOREACH(vu_fd_watch, &server->vu_fd_watches, next) {
471 aio_set_fd_handler(server->ctx, vu_fd_watch->fd,
472 NULL, NULL, NULL, NULL, vu_fd_watch);
473 }
474 }
475
476 server->ctx = NULL;
477
478 if (server->ioc) {
479 if (server->in_qio_channel_yield) {
480 /* Stop receiving the next vhost-user message */
481 qio_channel_wake_read(server->ioc);
482 }
483 }
484 }
485
486 bool vhost_user_server_start(VuServer *server,
487 SocketAddress *socket_addr,
488 AioContext *ctx,
489 uint16_t max_queues,
490 const VuDevIface *vu_iface,
491 Error **errp)
492 {
493 QEMUBH *bh;
494 QIONetListener *listener;
495
496 if (socket_addr->type != SOCKET_ADDRESS_TYPE_UNIX &&
497 socket_addr->type != SOCKET_ADDRESS_TYPE_FD) {
498 error_setg(errp, "Only socket address types 'unix' and 'fd' are supported");
499 return false;
500 }
501
502 listener = qio_net_listener_new();
503 if (qio_net_listener_open_sync(listener, socket_addr, 1,
504 errp) < 0) {
505 object_unref(OBJECT(listener));
506 return false;
507 }
508
509 bh = qemu_bh_new(restart_listener_bh, server);
510
511 /* zero out unspecified fields */
512 *server = (VuServer) {
513 .listener = listener,
514 .restart_listener_bh = bh,
515 .vu_iface = vu_iface,
516 .max_queues = max_queues,
517 .ctx = ctx,
518 };
519
520 qio_net_listener_set_name(server->listener, "vhost-user-backend-listener");
521
522 qio_net_listener_set_client_func(server->listener,
523 vu_accept,
524 server,
525 NULL);
526
527 QTAILQ_INIT(&server->vu_fd_watches);
528 return true;
529 }