xref: /openbmc/qemu/migration/colo.c (revision 8e6c718a)
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 "migration.h"
18 #include "qemu-file.h"
19 #include "savevm.h"
20 #include "migration/colo.h"
21 #include "block.h"
22 #include "io/channel-buffer.h"
23 #include "trace.h"
24 #include "qemu/error-report.h"
25 #include "qemu/main-loop.h"
26 #include "qemu/rcu.h"
27 #include "migration/failover.h"
28 #include "migration/ram.h"
29 #include "block/replication.h"
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 "sysemu/cpus.h"
35 #include "sysemu/runstate.h"
36 #include "net/filter.h"
37 #include "options.h"
38 
39 static bool vmstate_loading;
40 static Notifier packets_compare_notifier;
41 
42 /* User need to know colo mode after COLO failover */
43 static COLOMode last_colo_mode;
44 
45 #define COLO_BUFFER_BASE_SIZE (4 * 1024 * 1024)
46 
47 bool migration_in_colo_state(void)
48 {
49     MigrationState *s = migrate_get_current();
50 
51     return (s->state == MIGRATION_STATUS_COLO);
52 }
53 
54 bool migration_incoming_in_colo_state(void)
55 {
56     MigrationIncomingState *mis = migration_incoming_get_current();
57 
58     return mis && (mis->state == MIGRATION_STATUS_COLO);
59 }
60 
61 static bool colo_runstate_is_stopped(void)
62 {
63     return runstate_check(RUN_STATE_COLO) || !runstate_is_running();
64 }
65 
66 static void colo_checkpoint_notify(void *opaque)
67 {
68     MigrationState *s = opaque;
69     int64_t next_notify_time;
70 
71     qemu_event_set(&s->colo_checkpoint_event);
72     s->colo_checkpoint_time = qemu_clock_get_ms(QEMU_CLOCK_HOST);
73     next_notify_time = s->colo_checkpoint_time + migrate_checkpoint_delay();
74     timer_mod(s->colo_delay_timer, next_notify_time);
75 }
76 
77 void colo_checkpoint_delay_set(void)
78 {
79     if (migration_in_colo_state()) {
80         colo_checkpoint_notify(migrate_get_current());
81     }
82 }
83 
84 static void secondary_vm_do_failover(void)
85 {
86 /* COLO needs enable block-replication */
87     int old_state;
88     MigrationIncomingState *mis = migration_incoming_get_current();
89     Error *local_err = NULL;
90 
91     /* Can not do failover during the process of VM's loading VMstate, Or
92      * it will break the secondary VM.
93      */
94     if (vmstate_loading) {
95         old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
96                         FAILOVER_STATUS_RELAUNCH);
97         if (old_state != FAILOVER_STATUS_ACTIVE) {
98             error_report("Unknown error while do failover for secondary VM,"
99                          "old_state: %s", FailoverStatus_str(old_state));
100         }
101         return;
102     }
103 
104     migrate_set_state(&mis->state, MIGRATION_STATUS_COLO,
105                       MIGRATION_STATUS_COMPLETED);
106 
107     replication_stop_all(true, &local_err);
108     if (local_err) {
109         error_report_err(local_err);
110         local_err = NULL;
111     }
112 
113     /* Notify all filters of all NIC to do checkpoint */
114     colo_notify_filters_event(COLO_EVENT_FAILOVER, &local_err);
115     if (local_err) {
116         error_report_err(local_err);
117     }
118 
119     if (!autostart) {
120         error_report("\"-S\" qemu option will be ignored in secondary side");
121         /* recover runstate to normal migration finish state */
122         autostart = true;
123     }
124     /*
125      * Make sure COLO incoming thread not block in recv or send,
126      * If mis->from_src_file and mis->to_src_file use the same fd,
127      * The second shutdown() will return -1, we ignore this value,
128      * It is harmless.
