qemu_migration.c 183.6 KB
Newer Older
1 2 3
/*
 * qemu_migration.c: QEMU migration handling
 *
4
 * Copyright (C) 2006-2014 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 49
#include "datatypes.h"
#include "fdstream.h"
50
#include "viruuid.h"
51
#include "virtime.h"
52
#include "locking/domain_lock.h"
53
#include "rpc/virnetsocket.h"
54
#include "virstoragefile.h"
M
Martin Kletzander 已提交
55
#include "viruri.h"
56
#include "virhook.h"
57
#include "virstring.h"
58
#include "virtypedparam.h"
M
Michael R. Hines 已提交
59
#include "virprocess.h"
60
#include "nwfilter_conf.h"
61
#include "storage/storage_driver.h"
62 63 64

#define VIR_FROM_THIS VIR_FROM_QEMU

65 66
VIR_LOG_INIT("qemu.qemu_migration");

67 68 69 70 71 72 73 74 75 76 77 78 79
VIR_ENUM_IMPL(qemuMigrationJobPhase, QEMU_MIGRATION_PHASE_LAST,
              "none",
              "perform2",
              "begin3",
              "perform3",
              "perform3_done",
              "confirm3_cancelled",
              "confirm3",
              "prepare",
              "finish2",
              "finish3",
);

80
enum qemuMigrationCookieFlags {
81
    QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS,
82
    QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE,
83
    QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT,
84
    QEMU_MIGRATION_COOKIE_FLAG_NETWORK,
85
    QEMU_MIGRATION_COOKIE_FLAG_NBD,
86
    QEMU_MIGRATION_COOKIE_FLAG_STATS,
87 88 89 90 91 92 93

    QEMU_MIGRATION_COOKIE_FLAG_LAST
};

VIR_ENUM_DECL(qemuMigrationCookieFlag);
VIR_ENUM_IMPL(qemuMigrationCookieFlag,
              QEMU_MIGRATION_COOKIE_FLAG_LAST,
94 95 96 97
              "graphics",
              "lockstate",
              "persistent",
              "network",
98 99
              "nbd",
              "statistics");
100 101 102

enum qemuMigrationCookieFeatures {
    QEMU_MIGRATION_COOKIE_GRAPHICS  = (1 << QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS),
103
    QEMU_MIGRATION_COOKIE_LOCKSTATE = (1 << QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE),
104
    QEMU_MIGRATION_COOKIE_PERSISTENT = (1 << QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT),
105
    QEMU_MIGRATION_COOKIE_NETWORK = (1 << QEMU_MIGRATION_COOKIE_FLAG_NETWORK),
106
    QEMU_MIGRATION_COOKIE_NBD = (1 << QEMU_MIGRATION_COOKIE_FLAG_NBD),
107
    QEMU_MIGRATION_COOKIE_STATS = (1 << QEMU_MIGRATION_COOKIE_FLAG_STATS),
108 109 110 111 112 113 114 115 116 117 118 119
};

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

120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
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;
};

141 142 143 144
typedef struct _qemuMigrationCookieNBD qemuMigrationCookieNBD;
typedef qemuMigrationCookieNBD *qemuMigrationCookieNBDPtr;
struct _qemuMigrationCookieNBD {
    int port; /* on which port does NBD server listen for incoming data */
145 146 147 148 149 150

    size_t ndisks;  /* Number of items in @disk array */
    struct {
        char *target;                   /* Disk target */
        unsigned long long capacity;    /* And its capacity */
    } *disks;
151 152
};

153 154 155
typedef struct _qemuMigrationCookie qemuMigrationCookie;
typedef qemuMigrationCookie *qemuMigrationCookiePtr;
struct _qemuMigrationCookie {
E
Eric Blake 已提交
156 157
    unsigned int flags;
    unsigned int flagsMandatory;
158 159

    /* Host properties */
160 161 162 163
    unsigned char localHostuuid[VIR_UUID_BUFLEN];
    unsigned char remoteHostuuid[VIR_UUID_BUFLEN];
    char *localHostname;
    char *remoteHostname;
164 165 166 167

    /* Guest properties */
    unsigned char uuid[VIR_UUID_BUFLEN];
    char *name;
168

169 170 171 172
    /* If (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) */
    char *lockState;
    char *lockDriver;

173 174
    /* If (flags & QEMU_MIGRATION_COOKIE_GRAPHICS) */
    qemuMigrationCookieGraphicsPtr graphics;
175 176 177

    /* If (flags & QEMU_MIGRATION_COOKIE_PERSISTENT) */
    virDomainDefPtr persistent;
178 179 180

    /* If (flags & QEMU_MIGRATION_COOKIE_NETWORK) */
    qemuMigrationCookieNetworkPtr network;
181 182 183

    /* If (flags & QEMU_MIGRATION_COOKIE_NBD) */
    qemuMigrationCookieNBDPtr nbd;
184 185 186

    /* If (flags & QEMU_MIGRATION_COOKIE_STATS) */
    qemuDomainJobInfoPtr jobInfo;
187 188
};

189 190 191 192 193 194 195 196 197
static void qemuMigrationCookieGraphicsFree(qemuMigrationCookieGraphicsPtr grap)
{
    if (!grap)
        return;
    VIR_FREE(grap->listen);
    VIR_FREE(grap->tlsSubject);
    VIR_FREE(grap);
}

198

199 200 201
static void
qemuMigrationCookieNetworkFree(qemuMigrationCookieNetworkPtr network)
{
202
    size_t i;
203 204 205 206 207 208 209 210 211 212 213 214 215

    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);
}


216 217 218 219 220 221 222 223 224 225 226 227
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);
}


228 229 230 231 232
static void qemuMigrationCookieFree(qemuMigrationCookiePtr mig)
{
    if (!mig)
        return;

233 234
    qemuMigrationCookieGraphicsFree(mig->graphics);
    qemuMigrationCookieNetworkFree(mig->network);
235
    qemuMigrationCookieNBDFree(mig->nbd);
236

237 238
    VIR_FREE(mig->localHostname);
    VIR_FREE(mig->remoteHostname);
239
    VIR_FREE(mig->name);
240 241
    VIR_FREE(mig->lockState);
    VIR_FREE(mig->lockDriver);
242
    VIR_FREE(mig->jobInfo);
243 244 245 246
    VIR_FREE(mig);
}


247
#ifdef WITH_GNUTLS
248 249 250 251 252 253 254 255 256 257 258 259
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)
260
        goto error;
261 262

    if (virFileReadAll(certfile, 8192, &pemdata) < 0) {
263 264
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unable to read server cert %s"), certfile);
265 266 267 268 269
        goto error;
    }

    ret = gnutls_x509_crt_init(&cert);
    if (ret < 0) {
270 271 272
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot initialize cert object: %s"),
                       gnutls_strerror(ret));
273 274 275 276 277 278 279 280
        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) {
281 282 283
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot load cert data from %s: %s"),
                       certfile, gnutls_strerror(ret));
284 285 286 287 288
        goto error;
    }

    subjectlen = 1024;
    if (VIR_ALLOC_N(subject, subjectlen+1) < 0)
289
        goto error;
290 291 292 293 294 295 296 297 298

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

    VIR_FREE(certfile);
    VIR_FREE(pemdata);

    return subject;

299
 error:
300 301 302 303
    VIR_FREE(certfile);
    VIR_FREE(pemdata);
    return NULL;
}
304
#endif
305 306

static qemuMigrationCookieGraphicsPtr
307
qemuMigrationCookieGraphicsAlloc(virQEMUDriverPtr driver,
308 309 310 311
                                 virDomainGraphicsDefPtr def)
{
    qemuMigrationCookieGraphicsPtr mig = NULL;
    const char *listenAddr;
312
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
313 314

    if (VIR_ALLOC(mig) < 0)
315
        goto error;
316 317 318 319

    mig->type = def->type;
    if (mig->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
        mig->port = def->data.vnc.port;
320
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
321
        if (!listenAddr)
322
            listenAddr = cfg->vncListen;
323

324
#ifdef WITH_GNUTLS
325 326
        if (cfg->vncTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->vncTLSx509certdir)))
327
            goto error;
328
#endif
329 330
    } else {
        mig->port = def->data.spice.port;
331
        if (cfg->spiceTLS)
332 333 334
            mig->tlsPort = def->data.spice.tlsPort;
        else
            mig->tlsPort = -1;
335
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
336
        if (!listenAddr)
337
            listenAddr = cfg->spiceListen;
338

339
#ifdef WITH_GNUTLS
340 341
        if (cfg->spiceTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->spiceTLSx509certdir)))
342
            goto error;
343
#endif
344
    }
345 346
    if (VIR_STRDUP(mig->listen, listenAddr) < 0)
        goto error;
347

348
    virObjectUnref(cfg);
349 350
    return mig;

351
 error:
352
    qemuMigrationCookieGraphicsFree(mig);
353
    virObjectUnref(cfg);
354 355 356 357
    return NULL;
}


358
static qemuMigrationCookieNetworkPtr
359
qemuMigrationCookieNetworkAlloc(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
360 361 362
                                virDomainDefPtr def)
{
    qemuMigrationCookieNetworkPtr mig;
363
    size_t i;
364 365

    if (VIR_ALLOC(mig) < 0)
366
        goto error;
367 368 369 370

    mig->nnets = def->nnets;

    if (VIR_ALLOC_N(mig->net, def->nnets) <0)
371
        goto error;
372 373 374 375 376 377 378 379 380 381 382 383 384 385 386

    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:
387
               break;
388
            case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
389 390
                if (virNetDevOpenvswitchGetMigrateData(&mig->net[i].portdata,
                                                       netptr->ifname) != 0) {
J
Jiri Denemark 已提交
391 392 393
                        virReportError(VIR_ERR_INTERNAL_ERROR,
                                       _("Unable to run command to get OVS port data for "
                                         "interface %s"), netptr->ifname);
394 395 396
                        goto error;
                }
                break;
397 398 399 400 401 402 403
            default:
                break;
            }
        }
    }
    return mig;

404
 error:
405 406 407 408
    qemuMigrationCookieNetworkFree(mig);
    return NULL;
}

409 410 411
static qemuMigrationCookiePtr
qemuMigrationCookieNew(virDomainObjPtr dom)
{
J
Jiri Denemark 已提交
412
    qemuDomainObjPrivatePtr priv = dom->privateData;
413
    qemuMigrationCookiePtr mig = NULL;
J
Jiri Denemark 已提交
414
    const char *name;
415 416

    if (VIR_ALLOC(mig) < 0)
417
        goto error;
418

J
Jiri Denemark 已提交
419 420 421 422
    if (priv->origname)
        name = priv->origname;
    else
        name = dom->def->name;
423 424
    if (VIR_STRDUP(mig->name, name) < 0)
        goto error;
425 426
    memcpy(mig->uuid, dom->def->uuid, VIR_UUID_BUFLEN);

427
    if (!(mig->localHostname = virGetHostname()))
428
        goto error;
429
    if (virGetHostUUID(mig->localHostuuid) < 0) {
430 431
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Unable to obtain host UUID"));
432 433 434 435 436
        goto error;
    }

    return mig;

437
 error:
438 439 440 441 442
    qemuMigrationCookieFree(mig);
    return NULL;
}


443 444
static int
qemuMigrationCookieAddGraphics(qemuMigrationCookiePtr mig,
445
                               virQEMUDriverPtr driver,
446 447
                               virDomainObjPtr dom)
{
448 449
    size_t i = 0;

450
    if (mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) {
451 452
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration graphics data already present"));
453 454 455
        return -1;
    }

456 457 458 459 460 461 462 463
    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;
       }
464
    }
465 466 467 468 469

    return 0;
}


470 471
static int
qemuMigrationCookieAddLockstate(qemuMigrationCookiePtr mig,
472
                                virQEMUDriverPtr driver,
473 474 475 476 477
                                virDomainObjPtr dom)
{
    qemuDomainObjPrivatePtr priv = dom->privateData;

    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
478 479
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration lockstate data already present"));
480 481 482 483
        return -1;
    }

    if (virDomainObjGetState(dom, NULL) == VIR_DOMAIN_PAUSED) {
484
        if (VIR_STRDUP(mig->lockState, priv->lockState) < 0)
485 486 487 488 489 490
            return -1;
    } else {
        if (virDomainLockProcessInquire(driver->lockManager, dom, &mig->lockState) < 0)
            return -1;
    }

491
    if (VIR_STRDUP(mig->lockDriver, virLockManagerPluginGetName(driver->lockManager)) < 0) {
492 493 494 495 496 497 498 499 500 501 502
        VIR_FREE(mig->lockState);
        return -1;
    }

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

    return 0;
}


503 504 505 506 507
static int
qemuMigrationCookieAddPersistent(qemuMigrationCookiePtr mig,
                                 virDomainObjPtr dom)
{
    if (mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) {
508 509
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration persistent data already present"));
510 511 512 513 514 515 516 517 518 519 520 521 522
        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;
}


523 524
static int
qemuMigrationCookieAddNetwork(qemuMigrationCookiePtr mig,
525
                              virQEMUDriverPtr driver,
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543
                              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;
}

544

545 546
static int
qemuMigrationCookieAddNBD(qemuMigrationCookiePtr mig,
547
                          virQEMUDriverPtr driver,
548 549 550
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
551 552
    virHashTablePtr stats = NULL;
    size_t i;
553
    int ret = -1, rc;
554 555 556

    /* It is not a bug if there already is a NBD data */
    if (!mig->nbd &&
557
        VIR_ALLOC(mig->nbd) < 0)
558 559
        return -1;

560 561 562 563 564 565 566 567 568 569 570 571 572 573
    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);
574 575
            rc = qemuMonitorBlockStatsUpdateCapacity(priv->mon, stats, false);
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
576
                goto cleanup;
577
            if (rc < 0)
578 579 580 581 582 583 584 585 586 587 588 589 590 591
                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++;
    }

