xref: /openbmc/qemu/migration/colo.c (revision 59a3a1c0)
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 "qemu/main-loop.h"
27 #include "qemu/rcu.h"
28 #include "migration/failover.h"
29 #ifdef CONFIG_REPLICATION
30 #include "replication.h"
31 #endif
32 #include "net/colo-compare.h"
33 #include "net/colo.h"
34 #include "block/block.h"
35 #include "qapi/qapi-events-migration.h"
36 #include "qapi/qmp/qerror.h"
37 #include "sysemu/cpus.h"
38 #include "sysemu/runstate.h"
39 #include "net/filter.h"
40 
41 static bool vmstate_loading;
42 static Notifier packets_compare_notifier;
43 
44 /* User need to know colo mode after COLO failover */
45 static COLOMode last_colo_mode;
46 
47 #define COLO_BUFFER_BASE_SIZE (4 * 1024 * 1024)
48 
49 bool migration_in_colo_state(void)
50 {
51     MigrationState *s = migrate_get_current();
52 
53     return (s->state == MIGRATION_STATUS_COLO);
54 }
55 
56 bool migration_incoming_in_colo_state(void)
57 {
58     MigrationIncomingState *mis = migration_incoming_get_current();
59 
60     return mis && (mis->state == MIGRATION_STATUS_COLO);
61 }
62 
63 static bool colo_runstate_is_stopped(void)
64 {
65     return runstate_check(RUN_STATE_COLO) || !runstate_is_running();
66 }
67 
68 static void secondary_vm_do_failover(void)
69 {
70 /* COLO needs enable block-replication */
71 #ifdef CONFIG_REPLICATION
72     int old_state;
73     MigrationIncomingState *mis = migration_incoming_get_current();
74     Error *local_err = NULL;
75 
76     /* Can not do failover during the process of VM's loading VMstate, Or
77      * it will break the secondary VM.
78      */
79     if (vmstate_loading) {
80         old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
81                         FAILOVER_STATUS_RELAUNCH);
82         if (old_state != FAILOVER_STATUS_ACTIVE) {
83             error_report("Unknown error while do failover for secondary VM,"
84                          "old_state: %s", FailoverStatus_str(old_state));
85         }
86         return;
87     }
88 
89     migrate_set_state(&mis->state, MIGRATION_STATUS_COLO,
90                       MIGRATION_STATUS_COMPLETED);
91 
92     replication_stop_all(true, &local_err);
93     if (local_err) {
94         error_report_err(local_err);
95     }
96 
97     /* Notify all filters of all NIC to do checkpoint */
98     colo_notify_filters_event(COLO_EVENT_FAILOVER, &local_err);
99     if (local_err) {
100         error_report_err(local_err);
101     }
102 
103     if (!autostart) {
104         error_report("\"-S\" qemu option will be ignored in secondary side");
105         /* recover runstate to normal migration finish state */
106         autostart = true;
107     }
108     /*
109      * Make sure COLO incoming thread not block in recv or send,
110      * If mis->from_src_file and mis->to_src_file use the same fd,
111      * The second shutdown() will return -1, we ignore this value,
112      * It is harmless.
113      */
114     if (mis->from_src_file) {
115         qemu_file_shutdown(mis->from_src_file);
116     }
117     if (mis->to_src_file) {
118         qemu_file_shutdown(mis->to_src_file);
119     }
120 
121     old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
122                                    FAILOVER_STATUS_COMPLETED);
123     if (old_state != FAILOVER_STATUS_ACTIVE) {
124         error_report("Incorrect state (%s) while doing failover for "
125                      "secondary VM", FailoverStatus_str(old_state));
126         return;
127     }
128     /* Notify COLO incoming thread that failover work is finished */
129     qemu_sem_post(&mis->colo_incoming_sem);
130 
131     /* For Secondary VM, jump to incoming co */
132     if (mis->migration_incoming_co) {
133         qemu_coroutine_enter(mis->migration_incoming_co);
134     }
135 #else
136     abort();
137 #endif
138 }
139 
140 static void primary_vm_do_failover(void)
141 {
142 #ifdef CONFIG_REPLICATION
143     MigrationState *s = migrate_get_current();
144     int old_state;
145     Error *local_err = NULL;
146 
147     migrate_set_state(&s->state, MIGRATION_STATUS_COLO,
148                       MIGRATION_STATUS_COMPLETED);
149     /*
150      * kick COLO thread which might wait at
151      * qemu_sem_wait(&s->colo_checkpoint_sem).
