qemu_migration.c 187.5 KB
Newer Older
1 2 3
/*
 * qemu_migration.c: QEMU migration handling
 *
4
 * Copyright (C) 2006-2015 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23
 *
 */

#include <config.h>

J
Ján Tomko 已提交
24 25
#include <netdb.h>
#include <sys/socket.h>
26
#include <sys/time.h>
27
#include <fcntl.h>
28
#include <poll.h>
29 30

#include "qemu_migration.h"
31
#include "qemu_migration_cookie.h"
32
#include "qemu_migration_params.h"
33 34 35 36
#include "qemu_monitor.h"
#include "qemu_domain.h"
#include "qemu_process.h"
#include "qemu_capabilities.h"
37
#include "qemu_alias.h"
38
#include "qemu_cgroup.h"
39
#include "qemu_hotplug.h"
40
#include "qemu_blockjob.h"
41
#include "qemu_security.h"
42
#include "qemu_block.h"
43

44
#include "domain_audit.h"
45
#include "virlog.h"
46
#include "virerror.h"
47
#include "viralloc.h"
E
Eric Blake 已提交
48
#include "virfile.h"
49
#include "virnetdevopenvswitch.h"
50
#include "datatypes.h"
51
#include "virfdstream.h"
52
#include "viruuid.h"
53
#include "virtime.h"
54
#include "locking/domain_lock.h"
55
#include "rpc/virnetsocket.h"
56
#include "virstoragefile.h"
M
Martin Kletzander 已提交
57
#include "viruri.h"
58
#include "virhook.h"
59
#include "virstring.h"
60
#include "virtypedparam.h"
M
Michael R. Hines 已提交
61
#include "virprocess.h"
62
#include "nwfilter_conf.h"
63 64 65

#define VIR_FROM_THIS VIR_FROM_QEMU

66 67
VIR_LOG_INIT("qemu.qemu_migration");

68 69 70 71 72 73 74 75 76 77 78 79 80
VIR_ENUM_IMPL(qemuMigrationJobPhase, QEMU_MIGRATION_PHASE_LAST,
              "none",
              "perform2",
              "begin3",
              "perform3",
              "perform3_done",
              "confirm3_cancelled",
              "confirm3",
              "prepare",
              "finish2",
              "finish3",
);

81 82 83
static int
qemuMigrationJobStart(virQEMUDriverPtr driver,
                      virDomainObjPtr vm,
84 85
                      qemuDomainAsyncJob job,
                      unsigned long apiFlags)
86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112
    ATTRIBUTE_NONNULL(1) ATTRIBUTE_NONNULL(2) ATTRIBUTE_RETURN_CHECK;

static void
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMigrationJobPhase phase)
    ATTRIBUTE_NONNULL(1) ATTRIBUTE_NONNULL(2);

static void
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuMigrationJobPhase phase)
    ATTRIBUTE_NONNULL(1) ATTRIBUTE_NONNULL(2);

static void
qemuMigrationJobContinue(virDomainObjPtr obj)
    ATTRIBUTE_NONNULL(1);

static bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
                         qemuDomainAsyncJob job)
    ATTRIBUTE_NONNULL(1);

static void
qemuMigrationJobFinish(virQEMUDriverPtr driver,
                       virDomainObjPtr obj)
    ATTRIBUTE_NONNULL(1) ATTRIBUTE_NONNULL(2);
113

114
static void
115
qemuMigrationSrcStoreDomainState(virDomainObjPtr vm)
116 117 118 119 120 121 122 123 124 125
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    priv->preMigrationState = virDomainObjGetState(vm, NULL);

    VIR_DEBUG("Storing pre-migration state=%d domain=%p",
              priv->preMigrationState, vm);
}

/* Returns true if the domain was resumed, false otherwise */
static bool
126
qemuMigrationSrcRestoreDomainState(virQEMUDriverPtr driver, virDomainObjPtr vm)
127 128
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
129 130
    int reason;
    virDomainState state = virDomainObjGetState(vm, &reason);
131 132
    bool ret = false;

133 134 135 136 137
    VIR_DEBUG("driver=%p, vm=%p, pre-mig-state=%s, state=%s, reason=%s",
              driver, vm,
              virDomainStateTypeToString(priv->preMigrationState),
              virDomainStateTypeToString(state),
              virDomainStateReasonToString(state, reason));
138

139 140 141 142 143
    if (state != VIR_DOMAIN_PAUSED ||
        reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED)
        goto cleanup;

    if (priv->preMigrationState == VIR_DOMAIN_RUNNING) {
144 145 146 147 148 149
        /* This is basically the only restore possibility that's safe
         * and we should attempt to do */

        VIR_DEBUG("Restoring pre-migration state due to migration error");

        /* we got here through some sort of failure; start the domain again */
150
        if (qemuProcessStartCPUs(driver, vm,
151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166
                                 VIR_DOMAIN_RUNNING_MIGRATION_CANCELED,
                                 QEMU_ASYNC_JOB_MIGRATION_OUT) < 0) {
            /* Hm, we already know we are in error here.  We don't want to
             * overwrite the previous error, though, so we just throw something
             * to the logs and hope for the best */
            VIR_ERROR(_("Failed to resume guest %s after failure"), vm->def->name);
            goto cleanup;
        }
        ret = true;
    }

 cleanup:
    priv->preMigrationState = VIR_DOMAIN_NOSTATE;
    return ret;
}

167 168

static int
169 170 171
qemuMigrationDstPrecreateDisk(virConnectPtr conn,
                              virDomainDiskDefPtr disk,
                              unsigned long long capacity)
172 173 174 175 176 177 178 179 180 181 182 183
{
    int ret = -1;
    virStoragePoolPtr pool = NULL;
    virStorageVolPtr vol = NULL;
    char *volName = NULL, *basePath = NULL;
    char *volStr = NULL;
    virBuffer buf = VIR_BUFFER_INITIALIZER;
    const char *format = NULL;
    unsigned int flags = 0;

    VIR_DEBUG("Precreate disk type=%s", virStorageTypeToString(disk->src->type));

184
    switch ((virStorageType)disk->src->type) {
185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204
    case VIR_STORAGE_TYPE_FILE:
        if (!virDomainDiskGetSource(disk)) {
            VIR_DEBUG("Dropping sourceless disk '%s'",
                      disk->dst);
            return 0;
        }

        if (VIR_STRDUP(basePath, disk->src->path) < 0)
            goto cleanup;

        if (!(volName = strrchr(basePath, '/'))) {
            virReportError(VIR_ERR_INVALID_ARG,
                           _("malformed disk path: %s"),
                           disk->src->path);
            goto cleanup;
        }

        *volName = '\0';
        volName++;

205
        if (!(pool = virStoragePoolLookupByTargetPath(conn, basePath)))
206 207 208 209 210 211 212 213 214 215 216 217 218 219 220
            goto cleanup;
        format = virStorageFileFormatTypeToString(disk->src->format);
        if (disk->src->format == VIR_STORAGE_FILE_QCOW2)
            flags |= VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA;
        break;

    case VIR_STORAGE_TYPE_VOLUME:
        if (!(pool = virStoragePoolLookupByName(conn, disk->src->srcpool->pool)))
            goto cleanup;
        format = virStorageFileFormatTypeToString(disk->src->format);
        volName = disk->src->srcpool->volume;
        if (disk->src->format == VIR_STORAGE_FILE_QCOW2)
            flags |= VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA;
        break;

221 222 223 224 225
    case VIR_STORAGE_TYPE_NETWORK:
        VIR_DEBUG("Skipping creation of network disk '%s'",
                  disk->dst);
        return 0;

226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272
    case VIR_STORAGE_TYPE_BLOCK:
    case VIR_STORAGE_TYPE_DIR:
    case VIR_STORAGE_TYPE_NONE:
    case VIR_STORAGE_TYPE_LAST:
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot precreate storage for disk type '%s'"),
                       virStorageTypeToString(disk->src->type));
        goto cleanup;
    }

    if ((vol = virStorageVolLookupByName(pool, volName))) {
        VIR_DEBUG("Skipping creation of already existing volume of name '%s'",
                  volName);
        ret = 0;
        goto cleanup;
    }

    virBufferAddLit(&buf, "<volume>\n");
    virBufferAdjustIndent(&buf, 2);
    virBufferEscapeString(&buf, "<name>%s</name>\n", volName);
    virBufferAsprintf(&buf, "<capacity>%llu</capacity>\n", capacity);
    virBufferAddLit(&buf, "<target>\n");
    virBufferAdjustIndent(&buf, 2);
    virBufferAsprintf(&buf, "<format type='%s'/>\n", format);
    virBufferAdjustIndent(&buf, -2);
    virBufferAddLit(&buf, "</target>\n");
    virBufferAdjustIndent(&buf, -2);
    virBufferAddLit(&buf, "</volume>\n");

    if (!(volStr = virBufferContentAndReset(&buf))) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("unable to create volume XML"));
        goto cleanup;
    }

    if (!(vol = virStorageVolCreateXML(pool, volStr, flags)))
        goto cleanup;

    ret = 0;
 cleanup:
    VIR_FREE(basePath);
    VIR_FREE(volStr);
    virObjectUnref(vol);
    virObjectUnref(pool);
    return ret;
}

273
static bool
274 275
qemuMigrationAnyCopyDisk(virDomainDiskDef const *disk,
                         size_t nmigrate_disks, const char **migrate_disks)
276 277 278 279 280 281 282 283 284 285 286 287 288 289 290
{
    size_t i;

    /* Check if the disk alias is in the list */
    if (nmigrate_disks) {
        for (i = 0; i < nmigrate_disks; i++) {
            if (STREQ(disk->dst, migrate_disks[i]))
                return true;
        }
        return false;
    }

    /* Default is to migrate only non-shared non-readonly disks
     * with source */
    return !disk->src->shared && !disk->src->readonly &&
291
           !virStorageSourceIsEmpty(disk->src);
292 293
}

294 295

static int
296 297 298 299 300
qemuMigrationDstPrecreateStorage(virDomainObjPtr vm,
                                 qemuMigrationCookieNBDPtr nbd,
                                 size_t nmigrate_disks,
                                 const char **migrate_disks,
                                 bool incremental)
301 302 303
{
    int ret = -1;
    size_t i = 0;
304
    virConnectPtr conn;
305 306 307 308

    if (!nbd || !nbd->ndisks)
        return 0;

309 310 311
    if (!(conn = virGetConnectStorage()))
        return -1;

312 313 314 315
    for (i = 0; i < nbd->ndisks; i++) {
        virDomainDiskDefPtr disk;
        const char *diskSrcPath;

316
        VIR_DEBUG("Looking up disk target '%s' (capacity=%llu)",
317 318
                  nbd->disks[i].target, nbd->disks[i].capacity);

319 320
        if (!(disk = virDomainDiskByName(vm->def, nbd->disks[i].target,
                                         false))) {
321 322 323 324 325 326 327 328
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unable to find disk by target: %s"),
                           nbd->disks[i].target);
            goto cleanup;
        }

        diskSrcPath = virDomainDiskGetSource(disk);

329
        /* Skip disks we don't want to migrate and already existing disks. */
330
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks) ||
331 332 333 334
            (diskSrcPath && virFileExists(diskSrcPath))) {
            continue;
        }

335 336 337 338 339 340 341
        if (incremental) {
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                           _("pre-creation of storage targets for incremental "
                             "storage migration is not supported"));
            goto cleanup;
        }

342 343
        VIR_DEBUG("Proceeding with disk source %s", NULLSTR(diskSrcPath));

344
        if (qemuMigrationDstPrecreateDisk(conn, disk, nbd->disks[i].capacity) < 0)
345 346 347 348 349
            goto cleanup;
    }

    ret = 0;
 cleanup:
350
    virObjectUnref(conn);
351 352 353 354
    return ret;
}


355
/**
356
 * qemuMigrationDstStartNBDServer:
357 358 359 360 361 362 363 364 365 366 367
 * @driver: qemu driver
 * @vm: domain
 *
 * Starts NBD server. This is a newer method to copy
 * storage during migration than using 'blk' and 'inc'
 * arguments in 'migrate' monitor command.
 * Error is reported here.
 *
 * Returns 0 on success, -1 otherwise.
 */
static int
368 369 370 371 372
qemuMigrationDstStartNBDServer(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
                               const char *listenAddr,
                               size_t nmigrate_disks,
                               const char **migrate_disks,
373 374
                               int nbdPort,
                               const char *tls_alias)
375 376 377 378 379 380 381
{
    int ret = -1;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    unsigned short port = 0;
    char *diskAlias = NULL;
    size_t i;

382 383 384 385 386 387
    if (nbdPort < 0 || nbdPort > USHRT_MAX) {
        virReportError(VIR_ERR_INVALID_ARG, "%s",
                       _("nbd port must be in range 0-65535"));
        return -1;
    }

388 389 390
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];

391
        /* check whether disk should be migrated */
392
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
393 394
            continue;

395
        if (disk->src->readonly || virStorageSourceIsEmpty(disk->src)) {
396
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED,
397 398
                           _("Cannot migrate empty or read-only disk %s"),
                           disk->dst);
399 400 401
            goto cleanup;
        }

402
        VIR_FREE(diskAlias);
403
        if (!(diskAlias = qemuAliasDiskDriveFromDisk(disk)))
404 405 406 407 408 409
            goto cleanup;

        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
            goto cleanup;

410 411 412 413 414 415
        if (port == 0) {
            if (nbdPort)
                port = nbdPort;
            else if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
                goto exit_monitor;

416
            if (qemuMonitorNBDServerStart(priv->mon, listenAddr, port, tls_alias) < 0)
417
                goto exit_monitor;
418 419
        }

420 421 422
        if (qemuMonitorNBDServerAdd(priv->mon, diskAlias, true) < 0)
            goto exit_monitor;
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
423 424 425 426 427 428
            goto cleanup;
    }

    priv->nbdPort = port;
    ret = 0;

429
 cleanup:
430
    VIR_FREE(diskAlias);
431
    if (ret < 0 && nbdPort == 0)
432
        virPortAllocatorRelease(port);
433
    return ret;
434 435 436 437

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;
438 439
}

440 441

static int
442 443 444
qemuMigrationDstStopNBDServer(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              qemuMigrationCookiePtr mig)
445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (!mig->nbd)
        return 0;

    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        return -1;

    if (qemuMonitorNBDServerStop(priv->mon) < 0)
        VIR_WARN("Unable to stop NBD server");
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;

460
    virPortAllocatorRelease(priv->nbdPort);
461 462 463 464 465
    priv->nbdPort = 0;
    return 0;
}


466
static void
467
qemuMigrationNBDReportMirrorError(virDomainDiskDefPtr disk)
468
{
469 470 471
    qemuBlockJobDataPtr job = QEMU_DOMAIN_DISK_PRIVATE(disk)->blockjob;

    if (job->errmsg) {
472 473
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("migration of disk %s failed: %s"),
474
                       disk->dst, job->errmsg);
475 476 477 478 479 480 481
    } else {
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("migration of disk %s failed"), disk->dst);
    }
}


482
/**
483
 * qemuMigrationSrcNBDStorageCopyReady:
484 485
 * @vm: domain
 *
486 487
 * Check the status of all drives copied via qemuMigrationSrcNBDStorageCopy.
 * Any pending block job events for the mirrored disks will be processed.
488 489 490 491 492 493
 *
 * Returns 1 if all mirrors are "ready",
 *         0 if some mirrors are still performing initial sync,
 *        -1 on error.
 */
static int
494 495
qemuMigrationSrcNBDStorageCopyReady(virDomainObjPtr vm,
                                    qemuDomainAsyncJob asyncJob)
496 497
{
    size_t i;
498 499
    size_t notReady = 0;
    int status;
500 501 502

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
503
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
504

505
        if (!diskPriv->migrating)
506 507
            continue;

508
        status = qemuBlockJobUpdateDisk(vm, asyncJob, disk);
509
        if (status == VIR_DOMAIN_BLOCK_JOB_FAILED) {
510
            qemuMigrationNBDReportMirrorError(disk);
511 512
            return -1;
        }
513 514 515

        if (disk->mirrorState != VIR_DOMAIN_DISK_MIRROR_STATE_READY)
            notReady++;
516 517
    }

518 519 520 521 522 523 524
    if (notReady) {
        VIR_DEBUG("Waiting for %zu disk mirrors to get ready", notReady);
        return 0;
    } else {
        VIR_DEBUG("All disk mirrors are ready");
        return 1;
    }
525 526 527
}


528 529 530 531
/*
 * If @check is true, the function will report an error and return a different
 * code in case a block job fails. This way we can properly abort migration in
 * case some block jobs failed once all memory has already been transferred.
532
 *
533 534 535 536
 * Returns 1 if all mirrors are gone,
 *         0 if some mirrors are still active,
 *         -1 some mirrors failed but some are still active,
 *         -2 all mirrors are gone but some of them failed.
537 538
 */
static int
539 540 541
qemuMigrationSrcNBDCopyCancelled(virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 bool check)
542
{
543 544
    size_t i;
    size_t active = 0;
545
    size_t completed = 0;
546 547
    int status;
    bool failed = false;
548

549
 retry:
550 551 552
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
553

554 555
        if (!diskPriv->migrating)
            continue;
556

557
        status = qemuBlockJobUpdateDisk(vm, asyncJob, disk);
558 559 560
        switch (status) {
        case VIR_DOMAIN_BLOCK_JOB_FAILED:
            if (check) {
561
                qemuMigrationNBDReportMirrorError(disk);
562
                failed = true;
563
            }
M
Marc Hartmayer 已提交
564
            ATTRIBUTE_FALLTHROUGH;
565 566
        case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
567
            qemuBlockJobSyncEndDisk(vm, asyncJob, disk);
568 569
            diskPriv->migrating = false;
            break;
570

571 572
        default:
            active++;
573
        }
574 575 576 577 578 579 580 581 582 583 584 585 586

        if (status == VIR_DOMAIN_BLOCK_JOB_COMPLETED)
            completed++;
    }

    /* Updating completed block job drops the lock thus we have to recheck
     * block jobs for disks that reside before the disk(s) with completed
     * block job.
     */
    if (completed > 0) {
        completed = 0;
        active = 0;
        goto retry;
587
    }
588

589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604
    if (failed) {
        if (active) {
            VIR_DEBUG("Some disk mirrors failed; still waiting for %zu "
                      "disk mirrors to finish", active);
            return -1;
        } else {
            VIR_DEBUG("All disk mirrors are gone; some of them failed");
            return -2;
        }
    } else {
        if (active) {
            VIR_DEBUG("Waiting for %zu disk mirrors to finish", active);
            return 0;
        } else {
            VIR_DEBUG("All disk mirrors are gone");
            return 1;
605
        }
606
    }
607
}
608 609


610 611 612 613 614 615
/*
 * Returns 0 on success,
 *         1 when job is already completed or it failed and failNoJob is false,
 *         -1 on error or when job failed and failNoJob is true.
 */
static int
616 617 618 619 620
qemuMigrationSrcNBDCopyCancelOne(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 virDomainDiskDefPtr disk,
                                 bool failNoJob,
                                 qemuDomainAsyncJob asyncJob)
621 622 623 624 625 626 627
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char *diskAlias = NULL;
    int ret = -1;
    int status;
    int rv;

628
    status = qemuBlockJobUpdateDisk(vm, asyncJob, disk);
629 630 631 632
    switch (status) {
    case VIR_DOMAIN_BLOCK_JOB_FAILED:
    case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        if (failNoJob) {
633
            qemuMigrationNBDReportMirrorError(disk);
634
            goto cleanup;
635
        }
636
        ATTRIBUTE_FALLTHROUGH;
637
    case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
638 639
        ret = 1;
        goto cleanup;
640 641
    }

642
    if (!(diskAlias = qemuAliasDiskDriveFromDisk(disk)))
643 644
        return -1;

645
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
646 647
        goto cleanup;

648
    rv = qemuMonitorBlockJobCancel(priv->mon, diskAlias);
649 650 651 652 653

