qemu_migration.c 197.4 KB
Newer Older
1 2 3
/*
 * qemu_migration.c: QEMU migration handling
 *
4
 * Copyright (C) 2006-2015 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23
 *
 */

#include <config.h>

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

#include "qemu_migration.h"
#include "qemu_monitor.h"
#include "qemu_domain.h"
#include "qemu_process.h"
#include "qemu_capabilities.h"
39
#include "qemu_command.h"
40
#include "qemu_cgroup.h"
41
#include "qemu_hotplug.h"
42
#include "qemu_blockjob.h"
43

44
#include "domain_audit.h"
45
#include "virlog.h"
46
#include "virerror.h"
47
#include "viralloc.h"
E
Eric Blake 已提交
48
#include "virfile.h"
49
#include "virnetdevopenvswitch.h"
50 51
#include "datatypes.h"
#include "fdstream.h"
52
#include "viruuid.h"
53
#include "virtime.h"
54
#include "locking/domain_lock.h"
55
#include "rpc/virnetsocket.h"
56
#include "virstoragefile.h"
M
Martin Kletzander 已提交
57
#include "viruri.h"
58
#include "virhook.h"
59
#include "virstring.h"
60
#include "virtypedparam.h"
M
Michael R. Hines 已提交
61
#include "virprocess.h"
62
#include "nwfilter_conf.h"
63
#include "storage/storage_driver.h"
64 65 66

#define VIR_FROM_THIS VIR_FROM_QEMU

67 68
VIR_LOG_INIT("qemu.qemu_migration");

69 70 71 72 73 74 75 76 77 78 79 80 81
VIR_ENUM_IMPL(qemuMigrationJobPhase, QEMU_MIGRATION_PHASE_LAST,
              "none",
              "perform2",
              "begin3",
              "perform3",
              "perform3_done",
              "confirm3_cancelled",
              "confirm3",
              "prepare",
              "finish2",
              "finish3",
);

82
enum qemuMigrationCookieFlags {
83
    QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS,
84
    QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE,
85
    QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT,
86
    QEMU_MIGRATION_COOKIE_FLAG_NETWORK,
87
    QEMU_MIGRATION_COOKIE_FLAG_NBD,
88
    QEMU_MIGRATION_COOKIE_FLAG_STATS,
89 90 91 92 93 94 95

    QEMU_MIGRATION_COOKIE_FLAG_LAST
};

VIR_ENUM_DECL(qemuMigrationCookieFlag);
VIR_ENUM_IMPL(qemuMigrationCookieFlag,
              QEMU_MIGRATION_COOKIE_FLAG_LAST,
96 97 98 99
              "graphics",
              "lockstate",
              "persistent",
              "network",
100 101
              "nbd",
              "statistics");
102 103 104

enum qemuMigrationCookieFeatures {
    QEMU_MIGRATION_COOKIE_GRAPHICS  = (1 << QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS),
105
    QEMU_MIGRATION_COOKIE_LOCKSTATE = (1 << QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE),
106
    QEMU_MIGRATION_COOKIE_PERSISTENT = (1 << QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT),
107
    QEMU_MIGRATION_COOKIE_NETWORK = (1 << QEMU_MIGRATION_COOKIE_FLAG_NETWORK),
108
    QEMU_MIGRATION_COOKIE_NBD = (1 << QEMU_MIGRATION_COOKIE_FLAG_NBD),
109
    QEMU_MIGRATION_COOKIE_STATS = (1 << QEMU_MIGRATION_COOKIE_FLAG_STATS),
110 111 112 113 114 115 116 117 118 119 120 121
};

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

122 123 124 125 126 127
typedef struct _qemuMigrationCookieNetData qemuMigrationCookieNetData;
typedef qemuMigrationCookieNetData *qemuMigrationCookieNetDataPtr;
struct _qemuMigrationCookieNetData {
    int vporttype; /* enum virNetDevVPortProfile */

    /*
128
     * Array of pointers to saved data. Each VIF will have its own
129 130 131 132 133 134 135 136 137 138 139 140 141 142
     * 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;
};

143 144 145 146
typedef struct _qemuMigrationCookieNBD qemuMigrationCookieNBD;
typedef qemuMigrationCookieNBD *qemuMigrationCookieNBDPtr;
struct _qemuMigrationCookieNBD {
    int port; /* on which port does NBD server listen for incoming data */
147 148 149 150 151 152

    size_t ndisks;  /* Number of items in @disk array */
    struct {
        char *target;                   /* Disk target */
        unsigned long long capacity;    /* And its capacity */
    } *disks;
153 154
};

155 156 157
typedef struct _qemuMigrationCookie qemuMigrationCookie;
typedef qemuMigrationCookie *qemuMigrationCookiePtr;
struct _qemuMigrationCookie {
E
Eric Blake 已提交
158 159
    unsigned int flags;
    unsigned int flagsMandatory;
160 161

    /* Host properties */
162 163 164 165
    unsigned char localHostuuid[VIR_UUID_BUFLEN];
    unsigned char remoteHostuuid[VIR_UUID_BUFLEN];
    char *localHostname;
    char *remoteHostname;
166 167 168 169

    /* Guest properties */
    unsigned char uuid[VIR_UUID_BUFLEN];
    char *name;
170

171 172 173 174
    /* If (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) */
    char *lockState;
    char *lockDriver;

175 176
    /* If (flags & QEMU_MIGRATION_COOKIE_GRAPHICS) */
    qemuMigrationCookieGraphicsPtr graphics;
177 178 179

    /* If (flags & QEMU_MIGRATION_COOKIE_PERSISTENT) */
    virDomainDefPtr persistent;
180 181 182

    /* If (flags & QEMU_MIGRATION_COOKIE_NETWORK) */
    qemuMigrationCookieNetworkPtr network;
183 184 185

    /* If (flags & QEMU_MIGRATION_COOKIE_NBD) */
    qemuMigrationCookieNBDPtr nbd;
186 187 188

    /* If (flags & QEMU_MIGRATION_COOKIE_STATS) */
    qemuDomainJobInfoPtr jobInfo;
189 190
};

191 192 193 194 195 196 197 198 199
static void qemuMigrationCookieGraphicsFree(qemuMigrationCookieGraphicsPtr grap)
{
    if (!grap)
        return;
    VIR_FREE(grap->listen);
    VIR_FREE(grap->tlsSubject);
    VIR_FREE(grap);
}

200

201 202 203
static void
qemuMigrationCookieNetworkFree(qemuMigrationCookieNetworkPtr network)
{
204
    size_t i;
205 206 207 208 209 210 211 212 213 214 215 216 217

    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);
}


218 219 220 221 222 223 224 225 226 227 228 229
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);
}


230 231 232 233 234
static void qemuMigrationCookieFree(qemuMigrationCookiePtr mig)
{
    if (!mig)
        return;

235 236
    qemuMigrationCookieGraphicsFree(mig->graphics);
    qemuMigrationCookieNetworkFree(mig->network);
237
    qemuMigrationCookieNBDFree(mig->nbd);
238

239 240
    VIR_FREE(mig->localHostname);
    VIR_FREE(mig->remoteHostname);
241
    VIR_FREE(mig->name);
242 243
    VIR_FREE(mig->lockState);
    VIR_FREE(mig->lockDriver);
244
    VIR_FREE(mig->jobInfo);
245 246 247 248
    VIR_FREE(mig);
}


249
#ifdef WITH_GNUTLS
250 251 252 253 254 255 256 257 258 259 260 261
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)
262
        goto error;
263 264

    if (virFileReadAll(certfile, 8192, &pemdata) < 0) {
265 266
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unable to read server cert %s"), certfile);
267 268 269 270 271
        goto error;
    }

    ret = gnutls_x509_crt_init(&cert);
    if (ret < 0) {
272 273 274
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot initialize cert object: %s"),
                       gnutls_strerror(ret));
275 276 277 278 279 280 281 282
        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) {
283 284 285
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot load cert data from %s: %s"),
                       certfile, gnutls_strerror(ret));
286 287 288 289 290
        goto error;
    }

    subjectlen = 1024;
    if (VIR_ALLOC_N(subject, subjectlen+1) < 0)
291
        goto error;
292 293 294 295 296 297 298 299 300

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

    VIR_FREE(certfile);
    VIR_FREE(pemdata);

    return subject;

301
 error:
302 303 304 305
    VIR_FREE(certfile);
    VIR_FREE(pemdata);
    return NULL;
}
306
#endif
307 308

static qemuMigrationCookieGraphicsPtr
309
qemuMigrationCookieGraphicsAlloc(virQEMUDriverPtr driver,
310 311 312 313
                                 virDomainGraphicsDefPtr def)
{
    qemuMigrationCookieGraphicsPtr mig = NULL;
    const char *listenAddr;
314
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
315 316

    if (VIR_ALLOC(mig) < 0)
317
        goto error;
318 319 320 321

    mig->type = def->type;
    if (mig->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
        mig->port = def->data.vnc.port;
322
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
323
        if (!listenAddr)
324
            listenAddr = cfg->vncListen;
325

326
#ifdef WITH_GNUTLS
327 328
        if (cfg->vncTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->vncTLSx509certdir)))
329
            goto error;
330
#endif
331 332
    } else {
        mig->port = def->data.spice.port;
333
        if (cfg->spiceTLS)
334 335 336
            mig->tlsPort = def->data.spice.tlsPort;
        else
            mig->tlsPort = -1;
337
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
338
        if (!listenAddr)
339
            listenAddr = cfg->spiceListen;
340

341
#ifdef WITH_GNUTLS
342 343
        if (cfg->spiceTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->spiceTLSx509certdir)))
344
            goto error;
345
#endif
346
    }
347 348
    if (VIR_STRDUP(mig->listen, listenAddr) < 0)
        goto error;
349

350
    virObjectUnref(cfg);
351 352
    return mig;

353
 error:
354
    qemuMigrationCookieGraphicsFree(mig);
355
    virObjectUnref(cfg);
356 357 358 359
    return NULL;
}


360
static qemuMigrationCookieNetworkPtr
361
qemuMigrationCookieNetworkAlloc(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
362 363 364
                                virDomainDefPtr def)
{
    qemuMigrationCookieNetworkPtr mig;
365
    size_t i;
366 367

    if (VIR_ALLOC(mig) < 0)
368
        goto error;
369 370 371 372

    mig->nnets = def->nnets;

    if (VIR_ALLOC_N(mig->net, def->nnets) <0)
373
        goto error;
374 375 376 377 378 379 380 381 382 383 384 385 386 387 388

    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:
389
               break;
390
            case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
391 392
                if (virNetDevOpenvswitchGetMigrateData(&mig->net[i].portdata,
                                                       netptr->ifname) != 0) {
J
Jiri Denemark 已提交
393 394 395
                        virReportError(VIR_ERR_INTERNAL_ERROR,
                                       _("Unable to run command to get OVS port data for "
                                         "interface %s"), netptr->ifname);
396 397 398
                        goto error;
                }
                break;
399 400 401 402 403 404 405
            default:
                break;
            }
        }
    }
    return mig;

406
 error:
407 408 409 410
    qemuMigrationCookieNetworkFree(mig);
    return NULL;
}

411 412 413
static qemuMigrationCookiePtr
qemuMigrationCookieNew(virDomainObjPtr dom)
{
J
Jiri Denemark 已提交
414
    qemuDomainObjPrivatePtr priv = dom->privateData;
415
    qemuMigrationCookiePtr mig = NULL;
J
Jiri Denemark 已提交
416
    const char *name;
417 418

    if (VIR_ALLOC(mig) < 0)
419
        goto error;
420

J
Jiri Denemark 已提交
421 422 423 424
    if (priv->origname)
        name = priv->origname;
    else
        name = dom->def->name;
425 426
    if (VIR_STRDUP(mig->name, name) < 0)
        goto error;
427 428
    memcpy(mig->uuid, dom->def->uuid, VIR_UUID_BUFLEN);

429
    if (!(mig->localHostname = virGetHostname()))
430
        goto error;
431
    if (virGetHostUUID(mig->localHostuuid) < 0) {
432 433
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Unable to obtain host UUID"));
434 435 436 437 438
        goto error;
    }

    return mig;

439
 error:
440 441 442 443 444
    qemuMigrationCookieFree(mig);
    return NULL;
}


445 446
static int
qemuMigrationCookieAddGraphics(qemuMigrationCookiePtr mig,
447
                               virQEMUDriverPtr driver,
448 449
                               virDomainObjPtr dom)
{
450 451
    size_t i = 0;

452
    if (mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) {
453 454
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration graphics data already present"));
455 456 457
        return -1;
    }

458 459 460 461 462 463 464 465
    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;
       }
466
    }
467 468 469 470 471

    return 0;
}


472 473
static int
qemuMigrationCookieAddLockstate(qemuMigrationCookiePtr mig,
474
                                virQEMUDriverPtr driver,
475 476 477 478 479
                                virDomainObjPtr dom)
{
    qemuDomainObjPrivatePtr priv = dom->privateData;

    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
480 481
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration lockstate data already present"));
482 483 484 485
        return -1;
    }

    if (virDomainObjGetState(dom, NULL) == VIR_DOMAIN_PAUSED) {
486
        if (VIR_STRDUP(mig->lockState, priv->lockState) < 0)
487 488 489 490 491 492
            return -1;
    } else {
        if (virDomainLockProcessInquire(driver->lockManager, dom, &mig->lockState) < 0)
            return -1;
    }

493
    if (VIR_STRDUP(mig->lockDriver, virLockManagerPluginGetName(driver->lockManager)) < 0) {
494 495 496 497 498 499 500 501 502 503 504
        VIR_FREE(mig->lockState);
        return -1;
    }

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

    return 0;
}


505 506 507 508 509
static int
qemuMigrationCookieAddPersistent(qemuMigrationCookiePtr mig,
                                 virDomainObjPtr dom)
{
    if (mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) {
510 511
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration persistent data already present"));
512 513 514 515 516 517 518 519 520 521 522 523 524
        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;
}


525 526
static int
qemuMigrationCookieAddNetwork(qemuMigrationCookiePtr mig,
527
                              virQEMUDriverPtr driver,
528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545
                              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;
}

546

547 548
static int
qemuMigrationCookieAddNBD(qemuMigrationCookiePtr mig,
549
                          virQEMUDriverPtr driver,
550 551 552
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
553 554
    virHashTablePtr stats = NULL;
    size_t i;
555
    int ret = -1, rc;
556 557 558

    /* It is not a bug if there already is a NBD data */
    if (!mig->nbd &&
559
        VIR_ALLOC(mig->nbd) < 0)
560 561
        return -1;

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

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

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

575 576 577
            if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                               priv->job.asyncJob) < 0)
                goto cleanup;
578 579
            rc = qemuMonitorBlockStatsUpdateCapacity(priv->mon, stats, false);
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
580
                goto cleanup;
581
            if (rc < 0)
582 583 584 585 586 587 588 589 590 591 592 593 594 595
                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++;
    }

596 597 598
    mig->nbd->port = priv->nbdPort;
    mig->flags |= QEMU_MIGRATION_COOKIE_NBD;

599 600 601 602
    ret = 0;
 cleanup:
    virHashFree(stats);
    return ret;
603 604 605
}


606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624
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;
}