152      */
153     colo_checkpoint_notify(migrate_get_current());
154 
155     /*
156      * Wake up COLO thread which may blocked in recv() or send(),
157      * The s->rp_state.from_dst_file and s->to_dst_file may use the
158      * same fd, but we still shutdown the fd for twice, it is harmless.
159      */
160     if (s->to_dst_file) {
161         qemu_file_shutdown(s->to_dst_file);
162     }
163     if (s->rp_state.from_dst_file) {
164         qemu_file_shutdown(s->rp_state.from_dst_file);
165     }
166 
167     old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
168                                    FAILOVER_STATUS_COMPLETED);
169     if (old_state != FAILOVER_STATUS_ACTIVE) {
170         error_report("Incorrect state (%s) while doing failover for Primary VM",
171                      FailoverStatus_str(old_state));
172         return;
173     }
174 
175     replication_stop_all(true, &local_err);
176     if (local_err) {
177         error_report_err(local_err);
178         local_err = NULL;
179     }
180 
181     /* Notify COLO thread that failover work is finished */
182     qemu_sem_post(&s->colo_exit_sem);
183 #else
184     abort();
185 #endif
186 }
187 
188 COLOMode get_colo_mode(void)
189 {
190     if (migration_in_colo_state()) {
191         return COLO_MODE_PRIMARY;
192     } else if (migration_incoming_in_colo_state()) {
193         return COLO_MODE_SECONDARY;
194     } else {
195         return COLO_MODE_NONE;
196     }
197 }
198 
199 void colo_do_failover(void)
200 {
201     /* Make sure VM stopped while failover happened. */
202     if (!colo_runstate_is_stopped()) {
203         vm_stop_force_state(RUN_STATE_COLO);
204     }
205 
206     switch (get_colo_mode()) {
207     case COLO_MODE_PRIMARY:
208         primary_vm_do_failover();
209         break;
210     case COLO_MODE_SECONDARY:
211         secondary_vm_do_failover();
212         break;
213     default:
214         error_report("colo_do_failover failed because the colo mode"
215                      " could not be obtained");
216     }
217 }
218 
219 #ifdef CONFIG_REPLICATION
220 void qmp_xen_set_replication(bool enable, bool primary,
221                              bool has_failover, bool failover,
222                              Error **errp)
223 {
224     ReplicationMode mode = primary ?
225                            REPLICATION_MODE_PRIMARY :
226                            REPLICATION_MODE_SECONDARY;
227 
228     if (has_failover && enable) {
229         error_setg(errp, "Parameter 'failover' is only for"
230                    " stopping replication");
231         return;
232     }
233 
234     if (enable) {
235         replication_start_all(mode, errp);
236     } else {
237         if (!has_failover) {
238             failover = NULL;
239         }
240         replication_stop_all(failover, failover ? NULL : errp);
241     }
242 }
243 
244 ReplicationStatus *qmp_query_xen_replication_status(Error **errp)
245 {
246     Error *err = NULL;
247     ReplicationStatus *s = g_new0(ReplicationStatus, 1);
248 
249     replication_get_error_all(&err);
250     if (err) {
251         s->error = true;
252         s->has_desc = true;
253         s->desc = g_strdup(error_get_pretty(err));
254     } else {
255         s->error = false;
256     }
257 
258     error_free(err);
259     return s;
260 }
261 
262 void qmp_xen_colo_do_checkpoint(Error **errp)
263 {
264     replication_do_checkpoint_all(errp);
265     /* Notify all filters of all NIC to do checkpoint */
266     colo_notify_filters_event(COLO_EVENT_CHECKPOINT, errp);
267 }
268 #endif
269 
270 COLOStatus *qmp_query_colo_status(Error **errp)
271 {
272     COLOStatus *s = g_new0(COLOStatus, 1);
273 
274     s->mode = get_colo_mode();
275     s->last_mode = last_colo_mode;
276 
277     switch (failover_get_state()) {
278     case FAILOVER_STATUS_NONE:
279         s->reason = COLO_EXIT_REASON_NONE;
280         break;
281     case FAILOVER_STATUS_COMPLETED:
282         s->reason = COLO_EXIT_REASON_REQUEST;
283         break;
284     default:
285         if (migration_in_colo_state()) {
286             s->reason = COLO_EXIT_REASON_PROCESSING;
287         } else {
288             s->reason = COLO_EXIT_REASON_ERROR;
