qemu_migration.c 189.3 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 24
 *
 */

#include <config.h>

#include <sys/time.h>
25
#include <fcntl.h>
26
#include <poll.h>
27 28

#include "qemu_migration.h"
29
#include "qemu_migration_cookie.h"
30
#include "qemu_migration_params.h"
31 32 33 34
#include "qemu_monitor.h"
#include "qemu_domain.h"
#include "qemu_process.h"
#include "qemu_capabilities.h"
35
#include "qemu_alias.h"
36
#include "qemu_cgroup.h"
37
#include "qemu_hotplug.h"
38
#include "qemu_blockjob.h"
39
#include "qemu_security.h"
40
#include "qemu_slirp.h"
41
#include "qemu_block.h"
42

43
#include "domain_audit.h"
44
#include "virlog.h"
45
#include "virerror.h"
46
#include "viralloc.h"
E
Eric Blake 已提交
47
#include "virfile.h"
48
#include "virnetdevopenvswitch.h"
49
#include "datatypes.h"
50
#include "virfdstream.h"
51
#include "viruuid.h"
52
#include "virtime.h"
53
#include "locking/domain_lock.h"
54
#include "rpc/virnetsocket.h"
55
#include "virstoragefile.h"
M
Martin Kletzander 已提交
56
#include "viruri.h"
57
#include "virhook.h"
58
#include "virstring.h"
59
#include "virtypedparam.h"
M
Michael R. Hines 已提交
60
#include "virprocess.h"
61
#include "nwfilter_conf.h"
62
#include "virdomainsnapshotobjlist.h"
63
#include "virsocket.h"
J
Ján Tomko 已提交
64
#include "virutil.h"
65 66 67

#define VIR_FROM_THIS VIR_FROM_QEMU

68 69
VIR_LOG_INIT("qemu.qemu_migration");

70 71
VIR_ENUM_IMPL(qemuMigrationJobPhase,
              QEMU_MIGRATION_PHASE_LAST,
72 73 74 75 76 77 78 79 80 81 82 83
              "none",
              "perform2",
              "begin3",
              "perform3",
              "perform3_done",
              "confirm3_cancelled",
              "confirm3",
              "prepare",
              "finish2",
              "finish3",
);

84 85 86
static int
qemuMigrationJobStart(virQEMUDriverPtr driver,
                      virDomainObjPtr vm,
87 88
                      qemuDomainAsyncJob job,
                      unsigned long apiFlags)
89
    ATTRIBUTE_NONNULL(1) ATTRIBUTE_NONNULL(2) G_GNUC_WARN_UNUSED_RESULT;
90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115

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);
116

117
static void
118
qemuMigrationSrcStoreDomainState(virDomainObjPtr vm)
119 120 121 122 123 124 125 126 127 128
{
    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
129
qemuMigrationSrcRestoreDomainState(virQEMUDriverPtr driver, virDomainObjPtr vm)
130 131
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
132 133
    int reason;
    virDomainState state = virDomainObjGetState(vm, &reason);
134 135
    bool ret = false;

136 137 138 139 140
    VIR_DEBUG("driver=%p, vm=%p, pre-mig-state=%s, state=%s, reason=%s",
              driver, vm,
              virDomainStateTypeToString(priv->preMigrationState),
              virDomainStateTypeToString(state),
              virDomainStateReasonToString(state, reason));
141

142 143 144 145 146
    if (state != VIR_DOMAIN_PAUSED ||
        reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED)
        goto cleanup;

    if (priv->preMigrationState == VIR_DOMAIN_RUNNING) {
147 148 149 150 151 152
        /* 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 */
153
        if (qemuProcessStartCPUs(driver, vm,
154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169
                                 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;
}

170 171

static int
172 173 174
qemuMigrationDstPrecreateDisk(virConnectPtr conn,
                              virDomainDiskDefPtr disk,
                              unsigned long long capacity)
175 176 177 178 179 180 181 182 183 184 185 186
{
    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));

187
    switch ((virStorageType)disk->src->type) {
188 189 190 191 192 193 194
    case VIR_STORAGE_TYPE_FILE:
        if (!virDomainDiskGetSource(disk)) {
            VIR_DEBUG("Dropping sourceless disk '%s'",
                      disk->dst);
            return 0;
        }

195
        basePath = g_strdup(disk->src->path);
196 197 198 199 200 201 202 203 204 205 206

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

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

207
        if (!(pool = virStoragePoolLookupByTargetPath(conn, basePath)))
208 209 210 211 212 213 214 215 216 217 218 219 220 221 222
            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;

223 224 225 226 227
    case VIR_STORAGE_TYPE_NETWORK:
        VIR_DEBUG("Skipping creation of network disk '%s'",
                  disk->dst);
        return 0;

228 229
    case VIR_STORAGE_TYPE_BLOCK:
    case VIR_STORAGE_TYPE_DIR:
230
    case VIR_STORAGE_TYPE_NVME:
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275
    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;
}

276
static bool
277 278
qemuMigrationAnyCopyDisk(virDomainDiskDef const *disk,
                         size_t nmigrate_disks, const char **migrate_disks)
279 280 281 282 283 284 285 286 287 288 289 290 291 292 293
{
    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 &&
294
           !virStorageSourceIsEmpty(disk->src);
295 296
}

297 298

static int
299 300 301 302 303
qemuMigrationDstPrecreateStorage(virDomainObjPtr vm,
                                 qemuMigrationCookieNBDPtr nbd,
                                 size_t nmigrate_disks,
                                 const char **migrate_disks,
                                 bool incremental)
304 305 306
{
    int ret = -1;
    size_t i = 0;
307
    virConnectPtr conn;
308 309 310 311

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

312 313 314
    if (!(conn = virGetConnectStorage()))
        return -1;

315 316 317 318
    for (i = 0; i < nbd->ndisks; i++) {
        virDomainDiskDefPtr disk;
        const char *diskSrcPath;

319
        VIR_DEBUG("Looking up disk target '%s' (capacity=%llu)",
320 321
                  nbd->disks[i].target, nbd->disks[i].capacity);

322
        if (!(disk = virDomainDiskByTarget(vm->def, nbd->disks[i].target))) {
323 324 325 326 327 328 329 330
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unable to find disk by target: %s"),
                           nbd->disks[i].target);
            goto cleanup;
        }

        diskSrcPath = virDomainDiskGetSource(disk);

331
        /* Skip disks we don't want to migrate and already existing disks. */
332
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks) ||
333 334 335 336
            (diskSrcPath && virFileExists(diskSrcPath))) {
            continue;
        }

337 338 339 340 341 342 343
        if (incremental) {
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                           _("pre-creation of storage targets for incremental "
                             "storage migration is not supported"));
            goto cleanup;
        }

344 345
        VIR_DEBUG("Proceeding with disk source %s", NULLSTR(diskSrcPath));

346
        if (qemuMigrationDstPrecreateDisk(conn, disk, nbd->disks[i].capacity) < 0)
347 348 349 350 351
            goto cleanup;
    }

    ret = 0;
 cleanup:
352
    virObjectUnref(conn);
353 354 355 356
    return ret;
}


357
/**
358
 * qemuMigrationDstStartNBDServer:
359 360 361 362 363 364 365 366 367 368 369
 * @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
370 371 372 373 374
qemuMigrationDstStartNBDServer(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
                               const char *listenAddr,
                               size_t nmigrate_disks,
                               const char **migrate_disks,
375 376
                               int nbdPort,
                               const char *tls_alias)
377 378 379 380 381
{
    int ret = -1;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    unsigned short port = 0;
    size_t i;
382 383 384 385
    virStorageNetHostDef server = {
        .name = (char *)listenAddr, /* cast away const */
        .transport = VIR_STORAGE_NET_HOST_TRANS_TCP,
    };
386

387 388 389 390 391 392
    if (nbdPort < 0 || nbdPort > USHRT_MAX) {
        virReportError(VIR_ERR_INVALID_ARG, "%s",
                       _("nbd port must be in range 0-65535"));
        return -1;
    }

393 394
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
395
        g_autofree char *diskAlias = NULL;
396 397
        const char *exportname = NULL;
        const char *devicename = NULL;
398

399
        /* check whether disk should be migrated */
400
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
401 402
            continue;

403
        if (disk->src->readonly || virStorageSourceIsEmpty(disk->src)) {
404
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED,
405 406
                           _("Cannot migrate empty or read-only disk %s"),
                           disk->dst);
407 408 409
            goto cleanup;
        }

410
        if (!(diskAlias = qemuAliasDiskDriveFromDisk(disk)))
411 412
            goto cleanup;

413 414 415 416 417 418 419 420
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV)) {
            exportname = diskAlias;
            devicename = disk->src->nodeformat;
        } else {
            exportname = NULL;
            devicename = diskAlias;
        }

421 422 423 424
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
            goto cleanup;

425 426 427 428 429 430
        if (port == 0) {
            if (nbdPort)
                port = nbdPort;
            else if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
                goto exit_monitor;

431 432
            server.port = port;
            if (qemuMonitorNBDServerStart(priv->mon, &server, tls_alias) < 0)
433
                goto exit_monitor;
434 435
        }

436
        if (qemuMonitorNBDServerAdd(priv->mon, devicename, exportname, true, NULL) < 0)
437 438
            goto exit_monitor;
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
439 440 441 442 443 444
            goto cleanup;
    }

    priv->nbdPort = port;
    ret = 0;

445
 cleanup:
446
    if (ret < 0 && nbdPort == 0)
447
        virPortAllocatorRelease(port);
448
    return ret;
449 450 451 452

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;
453 454
}

455 456

static int
457 458 459
qemuMigrationDstStopNBDServer(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              qemuMigrationCookiePtr mig)
460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
{
    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;

475
    virPortAllocatorRelease(priv->nbdPort);
476 477 478 479 480
    priv->nbdPort = 0;
    return 0;
}


481
static void
482 483
qemuMigrationNBDReportMirrorError(qemuBlockJobDataPtr job,
                                  const char *diskdst)
484
{
485
    if (job->errmsg) {
486 487
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("migration of disk %s failed: %s"),
488
                       diskdst, job->errmsg);
489 490
    } else {
        virReportError(VIR_ERR_OPERATION_FAILED,
491
                       _("migration of disk %s failed"), diskdst);
492 493 494 495
    }
}


496
/**
497
 * qemuMigrationSrcNBDStorageCopyReady:
498 499
 * @vm: domain
 *
500 501
 * Check the status of all drives copied via qemuMigrationSrcNBDStorageCopy.
 * Any pending block job events for the mirrored disks will be processed.
502 503 504 505 506 507
 *
 * Returns 1 if all mirrors are "ready",
 *         0 if some mirrors are still performing initial sync,
 *        -1 on error.
 */
static int
508 509
qemuMigrationSrcNBDStorageCopyReady(virDomainObjPtr vm,
                                    qemuDomainAsyncJob asyncJob)
510 511
{
    size_t i;
512 513
    size_t notReady = 0;
    int status;
514 515 516

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
517
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
518
        qemuBlockJobDataPtr job;
519

520
        if (!diskPriv->migrating)
521 522
            continue;

523 524 525 526 527 528 529
        if (!(job = qemuBlockJobDiskGetJob(disk))) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("missing block job data for disk '%s'"), disk->dst);
            return -1;
        }

        status = qemuBlockJobUpdate(vm, job, asyncJob);
530
        if (status == VIR_DOMAIN_BLOCK_JOB_FAILED) {
531 532
            qemuMigrationNBDReportMirrorError(job, disk->dst);
            virObjectUnref(job);
533 534
            return -1;
        }
535

536 537
        virObjectUnref(job);

538 539
        if (disk->mirrorState != VIR_DOMAIN_DISK_MIRROR_STATE_READY)
            notReady++;
540 541
    }

542 543 544 545 546 547 548
    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;
    }
549 550 551
}


552 553 554 555
/*
 * 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.
556
 *
557 558 559 560
 * 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.
561 562
 */
static int
563 564 565
qemuMigrationSrcNBDCopyCancelled(virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 bool check)
566
{
567 568
    size_t i;
    size_t active = 0;
569
    size_t completed = 0;
570 571
    int status;
    bool failed = false;
572

573
 retry:
574 575 576
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
577
        qemuBlockJobDataPtr job;
578

579 580
        if (!diskPriv->migrating)
            continue;
581

582 583 584 585
        if (!(job = qemuBlockJobDiskGetJob(disk)))
            continue;

        status = qemuBlockJobUpdate(vm, job, asyncJob);
586 587 588
        switch (status) {
        case VIR_DOMAIN_BLOCK_JOB_FAILED:
            if (check) {
589
                qemuMigrationNBDReportMirrorError(job, disk->dst);
590
                failed = true;
591
            }
592
            G_GNUC_FALLTHROUGH;
593 594 595 596
        case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
            diskPriv->migrating = false;
            break;
597

598 599
        default:
            active++;
600
        }
601 602 603

        if (status == VIR_DOMAIN_BLOCK_JOB_COMPLETED)
            completed++;
604 605

        virObjectUnref(job);
606 607 608 609 610 611 612 613 614 615
    }

    /* 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;
616
    }
617

618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
    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;
634
        }
635
    }
636
}
637 638


639 640 641 642 643 644
/*
 * 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
645 646 647
qemuMigrationSrcNBDCopyCancelOne(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 virDomainDiskDefPtr disk,
648
                                 qemuBlockJobDataPtr job,
649 650
                                 bool failNoJob,
                                 qemuDomainAsyncJob asyncJob)
651 652 653 654 655
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int status;
    int rv;

656
    status = qemuBlockJobUpdate(vm, job, asyncJob);
657 658 659 660
    switch (status) {
    case VIR_DOMAIN_BLOCK_JOB_FAILED:
    case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        if (failNoJob) {
661
            qemuMigrationNBDReportMirrorError(job, disk->dst);
662
            return -1;
663
        }
664
        G_GNUC_FALLTHROUGH;
665
    case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
666
        return 1;
667 668
    }

669
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
670
        return -1;
671

672
    rv = qemuMonitorBlockJobCancel(priv->mon, job->name);
673 674

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rv < 0)
675
        return -1;
676

677
    return 0;
678 679 680 681
}


/**
682
 * qemuMigrationSrcNBDCopyCancel:
683 684
 * @driver: qemu driver
 * @vm: domain
685
 * @check: if true report an error when some of the mirrors fails
686
 *
687
 * Cancel all drive-mirrors started by qemuMigrationSrcNBDStorageCopy.
688 689 690
 * Any pending block job events for the affected disks will be processed and
 * synchronous block job terminated regardless of return value unless qemu
 * has crashed.
691 692 693 694
 *
 * Returns 0 on success, -1 otherwise.
 */
static int
695 696 697 698 699
qemuMigrationSrcNBDCopyCancel(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              bool check,
                              qemuDomainAsyncJob asyncJob,
                              virConnectPtr dconn)
700
{
701
    virErrorPtr err = NULL;
702
    int ret = -1;
703
    size_t i;
704 705 706 707
    int rv;
    bool failed = false;

    VIR_DEBUG("Cancelling drive mirrors for domain %s", vm->def->name);
708 709 710

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
711
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
712
        qemuBlockJobDataPtr job;
713

714 715
        if (!(job = qemuBlockJobDiskGetJob(disk)) ||
            !qemuBlockJobIsRunning(job))
716 717
            diskPriv->migrating = false;

718 719
        if (!diskPriv->migrating) {
            virObjectUnref(job);
720
            continue;
721
        }
722

723
        rv = qemuMigrationSrcNBDCopyCancelOne(driver, vm, disk, job,
724
                                              check, asyncJob);
725 726 727
        if (rv != 0) {
            if (rv < 0) {
                if (!err)
728
                    virErrorPreserveLast(&err);
729 730
                failed = true;
            }
731
            qemuBlockJobSyncEnd(vm, job, asyncJob);
732 733
            diskPriv->migrating = false;
        }
734 735

        virObjectUnref(job);
736 737
    }

738
    while ((rv = qemuMigrationSrcNBDCopyCancelled(vm, asyncJob, check)) != 1) {
739 740 741 742 743 744 745
        if (check && !failed &&
            dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            failed = true;
        }