129      */
130     if (mis->from_src_file) {
131         qemu_file_shutdown(mis->from_src_file);
132     }
133     if (mis->to_src_file) {
134         qemu_file_shutdown(mis->to_src_file);
135     }
136 
137     old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
138                                    FAILOVER_STATUS_COMPLETED);
139     if (old_state != FAILOVER_STATUS_ACTIVE) {
140         error_report("Incorrect state (%s) while doing failover for "
141                      "secondary VM", FailoverStatus_str(old_state));
142         return;
143     }
144     /* Notify COLO incoming thread that failover work is finished */
145     qemu_sem_post(&mis->colo_incoming_sem);
146 
147     /* For Secondary VM, jump to incoming co */
148     if (mis->colo_incoming_co) {
149         qemu_coroutine_enter(mis->colo_incoming_co);
150     }
151 }
152 
153 static void primary_vm_do_failover(void)
154 {
155     MigrationState *s = migrate_get_current();
156     int old_state;
157     Error *local_err = NULL;
158 
159     migrate_set_state(&s->state, MIGRATION_STATUS_COLO,
160                       MIGRATION_STATUS_COMPLETED);
161     /*
162      * kick COLO thread which might wait at
163      * qemu_sem_wait(&s->colo_checkpoint_sem).
164      */
165     colo_checkpoint_notify(s);
166 
167     /*
168      * Wake up COLO thread which may blocked in recv() or send(),
169      * The s->rp_state.from_dst_file and s->to_dst_file may use the
170      * same fd, but we still shutdown the fd for twice, it is harmless.
171      */
172     if (s->to_dst_file) {
173         qemu_file_shutdown(s->to_dst_file);
174     }
175     if (s->rp_state.from_dst_file) {
176         qemu_file_shutdown(s->rp_state.from_dst_file);
177     }
178 
179     old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
180                                    FAILOVER_STATUS_COMPLETED);
181     if (old_state != FAILOVER_STATUS_ACTIVE) {
182         error_report("Incorrect state (%s) while doing failover for Primary VM",
183                      FailoverStatus_str(old_state));
184         return;
185     }
186 
187     replication_stop_all(true, &local_err);
188     if (local_err) {
189         error_report_err(local_err);
190         local_err = NULL;
191     }
192 
193     /* Notify COLO thread that failover work is finished */
194     qemu_sem_post(&s->colo_exit_sem);
195 }
196 
197 COLOMode get_colo_mode(void)
198 {
199     if (migration_in_colo_state()) {
200         return COLO_MODE_PRIMARY;
201     } else if (migration_incoming_in_colo_state()) {
202         return COLO_MODE_SECONDARY;
203     } else {
204         return COLO_MODE_NONE;
205     }
206 }
207 
208 void colo_do_failover(void)
209 {
210     /* Make sure VM stopped while failover happened. */
211     if (!colo_runstate_is_stopped()) {
212         vm_stop_force_state(RUN_STATE_COLO);
213     }
214 
215     switch (last_colo_mode = get_colo_mode()) {
216     case COLO_MODE_PRIMARY:
217         primary_vm_do_failover();
218         break;
219     case COLO_MODE_SECONDARY:
220         secondary_vm_do_failover();
221         break;
222     default:
223         error_report("colo_do_failover failed because the colo mode"
224                      " could not be obtained");
