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 = qemuBlockJobUpdate(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 = qemuBlockJobUpdate(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
            qemuBlockJobSyncEnd(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 = qemuBlockJobUpdate(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
 *
680 681
 * Cancel all drive-mirrors started by qemuMigrationSrcNBDStorageCopy.
 * Any pending block job events for the affected disks will be processed.
682 683 684 685
 *
 * Returns 0 on success, -1 otherwise.
 */
static int
686 687 688 689 690
qemuMigrationSrcNBDCopyCancel(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              bool check,
                              qemuDomainAsyncJob asyncJob,
                              virConnectPtr dconn)
691
{
692
    virErrorPtr err = NULL;
693
    int ret = -1;
694
    size_t i;
695 696 697 698
    int rv;
    bool failed = false;

    VIR_DEBUG("Cancelling drive mirrors for domain %s", vm->def->name);
699 700 701

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
702
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
703

704
        if (!diskPriv->migrating)
705 706
            continue;

707 708
        rv = qemuMigrationSrcNBDCopyCancelOne(driver, vm, disk,
                                              check, asyncJob);
709 710 711 712 713 714
        if (rv != 0) {
            if (rv < 0) {
                if (!err)
                    err = virSaveLastError();
                failed = true;
            }
715
            qemuBlockJobSyncEnd(vm, asyncJob, disk);
716 717 718 719
            diskPriv->migrating = false;
        }
    }

720
    while ((rv = qemuMigrationSrcNBDCopyCancelled(vm, asyncJob, check)) != 1) {
721 722 723 724 725 726 727
        if (check && !failed &&
            dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            failed = true;
        }

728 729 730 731
        if (rv < 0) {
            failed = true;
            if (rv == -2)
                break;
732
        }
733

734 735 736 737 738
        if (failed && !err)
            err = virSaveLastError();

        if (virDomainObjWait(vm) < 0)
            goto cleanup;
739 740
    }

741 742 743 744 745 746 747 748 749 750 751 752 753
    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;
    }

754 755 756
    ret = failed ? -1 : 0;

 cleanup:
757 758 759 760 761
    if (err) {
        virSetError(err);
        virFreeError(err);
    }
    return ret;
762 763 764
}


765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843
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;
}


844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885
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;
}


886
/**
887
 * qemuMigrationSrcNBDStorageCopy:
888 889 890 891
 * @driver: qemu driver
 * @vm: domain
 * @mig: migration cookie
 * @host: where are we migrating to
892
 * @speed: bandwidth limit in MiB/s
893 894
 * @migrate_flags: migrate monitor command flags
 *
895 896 897
 * 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
898
 * to do the very same operation. On failure, the caller is
899 900
 * expected to call qemuMigrationSrcNBDCopyCancel to stop all
 * running copy operations.
901 902 903 904 905
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
906 907 908 909 910 911 912 913
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,
914 915 916
                               virConnectPtr dconn,
                               const char *tlsAlias,
                               unsigned int flags)
917 918 919 920
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int port;
921
    size_t i;
922
    char *diskAlias = NULL;
923
    unsigned long long mirror_speed = speed;
924
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
925
    int rv;
926
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
927 928

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
929

930 931 932 933 934 935 936 937
    if (mirror_speed > LLONG_MAX >> 20) {
        virReportError(VIR_ERR_OVERFLOW,
                       _("bandwidth must be less than %llu"),
                       LLONG_MAX >> 20);
        goto cleanup;
    }
    mirror_speed <<= 20;

938 939 940 941 942 943 944 945 946
    /* 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];
947
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
948
        int rc;
949

950
        /* check whether disk should be migrated */
951
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
952 953
            continue;

954
        if (!(diskAlias = qemuAliasDiskDriveFromDisk(disk)))
955 956
            goto cleanup;

957 958
        qemuBlockJobSyncBegin(disk);

959 960 961 962 963 964 965 966 967 968 969 970 971 972 973
        if (flags & VIR_MIGRATE_TLS) {
            rc = qemuMigrationSrcNBDStorageCopyBlockdev(driver, vm,
                                                        disk, diskAlias,
                                                        host, port,
                                                        mirror_speed,
                                                        mirror_flags,
                                                        tlsAlias);
        } else {
            rc = qemuMigrationSrcNBDStorageCopyDriveMirror(driver, vm, diskAlias,
                                                           host, port,
                                                           mirror_speed,
                                                           mirror_flags);
        }

        if (rc < 0) {
974
            qemuBlockJobSyncEnd(vm, QEMU_ASYNC_JOB_MIGRATION_OUT, disk);
975 976
            goto cleanup;
        }
977 978

        VIR_FREE(diskAlias);
979
        diskPriv->migrating = true;
980

981
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
982 983 984
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto cleanup;
        }
985
    }
986

987
    while ((rv = qemuMigrationSrcNBDStorageCopyReady(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) != 1) {
988 989
        if (rv < 0)
            goto cleanup;
990

991
        if (priv->job.abortJob) {
992
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
993 994 995 996
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
            goto cleanup;
997
        }
998

999 1000 1001 1002 1003 1004
        if (dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            goto cleanup;
        }

1005
        if (virDomainObjWait(vm) < 0)
1006
            goto cleanup;
1007 1008
    }

1009 1010
    qemuMigrationSrcFetchMirrorStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                     priv->job.current);
1011

1012
    /* Okay, all disks are ready. Modify migrate_flags */
1013 1014 1015 1016
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

1017
 cleanup:
1018
    virObjectUnref(cfg);
1019 1020 1021
    VIR_FREE(diskAlias);
    return ret;
}
1022

1023

1024
/**
1025
 * qemuMigrationSrcIsAllowedHostdev:
1026 1027 1028 1029 1030 1031
 * @def: domain definition
 *
 * Checks that @def does not contain any host devices unsupported accross
 * migrations. Returns true if the vm is allowed to migrate.
 */
static bool
1032
qemuMigrationSrcIsAllowedHostdev(const virDomainDef *def)
1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
{
    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;
}


1052
/**
1053
 * qemuMigrationSrcIsAllowed:
1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065
 * @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.
 */
1066
bool
1067 1068 1069 1070
qemuMigrationSrcIsAllowed(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          bool remote,
                          unsigned int flags)
1071
{
1072
    int nsnapshots;
1073
    int pauseReason;
1074
    size_t i;
1075

1076 1077 1078 1079 1080
    /* perform these checks only when migrating to remote hosts */
    if (remote) {
        nsnapshots = virDomainSnapshotObjListNum(vm->snapshots, NULL, 0);
        if (nsnapshots < 0)
            return false;
1081

1082 1083 1084 1085 1086
        if (nsnapshots > 0) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("cannot migrate domain with %d snapshots"),
                           nsnapshots);
            return false;
1087
        }
1088

1089
        /* cancel migration if disk I/O error is emitted while migrating */
1090
        if (flags & VIR_MIGRATE_ABORT_ON_ERROR &&
1091
            !(flags & VIR_MIGRATE_OFFLINE) &&
1092 1093
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
E
Eric Blake 已提交
1094
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1095
                           _("cannot migrate domain with I/O error"));
E
Eric Blake 已提交
1096 1097
            return false;
        }
1098 1099

    }
1100

1101 1102 1103 1104 1105 1106 1107
    /* 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;
        }
1108

1109

1110 1111
        if (qemuDomainHasBlockjob(vm, false)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1112
                           _("domain has active block job"));
1113 1114 1115
            return false;
        }

1116
        if (!qemuMigrationSrcIsAllowedHostdev(vm->def))
1117 1118
            return false;

1119
        if (vm->def->cpu) {
1120 1121 1122
            /* QEMU blocks migration and save with invariant TSC enabled
             * unless TSC frequency is explicitly set.
             */
1123 1124
            if (virCPUCheckFeature(vm->def->os.arch, vm->def->cpu,
                                   "invtsc") == 1) {
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
                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;
                }
1143
            }
J
Ján Tomko 已提交
1144 1145
        }

1146 1147 1148
        /* Verify that memory device config can be transferred reliably */
        for (i = 0; i < vm->def->nmems; i++) {
            virDomainMemoryDefPtr mem = vm->def->mems[i];
1149

1150 1151 1152 1153 1154
            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"));
1155

1156 1157
                return false;
            }
1158
        }
1159 1160 1161 1162 1163 1164

        if (vm->def->nshmems) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("migration with shmem device is not supported"));
            return false;
        }
1165 1166
    }

1167 1168 1169
    return true;
}

1170
static bool
1171 1172 1173 1174
qemuMigrationSrcIsSafe(virDomainDefPtr def,
                       size_t nmigrate_disks,
                       const char **migrate_disks,
                       unsigned int flags)
1175

1176
{
1177 1178
    bool storagemigration = flags & (VIR_MIGRATE_NON_SHARED_DISK |
                                     VIR_MIGRATE_NON_SHARED_INC);
1179
    size_t i;
1180
    int rc;
1181

1182
    for (i = 0; i < def->ndisks; i++) {
1183
        virDomainDiskDefPtr disk = def->disks[i];
1184
        const char *src = virDomainDiskGetSource(disk);
1185

1186 1187
        /* Disks without any source (i.e. floppies and CD-ROMs)
         * OR readonly are safe. */
1188
        if (virStorageSourceIsEmpty(disk->src) ||
1189
            disk->src->readonly)
1190
            continue;
1191

1192
        /* Disks which are migrated by qemu are safe too. */
1193
        if (storagemigration &&
1194
            qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1195
            continue;
1196

1197
        /* However, disks on local FS (e.g. ext4) are not safe. */
1198
        if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
1199
            if ((rc = virFileIsSharedFS(src)) < 0) {
1200
                return false;
1201 1202 1203 1204 1205
            } else if (rc == 0) {
                virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                               _("Migration without shared storage is unsafe"));
                return false;
            }
1206 1207 1208 1209 1210 1211 1212
            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;
1213
        }
1214

1215 1216 1217 1218 1219 1220 1221
        /* 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;

1222 1223
        virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                       _("Migration may lead to data corruption if disks"
1224
                         " use cache != none or cache != directsync"));
1225
        return false;
1226 1227 1228 1229 1230
    }

    return true;
}

1231
/** qemuMigrationSrcSetOffline
1232 1233 1234
 * Pause domain for non-live migration.
 */
int
1235 1236
qemuMigrationSrcSetOffline(virQEMUDriverPtr driver,
                           virDomainObjPtr vm)
1237 1238
{
    int ret;
1239
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
1240 1241
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
1242
    if (ret == 0) {
1243
        virObjectEventPtr event;
1244

1245
        event = virDomainEventLifecycleNewFromObj(vm,
1246 1247
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
1248
        virObjectEventStateQueue(driver->domainEventState, event);
1249 1250 1251 1252 1253
    }

    return ret;
}

1254 1255

void
1256 1257
qemuMigrationAnyPostcopyFailed(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287
{
    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);
1288
        virObjectEventStateQueue(driver->domainEventState, event);
1289 1290 1291 1292 1293 1294 1295
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
    }
}


1296
static int
1297
qemuMigrationSrcWaitForSpice(virDomainObjPtr vm)
1298 1299 1300
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

1301 1302
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION) ||
        !priv->job.spiceMigration)
1303 1304
        return 0;

1305
    VIR_DEBUG("Waiting for SPICE to finish migration");
1306 1307
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
1308 1309 1310 1311
            return -1;
    }
    return 0;
}
1312

1313 1314 1315 1316

static void
qemuMigrationUpdateJobType(qemuDomainJobInfoPtr jobInfo)
{
1317
    switch ((qemuMonitorMigrationStatus) jobInfo->stats.mig.status) {
1318 1319 1320 1321
    case QEMU_MONITOR_MIGRATION_STATUS_POSTCOPY:
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_POSTCOPY;
        break;

1322
    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
1323
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED;
1324 1325 1326
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
1327
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_NONE;
1328 1329 1330
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
1331
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1332 1333 1334
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
1335
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1336 1337
        break;

1338 1339 1340 1341 1342 1343 1344 1345
    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;

