migration.c 71.2 KB
Newer Older
A
aliguori 已提交
1 2 3 4 5 6 7 8 9 10 11
/*
 * QEMU live migration
 *
 * Copyright IBM, Corp. 2008
 *
 * Authors:
 *  Anthony Liguori   <aliguori@us.ibm.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
12 13
 * Contributions after 2012-01-13 are licensed under the terms of the
 * GNU GPL, version 2 or (at your option) any later version.
A
aliguori 已提交
14 15
 */

P
Peter Maydell 已提交
16
#include "qemu/osdep.h"
17
#include "qemu/cutils.h"
18
#include "qemu/error-report.h"
19
#include "migration/blocker.h"
20
#include "exec.h"
21
#include "fd.h"
22
#include "socket.h"
23
#include "rdma.h"
24
#include "ram.h"
J
Juan Quintela 已提交
25
#include "migration/global_state.h"
26
#include "migration/misc.h"
27
#include "migration.h"
28
#include "savevm.h"
29
#include "qemu-file-channel.h"
J
Juan Quintela 已提交
30
#include "qemu-file.h"
31
#include "migration/vmstate.h"
32
#include "block/block.h"
33
#include "qapi/qmp/qerror.h"
34
#include "qemu/rcu.h"
35
#include "block.h"
36
#include "postcopy-ram.h"
37
#include "qemu/thread.h"
L
Luiz Capitulino 已提交
38
#include "qmp-commands.h"
39
#include "trace.h"
40
#include "qapi-event.h"
41
#include "exec/target_page.h"
42
#include "io/channel-buffer.h"
43
#include "migration/colo.h"
44
#include "hw/boards.h"
P
Peter Xu 已提交
45
#include "monitor/monitor.h"
46

47
#define MAX_THROTTLE  (32 << 20)      /* Migration transfer speed throttling */
A
aliguori 已提交
48

J
Juan Quintela 已提交
49 50 51 52 53
/* Amount of time to allocate to each "chunk" of bandwidth-throttled
 * data. */
#define BUFFER_DELAY     100
#define XFER_LIMIT_RATIO (1000 / BUFFER_DELAY)

54 55 56 57
/* Time in milliseconds we are allowed to stop the source,
 * for sending the last part */
#define DEFAULT_MIGRATE_SET_DOWNTIME 300

58 59 60 61
/* Maximum migrate downtime set to 2000 seconds */
#define MAX_MIGRATE_DOWNTIME_SECONDS 2000
#define MAX_MIGRATE_DOWNTIME (MAX_MIGRATE_DOWNTIME_SECONDS * 1000)

62 63
/* Default compression thread count */
#define DEFAULT_MIGRATE_COMPRESS_THREAD_COUNT 8
64 65 66
/* Default decompression thread count, usually decompression is at
 * least 4 times as fast as compression.*/
#define DEFAULT_MIGRATE_DECOMPRESS_THREAD_COUNT 2
67 68
/*0: means nocompress, 1: best speed, ... 9: best compress ratio */
#define DEFAULT_MIGRATE_COMPRESS_LEVEL 1
69
/* Define default autoconverge cpu throttle migration parameters */
70 71
#define DEFAULT_MIGRATE_CPU_THROTTLE_INITIAL 20
#define DEFAULT_MIGRATE_CPU_THROTTLE_INCREMENT 10
72

73 74 75
/* Migration XBZRLE default cache size */
#define DEFAULT_MIGRATE_CACHE_SIZE (64 * 1024 * 1024)

76 77 78 79 80
/* The delay time (in ms) between two COLO checkpoints
 * Note: Please change this default value to 10000 when we support hybrid mode.
 */
#define DEFAULT_MIGRATE_X_CHECKPOINT_DELAY 200

81 82 83
static NotifierList migration_state_notifiers =
    NOTIFIER_LIST_INITIALIZER(migration_state_notifiers);

D
Dr. David Alan Gilbert 已提交
84 85
static bool deferred_incoming;

86 87 88 89 90 91 92 93 94 95 96 97
/* Messages sent on the return path from destination to source */
enum mig_rp_message_type {
    MIG_RP_MSG_INVALID = 0,  /* Must be 0 */
    MIG_RP_MSG_SHUT,         /* sibling will not send any more RP messages */
    MIG_RP_MSG_PONG,         /* Response to a PING; data (seq: be32 ) */

    MIG_RP_MSG_REQ_PAGES_ID, /* data (start: be64, len: be32, id: string) */
    MIG_RP_MSG_REQ_PAGES,    /* data (start: be64, len: be32) */

    MIG_RP_MSG_MAX
};

98 99 100 101
/* When we add fault tolerance, we could have several
   migrations at once.  For now we don't need to add
   dynamic creation of migration */

102 103
static MigrationState *current_migration;

104 105
static bool migration_object_check(MigrationState *ms, Error **errp);

106 107
void migration_object_init(void)
{
108
    MachineState *ms = MACHINE(qdev_get_machine());
109
    Error *err = NULL;
110

111 112 113
    /* This can only be called once. */
    assert(!current_migration);
    current_migration = MIGRATION_OBJ(object_new(TYPE_MIGRATION));
114

115 116 117 118 119
    if (!migration_object_check(current_migration, &err)) {
        error_report_err(err);
        exit(1);
    }

120 121 122 123 124 125 126 127
    /*
     * We cannot really do this in migration_instance_init() since at
     * that time global properties are not yet applied, then this
     * value will be definitely replaced by something else.
     */
    if (ms->enforce_config_section) {
        current_migration->send_configuration = true;
    }
128 129
}

130
/* For outgoing */
131
MigrationState *migrate_get_current(void)
132
{
133 134 135
    /* This can only be called after the object created. */
    assert(current_migration);
    return current_migration;
136 137
}

138 139
MigrationIncomingState *migration_incoming_get_current(void)
{
140 141
    static bool once;
    static MigrationIncomingState mis_current;
142

143 144 145 146 147 148 149 150
    if (!once) {
        mis_current.state = MIGRATION_STATUS_NONE;
        memset(&mis_current, 0, sizeof(MigrationIncomingState));
        qemu_mutex_init(&mis_current.rp_mutex);
        qemu_event_init(&mis_current.main_thread_load_event, false);
        once = true;
    }
    return &mis_current;
151 152 153 154
}

void migration_incoming_state_destroy(void)
{
155 156
    struct MigrationIncomingState *mis = migration_incoming_get_current();

157
    if (mis->to_src_file) {
158 159
        /* Tell source that we are done */
        migrate_send_rp_shut(mis, qemu_file_get_error(mis->from_src_file) != 0);
160 161 162 163
        qemu_fclose(mis->to_src_file);
        mis->to_src_file = NULL;
    }

164 165 166 167 168
    if (mis->from_src_file) {
        qemu_fclose(mis->from_src_file);
        mis->from_src_file = NULL;
    }

169
    qemu_event_reset(&mis->main_thread_load_event);
170 171
}

172 173 174 175 176 177 178
static void migrate_generate_event(int new_state)
{
    if (migrate_use_events()) {
        qapi_event_send_migration(new_state, &error_abort);
    }
}

D
Dr. David Alan Gilbert 已提交
179 180 181 182 183 184 185 186 187 188 189 190 191
/*
 * Called on -incoming with a defer: uri.
 * The migration can be started later after any parameters have been
 * changed.
 */
static void deferred_incoming_migration(Error **errp)
{
    if (deferred_incoming) {
        error_setg(errp, "Incoming migration already deferred");
    }
    deferred_incoming = true;
}

192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208
/*
 * Send a message on the return channel back to the source
 * of the migration.
 */
static void migrate_send_rp_message(MigrationIncomingState *mis,
                                    enum mig_rp_message_type message_type,
                                    uint16_t len, void *data)
{
    trace_migrate_send_rp_message((int)message_type, len);
    qemu_mutex_lock(&mis->rp_mutex);
    qemu_put_be16(mis->to_src_file, (unsigned int)message_type);
    qemu_put_be16(mis->to_src_file, len);
    qemu_put_buffer(mis->to_src_file, data, len);
    qemu_fflush(mis->to_src_file);
    qemu_mutex_unlock(&mis->rp_mutex);
}

209 210 211 212 213 214 215 216 217 218
/* Request a range of pages from the source VM at the given
 * start address.
 *   rbname: Name of the RAMBlock to request the page in, if NULL it's the same
 *           as the last request (a name must have been given previously)
 *   Start: Address offset within the RB
 *   Len: Length in bytes required - must be a multiple of pagesize
 */
void migrate_send_rp_req_pages(MigrationIncomingState *mis, const char *rbname,
                               ram_addr_t start, size_t len)
{
S
Stefan Weil 已提交
219
    uint8_t bufc[12 + 1 + 255]; /* start (8), len (4), rbname up to 256 */
220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237
    size_t msglen = 12; /* start + len */

    *(uint64_t *)bufc = cpu_to_be64((uint64_t)start);
    *(uint32_t *)(bufc + 8) = cpu_to_be32((uint32_t)len);

    if (rbname) {
        int rbname_len = strlen(rbname);
        assert(rbname_len < 256);

        bufc[msglen++] = rbname_len;
        memcpy(bufc + msglen, rbname, rbname_len);
        msglen += rbname_len;
        migrate_send_rp_message(mis, MIG_RP_MSG_REQ_PAGES_ID, msglen, bufc);
    } else {
        migrate_send_rp_message(mis, MIG_RP_MSG_REQ_PAGES, msglen, bufc);
    }
}

238
void qemu_start_incoming_migration(const char *uri, Error **errp)
A
aliguori 已提交
239
{
A
aliguori 已提交
240 241
    const char *p;

242
    qapi_event_send_migration(MIGRATION_STATUS_SETUP, &error_abort);
D
Dr. David Alan Gilbert 已提交
243 244 245
    if (!strcmp(uri, "defer")) {
        deferred_incoming_migration(errp);
    } else if (strstart(uri, "tcp:", &p)) {
246
        tcp_start_incoming_migration(p, errp);
M
Michael R. Hines 已提交
247
#ifdef CONFIG_RDMA
D
Dr. David Alan Gilbert 已提交
248
    } else if (strstart(uri, "rdma:", &p)) {
M
Michael R. Hines 已提交
249 250
        rdma_start_incoming_migration(p, errp);
#endif
D
Dr. David Alan Gilbert 已提交
251
    } else if (strstart(uri, "exec:", &p)) {
252
        exec_start_incoming_migration(p, errp);
D
Dr. David Alan Gilbert 已提交
253
    } else if (strstart(uri, "unix:", &p)) {
254
        unix_start_incoming_migration(p, errp);
D
Dr. David Alan Gilbert 已提交
255
    } else if (strstart(uri, "fd:", &p)) {
256
        fd_start_incoming_migration(p, errp);
D
Dr. David Alan Gilbert 已提交
257
    } else {
258
        error_setg(errp, "unknown migration protocol: %s", uri);
J
Juan Quintela 已提交
259
    }
A
aliguori 已提交
260 261
}

262 263 264 265 266
static void process_incoming_migration_bh(void *opaque)
{
    Error *local_err = NULL;
    MigrationIncomingState *mis = opaque;