746 747 748 749
        if (rv < 0) {
            failed = true;
            if (rv == -2)
                break;
750
        }
751

752
        if (failed && !err)
753
            virErrorPreserveLast(&err);
754 755 756

        if (virDomainObjWait(vm) < 0)
            goto cleanup;
757 758
    }

759 760 761 762 763 764 765 766 767
    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);
768
        virObjectUnref(diskPriv->migrSource);
769 770 771
        diskPriv->migrSource = NULL;
    }

772 773 774
    ret = failed ? -1 : 0;

 cleanup:
775
    virErrorRestore(&err);
776
    return ret;
777 778 779
}


780 781 782 783 784
static virStorageSourcePtr
qemuMigrationSrcNBDStorageCopyBlockdevPrepareSource(virDomainDiskDefPtr disk,
                                                    const char *host,
                                                    int port,
                                                    const char *tlsAlias)
785
{
786
    g_autoptr(virStorageSource) copysrc = NULL;
787

788
    if (!(copysrc = virStorageSourceNew()))
789
        return NULL;
790 791 792 793 794

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

795
    if (!(copysrc->backingStore = virStorageSourceNew()))
796
        return NULL;
797

798 799
    if (!(copysrc->path = qemuAliasDiskDriveFromDisk(disk)))
        return NULL;
800

801
    copysrc->hosts = g_new0(virStorageNetHostDef, 1);
802 803 804 805

    copysrc->nhosts = 1;
    copysrc->hosts->transport = VIR_STORAGE_NET_HOST_TRANS_TCP;
    copysrc->hosts->port = port;
806
    copysrc->hosts->name = g_strdup(host);
807

808
    copysrc->tlsAlias = g_strdup(tlsAlias);
809

810 811
    copysrc->nodestorage = g_strdup_printf("migration-%s-storage", disk->dst);
    copysrc->nodeformat = g_strdup_printf("migration-%s-format", disk->dst);
812

813 814 815 816 817 818 819 820
    return g_steal_pointer(&copysrc);
}


static int
qemuMigrationSrcNBDStorageCopyBlockdev(virQEMUDriverPtr driver,
                                       virDomainObjPtr vm,
                                       virDomainDiskDefPtr disk,
821 822 823
                                       const char *jobname,
                                       const char *sourcename,
                                       bool persistjob,
824 825 826 827 828 829 830 831 832 833 834
                                       const char *host,
                                       int port,
                                       unsigned long long mirror_speed,
                                       unsigned int mirror_shallow,
                                       const char *tlsAlias)
{
    g_autoptr(qemuBlockStorageSourceAttachData) data = NULL;
    qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
    int mon_ret = 0;
    g_autoptr(virStorageSource) copysrc = NULL;

835
    VIR_DEBUG("starting blockdev mirror for disk=%s to host=%s", disk->dst, host);
836 837 838 839

    if (!(copysrc = qemuMigrationSrcNBDStorageCopyBlockdevPrepareSource(disk, host, port, tlsAlias)))
        return -1;

840 841
    /* Migration via blockdev-mirror was supported sooner than the auto-read-only
     * feature was added to qemu */
842 843 844
    if (!(data = qemuBlockStorageSourceAttachPrepareBlockdev(copysrc,
                                                             copysrc->backingStore,
                                                             false)))
845
        return -1;
846 847 848

    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
849
        return -1;
850 851 852 853

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

    if (mon_ret == 0)
854 855
        mon_ret = qemuMonitorBlockdevMirror(qemuDomainGetMonitor(vm), jobname, persistjob,
                                            sourcename, copysrc->nodeformat,
856
                                            mirror_speed, 0, 0, mirror_shallow);
857 858 859 860 861

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

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || mon_ret < 0)
862
        return -1;
863

864
    diskPriv->migrSource = g_steal_pointer(&copysrc);
865

866
    return 0;
867 868 869
}


870 871 872 873 874 875 876
static int
qemuMigrationSrcNBDStorageCopyDriveMirror(virQEMUDriverPtr driver,
                                          virDomainObjPtr vm,
                                          const char *diskAlias,
                                          const char *host,
                                          int port,
                                          unsigned long long mirror_speed,
877
                                          bool mirror_shallow)
878
{
879
    g_autofree char *nbd_dest = NULL;
880 881 882
    int mon_ret;

    if (strchr(host, ':')) {
883 884
        nbd_dest = g_strdup_printf("nbd:[%s]:%d:exportname=%s", host, port,
                                   diskAlias);
885
    } else {
886 887
        nbd_dest = g_strdup_printf("nbd:%s:%d:exportname=%s", host, port,
                                   diskAlias);
888 889 890 891
    }

    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
892
        return -1;
893 894 895

    mon_ret = qemuMonitorDriveMirror(qemuDomainGetMonitor(vm),
                                     diskAlias, nbd_dest, "raw",
896
                                     mirror_speed, 0, 0, mirror_shallow, true);
897 898

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || mon_ret < 0)
899
        return -1;
900

901
    return 0;
902 903 904
}


905 906 907 908 909 910 911
static int
qemuMigrationSrcNBDStorageCopyOne(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  virDomainDiskDefPtr disk,
                                  const char *host,
                                  int port,
                                  unsigned long long mirror_speed,
912
                                  bool mirror_shallow,
913 914 915
                                  const char *tlsAlias,
                                  unsigned int flags)
{
916
    qemuDomainObjPrivatePtr priv = vm->privateData;
917
    qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
918
    qemuBlockJobDataPtr job = NULL;
919
    char *diskAlias = NULL;
920 921 922
    const char *jobname = NULL;
    const char *sourcename = NULL;
    bool persistjob = false;
923 924 925 926 927 928
    int rc;
    int ret = -1;

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

929
    if (!(job = qemuBlockJobDiskNew(vm, disk, QEMU_BLOCKJOB_TYPE_COPY, diskAlias)))
930 931
        goto cleanup;

932 933
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV)) {
        jobname = diskAlias;
934
        sourcename = qemuDomainDiskGetTopNodename(disk);
935 936 937 938 939 940 941
        persistjob = true;
    } else {
        jobname = NULL;
        sourcename = diskAlias;
        persistjob = false;
    }

942
    qemuBlockJobSyncBegin(job);
943

944 945
    if (flags & VIR_MIGRATE_TLS ||
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV)) {
946
        rc = qemuMigrationSrcNBDStorageCopyBlockdev(driver, vm,
947 948
                                                    disk, jobname,
                                                    sourcename, persistjob,
949 950
                                                    host, port,
                                                    mirror_speed,
951
                                                    mirror_shallow,
952 953 954 955 956
                                                    tlsAlias);
    } else {
        rc = qemuMigrationSrcNBDStorageCopyDriveMirror(driver, vm, diskAlias,
                                                       host, port,
                                                       mirror_speed,
957
                                                       mirror_shallow);
958 959
    }

960
    if (rc < 0)
961 962 963
        goto cleanup;

    diskPriv->migrating = true;
964
    qemuBlockJobStarted(job, vm);
965 966 967 968

    ret = 0;

 cleanup:
969
    qemuBlockJobStartupFinalize(vm, job);
970 971 972 973 974
    VIR_FREE(diskAlias);
    return ret;
}


975
/**
976
 * qemuMigrationSrcNBDStorageCopy:
977 978 979 980
 * @driver: qemu driver
 * @vm: domain
 * @mig: migration cookie
 * @host: where are we migrating to
981
 * @speed: bandwidth limit in MiB/s
982 983
 * @migrate_flags: migrate monitor command flags
 *
984 985 986
 * 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
987
 * to do the very same operation. On failure, the caller is
988 989
 * expected to call qemuMigrationSrcNBDCopyCancel to stop all
 * running copy operations.
990 991 992 993 994
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
995 996 997 998 999 1000 1001 1002
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,
1003 1004 1005
                               virConnectPtr dconn,
                               const char *tlsAlias,
                               unsigned int flags)
1006 1007 1008
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int port;
1009
    size_t i;
1010
    unsigned long long mirror_speed = speed;
1011
    bool mirror_shallow = *migrate_flags & QEMU_MONITOR_MIGRATE_NON_SHARED_INC;
1012
    int rv;
1013
    g_autoptr(virQEMUDriverConfig) cfg = virQEMUDriverGetConfig(driver);
1014 1015

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
1016

1017 1018 1019 1020
    if (mirror_speed > LLONG_MAX >> 20) {
        virReportError(VIR_ERR_OVERFLOW,
                       _("bandwidth must be less than %llu"),
                       LLONG_MAX >> 20);
1021
        return -1;
1022 1023 1024
    }
    mirror_speed <<= 20;

1025 1026 1027 1028 1029 1030 1031
    /* steal NBD port and thus prevent its propagation back to destination */
    port = mig->nbd->port;
    mig->nbd->port = 0;

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

1032
        /* check whether disk should be migrated */
1033
        if (!qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1034 1035
            continue;

1036
        if (qemuMigrationSrcNBDStorageCopyOne(driver, vm, disk, host, port,
1037
                                              mirror_speed, mirror_shallow,
1038
                                              tlsAlias, flags) < 0)
1039
            return -1;
1040

1041
        if (virDomainObjSave(vm, driver->xmlopt, cfg->stateDir) < 0) {
1042
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
1043
            return -1;
1044
        }
1045
    }
1046

1047
    while ((rv = qemuMigrationSrcNBDStorageCopyReady(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) != 1) {
1048
        if (rv < 0)
1049
            return -1;
1050

1051
        if (priv->job.abortJob) {
1052
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1053 1054 1055
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
1056
            return -1;
1057
        }
1058

1059 1060 1061
        if (dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
1062
            return -1;
1063 1064
        }

1065
        if (virDomainObjWait(vm) < 0)
1066
            return -1;
1067 1068
    }

1069 1070
    qemuMigrationSrcFetchMirrorStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                     priv->job.current);
1071

1072
    /* Okay, all disks are ready. Modify migrate_flags */
1073 1074 1075
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);

1076
    return 0;
1077
}
1078

1079

1080
/**
1081
 * qemuMigrationSrcIsAllowedHostdev:
1082 1083
 * @def: domain definition
 *
1084
 * Checks that @def does not contain any host devices unsupported across
1085 1086 1087
 * migrations. Returns true if the vm is allowed to migrate.
 */
static bool
1088
qemuMigrationSrcIsAllowedHostdev(const virDomainDef *def)
1089 1090 1091 1092 1093 1094 1095
{
    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];
1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143
        switch ((virDomainHostdevMode)hostdev->mode) {
        case VIR_DOMAIN_HOSTDEV_MODE_CAPABILITIES:
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                           _("cannot migrate a domain with <hostdev mode='capabilities'>"));
            return false;

        case VIR_DOMAIN_HOSTDEV_MODE_SUBSYS:
            switch ((virDomainHostdevSubsysType)hostdev->source.subsys.type) {
            case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB:
                /* USB devices can be "migrated" */
                continue;

            case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_SCSI:
            case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_SCSI_HOST:
            case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_MDEV:
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED,
                               _("cannot migrate a domain with <hostdev mode='subsystem' type='%s'>"),
                               virDomainHostdevSubsysTypeToString(hostdev->source.subsys.type));
                return false;

            case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI:
                /*
                 * if this is a network interface with <teaming
                 * type='transient'>, migration *is* allowed because
                 * the device will be auto-unplugged by QEMU during
                 * migration.
                 */
                if (hostdev->parentnet &&
                    hostdev->parentnet->teaming.type == VIR_DOMAIN_NET_TEAMING_TYPE_TRANSIENT) {
                    continue;
                }

                /* all other PCI hostdevs can't be migrated */
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED,
                               _("cannot migrate a domain with <hostdev mode='subsystem' type='%s'>"),
                               virDomainHostdevSubsysTypeToString(hostdev->source.subsys.type));
                return false;

            case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_LAST:
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("invalid hostdev subsystem type"));
                return false;
            }
            break;

        case VIR_DOMAIN_HOSTDEV_MODE_LAST:
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("invalid hostdev mode"));
1144 1145 1146 1147 1148 1149 1150 1151
            return false;
        }
    }

    return true;
}


1152
/**
1153
 * qemuMigrationSrcIsAllowed:
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165
 * @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.
 */
1166
bool
1167 1168 1169 1170
qemuMigrationSrcIsAllowed(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          bool remote,
                          unsigned int flags)
1171
{
1172
    int nsnapshots;
1173
    int pauseReason;
1174
    size_t i;
1175

1176 1177 1178 1179 1180
    /* perform these checks only when migrating to remote hosts */
    if (remote) {
        nsnapshots = virDomainSnapshotObjListNum(vm->snapshots, NULL, 0);
        if (nsnapshots < 0)
            return false;
1181

1182 1183 1184 1185 1186
        if (nsnapshots > 0) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("cannot migrate domain with %d snapshots"),
                           nsnapshots);
            return false;
1187
        }
1188 1189
    }

1190 1191
    /* following checks don't make sense for offline migration */
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206
        if (remote) {
            /* cancel migration if disk I/O error is emitted while migrating */
            if (flags & VIR_MIGRATE_ABORT_ON_ERROR &&
                virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
                pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("cannot migrate domain with I/O error"));
                return false;
            }

            if (qemuProcessAutoDestroyActive(driver, vm)) {
                virReportError(VIR_ERR_OPERATION_INVALID,
                               "%s", _("domain is marked for auto destroy"));
                return false;
            }
1207
        }
1208

1209

1210 1211
        if (qemuDomainHasBlockjob(vm, false)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1212
                           _("domain has active block job"));
1213 1214 1215
            return false;
        }

1216
        if (!qemuMigrationSrcIsAllowedHostdev(vm->def))
1217 1218
            return false;

1219
        if (vm->def->cpu) {
1220 1221 1222
            /* QEMU blocks migration and save with invariant TSC enabled
             * unless TSC frequency is explicitly set.
             */
1223 1224
            if (virCPUCheckFeature(vm->def->os.arch, vm->def->cpu,
                                   "invtsc") == 1) {
1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242
                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;
                }
1243
            }
J
Ján Tomko 已提交
1244 1245
        }

1246 1247 1248
        /* Verify that memory device config can be transferred reliably */
        for (i = 0; i < vm->def->nmems; i++) {
            virDomainMemoryDefPtr mem = vm->def->mems[i];
1249

1250 1251 1252 1253 1254
            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"));
1255

1256 1257
                return false;
            }
1258
        }
1259 1260 1261 1262 1263 1264

        if (vm->def->nshmems) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("migration with shmem device is not supported"));
            return false;
        }
1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275

        for (i = 0; i < vm->def->nnets; i++) {
            virDomainNetDefPtr net = vm->def->nets[i];
            qemuSlirpPtr slirp = QEMU_DOMAIN_NETWORK_PRIVATE(net)->slirp;

            if (slirp && !qemuSlirpHasFeature(slirp, QEMU_SLIRP_FEATURE_MIGRATE)) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("a slirp-helper cannot be migrated"));
                return false;
            }
        }
1276 1277 1278 1279 1280 1281 1282 1283 1284 1285

        for (i = 0; i < vm->def->nfss; i++) {
            virDomainFSDefPtr fs = vm->def->fss[i];

            if (fs->fsdriver == VIR_DOMAIN_FS_DRIVER_TYPE_VIRTIOFS) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("migration with virtiofs device is not supported"));
                return false;
            }
        }
1286 1287
    }

1288 1289 1290
    return true;
}

1291
static bool
1292
qemuMigrationSrcIsSafe(virDomainDefPtr def,
1293
                       virQEMUCapsPtr qemuCaps,
1294 1295 1296
                       size_t nmigrate_disks,
                       const char **migrate_disks,
                       unsigned int flags)
1297

