xref: /openbmc/qemu/system/runstate.c (revision 1136309df5e3de7d031a137221318f48784cca5d)
1 /*
2  * QEMU main system emulation loop
3  *
4  * Copyright (c) 2003-2020 QEMU contributors
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 "audio/audio.h"
27 #include "block/block.h"
28 #include "block/export.h"
29 #include "chardev/char.h"
30 #include "crypto/cipher.h"
31 #include "crypto/init.h"
32 #include "exec/cpu-common.h"
33 #include "gdbstub/syscalls.h"
34 #include "hw/boards.h"
35 #include "hw/resettable.h"
36 #include "migration/misc.h"
37 #include "migration/postcopy-ram.h"
38 #include "monitor/monitor.h"
39 #include "net/net.h"
40 #include "net/vhost_net.h"
41 #include "qapi/error.h"
42 #include "qapi/qapi-commands-run-state.h"
43 #include "qapi/qapi-events-run-state.h"
44 #include "qemu/accel.h"
45 #include "qemu/error-report.h"
46 #include "qemu/job.h"
47 #include "qemu/log.h"
48 #include "qemu/module.h"
49 #include "qemu/sockets.h"
50 #include "qemu/timer.h"
51 #include "qemu/thread.h"
52 #include "qom/object.h"
53 #include "qom/object_interfaces.h"
54 #include "system/cpus.h"
55 #include "system/qtest.h"
56 #include "system/replay.h"
57 #include "system/reset.h"
58 #include "system/runstate.h"
59 #include "system/runstate-action.h"
60 #include "system/system.h"
61 #include "system/tpm.h"
62 #include "trace.h"
63 
64 static NotifierList exit_notifiers =
65     NOTIFIER_LIST_INITIALIZER(exit_notifiers);
66 
67 static RunState current_run_state = RUN_STATE_PRELAUNCH;
68 
69 /* We use RUN_STATE__MAX but any invalid value will do */
70 static RunState vmstop_requested = RUN_STATE__MAX;
71 static QemuMutex vmstop_lock;
72 
73 typedef struct {
74     RunState from;
75     RunState to;
76 } RunStateTransition;
77 
78 static const RunStateTransition runstate_transitions_def[] = {
79     { RUN_STATE_PRELAUNCH, RUN_STATE_INMIGRATE },
80     { RUN_STATE_PRELAUNCH, RUN_STATE_SUSPENDED },
81 
82     { RUN_STATE_DEBUG, RUN_STATE_RUNNING },
83     { RUN_STATE_DEBUG, RUN_STATE_FINISH_MIGRATE },
84     { RUN_STATE_DEBUG, RUN_STATE_PRELAUNCH },
85 
86     { RUN_STATE_INMIGRATE, RUN_STATE_INTERNAL_ERROR },
87     { RUN_STATE_INMIGRATE, RUN_STATE_IO_ERROR },
88     { RUN_STATE_INMIGRATE, RUN_STATE_PAUSED },
89     { RUN_STATE_INMIGRATE, RUN_STATE_RUNNING },
90     { RUN_STATE_INMIGRATE, RUN_STATE_SHUTDOWN },
91     { RUN_STATE_INMIGRATE, RUN_STATE_SUSPENDED },
92     { RUN_STATE_INMIGRATE, RUN_STATE_WATCHDOG },
93     { RUN_STATE_INMIGRATE, RUN_STATE_GUEST_PANICKED },
94     { RUN_STATE_INMIGRATE, RUN_STATE_FINISH_MIGRATE },
95     { RUN_STATE_INMIGRATE, RUN_STATE_PRELAUNCH },
96     { RUN_STATE_INMIGRATE, RUN_STATE_POSTMIGRATE },
97     { RUN_STATE_INMIGRATE, RUN_STATE_COLO },
98 
99     { RUN_STATE_INTERNAL_ERROR, RUN_STATE_PAUSED },
100     { RUN_STATE_INTERNAL_ERROR, RUN_STATE_FINISH_MIGRATE },
101     { RUN_STATE_INTERNAL_ERROR, RUN_STATE_PRELAUNCH },
102 
103     { RUN_STATE_IO_ERROR, RUN_STATE_RUNNING },
104     { RUN_STATE_IO_ERROR, RUN_STATE_FINISH_MIGRATE },
105     { RUN_STATE_IO_ERROR, RUN_STATE_PRELAUNCH },
106 
107     { RUN_STATE_PAUSED, RUN_STATE_RUNNING },
108     { RUN_STATE_PAUSED, RUN_STATE_FINISH_MIGRATE },
109     { RUN_STATE_PAUSED, RUN_STATE_POSTMIGRATE },
110     { RUN_STATE_PAUSED, RUN_STATE_PRELAUNCH },
111     { RUN_STATE_PAUSED, RUN_STATE_COLO},
112     { RUN_STATE_PAUSED, RUN_STATE_SUSPENDED},
113 
114     { RUN_STATE_POSTMIGRATE, RUN_STATE_RUNNING },
115     { RUN_STATE_POSTMIGRATE, RUN_STATE_FINISH_MIGRATE },
116     { RUN_STATE_POSTMIGRATE, RUN_STATE_PRELAUNCH },
117 
118     { RUN_STATE_PRELAUNCH, RUN_STATE_RUNNING },
119     { RUN_STATE_PRELAUNCH, RUN_STATE_FINISH_MIGRATE },
120     { RUN_STATE_PRELAUNCH, RUN_STATE_INMIGRATE },
121 
122     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_RUNNING },