267 268
    /* Make sure all file formats flush their mutable metadata.
     * If we get an error here, just don't restart the VM yet. */
269
    bdrv_invalidate_cache_all(&local_err);
270
    if (local_err) {
271
        error_report_err(local_err);
272 273 274 275
        local_err = NULL;
        autostart = false;
    }

276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306
    /*
     * This must happen after all error conditions are dealt with and
     * we're sure the VM is going to be running on this host.
     */
    qemu_announce_self();

    /* If global state section was not received or we are in running
       state, we need to obey autostart. Any other state is set with
       runstate_set. */

    if (!global_state_received() ||
        global_state_get_runstate() == RUN_STATE_RUNNING) {
        if (autostart) {
            vm_start();
        } else {
            runstate_set(RUN_STATE_PAUSED);
        }
    } else {
        runstate_set(global_state_get_runstate());
    }
    /*
     * This must happen after any state changes since as soon as an external
     * observer sees this event they might start to prod at the VM assuming
     * it's ready to use.
     */
    migrate_set_state(&mis->state, MIGRATION_STATUS_ACTIVE,
                      MIGRATION_STATUS_COMPLETED);
    qemu_bh_delete(mis->bh);
    migration_incoming_state_destroy();
}

307
static void process_incoming_migration_co(void *opaque)
308
{
309
    MigrationIncomingState *mis = migration_incoming_get_current();
310
    PostcopyState ps;
311 312
    int ret;

313
    assert(mis->from_src_file);
314
    mis->largest_page_size = qemu_ram_pagesize_largest();
315
    postcopy_state_set(POSTCOPY_INCOMING_NONE);
316 317
    migrate_set_state(&mis->state, MIGRATION_STATUS_NONE,
                      MIGRATION_STATUS_ACTIVE);
318
    ret = qemu_loadvm_state(mis->from_src_file);
319

320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340
    ps = postcopy_state_get();
    trace_process_incoming_migration_co_end(ret, ps);
    if (ps != POSTCOPY_INCOMING_NONE) {
        if (ps == POSTCOPY_INCOMING_ADVISE) {
            /*
             * Where a migration had postcopy enabled (and thus went to advise)
             * but managed to complete within the precopy period, we can use
             * the normal exit.
             */
            postcopy_ram_incoming_cleanup(mis);
        } else if (ret >= 0) {
            /*
             * Postcopy was started, cleanup should happen at the end of the
             * postcopy thread.
             */
            trace_process_incoming_migration_co_postcopy_end_main();
            return;
        }
        /* Else if something went wrong then just fall out of the normal exit */
    }

341 342 343 344 345 346 347 348 349 350 351 352
    /* we get COLO info, and know if we are in COLO mode */
    if (!ret && migration_incoming_enable_colo()) {
        mis->migration_incoming_co = qemu_coroutine_self();
        qemu_thread_create(&mis->colo_incoming_thread, "COLO incoming",
             colo_process_incoming_thread, mis, QEMU_THREAD_JOINABLE);
        mis->have_colo_incoming_thread = true;
        qemu_coroutine_yield();

        /* Wait checkpoint incoming thread exit before free resource */
        qemu_thread_join(&mis->colo_incoming_thread);
    }

353
    if (ret < 0) {
354 355
        migrate_set_state(&mis->state, MIGRATION_STATUS_ACTIVE,
                          MIGRATION_STATUS_FAILED);
356
        error_report("load of migration failed: %s", strerror(-ret));
357
        qemu_fclose(mis->from_src_file);
358
        exit(EXIT_FAILURE);
359
    }
360 361
    mis->bh = qemu_bh_new(process_incoming_migration_bh, mis);
    qemu_bh_schedule(mis->bh);
362 363
}

364
void migration_fd_process_incoming(QEMUFile *f)
365
{
366 367
    Coroutine *co = qemu_coroutine_create(process_incoming_migration_co, NULL);
    MigrationIncomingState *mis = migration_incoming_get_current();
368

369 370 371
    if (!mis->from_src_file) {
        mis->from_src_file = f;
    }
372
    qemu_file_set_blocking(f, false);
373
    qemu_coroutine_enter(co);
374 375
}

376 377 378 379 380 381 382 383 384 385 386
void migration_ioc_process_incoming(QIOChannel *ioc)
{
    MigrationIncomingState *mis = migration_incoming_get_current();

    if (!mis->from_src_file) {
        QEMUFile *f = qemu_fopen_channel_input(ioc);
        migration_fd_process_incoming(f);
    }
    /* We still only have a single channel.  Nothing to do here yet */
}

387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413
/*
 * Send a 'SHUT' message on the return channel with the given value
 * to indicate that we've finished with the RP.  Non-0 value indicates
 * error.
 */
void migrate_send_rp_shut(MigrationIncomingState *mis,
                          uint32_t value)
{
    uint32_t buf;

    buf = cpu_to_be32(value);
    migrate_send_rp_message(mis, MIG_RP_MSG_SHUT, sizeof(buf), &buf);
}

/*
 * Send a 'PONG' message on the return channel with the given value
 * (normally in response to a 'PING')
 */
void migrate_send_rp_pong(MigrationIncomingState *mis,
                          uint32_t value)
{
    uint32_t buf;

    buf = cpu_to_be32(value);
    migrate_send_rp_message(mis, MIG_RP_MSG_PONG, sizeof(buf), &buf);
}

O
Orit Wasserman 已提交
414 415 416 417 418 419 420
MigrationCapabilityStatusList *qmp_query_migrate_capabilities(Error **errp)
{
    MigrationCapabilityStatusList *head = NULL;
    MigrationCapabilityStatusList *caps;
    MigrationState *s = migrate_get_current();
    int i;

421
    caps = NULL; /* silence compiler warning */
422
    for (i = 0; i < MIGRATION_CAPABILITY__MAX; i++) {
423 424 425 426 427
#ifndef CONFIG_LIVE_BLOCK_MIGRATION
        if (i == MIGRATION_CAPABILITY_BLOCK) {
            continue;
        }
#endif
O
Orit Wasserman 已提交
428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443
        if (head == NULL) {
            head = g_malloc0(sizeof(*caps));
            caps = head;
        } else {
            caps->next = g_malloc0(sizeof(*caps));
            caps = caps->next;
        }
        caps->value =
            g_malloc(sizeof(*caps->value));
        caps->value->capability = i;
        caps->value->state = s->enabled_capabilities[i];
    }

    return head;
}

444 445 446 447 448
MigrationParameters *qmp_query_migrate_parameters(Error **errp)
{
    MigrationParameters *params;
    MigrationState *s = migrate_get_current();

449
    /* TODO use QAPI_CLONE() instead of duplicating it inline */
450
    params = g_malloc0(sizeof(*params));
451
    params->has_compress_level = true;
452
    params->compress_level = s->parameters.compress_level;
453
    params->has_compress_threads = true;
454
    params->compress_threads = s->parameters.compress_threads;
455
    params->has_decompress_threads = true;
456
    params->decompress_threads = s->parameters.decompress_threads;
457
    params->has_cpu_throttle_initial = true;
458
    params->cpu_throttle_initial = s->parameters.cpu_throttle_initial;
459
    params->has_cpu_throttle_increment = true;
460
    params->cpu_throttle_increment = s->parameters.cpu_throttle_increment;
461
    params->has_tls_creds = true;
462
    params->tls_creds = g_strdup(s->parameters.tls_creds);
463
    params->has_tls_hostname = true;
464
    params->tls_hostname = g_strdup(s->parameters.tls_hostname);
465 466 467 468
    params->has_max_bandwidth = true;
    params->max_bandwidth = s->parameters.max_bandwidth;
    params->has_downtime_limit = true;
    params->downtime_limit = s->parameters.downtime_limit;
469
    params->has_x_checkpoint_delay = true;
470
    params->x_checkpoint_delay = s->parameters.x_checkpoint_delay;
471 472
    params->has_block_incremental = true;
    params->block_incremental = s->parameters.block_incremental;
473 474 475 476

    return params;
}

477 478 479 480 481 482 483 484
/*
 * Return true if we're already in the middle of a migration
 * (i.e. any of the active or setup states)
 */
static bool migration_is_setup_or_active(int state)
{
    switch (state) {
    case MIGRATION_STATUS_ACTIVE:
485
    case MIGRATION_STATUS_POSTCOPY_ACTIVE:
486 487 488 489 490 491 492 493 494
    case MIGRATION_STATUS_SETUP:
        return true;

    default:
        return false;

    }
}

495 496 497 498
static void populate_ram_info(MigrationInfo *info, MigrationState *s)
{
    info->has_ram = true;
    info->ram = g_malloc0(sizeof(*info->ram));
499
    info->ram->transferred = ram_counters.transferred;
500
    info->ram->total = ram_bytes_total();
501
    info->ram->duplicate = ram_counters.duplicate;
502 503
    /* legacy value.  It is not used anymore */
    info->ram->skipped = 0;
504 505
    info->ram->normal = ram_counters.normal;
    info->ram->normal_bytes = ram_counters.normal *
506
        qemu_target_page_size();
507
    info->ram->mbps = s->mbps;
508 509
    info->ram->dirty_sync_count = ram_counters.dirty_sync_count;
    info->ram->postcopy_requests = ram_counters.postcopy_requests;
510
    info->ram->page_size = qemu_target_page_size();
511

512 513 514 515
    if (migrate_use_xbzrle()) {
        info->has_xbzrle_cache = true;
        info->xbzrle_cache = g_malloc0(sizeof(*info->xbzrle_cache));
        info->xbzrle_cache->cache_size = migrate_xbzrle_cache_size();
516 517 518 519 520
        info->xbzrle_cache->bytes = xbzrle_counters.bytes;
        info->xbzrle_cache->pages = xbzrle_counters.pages;
        info->xbzrle_cache->cache_miss = xbzrle_counters.cache_miss;
        info->xbzrle_cache->cache_miss_rate = xbzrle_counters.cache_miss_rate;
        info->xbzrle_cache->overflow = xbzrle_counters.overflow;
521 522
    }

523 524 525 526 527
    if (cpu_throttle_active()) {
        info->has_cpu_throttle_percentage = true;
        info->cpu_throttle_percentage = cpu_throttle_get_percentage();
    }

528 529
    if (s->state != MIGRATION_STATUS_COMPLETED) {
        info->ram->remaining = ram_bytes_remaining();
530
        info->ram->dirty_pages_rate = ram_counters.dirty_pages_rate;
531 532 533
    }
}

534 535 536 537 538 539 540 541 542 543 544
static void populate_disk_info(MigrationInfo *info)
{
    if (blk_mig_active()) {
        info->has_disk = true;
        info->disk = g_malloc0(sizeof(*info->disk));
        info->disk->transferred = blk_mig_bytes_transferred();
        info->disk->remaining = blk_mig_bytes_remaining();
        info->disk->total = blk_mig_bytes_total();
    }
}

