qemu_migration.c 183.8 KB
Newer Older
1 2 3
/*
 * qemu_migration.c: QEMU migration handling
 *
4
 * Copyright (C) 2006-2015 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23
 *
 */

#include <config.h>

J
Ján Tomko 已提交
24 25
#include <netdb.h>
#include <sys/socket.h>
26
#include <sys/time.h>
27
#ifdef WITH_GNUTLS
28 29 30
# include <gnutls/gnutls.h>
# include <gnutls/x509.h>
#endif
31
#include <fcntl.h>
32
#include <poll.h>
33 34 35 36 37 38

#include "qemu_migration.h"
#include "qemu_monitor.h"
#include "qemu_domain.h"
#include "qemu_process.h"
#include "qemu_capabilities.h"
39
#include "qemu_command.h"
40
#include "qemu_cgroup.h"
41
#include "qemu_hotplug.h"
42

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

#define VIR_FROM_THIS VIR_FROM_QEMU

66 67
VIR_LOG_INIT("qemu.qemu_migration");

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

81
enum qemuMigrationCookieFlags {
82
    QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS,
83
    QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE,
84
    QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT,
85
    QEMU_MIGRATION_COOKIE_FLAG_NETWORK,
86
    QEMU_MIGRATION_COOKIE_FLAG_NBD,
87
    QEMU_MIGRATION_COOKIE_FLAG_STATS,
88 89 90 91 92 93 94

    QEMU_MIGRATION_COOKIE_FLAG_LAST
};

VIR_ENUM_DECL(qemuMigrationCookieFlag);
VIR_ENUM_IMPL(qemuMigrationCookieFlag,
              QEMU_MIGRATION_COOKIE_FLAG_LAST,
95 96 97 98
              "graphics",
              "lockstate",
              "persistent",
              "network",
99 100
              "nbd",
              "statistics");
101 102 103

enum qemuMigrationCookieFeatures {
    QEMU_MIGRATION_COOKIE_GRAPHICS  = (1 << QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS),
104
    QEMU_MIGRATION_COOKIE_LOCKSTATE = (1 << QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE),
105
    QEMU_MIGRATION_COOKIE_PERSISTENT = (1 << QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT),
106
    QEMU_MIGRATION_COOKIE_NETWORK = (1 << QEMU_MIGRATION_COOKIE_FLAG_NETWORK),
107
    QEMU_MIGRATION_COOKIE_NBD = (1 << QEMU_MIGRATION_COOKIE_FLAG_NBD),
108
    QEMU_MIGRATION_COOKIE_STATS = (1 << QEMU_MIGRATION_COOKIE_FLAG_STATS),
109 110 111 112 113 114 115 116 117 118 119 120
};

typedef struct _qemuMigrationCookieGraphics qemuMigrationCookieGraphics;
typedef qemuMigrationCookieGraphics *qemuMigrationCookieGraphicsPtr;
struct _qemuMigrationCookieGraphics {
    int type;
    int port;
    int tlsPort;
    char *listen;
    char *tlsSubject;
};

121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141
typedef struct _qemuMigrationCookieNetData qemuMigrationCookieNetData;
typedef qemuMigrationCookieNetData *qemuMigrationCookieNetDataPtr;
struct _qemuMigrationCookieNetData {
    int vporttype; /* enum virNetDevVPortProfile */

    /*
     * Array of pointers to saved data. Each VIF will have it's own
     * data to transfer.
     */
    char *portdata;
};

typedef struct _qemuMigrationCookieNetwork qemuMigrationCookieNetwork;
typedef qemuMigrationCookieNetwork *qemuMigrationCookieNetworkPtr;
struct _qemuMigrationCookieNetwork {
    /* How many virtual NICs are we saving data for? */
    int nnets;

    qemuMigrationCookieNetDataPtr net;
};

142 143 144 145
typedef struct _qemuMigrationCookieNBD qemuMigrationCookieNBD;
typedef qemuMigrationCookieNBD *qemuMigrationCookieNBDPtr;
struct _qemuMigrationCookieNBD {
    int port; /* on which port does NBD server listen for incoming data */
146 147 148 149 150 151

    size_t ndisks;  /* Number of items in @disk array */
    struct {
        char *target;                   /* Disk target */
        unsigned long long capacity;    /* And its capacity */
    } *disks;
152 153
};

154 155 156
typedef struct _qemuMigrationCookie qemuMigrationCookie;
typedef qemuMigrationCookie *qemuMigrationCookiePtr;
struct _qemuMigrationCookie {
E
Eric Blake 已提交
157 158
    unsigned int flags;
    unsigned int flagsMandatory;
159 160

    /* Host properties */
161 162 163 164
    unsigned char localHostuuid[VIR_UUID_BUFLEN];
    unsigned char remoteHostuuid[VIR_UUID_BUFLEN];
    char *localHostname;
    char *remoteHostname;
165 166 167 168

    /* Guest properties */
    unsigned char uuid[VIR_UUID_BUFLEN];
    char *name;
169

170 171 172 173
    /* If (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) */
    char *lockState;
    char *lockDriver;

174 175
    /* If (flags & QEMU_MIGRATION_COOKIE_GRAPHICS) */
    qemuMigrationCookieGraphicsPtr graphics;
176 177 178

    /* If (flags & QEMU_MIGRATION_COOKIE_PERSISTENT) */
    virDomainDefPtr persistent;
179 180 181

    /* If (flags & QEMU_MIGRATION_COOKIE_NETWORK) */
    qemuMigrationCookieNetworkPtr network;
182 183 184

    /* If (flags & QEMU_MIGRATION_COOKIE_NBD) */
    qemuMigrationCookieNBDPtr nbd;
185 186 187

    /* If (flags & QEMU_MIGRATION_COOKIE_STATS) */
    qemuDomainJobInfoPtr jobInfo;
188 189
};

190 191 192 193 194 195 196 197 198
static void qemuMigrationCookieGraphicsFree(qemuMigrationCookieGraphicsPtr grap)
{
    if (!grap)
        return;
    VIR_FREE(grap->listen);
    VIR_FREE(grap->tlsSubject);
    VIR_FREE(grap);
}

199

200 201 202
static void
qemuMigrationCookieNetworkFree(qemuMigrationCookieNetworkPtr network)
{
203
    size_t i;
204 205 206 207 208 209 210 211 212 213 214 215 216

    if (!network)
        return;

    if (network->net) {
        for (i = 0; i < network->nnets; i++)
            VIR_FREE(network->net[i].portdata);
    }
    VIR_FREE(network->net);
    VIR_FREE(network);
}


217 218 219 220 221 222 223 224 225 226 227 228
static void qemuMigrationCookieNBDFree(qemuMigrationCookieNBDPtr nbd)
{
    if (!nbd)
        return;

    while (nbd->ndisks)
        VIR_FREE(nbd->disks[--nbd->ndisks].target);
    VIR_FREE(nbd->disks);
    VIR_FREE(nbd);
}


229 230 231 232 233
static void qemuMigrationCookieFree(qemuMigrationCookiePtr mig)
{
    if (!mig)
        return;

234 235
    qemuMigrationCookieGraphicsFree(mig->graphics);
    qemuMigrationCookieNetworkFree(mig->network);
236
    qemuMigrationCookieNBDFree(mig->nbd);
237

238 239
    VIR_FREE(mig->localHostname);
    VIR_FREE(mig->remoteHostname);
240
    VIR_FREE(mig->name);
241 242
    VIR_FREE(mig->lockState);
    VIR_FREE(mig->lockDriver);
243
    VIR_FREE(mig->jobInfo);
244 245 246 247
    VIR_FREE(mig);
}


248
#ifdef WITH_GNUTLS
249 250 251 252 253 254 255 256 257 258 259 260
static char *
qemuDomainExtractTLSSubject(const char *certdir)
{
    char *certfile = NULL;
    char *subject = NULL;
    char *pemdata = NULL;
    gnutls_datum_t pemdatum;
    gnutls_x509_crt_t cert;
    int ret;
    size_t subjectlen;

    if (virAsprintf(&certfile, "%s/server-cert.pem", certdir) < 0)
261
        goto error;
262 263

    if (virFileReadAll(certfile, 8192, &pemdata) < 0) {
264 265
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unable to read server cert %s"), certfile);
266 267 268 269 270
        goto error;
    }

    ret = gnutls_x509_crt_init(&cert);
    if (ret < 0) {
271 272 273
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot initialize cert object: %s"),
                       gnutls_strerror(ret));
274 275 276 277 278 279 280 281
        goto error;
    }

    pemdatum.data = (unsigned char *)pemdata;
    pemdatum.size = strlen(pemdata);

    ret = gnutls_x509_crt_import(cert, &pemdatum, GNUTLS_X509_FMT_PEM);
    if (ret < 0) {
282 283 284
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot load cert data from %s: %s"),
                       certfile, gnutls_strerror(ret));
285 286 287 288 289
        goto error;
    }

    subjectlen = 1024;
    if (VIR_ALLOC_N(subject, subjectlen+1) < 0)
290
        goto error;
291 292 293 294 295 296 297 298 299

    gnutls_x509_crt_get_dn(cert, subject, &subjectlen);
    subject[subjectlen] = '\0';

    VIR_FREE(certfile);
    VIR_FREE(pemdata);

    return subject;

300
 error:
301 302 303 304
    VIR_FREE(certfile);
    VIR_FREE(pemdata);
    return NULL;
}
305
#endif
306 307

static qemuMigrationCookieGraphicsPtr
308
qemuMigrationCookieGraphicsAlloc(virQEMUDriverPtr driver,
309 310 311 312
                                 virDomainGraphicsDefPtr def)
{
    qemuMigrationCookieGraphicsPtr mig = NULL;
    const char *listenAddr;
313
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
314 315

    if (VIR_ALLOC(mig) < 0)
316
        goto error;
317 318 319 320

    mig->type = def->type;
    if (mig->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
        mig->port = def->data.vnc.port;
321
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
322
        if (!listenAddr)
323
            listenAddr = cfg->vncListen;
324

325
#ifdef WITH_GNUTLS
326 327
        if (cfg->vncTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->vncTLSx509certdir)))
328
            goto error;
329
#endif
330 331
    } else {
        mig->port = def->data.spice.port;
332
        if (cfg->spiceTLS)
333 334 335
            mig->tlsPort = def->data.spice.tlsPort;
        else
            mig->tlsPort = -1;
336
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
337
        if (!listenAddr)
338
            listenAddr = cfg->spiceListen;
339

340
#ifdef WITH_GNUTLS
341 342
        if (cfg->spiceTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->spiceTLSx509certdir)))
343
            goto error;
344
#endif
345
    }
346 347
    if (VIR_STRDUP(mig->listen, listenAddr) < 0)
        goto error;
348

349
    virObjectUnref(cfg);
350 351
    return mig;

352
 error:
353
    qemuMigrationCookieGraphicsFree(mig);
354
    virObjectUnref(cfg);
355 356 357 358
    return NULL;
}


359
static qemuMigrationCookieNetworkPtr
360
qemuMigrationCookieNetworkAlloc(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
361 362 363
                                virDomainDefPtr def)
{
    qemuMigrationCookieNetworkPtr mig;
364
    size_t i;
365 366

    if (VIR_ALLOC(mig) < 0)
367
        goto error;
368 369 370 371

    mig->nnets = def->nnets;

    if (VIR_ALLOC_N(mig->net, def->nnets) <0)
372
        goto error;
373 374 375 376 377 378 379 380 381 382 383 384 385 386 387

    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr netptr;
        virNetDevVPortProfilePtr vport;

        netptr = def->nets[i];
        vport = virDomainNetGetActualVirtPortProfile(netptr);

        if (vport) {
            mig->net[i].vporttype = vport->virtPortType;

            switch (vport->virtPortType) {
            case VIR_NETDEV_VPORT_PROFILE_NONE:
            case VIR_NETDEV_VPORT_PROFILE_8021QBG:
            case VIR_NETDEV_VPORT_PROFILE_8021QBH:
388
               break;
389
            case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
390 391
                if (virNetDevOpenvswitchGetMigrateData(&mig->net[i].portdata,
                                                       netptr->ifname) != 0) {
J
Jiri Denemark 已提交
392 393 394
                        virReportError(VIR_ERR_INTERNAL_ERROR,
                                       _("Unable to run command to get OVS port data for "
                                         "interface %s"), netptr->ifname);
395 396 397
                        goto error;
                }
                break;
398 399 400 401 402 403 404
            default:
                break;
            }
        }
    }
    return mig;

405
 error:
406 407 408 409
    qemuMigrationCookieNetworkFree(mig);
    return NULL;
}

410 411 412
static qemuMigrationCookiePtr
qemuMigrationCookieNew(virDomainObjPtr dom)
{
J
Jiri Denemark 已提交
413
    qemuDomainObjPrivatePtr priv = dom->privateData;
414
    qemuMigrationCookiePtr mig = NULL;
J
Jiri Denemark 已提交
415
    const char *name;
416 417

    if (VIR_ALLOC(mig) < 0)
418
        goto error;
419

J
Jiri Denemark 已提交
420 421 422 423
    if (priv->origname)
        name = priv->origname;
    else
        name = dom->def->name;
424 425
    if (VIR_STRDUP(mig->name, name) < 0)
        goto error;
426 427
    memcpy(mig->uuid, dom->def->uuid, VIR_UUID_BUFLEN);

428
    if (!(mig->localHostname = virGetHostname()))
429
        goto error;
430
    if (virGetHostUUID(mig->localHostuuid) < 0) {
431 432
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Unable to obtain host UUID"));
433 434 435 436 437
        goto error;
    }

    return mig;

438
 error:
439 440 441 442 443
    qemuMigrationCookieFree(mig);
    return NULL;
}


444 445
static int
qemuMigrationCookieAddGraphics(qemuMigrationCookiePtr mig,
446
                               virQEMUDriverPtr driver,
447 448
                               virDomainObjPtr dom)
{
449 450
    size_t i = 0;

451
    if (mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) {
452 453
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration graphics data already present"));
454 455 456
        return -1;
    }

457 458 459 460 461 462 463 464
    for (i = 0; i < dom->def->ngraphics; i++) {
       if (dom->def->graphics[i]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
           if (!(mig->graphics =
                 qemuMigrationCookieGraphicsAlloc(driver, dom->def->graphics[i])))
               return -1;
           mig->flags |= QEMU_MIGRATION_COOKIE_GRAPHICS;
           break;
       }
465
    }
466 467 468 469 470

    return 0;
}


471 472
static int
qemuMigrationCookieAddLockstate(qemuMigrationCookiePtr mig,
473
                                virQEMUDriverPtr driver,
474 475 476 477 478
                                virDomainObjPtr dom)
{
    qemuDomainObjPrivatePtr priv = dom->privateData;

    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
479 480
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration lockstate data already present"));
481 482 483 484
        return -1;
    }

    if (virDomainObjGetState(dom, NULL) == VIR_DOMAIN_PAUSED) {
485
        if (VIR_STRDUP(mig->lockState, priv->lockState) < 0)
486 487 488 489 490 491
            return -1;
    } else {
        if (virDomainLockProcessInquire(driver->lockManager, dom, &mig->lockState) < 0)
            return -1;
    }

492
    if (VIR_STRDUP(mig->lockDriver, virLockManagerPluginGetName(driver->lockManager)) < 0) {
493 494 495 496 497 498 499 500 501 502 503
        VIR_FREE(mig->lockState);
        return -1;
    }

    mig->flags |= QEMU_MIGRATION_COOKIE_LOCKSTATE;
    mig->flagsMandatory |= QEMU_MIGRATION_COOKIE_LOCKSTATE;

    return 0;
}


504 505 506 507 508
static int
qemuMigrationCookieAddPersistent(qemuMigrationCookiePtr mig,
                                 virDomainObjPtr dom)
{
    if (mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) {
509 510
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration persistent data already present"));
511 512 513 514 515 516 517 518 519 520 521 522 523
        return -1;
    }

    if (!dom->newDef)
        return 0;

    mig->persistent = dom->newDef;
    mig->flags |= QEMU_MIGRATION_COOKIE_PERSISTENT;
    mig->flagsMandatory |= QEMU_MIGRATION_COOKIE_PERSISTENT;
    return 0;
}


524 525
static int
qemuMigrationCookieAddNetwork(qemuMigrationCookiePtr mig,
526
                              virQEMUDriverPtr driver,
527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544
                              virDomainObjPtr dom)
{
    if (mig->flags & QEMU_MIGRATION_COOKIE_NETWORK) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Network migration data already present"));
        return -1;
    }

    if (dom->def->nnets > 0) {
        mig->network = qemuMigrationCookieNetworkAlloc(driver, dom->def);
        if (!mig->network)
            return -1;
        mig->flags |= QEMU_MIGRATION_COOKIE_NETWORK;
    }

    return 0;
}

545

546 547
static int
qemuMigrationCookieAddNBD(qemuMigrationCookiePtr mig,
548
                          virQEMUDriverPtr driver,
549 550 551
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
552 553
    virHashTablePtr stats = NULL;
    size_t i;
554
    int ret = -1, rc;
555 556 557

    /* It is not a bug if there already is a NBD data */
    if (!mig->nbd &&
558
        VIR_ALLOC(mig->nbd) < 0)
559 560
        return -1;

561 562 563 564 565 566 567 568 569 570 571 572 573 574
    if (vm->def->ndisks &&
        VIR_ALLOC_N(mig->nbd->disks, vm->def->ndisks) < 0)
        return -1;
    mig->nbd->ndisks = 0;

