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 "qemu/timer.h"
  15 #include "sysemu/sysemu.h"
  16 #include "migration/colo.h"
  17 #include "io/channel-buffer.h"
  18 #include "trace.h"
  19 #include "qemu/error-report.h"
  20 #include "qapi/error.h"
  21 #include "migration/failover.h"
  22 #include "replication.h"
  23 #include "qmp-commands.h"
  24
  25 static bool vmstate_loading;
  26
  27 #define COLO_BUFFER_BASE_SIZE (4 * 1024 * 1024)
  28
  29 bool colo_supported(void)
  30 {
  31     return true;
  32 }
  33
  34 bool migration_in_colo_state(void)
  35 {
  36     MigrationState *s = migrate_get_current();
  37
  38     return (s->state == MIGRATION_STATUS_COLO);
  39 }
  40
  41 bool migration_incoming_in_colo_state(void)
  42 {
  43     MigrationIncomingState *mis = migration_incoming_get_current();
  44
  45     return mis && (mis->state == MIGRATION_STATUS_COLO);
  46 }
  47
  48 static bool colo_runstate_is_stopped(void)
  49 {
  50     return runstate_check(RUN_STATE_COLO) || !runstate_is_running();
  51 }
  52
  53 static void secondary_vm_do_failover(void)
  54 {
  55     int old_state;
  56     MigrationIncomingState *mis = migration_incoming_get_current();
  57
  58     /* Can not do failover during the process of VM's loading VMstate, Or
  59      * it will break the secondary VM.
  60      */
  61     if (vmstate_loading) {
  62         old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
  63                         FAILOVER_STATUS_RELAUNCH);
  64         if (old_state != FAILOVER_STATUS_ACTIVE) {
  65             error_report("Unknown error while do failover for secondary VM,"
  66                          "old_state: %s", FailoverStatus_lookup[old_state]);
  67         }
  68         return;
  69     }
  70
  71     migrate_set_state(&mis->state, MIGRATION_STATUS_COLO,
  72                       MIGRATION_STATUS_COMPLETED);
  73
  74     if (!autostart) {
  75         error_report("\"-S\" qemu option will be ignored in secondary side");
  76         /* recover runstate to normal migration finish state */
  77         autostart = true;
  78     }
  79     /*
  80      * Make sure COLO incoming thread not block in recv or send,
  81      * If mis->from_src_file and mis->to_src_file use the same fd,
  82      * The second shutdown() will return -1, we ignore this value,
  83      * It is harmless.
  84      */
  85     if (mis->from_src_file) {
  86         qemu_file_shutdown(mis->from_src_file);
  87     }
  88     if (mis->to_src_file) {
  89         qemu_file_shutdown(mis->to_src_file);
  90     }
  91
  92     old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
  93                                    FAILOVER_STATUS_COMPLETED);
  94     if (old_state != FAILOVER_STATUS_ACTIVE) {
  95         error_report("Incorrect state (%s) while doing failover for "
  96                      "secondary VM", FailoverStatus_lookup[old_state]);
  97         return;
  98     }
  99     /* Notify COLO incoming thread that failover work is finished */
 100     qemu_sem_post(&mis->colo_incoming_sem);
 101     /* For Secondary VM, jump to incoming co */
 102     if (mis->migration_incoming_co) {
 103         qemu_coroutine_enter(mis->migration_incoming_co);
 104     }
 105 }
 106
 107 static void primary_vm_do_failover(void)
 108 {
 109     MigrationState *s = migrate_get_current();
 110     int old_state;
 111
 112     migrate_set_state(&s->state, MIGRATION_STATUS_COLO,
 113                       MIGRATION_STATUS_COMPLETED);
 114
 115     /*
 116      * Wake up COLO thread which may blocked in recv() or send(),
 117      * The s->rp_state.from_dst_file and s->to_dst_file may use the
 118      * same fd, but we still shutdown the fd for twice, it is harmless.
