Merge tag 'misc-next-pull-request' of https://gitlab.com/berrange/qemu into staging
[qemu.git] / softmmu / cpus.c
blob23b30484b2f8cf7618b03ba48d4a4967790ef98d
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 "monitor/monitor.h"
27 #include "qemu/coroutine-tls.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/cpu-common.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"
47 #include "trace.h"
49 #ifdef CONFIG_LINUX
51 #include <sys/prctl.h>
53 #ifndef PR_MCE_KILL
54 #define PR_MCE_KILL 33
55 #endif
57 #ifndef PR_MCE_KILL_SET
58 #define PR_MCE_KILL_SET 1
59 #endif
61 #ifndef PR_MCE_KILL_EARLY
62 #define PR_MCE_KILL_EARLY 1
63 #endif
65 #endif /* CONFIG_LINUX */
67 static QemuMutex qemu_global_mutex;
70 * The chosen accelerator is supposed to register this.
72 static const AccelOpsClass *cpus_accel;
74 bool cpu_is_stopped(CPUState *cpu)
76 return cpu->stopped || !runstate_is_running();
79 bool cpu_work_list_empty(CPUState *cpu)
81 return QSIMPLEQ_EMPTY_ATOMIC(&cpu->work_list);
84 bool cpu_thread_is_idle(CPUState *cpu)
86 if (cpu->stop || !cpu_work_list_empty(cpu)) {
87 return false;
89 if (cpu_is_stopped(cpu)) {
90 return true;
92 if (!cpu->halted || cpu_has_work(cpu)) {
93 return false;
95 if (cpus_accel->cpu_thread_is_idle) {
96 return cpus_accel->cpu_thread_is_idle(cpu);
98 return true;
101 bool all_cpu_threads_idle(void)
103 CPUState *cpu;
105 CPU_FOREACH(cpu) {
106 if (!cpu_thread_is_idle(cpu)) {
107 return false;
110 return true;
113 /***********************************************************/
114 void hw_error(const char *fmt, ...)
116 va_list ap;
117 CPUState *cpu;
119 va_start(ap, fmt);
120 fprintf(stderr, "qemu: hardware error: ");
121 vfprintf(stderr, fmt, ap);
122 fprintf(stderr, "\n");
123 CPU_FOREACH(cpu) {
124 fprintf(stderr, "CPU #%d:\n", cpu->cpu_index);
125 cpu_dump_state(cpu, stderr, CPU_DUMP_FPU);
127 va_end(ap);
128 abort();
131 void cpu_synchronize_all_states(void)
133 CPUState *cpu;
135 CPU_FOREACH(cpu) {
136 cpu_synchronize_state(cpu);
140 void cpu_synchronize_all_post_reset(void)
142 CPUState *cpu;
144 CPU_FOREACH(cpu) {
145 cpu_synchronize_post_reset(cpu);
149 void cpu_synchronize_all_post_init(void)
151 CPUState *cpu;
153 CPU_FOREACH(cpu) {
154 cpu_synchronize_post_init(cpu);
158 void cpu_synchronize_all_pre_loadvm(void)
160 CPUState *cpu;
162 CPU_FOREACH(cpu) {
163 cpu_synchronize_pre_loadvm(cpu);
167 void cpu_synchronize_state(CPUState *cpu)
169 if (cpus_accel->synchronize_state) {
170 cpus_accel->synchronize_state(cpu);
174 void cpu_synchronize_post_reset(CPUState *cpu)
176 if (cpus_accel->synchronize_post_reset) {
177 cpus_accel->synchronize_post_reset(cpu);
181 void cpu_synchronize_post_init(CPUState *cpu)
183 if (cpus_accel->synchronize_post_init) {
184 cpus_accel->synchronize_post_init(cpu);
188 void cpu_synchronize_pre_loadvm(CPUState *cpu)
190 if (cpus_accel->synchronize_pre_loadvm) {
191 cpus_accel->synchronize_pre_loadvm(cpu);
195 bool cpus_are_resettable(void)
197 if (cpus_accel->cpus_are_resettable) {
198 return cpus_accel->cpus_are_resettable();
200 return true;
203 int64_t cpus_get_virtual_clock(void)
206 * XXX
208 * need to check that cpus_accel is not NULL, because qcow2 calls
209 * qemu_get_clock_ns(CLOCK_VIRTUAL) without any accel initialized and
210 * with ticks disabled in some io-tests:
211 * 030 040 041 060 099 120 127 140 156 161 172 181 191 192 195 203 229 249 256 267
213 * is this expected?