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

        if (!stats) {
            if (!(stats = virHashCreate(10, virHashValueFree)))
                goto cleanup;

            qemuDomainObjEnterMonitor(driver, vm);
575 576
            rc = qemuMonitorBlockStatsUpdateCapacity(priv->mon, stats, false);
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
577
                goto cleanup;
578
            if (rc < 0)
579 580 581 582 583 584 585 586 587 588 589 590 591 592
                goto cleanup;
        }

        if (!disk->info.alias ||
            !(entry = virHashLookup(stats, disk->info.alias)))
            continue;

        if (VIR_STRDUP(mig->nbd->disks[mig->nbd->ndisks].target,
                       disk->dst) < 0)
            goto cleanup;
        mig->nbd->disks[mig->nbd->ndisks].capacity = entry->capacity;
        mig->nbd->ndisks++;
    }

593 594 595
    mig->nbd->port = priv->nbdPort;
    mig->flags |= QEMU_MIGRATION_COOKIE_NBD;

596 597 598 599
    ret = 0;
 cleanup:
    virHashFree(stats);
    return ret;
600 601 602
}


603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621
static int
qemuMigrationCookieAddStatistics(qemuMigrationCookiePtr mig,
                                 virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (!priv->job.completed)
        return 0;

    if (!mig->jobInfo && VIR_ALLOC(mig->jobInfo) < 0)
        return -1;

    *mig->jobInfo = *priv->job.completed;
    mig->flags |= QEMU_MIGRATION_COOKIE_STATS;

    return 0;
}


622 623 624
static void qemuMigrationCookieGraphicsXMLFormat(virBufferPtr buf,
                                                 qemuMigrationCookieGraphicsPtr grap)
{
625
    virBufferAsprintf(buf, "<graphics type='%s' port='%d' listen='%s'",
626 627 628 629 630 631
                      virDomainGraphicsTypeToString(grap->type),
                      grap->port, grap->listen);
    if (grap->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE)
        virBufferAsprintf(buf, " tlsPort='%d'", grap->tlsPort);
    if (grap->tlsSubject) {
        virBufferAddLit(buf, ">\n");
632 633 634 635
        virBufferAdjustIndent(buf, 2);
        virBufferEscapeString(buf, "<cert info='subject' value='%s'/>\n", grap->tlsSubject);
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</graphics>\n");
636 637 638 639 640 641
    } else {
        virBufferAddLit(buf, "/>\n");
    }
}


642 643 644 645
static void
qemuMigrationCookieNetworkXMLFormat(virBufferPtr buf,
                                    qemuMigrationCookieNetworkPtr optr)
{
646
    size_t i;
647 648 649 650 651 652
    bool empty = true;

    for (i = 0; i < optr->nnets; i++) {
        /* If optr->net[i].vporttype is not set, there is nothing to transfer */
        if (optr->net[i].vporttype != VIR_NETDEV_VPORT_PROFILE_NONE) {
            if (empty) {
653 654
                virBufferAddLit(buf, "<network>\n");
                virBufferAdjustIndent(buf, 2);
655 656
                empty = false;
            }
657
            virBufferAsprintf(buf, "<interface index='%zu' vporttype='%s'",
658 659 660
                              i, virNetDevVPortTypeToString(optr->net[i].vporttype));
            if (optr->net[i].portdata) {
                virBufferAddLit(buf, ">\n");
661 662
                virBufferAdjustIndent(buf, 2);
                virBufferEscapeString(buf, "<portdata>%s</portdata>\n",
663
                                      optr->net[i].portdata);
664 665
                virBufferAdjustIndent(buf, -2);
                virBufferAddLit(buf, "</interface>\n");
666 667 668 669 670
            } else {
                virBufferAddLit(buf, "/>\n");
            }
        }
    }
671 672 673 674
    if (!empty) {
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</network>\n");
    }
675 676 677
}


678 679 680 681 682 683 684 685 686
static void
qemuMigrationCookieStatisticsXMLFormat(virBufferPtr buf,
                                       qemuDomainJobInfoPtr jobInfo)
{
    qemuMonitorMigrationStatus *status = &jobInfo->status;

    virBufferAddLit(buf, "<statistics>\n");
    virBufferAdjustIndent(buf, 2);

687 688 689
    virBufferAsprintf(buf, "<started>%llu</started>\n", jobInfo->started);
    virBufferAsprintf(buf, "<stopped>%llu</stopped>\n", jobInfo->stopped);

690 691 692 693 694 695 696 697 698 699
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_TIME_ELAPSED,
                      jobInfo->timeElapsed);
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_TIME_REMAINING,
                      jobInfo->timeRemaining);
    if (status->downtime_set)
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_DOWNTIME,
                          status->downtime);
700 701 702 703
    if (status->setup_time_set)
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_SETUP_TIME,
                          status->setup_time);
704 705 706 707 708 709 710 711 712 713

    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_TOTAL,
                      status->ram_total);
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_PROCESSED,
                      status->ram_transferred);
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_REMAINING,
                      status->ram_remaining);
714 715 716
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_BPS,
                      status->ram_bps);
717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738

    if (status->ram_duplicate_set) {
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_MEMORY_CONSTANT,
                          status->ram_duplicate);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_MEMORY_NORMAL,
                          status->ram_normal);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_MEMORY_NORMAL_BYTES,
                          status->ram_normal_bytes);
    }

    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_TOTAL,
                      status->disk_total);
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_PROCESSED,
                      status->disk_transferred);
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_REMAINING,
                      status->disk_remaining);
739 740 741
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_BPS,
                      status->disk_bps);
742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765

    if (status->xbzrle_set) {
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_COMPRESSION_CACHE,
                          status->xbzrle_cache_size);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_COMPRESSION_BYTES,
                          status->xbzrle_bytes);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_COMPRESSION_PAGES,
                          status->xbzrle_pages);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_COMPRESSION_CACHE_MISSES,
                          status->xbzrle_cache_miss);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_COMPRESSION_OVERFLOW,
                          status->xbzrle_overflow);
    }

    virBufferAdjustIndent(buf, -2);
    virBufferAddLit(buf, "</statistics>\n");
}


766
static int
767
qemuMigrationCookieXMLFormat(virQEMUDriverPtr driver,
768
                             virBufferPtr buf,
769
                             qemuMigrationCookiePtr mig)
770 771 772
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
    char hostuuidstr[VIR_UUID_STRING_BUFLEN];
773
    size_t i;
774 775

    virUUIDFormat(mig->uuid, uuidstr);
776
    virUUIDFormat(mig->localHostuuid, hostuuidstr);
777

778
    virBufferAddLit(buf, "<qemu-migration>\n");
779 780 781 782 783
    virBufferAdjustIndent(buf, 2);
    virBufferEscapeString(buf, "<name>%s</name>\n", mig->name);
    virBufferAsprintf(buf, "<uuid>%s</uuid>\n", uuidstr);
    virBufferEscapeString(buf, "<hostname>%s</hostname>\n", mig->localHostname);
    virBufferAsprintf(buf, "<hostuuid>%s</hostuuid>\n", hostuuidstr);
784

785
    for (i = 0; i < QEMU_MIGRATION_COOKIE_FLAG_LAST; i++) {
786
        if (mig->flagsMandatory & (1 << i))
787
            virBufferAsprintf(buf, "<feature name='%s'/>\n",
788 789 790
                              qemuMigrationCookieFlagTypeToString(i));
    }

791 792
    if ((mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        mig->graphics)
793 794
        qemuMigrationCookieGraphicsXMLFormat(buf, mig->graphics);

795 796
    if ((mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        mig->lockState) {
797
        virBufferAsprintf(buf, "<lockstate driver='%s'>\n",
798
                          mig->lockDriver);
799 800
        virBufferAdjustIndent(buf, 2);
        virBufferAsprintf(buf, "<leases>%s</leases>\n",
801
                          mig->lockState);
802 803
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</lockstate>\n");
804 805
    }

806 807
    if ((mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        mig->persistent) {
808 809 810
        if (qemuDomainDefFormatBuf(driver,
                                   mig->persistent,
                                   VIR_DOMAIN_XML_INACTIVE |
811 812
                                   VIR_DOMAIN_XML_SECURE |
                                   VIR_DOMAIN_XML_MIGRATABLE,
813
                                   buf) < 0)
814
            return -1;
815 816
    }

817 818 819
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NETWORK) && mig->network)
        qemuMigrationCookieNetworkXMLFormat(buf, mig->network);

820
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NBD) && mig->nbd) {
821
        virBufferAddLit(buf, "<nbd");
822 823
        if (mig->nbd->port)
            virBufferAsprintf(buf, " port='%d'", mig->nbd->port);
824 825 826 827 828 829 830 831 832 833 834 835 836 837
        if (mig->nbd->ndisks) {
            virBufferAddLit(buf, ">\n");
            virBufferAdjustIndent(buf, 2);
            for (i = 0; i < mig->nbd->ndisks; i++) {
                virBufferEscapeString(buf, "<disk target='%s'",
                                      mig->nbd->disks[i].target);
                virBufferAsprintf(buf, " capacity='%llu'/>\n",
                                  mig->nbd->disks[i].capacity);
            }
            virBufferAdjustIndent(buf, -2);
            virBufferAddLit(buf, "</nbd>\n");
        } else {
            virBufferAddLit(buf, "/>\n");
        }
838 839
    }

840 841 842
    if (mig->flags & QEMU_MIGRATION_COOKIE_STATS && mig->jobInfo)
        qemuMigrationCookieStatisticsXMLFormat(buf, mig->jobInfo);

843
    virBufferAdjustIndent(buf, -2);
844
    virBufferAddLit(buf, "</qemu-migration>\n");
845
    return 0;
846 847 848
}


849
static char *qemuMigrationCookieXMLFormatStr(virQEMUDriverPtr driver,
850
                                             qemuMigrationCookiePtr mig)
851 852 853
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;

854
    if (qemuMigrationCookieXMLFormat(driver, &buf, mig) < 0) {
855 856 857
        virBufferFreeAndReset(&buf);
        return NULL;
    }
858

859
    if (virBufferCheckError(&buf) < 0)
860 861 862 863 864 865
        return NULL;

    return virBufferContentAndReset(&buf);
}


866 867 868 869 870 871 872
static qemuMigrationCookieGraphicsPtr
qemuMigrationCookieGraphicsXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieGraphicsPtr grap;
    char *tmp;

    if (VIR_ALLOC(grap) < 0)
873
        goto error;
874 875

    if (!(tmp = virXPathString("string(./graphics/@type)", ctxt))) {
876 877
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing type attribute in migration data"));
878 879 880
        goto error;
    }
    if ((grap->type = virDomainGraphicsTypeFromString(tmp)) < 0) {
881 882
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unknown graphics type %s"), tmp);
883 884 885
        VIR_FREE(tmp);
        goto error;
    }
E
Eric Blake 已提交
886
    VIR_FREE(tmp);
887
    if (virXPathInt("string(./graphics/@port)", ctxt, &grap->port) < 0) {
888 889
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing port attribute in migration data"));
890 891 892 893
        goto error;
    }
    if (grap->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        if (virXPathInt("string(./graphics/@tlsPort)", ctxt, &grap->tlsPort) < 0) {
894 895
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing tlsPort attribute in migration data"));
896 897 898 899
            goto error;
        }
    }
    if (!(grap->listen = virXPathString("string(./graphics/@listen)", ctxt))) {
900 901
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing listen attribute in migration data"));
902 903 904
        goto error;
    }
    /* Optional */
905
    grap->tlsSubject = virXPathString("string(./graphics/cert[@info='subject']/@value)", ctxt);
906 907 908

    return grap;

909
 error:
910 911 912 913 914
    qemuMigrationCookieGraphicsFree(grap);
    return NULL;
}


915 916 917 918
static qemuMigrationCookieNetworkPtr
qemuMigrationCookieNetworkXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieNetworkPtr optr;
919
    size_t i;
920 921 922 923 924 925
    int n;
    xmlNodePtr *interfaces = NULL;
    char *vporttype;
    xmlNodePtr save_ctxt = ctxt->node;

    if (VIR_ALLOC(optr) < 0)
926
        goto error;
927 928 929 930 931 932 933 934

    if ((n = virXPathNodeSet("./network/interface", ctxt, &interfaces)) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing interface information"));
        goto error;
    }

    optr->nnets = n;
935
    if (VIR_ALLOC_N(optr->net, optr->nnets) < 0)
936
        goto error;
937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952

    for (i = 0; i < n; i++) {
        /* portdata is optional, and may not exist */
        ctxt->node = interfaces[i];
        optr->net[i].portdata = virXPathString("string(./portdata[1])", ctxt);

        if (!(vporttype = virXMLPropString(interfaces[i], "vporttype"))) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing vporttype attribute in migration data"));
            goto error;
        }
        optr->net[i].vporttype = virNetDevVPortTypeFromString(vporttype);
    }

    VIR_FREE(interfaces);

953
 cleanup:
954 955 956
    ctxt->node = save_ctxt;
    return optr;

957
 error:
958 959 960 961 962 963 964
    VIR_FREE(interfaces);
    qemuMigrationCookieNetworkFree(optr);
    optr = NULL;
    goto cleanup;
}


965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
static qemuMigrationCookieNBDPtr
qemuMigrationCookieNBDXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieNBDPtr ret = NULL;
    char *port = NULL, *capacity = NULL;
    size_t i;
    int n;
    xmlNodePtr *disks = NULL;
    xmlNodePtr save_ctxt = ctxt->node;

    if (VIR_ALLOC(ret) < 0)
        goto error;

    port = virXPathString("string(./nbd/@port)", ctxt);
    if (port && virStrToLong_i(port, NULL, 10, &ret->port) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Malformed nbd port '%s'"),
                       port);
        goto error;
    }

    /* Now check if source sent a list of disks to prealloc. We might be
     * talking to an older server, so it's not an error if the list is
     * missing. */
    if ((n = virXPathNodeSet("./nbd/disk", ctxt, &disks)) > 0) {
        if (VIR_ALLOC_N(ret->disks, n) < 0)
            goto error;
        ret->ndisks = n;

        for (i = 0; i < n; i++) {
            ctxt->node = disks[i];
            VIR_FREE(capacity);

            if (!(ret->disks[i].target = virXPathString("string(./@target)", ctxt))) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("Malformed disk target"));
                goto error;
            }

            capacity = virXPathString("string(./@capacity)", ctxt);
            if (!capacity ||
                virStrToLong_ull(capacity, NULL, 10,
                                 &ret->disks[i].capacity) < 0) {
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Malformed disk capacity: '%s'"),
                               NULLSTR(capacity));
                goto error;
            }
        }
    }

 cleanup:
    VIR_FREE(port);
    VIR_FREE(capacity);
    VIR_FREE(disks);
    ctxt->node = save_ctxt;
    return ret;
 error:
    qemuMigrationCookieNBDFree(ret);
    ret = NULL;
    goto cleanup;
}


1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044
static qemuDomainJobInfoPtr
qemuMigrationCookieStatisticsXMLParse(xmlXPathContextPtr ctxt)
{
    qemuDomainJobInfoPtr jobInfo = NULL;
    qemuMonitorMigrationStatus *status;
    xmlNodePtr save_ctxt = ctxt->node;

    if (!(ctxt->node = virXPathNode("./statistics", ctxt)))
        goto cleanup;

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

    status = &jobInfo->status;
    jobInfo->type = VIR_DOMAIN_JOB_COMPLETED;

1045 1046 1047
    virXPathULongLong("string(./started[1])", ctxt, &jobInfo->started);
    virXPathULongLong("string(./stopped[1])", ctxt, &jobInfo->stopped);

1048 1049 1050 1051 1052 1053 1054
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_TIME_ELAPSED "[1])",
                      ctxt, &jobInfo->timeElapsed);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_TIME_REMAINING "[1])",
                      ctxt, &jobInfo->timeRemaining);
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_DOWNTIME "[1])",
                          ctxt, &status->downtime) == 0)
        status->downtime_set = true;
1055 1056 1057
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_SETUP_TIME "[1])",
                          ctxt, &status->setup_time) == 0)
        status->setup_time_set = true;
1058 1059 1060 1061 1062 1063 1064

    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_TOTAL "[1])",
                      ctxt, &status->ram_total);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_PROCESSED "[1])",
                      ctxt, &status->ram_transferred);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_REMAINING "[1])",
                      ctxt, &status->ram_remaining);
1065 1066
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_BPS "[1])",
                      ctxt, &status->ram_bps);
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081

    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_CONSTANT "[1])",
                          ctxt, &status->ram_duplicate) == 0)
        status->ram_duplicate_set = true;
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_NORMAL "[1])",
                      ctxt, &status->ram_normal);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_NORMAL_BYTES "[1])",
                      ctxt, &status->ram_normal_bytes);

    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_TOTAL "[1])",
                      ctxt, &status->disk_total);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_PROCESSED "[1])",
                      ctxt, &status->disk_transferred);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_REMAINING "[1])",
                      ctxt, &status->disk_remaining);
1082 1083
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_BPS "[1])",
                      ctxt, &status->disk_bps);
1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102