123     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_PAUSED },
124     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_POSTMIGRATE },
125     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_PRELAUNCH },
126     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_COLO },
127     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_INTERNAL_ERROR },
128     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_IO_ERROR },
129     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_SHUTDOWN },
130     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_SUSPENDED },
131     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_WATCHDOG },
132     { RUN_STATE_FINISH_MIGRATE, RUN_STATE_GUEST_PANICKED },
133 
134     { RUN_STATE_RESTORE_VM, RUN_STATE_RUNNING },
135     { RUN_STATE_RESTORE_VM, RUN_STATE_PRELAUNCH },
136     { RUN_STATE_RESTORE_VM, RUN_STATE_SUSPENDED },
137 
138     { RUN_STATE_COLO, RUN_STATE_RUNNING },
139     { RUN_STATE_COLO, RUN_STATE_PRELAUNCH },
140     { RUN_STATE_COLO, RUN_STATE_SHUTDOWN},
141 
142     { RUN_STATE_RUNNING, RUN_STATE_DEBUG },
143     { RUN_STATE_RUNNING, RUN_STATE_INTERNAL_ERROR },
144     { RUN_STATE_RUNNING, RUN_STATE_IO_ERROR },
145     { RUN_STATE_RUNNING, RUN_STATE_PAUSED },
146     { RUN_STATE_RUNNING, RUN_STATE_FINISH_MIGRATE },
147     { RUN_STATE_RUNNING, RUN_STATE_RESTORE_VM },
148     { RUN_STATE_RUNNING, RUN_STATE_SAVE_VM },
149     { RUN_STATE_RUNNING, RUN_STATE_SHUTDOWN },
150     { RUN_STATE_RUNNING, RUN_STATE_WATCHDOG },
151     { RUN_STATE_RUNNING, RUN_STATE_GUEST_PANICKED },
152     { RUN_STATE_RUNNING, RUN_STATE_COLO},
153 
154     { RUN_STATE_SAVE_VM, RUN_STATE_RUNNING },
155     { RUN_STATE_SAVE_VM, RUN_STATE_SUSPENDED },
156 
157     { RUN_STATE_SHUTDOWN, RUN_STATE_PAUSED },
158     { RUN_STATE_SHUTDOWN, RUN_STATE_FINISH_MIGRATE },
159     { RUN_STATE_SHUTDOWN, RUN_STATE_PRELAUNCH },
160     { RUN_STATE_SHUTDOWN, RUN_STATE_COLO },
161 
162     { RUN_STATE_DEBUG, RUN_STATE_SUSPENDED },
163     { RUN_STATE_RUNNING, RUN_STATE_SUSPENDED },
164     { RUN_STATE_SUSPENDED, RUN_STATE_RUNNING },
165     { RUN_STATE_SUSPENDED, RUN_STATE_FINISH_MIGRATE },
166     { RUN_STATE_SUSPENDED, RUN_STATE_PRELAUNCH },
167     { RUN_STATE_SUSPENDED, RUN_STATE_COLO},
168     { RUN_STATE_SUSPENDED, RUN_STATE_PAUSED},
169     { RUN_STATE_SUSPENDED, RUN_STATE_SAVE_VM },
170     { RUN_STATE_SUSPENDED, RUN_STATE_RESTORE_VM },
171     { RUN_STATE_SUSPENDED, RUN_STATE_SHUTDOWN },
172 
173     { RUN_STATE_WATCHDOG, RUN_STATE_RUNNING },
174     { RUN_STATE_WATCHDOG, RUN_STATE_FINISH_MIGRATE },
175     { RUN_STATE_WATCHDOG, RUN_STATE_PRELAUNCH },
176     { RUN_STATE_WATCHDOG, RUN_STATE_COLO},
177 
178     { RUN_STATE_GUEST_PANICKED, RUN_STATE_RUNNING },
179     { RUN_STATE_GUEST_PANICKED, RUN_STATE_FINISH_MIGRATE },
180     { RUN_STATE_GUEST_PANICKED, RUN_STATE_PRELAUNCH },
181 
182     { RUN_STATE__MAX, RUN_STATE__MAX },
183 };
184 
185 static const RunStateTransition replay_play_runstate_transitions_def[] = {
186     { RUN_STATE_SHUTDOWN, RUN_STATE_RUNNING},
187 
188     { RUN_STATE__MAX, RUN_STATE__MAX },
189 };
190 
191 static bool runstate_valid_transitions[RUN_STATE__MAX][RUN_STATE__MAX];
192 
193 bool runstate_check(RunState state)
194 {
195     return current_run_state == state;
196 }
197 
198 static void transitions_set_valid(const RunStateTransition *rst)
199 {
200     const RunStateTransition *p;
201 
202     for (p = rst; p->from != RUN_STATE__MAX; p++) {
203         runstate_valid_transitions[p->from][p->to] = true;
204     }
205 }
206 
207 void runstate_replay_enable(void)
208 {
209     assert(replay_mode != REPLAY_MODE_NONE);
210 
211     if (replay_mode == REPLAY_MODE_PLAY) {
212         /*
213          * When reverse-debugging, it is possible to move state from
214          * shutdown to running.