215 * XXX
217 if (cpus_accel && cpus_accel->get_virtual_clock) {
218 return cpus_accel->get_virtual_clock();
220 return cpu_get_clock();
224 * return the time elapsed in VM between vm_start and vm_stop. Unless
225 * icount is active, cpus_get_elapsed_ticks() uses units of the host CPU cycle
226 * counter.
228 int64_t cpus_get_elapsed_ticks(void)
230 if (cpus_accel->get_elapsed_ticks) {
231 return cpus_accel->get_elapsed_ticks();
233 return cpu_get_ticks();
236 static void generic_handle_interrupt(CPUState *cpu, int mask)
238 cpu->interrupt_request |= mask;
240 if (!qemu_cpu_is_self(cpu)) {
241 qemu_cpu_kick(cpu);
245 void cpu_interrupt(CPUState *cpu, int mask)
247 if (cpus_accel->handle_interrupt) {
248 cpus_accel->handle_interrupt(cpu, mask);
249 } else {
250 generic_handle_interrupt(cpu, mask);
254 static int do_vm_stop(RunState state, bool send_stop)
256 int ret = 0;
258 if (runstate_is_running()) {
259 runstate_set(state);
260 cpu_disable_ticks();
261 pause_all_vcpus();
262 vm_state_notify(0, state);
263 if (send_stop) {
264 qapi_event_send_stop();
268 bdrv_drain_all();
269 ret = bdrv_flush_all();
270 trace_vm_stop_flush_all(ret);
272 return ret;
275 /* Special vm_stop() variant for terminating the process. Historically clients
276 * did not expect a QMP STOP event and so we need to retain compatibility.
278 int vm_shutdown(void)
280 return do_vm_stop(RUN_STATE_SHUTDOWN, false);
283 bool cpu_can_run(CPUState *cpu)
285 if (cpu->stop) {
286 return false;
288 if (cpu_is_stopped(cpu)) {
289 return false;
291 return true;
294 void cpu_handle_guest_debug(CPUState *cpu)
296 if (replay_running_debug()) {
297 if (!cpu->singlestep_enabled) {
299 * Report about the breakpoint and
300 * make a single step to skip it
302 replay_breakpoint();
303 cpu_single_step(cpu, SSTEP_ENABLE);
304 } else {
305 cpu_single_step(cpu, 0);
307 } else {
308 gdb_set_stop_cpu(cpu);
309 qemu_system_debug_request();
310 cpu->stopped = true;
314 #ifdef CONFIG_LINUX
315 static void sigbus_reraise(void)
317 sigset_t set;
318 struct sigaction action;
320 memset(&action, 0, sizeof(action));
321 action.sa_handler = SIG_DFL;
322 if (!sigaction(SIGBUS, &action, NULL)) {
323 raise(SIGBUS);
324 sigemptyset(&set);
325 sigaddset(&set, SIGBUS);
326 pthread_sigmask(SIG_UNBLOCK, &set, NULL);
328 perror("Failed to re-raise SIGBUS!");
329 abort();
332 static void sigbus_handler(int n, siginfo_t *siginfo, void *ctx)
334 if (siginfo->si_code != BUS_MCEERR_AO && siginfo->si_code != BUS_MCEERR_AR) {
335 sigbus_reraise();
338 if (current_cpu) {
339 /* Called asynchronously in VCPU thread. */
340 if (kvm_on_sigbus_vcpu(current_cpu, siginfo->si_code, siginfo->si_addr)) {
341 sigbus_reraise();
343 } else {
344 /* Called synchronously (via signalfd) in main thread. */
345 if (kvm_on_sigbus(siginfo->si_code, siginfo->si_addr)) {
346 sigbus_reraise();
351 static void qemu_init_sigbus(void)
353 struct sigaction action;
356 * ALERT: when modifying this, take care that SIGBUS forwarding in
357 * os_mem_prealloc() will continue working as expected.