592 593 594
    mig->nbd->port = priv->nbdPort;
    mig->flags |= QEMU_MIGRATION_COOKIE_NBD;

595 596 597 598
    ret = 0;
 cleanup:
    virHashFree(stats);
    return ret;
599 600 601
}


602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
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;
}


621 622 623
static void qemuMigrationCookieGraphicsXMLFormat(virBufferPtr buf,
                                                 qemuMigrationCookieGraphicsPtr grap)
{
624
    virBufferAsprintf(buf, "<graphics type='%s' port='%d' listen='%s'",
625 626 627 628 629 630
                      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");
631 632 633 634
        virBufferAdjustIndent(buf, 2);
        virBufferEscapeString(buf, "<cert info='subject' value='%s'/>\n", grap->tlsSubject);
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</graphics>\n");
635 636 637 638 639 640
    } else {
        virBufferAddLit(buf, "/>\n");
    }
}


641 642 643 644
static void
qemuMigrationCookieNetworkXMLFormat(virBufferPtr buf,
                                    qemuMigrationCookieNetworkPtr optr)
{
645
    size_t i;
646 647 648 649 650 651
    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) {
652 653
                virBufferAddLit(buf, "<network>\n");
                virBufferAdjustIndent(buf, 2);
654 655
                empty = false;
            }
656
            virBufferAsprintf(buf, "<interface index='%zu' vporttype='%s'",
657 658 659
                              i, virNetDevVPortTypeToString(optr->net[i].vporttype));
            if (optr->net[i].portdata) {
                virBufferAddLit(buf, ">\n");
660 661
                virBufferAdjustIndent(buf, 2);
                virBufferEscapeString(buf, "<portdata>%s</portdata>\n",
662
                                      optr->net[i].portdata);
663 664
                virBufferAdjustIndent(buf, -2);
                virBufferAddLit(buf, "</interface>\n");
665 666 667 668 669
            } else {
                virBufferAddLit(buf, "/>\n");
            }
        }
    }
670 671 672 673
    if (!empty) {
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</network>\n");
    }
674 675 676
}


677 678 679 680 681 682 683 684 685
static void
qemuMigrationCookieStatisticsXMLFormat(virBufferPtr buf,
                                       qemuDomainJobInfoPtr jobInfo)
{
    qemuMonitorMigrationStatus *status = &jobInfo->status;

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

686 687 688
    virBufferAsprintf(buf, "<started>%llu</started>\n", jobInfo->started);
    virBufferAsprintf(buf, "<stopped>%llu</stopped>\n", jobInfo->stopped);

689 690 691 692 693 694 695 696 697 698
    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$s>\n",
                          VIR_DOMAIN_JOB_DOWNTIME,
                          status->downtime);
699 700 701 702
    if (status->setup_time_set)
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_SETUP_TIME,
                          status->setup_time);
703 704 705 706 707 708 709 710 711 712

    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$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_REMAINING,
                      status->ram_remaining);
713 714 715
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_BPS,
                      status->ram_bps);
716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737

    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$s>\n",
                      VIR_DOMAIN_JOB_DISK_REMAINING,
                      status->disk_remaining);
738 739 740
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_BPS,
                      status->disk_bps);
741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764

    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");
}


765
static int
766
qemuMigrationCookieXMLFormat(virQEMUDriverPtr driver,
767
                             virBufferPtr buf,
768
                             qemuMigrationCookiePtr mig)
769 770 771
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
    char hostuuidstr[VIR_UUID_STRING_BUFLEN];
772
    size_t i;
773 774

    virUUIDFormat(mig->uuid, uuidstr);
775
    virUUIDFormat(mig->localHostuuid, hostuuidstr);
776

777
    virBufferAddLit(buf, "<qemu-migration>\n");
778 779 780 781 782
    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);
783

784
    for (i = 0; i < QEMU_MIGRATION_COOKIE_FLAG_LAST; i++) {
785
        if (mig->flagsMandatory & (1 << i))
786
            virBufferAsprintf(buf, "<feature name='%s'/>\n",
787 788 789
                              qemuMigrationCookieFlagTypeToString(i));
    }

790 791
    if ((mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        mig->graphics)
792 793
        qemuMigrationCookieGraphicsXMLFormat(buf, mig->graphics);

794 795
    if ((mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        mig->lockState) {
796
        virBufferAsprintf(buf, "<lockstate driver='%s'>\n",
797
                          mig->lockDriver);
798 799
        virBufferAdjustIndent(buf, 2);
        virBufferAsprintf(buf, "<leases>%s</leases>\n",
800
                          mig->lockState);
801 802
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</lockstate>\n");
803 804
    }

805 806
    if ((mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        mig->persistent) {
807 808 809
        if (qemuDomainDefFormatBuf(driver,
                                   mig->persistent,
                                   VIR_DOMAIN_XML_INACTIVE |
810 811
                                   VIR_DOMAIN_XML_SECURE |
                                   VIR_DOMAIN_XML_MIGRATABLE,
812
                                   buf) < 0)
813
            return -1;
814 815
    }

816 817 818
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NETWORK) && mig->network)
        qemuMigrationCookieNetworkXMLFormat(buf, mig->network);

819
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NBD) && mig->nbd) {
820
        virBufferAddLit(buf, "<nbd");
821 822
        if (mig->nbd->port)
            virBufferAsprintf(buf, " port='%d'", mig->nbd->port);
823 824 825 826 827 828 829 830 831 832 833 834 835 836
        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");
        }
837 838
    }

839 840 841
    if (mig->flags & QEMU_MIGRATION_COOKIE_STATS && mig->jobInfo)
        qemuMigrationCookieStatisticsXMLFormat(buf, mig->jobInfo);

842
    virBufferAdjustIndent(buf, -2);
843
    virBufferAddLit(buf, "</qemu-migration>\n");
844
    return 0;
845 846 847
}


848
static char *qemuMigrationCookieXMLFormatStr(virQEMUDriverPtr driver,
849
                                             qemuMigrationCookiePtr mig)
850 851 852
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;

853
    if (qemuMigrationCookieXMLFormat(driver, &buf, mig) < 0) {
854 855 856
        virBufferFreeAndReset(&buf);
        return NULL;
    }
857

858
    if (virBufferCheckError(&buf) < 0)
859 860 861 862 863 864
        return NULL;

    return virBufferContentAndReset(&buf);
}


865 866 867 868 869 870 871
static qemuMigrationCookieGraphicsPtr
qemuMigrationCookieGraphicsXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieGraphicsPtr grap;
    char *tmp;

    if (VIR_ALLOC(grap) < 0)
872
        goto error;
873 874

    if (!(tmp = virXPathString("string(./graphics/@type)", ctxt))) {
875 876
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing type attribute in migration data"));
877 878 879
        goto error;
    }
    if ((grap->type = virDomainGraphicsTypeFromString(tmp)) < 0) {
880 881
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unknown graphics type %s"), tmp);
882 883 884
        VIR_FREE(tmp);
        goto error;
    }
E
Eric Blake 已提交
885
    VIR_FREE(tmp);
886
    if (virXPathInt("string(./graphics/@port)", ctxt, &grap->port) < 0) {
887 888
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing port attribute in migration data"));
889 890 891 892
        goto error;
    }
    if (grap->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        if (virXPathInt("string(./graphics/@tlsPort)", ctxt, &grap->tlsPort) < 0) {
893 894
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing tlsPort attribute in migration data"));
895 896 897 898
            goto error;
        }
    }
    if (!(grap->listen = virXPathString("string(./graphics/@listen)", ctxt))) {
899 900
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing listen attribute in migration data"));
901 902 903
        goto error;
    }
    /* Optional */
904
    grap->tlsSubject = virXPathString("string(./graphics/cert[@info='subject']/@value)", ctxt);
905 906 907

    return grap;

908
 error:
909 910 911 912 913
    qemuMigrationCookieGraphicsFree(grap);
    return NULL;
}


914 915 916 917
static qemuMigrationCookieNetworkPtr
qemuMigrationCookieNetworkXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieNetworkPtr optr;
918
    size_t i;
919 920 921 922 923 924
    int n;
    xmlNodePtr *interfaces = NULL;
    char *vporttype;
    xmlNodePtr save_ctxt = ctxt->node;

    if (VIR_ALLOC(optr) < 0)
925
        goto error;
926 927 928 929 930 931 932 933

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

    optr->nnets = n;
934
    if (VIR_ALLOC_N(optr->net, optr->nnets) < 0)
935
        goto error;
936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951

    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);

952
 cleanup:
953 954 955
    ctxt->node = save_ctxt;
    return optr;

956
 error:
957 958 959 960 961 962 963
    VIR_FREE(interfaces);
    qemuMigrationCookieNetworkFree(optr);
    optr = NULL;
    goto cleanup;
}


964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 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
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;
}


1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
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;

1044 1045 1046
    virXPathULongLong("string(./started[1])", ctxt, &jobInfo->started);
    virXPathULongLong("string(./stopped[1])", ctxt, &jobInfo->stopped);

1047 1048 1049 1050 1051 1052 1053
    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;
1054 1055 1056
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_SETUP_TIME "[1])",
                          ctxt, &status->setup_time) == 0)
        status->setup_time_set = true;
1057 1058 1059 1060 1061 1062 1063

    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);
1064 1065
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_BPS "[1])",
                      ctxt, &status->ram_bps);
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080

    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);
1081 1082
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_BPS "[1])",
                      ctxt, &status->disk_bps);
1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101

    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;
}


1102 1103
static int
qemuMigrationCookieXMLParse(qemuMigrationCookiePtr mig,
1104
                            virQEMUDriverPtr driver,
1105
                            xmlDocPtr doc,
1106
                            xmlXPathContextPtr ctxt,
E
Eric Blake 已提交
1107
                            unsigned int flags)
1108 1109
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
1110
    char *tmp = NULL;
1111
    xmlNodePtr *nodes = NULL;
1112 1113
    size_t i;
    int n;
1114 1115 1116 1117
    virCapsPtr caps = NULL;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto error;
1118 1119 1120 1121 1122 1123 1124 1125

    /* 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))) {
1126 1127
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing name element in migration data"));
1128 1129 1130
        goto error;
    }
    if (STRNEQ(tmp, mig->name)) {
1131 1132 1133
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected name %s vs %s"),
                       tmp, mig->name);
1134 1135 1136 1137 1138 1139 1140
        goto error;
    }
    VIR_FREE(tmp);

    /* Extract domain uuid */
    tmp = virXPathString("string(./uuid[1])", ctxt);
    if (!tmp) {
1141 1142
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing uuid element in migration data"));
1143 1144 1145 1146
        goto error;
    }
    virUUIDFormat(mig->uuid, uuidstr);
    if (STRNEQ(tmp, uuidstr)) {
1147 1148 1149
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected UUID %s vs %s"),
                       tmp, uuidstr);
1150
        goto error;
1151 1152 1153 1154
    }
    VIR_FREE(tmp);

    /* Check & forbid "localhost" migration */
1155
    if (!(mig->remoteHostname = virXPathString("string(./hostname[1])", ctxt))) {
1156 1157
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostname element in migration data"));
1158 1159
        goto error;
    }
1160
    if (STREQ(mig->remoteHostname, mig->localHostname)) {
1161 1162 1163
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Attempt to migrate guest to the same host %s"),
                       mig->remoteHostname);
1164 1165 1166 1167
        goto error;
    }

    if (!(tmp = virXPathString("string(./hostuuid[1])", ctxt))) {
1168 1169 1170
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostuuid element in migration data"));
        goto error;
1171
    }
1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183
    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);
1184

1185 1186
    /* Check to ensure all mandatory features from XML are also
     * present in 'flags' */
1187
    if ((n = virXPathNodeSet("./feature", ctxt, &nodes)) < 0)
1188 1189
        goto error;

1190
    for (i = 0; i < n; i++) {
1191 1192 1193
        int val;
        char *str = virXMLPropString(nodes[i], "name");
        if (!str) {
1194 1195
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing feature name"));
1196 1197 1198 1199
            goto error;
        }

        if ((val = qemuMigrationCookieFlagTypeFromString(str)) < 0) {
1200 1201 1202
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unknown migration cookie feature %s"),
                           str);
1203 1204 1205 1206 1207
            VIR_FREE(str);
            goto error;
        }

        if ((flags & (1 << val)) == 0) {
1208 1209 1210
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unsupported migration cookie feature %s"),
                           str);
1211 1212 1213 1214 1215 1216
            VIR_FREE(str);
        }
        VIR_FREE(str);
    }
    VIR_FREE(nodes);

1217 1218 1219 1220 1221
    if ((flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        virXPathBoolean("count(./graphics) > 0", ctxt) &&
        (!(mig->graphics = qemuMigrationCookieGraphicsXMLParse(ctxt))))
        goto error;

1222 1223 1224 1225
    if ((flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        virXPathBoolean("count(./lockstate) > 0", ctxt)) {
        mig->lockDriver = virXPathString("string(./lockstate[1]/@driver)", ctxt);
        if (!mig->lockDriver) {
1226 1227
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Missing lock driver name in migration cookie"));
1228 1229 1230 1231 1232 1233 1234
            goto error;
        }
        mig->lockState = virXPathString("string(./lockstate[1]/leases[1])", ctxt);
        if (mig->lockState && STREQ(mig->lockState, ""))
            VIR_FREE(mig->lockState);
    }

