]>
Commit | Line | Data |
---|---|---|
296af7c9 BS |
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 | */ | |
24 | ||
25 | /* Needed early for CONFIG_BSD etc. */ | |
26 | #include "config-host.h" | |
27 | ||
28 | #include "monitor.h" | |
29 | #include "sysemu.h" | |
30 | #include "gdbstub.h" | |
31 | #include "dma.h" | |
32 | #include "kvm.h" | |
262ea18e | 33 | #include "exec-all.h" |
296af7c9 BS |
34 | |
35 | #include "cpus.h" | |
a8486bc9 | 36 | #include "compatfd.h" |
c0532a76 MT |
37 | #ifdef CONFIG_LINUX |
38 | #include <sys/prctl.h> | |
39 | #endif | |
296af7c9 | 40 | |
7277e027 BS |
41 | #ifdef SIGRTMIN |
42 | #define SIG_IPI (SIGRTMIN+4) | |
43 | #else | |
44 | #define SIG_IPI SIGUSR1 | |
45 | #endif | |
46 | ||
c0532a76 MT |
47 | #ifndef PR_MCE_KILL |
48 | #define PR_MCE_KILL 33 | |
49 | #endif | |
50 | ||
296af7c9 BS |
51 | static CPUState *next_cpu; |
52 | ||
53 | /***********************************************************/ | |
54 | void hw_error(const char *fmt, ...) | |
55 | { | |
56 | va_list ap; | |
57 | CPUState *env; | |
58 | ||
59 | va_start(ap, fmt); | |
60 | fprintf(stderr, "qemu: hardware error: "); | |
61 | vfprintf(stderr, fmt, ap); | |
62 | fprintf(stderr, "\n"); | |
63 | for(env = first_cpu; env != NULL; env = env->next_cpu) { | |
64 | fprintf(stderr, "CPU #%d:\n", env->cpu_index); | |
65 | #ifdef TARGET_I386 | |
66 | cpu_dump_state(env, stderr, fprintf, X86_DUMP_FPU); | |
67 | #else | |
68 | cpu_dump_state(env, stderr, fprintf, 0); | |
69 | #endif | |
70 | } | |
71 | va_end(ap); | |
72 | abort(); | |
73 | } | |
74 | ||
75 | void cpu_synchronize_all_states(void) | |
76 | { | |
77 | CPUState *cpu; | |
78 | ||
79 | for (cpu = first_cpu; cpu; cpu = cpu->next_cpu) { | |
80 | cpu_synchronize_state(cpu); | |
81 | } | |
82 | } | |
83 | ||
84 | void cpu_synchronize_all_post_reset(void) | |
85 | { | |
86 | CPUState *cpu; | |
87 | ||
88 | for (cpu = first_cpu; cpu; cpu = cpu->next_cpu) { | |
89 | cpu_synchronize_post_reset(cpu); | |
90 | } | |
91 | } | |
92 | ||
93 | void cpu_synchronize_all_post_init(void) | |
94 | { | |
95 | CPUState *cpu; | |
96 | ||
97 | for (cpu = first_cpu; cpu; cpu = cpu->next_cpu) { | |
98 | cpu_synchronize_post_init(cpu); | |
99 | } | |
100 | } | |
101 | ||
3ae9501c MT |
102 | int cpu_is_stopped(CPUState *env) |
103 | { | |
104 | return !vm_running || env->stopped; | |
105 | } | |
106 | ||
296af7c9 BS |
107 | static void do_vm_stop(int reason) |
108 | { | |
109 | if (vm_running) { | |
110 | cpu_disable_ticks(); | |
111 | vm_running = 0; | |
112 | pause_all_vcpus(); | |
113 | vm_state_notify(0, reason); | |
55df6f33 MT |
114 | qemu_aio_flush(); |
115 | bdrv_flush_all(); | |
296af7c9 BS |
116 | monitor_protocol_event(QEVENT_STOP, NULL); |
117 | } | |
118 | } | |
119 | ||
120 | static int cpu_can_run(CPUState *env) | |
121 | { | |
122 | if (env->stop) | |
123 | return 0; | |
55274a30 | 124 | if (env->stopped || !vm_running) |
296af7c9 BS |
125 | return 0; |
126 | return 1; | |
127 | } | |
128 | ||
129 | static int cpu_has_work(CPUState *env) | |
130 | { | |
131 | if (env->stop) | |
132 | return 1; | |
e82bcec2 MT |
133 | if (env->queued_work_first) |
134 | return 1; | |
55274a30 | 135 | if (env->stopped || !vm_running) |
296af7c9 BS |
136 | return 0; |
137 | if (!env->halted) | |
138 | return 1; | |
139 | if (qemu_cpu_has_work(env)) | |
140 | return 1; | |
141 | return 0; | |
142 | } | |
143 | ||
472fb0c4 | 144 | static int any_cpu_has_work(void) |
296af7c9 BS |
145 | { |
146 | CPUState *env; | |
147 | ||
148 | for (env = first_cpu; env != NULL; env = env->next_cpu) | |
149 | if (cpu_has_work(env)) | |
150 | return 1; | |
151 | return 0; | |
152 | } | |
153 | ||
3c638d06 JK |
154 | static void cpu_debug_handler(CPUState *env) |
155 | { | |
156 | gdb_set_stop_cpu(env); | |
157 | debug_requested = EXCP_DEBUG; | |