625 626 627
static void qemuMigrationCookieGraphicsXMLFormat(virBufferPtr buf,
                                                 qemuMigrationCookieGraphicsPtr grap)
{
628
    virBufferAsprintf(buf, "<graphics type='%s' port='%d' listen='%s'",
629 630 631 632 633 634
                      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");
635 636 637 638
        virBufferAdjustIndent(buf, 2);
        virBufferEscapeString(buf, "<cert info='subject' value='%s'/>\n", grap->tlsSubject);
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</graphics>\n");
639 640 641 642 643 644
    } else {
        virBufferAddLit(buf, "/>\n");
    }
}


645 646 647 648
static void
qemuMigrationCookieNetworkXMLFormat(virBufferPtr buf,
                                    qemuMigrationCookieNetworkPtr optr)
{
649
    size_t i;
650 651 652 653 654 655
    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) {
656 657
                virBufferAddLit(buf, "<network>\n");
                virBufferAdjustIndent(buf, 2);
658 659
                empty = false;
            }
660
            virBufferAsprintf(buf, "<interface index='%zu' vporttype='%s'",
661 662 663
                              i, virNetDevVPortTypeToString(optr->net[i].vporttype));
            if (optr->net[i].portdata) {
                virBufferAddLit(buf, ">\n");
664 665
                virBufferAdjustIndent(buf, 2);
                virBufferEscapeString(buf, "<portdata>%s</portdata>\n",
666
                                      optr->net[i].portdata);
667 668
                virBufferAdjustIndent(buf, -2);
                virBufferAddLit(buf, "</interface>\n");
669 670 671 672 673
            } else {
                virBufferAddLit(buf, "/>\n");
            }
        }
    }
674 675 676 677
    if (!empty) {
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</network>\n");
    }
678 679 680
}


681 682 683 684 685 686 687 688 689
static void
qemuMigrationCookieStatisticsXMLFormat(virBufferPtr buf,
                                       qemuDomainJobInfoPtr jobInfo)
{
    qemuMonitorMigrationStatus *status = &jobInfo->status;

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

690 691
    virBufferAsprintf(buf, "<started>%llu</started>\n", jobInfo->started);
    virBufferAsprintf(buf, "<stopped>%llu</stopped>\n", jobInfo->stopped);
692 693 694
    virBufferAsprintf(buf, "<sent>%llu</sent>\n", jobInfo->sent);
    if (jobInfo->timeDeltaSet)
        virBufferAsprintf(buf, "<delta>%lld</delta>\n", jobInfo->timeDelta);
695

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
    virXPathULongLong("string(./started[1])", ctxt, &jobInfo->started);
    virXPathULongLong("string(./stopped[1])", ctxt, &jobInfo->stopped);
1053 1054 1055
    virXPathULongLong("string(./sent[1])", ctxt, &jobInfo->sent);
    if (virXPathLongLong("string(./delta[1])", ctxt, &jobInfo->timeDelta) == 0)
        jobInfo->timeDeltaSet = true;
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);
1061

1062 1063 1064
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_DOWNTIME "[1])",
                          ctxt, &status->downtime) == 0)
        status->downtime_set = true;
1065 1066 1067
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_SETUP_TIME "[1])",
                          ctxt, &status->setup_time) == 0)
        status->setup_time_set = true;
1068 1069 1070 1071 1072 1073 1074

    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);
1075 1076
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_BPS "[1])",
                      ctxt, &status->ram_bps);
1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091

    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);
1092 1093
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_BPS "[1])",
                      ctxt, &status->disk_bps);
1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112

    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;
}


1113 1114
static int
qemuMigrationCookieXMLParse(qemuMigrationCookiePtr mig,
1115
                            virQEMUDriverPtr driver,
1116
                            xmlDocPtr doc,
1117
                            xmlXPathContextPtr ctxt,
E
Eric Blake 已提交
1118
                            unsigned int flags)
1119 1120
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
1121
    char *tmp = NULL;
1122
    xmlNodePtr *nodes = NULL;
1123 1124
    size_t i;
    int n;
1125 1126 1127 1128
    virCapsPtr caps = NULL;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto error;
1129 1130 1131 1132 1133 1134 1135 1136

    /* 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))) {
1137 1138
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing name element in migration data"));
1139 1140 1141
        goto error;
    }
    if (STRNEQ(tmp, mig->name)) {
1142 1143 1144
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected name %s vs %s"),
                       tmp, mig->name);
1145 1146 1147 1148 1149 1150 1151
        goto error;
    }
    VIR_FREE(tmp);

    /* Extract domain uuid */
    tmp = virXPathString("string(./uuid[1])", ctxt);
    if (!tmp) {
1152 1153
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing uuid element in migration data"));
1154 1155 1156 1157
        goto error;
    }
    virUUIDFormat(mig->uuid, uuidstr);
    if (STRNEQ(tmp, uuidstr)) {
1158 1159 1160
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected UUID %s vs %s"),
                       tmp, uuidstr);
1161
        goto error;
1162 1163 1164 1165
    }
    VIR_FREE(tmp);

    /* Check & forbid "localhost" migration */
1166
    if (!(mig->remoteHostname = virXPathString("string(./hostname[1])", ctxt))) {
1167 1168
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostname element in migration data"));
1169 1170
        goto error;
    }
1171
    if (STREQ(mig->remoteHostname, mig->localHostname)) {
1172 1173 1174
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Attempt to migrate guest to the same host %s"),
                       mig->remoteHostname);
1175 1176 1177 1178
        goto error;
    }

    if (!(tmp = virXPathString("string(./hostuuid[1])", ctxt))) {
1179 1180 1181
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostuuid element in migration data"));
        goto error;
1182
    }
1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194
    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);
1195

1196 1197
    /* Check to ensure all mandatory features from XML are also
     * present in 'flags' */
1198
    if ((n = virXPathNodeSet("./feature", ctxt, &nodes)) < 0)
1199 1200
        goto error;

1201
    for (i = 0; i < n; i++) {
1202 1203 1204
        int val;
        char *str = virXMLPropString(nodes[i], "name");
        if (!str) {
1205 1206
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing feature name"));
1207 1208 1209 1210
            goto error;
        }

        if ((val = qemuMigrationCookieFlagTypeFromString(str)) < 0) {
1211 1212 1213
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unknown migration cookie feature %s"),
                           str);
1214 1215 1216 1217 1218
            VIR_FREE(str);
            goto error;
        }

        if ((flags & (1 << val)) == 0) {
1219 1220 1221
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unsupported migration cookie feature %s"),
                           str);
1222 1223 1224 1225 1226 1227
            VIR_FREE(str);
        }
        VIR_FREE(str);
    }
    VIR_FREE(nodes);

1228 1229 1230 1231 1232
    if ((flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        virXPathBoolean("count(./graphics) > 0", ctxt) &&
        (!(mig->graphics = qemuMigrationCookieGraphicsXMLParse(ctxt))))
        goto error;

1233 1234 1235 1236
    if ((flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        virXPathBoolean("count(./lockstate) > 0", ctxt)) {
        mig->lockDriver = virXPathString("string(./lockstate[1]/@driver)", ctxt);
        if (!mig->lockDriver) {
1237 1238
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Missing lock driver name in migration cookie"));
1239 1240 1241 1242 1243 1244 1245
            goto error;
        }
        mig->lockState = virXPathString("string(./lockstate[1]/leases[1])", ctxt);
        if (mig->lockState && STREQ(mig->lockState, ""))
            VIR_FREE(mig->lockState);
    }

1246 1247 1248
    if ((flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        virXPathBoolean("count(./domain) > 0", ctxt)) {
        if ((n = virXPathNodeSet("./domain", ctxt, &nodes)) > 1) {
1249 1250 1251 1252
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Too many domain elements in "
                             "migration cookie: %d"),
                           n);
1253 1254
            goto error;
        }
1255 1256
        mig->persistent = virDomainDefParseNode(doc, nodes[0],
                                                caps, driver->xmlopt,
1257
                                                VIR_DOMAIN_DEF_PARSE_INACTIVE);
1258 1259 1260 1261 1262 1263 1264 1265
        if (!mig->persistent) {
            /* virDomainDefParseNode already reported
             * an error for us */
            goto error;
        }
        VIR_FREE(nodes);
    }

1266 1267 1268 1269 1270
    if ((flags & QEMU_MIGRATION_COOKIE_NETWORK) &&
        virXPathBoolean("count(./network) > 0", ctxt) &&
        (!(mig->network = qemuMigrationCookieNetworkXMLParse(ctxt))))
        goto error;

1271
    if (flags & QEMU_MIGRATION_COOKIE_NBD &&
1272 1273 1274
        virXPathBoolean("boolean(./nbd)", ctxt) &&
        (!(mig->nbd = qemuMigrationCookieNBDXMLParse(ctxt))))
        goto error;
1275

1276 1277 1278 1279 1280
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        virXPathBoolean("boolean(./statistics)", ctxt) &&
        (!(mig->jobInfo = qemuMigrationCookieStatisticsXMLParse(ctxt))))
        goto error;

1281
    virObjectUnref(caps);
1282 1283
    return 0;

1284
 error:
1285
    VIR_FREE(tmp);
1286
    VIR_FREE(nodes);
1287
    virObjectUnref(caps);
1288 1289 1290 1291 1292 1293
    return -1;
}


static int
qemuMigrationCookieXMLParseStr(qemuMigrationCookiePtr mig,
1294
                               virQEMUDriverPtr driver,
1295
                               const char *xml,
E
Eric Blake 已提交
1296
                               unsigned int flags)
1297 1298 1299
{
    xmlDocPtr doc = NULL;
    xmlXPathContextPtr ctxt = NULL;
1300
    int ret = -1;
1301 1302 1303

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

1304
    if (!(doc = virXMLParseStringCtxt(xml, _("(qemu_migration_cookie)"), &ctxt)))
1305 1306
        goto cleanup;

1307
    ret = qemuMigrationCookieXMLParse(mig, driver, doc, ctxt, flags);
1308

1309
 cleanup:
1310 1311 1312 1313 1314 1315 1316 1317 1318
    xmlXPathFreeContext(ctxt);
    xmlFreeDoc(doc);

    return ret;
}


static int
qemuMigrationBakeCookie(qemuMigrationCookiePtr mig,
1319
                        virQEMUDriverPtr driver,
1320
                        virDomainObjPtr dom,
1321 1322
                        char **cookieout,
                        int *cookieoutlen,
E
Eric Blake 已提交
1323
                        unsigned int flags)
1324
{
1325 1326
    if (!cookieout || !cookieoutlen)
        return 0;
1327 1328 1329

    *cookieoutlen = 0;

1330 1331 1332 1333
    if (flags & QEMU_MIGRATION_COOKIE_GRAPHICS &&
        qemuMigrationCookieAddGraphics(mig, driver, dom) < 0)
        return -1;

1334 1335 1336 1337
    if (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE &&
        qemuMigrationCookieAddLockstate(mig, driver, dom) < 0)
        return -1;

1338 1339 1340 1341
    if (flags & QEMU_MIGRATION_COOKIE_PERSISTENT &&
        qemuMigrationCookieAddPersistent(mig, dom) < 0)
        return -1;

1342 1343 1344 1345 1346
    if (flags & QEMU_MIGRATION_COOKIE_NETWORK &&
        qemuMigrationCookieAddNetwork(mig, driver, dom) < 0) {
        return -1;
    }

1347 1348 1349 1350
    if ((flags & QEMU_MIGRATION_COOKIE_NBD) &&
        qemuMigrationCookieAddNBD(mig, driver, dom) < 0)
        return -1;

1351 1352 1353 1354
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        qemuMigrationCookieAddStatistics(mig, dom) < 0)
        return -1;

1355
    if (!(*cookieout = qemuMigrationCookieXMLFormatStr(driver, mig)))
1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
        return -1;

    *cookieoutlen = strlen(*cookieout) + 1;

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

    return 0;
}


static qemuMigrationCookiePtr
1367
qemuMigrationEatCookie(virQEMUDriverPtr driver,
1368
                       virDomainObjPtr dom,
1369 1370
                       const char *cookiein,
                       int cookieinlen,
E
Eric Blake 已提交
1371
                       unsigned int flags)
1372 1373 1374 1375 1376 1377
{
    qemuMigrationCookiePtr mig = NULL;

    /* Parse & validate incoming cookie (if any) */
    if (cookiein && cookieinlen &&
        cookiein[cookieinlen-1] != '\0') {
1378 1379
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration cookie was not NULL terminated"));
1380 1381 1382 1383 1384 1385 1386 1387 1388 1389
        goto error;
    }

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

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

    if (cookiein && cookieinlen &&
        qemuMigrationCookieXMLParseStr(mig,
1390
                                       driver,
1391 1392 1393 1394
                                       cookiein,
                                       flags) < 0)
        goto error;

1395 1396 1397 1398 1399 1400 1401 1402
    if (flags & QEMU_MIGRATION_COOKIE_PERSISTENT &&
        mig->persistent &&
        STRNEQ(dom->def->name, mig->persistent->name)) {
        VIR_FREE(mig->persistent->name);
        if (VIR_STRDUP(mig->persistent->name, dom->def->name) < 0)
            goto error;
    }

1403 1404 1405
    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
        if (!mig->lockDriver) {
            if (virLockManagerPluginUsesState(driver->lockManager)) {
1406 1407 1408
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Missing %s lock state for migration cookie"),
                               virLockManagerPluginGetName(driver->lockManager));
1409 1410 1411 1412
                goto error;
            }
        } else if (STRNEQ(mig->lockDriver,
                          virLockManagerPluginGetName(driver->lockManager))) {
1413 1414 1415 1416
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Source host lock driver %s different from target %s"),
                           mig->lockDriver,
                           virLockManagerPluginGetName(driver->lockManager));
1417 1418 1419 1420
            goto error;
        }
    }

1421 1422
    return mig;

1423
 error:
1424 1425 1426
    qemuMigrationCookieFree(mig);
    return NULL;
}
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 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474
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;
}

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

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;

1529 1530 1531 1532 1533
    case VIR_STORAGE_TYPE_NETWORK:
        VIR_DEBUG("Skipping creation of network disk '%s'",
                  disk->dst);
        return 0;

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

1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602
static bool
qemuMigrateDisk(virDomainDiskDef const *disk,
                size_t nmigrate_disks, const char **migrate_disks)
{
    size_t i;

    /* Check if the disk alias is in the list */
    if (nmigrate_disks) {
        for (i = 0; i < nmigrate_disks; i++) {
            if (STREQ(disk->dst, migrate_disks[i]))
                return true;
        }
        return false;
    }

    /* Default is to migrate only non-shared non-readonly disks
     * with source */
    return !disk->src->shared && !disk->src->readonly &&
           virDomainDiskGetSource(disk);
}

1603 1604 1605 1606 1607

static int
qemuMigrationPrecreateStorage(virConnectPtr conn,
                              virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                              virDomainObjPtr vm,
1608 1609 1610
                              qemuMigrationCookieNBDPtr nbd,
                              size_t nmigrate_disks,
                              const char **migrate_disks)
1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
{
    int ret = -1;
    size_t i = 0;

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

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

1622
        VIR_DEBUG("Looking up disk target '%s' (capacity=%llu)",
1623 1624
                  nbd->disks[i].target, nbd->disks[i].capacity);

1625 1626
        if (!(disk = virDomainDiskByName(vm->def, nbd->disks[i].target,
                                         false))) {
1627 1628 1629 1630 1631 1632 1633 1634
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unable to find disk by target: %s"),
                           nbd->disks[i].target);
            goto cleanup;
        }

        diskSrcPath = virDomainDiskGetSource(disk);

