qemu_migration.c 162.7 KB
Newer Older
1 2 3
/*
 * qemu_migration.c: QEMU migration handling
 *
4
 * Copyright (C) 2006-2014 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23
 *
 */

#include <config.h>

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

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

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

#define VIR_FROM_THIS VIR_FROM_QEMU

62 63
VIR_LOG_INIT("qemu.qemu_migration");

64 65 66 67 68 69 70 71 72 73 74 75 76
VIR_ENUM_IMPL(qemuMigrationJobPhase, QEMU_MIGRATION_PHASE_LAST,
              "none",
              "perform2",
              "begin3",
              "perform3",
              "perform3_done",
              "confirm3_cancelled",
              "confirm3",
              "prepare",
              "finish2",
              "finish3",
);

77
enum qemuMigrationCookieFlags {
78
    QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS,
79
    QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE,
80
    QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT,
81
    QEMU_MIGRATION_COOKIE_FLAG_NETWORK,
82
    QEMU_MIGRATION_COOKIE_FLAG_NBD,
83 84 85 86 87 88 89

    QEMU_MIGRATION_COOKIE_FLAG_LAST
};

VIR_ENUM_DECL(qemuMigrationCookieFlag);
VIR_ENUM_IMPL(qemuMigrationCookieFlag,
              QEMU_MIGRATION_COOKIE_FLAG_LAST,
90 91 92 93 94
              "graphics",
              "lockstate",
              "persistent",
              "network",
              "nbd");
95 96 97

enum qemuMigrationCookieFeatures {
    QEMU_MIGRATION_COOKIE_GRAPHICS  = (1 << QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS),
98
    QEMU_MIGRATION_COOKIE_LOCKSTATE = (1 << QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE),
99
    QEMU_MIGRATION_COOKIE_PERSISTENT = (1 << QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT),
100
    QEMU_MIGRATION_COOKIE_NETWORK = (1 << QEMU_MIGRATION_COOKIE_FLAG_NETWORK),
101
    QEMU_MIGRATION_COOKIE_NBD = (1 << QEMU_MIGRATION_COOKIE_FLAG_NBD),
102 103 104 105 106 107 108 109 110 111 112 113
};

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

114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134
typedef struct _qemuMigrationCookieNetData qemuMigrationCookieNetData;
typedef qemuMigrationCookieNetData *qemuMigrationCookieNetDataPtr;
struct _qemuMigrationCookieNetData {
    int vporttype; /* enum virNetDevVPortProfile */

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

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

    qemuMigrationCookieNetDataPtr net;
};

135 136 137 138 139 140
typedef struct _qemuMigrationCookieNBD qemuMigrationCookieNBD;
typedef qemuMigrationCookieNBD *qemuMigrationCookieNBDPtr;
struct _qemuMigrationCookieNBD {
    int port; /* on which port does NBD server listen for incoming data */
};

141 142 143
typedef struct _qemuMigrationCookie qemuMigrationCookie;
typedef qemuMigrationCookie *qemuMigrationCookiePtr;
struct _qemuMigrationCookie {
E
Eric Blake 已提交
144 145
    unsigned int flags;
    unsigned int flagsMandatory;
146 147

    /* Host properties */
148 149 150 151
    unsigned char localHostuuid[VIR_UUID_BUFLEN];
    unsigned char remoteHostuuid[VIR_UUID_BUFLEN];
    char *localHostname;
    char *remoteHostname;
152 153 154 155

    /* Guest properties */
    unsigned char uuid[VIR_UUID_BUFLEN];
    char *name;
156

157 158 159 160
    /* If (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) */
    char *lockState;
    char *lockDriver;

161 162
    /* If (flags & QEMU_MIGRATION_COOKIE_GRAPHICS) */
    qemuMigrationCookieGraphicsPtr graphics;
163 164 165

    /* If (flags & QEMU_MIGRATION_COOKIE_PERSISTENT) */
    virDomainDefPtr persistent;
166 167 168

    /* If (flags & QEMU_MIGRATION_COOKIE_NETWORK) */
    qemuMigrationCookieNetworkPtr network;
169 170 171

    /* If (flags & QEMU_MIGRATION_COOKIE_NBD) */
    qemuMigrationCookieNBDPtr nbd;
172 173
};

174 175 176 177 178 179 180 181 182
static void qemuMigrationCookieGraphicsFree(qemuMigrationCookieGraphicsPtr grap)
{
    if (!grap)
        return;
    VIR_FREE(grap->listen);
    VIR_FREE(grap->tlsSubject);
    VIR_FREE(grap);
}

183

184 185 186
static void
qemuMigrationCookieNetworkFree(qemuMigrationCookieNetworkPtr network)
{
187
    size_t i;
188 189 190 191 192 193 194 195 196 197 198 199 200

    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);
}


201 202 203 204 205
static void qemuMigrationCookieFree(qemuMigrationCookiePtr mig)
{
    if (!mig)
        return;

206 207
    qemuMigrationCookieGraphicsFree(mig->graphics);
    qemuMigrationCookieNetworkFree(mig->network);
208

209 210
    VIR_FREE(mig->localHostname);
    VIR_FREE(mig->remoteHostname);
211
    VIR_FREE(mig->name);
212 213
    VIR_FREE(mig->lockState);
    VIR_FREE(mig->lockDriver);
214
    VIR_FREE(mig->nbd);
215 216 217 218
    VIR_FREE(mig);
}


219
#ifdef WITH_GNUTLS
220 221 222 223 224 225 226 227 228 229 230 231
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)
232
        goto error;
233 234

    if (virFileReadAll(certfile, 8192, &pemdata) < 0) {
235 236
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unable to read server cert %s"), certfile);
237 238 239 240 241
        goto error;
    }

    ret = gnutls_x509_crt_init(&cert);
    if (ret < 0) {
242 243 244
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot initialize cert object: %s"),
                       gnutls_strerror(ret));
245 246 247 248 249 250 251 252
        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) {
253 254 255
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot load cert data from %s: %s"),
                       certfile, gnutls_strerror(ret));
256 257 258 259 260
        goto error;
    }

    subjectlen = 1024;
    if (VIR_ALLOC_N(subject, subjectlen+1) < 0)
261
        goto error;
262 263 264 265 266 267 268 269 270