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rv < 0)
        goto cleanup;

    ret = 0;
654 655 656 657 658 659 660 661

 cleanup:
    VIR_FREE(diskAlias);
    return ret;
}


/**
662
 * qemuMigrationSrcNBDCopyCancel:
663 664
 * @driver: qemu driver
 * @vm: domain
665
 * @check: if true report an error when some of the mirrors fails
666
 *
667
 * Cancel all drive-mirrors started by qemuMigrationSrcNBDStorageCopy.
668 669 670
 * Any pending block job events for the affected disks will be processed and
 * synchronous block job terminated regardless of return value unless qemu
 * has crashed.
671 672 673 674
 *
 * Returns 0 on success, -1 otherwise.
 */
static int
675 676 677 678 679
qemuMigrationSrcNBDCopyCancel(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              bool check,
                              qemuDomainAsyncJob asyncJob,
                              virConnectPtr dconn)
680
{
681
    virErrorPtr err = NULL;
682
    int ret = -1;
683
    size_t i;
684 685 686 687
    int rv;
    bool failed = false;

    VIR_DEBUG("Cancelling drive mirrors for domain %s", vm->def->name);
688 689 690

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
691
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
692
        qemuBlockJobDataPtr job;
693

694 695
        if (!(job = qemuBlockJobDiskGetJob(disk)) ||
            !qemuBlockJobIsRunning(job))
696 697
            diskPriv->migrating = false;

698 699
        if (!diskPriv->migrating) {
            virObjectUnref(job);
700
            continue;
701
        }
702

703 704
        rv = qemuMigrationSrcNBDCopyCancelOne(driver, vm, disk,
                                              check, asyncJob);
705 706 707 708 709 710
        if (rv != 0) {
            if (rv < 0) {
                if (!err)
                    err = virSaveLastError();
                failed = true;
            }
711
            qemuBlockJobSyncEndDisk(vm, asyncJob, disk);
712 713
            diskPriv->migrating = false;
        }
714 715

        virObjectUnref(job);
716 717
    }

718
    while ((rv = qemuMigrationSrcNBDCopyCancelled(vm, asyncJob, check)) != 1) {
719 720 721 722 723 724 725
        if (check && !failed &&
            dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            failed = true;
        }

726 727 728 729
        if (rv < 0) {
            failed = true;
            if (rv == -2)
                break;
730
        }
731

732 733 734 735 736
        if (failed && !err)
            err = virSaveLastError();

        if (virDomainObjWait(vm) < 0)
            goto cleanup;
737 738
    }

739 740 741 742 743 744 745 746 747 748 749 750 751
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);

        if (!diskPriv->migrSource)
            continue;

        qemuBlockStorageSourceDetachOneBlockdev(driver, vm, asyncJob,
                                                diskPriv->migrSource);
        virStorageSourceFree(diskPriv->migrSource);
        diskPriv->migrSource = NULL;
    }

752 753 754
    ret = failed ? -1 : 0;

 cleanup:
755 756 757 758 759
    if (err) {
        virSetError(err);
        virFreeError(err);
    }
    return ret;
760 761 762
}


763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841
static int
qemuMigrationSrcNBDStorageCopyBlockdev(virQEMUDriverPtr driver,
                                       virDomainObjPtr vm,
                                       virDomainDiskDefPtr disk,
                                       const char *diskAlias,
                                       const char *host,
                                       int port,
                                       unsigned long long mirror_speed,
                                       unsigned int mirror_flags,
                                       const char *tlsAlias)
{
    qemuBlockStorageSourceAttachDataPtr data = NULL;
    qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
    virStorageSourcePtr copysrc = NULL;
    int mon_ret = 0;
    int ret = -1;

    VIR_DEBUG("starting blockdev mirror for disk=%s to host=%s", diskAlias, host);

    if (VIR_ALLOC(copysrc) < 0)
        goto cleanup;

    copysrc->type = VIR_STORAGE_TYPE_NETWORK;
    copysrc->protocol = VIR_STORAGE_NET_PROTOCOL_NBD;
    copysrc->format = VIR_STORAGE_FILE_RAW;

    if (VIR_ALLOC(copysrc->backingStore) < 0)
        goto cleanup;

    if (VIR_STRDUP(copysrc->path, diskAlias) < 0)
        goto cleanup;

    if (VIR_ALLOC_N(copysrc->hosts, 1) < 0)
        goto cleanup;

    copysrc->nhosts = 1;
    copysrc->hosts->transport = VIR_STORAGE_NET_HOST_TRANS_TCP;
    copysrc->hosts->port = port;
    if (VIR_STRDUP(copysrc->hosts->name, host) < 0)
        goto cleanup;

    if (VIR_STRDUP(copysrc->tlsAlias, tlsAlias) < 0)
        goto cleanup;

    if (virAsprintf(&copysrc->nodestorage, "migration-%s-storage", disk->dst) < 0 ||
        virAsprintf(&copysrc->nodeformat, "migration-%s-format", disk->dst) < 0)
        goto cleanup;

    if (!(data = qemuBlockStorageSourceAttachPrepareBlockdev(copysrc)))
        goto cleanup;

    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

    mon_ret = qemuBlockStorageSourceAttachApply(qemuDomainGetMonitor(vm), data);

    if (mon_ret == 0)
        mon_ret = qemuMonitorBlockdevMirror(qemuDomainGetMonitor(vm), NULL,
                                            diskAlias, copysrc->nodeformat,
                                            mirror_speed, 0, 0, mirror_flags);

    if (mon_ret != 0)
        qemuBlockStorageSourceAttachRollback(qemuDomainGetMonitor(vm), data);

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || mon_ret < 0)
        goto cleanup;

    VIR_STEAL_PTR(diskPriv->migrSource, copysrc);

    ret = 0;

 cleanup:
    qemuBlockStorageSourceAttachDataFree(data);
    virStorageSourceFree(copysrc);
    return ret;
}


842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883
static int
qemuMigrationSrcNBDStorageCopyDriveMirror(virQEMUDriverPtr driver,
                                          virDomainObjPtr vm,
                                          const char *diskAlias,
                                          const char *host,
                                          int port,
                                          unsigned long long mirror_speed,
                                          unsigned int mirror_flags)
{
    char *nbd_dest = NULL;
    int mon_ret;
    int ret = -1;

    if (strchr(host, ':')) {
        if (virAsprintf(&nbd_dest, "nbd:[%s]:%d:exportname=%s",
                        host, port, diskAlias) < 0)
            goto cleanup;
    } else {
        if (virAsprintf(&nbd_dest, "nbd:%s:%d:exportname=%s",
                        host, port, diskAlias) < 0)
            goto cleanup;
    }

    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

    mon_ret = qemuMonitorDriveMirror(qemuDomainGetMonitor(vm),
                                     diskAlias, nbd_dest, "raw",
                                     mirror_speed, 0, 0, mirror_flags);

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || mon_ret < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    VIR_FREE(nbd_dest);
    return ret;
}


884 885 886 887 888 889 890 891 892 893 894 895
static int
qemuMigrationSrcNBDStorageCopyOne(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  virDomainDiskDefPtr disk,
                                  const char *host,
                                  int port,
                                  unsigned long long mirror_speed,
                                  unsigned int mirror_flags,
                                  const char *tlsAlias,
                                  unsigned int flags)
{
    qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
896
    qemuBlockJobDataPtr job = NULL;
897 898 899 900 901 902 903
    char *diskAlias = NULL;
    int rc;
    int ret = -1;

    if (!(diskAlias = qemuAliasDiskDriveFromDisk(disk)))
        goto cleanup;

904
    if (!(job = qemuBlockJobDiskNew(disk, QEMU_BLOCKJOB_TYPE_COPY)))
905 906
        goto cleanup;

907
    qemuBlockJobSyncBegin(job);
908 909 910 911 912 913 914 915 916 917 918 919 920 921 922

    if (flags & VIR_MIGRATE_TLS) {
        rc = qemuMigrationSrcNBDStorageCopyBlockdev(driver, vm,
                                                    disk, diskAlias,
                                                    host, port,
                                                    mirror_speed,
                                                    mirror_flags,
                                                    tlsAlias);
    } else {
        rc = qemuMigrationSrcNBDStorageCopyDriveMirror(driver, vm, diskAlias,
                                                       host, port,
                                                       mirror_speed,
                                                       mirror_flags);
    }

923
    if (rc < 0)
924 925 926
        goto cleanup;

    diskPriv->migrating = true;
927
    qemuBlockJobStarted(job);
928 929 930 931

    ret = 0;

 cleanup:
932
    qemuBlockJobStartupFinalize(job);
933 934 935 936 937
    VIR_FREE(diskAlias);
    return ret;
}


938
/**
939
 * qemuMigrationSrcNBDStorageCopy:
940 941 942 943
 * @driver: qemu driver
 * @vm: domain
 * @mig: migration cookie
 * @host: where are we migrating to
944
 * @speed: bandwidth limit in MiB/s
945 946
 * @migrate_flags: migrate monitor command flags
 *
947 948 949
 * Migrate non-shared storage using the NBD protocol to the server running
 * inside the qemu process on dst and wait until the copy converges.
 * On success update @migrate_flags so we don't tell 'migrate' command
950
 * to do the very same operation. On failure, the caller is
951 952
 * expected to call qemuMigrationSrcNBDCopyCancel to stop all
 * running copy operations.
953 954 955 956 957
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
958 959 960 961 962 963 964 965
qemuMigrationSrcNBDStorageCopy(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
                               qemuMigrationCookiePtr mig,
                               const char *host,
                               unsigned long speed,
                               unsigned int *migrate_flags,
                               size_t nmigrate_disks,
                               const char **migrate_disks,
966 967 968
                               virConnectPtr dconn,
                               const char *tlsAlias,
                               unsigned int flags)
969 970 971 972
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int port;
973
    size_t i;
974
    unsigned long long mirror_speed = speed;
975
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
976
    int rv;
977
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
978 979

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
980

981 982 983 984 985 986 987 988
    if (mirror_speed > LLONG_MAX >> 20) {
        virReportError(VIR_ERR_OVERFLOW,
                       _("bandwidth must be less than %llu"),
                       LLONG_MAX >> 20);
        goto cleanup;
    }
    mirror_speed <<= 20;

989 990 991 992 993 994 995 996 997 998
    /* steal NBD port and thus prevent its propagation back to destination */
    port = mig->nbd->port;
    mig->nbd->port = 0;

    if (*migrate_flags & QEMU_MONITOR_MIGRATE_NON_SHARED_INC)
        mirror_flags |= VIR_DOMAIN_BLOCK_REBASE_SHALLOW;

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];

999
        /* check whether disk should be migrated */
1000
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1001 1002
            continue;

1003 1004 1005
        if (qemuMigrationSrcNBDStorageCopyOne(driver, vm, disk, host, port,
                                              mirror_speed, mirror_flags,
                                              tlsAlias, flags) < 0)
1006 1007
            goto cleanup;

1008
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1009 1010 1011
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto cleanup;
        }
1012
    }
1013

1014
    while ((rv = qemuMigrationSrcNBDStorageCopyReady(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) != 1) {
1015 1016
        if (rv < 0)
            goto cleanup;
1017

1018
        if (priv->job.abortJob) {
1019
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1020 1021 1022 1023
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
            goto cleanup;
1024
        }
1025

1026 1027 1028 1029 1030 1031
        if (dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            goto cleanup;
        }

1032
        if (virDomainObjWait(vm) < 0)
1033
            goto cleanup;
1034 1035
    }

1036 1037
    qemuMigrationSrcFetchMirrorStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                     priv->job.current);
1038

1039
    /* Okay, all disks are ready. Modify migrate_flags */
1040 1041 1042 1043
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

1044
 cleanup:
1045
    virObjectUnref(cfg);
1046 1047
    return ret;
}
1048

1049

1050
/**
1051
 * qemuMigrationSrcIsAllowedHostdev:
1052 1053
 * @def: domain definition
 *
1054
 * Checks that @def does not contain any host devices unsupported across
1055 1056 1057
 * migrations. Returns true if the vm is allowed to migrate.
 */
static bool
1058
qemuMigrationSrcIsAllowedHostdev(const virDomainDef *def)
1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077
{
    size_t i;

    /* Migration with USB host devices is allowed, all other devices are
     * forbidden. */
    for (i = 0; i < def->nhostdevs; i++) {
        virDomainHostdevDefPtr hostdev = def->hostdevs[i];
        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS ||
            hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("domain has assigned non-USB host devices"));
            return false;
        }
    }

    return true;
}


1078
/**
1079
 * qemuMigrationSrcIsAllowed:
1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
 * @driver: qemu driver struct
 * @vm: domain object
 * @remote: migration is remote
 * @flags: migration flags (see struct virDomainMigrateFlags)
 *
 * Validates that the configuration of @vm can be migrated in various
 * situations. If @remote is true, the migration happens to remote host. @flags
 * is used to check various special migration types according to the request.
 *
 * Returns true if migration is supported. Reports libvirt error and returns
 * false otherwise.
 */
1092
bool
1093 1094 1095 1096
qemuMigrationSrcIsAllowed(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          bool remote,
                          unsigned int flags)
1097
{
1098
    int nsnapshots;
1099
    int pauseReason;
1100
    size_t i;
1101

1102 1103 1104 1105 1106
    /* perform these checks only when migrating to remote hosts */
    if (remote) {
        nsnapshots = virDomainSnapshotObjListNum(vm->snapshots, NULL, 0);
        if (nsnapshots < 0)
            return false;
1107

1108 1109 1110 1111 1112
        if (nsnapshots > 0) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("cannot migrate domain with %d snapshots"),
                           nsnapshots);
            return false;
1113
        }
1114

1115
        /* cancel migration if disk I/O error is emitted while migrating */
1116
        if (flags & VIR_MIGRATE_ABORT_ON_ERROR &&
1117
            !(flags & VIR_MIGRATE_OFFLINE) &&
1118 1119
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
E
Eric Blake 已提交
1120
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1121
                           _("cannot migrate domain with I/O error"));
E
Eric Blake 已提交
1122 1123
            return false;
        }
1124 1125

    }
1126

1127 1128 1129 1130 1131 1132 1133
    /* following checks don't make sense for offline migration */
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        if (qemuProcessAutoDestroyActive(driver, vm)) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("domain is marked for auto destroy"));
            return false;
        }
1134

1135

1136 1137
        if (qemuDomainHasBlockjob(vm, false)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1138
                           _("domain has active block job"));
1139 1140 1141
            return false;
        }

1142
        if (!qemuMigrationSrcIsAllowedHostdev(vm->def))
1143 1144
            return false;

1145
        if (vm->def->cpu) {
1146 1147 1148
            /* QEMU blocks migration and save with invariant TSC enabled
             * unless TSC frequency is explicitly set.
             */
1149 1150
            if (virCPUCheckFeature(vm->def->os.arch, vm->def->cpu,
                                   "invtsc") == 1) {
1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168
                bool block = true;

                for (i = 0; i < vm->def->clock.ntimers; i++) {
                    virDomainTimerDefPtr timer = vm->def->clock.timers[i];

                    if (timer->name == VIR_DOMAIN_TIMER_NAME_TSC &&
                        timer->frequency > 0) {
                        block = false;
                        break;
                    }
                }

                if (block) {
                    virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                                   _("domain has 'invtsc' CPU feature but "
                                     "TSC frequency is not specified"));
                    return false;
                }
1169
            }
J
Ján Tomko 已提交
1170 1171
        }

1172 1173 1174
        /* Verify that memory device config can be transferred reliably */
        for (i = 0; i < vm->def->nmems; i++) {
            virDomainMemoryDefPtr mem = vm->def->mems[i];
1175

1176 1177 1178 1179 1180
            if (mem->model == VIR_DOMAIN_MEMORY_MODEL_DIMM &&
                mem->info.type != VIR_DOMAIN_DEVICE_ADDRESS_TYPE_DIMM) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("domain's dimm info lacks slot ID "
                                 "or base address"));
1181

1182 1183
                return false;
            }
1184
        }
1185 1186 1187 1188 1189 1190

        if (vm->def->nshmems) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("migration with shmem device is not supported"));
            return false;
        }
1191 1192
    }

1193 1194 1195
    return true;
}

1196
static bool
1197 1198 1199 1200
qemuMigrationSrcIsSafe(virDomainDefPtr def,
                       size_t nmigrate_disks,
                       const char **migrate_disks,
                       unsigned int flags)
1201

1202
{
1203 1204
    bool storagemigration = flags & (VIR_MIGRATE_NON_SHARED_DISK |
                                     VIR_MIGRATE_NON_SHARED_INC);
1205
    size_t i;
1206
    int rc;
1207

1208
    for (i = 0; i < def->ndisks; i++) {
1209
        virDomainDiskDefPtr disk = def->disks[i];
1210
        const char *src = virDomainDiskGetSource(disk);
1211

1212 1213
        /* Disks without any source (i.e. floppies and CD-ROMs)
         * OR readonly are safe. */
1214
        if (virStorageSourceIsEmpty(disk->src) ||
1215
            disk->src->readonly)
1216
            continue;
1217

1218
        /* Disks which are migrated by qemu are safe too. */
1219
        if (storagemigration &&
1220
            qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1221
            continue;
1222

1223
        /* However, disks on local FS (e.g. ext4) are not safe. */
1224
        if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
1225
            if ((rc = virFileIsSharedFS(src)) < 0) {
1226
                return false;
1227 1228 1229 1230 1231
            } else if (rc == 0) {
                virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                               _("Migration without shared storage is unsafe"));
                return false;
            }
1232 1233 1234 1235 1236 1237 1238
            if ((rc = virStorageFileIsClusterFS(src)) < 0)
                return false;
            else if (rc == 1)
                continue;
        } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                   disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
            continue;
1239
        }
1240

1241 1242 1243 1244 1245 1246 1247
        /* Our code elsewhere guarantees shared disks are either readonly (in
         * which case cache mode doesn't matter) or used with cache=none or used with cache=directsync */
        if (disk->src->shared ||
            disk->cachemode == VIR_DOMAIN_DISK_CACHE_DISABLE ||
            disk->cachemode == VIR_DOMAIN_DISK_CACHE_DIRECTSYNC)
            continue;

1248 1249
        virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                       _("Migration may lead to data corruption if disks"
1250
                         " use cache != none or cache != directsync"));
1251
        return false;
1252 1253 1254 1255 1256
    }

    return true;
}

1257
/** qemuMigrationSrcSetOffline
1258 1259 1260
 * Pause domain for non-live migration.
 */
int
1261 1262
qemuMigrationSrcSetOffline(virQEMUDriverPtr driver,
                           virDomainObjPtr vm)
1263 1264
{
    int ret;
1265
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
1266 1267
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
1268
    if (ret == 0) {
1269
        virObjectEventPtr event;
1270

1271
        event = virDomainEventLifecycleNewFromObj(vm,
1272 1273
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
1274
        virObjectEventStateQueue(driver->domainEventState, event);
1275 1276 1277 1278 1279
    }

    return ret;
}

1280 1281

void
1282 1283
qemuMigrationAnyPostcopyFailed(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313
{
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    if (state != VIR_DOMAIN_PAUSED &&
        state != VIR_DOMAIN_RUNNING)
        return;

    if (state == VIR_DOMAIN_PAUSED &&
        reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED)
        return;

    VIR_WARN("Migration of domain %s failed during post-copy; "
             "leaving the domain paused", vm->def->name);

    if (state == VIR_DOMAIN_RUNNING) {
        virObjectEventPtr event;

        if (qemuProcessStopCPUs(driver, vm,
                                VIR_DOMAIN_PAUSED_POSTCOPY_FAILED,
                                QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
            VIR_WARN("Unable to pause guest CPUs for %s", vm->def->name);
            return;
        }

        event = virDomainEventLifecycleNewFromObj(vm,
                                VIR_DOMAIN_EVENT_SUSPENDED,
                                VIR_DOMAIN_EVENT_SUSPENDED_POSTCOPY_FAILED);
1314
        virObjectEventStateQueue(driver->domainEventState, event);
1315 1316 1317 1318 1319 1320 1321
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
    }
}


1322
static int
1323
qemuMigrationSrcWaitForSpice(virDomainObjPtr vm)
1324 1325 1326
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

1327 1328
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION) ||
        !priv->job.spiceMigration)
1329 1330
        return 0;

1331
    VIR_DEBUG("Waiting for SPICE to finish migration");
1332 1333
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
1334 1335 1336 1337
            return -1;
    }
    return 0;
}
1338

1339 1340 1341 1342

static void
qemuMigrationUpdateJobType(qemuDomainJobInfoPtr jobInfo)
{
1343
    switch ((qemuMonitorMigrationStatus) jobInfo->stats.mig.status) {
1344 1345 1346 1347
    case QEMU_MONITOR_MIGRATION_STATUS_POSTCOPY:
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_POSTCOPY;
        break;

