]>
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 | |
d0f294ce JK |
230 | /* If we have signalfd, we mask out the signals we want to handle and then |
231 | * use signalfd to listen for them. We rely on whatever the current signal | |
232 | * handler is to dispatch the signals when we receive them. | |
233 | */ | |
234 | static void sigfd_handler(void *opaque) | |
235 | { | |
236 | int fd = (unsigned long) opaque; | |
237 | struct qemu_signalfd_siginfo info; | |
238 | struct sigaction action; | |
239 | ssize_t len; | |
240 | ||
241 | while (1) { | |
242 | do { | |
243 | len = read(fd, &info, sizeof(info)); | |
244 | } while (len == -1 && errno == EINTR); | |
245 | ||
246 | if (len == -1 && errno == EAGAIN) { | |
247 | break; | |
248 | } | |
249 | ||
250 | if (len != sizeof(info)) { | |
251 | printf("read from sigfd returned %zd: %m\n", len); | |
252 | return; | |
253 | } | |
254 | ||
255 | sigaction(info.ssi_signo, NULL, &action); | |
256 | if ((action.sa_flags & SA_SIGINFO) && action.sa_sigaction) { | |
257 | action.sa_sigaction(info.ssi_signo, | |
258 | (siginfo_t *)&info, NULL); | |
259 | } else if (action.sa_handler) { | |
260 | action.sa_handler(info.ssi_signo); | |
261 | } | |
262 | } | |
263 | } | |
264 | ||
265 | static int qemu_signalfd_init(sigset_t mask) | |
266 | { | |
267 | int sigfd; | |
268 | ||
269 | sigfd = qemu_signalfd(&mask); | |
270 | if (sigfd == -1) { | |
271 | fprintf(stderr, "failed to create signalfd\n"); | |
272 | return -errno; | |
273 | } | |
274 | ||
275 | fcntl_setfl(sigfd, O_NONBLOCK); | |
276 | ||
277 | qemu_set_fd_handler2(sigfd, NULL, sigfd_handler, NULL, | |
278 | (void *)(unsigned long) sigfd); | |
279 | ||
280 | return 0; | |
281 | } | |
282 | ||
9a36085b JK |
283 | static void sigbus_reraise(void); |
284 | ||
285 | static void qemu_kvm_eat_signals(CPUState *env) | |
286 | { | |
287 | struct timespec ts = { 0, 0 }; | |
288 | siginfo_t siginfo; | |
289 | sigset_t waitset; | |
290 | sigset_t chkset; | |
291 | int r; | |
292 | ||
293 | sigemptyset(&waitset); | |
294 | sigaddset(&waitset, SIG_IPI); | |
295 | sigaddset(&waitset, SIGBUS); | |
296 | ||
297 | do { | |
298 | r = sigtimedwait(&waitset, &siginfo, &ts); | |
299 | if (r == -1 && !(errno == EAGAIN || errno == EINTR)) { | |
300 | perror("sigtimedwait"); | |
301 | exit(1); | |
302 | } | |
303 | ||
304 | switch (r) { | |
305 | #ifdef CONFIG_IOTHREAD | |
306 | case SIGBUS: | |
307 | if (kvm_on_sigbus_vcpu(env, siginfo.si_code, siginfo.si_addr)) { | |
308 | sigbus_reraise(); | |
309 | } | |
310 | break; | |
311 | #endif | |
312 | default: | |
313 | break; | |
314 | } | |
315 | ||
316 | r = sigpending(&chkset); | |
317 | if (r == -1) { | |
318 | perror("sigpending"); | |
319 | exit(1); | |
320 | } | |
321 | } while (sigismember(&chkset, SIG_IPI) || sigismember(&chkset, SIGBUS)); | |
322 | } | |
323 | ||
55f8d6ac JK |
324 | #else /* _WIN32 */ |
325 | ||
296af7c9 BS |
326 | HANDLE qemu_event_handle; |
327 | ||
328 | static void dummy_event_handler(void *opaque) | |
329 | { | |
330 | } | |
331 | ||
332 | static int qemu_event_init(void) | |
333 | { | |