289         }
290     }
291 
292     return s;
293 }
294 
295 static void colo_send_message(QEMUFile *f, COLOMessage msg,
296                               Error **errp)
297 {
298     int ret;
299 
300     if (msg >= COLO_MESSAGE__MAX) {
301         error_setg(errp, "%s: Invalid message", __func__);
302         return;
303     }
304     qemu_put_be32(f, msg);
305     qemu_fflush(f);
306 
307     ret = qemu_file_get_error(f);
308     if (ret < 0) {
309         error_setg_errno(errp, -ret, "Can't send COLO message");
310     }
311     trace_colo_send_message(COLOMessage_str(msg));
312 }
313 
314 static void colo_send_message_value(QEMUFile *f, COLOMessage msg,
315                                     uint64_t value, Error **errp)
316 {
317     Error *local_err = NULL;
318     int ret;
319 
320     colo_send_message(f, msg, &local_err);
321     if (local_err) {
322         error_propagate(errp, local_err);
323         return;
324     }
325     qemu_put_be64(f, value);
326     qemu_fflush(f);
327 
328     ret = qemu_file_get_error(f);
329     if (ret < 0) {
330         error_setg_errno(errp, -ret, "Failed to send value for message:%s",
331                          COLOMessage_str(msg));
332     }
333 }
334 
335 static COLOMessage colo_receive_message(QEMUFile *f, Error **errp)
336 {
337     COLOMessage msg;
338     int ret;
339 
340     msg = qemu_get_be32(f);
341     ret = qemu_file_get_error(f);
342     if (ret < 0) {
343         error_setg_errno(errp, -ret, "Can't receive COLO message");
344         return msg;
345     }
346     if (msg >= COLO_MESSAGE__MAX) {
347         error_setg(errp, "%s: Invalid message", __func__);
348         return msg;
349     }
350     trace_colo_receive_message(COLOMessage_str(msg));
351     return msg;
352 }
353 
354 static void colo_receive_check_message(QEMUFile *f, COLOMessage expect_msg,
355                                        Error **errp)
356 {
357     COLOMessage msg;
358     Error *local_err = NULL;
359 
360     msg = colo_receive_message(f, &local_err);
361     if (local_err) {
362         error_propagate(errp, local_err);
363         return;
364     }
365     if (msg != expect_msg) {
366         error_setg(errp, "Unexpected COLO message %d, expected %d",
367                           msg, expect_msg);
368     }
369 }
370 
371 static uint64_t colo_receive_message_value(QEMUFile *f, uint32_t expect_msg,
372                                            Error **errp)
373 {
374     Error *local_err = NULL;
375     uint64_t value;
376     int ret;
377 
378     colo_receive_check_message(f, expect_msg, &local_err);
379     if (local_err) {
380         error_propagate(errp, local_err);
381         return 0;
382     }
383 
384     value = qemu_get_be64(f);
385     ret = qemu_file_get_error(f);
386     if (ret < 0) {
387         error_setg_errno(errp, -ret, "Failed to get value for COLO message: %s",
388                          COLOMessage_str(expect_msg));
389     }
390     return value;
391 }
392 
393 static int colo_do_checkpoint_transaction(MigrationState *s,
394                                           QIOChannelBuffer *bioc,
395                                           QEMUFile *fb)
396 {
397     Error *local_err = NULL;
398     int ret = -1;
399 
400     colo_send_message(s->to_dst_file, COLO_MESSAGE_CHECKPOINT_REQUEST,
401                       &local_err);
402     if (local_err) {
403         goto out;
404     }
405 
406     colo_receive_check_message(s->rp_state.from_dst_file,
407                     COLO_MESSAGE_CHECKPOINT_REPLY, &local_err);
408     if (local_err) {
409         goto out;
410     }
411     /* Reset channel-buffer directly */
412     qio_channel_io_seek(QIO_CHANNEL(bioc), 0, 0, NULL);
413     bioc->usage = 0;
414 
415     qemu_mutex_lock_iothread();
416     if (failover_get_state() != FAILOVER_STATUS_NONE) {
417         qemu_mutex_unlock_iothread();
418         goto out;
419     }
420     vm_stop_force_state(RUN_STATE_COLO);
421     qemu_mutex_unlock_iothread();
422     trace_colo_vm_state_change("run", "stop");
423     /*
424      * Failover request bh could be called after vm_stop_force_state(),
425      * So we need check failover_request_is_active() again.