215          */
216         transitions_set_valid(&replay_play_runstate_transitions_def[0]);
217     }
218 }
219 
220 static void runstate_init(void)
221 {
222     memset(&runstate_valid_transitions, 0, sizeof(runstate_valid_transitions));
223 
224     transitions_set_valid(&runstate_transitions_def[0]);
225 
226     qemu_mutex_init(&vmstop_lock);
227 }
228 
229 /* This function will abort() on invalid state transitions */
230 void runstate_set(RunState new_state)
231 {
232     assert(new_state < RUN_STATE__MAX);
233 
234     trace_runstate_set(current_run_state, RunState_str(current_run_state),
235                        new_state, RunState_str(new_state));
236 
237     if (current_run_state == new_state) {
238         return;
239     }
240 
241     if (!runstate_valid_transitions[current_run_state][new_state]) {
242         error_report("invalid runstate transition: '%s' -> '%s'",
243                      RunState_str(current_run_state),
244                      RunState_str(new_state));
245         abort();
246     }
247 
248     current_run_state = new_state;
249 }
250 
251 RunState runstate_get(void)
252 {
253     return current_run_state;
254 }
255 
256 bool runstate_is_running(void)
257 {
258     return runstate_check(RUN_STATE_RUNNING);
259 }
260 
261 bool runstate_needs_reset(void)
262 {
263     return runstate_check(RUN_STATE_INTERNAL_ERROR) ||
264         runstate_check(RUN_STATE_SHUTDOWN);
265 }
266 
267 StatusInfo *qmp_query_status(Error **errp)
268 {
269     StatusInfo *info = g_malloc0(sizeof(*info));
270 
271     info->running = runstate_is_running();
272     info->status = current_run_state;
273 
274     return info;
275 }
276 
277 bool qemu_vmstop_requested(RunState *r)
278 {
279     qemu_mutex_lock(&vmstop_lock);
280     *r = vmstop_requested;
281     vmstop_requested = RUN_STATE__MAX;
282     qemu_mutex_unlock(&vmstop_lock);
283     return *r < RUN_STATE__MAX;
284 }
285 
286 void qemu_system_vmstop_request_prepare(void)
287 {
288     qemu_mutex_lock(&vmstop_lock);
289 }
290 
291 void qemu_system_vmstop_request(RunState state)
292 {
293     vmstop_requested = state;
294     qemu_mutex_unlock(&vmstop_lock);
295     qemu_notify_event();
296 }
297 struct VMChangeStateEntry {
298     VMChangeStateHandler *cb;
299     VMChangeStateHandler *prepare_cb;
300     VMChangeStateHandlerWithRet *cb_ret;
301     void *opaque;
302     QTAILQ_ENTRY(VMChangeStateEntry) entries;
303     int priority;
304 };
305 
306 static QTAILQ_HEAD(, VMChangeStateEntry) vm_change_state_head =
307     QTAILQ_HEAD_INITIALIZER(vm_change_state_head);
308 
309 /**
310  * qemu_add_vm_change_state_handler_prio:
311  * @cb: the callback to invoke
312  * @opaque: user data passed to the callback
313  * @priority: low priorities execute first when the vm runs and the reverse is
314  *            true when the vm stops
315  *
316  * Register a callback function that is invoked when the vm starts or stops
317  * running.
318  *
319  * Returns: an entry to be freed using qemu_del_vm_change_state_handler()
320  */
321 VMChangeStateEntry *qemu_add_vm_change_state_handler_prio(
322         VMChangeStateHandler *cb, void *opaque, int priority)
323 {
324     return qemu_add_vm_change_state_handler_prio_full(cb, NULL, NULL,
325                                                       opaque, priority);
326 }
327 
328 /**
329  * qemu_add_vm_change_state_handler_prio_full:
330  * @cb: the main callback to invoke
331  * @prepare_cb: a callback to invoke before the main callback
332  * @cb_ret: the main callback to invoke with return value
333  * @opaque: user data passed to the callbacks
334  * @priority: low priorities execute first when the vm runs and the reverse is
335  *            true when the vm stops
336  *
337  * Register a main callback function and an optional prepare callback function
338  * that are invoked when the vm starts or stops running. The main callback and
339  * the prepare callback are called in two separate phases: First all prepare
340  * callbacks are called and only then all main callbacks are called. As its
341  * name suggests, the prepare callback can be used to do some preparatory work
342  * before invoking the main callback.
