qemu_migration.c 200.6 KB
Newer Older
1 2 3
/*
 * qemu_migration.c: QEMU migration handling
 *
4
 * Copyright (C) 2006-2015 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23
 *
 */

#include <config.h>

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

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

42
#include "domain_audit.h"
43
#include "virlog.h"
44
#include "virerror.h"
45
#include "viralloc.h"
E
Eric Blake 已提交
46
#include "virfile.h"
47
#include "virnetdevopenvswitch.h"
48
#include "datatypes.h"
49
#include "virfdstream.h"
50
#include "viruuid.h"
51
#include "virtime.h"
52
#include "locking/domain_lock.h"
53
#include "rpc/virnetsocket.h"
54
#include "virstoragefile.h"
M
Martin Kletzander 已提交
55
#include "viruri.h"
56
#include "virhook.h"
57
#include "virstring.h"
58
#include "virtypedparam.h"
M
Michael R. Hines 已提交
59
#include "virprocess.h"
60
#include "nwfilter_conf.h"
61 62 63

#define VIR_FROM_THIS VIR_FROM_QEMU

64 65
VIR_LOG_INIT("qemu.qemu_migration");

66 67 68 69 70 71 72 73 74 75 76 77 78
VIR_ENUM_IMPL(qemuMigrationJobPhase, QEMU_MIGRATION_PHASE_LAST,
              "none",
              "perform2",
              "begin3",
              "perform3",
              "perform3_done",
              "confirm3_cancelled",
              "confirm3",
              "prepare",
              "finish2",
              "finish3",
);

79 80 81 82 83
VIR_ENUM_IMPL(qemuMigrationCompressMethod, QEMU_MIGRATION_COMPRESS_LAST,
              "xbzrle",
              "mt",
);

84 85
#define QEMU_MIGRATION_TLS_ALIAS_BASE "libvirt_migrate"

86

87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109
/* qemuMigrationCheckTLSCreds
 * @driver: pointer to qemu driver
 * @vm: domain object
 * @asyncJob: migration job to join
 *
 * Query the migration parameters looking for the 'tls-creds' parameter.
 * If found, then we can support setting or clearing the parameters and thus
 * can support TLS for migration.
 *
 * Returns 0 if we were able to successfully fetch the params and
 * additionally if the tls-creds parameter exists, saves it in the
 * private domain structure. Returns -1 on failure.
 */
static int
qemuMigrationCheckTLSCreds(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuDomainAsyncJob asyncJob)
{
    int ret = -1;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuMonitorMigrationParams migParams = { 0 };

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
110
        return -1;
111 112 113 114 115

    if (qemuMonitorGetMigrationParams(priv->mon, &migParams) < 0)
        goto cleanup;

    /* NB: Could steal NULL pointer too! Let caller decide what to do. */
116
    VIR_STEAL_PTR(priv->migTLSAlias, migParams.tlsCreds);
117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214

    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

    qemuMigrationParamsClear(&migParams);

    return ret;
}


/* qemuMigrationCheckSetupTLS
 * @conn: Connection pointer
 * @driver: pointer to qemu driver
 * @vm: domain object
 * @cfg: configuration pointer
 * @asyncJob: migration job to join
 *
 * Check if TLS is possible and set up the environment. Assumes the caller
 * desires to use TLS (e.g. caller found VIR_MIGRATE_TLS flag).
 *
 * Ensure the qemu.conf has been properly configured to add an entry for
 * "migrate_tls_x509_cert_dir". Also check if the "tls-creds" parameter
 * was present from a query of migration parameters
 *
 * Returns 0 on success, -1 on error/failure
 */
static int
qemuMigrationCheckSetupTLS(virConnectPtr conn,
                           virQEMUDriverPtr driver,
                           virQEMUDriverConfigPtr cfg,
                           virDomainObjPtr vm,
                           qemuDomainAsyncJob asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (!cfg->migrateTLSx509certdir) {
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("host migration TLS directory not configured"));
        return -1;
    }

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

    if (!priv->migTLSAlias) {
        virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                       _("TLS migration is not supported with this "
                         "QEMU binary"));
        return -1;
    }

    /* If there's a secret, then grab/store it now using the connection */
    if (cfg->migrateTLSx509secretUUID &&
        !(priv->migSecinfo =
          qemuDomainSecretInfoTLSNew(conn, priv, QEMU_MIGRATION_TLS_ALIAS_BASE,
                                     cfg->migrateTLSx509secretUUID)))
        return -1;

    return 0;
}


/* qemuMigrationAddTLSObjects
 * @driver: pointer to qemu driver
 * @vm: domain object
 * @cfg: configuration pointer
 * @tlsListen: server or client
 * @asyncJob: Migration job to join
 * @tlsAlias: alias to be generated for TLS object
 * @secAlias: alias to be generated for a secinfo object
 * @migParams: migration parameters to set
 *
 * Create the TLS objects for the migration and set the migParams value
 *
 * Returns 0 on success, -1 on failure
 */
static int
qemuMigrationAddTLSObjects(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           virQEMUDriverConfigPtr cfg,
                           bool tlsListen,
                           qemuDomainAsyncJob asyncJob,
                           char **tlsAlias,
                           char **secAlias,
                           qemuMonitorMigrationParamsPtr migParams)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virJSONValuePtr tlsProps = NULL;
    virJSONValuePtr secProps = NULL;

    if (qemuDomainGetTLSObjects(priv->qemuCaps, priv->migSecinfo,
                                cfg->migrateTLSx509certdir, tlsListen,
                                cfg->migrateTLSx509verify,
                                QEMU_MIGRATION_TLS_ALIAS_BASE,
                                &tlsProps, tlsAlias, &secProps, secAlias) < 0)
215
        goto error;
216 217 218 219 220 221 222 223 224

    /* Ensure the domain doesn't already have the TLS objects defined...
     * This should prevent any issues just in case some cleanup wasn't
     * properly completed (both src and dst use the same alias) or
     * some other error path between now and perform . */
    qemuDomainDelTLSObjects(driver, vm, asyncJob, *secAlias, *tlsAlias);

    if (qemuDomainAddTLSObjects(driver, vm, asyncJob, *secAlias, &secProps,
                                *tlsAlias, &tlsProps) < 0)
225
        goto error;
226

227
    if (VIR_STRDUP(migParams->tlsCreds, *tlsAlias) < 0)
228
        goto error;
229 230

    return 0;
231 232 233 234 235

 error:
    virJSONValueFree(tlsProps);
    virJSONValueFree(secProps);
    return -1;
236 237 238
}


239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254
static void
qemuMigrationStoreDomainState(virDomainObjPtr vm)
{
    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
qemuMigrationRestoreDomainState(virConnectPtr conn, virDomainObjPtr vm)
{
    virQEMUDriverPtr driver = conn->privateData;
    qemuDomainObjPrivatePtr priv = vm->privateData;
255 256
    int reason;
    virDomainState state = virDomainObjGetState(vm, &reason);
257 258
    bool ret = false;

259 260 261 262 263
    VIR_DEBUG("driver=%p, vm=%p, pre-mig-state=%s, state=%s, reason=%s",
              driver, vm,
              virDomainStateTypeToString(priv->preMigrationState),
              virDomainStateTypeToString(state),
              virDomainStateReasonToString(state, reason));
264

265 266 267 268 269
    if (state != VIR_DOMAIN_PAUSED ||
        reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED)
        goto cleanup;

    if (priv->preMigrationState == VIR_DOMAIN_RUNNING) {
270 271 272 273 274 275
        /* 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 */
276
        if (qemuProcessStartCPUs(driver, vm,
277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292
                                 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;
}

293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330

static int
qemuMigrationPrecreateDisk(virConnectPtr conn,
                           virDomainDiskDefPtr disk,
                           unsigned long long capacity)
{
    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));

    switch ((virStorageType) disk->src->type) {
    case VIR_STORAGE_TYPE_FILE:
        if (!virDomainDiskGetSource(disk)) {
            VIR_DEBUG("Dropping sourceless disk '%s'",
                      disk->dst);
            return 0;
        }

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

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

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

331
        if (!(pool = virStoragePoolLookupByTargetPath(conn, basePath)))
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346
            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;

347 348 349 350 351
    case VIR_STORAGE_TYPE_NETWORK:
        VIR_DEBUG("Skipping creation of network disk '%s'",
                  disk->dst);
        return 0;

352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398
    case VIR_STORAGE_TYPE_BLOCK:
    case VIR_STORAGE_TYPE_DIR:
    case VIR_STORAGE_TYPE_NONE:
    case VIR_STORAGE_TYPE_LAST:
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot precreate storage for disk type '%s'"),
                       virStorageTypeToString(disk->src->type));
        goto cleanup;
    }

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

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

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

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

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

399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416
static bool
qemuMigrateDisk(virDomainDiskDef const *disk,
                size_t nmigrate_disks, const char **migrate_disks)
{
    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 &&
417
           !virStorageSourceIsEmpty(disk->src);
418 419
}

420 421 422 423 424

static int
qemuMigrationPrecreateStorage(virConnectPtr conn,
                              virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                              virDomainObjPtr vm,
425 426
                              qemuMigrationCookieNBDPtr nbd,
                              size_t nmigrate_disks,
427 428
                              const char **migrate_disks,
                              bool incremental)
429 430 431 432 433 434 435 436 437 438 439
{
    int ret = -1;
    size_t i = 0;

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

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

440
        VIR_DEBUG("Looking up disk target '%s' (capacity=%llu)",
441 442
                  nbd->disks[i].target, nbd->disks[i].capacity);

443 444
        if (!(disk = virDomainDiskByName(vm->def, nbd->disks[i].target,
                                         false))) {
445 446 447 448 449 450 451 452
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unable to find disk by target: %s"),
                           nbd->disks[i].target);
            goto cleanup;
        }

        diskSrcPath = virDomainDiskGetSource(disk);

453 454
        /* Skip disks we don't want to migrate and already existing disks. */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks) ||
455 456 457 458
            (diskSrcPath && virFileExists(diskSrcPath))) {
            continue;
        }

459 460 461 462 463 464 465
        if (incremental) {
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                           _("pre-creation of storage targets for incremental "
                             "storage migration is not supported"));
            goto cleanup;
        }

466 467 468 469 470 471 472 473 474 475 476 477
        VIR_DEBUG("Proceeding with disk source %s", NULLSTR(diskSrcPath));

        if (qemuMigrationPrecreateDisk(conn, disk, nbd->disks[i].capacity) < 0)
            goto cleanup;
    }

    ret = 0;
 cleanup:
    return ret;
}


478 479 480 481 482 483 484 485 486 487 488 489 490 491
/**
 * qemuMigrationStartNBDServer:
 * @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
qemuMigrationStartNBDServer(virQEMUDriverPtr driver,
J
Ján Tomko 已提交
492
                            virDomainObjPtr vm,
493 494
                            const char *listenAddr,
                            size_t nmigrate_disks,
495 496
                            const char **migrate_disks,
                            int nbdPort)
497 498 499 500 501 502 503
{
    int ret = -1;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    unsigned short port = 0;
    char *diskAlias = NULL;
    size_t i;

504 505 506 507 508 509
    if (nbdPort < 0 || nbdPort > USHRT_MAX) {
        virReportError(VIR_ERR_INVALID_ARG, "%s",
                       _("nbd port must be in range 0-65535"));
        return -1;
    }

510 511 512
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];

513 514
        /* check whether disk should be migrated */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks))
515 516
            continue;

517
        if (disk->src->readonly || virStorageSourceIsEmpty(disk->src)) {
518
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED,
519 520
                           _("Cannot migrate empty or read-only disk %s"),
                           disk->dst);
521 522 523
            goto cleanup;
        }

524
        VIR_FREE(diskAlias);
525
        if (!(diskAlias = qemuAliasFromDisk(disk)))
526 527 528 529 530 531
            goto cleanup;

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

532 533 534 535 536 537 538 539
        if (port == 0) {
            if (nbdPort)
                port = nbdPort;
            else if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
                goto exit_monitor;

            if (qemuMonitorNBDServerStart(priv->mon, listenAddr, port) < 0)
                goto exit_monitor;
540 541
        }

542 543 544
        if (qemuMonitorNBDServerAdd(priv->mon, diskAlias, true) < 0)
            goto exit_monitor;
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
545 546 547 548 549 550
            goto cleanup;
    }

    priv->nbdPort = port;
    ret = 0;

551
 cleanup:
552
    VIR_FREE(diskAlias);
553
    if (ret < 0 && nbdPort == 0)
554
        virPortAllocatorRelease(driver->migrationPorts, port);
555
    return ret;
556 557 558 559

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;
560 561
}

562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588

static int
qemuMigrationStopNBDServer(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuMigrationCookiePtr mig)
{
    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;

    virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
    priv->nbdPort = 0;
    return 0;
}


/**
589
 * qemuMigrationDriveMirrorReady:
590 591 592 593 594 595 596 597 598 599 600 601
 * @driver: qemu driver
 * @vm: domain
 *
 * Check the status of all drive-mirrors started by
 * qemuMigrationDriveMirror. Any pending block job events
 * for the mirrored disks will be processed.
 *
 * Returns 1 if all mirrors are "ready",
 *         0 if some mirrors are still performing initial sync,
 *        -1 on error.
 */
static int
602
qemuMigrationDriveMirrorReady(virQEMUDriverPtr driver,
603 604
                              virDomainObjPtr vm,
                              qemuDomainAsyncJob asyncJob)
605 606
{
    size_t i;
607 608
    size_t notReady = 0;
    int status;
609 610 611

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
612
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
613
        char *error = NULL;
614

615
        if (!diskPriv->migrating)
616 617
            continue;

618
        status = qemuBlockJobUpdate(driver, vm, asyncJob, disk, &error);
619
        if (status == VIR_DOMAIN_BLOCK_JOB_FAILED) {
620 621 622 623 624 625 626 627 628
            if (error) {
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("migration of disk %s failed: %s"),
                               disk->dst, error);
                VIR_FREE(error);
            } else {
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("migration of disk %s failed"), disk->dst);
            }
629 630
            return -1;
        }
631
        VIR_FREE(error);
632 633 634

        if (disk->mirrorState != VIR_DOMAIN_DISK_MIRROR_STATE_READY)
            notReady++;
635 636
    }

637 638 639 640 641 642 643
    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;
    }
644 645 646
}


647 648 649 650
/*
 * 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.
651
 *
652 653 654 655
 * 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.
656 657
 */
static int
658
qemuMigrationDriveMirrorCancelled(virQEMUDriverPtr driver,
659
                                  virDomainObjPtr vm,
660
                                  qemuDomainAsyncJob asyncJob,
661
                                  bool check)
662
{
663 664
    size_t i;
    size_t active = 0;
665
    size_t completed = 0;
666 667
    int status;
    bool failed = false;
668

669
 retry:
670 671 672
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
673
        char *error = NULL;
674

675 676
        if (!diskPriv->migrating)
            continue;
677

678
        status = qemuBlockJobUpdate(driver, vm, asyncJob, disk, &error);
679 680 681
        switch (status) {
        case VIR_DOMAIN_BLOCK_JOB_FAILED:
            if (check) {
682 683 684 685 686 687 688 689
                if (error) {
                    virReportError(VIR_ERR_OPERATION_FAILED,
                                   _("migration of disk %s failed: %s"),
                                   disk->dst, error);
                } else {
                    virReportError(VIR_ERR_OPERATION_FAILED,
                                   _("migration of disk %s failed"), disk->dst);
                }
690
                failed = true;
691
            }
M
Marc Hartmayer 已提交
692
            ATTRIBUTE_FALLTHROUGH;
693 694
        case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
695
            qemuBlockJobSyncEnd(driver, vm, asyncJob, disk);
696 697
            diskPriv->migrating = false;
            break;
698

699 700
        default:
            active++;
701
        }
702 703 704

        if (status == VIR_DOMAIN_BLOCK_JOB_COMPLETED)
            completed++;
705 706

        VIR_FREE(error);
707 708 709 710 711 712 713 714 715 716
    }

    /* 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;
717
    }
718

719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
    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;
735
        }
736
    }
737
}
738 739


740 741 742 743 744 745 746 747 748
/*
 * 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
qemuMigrationCancelOneDriveMirror(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  virDomainDiskDefPtr disk,
749 750
                                  bool failNoJob,
                                  qemuDomainAsyncJob asyncJob)
751 752 753
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char *diskAlias = NULL;
754
    char *error = NULL;
755 756 757 758
    int ret = -1;
    int status;
    int rv;

759
    status = qemuBlockJobUpdate(driver, vm, asyncJob, disk, &error);
760 761 762 763
    switch (status) {
    case VIR_DOMAIN_BLOCK_JOB_FAILED:
    case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        if (failNoJob) {
764 765 766 767 768 769 770 771 772
            if (error) {
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("migration of disk %s failed: %s"),
                               disk->dst, error);
            } else {
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("migration of disk %s failed"), disk->dst);
            }
            goto cleanup;
773
        }
774
        ATTRIBUTE_FALLTHROUGH;
775
    case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
776 777
        ret = 1;
        goto cleanup;
778 779
    }

780
    if (!(diskAlias = qemuAliasFromDisk(disk)))
781 782
        return -1;

783
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
784 785
        goto cleanup;

786
    rv = qemuMonitorBlockJobCancel(priv->mon, diskAlias);
787 788 789 790 791

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

    ret = 0;
792 793 794

 cleanup:
    VIR_FREE(diskAlias);
795
    VIR_FREE(error);
796 797 798 799 800 801 802 803
    return ret;
}


/**
 * qemuMigrationCancelDriveMirror:
 * @driver: qemu driver
 * @vm: domain
804
 * @check: if true report an error when some of the mirrors fails
805 806 807 808 809 810 811 812 813
 *
 * Cancel all drive-mirrors started by qemuMigrationDriveMirror.
 * Any pending block job events for the affected disks will be
 * processed.
 *
 * Returns 0 on success, -1 otherwise.
 */
static int
qemuMigrationCancelDriveMirror(virQEMUDriverPtr driver,
814
                               virDomainObjPtr vm,
815
                               bool check,
816 817
                               qemuDomainAsyncJob asyncJob,
                               virConnectPtr dconn)
818
{
819
    virErrorPtr err = NULL;
820
    int ret = -1;
821
    size_t i;
822 823 824 825
    int rv;
    bool failed = false;

    VIR_DEBUG("Cancelling drive mirrors for domain %s", vm->def->name);
826 827 828

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
829
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
830

831
        if (!diskPriv->migrating)
832 833
            continue;