158 | vm_stop(EXCP_DEBUG); | |
159 | } | |
160 | ||
296af7c9 BS |
161 | #ifndef _WIN32 |
162 | static int io_thread_fd = -1; | |
163 | ||
164 | static void qemu_event_increment(void) | |
165 | { | |
166 | /* Write 8 bytes to be compatible with eventfd. */ | |
26a82330 | 167 | static const uint64_t val = 1; |
296af7c9 BS |
168 | ssize_t ret; |
169 | ||
170 | if (io_thread_fd == -1) | |
171 | return; | |
172 | ||
173 | do { | |
174 | ret = write(io_thread_fd, &val, sizeof(val)); | |
175 | } while (ret < 0 && errno == EINTR); | |
176 | ||
177 | /* EAGAIN is fine, a read must be pending. */ | |
178 | if (ret < 0 && errno != EAGAIN) { | |
179 | fprintf(stderr, "qemu_event_increment: write() filed: %s\n", | |
180 | strerror(errno)); | |
181 | exit (1); | |
182 | } | |
183 | } | |
184 | ||
185 | static void qemu_event_read(void *opaque) | |
186 | { | |
187 | int fd = (unsigned long)opaque; | |
188 | ssize_t len; | |
189 | char buffer[512]; | |
190 | ||
191 | /* Drain the notify pipe. For eventfd, only 8 bytes will be read. */ | |
192 | do { | |
193 | len = read(fd, buffer, sizeof(buffer)); | |
194 | } while ((len == -1 && errno == EINTR) || len == sizeof(buffer)); | |
195 | } | |
196 | ||
197 | static int qemu_event_init(void) | |
198 | { | |
199 | int err; | |
200 | int fds[2]; | |
201 | ||
202 | err = qemu_eventfd(fds); | |
203 | if (err == -1) | |
204 | return -errno; | |
205 | ||
206 | err = fcntl_setfl(fds[0], O_NONBLOCK); | |
207 | if (err < 0) | |
208 | goto fail; | |
209 | ||
210 | err = fcntl_setfl(fds[1], O_NONBLOCK); | |
211 | if (err < 0) | |
212 | goto fail; | |
213 | ||
214 | qemu_set_fd_handler2(fds[0], NULL, qemu_event_read, NULL, | |
215 | (void *)(unsigned long)fds[0]); | |
216 | ||
217 | io_thread_fd = fds[1]; | |
218 | return 0; | |
219 | ||
220 | fail: | |
221 | close(fds[0]); | |
222 | close(fds[1]); | |
223 | return err; | |
224 | } | |
55f8d6ac | 225 | |
55f8d6ac JK |
226 | static void dummy_signal(int sig) |
227 | { | |
228 | } | |
55f8d6ac | 229 | |
9a36085b JK |
230 | static void sigbus_reraise(void); |
231 | ||
232 | static void qemu_kvm_eat_signals(CPUState *env) | |
233 | { | |
234 | struct timespec ts = { 0, 0 }; | |
235 | siginfo_t siginfo; | |
236 | sigset_t waitset; | |
237 | sigset_t chkset; | |
238 | int r; | |
239 | ||
240 | sigemptyset(&waitset); | |
241 | sigaddset(&waitset, SIG_IPI); | |
242 | sigaddset(&waitset, SIGBUS); | |
243 | ||
244 | do { | |
245 | r = sigtimedwait(&waitset, &siginfo, &ts); | |
246 | if (r == -1 && !(errno == EAGAIN || errno == EINTR)) { | |
247 | perror("sigtimedwait"); | |
248 | exit(1); | |
249 | } | |
250 | ||
251 | switch (r) { | |
252 | #ifdef CONFIG_IOTHREAD | |
253 | case SIGBUS: | |
254 | if (kvm_on_sigbus_vcpu(env, siginfo.si_code, siginfo.si_addr)) { | |
255 | sigbus_reraise(); | |
256 | } | |
257 | break; | |
258 | #endif | |
259 | default: | |
260 | break; | |
261 | } | |
262 | ||
263 | r = sigpending(&chkset); | |
264 | if (r == -1) { | |
265 | perror("sigpending"); | |
266 | exit(1); | |
267 | } | |
268 | } while (sigismember(&chkset, SIG_IPI) || sigismember(&chkset, SIGBUS)); | |
269 | } | |
270 | ||
55f8d6ac JK |
271 | #else /* _WIN32 */ |
272 | ||
296af7c9 BS |
273 | HANDLE qemu_event_handle; |
274 | ||
275 | static void dummy_event_handler(void *opaque) | |
276 | { | |
277 | } | |
278 | ||
279 | static int qemu_event_init(void) | |
280 | { | |
281 | qemu_event_handle = CreateEvent(NULL, FALSE, FALSE, NULL); | |
282 | if (!qemu_event_handle) { | |
283 | fprintf(stderr, "Failed CreateEvent: %ld\n", GetLastError()); | |
284 | return -1; | |
285 | } | |
286 | qemu_add_wait_object(qemu_event_handle, dummy_event_handler, NULL); | |
287 | return 0; | |
288 | } | |
289 | ||
290 | static void qemu_event_increment(void) | |
291 | { | |
292 | if (!SetEvent(qemu_event_handle)) { | |