L
Luiz Capitulino 已提交
545
MigrationInfo *qmp_query_migrate(Error **errp)
A
aliguori 已提交
546
{
L
Luiz Capitulino 已提交
547
    MigrationInfo *info = g_malloc0(sizeof(*info));
548 549 550
    MigrationState *s = migrate_get_current();

    switch (s->state) {
551
    case MIGRATION_STATUS_NONE:
552 553
        /* no migration has happened ever */
        break;
554
    case MIGRATION_STATUS_SETUP:
555
        info->has_status = true;
556
        info->has_total_time = false;
557
        break;
558 559
    case MIGRATION_STATUS_ACTIVE:
    case MIGRATION_STATUS_CANCELLING:
560
    case MIGRATION_STATUS_POSTCOPY_ACTIVE:
561
         /* TODO add some postcopy stats */
562 563 564 565 566 567 568 569 570
        info->has_status = true;
        info->has_total_time = true;
        info->total_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME)
            - s->total_time;
        info->has_expected_downtime = true;
        info->expected_downtime = s->expected_downtime;
        info->has_setup_time = true;
        info->setup_time = s->setup_time;

571
        populate_ram_info(info, s);
572
        populate_disk_info(info);
573
        break;
574 575 576 577
    case MIGRATION_STATUS_COLO:
        info->has_status = true;
        /* TODO: display COLO specific information (checkpoint info etc.) */
        break;
578
    case MIGRATION_STATUS_COMPLETED:
L
Luiz Capitulino 已提交
579
        info->has_status = true;
580
        info->has_total_time = true;
581
        info->total_time = s->total_time;
582 583
        info->has_downtime = true;
        info->downtime = s->downtime;
584 585
        info->has_setup_time = true;
        info->setup_time = s->setup_time;
J
Juan Quintela 已提交
586

587
        populate_ram_info(info, s);
588
        break;
589
    case MIGRATION_STATUS_FAILED:
L
Luiz Capitulino 已提交
590
        info->has_status = true;
591 592 593 594
        if (s->error) {
            info->has_error_desc = true;
            info->error_desc = g_strdup(error_get_pretty(s->error));
        }
595
        break;
596
    case MIGRATION_STATUS_CANCELLED:
L
Luiz Capitulino 已提交
597
        info->has_status = true;
598
        break;
A
aliguori 已提交
599
    }
600
    info->status = s->state;
L
Luiz Capitulino 已提交
601 602

    return info;
A
aliguori 已提交
603 604
}

605 606 607 608 609 610 611 612 613 614 615 616
/**
 * @migration_caps_check - check capability validity
 *
 * @cap_list: old capability list, array of bool
 * @params: new capabilities to be applied soon
 * @errp: set *errp if the check failed, with reason
 *
 * Returns true if check passed, otherwise false.
 */
static bool migrate_caps_check(bool *cap_list,
                               MigrationCapabilityStatusList *params,
                               Error **errp)
O
Orit Wasserman 已提交
617 618
{
    MigrationCapabilityStatusList *cap;
619
    bool old_postcopy_cap;
O
Orit Wasserman 已提交
620

621
    old_postcopy_cap = cap_list[MIGRATION_CAPABILITY_POSTCOPY_RAM];
O
Orit Wasserman 已提交
622 623

    for (cap = params; cap; cap = cap->next) {
624 625 626
        cap_list[cap->value->capability] = cap->value->state;
    }

627
#ifndef CONFIG_LIVE_BLOCK_MIGRATION
628 629 630 631 632
    if (cap_list[MIGRATION_CAPABILITY_BLOCK]) {
        error_setg(errp, "QEMU compiled without old-style (blk/-b, inc/-i) "
                   "block migration");
        error_append_hint(errp, "Use drive_mirror+NBD instead.\n");
        return false;
O
Orit Wasserman 已提交
633
    }
634
#endif
635

636 637
    if (cap_list[MIGRATION_CAPABILITY_POSTCOPY_RAM]) {
        if (cap_list[MIGRATION_CAPABILITY_COMPRESS]) {
638 639 640 641 642
            /* The decompression threads asynchronously write into RAM
             * rather than use the atomic copies needed to avoid
             * userfaulting.  It should be possible to fix the decompression
             * threads for compatibility in future.
             */
643 644 645
            error_setg(errp, "Postcopy is not currently compatible "
                       "with compression");
            return false;
646
        }
647

648 649 650 651 652 653 654 655 656
        /* This check is reasonably expensive, so only when it's being
         * set the first time, also it's only the destination that needs
         * special support.
         */
        if (!old_postcopy_cap && runstate_check(RUN_STATE_INMIGRATE) &&
            !postcopy_ram_supported_by_host()) {
            /* postcopy_ram_supported_by_host will have emitted a more
             * detailed message
             */
657 658
            error_setg(errp, "Postcopy is not supported");
            return false;
659
        }
660
    }
661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682

    return true;
}

void qmp_migrate_set_capabilities(MigrationCapabilityStatusList *params,
                                  Error **errp)
{
    MigrationState *s = migrate_get_current();
    MigrationCapabilityStatusList *cap;

    if (migration_is_setup_or_active(s->state)) {
        error_setg(errp, QERR_MIGRATION_ACTIVE);
        return;
    }

    if (!migrate_caps_check(s->enabled_capabilities, params, errp)) {
        return;
    }

    for (cap = params; cap; cap = cap->next) {
        s->enabled_capabilities[cap->value->capability] = cap->value->state;
    }
O
Orit Wasserman 已提交
683 684
}

685 686 687 688 689
/*
 * Check whether the parameters are valid. Error will be put into errp
 * (if provided). Return true if valid, otherwise false.
 */
static bool migrate_params_check(MigrationParameters *params, Error **errp)
690
{
691 692
    if (params->has_compress_level &&
        (params->compress_level < 0 || params->compress_level > 9)) {
693 694
        error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "compress_level",
                   "is invalid, it should be in the range of 0 to 9");
695
        return false;
696
    }
697

698 699
    if (params->has_compress_threads &&
        (params->compress_threads < 1 || params->compress_threads > 255)) {
700 701 702
        error_setg(errp, QERR_INVALID_PARAMETER_VALUE,
                   "compress_threads",
                   "is invalid, it should be in the range of 1 to 255");
703
        return false;
704
    }
705

706 707
    if (params->has_decompress_threads &&
        (params->decompress_threads < 1 || params->decompress_threads > 255)) {
708 709 710
        error_setg(errp, QERR_INVALID_PARAMETER_VALUE,
                   "decompress_threads",
                   "is invalid, it should be in the range of 1 to 255");
711
        return false;
712
    }
713

714 715 716
    if (params->has_cpu_throttle_initial &&
        (params->cpu_throttle_initial < 1 ||
         params->cpu_throttle_initial > 99)) {
717
        error_setg(errp, QERR_INVALID_PARAMETER_VALUE,
718
                   "cpu_throttle_initial",
719
                   "an integer in the range of 1 to 99");
720
        return false;
721
    }
722

723 724 725
    if (params->has_cpu_throttle_increment &&
        (params->cpu_throttle_increment < 1 ||
         params->cpu_throttle_increment > 99)) {
726
        error_setg(errp, QERR_INVALID_PARAMETER_VALUE,
727
                   "cpu_throttle_increment",
728
                   "an integer in the range of 1 to 99");
729
        return false;
730
    }
731

732 733 734 735
    if (params->has_max_bandwidth &&
        (params->max_bandwidth < 0 || params->max_bandwidth > SIZE_MAX)) {
        error_setg(errp, "Parameter 'max_bandwidth' expects an integer in the"
                         " range of 0 to %zu bytes/second", SIZE_MAX);
736
        return false;
737
    }
738

739
    if (params->has_downtime_limit &&
740 741 742 743 744
        (params->downtime_limit < 0 ||
         params->downtime_limit > MAX_MIGRATE_DOWNTIME)) {
        error_setg(errp, "Parameter 'downtime_limit' expects an integer in "
                         "the range of 0 to %d milliseconds",
                         MAX_MIGRATE_DOWNTIME);
745
        return false;
746
    }
747

748 749 750 751
    if (params->has_x_checkpoint_delay && (params->x_checkpoint_delay < 0)) {
        error_setg(errp, QERR_INVALID_PARAMETER_VALUE,
                    "x_checkpoint_delay",
                    "is invalid, it should be positive");
752 753 754 755 756 757
        return false;
    }

    return true;
}

758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785
static void migrate_params_test_apply(MigrateSetParameters *params,
                                      MigrationParameters *dest)
{
    *dest = migrate_get_current()->parameters;

    /* TODO use QAPI_CLONE() instead of duplicating it inline */

    if (params->has_compress_level) {
        dest->compress_level = params->compress_level;
    }

    if (params->has_compress_threads) {
        dest->compress_threads = params->compress_threads;
    }

    if (params->has_decompress_threads) {
        dest->decompress_threads = params->decompress_threads;
    }

    if (params->has_cpu_throttle_initial) {
        dest->cpu_throttle_initial = params->cpu_throttle_initial;
    }

    if (params->has_cpu_throttle_increment) {
        dest->cpu_throttle_increment = params->cpu_throttle_increment;
    }

    if (params->has_tls_creds) {
786 787
        assert(params->tls_creds->type == QTYPE_QSTRING);
        dest->tls_creds = g_strdup(params->tls_creds->u.s);
788 789 790
    }

    if (params->has_tls_hostname) {
791 792
        assert(params->tls_hostname->type == QTYPE_QSTRING);
        dest->tls_hostname = g_strdup(params->tls_hostname->u.s);
793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812
    }

    if (params->has_max_bandwidth) {
        dest->max_bandwidth = params->max_bandwidth;
    }

    if (params->has_downtime_limit) {
        dest->downtime_limit = params->downtime_limit;
    }

    if (params->has_x_checkpoint_delay) {
        dest->x_checkpoint_delay = params->x_checkpoint_delay;
    }

    if (params->has_block_incremental) {
        dest->block_incremental = params->block_incremental;
    }
}

static void migrate_params_apply(MigrateSetParameters *params)
813 814 815
{
    MigrationState *s = migrate_get_current();

816 817
    /* TODO use QAPI_CLONE() instead of duplicating it inline */

818 819
    if (params->has_compress_level) {
        s->parameters.compress_level = params->compress_level;
820
    }
821

822 823
    if (params->has_compress_threads) {
        s->parameters.compress_threads = params->compress_threads;
824
    }
825

826 827
    if (params->has_decompress_threads) {
        s->parameters.decompress_threads = params->decompress_threads;
828
    }
829

830 831
    if (params->has_cpu_throttle_initial) {
        s->parameters.cpu_throttle_initial = params->cpu_throttle_initial;
832
    }
833

834 835
    if (params->has_cpu_throttle_increment) {
        s->parameters.cpu_throttle_increment = params->cpu_throttle_increment;
836
    }
837

838
    if (params->has_tls_creds) {
839
        g_free(s->parameters.tls_creds);
840 841
        assert(params->tls_creds->type == QTYPE_QSTRING);
        s->parameters.tls_creds = g_strdup(params->tls_creds->u.s);
842
    }
843

844
    if (params->has_tls_hostname) {
845
        g_free(s->parameters.tls_hostname);
846 847
        assert(params->tls_hostname->type == QTYPE_QSTRING);
        s->parameters.tls_hostname = g_strdup(params->tls_hostname->u.s);
848
    }
849

