qemu_migration.c 187.2 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);
461 462 463 464 465 466
    priv->nbdPort = 0;
    return 0;
}


/**
467
 * qemuMigrationSrcNBDStorageCopyReady:
468 469
 * @vm: domain
 *
470 471
 * Check the status of all drives copied via qemuMigrationSrcNBDStorageCopy.
 * Any pending block job events for the mirrored disks will be processed.
472 473 474 475 476 477
 *
 * Returns 1 if all mirrors are "ready",
 *         0 if some mirrors are still performing initial sync,
 *        -1 on error.
 */
static int
478 479
qemuMigrationSrcNBDStorageCopyReady(virDomainObjPtr vm,
                                    qemuDomainAsyncJob asyncJob)
480 481
{
    size_t i;
482 483
    size_t notReady = 0;
    int status;
484 485 486

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
487
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
488
        char *error = NULL;
489

490
        if (!diskPriv->migrating)
491 492
            continue;

493
        status = qemuBlockJobUpdateDisk(vm, asyncJob, disk, &error);
494
        if (status == VIR_DOMAIN_BLOCK_JOB_FAILED) {
495 496 497 498 499 500 501 502 503
            if (error) {
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("migration of disk %s failed: %s"),
                               disk->dst, error);
                VIR_FREE(error);
            } else {
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("migration of disk %s failed"), disk->dst);
            }
504 505
            return -1;
        }
506
        VIR_FREE(error);
507 508 509

        if (disk->mirrorState != VIR_DOMAIN_DISK_MIRROR_STATE_READY)
            notReady++;
510 511
    }

512 513 514 515 516 517 518
    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;
    }
519 520 521
}


522 523 524 525
/*
 * 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.
526
 *
527 528 529 530
 * 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.
531 532
 */
static int
533 534 535
qemuMigrationSrcNBDCopyCancelled(virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 bool check)
536
{
537 538
    size_t i;
    size_t active = 0;
539
    size_t completed = 0;
540 541
    int status;
    bool failed = false;
542

543
 retry:
544 545 546
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
547
        char *error = NULL;
548

549 550
        if (!diskPriv->migrating)
            continue;
551

552
        status = qemuBlockJobUpdateDisk(vm, asyncJob, disk, &error);
553 554 555
        switch (status) {
        case VIR_DOMAIN_BLOCK_JOB_FAILED:
            if (check) {
556 557 558 559 560 561 562 563
                if (error) {
                    virReportError(VIR_ERR_OPERATION_FAILED,
                                   _("migration of disk %s failed: %s"),
                                   disk->dst, error);
                } else {
                    virReportError(VIR_ERR_OPERATION_FAILED,
                                   _("migration of disk %s failed"), disk->dst);
                }
564
                failed = true;
565
            }
M
Marc Hartmayer 已提交
566
            ATTRIBUTE_FALLTHROUGH;
567 568
        case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
569
            qemuBlockJobSyncEndDisk(vm, asyncJob, disk);
570 571
            diskPriv->migrating = false;
            break;
572

573 574
        default:
            active++;
575
        }
576 577 578

        if (status == VIR_DOMAIN_BLOCK_JOB_COMPLETED)
            completed++;
579 580

        VIR_FREE(error);
581 582 583 584 585 586 587 588 589 590
    }

    /* 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;
591
    }
592

593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608
    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;
609
        }
610
    }
611
}
612 613


614 615 616 617 618 619
/*
 * 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
620 621 622 623 624
qemuMigrationSrcNBDCopyCancelOne(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 virDomainDiskDefPtr disk,
                                 bool failNoJob,
                                 qemuDomainAsyncJob asyncJob)
625 626 627
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char *diskAlias = NULL;
628
    char *error = NULL;
629 630 631 632
    int ret = -1;
    int status;
    int rv;

633
    status = qemuBlockJobUpdateDisk(vm, asyncJob, disk, &error);
634 635 636 637
    switch (status) {
    case VIR_DOMAIN_BLOCK_JOB_FAILED:
    case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        if (failNoJob) {
638 639 640 641 642 643 644 645 646
            if (error) {
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("migration of disk %s failed: %s"),
                               disk->dst, error);
            } else {
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("migration of disk %s failed"), disk->dst);
            }
            goto cleanup;
647
        }
648
        ATTRIBUTE_FALLTHROUGH;
649
    case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
650 651
        ret = 1;
        goto cleanup;
652 653
    }

654
    if (!(diskAlias = qemuAliasDiskDriveFromDisk(disk)))
655 656
        return -1;

657
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
658 659
        goto cleanup;

660
    rv = qemuMonitorBlockJobCancel(priv->mon, diskAlias);
661 662 663 664 665

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

    ret = 0;
666 667 668

 cleanup:
    VIR_FREE(diskAlias);
669
    VIR_FREE(error);
670 671 672 673 674
    return ret;
}


/**
675
 * qemuMigrationSrcNBDCopyCancel:
676 677
 * @driver: qemu driver
 * @vm: domain
678
 * @check: if true report an error when some of the mirrors fails
679
 *
680
 * Cancel all drive-mirrors started by qemuMigrationSrcNBDStorageCopy.
681 682 683
 * 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.
684 685 686 687
 *
 * Returns 0 on success, -1 otherwise.
 */
static int
688 689 690 691 692
qemuMigrationSrcNBDCopyCancel(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              bool check,
                              qemuDomainAsyncJob asyncJob,
                              virConnectPtr dconn)
693
{
694
    virErrorPtr err = NULL;
695
    int ret = -1;
696
    size_t i;
697 698 699 700
    int rv;
    bool failed = false;

    VIR_DEBUG("Cancelling drive mirrors for domain %s", vm->def->name);
701 702 703

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
704
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
705

706 707 708 709 710
        if (!diskPriv->blockjob->started) {
            qemuBlockJobSyncEndDisk(vm, asyncJob, disk);
            diskPriv->migrating = false;
        }

711
        if (!diskPriv->migrating)
712 713
            continue;

714 715
        rv = qemuMigrationSrcNBDCopyCancelOne(driver, vm, disk,
                                              check, asyncJob);
716 717 718 719 720 721
        if (rv != 0) {
            if (rv < 0) {
                if (!err)
                    err = virSaveLastError();
                failed = true;
            }
722
            qemuBlockJobSyncEndDisk(vm, asyncJob, disk);
723 724 725 726
            diskPriv->migrating = false;
        }
    }

727
    while ((rv = qemuMigrationSrcNBDCopyCancelled(vm, asyncJob, check)) != 1) {
728 729 730 731 732 733 734
        if (check && !failed &&
            dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            failed = true;
        }

735 736 737 738
        if (rv < 0) {
            failed = true;
            if (rv == -2)
                break;
739
        }
740

741 742 743 744 745
        if (failed && !err)
            err = virSaveLastError();

        if (virDomainObjWait(vm) < 0)
            goto cleanup;
746 747
    }

748 749 750 751 752 753 754 755 756 757 758 759 760
    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;
    }

761 762 763
    ret = failed ? -1 : 0;

 cleanup:
764 765 766 767 768
    if (err) {
        virSetError(err);
        virFreeError(err);
    }
    return ret;
769 770 771
}


772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850
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;
}


851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892
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;
}


893
/**
894
 * qemuMigrationSrcNBDStorageCopy:
895 896 897 898
 * @driver: qemu driver
 * @vm: domain
 * @mig: migration cookie
 * @host: where are we migrating to
899
 * @speed: bandwidth limit in MiB/s
900 901
 * @migrate_flags: migrate monitor command flags
 *
902 903 904
 * 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
905
 * to do the very same operation. On failure, the caller is
906 907
 * expected to call qemuMigrationSrcNBDCopyCancel to stop all
 * running copy operations.
908 909 910 911 912
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
913 914 915 916 917 918 919 920
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,
921 922 923
                               virConnectPtr dconn,
                               const char *tlsAlias,
                               unsigned int flags)
924 925 926 927
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int port;
928
    size_t i;
929
    char *diskAlias = NULL;
930
    unsigned long long mirror_speed = speed;
931
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
932
    int rv;
933
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
934 935

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
936

937 938 939 940 941 942 943 944
    if (mirror_speed > LLONG_MAX >> 20) {
        virReportError(VIR_ERR_OVERFLOW,
                       _("bandwidth must be less than %llu"),
                       LLONG_MAX >> 20);
        goto cleanup;
    }
    mirror_speed <<= 20;

945 946 947 948 949 950 951 952 953
    /* 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];
954
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
955
        int rc;
956

957
        /* check whether disk should be migrated */
958
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
959 960
            continue;

961
        if (!(diskAlias = qemuAliasDiskDriveFromDisk(disk)))
962 963
            goto cleanup;

964
        qemuBlockJobSyncBeginDisk(disk);
965

966 967 968 969 970 971 972 973 974 975 976 977 978 979 980
        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);
        }

        if (rc < 0) {
981
            qemuBlockJobSyncEndDisk(vm, QEMU_ASYNC_JOB_MIGRATION_OUT, disk);
982 983
            goto cleanup;
        }
984 985

        VIR_FREE(diskAlias);
986
        diskPriv->migrating = true;
987
        diskPriv->blockjob->started = true;
988

989
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
990 991 992
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto cleanup;
        }
993
    }
994

995
    while ((rv = qemuMigrationSrcNBDStorageCopyReady(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) != 1) {
996 997
        if (rv < 0)
            goto cleanup;
998

999
        if (priv->job.abortJob) {
1000
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1001 1002 1003 1004
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
            goto cleanup;
1005
        }
1006

1007 1008 1009 1010 1011 1012
        if (dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            goto cleanup;
        }

1013
        if (virDomainObjWait(vm) < 0)
1014
            goto cleanup;
1015 1016
    }

1017 1018
    qemuMigrationSrcFetchMirrorStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                     priv->job.current);
1019

1020
    /* Okay, all disks are ready. Modify migrate_flags */
1021 1022 1023 1024
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

1025
 cleanup:
1026
    virObjectUnref(cfg);
1027 1028 1029
    VIR_FREE(diskAlias);
    return ret;
}
1030

1031

1032
/**
1033
 * qemuMigrationSrcIsAllowedHostdev:
1034 1035
 * @def: domain definition
 *
1036
 * Checks that @def does not contain any host devices unsupported across
1037 1038 1039
 * migrations. Returns true if the vm is allowed to migrate.
 */
static bool
1040
qemuMigrationSrcIsAllowedHostdev(const virDomainDef *def)
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
{
    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;
}


1060
/**
1061
 * qemuMigrationSrcIsAllowed:
1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073
 * @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.
 */
1074
bool
1075 1076 1077 1078
qemuMigrationSrcIsAllowed(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          bool remote,
                          unsigned int flags)
1079
{
1080
    int nsnapshots;
1081
    int pauseReason;
1082
    size_t i;
1083

1084 1085 1086 1087 1088
    /* perform these checks only when migrating to remote hosts */
    if (remote) {
        nsnapshots = virDomainSnapshotObjListNum(vm->snapshots, NULL, 0);
        if (nsnapshots < 0)
            return false;
1089

1090 1091 1092 1093 1094
        if (nsnapshots > 0) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("cannot migrate domain with %d snapshots"),
                           nsnapshots);
            return false;
1095
        }
1096

1097
        /* cancel migration if disk I/O error is emitted while migrating */
1098
        if (flags & VIR_MIGRATE_ABORT_ON_ERROR &&
1099
            !(flags & VIR_MIGRATE_OFFLINE) &&
1100 1101
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
E
Eric Blake 已提交
1102
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1103
                           _("cannot migrate domain with I/O error"));
E
Eric Blake 已提交
1104 1105
            return false;
        }
1106 1107

    }
1108

1109 1110 1111 1112 1113 1114 1115
    /* 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;
        }
1116

1117

1118 1119
        if (qemuDomainHasBlockjob(vm, false)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1120
                           _("domain has active block job"));
1121 1122 1123
            return false;
        }

1124
        if (!qemuMigrationSrcIsAllowedHostdev(vm->def))
1125 1126
            return false;

1127
        if (vm->def->cpu) {
1128 1129 1130
            /* QEMU blocks migration and save with invariant TSC enabled
             * unless TSC frequency is explicitly set.
             */
1131 1132
            if (virCPUCheckFeature(vm->def->os.arch, vm->def->cpu,
                                   "invtsc") == 1) {
1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150
                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;
                }
1151
            }
J
Ján Tomko 已提交
1152 1153
        }

1154 1155 1156
        /* Verify that memory device config can be transferred reliably */
        for (i = 0; i < vm->def->nmems; i++) {
            virDomainMemoryDefPtr mem = vm->def->mems[i];
1157

