qemu_migration.c 174.5 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 62 63

#define VIR_FROM_THIS VIR_FROM_QEMU

64 65
VIR_LOG_INIT("qemu.qemu_migration");

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

79
enum qemuMigrationCookieFlags {
80
    QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS,
81
    QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE,
82
    QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT,
83
    QEMU_MIGRATION_COOKIE_FLAG_NETWORK,
84
    QEMU_MIGRATION_COOKIE_FLAG_NBD,
85
    QEMU_MIGRATION_COOKIE_FLAG_STATS,
86 87 88 89 90 91 92

    QEMU_MIGRATION_COOKIE_FLAG_LAST
};

VIR_ENUM_DECL(qemuMigrationCookieFlag);
VIR_ENUM_IMPL(qemuMigrationCookieFlag,
              QEMU_MIGRATION_COOKIE_FLAG_LAST,
93 94 95 96
              "graphics",
              "lockstate",
              "persistent",
              "network",
97 98
              "nbd",
              "statistics");
99 100 101

enum qemuMigrationCookieFeatures {
    QEMU_MIGRATION_COOKIE_GRAPHICS  = (1 << QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS),
102
    QEMU_MIGRATION_COOKIE_LOCKSTATE = (1 << QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE),
103
    QEMU_MIGRATION_COOKIE_PERSISTENT = (1 << QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT),
104
    QEMU_MIGRATION_COOKIE_NETWORK = (1 << QEMU_MIGRATION_COOKIE_FLAG_NETWORK),
105
    QEMU_MIGRATION_COOKIE_NBD = (1 << QEMU_MIGRATION_COOKIE_FLAG_NBD),
106
    QEMU_MIGRATION_COOKIE_STATS = (1 << QEMU_MIGRATION_COOKIE_FLAG_STATS),
107 108 109 110 111 112 113 114 115 116 117 118
};

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

119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139
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;
};

140 141 142 143 144 145
typedef struct _qemuMigrationCookieNBD qemuMigrationCookieNBD;
typedef qemuMigrationCookieNBD *qemuMigrationCookieNBDPtr;
struct _qemuMigrationCookieNBD {
    int port; /* on which port does NBD server listen for incoming data */
};

146 147 148
typedef struct _qemuMigrationCookie qemuMigrationCookie;
typedef qemuMigrationCookie *qemuMigrationCookiePtr;
struct _qemuMigrationCookie {
E
Eric Blake 已提交
149 150
    unsigned int flags;
    unsigned int flagsMandatory;
151 152

    /* Host properties */
153 154 155 156
    unsigned char localHostuuid[VIR_UUID_BUFLEN];
    unsigned char remoteHostuuid[VIR_UUID_BUFLEN];
    char *localHostname;
    char *remoteHostname;
157 158 159 160

    /* Guest properties */
    unsigned char uuid[VIR_UUID_BUFLEN];
    char *name;
161

162 163 164 165
    /* If (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) */
    char *lockState;
    char *lockDriver;

166 167
    /* If (flags & QEMU_MIGRATION_COOKIE_GRAPHICS) */
    qemuMigrationCookieGraphicsPtr graphics;
168 169 170

    /* If (flags & QEMU_MIGRATION_COOKIE_PERSISTENT) */
    virDomainDefPtr persistent;
171 172 173

    /* If (flags & QEMU_MIGRATION_COOKIE_NETWORK) */
    qemuMigrationCookieNetworkPtr network;
174 175 176

    /* If (flags & QEMU_MIGRATION_COOKIE_NBD) */
    qemuMigrationCookieNBDPtr nbd;
177 178 179

    /* If (flags & QEMU_MIGRATION_COOKIE_STATS) */
    qemuDomainJobInfoPtr jobInfo;
180 181
};

182 183 184 185 186 187 188 189 190
static void qemuMigrationCookieGraphicsFree(qemuMigrationCookieGraphicsPtr grap)
{
    if (!grap)
        return;
    VIR_FREE(grap->listen);
    VIR_FREE(grap->tlsSubject);
    VIR_FREE(grap);
}

191

192 193 194
static void
qemuMigrationCookieNetworkFree(qemuMigrationCookieNetworkPtr network)
{
195
    size_t i;
196 197 198 199 200 201 202 203 204 205 206 207 208

    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);
}


209 210 211 212 213
static void qemuMigrationCookieFree(qemuMigrationCookiePtr mig)
{
    if (!mig)
        return;

214 215
    qemuMigrationCookieGraphicsFree(mig->graphics);
    qemuMigrationCookieNetworkFree(mig->network);
216

217 218
    VIR_FREE(mig->localHostname);
    VIR_FREE(mig->remoteHostname);
219
    VIR_FREE(mig->name);
220 221
    VIR_FREE(mig->lockState);
    VIR_FREE(mig->lockDriver);
222
    VIR_FREE(mig->nbd);
223
    VIR_FREE(mig->jobInfo);
224 225 226 227
    VIR_FREE(mig);
}


228
#ifdef WITH_GNUTLS
229 230 231 232 233 234 235 236 237 238 239 240
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)
241
        goto error;
242 243

    if (virFileReadAll(certfile, 8192, &pemdata) < 0) {
244 245
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unable to read server cert %s"), certfile);
246 247 248 249 250
        goto error;
    }

    ret = gnutls_x509_crt_init(&cert);
    if (ret < 0) {
251 252 253
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot initialize cert object: %s"),
                       gnutls_strerror(ret));
254 255 256 257 258 259 260 261
        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) {
262 263 264
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot load cert data from %s: %s"),
                       certfile, gnutls_strerror(ret));
265 266 267 268 269
        goto error;
    }

    subjectlen = 1024;
    if (VIR_ALLOC_N(subject, subjectlen+1) < 0)
270
        goto error;
271 272 273 274 275 276 277 278 279

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

    VIR_FREE(certfile);
    VIR_FREE(pemdata);

    return subject;

280
 error:
281 282 283 284
    VIR_FREE(certfile);
    VIR_FREE(pemdata);
    return NULL;
}
285
#endif
286 287

static qemuMigrationCookieGraphicsPtr
288
qemuMigrationCookieGraphicsAlloc(virQEMUDriverPtr driver,
289 290 291 292
                                 virDomainGraphicsDefPtr def)
{
    qemuMigrationCookieGraphicsPtr mig = NULL;
    const char *listenAddr;
293
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
294 295

    if (VIR_ALLOC(mig) < 0)
296
        goto error;
297 298 299 300

    mig->type = def->type;
    if (mig->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
        mig->port = def->data.vnc.port;
301
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
302
        if (!listenAddr)
303
            listenAddr = cfg->vncListen;
304

305
#ifdef WITH_GNUTLS
306 307
        if (cfg->vncTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->vncTLSx509certdir)))
308
            goto error;
309
#endif
310 311
    } else {
        mig->port = def->data.spice.port;
312
        if (cfg->spiceTLS)
313 314 315
            mig->tlsPort = def->data.spice.tlsPort;
        else
            mig->tlsPort = -1;
316
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
317
        if (!listenAddr)
318
            listenAddr = cfg->spiceListen;
319

320
#ifdef WITH_GNUTLS
321 322
        if (cfg->spiceTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->spiceTLSx509certdir)))
323
            goto error;
324
#endif
325
    }
326 327
    if (VIR_STRDUP(mig->listen, listenAddr) < 0)
        goto error;
328

329
    virObjectUnref(cfg);
330 331
    return mig;

332
 error:
333
    qemuMigrationCookieGraphicsFree(mig);
334
    virObjectUnref(cfg);
335 336 337 338
    return NULL;
}


339
static qemuMigrationCookieNetworkPtr
340
qemuMigrationCookieNetworkAlloc(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
341 342 343
                                virDomainDefPtr def)
{
    qemuMigrationCookieNetworkPtr mig;
344
    size_t i;
345 346

    if (VIR_ALLOC(mig) < 0)
347
        goto error;
348 349 350 351

    mig->nnets = def->nnets;

    if (VIR_ALLOC_N(mig->net, def->nnets) <0)
352
        goto error;
353 354 355 356 357 358 359 360 361 362 363 364 365 366 367

    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:
368
               break;
369
            case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
370 371 372 373 374 375 376 377
                if (virNetDevOpenvswitchGetMigrateData(&mig->net[i].portdata,
                                                       netptr->ifname) != 0) {
                        virReportSystemError(VIR_ERR_INTERNAL_ERROR,
                                             _("Unable to run command to get OVS port data for "
                                             "interface %s"), netptr->ifname);
                        goto error;
                }
                break;
378 379 380 381 382 383 384
            default:
                break;
            }
        }
    }
    return mig;

385
 error:
386 387 388 389
    qemuMigrationCookieNetworkFree(mig);
    return NULL;
}

390 391 392
static qemuMigrationCookiePtr
qemuMigrationCookieNew(virDomainObjPtr dom)
{
J
Jiri Denemark 已提交
393
    qemuDomainObjPrivatePtr priv = dom->privateData;
394
    qemuMigrationCookiePtr mig = NULL;
J
Jiri Denemark 已提交
395
    const char *name;
396 397

    if (VIR_ALLOC(mig) < 0)
398
        goto error;
399

J
Jiri Denemark 已提交
400 401 402 403
    if (priv->origname)
        name = priv->origname;
    else
        name = dom->def->name;
404 405
    if (VIR_STRDUP(mig->name, name) < 0)
        goto error;
406 407
    memcpy(mig->uuid, dom->def->uuid, VIR_UUID_BUFLEN);

408
    if (!(mig->localHostname = virGetHostname()))
409
        goto error;
410
    if (virGetHostUUID(mig->localHostuuid) < 0) {
411 412
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Unable to obtain host UUID"));
413 414 415 416 417
        goto error;
    }

    return mig;

418
 error:
419 420 421 422 423
    qemuMigrationCookieFree(mig);
    return NULL;
}


424 425
static int
qemuMigrationCookieAddGraphics(qemuMigrationCookiePtr mig,
426
                               virQEMUDriverPtr driver,
427 428
                               virDomainObjPtr dom)
{
429 430
    size_t i = 0;

431
    if (mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) {
432 433
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration graphics data already present"));
434 435 436
        return -1;
    }

437 438 439 440 441 442 443 444
    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;
       }
445
    }
446 447 448 449 450

    return 0;
}


451 452
static int
qemuMigrationCookieAddLockstate(qemuMigrationCookiePtr mig,
453
                                virQEMUDriverPtr driver,
454 455 456 457 458
                                virDomainObjPtr dom)
{
    qemuDomainObjPrivatePtr priv = dom->privateData;

    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
459 460
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration lockstate data already present"));
461 462 463 464
        return -1;
    }

    if (virDomainObjGetState(dom, NULL) == VIR_DOMAIN_PAUSED) {
465
        if (VIR_STRDUP(mig->lockState, priv->lockState) < 0)
466 467 468 469 470 471
            return -1;
    } else {
        if (virDomainLockProcessInquire(driver->lockManager, dom, &mig->lockState) < 0)
            return -1;
    }

472
    if (VIR_STRDUP(mig->lockDriver, virLockManagerPluginGetName(driver->lockManager)) < 0) {
473 474 475 476 477 478 479 480 481 482 483
        VIR_FREE(mig->lockState);
        return -1;
    }

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

    return 0;
}


484 485 486 487 488
static int
qemuMigrationCookieAddPersistent(qemuMigrationCookiePtr mig,
                                 virDomainObjPtr dom)
{
    if (mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) {
489 490
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration persistent data already present"));
491 492 493 494 495 496 497 498 499 500 501 502 503
        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;
}


504 505
static int
qemuMigrationCookieAddNetwork(qemuMigrationCookiePtr mig,
506
                              virQEMUDriverPtr driver,
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
                              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;
}

525

526 527 528 529 530 531 532 533 534
static int
qemuMigrationCookieAddNBD(qemuMigrationCookiePtr mig,
                          virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    /* It is not a bug if there already is a NBD data */
    if (!mig->nbd &&
535
        VIR_ALLOC(mig->nbd) < 0)
536 537 538 539 540 541 542 543 544
        return -1;

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

    return 0;
}


545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563
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;
}


564 565 566
static void qemuMigrationCookieGraphicsXMLFormat(virBufferPtr buf,
                                                 qemuMigrationCookieGraphicsPtr grap)
{
567
    virBufferAsprintf(buf, "<graphics type='%s' port='%d' listen='%s'",
568 569 570 571 572 573
                      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");
574 575 576 577
        virBufferAdjustIndent(buf, 2);
        virBufferEscapeString(buf, "<cert info='subject' value='%s'/>\n", grap->tlsSubject);
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</graphics>\n");
578 579 580 581 582 583
    } else {
        virBufferAddLit(buf, "/>\n");
    }
}


584 585 586 587
static void
qemuMigrationCookieNetworkXMLFormat(virBufferPtr buf,
                                    qemuMigrationCookieNetworkPtr optr)
{
588
    size_t i;
589 590 591 592 593 594
    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) {
595 596
                virBufferAddLit(buf, "<network>\n");
                virBufferAdjustIndent(buf, 2);
597 598
                empty = false;
            }
599
            virBufferAsprintf(buf, "<interface index='%zu' vporttype='%s'",
600 601 602
                              i, virNetDevVPortTypeToString(optr->net[i].vporttype));
            if (optr->net[i].portdata) {
                virBufferAddLit(buf, ">\n");
603 604
                virBufferAdjustIndent(buf, 2);
                virBufferEscapeString(buf, "<portdata>%s</portdata>\n",
605
                                      optr->net[i].portdata);
606 607
                virBufferAdjustIndent(buf, -2);
                virBufferAddLit(buf, "</interface>\n");
608 609 610 611 612
            } else {
                virBufferAddLit(buf, "/>\n");
            }
        }
    }
613 614 615 616
    if (!empty) {
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</network>\n");
    }
617 618 619
}


620 621 622 623 624 625 626 627 628
static void
qemuMigrationCookieStatisticsXMLFormat(virBufferPtr buf,
                                       qemuDomainJobInfoPtr jobInfo)
{
    qemuMonitorMigrationStatus *status = &jobInfo->status;

