qemu_migration.c 183.3 KB
Newer Older
1 2 3
/*
 * qemu_migration.c: QEMU migration handling
 *
4
 * Copyright (C) 2006-2014 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23
 *
 */

#include <config.h>

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

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

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

#define VIR_FROM_THIS VIR_FROM_QEMU

65 66
VIR_LOG_INIT("qemu.qemu_migration");

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

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

    QEMU_MIGRATION_COOKIE_FLAG_LAST
};

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

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

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

120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
typedef struct _qemuMigrationCookieNetData qemuMigrationCookieNetData;
typedef qemuMigrationCookieNetData *qemuMigrationCookieNetDataPtr;
struct _qemuMigrationCookieNetData {
    int vporttype; /* enum virNetDevVPortProfile */

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

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

    qemuMigrationCookieNetDataPtr net;
};

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

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

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

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

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

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

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

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

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

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

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

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

198

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

    if (!network)
        return;

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


216 217 218 219 220 221 222 223 224 225 226 227
static void qemuMigrationCookieNBDFree(qemuMigrationCookieNBDPtr nbd)
{
    if (!nbd)
        return;

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


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

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

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


247
#ifdef WITH_GNUTLS
248 249 250 251 252 253 254 255 256 257 258 259
static char *
qemuDomainExtractTLSSubject(const char *certdir)
{
    char *certfile = NULL;
    char *subject = NULL;
    char *pemdata = NULL;
    gnutls_datum_t pemdatum;
    gnutls_x509_crt_t cert;
    int ret;
    size_t subjectlen;

    if (virAsprintf(&certfile, "%s/server-cert.pem", certdir) < 0)
260
        goto error;
261 262

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

    ret = gnutls_x509_crt_init(&cert);
    if (ret < 0) {
270 271 272
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot initialize cert object: %s"),
                       gnutls_strerror(ret));
273 274 275 276 277 278 279 280
        goto error;
    }

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

    ret = gnutls_x509_crt_import(cert, &pemdatum, GNUTLS_X509_FMT_PEM);
    if (ret < 0) {
281 282 283
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot load cert data from %s: %s"),
                       certfile, gnutls_strerror(ret));
284 285 286 287 288
        goto error;
    }

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

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

    VIR_FREE(certfile);
    VIR_FREE(pemdata);

    return subject;

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

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

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

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

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

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

348
    virObjectUnref(cfg);
349 350
    return mig;

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


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

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

    mig->nnets = def->nnets;

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

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

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

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

            switch (vport->virtPortType) {
            case VIR_NETDEV_VPORT_PROFILE_NONE:
            case VIR_NETDEV_VPORT_PROFILE_8021QBG:
            case VIR_NETDEV_VPORT_PROFILE_8021QBH:
387
               break;
388
            case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
389 390
                if (virNetDevOpenvswitchGetMigrateData(&mig->net[i].portdata,
                                                       netptr->ifname) != 0) {
J
Jiri Denemark 已提交
391 392 393
                        virReportError(VIR_ERR_INTERNAL_ERROR,
                                       _("Unable to run command to get OVS port data for "
                                         "interface %s"), netptr->ifname);
394 395 396
                        goto error;
                }
                break;
397 398 399 400 401 402 403
            default:
                break;
            }
        }
    }
    return mig;

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

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

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

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

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

    return mig;

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


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

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

456 457 458 459 460 461 462 463
    for (i = 0; i < dom->def->ngraphics; i++) {
       if (dom->def->graphics[i]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
           if (!(mig->graphics =
                 qemuMigrationCookieGraphicsAlloc(driver, dom->def->graphics[i])))
               return -1;
           mig->flags |= QEMU_MIGRATION_COOKIE_GRAPHICS;
           break;
       }
464
    }
465 466 467 468 469

    return 0;
}


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

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

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

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

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

    return 0;
}


503 504 505 506 507
static int
qemuMigrationCookieAddPersistent(qemuMigrationCookiePtr mig,
                                 virDomainObjPtr dom)
{
    if (mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) {
508 509
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration persistent data already present"));
510 511 512 513 514 515 516 517 518 519 520 521 522
        return -1;
    }

    if (!dom->newDef)
        return 0;

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


523 524
static int
qemuMigrationCookieAddNetwork(qemuMigrationCookiePtr mig,
525
                              virQEMUDriverPtr driver,
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543
                              virDomainObjPtr dom)
{
    if (mig->flags & QEMU_MIGRATION_COOKIE_NETWORK) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Network migration data already present"));
        return -1;
    }

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

    return 0;
}

544

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

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

560 561 562 563 564 565 566 567 568 569 570 571 572 573
    if (vm->def->ndisks &&
        VIR_ALLOC_N(mig->nbd->disks, vm->def->ndisks) < 0)
        return -1;
    mig->nbd->ndisks = 0;

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

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

            qemuDomainObjEnterMonitor(driver, vm);
574 575
            if (qemuMonitorBlockStatsUpdateCapacity(priv->mon, stats,
                                                    false) < 0) {
576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598
                qemuDomainObjExitMonitor(driver, vm);
                goto cleanup;
            }
            qemuDomainObjExitMonitor(driver, vm);

            if (!virDomainObjIsActive(vm)) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("domain exited meanwhile"));
                goto cleanup;
            }
        }

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

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

599 600 601
    mig->nbd->port = priv->nbdPort;
    mig->flags |= QEMU_MIGRATION_COOKIE_NBD;

602 603 604 605
    ret = 0;
 cleanup:
    virHashFree(stats);
    return ret;
606 607 608
}


609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627
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;
}


628 629 630
static void qemuMigrationCookieGraphicsXMLFormat(virBufferPtr buf,
                                                 qemuMigrationCookieGraphicsPtr grap)
{
631
    virBufferAsprintf(buf, "<graphics type='%s' port='%d' listen='%s'",
632 633 634 635 636 637
                      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");
638 639 640 641
        virBufferAdjustIndent(buf, 2);
        virBufferEscapeString(buf, "<cert info='subject' value='%s'/>\n", grap->tlsSubject);
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</graphics>\n");
642 643 644 645 646 647
    } else {
        virBufferAddLit(buf, "/>\n");
    }
}


648 649 650 651
static void
qemuMigrationCookieNetworkXMLFormat(virBufferPtr buf,
                                    qemuMigrationCookieNetworkPtr optr)
{
652
    size_t i;
653 654 655 656 657 658
    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) {
659 660
                virBufferAddLit(buf, "<network>\n");
                virBufferAdjustIndent(buf, 2);
661 662
                empty = false;
            }
663
            virBufferAsprintf(buf, "<interface index='%zu' vporttype='%s'",
664 665 666
                              i, virNetDevVPortTypeToString(optr->net[i].vporttype));
            if (optr->net[i].portdata) {
                virBufferAddLit(buf, ">\n");
667 668
                virBufferAdjustIndent(buf, 2);
                virBufferEscapeString(buf, "<portdata>%s</portdata>\n",
669
                                      optr->net[i].portdata);
670 671
                virBufferAdjustIndent(buf, -2);
                virBufferAddLit(buf, "</interface>\n");
672 673 674 675 676
            } else {
                virBufferAddLit(buf, "/>\n");
            }
        }
    }
677 678 679 680
    if (!empty) {
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</network>\n");
    }
681 682 683
}


684 685 686 687 688 689 690 691 692
static void
qemuMigrationCookieStatisticsXMLFormat(virBufferPtr buf,
                                       qemuDomainJobInfoPtr jobInfo)
{
    qemuMonitorMigrationStatus *status = &jobInfo->status;

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

693 694 695
    virBufferAsprintf(buf, "<started>%llu</started>\n", jobInfo->started);
    virBufferAsprintf(buf, "<stopped>%llu</stopped>\n", jobInfo->stopped);

696 697 698 699 700 701 702 703 704 705
    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$s>\n",
                          VIR_DOMAIN_JOB_DOWNTIME,
                          status->downtime);
706 707 708 709
    if (status->setup_time_set)
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_SETUP_TIME,
                          status->setup_time);
710 711 712 713 714 715 716 717 718 719

    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$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_REMAINING,
                      status->ram_remaining);
720 721 722
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_BPS,
                      status->ram_bps);
723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744

    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$s>\n",
                      VIR_DOMAIN_JOB_DISK_REMAINING,
                      status->disk_remaining);
745 746 747
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_BPS,
                      status->disk_bps);
748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771

    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");
}


772
static int
773
qemuMigrationCookieXMLFormat(virQEMUDriverPtr driver,
774
                             virBufferPtr buf,
775
                             qemuMigrationCookiePtr mig)
776 777 778
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
    char hostuuidstr[VIR_UUID_STRING_BUFLEN];
779
    size_t i;
780 781

    virUUIDFormat(mig->uuid, uuidstr);
782
    virUUIDFormat(mig->localHostuuid, hostuuidstr);
783

784
    virBufferAddLit(buf, "<qemu-migration>\n");
785 786 787 788 789
    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);
790

791
    for (i = 0; i < QEMU_MIGRATION_COOKIE_FLAG_LAST; i++) {
792
        if (mig->flagsMandatory & (1 << i))
793
            virBufferAsprintf(buf, "<feature name='%s'/>\n",
794 795 796
                              qemuMigrationCookieFlagTypeToString(i));
    }

797 798
    if ((mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        mig->graphics)
799 800
        qemuMigrationCookieGraphicsXMLFormat(buf, mig->graphics);

801 802
    if ((mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        mig->lockState) {
803
        virBufferAsprintf(buf, "<lockstate driver='%s'>\n",
804
                          mig->lockDriver);
805 806
        virBufferAdjustIndent(buf, 2);
        virBufferAsprintf(buf, "<leases>%s</leases>\n",
807
                          mig->lockState);
808 809
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</lockstate>\n");
810 811
    }

812 813
    if ((mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        mig->persistent) {
814 815 816
        if (qemuDomainDefFormatBuf(driver,
                                   mig->persistent,
                                   VIR_DOMAIN_XML_INACTIVE |
817 818
                                   VIR_DOMAIN_XML_SECURE |
                                   VIR_DOMAIN_XML_MIGRATABLE,
819
                                   buf) < 0)
820
            return -1;
821 822
    }

823 824 825
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NETWORK) && mig->network)
        qemuMigrationCookieNetworkXMLFormat(buf, mig->network);

826
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NBD) && mig->nbd) {
827
        virBufferAddLit(buf, "<nbd");
828 829
        if (mig->nbd->port)
            virBufferAsprintf(buf, " port='%d'", mig->nbd->port);
830 831 832 833 834 835 836 837 838 839 840 841 842 843
        if (mig->nbd->ndisks) {
            virBufferAddLit(buf, ">\n");
            virBufferAdjustIndent(buf, 2);
            for (i = 0; i < mig->nbd->ndisks; i++) {
                virBufferEscapeString(buf, "<disk target='%s'",
                                      mig->nbd->disks[i].target);
                virBufferAsprintf(buf, " capacity='%llu'/>\n",
                                  mig->nbd->disks[i].capacity);
            }
            virBufferAdjustIndent(buf, -2);
            virBufferAddLit(buf, "</nbd>\n");
        } else {
            virBufferAddLit(buf, "/>\n");
        }
844 845
    }

846 847 848
    if (mig->flags & QEMU_MIGRATION_COOKIE_STATS && mig->jobInfo)
        qemuMigrationCookieStatisticsXMLFormat(buf, mig->jobInfo);

849
    virBufferAdjustIndent(buf, -2);
850
    virBufferAddLit(buf, "</qemu-migration>\n");
851
    return 0;
852 853 854
}


855
static char *qemuMigrationCookieXMLFormatStr(virQEMUDriverPtr driver,
856
                                             qemuMigrationCookiePtr mig)
857 858 859
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;

860
    if (qemuMigrationCookieXMLFormat(driver, &buf, mig) < 0) {
861 862 863
        virBufferFreeAndReset(&buf);
        return NULL;
    }
864

865
    if (virBufferCheckError(&buf) < 0)
866 867 868 869 870 871
        return NULL;

    return virBufferContentAndReset(&buf);
}


872 873 874 875 876 877 878
static qemuMigrationCookieGraphicsPtr
qemuMigrationCookieGraphicsXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieGraphicsPtr grap;
    char *tmp;

    if (VIR_ALLOC(grap) < 0)
879
        goto error;
880 881

    if (!(tmp = virXPathString("string(./graphics/@type)", ctxt))) {
882 883
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing type attribute in migration data"));
884 885 886
        goto error;
    }
    if ((grap->type = virDomainGraphicsTypeFromString(tmp)) < 0) {
887 888
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unknown graphics type %s"), tmp);
889 890 891
        VIR_FREE(tmp);
        goto error;
    }
E
Eric Blake 已提交
892
    VIR_FREE(tmp);
893
    if (virXPathInt("string(./graphics/@port)", ctxt, &grap->port) < 0) {
894 895
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing port attribute in migration data"));
896 897 898 899
        goto error;
    }
    if (grap->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        if (virXPathInt("string(./graphics/@tlsPort)", ctxt, &grap->tlsPort) < 0) {
900 901
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing tlsPort attribute in migration data"));
902 903 904 905
            goto error;
        }
    }
    if (!(grap->listen = virXPathString("string(./graphics/@listen)", ctxt))) {
906 907
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing listen attribute in migration data"));
908 909 910
        goto error;
    }
    /* Optional */
911
    grap->tlsSubject = virXPathString("string(./graphics/cert[@info='subject']/@value)", ctxt);
912 913 914

    return grap;