850 851 852 853 854 855 856
    if (params->has_max_bandwidth) {
        s->parameters.max_bandwidth = params->max_bandwidth;
        if (s->to_dst_file) {
            qemu_file_set_rate_limit(s->to_dst_file,
                                s->parameters.max_bandwidth / XFER_LIMIT_RATIO);
        }
    }
857

858 859 860
    if (params->has_downtime_limit) {
        s->parameters.downtime_limit = params->downtime_limit;
    }
861 862 863

    if (params->has_x_checkpoint_delay) {
        s->parameters.x_checkpoint_delay = params->x_checkpoint_delay;
864 865 866
        if (migration_in_colo_state()) {
            colo_checkpoint_notify(s);
        }
867
    }
868

869 870 871
    if (params->has_block_incremental) {
        s->parameters.block_incremental = params->block_incremental;
    }
872 873
}

874
void qmp_migrate_set_parameters(MigrateSetParameters *params, Error **errp)
875
{
876 877
    MigrationParameters tmp;

878 879 880 881 882 883 884 885 886 887 888 889 890 891 892
    /* TODO Rewrite "" to null instead */
    if (params->has_tls_creds
        && params->tls_creds->type == QTYPE_QNULL) {
        QDECREF(params->tls_creds->u.n);
        params->tls_creds->type = QTYPE_QSTRING;
        params->tls_creds->u.s = strdup("");
    }
    /* TODO Rewrite "" to null instead */
    if (params->has_tls_hostname
        && params->tls_hostname->type == QTYPE_QNULL) {
        QDECREF(params->tls_hostname->u.n);
        params->tls_hostname->type = QTYPE_QSTRING;
        params->tls_hostname->u.s = strdup("");
    }

893 894 895
    migrate_params_test_apply(params, &tmp);

    if (!migrate_params_check(&tmp, errp)) {
896 897 898 899 900 901 902
        /* Invalid parameter */
        return;
    }

    migrate_params_apply(params);
}

903

904 905 906 907 908
void qmp_migrate_start_postcopy(Error **errp)
{
    MigrationState *s = migrate_get_current();

    if (!migrate_postcopy_ram()) {
909
        error_setg(errp, "Enable postcopy with migrate_set_capability before"
910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925
                         " the start of migration");
        return;
    }

    if (s->state == MIGRATION_STATUS_NONE) {
        error_setg(errp, "Postcopy must be started after migration has been"
                         " started");
        return;
    }
    /*
     * we don't error if migration has finished since that would be racy
     * with issuing this command.
     */
    atomic_set(&s->start_postcopy, true);
}

926 927
/* shared migration helpers */

928
void migrate_set_state(int *state, int old_state, int new_state)
929
{
930
    assert(new_state < MIGRATION_STATUS__MAX);
931
    if (atomic_cmpxchg(state, old_state, new_state) == old_state) {
932
        trace_migrate_set_state(MigrationStatus_str(new_state));
933
        migrate_generate_event(new_state);
934 935 936
    }
}

P
Peter Xu 已提交
937 938 939 940
static MigrationCapabilityStatusList *migrate_cap_add(
    MigrationCapabilityStatusList *list,
    MigrationCapability index,
    bool state)
941 942 943 944 945
{
    MigrationCapabilityStatusList *cap;

    cap = g_new0(MigrationCapabilityStatusList, 1);
    cap->value = g_new0(MigrationCapabilityStatus, 1);
P
Peter Xu 已提交
946 947 948 949 950 951 952 953 954 955 956 957
    cap->value->capability = index;
    cap->value->state = state;
    cap->next = list;

    return cap;
}

void migrate_set_block_enabled(bool value, Error **errp)
{
    MigrationCapabilityStatusList *cap;

    cap = migrate_cap_add(NULL, MIGRATION_CAPABILITY_BLOCK, value);
958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976
    qmp_migrate_set_capabilities(cap, errp);
    qapi_free_MigrationCapabilityStatusList(cap);
}

static void migrate_set_block_incremental(MigrationState *s, bool value)
{
    s->parameters.block_incremental = value;
}

static void block_cleanup_parameters(MigrationState *s)
{
    if (s->must_remove_block_options) {
        /* setting to false can never fail */
        migrate_set_block_enabled(false, &error_abort);
        migrate_set_block_incremental(s, false);
        s->must_remove_block_options = false;
    }
}

977
static void migrate_fd_cleanup(void *opaque)
978
{
979 980 981 982 983
    MigrationState *s = opaque;

    qemu_bh_delete(s->cleanup_bh);
    s->cleanup_bh = NULL;

984
    if (s->to_dst_file) {
985
        trace_migrate_fd_cleanup();
986
        qemu_mutex_unlock_iothread();
987 988 989 990
        if (s->migration_thread_running) {
            qemu_thread_join(&s->thread);
            s->migration_thread_running = false;
        }
991 992
        qemu_mutex_lock_iothread();

993 994
        qemu_fclose(s->to_dst_file);
        s->to_dst_file = NULL;
995 996
    }

997 998
    assert((s->state != MIGRATION_STATUS_ACTIVE) &&
           (s->state != MIGRATION_STATUS_POSTCOPY_ACTIVE));
999

1000
    if (s->state == MIGRATION_STATUS_CANCELLING) {
1001
        migrate_set_state(&s->state, MIGRATION_STATUS_CANCELLING,
1002
                          MIGRATION_STATUS_CANCELLED);
1003
    }
1004 1005

    notifier_list_notify(&migration_state_notifiers, s);
1006
    block_cleanup_parameters(s);
1007 1008
}

1009
void migrate_fd_error(MigrationState *s, const Error *error)
1010
{
1011
    trace_migrate_fd_error(error_get_pretty(error));
1012
    assert(s->to_dst_file == NULL);
1013 1014
    migrate_set_state(&s->state, MIGRATION_STATUS_SETUP,
                      MIGRATION_STATUS_FAILED);
1015 1016 1017
    if (!s->error) {
        s->error = error_copy(error);
    }
1018
    notifier_list_notify(&migration_state_notifiers, s);
1019
    block_cleanup_parameters(s);
1020 1021
}

1022
static void migrate_fd_cancel(MigrationState *s)
1023
{
1024
    int old_state ;
1025
    QEMUFile *f = migrate_get_current()->to_dst_file;
1026
    trace_migrate_fd_cancel();
1027

1028 1029 1030 1031 1032
    if (s->rp_state.from_dst_file) {
        /* shutdown the rp socket, so causing the rp thread to shutdown */
        qemu_file_shutdown(s->rp_state.from_dst_file);
    }

1033 1034
    do {
        old_state = s->state;
1035
        if (!migration_is_setup_or_active(old_state)) {
1036 1037
            break;
        }
1038
        migrate_set_state(&s->state, old_state, MIGRATION_STATUS_CANCELLING);
1039
    } while (s->state != MIGRATION_STATUS_CANCELLING);
1040 1041 1042 1043 1044 1045 1046 1047

    /*
     * If we're unlucky the migration code might be stuck somewhere in a
     * send/write while the network has failed and is waiting to timeout;
     * if we've got shutdown(2) available then we can force it to quit.
     * The outgoing qemu file gets closed in migrate_fd_cleanup that is
     * called in a bh, so there is no race against this cancel.
     */
1048
    if (s->state == MIGRATION_STATUS_CANCELLING && f) {
1049 1050
        qemu_file_shutdown(f);
    }
1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
    if (s->state == MIGRATION_STATUS_CANCELLING && s->block_inactive) {
        Error *local_err = NULL;

        bdrv_invalidate_cache_all(&local_err);
        if (local_err) {
            error_report_err(local_err);
        } else {
            s->block_inactive = false;
        }
    }
1061
    block_cleanup_parameters(s);
1062 1063
}

1064 1065 1066 1067 1068 1069 1070
void add_migration_state_change_notifier(Notifier *notify)
{
    notifier_list_add(&migration_state_notifiers, notify);
}

void remove_migration_state_change_notifier(Notifier *notify)
{
P
Paolo Bonzini 已提交
1071
    notifier_remove(notify);
1072 1073
}

S
Stefan Hajnoczi 已提交
1074
bool migration_in_setup(MigrationState *s)
1075
{
1076
    return s->state == MIGRATION_STATUS_SETUP;
1077 1078
}

1079
bool migration_has_finished(MigrationState *s)
1080
{
1081
    return s->state == MIGRATION_STATUS_COMPLETED;
1082
}
1083

1084 1085
bool migration_has_failed(MigrationState *s)
{
1086 1087
    return (s->state == MIGRATION_STATUS_CANCELLED ||
            s->state == MIGRATION_STATUS_FAILED);
1088 1089
}

1090
bool migration_in_postcopy(void)
1091
{
1092 1093
    MigrationState *s = migrate_get_current();

1094 1095 1096
    return (s->state == MIGRATION_STATUS_POSTCOPY_ACTIVE);
}

1097 1098
bool migration_in_postcopy_after_devices(MigrationState *s)
{
1099
    return migration_in_postcopy() && s->postcopy_after_devices;
1100 1101
}

1102
bool migration_is_idle(void)
1103
{
1104
    MigrationState *s = migrate_get_current();
1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124

    switch (s->state) {
    case MIGRATION_STATUS_NONE:
    case MIGRATION_STATUS_CANCELLED:
    case MIGRATION_STATUS_COMPLETED:
    case MIGRATION_STATUS_FAILED:
        return true;
    case MIGRATION_STATUS_SETUP:
    case MIGRATION_STATUS_CANCELLING:
    case MIGRATION_STATUS_ACTIVE:
    case MIGRATION_STATUS_POSTCOPY_ACTIVE:
    case MIGRATION_STATUS_COLO:
        return false;
    case MIGRATION_STATUS__MAX:
        g_assert_not_reached();
    }

    return false;
}

1125
MigrationState *migrate_init(void)
1126
{
1127
    MigrationState *s = migrate_get_current();
O
Orit Wasserman 已提交
1128

1129 1130 1131 1132 1133 1134 1135 1136
    /*
     * Reinitialise all migration state, except
     * parameters/capabilities that the user set, and
     * locks.
     */
    s->bytes_xfer = 0;
    s->xfer_limit = 0;
    s->cleanup_bh = 0;
1137
    s->to_dst_file = NULL;
1138 1139 1140 1141 1142 1143 1144 1145
    s->state = MIGRATION_STATUS_NONE;
    s->rp_state.from_dst_file = NULL;
    s->rp_state.error = false;
    s->mbps = 0.0;
    s->downtime = 0;
    s->expected_downtime = 0;
    s->setup_time = 0;
    s->start_postcopy = false;
1146
    s->postcopy_after_devices = false;
1147
    s->migration_thread_running = false;
1148 1149
    error_free(s->error);
    s->error = NULL;
1150

1151
    migrate_set_state(&s->state, MIGRATION_STATUS_NONE, MIGRATION_STATUS_SETUP);
1152

1153
    s->total_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
1154 1155
    return s;
}
1156

A
Anthony Liguori 已提交
1157 1158
static GSList *migration_blockers;

1159
int migrate_add_blocker(Error *reason, Error **errp)
A
Anthony Liguori 已提交
1160
{
1161
    if (migrate_get_current()->only_migratable) {
1162 1163 1164 1165 1166 1167
        error_propagate(errp, error_copy(reason));
        error_prepend(errp, "disallowing migration blocker "
                          "(--only_migratable) for: ");
        return -EACCES;
    }