1348
    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
1349
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED;
1350 1351 1352
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
1353
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_NONE;
1354 1355 1356
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
1357
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1358 1359 1360
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
1361
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1362 1363
        break;

1364 1365 1366 1367 1368 1369 1370 1371
    case QEMU_MONITOR_MIGRATION_STATUS_PRE_SWITCHOVER:
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_PAUSED;
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_DEVICE:
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_MIGRATING;
        break;

1372 1373 1374
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLING:
1375
    case QEMU_MONITOR_MIGRATION_STATUS_LAST:
1376 1377 1378 1379 1380 1381
        break;
    }
}


int
1382 1383 1384 1385 1386
qemuMigrationAnyFetchStats(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuDomainAsyncJob asyncJob,
                           qemuDomainJobInfoPtr jobInfo,
                           char **error)
1387 1388
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1389
    qemuMonitorMigrationStats stats;
1390 1391 1392 1393 1394
    int rv;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;

1395
    rv = qemuMonitorGetMigrationStats(priv->mon, &stats, error);
1396 1397 1398 1399

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rv < 0)
        return -1;

1400
    jobInfo->stats.mig = stats;
1401 1402

    return 0;
1403 1404 1405
}


1406 1407 1408 1409 1410 1411 1412
static const char *
qemuMigrationJobName(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
1413
        return _("migration out job");
1414 1415 1416 1417
    case QEMU_ASYNC_JOB_SAVE:
        return _("domain save job");
    case QEMU_ASYNC_JOB_DUMP:
        return _("domain core dump job");
1418 1419 1420 1421 1422 1423 1424 1425 1426
    case QEMU_ASYNC_JOB_NONE:
        return _("undefined");
    case QEMU_ASYNC_JOB_MIGRATION_IN:
        return _("migration in job");
    case QEMU_ASYNC_JOB_SNAPSHOT:
        return _("snapshot job");
    case QEMU_ASYNC_JOB_START:
        return _("start job");
    case QEMU_ASYNC_JOB_LAST:
1427 1428 1429 1430 1431 1432
    default:
        return _("job");
    }
}


1433
static int
1434
qemuMigrationJobCheckStatus(virQEMUDriverPtr driver,
1435
                            virDomainObjPtr vm,
1436
                            qemuDomainAsyncJob asyncJob)
1437 1438 1439
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1440
    char *error = NULL;
1441
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1442
    int ret = -1;
1443

1444
    if (!events ||
1445
        jobInfo->stats.mig.status == QEMU_MONITOR_MIGRATION_STATUS_ERROR) {
1446
        if (qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, &error) < 0)
1447 1448
            return -1;
    }
1449

1450 1451
    qemuMigrationUpdateJobType(jobInfo);

1452 1453
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_NONE:
1454 1455
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("is not active"));
1456
        goto cleanup;
1457

1458
    case QEMU_DOMAIN_JOB_STATUS_FAILED:
1459
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
1460 1461 1462
                       qemuMigrationJobName(vm),
                       error ? error : _("unexpectedly failed"));
        goto cleanup;
1463

1464
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
1465 1466
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("canceled by client"));
1467
        goto cleanup;
1468

1469 1470
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
1471
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
1472
    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
1473
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1474
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1475 1476
        break;
    }
1477 1478 1479 1480 1481 1482

    ret = 0;

 cleanup:
    VIR_FREE(error);
    return ret;
1483 1484 1485
}


1486 1487
enum qemuMigrationCompletedFlags {
    QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR = (1 << 0),
1488
    /* This flag should only be set when run on src host */
1489
    QEMU_MIGRATION_COMPLETED_CHECK_STORAGE  = (1 << 1),
1490
    QEMU_MIGRATION_COMPLETED_POSTCOPY       = (1 << 2),
1491
    QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER = (1 << 3),
1492 1493
};

1494

1495 1496 1497 1498 1499 1500 1501
/**
 * Returns 1 if migration completed successfully,
 *         0 if the domain is still being migrated,
 *         -1 migration failed,
 *         -2 something else failed, we need to cancel migration.
 */
static int
1502 1503 1504 1505 1506
qemuMigrationAnyCompleted(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          qemuDomainAsyncJob asyncJob,
                          virConnectPtr dconn,
                          unsigned int flags)
1507 1508 1509 1510 1511
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    int pauseReason;

1512
    if (qemuMigrationJobCheckStatus(driver, vm, asyncJob) < 0)
1513 1514
        goto error;

1515
    /* This flag should only be set when run on src host */
1516
    if (flags & QEMU_MIGRATION_COMPLETED_CHECK_STORAGE &&
1517
        qemuMigrationSrcNBDStorageCopyReady(vm, asyncJob) < 0)
1518 1519
        goto error;

1520
    if (flags & QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR &&
1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533
        virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
        pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("failed due to I/O error"));
        goto error;
    }

    if (dconn && virConnectIsAlive(dconn) <= 0) {
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Lost connection to destination host"));
        goto error;
    }

1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
    /* Migration was paused before serializing device state, let's return to
     * the caller so that it can finish all block jobs, resume migration, and
     * wait again for the real end of the migration.
     */
    if (flags & QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER &&
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_PAUSED) {
        VIR_DEBUG("Migration paused before switchover");
        return 1;
    }

1544 1545 1546 1547 1548
    /* In case of postcopy the source considers migration completed at the
     * moment it switched from active to postcopy-active state. The destination
     * will continue waiting until the migrate state changes to completed.
     */
    if (flags & QEMU_MIGRATION_COMPLETED_POSTCOPY &&
1549
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
1550 1551 1552 1553
        VIR_DEBUG("Migration switched to post-copy");
        return 1;
    }

1554
    if (jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
1555 1556 1557 1558 1559
        return 1;
    else
        return 0;

 error:
1560 1561 1562
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1563
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1564
        /* The migration was aborted by us rather than QEMU itself. */
1565
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1566
        return -2;
1567 1568 1569

    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
        /* Something failed after QEMU already finished the migration. */
1570
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1571
        return -1;
1572 1573 1574 1575

    case QEMU_DOMAIN_JOB_STATUS_FAILED:
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
        /* QEMU aborted the migration. */
1576
        return -1;
1577 1578 1579 1580 1581 1582

    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_NONE:
        /* Impossible. */
        break;
1583
    }
1584 1585

    return -1;
1586 1587 1588
}


1589 1590 1591
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
1592
static int
1593 1594 1595 1596 1597
qemuMigrationSrcWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  virConnectPtr dconn,
                                  unsigned int flags)
1598
{
1599
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
1600
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1601
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1602
    int rv;
1603

1604 1605
    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_MIGRATING;

1606 1607
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           dconn, flags)) != 1) {
1608 1609
        if (rv < 0)
            return rv;
1610

1611 1612
        if (events) {
            if (virDomainObjWait(vm) < 0) {
1613 1614
                if (virDomainObjIsActive(vm))
                    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1615 1616 1617 1618 1619 1620 1621 1622 1623 1624
                return -2;
            }
        } else {
            /* Poll every 50ms for progress & to allow cancellation */
            struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };

            virObjectUnlock(vm);
            nanosleep(&ts, NULL);
            virObjectLock(vm);
        }
1625 1626
    }

1627
    if (events)
1628
        ignore_value(qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, NULL));
1629

1630
    qemuDomainJobInfoUpdateTime(jobInfo);
1631 1632
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
1633
    if (VIR_ALLOC(priv->job.completed) == 0) {
1634
        *priv->job.completed = *jobInfo;
1635 1636
        priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
    }
1637

1638 1639 1640 1641
    if (asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT &&
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;

1642
    return 0;
1643 1644 1645
}


1646
static int
1647 1648 1649 1650
qemuMigrationDstWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  bool postcopy)
1651 1652
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1653
    unsigned int flags = 0;
1654 1655 1656 1657 1658 1659 1660
    int rv;

    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT))
        return 0;

    VIR_DEBUG("Waiting for incoming migration to complete");

1661 1662 1663
    if (postcopy)
        flags = QEMU_MIGRATION_COMPLETED_POSTCOPY;

1664 1665
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           NULL, flags)) != 1) {
1666 1667 1668 1669 1670 1671 1672 1673
        if (rv < 0 || virDomainObjWait(vm) < 0)
            return -1;
    }

    return 0;
}


1674
static int
1675 1676 1677 1678
qemuMigrationSrcGraphicsRelocate(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuMigrationCookiePtr cookie,
                                 const char *graphicsuri)
1679 1680
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1681 1682
    int ret = -1;
    const char *listenAddress = NULL;
1683
    virSocketAddr addr;
1684 1685 1686 1687 1688
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
1689

1690
    if (!cookie || (!cookie->graphics && !graphicsuri))
1691 1692
        return 0;

1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711
    if (graphicsuri && !(uri = virURIParse(graphicsuri)))
        goto cleanup;

    if (cookie->graphics) {
        type = cookie->graphics->type;

        listenAddress = cookie->graphics->listen;

        if (!listenAddress ||
            (virSocketAddrParse(&addr, listenAddress, AF_UNSPEC) > 0 &&
             virSocketAddrIsWildcard(&addr)))
            listenAddress = cookie->remoteHostname;

        port = cookie->graphics->port;
        tlsPort = cookie->graphics->tlsPort;
        tlsSubject = cookie->graphics->tlsSubject;
    }

    if (uri) {
1712
        size_t i;
1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739

        if ((type = virDomainGraphicsTypeFromString(uri->scheme)) < 0) {
            virReportError(VIR_ERR_INVALID_ARG,
                           _("unknown graphics type %s"), uri->scheme);
            goto cleanup;
        }

        if (uri->server)
            listenAddress = uri->server;
        if (uri->port > 0)
            port = uri->port;

        for (i = 0; i < uri->paramsCount; i++) {
            virURIParamPtr param = uri->params + i;

            if (STRCASEEQ(param->name, "tlsPort")) {
                if (virStrToLong_i(param->value, NULL, 10, &tlsPort) < 0) {
                    virReportError(VIR_ERR_INVALID_ARG,
                                   _("invalid tlsPort number: %s"),
                                   param->value);
                    goto cleanup;
                }
            } else if (STRCASEEQ(param->name, "tlsSubject")) {
                tlsSubject = param->value;
            }
        }
    }
1740 1741 1742 1743

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
1744 1745 1746 1747
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
1748

1749 1750 1751 1752 1753 1754 1755 1756 1757
    /* Older libvirt sends port == 0 for listen type='none' graphics. It's
     * safe to ignore such requests since relocation to unknown port does
     * not make sense in general.
     */
    if (port <= 0 && tlsPort <= 0) {
        ret = 0;
        goto cleanup;
    }

1758 1759 1760 1761
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
1762
        priv->job.spiceMigration = !ret;
1763 1764
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
1765
    }
1766

1767
 cleanup:
1768
    virURIFree(uri);
1769 1770 1771 1772
    return ret;
}


1773
static int
1774 1775 1776
qemuMigrationDstOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            qemuMigrationCookiePtr cookie)
1777
{
1778 1779
    virDomainNetDefPtr netptr;
    int ret = -1;
1780
    size_t i;
1781 1782 1783 1784 1785 1786 1787 1788

    for (i = 0; i < cookie->network->nnets; i++) {
        netptr = vm->def->nets[i];

        switch (cookie->network->net[i].vporttype) {
        case VIR_NETDEV_VPORT_PROFILE_NONE:
        case VIR_NETDEV_VPORT_PROFILE_8021QBG:
        case VIR_NETDEV_VPORT_PROFILE_8021QBH:
1789
           break;
1790
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
1791 1792
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
1793 1794 1795
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
1796 1797 1798
                goto cleanup;
            }
            break;
1799 1800 1801 1802 1803
        default:
            break;
        }
    }

1804
    ret = 0;
1805
 cleanup:
1806 1807 1808 1809
    return ret;
}


1810
int
1811 1812
qemuMigrationDstCheckProtocol(virQEMUCapsPtr qemuCaps,
                              const char *migrateFrom)
1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835
{
    if (STRPREFIX(migrateFrom, "rdma")) {
        if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_MIGRATE_RDMA)) {
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                           _("incoming RDMA migration is not supported "
                             "with this QEMU binary"));
            return -1;
        }
    } else if (!STRPREFIX(migrateFrom, "tcp") &&
               !STRPREFIX(migrateFrom, "exec") &&
               !STRPREFIX(migrateFrom, "fd") &&
               !STRPREFIX(migrateFrom, "unix") &&
               STRNEQ(migrateFrom, "stdio")) {
        virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                       _("unknown migration protocol"));
        return -1;
    }

    return 0;
}


char *
1836 1837
qemuMigrationDstGetURI(const char *migrateFrom,
                       int migrateFd)
1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849
{
    char *uri = NULL;

    if (STREQ(migrateFrom, "stdio"))
        ignore_value(virAsprintf(&uri, "fd:%d", migrateFd));
    else
        ignore_value(VIR_STRDUP(uri, migrateFrom));

    return uri;
}


1850
int
1851 1852 1853 1854
qemuMigrationDstRun(virQEMUDriverPtr driver,
                    virDomainObjPtr vm,
                    const char *uri,
                    qemuDomainAsyncJob asyncJob)
1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int rv;

    VIR_DEBUG("Setting up incoming migration with URI %s", uri);

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;

    rv = qemuMonitorMigrateIncoming(priv->mon, uri);

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rv < 0)
        goto cleanup;

    if (asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN) {
1871
        /* qemuMigrationDstWaitForCompletion is called from the Finish phase */
1872 1873 1874 1875
        ret = 0;
        goto cleanup;
    }

1876
    if (qemuMigrationDstWaitForCompletion(driver, vm, asyncJob, false) < 0)
1877 1878 1879 1880 1881 1882 1883 1884 1885
        goto cleanup;

    ret = 0;

 cleanup:
    return ret;
}


1886 1887 1888 1889 1890 1891
/* This is called for outgoing non-p2p migrations when a connection to the
 * client which initiated the migration was closed but we were waiting for it
 * to follow up with the next phase, that is, in between
 * qemuDomainMigrateBegin3 and qemuDomainMigratePerform3 or
 * qemuDomainMigratePerform3 and qemuDomainMigrateConfirm3.
 */
1892
static void
1893 1894 1895
qemuMigrationSrcCleanup(virDomainObjPtr vm,
                        virConnectPtr conn,
                        void *opaque)
1896
{
1897
    virQEMUDriverPtr driver = opaque;
1898 1899 1900 1901 1902 1903 1904 1905 1906
    qemuDomainObjPrivatePtr priv = vm->privateData;

    VIR_DEBUG("vm=%s, conn=%p, asyncJob=%s, phase=%s",
              vm->def->name, conn,
              qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
              qemuDomainAsyncJobPhaseToString(priv->job.asyncJob,
                                              priv->job.phase));

    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_OUT))
1907
        return;
1908 1909 1910 1911 1912

    VIR_DEBUG("The connection which started outgoing migration of domain %s"
              " was closed; canceling the migration",
              vm->def->name);

1913
    switch ((qemuMigrationJobPhase) priv->job.phase) {
1914 1915 1916 1917 1918 1919 1920 1921 1922
    case QEMU_MIGRATION_PHASE_BEGIN3:
        /* just forget we were about to migrate */
        qemuDomainObjDiscardAsyncJob(driver, vm);
        break;

    case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
        VIR_WARN("Migration of domain %s finished but we don't know if the"
                 " domain was successfully started on destination or not",
                 vm->def->name);
1923
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
1924
                                 priv->job.migParams, priv->job.apiFlags);
1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946
        /* clear the job and let higher levels decide what to do */
        qemuDomainObjDiscardAsyncJob(driver, vm);
        break;

    case QEMU_MIGRATION_PHASE_PERFORM3:
        /* cannot be seen without an active migration API; unreachable */
    case QEMU_MIGRATION_PHASE_CONFIRM3:
    case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
        /* all done; unreachable */
    case QEMU_MIGRATION_PHASE_PREPARE:
    case QEMU_MIGRATION_PHASE_FINISH2:
    case QEMU_MIGRATION_PHASE_FINISH3:
        /* incoming migration; unreachable */
    case QEMU_MIGRATION_PHASE_PERFORM2:
        /* single phase outgoing migration; unreachable */
    case QEMU_MIGRATION_PHASE_NONE:
    case QEMU_MIGRATION_PHASE_LAST:
        /* unreachable */
        ;
    }
}

1947

1948
/* The caller is supposed to lock the vm and start a migration job. */
1949
static char *
1950 1951 1952 1953 1954 1955 1956 1957 1958
qemuMigrationSrcBeginPhase(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           const char *xmlin,
                           const char *dname,
                           char **cookieout,
                           int *cookieoutlen,
                           size_t nmigrate_disks,
                           const char **migrate_disks,
                           unsigned long flags)
1959 1960 1961
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
1962
    virDomainDefPtr def = NULL;
1963
    qemuDomainObjPrivatePtr priv = vm->privateData;
1964
    virCapsPtr caps = NULL;
1965
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
1966

1967
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
1968
              " cookieout=%p, cookieoutlen=%p,"
1969
              " nmigrate_disks=%zu, migrate_disks=%p, flags=0x%lx",
1970
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
1971 1972
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
1973

1974 1975 1976
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

1977 1978 1979 1980 1981 1982
    /* Only set the phase if we are inside QEMU_ASYNC_JOB_MIGRATION_OUT.
     * Otherwise we will start the async job later in the perform phase losing
     * change protection.
     */
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_MIGRATION_OUT)
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_BEGIN3);
1983

1984
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
1985 1986
        goto cleanup;

1987
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
1988
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
1989 1990
        goto cleanup;

1991 1992 1993 1994 1995 1996 1997 1998 1999
    if (flags & VIR_MIGRATE_POSTCOPY &&
        (!(flags & VIR_MIGRATE_LIVE) ||
         flags & VIR_MIGRATE_PAUSED)) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("post-copy migration is not supported with non-live "
                         "or paused migration"));
        goto cleanup;
    }

2000 2001 2002 2003 2004 2005
    if (flags & VIR_MIGRATE_POSTCOPY && flags & VIR_MIGRATE_TUNNELLED) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("post-copy is not supported with tunnelled migration"));
        goto cleanup;
    }

2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048
    if (flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC)) {
        bool has_drive_mirror =  virQEMUCapsGet(priv->qemuCaps,
                                                QEMU_CAPS_DRIVE_MIRROR);

        if (nmigrate_disks) {
            if (has_drive_mirror) {
                size_t i, j;
                /* Check user requested only known disk targets. */
                for (i = 0; i < nmigrate_disks; i++) {
                    for (j = 0; j < vm->def->ndisks; j++) {
                        if (STREQ(vm->def->disks[j]->dst, migrate_disks[i]))
                            break;
                    }

                    if (j == vm->def->ndisks) {
                        virReportError(VIR_ERR_INVALID_ARG,
                                       _("disk target %s not found"),
                                       migrate_disks[i]);
                        goto cleanup;
                    }
                }

                if (flags & VIR_MIGRATE_TUNNELLED) {
                    virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                                   _("Selecting disks to migrate is not "
                                     "implemented for tunnelled migration"));
                    goto cleanup;
                }
            } else {
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                               _("qemu does not support drive-mirror command"));
                goto cleanup;
            }
        }

        if (has_drive_mirror) {
            /* TODO support NBD for TUNNELLED migration */
            if (flags & VIR_MIGRATE_TUNNELLED) {
                VIR_WARN("NBD in tunnelled migration is currently not supported");
            } else {
                cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
                priv->nbdPort = 0;
            }
2049 2050 2051
        }
    }

2052
    if (virDomainDefHasMemoryHotplug(vm->def) ||
2053
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
2054
         vm->newDef && virDomainDefHasMemoryHotplug(vm->newDef)))
2055 2056
        cookieFlags |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

2057 2058 2059 2060 2061
    if (!qemuDomainVcpuHotplugIsInOrder(vm->def) ||
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
         vm->newDef && !qemuDomainVcpuHotplugIsInOrder(vm->newDef)))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU_HOTPLUG;

2062 2063 2064
    if (priv->origCPU)
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU;

2065 2066
    cookieFlags |= QEMU_MIGRATION_COOKIE_ALLOW_REBOOT;

2067 2068 2069
    if (!(flags & VIR_MIGRATE_OFFLINE))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CAPS;

2070 2071
    if (!(mig = qemuMigrationEatCookie(driver, vm->def,
                                       priv->origname, priv, NULL, 0, 0)))
2072 2073 2074
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
2075
                                QEMU_MIGRATION_SOURCE,
2076
                                cookieout, cookieoutlen,
2077
                                cookieFlags) < 0)
2078 2079
        goto cleanup;