1635 1636
        /* Skip disks we don't want to migrate and already existing disks. */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks) ||
1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652
            (diskSrcPath && virFileExists(diskSrcPath))) {
            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;
}


1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666
/**
 * 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
1667
                            virDomainObjPtr vm,
1668 1669 1670
                            const char *listenAddr,
                            size_t nmigrate_disks,
                            const char **migrate_disks)
1671 1672 1673 1674 1675 1676 1677 1678 1679 1680
{
    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];

1681 1682
        /* check whether disk should be migrated */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks))
1683 1684 1685 1686
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1687
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1688 1689 1690 1691 1692 1693 1694
            goto cleanup;

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

        if (!port &&
1695
            ((virPortAllocatorAcquire(driver->migrationPorts, &port) < 0) ||
1696
             (qemuMonitorNBDServerStart(priv->mon, listenAddr, port) < 0))) {
1697
            goto exit_monitor;
1698 1699
        }

1700 1701 1702
        if (qemuMonitorNBDServerAdd(priv->mon, diskAlias, true) < 0)
            goto exit_monitor;
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
1703 1704 1705 1706 1707 1708
            goto cleanup;
    }

    priv->nbdPort = port;
    ret = 0;

1709
 cleanup:
1710
    VIR_FREE(diskAlias);
1711
    if (ret < 0)
1712
        virPortAllocatorRelease(driver->migrationPorts, port);
1713
    return ret;
1714 1715 1716 1717

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;
1718 1719
}

1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746

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

    if (!mig->nbd)
        return 0;

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

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

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


/**
1747
 * qemuMigrationDriveMirrorReady:
1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759
 * @driver: qemu driver
 * @vm: domain
 *
 * Check the status of all drive-mirrors started by
 * qemuMigrationDriveMirror. Any pending block job events
 * for the mirrored disks will be processed.
 *
 * Returns 1 if all mirrors are "ready",
 *         0 if some mirrors are still performing initial sync,
 *        -1 on error.
 */
static int
1760
qemuMigrationDriveMirrorReady(virQEMUDriverPtr driver,
1761 1762 1763
                              virDomainObjPtr vm)
{
    size_t i;
1764 1765
    size_t notReady = 0;
    int status;
1766 1767 1768

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
1769
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1770

1771
        if (!diskPriv->migrating)
1772 1773
            continue;

1774 1775
        status = qemuBlockJobUpdate(driver, vm, disk);
        if (status == VIR_DOMAIN_BLOCK_JOB_FAILED) {
1776 1777 1778 1779 1780
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("migration of disk %s failed"),
                           disk->dst);
            return -1;
        }
1781 1782 1783

        if (disk->mirrorState != VIR_DOMAIN_DISK_MIRROR_STATE_READY)
            notReady++;
1784 1785
    }

1786 1787 1788 1789 1790 1791 1792
    if (notReady) {
        VIR_DEBUG("Waiting for %zu disk mirrors to get ready", notReady);
        return 0;
    } else {
        VIR_DEBUG("All disk mirrors are ready");
        return 1;
    }
1793 1794 1795
}


1796 1797 1798 1799
/*
 * If @check is true, the function will report an error and return a different
 * code in case a block job fails. This way we can properly abort migration in
 * case some block jobs failed once all memory has already been transferred.
1800
 *
1801 1802 1803 1804
 * Returns 1 if all mirrors are gone,
 *         0 if some mirrors are still active,
 *         -1 some mirrors failed but some are still active,
 *         -2 all mirrors are gone but some of them failed.
1805 1806
 */
static int
1807
qemuMigrationDriveMirrorCancelled(virQEMUDriverPtr driver,
1808
                                  virDomainObjPtr vm,
1809
                                  bool check)
1810
{
1811 1812 1813 1814
    size_t i;
    size_t active = 0;
    int status;
    bool failed = false;
1815

1816 1817 1818
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1819

1820 1821
        if (!diskPriv->migrating)
            continue;
1822

1823 1824 1825 1826
        status = qemuBlockJobUpdate(driver, vm, disk);
        switch (status) {
        case VIR_DOMAIN_BLOCK_JOB_FAILED:
            if (check) {
1827
                virReportError(VIR_ERR_OPERATION_FAILED,
1828
                               _("migration of disk %s failed"),
1829
                               disk->dst);
1830
                failed = true;
1831
            }
1832 1833 1834 1835 1836 1837
            /* fallthrough */
        case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
            qemuBlockJobSyncEnd(driver, vm, disk);
            diskPriv->migrating = false;
            break;
1838

1839 1840
        default:
            active++;
1841
        }
1842
    }
1843

1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859
    if (failed) {
        if (active) {
            VIR_DEBUG("Some disk mirrors failed; still waiting for %zu "
                      "disk mirrors to finish", active);
            return -1;
        } else {
            VIR_DEBUG("All disk mirrors are gone; some of them failed");
            return -2;
        }
    } else {
        if (active) {
            VIR_DEBUG("Waiting for %zu disk mirrors to finish", active);
            return 0;
        } else {
            VIR_DEBUG("All disk mirrors are gone");
            return 1;
1860
        }
1861
    }
1862
}
1863 1864


1865 1866 1867 1868 1869 1870 1871 1872 1873
/*
 * Returns 0 on success,
 *         1 when job is already completed or it failed and failNoJob is false,
 *         -1 on error or when job failed and failNoJob is true.
 */
static int
qemuMigrationCancelOneDriveMirror(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  virDomainDiskDefPtr disk,
1874 1875
                                  bool failNoJob,
                                  qemuDomainAsyncJob asyncJob)
1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char *diskAlias = NULL;
    int ret = -1;
    int status;
    int rv;

    status = qemuBlockJobUpdate(driver, vm, disk);
    switch (status) {
    case VIR_DOMAIN_BLOCK_JOB_FAILED:
    case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        if (failNoJob) {
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("migration of disk %s failed"),
                           disk->dst);
            return -1;
        }
        return 1;

    case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
        return 1;
    }

    if (virAsprintf(&diskAlias, "%s%s",
                    QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
        return -1;

1903
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
1904 1905 1906 1907 1908 1909 1910 1911
        goto cleanup;

    rv = qemuMonitorBlockJobCancel(priv->mon, diskAlias, true);

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

    ret = 0;
1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922

 cleanup:
    VIR_FREE(diskAlias);
    return ret;
}


/**
 * qemuMigrationCancelDriveMirror:
 * @driver: qemu driver
 * @vm: domain
1923
 * @check: if true report an error when some of the mirrors fails
1924 1925 1926 1927 1928 1929 1930 1931 1932
 *
 * Cancel all drive-mirrors started by qemuMigrationDriveMirror.
 * Any pending block job events for the affected disks will be
 * processed.
 *
 * Returns 0 on success, -1 otherwise.
 */
static int
qemuMigrationCancelDriveMirror(virQEMUDriverPtr driver,
1933
                               virDomainObjPtr vm,
1934 1935
                               bool check,
                               qemuDomainAsyncJob asyncJob)
1936
{
1937
    virErrorPtr err = NULL;
1938
    int ret = -1;
1939
    size_t i;
1940 1941 1942 1943
    int rv;
    bool failed = false;

    VIR_DEBUG("Cancelling drive mirrors for domain %s", vm->def->name);
1944 1945 1946

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
1947
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1948

1949
        if (!diskPriv->migrating)
1950 1951
            continue;

1952 1953
        rv = qemuMigrationCancelOneDriveMirror(driver, vm, disk,
                                               check, asyncJob);
1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969
        if (rv != 0) {
            if (rv < 0) {
                if (!err)
                    err = virSaveLastError();
                failed = true;
            }
            qemuBlockJobSyncEnd(driver, vm, disk);
            diskPriv->migrating = false;
        }
    }

    while ((rv = qemuMigrationDriveMirrorCancelled(driver, vm, check)) != 1) {
        if (rv < 0) {
            failed = true;
            if (rv == -2)
                break;
1970
        }
1971

1972 1973 1974 1975 1976
        if (failed && !err)
            err = virSaveLastError();

        if (virDomainObjWait(vm) < 0)
            goto cleanup;
1977 1978
    }

1979 1980 1981
    ret = failed ? -1 : 0;

 cleanup:
1982 1983 1984 1985 1986
    if (err) {
        virSetError(err);
        virFreeError(err);
    }
    return ret;
1987 1988 1989
}


1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
/**
 * 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
2001 2002 2003 2004 2005
 * simultaneously to both source and destination. On success,
 * update @migrate_flags so we don't tell 'migrate' command
 * to do the very same operation. On failure, the caller is
 * expected to call qemuMigrationCancelDriveMirror to stop all
 * running mirrors.
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
qemuMigrationDriveMirror(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMigrationCookiePtr mig,
                         const char *host,
                         unsigned long speed,
2016 2017 2018
                         unsigned int *migrate_flags,
                         size_t nmigrate_disks,
                         const char **migrate_disks)
2019 2020 2021 2022
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int port;
2023
    size_t i;
2024 2025
    char *diskAlias = NULL;
    char *nbd_dest = NULL;
2026
    char *hoststr = NULL;
2027
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
2028
    int rv;
2029
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2030 2031

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
2032 2033 2034 2035 2036

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

2037 2038
    /* escape literal IPv6 address */
    if (strchr(host, ':')) {
2039
        if (virAsprintf(&hoststr, "[%s]", host) < 0)
2040
            goto cleanup;
2041
    } else if (VIR_STRDUP(hoststr, host) < 0) {
2042
        goto cleanup;
2043 2044
    }

2045 2046 2047 2048 2049
    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];
2050
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
2051
        int mon_ret;
2052

2053 2054
        /* check whether disk should be migrated */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks))
2055 2056 2057 2058 2059
            continue;

        if ((virAsprintf(&diskAlias, "%s%s",
                         QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0) ||
            (virAsprintf(&nbd_dest, "nbd:%s:%d:exportname=%s",
2060
                         hoststr, port, diskAlias) < 0))
2061 2062
            goto cleanup;

2063
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
2064
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
2065 2066
            goto cleanup;

2067
        qemuBlockJobSyncBegin(disk);
2068
        /* Force "raw" format for NBD export */
2069
        mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest,
2070
                                         "raw", speed, 0, 0, mirror_flags);
2071 2072
        VIR_FREE(diskAlias);
        VIR_FREE(nbd_dest);
2073

2074
        if (qemuDomainObjExitMonitor(driver, vm) < 0 || mon_ret < 0) {
2075
            qemuBlockJobSyncEnd(driver, vm, disk);
2076 2077
            goto cleanup;
        }
2078
        diskPriv->migrating = true;
2079 2080 2081 2082 2083

        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto cleanup;
        }
2084
    }
2085

2086 2087 2088
    while ((rv = qemuMigrationDriveMirrorReady(driver, vm)) != 1) {
        if (rv < 0)
            goto cleanup;
2089

2090
        if (priv->job.abortJob) {
2091 2092 2093 2094 2095
            priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
            goto cleanup;
2096
        }
2097

2098
        if (virDomainObjWait(vm) < 0)
2099
            goto cleanup;
2100 2101
    }

2102
    /* Okay, all disks are ready. Modify migrate_flags */
2103 2104 2105 2106
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

2107
 cleanup:
2108
    virObjectUnref(cfg);
2109 2110
    VIR_FREE(diskAlias);
    VIR_FREE(nbd_dest);
2111
    VIR_FREE(hoststr);
2112 2113
    return ret;
}
2114

2115

2116 2117 2118 2119 2120 2121 2122 2123 2124
/* 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. */
2125
bool
2126
qemuMigrationIsAllowed(virQEMUDriverPtr driver, virDomainObjPtr vm,
2127
                       virDomainDefPtr def, bool remote, bool abort_on_error)
2128
{
2129
    int nsnapshots;
2130
    int pauseReason;
2131
    size_t i;
2132 2133 2134

    if (vm) {
        if (qemuProcessAutoDestroyActive(driver, vm)) {
2135 2136
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("domain is marked for auto destroy"));
2137 2138
            return false;
        }
2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151

        /* 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;
            }
2152 2153

            /* cancel migration if disk I/O error is emitted while migrating */
2154 2155
            if (abort_on_error &&
                virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
2156 2157 2158 2159 2160 2161
                pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("cannot migrate domain with I/O error"));
                return false;
            }

2162
        }
2163

2164
        if (qemuDomainHasBlockjob(vm, false)) {
E
Eric Blake 已提交
2165
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
2166
                           _("domain has an active block job"));
E
Eric Blake 已提交
2167 2168
            return false;
        }
2169 2170 2171

        def = vm->def;
    }
2172 2173

    /* Migration with USB host devices is allowed, all other devices are
2174
     * forbidden. */
2175 2176 2177 2178
    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) {
2179 2180 2181
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("domain has assigned non-USB host devices"));
            return false;
2182 2183
        }
    }
2184

2185 2186 2187
    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
2188

2189 2190
            if (feature->policy != VIR_CPU_FEATURE_REQUIRE)
                continue;
J
Ján Tomko 已提交
2191

2192 2193 2194 2195 2196 2197 2198
            /* 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
2199 2200 2201
        }
    }

2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215
    /* Verify that memory device config can be transferred reliably */
    for (i = 0; i < def->nmems; i++) {
        virDomainMemoryDefPtr mem = def->mems[i];

        if (mem->model == VIR_DOMAIN_MEMORY_MODEL_DIMM &&
            mem->info.type != VIR_DOMAIN_DEVICE_ADDRESS_TYPE_DIMM) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("domain's dimm info lacks slot ID "
                             "or base address"));

            return false;
        }
    }

2216 2217 2218
    return true;
}

2219
static bool
2220 2221 2222
qemuMigrationIsSafe(virDomainDefPtr def,
                    size_t nmigrate_disks,
                    const char **migrate_disks)
2223
{
2224
    size_t i;
2225

2226
    for (i = 0; i < def->ndisks; i++) {
2227
        virDomainDiskDefPtr disk = def->disks[i];
2228
        const char *src = virDomainDiskGetSource(disk);
2229

2230 2231
        /* Our code elsewhere guarantees shared disks are either readonly (in
         * which case cache mode doesn't matter) or used with cache=none */
2232
        if (qemuMigrateDisk(disk, nmigrate_disks, migrate_disks) &&
2233
            disk->cachemode != VIR_DOMAIN_DISK_CACHE_DISABLE) {
2234
            int rc;
2235

E
Eric Blake 已提交
2236
            if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
2237
                if ((rc = virFileIsSharedFS(src)) < 0)
2238 2239
                    return false;
                else if (rc == 0)
2240
                    continue;
2241
                if ((rc = virStorageFileIsClusterFS(src)) < 0)
2242
                    return false;
2243 2244
                else if (rc == 1)
                    continue;
2245 2246
            } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                       disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
2247
                continue;
2248
            }
2249

2250 2251 2252
            virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                           _("Migration may lead to data corruption if disks"
                             " use cache != none"));
2253 2254 2255 2256 2257 2258 2259
            return false;
        }
    }

    return true;
}

2260 2261 2262 2263
/** qemuMigrationSetOffline
 * Pause domain for non-live migration.
 */
int
2264
qemuMigrationSetOffline(virQEMUDriverPtr driver,
2265 2266 2267
                        virDomainObjPtr vm)
{
    int ret;
2268
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
2269 2270
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
2271
    if (ret == 0) {
2272
        virObjectEventPtr event;
2273