915
 error:
916 917 918 919 920
    qemuMigrationCookieGraphicsFree(grap);
    return NULL;
}


921 922 923 924
static qemuMigrationCookieNetworkPtr
qemuMigrationCookieNetworkXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieNetworkPtr optr;
925
    size_t i;
926 927 928 929 930 931
    int n;
    xmlNodePtr *interfaces = NULL;
    char *vporttype;
    xmlNodePtr save_ctxt = ctxt->node;

    if (VIR_ALLOC(optr) < 0)
932
        goto error;
933 934 935 936 937 938 939 940

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

    optr->nnets = n;
941
    if (VIR_ALLOC_N(optr->net, optr->nnets) < 0)
942
        goto error;
943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958

    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);

959
 cleanup:
960 961 962
    ctxt->node = save_ctxt;
    return optr;

963
 error:
964 965 966 967 968 969 970
    VIR_FREE(interfaces);
    qemuMigrationCookieNetworkFree(optr);
    optr = NULL;
    goto cleanup;
}


971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034
static qemuMigrationCookieNBDPtr
qemuMigrationCookieNBDXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieNBDPtr ret = NULL;
    char *port = NULL, *capacity = NULL;
    size_t i;
    int n;
    xmlNodePtr *disks = NULL;
    xmlNodePtr save_ctxt = ctxt->node;

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

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

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

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

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

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

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


1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
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;

1051 1052 1053
    virXPathULongLong("string(./started[1])", ctxt, &jobInfo->started);
    virXPathULongLong("string(./stopped[1])", ctxt, &jobInfo->stopped);

1054 1055 1056 1057 1058 1059 1060
    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;
1061 1062 1063
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_SETUP_TIME "[1])",
                          ctxt, &status->setup_time) == 0)
        status->setup_time_set = true;
1064 1065 1066 1067 1068 1069 1070

    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);
1071 1072
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_BPS "[1])",
                      ctxt, &status->ram_bps);
1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087

    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);
1088 1089
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_BPS "[1])",
                      ctxt, &status->disk_bps);
1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108

    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;
}


1109 1110
static int
qemuMigrationCookieXMLParse(qemuMigrationCookiePtr mig,
1111
                            virQEMUDriverPtr driver,
1112
                            xmlDocPtr doc,
1113
                            xmlXPathContextPtr ctxt,
E
Eric Blake 已提交
1114
                            unsigned int flags)
1115 1116
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
1117
    char *tmp = NULL;
1118
    xmlNodePtr *nodes = NULL;
1119 1120
    size_t i;
    int n;
1121 1122 1123 1124
    virCapsPtr caps = NULL;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto error;
1125 1126 1127 1128 1129 1130 1131 1132

    /* 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))) {
1133 1134
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing name element in migration data"));
1135 1136 1137
        goto error;
    }
    if (STRNEQ(tmp, mig->name)) {
1138 1139 1140
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected name %s vs %s"),
                       tmp, mig->name);
1141 1142 1143 1144 1145 1146 1147
        goto error;
    }
    VIR_FREE(tmp);

    /* Extract domain uuid */
    tmp = virXPathString("string(./uuid[1])", ctxt);
    if (!tmp) {
1148 1149
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing uuid element in migration data"));
1150 1151 1152 1153
        goto error;
    }
    virUUIDFormat(mig->uuid, uuidstr);
    if (STRNEQ(tmp, uuidstr)) {
1154 1155 1156
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected UUID %s vs %s"),
                       tmp, uuidstr);
1157 1158 1159 1160
    }
    VIR_FREE(tmp);

    /* Check & forbid "localhost" migration */
1161
    if (!(mig->remoteHostname = virXPathString("string(./hostname[1])", ctxt))) {
1162 1163
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostname element in migration data"));
1164 1165
        goto error;
    }
1166
    if (STREQ(mig->remoteHostname, mig->localHostname)) {
1167 1168 1169
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Attempt to migrate guest to the same host %s"),
                       mig->remoteHostname);
1170 1171 1172 1173
        goto error;
    }

    if (!(tmp = virXPathString("string(./hostuuid[1])", ctxt))) {
1174 1175 1176
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostuuid element in migration data"));
        goto error;
1177
    }
1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189
    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);
1190

1191 1192
    /* Check to ensure all mandatory features from XML are also
     * present in 'flags' */
1193
    if ((n = virXPathNodeSet("./feature", ctxt, &nodes)) < 0)
1194 1195
        goto error;

1196
    for (i = 0; i < n; i++) {
1197 1198 1199
        int val;
        char *str = virXMLPropString(nodes[i], "name");
        if (!str) {
1200 1201
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing feature name"));
1202 1203 1204 1205
            goto error;
        }

        if ((val = qemuMigrationCookieFlagTypeFromString(str)) < 0) {
1206 1207 1208
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unknown migration cookie feature %s"),
                           str);
1209 1210 1211 1212 1213
            VIR_FREE(str);
            goto error;
        }

        if ((flags & (1 << val)) == 0) {
1214 1215 1216
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unsupported migration cookie feature %s"),
                           str);
1217 1218 1219 1220 1221 1222
            VIR_FREE(str);
        }
        VIR_FREE(str);
    }
    VIR_FREE(nodes);

1223 1224 1225 1226 1227
    if ((flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        virXPathBoolean("count(./graphics) > 0", ctxt) &&
        (!(mig->graphics = qemuMigrationCookieGraphicsXMLParse(ctxt))))
        goto error;

1228 1229 1230 1231
    if ((flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        virXPathBoolean("count(./lockstate) > 0", ctxt)) {
        mig->lockDriver = virXPathString("string(./lockstate[1]/@driver)", ctxt);
        if (!mig->lockDriver) {
1232 1233
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Missing lock driver name in migration cookie"));
1234 1235 1236 1237 1238 1239 1240
            goto error;
        }
        mig->lockState = virXPathString("string(./lockstate[1]/leases[1])", ctxt);
        if (mig->lockState && STREQ(mig->lockState, ""))
            VIR_FREE(mig->lockState);
    }

1241 1242 1243
    if ((flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        virXPathBoolean("count(./domain) > 0", ctxt)) {
        if ((n = virXPathNodeSet("./domain", ctxt, &nodes)) > 1) {
1244 1245 1246 1247
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Too many domain elements in "
                             "migration cookie: %d"),
                           n);
1248 1249
            goto error;
        }
1250 1251
        mig->persistent = virDomainDefParseNode(doc, nodes[0],
                                                caps, driver->xmlopt,
1252
                                                -1, VIR_DOMAIN_DEF_PARSE_INACTIVE);
1253 1254 1255 1256 1257 1258 1259 1260
        if (!mig->persistent) {
            /* virDomainDefParseNode already reported
             * an error for us */
            goto error;
        }
        VIR_FREE(nodes);
    }

1261 1262 1263 1264 1265
    if ((flags & QEMU_MIGRATION_COOKIE_NETWORK) &&
        virXPathBoolean("count(./network) > 0", ctxt) &&
        (!(mig->network = qemuMigrationCookieNetworkXMLParse(ctxt))))
        goto error;

1266
    if (flags & QEMU_MIGRATION_COOKIE_NBD &&
1267 1268 1269
        virXPathBoolean("boolean(./nbd)", ctxt) &&
        (!(mig->nbd = qemuMigrationCookieNBDXMLParse(ctxt))))
        goto error;
1270

1271 1272 1273 1274 1275
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        virXPathBoolean("boolean(./statistics)", ctxt) &&
        (!(mig->jobInfo = qemuMigrationCookieStatisticsXMLParse(ctxt))))
        goto error;

1276
    virObjectUnref(caps);
1277 1278
    return 0;

1279
 error:
1280
    VIR_FREE(tmp);
1281
    VIR_FREE(nodes);
1282
    virObjectUnref(caps);
1283 1284 1285 1286 1287 1288
    return -1;
}


static int
qemuMigrationCookieXMLParseStr(qemuMigrationCookiePtr mig,
1289
                               virQEMUDriverPtr driver,
1290
                               const char *xml,
E
Eric Blake 已提交
1291
                               unsigned int flags)
1292 1293 1294
{
    xmlDocPtr doc = NULL;
    xmlXPathContextPtr ctxt = NULL;
1295
    int ret = -1;
1296 1297 1298

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

1299
    if (!(doc = virXMLParseStringCtxt(xml, _("(qemu_migration_cookie)"), &ctxt)))
1300 1301
        goto cleanup;

1302
    ret = qemuMigrationCookieXMLParse(mig, driver, doc, ctxt, flags);
1303

1304
 cleanup:
1305 1306 1307 1308 1309 1310 1311 1312 1313
    xmlXPathFreeContext(ctxt);
    xmlFreeDoc(doc);

    return ret;
}


static int
qemuMigrationBakeCookie(qemuMigrationCookiePtr mig,
1314
                        virQEMUDriverPtr driver,
1315
                        virDomainObjPtr dom,
1316 1317
                        char **cookieout,
                        int *cookieoutlen,
E
Eric Blake 已提交
1318
                        unsigned int flags)
1319
{
1320 1321
    if (!cookieout || !cookieoutlen)
        return 0;
1322 1323 1324

    *cookieoutlen = 0;

1325 1326 1327 1328
    if (flags & QEMU_MIGRATION_COOKIE_GRAPHICS &&
        qemuMigrationCookieAddGraphics(mig, driver, dom) < 0)
        return -1;

1329 1330 1331 1332
    if (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE &&
        qemuMigrationCookieAddLockstate(mig, driver, dom) < 0)
        return -1;

1333 1334 1335 1336
    if (flags & QEMU_MIGRATION_COOKIE_PERSISTENT &&
        qemuMigrationCookieAddPersistent(mig, dom) < 0)
        return -1;

1337 1338 1339 1340 1341
    if (flags & QEMU_MIGRATION_COOKIE_NETWORK &&
        qemuMigrationCookieAddNetwork(mig, driver, dom) < 0) {
        return -1;
    }

1342 1343 1344 1345
    if ((flags & QEMU_MIGRATION_COOKIE_NBD) &&
        qemuMigrationCookieAddNBD(mig, driver, dom) < 0)
        return -1;

1346 1347 1348 1349
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        qemuMigrationCookieAddStatistics(mig, dom) < 0)
        return -1;

1350
    if (!(*cookieout = qemuMigrationCookieXMLFormatStr(driver, mig)))
1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361
        return -1;

    *cookieoutlen = strlen(*cookieout) + 1;

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

    return 0;
}


static qemuMigrationCookiePtr
1362
qemuMigrationEatCookie(virQEMUDriverPtr driver,
1363
                       virDomainObjPtr dom,
1364 1365
                       const char *cookiein,
                       int cookieinlen,
E
Eric Blake 已提交
1366
                       unsigned int flags)
1367 1368 1369 1370 1371 1372
{
    qemuMigrationCookiePtr mig = NULL;

    /* Parse & validate incoming cookie (if any) */
    if (cookiein && cookieinlen &&
        cookiein[cookieinlen-1] != '\0') {
1373 1374
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration cookie was not NULL terminated"));
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384
        goto error;
    }

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

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

    if (cookiein && cookieinlen &&
        qemuMigrationCookieXMLParseStr(mig,
1385
                                       driver,
1386 1387 1388 1389
                                       cookiein,
                                       flags) < 0)
        goto error;

1390 1391 1392
    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
        if (!mig->lockDriver) {
            if (virLockManagerPluginUsesState(driver->lockManager)) {
1393 1394 1395
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Missing %s lock state for migration cookie"),
                               virLockManagerPluginGetName(driver->lockManager));
1396 1397 1398 1399
                goto error;
            }
        } else if (STRNEQ(mig->lockDriver,
                          virLockManagerPluginGetName(driver->lockManager))) {
1400 1401 1402 1403
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Source host lock driver %s different from target %s"),
                           mig->lockDriver,
                           virLockManagerPluginGetName(driver->lockManager));
1404 1405 1406 1407
            goto error;
        }
    }

1408 1409
    return mig;

1410
 error:
1411 1412 1413
    qemuMigrationCookieFree(mig);
    return NULL;
}
1414

1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461
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;
}

1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614

static int
qemuMigrationPrecreateDisk(virConnectPtr conn,
                           virDomainDiskDefPtr disk,
                           unsigned long long capacity)
{
    int ret = -1;
    virStoragePoolPtr pool = NULL;
    virStorageVolPtr vol = NULL;
    char *volName = NULL, *basePath = NULL;
    char *volStr = NULL;
    virBuffer buf = VIR_BUFFER_INITIALIZER;
    const char *format = NULL;
    unsigned int flags = 0;

