qemu_migration.c 188.0 KB
Newer Older
1 2 3
/*
 * qemu_migration.c: QEMU migration handling
 *
4
 * Copyright (C) 2006-2015 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23
 *
 */

#include <config.h>

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

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

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

#define VIR_FROM_THIS VIR_FROM_QEMU

66 67
VIR_LOG_INIT("qemu.qemu_migration");

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

167 168

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

294 295

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

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

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

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

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

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

        diskSrcPath = virDomainDiskGetSource(disk);

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

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

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

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

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


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

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

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

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

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

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

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

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

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

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

    priv->nbdPort = port;
    ret = 0;

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

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

440 441

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

    if (!mig->nbd)
        return 0;

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

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

460
    virPortAllocatorRelease(priv->nbdPort);
461 462 463 464 465 466
    priv->nbdPort = 0;
    return 0;
}


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

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

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

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

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

512 513 514 515 516 517 518
    if (notReady) {
        VIR_DEBUG("Waiting for %zu disk mirrors to get ready", notReady);
        return 0;
    } else {
        VIR_DEBUG("All disk mirrors are ready");
        return 1;
    }
519 520 521
}


522 523 524 525
/*
 * If @check is true, the function will report an error and return a different
 * code in case a block job fails. This way we can properly abort migration in
 * case some block jobs failed once all memory has already been transferred.
526
 *
527 528 529 530
 * Returns 1 if all mirrors are gone,
 *         0 if some mirrors are still active,
 *         -1 some mirrors failed but some are still active,
 *         -2 all mirrors are gone but some of them failed.
531 532
 */
static int
533 534 535
qemuMigrationSrcNBDCopyCancelled(virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 bool check)
536
{
537 538
    size_t i;
    size_t active = 0;
539
    size_t completed = 0;
540 541
    int status;
    bool failed = false;
542

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

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

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

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

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

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

    /* Updating completed block job drops the lock thus we have to recheck
     * block jobs for disks that reside before the disk(s) with completed
     * block job.
     */
    if (completed > 0) {
        completed = 0;
        active = 0;
        goto retry;
591
    }
592

593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608
    if (failed) {
        if (active) {
            VIR_DEBUG("Some disk mirrors failed; still waiting for %zu "
                      "disk mirrors to finish", active);
            return -1;
        } else {
            VIR_DEBUG("All disk mirrors are gone; some of them failed");
            return -2;
        }
    } else {
        if (active) {
            VIR_DEBUG("Waiting for %zu disk mirrors to finish", active);
            return 0;
        } else {
            VIR_DEBUG("All disk mirrors are gone");
            return 1;
609
        }
610
    }
611
}
612 613


614 615 616 617 618 619
/*
 * Returns 0 on success,
 *         1 when job is already completed or it failed and failNoJob is false,
 *         -1 on error or when job failed and failNoJob is true.
 */
static int
620 621 622 623 624
qemuMigrationSrcNBDCopyCancelOne(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 virDomainDiskDefPtr disk,
                                 bool failNoJob,
                                 qemuDomainAsyncJob asyncJob)
625 626 627
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char *diskAlias = NULL;
628
    char *error = NULL;
629 630 631 632
    int ret = -1;
    int status;
    int rv;

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

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

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

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

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

    ret = 0;
666 667 668

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


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

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

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

706
        if (!diskPriv->blockjob->started)
707 708
            diskPriv->migrating = false;

709
        if (!diskPriv->migrating)
710 711
            continue;

712 713
        rv = qemuMigrationSrcNBDCopyCancelOne(driver, vm, disk,
                                              check, asyncJob);
714 715 716 717 718 719
        if (rv != 0) {
            if (rv < 0) {
                if (!err)
                    err = virSaveLastError();
                failed = true;
            }
720
            qemuBlockJobSyncEndDisk(vm, asyncJob, disk);
721 722 723 724
            diskPriv->migrating = false;
        }
    }

725
    while ((rv = qemuMigrationSrcNBDCopyCancelled(vm, asyncJob, check)) != 1) {
726 727 728 729 730 731 732
        if (check && !failed &&
            dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            failed = true;
        }

733 734 735 736
        if (rv < 0) {
            failed = true;
            if (rv == -2)
                break;
737
        }
738

739 740 741 742 743
        if (failed && !err)
            err = virSaveLastError();

        if (virDomainObjWait(vm) < 0)
            goto cleanup;
744 745
    }

746 747 748 749 750 751 752 753 754 755 756 757 758
    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;
    }

759 760 761
    ret = failed ? -1 : 0;

 cleanup:
762 763 764 765 766
    if (err) {
        virSetError(err);
        virFreeError(err);
    }
    return ret;
767 768 769
}


770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848
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;
}


849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890
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;
}


891 892 893 894 895 896 897 898 899 900 901 902
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);
903
    qemuBlockJobDataPtr job = NULL;
904 905 906 907 908 909 910
    char *diskAlias = NULL;
    int rc;
    int ret = -1;

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

911
    if (!(job = qemuBlockJobDiskNew(disk, QEMU_BLOCKJOB_TYPE_COPY)))
912 913
        goto cleanup;

914
    qemuBlockJobSyncBegin(job);
915 916 917 918 919 920 921 922 923 924 925 926 927 928 929

    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);
    }

930
    if (rc < 0)
931 932 933
        goto cleanup;

    diskPriv->migrating = true;
934
    qemuBlockJobStarted(job);
935 936 937 938

    ret = 0;

 cleanup:
939
    qemuBlockJobStartupFinalize(job);
940 941 942 943 944
    VIR_FREE(diskAlias);
    return ret;
}


945
/**
946
 * qemuMigrationSrcNBDStorageCopy:
947 948 949 950
 * @driver: qemu driver
 * @vm: domain
 * @mig: migration cookie
 * @host: where are we migrating to
951
 * @speed: bandwidth limit in MiB/s
952 953
 * @migrate_flags: migrate monitor command flags
 *
954 955 956
 * 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
957
 * to do the very same operation. On failure, the caller is
958 959
 * expected to call qemuMigrationSrcNBDCopyCancel to stop all
 * running copy operations.
960 961 962 963 964
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
965 966 967 968 969 970 971 972
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,
973 974 975
                               virConnectPtr dconn,
                               const char *tlsAlias,
                               unsigned int flags)
976 977 978 979
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int port;
980
    size_t i;
981
    unsigned long long mirror_speed = speed;
982
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
983
    int rv;
984
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
985 986

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
987

988 989 990 991 992 993 994 995
    if (mirror_speed > LLONG_MAX >> 20) {
        virReportError(VIR_ERR_OVERFLOW,
                       _("bandwidth must be less than %llu"),
                       LLONG_MAX >> 20);
        goto cleanup;
    }
    mirror_speed <<= 20;

996 997 998 999 1000 1001 1002 1003 1004 1005
    /* 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];

1006
        /* check whether disk should be migrated */
1007
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1008 1009
            continue;

1010 1011 1012
        if (qemuMigrationSrcNBDStorageCopyOne(driver, vm, disk, host, port,
                                              mirror_speed, mirror_flags,
                                              tlsAlias, flags) < 0)
1013 1014
            goto cleanup;

1015
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1016 1017 1018
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto cleanup;
        }
1019
    }
1020

1021
    while ((rv = qemuMigrationSrcNBDStorageCopyReady(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) != 1) {
1022 1023
        if (rv < 0)
            goto cleanup;
1024

1025
        if (priv->job.abortJob) {
1026
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1027 1028 1029 1030
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
            goto cleanup;
1031
        }
1032

1033 1034 1035 1036 1037 1038
        if (dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            goto cleanup;
        }

1039
        if (virDomainObjWait(vm) < 0)
1040
            goto cleanup;
1041 1042
    }

1043 1044
    qemuMigrationSrcFetchMirrorStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                     priv->job.current);
1045

1046
    /* Okay, all disks are ready. Modify migrate_flags */
1047 1048 1049 1050
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

1051
 cleanup:
1052
    virObjectUnref(cfg);
1053 1054
    return ret;
}
1055

1056

1057
/**
1058
 * qemuMigrationSrcIsAllowedHostdev:
1059 1060
 * @def: domain definition
 *
1061
 * Checks that @def does not contain any host devices unsupported across
1062 1063 1064
 * migrations. Returns true if the vm is allowed to migrate.
 */
static bool
1065
qemuMigrationSrcIsAllowedHostdev(const virDomainDef *def)
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
{
    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;
}


1085
/**
1086
 * qemuMigrationSrcIsAllowed:
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098
 * @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.
 */
1099
bool
1100 1101 1102 1103
qemuMigrationSrcIsAllowed(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          bool remote,
                          unsigned int flags)
1104
{
1105
    int nsnapshots;
1106
    int pauseReason;
1107
    size_t i;
1108

1109 1110 1111 1112 1113
    /* perform these checks only when migrating to remote hosts */
    if (remote) {
        nsnapshots = virDomainSnapshotObjListNum(vm->snapshots, NULL, 0);
        if (nsnapshots < 0)
            return false;
1114

1115 1116 1117 1118 1119
        if (nsnapshots > 0) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("cannot migrate domain with %d snapshots"),
                           nsnapshots);
            return false;
1120
        }
1121

1122
        /* cancel migration if disk I/O error is emitted while migrating */
1123
        if (flags & VIR_MIGRATE_ABORT_ON_ERROR &&
1124
            !(flags & VIR_MIGRATE_OFFLINE) &&
1125 1126
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
E
Eric Blake 已提交
1127
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1128
                           _("cannot migrate domain with I/O error"));
E
Eric Blake 已提交
1129 1130
            return false;
        }
1131 1132

    }
1133

1134 1135 1136 1137 1138 1139 1140
    /* 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;
        }
1141

1142

1143 1144
        if (qemuDomainHasBlockjob(vm, false)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1145
                           _("domain has active block job"));
1146 1147 1148
            return false;
        }

1149
        if (!qemuMigrationSrcIsAllowedHostdev(vm->def))
1150 1151
            return false;

1152
        if (vm->def->cpu) {
1153 1154 1155
            /* QEMU blocks migration and save with invariant TSC enabled
             * unless TSC frequency is explicitly set.
             */
1156 1157
            if (virCPUCheckFeature(vm->def->os.arch, vm->def->cpu,
                                   "invtsc") == 1) {
1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175
                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;
                }
1176
            }
J
Ján Tomko 已提交
1177 1178
        }

1179 1180 1181
        /* Verify that memory device config can be transferred reliably */
        for (i = 0; i < vm->def->nmems; i++) {
            virDomainMemoryDefPtr mem = vm->def->mems[i];
1182

1183 1184 1185 1186 1187
            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"));
1188

1189 1190
                return false;
            }
1191
        }
1192 1193 1194 1195 1196 1197

        if (vm->def->nshmems) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("migration with shmem device is not supported"));
            return false;
        }
1198 1199
    }

1200 1201 1202
    return true;
}

1203
static bool
1204 1205 1206 1207
qemuMigrationSrcIsSafe(virDomainDefPtr def,
                       size_t nmigrate_disks,
                       const char **migrate_disks,
                       unsigned int flags)
1208

1209
{
1210 1211
    bool storagemigration = flags & (VIR_MIGRATE_NON_SHARED_DISK |
                                     VIR_MIGRATE_NON_SHARED_INC);
1212
    size_t i;
1213
    int rc;
1214

1215
    for (i = 0; i < def->ndisks; i++) {
1216
        virDomainDiskDefPtr disk = def->disks[i];
1217
        const char *src = virDomainDiskGetSource(disk);
1218

1219 1220
        /* Disks without any source (i.e. floppies and CD-ROMs)
         * OR readonly are safe. */
1221
        if (virStorageSourceIsEmpty(disk->src) ||
1222
            disk->src->readonly)
1223
            continue;
1224

1225
        /* Disks which are migrated by qemu are safe too. */
1226
        if (storagemigration &&
1227
            qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1228
            continue;
1229

1230
        /* However, disks on local FS (e.g. ext4) are not safe. */
1231
        if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
1232
            if ((rc = virFileIsSharedFS(src)) < 0) {
1233
                return false;
1234 1235 1236 1237 1238
            } else if (rc == 0) {
                virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                               _("Migration without shared storage is unsafe"));
                return false;
            }
1239 1240 1241 1242 1243 1244 1245
            if ((rc = virStorageFileIsClusterFS(src)) < 0)
                return false;
            else if (rc == 1)
                continue;
        } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                   disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
            continue;
1246
        }
1247