2274
        event = virDomainEventLifecycleNewFromObj(vm,
2275 2276 2277 2278 2279 2280 2281 2282 2283 2284
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
        if (event)
            qemuDomainEventQueue(driver, event);
    }

    return ret;
}


2285 2286 2287
static int
qemuMigrationSetCompression(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
2288
                            bool state,
2289
                            qemuDomainAsyncJob job)
2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302
{
    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;
2303 2304 2305
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321
    } 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,
2322 2323
                QEMU_MONITOR_MIGRATION_CAPS_XBZRLE,
                state);
2324

2325
 cleanup:
2326 2327
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2328 2329 2330
    return ret;
}

2331 2332 2333
static int
qemuMigrationSetAutoConverge(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
2334
                             bool state,
2335
                             qemuDomainAsyncJob job)
2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348
{
    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;
2349 2350 2351
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2352 2353 2354 2355 2356 2357 2358 2359 2360 2361
    } 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,
2362 2363
                QEMU_MONITOR_MIGRATION_CAPS_AUTO_CONVERGE,
                state);
2364

2365
 cleanup:
2366 2367
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2368 2369 2370 2371
    return ret;
}


2372 2373 2374
static int
qemuMigrationSetPinAll(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
2375
                       bool state,
2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389
                       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;
2390 2391 2392
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408
    } 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,
2409 2410
                QEMU_MONITOR_MIGRATION_CAPS_RDMA_PIN_ALL,
                state);
2411 2412

 cleanup:
2413 2414
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2415 2416 2417
    return ret;
}

2418
static int
2419
qemuMigrationWaitForSpice(virDomainObjPtr vm)
2420 2421 2422
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool wait_for_spice = false;
2423
    size_t i = 0;
2424

2425 2426 2427 2428 2429 2430 2431
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION))
        return 0;

    for (i = 0; i < vm->def->ngraphics; i++) {
        if (vm->def->graphics[i]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            wait_for_spice = true;
            break;
2432 2433
        }
    }
2434 2435 2436 2437

    if (!wait_for_spice)
        return 0;

2438 2439
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
2440 2441 2442 2443
            return -1;
    }
    return 0;
}
2444

2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496

static void
qemuMigrationUpdateJobType(qemuDomainJobInfoPtr jobInfo)
{
    switch (jobInfo->status.status) {
    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
        jobInfo->type = VIR_DOMAIN_JOB_COMPLETED;
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
        jobInfo->type = VIR_DOMAIN_JOB_NONE;
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
        jobInfo->type = VIR_DOMAIN_JOB_CANCELLED;
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLING:
        break;
    }
}


int
qemuMigrationFetchJobStatus(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
                            qemuDomainAsyncJob asyncJob,
                            qemuDomainJobInfoPtr jobInfo)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rv;

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

    memset(&jobInfo->status, 0, sizeof(jobInfo->status));
    rv = qemuMonitorGetMigrationStatus(priv->mon, &jobInfo->status);

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

    qemuMigrationUpdateJobType(jobInfo);
    return qemuDomainJobInfoUpdateTime(jobInfo);
}


2497
static int
2498
qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver,
2499
                             virDomainObjPtr vm,
2500
                             qemuDomainAsyncJob asyncJob)
2501
{
2502
    qemuDomainObjPrivatePtr priv = vm->privateData;
2503 2504
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    qemuDomainJobInfo newInfo = *jobInfo;
2505

2506 2507
    if (qemuMigrationFetchJobStatus(driver, vm, asyncJob, &newInfo) < 0)
        return -1;
2508

2509 2510 2511
    *jobInfo = newInfo;
    return 0;
}
2512

2513

2514 2515 2516 2517 2518 2519 2520 2521
static int
qemuMigrationCheckJobStatus(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
                            const char *job,
                            qemuDomainAsyncJob asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
2522

2523 2524
    if (qemuMigrationUpdateJobStatus(driver, vm, asyncJob) < 0)
        return -1;
2525

2526 2527
    switch (jobInfo->type) {
    case VIR_DOMAIN_JOB_NONE:
J
Jiri Denemark 已提交
2528 2529
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("is not active"));
2530
        return -1;
2531

2532
    case VIR_DOMAIN_JOB_FAILED:
2533 2534
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("unexpectedly failed"));
2535
        return -1;
2536

2537
    case VIR_DOMAIN_JOB_CANCELLED:
2538 2539
        virReportError(VIR_ERR_OPERATION_ABORTED,
                       _("%s: %s"), job, _("canceled by client"));
2540 2541 2542 2543 2544 2545
        return -1;

    case VIR_DOMAIN_JOB_BOUNDED:
    case VIR_DOMAIN_JOB_UNBOUNDED:
    case VIR_DOMAIN_JOB_COMPLETED:
    case VIR_DOMAIN_JOB_LAST:
2546 2547
        break;
    }
2548
    return 0;
2549 2550 2551
}


2552 2553 2554
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
2555
static int
J
Jiri Denemark 已提交
2556 2557
qemuMigrationWaitForCompletion(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
2558
                               qemuDomainAsyncJob asyncJob,
J
Jiri Denemark 已提交
2559
                               virConnectPtr dconn,
2560 2561
                               bool abort_on_error,
                               bool storage)
2562
{
2563
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
2564
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
2565
    const char *job;
2566
    int pauseReason;
2567
    int ret = -1;
2568

2569 2570
    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
2571 2572
        job = _("migration job");
        break;
2573
    case QEMU_ASYNC_JOB_SAVE:
2574 2575
        job = _("domain save job");
        break;
2576
    case QEMU_ASYNC_JOB_DUMP:
2577 2578 2579 2580 2581
        job = _("domain core dump job");
        break;
    default:
        job = _("job");
    }
2582

J
Jiri Denemark 已提交
2583
    jobInfo->type = VIR_DOMAIN_JOB_UNBOUNDED;
2584

2585
    while (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
2586 2587 2588
        /* Poll every 50ms for progress & to allow cancellation */
        struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };

2589
        if (qemuMigrationCheckJobStatus(driver, vm, job, asyncJob) < 0)
2590
            goto error;
2591

2592 2593 2594 2595
        if (storage &&
            qemuMigrationDriveMirrorReady(driver, vm) < 0)
            break;

2596
        /* cancel migration if disk I/O error is emitted while migrating */
2597
        if (abort_on_error &&
2598
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
2599 2600 2601
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("%s: %s"), job, _("failed due to I/O error"));
2602
            goto error;
2603
        }
2604

2605
        if (dconn && virConnectIsAlive(dconn) <= 0) {
2606 2607
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
2608
            goto error;
2609 2610
        }

2611 2612 2613 2614 2615
        if (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
            virObjectUnlock(vm);
            nanosleep(&ts, NULL);
            virObjectLock(vm);
        }
2616 2617
    }

2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
    if (VIR_ALLOC(priv->job.completed) == 0)
        *priv->job.completed = *jobInfo;
    return 0;

 error:
    /* Check if the migration was aborted by us rather than QEMU itself. */
    if (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED ||
        jobInfo->type == VIR_DOMAIN_JOB_COMPLETED) {
        if (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED)
            ret = -2;
J
Jiri Denemark 已提交
2630
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
2631
    }
2632
    return ret;
2633 2634 2635
}


2636
static int
2637
qemuDomainMigrateGraphicsRelocate(virQEMUDriverPtr driver,
2638
                                  virDomainObjPtr vm,
2639 2640
                                  qemuMigrationCookiePtr cookie,
                                  const char *graphicsuri)
2641 2642
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2643 2644
    int ret = -1;
    const char *listenAddress = NULL;
2645
    virSocketAddr addr;
2646 2647 2648 2649 2650
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
2651

2652
    if (!cookie || (!cookie->graphics && !graphicsuri))
2653 2654
        return 0;

2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673
    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) {
2674
        size_t i;
2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701

        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;
            }
        }
    }
2702 2703 2704 2705

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
2706 2707 2708 2709
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
2710

2711 2712 2713 2714
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
2715 2716
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
2717
    }
2718

2719
 cleanup:
2720
    virURIFree(uri);
2721 2722 2723 2724
    return ret;
}


2725
static int
2726
qemuDomainMigrateOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
2727 2728 2729
                             virDomainObjPtr vm,
                             qemuMigrationCookiePtr cookie)
{
2730 2731
    virDomainNetDefPtr netptr;
    int ret = -1;
2732
    size_t i;
2733 2734 2735 2736 2737 2738 2739 2740

    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:
2741
           break;
2742
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
2743 2744
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
2745 2746 2747
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
2748 2749 2750
                goto cleanup;
            }
            break;
2751 2752 2753 2754 2755
        default:
            break;
        }
    }

2756
    ret = 0;
2757
 cleanup:
2758 2759 2760 2761
    return ret;
}


2762 2763 2764 2765 2766 2767
/* 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.
 */
2768 2769 2770 2771
static virDomainObjPtr
qemuMigrationCleanup(virDomainObjPtr vm,
                     virConnectPtr conn,
                     void *opaque)
2772
{
2773
    virQEMUDriverPtr driver = opaque;
2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788
    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);

2789
    switch ((qemuMigrationJobPhase) priv->job.phase) {
2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819
    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 */
        ;
    }

2820
 cleanup:
2821 2822 2823
    return vm;
}

2824

2825
/* The caller is supposed to lock the vm and start a migration job. */
2826 2827 2828 2829 2830 2831 2832
static char *
qemuMigrationBeginPhase(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dname,
                        char **cookieout,
                        int *cookieoutlen,
2833 2834
                        size_t nmigrate_disks,
                        const char **migrate_disks,
2835
                        unsigned long flags)
2836 2837 2838
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
2839
    virDomainDefPtr def = NULL;
2840
    qemuDomainObjPrivatePtr priv = vm->privateData;
2841
    virCapsPtr caps = NULL;
2842
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
2843
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2844

2845
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
2846 2847
              " cookieout=%p, cookieoutlen=%p,"
              " nmigrate_disks=%zu, migrate_disks=%p, flags=%lx",
2848
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
2849 2850
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
2851

2852 2853 2854
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2855 2856 2857 2858 2859 2860
    /* 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);
2861

2862
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
2863 2864
        goto cleanup;

2865 2866
    if (!(flags & VIR_MIGRATE_UNSAFE) &&
        !qemuMigrationIsSafe(vm->def, nmigrate_disks, migrate_disks))
2867 2868
        goto cleanup;

2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912
    if (flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC)) {
        bool has_drive_mirror =  virQEMUCapsGet(priv->qemuCaps,
                                                QEMU_CAPS_DRIVE_MIRROR);

        if (nmigrate_disks) {
            if (has_drive_mirror) {
                size_t i, j;

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

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

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

        if (has_drive_mirror) {
            /* TODO support NBD for TUNNELLED migration */
            if (flags & VIR_MIGRATE_TUNNELLED) {
                VIR_WARN("NBD in tunnelled migration is currently not supported");
            } else {
                cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
                priv->nbdPort = 0;
            }
2913 2914 2915
        }
    }

2916
    if (!(mig = qemuMigrationEatCookie(driver, vm, NULL, 0, 0)))
2917 2918 2919 2920
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
                                cookieout, cookieoutlen,
2921
                                cookieFlags) < 0)
2922 2923
        goto cleanup;

L
liguang 已提交
2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945
    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;
        }
    }

2946
    if (xmlin) {
2947
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt,
2948
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE)))
2949 2950
            goto cleanup;

2951
        if (!qemuDomainDefCheckABIStability(driver, vm->def, def))
2952 2953
            goto cleanup;

2954
        rv = qemuDomainDefFormatLive(driver, def, false, true);
2955
    } else {
2956
        rv = qemuDomainDefFormatLive(driver, vm->def, false, true);
2957
    }
2958

2959
 cleanup:
2960
    qemuMigrationCookieFree(mig);
2961
    virObjectUnref(caps);
2962
    virDomainDefFree(def);
2963 2964 2965
    return rv;
}

2966 2967 2968 2969 2970 2971 2972
char *
qemuMigrationBegin(virConnectPtr conn,
                   virDomainObjPtr vm,
                   const char *xmlin,
                   const char *dname,
                   char **cookieout,
                   int *cookieoutlen,
2973 2974
                   size_t nmigrate_disks,
                   const char **migrate_disks,
2975 2976 2977 2978
                   unsigned long flags)
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
2979
    qemuDomainAsyncJob asyncJob;
2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990

    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;
    }

2991 2992
    qemuMigrationStoreDomainState(vm);

2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007
    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,
3008
                                        nmigrate_disks, migrate_disks, flags)))
3009 3010 3011 3012 3013 3014 3015
        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.
         */
3016
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
3017 3018
                                 qemuMigrationCleanup) < 0) {
            VIR_FREE(xml);
3019
            goto endjob;
3020
        }
3021
        qemuMigrationJobContinue(vm);
3022 3023 3024 3025
    } else {
        goto endjob;
    }

3026
 cleanup:
M
Michal Privoznik 已提交
3027
    virDomainObjEndAPI(&vm);
3028 3029
    return xml;

3030
 endjob:
3031 3032 3033 3034
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
3035 3036 3037
    goto cleanup;
}

3038

3039 3040
/* Prepare is the first step, and it runs on the destination host.
 */
3041

3042
static void
3043
qemuMigrationPrepareCleanup(virQEMUDriverPtr driver,
3044 3045 3046 3047 3048 3049 3050 3051 3052 3053
                            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));

3054 3055 3056
    virPortAllocatorRelease(driver->migrationPorts, priv->migrationPort);
    priv->migrationPort = 0;

3057 3058 3059 3060 3061
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

3062
static int
3063
qemuMigrationPrepareAny(virQEMUDriverPtr driver,
3064 3065 3066 3067 3068
                        virConnectPtr dconn,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
3069
                        virDomainDefPtr *def,
3070
                        const char *origname,
L
liguang 已提交
3071
                        virStreamPtr st,
3072
                        const char *protocol,
3073 3074
                        unsigned short port,
                        bool autoPort,
3075
                        const char *listenAddress,
3076 3077
                        size_t nmigrate_disks,
                        const char **migrate_disks,
L
liguang 已提交
3078
                        unsigned long flags)
3079 3080
{
    virDomainObjPtr vm = NULL;
3081
    virObjectEventPtr event = NULL;
3082
    int ret = -1;
3083
    int dataFD[2] = { -1, -1 };
3084
    qemuDomainObjPrivatePtr priv = NULL;
3085
    qemuMigrationCookiePtr mig = NULL;
3086
    bool tunnel = !!st;
J
Jiri Denemark 已提交
3087
    char *xmlout = NULL;
L
liguang 已提交
3088
    unsigned int cookieFlags;
3089
    virCapsPtr caps = NULL;
J
Ján Tomko 已提交
3090
    char *migrateFrom = NULL;
3091
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3092
    bool taint_hook = false;
3093

3094 3095
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117
    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;
        }
    }

3118 3119 3120
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

3121
    if (!qemuMigrationIsAllowed(driver, NULL, *def, true, abort_on_error))
3122 3123
        goto cleanup;

J
Jiri Denemark 已提交
3124 3125 3126 3127 3128
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

3129
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
3130 3131
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
3132 3133
            goto cleanup;

3134
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
3135 3136 3137 3138 3139 3140 3141
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
3142
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
3143 3144 3145 3146 3147 3148
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
3149
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt,
3150
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE);
J
Jiri Denemark 已提交
3151 3152 3153
                if (!newdef)
                    goto cleanup;

3154
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
3155 3156 3157 3158
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

