overall/alpha tcg cpus|hppa: Fix Lesser GPL version number
[qemu.git] / softmmu / cpus.c
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 #include "qemu/osdep.h"
26 #include "qemu-common.h"
27 #include "monitor/monitor.h"
28 #include "qapi/error.h"
29 #include "qapi/qapi-commands-machine.h"
30 #include "qapi/qapi-commands-misc.h"
31 #include "qapi/qapi-events-run-state.h"
32 #include "qapi/qmp/qerror.h"
33 #include "exec/gdbstub.h"
34 #include "sysemu/hw_accel.h"
35 #include "exec/exec-all.h"
36 #include "qemu/thread.h"
37 #include "qemu/plugin.h"
38 #include "sysemu/cpus.h"
39 #include "qemu/guest-random.h"
40 #include "hw/nmi.h"
41 #include "sysemu/replay.h"
42 #include "sysemu/runstate.h"
43 #include "sysemu/cpu-timers.h"
44 #include "hw/boards.h"
45 #include "hw/hw.h"
46
47 #ifdef CONFIG_LINUX
48
49 #include <sys/prctl.h>
50
51 #ifndef PR_MCE_KILL
52 #define PR_MCE_KILL 33
53 #endif
54
55 #ifndef PR_MCE_KILL_SET
56 #define PR_MCE_KILL_SET 1
57 #endif
58
59 #ifndef PR_MCE_KILL_EARLY
60 #define PR_MCE_KILL_EARLY 1
61 #endif
62
63 #endif /* CONFIG_LINUX */
64
65 static QemuMutex qemu_global_mutex;
66
67 bool cpu_is_stopped(CPUState *cpu)
68 {
69 return cpu->stopped || !runstate_is_running();
70 }
71
72 bool cpu_work_list_empty(CPUState *cpu)
73 {
74 bool ret;
75
76 qemu_mutex_lock(&cpu->work_mutex);
77 ret = QSIMPLEQ_EMPTY(&cpu->work_list);
78 qemu_mutex_unlock(&cpu->work_mutex);
79 return ret;
80 }
81
82 bool cpu_thread_is_idle(CPUState *cpu)
83 {
84 if (cpu->stop || !cpu_work_list_empty(cpu)) {
85 return false;
86 }
87 if (cpu_is_stopped(cpu)) {
88 return true;
89 }
90 if (!cpu->halted || cpu_has_work(cpu) ||
91 kvm_halt_in_kernel()) {
92 return false;
93 }
94 return true;
95 }
96
97 bool all_cpu_threads_idle(void)
98 {
99 CPUState *cpu;
100
101 CPU_FOREACH(cpu) {
102 if (!cpu_thread_is_idle(cpu)) {
103 return false;
104 }
105 }
106 return true;
107 }
108
109 /***********************************************************/
110 void hw_error(const char *fmt, ...)
111 {
112 va_list ap;
113 CPUState *cpu;
114
115 va_start(ap, fmt);
116 fprintf(stderr, "qemu: hardware error: ");
117 vfprintf(stderr, fmt, ap);
118 fprintf(stderr, "\n");
119 CPU_FOREACH(cpu) {
120 fprintf(stderr, "CPU #%d:\n", cpu->cpu_index);
121 cpu_dump_state(cpu, stderr, CPU_DUMP_FPU);
122 }
123 va_end(ap);
124 abort();
125 }
126
127 /*
128 * The chosen accelerator is supposed to register this.