 119      */
 120     if (s->to_dst_file) {
 121         qemu_file_shutdown(s->to_dst_file);
 122     }
 123     if (s->rp_state.from_dst_file) {
 124         qemu_file_shutdown(s->rp_state.from_dst_file);
 125     }
 126
 127     old_state = failover_set_state(FAILOVER_STATUS_ACTIVE,
 128                                    FAILOVER_STATUS_COMPLETED);
 129     if (old_state != FAILOVER_STATUS_ACTIVE) {
 130         error_report("Incorrect state (%s) while doing failover for Primary VM",
 131                      FailoverStatus_lookup[old_state]);
 132         return;
 133     }
 134     /* Notify COLO thread that failover work is finished */
 135     qemu_sem_post(&s->colo_exit_sem);
 136 }
 137
 138 void colo_do_failover(MigrationState *s)
 139 {
 140     /* Make sure VM stopped while failover happened. */
 141     if (!colo_runstate_is_stopped()) {
 142         vm_stop_force_state(RUN_STATE_COLO);
 143     }
 144
 145     if (get_colo_mode() == COLO_MODE_PRIMARY) {
 146         primary_vm_do_failover();
 147     } else {
 148         secondary_vm_do_failover();
 149     }
 150 }
 151
 152 void qmp_xen_set_replication(bool enable, bool primary,
 153                              bool has_failover, bool failover,
 154                              Error **errp)
 155 {
 156 #ifdef CONFIG_REPLICATION
 157     ReplicationMode mode = primary ?
 158                            REPLICATION_MODE_PRIMARY :
 159                            REPLICATION_MODE_SECONDARY;
 160
 161     if (has_failover && enable) {
 162         error_setg(errp, "Parameter 'failover' is only for"
 163                    " stopping replication");
 164         return;
 165     }
 166
 167     if (enable) {
 168         replication_start_all(mode, errp);
 169     } else {
 170         if (!has_failover) {
 171             failover = NULL;
 172         }
 173         replication_stop_all(failover, failover ? NULL : errp);
 174     }
 175 #else
 176     abort();
 177 #endif
 178 }
 179
 180 ReplicationStatus *qmp_query_xen_replication_status(Error **errp)
 181 {
 182 #ifdef CONFIG_REPLICATION
 183     Error *err = NULL;
 184     ReplicationStatus *s = g_new0(ReplicationStatus, 1);
 185
 186     replication_get_error_all(&err);
 187     if (err) {
 188         s->error = true;
 189         s->has_desc = true;
 190         s->desc = g_strdup(error_get_pretty(err));
 191     } else {
 192         s->error = false;
 193     }
 194
 195     error_free(err);
 196     return s;
 197 #else
 198     abort();
 199 #endif
 200 }
 201
 202 void qmp_xen_colo_do_checkpoint(Error **errp)
 203 {
 204 #ifdef CONFIG_REPLICATION
 205     replication_do_checkpoint_all(errp);
 206 #else
 207     abort();
 208 #endif
 209 }
 210
 211 static void colo_send_message(QEMUFile *f, COLOMessage msg,
 212                               Error **errp)
 213 {
 214     int ret;
 215
 216     if (msg >= COLO_MESSAGE__MAX) {
 217         error_setg(errp, "%s: Invalid message", __func__);
 218         return;
 219     }
 220     qemu_put_be32(f, msg);
 221     qemu_fflush(f);
 222
 223     ret = qemu_file_get_error(f);
 224     if (ret < 0) {
 225         error_setg_errno(errp, -ret, "Can't send COLO message");
 226     }
 227     trace_colo_send_message(COLOMessage_lookup[msg]);
 228 }
 229
 230 static void colo_send_message_value(QEMUFile *f, COLOMessage msg,
 231                                     uint64_t value, Error **errp)
 232 {
 233     Error *local_err = NULL;
 234     int ret;
 235