3159 3160
                virDomainDefFree(*def);
                *def = newdef;
3161 3162 3163 3164
                /* 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
3165 3166 3167 3168
            }
        }
    }

J
Ján Tomko 已提交
3169 3170 3171 3172
    if (tunnel) {
        /* QEMU will be started with -incoming stdio
         * (which qemu_command might convert to exec:cat or fd:n)
         */
3173
        if (VIR_STRDUP(migrateFrom, "stdio") < 0)
J
Ján Tomko 已提交
3174 3175
            goto cleanup;
    } else {
3176 3177 3178
        bool encloseAddress = false;
        bool hostIPv6Capable = false;
        bool qemuIPv6Capable = false;
J
Ján Tomko 已提交
3179 3180 3181 3182
        virQEMUCapsPtr qemuCaps = NULL;
        struct addrinfo *info = NULL;
        struct addrinfo hints = { .ai_flags = AI_ADDRCONFIG,
                                  .ai_socktype = SOCK_STREAM };
3183
        const char *incFormat;
J
Ján Tomko 已提交
3184

3185 3186 3187 3188
        if (getaddrinfo("::", NULL, &hints, &info) == 0) {
            freeaddrinfo(info);
            hostIPv6Capable = true;
        }
J
Ján Tomko 已提交
3189
        if (!(qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
3190 3191
                                                    (*def)->emulator,
                                                    (*def)->os.machine)))
J
Ján Tomko 已提交
3192 3193
            goto cleanup;

3194 3195 3196 3197
        qemuIPv6Capable = virQEMUCapsGet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
        virObjectUnref(qemuCaps);

        if (listenAddress) {
3198 3199 3200 3201 3202 3203 3204 3205 3206
            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"));
3207 3208
                    goto cleanup;
                }
3209 3210
                /* IPv6 address must be escaped in brackets on the cmd line */
                encloseAddress = true;
3211
            } else {
3212
                /* listenAddress is a hostname or IPv4 */
3213
            }
3214
        } else if (qemuIPv6Capable && hostIPv6Capable) {
3215 3216 3217
            /* Listen on :: instead of 0.0.0.0 if QEMU understands it
             * and there is at least one IPv6 address configured
             */
3218 3219 3220 3221
            listenAddress = "::";
            encloseAddress = true;
        } else {
            listenAddress = "0.0.0.0";
J
Ján Tomko 已提交
3222 3223
        }

3224 3225 3226 3227
        /* 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
3228
         */
3229 3230 3231 3232 3233 3234
        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
3235 3236 3237
            goto cleanup;
    }

3238
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
3239
                                   driver->xmlopt,
3240 3241 3242
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
3243
        goto cleanup;
3244

3245
    virObjectRef(vm);
3246
    *def = NULL;
3247
    priv = vm->privateData;
3248 3249
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
3250

3251 3252 3253 3254 3255
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

3256
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
3257 3258
                                       QEMU_MIGRATION_COOKIE_LOCKSTATE |
                                       QEMU_MIGRATION_COOKIE_NBD)))
3259 3260
        goto cleanup;

3261 3262
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
3263 3264 3265 3266 3267 3268
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

3269 3270
    if (qemuMigrationPrecreateStorage(dconn, driver, vm, mig->nbd,
                                      nmigrate_disks, migrate_disks) < 0)
3271 3272
        goto cleanup;

3273
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
3274
        goto cleanup;
3275
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
3276 3277 3278 3279

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

L
liguang 已提交
3280 3281 3282
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

3283 3284
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
3285 3286
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
3287 3288 3289 3290
        goto endjob;
    }

    /* Start the QEMU daemon, with the same command-line arguments plus
3291
     * -incoming $migrateFrom
3292
     */
3293 3294
    if (qemuProcessStart(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                         migrateFrom, dataFD[0], NULL, NULL,
3295 3296
                         VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
                         VIR_QEMU_PROCESS_START_PAUSED |
3297
                         VIR_QEMU_PROCESS_START_AUTODESTROY) < 0) {
3298
        virDomainAuditStart(vm, "migrated", false);
3299 3300 3301
        goto endjob;
    }

3302 3303 3304 3305
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
3306
            goto stop;
3307
        }
3308
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
3309 3310
    }

3311 3312
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
3313 3314 3315
                                    QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stop;

3316
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
3317 3318 3319 3320
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
        goto stop;
    }

3321 3322 3323
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
                               QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
3324 3325
        goto stop;

3326 3327 3328 3329 3330 3331 3332 3333
    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");
    }
3334

3335
 done:
L
liguang 已提交
3336 3337 3338 3339 3340
    if (flags & VIR_MIGRATE_OFFLINE)
        cookieFlags = 0;
    else
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS;

3341 3342 3343
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
3344 3345
        if (qemuMigrationStartNBDServer(driver, vm, listenAddress,
                                        nmigrate_disks, migrate_disks) < 0) {
3346 3347
            /* error already reported */
            goto endjob;
3348
        }
3349
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
3350 3351 3352 3353
    }

    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
3354 3355 3356 3357 3358 3359 3360
        /* 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");
    }

3361 3362 3363
    if (qemuDomainCleanupAdd(vm, qemuMigrationPrepareCleanup) < 0)
        goto endjob;

L
liguang 已提交
3364 3365
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
3366
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
3367 3368 3369
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
3370

3371 3372 3373 3374
    /* 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.
     */
3375
    qemuMigrationJobContinue(vm);
3376

3377 3378
    if (autoPort)
        priv->migrationPort = port;
3379
    ret = 0;
3380

3381
 cleanup:
J
Ján Tomko 已提交
3382
    VIR_FREE(migrateFrom);
J
Jiri Denemark 已提交
3383
    VIR_FREE(xmlout);
3384 3385
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
3386 3387 3388
    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 */
3389
        VIR_FREE(priv->origname);
3390 3391 3392
        virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
        priv->nbdPort = 0;
        qemuDomainRemoveInactive(driver, vm);
3393
    }
M
Michal Privoznik 已提交
3394
    virDomainObjEndAPI(&vm);
3395 3396
    if (event)
        qemuDomainEventQueue(driver, event);
3397
    qemuMigrationCookieFree(mig);
3398
    virObjectUnref(caps);
3399
    virNWFilterUnlockFilterUpdates();
3400
    return ret;
3401

3402
 stop:
3403 3404 3405
    virDomainAuditStart(vm, "migrated", false);
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, 0);

3406
 endjob:
3407
    qemuMigrationJobFinish(driver, vm);
3408
    goto cleanup;
3409 3410 3411
}


3412 3413 3414 3415 3416
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
3417
qemuMigrationPrepareTunnel(virQEMUDriverPtr driver,
3418 3419 3420 3421 3422 3423
                           virConnectPtr dconn,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virStreamPtr st,
3424
                           virDomainDefPtr *def,
3425
                           const char *origname,
L
liguang 已提交
3426
                           unsigned long flags)
3427 3428 3429 3430
{
    int ret;

    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
3431
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
3432
              "origname=%s, flags=%lx",
3433
              driver, dconn, NULLSTR(cookiein), cookieinlen,
3434
              cookieout, cookieoutlen, st, *def, origname, flags);
3435

3436 3437 3438 3439 3440 3441
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

3442
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3443
                                  cookieout, cookieoutlen, def, origname,
3444
                                  st, NULL, 0, false, NULL, 0, NULL, flags);
3445 3446 3447 3448
    return ret;
}


3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471
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;
}


3472
int
3473
qemuMigrationPrepareDirect(virQEMUDriverPtr driver,
3474
                           virConnectPtr dconn,
3475 3476 3477 3478
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
3479 3480
                           const char *uri_in,
                           char **uri_out,
3481
                           virDomainDefPtr *def,
3482
                           const char *origname,
3483
                           const char *listenAddress,
3484 3485
                           size_t nmigrate_disks,
                           const char **migrate_disks,
L
liguang 已提交
3486
                           unsigned long flags)
3487
{
3488 3489
    unsigned short port = 0;
    bool autoPort = true;
3490 3491
    char *hostname = NULL;
    int ret = -1;
3492
    virURIPtr uri = NULL;
3493 3494
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
3495

3496 3497
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
3498 3499
              "def=%p, origname=%s, listenAddress=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p, flags=%lx",
3500 3501
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
3502 3503
              *def, origname, NULLSTR(listenAddress),
              nmigrate_disks, migrate_disks, flags);
3504

3505 3506
    *uri_out = NULL;

3507 3508 3509
    /* 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
3510 3511 3512
     * 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".
3513 3514 3515 3516 3517 3518
     *
     * 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) {
3519 3520 3521
        bool encloseAddress = false;
        const char *incFormat;

3522
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3523
            goto cleanup;
3524

3525
        if (migrateHost != NULL) {
3526 3527
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
3528

3529
            if (VIR_STRDUP(hostname, migrateHost) < 0)
3530 3531 3532 3533 3534
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
3535 3536

        if (STRPREFIX(hostname, "localhost")) {
3537 3538 3539
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
3540 3541 3542 3543 3544
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
3545
         * compatibility with old targets. We at least make the
3546 3547
         * new targets accept both syntaxes though.
         */
3548 3549 3550 3551 3552 3553
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
3554 3555
            goto cleanup;
    } else {
3556
        bool well_formed_uri;
J
Ján Tomko 已提交
3557

3558 3559
        if (!(uri = qemuMigrationParseURI(uri_in, &well_formed_uri)))
            goto cleanup;
J
Ján Tomko 已提交
3560

3561 3562 3563 3564 3565 3566 3567
        if (uri->scheme == NULL) {
            virReportError(VIR_ERR_INVALID_ARG,
                           _("missing scheme in migration URI: %s"),
                           uri_in);
            goto cleanup;
        }

M
Michael R. Hines 已提交
3568 3569
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
3570 3571 3572
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
3573 3574 3575 3576 3577 3578 3579 3580 3581 3582
            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) {
3583
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3584
                goto cleanup;
3585

3586
            /* Send well-formed URI only if uri_in was well-formed */
3587 3588 3589 3590 3591 3592 3593 3594
            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;
            }
3595
        } else {
3596 3597
            port = uri->port;
            autoPort = false;
3598 3599 3600 3601 3602 3603
        }
    }

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

3604
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3605
                                  cookieout, cookieoutlen, def, origname,
3606
                                  NULL, uri ? uri->scheme : "tcp",
3607 3608
                                  port, autoPort, listenAddress,
                                  nmigrate_disks, migrate_disks, flags);
3609
 cleanup:
3610
    virURIFree(uri);
3611
    VIR_FREE(hostname);
3612
    virObjectUnref(cfg);
3613
    if (ret != 0) {
3614
        VIR_FREE(*uri_out);
3615 3616 3617
        if (autoPort)
            virPortAllocatorRelease(driver->migrationPorts, port);
    }
3618 3619 3620 3621
    return ret;
}


3622 3623 3624
virDomainDefPtr
qemuMigrationPrepareDef(virQEMUDriverPtr driver,
                        const char *dom_xml,
3625 3626
                        const char *dname,
                        char **origname)
3627 3628 3629
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
3630
    char *name = NULL;
3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641

    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,
3642
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE)))
3643 3644 3645
        goto cleanup;

    if (dname) {
3646
        name = def->name;
3647 3648 3649 3650 3651 3652
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

3653
 cleanup:
3654
    virObjectUnref(caps);
3655 3656 3657 3658
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
3659 3660 3661 3662
    return def;
}


3663 3664 3665 3666 3667 3668 3669 3670 3671 3672
static int
qemuMigrationConfirmPhase(virQEMUDriverPtr driver,
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *cookiein,
                          int cookieinlen,
                          unsigned int flags,
                          int retcode)
{
    qemuMigrationCookiePtr mig;
3673
    virObjectEventPtr event = NULL;
3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688
    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);

3689 3690
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                       QEMU_MIGRATION_COOKIE_STATS)))
3691 3692
        goto cleanup;

3693 3694 3695
    /* Update total times with the values sent by the destination daemon */
    if (mig->jobInfo) {
        qemuDomainObjPrivatePtr priv = vm->privateData;
3696 3697 3698
        VIR_FREE(priv->job.completed);
        priv->job.completed = mig->jobInfo;
        mig->jobInfo = NULL;
3699 3700
    }

3701 3702 3703 3704 3705 3706 3707 3708 3709
    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 */
3710
        qemuMigrationWaitForSpice(vm);
3711 3712 3713 3714 3715

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

3716
        event = virDomainEventLifecycleNewFromObj(vm,
3717 3718 3719
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
    } else {
3720
        virErrorPtr orig_err = virSaveLastError();
3721 3722

        /* cancel any outstanding NBD jobs */
3723 3724
        qemuMigrationCancelDriveMirror(driver, vm, false,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT);
3725 3726 3727

        virSetError(orig_err);
        virFreeError(orig_err);
3728

3729 3730 3731 3732
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
3733 3734 3735 3736 3737 3738 3739 3740
        }

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

3741
 done:
3742 3743 3744
    qemuMigrationCookieFree(mig);
    rv = 0;

3745
 cleanup:
3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760
    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;
3761
    qemuMigrationJobPhase phase;
3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775
    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);
3776 3777
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
3778 3779 3780 3781 3782

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

3783 3784 3785
    qemuMigrationJobFinish(driver, vm);
    if (!virDomainObjIsActive(vm) &&
        (!vm->persistent || (flags & VIR_MIGRATE_UNDEFINE_SOURCE))) {
3786 3787 3788 3789 3790
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
        qemuDomainRemoveInactive(driver, vm);
    }

3791
 cleanup:
M
Michal Privoznik 已提交
3792
    virDomainObjEndAPI(&vm);
3793 3794 3795 3796 3797
    virObjectUnref(cfg);
    return ret;
}


3798 3799
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3800
    MIGRATION_DEST_CONNECT_HOST,
3801
    MIGRATION_DEST_UNIX,
3802
    MIGRATION_DEST_FD,
3803
};
3804

3805 3806 3807 3808
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3809

3810 3811 3812 3813 3814 3815
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3816
            const char *protocol;
3817 3818 3819 3820 3821
            const char *name;
            int port;
        } host;

        struct {
3822
            char *file;
3823 3824
            int sock;
        } unix_socket;
3825 3826 3827 3828 3829

        struct {
            int qemu;
            int local;
        } fd;
3830 3831 3832 3833 3834 3835 3836
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3837 3838 3839

#define TUNNEL_SEND_BUF_SIZE 65536

3840 3841 3842 3843 3844 3845 3846
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3847 3848
    int wakeupRecvFD;
    int wakeupSendFD;
3849 3850 3851
};

static void qemuMigrationIOFunc(void *arg)
3852
{
3853
    qemuMigrationIOThreadPtr data = arg;
3854 3855 3856 3857 3858 3859 3860
    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);
3861

3862
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3863
        goto abrt;
3864

3865 3866 3867
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3868
    for (;;) {
3869 3870 3871 3872 3873 3874 3875 3876 3877 3878
        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;
3879
            virReportSystemError(errno, "%s",
3880 3881
                                 _("poll failed in migration tunnel"));
            goto abrt;
3882
        }
3883 3884 3885 3886 3887 3888 3889

        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");
3890
            break;
3891
        }
3892

3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908
        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;
            }
3909 3910
        }

3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927
        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;
            }
        }
    }
3928

3929 3930
    if (virStreamFinish(data->st) < 0)
        goto error;
3931

3932 3933
    VIR_FREE(buffer);

3934 3935
    return;