1168
    if (migration_is_idle()) {
1169 1170 1171 1172 1173 1174 1175 1176
        migration_blockers = g_slist_prepend(migration_blockers, reason);
        return 0;
    }

    error_propagate(errp, error_copy(reason));
    error_prepend(errp, "disallowing migration blocker (migration in "
                      "progress) for: ");
    return -EBUSY;
A
Anthony Liguori 已提交
1177 1178 1179 1180 1181 1182 1183
}

void migrate_del_blocker(Error *reason)
{
    migration_blockers = g_slist_remove(migration_blockers, reason);
}

D
Dr. David Alan Gilbert 已提交
1184 1185 1186
void qmp_migrate_incoming(const char *uri, Error **errp)
{
    Error *local_err = NULL;
1187
    static bool once = true;
D
Dr. David Alan Gilbert 已提交
1188 1189

    if (!deferred_incoming) {
1190
        error_setg(errp, "For use with '-incoming defer'");
D
Dr. David Alan Gilbert 已提交
1191 1192
        return;
    }
1193 1194 1195
    if (!once) {
        error_setg(errp, "The incoming migration has already been started");
    }
D
Dr. David Alan Gilbert 已提交
1196 1197 1198 1199 1200 1201 1202 1203

    qemu_start_incoming_migration(uri, &local_err);

    if (local_err) {
        error_propagate(errp, local_err);
        return;
    }

1204
    once = false;
D
Dr. David Alan Gilbert 已提交
1205 1206
}

1207 1208 1209 1210 1211 1212 1213
bool migration_is_blocked(Error **errp)
{
    if (qemu_savevm_state_blocked(errp)) {
        return true;
    }

    if (migration_blockers) {
1214
        error_propagate(errp, error_copy(migration_blockers->data));
1215 1216 1217 1218 1219 1220
        return true;
    }

    return false;
}

L
Luiz Capitulino 已提交
1221 1222 1223
void qmp_migrate(const char *uri, bool has_blk, bool blk,
                 bool has_inc, bool inc, bool has_detach, bool detach,
                 Error **errp)
1224
{
1225
    Error *local_err = NULL;
1226
    MigrationState *s = migrate_get_current();
1227 1228
    const char *p;

1229
    if (migration_is_setup_or_active(s->state) ||
1230 1231
        s->state == MIGRATION_STATUS_CANCELLING ||
        s->state == MIGRATION_STATUS_COLO) {
1232
        error_setg(errp, QERR_MIGRATION_ACTIVE);
L
Luiz Capitulino 已提交
1233
        return;
1234
    }
1235 1236 1237 1238 1239
    if (runstate_check(RUN_STATE_INMIGRATE)) {
        error_setg(errp, "Guest is waiting for an incoming migration");
        return;
    }

1240
    if (migration_is_blocked(errp)) {
L
Luiz Capitulino 已提交
1241
        return;
A
Anthony Liguori 已提交
1242 1243
    }

1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261
    if ((has_blk && blk) || (has_inc && inc)) {
        if (migrate_use_block() || migrate_use_block_incremental()) {
            error_setg(errp, "Command options are incompatible with "
                       "current migration capabilities");
            return;
        }
        migrate_set_block_enabled(true, &local_err);
        if (local_err) {
            error_propagate(errp, local_err);
            return;
        }
        s->must_remove_block_options = true;
    }

    if (has_inc && inc) {
        migrate_set_block_incremental(s, true);
    }

1262
    s = migrate_init();
1263 1264

    if (strstart(uri, "tcp:", &p)) {
1265
        tcp_start_outgoing_migration(s, p, &local_err);
M
Michael R. Hines 已提交
1266
#ifdef CONFIG_RDMA
1267
    } else if (strstart(uri, "rdma:", &p)) {
M
Michael R. Hines 已提交
1268 1269
        rdma_start_outgoing_migration(s, p, &local_err);
#endif
1270
    } else if (strstart(uri, "exec:", &p)) {
1271
        exec_start_outgoing_migration(s, p, &local_err);
1272
    } else if (strstart(uri, "unix:", &p)) {
1273
        unix_start_outgoing_migration(s, p, &local_err);
1274
    } else if (strstart(uri, "fd:", &p)) {
1275
        fd_start_outgoing_migration(s, p, &local_err);
1276
    } else {
1277 1278
        error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "uri",
                   "a valid migration protocol");
1279 1280
        migrate_set_state(&s->state, MIGRATION_STATUS_SETUP,
                          MIGRATION_STATUS_FAILED);
L
Luiz Capitulino 已提交
1281
        return;
1282 1283
    }

1284
    if (local_err) {
1285
        migrate_fd_error(s, local_err);
1286
        error_propagate(errp, local_err);
L
Luiz Capitulino 已提交
1287
        return;
1288
    }
1289 1290
}

L
Luiz Capitulino 已提交
1291
void qmp_migrate_cancel(Error **errp)
1292
{
1293
    migrate_fd_cancel(migrate_get_current());
1294 1295
}

1296 1297 1298
void qmp_migrate_set_cache_size(int64_t value, Error **errp)
{
    MigrationState *s = migrate_get_current();
1299
    int64_t new_size;
1300 1301 1302

    /* Check for truncation */
    if (value != (size_t)value) {
1303 1304
        error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "cache size",
                   "exceeding address space");
1305 1306 1307
        return;
    }

1308 1309
    /* Cache should not be larger than guest ram size */
    if (value > ram_bytes_total()) {
1310 1311
        error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "cache size",
                   "exceeds guest ram size ");
1312 1313 1314
        return;
    }

1315 1316
    new_size = xbzrle_cache_resize(value);
    if (new_size < 0) {
1317 1318
        error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "cache size",
                   "is smaller than page size");
1319 1320 1321 1322
        return;
    }

    s->xbzrle_cache_size = new_size;
1323 1324 1325 1326 1327 1328 1329
}

int64_t qmp_query_migrate_cache_size(Error **errp)
{
    return migrate_xbzrle_cache_size();
}

L
Luiz Capitulino 已提交
1330
void qmp_migrate_set_speed(int64_t value, Error **errp)
1331
{
1332
    MigrateSetParameters p = {
1333 1334 1335
        .has_max_bandwidth = true,
        .max_bandwidth = value,
    };
1336

1337
    qmp_migrate_set_parameters(&p, errp);
1338 1339
}

1340
void qmp_migrate_set_downtime(double value, Error **errp)
1341
{
1342 1343 1344 1345 1346 1347 1348
    if (value < 0 || value > MAX_MIGRATE_DOWNTIME_SECONDS) {
        error_setg(errp, "Parameter 'downtime_limit' expects an integer in "
                         "the range of 0 to %d seconds",
                         MAX_MIGRATE_DOWNTIME_SECONDS);
        return;
    }

1349 1350 1351
    value *= 1000; /* Convert to milliseconds */
    value = MAX(0, MIN(INT64_MAX, value));

1352
    MigrateSetParameters p = {
1353 1354 1355 1356 1357
        .has_downtime_limit = true,
        .downtime_limit = value,
    };

    qmp_migrate_set_parameters(&p, errp);
1358
}
1359

1360 1361 1362 1363 1364 1365 1366 1367 1368
bool migrate_release_ram(void)
{
    MigrationState *s;

    s = migrate_get_current();

    return s->enabled_capabilities[MIGRATION_CAPABILITY_RELEASE_RAM];
}

1369 1370 1371 1372 1373 1374
bool migrate_postcopy_ram(void)
{
    MigrationState *s;

    s = migrate_get_current();

D
Dr. David Alan Gilbert 已提交
1375
    return s->enabled_capabilities[MIGRATION_CAPABILITY_POSTCOPY_RAM];
1376 1377
}

1378 1379 1380 1381 1382 1383 1384 1385 1386
bool migrate_auto_converge(void)
{
    MigrationState *s;

    s = migrate_get_current();

    return s->enabled_capabilities[MIGRATION_CAPABILITY_AUTO_CONVERGE];
}

1387 1388 1389 1390 1391 1392 1393 1394 1395
bool migrate_zero_blocks(void)
{
    MigrationState *s;

    s = migrate_get_current();

    return s->enabled_capabilities[MIGRATION_CAPABILITY_ZERO_BLOCKS];
}

1396 1397
bool migrate_use_compression(void)
{
1398 1399 1400 1401 1402
    MigrationState *s;

    s = migrate_get_current();

    return s->enabled_capabilities[MIGRATION_CAPABILITY_COMPRESS];
1403 1404 1405 1406 1407 1408 1409 1410
}

int migrate_compress_level(void)
{
    MigrationState *s;

    s = migrate_get_current();

1411
    return s->parameters.compress_level;
1412 1413 1414 1415 1416 1417 1418 1419
}

int migrate_compress_threads(void)
{
    MigrationState *s;

    s = migrate_get_current();

1420
    return s->parameters.compress_threads;
1421 1422
}

1423 1424 1425 1426 1427 1428
int migrate_decompress_threads(void)
{
    MigrationState *s;

    s = migrate_get_current();

1429
    return s->parameters.decompress_threads;
1430 1431
}

1432 1433 1434 1435 1436 1437 1438 1439 1440
bool migrate_use_events(void)
{
    MigrationState *s;

    s = migrate_get_current();

    return s->enabled_capabilities[MIGRATION_CAPABILITY_EVENTS];
}

1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457
int migrate_use_xbzrle(void)
{
    MigrationState *s;

    s = migrate_get_current();

    return s->enabled_capabilities[MIGRATION_CAPABILITY_XBZRLE];
}

int64_t migrate_xbzrle_cache_size(void)
{
    MigrationState *s;

    s = migrate_get_current();

    return s->xbzrle_cache_size;
}
1458

1459 1460 1461 1462 1463 1464 1465 1466 1467
bool migrate_use_block(void)
{
    MigrationState *s;

    s = migrate_get_current();

    return s->enabled_capabilities[MIGRATION_CAPABILITY_BLOCK];
}

1468 1469 1470 1471 1472 1473 1474 1475 1476
bool migrate_use_return_path(void)
{
    MigrationState *s;

    s = migrate_get_current();

    return s->enabled_capabilities[MIGRATION_CAPABILITY_RETURN_PATH];
}

1477 1478 1479 1480 1481 1482 1483 1484 1485
bool migrate_use_block_incremental(void)
{
    MigrationState *s;

    s = migrate_get_current();

    return s->parameters.block_incremental;
}

1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502
/* migration thread support */
/*
 * Something bad happened to the RP stream, mark an error
 * The caller shall print or trace something to indicate why
 */
static void mark_source_rp_bad(MigrationState *s)
{
    s->rp_state.error = true;
}

static struct rp_cmd_args {
    ssize_t     len; /* -1 = variable */
    const char *name;
} rp_cmd_args[] = {
    [MIG_RP_MSG_INVALID]        = { .len = -1, .name = "INVALID" },
    [MIG_RP_MSG_SHUT]           = { .len =  4, .name = "SHUT" },
    [MIG_RP_MSG_PONG]           = { .len =  4, .name = "PONG" },
1503 1504
    [MIG_RP_MSG_REQ_PAGES]      = { .len = 12, .name = "REQ_PAGES" },
    [MIG_RP_MSG_REQ_PAGES_ID]   = { .len = -1, .name = "REQ_PAGES_ID" },
1505 1506 1507
    [MIG_RP_MSG_MAX]            = { .len = -1, .name = "MAX" },
};