1346 1347 1348
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLING:
1349
    case QEMU_MONITOR_MIGRATION_STATUS_LAST:
1350 1351 1352 1353 1354 1355
        break;
    }
}


int
1356 1357 1358 1359 1360
qemuMigrationAnyFetchStats(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuDomainAsyncJob asyncJob,
                           qemuDomainJobInfoPtr jobInfo,
                           char **error)
1361 1362
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1363
    qemuMonitorMigrationStats stats;
1364 1365 1366 1367 1368
    int rv;

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

1369
    rv = qemuMonitorGetMigrationStats(priv->mon, &stats, error);
1370 1371 1372 1373

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

1374
    jobInfo->stats.mig = stats;
1375 1376

    return 0;
1377 1378 1379
}


1380 1381 1382 1383 1384 1385 1386
static const char *
qemuMigrationJobName(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
1387
        return _("migration out job");
1388 1389 1390 1391
    case QEMU_ASYNC_JOB_SAVE:
        return _("domain save job");
    case QEMU_ASYNC_JOB_DUMP:
        return _("domain core dump job");
1392 1393 1394 1395 1396 1397 1398 1399 1400
    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:
1401 1402 1403 1404 1405 1406
    default:
        return _("job");
    }
}


1407
static int
1408
qemuMigrationJobCheckStatus(virQEMUDriverPtr driver,
1409
                            virDomainObjPtr vm,
1410
                            qemuDomainAsyncJob asyncJob)
1411 1412 1413
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1414
    char *error = NULL;
1415
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1416
    int ret = -1;
1417

1418
    if (!events ||
1419
        jobInfo->stats.mig.status == QEMU_MONITOR_MIGRATION_STATUS_ERROR) {
1420
        if (qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, &error) < 0)
1421 1422
            return -1;
    }
1423

1424 1425
    qemuMigrationUpdateJobType(jobInfo);

1426 1427
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_NONE:
1428 1429
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("is not active"));
1430
        goto cleanup;
1431

1432
    case QEMU_DOMAIN_JOB_STATUS_FAILED:
1433
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
1434 1435 1436
                       qemuMigrationJobName(vm),
                       error ? error : _("unexpectedly failed"));
        goto cleanup;
1437

1438
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
1439 1440
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("canceled by client"));
1441
        goto cleanup;
1442

1443 1444
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
1445
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
1446
    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
1447
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1448
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1449 1450
        break;
    }
1451 1452 1453 1454 1455 1456

    ret = 0;

 cleanup:
    VIR_FREE(error);
    return ret;
1457 1458 1459
}


1460 1461
enum qemuMigrationCompletedFlags {
    QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR = (1 << 0),
1462
    /* This flag should only be set when run on src host */
1463
    QEMU_MIGRATION_COMPLETED_CHECK_STORAGE  = (1 << 1),
1464
    QEMU_MIGRATION_COMPLETED_POSTCOPY       = (1 << 2),
1465
    QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER = (1 << 3),
1466 1467
};

1468

1469 1470 1471 1472 1473 1474 1475
/**
 * 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
1476 1477 1478 1479 1480
qemuMigrationAnyCompleted(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          qemuDomainAsyncJob asyncJob,
                          virConnectPtr dconn,
                          unsigned int flags)
1481 1482 1483 1484 1485
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    int pauseReason;

1486
    if (qemuMigrationJobCheckStatus(driver, vm, asyncJob) < 0)
1487 1488
        goto error;

1489
    /* This flag should only be set when run on src host */
1490
    if (flags & QEMU_MIGRATION_COMPLETED_CHECK_STORAGE &&
1491
        qemuMigrationSrcNBDStorageCopyReady(vm, asyncJob) < 0)
1492 1493
        goto error;

1494
    if (flags & QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR &&
1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507
        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;
    }

1508 1509 1510 1511 1512 1513 1514 1515 1516 1517
    /* 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;
    }

1518 1519 1520 1521 1522
    /* 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 &&
1523
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
1524 1525 1526 1527
        VIR_DEBUG("Migration switched to post-copy");
        return 1;
    }

1528
    if (jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
1529 1530 1531 1532 1533
        return 1;
    else
        return 0;

 error:
1534 1535 1536
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1537
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1538
        /* The migration was aborted by us rather than QEMU itself. */
1539
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1540
        return -2;
1541 1542 1543

    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
        /* Something failed after QEMU already finished the migration. */
1544
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1545
        return -1;
1546 1547 1548 1549

    case QEMU_DOMAIN_JOB_STATUS_FAILED:
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
        /* QEMU aborted the migration. */
1550
        return -1;
1551 1552 1553 1554 1555 1556

    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_NONE:
        /* Impossible. */
        break;
1557
    }
1558 1559

    return -1;
1560 1561 1562
}


1563 1564 1565
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
1566
static int
1567 1568 1569 1570 1571
qemuMigrationSrcWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  virConnectPtr dconn,
                                  unsigned int flags)
1572
{
1573
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
1574
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1575
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1576
    int rv;
1577

1578 1579
    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_MIGRATING;

1580 1581
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           dconn, flags)) != 1) {
1582 1583
        if (rv < 0)
            return rv;
1584

1585 1586
        if (events) {
            if (virDomainObjWait(vm) < 0) {
1587 1588
                if (virDomainObjIsActive(vm))
                    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
                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);
        }
1599 1600
    }

1601
    if (events)
1602
        ignore_value(qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, NULL));
1603

1604
    qemuDomainJobInfoUpdateTime(jobInfo);
1605 1606
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
1607
    if (VIR_ALLOC(priv->job.completed) == 0) {
1608
        *priv->job.completed = *jobInfo;
1609 1610
        priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
    }
1611

1612 1613 1614 1615
    if (asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT &&
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;

1616
    return 0;
1617 1618 1619
}


1620
static int
1621 1622 1623 1624
qemuMigrationDstWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  bool postcopy)
1625 1626
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1627
    unsigned int flags = 0;
1628 1629 1630 1631 1632 1633 1634
    int rv;

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

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

1635 1636 1637
    if (postcopy)
        flags = QEMU_MIGRATION_COMPLETED_POSTCOPY;

1638 1639
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           NULL, flags)) != 1) {
1640 1641 1642 1643 1644 1645 1646 1647
        if (rv < 0 || virDomainObjWait(vm) < 0)
            return -1;
    }

    return 0;
}


1648
static int
1649 1650 1651 1652
qemuMigrationSrcGraphicsRelocate(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuMigrationCookiePtr cookie,
                                 const char *graphicsuri)
1653 1654
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1655 1656
    int ret = -1;
    const char *listenAddress = NULL;
1657
    virSocketAddr addr;
1658 1659 1660 1661 1662
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
1663

1664
    if (!cookie || (!cookie->graphics && !graphicsuri))
1665 1666
        return 0;

1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685
    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) {
1686
        size_t i;
1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713

        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;
            }
        }
    }
1714 1715 1716 1717

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
1718 1719 1720 1721
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
1722

1723 1724 1725 1726 1727 1728 1729 1730 1731
    /* 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;
    }

1732 1733 1734 1735
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
1736
        priv->job.spiceMigration = !ret;
1737 1738
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
1739
    }
1740

1741
 cleanup:
1742
    virURIFree(uri);
1743 1744 1745 1746
    return ret;
}


1747
static int
1748 1749 1750
qemuMigrationDstOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            qemuMigrationCookiePtr cookie)
1751
{
1752 1753
    virDomainNetDefPtr netptr;
    int ret = -1;
1754
    size_t i;
1755 1756 1757 1758 1759 1760 1761 1762

    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:
1763
           break;
1764
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
1765 1766
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
1767 1768 1769
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
1770 1771 1772
                goto cleanup;
            }
            break;
1773 1774 1775 1776 1777
        default:
            break;
        }
    }

1778
    ret = 0;
1779
 cleanup:
1780 1781 1782 1783
    return ret;
}


1784
int
1785 1786
qemuMigrationDstCheckProtocol(virQEMUCapsPtr qemuCaps,
                              const char *migrateFrom)
1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809
{
    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 *
1810 1811
qemuMigrationDstGetURI(const char *migrateFrom,
                       int migrateFd)
1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823
{
    char *uri = NULL;

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

    return uri;
}


1824
int
1825 1826 1827 1828
qemuMigrationDstRun(virQEMUDriverPtr driver,
                    virDomainObjPtr vm,
                    const char *uri,
                    qemuDomainAsyncJob asyncJob)
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844
{
    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) {
1845
        /* qemuMigrationDstWaitForCompletion is called from the Finish phase */
1846 1847 1848 1849
        ret = 0;
        goto cleanup;
    }

1850
    if (qemuMigrationDstWaitForCompletion(driver, vm, asyncJob, false) < 0)
1851 1852 1853 1854 1855 1856 1857 1858 1859
        goto cleanup;

    ret = 0;

 cleanup:
    return ret;
}


1860 1861 1862 1863 1864 1865
/* 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.
 */
1866
static void
1867 1868 1869
qemuMigrationSrcCleanup(virDomainObjPtr vm,
                        virConnectPtr conn,
                        void *opaque)
1870
{
1871
    virQEMUDriverPtr driver = opaque;
1872 1873 1874 1875 1876 1877 1878 1879 1880
    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))
1881
        return;
1882 1883 1884 1885 1886

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

1887
    switch ((qemuMigrationJobPhase) priv->job.phase) {
1888 1889 1890 1891 1892 1893 1894 1895 1896
    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);
1897
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
1898
                                 priv->job.migParams, priv->job.apiFlags);
1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
        /* 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 */
        ;
    }
}

1921

1922
/* The caller is supposed to lock the vm and start a migration job. */
1923
static char *
1924 1925 1926 1927 1928 1929 1930 1931 1932
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)
1933 1934 1935
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
1936
    virDomainDefPtr def = NULL;
1937
    qemuDomainObjPrivatePtr priv = vm->privateData;
1938
    virCapsPtr caps = NULL;
1939
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
1940

1941
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
1942
              " cookieout=%p, cookieoutlen=%p,"
1943
              " nmigrate_disks=%zu, migrate_disks=%p, flags=0x%lx",
1944
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
1945 1946
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
1947

1948 1949 1950
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

1951 1952 1953 1954 1955 1956
    /* 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);
1957

1958
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
1959 1960
        goto cleanup;

1961
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
1962
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
1963 1964
        goto cleanup;

1965 1966 1967 1968 1969 1970 1971 1972 1973
    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;
    }

1974 1975 1976 1977 1978 1979
    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;
    }

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
    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;
            }
2023 2024 2025
        }
    }

2026
    if (virDomainDefHasMemoryHotplug(vm->def) ||
2027
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
2028
         vm->newDef && virDomainDefHasMemoryHotplug(vm->newDef)))
2029 2030
        cookieFlags |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

2031 2032 2033 2034 2035
    if (!qemuDomainVcpuHotplugIsInOrder(vm->def) ||
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
         vm->newDef && !qemuDomainVcpuHotplugIsInOrder(vm->newDef)))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU_HOTPLUG;

2036 2037 2038
    if (priv->origCPU)
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU;

2039 2040
    cookieFlags |= QEMU_MIGRATION_COOKIE_ALLOW_REBOOT;

2041 2042 2043
    if (!(flags & VIR_MIGRATE_OFFLINE))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CAPS;

2044 2045
    if (!(mig = qemuMigrationEatCookie(driver, vm->def,
                                       priv->origname, priv, NULL, 0, 0)))
2046 2047 2048
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
2049
                                QEMU_MIGRATION_SOURCE,
2050
                                cookieout, cookieoutlen,
2051
                                cookieFlags) < 0)
2052 2053
        goto cleanup;

L
liguang 已提交
2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075
    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;
        }
    }

2076
    if (xmlin) {
2077
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt, priv->qemuCaps,
2078 2079
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                            VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2080 2081
            goto cleanup;

2082
        if (!qemuDomainCheckABIStability(driver, vm, def))
2083 2084
            goto cleanup;

2085
        rv = qemuDomainDefFormatLive(driver, def, NULL, false, true);
2086
    } else {
2087 2088
        rv = qemuDomainDefFormatLive(driver, vm->def, priv->origCPU,
                                     false, true);
2089
    }
2090

