MAINTAINERS: Cover docs/igd-assign.txt in VFIO section
[qemu/ar7.git] / softmmu / cpus.c
bloba7ee431187adb67137c0c9b0e1605d50f2ba5f89
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 #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 "sysemu/whpx.h"
45 #include "hw/boards.h"
46 #include "hw/hw.h"
48 #ifdef CONFIG_LINUX
50 #include <sys/prctl.h>
52 #ifndef PR_MCE_KILL
53 #define PR_MCE_KILL 33
54 #endif
56 #ifndef PR_MCE_KILL_SET
57 #define PR_MCE_KILL_SET 1
58 #endif
60 #ifndef PR_MCE_KILL_EARLY
61 #define PR_MCE_KILL_EARLY 1
62 #endif
64 #endif /* CONFIG_LINUX */
66 static QemuMutex qemu_global_mutex;
68 bool cpu_is_stopped(CPUState *cpu)
70 return cpu->stopped || !runstate_is_running();
73 bool cpu_work_list_empty(CPUState *cpu)
75 bool ret;
77 qemu_mutex_lock(&cpu->work_mutex);
78 ret = QSIMPLEQ_EMPTY(&cpu->work_list);
79 qemu_mutex_unlock(&cpu->work_mutex);
80 return ret;
83 bool cpu_thread_is_idle(CPUState *cpu)
85 if (cpu->stop || !cpu_work_list_empty(cpu)) {
86 return false;
88 if (cpu_is_stopped(cpu)) {
89 return true;
91 if (!cpu->halted || cpu_has_work(cpu) ||
92 kvm_halt_in_kernel() || whpx_apic_in_platform()) {
93 return false;
95 return true;
98 bool all_cpu_threads_idle(void)
100 CPUState *cpu;
102 CPU_FOREACH(cpu) {
103 if (!cpu_thread_is_idle(cpu)) {
104 return false;
107 return true;
110 /***********************************************************/
111 void hw_error(const char *fmt, ...)
113 va_list ap;
114 CPUState *cpu;
116 va_start(ap, fmt);
117 fprintf(stderr, "qemu: hardware error: ");
118 vfprintf(stderr, fmt, ap);
119 fprintf(stderr, "\n");
120 CPU_FOREACH(cpu) {
121 fprintf(stderr, "CPU #%d:\n", cpu->cpu_index);
122 cpu_dump_state(cpu, stderr, CPU_DUMP_FPU);
124 va_end(ap);
125 abort();
129 * The chosen accelerator is supposed to register this.
131 static const AccelOpsClass *cpus_accel;
133 void cpu_synchronize_all_states(void)
135 CPUState *cpu;
137 CPU_FOREACH(cpu) {
138 cpu_synchronize_state(cpu);
142 void cpu_synchronize_all_post_reset(void)
144 CPUState *cpu;
146 CPU_FOREACH(cpu) {
147 cpu_synchronize_post_reset(cpu);
151 void cpu_synchronize_all_post_init(void)
153 CPUState *cpu;
155 CPU_FOREACH(cpu) {
156 cpu_synchronize_post_init(cpu);
160 void cpu_synchronize_all_pre_loadvm(void)
162 CPUState *cpu;
164 CPU_FOREACH(cpu) {
165 cpu_synchronize_pre_loadvm(cpu);
169 void cpu_synchronize_state(CPUState *cpu)
171 if (cpus_accel->synchronize_state) {
172 cpus_accel->synchronize_state(cpu);
176 void cpu_synchronize_post_reset(CPUState *cpu)
178 if (cpus_accel->synchronize_post_reset) {
179 cpus_accel->synchronize_post_reset(cpu);
183 void cpu_synchronize_post_init(CPUState *cpu)
185 if (cpus_accel->synchronize_post_init) {
186 cpus_accel->synchronize_post_init(cpu);
190 void cpu_synchronize_pre_loadvm(CPUState *cpu)
192 if (cpus_accel->synchronize_pre_loadvm) {
193 cpus_accel->synchronize_pre_loadvm(cpu);
197 bool cpus_are_resettable(void)
199 return cpu_check_are_resettable();
202 int64_t cpus_get_virtual_clock(void)
205 * XXX
207 * need to check that cpus_accel is not NULL, because qcow2 calls
208 * qemu_get_clock_ns(CLOCK_VIRTUAL) without any accel initialized and
209 * with ticks disabled in some io-tests:
210 * 030 040 041 060 099 120 127 140 156 161 172 181 191 192 195 203 229 249 256 267
212 * is this expected?
