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
}


778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856
static int
qemuMigrationSrcNBDStorageCopyBlockdev(virQEMUDriverPtr driver,
                                       virDomainObjPtr vm,
                                       virDomainDiskDefPtr disk,
                                       const char *diskAlias,
                                       const char *host,
                                       int port,
                                       unsigned long long mirror_speed,
                                       unsigned int mirror_flags,
                                       const char *tlsAlias)
{
    qemuBlockStorageSourceAttachDataPtr data = NULL;
    qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
    virStorageSourcePtr copysrc = NULL;
    int mon_ret = 0;
    int ret = -1;

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    VIR_STEAL_PTR(diskPriv->migrSource, copysrc);

    ret = 0;

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


857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898
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;
}


899 900 901 902 903 904 905 906 907 908 909 910
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);
911
    qemuBlockJobDataPtr job = NULL;
912 913 914 915 916 917 918
    char *diskAlias = NULL;
    int rc;
    int ret = -1;

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

919
    if (!(job = qemuBlockJobDiskNew(disk, QEMU_BLOCKJOB_TYPE_COPY, diskAlias)))
920 921
        goto cleanup;

922
    qemuBlockJobSyncBegin(job);
923 924 925 926 927 928 929 930 931 932 933 934 935 936 937

    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);
    }

938
    if (rc < 0)
939 940 941
        goto cleanup;

    diskPriv->migrating = true;
942
    qemuBlockJobStarted(job);
943 944 945 946

    ret = 0;

 cleanup:
947
    qemuBlockJobStartupFinalize(job);
948 949 950 951 952
    VIR_FREE(diskAlias);
    return ret;
}


953
/**
954
 * qemuMigrationSrcNBDStorageCopy:
955 956 957 958
 * @driver: qemu driver
 * @vm: domain
 * @mig: migration cookie
 * @host: where are we migrating to
959
 * @speed: bandwidth limit in MiB/s
960 961
 * @migrate_flags: migrate monitor command flags
 *
962 963 964
 * 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
965
 * to do the very same operation. On failure, the caller is
966 967
 * expected to call qemuMigrationSrcNBDCopyCancel to stop all
 * running copy operations.
968 969 970 971 972
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
973 974 975 976 977 978 979 980
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,
981 982 983
                               virConnectPtr dconn,
                               const char *tlsAlias,
                               unsigned int flags)
984 985 986 987
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int port;
988
    size_t i;
989
    unsigned long long mirror_speed = speed;
990
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
991
    int rv;
992
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
993 994

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
995

996 997 998 999 1000 1001 1002 1003
    if (mirror_speed > LLONG_MAX >> 20) {
        virReportError(VIR_ERR_OVERFLOW,
                       _("bandwidth must be less than %llu"),
                       LLONG_MAX >> 20);
        goto cleanup;
    }
    mirror_speed <<= 20;

1004 1005 1006 1007 1008 1009 1010 1011 1012 1013
    /* 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];

1014
        /* check whether disk should be migrated */
1015
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1016 1017
            continue;

1018 1019 1020
        if (qemuMigrationSrcNBDStorageCopyOne(driver, vm, disk, host, port,
                                              mirror_speed, mirror_flags,
                                              tlsAlias, flags) < 0)
1021 1022
            goto cleanup;

1023
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1024 1025 1026
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto cleanup;
        }
1027
    }
1028

1029
    while ((rv = qemuMigrationSrcNBDStorageCopyReady(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) != 1) {
1030 1031
        if (rv < 0)
            goto cleanup;
1032

1033
        if (priv->job.abortJob) {
1034
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1035 1036 1037 1038
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
            goto cleanup;
1039
        }
1040

1041 1042 1043 1044 1045 1046
        if (dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            goto cleanup;
        }

1047
        if (virDomainObjWait(vm) < 0)
1048
            goto cleanup;
1049 1050
    }

1051 1052
    qemuMigrationSrcFetchMirrorStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                     priv->job.current);
1053

1054
    /* Okay, all disks are ready. Modify migrate_flags */
1055 1056 1057 1058
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

1059
 cleanup:
1060
    virObjectUnref(cfg);
1061 1062
    return ret;
}
1063

1064

1065
/**
1066
 * qemuMigrationSrcIsAllowedHostdev:
1067 1068
 * @def: domain definition
 *
1069
 * Checks that @def does not contain any host devices unsupported across
1070 1071 1072
 * migrations. Returns true if the vm is allowed to migrate.
 */
static bool
1073
qemuMigrationSrcIsAllowedHostdev(const virDomainDef *def)
1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
{
    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;
}


1093
/**
1094
 * qemuMigrationSrcIsAllowed:
1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
 * @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.
 */
1107
bool
1108 1109 1110 1111
qemuMigrationSrcIsAllowed(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          bool remote,
                          unsigned int flags)
1112
{
1113
    int nsnapshots;
1114
    int pauseReason;
1115
    size_t i;
1116

1117 1118 1119 1120 1121
    /* perform these checks only when migrating to remote hosts */
    if (remote) {
        nsnapshots = virDomainSnapshotObjListNum(vm->snapshots, NULL, 0);
        if (nsnapshots < 0)
            return false;
1122

1123 1124 1125 1126 1127
        if (nsnapshots > 0) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("cannot migrate domain with %d snapshots"),
                           nsnapshots);
            return false;
1128
        }
1129

1130
        /* cancel migration if disk I/O error is emitted while migrating */
1131
        if (flags & VIR_MIGRATE_ABORT_ON_ERROR &&
1132
            !(flags & VIR_MIGRATE_OFFLINE) &&
1133 1134
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
E
Eric Blake 已提交
1135
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1136
                           _("cannot migrate domain with I/O error"));
E
Eric Blake 已提交
1137 1138
            return false;
        }
1139 1140

    }
1141

1142 1143 1144 1145 1146 1147 1148
    /* 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;
        }
1149

1150

1151 1152
        if (qemuDomainHasBlockjob(vm, false)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1153
                           _("domain has active block job"));
1154 1155 1156
            return false;
        }

1157
        if (!qemuMigrationSrcIsAllowedHostdev(vm->def))
1158 1159
            return false;

1160
        if (vm->def->cpu) {
1161 1162 1163
            /* QEMU blocks migration and save with invariant TSC enabled
             * unless TSC frequency is explicitly set.
             */
1164 1165
            if (virCPUCheckFeature(vm->def->os.arch, vm->def->cpu,
                                   "invtsc") == 1) {
1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183
                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;
                }
1184
            }
J
Ján Tomko 已提交
1185 1186
        }

1187 1188 1189
        /* Verify that memory device config can be transferred reliably */
        for (i = 0; i < vm->def->nmems; i++) {
            virDomainMemoryDefPtr mem = vm->def->mems[i];
1190

1191 1192 1193 1194 1195
            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"));
1196

1197 1198
                return false;
            }
1199
        }
1200 1201 1202 1203 1204 1205

        if (vm->def->nshmems) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("migration with shmem device is not supported"));
            return false;
        }
1206 1207
    }

1208 1209 1210
    return true;
}

1211
static bool
1212 1213 1214 1215
qemuMigrationSrcIsSafe(virDomainDefPtr def,
                       size_t nmigrate_disks,
                       const char **migrate_disks,
                       unsigned int flags)
1216

1217
{
1218 1219
    bool storagemigration = flags & (VIR_MIGRATE_NON_SHARED_DISK |
                                     VIR_MIGRATE_NON_SHARED_INC);
1220
    size_t i;
1221
    int rc;
1222

1223
    for (i = 0; i < def->ndisks; i++) {
1224
        virDomainDiskDefPtr disk = def->disks[i];
1225
        const char *src = virDomainDiskGetSource(disk);
1226

1227 1228
        /* Disks without any source (i.e. floppies and CD-ROMs)
         * OR readonly are safe. */
1229
        if (virStorageSourceIsEmpty(disk->src) ||
1230
            disk->src->readonly)
1231
            continue;
1232

1233
        /* Disks which are migrated by qemu are safe too. */
1234
        if (storagemigration &&
1235
            qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1236
            continue;
1237

1238
        /* However, disks on local FS (e.g. ext4) are not safe. */
1239
        if (virStorageSourceGetActualType(disk->src) == VIR_STORAGE_TYPE_FILE) {
1240
            if ((rc = virFileIsSharedFS(src)) < 0) {
1241
                return false;
1242 1243 1244 1245 1246
            } else if (rc == 0) {
                virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                               _("Migration without shared storage is unsafe"));
                return false;
            }
1247 1248 1249 1250
            if ((rc = virStorageFileIsClusterFS(src)) < 0)
                return false;
            else if (rc == 1)
                continue;
1251 1252
        } else if (virStorageSourceGetActualType(disk->src) == VIR_STORAGE_TYPE_NETWORK) {
            /* But network disks are safe again. */
1253
            continue;
1254
        }
1255

1256 1257 1258 1259 1260 1261 1262
        /* 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;

1263 1264
        virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                       _("Migration may lead to data corruption if disks"
1265
                         " use cache != none or cache != directsync"));
1266
        return false;
1267 1268 1269 1270 1271
    }

    return true;
}

1272
/** qemuMigrationSrcSetOffline
1273 1274 1275
 * Pause domain for non-live migration.
 */
int
1276 1277
qemuMigrationSrcSetOffline(virQEMUDriverPtr driver,
                           virDomainObjPtr vm)
1278 1279
{
    int ret;
1280
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
1281 1282
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
1283
    if (ret == 0) {
1284
        virObjectEventPtr event;
1285

1286
        event = virDomainEventLifecycleNewFromObj(vm,
1287 1288
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
1289
        virObjectEventStateQueue(driver->domainEventState, event);
1290 1291 1292 1293 1294
    }

    return ret;
}

1295 1296

void
1297 1298
qemuMigrationAnyPostcopyFailed(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
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 1328
{
    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);
1329
        virObjectEventStateQueue(driver->domainEventState, event);
1330 1331 1332 1333 1334 1335 1336
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
    }
}


1337
static int
1338
qemuMigrationSrcWaitForSpice(virDomainObjPtr vm)
1339 1340 1341
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

1342
    if (!priv->job.spiceMigration)
1343 1344
        return 0;

1345
    VIR_DEBUG("Waiting for SPICE to finish migration");
1346 1347
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
1348 1349 1350 1351
            return -1;
    }
    return 0;
}
1352

1353 1354 1355 1356

static void
qemuMigrationUpdateJobType(qemuDomainJobInfoPtr jobInfo)
{
1357
    switch ((qemuMonitorMigrationStatus) jobInfo->stats.mig.status) {
1358 1359 1360 1361
    case QEMU_MONITOR_MIGRATION_STATUS_POSTCOPY:
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_POSTCOPY;
        break;

1362
    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
1363
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED;
1364 1365 1366
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
1367
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_NONE;
1368 1369 1370
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
1371
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1372 1373 1374
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
1375
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1376 1377
        break;

1378 1379 1380 1381 1382 1383 1384 1385
    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;

1386 1387 1388
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLING:
1389
    case QEMU_MONITOR_MIGRATION_STATUS_LAST:
1390 1391 1392 1393 1394 1395
        break;
    }
}


int
1396 1397 1398 1399 1400
qemuMigrationAnyFetchStats(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuDomainAsyncJob asyncJob,
                           qemuDomainJobInfoPtr jobInfo,
                           char **error)
1401 1402
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1403
    qemuMonitorMigrationStats stats;
1404 1405 1406 1407 1408
    int rv;

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

1409
    rv = qemuMonitorGetMigrationStats(priv->mon, &stats, error);
1410 1411 1412 1413

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

1414
    jobInfo->stats.mig = stats;
1415 1416

    return 0;
1417 1418 1419
}


1420 1421 1422 1423 1424 1425 1426
static const char *
qemuMigrationJobName(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
1427
        return _("migration out job");
1428 1429 1430 1431
    case QEMU_ASYNC_JOB_SAVE:
        return _("domain save job");
    case QEMU_ASYNC_JOB_DUMP:
        return _("domain core dump job");
1432 1433 1434 1435 1436 1437 1438 1439 1440
    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:
1441 1442 1443 1444 1445 1446
    default:
        return _("job");
    }
}