3936
 abrt:
3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3948
 error:
3949 3950
    virCopyLastError(&data->err);
    virResetLastError();
3951
    VIR_FREE(buffer);
3952 3953 3954 3955 3956 3957 3958
}


static qemuMigrationIOThreadPtr
qemuMigrationStartTunnel(virStreamPtr st,
                         int sock)
{
3959 3960
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3961

3962 3963 3964 3965
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3966 3967
    }

3968
    if (VIR_ALLOC(io) < 0)
3969
        goto error;
3970

3971 3972
    io->st = st;
    io->sock = sock;
3973 3974
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3975 3976 3977 3978 3979 3980

    if (virThreadCreate(&io->thread, true,
                        qemuMigrationIOFunc,
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3981
        goto error;
3982 3983 3984
    }

    return io;
3985

3986
 error:
3987 3988 3989 3990
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3991 3992 3993
}

static int
3994
qemuMigrationStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3995 3996
{
    int rv = -1;
3997 3998 3999 4000 4001 4002 4003 4004 4005
    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;
    }

4006 4007 4008 4009
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
4010 4011 4012 4013
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
4014 4015 4016 4017 4018 4019
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

4020
 cleanup:
4021 4022
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
4023 4024
    VIR_FREE(io);
    return rv;
4025 4026
}

4027
static int
4028
qemuMigrationConnect(virQEMUDriverPtr driver,
4029 4030 4031 4032 4033 4034 4035 4036 4037
                     virDomainObjPtr vm,
                     qemuMigrationSpecPtr spec)
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
4038
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
4039 4040 4041 4042 4043 4044 4045
        return -1;

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

    if (virSecurityManagerSetSocketLabel(driver->securityManager, vm->def) < 0)
        goto cleanup;
4046 4047 4048
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
4049
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
4050
        virObjectUnref(sock);
4051 4052 4053 4054 4055
    }
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0 ||
        spec->dest.fd.qemu == -1)
        goto cleanup;

4056 4057 4058 4059 4060 4061 4062
    /* 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;
    }

4063 4064
    ret = 0;

4065
 cleanup:
4066 4067 4068 4069 4070 4071
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

4072
static int
4073
qemuMigrationRun(virQEMUDriverPtr driver,
4074 4075 4076 4077 4078 4079 4080
                 virDomainObjPtr vm,
                 const char *cookiein,
                 int cookieinlen,
                 char **cookieout,
                 int *cookieoutlen,
                 unsigned long flags,
                 unsigned long resource,
4081
                 qemuMigrationSpecPtr spec,
4082
                 virConnectPtr dconn,
4083 4084 4085
                 const char *graphicsuri,
                 size_t nmigrate_disks,
                 const char **migrate_disks)
4086
{
4087
    int ret = -1;
4088 4089
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
4090
    qemuMigrationCookiePtr mig = NULL;
4091
    qemuMigrationIOThreadPtr iothread = NULL;
4092
    int fd = -1;
4093
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
4094
    virErrorPtr orig_err = NULL;
4095
    unsigned int cookieFlags = 0;
4096
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
4097
    int rc;
4098 4099 4100

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
4101 4102
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
4103 4104
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
4105
              spec, spec->destType, spec->fwdType, dconn,
4106
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4107

4108 4109 4110 4111 4112 4113 4114 4115 4116 4117
    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;
    }

4118 4119
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
4120 4121 4122 4123
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
4124 4125 4126
        return -1;
    }

4127 4128 4129
    mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                 cookieFlags | QEMU_MIGRATION_COOKIE_GRAPHICS);
    if (!mig)
4130 4131
        goto cleanup;

4132
    if (qemuDomainMigrateGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
4133 4134
        VIR_WARN("unable to provide data for graphics client relocation");

4135 4136 4137 4138 4139 4140 4141
    if (migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                         QEMU_MONITOR_MIGRATE_NON_SHARED_INC)) {
        if (mig->nbd) {
            /* This will update migrate_flags on success */
            if (qemuMigrationDriveMirror(driver, vm, mig,
                                         spec->dest.host.name,
                                         migrate_speed,
4142 4143 4144
                                         &migrate_flags,
                                         nmigrate_disks,
                                         migrate_disks) < 0) {
4145 4146 4147 4148 4149 4150 4151 4152
                goto cleanup;
            }
        } else {
            /* Destination doesn't support NBD server.
             * Fall back to previous implementation. */
            VIR_DEBUG("Destination doesn't support NBD server "
                      "Falling back to previous implementation.");
        }
4153 4154
    }

4155
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
4156 4157 4158 4159 4160 4161
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

4162 4163
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
4164 4165 4166
                                    QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

4167 4168
    if (qemuMigrationSetAutoConverge(driver, vm,
                                     flags & VIR_MIGRATE_AUTO_CONVERGE,
4169 4170 4171
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

4172 4173
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
4174 4175 4176
                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

4177 4178
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4179 4180
        goto cleanup;

4181
    if (priv->job.abortJob) {
4182 4183
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
4184
         * priv->job.abortJob will not change */
4185
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
4186
        priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
4187 4188 4189 4190 4191 4192
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
        goto cleanup;
    }

4193 4194
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
4195

4196 4197
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
4198
        qemuMigrationConnect(driver, vm, spec) < 0) {
4199
        goto exit_monitor;
4200
    }
4201

4202 4203
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
4204 4205
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
4206
            goto exit_monitor;
M
Michael R. Hines 已提交
4207
        }
4208
        ret = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
4209
                                       spec->dest.host.protocol,
4210 4211 4212 4213
                                       spec->dest.host.name,
                                       spec->dest.host.port);
        break;

4214 4215 4216 4217
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

4218
    case MIGRATION_DEST_UNIX:
4219
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX)) {
4220 4221 4222 4223 4224 4225 4226 4227 4228
            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;
4229 4230

    case MIGRATION_DEST_FD:
4231
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
4232
            fd = spec->dest.fd.local;
4233 4234
            spec->dest.fd.local = -1;
        }
4235 4236 4237 4238
        ret = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                     spec->dest.fd.qemu);
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
4239
    }
4240 4241
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
4242
    if (ret < 0)
4243 4244
        goto cleanup;
    ret = -1;
4245 4246 4247 4248

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

4249 4250 4251 4252 4253
    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.
         */
4254 4255
        if (qemuMigrationCheckJobStatus(driver, vm, _("migration job"),
                                        QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4256
            goto cancel;
4257

4258 4259 4260 4261 4262 4263 4264
        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;
        }
4265 4266
    }

4267 4268
    if (spec->fwdType != MIGRATION_FWD_DIRECT &&
        !(iothread = qemuMigrationStartTunnel(spec->fwd.stream, fd)))
4269 4270
        goto cancel;

4271 4272
    rc = qemuMigrationWaitForCompletion(driver, vm,
                                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4273
                                        dconn, abort_on_error, !!mig->nbd);
4274 4275 4276
    if (rc == -2)
        goto cancel;
    else if (rc == -1)
4277
        goto cleanup;
4278

4279 4280 4281 4282 4283 4284 4285 4286 4287
    /* 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) {
4288 4289
        if (qemuMigrationSetOffline(driver, vm) < 0) {
            priv->job.current->type = VIR_DOMAIN_JOB_FAILED;
4290
            goto cleanup;
4291
        }
4292 4293
    }

4294
    ret = 0;
4295

4296
 cleanup:
4297 4298 4299
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

4300
    /* cancel any outstanding NBD jobs */
4301
    if (mig && mig->nbd) {
4302 4303
        if (qemuMigrationCancelDriveMirror(driver, vm, ret == 0,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4304 4305
            ret = -1;
    }
4306

4307
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
4308
        if (iothread && qemuMigrationStopTunnel(iothread, ret < 0) < 0)
4309
            ret = -1;
4310
        VIR_FORCE_CLOSE(fd);
4311
    }
4312

4313
    if (priv->job.completed) {
4314
        qemuDomainJobInfoUpdateTime(priv->job.completed);
4315
        qemuDomainJobInfoUpdateDowntime(priv->job.completed);
4316
        ignore_value(virTimeMillisNow(&priv->job.completed->sent));
4317
    }
4318

4319 4320 4321
    if (priv->job.current->type == VIR_DOMAIN_JOB_UNBOUNDED)
        priv->job.current->type = VIR_DOMAIN_JOB_FAILED;

4322 4323
    cookieFlags |= QEMU_MIGRATION_COOKIE_NETWORK |
                   QEMU_MIGRATION_COOKIE_STATS;
4324 4325
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookieFlags |= QEMU_MIGRATION_COOKIE_PERSISTENT;
4326
    if (ret == 0 &&
4327 4328
        qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
4329
        VIR_WARN("Unable to encode migration cookie");
4330
    }
4331

4332 4333
    qemuMigrationCookieFree(mig);

4334 4335 4336 4337 4338
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4339 4340
    return ret;

4341 4342 4343 4344
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;

4345
 cancel:
4346 4347
    orig_err = virSaveLastError();

4348
    if (virDomainObjIsActive(vm)) {
4349 4350
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
4351
            qemuMonitorMigrateCancel(priv->mon);
4352
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
4353
        }
4354
    }
4355 4356 4357
    goto cleanup;
}

4358
/* Perform migration using QEMU's native migrate support,
4359 4360
 * not encrypted obviously
 */
4361
static int doNativeMigrate(virQEMUDriverPtr driver,
4362 4363 4364 4365 4366 4367 4368
                           virDomainObjPtr vm,
                           const char *uri,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4369
                           unsigned long resource,
4370
                           virConnectPtr dconn,
4371 4372 4373
                           const char *graphicsuri,
                           size_t nmigrate_disks,
                           const char **migrate_disks)
4374
{
4375
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
4376
    virURIPtr uribits = NULL;
4377
    int ret = -1;
4378 4379 4380
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
4381
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
4382
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
4383
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
4384
              cookieout, cookieoutlen, flags, resource,
4385
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4386

4387
    if (!(uribits = qemuMigrationParseURI(uri, NULL)))
4388 4389
        return -1;

4390 4391 4392 4393 4394 4395 4396
    if (uribits->scheme == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("missing scheme in migration URI: %s"),
                       uri);
        goto cleanup;
    }

M
Michael R. Hines 已提交
4397 4398 4399 4400 4401
    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"));
4402
            goto cleanup;
M
Michael R. Hines 已提交
4403
        }
4404
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
4405 4406 4407
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
4408
            goto cleanup;
M
Michael R. Hines 已提交
4409 4410 4411 4412 4413
        }
    }

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
        STRNEQ(uribits->scheme, "rdma"))
4414 4415
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
4416
        spec.destType = MIGRATION_DEST_HOST;
4417
    spec.dest.host.protocol = uribits->scheme;
4418 4419 4420
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
4421

4422
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4423
                           cookieoutlen, flags, resource, &spec, dconn,
4424
                           graphicsuri, nmigrate_disks, migrate_disks);
4425 4426 4427 4428

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

4429
 cleanup:
4430
    virURIFree(uribits);
4431 4432 4433 4434 4435

    return ret;
}


4436
static int doTunnelMigrate(virQEMUDriverPtr driver,
4437 4438 4439 4440 4441 4442 4443
                           virDomainObjPtr vm,
                           virStreamPtr st,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4444
                           unsigned long resource,
4445
                           virConnectPtr dconn,
4446 4447 4448
                           const char *graphicsuri,
                           size_t nmigrate_disks,
                           const char **migrate_disks)
4449 4450
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
4451
    virNetSocketPtr sock = NULL;
4452 4453
    int ret = -1;
    qemuMigrationSpec spec;
4454
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4455 4456

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
4457
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
4458
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
4459
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
4460
              cookieout, cookieoutlen, flags, resource,
4461
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4462

4463 4464 4465
    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)) {
4466 4467
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Source qemu is too old to support tunnelled migration"));
4468
        virObjectUnref(cfg);
4469
        return -1;
4470 4471 4472 4473 4474
    }

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

4475
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD)) {
4476 4477 4478 4479 4480 4481
        int fds[2];

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

4482
        if (pipe2(fds, O_CLOEXEC) == 0) {
4483 4484 4485 4486
            spec.dest.fd.qemu = fds[1];
            spec.dest.fd.local = fds[0];
        }
        if (spec.dest.fd.qemu == -1 ||
4487
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
4488
                                              spec.dest.fd.qemu) < 0) {
4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499
            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",
4500
                        cfg->libDir, vm->def->name) < 0)
4501 4502 4503
            goto cleanup;

        if (virNetSocketNewListenUNIX(spec.dest.unix_socket.file, 0700,
4504
                                      cfg->user, cfg->group,
4505 4506 4507 4508 4509 4510 4511
                                      &sock) < 0 ||
            virNetSocketListen(sock, 1) < 0)
            goto cleanup;

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

4512
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4513
                           cookieoutlen, flags, resource, &spec, dconn,
4514
                           graphicsuri, nmigrate_disks, migrate_disks);
4515

4516
 cleanup:
4517 4518 4519 4520
    if (spec.destType == MIGRATION_DEST_FD) {
        VIR_FORCE_CLOSE(spec.dest.fd.qemu);
        VIR_FORCE_CLOSE(spec.dest.fd.local);
    } else {
4521
        virObjectUnref(sock);
4522 4523
        VIR_FREE(spec.dest.unix_socket.file);
    }
4524

4525
    virObjectUnref(cfg);
4526 4527 4528 4529
    return ret;
}


4530 4531 4532 4533
/* 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 */
4534
static int doPeer2PeerMigrate2(virQEMUDriverPtr driver,
4535
                               virConnectPtr sconn ATTRIBUTE_UNUSED,
4536 4537
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
4538
                               const char *dconnuri,
4539 4540 4541
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
4542 4543 4544
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
4545
    char *cookie = NULL;
4546 4547 4548
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
4549
    bool cancelled;
4550
    virStreamPtr st = NULL;
4551
    unsigned long destflags;
4552

4553
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
4554
              "flags=%lx, dname=%s, resource=%lu",
4555 4556
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
4557

4558 4559 4560 4561 4562
    /* 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,
4563 4564
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
4565 4566 4567 4568 4569
        return -1;

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

4570 4571
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4572

4573 4574 4575 4576 4577 4578 4579 4580 4581 4582
    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;

4583
        qemuDomainObjEnterRemote(vm);
4584
        ret = dconn->driver->domainMigratePrepareTunnel
4585
            (dconn, st, destflags, dname, resource, dom_xml);
4586
        qemuDomainObjExitRemote(vm);
4587
    } else {
4588
        qemuDomainObjEnterRemote(vm);
4589 4590
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
4591
             destflags, dname, resource, dom_xml);
4592
        qemuDomainObjExitRemote(vm);
4593 4594 4595
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
4596 4597 4598
        goto cleanup;

    /* the domain may have shutdown or crashed while we had the locks dropped
4599
     * in qemuDomainObjEnterRemote, so check again
4600 4601
     */
    if (!virDomainObjIsActive(vm)) {
4602 4603
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4604 4605 4606
        goto cleanup;
    }

4607 4608
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
4609 4610
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
4611
        cancelled = true;
4612
        orig_err = virSaveLastError();
4613
        goto finish;
4614 4615
    }

4616 4617 4618 4619
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
4620
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4621
    if (flags & VIR_MIGRATE_TUNNELLED)
4622 4623
        ret = doTunnelMigrate(driver, vm, st,
                              NULL, 0, NULL, NULL,
4624 4625
                              flags, resource, dconn,
                              NULL, 0, NULL);
4626 4627 4628 4629
    else
        ret = doNativeMigrate(driver, vm, uri_out,
                              cookie, cookielen,
                              NULL, NULL, /* No out cookie with v2 migration */
4630
                              flags, resource, dconn, NULL, 0, NULL);
4631 4632 4633 4634