1158 1159 1160 1161 1162
            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"));
1163

1164 1165
                return false;
            }
1166
        }
1167 1168 1169 1170 1171 1172

        if (vm->def->nshmems) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("migration with shmem device is not supported"));
            return false;
        }
1173 1174
    }

1175 1176 1177
    return true;
}

1178
static bool
1179 1180 1181 1182
qemuMigrationSrcIsSafe(virDomainDefPtr def,
                       size_t nmigrate_disks,
                       const char **migrate_disks,
                       unsigned int flags)
1183

1184
{
1185 1186
    bool storagemigration = flags & (VIR_MIGRATE_NON_SHARED_DISK |
                                     VIR_MIGRATE_NON_SHARED_INC);
1187
    size_t i;
1188
    int rc;
1189

1190
    for (i = 0; i < def->ndisks; i++) {
1191
        virDomainDiskDefPtr disk = def->disks[i];
1192
        const char *src = virDomainDiskGetSource(disk);
1193

1194 1195
        /* Disks without any source (i.e. floppies and CD-ROMs)
         * OR readonly are safe. */
1196
        if (virStorageSourceIsEmpty(disk->src) ||
1197
            disk->src->readonly)
1198
            continue;
1199

1200
        /* Disks which are migrated by qemu are safe too. */
1201
        if (storagemigration &&
1202
            qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1203
            continue;
1204

1205
        /* However, disks on local FS (e.g. ext4) are not safe. */
1206
        if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
1207
            if ((rc = virFileIsSharedFS(src)) < 0) {
1208
                return false;
1209 1210 1211 1212 1213
            } else if (rc == 0) {
                virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                               _("Migration without shared storage is unsafe"));
                return false;
            }
1214 1215 1216 1217 1218 1219 1220
            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;
1221
        }
1222

1223 1224 1225 1226 1227 1228 1229
        /* 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;

1230 1231
        virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                       _("Migration may lead to data corruption if disks"
1232
                         " use cache != none or cache != directsync"));
1233
        return false;
1234 1235 1236 1237 1238
    }

    return true;
}

1239
/** qemuMigrationSrcSetOffline
1240 1241 1242
 * Pause domain for non-live migration.
 */
int
1243 1244
qemuMigrationSrcSetOffline(virQEMUDriverPtr driver,
                           virDomainObjPtr vm)
1245 1246
{
    int ret;
1247
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
1248 1249
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
1250
    if (ret == 0) {
1251
        virObjectEventPtr event;
1252

1253
        event = virDomainEventLifecycleNewFromObj(vm,
1254 1255
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
1256
        virObjectEventStateQueue(driver->domainEventState, event);
1257 1258 1259 1260 1261
    }

    return ret;
}

1262 1263

void
1264 1265
qemuMigrationAnyPostcopyFailed(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
{
    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);
1296
        virObjectEventStateQueue(driver->domainEventState, event);
1297 1298 1299 1300 1301 1302 1303
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
    }
}


1304
static int
1305
qemuMigrationSrcWaitForSpice(virDomainObjPtr vm)
1306 1307 1308
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

1309 1310
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION) ||
        !priv->job.spiceMigration)
1311 1312
        return 0;

1313
    VIR_DEBUG("Waiting for SPICE to finish migration");
1314 1315
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
1316 1317 1318 1319
            return -1;
    }
    return 0;
}
1320

1321 1322 1323 1324

static void
qemuMigrationUpdateJobType(qemuDomainJobInfoPtr jobInfo)
{
1325
    switch ((qemuMonitorMigrationStatus) jobInfo->stats.mig.status) {
1326 1327 1328 1329
    case QEMU_MONITOR_MIGRATION_STATUS_POSTCOPY:
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_POSTCOPY;
        break;

1330
    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
1331
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED;
1332 1333 1334
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
1335
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_NONE;
1336 1337 1338
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
1339
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1340 1341 1342
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
1343
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1344 1345
        break;

1346 1347 1348 1349 1350 1351 1352 1353
    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;

1354 1355 1356
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLING:
1357
    case QEMU_MONITOR_MIGRATION_STATUS_LAST:
1358 1359 1360 1361 1362 1363
        break;
    }
}


int
1364 1365 1366 1367 1368
qemuMigrationAnyFetchStats(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuDomainAsyncJob asyncJob,
                           qemuDomainJobInfoPtr jobInfo,
                           char **error)
1369 1370
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1371
    qemuMonitorMigrationStats stats;
1372 1373 1374 1375 1376
    int rv;

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

1377
    rv = qemuMonitorGetMigrationStats(priv->mon, &stats, error);
1378 1379 1380 1381

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

1382
    jobInfo->stats.mig = stats;
1383 1384

    return 0;
1385 1386 1387
}


1388 1389 1390 1391 1392 1393 1394
static const char *
qemuMigrationJobName(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
1395
        return _("migration out job");
1396 1397 1398 1399
    case QEMU_ASYNC_JOB_SAVE:
        return _("domain save job");
    case QEMU_ASYNC_JOB_DUMP:
        return _("domain core dump job");
1400 1401 1402 1403 1404 1405 1406 1407 1408
    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:
1409 1410 1411 1412 1413 1414
    default:
        return _("job");
    }
}


1415
static int
1416
qemuMigrationJobCheckStatus(virQEMUDriverPtr driver,
1417
                            virDomainObjPtr vm,
1418
                            qemuDomainAsyncJob asyncJob)
1419 1420 1421
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1422
    char *error = NULL;
1423
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1424
    int ret = -1;
1425

1426
    if (!events ||
1427
        jobInfo->stats.mig.status == QEMU_MONITOR_MIGRATION_STATUS_ERROR) {
1428
        if (qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, &error) < 0)
1429 1430
            return -1;
    }
1431

1432 1433
    qemuMigrationUpdateJobType(jobInfo);

1434 1435
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_NONE:
1436 1437
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("is not active"));
1438
        goto cleanup;
1439

1440
    case QEMU_DOMAIN_JOB_STATUS_FAILED:
1441
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
1442 1443 1444
                       qemuMigrationJobName(vm),
                       error ? error : _("unexpectedly failed"));
        goto cleanup;
1445

1446
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
1447 1448
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("canceled by client"));
1449
        goto cleanup;
1450

1451 1452
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
1453
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
1454
    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
1455
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1456
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1457 1458
        break;
    }
1459 1460 1461 1462 1463 1464

    ret = 0;

 cleanup:
    VIR_FREE(error);
    return ret;
1465 1466 1467
}


1468 1469
enum qemuMigrationCompletedFlags {
    QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR = (1 << 0),
1470
    /* This flag should only be set when run on src host */
1471
    QEMU_MIGRATION_COMPLETED_CHECK_STORAGE  = (1 << 1),
1472
    QEMU_MIGRATION_COMPLETED_POSTCOPY       = (1 << 2),
1473
    QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER = (1 << 3),
1474 1475
};

1476

1477 1478 1479 1480 1481 1482 1483
/**
 * 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
1484 1485 1486 1487 1488
qemuMigrationAnyCompleted(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          qemuDomainAsyncJob asyncJob,
                          virConnectPtr dconn,
                          unsigned int flags)
1489 1490 1491 1492 1493
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    int pauseReason;

1494
    if (qemuMigrationJobCheckStatus(driver, vm, asyncJob) < 0)
1495 1496
        goto error;

1497
    /* This flag should only be set when run on src host */
1498
    if (flags & QEMU_MIGRATION_COMPLETED_CHECK_STORAGE &&
1499
        qemuMigrationSrcNBDStorageCopyReady(vm, asyncJob) < 0)
1500 1501
        goto error;

1502
    if (flags & QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR &&
1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515
        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;
    }

1516 1517 1518 1519 1520 1521 1522 1523 1524 1525
    /* 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;
    }

1526 1527 1528 1529 1530
    /* 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 &&
1531
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
1532 1533 1534 1535
        VIR_DEBUG("Migration switched to post-copy");
        return 1;
    }

1536
    if (jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
1537 1538 1539 1540 1541
        return 1;
    else
        return 0;

 error:
1542 1543 1544
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1545
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1546
        /* The migration was aborted by us rather than QEMU itself. */
1547
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1548
        return -2;
1549 1550 1551

    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
        /* Something failed after QEMU already finished the migration. */
1552
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1553
        return -1;
1554 1555 1556 1557

    case QEMU_DOMAIN_JOB_STATUS_FAILED:
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
        /* QEMU aborted the migration. */
1558
        return -1;
1559 1560 1561 1562 1563 1564

    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_NONE:
        /* Impossible. */
        break;
1565
    }
1566 1567

    return -1;
1568 1569 1570
}


1571 1572 1573
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
1574
static int
1575 1576 1577 1578 1579
qemuMigrationSrcWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  virConnectPtr dconn,
                                  unsigned int flags)
1580
{
1581
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
1582
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1583
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1584
    int rv;
1585

1586 1587
    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_MIGRATING;

1588 1589
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           dconn, flags)) != 1) {
1590 1591
        if (rv < 0)
            return rv;
1592

1593 1594
        if (events) {
            if (virDomainObjWait(vm) < 0) {
1595 1596
                if (virDomainObjIsActive(vm))
                    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1597 1598 1599 1600 1601 1602 1603 1604 1605 1606
                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);
        }
1607 1608
    }

1609
    if (events)
1610
        ignore_value(qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, NULL));
1611

1612
    qemuDomainJobInfoUpdateTime(jobInfo);
1613 1614
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
1615
    if (VIR_ALLOC(priv->job.completed) == 0) {
1616
        *priv->job.completed = *jobInfo;
1617 1618
        priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
    }
1619

1620 1621 1622 1623
    if (asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT &&
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;

1624
    return 0;
1625 1626 1627
}


1628
static int
1629 1630 1631 1632
qemuMigrationDstWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  bool postcopy)
1633 1634
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1635
    unsigned int flags = 0;
1636 1637 1638 1639 1640 1641 1642
    int rv;

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

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

1643 1644 1645
    if (postcopy)
        flags = QEMU_MIGRATION_COMPLETED_POSTCOPY;

1646 1647
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           NULL, flags)) != 1) {
1648 1649 1650 1651 1652 1653 1654 1655
        if (rv < 0 || virDomainObjWait(vm) < 0)
            return -1;
    }

    return 0;
}


1656
static int
1657 1658 1659 1660
qemuMigrationSrcGraphicsRelocate(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuMigrationCookiePtr cookie,
                                 const char *graphicsuri)
1661 1662
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1663 1664
    int ret = -1;
    const char *listenAddress = NULL;
1665
    virSocketAddr addr;
1666 1667 1668 1669 1670
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
1671

1672
    if (!cookie || (!cookie->graphics && !graphicsuri))
1673 1674
        return 0;

1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
    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) {
1694
        size_t i;
1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721

        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;
            }
        }
    }
1722 1723 1724 1725

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
1726 1727 1728 1729
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
1730

1731 1732 1733 1734 1735 1736 1737 1738 1739
    /* 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;
    }

1740 1741 1742 1743
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
1744
        priv->job.spiceMigration = !ret;
1745 1746
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
1747
    }
1748

1749
 cleanup:
1750
    virURIFree(uri);
1751 1752 1753 1754
    return ret;
}


1755
static int
1756 1757 1758
qemuMigrationDstOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            qemuMigrationCookiePtr cookie)
1759
{
1760 1761
    virDomainNetDefPtr netptr;
    int ret = -1;
1762
    size_t i;
1763 1764 1765 1766 1767 1768 1769 1770

    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:
1771
           break;
1772
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
1773 1774
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
1775 1776 1777
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
1778 1779 1780
                goto cleanup;
            }
            break;
1781 1782 1783 1784 1785
        default:
            break;
        }
    }

1786
    ret = 0;
1787
 cleanup:
1788 1789 1790 1791
    return ret;
}


1792
int
1793 1794
qemuMigrationDstCheckProtocol(virQEMUCapsPtr qemuCaps,
                              const char *migrateFrom)
1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817
{
    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 *
1818 1819
qemuMigrationDstGetURI(const char *migrateFrom,
                       int migrateFd)
1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831
{
    char *uri = NULL;

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

    return uri;
}


1832
int
1833 1834 1835 1836
qemuMigrationDstRun(virQEMUDriverPtr driver,
                    virDomainObjPtr vm,
                    const char *uri,
                    qemuDomainAsyncJob asyncJob)
1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
{
    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) {
1853
        /* qemuMigrationDstWaitForCompletion is called from the Finish phase */
1854 1855 1856 1857
        ret = 0;
        goto cleanup;
    }