1447
static int
1448
qemuMigrationJobCheckStatus(virQEMUDriverPtr driver,
1449
                            virDomainObjPtr vm,
1450
                            qemuDomainAsyncJob asyncJob)
1451 1452 1453
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1454
    char *error = NULL;
1455
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1456
    int ret = -1;
1457

1458
    if (!events ||
1459
        jobInfo->stats.mig.status == QEMU_MONITOR_MIGRATION_STATUS_ERROR) {
1460
        if (qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, &error) < 0)
1461 1462
            return -1;
    }
1463

1464 1465
    qemuMigrationUpdateJobType(jobInfo);

1466 1467
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_NONE:
1468 1469
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("is not active"));
1470
        goto cleanup;
1471

1472
    case QEMU_DOMAIN_JOB_STATUS_FAILED:
1473
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
1474 1475 1476
                       qemuMigrationJobName(vm),
                       error ? error : _("unexpectedly failed"));
        goto cleanup;
1477

1478
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
1479 1480
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("canceled by client"));
1481
        goto cleanup;
1482

1483 1484
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
1485
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
1486
    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
1487
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1488
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1489 1490
        break;
    }
1491 1492 1493 1494 1495 1496

    ret = 0;

 cleanup:
    VIR_FREE(error);
    return ret;
1497 1498 1499
}


1500 1501
enum qemuMigrationCompletedFlags {
    QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR = (1 << 0),
1502
    /* This flag should only be set when run on src host */
1503
    QEMU_MIGRATION_COMPLETED_CHECK_STORAGE  = (1 << 1),
1504
    QEMU_MIGRATION_COMPLETED_POSTCOPY       = (1 << 2),
1505
    QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER = (1 << 3),
1506 1507
};

1508

1509 1510 1511 1512 1513 1514 1515
/**
 * 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
1516 1517 1518 1519 1520
qemuMigrationAnyCompleted(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          qemuDomainAsyncJob asyncJob,
                          virConnectPtr dconn,
                          unsigned int flags)
1521 1522 1523 1524 1525
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    int pauseReason;

1526
    if (qemuMigrationJobCheckStatus(driver, vm, asyncJob) < 0)
1527 1528
        goto error;

1529
    /* This flag should only be set when run on src host */
1530
    if (flags & QEMU_MIGRATION_COMPLETED_CHECK_STORAGE &&
1531
        qemuMigrationSrcNBDStorageCopyReady(vm, asyncJob) < 0)
1532 1533
        goto error;

1534
    if (flags & QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR &&
1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547
        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;
    }

1548 1549 1550 1551 1552 1553 1554 1555 1556 1557
    /* 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;
    }

1558 1559 1560 1561 1562
    /* 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 &&
1563
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
1564 1565 1566 1567
        VIR_DEBUG("Migration switched to post-copy");
        return 1;
    }

1568
    if (jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
1569 1570 1571 1572 1573
        return 1;
    else
        return 0;

 error:
1574 1575 1576
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1577
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1578
        /* The migration was aborted by us rather than QEMU itself. */
1579
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1580
        return -2;
1581 1582 1583

    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
        /* Something failed after QEMU already finished the migration. */
1584
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1585
        return -1;
1586 1587 1588 1589

    case QEMU_DOMAIN_JOB_STATUS_FAILED:
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
        /* QEMU aborted the migration. */
1590
        return -1;
1591 1592 1593 1594 1595 1596

    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_NONE:
        /* Impossible. */
        break;
1597
    }
1598 1599

    return -1;
1600 1601 1602
}


1603 1604 1605
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
1606
static int
1607 1608 1609 1610 1611
qemuMigrationSrcWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  virConnectPtr dconn,
                                  unsigned int flags)
1612
{
1613
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
1614
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1615
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1616
    int rv;
1617

1618 1619
    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_MIGRATING;

1620 1621
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           dconn, flags)) != 1) {
1622 1623
        if (rv < 0)
            return rv;
1624

1625 1626
        if (events) {
            if (virDomainObjWait(vm) < 0) {
1627 1628
                if (virDomainObjIsActive(vm))
                    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1629 1630 1631 1632 1633 1634 1635 1636 1637 1638
                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);
        }
1639 1640
    }

1641
    if (events)
1642
        ignore_value(qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, NULL));
1643

1644
    qemuDomainJobInfoUpdateTime(jobInfo);
1645 1646
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
1647
    if (VIR_ALLOC(priv->job.completed) == 0) {
1648
        *priv->job.completed = *jobInfo;
1649 1650
        priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
    }
1651

1652 1653 1654 1655
    if (asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT &&
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;

1656
    return 0;
1657 1658 1659
}


1660
static int
1661 1662 1663 1664
qemuMigrationDstWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  bool postcopy)
1665 1666
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1667
    unsigned int flags = 0;
1668 1669 1670 1671 1672 1673 1674
    int rv;

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

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

1675 1676 1677
    if (postcopy)
        flags = QEMU_MIGRATION_COMPLETED_POSTCOPY;

1678 1679
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           NULL, flags)) != 1) {
1680 1681 1682 1683 1684 1685 1686 1687
        if (rv < 0 || virDomainObjWait(vm) < 0)
            return -1;
    }

    return 0;
}


1688
static int
1689 1690 1691 1692
qemuMigrationSrcGraphicsRelocate(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuMigrationCookiePtr cookie,
                                 const char *graphicsuri)
1693 1694
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1695 1696
    int ret = -1;
    const char *listenAddress = NULL;
1697
    virSocketAddr addr;
1698 1699 1700 1701 1702
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
1703

1704
    if (!cookie || (!cookie->graphics && !graphicsuri))
1705 1706
        return 0;

1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725
    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) {
1726
        size_t i;
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 1753

        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;
            }
        }
    }
1754 1755 1756 1757

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
1758 1759 1760 1761
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
1762

1763 1764 1765 1766 1767 1768 1769 1770 1771
    /* 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;
    }

1772 1773 1774 1775
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
1776
        priv->job.spiceMigration = !ret;
1777 1778
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
1779
    }
1780

1781
 cleanup:
1782
    virURIFree(uri);
1783 1784 1785 1786
    return ret;
}


1787
static int
1788 1789 1790
qemuMigrationDstOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            qemuMigrationCookiePtr cookie)
1791
{
1792 1793
    virDomainNetDefPtr netptr;
    int ret = -1;
1794
    size_t i;
1795 1796 1797 1798 1799 1800 1801 1802

    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:
1803
           break;
1804
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
1805 1806
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
1807 1808 1809
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
1810 1811 1812
                goto cleanup;
            }
            break;
1813 1814 1815 1816 1817
        default:
            break;
        }
    }

1818
    ret = 0;
1819
 cleanup:
1820 1821 1822 1823
    return ret;
}


1824
int
1825 1826
qemuMigrationDstCheckProtocol(virQEMUCapsPtr qemuCaps,
                              const char *migrateFrom)
1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849
{
    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 *
1850 1851
qemuMigrationDstGetURI(const char *migrateFrom,
                       int migrateFd)
1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863
{
    char *uri = NULL;

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

    return uri;
}


1864
int
1865 1866 1867 1868
qemuMigrationDstRun(virQEMUDriverPtr driver,
                    virDomainObjPtr vm,
                    const char *uri,
                    qemuDomainAsyncJob asyncJob)
1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884
{
    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) {
1885
        /* qemuMigrationDstWaitForCompletion is called from the Finish phase */
1886 1887 1888 1889
        ret = 0;
        goto cleanup;
    }

1890
    if (qemuMigrationDstWaitForCompletion(driver, vm, asyncJob, false) < 0)
1891 1892 1893 1894 1895 1896 1897 1898 1899
        goto cleanup;

    ret = 0;

 cleanup:
    return ret;
}


1900 1901 1902 1903 1904 1905
/* 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.
 */
1906
static void
1907 1908 1909
qemuMigrationSrcCleanup(virDomainObjPtr vm,
                        virConnectPtr conn,
                        void *opaque)
1910
{
1911
    virQEMUDriverPtr driver = opaque;
1912 1913 1914 1915 1916 1917 1918 1919 1920
    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))
1921
        return;
1922 1923 1924 1925 1926

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

1927
    switch ((qemuMigrationJobPhase) priv->job.phase) {
1928 1929 1930 1931 1932 1933 1934 1935 1936
    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);
1937
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
1938
                                 priv->job.migParams, priv->job.apiFlags);
1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960
        /* 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 */
        ;
    }
}

1961

1962
/* The caller is supposed to lock the vm and start a migration job. */
1963
static char *
1964 1965 1966 1967 1968 1969 1970 1971 1972
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)
1973 1974 1975
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
1976
    virDomainDefPtr def = NULL;
1977
    qemuDomainObjPrivatePtr priv = vm->privateData;
1978
    virCapsPtr caps = NULL;
1979
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
1980

1981
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
1982
              " cookieout=%p, cookieoutlen=%p,"
1983
              " nmigrate_disks=%zu, migrate_disks=%p, flags=0x%lx",
1984
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
1985 1986
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
1987

1988 1989 1990
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

1991 1992 1993 1994 1995 1996
    /* 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);
1997

1998
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
1999 2000
        goto cleanup;

2001
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
2002
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
2003 2004
        goto cleanup;

2005 2006 2007 2008 2009 2010 2011 2012 2013
    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;
    }

2014 2015 2016 2017 2018 2019
    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;
    }

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 2062
    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;
            }
2063 2064 2065
        }
    }

2066
    if (virDomainDefHasMemoryHotplug(vm->def) ||
2067
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
2068
         vm->newDef && virDomainDefHasMemoryHotplug(vm->newDef)))
2069 2070
        cookieFlags |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

2071 2072 2073 2074 2075
    if (!qemuDomainVcpuHotplugIsInOrder(vm->def) ||
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
         vm->newDef && !qemuDomainVcpuHotplugIsInOrder(vm->newDef)))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU_HOTPLUG;

2076 2077 2078
    if (priv->origCPU)
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU;

2079 2080
    cookieFlags |= QEMU_MIGRATION_COOKIE_ALLOW_REBOOT;

2081 2082 2083
    if (!(flags & VIR_MIGRATE_OFFLINE))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CAPS;

2084 2085
    if (!(mig = qemuMigrationEatCookie(driver, vm->def,
                                       priv->origname, priv, NULL, 0, 0)))
2086 2087 2088
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
2089
                                QEMU_MIGRATION_SOURCE,
2090
                                cookieout, cookieoutlen,
2091
                                cookieFlags) < 0)
2092 2093
        goto cleanup;

L
liguang 已提交
2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115
    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;
        }
    }

2116
    if (xmlin) {
2117
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt, priv->qemuCaps,
2118 2119
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                            VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2120 2121
            goto cleanup;

2122
        if (!qemuDomainCheckABIStability(driver, vm, def))
2123 2124
            goto cleanup;

2125
        rv = qemuDomainDefFormatLive(driver, def, NULL, false, true);
2126
    } else {
2127 2128
        rv = qemuDomainDefFormatLive(driver, vm->def, priv->origCPU,
                                     false, true);
2129
    }
2130

2131
 cleanup:
2132
    qemuMigrationCookieFree(mig);
2133
    virObjectUnref(caps);
2134
    virDomainDefFree(def);
2135 2136 2137
    return rv;
}

2138
char *
2139 2140 2141 2142 2143 2144 2145 2146 2147
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)
2148 2149 2150
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
2151
    qemuDomainAsyncJob asyncJob;
2152 2153

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
2154 2155
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
2156 2157 2158 2159 2160 2161 2162 2163
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_MIGRATION_OUT;
    } else {
        if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_NONE;
    }

2164
    qemuMigrationSrcStoreDomainState(vm);
2165

2166
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
2167 2168 2169 2170 2171 2172
        goto endjob;