214 * XXX
216 if (cpus_accel && cpus_accel->get_virtual_clock) {
217 return cpus_accel->get_virtual_clock();
219 return cpu_get_clock();
223 * return the time elapsed in VM between vm_start and vm_stop. Unless
224 * icount is active, cpus_get_elapsed_ticks() uses units of the host CPU cycle
225 * counter.
227 int64_t cpus_get_elapsed_ticks(void)
229 if (cpus_accel->get_elapsed_ticks) {
230 return cpus_accel->get_elapsed_ticks();
232 return cpu_get_ticks();
235 static void generic_handle_interrupt(CPUState *cpu, int mask)
237 cpu->interrupt_request |= mask;
239 if (!qemu_cpu_is_self(cpu)) {
240 qemu_cpu_kick(cpu);
244 void cpu_interrupt(CPUState *cpu, int mask)
246 if (cpus_accel->handle_interrupt) {
247 cpus_accel->handle_interrupt(cpu, mask);
248 } else {
249 generic_handle_interrupt(cpu, mask);
253 static int do_vm_stop(RunState state, bool send_stop)
255 int ret = 0;
257 if (runstate_is_running()) {
258 runstate_set(state);
259 cpu_disable_ticks();
260 pause_all_vcpus();
261 vm_state_notify(0, state);
262 if (send_stop) {
263 qapi_event_send_stop();
267 bdrv_drain_all();
268 ret = bdrv_flush_all();
270 return ret;
273 /* Special vm_stop() variant for terminating the process. Historically clients
274 * did not expect a QMP STOP event and so we need to retain compatibility.
276 int vm_shutdown(void)
278 return do_vm_stop(RUN_STATE_SHUTDOWN, false);
281 bool cpu_can_run(CPUState *cpu)
283 if (cpu->stop) {
284 return false;
286 if (cpu_is_stopped(cpu)) {
287 return false;
289 return true;
292 void cpu_handle_guest_debug(CPUState *cpu)
294 if (replay_running_debug()) {
295 if (!cpu->singlestep_enabled) {
297 * Report about the breakpoint and
298 * make a single step to skip it
300 replay_breakpoint();
301 cpu_single_step(cpu, SSTEP_ENABLE);
302 } else {
303 cpu_single_step(cpu, 0);
305 } else {
306 gdb_set_stop_cpu(cpu);
307 qemu_system_debug_request();
308 cpu->stopped = true;
312 #ifdef CONFIG_LINUX
313 static void sigbus_reraise(void)
315 sigset_t set;
316 struct sigaction action;
318 memset(&action, 0, sizeof(action));
319 action.sa_handler = SIG_DFL;
320 if (!sigaction(SIGBUS, &action, NULL)) {
321 raise(SIGBUS);
322 sigemptyset(&set);
323 sigaddset(&set, SIGBUS);
324 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
326 perror("Failed to re-raise SIGBUS!\n");
327 abort();
330 static void sigbus_handler(int n, siginfo_t *siginfo, void *ctx)
332 if (siginfo->si_code != BUS_MCEERR_AO && siginfo->si_code != BUS_MCEERR_AR) {
333 sigbus_reraise();
336 if (current_cpu) {
337 /* Called asynchronously in VCPU thread. */
338 if (kvm_on_sigbus_vcpu(current_cpu, siginfo->si_code, siginfo->si_addr)) {
339 sigbus_reraise();
341 } else {
342 /* Called synchronously (via signalfd) in main thread. */
343 if (kvm_on_sigbus(siginfo->si_code, siginfo->si_addr)) {
344 sigbus_reraise();