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

    VIR_FREE(certfile);
    VIR_FREE(pemdata);

    return subject;

271
 error:
272 273 274 275
    VIR_FREE(certfile);
    VIR_FREE(pemdata);
    return NULL;
}
276
#endif
277 278

static qemuMigrationCookieGraphicsPtr
279
qemuMigrationCookieGraphicsAlloc(virQEMUDriverPtr driver,
280 281 282 283
                                 virDomainGraphicsDefPtr def)
{
    qemuMigrationCookieGraphicsPtr mig = NULL;
    const char *listenAddr;
284
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
285 286

    if (VIR_ALLOC(mig) < 0)
287
        goto error;
288 289 290 291

    mig->type = def->type;
    if (mig->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
        mig->port = def->data.vnc.port;
292
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
293
        if (!listenAddr)
294
            listenAddr = cfg->vncListen;
295

296
#ifdef WITH_GNUTLS
297 298
        if (cfg->vncTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->vncTLSx509certdir)))
299
            goto error;
300
#endif
301 302
    } else {
        mig->port = def->data.spice.port;
303
        if (cfg->spiceTLS)
304 305 306
            mig->tlsPort = def->data.spice.tlsPort;
        else
            mig->tlsPort = -1;
307
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
308
        if (!listenAddr)
309
            listenAddr = cfg->spiceListen;
310

311
#ifdef WITH_GNUTLS
312 313
        if (cfg->spiceTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->spiceTLSx509certdir)))
314
            goto error;
315
#endif
316
    }
317 318
    if (VIR_STRDUP(mig->listen, listenAddr) < 0)
        goto error;
319

320
    virObjectUnref(cfg);
321 322
    return mig;

323
 error:
324
    qemuMigrationCookieGraphicsFree(mig);
325
    virObjectUnref(cfg);
326 327 328 329
    return NULL;
}


330
static qemuMigrationCookieNetworkPtr
331
qemuMigrationCookieNetworkAlloc(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
332 333 334
                                virDomainDefPtr def)
{
    qemuMigrationCookieNetworkPtr mig;
335
    size_t i;
336 337

    if (VIR_ALLOC(mig) < 0)
338
        goto error;
339 340 341 342

    mig->nnets = def->nnets;

    if (VIR_ALLOC_N(mig->net, def->nnets) <0)
343
        goto error;
344 345 346 347 348 349 350 351 352 353 354 355 356 357 358

    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:
359
               break;
360
            case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
361 362 363 364 365 366 367 368
                if (virNetDevOpenvswitchGetMigrateData(&mig->net[i].portdata,
                                                       netptr->ifname) != 0) {
                        virReportSystemError(VIR_ERR_INTERNAL_ERROR,
                                             _("Unable to run command to get OVS port data for "
                                             "interface %s"), netptr->ifname);
                        goto error;
                }
                break;
369 370 371 372 373 374 375
            default:
                break;
            }
        }
    }
    return mig;

376
 error:
377 378 379 380
    qemuMigrationCookieNetworkFree(mig);
    return NULL;
}

381 382 383
static qemuMigrationCookiePtr
qemuMigrationCookieNew(virDomainObjPtr dom)
{
J
Jiri Denemark 已提交
384
    qemuDomainObjPrivatePtr priv = dom->privateData;
385
    qemuMigrationCookiePtr mig = NULL;
J
Jiri Denemark 已提交
386
    const char *name;
387 388

    if (VIR_ALLOC(mig) < 0)
389
        goto error;
390

J
Jiri Denemark 已提交
391 392 393 394
    if (priv->origname)
        name = priv->origname;
    else
        name = dom->def->name;
395 396
    if (VIR_STRDUP(mig->name, name) < 0)
        goto error;
397 398
    memcpy(mig->uuid, dom->def->uuid, VIR_UUID_BUFLEN);

399
    if (!(mig->localHostname = virGetHostname()))
400
        goto error;
401
    if (virGetHostUUID(mig->localHostuuid) < 0) {
402 403
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Unable to obtain host UUID"));
404 405 406 407 408
        goto error;
    }

    return mig;

409
 error:
410 411 412 413 414
    qemuMigrationCookieFree(mig);
    return NULL;
}


415 416
static int
qemuMigrationCookieAddGraphics(qemuMigrationCookiePtr mig,
417
                               virQEMUDriverPtr driver,
418 419
                               virDomainObjPtr dom)
{
420 421
    size_t i = 0;

422
    if (mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) {
423 424
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration graphics data already present"));
425 426 427
        return -1;
    }

428 429 430 431 432 433 434 435
    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;
       }
436
    }
437 438 439 440 441

    return 0;
}


442 443
static int
qemuMigrationCookieAddLockstate(qemuMigrationCookiePtr mig,
444
                                virQEMUDriverPtr driver,
445 446 447 448 449
                                virDomainObjPtr dom)
{
    qemuDomainObjPrivatePtr priv = dom->privateData;

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

    if (virDomainObjGetState(dom, NULL) == VIR_DOMAIN_PAUSED) {
456
        if (VIR_STRDUP(mig->lockState, priv->lockState) < 0)
457 458 459 460 461 462
            return -1;
    } else {
        if (virDomainLockProcessInquire(driver->lockManager, dom, &mig->lockState) < 0)
            return -1;
    }

463
    if (VIR_STRDUP(mig->lockDriver, virLockManagerPluginGetName(driver->lockManager)) < 0) {
464 465 466 467 468 469 470 471 472 473 474
        VIR_FREE(mig->lockState);
        return -1;
    }

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

    return 0;
}


475 476 477 478 479
static int
qemuMigrationCookieAddPersistent(qemuMigrationCookiePtr mig,
                                 virDomainObjPtr dom)
{
    if (mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) {
480 481
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration persistent data already present"));
482 483 484 485 486 487 488 489 490 491 492 493 494
        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;
}


