qcow2: Return right error code in write_refcount_block_entries
[qemu.git] / cpus.c
blob8341f6c201be07d941f21d6830ec4ed2c87530d0
1 /*
2 * QEMU System Emulator
4 * Copyright (c) 2003-2008 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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.
25 /* Needed early for CONFIG_BSD etc. */
26 #include "config-host.h"
28 #include "monitor.h"
29 #include "sysemu.h"
30 #include "gdbstub.h"
31 #include "dma.h"
32 #include "kvm.h"
34 #include "cpus.h"
36 #ifdef SIGRTMIN
37 #define SIG_IPI (SIGRTMIN+4)
38 #else
39 #define SIG_IPI SIGUSR1
40 #endif
42 static CPUState *cur_cpu;
43 static CPUState *next_cpu;
45 /***********************************************************/
46 void hw_error(const char *fmt, ...)
48 va_list ap;
49 CPUState *env;
51 va_start(ap, fmt);
52 fprintf(stderr, "qemu: hardware error: ");
53 vfprintf(stderr, fmt, ap);
54 fprintf(stderr, "\n");
55 for(env = first_cpu; env != NULL; env = env->next_cpu) {
56 fprintf(stderr, "CPU #%d:\n", env->cpu_index);
57 #ifdef TARGET_I386
58 cpu_dump_state(env, stderr, fprintf, X86_DUMP_FPU);
59 #else
60 cpu_dump_state(env, stderr, fprintf, 0);
61 #endif
63 va_end(ap);
64 abort();
67 void cpu_synchronize_all_states(void)
69 CPUState *cpu;
71 for (cpu = first_cpu; cpu; cpu = cpu->next_cpu) {
72 cpu_synchronize_state(cpu);
76 void cpu_synchronize_all_post_reset(void)
78 CPUState *cpu;
80 for (cpu = first_cpu; cpu; cpu = cpu->next_cpu) {
81 cpu_synchronize_post_reset(cpu);
85 void cpu_synchronize_all_post_init(void)
87 CPUState *cpu;
89 for (cpu = first_cpu; cpu; cpu = cpu->next_cpu) {
90 cpu_synchronize_post_init(cpu);
94 int cpu_is_stopped(CPUState *env)
96 return !vm_running || env->stopped;
99 static void do_vm_stop(int reason)
101 if (vm_running) {
102 cpu_disable_ticks();
103 vm_running = 0;
104 pause_all_vcpus();
105 vm_state_notify(0, reason);
106 monitor_protocol_event(QEVENT_STOP, NULL);
110 static int cpu_can_run(CPUState *env)
112 if (env->stop)
113 return 0;
114 if (env->stopped || !vm_running)
115 return 0;
116 return 1;
119 static int cpu_has_work(CPUState *env)
121 if (env->stop)
122 return 1;
123 if (env->queued_work_first)
124 return 1;
125 if (env->stopped || !vm_running)
126 return 0;
127 if (!env->halted)
128 return 1;
129 if (qemu_cpu_has_work(env))
130 return 1;
131 return 0;
134 static int tcg_has_work(void)
136 CPUState *env;
138 for (env = first_cpu; env != NULL; env = env->next_cpu)
139 if (cpu_has_work(env))
140 return 1;
141 return 0;
144 #ifndef _WIN32
145 static int io_thread_fd = -1;
147 static void qemu_event_increment(void)
149 /* Write 8 bytes to be compatible with eventfd. */
150 static const uint64_t val = 1;
151 ssize_t ret;
153 if (io_thread_fd == -1)
154 return;
156 do {
157 ret = write(io_thread_fd, &val, sizeof(val));
158 } while (ret < 0 && errno == EINTR);
160 /* EAGAIN is fine, a read must be pending. */
161 if (ret < 0 && errno != EAGAIN) {
162 fprintf(stderr, "qemu_event_increment: write() filed: %s\n",
163 strerror(errno));
164 exit (1);
168 static void qemu_event_read(void *opaque)
170 int fd = (unsigned long)opaque;
171 ssize_t len;
172 char buffer[512];
174 /* Drain the notify pipe. For eventfd, only 8 bytes will be read. */
175 do {
176 len = read(fd, buffer, sizeof(buffer));
177 } while ((len == -1 && errno == EINTR) || len == sizeof(buffer));
180 static int qemu_event_init(void)
182 int err;
183 int fds[2];
185 err = qemu_eventfd(fds);
186 if (err == -1)
187 return -errno;
189 err = fcntl_setfl(fds[0], O_NONBLOCK);
190 if (err < 0)
191 goto fail;
193 err = fcntl_setfl(fds[1], O_NONBLOCK);
194 if (err < 0)
195 goto fail;
197 qemu_set_fd_handler2(fds[0], NULL, qemu_event_read, NULL,
198 (void *)(unsigned long)fds[0]);
200 io_thread_fd = fds[1];
201 return 0;
203 fail:
204 close(fds[0]);
205 close(fds[1]);
206 return err;
208 #else
209 HANDLE qemu_event_handle;