349 static void qemu_init_sigbus(void)
351 struct sigaction action;
353 memset(&action, 0, sizeof(action));
354 action.sa_flags = SA_SIGINFO;
355 action.sa_sigaction = sigbus_handler;
356 sigaction(SIGBUS, &action, NULL);
358 prctl(PR_MCE_KILL, PR_MCE_KILL_SET, PR_MCE_KILL_EARLY, 0, 0);
360 #else /* !CONFIG_LINUX */
361 static void qemu_init_sigbus(void)
364 #endif /* !CONFIG_LINUX */
366 static QemuThread io_thread;
368 /* cpu creation */
369 static QemuCond qemu_cpu_cond;
370 /* system init */
371 static QemuCond qemu_pause_cond;
373 void qemu_init_cpu_loop(void)
375 qemu_init_sigbus();
376 qemu_cond_init(&qemu_cpu_cond);
377 qemu_cond_init(&qemu_pause_cond);
378 qemu_mutex_init(&qemu_global_mutex);
380 qemu_thread_get_self(&io_thread);
383 void run_on_cpu(CPUState *cpu, run_on_cpu_func func, run_on_cpu_data data)
385 do_run_on_cpu(cpu, func, data, &qemu_global_mutex);
388 static void qemu_cpu_stop(CPUState *cpu, bool exit)
390 g_assert(qemu_cpu_is_self(cpu));
391 cpu->stop = false;
392 cpu->stopped = true;
393 if (exit) {
394 cpu_exit(cpu);
396 qemu_cond_broadcast(&qemu_pause_cond);
399 void qemu_wait_io_event_common(CPUState *cpu)
401 qatomic_mb_set(&cpu->thread_kicked, false);
402 if (cpu->stop) {
403 qemu_cpu_stop(cpu, false);
405 process_queued_cpu_work(cpu);
408 void qemu_wait_io_event(CPUState *cpu)
410 bool slept = false;
412 while (cpu_thread_is_idle(cpu)) {
413 if (!slept) {
414 slept = true;
415 qemu_plugin_vcpu_idle_cb(cpu);
417 qemu_cond_wait(cpu->halt_cond, &qemu_global_mutex);
419 if (slept) {
420 qemu_plugin_vcpu_resume_cb(cpu);
423 #ifdef _WIN32
424 /* Eat dummy APC queued by cpus_kick_thread. */
425 if (hax_enabled()) {
426 SleepEx(0, TRUE);
428 #endif
429 qemu_wait_io_event_common(cpu);
432 void cpus_kick_thread(CPUState *cpu)
434 #ifndef _WIN32
435 int err;
437 if (cpu->thread_kicked) {
438 return;
440 cpu->thread_kicked = true;
441 err = pthread_kill(cpu->thread->thread, SIG_IPI);
442 if (err && err != ESRCH) {
443 fprintf(stderr, "qemu:%s: %s", __func__, strerror(err));
444 exit(1);
446 #endif
449 void qemu_cpu_kick(CPUState *cpu)
451 qemu_cond_broadcast(cpu->halt_cond);
452 if (cpus_accel->kick_vcpu_thread) {
453 cpus_accel->kick_vcpu_thread(cpu);
454 } else { /* default */
455 cpus_kick_thread(cpu);
459 void qemu_cpu_kick_self(void)
461 assert(current_cpu);
462 cpus_kick_thread(current_cpu);
465 bool qemu_cpu_is_self(CPUState *cpu)
467 return qemu_thread_is_self(cpu->thread);
470 bool qemu_in_vcpu_thread(void)
472 return current_cpu && qemu_cpu_is_self(current_cpu);
475 static __thread bool iothread_locked = false;
477 bool qemu_mutex_iothread_locked(void)
479 return iothread_locked;
483 * The BQL is taken from so many places that it is worth profiling the
484 * callers directly, instead of funneling them all through a single function.
