xref: /openbmc/qemu/migration/colo.c (revision 58ea30f5)
1 /*
2  * COarse-grain LOck-stepping Virtual Machines for Non-stop Service (COLO)
3  * (a.k.a. Fault Tolerance or Continuous Replication)
4  *
5  * Copyright (c) 2016 HUAWEI TECHNOLOGIES CO., LTD.
6  * Copyright (c) 2016 FUJITSU LIMITED
7  * Copyright (c) 2016 Intel Corporation
8  *
9  * This work is licensed under the terms of the GNU GPL, version 2 or
10  * later.  See the COPYING file in the top-level directory.
11  */
12 
13 #include "qemu/osdep.h"
14 #include "sysemu/sysemu.h"
15 #include "qapi/error.h"
16 #include "qapi/qapi-commands-migration.h"
17 #include "qemu-file-channel.h"
18 #include "migration.h"
19 #include "qemu-file.h"
20 #include "savevm.h"
21 #include "migration/colo.h"
22 #include "block.h"
23 #include "io/channel-buffer.h"
24 #include "trace.h"
25 #include "qemu/error-report.h"
26 #include "migration/failover.h"
27 #ifdef CONFIG_REPLICATION
28 #include "replication.h"
29 #endif
30 #include "net/colo-compare.h"
31 #include "net/colo.h"
32 #include "block/block.h"
33 #include "qapi/qapi-events-migration.h"
34 #include "qapi/qmp/qerror.h"
35 #include "sysemu/cpus.h"
36 #include "net/filter.h"
37 
38 static bool vmstate_loading;
39 static Notifier packets_compare_notifier;
40 
41 /* User need to know colo mode after COLO failover */
42 static COLOMode last_colo_mode;
43 
44 #define COLO_BUFFER_BASE_SIZE (4 * 1024 * 1024)
45 
46 bool migration_in_colo_state(void)
47 {
48     MigrationState *s = migrate_get_current();
49 
50     return (s->state == MIGRATION_STATUS_COLO);
51 }
52 
53 bool migration_incoming_in_colo_state(void)
54 {
55     MigrationIncomingState *mis = migration_incoming_get_current();
56 
57     return mis && (mis->state == MIGRATION_STATUS_COLO);
58 }
59 
60 static bool colo_runstate_is_stopped(void)
61 {
62     return runstate_check(RUN_STATE_COLO) || !runstate_is_running();
63 }
64 
65 static void secondary_vm_do_failover(void)
66 {
67 /* COLO needs enable block-replication */
68 #ifdef CONFIG_REPLICATION
69     int old_state;
70     MigrationIncomingState *mis = migration_incoming_get_current();
71     Error *local_err = NULL;
72 
73     /* Can not do failover during the process of VM's loading VMstate, Or
74      * it will break the secondary VM.
75      */
76     if (vmstate_loading) {
77         old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
78                         FAILOVER_STATUS_RELAUNCH);
79         if (old_state != FAILOVER_STATUS_ACTIVE) {
80             error_report("Unknown error while do failover for secondary VM,"
81                          "old_state: %s", FailoverStatus_str(old_state));
82         }
83         return;
84     }
85 
86     migrate_set_state(&mis->state, MIGRATION_STATUS_COLO,
87                       MIGRATION_STATUS_COMPLETED);
88 
89     replication_stop_all(true, &local_err);
90     if (local_err) {
91         error_report_err(local_err);
92     }
93 
94     /* Notify all filters of all NIC to do checkpoint */
95     colo_notify_filters_event(COLO_EVENT_FAILOVER, &local_err);
96     if (local_err) {
97         error_report_err(local_err);
98     }
99 
100     if (!autostart) {
101         error_report("\"-S\" qemu option will be ignored in secondary side");
102         /* recover runstate to normal migration finish state */
103         autostart = true;
104     }
105     /*
106      * Make sure COLO incoming thread not block in recv or send,
107      * If mis->from_src_file and mis->to_src_file use the same fd,
108      * The second shutdown() will return -1, we ignore this value,
109      * It is harmless.