495 496
static int
qemuMigrationCookieAddNetwork(qemuMigrationCookiePtr mig,
497
                              virQEMUDriverPtr driver,
498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515
                              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;
}

516

517 518 519 520 521 522 523 524 525
static int
qemuMigrationCookieAddNBD(qemuMigrationCookiePtr mig,
                          virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    /* It is not a bug if there already is a NBD data */
    if (!mig->nbd &&
526
        VIR_ALLOC(mig->nbd) < 0)
527 528 529 530 531 532 533 534 535
        return -1;

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

    return 0;
}


536 537 538
static void qemuMigrationCookieGraphicsXMLFormat(virBufferPtr buf,
                                                 qemuMigrationCookieGraphicsPtr grap)
{
539
    virBufferAsprintf(buf, "<graphics type='%s' port='%d' listen='%s'",
540 541 542 543 544 545
                      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");
546 547 548 549
        virBufferAdjustIndent(buf, 2);
        virBufferEscapeString(buf, "<cert info='subject' value='%s'/>\n", grap->tlsSubject);
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</graphics>\n");
550 551 552 553 554 555
    } else {
        virBufferAddLit(buf, "/>\n");
    }
}


556 557 558 559
static void
qemuMigrationCookieNetworkXMLFormat(virBufferPtr buf,
                                    qemuMigrationCookieNetworkPtr optr)
{
560
    size_t i;
561 562 563 564 565 566
    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) {
567 568
                virBufferAddLit(buf, "<network>\n");
                virBufferAdjustIndent(buf, 2);
569 570
                empty = false;
            }
571
            virBufferAsprintf(buf, "<interface index='%zu' vporttype='%s'",
572 573 574
                              i, virNetDevVPortTypeToString(optr->net[i].vporttype));
            if (optr->net[i].portdata) {
                virBufferAddLit(buf, ">\n");
575 576
                virBufferAdjustIndent(buf, 2);
                virBufferEscapeString(buf, "<portdata>%s</portdata>\n",
577
                                      optr->net[i].portdata);
578 579
                virBufferAdjustIndent(buf, -2);
                virBufferAddLit(buf, "</interface>\n");
580 581 582 583 584
            } else {
                virBufferAddLit(buf, "/>\n");
            }
        }
    }
585 586 587 588
    if (!empty) {
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</network>\n");
    }
589 590 591
}


592
static int
593
qemuMigrationCookieXMLFormat(virQEMUDriverPtr driver,
594
                             virBufferPtr buf,
595
                             qemuMigrationCookiePtr mig)
596 597 598
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
    char hostuuidstr[VIR_UUID_STRING_BUFLEN];
599
    size_t i;
600 601

    virUUIDFormat(mig->uuid, uuidstr);
602
    virUUIDFormat(mig->localHostuuid, hostuuidstr);
603

604
    virBufferAddLit(buf, "<qemu-migration>\n");
605 606 607 608 609
    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);
610

611
    for (i = 0; i < QEMU_MIGRATION_COOKIE_FLAG_LAST; i++) {
612
        if (mig->flagsMandatory & (1 << i))
613
            virBufferAsprintf(buf, "<feature name='%s'/>\n",
614 615 616
                              qemuMigrationCookieFlagTypeToString(i));
    }

617 618
    if ((mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        mig->graphics)
619 620
        qemuMigrationCookieGraphicsXMLFormat(buf, mig->graphics);

621 622
    if ((mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        mig->lockState) {
623
        virBufferAsprintf(buf, "<lockstate driver='%s'>\n",
624
                          mig->lockDriver);
625 626
        virBufferAdjustIndent(buf, 2);
        virBufferAsprintf(buf, "<leases>%s</leases>\n",
627
                          mig->lockState);
628 629
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</lockstate>\n");
630 631
    }

632 633
    if ((mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        mig->persistent) {
634 635 636
        if (qemuDomainDefFormatBuf(driver,
                                   mig->persistent,
                                   VIR_DOMAIN_XML_INACTIVE |
637 638
                                   VIR_DOMAIN_XML_SECURE |
                                   VIR_DOMAIN_XML_MIGRATABLE,
639
                                   buf) < 0)
640
            return -1;
641 642
    }

643 644 645
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NETWORK) && mig->network)
        qemuMigrationCookieNetworkXMLFormat(buf, mig->network);

646
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NBD) && mig->nbd) {
647
        virBufferAddLit(buf, "<nbd");
648 649 650 651 652
        if (mig->nbd->port)
            virBufferAsprintf(buf, " port='%d'", mig->nbd->port);
        virBufferAddLit(buf, "/>\n");
    }

653
    virBufferAdjustIndent(buf, -2);
654
    virBufferAddLit(buf, "</qemu-migration>\n");
655
    return 0;
656 657 658
}


659
static char *qemuMigrationCookieXMLFormatStr(virQEMUDriverPtr driver,
660
                                             qemuMigrationCookiePtr mig)
661 662 663
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;

664
    if (qemuMigrationCookieXMLFormat(driver, &buf, mig) < 0) {
665 666 667
        virBufferFreeAndReset(&buf);
        return NULL;
    }
668