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

629 630 631
    virBufferAsprintf(buf, "<started>%llu</started>\n", jobInfo->started);
    virBufferAsprintf(buf, "<stopped>%llu</stopped>\n", jobInfo->stopped);

632 633 634 635 636 637 638 639 640 641
    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$s>\n",
                          VIR_DOMAIN_JOB_DOWNTIME,
                          status->downtime);
642 643 644 645
    if (status->setup_time_set)
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_SETUP_TIME,
                          status->setup_time);
646 647 648 649 650 651 652 653 654 655

    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$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_REMAINING,
                      status->ram_remaining);
656 657 658
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_BPS,
                      status->ram_bps);
659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680

    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$s>\n",
                      VIR_DOMAIN_JOB_DISK_REMAINING,
                      status->disk_remaining);
681 682 683
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_BPS,
                      status->disk_bps);
684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707

    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");
}


708
static int
709
qemuMigrationCookieXMLFormat(virQEMUDriverPtr driver,
710
                             virBufferPtr buf,
711
                             qemuMigrationCookiePtr mig)
712 713 714
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
    char hostuuidstr[VIR_UUID_STRING_BUFLEN];
715
    size_t i;
716 717

    virUUIDFormat(mig->uuid, uuidstr);
718
    virUUIDFormat(mig->localHostuuid, hostuuidstr);
719

720
    virBufferAddLit(buf, "<qemu-migration>\n");
721 722 723 724 725
    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);
726

727
    for (i = 0; i < QEMU_MIGRATION_COOKIE_FLAG_LAST; i++) {
728
        if (mig->flagsMandatory & (1 << i))
729
            virBufferAsprintf(buf, "<feature name='%s'/>\n",
730 731 732
                              qemuMigrationCookieFlagTypeToString(i));
    }

733 734
    if ((mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        mig->graphics)
735 736
        qemuMigrationCookieGraphicsXMLFormat(buf, mig->graphics);

737 738
    if ((mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        mig->lockState) {
739
        virBufferAsprintf(buf, "<lockstate driver='%s'>\n",
740
                          mig->lockDriver);
741 742
        virBufferAdjustIndent(buf, 2);
        virBufferAsprintf(buf, "<leases>%s</leases>\n",
743
                          mig->lockState);
744 745
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</lockstate>\n");
746 747
    }

748 749
    if ((mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        mig->persistent) {
750 751 752
        if (qemuDomainDefFormatBuf(driver,
                                   mig->persistent,
                                   VIR_DOMAIN_XML_INACTIVE |
753 754
                                   VIR_DOMAIN_XML_SECURE |
                                   VIR_DOMAIN_XML_MIGRATABLE,
755
                                   buf) < 0)
756
            return -1;
757 758
    }

759 760 761
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NETWORK) && mig->network)
        qemuMigrationCookieNetworkXMLFormat(buf, mig->network);

762
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NBD) && mig->nbd) {
763
        virBufferAddLit(buf, "<nbd");
764 765 766 767 768
        if (mig->nbd->port)
            virBufferAsprintf(buf, " port='%d'", mig->nbd->port);
        virBufferAddLit(buf, "/>\n");
    }

769 770 771
    if (mig->flags & QEMU_MIGRATION_COOKIE_STATS && mig->jobInfo)
        qemuMigrationCookieStatisticsXMLFormat(buf, mig->jobInfo);

772
    virBufferAdjustIndent(buf, -2);
773
    virBufferAddLit(buf, "</qemu-migration>\n");
774
    return 0;
775 776 777
}


778
static char *qemuMigrationCookieXMLFormatStr(virQEMUDriverPtr driver,
779
                                             qemuMigrationCookiePtr mig)
780 781 782
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;

783
    if (qemuMigrationCookieXMLFormat(driver, &buf, mig) < 0) {
784 785 786
        virBufferFreeAndReset(&buf);
        return NULL;
    }
787

788
    if (virBufferCheckError(&buf) < 0)
789 790 791 792 793 794
        return NULL;

    return virBufferContentAndReset(&buf);
}


795 796 797 798 799 800 801
static qemuMigrationCookieGraphicsPtr
qemuMigrationCookieGraphicsXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieGraphicsPtr grap;
    char *tmp;

    if (VIR_ALLOC(grap) < 0)
802
        goto error;
803 804

    if (!(tmp = virXPathString("string(./graphics/@type)", ctxt))) {
805 806
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing type attribute in migration data"));
807 808 809
        goto error;
    }
    if ((grap->type = virDomainGraphicsTypeFromString(tmp)) < 0) {
810 811
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unknown graphics type %s"), tmp);
812 813 814
        VIR_FREE(tmp);
        goto error;
    }
E
Eric Blake 已提交
815
    VIR_FREE(tmp);
816
    if (virXPathInt("string(./graphics/@port)", ctxt, &grap->port) < 0) {
817 818
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing port attribute in migration data"));
819 820 821 822
        goto error;
    }
    if (grap->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        if (virXPathInt("string(./graphics/@tlsPort)", ctxt, &grap->tlsPort) < 0) {
823 824
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing tlsPort attribute in migration data"));
825 826 827 828
            goto error;
        }
    }
    if (!(grap->listen = virXPathString("string(./graphics/@listen)", ctxt))) {
829 830
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing listen attribute in migration data"));
831 832 833
        goto error;
    }
    /* Optional */
834
    grap->tlsSubject = virXPathString("string(./graphics/cert[@info='subject']/@value)", ctxt);
835 836 837

    return grap;

838
 error:
839 840 841 842 843
    qemuMigrationCookieGraphicsFree(grap);
    return NULL;
}


844 845 846 847
static qemuMigrationCookieNetworkPtr
qemuMigrationCookieNetworkXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieNetworkPtr optr;
848
    size_t i;
849 850 851 852 853 854
    int n;
    xmlNodePtr *interfaces = NULL;
    char *vporttype;
    xmlNodePtr save_ctxt = ctxt->node;

    if (VIR_ALLOC(optr) < 0)
855
        goto error;
856 857 858 859 860 861 862 863

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

    optr->nnets = n;
864
    if (VIR_ALLOC_N(optr->net, optr->nnets) < 0)
865
        goto error;
866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881

    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);

882
 cleanup:
883 884 885
    ctxt->node = save_ctxt;
    return optr;

886
 error:
887 888 889 890 891 892 893
    VIR_FREE(interfaces);
    qemuMigrationCookieNetworkFree(optr);
    optr = NULL;
    goto cleanup;
}


894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909
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;

910 911 912
    virXPathULongLong("string(./started[1])", ctxt, &jobInfo->started);
    virXPathULongLong("string(./stopped[1])", ctxt, &jobInfo->stopped);

913 914 915 916 917 918 919
    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;
920 921 922
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_SETUP_TIME "[1])",
                          ctxt, &status->setup_time) == 0)
        status->setup_time_set = true;
923 924 925 926 927 928 929

    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);
930 931
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_BPS "[1])",
                      ctxt, &status->ram_bps);
932 933 934 935 936 937 938 939 940 941 942 943 944 945 946

    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);
947 948
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_BPS "[1])",
                      ctxt, &status->disk_bps);
949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967

    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;
}


968 969
static int
qemuMigrationCookieXMLParse(qemuMigrationCookiePtr mig,
970
                            virQEMUDriverPtr driver,
971
                            xmlDocPtr doc,
972
                            xmlXPathContextPtr ctxt,
E
Eric Blake 已提交
973
                            unsigned int flags)
974 975
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
976
    char *tmp = NULL;
977
    xmlNodePtr *nodes = NULL;
978 979
    size_t i;
    int n;
980 981 982 983
    virCapsPtr caps = NULL;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto error;
984 985 986 987 988 989 990 991

    /* 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))) {
992 993
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing name element in migration data"));
994 995 996
        goto error;
    }
    if (STRNEQ(tmp, mig->name)) {
997 998 999
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected name %s vs %s"),
                       tmp, mig->name);
1000 1001 1002 1003 1004 1005 1006
        goto error;
    }
    VIR_FREE(tmp);

    /* Extract domain uuid */
    tmp = virXPathString("string(./uuid[1])", ctxt);
    if (!tmp) {
1007 1008
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing uuid element in migration data"));
1009 1010 1011 1012
        goto error;
    }
    virUUIDFormat(mig->uuid, uuidstr);
    if (STRNEQ(tmp, uuidstr)) {
1013 1014 1015
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected UUID %s vs %s"),
                       tmp, uuidstr);
1016 1017 1018 1019
    }
    VIR_FREE(tmp);

    /* Check & forbid "localhost" migration */
1020
    if (!(mig->remoteHostname = virXPathString("string(./hostname[1])", ctxt))) {
1021 1022
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostname element in migration data"));
1023 1024
        goto error;
    }
1025
    if (STREQ(mig->remoteHostname, mig->localHostname)) {
1026 1027 1028
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Attempt to migrate guest to the same host %s"),
                       mig->remoteHostname);
1029 1030 1031 1032
        goto error;
    }

    if (!(tmp = virXPathString("string(./hostuuid[1])", ctxt))) {
1033 1034 1035
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostuuid element in migration data"));
        goto error;
1036
    }
1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
    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);
1049

1050 1051
    /* Check to ensure all mandatory features from XML are also
     * present in 'flags' */
1052
    if ((n = virXPathNodeSet("./feature", ctxt, &nodes)) < 0)
1053 1054
        goto error;

1055
    for (i = 0; i < n; i++) {
1056 1057 1058
        int val;
        char *str = virXMLPropString(nodes[i], "name");
        if (!str) {
1059 1060
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing feature name"));
1061 1062 1063 1064
            goto error;
        }

        if ((val = qemuMigrationCookieFlagTypeFromString(str)) < 0) {
1065 1066 1067
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unknown migration cookie feature %s"),
                           str);
1068 1069 1070 1071 1072
            VIR_FREE(str);
            goto error;
        }

        if ((flags & (1 << val)) == 0) {
1073 1074 1075
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unsupported migration cookie feature %s"),
                           str);
1076 1077 1078 1079 1080 1081
            VIR_FREE(str);
        }
        VIR_FREE(str);
    }
    VIR_FREE(nodes);

1082 1083 1084 1085 1086
    if ((flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        virXPathBoolean("count(./graphics) > 0", ctxt) &&
        (!(mig->graphics = qemuMigrationCookieGraphicsXMLParse(ctxt))))
        goto error;

1087 1088 1089 1090
    if ((flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        virXPathBoolean("count(./lockstate) > 0", ctxt)) {
        mig->lockDriver = virXPathString("string(./lockstate[1]/@driver)", ctxt);
        if (!mig->lockDriver) {
1091 1092
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Missing lock driver name in migration cookie"));
1093 1094 1095 1096 1097 1098 1099
            goto error;
        }
        mig->lockState = virXPathString("string(./lockstate[1]/leases[1])", ctxt);
        if (mig->lockState && STREQ(mig->lockState, ""))
            VIR_FREE(mig->lockState);
    }

1100 1101 1102
    if ((flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        virXPathBoolean("count(./domain) > 0", ctxt)) {
        if ((n = virXPathNodeSet("./domain", ctxt, &nodes)) > 1) {
1103 1104 1105 1106
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Too many domain elements in "
                             "migration cookie: %d"),
                           n);
1107 1108
            goto error;
        }
1109 1110
        mig->persistent = virDomainDefParseNode(doc, nodes[0],
                                                caps, driver->xmlopt,
1111 1112 1113 1114 1115 1116 1117 1118 1119
                                                -1, VIR_DOMAIN_XML_INACTIVE);
        if (!mig->persistent) {
            /* virDomainDefParseNode already reported
             * an error for us */
            goto error;
        }
        VIR_FREE(nodes);
    }

1120 1121 1122 1123 1124
    if ((flags & QEMU_MIGRATION_COOKIE_NETWORK) &&
        virXPathBoolean("count(./network) > 0", ctxt) &&
        (!(mig->network = qemuMigrationCookieNetworkXMLParse(ctxt))))
        goto error;

1125 1126 1127 1128
    if (flags & QEMU_MIGRATION_COOKIE_NBD &&
        virXPathBoolean("boolean(./nbd)", ctxt)) {
        char *port;

1129
        if (VIR_ALLOC(mig->nbd) < 0)
1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
            goto error;

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

1143 1144 1145 1146 1147
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        virXPathBoolean("boolean(./statistics)", ctxt) &&
        (!(mig->jobInfo = qemuMigrationCookieStatisticsXMLParse(ctxt))))
        goto error;

1148
    virObjectUnref(caps);
1149 1150
    return 0;

1151
 error:
1152
    VIR_FREE(tmp);
1153
    VIR_FREE(nodes);
1154
    virObjectUnref(caps);
1155 1156 1157 1158 1159 1160
    return -1;
}


static int
qemuMigrationCookieXMLParseStr(qemuMigrationCookiePtr mig,
1161
                               virQEMUDriverPtr driver,
1162
                               const char *xml,
E
Eric Blake 已提交
1163
                               unsigned int flags)
1164 1165 1166
{
    xmlDocPtr doc = NULL;
    xmlXPathContextPtr ctxt = NULL;
1167
    int ret = -1;
1168 1169 1170

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

1171
    if (!(doc = virXMLParseStringCtxt(xml, _("(qemu_migration_cookie)"), &ctxt)))
1172 1173
        goto cleanup;

1174
    ret = qemuMigrationCookieXMLParse(mig, driver, doc, ctxt, flags);
1175

1176
 cleanup:
1177 1178 1179 1180 1181 1182 1183 1184 1185
    xmlXPathFreeContext(ctxt);
    xmlFreeDoc(doc);

    return ret;
}


static int
qemuMigrationBakeCookie(qemuMigrationCookiePtr mig,
1186
                        virQEMUDriverPtr driver,
1187
                        virDomainObjPtr dom,
1188 1189
                        char **cookieout,
                        int *cookieoutlen,
E
Eric Blake 已提交
1190
                        unsigned int flags)
1191
{
1192 1193
    if (!cookieout || !cookieoutlen)
        return 0;
1194 1195 1196