110      */
111     if (mis->from_src_file) {
112         qemu_file_shutdown(mis->from_src_file);
113     }
114     if (mis->to_src_file) {
115         qemu_file_shutdown(mis->to_src_file);
116     }
117 
118     old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
119                                    FAILOVER_STATUS_COMPLETED);
120     if (old_state != FAILOVER_STATUS_ACTIVE) {
121         error_report("Incorrect state (%s) while doing failover for "
122                      "secondary VM", FailoverStatus_str(old_state));
123         return;
124     }
125     /* Notify COLO incoming thread that failover work is finished */
126     qemu_sem_post(&mis->colo_incoming_sem);
127 
128     /* For Secondary VM, jump to incoming co */
129     if (mis->migration_incoming_co) {
130         qemu_coroutine_enter(mis->migration_incoming_co);
131     }
132 #else
133     abort();
134 #endif
135 }
136 
137 static void primary_vm_do_failover(void)
138 {
139 #ifdef CONFIG_REPLICATION
140     MigrationState *s = migrate_get_current();
141     int old_state;
142     Error *local_err = NULL;
143 
144     migrate_set_state(&s->state, MIGRATION_STATUS_COLO,
145                       MIGRATION_STATUS_COMPLETED);
146     /*
147      * kick COLO thread which might wait at
148      * qemu_sem_wait(&s->colo_checkpoint_sem).
149      */
150     colo_checkpoint_notify(migrate_get_current());
151 
152     /*
153      * Wake up COLO thread which may blocked in recv() or send(),
154      * The s->rp_state.from_dst_file and s->to_dst_file may use the
155      * same fd, but we still shutdown the fd for twice, it is harmless.
156      */
157     if (s->to_dst_file) {
158         qemu_file_shutdown(s->to_dst_file);
159     }
160     if (s->rp_state.from_dst_file) {
161         qemu_file_shutdown(s->rp_state.from_dst_file);
162     }
163 
164     old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
165                                    FAILOVER_STATUS_COMPLETED);
166     if (old_state != FAILOVER_STATUS_ACTIVE) {
167         error_report("Incorrect state (%s) while doing failover for Primary VM",
168                      FailoverStatus_str(old_state));
169         return;
170     }
171 
172     replication_stop_all(true, &local_err);
173     if (local_err) {
174         error_report_err(local_err);
175         local_err = NULL;
176     }
177 
178     /* Notify COLO thread that failover work is finished */
179     qemu_sem_post(&s->colo_exit_sem);
180 #else
181     abort();
182 #endif
183 }
184 
185 COLOMode get_colo_mode(void)
186 {
187     if (migration_in_colo_state()) {
188         return COLO_MODE_PRIMARY;
189     } else if (migration_incoming_in_colo_state()) {
190         return COLO_MODE_SECONDARY;
191     } else {
192         return COLO_MODE_NONE;
193     }
194 }
195 
196 void colo_do_failover(MigrationState *s)
197 {
198     /* Make sure VM stopped while failover happened. */
199     if (!colo_runstate_is_stopped()) {
200         vm_stop_force_state(RUN_STATE_COLO);
201     }
202 
203     switch (get_colo_mode()) {
204     case COLO_MODE_PRIMARY:
205         primary_vm_do_failover();
206         break;
207     case COLO_MODE_SECONDARY:
208         secondary_vm_do_failover();
209         break;
210     default:
211         error_report("colo_do_failover failed because the colo mode"
212                      " could not be obtained");