426      */
427     if (failover_get_state() != FAILOVER_STATUS_NONE) {
428         goto out;
429     }
430 
431     colo_notify_compares_event(NULL, COLO_EVENT_CHECKPOINT, &local_err);
432     if (local_err) {
433         goto out;
434     }
435 
436     /* Disable block migration */
437     migrate_set_block_enabled(false, &local_err);
438     qemu_mutex_lock_iothread();
439 
440 #ifdef CONFIG_REPLICATION
441     replication_do_checkpoint_all(&local_err);
442     if (local_err) {
443         qemu_mutex_unlock_iothread();
444         goto out;
445     }
446 #else
447         abort();
448 #endif
449 
450     colo_send_message(s->to_dst_file, COLO_MESSAGE_VMSTATE_SEND, &local_err);
451     if (local_err) {
452         qemu_mutex_unlock_iothread();
453         goto out;
454     }
455     /* Note: device state is saved into buffer */
456     ret = qemu_save_device_state(fb);
457 
458     qemu_mutex_unlock_iothread();
459     if (ret < 0) {
460         goto out;
461     }
462     /*
463      * Only save VM's live state, which not including device state.
464      * TODO: We may need a timeout mechanism to prevent COLO process
465      * to be blocked here.
466      */
467     qemu_savevm_live_state(s->to_dst_file);
468 
469     qemu_fflush(fb);
470 
471     /*
472      * We need the size of the VMstate data in Secondary side,
473      * With which we can decide how much data should be read.
474      */
475     colo_send_message_value(s->to_dst_file, COLO_MESSAGE_VMSTATE_SIZE,
476                             bioc->usage, &local_err);
477     if (local_err) {
478         goto out;
479     }
480 
481     qemu_put_buffer(s->to_dst_file, bioc->data, bioc->usage);
482     qemu_fflush(s->to_dst_file);
483     ret = qemu_file_get_error(s->to_dst_file);
484     if (ret < 0) {
485         goto out;
486     }
487 
488     colo_receive_check_message(s->rp_state.from_dst_file,
489                        COLO_MESSAGE_VMSTATE_RECEIVED, &local_err);
490     if (local_err) {
491         goto out;
492     }
493 
494     colo_receive_check_message(s->rp_state.from_dst_file,
495                        COLO_MESSAGE_VMSTATE_LOADED, &local_err);
496     if (local_err) {
497         goto out;
498     }
499 
500     ret = 0;
501 
502     qemu_mutex_lock_iothread();
503     vm_start();
504     qemu_mutex_unlock_iothread();
505     trace_colo_vm_state_change("stop", "run");
506 
507 out:
508     if (local_err) {
509         error_report_err(local_err);
510     }
511     return ret;
512 }
513 
514 static void colo_compare_notify_checkpoint(Notifier *notifier, void *data)
515 {
516     colo_checkpoint_notify(data);
517 }
518 
519 static void colo_process_checkpoint(MigrationState *s)
520 {
521     QIOChannelBuffer *bioc;
522     QEMUFile *fb = NULL;
523     int64_t current_time = qemu_clock_get_ms(QEMU_CLOCK_HOST);
524     Error *local_err = NULL;
525     int ret;
526 
527     last_colo_mode = get_colo_mode();
528     if (last_colo_mode != COLO_MODE_PRIMARY) {
529         error_report("COLO mode must be COLO_MODE_PRIMARY");
530         return;
531     }
532 
533     failover_init_state();
534 
535     s->rp_state.from_dst_file = qemu_file_get_return_path(s->to_dst_file);
536     if (!s->rp_state.from_dst_file) {
537         error_report("Open QEMUFile from_dst_file failed");
538         goto out;
539     }
540 
541     packets_compare_notifier.notify = colo_compare_notify_checkpoint;
542     colo_compare_register_notifier(&packets_compare_notifier);
543 
544     /*
545      * Wait for Secondary finish loading VM states and enter COLO
546      * restore.