1508 1509 1510 1511 1512 1513 1514 1515
/*
 * Process a request for pages received on the return path,
 * We're allowed to send more than requested (e.g. to round to our page size)
 * and we don't need to send pages that have already been sent.
 */
static void migrate_handle_rp_req_pages(MigrationState *ms, const char* rbname,
                                       ram_addr_t start, size_t len)
{
1516 1517
    long our_host_ps = getpagesize();

1518
    trace_migrate_handle_rp_req_pages(rbname, start, len);
1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531

    /*
     * Since we currently insist on matching page sizes, just sanity check
     * we're being asked for whole host pages.
     */
    if (start & (our_host_ps-1) ||
       (len & (our_host_ps-1))) {
        error_report("%s: Misaligned page request, start: " RAM_ADDR_FMT
                     " len: %zd", __func__, start, len);
        mark_source_rp_bad(ms);
        return;
    }

1532
    if (ram_save_queue_pages(rbname, start, len)) {
1533 1534
        mark_source_rp_bad(ms);
    }
1535 1536
}

1537 1538 1539 1540 1541 1542 1543 1544 1545
/*
 * Handles messages sent on the return path towards the source VM
 *
 */
static void *source_return_path_thread(void *opaque)
{
    MigrationState *ms = opaque;
    QEMUFile *rp = ms->rp_state.from_dst_file;
    uint16_t header_len, header_type;
1546
    uint8_t buf[512];
1547
    uint32_t tmp32, sibling_error;
1548 1549
    ram_addr_t start = 0; /* =0 to silence warning */
    size_t  len = 0, expected_len;
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568
    int res;

    trace_source_return_path_thread_entry();
    while (!ms->rp_state.error && !qemu_file_get_error(rp) &&
           migration_is_setup_or_active(ms->state)) {
        trace_source_return_path_thread_loop_top();
        header_type = qemu_get_be16(rp);
        header_len = qemu_get_be16(rp);

        if (header_type >= MIG_RP_MSG_MAX ||
            header_type == MIG_RP_MSG_INVALID) {
            error_report("RP: Received invalid message 0x%04x length 0x%04x",
                    header_type, header_len);
            mark_source_rp_bad(ms);
            goto out;
        }

        if ((rp_cmd_args[header_type].len != -1 &&
            header_len != rp_cmd_args[header_type].len) ||
1569
            header_len > sizeof(buf)) {
1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590
            error_report("RP: Received '%s' message (0x%04x) with"
                    "incorrect length %d expecting %zu",
                    rp_cmd_args[header_type].name, header_type, header_len,
                    (size_t)rp_cmd_args[header_type].len);
            mark_source_rp_bad(ms);
            goto out;
        }

        /* We know we've got a valid header by this point */
        res = qemu_get_buffer(rp, buf, header_len);
        if (res != header_len) {
            error_report("RP: Failed reading data for message 0x%04x"
                         " read %d expected %d",
                         header_type, res, header_len);
            mark_source_rp_bad(ms);
            goto out;
        }

        /* OK, we have the message and the data */
        switch (header_type) {
        case MIG_RP_MSG_SHUT:
1591
            sibling_error = ldl_be_p(buf);
1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604
            trace_source_return_path_thread_shut(sibling_error);
            if (sibling_error) {
                error_report("RP: Sibling indicated error %d", sibling_error);
                mark_source_rp_bad(ms);
            }
            /*
             * We'll let the main thread deal with closing the RP
             * we could do a shutdown(2) on it, but we're the only user
             * anyway, so there's nothing gained.
             */
            goto out;

        case MIG_RP_MSG_PONG:
1605
            tmp32 = ldl_be_p(buf);
1606 1607 1608
            trace_source_return_path_thread_pong(tmp32);
            break;

1609
        case MIG_RP_MSG_REQ_PAGES:
1610 1611
            start = ldq_be_p(buf);
            len = ldl_be_p(buf + 8);
1612 1613 1614 1615 1616 1617 1618
            migrate_handle_rp_req_pages(ms, NULL, start, len);
            break;

        case MIG_RP_MSG_REQ_PAGES_ID:
            expected_len = 12 + 1; /* header + termination */

            if (header_len >= expected_len) {
1619 1620
                start = ldq_be_p(buf);
                len = ldl_be_p(buf + 8);
1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
                /* Now we expect an idstr */
                tmp32 = buf[12]; /* Length of the following idstr */
                buf[13 + tmp32] = '\0';
                expected_len += tmp32;
            }
            if (header_len != expected_len) {
                error_report("RP: Req_Page_id with length %d expecting %zd",
                        header_len, expected_len);
                mark_source_rp_bad(ms);
                goto out;
            }
            migrate_handle_rp_req_pages(ms, (char *)&buf[13], start, len);
            break;

1635 1636 1637 1638
        default:
            break;
        }
    }
D
Dr. David Alan Gilbert 已提交
1639
    if (qemu_file_get_error(rp)) {
1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653
        trace_source_return_path_thread_bad_end();
        mark_source_rp_bad(ms);
    }

    trace_source_return_path_thread_end();
out:
    ms->rp_state.from_dst_file = NULL;
    qemu_fclose(rp);
    return NULL;
}

static int open_return_path_on_source(MigrationState *ms)
{

1654
    ms->rp_state.from_dst_file = qemu_file_get_return_path(ms->to_dst_file);
1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675
    if (!ms->rp_state.from_dst_file) {
        return -1;
    }

    trace_open_return_path_on_source();
    qemu_thread_create(&ms->rp_state.rp_thread, "return path",
                       source_return_path_thread, ms, QEMU_THREAD_JOINABLE);

    trace_open_return_path_on_source_continue();

    return 0;
}

/* Returns 0 if the RP was ok, otherwise there was an error on the RP */
static int await_return_path_close_on_source(MigrationState *ms)
{
    /*
     * If this is a normal exit then the destination will send a SHUT and the
     * rp_thread will exit, however if there's an error we need to cause
     * it to exit.
     */
1676
    if (qemu_file_get_error(ms->to_dst_file) && ms->rp_state.from_dst_file) {
1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689
        /*
         * shutdown(2), if we have it, will cause it to unblock if it's stuck
         * waiting for the destination.
         */
        qemu_file_shutdown(ms->rp_state.from_dst_file);
        mark_source_rp_bad(ms);
    }
    trace_await_return_path_close_on_source_joining();
    qemu_thread_join(&ms->rp_state.rp_thread);
    trace_await_return_path_close_on_source_close();
    return ms->rp_state.error;
}

1690 1691 1692 1693 1694 1695 1696
/*
 * Switch from normal iteration to postcopy
 * Returns non-0 on error
 */
static int postcopy_start(MigrationState *ms, bool *old_vm_running)
{
    int ret;
1697 1698
    QIOChannelBuffer *bioc;
    QEMUFile *fb;
1699
    int64_t time_at_stop = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
1700
    bool restart_block = false;
1701
    migrate_set_state(&ms->state, MIGRATION_STATUS_ACTIVE,
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711
                      MIGRATION_STATUS_POSTCOPY_ACTIVE);

    trace_postcopy_start();
    qemu_mutex_lock_iothread();
    trace_postcopy_start_set_run();

    qemu_system_wakeup_request(QEMU_WAKEUP_REASON_OTHER);
    *old_vm_running = runstate_is_running();
    global_state_store();
    ret = vm_stop_force_state(RUN_STATE_FINISH_MIGRATE);
1712 1713 1714
    if (ret < 0) {
        goto fail;
    }
1715

1716
    ret = bdrv_inactivate_all();
1717 1718 1719
    if (ret < 0) {
        goto fail;
    }
1720
    restart_block = true;
1721

1722 1723 1724 1725
    /*
     * Cause any non-postcopiable, but iterative devices to
     * send out their final data.
     */
1726
    qemu_savevm_state_complete_precopy(ms->to_dst_file, true, false);
1727

1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743
    /*
     * in Finish migrate and with the io-lock held everything should
     * be quiet, but we've potentially still got dirty pages and we
     * need to tell the destination to throw any pages it's already received
     * that are dirty
     */
    if (ram_postcopy_send_discard_bitmap(ms)) {
        error_report("postcopy send discard bitmap failed");
        goto fail;
    }

    /*
     * send rest of state - note things that are doing postcopy
     * will notice we're in POSTCOPY_ACTIVE and not actually
     * wrap their state up here
     */
1744
    qemu_file_set_rate_limit(ms->to_dst_file, INT64_MAX);
1745
    /* Ping just for debugging, helps line traces up */
1746
    qemu_savevm_send_ping(ms->to_dst_file, 2);
1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758

    /*
     * While loading the device state we may trigger page transfer
     * requests and the fd must be free to process those, and thus
     * the destination must read the whole device state off the fd before
     * it starts processing it.  Unfortunately the ad-hoc migration format
     * doesn't allow the destination to know the size to read without fully
     * parsing it through each devices load-state code (especially the open
     * coded devices that use get/put).
     * So we wrap the device state up in a package with a length at the start;
     * to do this we use a qemu_buf to hold the whole of the device state.
     */
1759
    bioc = qio_channel_buffer_new(4096);
1760
    qio_channel_set_name(QIO_CHANNEL(bioc), "migration-postcopy-buffer");
1761 1762
    fb = qemu_fopen_channel_output(QIO_CHANNEL(bioc));
    object_unref(OBJECT(bioc));
1763

1764 1765 1766 1767 1768 1769
    /*
     * Make sure the receiver can get incoming pages before we send the rest
     * of the state
     */
    qemu_savevm_send_postcopy_listen(fb);

1770
    qemu_savevm_state_complete_precopy(fb, false, false);
1771 1772 1773 1774 1775 1776
    qemu_savevm_send_ping(fb, 3);

    qemu_savevm_send_postcopy_run(fb);

    /* <><> end of stuff going into the package */

1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788
    /* Last point of recovery; as soon as we send the package the destination
     * can open devices and potentially start running.
     * Lets just check again we've not got any errors.
     */
    ret = qemu_file_get_error(ms->to_dst_file);
    if (ret) {
        error_report("postcopy_start: Migration stream errored (pre package)");
        goto fail_closefb;
    }

    restart_block = false;

1789
    /* Now send that blob */
1790
    if (qemu_savevm_send_packaged(ms->to_dst_file, bioc->data, bioc->usage)) {
1791 1792 1793
        goto fail_closefb;
    }
    qemu_fclose(fb);
1794 1795 1796 1797 1798 1799 1800 1801

    /* Send a notify to give a chance for anything that needs to happen
     * at the transition to postcopy and after the device state; in particular
     * spice needs to trigger a transition now
     */
    ms->postcopy_after_devices = true;
    notifier_list_notify(&migration_state_notifiers, ms);

1802 1803 1804 1805 1806 1807 1808 1809
    ms->downtime =  qemu_clock_get_ms(QEMU_CLOCK_REALTIME) - time_at_stop;

    qemu_mutex_unlock_iothread();

