]> git.proxmox.com Git - qemu.git/blame - main-loop.c
vma: implement experimental nbd server
[qemu.git] / main-loop.c
CommitLineData
d3b12f5d
PB
1/*
2 * QEMU System Emulator
3 *
4 * Copyright (c) 2003-2008 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
d3b12f5d 24
0ec024f6 25#include "qemu-common.h"
1de7afc9 26#include "qemu/timer.h"
0ec024f6 27#include "slirp/slirp.h"
1de7afc9 28#include "qemu/main-loop.h"
737e150e 29#include "block/aio.h"
d3b12f5d
PB
30
31#ifndef _WIN32
32
1de7afc9 33#include "qemu/compatfd.h"
0ec024f6 34
d3b12f5d
PB
35/* If we have signalfd, we mask out the signals we want to handle and then
36 * use signalfd to listen for them. We rely on whatever the current signal
37 * handler is to dispatch the signals when we receive them.
38 */
39static void sigfd_handler(void *opaque)
40{
41 int fd = (intptr_t)opaque;
42 struct qemu_signalfd_siginfo info;
43 struct sigaction action;
44 ssize_t len;
45
46 while (1) {
47 do {
48 len = read(fd, &info, sizeof(info));
49 } while (len == -1 && errno == EINTR);
50
51 if (len == -1 && errno == EAGAIN) {
52 break;
53 }
54
55 if (len != sizeof(info)) {
56 printf("read from sigfd returned %zd: %m\n", len);
57 return;
58 }
59
60 sigaction(info.ssi_signo, NULL, &action);
61 if ((action.sa_flags & SA_SIGINFO) && action.sa_sigaction) {
62 action.sa_sigaction(info.ssi_signo,
63 (siginfo_t *)&info, NULL);
64 } else if (action.sa_handler) {
65 action.sa_handler(info.ssi_signo);
66 }
67 }
68}
69
70static int qemu_signal_init(void)
71{
72 int sigfd;
73 sigset_t set;
74
75 /*
76 * SIG_IPI must be blocked in the main thread and must not be caught
77 * by sigwait() in the signal thread. Otherwise, the cpu thread will
78 * not catch it reliably.
79 */
80 sigemptyset(&set);
81 sigaddset(&set, SIG_IPI);
d3b12f5d
PB
82 sigaddset(&set, SIGIO);
83 sigaddset(&set, SIGALRM);
84 sigaddset(&set, SIGBUS);
85 pthread_sigmask(SIG_BLOCK, &set, NULL);
86
4aa7534d 87 sigdelset(&set, SIG_IPI);
d3b12f5d
PB
88 sigfd = qemu_signalfd(&set);
89 if (sigfd == -1) {
90 fprintf(stderr, "failed to create signalfd\n");
91 return -errno;
92 }
93
94 fcntl_setfl(sigfd, O_NONBLOCK);
95
96 qemu_set_fd_handler2(sigfd, NULL, sigfd_handler, NULL,
97 (void *)(intptr_t)sigfd);
98
99 return 0;
100}
101
102#else /* _WIN32 */
103
4c8d0d27 104static int qemu_signal_init(void)
d3b12f5d 105{
d3b12f5d
PB
106 return 0;
107}
4c8d0d27
PB
108#endif
109
110static AioContext *qemu_aio_context;
d3b12f5d
PB
111
112void qemu_notify_event(void)
113{
4c8d0d27 114 if (!qemu_aio_context) {
ee77dfb2
MR
115 return;
116 }
4c8d0d27 117 aio_notify(qemu_aio_context);
d3b12f5d
PB
118}
119
172061a0 120int qemu_init_main_loop(void)
d3b12f5d
PB
121{
122 int ret;
82cbbdc6 123 GSource *src;
d3b12f5d 124
172061a0 125 init_clocks();
f9ab4654
PB
126 if (init_timer_alarm() < 0) {
127 fprintf(stderr, "could not initialize alarm timer\n");
128 exit(1);
129 }
172061a0 130
d3b12f5d
PB
131 ret = qemu_signal_init();
132 if (ret) {
133 return ret;
134 }
135