1235 1236 1237
    if ((flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        virXPathBoolean("count(./domain) > 0", ctxt)) {
        if ((n = virXPathNodeSet("./domain", ctxt, &nodes)) > 1) {
1238 1239 1240 1241
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Too many domain elements in "
                             "migration cookie: %d"),
                           n);
1242 1243
            goto error;
        }
1244 1245
        mig->persistent = virDomainDefParseNode(doc, nodes[0],
                                                caps, driver->xmlopt,
1246
                                                -1, VIR_DOMAIN_DEF_PARSE_INACTIVE);
1247 1248 1249 1250 1251 1252 1253 1254
        if (!mig->persistent) {
            /* virDomainDefParseNode already reported
             * an error for us */
            goto error;
        }
        VIR_FREE(nodes);
    }

1255 1256 1257 1258 1259
    if ((flags & QEMU_MIGRATION_COOKIE_NETWORK) &&
        virXPathBoolean("count(./network) > 0", ctxt) &&
        (!(mig->network = qemuMigrationCookieNetworkXMLParse(ctxt))))
        goto error;

1260
    if (flags & QEMU_MIGRATION_COOKIE_NBD &&
1261 1262 1263
        virXPathBoolean("boolean(./nbd)", ctxt) &&
        (!(mig->nbd = qemuMigrationCookieNBDXMLParse(ctxt))))
        goto error;
1264

1265 1266 1267 1268 1269
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        virXPathBoolean("boolean(./statistics)", ctxt) &&
        (!(mig->jobInfo = qemuMigrationCookieStatisticsXMLParse(ctxt))))
        goto error;

1270
    virObjectUnref(caps);
1271 1272
    return 0;

1273
 error:
1274
    VIR_FREE(tmp);
1275
    VIR_FREE(nodes);
1276
    virObjectUnref(caps);
1277 1278 1279 1280 1281 1282
    return -1;
}


static int
qemuMigrationCookieXMLParseStr(qemuMigrationCookiePtr mig,
1283
                               virQEMUDriverPtr driver,
1284
                               const char *xml,
E
Eric Blake 已提交
1285
                               unsigned int flags)
1286 1287 1288
{
    xmlDocPtr doc = NULL;
    xmlXPathContextPtr ctxt = NULL;
1289
    int ret = -1;
1290 1291 1292

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

1293
    if (!(doc = virXMLParseStringCtxt(xml, _("(qemu_migration_cookie)"), &ctxt)))
1294 1295
        goto cleanup;

1296
    ret = qemuMigrationCookieXMLParse(mig, driver, doc, ctxt, flags);
1297

1298
 cleanup:
1299 1300 1301 1302 1303 1304 1305 1306 1307
    xmlXPathFreeContext(ctxt);
    xmlFreeDoc(doc);

    return ret;
}


static int
qemuMigrationBakeCookie(qemuMigrationCookiePtr mig,
1308
                        virQEMUDriverPtr driver,
1309
                        virDomainObjPtr dom,
1310 1311
                        char **cookieout,
                        int *cookieoutlen,
E
Eric Blake 已提交
1312
                        unsigned int flags)
1313
{
1314 1315
    if (!cookieout || !cookieoutlen)
        return 0;
1316 1317 1318

    *cookieoutlen = 0;

1319 1320 1321 1322
    if (flags & QEMU_MIGRATION_COOKIE_GRAPHICS &&
        qemuMigrationCookieAddGraphics(mig, driver, dom) < 0)
        return -1;

1323 1324 1325 1326
    if (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE &&
        qemuMigrationCookieAddLockstate(mig, driver, dom) < 0)
        return -1;

1327 1328 1329 1330
    if (flags & QEMU_MIGRATION_COOKIE_PERSISTENT &&
        qemuMigrationCookieAddPersistent(mig, dom) < 0)
        return -1;

1331 1332 1333 1334 1335
    if (flags & QEMU_MIGRATION_COOKIE_NETWORK &&
        qemuMigrationCookieAddNetwork(mig, driver, dom) < 0) {
        return -1;
    }

1336 1337 1338 1339
    if ((flags & QEMU_MIGRATION_COOKIE_NBD) &&
        qemuMigrationCookieAddNBD(mig, driver, dom) < 0)
        return -1;

1340 1341 1342 1343
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        qemuMigrationCookieAddStatistics(mig, dom) < 0)
        return -1;

1344
    if (!(*cookieout = qemuMigrationCookieXMLFormatStr(driver, mig)))
1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355
        return -1;

    *cookieoutlen = strlen(*cookieout) + 1;

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

    return 0;
}


static qemuMigrationCookiePtr
1356
qemuMigrationEatCookie(virQEMUDriverPtr driver,
1357
                       virDomainObjPtr dom,
1358 1359
                       const char *cookiein,
                       int cookieinlen,
E
Eric Blake 已提交
1360
                       unsigned int flags)
1361 1362 1363 1364 1365 1366
{
    qemuMigrationCookiePtr mig = NULL;

    /* Parse & validate incoming cookie (if any) */
    if (cookiein && cookieinlen &&
        cookiein[cookieinlen-1] != '\0') {
1367 1368
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration cookie was not NULL terminated"));
1369 1370 1371 1372 1373 1374 1375 1376 1377 1378
        goto error;
    }

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

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

    if (cookiein && cookieinlen &&
        qemuMigrationCookieXMLParseStr(mig,
1379
                                       driver,
1380 1381 1382 1383
                                       cookiein,
                                       flags) < 0)
        goto error;

1384 1385 1386
    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
        if (!mig->lockDriver) {
            if (virLockManagerPluginUsesState(driver->lockManager)) {
1387 1388 1389
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Missing %s lock state for migration cookie"),
                               virLockManagerPluginGetName(driver->lockManager));
1390 1391 1392 1393
                goto error;
            }
        } else if (STRNEQ(mig->lockDriver,
                          virLockManagerPluginGetName(driver->lockManager))) {
1394 1395 1396 1397
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Source host lock driver %s different from target %s"),
                           mig->lockDriver,
                           virLockManagerPluginGetName(driver->lockManager));
1398 1399 1400 1401
            goto error;
        }
    }

1402 1403
    return mig;

1404
 error:
1405 1406 1407
    qemuMigrationCookieFree(mig);
    return NULL;
}
1408

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
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;
}

1456 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

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;
}


1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
/**
 * 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
1623 1624
                            virDomainObjPtr vm,
                            const char *listenAddr)
1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635
{
    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 */
1636 1637
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1638 1639 1640 1641
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1642
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1643 1644 1645 1646 1647 1648 1649
            goto cleanup;

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

        if (!port &&
1650
            ((virPortAllocatorAcquire(driver->migrationPorts, &port) < 0) ||
1651
             (qemuMonitorNBDServerStart(priv->mon, listenAddr, port) < 0))) {
1652
            goto exit_monitor;
1653 1654
        }

1655 1656 1657
        if (qemuMonitorNBDServerAdd(priv->mon, diskAlias, true) < 0)
            goto exit_monitor;
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
1658 1659 1660 1661 1662 1663
            goto cleanup;
    }

    priv->nbdPort = port;
    ret = 0;

1664
 cleanup:
1665
    VIR_FREE(diskAlias);
1666
    if (ret < 0)
1667
        virPortAllocatorRelease(driver->migrationPorts, port);
1668
    return ret;
1669 1670 1671 1672

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;
1673 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
/**
 * 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;
1709
    char *hoststr = NULL;
1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728
    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;

1729 1730
    /* escape literal IPv6 address */
    if (strchr(host, ':')) {
1731
        if (virAsprintf(&hoststr, "[%s]", host) < 0)
1732 1733 1734 1735 1736
            goto error;
    } else if (VIR_STRDUP(hoststr, host) < 0) {
        goto error;
    }

1737 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];
        virDomainBlockJobInfo info;

        /* 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 1775 1776 1777 1778 1779 1780 1781 1782 1783

        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 };

            memset(&info, 0, sizeof(info));

            if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
                goto error;
            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 */
1784
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
1785
                priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
1786 1787 1788 1789 1790
                virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                               qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                               _("canceled by client"));
                goto error;
            }
1791 1792
            mon_ret = qemuMonitorBlockJobInfo(priv->mon, diskAlias, &info,
                                              NULL);
1793 1794
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                goto error;
1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820

            if (mon_ret < 0)
                goto error;

            if (info.cur == info.end) {
                VIR_DEBUG("Drive mirroring of '%s' completed", diskAlias);
                break;
            }

            /* XXX Frankly speaking, we should listen to the events,
             * instead of doing this. But this works for now and we
             * are doing something similar in migration itself anyway */

            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;

1821
 cleanup:
1822 1823
    VIR_FREE(diskAlias);
    VIR_FREE(nbd_dest);
1824
    VIR_FREE(hoststr);
1825 1826
    return ret;

1827
 error:
1828 1829 1830 1831 1832 1833 1834
    /* don't overwrite any errors */
    err = virSaveLastError();
    /* cancel any outstanding jobs */
    while (lastGood) {
        virDomainDiskDefPtr disk = vm->def->disks[--lastGood];

        /* skip shared, RO disks */
1835 1836
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1837 1838 1839 1840
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1841
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1842 1843 1844
            continue;
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
1845
            if (qemuMonitorBlockJob(priv->mon, diskAlias, NULL, NULL, 0,
1846
                                    BLOCK_JOB_ABORT, true) < 0) {
1847 1848
                VIR_WARN("Unable to cancel block-job on '%s'", diskAlias);
            }
1849 1850
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                break;
1851 1852 1853 1854 1855 1856 1857 1858 1859
        } else {
            VIR_WARN("Unable to enter monitor. No block job cancelled");
        }
    }
    if (err)
        virSetError(err);
    virFreeError(err);
    goto cleanup;
}
1860

1861

1862
static int
1863 1864 1865 1866 1867 1868 1869
qemuMigrationStopNBDServer(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuMigrationCookiePtr mig)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (!mig->nbd)
1870
        return 0;
1871 1872 1873

    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
1874
        return -1;
1875 1876 1877

    if (qemuMonitorNBDServerStop(priv->mon) < 0)
        VIR_WARN("Unable to stop NBD server");
1878 1879
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
1880

1881
    virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
1882
    priv->nbdPort = 0;
1883
    return 0;
1884 1885
}

1886
static int
1887 1888 1889 1890 1891 1892 1893
qemuMigrationCancelDriveMirror(qemuMigrationCookiePtr mig,
                               virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    size_t i;
    char *diskAlias = NULL;
1894
    int ret = 0;
1895 1896 1897 1898 1899 1900 1901

    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 */
1902 1903
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1904 1905 1906 1907
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1908
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1909 1910 1911 1912 1913 1914
            goto cleanup;

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

1915
        if (qemuMonitorBlockJob(priv->mon, diskAlias, NULL, NULL, 0,
1916
                                BLOCK_JOB_ABORT, true) < 0)
1917
            VIR_WARN("Unable to stop block job on %s", diskAlias);
1918 1919 1920 1921
        if (qemuDomainObjExitMonitor(driver, vm) < 0) {
            ret = -1;
            goto cleanup;
        }
1922 1923
    }

1924
 cleanup:
1925
    VIR_FREE(diskAlias);
1926
    return ret;
1927 1928
}

1929 1930 1931 1932 1933 1934 1935 1936 1937
/* 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. */
1938
bool
1939
qemuMigrationIsAllowed(virQEMUDriverPtr driver, virDomainObjPtr vm,
1940
                       virDomainDefPtr def, bool remote, bool abort_on_error)
1941
{
1942
    int nsnapshots;
1943
    int pauseReason;
1944
    bool forbid;
1945
    size_t i;
1946 1947 1948

    if (vm) {
        if (qemuProcessAutoDestroyActive(driver, vm)) {
1949 1950
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("domain is marked for auto destroy"));
1951 1952
            return false;
        }
1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965

        /* 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;
            }
1966 1967

            /* cancel migration if disk I/O error is emitted while migrating */
1968 1969
            if (abort_on_error &&
                virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
1970 1971 1972 1973 1974 1975
                pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("cannot migrate domain with I/O error"));
                return false;
            }

1976
        }
1977

E
Eric Blake 已提交
1978 1979
        if (virDomainHasDiskMirror(vm)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1980
                           _("domain has an active block job"));
E
Eric Blake 已提交
1981 1982
            return false;
        }
1983 1984 1985

        def = vm->def;
    }
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

    /* 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",
2001
                       _("domain has assigned non-USB host devices"));
2002 2003 2004
        return false;
    }

2005 2006 2007
    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
2008

2009 2010
            if (feature->policy != VIR_CPU_FEATURE_REQUIRE)
                continue;
J
Ján Tomko 已提交
2011

2012 2013 2014 2015 2016 2017 2018
            /* 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
2019 2020 2021
        }
    }

2022 2023 2024
    return true;
}

2025 2026 2027
static bool
qemuMigrationIsSafe(virDomainDefPtr def)
{
2028
    size_t i;
2029

2030
    for (i = 0; i < def->ndisks; i++) {
2031
        virDomainDiskDefPtr disk = def->disks[i];
2032
        const char *src = virDomainDiskGetSource(disk);
2033

2034 2035
        /* Our code elsewhere guarantees shared disks are either readonly (in
         * which case cache mode doesn't matter) or used with cache=none */
2036
        if (src &&
2037 2038
            !disk->src->shared &&
            !disk->src->readonly &&
2039
            disk->cachemode != VIR_DOMAIN_DISK_CACHE_DISABLE) {
2040
            int rc;
2041

E
Eric Blake 已提交
2042
            if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
2043
                if ((rc = virFileIsSharedFS(src)) < 0)
2044 2045
                    return false;
                else if (rc == 0)
2046
                    continue;
2047
                if ((rc = virStorageFileIsClusterFS(src)) < 0)
2048
                    return false;
2049 2050
                else if (rc == 1)
                    continue;
2051 2052
            } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                       disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
2053
                continue;
2054
            }
2055