359 memset(&action, 0, sizeof(action));
360 action.sa_flags = SA_SIGINFO;
361 action.sa_sigaction = sigbus_handler;
362 sigaction(SIGBUS, &action, NULL);
364 prctl(PR_MCE_KILL, PR_MCE_KILL_SET, PR_MCE_KILL_EARLY, 0, 0);
366 #else /* !CONFIG_LINUX */
367 static void qemu_init_sigbus(void)
370 #endif /* !CONFIG_LINUX */
372 static QemuThread io_thread;
374 /* cpu creation */
375 static QemuCond qemu_cpu_cond;
376 /* system init */
377 static QemuCond qemu_pause_cond;
379 void qemu_init_cpu_loop(void)
381 qemu_init_sigbus();
382 qemu_cond_init(&qemu_cpu_cond);
383 qemu_cond_init(&qemu_pause_cond);
384 qemu_mutex_init(&qemu_global_mutex);
386 qemu_thread_get_self(&io_thread);
389 void run_on_cpu(CPUState *cpu, run_on_cpu_func func, run_on_cpu_data data)
391 do_run_on_cpu(cpu, func, data, &qemu_global_mutex);
394 static void qemu_cpu_stop(CPUState *cpu, bool exit)
396 g_assert(qemu_cpu_is_self(cpu));
397 cpu->stop = false;
398 cpu->stopped = true;
399 if (exit) {
400 cpu_exit(cpu);
402 qemu_cond_broadcast(&qemu_pause_cond);
405 void qemu_wait_io_event_common(CPUState *cpu)
407 qatomic_mb_set(&cpu->thread_kicked, false);
408 if (cpu->stop) {
409 qemu_cpu_stop(cpu, false);
411 process_queued_cpu_work(cpu);
414 void qemu_wait_io_event(CPUState *cpu)
416 bool slept = false;
418 while (cpu_thread_is_idle(cpu)) {
419 if (!slept) {
420 slept = true;
421 qemu_plugin_vcpu_idle_cb(cpu);
423 qemu_cond_wait(cpu->halt_cond, &qemu_global_mutex);
425 if (slept) {
426 qemu_plugin_vcpu_resume_cb(cpu);
429 #ifdef _WIN32
430 /* Eat dummy APC queued by cpus_kick_thread. */
431 if (hax_enabled()) {
432 SleepEx(0, TRUE);
434 #endif
435 qemu_wait_io_event_common(cpu);
438 void cpus_kick_thread(CPUState *cpu)
440 #ifndef _WIN32
441 int err;
443 if (cpu->thread_kicked) {
444 return;
446 cpu->thread_kicked = true;
447 err = pthread_kill(cpu->thread->thread, SIG_IPI);
448 if (err && err != ESRCH) {
449 fprintf(stderr, "qemu:%s: %s", __func__, strerror(err));
450 exit(1);
452 #endif
455 void qemu_cpu_kick(CPUState *cpu)
457 qemu_cond_broadcast(cpu->halt_cond);
458 if (cpus_accel->kick_vcpu_thread) {
459 cpus_accel->kick_vcpu_thread(cpu);
460 } else { /* default */
461 cpus_kick_thread(cpu);
465 void qemu_cpu_kick_self(void)
467 assert(current_cpu);
468 cpus_kick_thread(current_cpu);
471 bool qemu_cpu_is_self(CPUState *cpu)
473 return qemu_thread_is_self(cpu->thread);
476 bool qemu_in_vcpu_thread(void)
478 return current_cpu && qemu_cpu_is_self(current_cpu);
481 QEMU_DEFINE_STATIC_CO_TLS(bool, iothread_locked)
483 bool qemu_mutex_iothread_locked(void)
485 return get_iothread_locked();
488 bool qemu_in_main_thread(void)
490 return qemu_mutex_iothread_locked();
494 * The BQL is taken from so many places that it is worth profiling the
495 * callers directly, instead of funneling them all through a single function.