225     }
226 }
227 
228 void qmp_xen_set_replication(bool enable, bool primary,
229                              bool has_failover, bool failover,
230                              Error **errp)
231 {
232     ReplicationMode mode = primary ?
233                            REPLICATION_MODE_PRIMARY :
234                            REPLICATION_MODE_SECONDARY;
235 
236     if (has_failover && enable) {
237         error_setg(errp, "Parameter 'failover' is only for"
238                    " stopping replication");
239         return;
240     }
241 
242     if (enable) {
243         replication_start_all(mode, errp);
244     } else {
245         if (!has_failover) {
246             failover = NULL;
247         }
248         replication_stop_all(failover, failover ? NULL : errp);
249     }
250 }
251 
252 ReplicationStatus *qmp_query_xen_replication_status(Error **errp)
253 {
254     Error *err = NULL;
255     ReplicationStatus *s = g_new0(ReplicationStatus, 1);
256 
257     replication_get_error_all(&err);
258     if (err) {
259         s->error = true;
260         s->desc = g_strdup(error_get_pretty(err));
261     } else {
262         s->error = false;
263     }
264 
265     error_free(err);
266     return s;
267 }
268 
269 void qmp_xen_colo_do_checkpoint(Error **errp)
270 {
271     Error *err = NULL;
272 
273     replication_do_checkpoint_all(&err);
274     if (err) {
275         error_propagate(errp, err);
276         return;
277     }
278     /* Notify all filters of all NIC to do checkpoint */
279     colo_notify_filters_event(COLO_EVENT_CHECKPOINT, errp);
280 }
281 
282 COLOStatus *qmp_query_colo_status(Error **errp)
283 {
284     COLOStatus *s = g_new0(COLOStatus, 1);
285 
286     s->mode = get_colo_mode();
287     s->last_mode = last_colo_mode;
288 
289     switch (failover_get_state()) {
290     case FAILOVER_STATUS_NONE:
291         s->reason = COLO_EXIT_REASON_NONE;
292         break;
293     case FAILOVER_STATUS_COMPLETED:
294         s->reason = COLO_EXIT_REASON_REQUEST;
295         break;
296     default:
297         if (migration_in_colo_state()) {
298             s->reason = COLO_EXIT_REASON_PROCESSING;
299         } else {
300             s->reason = COLO_EXIT_REASON_ERROR;
301         }
302     }
303 
304     return s;
305 }
306 
307 static void colo_send_message(QEMUFile *f, COLOMessage msg,
308                               Error **errp)
309 {
310     int ret;
311 
312     if (msg >= COLO_MESSAGE__MAX) {
313         error_setg(errp, "%s: Invalid message", __func__);
314         return;
315     }
316     qemu_put_be32(f, msg);
317     qemu_fflush(f);
318 
319     ret = qemu_file_get_error(f);
320     if (ret < 0) {
321         error_setg_errno(errp, -ret, "Can't send COLO message");
322     }
323     trace_colo_send_message(COLOMessage_str(msg));
324 }
325 
326 static void colo_send_message_value(QEMUFile *f, COLOMessage msg,
327                                     uint64_t value, Error **errp)
328 {
329     Error *local_err = NULL;
330     int ret;
331 
332     colo_send_message(f, msg, &local_err);
333     if (local_err) {
334         error_propagate(errp, local_err);
335         return;
336     }
337     qemu_put_be64(f, value);
338     qemu_fflush(f);
339 
340     ret = qemu_file_get_error(f);
341     if (ret < 0) {
342         error_setg_errno(errp, -ret, "Failed to send value for message:%s",
343                          COLOMessage_str(msg));
344     }
345 }
346 
347 static COLOMessage colo_receive_message(QEMUFile *f, Error **errp)
348 {
349     COLOMessage msg;
350     int ret;
351 
352     msg = qemu_get_be32(f);
353     ret = qemu_file_get_error(f);