834 835
        rv = qemuMigrationCancelOneDriveMirror(driver, vm, disk,
                                               check, asyncJob);
836 837 838 839 840 841
        if (rv != 0) {
            if (rv < 0) {
                if (!err)
                    err = virSaveLastError();
                failed = true;
            }
842
            qemuBlockJobSyncEnd(driver, vm, asyncJob, disk);
843 844 845 846
            diskPriv->migrating = false;
        }
    }

847 848
    while ((rv = qemuMigrationDriveMirrorCancelled(driver, vm, asyncJob,
                                                   check)) != 1) {
849 850 851 852 853 854 855
        if (check && !failed &&
            dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            failed = true;
        }

856 857 858 859
        if (rv < 0) {
            failed = true;
            if (rv == -2)
                break;
860
        }
861

862 863 864 865 866
        if (failed && !err)
            err = virSaveLastError();

        if (virDomainObjWait(vm) < 0)
            goto cleanup;
867 868
    }

869 870 871
    ret = failed ? -1 : 0;

 cleanup:
872 873 874 875 876
    if (err) {
        virSetError(err);
        virFreeError(err);
    }
    return ret;
877 878 879
}


880 881 882 883 884 885
/**
 * qemuMigrationDriveMirror:
 * @driver: qemu driver
 * @vm: domain
 * @mig: migration cookie
 * @host: where are we migrating to
886
 * @speed: bandwidth limit in MiB/s
887 888 889 890
 * @migrate_flags: migrate monitor command flags
 *
 * Run drive-mirror to feed NBD server running on dst and wait
 * till the process switches into another phase where writes go
891 892 893 894 895
 * simultaneously to both source and destination. On success,
 * update @migrate_flags so we don't tell 'migrate' command
 * to do the very same operation. On failure, the caller is
 * expected to call qemuMigrationCancelDriveMirror to stop all
 * running mirrors.
896 897 898 899 900 901 902 903 904 905
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
qemuMigrationDriveMirror(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMigrationCookiePtr mig,
                         const char *host,
                         unsigned long speed,
906 907
                         unsigned int *migrate_flags,
                         size_t nmigrate_disks,
908 909
                         const char **migrate_disks,
                         virConnectPtr dconn)
910 911 912 913
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int port;
914
    size_t i;
915 916
    char *diskAlias = NULL;
    char *nbd_dest = NULL;
917
    char *hoststr = NULL;
918
    unsigned long long mirror_speed = speed;
919
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
920
    int rv;
921
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
922 923

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
924

925 926 927 928 929 930 931 932
    if (mirror_speed > LLONG_MAX >> 20) {
        virReportError(VIR_ERR_OVERFLOW,
                       _("bandwidth must be less than %llu"),
                       LLONG_MAX >> 20);
        goto cleanup;
    }
    mirror_speed <<= 20;

933 934 935 936
    /* steal NBD port and thus prevent its propagation back to destination */
    port = mig->nbd->port;
    mig->nbd->port = 0;

937 938
    /* escape literal IPv6 address */
    if (strchr(host, ':')) {
939
        if (virAsprintf(&hoststr, "[%s]", host) < 0)
940
            goto cleanup;
941
    } else if (VIR_STRDUP(hoststr, host) < 0) {
942
        goto cleanup;
943 944
    }

945 946 947 948 949
    if (*migrate_flags & QEMU_MONITOR_MIGRATE_NON_SHARED_INC)
        mirror_flags |= VIR_DOMAIN_BLOCK_REBASE_SHALLOW;

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
950
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
951
        int mon_ret;
952

953 954
        /* check whether disk should be migrated */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks))
955 956
            continue;

957
        if (!(diskAlias = qemuAliasFromDisk(disk)) ||
958
            (virAsprintf(&nbd_dest, "nbd:%s:%d:exportname=%s",
959
                         hoststr, port, diskAlias) < 0))
960 961
            goto cleanup;

962
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
963
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
964 965
            goto cleanup;

966
        qemuBlockJobSyncBegin(disk);
967
        /* Force "raw" format for NBD export */
968
        mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest,
969
                                         "raw", mirror_speed, 0, 0, mirror_flags);
970 971
        VIR_FREE(diskAlias);
        VIR_FREE(nbd_dest);
972

973
        if (qemuDomainObjExitMonitor(driver, vm) < 0 || mon_ret < 0) {
974
            qemuBlockJobSyncEnd(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT, disk);
975 976
            goto cleanup;
        }
977
        diskPriv->migrating = true;
978

979
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
980 981 982
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto cleanup;
        }
983
    }
984

985 986
    while ((rv = qemuMigrationDriveMirrorReady(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_OUT)) != 1) {
987 988
        if (rv < 0)
            goto cleanup;
989

990
        if (priv->job.abortJob) {
991
            priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
992 993 994 995
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
            goto cleanup;
996
        }
997

998 999 1000 1001 1002 1003
        if (dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            goto cleanup;
        }

1004
        if (virDomainObjWait(vm) < 0)
1005
            goto cleanup;
1006 1007
    }

1008 1009 1010
    qemuMigrationFetchMirrorStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
                                  priv->job.current);

1011
    /* Okay, all disks are ready. Modify migrate_flags */
1012 1013 1014 1015
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

1016
 cleanup:
1017
    virObjectUnref(cfg);
1018 1019
    VIR_FREE(diskAlias);
    VIR_FREE(nbd_dest);
1020
    VIR_FREE(hoststr);
1021 1022
    return ret;
}
1023

1024

1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
/**
 * qemuMigrationIsAllowedHostdev:
 * @def: domain definition
 *
 * Checks that @def does not contain any host devices unsupported accross
 * migrations. Returns true if the vm is allowed to migrate.
 */
static bool
qemuMigrationIsAllowedHostdev(const virDomainDef *def)
{
    size_t i;

    /* Migration with USB host devices is allowed, all other devices are
     * forbidden. */
    for (i = 0; i < def->nhostdevs; i++) {
        virDomainHostdevDefPtr hostdev = def->hostdevs[i];
        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS ||
            hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("domain has assigned non-USB host devices"));
            return false;
        }
    }

    return true;
}


1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
/**
 * qemuMigrationIsAllowed:
 * @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.
 */
1067
bool
1068 1069 1070
qemuMigrationIsAllowed(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
                       bool remote,
1071
                       unsigned int flags)
1072
{
1073
    int nsnapshots;
1074
    int pauseReason;
1075
    size_t i;
1076

1077 1078 1079 1080 1081
    /* perform these checks only when migrating to remote hosts */
    if (remote) {
        nsnapshots = virDomainSnapshotObjListNum(vm->snapshots, NULL, 0);
        if (nsnapshots < 0)
            return false;
1082

1083 1084 1085 1086 1087
        if (nsnapshots > 0) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("cannot migrate domain with %d snapshots"),
                           nsnapshots);
            return false;
1088
        }
1089

1090
        /* cancel migration if disk I/O error is emitted while migrating */
1091
        if (flags & VIR_MIGRATE_ABORT_ON_ERROR &&
1092
            !(flags & VIR_MIGRATE_OFFLINE) &&
1093 1094
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
E
Eric Blake 已提交
1095
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1096
                           _("cannot migrate domain with I/O error"));
E
Eric Blake 已提交
1097 1098
            return false;
        }
1099 1100

    }
1101

1102 1103 1104 1105 1106 1107 1108
    /* following checks don't make sense for offline migration */
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        if (qemuProcessAutoDestroyActive(driver, vm)) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("domain is marked for auto destroy"));
            return false;
        }
1109

1110

1111 1112
        if (qemuDomainHasBlockjob(vm, false)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1113
                           _("domain has active block job"));
1114 1115 1116 1117 1118 1119
            return false;
        }

        if (!qemuMigrationIsAllowedHostdev(vm->def))
            return false;

1120
        if (vm->def->cpu) {
1121 1122 1123
            /* QEMU blocks migration and save with invariant TSC enabled
             * unless TSC frequency is explicitly set.
             */
1124 1125
            if (virCPUCheckFeature(vm->def->os.arch, vm->def->cpu,
                                   "invtsc") == 1) {
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143
                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;
                }
1144
            }
J
Ján Tomko 已提交
1145 1146
        }

1147 1148 1149
        /* Verify that memory device config can be transferred reliably */
        for (i = 0; i < vm->def->nmems; i++) {
            virDomainMemoryDefPtr mem = vm->def->mems[i];
1150

1151 1152 1153 1154 1155
            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"));
1156

1157 1158
                return false;
            }
1159
        }
1160 1161 1162 1163 1164 1165

        if (vm->def->nshmems) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("migration with shmem device is not supported"));
            return false;
        }
1166 1167
    }

1168 1169 1170
    return true;
}

1171
static bool
1172 1173
qemuMigrationIsSafe(virDomainDefPtr def,
                    size_t nmigrate_disks,
1174 1175 1176
                    const char **migrate_disks,
                    unsigned int flags)

1177
{
1178 1179
    bool storagemigration = flags & (VIR_MIGRATE_NON_SHARED_DISK |
                                     VIR_MIGRATE_NON_SHARED_INC);
1180
    size_t i;
1181
    int rc;
1182

1183
    for (i = 0; i < def->ndisks; i++) {
1184
        virDomainDiskDefPtr disk = def->disks[i];
1185
        const char *src = virDomainDiskGetSource(disk);
1186

1187
        /* Our code elsewhere guarantees shared disks are either readonly (in
1188
         * which case cache mode doesn't matter) or used with cache=none or used with cache=directsync */
1189 1190 1191
        if (virStorageSourceIsEmpty(disk->src) ||
            disk->src->readonly ||
            disk->src->shared ||
1192 1193
            disk->cachemode == VIR_DOMAIN_DISK_CACHE_DISABLE ||
            disk->cachemode == VIR_DOMAIN_DISK_CACHE_DIRECTSYNC)
1194
            continue;
1195

1196 1197 1198 1199
        /* disks which are migrated by qemu are safe too */
        if (storagemigration &&
            qemuMigrateDisk(disk, nmigrate_disks, migrate_disks))
            continue;
1200

1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212
        if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
            if ((rc = virFileIsSharedFS(src)) < 0)
                return false;
            else if (rc == 0)
                continue;
            if ((rc = virStorageFileIsClusterFS(src)) < 0)
                return false;
            else if (rc == 1)
                continue;
        } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                   disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
            continue;
1213
        }
1214 1215 1216

        virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                       _("Migration may lead to data corruption if disks"
1217
                         " use cache != none or cache != directsync"));
1218
        return false;
1219 1220 1221 1222 1223
    }

    return true;
}

1224 1225 1226 1227
/** qemuMigrationSetOffline
 * Pause domain for non-live migration.
 */
int
1228
qemuMigrationSetOffline(virQEMUDriverPtr driver,
1229 1230 1231
                        virDomainObjPtr vm)
{
    int ret;
1232
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
1233 1234
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
1235
    if (ret == 0) {
1236
        virObjectEventPtr event;
1237

1238
        event = virDomainEventLifecycleNewFromObj(vm,
1239 1240
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
1241
        qemuDomainEventQueue(driver, event);
1242 1243 1244 1245 1246
    }

    return ret;
}

1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288

void
qemuMigrationPostcopyFailed(virQEMUDriverPtr driver,
                            virDomainObjPtr vm)
{
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    if (state != VIR_DOMAIN_PAUSED &&
        state != VIR_DOMAIN_RUNNING)
        return;

    if (state == VIR_DOMAIN_PAUSED &&
        reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED)
        return;

    VIR_WARN("Migration of domain %s failed during post-copy; "
             "leaving the domain paused", vm->def->name);

    if (state == VIR_DOMAIN_RUNNING) {
        virObjectEventPtr event;

        if (qemuProcessStopCPUs(driver, vm,
                                VIR_DOMAIN_PAUSED_POSTCOPY_FAILED,
                                QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
            VIR_WARN("Unable to pause guest CPUs for %s", vm->def->name);
            return;
        }

        event = virDomainEventLifecycleNewFromObj(vm,
                                VIR_DOMAIN_EVENT_SUSPENDED,
                                VIR_DOMAIN_EVENT_SUSPENDED_POSTCOPY_FAILED);
        qemuDomainEventQueue(driver, event);
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
    }
}


1289
static int
1290
qemuMigrationSetOption(virQEMUDriverPtr driver,
1291
                       virDomainObjPtr vm,
1292
                       qemuMonitorMigrationCaps capability,
1293
                       bool state,
1294 1295 1296 1297 1298
                       qemuDomainAsyncJob job)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret;

1299 1300 1301 1302 1303
    if (!qemuMigrationCapsGet(vm, capability)) {
        if (!state) {
            /* Unsupported but we want it off anyway */
            return 0;
        }
1304 1305

        if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
1306 1307 1308 1309
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("Migration option '%s' is not supported by "
                             "target QEMU binary"),
                           qemuMonitorMigrationCapsTypeToString(capability));
1310
        } else {
1311 1312 1313 1314
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("Migration option '%s' is not supported by "
                             "source QEMU binary"),
                           qemuMonitorMigrationCapsTypeToString(capability));
1315
        }
1316
        return -1;
1317 1318
    }

1319 1320 1321
    if (qemuDomainObjEnterMonitorAsync(driver, vm, job) < 0)
        return -1;

1322
    ret = qemuMonitorSetMigrationCapability(priv->mon, capability, state);
1323

1324 1325
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
1326

1327 1328 1329
    return ret;
}

1330 1331 1332 1333 1334 1335 1336

static int
qemuMigrationSetPostCopy(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         bool state,
                         qemuDomainAsyncJob job)
{
1337 1338
    qemuDomainObjPrivatePtr priv = vm->privateData;

1339 1340 1341 1342 1343
    if (qemuMigrationSetOption(driver, vm,
                               QEMU_MONITOR_MIGRATION_CAPS_POSTCOPY,
                               state, job) < 0)
        return -1;

1344
    priv->job.postcopyEnabled = state;
1345 1346 1347 1348
    return 0;
}


1349
static int
1350
qemuMigrationWaitForSpice(virDomainObjPtr vm)
1351 1352 1353
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

1354 1355
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION) ||
        !priv->job.spiceMigration)
1356 1357
        return 0;

1358
    VIR_DEBUG("Waiting for SPICE to finish migration");
1359 1360
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
1361 1362 1363 1364
            return -1;
    }
    return 0;
}
1365

1366 1367 1368 1369

static void
qemuMigrationUpdateJobType(qemuDomainJobInfoPtr jobInfo)
{
1370
    switch ((qemuMonitorMigrationStatus) jobInfo->stats.mig.status) {
1371 1372 1373 1374
    case QEMU_MONITOR_MIGRATION_STATUS_POSTCOPY:
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_POSTCOPY;
        break;

1375
    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
1376
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED;
1377 1378 1379
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
1380
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_NONE;
1381 1382 1383
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
1384
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1385 1386 1387
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
1388
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
1389 1390
        break;

1391 1392 1393 1394 1395 1396 1397 1398
    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;

1399 1400 1401
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLING:
1402
    case QEMU_MONITOR_MIGRATION_STATUS_LAST:
1403 1404 1405 1406 1407 1408
        break;
    }
}


int
1409 1410 1411
qemuMigrationFetchStats(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuDomainAsyncJob asyncJob,
1412 1413
                        qemuDomainJobInfoPtr jobInfo,
                        char **error)
1414 1415
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1416
    qemuMonitorMigrationStats stats;
1417 1418 1419 1420 1421
    int rv;

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

1422
    rv = qemuMonitorGetMigrationStats(priv->mon, &stats, error);
1423 1424 1425 1426

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

1427
    jobInfo->stats.mig = stats;
1428 1429

    return 0;
1430 1431 1432
}


1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450
static const char *
qemuMigrationJobName(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
        return _("migration job");
    case QEMU_ASYNC_JOB_SAVE:
        return _("domain save job");
    case QEMU_ASYNC_JOB_DUMP:
        return _("domain core dump job");
    default:
        return _("job");
    }
}


1451 1452 1453
static int
qemuMigrationCheckJobStatus(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
1454
                            qemuDomainAsyncJob asyncJob)
1455 1456 1457
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1458
    char *error = NULL;
1459
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1460
    int ret = -1;
1461

1462
    if (!events ||
1463
        jobInfo->stats.mig.status == QEMU_MONITOR_MIGRATION_STATUS_ERROR) {
1464 1465 1466
        if (qemuMigrationFetchStats(driver, vm, asyncJob, jobInfo, &error) < 0)
            return -1;
    }
1467

1468 1469
    qemuMigrationUpdateJobType(jobInfo);

1470 1471
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_NONE:
1472 1473
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("is not active"));
1474
        goto cleanup;
1475

1476
    case QEMU_DOMAIN_JOB_STATUS_FAILED:
1477
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
1478 1479 1480
                       qemuMigrationJobName(vm),
                       error ? error : _("unexpectedly failed"));
        goto cleanup;
1481

1482
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
1483 1484
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("canceled by client"));
1485
        goto cleanup;
1486

1487 1488
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
1489
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
1490
    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
1491
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1492
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1493 1494
        break;
    }
1495 1496 1497 1498 1499 1500

    ret = 0;

 cleanup:
    VIR_FREE(error);
    return ret;
1501 1502 1503
}


1504 1505 1506
enum qemuMigrationCompletedFlags {
    QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR = (1 << 0),
    QEMU_MIGRATION_COMPLETED_CHECK_STORAGE  = (1 << 1),
1507
    QEMU_MIGRATION_COMPLETED_POSTCOPY       = (1 << 2),
1508
    QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER = (1 << 3),
1509 1510
};

1511

1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522
/**
 * 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
qemuMigrationCompleted(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
                       qemuDomainAsyncJob asyncJob,
                       virConnectPtr dconn,
1523
                       unsigned int flags)
1524 1525 1526 1527 1528
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    int pauseReason;

1529
    if (qemuMigrationCheckJobStatus(driver, vm, asyncJob) < 0)
1530 1531
        goto error;

1532
    if (flags & QEMU_MIGRATION_COMPLETED_CHECK_STORAGE &&
1533
        qemuMigrationDriveMirrorReady(driver, vm, asyncJob) < 0)
1534 1535
        goto error;

1536
    if (flags & QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR &&
1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
        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;
    }

1550 1551 1552 1553 1554 1555 1556 1557 1558 1559
    /* 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;
    }

1560 1561 1562 1563 1564
    /* 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 &&
1565
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
1566 1567 1568 1569
        VIR_DEBUG("Migration switched to post-copy");
        return 1;
    }

