qemu_migration.c 187.8 KB
Newer Older
1 2 3
/*
 * qemu_migration.c: QEMU migration handling
 *
4
 * Copyright (C) 2006-2015 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23
 *
 */

#include <config.h>

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

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

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

#define VIR_FROM_THIS VIR_FROM_QEMU

66 67
VIR_LOG_INIT("qemu.qemu_migration");

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

167 168

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

294 295

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

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

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

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

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

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

        diskSrcPath = virDomainDiskGetSource(disk);

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

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

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

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

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


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

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

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

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

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

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

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

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

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

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

    priv->nbdPort = port;
    ret = 0;

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

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

440 441

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

    if (!mig->nbd)
        return 0;

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

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

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


466
static void
467 468
qemuMigrationNBDReportMirrorError(qemuBlockJobDataPtr job,
                                  const char *diskdst)
469
{
470
    if (job->errmsg) {
471 472
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("migration of disk %s failed: %s"),
473
                       diskdst, job->errmsg);
474 475
    } else {
        virReportError(VIR_ERR_OPERATION_FAILED,
476
                       _("migration of disk %s failed"), diskdst);
477 478 479 480
    }
}


481
/**
482
 * qemuMigrationSrcNBDStorageCopyReady:
483 484
 * @vm: domain
 *
485 486
 * Check the status of all drives copied via qemuMigrationSrcNBDStorageCopy.
 * Any pending block job events for the mirrored disks will be processed.
487 488 489 490 491 492
 *
 * Returns 1 if all mirrors are "ready",
 *         0 if some mirrors are still performing initial sync,
 *        -1 on error.
 */
static int
493 494
qemuMigrationSrcNBDStorageCopyReady(virDomainObjPtr vm,
                                    qemuDomainAsyncJob asyncJob)
495 496
{
    size_t i;
497 498
    size_t notReady = 0;
    int status;
499 500 501

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
502
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
503
        qemuBlockJobDataPtr job;
504

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

508 509 510 511 512 513 514
        if (!(job = qemuBlockJobDiskGetJob(disk))) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("missing block job data for disk '%s'"), disk->dst);
            return -1;
        }

        status = qemuBlockJobUpdate(vm, job, asyncJob);
515
        if (status == VIR_DOMAIN_BLOCK_JOB_FAILED) {
516 517
            qemuMigrationNBDReportMirrorError(job, disk->dst);
            virObjectUnref(job);
518 519
            return -1;
        }
520

521 522
        virObjectUnref(job);

523 524
        if (disk->mirrorState != VIR_DOMAIN_DISK_MIRROR_STATE_READY)
            notReady++;
525 526
    }

527 528 529 530 531 532 533
    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;
    }
534 535 536
}


537 538 539 540
/*
 * 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.
541
 *
542 543 544 545
 * 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.
546 547
 */
static int
548 549 550
qemuMigrationSrcNBDCopyCancelled(virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 bool check)
551
{
552 553
    size_t i;
    size_t active = 0;
554
    size_t completed = 0;
555 556
    int status;
    bool failed = false;
557

558
 retry:
559 560 561
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
562
        qemuBlockJobDataPtr job;
563

564 565
        if (!diskPriv->migrating)
            continue;
566

567 568 569 570
        if (!(job = qemuBlockJobDiskGetJob(disk)))
            continue;

        status = qemuBlockJobUpdate(vm, job, asyncJob);
571 572 573
        switch (status) {
        case VIR_DOMAIN_BLOCK_JOB_FAILED:
            if (check) {
574
                qemuMigrationNBDReportMirrorError(job, disk->dst);
575
                failed = true;
576
            }
M
Marc Hartmayer 已提交
577
            ATTRIBUTE_FALLTHROUGH;
578 579 580 581
        case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
            diskPriv->migrating = false;
            break;
582

583 584
        default:
            active++;
585
        }
586 587 588

        if (status == VIR_DOMAIN_BLOCK_JOB_COMPLETED)
            completed++;
589 590

        virObjectUnref(job);
591 592 593 594 595 596 597 598 599 600
    }

    /* 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;
601
    }
602

603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
    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;
619
        }
620
    }
621
}
622 623


624 625 626 627 628 629
/*
 * 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
630 631 632
qemuMigrationSrcNBDCopyCancelOne(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 virDomainDiskDefPtr disk,
633
                                 qemuBlockJobDataPtr job,
634 635
                                 bool failNoJob,
                                 qemuDomainAsyncJob asyncJob)
636 637 638 639 640 641 642
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char *diskAlias = NULL;
    int ret = -1;
    int status;
    int rv;

643
    status = qemuBlockJobUpdate(vm, job, asyncJob);
644 645 646 647
    switch (status) {
    case VIR_DOMAIN_BLOCK_JOB_FAILED:
    case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        if (failNoJob) {
648
            qemuMigrationNBDReportMirrorError(job, disk->dst);
649
            goto cleanup;
650
        }
651
        ATTRIBUTE_FALLTHROUGH;
652
    case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
653 654
        ret = 1;
        goto cleanup;
655 656
    }

657
    if (!(diskAlias = qemuAliasDiskDriveFromDisk(disk)))
658 659
        return -1;

660
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
661 662
        goto cleanup;

663
    rv = qemuMonitorBlockJobCancel(priv->mon, diskAlias);
664 665 666 667 668

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

    ret = 0;
669 670 671 672 673 674 675 676

 cleanup:
    VIR_FREE(diskAlias);
    return ret;
}


/**
677
 * qemuMigrationSrcNBDCopyCancel:
678 679
 * @driver: qemu driver
 * @vm: domain
680
 * @check: if true report an error when some of the mirrors fails
681
 *
682
 * Cancel all drive-mirrors started by qemuMigrationSrcNBDStorageCopy.
683 684 685
 * 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.
686 687 688 689
 *
 * Returns 0 on success, -1 otherwise.
 */
static int
690 691 692 693 694
qemuMigrationSrcNBDCopyCancel(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              bool check,
                              qemuDomainAsyncJob asyncJob,
                              virConnectPtr dconn)
695
{
696
    virErrorPtr err = NULL;
697
    int ret = -1;
698
    size_t i;
699 700 701 702
    int rv;
    bool failed = false;

    VIR_DEBUG("Cancelling drive mirrors for domain %s", vm->def->name);
703 704 705

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
706
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
707
        qemuBlockJobDataPtr job;
708

709 710
        if (!(job = qemuBlockJobDiskGetJob(disk)) ||
            !qemuBlockJobIsRunning(job))
711 712
            diskPriv->migrating = false;

713 714
        if (!diskPriv->migrating) {
            virObjectUnref(job);
715
            continue;
716
        }
717

718
        rv = qemuMigrationSrcNBDCopyCancelOne(driver, vm, disk, job,
719
                                              check, asyncJob);
720 721 722 723 724 725
        if (rv != 0) {
            if (rv < 0) {
                if (!err)
                    err = virSaveLastError();
                failed = true;
            }
726
            qemuBlockJobSyncEnd(vm, job, asyncJob);
727 728
            diskPriv->migrating = false;
        }
729 730

        virObjectUnref(job);
731 732
    }

733
    while ((rv = qemuMigrationSrcNBDCopyCancelled(vm, asyncJob, check)) != 1) {
734 735 736 737 738 739 740
        if (check && !failed &&
            dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            failed = true;
        }

741 742 743 744
        if (rv < 0) {
            failed = true;
            if (rv == -2)
                break;
745
        }
746

747 748 749 750 751
        if (failed && !err)
            err = virSaveLastError();

        if (virDomainObjWait(vm) < 0)
            goto cleanup;
752 753
    }

754 755 756 757 758 759 760 761 762 763 764 765 766
    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;
    }

767 768 769
    ret = failed ? -1 : 0;

 cleanup:
770 771 772 773 774
    if (err) {
        virSetError(err);
        virFreeError(err);
    }
    return ret;
775 776 777
}


778 779 780 781 782 783 784 785 786 787 788 789 790 791 792
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);
    int mon_ret = 0;
    int ret = -1;
793
    VIR_AUTOUNREF(virStorageSourcePtr) copysrc = NULL;
794 795 796

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

797
    if (!(copysrc = virStorageSourceNew()))
798 799 800 801 802 803
        goto cleanup;

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

804
    if (!(copysrc->backingStore = virStorageSourceNew()))
805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855
        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);
    return ret;
}


856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897
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;
}


898 899 900 901 902 903 904 905 906 907 908 909
static int
qemuMigrationSrcNBDStorageCopyOne(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  virDomainDiskDefPtr disk,
                                  const char *host,
                                  int port,
                                  unsigned long long mirror_speed,
                                  unsigned int mirror_flags,
                                  const char *tlsAlias,
                                  unsigned int flags)
{
    qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
910
    qemuBlockJobDataPtr job = NULL;
911 912 913 914 915 916 917
    char *diskAlias = NULL;
    int rc;
    int ret = -1;

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

918
    if (!(job = qemuBlockJobDiskNew(disk, QEMU_BLOCKJOB_TYPE_COPY, diskAlias)))
919 920
        goto cleanup;

921
    qemuBlockJobSyncBegin(job);
922 923 924 925 926 927 928 929 930 931 932 933 934 935 936

    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);
    }

937
    if (rc < 0)
938 939 940
        goto cleanup;

    diskPriv->migrating = true;
941
    qemuBlockJobStarted(job);
942 943 944 945

    ret = 0;

 cleanup:
946
    qemuBlockJobStartupFinalize(job);
947 948 949 950 951
    VIR_FREE(diskAlias);
    return ret;
}


952
/**
953
 * qemuMigrationSrcNBDStorageCopy:
954 955 956 957
 * @driver: qemu driver
 * @vm: domain
 * @mig: migration cookie
 * @host: where are we migrating to
958
 * @speed: bandwidth limit in MiB/s
959 960
 * @migrate_flags: migrate monitor command flags
 *
961 962 963
 * 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
964
 * to do the very same operation. On failure, the caller is
965 966
 * expected to call qemuMigrationSrcNBDCopyCancel to stop all
 * running copy operations.
967 968 969 970 971
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
972 973 974 975 976 977 978 979
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,
980 981 982
                               virConnectPtr dconn,
                               const char *tlsAlias,
                               unsigned int flags)
983 984 985 986
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int port;
987
    size_t i;
988
    unsigned long long mirror_speed = speed;
989
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
990
    int rv;
991
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
992 993

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
994

995 996 997 998 999 1000 1001 1002
    if (mirror_speed > LLONG_MAX >> 20) {
        virReportError(VIR_ERR_OVERFLOW,
                       _("bandwidth must be less than %llu"),
                       LLONG_MAX >> 20);
        goto cleanup;
    }
    mirror_speed <<= 20;

1003 1004 1005 1006 1007 1008 1009 1010 1011 1012
    /* 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];

1013
        /* check whether disk should be migrated */
1014
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1015 1016
            continue;

1017 1018 1019
        if (qemuMigrationSrcNBDStorageCopyOne(driver, vm, disk, host, port,
                                              mirror_speed, mirror_flags,
                                              tlsAlias, flags) < 0)
1020 1021
            goto cleanup;

1022
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1023 1024 1025
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto cleanup;
        }
1026
    }
1027

1028
    while ((rv = qemuMigrationSrcNBDStorageCopyReady(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) != 1) {
1029 1030
        if (rv < 0)
            goto cleanup;
1031

1032
        if (priv->job.abortJob) {
1033
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1034 1035 1036 1037
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
            goto cleanup;
1038
        }
1039

1040 1041 1042 1043 1044 1045
        if (dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            goto cleanup;
        }

1046
        if (virDomainObjWait(vm) < 0)
1047
            goto cleanup;
1048 1049
    }

1050 1051
    qemuMigrationSrcFetchMirrorStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                     priv->job.current);
1052

1053
    /* Okay, all disks are ready. Modify migrate_flags */
1054 1055 1056 1057
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

1058
 cleanup:
1059
    virObjectUnref(cfg);
1060 1061
    return ret;
}
1062

1063

1064
/**
1065
 * qemuMigrationSrcIsAllowedHostdev:
1066 1067
 * @def: domain definition
 *
1068
 * Checks that @def does not contain any host devices unsupported across
1069 1070 1071
 * migrations. Returns true if the vm is allowed to migrate.
 */
static bool
1072
qemuMigrationSrcIsAllowedHostdev(const virDomainDef *def)
1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
{
    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;
}


1092
/**
1093
 * qemuMigrationSrcIsAllowed:
1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105
 * @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.
 */
1106
bool
1107 1108 1109 1110
qemuMigrationSrcIsAllowed(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          bool remote,
                          unsigned int flags)
1111
{
1112
    int nsnapshots;
1113
    int pauseReason;
1114
    size_t i;
1115

1116 1117 1118 1119 1120
    /* perform these checks only when migrating to remote hosts */
    if (remote) {
        nsnapshots = virDomainSnapshotObjListNum(vm->snapshots, NULL, 0);
        if (nsnapshots < 0)
            return false;
1121