    *cookieoutlen = 0;

1197 1198 1199 1200
    if (flags & QEMU_MIGRATION_COOKIE_GRAPHICS &&
        qemuMigrationCookieAddGraphics(mig, driver, dom) < 0)
        return -1;

1201 1202 1203 1204
    if (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE &&
        qemuMigrationCookieAddLockstate(mig, driver, dom) < 0)
        return -1;

1205 1206 1207 1208
    if (flags & QEMU_MIGRATION_COOKIE_PERSISTENT &&
        qemuMigrationCookieAddPersistent(mig, dom) < 0)
        return -1;

1209 1210 1211 1212 1213
    if (flags & QEMU_MIGRATION_COOKIE_NETWORK &&
        qemuMigrationCookieAddNetwork(mig, driver, dom) < 0) {
        return -1;
    }

1214 1215 1216 1217
    if ((flags & QEMU_MIGRATION_COOKIE_NBD) &&
        qemuMigrationCookieAddNBD(mig, driver, dom) < 0)
        return -1;

1218 1219 1220 1221
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        qemuMigrationCookieAddStatistics(mig, dom) < 0)
        return -1;

1222
    if (!(*cookieout = qemuMigrationCookieXMLFormatStr(driver, mig)))
1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233
        return -1;

    *cookieoutlen = strlen(*cookieout) + 1;

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

    return 0;
}


static qemuMigrationCookiePtr
1234
qemuMigrationEatCookie(virQEMUDriverPtr driver,
1235
                       virDomainObjPtr dom,
1236 1237
                       const char *cookiein,
                       int cookieinlen,
E
Eric Blake 已提交
1238
                       unsigned int flags)
1239 1240 1241 1242 1243 1244
{
    qemuMigrationCookiePtr mig = NULL;

    /* Parse & validate incoming cookie (if any) */
    if (cookiein && cookieinlen &&
        cookiein[cookieinlen-1] != '\0') {
1245 1246
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration cookie was not NULL terminated"));
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
        goto error;
    }

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

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

    if (cookiein && cookieinlen &&
        qemuMigrationCookieXMLParseStr(mig,
1257
                                       driver,
1258 1259 1260 1261
                                       cookiein,
                                       flags) < 0)
        goto error;

1262 1263 1264
    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
        if (!mig->lockDriver) {
            if (virLockManagerPluginUsesState(driver->lockManager)) {
1265 1266 1267
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Missing %s lock state for migration cookie"),
                               virLockManagerPluginGetName(driver->lockManager));
1268 1269 1270 1271
                goto error;
            }
        } else if (STRNEQ(mig->lockDriver,
                          virLockManagerPluginGetName(driver->lockManager))) {
1272 1273 1274 1275
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Source host lock driver %s different from target %s"),
                           mig->lockDriver,
                           virLockManagerPluginGetName(driver->lockManager));
1276 1277 1278 1279
            goto error;
        }
    }

1280 1281
    return mig;

1282
 error:
1283 1284 1285
    qemuMigrationCookieFree(mig);
    return NULL;
}
1286

1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333
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;
}

1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347
/**
 * 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
1348 1349
                            virDomainObjPtr vm,
                            const char *listenAddr)
1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360
{
    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 */
1361 1362
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1363 1364 1365 1366
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1367
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1368 1369 1370 1371 1372 1373 1374
            goto cleanup;

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

        if (!port &&
1375
            ((virPortAllocatorAcquire(driver->migrationPorts, &port) < 0) ||
1376
             (qemuMonitorNBDServerStart(priv->mon, listenAddr, port) < 0))) {
1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390
            qemuDomainObjExitMonitor(driver, vm);
            goto cleanup;
        }

        if (qemuMonitorNBDServerAdd(priv->mon, diskAlias, true) < 0) {
            qemuDomainObjExitMonitor(driver, vm);
            goto cleanup;
        }
        qemuDomainObjExitMonitor(driver, vm);
    }

    priv->nbdPort = port;
    ret = 0;

1391
 cleanup:
1392
    VIR_FREE(diskAlias);
1393
    if (ret < 0)
1394
        virPortAllocatorRelease(driver->migrationPorts, port);
1395 1396 1397
    return ret;
}

1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431
/**
 * 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;
1432
    char *hoststr = NULL;
1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451
    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;

1452 1453
    /* escape literal IPv6 address */
    if (strchr(host, ':')) {
1454
        if (virAsprintf(&hoststr, "[%s]", host) < 0)
1455 1456 1457 1458 1459
            goto error;
    } else if (VIR_STRDUP(hoststr, host) < 0) {
        goto error;
    }

1460 1461 1462 1463 1464 1465 1466 1467
    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 */
1468 1469
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1470 1471 1472 1473 1474 1475 1476
            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",
1477
                         hoststr, port, diskAlias) < 0))
1478 1479 1480 1481 1482 1483
            goto error;

        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            goto error;
        mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest,
1484
                                         NULL, speed, 0, 0, mirror_flags);
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
        qemuDomainObjExitMonitor(driver, vm);

        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 */
                qemuDomainObjExitMonitor(driver, vm);
                virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                               qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                               _("canceled by client"));
                goto error;
            }
1512 1513
            mon_ret = qemuMonitorBlockJobInfo(priv->mon, diskAlias, &info,
                                              NULL);
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
            qemuDomainObjExitMonitor(driver, vm);

            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;

1541
 cleanup:
1542 1543
    VIR_FREE(diskAlias);
    VIR_FREE(nbd_dest);
1544
    VIR_FREE(hoststr);
1545 1546
    return ret;

1547
 error:
1548 1549 1550 1551 1552 1553 1554
    /* don't overwrite any errors */
    err = virSaveLastError();
    /* cancel any outstanding jobs */
    while (lastGood) {
        virDomainDiskDefPtr disk = vm->def->disks[--lastGood];

        /* skip shared, RO disks */
1555 1556
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1557 1558 1559 1560
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1561
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1562 1563 1564
            continue;
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
1565
            if (qemuMonitorBlockJob(priv->mon, diskAlias, NULL, NULL, 0,
1566
                                    BLOCK_JOB_ABORT, true) < 0) {
1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578
                VIR_WARN("Unable to cancel block-job on '%s'", diskAlias);
            }
            qemuDomainObjExitMonitor(driver, vm);
        } else {
            VIR_WARN("Unable to enter monitor. No block job cancelled");
        }
    }
    if (err)
        virSetError(err);
    virFreeError(err);
    goto cleanup;
}
1579

1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599

static void
qemuMigrationStopNBDServer(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           qemuMigrationCookiePtr mig)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (!mig->nbd)
        return;

    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        return;

    if (qemuMonitorNBDServerStop(priv->mon) < 0)
        VIR_WARN("Unable to stop NBD server");

    qemuDomainObjExitMonitor(driver, vm);

1600
    virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
1601 1602 1603
    priv->nbdPort = 0;
}

1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618
static void
qemuMigrationCancelDriveMirror(qemuMigrationCookiePtr mig,
                               virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    size_t i;
    char *diskAlias = NULL;

    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 */
1619 1620
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1621 1622 1623 1624
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1625
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1626 1627 1628 1629 1630 1631
            goto cleanup;

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

1632
        if (qemuMonitorBlockJob(priv->mon, diskAlias, NULL, NULL, 0,
1633
                                BLOCK_JOB_ABORT, true) < 0)
1634 1635 1636 1637
            VIR_WARN("Unable to stop block job on %s", diskAlias);
        qemuDomainObjExitMonitor(driver, vm);
    }

1638
 cleanup:
1639 1640 1641 1642
    VIR_FREE(diskAlias);
    return;
}

1643 1644 1645 1646 1647 1648 1649 1650 1651
/* 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. */
1652
bool
1653
qemuMigrationIsAllowed(virQEMUDriverPtr driver, virDomainObjPtr vm,
1654
                       virDomainDefPtr def, bool remote, bool abort_on_error)
1655
{
1656
    int nsnapshots;
1657
    int pauseReason;
1658
    bool forbid;
1659
    size_t i;
1660 1661 1662

    if (vm) {
        if (qemuProcessAutoDestroyActive(driver, vm)) {
1663 1664
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("domain is marked for auto destroy"));
1665 1666
            return false;
        }
1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679

        /* 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;
            }
1680 1681

            /* cancel migration if disk I/O error is emitted while migrating */
1682 1683
            if (abort_on_error &&
                virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
1684 1685 1686 1687 1688 1689
                pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("cannot migrate domain with I/O error"));
                return false;
            }

1690
        }
1691

E
Eric Blake 已提交
1692 1693
        if (virDomainHasDiskMirror(vm)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1694
                           _("domain has an active block job"));
E
Eric Blake 已提交
1695 1696
            return false;
        }
1697 1698 1699

        def = vm->def;
    }
1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714

    /* 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",
1715
                       _("domain has assigned non-USB host devices"));
1716 1717 1718
        return false;
    }

1719 1720 1721
    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
1722

1723 1724
            if (feature->policy != VIR_CPU_FEATURE_REQUIRE)
                continue;
J
Ján Tomko 已提交
1725

1726 1727 1728 1729 1730 1731 1732
            /* 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
1733 1734 1735
        }
    }

1736 1737 1738
    return true;
}

1739 1740 1741
static bool
qemuMigrationIsSafe(virDomainDefPtr def)
{
1742
    size_t i;
1743

1744
    for (i = 0; i < def->ndisks; i++) {
1745
        virDomainDiskDefPtr disk = def->disks[i];
1746
        const char *src = virDomainDiskGetSource(disk);
1747

1748 1749
        /* Our code elsewhere guarantees shared disks are either readonly (in
         * which case cache mode doesn't matter) or used with cache=none */
1750
        if (src &&
1751 1752
            !disk->src->shared &&
            !disk->src->readonly &&
1753
            disk->cachemode != VIR_DOMAIN_DISK_CACHE_DISABLE) {
1754
            int rc;
1755

E
Eric Blake 已提交
1756
            if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
1757
                if ((rc = virFileIsSharedFS(src)) < 0)
1758 1759
                    return false;
                else if (rc == 0)
1760
                    continue;
1761
                if ((rc = virStorageFileIsClusterFS(src)) < 0)
1762
                    return false;
1763 1764
                else if (rc == 1)
                    continue;
1765 1766
            } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                       disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
1767
                continue;
1768
            }
1769

1770 1771 1772
            virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                           _("Migration may lead to data corruption if disks"
                             " use cache != none"));
1773 1774 1775 1776 1777 1778 1779
            return false;
        }
    }

    return true;
}

1780 1781 1782 1783
/** qemuMigrationSetOffline
 * Pause domain for non-live migration.
 */
int
1784
qemuMigrationSetOffline(virQEMUDriverPtr driver,
1785 1786 1787
                        virDomainObjPtr vm)
{
    int ret;
1788
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
1789 1790
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
1791
    if (ret == 0) {
1792
        virObjectEventPtr event;
1793

1794
        event = virDomainEventLifecycleNewFromObj(vm,
1795 1796 1797 1798 1799 1800 1801 1802 1803 1804
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
        if (event)
            qemuDomainEventQueue(driver, event);
    }

    return ret;
}


1805 1806 1807
static int
qemuMigrationSetCompression(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
1808
                            qemuDomainAsyncJob job)
1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839
{
    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;
    } 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,
                QEMU_MONITOR_MIGRATION_CAPS_XBZRLE);

1840
 cleanup:
1841 1842 1843 1844
    qemuDomainObjExitMonitor(driver, vm);
    return ret;
}

1845 1846 1847
static int
qemuMigrationSetAutoConverge(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
1848
                             qemuDomainAsyncJob job)
1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
{
    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;
    } 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,
                QEMU_MONITOR_MIGRATION_CAPS_AUTO_CONVERGE);

1874
 cleanup:
1875 1876 1877 1878 1879
    qemuDomainObjExitMonitor(driver, vm);
    return ret;
}


1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919
static int
qemuMigrationSetPinAll(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
                       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;
    } 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,
                QEMU_MONITOR_MIGRATION_CAPS_RDMA_PIN_ALL);

 cleanup:
    qemuDomainObjExitMonitor(driver, vm);
    return ret;
}

1920 1921 1922 1923 1924 1925 1926
static int
qemuMigrationWaitForSpice(virQEMUDriverPtr driver,
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool wait_for_spice = false;
    bool spice_migrated = false;
1927
    size_t i = 0;
1928

1929 1930 1931 1932 1933 1934 1935 1936
    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;
            }
        }
    }
1937 1938 1939 1940 1941 1942 1943 1944

    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 };

1945 1946 1947 1948
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            return -1;

1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961
        if (qemuMonitorGetSpiceMigrationStatus(priv->mon,
                                               &spice_migrated) < 0) {
            qemuDomainObjExitMonitor(driver, vm);
            return -1;
        }
        qemuDomainObjExitMonitor(driver, vm);
        virObjectUnlock(vm);
        nanosleep(&ts, NULL);
        virObjectLock(vm);
    }

    return 0;
}
1962

1963
static int
1964
qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver,
1965
                             virDomainObjPtr vm,
1966
                             const char *job,
1967
                             qemuDomainAsyncJob asyncJob)
1968
{
1969
    qemuDomainObjPrivatePtr priv = vm->privateData;
1970
    qemuMonitorMigrationStatus status;
J
Jiri Denemark 已提交
1971 1972
    qemuDomainJobInfoPtr jobInfo;
    int ret;
1973 1974

    memset(&status, 0, sizeof(status));
1975

1976 1977
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
    if (ret < 0) {
1978 1979 1980
        /* Guest already exited or waiting for the job timed out; nothing
         * further to update. */
        return ret;
1981
    }
1982
    ret = qemuMonitorGetMigrationStatus(priv->mon, &status);
1983