1570
    if (jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
1571 1572 1573 1574 1575
        return 1;
    else
        return 0;

 error:
1576 1577 1578
    switch (jobInfo->status) {
    case QEMU_DOMAIN_JOB_STATUS_MIGRATING:
    case QEMU_DOMAIN_JOB_STATUS_POSTCOPY:
1579
    case QEMU_DOMAIN_JOB_STATUS_PAUSED:
1580
        /* The migration was aborted by us rather than QEMU itself. */
1581
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1582
        return -2;
1583 1584 1585

    case QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED:
        /* Something failed after QEMU already finished the migration. */
1586
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1587
        return -1;
1588 1589 1590 1591

    case QEMU_DOMAIN_JOB_STATUS_FAILED:
    case QEMU_DOMAIN_JOB_STATUS_CANCELED:
        /* QEMU aborted the migration. */
1592
        return -1;
1593 1594 1595 1596 1597 1598

    case QEMU_DOMAIN_JOB_STATUS_ACTIVE:
    case QEMU_DOMAIN_JOB_STATUS_COMPLETED:
    case QEMU_DOMAIN_JOB_STATUS_NONE:
        /* Impossible. */
        break;
1599
    }
1600 1601

    return -1;
1602 1603 1604
}


1605 1606 1607
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
1608
static int
J
Jiri Denemark 已提交
1609 1610
qemuMigrationWaitForCompletion(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
1611
                               qemuDomainAsyncJob asyncJob,
J
Jiri Denemark 已提交
1612
                               virConnectPtr dconn,
1613
                               unsigned int flags)
1614
{
1615
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
1616
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
1617
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
1618
    int rv;
1619

1620 1621
    jobInfo->status = QEMU_DOMAIN_JOB_STATUS_MIGRATING;

1622 1623
    while ((rv = qemuMigrationCompleted(driver, vm, asyncJob,
                                        dconn, flags)) != 1) {
1624 1625
        if (rv < 0)
            return rv;
1626

1627 1628
        if (events) {
            if (virDomainObjWait(vm) < 0) {
1629
                jobInfo->status = QEMU_DOMAIN_JOB_STATUS_FAILED;
1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
                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);
        }
1640 1641
    }

1642
    if (events)
1643
        ignore_value(qemuMigrationFetchStats(driver, vm, asyncJob, jobInfo, NULL));
1644

1645
    qemuDomainJobInfoUpdateTime(jobInfo);
1646 1647
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
1648
    if (VIR_ALLOC(priv->job.completed) == 0) {
1649
        *priv->job.completed = *jobInfo;
1650 1651
        priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
    }
1652

1653 1654 1655 1656
    if (asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT &&
        jobInfo->status == QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED)
        jobInfo->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;

1657
    return 0;
1658 1659 1660
}


1661 1662 1663
static int
qemuMigrationWaitForDestCompletion(virQEMUDriverPtr driver,
                                   virDomainObjPtr vm,
1664 1665
                                   qemuDomainAsyncJob asyncJob,
                                   bool postcopy)
1666 1667
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1668
    unsigned int flags = 0;
1669 1670 1671 1672 1673 1674 1675
    int rv;

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

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

1676 1677 1678 1679 1680
    if (postcopy)
        flags = QEMU_MIGRATION_COMPLETED_POSTCOPY;

    while ((rv = qemuMigrationCompleted(driver, vm, asyncJob,
                                        NULL, flags)) != 1) {
1681 1682 1683 1684 1685 1686 1687 1688
        if (rv < 0 || virDomainObjWait(vm) < 0)
            return -1;
    }

    return 0;
}


1689
static int
1690
qemuDomainMigrateGraphicsRelocate(virQEMUDriverPtr driver,
1691
                                  virDomainObjPtr vm,
1692 1693
                                  qemuMigrationCookiePtr cookie,
                                  const char *graphicsuri)
1694 1695
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1696 1697
    int ret = -1;
    const char *listenAddress = NULL;
1698
    virSocketAddr addr;
1699 1700 1701 1702 1703
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
1704

1705
    if (!cookie || (!cookie->graphics && !graphicsuri))
1706 1707
        return 0;

1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726
    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) {
1727
        size_t i;
1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754

        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;
            }
        }
    }
1755 1756 1757 1758

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
1759 1760 1761 1762
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
1763

1764 1765 1766 1767 1768 1769 1770 1771 1772
    /* 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;
    }

1773 1774 1775 1776
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
1777
        priv->job.spiceMigration = !ret;
1778 1779
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
1780
    }
1781

1782
 cleanup:
1783
    virURIFree(uri);
1784 1785 1786 1787
    return ret;
}


1788
static int
1789
qemuDomainMigrateOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
1790 1791 1792
                             virDomainObjPtr vm,
                             qemuMigrationCookiePtr cookie)
{
1793 1794
    virDomainNetDefPtr netptr;
    int ret = -1;
1795
    size_t i;
1796 1797 1798 1799 1800 1801 1802 1803

    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:
1804
           break;
1805
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
1806 1807
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
1808 1809 1810
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
1811 1812 1813
                goto cleanup;
            }
            break;
1814 1815 1816 1817 1818
        default:
            break;
        }
    }

1819
    ret = 0;
1820
 cleanup:
1821 1822 1823 1824
    return ret;
}


1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864
int
qemuMigrationCheckIncoming(virQEMUCapsPtr qemuCaps,
                           const char *migrateFrom)
{
    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 *
qemuMigrationIncomingURI(const char *migrateFrom,
                         int migrateFd)
{
    char *uri = NULL;

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

    return uri;
}


1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890
int
qemuMigrationRunIncoming(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         const char *uri,
                         qemuDomainAsyncJob asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int rv;

    VIR_DEBUG("Setting up incoming migration with URI %s", uri);

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

    rv = qemuMonitorMigrateIncoming(priv->mon, uri);

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

    if (asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN) {
        /* qemuMigrationWaitForDestCompletion is called from the Finish phase */
        ret = 0;
        goto cleanup;
    }

1891
    if (qemuMigrationWaitForDestCompletion(driver, vm, asyncJob, false) < 0)
1892 1893 1894 1895 1896 1897 1898 1899 1900
        goto cleanup;

    ret = 0;

 cleanup:
    return ret;
}


1901 1902 1903 1904 1905 1906
/* 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.
 */
1907 1908 1909 1910
static virDomainObjPtr
qemuMigrationCleanup(virDomainObjPtr vm,
                     virConnectPtr conn,
                     void *opaque)
1911
{
1912
    virQEMUDriverPtr driver = opaque;
1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
    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))
        goto cleanup;

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

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

1959
 cleanup:
1960 1961 1962
    return vm;
}

1963

1964
/* The caller is supposed to lock the vm and start a migration job. */
1965 1966 1967 1968 1969 1970 1971
static char *
qemuMigrationBeginPhase(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dname,
                        char **cookieout,
                        int *cookieoutlen,
1972 1973
                        size_t nmigrate_disks,
                        const char **migrate_disks,
1974
                        unsigned long flags)
1975 1976 1977
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
1978
    virDomainDefPtr def = NULL;
1979
    qemuDomainObjPrivatePtr priv = vm->privateData;
1980
    virCapsPtr caps = NULL;
1981
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
1982

1983
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
1984
              " cookieout=%p, cookieoutlen=%p,"
1985
              " nmigrate_disks=%zu, migrate_disks=%p, flags=0x%lx",
1986
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
1987 1988
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
1989

1990 1991 1992
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

1993 1994 1995 1996 1997 1998
    /* 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);
1999

2000
    if (!qemuMigrationIsAllowed(driver, vm, true, flags))
2001 2002
        goto cleanup;

2003
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
2004
        !qemuMigrationIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
2005 2006
        goto cleanup;

2007 2008 2009 2010 2011 2012 2013 2014 2015
    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;
    }

2016 2017 2018 2019 2020 2021
    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;
    }

2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064
    if (flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC)) {
        bool has_drive_mirror =  virQEMUCapsGet(priv->qemuCaps,
                                                QEMU_CAPS_DRIVE_MIRROR);

        if (nmigrate_disks) {
            if (has_drive_mirror) {
                size_t i, j;
                /* Check user requested only known disk targets. */
                for (i = 0; i < nmigrate_disks; i++) {
                    for (j = 0; j < vm->def->ndisks; j++) {
                        if (STREQ(vm->def->disks[j]->dst, migrate_disks[i]))
                            break;
                    }

                    if (j == vm->def->ndisks) {
                        virReportError(VIR_ERR_INVALID_ARG,
                                       _("disk target %s not found"),
                                       migrate_disks[i]);
                        goto cleanup;
                    }
                }

                if (flags & VIR_MIGRATE_TUNNELLED) {
                    virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                                   _("Selecting disks to migrate is not "
                                     "implemented for tunnelled migration"));
                    goto cleanup;
                }
            } else {
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                               _("qemu does not support drive-mirror command"));
                goto cleanup;
            }
        }

        if (has_drive_mirror) {
            /* TODO support NBD for TUNNELLED migration */
            if (flags & VIR_MIGRATE_TUNNELLED) {
                VIR_WARN("NBD in tunnelled migration is currently not supported");
            } else {
                cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
                priv->nbdPort = 0;
            }
2065 2066 2067
        }
    }

2068
    if (virDomainDefHasMemoryHotplug(vm->def) ||
2069
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
2070
         vm->newDef && virDomainDefHasMemoryHotplug(vm->newDef)))
2071 2072
        cookieFlags |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

2073 2074 2075 2076 2077
    if (!qemuDomainVcpuHotplugIsInOrder(vm->def) ||
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
         vm->newDef && !qemuDomainVcpuHotplugIsInOrder(vm->newDef)))
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU_HOTPLUG;

2078 2079 2080
    if (priv->origCPU)
        cookieFlags |= QEMU_MIGRATION_COOKIE_CPU;

2081 2082
    cookieFlags |= QEMU_MIGRATION_COOKIE_ALLOW_REBOOT;

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

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

L
liguang 已提交
2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112
    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;
        }
    }

2113
    if (xmlin) {
2114
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt, priv->qemuCaps,
2115 2116
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                            VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
2117 2118
            goto cleanup;

2119
        if (!qemuDomainCheckABIStability(driver, vm, def))
2120 2121
            goto cleanup;

2122
        rv = qemuDomainDefFormatLive(driver, def, NULL, false, true);
2123
    } else {
2124 2125
        rv = qemuDomainDefFormatLive(driver, vm->def, priv->origCPU,
                                     false, true);
2126
    }
2127

2128
 cleanup:
2129
    qemuMigrationCookieFree(mig);
2130
    virObjectUnref(caps);
2131
    virDomainDefFree(def);
2132 2133 2134
    return rv;
}

2135 2136 2137 2138 2139 2140 2141
char *
qemuMigrationBegin(virConnectPtr conn,
                   virDomainObjPtr vm,
                   const char *xmlin,
                   const char *dname,
                   char **cookieout,
                   int *cookieoutlen,
2142 2143
                   size_t nmigrate_disks,
                   const char **migrate_disks,
2144 2145 2146
                   unsigned long flags)
{
    virQEMUDriverPtr driver = conn->privateData;
2147
    virQEMUDriverConfigPtr cfg = NULL;
2148
    char *xml = NULL;
2149
    qemuDomainAsyncJob asyncJob;
2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
        if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_MIGRATION_OUT;
    } else {
        if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
            goto cleanup;
        asyncJob = QEMU_ASYNC_JOB_NONE;
    }

2161 2162
    qemuMigrationStoreDomainState(vm);

2163 2164 2165 2166 2167 2168 2169 2170 2171 2172
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
        goto endjob;
    }

    /* Check if there is any ejected media.
     * We don't want to require them on the destination.
     */
    if (!(flags & VIR_MIGRATE_OFFLINE) &&
2173
        qemuProcessRefreshDisks(driver, vm, asyncJob) < 0)
2174 2175 2176 2177
        goto endjob;

    if (!(xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
                                        cookieout, cookieoutlen,
2178
                                        nmigrate_disks, migrate_disks, flags)))
2179 2180
        goto endjob;

2181 2182 2183 2184 2185 2186
    if (flags & VIR_MIGRATE_TLS) {
        cfg = virQEMUDriverGetConfig(driver);
        if (qemuMigrationCheckSetupTLS(conn, driver, cfg, vm, asyncJob) < 0)
            goto endjob;
    }

2187 2188 2189 2190 2191
    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.
         */
2192
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
2193 2194
                                 qemuMigrationCleanup) < 0) {
            VIR_FREE(xml);
2195
            goto endjob;
2196
        }
2197
        qemuMigrationJobContinue(vm);
2198 2199 2200 2201
    } else {
        goto endjob;
    }

2202
 cleanup:
2203
    virObjectUnref(cfg);
M
Michal Privoznik 已提交
2204
    virDomainObjEndAPI(&vm);
2205 2206
    return xml;

2207
 endjob:
2208 2209 2210 2211
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2212 2213 2214
    goto cleanup;
}

2215

2216 2217
/* Prepare is the first step, and it runs on the destination host.
 */
2218

2219
static void
2220
qemuMigrationPrepareCleanup(virQEMUDriverPtr driver,
2221 2222 2223 2224 2225 2226 2227 2228 2229 2230
                            virDomainObjPtr vm)
{
    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));

2231 2232 2233
    virPortAllocatorRelease(driver->migrationPorts, priv->migrationPort);
    priv->migrationPort = 0;

2234 2235 2236 2237 2238
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2239
static qemuProcessIncomingDefPtr
2240 2241 2242 2243
qemuMigrationPrepareIncoming(virDomainObjPtr vm,
                             bool tunnel,
                             const char *protocol,
                             const char *listenAddress,
2244 2245
                             unsigned short port,
                             int fd)
2246 2247
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2248
    qemuProcessIncomingDefPtr inc = NULL;
2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310
    char *migrateFrom = NULL;

    if (tunnel) {
        if (VIR_STRDUP(migrateFrom, "stdio") < 0)
            goto cleanup;
    } else {
        bool encloseAddress = false;
        bool hostIPv6Capable = false;
        bool qemuIPv6Capable = 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;
        }
        qemuIPv6Capable = virQEMUCapsGet(priv->qemuCaps,
                                         QEMU_CAPS_IPV6_MIGRATION);

        if (listenAddress) {
            if (virSocketAddrNumericFamily(listenAddress) == AF_INET6) {
                if (!qemuIPv6Capable) {
                    virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                                   _("qemu isn't capable of IPv6"));
                    goto cleanup;
                }
                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 */
            }
        } else if (qemuIPv6Capable && hostIPv6Capable) {
            /* Listen on :: instead of 0.0.0.0 if QEMU understands it
             * and there is at least one IPv6 address configured
             */
            listenAddress = "::";
            encloseAddress = true;
        } else {
            listenAddress = "0.0.0.0";
        }

        /* QEMU will be started with
         *   -incoming protocol:[<IPv6 addr>]:port,
         *   -incoming protocol:<IPv4 addr>:port, or
         *   -incoming protocol:<hostname>:port
         */
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";
        if (virAsprintf(&migrateFrom, incFormat,
                        protocol, listenAddress, port) < 0)
            goto cleanup;
    }

2311 2312
    inc = qemuProcessIncomingDefNew(priv->qemuCaps, listenAddress,
                                    migrateFrom, fd, NULL);
2313

2314
 cleanup:
2315 2316
    VIR_FREE(migrateFrom);
    return inc;
2317 2318
}

2319 2320 2321 2322
static int
qemuMigrationSetCompression(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
                            qemuDomainAsyncJob job,
2323 2324
                            qemuMigrationCompressionPtr compression,
                            qemuMonitorMigrationParamsPtr migParams)
2325
{
2326 2327 2328
    int ret = -1;
    qemuDomainObjPrivatePtr priv = vm->privateData;

2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342
    if (qemuMigrationSetOption(driver, vm,
                               QEMU_MONITOR_MIGRATION_CAPS_XBZRLE,
                               compression->methods &
                                   (1ULL << QEMU_MIGRATION_COMPRESS_XBZRLE),
                               job) < 0)
        return -1;

    if (qemuMigrationSetOption(driver, vm,
                               QEMU_MONITOR_MIGRATION_CAPS_COMPRESS,
                               compression->methods &
                                   (1ULL << QEMU_MIGRATION_COMPRESS_MT),
                               job) < 0)
        return -1;

2343 2344 2345
    if (qemuDomainObjEnterMonitorAsync(driver, vm, job) < 0)
        return -1;

2346 2347
    migParams->compressLevel_set = compression->level_set;
    migParams->compressLevel = compression->level;
2348

2349 2350
    migParams->compressThreads_set = compression->threads_set;
    migParams->compressThreads = compression->threads;
2351

2352 2353
    migParams->decompressThreads_set = compression->dthreads_set;
    migParams->decompressThreads = compression->dthreads;
2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366

    if (compression->xbzrle_cache_set &&
        qemuMonitorSetMigrationCacheSize(priv->mon,
                                         compression->xbzrle_cache) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

    return ret;
2367 2368
}

2369

2370 2371 2372 2373 2374 2375
void
qemuMigrationParamsClear(qemuMonitorMigrationParamsPtr migParams)
{
    if (!migParams)
        return;

2376 2377
    VIR_FREE(migParams->tlsCreds);
    VIR_FREE(migParams->tlsHostname);
2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391
}


void
qemuMigrationParamsFree(qemuMonitorMigrationParamsPtr *migParams)
{
    if (!*migParams)
        return;

    qemuMigrationParamsClear(*migParams);
    VIR_FREE(*migParams);
}


2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417
/* qemuMigrationSetEmptyTLSParams
 * @driver: pointer to qemu driver
 * @vm: domain object
 * @asyncJob: migration job to join
 * @migParams: Pointer to a migration parameters block
 *
 * If we support setting the tls-creds, then set both tls-creds and
 * tls-hostname to the empty string ("") which indicates to not use
 * TLS on this migration.
 *
 * Returns 0 on success, -1 on failure
 */
static int
qemuMigrationSetEmptyTLSParams(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
                               qemuDomainAsyncJob asyncJob,
                               qemuMonitorMigrationParamsPtr migParams)
{
   qemuDomainObjPrivatePtr priv = vm->privateData;

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

   if (!priv->migTLSAlias)
       return 0;

2418 2419
   if (VIR_STRDUP(migParams->tlsCreds, "") < 0 ||
       VIR_STRDUP(migParams->tlsHostname, "") < 0)
2420 2421 2422 2423 2424 2425
       return -1;

    return 0;
}


2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438
qemuMonitorMigrationParamsPtr
qemuMigrationParams(virTypedParameterPtr params,
                    int nparams,
                    unsigned long flags)
{
    qemuMonitorMigrationParamsPtr migParams;

    if (VIR_ALLOC(migParams) < 0)
        return NULL;

    if (!params)
        return migParams;