211 static void dummy_event_handler(void *opaque)
215 static int qemu_event_init(void)
217 qemu_event_handle = CreateEvent(NULL, FALSE, FALSE, NULL);
218 if (!qemu_event_handle) {
219 fprintf(stderr, "Failed CreateEvent: %ld\n", GetLastError());
220 return -1;
222 qemu_add_wait_object(qemu_event_handle, dummy_event_handler, NULL);
223 return 0;
226 static void qemu_event_increment(void)
228 if (!SetEvent(qemu_event_handle)) {
229 fprintf(stderr, "qemu_event_increment: SetEvent failed: %ld\n",
230 GetLastError());
231 exit (1);
234 #endif
236 #ifndef CONFIG_IOTHREAD
237 int qemu_init_main_loop(void)
239 return qemu_event_init();
242 void qemu_main_loop_start(void)
246 void qemu_init_vcpu(void *_env)
248 CPUState *env = _env;
250 env->nr_cores = smp_cores;
251 env->nr_threads = smp_threads;
252 if (kvm_enabled())
253 kvm_init_vcpu(env);
254 return;
257 int qemu_cpu_self(void *env)
259 return 1;
262 void run_on_cpu(CPUState *env, void (*func)(void *data), void *data)
264 func(data);
267 void resume_all_vcpus(void)
271 void pause_all_vcpus(void)
275 void qemu_cpu_kick(void *env)
277 return;
280 void qemu_notify_event(void)
282 CPUState *env = cpu_single_env;
284 qemu_event_increment ();
285 if (env) {
286 cpu_exit(env);
288 if (next_cpu && env != next_cpu) {
289 cpu_exit(next_cpu);
293 void qemu_mutex_lock_iothread(void) {}
294 void qemu_mutex_unlock_iothread(void) {}
296 void vm_stop(int reason)
298 do_vm_stop(reason);
301 #else /* CONFIG_IOTHREAD */
303 #include "qemu-thread.h"
305 QemuMutex qemu_global_mutex;
306 static QemuMutex qemu_fair_mutex;
308 static QemuThread io_thread;
310 static QemuThread *tcg_cpu_thread;
311 static QemuCond *tcg_halt_cond;
313 static int qemu_system_ready;
314 /* cpu creation */
315 static QemuCond qemu_cpu_cond;
316 /* system init */
317 static QemuCond qemu_system_cond;
318 static QemuCond qemu_pause_cond;
319 static QemuCond qemu_work_cond;
321 static void tcg_block_io_signals(void);
322 static void kvm_block_io_signals(CPUState *env);
323 static void unblock_io_signals(void);
325 int qemu_init_main_loop(void)
327 int ret;
329 ret = qemu_event_init();
330 if (ret)
331 return ret;
333 qemu_cond_init(&qemu_pause_cond);
334 qemu_mutex_init(&qemu_fair_mutex);
335 qemu_mutex_init(&qemu_global_mutex);
336 qemu_mutex_lock(&qemu_global_mutex);
338 unblock_io_signals();
339 qemu_thread_self(&io_thread);
341 return 0;
344 void qemu_main_loop_start(void)
346 qemu_system_ready = 1;
347 qemu_cond_broadcast(&qemu_system_cond);
350 void run_on_cpu(CPUState *env, void (*func)(void *data), void *data)
352 struct qemu_work_item wi;
354 if (qemu_cpu_self(env)) {
355 func(data);
356 return;
359 wi.func = func;
360 wi.data = data;
361 if (!env->queued_work_first)
362 env->queued_work_first = &wi;
363 else
364 env->queued_work_last->next = &wi;
365 env->queued_work_last = &wi;
366 wi.next = NULL;
367 wi.done = false;
369 qemu_cpu_kick(env);
370 while (!wi.done) {
371 CPUState *self_env = cpu_single_env;
373 qemu_cond_wait(&qemu_work_cond, &qemu_global_mutex);
374 cpu_single_env = self_env;
378 static void flush_queued_work(CPUState *env)
380 struct qemu_work_item *wi;
382 if (!env->queued_work_first)
383 return;
385 while ((wi = env->queued_work_first)) {
386 env->queued_work_first = wi->next;
387 wi->func(wi->data);
388 wi->done = true;
390 env->queued_work_last = NULL;
391 qemu_cond_broadcast(&qemu_work_cond);
394 static void qemu_wait_io_event_common(CPUState *env)
396 if (env->stop) {
397 env->stop = 0;
398 env->stopped = 1;
399 qemu_cond_signal(&qemu_pause_cond);
401 flush_queued_work(env);
404 static void qemu_wait_io_event(CPUState *env)
406 while (!tcg_has_work())
407 qemu_cond_timedwait(env->halt_cond, &qemu_global_mutex, 1000);
409 qemu_mutex_unlock(&qemu_global_mutex);
412 * Users of qemu_global_mutex can be starved, having no chance
413 * to acquire it since this path will get to it first.