1984
    qemuDomainObjExitMonitor(driver, vm);
1985

J
Jiri Denemark 已提交
1986 1987
    if (ret < 0 ||
        qemuDomainJobInfoUpdateTime(priv->job.current) < 0)
1988
        return -1;
1989

1990
    ret = -1;
J
Jiri Denemark 已提交
1991 1992 1993 1994 1995
    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
1996
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
1997 1998 1999 2000
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
        ret = 0;
        break;

J
Jiri Denemark 已提交
2001 2002 2003 2004
    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
        jobInfo->type = VIR_DOMAIN_JOB_NONE;
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("is not active"));
2005 2006 2007
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
J
Jiri Denemark 已提交
2008
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
2009 2010
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("unexpectedly failed"));
2011 2012 2013
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
J
Jiri Denemark 已提交
2014
        jobInfo->type = VIR_DOMAIN_JOB_CANCELLED;
2015 2016
        virReportError(VIR_ERR_OPERATION_ABORTED,
                       _("%s: %s"), job, _("canceled by client"));
2017 2018
        break;
    }
J
Jiri Denemark 已提交
2019
    jobInfo->status = status;
2020 2021 2022 2023 2024

    return ret;
}


2025 2026 2027
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
2028
static int
J
Jiri Denemark 已提交
2029 2030
qemuMigrationWaitForCompletion(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
2031
                               qemuDomainAsyncJob asyncJob,
J
Jiri Denemark 已提交
2032 2033
                               virConnectPtr dconn,
                               bool abort_on_error)
2034
{
2035
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
2036
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
2037
    const char *job;
2038
    int pauseReason;
2039

2040 2041
    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
2042 2043
        job = _("migration job");
        break;
2044
    case QEMU_ASYNC_JOB_SAVE:
2045 2046
        job = _("domain save job");
        break;
2047
    case QEMU_ASYNC_JOB_DUMP:
2048 2049 2050 2051 2052
        job = _("domain core dump job");
        break;
    default:
        job = _("job");
    }
2053

J
Jiri Denemark 已提交
2054
    jobInfo->type = VIR_DOMAIN_JOB_UNBOUNDED;
2055

J
Jiri Denemark 已提交
2056
    while (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
2057 2058 2059
        /* Poll every 50ms for progress & to allow cancellation */
        struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };

2060
        if (qemuMigrationUpdateJobStatus(driver, vm, job, asyncJob) == -1)
2061
            break;
2062

2063
        /* cancel migration if disk I/O error is emitted while migrating */
2064
        if (abort_on_error &&
2065
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
2066 2067 2068 2069 2070
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("%s: %s"), job, _("failed due to I/O error"));
            break;
        }
2071

2072
        if (dconn && virConnectIsAlive(dconn) <= 0) {
2073 2074
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
2075
            break;
2076 2077
        }

2078
        virObjectUnlock(vm);
2079 2080 2081

        nanosleep(&ts, NULL);

2082
        virObjectLock(vm);
2083 2084
    }

J
Jiri Denemark 已提交
2085
    if (jobInfo->type == VIR_DOMAIN_JOB_COMPLETED) {
2086
        qemuDomainJobInfoUpdateDowntime(jobInfo);
2087 2088 2089
        VIR_FREE(priv->job.completed);
        if (VIR_ALLOC(priv->job.completed) == 0)
            *priv->job.completed = *jobInfo;
2090
        return 0;
J
Jiri Denemark 已提交
2091
    } else if (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
2092 2093 2094
        /* 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
2095
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
2096 2097
        return -2;
    } else {
2098
        return -1;
2099
    }
2100 2101 2102
}


2103
static int
2104
qemuDomainMigrateGraphicsRelocate(virQEMUDriverPtr driver,
2105
                                  virDomainObjPtr vm,
2106 2107
                                  qemuMigrationCookiePtr cookie,
                                  const char *graphicsuri)
2108 2109
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2110 2111
    int ret = -1;
    const char *listenAddress = NULL;
2112
    virSocketAddr addr;
2113 2114 2115 2116 2117
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
2118

2119
    if (!cookie || (!cookie->graphics && !graphicsuri))
2120 2121
        return 0;

2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140
    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) {
2141
        size_t i;
2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168

        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;
            }
        }
    }
2169 2170 2171 2172

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
2173 2174 2175 2176
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
2177

2178 2179 2180 2181
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
2182
        qemuDomainObjExitMonitor(driver, vm);
2183
    }
2184

2185
 cleanup:
2186
    virURIFree(uri);
2187 2188 2189 2190
    return ret;
}


2191
static int
2192
qemuDomainMigrateOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
2193 2194 2195
                             virDomainObjPtr vm,
                             qemuMigrationCookiePtr cookie)
{
2196 2197
    virDomainNetDefPtr netptr;
    int ret = -1;
2198
    size_t i;
2199 2200 2201 2202 2203 2204 2205 2206

    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:
2207
           break;
2208
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
2209 2210 2211 2212 2213 2214 2215 2216
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
                virReportSystemError(VIR_ERR_INTERNAL_ERROR,
                                     _("Unable to run command to set OVS port data for "
                                     "interface %s"), netptr->ifname);
                goto cleanup;
            }
            break;
2217 2218 2219 2220 2221
        default:
            break;
        }
    }

2222
    ret = 0;
2223
 cleanup:
2224 2225 2226 2227
    return ret;
}


2228 2229 2230 2231 2232 2233
/* 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.
 */
2234 2235 2236 2237
static virDomainObjPtr
qemuMigrationCleanup(virDomainObjPtr vm,
                     virConnectPtr conn,
                     void *opaque)
2238
{
2239
    virQEMUDriverPtr driver = opaque;
2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254
    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);

2255
    switch ((qemuMigrationJobPhase) priv->job.phase) {
2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285
    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 */
        ;
    }

2286
 cleanup:
2287 2288 2289
    return vm;
}

2290

2291
/* The caller is supposed to lock the vm and start a migration job. */
2292 2293
static char
*qemuMigrationBeginPhase(virQEMUDriverPtr driver,
2294
                         virDomainObjPtr vm,
2295
                         const char *xmlin,
2296
                         const char *dname,
2297
                         char **cookieout,
2298 2299
                         int *cookieoutlen,
                         unsigned long flags)
2300 2301 2302
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
2303
    virDomainDefPtr def = NULL;
2304
    qemuDomainObjPrivatePtr priv = vm->privateData;
2305
    virCapsPtr caps = NULL;
2306
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
2307
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2308

2309
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
2310
              " cookieout=%p, cookieoutlen=%p, flags=%lx",
2311
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
2312
              cookieout, cookieoutlen, flags);
2313

2314 2315 2316
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2317 2318 2319 2320 2321 2322
    /* 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);
2323

2324
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
2325 2326
        goto cleanup;

2327 2328 2329
    if (!(flags & VIR_MIGRATE_UNSAFE) && !qemuMigrationIsSafe(vm->def))
        goto cleanup;

2330
    if (flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
2331
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DRIVE_MIRROR)) {
2332
        /* TODO support NBD for TUNNELLED migration */
2333
        if (flags & VIR_MIGRATE_TUNNELLED) {
2334 2335 2336 2337
            VIR_WARN("NBD in tunnelled migration is currently not supported");
        } else {
            cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
            priv->nbdPort = 0;
2338 2339 2340
        }
    }

2341
    if (!(mig = qemuMigrationEatCookie(driver, vm, NULL, 0, 0)))
2342 2343 2344 2345
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
                                cookieout, cookieoutlen,
2346
                                cookieFlags) < 0)
2347 2348
        goto cleanup;

L
liguang 已提交
2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370
    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;
        }
    }

2371
    if (xmlin) {
2372
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt,
M
Matthias Bolte 已提交
2373
                                            QEMU_EXPECTED_VIRT_TYPES,
2374 2375 2376
                                            VIR_DOMAIN_XML_INACTIVE)))
            goto cleanup;

2377
        if (!qemuDomainDefCheckABIStability(driver, vm->def, def))
2378 2379
            goto cleanup;

2380
        rv = qemuDomainDefFormatLive(driver, def, false, true);
2381
    } else {
2382
        rv = qemuDomainDefFormatLive(driver, vm->def, false, true);
2383
    }
2384

2385
 cleanup:
2386
    qemuMigrationCookieFree(mig);
2387
    virObjectUnref(caps);
2388
    virDomainDefFree(def);
2389 2390 2391
    return rv;
}

2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402
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;
2403
    qemuDomainAsyncJob asyncJob;
2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414

    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;
    }

2415 2416
    qemuMigrationStoreDomainState(vm);

2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439
    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.
         */
2440 2441
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                 qemuMigrationCleanup) < 0)
2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454
            goto endjob;
        if (qemuMigrationJobContinue(vm) == 0) {
            vm = NULL;
            virReportError(VIR_ERR_OPERATION_FAILED,
                           "%s", _("domain disappeared"));
            VIR_FREE(xml);
            if (cookieout)
                VIR_FREE(*cookieout);
        }
    } else {
        goto endjob;
    }

2455
 cleanup:
2456 2457 2458 2459
    if (vm)
        virObjectUnlock(vm);
    return xml;

2460
 endjob:
2461 2462 2463 2464
    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
        if (qemuMigrationJobFinish(driver, vm) == 0)
            vm = NULL;
    } else {
E
Eric Blake 已提交
2465
        if (!qemuDomainObjEndJob(driver, vm))
2466 2467 2468 2469 2470
            vm = NULL;
    }
    goto cleanup;
}

2471

2472 2473
/* Prepare is the first step, and it runs on the destination host.
 */
2474

2475
static void
2476
qemuMigrationPrepareCleanup(virQEMUDriverPtr driver,
2477 2478 2479 2480 2481 2482 2483 2484 2485 2486
                            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));

2487 2488 2489
    virPortAllocatorRelease(driver->migrationPorts, priv->migrationPort);
    priv->migrationPort = 0;

2490 2491 2492 2493 2494
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2495
static int
2496
qemuMigrationPrepareAny(virQEMUDriverPtr driver,
2497 2498 2499 2500 2501
                        virConnectPtr dconn,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
2502
                        virDomainDefPtr *def,
2503
                        const char *origname,
L
liguang 已提交
2504
                        virStreamPtr st,
2505
                        const char *protocol,
2506 2507
                        unsigned short port,
                        bool autoPort,
2508
                        const char *listenAddress,
L
liguang 已提交
2509
                        unsigned long flags)
2510 2511
{
    virDomainObjPtr vm = NULL;
2512
    virObjectEventPtr event = NULL;
2513
    int ret = -1;
2514
    int dataFD[2] = { -1, -1 };
2515
    qemuDomainObjPrivatePtr priv = NULL;
J
Jiri Denemark 已提交
2516
    unsigned long long now;
2517
    qemuMigrationCookiePtr mig = NULL;
2518
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2519
    char *xmlout = NULL;
L
liguang 已提交
2520
    unsigned int cookieFlags;
2521
    virCapsPtr caps = NULL;
J
Ján Tomko 已提交
2522
    char *migrateFrom = NULL;
2523
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2524
    bool taint_hook = false;
2525

2526
    if (virTimeMillisNow(&now) < 0)
2527 2528
        return -1;

2529 2530
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552
    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;
        }
    }

2553 2554 2555
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2556
    if (!qemuMigrationIsAllowed(driver, NULL, *def, true, abort_on_error))
2557 2558
        goto cleanup;

J
Jiri Denemark 已提交
2559 2560 2561 2562 2563
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2564
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2565 2566
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2567 2568
            goto cleanup;

2569
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2570 2571 2572 2573 2574 2575 2576
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2577
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2578 2579 2580 2581 2582 2583
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2584
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt,
J
Jiri Denemark 已提交
2585 2586 2587 2588 2589
                                                 QEMU_EXPECTED_VIRT_TYPES,
                                                 VIR_DOMAIN_XML_INACTIVE);
                if (!newdef)
                    goto cleanup;

2590
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2591 2592 2593 2594
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2595 2596
                virDomainDefFree(*def);
                *def = newdef;
2597 2598 2599 2600
                /* 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
2601 2602 2603 2604
            }
        }
    }

J
Ján Tomko 已提交
2605 2606 2607 2608
    if (tunnel) {
        /* QEMU will be started with -incoming stdio
         * (which qemu_command might convert to exec:cat or fd:n)
         */
2609
        if (VIR_STRDUP(migrateFrom, "stdio") < 0)
J
Ján Tomko 已提交
2610 2611
            goto cleanup;
    } else {
2612 2613 2614
        bool encloseAddress = false;
        bool hostIPv6Capable = false;
        bool qemuIPv6Capable = false;
J
Ján Tomko 已提交
2615 2616 2617 2618
        virQEMUCapsPtr qemuCaps = NULL;
        struct addrinfo *info = NULL;
        struct addrinfo hints = { .ai_flags = AI_ADDRCONFIG,
                                  .ai_socktype = SOCK_STREAM };
2619
        const char *incFormat;
J
Ján Tomko 已提交
2620

2621 2622 2623 2624
        if (getaddrinfo("::", NULL, &hints, &info) == 0) {
            freeaddrinfo(info);
            hostIPv6Capable = true;
        }
J
Ján Tomko 已提交
2625
        if (!(qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
2626
                                                    (*def)->emulator)))
J
Ján Tomko 已提交
2627 2628
            goto cleanup;

2629 2630 2631 2632
        qemuIPv6Capable = virQEMUCapsGet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
        virObjectUnref(qemuCaps);

        if (listenAddress) {
2633 2634 2635 2636 2637 2638 2639 2640 2641
            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"));
2642 2643
                    goto cleanup;
                }
2644 2645
                /* IPv6 address must be escaped in brackets on the cmd line */
                encloseAddress = true;
2646
            } else {
2647
                /* listenAddress is a hostname or IPv4 */
2648
            }
2649
        } else if (qemuIPv6Capable && hostIPv6Capable) {
2650 2651 2652
            /* Listen on :: instead of 0.0.0.0 if QEMU understands it
             * and there is at least one IPv6 address configured
             */
2653 2654 2655 2656
            listenAddress = "::";
            encloseAddress = true;
        } else {
            listenAddress = "0.0.0.0";
J
Ján Tomko 已提交
2657 2658
        }