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4635

4636 4637 4638
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4639
    cancelled = ret < 0;
4640

4641
 finish:
4642 4643 4644 4645
    /* 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.
     */
4646
    dname = dname ? dname : vm->def->name;
4647
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4648
    qemuDomainObjEnterRemote(vm);
4649
    ddomain = dconn->driver->domainMigrateFinish2
4650
        (dconn, dname, cookie, cookielen,
4651
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4652
    qemuDomainObjExitRemote(vm);
4653 4654
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4655

4656
 cleanup:
4657
    if (ddomain) {
4658
        virObjectUnref(ddomain);
4659 4660 4661 4662
        ret = 0;
    } else {
        ret = -1;
    }
4663

4664
    virObjectUnref(st);
4665 4666 4667 4668 4669 4670

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4671
    VIR_FREE(cookie);
4672 4673

    return ret;
4674 4675 4676
}


4677 4678 4679 4680
/* 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 */
4681 4682 4683 4684 4685 4686 4687 4688 4689
static int
doPeer2PeerMigrate3(virQEMUDriverPtr driver,
                    virConnectPtr sconn,
                    virConnectPtr dconn,
                    const char *dconnuri,
                    virDomainObjPtr vm,
                    const char *xmlin,
                    const char *dname,
                    const char *uri,
4690
                    const char *graphicsuri,
4691
                    const char *listenAddress,
4692 4693
                    size_t nmigrate_disks,
                    const char **migrate_disks,
4694 4695 4696
                    unsigned long long bandwidth,
                    bool useParams,
                    unsigned long flags)
4697 4698 4699 4700 4701 4702 4703 4704 4705 4706
{
    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;
4707
    bool cancelled = true;
4708
    virStreamPtr st = NULL;
4709
    unsigned long destflags;
4710 4711 4712
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4713
    size_t i;
4714 4715

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4716
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4717 4718
              "nmigrate_disks=%zu, migrate_disks=%p, bandwidth=%llu, "
              "useParams=%d, flags=%lx",
4719
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4720
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4721 4722
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks,
              bandwidth, useParams, flags);
4723

4724 4725 4726 4727 4728
    /* 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.  */

4729
    dom_xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
4730 4731
                                      &cookieout, &cookieoutlen,
                                      nmigrate_disks, migrate_disks, flags);
4732 4733 4734
    if (!dom_xml)
        goto cleanup;

4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754
    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;
4755 4756 4757 4758 4759 4760

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4761 4762 4763 4764 4765
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4766 4767 4768 4769 4770
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4771 4772
    }

4773 4774 4775
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4776 4777
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4778

4779 4780 4781 4782 4783 4784 4785 4786 4787
    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;

4788
        qemuDomainObjEnterRemote(vm);
4789 4790 4791 4792 4793 4794 4795 4796 4797
        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);
        }
4798
        qemuDomainObjExitRemote(vm);
4799
    } else {
4800
        qemuDomainObjEnterRemote(vm);
4801 4802 4803 4804 4805 4806 4807 4808 4809
        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);
        }
4810
        qemuDomainObjExitRemote(vm);
4811 4812 4813 4814 4815
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

L
liguang 已提交
4816 4817 4818 4819
    if (flags & VIR_MIGRATE_OFFLINE) {
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4820
        cancelled = false;
L
liguang 已提交
4821 4822 4823
        goto finish;
    }

4824 4825 4826 4827
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4828 4829
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4830
            goto finish;
4831
        }
4832
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4833 4834
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4835
        orig_err = virSaveLastError();
4836 4837 4838 4839 4840 4841 4842 4843
        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.
     */
4844
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4845
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4846 4847 4848 4849 4850
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4851
    if (flags & VIR_MIGRATE_TUNNELLED) {
4852 4853 4854
        ret = doTunnelMigrate(driver, vm, st,
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4855 4856
                              flags, bandwidth, dconn, graphicsuri,
                              nmigrate_disks, migrate_disks);
4857 4858
    } else {
        ret = doNativeMigrate(driver, vm, uri,
4859 4860
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4861 4862
                              flags, bandwidth, dconn, graphicsuri,
                              nmigrate_disks, migrate_disks);
4863
    }
4864 4865

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4866
    if (ret < 0) {
4867
        orig_err = virSaveLastError();
4868 4869 4870 4871
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4872 4873 4874 4875

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4876
    cancelled = ret < 0;
4877

4878
 finish:
4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890
    /*
     * 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;
4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913

    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);
    }
4914 4915
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4916

4917 4918 4919 4920 4921 4922 4923
    /* 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.
4924
     */
4925
    cancelled = ddomain == NULL;
4926

4927 4928 4929 4930 4931 4932
    /* 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();

4933 4934 4935 4936
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4937
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4938 4939 4940 4941 4942
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4943 4944 4945
    ret = qemuMigrationConfirmPhase(driver, sconn, vm,
                                    cookiein, cookieinlen,
                                    flags, cancelled);
4946 4947 4948 4949
    /* 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.
     */
4950 4951 4952
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4953 4954 4955

 cleanup:
    if (ddomain) {
4956
        virObjectUnref(ddomain);
4957 4958 4959 4960 4961
        ret = 0;
    } else {
        ret = -1;
    }

4962
    virObjectUnref(st);
4963 4964 4965 4966 4967 4968 4969 4970

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4971
    virTypedParamsFree(params, nparams);
4972 4973 4974 4975
    return ret;
}


4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4988
static int doPeer2PeerMigrate(virQEMUDriverPtr driver,
4989
                              virConnectPtr sconn,
4990
                              virDomainObjPtr vm,
4991
                              const char *xmlin,
4992
                              const char *dconnuri,
4993
                              const char *uri,
4994
                              const char *graphicsuri,
4995
                              const char *listenAddress,
4996 4997
                              size_t nmigrate_disks,
                              const char **migrate_disks,
4998 4999
                              unsigned long flags,
                              const char *dname,
5000 5001
                              unsigned long resource,
                              bool *v3proto)
5002 5003 5004 5005
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
5006
    virErrorPtr orig_err = NULL;
5007
    bool offline = false;
5008
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5009
    bool useParams;
5010

5011 5012 5013
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, uri=%s, "
              "graphicsuri=%s, listenAddress=%s, nmigrate_disks=%zu, "
              "migrate_disks=%p, flags=%lx, dname=%s, resource=%lu",
5014
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
5015
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
5016
              nmigrate_disks, migrate_disks, flags, NULLSTR(dname), resource);
5017

5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031
    if (flags & VIR_MIGRATE_TUNNELLED && uri) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("migration URI is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

    if (flags & VIR_MIGRATE_TUNNELLED && listenAddress) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("listen address is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

5032 5033 5034 5035
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

5036
    qemuDomainObjEnterRemote(vm);
5037
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
5038
    qemuDomainObjExitRemote(vm);
5039
    if (dconn == NULL) {
5040
        virReportError(VIR_ERR_OPERATION_FAILED,
5041 5042
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
5043
        virObjectUnref(cfg);
5044 5045 5046
        return -1;
    }

5047 5048
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
5049 5050
        goto cleanup;

5051
    qemuDomainObjEnterRemote(vm);
5052 5053
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
5054
        /* v3proto reflects whether the caller used Perform3, but with
5055
         * p2p migrate, regardless of whether Perform2 or Perform3
5056 5057 5058 5059
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
5060 5061
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
L
liguang 已提交
5062 5063 5064
    if (flags & VIR_MIGRATE_OFFLINE)
        offline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                           VIR_DRV_FEATURE_MIGRATION_OFFLINE);
5065
    qemuDomainObjExitRemote(vm);
5066

5067
    if (!p2p) {
5068 5069
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
5070 5071 5072
        goto cleanup;
    }

5073 5074
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
5075
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
5076 5077 5078 5079 5080 5081
        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
5082 5083 5084 5085 5086 5087 5088
    if (flags & VIR_MIGRATE_OFFLINE && !offline) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

5089
    /* domain may have been stopped while we were talking to remote daemon */
L
liguang 已提交
5090
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
5091 5092
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
5093 5094 5095
        goto cleanup;
    }

5096 5097 5098 5099 5100 5101 5102
    /* 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;

5103 5104
    if (*v3proto) {
        ret = doPeer2PeerMigrate3(driver, sconn, dconn, dconnuri, vm, xmlin,
5105
                                  dname, uri, graphicsuri, listenAddress,
5106 5107
                                  nmigrate_disks, migrate_disks, resource,
                                  useParams, flags);
5108
    } else {
5109
        ret = doPeer2PeerMigrate2(driver, sconn, dconn, vm,
5110
                                  dconnuri, flags, dname, resource);
5111
    }
5112

5113
 cleanup:
5114
    orig_err = virSaveLastError();
5115
    qemuDomainObjEnterRemote(vm);
5116
    virObjectUnref(dconn);
5117
    qemuDomainObjExitRemote(vm);
5118 5119 5120 5121
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5122
    virObjectUnref(cfg);
5123 5124 5125 5126
    return ret;
}


5127 5128 5129 5130 5131 5132
/*
 * 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
5133
qemuMigrationPerformJob(virQEMUDriverPtr driver,
5134 5135 5136 5137 5138
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dconnuri,
                        const char *uri,
5139
                        const char *graphicsuri,
5140
                        const char *listenAddress,
5141 5142
                        size_t nmigrate_disks,
                        const char **migrate_disks,
5143
                        const char *cookiein,
5144 5145 5146 5147 5148 5149 5150
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
5151
{
5152
    virObjectEventPtr event = NULL;
5153
    int ret = -1;
5154
    virErrorPtr orig_err = NULL;
5155
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5156
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
5157

5158
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
5159 5160
        goto cleanup;

L
liguang 已提交
5161
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
5162 5163
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
5164 5165 5166
        goto endjob;
    }

5167
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
5168
        goto endjob;
5169

5170 5171
    if (!(flags & VIR_MIGRATE_UNSAFE) &&
        !qemuMigrationIsSafe(vm->def, nmigrate_disks, migrate_disks))
5172
        goto endjob;
5173

5174
    qemuMigrationStoreDomainState(vm);
5175 5176

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
5177
        ret = doPeer2PeerMigrate(driver, conn, vm, xmlin,
5178
                                 dconnuri, uri, graphicsuri, listenAddress,
5179
                                 nmigrate_disks, migrate_disks,
5180
                                 flags, dname, resource, &v3proto);
5181
    } else {
5182 5183 5184
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
        ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                              cookieout, cookieoutlen,
5185
                              flags, resource, NULL, NULL, 0, NULL);
5186
    }
5187 5188
    if (ret < 0)
        goto endjob;
5189

5190 5191 5192 5193
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
5194
    if (!v3proto) {
5195 5196
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
5197
        virDomainAuditStop(vm, "migrated");
5198
        event = virDomainEventLifecycleNewFromObj(vm,
5199 5200
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
5201 5202
    }

5203
 endjob:
5204 5205 5206
    if (ret < 0)
        orig_err = virSaveLastError();

5207
    if (qemuMigrationRestoreDomainState(conn, vm)) {
5208
        event = virDomainEventLifecycleNewFromObj(vm,
5209 5210 5211
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
5212

5213 5214 5215 5216
    qemuMigrationJobFinish(driver, vm);
    if (!virDomainObjIsActive(vm) &&
        (!vm->persistent ||
         (ret == 0 && (flags & VIR_MIGRATE_UNDEFINE_SOURCE)))) {
5217
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
5218
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
5219
        qemuDomainRemoveInactive(driver, vm);
5220 5221
    }

5222 5223 5224 5225 5226
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

5227
 cleanup:
M
Michal Privoznik 已提交
5228
    virDomainObjEndAPI(&vm);
5229 5230
    if (event)
        qemuDomainEventQueue(driver, event);
5231
    virObjectUnref(cfg);
5232 5233 5234 5235 5236 5237 5238
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
5239
qemuMigrationPerformPhase(virQEMUDriverPtr driver,
5240 5241 5242
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *uri,
5243
                          const char *graphicsuri,
5244 5245
                          size_t nmigrate_disks,
                          const char **migrate_disks,
5246 5247 5248 5249 5250 5251 5252
                          const char *cookiein,
                          int cookieinlen,
                          char **cookieout,
                          int *cookieoutlen,
                          unsigned long flags,
                          unsigned long resource)
{
5253
    virObjectEventPtr event = NULL;
5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264
    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);
5265 5266
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
5267 5268 5269

    ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                          cookieout, cookieoutlen,
5270 5271
                          flags, resource, NULL, graphicsuri,
                          nmigrate_disks, migrate_disks);
5272

5273 5274 5275 5276 5277
    if (ret < 0) {
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
5278
        }
5279
        goto endjob;
5280
    }
5281 5282 5283

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

5284 5285
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                             qemuMigrationCleanup) < 0)
5286 5287
        goto endjob;

5288
 endjob:
5289
    if (ret < 0)
5290
        qemuMigrationJobFinish(driver, vm);
5291
    else
5292 5293
        qemuMigrationJobContinue(vm);
    if (!virDomainObjIsActive(vm) && !vm->persistent)
5294
        qemuDomainRemoveInactive(driver, vm);
5295

5296
 cleanup:
M
Michal Privoznik 已提交
5297
    virDomainObjEndAPI(&vm);
5298 5299 5300 5301 5302
    if (event)
        qemuDomainEventQueue(driver, event);
    return ret;
}

5303
int
5304
qemuMigrationPerform(virQEMUDriverPtr driver,
5305 5306 5307 5308 5309
                     virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *xmlin,
                     const char *dconnuri,
                     const char *uri,
5310
                     const char *graphicsuri,
5311
                     const char *listenAddress,
5312 5313
                     size_t nmigrate_disks,
                     const char **migrate_disks,
5314 5315 5316 5317 5318 5319 5320 5321 5322 5323
                     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, "
5324 5325
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p, "
5326 5327
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
              "flags=%lx, dname=%s, resource=%lu, v3proto=%d",
5328
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
5329
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
5330 5331
              nmigrate_disks, migrate_disks, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, NULLSTR(dname), resource, v3proto);
5332 5333 5334

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
5335 5336
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
5337 5338 5339 5340
            return -1;
        }

        return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri, uri,
5341
                                       graphicsuri, listenAddress,
5342
                                       nmigrate_disks, migrate_disks,
5343
                                       cookiein, cookieinlen,
5344 5345
                                       cookieout, cookieoutlen,
                                       flags, dname, resource, v3proto);
5346 5347
    } else {
        if (dconnuri) {
5348 5349
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
5350 5351 5352 5353 5354
            return -1;
        }

        if (v3proto) {
            return qemuMigrationPerformPhase(driver, conn, vm, uri,
5355
                                             graphicsuri,
5356
                                             nmigrate_disks, migrate_disks,
5357 5358
                                             cookiein, cookieinlen,
                                             cookieout, cookieoutlen,
5359
                                             flags, resource);
5360 5361
        } else {
            return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri,
5362
                                           uri, graphicsuri, listenAddress,
5363
                                           nmigrate_disks, migrate_disks,
5364
                                           cookiein, cookieinlen,
5365 5366 5367 5368 5369
                                           cookieout, cookieoutlen, flags,
                                           dname, resource, v3proto);
        }
    }
}
5370