f627aab1 136 qemu_aio_context = aio_context_new();
82cbbdc6
PB
137 src = aio_get_g_source(qemu_aio_context);
138 g_source_attach(src, NULL);
139 g_source_unref(src);
d3b12f5d
PB
140 return 0;
141}
142
15455536
PB
143static fd_set rfds, wfds, xfds;
144static int nfds;
06ac7d49 145static GPollFD poll_fds[1024 * 2]; /* this is probably overkill */
d3b12f5d
PB
146static int n_poll_fds;
147static int max_priority;
148
ea26ce76 149#ifndef _WIN32
d3b12f5d 150static void glib_select_fill(int *max_fd, fd_set *rfds, fd_set *wfds,
7c7db755 151 fd_set *xfds, uint32_t *cur_timeout)
d3b12f5d
PB
152{
153 GMainContext *context = g_main_context_default();
154 int i;
4dae83ae 155 int timeout = 0;
d3b12f5d
PB
156
157 g_main_context_prepare(context, &max_priority);
158
159 n_poll_fds = g_main_context_query(context, max_priority, &timeout,
160 poll_fds, ARRAY_SIZE(poll_fds));
161 g_assert(n_poll_fds <= ARRAY_SIZE(poll_fds));
162
163 for (i = 0; i < n_poll_fds; i++) {
164 GPollFD *p = &poll_fds[i];
165
166 if ((p->events & G_IO_IN)) {
167 FD_SET(p->fd, rfds);
168 *max_fd = MAX(*max_fd, p->fd);
169 }
170 if ((p->events & G_IO_OUT)) {
171 FD_SET(p->fd, wfds);
172 *max_fd = MAX(*max_fd, p->fd);
173 }
174 if ((p->events & G_IO_ERR)) {
175 FD_SET(p->fd, xfds);
176 *max_fd = MAX(*max_fd, p->fd);
177 }
178 }
179
4dae83ae
PB
180 if (timeout >= 0 && timeout < *cur_timeout) {
181 *cur_timeout = timeout;
d3b12f5d
PB
182 }
183}
184
185static void glib_select_poll(fd_set *rfds, fd_set *wfds, fd_set *xfds,
186 bool err)
187{
188 GMainContext *context = g_main_context_default();
189
190 if (!err) {
191 int i;
192
193 for (i = 0; i < n_poll_fds; i++) {
194 GPollFD *p = &poll_fds[i];
195
196 if ((p->events & G_IO_IN) && FD_ISSET(p->fd, rfds)) {
197 p->revents |= G_IO_IN;
198 }
199 if ((p->events & G_IO_OUT) && FD_ISSET(p->fd, wfds)) {
200 p->revents |= G_IO_OUT;
201 }
202 if ((p->events & G_IO_ERR) && FD_ISSET(p->fd, xfds)) {
203 p->revents |= G_IO_ERR;
204 }
205 }
206 }
207
208 if (g_main_context_check(context, max_priority, poll_fds, n_poll_fds)) {
209 g_main_context_dispatch(context);
210 }
211}
212
7c7db755 213static int os_host_main_loop_wait(uint32_t timeout)
15455536 214{
7c7db755 215 struct timeval tv, *tvarg = NULL;
15455536
PB
216 int ret;
217
218 glib_select_fill(&nfds, &rfds, &wfds, &xfds, &timeout);
219
7c7db755
SS
220 if (timeout < UINT32_MAX) {
221 tvarg = &tv;
222 tv.tv_sec = timeout / 1000;
223 tv.tv_usec = (timeout % 1000) * 1000;
224 }
225
15455536
PB
226 if (timeout > 0) {
227 qemu_mutex_unlock_iothread();
228 }
229
7c7db755 230 ret = select(nfds + 1, &rfds, &wfds, &xfds, tvarg);
15455536
PB
231
232 if (timeout > 0) {
233 qemu_mutex_lock_iothread();
234 }
235
236 glib_select_poll(&rfds, &wfds, &xfds, (ret < 0));
237 return ret;
238}
239#else
d3b12f5d
PB
240/***********************************************************/
241/* Polling handling */
242
243typedef struct PollingEntry {
244 PollingFunc *func;
245 void *opaque;
246 struct PollingEntry *next;
247} PollingEntry;