2056 2057 2058
            virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                           _("Migration may lead to data corruption if disks"
                             " use cache != none"));
2059 2060 2061 2062 2063 2064 2065
            return false;
        }
    }

    return true;
}

2066 2067 2068 2069
/** qemuMigrationSetOffline
 * Pause domain for non-live migration.
 */
int
2070
qemuMigrationSetOffline(virQEMUDriverPtr driver,
2071 2072 2073
                        virDomainObjPtr vm)
{
    int ret;
2074
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
2075 2076
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
2077
    if (ret == 0) {
2078
        virObjectEventPtr event;
2079

2080
        event = virDomainEventLifecycleNewFromObj(vm,
2081 2082 2083 2084 2085 2086 2087 2088 2089 2090
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
        if (event)
            qemuDomainEventQueue(driver, event);
    }

    return ret;
}


2091 2092 2093
static int
qemuMigrationSetCompression(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
2094
                            bool state,
2095
                            qemuDomainAsyncJob job)
2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108
{
    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;
2109 2110 2111
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127
    } 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,
2128 2129
                QEMU_MONITOR_MIGRATION_CAPS_XBZRLE,
                state);
2130

2131
 cleanup:
2132 2133
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2134 2135 2136
    return ret;
}

2137 2138 2139
static int
qemuMigrationSetAutoConverge(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
2140
                             bool state,
2141
                             qemuDomainAsyncJob job)
2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154
{
    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;
2155 2156 2157
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2158 2159 2160 2161 2162 2163 2164 2165 2166 2167
    } 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,
2168 2169
                QEMU_MONITOR_MIGRATION_CAPS_AUTO_CONVERGE,
                state);
2170

2171
 cleanup:
2172 2173
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2174 2175 2176 2177
    return ret;
}


2178 2179 2180
static int
qemuMigrationSetPinAll(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
2181
                       bool state,
2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195
                       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;
2196 2197 2198
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214
    } 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,
2215 2216
                QEMU_MONITOR_MIGRATION_CAPS_RDMA_PIN_ALL,
                state);
2217 2218

 cleanup:
2219 2220
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2221 2222 2223
    return ret;
}

2224 2225 2226 2227 2228 2229 2230
static int
qemuMigrationWaitForSpice(virQEMUDriverPtr driver,
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool wait_for_spice = false;
    bool spice_migrated = false;
2231
    size_t i = 0;
2232
    int rc;
2233

2234 2235 2236 2237 2238 2239 2240 2241
    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;
            }
        }
    }
2242 2243 2244 2245 2246 2247 2248 2249

    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 };

2250 2251 2252 2253
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            return -1;

2254 2255 2256 2257
        rc = qemuMonitorGetSpiceMigrationStatus(priv->mon, &spice_migrated);
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
        if (rc < 0)
2258 2259 2260 2261 2262 2263 2264 2265
            return -1;
        virObjectUnlock(vm);
        nanosleep(&ts, NULL);
        virObjectLock(vm);
    }

    return 0;
}
2266

2267
static int
2268
qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver,
2269
                             virDomainObjPtr vm,
2270
                             const char *job,
2271
                             qemuDomainAsyncJob asyncJob)
2272
{
2273
    qemuDomainObjPrivatePtr priv = vm->privateData;
2274
    qemuMonitorMigrationStatus status;
J
Jiri Denemark 已提交
2275 2276
    qemuDomainJobInfoPtr jobInfo;
    int ret;
2277 2278

    memset(&status, 0, sizeof(status));
2279

2280 2281
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
    if (ret < 0) {
2282 2283 2284
        /* Guest already exited or waiting for the job timed out; nothing
         * further to update. */
        return ret;
2285
    }
2286
    ret = qemuMonitorGetMigrationStatus(priv->mon, &status);
2287

2288 2289
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2290

J
Jiri Denemark 已提交
2291 2292
    if (ret < 0 ||
        qemuDomainJobInfoUpdateTime(priv->job.current) < 0)
2293
        return -1;
2294

2295
    ret = -1;
J
Jiri Denemark 已提交
2296 2297 2298 2299 2300
    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
2301
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
2302 2303 2304 2305
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
        ret = 0;
        break;

J
Jiri Denemark 已提交
2306 2307 2308 2309
    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
        jobInfo->type = VIR_DOMAIN_JOB_NONE;
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("is not active"));
2310 2311 2312
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
J
Jiri Denemark 已提交
2313
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
2314 2315
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("unexpectedly failed"));
2316 2317 2318
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
J
Jiri Denemark 已提交
2319
        jobInfo->type = VIR_DOMAIN_JOB_CANCELLED;
2320 2321
        virReportError(VIR_ERR_OPERATION_ABORTED,
                       _("%s: %s"), job, _("canceled by client"));
2322 2323
        break;
    }
J
Jiri Denemark 已提交
2324
    jobInfo->status = status;
2325 2326 2327 2328 2329

    return ret;
}


2330 2331 2332
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
2333
static int
J
Jiri Denemark 已提交
2334 2335
qemuMigrationWaitForCompletion(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
2336
                               qemuDomainAsyncJob asyncJob,
J
Jiri Denemark 已提交
2337 2338
                               virConnectPtr dconn,
                               bool abort_on_error)
2339
{
2340
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
2341
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
2342
    const char *job;
2343
    int pauseReason;
2344

2345 2346
    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
2347 2348
        job = _("migration job");
        break;
2349
    case QEMU_ASYNC_JOB_SAVE:
2350 2351
        job = _("domain save job");
        break;
2352
    case QEMU_ASYNC_JOB_DUMP:
2353 2354 2355 2356 2357
        job = _("domain core dump job");
        break;
    default:
        job = _("job");
    }
2358

J
Jiri Denemark 已提交
2359
    jobInfo->type = VIR_DOMAIN_JOB_UNBOUNDED;
2360

J
Jiri Denemark 已提交
2361
    while (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
2362 2363 2364
        /* Poll every 50ms for progress & to allow cancellation */
        struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };

2365
        if (qemuMigrationUpdateJobStatus(driver, vm, job, asyncJob) == -1)
2366
            break;
2367

2368
        /* cancel migration if disk I/O error is emitted while migrating */
2369
        if (abort_on_error &&
2370
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
2371 2372 2373 2374 2375
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("%s: %s"), job, _("failed due to I/O error"));
            break;
        }
2376

2377
        if (dconn && virConnectIsAlive(dconn) <= 0) {
2378 2379
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
2380
            break;
2381 2382
        }

2383
        virObjectUnlock(vm);
2384 2385 2386

        nanosleep(&ts, NULL);

2387
        virObjectLock(vm);
2388 2389
    }

J
Jiri Denemark 已提交
2390
    if (jobInfo->type == VIR_DOMAIN_JOB_COMPLETED) {
2391
        qemuDomainJobInfoUpdateDowntime(jobInfo);
2392 2393 2394
        VIR_FREE(priv->job.completed);
        if (VIR_ALLOC(priv->job.completed) == 0)
            *priv->job.completed = *jobInfo;
2395
        return 0;
J
Jiri Denemark 已提交
2396
    } else if (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
2397 2398 2399
        /* 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
2400
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
2401 2402
        return -2;
    } else {
2403
        return -1;
2404
    }
2405 2406 2407
}


2408
static int
2409
qemuDomainMigrateGraphicsRelocate(virQEMUDriverPtr driver,
2410
                                  virDomainObjPtr vm,
2411 2412
                                  qemuMigrationCookiePtr cookie,
                                  const char *graphicsuri)
2413 2414
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2415 2416
    int ret = -1;
    const char *listenAddress = NULL;
2417
    virSocketAddr addr;
2418 2419 2420 2421 2422
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
2423

2424
    if (!cookie || (!cookie->graphics && !graphicsuri))
2425 2426
        return 0;

2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445
    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) {
2446
        size_t i;
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 2473

        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;
            }
        }
    }
2474 2475 2476 2477

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
2478 2479 2480 2481
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
2482

2483 2484 2485 2486
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
2487 2488
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
2489
    }
2490

2491
 cleanup:
2492
    virURIFree(uri);
2493 2494 2495 2496
    return ret;
}


2497
static int
2498
qemuDomainMigrateOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
2499 2500 2501
                             virDomainObjPtr vm,
                             qemuMigrationCookiePtr cookie)
{
2502 2503
    virDomainNetDefPtr netptr;
    int ret = -1;
2504
    size_t i;
2505 2506 2507 2508 2509 2510 2511 2512

    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:
2513
           break;
2514
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
2515 2516
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
2517 2518 2519
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
2520 2521 2522
                goto cleanup;
            }
            break;
2523 2524 2525 2526 2527
        default:
            break;
        }
    }

2528
    ret = 0;
2529
 cleanup:
2530 2531 2532 2533
    return ret;
}


2534 2535 2536 2537 2538 2539
/* 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.
 */
2540 2541 2542 2543
static virDomainObjPtr
qemuMigrationCleanup(virDomainObjPtr vm,
                     virConnectPtr conn,
                     void *opaque)
2544
{
2545
    virQEMUDriverPtr driver = opaque;
2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560
    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);

2561
    switch ((qemuMigrationJobPhase) priv->job.phase) {
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 2591
    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 */
        ;
    }

2592
 cleanup:
2593 2594 2595
    return vm;
}

2596

2597
/* The caller is supposed to lock the vm and start a migration job. */
2598 2599
static char
*qemuMigrationBeginPhase(virQEMUDriverPtr driver,
2600
                         virDomainObjPtr vm,
2601
                         const char *xmlin,
2602
                         const char *dname,
2603
                         char **cookieout,
2604 2605
                         int *cookieoutlen,
                         unsigned long flags)
2606 2607 2608
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
2609
    virDomainDefPtr def = NULL;
2610
    qemuDomainObjPrivatePtr priv = vm->privateData;
2611
    virCapsPtr caps = NULL;
2612
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
2613
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2614

2615
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
2616
              " cookieout=%p, cookieoutlen=%p, flags=%lx",
2617
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
2618
              cookieout, cookieoutlen, flags);
2619

2620 2621 2622
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2623 2624 2625 2626 2627 2628
    /* 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);
2629

2630
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
2631 2632
        goto cleanup;

2633 2634 2635
    if (!(flags & VIR_MIGRATE_UNSAFE) && !qemuMigrationIsSafe(vm->def))
        goto cleanup;

2636
    if (flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
2637
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DRIVE_MIRROR)) {
2638
        /* TODO support NBD for TUNNELLED migration */
2639
        if (flags & VIR_MIGRATE_TUNNELLED) {
2640 2641 2642 2643
            VIR_WARN("NBD in tunnelled migration is currently not supported");
        } else {
            cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
            priv->nbdPort = 0;
2644 2645 2646
        }
    }

2647
    if (!(mig = qemuMigrationEatCookie(driver, vm, NULL, 0, 0)))
2648 2649 2650 2651
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
                                cookieout, cookieoutlen,
2652
                                cookieFlags) < 0)
2653 2654
        goto cleanup;

L
liguang 已提交
2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676
    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;
        }
    }

2677
    if (xmlin) {
2678
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt,
M
Matthias Bolte 已提交
2679
                                            QEMU_EXPECTED_VIRT_TYPES,
2680
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE)))
2681 2682
            goto cleanup;

2683
        if (!qemuDomainDefCheckABIStability(driver, vm->def, def))
2684 2685
            goto cleanup;

2686
        rv = qemuDomainDefFormatLive(driver, def, false, true);
2687
    } else {
2688
        rv = qemuDomainDefFormatLive(driver, vm->def, false, true);
2689
    }
2690

2691
 cleanup:
2692
    qemuMigrationCookieFree(mig);
2693
    virObjectUnref(caps);
2694
    virDomainDefFree(def);
2695 2696 2697
    return rv;
}

2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708
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;
2709
    qemuDomainAsyncJob asyncJob;
2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720

    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;
    }

2721 2722
    qemuMigrationStoreDomainState(vm);

2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745
    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.
         */
2746 2747
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                 qemuMigrationCleanup) < 0)
2748
            goto endjob;
2749
        qemuMigrationJobContinue(vm);
2750 2751 2752 2753
    } else {
        goto endjob;
    }

2754
 cleanup:
2755
    qemuDomObjEndAPI(&vm);
2756 2757
    return xml;

2758
 endjob:
2759 2760 2761 2762
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2763 2764 2765
    goto cleanup;
}

2766

2767 2768
/* Prepare is the first step, and it runs on the destination host.
 */
2769

2770
static void
2771
qemuMigrationPrepareCleanup(virQEMUDriverPtr driver,
2772 2773 2774 2775 2776 2777 2778 2779 2780 2781
                            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));

2782 2783 2784
    virPortAllocatorRelease(driver->migrationPorts, priv->migrationPort);
    priv->migrationPort = 0;

2785 2786 2787 2788 2789
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2790
static int
2791
qemuMigrationPrepareAny(virQEMUDriverPtr driver,
2792 2793 2794 2795 2796
                        virConnectPtr dconn,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
2797
                        virDomainDefPtr *def,
2798
                        const char *origname,
L
liguang 已提交
2799
                        virStreamPtr st,
2800
                        const char *protocol,
2801 2802
                        unsigned short port,
                        bool autoPort,
2803
                        const char *listenAddress,
L
liguang 已提交
2804
                        unsigned long flags)
2805 2806
{
    virDomainObjPtr vm = NULL;
2807
    virObjectEventPtr event = NULL;
2808
    int ret = -1;
2809
    int dataFD[2] = { -1, -1 };
2810
    qemuDomainObjPrivatePtr priv = NULL;
J
Jiri Denemark 已提交
2811
    unsigned long long now;
2812
    qemuMigrationCookiePtr mig = NULL;
2813
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2814
    char *xmlout = NULL;
L
liguang 已提交
2815
    unsigned int cookieFlags;
2816
    virCapsPtr caps = NULL;
J
Ján Tomko 已提交
2817
    char *migrateFrom = NULL;
2818
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2819
    bool taint_hook = false;
2820