293 | fprintf(stderr, "qemu_event_increment: SetEvent failed: %ld\n", | |
294 | GetLastError()); | |
295 | exit (1); | |
296 | } | |
297 | } | |
9a36085b JK |
298 | |
299 | static void qemu_kvm_eat_signals(CPUState *env) | |
300 | { | |
301 | } | |
55f8d6ac | 302 | #endif /* _WIN32 */ |
296af7c9 BS |
303 | |
304 | #ifndef CONFIG_IOTHREAD | |
ff48eb5f JK |
305 | static void qemu_kvm_init_cpu_signals(CPUState *env) |
306 | { | |
307 | #ifndef _WIN32 | |
308 | int r; | |
309 | sigset_t set; | |
310 | struct sigaction sigact; | |
311 | ||
312 | memset(&sigact, 0, sizeof(sigact)); | |
313 | sigact.sa_handler = dummy_signal; | |
314 | sigaction(SIG_IPI, &sigact, NULL); | |
315 | ||
316 | sigemptyset(&set); | |
317 | sigaddset(&set, SIG_IPI); | |
318 | pthread_sigmask(SIG_BLOCK, &set, NULL); | |
319 | ||
320 | pthread_sigmask(SIG_BLOCK, NULL, &set); | |
321 | sigdelset(&set, SIG_IPI); | |
322 | sigdelset(&set, SIGBUS); | |
323 | r = kvm_set_signal_mask(env, &set); | |
324 | if (r) { | |
325 | fprintf(stderr, "kvm_set_signal_mask: %s\n", strerror(-r)); | |
326 | exit(1); | |
327 | } | |
328 | #endif | |
329 | } | |
330 | ||
296af7c9 BS |
331 | int qemu_init_main_loop(void) |
332 | { | |
3c638d06 JK |
333 | cpu_set_debug_excp_handler(cpu_debug_handler); |
334 | ||
296af7c9 BS |
335 | return qemu_event_init(); |
336 | } | |
337 | ||
7277e027 BS |
338 | void qemu_main_loop_start(void) |
339 | { | |
340 | } | |
341 | ||
296af7c9 BS |
342 | void qemu_init_vcpu(void *_env) |
343 | { | |
344 | CPUState *env = _env; | |
84b4915d | 345 | int r; |
296af7c9 BS |
346 | |
347 | env->nr_cores = smp_cores; | |
348 | env->nr_threads = smp_threads; | |
84b4915d JK |
349 | |
350 | if (kvm_enabled()) { | |
351 | r = kvm_init_vcpu(env); | |
352 | if (r < 0) { | |
353 | fprintf(stderr, "kvm_init_vcpu failed: %s\n", strerror(-r)); | |
354 | exit(1); | |
355 | } | |
ff48eb5f | 356 | qemu_kvm_init_cpu_signals(env); |
84b4915d | 357 | } |
296af7c9 BS |
358 | } |
359 | ||
360 | int qemu_cpu_self(void *env) | |
361 | { | |
362 | return 1; | |
363 | } | |
364 | ||
e82bcec2 MT |
365 | void run_on_cpu(CPUState *env, void (*func)(void *data), void *data) |
366 | { | |
367 | func(data); | |
368 | } | |
369 | ||
296af7c9 BS |
370 | void resume_all_vcpus(void) |
371 | { | |
372 | } | |
373 | ||
374 | void pause_all_vcpus(void) | |
375 | { | |
376 | } | |
377 | ||
378 | void qemu_cpu_kick(void *env) | |
379 | { | |
380 | return; | |
381 | } | |
382 | ||
383 | void qemu_notify_event(void) | |
384 | { | |
385 | CPUState *env = cpu_single_env; | |
386 | ||
387 | qemu_event_increment (); | |
388 | if (env) { | |
389 | cpu_exit(env); | |
390 | } | |
391 | if (next_cpu && env != next_cpu) { | |
392 | cpu_exit(next_cpu); | |
393 | } | |
38145df2 | 394 | exit_request = 1; |
296af7c9 BS |
395 | } |
396 | ||
397 | void qemu_mutex_lock_iothread(void) {} | |
398 | void qemu_mutex_unlock_iothread(void) {} | |
399 | ||
b4a3d965 JK |
400 | void cpu_stop_current(void) |
401 | { | |
402 | } | |
403 | ||
296af7c9 BS |
404 | void vm_stop(int reason) |
405 | { | |
406 | do_vm_stop(reason); | |
407 | } | |
408 | ||
409 | #else /* CONFIG_IOTHREAD */ | |
410 | ||
411 | #include "qemu-thread.h" | |
412 | ||
413 | QemuMutex qemu_global_mutex; | |
414 | static QemuMutex qemu_fair_mutex; | |
415 | ||
416 | static QemuThread io_thread; | |
417 | ||
418 | static QemuThread *tcg_cpu_thread; | |
419 | static QemuCond *tcg_halt_cond; | |
420 | ||
421 | static int qemu_system_ready; | |
422 | /* cpu creation */ | |
423 | static QemuCond qemu_cpu_cond; | |
424 | /* system init */ | |
425 | static QemuCond qemu_system_cond; | |
426 | static QemuCond qemu_pause_cond; | |
e82bcec2 | 427 | static QemuCond qemu_work_cond; |
296af7c9 | 428 | |
a8486bc9 MT |
429 | /* If we have signalfd, we mask out the signals we want to handle and then |