    /*
     * Although this ping is just for debug, it could potentially be
     * used for getting a better measurement of downtime at the source.
     */
1810
    qemu_savevm_send_ping(ms->to_dst_file, 4);
1811

1812 1813 1814 1815
    if (migrate_release_ram()) {
        ram_postcopy_migrated_memory_release(ms);
    }

1816
    ret = qemu_file_get_error(ms->to_dst_file);
1817 1818
    if (ret) {
        error_report("postcopy_start: Migration stream errored");
1819
        migrate_set_state(&ms->state, MIGRATION_STATUS_POSTCOPY_ACTIVE,
1820 1821 1822 1823 1824 1825 1826 1827
                              MIGRATION_STATUS_FAILED);
    }

    return ret;

fail_closefb:
    qemu_fclose(fb);
fail:
1828
    migrate_set_state(&ms->state, MIGRATION_STATUS_POSTCOPY_ACTIVE,
1829
                          MIGRATION_STATUS_FAILED);
1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840
    if (restart_block) {
        /* A failure happened early enough that we know the destination hasn't
         * accessed block devices, so we're safe to recover.
         */
        Error *local_err = NULL;

        bdrv_invalidate_cache_all(&local_err);
        if (local_err) {
            error_report_err(local_err);
        }
    }
1841 1842 1843 1844
    qemu_mutex_unlock_iothread();
    return -1;
}

1845 1846 1847 1848 1849
/**
 * migration_completion: Used by migration_thread when there's not much left.
 *   The caller 'breaks' the loop when this returns.
 *
 * @s: Current migration state
1850
 * @current_active_state: The migration state we expect to be in
1851 1852 1853
 * @*old_vm_running: Pointer to old_vm_running flag
 * @*start_time: Pointer to time to update
 */
1854 1855
static void migration_completion(MigrationState *s, int current_active_state,
                                 bool *old_vm_running,
1856 1857 1858 1859
                                 int64_t *start_time)
{
    int ret;

1860 1861 1862 1863 1864 1865 1866 1867
    if (s->state == MIGRATION_STATUS_ACTIVE) {
        qemu_mutex_lock_iothread();
        *start_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
        qemu_system_wakeup_request(QEMU_WAKEUP_REASON_OTHER);
        *old_vm_running = runstate_is_running();
        ret = global_state_store();

        if (!ret) {
1868
            bool inactivate = !migrate_colo_enabled();
1869
            ret = vm_stop_force_state(RUN_STATE_FINISH_MIGRATE);
1870 1871
            if (ret >= 0) {
                qemu_file_set_rate_limit(s->to_dst_file, INT64_MAX);
1872 1873
                ret = qemu_savevm_state_complete_precopy(s->to_dst_file, false,
                                                         inactivate);
1874
            }
1875 1876
            if (inactivate && ret >= 0) {
                s->block_inactive = true;
1877 1878 1879
            }
        }
        qemu_mutex_unlock_iothread();
1880

1881 1882
        if (ret < 0) {
            goto fail;
1883
        }
1884 1885 1886
    } else if (s->state == MIGRATION_STATUS_POSTCOPY_ACTIVE) {
        trace_migration_completion_postcopy_end();

1887
        qemu_savevm_state_complete_postcopy(s->to_dst_file);
1888
        trace_migration_completion_postcopy_end_after_complete();
1889 1890
    }

1891 1892 1893 1894 1895 1896
    /*
     * If rp was opened we must clean up the thread before
     * cleaning everything else up (since if there are no failures
     * it will wait for the destination to send it's status in
     * a SHUT command).
     */
P
Peter Xu 已提交
1897
    if (s->rp_state.from_dst_file) {
1898
        int rp_error;
P
Peter Xu 已提交
1899
        trace_migration_return_path_end_before();
1900
        rp_error = await_return_path_close_on_source(s);
P
Peter Xu 已提交
1901
        trace_migration_return_path_end_after(rp_error);
1902
        if (rp_error) {
1903
            goto fail_invalidate;
1904
        }
1905 1906
    }

1907
    if (qemu_file_get_error(s->to_dst_file)) {
1908
        trace_migration_completion_file_err();
1909
        goto fail_invalidate;
1910 1911
    }

1912 1913 1914 1915 1916
    if (!migrate_colo_enabled()) {
        migrate_set_state(&s->state, current_active_state,
                          MIGRATION_STATUS_COMPLETED);
    }

1917 1918
    return;

1919 1920 1921 1922 1923 1924 1925
fail_invalidate:
    /* If not doing postcopy, vm_start() will be called: let's regain
     * control on images.
     */
    if (s->state == MIGRATION_STATUS_ACTIVE) {
        Error *local_err = NULL;

1926
        qemu_mutex_lock_iothread();
1927 1928 1929
        bdrv_invalidate_cache_all(&local_err);
        if (local_err) {
            error_report_err(local_err);
1930 1931
        } else {
            s->block_inactive = false;
1932
        }
1933
        qemu_mutex_unlock_iothread();
1934 1935
    }

1936
fail:
1937 1938
    migrate_set_state(&s->state, current_active_state,
                      MIGRATION_STATUS_FAILED);
1939 1940
}

1941 1942 1943 1944 1945 1946
bool migrate_colo_enabled(void)
{
    MigrationState *s = migrate_get_current();
    return s->enabled_capabilities[MIGRATION_CAPABILITY_X_COLO];
}

1947 1948 1949 1950
/*
 * Master migration thread on the source VM.
 * It drives the migration and pumps the data down the outgoing channel.
 */
J
Juan Quintela 已提交
1951
static void *migration_thread(void *opaque)
1952
{
1953
    MigrationState *s = opaque;
1954
    /* Used by the bandwidth calcs, updated later */
1955 1956
    int64_t initial_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
    int64_t setup_start = qemu_clock_get_ms(QEMU_CLOCK_HOST);
1957
    int64_t initial_bytes = 0;
1958 1959 1960 1961 1962 1963
    /*
     * The final stage happens when the remaining data is smaller than
     * this threshold; it's calculated from the requested downtime and
     * measured bandwidth
     */
    int64_t threshold_size = 0;
1964
    int64_t start_time = initial_time;
1965
    int64_t end_time;
1966
    bool old_vm_running = false;
1967 1968 1969
    bool entered_postcopy = false;
    /* The active state we expect to be in; ACTIVE or POSTCOPY_ACTIVE */
    enum MigrationStatus current_active_state = MIGRATION_STATUS_ACTIVE;
1970
    bool enable_colo = migrate_colo_enabled();
1971

1972 1973
    rcu_register_thread();

1974
    qemu_savevm_state_header(s->to_dst_file);
1975

1976 1977 1978 1979 1980
    /*
     * If we opened the return path, we need to make sure dst has it
     * opened as well.
     */
    if (s->rp_state.from_dst_file) {
1981
        /* Now tell the dest that it should open its end so it can reply */
1982
        qemu_savevm_send_open_return_path(s->to_dst_file);
1983 1984

        /* And do a ping that will make stuff easier to debug */
1985
        qemu_savevm_send_ping(s->to_dst_file, 1);
P
Peter Xu 已提交
1986
    }
1987

P
Peter Xu 已提交
1988
    if (migrate_postcopy_ram()) {
1989 1990 1991 1992 1993
        /*
         * Tell the destination that we *might* want to do postcopy later;
         * if the other end can't do postcopy it should fail now, nice and
         * early.
         */
1994
        qemu_savevm_send_postcopy_advise(s->to_dst_file);
1995 1996
    }

1997
    qemu_savevm_state_setup(s->to_dst_file);
1998

1999
    s->setup_time = qemu_clock_get_ms(QEMU_CLOCK_HOST) - setup_start;
2000 2001
    migrate_set_state(&s->state, MIGRATION_STATUS_SETUP,
                      MIGRATION_STATUS_ACTIVE);
2002

2003 2004 2005 2006
    trace_migration_thread_setup_complete();

    while (s->state == MIGRATION_STATUS_ACTIVE ||
           s->state == MIGRATION_STATUS_POSTCOPY_ACTIVE) {
2007
        int64_t current_time;
2008
        uint64_t pending_size;
2009

2010
        if (!qemu_file_rate_limit(s->to_dst_file)) {
2011 2012
            uint64_t pend_post, pend_nonpost;

2013 2014
            qemu_savevm_state_pending(s->to_dst_file, threshold_size,
                                      &pend_nonpost, &pend_post);
2015
            pending_size = pend_nonpost + pend_post;
2016
            trace_migrate_pending(pending_size, threshold_size,
2017
                                  pend_post, pend_nonpost);
2018
            if (pending_size && pending_size >= threshold_size) {
2019 2020 2021 2022
                /* Still a significant amount to transfer */

                if (migrate_postcopy_ram() &&
                    s->state != MIGRATION_STATUS_POSTCOPY_ACTIVE &&
2023
                    pend_nonpost <= threshold_size &&
2024 2025 2026 2027 2028 2029 2030 2031 2032 2033
                    atomic_read(&s->start_postcopy)) {

                    if (!postcopy_start(s, &old_vm_running)) {
                        current_active_state = MIGRATION_STATUS_POSTCOPY_ACTIVE;
                        entered_postcopy = true;
                    }

                    continue;
                }
                /* Just another iteration step */
2034
                qemu_savevm_state_iterate(s->to_dst_file, entered_postcopy);
2035
            } else {
2036
                trace_migration_thread_low_pending(pending_size);
2037
                migration_completion(s, current_active_state,
2038
                                     &old_vm_running, &start_time);
2039
                break;
2040 2041
            }
        }
2042

2043
        if (qemu_file_get_error(s->to_dst_file)) {
2044 2045
            migrate_set_state(&s->state, current_active_state,
                              MIGRATION_STATUS_FAILED);
2046
            trace_migration_thread_file_err();
2047 2048
            break;
        }
2049
        current_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
2050
        if (current_time >= initial_time + BUFFER_DELAY) {
2051 2052
            uint64_t transferred_bytes = qemu_ftell(s->to_dst_file) -
                                         initial_bytes;
2053
            uint64_t time_spent = current_time - initial_time;
2054
            double bandwidth = (double)transferred_bytes / time_spent;
2055
            threshold_size = bandwidth * s->parameters.downtime_limit;
2056

2057 2058
            s->mbps = (((double) transferred_bytes * 8.0) /
                    ((double) time_spent / 1000.0)) / 1000.0 / 1000.0;
2059

2060
            trace_migrate_transferred(transferred_bytes, time_spent,
2061
                                      bandwidth, threshold_size);
2062 2063
            /* if we haven't sent anything, we don't want to recalculate
               10000 is a small enough number for our purposes */
2064 2065
            if (ram_counters.dirty_pages_rate && transferred_bytes > 10000) {
                s->expected_downtime = ram_counters.dirty_pages_rate *
2066
                    qemu_target_page_size() / bandwidth;
2067
            }
2068

2069
            qemu_file_reset_rate_limit(s->to_dst_file);
2070
            initial_time = current_time;
2071
            initial_bytes = qemu_ftell(s->to_dst_file);
2072
        }
2073
        if (qemu_file_rate_limit(s->to_dst_file)) {
2074 2075 2076
            /* usleep expects microseconds */
            g_usleep((initial_time + BUFFER_DELAY - current_time)*1000);
        }
2077 2078
    }