    /* Check if there is any ejected media.
     * We don't want to require them on the destination.
     */
    if (!(flags & VIR_MIGRATE_OFFLINE) &&
2173
        qemuProcessRefreshDisks(driver, vm, asyncJob) < 0)
2174 2175
        goto endjob;

2176 2177 2178
    if (!(xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                           cookieout, cookieoutlen,
                                           nmigrate_disks, migrate_disks, flags)))
2179 2180 2181 2182 2183 2184 2185
        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.
         */
2186
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
2187
                                 qemuMigrationSrcCleanup) < 0) {
2188
            VIR_FREE(xml);
2189
            goto endjob;
2190
        }
2191
        qemuMigrationJobContinue(vm);
2192 2193 2194 2195
    } else {
        goto endjob;
    }

2196
 cleanup:
M
Michal Privoznik 已提交
2197
    virDomainObjEndAPI(&vm);
2198 2199
    return xml;

2200
 endjob:
2201 2202 2203 2204
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2205 2206 2207
    goto cleanup;
}

2208

2209 2210
/* Prepare is the first step, and it runs on the destination host.
 */
2211

2212
static void
2213 2214
qemuMigrationDstPrepareCleanup(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
2215 2216 2217 2218 2219 2220 2221 2222 2223
{
    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));

2224
    virPortAllocatorRelease(priv->migrationPort);
2225 2226
    priv->migrationPort = 0;

2227 2228 2229 2230 2231
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2232
static qemuProcessIncomingDefPtr
2233 2234 2235 2236 2237 2238
qemuMigrationDstPrepare(virDomainObjPtr vm,
                        bool tunnel,
                        const char *protocol,
                        const char *listenAddress,
                        unsigned short port,
                        int fd)
2239 2240
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2241
    qemuProcessIncomingDefPtr inc = NULL;
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 2271
    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 */
            }
2272
        } else if (hostIPv6Capable) {
2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295
            /* 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;
    }

2296 2297
    inc = qemuProcessIncomingDefNew(priv->qemuCaps, listenAddress,
                                    migrateFrom, fd, NULL);
2298

2299
 cleanup:
2300 2301
    VIR_FREE(migrateFrom);
    return inc;
2302 2303
}

2304
static int
2305
qemuMigrationDstPrepareAny(virQEMUDriverPtr driver,
2306
                           virConnectPtr dconn,
2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320
                           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,
2321
                           qemuMigrationParamsPtr migParams,
2322
                           unsigned long flags)
2323 2324
{
    virDomainObjPtr vm = NULL;
2325
    virObjectEventPtr event = NULL;
2326
    virErrorPtr origErr;
2327
    int ret = -1;
2328
    int dataFD[2] = { -1, -1 };
2329
    qemuDomainObjPrivatePtr priv = NULL;
2330
    qemuMigrationCookiePtr mig = NULL;
2331
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2332
    char *xmlout = NULL;
L
liguang 已提交
2333
    unsigned int cookieFlags;
2334
    unsigned int startFlags;
2335
    virCapsPtr caps = NULL;
2336
    qemuProcessIncomingDefPtr incoming = NULL;
2337
    bool taint_hook = false;
2338 2339 2340
    bool stopProcess = false;
    bool relabel = false;
    int rv;
2341
    char *tlsAlias = NULL;
2342

2343 2344
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364
    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;
        }
2365 2366
        cookieFlags = 0;
    } else {
2367 2368
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS |
                      QEMU_MIGRATION_COOKIE_CAPS;
L
liguang 已提交
2369 2370
    }

2371 2372 2373 2374 2375 2376 2377 2378 2379
    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;
    }

2380 2381 2382 2383 2384 2385
    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;
    }

2386 2387 2388
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2389
    if (!qemuMigrationSrcIsAllowedHostdev(*def))
2390 2391
        goto cleanup;

J
Jiri Denemark 已提交
2392 2393 2394 2395 2396
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2397
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2398 2399
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2400 2401
            goto cleanup;

2402
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2403 2404 2405 2406 2407 2408 2409
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2410
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2411 2412 2413 2414 2415 2416
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2417
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt, NULL,
2418 2419
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                                 VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE);
J
Jiri Denemark 已提交
2420 2421 2422
                if (!newdef)
                    goto cleanup;

2423
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2424 2425 2426 2427
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2428 2429
                virDomainDefFree(*def);
                *def = newdef;
2430 2431 2432 2433
                /* 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
2434 2435 2436 2437
            }
        }
    }

2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451
    /* 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;

2452
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2453
                                   driver->xmlopt,
2454 2455 2456
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2457
        goto cleanup;
2458
    *def = NULL;
2459

2460
    priv = vm->privateData;
2461 2462
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2463

2464 2465 2466 2467 2468
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2469 2470
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
2471 2472 2473 2474 2475 2476
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2477 2478 2479
    if (qemuMigrationDstPrecreateStorage(vm, mig->nbd,
                                         nmigrate_disks, migrate_disks,
                                         !!(flags & VIR_MIGRATE_NON_SHARED_INC)) < 0)
2480 2481
        goto cleanup;

2482 2483
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                              flags) < 0)
2484
        goto cleanup;
2485
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2486 2487 2488 2489

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

L
liguang 已提交
2490 2491 2492
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2493 2494
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
2495 2496
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
2497
        goto stopjob;
2498 2499
    }

2500 2501
    startFlags = VIR_QEMU_PROCESS_START_AUTODESTROY;

2502
    if (qemuProcessInit(driver, vm, mig->cpu, QEMU_ASYNC_JOB_MIGRATION_IN,
2503
                        true, startFlags) < 0)
2504
        goto stopjob;
2505
    stopProcess = true;
2506

2507 2508
    priv->allowReboot = mig->allowReboot;

2509 2510 2511
    if (!(incoming = qemuMigrationDstPrepare(vm, tunnel, protocol,
                                             listenAddress, port,
                                             dataFD[0])))
2512
        goto stopjob;
2513

2514
    if (qemuProcessPrepareDomain(driver, vm, startFlags) < 0)
2515 2516
        goto stopjob;

2517
    if (qemuProcessPrepareHost(driver, vm, startFlags) < 0)
2518 2519
        goto stopjob;

2520
    rv = qemuProcessLaunch(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2521 2522
                           incoming, NULL,
                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
2523
                           startFlags);
2524 2525 2526
    if (rv < 0) {
        if (rv == -2)
            relabel = true;
2527
        goto stopjob;
2528
    }
2529
    relabel = true;
2530

2531 2532 2533 2534
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
2535
            goto stopjob;
2536
        }
2537
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
2538 2539
    }

2540
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
2541
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
2542
        goto stopjob;
M
Michael R. Hines 已提交
2543 2544
    }

2545
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2546
                                 migParams, mig->caps->automatic) < 0)
2547 2548 2549 2550 2551 2552 2553
        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,
2554
                                         &tlsAlias, NULL,
2555 2556 2557 2558 2559 2560 2561
                                         migParams) < 0)
            goto stopjob;
    } else {
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
            goto stopjob;
    }

2562 2563
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                                 migParams) < 0)
2564 2565
        goto stopjob;

2566 2567 2568
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580
        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;
        }

2581 2582
        if (qemuMigrationDstStartNBDServer(driver, vm, incoming->address,
                                           nmigrate_disks, migrate_disks,
2583
                                           nbdPort, nbdTLSAlias) < 0) {
2584
            goto stopjob;
2585
        }
2586
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
2587 2588
    }

2589 2590 2591 2592 2593 2594 2595 2596 2597
    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");
    }

2598
    if (incoming->deferredURI &&
2599 2600
        qemuMigrationDstRun(driver, vm, incoming->deferredURI,
                            QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2601 2602
        goto stopjob;

2603
    if (qemuProcessFinishStartup(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2604 2605 2606
                                 false, VIR_DOMAIN_PAUSED_MIGRATION) < 0)
        goto stopjob;

2607
 done:
2608 2609 2610
    if (qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_DESTINATION,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
2611 2612 2613 2614 2615 2616 2617
        /* 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");
    }

2618
    if (qemuDomainCleanupAdd(vm, qemuMigrationDstPrepareCleanup) < 0)
2619
        goto stopjob;
2620

L
liguang 已提交
2621 2622
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
2623
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
2624 2625 2626
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
2627

2628 2629 2630 2631
    /* 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.
     */
2632
    qemuMigrationJobContinue(vm);
2633

2634 2635
    if (autoPort)
        priv->migrationPort = port;
2636 2637 2638 2639 2640
    /* 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;
2641
    ret = 0;
2642

2643
 cleanup:
2644
    virErrorPreserveLast(&origErr);
2645
    VIR_FREE(tlsAlias);
2646
    qemuProcessIncomingDefFree(incoming);
J
Jiri Denemark 已提交
2647
    VIR_FREE(xmlout);
2648 2649
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
2650 2651 2652
    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 */
2653
        VIR_FREE(priv->origname);
2654 2655 2656 2657
        /* release if port is auto selected which is not the case if
         * it is given in parameters
         */
        if (nbdPort == 0)
2658
            virPortAllocatorRelease(priv->nbdPort);
2659
        priv->nbdPort = 0;
2660
        virDomainObjRemoveTransientDef(vm);
2661
        qemuDomainRemoveInactiveJob(driver, vm);
2662
    }
M
Michal Privoznik 已提交
2663
    virDomainObjEndAPI(&vm);
2664
    virObjectEventStateQueue(driver->domainEventState, event);
2665
    qemuMigrationCookieFree(mig);
2666
    virObjectUnref(caps);
2667
    virNWFilterUnlockFilterUpdates();
2668
    virErrorRestore(&origErr);
2669
    return ret;
2670

2671
 stopjob:
2672
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2673
                             priv->job.migParams, priv->job.apiFlags);
2674

2675 2676 2677 2678
    if (stopProcess) {
        unsigned int stopFlags = VIR_QEMU_PROCESS_STOP_MIGRATED;
        if (!relabel)
            stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
2679
        virDomainAuditStart(vm, "migrated", false);
2680 2681
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        QEMU_ASYNC_JOB_MIGRATION_IN, stopFlags);
2682
    }
2683

2684
    qemuMigrationJobFinish(driver, vm);
2685
    goto cleanup;
2686 2687 2688
}


2689 2690 2691 2692 2693
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
2694
qemuMigrationDstPrepareTunnel(virQEMUDriverPtr driver,
2695
                              virConnectPtr dconn,
2696 2697 2698 2699 2700 2701 2702
                              const char *cookiein,
                              int cookieinlen,
                              char **cookieout,
                              int *cookieoutlen,
                              virStreamPtr st,
                              virDomainDefPtr *def,
                              const char *origname,
2703
                              qemuMigrationParamsPtr migParams,
2704
                              unsigned long flags)
2705
{
2706
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2707
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
2708
              "origname=%s, flags=0x%lx",
2709
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2710
              cookieout, cookieoutlen, st, *def, origname, flags);
2711

2712 2713 2714 2715 2716 2717
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

2718 2719 2720
    return qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
                                      cookieout, cookieoutlen, def, origname,
                                      st, NULL, 0, false, NULL, 0, NULL, 0,
2721
                                      migParams, flags);
2722 2723 2724
}


2725
static virURIPtr
2726
qemuMigrationAnyParseURI(const char *uri, bool *wellFormed)
2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747
{
    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;
}


2748
int
2749
qemuMigrationDstPrepareDirect(virQEMUDriverPtr driver,
2750
                              virConnectPtr dconn,
2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762
                              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,
2763
                              qemuMigrationParamsPtr migParams,
2764
                              unsigned long flags)
2765
{
2766 2767
    unsigned short port = 0;
    bool autoPort = true;
2768 2769
    char *hostname = NULL;
    int ret = -1;
2770
    virURIPtr uri = NULL;
2771 2772
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
2773

2774
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2775
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
2776
              "def=%p, origname=%s, listenAddress=%s, "
2777
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, flags=0x%lx",
2778
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2779
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
2780
              *def, origname, NULLSTR(listenAddress),
2781
              nmigrate_disks, migrate_disks, nbdPort, flags);
2782