129 */
130 static const CpusAccel *cpus_accel;
131
132 void cpu_synchronize_all_states(void)
133 {
134 CPUState *cpu;
135
136 CPU_FOREACH(cpu) {
137 cpu_synchronize_state(cpu);
138 }
139 }
140
141 void cpu_synchronize_all_post_reset(void)
142 {
143 CPUState *cpu;
144
145 CPU_FOREACH(cpu) {
146 cpu_synchronize_post_reset(cpu);
147 }
148 }
149
150 void cpu_synchronize_all_post_init(void)
151 {
152 CPUState *cpu;
153
154 CPU_FOREACH(cpu) {
155 cpu_synchronize_post_init(cpu);
156 }
157 }
158
159 void cpu_synchronize_all_pre_loadvm(void)
160 {
161 CPUState *cpu;
162
163 CPU_FOREACH(cpu) {
164 cpu_synchronize_pre_loadvm(cpu);
165 }
166 }
167
168 void cpu_synchronize_state(CPUState *cpu)
169 {
170 if (cpus_accel->synchronize_state) {
171 cpus_accel->synchronize_state(cpu);
172 }
173 }
174
175 void cpu_synchronize_post_reset(CPUState *cpu)
176 {
177 if (cpus_accel->synchronize_post_reset) {
178 cpus_accel->synchronize_post_reset(cpu);
179 }
180 }
181
182 void cpu_synchronize_post_init(CPUState *cpu)
183 {
184 if (cpus_accel->synchronize_post_init) {
185 cpus_accel->synchronize_post_init(cpu);
186 }
187 }
188
189 void cpu_synchronize_pre_loadvm(CPUState *cpu)
190 {
191 if (cpus_accel->synchronize_pre_loadvm) {
192 cpus_accel->synchronize_pre_loadvm(cpu);
193 }
194 }
195
196 int64_t cpus_get_virtual_clock(void)
197 {
198 /*
199 * XXX
200 *
201 * need to check that cpus_accel is not NULL, because qcow2 calls
202 * qemu_get_clock_ns(CLOCK_VIRTUAL) without any accel initialized and
203 * with ticks disabled in some io-tests:
204 * 030 040 041 060 099 120 127 140 156 161 172 181 191 192 195 203 229 249 256 267
205 *
206 * is this expected?
207 *
208 * XXX
209 */
210 if (cpus_accel && cpus_accel->get_virtual_clock) {
211 return cpus_accel->get_virtual_clock();
212 }
213 return cpu_get_clock();
214 }
215
216 /*
217 * return the time elapsed in VM between vm_start and vm_stop. Unless
218 * icount is active, cpus_get_elapsed_ticks() uses units of the host CPU cycle
219 * counter.
220 */
221 int64_t cpus_get_elapsed_ticks(void)
222 {
223 if (cpus_accel->get_elapsed_ticks) {
224 return cpus_accel->get_elapsed_ticks();
225 }
226 return cpu_get_ticks();
227 }
228
229 static void generic_handle_interrupt(CPUState *cpu, int mask)
230 {
231 cpu->interrupt_request |= mask;
232
233 if (!qemu_cpu_is_self(cpu)) {
234 qemu_cpu_kick(cpu);
235 }
236 }
237
238 void cpu_interrupt(CPUState *cpu, int mask)
239 {
240 if (cpus_accel->handle_interrupt) {
241 cpus_accel->handle_interrupt(cpu, mask);
242 } else {
243 generic_handle_interrupt(cpu, mask);
244 }
245 }
246
247 static int do_vm_stop(RunState state, bool send_stop)
248 {
249 int ret = 0;
250
251 if (runstate_is_running()) {
252 runstate_set(state);
253 cpu_disable_ticks();
254 pause_all_vcpus();
255 vm_state_notify(0, state);
256 if (send_stop) {
257 qapi_event_send_stop();
258 }
259 }
260
261 bdrv_drain_all();
262 ret = bdrv_flush_all();
263
264 return ret;
265 }
266
267 /* Special vm_stop() variant for terminating the process. Historically clients
268 * did not expect a QMP STOP event and so we need to retain compatibility.