497 void qemu_mutex_lock_iothread_impl(const char *file, int line)
499 QemuMutexLockFunc bql_lock = qatomic_read(&qemu_bql_mutex_lock_func);
501 g_assert(!qemu_mutex_iothread_locked());
502 bql_lock(&qemu_global_mutex, file, line);
503 set_iothread_locked(true);
506 void qemu_mutex_unlock_iothread(void)
508 g_assert(qemu_mutex_iothread_locked());
509 set_iothread_locked(false);
510 qemu_mutex_unlock(&qemu_global_mutex);
513 void qemu_cond_wait_iothread(QemuCond *cond)
515 qemu_cond_wait(cond, &qemu_global_mutex);
518 void qemu_cond_timedwait_iothread(QemuCond *cond, int ms)
520 qemu_cond_timedwait(cond, &qemu_global_mutex, ms);
523 /* signal CPU creation */
524 void cpu_thread_signal_created(CPUState *cpu)
526 cpu->created = true;
527 qemu_cond_signal(&qemu_cpu_cond);
530 /* signal CPU destruction */
531 void cpu_thread_signal_destroyed(CPUState *cpu)
533 cpu->created = false;
534 qemu_cond_signal(&qemu_cpu_cond);
538 static bool all_vcpus_paused(void)
540 CPUState *cpu;
542 CPU_FOREACH(cpu) {
543 if (!cpu->stopped) {
544 return false;
548 return true;
551 void pause_all_vcpus(void)
553 CPUState *cpu;
555 qemu_clock_enable(QEMU_CLOCK_VIRTUAL, false);
556 CPU_FOREACH(cpu) {
557 if (qemu_cpu_is_self(cpu)) {
558 qemu_cpu_stop(cpu, true);
559 } else {
560 cpu->stop = true;
561 qemu_cpu_kick(cpu);
565 /* We need to drop the replay_lock so any vCPU threads woken up
566 * can finish their replay tasks
568 replay_mutex_unlock();
570 while (!all_vcpus_paused()) {
571 qemu_cond_wait(&qemu_pause_cond, &qemu_global_mutex);
572 CPU_FOREACH(cpu) {
573 qemu_cpu_kick(cpu);
577 qemu_mutex_unlock_iothread();
578 replay_mutex_lock();
579 qemu_mutex_lock_iothread();
582 void cpu_resume(CPUState *cpu)
584 cpu->stop = false;
585 cpu->stopped = false;
586 qemu_cpu_kick(cpu);
589 void resume_all_vcpus(void)
591 CPUState *cpu;
593 if (!runstate_is_running()) {
594 return;
597 qemu_clock_enable(QEMU_CLOCK_VIRTUAL, true);
598 CPU_FOREACH(cpu) {
599 cpu_resume(cpu);
603 void cpu_remove_sync(CPUState *cpu)
605 cpu->stop = true;
606 cpu->unplug = true;
607 qemu_cpu_kick(cpu);
608 qemu_mutex_unlock_iothread();
609 qemu_thread_join(cpu->thread);
610 qemu_mutex_lock_iothread();
613 void cpus_register_accel(const AccelOpsClass *ops)
615 assert(ops != NULL);
616 assert(ops->create_vcpu_thread != NULL); /* mandatory */
617 cpus_accel = ops;
620 void qemu_init_vcpu(CPUState *cpu)
622 MachineState *ms = MACHINE(qdev_get_machine());
624 cpu->nr_cores = ms->smp.cores;
625 cpu->nr_threads = ms->smp.threads;
626 cpu->stopped = true;
627 cpu->random_seed = qemu_guest_random_seed_thread_part1();
629 if (!cpu->as) {
630 /* If the target cpu hasn't set up any address spaces itself,
631 * give it the default one.
633 cpu->num_ases = 1;
634 cpu_address_space_init(cpu, 0, "cpu-memory", cpu->memory);
637 /* accelerators all implement the AccelOpsClass */
638 g_assert(cpus_accel != NULL && cpus_accel->create_vcpu_thread != NULL);
639 cpus_accel->create_vcpu_thread(cpu);
641 while (!cpu->created) {
642 qemu_cond_wait(&qemu_cpu_cond, &qemu_global_mutex);
646 void cpu_stop_current(void)
648 if (current_cpu) {
649 current_cpu->stop = true;
650 cpu_exit(current_cpu);
654 int vm_stop(RunState state)
656 if (qemu_in_vcpu_thread()) {
657 qemu_system_vmstop_request_prepare();
658 qemu_system_vmstop_request(state);
660 * FIXME: should not return to device code in case
661 * vm_stop() has been requested.