1122 1123 1124 1125 1126
        if (nsnapshots > 0) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("cannot migrate domain with %d snapshots"),
                           nsnapshots);
            return false;
1127
        }
1128

1129
        /* cancel migration if disk I/O error is emitted while migrating */
1130
        if (flags & VIR_MIGRATE_ABORT_ON_ERROR &&
1131
            !(flags & VIR_MIGRATE_OFFLINE) &&
1132 1133
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
E
Eric Blake 已提交
1134
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1135
                           _("cannot migrate domain with I/O error"));
E
Eric Blake 已提交
1136 1137
            return false;
        }
1138 1139

    }
1140

1141 1142 1143 1144 1145 1146 1147
    /* 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;
        }
1148

1149

1150 1151
        if (qemuDomainHasBlockjob(vm, false)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1152
                           _("domain has active block job"));
1153 1154 1155
            return false;
        }

1156
        if (!qemuMigrationSrcIsAllowedHostdev(vm->def))
1157 1158
            return false;

1159
        if (vm->def->cpu) {
1160 1161 1162
            /* QEMU blocks migration and save with invariant TSC enabled
             * unless TSC frequency is explicitly set.
             */
1163 1164
            if (virCPUCheckFeature(vm->def->os.arch, vm->def->cpu,
                                   "invtsc") == 1) {
1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182
                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;
                }
1183
            }
J
Ján Tomko 已提交
1184 1185
        }

1186 1187 1188
        /* Verify that memory device config can be transferred reliably */
        for (i = 0; i < vm->def->nmems; i++) {
            virDomainMemoryDefPtr mem = vm->def->mems[i];
1189

1190 1191 1192 1193 1194
            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"));
1195

1196 1197
                return false;
            }
1198
        }
1199 1200 1201 1202 1203 1204

        if (vm->def->nshmems) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("migration with shmem device is not supported"));
            return false;
        }
1205 1206
    }

1207 1208 1209
    return true;
}

1210
static bool
1211 1212 1213 1214
qemuMigrationSrcIsSafe(virDomainDefPtr def,
                       size_t nmigrate_disks,
                       const char **migrate_disks,
                       unsigned int flags)
1215

1216
{
1217 1218
    bool storagemigration = flags & (VIR_MIGRATE_NON_SHARED_DISK |
                                     VIR_MIGRATE_NON_SHARED_INC);
1219
    size_t i;
1220
    int rc;
1221

1222
    for (i = 0; i < def->ndisks; i++) {
1223
        virDomainDiskDefPtr disk = def->disks[i];
1224
        const char *src = virDomainDiskGetSource(disk);
1225

1226 1227
        /* Disks without any source (i.e. floppies and CD-ROMs)
         * OR readonly are safe. */
1228
        if (virStorageSourceIsEmpty(disk->src) ||
1229
            disk->src->readonly)
1230
            continue;
1231

1232
        /* Disks which are migrated by qemu are safe too. */
1233
        if (storagemigration &&
1234
            qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1235
            continue;
1236

1237
        /* However, disks on local FS (e.g. ext4) are not safe. */
1238
        if (virStorageSourceGetActualType(disk->src) == VIR_STORAGE_TYPE_FILE) {
1239
            if ((rc = virFileIsSharedFS(src)) < 0) {
1240
                return false;
1241 1242 1243 1244 1245
            } else if (rc == 0) {
                virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                               _("Migration without shared storage is unsafe"));
                return false;
            }
1246 1247 1248 1249
            if ((rc = virStorageFileIsClusterFS(src)) < 0)
                return false;
            else if (rc == 1)
                continue;
1250 1251
        } else if (virStorageSourceGetActualType(disk->src) == VIR_STORAGE_TYPE_NETWORK) {
            /* But network disks are safe again. */
1252
            continue;
1253
        }
1254

1255 1256 1257 1258 1259 1260 1261
        /* 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;

1262 1263
        virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                       _("Migration may lead to data corruption if disks"
1264
                         " use cache != none or cache != directsync"));
1265
        return false;
1266 1267 1268 1269 1270
    }

    return true;
}

1271
/** qemuMigrationSrcSetOffline
1272 1273 1274
 * Pause domain for non-live migration.
 */
int
1275 1276
qemuMigrationSrcSetOffline(virQEMUDriverPtr driver,
                           virDomainObjPtr vm)
1277 1278
{
    int ret;
1279
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
1280 1281
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
1282
    if (ret == 0) {
1283
        virObjectEventPtr event;
1284

1285
        event = virDomainEventLifecycleNewFromObj(vm,
1286 1287
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
1288
        virObjectEventStateQueue(driver->domainEventState, event);
1289 1290 1291 1292 1293
    }

    return ret;
}

1294 1295

void
1296 1297
qemuMigrationAnyPostcopyFailed(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327
{
    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);
1328
        virObjectEventStateQueue(driver->domainEventState, event);
1329 1330 1331 1332 1333 1334 1335
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
    }
}


1336
static int
1337
qemuMigrationSrcWaitForSpice(virDomainObjPtr vm)
1338 1339 1340
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

1341
    if (!priv->job.spiceMigration)
1342 1343
        return 0;

1344
    VIR_DEBUG("Waiting for SPICE to finish migration");
1345 1346
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
1347 1348 1349 1350
            return -1;
    }
    return 0;
}
1351

1352 1353 1354 1355

static void
qemuMigrationUpdateJobType(qemuDomainJobInfoPtr jobInfo)
{
1356
    switch ((qemuMonitorMigrationStatus) jobInfo->stats.mig.status) {
1357 1358 1359 1360
    case QEMU_MONITOR_MIGRATION_STATUS_POSTCOPY:
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_POSTCOPY;
        break;

1361
    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
1362
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED;
1363 1364 1365
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
1366
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_NONE;
1367 1368 1369
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
1370
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1371 1372 1373
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
1374
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1375 1376
        break;

1377 1378 1379 1380 1381 1382 1383 1384
    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;

1385 1386 1387
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLING:
1388
    case QEMU_MONITOR_MIGRATION_STATUS_LAST:
1389 1390 1391 1392 1393 1394
        break;
    }
}


int
1395 1396 1397 1398 1399
qemuMigrationAnyFetchStats(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuDomainAsyncJob asyncJob,
                           qemuDomainJobInfoPtr jobInfo,
                           char **error)
1400 1401
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1402
    qemuMonitorMigrationStats stats;
1403 1404 1405 1406 1407
    int rv;

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

1408
    rv = qemuMonitorGetMigrationStats(priv->mon, &stats, error);
1409 1410 1411 1412

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

1413
    jobInfo->stats.mig = stats;
1414 1415

    return 0;
1416 1417 1418
}


1419 1420 1421 1422 1423 1424 1425
static const char *
qemuMigrationJobName(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
1426
        return _("migration out job");
1427 1428 1429 1430
    case QEMU_ASYNC_JOB_SAVE:
        return _("domain save job");
    case QEMU_ASYNC_JOB_DUMP:
        return _("domain core dump job");
1431 1432 1433 1434 1435 1436 1437 1438 1439
    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:
1440 1441 1442 1443 1444 1445
    default:
        return _("job");
    }
}


1446
static int
1447
qemuMigrationJobCheckStatus(virQEMUDriverPtr driver,
1448
                            virDomainObjPtr vm,
1449
                            qemuDomainAsyncJob asyncJob)
1450 1451 1452
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1453
    char *error = NULL;
1454
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1455
    int ret = -1;
1456

1457
    if (!events ||
1458
        jobInfo->stats.mig.status == QEMU_MONITOR_MIGRATION_STATUS_ERROR) {
1459
        if (qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, &error) < 0)
1460 1461
            return -1;
    }
1462

1463 1464
    qemuMigrationUpdateJobType(jobInfo);

1465 1466
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_NONE:
1467 1468
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("is not active"));
1469
        goto cleanup;
1470

1471
    case QEMU_DOMAIN_JOB_STATUS_FAILED:
1472
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
1473 1474 1475
                       qemuMigrationJobName(vm),
                       error ? error : _("unexpectedly failed"));
        goto cleanup;
1476

1477
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
1478 1479
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("canceled by client"));
1480
        goto cleanup;
1481

1482 1483
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
1484
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
1485
    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
1486
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1487
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1488 1489
        break;
    }
1490 1491 1492 1493 1494 1495

    ret = 0;

 cleanup:
    VIR_FREE(error);
    return ret;
1496 1497 1498
}


1499 1500
enum qemuMigrationCompletedFlags {
    QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR = (1 << 0),
1501
    /* This flag should only be set when run on src host */
1502
    QEMU_MIGRATION_COMPLETED_CHECK_STORAGE  = (1 << 1),
1503
    QEMU_MIGRATION_COMPLETED_POSTCOPY       = (1 << 2),
1504
    QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER = (1 << 3),
1505 1506
};

1507

1508 1509 1510 1511 1512 1513 1514
/**
 * 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
1515 1516 1517 1518 1519
qemuMigrationAnyCompleted(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          qemuDomainAsyncJob asyncJob,
                          virConnectPtr dconn,
                          unsigned int flags)
1520 1521 1522 1523 1524
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    int pauseReason;

1525
    if (qemuMigrationJobCheckStatus(driver, vm, asyncJob) < 0)
1526 1527
        goto error;

1528
    /* This flag should only be set when run on src host */
1529
    if (flags & QEMU_MIGRATION_COMPLETED_CHECK_STORAGE &&
1530
        qemuMigrationSrcNBDStorageCopyReady(vm, asyncJob) < 0)
1531 1532
        goto error;

1533
    if (flags & QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR &&
1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546
        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;
    }

1547 1548 1549 1550 1551 1552 1553 1554 1555 1556
    /* 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;
    }

1557 1558 1559 1560 1561
    /* 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 &&
1562
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
1563 1564 1565 1566
        VIR_DEBUG("Migration switched to post-copy");
        return 1;
    }

1567
    if (jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
1568 1569 1570 1571 1572
        return 1;
    else
        return 0;

 error:
1573 1574 1575
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1576
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1577
        /* The migration was aborted by us rather than QEMU itself. */
1578
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1579
        return -2;
1580 1581 1582

    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
        /* Something failed after QEMU already finished the migration. */
1583
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1584
        return -1;
1585 1586 1587 1588

    case QEMU_DOMAIN_JOB_STATUS_FAILED:
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
        /* QEMU aborted the migration. */
1589
        return -1;
1590 1591 1592 1593 1594 1595

    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_NONE:
        /* Impossible. */
        break;
1596
    }
1597 1598

    return -1;
1599 1600 1601
}


1602 1603 1604
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
1605
static int
1606 1607 1608 1609 1610
qemuMigrationSrcWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  virConnectPtr dconn,
                                  unsigned int flags)
1611
{
1612
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
1613
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1614
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1615
    int rv;
1616

1617 1618
    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_MIGRATING;

1619 1620
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           dconn, flags)) != 1) {
1621 1622
        if (rv < 0)
            return rv;
1623

1624 1625
        if (events) {
            if (virDomainObjWait(vm) < 0) {
1626 1627
                if (virDomainObjIsActive(vm))
                    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1628 1629 1630 1631 1632 1633 1634 1635 1636 1637
                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);
        }
1638 1639
    }

1640
    if (events)
1641
        ignore_value(qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, NULL));
1642

1643
    qemuDomainJobInfoUpdateTime(jobInfo);
1644 1645
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
1646
    if (VIR_ALLOC(priv->job.completed) == 0) {
1647
        *priv->job.completed = *jobInfo;
1648 1649
        priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
    }
1650

1651 1652 1653 1654
    if (asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT &&
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;

1655
    return 0;
1656 1657 1658
}


1659
static int
1660 1661 1662 1663
qemuMigrationDstWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  bool postcopy)
1664 1665
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1666
    unsigned int flags = 0;
1667 1668 1669 1670 1671 1672 1673
    int rv;

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

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

1674 1675 1676
    if (postcopy)
        flags = QEMU_MIGRATION_COMPLETED_POSTCOPY;

1677 1678
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           NULL, flags)) != 1) {
1679 1680 1681 1682 1683 1684 1685 1686
        if (rv < 0 || virDomainObjWait(vm) < 0)
            return -1;
    }

    return 0;
}


1687
static int
1688 1689 1690 1691
qemuMigrationSrcGraphicsRelocate(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuMigrationCookiePtr cookie,
                                 const char *graphicsuri)
1692 1693
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1694 1695
    int ret = -1;
    const char *listenAddress = NULL;
1696
    virSocketAddr addr;
1697 1698 1699 1700 1701
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
1702

1703
    if (!cookie || (!cookie->graphics && !graphicsuri))
1704 1705
        return 0;

1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724
    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) {
1725
        size_t i;
1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752

        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;
            }
        }
    }
1753 1754 1755 1756

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
1757 1758 1759 1760
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
1761

1762 1763 1764 1765 1766 1767 1768 1769 1770
    /* 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;
    }