    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_COMPRESSION_CACHE "[1])",
                          ctxt, &status->xbzrle_cache_size) == 0)
        status->xbzrle_set = true;
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_COMPRESSION_BYTES "[1])",
                      ctxt, &status->xbzrle_bytes);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_COMPRESSION_PAGES "[1])",
                      ctxt, &status->xbzrle_pages);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_COMPRESSION_CACHE_MISSES "[1])",
                      ctxt, &status->xbzrle_cache_miss);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_COMPRESSION_OVERFLOW "[1])",
                      ctxt, &status->xbzrle_overflow);

 cleanup:
    ctxt->node = save_ctxt;
    return jobInfo;
}


1103 1104
static int
qemuMigrationCookieXMLParse(qemuMigrationCookiePtr mig,
1105
                            virQEMUDriverPtr driver,
1106
                            xmlDocPtr doc,
1107
                            xmlXPathContextPtr ctxt,
E
Eric Blake 已提交
1108
                            unsigned int flags)
1109 1110
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
1111
    char *tmp = NULL;
1112
    xmlNodePtr *nodes = NULL;
1113 1114
    size_t i;
    int n;
1115 1116 1117 1118
    virCapsPtr caps = NULL;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto error;
1119 1120 1121 1122 1123 1124 1125 1126

    /* We don't store the uuid, name, hostname, or hostuuid
     * values. We just compare them to local data to do some
     * sanity checking on migration operation
     */

    /* Extract domain name */
    if (!(tmp = virXPathString("string(./name[1])", ctxt))) {
1127 1128
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing name element in migration data"));
1129 1130 1131
        goto error;
    }
    if (STRNEQ(tmp, mig->name)) {
1132 1133 1134
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected name %s vs %s"),
                       tmp, mig->name);
1135 1136 1137 1138 1139 1140 1141
        goto error;
    }
    VIR_FREE(tmp);

    /* Extract domain uuid */
    tmp = virXPathString("string(./uuid[1])", ctxt);
    if (!tmp) {
1142 1143
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing uuid element in migration data"));
1144 1145 1146 1147
        goto error;
    }
    virUUIDFormat(mig->uuid, uuidstr);
    if (STRNEQ(tmp, uuidstr)) {
1148 1149 1150
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected UUID %s vs %s"),
                       tmp, uuidstr);
1151
        goto error;
1152 1153 1154 1155
    }
    VIR_FREE(tmp);

    /* Check & forbid "localhost" migration */
1156
    if (!(mig->remoteHostname = virXPathString("string(./hostname[1])", ctxt))) {
1157 1158
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostname element in migration data"));
1159 1160
        goto error;
    }
1161
    if (STREQ(mig->remoteHostname, mig->localHostname)) {
1162 1163 1164
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Attempt to migrate guest to the same host %s"),
                       mig->remoteHostname);
1165 1166 1167 1168
        goto error;
    }

    if (!(tmp = virXPathString("string(./hostuuid[1])", ctxt))) {
1169 1170 1171
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostuuid element in migration data"));
        goto error;
1172
    }
1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184
    if (virUUIDParse(tmp, mig->remoteHostuuid) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("malformed hostuuid element in migration data"));
        goto error;
    }
    if (memcmp(mig->remoteHostuuid, mig->localHostuuid, VIR_UUID_BUFLEN) == 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Attempt to migrate guest to the same host %s"),
                       tmp);
        goto error;
    }
    VIR_FREE(tmp);
1185

1186 1187
    /* Check to ensure all mandatory features from XML are also
     * present in 'flags' */
1188
    if ((n = virXPathNodeSet("./feature", ctxt, &nodes)) < 0)
1189 1190
        goto error;

1191
    for (i = 0; i < n; i++) {
1192 1193 1194
        int val;
        char *str = virXMLPropString(nodes[i], "name");
        if (!str) {
1195 1196
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing feature name"));
1197 1198 1199 1200
            goto error;
        }

        if ((val = qemuMigrationCookieFlagTypeFromString(str)) < 0) {
1201 1202 1203
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unknown migration cookie feature %s"),
                           str);
1204 1205 1206 1207 1208
            VIR_FREE(str);
            goto error;
        }

        if ((flags & (1 << val)) == 0) {
1209 1210 1211
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unsupported migration cookie feature %s"),
                           str);
1212 1213 1214 1215 1216 1217
            VIR_FREE(str);
        }
        VIR_FREE(str);
    }
    VIR_FREE(nodes);

1218 1219 1220 1221 1222
    if ((flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        virXPathBoolean("count(./graphics) > 0", ctxt) &&
        (!(mig->graphics = qemuMigrationCookieGraphicsXMLParse(ctxt))))
        goto error;

1223 1224 1225 1226
    if ((flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        virXPathBoolean("count(./lockstate) > 0", ctxt)) {
        mig->lockDriver = virXPathString("string(./lockstate[1]/@driver)", ctxt);
        if (!mig->lockDriver) {
1227 1228
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Missing lock driver name in migration cookie"));
1229 1230 1231 1232 1233 1234 1235
            goto error;
        }
        mig->lockState = virXPathString("string(./lockstate[1]/leases[1])", ctxt);
        if (mig->lockState && STREQ(mig->lockState, ""))
            VIR_FREE(mig->lockState);
    }

1236 1237 1238
    if ((flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        virXPathBoolean("count(./domain) > 0", ctxt)) {
        if ((n = virXPathNodeSet("./domain", ctxt, &nodes)) > 1) {
1239 1240 1241 1242
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Too many domain elements in "
                             "migration cookie: %d"),
                           n);
1243 1244
            goto error;
        }
1245 1246
        mig->persistent = virDomainDefParseNode(doc, nodes[0],
                                                caps, driver->xmlopt,
1247
                                                -1, VIR_DOMAIN_DEF_PARSE_INACTIVE);
1248 1249 1250 1251 1252 1253 1254 1255
        if (!mig->persistent) {
            /* virDomainDefParseNode already reported
             * an error for us */
            goto error;
        }
        VIR_FREE(nodes);
    }

1256 1257 1258 1259 1260
    if ((flags & QEMU_MIGRATION_COOKIE_NETWORK) &&
        virXPathBoolean("count(./network) > 0", ctxt) &&
        (!(mig->network = qemuMigrationCookieNetworkXMLParse(ctxt))))
        goto error;

1261
    if (flags & QEMU_MIGRATION_COOKIE_NBD &&
1262 1263 1264
        virXPathBoolean("boolean(./nbd)", ctxt) &&
        (!(mig->nbd = qemuMigrationCookieNBDXMLParse(ctxt))))
        goto error;
1265

1266 1267 1268 1269 1270
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        virXPathBoolean("boolean(./statistics)", ctxt) &&
        (!(mig->jobInfo = qemuMigrationCookieStatisticsXMLParse(ctxt))))
        goto error;

1271
    virObjectUnref(caps);
1272 1273
    return 0;

1274
 error:
1275
    VIR_FREE(tmp);
1276
    VIR_FREE(nodes);
1277
    virObjectUnref(caps);
1278 1279 1280 1281 1282 1283
    return -1;
}


static int
qemuMigrationCookieXMLParseStr(qemuMigrationCookiePtr mig,
1284
                               virQEMUDriverPtr driver,
1285
                               const char *xml,
E
Eric Blake 已提交
1286
                               unsigned int flags)
1287 1288 1289
{
    xmlDocPtr doc = NULL;
    xmlXPathContextPtr ctxt = NULL;
1290
    int ret = -1;
1291 1292 1293

    VIR_DEBUG("xml=%s", NULLSTR(xml));

1294
    if (!(doc = virXMLParseStringCtxt(xml, _("(qemu_migration_cookie)"), &ctxt)))
1295 1296
        goto cleanup;

1297
    ret = qemuMigrationCookieXMLParse(mig, driver, doc, ctxt, flags);
1298

1299
 cleanup:
1300 1301 1302 1303 1304 1305 1306 1307 1308
    xmlXPathFreeContext(ctxt);
    xmlFreeDoc(doc);

    return ret;
}


static int
qemuMigrationBakeCookie(qemuMigrationCookiePtr mig,
1309
                        virQEMUDriverPtr driver,
1310
                        virDomainObjPtr dom,
1311 1312
                        char **cookieout,
                        int *cookieoutlen,
E
Eric Blake 已提交
1313
                        unsigned int flags)
1314
{
1315 1316
    if (!cookieout || !cookieoutlen)
        return 0;
1317 1318 1319

    *cookieoutlen = 0;

1320 1321 1322 1323
    if (flags & QEMU_MIGRATION_COOKIE_GRAPHICS &&
        qemuMigrationCookieAddGraphics(mig, driver, dom) < 0)
        return -1;

1324 1325 1326 1327
    if (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE &&
        qemuMigrationCookieAddLockstate(mig, driver, dom) < 0)
        return -1;

1328 1329 1330 1331
    if (flags & QEMU_MIGRATION_COOKIE_PERSISTENT &&
        qemuMigrationCookieAddPersistent(mig, dom) < 0)
        return -1;

1332 1333 1334 1335 1336
    if (flags & QEMU_MIGRATION_COOKIE_NETWORK &&
        qemuMigrationCookieAddNetwork(mig, driver, dom) < 0) {
        return -1;
    }

1337 1338 1339 1340
    if ((flags & QEMU_MIGRATION_COOKIE_NBD) &&
        qemuMigrationCookieAddNBD(mig, driver, dom) < 0)
        return -1;

1341 1342 1343 1344
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        qemuMigrationCookieAddStatistics(mig, dom) < 0)
        return -1;

1345
    if (!(*cookieout = qemuMigrationCookieXMLFormatStr(driver, mig)))
1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356
        return -1;

    *cookieoutlen = strlen(*cookieout) + 1;

    VIR_DEBUG("cookielen=%d cookie=%s", *cookieoutlen, *cookieout);

    return 0;
}


static qemuMigrationCookiePtr
1357
qemuMigrationEatCookie(virQEMUDriverPtr driver,
1358
                       virDomainObjPtr dom,
1359 1360
                       const char *cookiein,
                       int cookieinlen,
E
Eric Blake 已提交
1361
                       unsigned int flags)
1362 1363 1364 1365 1366 1367
{
    qemuMigrationCookiePtr mig = NULL;

    /* Parse & validate incoming cookie (if any) */
    if (cookiein && cookieinlen &&
        cookiein[cookieinlen-1] != '\0') {
1368 1369
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration cookie was not NULL terminated"));
1370 1371 1372 1373 1374 1375 1376 1377 1378 1379
        goto error;
    }

    VIR_DEBUG("cookielen=%d cookie='%s'", cookieinlen, NULLSTR(cookiein));

    if (!(mig = qemuMigrationCookieNew(dom)))
        return NULL;

    if (cookiein && cookieinlen &&
        qemuMigrationCookieXMLParseStr(mig,
1380
                                       driver,
1381 1382 1383 1384
                                       cookiein,
                                       flags) < 0)
        goto error;

1385 1386 1387
    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
        if (!mig->lockDriver) {
            if (virLockManagerPluginUsesState(driver->lockManager)) {
1388 1389 1390
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Missing %s lock state for migration cookie"),
                               virLockManagerPluginGetName(driver->lockManager));
1391 1392 1393 1394
                goto error;
            }
        } else if (STRNEQ(mig->lockDriver,
                          virLockManagerPluginGetName(driver->lockManager))) {
1395 1396 1397 1398
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Source host lock driver %s different from target %s"),
                           mig->lockDriver,
                           virLockManagerPluginGetName(driver->lockManager));
1399 1400 1401 1402
            goto error;
        }
    }

1403 1404
    return mig;

1405
 error:
1406 1407 1408
    qemuMigrationCookieFree(mig);
    return NULL;
}
1409

1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456
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;
    int state = virDomainObjGetState(vm, NULL);
    bool ret = false;

    VIR_DEBUG("driver=%p, vm=%p, pre-mig-state=%d, state=%d",
              driver, vm, priv->preMigrationState, state);

    if (state == VIR_DOMAIN_PAUSED &&
        priv->preMigrationState == VIR_DOMAIN_RUNNING) {
        /* 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 */
        if (qemuProcessStartCPUs(driver, vm, conn,
                                 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;
}

1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609

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++;

        if (!(pool = storagePoolLookupByTargetPath(conn, basePath)))
            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;

    case VIR_STORAGE_TYPE_BLOCK:
    case VIR_STORAGE_TYPE_DIR:
    case VIR_STORAGE_TYPE_NETWORK:
    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;
        break;
    }

    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;
}


static int
qemuMigrationPrecreateStorage(virConnectPtr conn,
                              virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                              virDomainObjPtr vm,
                              qemuMigrationCookieNBDPtr nbd)
{
    int ret = -1;
    size_t i = 0;

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

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

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

        if ((indx = virDomainDiskIndexByName(vm->def,
                                             nbd->disks[i].target, false)) < 0) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unable to find disk by target: %s"),
                           nbd->disks[i].target);
            goto cleanup;
        }

        disk = vm->def->disks[indx];
        diskSrcPath = virDomainDiskGetSource(disk);

        if (disk->src->shared || disk->src->readonly ||
            (diskSrcPath && virFileExists(diskSrcPath))) {
            /* Skip shared, read-only and already existing disks. */
            continue;
        }

        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;
}


1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623
/**
 * 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
1624 1625
                            virDomainObjPtr vm,
                            const char *listenAddr)
1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636
{
    int ret = -1;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    unsigned short port = 0;
    char *diskAlias = NULL;
    size_t i;

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

        /* skip shared, RO and source-less disks */
1637 1638
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1639 1640 1641 1642
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1643
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1644 1645 1646 1647 1648 1649 1650
            goto cleanup;

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

        if (!port &&
1651
            ((virPortAllocatorAcquire(driver->migrationPorts, &port) < 0) ||
1652
             (qemuMonitorNBDServerStart(priv->mon, listenAddr, port) < 0))) {
1653
            goto exit_monitor;
1654 1655
        }

1656 1657 1658
        if (qemuMonitorNBDServerAdd(priv->mon, diskAlias, true) < 0)
            goto exit_monitor;
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
1659 1660 1661 1662 1663 1664
            goto cleanup;
    }

    priv->nbdPort = port;
    ret = 0;

1665
 cleanup:
1666
    VIR_FREE(diskAlias);
1667
    if (ret < 0)
1668
        virPortAllocatorRelease(driver->migrationPorts, port);
1669
    return ret;
1670 1671 1672 1673

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;
1674 1675
}

1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709
/**
 * qemuMigrationDriveMirror:
 * @driver: qemu driver
 * @vm: domain
 * @mig: migration cookie
 * @host: where are we migrating to
 * @speed: how much should the copying be limited
 * @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
 * simultaneously to both source and destination. And this switch
 * is what we are waiting for before proceeding with the next
 * disk. On success, update @migrate_flags so we don't tell
 * 'migrate' command to do the very same operation.
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
qemuMigrationDriveMirror(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMigrationCookiePtr mig,
                         const char *host,
                         unsigned long speed,
                         unsigned int *migrate_flags)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int mon_ret;
    int port;
    size_t i, lastGood = 0;
    char *diskAlias = NULL;
    char *nbd_dest = NULL;
1710
    char *hoststr = NULL;
1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
    virErrorPtr err = NULL;

    if (!(*migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                            QEMU_MONITOR_MIGRATE_NON_SHARED_INC)))
        return 0;

    if (!mig->nbd) {
        /* Destination doesn't support NBD server.
         * Fall back to previous implementation. */
        VIR_DEBUG("Destination doesn't support NBD server "
                  "Falling back to previous implementation.");
        return 0;
    }

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

1730 1731
    /* escape literal IPv6 address */
    if (strchr(host, ':')) {
1732
        if (virAsprintf(&hoststr, "[%s]", host) < 0)
1733 1734 1735 1736 1737
            goto error;
    } else if (VIR_STRDUP(hoststr, host) < 0) {
        goto error;
    }

1738 1739 1740 1741 1742 1743 1744
    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];

        /* skip shared, RO and source-less disks */
1745 1746
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1747 1748 1749 1750 1751 1752 1753
            continue;

        VIR_FREE(diskAlias);
        VIR_FREE(nbd_dest);
        if ((virAsprintf(&diskAlias, "%s%s",
                         QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0) ||
            (virAsprintf(&nbd_dest, "nbd:%s:%d:exportname=%s",
1754
                         hoststr, port, diskAlias) < 0))
1755 1756 1757 1758 1759 1760
            goto error;

        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            goto error;
        mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest,
1761
                                         NULL, speed, 0, 0, mirror_flags);
1762 1763
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            goto error;
1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774

        if (mon_ret < 0)
            goto error;

        lastGood = i;

        /* wait for completion */
        while (true) {
            /* Poll every 500ms for progress & to allow cancellation */
            struct timespec ts = { .tv_sec = 0, .tv_nsec = 500 * 1000 * 1000ull };

1775 1776 1777 1778 1779
            /* Explicitly check if domain is still alive. Maybe qemu
             * died meanwhile so we won't see any event at all. */
            if (!virDomainObjIsActive(vm)) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("guest unexpectedly quit"));
1780
                goto error;
1781 1782 1783 1784 1785
            }

            /* The following check should be race free as long as the variable
             * is set only with domain object locked. And here we have the
             * domain object locked too. */
1786
            if (priv->job.asyncAbort) {
1787
                priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
1788 1789 1790 1791 1792 1793
                virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                               qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                               _("canceled by client"));
                goto error;
            }