343  *
344  * Returns: an entry to be freed using qemu_del_vm_change_state_handler()
345  */
346 VMChangeStateEntry *
347 qemu_add_vm_change_state_handler_prio_full(VMChangeStateHandler *cb,
348                                            VMChangeStateHandler *prepare_cb,
349                                            VMChangeStateHandlerWithRet *cb_ret,
350                                            void *opaque, int priority)
351 {
352     VMChangeStateEntry *e;
353     VMChangeStateEntry *other;
354 
355     e = g_malloc0(sizeof(*e));
356     e->cb = cb;
357     e->prepare_cb = prepare_cb;
358     e->cb_ret = cb_ret;
359     e->opaque = opaque;
360     e->priority = priority;
361 
362     /* Keep list sorted in ascending priority order */
363     QTAILQ_FOREACH(other, &vm_change_state_head, entries) {
364         if (priority < other->priority) {
365             QTAILQ_INSERT_BEFORE(other, e, entries);
366             return e;
367         }
368     }
369 
370     QTAILQ_INSERT_TAIL(&vm_change_state_head, e, entries);
371     return e;
372 }
373 
374 VMChangeStateEntry *qemu_add_vm_change_state_handler(VMChangeStateHandler *cb,
375                                                      void *opaque)
376 {
377     return qemu_add_vm_change_state_handler_prio(cb, opaque, 0);
378 }
379 
380 void qemu_del_vm_change_state_handler(VMChangeStateEntry *e)
381 {
382     QTAILQ_REMOVE(&vm_change_state_head, e, entries);
383     g_free(e);
384 }
385 
386 int vm_state_notify(bool running, RunState state)
387 {
388     VMChangeStateEntry *e, *next;
389     int ret = 0;
390 
391     trace_vm_state_notify(running, state, RunState_str(state));
392 
393     if (running) {
394         QTAILQ_FOREACH_SAFE(e, &vm_change_state_head, entries, next) {
395             if (e->prepare_cb) {
396                 e->prepare_cb(e->opaque, running, state);
397             }
398         }
399 
400         QTAILQ_FOREACH_SAFE(e, &vm_change_state_head, entries, next) {
401             if (e->cb) {
402                 e->cb(e->opaque, running, state);
403             } else if (e->cb_ret) {
404                 /*
405                  * Here ignore the return value of cb_ret because
406                  * we only care about the stopping the device during
407                  * the VM live migration to indicate whether the
408                  * connection between qemu and backend is normal.
409                  */
410                 e->cb_ret(e->opaque, running, state);
411             }
412         }
413     } else {
414         QTAILQ_FOREACH_REVERSE_SAFE(e, &vm_change_state_head, entries, next) {
415             if (e->prepare_cb) {
416                 e->prepare_cb(e->opaque, running, state);
417             }
418         }
419 
420         QTAILQ_FOREACH_REVERSE_SAFE(e, &vm_change_state_head, entries, next) {
421             if (e->cb) {
422                 e->cb(e->opaque, running, state);
423             } else if (e->cb_ret) {
424                 /*
425                  * We should execute all registered callbacks even if
426                  * one of them returns failure, otherwise, some cleanup
427                  * work of the device will be skipped.
428                  */
429                 ret |= e->cb_ret(e->opaque, running, state);
430             }
431         }
432     }
433     return ret;
434 }
435 
436 static ShutdownCause reset_requested;
437 static ShutdownCause shutdown_requested;
438 static int shutdown_exit_code = EXIT_SUCCESS;
439 static int shutdown_signal;
440 static bool force_shutdown;
441 static pid_t shutdown_pid;
442 static int powerdown_requested;
443 static int debug_requested;
444 static int suspend_requested;
445 static WakeupReason wakeup_reason;
446 static NotifierList powerdown_notifiers =
447     NOTIFIER_LIST_INITIALIZER(powerdown_notifiers);
448 static NotifierList suspend_notifiers =
449     NOTIFIER_LIST_INITIALIZER(suspend_notifiers);
450 static NotifierList wakeup_notifiers =
451     NOTIFIER_LIST_INITIALIZER(wakeup_notifiers);
452 static NotifierList shutdown_notifiers =
453     NOTIFIER_LIST_INITIALIZER(shutdown_notifiers);
454 static uint32_t wakeup_reason_mask = ~(1 << QEMU_WAKEUP_REASON_NONE);
455 
456 ShutdownCause qemu_shutdown_requested_get(void)
457 {
458     return shutdown_requested;
459 }
460 
461 bool qemu_force_shutdown_requested(void)
462 {
463     return force_shutdown;
464 }
465 
466 ShutdownCause qemu_reset_requested_get(void)
467 {
468     return reset_requested;
469 }
470 
471 static int qemu_shutdown_requested(void)
472 {
473     return qatomic_xchg(&shutdown_requested, SHUTDOWN_CAUSE_NONE);
474 }
475 
476 static void qemu_kill_report(void)
477 {
478     if (!qtest_driver() && shutdown_signal) {
479         if (shutdown_pid == 0) {
480             /* This happens for eg ^C at the terminal, so it's worth
481              * avoiding printing an odd message in that case.