430 | * use signalfd to listen for them. We rely on whatever the current signal | |
431 | * handler is to dispatch the signals when we receive them. | |
432 | */ | |
433 | static void sigfd_handler(void *opaque) | |
434 | { | |
435 | int fd = (unsigned long) opaque; | |
436 | struct qemu_signalfd_siginfo info; | |
437 | struct sigaction action; | |
438 | ssize_t len; | |
439 | ||
440 | while (1) { | |
441 | do { | |
442 | len = read(fd, &info, sizeof(info)); | |
443 | } while (len == -1 && errno == EINTR); | |
444 | ||
445 | if (len == -1 && errno == EAGAIN) { | |
446 | break; | |
447 | } | |
448 | ||
449 | if (len != sizeof(info)) { | |
450 | printf("read from sigfd returned %zd: %m\n", len); | |
451 | return; | |
452 | } | |
453 | ||
454 | sigaction(info.ssi_signo, NULL, &action); | |
455 | if ((action.sa_flags & SA_SIGINFO) && action.sa_sigaction) { | |
456 | action.sa_sigaction(info.ssi_signo, | |
457 | (siginfo_t *)&info, NULL); | |
458 | } else if (action.sa_handler) { | |
459 | action.sa_handler(info.ssi_signo); | |
460 | } | |
461 | } | |
462 | } | |
463 | ||
55f8d6ac JK |
464 | static void cpu_signal(int sig) |
465 | { | |
466 | if (cpu_single_env) { | |
467 | cpu_exit(cpu_single_env); | |
468 | } | |
469 | exit_request = 1; | |
470 | } | |
471 | ||
472 | static void qemu_kvm_init_cpu_signals(CPUState *env) | |
473 | { | |
474 | int r; | |
475 | sigset_t set; | |
476 | struct sigaction sigact; | |
477 | ||
478 | memset(&sigact, 0, sizeof(sigact)); | |
479 | sigact.sa_handler = dummy_signal; | |
480 | sigaction(SIG_IPI, &sigact, NULL); | |
481 | ||
482 | pthread_sigmask(SIG_BLOCK, NULL, &set); | |
483 | sigdelset(&set, SIG_IPI); | |
484 | sigdelset(&set, SIGBUS); | |
485 | r = kvm_set_signal_mask(env, &set); | |
486 | if (r) { | |
487 | fprintf(stderr, "kvm_set_signal_mask: %s\n", strerror(-r)); | |
488 | exit(1); | |
489 | } | |
490 | } | |
491 | ||
492 | static void qemu_tcg_init_cpu_signals(void) | |
493 | { | |
494 | sigset_t set; | |
495 | struct sigaction sigact; | |
496 | ||
497 | memset(&sigact, 0, sizeof(sigact)); | |
498 | sigact.sa_handler = cpu_signal; | |
499 | sigaction(SIG_IPI, &sigact, NULL); | |
500 | ||
501 | sigemptyset(&set); | |
502 | sigaddset(&set, SIG_IPI); | |
503 | pthread_sigmask(SIG_UNBLOCK, &set, NULL); | |
504 | } | |
505 | ||
506 | static void sigbus_handler(int n, struct qemu_signalfd_siginfo *siginfo, | |
507 | void *ctx); | |
508 | ||
509 | static sigset_t block_io_signals(void) | |
510 | { | |
511 | sigset_t set; | |
512 | struct sigaction action; | |
513 | ||
514 | /* SIGUSR2 used by posix-aio-compat.c */ | |
515 | sigemptyset(&set); | |
516 | sigaddset(&set, SIGUSR2); | |
517 | pthread_sigmask(SIG_UNBLOCK, &set, NULL); | |
518 | ||
519 | sigemptyset(&set); | |
520 | sigaddset(&set, SIGIO); | |
521 | sigaddset(&set, SIGALRM); | |
522 | sigaddset(&set, SIG_IPI); | |
523 | sigaddset(&set, SIGBUS); | |
524 | pthread_sigmask(SIG_BLOCK, &set, NULL); | |
525 | ||
526 | memset(&action, 0, sizeof(action)); | |
527 | action.sa_flags = SA_SIGINFO; | |
528 | action.sa_sigaction = (void (*)(int, siginfo_t*, void*))sigbus_handler; | |
529 | sigaction(SIGBUS, &action, NULL); | |
530 | prctl(PR_MCE_KILL, 1, 1, 0, 0); | |
531 | ||
532 | return set; | |
533 | } | |
534 | ||
a8486bc9 MT |
535 | static int qemu_signalfd_init(sigset_t mask) |
536 | { | |
537 | int sigfd; | |
538 | ||
539 | sigfd = qemu_signalfd(&mask); | |
540 | if (sigfd == -1) { | |
541 | fprintf(stderr, "failed to create signalfd\n"); | |
542 | return -errno; | |
543 | } | |
544 | ||
545 | fcntl_setfl(sigfd, O_NONBLOCK); | |
546 | ||
547 | qemu_set_fd_handler2(sigfd, NULL, sigfd_handler, NULL, | |
548 | (void *)(unsigned long) sigfd); | |
549 | ||
550 | return 0; | |
551 | } | |
296af7c9 BS |
552 | |