547      */
548     colo_receive_check_message(s->rp_state.from_dst_file,
549                        COLO_MESSAGE_CHECKPOINT_READY, &local_err);
550     if (local_err) {
551         goto out;
552     }
553     bioc = qio_channel_buffer_new(COLO_BUFFER_BASE_SIZE);
554     fb = qemu_fopen_channel_output(QIO_CHANNEL(bioc));
555     object_unref(OBJECT(bioc));
556 
557     qemu_mutex_lock_iothread();
558 #ifdef CONFIG_REPLICATION
559     replication_start_all(REPLICATION_MODE_PRIMARY, &local_err);
560     if (local_err) {
561         qemu_mutex_unlock_iothread();
562         goto out;
563     }
564 #else
565         abort();
566 #endif
567 
568     vm_start();
569     qemu_mutex_unlock_iothread();
570     trace_colo_vm_state_change("stop", "run");
571 
572     timer_mod(s->colo_delay_timer,
573             current_time + s->parameters.x_checkpoint_delay);
574 
575     while (s->state == MIGRATION_STATUS_COLO) {
576         if (failover_get_state() != FAILOVER_STATUS_NONE) {
577             error_report("failover request");
578             goto out;
579         }
580 
581         qemu_sem_wait(&s->colo_checkpoint_sem);
582 
583         if (s->state != MIGRATION_STATUS_COLO) {
584             goto out;
585         }
586         ret = colo_do_checkpoint_transaction(s, bioc, fb);
587         if (ret < 0) {
588             goto out;
589         }
590     }
591 
592 out:
593     /* Throw the unreported error message after exited from loop */
594     if (local_err) {
595         error_report_err(local_err);
596     }
597 
598     if (fb) {
599         qemu_fclose(fb);
600     }
601 
602     /*
603      * There are only two reasons we can get here, some error happened
604      * or the user triggered failover.
605      */
606     switch (failover_get_state()) {
607     case FAILOVER_STATUS_COMPLETED:
608         qapi_event_send_colo_exit(COLO_MODE_PRIMARY,
609                                   COLO_EXIT_REASON_REQUEST);
610         break;
611     default:
612         qapi_event_send_colo_exit(COLO_MODE_PRIMARY,
613                                   COLO_EXIT_REASON_ERROR);
614     }
615 
616     /* Hope this not to be too long to wait here */
617     qemu_sem_wait(&s->colo_exit_sem);
618     qemu_sem_destroy(&s->colo_exit_sem);
619 
620     /*
621      * It is safe to unregister notifier after failover finished.
622      * Besides, colo_delay_timer and colo_checkpoint_sem can't be
623      * released befor unregister notifier, or there will be use-after-free
624      * error.
625      */
626     colo_compare_unregister_notifier(&packets_compare_notifier);
627     timer_del(s->colo_delay_timer);
628     timer_free(s->colo_delay_timer);
629     qemu_sem_destroy(&s->colo_checkpoint_sem);
630 
631     /*
632      * Must be called after failover BH is completed,
633      * Or the failover BH may shutdown the wrong fd that
634      * re-used by other threads after we release here.