 236     colo_send_message(f, msg, &local_err);
 237     if (local_err) {
 238         error_propagate(errp, local_err);
 239         return;
 240     }
 241     qemu_put_be64(f, value);
 242     qemu_fflush(f);
 243
 244     ret = qemu_file_get_error(f);
 245     if (ret < 0) {
 246         error_setg_errno(errp, -ret, "Failed to send value for message:%s",
 247                          COLOMessage_lookup[msg]);
 248     }
 249 }
 250
 251 static COLOMessage colo_receive_message(QEMUFile *f, Error **errp)
 252 {
 253     COLOMessage msg;
 254     int ret;
 255
 256     msg = qemu_get_be32(f);
 257     ret = qemu_file_get_error(f);
 258     if (ret < 0) {
 259         error_setg_errno(errp, -ret, "Can't receive COLO message");
 260         return msg;
 261     }
 262     if (msg >= COLO_MESSAGE__MAX) {
 263         error_setg(errp, "%s: Invalid message", __func__);
 264         return msg;
 265     }
 266     trace_colo_receive_message(COLOMessage_lookup[msg]);
 267     return msg;
 268 }
 269
 270 static void colo_receive_check_message(QEMUFile *f, COLOMessage expect_msg,
 271                                        Error **errp)
 272 {
 273     COLOMessage msg;
 274     Error *local_err = NULL;
 275
 276     msg = colo_receive_message(f, &local_err);
 277     if (local_err) {
 278         error_propagate(errp, local_err);
 279         return;
 280     }
 281     if (msg != expect_msg) {
 282         error_setg(errp, "Unexpected COLO message %d, expected %d",
 283                           msg, expect_msg);
 284     }
 285 }
 286
 287 static uint64_t colo_receive_message_value(QEMUFile *f, uint32_t expect_msg,
 288                                            Error **errp)
 289 {
 290     Error *local_err = NULL;
 291     uint64_t value;
 292     int ret;
 293
 294     colo_receive_check_message(f, expect_msg, &local_err);
 295     if (local_err) {
 296         error_propagate(errp, local_err);
 297         return 0;
 298     }
 299
 300     value = qemu_get_be64(f);
 301     ret = qemu_file_get_error(f);
 302     if (ret < 0) {
 303         error_setg_errno(errp, -ret, "Failed to get value for COLO message: %s",
 304                          COLOMessage_lookup[expect_msg]);
 305     }
 306     return value;
 307 }
 308
 309 static int colo_do_checkpoint_transaction(MigrationState *s,
 310                                           QIOChannelBuffer *bioc,
 311                                           QEMUFile *fb)
 312 {
 313     Error *local_err = NULL;
 314     int ret = -1;
 315
 316     colo_send_message(s->to_dst_file, COLO_MESSAGE_CHECKPOINT_REQUEST,
 317                       &local_err);
 318     if (local_err) {
 319         goto out;
 320     }
 321
 322     colo_receive_check_message(s->rp_state.from_dst_file,
 323                     COLO_MESSAGE_CHECKPOINT_REPLY, &local_err);
 324     if (local_err) {
 325         goto out;
 326     }
 327     /* Reset channel-buffer directly */
 328     qio_channel_io_seek(QIO_CHANNEL(bioc), 0, 0, NULL);
 329     bioc->usage = 0;
 330
 331     qemu_mutex_lock_iothread();
 332     if (failover_get_state() != FAILOVER_STATUS_NONE) {
 333         qemu_mutex_unlock_iothread();
 334         goto out;
 335     }
 336     vm_stop_force_state(RUN_STATE_COLO);
 337     qemu_mutex_unlock_iothread();
 338     trace_colo_vm_state_change("run", "stop");
 339     /*
 340      * Failover request bh could be called after vm_stop_force_state(),
 341      * So we need check failover_request_is_active() again.