2091
 cleanup:
2092
    qemuMigrationCookieFree(mig);
2093
    virObjectUnref(caps);
2094
    virDomainDefFree(def);
2095 2096 2097
    return rv;
}

2098
char *
2099 2100 2101 2102 2103 2104 2105 2106 2107
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)
2108 2109 2110
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
2111
    qemuDomainAsyncJob asyncJob;
2112 2113

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
2114 2115
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
2116 2117 2118 2119 2120 2121 2122 2123
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_MIGRATION_OUT;
    } else {
        if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_NONE;
    }

2124
    qemuMigrationSrcStoreDomainState(vm);
2125

2126
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
2127 2128 2129 2130 2131 2132
        goto endjob;

    /* Check if there is any ejected media.
     * We don't want to require them on the destination.
     */
    if (!(flags & VIR_MIGRATE_OFFLINE) &&
2133
        qemuProcessRefreshDisks(driver, vm, asyncJob) < 0)
2134 2135
        goto endjob;

2136 2137 2138
    if (!(xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                           cookieout, cookieoutlen,
                                           nmigrate_disks, migrate_disks, flags)))
2139 2140 2141 2142 2143 2144 2145
        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.
         */
2146
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
2147
                                 qemuMigrationSrcCleanup) < 0) {
2148
            VIR_FREE(xml);
2149
            goto endjob;
2150
        }
2151
        qemuMigrationJobContinue(vm);
2152 2153 2154 2155
    } else {
        goto endjob;
    }

2156
 cleanup:
M
Michal Privoznik 已提交
2157
    virDomainObjEndAPI(&vm);
2158 2159
    return xml;

2160
 endjob:
2161 2162 2163 2164
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2165 2166 2167
    goto cleanup;
}

2168

2169 2170
/* Prepare is the first step, and it runs on the destination host.
 */
2171

2172
static void
2173 2174
qemuMigrationDstPrepareCleanup(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
2175 2176 2177 2178 2179 2180 2181 2182 2183
{
    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));

2184
    virPortAllocatorRelease(priv->migrationPort);
2185 2186
    priv->migrationPort = 0;

2187 2188 2189 2190 2191
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2192
static qemuProcessIncomingDefPtr
2193 2194 2195 2196 2197 2198
qemuMigrationDstPrepare(virDomainObjPtr vm,
                        bool tunnel,
                        const char *protocol,
                        const char *listenAddress,
                        unsigned short port,
                        int fd)
2199 2200
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2201
    qemuProcessIncomingDefPtr inc = NULL;
2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231
    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 */
            }
2232
        } else if (hostIPv6Capable) {
2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255
            /* 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;
    }

2256 2257
    inc = qemuProcessIncomingDefNew(priv->qemuCaps, listenAddress,
                                    migrateFrom, fd, NULL);
2258

2259
 cleanup:
2260 2261
    VIR_FREE(migrateFrom);
    return inc;
2262 2263
}

2264
static int
2265
qemuMigrationDstPrepareAny(virQEMUDriverPtr driver,
2266
                           virConnectPtr dconn,
2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280
                           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,
2281
                           qemuMigrationParamsPtr migParams,
2282
                           unsigned long flags)
2283 2284
{
    virDomainObjPtr vm = NULL;
2285
    virObjectEventPtr event = NULL;
2286
    virErrorPtr origErr;
2287
    int ret = -1;
2288
    int dataFD[2] = { -1, -1 };
2289
    qemuDomainObjPrivatePtr priv = NULL;
2290
    qemuMigrationCookiePtr mig = NULL;
2291
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2292
    char *xmlout = NULL;
L
liguang 已提交
2293
    unsigned int cookieFlags;
2294
    unsigned int startFlags;
2295
    virCapsPtr caps = NULL;
2296
    qemuProcessIncomingDefPtr incoming = NULL;
2297
    bool taint_hook = false;
2298 2299 2300
    bool stopProcess = false;
    bool relabel = false;
    int rv;
2301
    char *tlsAlias = NULL;
2302

2303 2304
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324
    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;
        }
2325 2326
        cookieFlags = 0;
    } else {
2327 2328
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS |
                      QEMU_MIGRATION_COOKIE_CAPS;
L
liguang 已提交
2329 2330
    }

2331 2332 2333 2334 2335 2336 2337 2338 2339
    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;
    }

2340 2341 2342 2343 2344 2345
    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;
    }

2346 2347 2348
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2349
    if (!qemuMigrationSrcIsAllowedHostdev(*def))
2350 2351
        goto cleanup;

J
Jiri Denemark 已提交
2352 2353 2354 2355 2356
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2357
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2358 2359
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2360 2361
            goto cleanup;

2362
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2363 2364 2365 2366 2367 2368 2369
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2370
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2371 2372 2373 2374 2375 2376
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2377
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt, NULL,
2378 2379
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                                 VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE);
J
Jiri Denemark 已提交
2380 2381 2382
                if (!newdef)
                    goto cleanup;

2383
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2384 2385 2386 2387
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2388 2389
                virDomainDefFree(*def);
                *def = newdef;
2390 2391 2392 2393
                /* 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
2394 2395 2396 2397
            }
        }
    }

2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411
    /* 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;

2412
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2413
                                   driver->xmlopt,
2414 2415 2416
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2417
        goto cleanup;
2418
    *def = NULL;
2419

2420
    priv = vm->privateData;
2421 2422
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2423

2424 2425 2426 2427 2428
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2429 2430
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
2431 2432 2433 2434 2435 2436
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2437 2438 2439
    if (qemuMigrationDstPrecreateStorage(vm, mig->nbd,
                                         nmigrate_disks, migrate_disks,
                                         !!(flags & VIR_MIGRATE_NON_SHARED_INC)) < 0)
2440 2441
        goto cleanup;

2442 2443
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                              flags) < 0)
2444
        goto cleanup;
2445
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2446 2447 2448 2449

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

L
liguang 已提交
2450 2451 2452
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2453 2454
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
2455 2456
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
2457
        goto stopjob;
2458 2459
    }

2460 2461
    startFlags = VIR_QEMU_PROCESS_START_AUTODESTROY;

2462
    if (qemuProcessInit(driver, vm, mig->cpu, QEMU_ASYNC_JOB_MIGRATION_IN,
2463
                        true, startFlags) < 0)
2464
        goto stopjob;
2465
    stopProcess = true;
2466

2467 2468
    priv->allowReboot = mig->allowReboot;

2469 2470 2471
    if (!(incoming = qemuMigrationDstPrepare(vm, tunnel, protocol,
                                             listenAddress, port,
                                             dataFD[0])))
2472
        goto stopjob;
2473

2474
    if (qemuProcessPrepareDomain(driver, vm, startFlags) < 0)
2475 2476
        goto stopjob;

2477
    if (qemuProcessPrepareHost(driver, vm, startFlags) < 0)
2478 2479
        goto stopjob;

2480
    rv = qemuProcessLaunch(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2481 2482
                           incoming, NULL,
                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
2483
                           startFlags);
2484 2485 2486
    if (rv < 0) {
        if (rv == -2)
            relabel = true;
2487
        goto stopjob;
2488
    }
2489
    relabel = true;
2490

2491 2492 2493 2494
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
2495
            goto stopjob;
2496
        }
2497
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
2498 2499
    }

2500
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
2501
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
2502
        goto stopjob;
M
Michael R. Hines 已提交
2503 2504
    }

2505
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2506
                                 migParams, mig->caps->automatic) < 0)
2507 2508 2509 2510 2511 2512 2513
        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,
2514
                                         &tlsAlias, NULL,
2515 2516 2517 2518 2519 2520 2521
                                         migParams) < 0)
            goto stopjob;
    } else {
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
            goto stopjob;
    }

2522 2523
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                                 migParams) < 0)
2524 2525
        goto stopjob;

2526 2527 2528
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540
        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;
        }

2541 2542
        if (qemuMigrationDstStartNBDServer(driver, vm, incoming->address,
                                           nmigrate_disks, migrate_disks,
2543
                                           nbdPort, nbdTLSAlias) < 0) {
2544
            goto stopjob;
2545
        }
2546
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
2547 2548
    }

2549 2550 2551 2552 2553 2554 2555 2556 2557
    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");
    }

2558
    if (incoming->deferredURI &&
2559 2560
        qemuMigrationDstRun(driver, vm, incoming->deferredURI,
                            QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2561 2562
        goto stopjob;

2563
    if (qemuProcessFinishStartup(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2564 2565 2566
                                 false, VIR_DOMAIN_PAUSED_MIGRATION) < 0)
        goto stopjob;

2567
 done:
2568 2569 2570
    if (qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_DESTINATION,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
2571 2572 2573 2574 2575 2576 2577
        /* 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");
    }

2578
    if (qemuDomainCleanupAdd(vm, qemuMigrationDstPrepareCleanup) < 0)
2579
        goto stopjob;
2580

L
liguang 已提交
2581 2582
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
2583
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
2584 2585 2586
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
2587

2588 2589 2590 2591
    /* 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.
     */
2592
    qemuMigrationJobContinue(vm);
2593

2594 2595
    if (autoPort)
        priv->migrationPort = port;
2596 2597 2598 2599 2600
    /* 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;
2601
    ret = 0;
2602

2603
 cleanup:
2604
    virErrorPreserveLast(&origErr);
2605
    VIR_FREE(tlsAlias);
2606
    qemuProcessIncomingDefFree(incoming);
J
Jiri Denemark 已提交
2607
    VIR_FREE(xmlout);
2608 2609
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
2610 2611 2612
    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 */
2613
        VIR_FREE(priv->origname);
2614 2615 2616 2617
        /* release if port is auto selected which is not the case if
         * it is given in parameters
         */
        if (nbdPort == 0)
2618
            virPortAllocatorRelease(priv->nbdPort);
2619
        priv->nbdPort = 0;
2620
        virDomainObjRemoveTransientDef(vm);
2621
        qemuDomainRemoveInactiveJob(driver, vm);
2622
    }
M
Michal Privoznik 已提交
2623
    virDomainObjEndAPI(&vm);
2624
    virObjectEventStateQueue(driver->domainEventState, event);
2625
    qemuMigrationCookieFree(mig);
2626
    virObjectUnref(caps);
2627
    virNWFilterUnlockFilterUpdates();
2628
    virErrorRestore(&origErr);
2629
    return ret;
2630

2631
 stopjob:
2632
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2633
                             priv->job.migParams, priv->job.apiFlags);
2634

2635 2636 2637 2638
    if (stopProcess) {
        unsigned int stopFlags = VIR_QEMU_PROCESS_STOP_MIGRATED;
        if (!relabel)
            stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
2639
        virDomainAuditStart(vm, "migrated", false);
2640 2641
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        QEMU_ASYNC_JOB_MIGRATION_IN, stopFlags);
2642
    }
2643

2644
    qemuMigrationJobFinish(driver, vm);
2645
    goto cleanup;
2646 2647 2648
}


2649 2650 2651 2652 2653
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
2654
qemuMigrationDstPrepareTunnel(virQEMUDriverPtr driver,
2655
                              virConnectPtr dconn,
2656 2657 2658 2659 2660 2661 2662
                              const char *cookiein,
                              int cookieinlen,
                              char **cookieout,
                              int *cookieoutlen,
                              virStreamPtr st,
                              virDomainDefPtr *def,
                              const char *origname,
2663
                              qemuMigrationParamsPtr migParams,
2664
                              unsigned long flags)
2665
{
2666
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2667
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
2668
              "origname=%s, flags=0x%lx",
2669
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2670
              cookieout, cookieoutlen, st, *def, origname, flags);
2671

2672 2673 2674 2675 2676 2677
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

2678 2679 2680
    return qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
                                      cookieout, cookieoutlen, def, origname,
                                      st, NULL, 0, false, NULL, 0, NULL, 0,
2681
                                      migParams, flags);
2682 2683 2684
}


2685
static virURIPtr
2686
qemuMigrationAnyParseURI(const char *uri, bool *wellFormed)
2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707
{
    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;
}