1248 1249 1250 1251 1252 1253 1254
        /* 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;

1255 1256
        virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                       _("Migration may lead to data corruption if disks"
1257
                         " use cache != none or cache != directsync"));
1258
        return false;
1259 1260 1261 1262 1263
    }

    return true;
}

1264
/** qemuMigrationSrcSetOffline
1265 1266 1267
 * Pause domain for non-live migration.
 */
int
1268 1269
qemuMigrationSrcSetOffline(virQEMUDriverPtr driver,
                           virDomainObjPtr vm)
1270 1271
{
    int ret;
1272
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
1273 1274
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
1275
    if (ret == 0) {
1276
        virObjectEventPtr event;
1277

1278
        event = virDomainEventLifecycleNewFromObj(vm,
1279 1280
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
1281
        virObjectEventStateQueue(driver->domainEventState, event);
1282 1283 1284 1285 1286
    }

    return ret;
}

1287 1288

void
1289 1290
qemuMigrationAnyPostcopyFailed(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320
{
    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);
1321
        virObjectEventStateQueue(driver->domainEventState, event);
1322 1323 1324 1325 1326 1327 1328
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
    }
}


1329
static int
1330
qemuMigrationSrcWaitForSpice(virDomainObjPtr vm)
1331 1332 1333
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

1334 1335
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION) ||
        !priv->job.spiceMigration)
1336 1337
        return 0;

1338
    VIR_DEBUG("Waiting for SPICE to finish migration");
1339 1340
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
1341 1342 1343 1344
            return -1;
    }
    return 0;
}
1345

1346 1347 1348 1349

static void
qemuMigrationUpdateJobType(qemuDomainJobInfoPtr jobInfo)
{
1350
    switch ((qemuMonitorMigrationStatus) jobInfo->stats.mig.status) {
1351 1352 1353 1354
    case QEMU_MONITOR_MIGRATION_STATUS_POSTCOPY:
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_POSTCOPY;
        break;

1355
    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
1356
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED;
1357 1358 1359
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
1360
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_NONE;
1361 1362 1363
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
1364
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1365 1366 1367
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
1368
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1369 1370
        break;

1371 1372 1373 1374 1375 1376 1377 1378
    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;

1379 1380 1381
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLING:
1382
    case QEMU_MONITOR_MIGRATION_STATUS_LAST:
1383 1384 1385 1386 1387 1388
        break;
    }
}


int
1389 1390 1391 1392 1393
qemuMigrationAnyFetchStats(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuDomainAsyncJob asyncJob,
                           qemuDomainJobInfoPtr jobInfo,
                           char **error)
1394 1395
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1396
    qemuMonitorMigrationStats stats;
1397 1398 1399 1400 1401
    int rv;

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

1402
    rv = qemuMonitorGetMigrationStats(priv->mon, &stats, error);
1403 1404 1405 1406

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

1407
    jobInfo->stats.mig = stats;
1408 1409

    return 0;
1410 1411 1412
}


1413 1414 1415 1416 1417 1418 1419
static const char *
qemuMigrationJobName(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
1420
        return _("migration out job");
1421 1422 1423 1424
    case QEMU_ASYNC_JOB_SAVE:
        return _("domain save job");
    case QEMU_ASYNC_JOB_DUMP:
        return _("domain core dump job");
1425 1426 1427 1428 1429 1430 1431 1432 1433
    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:
1434 1435 1436 1437 1438 1439
    default:
        return _("job");
    }
}


1440
static int
1441
qemuMigrationJobCheckStatus(virQEMUDriverPtr driver,
1442
                            virDomainObjPtr vm,
1443
                            qemuDomainAsyncJob asyncJob)
1444 1445 1446
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1447
    char *error = NULL;
1448
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1449
    int ret = -1;
1450

1451
    if (!events ||
1452
        jobInfo->stats.mig.status == QEMU_MONITOR_MIGRATION_STATUS_ERROR) {
1453
        if (qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, &error) < 0)
1454 1455
            return -1;
    }
1456

1457 1458
    qemuMigrationUpdateJobType(jobInfo);

1459 1460
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_NONE:
1461 1462
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("is not active"));
1463
        goto cleanup;
1464

1465
    case QEMU_DOMAIN_JOB_STATUS_FAILED:
1466
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
1467 1468 1469
                       qemuMigrationJobName(vm),
                       error ? error : _("unexpectedly failed"));
        goto cleanup;
1470

1471
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
1472 1473
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("canceled by client"));
1474
        goto cleanup;
1475

1476 1477
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
1478
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
1479
    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
1480
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1481
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1482 1483
        break;
    }
1484 1485 1486 1487 1488 1489

    ret = 0;

 cleanup:
    VIR_FREE(error);
    return ret;
1490 1491 1492
}


1493 1494
enum qemuMigrationCompletedFlags {
    QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR = (1 << 0),
1495
    /* This flag should only be set when run on src host */
1496
    QEMU_MIGRATION_COMPLETED_CHECK_STORAGE  = (1 << 1),
1497
    QEMU_MIGRATION_COMPLETED_POSTCOPY       = (1 << 2),
1498
    QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER = (1 << 3),
1499 1500
};

1501

1502 1503 1504 1505 1506 1507 1508
/**
 * 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
1509 1510 1511 1512 1513
qemuMigrationAnyCompleted(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          qemuDomainAsyncJob asyncJob,
                          virConnectPtr dconn,
                          unsigned int flags)
1514 1515 1516 1517 1518
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    int pauseReason;

1519
    if (qemuMigrationJobCheckStatus(driver, vm, asyncJob) < 0)
1520 1521
        goto error;

1522
    /* This flag should only be set when run on src host */
1523
    if (flags & QEMU_MIGRATION_COMPLETED_CHECK_STORAGE &&
1524
        qemuMigrationSrcNBDStorageCopyReady(vm, asyncJob) < 0)
1525 1526
        goto error;

1527
    if (flags & QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR &&
1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540
        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;
    }

1541 1542 1543 1544 1545 1546 1547 1548 1549 1550
    /* 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;
    }

1551 1552 1553 1554 1555
    /* 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 &&
1556
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
1557 1558 1559 1560
        VIR_DEBUG("Migration switched to post-copy");
        return 1;
    }

1561
    if (jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
1562 1563 1564 1565 1566
        return 1;
    else
        return 0;

 error:
1567 1568 1569
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1570
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1571
        /* The migration was aborted by us rather than QEMU itself. */
1572
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1573
        return -2;
1574 1575 1576

    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
        /* Something failed after QEMU already finished the migration. */
1577
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1578
        return -1;
1579 1580 1581 1582

    case QEMU_DOMAIN_JOB_STATUS_FAILED:
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
        /* QEMU aborted the migration. */
1583
        return -1;
1584 1585 1586 1587 1588 1589

    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_NONE:
        /* Impossible. */
        break;
1590
    }
1591 1592

    return -1;
1593 1594 1595
}


1596 1597 1598
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
1599
static int
1600 1601 1602 1603 1604
qemuMigrationSrcWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  virConnectPtr dconn,
                                  unsigned int flags)
1605
{
1606
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
1607
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1608
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1609
    int rv;
1610

1611 1612
    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_MIGRATING;

1613 1614
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           dconn, flags)) != 1) {
1615 1616
        if (rv < 0)
            return rv;
1617

1618 1619
        if (events) {
            if (virDomainObjWait(vm) < 0) {
1620 1621
                if (virDomainObjIsActive(vm))
                    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1622 1623 1624 1625 1626 1627 1628 1629 1630 1631
                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);
        }
1632 1633
    }

1634
    if (events)
1635
        ignore_value(qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, NULL));
1636

1637
    qemuDomainJobInfoUpdateTime(jobInfo);
1638 1639
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
1640
    if (VIR_ALLOC(priv->job.completed) == 0) {
1641
        *priv->job.completed = *jobInfo;
1642 1643
        priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
    }
1644

1645 1646 1647 1648
    if (asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT &&
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;

1649
    return 0;
1650 1651 1652
}


1653
static int
1654 1655 1656 1657
qemuMigrationDstWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  bool postcopy)
1658 1659
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1660
    unsigned int flags = 0;
1661 1662 1663 1664 1665 1666 1667
    int rv;

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

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

1668 1669 1670
    if (postcopy)
        flags = QEMU_MIGRATION_COMPLETED_POSTCOPY;

1671 1672
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           NULL, flags)) != 1) {
1673 1674 1675 1676 1677 1678 1679 1680
        if (rv < 0 || virDomainObjWait(vm) < 0)
            return -1;
    }

    return 0;
}


1681
static int
1682 1683 1684 1685
qemuMigrationSrcGraphicsRelocate(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuMigrationCookiePtr cookie,
                                 const char *graphicsuri)
1686 1687
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1688 1689
    int ret = -1;
    const char *listenAddress = NULL;
1690
    virSocketAddr addr;
1691 1692 1693 1694 1695
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
1696

1697
    if (!cookie || (!cookie->graphics && !graphicsuri))
1698 1699
        return 0;

1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718
    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) {
1719
        size_t i;
1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746

        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;
            }
        }
    }
1747 1748 1749 1750

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
1751 1752 1753 1754
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
1755

1756 1757 1758 1759 1760 1761 1762 1763 1764
    /* 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;
    }

1765 1766 1767 1768
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
1769
        priv->job.spiceMigration = !ret;
1770 1771
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
1772
    }
1773

1774
 cleanup:
1775
    virURIFree(uri);
1776 1777 1778 1779
    return ret;
}


1780
static int
1781 1782 1783
qemuMigrationDstOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            qemuMigrationCookiePtr cookie)
1784
{
1785 1786
    virDomainNetDefPtr netptr;
    int ret = -1;
1787
    size_t i;
1788 1789 1790 1791 1792 1793 1794 1795

    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:
1796
           break;
1797
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
1798 1799
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
1800 1801 1802
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
1803 1804 1805
                goto cleanup;
            }
            break;
1806 1807 1808 1809 1810
        default:
            break;
        }
    }

1811
    ret = 0;
1812
 cleanup:
1813 1814 1815 1816
    return ret;
}


1817
int
1818 1819
qemuMigrationDstCheckProtocol(virQEMUCapsPtr qemuCaps,
                              const char *migrateFrom)
1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842
{
    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 *
1843 1844
qemuMigrationDstGetURI(const char *migrateFrom,
                       int migrateFd)
1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856
{
    char *uri = NULL;

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

    return uri;
}


1857
int
1858 1859 1860 1861
qemuMigrationDstRun(virQEMUDriverPtr driver,
                    virDomainObjPtr vm,
                    const char *uri,
                    qemuDomainAsyncJob asyncJob)
1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877
{
    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) {
1878
        /* qemuMigrationDstWaitForCompletion is called from the Finish phase */
1879 1880 1881 1882
        ret = 0;
        goto cleanup;
    }

1883
    if (qemuMigrationDstWaitForCompletion(driver, vm, asyncJob, false) < 0)
1884 1885 1886 1887 1888 1889 1890 1891 1892
        goto cleanup;

    ret = 0;

 cleanup:
    return ret;
}


1893 1894 1895 1896 1897 1898
/* 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.
 */
1899
static void
1900 1901 1902
qemuMigrationSrcCleanup(virDomainObjPtr vm,
                        virConnectPtr conn,
                        void *opaque)
1903
{
1904
    virQEMUDriverPtr driver = opaque;
1905 1906 1907 1908 1909 1910 1911 1912 1913
    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))
1914
        return;
1915 1916 1917 1918 1919

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

1920
    switch ((qemuMigrationJobPhase) priv->job.phase) {
1921 1922 1923 1924 1925 1926 1927 1928 1929
    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);
1930
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
1931
                                 priv->job.migParams, priv->job.apiFlags);
1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953
        /* 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 */
        ;
    }
}

1954

1955
/* The caller is supposed to lock the vm and start a migration job. */
1956
static char *
1957 1958 1959 1960 1961 1962 1963 1964 1965
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)
1966 1967 1968
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
1969
    virDomainDefPtr def = NULL;
1970
    qemuDomainObjPrivatePtr priv = vm->privateData;
1971
    virCapsPtr caps = NULL;
1972
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
1973

1974
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
1975
              " cookieout=%p, cookieoutlen=%p,"
1976
              " nmigrate_disks=%zu, migrate_disks=%p, flags=0x%lx",
1977
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
1978 1979
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
1980