 342      */
 343     if (failover_get_state() != FAILOVER_STATUS_NONE) {
 344         goto out;
 345     }
 346
 347     /* Disable block migration */
 348     s->params.blk = 0;
 349     s->params.shared = 0;
 350     qemu_savevm_state_header(fb);
 351     qemu_savevm_state_begin(fb, &s->params);
 352     qemu_mutex_lock_iothread();
 353     qemu_savevm_state_complete_precopy(fb, false);
 354     qemu_mutex_unlock_iothread();
 355
 356     qemu_fflush(fb);
 357
 358     colo_send_message(s->to_dst_file, COLO_MESSAGE_VMSTATE_SEND, &local_err);
 359     if (local_err) {
 360         goto out;
 361     }
 362     /*
 363      * We need the size of the VMstate data in Secondary side,
 364      * With which we can decide how much data should be read.
 365      */
 366     colo_send_message_value(s->to_dst_file, COLO_MESSAGE_VMSTATE_SIZE,
 367                             bioc->usage, &local_err);
 368     if (local_err) {
 369         goto out;
 370     }
 371
 372     qemu_put_buffer(s->to_dst_file, bioc->data, bioc->usage);
 373     qemu_fflush(s->to_dst_file);
 374     ret = qemu_file_get_error(s->to_dst_file);
 375     if (ret < 0) {
 376         goto out;
 377     }
 378
 379     colo_receive_check_message(s->rp_state.from_dst_file,
 380                        COLO_MESSAGE_VMSTATE_RECEIVED, &local_err);
 381     if (local_err) {
 382         goto out;
 383     }
 384
 385     colo_receive_check_message(s->rp_state.from_dst_file,
 386                        COLO_MESSAGE_VMSTATE_LOADED, &local_err);
 387     if (local_err) {
 388         goto out;
 389     }
 390
 391     ret = 0;
 392
 393     qemu_mutex_lock_iothread();
 394     vm_start();
 395     qemu_mutex_unlock_iothread();
 396     trace_colo_vm_state_change("stop", "run");
 397
 398 out:
 399     if (local_err) {
 400         error_report_err(local_err);
 401     }
 402     return ret;
 403 }
 404
 405 static void colo_process_checkpoint(MigrationState *s)
 406 {
 407     QIOChannelBuffer *bioc;
 408     QEMUFile *fb = NULL;
 409     int64_t current_time = qemu_clock_get_ms(QEMU_CLOCK_HOST);
 410     Error *local_err = NULL;
 411     int ret;
 412
 413     failover_init_state();
 414
 415     s->rp_state.from_dst_file = qemu_file_get_return_path(s->to_dst_file);
 416     if (!s->rp_state.from_dst_file) {
 417         error_report("Open QEMUFile from_dst_file failed");
 418         goto out;
 419     }
 420
 421     /*
 422      * Wait for Secondary finish loading VM states and enter COLO
 423      * restore.
 424      */
 425     colo_receive_check_message(s->rp_state.from_dst_file,
 426                        COLO_MESSAGE_CHECKPOINT_READY, &local_err);
 427     if (local_err) {
 428         goto out;
 429     }
 430     bioc = qio_channel_buffer_new(COLO_BUFFER_BASE_SIZE);
 431     fb = qemu_fopen_channel_output(QIO_CHANNEL(bioc));
 432     object_unref(OBJECT(bioc));
 433
 434     qemu_mutex_lock_iothread();
 435     vm_start();
 436     qemu_mutex_unlock_iothread();
 437     trace_colo_vm_state_change("stop", "run");
 438
 439     timer_mod(s->colo_delay_timer,
 440             current_time + s->parameters.x_checkpoint_delay);
 441
 442     while (s->state == MIGRATION_STATUS_COLO) {
 443         if (failover_get_state() != FAILOVER_STATUS_NONE) {
 444             error_report("failover request");
 445             goto out;
 446         }
 447
 448         qemu_sem_wait(&s->colo_checkpoint_sem);
 449
 450         ret = colo_do_checkpoint_transaction(s, bioc, fb);
 451         if (ret < 0) {
 452             goto out;
 453         }
 454     }
 455
 456 out:
 457     /* Throw the unreported error message after exited from loop */
 458     if (local_err) {
 459         error_report_err(local_err);
 460     }
 461
 462     if (fb) {
 463         qemu_fclose(fb);
 464     }
 465
 466     timer_del(s->colo_delay_timer);
 467
 468     /* Hope this not to be too long to wait here */
 469     qemu_sem_wait(&s->colo_exit_sem);
 470     qemu_sem_destroy(&s->colo_exit_sem);
 471     /*
 472      * Must be called after failover BH is completed,
 473      * Or the failover BH may shutdown the wrong fd that
 474      * re-used by other threads after we release here.