553 | int qemu_init_main_loop(void) | |
554 | { | |
555 | int ret; | |
a8486bc9 | 556 | sigset_t blocked_signals; |
296af7c9 | 557 | |
3c638d06 JK |
558 | cpu_set_debug_excp_handler(cpu_debug_handler); |
559 | ||
a8486bc9 MT |
560 | blocked_signals = block_io_signals(); |
561 | ||
562 | ret = qemu_signalfd_init(blocked_signals); | |
563 | if (ret) | |
564 | return ret; | |
565 | ||
566 | /* Note eventfd must be drained before signalfd handlers run */ | |
296af7c9 BS |
567 | ret = qemu_event_init(); |
568 | if (ret) | |
569 | return ret; | |
570 | ||
571 | qemu_cond_init(&qemu_pause_cond); | |
f8ca7b43 | 572 | qemu_cond_init(&qemu_system_cond); |
296af7c9 BS |
573 | qemu_mutex_init(&qemu_fair_mutex); |
574 | qemu_mutex_init(&qemu_global_mutex); | |
575 | qemu_mutex_lock(&qemu_global_mutex); | |
576 | ||
296af7c9 BS |
577 | qemu_thread_self(&io_thread); |
578 | ||
579 | return 0; | |
580 | } | |
581 | ||
7277e027 BS |
582 | void qemu_main_loop_start(void) |
583 | { | |
584 | qemu_system_ready = 1; | |
585 | qemu_cond_broadcast(&qemu_system_cond); | |
586 | } | |
587 | ||
e82bcec2 MT |
588 | void run_on_cpu(CPUState *env, void (*func)(void *data), void *data) |
589 | { | |
590 | struct qemu_work_item wi; | |
591 | ||
592 | if (qemu_cpu_self(env)) { | |
593 | func(data); | |
594 | return; | |
595 | } | |
596 | ||
597 | wi.func = func; | |
598 | wi.data = data; | |
599 | if (!env->queued_work_first) | |
600 | env->queued_work_first = &wi; | |
601 | else | |
602 | env->queued_work_last->next = &wi; | |
603 | env->queued_work_last = &wi; | |
604 | wi.next = NULL; | |
605 | wi.done = false; | |
606 | ||
607 | qemu_cpu_kick(env); | |
608 | while (!wi.done) { | |
609 | CPUState *self_env = cpu_single_env; | |
610 | ||
611 | qemu_cond_wait(&qemu_work_cond, &qemu_global_mutex); | |
612 | cpu_single_env = self_env; | |
613 | } | |
614 | } | |
615 | ||
616 | static void flush_queued_work(CPUState *env) | |
617 | { | |
618 | struct qemu_work_item *wi; | |
619 | ||
620 | if (!env->queued_work_first) | |
621 | return; | |
622 | ||
623 | while ((wi = env->queued_work_first)) { | |
624 | env->queued_work_first = wi->next; | |
625 | wi->func(wi->data); | |
626 | wi->done = true; | |
627 | } | |
628 | env->queued_work_last = NULL; | |
629 | qemu_cond_broadcast(&qemu_work_cond); | |
630 | } | |
631 | ||
296af7c9 BS |
632 | static void qemu_wait_io_event_common(CPUState *env) |
633 | { | |
634 | if (env->stop) { | |
635 | env->stop = 0; | |
636 | env->stopped = 1; | |
637 | qemu_cond_signal(&qemu_pause_cond); | |
638 | } | |
e82bcec2 | 639 | flush_queued_work(env); |
aa2c364b | 640 | env->thread_kicked = false; |
296af7c9 BS |
641 | } |
642 | ||
6cabe1f3 | 643 | static void qemu_tcg_wait_io_event(void) |
296af7c9 | 644 | { |
6cabe1f3 JK |
645 | CPUState *env; |
646 | ||
472fb0c4 | 647 | while (!any_cpu_has_work()) |
6cabe1f3 | 648 | qemu_cond_timedwait(tcg_halt_cond, &qemu_global_mutex, 1000); |
296af7c9 BS |
649 | |
650 | qemu_mutex_unlock(&qemu_global_mutex); | |
651 | ||
652 | /* | |
653 | * Users of qemu_global_mutex can be starved, having no chance | |
654 | * to acquire it since this path will get to it first. | |
655 | * So use another lock to provide fairness. | |
656 | */ | |
657 | qemu_mutex_lock(&qemu_fair_mutex); | |
658 | qemu_mutex_unlock(&qemu_fair_mutex); | |
659 | ||
660 | qemu_mutex_lock(&qemu_global_mutex); | |
6cabe1f3 JK |
661 | |
662 | for (env = first_cpu; env != NULL; env = env->next_cpu) { | |
663 | qemu_wait_io_event_common(env); | |
664 | } | |
296af7c9 BS |
665 | } |
666 | ||
c0532a76 MT |
667 | static void sigbus_reraise(void) |
668 | { | |
669 | sigset_t set; | |
670 | struct sigaction action; | |
671 | ||
672 | memset(&action, 0, sizeof(action)); | |
673 | action.sa_handler = SIG_DFL; | |
674 | if (!sigaction(SIGBUS, &action, NULL)) { | |
675 | raise(SIGBUS); | |