2708
int
2709
qemuMigrationDstPrepareDirect(virQEMUDriverPtr driver,
2710
                              virConnectPtr dconn,
2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722
                              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,
2723
                              qemuMigrationParamsPtr migParams,
2724
                              unsigned long flags)
2725
{
2726 2727
    unsigned short port = 0;
    bool autoPort = true;
2728 2729
    char *hostname = NULL;
    int ret = -1;
2730
    virURIPtr uri = NULL;
2731 2732
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
2733

2734
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2735
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
2736
              "def=%p, origname=%s, listenAddress=%s, "
2737
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, flags=0x%lx",
2738
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2739
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
2740
              *def, origname, NULLSTR(listenAddress),
2741
              nmigrate_disks, migrate_disks, nbdPort, flags);
2742

2743 2744
    *uri_out = NULL;

2745 2746 2747
    /* 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
2748 2749 2750
     * 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".
2751 2752 2753 2754 2755 2756
     *
     * 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) {
2757 2758 2759
        bool encloseAddress = false;
        const char *incFormat;

2760
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2761
            goto cleanup;
2762

2763
        if (migrateHost != NULL) {
2764 2765
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
2766

2767
            if (VIR_STRDUP(hostname, migrateHost) < 0)
2768 2769 2770 2771 2772
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
2773 2774

        if (STRPREFIX(hostname, "localhost")) {
2775 2776 2777
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
2778 2779 2780 2781 2782
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
2783
         * compatibility with old targets. We at least make the
2784 2785
         * new targets accept both syntaxes though.
         */
2786 2787 2788 2789 2790 2791
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
2792 2793
            goto cleanup;
    } else {
2794
        bool well_formed_uri;
J
Ján Tomko 已提交
2795

2796
        if (!(uri = qemuMigrationAnyParseURI(uri_in, &well_formed_uri)))
2797
            goto cleanup;
J
Ján Tomko 已提交
2798

2799 2800 2801 2802 2803 2804 2805
        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
2806 2807
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
2808 2809 2810
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
2811 2812 2813 2814 2815 2816 2817 2818 2819 2820
            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) {
2821
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2822
                goto cleanup;
2823

2824
            /* Send well-formed URI only if uri_in was well-formed */
2825 2826 2827 2828 2829 2830 2831 2832
            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;
            }
2833
        } else {
2834 2835
            port = uri->port;
            autoPort = false;
2836 2837 2838 2839 2840 2841
        }
    }

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

2842
    ret = qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
2843 2844 2845 2846
                                     cookieout, cookieoutlen, def, origname,
                                     NULL, uri ? uri->scheme : "tcp",
                                     port, autoPort, listenAddress,
                                     nmigrate_disks, migrate_disks, nbdPort,
2847
                                     migParams, flags);
2848
 cleanup:
2849
    virURIFree(uri);
2850
    VIR_FREE(hostname);
2851
    virObjectUnref(cfg);
2852
    if (ret != 0) {
2853
        VIR_FREE(*uri_out);
2854
        if (autoPort)
2855
            virPortAllocatorRelease(port);
2856
    }
2857 2858 2859 2860
    return ret;
}


2861
virDomainDefPtr
2862 2863 2864 2865
qemuMigrationAnyPrepareDef(virQEMUDriverPtr driver,
                           const char *dom_xml,
                           const char *dname,
                           char **origname)
2866 2867 2868
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
2869
    char *name = NULL;
2870 2871 2872 2873 2874 2875 2876 2877 2878 2879

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

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

2880
    if (!(def = virDomainDefParseString(dom_xml, caps, driver->xmlopt, NULL,
2881 2882
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                        VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2883 2884 2885
        goto cleanup;

    if (dname) {
2886
        name = def->name;
2887 2888 2889 2890 2891 2892
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

2893
 cleanup:
2894
    virObjectUnref(caps);
2895 2896 2897 2898
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
2899 2900 2901 2902
    return def;
}


2903
static int
2904 2905 2906 2907 2908 2909
qemuMigrationSrcConfirmPhase(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
                             const char *cookiein,
                             int cookieinlen,
                             unsigned int flags,
                             int retcode)
2910 2911
{
    qemuMigrationCookiePtr mig;
2912
    virObjectEventPtr event;
2913 2914
    int rv = -1;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2915 2916
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = NULL;
2917

2918
    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
2919
              "flags=0x%x, retcode=%d",
2920
              driver, vm, NULLSTR(cookiein), cookieinlen,
2921 2922 2923 2924 2925 2926 2927 2928 2929
              flags, retcode);

    virCheckFlags(QEMU_MIGRATION_FLAGS, -1);

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

2930 2931
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen,
2932
                                       QEMU_MIGRATION_COOKIE_STATS)))
2933 2934
        goto cleanup;

2935 2936 2937
    if (retcode == 0)
        jobInfo = priv->job.completed;
    else
2938
        VIR_FREE(priv->job.completed);
2939 2940 2941

    /* Update times with the values sent by the destination daemon */
    if (mig->jobInfo && jobInfo) {
2942 2943 2944 2945 2946 2947 2948 2949
        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 &&
2950 2951
            qemuMigrationAnyFetchStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                       jobInfo, NULL) < 0)
2952 2953
            VIR_WARN("Could not refresh migration statistics");

2954 2955 2956
        qemuDomainJobInfoUpdateTime(jobInfo);
        jobInfo->timeDeltaSet = mig->jobInfo->timeDeltaSet;
        jobInfo->timeDelta = mig->jobInfo->timeDelta;
2957 2958
        jobInfo->stats.mig.downtime_set = mig->jobInfo->stats.mig.downtime_set;
        jobInfo->stats.mig.downtime = mig->jobInfo->stats.mig.downtime;
2959 2960
    }

2961 2962 2963
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2964 2965
    /* 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.
2966 2967 2968 2969
     */
    if (retcode == 0) {
        /* If guest uses SPICE and supports seamless migration we have to hold
         * up domain shutdown until SPICE server transfers its data */
2970
        qemuMigrationSrcWaitForSpice(vm);
2971 2972

        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
2973
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
2974 2975 2976
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
        virDomainAuditStop(vm, "migrated");

2977
        event = virDomainEventLifecycleNewFromObj(vm,
2978 2979
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
2980
        virObjectEventStateQueue(driver->domainEventState, event);
2981
        qemuDomainEventEmitJobCompleted(driver, vm);
2982
    } else {
2983
        virErrorPtr orig_err = virSaveLastError();
2984
        int reason;
2985 2986

        /* cancel any outstanding NBD jobs */
2987 2988
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT, NULL);
2989 2990 2991

        virSetError(orig_err);
        virFreeError(orig_err);
2992

2993
        if (virDomainObjGetState(vm, &reason) == VIR_DOMAIN_PAUSED &&
2994
            reason == VIR_DOMAIN_PAUSED_POSTCOPY)
2995
            qemuMigrationAnyPostcopyFailed(driver, vm);
2996 2997
        else
            qemuMigrationSrcRestoreDomainState(driver, vm);
2998

2999
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3000
                                 priv->job.migParams, priv->job.apiFlags);
3001

3002
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
3003 3004 3005
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
    }

3006
 done:
3007 3008 3009
    qemuMigrationCookieFree(mig);
    rv = 0;

3010
 cleanup:
3011 3012 3013 3014 3015
    virObjectUnref(cfg);
    return rv;
}

int
3016 3017 3018 3019 3020 3021
qemuMigrationSrcConfirm(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *cookiein,
                        int cookieinlen,
                        unsigned int flags,
                        int cancelled)
3022
{
3023
    qemuMigrationJobPhase phase;
3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037
    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);
3038
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
3039
                           qemuMigrationSrcCleanup);
3040

3041 3042 3043
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
3044

3045
    qemuMigrationJobFinish(driver, vm);
3046
    if (!virDomainObjIsActive(vm)) {
3047
        if (!cancelled && ret == 0 && flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
3048
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
3049 3050
            vm->persistent = 0;
        }
3051
        qemuDomainRemoveInactiveJob(driver, vm);
3052 3053
    }

3054
 cleanup:
M
Michal Privoznik 已提交
3055
    virDomainObjEndAPI(&vm);
3056 3057 3058 3059 3060
    virObjectUnref(cfg);
    return ret;
}


3061 3062
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3063
    MIGRATION_DEST_CONNECT_HOST,
3064
    MIGRATION_DEST_FD,
3065
};
3066

3067 3068 3069 3070
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3071

3072 3073 3074 3075 3076 3077
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3078
            const char *protocol;
3079 3080 3081 3082
            const char *name;
            int port;
        } host;

3083 3084 3085 3086
        struct {
            int qemu;
            int local;
        } fd;
3087 3088 3089 3090 3091 3092 3093
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3094 3095 3096

#define TUNNEL_SEND_BUF_SIZE 65536

3097 3098 3099 3100 3101 3102 3103
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3104 3105
    int wakeupRecvFD;
    int wakeupSendFD;
3106 3107
};

3108
static void qemuMigrationSrcIOFunc(void *arg)
3109
{
3110
    qemuMigrationIOThreadPtr data = arg;
3111 3112 3113 3114 3115 3116 3117
    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);
3118

3119
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3120
        goto abrt;
3121

3122 3123 3124
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3125
    for (;;) {
3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
        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;
3136
            virReportSystemError(errno, "%s",
3137 3138
                                 _("poll failed in migration tunnel"));
            goto abrt;
3139
        }
3140 3141 3142 3143 3144 3145 3146

        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");
3147
            break;
3148
        }
3149

3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165
        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;
            }
3166 3167
        }

3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184
        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;
            }
        }
    }
3185

3186 3187
    if (virStreamFinish(data->st) < 0)
        goto error;
3188

3189
    VIR_FORCE_CLOSE(data->sock);
3190 3191
    VIR_FREE(buffer);

3192 3193
    return;

3194
 abrt:
3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3206
 error:
3207 3208 3209 3210 3211
    /* 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);
3212
    virResetLastError();
3213
    VIR_FREE(buffer);
3214 3215 3216 3217
}


static qemuMigrationIOThreadPtr
3218 3219
qemuMigrationSrcStartTunnel(virStreamPtr st,
                            int sock)
3220
{
3221 3222
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3223

3224 3225 3226 3227
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3228 3229
    }

3230
    if (VIR_ALLOC(io) < 0)
3231
        goto error;
3232

3233 3234
    io->st = st;
    io->sock = sock;
3235 3236
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3237 3238

    if (virThreadCreate(&io->thread, true,
3239
                        qemuMigrationSrcIOFunc,
3240 3241 3242
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3243
        goto error;
3244 3245 3246
    }

    return io;
3247

3248
 error:
3249 3250 3251 3252
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3253 3254 3255
}

static int
3256
qemuMigrationSrcStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3257 3258
{
    int rv = -1;
3259 3260 3261 3262 3263 3264 3265 3266 3267
    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;
    }

3268 3269 3270 3271
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3272 3273 3274 3275
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3276 3277 3278 3279 3280 3281
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3282
 cleanup:
3283 3284
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3285 3286
    VIR_FREE(io);
    return rv;
3287 3288
}

3289
static int
3290 3291 3292
qemuMigrationSrcConnect(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationSpecPtr spec)
3293 3294 3295 3296 3297 3298 3299
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3300
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3301 3302 3303 3304 3305
        return -1;

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

3306
    if (qemuSecuritySetSocketLabel(driver->securityManager, vm->def) < 0)
3307
        goto cleanup;
3308 3309 3310
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
3311
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
3312
        virObjectUnref(sock);
3313
    }
3314
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0 ||
3315 3316 3317
        spec->dest.fd.qemu == -1)
        goto cleanup;

3318 3319 3320 3321 3322 3323 3324
    /* 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;
    }

3325 3326
    ret = 0;

3327
 cleanup:
3328 3329 3330 3331 3332 3333
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3334 3335

static int
3336 3337 3338 3339
qemuMigrationSrcContinue(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMonitorMigrationStatus status,
                         qemuDomainAsyncJob asyncJob)
3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355
{
    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;
}