5371
static int
5372 5373
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def)
{
5374
    size_t i;
5375 5376 5377 5378 5379
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
5380
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5381
            if (virNetDevVPortProfileAssociate(net->ifname,
5382
                                               virDomainNetGetActualVirtPortProfile(net),
5383
                                               &net->mac,
5384
                                               virDomainNetGetActualDirectDev(net),
5385
                                               -1,
5386
                                               def->uuid,
5387 5388
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
5389 5390 5391
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
5392
                goto err_exit;
5393
            }
5394
            last_good_net = i;
5395
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
5396

5397
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
5398 5399 5400 5401
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
5402 5403 5404
        }
    }

5405
    return 0;
5406

5407
 err_exit:
5408
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
5409
        net = def->nets[i];
5410
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5411
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
5412
                                                           virDomainNetGetActualVirtPortProfile(net),
5413
                                                           &net->mac,
5414
                                                           virDomainNetGetActualDirectDev(net),
5415
                                                           -1,
5416
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
5417 5418
        }
    }
5419
    return -1;
5420 5421 5422 5423
}


virDomainPtr
5424
qemuMigrationFinish(virQEMUDriverPtr driver,
5425 5426
                    virConnectPtr dconn,
                    virDomainObjPtr vm,
5427 5428 5429 5430
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
5431
                    unsigned long flags,
5432 5433
                    int retcode,
                    bool v3proto)
5434 5435
{
    virDomainPtr dom = NULL;
5436
    virObjectEventPtr event = NULL;
5437
    bool newVM = true;
5438
    qemuMigrationCookiePtr mig = NULL;
5439
    virErrorPtr orig_err = NULL;
5440
    int cookie_flags = 0;
J
Jiri Denemark 已提交
5441
    qemuDomainObjPrivatePtr priv = vm->privateData;
5442
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5443
    virCapsPtr caps = NULL;
5444
    unsigned short port;
5445

5446
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
5447
              "cookieout=%p, cookieoutlen=%p, flags=%lx, retcode=%d",
5448 5449
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
5450

5451 5452 5453
    port = priv->migrationPort;
    priv->migrationPort = 0;

5454 5455 5456
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

5457
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
5458 5459
        goto cleanup;

5460 5461 5462
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
5463

5464
    qemuDomainCleanupRemove(vm, qemuMigrationPrepareCleanup);
5465
    VIR_FREE(priv->job.completed);
5466

5467
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
5468 5469
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
5470 5471 5472 5473 5474
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookie_flags |= QEMU_MIGRATION_COOKIE_PERSISTENT;

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein,
                                       cookieinlen, cookie_flags)))
5475
        goto endjob;
5476 5477 5478 5479 5480

    /* 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
5481
        if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
5482 5483
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("guest unexpectedly quit"));
5484
            goto endjob;
5485 5486
        }

5487
        if (mig->jobInfo) {
5488 5489
            qemuDomainJobInfoPtr jobInfo = mig->jobInfo;
            priv->job.completed = jobInfo;
5490
            mig->jobInfo = NULL;
5491 5492 5493 5494
            if (jobInfo->sent && virTimeMillisNow(&jobInfo->received) == 0) {
                jobInfo->timeDelta = jobInfo->received - jobInfo->sent;
                jobInfo->timeDeltaSet = true;
            }
5495 5496
        }

L
liguang 已提交
5497 5498 5499 5500 5501
        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");
5502
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5503 5504 5505 5506 5507 5508 5509
                                                 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");
5510
        }
5511

5512 5513
        if (qemuMigrationStopNBDServer(driver, vm, mig) < 0)
            goto endjob;
5514

5515
        if (flags & VIR_MIGRATE_PERSIST_DEST) {
5516
            virDomainDefPtr vmdef;
5517
            if (vm->persistent)
5518
                newVM = false;
5519
            vm->persistent = 1;
5520 5521 5522
            if (mig->persistent)
                vm->newDef = vmdef = mig->persistent;
            else
5523
                vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm);
5524
            if (!vmdef || virDomainSaveConfig(cfg->configDir, vmdef) < 0) {
5525 5526 5527 5528 5529 5530 5531 5532 5533
                /* 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.
                 */
5534 5535

                /*
5536 5537
                 * However, in v3 protocol, the source VM is still available
                 * to restart during confirm() step, so we kill it off now.
5538 5539
                 */
                if (v3proto) {
L
liguang 已提交
5540 5541 5542 5543 5544
                    if (!(flags & VIR_MIGRATE_OFFLINE)) {
                        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                        VIR_QEMU_PROCESS_STOP_MIGRATED);
                        virDomainAuditStop(vm, "failed");
                    }
5545 5546
                    if (newVM)
                        vm->persistent = 0;
5547
                }
A
Alex Jia 已提交
5548
                if (!vmdef)
5549 5550
                    virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                                   _("can't get vmdef"));
5551 5552 5553
                goto endjob;
            }

5554
            event = virDomainEventLifecycleNewFromObj(vm,
5555 5556 5557 5558 5559 5560 5561 5562 5563
                                             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
5564
        if (!(flags & VIR_MIGRATE_PAUSED) && !(flags & VIR_MIGRATE_OFFLINE)) {
5565 5566 5567 5568
            /* 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
5569
            if (qemuProcessStartCPUs(driver, vm, dconn,
5570 5571
                                     VIR_DOMAIN_RUNNING_MIGRATED,
                                     QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5572
                if (virGetLastError() == NULL)
5573 5574
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   "%s", _("resume operation failed"));
5575 5576 5577 5578 5579
                /* Need to save the current error, in case shutting
                 * down the process overwrites it
                 */
                orig_err = virSaveLastError();

5580 5581 5582 5583 5584 5585 5586 5587 5588
                /*
                 * 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) {
5589 5590
                    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                    VIR_QEMU_PROCESS_STOP_MIGRATED);
5591
                    virDomainAuditStop(vm, "failed");
5592
                    event = virDomainEventLifecycleNewFromObj(vm,
5593 5594 5595
                                                     VIR_DOMAIN_EVENT_STOPPED,
                                                     VIR_DOMAIN_EVENT_STOPPED_FAILED);
                }
5596 5597
                goto endjob;
            }
5598 5599 5600 5601
            if (priv->job.completed) {
                qemuDomainJobInfoUpdateTime(priv->job.completed);
                qemuDomainJobInfoUpdateDowntime(priv->job.completed);
            }
5602 5603
        }

5604
        dom = virGetDomain(dconn, vm->def->name, vm->def->uuid);
5605

L
liguang 已提交
5606
        if (!(flags & VIR_MIGRATE_OFFLINE)) {
5607
            event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5608 5609 5610 5611 5612 5613 5614
                                             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);
5615
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
5616 5617 5618
                                                 VIR_DOMAIN_EVENT_SUSPENDED,
                                                 VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);
            }
5619
        }
L
liguang 已提交
5620 5621

        if (virDomainObjIsActive(vm) &&
5622
            virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
5623 5624 5625
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto endjob;
        }
5626 5627 5628

        /* Guest is successfully running, so cancel previous auto destroy */
        qemuProcessAutoDestroyRemove(driver, vm);
L
liguang 已提交
5629
    } else if (!(flags & VIR_MIGRATE_OFFLINE)) {
5630 5631
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
5632
        virDomainAuditStop(vm, "failed");
5633
        event = virDomainEventLifecycleNewFromObj(vm,
5634 5635 5636 5637
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_FAILED);
    }

5638 5639
    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen,
                                QEMU_MIGRATION_COOKIE_STATS) < 0)
5640 5641
        VIR_WARN("Unable to encode migration cookie");

5642
 endjob:
5643 5644
    qemuMigrationJobFinish(driver, vm);
    if (!vm->persistent && !virDomainObjIsActive(vm))
5645
        qemuDomainRemoveInactive(driver, vm);
5646

5647
 cleanup:
5648
    virPortAllocatorRelease(driver->migrationPorts, port);
5649 5650 5651
    if (priv->mon)
        qemuMonitorSetDomainLog(priv->mon, -1);
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5652
    virDomainObjEndAPI(&vm);
5653 5654
    if (event)
        qemuDomainEventQueue(driver, event);
5655
    qemuMigrationCookieFree(mig);
5656 5657 5658 5659
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5660
    virObjectUnref(caps);
5661
    virObjectUnref(cfg);
5662 5663
    return dom;
}
5664

5665

5666
/* Helper function called while vm is active.  */
5667
int
5668
qemuMigrationToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
5669 5670
                    int fd, off_t offset, const char *path,
                    const char *compressor,
E
Eric Blake 已提交
5671
                    bool bypassSecurityDriver,
5672
                    qemuDomainAsyncJob asyncJob)
5673 5674 5675
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5676
    int ret = -1;
5677
    bool restoreLabel = false;
5678 5679
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5680
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5681
    char *errbuf = NULL;
5682
    virErrorPtr orig_err = NULL;
5683 5684 5685 5686 5687

    /* 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,
5688 5689
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5690 5691
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5692
    }
5693

5694 5695 5696 5697 5698 5699 5700
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5701
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5702
        (!compressor || pipe(pipeFD) == 0)) {
5703
        /* All right! We can use fd migration, which means that qemu
5704 5705 5706
         * 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.  */
5707
        if (virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
5708
                                              compressor ? pipeFD[1] : fd) < 0)
5709
            goto cleanup;
5710 5711 5712
        bypassSecurityDriver = true;
    } else {
        /* Phooey - we have to fall back on exec migration, where qemu
E
Eric Blake 已提交
5713 5714
         * has to popen() the file by name, and block devices have to be
         * given cgroup ACL permission.  We might also stumble on
5715 5716
         * a race present in some qemu versions where it does a wait()
         * that botches pclose.  */
5717 5718
        if (virCgroupHasController(priv->cgroup,
                                   VIR_CGROUP_CONTROLLER_DEVICES)) {
5719 5720 5721 5722
            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
5723
                /* path was not a device, no further need for cgroup */
5724
            } else if (rv < 0) {
5725 5726
                goto cleanup;
            }
5727
        }
5728 5729
        if ((!bypassSecurityDriver) &&
            virSecurityManagerSetSavedStateLabel(driver->securityManager,
5730
                                                 vm->def, path) < 0)
5731
            goto cleanup;
5732
        restoreLabel = true;
5733 5734
    }

5735
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5736 5737
        goto cleanup;

5738 5739 5740
    if (!compressor) {
        const char *args[] = { "cat", NULL };

5741
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757
            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
        };
5758 5759 5760 5761
        if (pipeFD[0] != -1) {
            cmd = virCommandNewArgs(args);
            virCommandSetInputFD(cmd, pipeFD[0]);
            virCommandSetOutputFD(cmd, &fd);
5762 5763
            virCommandSetErrorBuffer(cmd, &errbuf);
            virCommandDoAsyncIO(cmd);
5764 5765 5766
            if (virSetCloseExec(pipeFD[1]) < 0) {
                virReportSystemError(errno, "%s",
                                     _("Unable to set cloexec flag"));
5767
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5768 5769 5770
                goto cleanup;
            }
            if (virCommandRunAsync(cmd, NULL) < 0) {
5771
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5772 5773 5774 5775 5776 5777 5778
                goto cleanup;
            }
            rc = qemuMonitorMigrateToFd(priv->mon,
                                        QEMU_MONITOR_MIGRATE_BACKGROUND,
                                        pipeFD[1]);
            if (VIR_CLOSE(pipeFD[0]) < 0 ||
                VIR_CLOSE(pipeFD[1]) < 0)
5779
                VIR_WARN("failed to close intermediate pipe");
5780 5781 5782 5783 5784
        } else {
            rc = qemuMonitorMigrateToFile(priv->mon,
                                          QEMU_MONITOR_MIGRATE_BACKGROUND,
                                          args, path, offset);
        }
5785
    }
5786
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5787
        goto cleanup;
5788 5789 5790
    if (rc < 0)
        goto cleanup;

5791 5792
    rc = qemuMigrationWaitForCompletion(driver, vm, asyncJob,
                                        NULL, false, false);
5793

5794 5795 5796 5797
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5798 5799
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5800
                qemuMonitorMigrateCancel(priv->mon);
5801
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5802 5803
            }
        }
5804
        goto cleanup;
5805
    }
5806

5807 5808 5809
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5810 5811
    ret = 0;

5812
 cleanup:
5813 5814 5815
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5816
    /* Restore max migration bandwidth */
5817 5818
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5819 5820
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5821
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5822 5823
    }

5824 5825
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
5826 5827 5828 5829 5830
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
5831 5832
    if (restoreLabel && (!bypassSecurityDriver) &&
        virSecurityManagerRestoreSavedStateLabel(driver->securityManager,
5833
                                                 vm->def, path) < 0)
5834 5835
        VIR_WARN("failed to restore save state label on %s", path);

5836 5837
    if (virCgroupHasController(priv->cgroup,
                               VIR_CGROUP_CONTROLLER_DEVICES)) {
5838 5839 5840
        int rv = virCgroupDenyDevicePath(priv->cgroup, path,
                                         VIR_CGROUP_DEVICE_RWM);
        virDomainAuditCgroupPath(vm, priv->cgroup, "deny", path, "rwm", rv == 0);
5841
    }
5842 5843 5844 5845 5846 5847

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

5848 5849
    return ret;
}
5850

5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928

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

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

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

    qemuDomainObjEnterMonitor(driver, vm);

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

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

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

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

        if (!diskPriv->migrating)
            continue;

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

    if (qemuMigrationCancelDriveMirror(driver, vm, false,
                                       QEMU_ASYNC_JOB_NONE) < 0)
        goto endsyncjob;

    ret = 0;

 cleanup:
    virHashFree(blockJobs);
    return ret;

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

            if (diskPriv->migrating) {
                qemuBlockJobSyncEnd(driver, vm, disk);
                diskPriv->migrating = false;
            }
        }
    }
    goto cleanup;
}


5929
int
5930
qemuMigrationJobStart(virQEMUDriverPtr driver,
5931
                      virDomainObjPtr vm,
5932
                      qemuDomainAsyncJob job)
5933 5934 5935
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

5936
    if (qemuDomainObjBeginAsyncJob(driver, vm, job) < 0)
5937 5938
        return -1;

5939
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
5940
        qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
5941
    } else {
5942 5943 5944
        qemuDomainObjSetAsyncJobMask(vm, (QEMU_JOB_DEFAULT_MASK |
                                          JOB_MASK(QEMU_JOB_SUSPEND) |
                                          JOB_MASK(QEMU_JOB_MIGRATION_OP)));
5945
    }
5946

J
Jiri Denemark 已提交
5947
    priv->job.current->type = VIR_DOMAIN_JOB_UNBOUNDED;
5948 5949 5950 5951 5952

    return 0;
}

void
5953
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
5954
                         virDomainObjPtr vm,
5955
                         qemuMigrationJobPhase phase)
5956 5957 5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969
{
    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
5970
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
5971
                           virDomainObjPtr vm,
5972
                           qemuMigrationJobPhase phase)
5973 5974 5975 5976
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

5977
void
5978 5979
qemuMigrationJobContinue(virDomainObjPtr vm)
{
5980
    qemuDomainObjReleaseAsyncJob(vm);
5981 5982 5983 5984
}

bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
5985
                         qemuDomainAsyncJob job)
5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996
{
    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");

5997
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
5998 5999 6000 6001 6002
        return false;
    }
    return true;
}

6003
void
6004
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
6005
{
6006
    qemuDomainObjEndAsyncJob(driver, vm);
6007
}