334 | qemu_event_handle = CreateEvent(NULL, FALSE, FALSE, NULL); | |
335 | if (!qemu_event_handle) { | |
336 | fprintf(stderr, "Failed CreateEvent: %ld\n", GetLastError()); | |
337 | return -1; | |
338 | } | |
339 | qemu_add_wait_object(qemu_event_handle, dummy_event_handler, NULL); | |
340 | return 0; | |
341 | } | |
342 | ||
343 | static void qemu_event_increment(void) | |
344 | { | |
345 | if (!SetEvent(qemu_event_handle)) { | |
346 | fprintf(stderr, "qemu_event_increment: SetEvent failed: %ld\n", | |
347 | GetLastError()); | |
348 | exit (1); | |
349 | } | |
350 | } | |
9a36085b JK |
351 | |
352 | static void qemu_kvm_eat_signals(CPUState *env) | |
353 | { | |
354 | } | |
55f8d6ac | 355 | #endif /* _WIN32 */ |
296af7c9 BS |
356 | |
357 | #ifndef CONFIG_IOTHREAD | |
ff48eb5f JK |
358 | static void qemu_kvm_init_cpu_signals(CPUState *env) |
359 | { | |
360 | #ifndef _WIN32 | |
361 | int r; | |
362 | sigset_t set; | |
363 | struct sigaction sigact; | |
364 | ||
365 | memset(&sigact, 0, sizeof(sigact)); | |
366 | sigact.sa_handler = dummy_signal; | |
367 | sigaction(SIG_IPI, &sigact, NULL); | |
368 | ||
369 | sigemptyset(&set); | |
370 | sigaddset(&set, SIG_IPI); | |
371 | pthread_sigmask(SIG_BLOCK, &set, NULL); | |
372 | ||
373 | pthread_sigmask(SIG_BLOCK, NULL, &set); | |
374 | sigdelset(&set, SIG_IPI); | |
375 | sigdelset(&set, SIGBUS); | |
376 | r = kvm_set_signal_mask(env, &set); | |
377 | if (r) { | |
378 | fprintf(stderr, "kvm_set_signal_mask: %s\n", strerror(-r)); | |
379 | exit(1); | |
380 | } | |
381 | #endif | |
382 | } | |
383 | ||
296af7c9 BS |
384 | int qemu_init_main_loop(void) |
385 | { | |
d0f294ce JK |
386 | #ifndef _WIN32 |
387 | sigset_t blocked_signals; | |
388 | int ret; | |
389 | ||
390 | sigemptyset(&blocked_signals); | |
391 | ||
392 | ret = qemu_signalfd_init(blocked_signals); | |
393 | if (ret) { | |
394 | return ret; | |
395 | } | |
396 | #endif | |
3c638d06 JK |
397 | cpu_set_debug_excp_handler(cpu_debug_handler); |
398 | ||
296af7c9 BS |
399 | return qemu_event_init(); |
400 | } | |
401 | ||
7277e027 BS |
402 | void qemu_main_loop_start(void) |
403 | { | |
404 | } | |
405 | ||
296af7c9 BS |
406 | void qemu_init_vcpu(void *_env) |
407 | { | |
408 | CPUState *env = _env; | |
84b4915d | 409 | int r; |
296af7c9 BS |
410 | |
411 | env->nr_cores = smp_cores; | |
412 | env->nr_threads = smp_threads; | |
84b4915d JK |
413 | |
414 | if (kvm_enabled()) { | |
415 | r = kvm_init_vcpu(env); | |
416 | if (r < 0) { | |
417 | fprintf(stderr, "kvm_init_vcpu failed: %s\n", strerror(-r)); | |
418 | exit(1); | |
419 | } | |
ff48eb5f | 420 | qemu_kvm_init_cpu_signals(env); |
84b4915d | 421 | } |
296af7c9 BS |
422 | } |
423 | ||
424 | int qemu_cpu_self(void *env) | |
425 | { | |
426 | return 1; | |
427 | } | |
428 | ||
e82bcec2 MT |
429 | void run_on_cpu(CPUState *env, void (*func)(void *data), void *data) |
430 | { | |
431 | func(data); | |
432 | } | |
433 | ||
296af7c9 BS |
434 | void resume_all_vcpus(void) |
435 | { | |
436 | } | |
437 | ||
438 | void pause_all_vcpus(void) | |
439 | { | |
440 | } | |
441 | ||
442 | void qemu_cpu_kick(void *env) | |
443 | { | |
444 | return; | |
445 | } | |