1794
            if (disk->mirrorState == VIR_DOMAIN_DISK_MIRROR_STATE_READY) {
1795 1796
                VIR_DEBUG("Drive mirroring of '%s' completed", diskAlias);
                break;
1797 1798 1799 1800 1801
            } else if (disk->mirrorState == VIR_DOMAIN_DISK_MIRROR_STATE_ABORT) {
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("migration of disk %s failed"),
                               disk->dst);
                goto error;
1802 1803
            }

1804
            /* XXX Turn this into virCond someday. */
1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818

            virObjectUnlock(vm);

            nanosleep(&ts, NULL);

            virObjectLock(vm);
        }
    }

    /* Okay, copied. Modify migrate_flags */
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

1819
 cleanup:
1820 1821
    VIR_FREE(diskAlias);
    VIR_FREE(nbd_dest);
1822
    VIR_FREE(hoststr);
1823 1824
    return ret;

1825
 error:
1826 1827 1828 1829 1830 1831 1832
    /* don't overwrite any errors */
    err = virSaveLastError();
    /* cancel any outstanding jobs */
    while (lastGood) {
        virDomainDiskDefPtr disk = vm->def->disks[--lastGood];

        /* skip shared, RO disks */
1833 1834
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1835 1836 1837 1838
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1839
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1840 1841 1842
            continue;
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
1843
            if (qemuMonitorBlockJob(priv->mon, diskAlias, NULL, NULL, 0,
1844
                                    BLOCK_JOB_ABORT, true) < 0) {
1845 1846
                VIR_WARN("Unable to cancel block-job on '%s'", diskAlias);
            }
1847 1848
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                break;
1849 1850 1851 1852 1853 1854 1855 1856 1857
        } else {
            VIR_WARN("Unable to enter monitor. No block job cancelled");
        }
    }
    if (err)
        virSetError(err);
    virFreeError(err);
    goto cleanup;
}
1858

1859

1860
static int
1861 1862 1863 1864 1865 1866 1867
qemuMigrationStopNBDServer(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuMigrationCookiePtr mig)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (!mig->nbd)
1868
        return 0;
1869 1870 1871

    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
1872
        return -1;
1873 1874 1875

    if (qemuMonitorNBDServerStop(priv->mon) < 0)
        VIR_WARN("Unable to stop NBD server");
1876 1877
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
1878

1879
    virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
1880
    priv->nbdPort = 0;
1881
    return 0;
1882 1883
}

1884
static int
1885 1886 1887 1888 1889 1890 1891
qemuMigrationCancelDriveMirror(qemuMigrationCookiePtr mig,
                               virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    size_t i;
    char *diskAlias = NULL;
1892
    int ret = 0;
1893 1894 1895 1896 1897 1898 1899

    VIR_DEBUG("mig=%p nbdPort=%d", mig->nbd, priv->nbdPort);

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

        /* skip shared, RO and source-less disks */
1900 1901
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1902 1903 1904 1905
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1906
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1907 1908 1909 1910 1911 1912
            goto cleanup;

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

1913
        if (qemuMonitorBlockJob(priv->mon, diskAlias, NULL, NULL, 0,
1914
                                BLOCK_JOB_ABORT, true) < 0)
1915
            VIR_WARN("Unable to stop block job on %s", diskAlias);
1916 1917 1918 1919
        if (qemuDomainObjExitMonitor(driver, vm) < 0) {
            ret = -1;
            goto cleanup;
        }
1920 1921
    }

1922
 cleanup:
1923
    VIR_FREE(diskAlias);
1924
    return ret;
1925 1926
}

1927 1928 1929 1930 1931 1932 1933 1934 1935
/* Validate whether the domain is safe to migrate.  If vm is NULL,
 * then this is being run in the v2 Prepare stage on the destination
 * (where we only have the target xml); if vm is provided, then this
 * is being run in either v2 Perform or v3 Begin (where we also have
 * access to all of the domain's metadata, such as whether it is
 * marked autodestroy or has snapshots).  While it would be nice to
 * assume that checking on source is sufficient to prevent ever
 * talking to the destination in the first place, we are stuck with
 * the fact that older servers did not do checks on the source. */
1936
bool
1937
qemuMigrationIsAllowed(virQEMUDriverPtr driver, virDomainObjPtr vm,
1938
                       virDomainDefPtr def, bool remote, bool abort_on_error)
1939
{
1940
    int nsnapshots;
1941
    int pauseReason;
1942
    bool forbid;
1943
    size_t i;
1944 1945 1946

    if (vm) {
        if (qemuProcessAutoDestroyActive(driver, vm)) {
1947 1948
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("domain is marked for auto destroy"));
1949 1950
            return false;
        }
1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963

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

            if (nsnapshots > 0) {
                virReportError(VIR_ERR_OPERATION_INVALID,
                               _("cannot migrate domain with %d snapshots"),
                               nsnapshots);
                return false;
            }
1964 1965

            /* cancel migration if disk I/O error is emitted while migrating */
1966 1967
            if (abort_on_error &&
                virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
1968 1969 1970 1971 1972 1973
                pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("cannot migrate domain with I/O error"));
                return false;
            }

1974
        }
1975

E
Eric Blake 已提交
1976 1977
        if (virDomainHasDiskMirror(vm)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1978
                           _("domain has an active block job"));
E
Eric Blake 已提交
1979 1980
            return false;
        }
1981 1982 1983

        def = vm->def;
    }
1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998

    /* Migration with USB host devices is allowed, all other devices are
     * forbidden.
     */
    forbid = false;
    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) {
            forbid = true;
            break;
        }
    }
    if (forbid) {
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1999
                       _("domain has assigned non-USB host devices"));
2000 2001 2002
        return false;
    }

2003 2004 2005
    if (def->cpu && def->cpu->mode != VIR_CPU_MODE_HOST_PASSTHROUGH) {
        for (i = 0; i < def->cpu->nfeatures; i++) {
            virCPUFeatureDefPtr feature = &def->cpu->features[i];
J
Ján Tomko 已提交
2006

2007 2008
            if (feature->policy != VIR_CPU_FEATURE_REQUIRE)
                continue;
J
Ján Tomko 已提交
2009

2010 2011 2012 2013 2014 2015 2016
            /* QEMU blocks migration and save with invariant TSC enabled */
            if (STREQ(feature->name, "invtsc")) {
                virReportError(VIR_ERR_OPERATION_INVALID,
                               _("domain has CPU feature: %s"),
                               feature->name);
                return false;
            }
J
Ján Tomko 已提交
2017 2018 2019
        }
    }

2020 2021 2022
    return true;
}

2023 2024 2025
static bool
qemuMigrationIsSafe(virDomainDefPtr def)
{
2026
    size_t i;
2027

2028
    for (i = 0; i < def->ndisks; i++) {
2029
        virDomainDiskDefPtr disk = def->disks[i];
2030
        const char *src = virDomainDiskGetSource(disk);
2031

2032 2033
        /* Our code elsewhere guarantees shared disks are either readonly (in
         * which case cache mode doesn't matter) or used with cache=none */
2034
        if (src &&
2035 2036
            !disk->src->shared &&
            !disk->src->readonly &&
2037
            disk->cachemode != VIR_DOMAIN_DISK_CACHE_DISABLE) {
2038
            int rc;
2039

E
Eric Blake 已提交
2040
            if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
2041
                if ((rc = virFileIsSharedFS(src)) < 0)
2042 2043
                    return false;
                else if (rc == 0)
2044
                    continue;
2045
                if ((rc = virStorageFileIsClusterFS(src)) < 0)
2046
                    return false;
2047 2048
                else if (rc == 1)
                    continue;
2049 2050
            } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                       disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
2051
                continue;
2052
            }
2053

2054 2055 2056
            virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                           _("Migration may lead to data corruption if disks"
                             " use cache != none"));
2057 2058 2059 2060 2061 2062 2063
            return false;
        }
    }

    return true;
}

2064 2065 2066 2067
/** qemuMigrationSetOffline
 * Pause domain for non-live migration.
 */
int
2068
qemuMigrationSetOffline(virQEMUDriverPtr driver,
2069 2070 2071
                        virDomainObjPtr vm)
{
    int ret;
2072
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
2073 2074
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
2075
    if (ret == 0) {
2076
        virObjectEventPtr event;
2077

2078
        event = virDomainEventLifecycleNewFromObj(vm,
2079 2080 2081 2082 2083 2084 2085 2086 2087 2088
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
        if (event)
            qemuDomainEventQueue(driver, event);
    }

    return ret;
}


2089 2090 2091
static int
qemuMigrationSetCompression(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
2092
                            bool state,
2093
                            qemuDomainAsyncJob job)
2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret;

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

    ret = qemuMonitorGetMigrationCapability(
                priv->mon,
                QEMU_MONITOR_MIGRATION_CAPS_XBZRLE);

    if (ret < 0) {
        goto cleanup;
2107 2108 2109
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125
    } else if (ret == 0) {
        if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                           _("Compressed migration is not supported by "
                             "target QEMU binary"));
        } else {
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                           _("Compressed migration is not supported by "
                             "source QEMU binary"));
        }
        ret = -1;
        goto cleanup;
    }

    ret = qemuMonitorSetMigrationCapability(
                priv->mon,
2126 2127
                QEMU_MONITOR_MIGRATION_CAPS_XBZRLE,
                state);
2128

2129
 cleanup:
2130 2131
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2132 2133 2134
    return ret;
}

2135 2136 2137
static int
qemuMigrationSetAutoConverge(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
2138
                             bool state,
2139
                             qemuDomainAsyncJob job)
2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret;

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

    ret = qemuMonitorGetMigrationCapability(
                priv->mon,
                QEMU_MONITOR_MIGRATION_CAPS_AUTO_CONVERGE);

    if (ret < 0) {
        goto cleanup;
2153 2154 2155
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2156 2157 2158 2159 2160 2161 2162 2163 2164 2165
    } else if (ret == 0) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("Auto-Converge is not supported by "
                         "QEMU binary"));
        ret = -1;
        goto cleanup;
    }

    ret = qemuMonitorSetMigrationCapability(
                priv->mon,
2166 2167
                QEMU_MONITOR_MIGRATION_CAPS_AUTO_CONVERGE,
                state);
2168

2169
 cleanup:
2170 2171
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2172 2173 2174 2175
    return ret;
}


2176 2177 2178
static int
qemuMigrationSetPinAll(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
2179
                       bool state,
2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193
                       qemuDomainAsyncJob job)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret;

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

    ret = qemuMonitorGetMigrationCapability(
                priv->mon,
                QEMU_MONITOR_MIGRATION_CAPS_RDMA_PIN_ALL);

    if (ret < 0) {
        goto cleanup;
2194 2195 2196
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212
    } else if (ret == 0) {
        if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                           _("rdma pinning migration is not supported by "
                             "target QEMU binary"));
        } else {
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                           _("rdma pinning migration is not supported by "
                             "source QEMU binary"));
        }
        ret = -1;
        goto cleanup;
    }

    ret = qemuMonitorSetMigrationCapability(
                priv->mon,
2213 2214
                QEMU_MONITOR_MIGRATION_CAPS_RDMA_PIN_ALL,
                state);
2215 2216

 cleanup:
2217 2218
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2219 2220 2221
    return ret;
}

2222 2223 2224 2225 2226 2227 2228
static int
qemuMigrationWaitForSpice(virQEMUDriverPtr driver,
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool wait_for_spice = false;
    bool spice_migrated = false;
2229
    size_t i = 0;
2230
    int rc;
2231

2232 2233 2234 2235 2236 2237 2238 2239
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION)) {
        for (i = 0; i < vm->def->ngraphics; i++) {
            if (vm->def->graphics[i]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
                wait_for_spice = true;
                break;
            }
        }
    }
2240 2241 2242 2243 2244 2245 2246 2247

    if (!wait_for_spice)
        return 0;

    while (!spice_migrated) {
        /* Poll every 50ms for progress & to allow cancellation */
        struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };

2248 2249 2250 2251
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            return -1;

2252 2253 2254 2255
        rc = qemuMonitorGetSpiceMigrationStatus(priv->mon, &spice_migrated);
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
        if (rc < 0)
2256 2257 2258 2259 2260 2261 2262 2263
            return -1;
        virObjectUnlock(vm);
        nanosleep(&ts, NULL);
        virObjectLock(vm);
    }

    return 0;
}
2264

2265
static int
2266
qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver,
2267
                             virDomainObjPtr vm,
2268
                             const char *job,
2269
                             qemuDomainAsyncJob asyncJob)
2270
{
2271
    qemuDomainObjPrivatePtr priv = vm->privateData;
2272
    qemuMonitorMigrationStatus status;
J
Jiri Denemark 已提交
2273 2274
    qemuDomainJobInfoPtr jobInfo;
    int ret;
2275 2276

    memset(&status, 0, sizeof(status));
2277

2278 2279
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
    if (ret < 0) {
2280 2281 2282
        /* Guest already exited or waiting for the job timed out; nothing
         * further to update. */
        return ret;
2283
    }
2284
    ret = qemuMonitorGetMigrationStatus(priv->mon, &status);
2285

2286 2287
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2288

J
Jiri Denemark 已提交
2289 2290
    if (ret < 0 ||
        qemuDomainJobInfoUpdateTime(priv->job.current) < 0)
2291
        return -1;
2292

2293
    ret = -1;
J
Jiri Denemark 已提交
2294 2295 2296 2297 2298
    jobInfo = priv->job.current;
    switch (status.status) {
    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
        jobInfo->type = VIR_DOMAIN_JOB_COMPLETED;
        /* fall through */
M
Michael Avdienko 已提交
2299
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
2300
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
2301
    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLING:
2302 2303 2304
        ret = 0;
        break;

J
Jiri Denemark 已提交
2305 2306 2307 2308
    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
        jobInfo->type = VIR_DOMAIN_JOB_NONE;
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("is not active"));
2309 2310 2311
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
J
Jiri Denemark 已提交
2312
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
2313 2314
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("unexpectedly failed"));
2315 2316 2317
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
J
Jiri Denemark 已提交
2318
        jobInfo->type = VIR_DOMAIN_JOB_CANCELLED;
2319 2320
        virReportError(VIR_ERR_OPERATION_ABORTED,
                       _("%s: %s"), job, _("canceled by client"));
2321 2322
        break;
    }
J
Jiri Denemark 已提交
2323
    jobInfo->status = status;
2324 2325 2326 2327 2328

    return ret;
}


2329 2330 2331
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
2332
static int
J
Jiri Denemark 已提交
2333 2334
qemuMigrationWaitForCompletion(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
2335
                               qemuDomainAsyncJob asyncJob,
J
Jiri Denemark 已提交
2336 2337
                               virConnectPtr dconn,
                               bool abort_on_error)
2338
{
2339
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
2340
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
2341
    const char *job;
2342
    int pauseReason;
2343

2344 2345
    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
2346 2347
        job = _("migration job");
        break;
2348
    case QEMU_ASYNC_JOB_SAVE:
2349 2350
        job = _("domain save job");
        break;
2351
    case QEMU_ASYNC_JOB_DUMP:
2352 2353 2354 2355 2356
        job = _("domain core dump job");
        break;
    default:
        job = _("job");
    }
2357

J
Jiri Denemark 已提交
2358
    jobInfo->type = VIR_DOMAIN_JOB_UNBOUNDED;
2359

J
Jiri Denemark 已提交
2360
    while (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
2361 2362 2363
        /* Poll every 50ms for progress & to allow cancellation */
        struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };

2364
        if (qemuMigrationUpdateJobStatus(driver, vm, job, asyncJob) == -1)
2365
            break;
2366

2367
        /* cancel migration if disk I/O error is emitted while migrating */
2368
        if (abort_on_error &&
2369
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
2370 2371 2372 2373 2374
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("%s: %s"), job, _("failed due to I/O error"));
            break;
        }
2375

2376
        if (dconn && virConnectIsAlive(dconn) <= 0) {
2377 2378
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
2379
            break;
2380 2381
        }

2382
        virObjectUnlock(vm);
2383 2384 2385

        nanosleep(&ts, NULL);

2386
        virObjectLock(vm);
2387 2388
    }

J
Jiri Denemark 已提交
2389
    if (jobInfo->type == VIR_DOMAIN_JOB_COMPLETED) {
2390
        qemuDomainJobInfoUpdateDowntime(jobInfo);
2391 2392 2393
        VIR_FREE(priv->job.completed);
        if (VIR_ALLOC(priv->job.completed) == 0)
            *priv->job.completed = *jobInfo;
2394
        return 0;
J
Jiri Denemark 已提交
2395
    } else if (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
2396 2397 2398
        /* The migration was aborted by us rather than QEMU itself so let's
         * update the job type and notify the caller to send migrate_cancel.
         */
J
Jiri Denemark 已提交
2399
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
2400 2401
        return -2;
    } else {
2402
        return -1;
2403
    }
2404 2405 2406
}


2407
static int
2408
qemuDomainMigrateGraphicsRelocate(virQEMUDriverPtr driver,
2409
                                  virDomainObjPtr vm,
2410 2411
                                  qemuMigrationCookiePtr cookie,
                                  const char *graphicsuri)
2412 2413
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2414 2415
    int ret = -1;
    const char *listenAddress = NULL;
2416
    virSocketAddr addr;
2417 2418 2419 2420 2421
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
2422

2423
    if (!cookie || (!cookie->graphics && !graphicsuri))
2424 2425
        return 0;

2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444
    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) {
2445
        size_t i;
2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472

        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;
            }
        }
    }
2473 2474 2475 2476