2439 2440 2441 2442 2443 2444 2445 2446 2447 2448
#define GET(PARAM, VAR) \
    do { \
        int rc; \
        if ((rc = virTypedParamsGetInt(params, nparams, \
                                       VIR_MIGRATE_PARAM_ ## PARAM, \
                                       &migParams->VAR)) < 0) \
            goto error; \
 \
        if (rc == 1) \
            migParams->VAR ## _set = true; \
2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466
    } while (0)

    GET(AUTO_CONVERGE_INITIAL, cpuThrottleInitial);
    GET(AUTO_CONVERGE_INCREMENT, cpuThrottleIncrement);

#undef GET

    if ((migParams->cpuThrottleInitial_set ||
         migParams->cpuThrottleIncrement_set) &&
        !(flags & VIR_MIGRATE_AUTO_CONVERGE)) {
        virReportError(VIR_ERR_INVALID_ARG, "%s",
                       _("Turn auto convergence on to tune it"));
        goto error;
    }

    return migParams;

 error:
2467
    qemuMigrationParamsFree(&migParams);
2468 2469 2470 2471
    return NULL;
}


2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496
static int
qemuMigrationSetParams(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
                       qemuDomainAsyncJob job,
                       qemuMonitorMigrationParamsPtr migParams)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;

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

    if (qemuMonitorSetMigrationParams(priv->mon, migParams) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

    return ret;
}


2497 2498 2499 2500 2501 2502
/* qemuMigrationResetTLS
 * @driver: pointer to qemu driver
 * @vm: domain object
 * @asyncJob: migration job to join
 *
 * Deconstruct all the setup possibly done for TLS - delete the TLS and
2503
 * security objects, free the secinfo, and reset the migration params to "".
2504 2505 2506
 *
 * Returns 0 on success, -1 on failure
 */
2507
static int
2508 2509
qemuMigrationResetTLS(virQEMUDriverPtr driver,
                      virDomainObjPtr vm,
2510
                      qemuDomainAsyncJob asyncJob)
2511 2512
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2513 2514
    char *tlsAlias = NULL;
    char *secAlias = NULL;
2515
    qemuMonitorMigrationParams migParams = { 0 };
R
Roman Bogorodskiy 已提交
2516
    int ret = -1;
2517

2518
    if (qemuMigrationCheckTLSCreds(driver, vm, asyncJob) < 0)
2519 2520
        return -1;

2521 2522
    /* If the tls-creds doesn't exist or if they're set to "" then there's
     * nothing to do since we never set anything up */
2523 2524 2525 2526 2527 2528
    if (!priv->migTLSAlias || !*priv->migTLSAlias)
        return 0;

    /* NB: If either or both fail to allocate memory we can still proceed
     *     since the next time we migrate another deletion attempt will be
     *     made after successfully generating the aliases. */
2529 2530
    tlsAlias = qemuAliasTLSObjFromSrcAlias(QEMU_MIGRATION_TLS_ALIAS_BASE);
    secAlias = qemuDomainGetSecretAESAlias(QEMU_MIGRATION_TLS_ALIAS_BASE, false);
2531 2532 2533 2534

    qemuDomainDelTLSObjects(driver, vm, asyncJob, secAlias, tlsAlias);
    qemuDomainSecretInfoFree(&priv->migSecinfo);

2535 2536
    if (VIR_STRDUP(migParams.tlsCreds, "") < 0 ||
        VIR_STRDUP(migParams.tlsHostname, "") < 0 ||
2537 2538 2539 2540 2541 2542
        qemuMigrationSetParams(driver, vm, asyncJob, &migParams) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
2543 2544
    VIR_FREE(tlsAlias);
    VIR_FREE(secAlias);
2545 2546 2547 2548 2549 2550
    qemuMigrationParamsClear(&migParams);

    return ret;
}


2551
static int
2552
qemuMigrationPrepareAny(virQEMUDriverPtr driver,
2553 2554 2555 2556 2557
                        virConnectPtr dconn,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
2558
                        virDomainDefPtr *def,
2559
                        const char *origname,
L
liguang 已提交
2560
                        virStreamPtr st,
2561
                        const char *protocol,
2562 2563
                        unsigned short port,
                        bool autoPort,
2564
                        const char *listenAddress,
2565 2566
                        size_t nmigrate_disks,
                        const char **migrate_disks,
2567
                        int nbdPort,
2568
                        qemuMigrationCompressionPtr compression,
L
liguang 已提交
2569
                        unsigned long flags)
2570 2571
{
    virDomainObjPtr vm = NULL;
2572
    virObjectEventPtr event = NULL;
2573
    virQEMUDriverConfigPtr cfg = NULL;
2574
    int ret = -1;
2575
    int dataFD[2] = { -1, -1 };
2576
    qemuDomainObjPrivatePtr priv = NULL;
2577
    qemuMigrationCookiePtr mig = NULL;
2578
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2579
    char *xmlout = NULL;
L
liguang 已提交
2580
    unsigned int cookieFlags;
2581
    unsigned int startFlags;
2582
    virCapsPtr caps = NULL;
2583
    qemuProcessIncomingDefPtr incoming = NULL;
2584
    bool taint_hook = false;
2585 2586 2587
    bool stopProcess = false;
    bool relabel = false;
    int rv;
2588 2589
    char *tlsAlias = NULL;
    char *secAlias = NULL;
2590
    qemuMonitorMigrationParams migParams = { 0 };
2591

2592 2593
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613
    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;
        }
2614 2615 2616
        cookieFlags = 0;
    } else {
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS;
L
liguang 已提交
2617 2618
    }

2619 2620 2621 2622 2623 2624 2625 2626 2627
    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;
    }

2628 2629 2630 2631 2632 2633
    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;
    }

2634 2635 2636
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2637
    if (!qemuMigrationIsAllowedHostdev(*def))
2638 2639
        goto cleanup;

J
Jiri Denemark 已提交
2640 2641 2642 2643 2644
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2645
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2646 2647
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2648 2649
            goto cleanup;

2650
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2651 2652 2653 2654 2655 2656 2657
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2658
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2659 2660 2661 2662 2663 2664
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2665
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt, NULL,
2666 2667
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                                 VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE);
J
Jiri Denemark 已提交
2668 2669 2670
                if (!newdef)
                    goto cleanup;

2671
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2672 2673 2674 2675
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2676 2677
                virDomainDefFree(*def);
                *def = newdef;
2678 2679 2680 2681
                /* 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
2682 2683 2684 2685
            }
        }
    }

2686
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2687
                                   driver->xmlopt,
2688 2689 2690
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2691
        goto cleanup;
2692

2693
    virObjectRef(vm);
2694
    *def = NULL;
2695
    priv = vm->privateData;
2696 2697
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2698

2699 2700 2701 2702 2703
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2704
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
2705
                                       QEMU_MIGRATION_COOKIE_LOCKSTATE |
2706
                                       QEMU_MIGRATION_COOKIE_NBD |
2707
                                       QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG |
2708
                                       QEMU_MIGRATION_COOKIE_CPU_HOTPLUG |
2709 2710
                                       QEMU_MIGRATION_COOKIE_CPU |
                                       QEMU_MIGRATION_COOKIE_ALLOW_REBOOT)))
2711 2712
        goto cleanup;

2713 2714
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
2715 2716 2717 2718 2719 2720
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2721
    if (qemuMigrationPrecreateStorage(dconn, driver, vm, mig->nbd,
2722 2723
                                      nmigrate_disks, migrate_disks,
                                      !!(flags & VIR_MIGRATE_NON_SHARED_INC)) < 0)
2724 2725
        goto cleanup;

2726
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2727
        goto cleanup;
2728
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2729 2730 2731 2732

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

L
liguang 已提交
2733 2734 2735
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2736 2737
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
2738 2739
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
2740
        goto stopjob;
2741 2742
    }

2743 2744
    startFlags = VIR_QEMU_PROCESS_START_AUTODESTROY;

2745
    if (qemuProcessInit(driver, vm, mig->cpu, QEMU_ASYNC_JOB_MIGRATION_IN,
2746
                        true, startFlags) < 0)
2747
        goto stopjob;
2748
    stopProcess = true;
2749

2750 2751
    priv->allowReboot = mig->allowReboot;

2752 2753 2754
    if (!(incoming = qemuMigrationPrepareIncoming(vm, tunnel, protocol,
                                                  listenAddress, port,
                                                  dataFD[0])))
2755
        goto stopjob;
2756

2757
    if (qemuProcessPrepareDomain(dconn, driver, vm, startFlags) < 0)
2758 2759
        goto stopjob;

2760
    if (qemuProcessPrepareHost(driver, vm, startFlags) < 0)
2761 2762
        goto stopjob;

2763 2764 2765
    rv = qemuProcessLaunch(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                           incoming, NULL,
                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
2766
                           startFlags);
2767 2768 2769
    if (rv < 0) {
        if (rv == -2)
            relabel = true;
2770
        goto stopjob;
2771
    }
2772
    relabel = true;
2773

2774 2775 2776 2777
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
2778
            goto stopjob;
2779
        }
2780
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
2781 2782
    }

2783
    if (qemuMigrationSetCompression(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2784
                                    compression, &migParams) < 0)
2785
        goto stopjob;
2786

2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800
    /* Migrations using TLS need to add the "tls-creds-x509" object and
     * set the migration TLS parameters */
    if (flags & VIR_MIGRATE_TLS) {
        cfg = virQEMUDriverGetConfig(driver);
        if (qemuMigrationCheckSetupTLS(dconn, driver, cfg, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
            goto stopjob;

        if (qemuMigrationAddTLSObjects(driver, vm, cfg, true,
                                       QEMU_ASYNC_JOB_MIGRATION_IN,
                                       &tlsAlias, &secAlias, &migParams) < 0)
            goto stopjob;

        /* Force reset of 'tls-hostname', it's a source only parameter */
2801
        if (VIR_STRDUP(migParams.tlsHostname, "") < 0)
2802 2803 2804 2805 2806 2807 2808 2809 2810
            goto stopjob;

    } else {
        if (qemuMigrationSetEmptyTLSParams(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_IN,
                                           &migParams) < 0)
            goto stopjob;
    }

2811
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
2812
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
2813
        goto stopjob;
M
Michael R. Hines 已提交
2814 2815
    }

2816 2817
    if (qemuMigrationSetOption(driver, vm,
                               QEMU_MONITOR_MIGRATION_CAPS_RDMA_PIN_ALL,
2818 2819
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
                               QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2820
        goto stopjob;
2821

2822 2823 2824 2825 2826
    if (qemuMigrationSetPostCopy(driver, vm,
                                 flags & VIR_MIGRATE_POSTCOPY,
                                 QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stopjob;

2827 2828 2829 2830
    if (qemuMigrationSetParams(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                               &migParams) < 0)
        goto stopjob;

2831 2832 2833
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
2834
        if (qemuMigrationStartNBDServer(driver, vm, incoming->address,
2835 2836
                                        nmigrate_disks, migrate_disks,
                                        nbdPort) < 0) {
2837
            goto stopjob;
2838
        }
2839
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
2840 2841
    }

2842 2843 2844 2845 2846 2847 2848 2849 2850
    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");
    }

2851 2852 2853 2854 2855
    if (incoming->deferredURI &&
        qemuMigrationRunIncoming(driver, vm, incoming->deferredURI,
                                 QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stopjob;

2856
    if (qemuProcessFinishStartup(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
2857 2858 2859
                                 false, VIR_DOMAIN_PAUSED_MIGRATION) < 0)
        goto stopjob;

2860
 done:
2861 2862
    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
2863 2864 2865 2866 2867 2868 2869
        /* 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");
    }

2870
    if (qemuDomainCleanupAdd(vm, qemuMigrationPrepareCleanup) < 0)
2871
        goto stopjob;
2872

L
liguang 已提交
2873 2874
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
2875
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
2876 2877 2878
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
2879

2880 2881 2882 2883
    /* 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.
     */
2884
    qemuMigrationJobContinue(vm);
2885

2886 2887
    if (autoPort)
        priv->migrationPort = port;
2888 2889 2890 2891 2892
    /* 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;
2893
    ret = 0;
2894

2895
 cleanup:
2896 2897 2898
    VIR_FREE(tlsAlias);
    VIR_FREE(secAlias);
    virObjectUnref(cfg);
2899
    qemuProcessIncomingDefFree(incoming);
J
Jiri Denemark 已提交
2900
    VIR_FREE(xmlout);
2901 2902
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
2903 2904 2905
    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 */
2906
        VIR_FREE(priv->origname);
2907 2908 2909 2910 2911
        /* release if port is auto selected which is not the case if
         * it is given in parameters
         */
        if (nbdPort == 0)
            virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
2912
        priv->nbdPort = 0;
2913
        virDomainObjRemoveTransientDef(vm);
2914
        qemuDomainRemoveInactiveJob(driver, vm);
2915
    }
2916
    qemuMigrationParamsClear(&migParams);
M
Michal Privoznik 已提交
2917
    virDomainObjEndAPI(&vm);
2918
    qemuDomainEventQueue(driver, event);
2919
    qemuMigrationCookieFree(mig);
2920
    virObjectUnref(caps);
2921
    virNWFilterUnlockFilterUpdates();
2922
    return ret;
2923

2924
 stopjob:
2925
    qemuMigrationReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN);
2926

2927 2928 2929 2930
    if (stopProcess) {
        unsigned int stopFlags = VIR_QEMU_PROCESS_STOP_MIGRATED;
        if (!relabel)
            stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
2931
        virDomainAuditStart(vm, "migrated", false);
2932 2933
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        QEMU_ASYNC_JOB_MIGRATION_IN, stopFlags);
2934
    }
2935

2936
    qemuMigrationJobFinish(driver, vm);
2937
    goto cleanup;
2938 2939 2940
}


2941 2942 2943 2944 2945
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
2946
qemuMigrationPrepareTunnel(virQEMUDriverPtr driver,
2947 2948 2949 2950 2951 2952
                           virConnectPtr dconn,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virStreamPtr st,
2953
                           virDomainDefPtr *def,
2954
                           const char *origname,
L
liguang 已提交
2955
                           unsigned long flags)
2956
{
2957
    qemuMigrationCompressionPtr compression = NULL;
2958 2959 2960
    int ret;

    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2961
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
2962
              "origname=%s, flags=0x%lx",
2963
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2964
              cookieout, cookieoutlen, st, *def, origname, flags);
2965

2966 2967 2968 2969 2970 2971
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

2972 2973 2974
    if (!(compression = qemuMigrationCompressionParse(NULL, 0, flags)))
        return -1;

2975
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
2976
                                  cookieout, cookieoutlen, def, origname,
2977 2978 2979
                                  st, NULL, 0, false, NULL, 0, NULL, 0,
                                  compression, flags);
    VIR_FREE(compression);
2980 2981 2982 2983
    return ret;
}


2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006
static virURIPtr
qemuMigrationParseURI(const char *uri, bool *wellFormed)
{
    char *tmp = NULL;
    virURIPtr parsed;

    /* For compatibility reasons tcp://... URIs are sent as tcp:...
     * We need to transform them to a well-formed URI before parsing. */
    if (STRPREFIX(uri, "tcp:") && !STRPREFIX(uri + 4, "//")) {
        if (virAsprintf(&tmp, "tcp://%s", uri + 4) < 0)
            return NULL;
        uri = tmp;
    }

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

    return parsed;
}


3007
int
3008
qemuMigrationPrepareDirect(virQEMUDriverPtr driver,
3009
                           virConnectPtr dconn,
3010 3011 3012 3013
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
3014 3015
                           const char *uri_in,
                           char **uri_out,
3016
                           virDomainDefPtr *def,
3017
                           const char *origname,
3018
                           const char *listenAddress,
3019 3020
                           size_t nmigrate_disks,
                           const char **migrate_disks,
3021
                           int nbdPort,
3022
                           qemuMigrationCompressionPtr compression,
L
liguang 已提交
3023
                           unsigned long flags)
3024
{
3025 3026
    unsigned short port = 0;
    bool autoPort = true;
3027 3028
    char *hostname = NULL;
    int ret = -1;
3029
    virURIPtr uri = NULL;
3030 3031
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
3032

3033 3034
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
3035
              "def=%p, origname=%s, listenAddress=%s, "
3036
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, flags=0x%lx",
3037 3038
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
3039
              *def, origname, NULLSTR(listenAddress),
3040
              nmigrate_disks, migrate_disks, nbdPort, flags);
3041

3042 3043
    *uri_out = NULL;

3044 3045 3046
    /* 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
3047 3048 3049
     * 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".
3050 3051 3052 3053 3054 3055
     *
     * 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) {
3056 3057 3058
        bool encloseAddress = false;
        const char *incFormat;

3059
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3060
            goto cleanup;
3061

3062
        if (migrateHost != NULL) {
3063 3064
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
3065

3066
            if (VIR_STRDUP(hostname, migrateHost) < 0)
3067 3068 3069 3070 3071
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
3072 3073

        if (STRPREFIX(hostname, "localhost")) {
3074 3075 3076
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
3077 3078 3079 3080 3081
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
3082
         * compatibility with old targets. We at least make the
3083 3084
         * new targets accept both syntaxes though.
         */
3085 3086 3087 3088 3089 3090
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
3091 3092
            goto cleanup;
    } else {
3093
        bool well_formed_uri;
J
Ján Tomko 已提交
3094

3095 3096
        if (!(uri = qemuMigrationParseURI(uri_in, &well_formed_uri)))
            goto cleanup;
J
Ján Tomko 已提交
3097

3098 3099 3100 3101 3102 3103 3104
        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
3105 3106
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
3107 3108 3109
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
3110 3111 3112 3113 3114 3115 3116 3117 3118 3119
            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) {
3120
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3121
                goto cleanup;
3122

3123
            /* Send well-formed URI only if uri_in was well-formed */
3124 3125 3126 3127 3128 3129 3130 3131
            if (well_formed_uri) {
                uri->port = port;
                if (!(*uri_out = virURIFormat(uri)))
                    goto cleanup;
            } else {
                if (virAsprintf(uri_out, "%s:%d", uri_in, port) < 0)
                    goto cleanup;
            }
3132
        } else {
3133 3134
            port = uri->port;
            autoPort = false;
3135 3136 3137 3138 3139 3140
        }
    }