1298
{
1299 1300
    bool storagemigration = flags & (VIR_MIGRATE_NON_SHARED_DISK |
                                     VIR_MIGRATE_NON_SHARED_INC);
1301
    size_t i;
1302
    int rc;
1303

1304
    for (i = 0; i < def->ndisks; i++) {
1305
        virDomainDiskDefPtr disk = def->disks[i];
1306
        const char *src = virDomainDiskGetSource(disk);
1307 1308
        int actualType = virStorageSourceGetActualType(disk->src);
        bool unsafe = false;
1309

1310 1311
        /* Disks without any source (i.e. floppies and CD-ROMs)
         * OR readonly are safe. */
1312
        if (virStorageSourceIsEmpty(disk->src) ||
1313
            disk->src->readonly)
1314
            continue;
1315

1316
        /* Disks which are migrated by qemu are safe too. */
1317
        if (storagemigration &&
1318
            qemuMigrationAnyCopyDisk(disk, nmigrate_disks, migrate_disks))
1319
            continue;
1320

1321
        /* However, disks on local FS (e.g. ext4) are not safe. */
1322 1323
        switch ((virStorageType) actualType) {
        case VIR_STORAGE_TYPE_FILE:
1324
            if ((rc = virFileIsSharedFS(src)) < 0) {
1325
                return false;
1326
            } else if (rc == 0) {
1327
                unsafe = true;
1328
            }
1329 1330 1331 1332
            if ((rc = virStorageFileIsClusterFS(src)) < 0)
                return false;
            else if (rc == 1)
                continue;
1333 1334
            break;
        case VIR_STORAGE_TYPE_NETWORK:
1335
            /* But network disks are safe again. */
1336
            continue;
1337

1338 1339 1340 1341
        case VIR_STORAGE_TYPE_NVME:
            unsafe = true;
            break;

1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353
        case VIR_STORAGE_TYPE_NONE:
        case VIR_STORAGE_TYPE_BLOCK:
        case VIR_STORAGE_TYPE_DIR:
        case VIR_STORAGE_TYPE_VOLUME:
        case VIR_STORAGE_TYPE_LAST:
            break;
        }

        if (unsafe) {
            virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                           _("Migration without shared storage is unsafe"));
            return false;
1354
        }
1355

1356 1357 1358 1359 1360 1361 1362
        /* 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;

1363 1364 1365 1366 1367
        if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_MIGRATION_FILE_DROP_CACHE)) {
            VIR_DEBUG("QEMU supports flushing caches; migration is safe");
            continue;
        }

1368 1369
        virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                       _("Migration may lead to data corruption if disks"
1370
                         " use cache other than none or directsync"));
1371
        return false;
1372 1373 1374 1375 1376
    }

    return true;
}

1377 1378

void
1379 1380
qemuMigrationAnyPostcopyFailed(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400
{
    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) {
        if (qemuProcessStopCPUs(driver, vm,
                                VIR_DOMAIN_PAUSED_POSTCOPY_FAILED,
1401
                                QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
1402 1403 1404 1405 1406 1407 1408 1409
            VIR_WARN("Unable to pause guest CPUs for %s", vm->def->name);
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
    }
}


1410
static int
1411
qemuMigrationSrcWaitForSpice(virDomainObjPtr vm)
1412 1413 1414
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

1415
    if (!priv->job.spiceMigration)
1416 1417
        return 0;

1418
    VIR_DEBUG("Waiting for SPICE to finish migration");
1419 1420
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
1421 1422 1423 1424
            return -1;
    }
    return 0;
}
1425

1426 1427 1428 1429

static void
qemuMigrationUpdateJobType(qemuDomainJobInfoPtr jobInfo)
{
1430
    switch ((qemuMonitorMigrationStatus) jobInfo->stats.mig.status) {
1431 1432 1433 1434
    case QEMU_MONITOR_MIGRATION_STATUS_POSTCOPY:
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_POSTCOPY;
        break;

1435
    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
1436
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED;
1437 1438 1439
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
1440
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_NONE;
1441 1442 1443
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
1444
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1445 1446 1447
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
1448
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1449 1450
        break;

1451 1452 1453 1454 1455 1456 1457 1458
    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;

1459 1460 1461
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLING:
1462
    case QEMU_MONITOR_MIGRATION_STATUS_WAIT_UNPLUG:
1463
    case QEMU_MONITOR_MIGRATION_STATUS_LAST:
1464 1465 1466 1467 1468 1469
        break;
    }
}


int
1470 1471 1472 1473 1474
qemuMigrationAnyFetchStats(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuDomainAsyncJob asyncJob,
                           qemuDomainJobInfoPtr jobInfo,
                           char **error)
1475 1476
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1477
    qemuMonitorMigrationStats stats;
1478 1479 1480 1481 1482
    int rv;

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

1483
    rv = qemuMonitorGetMigrationStats(priv->mon, &stats, error);
1484 1485 1486 1487

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

1488
    jobInfo->stats.mig = stats;
1489 1490

    return 0;
1491 1492 1493
}


1494 1495 1496 1497 1498 1499 1500
static const char *
qemuMigrationJobName(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
1501
        return _("migration out job");
1502 1503 1504 1505
    case QEMU_ASYNC_JOB_SAVE:
        return _("domain save job");
    case QEMU_ASYNC_JOB_DUMP:
        return _("domain core dump job");
1506 1507 1508 1509 1510 1511 1512 1513
    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");
1514 1515
    case QEMU_ASYNC_JOB_BACKUP:
        return _("backup job");
1516
    case QEMU_ASYNC_JOB_LAST:
1517 1518 1519 1520 1521 1522
    default:
        return _("job");
    }
}


1523
static int
1524
qemuMigrationJobCheckStatus(virQEMUDriverPtr driver,
1525
                            virDomainObjPtr vm,
1526
                            qemuDomainAsyncJob asyncJob)
1527 1528 1529
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1530
    char *error = NULL;
1531
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1532
    int ret = -1;
1533

1534
    if (!events ||
1535
        jobInfo->stats.mig.status == QEMU_MONITOR_MIGRATION_STATUS_ERROR) {
1536
        if (qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, &error) < 0)
1537 1538
            return -1;
    }
1539

1540 1541
    qemuMigrationUpdateJobType(jobInfo);

1542 1543
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_NONE:
1544 1545
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("is not active"));
1546
        goto cleanup;
1547

1548
    case QEMU_DOMAIN_JOB_STATUS_FAILED:
1549
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
1550 1551 1552
                       qemuMigrationJobName(vm),
                       error ? error : _("unexpectedly failed"));
        goto cleanup;
1553

1554
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
1555 1556
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("canceled by client"));
1557
        goto cleanup;
1558

1559 1560
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
1561
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
1562
    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
1563
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1564
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1565 1566
        break;
    }
1567 1568 1569 1570 1571 1572

    ret = 0;

 cleanup:
    VIR_FREE(error);
    return ret;
1573 1574 1575
}


1576 1577
enum qemuMigrationCompletedFlags {
    QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR = (1 << 0),
1578
    /* This flag should only be set when run on src host */
1579
    QEMU_MIGRATION_COMPLETED_CHECK_STORAGE  = (1 << 1),
1580
    QEMU_MIGRATION_COMPLETED_POSTCOPY       = (1 << 2),
1581
    QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER = (1 << 3),
1582 1583
};

1584

1585 1586 1587 1588 1589 1590 1591
/**
 * 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
1592 1593 1594 1595 1596
qemuMigrationAnyCompleted(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          qemuDomainAsyncJob asyncJob,
                          virConnectPtr dconn,
                          unsigned int flags)
1597 1598 1599 1600 1601
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    int pauseReason;

1602
    if (qemuMigrationJobCheckStatus(driver, vm, asyncJob) < 0)
1603 1604
        goto error;

1605
    /* This flag should only be set when run on src host */
1606
    if (flags & QEMU_MIGRATION_COMPLETED_CHECK_STORAGE &&
1607
        qemuMigrationSrcNBDStorageCopyReady(vm, asyncJob) < 0)
1608 1609
        goto error;

1610
    if (flags & QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR &&
1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623
        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;
    }

1624 1625 1626 1627 1628 1629 1630 1631 1632 1633
    /* 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;
    }

1634 1635 1636 1637 1638
    /* 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 &&
1639
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
1640 1641 1642 1643
        VIR_DEBUG("Migration switched to post-copy");
        return 1;
    }

1644
    if (jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
1645 1646 1647 1648 1649
        return 1;
    else
        return 0;

 error:
1650 1651 1652
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1653
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1654
        /* The migration was aborted by us rather than QEMU itself. */
1655
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1656
        return -2;
1657 1658 1659

    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
        /* Something failed after QEMU already finished the migration. */
1660
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1661
        return -1;
1662 1663 1664 1665

    case QEMU_DOMAIN_JOB_STATUS_FAILED:
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
        /* QEMU aborted the migration. */
1666
        return -1;
1667 1668 1669 1670 1671 1672

    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_NONE:
        /* Impossible. */
        break;
1673
    }
1674 1675

    return -1;
1676 1677 1678
}


1679 1680 1681
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
1682
static int
1683 1684 1685 1686 1687
qemuMigrationSrcWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  virConnectPtr dconn,
                                  unsigned int flags)
1688
{
1689
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
1690
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1691
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1692
    int rv;
1693

1694 1695
    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_MIGRATING;

1696 1697
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           dconn, flags)) != 1) {
1698 1699
        if (rv < 0)
            return rv;
1700

1701 1702
        if (events) {
            if (virDomainObjWait(vm) < 0) {
1703 1704
                if (virDomainObjIsActive(vm))
                    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1705 1706 1707 1708 1709 1710 1711 1712 1713 1714
                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);
        }
1715 1716
    }

1717
    if (events)
1718
        ignore_value(qemuMigrationAnyFetchStats(driver, vm, asyncJob, jobInfo, NULL));
1719

1720
    qemuDomainJobInfoUpdateTime(jobInfo);
1721 1722
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
1723
    if (VIR_ALLOC(priv->job.completed) == 0) {
1724
        *priv->job.completed = *jobInfo;
1725 1726
        priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
    }
1727

1728 1729 1730 1731
    if (asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT &&
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;

1732
    return 0;
1733 1734 1735
}


1736
static int
1737 1738 1739 1740
qemuMigrationDstWaitForCompletion(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob,
                                  bool postcopy)
1741 1742
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1743
    unsigned int flags = 0;
1744 1745 1746 1747 1748 1749 1750
    int rv;

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

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

1751 1752 1753
    if (postcopy)
        flags = QEMU_MIGRATION_COMPLETED_POSTCOPY;

1754 1755
    while ((rv = qemuMigrationAnyCompleted(driver, vm, asyncJob,
                                           NULL, flags)) != 1) {
1756 1757 1758 1759 1760 1761 1762 1763
        if (rv < 0 || virDomainObjWait(vm) < 0)
            return -1;
    }

    return 0;
}


1764
static int
1765 1766 1767 1768
qemuMigrationSrcGraphicsRelocate(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuMigrationCookiePtr cookie,
                                 const char *graphicsuri)
1769 1770
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1771 1772
    int ret = -1;
    const char *listenAddress = NULL;
1773
    virSocketAddr addr;
1774 1775 1776 1777 1778
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
1779

1780
    if (!cookie || (!cookie->graphics && !graphicsuri))
1781 1782
        return 0;

1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801
    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) {
1802
        size_t i;
1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829

        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;
            }
        }
    }
1830 1831 1832 1833

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
1834 1835 1836 1837
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
1838

1839 1840 1841 1842 1843 1844 1845 1846 1847
    /* 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;
    }

1848 1849 1850 1851
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
1852
        priv->job.spiceMigration = !ret;
1853 1854
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
1855
    }
1856

1857
 cleanup:
1858
    virURIFree(uri);
1859 1860 1861 1862
    return ret;
}


1863
static int
J
Ján Tomko 已提交
1864
qemuMigrationDstOPDRelocate(virQEMUDriverPtr driver G_GNUC_UNUSED,
1865 1866
                            virDomainObjPtr vm,
                            qemuMigrationCookiePtr cookie)
1867
{
1868
    virDomainNetDefPtr netptr;
1869
    size_t i;
1870 1871 1872 1873 1874 1875 1876 1877

    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:
1878
           break;
1879
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
1880 1881
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
1882 1883 1884
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
1885
                return -1;
1886 1887
            }
            break;
1888 1889 1890 1891 1892
        default:
            break;
        }
    }

1893
    return 0;
1894 1895 1896
}


1897
int
1898 1899
qemuMigrationDstCheckProtocol(virQEMUCapsPtr qemuCaps,
                              const char *migrateFrom)
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922
{
    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 *
1923 1924
qemuMigrationDstGetURI(const char *migrateFrom,
                       int migrateFd)
1925 1926 1927 1928
{
    char *uri = NULL;

    if (STREQ(migrateFrom, "stdio"))
1929
        uri = g_strdup_printf("fd:%d", migrateFd);
1930
    else
1931
        uri = g_strdup(migrateFrom);
1932 1933 1934 1935 1936

    return uri;
}


1937
int
1938 1939 1940 1941
qemuMigrationDstRun(virQEMUDriverPtr driver,
                    virDomainObjPtr vm,
                    const char *uri,
                    qemuDomainAsyncJob asyncJob)
1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    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)
1954
        return -1;
1955 1956

    if (asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN) {
1957
        /* qemuMigrationDstWaitForCompletion is called from the Finish phase */
1958
        return 0;
1959 1960
    }

1961
    if (qemuMigrationDstWaitForCompletion(driver, vm, asyncJob, false) < 0)
1962
        return -1;
1963

1964
    return 0;
1965 1966 1967
}


1968 1969 1970 1971 1972 1973
/* 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.
 */
1974
static void
1975 1976 1977
qemuMigrationSrcCleanup(virDomainObjPtr vm,
                        virConnectPtr conn,
                        void *opaque)
1978
{
1979
    virQEMUDriverPtr driver = opaque;
1980 1981 1982 1983 1984 1985 1986 1987 1988
    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))
1989
        return;
1990 1991 1992 1993 1994

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

1995
    switch ((qemuMigrationJobPhase) priv->job.phase) {
1996 1997 1998 1999 2000 2001 2002 2003 2004
    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);
2005
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
2006
                                 priv->job.migParams, priv->job.apiFlags);
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028
        /* 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 */
        ;
    }
}

2029

2030
/* The caller is supposed to lock the vm and start a migration job. */
2031
static char *
2032 2033 2034 2035 2036 2037 2038 2039 2040
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)
2041 2042 2043
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
2044
    virDomainDefPtr def = NULL;
2045
    qemuDomainObjPrivatePtr priv = vm->privateData;
2046
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
2047

2048
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
2049
              " cookieout=%p, cookieoutlen=%p,"
2050
              " nmigrate_disks=%zu, migrate_disks=%p, flags=0x%lx",
2051
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
2052 2053
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
2054

2055 2056 2057 2058 2059 2060
    /* 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);
2061

2062
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
2063 2064
        goto cleanup;

2065
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
2066 2067
        !qemuMigrationSrcIsSafe(vm->def, priv->qemuCaps,
                                nmigrate_disks, migrate_disks, flags))
2068 2069
        goto cleanup;

2070 2071 2072 2073 2074 2075 2076 2077 2078
    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;
    }

2079 2080 2081 2082 2083 2084
    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;
    }

2085 2086
    if (flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC)) {
        if (nmigrate_disks) {
2087 2088 2089 2090 2091 2092
            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;
2093 2094
                }

2095 2096 2097 2098
                if (j == vm->def->ndisks) {
                    virReportError(VIR_ERR_INVALID_ARG,
                                   _("disk target %s not found"),
                                   migrate_disks[i]);
2099 2100
                    goto cleanup;
                }
2101 2102 2103
            }

            if (flags & VIR_MIGRATE_TUNNELLED) {
2104
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
2105 2106
                               _("Selecting disks to migrate is not "
                                 "implemented for tunnelled migration"));
2107 2108 2109 2110
                goto cleanup;
            }
        }