2783 2784
    *uri_out = NULL;

2785 2786 2787
    /* 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
2788 2789 2790
     * 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".
2791 2792 2793 2794 2795 2796
     *
     * 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) {
2797 2798 2799
        bool encloseAddress = false;
        const char *incFormat;

2800
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2801
            goto cleanup;
2802

2803
        if (migrateHost != NULL) {
2804 2805
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
2806

2807
            if (VIR_STRDUP(hostname, migrateHost) < 0)
2808 2809 2810 2811 2812
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
2813 2814

        if (STRPREFIX(hostname, "localhost")) {
2815 2816 2817
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
2818 2819 2820 2821 2822
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
2823
         * compatibility with old targets. We at least make the
2824 2825
         * new targets accept both syntaxes though.
         */
2826 2827 2828 2829 2830 2831
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
2832 2833
            goto cleanup;
    } else {
2834
        bool well_formed_uri;
J
Ján Tomko 已提交
2835

2836
        if (!(uri = qemuMigrationAnyParseURI(uri_in, &well_formed_uri)))
2837
            goto cleanup;
J
Ján Tomko 已提交
2838

2839 2840 2841 2842 2843 2844 2845
        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
2846 2847
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
2848 2849 2850
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
2851 2852 2853 2854 2855 2856 2857 2858 2859 2860
            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) {
2861
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2862
                goto cleanup;
2863

2864
            /* Send well-formed URI only if uri_in was well-formed */
2865 2866 2867 2868 2869 2870 2871 2872
            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;
            }
2873
        } else {
2874 2875
            port = uri->port;
            autoPort = false;
2876 2877 2878 2879 2880 2881
        }
    }

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

2882
    ret = qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
2883 2884 2885 2886
                                     cookieout, cookieoutlen, def, origname,
                                     NULL, uri ? uri->scheme : "tcp",
                                     port, autoPort, listenAddress,
                                     nmigrate_disks, migrate_disks, nbdPort,
2887
                                     migParams, flags);
2888
 cleanup:
2889
    virURIFree(uri);
2890
    VIR_FREE(hostname);
2891
    virObjectUnref(cfg);
2892
    if (ret != 0) {
2893
        VIR_FREE(*uri_out);
2894
        if (autoPort)
2895
            virPortAllocatorRelease(port);
2896
    }
2897 2898 2899 2900
    return ret;
}


2901
virDomainDefPtr
2902 2903 2904 2905
qemuMigrationAnyPrepareDef(virQEMUDriverPtr driver,
                           const char *dom_xml,
                           const char *dname,
                           char **origname)
2906 2907 2908
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
2909
    char *name = NULL;
2910 2911 2912 2913 2914 2915 2916 2917 2918 2919

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

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

2920
    if (!(def = virDomainDefParseString(dom_xml, caps, driver->xmlopt, NULL,
2921 2922
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                        VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2923 2924 2925
        goto cleanup;

    if (dname) {
2926
        name = def->name;
2927 2928 2929 2930 2931 2932
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

2933
 cleanup:
2934
    virObjectUnref(caps);
2935 2936 2937 2938
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
2939 2940 2941 2942
    return def;
}


2943
static int
2944 2945 2946 2947 2948 2949
qemuMigrationSrcConfirmPhase(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
                             const char *cookiein,
                             int cookieinlen,
                             unsigned int flags,
                             int retcode)
2950 2951
{
    qemuMigrationCookiePtr mig;
2952
    virObjectEventPtr event;
2953 2954
    int rv = -1;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2955 2956
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = NULL;
2957

2958
    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
2959
              "flags=0x%x, retcode=%d",
2960
              driver, vm, NULLSTR(cookiein), cookieinlen,
2961 2962 2963 2964 2965 2966 2967 2968 2969
              flags, retcode);

    virCheckFlags(QEMU_MIGRATION_FLAGS, -1);

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

2970 2971
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen,
2972
                                       QEMU_MIGRATION_COOKIE_STATS)))
2973 2974
        goto cleanup;

2975 2976 2977
    if (retcode == 0)
        jobInfo = priv->job.completed;
    else
2978
        VIR_FREE(priv->job.completed);
2979 2980 2981

    /* Update times with the values sent by the destination daemon */
    if (mig->jobInfo && jobInfo) {
2982 2983 2984 2985 2986 2987 2988 2989
        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 &&
2990 2991
            qemuMigrationAnyFetchStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                       jobInfo, NULL) < 0)
2992 2993
            VIR_WARN("Could not refresh migration statistics");

2994 2995 2996
        qemuDomainJobInfoUpdateTime(jobInfo);
        jobInfo->timeDeltaSet = mig->jobInfo->timeDeltaSet;
        jobInfo->timeDelta = mig->jobInfo->timeDelta;
2997 2998
        jobInfo->stats.mig.downtime_set = mig->jobInfo->stats.mig.downtime_set;
        jobInfo->stats.mig.downtime = mig->jobInfo->stats.mig.downtime;
2999 3000
    }

3001 3002 3003
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

3004 3005
    /* 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.
3006 3007 3008 3009
     */
    if (retcode == 0) {
        /* If guest uses SPICE and supports seamless migration we have to hold
         * up domain shutdown until SPICE server transfers its data */
3010
        qemuMigrationSrcWaitForSpice(vm);
3011 3012

        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
3013
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
3014 3015 3016
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
        virDomainAuditStop(vm, "migrated");

3017
        event = virDomainEventLifecycleNewFromObj(vm,
3018 3019
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
3020
        virObjectEventStateQueue(driver->domainEventState, event);
3021
        qemuDomainEventEmitJobCompleted(driver, vm);
3022
    } else {
3023
        virErrorPtr orig_err = virSaveLastError();
3024
        int reason;
3025 3026

        /* cancel any outstanding NBD jobs */
3027 3028
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT, NULL);
3029 3030 3031

        virSetError(orig_err);
        virFreeError(orig_err);
3032

3033
        if (virDomainObjGetState(vm, &reason) == VIR_DOMAIN_PAUSED &&
3034
            reason == VIR_DOMAIN_PAUSED_POSTCOPY)
3035
            qemuMigrationAnyPostcopyFailed(driver, vm);
3036 3037
        else
            qemuMigrationSrcRestoreDomainState(driver, vm);
3038

3039
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3040
                                 priv->job.migParams, priv->job.apiFlags);
3041

3042
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
3043 3044 3045
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
    }

3046
 done:
3047 3048 3049
    qemuMigrationCookieFree(mig);
    rv = 0;

3050
 cleanup:
3051 3052 3053 3054 3055
    virObjectUnref(cfg);
    return rv;
}

int
3056 3057 3058 3059 3060 3061
qemuMigrationSrcConfirm(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *cookiein,
                        int cookieinlen,
                        unsigned int flags,
                        int cancelled)
3062
{
3063
    qemuMigrationJobPhase phase;
3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077
    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);
3078
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
3079
                           qemuMigrationSrcCleanup);
3080

3081 3082 3083
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
3084

3085
    qemuMigrationJobFinish(driver, vm);
3086
    if (!virDomainObjIsActive(vm)) {
3087
        if (!cancelled && ret == 0 && flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
3088
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
3089 3090
            vm->persistent = 0;
        }
3091
        qemuDomainRemoveInactiveJob(driver, vm);
3092 3093
    }

3094
 cleanup:
M
Michal Privoznik 已提交
3095
    virDomainObjEndAPI(&vm);
3096 3097 3098 3099 3100
    virObjectUnref(cfg);
    return ret;
}


3101 3102
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3103
    MIGRATION_DEST_CONNECT_HOST,
3104
    MIGRATION_DEST_FD,
3105
};
3106

3107 3108 3109 3110
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3111

3112 3113 3114 3115 3116 3117
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3118
            const char *protocol;
3119 3120 3121 3122
            const char *name;
            int port;
        } host;

3123 3124 3125 3126
        struct {
            int qemu;
            int local;
        } fd;
3127 3128 3129 3130 3131 3132 3133
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3134 3135 3136

#define TUNNEL_SEND_BUF_SIZE 65536

3137 3138 3139 3140 3141 3142 3143
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3144 3145
    int wakeupRecvFD;
    int wakeupSendFD;
3146 3147
};

3148
static void qemuMigrationSrcIOFunc(void *arg)
3149
{
3150
    qemuMigrationIOThreadPtr data = arg;
3151 3152 3153 3154 3155 3156 3157
    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);
3158

3159
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3160
        goto abrt;
3161

3162 3163 3164
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3165
    for (;;) {
3166 3167 3168 3169 3170 3171 3172 3173 3174 3175
        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;
3176
            virReportSystemError(errno, "%s",
3177 3178
                                 _("poll failed in migration tunnel"));
            goto abrt;
3179
        }
3180 3181 3182 3183 3184 3185 3186

        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");
3187
            break;
3188
        }
3189

3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205
        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;
            }
3206 3207
        }

3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224
        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;
            }
        }
    }
3225

3226 3227
    if (virStreamFinish(data->st) < 0)
        goto error;
3228

3229
    VIR_FORCE_CLOSE(data->sock);
3230 3231
    VIR_FREE(buffer);

3232 3233
    return;

3234
 abrt:
3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3246
 error:
3247 3248 3249 3250 3251
    /* 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);
3252
    virResetLastError();
3253
    VIR_FREE(buffer);
3254 3255 3256 3257
}


static qemuMigrationIOThreadPtr
3258 3259
qemuMigrationSrcStartTunnel(virStreamPtr st,
                            int sock)
3260
{
3261 3262
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3263

3264 3265 3266 3267
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3268 3269
    }

3270
    if (VIR_ALLOC(io) < 0)
3271
        goto error;
3272

3273 3274
    io->st = st;
    io->sock = sock;
3275 3276
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3277 3278

    if (virThreadCreate(&io->thread, true,
3279
                        qemuMigrationSrcIOFunc,
3280 3281 3282
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3283
        goto error;
3284 3285 3286
    }

    return io;
3287

3288
 error:
3289 3290 3291 3292
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3293 3294 3295
}

static int
3296
qemuMigrationSrcStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3297 3298
{
    int rv = -1;
3299 3300 3301 3302 3303 3304 3305 3306 3307
    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;
    }

3308 3309 3310 3311
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3312 3313 3314 3315
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3316 3317 3318 3319 3320 3321
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3322
 cleanup:
3323 3324
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3325 3326
    VIR_FREE(io);
    return rv;
3327 3328
}

3329
static int
3330 3331 3332
qemuMigrationSrcConnect(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationSpecPtr spec)
3333 3334 3335 3336 3337 3338 3339
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3340
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3341 3342 3343 3344 3345
        return -1;

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

3346
    if (qemuSecuritySetSocketLabel(driver->securityManager, vm->def) < 0)
3347
        goto cleanup;
3348 3349 3350
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
3351
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
3352
        virObjectUnref(sock);
3353
    }
3354
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0 ||
3355 3356 3357
        spec->dest.fd.qemu == -1)
        goto cleanup;

3358 3359 3360 3361 3362 3363 3364
    /* 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;
    }

3365 3366
    ret = 0;

3367
 cleanup:
3368 3369 3370 3371 3372 3373
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3374 3375

static int
3376 3377 3378 3379
qemuMigrationSrcContinue(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMonitorMigrationStatus status,
                         qemuDomainAsyncJob asyncJob)
3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395
{
    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;
}


3396
static int
3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410
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,
3411
                    qemuMigrationParamsPtr migParams)
3412
{
3413
    int ret = -1;
3414 3415
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3416
    qemuMigrationCookiePtr mig = NULL;
3417
    char *tlsAlias = NULL;
3418
    qemuMigrationIOThreadPtr iothread = NULL;
3419
    int fd = -1;
3420
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3421
    virErrorPtr orig_err = NULL;
3422
    unsigned int cookieFlags = 0;
3423
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3424
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
3425
    bool cancel = false;
3426
    unsigned int waitFlags;
3427
    virDomainDefPtr persistDef = NULL;
3428
    char *timestamp;
3429
    int rc;
3430 3431