1981 1982 1983
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

1984 1985 1986 1987 1988 1989
    /* 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);
1990

1991
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
1992 1993
        goto cleanup;

1994
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
1995
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
1996 1997
        goto cleanup;

1998 1999 2000 2001 2002 2003 2004 2005 2006
    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;
    }

2007 2008 2009 2010 2011 2012
    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;
    }

2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
    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;
            }
2056 2057 2058
        }
    }

2059
    if (virDomainDefHasMemoryHotplug(vm->def) ||
2060
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
2061
         vm->newDef && virDomainDefHasMemoryHotplug(vm->newDef)))
2062 2063
        cookieFlags |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

2064 2065 2066 2067 2068
    if (!qemuDomainVcpuHotplugIsInOrder(vm->def) ||
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
         vm->newDef && !qemuDomainVcpuHotplugIsInOrder(vm->newDef)))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU_HOTPLUG;

2069 2070 2071
    if (priv->origCPU)
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU;

2072 2073
    cookieFlags |= QEMU_MIGRATION_COOKIE_ALLOW_REBOOT;

2074 2075 2076
    if (!(flags & VIR_MIGRATE_OFFLINE))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CAPS;

2077 2078
    if (!(mig = qemuMigrationEatCookie(driver, vm->def,
                                       priv->origname, priv, NULL, 0, 0)))
2079 2080 2081
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
2082
                                QEMU_MIGRATION_SOURCE,
2083
                                cookieout, cookieoutlen,
2084
                                cookieFlags) < 0)
2085 2086
        goto cleanup;

L
liguang 已提交
2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108
    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;
        }
    }

2109
    if (xmlin) {
2110
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt, priv->qemuCaps,
2111 2112
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                            VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2113 2114
            goto cleanup;

2115
        if (!qemuDomainCheckABIStability(driver, vm, def))
2116 2117
            goto cleanup;

2118
        rv = qemuDomainDefFormatLive(driver, def, NULL, false, true);
2119
    } else {
2120 2121
        rv = qemuDomainDefFormatLive(driver, vm->def, priv->origCPU,
                                     false, true);
2122
    }
2123

2124
 cleanup:
2125
    qemuMigrationCookieFree(mig);
2126
    virObjectUnref(caps);
2127
    virDomainDefFree(def);
2128 2129 2130
    return rv;
}

2131
char *
2132 2133 2134 2135 2136 2137 2138 2139 2140
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)
2141 2142 2143
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
2144
    qemuDomainAsyncJob asyncJob;
2145 2146

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
2147 2148
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
2149 2150 2151 2152 2153 2154 2155 2156
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_MIGRATION_OUT;
    } else {
        if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_NONE;
    }

2157
    qemuMigrationSrcStoreDomainState(vm);
2158

2159
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
2160 2161 2162 2163 2164 2165
        goto endjob;

    /* Check if there is any ejected media.
     * We don't want to require them on the destination.
     */
    if (!(flags & VIR_MIGRATE_OFFLINE) &&
2166
        qemuProcessRefreshDisks(driver, vm, asyncJob) < 0)
2167 2168
        goto endjob;

2169 2170 2171
    if (!(xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                           cookieout, cookieoutlen,
                                           nmigrate_disks, migrate_disks, flags)))
2172 2173 2174 2175 2176 2177 2178
        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.
         */
2179
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
2180
                                 qemuMigrationSrcCleanup) < 0) {
2181
            VIR_FREE(xml);
2182
            goto endjob;
2183
        }
2184
        qemuMigrationJobContinue(vm);
2185 2186 2187 2188
    } else {
        goto endjob;
    }

2189
 cleanup:
M
Michal Privoznik 已提交
2190
    virDomainObjEndAPI(&vm);
2191 2192
    return xml;

2193
 endjob:
2194 2195 2196 2197
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2198 2199 2200
    goto cleanup;
}

2201

2202 2203
/* Prepare is the first step, and it runs on the destination host.
 */
2204

2205
static void
2206 2207
qemuMigrationDstPrepareCleanup(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
2208 2209 2210 2211 2212 2213 2214 2215 2216
{
    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));

2217
    virPortAllocatorRelease(priv->migrationPort);
2218 2219
    priv->migrationPort = 0;

2220 2221 2222 2223 2224
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2225
static qemuProcessIncomingDefPtr
2226 2227 2228 2229 2230 2231
qemuMigrationDstPrepare(virDomainObjPtr vm,
                        bool tunnel,
                        const char *protocol,
                        const char *listenAddress,
                        unsigned short port,
                        int fd)
2232 2233
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2234
    qemuProcessIncomingDefPtr inc = NULL;
2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264
    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 */
            }
2265
        } else if (hostIPv6Capable) {
2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288
            /* 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;
    }

2289 2290
    inc = qemuProcessIncomingDefNew(priv->qemuCaps, listenAddress,
                                    migrateFrom, fd, NULL);
2291

2292
 cleanup:
2293 2294
    VIR_FREE(migrateFrom);
    return inc;
2295 2296
}

2297
static int
2298
qemuMigrationDstPrepareAny(virQEMUDriverPtr driver,
2299
                           virConnectPtr dconn,
2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313
                           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,
2314
                           qemuMigrationParamsPtr migParams,
2315
                           unsigned long flags)
2316 2317
{
    virDomainObjPtr vm = NULL;
2318
    virObjectEventPtr event = NULL;
2319
    virErrorPtr origErr;
2320
    int ret = -1;
2321
    int dataFD[2] = { -1, -1 };
2322
    qemuDomainObjPrivatePtr priv = NULL;
2323
    qemuMigrationCookiePtr mig = NULL;
2324
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2325
    char *xmlout = NULL;
L
liguang 已提交
2326
    unsigned int cookieFlags;
2327
    unsigned int startFlags;
2328
    virCapsPtr caps = NULL;
2329
    qemuProcessIncomingDefPtr incoming = NULL;
2330
    bool taint_hook = false;
2331 2332 2333
    bool stopProcess = false;
    bool relabel = false;
    int rv;
2334
    char *tlsAlias = NULL;
2335

2336 2337
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357
    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;
        }
2358 2359
        cookieFlags = 0;
    } else {
2360 2361
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS |
                      QEMU_MIGRATION_COOKIE_CAPS;
L
liguang 已提交
2362 2363
    }

2364 2365 2366 2367 2368 2369 2370 2371 2372
    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;
    }

2373 2374 2375 2376 2377 2378
    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;
    }

2379 2380 2381
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2382
    if (!qemuMigrationSrcIsAllowedHostdev(*def))
2383 2384
        goto cleanup;

J
Jiri Denemark 已提交
2385 2386 2387 2388 2389
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2390
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2391 2392
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2393 2394
            goto cleanup;

2395
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2396 2397 2398 2399 2400 2401 2402
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2403
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2404 2405 2406 2407 2408 2409
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2410
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt, NULL,
2411 2412
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                                 VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE);
J
Jiri Denemark 已提交
2413 2414 2415
                if (!newdef)
                    goto cleanup;

2416
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2417 2418 2419 2420
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2421 2422
                virDomainDefFree(*def);
                *def = newdef;
2423 2424 2425 2426
                /* 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
2427 2428 2429 2430
            }
        }
    }

2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444
    /* 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;

2445
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2446
                                   driver->xmlopt,
2447 2448 2449
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2450
        goto cleanup;
2451
    *def = NULL;
2452

2453
    priv = vm->privateData;
2454 2455
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2456

2457 2458 2459 2460 2461
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2462 2463
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
2464 2465 2466 2467 2468 2469
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2470 2471 2472
    if (qemuMigrationDstPrecreateStorage(vm, mig->nbd,
                                         nmigrate_disks, migrate_disks,
                                         !!(flags & VIR_MIGRATE_NON_SHARED_INC)) < 0)
2473 2474
        goto cleanup;

2475 2476
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                              flags) < 0)
2477
        goto cleanup;
2478
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2479 2480 2481 2482

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

L
liguang 已提交
2483 2484 2485
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2486 2487
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
2488 2489
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
2490
        goto stopjob;
2491 2492
    }

2493 2494
    startFlags = VIR_QEMU_PROCESS_START_AUTODESTROY;

2495
    if (qemuProcessInit(driver, vm, mig->cpu, QEMU_ASYNC_JOB_MIGRATION_IN,
2496
                        true, startFlags) < 0)
2497
        goto stopjob;
2498
    stopProcess = true;
2499

2500 2501
    priv->allowReboot = mig->allowReboot;

2502 2503 2504
    if (!(incoming = qemuMigrationDstPrepare(vm, tunnel, protocol,
                                             listenAddress, port,
                                             dataFD[0])))
2505
        goto stopjob;
2506

2507
    if (qemuProcessPrepareDomain(driver, vm, startFlags) < 0)
2508 2509
        goto stopjob;

2510
    if (qemuProcessPrepareHost(driver, vm, startFlags) < 0)
2511 2512
        goto stopjob;

2513
    rv = qemuProcessLaunch(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2514 2515
                           incoming, NULL,
                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
2516
                           startFlags);
2517 2518 2519
    if (rv < 0) {
        if (rv == -2)
            relabel = true;
2520
        goto stopjob;
2521
    }
2522
    relabel = true;
2523

2524 2525 2526 2527
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
2528
            goto stopjob;
2529
        }
2530
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
2531 2532
    }

2533
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
2534
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
2535
        goto stopjob;
M
Michael R. Hines 已提交
2536 2537
    }

2538
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2539
                                 migParams, mig->caps->automatic) < 0)
2540 2541 2542 2543 2544 2545 2546
        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,
2547
                                         &tlsAlias, NULL,
2548 2549 2550 2551 2552 2553 2554
                                         migParams) < 0)
            goto stopjob;
    } else {
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
            goto stopjob;
    }

2555 2556
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                                 migParams) < 0)
2557 2558
        goto stopjob;

2559 2560 2561
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573
        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;
        }

2574 2575
        if (qemuMigrationDstStartNBDServer(driver, vm, incoming->address,
                                           nmigrate_disks, migrate_disks,
2576
                                           nbdPort, nbdTLSAlias) < 0) {
2577
            goto stopjob;
2578
        }
2579
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
2580 2581
    }

2582 2583 2584 2585 2586 2587 2588 2589 2590
    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");
    }

2591
    if (incoming->deferredURI &&
2592 2593
        qemuMigrationDstRun(driver, vm, incoming->deferredURI,
                            QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2594 2595
        goto stopjob;

2596
    if (qemuProcessFinishStartup(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2597 2598 2599
                                 false, VIR_DOMAIN_PAUSED_MIGRATION) < 0)
        goto stopjob;

2600
 done:
2601 2602 2603
    if (qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_DESTINATION,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
2604 2605 2606 2607 2608 2609 2610
        /* 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");
    }

2611
    if (qemuDomainCleanupAdd(vm, qemuMigrationDstPrepareCleanup) < 0)
2612
        goto stopjob;
2613

L
liguang 已提交
2614 2615
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
2616
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
2617 2618 2619
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
2620

2621 2622 2623 2624
    /* 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.
     */
2625
    qemuMigrationJobContinue(vm);
2626

2627 2628
    if (autoPort)
        priv->migrationPort = port;
2629 2630 2631 2632 2633
    /* 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;
2634
    ret = 0;
2635

2636
 cleanup:
2637
    virErrorPreserveLast(&origErr);
2638
    VIR_FREE(tlsAlias);
2639
    qemuProcessIncomingDefFree(incoming);
J
Jiri Denemark 已提交
2640
    VIR_FREE(xmlout);
2641 2642
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
2643 2644 2645
    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 */
2646
        VIR_FREE(priv->origname);
2647 2648 2649 2650
        /* release if port is auto selected which is not the case if
         * it is given in parameters
         */
        if (nbdPort == 0)
2651
            virPortAllocatorRelease(priv->nbdPort);
2652
        priv->nbdPort = 0;
2653
        virDomainObjRemoveTransientDef(vm);
2654
        qemuDomainRemoveInactiveJob(driver, vm);
2655
    }
M
Michal Privoznik 已提交
2656
    virDomainObjEndAPI(&vm);
2657
    virObjectEventStateQueue(driver->domainEventState, event);
2658
    qemuMigrationCookieFree(mig);
2659
    virObjectUnref(caps);
2660
    virNWFilterUnlockFilterUpdates();
2661
    virErrorRestore(&origErr);
2662
    return ret;
2663