269 */
270 int vm_shutdown(void)
271 {
272 return do_vm_stop(RUN_STATE_SHUTDOWN, false);
273 }
274
275 bool cpu_can_run(CPUState *cpu)
276 {
277 if (cpu->stop) {
278 return false;
279 }
280 if (cpu_is_stopped(cpu)) {
281 return false;
282 }
283 return true;
284 }
285
286 void cpu_handle_guest_debug(CPUState *cpu)
287 {
288 if (replay_running_debug()) {
289 if (!cpu->singlestep_enabled) {
290 /*
291 * Report about the breakpoint and
292 * make a single step to skip it
293 */
294 replay_breakpoint();
295 cpu_single_step(cpu, SSTEP_ENABLE);
296 } else {
297 cpu_single_step(cpu, 0);
298 }
299 } else {
300 gdb_set_stop_cpu(cpu);
301 qemu_system_debug_request();
302 cpu->stopped = true;
303 }
304 }
305
306 #ifdef CONFIG_LINUX
307 static void sigbus_reraise(void)
308 {
309 sigset_t set;
310 struct sigaction action;
311
312 memset(&action, 0, sizeof(action));
313 action.sa_handler = SIG_DFL;
314 if (!sigaction(SIGBUS, &action, NULL)) {
315 raise(SIGBUS);
316 sigemptyset(&set);
317 sigaddset(&set, SIGBUS);
318 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
319 }
320 perror("Failed to re-raise SIGBUS!\n");
321 abort();
322 }
323
324 static void sigbus_handler(int n, siginfo_t *siginfo, void *ctx)
325 {
326 if (siginfo->si_code != BUS_MCEERR_AO && siginfo->si_code != BUS_MCEERR_AR) {
327 sigbus_reraise();
328 }
329
330 if (current_cpu) {
331 /* Called asynchronously in VCPU thread. */
332 if (kvm_on_sigbus_vcpu(current_cpu, siginfo->si_code, siginfo->si_addr)) {
333 sigbus_reraise();
334 }
335 } else {
336 /* Called synchronously (via signalfd) in main thread. */
337 if (kvm_on_sigbus(siginfo->si_code, siginfo->si_addr)) {
338 sigbus_reraise();
339 }
340 }
341 }
342
343 static void qemu_init_sigbus(void)
344 {
345 struct sigaction action;
346
347 memset(&action, 0, sizeof(action));
348 action.sa_flags = SA_SIGINFO;
349 action.sa_sigaction = sigbus_handler;
350 sigaction(SIGBUS, &action, NULL);
351
352 prctl(PR_MCE_KILL, PR_MCE_KILL_SET, PR_MCE_KILL_EARLY, 0, 0);
353 }
354 #else /* !CONFIG_LINUX */
355 static void qemu_init_sigbus(void)
356 {
357 }
358 #endif /* !CONFIG_LINUX */
359
360 static QemuThread io_thread;
361
362 /* cpu creation */
363 static QemuCond qemu_cpu_cond;
364 /* system init */
365 static QemuCond qemu_pause_cond;
366
367 void qemu_init_cpu_loop(void)
368 {
369 qemu_init_sigbus();
370 qemu_cond_init(&qemu_cpu_cond);
371 qemu_cond_init(&qemu_pause_cond);
372 qemu_mutex_init(&qemu_global_mutex);
373
374 qemu_thread_get_self(&io_thread);
375 }
376
377 void run_on_cpu(CPUState *cpu, run_on_cpu_func func, run_on_cpu_data data)
378 {
379 do_run_on_cpu(cpu, func, data, &qemu_global_mutex);
380 }
381
382 static void qemu_cpu_stop(CPUState *cpu, bool exit)
383 {
384 g_assert(qemu_cpu_is_self(cpu));
385 cpu->stop = false;
386 cpu->stopped = true;
387 if (exit) {
388 cpu_exit(cpu);
389 }
390 qemu_cond_broadcast(&qemu_pause_cond);
391 }
392
393 void qemu_wait_io_event_common(CPUState *cpu)
394 {