213     }
214 }
215 
216 #ifdef CONFIG_REPLICATION
217 void qmp_xen_set_replication(bool enable, bool primary,
218                              bool has_failover, bool failover,
219                              Error **errp)
220 {
221     ReplicationMode mode = primary ?
222                            REPLICATION_MODE_PRIMARY :
223                            REPLICATION_MODE_SECONDARY;
224 
225     if (has_failover && enable) {
226         error_setg(errp, "Parameter 'failover' is only for"
227                    " stopping replication");
228         return;
229     }
230 
231     if (enable) {
232         replication_start_all(mode, errp);
233     } else {
234         if (!has_failover) {
235             failover = NULL;
236         }
237         replication_stop_all(failover, failover ? NULL : errp);
238     }
239 }
240 
241 ReplicationStatus *qmp_query_xen_replication_status(Error **errp)
242 {
243     Error *err = NULL;
244     ReplicationStatus *s = g_new0(ReplicationStatus, 1);
245 
246     replication_get_error_all(&err);
247     if (err) {
248         s->error = true;
249         s->has_desc = true;
250         s->desc = g_strdup(error_get_pretty(err));
251     } else {
252         s->error = false;
253     }
254 
255     error_free(err);
256     return s;
257 }
258 
259 void qmp_xen_colo_do_checkpoint(Error **errp)
260 {
261     replication_do_checkpoint_all(errp);
262 }
263 #endif
264 
265 COLOStatus *qmp_query_colo_status(Error **errp)
266 {
267     COLOStatus *s = g_new0(COLOStatus, 1);
268 
269     s->mode = get_colo_mode();
270     s->last_mode = last_colo_mode;
271 
272     switch (failover_get_state()) {
273     case FAILOVER_STATUS_NONE:
274         s->reason = COLO_EXIT_REASON_NONE;
275         break;
276     case FAILOVER_STATUS_COMPLETED:
277         s->reason = COLO_EXIT_REASON_REQUEST;
278         break;
279     default:
280         if (migration_in_colo_state()) {
281             s->reason = COLO_EXIT_REASON_PROCESSING;
282         } else {
283             s->reason = COLO_EXIT_REASON_ERROR;
284         }
285     }
286 
287     return s;
288 }
289 
290 static void colo_send_message(QEMUFile *f, COLOMessage msg,
291                               Error **errp)
292 {
293     int ret;
294 
295     if (msg >= COLO_MESSAGE__MAX) {
296         error_setg(errp, "%s: Invalid message", __func__);
297         return;
298     }
299     qemu_put_be32(f, msg);
300     qemu_fflush(f);
301 
302     ret = qemu_file_get_error(f);
303     if (ret < 0) {
304         error_setg_errno(errp, -ret, "Can't send COLO message");
305     }
306     trace_colo_send_message(COLOMessage_str(msg));
307 }
308 
309 static void colo_send_message_value(QEMUFile *f, COLOMessage msg,
310                                     uint64_t value, Error **errp)
311 {
312     Error *local_err = NULL;
313     int ret;
314 
315     colo_send_message(f, msg, &local_err);
316     if (local_err) {
317         error_propagate(errp, local_err);
318         return;
319     }
320     qemu_put_be64(f, value);
321     qemu_fflush(f);
322 
323     ret = qemu_file_get_error(f);
324     if (ret < 0) {
325         error_setg_errno(errp, -ret, "Failed to send value for message:%s",
326                          COLOMessage_str(msg));
327     }
328 }
329 
330 static COLOMessage colo_receive_message(QEMUFile *f, Error **errp)
331 {
332     COLOMessage msg;
333     int ret;
334 
335     msg = qemu_get_be32(f);
336     ret = qemu_file_get_error(f);
337     if (ret < 0) {
338         error_setg_errno(errp, -ret, "Can't receive COLO message");
339         return msg;
340     }
341     if (msg >= COLO_MESSAGE__MAX) {
342         error_setg(errp, "%s: Invalid message", __func__);
343         return msg;
344     }
345     trace_colo_receive_message(COLOMessage_str(msg));
346     return msg;
347 }
348 
349 static void colo_receive_check_message(QEMUFile *f, COLOMessage expect_msg,
350                                        Error **errp)
351 {
352     COLOMessage msg;
353     Error *local_err = NULL;
354 
355     msg = colo_receive_message(f, &local_err);
356     if (local_err) {
357         error_propagate(errp, local_err);
358         return;
359     }
360     if (msg != expect_msg) {
361         error_setg(errp, "Unexpected COLO message %d, expected %d",
362                           msg, expect_msg);
363     }
364 }
365 
366 static uint64_t colo_receive_message_value(QEMUFile *f, uint32_t expect_msg,
367                                            Error **errp)
368 {
369     Error *local_err = NULL;
370     uint64_t value;
371     int ret;
372 
373     colo_receive_check_message(f, expect_msg, &local_err);
374     if (local_err) {
375         error_propagate(errp, local_err);
376         return 0;
377     }
378 
379     value = qemu_get_be64(f);
380     ret = qemu_file_get_error(f);
381     if (ret < 0) {
382         error_setg_errno(errp, -ret, "Failed to get value for COLO message: %s",
383                          COLOMessage_str(expect_msg));
384     }
385     return value;
386 }
387 
388 static int colo_do_checkpoint_transaction(MigrationState *s,
389                                           QIOChannelBuffer *bioc,
390                                           QEMUFile *fb)
391 {
392     Error *local_err = NULL;
393     int ret = -1;
394 
395     colo_send_message(s->to_dst_file, COLO_MESSAGE_CHECKPOINT_REQUEST,
396                       &local_err);
397     if (local_err) {
398         goto out;
399     }
400 
401     colo_receive_check_message(s->rp_state.from_dst_file,
402                     COLO_MESSAGE_CHECKPOINT_REPLY, &local_err);
403     if (local_err) {
404         goto out;
405     }
406     /* Reset channel-buffer directly */
407     qio_channel_io_seek(QIO_CHANNEL(bioc), 0, 0, NULL);
408     bioc->usage = 0;
409 
410     qemu_mutex_lock_iothread();
411     if (failover_get_state() != FAILOVER_STATUS_NONE) {
412         qemu_mutex_unlock_iothread();
413         goto out;
414     }
415     vm_stop_force_state(RUN_STATE_COLO);
416     qemu_mutex_unlock_iothread();
417     trace_colo_vm_state_change("run", "stop");
418     /*
419      * Failover request bh could be called after vm_stop_force_state(),
420      * So we need check failover_request_is_active() again.