482              */
483             error_report("terminating on signal %d", shutdown_signal);
484         } else {
485             char *shutdown_cmd = qemu_get_pid_name(shutdown_pid);
486 
487             error_report("terminating on signal %d from pid " FMT_pid " (%s)",
488                          shutdown_signal, shutdown_pid,
489                          shutdown_cmd ? shutdown_cmd : "<unknown process>");
490             g_free(shutdown_cmd);
491         }
492         shutdown_signal = 0;
493     }
494 }
495 
496 static ShutdownCause qemu_reset_requested(void)
497 {
498     ShutdownCause r = reset_requested;
499 
500     if (r && replay_checkpoint(CHECKPOINT_RESET_REQUESTED)) {
501         reset_requested = SHUTDOWN_CAUSE_NONE;
502         return r;
503     }
504     return SHUTDOWN_CAUSE_NONE;
505 }
506 
507 static int qemu_suspend_requested(void)
508 {
509     int r = suspend_requested;
510     if (r && replay_checkpoint(CHECKPOINT_SUSPEND_REQUESTED)) {
511         suspend_requested = 0;
512         return r;
513     }
514     return false;
515 }
516 
517 static WakeupReason qemu_wakeup_requested(void)
518 {
519     return wakeup_reason;
520 }
521 
522 static int qemu_powerdown_requested(void)
523 {
524     int r = powerdown_requested;
525     powerdown_requested = 0;
526     return r;
527 }
528 
529 static int qemu_debug_requested(void)
530 {
531     int r = debug_requested;
532     debug_requested = 0;
533     return r;
534 }
535 
536 /*
537  * Reset the VM. Issue an event unless @reason is SHUTDOWN_CAUSE_NONE.
538  */
539 void qemu_system_reset(ShutdownCause reason)
540 {
541     MachineClass *mc;
542     ResetType type;
543 
544     mc = current_machine ? MACHINE_GET_CLASS(current_machine) : NULL;
545 
546     cpu_synchronize_all_states();
547 
548     switch (reason) {
549     case SHUTDOWN_CAUSE_SNAPSHOT_LOAD:
550         type = RESET_TYPE_SNAPSHOT_LOAD;
551         break;
552     default:
553         type = RESET_TYPE_COLD;
554     }
555     if (mc && mc->reset) {
556         mc->reset(current_machine, type);
557     } else {
558         qemu_devices_reset(type);
559     }
560     switch (reason) {
561     case SHUTDOWN_CAUSE_NONE:
562     case SHUTDOWN_CAUSE_SUBSYSTEM_RESET:
563     case SHUTDOWN_CAUSE_SNAPSHOT_LOAD:
564         break;
565     default:
566         qapi_event_send_reset(shutdown_caused_by_guest(reason), reason);
567     }
568 
569     /*
570      * Some boards use the machine reset callback to point CPUs to the firmware
571      * entry point.  Assume that this is not the case for boards that support
572      * non-resettable CPUs (currently used only for confidential guests), in
573      * which case cpu_synchronize_all_post_init() is enough because
574      * it does _more_  than cpu_synchronize_all_post_reset().
575      */
576     if (cpus_are_resettable()) {
577         cpu_synchronize_all_post_reset();
578     } else {
579         assert(runstate_check(RUN_STATE_PRELAUNCH));
580     }
581 
582     vm_set_suspended(false);
583 }
584 
585 /*
586  * Wake the VM after suspend.
587  */
588 static void qemu_system_wakeup(void)
589 {
590     MachineClass *mc;
591 
592     mc = current_machine ? MACHINE_GET_CLASS(current_machine) : NULL;
593 
594     if (mc && mc->wakeup) {
595         mc->wakeup(current_machine);
596     }
597 }
598 
599 static char *tdx_parse_panic_message(char *message)
600 {
601     bool printable = false;
602     char *buf = NULL;
603     int len = 0, i;
604 
605     /*
606      * Although message is defined as a json string, we shouldn't
607      * unconditionally treat it as is because the guest generated it and
608      * it's not necessarily trustable.