414 * So use another lock to provide fairness.
416 qemu_mutex_lock(&qemu_fair_mutex);
417 qemu_mutex_unlock(&qemu_fair_mutex);
419 qemu_mutex_lock(&qemu_global_mutex);
420 qemu_wait_io_event_common(env);
423 static void qemu_kvm_eat_signal(CPUState *env, int timeout)
425 struct timespec ts;
426 int r, e;
427 siginfo_t siginfo;
428 sigset_t waitset;
430 ts.tv_sec = timeout / 1000;
431 ts.tv_nsec = (timeout % 1000) * 1000000;
433 sigemptyset(&waitset);
434 sigaddset(&waitset, SIG_IPI);
436 qemu_mutex_unlock(&qemu_global_mutex);
437 r = sigtimedwait(&waitset, &siginfo, &ts);
438 e = errno;
439 qemu_mutex_lock(&qemu_global_mutex);
441 if (r == -1 && !(e == EAGAIN || e == EINTR)) {
442 fprintf(stderr, "sigtimedwait: %s\n", strerror(e));
443 exit(1);
447 static void qemu_kvm_wait_io_event(CPUState *env)
449 while (!cpu_has_work(env))
450 qemu_cond_timedwait(env->halt_cond, &qemu_global_mutex, 1000);
452 qemu_kvm_eat_signal(env, 0);
453 qemu_wait_io_event_common(env);
456 static int qemu_cpu_exec(CPUState *env);
458 static void *kvm_cpu_thread_fn(void *arg)
460 CPUState *env = arg;
462 qemu_mutex_lock(&qemu_global_mutex);
463 qemu_thread_self(env->thread);
464 if (kvm_enabled())
465 kvm_init_vcpu(env);
467 kvm_block_io_signals(env);
469 /* signal CPU creation */
470 env->created = 1;
471 qemu_cond_signal(&qemu_cpu_cond);
473 /* and wait for machine initialization */
474 while (!qemu_system_ready)
475 qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100);
477 while (1) {
478 if (cpu_can_run(env))
479 qemu_cpu_exec(env);
480 qemu_kvm_wait_io_event(env);
483 return NULL;
486 static void *tcg_cpu_thread_fn(void *arg)
488 CPUState *env = arg;
490 tcg_block_io_signals();
491 qemu_thread_self(env->thread);
493 /* signal CPU creation */
494 qemu_mutex_lock(&qemu_global_mutex);
495 for (env = first_cpu; env != NULL; env = env->next_cpu)
496 env->created = 1;
497 qemu_cond_signal(&qemu_cpu_cond);
499 /* and wait for machine initialization */
500 while (!qemu_system_ready)
501 qemu_cond_timedwait(&qemu_system_cond, &qemu_global_mutex, 100);
503 while (1) {
504 tcg_cpu_exec();
505 qemu_wait_io_event(cur_cpu);
508 return NULL;
511 void qemu_cpu_kick(void *_env)
513 CPUState *env = _env;
514 qemu_cond_broadcast(env->halt_cond);
515 qemu_thread_signal(env->thread, SIG_IPI);
518 int qemu_cpu_self(void *_env)
520 CPUState *env = _env;
521 QemuThread this;
523 qemu_thread_self(&this);
525 return qemu_thread_equal(&this, env->thread);
528 static void cpu_signal(int sig)
530 if (cpu_single_env)
531 cpu_exit(cpu_single_env);
532 exit_request = 1;
535 static void tcg_block_io_signals(void)
537 sigset_t set;
538 struct sigaction sigact;
540 sigemptyset(&set);
541 sigaddset(&set, SIGUSR2);
542 sigaddset(&set, SIGIO);
543 sigaddset(&set, SIGALRM);
544 sigaddset(&set, SIGCHLD);
545 pthread_sigmask(SIG_BLOCK, &set, NULL);
547 sigemptyset(&set);
548 sigaddset(&set, SIG_IPI);
549 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
551 memset(&sigact, 0, sizeof(sigact));
552 sigact.sa_handler = cpu_signal;
553 sigaction(SIG_IPI, &sigact, NULL);
556 static void dummy_signal(int sig)
560 static void kvm_block_io_signals(CPUState *env)
562 int r;
563 sigset_t set;
564 struct sigaction sigact;
566 sigemptyset(&set);
567 sigaddset(&set, SIGUSR2);
568 sigaddset(&set, SIGIO);
569 sigaddset(&set, SIGALRM);
570 sigaddset(&set, SIGCHLD);
571 sigaddset(&set, SIG_IPI);
572 pthread_sigmask(SIG_BLOCK, &set, NULL);
