migration/colo.c

   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(void)
 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 }