486 void qemu_mutex_lock_iothread_impl(const char *file, int line)
488 QemuMutexLockFunc bql_lock = qatomic_read(&qemu_bql_mutex_lock_func);
490 g_assert(!qemu_mutex_iothread_locked());
491 bql_lock(&qemu_global_mutex, file, line);
492 iothread_locked = true;
495 void qemu_mutex_unlock_iothread(void)
497 g_assert(qemu_mutex_iothread_locked());
498 iothread_locked = false;
499 qemu_mutex_unlock(&qemu_global_mutex);
502 void qemu_cond_wait_iothread(QemuCond *cond)
504 qemu_cond_wait(cond, &qemu_global_mutex);
507 void qemu_cond_timedwait_iothread(QemuCond *cond, int ms)
509 qemu_cond_timedwait(cond, &qemu_global_mutex, ms);
512 /* signal CPU creation */
513 void cpu_thread_signal_created(CPUState *cpu)
515 cpu->created = true;
516 qemu_cond_signal(&qemu_cpu_cond);
519 /* signal CPU destruction */
520 void cpu_thread_signal_destroyed(CPUState *cpu)
522 cpu->created = false;
523 qemu_cond_signal(&qemu_cpu_cond);
527 static bool all_vcpus_paused(void)
529 CPUState *cpu;
531 CPU_FOREACH(cpu) {
532 if (!cpu->stopped) {
533 return false;
537 return true;
540 void pause_all_vcpus(void)
542 CPUState *cpu;
544 qemu_clock_enable(QEMU_CLOCK_VIRTUAL, false);
545 CPU_FOREACH(cpu) {
546 if (qemu_cpu_is_self(cpu)) {
547 qemu_cpu_stop(cpu, true);
548 } else {
549 cpu->stop = true;
550 qemu_cpu_kick(cpu);
554 /* We need to drop the replay_lock so any vCPU threads woken up
555 * can finish their replay tasks
557 replay_mutex_unlock();
559 while (!all_vcpus_paused()) {
560 qemu_cond_wait(&qemu_pause_cond, &qemu_global_mutex);
561 CPU_FOREACH(cpu) {
562 qemu_cpu_kick(cpu);
566 qemu_mutex_unlock_iothread();
567 replay_mutex_lock();
568 qemu_mutex_lock_iothread();
571 void cpu_resume(CPUState *cpu)
573 cpu->stop = false;
574 cpu->stopped = false;
575 qemu_cpu_kick(cpu);
578 void resume_all_vcpus(void)
580 CPUState *cpu;
582 if (!runstate_is_running()) {
583 return;
586 qemu_clock_enable(QEMU_CLOCK_VIRTUAL, true);
587 CPU_FOREACH(cpu) {
588 cpu_resume(cpu);
592 void cpu_remove_sync(CPUState *cpu)
594 cpu->stop = true;
595 cpu->unplug = true;
596 qemu_cpu_kick(cpu);
597 qemu_mutex_unlock_iothread();
598 qemu_thread_join(cpu->thread);
599 qemu_mutex_lock_iothread();
602 void cpus_register_accel(const AccelOpsClass *ops)
604 assert(ops != NULL);
605 assert(ops->create_vcpu_thread != NULL); /* mandatory */
606 cpus_accel = ops;
609 void qemu_init_vcpu(CPUState *cpu)
611 MachineState *ms = MACHINE(qdev_get_machine());
613 cpu->nr_cores = ms->smp.cores;
614 cpu->nr_threads = ms->smp.threads;
615 cpu->stopped = true;
616 cpu->random_seed = qemu_guest_random_seed_thread_part1();
618 if (!cpu->as) {
619 /* If the target cpu hasn't set up any address spaces itself,
620 * give it the default one.
622 cpu->num_ases = 1;
623 cpu_address_space_init(cpu, 0, "cpu-memory", cpu->memory);
626 /* accelerators all implement the AccelOpsClass */
627 g_assert(cpus_accel != NULL && cpus_accel->create_vcpu_thread != NULL);
628 cpus_accel->create_vcpu_thread(cpu);
630 while (!cpu->created) {
631 qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex);
635 void cpu_stop_current(void)
637 if (current_cpu) {
638 current_cpu->stop = true;
639 cpu_exit(current_cpu);
643 int vm_stop(RunState state)
645 if (qemu_in_vcpu_thread()) {
646 qemu_system_vmstop_request_prepare();
647 qemu_system_vmstop_request(state);
649 * FIXME: should not return to device code in case
650 * vm_stop() has been requested.