2664
 stopjob:
2665
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2666
                             priv->job.migParams, priv->job.apiFlags);
2667

2668 2669 2670 2671
    if (stopProcess) {
        unsigned int stopFlags = VIR_QEMU_PROCESS_STOP_MIGRATED;
        if (!relabel)
            stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
2672
        virDomainAuditStart(vm, "migrated", false);
2673 2674
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        QEMU_ASYNC_JOB_MIGRATION_IN, stopFlags);
2675
    }
2676

2677
    qemuMigrationJobFinish(driver, vm);
2678
    goto cleanup;
2679 2680 2681
}


2682 2683 2684 2685 2686
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
2687
qemuMigrationDstPrepareTunnel(virQEMUDriverPtr driver,
2688
                              virConnectPtr dconn,
2689 2690 2691 2692 2693 2694 2695
                              const char *cookiein,
                              int cookieinlen,
                              char **cookieout,
                              int *cookieoutlen,
                              virStreamPtr st,
                              virDomainDefPtr *def,
                              const char *origname,
2696
                              qemuMigrationParamsPtr migParams,
2697
                              unsigned long flags)
2698
{
2699
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2700
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
2701
              "origname=%s, flags=0x%lx",
2702
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2703
              cookieout, cookieoutlen, st, *def, origname, flags);
2704

2705 2706 2707 2708 2709 2710
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

2711 2712 2713
    return qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
                                      cookieout, cookieoutlen, def, origname,
                                      st, NULL, 0, false, NULL, 0, NULL, 0,
2714
                                      migParams, flags);
2715 2716 2717
}


2718
static virURIPtr
2719
qemuMigrationAnyParseURI(const char *uri, bool *wellFormed)
2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740
{
    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;
}


2741
int
2742
qemuMigrationDstPrepareDirect(virQEMUDriverPtr driver,
2743
                              virConnectPtr dconn,
2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755
                              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,
2756
                              qemuMigrationParamsPtr migParams,
2757
                              unsigned long flags)
2758
{
2759 2760
    unsigned short port = 0;
    bool autoPort = true;
2761 2762
    char *hostname = NULL;
    int ret = -1;
2763
    virURIPtr uri = NULL;
2764 2765
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
2766

2767
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2768
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
2769
              "def=%p, origname=%s, listenAddress=%s, "
2770
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, flags=0x%lx",
2771
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2772
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
2773
              *def, origname, NULLSTR(listenAddress),
2774
              nmigrate_disks, migrate_disks, nbdPort, flags);
2775

2776 2777
    *uri_out = NULL;

2778 2779 2780
    /* 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
2781 2782 2783
     * 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".
2784 2785 2786 2787 2788 2789
     *
     * 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) {
2790 2791 2792
        bool encloseAddress = false;
        const char *incFormat;

2793
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2794
            goto cleanup;
2795

2796
        if (migrateHost != NULL) {
2797 2798
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
2799

2800
            if (VIR_STRDUP(hostname, migrateHost) < 0)
2801 2802 2803 2804 2805
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
2806 2807

        if (STRPREFIX(hostname, "localhost")) {
2808 2809 2810
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
2811 2812 2813 2814 2815
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
2816
         * compatibility with old targets. We at least make the
2817 2818
         * new targets accept both syntaxes though.
         */
2819 2820 2821 2822 2823 2824
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
2825 2826
            goto cleanup;
    } else {
2827
        bool well_formed_uri;
J
Ján Tomko 已提交
2828

2829
        if (!(uri = qemuMigrationAnyParseURI(uri_in, &well_formed_uri)))
2830
            goto cleanup;
J
Ján Tomko 已提交
2831

2832 2833 2834 2835 2836 2837 2838
        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
2839 2840
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
2841 2842 2843
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
2844 2845 2846 2847 2848 2849 2850 2851 2852 2853
            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) {
2854
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2855
                goto cleanup;
2856

2857
            /* Send well-formed URI only if uri_in was well-formed */
2858 2859 2860 2861 2862 2863 2864 2865
            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;
            }
2866
        } else {
2867 2868
            port = uri->port;
            autoPort = false;
2869 2870 2871 2872 2873 2874
        }
    }

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

2875
    ret = qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
2876 2877 2878 2879
                                     cookieout, cookieoutlen, def, origname,
                                     NULL, uri ? uri->scheme : "tcp",
                                     port, autoPort, listenAddress,
                                     nmigrate_disks, migrate_disks, nbdPort,
2880
                                     migParams, flags);
2881
 cleanup:
2882
    virURIFree(uri);
2883
    VIR_FREE(hostname);
2884
    virObjectUnref(cfg);
2885
    if (ret != 0) {
2886
        VIR_FREE(*uri_out);
2887
        if (autoPort)
2888
            virPortAllocatorRelease(port);
2889
    }
2890 2891 2892 2893
    return ret;
}


2894
virDomainDefPtr
2895 2896 2897 2898
qemuMigrationAnyPrepareDef(virQEMUDriverPtr driver,
                           const char *dom_xml,
                           const char *dname,
                           char **origname)
2899 2900 2901
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
2902
    char *name = NULL;
2903 2904 2905 2906 2907 2908 2909 2910 2911 2912

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

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

2913
    if (!(def = virDomainDefParseString(dom_xml, caps, driver->xmlopt, NULL,
2914 2915
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                        VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2916 2917 2918
        goto cleanup;

    if (dname) {
2919
        name = def->name;
2920 2921 2922 2923 2924 2925
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

2926
 cleanup:
2927
    virObjectUnref(caps);
2928 2929 2930 2931
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
2932 2933 2934 2935
    return def;
}


2936
static int
2937 2938 2939 2940 2941 2942
qemuMigrationSrcConfirmPhase(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
                             const char *cookiein,
                             int cookieinlen,
                             unsigned int flags,
                             int retcode)
2943 2944
{
    qemuMigrationCookiePtr mig;
2945
    virObjectEventPtr event;
2946 2947
    int rv = -1;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2948 2949
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = NULL;
2950

2951
    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
2952
              "flags=0x%x, retcode=%d",
2953
              driver, vm, NULLSTR(cookiein), cookieinlen,
2954 2955 2956 2957 2958 2959 2960 2961 2962
              flags, retcode);

    virCheckFlags(QEMU_MIGRATION_FLAGS, -1);

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

2963 2964
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen,
2965
                                       QEMU_MIGRATION_COOKIE_STATS)))
2966 2967
        goto cleanup;

2968 2969 2970
    if (retcode == 0)
        jobInfo = priv->job.completed;
    else
2971
        VIR_FREE(priv->job.completed);
2972 2973 2974

    /* Update times with the values sent by the destination daemon */
    if (mig->jobInfo && jobInfo) {
2975 2976 2977 2978 2979 2980 2981 2982
        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 &&
2983 2984
            qemuMigrationAnyFetchStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                       jobInfo, NULL) < 0)
2985 2986
            VIR_WARN("Could not refresh migration statistics");

2987 2988 2989
        qemuDomainJobInfoUpdateTime(jobInfo);
        jobInfo->timeDeltaSet = mig->jobInfo->timeDeltaSet;
        jobInfo->timeDelta = mig->jobInfo->timeDelta;
2990 2991
        jobInfo->stats.mig.downtime_set = mig->jobInfo->stats.mig.downtime_set;
        jobInfo->stats.mig.downtime = mig->jobInfo->stats.mig.downtime;
2992 2993
    }

2994 2995 2996
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2997 2998
    /* 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.
2999 3000 3001 3002
     */
    if (retcode == 0) {
        /* If guest uses SPICE and supports seamless migration we have to hold
         * up domain shutdown until SPICE server transfers its data */
3003
        qemuMigrationSrcWaitForSpice(vm);
3004 3005

        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
3006
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
3007 3008 3009
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
        virDomainAuditStop(vm, "migrated");

3010
        event = virDomainEventLifecycleNewFromObj(vm,
3011 3012
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
3013
        virObjectEventStateQueue(driver->domainEventState, event);
3014
        qemuDomainEventEmitJobCompleted(driver, vm);
3015
    } else {
3016
        virErrorPtr orig_err = virSaveLastError();
3017
        int reason;
3018 3019

        /* cancel any outstanding NBD jobs */
3020 3021
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT, NULL);
3022 3023 3024

        virSetError(orig_err);
        virFreeError(orig_err);
3025

3026
        if (virDomainObjGetState(vm, &reason) == VIR_DOMAIN_PAUSED &&
3027
            reason == VIR_DOMAIN_PAUSED_POSTCOPY)
3028
            qemuMigrationAnyPostcopyFailed(driver, vm);
3029 3030
        else
            qemuMigrationSrcRestoreDomainState(driver, vm);
3031

3032
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3033
                                 priv->job.migParams, priv->job.apiFlags);
3034

3035
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
3036 3037 3038
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
    }

3039
 done:
3040 3041 3042
    qemuMigrationCookieFree(mig);
    rv = 0;

3043
 cleanup:
3044 3045 3046 3047 3048
    virObjectUnref(cfg);
    return rv;
}

int
3049 3050 3051 3052 3053 3054
qemuMigrationSrcConfirm(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *cookiein,
                        int cookieinlen,
                        unsigned int flags,
                        int cancelled)
3055
{
3056
    qemuMigrationJobPhase phase;
3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070
    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);
3071
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
3072
                           qemuMigrationSrcCleanup);
3073

3074 3075 3076
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
3077

3078
    qemuMigrationJobFinish(driver, vm);
3079
    if (!virDomainObjIsActive(vm)) {
3080
        if (!cancelled && ret == 0 && flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
3081
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
3082 3083
            vm->persistent = 0;
        }
3084
        qemuDomainRemoveInactiveJob(driver, vm);
3085 3086
    }

3087
 cleanup:
M
Michal Privoznik 已提交
3088
    virDomainObjEndAPI(&vm);
3089 3090 3091 3092 3093
    virObjectUnref(cfg);
    return ret;
}


3094 3095
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3096
    MIGRATION_DEST_CONNECT_HOST,
3097
    MIGRATION_DEST_FD,
3098
};
3099

3100 3101 3102 3103
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3104

3105 3106 3107 3108 3109 3110
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3111
            const char *protocol;
3112 3113 3114 3115
            const char *name;
            int port;
        } host;

3116 3117 3118 3119
        struct {
            int qemu;
            int local;
        } fd;
3120 3121 3122 3123 3124 3125 3126
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3127 3128 3129

#define TUNNEL_SEND_BUF_SIZE 65536

3130 3131 3132 3133 3134 3135 3136
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3137 3138
    int wakeupRecvFD;
    int wakeupSendFD;
3139 3140
};

3141
static void qemuMigrationSrcIOFunc(void *arg)
3142
{
3143
    qemuMigrationIOThreadPtr data = arg;
3144 3145 3146 3147 3148 3149 3150
    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);
3151

3152
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3153
        goto abrt;
3154

3155 3156 3157
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3158
    for (;;) {
3159 3160 3161 3162 3163 3164 3165 3166 3167 3168
        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;
3169
            virReportSystemError(errno, "%s",
3170 3171
                                 _("poll failed in migration tunnel"));
            goto abrt;
3172
        }
3173 3174 3175 3176 3177 3178 3179

        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");
3180
            break;
3181
        }
3182

3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198
        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;
            }
3199 3200
        }

3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217
        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;
            }
        }
    }
3218

3219 3220
    if (virStreamFinish(data->st) < 0)
        goto error;
3221

3222
    VIR_FORCE_CLOSE(data->sock);
3223 3224
    VIR_FREE(buffer);

3225 3226
    return;

3227
 abrt:
3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3239
 error:
3240 3241 3242 3243 3244
    /* 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);
3245
    virResetLastError();
3246
    VIR_FREE(buffer);
3247 3248 3249 3250
}


static qemuMigrationIOThreadPtr
3251 3252
qemuMigrationSrcStartTunnel(virStreamPtr st,
                            int sock)
3253
{
3254 3255
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3256

3257 3258 3259 3260
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3261 3262
    }