    VIR_DEBUG("Precreate disk type=%s", virStorageTypeToString(disk->src->type));

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

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

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

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

        if (!(pool = storagePoolLookupByTargetPath(conn, basePath)))
            goto cleanup;
        format = virStorageFileFormatTypeToString(disk->src->format);
        if (disk->src->format == VIR_STORAGE_FILE_QCOW2)
            flags |= VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA;
        break;

    case VIR_STORAGE_TYPE_VOLUME:
        if (!(pool = virStoragePoolLookupByName(conn, disk->src->srcpool->pool)))
            goto cleanup;
        format = virStorageFileFormatTypeToString(disk->src->format);
        volName = disk->src->srcpool->volume;
        if (disk->src->format == VIR_STORAGE_FILE_QCOW2)
            flags |= VIR_STORAGE_VOL_CREATE_PREALLOC_METADATA;
        break;

    case VIR_STORAGE_TYPE_BLOCK:
    case VIR_STORAGE_TYPE_DIR:
    case VIR_STORAGE_TYPE_NETWORK:
    case VIR_STORAGE_TYPE_NONE:
    case VIR_STORAGE_TYPE_LAST:
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot precreate storage for disk type '%s'"),
                       virStorageTypeToString(disk->src->type));
        goto cleanup;
        break;
    }

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

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

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

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

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


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

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

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

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

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

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

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

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

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

    ret = 0;
 cleanup:
    return ret;
}


1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628
/**
 * 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
1629 1630
                            virDomainObjPtr vm,
                            const char *listenAddr)
1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641
{
    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 */
1642 1643
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1644 1645 1646 1647
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1648
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1649 1650 1651 1652 1653 1654 1655
            goto cleanup;

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

        if (!port &&
1656
            ((virPortAllocatorAcquire(driver->migrationPorts, &port) < 0) ||
1657
             (qemuMonitorNBDServerStart(priv->mon, listenAddr, port) < 0))) {
1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671
            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;

1672
 cleanup:
1673
    VIR_FREE(diskAlias);
1674
    if (ret < 0)
1675
        virPortAllocatorRelease(driver->migrationPorts, port);
1676 1677 1678
    return ret;
}

1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712
/**
 * 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;
1713
    char *hoststr = NULL;
1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732
    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;

1733 1734
    /* escape literal IPv6 address */
    if (strchr(host, ':')) {
1735
        if (virAsprintf(&hoststr, "[%s]", host) < 0)
1736 1737 1738 1739 1740
            goto error;
    } else if (VIR_STRDUP(hoststr, host) < 0) {
        goto error;
    }

1741 1742 1743 1744 1745 1746 1747 1748
    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 */
1749 1750
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1751 1752 1753 1754 1755 1756 1757
            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",
1758
                         hoststr, port, diskAlias) < 0))
1759 1760 1761 1762 1763 1764
            goto error;

        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            goto error;
        mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest,
1765
                                         NULL, speed, 0, 0, mirror_flags);
1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787
        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);
1788
                priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
1789 1790 1791 1792 1793
                virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                               qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                               _("canceled by client"));
                goto error;
            }
1794 1795
            mon_ret = qemuMonitorBlockJobInfo(priv->mon, diskAlias, &info,
                                              NULL);
1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822
            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;

1823
 cleanup:
1824 1825
    VIR_FREE(diskAlias);
    VIR_FREE(nbd_dest);
1826
    VIR_FREE(hoststr);
1827 1828
    return ret;

1829
 error:
1830 1831 1832 1833 1834 1835 1836
    /* don't overwrite any errors */
    err = virSaveLastError();
    /* cancel any outstanding jobs */
    while (lastGood) {
        virDomainDiskDefPtr disk = vm->def->disks[--lastGood];

        /* skip shared, RO disks */
1837 1838
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1839 1840 1841 1842
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1843
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1844 1845 1846
            continue;
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
1847
            if (qemuMonitorBlockJob(priv->mon, diskAlias, NULL, NULL, 0,
1848
                                    BLOCK_JOB_ABORT, true) < 0) {
1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
                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;
}
1861

1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881

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);

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

1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900
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 */
1901 1902
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1903 1904 1905 1906
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1907
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1908 1909 1910 1911 1912 1913
            goto cleanup;

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

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

1920
 cleanup:
1921 1922 1923 1924
    VIR_FREE(diskAlias);
    return;
}

1925 1926 1927 1928 1929 1930 1931 1932 1933
/* 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. */
1934
bool
1935
qemuMigrationIsAllowed(virQEMUDriverPtr driver, virDomainObjPtr vm,
1936
                       virDomainDefPtr def, bool remote, bool abort_on_error)
1937
{
1938
    int nsnapshots;
1939
    int pauseReason;
1940
    bool forbid;
1941
    size_t i;
1942 1943 1944

    if (vm) {
        if (qemuProcessAutoDestroyActive(driver, vm)) {
1945 1946
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("domain is marked for auto destroy"));
1947 1948
            return false;
        }
1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961

        /* 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;
            }
1962 1963

            /* cancel migration if disk I/O error is emitted while migrating */
1964 1965
            if (abort_on_error &&
                virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
1966 1967 1968 1969 1970 1971
                pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("cannot migrate domain with I/O error"));
                return false;
            }

1972
        }
1973

E
Eric Blake 已提交
1974 1975
        if (virDomainHasDiskMirror(vm)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1976
                           _("domain has an active block job"));
E
Eric Blake 已提交
1977 1978
            return false;
        }
1979 1980 1981

        def = vm->def;
    }
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996

    /* 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",
1997
                       _("domain has assigned non-USB host devices"));
1998 1999 2000
        return false;
    }

2001 2002 2003
    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
2004

2005 2006
            if (feature->policy != VIR_CPU_FEATURE_REQUIRE)
                continue;
J
Ján Tomko 已提交
2007

2008 2009 2010 2011 2012 2013 2014
            /* 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
2015 2016 2017
        }
    }

2018 2019 2020
    return true;
}

2021 2022 2023
static bool
qemuMigrationIsSafe(virDomainDefPtr def)
{
2024
    size_t i;
2025

2026
    for (i = 0; i < def->ndisks; i++) {
2027
        virDomainDiskDefPtr disk = def->disks[i];
2028
        const char *src = virDomainDiskGetSource(disk);
2029

2030 2031
        /* Our code elsewhere guarantees shared disks are either readonly (in
         * which case cache mode doesn't matter) or used with cache=none */
2032
        if (src &&
2033 2034
            !disk->src->shared &&
            !disk->src->readonly &&
2035
            disk->cachemode != VIR_DOMAIN_DISK_CACHE_DISABLE) {
2036
            int rc;
2037

E
Eric Blake 已提交
2038
            if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
2039
                if ((rc = virFileIsSharedFS(src)) < 0)
2040 2041
                    return false;
                else if (rc == 0)
2042
                    continue;
2043
                if ((rc = virStorageFileIsClusterFS(src)) < 0)
2044
                    return false;
2045 2046
                else if (rc == 1)
                    continue;
2047 2048
            } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                       disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
2049
                continue;
2050
            }
2051

2052 2053 2054
            virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                           _("Migration may lead to data corruption if disks"
                             " use cache != none"));
2055 2056 2057 2058 2059 2060 2061
            return false;
        }
    }

    return true;
}

2062 2063 2064 2065
/** qemuMigrationSetOffline
 * Pause domain for non-live migration.
 */
int
2066
qemuMigrationSetOffline(virQEMUDriverPtr driver,
2067 2068 2069
                        virDomainObjPtr vm)
{
    int ret;
2070
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
2071 2072
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
2073
    if (ret == 0) {
2074
        virObjectEventPtr event;
2075

2076
        event = virDomainEventLifecycleNewFromObj(vm,
2077 2078 2079 2080 2081 2082 2083 2084 2085 2086
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
        if (event)
            qemuDomainEventQueue(driver, event);
    }

    return ret;
}


2087 2088 2089
static int
qemuMigrationSetCompression(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
2090
                            bool state,
2091
                            qemuDomainAsyncJob job)
2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104
{
    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;
2105 2106 2107
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123
    } 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,
2124 2125
                QEMU_MONITOR_MIGRATION_CAPS_XBZRLE,
                state);
2126

2127
 cleanup:
2128 2129 2130 2131
    qemuDomainObjExitMonitor(driver, vm);
    return ret;
}

2132 2133 2134
static int
qemuMigrationSetAutoConverge(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
2135
                             bool state,
2136
                             qemuDomainAsyncJob job)
2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149
{
    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;
2150 2151 2152
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2153 2154 2155 2156 2157 2158 2159 2160 2161 2162
    } 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,
2163 2164
                QEMU_MONITOR_MIGRATION_CAPS_AUTO_CONVERGE,
                state);
2165

2166
 cleanup:
2167 2168 2169 2170 2171
    qemuDomainObjExitMonitor(driver, vm);
    return ret;
}


2172 2173 2174
static int
qemuMigrationSetPinAll(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
2175
                       bool state,
2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189
                       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;
2190 2191 2192
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208
    } 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,
2209 2210
                QEMU_MONITOR_MIGRATION_CAPS_RDMA_PIN_ALL,
                state);
2211 2212 2213 2214 2215 2216

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

2217 2218 2219 2220 2221 2222 2223
static int
qemuMigrationWaitForSpice(virQEMUDriverPtr driver,
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool wait_for_spice = false;
    bool spice_migrated = false;
2224
    size_t i = 0;
2225

2226 2227 2228 2229 2230 2231 2232 2233
    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;
            }
        }
    }
2234 2235 2236 2237 2238 2239 2240 2241

    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 };

2242 2243 2244 2245
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            return -1;

2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258
        if (qemuMonitorGetSpiceMigrationStatus(priv->mon,
                                               &spice_migrated) < 0) {
            qemuDomainObjExitMonitor(driver, vm);
            return -1;
        }
        qemuDomainObjExitMonitor(driver, vm);
        virObjectUnlock(vm);
        nanosleep(&ts, NULL);
        virObjectLock(vm);
    }

    return 0;
}
2259

2260
static int
2261
qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver,
2262
                             virDomainObjPtr vm,
2263
                             const char *job,
2264
                             qemuDomainAsyncJob asyncJob)
2265
{
2266
    qemuDomainObjPrivatePtr priv = vm->privateData;
2267
    qemuMonitorMigrationStatus status;
J
Jiri Denemark 已提交
2268 2269
    qemuDomainJobInfoPtr jobInfo;
    int ret;
2270 2271

    memset(&status, 0, sizeof(status));
2272

2273 2274
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
    if (ret < 0) {
2275 2276 2277
        /* Guest already exited or waiting for the job timed out; nothing
         * further to update. */
        return ret;
2278
    }
2279
    ret = qemuMonitorGetMigrationStatus(priv->mon, &status);
2280

2281
    qemuDomainObjExitMonitor(driver, vm);
2282

J
Jiri Denemark 已提交
2283 2284
    if (ret < 0 ||
        qemuDomainJobInfoUpdateTime(priv->job.current) < 0)
2285
        return -1;
2286

2287
    ret = -1;
J
Jiri Denemark 已提交
2288 2289 2290 2291 2292
    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
2293
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
2294 2295 2296 2297
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
        ret = 0;
        break;

J
Jiri Denemark 已提交
2298 2299 2300 2301
    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
        jobInfo->type = VIR_DOMAIN_JOB_NONE;
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("is not active"));
2302 2303 2304
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
J
Jiri Denemark 已提交
2305
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
2306 2307
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("unexpectedly failed"));
2308 2309 2310
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
J
Jiri Denemark 已提交
2311
        jobInfo->type = VIR_DOMAIN_JOB_CANCELLED;
2312 2313
        virReportError(VIR_ERR_OPERATION_ABORTED,
                       _("%s: %s"), job, _("canceled by client"));
2314 2315
        break;
    }
J
Jiri Denemark 已提交
2316
    jobInfo->status = status;
2317 2318 2319 2320 2321

    return ret;
}


2322 2323 2324
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
2325
static int
J
Jiri Denemark 已提交
2326 2327
qemuMigrationWaitForCompletion(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
2328
                               qemuDomainAsyncJob asyncJob,
J
Jiri Denemark 已提交
2329 2330
                               virConnectPtr dconn,
                               bool abort_on_error)
2331
{
2332
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
2333
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
2334
    const char *job;
2335
    int pauseReason;
2336

2337 2338
    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
2339 2340
        job = _("migration job");
        break;
2341
    case QEMU_ASYNC_JOB_SAVE:
2342 2343
        job = _("domain save job");
        break;
2344
    case QEMU_ASYNC_JOB_DUMP:
2345 2346 2347 2348 2349
        job = _("domain core dump job");
        break;
    default:
        job = _("job");
    }
2350

J
Jiri Denemark 已提交
2351
    jobInfo->type = VIR_DOMAIN_JOB_UNBOUNDED;
2352

J
Jiri Denemark 已提交
2353
    while (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
2354 2355 2356
        /* Poll every 50ms for progress & to allow cancellation */
        struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };

2357
        if (qemuMigrationUpdateJobStatus(driver, vm, job, asyncJob) == -1)
2358
            break;
2359

2360
        /* cancel migration if disk I/O error is emitted while migrating */
2361
        if (abort_on_error &&
2362
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
2363 2364 2365 2366 2367
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("%s: %s"), job, _("failed due to I/O error"));
            break;
        }
2368

2369
        if (dconn && virConnectIsAlive(dconn) <= 0) {
2370 2371
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
2372
            break;
2373 2374
        }

2375
        virObjectUnlock(vm);
2376 2377 2378

        nanosleep(&ts, NULL);

2379
        virObjectLock(vm);
2380 2381
    }

J
Jiri Denemark 已提交
2382
    if (jobInfo->type == VIR_DOMAIN_JOB_COMPLETED) {
2383
        qemuDomainJobInfoUpdateDowntime(jobInfo);
2384 2385 2386
        VIR_FREE(priv->job.completed);
        if (VIR_ALLOC(priv->job.completed) == 0)
            *priv->job.completed = *jobInfo;
2387
        return 0;
J
Jiri Denemark 已提交
2388
    } else if (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
2389 2390 2391
        /* 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
2392
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
2393 2394
        return -2;
    } else {
2395
        return -1;
2396
    }
2397 2398 2399
}


2400
static int
2401
qemuDomainMigrateGraphicsRelocate(virQEMUDriverPtr driver,
2402
                                  virDomainObjPtr vm,
2403 2404
                                  qemuMigrationCookiePtr cookie,
                                  const char *graphicsuri)
2405 2406
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2407 2408
    int ret = -1;
    const char *listenAddress = NULL;
2409
    virSocketAddr addr;
2410 2411 2412 2413 2414
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
2415

2416
    if (!cookie || (!cookie->graphics && !graphicsuri))
2417 2418
        return 0;

2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437
    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) {
2438
        size_t i;
2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465

        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;
            }
        }
    }
2466 2467 2468 2469

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
2470 2471 2472 2473
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
2474

2475 2476 2477 2478
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
2479
        qemuDomainObjExitMonitor(driver, vm);
2480
    }
2481

2482
 cleanup:
2483
    virURIFree(uri);
2484 2485 2486 2487
    return ret;
}


2488
static int
2489
qemuDomainMigrateOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
2490 2491 2492
                             virDomainObjPtr vm,
                             qemuMigrationCookiePtr cookie)
{
2493 2494
    virDomainNetDefPtr netptr;
    int ret = -1;
2495
    size_t i;
2496 2497 2498 2499 2500 2501 2502 2503

    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:
2504
           break;
2505
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
2506 2507
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
2508 2509 2510
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
2511 2512 2513
                goto cleanup;
            }
            break;
2514 2515 2516 2517 2518
        default:
            break;
        }
    }

2519
    ret = 0;
2520
 cleanup:
2521 2522 2523 2524
    return ret;
}


2525 2526 2527 2528 2529 2530
/* 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.
 */
2531 2532 2533 2534
static virDomainObjPtr
qemuMigrationCleanup(virDomainObjPtr vm,
                     virConnectPtr conn,
                     void *opaque)
2535
{
2536
    virQEMUDriverPtr driver = opaque;
2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551
    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);

2552
    switch ((qemuMigrationJobPhase) priv->job.phase) {
2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582
    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 */
        ;
    }

2583
 cleanup:
2584 2585 2586
    return vm;
}

2587

2588
/* The caller is supposed to lock the vm and start a migration job. */
2589 2590
static char
*qemuMigrationBeginPhase(virQEMUDriverPtr driver,
2591
                         virDomainObjPtr vm,
2592
                         const char *xmlin,
2593
                         const char *dname,
2594
                         char **cookieout,
2595 2596
                         int *cookieoutlen,
                         unsigned long flags)
2597 2598 2599
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
2600
    virDomainDefPtr def = NULL;
2601
    qemuDomainObjPrivatePtr priv = vm->privateData;
2602
    virCapsPtr caps = NULL;
2603
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
2604
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2605

2606
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
2607
              " cookieout=%p, cookieoutlen=%p, flags=%lx",
2608
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
2609
              cookieout, cookieoutlen, flags);
2610

2611 2612 2613
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2614 2615 2616 2617 2618 2619
    /* 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);
2620

2621
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
2622 2623
        goto cleanup;

2624 2625 2626
    if (!(flags & VIR_MIGRATE_UNSAFE) && !qemuMigrationIsSafe(vm->def))
        goto cleanup;

2627
    if (flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
2628
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DRIVE_MIRROR)) {
2629
        /* TODO support NBD for TUNNELLED migration */
2630
        if (flags & VIR_MIGRATE_TUNNELLED) {
2631 2632 2633 2634
            VIR_WARN("NBD in tunnelled migration is currently not supported");
        } else {
            cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
            priv->nbdPort = 0;
2635 2636 2637
        }
    }

2638
    if (!(mig = qemuMigrationEatCookie(driver, vm, NULL, 0, 0)))
2639 2640 2641 2642
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
                                cookieout, cookieoutlen,
2643
                                cookieFlags) < 0)
2644 2645
        goto cleanup;

L
liguang 已提交
2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667
    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;
        }
    }

2668
    if (xmlin) {
2669
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt,
M
Matthias Bolte 已提交
2670
                                            QEMU_EXPECTED_VIRT_TYPES,
2671
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE)))
2672 2673
            goto cleanup;

2674
        if (!qemuDomainDefCheckABIStability(driver, vm->def, def))
2675 2676
            goto cleanup;

2677
        rv = qemuDomainDefFormatLive(driver, def, false, true);
2678
    } else {
2679
        rv = qemuDomainDefFormatLive(driver, vm->def, false, true);
2680
    }
2681

2682
 cleanup:
2683
    qemuMigrationCookieFree(mig);
2684
    virObjectUnref(caps);
2685
    virDomainDefFree(def);
2686 2687 2688
    return rv;
}

2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699
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;
2700
    qemuDomainAsyncJob asyncJob;
2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711

    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;
    }

2712 2713
    qemuMigrationStoreDomainState(vm);

2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736
    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.
         */
2737 2738
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                 qemuMigrationCleanup) < 0)
2739
            goto endjob;
2740
        qemuMigrationJobContinue(vm);
2741 2742 2743 2744
    } else {
        goto endjob;
    }

2745
 cleanup:
2746
    qemuDomObjEndAPI(&vm);
2747 2748
    return xml;

2749
 endjob:
2750 2751 2752 2753
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
2754 2755 2756
    goto cleanup;
}

2757

2758 2759
/* Prepare is the first step, and it runs on the destination host.
 */
2760

2761
static void
2762
qemuMigrationPrepareCleanup(virQEMUDriverPtr driver,
2763 2764 2765 2766 2767 2768 2769 2770 2771 2772
                            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));

2773 2774 2775
    virPortAllocatorRelease(driver->migrationPorts, priv->migrationPort);
    priv->migrationPort = 0;

2776 2777 2778 2779 2780
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2781
static int
2782
qemuMigrationPrepareAny(virQEMUDriverPtr driver,
2783 2784 2785 2786 2787
                        virConnectPtr dconn,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
2788
                        virDomainDefPtr *def,
2789
                        const char *origname,
L
liguang 已提交
2790
                        virStreamPtr st,
2791
                        const char *protocol,
2792 2793
                        unsigned short port,
                        bool autoPort,
2794
                        const char *listenAddress,
L
liguang 已提交
2795
                        unsigned long flags)
2796 2797
{
    virDomainObjPtr vm = NULL;
2798
    virObjectEventPtr event = NULL;
2799
    int ret = -1;
2800
    int dataFD[2] = { -1, -1 };
2801
    qemuDomainObjPrivatePtr priv = NULL;
J
Jiri Denemark 已提交
2802
    unsigned long long now;
2803
    qemuMigrationCookiePtr mig = NULL;
2804
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2805
    char *xmlout = NULL;
L
liguang 已提交
2806
    unsigned int cookieFlags;
2807
    virCapsPtr caps = NULL;
J
Ján Tomko 已提交
2808
    char *migrateFrom = NULL;
2809
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2810
    bool taint_hook = false;
2811

2812
    if (virTimeMillisNow(&now) < 0)
2813 2814
        return -1;

2815 2816
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838
    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;
        }
    }

2839 2840 2841
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2842
    if (!qemuMigrationIsAllowed(driver, NULL, *def, true, abort_on_error))
2843 2844
        goto cleanup;

J
Jiri Denemark 已提交
2845 2846 2847 2848 2849
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2850
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2851 2852
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2853 2854
            goto cleanup;

2855
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2856 2857 2858 2859 2860 2861 2862
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
2863
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
2864 2865 2866 2867 2868 2869
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2870
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt,
J
Jiri Denemark 已提交
2871
                                                 QEMU_EXPECTED_VIRT_TYPES,
2872
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE);
J
Jiri Denemark 已提交
2873 2874 2875
                if (!newdef)
                    goto cleanup;

2876
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2877 2878 2879 2880
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2881 2882
                virDomainDefFree(*def);
                *def = newdef;
2883 2884 2885 2886
                /* 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
2887 2888 2889 2890
            }
        }
    }

J
Ján Tomko 已提交
2891 2892 2893 2894
    if (tunnel) {
        /* QEMU will be started with -incoming stdio
         * (which qemu_command might convert to exec:cat or fd:n)
         */
2895
        if (VIR_STRDUP(migrateFrom, "stdio") < 0)
J
Ján Tomko 已提交
2896 2897
            goto cleanup;
    } else {
2898 2899 2900
        bool encloseAddress = false;
        bool hostIPv6Capable = false;
        bool qemuIPv6Capable = false;
J
Ján Tomko 已提交
2901 2902 2903 2904
        virQEMUCapsPtr qemuCaps = NULL;
        struct addrinfo *info = NULL;
        struct addrinfo hints = { .ai_flags = AI_ADDRCONFIG,
                                  .ai_socktype = SOCK_STREAM };
2905
        const char *incFormat;
J
Ján Tomko 已提交
2906

2907 2908 2909 2910
        if (getaddrinfo("::", NULL, &hints, &info) == 0) {
            freeaddrinfo(info);
            hostIPv6Capable = true;
        }
J
Ján Tomko 已提交
2911
        if (!(qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
2912
                                                    (*def)->emulator)))
J
Ján Tomko 已提交
2913 2914
            goto cleanup;

2915 2916 2917 2918
        qemuIPv6Capable = virQEMUCapsGet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
        virObjectUnref(qemuCaps);

        if (listenAddress) {
2919 2920 2921 2922 2923 2924 2925 2926 2927
            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"));
2928 2929
                    goto cleanup;
                }
2930 2931
                /* IPv6 address must be escaped in brackets on the cmd line */
                encloseAddress = true;
2932
            } else {
2933
                /* listenAddress is a hostname or IPv4 */
2934
            }
2935
        } else if (qemuIPv6Capable && hostIPv6Capable) {
2936 2937 2938
            /* Listen on :: instead of 0.0.0.0 if QEMU understands it
             * and there is at least one IPv6 address configured
             */
2939 2940 2941 2942
            listenAddress = "::";
            encloseAddress = true;
        } else {
            listenAddress = "0.0.0.0";
J
Ján Tomko 已提交
2943 2944
        }

2945 2946 2947 2948
        /* 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
2949
         */
2950 2951 2952 2953 2954 2955
        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
2956 2957 2958
            goto cleanup;
    }

2959
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2960
                                   driver->xmlopt,
2961 2962 2963
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2964
        goto cleanup;
2965

2966
    virObjectRef(vm);
2967
    *def = NULL;
2968
    priv = vm->privateData;
2969 2970
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2971

2972 2973 2974 2975 2976
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2977
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
2978 2979
                                       QEMU_MIGRATION_COOKIE_LOCKSTATE |
                                       QEMU_MIGRATION_COOKIE_NBD)))
2980 2981
        goto cleanup;

2982
    if (STREQ_NULLABLE(protocol, "rdma") && !vm->def->mem.hard_limit) {
M
Michael R. Hines 已提交
2983 2984 2985 2986 2987 2988
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

2989 2990 2991
    if (qemuMigrationPrecreateStorage(dconn, driver, vm, mig->nbd) < 0)
        goto cleanup;

2992
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2993
        goto cleanup;
2994
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2995 2996 2997 2998

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

L
liguang 已提交
2999 3000 3001
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

3002 3003
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
3004 3005
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
3006 3007 3008 3009
        goto endjob;
    }

    /* Start the QEMU daemon, with the same command-line arguments plus
3010
     * -incoming $migrateFrom
3011
     */
3012 3013
    if (qemuProcessStart(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                         migrateFrom, dataFD[0], NULL, NULL,
3014 3015
                         VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
                         VIR_QEMU_PROCESS_START_PAUSED |
3016
                         VIR_QEMU_PROCESS_START_AUTODESTROY) < 0) {
3017
        virDomainAuditStart(vm, "migrated", false);
3018 3019 3020
        goto endjob;
    }