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

3141
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3142
                                  cookieout, cookieoutlen, def, origname,
3143
                                  NULL, uri ? uri->scheme : "tcp",
3144
                                  port, autoPort, listenAddress,
3145 3146
                                  nmigrate_disks, migrate_disks, nbdPort,
                                  compression, flags);
3147
 cleanup:
3148
    virURIFree(uri);
3149
    VIR_FREE(hostname);
3150
    virObjectUnref(cfg);
3151
    if (ret != 0) {
3152
        VIR_FREE(*uri_out);
3153 3154 3155
        if (autoPort)
            virPortAllocatorRelease(driver->migrationPorts, port);
    }
3156 3157 3158 3159
    return ret;
}


3160 3161 3162
virDomainDefPtr
qemuMigrationPrepareDef(virQEMUDriverPtr driver,
                        const char *dom_xml,
3163 3164
                        const char *dname,
                        char **origname)
3165 3166 3167
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
3168
    char *name = NULL;
3169 3170 3171 3172 3173 3174 3175 3176 3177 3178

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

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

3179
    if (!(def = virDomainDefParseString(dom_xml, caps, driver->xmlopt, NULL,
3180 3181
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                        VIR_DOMAIN_DEF_PARSE_SKIP_VALIDATE)))
3182 3183 3184
        goto cleanup;

    if (dname) {
3185
        name = def->name;
3186 3187 3188 3189 3190 3191
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

3192
 cleanup:
3193
    virObjectUnref(caps);
3194 3195 3196 3197
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
3198 3199 3200 3201
    return def;
}


3202 3203 3204 3205 3206 3207 3208 3209 3210 3211
static int
qemuMigrationConfirmPhase(virQEMUDriverPtr driver,
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *cookiein,
                          int cookieinlen,
                          unsigned int flags,
                          int retcode)
{
    qemuMigrationCookiePtr mig;
3212
    virObjectEventPtr event;
3213 3214
    int rv = -1;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3215 3216
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = NULL;
3217 3218

    VIR_DEBUG("driver=%p, conn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
3219
              "flags=0x%x, retcode=%d",
3220 3221 3222 3223 3224 3225 3226 3227 3228 3229
              driver, conn, vm, NULLSTR(cookiein), cookieinlen,
              flags, retcode);

    virCheckFlags(QEMU_MIGRATION_FLAGS, -1);

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

3230 3231
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                       QEMU_MIGRATION_COOKIE_STATS)))
3232 3233
        goto cleanup;

3234 3235 3236
    if (retcode == 0)
        jobInfo = priv->job.completed;
    else
3237
        VIR_FREE(priv->job.completed);
3238 3239 3240

    /* Update times with the values sent by the destination daemon */
    if (mig->jobInfo && jobInfo) {
3241 3242 3243 3244 3245 3246 3247 3248
        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 &&
3249
            qemuMigrationFetchStats(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3250
                                    jobInfo, NULL) < 0)
3251 3252
            VIR_WARN("Could not refresh migration statistics");

3253 3254 3255
        qemuDomainJobInfoUpdateTime(jobInfo);
        jobInfo->timeDeltaSet = mig->jobInfo->timeDeltaSet;
        jobInfo->timeDelta = mig->jobInfo->timeDelta;
3256 3257
        jobInfo->stats.mig.downtime_set = mig->jobInfo->stats.mig.downtime_set;
        jobInfo->stats.mig.downtime = mig->jobInfo->stats.mig.downtime;
3258 3259
    }

3260 3261 3262
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

3263 3264
    /* 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.
3265 3266 3267 3268
     */
    if (retcode == 0) {
        /* If guest uses SPICE and supports seamless migration we have to hold
         * up domain shutdown until SPICE server transfers its data */
3269
        qemuMigrationWaitForSpice(vm);
3270 3271

        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
3272
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
3273 3274 3275
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
        virDomainAuditStop(vm, "migrated");

3276
        event = virDomainEventLifecycleNewFromObj(vm,
3277 3278
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
3279 3280
        qemuDomainEventQueue(driver, event);
        qemuDomainEventEmitJobCompleted(driver, vm);
3281
    } else {
3282
        virErrorPtr orig_err = virSaveLastError();
3283
        int reason;
3284 3285

        /* cancel any outstanding NBD jobs */
3286
        qemuMigrationCancelDriveMirror(driver, vm, false,
3287
                                       QEMU_ASYNC_JOB_MIGRATION_OUT, NULL);
3288 3289 3290

        virSetError(orig_err);
        virFreeError(orig_err);
3291

3292 3293 3294 3295
        if (virDomainObjGetState(vm, &reason) == VIR_DOMAIN_PAUSED &&
            reason == VIR_DOMAIN_PAUSED_POSTCOPY) {
            qemuMigrationPostcopyFailed(driver, vm);
        } else if (qemuMigrationRestoreDomainState(conn, vm)) {
3296 3297 3298
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
3299
            qemuDomainEventQueue(driver, event);
3300 3301
        }

3302
        qemuMigrationReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT);
3303

3304
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
3305 3306 3307
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
    }

3308
 done:
3309 3310 3311
    qemuMigrationCookieFree(mig);
    rv = 0;

3312
 cleanup:
3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325
    virObjectUnref(cfg);
    return rv;
}

int
qemuMigrationConfirm(virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *cookiein,
                     int cookieinlen,
                     unsigned int flags,
                     int cancelled)
{
    virQEMUDriverPtr driver = conn->privateData;
3326
    qemuMigrationJobPhase phase;
3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340
    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);
3341 3342
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
3343 3344 3345 3346 3347

    ret = qemuMigrationConfirmPhase(driver, conn, vm,
                                    cookiein, cookieinlen,
                                    flags, cancelled);

3348
    qemuMigrationJobFinish(driver, vm);
3349
    if (!virDomainObjIsActive(vm)) {
3350
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
3351
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
3352 3353
            vm->persistent = 0;
        }
3354
        qemuDomainRemoveInactiveJob(driver, vm);
3355 3356
    }

3357
 cleanup:
M
Michal Privoznik 已提交
3358
    virDomainObjEndAPI(&vm);
3359 3360 3361 3362 3363
    virObjectUnref(cfg);
    return ret;
}


3364 3365
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3366
    MIGRATION_DEST_CONNECT_HOST,
3367
    MIGRATION_DEST_FD,
3368
};
3369

3370 3371 3372 3373
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3374

3375 3376 3377 3378 3379 3380
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3381
            const char *protocol;
3382 3383 3384 3385
            const char *name;
            int port;
        } host;

3386 3387 3388 3389
        struct {
            int qemu;
            int local;
        } fd;
3390 3391 3392 3393 3394 3395 3396
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3397 3398 3399

#define TUNNEL_SEND_BUF_SIZE 65536

3400 3401 3402 3403 3404 3405 3406
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3407 3408
    int wakeupRecvFD;
    int wakeupSendFD;
3409 3410 3411
};

static void qemuMigrationIOFunc(void *arg)
3412
{
3413
    qemuMigrationIOThreadPtr data = arg;
3414 3415 3416 3417 3418 3419 3420
    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);
3421

3422
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3423
        goto abrt;
3424

3425 3426 3427
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3428
    for (;;) {
3429 3430 3431 3432 3433 3434 3435 3436 3437 3438
        int ret;

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

        ret = poll(fds, ARRAY_CARDINALITY(fds), timeout);

        if (ret < 0) {
            if (errno == EAGAIN || errno == EINTR)
                continue;
3439
            virReportSystemError(errno, "%s",
3440 3441
                                 _("poll failed in migration tunnel"));
            goto abrt;
3442
        }
3443 3444 3445 3446 3447 3448 3449

        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");
3450
            break;
3451
        }
3452

3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468
        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;
            }
3469 3470
        }

3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487
        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;
            }
        }
    }
3488

3489 3490
    if (virStreamFinish(data->st) < 0)
        goto error;
3491

3492
    VIR_FORCE_CLOSE(data->sock);
3493 3494
    VIR_FREE(buffer);

3495 3496
    return;

3497
 abrt:
3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3509
 error:
3510 3511 3512 3513 3514
    /* 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);
3515
    virResetLastError();
3516
    VIR_FREE(buffer);
3517 3518 3519 3520 3521 3522 3523
}


static qemuMigrationIOThreadPtr
qemuMigrationStartTunnel(virStreamPtr st,
                         int sock)
{
3524 3525
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3526

3527 3528 3529 3530
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3531 3532
    }

3533
    if (VIR_ALLOC(io) < 0)
3534
        goto error;
3535

3536 3537
    io->st = st;
    io->sock = sock;
3538 3539
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3540 3541 3542 3543 3544 3545

    if (virThreadCreate(&io->thread, true,
                        qemuMigrationIOFunc,
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3546
        goto error;
3547 3548 3549
    }

    return io;
3550

3551
 error:
3552 3553 3554 3555
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3556 3557 3558
}

static int
3559
qemuMigrationStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3560 3561
{
    int rv = -1;
3562 3563 3564 3565 3566 3567 3568 3569 3570
    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;
    }

3571 3572 3573 3574
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3575 3576 3577 3578
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3579 3580 3581 3582 3583 3584
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3585
 cleanup:
3586 3587
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3588 3589
    VIR_FREE(io);
    return rv;
3590 3591
}

3592
static int
3593
qemuMigrationConnect(virQEMUDriverPtr driver,
3594 3595 3596 3597 3598 3599 3600 3601 3602
                     virDomainObjPtr vm,
                     qemuMigrationSpecPtr spec)
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3603
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3604 3605 3606 3607 3608
        return -1;

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

3609
    if (qemuSecuritySetSocketLabel(driver->securityManager, vm->def) < 0)
3610
        goto cleanup;
3611 3612 3613
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
3614
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
3615
        virObjectUnref(sock);
3616
    }
3617
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0 ||
3618 3619 3620
        spec->dest.fd.qemu == -1)
        goto cleanup;

3621 3622 3623 3624 3625 3626 3627
    /* 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;
    }

3628 3629
    ret = 0;

3630
 cleanup:
3631 3632 3633 3634 3635 3636
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658

static int
qemuMigrationContinue(virQEMUDriverPtr driver,
                      virDomainObjPtr vm,
                      qemuMonitorMigrationStatus status,
                      qemuDomainAsyncJob asyncJob)
{
    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;
}


3659
static int
3660
qemuMigrationRun(virQEMUDriverPtr driver,
3661
                 virDomainObjPtr vm,
3662
                 const char *persist_xml,
3663 3664 3665 3666 3667 3668
                 const char *cookiein,
                 int cookieinlen,
                 char **cookieout,
                 int *cookieoutlen,
                 unsigned long flags,
                 unsigned long resource,
3669
                 qemuMigrationSpecPtr spec,
3670
                 virConnectPtr dconn,
3671 3672
                 const char *graphicsuri,
                 size_t nmigrate_disks,
3673
                 const char **migrate_disks,
3674 3675
                 qemuMigrationCompressionPtr compression,
                 qemuMonitorMigrationParamsPtr migParams)
3676
{
3677
    int ret = -1;
3678
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
3679
    virQEMUDriverConfigPtr cfg = NULL;
3680
    qemuDomainObjPrivatePtr priv = vm->privateData;
3681
    qemuMigrationCookiePtr mig = NULL;
3682 3683
    char *tlsAlias = NULL;
    char *secAlias = NULL;
3684
    qemuMigrationIOThreadPtr iothread = NULL;
3685
    int fd = -1;
3686
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3687
    virErrorPtr orig_err = NULL;
3688
    unsigned int cookieFlags = 0;
3689
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3690
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
3691
    bool cancel = false;
3692
    unsigned int waitFlags;
3693
    virDomainDefPtr persistDef = NULL;
3694
    char *timestamp;
3695
    int rc;
3696 3697

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
3698
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
3699 3700
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
3701 3702
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3703
              spec, spec->destType, spec->fwdType, dconn,
3704
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
3705

3706 3707 3708 3709 3710 3711 3712 3713 3714 3715
    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;
    }

3716 3717
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3718 3719 3720 3721
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3722 3723 3724
        return -1;
    }

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

3728 3729
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
        if (persist_xml) {
3730 3731
            if (!(persistDef = qemuMigrationPrepareDef(driver, persist_xml,
                                                       NULL, NULL)))
3732
                goto error;
3733 3734 3735
        } else {
            virDomainDefPtr def = vm->newDef ? vm->newDef : vm->def;
            if (!(persistDef = qemuDomainDefCopy(driver, def,
3736 3737
                                                 VIR_DOMAIN_XML_SECURE |
                                                 VIR_DOMAIN_XML_MIGRATABLE)))
3738
                goto error;
3739 3740 3741
        }
    }

3742 3743 3744
    mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                 cookieFlags | QEMU_MIGRATION_COOKIE_GRAPHICS);
    if (!mig)
3745
        goto error;
3746

3747
    if (qemuDomainMigrateGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3748 3749
        VIR_WARN("unable to provide data for graphics client relocation");

3750 3751 3752 3753 3754 3755 3756 3757
    if (flags & VIR_MIGRATE_TLS) {
        cfg = virQEMUDriverGetConfig(driver);

        /* Begin/CheckSetupTLS already set up migTLSAlias, the following
         * assumes that and adds the TLS objects to the domain. */
        if (qemuMigrationAddTLSObjects(driver, vm, cfg, false,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT,
                                       &tlsAlias, &secAlias, migParams) < 0)
3758
            goto error;
3759 3760 3761 3762 3763

        /* We need to add tls-hostname whenever QEMU itself does not
         * connect directly to the destination. */
        if (spec->destType == MIGRATION_DEST_CONNECT_HOST ||
            spec->destType == MIGRATION_DEST_FD) {
3764
            if (VIR_STRDUP(migParams->tlsHostname, spec->dest.host.name) < 0)
3765
                goto error;
3766 3767
        } else {
            /* Be sure there's nothing from a previous migration */
3768
            if (VIR_STRDUP(migParams->tlsHostname, "") < 0)
3769
                goto error;
3770 3771 3772 3773 3774
        }
    } else {
        if (qemuMigrationSetEmptyTLSParams(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT,
                                           migParams) < 0)
3775
            goto error;
3776 3777
    }

3778 3779 3780 3781 3782 3783 3784
    if (migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                         QEMU_MONITOR_MIGRATE_NON_SHARED_INC)) {
        if (mig->nbd) {
            /* This will update migrate_flags on success */
            if (qemuMigrationDriveMirror(driver, vm, mig,
                                         spec->dest.host.name,
                                         migrate_speed,
3785 3786
                                         &migrate_flags,
                                         nmigrate_disks,
3787 3788
                                         migrate_disks,
                                         dconn) < 0) {
3789
                goto error;
3790 3791 3792 3793 3794 3795 3796
            }
        } 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.");
        }
3797 3798
    }

3799
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3800 3801 3802
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
3803
            goto error;
3804 3805
    }

3806
    if (qemuMigrationSetCompression(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3807
                                    compression, migParams) < 0)
3808
        goto error;
3809

3810 3811 3812 3813
    if (qemuMigrationSetOption(driver, vm,
                               QEMU_MONITOR_MIGRATION_CAPS_AUTO_CONVERGE,
                               flags & VIR_MIGRATE_AUTO_CONVERGE,
                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3814
        goto error;
3815

3816 3817
    if (qemuMigrationSetOption(driver, vm,
                               QEMU_MONITOR_MIGRATION_CAPS_RDMA_PIN_ALL,
3818
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
3819
                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3820
        goto error;
3821

3822 3823 3824
    if (qemuMigrationSetPostCopy(driver, vm,
                                 flags & VIR_MIGRATE_POSTCOPY,
                                 QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3825
        goto error;
3826

3827 3828 3829 3830 3831 3832
    if (qemuMigrationCapsGet(vm, QEMU_MONITOR_MIGRATION_CAPS_PAUSE_BEFORE_SWITCHOVER) &&
        qemuMigrationSetOption(driver, vm,
                               QEMU_MONITOR_MIGRATION_CAPS_PAUSE_BEFORE_SWITCHOVER,
                               true, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto error;

3833
    if (qemuMigrationSetParams(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT,
3834
                               migParams) < 0)
3835
        goto error;
3836

3837 3838
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3839
        goto error;
3840

3841
    if (priv->job.abortJob) {
3842 3843
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
3844
         * priv->job.abortJob will not change */
3845
        priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_CANCELED;
3846 3847 3848
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
3849
        goto exit_monitor;
3850 3851
    }

3852 3853
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
3854

3855 3856
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3857
        qemuMigrationConnect(driver, vm, spec) < 0) {
3858
        goto exit_monitor;
3859
    }
3860

3861 3862 3863 3864 3865 3866
    /* log start of migration */
    if ((timestamp = virTimeStringNow()) != NULL) {
        qemuDomainLogAppendMessage(driver, vm, "%s: initiating migration\n", timestamp);
        VIR_FREE(timestamp);
    }

3867
    rc = -1;
3868 3869
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3870 3871
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
3872
            goto exit_monitor;
M
Michael R. Hines 已提交
3873
        }
3874 3875 3876 3877
        rc = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
                                      spec->dest.host.protocol,
                                      spec->dest.host.name,
                                      spec->dest.host.port);
3878 3879
        break;

3880 3881 3882 3883
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3884
    case MIGRATION_DEST_FD:
3885
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3886
            fd = spec->dest.fd.local;
3887 3888
            spec->dest.fd.local = -1;
        }
3889 3890
        rc = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                    spec->dest.fd.qemu);
3891 3892
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3893
    }
3894 3895

    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
3896
        goto error;
3897 3898 3899

    /* From this point onwards we *must* call cancel to abort the
     * migration on source if anything goes wrong */
3900
    cancel = true;
3901

3902 3903
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
        if (!(iothread = qemuMigrationStartTunnel(spec->fwd.stream, fd)))
3904
            goto error;
3905 3906 3907 3908 3909
        /* If we've created a tunnel, then the 'fd' will be closed in the
         * qemuMigrationIOFunc as data->sock.
         */
        fd = -1;
    }
3910

3911
    waitFlags = QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;
3912 3913 3914 3915 3916 3917 3918
    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;

3919 3920
    rc = qemuMigrationWaitForCompletion(driver, vm,
                                        QEMU_ASYNC_JOB_MIGRATION_OUT,
3921
                                        dconn, waitFlags);
3922
    if (rc == -2) {
3923
        goto error;
3924 3925 3926 3927 3928
    } else if (rc == -1) {
        /* QEMU reported failed migration, nothing to cancel anymore */
        cancel = false;
        goto error;
    }
3929