421      */
422     if (failover_get_state() != FAILOVER_STATUS_NONE) {
423         goto out;
424     }
425 
426     colo_notify_compares_event(NULL, COLO_EVENT_CHECKPOINT, &local_err);
427     if (local_err) {
428         goto out;
429     }
430 
431     /* Disable block migration */
432     migrate_set_block_enabled(false, &local_err);
433     qemu_mutex_lock_iothread();
434 
435 #ifdef CONFIG_REPLICATION
436     replication_do_checkpoint_all(&local_err);
437     if (local_err) {
438         qemu_mutex_unlock_iothread();
439         goto out;
440     }
441 #else
442         abort();
443 #endif
444 
445     colo_send_message(s->to_dst_file, COLO_MESSAGE_VMSTATE_SEND, &local_err);
446     if (local_err) {
447         qemu_mutex_unlock_iothread();
448         goto out;
449     }
450     /* Note: device state is saved into buffer */
451     ret = qemu_save_device_state(fb);
452 
453     qemu_mutex_unlock_iothread();
454     if (ret < 0) {
455         goto out;
456     }
457     /*
458      * Only save VM's live state, which not including device state.
459      * TODO: We may need a timeout mechanism to prevent COLO process
460      * to be blocked here.
461      */
462     qemu_savevm_live_state(s->to_dst_file);
463 
464     qemu_fflush(fb);
465 
466     /*
467      * We need the size of the VMstate data in Secondary side,
468      * With which we can decide how much data should be read.
469      */
470     colo_send_message_value(s->to_dst_file, COLO_MESSAGE_VMSTATE_SIZE,
471                             bioc->usage, &local_err);
472     if (local_err) {
473         goto out;
474     }
475 
476     qemu_put_buffer(s->to_dst_file, bioc->data, bioc->usage);
477     qemu_fflush(s->to_dst_file);
478     ret = qemu_file_get_error(s->to_dst_file);
479     if (ret < 0) {
480         goto out;
481     }
482 
483     colo_receive_check_message(s->rp_state.from_dst_file,
484                        COLO_MESSAGE_VMSTATE_RECEIVED, &local_err);
485     if (local_err) {
486         goto out;
487     }
488 
489     colo_receive_check_message(s->rp_state.from_dst_file,
490                        COLO_MESSAGE_VMSTATE_LOADED, &local_err);
491     if (local_err) {
492         goto out;
493     }
494 
495     ret = 0;
496 
497     qemu_mutex_lock_iothread();
498     vm_start();
499     qemu_mutex_unlock_iothread();
500     trace_colo_vm_state_change("stop", "run");
501 
502 out:
503     if (local_err) {
504         error_report_err(local_err);
505     }
506     return ret;
507 }
508 
509 static void colo_compare_notify_checkpoint(Notifier *notifier, void *data)
510 {
511     colo_checkpoint_notify(data);
512 }
513 
514 static void colo_process_checkpoint(MigrationState *s)
515 {
516     QIOChannelBuffer *bioc;
517     QEMUFile *fb = NULL;
518     int64_t current_time = qemu_clock_get_ms(QEMU_CLOCK_HOST);
519     Error *local_err = NULL;
520     int ret;
521 
522     last_colo_mode = get_colo_mode();
523     if (last_colo_mode != COLO_MODE_PRIMARY) {
524         error_report("COLO mode must be COLO_MODE_PRIMARY");
525         return;
526     }
527 
528     failover_init_state();
529 
530     s->rp_state.from_dst_file = qemu_file_get_return_path(s->to_dst_file);
531     if (!s->rp_state.from_dst_file) {
532         error_report("Open QEMUFile from_dst_file failed");
533         goto out;
534     }
535 
536     packets_compare_notifier.notify = colo_compare_notify_checkpoint;
537     colo_compare_register_notifier(&packets_compare_notifier);
538 
539     /*
540      * Wait for Secondary finish loading VM states and enter COLO
541      * restore.