2111 2112 2113 2114 2115 2116
        /* 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;
2117 2118 2119
        }
    }

2120
    if (virDomainDefHasMemoryHotplug(vm->def) ||
2121
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
2122
         vm->newDef && virDomainDefHasMemoryHotplug(vm->newDef)))
2123 2124
        cookieFlags |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

2125 2126 2127 2128 2129
    if (!qemuDomainVcpuHotplugIsInOrder(vm->def) ||
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
         vm->newDef && !qemuDomainVcpuHotplugIsInOrder(vm->newDef)))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU_HOTPLUG;

2130 2131 2132
    if (priv->origCPU)
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU;

2133 2134
    cookieFlags |= QEMU_MIGRATION_COOKIE_ALLOW_REBOOT;

2135 2136 2137
    if (!(flags & VIR_MIGRATE_OFFLINE))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CAPS;

2138 2139
    if (!(mig = qemuMigrationEatCookie(driver, vm->def,
                                       priv->origname, priv, NULL, 0, 0)))
2140 2141 2142
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
2143
                                QEMU_MIGRATION_SOURCE,
2144
                                cookieout, cookieoutlen,
2145
                                cookieFlags) < 0)
2146 2147
        goto cleanup;

L
liguang 已提交
2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169
    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;
        }
    }

2170
    if (xmlin) {
2171
        if (!(def = virDomainDefParseString(xmlin, driver->xmlopt, priv->qemuCaps,
2172 2173
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                            VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2174 2175
            goto cleanup;

2176
        if (!qemuDomainCheckABIStability(driver, vm, def))
2177 2178
            goto cleanup;

2179
        rv = qemuDomainDefFormatLive(driver, priv->qemuCaps, def, NULL, false, true);
2180
    } else {
2181
        rv = qemuDomainDefFormatLive(driver, priv->qemuCaps, vm->def, priv->origCPU,
2182
                                     false, true);
2183
    }
2184

2185
 cleanup:
2186
    qemuMigrationCookieFree(mig);
2187
    virDomainDefFree(def);
2188 2189 2190
    return rv;
}

2191
char *
2192 2193 2194 2195 2196 2197 2198 2199 2200
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)
2201 2202 2203
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
2204
    qemuDomainAsyncJob asyncJob;
2205 2206

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
2207 2208
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
2209 2210 2211 2212 2213 2214 2215 2216
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_MIGRATION_OUT;
    } else {
        if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_NONE;
    }

2217
    qemuMigrationSrcStoreDomainState(vm);
2218

2219
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
2220 2221 2222 2223 2224 2225
        goto endjob;

    /* Check if there is any ejected media.
     * We don't want to require them on the destination.
     */
    if (!(flags & VIR_MIGRATE_OFFLINE) &&
2226
        qemuProcessRefreshDisks(driver, vm, asyncJob) < 0)
2227 2228
        goto endjob;

2229 2230 2231
    if (!(xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                           cookieout, cookieoutlen,
                                           nmigrate_disks, migrate_disks, flags)))
2232 2233 2234 2235 2236 2237 2238
        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.
         */
2239
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
2240
                                 qemuMigrationSrcCleanup) < 0) {
2241
            VIR_FREE(xml);
2242
            goto endjob;
2243
        }
2244
        qemuMigrationJobContinue(vm);
2245 2246 2247 2248
    } else {
        goto endjob;
    }

2249
 cleanup:
M
Michal Privoznik 已提交
2250
    virDomainObjEndAPI(&vm);
2251 2252
    return xml;

2253
 endjob:
2254 2255 2256 2257
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2258 2259 2260
    goto cleanup;
}

2261

2262 2263
/* Prepare is the first step, and it runs on the destination host.
 */
2264

2265
static void
2266 2267
qemuMigrationDstPrepareCleanup(virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
2268 2269 2270 2271 2272 2273 2274 2275 2276
{
    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));

2277
    virPortAllocatorRelease(priv->migrationPort);
2278 2279
    priv->migrationPort = 0;

2280 2281 2282 2283 2284
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2285
static qemuProcessIncomingDefPtr
2286 2287 2288 2289 2290 2291
qemuMigrationDstPrepare(virDomainObjPtr vm,
                        bool tunnel,
                        const char *protocol,
                        const char *listenAddress,
                        unsigned short port,
                        int fd)
2292 2293
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2294
    qemuProcessIncomingDefPtr inc = NULL;
2295 2296 2297
    char *migrateFrom = NULL;

    if (tunnel) {
2298
        migrateFrom = g_strdup("stdio");
2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323
    } 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 */
            }
2324
        } else if (hostIPv6Capable) {
2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342
            /* 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";
2343
        migrateFrom = g_strdup_printf(incFormat, protocol, listenAddress, port);
2344 2345
    }

2346 2347
    inc = qemuProcessIncomingDefNew(priv->qemuCaps, listenAddress,
                                    migrateFrom, fd, NULL);
2348

2349
 cleanup:
2350 2351
    VIR_FREE(migrateFrom);
    return inc;
2352 2353
}

2354
static int
2355
qemuMigrationDstPrepareAny(virQEMUDriverPtr driver,
2356
                           virConnectPtr dconn,
2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370
                           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,
2371
                           qemuMigrationParamsPtr migParams,
2372
                           unsigned long flags)
2373 2374
{
    virDomainObjPtr vm = NULL;
2375
    virObjectEventPtr event = NULL;
2376
    virErrorPtr origErr;
2377
    int ret = -1;
2378
    int dataFD[2] = { -1, -1 };
2379
    qemuDomainObjPrivatePtr priv = NULL;
2380
    qemuMigrationCookiePtr mig = NULL;
2381
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2382
    char *xmlout = NULL;
L
liguang 已提交
2383
    unsigned int cookieFlags;
2384
    unsigned int startFlags;
2385
    qemuProcessIncomingDefPtr incoming = NULL;
2386
    bool taint_hook = false;
2387 2388 2389
    bool stopProcess = false;
    bool relabel = false;
    int rv;
2390
    char *tlsAlias = NULL;
2391

2392 2393
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413
    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;
        }
2414 2415
        cookieFlags = 0;
    } else {
2416 2417
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS |
                      QEMU_MIGRATION_COOKIE_CAPS;
L
liguang 已提交
2418 2419
    }

2420 2421 2422 2423 2424 2425 2426 2427 2428
    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;
    }

2429 2430 2431 2432 2433 2434
    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;
    }

2435
    if (!qemuMigrationSrcIsAllowedHostdev(*def))
2436 2437
        goto cleanup;

J
Jiri Denemark 已提交
2438 2439 2440 2441 2442
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2443
        if (!(xml = qemuDomainDefFormatXML(driver, NULL, *def,
2444 2445
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2446 2447
            goto cleanup;

2448
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2449 2450 2451 2452 2453 2454 2455
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2456
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2457 2458 2459 2460 2461 2462
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2463
                newdef = virDomainDefParseString(xmlout, driver->xmlopt, NULL,
2464 2465
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                                 VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE);
J
Jiri Denemark 已提交
2466 2467 2468
                if (!newdef)
                    goto cleanup;

2469
                if (!qemuDomainDefCheckABIStability(driver, NULL, *def, newdef)) {
J
Jiri Denemark 已提交
2470 2471 2472 2473
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2474 2475
                virDomainDefFree(*def);
                *def = newdef;
2476 2477 2478 2479
                /* 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
2480 2481 2482 2483
            }
        }
    }

2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497
    /* 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;

2498
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2499
                                   driver->xmlopt,
2500 2501 2502
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2503
        goto cleanup;
2504
    *def = NULL;
2505

2506
    priv = vm->privateData;
2507
    priv->origname = g_strdup(origname);
2508

2509 2510 2511 2512 2513
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2514 2515
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
2516 2517 2518 2519 2520 2521
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2522 2523 2524
    if (qemuMigrationDstPrecreateStorage(vm, mig->nbd,
                                         nmigrate_disks, migrate_disks,
                                         !!(flags & VIR_MIGRATE_NON_SHARED_INC)) < 0)
2525 2526
        goto cleanup;

2527 2528
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                              flags) < 0)
2529
        goto cleanup;
2530
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2531 2532 2533 2534

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

L
liguang 已提交
2535 2536 2537
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2538
    if (tunnel &&
2539
        virPipe(dataFD) < 0)
2540
        goto stopjob;
2541

2542 2543
    startFlags = VIR_QEMU_PROCESS_START_AUTODESTROY;

2544
    if (qemuProcessInit(driver, vm, mig->cpu, QEMU_ASYNC_JOB_MIGRATION_IN,
2545
                        true, startFlags) < 0)
2546
        goto stopjob;
2547
    stopProcess = true;
2548

2549 2550
    priv->allowReboot = mig->allowReboot;

2551 2552 2553
    if (!(incoming = qemuMigrationDstPrepare(vm, tunnel, protocol,
                                             listenAddress, port,
                                             dataFD[0])))
2554
        goto stopjob;
2555

2556
    if (qemuProcessPrepareDomain(driver, vm, startFlags) < 0)
2557 2558
        goto stopjob;

2559
    if (qemuProcessPrepareHost(driver, vm, startFlags) < 0)
2560 2561
        goto stopjob;

2562
    rv = qemuProcessLaunch(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2563 2564
                           incoming, NULL,
                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
2565
                           startFlags);
2566 2567 2568
    if (rv < 0) {
        if (rv == -2)
            relabel = true;
2569
        goto stopjob;
2570
    }
2571
    relabel = true;
2572

2573 2574 2575 2576
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
2577
            goto stopjob;
2578
        }
2579
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
2580 2581
    }

2582
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
2583
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
2584
        goto stopjob;
M
Michael R. Hines 已提交
2585 2586
    }

2587
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2588
                                 migParams, mig->caps->automatic) < 0)
2589 2590 2591 2592 2593 2594 2595
        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,
2596
                                         &tlsAlias, NULL,
2597 2598 2599 2600 2601 2602 2603
                                         migParams) < 0)
            goto stopjob;
    } else {
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
            goto stopjob;
    }

2604 2605
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                                 migParams) < 0)
2606 2607
        goto stopjob;

2608 2609 2610
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622
        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;
        }

2623 2624
        if (qemuMigrationDstStartNBDServer(driver, vm, incoming->address,
                                           nmigrate_disks, migrate_disks,
2625
                                           nbdPort, nbdTLSAlias) < 0) {
2626
            goto stopjob;
2627
        }
2628
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
2629 2630
    }

2631 2632 2633 2634 2635 2636 2637 2638 2639
    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");
    }

2640
    if (incoming->deferredURI &&
2641 2642
        qemuMigrationDstRun(driver, vm, incoming->deferredURI,
                            QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2643 2644
        goto stopjob;

2645
    if (qemuProcessFinishStartup(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2646 2647 2648
                                 false, VIR_DOMAIN_PAUSED_MIGRATION) < 0)
        goto stopjob;

2649
 done:
2650 2651 2652
    if (qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_DESTINATION,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
2653 2654 2655 2656 2657 2658 2659
        /* 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");
    }

2660
    if (qemuDomainCleanupAdd(vm, qemuMigrationDstPrepareCleanup) < 0)
2661
        goto stopjob;
2662

L
liguang 已提交
2663 2664
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
2665
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
2666 2667 2668
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
2669

2670 2671 2672 2673
    /* 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.
     */
2674
    qemuMigrationJobContinue(vm);
2675

2676 2677
    if (autoPort)
        priv->migrationPort = port;
2678 2679 2680 2681 2682
    /* 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;
2683
    ret = 0;
2684

2685
 cleanup:
2686
    virErrorPreserveLast(&origErr);
2687
    VIR_FREE(tlsAlias);
2688
    qemuProcessIncomingDefFree(incoming);
J
Jiri Denemark 已提交
2689
    VIR_FREE(xmlout);
2690 2691
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
2692 2693 2694
    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 */
2695
        VIR_FREE(priv->origname);
2696 2697 2698 2699
        /* release if port is auto selected which is not the case if
         * it is given in parameters
         */
        if (nbdPort == 0)
2700
            virPortAllocatorRelease(priv->nbdPort);
2701
        priv->nbdPort = 0;
2702
        virDomainObjRemoveTransientDef(vm);
2703
        qemuDomainRemoveInactiveJob(driver, vm);
2704
    }
M
Michal Privoznik 已提交
2705
    virDomainObjEndAPI(&vm);
2706
    virObjectEventStateQueue(driver->domainEventState, event);
2707
    qemuMigrationCookieFree(mig);
2708
    virNWFilterUnlockFilterUpdates();
2709
    virErrorRestore(&origErr);
2710
    return ret;
2711

2712
 stopjob:
2713
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2714
                             priv->job.migParams, priv->job.apiFlags);
2715

2716 2717 2718 2719
    if (stopProcess) {
        unsigned int stopFlags = VIR_QEMU_PROCESS_STOP_MIGRATED;
        if (!relabel)
            stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
2720
        virDomainAuditStart(vm, "migrated", false);
2721 2722
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        QEMU_ASYNC_JOB_MIGRATION_IN, stopFlags);
2723
    }
2724

2725
    qemuMigrationJobFinish(driver, vm);
2726
    goto cleanup;
2727 2728 2729
}


2730 2731 2732 2733 2734
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
2735
qemuMigrationDstPrepareTunnel(virQEMUDriverPtr driver,
2736
                              virConnectPtr dconn,
2737 2738 2739 2740 2741 2742 2743
                              const char *cookiein,
                              int cookieinlen,
                              char **cookieout,
                              int *cookieoutlen,
                              virStreamPtr st,
                              virDomainDefPtr *def,
                              const char *origname,
2744
                              qemuMigrationParamsPtr migParams,
2745
                              unsigned long flags)
2746
{
2747
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2748
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
2749
              "origname=%s, flags=0x%lx",
2750
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2751
              cookieout, cookieoutlen, st, *def, origname, flags);
2752

2753 2754 2755 2756 2757 2758
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

2759 2760 2761
    return qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
                                      cookieout, cookieoutlen, def, origname,
                                      st, NULL, 0, false, NULL, 0, NULL, 0,
2762
                                      migParams, flags);
2763 2764 2765
}


2766
static virURIPtr
2767
qemuMigrationAnyParseURI(const char *uri, bool *wellFormed)
2768 2769 2770 2771 2772 2773 2774
{
    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, "//")) {
2775
        tmp = g_strdup_printf("tcp://%s", uri + 4);
2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787
        uri = tmp;
    }

    parsed = virURIParse(uri);
    if (parsed && wellFormed)
        *wellFormed = !tmp;
    VIR_FREE(tmp);

    return parsed;
}


2788
int
2789
qemuMigrationDstPrepareDirect(virQEMUDriverPtr driver,
2790
                              virConnectPtr dconn,
2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802
                              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,
2803
                              qemuMigrationParamsPtr migParams,
2804
                              unsigned long flags)
2805
{
2806 2807
    unsigned short port = 0;
    bool autoPort = true;
2808 2809
    char *hostname = NULL;
    int ret = -1;
2810
    virURIPtr uri = NULL;
2811 2812
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
2813

2814
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2815
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
2816
              "def=%p, origname=%s, listenAddress=%s, "
2817
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, flags=0x%lx",
2818
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2819
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
2820
              *def, origname, NULLSTR(listenAddress),
2821
              nmigrate_disks, migrate_disks, nbdPort, flags);
2822

2823 2824
    *uri_out = NULL;

2825 2826 2827
    /* 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
2828 2829 2830
     * 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".
2831 2832 2833 2834 2835 2836
     *
     * 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) {
2837 2838 2839
        bool encloseAddress = false;
        const char *incFormat;

2840
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2841
            goto cleanup;
2842

2843
        if (migrateHost != NULL) {
2844 2845
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
2846

2847
            hostname = g_strdup(migrateHost);
2848 2849 2850 2851
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
2852 2853

        if (STRPREFIX(hostname, "localhost")) {
2854 2855 2856
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
2857 2858 2859 2860 2861
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
2862
         * compatibility with old targets. We at least make the
2863 2864
         * new targets accept both syntaxes though.
         */
2865 2866 2867 2868 2869
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