676 | sigemptyset(&set); | |
677 | sigaddset(&set, SIGBUS); | |
678 | sigprocmask(SIG_UNBLOCK, &set, NULL); | |
679 | } | |
680 | perror("Failed to re-raise SIGBUS!\n"); | |
681 | abort(); | |
682 | } | |
683 | ||
684 | static void sigbus_handler(int n, struct qemu_signalfd_siginfo *siginfo, | |
685 | void *ctx) | |
686 | { | |
a1b87fe0 | 687 | if (kvm_on_sigbus(siginfo->ssi_code, (void *)(intptr_t)siginfo->ssi_addr)) { |
c0532a76 | 688 | sigbus_reraise(); |
a1b87fe0 | 689 | } |
c0532a76 MT |
690 | } |
691 | ||
296af7c9 BS |
692 | static void qemu_kvm_wait_io_event(CPUState *env) |
693 | { | |
694 | while (!cpu_has_work(env)) | |
695 | qemu_cond_timedwait(env->halt_cond, &qemu_global_mutex, 1000); | |
696 | ||
5db5bdac | 697 | qemu_kvm_eat_signals(env); |
296af7c9 BS |
698 | qemu_wait_io_event_common(env); |
699 | } | |
700 | ||
701 | static int qemu_cpu_exec(CPUState *env); | |
702 | ||
703 | static void *kvm_cpu_thread_fn(void *arg) | |
704 | { | |
705 | CPUState *env = arg; | |
84b4915d | 706 | int r; |
296af7c9 | 707 | |
6164e6d6 | 708 | qemu_mutex_lock(&qemu_global_mutex); |
296af7c9 | 709 | qemu_thread_self(env->thread); |
d31ae052 | 710 | |
84b4915d JK |
711 | r = kvm_init_vcpu(env); |
712 | if (r < 0) { | |
713 | fprintf(stderr, "kvm_init_vcpu failed: %s\n", strerror(-r)); | |
714 | exit(1); | |
715 | } | |
296af7c9 | 716 | |
55f8d6ac | 717 | qemu_kvm_init_cpu_signals(env); |
296af7c9 BS |
718 | |
719 | /* signal CPU creation */ | |
296af7c9 BS |
720 | env->created = 1; |
721 | qemu_cond_signal(&qemu_cpu_cond); | |
722 | ||
723 | /* and wait for machine initialization */ | |
724 | while (!qemu_system_ready) | |
725 | qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100); | |
726 | ||
727 | while (1) { | |
728 | if (cpu_can_run(env)) | |
729 | qemu_cpu_exec(env); | |
730 | qemu_kvm_wait_io_event(env); | |
731 | } | |
732 | ||
733 | return NULL; | |
734 | } | |
735 | ||
736 | static void *tcg_cpu_thread_fn(void *arg) | |
737 | { | |
738 | CPUState *env = arg; | |
739 | ||
55f8d6ac | 740 | qemu_tcg_init_cpu_signals(); |
296af7c9 BS |
741 | qemu_thread_self(env->thread); |
742 | ||
743 | /* signal CPU creation */ | |
744 | qemu_mutex_lock(&qemu_global_mutex); | |
745 | for (env = first_cpu; env != NULL; env = env->next_cpu) | |
746 | env->created = 1; | |
747 | qemu_cond_signal(&qemu_cpu_cond); | |
748 | ||
749 | /* and wait for machine initialization */ | |
750 | while (!qemu_system_ready) | |
751 | qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100); | |
752 | ||
753 | while (1) { | |
472fb0c4 | 754 | cpu_exec_all(); |
6cabe1f3 | 755 | qemu_tcg_wait_io_event(); |
296af7c9 BS |
756 | } |
757 | ||
758 | return NULL; | |
759 | } | |
760 | ||
761 | void qemu_cpu_kick(void *_env) | |
762 | { | |
763 | CPUState *env = _env; | |
764 | qemu_cond_broadcast(env->halt_cond); | |
aa2c364b JK |
765 | if (!env->thread_kicked) { |
766 | qemu_thread_signal(env->thread, SIG_IPI); | |
767 | env->thread_kicked = true; | |
768 | } | |
296af7c9 BS |
769 | } |
770 | ||
771 | int qemu_cpu_self(void *_env) | |
772 | { | |
773 | CPUState *env = _env; | |
774 | QemuThread this; | |
775 | ||
776 | qemu_thread_self(&this); | |
777 | ||
778 | return qemu_thread_equal(&this, env->thread); | |
779 | } | |
780 | ||
296af7c9 BS |
781 | void qemu_mutex_lock_iothread(void) |
782 | { | |
783 | if (kvm_enabled()) { | |
296af7c9 | 784 | qemu_mutex_lock(&qemu_global_mutex); |
1a28cac3 MT |
785 | } else { |
786 | qemu_mutex_lock(&qemu_fair_mutex); | |
787 | if (qemu_mutex_trylock(&qemu_global_mutex)) { | |
788 | qemu_thread_signal(tcg_cpu_thread, SIG_IPI); | |
789 | qemu_mutex_lock(&qemu_global_mutex); | |
790 | } | |
791 | qemu_mutex_unlock(&qemu_fair_mutex); | |
792 | } | |
296af7c9 BS |
793 | } |
794 | ||
795 | void qemu_mutex_unlock_iothread(void) | |