L
liguang 已提交
2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101
    if (flags & VIR_MIGRATE_OFFLINE) {
        if (flags & (VIR_MIGRATE_NON_SHARED_DISK |
                     VIR_MIGRATE_NON_SHARED_INC)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("offline migration cannot handle "
                             "non-shared storage"));
            goto cleanup;
        }
        if (!(flags & VIR_MIGRATE_PERSIST_DEST)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("offline migration must be specified with "
                             "the persistent flag set"));
            goto cleanup;
        }
        if (flags & VIR_MIGRATE_TUNNELLED) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("tunnelled offline migration does not "
                             "make sense"));
            goto cleanup;
        }
    }

2102
    if (xmlin) {
2103
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt, priv->qemuCaps,
2104 2105
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                            VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2106 2107
            goto cleanup;

2108
        if (!qemuDomainCheckABIStability(driver, vm, def))
2109 2110
            goto cleanup;

2111
        rv = qemuDomainDefFormatLive(driver, def, NULL, false, true);
2112
    } else {
2113 2114
        rv = qemuDomainDefFormatLive(driver, vm->def, priv->origCPU,
                                     false, true);
2115
    }
2116

2117
 cleanup:
2118
    qemuMigrationCookieFree(mig);
2119
    virObjectUnref(caps);
2120
    virDomainDefFree(def);
2121 2122 2123
    return rv;
}

2124
char *
2125 2126 2127 2128 2129 2130 2131 2132 2133
qemuMigrationSrcBegin(virConnectPtr conn,
                      virDomainObjPtr vm,
                      const char *xmlin,
                      const char *dname,
                      char **cookieout,
                      int *cookieoutlen,
                      size_t nmigrate_disks,
                      const char **migrate_disks,
                      unsigned long flags)
2134 2135 2136
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
2137
    qemuDomainAsyncJob asyncJob;
2138 2139

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
2140 2141
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
2142 2143 2144 2145 2146 2147 2148 2149
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_MIGRATION_OUT;
    } else {
        if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_NONE;
    }

2150
    qemuMigrationSrcStoreDomainState(vm);
2151

2152
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
2153 2154 2155 2156 2157 2158
        goto endjob;

    /* Check if there is any ejected media.
     * We don't want to require them on the destination.
     */
    if (!(flags & VIR_MIGRATE_OFFLINE) &&
2159
        qemuProcessRefreshDisks(driver, vm, asyncJob) < 0)
2160 2161
        goto endjob;

2162 2163 2164
    if (!(xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                           cookieout, cookieoutlen,
                                           nmigrate_disks, migrate_disks, flags)))
2165 2166 2167 2168 2169 2170 2171
        goto endjob;

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
        /* We keep the job active across API calls until the confirm() call.
         * This prevents any other APIs being invoked while migration is taking
         * place.
         */
2172
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
2173
                                 qemuMigrationSrcCleanup) < 0) {
2174
            VIR_FREE(xml);
2175
            goto endjob;
2176
        }
2177
        qemuMigrationJobContinue(vm);
2178 2179 2180 2181
    } else {
        goto endjob;
    }

2182
 cleanup:
M
Michal Privoznik 已提交
2183
    virDomainObjEndAPI(&vm);
2184 2185
    return xml;

2186
 endjob:
2187 2188 2189 2190
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2191 2192 2193
    goto cleanup;
}

2194

2195 2196
/* Prepare is the first step, and it runs on the destination host.
 */
2197

2198
static void
2199 2200
qemuMigrationDstPrepareCleanup(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
2201 2202 2203 2204 2205 2206 2207 2208 2209
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    VIR_DEBUG("driver=%p, vm=%s, job=%s, asyncJob=%s",
              driver,
              vm->def->name,
              qemuDomainJobTypeToString(priv->job.active),
              qemuDomainAsyncJobTypeToString(priv->job.asyncJob));

2210
    virPortAllocatorRelease(priv->migrationPort);
2211 2212
    priv->migrationPort = 0;

2213 2214 2215 2216 2217
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2218
static qemuProcessIncomingDefPtr
2219 2220 2221 2222 2223 2224
qemuMigrationDstPrepare(virDomainObjPtr vm,
                        bool tunnel,
                        const char *protocol,
                        const char *listenAddress,
                        unsigned short port,
                        int fd)
2225 2226
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2227
    qemuProcessIncomingDefPtr inc = NULL;
2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257
    char *migrateFrom = NULL;

    if (tunnel) {
        if (VIR_STRDUP(migrateFrom, "stdio") < 0)
            goto cleanup;
    } else {
        bool encloseAddress = false;
        bool hostIPv6Capable = false;
        struct addrinfo *info = NULL;
        struct addrinfo hints = { .ai_flags = AI_ADDRCONFIG,
                                  .ai_socktype = SOCK_STREAM };
        const char *incFormat;

        if (getaddrinfo("::", NULL, &hints, &info) == 0) {
            freeaddrinfo(info);
            hostIPv6Capable = true;
        }

        if (listenAddress) {
            if (virSocketAddrNumericFamily(listenAddress) == AF_INET6) {
                if (!hostIPv6Capable) {
                    virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                                   _("host isn't capable of IPv6"));
                    goto cleanup;
                }
                /* IPv6 address must be escaped in brackets on the cmd line */
                encloseAddress = true;
            } else {
                /* listenAddress is a hostname or IPv4 */
            }
2258
        } else if (hostIPv6Capable) {
2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281
            /* Listen on :: instead of 0.0.0.0 if QEMU understands it
             * and there is at least one IPv6 address configured
             */
            listenAddress = "::";
            encloseAddress = true;
        } else {
            listenAddress = "0.0.0.0";
        }

        /* QEMU will be started with
         *   -incoming protocol:[<IPv6 addr>]:port,
         *   -incoming protocol:<IPv4 addr>:port, or
         *   -incoming protocol:<hostname>:port
         */
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";
        if (virAsprintf(&migrateFrom, incFormat,
                        protocol, listenAddress, port) < 0)
            goto cleanup;
    }

2282 2283
    inc = qemuProcessIncomingDefNew(priv->qemuCaps, listenAddress,
                                    migrateFrom, fd, NULL);
2284

2285
 cleanup:
2286 2287
    VIR_FREE(migrateFrom);
    return inc;
2288 2289
}

2290
static int
2291
qemuMigrationDstPrepareAny(virQEMUDriverPtr driver,
2292
                           virConnectPtr dconn,
2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virDomainDefPtr *def,
                           const char *origname,
                           virStreamPtr st,
                           const char *protocol,
                           unsigned short port,
                           bool autoPort,
                           const char *listenAddress,
                           size_t nmigrate_disks,
                           const char **migrate_disks,
                           int nbdPort,
2307
                           qemuMigrationParamsPtr migParams,
2308
                           unsigned long flags)
2309 2310
{
    virDomainObjPtr vm = NULL;
2311
    virObjectEventPtr event = NULL;
2312
    virErrorPtr origErr;
2313
    int ret = -1;
2314
    int dataFD[2] = { -1, -1 };
2315
    qemuDomainObjPrivatePtr priv = NULL;
2316
    qemuMigrationCookiePtr mig = NULL;
2317
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2318
    char *xmlout = NULL;
L
liguang 已提交
2319
    unsigned int cookieFlags;
2320
    unsigned int startFlags;
2321
    virCapsPtr caps = NULL;
2322
    qemuProcessIncomingDefPtr incoming = NULL;
2323
    bool taint_hook = false;
2324 2325 2326
    bool stopProcess = false;
    bool relabel = false;
    int rv;
2327
    char *tlsAlias = NULL;
2328

2329 2330
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350
    if (flags & VIR_MIGRATE_OFFLINE) {
        if (flags & (VIR_MIGRATE_NON_SHARED_DISK |
                     VIR_MIGRATE_NON_SHARED_INC)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("offline migration cannot handle "
                             "non-shared storage"));
            goto cleanup;
        }
        if (!(flags & VIR_MIGRATE_PERSIST_DEST)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("offline migration must be specified with "
                             "the persistent flag set"));
            goto cleanup;
        }
        if (tunnel) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("tunnelled offline migration does not "
                             "make sense"));
            goto cleanup;
        }
2351 2352
        cookieFlags = 0;
    } else {
2353 2354
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS |
                      QEMU_MIGRATION_COOKIE_CAPS;
L
liguang 已提交
2355 2356
    }

2357 2358 2359 2360 2361 2362 2363 2364 2365
    if (flags & VIR_MIGRATE_POSTCOPY &&
        (!(flags & VIR_MIGRATE_LIVE) ||
         flags & VIR_MIGRATE_PAUSED)) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("post-copy migration is not supported with non-live "
                         "or paused migration"));
        goto cleanup;
    }

2366 2367 2368 2369 2370 2371
    if (flags & VIR_MIGRATE_POSTCOPY && flags & VIR_MIGRATE_TUNNELLED) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("post-copy is not supported with tunnelled migration"));
        goto cleanup;
    }

2372 2373 2374
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2375
    if (!qemuMigrationSrcIsAllowedHostdev(*def))
2376 2377
        goto cleanup;

J
Jiri Denemark 已提交
2378 2379 2380 2381 2382
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2383
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2384 2385
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2386 2387
            goto cleanup;

2388
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2389 2390 2391 2392 2393 2394 2395
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2396
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2397 2398 2399 2400 2401 2402
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2403
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt, NULL,
2404 2405
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                                 VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE);
J
Jiri Denemark 已提交
2406 2407 2408
                if (!newdef)
                    goto cleanup;

2409
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2410 2411 2412 2413
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2414 2415
                virDomainDefFree(*def);
                *def = newdef;
2416 2417 2418 2419
                /* We should taint the domain here. However, @vm and therefore
                 * privateData too are still NULL, so just notice the fact and
                 * taint it later. */
                taint_hook = true;
J
Jiri Denemark 已提交
2420 2421 2422 2423
            }
        }
    }

2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437
    /* Parse cookie earlier than adding the domain onto the
     * domain list. Parsing/validation may fail and there's no
     * point in having the domain in the list at that point. */
    if (!(mig = qemuMigrationEatCookie(driver, *def, origname, NULL,
                                       cookiein, cookieinlen,
                                       QEMU_MIGRATION_COOKIE_LOCKSTATE |
                                       QEMU_MIGRATION_COOKIE_NBD |
                                       QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG |
                                       QEMU_MIGRATION_COOKIE_CPU_HOTPLUG |
                                       QEMU_MIGRATION_COOKIE_CPU |
                                       QEMU_MIGRATION_COOKIE_ALLOW_REBOOT |
                                       QEMU_MIGRATION_COOKIE_CAPS)))
        goto cleanup;

2438
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2439
                                   driver->xmlopt,
2440 2441 2442
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2443
        goto cleanup;
2444
    *def = NULL;
2445

2446
    priv = vm->privateData;
2447 2448
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2449

2450 2451 2452 2453 2454
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2455 2456
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
2457 2458 2459 2460 2461 2462
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2463 2464 2465
    if (qemuMigrationDstPrecreateStorage(vm, mig->nbd,
                                         nmigrate_disks, migrate_disks,
                                         !!(flags & VIR_MIGRATE_NON_SHARED_INC)) < 0)
2466 2467
        goto cleanup;

2468 2469
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                              flags) < 0)
2470
        goto cleanup;
2471
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2472 2473 2474 2475

    /* Domain starts inactive, even if the domain XML had an id field. */
    vm->def->id = -1;

L
liguang 已提交
2476 2477 2478
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2479 2480
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
2481 2482
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
2483
        goto stopjob;
2484 2485
    }

2486 2487
    startFlags = VIR_QEMU_PROCESS_START_AUTODESTROY;

2488
    if (qemuProcessInit(driver, vm, mig->cpu, QEMU_ASYNC_JOB_MIGRATION_IN,
2489
                        true, startFlags) < 0)
2490
        goto stopjob;
2491
    stopProcess = true;
2492

2493 2494
    priv->allowReboot = mig->allowReboot;

2495 2496 2497
    if (!(incoming = qemuMigrationDstPrepare(vm, tunnel, protocol,
                                             listenAddress, port,
                                             dataFD[0])))
2498
        goto stopjob;
2499

2500
    if (qemuProcessPrepareDomain(driver, vm, startFlags) < 0)
2501 2502
        goto stopjob;

2503
    if (qemuProcessPrepareHost(driver, vm, startFlags) < 0)
2504 2505
        goto stopjob;

2506
    rv = qemuProcessLaunch(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2507 2508
                           incoming, NULL,
                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
2509
                           startFlags);
2510 2511 2512
    if (rv < 0) {
        if (rv == -2)
            relabel = true;
2513
        goto stopjob;
2514
    }
2515
    relabel = true;
2516

2517 2518 2519 2520
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
2521
            goto stopjob;
2522
        }
2523
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
2524 2525
    }

2526
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
2527
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
2528
        goto stopjob;
M
Michael R. Hines 已提交
2529 2530
    }

2531
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2532
                                 migParams, mig->caps->automatic) < 0)
2533 2534 2535 2536 2537 2538 2539
        goto stopjob;

    /* Migrations using TLS need to add the "tls-creds-x509" object and
     * set the migration TLS parameters */
    if (flags & VIR_MIGRATE_TLS) {
        if (qemuMigrationParamsEnableTLS(driver, vm, true,
                                         QEMU_ASYNC_JOB_MIGRATION_IN,
2540
                                         &tlsAlias, NULL,
2541 2542 2543 2544 2545 2546 2547
                                         migParams) < 0)
            goto stopjob;
    } else {
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
            goto stopjob;
    }

2548 2549
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                                 migParams) < 0)
2550 2551
        goto stopjob;

2552 2553 2554
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566
        const char *nbdTLSAlias = NULL;

        if (flags & VIR_MIGRATE_TLS) {
            if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_TLS)) {
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                               _("QEMU NBD server does not support TLS transport"));
                goto stopjob;
            }

            nbdTLSAlias = tlsAlias;
        }

2567 2568
        if (qemuMigrationDstStartNBDServer(driver, vm, incoming->address,
                                           nmigrate_disks, migrate_disks,
2569
                                           nbdPort, nbdTLSAlias) < 0) {
2570
            goto stopjob;
2571
        }
2572
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
2573 2574
    }

2575 2576 2577 2578 2579 2580 2581 2582 2583
    if (mig->lockState) {
        VIR_DEBUG("Received lockstate %s", mig->lockState);
        VIR_FREE(priv->lockState);
        priv->lockState = mig->lockState;
        mig->lockState = NULL;
    } else {
        VIR_DEBUG("Received no lockstate");
    }

2584
    if (incoming->deferredURI &&
2585 2586
        qemuMigrationDstRun(driver, vm, incoming->deferredURI,
                            QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2587 2588
        goto stopjob;

2589
    if (qemuProcessFinishStartup(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2590 2591 2592
                                 false, VIR_DOMAIN_PAUSED_MIGRATION) < 0)
        goto stopjob;

2593
 done:
2594 2595 2596
    if (qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_DESTINATION,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
2597 2598 2599 2600 2601 2602 2603
        /* We could tear down the whole guest here, but
         * cookie data is (so far) non-critical, so that
         * seems a little harsh. We'll just warn for now.
         */
        VIR_WARN("Unable to encode migration cookie");
    }

2604
    if (qemuDomainCleanupAdd(vm, qemuMigrationDstPrepareCleanup) < 0)
2605
        goto stopjob;
2606

L
liguang 已提交
2607 2608
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
2609
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
2610 2611 2612
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
2613

2614 2615 2616 2617
    /* We keep the job active across API calls until the finish() call.
     * This prevents any other APIs being invoked while incoming
     * migration is taking place.
     */
2618
    qemuMigrationJobContinue(vm);
2619

2620 2621
    if (autoPort)
        priv->migrationPort = port;
2622 2623 2624 2625 2626
    /* in this case port is not auto selected and we don't need to manage it
     * anymore after cookie is baked
     */
    if (nbdPort != 0)
        priv->nbdPort = 0;
2627
    ret = 0;
2628

2629
 cleanup:
2630
    virErrorPreserveLast(&origErr);
2631
    VIR_FREE(tlsAlias);
2632
    qemuProcessIncomingDefFree(incoming);
J
Jiri Denemark 已提交
2633
    VIR_FREE(xmlout);
2634 2635
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
2636 2637 2638
    if (ret < 0 && priv) {
        /* priv is set right after vm is added to the list of domains
         * and there is no 'goto cleanup;' in the middle of those */
2639
        VIR_FREE(priv->origname);
2640 2641 2642 2643
        /* release if port is auto selected which is not the case if
         * it is given in parameters
         */
        if (nbdPort == 0)
2644
            virPortAllocatorRelease(priv->nbdPort);
2645
        priv->nbdPort = 0;
2646
        virDomainObjRemoveTransientDef(vm);
2647
        qemuDomainRemoveInactiveJob(driver, vm);
2648
    }
M
Michal Privoznik 已提交
2649
    virDomainObjEndAPI(&vm);
2650
    virObjectEventStateQueue(driver->domainEventState, event);
2651
    qemuMigrationCookieFree(mig);
2652
    virObjectUnref(caps);
2653
    virNWFilterUnlockFilterUpdates();
2654
    virErrorRestore(&origErr);
2655
    return ret;
2656

2657
 stopjob:
2658
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2659
                             priv->job.migParams, priv->job.apiFlags);
2660

2661 2662 2663 2664
    if (stopProcess) {
        unsigned int stopFlags = VIR_QEMU_PROCESS_STOP_MIGRATED;
        if (!relabel)
            stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
2665
        virDomainAuditStart(vm, "migrated", false);
2666 2667
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        QEMU_ASYNC_JOB_MIGRATION_IN, stopFlags);
2668
    }
2669

2670
    qemuMigrationJobFinish(driver, vm);
2671
    goto cleanup;
2672 2673 2674
}


2675 2676 2677 2678 2679
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
2680
qemuMigrationDstPrepareTunnel(virQEMUDriverPtr driver,
2681
                              virConnectPtr dconn,
2682 2683 2684 2685 2686 2687 2688
                              const char *cookiein,
                              int cookieinlen,
                              char **cookieout,
                              int *cookieoutlen,
                              virStreamPtr st,
                              virDomainDefPtr *def,
                              const char *origname,
2689
                              qemuMigrationParamsPtr migParams,
2690
                              unsigned long flags)
2691
{
2692
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2693
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
2694
              "origname=%s, flags=0x%lx",
2695
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2696
              cookieout, cookieoutlen, st, *def, origname, flags);
2697

2698 2699 2700 2701 2702 2703
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

2704 2705 2706
    return qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
                                      cookieout, cookieoutlen, def, origname,
                                      st, NULL, 0, false, NULL, 0, NULL, 0,
2707
                                      migParams, flags);
2708 2709 2710
}


2711
static virURIPtr
2712
qemuMigrationAnyParseURI(const char *uri, bool *wellFormed)
2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733
{
    char *tmp = NULL;
    virURIPtr parsed;

    /* For compatibility reasons tcp://... URIs are sent as tcp:...
     * We need to transform them to a well-formed URI before parsing. */
    if (STRPREFIX(uri, "tcp:") && !STRPREFIX(uri + 4, "//")) {
        if (virAsprintf(&tmp, "tcp://%s", uri + 4) < 0)
            return NULL;
        uri = tmp;
    }

    parsed = virURIParse(uri);
    if (parsed && wellFormed)
        *wellFormed = !tmp;
    VIR_FREE(tmp);

    return parsed;
}


2734
int
2735
qemuMigrationDstPrepareDirect(virQEMUDriverPtr driver,
2736
                              virConnectPtr dconn,
2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748
                              const char *cookiein,
                              int cookieinlen,
                              char **cookieout,
                              int *cookieoutlen,
                              const char *uri_in,
                              char **uri_out,
                              virDomainDefPtr *def,
                              const char *origname,
                              const char *listenAddress,
                              size_t nmigrate_disks,
                              const char **migrate_disks,
                              int nbdPort,
2749
                              qemuMigrationParamsPtr migParams,
2750
                              unsigned long flags)
2751
{
2752 2753
    unsigned short port = 0;
    bool autoPort = true;
2754 2755
    char *hostname = NULL;
    int ret = -1;
2756
    virURIPtr uri = NULL;
2757 2758
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
2759