2821
    if (virTimeMillisNow(&now) < 0)
2822 2823
        return -1;

2824 2825
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847
    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;
        }
    }

2848 2849 2850
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2851
    if (!qemuMigrationIsAllowed(driver, NULL, *def, true, abort_on_error))
2852 2853
        goto cleanup;

J
Jiri Denemark 已提交
2854 2855 2856 2857 2858
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2859
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2860 2861
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2862 2863
            goto cleanup;

2864
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2865 2866 2867 2868 2869 2870 2871
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2872
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2873 2874 2875 2876 2877 2878
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2879
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt,
J
Jiri Denemark 已提交
2880
                                                 QEMU_EXPECTED_VIRT_TYPES,
2881
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE);
J
Jiri Denemark 已提交
2882 2883 2884
                if (!newdef)
                    goto cleanup;

2885
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2886 2887 2888 2889
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2890 2891
                virDomainDefFree(*def);
                *def = newdef;
2892 2893 2894 2895
                /* 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
2896 2897 2898 2899
            }
        }
    }

J
Ján Tomko 已提交
2900 2901 2902 2903
    if (tunnel) {
        /* QEMU will be started with -incoming stdio
         * (which qemu_command might convert to exec:cat or fd:n)
         */
2904
        if (VIR_STRDUP(migrateFrom, "stdio") < 0)
J
Ján Tomko 已提交
2905 2906
            goto cleanup;
    } else {
2907 2908 2909
        bool encloseAddress = false;
        bool hostIPv6Capable = false;
        bool qemuIPv6Capable = false;
J
Ján Tomko 已提交
2910 2911 2912 2913
        virQEMUCapsPtr qemuCaps = NULL;
        struct addrinfo *info = NULL;
        struct addrinfo hints = { .ai_flags = AI_ADDRCONFIG,
                                  .ai_socktype = SOCK_STREAM };
2914
        const char *incFormat;
J
Ján Tomko 已提交
2915

2916 2917 2918 2919
        if (getaddrinfo("::", NULL, &hints, &info) == 0) {
            freeaddrinfo(info);
            hostIPv6Capable = true;
        }
J
Ján Tomko 已提交
2920
        if (!(qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
2921
                                                    (*def)->emulator)))
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
    if (STREQ_NULLABLE(protocol, "rdma") && !vm->def->mem.hard_limit) {
M
Michael R. Hines 已提交
2992 2993 2994 2995 2996 2997
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2998 2999 3000
    if (qemuMigrationPrecreateStorage(dconn, driver, vm, mig->nbd) < 0)
        goto cleanup;

3001
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
3002
        goto cleanup;
3003
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
3004 3005 3006 3007

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

L
liguang 已提交
3008 3009 3010
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

3011 3012
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
3013 3014
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
3015 3016 3017 3018
        goto endjob;
    }

    /* Start the QEMU daemon, with the same command-line arguments plus
3019
     * -incoming $migrateFrom
3020
     */
3021 3022
    if (qemuProcessStart(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                         migrateFrom, dataFD[0], NULL, NULL,
3023 3024
                         VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
                         VIR_QEMU_PROCESS_START_PAUSED |
3025
                         VIR_QEMU_PROCESS_START_AUTODESTROY) < 0) {
3026
        virDomainAuditStart(vm, "migrated", false);
3027 3028 3029
        goto endjob;
    }

3030 3031 3032 3033
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
3034
            goto stop;
3035
        }
3036
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
3037 3038
    }

3039 3040
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
3041 3042 3043
                                    QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stop;

3044
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
3045 3046 3047 3048
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
        goto stop;
    }

3049 3050 3051
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
                               QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
3052 3053
        goto stop;

3054 3055 3056 3057 3058 3059 3060 3061
    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");
    }
3062

3063
 done:
L
liguang 已提交
3064 3065 3066 3067 3068
    if (flags & VIR_MIGRATE_OFFLINE)
        cookieFlags = 0;
    else
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS;

3069 3070 3071
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
3072
        if (qemuMigrationStartNBDServer(driver, vm, listenAddress) < 0) {
3073 3074
            /* error already reported */
            goto endjob;
3075
        }
3076
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
3077 3078 3079 3080
    }

    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
3081 3082 3083 3084 3085 3086 3087
        /* 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");
    }

3088 3089 3090
    if (qemuDomainCleanupAdd(vm, qemuMigrationPrepareCleanup) < 0)
        goto endjob;

L
liguang 已提交
3091 3092
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
3093
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
3094 3095 3096
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
3097

3098 3099 3100 3101
    /* 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.
     */
3102
    qemuMigrationJobContinue(vm);
3103

3104 3105
    if (autoPort)
        priv->migrationPort = port;
3106
    ret = 0;
3107

3108
 cleanup:
J
Ján Tomko 已提交
3109
    VIR_FREE(migrateFrom);
J
Jiri Denemark 已提交
3110
    VIR_FREE(xmlout);
3111 3112
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
3113 3114 3115
    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 */
3116
        VIR_FREE(priv->origname);
3117 3118 3119
        virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
        priv->nbdPort = 0;
        qemuDomainRemoveInactive(driver, vm);
3120
    }
3121
    qemuDomObjEndAPI(&vm);
3122 3123
    if (event)
        qemuDomainEventQueue(driver, event);
3124
    qemuMigrationCookieFree(mig);
3125
    virObjectUnref(caps);
3126
    virNWFilterUnlockFilterUpdates();
3127
    return ret;
3128

3129
 stop:
3130 3131 3132
    virDomainAuditStart(vm, "migrated", false);
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, 0);

3133
 endjob:
3134
    qemuMigrationJobFinish(driver, vm);
3135
    goto cleanup;
3136 3137 3138
}


3139 3140 3141 3142 3143
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
3144
qemuMigrationPrepareTunnel(virQEMUDriverPtr driver,
3145 3146 3147 3148 3149 3150
                           virConnectPtr dconn,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virStreamPtr st,
3151
                           virDomainDefPtr *def,
3152
                           const char *origname,
L
liguang 已提交
3153
                           unsigned long flags)
3154 3155 3156 3157
{
    int ret;

    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
3158
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
3159
              "origname=%s, flags=%lx",
3160
              driver, dconn, NULLSTR(cookiein), cookieinlen,
3161
              cookieout, cookieoutlen, st, *def, origname, flags);
3162

3163 3164 3165 3166 3167 3168
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

3169
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3170
                                  cookieout, cookieoutlen, def, origname,
3171
                                  st, NULL, 0, false, NULL, flags);
3172 3173 3174 3175
    return ret;
}


3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198
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;
}


3199
int
3200
qemuMigrationPrepareDirect(virQEMUDriverPtr driver,
3201
                           virConnectPtr dconn,
3202 3203 3204 3205
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
3206 3207
                           const char *uri_in,
                           char **uri_out,
3208
                           virDomainDefPtr *def,
3209
                           const char *origname,
3210
                           const char *listenAddress,
L
liguang 已提交
3211
                           unsigned long flags)
3212
{
3213 3214
    unsigned short port = 0;
    bool autoPort = true;
3215 3216
    char *hostname = NULL;
    int ret = -1;
3217
    virURIPtr uri = NULL;
3218 3219
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
3220

3221 3222
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
3223
              "def=%p, origname=%s, listenAddress=%s, flags=%lx",
3224 3225
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
3226
              *def, origname, NULLSTR(listenAddress), flags);
3227

3228 3229
    *uri_out = NULL;

3230 3231 3232
    /* 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
3233 3234 3235
     * 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".
3236 3237 3238 3239 3240 3241
     *
     * 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) {
3242 3243 3244
        bool encloseAddress = false;
        const char *incFormat;

3245
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3246
            goto cleanup;
3247

3248
        if (migrateHost != NULL) {
3249 3250
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
3251

3252
            if (VIR_STRDUP(hostname, migrateHost) < 0)
3253 3254 3255 3256 3257
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
3258 3259

        if (STRPREFIX(hostname, "localhost")) {
3260 3261 3262
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
3263 3264 3265 3266 3267
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
3268
         * compatibility with old targets. We at least make the
3269 3270
         * new targets accept both syntaxes though.
         */
3271 3272 3273 3274 3275 3276
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
3277 3278
            goto cleanup;
    } else {
3279
        bool well_formed_uri;
J
Ján Tomko 已提交
3280

3281 3282
        if (!(uri = qemuMigrationParseURI(uri_in, &well_formed_uri)))
            goto cleanup;
J
Ján Tomko 已提交
3283

3284 3285 3286 3287 3288 3289 3290
        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
3291 3292
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
3293 3294 3295
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
3296 3297 3298 3299 3300 3301 3302 3303 3304 3305
            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) {
3306
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3307
                goto cleanup;
3308

3309
            /* Send well-formed URI only if uri_in was well-formed */
3310 3311 3312 3313 3314 3315 3316 3317
            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;
            }
3318
        } else {
3319 3320
            port = uri->port;
            autoPort = false;
3321 3322 3323 3324 3325 3326
        }
    }

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

3327
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3328
                                  cookieout, cookieoutlen, def, origname,
3329 3330
                                  NULL, uri ? uri->scheme : "tcp",
                                  port, autoPort, listenAddress, flags);
3331
 cleanup:
3332
    virURIFree(uri);
3333
    VIR_FREE(hostname);
3334
    virObjectUnref(cfg);
3335
    if (ret != 0) {
3336
        VIR_FREE(*uri_out);
3337 3338 3339
        if (autoPort)
            virPortAllocatorRelease(driver->migrationPorts, port);
    }
3340 3341 3342 3343
    return ret;
}


3344 3345 3346
virDomainDefPtr
qemuMigrationPrepareDef(virQEMUDriverPtr driver,
                        const char *dom_xml,
3347 3348
                        const char *dname,
                        char **origname)
3349 3350 3351
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
3352
    char *name = NULL;
3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364

    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,
3365
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE)))
3366 3367 3368
        goto cleanup;

    if (dname) {
3369
        name = def->name;
3370 3371 3372 3373 3374 3375
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

3376
 cleanup:
3377
    virObjectUnref(caps);
3378 3379 3380 3381
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
3382 3383 3384 3385
    return def;
}


3386 3387 3388 3389 3390 3391 3392 3393 3394 3395
static int
qemuMigrationConfirmPhase(virQEMUDriverPtr driver,
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *cookiein,
                          int cookieinlen,
                          unsigned int flags,
                          int retcode)
{
    qemuMigrationCookiePtr mig;
3396
    virObjectEventPtr event = NULL;
3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411
    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);

3412 3413
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                       QEMU_MIGRATION_COOKIE_STATS)))
3414 3415
        goto cleanup;

3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432
    /* 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;
        }
    }

3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447
    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");

3448
        event = virDomainEventLifecycleNewFromObj(vm,
3449 3450 3451 3452 3453 3454 3455
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
    } else {

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

3456 3457 3458 3459
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
3460 3461 3462 3463 3464 3465 3466 3467
        }

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

3468
 done:
3469 3470 3471
    qemuMigrationCookieFree(mig);
    rv = 0;

3472
 cleanup:
3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487
    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;
3488
    qemuMigrationJobPhase phase;
3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502
    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);
3503 3504
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
3505 3506 3507 3508 3509

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

3510 3511 3512
    qemuMigrationJobFinish(driver, vm);
    if (!virDomainObjIsActive(vm) &&
        (!vm->persistent || (flags & VIR_MIGRATE_UNDEFINE_SOURCE))) {
3513 3514 3515 3516 3517
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
        qemuDomainRemoveInactive(driver, vm);
    }

3518
 cleanup:
3519
    qemuDomObjEndAPI(&vm);
3520 3521 3522 3523 3524
    virObjectUnref(cfg);
    return ret;
}


3525 3526
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3527
    MIGRATION_DEST_CONNECT_HOST,
3528
    MIGRATION_DEST_UNIX,
3529
    MIGRATION_DEST_FD,
3530
};
3531

3532 3533 3534 3535
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3536

3537 3538 3539 3540 3541 3542
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3543
            const char *protocol;
3544 3545 3546 3547 3548
            const char *name;
            int port;
        } host;

        struct {
3549
            char *file;
3550 3551
            int sock;
        } unix_socket;
3552 3553 3554 3555 3556

        struct {
            int qemu;
            int local;
        } fd;
3557 3558 3559 3560 3561 3562 3563
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3564 3565 3566

#define TUNNEL_SEND_BUF_SIZE 65536

3567 3568 3569 3570 3571 3572 3573
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3574 3575
    int wakeupRecvFD;
    int wakeupSendFD;
3576 3577 3578
};

static void qemuMigrationIOFunc(void *arg)
3579
{
3580
    qemuMigrationIOThreadPtr data = arg;
3581 3582 3583 3584 3585 3586 3587
    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);
3588

3589
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3590
        goto abrt;
3591

3592 3593 3594
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3595
    for (;;) {
3596 3597 3598 3599 3600 3601 3602 3603 3604 3605
        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;
3606
            virReportSystemError(errno, "%s",
3607 3608
                                 _("poll failed in migration tunnel"));
            goto abrt;
3609
        }
3610 3611 3612 3613 3614 3615 3616

        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");
3617
            break;
3618
        }
3619

3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635
        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;
            }
3636 3637
        }

3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654
        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;
            }
        }
    }
3655

3656 3657
    if (virStreamFinish(data->st) < 0)
        goto error;
3658

3659 3660
    VIR_FREE(buffer);

3661 3662
    return;

3663
 abrt:
3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3675
 error:
3676 3677
    virCopyLastError(&data->err);
    virResetLastError();
3678
    VIR_FREE(buffer);
3679 3680 3681 3682 3683 3684 3685
}


static qemuMigrationIOThreadPtr
qemuMigrationStartTunnel(virStreamPtr st,
                         int sock)
{
3686 3687
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3688