2659 2660 2661 2662
        /* 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
2663
         */
2664 2665 2666 2667 2668 2669
        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
2670 2671 2672
            goto cleanup;
    }

2673
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2674
                                   driver->xmlopt,
2675 2676 2677
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2678
        goto cleanup;
2679

2680
    *def = NULL;
2681
    priv = vm->privateData;
2682 2683
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2684

2685 2686 2687 2688 2689
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2690
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
2691 2692
                                       QEMU_MIGRATION_COOKIE_LOCKSTATE |
                                       QEMU_MIGRATION_COOKIE_NBD)))
2693 2694
        goto cleanup;

M
Michael R. Hines 已提交
2695 2696 2697 2698 2699 2700 2701
    if (STREQ(protocol, "rdma") && !vm->def->mem.hard_limit) {
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2702
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2703
        goto cleanup;
2704
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2705 2706 2707 2708

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

L
liguang 已提交
2709 2710 2711
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2712 2713
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
2714 2715
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
2716 2717 2718 2719
        goto endjob;
    }

    /* Start the QEMU daemon, with the same command-line arguments plus
2720
     * -incoming $migrateFrom
2721
     */
2722 2723
    if (qemuProcessStart(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                         migrateFrom, dataFD[0], NULL, NULL,
2724 2725
                         VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
                         VIR_QEMU_PROCESS_START_PAUSED |
2726
                         VIR_QEMU_PROCESS_START_AUTODESTROY) < 0) {
2727
        virDomainAuditStart(vm, "migrated", false);
2728 2729 2730
        goto endjob;
    }

2731 2732 2733 2734
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
2735
            goto stop;
2736
        }
2737
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
2738 2739
    }

2740 2741 2742 2743 2744
    if (flags & VIR_MIGRATE_COMPRESSED &&
        qemuMigrationSetCompression(driver, vm,
                                    QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stop;

M
Michael R. Hines 已提交
2745 2746 2747 2748 2749
    if (STREQ(protocol, "rdma") &&
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
        goto stop;
    }

2750 2751 2752 2753
    if (flags & VIR_MIGRATE_RDMA_PIN_ALL &&
        qemuMigrationSetPinAll(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stop;

2754 2755 2756 2757 2758 2759 2760 2761
    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");
    }
2762

2763
 done:
L
liguang 已提交
2764 2765 2766 2767 2768
    if (flags & VIR_MIGRATE_OFFLINE)
        cookieFlags = 0;
    else
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS;

2769 2770 2771
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
2772
        if (qemuMigrationStartNBDServer(driver, vm, listenAddress) < 0) {
2773 2774
            /* error already reported */
            goto endjob;
2775
        }
2776
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
2777 2778 2779 2780
    }

    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
2781 2782 2783 2784 2785 2786 2787
        /* 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");
    }

2788 2789 2790
    if (qemuDomainCleanupAdd(vm, qemuMigrationPrepareCleanup) < 0)
        goto endjob;

L
liguang 已提交
2791 2792
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
2793
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
2794 2795 2796
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
2797

2798 2799 2800 2801
    /* 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.
     */
2802
    if (!qemuMigrationJobContinue(vm)) {
2803
        vm = NULL;
2804 2805
        virReportError(VIR_ERR_OPERATION_FAILED,
                       "%s", _("domain disappeared"));
2806
        goto cleanup;
2807
    }
2808

2809 2810
    if (autoPort)
        priv->migrationPort = port;
2811
    ret = 0;
2812

2813
 cleanup:
J
Ján Tomko 已提交
2814
    VIR_FREE(migrateFrom);
J
Jiri Denemark 已提交
2815
    VIR_FREE(xmlout);
2816 2817
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
2818
    if (vm) {
2819
        if (ret < 0) {
2820
            virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
2821 2822
            priv->nbdPort = 0;
        }
2823 2824 2825 2826
        if (ret >= 0 || vm->persistent)
            virObjectUnlock(vm);
        else
            qemuDomainRemoveInactive(driver, vm);
2827
    }
2828 2829
    if (event)
        qemuDomainEventQueue(driver, event);
2830
    qemuMigrationCookieFree(mig);
2831
    virObjectUnref(caps);
2832
    virNWFilterUnlockFilterUpdates();
2833
    return ret;
2834

2835
 stop:
2836 2837 2838
    virDomainAuditStart(vm, "migrated", false);
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, 0);

2839
 endjob:
2840
    if (!qemuMigrationJobFinish(driver, vm)) {
2841 2842 2843
        vm = NULL;
    }
    goto cleanup;
2844 2845 2846
}


2847 2848 2849 2850 2851
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
2852
qemuMigrationPrepareTunnel(virQEMUDriverPtr driver,
2853 2854 2855 2856 2857 2858
                           virConnectPtr dconn,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virStreamPtr st,
2859
                           virDomainDefPtr *def,
2860
                           const char *origname,
L
liguang 已提交
2861
                           unsigned long flags)
2862 2863 2864 2865
{
    int ret;

    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2866
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
2867
              "origname=%s, flags=%lx",
2868
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2869
              cookieout, cookieoutlen, st, *def, origname, flags);
2870

2871 2872 2873 2874 2875 2876
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

2877
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
2878
                                  cookieout, cookieoutlen, def, origname,
2879
                                  st, NULL, 0, false, NULL, flags);
2880 2881 2882 2883
    return ret;
}


2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906
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;
}


2907
int
2908
qemuMigrationPrepareDirect(virQEMUDriverPtr driver,
2909
                           virConnectPtr dconn,
2910 2911 2912 2913
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
2914 2915
                           const char *uri_in,
                           char **uri_out,
2916
                           virDomainDefPtr *def,
2917
                           const char *origname,
2918
                           const char *listenAddress,
L
liguang 已提交
2919
                           unsigned long flags)
2920
{
2921 2922
    unsigned short port = 0;
    bool autoPort = true;
2923 2924
    char *hostname = NULL;
    int ret = -1;
2925
    virURIPtr uri = NULL;
2926 2927
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
2928

2929 2930
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
2931
              "def=%p, origname=%s, listenAddress=%s, flags=%lx",
2932 2933
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
2934
              *def, origname, NULLSTR(listenAddress), flags);
2935

2936 2937
    *uri_out = NULL;

2938 2939 2940
    /* 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
2941 2942 2943
     * 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".
2944 2945 2946 2947 2948 2949
     *
     * 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) {
2950 2951 2952
        bool encloseAddress = false;
        const char *incFormat;

2953
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2954
            goto cleanup;
2955

2956
        if (migrateHost != NULL) {
2957 2958
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
2959

2960
            if (VIR_STRDUP(hostname, migrateHost) < 0)
2961 2962 2963 2964 2965
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
2966 2967

        if (STRPREFIX(hostname, "localhost")) {
2968 2969 2970
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
2971 2972 2973 2974 2975
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
2976
         * compatibility with old targets. We at least make the
2977 2978
         * new targets accept both syntaxes though.
         */
2979 2980 2981 2982 2983 2984
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
2985 2986
            goto cleanup;
    } else {
2987
        bool well_formed_uri;
J
Ján Tomko 已提交
2988

2989 2990
        if (!(uri = qemuMigrationParseURI(uri_in, &well_formed_uri)))
            goto cleanup;
J
Ján Tomko 已提交
2991

M
Michael R. Hines 已提交
2992 2993
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
2994 2995 2996
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
2997 2998 2999 3000 3001 3002 3003 3004 3005 3006
            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) {
3007
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3008
                goto cleanup;
3009

3010
            /* Send well-formed URI only if uri_in was well-formed */
3011 3012 3013 3014 3015 3016 3017 3018
            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;
            }
3019
        } else {
3020 3021
            port = uri->port;
            autoPort = false;
3022 3023 3024 3025 3026 3027
        }
    }

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

3028
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3029
                                  cookieout, cookieoutlen, def, origname,
3030 3031
                                  NULL, uri ? uri->scheme : "tcp",
                                  port, autoPort, listenAddress, flags);
3032
 cleanup:
3033
    virURIFree(uri);
3034
    VIR_FREE(hostname);
3035
    virObjectUnref(cfg);
3036
    if (ret != 0) {
3037
        VIR_FREE(*uri_out);
3038 3039 3040
        if (autoPort)
            virPortAllocatorRelease(driver->migrationPorts, port);
    }
3041 3042 3043 3044
    return ret;
}


3045 3046 3047
virDomainDefPtr
qemuMigrationPrepareDef(virQEMUDriverPtr driver,
                        const char *dom_xml,
3048 3049
                        const char *dname,
                        char **origname)
3050 3051 3052
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
3053
    char *name = NULL;
3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069

    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,
                                        VIR_DOMAIN_XML_INACTIVE)))
        goto cleanup;

    if (dname) {
3070
        name = def->name;
3071 3072 3073 3074 3075 3076
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

3077
 cleanup:
3078
    virObjectUnref(caps);
3079 3080 3081 3082
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
3083 3084 3085 3086
    return def;
}


3087 3088 3089 3090 3091 3092 3093 3094 3095 3096
static int
qemuMigrationConfirmPhase(virQEMUDriverPtr driver,
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *cookiein,
                          int cookieinlen,
                          unsigned int flags,
                          int retcode)
{
    qemuMigrationCookiePtr mig;
3097
    virObjectEventPtr event = NULL;
3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112
    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);

3113 3114
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                       QEMU_MIGRATION_COOKIE_STATS)))
3115 3116
        goto cleanup;

3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133
    /* 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;
        }
    }

3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148
    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");

3149
        event = virDomainEventLifecycleNewFromObj(vm,
3150 3151 3152 3153 3154 3155 3156
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
    } else {

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

3157 3158 3159 3160
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
3161 3162 3163 3164 3165 3166 3167 3168
        }

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

3169
 done:
3170 3171 3172
    qemuMigrationCookieFree(mig);
    rv = 0;

3173
 cleanup:
3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188
    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;
3189
    qemuMigrationJobPhase phase;
3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203
    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);
3204 3205
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220

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

    if (qemuMigrationJobFinish(driver, vm) == 0) {
        vm = NULL;
    } else if (!virDomainObjIsActive(vm) &&
               (!vm->persistent || (flags & VIR_MIGRATE_UNDEFINE_SOURCE))) {
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
        qemuDomainRemoveInactive(driver, vm);
        vm = NULL;
    }

3221
 cleanup:
3222 3223 3224 3225 3226 3227 3228
    if (vm)
        virObjectUnlock(vm);
    virObjectUnref(cfg);
    return ret;
}


3229 3230
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3231
    MIGRATION_DEST_CONNECT_HOST,
3232
    MIGRATION_DEST_UNIX,
3233
    MIGRATION_DEST_FD,
3234
};
3235

3236 3237 3238 3239
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3240

3241 3242 3243 3244 3245 3246
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3247
            const char *protocol;
3248 3249 3250 3251 3252
            const char *name;
            int port;
        } host;

        struct {
3253
            char *file;
3254 3255
            int sock;
        } unix_socket;
3256 3257 3258 3259 3260

        struct {
            int qemu;
            int local;
        } fd;
3261 3262 3263 3264 3265 3266 3267
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3268 3269 3270

#define TUNNEL_SEND_BUF_SIZE 65536

3271 3272 3273 3274 3275 3276 3277
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3278 3279
    int wakeupRecvFD;
    int wakeupSendFD;
3280 3281 3282
};

static void qemuMigrationIOFunc(void *arg)
3283
{
3284
    qemuMigrationIOThreadPtr data = arg;
3285 3286 3287 3288 3289 3290 3291
    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);
3292

3293
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3294
        goto abrt;
3295

3296 3297 3298
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3299
    for (;;) {
3300 3301 3302 3303 3304 3305 3306 3307 3308 3309
        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;
3310
            virReportSystemError(errno, "%s",
3311 3312
                                 _("poll failed in migration tunnel"));
            goto abrt;
3313
        }
3314 3315 3316 3317 3318 3319 3320

        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");
3321
            break;
3322
        }
3323

3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339
        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;
            }
3340 3341
        }

3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358
        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;
            }
        }
    }
3359

3360 3361
    if (virStreamFinish(data->st) < 0)
        goto error;
3362

3363 3364
    VIR_FREE(buffer);

3365 3366
    return;

3367
 abrt:
3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3379
 error:
3380 3381
    virCopyLastError(&data->err);
    virResetLastError();
3382
    VIR_FREE(buffer);
3383 3384 3385 3386 3387 3388 3389
}


static qemuMigrationIOThreadPtr
qemuMigrationStartTunnel(virStreamPtr st,
                         int sock)
{
3390 3391
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3392

3393 3394 3395 3396
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3397 3398
    }

3399
    if (VIR_ALLOC(io) < 0)
3400
        goto error;
3401

3402 3403
    io->st = st;
    io->sock = sock;
3404 3405
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3406 3407 3408 3409 3410 3411

    if (virThreadCreate(&io->thread, true,
                        qemuMigrationIOFunc,
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3412
        goto error;
3413 3414 3415
    }

    return io;
3416

3417
 error:
3418 3419 3420 3421
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3422 3423 3424
}

static int
3425
qemuMigrationStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3426 3427
{
    int rv = -1;
3428 3429 3430 3431 3432 3433 3434 3435 3436
    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;
    }

3437 3438 3439 3440
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3441 3442 3443 3444
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3445 3446 3447 3448 3449 3450
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3451
 cleanup:
3452 3453
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3454 3455
    VIR_FREE(io);
    return rv;
3456 3457
}

3458
static int
3459
qemuMigrationConnect(virQEMUDriverPtr driver,
3460 3461 3462 3463 3464 3465 3466 3467 3468
                     virDomainObjPtr vm,
                     qemuMigrationSpecPtr spec)
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3469
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3470 3471 3472 3473 3474 3475 3476 3477 3478
        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);
3479
        virObjectUnref(sock);
3480 3481 3482 3483 3484
    }
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0 ||
        spec->dest.fd.qemu == -1)
        goto cleanup;