542      */
543     colo_receive_check_message(s->rp_state.from_dst_file,
544                        COLO_MESSAGE_CHECKPOINT_READY, &local_err);
545     if (local_err) {
546         goto out;
547     }
548     bioc = qio_channel_buffer_new(COLO_BUFFER_BASE_SIZE);
549     fb = qemu_fopen_channel_output(QIO_CHANNEL(bioc));
550     object_unref(OBJECT(bioc));
551 
552     qemu_mutex_lock_iothread();
553 #ifdef CONFIG_REPLICATION
554     replication_start_all(REPLICATION_MODE_PRIMARY, &local_err);
555     if (local_err) {
556         qemu_mutex_unlock_iothread();
557         goto out;
558     }
559 #else
560         abort();
561 #endif
562 
563     vm_start();
564     qemu_mutex_unlock_iothread();
565     trace_colo_vm_state_change("stop", "run");
566 
567     timer_mod(s->colo_delay_timer,
568             current_time + s->parameters.x_checkpoint_delay);
569 
570     while (s->state == MIGRATION_STATUS_COLO) {
571         if (failover_get_state() != FAILOVER_STATUS_NONE) {
572             error_report("failover request");
573             goto out;
574         }
575 
576         qemu_sem_wait(&s->colo_checkpoint_sem);
577 
578         if (s->state != MIGRATION_STATUS_COLO) {
579             goto out;
580         }
581         ret = colo_do_checkpoint_transaction(s, bioc, fb);
582         if (ret < 0) {
583             goto out;
584         }
585     }
586 
587 out:
588     /* Throw the unreported error message after exited from loop */
589     if (local_err) {
590         error_report_err(local_err);
591     }
592 
593     if (fb) {
594         qemu_fclose(fb);
595     }
596 
597     /*
598      * There are only two reasons we can get here, some error happened
599      * or the user triggered failover.
600      */
601     switch (failover_get_state()) {
602     case FAILOVER_STATUS_COMPLETED:
603         qapi_event_send_colo_exit(COLO_MODE_PRIMARY,
604                                   COLO_EXIT_REASON_REQUEST);
605         break;
606     default:
607         qapi_event_send_colo_exit(COLO_MODE_PRIMARY,
608                                   COLO_EXIT_REASON_ERROR);
609     }
610 
611     /* Hope this not to be too long to wait here */
612     qemu_sem_wait(&s->colo_exit_sem);
613     qemu_sem_destroy(&s->colo_exit_sem);
614 
615     /*
616      * It is safe to unregister notifier after failover finished.
617      * Besides, colo_delay_timer and colo_checkpoint_sem can't be
618      * released befor unregister notifier, or there will be use-after-free
619      * error.
620      */
621     colo_compare_unregister_notifier(&packets_compare_notifier);
622     timer_del(s->colo_delay_timer);
623     timer_free(s->colo_delay_timer);
624     qemu_sem_destroy(&s->colo_checkpoint_sem);
625 
626     /*
627      * Must be called after failover BH is completed,
628      * Or the failover BH may shutdown the wrong fd that
629      * re-used by other threads after we release here.