3689 3690 3691 3692
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3693 3694
    }

3695
    if (VIR_ALLOC(io) < 0)
3696
        goto error;
3697

3698 3699
    io->st = st;
    io->sock = sock;
3700 3701
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3702 3703 3704 3705 3706 3707

    if (virThreadCreate(&io->thread, true,
                        qemuMigrationIOFunc,
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3708
        goto error;
3709 3710 3711
    }

    return io;
3712

3713
 error:
3714 3715 3716 3717
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3718 3719 3720
}

static int
3721
qemuMigrationStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3722 3723
{
    int rv = -1;
3724 3725 3726 3727 3728 3729 3730 3731 3732
    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;
    }

3733 3734 3735 3736
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3737 3738 3739 3740
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3741 3742 3743 3744 3745 3746
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3747
 cleanup:
3748 3749
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3750 3751
    VIR_FREE(io);
    return rv;
3752 3753
}

3754
static int
3755
qemuMigrationConnect(virQEMUDriverPtr driver,
3756 3757 3758 3759 3760 3761 3762 3763 3764
                     virDomainObjPtr vm,
                     qemuMigrationSpecPtr spec)
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3765
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3766 3767 3768 3769 3770 3771 3772 3773 3774
        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);
3775
        virObjectUnref(sock);
3776 3777 3778 3779 3780
    }
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0 ||
        spec->dest.fd.qemu == -1)
        goto cleanup;

3781 3782 3783 3784 3785 3786 3787
    /* 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;
    }

3788 3789
    ret = 0;

3790
 cleanup:
3791 3792 3793 3794 3795 3796
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3797
static int
3798
qemuMigrationRun(virQEMUDriverPtr driver,
3799 3800 3801 3802 3803 3804 3805
                 virDomainObjPtr vm,
                 const char *cookiein,
                 int cookieinlen,
                 char **cookieout,
                 int *cookieoutlen,
                 unsigned long flags,
                 unsigned long resource,
3806
                 qemuMigrationSpecPtr spec,
3807 3808
                 virConnectPtr dconn,
                 const char *graphicsuri)
3809
{
3810
    int ret = -1;
3811 3812
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3813
    qemuMigrationCookiePtr mig = NULL;
3814
    qemuMigrationIOThreadPtr iothread = NULL;
3815
    int fd = -1;
3816
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3817
    virErrorPtr orig_err = NULL;
3818
    unsigned int cookieFlags = 0;
3819
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3820
    int rc;
3821 3822 3823

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
3824
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s",
3825 3826
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3827 3828
              spec, spec->destType, spec->fwdType, dconn,
              NULLSTR(graphicsuri));
3829

3830 3831 3832 3833 3834 3835 3836 3837 3838 3839
    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;
    }

3840 3841
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3842 3843 3844 3845
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3846 3847 3848
        return -1;
    }

3849 3850 3851
    mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                 cookieFlags | QEMU_MIGRATION_COOKIE_GRAPHICS);
    if (!mig)
3852 3853
        goto cleanup;

3854
    if (qemuDomainMigrateGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3855 3856
        VIR_WARN("unable to provide data for graphics client relocation");

3857 3858 3859 3860 3861 3862 3863
    /* 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;
    }

3864
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3865 3866 3867 3868 3869 3870
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

3871 3872
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
3873 3874 3875
                                    QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3876 3877
    if (qemuMigrationSetAutoConverge(driver, vm,
                                     flags & VIR_MIGRATE_AUTO_CONVERGE,
3878 3879 3880
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3881 3882
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
3883 3884 3885
                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3886 3887
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3888 3889
        goto cleanup;

3890 3891 3892 3893
    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 */
3894
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
3895
        priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
3896 3897 3898 3899 3900 3901
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
        goto cleanup;
    }

3902 3903
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
3904

3905 3906
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3907
        qemuMigrationConnect(driver, vm, spec) < 0) {
3908
        goto exit_monitor;
3909
    }
3910

3911 3912
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3913 3914
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
3915
            goto exit_monitor;
M
Michael R. Hines 已提交
3916
        }
3917
        ret = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
3918
                                       spec->dest.host.protocol,
3919 3920 3921 3922
                                       spec->dest.host.name,
                                       spec->dest.host.port);
        break;

3923 3924 3925 3926
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3927
    case MIGRATION_DEST_UNIX:
3928
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX)) {
3929 3930 3931 3932 3933 3934 3935 3936 3937
            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;
3938 3939

    case MIGRATION_DEST_FD:
3940
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3941
            fd = spec->dest.fd.local;
3942 3943
            spec->dest.fd.local = -1;
        }
3944 3945 3946 3947
        ret = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                     spec->dest.fd.qemu);
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3948
    }
3949 3950
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3951
    if (ret < 0)
3952 3953
        goto cleanup;
    ret = -1;
3954 3955 3956 3957

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

3958 3959 3960 3961 3962 3963 3964 3965
    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;
3966

3967 3968 3969 3970 3971 3972 3973
        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;
        }
3974 3975
    }

3976 3977
    if (spec->fwdType != MIGRATION_FWD_DIRECT &&
        !(iothread = qemuMigrationStartTunnel(spec->fwd.stream, fd)))
3978 3979
        goto cancel;

3980 3981 3982 3983 3984 3985
    rc = qemuMigrationWaitForCompletion(driver, vm,
                                        QEMU_ASYNC_JOB_MIGRATION_OUT,
                                        dconn, abort_on_error);
    if (rc == -2)
        goto cancel;
    else if (rc == -1)
3986
        goto cleanup;
3987

3988 3989 3990 3991 3992 3993 3994 3995 3996
    /* 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) {
3997 3998
        if (qemuMigrationSetOffline(driver, vm) < 0) {
            priv->job.current->type = VIR_DOMAIN_JOB_FAILED;
3999
            goto cleanup;
4000
        }
4001 4002
    }

4003
    ret = 0;
4004

4005
 cleanup:
4006 4007 4008
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

4009
    /* cancel any outstanding NBD jobs */
4010
    if (mig)
4011
        ignore_value(qemuMigrationCancelDriveMirror(mig, driver, vm));
4012

4013
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
4014
        if (iothread && qemuMigrationStopTunnel(iothread, ret < 0) < 0)
4015
            ret = -1;
4016
        VIR_FORCE_CLOSE(fd);
4017
    }
4018

4019
    if (priv->job.completed) {
4020
        qemuDomainJobInfoUpdateTime(priv->job.completed);
4021 4022
        qemuDomainJobInfoUpdateDowntime(priv->job.completed);
    }
4023

4024 4025 4026
    if (priv->job.current->type == VIR_DOMAIN_JOB_UNBOUNDED)
        priv->job.current->type = VIR_DOMAIN_JOB_FAILED;

4027 4028
    cookieFlags |= QEMU_MIGRATION_COOKIE_NETWORK |
                   QEMU_MIGRATION_COOKIE_STATS;
4029 4030
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookieFlags |= QEMU_MIGRATION_COOKIE_PERSISTENT;
4031
    if (ret == 0 &&
4032 4033
        qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
4034
        VIR_WARN("Unable to encode migration cookie");
4035
    }
4036

4037 4038
    qemuMigrationCookieFree(mig);

4039 4040 4041 4042 4043
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4044 4045
    return ret;

4046 4047 4048 4049
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;

4050
 cancel:
4051 4052
    orig_err = virSaveLastError();

4053
    if (virDomainObjIsActive(vm)) {
4054 4055
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
4056
            qemuMonitorMigrateCancel(priv->mon);
4057
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
4058
        }
4059
    }
4060 4061 4062
    goto cleanup;
}

4063
/* Perform migration using QEMU's native migrate support,
4064 4065
 * not encrypted obviously
 */
4066
static int doNativeMigrate(virQEMUDriverPtr driver,
4067 4068 4069 4070 4071 4072 4073
                           virDomainObjPtr vm,
                           const char *uri,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4074
                           unsigned long resource,
4075 4076
                           virConnectPtr dconn,
                           const char *graphicsuri)
4077
{
4078
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
4079
    virURIPtr uribits = NULL;
4080
    int ret = -1;
4081 4082 4083
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
4084 4085
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "graphicsuri=%s",
4086
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
4087 4088
              cookieout, cookieoutlen, flags, resource,
              NULLSTR(graphicsuri));
4089

4090
    if (!(uribits = qemuMigrationParseURI(uri, NULL)))
4091 4092
        return -1;

4093 4094 4095 4096 4097 4098 4099
    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
4100 4101 4102 4103 4104
    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"));
4105
            goto cleanup;
M
Michael R. Hines 已提交
4106 4107 4108 4109 4110
        }
        if (!vm->def->mem.hard_limit) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
4111
            goto cleanup;
M
Michael R. Hines 已提交
4112 4113 4114 4115 4116
        }
    }

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
        STRNEQ(uribits->scheme, "rdma"))
4117 4118
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
4119
        spec.destType = MIGRATION_DEST_HOST;
4120
    spec.dest.host.protocol = uribits->scheme;
4121 4122 4123
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
4124

4125
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4126 4127
                           cookieoutlen, flags, resource, &spec, dconn,
                           graphicsuri);
4128 4129 4130 4131

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

4132
 cleanup:
4133
    virURIFree(uribits);
4134 4135 4136 4137 4138

    return ret;
}


4139
static int doTunnelMigrate(virQEMUDriverPtr driver,
4140 4141 4142 4143 4144 4145 4146
                           virDomainObjPtr vm,
                           virStreamPtr st,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4147
                           unsigned long resource,
4148 4149
                           virConnectPtr dconn,
                           const char *graphicsuri)
4150 4151
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
4152
    virNetSocketPtr sock = NULL;
4153 4154
    int ret = -1;
    qemuMigrationSpec spec;
4155
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4156 4157

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
4158 4159
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "graphicsuri=%s",
4160
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
4161 4162
              cookieout, cookieoutlen, flags, resource,
              NULLSTR(graphicsuri));
4163

4164 4165 4166
    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)) {
4167 4168
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Source qemu is too old to support tunnelled migration"));
4169
        virObjectUnref(cfg);
4170
        return -1;
4171 4172 4173 4174 4175
    }

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

4176
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD)) {
4177 4178 4179 4180 4181 4182
        int fds[2];

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

4183
        if (pipe2(fds, O_CLOEXEC) == 0) {
4184 4185 4186 4187
            spec.dest.fd.qemu = fds[1];
            spec.dest.fd.local = fds[0];
        }
        if (spec.dest.fd.qemu == -1 ||
4188
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
4189
                                              spec.dest.fd.qemu) < 0) {
4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200
            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",
4201
                        cfg->libDir, vm->def->name) < 0)
4202 4203 4204
            goto cleanup;

        if (virNetSocketNewListenUNIX(spec.dest.unix_socket.file, 0700,
4205
                                      cfg->user, cfg->group,
4206 4207 4208 4209 4210 4211 4212
                                      &sock) < 0 ||
            virNetSocketListen(sock, 1) < 0)
            goto cleanup;

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

4213
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4214 4215
                           cookieoutlen, flags, resource, &spec, dconn,
                           graphicsuri);
4216

4217
 cleanup:
4218 4219 4220 4221
    if (spec.destType == MIGRATION_DEST_FD) {
        VIR_FORCE_CLOSE(spec.dest.fd.qemu);
        VIR_FORCE_CLOSE(spec.dest.fd.local);
    } else {
4222
        virObjectUnref(sock);
4223 4224
        VIR_FREE(spec.dest.unix_socket.file);
    }
4225

4226
    virObjectUnref(cfg);
4227 4228 4229 4230
    return ret;
}


4231 4232 4233 4234
/* 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 */
4235
static int doPeer2PeerMigrate2(virQEMUDriverPtr driver,
4236
                               virConnectPtr sconn ATTRIBUTE_UNUSED,
4237 4238
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
4239
                               const char *dconnuri,
4240 4241 4242
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
4243 4244 4245
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
4246
    char *cookie = NULL;
4247 4248 4249
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
4250
    bool cancelled;
4251
    virStreamPtr st = NULL;
4252
    unsigned long destflags;
4253

4254
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
4255
              "flags=%lx, dname=%s, resource=%lu",
4256 4257
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
4258

4259 4260 4261 4262 4263
    /* 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,
4264 4265
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
4266 4267 4268 4269 4270
        return -1;

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

4271 4272
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4273

4274 4275 4276 4277 4278 4279 4280 4281 4282 4283
    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;

4284
        qemuDomainObjEnterRemote(vm);
4285
        ret = dconn->driver->domainMigratePrepareTunnel
4286
            (dconn, st, destflags, dname, resource, dom_xml);
4287
        qemuDomainObjExitRemote(vm);
4288
    } else {
4289
        qemuDomainObjEnterRemote(vm);
4290 4291
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
4292
             destflags, dname, resource, dom_xml);
4293
        qemuDomainObjExitRemote(vm);
4294 4295 4296
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
4297 4298 4299
        goto cleanup;

    /* the domain may have shutdown or crashed while we had the locks dropped
4300
     * in qemuDomainObjEnterRemote, so check again
4301 4302
     */
    if (!virDomainObjIsActive(vm)) {
4303 4304
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4305 4306 4307
        goto cleanup;
    }

4308 4309
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
4310 4311
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
4312
        cancelled = true;
4313
        orig_err = virSaveLastError();
4314
        goto finish;
4315 4316
    }

4317 4318 4319 4320
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
4321
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4322
    if (flags & VIR_MIGRATE_TUNNELLED)
4323 4324
        ret = doTunnelMigrate(driver, vm, st,
                              NULL, 0, NULL, NULL,
4325
                              flags, resource, dconn, NULL);
4326 4327 4328 4329
    else
        ret = doNativeMigrate(driver, vm, uri_out,
                              cookie, cookielen,
                              NULL, NULL, /* No out cookie with v2 migration */
4330
                              flags, resource, dconn, NULL);
4331 4332 4333 4334