609      */
610     if (message) {
611         /* The caller guarantees the NULL-terminated string. */
612         len = strlen(message);
613 
614         printable = len > 0;
615         for (i = 0; i < len; i++) {
616             if (!(0x20 <= message[i] && message[i] <= 0x7e)) {
617                 printable = false;
618                 break;
619             }
620         }
621     }
622 
623     if (len == 0) {
624         buf = g_malloc(1);
625         buf[0] = '\0';
626     } else {
627         if (!printable) {
628             /* 3 = length of "%02x " */
629             buf = g_malloc(len * 3);
630             for (i = 0; i < len; i++) {
631                 if (message[i] == '\0') {
632                     break;
633                 } else {
634                     sprintf(buf + 3 * i, "%02x ", message[i]);
635                 }
636             }
637             if (i > 0) {
638                 /* replace the last ' '(space) to NULL */
639                 buf[i * 3 - 1] = '\0';
640             } else {
641                 buf[0] = '\0';
642             }
643         } else {
644             buf = g_strdup(message);
645         }
646     }
647 
648     return buf;
649 }
650 
651 void qemu_system_guest_panicked(GuestPanicInformation *info)
652 {
653     qemu_log_mask(LOG_GUEST_ERROR, "Guest crashed");
654 
655     if (current_cpu) {
656         current_cpu->crash_occurred = true;
657     }
658     /*
659      * TODO:  Currently the available panic actions are: none, pause, and
660      * shutdown, but in principle debug and reset could be supported as well.
661      * Investigate any potential use cases for the unimplemented actions.
662      */
663     if (panic_action == PANIC_ACTION_PAUSE
664         || (panic_action == PANIC_ACTION_SHUTDOWN && shutdown_action == SHUTDOWN_ACTION_PAUSE)) {
665         qapi_event_send_guest_panicked(GUEST_PANIC_ACTION_PAUSE, info);
666         vm_stop(RUN_STATE_GUEST_PANICKED);
667     } else if (panic_action == PANIC_ACTION_SHUTDOWN ||
668                panic_action == PANIC_ACTION_EXIT_FAILURE) {
669         qapi_event_send_guest_panicked(GUEST_PANIC_ACTION_POWEROFF, info);
670         vm_stop(RUN_STATE_GUEST_PANICKED);
671         qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_PANIC);
672     } else {
673         qapi_event_send_guest_panicked(GUEST_PANIC_ACTION_RUN, info);
674     }
675 
676     if (info) {
677         if (info->type == GUEST_PANIC_INFORMATION_TYPE_HYPER_V) {
678             qemu_log_mask(LOG_GUEST_ERROR, "\nHV crash parameters: (%#"PRIx64
679                           " %#"PRIx64" %#"PRIx64" %#"PRIx64" %#"PRIx64")\n",
680                           info->u.hyper_v.arg1,
681                           info->u.hyper_v.arg2,
682                           info->u.hyper_v.arg3,
683                           info->u.hyper_v.arg4,
684                           info->u.hyper_v.arg5);
685         } else if (info->type == GUEST_PANIC_INFORMATION_TYPE_S390) {
686             qemu_log_mask(LOG_GUEST_ERROR, " on cpu %d: %s\n"
687                           "PSW: 0x%016" PRIx64 " 0x%016" PRIx64"\n",
688                           info->u.s390.core,
689                           S390CrashReason_str(info->u.s390.reason),
690                           info->u.s390.psw_mask,
691                           info->u.s390.psw_addr);
692         } else if (info->type == GUEST_PANIC_INFORMATION_TYPE_TDX) {
693             char *message = tdx_parse_panic_message(info->u.tdx.message);
694             qemu_log_mask(LOG_GUEST_ERROR,
695                           "\nTDX guest reports fatal error."