630      */
631     if (s->rp_state.from_dst_file) {
632         qemu_fclose(s->rp_state.from_dst_file);
633     }
634 }
635 
636 void colo_checkpoint_notify(void *opaque)
637 {
638     MigrationState *s = opaque;
639     int64_t next_notify_time;
640 
641     qemu_sem_post(&s->colo_checkpoint_sem);
642     s->colo_checkpoint_time = qemu_clock_get_ms(QEMU_CLOCK_HOST);
643     next_notify_time = s->colo_checkpoint_time +
644                     s->parameters.x_checkpoint_delay;
645     timer_mod(s->colo_delay_timer, next_notify_time);
646 }
647 
648 void migrate_start_colo_process(MigrationState *s)
649 {
650     qemu_mutex_unlock_iothread();
651     qemu_sem_init(&s->colo_checkpoint_sem, 0);
652     s->colo_delay_timer =  timer_new_ms(QEMU_CLOCK_HOST,
653                                 colo_checkpoint_notify, s);
654 
655     qemu_sem_init(&s->colo_exit_sem, 0);
656     migrate_set_state(&s->state, MIGRATION_STATUS_ACTIVE,
657                       MIGRATION_STATUS_COLO);
658     colo_process_checkpoint(s);
659     qemu_mutex_lock_iothread();
660 }
661 
662 static void colo_wait_handle_message(QEMUFile *f, int *checkpoint_request,
663                                      Error **errp)
664 {
665     COLOMessage msg;
666     Error *local_err = NULL;
667 
668     msg = colo_receive_message(f, &local_err);
669     if (local_err) {
670         error_propagate(errp, local_err);
671         return;
672     }
673 
674     switch (msg) {
675     case COLO_MESSAGE_CHECKPOINT_REQUEST:
676         *checkpoint_request = 1;
677         break;
678     default:
679         *checkpoint_request = 0;
680         error_setg(errp, "Got unknown COLO message: %d", msg);
681         break;
682     }
683 }
684 
685 void *colo_process_incoming_thread(void *opaque)
686 {
687     MigrationIncomingState *mis = opaque;
688     QEMUFile *fb = NULL;
689     QIOChannelBuffer *bioc = NULL; /* Cache incoming device state */
690     uint64_t total_size;
691     uint64_t value;
692     Error *local_err = NULL;
693     int ret;
694 
695     rcu_register_thread();
696     qemu_sem_init(&mis->colo_incoming_sem, 0);
697 
698     migrate_set_state(&mis->state, MIGRATION_STATUS_ACTIVE,
699                       MIGRATION_STATUS_COLO);
700 
701     last_colo_mode = get_colo_mode();
702     if (last_colo_mode != COLO_MODE_SECONDARY) {
703         error_report("COLO mode must be COLO_MODE_SECONDARY");
704         return NULL;
705     }
706 
707     failover_init_state();
708 
709     mis->to_src_file = qemu_file_get_return_path(mis->from_src_file);
710     if (!mis->to_src_file) {
711         error_report("COLO incoming thread: Open QEMUFile to_src_file failed");
712         goto out;
713     }
714     /*
715      * Note: the communication between Primary side and Secondary side
716      * should be sequential, we set the fd to unblocked in migration incoming
717      * coroutine, and here we are in the COLO incoming thread, so it is ok to
718      * set the fd back to blocked.
719      */
720     qemu_file_set_blocking(mis->from_src_file, true);
721 
722     bioc = qio_channel_buffer_new(COLO_BUFFER_BASE_SIZE);
723     fb = qemu_fopen_channel_input(QIO_CHANNEL(bioc));
724     object_unref(OBJECT(bioc));
725 
726     qemu_mutex_lock_iothread();
727 #ifdef CONFIG_REPLICATION
728     replication_start_all(REPLICATION_MODE_SECONDARY, &local_err);
729     if (local_err) {
730         qemu_mutex_unlock_iothread();
731         goto out;
732     }
733 #else
734         abort();
735 #endif
736     vm_start();
737     trace_colo_vm_state_change("stop", "run");
738     qemu_mutex_unlock_iothread();
739 
740     colo_send_message(mis->to_src_file, COLO_MESSAGE_CHECKPOINT_READY,
741                       &local_err);
742     if (local_err) {
743         goto out;
744     }
745 
746     while (mis->state == MIGRATION_STATUS_COLO) {
747         int request = 0;
748 
749         colo_wait_handle_message(mis->from_src_file, &request, &local_err);
750         if (local_err) {
751             goto out;
752         }
753         assert(request);
754         if (failover_get_state() != FAILOVER_STATUS_NONE) {
755             error_report("failover request");
756             goto out;
757         }
758 
759         qemu_mutex_lock_iothread();
760         vm_stop_force_state(RUN_STATE_COLO);
761         trace_colo_vm_state_change("run", "stop");
762         qemu_mutex_unlock_iothread();
763 
764         /* FIXME: This is unnecessary for periodic checkpoint mode */
765         colo_send_message(mis->to_src_file, COLO_MESSAGE_CHECKPOINT_REPLY,
766                      &local_err);
767         if (local_err) {
768             goto out;
769         }
770 
771         colo_receive_check_message(mis->from_src_file,
772                            COLO_MESSAGE_VMSTATE_SEND, &local_err);
773         if (local_err) {
774             goto out;
775         }
776 
777         qemu_mutex_lock_iothread();
778         cpu_synchronize_all_pre_loadvm();
779         ret = qemu_loadvm_state_main(mis->from_src_file, mis);
780         qemu_mutex_unlock_iothread();
781 
782         if (ret < 0) {
783             error_report("Load VM's live state (ram) error");
784             goto out;
785         }
786 
787         value = colo_receive_message_value(mis->from_src_file,
788                                  COLO_MESSAGE_VMSTATE_SIZE, &local_err);
789         if (local_err) {
790             goto out;
791         }
792 
793         /*
794          * Read VM device state data into channel buffer,
795          * It's better to re-use the memory allocated.