2870
        *uri_out = g_strdup_printf(incFormat, "tcp", hostname, port);
2871
    } else {
2872
        bool well_formed_uri;
J
Ján Tomko 已提交
2873

2874
        if (!(uri = qemuMigrationAnyParseURI(uri_in, &well_formed_uri)))
2875
            goto cleanup;
J
Ján Tomko 已提交
2876

2877 2878 2879 2880 2881 2882 2883
        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
2884 2885
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
2886 2887 2888
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
2889 2890 2891 2892 2893 2894 2895 2896 2897 2898
            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) {
2899
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2900
                goto cleanup;
2901

2902
            /* Send well-formed URI only if uri_in was well-formed */
2903 2904 2905 2906 2907
            if (well_formed_uri) {
                uri->port = port;
                if (!(*uri_out = virURIFormat(uri)))
                    goto cleanup;
            } else {
2908
                *uri_out = g_strdup_printf("%s:%d", uri_in, port);
2909
            }
2910
        } else {
2911 2912
            port = uri->port;
            autoPort = false;
2913 2914 2915 2916 2917 2918
        }
    }

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

2919
    ret = qemuMigrationDstPrepareAny(driver, dconn, cookiein, cookieinlen,
2920 2921 2922 2923
                                     cookieout, cookieoutlen, def, origname,
                                     NULL, uri ? uri->scheme : "tcp",
                                     port, autoPort, listenAddress,
                                     nmigrate_disks, migrate_disks, nbdPort,
2924
                                     migParams, flags);
2925
 cleanup:
2926
    virURIFree(uri);
2927
    VIR_FREE(hostname);
2928
    virObjectUnref(cfg);
2929
    if (ret != 0) {
2930
        VIR_FREE(*uri_out);
2931
        if (autoPort)
2932
            virPortAllocatorRelease(port);
2933
    }
2934 2935 2936 2937
    return ret;
}


2938
virDomainDefPtr
2939
qemuMigrationAnyPrepareDef(virQEMUDriverPtr driver,
2940
                           virQEMUCapsPtr qemuCaps,
2941 2942 2943
                           const char *dom_xml,
                           const char *dname,
                           char **origname)
2944 2945
{
    virDomainDefPtr def;
2946
    char *name = NULL;
2947 2948 2949 2950 2951 2952 2953

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

2954
    if (!(def = virDomainDefParseString(dom_xml, driver->xmlopt,
2955
                                        qemuCaps,
2956 2957
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                        VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2958 2959 2960
        goto cleanup;

    if (dname) {
2961
        name = def->name;
2962
        def->name = g_strdup(dname);
2963 2964
    }

2965
 cleanup:
2966 2967 2968 2969
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
2970 2971 2972 2973
    return def;
}


2974
static int
2975 2976 2977 2978 2979 2980
qemuMigrationSrcConfirmPhase(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
                             const char *cookiein,
                             int cookieinlen,
                             unsigned int flags,
                             int retcode)
2981 2982
{
    qemuMigrationCookiePtr mig;
2983
    virObjectEventPtr event;
2984 2985
    int rv = -1;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2986 2987
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = NULL;
2988

2989
    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
2990
              "flags=0x%x, retcode=%d",
2991
              driver, vm, NULLSTR(cookiein), cookieinlen,
2992 2993 2994 2995 2996 2997 2998 2999 3000
              flags, retcode);

    virCheckFlags(QEMU_MIGRATION_FLAGS, -1);

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

3001 3002
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen,
3003
                                       QEMU_MIGRATION_COOKIE_STATS)))
3004 3005
        goto cleanup;

3006 3007 3008
    if (retcode == 0)
        jobInfo = priv->job.completed;
    else
3009
        VIR_FREE(priv->job.completed);
3010 3011 3012

    /* Update times with the values sent by the destination daemon */
    if (mig->jobInfo && jobInfo) {
3013 3014 3015 3016 3017 3018 3019 3020
        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 &&
3021 3022
            qemuMigrationAnyFetchStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                       jobInfo, NULL) < 0)
3023 3024
            VIR_WARN("Could not refresh migration statistics");

3025 3026 3027
        qemuDomainJobInfoUpdateTime(jobInfo);
        jobInfo->timeDeltaSet = mig->jobInfo->timeDeltaSet;
        jobInfo->timeDelta = mig->jobInfo->timeDelta;
3028 3029
        jobInfo->stats.mig.downtime_set = mig->jobInfo->stats.mig.downtime_set;
        jobInfo->stats.mig.downtime = mig->jobInfo->stats.mig.downtime;
3030 3031
    }

3032 3033 3034
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

3035 3036
    /* 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.
3037 3038 3039 3040
     */
    if (retcode == 0) {
        /* If guest uses SPICE and supports seamless migration we have to hold
         * up domain shutdown until SPICE server transfers its data */
3041
        qemuMigrationSrcWaitForSpice(vm);
3042 3043

        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
3044
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
3045 3046 3047
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
        virDomainAuditStop(vm, "migrated");

3048
        event = virDomainEventLifecycleNewFromObj(vm,
3049 3050
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
3051
        virObjectEventStateQueue(driver->domainEventState, event);
3052
        qemuDomainEventEmitJobCompleted(driver, vm);
3053
    } else {
3054
        virErrorPtr orig_err;
3055
        int reason;
3056

3057 3058
        virErrorPreserveLast(&orig_err);

3059
        /* cancel any outstanding NBD jobs */
3060 3061
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT, NULL);
3062

3063
        virErrorRestore(&orig_err);
3064

3065
        if (virDomainObjGetState(vm, &reason) == VIR_DOMAIN_PAUSED &&
3066
            reason == VIR_DOMAIN_PAUSED_POSTCOPY)
3067
            qemuMigrationAnyPostcopyFailed(driver, vm);
3068 3069
        else
            qemuMigrationSrcRestoreDomainState(driver, vm);
3070

3071
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3072
                                 priv->job.migParams, priv->job.apiFlags);
3073

3074
        if (virDomainObjSave(vm, driver->xmlopt, cfg->stateDir) < 0)
3075 3076 3077
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
    }

3078
 done:
3079 3080 3081
    qemuMigrationCookieFree(mig);
    rv = 0;

3082
 cleanup:
3083 3084 3085 3086 3087
    virObjectUnref(cfg);
    return rv;
}

int
3088 3089 3090 3091 3092 3093
qemuMigrationSrcConfirm(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *cookiein,
                        int cookieinlen,
                        unsigned int flags,
                        int cancelled)
3094
{
3095
    qemuMigrationJobPhase phase;
3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109
    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);
3110
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
3111
                           qemuMigrationSrcCleanup);
3112

3113 3114 3115
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
3116

3117
    qemuMigrationJobFinish(driver, vm);
3118
    if (!virDomainObjIsActive(vm)) {
3119
        if (!cancelled && ret == 0 && flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
3120
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
3121 3122
            vm->persistent = 0;
        }
3123
        qemuDomainRemoveInactiveJob(driver, vm);
3124 3125
    }

3126
 cleanup:
M
Michal Privoznik 已提交
3127
    virDomainObjEndAPI(&vm);
3128 3129 3130 3131 3132
    virObjectUnref(cfg);
    return ret;
}


3133 3134
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3135
    MIGRATION_DEST_CONNECT_HOST,
3136
    MIGRATION_DEST_FD,
3137
};
3138

3139 3140 3141 3142
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3143

3144 3145 3146 3147 3148 3149
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3150
            const char *protocol;
3151 3152 3153 3154
            const char *name;
            int port;
        } host;

3155 3156 3157 3158
        struct {
            int qemu;
            int local;
        } fd;
3159 3160 3161 3162 3163 3164 3165
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3166 3167 3168

#define TUNNEL_SEND_BUF_SIZE 65536

3169 3170 3171 3172 3173 3174 3175
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3176 3177
    int wakeupRecvFD;
    int wakeupSendFD;
3178 3179
};

3180
static void qemuMigrationSrcIOFunc(void *arg)
3181
{
3182
    qemuMigrationIOThreadPtr data = arg;
3183 3184 3185 3186 3187 3188 3189
    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);
3190

3191
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3192
        goto abrt;
3193

3194 3195 3196
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3197
    for (;;) {
3198 3199 3200 3201 3202
        int ret;

        fds[0].events = fds[1].events = POLLIN;
        fds[0].revents = fds[1].revents = 0;

3203
        ret = poll(fds, G_N_ELEMENTS(fds), timeout);
3204 3205 3206 3207

        if (ret < 0) {
            if (errno == EAGAIN || errno == EINTR)
                continue;
3208
            virReportSystemError(errno, "%s",
3209 3210
                                 _("poll failed in migration tunnel"));
            goto abrt;
3211
        }
3212 3213 3214 3215 3216 3217 3218

        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");
3219
            break;
3220
        }
3221

3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237
        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;
            }
3238 3239
        }

3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256
        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;
            }
        }
    }
3257

3258 3259
    if (virStreamFinish(data->st) < 0)
        goto error;
3260

3261
    VIR_FORCE_CLOSE(data->sock);
3262 3263
    VIR_FREE(buffer);

3264 3265
    return;

3266
 abrt:
3267
    virErrorPreserveLast(&err);
3268 3269 3270 3271 3272
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
3273
    virErrorRestore(&err);
3274

3275
 error:
3276 3277 3278 3279 3280
    /* 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);
3281
    virResetLastError();
3282
    VIR_FREE(buffer);
3283 3284 3285 3286
}


static qemuMigrationIOThreadPtr
3287 3288
qemuMigrationSrcStartTunnel(virStreamPtr st,
                            int sock)
3289
{
3290 3291
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3292

3293
    if (virPipe(wakeupFD) < 0)
3294
        goto error;
3295

3296
    if (VIR_ALLOC(io) < 0)
3297
        goto error;
3298

3299 3300
    io->st = st;
    io->sock = sock;
3301 3302
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3303

3304 3305 3306 3307 3308
    if (virThreadCreateFull(&io->thread, true,
                            qemuMigrationSrcIOFunc,
                            "qemu-mig-tunnel",
                            false,
                            io) < 0) {
3309 3310
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3311
        goto error;
3312 3313 3314
    }

    return io;
3315

3316
 error:
3317 3318 3319 3320
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3321 3322 3323
}

static int
3324
qemuMigrationSrcStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3325 3326
{
    int rv = -1;
3327 3328 3329 3330 3331 3332 3333 3334 3335
    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;
    }

3336 3337 3338 3339
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3340 3341 3342 3343
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3344 3345 3346 3347 3348 3349
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3350
 cleanup:
3351 3352
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3353 3354
    VIR_FREE(io);
    return rv;
3355 3356
}

3357
static int
3358 3359 3360
qemuMigrationSrcConnect(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationSpecPtr spec)
3361 3362 3363 3364 3365 3366 3367
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3368
    port = g_strdup_printf("%d", spec->dest.host.port);
3369 3370 3371 3372

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

3373
    if (qemuSecuritySetSocketLabel(driver->securityManager, vm->def) < 0)
3374
        goto cleanup;
3375 3376 3377
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
3378
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
3379
        virObjectUnref(sock);
3380
    }
3381
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0 ||
3382 3383 3384
        spec->dest.fd.qemu == -1)
        goto cleanup;

3385 3386 3387 3388 3389 3390 3391
    /* 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;
    }

3392 3393
    ret = 0;

3394
 cleanup:
3395 3396 3397 3398 3399 3400
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3401 3402

static int
3403 3404 3405 3406
qemuMigrationSrcContinue(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMonitorMigrationStatus status,
                         qemuDomainAsyncJob asyncJob)
3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422
{
    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;
}


3423
static int
3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437
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,
3438
                    qemuMigrationParamsPtr migParams)
3439
{
3440
    int ret = -1;
3441 3442
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3443
    qemuMigrationCookiePtr mig = NULL;
3444
    char *tlsAlias = NULL;
3445
    qemuMigrationIOThreadPtr iothread = NULL;
3446
    int fd = -1;
3447
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3448
    virErrorPtr orig_err = NULL;
3449
    unsigned int cookieFlags = 0;
3450
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3451
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
3452
    bool cancel = false;
3453
    unsigned int waitFlags;
3454
    virDomainDefPtr persistDef = NULL;
3455
    char *timestamp;
3456
    int rc;
3457 3458

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
3459
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3460 3461
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
3462 3463
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3464
              spec, spec->destType, spec->fwdType, dconn,
3465
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3466

3467 3468 3469 3470 3471 3472 3473 3474 3475 3476
    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;
    }

3477 3478
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3479 3480 3481 3482
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3483 3484 3485
        return -1;
    }

3486 3487 3488
    if (events)
        priv->signalIOError = abort_on_error;

3489 3490
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
        if (persist_xml) {
3491 3492 3493
            if (!(persistDef = qemuMigrationAnyPrepareDef(driver,
                                                          priv->qemuCaps,
                                                          persist_xml,
3494
                                                          NULL, NULL)))
3495
                goto error;
3496 3497
        } else {
            virDomainDefPtr def = vm->newDef ? vm->newDef : vm->def;
3498
            if (!(persistDef = qemuDomainDefCopy(driver, priv->qemuCaps, def,
3499 3500
                                                 VIR_DOMAIN_XML_SECURE |
                                                 VIR_DOMAIN_XML_MIGRATABLE)))
3501
                goto error;
3502 3503 3504
        }
    }

3505 3506
    mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                 cookiein, cookieinlen,
3507 3508 3509
                                 cookieFlags |
                                 QEMU_MIGRATION_COOKIE_GRAPHICS |
                                 QEMU_MIGRATION_COOKIE_CAPS);
3510
    if (!mig)
3511
        goto error;
3512

3513
    if (qemuMigrationSrcGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3514 3515
        VIR_WARN("unable to provide data for graphics client relocation");

3516
    if (qemuMigrationParamsCheck(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3517
                                 migParams, mig->caps->automatic) < 0)
3518 3519
        goto error;

3520
    if (flags & VIR_MIGRATE_TLS) {
3521
        const char *hostname = NULL;
3522 3523 3524 3525

        /* We need to add tls-hostname whenever QEMU itself does not
         * connect directly to the destination. */
        if (spec->destType == MIGRATION_DEST_CONNECT_HOST ||
3526 3527 3528
            spec->destType == MIGRATION_DEST_FD)
            hostname = spec->dest.host.name;

3529
        if (qemuMigrationParamsEnableTLS(driver, vm, false,
3530
                                         QEMU_ASYNC_JOB_MIGRATION_OUT,
3531
                                         &tlsAlias, hostname,
3532 3533
                                         migParams) < 0)
            goto error;
3534
    } else {
3535
        if (qemuMigrationParamsDisableTLS(vm, migParams) < 0)
3536
            goto error;
3537 3538
    }

3539 3540 3541 3542
    if (qemuMigrationParamsApply(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                 migParams) < 0)
        goto error;

3543 3544 3545
    if (migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                         QEMU_MONITOR_MIGRATE_NON_SHARED_INC)) {
        if (mig->nbd) {
3546 3547 3548 3549
            /* 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. */
3550 3551
            if (flags & VIR_MIGRATE_TLS &&
                !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV_DEL)) {
3552 3553 3554 3555 3556
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                               _("NBD migration with TLS is not supported"));
                goto error;
            }

3557
            /* This will update migrate_flags on success */
3558 3559 3560 3561 3562 3563
            if (qemuMigrationSrcNBDStorageCopy(driver, vm, mig,
                                               spec->dest.host.name,
                                               migrate_speed,
                                               &migrate_flags,
                                               nmigrate_disks,
                                               migrate_disks,
3564
                                               dconn, tlsAlias, flags) < 0) {
3565
                goto error;
3566 3567 3568 3569 3570 3571 3572
            }
        } 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.");
        }
3573 3574
    }