3356
static int
3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370
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,
3371
                    qemuMigrationParamsPtr migParams)
3372
{
3373
    int ret = -1;
3374 3375
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3376
    qemuMigrationCookiePtr mig = NULL;
3377
    char *tlsAlias = NULL;
3378
    qemuMigrationIOThreadPtr iothread = NULL;
3379
    int fd = -1;
3380
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3381
    virErrorPtr orig_err = NULL;
3382
    unsigned int cookieFlags = 0;
3383
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3384
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
3385
    bool cancel = false;
3386
    unsigned int waitFlags;
3387
    virDomainDefPtr persistDef = NULL;
3388
    char *timestamp;
3389
    int rc;
3390 3391

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
3392
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3393 3394
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
3395 3396
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3397
              spec, spec->destType, spec->fwdType, dconn,
3398
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3399

3400 3401 3402 3403 3404 3405 3406 3407 3408 3409
    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;
    }

3410 3411
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3412 3413 3414 3415
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3416 3417 3418
        return -1;
    }

3419 3420 3421
    if (events)
        priv->signalIOError = abort_on_error;

3422 3423
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
        if (persist_xml) {
3424 3425
            if (!(persistDef = qemuMigrationAnyPrepareDef(driver, persist_xml,
                                                          NULL, NULL)))
3426
                goto error;
3427 3428 3429
        } else {
            virDomainDefPtr def = vm->newDef ? vm->newDef : vm->def;
            if (!(persistDef = qemuDomainDefCopy(driver, def,
3430 3431
                                                 VIR_DOMAIN_XML_SECURE |
                                                 VIR_DOMAIN_XML_MIGRATABLE)))
3432
                goto error;
3433 3434 3435
        }
    }

3436 3437
    mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                 cookiein, cookieinlen,
3438 3439 3440
                                 cookieFlags |
                                 QEMU_MIGRATION_COOKIE_GRAPHICS |
                                 QEMU_MIGRATION_COOKIE_CAPS);
3441
    if (!mig)
3442
        goto error;
3443

3444
    if (qemuMigrationSrcGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3445 3446
        VIR_WARN("unable to provide data for graphics client relocation");

3447
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3448
                                 migParams, mig->caps->automatic) < 0)
3449 3450
        goto error;

3451
    if (flags & VIR_MIGRATE_TLS) {
3452
        const char *hostname = NULL;
3453 3454 3455 3456

        /* We need to add tls-hostname whenever QEMU itself does not
         * connect directly to the destination. */
        if (spec->destType == MIGRATION_DEST_CONNECT_HOST ||
3457 3458 3459
            spec->destType == MIGRATION_DEST_FD)
            hostname = spec->dest.host.name;

3460
        if (qemuMigrationParamsEnableTLS(driver, vm, false,
3461
                                         QEMU_ASYNC_JOB_MIGRATION_OUT,
3462
                                         &tlsAlias, hostname,
3463 3464
                                         migParams) < 0)
            goto error;
3465
    } else {
3466
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
3467
            goto error;
3468 3469
    }

3470 3471 3472 3473
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                 migParams) < 0)
        goto error;

3474 3475 3476
    if (migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                         QEMU_MONITOR_MIGRATE_NON_SHARED_INC)) {
        if (mig->nbd) {
3477 3478 3479 3480
            /* 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. */
3481 3482
            if (flags & VIR_MIGRATE_TLS &&
                !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV_DEL)) {
3483 3484 3485 3486 3487
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                               _("NBD migration with TLS is not supported"));
                goto error;
            }

3488
            /* This will update migrate_flags on success */
3489 3490 3491 3492 3493 3494
            if (qemuMigrationSrcNBDStorageCopy(driver, vm, mig,
                                               spec->dest.host.name,
                                               migrate_speed,
                                               &migrate_flags,
                                               nmigrate_disks,
                                               migrate_disks,
3495
                                               dconn, tlsAlias, flags) < 0) {
3496
                goto error;
3497 3498 3499 3500 3501 3502 3503
            }
        } 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.");
        }
3504 3505
    }

3506
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3507 3508
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
3509
        if (qemuMigrationSrcSetOffline(driver, vm) < 0)
3510
            goto error;
3511 3512
    }

3513 3514
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3515
        goto error;
3516

3517
    if (priv->job.abortJob) {
3518 3519
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
3520
         * priv->job.abortJob will not change */
3521
        priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
3522 3523 3524
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
3525
        goto exit_monitor;
3526 3527
    }

3528 3529
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
3530

3531 3532
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3533
        qemuMigrationSrcConnect(driver, vm, spec) < 0) {
3534
        goto exit_monitor;
3535
    }
3536

3537 3538 3539 3540 3541 3542
    /* log start of migration */
    if ((timestamp = virTimeStringNow()) != NULL) {
        qemuDomainLogAppendMessage(driver, vm, "%s: initiating migration\n", timestamp);
        VIR_FREE(timestamp);
    }

3543
    rc = -1;
3544 3545
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3546 3547
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
3548
            goto exit_monitor;
M
Michael R. Hines 已提交
3549
        }
3550 3551 3552 3553
        rc = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
                                      spec->dest.host.protocol,
                                      spec->dest.host.name,
                                      spec->dest.host.port);
3554 3555
        break;

3556 3557 3558 3559
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3560
    case MIGRATION_DEST_FD:
3561
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3562
            fd = spec->dest.fd.local;
3563 3564
            spec->dest.fd.local = -1;
        }
3565 3566
        rc = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                    spec->dest.fd.qemu);
3567 3568
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3569
    }
3570 3571

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
3572
        goto error;
3573 3574 3575

    /* From this point onwards we *must* call cancel to abort the
     * migration on source if anything goes wrong */
3576
    cancel = true;
3577

3578
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3579
        if (!(iothread = qemuMigrationSrcStartTunnel(spec->fwd.stream, fd)))
3580
            goto error;
3581 3582 3583 3584 3585
        /* If we've created a tunnel, then the 'fd' will be closed in the
         * qemuMigrationIOFunc as data->sock.
         */
        fd = -1;
    }
3586

3587
    waitFlags = QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;
3588 3589 3590 3591 3592 3593 3594
    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;

3595 3596 3597
    rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT,
                                           dconn, waitFlags);
3598
    if (rc == -2) {
3599
        goto error;
3600 3601 3602 3603 3604
    } else if (rc == -1) {
        /* QEMU reported failed migration, nothing to cancel anymore */
        cancel = false;
        goto error;
    }
3605

3606 3607 3608
    /* 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.
3609
     */
3610 3611 3612 3613 3614
    if (priv->monJSON) {
        while (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            priv->signalStop = true;
            rc = virDomainObjWait(vm);
            priv->signalStop = false;
3615
            if (rc < 0)
3616
                goto error;
3617
        }
3618
    } else if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING &&
3619
               qemuMigrationSrcSetOffline(driver, vm) < 0) {
3620
        goto error;
3621
    }
3622

J
Ján Tomko 已提交
3623
    if (mig->nbd &&
3624 3625 3626
        qemuMigrationSrcNBDCopyCancel(driver, vm, true,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT,
                                      dconn) < 0)
3627
        goto error;
3628

3629 3630 3631 3632 3633
    /* 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) {
3634 3635 3636
        if (qemuMigrationSrcContinue(driver, vm,
                                     QEMU_MONITOR_MIGRATION_STATUS_PRE_SWITCHOVER,
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3637 3638 3639 3640
            goto error;

        waitFlags ^= QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;

3641 3642 3643
        rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_OUT,
                                               dconn, waitFlags);
3644 3645 3646 3647 3648 3649 3650 3651 3652
        if (rc == -2) {
            goto error;
        } else if (rc == -1) {
            /* QEMU reported failed migration, nothing to cancel anymore */
            cancel = false;
            goto error;
        }
    }

3653 3654 3655 3656
    if (iothread) {
        qemuMigrationIOThreadPtr io;

        VIR_STEAL_PTR(io, iothread);
3657
        if (qemuMigrationSrcStopTunnel(io, false) < 0)
3658
            goto error;
3659 3660 3661
    }

    if (priv->job.completed) {
3662
        priv->job.completed->stopped = priv->job.current->stopped;
3663 3664 3665 3666 3667 3668 3669 3670 3671
        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 ||
3672 3673 3674
        qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_SOURCE,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
3675 3676
        VIR_WARN("Unable to encode migration cookie");
    }
3677

3678
    ret = 0;
3679

3680
 cleanup:
3681
    VIR_FREE(tlsAlias);
3682
    VIR_FORCE_CLOSE(fd);
3683
    virDomainDefFree(persistDef);
3684 3685
    qemuMigrationCookieFree(mig);

3686 3687 3688
    if (events)
        priv->signalIOError = false;

3689 3690 3691 3692 3693
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

3694 3695
    return ret;

3696
 error:
3697 3698
    orig_err = virSaveLastError();

3699 3700 3701 3702 3703 3704 3705 3706
    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));
        }
3707

3708 3709
        /* cancel any outstanding NBD jobs */
        if (mig && mig->nbd)
3710 3711 3712
            qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                          QEMU_ASYNC_JOB_MIGRATION_OUT,
                                          dconn);
3713 3714 3715 3716

        if (priv->job.current->status != QEMU_DOMAIN_JOB_STATUS_CANCELED)
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
    }
3717 3718

    if (iothread)
3719
        qemuMigrationSrcStopTunnel(iothread, true);
3720 3721 3722

    goto cleanup;

3723 3724
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
3725
    goto error;
3726 3727
}

3728
/* Perform migration using QEMU's native migrate support,
3729 3730
 * not encrypted obviously
 */
3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745
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,
3746
                              qemuMigrationParamsPtr migParams)
3747
{
3748
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
3749
    virURIPtr uribits = NULL;
3750
    int ret = -1;
3751 3752 3753
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
3754
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3755
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
3756
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
3757
              cookieout, cookieoutlen, flags, resource,
3758
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3759

3760
    if (!(uribits = qemuMigrationAnyParseURI(uri, NULL)))
3761 3762
        return -1;

3763 3764 3765 3766 3767 3768 3769
    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
3770 3771 3772 3773 3774
    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"));
3775
            goto cleanup;
M
Michael R. Hines 已提交
3776
        }
3777
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
3778 3779 3780
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
3781
            goto cleanup;
M
Michael R. Hines 已提交
3782 3783 3784
        }
    }

3785
    if (STRNEQ(uribits->scheme, "rdma"))
3786 3787
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
3788
        spec.destType = MIGRATION_DEST_HOST;
3789
    spec.dest.host.protocol = uribits->scheme;
3790 3791 3792
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
3793

3794 3795 3796
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen, cookieout,
                              cookieoutlen, flags, resource, &spec, dconn,
                              graphicsuri, nmigrate_disks, migrate_disks,
3797
                              migParams);
3798 3799 3800 3801

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

3802
 cleanup:
3803
    virURIFree(uribits);
3804 3805 3806 3807 3808

    return ret;
}


3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823
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,
3824
                              qemuMigrationParamsPtr migParams)
3825 3826 3827
{
    int ret = -1;
    qemuMigrationSpec spec;
3828
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3829
    int fds[2] = { -1, -1 };
3830 3831

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
3832
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3833
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
3834
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
3835
              cookieout, cookieoutlen, flags, resource,
3836
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3837 3838 3839 3840

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

3841

3842 3843 3844
    spec.destType = MIGRATION_DEST_FD;
    spec.dest.fd.qemu = -1;
    spec.dest.fd.local = -1;
3845

3846 3847 3848 3849 3850
    if (pipe2(fds, O_CLOEXEC) == 0) {
        spec.dest.fd.qemu = fds[1];
        spec.dest.fd.local = fds[0];
    }
    if (spec.dest.fd.qemu == -1 ||
3851 3852
        qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    spec.dest.fd.qemu) < 0) {
3853 3854 3855
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
        goto cleanup;
3856 3857
    }

3858 3859 3860
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen,
                              cookieout, cookieoutlen, flags, resource, &spec,
                              dconn, graphicsuri, nmigrate_disks, migrate_disks,
3861
                              migParams);
3862

3863
 cleanup:
3864 3865
    VIR_FORCE_CLOSE(spec.dest.fd.qemu);
    VIR_FORCE_CLOSE(spec.dest.fd.local);
3866