395 qatomic_mb_set(&cpu->thread_kicked, false);
396 if (cpu->stop) {
397 qemu_cpu_stop(cpu, false);
398 }
399 process_queued_cpu_work(cpu);
400 }
401
402 void qemu_wait_io_event(CPUState *cpu)
403 {
404 bool slept = false;
405
406 while (cpu_thread_is_idle(cpu)) {
407 if (!slept) {
408 slept = true;
409 qemu_plugin_vcpu_idle_cb(cpu);
410 }
411 qemu_cond_wait(cpu->halt_cond, &qemu_global_mutex);
412 }
413 if (slept) {
414 qemu_plugin_vcpu_resume_cb(cpu);
415 }
416
417 #ifdef _WIN32
418 /* Eat dummy APC queued by cpus_kick_thread. */
419 if (hax_enabled()) {
420 SleepEx(0, TRUE);
421 }
422 #endif
423 qemu_wait_io_event_common(cpu);
424 }
425
426 void cpus_kick_thread(CPUState *cpu)
427 {
428 #ifndef _WIN32
429 int err;
430
431 if (cpu->thread_kicked) {
432 return;
433 }
434 cpu->thread_kicked = true;
435 err = pthread_kill(cpu->thread->thread, SIG_IPI);
436 if (err && err != ESRCH) {
437 fprintf(stderr, "qemu:%s: %s", __func__, strerror(err));
438 exit(1);
439 }
440 #endif
441 }
442
443 void qemu_cpu_kick(CPUState *cpu)
444 {
445 qemu_cond_broadcast(cpu->halt_cond);
446 if (cpus_accel->kick_vcpu_thread) {
447 cpus_accel->kick_vcpu_thread(cpu);
448 } else { /* default */
449 cpus_kick_thread(cpu);
450 }
451 }
452
453 void qemu_cpu_kick_self(void)
454 {
455 assert(current_cpu);
456 cpus_kick_thread(current_cpu);
457 }
458
459 bool qemu_cpu_is_self(CPUState *cpu)
460 {
461 return qemu_thread_is_self(cpu->thread);
462 }
463
464 bool qemu_in_vcpu_thread(void)
465 {
466 return current_cpu && qemu_cpu_is_self(current_cpu);
467 }
468
469 static __thread bool iothread_locked = false;
470
471 bool qemu_mutex_iothread_locked(void)
472 {
473 return iothread_locked;
474 }
475
476 /*
477 * The BQL is taken from so many places that it is worth profiling the
478 * callers directly, instead of funneling them all through a single function.
479 */
480 void qemu_mutex_lock_iothread_impl(const char *file, int line)
481 {
482 QemuMutexLockFunc bql_lock = qatomic_read(&qemu_bql_mutex_lock_func);
483
484 g_assert(!qemu_mutex_iothread_locked());
485 bql_lock(&qemu_global_mutex, file, line);
486 iothread_locked = true;
487 }
488
489 void qemu_mutex_unlock_iothread(void)
490 {
491 g_assert(qemu_mutex_iothread_locked());
492 iothread_locked = false;
493 qemu_mutex_unlock(&qemu_global_mutex);
494 }
495
496 void qemu_cond_wait_iothread(QemuCond *cond)
497 {
498 qemu_cond_wait(cond, &qemu_global_mutex);
499 }
500
501 void qemu_cond_timedwait_iothread(QemuCond *cond, int ms)
502 {
503 qemu_cond_timedwait(cond, &qemu_global_mutex, ms);
504 }
505
506 /* signal CPU creation */
507 void cpu_thread_signal_created(CPUState *cpu)
508 {
509 cpu->created = true;
510 qemu_cond_signal(&qemu_cpu_cond);
511 }
512
513 /* signal CPU destruction */
514 void cpu_thread_signal_destroyed(CPUState *cpu)
515 {
516 cpu->created = false;
517 qemu_cond_signal(&qemu_cpu_cond);
518 }
519
520
521 static bool all_vcpus_paused(void)
522 {
523 CPUState *cpu;
524
525 CPU_FOREACH(cpu) {