652 cpu_stop_current();
653 return 0;
656 return do_vm_stop(state, true);
660 * Prepare for (re)starting the VM.
661 * Returns -1 if the vCPUs are not to be restarted (e.g. if they are already
662 * running or in case of an error condition), 0 otherwise.
664 int vm_prepare_start(void)
666 RunState requested;
668 qemu_vmstop_requested(&requested);
669 if (runstate_is_running() && requested == RUN_STATE__MAX) {
670 return -1;
673 /* Ensure that a STOP/RESUME pair of events is emitted if a
674 * vmstop request was pending. The BLOCK_IO_ERROR event, for
675 * example, according to documentation is always followed by
676 * the STOP event.
678 if (runstate_is_running()) {
679 qapi_event_send_stop();
680 qapi_event_send_resume();
681 return -1;
684 /* We are sending this now, but the CPUs will be resumed shortly later */
685 qapi_event_send_resume();
687 cpu_enable_ticks();
688 runstate_set(RUN_STATE_RUNNING);
689 vm_state_notify(1, RUN_STATE_RUNNING);
690 return 0;
693 void vm_start(void)
695 if (!vm_prepare_start()) {
696 resume_all_vcpus();
700 /* does a state transition even if the VM is already stopped,
701 current state is forgotten forever */
702 int vm_stop_force_state(RunState state)
704 if (runstate_is_running()) {
705 return vm_stop(state);
706 } else {
707 runstate_set(state);
709 bdrv_drain_all();
710 /* Make sure to return an error if the flush in a previous vm_stop()
711 * failed. */
712 return bdrv_flush_all();
716 void list_cpus(const char *optarg)
718 /* XXX: implement xxx_cpu_list for targets that still miss it */
719 #if defined(cpu_list)
720 cpu_list();
721 #endif
724 void qmp_memsave(int64_t addr, int64_t size, const char *filename,
725 bool has_cpu, int64_t cpu_index, Error **errp)
727 FILE *f;
728 uint32_t l;
729 CPUState *cpu;
730 uint8_t buf[1024];
731 int64_t orig_addr = addr, orig_size = size;
733 if (!has_cpu) {
734 cpu_index = 0;
737 cpu = qemu_get_cpu(cpu_index);
738 if (cpu == NULL) {
739 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "cpu-index",
740 "a CPU number");
741 return;
744 f = fopen(filename, "wb");
745 if (!f) {
746 error_setg_file_open(errp, errno, filename);
747 return;
750 while (size != 0) {
751 l = sizeof(buf);
752 if (l > size)
753 l = size;
754 if (cpu_memory_rw_debug(cpu, addr, buf, l, 0) != 0) {
755 error_setg(errp, "Invalid addr 0x%016" PRIx64 "/size %" PRId64
756 " specified", orig_addr, orig_size);
757 goto exit;
759 if (fwrite(buf, 1, l, f) != l) {
760 error_setg(errp, QERR_IO_ERROR);
761 goto exit;
763 addr += l;
764 size -= l;
767 exit:
768 fclose(f);
771 void qmp_pmemsave(int64_t addr, int64_t size, const char *filename,
772 Error **errp)
774 FILE *f;
775 uint32_t l;
776 uint8_t buf[1024];
778 f = fopen(filename, "wb");
779 if (!f) {
780 error_setg_file_open(errp, errno, filename);
781 return;
784 while (size != 0) {
785 l = sizeof(buf);
786 if (l > size)
787 l = size;
788 cpu_physical_memory_read(addr, buf, l);
789 if (fwrite(buf, 1, l, f) != l) {
790 error_setg(errp, QERR_IO_ERROR);
791 goto exit;
793 addr += l;
794 size -= l;
797 exit:
798 fclose(f);
801 void qmp_inject_nmi(Error **errp)
803 nmi_monitor_handle(monitor_get_cpu_index(monitor_cur()), errp);