446 | ||
447 | void qemu_notify_event(void) | |
448 | { | |
449 | CPUState *env = cpu_single_env; | |
450 | ||
451 | qemu_event_increment (); | |
452 | if (env) { | |
453 | cpu_exit(env); | |
454 | } | |
455 | if (next_cpu && env != next_cpu) { | |
456 | cpu_exit(next_cpu); | |
457 | } | |
38145df2 | 458 | exit_request = 1; |
296af7c9 BS |
459 | } |
460 | ||
461 | void qemu_mutex_lock_iothread(void) {} | |
462 | void qemu_mutex_unlock_iothread(void) {} | |
463 | ||
b4a3d965 JK |
464 | void cpu_stop_current(void) |
465 | { | |
466 | } | |
467 | ||
296af7c9 BS |
468 | void vm_stop(int reason) |
469 | { | |
470 | do_vm_stop(reason); | |
471 | } | |
472 | ||
473 | #else /* CONFIG_IOTHREAD */ | |
474 | ||
475 | #include "qemu-thread.h" | |
476 | ||
477 | QemuMutex qemu_global_mutex; | |
478 | static QemuMutex qemu_fair_mutex; | |
479 | ||
480 | static QemuThread io_thread; | |
481 | ||
482 | static QemuThread *tcg_cpu_thread; | |
483 | static QemuCond *tcg_halt_cond; | |
484 | ||
485 | static int qemu_system_ready; | |
486 | /* cpu creation */ | |
487 | static QemuCond qemu_cpu_cond; | |
488 | /* system init */ | |
489 | static QemuCond qemu_system_cond; | |
490 | static QemuCond qemu_pause_cond; | |
e82bcec2 | 491 | static QemuCond qemu_work_cond; |
296af7c9 | 492 | |
55f8d6ac JK |
493 | static void cpu_signal(int sig) |
494 | { | |
495 | if (cpu_single_env) { | |
496 | cpu_exit(cpu_single_env); | |
497 | } | |
498 | exit_request = 1; | |
499 | } | |
500 | ||
501 | static void qemu_kvm_init_cpu_signals(CPUState *env) | |
502 | { | |
503 | int r; | |
504 | sigset_t set; | |
505 | struct sigaction sigact; | |
506 | ||
507 | memset(&sigact, 0, sizeof(sigact)); | |
508 | sigact.sa_handler = dummy_signal; | |
509 | sigaction(SIG_IPI, &sigact, NULL); | |
510 | ||
511 | pthread_sigmask(SIG_BLOCK, NULL, &set); | |
512 | sigdelset(&set, SIG_IPI); | |
513 | sigdelset(&set, SIGBUS); | |
514 | r = kvm_set_signal_mask(env, &set); | |
515 | if (r) { | |
516 | fprintf(stderr, "kvm_set_signal_mask: %s\n", strerror(-r)); | |
517 | exit(1); | |
518 | } | |
519 | } | |
520 | ||
521 | static void qemu_tcg_init_cpu_signals(void) | |
522 | { | |
523 | sigset_t set; | |
524 | struct sigaction sigact; | |
525 | ||
526 | memset(&sigact, 0, sizeof(sigact)); | |
527 | sigact.sa_handler = cpu_signal; | |
528 | sigaction(SIG_IPI, &sigact, NULL); | |
529 | ||
530 | sigemptyset(&set); | |
531 | sigaddset(&set, SIG_IPI); | |
532 | pthread_sigmask(SIG_UNBLOCK, &set, NULL); | |
533 | } | |
534 | ||
535 | static void sigbus_handler(int n, struct qemu_signalfd_siginfo *siginfo, | |
536 | void *ctx); | |
537 | ||
538 | static sigset_t block_io_signals(void) | |
539 | { | |
540 | sigset_t set; | |
541 | struct sigaction action; | |
542 | ||
543 | /* SIGUSR2 used by posix-aio-compat.c */ | |
544 | sigemptyset(&set); | |
545 | sigaddset(&set, SIGUSR2); | |
546 | pthread_sigmask(SIG_UNBLOCK, &set, NULL); | |
547 | ||
548 | sigemptyset(&set); | |
549 | sigaddset(&set, SIGIO); | |
550 | sigaddset(&set, SIGALRM); | |
551 | sigaddset(&set, SIG_IPI); | |
552 | sigaddset(&set, SIGBUS); | |
553 | pthread_sigmask(SIG_BLOCK, &set, NULL); | |