2760
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2761
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
2762
              "def=%p, origname=%s, listenAddress=%s, "
2763
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, flags=0x%lx",
2764
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2765
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
2766
              *def, origname, NULLSTR(listenAddress),
2767
              nmigrate_disks, migrate_disks, nbdPort, flags);
2768

2769 2770
    *uri_out = NULL;

2771 2772 2773
    /* The URI passed in may be NULL or a string "tcp://somehostname:port".
     *
     * If the URI passed in is NULL then we allocate a port number
2774 2775 2776
     * from our pool of port numbers, and if the migrateHost is configured,
     * we return a URI of "tcp://migrateHost:port", otherwise return a URI
     * of "tcp://ourhostname:port".
2777 2778 2779 2780 2781 2782
     *
     * If the URI passed in is not NULL then we try to parse out the
     * port number and use that (note that the hostname is assumed
     * to be a correct hostname which refers to the target machine).
     */
    if (uri_in == NULL) {
2783 2784 2785
        bool encloseAddress = false;
        const char *incFormat;

2786
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2787
            goto cleanup;
2788

2789
        if (migrateHost != NULL) {
2790 2791
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
2792

2793
            if (VIR_STRDUP(hostname, migrateHost) < 0)
2794 2795 2796 2797 2798
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
2799 2800

        if (STRPREFIX(hostname, "localhost")) {
2801 2802 2803
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
2804 2805 2806 2807 2808
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
2809
         * compatibility with old targets. We at least make the
2810 2811
         * new targets accept both syntaxes though.
         */
2812 2813 2814 2815 2816 2817
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
2818 2819
            goto cleanup;
    } else {
2820
        bool well_formed_uri;
J
Ján Tomko 已提交
2821

2822
        if (!(uri = qemuMigrationAnyParseURI(uri_in, &well_formed_uri)))
2823
            goto cleanup;
J
Ján Tomko 已提交
2824

2825 2826 2827 2828 2829 2830 2831
        if (uri->scheme == NULL) {
            virReportError(VIR_ERR_INVALID_ARG,
                           _("missing scheme in migration URI: %s"),
                           uri_in);
            goto cleanup;
        }

M
Michael R. Hines 已提交
2832 2833
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
2834 2835 2836
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
2837 2838 2839 2840 2841 2842 2843 2844 2845 2846
            goto cleanup;
        }

        if (uri->server == NULL) {
            virReportError(VIR_ERR_INVALID_ARG, _("missing host in migration"
                                                  " URI: %s"), uri_in);
            goto cleanup;
        }

        if (uri->port == 0) {
2847
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2848
                goto cleanup;
2849

2850
            /* Send well-formed URI only if uri_in was well-formed */
2851 2852 2853 2854 2855 2856 2857 2858
            if (well_formed_uri) {
                uri->port = port;
                if (!(*uri_out = virURIFormat(uri)))
                    goto cleanup;
            } else {
                if (virAsprintf(uri_out, "%s:%d", uri_in, port) < 0)
                    goto cleanup;
            }
2859
        } else {
2860 2861
            port = uri->port;
            autoPort = false;
2862 2863 2864 2865 2866 2867
        }
    }

    if (*uri_out)
        VIR_DEBUG("Generated uri_out=%s", *uri_out);

2868
    ret = qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
2869 2870 2871 2872
                                     cookieout, cookieoutlen, def, origname,
                                     NULL, uri ? uri->scheme : "tcp",
                                     port, autoPort, listenAddress,
                                     nmigrate_disks, migrate_disks, nbdPort,
2873
                                     migParams, flags);
2874
 cleanup:
2875
    virURIFree(uri);
2876
    VIR_FREE(hostname);
2877
    virObjectUnref(cfg);
2878
    if (ret != 0) {
2879
        VIR_FREE(*uri_out);
2880
        if (autoPort)
2881
            virPortAllocatorRelease(port);
2882
    }
2883 2884 2885 2886
    return ret;
}


2887
virDomainDefPtr
2888 2889 2890 2891
qemuMigrationAnyPrepareDef(virQEMUDriverPtr driver,
                           const char *dom_xml,
                           const char *dname,
                           char **origname)
2892 2893 2894
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
2895
    char *name = NULL;
2896 2897 2898 2899 2900 2901 2902 2903 2904 2905

    if (!dom_xml) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("no domain XML passed"));
        return NULL;
    }

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        return NULL;

2906
    if (!(def = virDomainDefParseString(dom_xml, caps, driver->xmlopt, NULL,
2907 2908
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                        VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2909 2910 2911
        goto cleanup;

    if (dname) {
2912
        name = def->name;
2913 2914 2915 2916 2917 2918
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

2919
 cleanup:
2920
    virObjectUnref(caps);
2921 2922 2923 2924
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
2925 2926 2927 2928
    return def;
}


2929
static int
2930 2931 2932 2933 2934 2935
qemuMigrationSrcConfirmPhase(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
                             const char *cookiein,
                             int cookieinlen,
                             unsigned int flags,
                             int retcode)
2936 2937
{
    qemuMigrationCookiePtr mig;
2938
    virObjectEventPtr event;
2939 2940
    int rv = -1;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2941 2942
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = NULL;
2943

2944
    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
2945
              "flags=0x%x, retcode=%d",
2946
              driver, vm, NULLSTR(cookiein), cookieinlen,
2947 2948 2949 2950 2951 2952 2953 2954 2955
              flags, retcode);

    virCheckFlags(QEMU_MIGRATION_FLAGS, -1);

    qemuMigrationJobSetPhase(driver, vm,
                             retcode == 0
                             ? QEMU_MIGRATION_PHASE_CONFIRM3
                             : QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED);

2956 2957
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen,
2958
                                       QEMU_MIGRATION_COOKIE_STATS)))
2959 2960
        goto cleanup;

2961 2962 2963
    if (retcode == 0)
        jobInfo = priv->job.completed;
    else
2964
        VIR_FREE(priv->job.completed);
2965 2966 2967

    /* Update times with the values sent by the destination daemon */
    if (mig->jobInfo && jobInfo) {
2968 2969 2970 2971 2972 2973 2974 2975
        int reason;

        /* We need to refresh migration statistics after a completed post-copy
         * migration since priv->job.completed contains obsolete data from the
         * time we switched to post-copy mode.
         */
        if (virDomainObjGetState(vm, &reason) == VIR_DOMAIN_PAUSED &&
            reason == VIR_DOMAIN_PAUSED_POSTCOPY &&
2976 2977
            qemuMigrationAnyFetchStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                       jobInfo, NULL) < 0)
2978 2979
            VIR_WARN("Could not refresh migration statistics");

2980 2981 2982
        qemuDomainJobInfoUpdateTime(jobInfo);
        jobInfo->timeDeltaSet = mig->jobInfo->timeDeltaSet;
        jobInfo->timeDelta = mig->jobInfo->timeDelta;
2983 2984
        jobInfo->stats.mig.downtime_set = mig->jobInfo->stats.mig.downtime_set;
        jobInfo->stats.mig.downtime = mig->jobInfo->stats.mig.downtime;
2985 2986
    }

2987 2988 2989
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2990 2991
    /* Did the migration go as planned?  If yes, kill off the domain object.
     * If something failed, resume CPUs, but only if we didn't use post-copy.
2992 2993 2994 2995
     */
    if (retcode == 0) {
        /* If guest uses SPICE and supports seamless migration we have to hold
         * up domain shutdown until SPICE server transfers its data */
2996
        qemuMigrationSrcWaitForSpice(vm);
2997 2998

        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
2999
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
3000 3001 3002
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
        virDomainAuditStop(vm, "migrated");

3003
        event = virDomainEventLifecycleNewFromObj(vm,
3004 3005
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
3006
        virObjectEventStateQueue(driver->domainEventState, event);
3007
        qemuDomainEventEmitJobCompleted(driver, vm);
3008
    } else {
3009
        virErrorPtr orig_err = virSaveLastError();
3010
        int reason;
3011 3012

        /* cancel any outstanding NBD jobs */
3013 3014
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT, NULL);
3015 3016 3017

        virSetError(orig_err);
        virFreeError(orig_err);
3018

3019
        if (virDomainObjGetState(vm, &reason) == VIR_DOMAIN_PAUSED &&
3020
            reason == VIR_DOMAIN_PAUSED_POSTCOPY)
3021
            qemuMigrationAnyPostcopyFailed(driver, vm);
3022 3023
        else
            qemuMigrationSrcRestoreDomainState(driver, vm);
3024

3025
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3026
                                 priv->job.migParams, priv->job.apiFlags);
3027

3028
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
3029 3030 3031
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
    }

3032
 done:
3033 3034 3035
    qemuMigrationCookieFree(mig);
    rv = 0;

3036
 cleanup:
3037 3038 3039 3040 3041
    virObjectUnref(cfg);
    return rv;
}

int
3042 3043 3044 3045 3046 3047
qemuMigrationSrcConfirm(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *cookiein,
                        int cookieinlen,
                        unsigned int flags,
                        int cancelled)
3048
{
3049
    qemuMigrationJobPhase phase;
3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063
    virQEMUDriverConfigPtr cfg = NULL;
    int ret = -1;

    cfg = virQEMUDriverGetConfig(driver);

    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_OUT))
        goto cleanup;

    if (cancelled)
        phase = QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED;
    else
        phase = QEMU_MIGRATION_PHASE_CONFIRM3;

    qemuMigrationJobStartPhase(driver, vm, phase);
3064
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
3065
                           qemuMigrationSrcCleanup);
3066

3067 3068 3069
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
3070

3071
    qemuMigrationJobFinish(driver, vm);
3072
    if (!virDomainObjIsActive(vm)) {
3073
        if (!cancelled && ret == 0 && flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
3074
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
3075 3076
            vm->persistent = 0;
        }
3077
        qemuDomainRemoveInactiveJob(driver, vm);
3078 3079
    }

3080
 cleanup:
M
Michal Privoznik 已提交
3081
    virDomainObjEndAPI(&vm);
3082 3083 3084 3085 3086
    virObjectUnref(cfg);
    return ret;
}


3087 3088
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3089
    MIGRATION_DEST_CONNECT_HOST,
3090
    MIGRATION_DEST_FD,
3091
};
3092

3093 3094 3095 3096
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3097

3098 3099 3100 3101 3102 3103
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3104
            const char *protocol;
3105 3106 3107 3108
            const char *name;
            int port;
        } host;

3109 3110 3111 3112
        struct {
            int qemu;
            int local;
        } fd;
3113 3114 3115 3116 3117 3118 3119
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3120 3121 3122

#define TUNNEL_SEND_BUF_SIZE 65536

3123 3124 3125 3126 3127 3128 3129
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3130 3131
    int wakeupRecvFD;
    int wakeupSendFD;
3132 3133
};

3134
static void qemuMigrationSrcIOFunc(void *arg)
3135
{
3136
    qemuMigrationIOThreadPtr data = arg;
3137 3138 3139 3140 3141 3142 3143
    char *buffer = NULL;
    struct pollfd fds[2];
    int timeout = -1;
    virErrorPtr err = NULL;

    VIR_DEBUG("Running migration tunnel; stream=%p, sock=%d",
              data->st, data->sock);
3144

3145
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3146
        goto abrt;
3147

3148 3149 3150
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3151
    for (;;) {
3152 3153 3154 3155 3156 3157 3158 3159 3160 3161
        int ret;

        fds[0].events = fds[1].events = POLLIN;
        fds[0].revents = fds[1].revents = 0;

        ret = poll(fds, ARRAY_CARDINALITY(fds), timeout);

        if (ret < 0) {
            if (errno == EAGAIN || errno == EINTR)
                continue;
3162
            virReportSystemError(errno, "%s",
3163 3164
                                 _("poll failed in migration tunnel"));
            goto abrt;
3165
        }
3166 3167 3168 3169 3170 3171 3172

        if (ret == 0) {
            /* We were asked to gracefully stop but reading would block. This
             * can only happen if qemu told us migration finished but didn't
             * close the migration fd. We handle this in the same way as EOF.
             */
            VIR_DEBUG("QEMU forgot to close migration fd");
3173
            break;
3174
        }
3175

3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191
        if (fds[1].revents & (POLLIN | POLLERR | POLLHUP)) {
            char stop = 0;

            if (saferead(data->wakeupRecvFD, &stop, 1) != 1) {
                virReportSystemError(errno, "%s",
                                     _("failed to read from wakeup fd"));
                goto abrt;
            }

            VIR_DEBUG("Migration tunnel was asked to %s",
                      stop ? "abort" : "finish");
            if (stop) {
                goto abrt;
            } else {
                timeout = 0;
            }
3192 3193
        }

3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210
        if (fds[0].revents & (POLLIN | POLLERR | POLLHUP)) {
            int nbytes;

            nbytes = saferead(data->sock, buffer, TUNNEL_SEND_BUF_SIZE);
            if (nbytes > 0) {
                if (virStreamSend(data->st, buffer, nbytes) < 0)
                    goto error;
            } else if (nbytes < 0) {
                virReportSystemError(errno, "%s",
                        _("tunnelled migration failed to read from qemu"));
                goto abrt;
            } else {
                /* EOF; get out of here */
                break;
            }
        }
    }
3211

3212 3213
    if (virStreamFinish(data->st) < 0)
        goto error;
3214

3215
    VIR_FORCE_CLOSE(data->sock);
3216 3217
    VIR_FREE(buffer);

3218 3219
    return;

3220
 abrt:
3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3232
 error:
3233 3234 3235 3236 3237
    /* Let the source qemu know that the transfer cant continue anymore.
     * Don't copy the error for EPIPE as destination has the actual error. */
    VIR_FORCE_CLOSE(data->sock);
    if (!virLastErrorIsSystemErrno(EPIPE))
        virCopyLastError(&data->err);
3238
    virResetLastError();
3239
    VIR_FREE(buffer);
3240 3241 3242 3243
}


static qemuMigrationIOThreadPtr
3244 3245
qemuMigrationSrcStartTunnel(virStreamPtr st,
                            int sock)
3246
{
3247 3248
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3249

3250 3251 3252 3253
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3254 3255
    }

3256
    if (VIR_ALLOC(io) < 0)
3257
        goto error;
3258

3259 3260
    io->st = st;
    io->sock = sock;
3261 3262
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3263 3264

    if (virThreadCreate(&io->thread, true,
3265
                        qemuMigrationSrcIOFunc,
3266 3267 3268
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3269
        goto error;
3270 3271 3272
    }

    return io;
3273

3274
 error:
3275 3276 3277 3278
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3279 3280 3281
}

static int
3282
qemuMigrationSrcStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3283 3284
{
    int rv = -1;
3285 3286 3287 3288 3289 3290 3291 3292 3293
    char stop = error ? 1 : 0;

    /* make sure the thread finishes its job and is joinable */
    if (safewrite(io->wakeupSendFD, &stop, 1) != 1) {
        virReportSystemError(errno, "%s",
                             _("failed to wakeup migration tunnel"));
        goto cleanup;
    }

3294 3295 3296 3297
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3298 3299 3300 3301
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3302 3303 3304 3305 3306 3307
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3308
 cleanup:
3309 3310
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3311 3312
    VIR_FREE(io);
    return rv;
3313 3314
}

3315
static int
3316 3317 3318
qemuMigrationSrcConnect(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationSpecPtr spec)
3319 3320 3321 3322 3323 3324 3325
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3326
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3327 3328 3329 3330 3331
        return -1;

    spec->destType = MIGRATION_DEST_FD;
    spec->dest.fd.qemu = -1;

3332
    if (qemuSecuritySetSocketLabel(driver->securityManager, vm->def) < 0)
3333
        goto cleanup;
3334 3335 3336
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
3337
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
3338
        virObjectUnref(sock);
3339
    }
3340
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0 ||
3341 3342 3343
        spec->dest.fd.qemu == -1)
        goto cleanup;

3344 3345 3346 3347 3348 3349 3350
    /* Migration expects a blocking FD */
    if (virSetBlocking(spec->dest.fd.qemu, true) < 0) {
        virReportSystemError(errno, _("Unable to set FD %d blocking"),
                             spec->dest.fd.qemu);
        goto cleanup;
    }

3351 3352
    ret = 0;

3353
 cleanup:
3354 3355 3356 3357 3358 3359
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3360 3361

static int
3362 3363 3364 3365
qemuMigrationSrcContinue(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMonitorMigrationStatus status,
                         qemuDomainAsyncJob asyncJob)
3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;

    ret = qemuMonitorMigrateContinue(priv->mon, status);

    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

    return ret;
}


3382
static int
3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396
qemuMigrationSrcRun(virQEMUDriverPtr driver,
                    virDomainObjPtr vm,
                    const char *persist_xml,
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
                    unsigned long flags,
                    unsigned long resource,
                    qemuMigrationSpecPtr spec,
                    virConnectPtr dconn,
                    const char *graphicsuri,
                    size_t nmigrate_disks,
                    const char **migrate_disks,
3397
                    qemuMigrationParamsPtr migParams)
3398
{
3399
    int ret = -1;
3400 3401
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3402
    qemuMigrationCookiePtr mig = NULL;
3403
    char *tlsAlias = NULL;
3404
    qemuMigrationIOThreadPtr iothread = NULL;
3405
    int fd = -1;
3406
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3407
    virErrorPtr orig_err = NULL;
3408
    unsigned int cookieFlags = 0;
3409
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3410
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
3411
    bool cancel = false;
3412
    unsigned int waitFlags;
3413
    virDomainDefPtr persistDef = NULL;
3414
    char *timestamp;
3415
    int rc;
3416 3417

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
3418
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3419 3420
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
3421 3422
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3423
              spec, spec->destType, spec->fwdType, dconn,
3424
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3425

3426 3427 3428 3429 3430 3431 3432 3433 3434 3435
    if (flags & VIR_MIGRATE_NON_SHARED_DISK) {
        migrate_flags |= QEMU_MONITOR_MIGRATE_NON_SHARED_DISK;
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
    }

    if (flags & VIR_MIGRATE_NON_SHARED_INC) {
        migrate_flags |= QEMU_MONITOR_MIGRATE_NON_SHARED_INC;
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
    }

3436 3437
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3438 3439 3440 3441
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3442 3443 3444
        return -1;
    }

3445 3446 3447
    if (events)
        priv->signalIOError = abort_on_error;

3448 3449
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
        if (persist_xml) {
3450 3451
            if (!(persistDef = qemuMigrationAnyPrepareDef(driver, persist_xml,
                                                          NULL, NULL)))
3452
                goto error;
3453 3454 3455
        } else {
            virDomainDefPtr def = vm->newDef ? vm->newDef : vm->def;
            if (!(persistDef = qemuDomainDefCopy(driver, def,
3456 3457
                                                 VIR_DOMAIN_XML_SECURE |
                                                 VIR_DOMAIN_XML_MIGRATABLE)))
3458
                goto error;
3459 3460 3461
        }
    }

3462 3463
    mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                 cookiein, cookieinlen,
3464 3465 3466
                                 cookieFlags |
                                 QEMU_MIGRATION_COOKIE_GRAPHICS |
                                 QEMU_MIGRATION_COOKIE_CAPS);
3467
    if (!mig)
3468
        goto error;
3469

3470
    if (qemuMigrationSrcGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3471 3472
        VIR_WARN("unable to provide data for graphics client relocation");

3473
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3474
                                 migParams, mig->caps->automatic) < 0)
3475 3476
        goto error;

3477
    if (flags & VIR_MIGRATE_TLS) {
3478
        const char *hostname = NULL;
3479 3480 3481 3482

        /* We need to add tls-hostname whenever QEMU itself does not
         * connect directly to the destination. */
        if (spec->destType == MIGRATION_DEST_CONNECT_HOST ||
3483 3484 3485
            spec->destType == MIGRATION_DEST_FD)
            hostname = spec->dest.host.name;

3486
        if (qemuMigrationParamsEnableTLS(driver, vm, false,
3487
                                         QEMU_ASYNC_JOB_MIGRATION_OUT,
3488
                                         &tlsAlias, hostname,
3489 3490
                                         migParams) < 0)
            goto error;
3491
    } else {
3492
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
3493
            goto error;
3494 3495
    }

3496 3497 3498 3499
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                 migParams) < 0)
        goto error;

3500 3501 3502
    if (migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                         QEMU_MONITOR_MIGRATE_NON_SHARED_INC)) {
        if (mig->nbd) {
3503 3504 3505 3506
            /* Currently libvirt does not support setting up of the NBD
             * non-shared storage migration with TLS. As we need to honour the
             * VIR_MIGRATE_TLS flag, we need to reject such migration until
             * we implement TLS for NBD. */
3507 3508
            if (flags & VIR_MIGRATE_TLS &&
                !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV_DEL)) {
3509 3510 3511 3512 3513
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                               _("NBD migration with TLS is not supported"));
                goto error;
            }