3485 3486 3487 3488 3489 3490 3491
    /* 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;
    }

3492 3493
    ret = 0;

3494
 cleanup:
3495 3496 3497 3498 3499 3500
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3501
static int
3502
qemuMigrationRun(virQEMUDriverPtr driver,
3503 3504 3505 3506 3507 3508 3509
                 virDomainObjPtr vm,
                 const char *cookiein,
                 int cookieinlen,
                 char **cookieout,
                 int *cookieoutlen,
                 unsigned long flags,
                 unsigned long resource,
3510
                 qemuMigrationSpecPtr spec,
3511 3512
                 virConnectPtr dconn,
                 const char *graphicsuri)
3513
{
3514
    int ret = -1;
3515 3516
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3517
    qemuMigrationCookiePtr mig = NULL;
3518
    qemuMigrationIOThreadPtr iothread = NULL;
3519
    int fd = -1;
3520
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3521
    virErrorPtr orig_err = NULL;
3522
    unsigned int cookieFlags = 0;
3523
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3524
    int rc;
3525 3526 3527

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
3528
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s",
3529 3530
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3531 3532
              spec, spec->destType, spec->fwdType, dconn,
              NULLSTR(graphicsuri));
3533

3534 3535 3536 3537 3538 3539 3540 3541 3542 3543
    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;
    }

3544 3545
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3546 3547 3548 3549
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3550 3551 3552
        return -1;
    }

3553 3554 3555
    mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                 cookieFlags | QEMU_MIGRATION_COOKIE_GRAPHICS);
    if (!mig)
3556 3557
        goto cleanup;

3558
    if (qemuDomainMigrateGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3559 3560
        VIR_WARN("unable to provide data for graphics client relocation");

3561 3562 3563 3564 3565 3566 3567
    /* 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;
    }

3568
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3569 3570 3571 3572 3573 3574
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

3575 3576 3577 3578 3579
    if (flags & VIR_MIGRATE_COMPRESSED &&
        qemuMigrationSetCompression(driver, vm,
                                    QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3580 3581 3582 3583 3584
    if (flags & VIR_MIGRATE_AUTO_CONVERGE &&
        qemuMigrationSetAutoConverge(driver, vm,
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3585 3586 3587 3588 3589
    if (flags & VIR_MIGRATE_RDMA_PIN_ALL &&
        qemuMigrationSetPinAll(driver, vm,
                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3590 3591
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3592 3593
        goto cleanup;

3594 3595 3596 3597
    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 */
3598
        qemuDomainObjExitMonitor(driver, vm);
3599 3600 3601 3602 3603 3604
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
        goto cleanup;
    }

3605
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0) {
3606
        qemuDomainObjExitMonitor(driver, vm);
3607 3608
        goto cleanup;
    }
3609

3610 3611
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3612
        qemuMigrationConnect(driver, vm, spec) < 0) {
3613
        qemuDomainObjExitMonitor(driver, vm);
3614
        goto cleanup;
3615
    }
3616

3617 3618
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3619 3620 3621 3622 3623
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
            qemuDomainObjExitMonitor(driver, vm);
            goto cleanup;
        }
3624
        ret = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
3625
                                       spec->dest.host.protocol,
3626 3627 3628 3629
                                       spec->dest.host.name,
                                       spec->dest.host.port);
        break;

3630 3631 3632 3633
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3634
    case MIGRATION_DEST_UNIX:
3635
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX)) {
3636 3637 3638 3639 3640 3641 3642 3643 3644
            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;
3645 3646

    case MIGRATION_DEST_FD:
3647
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3648
            fd = spec->dest.fd.local;
3649 3650
            spec->dest.fd.local = -1;
        }
3651 3652 3653 3654
        ret = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                     spec->dest.fd.qemu);
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3655
    }
3656
    qemuDomainObjExitMonitor(driver, vm);
3657
    if (ret < 0)
3658 3659
        goto cleanup;
    ret = -1;
3660 3661

    if (!virDomainObjIsActive(vm)) {
3662 3663
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
3664 3665 3666 3667 3668 3669
        goto cleanup;
    }

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

3670 3671 3672 3673 3674 3675 3676 3677
    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;
3678

3679 3680 3681 3682 3683 3684 3685
        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;
        }
3686 3687
    }

3688 3689
    if (spec->fwdType != MIGRATION_FWD_DIRECT &&
        !(iothread = qemuMigrationStartTunnel(spec->fwd.stream, fd)))
3690 3691
        goto cancel;

3692 3693 3694 3695 3696 3697
    rc = qemuMigrationWaitForCompletion(driver, vm,
                                        QEMU_ASYNC_JOB_MIGRATION_OUT,
                                        dconn, abort_on_error);
    if (rc == -2)
        goto cancel;
    else if (rc == -1)
3698
        goto cleanup;
3699

3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712
    /* 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) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

3713
    ret = 0;
3714

3715
 cleanup:
3716 3717 3718
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

3719
    /* cancel any outstanding NBD jobs */
3720 3721
    if (mig)
        qemuMigrationCancelDriveMirror(mig, driver, vm);
3722

3723
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3724
        if (iothread && qemuMigrationStopTunnel(iothread, ret < 0) < 0)
3725
            ret = -1;
3726
        VIR_FORCE_CLOSE(fd);
3727
    }
3728

3729
    if (priv->job.completed) {
3730
        qemuDomainJobInfoUpdateTime(priv->job.completed);
3731 3732
        qemuDomainJobInfoUpdateDowntime(priv->job.completed);
    }
3733

3734 3735
    cookieFlags |= QEMU_MIGRATION_COOKIE_NETWORK |
                   QEMU_MIGRATION_COOKIE_STATS;
3736 3737
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookieFlags |= QEMU_MIGRATION_COOKIE_PERSISTENT;
3738
    if (ret == 0 &&
3739 3740
        qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
3741
        VIR_WARN("Unable to encode migration cookie");
3742
    }
3743

3744 3745
    qemuMigrationCookieFree(mig);

3746 3747 3748 3749 3750
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

3751 3752
    return ret;

3753
 cancel:
3754 3755
    orig_err = virSaveLastError();

3756
    if (virDomainObjIsActive(vm)) {
3757 3758
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
3759
            qemuMonitorMigrateCancel(priv->mon);
3760
            qemuDomainObjExitMonitor(driver, vm);
3761
        }
3762
    }
3763 3764 3765
    goto cleanup;
}

3766
/* Perform migration using QEMU's native migrate support,
3767 3768
 * not encrypted obviously
 */
3769
static int doNativeMigrate(virQEMUDriverPtr driver,
3770 3771 3772 3773 3774 3775 3776
                           virDomainObjPtr vm,
                           const char *uri,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
3777
                           unsigned long resource,
3778 3779
                           virConnectPtr dconn,
                           const char *graphicsuri)
3780
{
3781
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
3782
    virURIPtr uribits = NULL;
3783
    int ret = -1;
3784 3785 3786
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
3787 3788
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "graphicsuri=%s",
3789
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
3790 3791
              cookieout, cookieoutlen, flags, resource,
              NULLSTR(graphicsuri));
3792

3793
    if (!(uribits = qemuMigrationParseURI(uri, NULL)))
3794 3795
        return -1;

M
Michael R. Hines 已提交
3796 3797 3798 3799 3800
    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"));
3801
            goto cleanup;
M
Michael R. Hines 已提交
3802 3803 3804 3805 3806
        }
        if (!vm->def->mem.hard_limit) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
3807
            goto cleanup;
M
Michael R. Hines 已提交
3808 3809 3810 3811 3812
        }
    }

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
        STRNEQ(uribits->scheme, "rdma"))
3813 3814
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
3815
        spec.destType = MIGRATION_DEST_HOST;
3816
    spec.dest.host.protocol = uribits->scheme;
3817 3818 3819
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
3820

3821
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
3822 3823
                           cookieoutlen, flags, resource, &spec, dconn,
                           graphicsuri);
3824 3825 3826 3827

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

3828
 cleanup:
3829
    virURIFree(uribits);
3830 3831 3832 3833 3834

    return ret;
}


3835
static int doTunnelMigrate(virQEMUDriverPtr driver,
3836 3837 3838 3839 3840 3841 3842
                           virDomainObjPtr vm,
                           virStreamPtr st,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
3843
                           unsigned long resource,
3844 3845
                           virConnectPtr dconn,
                           const char *graphicsuri)
3846 3847
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
3848
    virNetSocketPtr sock = NULL;
3849 3850
    int ret = -1;
    qemuMigrationSpec spec;
3851
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3852 3853

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
3854 3855
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "graphicsuri=%s",
3856
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
3857 3858
              cookieout, cookieoutlen, flags, resource,
              NULLSTR(graphicsuri));
3859

3860 3861 3862
    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)) {
3863 3864
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Source qemu is too old to support tunnelled migration"));
3865
        virObjectUnref(cfg);
3866
        return -1;
3867 3868 3869 3870 3871
    }

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

3872
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD)) {
3873 3874 3875 3876 3877 3878
        int fds[2];

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

3879
        if (pipe2(fds, O_CLOEXEC) == 0) {
3880 3881 3882 3883
            spec.dest.fd.qemu = fds[1];
            spec.dest.fd.local = fds[0];
        }
        if (spec.dest.fd.qemu == -1 ||
3884
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
3885
                                              spec.dest.fd.qemu) < 0) {
3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896
            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",
3897
                        cfg->libDir, vm->def->name) < 0)
3898 3899 3900
            goto cleanup;

        if (virNetSocketNewListenUNIX(spec.dest.unix_socket.file, 0700,
3901
                                      cfg->user, cfg->group,
3902 3903 3904 3905 3906 3907 3908
                                      &sock) < 0 ||
            virNetSocketListen(sock, 1) < 0)
            goto cleanup;

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

3909
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
3910 3911
                           cookieoutlen, flags, resource, &spec, dconn,
                           graphicsuri);
3912

3913
 cleanup:
3914 3915 3916 3917
    if (spec.destType == MIGRATION_DEST_FD) {
        VIR_FORCE_CLOSE(spec.dest.fd.qemu);
        VIR_FORCE_CLOSE(spec.dest.fd.local);
    } else {
3918
        virObjectUnref(sock);
3919 3920
        VIR_FREE(spec.dest.unix_socket.file);
    }
3921

3922
    virObjectUnref(cfg);
3923 3924 3925 3926
    return ret;
}


3927 3928 3929 3930
/* 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 */
3931
static int doPeer2PeerMigrate2(virQEMUDriverPtr driver,
3932
                               virConnectPtr sconn ATTRIBUTE_UNUSED,
3933 3934
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
3935
                               const char *dconnuri,
3936 3937 3938
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
3939 3940 3941
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
3942
    char *cookie = NULL;
3943 3944 3945
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
3946
    bool cancelled;
3947
    virStreamPtr st = NULL;
3948
    unsigned long destflags;
3949

3950
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
3951
              "flags=%lx, dname=%s, resource=%lu",
3952 3953
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
3954

3955 3956 3957 3958 3959
    /* 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,
3960 3961
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
3962 3963 3964 3965 3966
        return -1;

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

3967 3968
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
3969

3970 3971 3972 3973 3974 3975 3976 3977 3978 3979
    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;

3980
        qemuDomainObjEnterRemote(vm);
3981
        ret = dconn->driver->domainMigratePrepareTunnel
3982
            (dconn, st, destflags, dname, resource, dom_xml);
3983
        qemuDomainObjExitRemote(vm);
3984
    } else {
3985
        qemuDomainObjEnterRemote(vm);
3986 3987
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
3988
             destflags, dname, resource, dom_xml);
3989
        qemuDomainObjExitRemote(vm);
3990 3991 3992
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
3993 3994 3995
        goto cleanup;

    /* the domain may have shutdown or crashed while we had the locks dropped
3996
     * in qemuDomainObjEnterRemote, so check again
3997 3998
     */
    if (!virDomainObjIsActive(vm)) {
3999 4000
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4001 4002 4003
        goto cleanup;
    }

4004 4005
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
4006 4007
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
4008
        cancelled = true;
4009
        orig_err = virSaveLastError();
4010
        goto finish;
4011 4012
    }

4013 4014 4015 4016
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
4017
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4018
    if (flags & VIR_MIGRATE_TUNNELLED)
4019 4020
        ret = doTunnelMigrate(driver, vm, st,
                              NULL, 0, NULL, NULL,
4021
                              flags, resource, dconn, NULL);
4022 4023 4024 4025
    else
        ret = doNativeMigrate(driver, vm, uri_out,
                              cookie, cookielen,
                              NULL, NULL, /* No out cookie with v2 migration */
4026
                              flags, resource, dconn, NULL);
4027 4028 4029 4030

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4031

4032 4033 4034
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4035
    cancelled = ret < 0;
4036

4037
 finish:
4038 4039 4040 4041
    /* 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.
     */
4042
    dname = dname ? dname : vm->def->name;
4043
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4044
    qemuDomainObjEnterRemote(vm);
4045
    ddomain = dconn->driver->domainMigrateFinish2
4046
        (dconn, dname, cookie, cookielen,
4047
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4048
    qemuDomainObjExitRemote(vm);
4049 4050
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4051

4052
 cleanup:
4053
    if (ddomain) {
4054
        virObjectUnref(ddomain);
4055 4056 4057 4058
        ret = 0;
    } else {
        ret = -1;
    }
4059

4060
    virObjectUnref(st);
4061 4062 4063 4064 4065 4066

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4067
    VIR_FREE(cookie);
4068 4069

    return ret;
4070 4071 4072
}


4073 4074 4075 4076
/* 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 */
4077 4078 4079 4080 4081 4082 4083 4084 4085
static int
doPeer2PeerMigrate3(virQEMUDriverPtr driver,
                    virConnectPtr sconn,
                    virConnectPtr dconn,
                    const char *dconnuri,
                    virDomainObjPtr vm,
                    const char *xmlin,
                    const char *dname,
                    const char *uri,
4086
                    const char *graphicsuri,
4087
                    const char *listenAddress,
4088 4089 4090
                    unsigned long long bandwidth,
                    bool useParams,
                    unsigned long flags)
4091 4092 4093 4094 4095 4096 4097 4098 4099 4100
{
    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;
4101
    bool cancelled = true;
4102
    virStreamPtr st = NULL;
4103
    unsigned long destflags;
4104 4105 4106 4107 4108
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4109 4110
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
              "bandwidth=%llu, useParams=%d, flags=%lx",
4111
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4112 4113
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
              NULLSTR(listenAddress), bandwidth, useParams, flags);
4114