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
3432
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3433 3434
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
3435 3436
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3437
              spec, spec->destType, spec->fwdType, dconn,
3438
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3439

3440 3441 3442 3443 3444 3445 3446 3447 3448 3449
    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;
    }

3450 3451
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3452 3453 3454 3455
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3456 3457 3458
        return -1;
    }

3459 3460 3461
    if (events)
        priv->signalIOError = abort_on_error;

3462 3463
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
        if (persist_xml) {
3464 3465
            if (!(persistDef = qemuMigrationAnyPrepareDef(driver, persist_xml,
                                                          NULL, NULL)))
3466
                goto error;
3467 3468 3469
        } else {
            virDomainDefPtr def = vm->newDef ? vm->newDef : vm->def;
            if (!(persistDef = qemuDomainDefCopy(driver, def,
3470 3471
                                                 VIR_DOMAIN_XML_SECURE |
                                                 VIR_DOMAIN_XML_MIGRATABLE)))
3472
                goto error;
3473 3474 3475
        }
    }

3476 3477
    mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                 cookiein, cookieinlen,
3478 3479 3480
                                 cookieFlags |
                                 QEMU_MIGRATION_COOKIE_GRAPHICS |
                                 QEMU_MIGRATION_COOKIE_CAPS);
3481
    if (!mig)
3482
        goto error;
3483

3484
    if (qemuMigrationSrcGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3485 3486
        VIR_WARN("unable to provide data for graphics client relocation");

3487
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3488
                                 migParams, mig->caps->automatic) < 0)
3489 3490
        goto error;

3491
    if (flags & VIR_MIGRATE_TLS) {
3492
        const char *hostname = NULL;
3493 3494 3495 3496

        /* We need to add tls-hostname whenever QEMU itself does not
         * connect directly to the destination. */
        if (spec->destType == MIGRATION_DEST_CONNECT_HOST ||
3497 3498 3499
            spec->destType == MIGRATION_DEST_FD)
            hostname = spec->dest.host.name;

3500
        if (qemuMigrationParamsEnableTLS(driver, vm, false,
3501
                                         QEMU_ASYNC_JOB_MIGRATION_OUT,
3502
                                         &tlsAlias, hostname,
3503 3504
                                         migParams) < 0)
            goto error;
3505
    } else {
3506
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
3507
            goto error;
3508 3509
    }

3510 3511 3512 3513
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                 migParams) < 0)
        goto error;

3514 3515 3516
    if (migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                         QEMU_MONITOR_MIGRATE_NON_SHARED_INC)) {
        if (mig->nbd) {
3517 3518 3519 3520
            /* 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. */
3521 3522
            if (flags & VIR_MIGRATE_TLS &&
                !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV_DEL)) {
3523 3524 3525 3526 3527
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                               _("NBD migration with TLS is not supported"));
                goto error;
            }

3528
            /* This will update migrate_flags on success */
3529 3530 3531 3532 3533 3534
            if (qemuMigrationSrcNBDStorageCopy(driver, vm, mig,
                                               spec->dest.host.name,
                                               migrate_speed,
                                               &migrate_flags,
                                               nmigrate_disks,
                                               migrate_disks,
3535
                                               dconn, tlsAlias, flags) < 0) {
3536
                goto error;
3537 3538 3539 3540 3541 3542 3543
            }
        } 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.");
        }
3544 3545
    }

3546
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3547 3548
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
3549
        if (qemuMigrationSrcSetOffline(driver, vm) < 0)
3550
            goto error;
3551 3552
    }

3553 3554
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3555
        goto error;
3556

3557
    if (priv->job.abortJob) {
3558 3559
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
3560
         * priv->job.abortJob will not change */
3561
        priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
3562 3563 3564
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
3565
        goto exit_monitor;
3566 3567
    }

3568 3569
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
3570

3571 3572
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3573
        qemuMigrationSrcConnect(driver, vm, spec) < 0) {
3574
        goto exit_monitor;
3575
    }
3576

3577 3578 3579 3580 3581 3582
    /* log start of migration */
    if ((timestamp = virTimeStringNow()) != NULL) {
        qemuDomainLogAppendMessage(driver, vm, "%s: initiating migration\n", timestamp);
        VIR_FREE(timestamp);
    }

3583
    rc = -1;
3584 3585
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3586 3587
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
3588
            goto exit_monitor;
M
Michael R. Hines 已提交
3589
        }
3590 3591 3592 3593
        rc = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
                                      spec->dest.host.protocol,
                                      spec->dest.host.name,
                                      spec->dest.host.port);
3594 3595
        break;

3596 3597 3598 3599
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3600
    case MIGRATION_DEST_FD:
3601
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3602
            fd = spec->dest.fd.local;
3603 3604
            spec->dest.fd.local = -1;
        }
3605 3606
        rc = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                    spec->dest.fd.qemu);
3607 3608
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3609
    }
3610 3611

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
3612
        goto error;
3613 3614 3615

    /* From this point onwards we *must* call cancel to abort the
     * migration on source if anything goes wrong */
3616
    cancel = true;
3617

3618
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3619
        if (!(iothread = qemuMigrationSrcStartTunnel(spec->fwd.stream, fd)))
3620
            goto error;
3621 3622 3623 3624 3625
        /* If we've created a tunnel, then the 'fd' will be closed in the
         * qemuMigrationIOFunc as data->sock.
         */
        fd = -1;
    }
3626

3627
    waitFlags = QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;
3628 3629 3630 3631 3632 3633 3634
    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;

3635 3636 3637
    rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT,
                                           dconn, waitFlags);
3638
    if (rc == -2) {
3639
        goto error;
3640 3641 3642 3643 3644
    } else if (rc == -1) {
        /* QEMU reported failed migration, nothing to cancel anymore */
        cancel = false;
        goto error;
    }
3645

3646 3647 3648
    /* 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.
3649
     */
3650 3651 3652 3653 3654
    if (priv->monJSON) {
        while (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            priv->signalStop = true;
            rc = virDomainObjWait(vm);
            priv->signalStop = false;
3655
            if (rc < 0)
3656
                goto error;
3657
        }
3658
    } else if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING &&
3659
               qemuMigrationSrcSetOffline(driver, vm) < 0) {
3660
        goto error;
3661
    }
3662

J
Ján Tomko 已提交
3663
    if (mig->nbd &&
3664 3665 3666
        qemuMigrationSrcNBDCopyCancel(driver, vm, true,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT,
                                      dconn) < 0)
3667
        goto error;
3668

3669 3670 3671 3672 3673
    /* 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) {
3674 3675 3676
        if (qemuMigrationSrcContinue(driver, vm,
                                     QEMU_MONITOR_MIGRATION_STATUS_PRE_SWITCHOVER,
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3677 3678 3679 3680
            goto error;

        waitFlags ^= QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;

3681 3682 3683
        rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_OUT,
                                               dconn, waitFlags);
3684 3685 3686 3687 3688 3689 3690 3691 3692
        if (rc == -2) {
            goto error;
        } else if (rc == -1) {
            /* QEMU reported failed migration, nothing to cancel anymore */
            cancel = false;
            goto error;
        }
    }

3693 3694 3695 3696
    if (iothread) {
        qemuMigrationIOThreadPtr io;

        VIR_STEAL_PTR(io, iothread);
3697
        if (qemuMigrationSrcStopTunnel(io, false) < 0)
3698
            goto error;
3699 3700 3701
    }

    if (priv->job.completed) {
3702
        priv->job.completed->stopped = priv->job.current->stopped;
3703 3704 3705 3706 3707 3708 3709 3710 3711
        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 ||
3712 3713 3714
        qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_SOURCE,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
3715 3716
        VIR_WARN("Unable to encode migration cookie");
    }
3717

3718
    ret = 0;
3719

3720
 cleanup:
3721
    VIR_FREE(tlsAlias);
3722
    VIR_FORCE_CLOSE(fd);
3723
    virDomainDefFree(persistDef);
3724 3725
    qemuMigrationCookieFree(mig);

3726 3727 3728
    if (events)
        priv->signalIOError = false;

3729 3730 3731 3732 3733
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

3734 3735
    return ret;

3736
 error:
3737 3738
    orig_err = virSaveLastError();

3739 3740 3741 3742 3743 3744 3745 3746
    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));
        }
3747

3748 3749
        /* cancel any outstanding NBD jobs */
        if (mig && mig->nbd)
3750 3751 3752
            qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                          QEMU_ASYNC_JOB_MIGRATION_OUT,
                                          dconn);
3753 3754 3755 3756

        if (priv->job.current->status != QEMU_DOMAIN_JOB_STATUS_CANCELED)
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
    }
3757 3758

    if (iothread)
3759
        qemuMigrationSrcStopTunnel(iothread, true);
3760 3761 3762

    goto cleanup;

3763 3764
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
3765
    goto error;
3766 3767
}

3768
/* Perform migration using QEMU's native migrate support,
3769 3770
 * not encrypted obviously
 */
3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785
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,
3786
                              qemuMigrationParamsPtr migParams)
3787
{
3788
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
3789
    virURIPtr uribits = NULL;
3790
    int ret = -1;
3791 3792 3793
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
3794
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3795
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
3796
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
3797
              cookieout, cookieoutlen, flags, resource,
3798
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3799

3800
    if (!(uribits = qemuMigrationAnyParseURI(uri, NULL)))
3801 3802
        return -1;

3803 3804 3805 3806 3807 3808 3809
    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
3810 3811 3812 3813 3814
    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"));
3815
            goto cleanup;
M
Michael R. Hines 已提交
3816
        }
3817
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
3818 3819 3820
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
3821
            goto cleanup;
M
Michael R. Hines 已提交
3822 3823 3824
        }
    }

3825
    if (STRNEQ(uribits->scheme, "rdma"))
3826 3827
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
3828
        spec.destType = MIGRATION_DEST_HOST;
3829
    spec.dest.host.protocol = uribits->scheme;
3830 3831 3832
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
3833

3834 3835 3836
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen, cookieout,
                              cookieoutlen, flags, resource, &spec, dconn,
                              graphicsuri, nmigrate_disks, migrate_disks,
3837
                              migParams);
3838 3839 3840 3841

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

3842
 cleanup:
3843
    virURIFree(uribits);
3844 3845 3846 3847 3848

    return ret;
}


3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863
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,
3864
                              qemuMigrationParamsPtr migParams)
3865 3866 3867
{
    int ret = -1;
    qemuMigrationSpec spec;
3868
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3869
    int fds[2] = { -1, -1 };
3870 3871

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
3872
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3873
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
3874
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
3875
              cookieout, cookieoutlen, flags, resource,
3876
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3877 3878 3879 3880

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

3881

3882 3883 3884
    spec.destType = MIGRATION_DEST_FD;
    spec.dest.fd.qemu = -1;
    spec.dest.fd.local = -1;
3885

3886 3887 3888 3889 3890
    if (pipe2(fds, O_CLOEXEC) == 0) {
        spec.dest.fd.qemu = fds[1];
        spec.dest.fd.local = fds[0];
    }
    if (spec.dest.fd.qemu == -1 ||
3891 3892
        qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    spec.dest.fd.qemu) < 0) {
3893 3894 3895
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
        goto cleanup;
3896 3897
    }

3898 3899 3900
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen,
                              cookieout, cookieoutlen, flags, resource, &spec,
                              dconn, graphicsuri, nmigrate_disks, migrate_disks,
3901
                              migParams);
3902

3903
 cleanup:
3904 3905
    VIR_FORCE_CLOSE(spec.dest.fd.qemu);
    VIR_FORCE_CLOSE(spec.dest.fd.local);
3906