3930 3931 3932
    /* When migration completed, QEMU will have paused the CPUs for us.
     * Wait for the STOP event to be processed or explicitly stop CPUs
     * (for old QEMU which does not send events) to release the lock state.
3933
     */
3934 3935 3936 3937 3938
    if (priv->monJSON) {
        while (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            priv->signalStop = true;
            rc = virDomainObjWait(vm);
            priv->signalStop = false;
3939
            if (rc < 0)
3940
                goto error;
3941
        }
3942 3943
    } else if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING &&
               qemuMigrationSetOffline(driver, vm) < 0) {
3944
        goto error;
3945
    }
3946 3947 3948 3949 3950

    if (mig && mig->nbd &&
        qemuMigrationCancelDriveMirror(driver, vm, true,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT,
                                       dconn) < 0)
3951
        goto error;
3952

3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976
    /* 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) {
        if (qemuMigrationContinue(driver, vm,
                                  QEMU_MONITOR_MIGRATION_STATUS_PRE_SWITCHOVER,
                                  QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            goto error;

        waitFlags ^= QEMU_MIGRATION_COMPLETED_PRE_SWITCHOVER;

        rc = qemuMigrationWaitForCompletion(driver, vm,
                                            QEMU_ASYNC_JOB_MIGRATION_OUT,
                                            dconn, waitFlags);
        if (rc == -2) {
            goto error;
        } else if (rc == -1) {
            /* QEMU reported failed migration, nothing to cancel anymore */
            cancel = false;
            goto error;
        }
    }

3977 3978 3979 3980 3981
    if (iothread) {
        qemuMigrationIOThreadPtr io;

        VIR_STEAL_PTR(io, iothread);
        if (qemuMigrationStopTunnel(io, false) < 0)
3982
            goto error;
3983 3984 3985
    }

    if (priv->job.completed) {
3986
        priv->job.completed->stopped = priv->job.current->stopped;
3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999
        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 ||
        qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
        VIR_WARN("Unable to encode migration cookie");
    }
4000

4001
    ret = 0;
4002

4003
 cleanup:
4004 4005 4006
    VIR_FREE(tlsAlias);
    VIR_FREE(secAlias);
    virObjectUnref(cfg);
4007
    VIR_FORCE_CLOSE(fd);
4008
    virDomainDefFree(persistDef);
4009 4010
    qemuMigrationCookieFree(mig);

4011 4012 4013
    if (events)
        priv->signalIOError = false;

4014 4015 4016 4017 4018
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4019 4020
    return ret;

4021
 error:
4022 4023 4024 4025 4026 4027 4028 4029 4030 4031
    orig_err = virSaveLastError();

    if (cancel &&
        priv->job.current->status != QEMU_DOMAIN_JOB_STATUS_QEMU_COMPLETED &&
        virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        qemuMonitorMigrateCancel(priv->mon);
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
    }
4032 4033 4034 4035 4036 4037 4038 4039 4040 4041

    /* cancel any outstanding NBD jobs */
    if (mig && mig->nbd)
        qemuMigrationCancelDriveMirror(driver, vm, false,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT,
                                       dconn);

    if (iothread)
        qemuMigrationStopTunnel(iothread, true);

4042
    if (priv->job.current->status != QEMU_DOMAIN_JOB_STATUS_CANCELED)
4043 4044 4045 4046
        priv->job.current->status = QEMU_DOMAIN_JOB_STATUS_FAILED;

    goto cleanup;

4047 4048
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
4049
    goto error;
4050 4051
}

4052
/* Perform migration using QEMU's native migrate support,
4053 4054
 * not encrypted obviously
 */
4055
static int doNativeMigrate(virQEMUDriverPtr driver,
4056
                           virDomainObjPtr vm,
4057
                           const char *persist_xml,
4058 4059 4060 4061 4062 4063
                           const char *uri,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4064
                           unsigned long resource,
4065
                           virConnectPtr dconn,
4066 4067
                           const char *graphicsuri,
                           size_t nmigrate_disks,
4068
                           const char **migrate_disks,
4069 4070
                           qemuMigrationCompressionPtr compression,
                           qemuMonitorMigrationParamsPtr migParams)
4071
{
4072
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
4073
    virURIPtr uribits = NULL;
4074
    int ret = -1;
4075 4076 4077
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
4078
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
4079
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
4080
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
4081
              cookieout, cookieoutlen, flags, resource,
4082
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4083

4084
    if (!(uribits = qemuMigrationParseURI(uri, NULL)))
4085 4086
        return -1;

4087 4088 4089 4090 4091 4092 4093
    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
4094 4095 4096 4097 4098
    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"));
4099
            goto cleanup;
M
Michael R. Hines 已提交
4100
        }
4101
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
4102 4103 4104
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
4105
            goto cleanup;
M
Michael R. Hines 已提交
4106 4107 4108
        }
    }

4109
    if (STRNEQ(uribits->scheme, "rdma"))
4110 4111
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
4112
        spec.destType = MIGRATION_DEST_HOST;
4113
    spec.dest.host.protocol = uribits->scheme;
4114 4115 4116
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
4117

4118
    ret = qemuMigrationRun(driver, vm, persist_xml, cookiein, cookieinlen, cookieout,
4119
                           cookieoutlen, flags, resource, &spec, dconn,
4120
                           graphicsuri, nmigrate_disks, migrate_disks,
4121
                           compression, migParams);
4122 4123 4124 4125

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

4126
 cleanup:
4127
    virURIFree(uribits);
4128 4129 4130 4131 4132

    return ret;
}


4133
static int doTunnelMigrate(virQEMUDriverPtr driver,
4134 4135
                           virDomainObjPtr vm,
                           virStreamPtr st,
4136
                           const char *persist_xml,
4137 4138 4139 4140 4141
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4142
                           unsigned long resource,
4143
                           virConnectPtr dconn,
4144 4145
                           const char *graphicsuri,
                           size_t nmigrate_disks,
4146
                           const char **migrate_disks,
4147 4148
                           qemuMigrationCompressionPtr compression,
                           qemuMonitorMigrationParamsPtr migParams)
4149 4150 4151
{
    int ret = -1;
    qemuMigrationSpec spec;
4152
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4153
    int fds[2] = { -1, -1 };
4154 4155

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
4156
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, resource=%lu, "
4157
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
4158
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
4159
              cookieout, cookieoutlen, flags, resource,
4160
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4161 4162 4163 4164

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

4165

4166 4167 4168
    spec.destType = MIGRATION_DEST_FD;
    spec.dest.fd.qemu = -1;
    spec.dest.fd.local = -1;
4169

4170 4171 4172 4173 4174
    if (pipe2(fds, O_CLOEXEC) == 0) {
        spec.dest.fd.qemu = fds[1];
        spec.dest.fd.local = fds[0];
    }
    if (spec.dest.fd.qemu == -1 ||
4175 4176
        qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    spec.dest.fd.qemu) < 0) {
4177 4178 4179
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
        goto cleanup;
4180 4181
    }

4182 4183 4184
    ret = qemuMigrationRun(driver, vm, persist_xml, cookiein, cookieinlen,
                           cookieout, cookieoutlen, flags, resource, &spec,
                           dconn, graphicsuri, nmigrate_disks, migrate_disks,
4185
                           compression, migParams);
4186

4187
 cleanup:
4188 4189
    VIR_FORCE_CLOSE(spec.dest.fd.qemu);
    VIR_FORCE_CLOSE(spec.dest.fd.local);
4190

4191
    virObjectUnref(cfg);
4192 4193 4194 4195
    return ret;
}


4196 4197 4198 4199
/* 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 */
4200
static int doPeer2PeerMigrate2(virQEMUDriverPtr driver,
4201
                               virConnectPtr sconn ATTRIBUTE_UNUSED,
4202 4203
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
4204
                               const char *dconnuri,
4205 4206 4207
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
4208 4209 4210
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
4211
    char *cookie = NULL;
4212 4213 4214
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
4215
    bool cancelled;
4216
    virStreamPtr st = NULL;
4217
    unsigned long destflags;
4218
    qemuMigrationCompressionPtr compression = NULL;
4219
    qemuMonitorMigrationParams migParams = { 0 };
4220

4221
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
4222
              "flags=0x%lx, dname=%s, resource=%lu",
4223 4224
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
4225

4226 4227 4228 4229 4230
    /* 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,
4231 4232
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
4233 4234 4235 4236 4237
        return -1;

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

4238 4239
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4240

4241 4242 4243
    if (!(compression = qemuMigrationCompressionParse(NULL, 0, flags)))
        goto cleanup;

4244 4245 4246 4247 4248 4249 4250 4251 4252 4253
    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;

4254
        qemuDomainObjEnterRemote(vm);
4255
        ret = dconn->driver->domainMigratePrepareTunnel
4256
            (dconn, st, destflags, dname, resource, dom_xml);
4257
        qemuDomainObjExitRemote(vm);
4258
    } else {
4259
        qemuDomainObjEnterRemote(vm);
4260 4261
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
4262
             destflags, dname, resource, dom_xml);
4263
        qemuDomainObjExitRemote(vm);
4264 4265 4266
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
4267 4268 4269
        goto cleanup;

    /* the domain may have shutdown or crashed while we had the locks dropped
4270
     * in qemuDomainObjEnterRemote, so check again
4271 4272
     */
    if (!virDomainObjIsActive(vm)) {
4273 4274
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4275 4276 4277
        goto cleanup;
    }

4278 4279
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
4280 4281
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
4282
        cancelled = true;
4283
        orig_err = virSaveLastError();
4284
        goto finish;
4285 4286
    }

4287 4288 4289 4290
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
4291
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4292
    if (flags & VIR_MIGRATE_TUNNELLED)
4293
        ret = doTunnelMigrate(driver, vm, st, NULL,
4294
                              NULL, 0, NULL, NULL,
4295
                              flags, resource, dconn,
4296
                              NULL, 0, NULL, compression, &migParams);
4297
    else
4298
        ret = doNativeMigrate(driver, vm, NULL, uri_out,
4299 4300
                              cookie, cookielen,
                              NULL, NULL, /* No out cookie with v2 migration */
4301 4302
                              flags, resource, dconn, NULL, 0, NULL,
                              compression, &migParams);
4303 4304 4305 4306

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4307

4308 4309 4310
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4311
    cancelled = ret < 0;
4312

4313
 finish:
4314 4315 4316 4317
    /* 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.
     */
4318
    dname = dname ? dname : vm->def->name;
4319
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4320
    qemuDomainObjEnterRemote(vm);
4321
    ddomain = dconn->driver->domainMigrateFinish2
4322
        (dconn, dname, cookie, cookielen,
4323
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4324
    qemuDomainObjExitRemote(vm);
4325 4326
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4327

4328
 cleanup:
4329
    if (ddomain) {
4330
        virObjectUnref(ddomain);
4331 4332 4333 4334
        ret = 0;
    } else {
        ret = -1;
    }
4335

4336
    virObjectUnref(st);
4337 4338 4339 4340 4341

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4342
    qemuMigrationParamsClear(&migParams);
4343
    VIR_FREE(uri_out);
4344
    VIR_FREE(cookie);
4345
    VIR_FREE(compression);
4346 4347

    return ret;
4348 4349 4350
}


4351 4352 4353 4354
/* 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 */
4355 4356 4357 4358 4359 4360 4361
static int
doPeer2PeerMigrate3(virQEMUDriverPtr driver,
                    virConnectPtr sconn,
                    virConnectPtr dconn,
                    const char *dconnuri,
                    virDomainObjPtr vm,
                    const char *xmlin,
4362
                    const char *persist_xml,
4363 4364
                    const char *dname,
                    const char *uri,
4365
                    const char *graphicsuri,
4366
                    const char *listenAddress,
4367 4368
                    size_t nmigrate_disks,
                    const char **migrate_disks,
4369
                    int nbdPort,
4370
                    qemuMigrationCompressionPtr compression,
4371
                    qemuMonitorMigrationParamsPtr migParams,
4372 4373 4374
                    unsigned long long bandwidth,
                    bool useParams,
                    unsigned long flags)
4375 4376 4377 4378 4379 4380 4381 4382 4383 4384
{
    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;
4385
    bool cancelled = true;
4386
    virStreamPtr st = NULL;
4387
    unsigned long destflags;
4388 4389 4390
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4391
    size_t i;
4392 4393

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4394
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4395
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
4396
              "bandwidth=%llu, useParams=%d, flags=0x%lx",
4397
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4398
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4399
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks, nbdPort,
4400
              bandwidth, useParams, flags);
4401

4402 4403 4404 4405 4406
    /* 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.  */

4407
    dom_xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
4408 4409
                                      &cookieout, &cookieoutlen,
                                      nmigrate_disks, migrate_disks, flags);
4410 4411 4412
    if (!dom_xml)
        goto cleanup;

4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432
    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;
4433 4434 4435 4436 4437 4438

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4439 4440 4441 4442 4443
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4444 4445 4446 4447 4448
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4449 4450 4451 4452 4453
        if (nbdPort &&
            virTypedParamsAddInt(&params, &nparams, &maxparams,
                                 VIR_MIGRATE_PARAM_DISKS_PORT,
                                 nbdPort) < 0)
            goto cleanup;
4454 4455 4456 4457

        if (qemuMigrationCompressionDump(compression, &params, &nparams,
                                         &maxparams, &flags) < 0)
            goto cleanup;
4458 4459
    }

4460 4461 4462
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4463 4464
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4465

4466 4467 4468 4469 4470 4471 4472 4473 4474
    VIR_DEBUG("Prepare3 %p", dconn);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
    if (flags & VIR_MIGRATE_TUNNELLED) {
        if (!(st = virStreamNew(dconn, 0)))
            goto cleanup;

4475
        qemuDomainObjEnterRemote(vm);
4476 4477 4478 4479 4480 4481 4482 4483 4484
        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);
        }
4485
        qemuDomainObjExitRemote(vm);
4486
    } else {
4487
        qemuDomainObjEnterRemote(vm);
4488 4489 4490 4491 4492 4493 4494 4495 4496
        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);
        }
4497
        qemuDomainObjExitRemote(vm);
4498 4499 4500 4501 4502
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

L
liguang 已提交
4503 4504 4505 4506
    if (flags & VIR_MIGRATE_OFFLINE) {
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4507
        cancelled = false;
L
liguang 已提交
4508 4509 4510
        goto finish;
    }

4511 4512 4513 4514
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4515 4516
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4517
            goto finish;
4518
        }
4519
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4520 4521
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4522
        orig_err = virSaveLastError();
4523 4524 4525 4526 4527 4528 4529 4530
        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.
     */
4531
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4532
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4533 4534 4535 4536 4537
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4538
    if (flags & VIR_MIGRATE_TUNNELLED) {
4539
        ret = doTunnelMigrate(driver, vm, st, persist_xml,
4540 4541
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4542
                              flags, bandwidth, dconn, graphicsuri,
4543 4544
                              nmigrate_disks, migrate_disks, compression,
                              migParams);
4545
    } else {
4546
        ret = doNativeMigrate(driver, vm, persist_xml, uri,
4547 4548
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4549
                              flags, bandwidth, dconn, graphicsuri,
4550 4551
                              nmigrate_disks, migrate_disks, compression,
                              migParams);
4552
    }
4553 4554

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4555
    if (ret < 0) {
4556
        orig_err = virSaveLastError();
4557 4558 4559 4560
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4561 4562 4563 4564

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4565
    cancelled = ret < 0;
4566

4567
 finish:
4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579
    /*
     * The status code from the source is passed to the destination.
     * The dest can cleanup in the source indicated it failed to
     * send all migration data. Returns NULL for ddomain if
     * the dest was unable to complete migration.
     */
    VIR_DEBUG("Finish3 %p ret=%d", dconn, ret);
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602

    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);
            qemuDomainObjExitRemote(vm);
        }
    } else {
        dname = dname ? dname : vm->def->name;
        qemuDomainObjEnterRemote(vm);
        ddomain = dconn->driver->domainMigrateFinish3
            (dconn, dname, cookiein, cookieinlen, &cookieout, &cookieoutlen,
             dconnuri, uri, destflags, cancelled);
        qemuDomainObjExitRemote(vm);
    }
4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622

    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();
4623 4624
                if (err &&
                    err->domain == VIR_FROM_QEMU &&
4625 4626 4627 4628 4629 4630 4631
                    err->code != VIR_ERR_MIGRATE_FINISH_OK) {
                    virFreeError(orig_err);
                    orig_err = NULL;
                }
            }
        }
    }
4632

4633 4634 4635 4636 4637 4638 4639
    /* 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.
4640
     */
4641
    cancelled = ddomain == NULL;
4642

4643 4644 4645 4646 4647 4648
    /* If finish3 set an error, and we don't have an earlier
     * one we need to preserve it in case confirm3 overwrites
     */
    if (!orig_err)
        orig_err = virSaveLastError();

4649 4650 4651 4652
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4653
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4654 4655 4656 4657 4658
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4659 4660 4661
    ret = qemuMigrationConfirmPhase(driver, sconn, vm,
                                    cookiein, cookieinlen,
                                    flags, cancelled);
4662 4663 4664 4665
    /* 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.
     */
4666 4667 4668
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4669 4670 4671

 cleanup:
    if (ddomain) {
4672
        virObjectUnref(ddomain);
4673 4674 4675 4676 4677
        ret = 0;
    } else {
        ret = -1;
    }

4678
    virObjectUnref(st);
4679 4680 4681 4682 4683 4684 4685 4686

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4687
    virTypedParamsFree(params, nparams);
4688 4689 4690 4691
    return ret;
}


4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703
static void
qemuMigrationConnectionClosed(virConnectPtr conn,
                              int reason,
                              void *opaque)
{
    virDomainObjPtr vm = opaque;

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


4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4716
static int doPeer2PeerMigrate(virQEMUDriverPtr driver,
4717
                              virConnectPtr sconn,
4718
                              virDomainObjPtr vm,
4719
                              const char *xmlin,
4720
                              const char *persist_xml,
4721
                              const char *dconnuri,
4722
                              const char *uri,
4723
                              const char *graphicsuri,
4724
                              const char *listenAddress,
4725 4726
                              size_t nmigrate_disks,
                              const char **migrate_disks,
4727
                              int nbdPort,
4728
                              qemuMigrationCompressionPtr compression,
4729
                              qemuMonitorMigrationParamsPtr migParams,
4730 4731
                              unsigned long flags,
                              const char *dname,
4732 4733
                              unsigned long resource,
                              bool *v3proto)
4734 4735 4736 4737
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4738
    virErrorPtr orig_err = NULL;
4739
    bool offline = false;
4740
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4741
    bool useParams;
4742