554 | ||
555 | memset(&action, 0, sizeof(action)); | |
556 | action.sa_flags = SA_SIGINFO; | |
557 | action.sa_sigaction = (void (*)(int, siginfo_t*, void*))sigbus_handler; | |
558 | sigaction(SIGBUS, &action, NULL); | |
559 | prctl(PR_MCE_KILL, 1, 1, 0, 0); | |
560 | ||
561 | return set; | |
562 | } | |
563 | ||
296af7c9 BS |
564 | int qemu_init_main_loop(void) |
565 | { | |
566 | int ret; | |
a8486bc9 | 567 | sigset_t blocked_signals; |
296af7c9 | 568 | |
3c638d06 JK |
569 | cpu_set_debug_excp_handler(cpu_debug_handler); |
570 | ||
a8486bc9 MT |
571 | blocked_signals = block_io_signals(); |
572 | ||
573 | ret = qemu_signalfd_init(blocked_signals); | |
574 | if (ret) | |
575 | return ret; | |
576 | ||
577 | /* Note eventfd must be drained before signalfd handlers run */ | |
296af7c9 BS |
578 | ret = qemu_event_init(); |
579 | if (ret) | |
580 | return ret; | |
581 | ||
582 | qemu_cond_init(&qemu_pause_cond); | |
f8ca7b43 | 583 | qemu_cond_init(&qemu_system_cond); |
296af7c9 BS |
584 | qemu_mutex_init(&qemu_fair_mutex); |
585 | qemu_mutex_init(&qemu_global_mutex); | |
586 | qemu_mutex_lock(&qemu_global_mutex); | |
587 | ||
296af7c9 BS |
588 | qemu_thread_self(&io_thread); |
589 | ||
590 | return 0; | |
591 | } | |
592 | ||
7277e027 BS |
593 | void qemu_main_loop_start(void) |
594 | { | |
595 | qemu_system_ready = 1; | |
596 | qemu_cond_broadcast(&qemu_system_cond); | |
597 | } | |
598 | ||
e82bcec2 MT |
599 | void run_on_cpu(CPUState *env, void (*func)(void *data), void *data) |
600 | { | |
601 | struct qemu_work_item wi; | |
602 | ||
603 | if (qemu_cpu_self(env)) { | |
604 | func(data); | |
605 | return; | |
606 | } | |
607 | ||
608 | wi.func = func; | |
609 | wi.data = data; | |
610 | if (!env->queued_work_first) | |
611 | env->queued_work_first = &wi; | |
612 | else | |
613 | env->queued_work_last->next = &wi; | |
614 | env->queued_work_last = &wi; | |
615 | wi.next = NULL; | |
616 | wi.done = false; | |
617 | ||
618 | qemu_cpu_kick(env); | |
619 | while (!wi.done) { | |
620 | CPUState *self_env = cpu_single_env; | |
621 | ||
622 | qemu_cond_wait(&qemu_work_cond, &qemu_global_mutex); | |
623 | cpu_single_env = self_env; | |
624 | } | |
625 | } | |
626 | ||
627 | static void flush_queued_work(CPUState *env) | |
628 | { | |
629 | struct qemu_work_item *wi; | |
630 | ||
631 | if (!env->queued_work_first) | |
632 | return; | |
633 | ||
634 | while ((wi = env->queued_work_first)) { | |
635 | env->queued_work_first = wi->next; | |
636 | wi->func(wi->data); | |
637 | wi->done = true; | |
638 | } | |
639 | env->queued_work_last = NULL; | |
640 | qemu_cond_broadcast(&qemu_work_cond); | |
641 | } | |
642 | ||
296af7c9 BS |
643 | static void qemu_wait_io_event_common(CPUState *env) |
644 | { | |
645 | if (env->stop) { | |
646 | env->stop = 0; | |
647 | env->stopped = 1; | |
648 | qemu_cond_signal(&qemu_pause_cond); | |
649 | } | |
e82bcec2 | 650 | flush_queued_work(env); |
aa2c364b | 651 | env->thread_kicked = false; |
296af7c9 BS |
652 | } |
653 | ||
6cabe1f3 | 654 | static void qemu_tcg_wait_io_event(void) |
296af7c9 | 655 | { |
6cabe1f3 JK |
656 | CPUState *env; |