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
2477 2478 2479 2480
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
2481

2482 2483 2484 2485
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
2486 2487
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
2488
    }
2489

2490
 cleanup:
2491
    virURIFree(uri);
2492 2493 2494 2495
    return ret;
}


2496
static int
2497
qemuDomainMigrateOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
2498 2499 2500
                             virDomainObjPtr vm,
                             qemuMigrationCookiePtr cookie)
{
2501 2502
    virDomainNetDefPtr netptr;
    int ret = -1;
2503
    size_t i;
2504 2505 2506 2507 2508 2509 2510 2511

    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:
2512
           break;
2513
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
2514 2515
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
2516 2517 2518
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
2519 2520 2521
                goto cleanup;
            }
            break;
2522 2523 2524 2525 2526
        default:
            break;
        }
    }

2527
    ret = 0;
2528
 cleanup:
2529 2530 2531 2532
    return ret;
}


2533 2534 2535 2536 2537 2538
/* 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.
 */
2539 2540 2541 2542
static virDomainObjPtr
qemuMigrationCleanup(virDomainObjPtr vm,
                     virConnectPtr conn,
                     void *opaque)
2543
{
2544
    virQEMUDriverPtr driver = opaque;
2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559
    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);

2560
    switch ((qemuMigrationJobPhase) priv->job.phase) {
2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590
    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 */
        ;
    }

2591
 cleanup:
2592 2593 2594
    return vm;
}

2595

2596
/* The caller is supposed to lock the vm and start a migration job. */
2597 2598
static char
*qemuMigrationBeginPhase(virQEMUDriverPtr driver,
2599
                         virDomainObjPtr vm,
2600
                         const char *xmlin,
2601
                         const char *dname,
2602
                         char **cookieout,
2603 2604
                         int *cookieoutlen,
                         unsigned long flags)
2605 2606 2607
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
2608
    virDomainDefPtr def = NULL;
2609
    qemuDomainObjPrivatePtr priv = vm->privateData;
2610
    virCapsPtr caps = NULL;
2611
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
2612
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2613

2614
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
2615
              " cookieout=%p, cookieoutlen=%p, flags=%lx",
2616
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
2617
              cookieout, cookieoutlen, flags);
2618

2619 2620 2621
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2622 2623 2624 2625 2626 2627
    /* 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);
2628

2629
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
2630 2631
        goto cleanup;

2632 2633 2634
    if (!(flags & VIR_MIGRATE_UNSAFE) && !qemuMigrationIsSafe(vm->def))
        goto cleanup;

2635
    if (flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
2636
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DRIVE_MIRROR)) {
2637
        /* TODO support NBD for TUNNELLED migration */
2638
        if (flags & VIR_MIGRATE_TUNNELLED) {
2639 2640 2641 2642
            VIR_WARN("NBD in tunnelled migration is currently not supported");
        } else {
            cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
            priv->nbdPort = 0;
2643 2644 2645
        }
    }

2646
    if (!(mig = qemuMigrationEatCookie(driver, vm, NULL, 0, 0)))
2647 2648 2649 2650
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
                                cookieout, cookieoutlen,
2651
                                cookieFlags) < 0)
2652 2653
        goto cleanup;

L
liguang 已提交
2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675
    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;
        }
    }

2676
    if (xmlin) {
2677
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt,
M
Matthias Bolte 已提交
2678
                                            QEMU_EXPECTED_VIRT_TYPES,
2679
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE)))
2680 2681
            goto cleanup;

2682
        if (!qemuDomainDefCheckABIStability(driver, vm->def, def))
2683 2684
            goto cleanup;

2685
        rv = qemuDomainDefFormatLive(driver, def, false, true);
2686
    } else {
2687
        rv = qemuDomainDefFormatLive(driver, vm->def, false, true);
2688
    }
2689

2690
 cleanup:
2691
    qemuMigrationCookieFree(mig);
2692
    virObjectUnref(caps);
2693
    virDomainDefFree(def);
2694 2695 2696
    return rv;
}

2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707
char *
qemuMigrationBegin(virConnectPtr conn,
                   virDomainObjPtr vm,
                   const char *xmlin,
                   const char *dname,
                   char **cookieout,
                   int *cookieoutlen,
                   unsigned long flags)
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
2708
    qemuDomainAsyncJob asyncJob;
2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719

    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;
    }

2720 2721
    qemuMigrationStoreDomainState(vm);

2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744
    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) &&
        qemuDomainCheckEjectableMedia(driver, vm, asyncJob) < 0)
        goto endjob;

    if (!(xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
                                        cookieout, cookieoutlen,
                                        flags)))
        goto endjob;

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
        /* We keep the job active across API calls until the confirm() call.
         * This prevents any other APIs being invoked while migration is taking
         * place.
         */
2745 2746
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                 qemuMigrationCleanup) < 0)
2747
            goto endjob;
2748
        qemuMigrationJobContinue(vm);
2749 2750 2751 2752
    } else {
        goto endjob;
    }

2753
 cleanup:
2754
    qemuDomObjEndAPI(&vm);
2755 2756
    return xml;

2757
 endjob:
2758 2759 2760 2761
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2762 2763 2764
    goto cleanup;
}

2765

2766 2767
/* Prepare is the first step, and it runs on the destination host.
 */
2768

2769
static void
2770
qemuMigrationPrepareCleanup(virQEMUDriverPtr driver,
2771 2772 2773 2774 2775 2776 2777 2778 2779 2780
                            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));

2781 2782 2783
    virPortAllocatorRelease(driver->migrationPorts, priv->migrationPort);
    priv->migrationPort = 0;

2784 2785 2786 2787 2788
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2789
static int
2790
qemuMigrationPrepareAny(virQEMUDriverPtr driver,
2791 2792 2793 2794 2795
                        virConnectPtr dconn,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
2796
                        virDomainDefPtr *def,
2797
                        const char *origname,
L
liguang 已提交
2798
                        virStreamPtr st,
2799
                        const char *protocol,
2800 2801
                        unsigned short port,
                        bool autoPort,
2802
                        const char *listenAddress,
L
liguang 已提交
2803
                        unsigned long flags)
2804 2805
{
    virDomainObjPtr vm = NULL;
2806
    virObjectEventPtr event = NULL;
2807
    int ret = -1;
2808
    int dataFD[2] = { -1, -1 };
2809
    qemuDomainObjPrivatePtr priv = NULL;
J
Jiri Denemark 已提交
2810
    unsigned long long now;
2811
    qemuMigrationCookiePtr mig = NULL;
2812
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2813
    char *xmlout = NULL;
L
liguang 已提交
2814
    unsigned int cookieFlags;
2815
    virCapsPtr caps = NULL;
J
Ján Tomko 已提交
2816
    char *migrateFrom = NULL;
2817
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2818
    bool taint_hook = false;
2819

2820
    if (virTimeMillisNow(&now) < 0)
2821 2822
        return -1;

2823 2824
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846
    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;
        }
    }

2847 2848 2849
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2850
    if (!qemuMigrationIsAllowed(driver, NULL, *def, true, abort_on_error))
2851 2852
        goto cleanup;

J
Jiri Denemark 已提交
2853 2854 2855 2856 2857
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2858
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2859 2860
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2861 2862
            goto cleanup;

2863
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2864 2865 2866 2867 2868 2869 2870
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2871
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2872 2873 2874 2875 2876 2877
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2878
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt,
J
Jiri Denemark 已提交
2879
                                                 QEMU_EXPECTED_VIRT_TYPES,
2880
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE);
J
Jiri Denemark 已提交
2881 2882 2883
                if (!newdef)
                    goto cleanup;

2884
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2885 2886 2887 2888
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2889 2890
                virDomainDefFree(*def);
                *def = newdef;
2891 2892 2893 2894
                /* 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
2895 2896 2897 2898
            }
        }
    }

J
Ján Tomko 已提交
2899 2900 2901 2902
    if (tunnel) {
        /* QEMU will be started with -incoming stdio
         * (which qemu_command might convert to exec:cat or fd:n)
         */
2903
        if (VIR_STRDUP(migrateFrom, "stdio") < 0)
J
Ján Tomko 已提交
2904 2905
            goto cleanup;
    } else {
2906 2907 2908
        bool encloseAddress = false;
        bool hostIPv6Capable = false;
        bool qemuIPv6Capable = false;
J
Ján Tomko 已提交
2909 2910 2911 2912
        virQEMUCapsPtr qemuCaps = NULL;
        struct addrinfo *info = NULL;
        struct addrinfo hints = { .ai_flags = AI_ADDRCONFIG,
                                  .ai_socktype = SOCK_STREAM };
2913
        const char *incFormat;
J
Ján Tomko 已提交
2914

2915 2916 2917 2918
        if (getaddrinfo("::", NULL, &hints, &info) == 0) {
            freeaddrinfo(info);
            hostIPv6Capable = true;
        }
J
Ján Tomko 已提交
2919
        if (!(qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
2920 2921
                                                    (*def)->emulator,
                                                    (*def)->os.machine)))
J
Ján Tomko 已提交
2922 2923
            goto cleanup;

2924 2925 2926 2927
        qemuIPv6Capable = virQEMUCapsGet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
        virObjectUnref(qemuCaps);

        if (listenAddress) {
2928 2929 2930 2931 2932 2933 2934 2935 2936
            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"));
2937 2938
                    goto cleanup;
                }
2939 2940
                /* IPv6 address must be escaped in brackets on the cmd line */
                encloseAddress = true;
2941
            } else {
2942
                /* listenAddress is a hostname or IPv4 */
2943
            }
2944
        } else if (qemuIPv6Capable && hostIPv6Capable) {
2945 2946 2947
            /* Listen on :: instead of 0.0.0.0 if QEMU understands it
             * and there is at least one IPv6 address configured
             */
2948 2949 2950 2951
            listenAddress = "::";
            encloseAddress = true;
        } else {
            listenAddress = "0.0.0.0";
J
Ján Tomko 已提交
2952 2953
        }

2954 2955 2956 2957
        /* QEMU will be started with
         *   -incoming protocol:[<IPv6 addr>]:port,
         *   -incoming protocol:<IPv4 addr>:port, or
         *   -incoming protocol:<hostname>:port
J
Ján Tomko 已提交
2958
         */
2959 2960 2961 2962 2963 2964
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";
        if (virAsprintf(&migrateFrom, incFormat,
                        protocol, listenAddress, port) < 0)
J
Ján Tomko 已提交
2965 2966 2967
            goto cleanup;
    }

2968
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2969
                                   driver->xmlopt,
2970 2971 2972
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2973
        goto cleanup;
2974

2975
    virObjectRef(vm);
2976
    *def = NULL;
2977
    priv = vm->privateData;
2978 2979
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2980

2981 2982 2983 2984 2985
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2986
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
2987 2988
                                       QEMU_MIGRATION_COOKIE_LOCKSTATE |
                                       QEMU_MIGRATION_COOKIE_NBD)))
2989 2990
        goto cleanup;

2991 2992
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
2993 2994 2995 2996 2997 2998
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2999 3000 3001
    if (qemuMigrationPrecreateStorage(dconn, driver, vm, mig->nbd) < 0)
        goto cleanup;

3002
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
3003
        goto cleanup;
3004
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
3005 3006 3007 3008

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

L
liguang 已提交
3009 3010 3011
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

3012 3013
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
3014 3015
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
3016 3017 3018 3019
        goto endjob;
    }

    /* Start the QEMU daemon, with the same command-line arguments plus
3020
     * -incoming $migrateFrom
3021
     */
3022 3023
    if (qemuProcessStart(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                         migrateFrom, dataFD[0], NULL, NULL,
3024 3025
                         VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
                         VIR_QEMU_PROCESS_START_PAUSED |
3026
                         VIR_QEMU_PROCESS_START_AUTODESTROY) < 0) {
3027
        virDomainAuditStart(vm, "migrated", false);
3028 3029 3030
        goto endjob;
    }

3031 3032 3033 3034
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
3035
            goto stop;
3036
        }
3037
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
3038 3039
    }

3040 3041
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
3042 3043 3044
                                    QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stop;

3045
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
3046 3047 3048 3049
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
        goto stop;
    }

3050 3051 3052
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
                               QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
3053 3054
        goto stop;

3055 3056 3057 3058 3059 3060 3061 3062
    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");
    }
3063

3064
 done:
L
liguang 已提交
3065 3066 3067 3068 3069
    if (flags & VIR_MIGRATE_OFFLINE)
        cookieFlags = 0;
    else
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS;

3070 3071 3072
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
3073
        if (qemuMigrationStartNBDServer(driver, vm, listenAddress) < 0) {
3074 3075
            /* error already reported */
            goto endjob;
3076
        }
3077
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
3078 3079 3080 3081
    }

    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
3082 3083 3084 3085 3086 3087 3088
        /* 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");
    }

3089 3090 3091
    if (qemuDomainCleanupAdd(vm, qemuMigrationPrepareCleanup) < 0)
        goto endjob;

L
liguang 已提交
3092 3093
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
3094
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
3095 3096 3097
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
3098

3099 3100 3101 3102
    /* 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.
     */
3103
    qemuMigrationJobContinue(vm);
3104

3105 3106
    if (autoPort)
        priv->migrationPort = port;
3107
    ret = 0;
3108

3109
 cleanup:
J
Ján Tomko 已提交
3110
    VIR_FREE(migrateFrom);
J
Jiri Denemark 已提交
3111
    VIR_FREE(xmlout);
3112 3113
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
3114 3115 3116
    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 */
3117
        VIR_FREE(priv->origname);
3118 3119 3120
        virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
        priv->nbdPort = 0;
        qemuDomainRemoveInactive(driver, vm);
3121
    }
3122
    qemuDomObjEndAPI(&vm);
3123 3124
    if (event)
        qemuDomainEventQueue(driver, event);
3125
    qemuMigrationCookieFree(mig);
3126
    virObjectUnref(caps);
3127
    virNWFilterUnlockFilterUpdates();
3128
    return ret;
3129

3130
 stop:
3131 3132 3133
    virDomainAuditStart(vm, "migrated", false);
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, 0);

3134
 endjob:
3135
    qemuMigrationJobFinish(driver, vm);
3136
    goto cleanup;
3137 3138 3139
}


3140 3141 3142 3143 3144
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
3145
qemuMigrationPrepareTunnel(virQEMUDriverPtr driver,
3146 3147 3148 3149 3150 3151
                           virConnectPtr dconn,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virStreamPtr st,
3152
                           virDomainDefPtr *def,
3153
                           const char *origname,
L
liguang 已提交
3154
                           unsigned long flags)
3155 3156 3157 3158
{
    int ret;

    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
3159
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
3160
              "origname=%s, flags=%lx",
3161
              driver, dconn, NULLSTR(cookiein), cookieinlen,
3162
              cookieout, cookieoutlen, st, *def, origname, flags);
3163

3164 3165 3166 3167 3168 3169
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

3170
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3171
                                  cookieout, cookieoutlen, def, origname,
3172
                                  st, NULL, 0, false, NULL, flags);
3173 3174 3175 3176
    return ret;
}


3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199
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;
}


3200
int
3201
qemuMigrationPrepareDirect(virQEMUDriverPtr driver,
3202
                           virConnectPtr dconn,
3203 3204 3205 3206
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
3207 3208
                           const char *uri_in,
                           char **uri_out,
3209
                           virDomainDefPtr *def,
3210
                           const char *origname,
3211
                           const char *listenAddress,
L
liguang 已提交
3212
                           unsigned long flags)
3213
{
3214 3215
    unsigned short port = 0;
    bool autoPort = true;
3216 3217
    char *hostname = NULL;
    int ret = -1;
3218
    virURIPtr uri = NULL;
3219 3220
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
3221

3222 3223
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
3224
              "def=%p, origname=%s, listenAddress=%s, flags=%lx",
3225 3226
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
3227
              *def, origname, NULLSTR(listenAddress), flags);
3228

3229 3230
    *uri_out = NULL;

3231 3232 3233
    /* 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
3234 3235 3236
     * 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".
3237 3238 3239 3240 3241 3242
     *
     * 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) {
3243 3244 3245
        bool encloseAddress = false;
        const char *incFormat;

3246
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3247
            goto cleanup;
3248

3249
        if (migrateHost != NULL) {
3250 3251
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
3252

3253
            if (VIR_STRDUP(hostname, migrateHost) < 0)
3254 3255 3256 3257 3258
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
3259 3260

        if (STRPREFIX(hostname, "localhost")) {
3261 3262 3263
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
3264 3265 3266 3267 3268
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
3269
         * compatibility with old targets. We at least make the
3270 3271
         * new targets accept both syntaxes though.
         */
3272 3273 3274 3275 3276 3277
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
3278 3279
            goto cleanup;
    } else {
3280
        bool well_formed_uri;
J
Ján Tomko 已提交
3281

3282 3283
        if (!(uri = qemuMigrationParseURI(uri_in, &well_formed_uri)))
            goto cleanup;
J
Ján Tomko 已提交
3284

3285 3286 3287 3288 3289 3290 3291
        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
3292 3293
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
3294 3295 3296
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
3297 3298 3299 3300 3301 3302 3303 3304 3305 3306
            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) {
3307
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3308
                goto cleanup;
3309

3310
            /* Send well-formed URI only if uri_in was well-formed */
3311 3312 3313 3314 3315 3316 3317 3318
            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;
            }
3319
        } else {
3320 3321
            port = uri->port;
            autoPort = false;
3322 3323 3324 3325 3326 3327
        }
    }