3263
    if (VIR_ALLOC(io) < 0)
3264
        goto error;
3265

3266 3267
    io->st = st;
    io->sock = sock;
3268 3269
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3270 3271

    if (virThreadCreate(&io->thread, true,
3272
                        qemuMigrationSrcIOFunc,
3273 3274 3275
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3276
        goto error;
3277 3278 3279
    }

    return io;
3280

3281
 error:
3282 3283 3284 3285
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3286 3287 3288
}

static int
3289
qemuMigrationSrcStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3290 3291
{
    int rv = -1;
3292 3293 3294 3295 3296 3297 3298 3299 3300
    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;
    }

3301 3302 3303 3304
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3305 3306 3307 3308
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3309 3310 3311 3312 3313 3314
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3315
 cleanup:
3316 3317
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3318 3319
    VIR_FREE(io);
    return rv;
3320 3321
}

3322
static int
3323 3324 3325
qemuMigrationSrcConnect(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationSpecPtr spec)
3326 3327 3328 3329 3330 3331 3332
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3333
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3334 3335 3336 3337 3338
        return -1;

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

3339
    if (qemuSecuritySetSocketLabel(driver->securityManager, vm->def) < 0)
3340
        goto cleanup;
3341 3342 3343
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
3344
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
3345
        virObjectUnref(sock);
3346
    }
3347
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0 ||
3348 3349 3350
        spec->dest.fd.qemu == -1)
        goto cleanup;

3351 3352 3353 3354 3355 3356 3357
    /* 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;
    }

3358 3359
    ret = 0;

3360
 cleanup:
3361 3362 3363 3364 3365 3366
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3367 3368

static int
3369 3370 3371 3372
qemuMigrationSrcContinue(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMonitorMigrationStatus status,
                         qemuDomainAsyncJob asyncJob)
3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388
{
    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;
}


3389
static int
3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403
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,
3404
                    qemuMigrationParamsPtr migParams)
3405
{
3406
    int ret = -1;
3407 3408
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3409
    qemuMigrationCookiePtr mig = NULL;
3410
    char *tlsAlias = NULL;
3411
    qemuMigrationIOThreadPtr iothread = NULL;
3412
    int fd = -1;
3413
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3414
    virErrorPtr orig_err = NULL;
3415
    unsigned int cookieFlags = 0;
3416
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3417
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
3418
    bool cancel = false;
3419
    unsigned int waitFlags;
3420
    virDomainDefPtr persistDef = NULL;
3421
    char *timestamp;
3422
    int rc;
3423 3424

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
3425
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3426 3427
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
3428 3429
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3430
              spec, spec->destType, spec->fwdType, dconn,
3431
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3432

3433 3434 3435 3436 3437 3438 3439 3440 3441 3442
    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;
    }

3443 3444
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3445 3446 3447 3448
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3449 3450 3451
        return -1;
    }

3452 3453 3454
    if (events)
        priv->signalIOError = abort_on_error;

3455 3456
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
        if (persist_xml) {
3457 3458
            if (!(persistDef = qemuMigrationAnyPrepareDef(driver, persist_xml,
                                                          NULL, NULL)))
3459
                goto error;
3460 3461 3462
        } else {
            virDomainDefPtr def = vm->newDef ? vm->newDef : vm->def;
            if (!(persistDef = qemuDomainDefCopy(driver, def,
3463 3464
                                                 VIR_DOMAIN_XML_SECURE |
                                                 VIR_DOMAIN_XML_MIGRATABLE)))
3465
                goto error;
3466 3467 3468
        }
    }

3469 3470
    mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                 cookiein, cookieinlen,
3471 3472 3473
                                 cookieFlags |
                                 QEMU_MIGRATION_COOKIE_GRAPHICS |
                                 QEMU_MIGRATION_COOKIE_CAPS);
3474
    if (!mig)
3475
        goto error;
3476

3477
    if (qemuMigrationSrcGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3478 3479
        VIR_WARN("unable to provide data for graphics client relocation");

3480
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3481
                                 migParams, mig->caps->automatic) < 0)
3482 3483
        goto error;

3484
    if (flags & VIR_MIGRATE_TLS) {
3485
        const char *hostname = NULL;
3486 3487 3488 3489

        /* We need to add tls-hostname whenever QEMU itself does not
         * connect directly to the destination. */
        if (spec->destType == MIGRATION_DEST_CONNECT_HOST ||
3490 3491 3492
            spec->destType == MIGRATION_DEST_FD)
            hostname = spec->dest.host.name;

3493
        if (qemuMigrationParamsEnableTLS(driver, vm, false,
3494
                                         QEMU_ASYNC_JOB_MIGRATION_OUT,
3495
                                         &tlsAlias, hostname,
3496 3497
                                         migParams) < 0)
            goto error;
3498
    } else {
3499
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
3500
            goto error;
3501 3502
    }

3503 3504 3505 3506
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                 migParams) < 0)
        goto error;

3507 3508 3509
    if (migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                         QEMU_MONITOR_MIGRATE_NON_SHARED_INC)) {
        if (mig->nbd) {
3510 3511 3512 3513
            /* 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. */
3514 3515
            if (flags & VIR_MIGRATE_TLS &&
                !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV_DEL)) {
3516 3517 3518 3519 3520
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                               _("NBD migration with TLS is not supported"));
                goto error;
            }

3521
            /* This will update migrate_flags on success */
3522 3523 3524 3525 3526 3527
            if (qemuMigrationSrcNBDStorageCopy(driver, vm, mig,
                                               spec->dest.host.name,
                                               migrate_speed,
                                               &migrate_flags,
                                               nmigrate_disks,
                                               migrate_disks,
3528
                                               dconn, tlsAlias, flags) < 0) {
3529
                goto error;
3530 3531 3532 3533 3534 3535 3536
            }
        } 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.");
        }
3537 3538
    }

3539
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3540 3541
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
3542
        if (qemuMigrationSrcSetOffline(driver, vm) < 0)
3543
            goto error;
3544 3545
    }

3546 3547
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3548
        goto error;
3549

3550
    if (priv->job.abortJob) {
3551 3552
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
3553
         * priv->job.abortJob will not change */
3554
        priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
3555 3556 3557
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
3558
        goto exit_monitor;
3559 3560
    }

3561 3562
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
3563

3564 3565
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3566
        qemuMigrationSrcConnect(driver, vm, spec) < 0) {
3567
        goto exit_monitor;
3568
    }
3569

3570 3571 3572 3573 3574 3575
    /* log start of migration */
    if ((timestamp = virTimeStringNow()) != NULL) {
        qemuDomainLogAppendMessage(driver, vm, "%s: initiating migration\n", timestamp);
        VIR_FREE(timestamp);
    }

3576
    rc = -1;
3577 3578
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3579 3580
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
3581
            goto exit_monitor;
M
Michael R. Hines 已提交
3582
        }
3583 3584 3585 3586
        rc = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
                                      spec->dest.host.protocol,
                                      spec->dest.host.name,
                                      spec->dest.host.port);
3587 3588
        break;

3589 3590 3591 3592
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3593
    case MIGRATION_DEST_FD:
3594
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3595
            fd = spec->dest.fd.local;
3596 3597
            spec->dest.fd.local = -1;
        }
3598 3599
        rc = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                    spec->dest.fd.qemu);
3600 3601
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3602
    }
3603 3604

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
3605
        goto error;
3606 3607 3608

    /* From this point onwards we *must* call cancel to abort the
     * migration on source if anything goes wrong */
3609
    cancel = true;
3610

3611
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3612
        if (!(iothread = qemuMigrationSrcStartTunnel(spec->fwd.stream, fd)))
3613
            goto error;
3614 3615 3616 3617 3618
        /* If we've created a tunnel, then the 'fd' will be closed in the
         * qemuMigrationIOFunc as data->sock.
         */
        fd = -1;
    }
3619

3620
    waitFlags = QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;
3621 3622 3623 3624 3625 3626 3627
    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;

3628 3629 3630
    rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT,
                                           dconn, waitFlags);
3631
    if (rc == -2) {
3632
        goto error;
3633 3634 3635 3636 3637
    } else if (rc == -1) {
        /* QEMU reported failed migration, nothing to cancel anymore */
        cancel = false;
        goto error;
    }
3638

3639 3640 3641
    /* 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.
3642
     */
3643 3644 3645 3646 3647
    if (priv->monJSON) {
        while (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            priv->signalStop = true;
            rc = virDomainObjWait(vm);
            priv->signalStop = false;
3648
            if (rc < 0)
3649
                goto error;
3650
        }
3651
    } else if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING &&
3652
               qemuMigrationSrcSetOffline(driver, vm) < 0) {
3653
        goto error;
3654
    }
3655

J
Ján Tomko 已提交
3656
    if (mig->nbd &&
3657 3658 3659
        qemuMigrationSrcNBDCopyCancel(driver, vm, true,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT,
                                      dconn) < 0)
3660
        goto error;
3661

3662 3663 3664 3665 3666
    /* 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) {
3667 3668 3669
        if (qemuMigrationSrcContinue(driver, vm,
                                     QEMU_MONITOR_MIGRATION_STATUS_PRE_SWITCHOVER,
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3670 3671 3672 3673
            goto error;

        waitFlags ^= QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;

3674 3675 3676
        rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_OUT,
                                               dconn, waitFlags);
3677 3678 3679 3680 3681 3682 3683 3684 3685
        if (rc == -2) {
            goto error;
        } else if (rc == -1) {
            /* QEMU reported failed migration, nothing to cancel anymore */
            cancel = false;
            goto error;
        }
    }

3686 3687 3688 3689
    if (iothread) {
        qemuMigrationIOThreadPtr io;

        VIR_STEAL_PTR(io, iothread);
3690
        if (qemuMigrationSrcStopTunnel(io, false) < 0)
3691
            goto error;
3692 3693 3694
    }

    if (priv->job.completed) {
3695
        priv->job.completed->stopped = priv->job.current->stopped;
3696 3697 3698 3699 3700 3701 3702 3703 3704
        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 ||
3705 3706 3707
        qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_SOURCE,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
3708 3709
        VIR_WARN("Unable to encode migration cookie");
    }
3710

3711
    ret = 0;
3712

3713
 cleanup:
3714
    VIR_FREE(tlsAlias);
3715
    VIR_FORCE_CLOSE(fd);
3716
    virDomainDefFree(persistDef);
3717 3718
    qemuMigrationCookieFree(mig);

3719 3720 3721
    if (events)
        priv->signalIOError = false;

3722 3723 3724 3725 3726
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

3727 3728
    return ret;

3729
 error:
3730 3731
    orig_err = virSaveLastError();

3732 3733 3734 3735 3736 3737 3738 3739
    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));
        }
3740

3741 3742
        /* cancel any outstanding NBD jobs */
        if (mig && mig->nbd)
3743 3744 3745
            qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                          QEMU_ASYNC_JOB_MIGRATION_OUT,
                                          dconn);
3746 3747 3748 3749

        if (priv->job.current->status != QEMU_DOMAIN_JOB_STATUS_CANCELED)
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
    }
3750 3751

    if (iothread)
3752
        qemuMigrationSrcStopTunnel(iothread, true);
3753 3754 3755

    goto cleanup;

3756 3757
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
3758
    goto error;
3759 3760
}

3761
/* Perform migration using QEMU's native migrate support,
3762 3763
 * not encrypted obviously
 */
3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778
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,
3779
                              qemuMigrationParamsPtr migParams)
3780
{
3781
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
3782
    virURIPtr uribits = NULL;
3783
    int ret = -1;
3784 3785 3786
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
3787
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3788
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
3789
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
3790
              cookieout, cookieoutlen, flags, resource,
3791
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3792

3793
    if (!(uribits = qemuMigrationAnyParseURI(uri, NULL)))
3794 3795
        return -1;

3796 3797 3798 3799 3800 3801 3802
    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
3803 3804 3805 3806 3807
    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"));
3808
            goto cleanup;
M
Michael R. Hines 已提交
3809
        }
3810
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
3811 3812 3813
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
3814
            goto cleanup;
M
Michael R. Hines 已提交
3815 3816 3817
        }
    }

3818
    if (STRNEQ(uribits->scheme, "rdma"))
3819 3820
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
3821
        spec.destType = MIGRATION_DEST_HOST;
3822
    spec.dest.host.protocol = uribits->scheme;
3823 3824 3825
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
3826

3827 3828 3829
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen, cookieout,
                              cookieoutlen, flags, resource, &spec, dconn,
                              graphicsuri, nmigrate_disks, migrate_disks,
3830
                              migParams);
3831 3832 3833 3834