696                           " error code: 0x%" PRIx32 " error message:\"%s\"\n",
697                           info->u.tdx.error_code, message);
698             g_free(message);
699             if (info->u.tdx.gpa != -1ull) {
700                 qemu_log_mask(LOG_GUEST_ERROR, "Additional error information "
701                               "can be found at gpa page: 0x%" PRIx64 "\n",
702                               info->u.tdx.gpa);
703             }
704         }
705 
706         qapi_free_GuestPanicInformation(info);
707     }
708 }
709 
710 void qemu_system_guest_crashloaded(GuestPanicInformation *info)
711 {
712     qemu_log_mask(LOG_GUEST_ERROR, "Guest crash loaded");
713     qapi_event_send_guest_crashloaded(GUEST_PANIC_ACTION_RUN, info);
714     qapi_free_GuestPanicInformation(info);
715 }
716 
717 void qemu_system_guest_pvshutdown(void)
718 {
719     qapi_event_send_guest_pvshutdown();
720     qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);
721 }
722 
723 void qemu_system_reset_request(ShutdownCause reason)
724 {
725     if (reboot_action == REBOOT_ACTION_SHUTDOWN &&
726         reason != SHUTDOWN_CAUSE_SUBSYSTEM_RESET) {
727         shutdown_requested = reason;
728     } else if (!cpus_are_resettable()) {
729         error_report("cpus are not resettable, terminating");
730         shutdown_requested = reason;
731     } else {
732         reset_requested = reason;
733     }
734     cpu_stop_current();
735     qemu_notify_event();
736 }
737 
738 static void qemu_system_suspend(void)
739 {
740     pause_all_vcpus();
741     notifier_list_notify(&suspend_notifiers, NULL);
742     runstate_set(RUN_STATE_SUSPENDED);
743     qapi_event_send_suspend();
744 }
745 
746 void qemu_system_suspend_request(void)
747 {
748     if (runstate_check(RUN_STATE_SUSPENDED)) {
749         return;
750     }
751     suspend_requested = 1;
752     cpu_stop_current();
753     qemu_notify_event();
754 }
755 
756 void qemu_register_suspend_notifier(Notifier *notifier)
757 {
758     notifier_list_add(&suspend_notifiers, notifier);
759 }
760 
761 void qemu_system_wakeup_request(WakeupReason reason, Error **errp)
762 {
763     trace_system_wakeup_request(reason);
764 
765     if (!runstate_check(RUN_STATE_SUSPENDED)) {
766         error_setg(errp,
767                    "Unable to wake up: guest is not in suspended state");
768         return;
769     }
770     if (!(wakeup_reason_mask & (1 << reason))) {
771         return;
772     }
773     runstate_set(RUN_STATE_RUNNING);
774     wakeup_reason = reason;
775     qemu_notify_event();
776 }
777 
778 void qemu_system_wakeup_enable(WakeupReason reason, bool enabled)
779 {
780     if (enabled) {
781         wakeup_reason_mask |= (1 << reason);
782     } else {
783         wakeup_reason_mask &= ~(1 << reason);
784     }
785 }
786 
787 void qemu_register_wakeup_notifier(Notifier *notifier)
788 {
789     notifier_list_add(&wakeup_notifiers, notifier);
790 }
791 
792 static bool wakeup_suspend_enabled;
793 
794 void qemu_register_wakeup_support(void)
795 {
796     wakeup_suspend_enabled = true;
797 }
798 
799 bool qemu_wakeup_suspend_enabled(void)
800 {
801     return wakeup_suspend_enabled;
802 }
803 
804 void qemu_system_killed(int signal, pid_t pid)
805 {
806     shutdown_signal = signal;
807     shutdown_pid = pid;
808     shutdown_action = SHUTDOWN_ACTION_POWEROFF;
809 
810     /* Cannot call qemu_system_shutdown_request directly because
811      * we are in a signal handler.
812      */
813     shutdown_requested = SHUTDOWN_CAUSE_HOST_SIGNAL;
814     force_shutdown = true;
815     qemu_notify_event();
816 }
817 
818 void qemu_system_shutdown_request_with_code(ShutdownCause reason,
819                                             int exit_code)
820 {
821     shutdown_exit_code = exit_code;
822     qemu_system_shutdown_request(reason);
823 }
824 
825 void qemu_system_shutdown_request(ShutdownCause reason)
826 {
827     trace_qemu_system_shutdown_request(reason);
828     replay_shutdown_request(reason);
829     shutdown_requested = reason;
830     if (reason == SHUTDOWN_CAUSE_HOST_QMP_QUIT) {
831         force_shutdown = true;
832     }
833     qemu_notify_event();
834 }
835 
836 static void qemu_system_powerdown(void)
837 {
838     qapi_event_send_powerdown();
839     notifier_list_notify(&powerdown_notifiers, NULL);
840 }
841 
842 static void qemu_system_shutdown(ShutdownCause cause)
843 {
844     qapi_event_send_shutdown(shutdown_caused_by_guest(cause), cause);
845     notifier_list_notify(&shutdown_notifiers, &cause);
846 }
847 
848 void qemu_system_powerdown_request(void)
849 {
850     trace_qemu_system_powerdown_request();
851     powerdown_requested = 1;
852     qemu_notify_event();
853 }
854 
855 void qemu_register_powerdown_notifier(Notifier *notifier)
856 {
857     notifier_list_add(&powerdown_notifiers, notifier);
858 }
859 
860 void qemu_register_shutdown_notifier(Notifier *notifier)
861 {
862     notifier_list_add(&shutdown_notifiers, notifier);
863 }
864 
865 void qemu_system_debug_request(void)