3907
    virObjectUnref(cfg);
3908 3909 3910 3911
    return ret;
}


3912 3913 3914 3915
/* 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 */
3916 3917 3918 3919 3920 3921 3922 3923
static int
qemuMigrationSrcPerformPeer2Peer2(virQEMUDriverPtr driver,
                                  virConnectPtr sconn,
                                  virConnectPtr dconn,
                                  virDomainObjPtr vm,
                                  const char *dconnuri,
                                  unsigned long flags,
                                  const char *dname,
3924 3925
                                  unsigned long resource,
                                  qemuMigrationParamsPtr migParams)
3926 3927 3928
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
3929
    char *cookie = NULL;
3930 3931 3932
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
3933
    bool cancelled;
3934
    virStreamPtr st = NULL;
3935
    unsigned long destflags;
3936

3937
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
3938
              "flags=0x%lx, dname=%s, resource=%lu",
3939 3940
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
3941

3942 3943 3944 3945 3946
    /* 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,
3947 3948
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
3949 3950 3951 3952 3953
        return -1;

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

3954 3955
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
3956

3957 3958 3959 3960 3961 3962 3963 3964 3965 3966
    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;

3967
        qemuDomainObjEnterRemote(vm);
3968
        ret = dconn->driver->domainMigratePrepareTunnel
3969
            (dconn, st, destflags, dname, resource, dom_xml);
3970 3971
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3972
    } else {
3973
        qemuDomainObjEnterRemote(vm);
3974 3975
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
3976
             destflags, dname, resource, dom_xml);
3977 3978
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3979 3980 3981
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
3982 3983
        goto cleanup;

3984 3985
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
3986 3987
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
3988
        cancelled = true;
3989
        orig_err = virSaveLastError();
3990
        goto finish;
3991 3992
    }

3993 3994 3995 3996
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
3997
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
3998
    if (flags & VIR_MIGRATE_TUNNELLED)
3999 4000 4001
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, NULL,
                                            NULL, 0, NULL, NULL,
                                            flags, resource, dconn,
4002
                                            NULL, 0, NULL, migParams);
4003
    else
4004 4005 4006 4007
        ret = qemuMigrationSrcPerformNative(driver, vm, NULL, uri_out,
                                            cookie, cookielen,
                                            NULL, NULL, /* No out cookie with v2 migration */
                                            flags, resource, dconn, NULL, 0, NULL,
4008
                                            migParams);
4009 4010 4011 4012

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4013

4014 4015 4016
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4017
    cancelled = ret < 0;
4018

4019
 finish:
4020 4021 4022 4023
    /* 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.
     */
4024
    dname = dname ? dname : vm->def->name;
4025
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4026
    qemuDomainObjEnterRemote(vm);
4027
    ddomain = dconn->driver->domainMigrateFinish2
4028
        (dconn, dname, cookie, cookielen,
4029
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4030 4031
    /* The domain is already gone at this point */
    ignore_value(qemuDomainObjExitRemote(vm, false));
4032 4033
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4034

4035
 cleanup:
4036
    if (ddomain) {
4037
        virObjectUnref(ddomain);
4038 4039 4040 4041
        ret = 0;
    } else {
        ret = -1;
    }
4042

4043
    virObjectUnref(st);
4044 4045 4046 4047 4048 4049

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4050
    VIR_FREE(cookie);
4051 4052

    return ret;
4053 4054 4055
}


4056 4057 4058 4059
/* 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 */
4060
static int
4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074
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,
4075
                                  qemuMigrationParamsPtr migParams,
4076 4077 4078
                                  unsigned long long bandwidth,
                                  bool useParams,
                                  unsigned long flags)
4079 4080 4081 4082 4083 4084 4085 4086 4087 4088
{
    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;
4089
    bool cancelled = true;
4090
    virStreamPtr st = NULL;
4091
    unsigned long destflags;
4092 4093 4094
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4095
    size_t i;
4096
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4097 4098

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4099
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4100
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4101
              "bandwidth=%llu, useParams=%d, flags=0x%lx",
4102
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4103
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4104
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks, nbdPort,
4105
              bandwidth, useParams, flags);
4106

4107 4108 4109 4110 4111
    /* 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.  */

4112 4113 4114
    dom_xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                         &cookieout, &cookieoutlen,
                                         nmigrate_disks, migrate_disks, flags);
4115 4116 4117
    if (!dom_xml)
        goto cleanup;

4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137
    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;
4138 4139 4140 4141 4142 4143

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4144 4145 4146 4147 4148
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4149 4150 4151 4152 4153
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4154 4155 4156 4157 4158
        if (nbdPort &&
            virTypedParamsAddInt(&params, &nparams, &maxparams,
                                 VIR_MIGRATE_PARAM_DISKS_PORT,
                                 nbdPort) < 0)
            goto cleanup;
4159

4160 4161
        if (qemuMigrationParamsDump(migParams, &params, &nparams,
                                    &maxparams, &flags) < 0)
4162
            goto cleanup;
4163 4164
    }

4165 4166 4167
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4168 4169
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4170

4171
    VIR_DEBUG("Prepare3 %p", dconn);
4172
    VIR_STEAL_PTR(cookiein, cookieout);
4173 4174 4175 4176 4177 4178
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
    if (flags & VIR_MIGRATE_TUNNELLED) {
        if (!(st = virStreamNew(dconn, 0)))
            goto cleanup;

4179
        qemuDomainObjEnterRemote(vm);
4180 4181 4182 4183 4184 4185 4186 4187 4188
        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);
        }
4189 4190
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4191
    } else {
4192
        qemuDomainObjEnterRemote(vm);
4193 4194 4195 4196 4197 4198 4199 4200 4201
        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);
        }
4202 4203
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4204 4205 4206 4207 4208
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

4209
    if (offline) {
L
liguang 已提交
4210 4211 4212
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4213
        cancelled = false;
L
liguang 已提交
4214 4215 4216
        goto finish;
    }

4217 4218 4219 4220
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4221 4222
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4223
            goto finish;
4224
        }
4225
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4226 4227
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4228
        orig_err = virSaveLastError();
4229 4230 4231 4232 4233 4234 4235 4236
        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.
     */
4237
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4238
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4239
    VIR_FREE(cookiein);
4240
    VIR_STEAL_PTR(cookiein, cookieout);
4241 4242
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
4243
    if (flags & VIR_MIGRATE_TUNNELLED) {
4244 4245 4246 4247
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, persist_xml,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4248
                                            nmigrate_disks, migrate_disks,
4249
                                            migParams);
4250
    } else {
4251 4252 4253 4254
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4255
                                            nmigrate_disks, migrate_disks,
4256
                                            migParams);
4257
    }
4258 4259

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4260
    if (ret < 0) {
4261
        orig_err = virSaveLastError();
4262 4263 4264 4265
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4266 4267 4268 4269

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4270
    cancelled = ret < 0;
4271

4272
 finish:
4273 4274 4275 4276 4277 4278 4279 4280
    /*
     * 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);
4281
    VIR_STEAL_PTR(cookiein, cookieout);
4282 4283
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296

    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);
4297 4298
            if (qemuDomainObjExitRemote(vm, !offline) < 0)
                goto cleanup;
4299 4300 4301 4302 4303 4304 4305
        }
    } else {
        dname = dname ? dname : vm->def->name;
        qemuDomainObjEnterRemote(vm);
        ddomain = dconn->driver->domainMigrateFinish3
            (dconn, dname, cookiein, cookieinlen, &cookieout, &cookieoutlen,
             dconnuri, uri, destflags, cancelled);
4306 4307
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4308
    }
4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328

    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();
4329 4330
                if (err &&
                    err->domain == VIR_FROM_QEMU &&
4331 4332 4333 4334 4335 4336 4337
                    err->code != VIR_ERR_MIGRATE_FINISH_OK) {
                    virFreeError(orig_err);
                    orig_err = NULL;
                }
            }
        }
    }
4338

4339 4340 4341 4342 4343 4344 4345
    /* 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.
4346
     */
4347
    cancelled = ddomain == NULL;
4348

4349 4350 4351 4352 4353 4354
    /* 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();

4355 4356 4357 4358
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4359
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4360
    VIR_FREE(cookiein);
4361
    VIR_STEAL_PTR(cookiein, cookieout);
4362 4363
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
4364 4365 4366
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
4367 4368 4369 4370
    /* 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.
     */
4371 4372 4373
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4374 4375 4376

 cleanup:
    if (ddomain) {
4377
        virObjectUnref(ddomain);
4378 4379 4380 4381 4382
        ret = 0;
    } else {
        ret = -1;
    }

4383
    virObjectUnref(st);
4384 4385 4386 4387 4388 4389 4390 4391

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4392
    virTypedParamsFree(params, nparams);
4393 4394 4395 4396
    return ret;
}


4397
static void
4398 4399 4400
qemuMigrationSrcConnectionClosed(virConnectPtr conn,
                                 int reason,
                                 void *opaque)
4401 4402 4403 4404 4405 4406 4407 4408
{
    virDomainObjPtr vm = opaque;

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


4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433
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,
4434
                                 qemuMigrationParamsPtr migParams,
4435 4436 4437 4438
                                 unsigned long flags,
                                 const char *dname,
                                 unsigned long resource,
                                 bool *v3proto)
4439 4440 4441 4442
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4443
    virErrorPtr orig_err = NULL;
4444
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4445
    bool dstOffline = false;
4446
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4447
    bool useParams;
4448

4449 4450
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, uri=%s, "
              "graphicsuri=%s, listenAddress=%s, nmigrate_disks=%zu, "
4451
              "migrate_disks=%p, nbdPort=%d, flags=0x%lx, dname=%s, "
4452
              "resource=%lu",
4453
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4454
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4455 4456
              nmigrate_disks, migrate_disks, nbdPort, flags, NULLSTR(dname),
              resource);
4457

4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471
    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;
    }

4472 4473 4474 4475 4476 4477 4478
    if (flags & VIR_MIGRATE_TUNNELLED && nbdPort) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("disk port address is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

4479 4480 4481 4482
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

4483
    qemuDomainObjEnterRemote(vm);
4484
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4485 4486 4487
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;

4488
    if (dconn == NULL) {
4489
        virReportError(VIR_ERR_OPERATION_FAILED,
4490 4491
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4492
        virObjectUnref(cfg);
4493 4494 4495
        return -1;
    }

4496 4497
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4498 4499
        goto cleanup;

4500
    if (virConnectRegisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed,
4501 4502 4503 4504
                                        vm, NULL) < 0) {
        goto cleanup;
    }

4505
    qemuDomainObjEnterRemote(vm);
4506 4507
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4508
        /* v3proto reflects whether the caller used Perform3, but with
4509
         * p2p migrate, regardless of whether Perform2 or Perform3
4510 4511 4512 4513
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4514 4515
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
4516
    if (offline)
4517 4518
        dstOffline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                              VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4519 4520
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;
4521

4522
    if (!p2p) {
4523 4524
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4525 4526 4527
        goto cleanup;
    }

4528 4529
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4530
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
4531 4532 4533 4534 4535 4536
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("Migration APIs with extensible parameters are not "
                         "supported but extended parameters were passed"));
        goto cleanup;
    }

4537
    if (offline && !dstOffline) {
L
liguang 已提交
4538 4539 4540 4541 4542 4543
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4544 4545 4546 4547 4548 4549 4550
    /* 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;

4551
    if (*v3proto) {
4552 4553 4554
        ret = qemuMigrationSrcPerformPeer2Peer3(driver, sconn, dconn, dconnuri, vm, xmlin,
                                                persist_xml, dname, uri, graphicsuri,
                                                listenAddress, nmigrate_disks, migrate_disks,
4555
                                                nbdPort, migParams, resource,
4556
                                                useParams, flags);
4557
    } else {
4558
        ret = qemuMigrationSrcPerformPeer2Peer2(driver, sconn, dconn, vm,
4559
                                                dconnuri, flags, dname, resource,
4560
                                                migParams);
4561
    }
4562