3021 3022 3023 3024
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
3025
            goto stop;
3026
        }
3027
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
3028 3029
    }

3030 3031
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
3032 3033 3034
                                    QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stop;

3035
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
3036 3037 3038 3039
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
        goto stop;
    }

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

3045 3046 3047 3048 3049 3050 3051 3052
    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");
    }
3053

3054
 done:
L
liguang 已提交
3055 3056 3057 3058 3059
    if (flags & VIR_MIGRATE_OFFLINE)
        cookieFlags = 0;
    else
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS;

3060 3061 3062
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
3063
        if (qemuMigrationStartNBDServer(driver, vm, listenAddress) < 0) {
3064 3065
            /* error already reported */
            goto endjob;
3066
        }
3067
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
3068 3069 3070 3071
    }

    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
3072 3073 3074 3075 3076 3077 3078
        /* 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");
    }

3079 3080 3081
    if (qemuDomainCleanupAdd(vm, qemuMigrationPrepareCleanup) < 0)
        goto endjob;

L
liguang 已提交
3082 3083
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
3084
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
3085 3086 3087
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
3088

3089 3090 3091 3092
    /* 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.
     */
3093
    qemuMigrationJobContinue(vm);
3094

3095 3096
    if (autoPort)
        priv->migrationPort = port;
3097
    ret = 0;
3098

3099
 cleanup:
J
Ján Tomko 已提交
3100
    VIR_FREE(migrateFrom);
J
Jiri Denemark 已提交
3101
    VIR_FREE(xmlout);
3102 3103
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
3104 3105 3106
    if (ret < 0 && priv) {
        /* priv is set right after vm is added to the list of domains
         * and there is no 'goto cleanup;' in the middle of those */
3107 3108 3109
        virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
        priv->nbdPort = 0;
        qemuDomainRemoveInactive(driver, vm);
3110
    }
3111
    qemuDomObjEndAPI(&vm);
3112 3113
    if (event)
        qemuDomainEventQueue(driver, event);
3114
    qemuMigrationCookieFree(mig);
3115
    virObjectUnref(caps);
3116
    virNWFilterUnlockFilterUpdates();
3117
    return ret;
3118

3119
 stop:
3120 3121 3122
    virDomainAuditStart(vm, "migrated", false);
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, 0);

3123
 endjob:
3124
    qemuMigrationJobFinish(driver, vm);
3125
    goto cleanup;
3126 3127 3128
}


3129 3130 3131 3132 3133
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
3134
qemuMigrationPrepareTunnel(virQEMUDriverPtr driver,
3135 3136 3137 3138 3139 3140
                           virConnectPtr dconn,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virStreamPtr st,
3141
                           virDomainDefPtr *def,
3142
                           const char *origname,
L
liguang 已提交
3143
                           unsigned long flags)
3144 3145 3146 3147
{
    int ret;

    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
3148
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
3149
              "origname=%s, flags=%lx",
3150
              driver, dconn, NULLSTR(cookiein), cookieinlen,
3151
              cookieout, cookieoutlen, st, *def, origname, flags);
3152

3153 3154 3155 3156 3157 3158
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

3159
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3160
                                  cookieout, cookieoutlen, def, origname,
3161
                                  st, NULL, 0, false, NULL, flags);
3162 3163 3164 3165
    return ret;
}


3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188
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;
}


3189
int
3190
qemuMigrationPrepareDirect(virQEMUDriverPtr driver,
3191
                           virConnectPtr dconn,
3192 3193 3194 3195
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
3196 3197
                           const char *uri_in,
                           char **uri_out,
3198
                           virDomainDefPtr *def,
3199
                           const char *origname,
3200
                           const char *listenAddress,
L
liguang 已提交
3201
                           unsigned long flags)
3202
{
3203 3204
    unsigned short port = 0;
    bool autoPort = true;
3205 3206
    char *hostname = NULL;
    int ret = -1;
3207
    virURIPtr uri = NULL;
3208 3209
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
3210

3211 3212
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
3213
              "def=%p, origname=%s, listenAddress=%s, flags=%lx",
3214 3215
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
3216
              *def, origname, NULLSTR(listenAddress), flags);
3217

3218 3219
    *uri_out = NULL;

3220 3221 3222
    /* 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
3223 3224 3225
     * 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".
3226 3227 3228 3229 3230 3231
     *
     * 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) {
3232 3233 3234
        bool encloseAddress = false;
        const char *incFormat;

3235
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3236
            goto cleanup;
3237

3238
        if (migrateHost != NULL) {
3239 3240
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
3241

3242
            if (VIR_STRDUP(hostname, migrateHost) < 0)
3243 3244 3245 3246 3247
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
3248 3249

        if (STRPREFIX(hostname, "localhost")) {
3250 3251 3252
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
3253 3254 3255 3256 3257
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
3258
         * compatibility with old targets. We at least make the
3259 3260
         * new targets accept both syntaxes though.
         */
3261 3262 3263 3264 3265 3266
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
3267 3268
            goto cleanup;
    } else {
3269
        bool well_formed_uri;
J
Ján Tomko 已提交
3270

3271 3272
        if (!(uri = qemuMigrationParseURI(uri_in, &well_formed_uri)))
            goto cleanup;
J
Ján Tomko 已提交
3273

M
Michael R. Hines 已提交
3274 3275
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
3276 3277 3278
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
3279 3280 3281 3282 3283 3284 3285 3286 3287 3288
            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) {
3289
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3290
                goto cleanup;
3291

3292
            /* Send well-formed URI only if uri_in was well-formed */
3293 3294 3295 3296 3297 3298 3299 3300
            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;
            }
3301
        } else {
3302 3303
            port = uri->port;
            autoPort = false;
3304 3305 3306 3307 3308 3309
        }
    }

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

3310
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3311
                                  cookieout, cookieoutlen, def, origname,
3312 3313
                                  NULL, uri ? uri->scheme : "tcp",
                                  port, autoPort, listenAddress, flags);
3314
 cleanup:
3315
    virURIFree(uri);
3316
    VIR_FREE(hostname);
3317
    virObjectUnref(cfg);
3318
    if (ret != 0) {
3319
        VIR_FREE(*uri_out);
3320 3321 3322
        if (autoPort)
            virPortAllocatorRelease(driver->migrationPorts, port);
    }
3323 3324 3325 3326
    return ret;
}


3327 3328 3329
virDomainDefPtr
qemuMigrationPrepareDef(virQEMUDriverPtr driver,
                        const char *dom_xml,
3330 3331
                        const char *dname,
                        char **origname)
3332 3333 3334
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
3335
    char *name = NULL;
3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347

    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,
3348
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE)))
3349 3350 3351
        goto cleanup;

    if (dname) {
3352
        name = def->name;
3353 3354 3355 3356 3357 3358
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

3359
 cleanup:
3360
    virObjectUnref(caps);
3361 3362 3363 3364
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
3365 3366 3367 3368
    return def;
}


3369 3370 3371 3372 3373 3374 3375 3376 3377 3378
static int
qemuMigrationConfirmPhase(virQEMUDriverPtr driver,
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *cookiein,
                          int cookieinlen,
                          unsigned int flags,
                          int retcode)
{
    qemuMigrationCookiePtr mig;
3379
    virObjectEventPtr event = NULL;
3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394
    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);

3395 3396
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                       QEMU_MIGRATION_COOKIE_STATS)))
3397 3398
        goto cleanup;

3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415
    /* 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;
        }
    }

3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430
    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");

3431
        event = virDomainEventLifecycleNewFromObj(vm,
3432 3433 3434 3435 3436 3437 3438
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
    } else {

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

3439 3440 3441 3442
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
3443 3444 3445 3446 3447 3448 3449 3450
        }

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

3451
 done:
3452 3453 3454
    qemuMigrationCookieFree(mig);
    rv = 0;

3455
 cleanup:
3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470
    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;
3471
    qemuMigrationJobPhase phase;
3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485
    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);
3486 3487
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
3488 3489 3490 3491 3492

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

3493 3494 3495
    qemuMigrationJobFinish(driver, vm);
    if (!virDomainObjIsActive(vm) &&
        (!vm->persistent || (flags & VIR_MIGRATE_UNDEFINE_SOURCE))) {
3496 3497 3498 3499 3500
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
        qemuDomainRemoveInactive(driver, vm);
    }

3501
 cleanup:
3502
    qemuDomObjEndAPI(&vm);
3503 3504 3505 3506 3507
    virObjectUnref(cfg);
    return ret;
}


3508 3509
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3510
    MIGRATION_DEST_CONNECT_HOST,
3511
    MIGRATION_DEST_UNIX,
3512
    MIGRATION_DEST_FD,
3513
};
3514

3515 3516 3517 3518
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3519

3520 3521 3522 3523 3524 3525
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3526
            const char *protocol;
3527 3528 3529 3530 3531
            const char *name;
            int port;
        } host;

        struct {
3532
            char *file;
3533 3534
            int sock;
        } unix_socket;
3535 3536 3537 3538 3539

        struct {
            int qemu;
            int local;
        } fd;
3540 3541 3542 3543 3544 3545 3546
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3547 3548 3549

#define TUNNEL_SEND_BUF_SIZE 65536

3550 3551 3552 3553 3554 3555 3556
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3557 3558
    int wakeupRecvFD;
    int wakeupSendFD;
3559 3560 3561
};

static void qemuMigrationIOFunc(void *arg)
3562
{
3563
    qemuMigrationIOThreadPtr data = arg;
3564 3565 3566 3567 3568 3569 3570
    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);
3571

3572
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3573
        goto abrt;
3574

3575 3576 3577
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3578
    for (;;) {
3579 3580 3581 3582 3583 3584 3585 3586 3587 3588
        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;
3589
            virReportSystemError(errno, "%s",
3590 3591
                                 _("poll failed in migration tunnel"));
            goto abrt;
3592
        }
3593 3594 3595 3596 3597 3598 3599

        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");
3600
            break;
3601
        }
3602

3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618
        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;
            }
3619 3620
        }

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

3639 3640
    if (virStreamFinish(data->st) < 0)
        goto error;
3641

3642 3643
    VIR_FREE(buffer);

3644 3645
    return;

3646
 abrt:
3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3658
 error:
3659 3660
    virCopyLastError(&data->err);
    virResetLastError();
3661
    VIR_FREE(buffer);
3662 3663 3664 3665 3666 3667 3668
}


static qemuMigrationIOThreadPtr
qemuMigrationStartTunnel(virStreamPtr st,
                         int sock)
{
3669 3670
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3671

3672 3673 3674 3675
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3676 3677
    }

3678
    if (VIR_ALLOC(io) < 0)
3679
        goto error;
3680

3681 3682
    io->st = st;
    io->sock = sock;
3683 3684
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3685 3686 3687 3688 3689 3690

    if (virThreadCreate(&io->thread, true,
                        qemuMigrationIOFunc,
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3691
        goto error;
3692 3693 3694
    }

    return io;
3695

3696
 error:
3697 3698 3699 3700
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3701 3702 3703
}

static int
3704
qemuMigrationStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3705 3706
{
    int rv = -1;
3707 3708 3709 3710 3711 3712 3713 3714 3715
    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;
    }

3716 3717 3718 3719
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3720 3721 3722 3723
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3724 3725 3726 3727 3728 3729
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3730
 cleanup:
3731 3732
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3733 3734
    VIR_FREE(io);
    return rv;
3735 3736
}

3737
static int
3738
qemuMigrationConnect(virQEMUDriverPtr driver,
3739 3740 3741 3742 3743 3744 3745 3746 3747
                     virDomainObjPtr vm,
                     qemuMigrationSpecPtr spec)
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3748
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3749 3750 3751 3752 3753 3754 3755 3756 3757
        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);
3758
        virObjectUnref(sock);
3759 3760 3761 3762 3763
    }
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0 ||
        spec->dest.fd.qemu == -1)
        goto cleanup;

3764 3765 3766 3767 3768 3769 3770
    /* 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;
    }

3771 3772
    ret = 0;

3773
 cleanup:
3774 3775 3776 3777 3778 3779
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3780
static int
3781
qemuMigrationRun(virQEMUDriverPtr driver,
3782 3783 3784 3785 3786 3787 3788
                 virDomainObjPtr vm,
                 const char *cookiein,
                 int cookieinlen,
                 char **cookieout,
                 int *cookieoutlen,
                 unsigned long flags,
                 unsigned long resource,
3789
                 qemuMigrationSpecPtr spec,
3790 3791
                 virConnectPtr dconn,
                 const char *graphicsuri)
3792
{
3793
    int ret = -1;
3794 3795
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3796
    qemuMigrationCookiePtr mig = NULL;
3797
    qemuMigrationIOThreadPtr iothread = NULL;
3798
    int fd = -1;
3799
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3800
    virErrorPtr orig_err = NULL;
3801
    unsigned int cookieFlags = 0;
3802
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3803
    int rc;
3804 3805 3806

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
3807
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s",
3808 3809
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3810 3811
              spec, spec->destType, spec->fwdType, dconn,
              NULLSTR(graphicsuri));
3812