1771 1772 1773 1774
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
1775
        priv->job.spiceMigration = !ret;
1776 1777
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
1778
    }
1779

1780
 cleanup:
1781
    virURIFree(uri);
1782 1783 1784 1785
    return ret;
}


1786
static int
1787 1788 1789
qemuMigrationDstOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            qemuMigrationCookiePtr cookie)
1790
{
1791 1792
    virDomainNetDefPtr netptr;
    int ret = -1;
1793
    size_t i;
1794 1795 1796 1797 1798 1799 1800 1801

    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:
1802
           break;
1803
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
1804 1805
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
1806 1807 1808
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
1809 1810 1811
                goto cleanup;
            }
            break;
1812 1813 1814 1815 1816
        default:
            break;
        }
    }

1817
    ret = 0;
1818
 cleanup:
1819 1820 1821 1822
    return ret;
}


1823
int
1824 1825
qemuMigrationDstCheckProtocol(virQEMUCapsPtr qemuCaps,
                              const char *migrateFrom)
1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848
{
    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 *
1849 1850
qemuMigrationDstGetURI(const char *migrateFrom,
                       int migrateFd)
1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862
{
    char *uri = NULL;

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

    return uri;
}


1863
int
1864 1865 1866 1867
qemuMigrationDstRun(virQEMUDriverPtr driver,
                    virDomainObjPtr vm,
                    const char *uri,
                    qemuDomainAsyncJob asyncJob)
1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
{
    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) {
1884
        /* qemuMigrationDstWaitForCompletion is called from the Finish phase */
1885 1886 1887 1888
        ret = 0;
        goto cleanup;
    }

1889
    if (qemuMigrationDstWaitForCompletion(driver, vm, asyncJob, false) < 0)
1890 1891 1892 1893 1894 1895 1896 1897 1898
        goto cleanup;

    ret = 0;

 cleanup:
    return ret;
}


1899 1900 1901 1902 1903 1904
/* 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.
 */
1905
static void
1906 1907 1908
qemuMigrationSrcCleanup(virDomainObjPtr vm,
                        virConnectPtr conn,
                        void *opaque)
1909
{
1910
    virQEMUDriverPtr driver = opaque;
1911 1912 1913 1914 1915 1916 1917 1918 1919
    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))
1920
        return;
1921 1922 1923 1924 1925

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

1926
    switch ((qemuMigrationJobPhase) priv->job.phase) {
1927 1928 1929 1930 1931 1932 1933 1934 1935
    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);
1936
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
1937
                                 priv->job.migParams, priv->job.apiFlags);
1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959
        /* 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 */
        ;
    }
}

1960

1961
/* The caller is supposed to lock the vm and start a migration job. */
1962
static char *
1963 1964 1965 1966 1967 1968 1969 1970 1971
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)
1972 1973 1974
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
1975
    virDomainDefPtr def = NULL;
1976
    qemuDomainObjPrivatePtr priv = vm->privateData;
1977
    virCapsPtr caps = NULL;
1978
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
1979

1980
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
1981
              " cookieout=%p, cookieoutlen=%p,"
1982
              " nmigrate_disks=%zu, migrate_disks=%p, flags=0x%lx",
1983
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
1984 1985
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
1986

1987 1988 1989
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

1990 1991 1992 1993 1994 1995
    /* 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);
1996

1997
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
1998 1999
        goto cleanup;

2000
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
2001
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
2002 2003
        goto cleanup;

2004 2005 2006 2007 2008 2009 2010 2011 2012
    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;
    }

2013 2014 2015 2016 2017 2018
    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;
    }

2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061
    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;
            }
2062 2063 2064
        }
    }

2065
    if (virDomainDefHasMemoryHotplug(vm->def) ||
2066
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
2067
         vm->newDef && virDomainDefHasMemoryHotplug(vm->newDef)))
2068 2069
        cookieFlags |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

2070 2071 2072 2073 2074
    if (!qemuDomainVcpuHotplugIsInOrder(vm->def) ||
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
         vm->newDef && !qemuDomainVcpuHotplugIsInOrder(vm->newDef)))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU_HOTPLUG;

2075 2076 2077
    if (priv->origCPU)
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU;

2078 2079
    cookieFlags |= QEMU_MIGRATION_COOKIE_ALLOW_REBOOT;

2080 2081 2082
    if (!(flags & VIR_MIGRATE_OFFLINE))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CAPS;

2083 2084
    if (!(mig = qemuMigrationEatCookie(driver, vm->def,
                                       priv->origname, priv, NULL, 0, 0)))
2085 2086 2087
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
2088
                                QEMU_MIGRATION_SOURCE,
2089
                                cookieout, cookieoutlen,
2090
                                cookieFlags) < 0)
2091 2092
        goto cleanup;

L
liguang 已提交
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114
    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;
        }
    }

2115
    if (xmlin) {
2116
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt, priv->qemuCaps,
2117 2118
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                            VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2119 2120
            goto cleanup;

2121
        if (!qemuDomainCheckABIStability(driver, vm, def))
2122 2123
            goto cleanup;

2124
        rv = qemuDomainDefFormatLive(driver, def, NULL, false, true);
2125
    } else {
2126 2127
        rv = qemuDomainDefFormatLive(driver, vm->def, priv->origCPU,
                                     false, true);
2128
    }
2129

2130
 cleanup:
2131
    qemuMigrationCookieFree(mig);
2132
    virObjectUnref(caps);
2133
    virDomainDefFree(def);
2134 2135 2136
    return rv;
}

2137
char *
2138 2139 2140 2141 2142 2143 2144 2145 2146
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)
2147 2148 2149
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
2150
    qemuDomainAsyncJob asyncJob;
2151 2152

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
2153 2154
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
2155 2156 2157 2158 2159 2160 2161 2162
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_MIGRATION_OUT;
    } else {
        if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_NONE;
    }

2163
    qemuMigrationSrcStoreDomainState(vm);
2164

2165
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
2166 2167 2168 2169 2170 2171
        goto endjob;

    /* Check if there is any ejected media.
     * We don't want to require them on the destination.
     */
    if (!(flags & VIR_MIGRATE_OFFLINE) &&
2172
        qemuProcessRefreshDisks(driver, vm, asyncJob) < 0)
2173 2174
        goto endjob;

2175 2176 2177
    if (!(xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                           cookieout, cookieoutlen,
                                           nmigrate_disks, migrate_disks, flags)))
2178 2179 2180 2181 2182 2183 2184
        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.
         */
2185
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
2186
                                 qemuMigrationSrcCleanup) < 0) {
2187
            VIR_FREE(xml);
2188
            goto endjob;
2189
        }
2190
        qemuMigrationJobContinue(vm);
2191 2192 2193 2194
    } else {
        goto endjob;
    }

2195
 cleanup:
M
Michal Privoznik 已提交
2196
    virDomainObjEndAPI(&vm);
2197 2198
    return xml;

2199
 endjob:
2200 2201 2202 2203
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2204 2205 2206
    goto cleanup;
}

2207

2208 2209
/* Prepare is the first step, and it runs on the destination host.
 */
2210

2211
static void
2212 2213
qemuMigrationDstPrepareCleanup(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
2214 2215 2216 2217 2218 2219 2220 2221 2222
{
    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));

2223
    virPortAllocatorRelease(priv->migrationPort);
2224 2225
    priv->migrationPort = 0;

2226 2227 2228 2229 2230
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2231
static qemuProcessIncomingDefPtr
2232 2233 2234 2235 2236 2237
qemuMigrationDstPrepare(virDomainObjPtr vm,
                        bool tunnel,
                        const char *protocol,
                        const char *listenAddress,
                        unsigned short port,
                        int fd)
2238 2239
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2240
    qemuProcessIncomingDefPtr inc = NULL;
2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270
    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 */
            }
2271
        } else if (hostIPv6Capable) {
2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294
            /* 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;
    }

2295 2296
    inc = qemuProcessIncomingDefNew(priv->qemuCaps, listenAddress,
                                    migrateFrom, fd, NULL);
2297

2298
 cleanup:
2299 2300
    VIR_FREE(migrateFrom);
    return inc;
2301 2302
}

2303
static int
2304
qemuMigrationDstPrepareAny(virQEMUDriverPtr driver,
2305
                           virConnectPtr dconn,
2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
                           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,
2320
                           qemuMigrationParamsPtr migParams,
2321
                           unsigned long flags)
2322 2323
{
    virDomainObjPtr vm = NULL;
2324
    virObjectEventPtr event = NULL;
2325
    virErrorPtr origErr;
2326
    int ret = -1;
2327
    int dataFD[2] = { -1, -1 };
2328
    qemuDomainObjPrivatePtr priv = NULL;
2329
    qemuMigrationCookiePtr mig = NULL;
2330
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2331
    char *xmlout = NULL;
L
liguang 已提交
2332
    unsigned int cookieFlags;
2333
    unsigned int startFlags;
2334
    virCapsPtr caps = NULL;
2335
    qemuProcessIncomingDefPtr incoming = NULL;
2336
    bool taint_hook = false;
2337 2338 2339
    bool stopProcess = false;
    bool relabel = false;
    int rv;
2340
    char *tlsAlias = NULL;
2341

2342 2343
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363
    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;
        }
2364 2365
        cookieFlags = 0;
    } else {
2366 2367
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS |
                      QEMU_MIGRATION_COOKIE_CAPS;
L
liguang 已提交
2368 2369
    }

2370 2371 2372 2373 2374 2375 2376 2377 2378
    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;
    }

2379 2380 2381 2382 2383 2384
    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;
    }

2385 2386 2387
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2388
    if (!qemuMigrationSrcIsAllowedHostdev(*def))
2389 2390
        goto cleanup;

J
Jiri Denemark 已提交
2391 2392 2393 2394 2395
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2396
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2397 2398
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2399 2400
            goto cleanup;

2401
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2402 2403 2404 2405 2406 2407 2408
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2409
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2410 2411 2412 2413 2414 2415
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2416
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt, NULL,
2417 2418
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                                 VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE);
J
Jiri Denemark 已提交
2419 2420 2421
                if (!newdef)
                    goto cleanup;

2422
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2423 2424 2425 2426
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2427 2428
                virDomainDefFree(*def);
                *def = newdef;
2429 2430 2431 2432
                /* 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
2433 2434 2435 2436
            }
        }
    }

2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450
    /* 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;

2451
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2452
                                   driver->xmlopt,
2453 2454 2455
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2456
        goto cleanup;
2457
    *def = NULL;
2458

2459
    priv = vm->privateData;
2460 2461
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2462

2463 2464 2465 2466 2467
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2468 2469
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
2470 2471 2472 2473 2474 2475
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2476 2477 2478
    if (qemuMigrationDstPrecreateStorage(vm, mig->nbd,
                                         nmigrate_disks, migrate_disks,
                                         !!(flags & VIR_MIGRATE_NON_SHARED_INC)) < 0)
2479 2480
        goto cleanup;

2481 2482
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                              flags) < 0)
2483
        goto cleanup;
2484
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2485 2486 2487 2488

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

L
liguang 已提交
2489 2490 2491
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2492 2493
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
2494 2495
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
2496
        goto stopjob;
2497 2498
    }

2499 2500
    startFlags = VIR_QEMU_PROCESS_START_AUTODESTROY;

2501
    if (qemuProcessInit(driver, vm, mig->cpu, QEMU_ASYNC_JOB_MIGRATION_IN,
2502
                        true, startFlags) < 0)
2503
        goto stopjob;
2504
    stopProcess = true;
2505

2506 2507
    priv->allowReboot = mig->allowReboot;

2508 2509 2510
    if (!(incoming = qemuMigrationDstPrepare(vm, tunnel, protocol,
                                             listenAddress, port,
                                             dataFD[0])))
2511
        goto stopjob;
2512

2513
    if (qemuProcessPrepareDomain(driver, vm, startFlags) < 0)
2514 2515
        goto stopjob;

2516
    if (qemuProcessPrepareHost(driver, vm, startFlags) < 0)
2517 2518
        goto stopjob;

2519
    rv = qemuProcessLaunch(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2520 2521
                           incoming, NULL,
                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
2522
                           startFlags);
2523 2524 2525
    if (rv < 0) {
        if (rv == -2)
            relabel = true;
2526
        goto stopjob;
2527
    }
2528
    relabel = true;
2529

2530 2531 2532 2533
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
2534
            goto stopjob;
2535
        }
2536
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
2537 2538
    }

2539
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
2540
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
2541
        goto stopjob;
M
Michael R. Hines 已提交
2542 2543
    }

2544
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2545
                                 migParams, mig->caps->automatic) < 0)
2546 2547 2548 2549 2550 2551 2552
        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,
2553
                                         &tlsAlias, NULL,
2554 2555 2556 2557 2558 2559 2560
                                         migParams) < 0)
            goto stopjob;
    } else {
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
            goto stopjob;
    }