866 {
867     debug_requested = 1;
868     qemu_notify_event();
869 }
870 
871 static bool main_loop_should_exit(int *status)
872 {
873     RunState r;
874     ShutdownCause request;
875 
876     if (qemu_debug_requested()) {
877         vm_stop(RUN_STATE_DEBUG);
878     }
879     if (qemu_suspend_requested()) {
880         qemu_system_suspend();
881     }
882     request = qemu_shutdown_requested();
883     if (request) {
884         qemu_kill_report();
885         qemu_system_shutdown(request);
886         if (shutdown_action == SHUTDOWN_ACTION_PAUSE) {
887             vm_stop(RUN_STATE_SHUTDOWN);
888         } else {
889             if (shutdown_exit_code != EXIT_SUCCESS) {
890                 *status = shutdown_exit_code;
891             } else if (request == SHUTDOWN_CAUSE_GUEST_PANIC &&
892                 panic_action == PANIC_ACTION_EXIT_FAILURE) {
893                 *status = EXIT_FAILURE;
894             }
895             return true;
896         }
897     }
898     request = qemu_reset_requested();
899     if (request) {
900         pause_all_vcpus();
901         qemu_system_reset(request);
902         resume_all_vcpus();
903         /*
904          * runstate can change in pause_all_vcpus()
905          * as iothread mutex is unlocked
906          */
907         if (!runstate_check(RUN_STATE_RUNNING) &&
908                 !runstate_check(RUN_STATE_INMIGRATE) &&
909                 !runstate_check(RUN_STATE_FINISH_MIGRATE)) {
910             runstate_set(RUN_STATE_PRELAUNCH);
911         }
912     }
913     if (qemu_wakeup_requested()) {
914         pause_all_vcpus();
915         qemu_system_wakeup();
916         notifier_list_notify(&wakeup_notifiers, &wakeup_reason);
917         wakeup_reason = QEMU_WAKEUP_REASON_NONE;
918         resume_all_vcpus();
919         qapi_event_send_wakeup();
920     }
921     if (qemu_powerdown_requested()) {
922         qemu_system_powerdown();
923     }
924     if (qemu_vmstop_requested(&r)) {
925         vm_stop(r);
926     }
927     return false;
928 }
929 
930 int qemu_main_loop(void)
931 {
932     int status = EXIT_SUCCESS;
933 
934     while (!main_loop_should_exit(&status)) {
935         main_loop_wait(false);
936     }
937 
938     return status;
939 }
940 
941 void qemu_add_exit_notifier(Notifier *notify)
942 {
943     notifier_list_add(&exit_notifiers, notify);
944 }
945 
946 void qemu_remove_exit_notifier(Notifier *notify)
947 {
948     notifier_remove(notify);
949 }
950 
951 static void qemu_run_exit_notifiers(void)
952 {
953     BQL_LOCK_GUARD();
954     notifier_list_notify(&exit_notifiers, NULL);
955 }
956 
957 void qemu_init_subsystems(void)
958 {
959     Error *err = NULL;
960 
961     os_set_line_buffering();
962 
963     module_call_init(MODULE_INIT_TRACE);
964 
965     qemu_init_cpu_list();
966     qemu_init_cpu_loop();
967     bql_lock();
968 
969     atexit(qemu_run_exit_notifiers);
970 
971     module_call_init(MODULE_INIT_QOM);
972     module_call_init(MODULE_INIT_MIGRATION);
973 
974     runstate_init();
975     precopy_infrastructure_init();
976     postcopy_infrastructure_init();
977     monitor_init_globals();
978 
979     if (qcrypto_init(&err) < 0) {
980         error_reportf_err(err, "cannot initialize crypto: ");
981         exit(1);
982     }
983 
984     os_setup_early_signal_handling();
985 
986     bdrv_init_with_whitelist();
987     socket_init();
988 }
989 
990 
991 void qemu_cleanup(int status)
992 {
993     gdb_exit(status);
994 
995     /*
996      * cleaning up the migration object cancels any existing migration
997      * try to do this early so that it also stops using devices.
998      */
999     migration_shutdown();
1000 
1001     /*
1002      * Close the exports before draining the block layer. The export
1003      * drivers may have coroutines yielding on it, so we need to clean
1004      * them up before the drain, as otherwise they may be get stuck in
1005      * blk_wait_while_drained().
1006      */
1007     blk_exp_close_all();
1008 
1009 
1010     /* No more vcpu or device emulation activity beyond this point */
1011     vm_shutdown();
1012     replay_finish();
1013 
1014     /*
1015      * We must cancel all block jobs while the block layer is drained,
1016      * or cancelling will be affected by throttling and thus may block
1017      * for an extended period of time.
1018      * Begin the drained section after vm_shutdown() to avoid requests being
1019      * stuck in the BlockBackend's request queue.
1020      * We do not need to end this section, because we do not want any
1021      * requests happening from here on anyway.
1022      */
1023     bdrv_drain_all_begin();
1024     job_cancel_sync_all();
1025     bdrv_close_all();
1026 
1027     /* vhost-user must be cleaned up before chardevs.  */
1028     tpm_cleanup();
1029     net_cleanup();
1030     audio_cleanup();
1031     monitor_cleanup();
1032     qemu_chr_cleanup();
1033     user_creatable_cleanup();
1034     /* TODO: unref root container, check all devices are ok */
1035 }
1036