796 | { | |
797 | qemu_mutex_unlock(&qemu_global_mutex); | |
798 | } | |
799 | ||
800 | static int all_vcpus_paused(void) | |
801 | { | |
802 | CPUState *penv = first_cpu; | |
803 | ||
804 | while (penv) { | |
805 | if (!penv->stopped) | |
806 | return 0; | |
807 | penv = (CPUState *)penv->next_cpu; | |
808 | } | |
809 | ||
810 | return 1; | |
811 | } | |
812 | ||
813 | void pause_all_vcpus(void) | |
814 | { | |
815 | CPUState *penv = first_cpu; | |
816 | ||
817 | while (penv) { | |
818 | penv->stop = 1; | |
296af7c9 BS |
819 | qemu_cpu_kick(penv); |
820 | penv = (CPUState *)penv->next_cpu; | |
821 | } | |
822 | ||
823 | while (!all_vcpus_paused()) { | |
824 | qemu_cond_timedwait(&qemu_pause_cond, &qemu_global_mutex, 100); | |
825 | penv = first_cpu; | |
826 | while (penv) { | |
1fbb22e5 | 827 | qemu_cpu_kick(penv); |
296af7c9 BS |
828 | penv = (CPUState *)penv->next_cpu; |
829 | } | |
830 | } | |
831 | } | |
832 | ||
833 | void resume_all_vcpus(void) | |
834 | { | |
835 | CPUState *penv = first_cpu; | |
836 | ||
837 | while (penv) { | |
838 | penv->stop = 0; | |
839 | penv->stopped = 0; | |
296af7c9 BS |
840 | qemu_cpu_kick(penv); |
841 | penv = (CPUState *)penv->next_cpu; | |
842 | } | |
843 | } | |
844 | ||
845 | static void tcg_init_vcpu(void *_env) | |
846 | { | |
847 | CPUState *env = _env; | |
848 | /* share a single thread for all cpus with TCG */ | |
849 | if (!tcg_cpu_thread) { | |
850 | env->thread = qemu_mallocz(sizeof(QemuThread)); | |
851 | env->halt_cond = qemu_mallocz(sizeof(QemuCond)); | |
852 | qemu_cond_init(env->halt_cond); | |
853 | qemu_thread_create(env->thread, tcg_cpu_thread_fn, env); | |
854 | while (env->created == 0) | |
855 | qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100); | |
856 | tcg_cpu_thread = env->thread; | |
857 | tcg_halt_cond = env->halt_cond; | |
858 | } else { | |
859 | env->thread = tcg_cpu_thread; | |
860 | env->halt_cond = tcg_halt_cond; | |
861 | } | |
862 | } | |
863 | ||
864 | static void kvm_start_vcpu(CPUState *env) | |
865 | { | |
866 | env->thread = qemu_mallocz(sizeof(QemuThread)); | |
867 | env->halt_cond = qemu_mallocz(sizeof(QemuCond)); | |
868 | qemu_cond_init(env->halt_cond); | |
869 | qemu_thread_create(env->thread, kvm_cpu_thread_fn, env); | |
870 | while (env->created == 0) | |
871 | qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100); | |
872 | } | |
873 | ||
874 | void qemu_init_vcpu(void *_env) | |
875 | { | |
876 | CPUState *env = _env; | |
877 | ||
878 | env->nr_cores = smp_cores; | |
879 | env->nr_threads = smp_threads; | |
880 | if (kvm_enabled()) | |
881 | kvm_start_vcpu(env); | |
882 | else | |
883 | tcg_init_vcpu(env); | |
884 | } | |
885 | ||
886 | void qemu_notify_event(void) | |
887 | { | |
888 | qemu_event_increment(); | |
889 | } | |
890 | ||
891 | static void qemu_system_vmstop_request(int reason) | |
892 | { | |
893 | vmstop_requested = reason; | |
894 | qemu_notify_event(); | |
895 | } | |
896 | ||
b4a3d965 JK |
897 | void cpu_stop_current(void) |
898 | { | |
899 | if (cpu_single_env) { | |
900 | cpu_single_env->stopped = 1; | |
901 | cpu_exit(cpu_single_env); | |
902 | } | |
903 | } | |
904 | ||
296af7c9 BS |
905 | void vm_stop(int reason) |
906 | { | |
907 | QemuThread me; | |
908 | qemu_thread_self(&me); | |
909 | ||
910 | if (!qemu_thread_equal(&me, &io_thread)) { | |
911 | qemu_system_vmstop_request(reason); | |
912 | /* | |
913 | * FIXME: should not return to device code in case | |
914 | * vm_stop() has been requested. | |
915 | */ | |
b4a3d965 | 916 | cpu_stop_current(); |
296af7c9 BS |
917 | return; |
918 | } | |
919 | do_vm_stop(reason); | |
920 | } | |
921 | ||
922 | #endif | |
923 | ||
924 | static int qemu_cpu_exec(CPUState *env) | |
925 | { | |
926 | int ret; | |
927 | #ifdef CONFIG_PROFILER | |
928 | int64_t ti; | |
929 | #endif | |