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

3328
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3329
                                  cookieout, cookieoutlen, def, origname,
3330 3331
                                  NULL, uri ? uri->scheme : "tcp",
                                  port, autoPort, listenAddress, flags);
3332
 cleanup:
3333
    virURIFree(uri);
3334
    VIR_FREE(hostname);
3335
    virObjectUnref(cfg);
3336
    if (ret != 0) {
3337
        VIR_FREE(*uri_out);
3338 3339 3340
        if (autoPort)
            virPortAllocatorRelease(driver->migrationPorts, port);
    }
3341 3342 3343 3344
    return ret;
}


3345 3346 3347
virDomainDefPtr
qemuMigrationPrepareDef(virQEMUDriverPtr driver,
                        const char *dom_xml,
3348 3349
                        const char *dname,
                        char **origname)
3350 3351 3352
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
3353
    char *name = NULL;
3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365

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

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

    if (!(def = virDomainDefParseString(dom_xml, caps, driver->xmlopt,
                                        QEMU_EXPECTED_VIRT_TYPES,
3366
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE)))
3367 3368 3369
        goto cleanup;

    if (dname) {
3370
        name = def->name;
3371 3372 3373 3374 3375 3376
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

3377
 cleanup:
3378
    virObjectUnref(caps);
3379 3380 3381 3382
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
3383 3384 3385 3386
    return def;
}


3387 3388 3389 3390 3391 3392 3393 3394 3395 3396
static int
qemuMigrationConfirmPhase(virQEMUDriverPtr driver,
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *cookiein,
                          int cookieinlen,
                          unsigned int flags,
                          int retcode)
{
    qemuMigrationCookiePtr mig;
3397
    virObjectEventPtr event = NULL;
3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412
    int rv = -1;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);

    VIR_DEBUG("driver=%p, conn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
              "flags=%x, retcode=%d",
              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);

3413 3414
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                       QEMU_MIGRATION_COOKIE_STATS)))
3415 3416
        goto cleanup;

3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433
    /* Update total times with the values sent by the destination daemon */
    if (mig->jobInfo) {
        qemuDomainObjPrivatePtr priv = vm->privateData;
        if (priv->job.completed) {
            qemuDomainJobInfoPtr jobInfo = priv->job.completed;
            if (mig->jobInfo->status.downtime_set) {
                jobInfo->status.downtime = mig->jobInfo->status.downtime;
                jobInfo->status.downtime_set = true;
            }
            if (mig->jobInfo->timeElapsed)
                jobInfo->timeElapsed = mig->jobInfo->timeElapsed;
        } else {
            priv->job.completed = mig->jobInfo;
            mig->jobInfo = NULL;
        }
    }

3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

    /* Did the migration go as planned?  If yes, kill off the
     * domain object, but if no, resume CPUs
     */
    if (retcode == 0) {
        /* If guest uses SPICE and supports seamless migration we have to hold
         * up domain shutdown until SPICE server transfers its data */
        qemuMigrationWaitForSpice(driver, vm);

        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
        virDomainAuditStop(vm, "migrated");

3449
        event = virDomainEventLifecycleNewFromObj(vm,
3450 3451 3452 3453 3454 3455 3456
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
    } else {

        /* cancel any outstanding NBD jobs */
        qemuMigrationCancelDriveMirror(mig, driver, vm);

3457 3458 3459 3460
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
3461 3462 3463 3464 3465 3466 3467 3468
        }

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

3469
 done:
3470 3471 3472
    qemuMigrationCookieFree(mig);
    rv = 0;

3473
 cleanup:
3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488
    if (event)
        qemuDomainEventQueue(driver, event);
    virObjectUnref(cfg);
    return rv;
}

int
qemuMigrationConfirm(virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *cookiein,
                     int cookieinlen,
                     unsigned int flags,
                     int cancelled)
{
    virQEMUDriverPtr driver = conn->privateData;
3489
    qemuMigrationJobPhase phase;
3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503
    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);
3504 3505
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
3506 3507 3508 3509 3510

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

3511 3512 3513
    qemuMigrationJobFinish(driver, vm);
    if (!virDomainObjIsActive(vm) &&
        (!vm->persistent || (flags & VIR_MIGRATE_UNDEFINE_SOURCE))) {
3514 3515 3516 3517 3518
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
        qemuDomainRemoveInactive(driver, vm);
    }

3519
 cleanup:
3520
    qemuDomObjEndAPI(&vm);
3521 3522 3523 3524 3525
    virObjectUnref(cfg);
    return ret;
}


3526 3527
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3528
    MIGRATION_DEST_CONNECT_HOST,
3529
    MIGRATION_DEST_UNIX,
3530
    MIGRATION_DEST_FD,
3531
};
3532

3533 3534 3535 3536
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3537

3538 3539 3540 3541 3542 3543
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3544
            const char *protocol;
3545 3546 3547 3548 3549
            const char *name;
            int port;
        } host;

        struct {
3550
            char *file;
3551 3552
            int sock;
        } unix_socket;
3553 3554 3555 3556 3557

        struct {
            int qemu;
            int local;
        } fd;
3558 3559 3560 3561 3562 3563 3564
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3565 3566 3567

#define TUNNEL_SEND_BUF_SIZE 65536

3568 3569 3570 3571 3572 3573 3574
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3575 3576
    int wakeupRecvFD;
    int wakeupSendFD;
3577 3578 3579
};

static void qemuMigrationIOFunc(void *arg)
3580
{
3581
    qemuMigrationIOThreadPtr data = arg;
3582 3583 3584 3585 3586 3587 3588
    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);
3589

3590
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3591
        goto abrt;
3592

3593 3594 3595
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3596
    for (;;) {
3597 3598 3599 3600 3601 3602 3603 3604 3605 3606
        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;
3607
            virReportSystemError(errno, "%s",
3608 3609
                                 _("poll failed in migration tunnel"));
            goto abrt;
3610
        }
3611 3612 3613 3614 3615 3616 3617

        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");
3618
            break;
3619
        }
3620

3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636
        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;
            }
3637 3638
        }

3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655
        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;
            }
        }
    }
3656

3657 3658
    if (virStreamFinish(data->st) < 0)
        goto error;
3659

3660 3661
    VIR_FREE(buffer);

3662 3663
    return;

3664
 abrt:
3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3676
 error:
3677 3678
    virCopyLastError(&data->err);
    virResetLastError();
3679
    VIR_FREE(buffer);
3680 3681 3682 3683 3684 3685 3686
}


static qemuMigrationIOThreadPtr
qemuMigrationStartTunnel(virStreamPtr st,
                         int sock)
{
3687 3688
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3689

3690 3691 3692 3693
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3694 3695
    }

3696
    if (VIR_ALLOC(io) < 0)
3697
        goto error;
3698

3699 3700
    io->st = st;
    io->sock = sock;
3701 3702
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3703 3704 3705 3706 3707 3708

    if (virThreadCreate(&io->thread, true,
                        qemuMigrationIOFunc,
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3709
        goto error;
3710 3711 3712
    }

    return io;
3713

3714
 error:
3715 3716 3717 3718
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3719 3720 3721
}

static int
3722
qemuMigrationStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3723 3724
{
    int rv = -1;
3725 3726 3727 3728 3729 3730 3731 3732 3733
    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;
    }

3734 3735 3736 3737
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3738 3739 3740 3741
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3742 3743 3744 3745 3746 3747
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3748
 cleanup:
3749 3750
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3751 3752
    VIR_FREE(io);
    return rv;
3753 3754
}

3755
static int
3756
qemuMigrationConnect(virQEMUDriverPtr driver,
3757 3758 3759 3760 3761 3762 3763 3764 3765
                     virDomainObjPtr vm,
                     qemuMigrationSpecPtr spec)
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3766
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3767 3768 3769 3770 3771 3772 3773 3774 3775
        return -1;

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

    if (virSecurityManagerSetSocketLabel(driver->securityManager, vm->def) < 0)
        goto cleanup;
    if (virNetSocketNewConnectTCP(host, port, &sock) == 0) {
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
3776
        virObjectUnref(sock);
3777 3778 3779 3780 3781
    }
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0 ||
        spec->dest.fd.qemu == -1)
        goto cleanup;

3782 3783 3784 3785 3786 3787 3788
    /* 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;
    }

3789 3790
    ret = 0;

3791
 cleanup:
3792 3793 3794 3795 3796 3797
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3798
static int
3799
qemuMigrationRun(virQEMUDriverPtr driver,
3800 3801 3802 3803 3804 3805 3806
                 virDomainObjPtr vm,
                 const char *cookiein,
                 int cookieinlen,
                 char **cookieout,
                 int *cookieoutlen,
                 unsigned long flags,
                 unsigned long resource,
3807
                 qemuMigrationSpecPtr spec,
3808 3809
                 virConnectPtr dconn,
                 const char *graphicsuri)
3810
{
3811
    int ret = -1;
3812 3813
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3814
    qemuMigrationCookiePtr mig = NULL;
3815
    qemuMigrationIOThreadPtr iothread = NULL;
3816
    int fd = -1;
3817
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3818
    virErrorPtr orig_err = NULL;
3819
    unsigned int cookieFlags = 0;
3820
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3821
    int rc;
3822 3823 3824

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
3825
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s",
3826 3827
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3828 3829
              spec, spec->destType, spec->fwdType, dconn,
              NULLSTR(graphicsuri));
3830

3831 3832 3833 3834 3835 3836 3837 3838 3839 3840
    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;
    }

3841 3842
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3843 3844 3845 3846
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3847 3848 3849
        return -1;
    }

3850 3851 3852
    mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                 cookieFlags | QEMU_MIGRATION_COOKIE_GRAPHICS);
    if (!mig)
3853 3854
        goto cleanup;

3855
    if (qemuDomainMigrateGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3856 3857
        VIR_WARN("unable to provide data for graphics client relocation");

3858 3859 3860 3861 3862 3863 3864
    /* this will update migrate_flags on success */
    if (qemuMigrationDriveMirror(driver, vm, mig, spec->dest.host.name,
                                 migrate_speed, &migrate_flags) < 0) {
        /* error reported by helper func */
        goto cleanup;
    }

3865
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3866 3867 3868 3869 3870 3871
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

3872 3873
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
3874 3875 3876
                                    QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3877 3878
    if (qemuMigrationSetAutoConverge(driver, vm,
                                     flags & VIR_MIGRATE_AUTO_CONVERGE,
3879 3880 3881
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3882 3883
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
3884 3885 3886
                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3887 3888
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3889 3890
        goto cleanup;

3891 3892 3893 3894
    if (priv->job.asyncAbort) {
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
         * priv->job.asyncAbort will not change */
3895
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
3896
        priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
3897 3898 3899 3900 3901 3902
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
        goto cleanup;
    }

3903 3904
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
3905

3906 3907
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3908
        qemuMigrationConnect(driver, vm, spec) < 0) {
3909
        goto exit_monitor;
3910
    }
3911

3912 3913
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3914 3915
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
3916
            goto exit_monitor;
M
Michael R. Hines 已提交
3917
        }
3918
        ret = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
3919
                                       spec->dest.host.protocol,
3920 3921 3922 3923
                                       spec->dest.host.name,
                                       spec->dest.host.port);
        break;

3924 3925 3926 3927
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3928
    case MIGRATION_DEST_UNIX:
3929
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX)) {
3930 3931 3932 3933 3934 3935 3936 3937 3938
            ret = qemuMonitorMigrateToUnix(priv->mon, migrate_flags,
                                           spec->dest.unix_socket.file);
        } else {
            const char *args[] = {
                "nc", "-U", spec->dest.unix_socket.file, NULL
            };
            ret = qemuMonitorMigrateToCommand(priv->mon, migrate_flags, args);
        }
        break;
3939 3940

    case MIGRATION_DEST_FD:
3941
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3942
            fd = spec->dest.fd.local;
3943 3944
            spec->dest.fd.local = -1;
        }
3945 3946 3947 3948
        ret = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                     spec->dest.fd.qemu);
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3949
    }
3950 3951
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3952
    if (ret < 0)
3953 3954
        goto cleanup;
    ret = -1;
3955 3956 3957 3958

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

3959 3960 3961 3962 3963 3964 3965 3966
    if (spec->destType == MIGRATION_DEST_UNIX) {
        /* It is also possible that the migrate didn't fail initially, but
         * rather failed later on.  Check its status before waiting for a
         * connection from qemu which may never be initiated.
         */
        if (qemuMigrationUpdateJobStatus(driver, vm, _("migration job"),
                                         QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            goto cancel;
3967

3968 3969 3970 3971 3972 3973 3974
        while ((fd = accept(spec->dest.unix_socket.sock, NULL, NULL)) < 0) {
            if (errno == EAGAIN || errno == EINTR)
                continue;
            virReportSystemError(errno, "%s",
                                 _("failed to accept connection from qemu"));
            goto cancel;
        }
3975 3976
    }

3977 3978
    if (spec->fwdType != MIGRATION_FWD_DIRECT &&
        !(iothread = qemuMigrationStartTunnel(spec->fwd.stream, fd)))
3979 3980
        goto cancel;

3981 3982 3983 3984 3985 3986
    rc = qemuMigrationWaitForCompletion(driver, vm,
                                        QEMU_ASYNC_JOB_MIGRATION_OUT,
                                        dconn, abort_on_error);
    if (rc == -2)
        goto cancel;
    else if (rc == -1)
3987
        goto cleanup;
3988

3989 3990 3991 3992 3993 3994 3995 3996 3997
    /* When migration completed, QEMU will have paused the
     * CPUs for us, but unless we're using the JSON monitor
     * we won't have been notified of this, so might still
     * think we're running. For v2 protocol this doesn't
     * matter because we'll kill the VM soon, but for v3
     * this is important because we stay paused until the
     * confirm3 step, but need to release the lock state
     */
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
3998 3999
        if (qemuMigrationSetOffline(driver, vm) < 0) {
            priv->job.current->type = VIR_DOMAIN_JOB_FAILED;
4000
            goto cleanup;
4001
        }
4002 4003
    }

4004
    ret = 0;
4005

4006
 cleanup:
4007 4008 4009
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

4010
    /* cancel any outstanding NBD jobs */
4011
    if (mig)
4012
        ignore_value(qemuMigrationCancelDriveMirror(mig, driver, vm));
4013

4014
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
4015
        if (iothread && qemuMigrationStopTunnel(iothread, ret < 0) < 0)
4016
            ret = -1;
4017
        VIR_FORCE_CLOSE(fd);
4018
    }
4019

4020
    if (priv->job.completed) {
4021
        qemuDomainJobInfoUpdateTime(priv->job.completed);
4022 4023
        qemuDomainJobInfoUpdateDowntime(priv->job.completed);
    }
4024

4025 4026 4027
    if (priv->job.current->type == VIR_DOMAIN_JOB_UNBOUNDED)
        priv->job.current->type = VIR_DOMAIN_JOB_FAILED;

4028 4029
    cookieFlags |= QEMU_MIGRATION_COOKIE_NETWORK |
                   QEMU_MIGRATION_COOKIE_STATS;
4030 4031
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookieFlags |= QEMU_MIGRATION_COOKIE_PERSISTENT;
4032
    if (ret == 0 &&
4033 4034
        qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
4035
        VIR_WARN("Unable to encode migration cookie");
4036
    }
4037

4038 4039
    qemuMigrationCookieFree(mig);

4040 4041 4042 4043 4044
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4045 4046
    return ret;

4047 4048 4049 4050
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;

4051
 cancel:
4052 4053
    orig_err = virSaveLastError();

4054
    if (virDomainObjIsActive(vm)) {
4055 4056
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
4057
            qemuMonitorMigrateCancel(priv->mon);
4058
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
4059
        }
4060
    }
4061 4062 4063
    goto cleanup;
}

4064
/* Perform migration using QEMU's native migrate support,
4065 4066
 * not encrypted obviously
 */
4067
static int doNativeMigrate(virQEMUDriverPtr driver,
4068 4069 4070 4071 4072 4073 4074
                           virDomainObjPtr vm,
                           const char *uri,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4075
                           unsigned long resource,
4076 4077
                           virConnectPtr dconn,
                           const char *graphicsuri)
4078
{
4079
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
4080
    virURIPtr uribits = NULL;
4081
    int ret = -1;
4082 4083 4084
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
4085 4086
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "graphicsuri=%s",
4087
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
4088 4089
              cookieout, cookieoutlen, flags, resource,
              NULLSTR(graphicsuri));
4090

4091
    if (!(uribits = qemuMigrationParseURI(uri, NULL)))
4092 4093
        return -1;

4094 4095 4096 4097 4098 4099 4100
    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
4101 4102 4103 4104 4105
    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"));
4106
            goto cleanup;
M
Michael R. Hines 已提交
4107
        }
4108
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
4109 4110 4111
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
4112
            goto cleanup;
M
Michael R. Hines 已提交
4113 4114 4115 4116 4117
        }
    }

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
        STRNEQ(uribits->scheme, "rdma"))
4118 4119
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
4120
        spec.destType = MIGRATION_DEST_HOST;
4121
    spec.dest.host.protocol = uribits->scheme;
4122 4123 4124
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
4125

4126
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4127 4128
                           cookieoutlen, flags, resource, &spec, dconn,
                           graphicsuri);
4129 4130 4131 4132