635      */
636     if (s->rp_state.from_dst_file) {
637         qemu_fclose(s->rp_state.from_dst_file);
638     }
639 }
640 
641 void colo_checkpoint_notify(void *opaque)
642 {
643     MigrationState *s = opaque;
644     int64_t next_notify_time;
645 
646     qemu_sem_post(&s->colo_checkpoint_sem);
647     s->colo_checkpoint_time = qemu_clock_get_ms(QEMU_CLOCK_HOST);
648     next_notify_time = s->colo_checkpoint_time +
649                     s->parameters.x_checkpoint_delay;
650     timer_mod(s->colo_delay_timer, next_notify_time);
651 }
652 
653 void migrate_start_colo_process(MigrationState *s)
654 {
655     qemu_mutex_unlock_iothread();
656     qemu_sem_init(&s->colo_checkpoint_sem, 0);
657     s->colo_delay_timer =  timer_new_ms(QEMU_CLOCK_HOST,
658                                 colo_checkpoint_notify, s);
659 
660     qemu_sem_init(&s->colo_exit_sem, 0);
661     migrate_set_state(&s->state, MIGRATION_STATUS_ACTIVE,
662                       MIGRATION_STATUS_COLO);
663     colo_process_checkpoint(s);
664     qemu_mutex_lock_iothread();
665 }
666 
667 static void colo_wait_handle_message(QEMUFile *f, int *checkpoint_request,
668                                      Error **errp)
669 {
670     COLOMessage msg;
671     Error *local_err = NULL;
672 
673     msg = colo_receive_message(f, &local_err);
674     if (local_err) {
675         error_propagate(errp, local_err);
676         return;
677     }
678 
679     switch (msg) {
680     case COLO_MESSAGE_CHECKPOINT_REQUEST:
681         *checkpoint_request = 1;
682         break;
683     default:
684         *checkpoint_request = 0;
685         error_setg(errp, "Got unknown COLO message: %d", msg);
686         break;
687     }
688 }
689 
690 void *colo_process_incoming_thread(void *opaque)
691 {
692     MigrationIncomingState *mis = opaque;
693     QEMUFile *fb = NULL;
694     QIOChannelBuffer *bioc = NULL; /* Cache incoming device state */
695     uint64_t total_size;
696     uint64_t value;
697     Error *local_err = NULL;
698     int ret;
699 
700     rcu_register_thread();
701     qemu_sem_init(&mis->colo_incoming_sem, 0);
702 
703     migrate_set_state(&mis->state, MIGRATION_STATUS_ACTIVE,
704                       MIGRATION_STATUS_COLO);
705 
706     last_colo_mode = get_colo_mode();
707     if (last_colo_mode != COLO_MODE_SECONDARY) {
708         error_report("COLO mode must be COLO_MODE_SECONDARY");
709         return NULL;
710     }
711 
712     failover_init_state();
713 
714     mis->to_src_file = qemu_file_get_return_path(mis->from_src_file);
715     if (!mis->to_src_file) {
716         error_report("COLO incoming thread: Open QEMUFile to_src_file failed");
717         goto out;
718     }
719     /*
720      * Note: the communication between Primary side and Secondary side
721      * should be sequential, we set the fd to unblocked in migration incoming
722      * coroutine, and here we are in the COLO incoming thread, so it is ok to
723      * set the fd back to blocked.
724      */
725     qemu_file_set_blocking(mis->from_src_file, true);
726 
727     bioc = qio_channel_buffer_new(COLO_BUFFER_BASE_SIZE);
728     fb = qemu_fopen_channel_input(QIO_CHANNEL(bioc));
729     object_unref(OBJECT(bioc));
730 
731     qemu_mutex_lock_iothread();
732 #ifdef CONFIG_REPLICATION
733     replication_start_all(REPLICATION_MODE_SECONDARY, &local_err);
734     if (local_err) {
735         qemu_mutex_unlock_iothread();
736         goto out;
737     }
738 #else
739         abort();
740 #endif
741     vm_start();
742     trace_colo_vm_state_change("stop", "run");
743     qemu_mutex_unlock_iothread();
744 
745     colo_send_message(mis->to_src_file, COLO_MESSAGE_CHECKPOINT_READY,
746                       &local_err);
747     if (local_err) {
748         goto out;
749     }
750 
751     while (mis->state == MIGRATION_STATUS_COLO) {
752         int request = 0;
753 
754         colo_wait_handle_message(mis->from_src_file, &request, &local_err);
755         if (local_err) {
756             goto out;
757         }
758         assert(request);
759         if (failover_get_state() != FAILOVER_STATUS_NONE) {
760             error_report("failover request");
761             goto out;
762         }
763 
764         qemu_mutex_lock_iothread();
765         vm_stop_force_state(RUN_STATE_COLO);
766         trace_colo_vm_state_change("run", "stop");
767         qemu_mutex_unlock_iothread();
768 
769         /* FIXME: This is unnecessary for periodic checkpoint mode */
770         colo_send_message(mis->to_src_file, COLO_MESSAGE_CHECKPOINT_REPLY,
771                      &local_err);
772         if (local_err) {
773             goto out;
774         }
775 
776         colo_receive_check_message(mis->from_src_file,
777                            COLO_MESSAGE_VMSTATE_SEND, &local_err);
778         if (local_err) {
779             goto out;
780         }
781 
782         qemu_mutex_lock_iothread();
783         cpu_synchronize_all_pre_loadvm();
784         ret = qemu_loadvm_state_main(mis->from_src_file, mis);
785         qemu_mutex_unlock_iothread();
786 
787         if (ret < 0) {
788             error_report("Load VM's live state (ram) error");
789             goto out;
790         }
791 
792         value = colo_receive_message_value(mis->from_src_file,
793                                  COLO_MESSAGE_VMSTATE_SIZE, &local_err);
794         if (local_err) {
795             goto out;
796         }
797 
798         /*
799          * Read VM device state data into channel buffer,
800          * It's better to re-use the memory allocated.