657 | ||
472fb0c4 | 658 | while (!any_cpu_has_work()) |
6cabe1f3 | 659 | qemu_cond_timedwait(tcg_halt_cond, &qemu_global_mutex, 1000); |
296af7c9 BS |
660 | |
661 | qemu_mutex_unlock(&qemu_global_mutex); | |
662 | ||
663 | /* | |
664 | * Users of qemu_global_mutex can be starved, having no chance | |
665 | * to acquire it since this path will get to it first. | |
666 | * So use another lock to provide fairness. | |
667 | */ | |
668 | qemu_mutex_lock(&qemu_fair_mutex); | |
669 | qemu_mutex_unlock(&qemu_fair_mutex); | |
670 | ||
671 | qemu_mutex_lock(&qemu_global_mutex); | |
6cabe1f3 JK |
672 | |
673 | for (env = first_cpu; env != NULL; env = env->next_cpu) { | |
674 | qemu_wait_io_event_common(env); | |
675 | } | |
296af7c9 BS |
676 | } |
677 | ||
c0532a76 MT |
678 | static void sigbus_reraise(void) |
679 | { | |
680 | sigset_t set; | |
681 | struct sigaction action; | |
682 | ||
683 | memset(&action, 0, sizeof(action)); | |
684 | action.sa_handler = SIG_DFL; | |
685 | if (!sigaction(SIGBUS, &action, NULL)) { | |
686 | raise(SIGBUS); | |
687 | sigemptyset(&set); | |
688 | sigaddset(&set, SIGBUS); | |
689 | sigprocmask(SIG_UNBLOCK, &set, NULL); | |
690 | } | |
691 | perror("Failed to re-raise SIGBUS!\n"); | |
692 | abort(); | |
693 | } | |
694 | ||
695 | static void sigbus_handler(int n, struct qemu_signalfd_siginfo *siginfo, | |
696 | void *ctx) | |
697 | { | |
a1b87fe0 | 698 | if (kvm_on_sigbus(siginfo->ssi_code, (void *)(intptr_t)siginfo->ssi_addr)) { |
c0532a76 | 699 | sigbus_reraise(); |
a1b87fe0 | 700 | } |
c0532a76 MT |
701 | } |
702 | ||
296af7c9 BS |
703 | static void qemu_kvm_wait_io_event(CPUState *env) |
704 | { | |
705 | while (!cpu_has_work(env)) | |
706 | qemu_cond_timedwait(env->halt_cond, &qemu_global_mutex, 1000); | |
707 | ||
5db5bdac | 708 | qemu_kvm_eat_signals(env); |
296af7c9 BS |
709 | qemu_wait_io_event_common(env); |
710 | } | |
711 | ||
712 | static int qemu_cpu_exec(CPUState *env); | |
713 | ||
714 | static void *kvm_cpu_thread_fn(void *arg) | |
715 | { | |
716 | CPUState *env = arg; | |
84b4915d | 717 | int r; |
296af7c9 | 718 | |
6164e6d6 | 719 | qemu_mutex_lock(&qemu_global_mutex); |
296af7c9 | 720 | qemu_thread_self(env->thread); |
d31ae052 | 721 | |
84b4915d JK |
722 | r = kvm_init_vcpu(env); |
723 | if (r < 0) { | |
724 | fprintf(stderr, "kvm_init_vcpu failed: %s\n", strerror(-r)); | |
725 | exit(1); | |
726 | } | |
296af7c9 | 727 | |
55f8d6ac | 728 | qemu_kvm_init_cpu_signals(env); |
296af7c9 BS |
729 | |
730 | /* signal CPU creation */ | |
296af7c9 BS |
731 | env->created = 1; |
732 | qemu_cond_signal(&qemu_cpu_cond); | |
733 | ||
734 | /* and wait for machine initialization */ | |
735 | while (!qemu_system_ready) | |
736 | qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100); | |
737 | ||
738 | while (1) { | |
739 | if (cpu_can_run(env)) | |
740 | qemu_cpu_exec(env); | |
741 | qemu_kvm_wait_io_event(env); | |
742 | } | |
743 | ||
744 | return NULL; | |
745 | } | |
746 | ||
747 | static void *tcg_cpu_thread_fn(void *arg) | |
748 | { | |
749 | CPUState *env = arg; | |
750 | ||
55f8d6ac | 751 | qemu_tcg_init_cpu_signals(); |