669
    if (virBufferCheckError(&buf) < 0)
670 671 672 673 674 675
        return NULL;

    return virBufferContentAndReset(&buf);
}


676 677 678 679 680 681 682
static qemuMigrationCookieGraphicsPtr
qemuMigrationCookieGraphicsXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieGraphicsPtr grap;
    char *tmp;

    if (VIR_ALLOC(grap) < 0)
683
        goto error;
684 685

    if (!(tmp = virXPathString("string(./graphics/@type)", ctxt))) {
686 687
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing type attribute in migration data"));
688 689 690
        goto error;
    }
    if ((grap->type = virDomainGraphicsTypeFromString(tmp)) < 0) {
691 692
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unknown graphics type %s"), tmp);
693 694 695
        VIR_FREE(tmp);
        goto error;
    }
E
Eric Blake 已提交
696
    VIR_FREE(tmp);
697
    if (virXPathInt("string(./graphics/@port)", ctxt, &grap->port) < 0) {
698 699
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing port attribute in migration data"));
700 701 702 703
        goto error;
    }
    if (grap->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        if (virXPathInt("string(./graphics/@tlsPort)", ctxt, &grap->tlsPort) < 0) {
704 705
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing tlsPort attribute in migration data"));
706 707 708 709
            goto error;
        }
    }
    if (!(grap->listen = virXPathString("string(./graphics/@listen)", ctxt))) {
710 711
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing listen attribute in migration data"));
712 713 714
        goto error;
    }
    /* Optional */
715
    grap->tlsSubject = virXPathString("string(./graphics/cert[@info='subject']/@value)", ctxt);
716 717 718

    return grap;

719
 error:
720 721 722 723 724
    qemuMigrationCookieGraphicsFree(grap);
    return NULL;
}


725 726 727 728
static qemuMigrationCookieNetworkPtr
qemuMigrationCookieNetworkXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieNetworkPtr optr;
729
    size_t i;
730 731 732 733 734 735
    int n;
    xmlNodePtr *interfaces = NULL;
    char *vporttype;
    xmlNodePtr save_ctxt = ctxt->node;

    if (VIR_ALLOC(optr) < 0)
736
        goto error;
737 738 739 740 741 742 743 744

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

    optr->nnets = n;
745
    if (VIR_ALLOC_N(optr->net, optr->nnets) < 0)
746
        goto error;
747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762

    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);

763
 cleanup:
764 765 766
    ctxt->node = save_ctxt;
    return optr;

767
 error:
768 769 770 771 772 773 774
    VIR_FREE(interfaces);
    qemuMigrationCookieNetworkFree(optr);
    optr = NULL;
    goto cleanup;
}


775 776
static int
qemuMigrationCookieXMLParse(qemuMigrationCookiePtr mig,
777
                            virQEMUDriverPtr driver,
778
                            xmlDocPtr doc,
779
                            xmlXPathContextPtr ctxt,
E
Eric Blake 已提交
780
                            unsigned int flags)
781 782
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
783
    char *tmp = NULL;
784
    xmlNodePtr *nodes = NULL;
785 786
    size_t i;
    int n;
787 788 789 790
    virCapsPtr caps = NULL;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto error;
791 792 793 794 795 796 797 798

    /* 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))) {
799 800
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing name element in migration data"));
801 802 803
        goto error;
    }
    if (STRNEQ(tmp, mig->name)) {
804 805 806
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected name %s vs %s"),
                       tmp, mig->name);
807 808 809 810 811 812 813
        goto error;
    }
    VIR_FREE(tmp);

    /* Extract domain uuid */
    tmp = virXPathString("string(./uuid[1])", ctxt);
    if (!tmp) {
814 815
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing uuid element in migration data"));
816 817 818 819
        goto error;
    }
    virUUIDFormat(mig->uuid, uuidstr);
    if (STRNEQ(tmp, uuidstr)) {
820 821 822
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected UUID %s vs %s"),
                       tmp, uuidstr);
823 824 825 826
    }
    VIR_FREE(tmp);

    /* Check & forbid "localhost" migration */
827
    if (!(mig->remoteHostname = virXPathString("string(./hostname[1])", ctxt))) {
828 829
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostname element in migration data"));
830 831
        goto error;
    }
832
    if (STREQ(mig->remoteHostname, mig->localHostname)) {
833 834 835
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Attempt to migrate guest to the same host %s"),
                       mig->remoteHostname);
836 837 838 839
        goto error;
    }

    if (!(tmp = virXPathString("string(./hostuuid[1])", ctxt))) {
840 841 842
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostuuid element in migration data"));
        goto error;
843
    }
844 845 846 847 848 849 850 851 852 853 854 855
    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);
856

857 858
    /* Check to ensure all mandatory features from XML are also
     * present in 'flags' */
859
    if ((n = virXPathNodeSet("./feature", ctxt, &nodes)) < 0)
860 861
        goto error;

862
    for (i = 0; i < n; i++) {
863 864 865
        int val;
        char *str = virXMLPropString(nodes[i], "name");
        if (!str) {
866 867
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing feature name"));
868 869 870 871
            goto error;
        }

        if ((val = qemuMigrationCookieFlagTypeFromString(str)) < 0) {
872 873 874
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unknown migration cookie feature %s"),
                           str);
875 876 877 878 879
            VIR_FREE(str);
            goto error;
        }

        if ((flags & (1 << val)) == 0) {
880 881 882
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unsupported migration cookie feature %s"),
                           str);
883 884 885 886 887 888
            VIR_FREE(str);
        }
        VIR_FREE(str);
    }
    VIR_FREE(nodes);

889 890 891 892 893
    if ((flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        virXPathBoolean("count(./graphics) > 0", ctxt) &&
        (!(mig->graphics = qemuMigrationCookieGraphicsXMLParse(ctxt))))
        goto error;

894 895 896 897
    if ((flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        virXPathBoolean("count(./lockstate) > 0", ctxt)) {
        mig->lockDriver = virXPathString("string(./lockstate[1]/@driver)", ctxt);
        if (!mig->lockDriver) {
898 899
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Missing lock driver name in migration cookie"));
900 901 902 903 904 905 906
            goto error;
        }
        mig->lockState = virXPathString("string(./lockstate[1]/leases[1])", ctxt);
        if (mig->lockState && STREQ(mig->lockState, ""))
            VIR_FREE(mig->lockState);
    }