4115 4116 4117 4118 4119
    /* 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.  */

4120 4121
    dom_xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
                                      &cookieout, &cookieoutlen, flags);
4122 4123 4124
    if (!dom_xml)
        goto cleanup;

4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144
    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;
4145 4146 4147 4148 4149 4150

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4151 4152 4153 4154 4155
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4156 4157
    }

4158 4159 4160
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4161 4162
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4163

4164 4165 4166 4167 4168 4169 4170 4171 4172
    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;

4173
        qemuDomainObjEnterRemote(vm);
4174 4175 4176 4177 4178 4179 4180 4181 4182
        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);
        }
4183
        qemuDomainObjExitRemote(vm);
4184
    } else {
4185
        qemuDomainObjEnterRemote(vm);
4186 4187 4188 4189 4190 4191 4192 4193 4194
        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);
        }
4195
        qemuDomainObjExitRemote(vm);
4196 4197 4198 4199 4200
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

L
liguang 已提交
4201 4202 4203 4204
    if (flags & VIR_MIGRATE_OFFLINE) {
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4205
        cancelled = false;
L
liguang 已提交
4206 4207 4208
        goto finish;
    }

4209 4210 4211 4212
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4213 4214
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4215
            goto finish;
4216
        }
4217
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4218 4219
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4220
        orig_err = virSaveLastError();
4221 4222 4223 4224 4225 4226 4227 4228
        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.
     */
4229
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4230
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4231 4232 4233 4234 4235
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4236
    if (flags & VIR_MIGRATE_TUNNELLED) {
4237 4238 4239
        ret = doTunnelMigrate(driver, vm, st,
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4240
                              flags, bandwidth, dconn, graphicsuri);
4241 4242
    } else {
        ret = doNativeMigrate(driver, vm, uri,
4243 4244
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4245
                              flags, bandwidth, dconn, graphicsuri);
4246
    }
4247 4248

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4249
    if (ret < 0) {
4250
        orig_err = virSaveLastError();
4251 4252 4253 4254
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4255 4256 4257 4258

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4259
    cancelled = ret < 0;
4260

4261
 finish:
4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273
    /*
     * 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;
4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296

    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);
    }
4297 4298
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4299

4300 4301 4302 4303 4304 4305 4306
    /* 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.
4307
     */
4308
    cancelled = ddomain == NULL;
4309

4310 4311 4312 4313 4314 4315
    /* 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();

4316 4317 4318 4319
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4320
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4321 4322 4323 4324 4325
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4326 4327 4328
    ret = qemuMigrationConfirmPhase(driver, sconn, vm,
                                    cookiein, cookieinlen,
                                    flags, cancelled);
4329 4330 4331 4332
    /* 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.
     */
4333 4334 4335
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4336 4337 4338

 cleanup:
    if (ddomain) {
4339
        virObjectUnref(ddomain);
4340 4341 4342 4343 4344
        ret = 0;
    } else {
        ret = -1;
    }

4345
    virObjectUnref(st);
4346 4347 4348 4349 4350 4351 4352 4353

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4354
    virTypedParamsFree(params, nparams);
4355 4356 4357 4358
    return ret;
}


4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4371
static int doPeer2PeerMigrate(virQEMUDriverPtr driver,
4372
                              virConnectPtr sconn,
4373
                              virDomainObjPtr vm,
4374
                              const char *xmlin,
4375
                              const char *dconnuri,
4376
                              const char *uri,
4377
                              const char *graphicsuri,
4378
                              const char *listenAddress,
4379 4380
                              unsigned long flags,
                              const char *dname,
4381 4382
                              unsigned long resource,
                              bool *v3proto)
4383 4384 4385 4386
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4387
    virErrorPtr orig_err = NULL;
4388
    bool offline = false;
4389
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4390
    bool useParams;
4391

4392
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4393 4394
              "uri=%s, graphicsuri=%s, listenAddress=%s, flags=%lx, "
              "dname=%s, resource=%lu",
4395
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4396 4397
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
              flags, NULLSTR(dname), resource);
4398 4399 4400 4401 4402

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

4403
    qemuDomainObjEnterRemote(vm);
4404
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4405
    qemuDomainObjExitRemote(vm);
4406
    if (dconn == NULL) {
4407
        virReportError(VIR_ERR_OPERATION_FAILED,
4408 4409
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4410
        virObjectUnref(cfg);
4411 4412 4413
        return -1;
    }

4414 4415
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4416 4417
        goto cleanup;

4418
    qemuDomainObjEnterRemote(vm);
4419 4420
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4421
        /* v3proto reflects whether the caller used Perform3, but with
4422
         * p2p migrate, regardless of whether Perform2 or Perform3
4423 4424 4425 4426
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4427 4428
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
L
liguang 已提交
4429 4430 4431
    if (flags & VIR_MIGRATE_OFFLINE)
        offline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                           VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4432
    qemuDomainObjExitRemote(vm);
4433

4434
    if (!p2p) {
4435 4436
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4437 4438 4439
        goto cleanup;
    }

4440 4441
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4442
    if (!useParams && graphicsuri) {
4443 4444 4445 4446 4447 4448
        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
4449 4450 4451 4452 4453 4454 4455
    if (flags & VIR_MIGRATE_OFFLINE && !offline) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4456
    /* domain may have been stopped while we were talking to remote daemon */
L
liguang 已提交
4457
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4458 4459
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4460 4461 4462
        goto cleanup;
    }

4463 4464 4465 4466 4467 4468 4469
    /* 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;

4470 4471
    if (*v3proto) {
        ret = doPeer2PeerMigrate3(driver, sconn, dconn, dconnuri, vm, xmlin,
4472 4473
                                  dname, uri, graphicsuri, listenAddress,
                                  resource, useParams, flags);
4474
    } else {
4475
        ret = doPeer2PeerMigrate2(driver, sconn, dconn, vm,
4476
                                  dconnuri, flags, dname, resource);
4477
    }
4478

4479
 cleanup:
4480
    orig_err = virSaveLastError();
4481
    qemuDomainObjEnterRemote(vm);
4482
    virObjectUnref(dconn);
4483
    qemuDomainObjExitRemote(vm);
4484 4485 4486 4487
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4488
    virObjectUnref(cfg);
4489 4490 4491 4492
    return ret;
}


4493 4494 4495 4496 4497 4498
/*
 * 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
4499
qemuMigrationPerformJob(virQEMUDriverPtr driver,
4500 4501 4502 4503 4504
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dconnuri,
                        const char *uri,
4505
                        const char *graphicsuri,
4506
                        const char *listenAddress,
4507
                        const char *cookiein,
4508 4509 4510 4511 4512 4513 4514
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4515
{
4516
    virObjectEventPtr event = NULL;
4517
    int ret = -1;
4518
    virErrorPtr orig_err = NULL;
4519
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4520
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
4521

4522
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4523 4524
        goto cleanup;

L
liguang 已提交
4525
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4526 4527
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
4528 4529 4530
        goto endjob;
    }

4531
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
4532
        goto endjob;
4533

4534
    if (!(flags & VIR_MIGRATE_UNSAFE) && !qemuMigrationIsSafe(vm->def))
4535
        goto endjob;
4536

4537
    qemuMigrationStoreDomainState(vm);
4538 4539

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4540
        ret = doPeer2PeerMigrate(driver, conn, vm, xmlin,
4541 4542
                                 dconnuri, uri, graphicsuri, listenAddress,
                                 flags, dname, resource, &v3proto);
4543
    } else {
4544 4545 4546
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
        ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                              cookieout, cookieoutlen,
4547
                              flags, resource, NULL, NULL);
4548
    }
4549 4550
    if (ret < 0)
        goto endjob;
4551

4552 4553 4554 4555
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4556
    if (!v3proto) {
4557 4558
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4559
        virDomainAuditStop(vm, "migrated");
4560
        event = virDomainEventLifecycleNewFromObj(vm,
4561 4562
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4563 4564
    }

4565
 endjob:
4566 4567 4568
    if (ret < 0)
        orig_err = virSaveLastError();

4569
    if (qemuMigrationRestoreDomainState(conn, vm)) {
4570
        event = virDomainEventLifecycleNewFromObj(vm,
4571 4572 4573
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
4574

4575
    if (!qemuMigrationJobFinish(driver, vm)) {
4576 4577 4578 4579 4580
        vm = NULL;
    } else if (!virDomainObjIsActive(vm) &&
               (!vm->persistent ||
                (ret == 0 && (flags & VIR_MIGRATE_UNDEFINE_SOURCE)))) {
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
4581
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4582
        qemuDomainRemoveInactive(driver, vm);
4583 4584 4585
        vm = NULL;
    }

4586 4587 4588 4589 4590
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4591
 cleanup:
4592
    if (vm)
4593
        virObjectUnlock(vm);
4594 4595
    if (event)
        qemuDomainEventQueue(driver, event);
4596
    virObjectUnref(cfg);
4597 4598 4599 4600 4601 4602 4603
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4604
qemuMigrationPerformPhase(virQEMUDriverPtr driver,
4605 4606 4607
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *uri,
4608
                          const char *graphicsuri,
4609 4610 4611 4612 4613 4614 4615
                          const char *cookiein,
                          int cookieinlen,
                          char **cookieout,
                          int *cookieoutlen,
                          unsigned long flags,
                          unsigned long resource)
{
4616
    virObjectEventPtr event = NULL;
4617
    int ret = -1;
4618
    bool hasrefs;
4619 4620 4621 4622 4623 4624 4625 4626 4627 4628

    /* 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);
4629 4630
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
4631 4632 4633

    ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                          cookieout, cookieoutlen,
4634
                          flags, resource, NULL, graphicsuri);
4635

4636 4637 4638 4639 4640
    if (ret < 0) {
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
4641
        }
4642
        goto endjob;
4643
    }
4644 4645 4646

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4647 4648
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                             qemuMigrationCleanup) < 0)
4649 4650
        goto endjob;

4651
 endjob:
4652
    if (ret < 0)
4653
        hasrefs = qemuMigrationJobFinish(driver, vm);
4654
    else
4655 4656
        hasrefs = qemuMigrationJobContinue(vm);
    if (!hasrefs) {
4657 4658
        vm = NULL;
    } else if (!virDomainObjIsActive(vm) && !vm->persistent) {
4659
        qemuDomainRemoveInactive(driver, vm);
4660
        vm = NULL;
4661
    }
4662

4663
 cleanup:
4664
    if (vm)
4665
        virObjectUnlock(vm);
4666 4667 4668 4669 4670
    if (event)
        qemuDomainEventQueue(driver, event);
    return ret;
}

4671
int
4672
qemuMigrationPerform(virQEMUDriverPtr driver,
4673 4674 4675 4676 4677
                     virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *xmlin,
                     const char *dconnuri,
                     const char *uri,
4678
                     const char *graphicsuri,
4679
                     const char *listenAddress,
4680 4681 4682 4683 4684 4685 4686 4687 4688 4689
                     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, "
4690
              "uri=%s, graphicsuri=%s, listenAddress=%s"
4691 4692
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
              "flags=%lx, dname=%s, resource=%lu, v3proto=%d",
4693
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4694
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4695 4696
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4697 4698 4699

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4700 4701
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4702 4703 4704 4705
            return -1;
        }

        return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri, uri,
4706 4707
                                       graphicsuri, listenAddress,
                                       cookiein, cookieinlen,
4708 4709
                                       cookieout, cookieoutlen,
                                       flags, dname, resource, v3proto);
4710 4711
    } else {
        if (dconnuri) {
4712 4713
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
4714 4715 4716 4717 4718
            return -1;
        }

        if (v3proto) {
            return qemuMigrationPerformPhase(driver, conn, vm, uri,
4719
                                             graphicsuri,
4720 4721
                                             cookiein, cookieinlen,
                                             cookieout, cookieoutlen,
4722
                                             flags, resource);
4723 4724
        } else {
            return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri,
4725
                                           uri, graphicsuri, listenAddress,
4726
                                           cookiein, cookieinlen,
4727 4728 4729 4730 4731
                                           cookieout, cookieoutlen, flags,
                                           dname, resource, v3proto);
        }
    }
}
4732

4733
static int
4734 4735
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def)
{
4736
    size_t i;
4737 4738 4739 4740 4741
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
4742
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4743
            if (virNetDevVPortProfileAssociate(net->ifname,
4744
                                               virDomainNetGetActualVirtPortProfile(net),
4745
                                               &net->mac,
4746
                                               virDomainNetGetActualDirectDev(net),
4747
                                               -1,
4748
                                               def->uuid,
4749 4750
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
4751 4752 4753
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
4754
                goto err_exit;
4755 4756
            }
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
4757

4758
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
4759 4760 4761 4762
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
4763 4764 4765 4766
        }
        last_good_net = i;
    }

4767
    return 0;
4768

4769
 err_exit:
4770
    for (i = 0; last_good_net != -1 && i < last_good_net; i++) {
4771
        net = def->nets[i];
4772
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4773
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
4774
                                                           virDomainNetGetActualVirtPortProfile(net),
4775
                                                           &net->mac,
4776
                                                           virDomainNetGetActualDirectDev(net),
4777
                                                           -1,
4778
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
4779 4780
        }
    }
4781
    return -1;
4782 4783 4784 4785
}


virDomainPtr
4786
qemuMigrationFinish(virQEMUDriverPtr driver,
4787 4788
                    virConnectPtr dconn,
                    virDomainObjPtr vm,
4789 4790 4791 4792
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
4793
                    unsigned long flags,
4794 4795
                    int retcode,
                    bool v3proto)
4796 4797
{
    virDomainPtr dom = NULL;
4798
    virObjectEventPtr event = NULL;
4799
    bool newVM = true;
4800
    qemuMigrationCookiePtr mig = NULL;
4801
    virErrorPtr orig_err = NULL;
4802
    int cookie_flags = 0;
J
Jiri Denemark 已提交
4803
    qemuDomainObjPrivatePtr priv = vm->privateData;
4804
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4805
    virCapsPtr caps = NULL;
4806
    unsigned short port;
4807

