xref: /openbmc/qemu/util/vhost-user-server.c (revision 6c3a9247)
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