2561 2562
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                                 migParams) < 0)
2563 2564
        goto stopjob;

2565 2566 2567
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579
        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;
        }

2580 2581
        if (qemuMigrationDstStartNBDServer(driver, vm, incoming->address,
                                           nmigrate_disks, migrate_disks,
2582
                                           nbdPort, nbdTLSAlias) < 0) {
2583
            goto stopjob;
2584
        }
2585
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
2586 2587
    }

2588 2589 2590 2591 2592 2593 2594 2595 2596
    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");
    }

2597
    if (incoming->deferredURI &&
2598 2599
        qemuMigrationDstRun(driver, vm, incoming->deferredURI,
                            QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2600 2601
        goto stopjob;

2602
    if (qemuProcessFinishStartup(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2603 2604 2605
                                 false, VIR_DOMAIN_PAUSED_MIGRATION) < 0)
        goto stopjob;

2606
 done:
2607 2608 2609
    if (qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_DESTINATION,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
2610 2611 2612 2613 2614 2615 2616
        /* 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");
    }

2617
    if (qemuDomainCleanupAdd(vm, qemuMigrationDstPrepareCleanup) < 0)
2618
        goto stopjob;
2619

L
liguang 已提交
2620 2621
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
2622
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
2623 2624 2625
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
2626

2627 2628 2629 2630
    /* 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.
     */
2631
    qemuMigrationJobContinue(vm);
2632

2633 2634
    if (autoPort)
        priv->migrationPort = port;
2635 2636 2637 2638 2639
    /* 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;
2640
    ret = 0;
2641

2642
 cleanup:
2643
    virErrorPreserveLast(&origErr);
2644
    VIR_FREE(tlsAlias);
2645
    qemuProcessIncomingDefFree(incoming);
J
Jiri Denemark 已提交
2646
    VIR_FREE(xmlout);
2647 2648
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
2649 2650 2651
    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 */
2652
        VIR_FREE(priv->origname);
2653 2654 2655 2656
        /* release if port is auto selected which is not the case if
         * it is given in parameters
         */
        if (nbdPort == 0)
2657
            virPortAllocatorRelease(priv->nbdPort);
2658
        priv->nbdPort = 0;
2659
        virDomainObjRemoveTransientDef(vm);
2660
        qemuDomainRemoveInactiveJob(driver, vm);
2661
    }
M
Michal Privoznik 已提交
2662
    virDomainObjEndAPI(&vm);
2663
    virObjectEventStateQueue(driver->domainEventState, event);
2664
    qemuMigrationCookieFree(mig);
2665
    virObjectUnref(caps);
2666
    virNWFilterUnlockFilterUpdates();
2667
    virErrorRestore(&origErr);
2668
    return ret;
2669

2670
 stopjob:
2671
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2672
                             priv->job.migParams, priv->job.apiFlags);
2673

2674 2675 2676 2677
    if (stopProcess) {
        unsigned int stopFlags = VIR_QEMU_PROCESS_STOP_MIGRATED;
        if (!relabel)
            stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
2678
        virDomainAuditStart(vm, "migrated", false);
2679 2680
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        QEMU_ASYNC_JOB_MIGRATION_IN, stopFlags);
2681
    }
2682

2683
    qemuMigrationJobFinish(driver, vm);
2684
    goto cleanup;
2685 2686 2687
}


2688 2689 2690 2691 2692
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
2693
qemuMigrationDstPrepareTunnel(virQEMUDriverPtr driver,
2694
                              virConnectPtr dconn,
2695 2696 2697 2698 2699 2700 2701
                              const char *cookiein,
                              int cookieinlen,
                              char **cookieout,
                              int *cookieoutlen,
                              virStreamPtr st,
                              virDomainDefPtr *def,
                              const char *origname,
2702
                              qemuMigrationParamsPtr migParams,
2703
                              unsigned long flags)
2704
{
2705
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2706
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
2707
              "origname=%s, flags=0x%lx",
2708
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2709
              cookieout, cookieoutlen, st, *def, origname, flags);
2710

2711 2712 2713 2714 2715 2716
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

2717 2718 2719
    return qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
                                      cookieout, cookieoutlen, def, origname,
                                      st, NULL, 0, false, NULL, 0, NULL, 0,
2720
                                      migParams, flags);
2721 2722 2723
}


2724
static virURIPtr
2725
qemuMigrationAnyParseURI(const char *uri, bool *wellFormed)
2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746
{
    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;
}


2747
int
2748
qemuMigrationDstPrepareDirect(virQEMUDriverPtr driver,
2749
                              virConnectPtr dconn,
2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761
                              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,
2762
                              qemuMigrationParamsPtr migParams,
2763
                              unsigned long flags)
2764
{
2765 2766
    unsigned short port = 0;
    bool autoPort = true;
2767 2768
    char *hostname = NULL;
    int ret = -1;
2769
    virURIPtr uri = NULL;
2770 2771
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
2772

2773
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2774
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
2775
              "def=%p, origname=%s, listenAddress=%s, "
2776
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, flags=0x%lx",
2777
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2778
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
2779
              *def, origname, NULLSTR(listenAddress),
2780
              nmigrate_disks, migrate_disks, nbdPort, flags);
2781

2782 2783
    *uri_out = NULL;

2784 2785 2786
    /* 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
2787 2788 2789
     * 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".
2790 2791 2792 2793 2794 2795
     *
     * 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) {
2796 2797 2798
        bool encloseAddress = false;
        const char *incFormat;

2799
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2800
            goto cleanup;
2801

2802
        if (migrateHost != NULL) {
2803 2804
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
2805

2806
            if (VIR_STRDUP(hostname, migrateHost) < 0)
2807 2808 2809 2810 2811
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
2812 2813

        if (STRPREFIX(hostname, "localhost")) {
2814 2815 2816
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
2817 2818 2819 2820 2821
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
2822
         * compatibility with old targets. We at least make the
2823 2824
         * new targets accept both syntaxes though.
         */
2825 2826 2827 2828 2829 2830
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
2831 2832
            goto cleanup;
    } else {
2833
        bool well_formed_uri;
J
Ján Tomko 已提交
2834

2835
        if (!(uri = qemuMigrationAnyParseURI(uri_in, &well_formed_uri)))
2836
            goto cleanup;
J
Ján Tomko 已提交
2837

2838 2839 2840 2841 2842 2843 2844
        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
2845 2846
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
2847 2848 2849
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
2850 2851 2852 2853 2854 2855 2856 2857 2858 2859
            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) {
2860
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2861
                goto cleanup;
2862

2863
            /* Send well-formed URI only if uri_in was well-formed */
2864 2865 2866 2867 2868 2869 2870 2871
            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;
            }
2872
        } else {
2873 2874
            port = uri->port;
            autoPort = false;
2875 2876 2877 2878 2879 2880
        }
    }

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

2881
    ret = qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
2882 2883 2884 2885
                                     cookieout, cookieoutlen, def, origname,
                                     NULL, uri ? uri->scheme : "tcp",
                                     port, autoPort, listenAddress,
                                     nmigrate_disks, migrate_disks, nbdPort,
2886
                                     migParams, flags);
2887
 cleanup:
2888
    virURIFree(uri);
2889
    VIR_FREE(hostname);
2890
    virObjectUnref(cfg);
2891
    if (ret != 0) {
2892
        VIR_FREE(*uri_out);
2893
        if (autoPort)
2894
            virPortAllocatorRelease(port);
2895
    }
2896 2897 2898 2899
    return ret;
}


2900
virDomainDefPtr
2901 2902 2903 2904
qemuMigrationAnyPrepareDef(virQEMUDriverPtr driver,
                           const char *dom_xml,
                           const char *dname,
                           char **origname)
2905 2906 2907
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
2908
    char *name = NULL;
2909 2910 2911 2912 2913 2914 2915 2916 2917 2918

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

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

2919
    if (!(def = virDomainDefParseString(dom_xml, caps, driver->xmlopt, NULL,
2920 2921
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                        VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2922 2923 2924
        goto cleanup;

    if (dname) {
2925
        name = def->name;
2926 2927 2928 2929 2930 2931
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

2932
 cleanup:
2933
    virObjectUnref(caps);
2934 2935 2936 2937
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
2938 2939 2940 2941
    return def;
}


2942
static int
2943 2944 2945 2946 2947 2948
qemuMigrationSrcConfirmPhase(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
                             const char *cookiein,
                             int cookieinlen,
                             unsigned int flags,
                             int retcode)
2949 2950
{
    qemuMigrationCookiePtr mig;
2951
    virObjectEventPtr event;
2952 2953
    int rv = -1;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2954 2955
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = NULL;
2956

2957
    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
2958
              "flags=0x%x, retcode=%d",
2959
              driver, vm, NULLSTR(cookiein), cookieinlen,
2960 2961 2962 2963 2964 2965 2966 2967 2968
              flags, retcode);

    virCheckFlags(QEMU_MIGRATION_FLAGS, -1);

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

2969 2970
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen,
2971
                                       QEMU_MIGRATION_COOKIE_STATS)))
2972 2973
        goto cleanup;

2974 2975 2976
    if (retcode == 0)
        jobInfo = priv->job.completed;
    else
2977
        VIR_FREE(priv->job.completed);
2978 2979 2980

    /* Update times with the values sent by the destination daemon */
    if (mig->jobInfo && jobInfo) {
2981 2982 2983 2984 2985 2986 2987 2988
        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 &&
2989 2990
            qemuMigrationAnyFetchStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                       jobInfo, NULL) < 0)
2991 2992
            VIR_WARN("Could not refresh migration statistics");

2993 2994 2995
        qemuDomainJobInfoUpdateTime(jobInfo);
        jobInfo->timeDeltaSet = mig->jobInfo->timeDeltaSet;
        jobInfo->timeDelta = mig->jobInfo->timeDelta;
2996 2997
        jobInfo->stats.mig.downtime_set = mig->jobInfo->stats.mig.downtime_set;
        jobInfo->stats.mig.downtime = mig->jobInfo->stats.mig.downtime;
2998 2999
    }

3000 3001 3002
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

3003 3004
    /* 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.
3005 3006 3007 3008
     */
    if (retcode == 0) {
        /* If guest uses SPICE and supports seamless migration we have to hold
         * up domain shutdown until SPICE server transfers its data */
3009
        qemuMigrationSrcWaitForSpice(vm);
3010 3011

        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
3012
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
3013 3014 3015
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
        virDomainAuditStop(vm, "migrated");

3016
        event = virDomainEventLifecycleNewFromObj(vm,
3017 3018
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
3019
        virObjectEventStateQueue(driver->domainEventState, event);
3020
        qemuDomainEventEmitJobCompleted(driver, vm);
3021
    } else {
3022
        virErrorPtr orig_err = virSaveLastError();
3023
        int reason;
3024 3025

        /* cancel any outstanding NBD jobs */
3026 3027
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT, NULL);
3028 3029 3030

        virSetError(orig_err);
        virFreeError(orig_err);
3031

3032
        if (virDomainObjGetState(vm, &reason) == VIR_DOMAIN_PAUSED &&
3033
            reason == VIR_DOMAIN_PAUSED_POSTCOPY)
3034
            qemuMigrationAnyPostcopyFailed(driver, vm);
3035 3036
        else
            qemuMigrationSrcRestoreDomainState(driver, vm);
3037

3038
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3039
                                 priv->job.migParams, priv->job.apiFlags);
3040

3041
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
3042 3043 3044
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
    }

3045
 done:
3046 3047 3048
    qemuMigrationCookieFree(mig);
    rv = 0;

3049
 cleanup:
3050 3051 3052 3053 3054
    virObjectUnref(cfg);
    return rv;
}

int
3055 3056 3057 3058 3059 3060
qemuMigrationSrcConfirm(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *cookiein,
                        int cookieinlen,
                        unsigned int flags,
                        int cancelled)
3061
{
3062
    qemuMigrationJobPhase phase;
3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076
    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);
3077
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
3078
                           qemuMigrationSrcCleanup);
3079

3080 3081 3082
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
3083

3084
    qemuMigrationJobFinish(driver, vm);
3085
    if (!virDomainObjIsActive(vm)) {
3086
        if (!cancelled && ret == 0 && flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
3087
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
3088 3089
            vm->persistent = 0;
        }
3090
        qemuDomainRemoveInactiveJob(driver, vm);
3091 3092
    }

3093
 cleanup:
M
Michal Privoznik 已提交
3094
    virDomainObjEndAPI(&vm);
3095 3096 3097 3098 3099
    virObjectUnref(cfg);
    return ret;
}


3100 3101
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3102
    MIGRATION_DEST_CONNECT_HOST,
3103
    MIGRATION_DEST_FD,
3104
};
3105

3106 3107 3108 3109
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3110

3111 3112 3113 3114 3115 3116
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3117
            const char *protocol;
3118 3119 3120 3121
            const char *name;
            int port;
        } host;

3122 3123 3124 3125
        struct {
            int qemu;
            int local;
        } fd;
3126 3127 3128 3129 3130 3131 3132
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3133 3134 3135