354     if (ret < 0) {
355         error_setg_errno(errp, -ret, "Can't receive COLO message");
356         return msg;
357     }
358     if (msg >= COLO_MESSAGE__MAX) {
359         error_setg(errp, "%s: Invalid message", __func__);
360         return msg;
361     }
362     trace_colo_receive_message(COLOMessage_str(msg));
363     return msg;
364 }
365 
366 static void colo_receive_check_message(QEMUFile *f, COLOMessage expect_msg,
367                                        Error **errp)
368 {
369     COLOMessage msg;
370     Error *local_err = NULL;
371 
372     msg = colo_receive_message(f, &local_err);
373     if (local_err) {
374         error_propagate(errp, local_err);
375         return;
376     }
377     if (msg != expect_msg) {
378         error_setg(errp, "Unexpected COLO message %d, expected %d",
379                           msg, expect_msg);
380     }
381 }
382 
383 static uint64_t colo_receive_message_value(QEMUFile *f, uint32_t expect_msg,
384                                            Error **errp)
385 {
386     Error *local_err = NULL;
387     uint64_t value;
388     int ret;
389 
390     colo_receive_check_message(f, expect_msg, &local_err);
391     if (local_err) {
392         error_propagate(errp, local_err);
393         return 0;
394     }
395 
396     value = qemu_get_be64(f);
397     ret = qemu_file_get_error(f);
398     if (ret < 0) {
399         error_setg_errno(errp, -ret, "Failed to get value for COLO message: %s",
400                          COLOMessage_str(expect_msg));
401     }
402     return value;
403 }
404 
405 static int colo_do_checkpoint_transaction(MigrationState *s,
406                                           QIOChannelBuffer *bioc,
407                                           QEMUFile *fb)
408 {
409     Error *local_err = NULL;
410     int ret = -1;
411 
412     colo_send_message(s->to_dst_file, COLO_MESSAGE_CHECKPOINT_REQUEST,
413                       &local_err);
414     if (local_err) {
415         goto out;
416     }
417 
418     colo_receive_check_message(s->rp_state.from_dst_file,
419                     COLO_MESSAGE_CHECKPOINT_REPLY, &local_err);
420     if (local_err) {
421         goto out;
422     }
423     /* Reset channel-buffer directly */
424     qio_channel_io_seek(QIO_CHANNEL(bioc), 0, 0, NULL);
425     bioc->usage = 0;
426 
427     qemu_mutex_lock_iothread();
428     if (failover_get_state() != FAILOVER_STATUS_NONE) {
429         qemu_mutex_unlock_iothread();
430         goto out;
431     }
432     vm_stop_force_state(RUN_STATE_COLO);
433     qemu_mutex_unlock_iothread();
434     trace_colo_vm_state_change("run", "stop");
435     /*
436      * Failover request bh could be called after vm_stop_force_state(),
437      * So we need check failover_request_is_active() again.
438      */
439     if (failover_get_state() != FAILOVER_STATUS_NONE) {
440         goto out;
441     }
442     qemu_mutex_lock_iothread();
443 
444     replication_do_checkpoint_all(&local_err);
445     if (local_err) {
446         qemu_mutex_unlock_iothread();
447         goto out;
448     }
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     if (migrate_auto_converge()) {
464         mig_throttle_counter_reset();
465     }
466     /*
467      * Only save VM's live state, which not including device state.
468      * TODO: We may need a timeout mechanism to prevent COLO process
469      * to be blocked here.
470      */
471     qemu_savevm_live_state(s->to_dst_file);
472 
473     qemu_fflush(fb);
474 
475     /*
476      * We need the size of the VMstate data in Secondary side,
477      * With which we can decide how much data should be read.