907 908 909
    if ((flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        virXPathBoolean("count(./domain) > 0", ctxt)) {
        if ((n = virXPathNodeSet("./domain", ctxt, &nodes)) > 1) {
910 911 912 913
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Too many domain elements in "
                             "migration cookie: %d"),
                           n);
914 915
            goto error;
        }
916 917
        mig->persistent = virDomainDefParseNode(doc, nodes[0],
                                                caps, driver->xmlopt,
918 919 920 921 922 923 924 925 926
                                                -1, VIR_DOMAIN_XML_INACTIVE);
        if (!mig->persistent) {
            /* virDomainDefParseNode already reported
             * an error for us */
            goto error;
        }
        VIR_FREE(nodes);
    }

927 928 929 930 931
    if ((flags & QEMU_MIGRATION_COOKIE_NETWORK) &&
        virXPathBoolean("count(./network) > 0", ctxt) &&
        (!(mig->network = qemuMigrationCookieNetworkXMLParse(ctxt))))
        goto error;

932 933 934 935
    if (flags & QEMU_MIGRATION_COOKIE_NBD &&
        virXPathBoolean("boolean(./nbd)", ctxt)) {
        char *port;

936
        if (VIR_ALLOC(mig->nbd) < 0)
937 938 939 940 941 942 943 944 945 946 947 948 949
            goto error;

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

950
    virObjectUnref(caps);
951 952
    return 0;

953
 error:
954
    VIR_FREE(tmp);
955
    VIR_FREE(nodes);
956
    virObjectUnref(caps);
957 958 959 960 961 962
    return -1;
}


static int
qemuMigrationCookieXMLParseStr(qemuMigrationCookiePtr mig,
963
                               virQEMUDriverPtr driver,
964
                               const char *xml,
E
Eric Blake 已提交
965
                               unsigned int flags)
966 967 968
{
    xmlDocPtr doc = NULL;
    xmlXPathContextPtr ctxt = NULL;
969
    int ret = -1;
970 971 972

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

973
    if (!(doc = virXMLParseStringCtxt(xml, _("(qemu_migration_cookie)"), &ctxt)))
974 975
        goto cleanup;

976
    ret = qemuMigrationCookieXMLParse(mig, driver, doc, ctxt, flags);
977

978
 cleanup:
979 980 981 982 983 984 985 986 987
    xmlXPathFreeContext(ctxt);
    xmlFreeDoc(doc);

    return ret;
}


static int
qemuMigrationBakeCookie(qemuMigrationCookiePtr mig,
988
                        virQEMUDriverPtr driver,
989
                        virDomainObjPtr dom,
990 991
                        char **cookieout,
                        int *cookieoutlen,
E
Eric Blake 已提交
992
                        unsigned int flags)
993
{
994 995
    if (!cookieout || !cookieoutlen)
        return 0;
996 997 998

    *cookieoutlen = 0;

999 1000 1001 1002
    if (flags & QEMU_MIGRATION_COOKIE_GRAPHICS &&
        qemuMigrationCookieAddGraphics(mig, driver, dom) < 0)
        return -1;

1003 1004 1005 1006
    if (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE &&
        qemuMigrationCookieAddLockstate(mig, driver, dom) < 0)
        return -1;

1007 1008 1009 1010
    if (flags & QEMU_MIGRATION_COOKIE_PERSISTENT &&
        qemuMigrationCookieAddPersistent(mig, dom) < 0)
        return -1;

1011 1012 1013 1014 1015
    if (flags & QEMU_MIGRATION_COOKIE_NETWORK &&
        qemuMigrationCookieAddNetwork(mig, driver, dom) < 0) {
        return -1;
    }

1016 1017 1018 1019
    if ((flags & QEMU_MIGRATION_COOKIE_NBD) &&
        qemuMigrationCookieAddNBD(mig, driver, dom) < 0)
        return -1;

1020
    if (!(*cookieout = qemuMigrationCookieXMLFormatStr(driver, mig)))
1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031
        return -1;

    *cookieoutlen = strlen(*cookieout) + 1;

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

    return 0;
}


static qemuMigrationCookiePtr
1032
qemuMigrationEatCookie(virQEMUDriverPtr driver,
1033
                       virDomainObjPtr dom,
1034 1035
                       const char *cookiein,
                       int cookieinlen,
E
Eric Blake 已提交
1036
                       unsigned int flags)
1037 1038 1039 1040 1041 1042
{
    qemuMigrationCookiePtr mig = NULL;

    /* Parse & validate incoming cookie (if any) */
    if (cookiein && cookieinlen &&
        cookiein[cookieinlen-1] != '\0') {
1043 1044
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration cookie was not NULL terminated"));
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054
        goto error;
    }

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

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

    if (cookiein && cookieinlen &&
        qemuMigrationCookieXMLParseStr(mig,
1055
                                       driver,
1056 1057 1058 1059
                                       cookiein,
                                       flags) < 0)
        goto error;

1060 1061 1062
    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
        if (!mig->lockDriver) {
            if (virLockManagerPluginUsesState(driver->lockManager)) {
1063 1064 1065
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Missing %s lock state for migration cookie"),
                               virLockManagerPluginGetName(driver->lockManager));
1066 1067 1068 1069
                goto error;
            }
        } else if (STRNEQ(mig->lockDriver,
                          virLockManagerPluginGetName(driver->lockManager))) {
1070 1071 1072 1073
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Source host lock driver %s different from target %s"),
                           mig->lockDriver,
                           virLockManagerPluginGetName(driver->lockManager));
1074 1075 1076 1077
            goto error;
        }
    }

1078 1079
    return mig;

1080
 error:
1081 1082 1083
    qemuMigrationCookieFree(mig);
    return NULL;
}
1084

1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131
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;
}

1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145
/**
 * 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
1146 1147
                            virDomainObjPtr vm,
                            const char *listenAddr)
1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
{
    int ret = -1;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    unsigned short port = 0;
    char *diskAlias = NULL;
    size_t i;

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

        /* skip shared, RO and source-less disks */
1159 1160
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1161 1162 1163 1164
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1165
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1166 1167 1168 1169 1170 1171 1172
            goto cleanup;

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

        if (!port &&
1173
            ((virPortAllocatorAcquire(driver->migrationPorts, &port) < 0) ||
1174
             (qemuMonitorNBDServerStart(priv->mon, listenAddr, port) < 0))) {
1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
            qemuDomainObjExitMonitor(driver, vm);
            goto cleanup;
        }