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

3835
 cleanup:
3836
    virURIFree(uribits);
3837 3838 3839 3840 3841

    return ret;
}


3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856
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,
3857
                              qemuMigrationParamsPtr migParams)
3858 3859 3860
{
    int ret = -1;
    qemuMigrationSpec spec;
3861
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3862
    int fds[2] = { -1, -1 };
3863 3864

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
3865
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3866
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
3867
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
3868
              cookieout, cookieoutlen, flags, resource,
3869
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3870 3871 3872 3873

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

3874

3875 3876 3877
    spec.destType = MIGRATION_DEST_FD;
    spec.dest.fd.qemu = -1;
    spec.dest.fd.local = -1;
3878

3879 3880 3881 3882 3883
    if (pipe2(fds, O_CLOEXEC) == 0) {
        spec.dest.fd.qemu = fds[1];
        spec.dest.fd.local = fds[0];
    }
    if (spec.dest.fd.qemu == -1 ||
3884 3885
        qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    spec.dest.fd.qemu) < 0) {
3886 3887 3888
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
        goto cleanup;
3889 3890
    }

3891 3892 3893
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen,
                              cookieout, cookieoutlen, flags, resource, &spec,
                              dconn, graphicsuri, nmigrate_disks, migrate_disks,
3894
                              migParams);
3895

3896
 cleanup:
3897 3898
    VIR_FORCE_CLOSE(spec.dest.fd.qemu);
    VIR_FORCE_CLOSE(spec.dest.fd.local);
3899

3900
    virObjectUnref(cfg);
3901 3902 3903 3904
    return ret;
}


3905 3906 3907 3908
/* 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 */
3909 3910 3911 3912 3913 3914 3915 3916
static int
qemuMigrationSrcPerformPeer2Peer2(virQEMUDriverPtr driver,
                                  virConnectPtr sconn,
                                  virConnectPtr dconn,
                                  virDomainObjPtr vm,
                                  const char *dconnuri,
                                  unsigned long flags,
                                  const char *dname,
3917 3918
                                  unsigned long resource,
                                  qemuMigrationParamsPtr migParams)
3919 3920 3921
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
3922
    char *cookie = NULL;
3923 3924 3925
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
3926
    bool cancelled;
3927
    virStreamPtr st = NULL;
3928
    unsigned long destflags;
3929

3930
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
3931
              "flags=0x%lx, dname=%s, resource=%lu",
3932 3933
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
3934

3935 3936 3937 3938 3939
    /* 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,
3940 3941
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
3942 3943 3944 3945 3946
        return -1;

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

3947 3948
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
3949

3950 3951 3952 3953 3954 3955 3956 3957 3958 3959
    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;

3960
        qemuDomainObjEnterRemote(vm);
3961
        ret = dconn->driver->domainMigratePrepareTunnel
3962
            (dconn, st, destflags, dname, resource, dom_xml);
3963 3964
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3965
    } else {
3966
        qemuDomainObjEnterRemote(vm);
3967 3968
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
3969
             destflags, dname, resource, dom_xml);
3970 3971
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3972 3973 3974
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
3975 3976
        goto cleanup;

3977 3978
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
3979 3980
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
3981
        cancelled = true;
3982
        orig_err = virSaveLastError();
3983
        goto finish;
3984 3985
    }

3986 3987 3988 3989
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
3990
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
3991
    if (flags & VIR_MIGRATE_TUNNELLED)
3992 3993 3994
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, NULL,
                                            NULL, 0, NULL, NULL,
                                            flags, resource, dconn,
3995
                                            NULL, 0, NULL, migParams);
3996
    else
3997 3998 3999 4000
        ret = qemuMigrationSrcPerformNative(driver, vm, NULL, uri_out,
                                            cookie, cookielen,
                                            NULL, NULL, /* No out cookie with v2 migration */
                                            flags, resource, dconn, NULL, 0, NULL,
4001
                                            migParams);
4002 4003 4004 4005

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4006

4007 4008 4009
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4010
    cancelled = ret < 0;
4011

4012
 finish:
4013 4014 4015 4016
    /* 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.
     */
4017
    dname = dname ? dname : vm->def->name;
4018
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4019
    qemuDomainObjEnterRemote(vm);
4020
    ddomain = dconn->driver->domainMigrateFinish2
4021
        (dconn, dname, cookie, cookielen,
4022
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4023 4024
    /* The domain is already gone at this point */
    ignore_value(qemuDomainObjExitRemote(vm, false));
4025 4026
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4027

4028
 cleanup:
4029
    if (ddomain) {
4030
        virObjectUnref(ddomain);
4031 4032 4033 4034
        ret = 0;
    } else {
        ret = -1;
    }
4035

4036
    virObjectUnref(st);
4037 4038 4039 4040 4041 4042

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4043
    VIR_FREE(cookie);
4044 4045

    return ret;
4046 4047 4048
}


4049 4050 4051 4052
/* 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 */
4053
static int
4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067
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,
4068
                                  qemuMigrationParamsPtr migParams,
4069 4070 4071
                                  unsigned long long bandwidth,
                                  bool useParams,
                                  unsigned long flags)
4072 4073 4074 4075 4076 4077 4078 4079 4080 4081
{
    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;
4082
    bool cancelled = true;
4083
    virStreamPtr st = NULL;
4084
    unsigned long destflags;
4085 4086 4087
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4088
    size_t i;
4089
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4090 4091

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4092
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4093
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4094
              "bandwidth=%llu, useParams=%d, flags=0x%lx",
4095
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4096
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4097
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks, nbdPort,
4098
              bandwidth, useParams, flags);
4099

4100 4101 4102 4103 4104
    /* 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.  */

4105 4106 4107
    dom_xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                         &cookieout, &cookieoutlen,
                                         nmigrate_disks, migrate_disks, flags);
4108 4109 4110
    if (!dom_xml)
        goto cleanup;

4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130
    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;
4131 4132 4133 4134 4135 4136

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4137 4138 4139 4140 4141
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4142 4143 4144 4145 4146
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4147 4148 4149 4150 4151
        if (nbdPort &&
            virTypedParamsAddInt(&params, &nparams, &maxparams,
                                 VIR_MIGRATE_PARAM_DISKS_PORT,
                                 nbdPort) < 0)
            goto cleanup;
4152

4153 4154
        if (qemuMigrationParamsDump(migParams, &params, &nparams,
                                    &maxparams, &flags) < 0)
4155
            goto cleanup;
4156 4157
    }

4158 4159 4160
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4161 4162
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4163

4164 4165 4166 4167 4168 4169 4170 4171 4172
    VIR_DEBUG("Prepare3 %p", dconn);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
    if (flags & VIR_MIGRATE_TUNNELLED) {
        if (!(st = virStreamNew(dconn, 0)))
            goto cleanup;

4173
        qemuDomainObjEnterRemote(vm);
4174 4175 4176 4177 4178 4179 4180 4181 4182
        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);
        }
4183 4184
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4185
    } else {
4186
        qemuDomainObjEnterRemote(vm);
4187 4188 4189 4190 4191 4192 4193 4194 4195
        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);
        }
4196 4197
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4198 4199 4200 4201 4202
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

4203
    if (offline) {
L
liguang 已提交
4204 4205 4206
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4207
        cancelled = false;
L
liguang 已提交
4208 4209 4210
        goto finish;
    }

4211 4212 4213 4214
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4215 4216
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4217
            goto finish;
4218
        }
4219
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4220 4221
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4222
        orig_err = virSaveLastError();
4223 4224 4225 4226 4227 4228 4229 4230
        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.
     */
4231
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4232
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4233 4234 4235 4236 4237
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4238
    if (flags & VIR_MIGRATE_TUNNELLED) {
4239 4240 4241 4242
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, persist_xml,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4243
                                            nmigrate_disks, migrate_disks,
4244
                                            migParams);
4245
    } else {
4246 4247 4248 4249
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4250
                                            nmigrate_disks, migrate_disks,
4251
                                            migParams);
4252
    }
4253 4254

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4255
    if (ret < 0) {
4256
        orig_err = virSaveLastError();
4257 4258 4259 4260
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4261 4262 4263 4264

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4265
    cancelled = ret < 0;
4266

4267
 finish:
4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279
    /*
     * The status code from the source is passed to the destination.
     * The dest can cleanup in the source indicated it failed to
     * send all migration data. Returns NULL for ddomain if
     * the dest was unable to complete migration.
     */
    VIR_DEBUG("Finish3 %p ret=%d", dconn, ret);
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292

    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);
4293 4294
            if (qemuDomainObjExitRemote(vm, !offline) < 0)
                goto cleanup;
4295 4296 4297 4298 4299 4300 4301
        }
    } else {
        dname = dname ? dname : vm->def->name;
        qemuDomainObjEnterRemote(vm);
        ddomain = dconn->driver->domainMigrateFinish3
            (dconn, dname, cookiein, cookieinlen, &cookieout, &cookieoutlen,
             dconnuri, uri, destflags, cancelled);
4302 4303
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4304
    }
4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324

    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();
4325 4326
                if (err &&
                    err->domain == VIR_FROM_QEMU &&
4327 4328 4329 4330 4331 4332 4333
                    err->code != VIR_ERR_MIGRATE_FINISH_OK) {
                    virFreeError(orig_err);
                    orig_err = NULL;
                }
            }
        }
    }
4334

4335 4336 4337 4338 4339 4340 4341
    /* 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.
4342
     */
4343
    cancelled = ddomain == NULL;
4344

4345 4346 4347 4348 4349 4350
    /* 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();

4351 4352 4353 4354
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4355
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4356 4357 4358 4359 4360
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4361 4362 4363
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
4364 4365 4366 4367
    /* 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.
     */
4368 4369 4370
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4371 4372 4373

 cleanup:
    if (ddomain) {
4374
        virObjectUnref(ddomain);
4375 4376 4377 4378 4379
        ret = 0;
    } else {
        ret = -1;
    }

4380
    virObjectUnref(st);
4381 4382 4383 4384 4385 4386 4387 4388

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4389
    virTypedParamsFree(params, nparams);
4390 4391 4392 4393
    return ret;
}


4394
static void
4395 4396 4397
qemuMigrationSrcConnectionClosed(virConnectPtr conn,
                                 int reason,
                                 void *opaque)
4398 4399 4400 4401 4402 4403 4404 4405
{
    virDomainObjPtr vm = opaque;

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


4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430
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,
4431
                                 qemuMigrationParamsPtr migParams,
4432 4433 4434 4435
                                 unsigned long flags,
                                 const char *dname,
                                 unsigned long resource,
                                 bool *v3proto)
4436 4437 4438 4439
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4440
    virErrorPtr orig_err = NULL;
4441
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4442
    bool dstOffline = false;
4443
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4444
    bool useParams;
4445

4446 4447
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, uri=%s, "
              "graphicsuri=%s, listenAddress=%s, nmigrate_disks=%zu, "
4448
              "migrate_disks=%p, nbdPort=%d, flags=0x%lx, dname=%s, "
4449
              "resource=%lu",
4450
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4451
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4452 4453
              nmigrate_disks, migrate_disks, nbdPort, flags, NULLSTR(dname),
              resource);
4454

4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468
    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;
    }

4469 4470 4471 4472 4473 4474 4475
    if (flags & VIR_MIGRATE_TUNNELLED && nbdPort) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("disk port address is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

4476 4477 4478 4479
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

4480
    qemuDomainObjEnterRemote(vm);
4481
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4482 4483 4484
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;

4485
    if (dconn == NULL) {
4486
        virReportError(VIR_ERR_OPERATION_FAILED,
4487 4488
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4489
        virObjectUnref(cfg);
4490 4491 4492
        return -1;
    }

4493 4494
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4495 4496
        goto cleanup;

4497
    if (virConnectRegisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed,
4498 4499 4500 4501
                                        vm, NULL) < 0) {
        goto cleanup;
    }

4502
    qemuDomainObjEnterRemote(vm);
4503 4504
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4505
        /* v3proto reflects whether the caller used Perform3, but with
4506
         * p2p migrate, regardless of whether Perform2 or Perform3
4507 4508 4509 4510
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4511 4512
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
4513
    if (offline)
4514 4515
        dstOffline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                              VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4516 4517
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;
4518

4519
    if (!p2p) {
4520 4521
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4522 4523 4524
        goto cleanup;
    }

4525 4526
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4527
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
4528 4529 4530 4531 4532 4533
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("Migration APIs with extensible parameters are not "
                         "supported but extended parameters were passed"));
        goto cleanup;
    }