3514
            /* This will update migrate_flags on success */
3515 3516 3517 3518 3519 3520
            if (qemuMigrationSrcNBDStorageCopy(driver, vm, mig,
                                               spec->dest.host.name,
                                               migrate_speed,
                                               &migrate_flags,
                                               nmigrate_disks,
                                               migrate_disks,
3521
                                               dconn, tlsAlias, flags) < 0) {
3522
                goto error;
3523 3524 3525 3526 3527 3528 3529
            }
        } else {
            /* Destination doesn't support NBD server.
             * Fall back to previous implementation. */
            VIR_DEBUG("Destination doesn't support NBD server "
                      "Falling back to previous implementation.");
        }
3530 3531
    }

3532
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3533 3534
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
3535
        if (qemuMigrationSrcSetOffline(driver, vm) < 0)
3536
            goto error;
3537 3538
    }

3539 3540
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3541
        goto error;
3542

3543
    if (priv->job.abortJob) {
3544 3545
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
3546
         * priv->job.abortJob will not change */
3547
        priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
3548 3549 3550
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
3551
        goto exit_monitor;
3552 3553
    }

3554 3555
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
3556

3557 3558
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3559
        qemuMigrationSrcConnect(driver, vm, spec) < 0) {
3560
        goto exit_monitor;
3561
    }
3562

3563 3564 3565 3566 3567 3568
    /* log start of migration */
    if ((timestamp = virTimeStringNow()) != NULL) {
        qemuDomainLogAppendMessage(driver, vm, "%s: initiating migration\n", timestamp);
        VIR_FREE(timestamp);
    }

3569
    rc = -1;
3570 3571
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3572 3573
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
3574
            goto exit_monitor;
M
Michael R. Hines 已提交
3575
        }
3576 3577 3578 3579
        rc = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
                                      spec->dest.host.protocol,
                                      spec->dest.host.name,
                                      spec->dest.host.port);
3580 3581
        break;

3582 3583 3584 3585
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3586
    case MIGRATION_DEST_FD:
3587
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3588
            fd = spec->dest.fd.local;
3589 3590
            spec->dest.fd.local = -1;
        }
3591 3592
        rc = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                    spec->dest.fd.qemu);
3593 3594
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3595
    }
3596 3597

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
3598
        goto error;
3599 3600 3601

    /* From this point onwards we *must* call cancel to abort the
     * migration on source if anything goes wrong */
3602
    cancel = true;
3603

3604
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3605
        if (!(iothread = qemuMigrationSrcStartTunnel(spec->fwd.stream, fd)))
3606
            goto error;
3607 3608 3609 3610 3611
        /* If we've created a tunnel, then the 'fd' will be closed in the
         * qemuMigrationIOFunc as data->sock.
         */
        fd = -1;
    }
3612

3613
    waitFlags = QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;
3614 3615 3616 3617 3618 3619 3620
    if (abort_on_error)
        waitFlags |= QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR;
    if (mig->nbd)
        waitFlags |= QEMU_MIGRATION_COMPLETED_CHECK_STORAGE;
    if (flags & VIR_MIGRATE_POSTCOPY)
        waitFlags |= QEMU_MIGRATION_COMPLETED_POSTCOPY;

3621 3622 3623
    rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT,
                                           dconn, waitFlags);
3624
    if (rc == -2) {
3625
        goto error;
3626 3627 3628 3629 3630
    } else if (rc == -1) {
        /* QEMU reported failed migration, nothing to cancel anymore */
        cancel = false;
        goto error;
    }
3631

3632 3633 3634
    /* When migration completed, QEMU will have paused the CPUs for us.
     * Wait for the STOP event to be processed or explicitly stop CPUs
     * (for old QEMU which does not send events) to release the lock state.
3635
     */
3636 3637 3638 3639 3640
    if (priv->monJSON) {
        while (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            priv->signalStop = true;
            rc = virDomainObjWait(vm);
            priv->signalStop = false;
3641
            if (rc < 0)
3642
                goto error;
3643
        }
3644
    } else if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING &&
3645
               qemuMigrationSrcSetOffline(driver, vm) < 0) {
3646
        goto error;
3647
    }
3648

J
Ján Tomko 已提交
3649
    if (mig->nbd &&
3650 3651 3652
        qemuMigrationSrcNBDCopyCancel(driver, vm, true,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT,
                                      dconn) < 0)
3653
        goto error;
3654

3655 3656 3657 3658 3659
    /* When migration was paused before serializing device state we need to
     * resume it now once we finished all block jobs and wait for the real
     * end of the migration.
     */
    if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_PAUSED) {
3660 3661 3662
        if (qemuMigrationSrcContinue(driver, vm,
                                     QEMU_MONITOR_MIGRATION_STATUS_PRE_SWITCHOVER,
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3663 3664 3665 3666
            goto error;

        waitFlags ^= QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;

3667 3668 3669
        rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_OUT,
                                               dconn, waitFlags);
3670 3671 3672 3673 3674 3675 3676 3677 3678
        if (rc == -2) {
            goto error;
        } else if (rc == -1) {
            /* QEMU reported failed migration, nothing to cancel anymore */
            cancel = false;
            goto error;
        }
    }

3679 3680 3681 3682
    if (iothread) {
        qemuMigrationIOThreadPtr io;

        VIR_STEAL_PTR(io, iothread);
3683
        if (qemuMigrationSrcStopTunnel(io, false) < 0)
3684
            goto error;
3685 3686 3687
    }

    if (priv->job.completed) {
3688
        priv->job.completed->stopped = priv->job.current->stopped;
3689 3690 3691 3692 3693 3694 3695 3696 3697
        qemuDomainJobInfoUpdateTime(priv->job.completed);
        qemuDomainJobInfoUpdateDowntime(priv->job.completed);
        ignore_value(virTimeMillisNow(&priv->job.completed->sent));
    }

    cookieFlags |= QEMU_MIGRATION_COOKIE_NETWORK |
                   QEMU_MIGRATION_COOKIE_STATS;

    if (qemuMigrationCookieAddPersistent(mig, &persistDef) < 0 ||
3698 3699 3700
        qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_SOURCE,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
3701 3702
        VIR_WARN("Unable to encode migration cookie");
    }
3703

3704
    ret = 0;
3705

3706
 cleanup:
3707
    VIR_FREE(tlsAlias);
3708
    VIR_FORCE_CLOSE(fd);
3709
    virDomainDefFree(persistDef);
3710 3711
    qemuMigrationCookieFree(mig);

3712 3713 3714
    if (events)
        priv->signalIOError = false;

3715 3716 3717 3718 3719
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

3720 3721
    return ret;

3722
 error:
3723 3724
    orig_err = virSaveLastError();

3725 3726 3727 3728 3729 3730 3731 3732
    if (virDomainObjIsActive(vm)) {
        if (cancel &&
            priv->job.current->status != QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED &&
            qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
            qemuMonitorMigrateCancel(priv->mon);
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
        }
3733

3734 3735
        /* cancel any outstanding NBD jobs */
        if (mig && mig->nbd)
3736 3737 3738
            qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                          QEMU_ASYNC_JOB_MIGRATION_OUT,
                                          dconn);
3739 3740 3741 3742

        if (priv->job.current->status != QEMU_DOMAIN_JOB_STATUS_CANCELED)
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
    }
3743 3744

    if (iothread)
3745
        qemuMigrationSrcStopTunnel(iothread, true);
3746 3747 3748

    goto cleanup;

3749 3750
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
3751
    goto error;
3752 3753
}

3754
/* Perform migration using QEMU's native migrate support,
3755 3756
 * not encrypted obviously
 */
3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771
static int
qemuMigrationSrcPerformNative(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              const char *persist_xml,
                              const char *uri,
                              const char *cookiein,
                              int cookieinlen,
                              char **cookieout,
                              int *cookieoutlen,
                              unsigned long flags,
                              unsigned long resource,
                              virConnectPtr dconn,
                              const char *graphicsuri,
                              size_t nmigrate_disks,
                              const char **migrate_disks,
3772
                              qemuMigrationParamsPtr migParams)
3773
{
3774
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
3775
    virURIPtr uribits = NULL;
3776
    int ret = -1;
3777 3778 3779
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
3780
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3781
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
3782
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
3783
              cookieout, cookieoutlen, flags, resource,
3784
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3785

3786
    if (!(uribits = qemuMigrationAnyParseURI(uri, NULL)))
3787 3788
        return -1;

3789 3790 3791 3792 3793 3794 3795
    if (uribits->scheme == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("missing scheme in migration URI: %s"),
                       uri);
        goto cleanup;
    }

M
Michael R. Hines 已提交
3796 3797 3798 3799 3800
    if (STREQ(uribits->scheme, "rdma")) {
        if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_RDMA)) {
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                           _("outgoing RDMA migration is not supported "
                             "with this QEMU binary"));
3801
            goto cleanup;
M
Michael R. Hines 已提交
3802
        }
3803
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
3804 3805 3806
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
3807
            goto cleanup;
M
Michael R. Hines 已提交
3808 3809 3810
        }
    }

3811
    if (STRNEQ(uribits->scheme, "rdma"))
3812 3813
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
3814
        spec.destType = MIGRATION_DEST_HOST;
3815
    spec.dest.host.protocol = uribits->scheme;
3816 3817 3818
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
3819

3820 3821 3822
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen, cookieout,
                              cookieoutlen, flags, resource, &spec, dconn,
                              graphicsuri, nmigrate_disks, migrate_disks,
3823
                              migParams);
3824 3825 3826 3827

    if (spec.destType == MIGRATION_DEST_FD)
        VIR_FORCE_CLOSE(spec.dest.fd.qemu);

3828
 cleanup:
3829
    virURIFree(uribits);
3830 3831 3832 3833 3834

    return ret;
}


3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849
static int
qemuMigrationSrcPerformTunnel(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              virStreamPtr st,
                              const char *persist_xml,
                              const char *cookiein,
                              int cookieinlen,
                              char **cookieout,
                              int *cookieoutlen,
                              unsigned long flags,
                              unsigned long resource,
                              virConnectPtr dconn,
                              const char *graphicsuri,
                              size_t nmigrate_disks,
                              const char **migrate_disks,
3850
                              qemuMigrationParamsPtr migParams)
3851 3852 3853
{
    int ret = -1;
    qemuMigrationSpec spec;
3854
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3855
    int fds[2] = { -1, -1 };
3856 3857

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
3858
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3859
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
3860
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
3861
              cookieout, cookieoutlen, flags, resource,
3862
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3863 3864 3865 3866

    spec.fwdType = MIGRATION_FWD_STREAM;
    spec.fwd.stream = st;

3867

3868 3869 3870
    spec.destType = MIGRATION_DEST_FD;
    spec.dest.fd.qemu = -1;
    spec.dest.fd.local = -1;
3871

3872 3873 3874 3875 3876
    if (pipe2(fds, O_CLOEXEC) == 0) {
        spec.dest.fd.qemu = fds[1];
        spec.dest.fd.local = fds[0];
    }
    if (spec.dest.fd.qemu == -1 ||
3877 3878
        qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    spec.dest.fd.qemu) < 0) {
3879 3880 3881
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
        goto cleanup;
3882 3883
    }

3884 3885 3886
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen,
                              cookieout, cookieoutlen, flags, resource, &spec,
                              dconn, graphicsuri, nmigrate_disks, migrate_disks,
3887
                              migParams);
3888

3889
 cleanup:
3890 3891
    VIR_FORCE_CLOSE(spec.dest.fd.qemu);
    VIR_FORCE_CLOSE(spec.dest.fd.local);
3892

3893
    virObjectUnref(cfg);
3894 3895 3896 3897
    return ret;
}


3898 3899 3900 3901
/* This is essentially a re-impl of virDomainMigrateVersion2
 * from libvirt.c, but running in source libvirtd context,
 * instead of client app context & also adding in tunnel
 * handling */
3902 3903 3904 3905 3906 3907 3908 3909
static int
qemuMigrationSrcPerformPeer2Peer2(virQEMUDriverPtr driver,
                                  virConnectPtr sconn,
                                  virConnectPtr dconn,
                                  virDomainObjPtr vm,
                                  const char *dconnuri,
                                  unsigned long flags,
                                  const char *dname,
3910 3911
                                  unsigned long resource,
                                  qemuMigrationParamsPtr migParams)
3912 3913 3914
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
3915
    char *cookie = NULL;
3916 3917 3918
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
3919
    bool cancelled;
3920
    virStreamPtr st = NULL;
3921
    unsigned long destflags;
3922

3923
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
3924
              "flags=0x%lx, dname=%s, resource=%lu",
3925 3926
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
3927

3928 3929 3930 3931 3932
    /* In version 2 of the protocol, the prepare step is slightly
     * different.  We fetch the domain XML of the source domain
     * and pass it to Prepare2.
     */
    if (!(dom_xml = qemuDomainFormatXML(driver, vm,
3933 3934
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
3935 3936 3937 3938 3939
        return -1;

    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

3940 3941
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
3942

3943 3944 3945 3946 3947 3948 3949 3950 3951 3952
    VIR_DEBUG("Prepare2 %p", dconn);
    if (flags & VIR_MIGRATE_TUNNELLED) {
        /*
         * Tunnelled Migrate Version 2 does not support cookies
         * due to missing parameters in the prepareTunnel() API.
         */

        if (!(st = virStreamNew(dconn, 0)))
            goto cleanup;

3953
        qemuDomainObjEnterRemote(vm);
3954
        ret = dconn->driver->domainMigratePrepareTunnel
3955
            (dconn, st, destflags, dname, resource, dom_xml);
3956 3957
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3958
    } else {
3959
        qemuDomainObjEnterRemote(vm);
3960 3961
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
3962
             destflags, dname, resource, dom_xml);
3963 3964
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3965 3966 3967
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
3968 3969
        goto cleanup;

3970 3971
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
3972 3973
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
3974
        cancelled = true;
3975
        orig_err = virSaveLastError();
3976
        goto finish;
3977 3978
    }

3979 3980 3981 3982
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
3983
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
3984
    if (flags & VIR_MIGRATE_TUNNELLED)
3985 3986 3987
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, NULL,
                                            NULL, 0, NULL, NULL,
                                            flags, resource, dconn,
3988
                                            NULL, 0, NULL, migParams);
3989
    else
3990 3991 3992 3993
        ret = qemuMigrationSrcPerformNative(driver, vm, NULL, uri_out,
                                            cookie, cookielen,
                                            NULL, NULL, /* No out cookie with v2 migration */
                                            flags, resource, dconn, NULL, 0, NULL,
3994
                                            migParams);
3995 3996 3997 3998

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
3999

4000 4001 4002
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4003
    cancelled = ret < 0;
4004

4005
 finish:
4006 4007 4008 4009
    /* In version 2 of the migration protocol, we pass the
     * status code from the sender to the destination host,
     * so it can do any cleanup if the migration failed.
     */
4010
    dname = dname ? dname : vm->def->name;
4011
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4012
    qemuDomainObjEnterRemote(vm);
4013
    ddomain = dconn->driver->domainMigrateFinish2
4014
        (dconn, dname, cookie, cookielen,
4015
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4016 4017
    /* The domain is already gone at this point */
    ignore_value(qemuDomainObjExitRemote(vm, false));
4018 4019
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4020

4021
 cleanup:
4022
    if (ddomain) {
4023
        virObjectUnref(ddomain);
4024 4025 4026 4027
        ret = 0;
    } else {
        ret = -1;
    }
4028

4029
    virObjectUnref(st);
4030 4031 4032 4033 4034 4035

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4036
    VIR_FREE(cookie);
4037 4038

    return ret;
4039 4040 4041
}


4042 4043 4044 4045
/* This is essentially a re-impl of virDomainMigrateVersion3
 * from libvirt.c, but running in source libvirtd context,
 * instead of client app context & also adding in tunnel
 * handling */
4046
static int
4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060
qemuMigrationSrcPerformPeer2Peer3(virQEMUDriverPtr driver,
                                  virConnectPtr sconn,
                                  virConnectPtr dconn,
                                  const char *dconnuri,
                                  virDomainObjPtr vm,
                                  const char *xmlin,
                                  const char *persist_xml,
                                  const char *dname,
                                  const char *uri,
                                  const char *graphicsuri,
                                  const char *listenAddress,
                                  size_t nmigrate_disks,
                                  const char **migrate_disks,
                                  int nbdPort,
4061
                                  qemuMigrationParamsPtr migParams,
4062 4063 4064
                                  unsigned long long bandwidth,
                                  bool useParams,
                                  unsigned long flags)
4065 4066 4067 4068 4069 4070 4071 4072 4073 4074
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
    char *cookiein = NULL;
    char *cookieout = NULL;
    char *dom_xml = NULL;
    int cookieinlen = 0;
    int cookieoutlen = 0;
    int ret = -1;
    virErrorPtr orig_err = NULL;
4075
    bool cancelled = true;
4076
    virStreamPtr st = NULL;
4077
    unsigned long destflags;
4078 4079 4080
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4081
    size_t i;
4082
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4083 4084

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4085
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4086
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4087
              "bandwidth=%llu, useParams=%d, flags=0x%lx",
4088
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4089
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4090
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks, nbdPort,
4091
              bandwidth, useParams, flags);
4092

4093 4094 4095 4096 4097
    /* Unlike the virDomainMigrateVersion3 counterpart, we don't need
     * to worry about auto-setting the VIR_MIGRATE_CHANGE_PROTECTION
     * bit here, because we are already running inside the context of
     * a single job.  */

4098 4099 4100
    dom_xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                         &cookieout, &cookieoutlen,
                                         nmigrate_disks, migrate_disks, flags);
4101 4102 4103
    if (!dom_xml)
        goto cleanup;

4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123
    if (useParams) {
        if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_DEST_XML, dom_xml) < 0)
            goto cleanup;

        if (dname &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_DEST_NAME, dname) < 0)
            goto cleanup;

        if (uri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_URI, uri) < 0)
            goto cleanup;

        if (bandwidth &&
            virTypedParamsAddULLong(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_BANDWIDTH,
                                    bandwidth) < 0)
            goto cleanup;
4124 4125 4126 4127 4128 4129

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4130 4131 4132 4133 4134
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4135 4136 4137 4138 4139
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4140 4141 4142 4143 4144
        if (nbdPort &&
            virTypedParamsAddInt(&params, &nparams, &maxparams,
                                 VIR_MIGRATE_PARAM_DISKS_PORT,
                                 nbdPort) < 0)
            goto cleanup;
4145

4146 4147
        if (qemuMigrationParamsDump(migParams, &params, &nparams,
                                    &maxparams, &flags) < 0)
4148
            goto cleanup;
4149 4150
    }

4151 4152 4153
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4154 4155
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4156

4157 4158 4159 4160 4161 4162 4163 4164 4165
    VIR_DEBUG("Prepare3 %p", dconn);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
    if (flags & VIR_MIGRATE_TUNNELLED) {
        if (!(st = virStreamNew(dconn, 0)))
            goto cleanup;

4166
        qemuDomainObjEnterRemote(vm);
4167 4168 4169 4170 4171 4172 4173 4174 4175
        if (useParams) {
            ret = dconn->driver->domainMigratePrepareTunnel3Params
                (dconn, st, params, nparams, cookiein, cookieinlen,
                 &cookieout, &cookieoutlen, destflags);
        } else {
            ret = dconn->driver->domainMigratePrepareTunnel3
                (dconn, st, cookiein, cookieinlen, &cookieout, &cookieoutlen,
                 destflags, dname, bandwidth, dom_xml);
        }
4176 4177
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4178
    } else {
4179
        qemuDomainObjEnterRemote(vm);
4180 4181 4182 4183 4184 4185 4186 4187 4188
        if (useParams) {
            ret = dconn->driver->domainMigratePrepare3Params
                (dconn, params, nparams, cookiein, cookieinlen,
                 &cookieout, &cookieoutlen, &uri_out, destflags);
        } else {
            ret = dconn->driver->domainMigratePrepare3
                (dconn, cookiein, cookieinlen, &cookieout, &cookieoutlen,
                 uri, &uri_out, destflags, dname, bandwidth, dom_xml);
        }
4189 4190
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4191 4192 4193 4194 4195
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