#define TUNNEL_SEND_BUF_SIZE 65536

3136 3137 3138 3139 3140 3141 3142
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3143 3144
    int wakeupRecvFD;
    int wakeupSendFD;
3145 3146
};

3147
static void qemuMigrationSrcIOFunc(void *arg)
3148
{
3149
    qemuMigrationIOThreadPtr data = arg;
3150 3151 3152 3153 3154 3155 3156
    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);
3157

3158
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3159
        goto abrt;
3160

3161 3162 3163
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3164
    for (;;) {
3165 3166 3167 3168 3169 3170 3171 3172 3173 3174
        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;
3175
            virReportSystemError(errno, "%s",
3176 3177
                                 _("poll failed in migration tunnel"));
            goto abrt;
3178
        }
3179 3180 3181 3182 3183 3184 3185

        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");
3186
            break;
3187
        }
3188

3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204
        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;
            }
3205 3206
        }

3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223
        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;
            }
        }
    }
3224

3225 3226
    if (virStreamFinish(data->st) < 0)
        goto error;
3227

3228
    VIR_FORCE_CLOSE(data->sock);
3229 3230
    VIR_FREE(buffer);

3231 3232
    return;

3233
 abrt:
3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3245
 error:
3246 3247 3248 3249 3250
    /* 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);
3251
    virResetLastError();
3252
    VIR_FREE(buffer);
3253 3254 3255 3256
}


static qemuMigrationIOThreadPtr
3257 3258
qemuMigrationSrcStartTunnel(virStreamPtr st,
                            int sock)
3259
{
3260 3261
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3262

3263 3264 3265 3266
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3267 3268
    }

3269
    if (VIR_ALLOC(io) < 0)
3270
        goto error;
3271

3272 3273
    io->st = st;
    io->sock = sock;
3274 3275
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3276 3277

    if (virThreadCreate(&io->thread, true,
3278
                        qemuMigrationSrcIOFunc,
3279 3280 3281
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3282
        goto error;
3283 3284 3285
    }

    return io;
3286

3287
 error:
3288 3289 3290 3291
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3292 3293 3294
}

static int
3295
qemuMigrationSrcStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3296 3297
{
    int rv = -1;
3298 3299 3300 3301 3302 3303 3304 3305 3306
    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;
    }

3307 3308 3309 3310
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3311 3312 3313 3314
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3315 3316 3317 3318 3319 3320
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3321
 cleanup:
3322 3323
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3324 3325
    VIR_FREE(io);
    return rv;
3326 3327
}

3328
static int
3329 3330 3331
qemuMigrationSrcConnect(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationSpecPtr spec)
3332 3333 3334 3335 3336 3337 3338
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3339
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3340 3341 3342 3343 3344
        return -1;

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

3345
    if (qemuSecuritySetSocketLabel(driver->securityManager, vm->def) < 0)
3346
        goto cleanup;
3347 3348 3349
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
3350
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
3351
        virObjectUnref(sock);
3352
    }
3353
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0 ||
3354 3355 3356
        spec->dest.fd.qemu == -1)
        goto cleanup;

3357 3358 3359 3360 3361 3362 3363
    /* 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;
    }

3364 3365
    ret = 0;

3366
 cleanup:
3367 3368 3369 3370 3371 3372
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3373 3374

static int
3375 3376 3377 3378
qemuMigrationSrcContinue(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMonitorMigrationStatus status,
                         qemuDomainAsyncJob asyncJob)
3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394
{
    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;
}


3395
static int
3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409
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,
3410
                    qemuMigrationParamsPtr migParams)
3411
{
3412
    int ret = -1;
3413 3414
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3415
    qemuMigrationCookiePtr mig = NULL;
3416
    char *tlsAlias = NULL;
3417
    qemuMigrationIOThreadPtr iothread = NULL;
3418
    int fd = -1;
3419
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3420
    virErrorPtr orig_err = NULL;
3421
    unsigned int cookieFlags = 0;
3422
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3423
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
3424
    bool cancel = false;
3425
    unsigned int waitFlags;
3426
    virDomainDefPtr persistDef = NULL;
3427
    char *timestamp;
3428
    int rc;
3429 3430

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
3431
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3432 3433
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
3434 3435
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3436
              spec, spec->destType, spec->fwdType, dconn,
3437
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3438

3439 3440 3441 3442 3443 3444 3445 3446 3447 3448
    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;
    }

3449 3450
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3451 3452 3453 3454
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3455 3456 3457
        return -1;
    }

3458 3459 3460
    if (events)
        priv->signalIOError = abort_on_error;

3461 3462
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
        if (persist_xml) {
3463 3464
            if (!(persistDef = qemuMigrationAnyPrepareDef(driver, persist_xml,
                                                          NULL, NULL)))
3465
                goto error;
3466 3467 3468
        } else {
            virDomainDefPtr def = vm->newDef ? vm->newDef : vm->def;
            if (!(persistDef = qemuDomainDefCopy(driver, def,
3469 3470
                                                 VIR_DOMAIN_XML_SECURE |
                                                 VIR_DOMAIN_XML_MIGRATABLE)))
3471
                goto error;
3472 3473 3474
        }
    }

3475 3476
    mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                 cookiein, cookieinlen,
3477 3478 3479
                                 cookieFlags |
                                 QEMU_MIGRATION_COOKIE_GRAPHICS |
                                 QEMU_MIGRATION_COOKIE_CAPS);
3480
    if (!mig)
3481
        goto error;
3482

3483
    if (qemuMigrationSrcGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3484 3485
        VIR_WARN("unable to provide data for graphics client relocation");

3486
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3487
                                 migParams, mig->caps->automatic) < 0)
3488 3489
        goto error;

3490
    if (flags & VIR_MIGRATE_TLS) {
3491
        const char *hostname = NULL;
3492 3493 3494 3495

        /* We need to add tls-hostname whenever QEMU itself does not
         * connect directly to the destination. */
        if (spec->destType == MIGRATION_DEST_CONNECT_HOST ||
3496 3497 3498
            spec->destType == MIGRATION_DEST_FD)
            hostname = spec->dest.host.name;

3499
        if (qemuMigrationParamsEnableTLS(driver, vm, false,
3500
                                         QEMU_ASYNC_JOB_MIGRATION_OUT,
3501
                                         &tlsAlias, hostname,
3502 3503
                                         migParams) < 0)
            goto error;
3504
    } else {
3505
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
3506
            goto error;
3507 3508
    }

3509 3510 3511 3512
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                 migParams) < 0)
        goto error;

3513 3514 3515
    if (migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                         QEMU_MONITOR_MIGRATE_NON_SHARED_INC)) {
        if (mig->nbd) {
3516 3517 3518 3519
            /* 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. */
3520 3521
            if (flags & VIR_MIGRATE_TLS &&
                !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV_DEL)) {
3522 3523 3524 3525 3526
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                               _("NBD migration with TLS is not supported"));
                goto error;
            }

3527
            /* This will update migrate_flags on success */
3528 3529 3530 3531 3532 3533
            if (qemuMigrationSrcNBDStorageCopy(driver, vm, mig,
                                               spec->dest.host.name,
                                               migrate_speed,
                                               &migrate_flags,
                                               nmigrate_disks,
                                               migrate_disks,
3534
                                               dconn, tlsAlias, flags) < 0) {
3535
                goto error;
3536 3537 3538 3539 3540 3541 3542
            }
        } 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.");
        }
3543 3544
    }

3545
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3546 3547
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
3548
        if (qemuMigrationSrcSetOffline(driver, vm) < 0)
3549
            goto error;
3550 3551
    }

3552 3553
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3554
        goto error;
3555

3556
    if (priv->job.abortJob) {
3557 3558
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
3559
         * priv->job.abortJob will not change */
3560
        priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
3561 3562 3563
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
3564
        goto exit_monitor;
3565 3566
    }

3567 3568
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
3569

3570 3571
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3572
        qemuMigrationSrcConnect(driver, vm, spec) < 0) {
3573
        goto exit_monitor;
3574
    }
3575

3576 3577 3578 3579 3580 3581
    /* log start of migration */
    if ((timestamp = virTimeStringNow()) != NULL) {
        qemuDomainLogAppendMessage(driver, vm, "%s: initiating migration\n", timestamp);
        VIR_FREE(timestamp);
    }

3582
    rc = -1;
3583 3584
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3585 3586
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
3587
            goto exit_monitor;
M
Michael R. Hines 已提交
3588
        }
3589 3590 3591 3592
        rc = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
                                      spec->dest.host.protocol,
                                      spec->dest.host.name,
                                      spec->dest.host.port);
3593 3594
        break;

3595 3596 3597 3598
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3599
    case MIGRATION_DEST_FD:
3600
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3601
            fd = spec->dest.fd.local;
3602 3603
            spec->dest.fd.local = -1;
        }
3604 3605
        rc = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                    spec->dest.fd.qemu);
3606 3607
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3608
    }
3609 3610

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
3611
        goto error;
3612 3613 3614

    /* From this point onwards we *must* call cancel to abort the
     * migration on source if anything goes wrong */
3615
    cancel = true;
3616

3617
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3618
        if (!(iothread = qemuMigrationSrcStartTunnel(spec->fwd.stream, fd)))
3619
            goto error;
3620 3621 3622 3623 3624
        /* If we've created a tunnel, then the 'fd' will be closed in the
         * qemuMigrationIOFunc as data->sock.
         */
        fd = -1;
    }
3625

3626
    waitFlags = QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;
3627 3628 3629 3630 3631 3632 3633
    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;

3634 3635 3636
    rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT,
                                           dconn, waitFlags);
3637
    if (rc == -2) {
3638
        goto error;
3639 3640 3641 3642 3643
    } else if (rc == -1) {
        /* QEMU reported failed migration, nothing to cancel anymore */
        cancel = false;
        goto error;
    }
3644

3645 3646 3647
    /* 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.
3648
     */
3649 3650 3651 3652 3653
    if (priv->monJSON) {
        while (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            priv->signalStop = true;
            rc = virDomainObjWait(vm);
            priv->signalStop = false;
3654
            if (rc < 0)
3655
                goto error;
3656
        }
3657
    } else if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING &&
3658
               qemuMigrationSrcSetOffline(driver, vm) < 0) {
3659
        goto error;
3660
    }
3661

J
Ján Tomko 已提交
3662
    if (mig->nbd &&
3663 3664 3665
        qemuMigrationSrcNBDCopyCancel(driver, vm, true,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT,
                                      dconn) < 0)
3666
        goto error;
3667

3668 3669 3670 3671 3672
    /* 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) {
3673 3674 3675
        if (qemuMigrationSrcContinue(driver, vm,
                                     QEMU_MONITOR_MIGRATION_STATUS_PRE_SWITCHOVER,
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3676 3677 3678 3679
            goto error;

        waitFlags ^= QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;

3680 3681 3682
        rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_OUT,
                                               dconn, waitFlags);
3683 3684 3685 3686 3687 3688 3689 3690 3691
        if (rc == -2) {
            goto error;
        } else if (rc == -1) {
            /* QEMU reported failed migration, nothing to cancel anymore */
            cancel = false;
            goto error;
        }
    }

3692 3693 3694 3695
    if (iothread) {
        qemuMigrationIOThreadPtr io;

        VIR_STEAL_PTR(io, iothread);
3696
        if (qemuMigrationSrcStopTunnel(io, false) < 0)
3697
            goto error;
3698 3699 3700
    }

    if (priv->job.completed) {
3701
        priv->job.completed->stopped = priv->job.current->stopped;
3702 3703 3704 3705 3706 3707 3708 3709 3710
        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 ||
3711 3712 3713
        qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_SOURCE,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
3714 3715
        VIR_WARN("Unable to encode migration cookie");
    }
3716

3717
    ret = 0;
3718

3719
 cleanup:
3720
    VIR_FREE(tlsAlias);
3721
    VIR_FORCE_CLOSE(fd);
3722
    virDomainDefFree(persistDef);
3723 3724
    qemuMigrationCookieFree(mig);

3725 3726 3727
    if (events)
        priv->signalIOError = false;

3728 3729 3730 3731 3732
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

3733 3734
    return ret;

3735
 error:
3736 3737
    orig_err = virSaveLastError();

3738 3739 3740 3741 3742 3743 3744 3745
    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));
        }
3746

3747 3748
        /* cancel any outstanding NBD jobs */
        if (mig && mig->nbd)
3749 3750 3751
            qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                          QEMU_ASYNC_JOB_MIGRATION_OUT,
                                          dconn);
3752 3753 3754 3755

        if (priv->job.current->status != QEMU_DOMAIN_JOB_STATUS_CANCELED)
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
    }
3756 3757

    if (iothread)
3758
        qemuMigrationSrcStopTunnel(iothread, true);
3759 3760 3761

    goto cleanup;

3762 3763
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
3764
    goto error;
3765 3766
}