3813 3814 3815 3816 3817 3818 3819 3820 3821 3822
    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;
    }

3823 3824
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3825 3826 3827 3828
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3829 3830 3831
        return -1;
    }

3832 3833 3834
    mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                 cookieFlags | QEMU_MIGRATION_COOKIE_GRAPHICS);
    if (!mig)
3835 3836
        goto cleanup;

3837
    if (qemuDomainMigrateGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3838 3839
        VIR_WARN("unable to provide data for graphics client relocation");

3840 3841 3842 3843 3844 3845 3846
    /* 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;
    }

3847
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3848 3849 3850 3851 3852 3853
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

3854 3855
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
3856 3857 3858
                                    QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3859 3860
    if (qemuMigrationSetAutoConverge(driver, vm,
                                     flags & VIR_MIGRATE_AUTO_CONVERGE,
3861 3862 3863
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3864 3865
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
3866 3867 3868
                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3869 3870
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3871 3872
        goto cleanup;

3873 3874 3875 3876
    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 */
3877
        qemuDomainObjExitMonitor(driver, vm);
3878
        priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
3879 3880 3881 3882 3883 3884
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
        goto cleanup;
    }

3885
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0) {
3886
        qemuDomainObjExitMonitor(driver, vm);
3887 3888
        goto cleanup;
    }
3889

3890 3891
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3892
        qemuMigrationConnect(driver, vm, spec) < 0) {
3893
        qemuDomainObjExitMonitor(driver, vm);
3894
        goto cleanup;
3895
    }
3896

3897 3898
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
3899 3900 3901 3902 3903
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
            qemuDomainObjExitMonitor(driver, vm);
            goto cleanup;
        }
3904
        ret = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
3905
                                       spec->dest.host.protocol,
3906 3907 3908 3909
                                       spec->dest.host.name,
                                       spec->dest.host.port);
        break;

3910 3911 3912 3913
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3914
    case MIGRATION_DEST_UNIX:
3915
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX)) {
3916 3917 3918 3919 3920 3921 3922 3923 3924
            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;
3925 3926

    case MIGRATION_DEST_FD:
3927
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3928
            fd = spec->dest.fd.local;
3929 3930
            spec->dest.fd.local = -1;
        }
3931 3932 3933 3934
        ret = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                     spec->dest.fd.qemu);
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3935
    }
3936
    qemuDomainObjExitMonitor(driver, vm);
3937
    if (ret < 0)
3938 3939
        goto cleanup;
    ret = -1;
3940 3941

    if (!virDomainObjIsActive(vm)) {
3942 3943
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
3944 3945 3946 3947 3948 3949
        goto cleanup;
    }

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

3950 3951 3952 3953 3954 3955 3956 3957
    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;
3958

3959 3960 3961 3962 3963 3964 3965
        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;
        }
3966 3967
    }

3968 3969
    if (spec->fwdType != MIGRATION_FWD_DIRECT &&
        !(iothread = qemuMigrationStartTunnel(spec->fwd.stream, fd)))
3970 3971
        goto cancel;

3972 3973 3974 3975 3976 3977
    rc = qemuMigrationWaitForCompletion(driver, vm,
                                        QEMU_ASYNC_JOB_MIGRATION_OUT,
                                        dconn, abort_on_error);
    if (rc == -2)
        goto cancel;
    else if (rc == -1)
3978
        goto cleanup;
3979

3980 3981 3982 3983 3984 3985 3986 3987 3988
    /* 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) {
3989 3990
        if (qemuMigrationSetOffline(driver, vm) < 0) {
            priv->job.current->type = VIR_DOMAIN_JOB_FAILED;
3991
            goto cleanup;
3992
        }
3993 3994
    }

3995
    ret = 0;
3996

3997
 cleanup:
3998 3999 4000
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

4001
    /* cancel any outstanding NBD jobs */
4002 4003
    if (mig)
        qemuMigrationCancelDriveMirror(mig, driver, vm);
4004

4005
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
4006
        if (iothread && qemuMigrationStopTunnel(iothread, ret < 0) < 0)
4007
            ret = -1;
4008
        VIR_FORCE_CLOSE(fd);
4009
    }
4010

4011
    if (priv->job.completed) {
4012
        qemuDomainJobInfoUpdateTime(priv->job.completed);
4013 4014
        qemuDomainJobInfoUpdateDowntime(priv->job.completed);
    }
4015

4016 4017 4018
    if (priv->job.current->type == VIR_DOMAIN_JOB_UNBOUNDED)
        priv->job.current->type = VIR_DOMAIN_JOB_FAILED;

4019 4020
    cookieFlags |= QEMU_MIGRATION_COOKIE_NETWORK |
                   QEMU_MIGRATION_COOKIE_STATS;
4021 4022
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookieFlags |= QEMU_MIGRATION_COOKIE_PERSISTENT;
4023
    if (ret == 0 &&
4024 4025
        qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
4026
        VIR_WARN("Unable to encode migration cookie");
4027
    }
4028

4029 4030
    qemuMigrationCookieFree(mig);

4031 4032 4033 4034 4035
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4036 4037
    return ret;

4038
 cancel:
4039 4040
    orig_err = virSaveLastError();

4041
    if (virDomainObjIsActive(vm)) {
4042 4043
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
4044
            qemuMonitorMigrateCancel(priv->mon);
4045
            qemuDomainObjExitMonitor(driver, vm);
4046
        }
4047
    }
4048 4049 4050
    goto cleanup;
}

4051
/* Perform migration using QEMU's native migrate support,
4052 4053
 * not encrypted obviously
 */
4054
static int doNativeMigrate(virQEMUDriverPtr driver,
4055 4056 4057 4058 4059 4060 4061
                           virDomainObjPtr vm,
                           const char *uri,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4062
                           unsigned long resource,
4063 4064
                           virConnectPtr dconn,
                           const char *graphicsuri)
4065
{
4066
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
4067
    virURIPtr uribits = NULL;
4068
    int ret = -1;
4069 4070 4071
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
4072 4073
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "graphicsuri=%s",
4074
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
4075 4076
              cookieout, cookieoutlen, flags, resource,
              NULLSTR(graphicsuri));
4077

4078
    if (!(uribits = qemuMigrationParseURI(uri, NULL)))
4079 4080
        return -1;

M
Michael R. Hines 已提交
4081 4082 4083 4084 4085
    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"));
4086
            goto cleanup;
M
Michael R. Hines 已提交
4087 4088 4089 4090 4091
        }
        if (!vm->def->mem.hard_limit) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
4092
            goto cleanup;
M
Michael R. Hines 已提交
4093 4094 4095 4096 4097
        }
    }

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
        STRNEQ(uribits->scheme, "rdma"))
4098 4099
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
4100
        spec.destType = MIGRATION_DEST_HOST;
4101
    spec.dest.host.protocol = uribits->scheme;
4102 4103 4104
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
4105

4106
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4107 4108
                           cookieoutlen, flags, resource, &spec, dconn,
                           graphicsuri);
4109 4110 4111 4112

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

4113
 cleanup:
4114
    virURIFree(uribits);
4115 4116 4117 4118 4119

    return ret;
}


4120
static int doTunnelMigrate(virQEMUDriverPtr driver,
4121 4122 4123 4124 4125 4126 4127
                           virDomainObjPtr vm,
                           virStreamPtr st,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4128
                           unsigned long resource,
4129 4130
                           virConnectPtr dconn,
                           const char *graphicsuri)
4131 4132
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
4133
    virNetSocketPtr sock = NULL;
4134 4135
    int ret = -1;
    qemuMigrationSpec spec;
4136
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4137 4138

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
4139 4140
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "graphicsuri=%s",
4141
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
4142 4143
              cookieout, cookieoutlen, flags, resource,
              NULLSTR(graphicsuri));
4144

4145 4146 4147
    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)) {
4148 4149
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Source qemu is too old to support tunnelled migration"));
4150
        virObjectUnref(cfg);
4151
        return -1;
4152 4153 4154 4155 4156
    }

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

4157
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD)) {
4158 4159 4160 4161 4162 4163
        int fds[2];

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

4164
        if (pipe2(fds, O_CLOEXEC) == 0) {
4165 4166 4167 4168
            spec.dest.fd.qemu = fds[1];
            spec.dest.fd.local = fds[0];
        }
        if (spec.dest.fd.qemu == -1 ||
4169
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
4170
                                              spec.dest.fd.qemu) < 0) {
4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181
            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",
4182
                        cfg->libDir, vm->def->name) < 0)
4183 4184 4185
            goto cleanup;

        if (virNetSocketNewListenUNIX(spec.dest.unix_socket.file, 0700,
4186
                                      cfg->user, cfg->group,
4187 4188 4189 4190 4191 4192 4193
                                      &sock) < 0 ||
            virNetSocketListen(sock, 1) < 0)
            goto cleanup;

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

4194
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4195 4196
                           cookieoutlen, flags, resource, &spec, dconn,
                           graphicsuri);
4197

4198
 cleanup:
4199 4200 4201 4202
    if (spec.destType == MIGRATION_DEST_FD) {
        VIR_FORCE_CLOSE(spec.dest.fd.qemu);
        VIR_FORCE_CLOSE(spec.dest.fd.local);
    } else {
4203
        virObjectUnref(sock);
4204 4205
        VIR_FREE(spec.dest.unix_socket.file);
    }
4206

4207
    virObjectUnref(cfg);
4208 4209 4210 4211
    return ret;
}


4212 4213 4214 4215
/* 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 */
4216
static int doPeer2PeerMigrate2(virQEMUDriverPtr driver,
4217
                               virConnectPtr sconn ATTRIBUTE_UNUSED,
4218 4219
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
4220
                               const char *dconnuri,
4221 4222 4223
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
4224 4225 4226
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
4227
    char *cookie = NULL;
4228 4229 4230
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
4231
    bool cancelled;
4232
    virStreamPtr st = NULL;
4233
    unsigned long destflags;
4234

4235
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
4236
              "flags=%lx, dname=%s, resource=%lu",
4237 4238
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
4239

4240 4241 4242 4243 4244
    /* 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,
4245 4246
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
4247 4248 4249 4250 4251
        return -1;

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

4252 4253
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4254

4255 4256 4257 4258 4259 4260 4261 4262 4263 4264
    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;

4265
        qemuDomainObjEnterRemote(vm);
4266
        ret = dconn->driver->domainMigratePrepareTunnel
4267
            (dconn, st, destflags, dname, resource, dom_xml);
4268
        qemuDomainObjExitRemote(vm);
4269
    } else {
4270
        qemuDomainObjEnterRemote(vm);
4271 4272
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
4273
             destflags, dname, resource, dom_xml);
4274
        qemuDomainObjExitRemote(vm);
4275 4276 4277
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
4278 4279 4280
        goto cleanup;

    /* the domain may have shutdown or crashed while we had the locks dropped
4281
     * in qemuDomainObjEnterRemote, so check again
4282 4283
     */
    if (!virDomainObjIsActive(vm)) {
4284 4285
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4286 4287 4288
        goto cleanup;
    }

4289 4290
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
4291 4292
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
4293
        cancelled = true;
4294
        orig_err = virSaveLastError();
4295
        goto finish;
4296 4297
    }

4298 4299 4300 4301
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
4302
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4303
    if (flags & VIR_MIGRATE_TUNNELLED)
4304 4305
        ret = doTunnelMigrate(driver, vm, st,
                              NULL, 0, NULL, NULL,
4306
                              flags, resource, dconn, NULL);
4307 4308 4309 4310
    else
        ret = doNativeMigrate(driver, vm, uri_out,
                              cookie, cookielen,
                              NULL, NULL, /* No out cookie with v2 migration */
4311
                              flags, resource, dconn, NULL);
4312 4313 4314 4315

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4316

4317 4318 4319
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4320
    cancelled = ret < 0;
4321

4322
 finish:
4323 4324 4325 4326
    /* 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.
     */
4327
    dname = dname ? dname : vm->def->name;
4328
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4329
    qemuDomainObjEnterRemote(vm);
4330
    ddomain = dconn->driver->domainMigrateFinish2
4331
        (dconn, dname, cookie, cookielen,
4332
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4333
    qemuDomainObjExitRemote(vm);
4334 4335
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4336

4337
 cleanup:
4338
    if (ddomain) {
4339
        virObjectUnref(ddomain);
4340 4341 4342 4343
        ret = 0;
    } else {
        ret = -1;
    }
4344

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

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4352
    VIR_FREE(cookie);
4353 4354

    return ret;
4355 4356 4357
}


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

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

4400 4401 4402 4403 4404
    /* 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.  */

4405 4406
    dom_xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
                                      &cookieout, &cookieoutlen, flags);
4407 4408 4409
    if (!dom_xml)
        goto cleanup;

4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429
    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;
4430 4431 4432 4433 4434 4435

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4436 4437 4438 4439 4440
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4441 4442
    }

4443 4444 4445
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4446 4447
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4448

4449 4450 4451 4452 4453 4454 4455 4456 4457
    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;