4563
 cleanup:
4564
    orig_err = virSaveLastError();
4565
    qemuDomainObjEnterRemote(vm);
4566
    virConnectUnregisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed);
4567
    virObjectUnref(dconn);
4568
    ignore_value(qemuDomainObjExitRemote(vm, false));
4569 4570 4571 4572
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4573
    virObjectUnref(cfg);
4574 4575 4576 4577
    return ret;
}


4578 4579 4580 4581 4582 4583
/*
 * 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
4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595
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,
4596
                           qemuMigrationParamsPtr migParams,
4597 4598 4599 4600 4601 4602 4603 4604
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
                           const char *dname,
                           unsigned long resource,
                           bool v3proto)
4605
{
4606
    virObjectEventPtr event = NULL;
4607
    int ret = -1;
4608
    virErrorPtr orig_err = NULL;
4609
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4610
    qemuDomainObjPrivatePtr priv = vm->privateData;
4611

4612 4613
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                              flags) < 0)
4614 4615
        goto cleanup;

4616
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
4617 4618
        goto endjob;

4619
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
4620
        goto endjob;
4621

4622
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
4623
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
4624
        goto endjob;
4625

4626
    qemuMigrationSrcStoreDomainState(vm);
4627 4628

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4629 4630 4631
        ret = qemuMigrationSrcPerformPeer2Peer(driver, conn, vm, xmlin, persist_xml,
                                               dconnuri, uri, graphicsuri, listenAddress,
                                               nmigrate_disks, migrate_disks, nbdPort,
4632
                                               migParams, flags, dname, resource,
4633
                                               &v3proto);
4634
    } else {
4635
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4636 4637 4638
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                            cookieout, cookieoutlen,
                                            flags, resource, NULL, NULL, 0, NULL,
4639
                                            migParams);
4640
    }
4641 4642
    if (ret < 0)
        goto endjob;
4643

4644 4645 4646 4647
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4648
    if (!v3proto) {
4649
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
4650
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4651
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4652
        virDomainAuditStop(vm, "migrated");
4653
        event = virDomainEventLifecycleNewFromObj(vm,
4654 4655
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4656 4657
    }

4658
 endjob:
4659 4660 4661
    if (ret < 0)
        orig_err = virSaveLastError();

4662 4663 4664 4665
    /* v2 proto has no confirm phase so we need to reset migration parameters
     * here
     */
    if (!v3proto && ret < 0)
4666
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4667
                                 priv->job.migParams, priv->job.apiFlags);
4668

4669
    qemuMigrationSrcRestoreDomainState(driver, vm);
4670

4671
    qemuMigrationJobFinish(driver, vm);
4672
    if (!virDomainObjIsActive(vm) && ret == 0) {
4673
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
4674
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4675 4676
            vm->persistent = 0;
        }
4677
        qemuDomainRemoveInactiveJob(driver, vm);
4678 4679
    }

4680 4681 4682 4683 4684
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4685
 cleanup:
M
Michal Privoznik 已提交
4686
    virDomainObjEndAPI(&vm);
4687
    virObjectEventStateQueue(driver->domainEventState, event);
4688
    virObjectUnref(cfg);
4689 4690 4691 4692 4693 4694 4695
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4696 4697 4698 4699 4700 4701 4702 4703
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,
4704
                             qemuMigrationParamsPtr migParams,
4705 4706 4707 4708 4709 4710
                             const char *cookiein,
                             int cookieinlen,
                             char **cookieout,
                             int *cookieoutlen,
                             unsigned long flags,
                             unsigned long resource)
4711
{
4712
    qemuDomainObjPrivatePtr priv = vm->privateData;
4713 4714 4715 4716
    int ret = -1;

    /* If we didn't start the job in the begin phase, start it now. */
    if (!(flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
4717 4718
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
4719 4720 4721 4722 4723 4724
            goto cleanup;
    } else if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) {
        goto cleanup;
    }

    qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4725
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
4726
                           qemuMigrationSrcCleanup);
4727

4728 4729 4730
    ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                        cookieout, cookieoutlen,
                                        flags, resource, NULL, graphicsuri,
4731
                                        nmigrate_disks, migrate_disks, migParams);
4732

4733
    if (ret < 0) {
4734
        qemuMigrationSrcRestoreDomainState(driver, vm);
4735
        goto endjob;
4736
    }
4737 4738 4739

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4740
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
4741
                             qemuMigrationSrcCleanup) < 0)
4742 4743
        goto endjob;

4744
 endjob:
4745
    if (ret < 0) {
4746
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4747
                                 priv->job.migParams, priv->job.apiFlags);
4748
        qemuMigrationJobFinish(driver, vm);
4749
    } else {
4750
        qemuMigrationJobContinue(vm);
4751 4752
    }

4753
    if (!virDomainObjIsActive(vm))
4754
        qemuDomainRemoveInactiveJob(driver, vm);
4755

4756
 cleanup:
M
Michal Privoznik 已提交
4757
    virDomainObjEndAPI(&vm);
4758 4759 4760
    return ret;
}

4761
int
4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773
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,
4774
                        qemuMigrationParamsPtr migParams,
4775 4776 4777 4778 4779 4780 4781 4782
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4783 4784
{
    VIR_DEBUG("driver=%p, conn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4785
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
4786
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4787
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
4788
              "flags=0x%lx, dname=%s, resource=%lu, v3proto=%d",
4789
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4790
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4791 4792 4793
              nmigrate_disks, migrate_disks, nbdPort,
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4794 4795 4796

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4797 4798
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4799 4800 4801
            return -1;
        }

4802 4803 4804
        return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, dconnuri, uri,
                                          graphicsuri, listenAddress,
                                          nmigrate_disks, migrate_disks, nbdPort,
4805
                                          migParams,
4806 4807 4808
                                          cookiein, cookieinlen,
                                          cookieout, cookieoutlen,
                                          flags, dname, resource, v3proto);
4809 4810
    } else {
        if (dconnuri) {
4811 4812
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
4813 4814 4815 4816
            return -1;
        }

        if (v3proto) {
4817 4818 4819
            return qemuMigrationSrcPerformPhase(driver, conn, vm, persist_xml, uri,
                                                graphicsuri,
                                                nmigrate_disks, migrate_disks,
4820
                                                migParams,
4821 4822 4823
                                                cookiein, cookieinlen,
                                                cookieout, cookieoutlen,
                                                flags, resource);
4824
        } else {
4825 4826 4827
            return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, NULL,
                                              uri, graphicsuri, listenAddress,
                                              nmigrate_disks, migrate_disks, nbdPort,
4828
                                              migParams,
4829 4830 4831
                                              cookiein, cookieinlen,
                                              cookieout, cookieoutlen, flags,
                                              dname, resource, v3proto);
4832 4833 4834
        }
    }
}
4835

4836
static int
4837
qemuMigrationDstVPAssociatePortProfiles(virDomainDefPtr def)
4838
{
4839
    size_t i;
4840 4841 4842 4843 4844
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
4845
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4846
            if (virNetDevVPortProfileAssociate(net->ifname,
4847
                                               virDomainNetGetActualVirtPortProfile(net),
4848
                                               &net->mac,
4849
                                               virDomainNetGetActualDirectDev(net),
4850
                                               -1,
4851
                                               def->uuid,
4852 4853
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
4854 4855 4856
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
4857
                goto err_exit;
4858
            }
4859
            last_good_net = i;
4860
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
4861

4862
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
4863 4864 4865 4866
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
4867 4868 4869
        }
    }

4870
    return 0;
4871

4872
 err_exit:
4873
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
4874
        net = def->nets[i];
4875
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4876
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
4877
                                                           virDomainNetGetActualVirtPortProfile(net),
4878
                                                           &net->mac,
4879
                                                           virDomainNetGetActualDirectDev(net),
4880
                                                           -1,
4881
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
4882 4883
        }
    }
4884
    return -1;
4885 4886 4887
}


J
Jiri Denemark 已提交
4888
static int
4889 4890 4891 4892
qemuMigrationDstPersist(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationCookiePtr mig,
                        bool ignoreSaveError)
J
Jiri Denemark 已提交
4893 4894 4895 4896
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    virDomainDefPtr vmdef;
4897 4898
    virDomainDefPtr oldDef = NULL;
    unsigned int oldPersist = vm->persistent;
J
Jiri Denemark 已提交
4899 4900 4901 4902 4903 4904 4905
    virObjectEventPtr event;
    int ret = -1;

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

    vm->persistent = 1;
4906 4907
    oldDef = vm->newDef;
    vm->newDef = qemuMigrationCookieGetPersistent(mig);
J
Jiri Denemark 已提交
4908

4909 4910
    if (!(vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm)))
        goto error;
J
Jiri Denemark 已提交
4911

4912 4913
    if (virDomainSaveConfig(cfg->configDir, driver->caps, vmdef) < 0 &&
        !ignoreSaveError)
4914
        goto error;
J
Jiri Denemark 已提交
4915 4916 4917

    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_DEFINED,
4918 4919 4920
                                              oldPersist ?
                                              VIR_DOMAIN_EVENT_DEFINED_UPDATED :
                                              VIR_DOMAIN_EVENT_DEFINED_ADDED);
4921
    virObjectEventStateQueue(driver->domainEventState, event);
J
Jiri Denemark 已提交
4922 4923 4924 4925

    ret = 0;

 cleanup:
4926
    virDomainDefFree(oldDef);
J
Jiri Denemark 已提交
4927 4928 4929
    virObjectUnref(caps);
    virObjectUnref(cfg);
    return ret;
4930 4931 4932 4933 4934 4935 4936

 error:
    virDomainDefFree(vm->newDef);
    vm->persistent = oldPersist;
    vm->newDef = oldDef;
    oldDef = NULL;
    goto cleanup;
J
Jiri Denemark 已提交
4937 4938 4939
}


4940
virDomainPtr
4941 4942 4943 4944 4945 4946 4947 4948 4949 4950
qemuMigrationDstFinish(virQEMUDriverPtr driver,
                       virConnectPtr dconn,
                       virDomainObjPtr vm,
                       const char *cookiein,
                       int cookieinlen,
                       char **cookieout,
                       int *cookieoutlen,
                       unsigned long flags,
                       int retcode,
                       bool v3proto)
4951 4952
{
    virDomainPtr dom = NULL;
4953
    qemuMigrationCookiePtr mig = NULL;
4954
    virErrorPtr orig_err = NULL;
4955
    int cookie_flags = 0;
J
Jiri Denemark 已提交
4956
    qemuDomainObjPrivatePtr priv = vm->privateData;
4957
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4958
    unsigned short port;
4959 4960
    unsigned long long timeReceived = 0;
    virObjectEventPtr event;
4961
    qemuDomainJobInfoPtr jobInfo = NULL;
4962
    bool inPostCopy = false;
4963
    bool doKill = true;
4964

4965
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
4966
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, retcode=%d",
4967 4968
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
4969

4970 4971 4972
    port = priv->migrationPort;
    priv->migrationPort = 0;

4973
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN)) {
4974
        qemuMigrationDstErrorReport(driver, vm->def->name);
4975
        goto cleanup;
4976
    }
4977

4978 4979
    ignore_value(virTimeMillisNow(&timeReceived));

4980 4981 4982
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
4983

4984
    qemuDomainCleanupRemove(vm, qemuMigrationDstPrepareCleanup);
4985
    VIR_FREE(priv->job.completed);
4986

4987
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
4988 4989
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
4990 4991 4992
    /* 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;
4993

4994 4995
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen, cookie_flags)))
4996
        goto endjob;
4997

4998
    if (flags & VIR_MIGRATE_OFFLINE) {
4999
        if (retcode == 0 &&
5000
            qemuMigrationDstPersist(driver, vm, mig, false) == 0)
5001
            dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, -1);
5002 5003
        goto endjob;
    }
5004

5005 5006 5007 5008
    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
         */
5009
        qemuDomainCheckMonitor(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN);
5010 5011
        goto endjob;
    }
5012

5013 5014 5015
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
5016
        qemuMigrationDstErrorReport(driver, vm->def->name);
5017 5018
        goto endjob;
    }
5019