478      */
479     colo_send_message_value(s->to_dst_file, COLO_MESSAGE_VMSTATE_SIZE,
480                             bioc->usage, &local_err);
481     if (local_err) {
482         goto out;
483     }
484 
485     qemu_put_buffer(s->to_dst_file, bioc->data, bioc->usage);
486     qemu_fflush(s->to_dst_file);
487     ret = qemu_file_get_error(s->to_dst_file);
488     if (ret < 0) {
489         goto out;
490     }
491 
492     colo_receive_check_message(s->rp_state.from_dst_file,
493                        COLO_MESSAGE_VMSTATE_RECEIVED, &local_err);
494     if (local_err) {
495         goto out;
496     }
497 
498     qemu_event_reset(&s->colo_checkpoint_event);
499     colo_notify_compares_event(NULL, COLO_EVENT_CHECKPOINT, &local_err);
500     if (local_err) {
501         goto out;
502     }
503 
504     colo_receive_check_message(s->rp_state.from_dst_file,
505                        COLO_MESSAGE_VMSTATE_LOADED, &local_err);
506     if (local_err) {
507         goto out;
508     }
509 
510     ret = 0;
511 
512     qemu_mutex_lock_iothread();
513     vm_start();
514     qemu_mutex_unlock_iothread();
515     trace_colo_vm_state_change("stop", "run");
516 
517 out:
518     if (local_err) {
519         error_report_err(local_err);
520     }
521     return ret;
522 }
523 
524 static void colo_compare_notify_checkpoint(Notifier *notifier, void *data)
525 {
526     colo_checkpoint_notify(data);
527 }
528 
529 static void colo_process_checkpoint(MigrationState *s)
530 {
531     QIOChannelBuffer *bioc;
532     QEMUFile *fb = NULL;
533     Error *local_err = NULL;
534     int ret;
535 
536     if (get_colo_mode() != COLO_MODE_PRIMARY) {
537         error_report("COLO mode must be COLO_MODE_PRIMARY");
538         return;
539     }
540 
541     failover_init_state();
542 
543     s->rp_state.from_dst_file = qemu_file_get_return_path(s->to_dst_file);
544     if (!s->rp_state.from_dst_file) {
545         error_report("Open QEMUFile from_dst_file failed");
546         goto out;
547     }
548 
549     packets_compare_notifier.notify = colo_compare_notify_checkpoint;
550     colo_compare_register_notifier(&packets_compare_notifier);
551 
552     /*
553      * Wait for Secondary finish loading VM states and enter COLO
554      * restore.
555      */
556     colo_receive_check_message(s->rp_state.from_dst_file,
557                        COLO_MESSAGE_CHECKPOINT_READY, &local_err);
558     if (local_err) {
559         goto out;
560     }
561     bioc = qio_channel_buffer_new(COLO_BUFFER_BASE_SIZE);
562     fb = qemu_file_new_output(QIO_CHANNEL(bioc));
563     object_unref(OBJECT(bioc));
564 
565     qemu_mutex_lock_iothread();
566     replication_start_all(REPLICATION_MODE_PRIMARY, &local_err);
567     if (local_err) {
568         qemu_mutex_unlock_iothread();
569         goto out;
570     }
571 
572     vm_start();
573     qemu_mutex_unlock_iothread();
574     trace_colo_vm_state_change("stop", "run");
575 
576     timer_mod(s->colo_delay_timer, qemu_clock_get_ms(QEMU_CLOCK_HOST) +
577               migrate_checkpoint_delay());
578 
579     while (s->state == MIGRATION_STATUS_COLO) {
580         if (failover_get_state() != FAILOVER_STATUS_NONE) {
581             error_report("failover request");
582             goto out;
583         }
584 
585         qemu_event_wait(&s->colo_checkpoint_event);
586 
587         if (s->state != MIGRATION_STATUS_COLO) {
588             goto out;
589         }
590         ret = colo_do_checkpoint_transaction(s, bioc, fb);
591         if (ret < 0) {
592             goto out;
593         }
594     }
595 
596 out:
597     /* Throw the unreported error message after exited from loop */
598     if (local_err) {
599         error_report_err(local_err);
600     }
601 
602     if (fb) {
603         qemu_fclose(fb);
604     }
605 
606     /*
607      * There are only two reasons we can get here, some error happened
608      * or the user triggered failover.
609      */
610     switch (failover_get_state()) {
611     case FAILOVER_STATUS_COMPLETED:
612         qapi_event_send_colo_exit(COLO_MODE_PRIMARY,
613                                   COLO_EXIT_REASON_REQUEST);
614         break;
615     default:
616         qapi_event_send_colo_exit(COLO_MODE_PRIMARY,
617                                   COLO_EXIT_REASON_ERROR);
618     }
619 
620     /* Hope this not to be too long to wait here */
621     qemu_sem_wait(&s->colo_exit_sem);
622     qemu_sem_destroy(&s->colo_exit_sem);
623 
624     /*
625      * It is safe to unregister notifier after failover finished.