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4335

4336 4337 4338
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4339
    cancelled = ret < 0;
4340

4341
 finish:
4342 4343 4344 4345
    /* 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.
     */
4346
    dname = dname ? dname : vm->def->name;
4347
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4348
    qemuDomainObjEnterRemote(vm);
4349
    ddomain = dconn->driver->domainMigrateFinish2
4350
        (dconn, dname, cookie, cookielen,
4351
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4352
    qemuDomainObjExitRemote(vm);
4353 4354
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4355

4356
 cleanup:
4357
    if (ddomain) {
4358
        virObjectUnref(ddomain);
4359 4360 4361 4362
        ret = 0;
    } else {
        ret = -1;
    }
4363

4364
    virObjectUnref(st);
4365 4366 4367 4368 4369 4370

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4371
    VIR_FREE(cookie);
4372 4373

    return ret;
4374 4375 4376
}


4377 4378 4379 4380
/* 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 */
4381 4382 4383 4384 4385 4386 4387 4388 4389
static int
doPeer2PeerMigrate3(virQEMUDriverPtr driver,
                    virConnectPtr sconn,
                    virConnectPtr dconn,
                    const char *dconnuri,
                    virDomainObjPtr vm,
                    const char *xmlin,
                    const char *dname,
                    const char *uri,
4390
                    const char *graphicsuri,
4391
                    const char *listenAddress,
4392 4393 4394
                    unsigned long long bandwidth,
                    bool useParams,
                    unsigned long flags)
4395 4396 4397 4398 4399 4400 4401 4402 4403 4404
{
    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;
4405
    bool cancelled = true;
4406
    virStreamPtr st = NULL;
4407
    unsigned long destflags;
4408 4409 4410 4411 4412
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4413 4414
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
              "bandwidth=%llu, useParams=%d, flags=%lx",
4415
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4416 4417
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
              NULLSTR(listenAddress), bandwidth, useParams, flags);
4418

4419 4420 4421 4422 4423
    /* 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.  */

4424 4425
    dom_xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
                                      &cookieout, &cookieoutlen, flags);
4426 4427 4428
    if (!dom_xml)
        goto cleanup;

4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448
    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;
4449 4450 4451 4452 4453 4454

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4455 4456 4457 4458 4459
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4460 4461
    }

4462 4463 4464
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4465 4466
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4467

4468 4469 4470 4471 4472 4473 4474 4475 4476
    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;

4477
        qemuDomainObjEnterRemote(vm);
4478 4479 4480 4481 4482 4483 4484 4485 4486
        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);
        }
4487
        qemuDomainObjExitRemote(vm);
4488
    } else {
4489
        qemuDomainObjEnterRemote(vm);
4490 4491 4492 4493 4494 4495 4496 4497 4498
        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);
        }
4499
        qemuDomainObjExitRemote(vm);
4500 4501 4502 4503 4504
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

L
liguang 已提交
4505 4506 4507 4508
    if (flags & VIR_MIGRATE_OFFLINE) {
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4509
        cancelled = false;
L
liguang 已提交
4510 4511 4512
        goto finish;
    }

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

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4553
    if (ret < 0) {
4554
        orig_err = virSaveLastError();
4555 4556 4557 4558
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4559 4560 4561 4562

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4563
    cancelled = ret < 0;
4564

4565
 finish:
4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577
    /*
     * 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;
4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600

    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);
    }
4601 4602
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4603

4604 4605 4606 4607 4608 4609 4610
    /* 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.
4611
     */
4612
    cancelled = ddomain == NULL;
4613

4614 4615 4616 4617 4618 4619
    /* 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();

4620 4621 4622 4623
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4624
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4625 4626 4627 4628 4629
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4630 4631 4632
    ret = qemuMigrationConfirmPhase(driver, sconn, vm,
                                    cookiein, cookieinlen,
                                    flags, cancelled);
4633 4634 4635 4636
    /* 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.
     */
4637 4638 4639
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4640 4641 4642

 cleanup:
    if (ddomain) {
4643
        virObjectUnref(ddomain);
4644 4645 4646 4647 4648
        ret = 0;
    } else {
        ret = -1;
    }

4649
    virObjectUnref(st);
4650 4651 4652 4653 4654 4655 4656 4657

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4658
    virTypedParamsFree(params, nparams);
4659 4660 4661 4662
    return ret;
}


4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4675
static int doPeer2PeerMigrate(virQEMUDriverPtr driver,
4676
                              virConnectPtr sconn,
4677
                              virDomainObjPtr vm,
4678
                              const char *xmlin,
4679
                              const char *dconnuri,
4680
                              const char *uri,
4681
                              const char *graphicsuri,
4682
                              const char *listenAddress,
4683 4684
                              unsigned long flags,
                              const char *dname,
4685 4686
                              unsigned long resource,
                              bool *v3proto)
4687 4688 4689 4690
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4691
    virErrorPtr orig_err = NULL;
4692
    bool offline = false;
4693
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4694
    bool useParams;
4695

4696
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4697 4698
              "uri=%s, graphicsuri=%s, listenAddress=%s, flags=%lx, "
              "dname=%s, resource=%lu",
4699
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4700 4701
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
              flags, NULLSTR(dname), resource);
4702 4703 4704 4705 4706

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

4707
    qemuDomainObjEnterRemote(vm);
4708
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4709
    qemuDomainObjExitRemote(vm);
4710
    if (dconn == NULL) {
4711
        virReportError(VIR_ERR_OPERATION_FAILED,
4712 4713
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4714
        virObjectUnref(cfg);
4715 4716 4717
        return -1;
    }

4718 4719
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4720 4721
        goto cleanup;

4722
    qemuDomainObjEnterRemote(vm);
4723 4724
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4725
        /* v3proto reflects whether the caller used Perform3, but with
4726
         * p2p migrate, regardless of whether Perform2 or Perform3
4727 4728 4729 4730
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4731 4732
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
L
liguang 已提交
4733 4734 4735
    if (flags & VIR_MIGRATE_OFFLINE)
        offline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                           VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4736
    qemuDomainObjExitRemote(vm);
4737

4738
    if (!p2p) {
4739 4740
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4741 4742 4743
        goto cleanup;
    }

4744 4745
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4746
    if (!useParams && graphicsuri) {
4747 4748 4749 4750 4751 4752
        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
4753 4754 4755 4756 4757 4758 4759
    if (flags & VIR_MIGRATE_OFFLINE && !offline) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4760
    /* domain may have been stopped while we were talking to remote daemon */
L
liguang 已提交
4761
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4762 4763
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4764 4765 4766
        goto cleanup;
    }

4767 4768 4769 4770 4771 4772 4773
    /* 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;

4774 4775
    if (*v3proto) {
        ret = doPeer2PeerMigrate3(driver, sconn, dconn, dconnuri, vm, xmlin,
4776 4777
                                  dname, uri, graphicsuri, listenAddress,
                                  resource, useParams, flags);
4778
    } else {
4779
        ret = doPeer2PeerMigrate2(driver, sconn, dconn, vm,
4780
                                  dconnuri, flags, dname, resource);
4781
    }
4782

4783
 cleanup:
4784
    orig_err = virSaveLastError();
4785
    qemuDomainObjEnterRemote(vm);
4786
    virObjectUnref(dconn);
4787
    qemuDomainObjExitRemote(vm);
4788 4789 4790 4791
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4792
    virObjectUnref(cfg);
4793 4794 4795 4796
    return ret;
}


4797 4798 4799 4800 4801 4802
/*
 * 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
4803
qemuMigrationPerformJob(virQEMUDriverPtr driver,
4804 4805 4806 4807 4808
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dconnuri,
                        const char *uri,
4809
                        const char *graphicsuri,
4810
                        const char *listenAddress,
4811
                        const char *cookiein,
4812 4813 4814 4815 4816 4817 4818
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4819
{
4820
    virObjectEventPtr event = NULL;
4821
    int ret = -1;
4822
    virErrorPtr orig_err = NULL;
4823
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4824
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
4825

4826
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4827 4828
        goto cleanup;

L
liguang 已提交
4829
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4830 4831
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
4832 4833 4834
        goto endjob;
    }

4835
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
4836
        goto endjob;
4837

4838
    if (!(flags & VIR_MIGRATE_UNSAFE) && !qemuMigrationIsSafe(vm->def))
4839
        goto endjob;
4840

4841
    qemuMigrationStoreDomainState(vm);
4842 4843

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4844
        ret = doPeer2PeerMigrate(driver, conn, vm, xmlin,
4845 4846
                                 dconnuri, uri, graphicsuri, listenAddress,
                                 flags, dname, resource, &v3proto);
4847
    } else {
4848 4849 4850
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
        ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                              cookieout, cookieoutlen,
4851
                              flags, resource, NULL, NULL);
4852
    }
4853 4854
    if (ret < 0)
        goto endjob;
4855

4856 4857 4858 4859
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4860
    if (!v3proto) {
4861 4862
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4863
        virDomainAuditStop(vm, "migrated");
4864
        event = virDomainEventLifecycleNewFromObj(vm,
4865 4866
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4867 4868
    }

4869
 endjob:
4870 4871 4872
    if (ret < 0)
        orig_err = virSaveLastError();

4873
    if (qemuMigrationRestoreDomainState(conn, vm)) {
4874
        event = virDomainEventLifecycleNewFromObj(vm,
4875 4876 4877
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
4878

4879 4880 4881 4882
    qemuMigrationJobFinish(driver, vm);
    if (!virDomainObjIsActive(vm) &&
        (!vm->persistent ||
         (ret == 0 && (flags & VIR_MIGRATE_UNDEFINE_SOURCE)))) {
4883
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
4884
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4885
        qemuDomainRemoveInactive(driver, vm);
4886 4887
    }

4888 4889 4890 4891 4892
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4893
 cleanup:
4894
    qemuDomObjEndAPI(&vm);
4895 4896
    if (event)
        qemuDomainEventQueue(driver, event);
4897
    virObjectUnref(cfg);
4898 4899 4900 4901 4902 4903 4904
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4905
qemuMigrationPerformPhase(virQEMUDriverPtr driver,
4906 4907 4908
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *uri,
4909
                          const char *graphicsuri,
4910 4911 4912 4913 4914 4915 4916
                          const char *cookiein,
                          int cookieinlen,
                          char **cookieout,
                          int *cookieoutlen,
                          unsigned long flags,
                          unsigned long resource)
{
4917
    virObjectEventPtr event = NULL;
4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928
    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);
4929 4930
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
4931 4932 4933

    ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                          cookieout, cookieoutlen,
4934
                          flags, resource, NULL, graphicsuri);
4935

4936 4937 4938 4939 4940
    if (ret < 0) {
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
4941
        }
4942
        goto endjob;
4943
    }
4944 4945 4946

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4947 4948
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                             qemuMigrationCleanup) < 0)
4949 4950
        goto endjob;

4951
 endjob:
4952
    if (ret < 0)
4953
        qemuMigrationJobFinish(driver, vm);
4954
    else
4955 4956
        qemuMigrationJobContinue(vm);
    if (!virDomainObjIsActive(vm) && !vm->persistent)
4957
        qemuDomainRemoveInactive(driver, vm);
4958

4959
 cleanup:
4960
    qemuDomObjEndAPI(&vm);
4961 4962 4963 4964 4965
    if (event)
        qemuDomainEventQueue(driver, event);
    return ret;
}

4966
int
4967
qemuMigrationPerform(virQEMUDriverPtr driver,
4968 4969 4970 4971 4972
                     virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *xmlin,
                     const char *dconnuri,
                     const char *uri,
4973
                     const char *graphicsuri,
4974
                     const char *listenAddress,
4975 4976 4977 4978 4979 4980 4981 4982 4983 4984
                     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, "
4985
              "uri=%s, graphicsuri=%s, listenAddress=%s"
4986 4987
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
              "flags=%lx, dname=%s, resource=%lu, v3proto=%d",
4988
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4989
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4990 4991
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4992 4993 4994

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4995 4996
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4997 4998 4999 5000
            return -1;
        }

        return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri, uri,
5001 5002
                                       graphicsuri, listenAddress,
                                       cookiein, cookieinlen,
5003 5004
                                       cookieout, cookieoutlen,
                                       flags, dname, resource, v3proto);
5005 5006
    } else {
        if (dconnuri) {
5007 5008
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
5009 5010 5011 5012 5013
            return -1;
        }

        if (v3proto) {
            return qemuMigrationPerformPhase(driver, conn, vm, uri,
5014
                                             graphicsuri,
5015 5016
                                             cookiein, cookieinlen,
                                             cookieout, cookieoutlen,
5017
                                             flags, resource);
5018 5019
        } else {
            return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri,
5020
                                           uri, graphicsuri, listenAddress,
5021
                                           cookiein, cookieinlen,
5022 5023 5024 5025 5026
                                           cookieout, cookieoutlen, flags,
                                           dname, resource, v3proto);
        }
    }
}
5027

5028
static int
5029 5030
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def)
{
5031
    size_t i;
5032 5033 5034 5035 5036
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
5037
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5038
            if (virNetDevVPortProfileAssociate(net->ifname,
5039
                                               virDomainNetGetActualVirtPortProfile(net),
5040
                                               &net->mac,
5041
                                               virDomainNetGetActualDirectDev(net),
5042
                                               -1,
5043
                                               def->uuid,
5044 5045
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
5046 5047 5048
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
5049
                goto err_exit;
5050 5051
            }
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
5052

5053
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
5054 5055 5056 5057
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
5058 5059 5060 5061
        }
        last_good_net = i;
    }