526 if (!cpu->stopped) {
527 return false;
528 }
529 }
530
531 return true;
532 }
533
534 void pause_all_vcpus(void)
535 {
536 CPUState *cpu;
537
538 qemu_clock_enable(QEMU_CLOCK_VIRTUAL, false);
539 CPU_FOREACH(cpu) {
540 if (qemu_cpu_is_self(cpu)) {
541 qemu_cpu_stop(cpu, true);
542 } else {
543 cpu->stop = true;
544 qemu_cpu_kick(cpu);
545 }
546 }
547
548 /* We need to drop the replay_lock so any vCPU threads woken up
549 * can finish their replay tasks
550 */
551 replay_mutex_unlock();
552
553 while (!all_vcpus_paused()) {
554 qemu_cond_wait(&qemu_pause_cond, &qemu_global_mutex);
555 CPU_FOREACH(cpu) {
556 qemu_cpu_kick(cpu);
557 }
558 }
559
560 qemu_mutex_unlock_iothread();
561 replay_mutex_lock();
562 qemu_mutex_lock_iothread();
563 }
564
565 void cpu_resume(CPUState *cpu)
566 {
567 cpu->stop = false;
568 cpu->stopped = false;
569 qemu_cpu_kick(cpu);
570 }
571
572 void resume_all_vcpus(void)
573 {
574 CPUState *cpu;
575
576 if (!runstate_is_running()) {
577 return;
578 }
579
580 qemu_clock_enable(QEMU_CLOCK_VIRTUAL, true);
581 CPU_FOREACH(cpu) {
582 cpu_resume(cpu);
583 }
584 }
585
586 void cpu_remove_sync(CPUState *cpu)
587 {
588 cpu->stop = true;
589 cpu->unplug = true;
590 qemu_cpu_kick(cpu);
591 qemu_mutex_unlock_iothread();
592 qemu_thread_join(cpu->thread);
593 qemu_mutex_lock_iothread();
594 }
595
596 void cpus_register_accel(const CpusAccel *ca)
597 {
598 assert(ca != NULL);
599 assert(ca->create_vcpu_thread != NULL); /* mandatory */
600 cpus_accel = ca;
601 }
602
603 void qemu_init_vcpu(CPUState *cpu)
604 {
605 MachineState *ms = MACHINE(qdev_get_machine());
606
607 cpu->nr_cores = ms->smp.cores;
608 cpu->nr_threads = ms->smp.threads;
609 cpu->stopped = true;
610 cpu->random_seed = qemu_guest_random_seed_thread_part1();
611
612 if (!cpu->as) {
613 /* If the target cpu hasn't set up any address spaces itself,
614 * give it the default one.
615 */
616 cpu->num_ases = 1;
617 cpu_address_space_init(cpu, 0, "cpu-memory", cpu->memory);
618 }
619
620 /* accelerators all implement the CpusAccel interface */
621 g_assert(cpus_accel != NULL && cpus_accel->create_vcpu_thread != NULL);
622 cpus_accel->create_vcpu_thread(cpu);
623
624 while (!cpu->created) {
625 qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex);
626 }
627 }
628
629 void cpu_stop_current(void)
630 {
631 if (current_cpu) {
632 current_cpu->stop = true;
633 cpu_exit(current_cpu);
634 }
635 }
636
637 int vm_stop(RunState state)
638 {
639 if (qemu_in_vcpu_thread()) {
640 qemu_system_vmstop_request_prepare();
641 qemu_system_vmstop_request(state);
642 /*
643 * FIXME: should not return to device code in case
644 * vm_stop() has been requested.
645 */
646 cpu_stop_current();
647 return 0;
648 }
649
650 return do_vm_stop(state, true);
651 }
652
653 /**
654 * Prepare for (re)starting the VM.
655 * Returns -1 if the vCPUs are not to be restarted (e.g. if they are already
656 * running or in case of an error condition), 0 otherwise.