3575
    /* Before EnterMonitor, since already qemuProcessStopCPUs does that */
3576 3577
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
3578 3579
        if (qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                                QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3580
            goto error;
3581 3582
    }

3583 3584
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3585
        goto error;
3586

3587
    if (priv->job.abortJob) {
3588 3589
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
3590
         * priv->job.abortJob will not change */
3591
        priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
3592 3593 3594
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
3595
        goto exit_monitor;
3596 3597
    }

3598 3599
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
3600

3601 3602
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3603
        qemuMigrationSrcConnect(driver, vm, spec) < 0) {
3604
        goto exit_monitor;
3605
    }
3606

3607 3608 3609 3610 3611 3612
    /* log start of migration */
    if ((timestamp = virTimeStringNow()) != NULL) {
        qemuDomainLogAppendMessage(driver, vm, "%s: initiating migration\n", timestamp);
        VIR_FREE(timestamp);
    }

3613
    rc = -1;
3614 3615
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3616 3617
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
3618
            goto exit_monitor;
M
Michael R. Hines 已提交
3619
        }
3620 3621 3622 3623
        rc = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
                                      spec->dest.host.protocol,
                                      spec->dest.host.name,
                                      spec->dest.host.port);
3624 3625
        break;

3626 3627 3628 3629
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3630
    case MIGRATION_DEST_FD:
3631
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3632
            fd = spec->dest.fd.local;
3633 3634
            spec->dest.fd.local = -1;
        }
3635 3636
        rc = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                    spec->dest.fd.qemu);
3637 3638
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3639
    }
3640 3641

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
3642
        goto error;
3643 3644 3645

    /* From this point onwards we *must* call cancel to abort the
     * migration on source if anything goes wrong */
3646
    cancel = true;
3647

3648
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3649
        if (!(iothread = qemuMigrationSrcStartTunnel(spec->fwd.stream, fd)))
3650
            goto error;
3651 3652 3653 3654 3655
        /* If we've created a tunnel, then the 'fd' will be closed in the
         * qemuMigrationIOFunc as data->sock.
         */
        fd = -1;
    }
3656

3657
    waitFlags = QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;
3658 3659 3660 3661 3662 3663 3664
    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;

3665 3666 3667
    rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT,
                                           dconn, waitFlags);
3668
    if (rc == -2) {
3669
        goto error;
3670 3671 3672 3673 3674
    } else if (rc == -1) {
        /* QEMU reported failed migration, nothing to cancel anymore */
        cancel = false;
        goto error;
    }
3675

3676
    /* When migration completed, QEMU will have paused the CPUs for us.
J
Ján Tomko 已提交
3677
     * Wait for the STOP event to be processed to release the lock state.
3678
     */
J
Ján Tomko 已提交
3679 3680 3681 3682 3683 3684
    while (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        priv->signalStop = true;
        rc = virDomainObjWait(vm);
        priv->signalStop = false;
        if (rc < 0)
            goto error;
3685
    }
3686

J
Ján Tomko 已提交
3687
    if (mig->nbd &&
3688 3689 3690
        qemuMigrationSrcNBDCopyCancel(driver, vm, true,
                                      QEMU_ASYNC_JOB_MIGRATION_OUT,
                                      dconn) < 0)
3691
        goto error;
3692

3693 3694 3695 3696 3697
    /* 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) {
3698 3699 3700
        if (qemuMigrationSrcContinue(driver, vm,
                                     QEMU_MONITOR_MIGRATION_STATUS_PRE_SWITCHOVER,
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3701 3702 3703 3704
            goto error;

        waitFlags ^= QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;

3705 3706 3707
        rc = qemuMigrationSrcWaitForCompletion(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_OUT,
                                               dconn, waitFlags);
3708 3709 3710 3711 3712 3713 3714 3715 3716
        if (rc == -2) {
            goto error;
        } else if (rc == -1) {
            /* QEMU reported failed migration, nothing to cancel anymore */
            cancel = false;
            goto error;
        }
    }

3717 3718 3719
    if (iothread) {
        qemuMigrationIOThreadPtr io;

3720
        io = g_steal_pointer(&iothread);
3721
        if (qemuMigrationSrcStopTunnel(io, false) < 0)
3722
            goto error;
3723 3724 3725
    }

    if (priv->job.completed) {
3726
        priv->job.completed->stopped = priv->job.current->stopped;
3727 3728 3729 3730 3731 3732 3733 3734 3735
        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 ||
3736 3737 3738
        qemuMigrationBakeCookie(mig, driver, vm,
                                QEMU_MIGRATION_SOURCE,
                                cookieout, cookieoutlen, cookieFlags) < 0) {
3739 3740
        VIR_WARN("Unable to encode migration cookie");
    }
3741

3742
    ret = 0;
3743

3744
 cleanup:
3745
    VIR_FREE(tlsAlias);
3746
    VIR_FORCE_CLOSE(fd);
3747
    virDomainDefFree(persistDef);
3748 3749
    qemuMigrationCookieFree(mig);

3750 3751 3752
    if (events)
        priv->signalIOError = false;

3753
    virErrorRestore(&orig_err);
3754

3755 3756
    return ret;

3757
 error:
3758
    virErrorPreserveLast(&orig_err);
3759

3760 3761 3762 3763 3764 3765 3766 3767
    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));
        }
3768

3769 3770
        /* cancel any outstanding NBD jobs */
        if (mig && mig->nbd)
3771 3772 3773
            qemuMigrationSrcNBDCopyCancel(driver, vm, false,
                                          QEMU_ASYNC_JOB_MIGRATION_OUT,
                                          dconn);
3774 3775 3776 3777

        if (priv->job.current->status != QEMU_DOMAIN_JOB_STATUS_CANCELED)
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
    }
3778 3779

    if (iothread)
3780
        qemuMigrationSrcStopTunnel(iothread, true);
3781 3782 3783

    goto cleanup;

3784 3785
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
3786
    goto error;
3787 3788
}

3789
/* Perform migration using QEMU's native migrate support,
3790 3791
 * not encrypted obviously
 */
3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806
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,
3807
                              qemuMigrationParamsPtr migParams)
3808
{
3809
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
3810
    virURIPtr uribits = NULL;
3811
    int ret = -1;
3812 3813 3814
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
3815
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3816
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
3817
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
3818
              cookieout, cookieoutlen, flags, resource,
3819
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3820

3821
    if (!(uribits = qemuMigrationAnyParseURI(uri, NULL)))
3822 3823
        return -1;

3824 3825 3826 3827 3828 3829 3830
    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
3831 3832 3833 3834 3835
    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"));
3836
            goto cleanup;
M
Michael R. Hines 已提交
3837
        }
3838
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
3839 3840 3841
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
3842
            goto cleanup;
M
Michael R. Hines 已提交
3843 3844 3845
        }
    }

3846 3847 3848 3849
    /* RDMA and multi-fd migration requires QEMU to connect to the destination
     * itself.
     */
    if (STREQ(uribits->scheme, "rdma") || (flags & VIR_MIGRATE_PARALLEL))
3850
        spec.destType = MIGRATION_DEST_HOST;
3851 3852
    else
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
3853
    spec.dest.host.protocol = uribits->scheme;
3854 3855 3856
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
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 3864 3865

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

3866
 cleanup:
3867
    virURIFree(uribits);
3868 3869 3870 3871 3872

    return ret;
}


3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887
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,
3888
                              qemuMigrationParamsPtr migParams)
3889 3890 3891
{
    int ret = -1;
    qemuMigrationSpec spec;
3892
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3893
    int fds[2] = { -1, -1 };
3894 3895

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
3896
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3897
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
3898
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
3899
              cookieout, cookieoutlen, flags, resource,
3900
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3901 3902 3903 3904

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

3905

3906 3907 3908
    spec.destType = MIGRATION_DEST_FD;
    spec.dest.fd.qemu = -1;
    spec.dest.fd.local = -1;
3909

3910 3911 3912 3913 3914 3915
    if (virPipe(fds) < 0)
        goto cleanup;

    spec.dest.fd.qemu = fds[1];
    spec.dest.fd.local = fds[0];

3916
    if (spec.dest.fd.qemu == -1 ||
3917 3918
        qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    spec.dest.fd.qemu) < 0) {
3919 3920 3921
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
        goto cleanup;
3922 3923
    }

3924 3925 3926
    ret = qemuMigrationSrcRun(driver, vm, persist_xml, cookiein, cookieinlen,
                              cookieout, cookieoutlen, flags, resource, &spec,
                              dconn, graphicsuri, nmigrate_disks, migrate_disks,
3927
                              migParams);
3928

3929
 cleanup:
3930 3931
    VIR_FORCE_CLOSE(spec.dest.fd.qemu);
    VIR_FORCE_CLOSE(spec.dest.fd.local);
3932

3933
    virObjectUnref(cfg);
3934 3935 3936 3937
    return ret;
}


3938 3939 3940 3941
/* 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 */
3942 3943 3944 3945 3946 3947 3948 3949
static int
qemuMigrationSrcPerformPeer2Peer2(virQEMUDriverPtr driver,
                                  virConnectPtr sconn,
                                  virConnectPtr dconn,
                                  virDomainObjPtr vm,
                                  const char *dconnuri,
                                  unsigned long flags,
                                  const char *dname,
3950 3951
                                  unsigned long resource,
                                  qemuMigrationParamsPtr migParams)
3952 3953 3954
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
3955
    char *cookie = NULL;
3956 3957 3958
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
3959
    bool cancelled;
3960
    virStreamPtr st = NULL;
3961
    unsigned long destflags;
3962

3963
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
3964
              "flags=0x%lx, dname=%s, resource=%lu",
3965 3966
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
3967

3968 3969 3970 3971 3972
    /* 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,
3973 3974
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
3975 3976 3977 3978 3979
        return -1;

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

3980 3981
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
3982

3983 3984 3985 3986 3987 3988 3989 3990 3991 3992
    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;

3993
        qemuDomainObjEnterRemote(vm);
3994
        ret = dconn->driver->domainMigratePrepareTunnel
3995
            (dconn, st, destflags, dname, resource, dom_xml);
3996 3997
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
3998
    } else {
3999
        qemuDomainObjEnterRemote(vm);
4000 4001
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
4002
             destflags, dname, resource, dom_xml);
4003 4004
        if (qemuDomainObjExitRemote(vm, true) < 0)
            goto cleanup;
4005 4006 4007
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
4008 4009
        goto cleanup;

4010 4011
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
4012 4013
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
4014
        cancelled = true;
4015
        virErrorPreserveLast(&orig_err);
4016
        goto finish;
4017 4018
    }

4019 4020 4021 4022
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
4023
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4024
    if (flags & VIR_MIGRATE_TUNNELLED)
4025 4026 4027
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, NULL,
                                            NULL, 0, NULL, NULL,
                                            flags, resource, dconn,
4028
                                            NULL, 0, NULL, migParams);
4029
    else
4030 4031 4032 4033
        ret = qemuMigrationSrcPerformNative(driver, vm, NULL, uri_out,
                                            cookie, cookielen,
                                            NULL, NULL, /* No out cookie with v2 migration */
                                            flags, resource, dconn, NULL, 0, NULL,
4034
                                            migParams);
4035 4036 4037

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
4038
        virErrorPreserveLast(&orig_err);
4039

4040 4041 4042
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4043
    cancelled = ret < 0;
4044

4045
 finish:
4046 4047 4048 4049
    /* 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.
     */
4050
    dname = dname ? dname : vm->def->name;
4051
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4052
    qemuDomainObjEnterRemote(vm);
4053
    ddomain = dconn->driver->domainMigrateFinish2
4054
        (dconn, dname, cookie, cookielen,
4055
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4056 4057
    /* The domain is already gone at this point */
    ignore_value(qemuDomainObjExitRemote(vm, false));
4058 4059
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4060

4061
 cleanup:
4062
    if (ddomain) {
4063
        virObjectUnref(ddomain);
4064 4065 4066 4067
        ret = 0;
    } else {
        ret = -1;
    }
4068

4069
    virObjectUnref(st);
4070

4071
    virErrorRestore(&orig_err);
4072
    VIR_FREE(uri_out);
4073
    VIR_FREE(cookie);
4074 4075

    return ret;
4076 4077 4078
}


4079 4080 4081 4082
/* 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 */
4083
static int
4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097
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,
4098
                                  qemuMigrationParamsPtr migParams,
4099 4100 4101
                                  unsigned long long bandwidth,
                                  bool useParams,
                                  unsigned long flags)
4102 4103 4104 4105 4106 4107 4108 4109 4110 4111
{
    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;
4112
    bool cancelled = true;
4113
    virStreamPtr st = NULL;
4114
    unsigned long destflags;
4115 4116 4117
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4118
    size_t i;
4119
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4120 4121

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4122
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4123
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4124
              "bandwidth=%llu, useParams=%d, flags=0x%lx",
4125
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4126
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4127
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks, nbdPort,
4128
              bandwidth, useParams, flags);
4129

4130 4131 4132 4133 4134
    /* 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.  */

4135 4136 4137
    dom_xml = qemuMigrationSrcBeginPhase(driver, vm, xmlin, dname,
                                         &cookieout, &cookieoutlen,
                                         nmigrate_disks, migrate_disks, flags);
4138 4139 4140
    if (!dom_xml)
        goto cleanup;

4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160
    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;
4161 4162 4163 4164 4165 4166

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4167 4168 4169 4170 4171
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4172 4173 4174 4175 4176
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4177 4178 4179 4180 4181
        if (nbdPort &&
            virTypedParamsAddInt(&params, &nparams, &maxparams,
                                 VIR_MIGRATE_PARAM_DISKS_PORT,
                                 nbdPort) < 0)
            goto cleanup;
4182

4183 4184
        if (qemuMigrationParamsDump(migParams, &params, &nparams,
                                    &maxparams, &flags) < 0)
4185
            goto cleanup;
4186 4187
    }

4188 4189 4190
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4191 4192
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4193

4194
    VIR_DEBUG("Prepare3 %p", dconn);
4195
    cookiein = g_steal_pointer(&cookieout);
4196 4197 4198 4199 4200 4201
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
    if (flags & VIR_MIGRATE_TUNNELLED) {
        if (!(st = virStreamNew(dconn, 0)))
            goto cleanup;

4202
        qemuDomainObjEnterRemote(vm);
4203 4204 4205 4206 4207 4208 4209 4210 4211
        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);
        }
4212 4213
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4214
    } else {
4215
        qemuDomainObjEnterRemote(vm);
4216 4217 4218 4219 4220 4221 4222 4223 4224
        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);
        }
4225 4226
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4227 4228 4229 4230 4231
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

4232
    if (offline) {
L
liguang 已提交
4233 4234 4235
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4236
        cancelled = false;
L
liguang 已提交
4237 4238 4239
        goto finish;
    }

4240 4241 4242 4243
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4244
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
4245
            virErrorPreserveLast(&orig_err);
4246
            goto finish;
4247
        }
4248
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4249 4250
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4251
        virErrorPreserveLast(&orig_err);
4252 4253 4254 4255 4256 4257 4258 4259
        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.
     */
4260
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4261
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4262
    VIR_FREE(cookiein);
4263
    cookiein = g_steal_pointer(&cookieout);
4264 4265
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
4266
    if (flags & VIR_MIGRATE_TUNNELLED) {
4267 4268 4269 4270
        ret = qemuMigrationSrcPerformTunnel(driver, vm, st, persist_xml,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4271
                                            nmigrate_disks, migrate_disks,
4272
                                            migParams);
4273
    } else {
4274 4275 4276 4277
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri,
                                            cookiein, cookieinlen,
                                            &cookieout, &cookieoutlen,
                                            flags, bandwidth, dconn, graphicsuri,
4278
                                            nmigrate_disks, migrate_disks,
4279
                                            migParams);
4280
    }
4281 4282

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4283
    if (ret < 0) {
4284
        virErrorPreserveLast(&orig_err);
4285 4286 4287 4288
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4289 4290 4291 4292

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4293
    cancelled = ret < 0;
4294