3867
    virObjectUnref(cfg);
3868 3869 3870 3871
    return ret;
}


3872 3873 3874 3875
/* 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 */
3876 3877 3878 3879 3880 3881 3882 3883
static int
qemuMigrationSrcPerformPeer2Peer2(virQEMUDriverPtr driver,
                                  virConnectPtr sconn,
                                  virConnectPtr dconn,
                                  virDomainObjPtr vm,
                                  const char *dconnuri,
                                  unsigned long flags,
                                  const char *dname,
3884 3885
                                  unsigned long resource,
                                  qemuMigrationParamsPtr migParams)
3886 3887 3888
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
3889
    char *cookie = NULL;
3890 3891 3892
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
3893
    bool cancelled;
3894
    virStreamPtr st = NULL;
3895
    unsigned long destflags;
3896

3897
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
3898
              "flags=0x%lx, dname=%s, resource=%lu",
3899 3900
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
3901

3902 3903 3904 3905 3906
    /* 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,
3907 3908
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
3909 3910 3911 3912 3913
        return -1;

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

3914 3915
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
3916

3917 3918 3919 3920 3921 3922 3923 3924 3925 3926
    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;

3927
        qemuDomainObjEnterRemote(vm);
3928
        ret = dconn->driver->domainMigratePrepareTunnel
3929
            (dconn, st, destflags, dname, resource, dom_xml);
3930 3931
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3932
    } else {
3933
        qemuDomainObjEnterRemote(vm);
3934 3935
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
3936
             destflags, dname, resource, dom_xml);
3937 3938
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3939 3940 3941
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
3942 3943
        goto cleanup;

3944 3945
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
3946 3947
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
3948
        cancelled = true;
3949
        orig_err = virSaveLastError();
3950
        goto finish;
3951 3952
    }

3953 3954 3955 3956
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
3957
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
3958
    if (flags & VIR_MIGRATE_TUNNELLED)
3959 3960 3961
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, NULL,
                                            NULL, 0, NULL, NULL,
                                            flags, resource, dconn,
3962
                                            NULL, 0, NULL, migParams);
3963
    else
3964 3965 3966 3967
        ret = qemuMigrationSrcPerformNative(driver, vm, NULL, uri_out,
                                            cookie, cookielen,
                                            NULL, NULL, /* No out cookie with v2 migration */
                                            flags, resource, dconn, NULL, 0, NULL,
3968
                                            migParams);
3969 3970 3971 3972

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
3973

3974 3975 3976
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
3977
    cancelled = ret < 0;
3978

3979
 finish:
3980 3981 3982 3983
    /* 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.
     */
3984
    dname = dname ? dname : vm->def->name;
3985
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
3986
    qemuDomainObjEnterRemote(vm);
3987
    ddomain = dconn->driver->domainMigrateFinish2
3988
        (dconn, dname, cookie, cookielen,
3989
         uri_out ? uri_out : dconnuri, destflags, cancelled);
3990 3991
    /* The domain is already gone at this point */
    ignore_value(qemuDomainObjExitRemote(vm, false));
3992 3993
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
3994

3995
 cleanup:
3996
    if (ddomain) {
3997
        virObjectUnref(ddomain);
3998 3999 4000 4001
        ret = 0;
    } else {
        ret = -1;
    }
4002

4003
    virObjectUnref(st);
4004 4005 4006 4007 4008 4009

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4010
    VIR_FREE(cookie);
4011 4012

    return ret;
4013 4014 4015
}


4016 4017 4018 4019
/* 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 */
4020
static int
4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034
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,
4035
                                  qemuMigrationParamsPtr migParams,
4036 4037 4038
                                  unsigned long long bandwidth,
                                  bool useParams,
                                  unsigned long flags)
4039 4040 4041 4042 4043 4044 4045 4046 4047 4048
{
    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;
4049
    bool cancelled = true;
4050
    virStreamPtr st = NULL;
4051
    unsigned long destflags;
4052 4053 4054
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4055
    size_t i;
4056
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4057 4058

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4059
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4060
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4061
              "bandwidth=%llu, useParams=%d, flags=0x%lx",
4062
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4063
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4064
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks, nbdPort,
4065
              bandwidth, useParams, flags);
4066

4067 4068 4069 4070 4071
    /* 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.  */

4072 4073 4074
    dom_xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                         &cookieout, &cookieoutlen,
                                         nmigrate_disks, migrate_disks, flags);
4075 4076 4077
    if (!dom_xml)
        goto cleanup;

4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097
    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;
4098 4099 4100 4101 4102 4103

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4104 4105 4106 4107 4108
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4109 4110 4111 4112 4113
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4114 4115 4116 4117 4118
        if (nbdPort &&
            virTypedParamsAddInt(&params, &nparams, &maxparams,
                                 VIR_MIGRATE_PARAM_DISKS_PORT,
                                 nbdPort) < 0)
            goto cleanup;
4119

4120 4121
        if (qemuMigrationParamsDump(migParams, &params, &nparams,
                                    &maxparams, &flags) < 0)
4122
            goto cleanup;
4123 4124
    }

4125 4126 4127
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4128 4129
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4130

4131 4132 4133 4134 4135 4136 4137 4138 4139
    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;

4140
        qemuDomainObjEnterRemote(vm);
4141 4142 4143 4144 4145 4146 4147 4148 4149
        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);
        }
4150 4151
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4152
    } else {
4153
        qemuDomainObjEnterRemote(vm);
4154 4155 4156 4157 4158 4159 4160 4161 4162
        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);
        }
4163 4164
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4165 4166 4167 4168 4169
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

4170
    if (offline) {
L
liguang 已提交
4171 4172 4173
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4174
        cancelled = false;
L
liguang 已提交
4175 4176 4177
        goto finish;
    }

4178 4179 4180 4181
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4182 4183
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4184
            goto finish;
4185
        }
4186
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4187 4188
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4189
        orig_err = virSaveLastError();
4190 4191 4192 4193 4194 4195 4196 4197
        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.
     */
4198
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4199
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4200 4201 4202 4203 4204
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4205
    if (flags & VIR_MIGRATE_TUNNELLED) {
4206 4207 4208 4209
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, persist_xml,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4210
                                            nmigrate_disks, migrate_disks,
4211
                                            migParams);
4212
    } else {
4213 4214 4215 4216
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4217
                                            nmigrate_disks, migrate_disks,
4218
                                            migParams);
4219
    }
4220 4221

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4222
    if (ret < 0) {
4223
        orig_err = virSaveLastError();
4224 4225 4226 4227
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4228 4229 4230 4231

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4232
    cancelled = ret < 0;
4233

4234
 finish:
4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246
    /*
     * 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;
4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259

    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);
4260 4261
            if (qemuDomainObjExitRemote(vm, !offline) < 0)
                goto cleanup;
4262 4263 4264 4265 4266 4267 4268
        }
    } else {
        dname = dname ? dname : vm->def->name;
        qemuDomainObjEnterRemote(vm);
        ddomain = dconn->driver->domainMigrateFinish3
            (dconn, dname, cookiein, cookieinlen, &cookieout, &cookieoutlen,
             dconnuri, uri, destflags, cancelled);
4269 4270
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4271
    }
4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291

    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();
4292 4293
                if (err &&
                    err->domain == VIR_FROM_QEMU &&
4294 4295 4296 4297 4298 4299 4300
                    err->code != VIR_ERR_MIGRATE_FINISH_OK) {
                    virFreeError(orig_err);
                    orig_err = NULL;
                }
            }
        }
    }
4301

4302 4303 4304 4305 4306 4307 4308
    /* 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.
4309
     */
4310
    cancelled = ddomain == NULL;
4311

4312 4313 4314 4315 4316 4317
    /* 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();

4318 4319 4320 4321
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4322
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4323 4324 4325 4326 4327
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4328 4329 4330
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
4331 4332 4333 4334
    /* 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.
     */
4335 4336 4337
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4338 4339 4340

 cleanup:
    if (ddomain) {
4341
        virObjectUnref(ddomain);
4342 4343 4344 4345 4346
        ret = 0;
    } else {
        ret = -1;
    }

4347
    virObjectUnref(st);
4348 4349 4350 4351 4352 4353 4354 4355

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4356
    virTypedParamsFree(params, nparams);
4357 4358 4359 4360
    return ret;
}


4361
static void
4362 4363 4364
qemuMigrationSrcConnectionClosed(virConnectPtr conn,
                                 int reason,
                                 void *opaque)
4365 4366 4367 4368 4369 4370 4371 4372
{
    virDomainObjPtr vm = opaque;

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


4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397
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,
4398
                                 qemuMigrationParamsPtr migParams,
4399 4400 4401 4402
                                 unsigned long flags,
                                 const char *dname,
                                 unsigned long resource,
                                 bool *v3proto)
4403 4404 4405 4406
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4407
    virErrorPtr orig_err = NULL;
4408
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4409
    bool dstOffline = false;
4410
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4411
    bool useParams;
4412

4413 4414
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, uri=%s, "
              "graphicsuri=%s, listenAddress=%s, nmigrate_disks=%zu, "
4415
              "migrate_disks=%p, nbdPort=%d, flags=0x%lx, dname=%s, "
4416
              "resource=%lu",
4417
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4418
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4419 4420
              nmigrate_disks, migrate_disks, nbdPort, flags, NULLSTR(dname),
              resource);
4421

4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435
    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;
    }

4436 4437 4438 4439 4440 4441 4442
    if (flags & VIR_MIGRATE_TUNNELLED && nbdPort) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("disk port address is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

4443 4444 4445 4446
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

4447
    qemuDomainObjEnterRemote(vm);
4448
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4449 4450 4451
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;

4452
    if (dconn == NULL) {
4453
        virReportError(VIR_ERR_OPERATION_FAILED,
4454 4455
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4456
        virObjectUnref(cfg);
4457 4458 4459
        return -1;
    }

4460 4461
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4462 4463
        goto cleanup;

4464
    if (virConnectRegisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed,
4465 4466 4467 4468
                                        vm, NULL) < 0) {
        goto cleanup;
    }

4469
    qemuDomainObjEnterRemote(vm);
4470 4471
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4472
        /* v3proto reflects whether the caller used Perform3, but with
4473
         * p2p migrate, regardless of whether Perform2 or Perform3
4474 4475 4476 4477
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4478 4479
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
4480
    if (offline)
4481 4482
        dstOffline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                              VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4483 4484
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;
4485

4486
    if (!p2p) {
4487 4488
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4489 4490 4491
        goto cleanup;
    }

4492 4493
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4494
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
4495 4496 4497 4498 4499 4500
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("Migration APIs with extensible parameters are not "
                         "supported but extended parameters were passed"));
        goto cleanup;
    }

4501
    if (offline && !dstOffline) {
L
liguang 已提交
4502 4503 4504 4505 4506 4507
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4508 4509 4510 4511 4512 4513 4514
    /* 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;

4515
    if (*v3proto) {
4516 4517 4518
        ret = qemuMigrationSrcPerformPeer2Peer3(driver, sconn, dconn, dconnuri, vm, xmlin,
                                                persist_xml, dname, uri, graphicsuri,
                                                listenAddress, nmigrate_disks, migrate_disks,
4519
                                                nbdPort, migParams, resource,
4520
                                                useParams, flags);
4521
    } else {
4522
        ret = qemuMigrationSrcPerformPeer2Peer2(driver, sconn, dconn, vm,
4523
                                                dconnuri, flags, dname, resource,
4524
                                                migParams);
4525
    }
4526

4527
 cleanup:
4528
    orig_err = virSaveLastError();
4529
    qemuDomainObjEnterRemote(vm);
4530
    virConnectUnregisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed);
4531
    virObjectUnref(dconn);
4532
    ignore_value(qemuDomainObjExitRemote(vm, false));
4533 4534 4535 4536
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4537
    virObjectUnref(cfg);
4538 4539 4540 4541
    return ret;
}