1858
    if (qemuMigrationDstWaitForCompletion(driver, vm, asyncJob, false) < 0)
1859 1860 1861 1862 1863 1864 1865 1866 1867
        goto cleanup;

    ret = 0;

 cleanup:
    return ret;
}


1868 1869 1870 1871 1872 1873
/* 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.
 */
1874
static void
1875 1876 1877
qemuMigrationSrcCleanup(virDomainObjPtr vm,
                        virConnectPtr conn,
                        void *opaque)
1878
{
1879
    virQEMUDriverPtr driver = opaque;
1880 1881 1882 1883 1884 1885 1886 1887 1888
    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))
1889
        return;
1890 1891 1892 1893 1894

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

1895
    switch ((qemuMigrationJobPhase) priv->job.phase) {
1896 1897 1898 1899 1900 1901 1902 1903 1904
    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);
1905
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
1906
                                 priv->job.migParams, priv->job.apiFlags);
1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928
        /* 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 */
        ;
    }
}

1929

1930
/* The caller is supposed to lock the vm and start a migration job. */
1931
static char *
1932 1933 1934 1935 1936 1937 1938 1939 1940
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)
1941 1942 1943
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
1944
    virDomainDefPtr def = NULL;
1945
    qemuDomainObjPrivatePtr priv = vm->privateData;
1946
    virCapsPtr caps = NULL;
1947
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
1948

1949
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
1950
              " cookieout=%p, cookieoutlen=%p,"
1951
              " nmigrate_disks=%zu, migrate_disks=%p, flags=0x%lx",
1952
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
1953 1954
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
1955

1956 1957 1958
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

1959 1960 1961 1962 1963 1964
    /* 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);
1965

1966
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
1967 1968
        goto cleanup;

1969
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
1970
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
1971 1972
        goto cleanup;

1973 1974 1975 1976 1977 1978 1979 1980 1981
    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;
    }

1982 1983 1984 1985 1986 1987
    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;
    }

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 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
    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;
            }
2031 2032 2033
        }
    }

2034
    if (virDomainDefHasMemoryHotplug(vm->def) ||
2035
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
2036
         vm->newDef && virDomainDefHasMemoryHotplug(vm->newDef)))
2037 2038
        cookieFlags |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

2039 2040 2041 2042 2043
    if (!qemuDomainVcpuHotplugIsInOrder(vm->def) ||
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
         vm->newDef && !qemuDomainVcpuHotplugIsInOrder(vm->newDef)))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU_HOTPLUG;

2044 2045 2046
    if (priv->origCPU)
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU;

2047 2048
    cookieFlags |= QEMU_MIGRATION_COOKIE_ALLOW_REBOOT;

2049 2050 2051
    if (!(flags & VIR_MIGRATE_OFFLINE))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CAPS;

2052 2053
    if (!(mig = qemuMigrationEatCookie(driver, vm->def,
                                       priv->origname, priv, NULL, 0, 0)))
2054 2055 2056
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
2057
                                QEMU_MIGRATION_SOURCE,
2058
                                cookieout, cookieoutlen,
2059
                                cookieFlags) < 0)
2060 2061
        goto cleanup;

L
liguang 已提交
2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083
    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;
        }
    }

2084
    if (xmlin) {
2085
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt, priv->qemuCaps,
2086 2087
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                            VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2088 2089
            goto cleanup;

2090
        if (!qemuDomainCheckABIStability(driver, vm, def))
2091 2092
            goto cleanup;

2093
        rv = qemuDomainDefFormatLive(driver, def, NULL, false, true);
2094
    } else {
2095 2096
        rv = qemuDomainDefFormatLive(driver, vm->def, priv->origCPU,
                                     false, true);
2097
    }
2098

2099
 cleanup:
2100
    qemuMigrationCookieFree(mig);
2101
    virObjectUnref(caps);
2102
    virDomainDefFree(def);
2103 2104 2105
    return rv;
}

2106
char *
2107 2108 2109 2110 2111 2112 2113 2114 2115
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)
2116 2117 2118
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
2119
    qemuDomainAsyncJob asyncJob;
2120 2121

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
2122 2123
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
2124 2125 2126 2127 2128 2129 2130 2131
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_MIGRATION_OUT;
    } else {
        if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_NONE;
    }

2132
    qemuMigrationSrcStoreDomainState(vm);
2133

2134
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
2135 2136 2137 2138 2139 2140
        goto endjob;

    /* Check if there is any ejected media.
     * We don't want to require them on the destination.
     */
    if (!(flags & VIR_MIGRATE_OFFLINE) &&
2141
        qemuProcessRefreshDisks(driver, vm, asyncJob) < 0)
2142 2143
        goto endjob;

2144 2145 2146
    if (!(xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                           cookieout, cookieoutlen,
                                           nmigrate_disks, migrate_disks, flags)))
2147 2148 2149 2150 2151 2152 2153
        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.
         */
2154
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
2155
                                 qemuMigrationSrcCleanup) < 0) {
2156
            VIR_FREE(xml);
2157
            goto endjob;
2158
        }
2159
        qemuMigrationJobContinue(vm);
2160 2161 2162 2163
    } else {
        goto endjob;
    }

2164
 cleanup:
M
Michal Privoznik 已提交
2165
    virDomainObjEndAPI(&vm);
2166 2167
    return xml;

2168
 endjob:
2169 2170 2171 2172
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2173 2174 2175
    goto cleanup;
}

2176

2177 2178
/* Prepare is the first step, and it runs on the destination host.
 */
2179

2180
static void
2181 2182
qemuMigrationDstPrepareCleanup(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
2183 2184 2185 2186 2187 2188 2189 2190 2191
{
    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));

2192
    virPortAllocatorRelease(priv->migrationPort);
2193 2194
    priv->migrationPort = 0;

2195 2196 2197 2198 2199
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2200
static qemuProcessIncomingDefPtr
2201 2202 2203 2204 2205 2206
qemuMigrationDstPrepare(virDomainObjPtr vm,
                        bool tunnel,
                        const char *protocol,
                        const char *listenAddress,
                        unsigned short port,
                        int fd)
2207 2208
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2209
    qemuProcessIncomingDefPtr inc = NULL;
2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239
    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 */
            }
2240
        } else if (hostIPv6Capable) {
2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263
            /* 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;
    }

2264 2265
    inc = qemuProcessIncomingDefNew(priv->qemuCaps, listenAddress,
                                    migrateFrom, fd, NULL);
2266

2267
 cleanup:
2268 2269
    VIR_FREE(migrateFrom);
    return inc;
2270 2271
}

2272
static int
2273
qemuMigrationDstPrepareAny(virQEMUDriverPtr driver,
2274
                           virConnectPtr dconn,
2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288
                           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,
2289
                           qemuMigrationParamsPtr migParams,
2290
                           unsigned long flags)
2291 2292
{
    virDomainObjPtr vm = NULL;
2293
    virObjectEventPtr event = NULL;
2294
    virErrorPtr origErr;
2295
    int ret = -1;
2296
    int dataFD[2] = { -1, -1 };
2297
    qemuDomainObjPrivatePtr priv = NULL;
2298
    qemuMigrationCookiePtr mig = NULL;
2299
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2300
    char *xmlout = NULL;
L
liguang 已提交
2301
    unsigned int cookieFlags;
2302
    unsigned int startFlags;
2303
    virCapsPtr caps = NULL;
2304
    qemuProcessIncomingDefPtr incoming = NULL;
2305
    bool taint_hook = false;
2306 2307 2308
    bool stopProcess = false;
    bool relabel = false;
    int rv;
2309
    char *tlsAlias = NULL;
2310

2311 2312
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332
    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;
        }
2333 2334
        cookieFlags = 0;
    } else {
2335 2336
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS |
                      QEMU_MIGRATION_COOKIE_CAPS;
L
liguang 已提交
2337 2338
    }

2339 2340 2341 2342 2343 2344 2345 2346 2347
    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;
    }

2348 2349 2350 2351 2352 2353
    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;
    }

2354 2355 2356
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2357
    if (!qemuMigrationSrcIsAllowedHostdev(*def))
2358 2359
        goto cleanup;

J
Jiri Denemark 已提交
2360 2361 2362 2363 2364
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2365
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2366 2367
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2368 2369
            goto cleanup;

2370
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2371 2372 2373 2374 2375 2376 2377
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2378
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2379 2380 2381 2382 2383 2384
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2385
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt, NULL,
2386 2387
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                                 VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE);
J
Jiri Denemark 已提交
2388 2389 2390
                if (!newdef)
                    goto cleanup;

2391
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2392 2393 2394 2395
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2396 2397
                virDomainDefFree(*def);
                *def = newdef;
2398 2399 2400 2401
                /* 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
2402 2403 2404 2405
            }
        }
    }

2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419
    /* 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;

2420
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2421
                                   driver->xmlopt,
2422 2423 2424
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2425
        goto cleanup;
2426
    *def = NULL;
2427

2428
    priv = vm->privateData;
2429 2430
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2431

2432 2433 2434 2435 2436
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2437 2438
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
2439 2440 2441 2442 2443 2444
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2445 2446 2447
    if (qemuMigrationDstPrecreateStorage(vm, mig->nbd,
                                         nmigrate_disks, migrate_disks,
                                         !!(flags & VIR_MIGRATE_NON_SHARED_INC)) < 0)
2448 2449
        goto cleanup;

2450 2451
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                              flags) < 0)
2452
        goto cleanup;
2453
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2454 2455 2456 2457

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

L
liguang 已提交
2458 2459 2460
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2461 2462
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
2463 2464
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
2465
        goto stopjob;
2466 2467
    }

2468 2469
    startFlags = VIR_QEMU_PROCESS_START_AUTODESTROY;

2470
    if (qemuProcessInit(driver, vm, mig->cpu, QEMU_ASYNC_JOB_MIGRATION_IN,
2471
                        true, startFlags) < 0)
2472
        goto stopjob;
2473
    stopProcess = true;
2474

2475 2476
    priv->allowReboot = mig->allowReboot;

2477 2478 2479
    if (!(incoming = qemuMigrationDstPrepare(vm, tunnel, protocol,
                                             listenAddress, port,
                                             dataFD[0])))
2480
        goto stopjob;
2481

2482
    if (qemuProcessPrepareDomain(driver, vm, startFlags) < 0)
2483 2484
        goto stopjob;

2485
    if (qemuProcessPrepareHost(driver, vm, startFlags) < 0)
2486 2487
        goto stopjob;

2488
    rv = qemuProcessLaunch(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2489 2490
                           incoming, NULL,
                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
2491
                           startFlags);
2492 2493 2494
    if (rv < 0) {
        if (rv == -2)
            relabel = true;
2495
        goto stopjob;
2496
    }
2497
    relabel = true;
2498

2499 2500 2501 2502
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
2503
            goto stopjob;
2504
        }
2505
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
2506 2507
    }

2508
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
2509
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
2510
        goto stopjob;
M
Michael R. Hines 已提交
2511 2512
    }

2513
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2514
                                 migParams, mig->caps->automatic) < 0)
2515 2516 2517 2518 2519 2520 2521
        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,
2522
                                         &tlsAlias, NULL,
2523 2524 2525 2526 2527 2528 2529
                                         migParams) < 0)
            goto stopjob;
    } else {
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
            goto stopjob;
    }

2530 2531
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                                 migParams) < 0)
2532 2533
        goto stopjob;

2534 2535 2536
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548
        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;
        }

2549 2550
        if (qemuMigrationDstStartNBDServer(driver, vm, incoming->address,
                                           nmigrate_disks, migrate_disks,
2551
                                           nbdPort, nbdTLSAlias) < 0) {
2552
            goto stopjob;
2553
        }
2554
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
2555 2556
    }

2557 2558 2559 2560 2561 2562 2563 2564 2565
    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");
    }

2566
    if (incoming->deferredURI &&
2567 2568
        qemuMigrationDstRun(driver, vm, incoming->deferredURI,
                            QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2569 2570
        goto stopjob;

2571
    if (qemuProcessFinishStartup(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2572 2573 2574
                                 false, VIR_DOMAIN_PAUSED_MIGRATION) < 0)
        goto stopjob;

2575
 done:
2576 2577 2578
    if (qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_DESTINATION,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
2579 2580 2581 2582 2583 2584 2585
        /* 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");
    }

2586
    if (qemuDomainCleanupAdd(vm, qemuMigrationDstPrepareCleanup) < 0)
2587
        goto stopjob;
2588

L
liguang 已提交
2589 2590
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
2591
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
2592 2593 2594
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
2595