2079
    trace_migration_thread_after_loop();
2080 2081
    /* If we enabled cpu throttling for auto-converge, turn it off. */
    cpu_throttle_stop();
2082
    end_time = qemu_clock_get_ms(QEMU_CLOCK_REALTIME);
2083

2084
    qemu_mutex_lock_iothread();
2085 2086 2087 2088 2089 2090 2091
    /*
     * The resource has been allocated by migration will be reused in COLO
     * process, so don't release them.
     */
    if (!enable_colo) {
        qemu_savevm_state_cleanup();
    }
2092
    if (s->state == MIGRATION_STATUS_COMPLETED) {
2093
        uint64_t transferred_bytes = qemu_ftell(s->to_dst_file);
2094
        s->total_time = end_time - s->total_time;
2095 2096 2097
        if (!entered_postcopy) {
            s->downtime = end_time - start_time;
        }
2098 2099 2100 2101
        if (s->total_time) {
            s->mbps = (((double) transferred_bytes * 8.0) /
                       ((double) s->total_time)) / 1000;
        }
2102 2103
        runstate_set(RUN_STATE_POSTMIGRATE);
    } else {
2104 2105 2106 2107 2108 2109 2110 2111 2112
        if (s->state == MIGRATION_STATUS_ACTIVE && enable_colo) {
            migrate_start_colo_process(s);
            qemu_savevm_state_cleanup();
            /*
            * Fixme: we will run VM in COLO no matter its old running state.
            * After exited COLO, we will keep running.
            */
            old_vm_running = true;
        }
2113
        if (old_vm_running && !entered_postcopy) {
2114
            vm_start();
2115 2116 2117 2118
        } else {
            if (runstate_check(RUN_STATE_FINISH_MIGRATE)) {
                runstate_set(RUN_STATE_POSTMIGRATE);
            }
2119
        }
2120
    }
2121
    qemu_bh_schedule(s->cleanup_bh);
2122
    qemu_mutex_unlock_iothread();
2123

2124
    rcu_unregister_thread();
2125 2126 2127
    return NULL;
}

2128
void migrate_fd_connect(MigrationState *s)
2129
{
2130
    s->expected_downtime = s->parameters.downtime_limit;
2131
    s->cleanup_bh = qemu_bh_new(migrate_fd_cleanup, s);
2132

2133
    qemu_file_set_blocking(s->to_dst_file, true);
2134
    qemu_file_set_rate_limit(s->to_dst_file,
2135
                             s->parameters.max_bandwidth / XFER_LIMIT_RATIO);
2136

2137 2138 2139
    /* Notify before starting migration thread */
    notifier_list_notify(&migration_state_notifiers, s);

2140
    /*
2141 2142 2143
     * Open the return path. For postcopy, it is used exclusively. For
     * precopy, only if user specified "return-path" capability would
     * QEMU uses the return path.
2144
     */
2145
    if (migrate_postcopy_ram() || migrate_use_return_path()) {
2146 2147
        if (open_return_path_on_source(s)) {
            error_report("Unable to open return-path for postcopy");
2148
            migrate_set_state(&s->state, MIGRATION_STATUS_SETUP,
2149 2150 2151 2152 2153 2154
                              MIGRATION_STATUS_FAILED);
            migrate_fd_cleanup(s);
            return;
        }
    }

2155
    qemu_thread_create(&s->thread, "live_migration", migration_thread, s,
2156
                       QEMU_THREAD_JOINABLE);
2157
    s->migration_thread_running = true;
2158
}
2159

P
Peter Xu 已提交
2160 2161 2162 2163 2164 2165 2166 2167 2168 2169
void migration_global_dump(Monitor *mon)
{
    MigrationState *ms = migrate_get_current();

    monitor_printf(mon, "globals: store-global-state=%d, only_migratable=%d, "
                   "send-configuration=%d, send-section-footer=%d\n",
                   ms->store_global_state, ms->only_migratable,
                   ms->send_configuration, ms->send_section_footer);
}

2170 2171 2172
#define DEFINE_PROP_MIG_CAP(name, x)             \
    DEFINE_PROP_BOOL(name, MigrationState, enabled_capabilities[x], false)

2173 2174 2175
static Property migration_properties[] = {
    DEFINE_PROP_BOOL("store-global-state", MigrationState,
                     store_global_state, true),
2176
    DEFINE_PROP_BOOL("only-migratable", MigrationState, only_migratable, false),
2177 2178
    DEFINE_PROP_BOOL("send-configuration", MigrationState,
                     send_configuration, true),
P
Peter Xu 已提交
2179 2180
    DEFINE_PROP_BOOL("send-section-footer", MigrationState,
                     send_section_footer, true),
P
Peter Xu 已提交
2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205

    /* Migration parameters */
    DEFINE_PROP_INT64("x-compress-level", MigrationState,
                      parameters.compress_level,
                      DEFAULT_MIGRATE_COMPRESS_LEVEL),
    DEFINE_PROP_INT64("x-compress-threads", MigrationState,
                      parameters.compress_threads,
                      DEFAULT_MIGRATE_COMPRESS_THREAD_COUNT),
    DEFINE_PROP_INT64("x-decompress-threads", MigrationState,
                      parameters.decompress_threads,
                      DEFAULT_MIGRATE_DECOMPRESS_THREAD_COUNT),
    DEFINE_PROP_INT64("x-cpu-throttle-initial", MigrationState,
                      parameters.cpu_throttle_initial,
                      DEFAULT_MIGRATE_CPU_THROTTLE_INITIAL),
    DEFINE_PROP_INT64("x-cpu-throttle-increment", MigrationState,
                      parameters.cpu_throttle_increment,
                      DEFAULT_MIGRATE_CPU_THROTTLE_INCREMENT),
    DEFINE_PROP_INT64("x-max-bandwidth", MigrationState,
                      parameters.max_bandwidth, MAX_THROTTLE),
    DEFINE_PROP_INT64("x-downtime-limit", MigrationState,
                      parameters.downtime_limit,
                      DEFAULT_MIGRATE_SET_DOWNTIME),
    DEFINE_PROP_INT64("x-checkpoint-delay", MigrationState,
                      parameters.x_checkpoint_delay,
                      DEFAULT_MIGRATE_X_CHECKPOINT_DELAY),
2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219

    /* Migration capabilities */
    DEFINE_PROP_MIG_CAP("x-xbzrle", MIGRATION_CAPABILITY_XBZRLE),
    DEFINE_PROP_MIG_CAP("x-rdma-pin-all", MIGRATION_CAPABILITY_RDMA_PIN_ALL),
    DEFINE_PROP_MIG_CAP("x-auto-converge", MIGRATION_CAPABILITY_AUTO_CONVERGE),
    DEFINE_PROP_MIG_CAP("x-zero-blocks", MIGRATION_CAPABILITY_ZERO_BLOCKS),
    DEFINE_PROP_MIG_CAP("x-compress", MIGRATION_CAPABILITY_COMPRESS),
    DEFINE_PROP_MIG_CAP("x-events", MIGRATION_CAPABILITY_EVENTS),
    DEFINE_PROP_MIG_CAP("x-postcopy-ram", MIGRATION_CAPABILITY_POSTCOPY_RAM),
    DEFINE_PROP_MIG_CAP("x-colo", MIGRATION_CAPABILITY_X_COLO),
    DEFINE_PROP_MIG_CAP("x-release-ram", MIGRATION_CAPABILITY_RELEASE_RAM),
    DEFINE_PROP_MIG_CAP("x-block", MIGRATION_CAPABILITY_BLOCK),
    DEFINE_PROP_MIG_CAP("x-return-path", MIGRATION_CAPABILITY_RETURN_PATH),

2220 2221 2222
    DEFINE_PROP_END_OF_LIST(),
};

2223 2224 2225 2226 2227
static void migration_class_init(ObjectClass *klass, void *data)
{
    DeviceClass *dc = DEVICE_CLASS(klass);

    dc->user_creatable = false;
2228
    dc->props = migration_properties;
2229 2230
}

M
Marc-André Lureau 已提交
2231 2232 2233 2234 2235 2236 2237 2238 2239
static void migration_instance_finalize(Object *obj)
{
    MigrationState *ms = MIGRATION_OBJ(obj);
    MigrationParameters *params = &ms->parameters;

    g_free(params->tls_hostname);
    g_free(params->tls_creds);
}

2240 2241 2242
static void migration_instance_init(Object *obj)
{
    MigrationState *ms = MIGRATION_OBJ(obj);
2243
    MigrationParameters *params = &ms->parameters;
2244 2245 2246 2247

    ms->state = MIGRATION_STATUS_NONE;
    ms->xbzrle_cache_size = DEFAULT_MIGRATE_CACHE_SIZE;
    ms->mbps = -1;
2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269

    params->tls_hostname = g_strdup("");
    params->tls_creds = g_strdup("");

    /* Set has_* up only for parameter checks */
    params->has_compress_level = true;
    params->has_compress_threads = true;
    params->has_decompress_threads = true;
    params->has_cpu_throttle_initial = true;
    params->has_cpu_throttle_increment = true;
    params->has_max_bandwidth = true;
    params->has_downtime_limit = true;
    params->has_x_checkpoint_delay = true;
    params->has_block_incremental = true;
}

/*
 * Return true if check pass, false otherwise. Error will be put
 * inside errp if provided.
 */
static bool migration_object_check(MigrationState *ms, Error **errp)
{
2270 2271 2272 2273 2274
    MigrationCapabilityStatusList *head = NULL;
    /* Assuming all off */
    bool cap_list[MIGRATION_CAPABILITY__MAX] = { 0 }, ret;
    int i;

2275 2276 2277 2278
    if (!migrate_params_check(&ms->parameters, errp)) {
        return false;
    }

2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290
    for (i = 0; i < MIGRATION_CAPABILITY__MAX; i++) {
        if (ms->enabled_capabilities[i]) {
            head = migrate_cap_add(head, i, true);
        }
    }

    ret = migrate_caps_check(cap_list, head, errp);

    /* It works with head == NULL */
    qapi_free_MigrationCapabilityStatusList(head);

    return ret;
2291 2292 2293 2294
}

static const TypeInfo migration_type = {
    .name = TYPE_MIGRATION,
2295
    /*
P
Peter Xu 已提交
2296 2297 2298 2299 2300 2301
     * NOTE: TYPE_MIGRATION is not really a device, as the object is
     * not created using qdev_create(), it is not attached to the qdev
     * device tree, and it is never realized.
     *
     * TODO: Make this TYPE_OBJECT once QOM provides something like
     * TYPE_DEVICE's "-global" properties.
2302
     */
2303 2304 2305 2306 2307
    .parent = TYPE_DEVICE,
    .class_init = migration_class_init,
    .class_size = sizeof(MigrationClass),
    .instance_size = sizeof(MigrationState),
    .instance_init = migration_instance_init,
M
Marc-André Lureau 已提交
2308
    .instance_finalize = migration_instance_finalize,
2309 2310 2311 2312 2313 2314 2315 2316
};

static void register_migration_types(void)
{
    type_register_static(&migration_type);
}

type_init(register_migration_types);