248
249static PollingEntry *first_polling_entry;
250
251int qemu_add_polling_cb(PollingFunc *func, void *opaque)
252{
253 PollingEntry **ppe, *pe;
254 pe = g_malloc0(sizeof(PollingEntry));
255 pe->func = func;
256 pe->opaque = opaque;
257 for(ppe = &first_polling_entry; *ppe != NULL; ppe = &(*ppe)->next);
258 *ppe = pe;
259 return 0;
260}
261
262void qemu_del_polling_cb(PollingFunc *func, void *opaque)
263{
264 PollingEntry **ppe, *pe;
265 for(ppe = &first_polling_entry; *ppe != NULL; ppe = &(*ppe)->next) {
266 pe = *ppe;
267 if (pe->func == func && pe->opaque == opaque) {
268 *ppe = pe->next;
269 g_free(pe);
270 break;
271 }
272 }
273}
274
275/***********************************************************/
276/* Wait objects support */
277typedef struct WaitObjects {
278 int num;
06ac7d49 279 int revents[MAXIMUM_WAIT_OBJECTS + 1];
d3b12f5d
PB
280 HANDLE events[MAXIMUM_WAIT_OBJECTS + 1];
281 WaitObjectFunc *func[MAXIMUM_WAIT_OBJECTS + 1];
282 void *opaque[MAXIMUM_WAIT_OBJECTS + 1];
283} WaitObjects;
284
285static WaitObjects wait_objects = {0};
286
287int qemu_add_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
288{
289 WaitObjects *w = &wait_objects;
290 if (w->num >= MAXIMUM_WAIT_OBJECTS) {
291 return -1;
292 }
293 w->events[w->num] = handle;
294 w->func[w->num] = func;
295 w->opaque[w->num] = opaque;
06ac7d49 296 w->revents[w->num] = 0;
d3b12f5d
PB
297 w->num++;
298 return 0;
299}
300
301void qemu_del_wait_object(HANDLE handle, WaitObjectFunc *func, void *opaque)
302{
303 int i, found;
304 WaitObjects *w = &wait_objects;
305
306 found = 0;
307 for (i = 0; i < w->num; i++) {
308 if (w->events[i] == handle) {
309 found = 1;
310 }
311 if (found) {
312 w->events[i] = w->events[i + 1];
313 w->func[i] = w->func[i + 1];
314 w->opaque[i] = w->opaque[i + 1];
06ac7d49 315 w->revents[i] = w->revents[i + 1];
d3b12f5d
PB
316 }
317 }
318 if (found) {
319 w->num--;
320 }
321}
322
d3385eb4
PB
323void qemu_fd_register(int fd)
324{
4c8d0d27
PB
325 WSAEventSelect(fd, event_notifier_get_handle(&qemu_aio_context->notifier),
326 FD_READ | FD_ACCEPT | FD_CLOSE |
d3385eb4
PB
327 FD_CONNECT | FD_WRITE | FD_OOB);
328}
329
7c7db755 330static int os_host_main_loop_wait(uint32_t timeout)
d3b12f5d 331{
ea26ce76 332 GMainContext *context = g_main_context_default();
5e3bc735 333 int select_ret, g_poll_ret, ret, i;
d3b12f5d 334 PollingEntry *pe;
d3385eb4 335 WaitObjects *w = &wait_objects;
42fe1c24 336 gint poll_timeout;
15455536 337 static struct timeval tv0;
d3b12f5d
PB
338
339 /* XXX: need to suppress polling by better using win32 events */
340 ret = 0;
341 for (pe = first_polling_entry; pe != NULL; pe = pe->next) {
342 ret |= pe->func(pe->opaque);
343 }
d3385eb4
PB
344 if (ret != 0) {
345 return ret;
346 }
d3b12f5d 347
ea26ce76 348 g_main_context_prepare(context, &max_priority);
42fe1c24 349 n_poll_fds = g_main_context_query(context, max_priority, &poll_timeout,
ea26ce76
PB
350 poll_fds, ARRAY_SIZE(poll_fds));
351 g_assert(n_poll_fds <= ARRAY_SIZE(poll_fds));