4534
    if (offline && !dstOffline) {
L
liguang 已提交
4535 4536 4537 4538 4539 4540
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4541 4542 4543 4544 4545 4546 4547
    /* 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;

4548
    if (*v3proto) {
4549 4550 4551
        ret = qemuMigrationSrcPerformPeer2Peer3(driver, sconn, dconn, dconnuri, vm, xmlin,
                                                persist_xml, dname, uri, graphicsuri,
                                                listenAddress, nmigrate_disks, migrate_disks,
4552
                                                nbdPort, migParams, resource,
4553
                                                useParams, flags);
4554
    } else {
4555
        ret = qemuMigrationSrcPerformPeer2Peer2(driver, sconn, dconn, vm,
4556
                                                dconnuri, flags, dname, resource,
4557
                                                migParams);
4558
    }
4559

4560
 cleanup:
4561
    orig_err = virSaveLastError();
4562
    qemuDomainObjEnterRemote(vm);
4563
    virConnectUnregisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed);
4564
    virObjectUnref(dconn);
4565
    ignore_value(qemuDomainObjExitRemote(vm, false));
4566 4567 4568 4569
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4570
    virObjectUnref(cfg);
4571 4572 4573 4574
    return ret;
}


4575 4576 4577 4578 4579 4580
/*
 * 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
4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592
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,
4593
                           qemuMigrationParamsPtr migParams,
4594 4595 4596 4597 4598 4599 4600 4601
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
                           const char *dname,
                           unsigned long resource,
                           bool v3proto)
4602
{
4603
    virObjectEventPtr event = NULL;
4604
    int ret = -1;
4605
    virErrorPtr orig_err = NULL;
4606
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4607
    qemuDomainObjPrivatePtr priv = vm->privateData;
4608

4609 4610
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                              flags) < 0)
4611 4612
        goto cleanup;

4613
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
4614 4615
        goto endjob;

4616
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
4617
        goto endjob;
4618

4619
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
4620
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
4621
        goto endjob;
4622

4623
    qemuMigrationSrcStoreDomainState(vm);
4624 4625

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4626 4627 4628
        ret = qemuMigrationSrcPerformPeer2Peer(driver, conn, vm, xmlin, persist_xml,
                                               dconnuri, uri, graphicsuri, listenAddress,
                                               nmigrate_disks, migrate_disks, nbdPort,
4629
                                               migParams, flags, dname, resource,
4630
                                               &v3proto);
4631
    } else {
4632
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4633 4634 4635
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                            cookieout, cookieoutlen,
                                            flags, resource, NULL, NULL, 0, NULL,
4636
                                            migParams);
4637
    }
4638 4639
    if (ret < 0)
        goto endjob;
4640

4641 4642 4643 4644
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4645
    if (!v3proto) {
4646
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
4647
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4648
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4649
        virDomainAuditStop(vm, "migrated");
4650
        event = virDomainEventLifecycleNewFromObj(vm,
4651 4652
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4653 4654
    }

4655
 endjob:
4656 4657 4658
    if (ret < 0)
        orig_err = virSaveLastError();

4659 4660 4661 4662
    /* v2 proto has no confirm phase so we need to reset migration parameters
     * here
     */
    if (!v3proto && ret < 0)
4663
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4664
                                 priv->job.migParams, priv->job.apiFlags);
4665

4666
    qemuMigrationSrcRestoreDomainState(driver, vm);
4667

4668
    qemuMigrationJobFinish(driver, vm);
4669
    if (!virDomainObjIsActive(vm) && ret == 0) {
4670
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
4671
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4672 4673
            vm->persistent = 0;
        }
4674
        qemuDomainRemoveInactiveJob(driver, vm);
4675 4676
    }

4677 4678 4679 4680 4681
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4682
 cleanup:
M
Michal Privoznik 已提交
4683
    virDomainObjEndAPI(&vm);
4684
    virObjectEventStateQueue(driver->domainEventState, event);
4685
    virObjectUnref(cfg);
4686 4687 4688 4689 4690 4691 4692
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4693 4694 4695 4696 4697 4698 4699 4700
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,
4701
                             qemuMigrationParamsPtr migParams,
4702 4703 4704 4705 4706 4707
                             const char *cookiein,
                             int cookieinlen,
                             char **cookieout,
                             int *cookieoutlen,
                             unsigned long flags,
                             unsigned long resource)
4708
{
4709
    qemuDomainObjPrivatePtr priv = vm->privateData;
4710 4711 4712 4713
    int ret = -1;

    /* If we didn't start the job in the begin phase, start it now. */
    if (!(flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
4714 4715
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
4716 4717 4718 4719 4720 4721
            goto cleanup;
    } else if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) {
        goto cleanup;
    }

    qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4722
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
4723
                           qemuMigrationSrcCleanup);
4724

4725 4726 4727
    ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                        cookieout, cookieoutlen,
                                        flags, resource, NULL, graphicsuri,
4728
                                        nmigrate_disks, migrate_disks, migParams);
4729

4730
    if (ret < 0) {
4731
        qemuMigrationSrcRestoreDomainState(driver, vm);
4732
        goto endjob;
4733
    }
4734 4735 4736

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4737
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
4738
                             qemuMigrationSrcCleanup) < 0)
4739 4740
        goto endjob;

4741
 endjob:
4742
    if (ret < 0) {
4743
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4744
                                 priv->job.migParams, priv->job.apiFlags);
4745
        qemuMigrationJobFinish(driver, vm);
4746
    } else {
4747
        qemuMigrationJobContinue(vm);
4748 4749
    }

4750
    if (!virDomainObjIsActive(vm))
4751
        qemuDomainRemoveInactiveJob(driver, vm);
4752

4753
 cleanup:
M
Michal Privoznik 已提交
4754
    virDomainObjEndAPI(&vm);
4755 4756 4757
    return ret;
}

4758
int
4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770
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,
4771
                        qemuMigrationParamsPtr migParams,
4772 4773 4774 4775 4776 4777 4778 4779
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4780 4781
{
    VIR_DEBUG("driver=%p, conn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4782
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
4783
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4784
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
4785
              "flags=0x%lx, dname=%s, resource=%lu, v3proto=%d",
4786
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4787
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4788 4789 4790
              nmigrate_disks, migrate_disks, nbdPort,
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4791 4792 4793

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4794 4795
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4796 4797 4798
            return -1;
        }

4799 4800 4801
        return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, dconnuri, uri,
                                          graphicsuri, listenAddress,
                                          nmigrate_disks, migrate_disks, nbdPort,
4802
                                          migParams,
4803 4804 4805
                                          cookiein, cookieinlen,
                                          cookieout, cookieoutlen,
                                          flags, dname, resource, v3proto);
4806 4807
    } else {
        if (dconnuri) {
4808 4809
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
4810 4811 4812 4813
            return -1;
        }

        if (v3proto) {
4814 4815 4816
            return qemuMigrationSrcPerformPhase(driver, conn, vm, persist_xml, uri,
                                                graphicsuri,
                                                nmigrate_disks, migrate_disks,
4817
                                                migParams,
4818 4819 4820
                                                cookiein, cookieinlen,
                                                cookieout, cookieoutlen,
                                                flags, resource);
4821
        } else {
4822 4823 4824
            return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, NULL,
                                              uri, graphicsuri, listenAddress,
                                              nmigrate_disks, migrate_disks, nbdPort,
4825
                                              migParams,
4826 4827 4828
                                              cookiein, cookieinlen,
                                              cookieout, cookieoutlen, flags,
                                              dname, resource, v3proto);
4829 4830 4831
        }
    }
}
4832

4833
static int
4834
qemuMigrationDstVPAssociatePortProfiles(virDomainDefPtr def)
4835
{
4836
    size_t i;
4837 4838 4839 4840 4841
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
4842
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4843
            if (virNetDevVPortProfileAssociate(net->ifname,
4844
                                               virDomainNetGetActualVirtPortProfile(net),
4845
                                               &net->mac,
4846
                                               virDomainNetGetActualDirectDev(net),
4847
                                               -1,
4848
                                               def->uuid,
4849 4850
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
4851 4852 4853
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
4854
                goto err_exit;
4855
            }
4856
            last_good_net = i;
4857
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
4858

4859
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
4860 4861 4862 4863
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
4864 4865 4866
        }
    }

4867
    return 0;
4868

4869
 err_exit:
4870
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
4871
        net = def->nets[i];
4872
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4873
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
4874
                                                           virDomainNetGetActualVirtPortProfile(net),
4875
                                                           &net->mac,
4876
                                                           virDomainNetGetActualDirectDev(net),
4877
                                                           -1,
4878
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
4879 4880
        }
    }
4881
    return -1;
4882 4883 4884
}


J
Jiri Denemark 已提交
4885
static int
4886 4887 4888 4889
qemuMigrationDstPersist(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationCookiePtr mig,
                        bool ignoreSaveError)
J
Jiri Denemark 已提交
4890 4891 4892 4893
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    virDomainDefPtr vmdef;
4894 4895
    virDomainDefPtr oldDef = NULL;
    unsigned int oldPersist = vm->persistent;
J
Jiri Denemark 已提交
4896 4897 4898 4899 4900 4901 4902
    virObjectEventPtr event;
    int ret = -1;

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

    vm->persistent = 1;
4903 4904
    oldDef = vm->newDef;
    vm->newDef = qemuMigrationCookieGetPersistent(mig);
J
Jiri Denemark 已提交
4905

4906 4907
    if (!(vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm)))
        goto error;
J
Jiri Denemark 已提交
4908

4909 4910
    if (virDomainSaveConfig(cfg->configDir, driver->caps, vmdef) < 0 &&
        !ignoreSaveError)
4911
        goto error;
J
Jiri Denemark 已提交
4912 4913 4914

    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_DEFINED,
4915 4916 4917
                                              oldPersist ?
                                              VIR_DOMAIN_EVENT_DEFINED_UPDATED :
                                              VIR_DOMAIN_EVENT_DEFINED_ADDED);
4918
    virObjectEventStateQueue(driver->domainEventState, event);
J
Jiri Denemark 已提交
4919 4920 4921 4922

    ret = 0;

 cleanup:
4923
    virDomainDefFree(oldDef);
J
Jiri Denemark 已提交
4924 4925 4926
    virObjectUnref(caps);
    virObjectUnref(cfg);
    return ret;
4927 4928 4929 4930 4931 4932 4933

 error:
    virDomainDefFree(vm->newDef);
    vm->persistent = oldPersist;
    vm->newDef = oldDef;
    oldDef = NULL;
    goto cleanup;
J
Jiri Denemark 已提交
4934 4935 4936
}


4937
virDomainPtr
4938 4939 4940 4941 4942 4943 4944 4945 4946 4947
qemuMigrationDstFinish(virQEMUDriverPtr driver,
                       virConnectPtr dconn,
                       virDomainObjPtr vm,
                       const char *cookiein,
                       int cookieinlen,
                       char **cookieout,
                       int *cookieoutlen,
                       unsigned long flags,
                       int retcode,
                       bool v3proto)
4948 4949
{
    virDomainPtr dom = NULL;
4950
    qemuMigrationCookiePtr mig = NULL;
4951
    virErrorPtr orig_err = NULL;
4952
    int cookie_flags = 0;
J
Jiri Denemark 已提交
4953
    qemuDomainObjPrivatePtr priv = vm->privateData;
4954
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4955
    unsigned short port;
4956 4957
    unsigned long long timeReceived = 0;
    virObjectEventPtr event;
4958
    qemuDomainJobInfoPtr jobInfo = NULL;
4959
    bool inPostCopy = false;
4960
    bool doKill = true;
4961

4962
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
4963
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, retcode=%d",
4964 4965
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
4966

4967 4968 4969
    port = priv->migrationPort;
    priv->migrationPort = 0;

4970
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN)) {
4971
        qemuMigrationDstErrorReport(driver, vm->def->name);
4972
        goto cleanup;
4973
    }
4974

4975 4976
    ignore_value(virTimeMillisNow(&timeReceived));

4977 4978 4979
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
4980

4981
    qemuDomainCleanupRemove(vm, qemuMigrationDstPrepareCleanup);
4982
    VIR_FREE(priv->job.completed);
4983

4984
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
4985 4986
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
4987 4988 4989
    /* 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;
4990

4991 4992
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen, cookie_flags)))
4993
        goto endjob;
4994