2596 2597 2598 2599
    /* 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.
     */
2600
    qemuMigrationJobContinue(vm);
2601

2602 2603
    if (autoPort)
        priv->migrationPort = port;
2604 2605 2606 2607 2608
    /* 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;
2609
    ret = 0;
2610

2611
 cleanup:
2612
    virErrorPreserveLast(&origErr);
2613
    VIR_FREE(tlsAlias);
2614
    qemuProcessIncomingDefFree(incoming);
J
Jiri Denemark 已提交
2615
    VIR_FREE(xmlout);
2616 2617
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
2618 2619 2620
    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 */
2621
        VIR_FREE(priv->origname);
2622 2623 2624 2625
        /* release if port is auto selected which is not the case if
         * it is given in parameters
         */
        if (nbdPort == 0)
2626
            virPortAllocatorRelease(priv->nbdPort);
2627
        priv->nbdPort = 0;
2628
        virDomainObjRemoveTransientDef(vm);
2629
        qemuDomainRemoveInactiveJob(driver, vm);
2630
    }
M
Michal Privoznik 已提交
2631
    virDomainObjEndAPI(&vm);
2632
    virObjectEventStateQueue(driver->domainEventState, event);
2633
    qemuMigrationCookieFree(mig);
2634
    virObjectUnref(caps);
2635
    virNWFilterUnlockFilterUpdates();
2636
    virErrorRestore(&origErr);
2637
    return ret;
2638

2639
 stopjob:
2640
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2641
                             priv->job.migParams, priv->job.apiFlags);
2642

2643 2644 2645 2646
    if (stopProcess) {
        unsigned int stopFlags = VIR_QEMU_PROCESS_STOP_MIGRATED;
        if (!relabel)
            stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
2647
        virDomainAuditStart(vm, "migrated", false);
2648 2649
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        QEMU_ASYNC_JOB_MIGRATION_IN, stopFlags);
2650
    }
2651

2652
    qemuMigrationJobFinish(driver, vm);
2653
    goto cleanup;
2654 2655 2656
}


2657 2658 2659 2660 2661
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
2662
qemuMigrationDstPrepareTunnel(virQEMUDriverPtr driver,
2663
                              virConnectPtr dconn,
2664 2665 2666 2667 2668 2669 2670
                              const char *cookiein,
                              int cookieinlen,
                              char **cookieout,
                              int *cookieoutlen,
                              virStreamPtr st,
                              virDomainDefPtr *def,
                              const char *origname,
2671
                              qemuMigrationParamsPtr migParams,
2672
                              unsigned long flags)
2673
{
2674
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2675
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
2676
              "origname=%s, flags=0x%lx",
2677
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2678
              cookieout, cookieoutlen, st, *def, origname, flags);
2679

2680 2681 2682 2683 2684 2685
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

2686 2687 2688
    return qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
                                      cookieout, cookieoutlen, def, origname,
                                      st, NULL, 0, false, NULL, 0, NULL, 0,
2689
                                      migParams, flags);
2690 2691 2692
}


2693
static virURIPtr
2694
qemuMigrationAnyParseURI(const char *uri, bool *wellFormed)
2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715
{
    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;
}


2716
int
2717
qemuMigrationDstPrepareDirect(virQEMUDriverPtr driver,
2718
                              virConnectPtr dconn,
2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730
                              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,
2731
                              qemuMigrationParamsPtr migParams,
2732
                              unsigned long flags)
2733
{
2734 2735
    unsigned short port = 0;
    bool autoPort = true;
2736 2737
    char *hostname = NULL;
    int ret = -1;
2738
    virURIPtr uri = NULL;
2739 2740
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
2741

2742
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2743
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
2744
              "def=%p, origname=%s, listenAddress=%s, "
2745
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, flags=0x%lx",
2746
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2747
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
2748
              *def, origname, NULLSTR(listenAddress),
2749
              nmigrate_disks, migrate_disks, nbdPort, flags);
2750

2751 2752
    *uri_out = NULL;

2753 2754 2755
    /* 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
2756 2757 2758
     * 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".
2759 2760 2761 2762 2763 2764
     *
     * 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) {
2765 2766 2767
        bool encloseAddress = false;
        const char *incFormat;

2768
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2769
            goto cleanup;
2770

2771
        if (migrateHost != NULL) {
2772 2773
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
2774

2775
            if (VIR_STRDUP(hostname, migrateHost) < 0)
2776 2777 2778 2779 2780
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
2781 2782

        if (STRPREFIX(hostname, "localhost")) {
2783 2784 2785
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
2786 2787 2788 2789 2790
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
2791
         * compatibility with old targets. We at least make the
2792 2793
         * new targets accept both syntaxes though.
         */
2794 2795 2796 2797 2798 2799
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
2800 2801
            goto cleanup;
    } else {
2802
        bool well_formed_uri;
J
Ján Tomko 已提交
2803

2804
        if (!(uri = qemuMigrationAnyParseURI(uri_in, &well_formed_uri)))
2805
            goto cleanup;
J
Ján Tomko 已提交
2806

2807 2808 2809 2810 2811 2812 2813
        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
2814 2815
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
2816 2817 2818
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
2819 2820 2821 2822 2823 2824 2825 2826 2827 2828
            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) {
2829
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2830
                goto cleanup;
2831

2832
            /* Send well-formed URI only if uri_in was well-formed */
2833 2834 2835 2836 2837 2838 2839 2840
            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;
            }
2841
        } else {
2842 2843
            port = uri->port;
            autoPort = false;
2844 2845 2846 2847 2848 2849
        }
    }

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

2850
    ret = qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
2851 2852 2853 2854
                                     cookieout, cookieoutlen, def, origname,
                                     NULL, uri ? uri->scheme : "tcp",
                                     port, autoPort, listenAddress,
                                     nmigrate_disks, migrate_disks, nbdPort,
2855
                                     migParams, flags);
2856
 cleanup:
2857
    virURIFree(uri);
2858
    VIR_FREE(hostname);
2859
    virObjectUnref(cfg);
2860
    if (ret != 0) {
2861
        VIR_FREE(*uri_out);
2862
        if (autoPort)
2863
            virPortAllocatorRelease(port);
2864
    }
2865 2866 2867 2868
    return ret;
}


2869
virDomainDefPtr
2870 2871 2872 2873
qemuMigrationAnyPrepareDef(virQEMUDriverPtr driver,
                           const char *dom_xml,
                           const char *dname,
                           char **origname)
2874 2875 2876
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
2877
    char *name = NULL;
2878 2879 2880 2881 2882 2883 2884 2885 2886 2887

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

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

2888
    if (!(def = virDomainDefParseString(dom_xml, caps, driver->xmlopt, NULL,
2889 2890
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                        VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2891 2892 2893
        goto cleanup;

    if (dname) {
2894
        name = def->name;
2895 2896 2897 2898 2899 2900
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

2901
 cleanup:
2902
    virObjectUnref(caps);
2903 2904 2905 2906
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
2907 2908 2909 2910
    return def;
}


2911
static int
2912 2913 2914 2915 2916 2917
qemuMigrationSrcConfirmPhase(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
                             const char *cookiein,
                             int cookieinlen,
                             unsigned int flags,
                             int retcode)
2918 2919
{
    qemuMigrationCookiePtr mig;
2920
    virObjectEventPtr event;
2921 2922
    int rv = -1;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2923 2924
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = NULL;
2925

2926
    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
2927
              "flags=0x%x, retcode=%d",
2928
              driver, vm, NULLSTR(cookiein), cookieinlen,
2929 2930 2931 2932 2933 2934 2935 2936 2937
              flags, retcode);

    virCheckFlags(QEMU_MIGRATION_FLAGS, -1);

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

2938 2939
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen,
2940
                                       QEMU_MIGRATION_COOKIE_STATS)))
2941 2942
        goto cleanup;

2943 2944 2945
    if (retcode == 0)
        jobInfo = priv->job.completed;
    else
2946
        VIR_FREE(priv->job.completed);
2947 2948 2949

    /* Update times with the values sent by the destination daemon */
    if (mig->jobInfo && jobInfo) {
2950 2951 2952 2953 2954 2955 2956 2957
        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 &&
2958 2959
            qemuMigrationAnyFetchStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                       jobInfo, NULL) < 0)
2960 2961
            VIR_WARN("Could not refresh migration statistics");

2962 2963 2964
        qemuDomainJobInfoUpdateTime(jobInfo);
        jobInfo->timeDeltaSet = mig->jobInfo->timeDeltaSet;
        jobInfo->timeDelta = mig->jobInfo->timeDelta;
2965 2966
        jobInfo->stats.mig.downtime_set = mig->jobInfo->stats.mig.downtime_set;
        jobInfo->stats.mig.downtime = mig->jobInfo->stats.mig.downtime;
2967 2968
    }

2969 2970 2971
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2972 2973
    /* 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.
2974 2975 2976 2977
     */
    if (retcode == 0) {
        /* If guest uses SPICE and supports seamless migration we have to hold
         * up domain shutdown until SPICE server transfers its data */
2978
        qemuMigrationSrcWaitForSpice(vm);
2979 2980

        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
2981
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
2982 2983 2984
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
        virDomainAuditStop(vm, "migrated");

2985
        event = virDomainEventLifecycleNewFromObj(vm,
2986 2987
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
2988
        virObjectEventStateQueue(driver->domainEventState, event);
2989
        qemuDomainEventEmitJobCompleted(driver, vm);
2990
    } else {
2991
        virErrorPtr orig_err = virSaveLastError();
2992
        int reason;
2993 2994

        /* cancel any outstanding NBD jobs */
2995 2996
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT, NULL);
2997 2998 2999

        virSetError(orig_err);
        virFreeError(orig_err);
3000

3001
        if (virDomainObjGetState(vm, &reason) == VIR_DOMAIN_PAUSED &&
3002
            reason == VIR_DOMAIN_PAUSED_POSTCOPY)
3003
            qemuMigrationAnyPostcopyFailed(driver, vm);
3004 3005
        else
            qemuMigrationSrcRestoreDomainState(driver, vm);
3006

3007
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3008
                                 priv->job.migParams, priv->job.apiFlags);
3009

3010
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
3011 3012 3013
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
    }

3014
 done:
3015 3016 3017
    qemuMigrationCookieFree(mig);
    rv = 0;

3018
 cleanup:
3019 3020 3021 3022 3023
    virObjectUnref(cfg);
    return rv;
}

int
3024 3025 3026 3027 3028 3029
qemuMigrationSrcConfirm(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *cookiein,
                        int cookieinlen,
                        unsigned int flags,
                        int cancelled)
3030
{
3031
    qemuMigrationJobPhase phase;
3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045
    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);
3046
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
3047
                           qemuMigrationSrcCleanup);
3048

3049 3050 3051
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
3052

3053
    qemuMigrationJobFinish(driver, vm);
3054
    if (!virDomainObjIsActive(vm)) {
3055
        if (!cancelled && ret == 0 && flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
3056
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
3057 3058
            vm->persistent = 0;
        }
3059
        qemuDomainRemoveInactiveJob(driver, vm);
3060 3061
    }

3062
 cleanup:
M
Michal Privoznik 已提交
3063
    virDomainObjEndAPI(&vm);
3064 3065 3066 3067 3068
    virObjectUnref(cfg);
    return ret;
}


3069 3070
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3071
    MIGRATION_DEST_CONNECT_HOST,
3072
    MIGRATION_DEST_FD,
3073
};
3074

3075 3076 3077 3078
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3079

3080 3081 3082 3083 3084 3085
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3086
            const char *protocol;
3087 3088 3089 3090
            const char *name;
            int port;
        } host;

3091 3092 3093 3094
        struct {
            int qemu;
            int local;
        } fd;
3095 3096 3097 3098 3099 3100 3101
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3102 3103 3104

#define TUNNEL_SEND_BUF_SIZE 65536

3105 3106 3107 3108 3109 3110 3111
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3112 3113
    int wakeupRecvFD;
    int wakeupSendFD;
3114 3115
};

3116
static void qemuMigrationSrcIOFunc(void *arg)
3117
{
3118
    qemuMigrationIOThreadPtr data = arg;
3119 3120 3121 3122 3123 3124 3125
    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);
3126

3127
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3128
        goto abrt;
3129

3130 3131 3132
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3133
    for (;;) {
3134 3135 3136 3137 3138 3139 3140 3141 3142 3143
        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;
3144
            virReportSystemError(errno, "%s",
3145 3146
                                 _("poll failed in migration tunnel"));
            goto abrt;
3147
        }
3148 3149 3150 3151 3152 3153 3154