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

4133
 cleanup:
4134
    virURIFree(uribits);
4135 4136 4137 4138 4139

    return ret;
}


4140
static int doTunnelMigrate(virQEMUDriverPtr driver,
4141 4142 4143 4144 4145 4146 4147
                           virDomainObjPtr vm,
                           virStreamPtr st,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4148
                           unsigned long resource,
4149 4150
                           virConnectPtr dconn,
                           const char *graphicsuri)
4151 4152
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
4153
    virNetSocketPtr sock = NULL;
4154 4155
    int ret = -1;
    qemuMigrationSpec spec;
4156
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4157 4158

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
4159 4160
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "graphicsuri=%s",
4161
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
4162 4163
              cookieout, cookieoutlen, flags, resource,
              NULLSTR(graphicsuri));
4164

4165 4166 4167
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
        !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX) &&
        !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_EXEC)) {
4168 4169
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Source qemu is too old to support tunnelled migration"));
4170
        virObjectUnref(cfg);
4171
        return -1;
4172 4173 4174 4175 4176
    }

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

4177
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD)) {
4178 4179 4180 4181 4182 4183
        int fds[2];

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

4184
        if (pipe2(fds, O_CLOEXEC) == 0) {
4185 4186 4187 4188
            spec.dest.fd.qemu = fds[1];
            spec.dest.fd.local = fds[0];
        }
        if (spec.dest.fd.qemu == -1 ||
4189
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
4190
                                              spec.dest.fd.qemu) < 0) {
4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201
            virReportSystemError(errno, "%s",
                        _("cannot create pipe for tunnelled migration"));
            goto cleanup;
        }
    } else {
        spec.destType = MIGRATION_DEST_UNIX;
        spec.dest.unix_socket.sock = -1;
        spec.dest.unix_socket.file = NULL;

        if (virAsprintf(&spec.dest.unix_socket.file,
                        "%s/qemu.tunnelmigrate.src.%s",
4202
                        cfg->libDir, vm->def->name) < 0)
4203 4204 4205
            goto cleanup;

        if (virNetSocketNewListenUNIX(spec.dest.unix_socket.file, 0700,
4206
                                      cfg->user, cfg->group,
4207 4208 4209 4210 4211 4212 4213
                                      &sock) < 0 ||
            virNetSocketListen(sock, 1) < 0)
            goto cleanup;

        spec.dest.unix_socket.sock = virNetSocketGetFD(sock);
    }

4214
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4215 4216
                           cookieoutlen, flags, resource, &spec, dconn,
                           graphicsuri);
4217

4218
 cleanup:
4219 4220 4221 4222
    if (spec.destType == MIGRATION_DEST_FD) {
        VIR_FORCE_CLOSE(spec.dest.fd.qemu);
        VIR_FORCE_CLOSE(spec.dest.fd.local);
    } else {
4223
        virObjectUnref(sock);
4224 4225
        VIR_FREE(spec.dest.unix_socket.file);
    }
4226

4227
    virObjectUnref(cfg);
4228 4229 4230 4231
    return ret;
}


4232 4233 4234 4235
/* 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 */
4236
static int doPeer2PeerMigrate2(virQEMUDriverPtr driver,
4237
                               virConnectPtr sconn ATTRIBUTE_UNUSED,
4238 4239
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
4240
                               const char *dconnuri,
4241 4242 4243
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
4244 4245 4246
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
4247
    char *cookie = NULL;
4248 4249 4250
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
4251
    bool cancelled;
4252
    virStreamPtr st = NULL;
4253
    unsigned long destflags;
4254

4255
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
4256
              "flags=%lx, dname=%s, resource=%lu",
4257 4258
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
4259

4260 4261 4262 4263 4264
    /* 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,
4265 4266
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
4267 4268 4269 4270 4271
        return -1;

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

4272 4273
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4274

4275 4276 4277 4278 4279 4280 4281 4282 4283 4284
    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;

4285
        qemuDomainObjEnterRemote(vm);
4286
        ret = dconn->driver->domainMigratePrepareTunnel
4287
            (dconn, st, destflags, dname, resource, dom_xml);
4288
        qemuDomainObjExitRemote(vm);
4289
    } else {
4290
        qemuDomainObjEnterRemote(vm);
4291 4292
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
4293
             destflags, dname, resource, dom_xml);
4294
        qemuDomainObjExitRemote(vm);
4295 4296 4297
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
4298 4299 4300
        goto cleanup;

    /* the domain may have shutdown or crashed while we had the locks dropped
4301
     * in qemuDomainObjEnterRemote, so check again
4302 4303
     */
    if (!virDomainObjIsActive(vm)) {
4304 4305
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4306 4307 4308
        goto cleanup;
    }

4309 4310
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
4311 4312
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
4313
        cancelled = true;
4314
        orig_err = virSaveLastError();
4315
        goto finish;
4316 4317
    }

4318 4319 4320 4321
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
4322
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4323
    if (flags & VIR_MIGRATE_TUNNELLED)
4324 4325
        ret = doTunnelMigrate(driver, vm, st,
                              NULL, 0, NULL, NULL,
4326
                              flags, resource, dconn, NULL);
4327 4328 4329 4330
    else
        ret = doNativeMigrate(driver, vm, uri_out,
                              cookie, cookielen,
                              NULL, NULL, /* No out cookie with v2 migration */
4331
                              flags, resource, dconn, NULL);
4332 4333 4334 4335

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4336

4337 4338 4339
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4340
    cancelled = ret < 0;
4341

4342
 finish:
4343 4344 4345 4346
    /* 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.
     */
4347
    dname = dname ? dname : vm->def->name;
4348
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4349
    qemuDomainObjEnterRemote(vm);
4350
    ddomain = dconn->driver->domainMigrateFinish2
4351
        (dconn, dname, cookie, cookielen,
4352
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4353
    qemuDomainObjExitRemote(vm);
4354 4355
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4356

4357
 cleanup:
4358
    if (ddomain) {
4359
        virObjectUnref(ddomain);
4360 4361 4362 4363
        ret = 0;
    } else {
        ret = -1;
    }
4364

4365
    virObjectUnref(st);
4366 4367 4368 4369 4370 4371

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4372
    VIR_FREE(cookie);
4373 4374

    return ret;
4375 4376 4377
}


4378 4379 4380 4381
/* 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 */
4382 4383 4384 4385 4386 4387 4388 4389 4390
static int
doPeer2PeerMigrate3(virQEMUDriverPtr driver,
                    virConnectPtr sconn,
                    virConnectPtr dconn,
                    const char *dconnuri,
                    virDomainObjPtr vm,
                    const char *xmlin,
                    const char *dname,
                    const char *uri,
4391
                    const char *graphicsuri,
4392
                    const char *listenAddress,
4393 4394 4395
                    unsigned long long bandwidth,
                    bool useParams,
                    unsigned long flags)
4396 4397 4398 4399 4400 4401 4402 4403 4404 4405
{
    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;
4406
    bool cancelled = true;
4407
    virStreamPtr st = NULL;
4408
    unsigned long destflags;
4409 4410 4411 4412 4413
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4414 4415
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
              "bandwidth=%llu, useParams=%d, flags=%lx",
4416
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4417 4418
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
              NULLSTR(listenAddress), bandwidth, useParams, flags);
4419

4420 4421 4422 4423 4424
    /* 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.  */

4425 4426
    dom_xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
                                      &cookieout, &cookieoutlen, flags);
4427 4428 4429
    if (!dom_xml)
        goto cleanup;

4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449
    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;
4450 4451 4452 4453 4454 4455

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4456 4457 4458 4459 4460
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4461 4462
    }

4463 4464 4465
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4466 4467
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4468

4469 4470 4471 4472 4473 4474 4475 4476 4477
    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;

4478
        qemuDomainObjEnterRemote(vm);
4479 4480 4481 4482 4483 4484 4485 4486 4487
        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);
        }
4488
        qemuDomainObjExitRemote(vm);
4489
    } else {
4490
        qemuDomainObjEnterRemote(vm);
4491 4492 4493 4494 4495 4496 4497 4498 4499
        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);
        }
4500
        qemuDomainObjExitRemote(vm);
4501 4502 4503 4504 4505
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

L
liguang 已提交
4506 4507 4508 4509
    if (flags & VIR_MIGRATE_OFFLINE) {
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4510
        cancelled = false;
L
liguang 已提交
4511 4512 4513
        goto finish;
    }

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

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4554
    if (ret < 0) {
4555
        orig_err = virSaveLastError();
4556 4557 4558 4559
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4560 4561 4562 4563

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4564
    cancelled = ret < 0;
4565

4566
 finish:
4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578
    /*
     * 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;
4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601

    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);
    }
4602 4603
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4604

4605 4606 4607 4608 4609 4610 4611
    /* 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.
4612
     */
4613
    cancelled = ddomain == NULL;
4614

4615 4616 4617 4618 4619 4620
    /* 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();

4621 4622 4623 4624
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4625
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4626 4627 4628 4629 4630
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4631 4632 4633
    ret = qemuMigrationConfirmPhase(driver, sconn, vm,
                                    cookiein, cookieinlen,
                                    flags, cancelled);
4634 4635 4636 4637
    /* 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.
     */
4638 4639 4640
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4641 4642 4643

 cleanup:
    if (ddomain) {
4644
        virObjectUnref(ddomain);
4645 4646 4647 4648 4649
        ret = 0;
    } else {
        ret = -1;
    }

4650
    virObjectUnref(st);
4651 4652 4653 4654 4655 4656 4657 4658

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4659
    virTypedParamsFree(params, nparams);
4660 4661 4662 4663
    return ret;
}


4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4676
static int doPeer2PeerMigrate(virQEMUDriverPtr driver,
4677
                              virConnectPtr sconn,
4678
                              virDomainObjPtr vm,
4679
                              const char *xmlin,
4680
                              const char *dconnuri,
4681
                              const char *uri,
4682
                              const char *graphicsuri,
4683
                              const char *listenAddress,
4684 4685
                              unsigned long flags,
                              const char *dname,
4686 4687
                              unsigned long resource,
                              bool *v3proto)
4688 4689 4690 4691
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4692
    virErrorPtr orig_err = NULL;
4693
    bool offline = false;
4694
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4695
    bool useParams;
4696

4697
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4698 4699
              "uri=%s, graphicsuri=%s, listenAddress=%s, flags=%lx, "
              "dname=%s, resource=%lu",
4700
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4701 4702
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
              flags, NULLSTR(dname), resource);
4703 4704 4705 4706 4707

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

4708
    qemuDomainObjEnterRemote(vm);
4709
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4710
    qemuDomainObjExitRemote(vm);
4711
    if (dconn == NULL) {
4712
        virReportError(VIR_ERR_OPERATION_FAILED,
4713 4714
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4715
        virObjectUnref(cfg);
4716 4717 4718
        return -1;
    }

4719 4720
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4721 4722
        goto cleanup;

4723
    qemuDomainObjEnterRemote(vm);
4724 4725
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4726
        /* v3proto reflects whether the caller used Perform3, but with
4727
         * p2p migrate, regardless of whether Perform2 or Perform3
4728 4729 4730 4731
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4732 4733
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
L
liguang 已提交
4734 4735 4736
    if (flags & VIR_MIGRATE_OFFLINE)
        offline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                           VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4737
    qemuDomainObjExitRemote(vm);
4738

4739
    if (!p2p) {
4740 4741
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4742 4743 4744
        goto cleanup;
    }

4745 4746
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4747
    if (!useParams && graphicsuri) {
4748 4749 4750 4751 4752 4753
        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
4754 4755 4756 4757 4758 4759 4760
    if (flags & VIR_MIGRATE_OFFLINE && !offline) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4761
    /* domain may have been stopped while we were talking to remote daemon */
L
liguang 已提交
4762
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4763 4764
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4765 4766 4767
        goto cleanup;
    }

4768 4769 4770 4771 4772 4773 4774
    /* 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;

4775 4776
    if (*v3proto) {
        ret = doPeer2PeerMigrate3(driver, sconn, dconn, dconnuri, vm, xmlin,
4777 4778
                                  dname, uri, graphicsuri, listenAddress,
                                  resource, useParams, flags);
4779
    } else {
4780
        ret = doPeer2PeerMigrate2(driver, sconn, dconn, vm,
4781
                                  dconnuri, flags, dname, resource);
4782
    }
4783

4784
 cleanup:
4785
    orig_err = virSaveLastError();
4786
    qemuDomainObjEnterRemote(vm);
4787
    virObjectUnref(dconn);
4788
    qemuDomainObjExitRemote(vm);
4789 4790 4791 4792
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4793
    virObjectUnref(cfg);
4794 4795 4796 4797
    return ret;
}


4798 4799 4800 4801 4802 4803
/*
 * 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
4804
qemuMigrationPerformJob(virQEMUDriverPtr driver,
4805 4806 4807 4808 4809
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dconnuri,
                        const char *uri,
4810
                        const char *graphicsuri,
4811
                        const char *listenAddress,
4812
                        const char *cookiein,
4813 4814 4815 4816 4817 4818 4819
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4820
{
4821
    virObjectEventPtr event = NULL;
4822
    int ret = -1;
4823
    virErrorPtr orig_err = NULL;
4824
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4825
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
4826

4827
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4828 4829
        goto cleanup;

L
liguang 已提交
4830
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4831 4832
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
4833 4834 4835
        goto endjob;
    }

4836
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
4837
        goto endjob;
4838

4839
    if (!(flags & VIR_MIGRATE_UNSAFE) && !qemuMigrationIsSafe(vm->def))
4840
        goto endjob;
4841

4842
    qemuMigrationStoreDomainState(vm);
4843 4844

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4845
        ret = doPeer2PeerMigrate(driver, conn, vm, xmlin,
4846 4847
                                 dconnuri, uri, graphicsuri, listenAddress,
                                 flags, dname, resource, &v3proto);
4848
    } else {
4849 4850 4851
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
        ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                              cookieout, cookieoutlen,
4852
                              flags, resource, NULL, NULL);
4853
    }
4854 4855
    if (ret < 0)
        goto endjob;
4856

4857 4858 4859 4860
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4861
    if (!v3proto) {
4862 4863
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4864
        virDomainAuditStop(vm, "migrated");
4865
        event = virDomainEventLifecycleNewFromObj(vm,
4866 4867
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4868 4869
    }

4870
 endjob:
4871 4872 4873
    if (ret < 0)
        orig_err = virSaveLastError();

4874
    if (qemuMigrationRestoreDomainState(conn, vm)) {
4875
        event = virDomainEventLifecycleNewFromObj(vm,
4876 4877 4878
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
4879

4880 4881 4882 4883
    qemuMigrationJobFinish(driver, vm);
    if (!virDomainObjIsActive(vm) &&
        (!vm->persistent ||
         (ret == 0 && (flags & VIR_MIGRATE_UNDEFINE_SOURCE)))) {
4884
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
4885
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4886
        qemuDomainRemoveInactive(driver, vm);
4887 4888
    }

4889 4890 4891 4892 4893
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4894
 cleanup:
4895
    qemuDomObjEndAPI(&vm);
4896 4897
    if (event)
        qemuDomainEventQueue(driver, event);
4898
    virObjectUnref(cfg);
4899 4900 4901 4902 4903 4904 4905
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4906
qemuMigrationPerformPhase(virQEMUDriverPtr driver,
4907 4908 4909
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *uri,
4910
                          const char *graphicsuri,
4911 4912 4913 4914 4915 4916 4917
                          const char *cookiein,
                          int cookieinlen,
                          char **cookieout,
                          int *cookieoutlen,
                          unsigned long flags,
                          unsigned long resource)
{
4918
    virObjectEventPtr event = NULL;
4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929
    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);
4930 4931
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
4932 4933 4934

    ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                          cookieout, cookieoutlen,
4935
                          flags, resource, NULL, graphicsuri);
4936

4937 4938 4939 4940 4941
    if (ret < 0) {
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
4942
        }
4943
        goto endjob;
4944
    }
4945 4946 4947

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4948 4949
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                             qemuMigrationCleanup) < 0)
4950 4951
        goto endjob;

4952
 endjob:
4953
    if (ret < 0)
4954
        qemuMigrationJobFinish(driver, vm);
4955
    else
4956 4957
        qemuMigrationJobContinue(vm);
    if (!virDomainObjIsActive(vm) && !vm->persistent)
4958
        qemuDomainRemoveInactive(driver, vm);
4959

4960
 cleanup:
4961
    qemuDomObjEndAPI(&vm);
4962 4963 4964 4965 4966
    if (event)
        qemuDomainEventQueue(driver, event);
    return ret;
}

4967
int
4968
qemuMigrationPerform(virQEMUDriverPtr driver,
4969 4970 4971 4972 4973
                     virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *xmlin,
                     const char *dconnuri,
                     const char *uri,
4974
                     const char *graphicsuri,
4975
                     const char *listenAddress,
4976 4977 4978 4979 4980 4981 4982 4983 4984 4985
                     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, "
4986
              "uri=%s, graphicsuri=%s, listenAddress=%s"
4987 4988
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
              "flags=%lx, dname=%s, resource=%lu, v3proto=%d",
4989
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4990
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4991 4992
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4993 4994 4995