4196
    if (offline) {
L
liguang 已提交
4197 4198 4199
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4200
        cancelled = false;
L
liguang 已提交
4201 4202 4203
        goto finish;
    }

4204 4205 4206 4207
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4208 4209
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4210
            goto finish;
4211
        }
4212
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4213 4214
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4215
        orig_err = virSaveLastError();
4216 4217 4218 4219 4220 4221 4222 4223
        goto finish;
    }

    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete. The src VM should remain
     * running, but in paused state until the destination can
     * confirm migration completion.
     */
4224
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4225
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4226 4227 4228 4229 4230
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4231
    if (flags & VIR_MIGRATE_TUNNELLED) {
4232 4233 4234 4235
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, persist_xml,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4236
                                            nmigrate_disks, migrate_disks,
4237
                                            migParams);
4238
    } else {
4239 4240 4241 4242
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4243
                                            nmigrate_disks, migrate_disks,
4244
                                            migParams);
4245
    }
4246 4247

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4248
    if (ret < 0) {
4249
        orig_err = virSaveLastError();
4250 4251 4252 4253
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4254 4255 4256 4257

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4258
    cancelled = ret < 0;
4259

4260
 finish:
4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272
    /*
     * The status code from the source is passed to the destination.
     * The dest can cleanup in the source indicated it failed to
     * send all migration data. Returns NULL for ddomain if
     * the dest was unable to complete migration.
     */
    VIR_DEBUG("Finish3 %p ret=%d", dconn, ret);
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285

    if (useParams) {
        if (virTypedParamsGetString(params, nparams,
                                    VIR_MIGRATE_PARAM_DEST_NAME, NULL) <= 0 &&
            virTypedParamsReplaceString(&params, &nparams,
                                        VIR_MIGRATE_PARAM_DEST_NAME,
                                        vm->def->name) < 0) {
            ddomain = NULL;
        } else {
            qemuDomainObjEnterRemote(vm);
            ddomain = dconn->driver->domainMigrateFinish3Params
                (dconn, params, nparams, cookiein, cookieinlen,
                 &cookieout, &cookieoutlen, destflags, cancelled);
4286 4287
            if (qemuDomainObjExitRemote(vm, !offline) < 0)
                goto cleanup;
4288 4289 4290 4291 4292 4293 4294
        }
    } else {
        dname = dname ? dname : vm->def->name;
        qemuDomainObjEnterRemote(vm);
        ddomain = dconn->driver->domainMigrateFinish3
            (dconn, dname, cookiein, cookieinlen, &cookieout, &cookieoutlen,
             dconnuri, uri, destflags, cancelled);
4295 4296
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4297
    }
4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317

    if (cancelled) {
        if (ddomain) {
            VIR_ERROR(_("finish step ignored that migration was cancelled"));
        } else {
            /* If Finish reported a useful error, use it instead of the
             * original "migration unexpectedly failed" error.
             *
             * This is ugly but we can't do better with the APIs we have. We
             * only replace the error if Finish was called with cancelled == 1
             * and reported a real error (old libvirt would report an error
             * from RPC instead of MIGRATE_FINISH_OK), which only happens when
             * the domain died on destination. To further reduce a possibility
             * of false positives we also check that Perform returned
             * VIR_ERR_OPERATION_FAILED.
             */
            if (orig_err &&
                orig_err->domain == VIR_FROM_QEMU &&
                orig_err->code == VIR_ERR_OPERATION_FAILED) {
                virErrorPtr err = virGetLastError();
4318 4319
                if (err &&
                    err->domain == VIR_FROM_QEMU &&
4320 4321 4322 4323 4324 4325 4326
                    err->code != VIR_ERR_MIGRATE_FINISH_OK) {
                    virFreeError(orig_err);
                    orig_err = NULL;
                }
            }
        }
    }
4327

4328 4329 4330 4331 4332 4333 4334
    /* If ddomain is NULL, then we were unable to start
     * the guest on the target, and must restart on the
     * source. There is a small chance that the ddomain
     * is NULL due to an RPC failure, in which case
     * ddomain could in fact be running on the dest.
     * The lock manager plugins should take care of
     * safety in this scenario.
4335
     */
4336
    cancelled = ddomain == NULL;
4337

4338 4339 4340 4341 4342 4343
    /* If finish3 set an error, and we don't have an earlier
     * one we need to preserve it in case confirm3 overwrites
     */
    if (!orig_err)
        orig_err = virSaveLastError();

4344 4345 4346 4347
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4348
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4349 4350 4351 4352 4353
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4354 4355 4356
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
4357 4358 4359 4360
    /* If Confirm3 returns -1, there's nothing more we can
     * do, but fortunately worst case is that there is a
     * domain left in 'paused' state on source.
     */
4361 4362 4363
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4364 4365 4366

 cleanup:
    if (ddomain) {
4367
        virObjectUnref(ddomain);
4368 4369 4370 4371 4372
        ret = 0;
    } else {
        ret = -1;
    }

4373
    virObjectUnref(st);
4374 4375 4376 4377 4378 4379 4380 4381

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4382
    virTypedParamsFree(params, nparams);
4383 4384 4385 4386
    return ret;
}


4387
static void
4388 4389 4390
qemuMigrationSrcConnectionClosed(virConnectPtr conn,
                                 int reason,
                                 void *opaque)
4391 4392 4393 4394 4395 4396 4397 4398
{
    virDomainObjPtr vm = opaque;

    VIR_DEBUG("conn=%p, reason=%d, vm=%s", conn, reason, vm->def->name);
    virDomainObjBroadcast(vm);
}


4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


static virConnectAuth virConnectAuthConfig = {
    .credtype = virConnectCredType,
    .ncredtype = ARRAY_CARDINALITY(virConnectCredType),
};


4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423
static int
qemuMigrationSrcPerformPeer2Peer(virQEMUDriverPtr driver,
                                 virConnectPtr sconn,
                                 virDomainObjPtr vm,
                                 const char *xmlin,
                                 const char *persist_xml,
                                 const char *dconnuri,
                                 const char *uri,
                                 const char *graphicsuri,
                                 const char *listenAddress,
                                 size_t nmigrate_disks,
                                 const char **migrate_disks,
                                 int nbdPort,
4424
                                 qemuMigrationParamsPtr migParams,
4425 4426 4427 4428
                                 unsigned long flags,
                                 const char *dname,
                                 unsigned long resource,
                                 bool *v3proto)
4429 4430 4431 4432
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4433
    virErrorPtr orig_err = NULL;
4434
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4435
    bool dstOffline = false;
4436
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4437
    bool useParams;
4438

4439 4440
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, uri=%s, "
              "graphicsuri=%s, listenAddress=%s, nmigrate_disks=%zu, "
4441
              "migrate_disks=%p, nbdPort=%d, flags=0x%lx, dname=%s, "
4442
              "resource=%lu",
4443
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4444
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4445 4446
              nmigrate_disks, migrate_disks, nbdPort, flags, NULLSTR(dname),
              resource);
4447

4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461
    if (flags & VIR_MIGRATE_TUNNELLED && uri) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("migration URI is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

    if (flags & VIR_MIGRATE_TUNNELLED && listenAddress) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("listen address is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

4462 4463 4464 4465 4466 4467 4468
    if (flags & VIR_MIGRATE_TUNNELLED && nbdPort) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("disk port address is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

4469 4470 4471 4472
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

4473
    qemuDomainObjEnterRemote(vm);
4474
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4475 4476 4477
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;

4478
    if (dconn == NULL) {
4479
        virReportError(VIR_ERR_OPERATION_FAILED,
4480 4481
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4482
        virObjectUnref(cfg);
4483 4484 4485
        return -1;
    }

4486 4487
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4488 4489
        goto cleanup;

4490
    if (virConnectRegisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed,
4491 4492 4493 4494
                                        vm, NULL) < 0) {
        goto cleanup;
    }

4495
    qemuDomainObjEnterRemote(vm);
4496 4497
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4498
        /* v3proto reflects whether the caller used Perform3, but with
4499
         * p2p migrate, regardless of whether Perform2 or Perform3
4500 4501 4502 4503
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4504 4505
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
4506
    if (offline)
4507 4508
        dstOffline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                              VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4509 4510
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;
4511

4512
    if (!p2p) {
4513 4514
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4515 4516 4517
        goto cleanup;
    }

4518 4519
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4520
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
4521 4522 4523 4524 4525 4526
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("Migration APIs with extensible parameters are not "
                         "supported but extended parameters were passed"));
        goto cleanup;
    }

4527
    if (offline && !dstOffline) {
L
liguang 已提交
4528 4529 4530 4531 4532 4533
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4534 4535 4536 4537 4538 4539 4540
    /* Change protection is only required on the source side (us), and
     * only for v3 migration when begin and perform are separate jobs.
     * But peer-2-peer is already a single job, and we still want to
     * talk to older destinations that would reject the flag.
     * Therefore it is safe to clear the bit here.  */
    flags &= ~VIR_MIGRATE_CHANGE_PROTECTION;

4541
    if (*v3proto) {
4542 4543 4544
        ret = qemuMigrationSrcPerformPeer2Peer3(driver, sconn, dconn, dconnuri, vm, xmlin,
                                                persist_xml, dname, uri, graphicsuri,
                                                listenAddress, nmigrate_disks, migrate_disks,
4545
                                                nbdPort, migParams, resource,
4546
                                                useParams, flags);
4547
    } else {
4548
        ret = qemuMigrationSrcPerformPeer2Peer2(driver, sconn, dconn, vm,
4549
                                                dconnuri, flags, dname, resource,
4550
                                                migParams);
4551
    }
4552

4553
 cleanup:
4554
    orig_err = virSaveLastError();
4555
    qemuDomainObjEnterRemote(vm);
4556
    virConnectUnregisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed);
4557
    virObjectUnref(dconn);
4558
    ignore_value(qemuDomainObjExitRemote(vm, false));
4559 4560 4561 4562
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4563
    virObjectUnref(cfg);
4564 4565 4566 4567
    return ret;
}


4568 4569 4570 4571 4572 4573
/*
 * This implements perform part of the migration protocol when migration job
 * does not need to be active across several APIs, i.e., peer2peer migration or
 * perform phase of v2 non-peer2peer migration.
 */
static int
4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585
qemuMigrationSrcPerformJob(virQEMUDriverPtr driver,
                           virConnectPtr conn,
                           virDomainObjPtr vm,
                           const char *xmlin,
                           const char *persist_xml,
                           const char *dconnuri,
                           const char *uri,
                           const char *graphicsuri,
                           const char *listenAddress,
                           size_t nmigrate_disks,
                           const char **migrate_disks,
                           int nbdPort,
4586
                           qemuMigrationParamsPtr migParams,
4587 4588 4589 4590 4591 4592 4593 4594
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
                           const char *dname,
                           unsigned long resource,
                           bool v3proto)
4595
{
4596
    virObjectEventPtr event = NULL;
4597
    int ret = -1;
4598
    virErrorPtr orig_err = NULL;
4599
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4600
    qemuDomainObjPrivatePtr priv = vm->privateData;
4601

4602 4603
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                              flags) < 0)
4604 4605
        goto cleanup;

4606
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
4607 4608
        goto endjob;

4609
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
4610
        goto endjob;
4611

4612
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
4613
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
4614
        goto endjob;
4615

4616
    qemuMigrationSrcStoreDomainState(vm);
4617 4618

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4619 4620 4621
        ret = qemuMigrationSrcPerformPeer2Peer(driver, conn, vm, xmlin, persist_xml,
                                               dconnuri, uri, graphicsuri, listenAddress,
                                               nmigrate_disks, migrate_disks, nbdPort,
4622
                                               migParams, flags, dname, resource,
4623
                                               &v3proto);
4624
    } else {
4625
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4626 4627 4628
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                            cookieout, cookieoutlen,
                                            flags, resource, NULL, NULL, 0, NULL,
4629
                                            migParams);
4630
    }
4631 4632
    if (ret < 0)
        goto endjob;
4633

4634 4635 4636 4637
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4638
    if (!v3proto) {
4639
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
4640
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4641
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4642
        virDomainAuditStop(vm, "migrated");
4643
        event = virDomainEventLifecycleNewFromObj(vm,
4644 4645
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4646 4647
    }

4648
 endjob:
4649 4650 4651
    if (ret < 0)
        orig_err = virSaveLastError();

4652 4653 4654 4655
    /* v2 proto has no confirm phase so we need to reset migration parameters
     * here
     */
    if (!v3proto && ret < 0)
4656
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4657
                                 priv->job.migParams, priv->job.apiFlags);
4658

4659
    qemuMigrationSrcRestoreDomainState(driver, vm);
4660

4661
    qemuMigrationJobFinish(driver, vm);
4662
    if (!virDomainObjIsActive(vm) && ret == 0) {
4663
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
4664
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4665 4666
            vm->persistent = 0;
        }
4667
        qemuDomainRemoveInactiveJob(driver, vm);
4668 4669
    }

4670 4671 4672 4673 4674
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4675
 cleanup:
M
Michal Privoznik 已提交
4676
    virDomainObjEndAPI(&vm);
4677
    virObjectEventStateQueue(driver->domainEventState, event);
4678
    virObjectUnref(cfg);
4679 4680 4681 4682 4683 4684 4685
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4686 4687 4688 4689 4690 4691 4692 4693
qemuMigrationSrcPerformPhase(virQEMUDriverPtr driver,
                             virConnectPtr conn,
                             virDomainObjPtr vm,
                             const char *persist_xml,
                             const char *uri,
                             const char *graphicsuri,
                             size_t nmigrate_disks,
                             const char **migrate_disks,
4694
                             qemuMigrationParamsPtr migParams,
4695 4696 4697 4698 4699 4700
                             const char *cookiein,
                             int cookieinlen,
                             char **cookieout,
                             int *cookieoutlen,
                             unsigned long flags,
                             unsigned long resource)
4701
{
4702
    qemuDomainObjPrivatePtr priv = vm->privateData;
4703 4704 4705 4706
    int ret = -1;

    /* If we didn't start the job in the begin phase, start it now. */
    if (!(flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
4707 4708
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
4709 4710 4711 4712 4713 4714
            goto cleanup;
    } else if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) {
        goto cleanup;
    }

    qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4715
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
4716
                           qemuMigrationSrcCleanup);
4717

4718 4719 4720
    ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                        cookieout, cookieoutlen,
                                        flags, resource, NULL, graphicsuri,
4721
                                        nmigrate_disks, migrate_disks, migParams);
4722

4723
    if (ret < 0) {
4724
        qemuMigrationSrcRestoreDomainState(driver, vm);
4725
        goto endjob;
4726
    }
4727 4728 4729

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4730
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
4731
                             qemuMigrationSrcCleanup) < 0)
4732 4733
        goto endjob;

4734
 endjob:
4735
    if (ret < 0) {
4736
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4737
                                 priv->job.migParams, priv->job.apiFlags);
4738
        qemuMigrationJobFinish(driver, vm);
4739
    } else {
4740
        qemuMigrationJobContinue(vm);
4741 4742
    }

4743
    if (!virDomainObjIsActive(vm))
4744
        qemuDomainRemoveInactiveJob(driver, vm);
4745

4746
 cleanup:
M
Michal Privoznik 已提交
4747
    virDomainObjEndAPI(&vm);
4748 4749 4750
    return ret;
}

4751
int
4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763
qemuMigrationSrcPerform(virQEMUDriverPtr driver,
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *persist_xml,
                        const char *dconnuri,
                        const char *uri,
                        const char *graphicsuri,
                        const char *listenAddress,
                        size_t nmigrate_disks,
                        const char **migrate_disks,
                        int nbdPort,
4764
                        qemuMigrationParamsPtr migParams,
4765 4766 4767 4768 4769 4770 4771 4772
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4773 4774
{
    VIR_DEBUG("driver=%p, conn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4775
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
4776
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4777
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
4778
              "flags=0x%lx, dname=%s, resource=%lu, v3proto=%d",
4779
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4780
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4781 4782 4783
              nmigrate_disks, migrate_disks, nbdPort,
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4784 4785 4786

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4787 4788
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4789 4790 4791
            return -1;
        }

4792 4793 4794
        return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, dconnuri, uri,
                                          graphicsuri, listenAddress,
                                          nmigrate_disks, migrate_disks, nbdPort,
4795
                                          migParams,
4796 4797 4798
                                          cookiein, cookieinlen,
                                          cookieout, cookieoutlen,
                                          flags, dname, resource, v3proto);
4799 4800
    } else {
        if (dconnuri) {
4801 4802
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
4803 4804 4805 4806
            return -1;
        }

        if (v3proto) {
4807 4808 4809
            return qemuMigrationSrcPerformPhase(driver, conn, vm, persist_xml, uri,
                                                graphicsuri,
                                                nmigrate_disks, migrate_disks,
4810
                                                migParams,
4811 4812 4813
                                                cookiein, cookieinlen,
                                                cookieout, cookieoutlen,
                                                flags, resource);
4814
        } else {
4815 4816 4817
            return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, NULL,
                                              uri, graphicsuri, listenAddress,
                                              nmigrate_disks, migrate_disks, nbdPort,
4818
                                              migParams,
4819 4820 4821
                                              cookiein, cookieinlen,
                                              cookieout, cookieoutlen, flags,
                                              dname, resource, v3proto);
4822 4823 4824
        }
    }
}
4825

4826
static int
4827
qemuMigrationDstVPAssociatePortProfiles(virDomainDefPtr def)
4828
{
4829
    size_t i;
4830 4831 4832 4833 4834
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
4835
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4836
            if (virNetDevVPortProfileAssociate(net->ifname,
4837
                                               virDomainNetGetActualVirtPortProfile(net),
4838
                                               &net->mac,
4839
                                               virDomainNetGetActualDirectDev(net),
4840
                                               -1,
4841
                                               def->uuid,
4842 4843
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
4844 4845 4846
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
4847
                goto err_exit;
4848
            }
4849
            last_good_net = i;
4850
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
4851

4852
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
4853 4854 4855 4856
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
4857 4858 4859
        }
    }

4860
    return 0;
4861

4862
 err_exit:
4863
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
4864
        net = def->nets[i];
4865
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4866
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
4867
                                                           virDomainNetGetActualVirtPortProfile(net),
4868
                                                           &net->mac,
4869
                                                           virDomainNetGetActualDirectDev(net),
4870
                                                           -1,
4871
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
4872 4873
        }
    }
4874
    return -1;
4875 4876 4877
}


J
Jiri Denemark 已提交
4878
static int
4879 4880 4881 4882
qemuMigrationDstPersist(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationCookiePtr mig,
                        bool ignoreSaveError)
J
Jiri Denemark 已提交
4883 4884 4885 4886
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    virDomainDefPtr vmdef;
4887 4888
    virDomainDefPtr oldDef = NULL;
    unsigned int oldPersist = vm->persistent;
J
Jiri Denemark 已提交
4889 4890 4891 4892 4893 4894 4895
    virObjectEventPtr event;
    int ret = -1;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

    vm->persistent = 1;
4896 4897
    oldDef = vm->newDef;
    vm->newDef = qemuMigrationCookieGetPersistent(mig);
J
Jiri Denemark 已提交
4898

4899 4900
    if (!(vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm)))
        goto error;
J
Jiri Denemark 已提交
4901

4902 4903
    if (virDomainSaveConfig(cfg->configDir, driver->caps, vmdef) < 0 &&
        !ignoreSaveError)
4904
        goto error;
J
Jiri Denemark 已提交
4905 4906 4907

    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_DEFINED,
4908 4909 4910
                                              oldPersist ?
                                              VIR_DOMAIN_EVENT_DEFINED_UPDATED :
                                              VIR_DOMAIN_EVENT_DEFINED_ADDED);
4911
    virObjectEventStateQueue(driver->domainEventState, event);
J
Jiri Denemark 已提交
4912 4913 4914 4915

    ret = 0;

 cleanup:
4916
    virDomainDefFree(oldDef);
J
Jiri Denemark 已提交
4917 4918 4919
    virObjectUnref(caps);
    virObjectUnref(cfg);
    return ret;
4920 4921 4922 4923 4924 4925 4926

 error:
    virDomainDefFree(vm->newDef);
    vm->persistent = oldPersist;
    vm->newDef = oldDef;
    oldDef = NULL;
    goto cleanup;
J
Jiri Denemark 已提交
4927 4928 4929
}