3767
/* Perform migration using QEMU's native migrate support,
3768 3769
 * not encrypted obviously
 */
3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784
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,
3785
                              qemuMigrationParamsPtr migParams)
3786
{
3787
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
3788
    virURIPtr uribits = NULL;
3789
    int ret = -1;
3790 3791 3792
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
3793
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3794
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
3795
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
3796
              cookieout, cookieoutlen, flags, resource,
3797
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3798

3799
    if (!(uribits = qemuMigrationAnyParseURI(uri, NULL)))
3800 3801
        return -1;

3802 3803 3804 3805 3806 3807 3808
    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
3809 3810 3811 3812 3813
    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"));
3814
            goto cleanup;
M
Michael R. Hines 已提交
3815
        }
3816
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
3817 3818 3819
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
3820
            goto cleanup;
M
Michael R. Hines 已提交
3821 3822 3823
        }
    }

3824
    if (STRNEQ(uribits->scheme, "rdma"))
3825 3826
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
3827
        spec.destType = MIGRATION_DEST_HOST;
3828
    spec.dest.host.protocol = uribits->scheme;
3829 3830 3831
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
3832

3833 3834 3835
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen, cookieout,
                              cookieoutlen, flags, resource, &spec, dconn,
                              graphicsuri, nmigrate_disks, migrate_disks,
3836
                              migParams);
3837 3838 3839 3840

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

3841
 cleanup:
3842
    virURIFree(uribits);
3843 3844 3845 3846 3847

    return ret;
}


3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862
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,
3863
                              qemuMigrationParamsPtr migParams)
3864 3865 3866
{
    int ret = -1;
    qemuMigrationSpec spec;
3867
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3868
    int fds[2] = { -1, -1 };
3869 3870

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
3871
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3872
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
3873
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
3874
              cookieout, cookieoutlen, flags, resource,
3875
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3876 3877 3878 3879

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

3880

3881 3882 3883
    spec.destType = MIGRATION_DEST_FD;
    spec.dest.fd.qemu = -1;
    spec.dest.fd.local = -1;
3884

3885 3886 3887 3888 3889
    if (pipe2(fds, O_CLOEXEC) == 0) {
        spec.dest.fd.qemu = fds[1];
        spec.dest.fd.local = fds[0];
    }
    if (spec.dest.fd.qemu == -1 ||
3890 3891
        qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    spec.dest.fd.qemu) < 0) {
3892 3893 3894
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
        goto cleanup;
3895 3896
    }

3897 3898 3899
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen,
                              cookieout, cookieoutlen, flags, resource, &spec,
                              dconn, graphicsuri, nmigrate_disks, migrate_disks,
3900
                              migParams);
3901

3902
 cleanup:
3903 3904
    VIR_FORCE_CLOSE(spec.dest.fd.qemu);
    VIR_FORCE_CLOSE(spec.dest.fd.local);
3905

3906
    virObjectUnref(cfg);
3907 3908 3909 3910
    return ret;
}


3911 3912 3913 3914
/* 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 */
3915 3916 3917 3918 3919 3920 3921 3922
static int
qemuMigrationSrcPerformPeer2Peer2(virQEMUDriverPtr driver,
                                  virConnectPtr sconn,
                                  virConnectPtr dconn,
                                  virDomainObjPtr vm,
                                  const char *dconnuri,
                                  unsigned long flags,
                                  const char *dname,
3923 3924
                                  unsigned long resource,
                                  qemuMigrationParamsPtr migParams)
3925 3926 3927
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
3928
    char *cookie = NULL;
3929 3930 3931
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
3932
    bool cancelled;
3933
    virStreamPtr st = NULL;
3934
    unsigned long destflags;
3935

3936
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
3937
              "flags=0x%lx, dname=%s, resource=%lu",
3938 3939
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
3940

3941 3942 3943 3944 3945
    /* 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,
3946 3947
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
3948 3949 3950 3951 3952
        return -1;

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

3953 3954
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
3955

3956 3957 3958 3959 3960 3961 3962 3963 3964 3965
    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;

3966
        qemuDomainObjEnterRemote(vm);
3967
        ret = dconn->driver->domainMigratePrepareTunnel
3968
            (dconn, st, destflags, dname, resource, dom_xml);
3969 3970
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3971
    } else {
3972
        qemuDomainObjEnterRemote(vm);
3973 3974
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
3975
             destflags, dname, resource, dom_xml);
3976 3977
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3978 3979 3980
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
3981 3982
        goto cleanup;

3983 3984
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
3985 3986
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
3987
        cancelled = true;
3988
        orig_err = virSaveLastError();
3989
        goto finish;
3990 3991
    }

3992 3993 3994 3995
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
3996
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
3997
    if (flags & VIR_MIGRATE_TUNNELLED)
3998 3999 4000
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, NULL,
                                            NULL, 0, NULL, NULL,
                                            flags, resource, dconn,
4001
                                            NULL, 0, NULL, migParams);
4002
    else
4003 4004 4005 4006
        ret = qemuMigrationSrcPerformNative(driver, vm, NULL, uri_out,
                                            cookie, cookielen,
                                            NULL, NULL, /* No out cookie with v2 migration */
                                            flags, resource, dconn, NULL, 0, NULL,
4007
                                            migParams);
4008 4009 4010 4011

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4012

4013 4014 4015
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4016
    cancelled = ret < 0;
4017

4018
 finish:
4019 4020 4021 4022
    /* 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.
     */
4023
    dname = dname ? dname : vm->def->name;
4024
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4025
    qemuDomainObjEnterRemote(vm);
4026
    ddomain = dconn->driver->domainMigrateFinish2
4027
        (dconn, dname, cookie, cookielen,
4028
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4029 4030
    /* The domain is already gone at this point */
    ignore_value(qemuDomainObjExitRemote(vm, false));
4031 4032
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4033

4034
 cleanup:
4035
    if (ddomain) {
4036
        virObjectUnref(ddomain);
4037 4038 4039 4040
        ret = 0;
    } else {
        ret = -1;
    }
4041

4042
    virObjectUnref(st);
4043 4044 4045 4046 4047 4048

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4049
    VIR_FREE(cookie);
4050 4051

    return ret;
4052 4053 4054
}


4055 4056 4057 4058
/* 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 */
4059
static int
4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073
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,
4074
                                  qemuMigrationParamsPtr migParams,
4075 4076 4077
                                  unsigned long long bandwidth,
                                  bool useParams,
                                  unsigned long flags)
4078 4079 4080 4081 4082 4083 4084 4085 4086 4087
{
    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;
4088
    bool cancelled = true;
4089
    virStreamPtr st = NULL;
4090
    unsigned long destflags;
4091 4092 4093
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4094
    size_t i;
4095
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4096 4097

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4098
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4099
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4100
              "bandwidth=%llu, useParams=%d, flags=0x%lx",
4101
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4102
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4103
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks, nbdPort,
4104
              bandwidth, useParams, flags);
4105

4106 4107 4108 4109 4110
    /* 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.  */

4111 4112 4113
    dom_xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                         &cookieout, &cookieoutlen,
                                         nmigrate_disks, migrate_disks, flags);
4114 4115 4116
    if (!dom_xml)
        goto cleanup;

4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136
    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;
4137 4138 4139 4140 4141 4142

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4143 4144 4145 4146 4147
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4148 4149 4150 4151 4152
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4153 4154 4155 4156 4157
        if (nbdPort &&
            virTypedParamsAddInt(&params, &nparams, &maxparams,
                                 VIR_MIGRATE_PARAM_DISKS_PORT,
                                 nbdPort) < 0)
            goto cleanup;
4158

4159 4160
        if (qemuMigrationParamsDump(migParams, &params, &nparams,
                                    &maxparams, &flags) < 0)
4161
            goto cleanup;
4162 4163
    }

4164 4165 4166
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4167 4168
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4169

4170
    VIR_DEBUG("Prepare3 %p", dconn);
4171
    VIR_STEAL_PTR(cookiein, cookieout);
4172 4173 4174 4175 4176 4177
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
    if (flags & VIR_MIGRATE_TUNNELLED) {
        if (!(st = virStreamNew(dconn, 0)))
            goto cleanup;

4178
        qemuDomainObjEnterRemote(vm);
4179 4180 4181 4182 4183 4184 4185 4186 4187
        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);
        }
4188 4189
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4190
    } else {
4191
        qemuDomainObjEnterRemote(vm);
4192 4193 4194 4195 4196 4197 4198 4199 4200
        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);
        }
4201 4202
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4203 4204 4205 4206 4207
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

4208
    if (offline) {
L
liguang 已提交
4209 4210 4211
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4212
        cancelled = false;
L
liguang 已提交
4213 4214 4215
        goto finish;
    }

4216 4217 4218 4219
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4220 4221
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4222
            goto finish;
4223
        }
4224
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4225 4226
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4227
        orig_err = virSaveLastError();
4228 4229 4230 4231 4232 4233 4234 4235
        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.
     */
4236
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4237
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4238
    VIR_FREE(cookiein);
4239
    VIR_STEAL_PTR(cookiein, cookieout);
4240 4241
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
4242
    if (flags & VIR_MIGRATE_TUNNELLED) {
4243 4244 4245 4246
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, persist_xml,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4247
                                            nmigrate_disks, migrate_disks,
4248
                                            migParams);
4249
    } else {
4250 4251 4252 4253
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4254
                                            nmigrate_disks, migrate_disks,
4255
                                            migParams);
4256
    }
4257 4258

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4259
    if (ret < 0) {
4260
        orig_err = virSaveLastError();
4261 4262 4263 4264
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4265 4266 4267 4268

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4269
    cancelled = ret < 0;
4270

4271
 finish:
4272 4273 4274 4275 4276 4277 4278 4279
    /*
     * 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);
4280
    VIR_STEAL_PTR(cookiein, cookieout);
4281 4282
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295

    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);
4296 4297
            if (qemuDomainObjExitRemote(vm, !offline) < 0)
                goto cleanup;
4298 4299 4300 4301 4302 4303 4304
        }
    } else {
        dname = dname ? dname : vm->def->name;
        qemuDomainObjEnterRemote(vm);
        ddomain = dconn->driver->domainMigrateFinish3
            (dconn, dname, cookiein, cookieinlen, &cookieout, &cookieoutlen,
             dconnuri, uri, destflags, cancelled);
4305 4306
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4307
    }
4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327

    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();
4328 4329
                if (err &&
                    err->domain == VIR_FROM_QEMU &&
4330 4331 4332 4333 4334 4335 4336
                    err->code != VIR_ERR_MIGRATE_FINISH_OK) {
                    virFreeError(orig_err);
                    orig_err = NULL;
                }
            }
        }
    }
4337

4338 4339 4340 4341 4342 4343 4344
    /* 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.
4345
     */
4346
    cancelled = ddomain == NULL;
4347

4348 4349 4350 4351 4352 4353
    /* 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();

4354 4355 4356 4357
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4358
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4359
    VIR_FREE(cookiein);
4360
    VIR_STEAL_PTR(cookiein, cookieout);
4361 4362
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
4363 4364 4365
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
4366 4367 4368 4369
    /* 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.
     */
4370 4371 4372
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4373 4374 4375

 cleanup:
    if (ddomain) {
4376
        virObjectUnref(ddomain);
4377 4378 4379 4380 4381
        ret = 0;
    } else {
        ret = -1;
    }

4382
    virObjectUnref(st);
4383 4384 4385 4386 4387 4388 4389 4390

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4391
    virTypedParamsFree(params, nparams);
4392 4393 4394 4395
    return ret;
}


4396
static void
4397 4398 4399
qemuMigrationSrcConnectionClosed(virConnectPtr conn,
                                 int reason,
                                 void *opaque)
4400 4401 4402 4403 4404 4405 4406 4407
{
    virDomainObjPtr vm = opaque;

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


4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432
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,
4433
                                 qemuMigrationParamsPtr migParams,
4434 4435 4436 4437
                                 unsigned long flags,
                                 const char *dname,
                                 unsigned long resource,
                                 bool *v3proto)
4438 4439 4440 4441
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4442
    virErrorPtr orig_err = NULL;
4443
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4444
    bool dstOffline = false;
4445
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4446
    bool useParams;
4447

4448 4449
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, uri=%s, "
              "graphicsuri=%s, listenAddress=%s, nmigrate_disks=%zu, "
4450
              "migrate_disks=%p, nbdPort=%d, flags=0x%lx, dname=%s, "
4451
              "resource=%lu",
4452
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4453
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4454 4455
              nmigrate_disks, migrate_disks, nbdPort, flags, NULLSTR(dname),
              resource);
4456

4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470
    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;
    }

4471 4472 4473 4474 4475 4476 4477
    if (flags & VIR_MIGRATE_TUNNELLED && nbdPort) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("disk port address is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

4478 4479 4480 4481
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

4482
    qemuDomainObjEnterRemote(vm);
4483
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4484 4485 4486
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;