5020
    if (qemuMigrationDstVPAssociatePortProfiles(vm->def) < 0)
5021
        goto endjob;
5022

5023
    if (mig->network && qemuMigrationDstOPDRelocate(driver, vm, mig) < 0)
5024
        VIR_WARN("unable to provide network data for relocation");
5025

5026
    if (qemuMigrationDstStopNBDServer(driver, vm, mig) < 0)
5027
        goto endjob;
5028

5029 5030
    if (qemuRefreshVirtioChannelState(driver, vm,
                                      QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
5031 5032
        goto endjob;

5033 5034
    if (qemuConnectAgent(driver, vm) < 0)
        goto endjob;
5035

5036
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
5037
        if (qemuMigrationDstPersist(driver, vm, mig, !v3proto) < 0) {
5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048
            /* 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.
5049 5050 5051
             */
            if (v3proto)
                goto endjob;
5052
        }
5053
    }
5054

5055 5056 5057
    /* We need to wait for QEMU to process all data sent by the source
     * before starting guest CPUs.
     */
5058 5059 5060
    if (qemuMigrationDstWaitForCompletion(driver, vm,
                                          QEMU_ASYNC_JOB_MIGRATION_IN,
                                          !!(flags & VIR_MIGRATE_POSTCOPY)) < 0) {
5061 5062 5063 5064 5065 5066
        /* 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;
    }
5067

5068 5069 5070 5071 5072 5073 5074 5075 5076
    /* 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;
    }

5077
    if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY)
5078 5079
        inPostCopy = true;

5080 5081 5082 5083 5084
    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
         */
5085
        if (qemuProcessStartCPUs(driver, vm,
5086 5087
                                 inPostCopy ? VIR_DOMAIN_RUNNING_POSTCOPY
                                            : VIR_DOMAIN_RUNNING_MIGRATED,
5088
                                 QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5089
            if (virGetLastErrorCode() == VIR_ERR_OK)
5090 5091 5092 5093 5094 5095
                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();
5096

5097 5098 5099 5100 5101 5102 5103 5104 5105 5106
            /*
             * 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;
5107
        }
5108

5109
        if (inPostCopy)
5110
            doKill = false;
5111
    }
5112

5113
    if (mig->jobInfo) {
5114
        jobInfo = mig->jobInfo;
5115 5116 5117 5118 5119 5120
        mig->jobInfo = NULL;

        if (jobInfo->sent && timeReceived) {
            jobInfo->timeDelta = timeReceived - jobInfo->sent;
            jobInfo->received = timeReceived;
            jobInfo->timeDeltaSet = true;
5121
        }
5122 5123
        qemuDomainJobInfoUpdateTime(jobInfo);
        qemuDomainJobInfoUpdateDowntime(jobInfo);
5124
    }
L
liguang 已提交
5125

5126
    if (inPostCopy) {
5127 5128 5129
        if (qemuMigrationDstWaitForCompletion(driver, vm,
                                              QEMU_ASYNC_JOB_MIGRATION_IN,
                                              false) < 0) {
5130 5131 5132 5133 5134 5135 5136 5137 5138
            goto endjob;
        }
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            virDomainObjSetState(vm,
                                 VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_MIGRATED);
        }
    }

5139
    dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, vm->def->id);
5140

5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154
    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);
    }
5155

5156 5157 5158 5159 5160
    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);
5161
        virObjectEventStateQueue(driver->domainEventState, event);
5162
    }
5163

5164
    if (virDomainObjIsActive(vm) &&
5165
        virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
5166 5167 5168 5169 5170
        VIR_WARN("Failed to save status on vm %s", vm->def->name);

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

5171
 endjob:
5172
    if (!dom &&
5173 5174
        !(flags & VIR_MIGRATE_OFFLINE) &&
        virDomainObjIsActive(vm)) {
5175
        if (doKill) {
5176 5177 5178 5179 5180 5181 5182
            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);
5183
            virObjectEventStateQueue(driver->domainEventState, event);
5184
        } else {
5185
            qemuMigrationAnyPostcopyFailed(driver, vm);
5186
        }
5187 5188
    }

5189
    if (dom) {
5190 5191 5192
        if (jobInfo) {
            VIR_STEAL_PTR(priv->job.completed, jobInfo);
            priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
5193
            priv->job.completed->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;
5194
        }
5195

5196 5197 5198
        if (qemuMigrationBakeCookie(mig, driver, vm,
                                    QEMU_MIGRATION_DESTINATION,
                                    cookieout, cookieoutlen,
5199 5200
                                    QEMU_MIGRATION_COOKIE_STATS) < 0)
            VIR_WARN("Unable to encode migration cookie");
5201 5202 5203 5204 5205 5206

        /* Remove completed stats for post-copy, everything but timing fields
         * is obsolete anyway.
         */
        if (inPostCopy)
            VIR_FREE(priv->job.completed);
5207
    }
5208

5209
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
5210
                             priv->job.migParams, priv->job.apiFlags);
5211

5212
    qemuMigrationJobFinish(driver, vm);
5213
    if (!virDomainObjIsActive(vm))
5214
        qemuDomainRemoveInactiveJob(driver, vm);
5215

5216
 cleanup:
5217
    VIR_FREE(jobInfo);
5218
    virPortAllocatorRelease(port);
5219
    if (priv->mon)
5220
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
5221
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5222
    virDomainObjEndAPI(&vm);
5223
    qemuMigrationCookieFree(mig);
5224 5225 5226 5227
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5228
    virObjectUnref(cfg);
5229 5230 5231 5232

    /* Set a special error if Finish is expected to return NULL as a result of
     * successful call with retcode != 0
     */
5233
    if (retcode != 0 && !dom && virGetLastErrorCode() == VIR_ERR_OK)
5234
        virReportError(VIR_ERR_MIGRATE_FINISH_OK, NULL);
5235 5236
    return dom;
}
5237

5238

5239
/* Helper function called while vm is active.  */
5240
int
5241 5242 5243 5244
qemuMigrationSrcToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
                       int fd,
                       const char *compressor,
                       qemuDomainAsyncJob asyncJob)
5245 5246 5247
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5248
    int ret = -1;
5249 5250
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5251
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5252
    char *errbuf = NULL;
5253
    virErrorPtr orig_err = NULL;
5254 5255 5256 5257 5258

    /* 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,
5259 5260
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5261 5262
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5263
    }
5264

5265 5266 5267 5268 5269 5270 5271
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5272 5273 5274 5275
    if (compressor && pipe(pipeFD) < 0) {
        virReportSystemError(errno, "%s",
                             _("Failed to create pipe for migration"));
        return -1;
5276 5277
    }

5278 5279 5280 5281
    /* 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.  */
5282 5283
    if (qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    compressor ? pipeFD[1] : fd) < 0)
5284 5285
        goto cleanup;

5286
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5287 5288
        goto cleanup;

5289
    if (!compressor) {
5290 5291 5292
        rc = qemuMonitorMigrateToFd(priv->mon,
                                    QEMU_MONITOR_MIGRATE_BACKGROUND,
                                    fd);
5293 5294 5295 5296 5297 5298 5299
    } else {
        const char *prog = compressor;
        const char *args[] = {
            prog,
            "-c",
            NULL
        };
5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310

        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;
5311
        }
5312 5313 5314 5315 5316 5317 5318 5319 5320 5321
        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");
5322
    }
5323
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5324
        goto cleanup;
5325 5326 5327
    if (rc < 0)
        goto cleanup;

5328
    rc = qemuMigrationSrcWaitForCompletion(driver, vm, asyncJob, NULL, 0);
5329

5330 5331 5332 5333
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5334 5335
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5336
                qemuMonitorMigrateCancel(priv->mon);
5337
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5338 5339
            }
        }
5340
        goto cleanup;
5341
    }
5342

5343 5344 5345
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5346
    qemuDomainEventEmitJobCompleted(driver, vm);
5347 5348
    ret = 0;

5349
 cleanup:
5350 5351 5352
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5353
    /* Restore max migration bandwidth */
5354 5355
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5356 5357
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5358
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5359 5360
    }

5361 5362
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5363 5364 5365 5366 5367
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5368 5369 5370 5371 5372 5373

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

5374 5375
    return ret;
}
5376

5377 5378

int
5379 5380
qemuMigrationSrcCancel(virQEMUDriverPtr driver,
                       virDomainObjPtr vm)
5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391
{
    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));
5392
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5393 5394 5395 5396 5397
        goto cleanup;

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
5398
        qemuBlockJobDataPtr job;
5399

5400
        if (!(job = qemuBlockJobDiskGetJob(disk)) ||
5401
            !qemuBlockJobIsRunning(job))
5402
            diskPriv->migrating = false;
5403 5404

        if (diskPriv->migrating) {
5405
            qemuBlockJobSyncBegin(job);
5406
            storage = true;
5407 5408
        }

5409 5410
        virObjectUnref(job);
    }
5411 5412 5413

    if (storage &&
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
5414
                                      QEMU_ASYNC_JOB_NONE, NULL) < 0)
5415
        goto cleanup;
5416 5417 5418 5419 5420 5421 5422 5423

    ret = 0;

 cleanup:
    return ret;
}


5424
static int
5425
qemuMigrationJobStart(virQEMUDriverPtr driver,
5426
                      virDomainObjPtr vm,
5427 5428
                      qemuDomainAsyncJob job,
                      unsigned long apiFlags)
5429
{
5430
    qemuDomainObjPrivatePtr priv = vm->privateData;
5431 5432
    virDomainJobOperation op;
    unsigned long long mask;
5433

5434
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5435 5436
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_IN;
        mask = QEMU_JOB_NONE;
5437
    } else {
5438 5439 5440 5441
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_OUT;
        mask = QEMU_JOB_DEFAULT_MASK |
               JOB_MASK(QEMU_JOB_SUSPEND) |
               JOB_MASK(QEMU_JOB_MIGRATION_OP);
5442
    }
5443

5444
    if (qemuDomainObjBeginAsyncJob(driver, vm, job, op, apiFlags) < 0)
5445 5446
        return -1;

5447 5448
    priv->job.current->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;

5449
    qemuDomainObjSetAsyncJobMask(vm, mask);
5450 5451 5452
    return 0;
}

5453
static void
5454
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5455
                         virDomainObjPtr vm,
5456
                         qemuMigrationJobPhase phase)
5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469
{
    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);
}

5470
static void
5471
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5472
                           virDomainObjPtr vm,
5473
                           qemuMigrationJobPhase phase)
5474 5475 5476 5477
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5478
static void
5479 5480
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5481
    qemuDomainObjReleaseAsyncJob(vm);
5482 5483
}

5484
static bool
5485
qemuMigrationJobIsActive(virDomainObjPtr vm,
5486
                         qemuDomainAsyncJob job)
5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497
{
    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");

5498
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5499 5500 5501 5502 5503
        return false;
    }
    return true;
}

5504
static void
5505
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5506
{
5507
    qemuDomainObjEndAsyncJob(driver, vm);
5508
}
5509 5510 5511


static void
5512
qemuMigrationDstErrorFree(void *data,
5513 5514 5515 5516 5517 5518 5519
                       const void *name ATTRIBUTE_UNUSED)
{
    virErrorPtr err = data;
    virFreeError(err);
}

int
5520
qemuMigrationDstErrorInit(virQEMUDriverPtr driver)
5521
{
5522
    driver->migrationErrors = virHashAtomicNew(64, qemuMigrationDstErrorFree);
5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533
    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
5534 5535 5536
qemuMigrationDstErrorSave(virQEMUDriverPtr driver,
                          const char *name,
                          virErrorPtr err)
5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549
{
    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
5550 5551
qemuMigrationDstErrorReport(virQEMUDriverPtr driver,
                            const char *name)
5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562
{
    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);
}
5563 5564


5565
int
5566 5567 5568 5569
qemuMigrationSrcFetchMirrorStats(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 qemuDomainJobInfoPtr jobInfo)
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 5613
{
    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;
}