        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");
3155
            break;
3156
        }
3157

3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173
        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;
            }
3174 3175
        }

3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192
        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;
            }
        }
    }
3193

3194 3195
    if (virStreamFinish(data->st) < 0)
        goto error;
3196

3197
    VIR_FORCE_CLOSE(data->sock);
3198 3199
    VIR_FREE(buffer);

3200 3201
    return;

3202
 abrt:
3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3214
 error:
3215 3216 3217 3218 3219
    /* 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);
3220
    virResetLastError();
3221
    VIR_FREE(buffer);
3222 3223 3224 3225
}


static qemuMigrationIOThreadPtr
3226 3227
qemuMigrationSrcStartTunnel(virStreamPtr st,
                            int sock)
3228
{
3229 3230
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3231

3232 3233 3234 3235
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3236 3237
    }

3238
    if (VIR_ALLOC(io) < 0)
3239
        goto error;
3240

3241 3242
    io->st = st;
    io->sock = sock;
3243 3244
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3245 3246

    if (virThreadCreate(&io->thread, true,
3247
                        qemuMigrationSrcIOFunc,
3248 3249 3250
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3251
        goto error;
3252 3253 3254
    }

    return io;
3255

3256
 error:
3257 3258 3259 3260
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3261 3262 3263
}

static int
3264
qemuMigrationSrcStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3265 3266
{
    int rv = -1;
3267 3268 3269 3270 3271 3272 3273 3274 3275
    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;
    }

3276 3277 3278 3279
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3280 3281 3282 3283
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3284 3285 3286 3287 3288 3289
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3290
 cleanup:
3291 3292
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3293 3294
    VIR_FREE(io);
    return rv;
3295 3296
}

3297
static int
3298 3299 3300
qemuMigrationSrcConnect(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationSpecPtr spec)
3301 3302 3303 3304 3305 3306 3307
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3308
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3309 3310 3311 3312 3313
        return -1;

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

3314
    if (qemuSecuritySetSocketLabel(driver->securityManager, vm->def) < 0)
3315
        goto cleanup;
3316 3317 3318
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
3319
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
3320
        virObjectUnref(sock);
3321
    }
3322
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0 ||
3323 3324 3325
        spec->dest.fd.qemu == -1)
        goto cleanup;

3326 3327 3328 3329 3330 3331 3332
    /* 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;
    }

3333 3334
    ret = 0;

3335
 cleanup:
3336 3337 3338 3339 3340 3341
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3342 3343

static int
3344 3345 3346 3347
qemuMigrationSrcContinue(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMonitorMigrationStatus status,
                         qemuDomainAsyncJob asyncJob)
3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363
{
    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;
}


3364
static int
3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378
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,
3379
                    qemuMigrationParamsPtr migParams)
3380
{
3381
    int ret = -1;
3382 3383
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3384
    qemuMigrationCookiePtr mig = NULL;
3385
    char *tlsAlias = NULL;
3386
    qemuMigrationIOThreadPtr iothread = NULL;
3387
    int fd = -1;
3388
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3389
    virErrorPtr orig_err = NULL;
3390
    unsigned int cookieFlags = 0;
3391
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3392
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
3393
    bool cancel = false;
3394
    unsigned int waitFlags;
3395
    virDomainDefPtr persistDef = NULL;
3396
    char *timestamp;
3397
    int rc;
3398 3399

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
3400
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3401 3402
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
3403 3404
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3405
              spec, spec->destType, spec->fwdType, dconn,
3406
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3407

3408 3409 3410 3411 3412 3413 3414 3415 3416 3417
    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;
    }

3418 3419
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3420 3421 3422 3423
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3424 3425 3426
        return -1;
    }

3427 3428 3429
    if (events)
        priv->signalIOError = abort_on_error;

3430 3431
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
        if (persist_xml) {
3432 3433
            if (!(persistDef = qemuMigrationAnyPrepareDef(driver, persist_xml,
                                                          NULL, NULL)))
3434
                goto error;
3435 3436 3437
        } else {
            virDomainDefPtr def = vm->newDef ? vm->newDef : vm->def;
            if (!(persistDef = qemuDomainDefCopy(driver, def,
3438 3439
                                                 VIR_DOMAIN_XML_SECURE |
                                                 VIR_DOMAIN_XML_MIGRATABLE)))
3440
                goto error;
3441 3442 3443
        }
    }

3444 3445
    mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                 cookiein, cookieinlen,
3446 3447 3448
                                 cookieFlags |
                                 QEMU_MIGRATION_COOKIE_GRAPHICS |
                                 QEMU_MIGRATION_COOKIE_CAPS);
3449
    if (!mig)
3450
        goto error;
3451

3452
    if (qemuMigrationSrcGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3453 3454
        VIR_WARN("unable to provide data for graphics client relocation");

3455
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3456
                                 migParams, mig->caps->automatic) < 0)
3457 3458
        goto error;

3459
    if (flags & VIR_MIGRATE_TLS) {
3460
        const char *hostname = NULL;
3461 3462 3463 3464

        /* We need to add tls-hostname whenever QEMU itself does not
         * connect directly to the destination. */
        if (spec->destType == MIGRATION_DEST_CONNECT_HOST ||
3465 3466 3467
            spec->destType == MIGRATION_DEST_FD)
            hostname = spec->dest.host.name;

3468
        if (qemuMigrationParamsEnableTLS(driver, vm, false,
3469
                                         QEMU_ASYNC_JOB_MIGRATION_OUT,
3470
                                         &tlsAlias, hostname,
3471 3472
                                         migParams) < 0)
            goto error;
3473
    } else {
3474
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
3475
            goto error;
3476 3477
    }

3478 3479 3480 3481
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                 migParams) < 0)
        goto error;

3482 3483 3484
    if (migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                         QEMU_MONITOR_MIGRATE_NON_SHARED_INC)) {
        if (mig->nbd) {
3485 3486 3487 3488
            /* 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. */
3489 3490
            if (flags & VIR_MIGRATE_TLS &&
                !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV_DEL)) {
3491 3492 3493 3494 3495
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                               _("NBD migration with TLS is not supported"));
                goto error;
            }

3496
            /* This will update migrate_flags on success */
3497 3498 3499 3500 3501 3502
            if (qemuMigrationSrcNBDStorageCopy(driver, vm, mig,
                                               spec->dest.host.name,
                                               migrate_speed,
                                               &migrate_flags,
                                               nmigrate_disks,
                                               migrate_disks,
3503
                                               dconn, tlsAlias, flags) < 0) {
3504
                goto error;
3505 3506 3507 3508 3509 3510 3511
            }
        } 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.");
        }
3512 3513
    }

3514
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3515 3516
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
3517
        if (qemuMigrationSrcSetOffline(driver, vm) < 0)
3518
            goto error;
3519 3520
    }

3521 3522
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3523
        goto error;
3524

3525
    if (priv->job.abortJob) {
3526 3527
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
3528
         * priv->job.abortJob will not change */
3529
        priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
3530 3531 3532
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
3533
        goto exit_monitor;
3534 3535
    }

3536 3537
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
3538

3539 3540
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3541
        qemuMigrationSrcConnect(driver, vm, spec) < 0) {
3542
        goto exit_monitor;
3543
    }
3544

3545 3546 3547 3548 3549 3550
    /* log start of migration */
    if ((timestamp = virTimeStringNow()) != NULL) {
        qemuDomainLogAppendMessage(driver, vm, "%s: initiating migration\n", timestamp);
        VIR_FREE(timestamp);
    }

3551
    rc = -1;
3552 3553
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3554 3555
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
3556
            goto exit_monitor;
M
Michael R. Hines 已提交
3557
        }
3558 3559 3560 3561
        rc = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
                                      spec->dest.host.protocol,
                                      spec->dest.host.name,
                                      spec->dest.host.port);
3562 3563
        break;

3564 3565 3566 3567
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3568
    case MIGRATION_DEST_FD:
3569
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3570
            fd = spec->dest.fd.local;
3571 3572
            spec->dest.fd.local = -1;
        }
3573 3574
        rc = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                    spec->dest.fd.qemu);
3575 3576
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3577
    }
3578 3579

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
3580
        goto error;
3581 3582 3583

    /* From this point onwards we *must* call cancel to abort the
     * migration on source if anything goes wrong */
3584
    cancel = true;
3585

3586
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3587
        if (!(iothread = qemuMigrationSrcStartTunnel(spec->fwd.stream, fd)))
3588
            goto error;
3589 3590 3591 3592 3593
        /* If we've created a tunnel, then the 'fd' will be closed in the
         * qemuMigrationIOFunc as data->sock.
         */
        fd = -1;
    }
3594

3595
    waitFlags = QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;
3596 3597 3598 3599 3600 3601 3602
    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;

3603 3604 3605
    rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT,
                                           dconn, waitFlags);
3606
    if (rc == -2) {
3607
        goto error;
3608 3609 3610 3611 3612
    } else if (rc == -1) {
        /* QEMU reported failed migration, nothing to cancel anymore */
        cancel = false;
        goto error;
    }
3613

3614 3615 3616
    /* 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.
3617
     */
3618 3619 3620 3621 3622
    if (priv->monJSON) {
        while (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            priv->signalStop = true;
            rc = virDomainObjWait(vm);
            priv->signalStop = false;
3623
            if (rc < 0)
3624
                goto error;
3625
        }
3626
    } else if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING &&
3627
               qemuMigrationSrcSetOffline(driver, vm) < 0) {
3628
        goto error;
3629
    }
3630

J
Ján Tomko 已提交
3631
    if (mig->nbd &&
3632 3633 3634
        qemuMigrationSrcNBDCopyCancel(driver, vm, true,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT,
                                      dconn) < 0)
3635
        goto error;
3636

3637 3638 3639 3640 3641
    /* 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) {
3642 3643 3644
        if (qemuMigrationSrcContinue(driver, vm,
                                     QEMU_MONITOR_MIGRATION_STATUS_PRE_SWITCHOVER,
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3645 3646 3647 3648
            goto error;

        waitFlags ^= QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;

3649 3650 3651
        rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_OUT,
                                               dconn, waitFlags);
3652 3653 3654 3655 3656 3657 3658 3659 3660
        if (rc == -2) {
            goto error;
        } else if (rc == -1) {
            /* QEMU reported failed migration, nothing to cancel anymore */
            cancel = false;
            goto error;
        }
    }

3661 3662 3663 3664
    if (iothread) {
        qemuMigrationIOThreadPtr io;

        VIR_STEAL_PTR(io, iothread);
3665
        if (qemuMigrationSrcStopTunnel(io, false) < 0)
3666
            goto error;
3667 3668 3669
    }

    if (priv->job.completed) {
3670
        priv->job.completed->stopped = priv->job.current->stopped;
3671 3672 3673 3674 3675 3676 3677 3678 3679
        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 ||
3680 3681 3682
        qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_SOURCE,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
3683 3684
        VIR_WARN("Unable to encode migration cookie");
    }
3685

3686
    ret = 0;
3687

3688
 cleanup:
3689
    VIR_FREE(tlsAlias);
3690
    VIR_FORCE_CLOSE(fd);
3691
    virDomainDefFree(persistDef);
3692 3693
    qemuMigrationCookieFree(mig);

3694 3695 3696
    if (events)
        priv->signalIOError = false;

3697 3698 3699 3700 3701
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

3702 3703
    return ret;

3704
 error:
3705 3706
    orig_err = virSaveLastError();

3707 3708 3709 3710 3711 3712 3713 3714
    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));
        }
3715

3716 3717
        /* cancel any outstanding NBD jobs */
        if (mig && mig->nbd)
3718 3719 3720
            qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                          QEMU_ASYNC_JOB_MIGRATION_OUT,
                                          dconn);
3721 3722 3723 3724

        if (priv->job.current->status != QEMU_DOMAIN_JOB_STATUS_CANCELED)
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
    }
3725 3726

    if (iothread)
3727
        qemuMigrationSrcStopTunnel(iothread, true);
3728 3729 3730

    goto cleanup;

3731 3732
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
3733
    goto error;
3734 3735
}

3736
/* Perform migration using QEMU's native migrate support,
3737 3738
 * not encrypted obviously
 */
3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753
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,
3754
                              qemuMigrationParamsPtr migParams)
3755
{
3756
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
3757
    virURIPtr uribits = NULL;
3758
    int ret = -1;
3759 3760 3761
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
3762
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3763
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
3764
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
3765
              cookieout, cookieoutlen, flags, resource,
3766
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3767