4487
    if (dconn == NULL) {
4488
        virReportError(VIR_ERR_OPERATION_FAILED,
4489 4490
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4491
        virObjectUnref(cfg);
4492 4493 4494
        return -1;
    }

4495 4496
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4497 4498
        goto cleanup;

4499
    if (virConnectRegisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed,
4500 4501 4502 4503
                                        vm, NULL) < 0) {
        goto cleanup;
    }

4504
    qemuDomainObjEnterRemote(vm);
4505 4506
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4507
        /* v3proto reflects whether the caller used Perform3, but with
4508
         * p2p migrate, regardless of whether Perform2 or Perform3
4509 4510 4511 4512
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4513 4514
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
4515
    if (offline)
4516 4517
        dstOffline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                              VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4518 4519
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;
4520

4521
    if (!p2p) {
4522 4523
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4524 4525 4526
        goto cleanup;
    }

4527 4528
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4529
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
4530 4531 4532 4533 4534 4535
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("Migration APIs with extensible parameters are not "
                         "supported but extended parameters were passed"));
        goto cleanup;
    }

4536
    if (offline && !dstOffline) {
L
liguang 已提交
4537 4538 4539 4540 4541 4542
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4543 4544 4545 4546 4547 4548 4549
    /* 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;

4550
    if (*v3proto) {
4551 4552 4553
        ret = qemuMigrationSrcPerformPeer2Peer3(driver, sconn, dconn, dconnuri, vm, xmlin,
                                                persist_xml, dname, uri, graphicsuri,
                                                listenAddress, nmigrate_disks, migrate_disks,
4554
                                                nbdPort, migParams, resource,
4555
                                                useParams, flags);
4556
    } else {
4557
        ret = qemuMigrationSrcPerformPeer2Peer2(driver, sconn, dconn, vm,
4558
                                                dconnuri, flags, dname, resource,
4559
                                                migParams);
4560
    }
4561

4562
 cleanup:
4563
    orig_err = virSaveLastError();
4564
    qemuDomainObjEnterRemote(vm);
4565
    virConnectUnregisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed);
4566
    virObjectUnref(dconn);
4567
    ignore_value(qemuDomainObjExitRemote(vm, false));
4568 4569 4570 4571
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4572
    virObjectUnref(cfg);
4573 4574 4575 4576
    return ret;
}


4577 4578 4579 4580 4581 4582
/*
 * 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
4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594
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,
4595
                           qemuMigrationParamsPtr migParams,
4596 4597 4598 4599 4600 4601 4602 4603
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
                           const char *dname,
                           unsigned long resource,
                           bool v3proto)
4604
{
4605
    virObjectEventPtr event = NULL;
4606
    int ret = -1;
4607
    virErrorPtr orig_err = NULL;
4608
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4609
    qemuDomainObjPrivatePtr priv = vm->privateData;
4610

4611 4612
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                              flags) < 0)
4613 4614
        goto cleanup;

4615
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
4616 4617
        goto endjob;

4618
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
4619
        goto endjob;
4620

4621
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
4622
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
4623
        goto endjob;
4624

4625
    qemuMigrationSrcStoreDomainState(vm);
4626 4627

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4628 4629 4630
        ret = qemuMigrationSrcPerformPeer2Peer(driver, conn, vm, xmlin, persist_xml,
                                               dconnuri, uri, graphicsuri, listenAddress,
                                               nmigrate_disks, migrate_disks, nbdPort,
4631
                                               migParams, flags, dname, resource,
4632
                                               &v3proto);
4633
    } else {
4634
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4635 4636 4637
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                            cookieout, cookieoutlen,
                                            flags, resource, NULL, NULL, 0, NULL,
4638
                                            migParams);
4639
    }
4640 4641
    if (ret < 0)
        goto endjob;
4642

4643 4644 4645 4646
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4647
    if (!v3proto) {
4648
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
4649
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4650
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4651
        virDomainAuditStop(vm, "migrated");
4652
        event = virDomainEventLifecycleNewFromObj(vm,
4653 4654
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4655 4656
    }

4657
 endjob:
4658 4659 4660
    if (ret < 0)
        orig_err = virSaveLastError();

4661 4662 4663 4664
    /* v2 proto has no confirm phase so we need to reset migration parameters
     * here
     */
    if (!v3proto && ret < 0)
4665
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4666
                                 priv->job.migParams, priv->job.apiFlags);
4667

4668
    qemuMigrationSrcRestoreDomainState(driver, vm);
4669

4670
    qemuMigrationJobFinish(driver, vm);
4671
    if (!virDomainObjIsActive(vm) && ret == 0) {
4672
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
4673
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4674 4675
            vm->persistent = 0;
        }
4676
        qemuDomainRemoveInactiveJob(driver, vm);
4677 4678
    }

4679 4680 4681 4682 4683
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4684
 cleanup:
M
Michal Privoznik 已提交
4685
    virDomainObjEndAPI(&vm);
4686
    virObjectEventStateQueue(driver->domainEventState, event);
4687
    virObjectUnref(cfg);
4688 4689 4690 4691 4692 4693 4694
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4695 4696 4697 4698 4699 4700 4701 4702
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,
4703
                             qemuMigrationParamsPtr migParams,
4704 4705 4706 4707 4708 4709
                             const char *cookiein,
                             int cookieinlen,
                             char **cookieout,
                             int *cookieoutlen,
                             unsigned long flags,
                             unsigned long resource)
4710
{
4711
    qemuDomainObjPrivatePtr priv = vm->privateData;
4712 4713 4714 4715
    int ret = -1;

    /* If we didn't start the job in the begin phase, start it now. */
    if (!(flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
4716 4717
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
4718 4719 4720 4721 4722 4723
            goto cleanup;
    } else if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) {
        goto cleanup;
    }

    qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4724
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
4725
                           qemuMigrationSrcCleanup);
4726

4727 4728 4729
    ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                        cookieout, cookieoutlen,
                                        flags, resource, NULL, graphicsuri,
4730
                                        nmigrate_disks, migrate_disks, migParams);
4731

4732
    if (ret < 0) {
4733
        qemuMigrationSrcRestoreDomainState(driver, vm);
4734
        goto endjob;
4735
    }
4736 4737 4738

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4739
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
4740
                             qemuMigrationSrcCleanup) < 0)
4741 4742
        goto endjob;

4743
 endjob:
4744
    if (ret < 0) {
4745
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4746
                                 priv->job.migParams, priv->job.apiFlags);
4747
        qemuMigrationJobFinish(driver, vm);
4748
    } else {
4749
        qemuMigrationJobContinue(vm);
4750 4751
    }

4752
    if (!virDomainObjIsActive(vm))
4753
        qemuDomainRemoveInactiveJob(driver, vm);
4754

4755
 cleanup:
M
Michal Privoznik 已提交
4756
    virDomainObjEndAPI(&vm);
4757 4758 4759
    return ret;
}

4760
int
4761 4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772
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,
4773
                        qemuMigrationParamsPtr migParams,
4774 4775 4776 4777 4778 4779 4780 4781
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4782 4783
{
    VIR_DEBUG("driver=%p, conn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4784
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
4785
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4786
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
4787
              "flags=0x%lx, dname=%s, resource=%lu, v3proto=%d",
4788
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4789
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4790 4791 4792
              nmigrate_disks, migrate_disks, nbdPort,
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4793 4794 4795

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4796 4797
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4798 4799 4800
            return -1;
        }

4801 4802 4803
        return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, dconnuri, uri,
                                          graphicsuri, listenAddress,
                                          nmigrate_disks, migrate_disks, nbdPort,
4804
                                          migParams,
4805 4806 4807
                                          cookiein, cookieinlen,
                                          cookieout, cookieoutlen,
                                          flags, dname, resource, v3proto);
4808 4809
    } else {
        if (dconnuri) {
4810 4811
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
4812 4813 4814 4815
            return -1;
        }

        if (v3proto) {
4816 4817 4818
            return qemuMigrationSrcPerformPhase(driver, conn, vm, persist_xml, uri,
                                                graphicsuri,
                                                nmigrate_disks, migrate_disks,
4819
                                                migParams,
4820 4821 4822
                                                cookiein, cookieinlen,
                                                cookieout, cookieoutlen,
                                                flags, resource);
4823
        } else {
4824 4825 4826
            return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, NULL,
                                              uri, graphicsuri, listenAddress,
                                              nmigrate_disks, migrate_disks, nbdPort,
4827
                                              migParams,
4828 4829 4830
                                              cookiein, cookieinlen,
                                              cookieout, cookieoutlen, flags,
                                              dname, resource, v3proto);
4831 4832 4833
        }
    }
}
4834

4835
static int
4836
qemuMigrationDstVPAssociatePortProfiles(virDomainDefPtr def)
4837
{
4838
    size_t i;
4839 4840 4841 4842 4843
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
4844
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4845
            if (virNetDevVPortProfileAssociate(net->ifname,
4846
                                               virDomainNetGetActualVirtPortProfile(net),
4847
                                               &net->mac,
4848
                                               virDomainNetGetActualDirectDev(net),
4849
                                               -1,
4850
                                               def->uuid,
4851 4852
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
4853 4854 4855
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
4856
                goto err_exit;
4857
            }
4858
            last_good_net = i;
4859
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
4860

4861
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
4862 4863 4864 4865
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
4866 4867 4868
        }
    }

4869
    return 0;
4870

4871
 err_exit:
4872
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
4873
        net = def->nets[i];
4874
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4875
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
4876
                                                           virDomainNetGetActualVirtPortProfile(net),
4877
                                                           &net->mac,
4878
                                                           virDomainNetGetActualDirectDev(net),
4879
                                                           -1,
4880
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
4881 4882
        }
    }
4883
    return -1;
4884 4885 4886
}


J
Jiri Denemark 已提交
4887
static int
4888 4889 4890 4891
qemuMigrationDstPersist(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationCookiePtr mig,
                        bool ignoreSaveError)
J
Jiri Denemark 已提交
4892 4893 4894 4895
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    virDomainDefPtr vmdef;
4896 4897
    virDomainDefPtr oldDef = NULL;
    unsigned int oldPersist = vm->persistent;
J
Jiri Denemark 已提交
4898 4899 4900 4901 4902 4903 4904
    virObjectEventPtr event;
    int ret = -1;

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

    vm->persistent = 1;
4905 4906
    oldDef = vm->newDef;
    vm->newDef = qemuMigrationCookieGetPersistent(mig);
J
Jiri Denemark 已提交
4907

4908 4909
    if (!(vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm)))
        goto error;
J
Jiri Denemark 已提交
4910

4911 4912
    if (virDomainSaveConfig(cfg->configDir, driver->caps, vmdef) < 0 &&
        !ignoreSaveError)
4913
        goto error;
J
Jiri Denemark 已提交
4914 4915 4916

    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_DEFINED,
4917 4918 4919
                                              oldPersist ?
                                              VIR_DOMAIN_EVENT_DEFINED_UPDATED :
                                              VIR_DOMAIN_EVENT_DEFINED_ADDED);
4920
    virObjectEventStateQueue(driver->domainEventState, event);
J
Jiri Denemark 已提交
4921 4922 4923 4924

    ret = 0;

 cleanup:
4925
    virDomainDefFree(oldDef);
J
Jiri Denemark 已提交
4926 4927 4928
    virObjectUnref(caps);
    virObjectUnref(cfg);
    return ret;
4929 4930 4931 4932 4933 4934 4935

 error:
    virDomainDefFree(vm->newDef);
    vm->persistent = oldPersist;
    vm->newDef = oldDef;
    oldDef = NULL;
    goto cleanup;
J
Jiri Denemark 已提交
4936 4937 4938
}


4939
virDomainPtr
4940 4941 4942 4943 4944 4945 4946 4947 4948 4949
qemuMigrationDstFinish(virQEMUDriverPtr driver,
                       virConnectPtr dconn,
                       virDomainObjPtr vm,
                       const char *cookiein,
                       int cookieinlen,
                       char **cookieout,
                       int *cookieoutlen,
                       unsigned long flags,
                       int retcode,
                       bool v3proto)
4950 4951
{
    virDomainPtr dom = NULL;
4952
    qemuMigrationCookiePtr mig = NULL;
4953
    virErrorPtr orig_err = NULL;
4954
    int cookie_flags = 0;
J
Jiri Denemark 已提交
4955
    qemuDomainObjPrivatePtr priv = vm->privateData;
4956
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4957
    unsigned short port;
4958 4959
    unsigned long long timeReceived = 0;
    virObjectEventPtr event;
4960
    qemuDomainJobInfoPtr jobInfo = NULL;
4961
    bool inPostCopy = false;
4962
    bool doKill = true;
4963

4964
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
4965
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, retcode=%d",
4966 4967
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
4968

4969 4970 4971
    port = priv->migrationPort;
    priv->migrationPort = 0;

4972
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN)) {
4973
        qemuMigrationDstErrorReport(driver, vm->def->name);
4974
        goto cleanup;
4975
    }
4976