626      * Besides, colo_delay_timer and colo_checkpoint_sem can't be
627      * released before unregister notifier, or there will be use-after-free
628      * error.
629      */
630     colo_compare_unregister_notifier(&packets_compare_notifier);
631     timer_free(s->colo_delay_timer);
632     qemu_event_destroy(&s->colo_checkpoint_event);
633 
634     /*
635      * Must be called after failover BH is completed,
636      * Or the failover BH may shutdown the wrong fd that
637      * re-used by other threads after we release here.
638      */
639     if (s->rp_state.from_dst_file) {
640         qemu_fclose(s->rp_state.from_dst_file);
641         s->rp_state.from_dst_file = NULL;
642     }
643 }
644 
645 void migrate_start_colo_process(MigrationState *s)
646 {
647     qemu_mutex_unlock_iothread();
648     qemu_event_init(&s->colo_checkpoint_event, false);
649     s->colo_delay_timer =  timer_new_ms(QEMU_CLOCK_HOST,
650                                 colo_checkpoint_notify, s);
651 
652     qemu_sem_init(&s->colo_exit_sem, 0);
653     colo_process_checkpoint(s);
654     qemu_mutex_lock_iothread();
655 }
656 
657 static void colo_incoming_process_checkpoint(MigrationIncomingState *mis,
658                       QEMUFile *fb, QIOChannelBuffer *bioc, Error **errp)
659 {
660     uint64_t total_size;
661     uint64_t value;
662     Error *local_err = NULL;
663     int ret;
664 
665     qemu_mutex_lock_iothread();
666     vm_stop_force_state(RUN_STATE_COLO);
667     qemu_mutex_unlock_iothread();
668     trace_colo_vm_state_change("run", "stop");
669 
670     /* FIXME: This is unnecessary for periodic checkpoint mode */
671     colo_send_message(mis->to_src_file, COLO_MESSAGE_CHECKPOINT_REPLY,
672                  &local_err);
673     if (local_err) {
674         error_propagate(errp, local_err);
675         return;
676     }
677 
678     colo_receive_check_message(mis->from_src_file,
679                        COLO_MESSAGE_VMSTATE_SEND, &local_err);
680     if (local_err) {
681         error_propagate(errp, local_err);
682         return;
683     }
684 
685     qemu_mutex_lock_iothread();
686     cpu_synchronize_all_states();
687     ret = qemu_loadvm_state_main(mis->from_src_file, mis);
688     qemu_mutex_unlock_iothread();
689 
690     if (ret < 0) {
691         error_setg(errp, "Load VM's live state (ram) error");
692         return;
693     }
694 
695     value = colo_receive_message_value(mis->from_src_file,
696                              COLO_MESSAGE_VMSTATE_SIZE, &local_err);
697     if (local_err) {
698         error_propagate(errp, local_err);
699         return;
700     }
701 
702     /*
703      * Read VM device state data into channel buffer,
704      * It's better to re-use the memory allocated.
705      * Here we need to handle the channel buffer directly.