352
06ac7d49 353 for (i = 0; i < w->num; i++) {
58b9630d 354 poll_fds[n_poll_fds + i].fd = (DWORD_PTR)w->events[i];
ea26ce76 355 poll_fds[n_poll_fds + i].events = G_IO_IN;
06ac7d49
PB
356 }
357
3239ad04
SW
358 if (poll_timeout < 0 || timeout < poll_timeout) {
359 poll_timeout = timeout;
360 }
361
d3385eb4 362 qemu_mutex_unlock_iothread();
5e3bc735 363 g_poll_ret = g_poll(poll_fds, n_poll_fds + w->num, poll_timeout);
d3385eb4 364 qemu_mutex_lock_iothread();
5e3bc735 365 if (g_poll_ret > 0) {
06ac7d49 366 for (i = 0; i < w->num; i++) {
ea26ce76 367 w->revents[i] = poll_fds[n_poll_fds + i].revents;
d3385eb4 368 }
06ac7d49
PB
369 for (i = 0; i < w->num; i++) {
370 if (w->revents[i] && w->func[i]) {
371 w->func[i](w->opaque[i]);
d3b12f5d 372 }
d3b12f5d
PB
373 }
374 }
375
ea26ce76
PB
376 if (g_main_context_check(context, max_priority, poll_fds, n_poll_fds)) {
377 g_main_context_dispatch(context);
378 }
379
5e3bc735
FC
380 /* Call select after g_poll to avoid a useless iteration and therefore
381 * improve socket latency.
d3385eb4
PB
382 */
383
5e3bc735
FC
384 if (nfds >= 0) {
385 select_ret = select(nfds + 1, &rfds, &wfds, &xfds, &tv0);
386 if (select_ret != 0) {
387 timeout = 0;
388 }
389 }
390
391 return select_ret || g_poll_ret;
d3b12f5d
PB
392}
393#endif
394
395int main_loop_wait(int nonblocking)
396{
7c7db755
SS
397 int ret;
398 uint32_t timeout = UINT32_MAX;
d3b12f5d
PB
399
400 if (nonblocking) {
401 timeout = 0;
d3b12f5d
PB
402 }
403
d3b12f5d
PB
404 /* poll any events */
405 /* XXX: separate device handlers from system ones */
406 nfds = -1;
407 FD_ZERO(&rfds);
408 FD_ZERO(&wfds);
409 FD_ZERO(&xfds);
410
411#ifdef CONFIG_SLIRP
7c7db755 412 slirp_update_timeout(&timeout);
d3b12f5d
PB
413 slirp_select_fill(&nfds, &rfds, &wfds, &xfds);
414#endif
415 qemu_iohandler_fill(&nfds, &rfds, &wfds, &xfds);
15455536 416 ret = os_host_main_loop_wait(timeout);
d3b12f5d
PB
417 qemu_iohandler_poll(&rfds, &wfds, &xfds, ret);
418#ifdef CONFIG_SLIRP
419 slirp_select_poll(&rfds, &wfds, &xfds, (ret < 0));
420#endif
421
422 qemu_run_all_timers();
423
d3b12f5d
PB
424 return ret;
425}
f627aab1
PB
426
427/* Functions to operate on the main QEMU AioContext. */
428
429QEMUBH *qemu_bh_new(QEMUBHFunc *cb, void *opaque)
430{
431 return aio_bh_new(qemu_aio_context, cb, opaque);
432}
433
a915f4bc
PB
434bool qemu_aio_wait(void)
435{
7c0628b2 436 return aio_poll(qemu_aio_context, true);
a915f4bc
PB
437}
438
f42b2207 439#ifdef CONFIG_POSIX
a915f4bc
PB
440void qemu_aio_set_fd_handler(int fd,
441 IOHandler *io_read,
442 IOHandler *io_write,
443 AioFlushHandler *io_flush,
444 void *opaque)
445{
446 aio_set_fd_handler(qemu_aio_context, fd, io_read, io_write, io_flush,
447 opaque);
a915f4bc 448}
82cbbdc6 449#endif
a915f4bc 450
a915f4bc
PB
451void qemu_aio_set_event_notifier(EventNotifier *notifier,
452 EventNotifierHandler *io_read,
453 AioFlushEventNotifierHandler *io_flush)
454{
82cbbdc6 455 aio_set_event_notifier(qemu_aio_context, notifier, io_read, io_flush);
a915f4bc 456}