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4996 4997
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4998 4999 5000 5001
            return -1;
        }

        return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri, uri,
5002 5003
                                       graphicsuri, listenAddress,
                                       cookiein, cookieinlen,
5004 5005
                                       cookieout, cookieoutlen,
                                       flags, dname, resource, v3proto);
5006 5007
    } else {
        if (dconnuri) {
5008 5009
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
5010 5011 5012 5013 5014
            return -1;
        }

        if (v3proto) {
            return qemuMigrationPerformPhase(driver, conn, vm, uri,
5015
                                             graphicsuri,
5016 5017
                                             cookiein, cookieinlen,
                                             cookieout, cookieoutlen,
5018
                                             flags, resource);
5019 5020
        } else {
            return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri,
5021
                                           uri, graphicsuri, listenAddress,
5022
                                           cookiein, cookieinlen,
5023 5024 5025 5026 5027
                                           cookieout, cookieoutlen, flags,
                                           dname, resource, v3proto);
        }
    }
}
5028

5029
static int
5030 5031
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def)
{
5032
    size_t i;
5033 5034 5035 5036 5037
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
5038
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5039
            if (virNetDevVPortProfileAssociate(net->ifname,
5040
                                               virDomainNetGetActualVirtPortProfile(net),
5041
                                               &net->mac,
5042
                                               virDomainNetGetActualDirectDev(net),
5043
                                               -1,
5044
                                               def->uuid,
5045 5046
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
5047 5048 5049
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
5050
                goto err_exit;
5051 5052
            }
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
5053

5054
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
5055 5056 5057 5058
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
5059 5060 5061 5062
        }
        last_good_net = i;
    }

5063
    return 0;
5064

5065
 err_exit:
5066
    for (i = 0; last_good_net != -1 && i < last_good_net; i++) {
5067
        net = def->nets[i];
5068
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5069
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
5070
                                                           virDomainNetGetActualVirtPortProfile(net),
5071
                                                           &net->mac,
5072
                                                           virDomainNetGetActualDirectDev(net),
5073
                                                           -1,
5074
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
5075 5076
        }
    }
5077
    return -1;
5078 5079 5080 5081
}


virDomainPtr
5082
qemuMigrationFinish(virQEMUDriverPtr driver,
5083 5084
                    virConnectPtr dconn,
                    virDomainObjPtr vm,
5085 5086 5087 5088
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
5089
                    unsigned long flags,
5090 5091
                    int retcode,
                    bool v3proto)
5092 5093
{
    virDomainPtr dom = NULL;
5094
    virObjectEventPtr event = NULL;
5095
    bool newVM = true;
5096
    qemuMigrationCookiePtr mig = NULL;
5097
    virErrorPtr orig_err = NULL;
5098
    int cookie_flags = 0;
J
Jiri Denemark 已提交
5099
    qemuDomainObjPrivatePtr priv = vm->privateData;
5100
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5101
    virCapsPtr caps = NULL;
5102
    unsigned short port;
5103

5104
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
5105
              "cookieout=%p, cookieoutlen=%p, flags=%lx, retcode=%d",
5106 5107
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
5108

5109 5110 5111
    port = priv->migrationPort;
    priv->migrationPort = 0;

5112 5113 5114
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

5115
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
5116 5117
        goto cleanup;

5118 5119 5120
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
5121

5122
    qemuDomainCleanupRemove(vm, qemuMigrationPrepareCleanup);
5123
    VIR_FREE(priv->job.completed);
5124

5125
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
5126 5127
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
5128 5129 5130 5131 5132
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookie_flags |= QEMU_MIGRATION_COOKIE_PERSISTENT;

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein,
                                       cookieinlen, cookie_flags)))
5133
        goto endjob;
5134 5135 5136 5137 5138

    /* Did the migration go as planned?  If yes, return the domain
     * object, but if no, clean up the empty qemu process.
     */
    if (retcode == 0) {
L
liguang 已提交
5139
        if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
5140 5141
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("guest unexpectedly quit"));
5142
            goto endjob;
5143 5144
        }

5145 5146 5147 5148 5149
        if (mig->jobInfo) {
            priv->job.completed = mig->jobInfo;
            mig->jobInfo = NULL;
        }

L
liguang 已提交
5150 5151 5152 5153 5154
        if (!(flags & VIR_MIGRATE_OFFLINE)) {
            if (qemuMigrationVPAssociatePortProfiles(vm->def) < 0) {
                qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                VIR_QEMU_PROCESS_STOP_MIGRATED);
                virDomainAuditStop(vm, "failed");
5155
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5156 5157 5158 5159 5160 5161 5162
                                                 VIR_DOMAIN_EVENT_STOPPED,
                                                 VIR_DOMAIN_EVENT_STOPPED_FAILED);
                goto endjob;
            }
            if (mig->network)
                if (qemuDomainMigrateOPDRelocate(driver, vm, mig) < 0)
                    VIR_WARN("unable to provide network data for relocation");
5163
        }
5164

5165 5166
        if (qemuMigrationStopNBDServer(driver, vm, mig) < 0)
            goto endjob;
5167

5168
        if (flags & VIR_MIGRATE_PERSIST_DEST) {
5169
            virDomainDefPtr vmdef;
5170
            if (vm->persistent)
5171
                newVM = false;
5172
            vm->persistent = 1;
5173 5174 5175
            if (mig->persistent)
                vm->newDef = vmdef = mig->persistent;
            else
5176
                vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm);
5177
            if (!vmdef || virDomainSaveConfig(cfg->configDir, vmdef) < 0) {
5178 5179 5180 5181 5182 5183 5184 5185 5186
                /* 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.
                 * Return a NULL dom pointer, and hope that this is a rare
                 * situation and management tools are smart.
                 */
5187 5188

                /*
5189 5190
                 * However, in v3 protocol, the source VM is still available
                 * to restart during confirm() step, so we kill it off now.
5191 5192
                 */
                if (v3proto) {
L
liguang 已提交
5193 5194 5195 5196 5197
                    if (!(flags & VIR_MIGRATE_OFFLINE)) {
                        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                        VIR_QEMU_PROCESS_STOP_MIGRATED);
                        virDomainAuditStop(vm, "failed");
                    }
5198 5199
                    if (newVM)
                        vm->persistent = 0;
5200
                }
A
Alex Jia 已提交
5201
                if (!vmdef)
5202 5203
                    virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                                   _("can't get vmdef"));
5204 5205 5206
                goto endjob;
            }

5207
            event = virDomainEventLifecycleNewFromObj(vm,
5208 5209 5210 5211 5212 5213 5214 5215 5216
                                             VIR_DOMAIN_EVENT_DEFINED,
                                             newVM ?
                                             VIR_DOMAIN_EVENT_DEFINED_ADDED :
                                             VIR_DOMAIN_EVENT_DEFINED_UPDATED);
            if (event)
                qemuDomainEventQueue(driver, event);
            event = NULL;
        }

L
liguang 已提交
5217
        if (!(flags & VIR_MIGRATE_PAUSED) && !(flags & VIR_MIGRATE_OFFLINE)) {
5218 5219 5220 5221
            /* 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
             */
J
Jiri Denemark 已提交
5222
            if (qemuProcessStartCPUs(driver, vm, dconn,
5223 5224
                                     VIR_DOMAIN_RUNNING_MIGRATED,
                                     QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5225
                if (virGetLastError() == NULL)
5226 5227
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   "%s", _("resume operation failed"));
5228 5229 5230 5231 5232
                /* Need to save the current error, in case shutting
                 * down the process overwrites it
                 */
                orig_err = virSaveLastError();

5233 5234 5235 5236 5237 5238 5239 5240 5241
                /*
                 * 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) {
5242 5243
                    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                    VIR_QEMU_PROCESS_STOP_MIGRATED);
5244
                    virDomainAuditStop(vm, "failed");
5245
                    event = virDomainEventLifecycleNewFromObj(vm,
5246 5247 5248
                                                     VIR_DOMAIN_EVENT_STOPPED,
                                                     VIR_DOMAIN_EVENT_STOPPED_FAILED);
                }
5249 5250
                goto endjob;
            }
5251 5252 5253 5254
            if (priv->job.completed) {
                qemuDomainJobInfoUpdateTime(priv->job.completed);
                qemuDomainJobInfoUpdateDowntime(priv->job.completed);
            }
5255 5256
        }

5257
        dom = virGetDomain(dconn, vm->def->name, vm->def->uuid);
5258

L
liguang 已提交
5259
        if (!(flags & VIR_MIGRATE_OFFLINE)) {
5260
            event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5261 5262 5263 5264 5265 5266 5267
                                             VIR_DOMAIN_EVENT_RESUMED,
                                             VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
            if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
                virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                                     VIR_DOMAIN_PAUSED_USER);
                if (event)
                    qemuDomainEventQueue(driver, event);
5268
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5269 5270 5271
                                                 VIR_DOMAIN_EVENT_SUSPENDED,
                                                 VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);
            }
5272
        }
L
liguang 已提交
5273 5274

        if (virDomainObjIsActive(vm) &&
5275
            virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
5276 5277 5278
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto endjob;
        }
5279 5280 5281

        /* Guest is successfully running, so cancel previous auto destroy */
        qemuProcessAutoDestroyRemove(driver, vm);
L
liguang 已提交
5282
    } else if (!(flags & VIR_MIGRATE_OFFLINE)) {
5283 5284
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
5285
        virDomainAuditStop(vm, "failed");
5286
        event = virDomainEventLifecycleNewFromObj(vm,
5287 5288 5289 5290
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_FAILED);
    }

5291 5292
    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen,
                                QEMU_MIGRATION_COOKIE_STATS) < 0)
5293 5294
        VIR_WARN("Unable to encode migration cookie");

5295
 endjob:
5296 5297
    qemuMigrationJobFinish(driver, vm);
    if (!vm->persistent && !virDomainObjIsActive(vm))
5298
        qemuDomainRemoveInactive(driver, vm);
5299

5300
 cleanup:
5301
    virPortAllocatorRelease(driver->migrationPorts, port);
5302 5303 5304 5305
    if (priv->mon)
        qemuMonitorSetDomainLog(priv->mon, -1);
    VIR_FREE(priv->origname);
    qemuDomObjEndAPI(&vm);
5306 5307
    if (event)
        qemuDomainEventQueue(driver, event);
5308
    qemuMigrationCookieFree(mig);
5309 5310 5311 5312
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5313
    virObjectUnref(caps);
5314
    virObjectUnref(cfg);
5315 5316
    return dom;
}
5317

5318

5319
/* Helper function called while vm is active.  */
5320
int
5321
qemuMigrationToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
5322 5323
                    int fd, off_t offset, const char *path,
                    const char *compressor,
E
Eric Blake 已提交
5324
                    bool bypassSecurityDriver,
5325
                    qemuDomainAsyncJob asyncJob)
5326 5327 5328
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5329
    int ret = -1;
5330
    bool restoreLabel = false;
5331 5332
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5333
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5334
    char *errbuf = NULL;
5335
    virErrorPtr orig_err = NULL;
5336 5337 5338 5339 5340

    /* 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,
5341 5342
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5343 5344
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5345
    }
5346

5347 5348 5349 5350 5351 5352 5353
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5354
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5355
        (!compressor || pipe(pipeFD) == 0)) {
5356
        /* All right! We can use fd migration, which means that qemu
5357 5358 5359
         * 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.  */
5360
        if (virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
5361
                                              compressor ? pipeFD[1] : fd) < 0)
5362
            goto cleanup;
5363 5364 5365
        bypassSecurityDriver = true;
    } else {
        /* Phooey - we have to fall back on exec migration, where qemu
E
Eric Blake 已提交
5366 5367
         * has to popen() the file by name, and block devices have to be
         * given cgroup ACL permission.  We might also stumble on
5368 5369
         * a race present in some qemu versions where it does a wait()
         * that botches pclose.  */
5370 5371
        if (virCgroupHasController(priv->cgroup,
                                   VIR_CGROUP_CONTROLLER_DEVICES)) {
5372 5373 5374 5375
            int rv = virCgroupAllowDevicePath(priv->cgroup, path,
                                              VIR_CGROUP_DEVICE_RW);
            virDomainAuditCgroupPath(vm, priv->cgroup, "allow", path, "rw", rv == 0);
            if (rv == 1) {
E
Eric Blake 已提交
5376
                /* path was not a device, no further need for cgroup */
5377
            } else if (rv < 0) {
5378 5379
                goto cleanup;
            }
5380
        }
5381 5382
        if ((!bypassSecurityDriver) &&
            virSecurityManagerSetSavedStateLabel(driver->securityManager,
5383
                                                 vm->def, path) < 0)
5384
            goto cleanup;
5385
        restoreLabel = true;
5386 5387
    }

5388
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5389 5390
        goto cleanup;

5391 5392 5393
    if (!compressor) {
        const char *args[] = { "cat", NULL };

5394
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410
            priv->monConfig->type == VIR_DOMAIN_CHR_TYPE_UNIX) {
            rc = qemuMonitorMigrateToFd(priv->mon,
                                        QEMU_MONITOR_MIGRATE_BACKGROUND,
                                        fd);
        } else {
            rc = qemuMonitorMigrateToFile(priv->mon,
                                          QEMU_MONITOR_MIGRATE_BACKGROUND,
                                          args, path, offset);
        }
    } else {
        const char *prog = compressor;
        const char *args[] = {
            prog,
            "-c",
            NULL
        };
5411 5412 5413 5414
        if (pipeFD[0] != -1) {
            cmd = virCommandNewArgs(args);
            virCommandSetInputFD(cmd, pipeFD[0]);
            virCommandSetOutputFD(cmd, &fd);
5415 5416
            virCommandSetErrorBuffer(cmd, &errbuf);
            virCommandDoAsyncIO(cmd);
5417 5418 5419
            if (virSetCloseExec(pipeFD[1]) < 0) {
                virReportSystemError(errno, "%s",
                                     _("Unable to set cloexec flag"));
5420
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5421 5422 5423
                goto cleanup;
            }
            if (virCommandRunAsync(cmd, NULL) < 0) {
5424
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5425 5426 5427 5428 5429 5430 5431
                goto cleanup;
            }
            rc = qemuMonitorMigrateToFd(priv->mon,
                                        QEMU_MONITOR_MIGRATE_BACKGROUND,
                                        pipeFD[1]);
            if (VIR_CLOSE(pipeFD[0]) < 0 ||
                VIR_CLOSE(pipeFD[1]) < 0)
5432
                VIR_WARN("failed to close intermediate pipe");
5433 5434 5435 5436 5437
        } else {
            rc = qemuMonitorMigrateToFile(priv->mon,
                                          QEMU_MONITOR_MIGRATE_BACKGROUND,
                                          args, path, offset);
        }
5438
    }
5439
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5440
        goto cleanup;
5441 5442 5443
    if (rc < 0)
        goto cleanup;

5444
    rc = qemuMigrationWaitForCompletion(driver, vm, asyncJob, NULL, false);
5445

5446 5447 5448 5449
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5450 5451
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5452
                qemuMonitorMigrateCancel(priv->mon);
5453
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5454 5455
            }
        }
5456
        goto cleanup;
5457
    }
5458

5459 5460 5461
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5462 5463
    ret = 0;

5464
 cleanup:
5465 5466 5467
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5468
    /* Restore max migration bandwidth */
5469 5470
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5471 5472
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5473
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5474 5475
    }

5476 5477
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5478 5479 5480 5481 5482
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5483 5484
    if (restoreLabel && (!bypassSecurityDriver) &&
        virSecurityManagerRestoreSavedStateLabel(driver->securityManager,
5485
                                                 vm->def, path) < 0)
5486 5487
        VIR_WARN("failed to restore save state label on %s", path);

5488 5489
    if (virCgroupHasController(priv->cgroup,
                               VIR_CGROUP_CONTROLLER_DEVICES)) {
5490 5491 5492
        int rv = virCgroupDenyDevicePath(priv->cgroup, path,
                                         VIR_CGROUP_DEVICE_RWM);
        virDomainAuditCgroupPath(vm, priv->cgroup, "deny", path, "rwm", rv == 0);
5493
    }
5494 5495 5496 5497 5498 5499

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

5500 5501
    return ret;
}
5502 5503

int
5504
qemuMigrationJobStart(virQEMUDriverPtr driver,
5505
                      virDomainObjPtr vm,
5506
                      qemuDomainAsyncJob job)
5507 5508 5509
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

5510
    if (qemuDomainObjBeginAsyncJob(driver, vm, job) < 0)
5511 5512
        return -1;

5513
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5514
        qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
5515
    } else {
5516 5517 5518
        qemuDomainObjSetAsyncJobMask(vm, (QEMU_JOB_DEFAULT_MASK |
                                          JOB_MASK(QEMU_JOB_SUSPEND) |
                                          JOB_MASK(QEMU_JOB_MIGRATION_OP)));
5519
    }
5520

J
Jiri Denemark 已提交
5521
    priv->job.current->type = VIR_DOMAIN_JOB_UNBOUNDED;
5522 5523 5524 5525 5526

    return 0;
}

void
5527
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5528
                         virDomainObjPtr vm,
5529
                         qemuMigrationJobPhase phase)
5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543
{
    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
5544
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5545
                           virDomainObjPtr vm,
5546
                           qemuMigrationJobPhase phase)
5547 5548 5549 5550
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5551
void
5552 5553
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5554
    qemuDomainObjReleaseAsyncJob(vm);
5555 5556 5557 5558
}

bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
5559
                         qemuDomainAsyncJob job)
5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570
{
    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");

5571
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5572 5573 5574 5575 5576
        return false;
    }
    return true;
}

5577
void
5578
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5579
{
5580
    qemuDomainObjEndAsyncJob(driver, vm);
5581
}