706      */
707     if (value > bioc->capacity) {
708         bioc->capacity = value;
709         bioc->data = g_realloc(bioc->data, bioc->capacity);
710     }
711     total_size = qemu_get_buffer(mis->from_src_file, bioc->data, value);
712     if (total_size != value) {
713         error_setg(errp, "Got %" PRIu64 " VMState data, less than expected"
714                     " %" PRIu64, total_size, value);
715         return;
716     }
717     bioc->usage = total_size;
718     qio_channel_io_seek(QIO_CHANNEL(bioc), 0, 0, NULL);
719 
720     colo_send_message(mis->to_src_file, COLO_MESSAGE_VMSTATE_RECEIVED,
721                  &local_err);
722     if (local_err) {
723         error_propagate(errp, local_err);
724         return;
725     }
726 
727     qemu_mutex_lock_iothread();
728     vmstate_loading = true;
729     colo_flush_ram_cache();
730     ret = qemu_load_device_state(fb);
731     if (ret < 0) {
732         error_setg(errp, "COLO: load device state failed");
733         vmstate_loading = false;
734         qemu_mutex_unlock_iothread();
735         return;
736     }
737 
738     replication_get_error_all(&local_err);
739     if (local_err) {
740         error_propagate(errp, local_err);
741         vmstate_loading = false;
742         qemu_mutex_unlock_iothread();
743         return;
744     }
745 
746     /* discard colo disk buffer */
747     replication_do_checkpoint_all(&local_err);
748     if (local_err) {
749         error_propagate(errp, local_err);
750         vmstate_loading = false;
751         qemu_mutex_unlock_iothread();
752         return;
753     }
754     /* Notify all filters of all NIC to do checkpoint */
755     colo_notify_filters_event(COLO_EVENT_CHECKPOINT, &local_err);
756 
757     if (local_err) {
758         error_propagate(errp, local_err);
759         vmstate_loading = false;
760         qemu_mutex_unlock_iothread();
761         return;
762     }
763 
764     vmstate_loading = false;
765     vm_start();
766     qemu_mutex_unlock_iothread();
767     trace_colo_vm_state_change("stop", "run");
768 
769     if (failover_get_state() == FAILOVER_STATUS_RELAUNCH) {
770         return;
771     }
772 
773     colo_send_message(mis->to_src_file, COLO_MESSAGE_VMSTATE_LOADED,
774                  &local_err);
775     error_propagate(errp, local_err);
776 }
777 
778 static void colo_wait_handle_message(MigrationIncomingState *mis,
779                 QEMUFile *fb, QIOChannelBuffer *bioc, Error **errp)
780 {
781     COLOMessage msg;
782     Error *local_err = NULL;
783 
784     msg = colo_receive_message(mis->from_src_file, &local_err);
785     if (local_err) {
786         error_propagate(errp, local_err);
787         return;
788     }
789 
790     switch (msg) {
791     case COLO_MESSAGE_CHECKPOINT_REQUEST:
792         colo_incoming_process_checkpoint(mis, fb, bioc, errp);
793         break;
794     default:
795         error_setg(errp, "Got unknown COLO message: %d", msg);
796         break;
797     }
798 }
799 
800 void colo_shutdown(void)
801 {
802     MigrationIncomingState *mis = NULL;
803     MigrationState *s = NULL;
804 
805     switch (get_colo_mode()) {
806     case COLO_MODE_PRIMARY:
807         s = migrate_get_current();
808         qemu_event_set(&s->colo_checkpoint_event);
809         qemu_sem_post(&s->colo_exit_sem);
810         break;
811     case COLO_MODE_SECONDARY:
812         mis = migration_incoming_get_current();
813         qemu_sem_post(&mis->colo_incoming_sem);
814         break;
815     default:
816         break;
817     }
818 }
819 
820 static void *colo_process_incoming_thread(void *opaque)
821 {
822     MigrationIncomingState *mis = opaque;
823     QEMUFile *fb = NULL;
824     QIOChannelBuffer *bioc = NULL; /* Cache incoming device state */
825     Error *local_err = NULL;
826 
827     rcu_register_thread();
828     qemu_sem_init(&mis->colo_incoming_sem, 0);
829 
830     migrate_set_state(&mis->state, MIGRATION_STATUS_ACTIVE,
831                       MIGRATION_STATUS_COLO);
832 
833     if (get_colo_mode() != COLO_MODE_SECONDARY) {
834         error_report("COLO mode must be COLO_MODE_SECONDARY");
835         return NULL;
836     }
837 
838     failover_init_state();
839 
840     mis->to_src_file = qemu_file_get_return_path(mis->from_src_file);
841     if (!mis->to_src_file) {
842         error_report("COLO incoming thread: Open QEMUFile to_src_file failed");
843         goto out;
844     }
845     /*
846      * Note: the communication between Primary side and Secondary side
847      * should be sequential, we set the fd to unblocked in migration incoming
848      * coroutine, and here we are in the COLO incoming thread, so it is ok to
849      * set the fd back to blocked.