 475      */
 476     if (s->rp_state.from_dst_file) {
 477         qemu_fclose(s->rp_state.from_dst_file);
 478     }
 479 }
 480
 481 void colo_checkpoint_notify(void *opaque)
 482 {
 483     MigrationState *s = opaque;
 484     int64_t next_notify_time;
 485
 486     qemu_sem_post(&s->colo_checkpoint_sem);
 487     s->colo_checkpoint_time = qemu_clock_get_ms(QEMU_CLOCK_HOST);
 488     next_notify_time = s->colo_checkpoint_time +
 489                     s->parameters.x_checkpoint_delay;
 490     timer_mod(s->colo_delay_timer, next_notify_time);
 491 }
 492
 493 void migrate_start_colo_process(MigrationState *s)
 494 {
 495     qemu_mutex_unlock_iothread();
 496     qemu_sem_init(&s->colo_checkpoint_sem, 0);
 497     s->colo_delay_timer =  timer_new_ms(QEMU_CLOCK_HOST,
 498                                 colo_checkpoint_notify, s);
 499
 500     qemu_sem_init(&s->colo_exit_sem, 0);
 501     migrate_set_state(&s->state, MIGRATION_STATUS_ACTIVE,
 502                       MIGRATION_STATUS_COLO);
 503     colo_process_checkpoint(s);
 504     qemu_mutex_lock_iothread();
 505 }
 506
 507 static void colo_wait_handle_message(QEMUFile *f, int *checkpoint_request,
 508                                      Error **errp)
 509 {
 510     COLOMessage msg;
 511     Error *local_err = NULL;
 512
 513     msg = colo_receive_message(f, &local_err);
 514     if (local_err) {
 515         error_propagate(errp, local_err);
 516         return;
 517     }
 518
 519     switch (msg) {
 520     case COLO_MESSAGE_CHECKPOINT_REQUEST:
 521         *checkpoint_request = 1;
 522         break;
 523     default:
 524         *checkpoint_request = 0;
 525         error_setg(errp, "Got unknown COLO message: %d", msg);
 526         break;
 527     }
 528 }
 529
 530 void *colo_process_incoming_thread(void *opaque)
 531 {
 532     MigrationIncomingState *mis = opaque;
 533     QEMUFile *fb = NULL;
 534     QIOChannelBuffer *bioc = NULL; /* Cache incoming device state */
 535     uint64_t total_size;
 536     uint64_t value;
 537     Error *local_err = NULL;
 538
 539     qemu_sem_init(&mis->colo_incoming_sem, 0);
 540
 541     migrate_set_state(&mis->state, MIGRATION_STATUS_ACTIVE,
 542                       MIGRATION_STATUS_COLO);
 543
 544     failover_init_state();
 545
 546     mis->to_src_file = qemu_file_get_return_path(mis->from_src_file);
 547     if (!mis->to_src_file) {
 548         error_report("COLO incoming thread: Open QEMUFile to_src_file failed");
 549         goto out;
 550     }
 551     /*
 552      * Note: the communication between Primary side and Secondary side
 553      * should be sequential, we set the fd to unblocked in migration incoming
 554      * coroutine, and here we are in the COLO incoming thread, so it is ok to
 555      * set the fd back to blocked.