4930
virDomainPtr
4931 4932 4933 4934 4935 4936 4937 4938 4939 4940
qemuMigrationDstFinish(virQEMUDriverPtr driver,
                       virConnectPtr dconn,
                       virDomainObjPtr vm,
                       const char *cookiein,
                       int cookieinlen,
                       char **cookieout,
                       int *cookieoutlen,
                       unsigned long flags,
                       int retcode,
                       bool v3proto)
4941 4942
{
    virDomainPtr dom = NULL;
4943
    qemuMigrationCookiePtr mig = NULL;
4944
    virErrorPtr orig_err = NULL;
4945
    int cookie_flags = 0;
J
Jiri Denemark 已提交
4946
    qemuDomainObjPrivatePtr priv = vm->privateData;
4947
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4948
    unsigned short port;
4949 4950
    unsigned long long timeReceived = 0;
    virObjectEventPtr event;
4951
    qemuDomainJobInfoPtr jobInfo = NULL;
4952
    bool inPostCopy = false;
4953
    bool doKill = true;
4954

4955
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
4956
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, retcode=%d",
4957 4958
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
4959

4960 4961 4962
    port = priv->migrationPort;
    priv->migrationPort = 0;

4963
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN)) {
4964
        qemuMigrationDstErrorReport(driver, vm->def->name);
4965
        goto cleanup;
4966
    }
4967

4968 4969
    ignore_value(virTimeMillisNow(&timeReceived));

4970 4971 4972
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
4973

4974
    qemuDomainCleanupRemove(vm, qemuMigrationDstPrepareCleanup);
4975
    VIR_FREE(priv->job.completed);
4976

4977
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
4978 4979
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
4980 4981 4982
    /* Some older versions of libvirt always send persistent XML in the cookie
     * even though VIR_MIGRATE_PERSIST_DEST was not used. */
    cookie_flags |= QEMU_MIGRATION_COOKIE_PERSISTENT;
4983

4984 4985
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen, cookie_flags)))
4986
        goto endjob;
4987

4988
    if (flags & VIR_MIGRATE_OFFLINE) {
4989
        if (retcode == 0 &&
4990
            qemuMigrationDstPersist(driver, vm, mig, false) == 0)
4991
            dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, -1);
4992 4993
        goto endjob;
    }
4994

4995 4996 4997 4998
    if (retcode != 0) {
        /* Check for a possible error on the monitor in case Finish was called
         * earlier than monitor EOF handler got a chance to process the error
         */
4999
        qemuDomainCheckMonitor(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN);
5000 5001
        goto endjob;
    }
5002

5003 5004 5005
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
5006
        qemuMigrationDstErrorReport(driver, vm->def->name);
5007 5008
        goto endjob;
    }
5009

5010
    if (qemuMigrationDstVPAssociatePortProfiles(vm->def) < 0)
5011
        goto endjob;
5012

5013
    if (mig->network && qemuMigrationDstOPDRelocate(driver, vm, mig) < 0)
5014
        VIR_WARN("unable to provide network data for relocation");
5015

5016
    if (qemuMigrationDstStopNBDServer(driver, vm, mig) < 0)
5017
        goto endjob;
5018

5019 5020
    if (qemuRefreshVirtioChannelState(driver, vm,
                                      QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
5021 5022
        goto endjob;

5023 5024
    if (qemuConnectAgent(driver, vm) < 0)
        goto endjob;
5025

5026
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
5027
        if (qemuMigrationDstPersist(driver, vm, mig, !v3proto) < 0) {
5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038
            /* Hmpf.  Migration was successful, but making it persistent
             * was not.  If we report successful, then when this domain
             * shuts down, management tools are in for a surprise.  On the
             * other hand, if we report failure, then the management tools
             * might try to restart the domain on the source side, even
             * though the domain is actually running on the destination.
             * Pretend success and hope that this is a rare situation and
             * management tools are smart.
             *
             * However, in v3 protocol, the source VM is still available
             * to restart during confirm() step, so we kill it off now.
5039 5040 5041
             */
            if (v3proto)
                goto endjob;
5042
        }
5043
    }
5044

5045 5046 5047
    /* We need to wait for QEMU to process all data sent by the source
     * before starting guest CPUs.
     */
5048 5049 5050
    if (qemuMigrationDstWaitForCompletion(driver, vm,
                                          QEMU_ASYNC_JOB_MIGRATION_IN,
                                          !!(flags & VIR_MIGRATE_POSTCOPY)) < 0) {
5051 5052 5053 5054 5055 5056
        /* There's not much we can do for v2 protocol since the
         * original domain on the source host is already gone.
         */
        if (v3proto)
            goto endjob;
    }
5057

5058 5059 5060 5061 5062 5063 5064 5065 5066
    /* Now that the state data was transferred we can refresh the actual state
     * of the devices */
    if (qemuProcessRefreshState(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
        /* Similarly to the case above v2 protocol will not be able to recover
         * from this. Let's ignore this and perhaps stuff will not break. */
        if (v3proto)
            goto endjob;
    }

5067
    if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY)
5068 5069
        inPostCopy = true;

5070 5071 5072 5073 5074
    if (!(flags & VIR_MIGRATE_PAUSED)) {
        /* run 'cont' on the destination, which allows migration on qemu
         * >= 0.10.6 to work properly.  This isn't strictly necessary on
         * older qemu's, but it also doesn't hurt anything there
         */
5075
        if (qemuProcessStartCPUs(driver, vm,
5076 5077
                                 inPostCopy ? VIR_DOMAIN_RUNNING_POSTCOPY
                                            : VIR_DOMAIN_RUNNING_MIGRATED,
5078
                                 QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5079
            if (virGetLastErrorCode() == VIR_ERR_OK)
5080 5081 5082 5083 5084 5085
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("resume operation failed"));
            /* Need to save the current error, in case shutting
             * down the process overwrites it
             */
            orig_err = virSaveLastError();
5086

5087 5088 5089 5090 5091 5092 5093 5094 5095 5096
            /*
             * In v3 protocol, the source VM is still available to
             * restart during confirm() step, so we kill it off
             * now.
             * In v2 protocol, the source is dead, so we leave
             * target in paused state, in case admin can fix
             * things up.
             */
            if (v3proto)
                goto endjob;
5097
        }
5098

5099
        if (inPostCopy)
5100
            doKill = false;
5101
    }
5102

5103
    if (mig->jobInfo) {
5104
        jobInfo = mig->jobInfo;
5105 5106 5107 5108 5109 5110
        mig->jobInfo = NULL;

        if (jobInfo->sent && timeReceived) {
            jobInfo->timeDelta = timeReceived - jobInfo->sent;
            jobInfo->received = timeReceived;
            jobInfo->timeDeltaSet = true;
5111
        }
5112 5113
        qemuDomainJobInfoUpdateTime(jobInfo);
        qemuDomainJobInfoUpdateDowntime(jobInfo);
5114
    }
L
liguang 已提交
5115

5116
    if (inPostCopy) {
5117 5118 5119
        if (qemuMigrationDstWaitForCompletion(driver, vm,
                                              QEMU_ASYNC_JOB_MIGRATION_IN,
                                              false) < 0) {
5120 5121 5122 5123 5124 5125 5126 5127 5128
            goto endjob;
        }
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            virDomainObjSetState(vm,
                                 VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_MIGRATED);
        }
    }

5129
    dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, vm->def->id);
5130

5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144
    if (inPostCopy) {
        /* The only RESUME event during post-copy migration is triggered by
         * QEMU when the running domain moves from the source to the
         * destination host, but then the migration keeps running until all
         * modified memory is transferred from the source host. This will
         * result in VIR_DOMAIN_EVENT_RESUMED with RESUMED_POSTCOPY detail.
         * However, our API documentation says we need to fire another RESUMED
         * event at the very end of migration with RESUMED_MIGRATED detail.
         */
        event = virDomainEventLifecycleNewFromObj(vm,
                                                  VIR_DOMAIN_EVENT_RESUMED,
                                                  VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
        virObjectEventStateQueue(driver->domainEventState, event);
    }
5145

5146 5147 5148 5149 5150
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_USER);
        event = virDomainEventLifecycleNewFromObj(vm,
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);
5151
        virObjectEventStateQueue(driver->domainEventState, event);
5152
    }
5153

5154
    if (virDomainObjIsActive(vm) &&
5155
        virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
5156 5157 5158 5159 5160
        VIR_WARN("Failed to save status on vm %s", vm->def->name);

    /* Guest is successfully running, so cancel previous auto destroy */
    qemuProcessAutoDestroyRemove(driver, vm);

5161
 endjob:
5162
    if (!dom &&
5163 5164
        !(flags & VIR_MIGRATE_OFFLINE) &&
        virDomainObjIsActive(vm)) {
5165
        if (doKill) {
5166 5167 5168 5169 5170 5171 5172
            qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                            QEMU_ASYNC_JOB_MIGRATION_IN,
                            VIR_QEMU_PROCESS_STOP_MIGRATED);
            virDomainAuditStop(vm, "failed");
            event = virDomainEventLifecycleNewFromObj(vm,
                                VIR_DOMAIN_EVENT_STOPPED,
                                VIR_DOMAIN_EVENT_STOPPED_FAILED);
5173
            virObjectEventStateQueue(driver->domainEventState, event);
5174
        } else {
5175
            qemuMigrationAnyPostcopyFailed(driver, vm);
5176
        }
5177 5178
    }

5179
    if (dom) {
5180 5181 5182
        if (jobInfo) {
            VIR_STEAL_PTR(priv->job.completed, jobInfo);
            priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
5183
            priv->job.completed->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;
5184
        }
5185

5186 5187 5188
        if (qemuMigrationBakeCookie(mig, driver, vm,
                                    QEMU_MIGRATION_DESTINATION,
                                    cookieout, cookieoutlen,
5189 5190
                                    QEMU_MIGRATION_COOKIE_STATS) < 0)
            VIR_WARN("Unable to encode migration cookie");
5191 5192 5193 5194 5195 5196

        /* Remove completed stats for post-copy, everything but timing fields
         * is obsolete anyway.
         */
        if (inPostCopy)
            VIR_FREE(priv->job.completed);
5197
    }
5198

5199
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
5200
                             priv->job.migParams, priv->job.apiFlags);
5201

5202
    qemuMigrationJobFinish(driver, vm);
5203
    if (!virDomainObjIsActive(vm))
5204
        qemuDomainRemoveInactiveJob(driver, vm);
5205

5206
 cleanup:
5207
    VIR_FREE(jobInfo);
5208
    virPortAllocatorRelease(port);
5209
    if (priv->mon)
5210
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
5211
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5212
    virDomainObjEndAPI(&vm);
5213
    qemuMigrationCookieFree(mig);
5214 5215 5216 5217
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5218
    virObjectUnref(cfg);
5219 5220 5221 5222

    /* Set a special error if Finish is expected to return NULL as a result of
     * successful call with retcode != 0
     */
5223
    if (retcode != 0 && !dom && virGetLastErrorCode() == VIR_ERR_OK)
5224
        virReportError(VIR_ERR_MIGRATE_FINISH_OK, NULL);
5225 5226
    return dom;
}
5227

5228

5229
/* Helper function called while vm is active.  */
5230
int
5231 5232 5233 5234
qemuMigrationSrcToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
                       int fd,
                       const char *compressor,
                       qemuDomainAsyncJob asyncJob)
5235 5236 5237
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5238
    int ret = -1;
5239 5240
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5241
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5242
    char *errbuf = NULL;
5243
    virErrorPtr orig_err = NULL;
5244 5245 5246 5247 5248

    /* Increase migration bandwidth to unlimited since target is a file.
     * Failure to change migration speed is not fatal. */
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
        qemuMonitorSetMigrationSpeed(priv->mon,
5249 5250
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5251 5252
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5253
    }
5254

5255 5256 5257 5258 5259 5260 5261
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5262 5263 5264 5265
    if (compressor && pipe(pipeFD) < 0) {
        virReportSystemError(errno, "%s",
                             _("Failed to create pipe for migration"));
        return -1;
5266 5267
    }

5268 5269 5270 5271
    /* All right! We can use fd migration, which means that qemu
     * doesn't have to open() the file, so while we still have to
     * grant SELinux access, we can do it on fd and avoid cleanup
     * later, as well as skip futzing with cgroup.  */
5272 5273
    if (qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    compressor ? pipeFD[1] : fd) < 0)
5274 5275
        goto cleanup;

5276
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5277 5278
        goto cleanup;

5279
    if (!compressor) {
5280 5281 5282
        rc = qemuMonitorMigrateToFd(priv->mon,
                                    QEMU_MONITOR_MIGRATE_BACKGROUND,
                                    fd);
5283 5284 5285 5286 5287 5288 5289
    } else {
        const char *prog = compressor;
        const char *args[] = {
            prog,
            "-c",
            NULL
        };
5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300

        cmd = virCommandNewArgs(args);
        virCommandSetInputFD(cmd, pipeFD[0]);
        virCommandSetOutputFD(cmd, &fd);
        virCommandSetErrorBuffer(cmd, &errbuf);
        virCommandDoAsyncIO(cmd);
        if (virSetCloseExec(pipeFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("Unable to set cloexec flag"));
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
            goto cleanup;
5301
        }
5302 5303 5304 5305 5306 5307 5308 5309 5310 5311
        if (virCommandRunAsync(cmd, NULL) < 0) {
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
            goto cleanup;
        }
        rc = qemuMonitorMigrateToFd(priv->mon,
                                    QEMU_MONITOR_MIGRATE_BACKGROUND,
                                    pipeFD[1]);
        if (VIR_CLOSE(pipeFD[0]) < 0 ||
            VIR_CLOSE(pipeFD[1]) < 0)
            VIR_WARN("failed to close intermediate pipe");
5312
    }
5313
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5314
        goto cleanup;
5315 5316 5317
    if (rc < 0)
        goto cleanup;

5318
    rc = qemuMigrationSrcWaitForCompletion(driver, vm, asyncJob, NULL, 0);
5319

5320 5321 5322 5323
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5324 5325
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5326
                qemuMonitorMigrateCancel(priv->mon);
5327
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5328 5329
            }
        }
5330
        goto cleanup;
5331
    }
5332

5333 5334 5335
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5336
    qemuDomainEventEmitJobCompleted(driver, vm);
5337 5338
    ret = 0;

5339
 cleanup:
5340 5341 5342
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5343
    /* Restore max migration bandwidth */
5344 5345
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5346 5347
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5348
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5349 5350
    }

5351 5352
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5353 5354 5355 5356 5357
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5358 5359 5360 5361 5362 5363

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

5364 5365
    return ret;
}
5366

5367 5368

int
5369 5370
qemuMigrationSrcCancel(virQEMUDriverPtr driver,
                       virDomainObjPtr vm)
5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool storage = false;
    size_t i;
    int ret = -1;

    VIR_DEBUG("Canceling unfinished outgoing migration of domain %s",
              vm->def->name);

    qemuDomainObjEnterMonitor(driver, vm);
    ignore_value(qemuMonitorMigrateCancel(priv->mon));
5382
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5383 5384 5385 5386 5387
        goto cleanup;

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
5388
        qemuBlockJobDataPtr job;
5389

5390
        if (!(job = qemuBlockJobDiskGetJob(disk)) ||
5391
            !qemuBlockJobIsRunning(job))
5392
            diskPriv->migrating = false;
5393 5394

        if (diskPriv->migrating) {
5395
            qemuBlockJobSyncBegin(job);
5396
            storage = true;
5397 5398
        }

5399 5400
        virObjectUnref(job);
    }
5401 5402 5403

    if (storage &&
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
5404
                                      QEMU_ASYNC_JOB_NONE, NULL) < 0)
5405
        goto cleanup;
5406 5407 5408 5409 5410 5411 5412 5413

    ret = 0;

 cleanup:
    return ret;
}


5414
static int
5415
qemuMigrationJobStart(virQEMUDriverPtr driver,
5416
                      virDomainObjPtr vm,
5417 5418
                      qemuDomainAsyncJob job,
                      unsigned long apiFlags)
5419
{
5420
    qemuDomainObjPrivatePtr priv = vm->privateData;
5421 5422
    virDomainJobOperation op;
    unsigned long long mask;
5423

5424
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5425 5426
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_IN;
        mask = QEMU_JOB_NONE;
5427
    } else {
5428 5429 5430 5431
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_OUT;
        mask = QEMU_JOB_DEFAULT_MASK |
               JOB_MASK(QEMU_JOB_SUSPEND) |
               JOB_MASK(QEMU_JOB_MIGRATION_OP);
5432
    }
5433

5434
    if (qemuDomainObjBeginAsyncJob(driver, vm, job, op, apiFlags) < 0)
5435 5436
        return -1;

5437 5438
    priv->job.current->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;

5439
    qemuDomainObjSetAsyncJobMask(vm, mask);
5440 5441 5442
    return 0;
}

5443
static void
5444
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5445
                         virDomainObjPtr vm,
5446
                         qemuMigrationJobPhase phase)
5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (phase < priv->job.phase) {
        VIR_ERROR(_("migration protocol going backwards %s => %s"),
                  qemuMigrationJobPhaseTypeToString(priv->job.phase),
                  qemuMigrationJobPhaseTypeToString(phase));
        return;
    }

    qemuDomainObjSetJobPhase(driver, vm, phase);
}

5460
static void
5461
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5462
                           virDomainObjPtr vm,
5463
                           qemuMigrationJobPhase phase)
5464 5465 5466 5467
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5468
static void
5469 5470
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5471
    qemuDomainObjReleaseAsyncJob(vm);
5472 5473
}

5474
static bool
5475
qemuMigrationJobIsActive(virDomainObjPtr vm,
5476
                         qemuDomainAsyncJob job)
5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->job.asyncJob != job) {
        const char *msg;

        if (job == QEMU_ASYNC_JOB_MIGRATION_IN)
            msg = _("domain '%s' is not processing incoming migration");
        else
            msg = _("domain '%s' is not being migrated");

5488
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5489 5490 5491 5492 5493
        return false;
    }
    return true;
}

5494
static void
5495
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5496
{
5497
    qemuDomainObjEndAsyncJob(driver, vm);
5498
}
5499 5500 5501


static void
5502
qemuMigrationDstErrorFree(void *data,
5503 5504 5505 5506 5507 5508 5509
                       const void *name ATTRIBUTE_UNUSED)
{
    virErrorPtr err = data;
    virFreeError(err);
}

int
5510
qemuMigrationDstErrorInit(virQEMUDriverPtr driver)
5511
{
5512
    driver->migrationErrors = virHashAtomicNew(64, qemuMigrationDstErrorFree);
5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523
    if (driver->migrationErrors)
        return 0;
    else
        return -1;
}

/**
 * This function consumes @err; the caller should consider the @err pointer
 * invalid after calling this function.
 */
void
5524 5525 5526
qemuMigrationDstErrorSave(virQEMUDriverPtr driver,
                          const char *name,
                          virErrorPtr err)
5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539
{
    if (!err)
        return;

    VIR_DEBUG("Saving incoming migration error for domain %s: %s",
              name, err->message);
    if (virHashAtomicUpdate(driver->migrationErrors, name, err) < 0) {
        VIR_WARN("Failed to save migration error for domain '%s'", name);
        virFreeError(err);
    }
}

void
5540 5541
qemuMigrationDstErrorReport(virQEMUDriverPtr driver,
                            const char *name)
5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552
{
    virErrorPtr err;

    if (!(err = virHashAtomicSteal(driver->migrationErrors, name)))
        return;

    VIR_DEBUG("Restoring saved incoming migration error for domain %s: %s",
              name, err->message);
    virSetError(err);
    virFreeError(err);
}
5553 5554


5555
int
5556 5557 5558 5559
qemuMigrationSrcFetchMirrorStats(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 qemuDomainJobInfoPtr jobInfo)
5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603
{
    size_t i;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool nbd = false;
    virHashTablePtr blockinfo = NULL;
    qemuDomainMirrorStatsPtr stats = &jobInfo->mirrorStats;

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        if (QEMU_DOMAIN_DISK_PRIVATE(disk)->migrating) {
            nbd = true;
            break;
        }
    }

    if (!nbd)
        return 0;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;

    blockinfo = qemuMonitorGetAllBlockJobInfo(priv->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || !blockinfo)
        return -1;

    memset(stats, 0, sizeof(*stats));

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
        qemuMonitorBlockJobInfoPtr data;

        if (!diskPriv->migrating ||
            !(data = virHashLookup(blockinfo, disk->info.alias)))
            continue;

        stats->transferred += data->cur;
        stats->total += data->end;
    }

    virHashFree(blockinfo);
    return 0;
}