5062
    return 0;
5063

5064
 err_exit:
5065
    for (i = 0; last_good_net != -1 && i < last_good_net; i++) {
5066
        net = def->nets[i];
5067
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5068
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
5069
                                                           virDomainNetGetActualVirtPortProfile(net),
5070
                                                           &net->mac,
5071
                                                           virDomainNetGetActualDirectDev(net),
5072
                                                           -1,
5073
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
5074 5075
        }
    }
5076
    return -1;
5077 5078 5079 5080
}


virDomainPtr
5081
qemuMigrationFinish(virQEMUDriverPtr driver,
5082 5083
                    virConnectPtr dconn,
                    virDomainObjPtr vm,
5084 5085 5086 5087
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
5088
                    unsigned long flags,
5089 5090
                    int retcode,
                    bool v3proto)
5091 5092
{
    virDomainPtr dom = NULL;
5093
    virObjectEventPtr event = NULL;
5094
    bool newVM = true;
5095
    qemuMigrationCookiePtr mig = NULL;
5096
    virErrorPtr orig_err = NULL;
5097
    int cookie_flags = 0;
J
Jiri Denemark 已提交
5098
    qemuDomainObjPrivatePtr priv = vm->privateData;
5099
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5100
    virCapsPtr caps = NULL;
5101
    unsigned short port;
5102

5103
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
5104
              "cookieout=%p, cookieoutlen=%p, flags=%lx, retcode=%d",
5105 5106
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
5107

5108 5109 5110
    port = priv->migrationPort;
    priv->migrationPort = 0;

5111 5112 5113
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

5114
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
5115 5116
        goto cleanup;

5117 5118 5119
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
5120

5121
    qemuDomainCleanupRemove(vm, qemuMigrationPrepareCleanup);
5122
    VIR_FREE(priv->job.completed);
5123

5124
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
5125 5126
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
5127 5128 5129 5130 5131
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookie_flags |= QEMU_MIGRATION_COOKIE_PERSISTENT;

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein,
                                       cookieinlen, cookie_flags)))
5132
        goto endjob;
5133 5134 5135 5136 5137

    /* 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
5138
        if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
5139 5140
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("guest unexpectedly quit"));
5141
            goto endjob;
5142 5143
        }

5144 5145 5146 5147 5148
        if (mig->jobInfo) {
            priv->job.completed = mig->jobInfo;
            mig->jobInfo = NULL;
        }

L
liguang 已提交
5149 5150 5151 5152 5153
        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");
5154
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5155 5156 5157 5158 5159 5160 5161
                                                 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");
5162
        }
5163

5164 5165
        if (qemuMigrationStopNBDServer(driver, vm, mig) < 0)
            goto endjob;
5166

5167
        if (flags & VIR_MIGRATE_PERSIST_DEST) {
5168
            virDomainDefPtr vmdef;
5169
            if (vm->persistent)
5170
                newVM = false;
5171
            vm->persistent = 1;
5172 5173 5174
            if (mig->persistent)
                vm->newDef = vmdef = mig->persistent;
            else
5175
                vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm);
5176
            if (!vmdef || virDomainSaveConfig(cfg->configDir, vmdef) < 0) {
5177 5178 5179 5180 5181 5182 5183 5184 5185
                /* 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.
                 */
5186 5187

                /*
5188 5189
                 * However, in v3 protocol, the source VM is still available
                 * to restart during confirm() step, so we kill it off now.
5190 5191
                 */
                if (v3proto) {
L
liguang 已提交
5192 5193 5194 5195 5196
                    if (!(flags & VIR_MIGRATE_OFFLINE)) {
                        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                        VIR_QEMU_PROCESS_STOP_MIGRATED);
                        virDomainAuditStop(vm, "failed");
                    }
5197 5198
                    if (newVM)
                        vm->persistent = 0;
5199
                }
A
Alex Jia 已提交
5200
                if (!vmdef)
5201 5202
                    virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                                   _("can't get vmdef"));
5203 5204 5205
                goto endjob;
            }

5206
            event = virDomainEventLifecycleNewFromObj(vm,
5207 5208 5209 5210 5211 5212 5213 5214 5215
                                             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
5216
        if (!(flags & VIR_MIGRATE_PAUSED) && !(flags & VIR_MIGRATE_OFFLINE)) {
5217 5218 5219 5220
            /* 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
5221
            if (qemuProcessStartCPUs(driver, vm, dconn,
5222 5223
                                     VIR_DOMAIN_RUNNING_MIGRATED,
                                     QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5224
                if (virGetLastError() == NULL)
5225 5226
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   "%s", _("resume operation failed"));
5227 5228 5229 5230 5231
                /* Need to save the current error, in case shutting
                 * down the process overwrites it
                 */
                orig_err = virSaveLastError();

5232 5233 5234 5235 5236 5237 5238 5239 5240
                /*
                 * 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) {
5241 5242
                    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                    VIR_QEMU_PROCESS_STOP_MIGRATED);
5243
                    virDomainAuditStop(vm, "failed");
5244
                    event = virDomainEventLifecycleNewFromObj(vm,
5245 5246 5247
                                                     VIR_DOMAIN_EVENT_STOPPED,
                                                     VIR_DOMAIN_EVENT_STOPPED_FAILED);
                }
5248 5249
                goto endjob;
            }
5250 5251 5252 5253
            if (priv->job.completed) {
                qemuDomainJobInfoUpdateTime(priv->job.completed);
                qemuDomainJobInfoUpdateDowntime(priv->job.completed);
            }
5254 5255
        }

5256
        dom = virGetDomain(dconn, vm->def->name, vm->def->uuid);
5257

L
liguang 已提交
5258
        if (!(flags & VIR_MIGRATE_OFFLINE)) {
5259
            event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5260 5261 5262 5263 5264 5265 5266
                                             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);
5267
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5268 5269 5270
                                                 VIR_DOMAIN_EVENT_SUSPENDED,
                                                 VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);
            }
5271
        }
L
liguang 已提交
5272 5273

        if (virDomainObjIsActive(vm) &&
5274
            virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
5275 5276 5277
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto endjob;
        }
5278 5279 5280

        /* Guest is successfully running, so cancel previous auto destroy */
        qemuProcessAutoDestroyRemove(driver, vm);
L
liguang 已提交
5281
    } else if (!(flags & VIR_MIGRATE_OFFLINE)) {
5282 5283
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
5284
        virDomainAuditStop(vm, "failed");
5285
        event = virDomainEventLifecycleNewFromObj(vm,
5286 5287 5288 5289
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_FAILED);
    }

5290 5291
    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen,
                                QEMU_MIGRATION_COOKIE_STATS) < 0)
5292 5293
        VIR_WARN("Unable to encode migration cookie");

5294
 endjob:
5295 5296
    qemuMigrationJobFinish(driver, vm);
    if (!vm->persistent && !virDomainObjIsActive(vm))
5297
        qemuDomainRemoveInactive(driver, vm);
5298

5299
 cleanup:
5300
    virPortAllocatorRelease(driver->migrationPorts, port);
5301 5302 5303 5304
    if (priv->mon)
        qemuMonitorSetDomainLog(priv->mon, -1);
    VIR_FREE(priv->origname);
    qemuDomObjEndAPI(&vm);
5305 5306
    if (event)
        qemuDomainEventQueue(driver, event);
5307
    qemuMigrationCookieFree(mig);
5308 5309 5310 5311
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5312
    virObjectUnref(caps);
5313
    virObjectUnref(cfg);
5314 5315
    return dom;
}
5316

5317

5318
/* Helper function called while vm is active.  */
5319
int
5320
qemuMigrationToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
5321 5322
                    int fd, off_t offset, const char *path,
                    const char *compressor,
E
Eric Blake 已提交
5323
                    bool bypassSecurityDriver,
5324
                    qemuDomainAsyncJob asyncJob)
5325 5326 5327
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5328
    int ret = -1;
5329
    bool restoreLabel = false;
5330 5331
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5332
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5333
    char *errbuf = NULL;
5334
    virErrorPtr orig_err = NULL;
5335 5336 5337 5338 5339

    /* 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,
5340 5341
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5342 5343
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5344
    }
5345

5346 5347 5348 5349 5350 5351 5352
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5353
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5354
        (!compressor || pipe(pipeFD) == 0)) {
5355
        /* All right! We can use fd migration, which means that qemu
5356 5357 5358
         * 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.  */
5359
        if (virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
5360
                                              compressor ? pipeFD[1] : fd) < 0)
5361
            goto cleanup;
5362 5363 5364
        bypassSecurityDriver = true;
    } else {
        /* Phooey - we have to fall back on exec migration, where qemu
E
Eric Blake 已提交
5365 5366
         * has to popen() the file by name, and block devices have to be
         * given cgroup ACL permission.  We might also stumble on
5367 5368
         * a race present in some qemu versions where it does a wait()
         * that botches pclose.  */
5369 5370
        if (virCgroupHasController(priv->cgroup,
                                   VIR_CGROUP_CONTROLLER_DEVICES)) {
5371 5372 5373 5374
            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
5375
                /* path was not a device, no further need for cgroup */
5376
            } else if (rv < 0) {
5377 5378
                goto cleanup;
            }
5379
        }
5380 5381
        if ((!bypassSecurityDriver) &&
            virSecurityManagerSetSavedStateLabel(driver->securityManager,
5382
                                                 vm->def, path) < 0)
5383
            goto cleanup;
5384
        restoreLabel = true;
5385 5386
    }

5387
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5388 5389
        goto cleanup;

5390 5391 5392
    if (!compressor) {
        const char *args[] = { "cat", NULL };

5393
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409
            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
        };
5410 5411 5412 5413
        if (pipeFD[0] != -1) {
            cmd = virCommandNewArgs(args);
            virCommandSetInputFD(cmd, pipeFD[0]);
            virCommandSetOutputFD(cmd, &fd);
5414 5415
            virCommandSetErrorBuffer(cmd, &errbuf);
            virCommandDoAsyncIO(cmd);
5416 5417 5418
            if (virSetCloseExec(pipeFD[1]) < 0) {
                virReportSystemError(errno, "%s",
                                     _("Unable to set cloexec flag"));
5419
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5420 5421 5422
                goto cleanup;
            }
            if (virCommandRunAsync(cmd, NULL) < 0) {
5423
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5424 5425 5426 5427 5428 5429 5430
                goto cleanup;
            }
            rc = qemuMonitorMigrateToFd(priv->mon,
                                        QEMU_MONITOR_MIGRATE_BACKGROUND,
                                        pipeFD[1]);
            if (VIR_CLOSE(pipeFD[0]) < 0 ||
                VIR_CLOSE(pipeFD[1]) < 0)
5431
                VIR_WARN("failed to close intermediate pipe");
5432 5433 5434 5435 5436
        } else {
            rc = qemuMonitorMigrateToFile(priv->mon,
                                          QEMU_MONITOR_MIGRATE_BACKGROUND,
                                          args, path, offset);
        }
5437
    }
5438
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5439
        goto cleanup;
5440 5441 5442
    if (rc < 0)
        goto cleanup;

5443
    rc = qemuMigrationWaitForCompletion(driver, vm, asyncJob, NULL, false);
5444

5445 5446 5447 5448
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5449 5450
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5451
                qemuMonitorMigrateCancel(priv->mon);
5452
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5453 5454
            }
        }
5455
        goto cleanup;
5456
    }
5457

5458 5459 5460
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5461 5462
    ret = 0;

5463
 cleanup:
5464 5465 5466
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5467
    /* Restore max migration bandwidth */
5468 5469
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5470 5471
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5472
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5473 5474
    }

5475 5476
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5477 5478 5479 5480 5481
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5482 5483
    if (restoreLabel && (!bypassSecurityDriver) &&
        virSecurityManagerRestoreSavedStateLabel(driver->securityManager,
5484
                                                 vm->def, path) < 0)
5485 5486
        VIR_WARN("failed to restore save state label on %s", path);

5487 5488
    if (virCgroupHasController(priv->cgroup,
                               VIR_CGROUP_CONTROLLER_DEVICES)) {
5489 5490 5491
        int rv = virCgroupDenyDevicePath(priv->cgroup, path,
                                         VIR_CGROUP_DEVICE_RWM);
        virDomainAuditCgroupPath(vm, priv->cgroup, "deny", path, "rwm", rv == 0);
5492
    }
5493 5494 5495 5496 5497 5498

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

5499 5500
    return ret;
}
5501 5502

int
5503
qemuMigrationJobStart(virQEMUDriverPtr driver,
5504
                      virDomainObjPtr vm,
5505
                      qemuDomainAsyncJob job)
5506 5507 5508
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

5509
    if (qemuDomainObjBeginAsyncJob(driver, vm, job) < 0)
5510 5511
        return -1;

5512
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5513
        qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
5514
    } else {
5515 5516 5517
        qemuDomainObjSetAsyncJobMask(vm, (QEMU_JOB_DEFAULT_MASK |
                                          JOB_MASK(QEMU_JOB_SUSPEND) |
                                          JOB_MASK(QEMU_JOB_MIGRATION_OP)));
5518
    }
5519

J
Jiri Denemark 已提交
5520
    priv->job.current->type = VIR_DOMAIN_JOB_UNBOUNDED;
5521 5522 5523 5524 5525

    return 0;
}

void
5526
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5527
                         virDomainObjPtr vm,
5528
                         qemuMigrationJobPhase phase)
5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542
{
    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
5543
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5544
                           virDomainObjPtr vm,
5545
                           qemuMigrationJobPhase phase)
5546 5547 5548 5549
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5550
void
5551 5552
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5553
    qemuDomainObjReleaseAsyncJob(vm);
5554 5555 5556 5557
}

bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
5558
                         qemuDomainAsyncJob job)
5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569
{
    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");

5570
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5571 5572 5573 5574 5575
        return false;
    }
    return true;
}

5576
void
5577
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5578
{
5579
    qemuDomainObjEndAsyncJob(driver, vm);
5580
}