4995
    if (flags & VIR_MIGRATE_OFFLINE) {
4996
        if (retcode == 0 &&
4997
            qemuMigrationDstPersist(driver, vm, mig, false) == 0)
4998
            dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, -1);
4999 5000
        goto endjob;
    }
5001

5002 5003 5004 5005
    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
         */
5006
        qemuDomainCheckMonitor(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN);
5007 5008
        goto endjob;
    }
5009

5010 5011 5012
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
5013
        qemuMigrationDstErrorReport(driver, vm->def->name);
5014 5015
        goto endjob;
    }
5016

5017
    if (qemuMigrationDstVPAssociatePortProfiles(vm->def) < 0)
5018
        goto endjob;
5019

5020
    if (mig->network && qemuMigrationDstOPDRelocate(driver, vm, mig) < 0)
5021
        VIR_WARN("unable to provide network data for relocation");
5022

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

5026 5027
    if (qemuRefreshVirtioChannelState(driver, vm,
                                      QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
5028 5029
        goto endjob;

5030 5031
    if (qemuConnectAgent(driver, vm) < 0)
        goto endjob;
5032

5033
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
5034
        if (qemuMigrationDstPersist(driver, vm, mig, !v3proto) < 0) {
5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045
            /* 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.
5046 5047 5048
             */
            if (v3proto)
                goto endjob;
5049
        }
5050
    }
5051

5052 5053 5054
    /* We need to wait for QEMU to process all data sent by the source
     * before starting guest CPUs.
     */
5055 5056 5057
    if (qemuMigrationDstWaitForCompletion(driver, vm,
                                          QEMU_ASYNC_JOB_MIGRATION_IN,
                                          !!(flags & VIR_MIGRATE_POSTCOPY)) < 0) {
5058 5059 5060 5061 5062 5063
        /* 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;
    }
5064

5065 5066 5067 5068 5069 5070 5071 5072 5073
    /* 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;
    }

5074
    if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY)
5075 5076
        inPostCopy = true;

5077 5078 5079 5080 5081
    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
         */
5082
        if (qemuProcessStartCPUs(driver, vm,
5083 5084
                                 inPostCopy ? VIR_DOMAIN_RUNNING_POSTCOPY
                                            : VIR_DOMAIN_RUNNING_MIGRATED,
5085
                                 QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5086
            if (virGetLastErrorCode() == VIR_ERR_OK)
5087 5088 5089 5090 5091 5092
                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();
5093

5094 5095 5096 5097 5098 5099 5100 5101 5102 5103
            /*
             * 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;
5104
        }
5105

5106
        if (inPostCopy)
5107
            doKill = false;
5108
    }
5109

5110
    if (mig->jobInfo) {
5111
        jobInfo = mig->jobInfo;
5112 5113 5114 5115 5116 5117
        mig->jobInfo = NULL;

        if (jobInfo->sent && timeReceived) {
            jobInfo->timeDelta = timeReceived - jobInfo->sent;
            jobInfo->received = timeReceived;
            jobInfo->timeDeltaSet = true;
5118
        }
5119 5120
        qemuDomainJobInfoUpdateTime(jobInfo);
        qemuDomainJobInfoUpdateDowntime(jobInfo);
5121
    }
L
liguang 已提交
5122

5123
    if (inPostCopy) {
5124 5125 5126
        if (qemuMigrationDstWaitForCompletion(driver, vm,
                                              QEMU_ASYNC_JOB_MIGRATION_IN,
                                              false) < 0) {
5127 5128 5129 5130 5131 5132 5133 5134 5135
            goto endjob;
        }
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            virDomainObjSetState(vm,
                                 VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_MIGRATED);
        }
    }

5136
    dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, vm->def->id);
5137

5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151
    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);
    }
5152

5153 5154 5155 5156 5157
    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);
5158
        virObjectEventStateQueue(driver->domainEventState, event);
5159
    }
5160

5161
    if (virDomainObjIsActive(vm) &&
5162
        virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
5163 5164 5165 5166 5167
        VIR_WARN("Failed to save status on vm %s", vm->def->name);

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

5168
 endjob:
5169
    if (!dom &&
5170 5171
        !(flags & VIR_MIGRATE_OFFLINE) &&
        virDomainObjIsActive(vm)) {
5172
        if (doKill) {
5173 5174 5175 5176 5177 5178 5179
            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);
5180
            virObjectEventStateQueue(driver->domainEventState, event);
5181
        } else {
5182
            qemuMigrationAnyPostcopyFailed(driver, vm);
5183
        }
5184 5185
    }

5186
    if (dom) {
5187 5188 5189
        if (jobInfo) {
            VIR_STEAL_PTR(priv->job.completed, jobInfo);
            priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
5190
            priv->job.completed->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;
5191
        }
5192

5193 5194 5195
        if (qemuMigrationBakeCookie(mig, driver, vm,
                                    QEMU_MIGRATION_DESTINATION,
                                    cookieout, cookieoutlen,
5196 5197
                                    QEMU_MIGRATION_COOKIE_STATS) < 0)
            VIR_WARN("Unable to encode migration cookie");
5198 5199 5200 5201 5202 5203

        /* Remove completed stats for post-copy, everything but timing fields
         * is obsolete anyway.
         */
        if (inPostCopy)
            VIR_FREE(priv->job.completed);
5204
    }
5205

5206
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
5207
                             priv->job.migParams, priv->job.apiFlags);
5208

5209
    qemuMigrationJobFinish(driver, vm);
5210
    if (!virDomainObjIsActive(vm))
5211
        qemuDomainRemoveInactiveJob(driver, vm);
5212

5213
 cleanup:
5214
    VIR_FREE(jobInfo);
5215
    virPortAllocatorRelease(port);
5216
    if (priv->mon)
5217
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
5218
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5219
    virDomainObjEndAPI(&vm);
5220
    qemuMigrationCookieFree(mig);
5221 5222 5223 5224
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5225
    virObjectUnref(cfg);
5226 5227 5228 5229

    /* Set a special error if Finish is expected to return NULL as a result of
     * successful call with retcode != 0
     */
5230
    if (retcode != 0 && !dom && virGetLastErrorCode() == VIR_ERR_OK)
5231
        virReportError(VIR_ERR_MIGRATE_FINISH_OK, NULL);
5232 5233
    return dom;
}
5234

5235

5236
/* Helper function called while vm is active.  */
5237
int
5238 5239 5240 5241
qemuMigrationSrcToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
                       int fd,
                       const char *compressor,
                       qemuDomainAsyncJob asyncJob)
5242 5243 5244
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5245
    int ret = -1;
5246 5247
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5248
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5249
    char *errbuf = NULL;
5250
    virErrorPtr orig_err = NULL;
5251 5252 5253 5254 5255

    /* 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,
5256 5257
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5258 5259
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5260
    }
5261

5262 5263 5264 5265 5266 5267 5268
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5269 5270 5271 5272
    if (compressor && pipe(pipeFD) < 0) {
        virReportSystemError(errno, "%s",
                             _("Failed to create pipe for migration"));
        return -1;
5273 5274
    }

5275 5276 5277 5278
    /* 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.  */
5279 5280
    if (qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    compressor ? pipeFD[1] : fd) < 0)
5281 5282
        goto cleanup;

5283
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5284 5285
        goto cleanup;

5286
    if (!compressor) {
5287 5288 5289
        rc = qemuMonitorMigrateToFd(priv->mon,
                                    QEMU_MONITOR_MIGRATE_BACKGROUND,
                                    fd);
5290 5291 5292 5293 5294 5295 5296
    } else {
        const char *prog = compressor;
        const char *args[] = {
            prog,
            "-c",
            NULL
        };
5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307

        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;
5308
        }
5309 5310 5311 5312 5313 5314 5315 5316 5317 5318
        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");
5319
    }
5320
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5321
        goto cleanup;
5322 5323 5324
    if (rc < 0)
        goto cleanup;

5325
    rc = qemuMigrationSrcWaitForCompletion(driver, vm, asyncJob, NULL, 0);
5326

5327 5328 5329 5330
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5331 5332
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5333
                qemuMonitorMigrateCancel(priv->mon);
5334
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5335 5336
            }
        }
5337
        goto cleanup;
5338
    }
5339

5340 5341 5342
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5343
    qemuDomainEventEmitJobCompleted(driver, vm);
5344 5345
    ret = 0;

5346
 cleanup:
5347 5348 5349
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5350
    /* Restore max migration bandwidth */
5351 5352
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5353 5354
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5355
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5356 5357
    }

5358 5359
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5360 5361 5362 5363 5364
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5365 5366 5367 5368 5369 5370

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

5371 5372
    return ret;
}
5373

5374 5375

int
5376 5377
qemuMigrationSrcCancel(virQEMUDriverPtr driver,
                       virDomainObjPtr vm)
5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388
{
    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));
5389
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5390 5391 5392 5393 5394
        goto cleanup;

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
5395
        qemuBlockJobDataPtr job;
5396

5397 5398
        if (!(job = qemuBlockJobDiskGetJob(disk)) ||
            !job->started)
5399
            diskPriv->migrating = false;
5400 5401

        if (diskPriv->migrating) {
5402
            qemuBlockJobSyncBegin(job);
5403
            storage = true;
5404 5405
        }

5406 5407
        virObjectUnref(job);
    }
5408 5409 5410

    if (storage &&
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
5411
                                      QEMU_ASYNC_JOB_NONE, NULL) < 0)
5412
        goto cleanup;
5413 5414 5415 5416 5417 5418 5419 5420

    ret = 0;

 cleanup:
    return ret;
}


5421
static int
5422
qemuMigrationJobStart(virQEMUDriverPtr driver,
5423
                      virDomainObjPtr vm,
5424 5425
                      qemuDomainAsyncJob job,
                      unsigned long apiFlags)
5426
{
5427
    qemuDomainObjPrivatePtr priv = vm->privateData;
5428 5429
    virDomainJobOperation op;
    unsigned long long mask;
5430

5431
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5432 5433
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_IN;
        mask = QEMU_JOB_NONE;
5434
    } else {
5435 5436 5437 5438
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_OUT;
        mask = QEMU_JOB_DEFAULT_MASK |
               JOB_MASK(QEMU_JOB_SUSPEND) |
               JOB_MASK(QEMU_JOB_MIGRATION_OP);
5439
    }
5440

5441
    if (qemuDomainObjBeginAsyncJob(driver, vm, job, op, apiFlags) < 0)
5442 5443
        return -1;

5444 5445
    priv->job.current->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;

5446
    qemuDomainObjSetAsyncJobMask(vm, mask);
5447 5448 5449
    return 0;
}

5450
static void
5451
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5452
                         virDomainObjPtr vm,
5453
                         qemuMigrationJobPhase phase)
5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466
{
    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);
}

5467
static void
5468
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5469
                           virDomainObjPtr vm,
5470
                           qemuMigrationJobPhase phase)
5471 5472 5473 5474
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5475
static void
5476 5477
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5478
    qemuDomainObjReleaseAsyncJob(vm);
5479 5480
}

5481
static bool
5482
qemuMigrationJobIsActive(virDomainObjPtr vm,
5483
                         qemuDomainAsyncJob job)
5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494
{
    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");

5495
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5496 5497 5498 5499 5500
        return false;
    }
    return true;
}

5501
static void
5502
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5503
{
5504
    qemuDomainObjEndAsyncJob(driver, vm);
5505
}
5506 5507 5508


static void
5509
qemuMigrationDstErrorFree(void *data,
5510 5511 5512 5513 5514 5515 5516
                       const void *name ATTRIBUTE_UNUSED)
{
    virErrorPtr err = data;
    virFreeError(err);
}

int
5517
qemuMigrationDstErrorInit(virQEMUDriverPtr driver)
5518
{
5519
    driver->migrationErrors = virHashAtomicNew(64, qemuMigrationDstErrorFree);
5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530
    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
5531 5532 5533
qemuMigrationDstErrorSave(virQEMUDriverPtr driver,
                          const char *name,
                          virErrorPtr err)
5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546
{
    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
5547 5548
qemuMigrationDstErrorReport(virQEMUDriverPtr driver,
                            const char *name)
5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559
{
    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);
}
5560 5561


5562
int
5563 5564 5565 5566
qemuMigrationSrcFetchMirrorStats(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 qemuDomainJobInfoPtr jobInfo)
5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610
{
    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;
}