4977 4978
    ignore_value(virTimeMillisNow(&timeReceived));

4979 4980 4981
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
4982

4983
    qemuDomainCleanupRemove(vm, qemuMigrationDstPrepareCleanup);
4984
    VIR_FREE(priv->job.completed);
4985

4986
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
4987 4988
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
4989 4990 4991
    /* 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;
4992

4993 4994
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen, cookie_flags)))
4995
        goto endjob;
4996

4997
    if (flags & VIR_MIGRATE_OFFLINE) {
4998
        if (retcode == 0 &&
4999
            qemuMigrationDstPersist(driver, vm, mig, false) == 0)
5000
            dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, -1);
5001 5002
        goto endjob;
    }
5003

5004 5005 5006 5007
    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
         */
5008
        qemuDomainCheckMonitor(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN);
5009 5010
        goto endjob;
    }
5011

5012 5013 5014
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
5015
        qemuMigrationDstErrorReport(driver, vm->def->name);
5016 5017
        goto endjob;
    }
5018

5019
    if (qemuMigrationDstVPAssociatePortProfiles(vm->def) < 0)
5020
        goto endjob;
5021

5022
    if (mig->network && qemuMigrationDstOPDRelocate(driver, vm, mig) < 0)
5023
        VIR_WARN("unable to provide network data for relocation");
5024

5025
    if (qemuMigrationDstStopNBDServer(driver, vm, mig) < 0)
5026
        goto endjob;
5027

5028 5029
    if (qemuRefreshVirtioChannelState(driver, vm,
                                      QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
5030 5031
        goto endjob;

5032 5033
    if (qemuConnectAgent(driver, vm) < 0)
        goto endjob;
5034

5035
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
5036
        if (qemuMigrationDstPersist(driver, vm, mig, !v3proto) < 0) {
5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047
            /* 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.
5048 5049 5050
             */
            if (v3proto)
                goto endjob;
5051
        }
5052
    }
5053

5054 5055 5056
    /* We need to wait for QEMU to process all data sent by the source
     * before starting guest CPUs.
     */
5057 5058 5059
    if (qemuMigrationDstWaitForCompletion(driver, vm,
                                          QEMU_ASYNC_JOB_MIGRATION_IN,
                                          !!(flags & VIR_MIGRATE_POSTCOPY)) < 0) {
5060 5061 5062 5063 5064 5065
        /* 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;
    }
5066

5067 5068 5069 5070 5071 5072 5073 5074 5075
    /* 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;
    }

5076
    if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY)
5077 5078
        inPostCopy = true;

5079 5080 5081 5082 5083
    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
         */
5084
        if (qemuProcessStartCPUs(driver, vm,
5085 5086
                                 inPostCopy ? VIR_DOMAIN_RUNNING_POSTCOPY
                                            : VIR_DOMAIN_RUNNING_MIGRATED,
5087
                                 QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5088
            if (virGetLastErrorCode() == VIR_ERR_OK)
5089 5090 5091 5092 5093 5094
                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();
5095

5096 5097 5098 5099 5100 5101 5102 5103 5104 5105
            /*
             * 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;
5106
        }
5107

5108
        if (inPostCopy)
5109
            doKill = false;
5110
    }
5111

5112
    if (mig->jobInfo) {
5113
        jobInfo = mig->jobInfo;
5114 5115 5116 5117 5118 5119
        mig->jobInfo = NULL;

        if (jobInfo->sent && timeReceived) {
            jobInfo->timeDelta = timeReceived - jobInfo->sent;
            jobInfo->received = timeReceived;
            jobInfo->timeDeltaSet = true;
5120
        }
5121 5122
        qemuDomainJobInfoUpdateTime(jobInfo);
        qemuDomainJobInfoUpdateDowntime(jobInfo);
5123
    }
L
liguang 已提交
5124

5125
    if (inPostCopy) {
5126 5127 5128
        if (qemuMigrationDstWaitForCompletion(driver, vm,
                                              QEMU_ASYNC_JOB_MIGRATION_IN,
                                              false) < 0) {
5129 5130 5131 5132 5133 5134 5135 5136 5137
            goto endjob;
        }
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            virDomainObjSetState(vm,
                                 VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_MIGRATED);
        }
    }

5138
    dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, vm->def->id);
5139

5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153
    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);
    }
5154

5155 5156 5157 5158 5159
    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);
5160
        virObjectEventStateQueue(driver->domainEventState, event);
5161
    }
5162

5163
    if (virDomainObjIsActive(vm) &&
5164
        virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
5165 5166 5167 5168 5169
        VIR_WARN("Failed to save status on vm %s", vm->def->name);

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

5170
 endjob:
5171
    if (!dom &&
5172 5173
        !(flags & VIR_MIGRATE_OFFLINE) &&
        virDomainObjIsActive(vm)) {
5174
        if (doKill) {
5175 5176 5177 5178 5179 5180 5181
            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);
5182
            virObjectEventStateQueue(driver->domainEventState, event);
5183
        } else {
5184
            qemuMigrationAnyPostcopyFailed(driver, vm);
5185
        }
5186 5187
    }

5188
    if (dom) {
5189 5190 5191
        if (jobInfo) {
            VIR_STEAL_PTR(priv->job.completed, jobInfo);
            priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
5192
            priv->job.completed->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;
5193
        }
5194

5195 5196 5197
        if (qemuMigrationBakeCookie(mig, driver, vm,
                                    QEMU_MIGRATION_DESTINATION,
                                    cookieout, cookieoutlen,
5198 5199
                                    QEMU_MIGRATION_COOKIE_STATS) < 0)
            VIR_WARN("Unable to encode migration cookie");
5200 5201 5202 5203 5204 5205

        /* Remove completed stats for post-copy, everything but timing fields
         * is obsolete anyway.
         */
        if (inPostCopy)
            VIR_FREE(priv->job.completed);
5206
    }
5207

5208
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
5209
                             priv->job.migParams, priv->job.apiFlags);
5210

5211
    qemuMigrationJobFinish(driver, vm);
5212
    if (!virDomainObjIsActive(vm))
5213
        qemuDomainRemoveInactiveJob(driver, vm);
5214

5215
 cleanup:
5216
    VIR_FREE(jobInfo);
5217
    virPortAllocatorRelease(port);
5218
    if (priv->mon)
5219
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
5220
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5221
    virDomainObjEndAPI(&vm);
5222
    qemuMigrationCookieFree(mig);
5223 5224 5225 5226
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5227
    virObjectUnref(cfg);
5228 5229 5230 5231

    /* Set a special error if Finish is expected to return NULL as a result of
     * successful call with retcode != 0
     */
5232
    if (retcode != 0 && !dom && virGetLastErrorCode() == VIR_ERR_OK)
5233
        virReportError(VIR_ERR_MIGRATE_FINISH_OK, NULL);
5234 5235
    return dom;
}
5236

5237

5238
/* Helper function called while vm is active.  */
5239
int
5240 5241 5242 5243
qemuMigrationSrcToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
                       int fd,
                       const char *compressor,
                       qemuDomainAsyncJob asyncJob)
5244 5245 5246
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5247
    int ret = -1;
5248 5249
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5250
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5251
    char *errbuf = NULL;
5252
    virErrorPtr orig_err = NULL;
5253 5254 5255 5256 5257

    /* 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,
5258 5259
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5260 5261
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5262
    }
5263

5264 5265 5266 5267 5268 5269 5270
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5271 5272 5273 5274
    if (compressor && pipe(pipeFD) < 0) {
        virReportSystemError(errno, "%s",
                             _("Failed to create pipe for migration"));
        return -1;
5275 5276
    }

5277 5278 5279 5280
    /* 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.  */
5281 5282
    if (qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    compressor ? pipeFD[1] : fd) < 0)
5283 5284
        goto cleanup;

5285
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5286 5287
        goto cleanup;

5288
    if (!compressor) {
5289 5290 5291
        rc = qemuMonitorMigrateToFd(priv->mon,
                                    QEMU_MONITOR_MIGRATE_BACKGROUND,
                                    fd);
5292 5293 5294 5295 5296 5297 5298
    } else {
        const char *prog = compressor;
        const char *args[] = {
            prog,
            "-c",
            NULL
        };
5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309

        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;
5310
        }
5311 5312 5313 5314 5315 5316 5317 5318 5319 5320
        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");
5321
    }
5322
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5323
        goto cleanup;
5324 5325 5326
    if (rc < 0)
        goto cleanup;

5327
    rc = qemuMigrationSrcWaitForCompletion(driver, vm, asyncJob, NULL, 0);
5328

5329 5330 5331 5332
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5333 5334
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5335
                qemuMonitorMigrateCancel(priv->mon);
5336
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5337 5338
            }
        }
5339
        goto cleanup;
5340
    }
5341

5342 5343 5344
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5345
    qemuDomainEventEmitJobCompleted(driver, vm);
5346 5347
    ret = 0;

5348
 cleanup:
5349 5350 5351
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5352
    /* Restore max migration bandwidth */
5353 5354
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5355 5356
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5357
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5358 5359
    }

5360 5361
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5362 5363 5364 5365 5366
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5367 5368 5369 5370 5371 5372

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

5373 5374
    return ret;
}
5375

5376 5377

int
5378 5379
qemuMigrationSrcCancel(virQEMUDriverPtr driver,
                       virDomainObjPtr vm)
5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390
{
    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));
5391
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5392 5393 5394 5395 5396
        goto cleanup;

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
5397
        qemuBlockJobDataPtr job;
5398

5399
        if (!(job = qemuBlockJobDiskGetJob(disk)) ||
5400
            !qemuBlockJobIsRunning(job))
5401
            diskPriv->migrating = false;
5402 5403

        if (diskPriv->migrating) {
5404
            qemuBlockJobSyncBegin(job);
5405
            storage = true;
5406 5407
        }

5408 5409
        virObjectUnref(job);
    }
5410 5411 5412

    if (storage &&
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
5413
                                      QEMU_ASYNC_JOB_NONE, NULL) < 0)
5414
        goto cleanup;
5415 5416 5417 5418 5419 5420 5421 5422

    ret = 0;

 cleanup:
    return ret;
}


5423
static int
5424
qemuMigrationJobStart(virQEMUDriverPtr driver,
5425
                      virDomainObjPtr vm,
5426 5427
                      qemuDomainAsyncJob job,
                      unsigned long apiFlags)
5428
{
5429
    qemuDomainObjPrivatePtr priv = vm->privateData;
5430 5431
    virDomainJobOperation op;
    unsigned long long mask;
5432

5433
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5434 5435
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_IN;
        mask = QEMU_JOB_NONE;
5436
    } else {
5437 5438 5439 5440
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_OUT;
        mask = QEMU_JOB_DEFAULT_MASK |
               JOB_MASK(QEMU_JOB_SUSPEND) |
               JOB_MASK(QEMU_JOB_MIGRATION_OP);
5441
    }
5442

5443
    if (qemuDomainObjBeginAsyncJob(driver, vm, job, op, apiFlags) < 0)
5444 5445
        return -1;

5446 5447
    priv->job.current->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;

5448
    qemuDomainObjSetAsyncJobMask(vm, mask);
5449 5450 5451
    return 0;
}

5452
static void
5453
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5454
                         virDomainObjPtr vm,
5455
                         qemuMigrationJobPhase phase)
5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468
{
    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);
}

5469
static void
5470
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5471
                           virDomainObjPtr vm,
5472
                           qemuMigrationJobPhase phase)
5473 5474 5475 5476
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5477
static void
5478 5479
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5480
    qemuDomainObjReleaseAsyncJob(vm);
5481 5482
}

5483
static bool
5484
qemuMigrationJobIsActive(virDomainObjPtr vm,
5485
                         qemuDomainAsyncJob job)
5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496
{
    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");

5497
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5498 5499 5500 5501 5502
        return false;
    }
    return true;
}

5503
static void
5504
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5505
{
5506
    qemuDomainObjEndAsyncJob(driver, vm);
5507
}
5508 5509 5510


static void
5511
qemuMigrationDstErrorFree(void *data,
5512 5513 5514 5515 5516 5517 5518
                       const void *name ATTRIBUTE_UNUSED)
{
    virErrorPtr err = data;
    virFreeError(err);
}

int
5519
qemuMigrationDstErrorInit(virQEMUDriverPtr driver)
5520
{
5521
    driver->migrationErrors = virHashAtomicNew(64, qemuMigrationDstErrorFree);
5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532
    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
5533 5534 5535
qemuMigrationDstErrorSave(virQEMUDriverPtr driver,
                          const char *name,
                          virErrorPtr err)
5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548
{
    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
5549 5550
qemuMigrationDstErrorReport(virQEMUDriverPtr driver,
                            const char *name)
5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561
{
    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);
}
5562 5563


5564
int
5565 5566 5567 5568
qemuMigrationSrcFetchMirrorStats(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 qemuDomainJobInfoPtr jobInfo)
5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612
{
    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;
}