574 pthread_sigmask(SIG_BLOCK, NULL, &set);
575 sigdelset(&set, SIG_IPI);
577 memset(&sigact, 0, sizeof(sigact));
578 sigact.sa_handler = dummy_signal;
579 sigaction(SIG_IPI, &sigact, NULL);
581 r = kvm_set_signal_mask(env, &set);
582 if (r) {
583 fprintf(stderr, "kvm_set_signal_mask: %s\n", strerror(r));
584 exit(1);
588 static void unblock_io_signals(void)
590 sigset_t set;
592 sigemptyset(&set);
593 sigaddset(&set, SIGUSR2);
594 sigaddset(&set, SIGIO);
595 sigaddset(&set, SIGALRM);
596 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
598 sigemptyset(&set);
599 sigaddset(&set, SIG_IPI);
600 pthread_sigmask(SIG_BLOCK, &set, NULL);
603 void qemu_mutex_lock_iothread(void)
605 if (kvm_enabled()) {
606 qemu_mutex_lock(&qemu_fair_mutex);
607 qemu_mutex_lock(&qemu_global_mutex);
608 qemu_mutex_unlock(&qemu_fair_mutex);
609 } else {
610 qemu_mutex_lock(&qemu_fair_mutex);
611 if (qemu_mutex_trylock(&qemu_global_mutex)) {
612 qemu_thread_signal(tcg_cpu_thread, SIG_IPI);
613 qemu_mutex_lock(&qemu_global_mutex);
615 qemu_mutex_unlock(&qemu_fair_mutex);
619 void qemu_mutex_unlock_iothread(void)
621 qemu_mutex_unlock(&qemu_global_mutex);
624 static int all_vcpus_paused(void)
626 CPUState *penv = first_cpu;
628 while (penv) {
629 if (!penv->stopped)
630 return 0;
631 penv = (CPUState *)penv->next_cpu;
634 return 1;
637 void pause_all_vcpus(void)
639 CPUState *penv = first_cpu;
641 while (penv) {
642 penv->stop = 1;
643 qemu_cpu_kick(penv);
644 penv = (CPUState *)penv->next_cpu;
647 while (!all_vcpus_paused()) {
648 qemu_cond_timedwait(&qemu_pause_cond, &qemu_global_mutex, 100);
649 penv = first_cpu;
650 while (penv) {
651 qemu_cpu_kick(penv);
652 penv = (CPUState *)penv->next_cpu;
657 void resume_all_vcpus(void)
659 CPUState *penv = first_cpu;
661 while (penv) {
662 penv->stop = 0;
663 penv->stopped = 0;
664 qemu_cpu_kick(penv);
665 penv = (CPUState *)penv->next_cpu;
669 static void tcg_init_vcpu(void *_env)
671 CPUState *env = _env;
672 /* share a single thread for all cpus with TCG */
673 if (!tcg_cpu_thread) {
674 env->thread = qemu_mallocz(sizeof(QemuThread));
675 env->halt_cond = qemu_mallocz(sizeof(QemuCond));
676 qemu_cond_init(env->halt_cond);
677 qemu_thread_create(env->thread, tcg_cpu_thread_fn, env);
678 while (env->created == 0)
679 qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100);
680 tcg_cpu_thread = env->thread;
681 tcg_halt_cond = env->halt_cond;
682 } else {
683 env->thread = tcg_cpu_thread;
684 env->halt_cond = tcg_halt_cond;
688 static void kvm_start_vcpu(CPUState *env)
690 env->thread = qemu_mallocz(sizeof(QemuThread));
691 env->halt_cond = qemu_mallocz(sizeof(QemuCond));
692 qemu_cond_init(env->halt_cond);
693 qemu_thread_create(env->thread, kvm_cpu_thread_fn, env);
694 while (env->created == 0)
695 qemu_cond_timedwait(&qemu_cpu_cond, &qemu_global_mutex, 100);
698 void qemu_init_vcpu(void *_env)
700 CPUState *env = _env;
702 env->nr_cores = smp_cores;
703 env->nr_threads = smp_threads;
704 if (kvm_enabled())
705 kvm_start_vcpu(env);
706 else
707 tcg_init_vcpu(env);
710 void qemu_notify_event(void)
712 qemu_event_increment();
715 static void qemu_system_vmstop_request(int reason)
717 vmstop_requested = reason;
718 qemu_notify_event();
721 void vm_stop(int reason)
723 QemuThread me;
724 qemu_thread_self(&me);
726 if (!qemu_thread_equal(&me, &io_thread)) {
727 qemu_system_vmstop_request(reason);
729 * FIXME: should not return to device code in case
730 * vm_stop() has been requested.