4458
        qemuDomainObjEnterRemote(vm);
4459 4460 4461 4462 4463 4464 4465 4466 4467
        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);
        }
4468
        qemuDomainObjExitRemote(vm);
4469
    } else {
4470
        qemuDomainObjEnterRemote(vm);
4471 4472 4473 4474 4475 4476 4477 4478 4479
        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);
        }
4480
        qemuDomainObjExitRemote(vm);
4481 4482 4483 4484 4485
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

L
liguang 已提交
4486 4487 4488 4489
    if (flags & VIR_MIGRATE_OFFLINE) {
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4490
        cancelled = false;
L
liguang 已提交
4491 4492 4493
        goto finish;
    }

4494 4495 4496 4497
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4498 4499
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4500
            goto finish;
4501
        }
4502
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4503 4504
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4505
        orig_err = virSaveLastError();
4506 4507 4508 4509 4510 4511 4512 4513
        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.
     */
4514
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4515
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4516 4517 4518 4519 4520
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4521
    if (flags & VIR_MIGRATE_TUNNELLED) {
4522 4523 4524
        ret = doTunnelMigrate(driver, vm, st,
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4525
                              flags, bandwidth, dconn, graphicsuri);
4526 4527
    } else {
        ret = doNativeMigrate(driver, vm, uri,
4528 4529
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4530
                              flags, bandwidth, dconn, graphicsuri);
4531
    }
4532 4533

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4534
    if (ret < 0) {
4535
        orig_err = virSaveLastError();
4536 4537 4538 4539
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4540 4541 4542 4543

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4544
    cancelled = ret < 0;
4545

4546
 finish:
4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558
    /*
     * 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;
4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581

    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);
    }
4582 4583
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4584

4585 4586 4587 4588 4589 4590 4591
    /* 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.
4592
     */
4593
    cancelled = ddomain == NULL;
4594

4595 4596 4597 4598 4599 4600
    /* 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();

4601 4602 4603 4604
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4605
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4606 4607 4608 4609 4610
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4611 4612 4613
    ret = qemuMigrationConfirmPhase(driver, sconn, vm,
                                    cookiein, cookieinlen,
                                    flags, cancelled);
4614 4615 4616 4617
    /* 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.
     */
4618 4619 4620
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4621 4622 4623

 cleanup:
    if (ddomain) {
4624
        virObjectUnref(ddomain);
4625 4626 4627 4628 4629
        ret = 0;
    } else {
        ret = -1;
    }

4630
    virObjectUnref(st);
4631 4632 4633 4634 4635 4636 4637 4638

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4639
    virTypedParamsFree(params, nparams);
4640 4641 4642 4643
    return ret;
}


4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4656
static int doPeer2PeerMigrate(virQEMUDriverPtr driver,
4657
                              virConnectPtr sconn,
4658
                              virDomainObjPtr vm,
4659
                              const char *xmlin,
4660
                              const char *dconnuri,
4661
                              const char *uri,
4662
                              const char *graphicsuri,
4663
                              const char *listenAddress,
4664 4665
                              unsigned long flags,
                              const char *dname,
4666 4667
                              unsigned long resource,
                              bool *v3proto)
4668 4669 4670 4671
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4672
    virErrorPtr orig_err = NULL;
4673
    bool offline = false;
4674
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4675
    bool useParams;
4676

4677
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4678 4679
              "uri=%s, graphicsuri=%s, listenAddress=%s, flags=%lx, "
              "dname=%s, resource=%lu",
4680
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4681 4682
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
              flags, NULLSTR(dname), resource);
4683 4684 4685 4686 4687

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

4688
    qemuDomainObjEnterRemote(vm);
4689
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4690
    qemuDomainObjExitRemote(vm);
4691
    if (dconn == NULL) {
4692
        virReportError(VIR_ERR_OPERATION_FAILED,
4693 4694
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4695
        virObjectUnref(cfg);
4696 4697 4698
        return -1;
    }

4699 4700
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4701 4702
        goto cleanup;

4703
    qemuDomainObjEnterRemote(vm);
4704 4705
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4706
        /* v3proto reflects whether the caller used Perform3, but with
4707
         * p2p migrate, regardless of whether Perform2 or Perform3
4708 4709 4710 4711
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4712 4713
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
L
liguang 已提交
4714 4715 4716
    if (flags & VIR_MIGRATE_OFFLINE)
        offline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                           VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4717
    qemuDomainObjExitRemote(vm);
4718

4719
    if (!p2p) {
4720 4721
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4722 4723 4724
        goto cleanup;
    }

4725 4726
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4727
    if (!useParams && graphicsuri) {
4728 4729 4730 4731 4732 4733
        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
4734 4735 4736 4737 4738 4739 4740
    if (flags & VIR_MIGRATE_OFFLINE && !offline) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4741
    /* domain may have been stopped while we were talking to remote daemon */
L
liguang 已提交
4742
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4743 4744
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4745 4746 4747
        goto cleanup;
    }

4748 4749 4750 4751 4752 4753 4754
    /* 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;

4755 4756
    if (*v3proto) {
        ret = doPeer2PeerMigrate3(driver, sconn, dconn, dconnuri, vm, xmlin,
4757 4758
                                  dname, uri, graphicsuri, listenAddress,
                                  resource, useParams, flags);
4759
    } else {
4760
        ret = doPeer2PeerMigrate2(driver, sconn, dconn, vm,
4761
                                  dconnuri, flags, dname, resource);
4762
    }
4763

4764
 cleanup:
4765
    orig_err = virSaveLastError();
4766
    qemuDomainObjEnterRemote(vm);
4767
    virObjectUnref(dconn);
4768
    qemuDomainObjExitRemote(vm);
4769 4770 4771 4772
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4773
    virObjectUnref(cfg);
4774 4775 4776 4777
    return ret;
}


4778 4779 4780 4781 4782 4783
/*
 * 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
4784
qemuMigrationPerformJob(virQEMUDriverPtr driver,
4785 4786 4787 4788 4789
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dconnuri,
                        const char *uri,
4790
                        const char *graphicsuri,
4791
                        const char *listenAddress,
4792
                        const char *cookiein,
4793 4794 4795 4796 4797 4798 4799
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4800
{
4801
    virObjectEventPtr event = NULL;
4802
    int ret = -1;
4803
    virErrorPtr orig_err = NULL;
4804
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4805
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
4806

4807
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4808 4809
        goto cleanup;

L
liguang 已提交
4810
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4811 4812
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
4813 4814 4815
        goto endjob;
    }

4816
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
4817
        goto endjob;
4818

4819
    if (!(flags & VIR_MIGRATE_UNSAFE) && !qemuMigrationIsSafe(vm->def))
4820
        goto endjob;
4821

4822
    qemuMigrationStoreDomainState(vm);
4823 4824

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4825
        ret = doPeer2PeerMigrate(driver, conn, vm, xmlin,
4826 4827
                                 dconnuri, uri, graphicsuri, listenAddress,
                                 flags, dname, resource, &v3proto);
4828
    } else {
4829 4830 4831
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
        ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                              cookieout, cookieoutlen,
4832
                              flags, resource, NULL, NULL);
4833
    }
4834 4835
    if (ret < 0)
        goto endjob;
4836

4837 4838 4839 4840
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4841
    if (!v3proto) {
4842 4843
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4844
        virDomainAuditStop(vm, "migrated");
4845
        event = virDomainEventLifecycleNewFromObj(vm,
4846 4847
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4848 4849
    }

4850
 endjob:
4851 4852 4853
    if (ret < 0)
        orig_err = virSaveLastError();

4854
    if (qemuMigrationRestoreDomainState(conn, vm)) {
4855
        event = virDomainEventLifecycleNewFromObj(vm,
4856 4857 4858
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
4859

4860 4861 4862 4863
    qemuMigrationJobFinish(driver, vm);
    if (!virDomainObjIsActive(vm) &&
        (!vm->persistent ||
         (ret == 0 && (flags & VIR_MIGRATE_UNDEFINE_SOURCE)))) {
4864
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
4865
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4866
        qemuDomainRemoveInactive(driver, vm);
4867 4868
    }

4869 4870 4871 4872 4873
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4874
 cleanup:
4875
    qemuDomObjEndAPI(&vm);
4876 4877
    if (event)
        qemuDomainEventQueue(driver, event);
4878
    virObjectUnref(cfg);
4879 4880 4881 4882 4883 4884 4885
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4886
qemuMigrationPerformPhase(virQEMUDriverPtr driver,
4887 4888 4889
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *uri,
4890
                          const char *graphicsuri,
4891 4892 4893 4894 4895 4896 4897
                          const char *cookiein,
                          int cookieinlen,
                          char **cookieout,
                          int *cookieoutlen,
                          unsigned long flags,
                          unsigned long resource)
{
4898
    virObjectEventPtr event = NULL;
4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909
    int ret = -1;

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

    qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4910 4911
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
4912 4913 4914

    ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                          cookieout, cookieoutlen,
4915
                          flags, resource, NULL, graphicsuri);
4916

4917 4918 4919 4920 4921
    if (ret < 0) {
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
4922
        }
4923
        goto endjob;
4924
    }
4925 4926 4927

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4928 4929
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                             qemuMigrationCleanup) < 0)
4930 4931
        goto endjob;

4932
 endjob:
4933
    if (ret < 0)
4934
        qemuMigrationJobFinish(driver, vm);
4935
    else
4936 4937
        qemuMigrationJobContinue(vm);
    if (!virDomainObjIsActive(vm) && !vm->persistent)
4938
        qemuDomainRemoveInactive(driver, vm);
4939

4940
 cleanup:
4941
    qemuDomObjEndAPI(&vm);
4942 4943 4944 4945 4946
    if (event)
        qemuDomainEventQueue(driver, event);
    return ret;
}

4947
int
4948
qemuMigrationPerform(virQEMUDriverPtr driver,
4949 4950 4951 4952 4953
                     virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *xmlin,
                     const char *dconnuri,
                     const char *uri,
4954
                     const char *graphicsuri,
4955
                     const char *listenAddress,
4956 4957 4958 4959 4960 4961 4962 4963 4964 4965
                     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, "
4966
              "uri=%s, graphicsuri=%s, listenAddress=%s"
4967 4968
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
              "flags=%lx, dname=%s, resource=%lu, v3proto=%d",
4969
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4970
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4971 4972
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4973 4974 4975

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4976 4977
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4978 4979 4980 4981
            return -1;
        }

        return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri, uri,
4982 4983
                                       graphicsuri, listenAddress,
                                       cookiein, cookieinlen,
4984 4985
                                       cookieout, cookieoutlen,
                                       flags, dname, resource, v3proto);
4986 4987
    } else {
        if (dconnuri) {
4988 4989
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
4990 4991 4992 4993 4994
            return -1;
        }

        if (v3proto) {
            return qemuMigrationPerformPhase(driver, conn, vm, uri,
4995
                                             graphicsuri,
4996 4997
                                             cookiein, cookieinlen,
                                             cookieout, cookieoutlen,
4998
                                             flags, resource);
4999 5000
        } else {
            return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri,
5001
                                           uri, graphicsuri, listenAddress,
5002
                                           cookiein, cookieinlen,
5003 5004 5005 5006 5007
                                           cookieout, cookieoutlen, flags,
                                           dname, resource, v3proto);
        }
    }
}
5008

5009
static int
5010 5011
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def)
{
5012
    size_t i;
5013 5014 5015 5016 5017
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
5018
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5019
            if (virNetDevVPortProfileAssociate(net->ifname,
5020
                                               virDomainNetGetActualVirtPortProfile(net),
5021
                                               &net->mac,
5022
                                               virDomainNetGetActualDirectDev(net),
5023
                                               -1,
5024
                                               def->uuid,
5025 5026
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
5027 5028 5029
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
5030
                goto err_exit;
5031 5032
            }
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
5033

5034
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
5035 5036 5037 5038
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
5039 5040 5041 5042
        }
        last_good_net = i;
    }

5043
    return 0;
5044

5045
 err_exit:
5046
    for (i = 0; last_good_net != -1 && i < last_good_net; i++) {
5047
        net = def->nets[i];
5048
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5049
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
5050
                                                           virDomainNetGetActualVirtPortProfile(net),
5051
                                                           &net->mac,
5052
                                                           virDomainNetGetActualDirectDev(net),
5053
                                                           -1,
5054
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
5055 5056
        }
    }
5057
    return -1;
5058 5059 5060 5061
}


virDomainPtr
5062
qemuMigrationFinish(virQEMUDriverPtr driver,
5063 5064
                    virConnectPtr dconn,
                    virDomainObjPtr vm,
5065 5066 5067 5068
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
5069
                    unsigned long flags,
5070 5071
                    int retcode,
                    bool v3proto)
5072 5073
{
    virDomainPtr dom = NULL;
5074
    virObjectEventPtr event = NULL;
5075
    bool newVM = true;
5076
    qemuMigrationCookiePtr mig = NULL;
5077
    virErrorPtr orig_err = NULL;
5078
    int cookie_flags = 0;
J
Jiri Denemark 已提交
5079
    qemuDomainObjPrivatePtr priv = vm->privateData;
5080
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5081
    virCapsPtr caps = NULL;
5082
    unsigned short port;
5083