3768
    if (!(uribits = qemuMigrationAnyParseURI(uri, NULL)))
3769 3770
        return -1;

3771 3772 3773 3774 3775 3776 3777
    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
3778 3779 3780 3781 3782
    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"));
3783
            goto cleanup;
M
Michael R. Hines 已提交
3784
        }
3785
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
3786 3787 3788
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
3789
            goto cleanup;
M
Michael R. Hines 已提交
3790 3791 3792
        }
    }

3793
    if (STRNEQ(uribits->scheme, "rdma"))
3794 3795
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
3796
        spec.destType = MIGRATION_DEST_HOST;
3797
    spec.dest.host.protocol = uribits->scheme;
3798 3799 3800
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
3801

3802 3803 3804
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen, cookieout,
                              cookieoutlen, flags, resource, &spec, dconn,
                              graphicsuri, nmigrate_disks, migrate_disks,
3805
                              migParams);
3806 3807 3808 3809

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

3810
 cleanup:
3811
    virURIFree(uribits);
3812 3813 3814 3815 3816

    return ret;
}


3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831
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,
3832
                              qemuMigrationParamsPtr migParams)
3833 3834 3835
{
    int ret = -1;
    qemuMigrationSpec spec;
3836
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3837
    int fds[2] = { -1, -1 };
3838 3839

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
3840
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3841
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
3842
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
3843
              cookieout, cookieoutlen, flags, resource,
3844
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3845 3846 3847 3848

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

3849

3850 3851 3852
    spec.destType = MIGRATION_DEST_FD;
    spec.dest.fd.qemu = -1;
    spec.dest.fd.local = -1;
3853

3854 3855 3856 3857 3858
    if (pipe2(fds, O_CLOEXEC) == 0) {
        spec.dest.fd.qemu = fds[1];
        spec.dest.fd.local = fds[0];
    }
    if (spec.dest.fd.qemu == -1 ||
3859 3860
        qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    spec.dest.fd.qemu) < 0) {
3861 3862 3863
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
        goto cleanup;
3864 3865
    }

3866 3867 3868
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen,
                              cookieout, cookieoutlen, flags, resource, &spec,
                              dconn, graphicsuri, nmigrate_disks, migrate_disks,
3869
                              migParams);
3870

3871
 cleanup:
3872 3873
    VIR_FORCE_CLOSE(spec.dest.fd.qemu);
    VIR_FORCE_CLOSE(spec.dest.fd.local);
3874

3875
    virObjectUnref(cfg);
3876 3877 3878 3879
    return ret;
}


3880 3881 3882 3883
/* 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 */
3884 3885 3886 3887 3888 3889 3890 3891
static int
qemuMigrationSrcPerformPeer2Peer2(virQEMUDriverPtr driver,
                                  virConnectPtr sconn,
                                  virConnectPtr dconn,
                                  virDomainObjPtr vm,
                                  const char *dconnuri,
                                  unsigned long flags,
                                  const char *dname,
3892 3893
                                  unsigned long resource,
                                  qemuMigrationParamsPtr migParams)
3894 3895 3896
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
3897
    char *cookie = NULL;
3898 3899 3900
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
3901
    bool cancelled;
3902
    virStreamPtr st = NULL;
3903
    unsigned long destflags;
3904

3905
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
3906
              "flags=0x%lx, dname=%s, resource=%lu",
3907 3908
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
3909

3910 3911 3912 3913 3914
    /* 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,
3915 3916
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
3917 3918 3919 3920 3921
        return -1;

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

3922 3923
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
3924

3925 3926 3927 3928 3929 3930 3931 3932 3933 3934
    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;

3935
        qemuDomainObjEnterRemote(vm);
3936
        ret = dconn->driver->domainMigratePrepareTunnel
3937
            (dconn, st, destflags, dname, resource, dom_xml);
3938 3939
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3940
    } else {
3941
        qemuDomainObjEnterRemote(vm);
3942 3943
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
3944
             destflags, dname, resource, dom_xml);
3945 3946
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3947 3948 3949
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
3950 3951
        goto cleanup;

3952 3953
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
3954 3955
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
3956
        cancelled = true;
3957
        orig_err = virSaveLastError();
3958
        goto finish;
3959 3960
    }

3961 3962 3963 3964
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
3965
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
3966
    if (flags & VIR_MIGRATE_TUNNELLED)
3967 3968 3969
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, NULL,
                                            NULL, 0, NULL, NULL,
                                            flags, resource, dconn,
3970
                                            NULL, 0, NULL, migParams);
3971
    else
3972 3973 3974 3975
        ret = qemuMigrationSrcPerformNative(driver, vm, NULL, uri_out,
                                            cookie, cookielen,
                                            NULL, NULL, /* No out cookie with v2 migration */
                                            flags, resource, dconn, NULL, 0, NULL,
3976
                                            migParams);
3977 3978 3979 3980

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
3981

3982 3983 3984
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
3985
    cancelled = ret < 0;
3986

3987
 finish:
3988 3989 3990 3991
    /* 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.
     */
3992
    dname = dname ? dname : vm->def->name;
3993
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
3994
    qemuDomainObjEnterRemote(vm);
3995
    ddomain = dconn->driver->domainMigrateFinish2
3996
        (dconn, dname, cookie, cookielen,
3997
         uri_out ? uri_out : dconnuri, destflags, cancelled);
3998 3999
    /* The domain is already gone at this point */
    ignore_value(qemuDomainObjExitRemote(vm, false));
4000 4001
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4002

4003
 cleanup:
4004
    if (ddomain) {
4005
        virObjectUnref(ddomain);
4006 4007 4008 4009
        ret = 0;
    } else {
        ret = -1;
    }
4010

4011
    virObjectUnref(st);
4012 4013 4014 4015 4016 4017

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4018
    VIR_FREE(cookie);
4019 4020

    return ret;
4021 4022 4023
}


4024 4025 4026 4027
/* 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 */
4028
static int
4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042
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,
4043
                                  qemuMigrationParamsPtr migParams,
4044 4045 4046
                                  unsigned long long bandwidth,
                                  bool useParams,
                                  unsigned long flags)
4047 4048 4049 4050 4051 4052 4053 4054 4055 4056
{
    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;
4057
    bool cancelled = true;
4058
    virStreamPtr st = NULL;
4059
    unsigned long destflags;
4060 4061 4062
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4063
    size_t i;
4064
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4065 4066

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4067
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4068
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4069
              "bandwidth=%llu, useParams=%d, flags=0x%lx",
4070
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4071
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4072
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks, nbdPort,
4073
              bandwidth, useParams, flags);
4074

4075 4076 4077 4078 4079
    /* 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.  */

4080 4081 4082
    dom_xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                         &cookieout, &cookieoutlen,
                                         nmigrate_disks, migrate_disks, flags);
4083 4084 4085
    if (!dom_xml)
        goto cleanup;

4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105
    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;
4106 4107 4108 4109 4110 4111

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4112 4113 4114 4115 4116
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4117 4118 4119 4120 4121
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4122 4123 4124 4125 4126
        if (nbdPort &&
            virTypedParamsAddInt(&params, &nparams, &maxparams,
                                 VIR_MIGRATE_PARAM_DISKS_PORT,
                                 nbdPort) < 0)
            goto cleanup;
4127

4128 4129
        if (qemuMigrationParamsDump(migParams, &params, &nparams,
                                    &maxparams, &flags) < 0)
4130
            goto cleanup;
4131 4132
    }

4133 4134 4135
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4136 4137
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4138

4139 4140 4141 4142 4143 4144 4145 4146 4147
    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;

4148
        qemuDomainObjEnterRemote(vm);
4149 4150 4151 4152 4153 4154 4155 4156 4157
        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);
        }
4158 4159
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4160
    } else {
4161
        qemuDomainObjEnterRemote(vm);
4162 4163 4164 4165 4166 4167 4168 4169 4170
        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);
        }
4171 4172
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4173 4174 4175 4176 4177
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

4178
    if (offline) {
L
liguang 已提交
4179 4180 4181
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4182
        cancelled = false;
L
liguang 已提交
4183 4184 4185
        goto finish;
    }

4186 4187 4188 4189
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4190 4191
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4192
            goto finish;
4193
        }
4194
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4195 4196
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4197
        orig_err = virSaveLastError();
4198 4199 4200 4201 4202 4203 4204 4205
        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.
     */
4206
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4207
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4208 4209 4210 4211 4212
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4213
    if (flags & VIR_MIGRATE_TUNNELLED) {
4214 4215 4216 4217
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, persist_xml,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4218
                                            nmigrate_disks, migrate_disks,
4219
                                            migParams);
4220
    } else {
4221 4222 4223 4224
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4225
                                            nmigrate_disks, migrate_disks,
4226
                                            migParams);
4227
    }
4228 4229

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4230
    if (ret < 0) {
4231
        orig_err = virSaveLastError();
4232 4233 4234 4235
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4236 4237 4238 4239

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4240
    cancelled = ret < 0;
4241

4242
 finish:
4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254
    /*
     * 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;
4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267

    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);
4268 4269
            if (qemuDomainObjExitRemote(vm, !offline) < 0)
                goto cleanup;
4270 4271 4272 4273 4274 4275 4276
        }
    } else {
        dname = dname ? dname : vm->def->name;
        qemuDomainObjEnterRemote(vm);
        ddomain = dconn->driver->domainMigrateFinish3
            (dconn, dname, cookiein, cookieinlen, &cookieout, &cookieoutlen,
             dconnuri, uri, destflags, cancelled);
4277 4278
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4279
    }
4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299

    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();
4300 4301
                if (err &&
                    err->domain == VIR_FROM_QEMU &&
4302 4303 4304 4305 4306 4307 4308
                    err->code != VIR_ERR_MIGRATE_FINISH_OK) {
                    virFreeError(orig_err);
                    orig_err = NULL;
                }
            }
        }
    }
4309

4310 4311 4312 4313 4314 4315 4316
    /* 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.
4317
     */
4318
    cancelled = ddomain == NULL;
4319

4320 4321 4322 4323 4324 4325
    /* 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();

4326 4327 4328 4329
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4330
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4331 4332 4333 4334 4335
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4336 4337 4338
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
4339 4340 4341 4342
    /* 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.
     */
4343 4344 4345
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4346 4347 4348

 cleanup:
    if (ddomain) {
4349
        virObjectUnref(ddomain);
4350 4351 4352 4353 4354
        ret = 0;
    } else {
        ret = -1;
    }

4355
    virObjectUnref(st);
4356 4357 4358 4359 4360 4361 4362 4363

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4364
    virTypedParamsFree(params, nparams);
4365 4366 4367 4368
    return ret;
}


4369
static void
4370 4371 4372
qemuMigrationSrcConnectionClosed(virConnectPtr conn,
                                 int reason,
                                 void *opaque)
4373 4374 4375 4376 4377 4378 4379 4380
{
    virDomainObjPtr vm = opaque;

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


4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405
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,
4406
                                 qemuMigrationParamsPtr migParams,
4407 4408 4409 4410
                                 unsigned long flags,
                                 const char *dname,
                                 unsigned long resource,
                                 bool *v3proto)
4411 4412 4413 4414
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4415
    virErrorPtr orig_err = NULL;
4416
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4417
    bool dstOffline = false;
4418
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4419
    bool useParams;
4420

4421 4422
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, uri=%s, "
              "graphicsuri=%s, listenAddress=%s, nmigrate_disks=%zu, "
4423
              "migrate_disks=%p, nbdPort=%d, flags=0x%lx, dname=%s, "
4424
              "resource=%lu",
4425
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4426
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4427 4428
              nmigrate_disks, migrate_disks, nbdPort, flags, NULLSTR(dname),
              resource);
4429

4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443
    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;
    }

4444 4445 4446 4447 4448 4449 4450
    if (flags & VIR_MIGRATE_TUNNELLED && nbdPort) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("disk port address is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

4451 4452 4453 4454
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

4455
    qemuDomainObjEnterRemote(vm);
4456
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4457 4458 4459
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;

4460
    if (dconn == NULL) {
4461
        virReportError(VIR_ERR_OPERATION_FAILED,
4462 4463
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4464
        virObjectUnref(cfg);
4465 4466 4467
        return -1;
    }

4468 4469
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4470 4471
        goto cleanup;

4472
    if (virConnectRegisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed,
4473 4474 4475 4476
                                        vm, NULL) < 0) {
        goto cleanup;
    }

4477
    qemuDomainObjEnterRemote(vm);
4478 4479
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4480
        /* v3proto reflects whether the caller used Perform3, but with
4481
         * p2p migrate, regardless of whether Perform2 or Perform3
4482 4483 4484 4485
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4486 4487
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
4488
    if (offline)
4489 4490
        dstOffline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                              VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4491 4492
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;
4493