732 if (cpu_single_env) {
733 cpu_exit(cpu_single_env);
734 cpu_single_env->stop = 1;
736 return;
738 do_vm_stop(reason);
741 #endif
743 static int qemu_cpu_exec(CPUState *env)
745 int ret;
746 #ifdef CONFIG_PROFILER
747 int64_t ti;
748 #endif
750 #ifdef CONFIG_PROFILER
751 ti = profile_getclock();
752 #endif
753 if (use_icount) {
754 int64_t count;
755 int decr;
756 qemu_icount -= (env->icount_decr.u16.low + env->icount_extra);
757 env->icount_decr.u16.low = 0;
758 env->icount_extra = 0;
759 count = qemu_icount_round (qemu_next_deadline());
760 qemu_icount += count;
761 decr = (count > 0xffff) ? 0xffff : count;
762 count -= decr;
763 env->icount_decr.u16.low = decr;
764 env->icount_extra = count;
766 ret = cpu_exec(env);
767 #ifdef CONFIG_PROFILER
768 qemu_time += profile_getclock() - ti;
769 #endif
770 if (use_icount) {
771 /* Fold pending instructions back into the
772 instruction counter, and clear the interrupt flag. */
773 qemu_icount -= (env->icount_decr.u16.low
774 + env->icount_extra);
775 env->icount_decr.u32 = 0;
776 env->icount_extra = 0;
778 return ret;
781 bool tcg_cpu_exec(void)
783 int ret = 0;
785 if (next_cpu == NULL)
786 next_cpu = first_cpu;
787 for (; next_cpu != NULL; next_cpu = next_cpu->next_cpu) {
788 CPUState *env = cur_cpu = next_cpu;
790 qemu_clock_enable(vm_clock,
791 (cur_cpu->singlestep_enabled & SSTEP_NOTIMER) == 0);
793 if (qemu_alarm_pending())
794 break;
795 if (cpu_can_run(env))
796 ret = qemu_cpu_exec(env);
797 else if (env->stop)
798 break;
800 if (ret == EXCP_DEBUG) {
801 gdb_set_stop_cpu(env);
802 debug_requested = EXCP_DEBUG;
803 break;
806 return tcg_has_work();
809 void set_numa_modes(void)
811 CPUState *env;
812 int i;
814 for (env = first_cpu; env != NULL; env = env->next_cpu) {
815 for (i = 0; i < nb_numa_nodes; i++) {
816 if (node_cpumask[i] & (1 << env->cpu_index)) {
817 env->numa_node = i;
823 void set_cpu_log(const char *optarg)
825 int mask;
826 const CPULogItem *item;
828 mask = cpu_str_to_log_mask(optarg);
829 if (!mask) {
830 printf("Log items (comma separated):\n");
831 for (item = cpu_log_items; item->mask != 0; item++) {
832 printf("%-10s %s\n", item->name, item->help);
834 exit(1);
836 cpu_set_log(mask);
839 /* Return the virtual CPU time, based on the instruction counter. */
840 int64_t cpu_get_icount(void)
842 int64_t icount;
843 CPUState *env = cpu_single_env;;
845 icount = qemu_icount;
846 if (env) {
847 if (!can_do_io(env)) {
848 fprintf(stderr, "Bad clock read\n");
850 icount -= (env->icount_decr.u16.low + env->icount_extra);
852 return qemu_icount_bias + (icount << icount_time_shift);
855 void list_cpus(FILE *f, int (*cpu_fprintf)(FILE *f, const char *fmt, ...),
856 const char *optarg)
858 /* XXX: implement xxx_cpu_list for targets that still miss it */
859 #if defined(cpu_list_id)
860 cpu_list_id(f, cpu_fprintf, optarg);
861 #elif defined(cpu_list)
862 cpu_list(f, cpu_fprintf); /* deprecated */
863 #endif