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

    priv->nbdPort = port;
    ret = 0;

1189
 cleanup:
1190
    VIR_FREE(diskAlias);
1191
    if (ret < 0)
1192 1193 1194 1195
        virPortAllocatorRelease(driver->remotePorts, port);
    return ret;
}

1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
/**
 * qemuMigrationDriveMirror:
 * @driver: qemu driver
 * @vm: domain
 * @mig: migration cookie
 * @host: where are we migrating to
 * @speed: how much should the copying be limited
 * @migrate_flags: migrate monitor command flags
 *
 * Run drive-mirror to feed NBD server running on dst and wait
 * till the process switches into another phase where writes go
 * simultaneously to both source and destination. And this switch
 * is what we are waiting for before proceeding with the next
 * disk. On success, update @migrate_flags so we don't tell
 * 'migrate' command to do the very same operation.
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
qemuMigrationDriveMirror(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMigrationCookiePtr mig,
                         const char *host,
                         unsigned long speed,
                         unsigned int *migrate_flags)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int mon_ret;
    int port;
    size_t i, lastGood = 0;
    char *diskAlias = NULL;
    char *nbd_dest = NULL;
1230
    char *hoststr = NULL;
1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
    virErrorPtr err = NULL;

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

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

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

1250 1251
    /* escape literal IPv6 address */
    if (strchr(host, ':')) {
1252
        if (virAsprintf(&hoststr, "[%s]", host) < 0)
1253 1254 1255 1256 1257
            goto error;
    } else if (VIR_STRDUP(hoststr, host) < 0) {
        goto error;
    }

1258 1259 1260 1261 1262 1263 1264 1265
    if (*migrate_flags & QEMU_MONITOR_MIGRATE_NON_SHARED_INC)
        mirror_flags |= VIR_DOMAIN_BLOCK_REBASE_SHALLOW;

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

        /* skip shared, RO and source-less disks */
1266 1267
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1268 1269 1270 1271 1272 1273 1274
            continue;

        VIR_FREE(diskAlias);
        VIR_FREE(nbd_dest);
        if ((virAsprintf(&diskAlias, "%s%s",
                         QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0) ||
            (virAsprintf(&nbd_dest, "nbd:%s:%d:exportname=%s",
1275
                         hoststr, port, diskAlias) < 0))
1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309
            goto error;

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

        if (mon_ret < 0)
            goto error;

        lastGood = i;

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

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

            if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
                goto error;
            if (priv->job.asyncAbort) {
                /* explicitly do this *after* we entered the monitor,
                 * as this is a critical section so we are guaranteed
                 * priv->job.asyncAbort will not change */
                qemuDomainObjExitMonitor(driver, vm);
                virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                               qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                               _("canceled by client"));
                goto error;
            }
1310 1311
            mon_ret = qemuMonitorBlockJobInfo(priv->mon, diskAlias, &info,
                                              NULL);
1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338
            qemuDomainObjExitMonitor(driver, vm);

            if (mon_ret < 0)
                goto error;

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

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

            virObjectUnlock(vm);

            nanosleep(&ts, NULL);

            virObjectLock(vm);
        }
    }

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

1339
 cleanup:
1340 1341
    VIR_FREE(diskAlias);
    VIR_FREE(nbd_dest);
1342
    VIR_FREE(hoststr);
1343 1344
    return ret;

1345
 error:
1346 1347 1348 1349 1350 1351 1352
    /* don't overwrite any errors */
    err = virSaveLastError();
    /* cancel any outstanding jobs */
    while (lastGood) {
        virDomainDiskDefPtr disk = vm->def->disks[--lastGood];

        /* skip shared, RO disks */
1353 1354
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1355 1356 1357 1358
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1359
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1360 1361 1362
            continue;
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
1363
            if (qemuMonitorBlockJob(priv->mon, diskAlias, NULL, NULL, 0,
1364
                                    BLOCK_JOB_ABORT, true) < 0) {
1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376
                VIR_WARN("Unable to cancel block-job on '%s'", diskAlias);
            }
            qemuDomainObjExitMonitor(driver, vm);
        } else {
            VIR_WARN("Unable to enter monitor. No block job cancelled");
        }
    }
    if (err)
        virSetError(err);
    virFreeError(err);
    goto cleanup;
}
1377

1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401

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

    if (!mig->nbd)
        return;

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

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

    qemuDomainObjExitMonitor(driver, vm);

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

1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416
static void
qemuMigrationCancelDriveMirror(qemuMigrationCookiePtr mig,
                               virQEMUDriverPtr driver,
                               virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    size_t i;
    char *diskAlias = NULL;

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

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

        /* skip shared, RO and source-less disks */
1417 1418
        if (disk->src->shared || disk->src->readonly ||
            !virDomainDiskGetSource(disk))
1419 1420 1421 1422
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1423
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1424 1425 1426 1427 1428 1429
            goto cleanup;

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

1430
        if (qemuMonitorBlockJob(priv->mon, diskAlias, NULL, NULL, 0,
1431
                                BLOCK_JOB_ABORT, true) < 0)
1432 1433 1434 1435
            VIR_WARN("Unable to stop block job on %s", diskAlias);
        qemuDomainObjExitMonitor(driver, vm);
    }

1436
 cleanup:
1437 1438 1439 1440
    VIR_FREE(diskAlias);
    return;
}

1441 1442 1443 1444 1445 1446 1447 1448 1449
/* 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. */
1450
bool
1451
qemuMigrationIsAllowed(virQEMUDriverPtr driver, virDomainObjPtr vm,
1452
                       virDomainDefPtr def, bool remote, bool abort_on_error)
1453
{
1454
    int nsnapshots;
1455
    int pauseReason;
1456
    bool forbid;
1457
    size_t i;
1458 1459 1460

    if (vm) {
        if (qemuProcessAutoDestroyActive(driver, vm)) {
1461 1462
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("domain is marked for auto destroy"));
1463 1464
            return false;
        }
1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477