663 cpu_stop_current();
664 return 0;
667 return do_vm_stop(state, true);
671 * Prepare for (re)starting the VM.
672 * Returns -1 if the vCPUs are not to be restarted (e.g. if they are already
673 * running or in case of an error condition), 0 otherwise.
675 int vm_prepare_start(bool step_pending)
677 RunState requested;
679 qemu_vmstop_requested(&requested);
680 if (runstate_is_running() && requested == RUN_STATE__MAX) {
681 return -1;
684 /* Ensure that a STOP/RESUME pair of events is emitted if a
685 * vmstop request was pending. The BLOCK_IO_ERROR event, for
686 * example, according to documentation is always followed by
687 * the STOP event.
689 if (runstate_is_running()) {
690 qapi_event_send_stop();
691 qapi_event_send_resume();
692 return -1;
696 * WHPX accelerator needs to know whether we are going to step
697 * any CPUs, before starting the first one.
699 if (cpus_accel->synchronize_pre_resume) {
700 cpus_accel->synchronize_pre_resume(step_pending);
703 /* We are sending this now, but the CPUs will be resumed shortly later */
704 qapi_event_send_resume();
706 cpu_enable_ticks();
707 runstate_set(RUN_STATE_RUNNING);
708 vm_state_notify(1, RUN_STATE_RUNNING);
709 return 0;
712 void vm_start(void)
714 if (!vm_prepare_start(false)) {
715 resume_all_vcpus();
719 /* does a state transition even if the VM is already stopped,
720 current state is forgotten forever */
721 int vm_stop_force_state(RunState state)
723 if (runstate_is_running()) {
724 return vm_stop(state);
725 } else {
726 int ret;
727 runstate_set(state);
729 bdrv_drain_all();
730 /* Make sure to return an error if the flush in a previous vm_stop()
731 * failed. */
732 ret = bdrv_flush_all();
733 trace_vm_stop_flush_all(ret);
734 return ret;
738 void qmp_memsave(int64_t addr, int64_t size, const char *filename,
739 bool has_cpu, int64_t cpu_index, Error **errp)
741 FILE *f;
742 uint32_t l;
743 CPUState *cpu;
744 uint8_t buf[1024];
745 int64_t orig_addr = addr, orig_size = size;
747 if (!has_cpu) {
748 cpu_index = 0;
751 cpu = qemu_get_cpu(cpu_index);
752 if (cpu == NULL) {
753 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "cpu-index",
754 "a CPU number");
755 return;
758 f = fopen(filename, "wb");
759 if (!f) {
760 error_setg_file_open(errp, errno, filename);
761 return;
764 while (size != 0) {
765 l = sizeof(buf);
766 if (l > size)
767 l = size;
768 if (cpu_memory_rw_debug(cpu, addr, buf, l, 0) != 0) {
769 error_setg(errp, "Invalid addr 0x%016" PRIx64 "/size %" PRId64
770 " specified", orig_addr, orig_size);
771 goto exit;
773 if (fwrite(buf, 1, l, f) != l) {
774 error_setg(errp, QERR_IO_ERROR);
775 goto exit;
777 addr += l;
778 size -= l;
781 exit:
782 fclose(f);
785 void qmp_pmemsave(int64_t addr, int64_t size, const char *filename,
786 Error **errp)
788 FILE *f;
789 uint32_t l;
790 uint8_t buf[1024];
792 f = fopen(filename, "wb");
793 if (!f) {
794 error_setg_file_open(errp, errno, filename);
795 return;
798 while (size != 0) {
799 l = sizeof(buf);
800 if (l > size)
801 l = size;
802 cpu_physical_memory_read(addr, buf, l);
803 if (fwrite(buf, 1, l, f) != l) {
804 error_setg(errp, QERR_IO_ERROR);
805 goto exit;
807 addr += l;
808 size -= l;
811 exit:
812 fclose(f);
815 void qmp_inject_nmi(Error **errp)
817 nmi_monitor_handle(monitor_get_cpu_index(monitor_cur()), errp);