657 */
658 int vm_prepare_start(void)
659 {
660 RunState requested;
661
662 qemu_vmstop_requested(&requested);
663 if (runstate_is_running() && requested == RUN_STATE__MAX) {
664 return -1;
665 }
666
667 /* Ensure that a STOP/RESUME pair of events is emitted if a
668 * vmstop request was pending. The BLOCK_IO_ERROR event, for
669 * example, according to documentation is always followed by
670 * the STOP event.
671 */
672 if (runstate_is_running()) {
673 qapi_event_send_stop();
674 qapi_event_send_resume();
675 return -1;
676 }
677
678 /* We are sending this now, but the CPUs will be resumed shortly later */
679 qapi_event_send_resume();
680
681 cpu_enable_ticks();
682 runstate_set(RUN_STATE_RUNNING);
683 vm_state_notify(1, RUN_STATE_RUNNING);
684 return 0;
685 }
686
687 void vm_start(void)
688 {
689 if (!vm_prepare_start()) {
690 resume_all_vcpus();
691 }
692 }
693
694 /* does a state transition even if the VM is already stopped,
695 current state is forgotten forever */
696 int vm_stop_force_state(RunState state)
697 {
698 if (runstate_is_running()) {
699 return vm_stop(state);
700 } else {
701 runstate_set(state);
702
703 bdrv_drain_all();
704 /* Make sure to return an error if the flush in a previous vm_stop()
705 * failed. */
706 return bdrv_flush_all();
707 }
708 }
709
710 void list_cpus(const char *optarg)
711 {
712 /* XXX: implement xxx_cpu_list for targets that still miss it */
713 #if defined(cpu_list)
714 cpu_list();
715 #endif
716 }
717
718 void qmp_memsave(int64_t addr, int64_t size, const char *filename,
719 bool has_cpu, int64_t cpu_index, Error **errp)
720 {
721 FILE *f;
722 uint32_t l;
723 CPUState *cpu;
724 uint8_t buf[1024];
725 int64_t orig_addr = addr, orig_size = size;
726
727 if (!has_cpu) {
728 cpu_index = 0;
729 }
730
731 cpu = qemu_get_cpu(cpu_index);
732 if (cpu == NULL) {
733 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "cpu-index",
734 "a CPU number");
735 return;
736 }
737
738 f = fopen(filename, "wb");
739 if (!f) {
740 error_setg_file_open(errp, errno, filename);
741 return;
742 }
743
744 while (size != 0) {
745 l = sizeof(buf);
746 if (l > size)
747 l = size;
748 if (cpu_memory_rw_debug(cpu, addr, buf, l, 0) != 0) {
749 error_setg(errp, "Invalid addr 0x%016" PRIx64 "/size %" PRId64
750 " specified", orig_addr, orig_size);
751 goto exit;
752 }
753 if (fwrite(buf, 1, l, f) != l) {
754 error_setg(errp, QERR_IO_ERROR);
755 goto exit;
756 }
757 addr += l;
758 size -= l;
759 }
760
761 exit:
762 fclose(f);
763 }
764
765 void qmp_pmemsave(int64_t addr, int64_t size, const char *filename,
766 Error **errp)
767 {
768 FILE *f;
769 uint32_t l;
770 uint8_t buf[1024];
771
772 f = fopen(filename, "wb");
773 if (!f) {
774 error_setg_file_open(errp, errno, filename);
775 return;
776 }
777
778 while (size != 0) {
779 l = sizeof(buf);
780 if (l > size)
781 l = size;
782 cpu_physical_memory_read(addr, buf, l);
783 if (fwrite(buf, 1, l, f) != l) {
784 error_setg(errp, QERR_IO_ERROR);
785 goto exit;
786 }
787 addr += l;
788 size -= l;
789 }
790
791 exit:
792 fclose(f);
793 }
794
795 void qmp_inject_nmi(Error **errp)
796 {
797 nmi_monitor_handle(monitor_get_cpu_index(monitor_cur()), errp);
798 }
799