        /* 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;
            }
1478 1479

            /* cancel migration if disk I/O error is emitted while migrating */
1480 1481
            if (abort_on_error &&
                virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
1482 1483 1484 1485 1486 1487
                pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("cannot migrate domain with I/O error"));
                return false;
            }

1488
        }
1489

E
Eric Blake 已提交
1490 1491
        if (virDomainHasDiskMirror(vm)) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1492
                           _("domain has an active block job"));
E
Eric Blake 已提交
1493 1494
            return false;
        }
1495 1496 1497

        def = vm->def;
    }
1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512

    /* Migration with USB host devices is allowed, all other devices are
     * forbidden.
     */
    forbid = false;
    for (i = 0; i < def->nhostdevs; i++) {
        virDomainHostdevDefPtr hostdev = def->hostdevs[i];
        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS ||
            hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB) {
            forbid = true;
            break;
        }
    }
    if (forbid) {
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
1513
                       _("domain has assigned non-USB host devices"));
1514 1515 1516
        return false;
    }

1517
    for (i = 0; def->cpu && i < def->cpu->nfeatures; i++) {
J
Ján Tomko 已提交
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531
        virCPUFeatureDefPtr feature = &def->cpu->features[i];

        if (feature->policy != VIR_CPU_FEATURE_REQUIRE)
            continue;

        /* 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;
        }
    }

1532 1533 1534
    return true;
}

1535 1536 1537
static bool
qemuMigrationIsSafe(virDomainDefPtr def)
{
1538
    size_t i;
1539

1540
    for (i = 0; i < def->ndisks; i++) {
1541
        virDomainDiskDefPtr disk = def->disks[i];
1542
        const char *src = virDomainDiskGetSource(disk);
1543

1544 1545
        /* Our code elsewhere guarantees shared disks are either readonly (in
         * which case cache mode doesn't matter) or used with cache=none */
1546
        if (src &&
1547 1548
            !disk->src->shared &&
            !disk->src->readonly &&
1549
            disk->cachemode != VIR_DOMAIN_DISK_CACHE_DISABLE) {
1550
            int rc;
1551

E
Eric Blake 已提交
1552
            if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
1553
                if ((rc = virFileIsSharedFS(src)) < 0)
1554 1555
                    return false;
                else if (rc == 0)
1556
                    continue;
1557
                if ((rc = virStorageFileIsClusterFS(src)) < 0)
1558
                    return false;
1559 1560
                else if (rc == 1)
                    continue;
1561 1562
            } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                       disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
1563
                continue;
1564
            }
1565

1566 1567 1568
            virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                           _("Migration may lead to data corruption if disks"
                             " use cache != none"));
1569 1570 1571 1572 1573 1574 1575
            return false;
        }
    }

    return true;
}

1576 1577 1578 1579
/** qemuMigrationSetOffline
 * Pause domain for non-live migration.
 */
int
1580
qemuMigrationSetOffline(virQEMUDriverPtr driver,
1581 1582 1583
                        virDomainObjPtr vm)
{
    int ret;
1584
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
1585 1586
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
1587
    if (ret == 0) {
1588
        virObjectEventPtr event;
1589

1590
        event = virDomainEventLifecycleNewFromObj(vm,
1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
        if (event)
            qemuDomainEventQueue(driver, event);
    }

    return ret;
}


1601 1602 1603
static int
qemuMigrationSetCompression(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
1604
                            qemuDomainAsyncJob job)
1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret;

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

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

    if (ret < 0) {
        goto cleanup;
    } else if (ret == 0) {
        if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                           _("Compressed migration is not supported by "
                             "target QEMU binary"));
        } else {
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                           _("Compressed migration is not supported by "
                             "source QEMU binary"));
        }
        ret = -1;
        goto cleanup;
    }

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

1636
 cleanup:
1637 1638 1639 1640
    qemuDomainObjExitMonitor(driver, vm);
    return ret;
}

1641 1642 1643
static int
qemuMigrationSetAutoConverge(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
1644
                             qemuDomainAsyncJob job)
1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret;

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

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

    if (ret < 0) {
        goto cleanup;
    } else if (ret == 0) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("Auto-Converge is not supported by "
                         "QEMU binary"));
        ret = -1;
        goto cleanup;
    }

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

1670
 cleanup:
1671 1672 1673 1674 1675
    qemuDomainObjExitMonitor(driver, vm);
    return ret;
}


1676 1677 1678 1679 1680 1681 1682
static int
qemuMigrationWaitForSpice(virQEMUDriverPtr driver,
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool wait_for_spice = false;
    bool spice_migrated = false;
1683
    size_t i = 0;
1684

1685 1686 1687 1688 1689 1690 1691 1692
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION)) {
        for (i = 0; i < vm->def->ngraphics; i++) {
            if (vm->def->graphics[i]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
                wait_for_spice = true;
                break;
            }
        }
    }
1693 1694 1695 1696 1697 1698 1699 1700

    if (!wait_for_spice)
        return 0;

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

1701 1702 1703 1704
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            return -1;

1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717
        if (qemuMonitorGetSpiceMigrationStatus(priv->mon,
                                               &spice_migrated) < 0) {
            qemuDomainObjExitMonitor(driver, vm);
            return -1;
        }
        qemuDomainObjExitMonitor(driver, vm);
        virObjectUnlock(vm);
        nanosleep(&ts, NULL);
        virObjectLock(vm);
    }

    return 0;
}
1718

1719
static int
1720
qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver,
1721
                             virDomainObjPtr vm,
1722
                             const char *job,
1723
                             qemuDomainAsyncJob asyncJob)
1724
{
1725
    qemuDomainObjPrivatePtr priv = vm->privateData;
1726
    int ret;
1727 1728 1729
    qemuMonitorMigrationStatus status;

    memset(&status, 0, sizeof(status));
1730

1731 1732
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
    if (ret < 0) {
1733 1734 1735
        /* Guest already exited or waiting for the job timed out; nothing
         * further to update. */
        return ret;
1736
    }
1737
    ret = qemuMonitorGetMigrationStatus(priv->mon, &status);
1738