 556      */
 557     qemu_file_set_blocking(mis->from_src_file, true);
 558
 559     bioc = qio_channel_buffer_new(COLO_BUFFER_BASE_SIZE);
 560     fb = qemu_fopen_channel_input(QIO_CHANNEL(bioc));
 561     object_unref(OBJECT(bioc));
 562
 563     colo_send_message(mis->to_src_file, COLO_MESSAGE_CHECKPOINT_READY,
 564                       &local_err);
 565     if (local_err) {
 566         goto out;
 567     }
 568
 569     while (mis->state == MIGRATION_STATUS_COLO) {
 570         int request = 0;
 571
 572         colo_wait_handle_message(mis->from_src_file, &request, &local_err);
 573         if (local_err) {
 574             goto out;
 575         }
 576         assert(request);
 577         if (failover_get_state() != FAILOVER_STATUS_NONE) {
 578             error_report("failover request");
 579             goto out;
 580         }
 581
 582         /* FIXME: This is unnecessary for periodic checkpoint mode */
 583         colo_send_message(mis->to_src_file, COLO_MESSAGE_CHECKPOINT_REPLY,
 584                      &local_err);
 585         if (local_err) {
 586             goto out;
 587         }
 588
 589         colo_receive_check_message(mis->from_src_file,
 590                            COLO_MESSAGE_VMSTATE_SEND, &local_err);
 591         if (local_err) {
 592             goto out;
 593         }
 594
 595         value = colo_receive_message_value(mis->from_src_file,
 596                                  COLO_MESSAGE_VMSTATE_SIZE, &local_err);
 597         if (local_err) {
 598             goto out;
 599         }
 600
 601         /*
 602          * Read VM device state data into channel buffer,
 603          * It's better to re-use the memory allocated.
 604          * Here we need to handle the channel buffer directly.
 605          */
 606         if (value > bioc->capacity) {
 607             bioc->capacity = value;
 608             bioc->data = g_realloc(bioc->data, bioc->capacity);
 609         }
 610         total_size = qemu_get_buffer(mis->from_src_file, bioc->data, value);
 611         if (total_size != value) {
 612             error_report("Got %" PRIu64 " VMState data, less than expected"
 613                         " %" PRIu64, total_size, value);
 614             goto out;
 615         }
 616         bioc->usage = total_size;
 617         qio_channel_io_seek(QIO_CHANNEL(bioc), 0, 0, NULL);
 618
 619         colo_send_message(mis->to_src_file, COLO_MESSAGE_VMSTATE_RECEIVED,
 620                      &local_err);
 621         if (local_err) {
 622             goto out;
 623         }
 624
 625         qemu_mutex_lock_iothread();
 626         qemu_system_reset(VMRESET_SILENT);
 627         vmstate_loading = true;
 628         if (qemu_loadvm_state(fb) < 0) {
 629             error_report("COLO: loadvm failed");
 630             qemu_mutex_unlock_iothread();
 631             goto out;
 632         }
 633
 634         vmstate_loading = false;
 635         qemu_mutex_unlock_iothread();
 636
 637         if (failover_get_state() == FAILOVER_STATUS_RELAUNCH) {
 638             failover_set_state(FAILOVER_STATUS_RELAUNCH,
 639                             FAILOVER_STATUS_NONE);
 640             failover_request_active(NULL);
 641             goto out;
 642         }
 643
 644         colo_send_message(mis->to_src_file, COLO_MESSAGE_VMSTATE_LOADED,
 645                      &local_err);
 646         if (local_err) {
 647             goto out;
 648         }
 649     }
 650
 651 out:
 652     vmstate_loading = false;
 653     /* Throw the unreported error message after exited from loop */
 654     if (local_err) {
 655         error_report_err(local_err);
 656     }
 657
 658     if (fb) {
 659         qemu_fclose(fb);
 660     }
 661
 662     /* Hope this not to be too long to loop here */
 663     qemu_sem_wait(&mis->colo_incoming_sem);
 664     qemu_sem_destroy(&mis->colo_incoming_sem);
 665     /* Must be called after failover BH is completed */
 666     if (mis->to_src_file) {
 667         qemu_fclose(mis->to_src_file);
 668     }
 669     migration_incoming_exit_colo();
 670
 671     return NULL;
 672 }