930 | ||
931 | #ifdef CONFIG_PROFILER | |
932 | ti = profile_getclock(); | |
933 | #endif | |
934 | if (use_icount) { | |
935 | int64_t count; | |
936 | int decr; | |
937 | qemu_icount -= (env->icount_decr.u16.low + env->icount_extra); | |
938 | env->icount_decr.u16.low = 0; | |
939 | env->icount_extra = 0; | |
940 | count = qemu_icount_round (qemu_next_deadline()); | |
941 | qemu_icount += count; | |
942 | decr = (count > 0xffff) ? 0xffff : count; | |
943 | count -= decr; | |
944 | env->icount_decr.u16.low = decr; | |
945 | env->icount_extra = count; | |
946 | } | |
947 | ret = cpu_exec(env); | |
948 | #ifdef CONFIG_PROFILER | |
949 | qemu_time += profile_getclock() - ti; | |
950 | #endif | |
951 | if (use_icount) { | |
952 | /* Fold pending instructions back into the | |
953 | instruction counter, and clear the interrupt flag. */ | |
954 | qemu_icount -= (env->icount_decr.u16.low | |
955 | + env->icount_extra); | |
956 | env->icount_decr.u32 = 0; | |
957 | env->icount_extra = 0; | |
958 | } | |
959 | return ret; | |
960 | } | |
961 | ||
472fb0c4 | 962 | bool cpu_exec_all(void) |
296af7c9 | 963 | { |
9a36085b JK |
964 | int r; |
965 | ||
296af7c9 BS |
966 | if (next_cpu == NULL) |
967 | next_cpu = first_cpu; | |
c629a4bc | 968 | for (; next_cpu != NULL && !exit_request; next_cpu = next_cpu->next_cpu) { |
345f4426 | 969 | CPUState *env = next_cpu; |
296af7c9 BS |
970 | |
971 | qemu_clock_enable(vm_clock, | |
345f4426 | 972 | (env->singlestep_enabled & SSTEP_NOTIMER) == 0); |
296af7c9 BS |
973 | |
974 | if (qemu_alarm_pending()) | |
975 | break; | |
3c638d06 | 976 | if (cpu_can_run(env)) { |
9a36085b JK |
977 | r = qemu_cpu_exec(env); |
978 | if (kvm_enabled()) { | |
979 | qemu_kvm_eat_signals(env); | |
980 | } | |
981 | if (r == EXCP_DEBUG) { | |
3c638d06 JK |
982 | break; |
983 | } | |
984 | } else if (env->stop) { | |
296af7c9 BS |
985 | break; |
986 | } | |
987 | } | |
c629a4bc | 988 | exit_request = 0; |
472fb0c4 | 989 | return any_cpu_has_work(); |
296af7c9 BS |
990 | } |
991 | ||
992 | void set_numa_modes(void) | |
993 | { | |
994 | CPUState *env; | |
995 | int i; | |
996 | ||
997 | for (env = first_cpu; env != NULL; env = env->next_cpu) { | |
998 | for (i = 0; i < nb_numa_nodes; i++) { | |
999 | if (node_cpumask[i] & (1 << env->cpu_index)) { | |
1000 | env->numa_node = i; | |
1001 | } | |
1002 | } | |
1003 | } | |
1004 | } | |
1005 | ||
1006 | void set_cpu_log(const char *optarg) | |
1007 | { | |
1008 | int mask; | |
1009 | const CPULogItem *item; | |
1010 | ||
1011 | mask = cpu_str_to_log_mask(optarg); | |
1012 | if (!mask) { | |
1013 | printf("Log items (comma separated):\n"); | |
1014 | for (item = cpu_log_items; item->mask != 0; item++) { | |
1015 | printf("%-10s %s\n", item->name, item->help); | |
1016 | } | |
1017 | exit(1); | |
1018 | } | |
1019 | cpu_set_log(mask); | |
1020 | } | |
29e922b6 BS |
1021 | |
1022 | /* Return the virtual CPU time, based on the instruction counter. */ | |
1023 | int64_t cpu_get_icount(void) | |
1024 | { | |
1025 | int64_t icount; | |
1026 | CPUState *env = cpu_single_env;; | |
1027 | ||
1028 | icount = qemu_icount; | |
1029 | if (env) { | |
1030 | if (!can_do_io(env)) { | |
1031 | fprintf(stderr, "Bad clock read\n"); | |
1032 | } | |
1033 | icount -= (env->icount_decr.u16.low + env->icount_extra); | |
1034 | } | |
1035 | return qemu_icount_bias + (icount << icount_time_shift); | |
1036 | } | |
262353cb | 1037 | |
9a78eead | 1038 | void list_cpus(FILE *f, fprintf_function cpu_fprintf, const char *optarg) |
262353cb BS |
1039 | { |
1040 | /* XXX: implement xxx_cpu_list for targets that still miss it */ | |
1041 | #if defined(cpu_list_id) | |
1042 | cpu_list_id(f, cpu_fprintf, optarg); | |
1043 | #elif defined(cpu_list) | |
1044 | cpu_list(f, cpu_fprintf); /* deprecated */ | |
1045 | #endif | |
1046 | } |