296af7c9 BS |
752 | qemu_thread_self(env->thread); |
753 | ||
754 | /* signal CPU creation */ | |
755 | qemu_mutex_lock(&qemu_global_mutex); | |
756 | for (env = first_cpu; env != NULL; env = env->next_cpu) | |
757 | env->created = 1; | |
758 | qemu_cond_signal(&qemu_cpu_cond); | |
759 | ||
760 | /* and wait for machine initialization */ | |
761 | while (!qemu_system_ready) | |
762 | qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100); | |
763 | ||
764 | while (1) { | |
472fb0c4 | 765 | cpu_exec_all(); |
6cabe1f3 | 766 | qemu_tcg_wait_io_event(); |
296af7c9 BS |
767 | } |
768 | ||
769 | return NULL; | |
770 | } | |
771 | ||
772 | void qemu_cpu_kick(void *_env) | |
773 | { | |
774 | CPUState *env = _env; | |
775 | qemu_cond_broadcast(env->halt_cond); | |
aa2c364b JK |
776 | if (!env->thread_kicked) { |
777 | qemu_thread_signal(env->thread, SIG_IPI); | |
778 | env->thread_kicked = true; | |
779 | } | |
296af7c9 BS |
780 | } |
781 | ||
782 | int qemu_cpu_self(void *_env) | |
783 | { | |
784 | CPUState *env = _env; | |
785 | QemuThread this; | |
786 | ||
787 | qemu_thread_self(&this); | |
788 | ||
789 | return qemu_thread_equal(&this, env->thread); | |
790 | } | |
791 | ||
296af7c9 BS |
792 | void qemu_mutex_lock_iothread(void) |
793 | { | |
794 | if (kvm_enabled()) { | |
296af7c9 | 795 | qemu_mutex_lock(&qemu_global_mutex); |
1a28cac3 MT |
796 | } else { |
797 | qemu_mutex_lock(&qemu_fair_mutex); | |
798 | if (qemu_mutex_trylock(&qemu_global_mutex)) { | |
799 | qemu_thread_signal(tcg_cpu_thread, SIG_IPI); | |
800 | qemu_mutex_lock(&qemu_global_mutex); | |
801 | } | |
802 | qemu_mutex_unlock(&qemu_fair_mutex); | |
803 | } | |
296af7c9 BS |
804 | } |
805 | ||
806 | void qemu_mutex_unlock_iothread(void) | |
807 | { | |
808 | qemu_mutex_unlock(&qemu_global_mutex); | |
809 | } | |
810 | ||
811 | static int all_vcpus_paused(void) | |
812 | { | |
813 | CPUState *penv = first_cpu; | |
814 | ||
815 | while (penv) { | |
816 | if (!penv->stopped) | |
817 | return 0; | |
818 | penv = (CPUState *)penv->next_cpu; | |
819 | } | |
820 | ||
821 | return 1; | |
822 | } | |
823 | ||
824 | void pause_all_vcpus(void) | |
825 | { | |
826 | CPUState *penv = first_cpu; | |
827 | ||
828 | while (penv) { | |
829 | penv->stop = 1; | |
296af7c9 BS |
830 | qemu_cpu_kick(penv); |
831 | penv = (CPUState *)penv->next_cpu; | |
832 | } | |
833 | ||
834 | while (!all_vcpus_paused()) { | |
835 | qemu_cond_timedwait(&qemu_pause_cond, &qemu_global_mutex, 100); | |
836 | penv = first_cpu; | |
837 | while (penv) { | |
1fbb22e5 | 838 | qemu_cpu_kick(penv); |
296af7c9 BS |
839 | penv = (CPUState *)penv->next_cpu; |
840 | } | |
841 | } | |
842 | } | |
843 | ||
844 | void resume_all_vcpus(void) | |
845 | { | |
846 | CPUState *penv = first_cpu; | |
847 | ||
848 | while (penv) { | |
849 | penv->stop = 0; | |
850 | penv->stopped = 0; | |
296af7c9 BS |
851 | qemu_cpu_kick(penv); |
852 | penv = (CPUState *)penv->next_cpu; | |
853 | } | |
854 | } | |
855 | ||
856 | static void tcg_init_vcpu(void *_env) | |
857 | { | |
858 | CPUState *env = _env; | |
859 | /* share a single thread for all cpus with TCG */ | |