4494
    if (!p2p) {
4495 4496
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4497 4498 4499
        goto cleanup;
    }

4500 4501
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4502
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
4503 4504 4505 4506 4507 4508
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("Migration APIs with extensible parameters are not "
                         "supported but extended parameters were passed"));
        goto cleanup;
    }

4509
    if (offline && !dstOffline) {
L
liguang 已提交
4510 4511 4512 4513 4514 4515
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4516 4517 4518 4519 4520 4521 4522
    /* 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;

4523
    if (*v3proto) {
4524 4525 4526
        ret = qemuMigrationSrcPerformPeer2Peer3(driver, sconn, dconn, dconnuri, vm, xmlin,
                                                persist_xml, dname, uri, graphicsuri,
                                                listenAddress, nmigrate_disks, migrate_disks,
4527
                                                nbdPort, migParams, resource,
4528
                                                useParams, flags);
4529
    } else {
4530
        ret = qemuMigrationSrcPerformPeer2Peer2(driver, sconn, dconn, vm,
4531
                                                dconnuri, flags, dname, resource,
4532
                                                migParams);
4533
    }
4534

4535
 cleanup:
4536
    orig_err = virSaveLastError();
4537
    qemuDomainObjEnterRemote(vm);
4538
    virConnectUnregisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed);
4539
    virObjectUnref(dconn);
4540
    ignore_value(qemuDomainObjExitRemote(vm, false));
4541 4542 4543 4544
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4545
    virObjectUnref(cfg);
4546 4547 4548 4549
    return ret;
}


4550 4551 4552 4553 4554 4555
/*
 * 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
4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567
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,
4568
                           qemuMigrationParamsPtr migParams,
4569 4570 4571 4572 4573 4574 4575 4576
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
                           const char *dname,
                           unsigned long resource,
                           bool v3proto)
4577
{
4578
    virObjectEventPtr event = NULL;
4579
    int ret = -1;
4580
    virErrorPtr orig_err = NULL;
4581
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4582
    qemuDomainObjPrivatePtr priv = vm->privateData;
4583

4584 4585
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                              flags) < 0)
4586 4587
        goto cleanup;

4588
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
4589 4590
        goto endjob;

4591
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
4592
        goto endjob;
4593

4594
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
4595
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
4596
        goto endjob;
4597

4598
    qemuMigrationSrcStoreDomainState(vm);
4599 4600

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4601 4602 4603
        ret = qemuMigrationSrcPerformPeer2Peer(driver, conn, vm, xmlin, persist_xml,
                                               dconnuri, uri, graphicsuri, listenAddress,
                                               nmigrate_disks, migrate_disks, nbdPort,
4604
                                               migParams, flags, dname, resource,
4605
                                               &v3proto);
4606
    } else {
4607
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4608 4609 4610
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                            cookieout, cookieoutlen,
                                            flags, resource, NULL, NULL, 0, NULL,
4611
                                            migParams);
4612
    }
4613 4614
    if (ret < 0)
        goto endjob;
4615

4616 4617 4618 4619
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4620
    if (!v3proto) {
4621
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
4622
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4623
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4624
        virDomainAuditStop(vm, "migrated");
4625
        event = virDomainEventLifecycleNewFromObj(vm,
4626 4627
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4628 4629
    }

4630
 endjob:
4631 4632 4633
    if (ret < 0)
        orig_err = virSaveLastError();

4634 4635 4636 4637
    /* v2 proto has no confirm phase so we need to reset migration parameters
     * here
     */
    if (!v3proto && ret < 0)
4638
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4639
                                 priv->job.migParams, priv->job.apiFlags);
4640

4641
    qemuMigrationSrcRestoreDomainState(driver, vm);
4642

4643
    qemuMigrationJobFinish(driver, vm);
4644
    if (!virDomainObjIsActive(vm) && ret == 0) {
4645
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
4646
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4647 4648
            vm->persistent = 0;
        }
4649
        qemuDomainRemoveInactiveJob(driver, vm);
4650 4651
    }

4652 4653 4654 4655 4656
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4657
 cleanup:
M
Michal Privoznik 已提交
4658
    virDomainObjEndAPI(&vm);
4659
    virObjectEventStateQueue(driver->domainEventState, event);
4660
    virObjectUnref(cfg);
4661 4662 4663 4664 4665 4666 4667
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4668 4669 4670 4671 4672 4673 4674 4675
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,
4676
                             qemuMigrationParamsPtr migParams,
4677 4678 4679 4680 4681 4682
                             const char *cookiein,
                             int cookieinlen,
                             char **cookieout,
                             int *cookieoutlen,
                             unsigned long flags,
                             unsigned long resource)
4683
{
4684
    qemuDomainObjPrivatePtr priv = vm->privateData;
4685 4686 4687 4688
    int ret = -1;

    /* If we didn't start the job in the begin phase, start it now. */
    if (!(flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
4689 4690
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
4691 4692 4693 4694 4695 4696
            goto cleanup;
    } else if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) {
        goto cleanup;
    }

    qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4697
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
4698
                           qemuMigrationSrcCleanup);
4699

4700 4701 4702
    ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                        cookieout, cookieoutlen,
                                        flags, resource, NULL, graphicsuri,
4703
                                        nmigrate_disks, migrate_disks, migParams);
4704

4705
    if (ret < 0) {
4706
        qemuMigrationSrcRestoreDomainState(driver, vm);
4707
        goto endjob;
4708
    }
4709 4710 4711

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4712
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
4713
                             qemuMigrationSrcCleanup) < 0)
4714 4715
        goto endjob;

4716
 endjob:
4717
    if (ret < 0) {
4718
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4719
                                 priv->job.migParams, priv->job.apiFlags);
4720
        qemuMigrationJobFinish(driver, vm);
4721
    } else {
4722
        qemuMigrationJobContinue(vm);
4723 4724
    }

4725
    if (!virDomainObjIsActive(vm))
4726
        qemuDomainRemoveInactiveJob(driver, vm);
4727

4728
 cleanup:
M
Michal Privoznik 已提交
4729
    virDomainObjEndAPI(&vm);
4730 4731 4732
    return ret;
}

4733
int
4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745
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,
4746
                        qemuMigrationParamsPtr migParams,
4747 4748 4749 4750 4751 4752 4753 4754
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4755 4756
{
    VIR_DEBUG("driver=%p, conn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4757
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
4758
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4759
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
4760
              "flags=0x%lx, dname=%s, resource=%lu, v3proto=%d",
4761
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4762
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4763 4764 4765
              nmigrate_disks, migrate_disks, nbdPort,
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4766 4767 4768

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4769 4770
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4771 4772 4773
            return -1;
        }

4774 4775 4776
        return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, dconnuri, uri,
                                          graphicsuri, listenAddress,
                                          nmigrate_disks, migrate_disks, nbdPort,
4777
                                          migParams,
4778 4779 4780
                                          cookiein, cookieinlen,
                                          cookieout, cookieoutlen,
                                          flags, dname, resource, v3proto);
4781 4782
    } else {
        if (dconnuri) {
4783 4784
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
4785 4786 4787 4788
            return -1;
        }

        if (v3proto) {
4789 4790 4791
            return qemuMigrationSrcPerformPhase(driver, conn, vm, persist_xml, uri,
                                                graphicsuri,
                                                nmigrate_disks, migrate_disks,
4792
                                                migParams,
4793 4794 4795
                                                cookiein, cookieinlen,
                                                cookieout, cookieoutlen,
                                                flags, resource);
4796
        } else {
4797 4798 4799
            return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, NULL,
                                              uri, graphicsuri, listenAddress,
                                              nmigrate_disks, migrate_disks, nbdPort,
4800
                                              migParams,
4801 4802 4803
                                              cookiein, cookieinlen,
                                              cookieout, cookieoutlen, flags,
                                              dname, resource, v3proto);
4804 4805 4806
        }
    }
}
4807

4808
static int
4809
qemuMigrationDstVPAssociatePortProfiles(virDomainDefPtr def)
4810
{
4811
    size_t i;
4812 4813 4814 4815 4816
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
4817
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4818
            if (virNetDevVPortProfileAssociate(net->ifname,
4819
                                               virDomainNetGetActualVirtPortProfile(net),
4820
                                               &net->mac,
4821
                                               virDomainNetGetActualDirectDev(net),
4822
                                               -1,
4823
                                               def->uuid,
4824 4825
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
4826 4827 4828
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
4829
                goto err_exit;
4830
            }
4831
            last_good_net = i;
4832
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
4833

4834
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
4835 4836 4837 4838
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
4839 4840 4841
        }
    }

4842
    return 0;
4843

4844
 err_exit:
4845
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
4846
        net = def->nets[i];
4847
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4848
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
4849
                                                           virDomainNetGetActualVirtPortProfile(net),
4850
                                                           &net->mac,
4851
                                                           virDomainNetGetActualDirectDev(net),
4852
                                                           -1,
4853
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
4854 4855
        }
    }
4856
    return -1;
4857 4858 4859
}


J
Jiri Denemark 已提交
4860
static int
4861 4862 4863 4864
qemuMigrationDstPersist(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationCookiePtr mig,
                        bool ignoreSaveError)
J
Jiri Denemark 已提交
4865 4866 4867 4868
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    virDomainDefPtr vmdef;
4869 4870
    virDomainDefPtr oldDef = NULL;
    unsigned int oldPersist = vm->persistent;
J
Jiri Denemark 已提交
4871 4872 4873 4874 4875 4876 4877
    virObjectEventPtr event;
    int ret = -1;

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

    vm->persistent = 1;
4878 4879
    oldDef = vm->newDef;
    vm->newDef = qemuMigrationCookieGetPersistent(mig);
J
Jiri Denemark 已提交
4880

4881 4882
    if (!(vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm)))
        goto error;
J
Jiri Denemark 已提交
4883

4884 4885
    if (virDomainSaveConfig(cfg->configDir, driver->caps, vmdef) < 0 &&
        !ignoreSaveError)
4886
        goto error;
J
Jiri Denemark 已提交
4887 4888 4889

    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_DEFINED,
4890 4891 4892
                                              oldPersist ?
                                              VIR_DOMAIN_EVENT_DEFINED_UPDATED :
                                              VIR_DOMAIN_EVENT_DEFINED_ADDED);
4893
    virObjectEventStateQueue(driver->domainEventState, event);
J
Jiri Denemark 已提交
4894 4895 4896 4897

    ret = 0;

 cleanup:
4898
    virDomainDefFree(oldDef);
J
Jiri Denemark 已提交
4899 4900 4901
    virObjectUnref(caps);
    virObjectUnref(cfg);
    return ret;
4902 4903 4904 4905 4906 4907 4908

 error:
    virDomainDefFree(vm->newDef);
    vm->persistent = oldPersist;
    vm->newDef = oldDef;
    oldDef = NULL;
    goto cleanup;
J
Jiri Denemark 已提交
4909 4910 4911
}


4912
virDomainPtr
4913 4914 4915 4916 4917 4918 4919 4920 4921 4922
qemuMigrationDstFinish(virQEMUDriverPtr driver,
                       virConnectPtr dconn,
                       virDomainObjPtr vm,
                       const char *cookiein,
                       int cookieinlen,
                       char **cookieout,
                       int *cookieoutlen,
                       unsigned long flags,
                       int retcode,
                       bool v3proto)
4923 4924
{
    virDomainPtr dom = NULL;
4925
    qemuMigrationCookiePtr mig = NULL;
4926
    virErrorPtr orig_err = NULL;
4927
    int cookie_flags = 0;
J
Jiri Denemark 已提交
4928
    qemuDomainObjPrivatePtr priv = vm->privateData;
4929
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4930
    unsigned short port;
4931 4932
    unsigned long long timeReceived = 0;
    virObjectEventPtr event;
4933
    qemuDomainJobInfoPtr jobInfo = NULL;
4934
    bool inPostCopy = false;
4935
    bool doKill = true;
4936

4937
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
4938
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, retcode=%d",
4939 4940
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
4941

4942 4943 4944
    port = priv->migrationPort;
    priv->migrationPort = 0;

4945
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN)) {
4946
        qemuMigrationDstErrorReport(driver, vm->def->name);
4947
        goto cleanup;
4948
    }
4949

4950 4951
    ignore_value(virTimeMillisNow(&timeReceived));

4952 4953 4954
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
4955

4956
    qemuDomainCleanupRemove(vm, qemuMigrationDstPrepareCleanup);
4957
    VIR_FREE(priv->job.completed);
4958