4295
 finish:
4296 4297 4298 4299 4300 4301 4302 4303
    /*
     * 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);
4304
    cookiein = g_steal_pointer(&cookieout);
4305 4306
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319

    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);
4320 4321
            if (qemuDomainObjExitRemote(vm, !offline) < 0)
                goto cleanup;
4322 4323 4324 4325 4326 4327 4328
        }
    } else {
        dname = dname ? dname : vm->def->name;
        qemuDomainObjEnterRemote(vm);
        ddomain = dconn->driver->domainMigrateFinish3
            (dconn, dname, cookiein, cookieinlen, &cookieout, &cookieoutlen,
             dconnuri, uri, destflags, cancelled);
4329 4330
        if (qemuDomainObjExitRemote(vm, !offline) < 0)
            goto cleanup;
4331
    }
4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351

    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();
4352 4353
                if (err &&
                    err->domain == VIR_FROM_QEMU &&
4354 4355 4356 4357 4358 4359 4360
                    err->code != VIR_ERR_MIGRATE_FINISH_OK) {
                    virFreeError(orig_err);
                    orig_err = NULL;
                }
            }
        }
    }
4361

4362 4363 4364 4365 4366 4367 4368
    /* 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.
4369
     */
4370
    cancelled = ddomain == NULL;
4371

4372 4373 4374 4375
    /* 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)
4376
        virErrorPreserveLast(&orig_err);
4377

4378 4379 4380 4381
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4382
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4383
    VIR_FREE(cookiein);
4384
    cookiein = g_steal_pointer(&cookieout);
4385 4386
    cookieinlen = cookieoutlen;
    cookieoutlen = 0;
4387 4388 4389
    ret = qemuMigrationSrcConfirmPhase(driver, vm,
                                       cookiein, cookieinlen,
                                       flags, cancelled);
4390 4391 4392 4393
    /* 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.
     */
4394 4395 4396
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4397 4398 4399

 cleanup:
    if (ddomain) {
4400
        virObjectUnref(ddomain);
4401 4402 4403 4404 4405
        ret = 0;
    } else {
        ret = -1;
    }

4406
    virObjectUnref(st);
4407

4408
    virErrorRestore(&orig_err);
4409 4410 4411
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4412
    virTypedParamsFree(params, nparams);
4413 4414 4415 4416
    return ret;
}


4417
static void
4418 4419 4420
qemuMigrationSrcConnectionClosed(virConnectPtr conn,
                                 int reason,
                                 void *opaque)
4421 4422 4423 4424 4425 4426 4427 4428
{
    virDomainObjPtr vm = opaque;

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


4429 4430 4431 4432 4433 4434 4435 4436
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


static virConnectAuth virConnectAuthConfig = {
    .credtype = virConnectCredType,
4437
    .ncredtype = G_N_ELEMENTS(virConnectCredType),
4438 4439 4440
};


4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453
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,
4454
                                 qemuMigrationParamsPtr migParams,
4455 4456 4457 4458
                                 unsigned long flags,
                                 const char *dname,
                                 unsigned long resource,
                                 bool *v3proto)
4459 4460 4461 4462
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4463
    virErrorPtr orig_err = NULL;
4464
    bool offline = !!(flags & VIR_MIGRATE_OFFLINE);
4465
    bool dstOffline = false;
4466
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4467
    bool useParams;
4468

4469 4470
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, uri=%s, "
              "graphicsuri=%s, listenAddress=%s, nmigrate_disks=%zu, "
4471
              "migrate_disks=%p, nbdPort=%d, flags=0x%lx, dname=%s, "
4472
              "resource=%lu",
4473
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4474
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4475 4476
              nmigrate_disks, migrate_disks, nbdPort, flags, NULLSTR(dname),
              resource);
4477

4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491
    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;
    }

4492 4493 4494 4495 4496 4497 4498
    if (flags & VIR_MIGRATE_TUNNELLED && nbdPort) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("disk port address is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

4499 4500 4501 4502
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

4503
    qemuDomainObjEnterRemote(vm);
4504
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4505 4506 4507
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;

4508
    if (dconn == NULL) {
4509
        virReportError(VIR_ERR_OPERATION_FAILED,
4510 4511
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4512
        virObjectUnref(cfg);
4513 4514 4515
        return -1;
    }

4516 4517
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4518 4519
        goto cleanup;

4520
    if (virConnectRegisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed,
4521 4522 4523 4524
                                        vm, NULL) < 0) {
        goto cleanup;
    }

4525
    qemuDomainObjEnterRemote(vm);
4526 4527
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4528
        /* v3proto reflects whether the caller used Perform3, but with
4529
         * p2p migrate, regardless of whether Perform2 or Perform3
4530 4531 4532 4533
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4534 4535
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
4536
    if (offline)
4537 4538
        dstOffline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                              VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4539 4540
    if (qemuDomainObjExitRemote(vm, !offline) < 0)
        goto cleanup;
4541

4542
    if (!p2p) {
4543 4544
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4545 4546 4547
        goto cleanup;
    }

4548 4549
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4550
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
4551 4552 4553 4554 4555 4556
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("Migration APIs with extensible parameters are not "
                         "supported but extended parameters were passed"));
        goto cleanup;
    }

4557
    if (offline && !dstOffline) {
L
liguang 已提交
4558 4559 4560 4561 4562 4563
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4564 4565 4566 4567 4568 4569 4570
    /* 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;

4571
    if (*v3proto) {
4572 4573 4574
        ret = qemuMigrationSrcPerformPeer2Peer3(driver, sconn, dconn, dconnuri, vm, xmlin,
                                                persist_xml, dname, uri, graphicsuri,
                                                listenAddress, nmigrate_disks, migrate_disks,
4575
                                                nbdPort, migParams, resource,
4576
                                                useParams, flags);
4577
    } else {
4578
        ret = qemuMigrationSrcPerformPeer2Peer2(driver, sconn, dconn, vm,
4579
                                                dconnuri, flags, dname, resource,
4580
                                                migParams);
4581
    }
4582

4583
 cleanup:
4584
    virErrorPreserveLast(&orig_err);
4585
    qemuDomainObjEnterRemote(vm);
4586
    virConnectUnregisterCloseCallback(dconn, qemuMigrationSrcConnectionClosed);
4587
    virObjectUnref(dconn);
4588
    ignore_value(qemuDomainObjExitRemote(vm, false));
4589
    virErrorRestore(&orig_err);
4590
    virObjectUnref(cfg);
4591 4592 4593 4594
    return ret;
}


4595 4596 4597 4598 4599 4600
/*
 * 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
4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612
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,
4613
                           qemuMigrationParamsPtr migParams,
4614 4615 4616 4617 4618 4619 4620 4621
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
                           const char *dname,
                           unsigned long resource,
                           bool v3proto)
4622
{
4623
    virObjectEventPtr event = NULL;
4624
    int ret = -1;
4625
    virErrorPtr orig_err = NULL;
4626
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4627
    qemuDomainObjPrivatePtr priv = vm->privateData;
4628

4629 4630
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                              flags) < 0)
4631 4632
        goto cleanup;

4633
    if (!(flags & VIR_MIGRATE_OFFLINE) && virDomainObjCheckActive(vm) < 0)
4634 4635
        goto endjob;

4636
    if (!qemuMigrationSrcIsAllowed(driver, vm, true, flags))
4637
        goto endjob;
4638

4639
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
4640 4641
        !qemuMigrationSrcIsSafe(vm->def, priv->qemuCaps,
                                nmigrate_disks, migrate_disks, flags))
4642
        goto endjob;
4643

4644
    qemuMigrationSrcStoreDomainState(vm);
4645 4646

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4647 4648 4649
        ret = qemuMigrationSrcPerformPeer2Peer(driver, conn, vm, xmlin, persist_xml,
                                               dconnuri, uri, graphicsuri, listenAddress,
                                               nmigrate_disks, migrate_disks, nbdPort,
4650
                                               migParams, flags, dname, resource,
4651
                                               &v3proto);
4652
    } else {
4653
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4654 4655 4656
        ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                            cookieout, cookieoutlen,
                                            flags, resource, NULL, NULL, 0, NULL,
4657
                                            migParams);
4658
    }
4659 4660
    if (ret < 0)
        goto endjob;
4661

4662 4663 4664 4665
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4666
    if (!v3proto) {
4667
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
4668
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4669
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4670
        virDomainAuditStop(vm, "migrated");
4671
        event = virDomainEventLifecycleNewFromObj(vm,
4672 4673
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4674 4675
    }

4676
 endjob:
4677
    if (ret < 0)
4678
        virErrorPreserveLast(&orig_err);
4679

4680 4681 4682 4683
    /* v2 proto has no confirm phase so we need to reset migration parameters
     * here
     */
    if (!v3proto && ret < 0)
4684
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4685
                                 priv->job.migParams, priv->job.apiFlags);
4686

4687
    qemuMigrationSrcRestoreDomainState(driver, vm);
4688

4689
    qemuMigrationJobFinish(driver, vm);
4690
    if (!virDomainObjIsActive(vm) && ret == 0) {
4691
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
4692
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4693 4694
            vm->persistent = 0;
        }
4695
        qemuDomainRemoveInactiveJob(driver, vm);
4696 4697
    }

4698
    virErrorRestore(&orig_err);
4699

4700
 cleanup:
4701
    virObjectEventStateQueue(driver->domainEventState, event);
4702
    virObjectUnref(cfg);
4703 4704 4705 4706 4707 4708 4709
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4710 4711 4712 4713 4714 4715 4716 4717
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,
4718
                             qemuMigrationParamsPtr migParams,
4719 4720 4721 4722 4723 4724
                             const char *cookiein,
                             int cookieinlen,
                             char **cookieout,
                             int *cookieoutlen,
                             unsigned long flags,
                             unsigned long resource)
4725
{
4726
    qemuDomainObjPrivatePtr priv = vm->privateData;
4727 4728 4729 4730
    int ret = -1;

    /* If we didn't start the job in the begin phase, start it now. */
    if (!(flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
4731 4732
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  flags) < 0)
4733
            return ret;
4734
    } else if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_OUT)) {
4735
        return ret;
4736 4737 4738
    }

    qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4739
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
4740
                           qemuMigrationSrcCleanup);
4741

4742 4743 4744
    ret = qemuMigrationSrcPerformNative(driver, vm, persist_xml, uri, cookiein, cookieinlen,
                                        cookieout, cookieoutlen,
                                        flags, resource, NULL, graphicsuri,
4745
                                        nmigrate_disks, migrate_disks, migParams);
4746

4747
    if (ret < 0) {
4748
        qemuMigrationSrcRestoreDomainState(driver, vm);
4749
        goto endjob;
4750
    }
4751 4752 4753

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4754
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
4755
                             qemuMigrationSrcCleanup) < 0)
4756 4757
        goto endjob;

4758
 endjob:
4759
    if (ret < 0) {
4760
        qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
4761
                                 priv->job.migParams, priv->job.apiFlags);
4762
        qemuMigrationJobFinish(driver, vm);
4763
    } else {
4764
        qemuMigrationJobContinue(vm);
4765 4766
    }

4767
    if (!virDomainObjIsActive(vm))
4768
        qemuDomainRemoveInactiveJob(driver, vm);
4769 4770 4771 4772

    return ret;
}

4773
int
4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785
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,
4786
                        qemuMigrationParamsPtr migParams,
4787 4788 4789 4790 4791 4792 4793 4794
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4795 4796
{
    VIR_DEBUG("driver=%p, conn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4797
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
4798
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4799
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
4800
              "flags=0x%lx, dname=%s, resource=%lu, v3proto=%d",
4801
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4802
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4803 4804 4805
              nmigrate_disks, migrate_disks, nbdPort,
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4806 4807 4808

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4809 4810
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4811 4812 4813
            return -1;
        }

4814 4815 4816
        return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, dconnuri, uri,
                                          graphicsuri, listenAddress,
                                          nmigrate_disks, migrate_disks, nbdPort,
4817
                                          migParams,
4818 4819 4820
                                          cookiein, cookieinlen,
                                          cookieout, cookieoutlen,
                                          flags, dname, resource, v3proto);
4821 4822
    } else {
        if (dconnuri) {
4823 4824
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
4825 4826 4827 4828
            return -1;
        }

        if (v3proto) {
4829 4830 4831
            return qemuMigrationSrcPerformPhase(driver, conn, vm, persist_xml, uri,
                                                graphicsuri,
                                                nmigrate_disks, migrate_disks,
4832
                                                migParams,
4833 4834 4835
                                                cookiein, cookieinlen,
                                                cookieout, cookieoutlen,
                                                flags, resource);
4836
        } else {
4837 4838 4839
            return qemuMigrationSrcPerformJob(driver, conn, vm, xmlin, persist_xml, NULL,
                                              uri, graphicsuri, listenAddress,
                                              nmigrate_disks, migrate_disks, nbdPort,
4840
                                              migParams,
4841 4842 4843
                                              cookiein, cookieinlen,
                                              cookieout, cookieoutlen, flags,
                                              dname, resource, v3proto);
4844 4845 4846
        }
    }
}
4847

4848
static int
4849
qemuMigrationDstVPAssociatePortProfiles(virDomainDefPtr def)
4850
{
4851
    size_t i;
4852 4853 4854 4855 4856
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
4857
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4858
            if (virNetDevVPortProfileAssociate(net->ifname,
4859
                                               virDomainNetGetActualVirtPortProfile(net),
4860
                                               &net->mac,
4861
                                               virDomainNetGetActualDirectDev(net),
4862
                                               -1,
4863
                                               def->uuid,
4864 4865
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
4866 4867 4868
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
4869
                goto err_exit;
4870
            }
4871
            last_good_net = i;
4872
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
4873

4874
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
4875 4876 4877 4878
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
4879 4880 4881
        }
    }

4882
    return 0;
4883

4884
 err_exit:
4885
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
4886
        net = def->nets[i];
4887
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4888
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
4889
                                                           virDomainNetGetActualVirtPortProfile(net),
4890
                                                           &net->mac,
4891
                                                           virDomainNetGetActualDirectDev(net),
4892
                                                           -1,
4893
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
4894 4895
        }
    }
4896
    return -1;
4897 4898 4899
}


J
Jiri Denemark 已提交
4900
static int
4901 4902 4903 4904
qemuMigrationDstPersist(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuMigrationCookiePtr mig,
                        bool ignoreSaveError)
J
Jiri Denemark 已提交
4905 4906
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4907
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
4908
    virDomainDefPtr vmdef;
4909 4910
    virDomainDefPtr oldDef = NULL;
    unsigned int oldPersist = vm->persistent;
J
Jiri Denemark 已提交
4911 4912 4913 4914
    virObjectEventPtr event;
    int ret = -1;

    vm->persistent = 1;
4915 4916
    oldDef = vm->newDef;
    vm->newDef = qemuMigrationCookieGetPersistent(mig);
J
Jiri Denemark 已提交
4917

4918
    if (!(vmdef = virDomainObjGetPersistentDef(driver->xmlopt, vm,
4919
                                               priv->qemuCaps)))
4920
        goto error;
J
Jiri Denemark 已提交
4921

4922
    if (virDomainDefSave(vmdef, driver->xmlopt, cfg->configDir) < 0 &&
4923
        !ignoreSaveError)
4924
        goto error;
J
Jiri Denemark 已提交
4925 4926 4927

    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_DEFINED,
4928 4929 4930
                                              oldPersist ?
                                              VIR_DOMAIN_EVENT_DEFINED_UPDATED :
                                              VIR_DOMAIN_EVENT_DEFINED_ADDED);
4931
    virObjectEventStateQueue(driver->domainEventState, event);
J
Jiri Denemark 已提交
4932 4933 4934 4935

    ret = 0;

 cleanup:
4936
    virDomainDefFree(oldDef);
J
Jiri Denemark 已提交
4937 4938
    virObjectUnref(cfg);
    return ret;
4939 4940 4941 4942 4943 4944 4945

 error:
    virDomainDefFree(vm->newDef);
    vm->persistent = oldPersist;
    vm->newDef = oldDef;
    oldDef = NULL;
    goto cleanup;
J
Jiri Denemark 已提交
4946 4947 4948
}