860 | if (!tcg_cpu_thread) { | |
861 | env->thread = qemu_mallocz(sizeof(QemuThread)); | |
862 | env->halt_cond = qemu_mallocz(sizeof(QemuCond)); | |
863 | qemu_cond_init(env->halt_cond); | |
864 | qemu_thread_create(env->thread, tcg_cpu_thread_fn, env); | |
865 | while (env->created == 0) | |
866 | qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100); | |
867 | tcg_cpu_thread = env->thread; | |
868 | tcg_halt_cond = env->halt_cond; | |
869 | } else { | |
870 | env->thread = tcg_cpu_thread; | |
871 | env->halt_cond = tcg_halt_cond; | |
872 | } | |
873 | } | |
874 | ||
875 | static void kvm_start_vcpu(CPUState *env) | |
876 | { | |
877 | env->thread = qemu_mallocz(sizeof(QemuThread)); | |
878 | env->halt_cond = qemu_mallocz(sizeof(QemuCond)); | |
879 | qemu_cond_init(env->halt_cond); | |
880 | qemu_thread_create(env->thread, kvm_cpu_thread_fn, env); | |
881 | while (env->created == 0) | |
882 | qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100); | |
883 | } | |
884 | ||
885 | void qemu_init_vcpu(void *_env) | |
886 | { | |
887 | CPUState *env = _env; | |
888 | ||
889 | env->nr_cores = smp_cores; | |
890 | env->nr_threads = smp_threads; | |
891 | if (kvm_enabled()) | |
892 | kvm_start_vcpu(env); | |
893 | else | |
894 | tcg_init_vcpu(env); | |
895 | } | |
896 | ||
897 | void qemu_notify_event(void) | |
898 | { | |
899 | qemu_event_increment(); | |
900 | } | |
901 | ||
902 | static void qemu_system_vmstop_request(int reason) | |
903 | { | |
904 | vmstop_requested = reason; | |
905 | qemu_notify_event(); | |
906 | } | |
907 | ||
b4a3d965 JK |
908 | void cpu_stop_current(void) |
909 | { | |
910 | if (cpu_single_env) { | |
911 | cpu_single_env->stopped = 1; | |
912 | cpu_exit(cpu_single_env); | |
913 | } | |
914 | } | |
915 | ||
296af7c9 BS |
916 | void vm_stop(int reason) |
917 | { | |
918 | QemuThread me; | |
919 | qemu_thread_self(&me); | |
920 | ||
921 | if (!qemu_thread_equal(&me, &io_thread)) { | |
922 | qemu_system_vmstop_request(reason); | |
923 | /* | |
924 | * FIXME: should not return to device code in case | |
925 | * vm_stop() has been requested. | |
926 | */ | |
b4a3d965 | 927 | cpu_stop_current(); |
296af7c9 BS |
928 | return; |
929 | } | |
930 | do_vm_stop(reason); | |
931 | } | |
932 | ||
933 | #endif | |
934 | ||
935 | static int qemu_cpu_exec(CPUState *env) | |
936 | { | |
937 | int ret; | |
938 | #ifdef CONFIG_PROFILER | |
939 | int64_t ti; | |
940 | #endif | |
941 | ||
942 | #ifdef CONFIG_PROFILER | |
943 | ti = profile_getclock(); | |
944 | #endif | |
945 | if (use_icount) { | |
946 | int64_t count; | |
947 | int decr; | |
948 | qemu_icount -= (env->icount_decr.u16.low + env->icount_extra); | |
949 | env->icount_decr.u16.low = 0; | |
950 | env->icount_extra = 0; | |
951 | count = qemu_icount_round (qemu_next_deadline()); | |
952 | qemu_icount += count; | |
953 | decr = (count > 0xffff) ? 0xffff : count; | |
954 | count -= decr; | |
955 | env->icount_decr.u16.low = decr; | |
956 | env->icount_extra = count; | |
957 | } | |
958 | ret = cpu_exec(env); | |
959 | #ifdef CONFIG_PROFILER | |
960 | qemu_time += profile_getclock() - ti; | |
961 | #endif | |