801          * Here we need to handle the channel buffer directly.
802          */
803         if (value > bioc->capacity) {
804             bioc->capacity = value;
805             bioc->data = g_realloc(bioc->data, bioc->capacity);
806         }
807         total_size = qemu_get_buffer(mis->from_src_file, bioc->data, value);
808         if (total_size != value) {
809             error_report("Got %" PRIu64 " VMState data, less than expected"
810                         " %" PRIu64, total_size, value);
811             goto out;
812         }
813         bioc->usage = total_size;
814         qio_channel_io_seek(QIO_CHANNEL(bioc), 0, 0, NULL);
815 
816         colo_send_message(mis->to_src_file, COLO_MESSAGE_VMSTATE_RECEIVED,
817                      &local_err);
818         if (local_err) {
819             goto out;
820         }
821 
822         qemu_mutex_lock_iothread();
823         vmstate_loading = true;
824         ret = qemu_load_device_state(fb);
825         if (ret < 0) {
826             error_report("COLO: load device state failed");
827             qemu_mutex_unlock_iothread();
828             goto out;
829         }
830 
831 #ifdef CONFIG_REPLICATION
832         replication_get_error_all(&local_err);
833         if (local_err) {
834             qemu_mutex_unlock_iothread();
835             goto out;
836         }
837 
838         /* discard colo disk buffer */
839         replication_do_checkpoint_all(&local_err);
840         if (local_err) {
841             qemu_mutex_unlock_iothread();
842             goto out;
843         }
844 #else
845         abort();
846 #endif
847         /* Notify all filters of all NIC to do checkpoint */
848         colo_notify_filters_event(COLO_EVENT_CHECKPOINT, &local_err);
849 
850         if (local_err) {
851             qemu_mutex_unlock_iothread();
852             goto out;
853         }
854 
855         vmstate_loading = false;
856         vm_start();
857         trace_colo_vm_state_change("stop", "run");
858         qemu_mutex_unlock_iothread();
859 
860         if (failover_get_state() == FAILOVER_STATUS_RELAUNCH) {
861             failover_set_state(FAILOVER_STATUS_RELAUNCH,
862                             FAILOVER_STATUS_NONE);
863             failover_request_active(NULL);
864             goto out;
865         }
866 
867         colo_send_message(mis->to_src_file, COLO_MESSAGE_VMSTATE_LOADED,
868                      &local_err);
869         if (local_err) {
870             goto out;
871         }
872     }
873 
874 out:
875     vmstate_loading = false;
876     /* Throw the unreported error message after exited from loop */
877     if (local_err) {
878         error_report_err(local_err);
879     }
880 
881     /*
882      * There are only two reasons we can get here, some error happened
883      * or the user triggered failover.
884      */
885     switch (failover_get_state()) {
886     case FAILOVER_STATUS_COMPLETED:
887         qapi_event_send_colo_exit(COLO_MODE_SECONDARY,
888                                   COLO_EXIT_REASON_REQUEST);
889         break;
890     default:
891         qapi_event_send_colo_exit(COLO_MODE_SECONDARY,
892                                   COLO_EXIT_REASON_ERROR);
893     }
894 
895     if (fb) {
896         qemu_fclose(fb);
897     }
898 
899     /* Hope this not to be too long to loop here */
900     qemu_sem_wait(&mis->colo_incoming_sem);
901     qemu_sem_destroy(&mis->colo_incoming_sem);
902     /* Must be called after failover BH is completed */
903     if (mis->to_src_file) {
904         qemu_fclose(mis->to_src_file);
905         mis->to_src_file = NULL;
906     }
907 
908     rcu_unregister_thread();
909     return NULL;
910 }
911