4542 4543 4544 4545 4546 4547
/*
 * 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
4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559
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,
4560
                           qemuMigrationParamsPtr migParams,
4561 4562 4563 4564 4565 4566 4567 4568
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
                           const char *dname,
                           unsigned long resource,
                           bool v3proto)
4569
{
4570
    virObjectEventPtr event = NULL;
4571
    int ret = -1;
4572
    virErrorPtr orig_err = NULL;
4573
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4574
    qemuDomainObjPrivatePtr priv = vm->privateData;
4575

4576 4577
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                              flags) < 0)
4578 4579
        goto cleanup;

4580
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
4581 4582
        goto endjob;

4583
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
4584
        goto endjob;
4585

4586
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
4587
        !qemuMigrationSrcIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
4588
        goto endjob;
4589

4590
    qemuMigrationSrcStoreDomainState(vm);
4591 4592

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4593 4594 4595
        ret = qemuMigrationSrcPerformPeer2Peer(driver, conn, vm, xmlin, persist_xml,
                                               dconnuri, uri, graphicsuri, listenAddress,
                                               nmigrate_disks, migrate_disks, nbdPort,
4596
                                               migParams, flags, dname, resource,
4597
                                               &v3proto);
4598
    } else {
4599
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4600 4601 4602
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                            cookieout, cookieoutlen,
                                            flags, resource, NULL, NULL, 0, NULL,
4603
                                            migParams);
4604
    }
4605 4606
    if (ret < 0)
        goto endjob;
4607

4608 4609 4610 4611
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4612
    if (!v3proto) {
4613
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
4614
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4615
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4616
        virDomainAuditStop(vm, "migrated");
4617
        event = virDomainEventLifecycleNewFromObj(vm,
4618 4619
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4620 4621
    }

4622
 endjob:
4623 4624 4625
    if (ret < 0)
        orig_err = virSaveLastError();

4626 4627 4628 4629
    /* v2 proto has no confirm phase so we need to reset migration parameters
     * here
     */
    if (!v3proto && ret < 0)
4630
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4631
                                 priv->job.migParams, priv->job.apiFlags);
4632

4633
    qemuMigrationSrcRestoreDomainState(driver, vm);
4634

4635
    qemuMigrationJobFinish(driver, vm);
4636
    if (!virDomainObjIsActive(vm) && ret == 0) {
4637
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
4638
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4639 4640
            vm->persistent = 0;
        }
4641
        qemuDomainRemoveInactiveJob(driver, vm);
4642 4643
    }

4644 4645 4646 4647 4648
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4649
 cleanup:
M
Michal Privoznik 已提交
4650
    virDomainObjEndAPI(&vm);
4651
    virObjectEventStateQueue(driver->domainEventState, event);
4652
    virObjectUnref(cfg);
4653 4654 4655 4656 4657 4658 4659
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4660 4661 4662 4663 4664 4665 4666 4667
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,
4668
                             qemuMigrationParamsPtr migParams,
4669 4670 4671 4672 4673 4674
                             const char *cookiein,
                             int cookieinlen,
                             char **cookieout,
                             int *cookieoutlen,
                             unsigned long flags,
                             unsigned long resource)
4675
{
4676
    qemuDomainObjPrivatePtr priv = vm->privateData;
4677 4678 4679 4680
    int ret = -1;

    /* If we didn't start the job in the begin phase, start it now. */
    if (!(flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
4681 4682
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
4683 4684 4685 4686 4687 4688
            goto cleanup;
    } else if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) {
        goto cleanup;
    }

    qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4689
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
4690
                           qemuMigrationSrcCleanup);
4691

4692 4693 4694
    ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                        cookieout, cookieoutlen,
                                        flags, resource, NULL, graphicsuri,
4695
                                        nmigrate_disks, migrate_disks, migParams);
4696

4697
    if (ret < 0) {
4698
        qemuMigrationSrcRestoreDomainState(driver, vm);
4699
        goto endjob;
4700
    }
4701 4702 4703

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4704
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
4705
                             qemuMigrationSrcCleanup) < 0)
4706 4707
        goto endjob;

4708
 endjob:
4709
    if (ret < 0) {
4710
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4711
                                 priv->job.migParams, priv->job.apiFlags);
4712
        qemuMigrationJobFinish(driver, vm);
4713
    } else {
4714
        qemuMigrationJobContinue(vm);
4715 4716
    }

4717
    if (!virDomainObjIsActive(vm))
4718
        qemuDomainRemoveInactiveJob(driver, vm);
4719

4720
 cleanup:
M
Michal Privoznik 已提交
4721
    virDomainObjEndAPI(&vm);
4722 4723 4724
    return ret;
}

4725
int
4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737
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,
4738
                        qemuMigrationParamsPtr migParams,
4739 4740 4741 4742 4743 4744 4745 4746
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4747 4748
{
    VIR_DEBUG("driver=%p, conn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4749
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
4750
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4751
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
4752
              "flags=0x%lx, dname=%s, resource=%lu, v3proto=%d",
4753
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4754
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4755 4756 4757
              nmigrate_disks, migrate_disks, nbdPort,
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4758 4759 4760

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4761 4762
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4763 4764 4765
            return -1;
        }

4766 4767 4768
        return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, dconnuri, uri,
                                          graphicsuri, listenAddress,
                                          nmigrate_disks, migrate_disks, nbdPort,
4769
                                          migParams,
4770 4771 4772
                                          cookiein, cookieinlen,
                                          cookieout, cookieoutlen,
                                          flags, dname, resource, v3proto);
4773 4774
    } else {
        if (dconnuri) {
4775 4776
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
4777 4778 4779 4780
            return -1;
        }

        if (v3proto) {
4781 4782 4783
            return qemuMigrationSrcPerformPhase(driver, conn, vm, persist_xml, uri,
                                                graphicsuri,
                                                nmigrate_disks, migrate_disks,
4784
                                                migParams,
4785 4786 4787
                                                cookiein, cookieinlen,
                                                cookieout, cookieoutlen,
                                                flags, resource);
4788
        } else {
4789 4790 4791
            return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, NULL,
                                              uri, graphicsuri, listenAddress,
                                              nmigrate_disks, migrate_disks, nbdPort,
4792
                                              migParams,
4793 4794 4795
                                              cookiein, cookieinlen,
                                              cookieout, cookieoutlen, flags,
                                              dname, resource, v3proto);
4796 4797 4798
        }
    }
}
4799

4800
static int
4801
qemuMigrationDstVPAssociatePortProfiles(virDomainDefPtr def)
4802
{
4803
    size_t i;
4804 4805 4806 4807 4808
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
4809
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4810
            if (virNetDevVPortProfileAssociate(net->ifname,
4811
                                               virDomainNetGetActualVirtPortProfile(net),
4812
                                               &net->mac,
4813
                                               virDomainNetGetActualDirectDev(net),
4814
                                               -1,
4815
                                               def->uuid,
4816 4817
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
4818 4819 4820
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
4821
                goto err_exit;
4822
            }
4823
            last_good_net = i;
4824
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
4825

4826
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
4827 4828 4829 4830
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
4831 4832 4833
        }
    }

4834
    return 0;
4835

4836
 err_exit:
4837
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
4838
        net = def->nets[i];
4839
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4840
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
4841
                                                           virDomainNetGetActualVirtPortProfile(net),
4842
                                                           &net->mac,
4843
                                                           virDomainNetGetActualDirectDev(net),
4844
                                                           -1,
4845
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
4846 4847
        }
    }
4848
    return -1;
4849 4850 4851
}


J
Jiri Denemark 已提交
4852
static int
4853 4854 4855 4856
qemuMigrationDstPersist(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationCookiePtr mig,
                        bool ignoreSaveError)
J
Jiri Denemark 已提交
4857 4858 4859 4860
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    virDomainDefPtr vmdef;
4861 4862
    virDomainDefPtr oldDef = NULL;
    unsigned int oldPersist = vm->persistent;
J
Jiri Denemark 已提交
4863 4864 4865 4866 4867 4868 4869
    virObjectEventPtr event;
    int ret = -1;

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

    vm->persistent = 1;
4870 4871
    oldDef = vm->newDef;
    vm->newDef = qemuMigrationCookieGetPersistent(mig);
J
Jiri Denemark 已提交
4872

4873 4874
    if (!(vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm)))
        goto error;
J
Jiri Denemark 已提交
4875

4876 4877
    if (virDomainSaveConfig(cfg->configDir, driver->caps, vmdef) < 0 &&
        !ignoreSaveError)
4878
        goto error;
J
Jiri Denemark 已提交
4879 4880 4881

    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_DEFINED,
4882 4883 4884
                                              oldPersist ?
                                              VIR_DOMAIN_EVENT_DEFINED_UPDATED :
                                              VIR_DOMAIN_EVENT_DEFINED_ADDED);
4885
    virObjectEventStateQueue(driver->domainEventState, event);
J
Jiri Denemark 已提交
4886 4887 4888 4889

    ret = 0;

 cleanup:
4890
    virDomainDefFree(oldDef);
J
Jiri Denemark 已提交
4891 4892 4893
    virObjectUnref(caps);
    virObjectUnref(cfg);
    return ret;
4894 4895 4896 4897 4898 4899 4900

 error:
    virDomainDefFree(vm->newDef);
    vm->persistent = oldPersist;
    vm->newDef = oldDef;
    oldDef = NULL;
    goto cleanup;
J
Jiri Denemark 已提交
4901 4902 4903
}


4904
virDomainPtr
4905 4906 4907 4908 4909 4910 4911 4912 4913 4914
qemuMigrationDstFinish(virQEMUDriverPtr driver,
                       virConnectPtr dconn,
                       virDomainObjPtr vm,
                       const char *cookiein,
                       int cookieinlen,
                       char **cookieout,
                       int *cookieoutlen,
                       unsigned long flags,
                       int retcode,
                       bool v3proto)
4915 4916
{
    virDomainPtr dom = NULL;
4917
    qemuMigrationCookiePtr mig = NULL;
4918
    virErrorPtr orig_err = NULL;
4919
    int cookie_flags = 0;
J
Jiri Denemark 已提交
4920
    qemuDomainObjPrivatePtr priv = vm->privateData;
4921
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4922
    unsigned short port;
4923 4924
    unsigned long long timeReceived = 0;
    virObjectEventPtr event;
4925
    qemuDomainJobInfoPtr jobInfo = NULL;
4926
    bool inPostCopy = false;
4927
    bool doKill = true;
4928

4929
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
4930
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, retcode=%d",
4931 4932
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
4933

4934 4935 4936
    port = priv->migrationPort;
    priv->migrationPort = 0;

4937
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN)) {
4938
        qemuMigrationDstErrorReport(driver, vm->def->name);
4939
        goto cleanup;
4940
    }
4941

4942 4943
    ignore_value(virTimeMillisNow(&timeReceived));

4944 4945 4946
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
4947

4948
    qemuDomainCleanupRemove(vm, qemuMigrationDstPrepareCleanup);
4949
    VIR_FREE(priv->job.completed);
4950

4951
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
4952 4953
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
4954 4955 4956
    /* 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;
4957

4958 4959
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen, cookie_flags)))
4960
        goto endjob;
4961

4962
    if (flags & VIR_MIGRATE_OFFLINE) {
4963
        if (retcode == 0 &&
4964
            qemuMigrationDstPersist(driver, vm, mig, false) == 0)
4965
            dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, -1);
4966 4967
        goto endjob;
    }
4968

4969 4970 4971 4972
    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
         */
4973
        qemuDomainCheckMonitor(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN);
4974 4975
        goto endjob;
    }
4976

4977 4978 4979
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4980
        qemuMigrationDstErrorReport(driver, vm->def->name);
4981 4982
        goto endjob;
    }
4983

4984
    if (qemuMigrationDstVPAssociatePortProfiles(vm->def) < 0)
4985
        goto endjob;
4986

4987
    if (mig->network && qemuMigrationDstOPDRelocate(driver, vm, mig) < 0)
4988
        VIR_WARN("unable to provide network data for relocation");
4989

4990
    if (qemuMigrationDstStopNBDServer(driver, vm, mig) < 0)
4991
        goto endjob;
4992

4993 4994
    if (qemuRefreshVirtioChannelState(driver, vm,
                                      QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
4995 4996
        goto endjob;