4808
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
4809
              "cookieout=%p, cookieoutlen=%p, flags=%lx, retcode=%d",
4810 4811
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
4812

4813 4814 4815
    port = priv->migrationPort;
    priv->migrationPort = 0;

4816 4817 4818
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

4819
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
4820 4821
        goto cleanup;

4822 4823 4824
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
4825

4826
    qemuDomainCleanupRemove(vm, qemuMigrationPrepareCleanup);
4827
    VIR_FREE(priv->job.completed);
4828

4829
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
4830 4831
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
4832 4833 4834 4835 4836
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookie_flags |= QEMU_MIGRATION_COOKIE_PERSISTENT;

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein,
                                       cookieinlen, cookie_flags)))
4837
        goto endjob;
4838 4839 4840 4841 4842

    /* 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
4843
        if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4844 4845
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("guest unexpectedly quit"));
4846
            goto endjob;
4847 4848
        }

4849 4850 4851 4852 4853
        if (mig->jobInfo) {
            priv->job.completed = mig->jobInfo;
            mig->jobInfo = NULL;
        }

L
liguang 已提交
4854 4855 4856 4857 4858
        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");
4859
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
4860 4861 4862 4863 4864 4865 4866
                                                 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");
4867
        }
4868

4869 4870
        qemuMigrationStopNBDServer(driver, vm, mig);

4871
        if (flags & VIR_MIGRATE_PERSIST_DEST) {
4872
            virDomainDefPtr vmdef;
4873
            if (vm->persistent)
4874
                newVM = false;
4875
            vm->persistent = 1;
4876 4877 4878
            if (mig->persistent)
                vm->newDef = vmdef = mig->persistent;
            else
4879
                vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm);
4880
            if (!vmdef || virDomainSaveConfig(cfg->configDir, vmdef) < 0) {
4881 4882 4883 4884 4885 4886 4887 4888 4889
                /* 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.
                 */
4890 4891

                /*
4892 4893
                 * However, in v3 protocol, the source VM is still available
                 * to restart during confirm() step, so we kill it off now.
4894 4895
                 */
                if (v3proto) {
L
liguang 已提交
4896 4897 4898 4899 4900
                    if (!(flags & VIR_MIGRATE_OFFLINE)) {
                        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                        VIR_QEMU_PROCESS_STOP_MIGRATED);
                        virDomainAuditStop(vm, "failed");
                    }
4901 4902
                    if (newVM)
                        vm->persistent = 0;
4903
                }
A
Alex Jia 已提交
4904
                if (!vmdef)
4905 4906
                    virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                                   _("can't get vmdef"));
4907 4908 4909
                goto endjob;
            }

4910
            event = virDomainEventLifecycleNewFromObj(vm,
4911 4912 4913 4914 4915 4916 4917 4918 4919
                                             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
4920
        if (!(flags & VIR_MIGRATE_PAUSED) && !(flags & VIR_MIGRATE_OFFLINE)) {
4921 4922 4923 4924
            /* 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
4925
            if (qemuProcessStartCPUs(driver, vm, dconn,
4926 4927
                                     VIR_DOMAIN_RUNNING_MIGRATED,
                                     QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
4928
                if (virGetLastError() == NULL)
4929 4930
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   "%s", _("resume operation failed"));
4931 4932 4933 4934 4935
                /* Need to save the current error, in case shutting
                 * down the process overwrites it
                 */
                orig_err = virSaveLastError();

4936 4937 4938 4939 4940 4941 4942 4943 4944
                /*
                 * 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) {
4945 4946
                    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                    VIR_QEMU_PROCESS_STOP_MIGRATED);
4947
                    virDomainAuditStop(vm, "failed");
4948
                    event = virDomainEventLifecycleNewFromObj(vm,
4949 4950 4951
                                                     VIR_DOMAIN_EVENT_STOPPED,
                                                     VIR_DOMAIN_EVENT_STOPPED_FAILED);
                }
4952 4953
                goto endjob;
            }
4954 4955 4956 4957
            if (priv->job.completed) {
                qemuDomainJobInfoUpdateTime(priv->job.completed);
                qemuDomainJobInfoUpdateDowntime(priv->job.completed);
            }
4958 4959
        }

4960
        dom = virGetDomain(dconn, vm->def->name, vm->def->uuid);
4961

L
liguang 已提交
4962
        if (!(flags & VIR_MIGRATE_OFFLINE)) {
4963
            event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
4964 4965 4966 4967 4968 4969 4970
                                             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);
4971
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
4972 4973 4974
                                                 VIR_DOMAIN_EVENT_SUSPENDED,
                                                 VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);
            }
4975
        }
L
liguang 已提交
4976 4977

        if (virDomainObjIsActive(vm) &&
4978
            virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
4979 4980 4981
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto endjob;
        }
4982 4983 4984

        /* Guest is successfully running, so cancel previous auto destroy */
        qemuProcessAutoDestroyRemove(driver, vm);
L
liguang 已提交
4985
    } else if (!(flags & VIR_MIGRATE_OFFLINE)) {
4986 4987
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4988
        virDomainAuditStop(vm, "failed");
4989
        event = virDomainEventLifecycleNewFromObj(vm,
4990 4991 4992 4993
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_FAILED);
    }

4994 4995
    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen,
                                QEMU_MIGRATION_COOKIE_STATS) < 0)
4996 4997
        VIR_WARN("Unable to encode migration cookie");

4998
 endjob:
E
Eric Blake 已提交
4999 5000 5001
    if (qemuMigrationJobFinish(driver, vm) == 0) {
        vm = NULL;
    } else if (!vm->persistent && !virDomainObjIsActive(vm)) {
5002
        qemuDomainRemoveInactive(driver, vm);
E
Eric Blake 已提交
5003
        vm = NULL;
5004
    }
5005

5006
 cleanup:
5007
    virPortAllocatorRelease(driver->migrationPorts, port);
J
Jiri Denemark 已提交
5008
    if (vm) {
5009 5010
        if (priv->mon)
            qemuMonitorSetDomainLog(priv->mon, -1);
J
Jiri Denemark 已提交
5011
        VIR_FREE(priv->origname);
5012
        virObjectUnlock(vm);
J
Jiri Denemark 已提交
5013
    }
5014 5015
    if (event)
        qemuDomainEventQueue(driver, event);
5016
    qemuMigrationCookieFree(mig);
5017 5018 5019 5020
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5021
    virObjectUnref(caps);
5022
    virObjectUnref(cfg);
5023 5024
    return dom;
}
5025

5026

5027
/* Helper function called while vm is active.  */
5028
int
5029
qemuMigrationToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
5030 5031
                    int fd, off_t offset, const char *path,
                    const char *compressor,
E
Eric Blake 已提交
5032
                    bool bypassSecurityDriver,
5033
                    qemuDomainAsyncJob asyncJob)
5034 5035 5036
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5037
    int ret = -1;
5038
    bool restoreLabel = false;
5039 5040
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5041
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5042
    char *errbuf = NULL;
5043
    virErrorPtr orig_err = NULL;
5044 5045 5046 5047 5048

    /* 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,
5049 5050
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5051
        qemuDomainObjExitMonitor(driver, vm);
5052
    }
5053

5054 5055 5056 5057 5058 5059 5060
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5061
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5062
        (!compressor || pipe(pipeFD) == 0)) {
5063
        /* All right! We can use fd migration, which means that qemu
5064 5065 5066
         * 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.  */
5067
        if (virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
5068
                                              compressor ? pipeFD[1] : fd) < 0)
5069
            goto cleanup;
5070 5071 5072
        bypassSecurityDriver = true;
    } else {
        /* Phooey - we have to fall back on exec migration, where qemu
E
Eric Blake 已提交
5073 5074
         * has to popen() the file by name, and block devices have to be
         * given cgroup ACL permission.  We might also stumble on
5075 5076
         * a race present in some qemu versions where it does a wait()
         * that botches pclose.  */
5077 5078
        if (virCgroupHasController(priv->cgroup,
                                   VIR_CGROUP_CONTROLLER_DEVICES)) {
5079 5080 5081 5082
            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
5083
                /* path was not a device, no further need for cgroup */
5084
            } else if (rv < 0) {
5085 5086
                goto cleanup;
            }
5087
        }
5088 5089
        if ((!bypassSecurityDriver) &&
            virSecurityManagerSetSavedStateLabel(driver->securityManager,
5090
                                                 vm->def, path) < 0)
5091
            goto cleanup;
5092
        restoreLabel = true;
5093 5094
    }

5095
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5096 5097
        goto cleanup;

5098 5099 5100
    if (!compressor) {
        const char *args[] = { "cat", NULL };

5101
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117
            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
        };
5118 5119 5120 5121
        if (pipeFD[0] != -1) {
            cmd = virCommandNewArgs(args);
            virCommandSetInputFD(cmd, pipeFD[0]);
            virCommandSetOutputFD(cmd, &fd);
5122 5123
            virCommandSetErrorBuffer(cmd, &errbuf);
            virCommandDoAsyncIO(cmd);
5124 5125 5126
            if (virSetCloseExec(pipeFD[1]) < 0) {
                virReportSystemError(errno, "%s",
                                     _("Unable to set cloexec flag"));
5127
                qemuDomainObjExitMonitor(driver, vm);
5128 5129 5130
                goto cleanup;
            }
            if (virCommandRunAsync(cmd, NULL) < 0) {
5131
                qemuDomainObjExitMonitor(driver, vm);
5132 5133 5134 5135 5136 5137 5138
                goto cleanup;
            }
            rc = qemuMonitorMigrateToFd(priv->mon,
                                        QEMU_MONITOR_MIGRATE_BACKGROUND,
                                        pipeFD[1]);
            if (VIR_CLOSE(pipeFD[0]) < 0 ||
                VIR_CLOSE(pipeFD[1]) < 0)
5139
                VIR_WARN("failed to close intermediate pipe");
5140 5141 5142 5143 5144
        } else {
            rc = qemuMonitorMigrateToFile(priv->mon,
                                          QEMU_MONITOR_MIGRATE_BACKGROUND,
                                          args, path, offset);
        }
5145
    }
5146
    qemuDomainObjExitMonitor(driver, vm);
5147

5148 5149 5150 5151 5152 5153
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        goto cleanup;
    }

5154 5155 5156
    if (rc < 0)
        goto cleanup;

5157
    rc = qemuMigrationWaitForCompletion(driver, vm, asyncJob, NULL, false);
5158

5159 5160 5161 5162
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5163 5164
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5165 5166 5167 5168
                qemuMonitorMigrateCancel(priv->mon);
                qemuDomainObjExitMonitor(driver, vm);
            }
        }
5169
        goto cleanup;
5170
    }
5171

5172 5173 5174
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5175 5176
    ret = 0;

5177
 cleanup:
5178 5179 5180
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5181
    /* Restore max migration bandwidth */
5182 5183
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5184 5185
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5186
        qemuDomainObjExitMonitor(driver, vm);
5187 5188
    }

5189 5190
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5191 5192 5193 5194 5195
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5196 5197
    if (restoreLabel && (!bypassSecurityDriver) &&
        virSecurityManagerRestoreSavedStateLabel(driver->securityManager,
5198
                                                 vm->def, path) < 0)
5199 5200
        VIR_WARN("failed to restore save state label on %s", path);

5201 5202
    if (virCgroupHasController(priv->cgroup,
                               VIR_CGROUP_CONTROLLER_DEVICES)) {
5203 5204 5205
        int rv = virCgroupDenyDevicePath(priv->cgroup, path,
                                         VIR_CGROUP_DEVICE_RWM);
        virDomainAuditCgroupPath(vm, priv->cgroup, "deny", path, "rwm", rv == 0);
5206
    }
5207 5208 5209 5210 5211 5212

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

5213 5214
    return ret;
}
5215 5216

int
5217
qemuMigrationJobStart(virQEMUDriverPtr driver,
5218
                      virDomainObjPtr vm,
5219
                      qemuDomainAsyncJob job)
5220 5221 5222
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

5223
    if (qemuDomainObjBeginAsyncJob(driver, vm, job) < 0)
5224 5225
        return -1;

5226
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5227
        qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
5228
    } else {
5229 5230 5231
        qemuDomainObjSetAsyncJobMask(vm, (QEMU_JOB_DEFAULT_MASK |
                                          JOB_MASK(QEMU_JOB_SUSPEND) |
                                          JOB_MASK(QEMU_JOB_MIGRATION_OP)));
5232
    }
5233

J
Jiri Denemark 已提交
5234
    priv->job.current->type = VIR_DOMAIN_JOB_UNBOUNDED;
5235 5236 5237 5238 5239

    return 0;
}

void
5240
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5241
                         virDomainObjPtr vm,
5242
                         qemuMigrationJobPhase phase)
5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256
{
    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
5257
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5258
                           virDomainObjPtr vm,
5259
                           qemuMigrationJobPhase phase)
5260
{
5261
    virObjectRef(vm);
5262 5263 5264
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5265
bool
5266 5267
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5268
    qemuDomainObjReleaseAsyncJob(vm);
5269
    return virObjectUnref(vm);
5270 5271 5272 5273
}

bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
5274
                         qemuDomainAsyncJob job)
5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285
{
    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");

5286
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5287 5288 5289 5290 5291
        return false;
    }
    return true;
}

5292
bool
5293
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5294 5295 5296
{
    return qemuDomainObjEndAsyncJob(driver, vm);
}