962 | if (use_icount) { | |
963 | /* Fold pending instructions back into the | |
964 | instruction counter, and clear the interrupt flag. */ | |
965 | qemu_icount -= (env->icount_decr.u16.low | |
966 | + env->icount_extra); | |
967 | env->icount_decr.u32 = 0; | |
968 | env->icount_extra = 0; | |
969 | } | |
970 | return ret; | |
971 | } | |
972 | ||
472fb0c4 | 973 | bool cpu_exec_all(void) |
296af7c9 | 974 | { |
9a36085b JK |
975 | int r; |
976 | ||
296af7c9 BS |
977 | if (next_cpu == NULL) |
978 | next_cpu = first_cpu; | |
c629a4bc | 979 | for (; next_cpu != NULL && !exit_request; next_cpu = next_cpu->next_cpu) { |
345f4426 | 980 | CPUState *env = next_cpu; |
296af7c9 BS |
981 | |
982 | qemu_clock_enable(vm_clock, | |
345f4426 | 983 | (env->singlestep_enabled & SSTEP_NOTIMER) == 0); |
296af7c9 BS |
984 | |
985 | if (qemu_alarm_pending()) | |
986 | break; | |
3c638d06 | 987 | if (cpu_can_run(env)) { |
9a36085b JK |
988 | r = qemu_cpu_exec(env); |
989 | if (kvm_enabled()) { | |
990 | qemu_kvm_eat_signals(env); | |
991 | } | |
992 | if (r == EXCP_DEBUG) { | |
3c638d06 JK |
993 | break; |
994 | } | |
995 | } else if (env->stop) { | |
296af7c9 BS |
996 | break; |
997 | } | |
998 | } | |
c629a4bc | 999 | exit_request = 0; |
472fb0c4 | 1000 | return any_cpu_has_work(); |
296af7c9 BS |
1001 | } |
1002 | ||
1003 | void set_numa_modes(void) | |
1004 | { | |
1005 | CPUState *env; | |
1006 | int i; | |
1007 | ||
1008 | for (env = first_cpu; env != NULL; env = env->next_cpu) { | |
1009 | for (i = 0; i < nb_numa_nodes; i++) { | |
1010 | if (node_cpumask[i] & (1 << env->cpu_index)) { | |
1011 | env->numa_node = i; | |
1012 | } | |
1013 | } | |
1014 | } | |
1015 | } | |
1016 | ||
1017 | void set_cpu_log(const char *optarg) | |
1018 | { | |
1019 | int mask; | |
1020 | const CPULogItem *item; | |
1021 | ||
1022 | mask = cpu_str_to_log_mask(optarg); | |
1023 | if (!mask) { | |
1024 | printf("Log items (comma separated):\n"); | |
1025 | for (item = cpu_log_items; item->mask != 0; item++) { | |
1026 | printf("%-10s %s\n", item->name, item->help); | |
1027 | } | |
1028 | exit(1); | |
1029 | } | |
1030 | cpu_set_log(mask); | |
1031 | } | |
29e922b6 BS |
1032 | |
1033 | /* Return the virtual CPU time, based on the instruction counter. */ | |
1034 | int64_t cpu_get_icount(void) | |
1035 | { | |
1036 | int64_t icount; | |
1037 | CPUState *env = cpu_single_env;; | |
1038 | ||
1039 | icount = qemu_icount; | |
1040 | if (env) { | |
1041 | if (!can_do_io(env)) { | |
1042 | fprintf(stderr, "Bad clock read\n"); | |
1043 | } | |
1044 | icount -= (env->icount_decr.u16.low + env->icount_extra); | |
1045 | } | |
1046 | return qemu_icount_bias + (icount << icount_time_shift); | |
1047 | } | |
262353cb | 1048 | |
9a78eead | 1049 | void list_cpus(FILE *f, fprintf_function cpu_fprintf, const char *optarg) |
262353cb BS |
1050 | { |
1051 | /* XXX: implement xxx_cpu_list for targets that still miss it */ | |
1052 | #if defined(cpu_list_id) | |
1053 | cpu_list_id(f, cpu_fprintf, optarg); | |
1054 | #elif defined(cpu_list) | |
1055 | cpu_list(f, cpu_fprintf); /* deprecated */ | |
1056 | #endif | |
1057 | } |