4997 4998
    if (qemuConnectAgent(driver, vm) < 0)
        goto endjob;
4999

5000
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
5001
        if (qemuMigrationDstPersist(driver, vm, mig, !v3proto) < 0) {
5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012
            /* 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.
5013 5014 5015
             */
            if (v3proto)
                goto endjob;
5016
        }
5017
    }
5018

5019 5020 5021
    /* We need to wait for QEMU to process all data sent by the source
     * before starting guest CPUs.
     */
5022 5023 5024
    if (qemuMigrationDstWaitForCompletion(driver, vm,
                                          QEMU_ASYNC_JOB_MIGRATION_IN,
                                          !!(flags & VIR_MIGRATE_POSTCOPY)) < 0) {
5025 5026 5027 5028 5029 5030
        /* 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;
    }
5031

5032 5033 5034 5035 5036 5037 5038 5039 5040
    /* 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;
    }

5041
    if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY)
5042 5043
        inPostCopy = true;

5044 5045 5046 5047 5048
    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
         */
5049
        if (qemuProcessStartCPUs(driver, vm,
5050 5051
                                 inPostCopy ? VIR_DOMAIN_RUNNING_POSTCOPY
                                            : VIR_DOMAIN_RUNNING_MIGRATED,
5052
                                 QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5053
            if (virGetLastErrorCode() == VIR_ERR_OK)
5054 5055 5056 5057 5058 5059
                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();
5060

5061 5062 5063 5064 5065 5066 5067 5068 5069 5070
            /*
             * 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;
5071
        }
5072

5073
        if (inPostCopy)
5074
            doKill = false;
5075
    }
5076

5077
    if (mig->jobInfo) {
5078
        jobInfo = mig->jobInfo;
5079 5080 5081 5082 5083 5084
        mig->jobInfo = NULL;

        if (jobInfo->sent && timeReceived) {
            jobInfo->timeDelta = timeReceived - jobInfo->sent;
            jobInfo->received = timeReceived;
            jobInfo->timeDeltaSet = true;
5085
        }
5086 5087
        qemuDomainJobInfoUpdateTime(jobInfo);
        qemuDomainJobInfoUpdateDowntime(jobInfo);
5088
    }
L
liguang 已提交
5089

5090
    if (inPostCopy) {
5091 5092 5093
        if (qemuMigrationDstWaitForCompletion(driver, vm,
                                              QEMU_ASYNC_JOB_MIGRATION_IN,
                                              false) < 0) {
5094 5095 5096 5097 5098 5099 5100 5101 5102
            goto endjob;
        }
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            virDomainObjSetState(vm,
                                 VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_MIGRATED);
        }
    }

5103
    dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, vm->def->id);
5104

5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118
    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);
    }
5119

5120 5121 5122 5123 5124
    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);
5125
        virObjectEventStateQueue(driver->domainEventState, event);
5126
    }
5127

5128
    if (virDomainObjIsActive(vm) &&
5129
        virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
5130 5131 5132 5133 5134
        VIR_WARN("Failed to save status on vm %s", vm->def->name);

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

5135
 endjob:
5136
    if (!dom &&
5137 5138
        !(flags & VIR_MIGRATE_OFFLINE) &&
        virDomainObjIsActive(vm)) {
5139
        if (doKill) {
5140 5141 5142 5143 5144 5145 5146
            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);
5147
            virObjectEventStateQueue(driver->domainEventState, event);
5148
        } else {
5149
            qemuMigrationAnyPostcopyFailed(driver, vm);
5150
        }
5151 5152
    }

5153
    if (dom) {
5154 5155 5156
        if (jobInfo) {
            VIR_STEAL_PTR(priv->job.completed, jobInfo);
            priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
5157
            priv->job.completed->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;
5158
        }
5159

5160 5161 5162
        if (qemuMigrationBakeCookie(mig, driver, vm,
                                    QEMU_MIGRATION_DESTINATION,
                                    cookieout, cookieoutlen,
5163 5164
                                    QEMU_MIGRATION_COOKIE_STATS) < 0)
            VIR_WARN("Unable to encode migration cookie");
5165 5166 5167 5168 5169 5170

        /* Remove completed stats for post-copy, everything but timing fields
         * is obsolete anyway.
         */
        if (inPostCopy)
            VIR_FREE(priv->job.completed);
5171
    }
5172

5173
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
5174
                             priv->job.migParams, priv->job.apiFlags);
5175

5176
    qemuMigrationJobFinish(driver, vm);
5177
    if (!virDomainObjIsActive(vm))
5178
        qemuDomainRemoveInactiveJob(driver, vm);
5179

5180
 cleanup:
5181
    VIR_FREE(jobInfo);
5182
    virPortAllocatorRelease(port);
5183
    if (priv->mon)
5184
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
5185
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5186
    virDomainObjEndAPI(&vm);
5187
    qemuMigrationCookieFree(mig);
5188 5189 5190 5191
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5192
    virObjectUnref(cfg);
5193 5194 5195 5196

    /* Set a special error if Finish is expected to return NULL as a result of
     * successful call with retcode != 0
     */
5197
    if (retcode != 0 && !dom && virGetLastErrorCode() == VIR_ERR_OK)
5198
        virReportError(VIR_ERR_MIGRATE_FINISH_OK, NULL);
5199 5200
    return dom;
}
5201

5202

5203
/* Helper function called while vm is active.  */
5204
int
5205 5206 5207 5208
qemuMigrationSrcToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
                       int fd,
                       const char *compressor,
                       qemuDomainAsyncJob asyncJob)
5209 5210 5211
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5212
    int ret = -1;
5213 5214
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5215
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5216
    char *errbuf = NULL;
5217
    virErrorPtr orig_err = NULL;
5218 5219 5220 5221 5222

    /* 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,
5223 5224
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5225 5226
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5227
    }
5228

5229 5230 5231 5232 5233 5234 5235
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5236 5237 5238 5239
    if (compressor && pipe(pipeFD) < 0) {
        virReportSystemError(errno, "%s",
                             _("Failed to create pipe for migration"));
        return -1;
5240 5241
    }

5242 5243 5244 5245
    /* 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.  */
5246 5247
    if (qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    compressor ? pipeFD[1] : fd) < 0)
5248 5249
        goto cleanup;

5250
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5251 5252
        goto cleanup;

5253
    if (!compressor) {
5254 5255 5256
        rc = qemuMonitorMigrateToFd(priv->mon,
                                    QEMU_MONITOR_MIGRATE_BACKGROUND,
                                    fd);
5257 5258 5259 5260 5261 5262 5263
    } else {
        const char *prog = compressor;
        const char *args[] = {
            prog,
            "-c",
            NULL
        };
5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274

        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;
5275
        }
5276 5277 5278 5279 5280 5281 5282 5283 5284 5285
        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");
5286
    }
5287
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5288
        goto cleanup;
5289 5290 5291
    if (rc < 0)
        goto cleanup;

5292
    rc = qemuMigrationSrcWaitForCompletion(driver, vm, asyncJob, NULL, 0);
5293

5294 5295 5296 5297
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5298 5299
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5300
                qemuMonitorMigrateCancel(priv->mon);
5301
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5302 5303
            }
        }
5304
        goto cleanup;
5305
    }
5306

5307 5308 5309
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5310
    qemuDomainEventEmitJobCompleted(driver, vm);
5311 5312
    ret = 0;

5313
 cleanup:
5314 5315 5316
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5317
    /* Restore max migration bandwidth */
5318 5319
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5320 5321
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5322
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5323 5324
    }

5325 5326
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5327 5328 5329 5330 5331
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5332 5333 5334 5335 5336 5337

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

5338 5339
    return ret;
}
5340

5341 5342

int
5343 5344
qemuMigrationSrcCancel(virQEMUDriverPtr driver,
                       virDomainObjPtr vm)
5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virHashTablePtr blockJobs = NULL;
    bool storage = false;
    size_t i;
    int ret = -1;

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

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

    qemuDomainObjEnterMonitor(driver, vm);

    ignore_value(qemuMonitorMigrateCancel(priv->mon));
    if (storage)
        blockJobs = qemuMonitorGetAllBlockJobInfo(priv->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || (storage && !blockJobs))
        goto endsyncjob;

    if (!storage) {
        ret = 0;
        goto cleanup;
    }

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

        if (!diskPriv->migrating)
            continue;

        if (virHashLookup(blockJobs, disk->info.alias)) {
            VIR_DEBUG("Drive mirror on disk %s is still running", disk->dst);
        } else {
            VIR_DEBUG("Drive mirror on disk %s is gone", disk->dst);
5388
            qemuBlockJobSyncEnd(vm, QEMU_ASYNC_JOB_NONE, disk);
5389 5390 5391 5392
            diskPriv->migrating = false;
        }
    }

5393 5394
    if (qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                      QEMU_ASYNC_JOB_NONE, NULL) < 0)
5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409
        goto endsyncjob;

    ret = 0;

 cleanup:
    virHashFree(blockJobs);
    return ret;

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

            if (diskPriv->migrating) {
5410
                qemuBlockJobSyncEnd(vm, QEMU_ASYNC_JOB_NONE, disk);
5411 5412 5413 5414 5415 5416 5417 5418
                diskPriv->migrating = false;
            }
        }
    }
    goto cleanup;
}


5419
static int
5420
qemuMigrationJobStart(virQEMUDriverPtr driver,
5421
                      virDomainObjPtr vm,
5422 5423
                      qemuDomainAsyncJob job,
                      unsigned long apiFlags)
5424
{
5425
    qemuDomainObjPrivatePtr priv = vm->privateData;
5426 5427
    virDomainJobOperation op;
    unsigned long long mask;
5428

5429
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5430 5431
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_IN;
        mask = QEMU_JOB_NONE;
5432
    } else {
5433 5434 5435 5436
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_OUT;
        mask = QEMU_JOB_DEFAULT_MASK |
               JOB_MASK(QEMU_JOB_SUSPEND) |
               JOB_MASK(QEMU_JOB_MIGRATION_OP);
5437
    }
5438

5439
    if (qemuDomainObjBeginAsyncJob(driver, vm, job, op, apiFlags) < 0)
5440 5441
        return -1;

5442 5443
    priv->job.current->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;

5444
    qemuDomainObjSetAsyncJobMask(vm, mask);
5445 5446 5447
    return 0;
}

5448
static void
5449
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5450
                         virDomainObjPtr vm,
5451
                         qemuMigrationJobPhase phase)
5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464
{
    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);
}

5465
static void
5466
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5467
                           virDomainObjPtr vm,
5468
                           qemuMigrationJobPhase phase)
5469 5470 5471 5472
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5473
static void
5474 5475
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5476
    qemuDomainObjReleaseAsyncJob(vm);
5477 5478
}

5479
static bool
5480
qemuMigrationJobIsActive(virDomainObjPtr vm,
5481
                         qemuDomainAsyncJob job)
5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492
{
    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");

5493
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5494 5495 5496 5497 5498
        return false;
    }
    return true;
}

5499
static void
5500
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5501
{
5502
    qemuDomainObjEndAsyncJob(driver, vm);
5503
}
5504 5505 5506


static void
5507
qemuMigrationDstErrorFree(void *data,
5508 5509 5510 5511 5512 5513 5514
                       const void *name ATTRIBUTE_UNUSED)
{
    virErrorPtr err = data;
    virFreeError(err);
}

int
5515
qemuMigrationDstErrorInit(virQEMUDriverPtr driver)
5516
{
5517
    driver->migrationErrors = virHashAtomicNew(64, qemuMigrationDstErrorFree);
5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528
    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
5529 5530 5531
qemuMigrationDstErrorSave(virQEMUDriverPtr driver,
                          const char *name,
                          virErrorPtr err)
5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544
{
    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
5545 5546
qemuMigrationDstErrorReport(virQEMUDriverPtr driver,
                            const char *name)
5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557
{
    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);
}
5558 5559


5560
int
5561 5562 5563 5564
qemuMigrationSrcFetchMirrorStats(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 qemuDomainJobInfoPtr jobInfo)
5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608
{
    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;
}