796          * Here we need to handle the channel buffer directly.
797          */
798         if (value > bioc->capacity) {
799             bioc->capacity = value;
800             bioc->data = g_realloc(bioc->data, bioc->capacity);
801         }
802         total_size = qemu_get_buffer(mis->from_src_file, bioc->data, value);
803         if (total_size != value) {
804             error_report("Got %" PRIu64 " VMState data, less than expected"
805                         " %" PRIu64, total_size, value);
806             goto out;
807         }
808         bioc->usage = total_size;
809         qio_channel_io_seek(QIO_CHANNEL(bioc), 0, 0, NULL);
810 
811         colo_send_message(mis->to_src_file, COLO_MESSAGE_VMSTATE_RECEIVED,
812                      &local_err);
813         if (local_err) {
814             goto out;
815         }
816 
817         qemu_mutex_lock_iothread();
818         vmstate_loading = true;
819         ret = qemu_load_device_state(fb);
820         if (ret < 0) {
821             error_report("COLO: load device state failed");
822             qemu_mutex_unlock_iothread();
823             goto out;
824         }
825 
826 #ifdef CONFIG_REPLICATION
827         replication_get_error_all(&local_err);
828         if (local_err) {
829             qemu_mutex_unlock_iothread();
830             goto out;
831         }
832 
833         /* discard colo disk buffer */
834         replication_do_checkpoint_all(&local_err);
835         if (local_err) {
836             qemu_mutex_unlock_iothread();
837             goto out;
838         }
839 #else
840         abort();
841 #endif
842         /* Notify all filters of all NIC to do checkpoint */
843         colo_notify_filters_event(COLO_EVENT_CHECKPOINT, &local_err);
844 
845         if (local_err) {
846             qemu_mutex_unlock_iothread();
847             goto out;
848         }
849 
850         vmstate_loading = false;
851         vm_start();
852         trace_colo_vm_state_change("stop", "run");
853         qemu_mutex_unlock_iothread();
854 
855         if (failover_get_state() == FAILOVER_STATUS_RELAUNCH) {
856             failover_set_state(FAILOVER_STATUS_RELAUNCH,
857                             FAILOVER_STATUS_NONE);
858             failover_request_active(NULL);
859             goto out;
860         }
861 
862         colo_send_message(mis->to_src_file, COLO_MESSAGE_VMSTATE_LOADED,
863                      &local_err);
864         if (local_err) {
865             goto out;
866         }
867     }
868 
869 out:
870     vmstate_loading = false;
871     /* Throw the unreported error message after exited from loop */
872     if (local_err) {
873         error_report_err(local_err);
874     }
875 
876     /*
877      * There are only two reasons we can get here, some error happened
878      * or the user triggered failover.
879      */
880     switch (failover_get_state()) {
881     case FAILOVER_STATUS_COMPLETED:
882         qapi_event_send_colo_exit(COLO_MODE_SECONDARY,
883                                   COLO_EXIT_REASON_REQUEST);
884         break;
885     default:
886         qapi_event_send_colo_exit(COLO_MODE_SECONDARY,
887                                   COLO_EXIT_REASON_ERROR);
888     }
889 
890     if (fb) {
891         qemu_fclose(fb);
892     }
893 
894     /* Hope this not to be too long to loop here */
895     qemu_sem_wait(&mis->colo_incoming_sem);
896     qemu_sem_destroy(&mis->colo_incoming_sem);
897     /* Must be called after failover BH is completed */
898     if (mis->to_src_file) {
899         qemu_fclose(mis->to_src_file);
900         mis->to_src_file = NULL;
901     }
902 
903     rcu_unregister_thread();
904     return NULL;
905 }
906