4949
virDomainPtr
4950 4951 4952 4953 4954 4955 4956 4957 4958 4959
qemuMigrationDstFinish(virQEMUDriverPtr driver,
                       virConnectPtr dconn,
                       virDomainObjPtr vm,
                       const char *cookiein,
                       int cookieinlen,
                       char **cookieout,
                       int *cookieoutlen,
                       unsigned long flags,
                       int retcode,
                       bool v3proto)
4960 4961
{
    virDomainPtr dom = NULL;
4962
    qemuMigrationCookiePtr mig = NULL;
4963
    virErrorPtr orig_err = NULL;
4964
    int cookie_flags = 0;
J
Jiri Denemark 已提交
4965
    qemuDomainObjPrivatePtr priv = vm->privateData;
4966
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4967
    unsigned short port;
4968 4969
    unsigned long long timeReceived = 0;
    virObjectEventPtr event;
4970
    qemuDomainJobInfoPtr jobInfo = NULL;
4971
    bool inPostCopy = false;
4972
    bool doKill = true;
4973

4974
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
4975
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, retcode=%d",
4976 4977
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
4978

4979 4980 4981
    port = priv->migrationPort;
    priv->migrationPort = 0;

4982
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN)) {
4983
        qemuMigrationDstErrorReport(driver, vm->def->name);
4984
        goto cleanup;
4985
    }
4986

4987 4988
    ignore_value(virTimeMillisNow(&timeReceived));

4989 4990 4991
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
4992

4993
    qemuDomainCleanupRemove(vm, qemuMigrationDstPrepareCleanup);
4994
    VIR_FREE(priv->job.completed);
4995

4996
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
4997 4998
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
4999 5000 5001
    /* 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;
5002

5003 5004
    if (!(mig = qemuMigrationEatCookie(driver, vm->def, priv->origname, priv,
                                       cookiein, cookieinlen, cookie_flags)))
5005
        goto endjob;
5006

5007
    if (flags & VIR_MIGRATE_OFFLINE) {
5008
        if (retcode == 0 &&
5009
            qemuMigrationDstPersist(driver, vm, mig, false) == 0)
5010
            dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, -1);
5011 5012
        goto endjob;
    }
5013

5014 5015 5016 5017
    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
         */
5018
        qemuDomainCheckMonitor(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN);
5019 5020
        goto endjob;
    }
5021

5022 5023 5024
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
5025
        qemuMigrationDstErrorReport(driver, vm->def->name);
5026 5027
        goto endjob;
    }
5028

5029
    if (qemuMigrationDstVPAssociatePortProfiles(vm->def) < 0)
5030
        goto endjob;
5031

5032
    if (mig->network && qemuMigrationDstOPDRelocate(driver, vm, mig) < 0)
5033
        VIR_WARN("unable to provide network data for relocation");
5034

5035
    if (qemuMigrationDstStopNBDServer(driver, vm, mig) < 0)
5036
        goto endjob;
5037

5038 5039
    if (qemuRefreshVirtioChannelState(driver, vm,
                                      QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
5040 5041
        goto endjob;

5042 5043
    if (qemuConnectAgent(driver, vm) < 0)
        goto endjob;
5044

5045
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
5046
        if (qemuMigrationDstPersist(driver, vm, mig, !v3proto) < 0) {
5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057
            /* 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.
5058 5059 5060
             */
            if (v3proto)
                goto endjob;
5061
        }
5062
    }
5063

5064 5065 5066
    /* We need to wait for QEMU to process all data sent by the source
     * before starting guest CPUs.
     */
5067 5068 5069
    if (qemuMigrationDstWaitForCompletion(driver, vm,
                                          QEMU_ASYNC_JOB_MIGRATION_IN,
                                          !!(flags & VIR_MIGRATE_POSTCOPY)) < 0) {
5070 5071 5072 5073 5074 5075
        /* 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;
    }
5076

5077 5078 5079 5080 5081 5082 5083 5084 5085
    /* 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;
    }

5086
    if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY)
5087 5088
        inPostCopy = true;

5089 5090 5091 5092 5093
    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
         */
5094
        if (qemuProcessStartCPUs(driver, vm,
5095 5096
                                 inPostCopy ? VIR_DOMAIN_RUNNING_POSTCOPY
                                            : VIR_DOMAIN_RUNNING_MIGRATED,
5097
                                 QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5098
            if (virGetLastErrorCode() == VIR_ERR_OK)
5099 5100 5101 5102 5103
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("resume operation failed"));
            /* Need to save the current error, in case shutting
             * down the process overwrites it
             */
5104
            virErrorPreserveLast(&orig_err);
5105

5106 5107 5108 5109 5110 5111 5112 5113 5114 5115
            /*
             * 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;
5116
        }
5117

5118
        if (inPostCopy)
5119
            doKill = false;
5120
    }
5121

5122
    if (mig->jobInfo) {
5123
        jobInfo = mig->jobInfo;
5124 5125 5126 5127 5128 5129
        mig->jobInfo = NULL;

        if (jobInfo->sent && timeReceived) {
            jobInfo->timeDelta = timeReceived - jobInfo->sent;
            jobInfo->received = timeReceived;
            jobInfo->timeDeltaSet = true;
5130
        }
5131 5132
        qemuDomainJobInfoUpdateTime(jobInfo);
        qemuDomainJobInfoUpdateDowntime(jobInfo);
5133
    }
L
liguang 已提交
5134

5135
    if (inPostCopy) {
5136 5137 5138
        if (qemuMigrationDstWaitForCompletion(driver, vm,
                                              QEMU_ASYNC_JOB_MIGRATION_IN,
                                              false) < 0) {
5139 5140 5141 5142 5143 5144 5145 5146 5147
            goto endjob;
        }
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            virDomainObjSetState(vm,
                                 VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_MIGRATED);
        }
    }

5148
    dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, vm->def->id);
5149

5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163
    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);
    }
5164

5165 5166 5167 5168 5169
    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);
5170
        virObjectEventStateQueue(driver->domainEventState, event);
5171
    }
5172

5173
    if (virDomainObjIsActive(vm) &&
5174
        virDomainObjSave(vm, driver->xmlopt, cfg->stateDir) < 0)
5175 5176 5177 5178 5179
        VIR_WARN("Failed to save status on vm %s", vm->def->name);

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

5180
 endjob:
5181
    if (!dom &&
5182 5183
        !(flags & VIR_MIGRATE_OFFLINE) &&
        virDomainObjIsActive(vm)) {
5184
        if (doKill) {
5185 5186 5187 5188 5189 5190 5191
            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);
5192
            virObjectEventStateQueue(driver->domainEventState, event);
5193
        } else {
5194
            qemuMigrationAnyPostcopyFailed(driver, vm);
5195
        }
5196 5197
    }

5198
    if (dom) {
5199
        if (jobInfo) {
5200
            priv->job.completed = g_steal_pointer(&jobInfo);
5201
            priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
5202
            priv->job.completed->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;
5203
        }
5204

5205 5206 5207
        if (qemuMigrationBakeCookie(mig, driver, vm,
                                    QEMU_MIGRATION_DESTINATION,
                                    cookieout, cookieoutlen,
5208 5209
                                    QEMU_MIGRATION_COOKIE_STATS) < 0)
            VIR_WARN("Unable to encode migration cookie");
5210 5211 5212 5213 5214 5215

        /* Remove completed stats for post-copy, everything but timing fields
         * is obsolete anyway.
         */
        if (inPostCopy)
            VIR_FREE(priv->job.completed);
5216
    }
5217

5218
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
5219
                             priv->job.migParams, priv->job.apiFlags);
5220

5221
    qemuMigrationJobFinish(driver, vm);
5222
    if (!virDomainObjIsActive(vm))
5223
        qemuDomainRemoveInactiveJob(driver, vm);
5224

5225
 cleanup:
5226
    VIR_FREE(jobInfo);
5227
    virPortAllocatorRelease(port);
5228
    if (priv->mon)
5229
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
5230
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5231
    virDomainObjEndAPI(&vm);
5232
    qemuMigrationCookieFree(mig);
5233
    virErrorRestore(&orig_err);
5234
    virObjectUnref(cfg);
5235 5236 5237 5238

    /* Set a special error if Finish is expected to return NULL as a result of
     * successful call with retcode != 0
     */
5239
    if (retcode != 0 && !dom && virGetLastErrorCode() == VIR_ERR_OK)
5240
        virReportError(VIR_ERR_MIGRATE_FINISH_OK, NULL);
5241 5242
    return dom;
}
5243

5244

5245
/* Helper function called while vm is active.  */
5246
int
5247 5248
qemuMigrationSrcToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
                       int fd,
5249
                       virCommandPtr compressor,
5250
                       qemuDomainAsyncJob asyncJob)
5251 5252 5253
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5254
    int ret = -1;
5255
    int pipeFD[2] = { -1, -1 };
5256
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5257
    char *errbuf = NULL;
5258
    virErrorPtr orig_err = NULL;
5259 5260 5261 5262 5263

    /* 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,
5264 5265
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5266 5267
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5268
    }
5269

5270 5271 5272 5273 5274 5275 5276
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5277
    if (compressor && virPipe(pipeFD) < 0)
5278
        return -1;
5279

5280 5281 5282 5283
    /* 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.  */
5284 5285
    if (qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    compressor ? pipeFD[1] : fd) < 0)
5286 5287
        goto cleanup;

5288
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5289 5290
        goto cleanup;

5291
    if (!compressor) {
5292 5293 5294
        rc = qemuMonitorMigrateToFd(priv->mon,
                                    QEMU_MONITOR_MIGRATE_BACKGROUND,
                                    fd);
5295
    } else {
5296 5297 5298 5299
        virCommandSetInputFD(compressor, pipeFD[0]);
        virCommandSetOutputFD(compressor, &fd);
        virCommandSetErrorBuffer(compressor, &errbuf);
        virCommandDoAsyncIO(compressor);
5300 5301 5302 5303 5304
        if (virSetCloseExec(pipeFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("Unable to set cloexec flag"));
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
            goto cleanup;
5305
        }
5306
        if (virCommandRunAsync(compressor, NULL) < 0) {
5307 5308 5309 5310 5311 5312 5313 5314 5315
            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");
5316
    }
5317
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5318
        goto cleanup;
5319 5320 5321
    if (rc < 0)
        goto cleanup;

5322
    rc = qemuMigrationSrcWaitForCompletion(driver, vm, asyncJob, NULL, 0);
5323

5324 5325
    if (rc < 0) {
        if (rc == -2) {
5326
            virErrorPreserveLast(&orig_err);
5327
            virCommandAbort(compressor);
5328 5329
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5330
                qemuMonitorMigrateCancel(priv->mon);
5331
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5332 5333
            }
        }
5334
        goto cleanup;
5335
    }
5336

5337
    if (compressor && virCommandWait(compressor, NULL) < 0)
5338 5339
        goto cleanup;

5340
    qemuDomainEventEmitJobCompleted(driver, vm);
5341 5342
    ret = 0;

5343
 cleanup:
5344
    if (ret < 0 && !orig_err)
5345
        virErrorPreserveLast(&orig_err);
5346

5347
    /* Restore max migration bandwidth */
5348 5349
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5350 5351
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5352
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5353 5354
    }

5355 5356
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5357
    if (errbuf) {
5358 5359 5360
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
    }
5361

5362
    virErrorRestore(&orig_err);
5363

5364 5365
    return ret;
}
5366

5367 5368

int
5369 5370
qemuMigrationSrcCancel(virQEMUDriverPtr driver,
                       virDomainObjPtr vm)
5371 5372 5373 5374 5375 5376 5377 5378 5379 5380
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool storage = false;
    size_t i;

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

    qemuDomainObjEnterMonitor(driver, vm);
    ignore_value(qemuMonitorMigrateCancel(priv->mon));
5381
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5382
        return -1;
5383 5384 5385 5386

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
5387
        qemuBlockJobDataPtr job;
5388

5389
        if (!(job = qemuBlockJobDiskGetJob(disk)) ||
5390
            !qemuBlockJobIsRunning(job))
5391
            diskPriv->migrating = false;
5392 5393

        if (diskPriv->migrating) {
5394
            qemuBlockJobSyncBegin(job);
5395
            storage = true;
5396 5397
        }

5398 5399
        virObjectUnref(job);
    }
5400 5401 5402

    if (storage &&
        qemuMigrationSrcNBDCopyCancel(driver, vm, false,
5403
                                      QEMU_ASYNC_JOB_NONE, NULL) < 0)
5404
        return -1;
5405

5406
    return 0;
5407 5408 5409
}


5410
static int
5411
qemuMigrationJobStart(virQEMUDriverPtr driver,
5412
                      virDomainObjPtr vm,
5413 5414
                      qemuDomainAsyncJob job,
                      unsigned long apiFlags)
5415
{
5416
    qemuDomainObjPrivatePtr priv = vm->privateData;
5417 5418
    virDomainJobOperation op;
    unsigned long long mask;
5419

5420
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5421 5422
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_IN;
        mask = QEMU_JOB_NONE;
5423
    } else {
5424 5425 5426 5427
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_OUT;
        mask = QEMU_JOB_DEFAULT_MASK |
               JOB_MASK(QEMU_JOB_SUSPEND) |
               JOB_MASK(QEMU_JOB_MIGRATION_OP);
5428
    }
5429

5430
    if (qemuDomainObjBeginAsyncJob(driver, vm, job, op, apiFlags) < 0)
5431 5432
        return -1;

5433 5434
    priv->job.current->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;

5435
    qemuDomainObjSetAsyncJobMask(vm, mask);
5436 5437 5438
    return 0;
}

5439
static void
5440
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5441
                         virDomainObjPtr vm,
5442
                         qemuMigrationJobPhase phase)
5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455
{
    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);
}

5456
static void
5457
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5458
                           virDomainObjPtr vm,
5459
                           qemuMigrationJobPhase phase)
5460 5461 5462 5463
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5464
static void
5465 5466
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5467
    qemuDomainObjReleaseAsyncJob(vm);
5468 5469
}

5470
static bool
5471
qemuMigrationJobIsActive(virDomainObjPtr vm,
5472
                         qemuDomainAsyncJob job)
5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483
{
    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");

5484
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5485 5486 5487 5488 5489
        return false;
    }
    return true;
}

5490
static void
5491
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5492
{
5493
    qemuDomainObjEndAsyncJob(driver, vm);
5494
}
5495 5496 5497


static void
5498
qemuMigrationDstErrorFree(void *data)
5499 5500 5501 5502 5503 5504
{
    virErrorPtr err = data;
    virFreeError(err);
}

int
5505
qemuMigrationDstErrorInit(virQEMUDriverPtr driver)
5506
{
5507
    driver->migrationErrors = virHashAtomicNew(64, qemuMigrationDstErrorFree);
5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518
    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
5519 5520 5521
qemuMigrationDstErrorSave(virQEMUDriverPtr driver,
                          const char *name,
                          virErrorPtr err)
5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534
{
    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
5535 5536
qemuMigrationDstErrorReport(virQEMUDriverPtr driver,
                            const char *name)
5537 5538 5539 5540 5541 5542 5543 5544
{
    virErrorPtr err;

    if (!(err = virHashAtomicSteal(driver->migrationErrors, name)))
        return;

    VIR_DEBUG("Restoring saved incoming migration error for domain %s: %s",
              name, err->message);
5545
    virErrorRestore(&err);
5546
}
5547 5548


5549
int
5550 5551 5552 5553
qemuMigrationSrcFetchMirrorStats(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 qemuDomainAsyncJob asyncJob,
                                 qemuDomainJobInfoPtr jobInfo)
5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574
{
    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;

5575
    blockinfo = qemuMonitorGetAllBlockJobInfo(priv->mon, false);
5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597

    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;
}