5084
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
5085
              "cookieout=%p, cookieoutlen=%p, flags=%lx, retcode=%d",
5086 5087
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
5088

5089 5090 5091
    port = priv->migrationPort;
    priv->migrationPort = 0;

5092 5093 5094
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

5095
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
5096 5097
        goto cleanup;

5098 5099 5100
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
5101

5102
    qemuDomainCleanupRemove(vm, qemuMigrationPrepareCleanup);
5103
    VIR_FREE(priv->job.completed);
5104

5105
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
5106 5107
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
5108 5109 5110 5111 5112
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookie_flags |= QEMU_MIGRATION_COOKIE_PERSISTENT;

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein,
                                       cookieinlen, cookie_flags)))
5113
        goto endjob;
5114 5115 5116 5117 5118

    /* 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
5119
        if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
5120 5121
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("guest unexpectedly quit"));
5122
            goto endjob;
5123 5124
        }

5125 5126 5127 5128 5129
        if (mig->jobInfo) {
            priv->job.completed = mig->jobInfo;
            mig->jobInfo = NULL;
        }

L
liguang 已提交
5130 5131 5132 5133 5134
        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");
5135
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5136 5137 5138 5139 5140 5141 5142
                                                 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");
5143
        }
5144

5145 5146
        qemuMigrationStopNBDServer(driver, vm, mig);

5147
        if (flags & VIR_MIGRATE_PERSIST_DEST) {
5148
            virDomainDefPtr vmdef;
5149
            if (vm->persistent)
5150
                newVM = false;
5151
            vm->persistent = 1;
5152 5153 5154
            if (mig->persistent)
                vm->newDef = vmdef = mig->persistent;
            else
5155
                vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm);
5156
            if (!vmdef || virDomainSaveConfig(cfg->configDir, vmdef) < 0) {
5157 5158 5159 5160 5161 5162 5163 5164 5165
                /* 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.
                 */
5166 5167

                /*
5168 5169
                 * However, in v3 protocol, the source VM is still available
                 * to restart during confirm() step, so we kill it off now.
5170 5171
                 */
                if (v3proto) {
L
liguang 已提交
5172 5173 5174 5175 5176
                    if (!(flags & VIR_MIGRATE_OFFLINE)) {
                        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                        VIR_QEMU_PROCESS_STOP_MIGRATED);
                        virDomainAuditStop(vm, "failed");
                    }
5177 5178
                    if (newVM)
                        vm->persistent = 0;
5179
                }
A
Alex Jia 已提交
5180
                if (!vmdef)
5181 5182
                    virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                                   _("can't get vmdef"));
5183 5184 5185
                goto endjob;
            }

5186
            event = virDomainEventLifecycleNewFromObj(vm,
5187 5188 5189 5190 5191 5192 5193 5194 5195
                                             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
5196
        if (!(flags & VIR_MIGRATE_PAUSED) && !(flags & VIR_MIGRATE_OFFLINE)) {
5197 5198 5199 5200
            /* 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
5201
            if (qemuProcessStartCPUs(driver, vm, dconn,
5202 5203
                                     VIR_DOMAIN_RUNNING_MIGRATED,
                                     QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5204
                if (virGetLastError() == NULL)
5205 5206
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   "%s", _("resume operation failed"));
5207 5208 5209 5210 5211
                /* Need to save the current error, in case shutting
                 * down the process overwrites it
                 */
                orig_err = virSaveLastError();

5212 5213 5214 5215 5216 5217 5218 5219 5220
                /*
                 * 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) {
5221 5222
                    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                    VIR_QEMU_PROCESS_STOP_MIGRATED);
5223
                    virDomainAuditStop(vm, "failed");
5224
                    event = virDomainEventLifecycleNewFromObj(vm,
5225 5226 5227
                                                     VIR_DOMAIN_EVENT_STOPPED,
                                                     VIR_DOMAIN_EVENT_STOPPED_FAILED);
                }
5228 5229
                goto endjob;
            }
5230 5231 5232 5233
            if (priv->job.completed) {
                qemuDomainJobInfoUpdateTime(priv->job.completed);
                qemuDomainJobInfoUpdateDowntime(priv->job.completed);
            }
5234 5235
        }

5236
        dom = virGetDomain(dconn, vm->def->name, vm->def->uuid);
5237

L
liguang 已提交
5238
        if (!(flags & VIR_MIGRATE_OFFLINE)) {
5239
            event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5240 5241 5242 5243 5244 5245 5246
                                             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);
5247
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5248 5249 5250
                                                 VIR_DOMAIN_EVENT_SUSPENDED,
                                                 VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);
            }
5251
        }
L
liguang 已提交
5252 5253

        if (virDomainObjIsActive(vm) &&
5254
            virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
5255 5256 5257
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto endjob;
        }
5258 5259 5260

        /* Guest is successfully running, so cancel previous auto destroy */
        qemuProcessAutoDestroyRemove(driver, vm);
L
liguang 已提交
5261
    } else if (!(flags & VIR_MIGRATE_OFFLINE)) {
5262 5263
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
5264
        virDomainAuditStop(vm, "failed");
5265
        event = virDomainEventLifecycleNewFromObj(vm,
5266 5267 5268 5269
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_FAILED);
    }

5270 5271
    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen,
                                QEMU_MIGRATION_COOKIE_STATS) < 0)
5272 5273
        VIR_WARN("Unable to encode migration cookie");

5274
 endjob:
5275 5276
    qemuMigrationJobFinish(driver, vm);
    if (!vm->persistent && !virDomainObjIsActive(vm))
5277
        qemuDomainRemoveInactive(driver, vm);
5278

5279
 cleanup:
5280
    virPortAllocatorRelease(driver->migrationPorts, port);
5281 5282 5283 5284
    if (priv->mon)
        qemuMonitorSetDomainLog(priv->mon, -1);
    VIR_FREE(priv->origname);
    qemuDomObjEndAPI(&vm);
5285 5286
    if (event)
        qemuDomainEventQueue(driver, event);
5287
    qemuMigrationCookieFree(mig);
5288 5289 5290 5291
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5292
    virObjectUnref(caps);
5293
    virObjectUnref(cfg);
5294 5295
    return dom;
}
5296

5297

5298
/* Helper function called while vm is active.  */
5299
int
5300
qemuMigrationToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
5301 5302
                    int fd, off_t offset, const char *path,
                    const char *compressor,
E
Eric Blake 已提交
5303
                    bool bypassSecurityDriver,
5304
                    qemuDomainAsyncJob asyncJob)
5305 5306 5307
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5308
    int ret = -1;
5309
    bool restoreLabel = false;
5310 5311
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5312
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5313
    char *errbuf = NULL;
5314
    virErrorPtr orig_err = NULL;
5315 5316 5317 5318 5319

    /* 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,
5320 5321
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5322
        qemuDomainObjExitMonitor(driver, vm);
5323
    }
5324

5325 5326 5327 5328 5329 5330 5331
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5332
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5333
        (!compressor || pipe(pipeFD) == 0)) {
5334
        /* All right! We can use fd migration, which means that qemu
5335 5336 5337
         * 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.  */
5338
        if (virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
5339
                                              compressor ? pipeFD[1] : fd) < 0)
5340
            goto cleanup;
5341 5342 5343
        bypassSecurityDriver = true;
    } else {
        /* Phooey - we have to fall back on exec migration, where qemu
E
Eric Blake 已提交
5344 5345
         * has to popen() the file by name, and block devices have to be
         * given cgroup ACL permission.  We might also stumble on
5346 5347
         * a race present in some qemu versions where it does a wait()
         * that botches pclose.  */
5348 5349
        if (virCgroupHasController(priv->cgroup,
                                   VIR_CGROUP_CONTROLLER_DEVICES)) {
5350 5351 5352 5353
            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
5354
                /* path was not a device, no further need for cgroup */
5355
            } else if (rv < 0) {
5356 5357
                goto cleanup;
            }
5358
        }
5359 5360
        if ((!bypassSecurityDriver) &&
            virSecurityManagerSetSavedStateLabel(driver->securityManager,
5361
                                                 vm->def, path) < 0)
5362
            goto cleanup;
5363
        restoreLabel = true;
5364 5365
    }

5366
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5367 5368
        goto cleanup;

5369 5370 5371
    if (!compressor) {
        const char *args[] = { "cat", NULL };

5372
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388
            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
        };
5389 5390 5391 5392
        if (pipeFD[0] != -1) {
            cmd = virCommandNewArgs(args);
            virCommandSetInputFD(cmd, pipeFD[0]);
            virCommandSetOutputFD(cmd, &fd);
5393 5394
            virCommandSetErrorBuffer(cmd, &errbuf);
            virCommandDoAsyncIO(cmd);
5395 5396 5397
            if (virSetCloseExec(pipeFD[1]) < 0) {
                virReportSystemError(errno, "%s",
                                     _("Unable to set cloexec flag"));
5398
                qemuDomainObjExitMonitor(driver, vm);
5399 5400 5401
                goto cleanup;
            }
            if (virCommandRunAsync(cmd, NULL) < 0) {
5402
                qemuDomainObjExitMonitor(driver, vm);
5403 5404 5405 5406 5407 5408 5409
                goto cleanup;
            }
            rc = qemuMonitorMigrateToFd(priv->mon,
                                        QEMU_MONITOR_MIGRATE_BACKGROUND,
                                        pipeFD[1]);
            if (VIR_CLOSE(pipeFD[0]) < 0 ||
                VIR_CLOSE(pipeFD[1]) < 0)
5410
                VIR_WARN("failed to close intermediate pipe");
5411 5412 5413 5414 5415
        } else {
            rc = qemuMonitorMigrateToFile(priv->mon,
                                          QEMU_MONITOR_MIGRATE_BACKGROUND,
                                          args, path, offset);
        }
5416
    }
5417
    qemuDomainObjExitMonitor(driver, vm);
5418

5419 5420 5421 5422 5423 5424
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        goto cleanup;
    }

5425 5426 5427
    if (rc < 0)
        goto cleanup;

5428
    rc = qemuMigrationWaitForCompletion(driver, vm, asyncJob, NULL, false);
5429

5430 5431 5432 5433
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5434 5435
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5436 5437 5438 5439
                qemuMonitorMigrateCancel(priv->mon);
                qemuDomainObjExitMonitor(driver, vm);
            }
        }
5440
        goto cleanup;
5441
    }
5442

5443 5444 5445
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5446 5447
    ret = 0;

5448
 cleanup:
5449 5450 5451
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5452
    /* Restore max migration bandwidth */
5453 5454
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5455 5456
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5457
        qemuDomainObjExitMonitor(driver, vm);
5458 5459
    }

5460 5461
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5462 5463 5464 5465 5466
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5467 5468
    if (restoreLabel && (!bypassSecurityDriver) &&
        virSecurityManagerRestoreSavedStateLabel(driver->securityManager,
5469
                                                 vm->def, path) < 0)
5470 5471
        VIR_WARN("failed to restore save state label on %s", path);

5472 5473
    if (virCgroupHasController(priv->cgroup,
                               VIR_CGROUP_CONTROLLER_DEVICES)) {
5474 5475 5476
        int rv = virCgroupDenyDevicePath(priv->cgroup, path,
                                         VIR_CGROUP_DEVICE_RWM);
        virDomainAuditCgroupPath(vm, priv->cgroup, "deny", path, "rwm", rv == 0);
5477
    }
5478 5479 5480 5481 5482 5483

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

5484 5485
    return ret;
}
5486 5487

int
5488
qemuMigrationJobStart(virQEMUDriverPtr driver,
5489
                      virDomainObjPtr vm,
5490
                      qemuDomainAsyncJob job)
5491 5492 5493
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

5494
    if (qemuDomainObjBeginAsyncJob(driver, vm, job) < 0)
5495 5496
        return -1;

5497
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5498
        qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
5499
    } else {
5500 5501 5502
        qemuDomainObjSetAsyncJobMask(vm, (QEMU_JOB_DEFAULT_MASK |
                                          JOB_MASK(QEMU_JOB_SUSPEND) |
                                          JOB_MASK(QEMU_JOB_MIGRATION_OP)));
5503
    }
5504

J
Jiri Denemark 已提交
5505
    priv->job.current->type = VIR_DOMAIN_JOB_UNBOUNDED;
5506 5507 5508 5509 5510

    return 0;
}

void
5511
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5512
                         virDomainObjPtr vm,
5513
                         qemuMigrationJobPhase phase)
5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527
{
    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
5528
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5529
                           virDomainObjPtr vm,
5530
                           qemuMigrationJobPhase phase)
5531 5532 5533 5534
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5535
void
5536 5537
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5538
    qemuDomainObjReleaseAsyncJob(vm);
5539 5540 5541 5542
}

bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
5543
                         qemuDomainAsyncJob job)
5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554
{
    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");

5555
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5556 5557 5558 5559 5560
        return false;
    }
    return true;
}

5561
void
5562
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
5563
{
5564
    qemuDomainObjEndAsyncJob(driver, vm);
5565
}