4743 4744
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, uri=%s, "
              "graphicsuri=%s, listenAddress=%s, nmigrate_disks=%zu, "
4745
              "migrate_disks=%p, nbdPort=%d, flags=0x%lx, dname=%s, "
4746
              "resource=%lu",
4747
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4748
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4749 4750
              nmigrate_disks, migrate_disks, nbdPort, flags, NULLSTR(dname),
              resource);
4751

4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765
    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;
    }

4766 4767 4768 4769 4770 4771 4772
    if (flags & VIR_MIGRATE_TUNNELLED && nbdPort) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("disk port address is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

4773 4774 4775 4776
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

4777
    qemuDomainObjEnterRemote(vm);
4778
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4779
    qemuDomainObjExitRemote(vm);
4780
    if (dconn == NULL) {
4781
        virReportError(VIR_ERR_OPERATION_FAILED,
4782 4783
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4784
        virObjectUnref(cfg);
4785 4786 4787
        return -1;
    }

4788 4789
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4790 4791
        goto cleanup;

4792 4793 4794 4795 4796
    if (virConnectRegisterCloseCallback(dconn, qemuMigrationConnectionClosed,
                                        vm, NULL) < 0) {
        goto cleanup;
    }

4797
    qemuDomainObjEnterRemote(vm);
4798 4799
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4800
        /* v3proto reflects whether the caller used Perform3, but with
4801
         * p2p migrate, regardless of whether Perform2 or Perform3
4802 4803 4804 4805
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4806 4807
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
L
liguang 已提交
4808 4809 4810
    if (flags & VIR_MIGRATE_OFFLINE)
        offline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                           VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4811
    qemuDomainObjExitRemote(vm);
4812

4813
    if (!p2p) {
4814 4815
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4816 4817 4818
        goto cleanup;
    }

4819 4820
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4821
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
4822 4823 4824 4825 4826 4827
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("Migration APIs with extensible parameters are not "
                         "supported but extended parameters were passed"));
        goto cleanup;
    }

L
liguang 已提交
4828 4829 4830 4831 4832 4833 4834
    if (flags & VIR_MIGRATE_OFFLINE && !offline) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4835
    /* domain may have been stopped while we were talking to remote daemon */
L
liguang 已提交
4836
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4837 4838
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4839 4840 4841
        goto cleanup;
    }

4842 4843 4844 4845 4846 4847 4848
    /* 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;

4849 4850
    if (*v3proto) {
        ret = doPeer2PeerMigrate3(driver, sconn, dconn, dconnuri, vm, xmlin,
4851 4852
                                  persist_xml, dname, uri, graphicsuri,
                                  listenAddress, nmigrate_disks, migrate_disks,
4853 4854
                                  nbdPort, compression, migParams, resource,
                                  useParams, flags);
4855
    } else {
4856
        ret = doPeer2PeerMigrate2(driver, sconn, dconn, vm,
4857
                                  dconnuri, flags, dname, resource);
4858
    }
4859

4860
 cleanup:
4861
    orig_err = virSaveLastError();
4862
    qemuDomainObjEnterRemote(vm);
4863
    virConnectUnregisterCloseCallback(dconn, qemuMigrationConnectionClosed);
4864
    virObjectUnref(dconn);
4865
    qemuDomainObjExitRemote(vm);
4866 4867 4868 4869
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4870
    virObjectUnref(cfg);
4871 4872 4873 4874
    return ret;
}


4875 4876 4877 4878 4879 4880
/*
 * 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
4881
qemuMigrationPerformJob(virQEMUDriverPtr driver,
4882 4883 4884
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
4885
                        const char *persist_xml,
4886 4887
                        const char *dconnuri,
                        const char *uri,
4888
                        const char *graphicsuri,
4889
                        const char *listenAddress,
4890 4891
                        size_t nmigrate_disks,
                        const char **migrate_disks,
4892
                        int nbdPort,
4893
                        qemuMigrationCompressionPtr compression,
4894
                        qemuMonitorMigrationParamsPtr migParams,
4895
                        const char *cookiein,
4896 4897 4898 4899 4900 4901 4902
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4903
{
4904
    virObjectEventPtr event = NULL;
4905
    int ret = -1;
4906
    virErrorPtr orig_err = NULL;
4907
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4908

4909
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4910 4911
        goto cleanup;

L
liguang 已提交
4912
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4913 4914
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
4915 4916 4917
        goto endjob;
    }

4918
    if (!qemuMigrationIsAllowed(driver, vm, true, flags))
4919
        goto endjob;
4920

4921
    if (!(flags & (VIR_MIGRATE_UNSAFE | VIR_MIGRATE_OFFLINE)) &&
4922
        !qemuMigrationIsSafe(vm->def, nmigrate_disks, migrate_disks, flags))
4923
        goto endjob;
4924

4925
    qemuMigrationStoreDomainState(vm);
4926 4927

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4928
        ret = doPeer2PeerMigrate(driver, conn, vm, xmlin, persist_xml,
4929
                                 dconnuri, uri, graphicsuri, listenAddress,
4930
                                 nmigrate_disks, migrate_disks, nbdPort,
4931 4932
                                 compression, migParams, flags, dname, resource,
                                 &v3proto);
4933
    } else {
4934
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4935
        ret = doNativeMigrate(driver, vm, persist_xml, uri, cookiein, cookieinlen,
4936
                              cookieout, cookieoutlen,
4937
                              flags, resource, NULL, NULL, 0, NULL,
4938
                              compression, migParams);
4939
    }
4940 4941
    if (ret < 0)
        goto endjob;
4942

4943 4944 4945 4946
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4947
    if (!v3proto) {
4948
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
4949
                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4950
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4951
        virDomainAuditStop(vm, "migrated");
4952
        event = virDomainEventLifecycleNewFromObj(vm,
4953 4954
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4955 4956
    }

4957
 endjob:
4958 4959 4960
    if (ret < 0)
        orig_err = virSaveLastError();

4961 4962 4963 4964
    /* v2 proto has no confirm phase so we need to reset migration parameters
     * here
     */
    if (!v3proto && ret < 0)
4965
        qemuMigrationReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT);
4966

4967
    if (qemuMigrationRestoreDomainState(conn, vm)) {
4968
        event = virDomainEventLifecycleNewFromObj(vm,
4969 4970 4971
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
4972

4973
    qemuMigrationJobFinish(driver, vm);
4974
    if (!virDomainObjIsActive(vm) && ret == 0) {
4975
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE) {
4976
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4977 4978
            vm->persistent = 0;
        }
4979
        qemuDomainRemoveInactiveJob(driver, vm);
4980 4981
    }

4982 4983 4984 4985 4986
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4987
 cleanup:
M
Michal Privoznik 已提交
4988
    virDomainObjEndAPI(&vm);
4989
    qemuDomainEventQueue(driver, event);
4990
    virObjectUnref(cfg);
4991 4992 4993 4994 4995 4996 4997
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4998
qemuMigrationPerformPhase(virQEMUDriverPtr driver,
4999 5000
                          virConnectPtr conn,
                          virDomainObjPtr vm,
5001
                          const char *persist_xml,
5002
                          const char *uri,
5003
                          const char *graphicsuri,
5004 5005
                          size_t nmigrate_disks,
                          const char **migrate_disks,
5006
                          qemuMigrationCompressionPtr compression,
5007
                          qemuMonitorMigrationParamsPtr migParams,
5008 5009 5010 5011 5012 5013 5014
                          const char *cookiein,
                          int cookieinlen,
                          char **cookieout,
                          int *cookieoutlen,
                          unsigned long flags,
                          unsigned long resource)
{
5015
    virObjectEventPtr event = NULL;
5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026
    int ret = -1;

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

    qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
5027 5028
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
5029

5030
    ret = doNativeMigrate(driver, vm, persist_xml, uri, cookiein, cookieinlen,
5031
                          cookieout, cookieoutlen,
5032
                          flags, resource, NULL, graphicsuri,
5033
                          nmigrate_disks, migrate_disks, compression, migParams);
5034

5035 5036 5037 5038 5039
    if (ret < 0) {
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
5040
        }
5041
        goto endjob;
5042
    }
5043 5044 5045

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

5046 5047
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                             qemuMigrationCleanup) < 0)
5048 5049
        goto endjob;

5050
 endjob:
5051 5052
    if (ret < 0) {
        qemuMigrationReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT);
5053
        qemuMigrationJobFinish(driver, vm);
5054
    } else {
5055
        qemuMigrationJobContinue(vm);
5056 5057
    }

5058
    if (!virDomainObjIsActive(vm))
5059
        qemuDomainRemoveInactiveJob(driver, vm);
5060

5061
 cleanup:
M
Michal Privoznik 已提交
5062
    virDomainObjEndAPI(&vm);
5063
    qemuDomainEventQueue(driver, event);
5064 5065 5066
    return ret;
}

5067
int
5068
qemuMigrationPerform(virQEMUDriverPtr driver,
5069 5070 5071
                     virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *xmlin,
5072
                     const char *persist_xml,
5073 5074
                     const char *dconnuri,
                     const char *uri,
5075
                     const char *graphicsuri,
5076
                     const char *listenAddress,
5077 5078
                     size_t nmigrate_disks,
                     const char **migrate_disks,
5079
                     int nbdPort,
5080
                     qemuMigrationCompressionPtr compression,
5081
                     qemuMonitorMigrationParamsPtr migParams,
5082 5083 5084 5085 5086 5087 5088 5089 5090 5091
                     const char *cookiein,
                     int cookieinlen,
                     char **cookieout,
                     int *cookieoutlen,
                     unsigned long flags,
                     const char *dname,
                     unsigned long resource,
                     bool v3proto)
{
    VIR_DEBUG("driver=%p, conn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
5092
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
5093
              "nmigrate_disks=%zu, migrate_disks=%p, nbdPort=%d, "
5094
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
5095
              "flags=0x%lx, dname=%s, resource=%lu, v3proto=%d",
5096
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
5097
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
5098 5099 5100
              nmigrate_disks, migrate_disks, nbdPort,
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
5101 5102 5103

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
5104 5105
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
5106 5107 5108
            return -1;
        }

5109
        return qemuMigrationPerformJob(driver, conn, vm, xmlin, persist_xml, dconnuri, uri,
5110
                                       graphicsuri, listenAddress,
5111
                                       nmigrate_disks, migrate_disks, nbdPort,
5112 5113
                                       compression, migParams,
                                       cookiein, cookieinlen,
5114 5115
                                       cookieout, cookieoutlen,
                                       flags, dname, resource, v3proto);
5116 5117
    } else {
        if (dconnuri) {
5118 5119
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
5120 5121 5122 5123
            return -1;
        }

        if (v3proto) {
5124
            return qemuMigrationPerformPhase(driver, conn, vm, persist_xml, uri,
5125
                                             graphicsuri,
5126
                                             nmigrate_disks, migrate_disks,
5127 5128
                                             compression, migParams,
                                             cookiein, cookieinlen,
5129
                                             cookieout, cookieoutlen,
5130
                                             flags, resource);
5131
        } else {
5132
            return qemuMigrationPerformJob(driver, conn, vm, xmlin, persist_xml, NULL,
5133
                                           uri, graphicsuri, listenAddress,
5134
                                           nmigrate_disks, migrate_disks, nbdPort,
5135 5136
                                           compression, migParams,
                                           cookiein, cookieinlen,
5137 5138 5139 5140 5141
                                           cookieout, cookieoutlen, flags,
                                           dname, resource, v3proto);
        }
    }
}
5142

5143
static int
5144 5145
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def)
{
5146
    size_t i;
5147 5148 5149 5150 5151
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
5152
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5153
            if (virNetDevVPortProfileAssociate(net->ifname,
5154
                                               virDomainNetGetActualVirtPortProfile(net),
5155
                                               &net->mac,
5156
                                               virDomainNetGetActualDirectDev(net),
5157
                                               -1,
5158
                                               def->uuid,
5159 5160
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
5161 5162 5163
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
5164
                goto err_exit;
5165
            }
5166
            last_good_net = i;
5167
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
5168

5169
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
5170 5171 5172 5173
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
5174 5175 5176
        }
    }

5177
    return 0;
5178

5179
 err_exit:
5180
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
5181
        net = def->nets[i];
5182
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5183
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
5184
                                                           virDomainNetGetActualVirtPortProfile(net),
5185
                                                           &net->mac,
5186
                                                           virDomainNetGetActualDirectDev(net),
5187
                                                           -1,
5188
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
5189 5190
        }
    }
5191
    return -1;
5192 5193 5194
}


J
Jiri Denemark 已提交
5195 5196 5197
static int
qemuMigrationPersist(virQEMUDriverPtr driver,
                     virDomainObjPtr vm,
5198 5199
                     qemuMigrationCookiePtr mig,
                     bool ignoreSaveError)
J
Jiri Denemark 已提交
5200 5201 5202 5203
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    virDomainDefPtr vmdef;
5204 5205
    virDomainDefPtr oldDef = NULL;
    unsigned int oldPersist = vm->persistent;
J
Jiri Denemark 已提交
5206 5207 5208 5209 5210 5211 5212
    virObjectEventPtr event;
    int ret = -1;

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

    vm->persistent = 1;
5213 5214
    oldDef = vm->newDef;
    vm->newDef = qemuMigrationCookieGetPersistent(mig);
J
Jiri Denemark 已提交
5215

5216 5217
    if (!(vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm)))
        goto error;
J
Jiri Denemark 已提交
5218

5219 5220
    if (virDomainSaveConfig(cfg->configDir, driver->caps, vmdef) < 0 &&
        !ignoreSaveError)
5221
        goto error;
J
Jiri Denemark 已提交
5222 5223 5224

    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_DEFINED,
5225 5226 5227
                                              oldPersist ?
                                              VIR_DOMAIN_EVENT_DEFINED_UPDATED :
                                              VIR_DOMAIN_EVENT_DEFINED_ADDED);
5228
    qemuDomainEventQueue(driver, event);
J
Jiri Denemark 已提交
5229 5230 5231 5232

    ret = 0;

 cleanup:
5233
    virDomainDefFree(oldDef);
J
Jiri Denemark 已提交
5234 5235 5236
    virObjectUnref(caps);
    virObjectUnref(cfg);
    return ret;
5237 5238 5239 5240 5241 5242 5243

 error:
    virDomainDefFree(vm->newDef);
    vm->persistent = oldPersist;
    vm->newDef = oldDef;
    oldDef = NULL;
    goto cleanup;
J
Jiri Denemark 已提交
5244 5245 5246
}


5247
virDomainPtr
5248
qemuMigrationFinish(virQEMUDriverPtr driver,
5249 5250
                    virConnectPtr dconn,
                    virDomainObjPtr vm,
5251 5252 5253 5254
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
5255
                    unsigned long flags,
5256 5257
                    int retcode,
                    bool v3proto)
5258 5259
{
    virDomainPtr dom = NULL;
5260
    qemuMigrationCookiePtr mig = NULL;
5261
    virErrorPtr orig_err = NULL;
5262
    int cookie_flags = 0;
J
Jiri Denemark 已提交
5263
    qemuDomainObjPrivatePtr priv = vm->privateData;
5264
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5265
    unsigned short port;
5266 5267
    unsigned long long timeReceived = 0;
    virObjectEventPtr event;
5268
    qemuDomainJobInfoPtr jobInfo = NULL;
5269
    bool inPostCopy = false;
5270
    bool doKill = true;
5271

5272
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
5273
              "cookieout=%p, cookieoutlen=%p, flags=0x%lx, retcode=%d",
5274 5275
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
5276

5277 5278 5279
    port = priv->migrationPort;
    priv->migrationPort = 0;

5280 5281
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN)) {
        qemuMigrationErrorReport(driver, vm->def->name);
5282
        goto cleanup;
5283
    }
5284

5285 5286
    ignore_value(virTimeMillisNow(&timeReceived));

5287 5288 5289
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
5290

5291
    qemuDomainCleanupRemove(vm, qemuMigrationPrepareCleanup);
5292
    VIR_FREE(priv->job.completed);
5293

5294
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
5295 5296
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
5297 5298 5299
    /* 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;
5300 5301 5302

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein,
                                       cookieinlen, cookie_flags)))
5303
        goto endjob;
5304

5305
    if (flags & VIR_MIGRATE_OFFLINE) {
5306 5307
        if (retcode == 0 &&
            qemuMigrationPersist(driver, vm, mig, false) == 0)
5308
            dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, -1);
5309 5310
        goto endjob;
    }
5311

5312 5313 5314 5315
    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
         */
5316
        qemuDomainCheckMonitor(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN);
5317 5318
        goto endjob;
    }
5319

5320 5321 5322 5323 5324 5325
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        qemuMigrationErrorReport(driver, vm->def->name);
        goto endjob;
    }
5326

5327 5328
    if (qemuMigrationVPAssociatePortProfiles(vm->def) < 0)
        goto endjob;
5329

5330 5331
    if (mig->network && qemuDomainMigrateOPDRelocate(driver, vm, mig) < 0)
        VIR_WARN("unable to provide network data for relocation");
5332

5333 5334
    if (qemuMigrationStopNBDServer(driver, vm, mig) < 0)
        goto endjob;
5335

5336 5337
    if (qemuRefreshVirtioChannelState(driver, vm,
                                      QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
5338 5339
        goto endjob;

5340 5341
    if (qemuConnectAgent(driver, vm) < 0)
        goto endjob;
5342

5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355
    if (flags & VIR_MIGRATE_PERSIST_DEST) {
        if (qemuMigrationPersist(driver, vm, mig, !v3proto) < 0) {
            /* 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.
5356 5357 5358
             */
            if (v3proto)
                goto endjob;
5359
        }
5360
    }
5361

5362 5363 5364 5365
    /* We need to wait for QEMU to process all data sent by the source
     * before starting guest CPUs.
     */
    if (qemuMigrationWaitForDestCompletion(driver, vm,
5366 5367
                                           QEMU_ASYNC_JOB_MIGRATION_IN,
                                           !!(flags & VIR_MIGRATE_POSTCOPY)) < 0) {
5368 5369 5370 5371 5372 5373
        /* 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;
    }
5374

5375 5376 5377 5378 5379 5380 5381 5382 5383
    /* 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;
    }

5384
    if (priv->job.current->status == QEMU_DOMAIN_JOB_STATUS_POSTCOPY)
5385 5386
        inPostCopy = true;

5387 5388 5389 5390 5391
    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
         */
5392
        if (qemuProcessStartCPUs(driver, vm,
5393 5394
                                 inPostCopy ? VIR_DOMAIN_RUNNING_POSTCOPY
                                            : VIR_DOMAIN_RUNNING_MIGRATED,
5395 5396 5397 5398 5399 5400 5401 5402
                                 QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
            if (virGetLastError() == NULL)
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("resume operation failed"));
            /* Need to save the current error, in case shutting
             * down the process overwrites it
             */
            orig_err = virSaveLastError();
5403