4959
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
4960 4961
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
4962 4963 4964
    /* 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;
4965

4966 4967
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen, cookie_flags)))
4968
        goto endjob;
4969

4970
    if (flags & VIR_MIGRATE_OFFLINE) {
4971
        if (retcode == 0 &&
4972
            qemuMigrationDstPersist(driver, vm, mig, false) == 0)
4973
            dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, -1);
4974 4975
        goto endjob;
    }
4976

4977 4978 4979 4980
    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
         */
4981
        qemuDomainCheckMonitor(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN);
4982 4983
        goto endjob;
    }
4984

4985 4986 4987
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4988
        qemuMigrationDstErrorReport(driver, vm->def->name);
4989 4990
        goto endjob;
    }
4991

4992
    if (qemuMigrationDstVPAssociatePortProfiles(vm->def) < 0)
4993
        goto endjob;
4994

4995
    if (mig->network && qemuMigrationDstOPDRelocate(driver, vm, mig) < 0)
4996
        VIR_WARN("unable to provide network data for relocation");
4997

4998
    if (qemuMigrationDstStopNBDServer(driver, vm, mig) < 0)
4999
        goto endjob;
5000

5001 5002
    if (qemuRefreshVirtioChannelState(driver, vm,
                                      QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
5003 5004
        goto endjob;

5005 5006
    if (qemuConnectAgent(driver, vm) < 0)
        goto endjob;
5007

5008
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
5009
        if (qemuMigrationDstPersist(driver, vm, mig, !v3proto) < 0) {
5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020
            /* 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.
5021 5022 5023
             */
            if (v3proto)
                goto endjob;
5024
        }
5025
    }
5026

5027 5028 5029
    /* We need to wait for QEMU to process all data sent by the source
     * before starting guest CPUs.
     */
5030 5031 5032
    if (qemuMigrationDstWaitForCompletion(driver, vm,
                                          QEMU_ASYNC_JOB_MIGRATION_IN,
                                          !!(flags & VIR_MIGRATE_POSTCOPY)) < 0) {
5033 5034 5035 5036 5037 5038
        /* 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;
    }
5039

5040 5041 5042 5043 5044 5045 5046 5047 5048
    /* 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;
    }

5049
    if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY)
5050 5051
        inPostCopy = true;

5052 5053 5054 5055 5056
    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
         */
5057
        if (qemuProcessStartCPUs(driver, vm,
5058 5059
                                 inPostCopy ? VIR_DOMAIN_RUNNING_POSTCOPY
                                            : VIR_DOMAIN_RUNNING_MIGRATED,
5060
                                 QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5061
            if (virGetLastErrorCode() == VIR_ERR_OK)
5062 5063 5064 5065 5066 5067
                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();
5068

5069 5070 5071 5072 5073 5074 5075 5076 5077 5078
            /*
             * 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;
5079
        }
5080

5081
        if (inPostCopy)
5082
            doKill = false;
5083
    }
5084

5085
    if (mig->jobInfo) {
5086
        jobInfo = mig->jobInfo;
5087 5088 5089 5090 5091 5092
        mig->jobInfo = NULL;

        if (jobInfo->sent && timeReceived) {
            jobInfo->timeDelta = timeReceived - jobInfo->sent;
            jobInfo->received = timeReceived;
            jobInfo->timeDeltaSet = true;
5093
        }
5094 5095
        qemuDomainJobInfoUpdateTime(jobInfo);
        qemuDomainJobInfoUpdateDowntime(jobInfo);
5096
    }
L
liguang 已提交
5097

5098
    if (inPostCopy) {
5099 5100 5101
        if (qemuMigrationDstWaitForCompletion(driver, vm,
                                              QEMU_ASYNC_JOB_MIGRATION_IN,
                                              false) < 0) {
5102 5103 5104 5105 5106 5107 5108 5109 5110
            goto endjob;
        }
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            virDomainObjSetState(vm,
                                 VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_MIGRATED);
        }
    }

5111
    dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, vm->def->id);
5112

5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126
    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);
    }
5127

5128 5129 5130 5131 5132
    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);
5133
        virObjectEventStateQueue(driver->domainEventState, event);
5134
    }
5135

5136
    if (virDomainObjIsActive(vm) &&
5137
        virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
5138 5139 5140 5141 5142
        VIR_WARN("Failed to save status on vm %s", vm->def->name);

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

5143
 endjob:
5144
    if (!dom &&
5145 5146
        !(flags & VIR_MIGRATE_OFFLINE) &&
        virDomainObjIsActive(vm)) {
5147
        if (doKill) {
5148 5149 5150 5151 5152 5153 5154
            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);
5155
            virObjectEventStateQueue(driver->domainEventState, event);
5156
        } else {
5157
            qemuMigrationAnyPostcopyFailed(driver, vm);
5158
        }
5159 5160
    }

5161
    if (dom) {
5162 5163 5164
        if (jobInfo) {
            VIR_STEAL_PTR(priv->job.completed, jobInfo);
            priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
5165
            priv->job.completed->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;
5166
        }
5167

5168 5169 5170
        if (qemuMigrationBakeCookie(mig, driver, vm,
                                    QEMU_MIGRATION_DESTINATION,
                                    cookieout, cookieoutlen,
5171 5172
                                    QEMU_MIGRATION_COOKIE_STATS) < 0)
            VIR_WARN("Unable to encode migration cookie");
5173 5174 5175 5176 5177 5178

        /* Remove completed stats for post-copy, everything but timing fields
         * is obsolete anyway.
         */
        if (inPostCopy)
            VIR_FREE(priv->job.completed);
5179
    }
5180

5181
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
5182
                             priv->job.migParams, priv->job.apiFlags);
5183

5184
    qemuMigrationJobFinish(driver, vm);
5185
    if (!virDomainObjIsActive(vm))
5186
        qemuDomainRemoveInactiveJob(driver, vm);
5187

5188
 cleanup:
5189
    VIR_FREE(jobInfo);
5190
    virPortAllocatorRelease(port);
5191
    if (priv->mon)
5192
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
5193
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5194
    virDomainObjEndAPI(&vm);
5195
    qemuMigrationCookieFree(mig);
5196 5197 5198 5199
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5200
    virObjectUnref(cfg);
5201 5202 5203 5204

    /* Set a special error if Finish is expected to return NULL as a result of
     * successful call with retcode != 0
     */
5205
    if (retcode != 0 && !dom && virGetLastErrorCode() == VIR_ERR_OK)
5206
        virReportError(VIR_ERR_MIGRATE_FINISH_OK, NULL);
5207 5208
    return dom;
}
5209

5210

5211
/* Helper function called while vm is active.  */
5212
int
5213 5214 5215 5216
qemuMigrationSrcToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
                       int fd,
                       const char *compressor,
                       qemuDomainAsyncJob asyncJob)
5217 5218 5219
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5220
    int ret = -1;
5221 5222
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5223
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5224
    char *errbuf = NULL;
5225
    virErrorPtr orig_err = NULL;
5226 5227 5228 5229 5230

    /* 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,
5231 5232
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5233 5234
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5235
    }
5236

5237 5238 5239 5240 5241 5242 5243
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5244 5245 5246 5247
    if (compressor && pipe(pipeFD) < 0) {
        virReportSystemError(errno, "%s",
                             _("Failed to create pipe for migration"));
        return -1;
5248 5249
    }

5250 5251 5252 5253
    /* 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.  */
5254 5255
    if (qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    compressor ? pipeFD[1] : fd) < 0)
5256 5257
        goto cleanup;

5258
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5259 5260
        goto cleanup;

5261
    if (!compressor) {
5262 5263 5264
        rc = qemuMonitorMigrateToFd(priv->mon,
                                    QEMU_MONITOR_MIGRATE_BACKGROUND,
                                    fd);
5265 5266 5267 5268 5269 5270 5271
    } else {
        const char *prog = compressor;
        const char *args[] = {
            prog,
            "-c",
            NULL
        };
5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282

        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;
5283
        }
5284 5285 5286 5287 5288 5289 5290 5291 5292 5293
        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");
5294
    }
5295
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5296
        goto cleanup;
5297 5298 5299
    if (rc < 0)
        goto cleanup;

5300
    rc = qemuMigrationSrcWaitForCompletion(driver, vm, asyncJob, NULL, 0);
5301

5302 5303 5304 5305
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5306 5307
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5308
                qemuMonitorMigrateCancel(priv->mon);
5309
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5310 5311
            }
        }
5312
        goto cleanup;
5313
    }
5314

5315 5316 5317
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5318
    qemuDomainEventEmitJobCompleted(driver, vm);
5319 5320
    ret = 0;

5321
 cleanup:
5322 5323 5324
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5325
    /* Restore max migration bandwidth */
5326 5327
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5328 5329
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5330
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5331 5332
    }

5333 5334
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5335 5336 5337 5338 5339
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5340 5341 5342 5343 5344 5345

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

5346 5347
    return ret;
}
5348

5349 5350

int
5351 5352
qemuMigrationSrcCancel(virQEMUDriverPtr driver,
                       virDomainObjPtr vm)
5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363
{
    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));
5364
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5365 5366 5367 5368 5369 5370
        goto cleanup;

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

5371
        if (!diskPriv->blockjob->started)
5372
            diskPriv->migrating = false;
5373 5374 5375 5376

        if (diskPriv->migrating) {
            qemuBlockJobSyncBeginDisk(disk);
            storage = true;
5377 5378 5379
        }
    }

5380 5381 5382

    if (storage &&
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
5383
                                      QEMU_ASYNC_JOB_NONE, NULL) < 0)
5384
        goto cleanup;
5385 5386 5387 5388 5389 5390 5391 5392

    ret = 0;

 cleanup:
    return ret;
}


5393
static int
5394
qemuMigrationJobStart(virQEMUDriverPtr driver,
5395
                      virDomainObjPtr vm,
5396 5397
                      qemuDomainAsyncJob job,
                      unsigned long apiFlags)
5398
{
5399
    qemuDomainObjPrivatePtr priv = vm->privateData;
5400 5401
    virDomainJobOperation op;
    unsigned long long mask;
5402

5403
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5404 5405
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_IN;
        mask = QEMU_JOB_NONE;
5406
    } else {
5407 5408 5409 5410
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_OUT;
        mask = QEMU_JOB_DEFAULT_MASK |
               JOB_MASK(QEMU_JOB_SUSPEND) |
               JOB_MASK(QEMU_JOB_MIGRATION_OP);
5411
    }
5412

5413
    if (qemuDomainObjBeginAsyncJob(driver, vm, job, op, apiFlags) < 0)
5414 5415
        return -1;

5416 5417
    priv->job.current->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;

5418
    qemuDomainObjSetAsyncJobMask(vm, mask);
5419 5420 5421
    return 0;
}

5422
static void
5423
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5424
                         virDomainObjPtr vm,
5425
                         qemuMigrationJobPhase phase)
5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438
{
    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);
}

5439
static void
5440
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5441
                           virDomainObjPtr vm,
5442
                           qemuMigrationJobPhase phase)
5443 5444 5445 5446
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5447
static void
5448 5449
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5450
    qemuDomainObjReleaseAsyncJob(vm);
5451 5452
}

5453
static bool
5454
qemuMigrationJobIsActive(virDomainObjPtr vm,
5455
                         qemuDomainAsyncJob job)
5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466
{
    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");

5467
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5468 5469 5470 5471 5472
        return false;
    }
    return true;
}

5473
static void
5474
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5475
{
5476
    qemuDomainObjEndAsyncJob(driver, vm);
5477
}
5478 5479 5480


static void
5481
qemuMigrationDstErrorFree(void *data,
5482 5483 5484 5485 5486 5487 5488
                       const void *name ATTRIBUTE_UNUSED)
{
    virErrorPtr err = data;
    virFreeError(err);
}

int
5489
qemuMigrationDstErrorInit(virQEMUDriverPtr driver)
5490
{
5491
    driver->migrationErrors = virHashAtomicNew(64, qemuMigrationDstErrorFree);
5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502
    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
5503 5504 5505
qemuMigrationDstErrorSave(virQEMUDriverPtr driver,
                          const char *name,
                          virErrorPtr err)
5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518
{
    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
5519 5520
qemuMigrationDstErrorReport(virQEMUDriverPtr driver,
                            const char *name)
5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531
{
    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);
}
5532 5533


5534
int
5535 5536 5537 5538
qemuMigrationSrcFetchMirrorStats(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 qemuDomainJobInfoPtr jobInfo)
5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582
{
    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;
}