850      */
851     qemu_file_set_blocking(mis->from_src_file, true);
852 
853     colo_incoming_start_dirty_log();
854 
855     bioc = qio_channel_buffer_new(COLO_BUFFER_BASE_SIZE);
856     fb = qemu_file_new_input(QIO_CHANNEL(bioc));
857     object_unref(OBJECT(bioc));
858 
859     qemu_mutex_lock_iothread();
860     replication_start_all(REPLICATION_MODE_SECONDARY, &local_err);
861     if (local_err) {
862         qemu_mutex_unlock_iothread();
863         goto out;
864     }
865     vm_start();
866     qemu_mutex_unlock_iothread();
867     trace_colo_vm_state_change("stop", "run");
868 
869     colo_send_message(mis->to_src_file, COLO_MESSAGE_CHECKPOINT_READY,
870                       &local_err);
871     if (local_err) {
872         goto out;
873     }
874 
875     while (mis->state == MIGRATION_STATUS_COLO) {
876         colo_wait_handle_message(mis, fb, bioc, &local_err);
877         if (local_err) {
878             error_report_err(local_err);
879             break;
880         }
881 
882         if (failover_get_state() == FAILOVER_STATUS_RELAUNCH) {
883             failover_set_state(FAILOVER_STATUS_RELAUNCH,
884                             FAILOVER_STATUS_NONE);
885             failover_request_active(NULL);
886             break;
887         }
888 
889         if (failover_get_state() != FAILOVER_STATUS_NONE) {
890             error_report("failover request");
891             break;
892         }
893     }
894 
895 out:
896     /*
897      * There are only two reasons we can get here, some error happened
898      * or the user triggered failover.
899      */
900     switch (failover_get_state()) {
901     case FAILOVER_STATUS_COMPLETED:
902         qapi_event_send_colo_exit(COLO_MODE_SECONDARY,
903                                   COLO_EXIT_REASON_REQUEST);
904         break;
905     default:
906         qapi_event_send_colo_exit(COLO_MODE_SECONDARY,
907                                   COLO_EXIT_REASON_ERROR);
908     }
909 
910     if (fb) {
911         qemu_fclose(fb);
912     }
913 
914     /* Hope this not to be too long to loop here */
915     qemu_sem_wait(&mis->colo_incoming_sem);
916     qemu_sem_destroy(&mis->colo_incoming_sem);
917 
918     rcu_unregister_thread();
919     return NULL;
920 }
921 
922 int coroutine_fn colo_incoming_co(void)
923 {
924     MigrationIncomingState *mis = migration_incoming_get_current();
925     Error *local_err = NULL;
926     QemuThread th;
927 
928     assert(qemu_mutex_iothread_locked());
929 
930     if (!migration_incoming_colo_enabled()) {
931         return 0;
932     }
933 
934     /* Make sure all file formats throw away their mutable metadata */
935     bdrv_activate_all(&local_err);
936     if (local_err) {
937         error_report_err(local_err);
938         return -EINVAL;
939     }
940 
941     qemu_thread_create(&th, "COLO incoming", colo_process_incoming_thread,
942                        mis, QEMU_THREAD_JOINABLE);
943 
944     mis->colo_incoming_co = qemu_coroutine_self();
945     qemu_coroutine_yield();
946     mis->colo_incoming_co = NULL;
947 
948     qemu_mutex_unlock_iothread();
949     /* Wait checkpoint incoming thread exit before free resource */
950     qemu_thread_join(&th);
951     qemu_mutex_lock_iothread();
952 
953     /* We hold the global iothread lock, so it is safe here */
954     colo_release_ram_cache();
955 
956     return 0;
957 }
958