5404 5405 5406 5407 5408 5409 5410 5411 5412 5413
            /*
             * 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;
5414
        }
5415 5416

        if (inPostCopy) {
5417
            doKill = false;
5418 5419 5420 5421 5422
            event = virDomainEventLifecycleNewFromObj(vm,
                                        VIR_DOMAIN_EVENT_RESUMED,
                                        VIR_DOMAIN_EVENT_RESUMED_POSTCOPY);
            qemuDomainEventQueue(driver, event);
        }
5423
    }
5424

5425
    if (mig->jobInfo) {
5426
        jobInfo = mig->jobInfo;
5427 5428 5429 5430 5431 5432
        mig->jobInfo = NULL;

        if (jobInfo->sent && timeReceived) {
            jobInfo->timeDelta = timeReceived - jobInfo->sent;
            jobInfo->received = timeReceived;
            jobInfo->timeDeltaSet = true;
5433
        }
5434 5435
        qemuDomainJobInfoUpdateTime(jobInfo);
        qemuDomainJobInfoUpdateDowntime(jobInfo);
5436
    }
L
liguang 已提交
5437

5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450
    if (inPostCopy) {
        if (qemuMigrationWaitForDestCompletion(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_IN,
                                               false) < 0) {
            goto endjob;
        }
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
            virDomainObjSetState(vm,
                                 VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_MIGRATED);
        }
    }

5451
    dom = virGetDomain(dconn, vm->def->name, vm->def->uuid, vm->def->id);
5452

5453 5454 5455 5456
    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_RESUMED,
                                              VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    qemuDomainEventQueue(driver, event);
5457

5458 5459 5460 5461 5462 5463
    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);
        qemuDomainEventQueue(driver, event);
5464
    }
5465

5466
    if (virDomainObjIsActive(vm) &&
5467
        virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
5468 5469 5470 5471 5472
        VIR_WARN("Failed to save status on vm %s", vm->def->name);

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

5473
 endjob:
5474
    if (!dom &&
5475 5476
        !(flags & VIR_MIGRATE_OFFLINE) &&
        virDomainObjIsActive(vm)) {
5477
        if (doKill) {
5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488
            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);
            qemuDomainEventQueue(driver, event);
        } else {
            qemuMigrationPostcopyFailed(driver, vm);
        }
5489 5490
    }

5491
    if (dom) {
5492 5493 5494 5495
        if (jobInfo) {
            VIR_STEAL_PTR(priv->job.completed, jobInfo);
            priv->job.completed->status = QEMU_DOMAIN_JOB_STATUS_COMPLETED;
        }
5496

5497 5498 5499
        if (qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen,
                                    QEMU_MIGRATION_COOKIE_STATS) < 0)
            VIR_WARN("Unable to encode migration cookie");
5500 5501 5502 5503 5504 5505

        /* Remove completed stats for post-copy, everything but timing fields
         * is obsolete anyway.
         */
        if (inPostCopy)
            VIR_FREE(priv->job.completed);
5506
    }
5507

5508
    qemuMigrationReset(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN);
5509

5510
    qemuMigrationJobFinish(driver, vm);
5511
    if (!virDomainObjIsActive(vm))
5512
        qemuDomainRemoveInactiveJob(driver, vm);
5513

5514
 cleanup:
5515
    VIR_FREE(jobInfo);
5516
    virPortAllocatorRelease(driver->migrationPorts, port);
5517
    if (priv->mon)
5518
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
5519
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5520
    virDomainObjEndAPI(&vm);
5521
    qemuMigrationCookieFree(mig);
5522 5523 5524 5525
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5526
    virObjectUnref(cfg);
5527 5528 5529 5530 5531 5532

    /* Set a special error if Finish is expected to return NULL as a result of
     * successful call with retcode != 0
     */
    if (retcode != 0 && !dom && !virGetLastError())
        virReportError(VIR_ERR_MIGRATE_FINISH_OK, NULL);
5533 5534
    return dom;
}
5535

5536

5537
/* Helper function called while vm is active.  */
5538
int
5539
qemuMigrationToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
5540
                    int fd,
5541
                    const char *compressor,
5542
                    qemuDomainAsyncJob asyncJob)
5543 5544 5545
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5546
    int ret = -1;
5547 5548
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5549
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5550
    char *errbuf = NULL;
5551
    virErrorPtr orig_err = NULL;
5552 5553 5554 5555 5556

    /* 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,
5557 5558
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5559 5560
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5561
    }
5562

5563 5564 5565 5566 5567 5568 5569
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5570 5571 5572 5573
    if (compressor && pipe(pipeFD) < 0) {
        virReportSystemError(errno, "%s",
                             _("Failed to create pipe for migration"));
        return -1;
5574 5575
    }

5576 5577 5578 5579
    /* 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.  */
5580 5581
    if (qemuSecuritySetImageFDLabel(driver->securityManager, vm->def,
                                    compressor ? pipeFD[1] : fd) < 0)
5582 5583
        goto cleanup;

5584
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5585 5586
        goto cleanup;

5587
    if (!compressor) {
5588 5589 5590
        rc = qemuMonitorMigrateToFd(priv->mon,
                                    QEMU_MONITOR_MIGRATE_BACKGROUND,
                                    fd);
5591 5592 5593 5594 5595 5596 5597
    } else {
        const char *prog = compressor;
        const char *args[] = {
            prog,
            "-c",
            NULL
        };
5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608

        cmd = virCommandNewArgs(args);
        virCommandSetInputFD(cmd, pipeFD[0]);
        virCommandSetOutputFD(cmd, &fd);
        virCommandSetErrorBuffer(cmd, &errbuf);
        virCommandDoAsyncIO(cmd);
        if (virSetCloseExec(pipeFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("Unable to set cloexec flag"));
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
            goto cleanup;
5609
        }
5610 5611 5612 5613 5614 5615 5616 5617 5618 5619
        if (virCommandRunAsync(cmd, NULL) < 0) {
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
            goto cleanup;
        }
        rc = qemuMonitorMigrateToFd(priv->mon,
                                    QEMU_MONITOR_MIGRATE_BACKGROUND,
                                    pipeFD[1]);
        if (VIR_CLOSE(pipeFD[0]) < 0 ||
            VIR_CLOSE(pipeFD[1]) < 0)
            VIR_WARN("failed to close intermediate pipe");
5620
    }
5621
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5622
        goto cleanup;
5623 5624 5625
    if (rc < 0)
        goto cleanup;

5626
    rc = qemuMigrationWaitForCompletion(driver, vm, asyncJob, NULL, 0);
5627

5628 5629 5630 5631
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5632 5633
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5634
                qemuMonitorMigrateCancel(priv->mon);
5635
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5636 5637
            }
        }
5638
        goto cleanup;
5639
    }
5640

5641 5642 5643
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5644
    qemuDomainEventEmitJobCompleted(driver, vm);
5645 5646
    ret = 0;

5647
 cleanup:
5648 5649 5650
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5651
    /* Restore max migration bandwidth */
5652 5653
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5654 5655
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5656
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5657 5658
    }

5659 5660
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5661 5662 5663 5664 5665
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5666 5667 5668 5669 5670 5671

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

5672 5673
    return ret;
}
5674

5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721

int
qemuMigrationCancel(virQEMUDriverPtr driver,
                    virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virHashTablePtr blockJobs = NULL;
    bool storage = false;
    size_t i;
    int ret = -1;

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

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

    qemuDomainObjEnterMonitor(driver, vm);

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

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

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

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

        if (!diskPriv->migrating)
            continue;

        if (virHashLookup(blockJobs, disk->info.alias)) {
            VIR_DEBUG("Drive mirror on disk %s is still running", disk->dst);
        } else {
            VIR_DEBUG("Drive mirror on disk %s is gone", disk->dst);
5722
            qemuBlockJobSyncEnd(driver, vm, QEMU_ASYNC_JOB_NONE, disk);
5723 5724 5725 5726 5727
            diskPriv->migrating = false;
        }
    }

    if (qemuMigrationCancelDriveMirror(driver, vm, false,
5728
                                       QEMU_ASYNC_JOB_NONE, NULL) < 0)
5729 5730 5731 5732 5733 5734 5735 5736 5737 5738 5739 5740 5741 5742 5743
        goto endsyncjob;

    ret = 0;

 cleanup:
    virHashFree(blockJobs);
    return ret;

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

            if (diskPriv->migrating) {
5744
                qemuBlockJobSyncEnd(driver, vm, QEMU_ASYNC_JOB_NONE, disk);
5745 5746 5747 5748 5749 5750 5751 5752
                diskPriv->migrating = false;
            }
        }
    }
    goto cleanup;
}


5753
int
5754
qemuMigrationJobStart(virQEMUDriverPtr driver,
5755
                      virDomainObjPtr vm,
5756
                      qemuDomainAsyncJob job)
5757
{
5758
    qemuDomainObjPrivatePtr priv = vm->privateData;
5759 5760
    virDomainJobOperation op;
    unsigned long long mask;
5761

5762
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5763 5764
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_IN;
        mask = QEMU_JOB_NONE;
5765
    } else {
5766 5767 5768 5769
        op = VIR_DOMAIN_JOB_OPERATION_MIGRATION_OUT;
        mask = QEMU_JOB_DEFAULT_MASK |
               JOB_MASK(QEMU_JOB_SUSPEND) |
               JOB_MASK(QEMU_JOB_MIGRATION_OP);
5770
    }
5771

5772 5773 5774
    if (qemuDomainObjBeginAsyncJob(driver, vm, job, op) < 0)
        return -1;

5775 5776
    priv->job.current->statsType = QEMU_DOMAIN_JOB_STATS_TYPE_MIGRATION;

5777
    qemuDomainObjSetAsyncJobMask(vm, mask);
5778 5779 5780 5781
    return 0;
}

void
5782
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5783
                         virDomainObjPtr vm,
5784
                         qemuMigrationJobPhase phase)
5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798
{
    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);
}

void
5799
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5800
                           virDomainObjPtr vm,
5801
                           qemuMigrationJobPhase phase)
5802 5803 5804 5805
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5806
void
5807 5808
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5809
    qemuDomainObjReleaseAsyncJob(vm);
5810 5811 5812 5813
}

bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
5814
                         qemuDomainAsyncJob job)
5815 5816 5817 5818 5819 5820 5821 5822 5823 5824 5825
{
    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");

5826
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5827 5828 5829 5830 5831
        return false;
    }
    return true;
}

5832
void
5833
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5834
{
5835
    qemuDomainObjEndAsyncJob(driver, vm);
5836
}
5837 5838 5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890


static void
qemuMigrationErrorFree(void *data,
                       const void *name ATTRIBUTE_UNUSED)
{
    virErrorPtr err = data;
    virFreeError(err);
}

int
qemuMigrationErrorInit(virQEMUDriverPtr driver)
{
    driver->migrationErrors = virHashAtomicNew(64, qemuMigrationErrorFree);
    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
qemuMigrationErrorSave(virQEMUDriverPtr driver,
                       const char *name,
                       virErrorPtr err)
{
    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
qemuMigrationErrorReport(virQEMUDriverPtr driver,
                         const char *name)
{
    virErrorPtr err;

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

    VIR_DEBUG("Restoring saved incoming migration error for domain %s: %s",
              name, err->message);
    virSetError(err);
    virFreeError(err);
}
5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928


/* don't ever pass NULL params with non zero nparams */
qemuMigrationCompressionPtr
qemuMigrationCompressionParse(virTypedParameterPtr params,
                              int nparams,
                              unsigned long flags)
{
    size_t i;
    qemuMigrationCompressionPtr compression = NULL;

    if (VIR_ALLOC(compression) < 0)
        return NULL;

    for (i = 0; i < nparams; i++) {
        int method;

        if (STRNEQ(params[i].field, VIR_MIGRATE_PARAM_COMPRESSION))
            continue;

        method = qemuMigrationCompressMethodTypeFromString(params[i].value.s);
        if (method < 0) {
            virReportError(VIR_ERR_INVALID_ARG,
                           _("Unsupported compression method '%s'"),
                           params[i].value.s);
            goto error;
        }

        if (compression->methods & (1ULL << method)) {
            virReportError(VIR_ERR_INVALID_ARG,
                           _("Compression method '%s' is specified twice"),
                           params[i].value.s);
            goto error;
        }

        compression->methods |= 1ULL << method;
    }

5929 5930 5931 5932 5933 5934
#define GET_PARAM(PARAM, TYPE, VALUE) \
    do { \
        int rc; \
        const char *par = VIR_MIGRATE_PARAM_COMPRESSION_ ## PARAM; \
 \
        if ((rc = virTypedParamsGet ## TYPE(params, nparams, \
5935
                                            par, &compression->VALUE)) < 0) \
5936 5937 5938 5939
            goto error; \
 \
        if (rc == 1) \
            compression->VALUE ## _set = true; \
5940 5941 5942
    } while (0)

    if (params) {
5943 5944 5945 5946
        GET_PARAM(MT_LEVEL, Int, level);
        GET_PARAM(MT_THREADS, Int, threads);
        GET_PARAM(MT_DTHREADS, Int, dthreads);
        GET_PARAM(XBZRLE_CACHE, ULLong, xbzrle_cache);
5947 5948 5949 5950
    }

#undef GET_PARAM

5951 5952 5953
    if ((compression->level_set ||
         compression->threads_set ||
         compression->dthreads_set) &&
5954 5955 5956 5957 5958 5959 5960 5961 5962 5963 5964 5965 5966
        !(compression->methods & (1ULL << QEMU_MIGRATION_COMPRESS_MT))) {
        virReportError(VIR_ERR_INVALID_ARG, "%s",
                       _("Turn multithread compression on to tune it"));
        goto error;
    }

    if (compression->xbzrle_cache_set &&
        !(compression->methods & (1ULL << QEMU_MIGRATION_COMPRESS_XBZRLE))) {
        virReportError(VIR_ERR_INVALID_ARG, "%s",
                       _("Turn xbzrle compression on to tune it"));
        goto error;
    }

5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984 5985
    if (!compression->methods && (flags & VIR_MIGRATE_COMPRESSED))
        compression->methods = 1ULL << QEMU_MIGRATION_COMPRESS_XBZRLE;

    return compression;

 error:
    VIR_FREE(compression);
    return NULL;
}

int
qemuMigrationCompressionDump(qemuMigrationCompressionPtr compression,
                             virTypedParameterPtr *params,
                             int *nparams,
                             int *maxparams,
                             unsigned long *flags)
{
    size_t i;

5986 5987
    if (compression->methods == 1ULL << QEMU_MIGRATION_COMPRESS_XBZRLE &&
        !compression->xbzrle_cache_set) {
5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999
        *flags |= VIR_MIGRATE_COMPRESSED;
        return 0;
    }

    for (i = 0; i < QEMU_MIGRATION_COMPRESS_LAST; ++i) {
        if ((compression->methods & (1ULL << i)) &&
            virTypedParamsAddString(params, nparams, maxparams,
                                    VIR_MIGRATE_PARAM_COMPRESSION,
                                    qemuMigrationCompressMethodTypeToString(i)) < 0)
            return -1;
    }

6000
    if (compression->level_set &&
6001 6002
        virTypedParamsAddInt(params, nparams, maxparams,
                             VIR_MIGRATE_PARAM_COMPRESSION_MT_LEVEL,
6003
                             compression->level) < 0)
6004 6005
        return -1;

6006
    if (compression->threads_set &&
6007 6008
        virTypedParamsAddInt(params, nparams, maxparams,
                             VIR_MIGRATE_PARAM_COMPRESSION_MT_THREADS,
6009
                             compression->threads) < 0)
6010 6011
        return -1;

6012
    if (compression->dthreads_set &&
6013 6014
        virTypedParamsAddInt(params, nparams, maxparams,
                             VIR_MIGRATE_PARAM_COMPRESSION_MT_DTHREADS,
6015
                             compression->dthreads) < 0)
6016 6017 6018 6019 6020 6021 6022 6023
        return -1;

    if (compression->xbzrle_cache_set &&
        virTypedParamsAddULLong(params, nparams, maxparams,
                                VIR_MIGRATE_PARAM_COMPRESSION_XBZRLE_CACHE,
                                compression->xbzrle_cache) < 0)
        return -1;

6024 6025
    return 0;
}
6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038


/*
 * qemuMigrationReset:
 *
 * Reset all migration parameters so that the next job which internally uses
 * migration (save, managedsave, snapshots, dump) will not try to use them.
 */
void
qemuMigrationReset(virQEMUDriverPtr driver,
                   virDomainObjPtr vm,
                   qemuDomainAsyncJob job)
{
6039
    qemuMonitorMigrationCaps cap;
6040
    virErrorPtr err = virSaveLastError();
6041

6042
    if (!virDomainObjIsActive(vm))
6043
        goto cleanup;
6044

6045
    if (qemuMigrationResetTLS(driver, vm, job) < 0)
6046
        goto cleanup;
6047 6048

    for (cap = 0; cap < QEMU_MONITOR_MIGRATION_CAPS_LAST; cap++) {
6049 6050
        if (qemuMigrationCapsGet(vm, cap) &&
            qemuMigrationSetOption(driver, vm, cap, false, job) < 0)
6051 6052 6053 6054 6055 6056 6057
            goto cleanup;
    }

 cleanup:
    if (err) {
        virSetError(err);
        virFreeError(err);
6058
    }
6059
}
6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088 6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110


int
qemuMigrationFetchMirrorStats(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              qemuDomainAsyncJob asyncJob,
                              qemuDomainJobInfoPtr jobInfo)
{
    size_t i;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool nbd = false;
    virHashTablePtr blockinfo = NULL;
    qemuDomainMirrorStatsPtr stats = &jobInfo->mirrorStats;

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

    if (!nbd)
        return 0;

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

    blockinfo = qemuMonitorGetAllBlockJobInfo(priv->mon);

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

    memset(stats, 0, sizeof(*stats));

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

        if (!diskPriv->migrating ||
            !(data = virHashLookup(blockinfo, disk->info.alias)))
            continue;

        stats->transferred += data->cur;
        stats->total += data->end;
    }

    virHashFree(blockinfo);
    return 0;
}
6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121 6122 6123 6124


bool
qemuMigrationCapsGet(virDomainObjPtr vm,
                     qemuMonitorMigrationCaps cap)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool enabled = false;

    if (priv->migrationCaps)
        ignore_value(virBitmapGetBit(priv->migrationCaps, cap, &enabled));

    return enabled;
}