1739
    qemuDomainObjExitMonitor(driver, vm);
1740

1741 1742
    priv->job.status = status;

1743
    if (ret < 0 || virTimeMillisNow(&priv->job.info.timeElapsed) < 0)
1744
        return -1;
1745

1746
    priv->job.info.timeElapsed -= priv->job.start;
1747

1748
    ret = -1;
1749
    switch (priv->job.status.status) {
1750
    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
1751
        priv->job.info.type = VIR_DOMAIN_JOB_NONE;
1752 1753
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("is not active"));
1754 1755
        break;

M
Michael Avdienko 已提交
1756 1757 1758 1759
    case QEMU_MONITOR_MIGRATION_STATUS_SETUP:
        ret = 0;
        break;

1760
    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775
        priv->job.info.fileTotal = priv->job.status.disk_total;
        priv->job.info.fileRemaining = priv->job.status.disk_remaining;
        priv->job.info.fileProcessed = priv->job.status.disk_transferred;

        priv->job.info.memTotal = priv->job.status.ram_total;
        priv->job.info.memRemaining = priv->job.status.ram_remaining;
        priv->job.info.memProcessed = priv->job.status.ram_transferred;

        priv->job.info.dataTotal =
            priv->job.status.ram_total + priv->job.status.disk_total;
        priv->job.info.dataRemaining =
            priv->job.status.ram_remaining + priv->job.status.disk_remaining;
        priv->job.info.dataProcessed =
            priv->job.status.ram_transferred +
            priv->job.status.disk_transferred;
1776 1777 1778 1779 1780

        ret = 0;
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
1781
        priv->job.info.type = VIR_DOMAIN_JOB_COMPLETED;
1782 1783 1784 1785
        ret = 0;
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
1786
        priv->job.info.type = VIR_DOMAIN_JOB_FAILED;
1787 1788
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("%s: %s"), job, _("unexpectedly failed"));
1789 1790 1791
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
1792
        priv->job.info.type = VIR_DOMAIN_JOB_CANCELLED;
1793 1794
        virReportError(VIR_ERR_OPERATION_ABORTED,
                       _("%s: %s"), job, _("canceled by client"));
1795 1796 1797 1798 1799 1800 1801
        break;
    }

    return ret;
}


1802 1803 1804
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
1805
static int
1806
qemuMigrationWaitForCompletion(virQEMUDriverPtr driver, virDomainObjPtr vm,
1807
                               qemuDomainAsyncJob asyncJob,
1808
                               virConnectPtr dconn, bool abort_on_error)
1809
{
1810
    qemuDomainObjPrivatePtr priv = vm->privateData;
1811
    const char *job;
1812
    int pauseReason;
1813

1814 1815
    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
1816 1817
        job = _("migration job");
        break;
1818
    case QEMU_ASYNC_JOB_SAVE:
1819 1820
        job = _("domain save job");
        break;
1821
    case QEMU_ASYNC_JOB_DUMP:
1822 1823 1824 1825 1826
        job = _("domain core dump job");
        break;
    default:
        job = _("job");
    }
1827

1828
    priv->job.info.type = VIR_DOMAIN_JOB_UNBOUNDED;
1829

1830
    while (priv->job.info.type == VIR_DOMAIN_JOB_UNBOUNDED) {
1831 1832 1833
        /* Poll every 50ms for progress & to allow cancellation */
        struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };

1834
        if (qemuMigrationUpdateJobStatus(driver, vm, job, asyncJob) == -1)
1835
            break;
1836

1837
        /* cancel migration if disk I/O error is emitted while migrating */
1838
        if (abort_on_error &&
1839
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
1840 1841 1842 1843 1844
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("%s: %s"), job, _("failed due to I/O error"));
            break;
        }
1845

1846
        if (dconn && virConnectIsAlive(dconn) <= 0) {
1847 1848
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
1849
            break;
1850 1851
        }

1852
        virObjectUnlock(vm);
1853 1854 1855

        nanosleep(&ts, NULL);

1856
        virObjectLock(vm);
1857 1858
    }

1859
    if (priv->job.info.type == VIR_DOMAIN_JOB_COMPLETED) {
1860
        return 0;
1861 1862 1863 1864 1865 1866 1867
    } else if (priv->job.info.type == VIR_DOMAIN_JOB_UNBOUNDED) {
        /* The migration was aborted by us rather than QEMU itself so let's
         * update the job type and notify the caller to send migrate_cancel.
         */
        priv->job.info.type = VIR_DOMAIN_JOB_FAILED;
        return -2;
    } else {
1868
        return -1;
1869
    }
1870 1871 1872
}


1873
static int
1874
qemuDomainMigrateGraphicsRelocate(virQEMUDriverPtr driver,
1875
                                  virDomainObjPtr vm,
1876 1877
                                  qemuMigrationCookiePtr cookie,
                                  const char *graphicsuri)
1878 1879
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1880 1881
    int ret = -1;
    const char *listenAddress = NULL;
1882
    virSocketAddr addr;
1883 1884 1885 1886 1887
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
1888

1889
    if (!cookie || (!cookie->graphics && !graphicsuri))
1890 1891
        return 0;

1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910
    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) {
1911
        size_t i;
1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938

        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;
            }
        }
    }
1939 1940 1941 1942

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
1943 1944 1945 1946
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
1947

1948 1949 1950 1951
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
1952
        qemuDomainObjExitMonitor(driver, vm);
1953
    }
1954

1955
 cleanup:
1956
    virURIFree(uri);
1957 1958 1959 1960
    return ret;
}


1961
static int
1962
qemuDomainMigrateOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
1963 1964 1965
                             virDomainObjPtr vm,
                             qemuMigrationCookiePtr cookie)
{
1966 1967
    virDomainNetDefPtr netptr;
    int ret = -1;
1968
    size_t i;
1969 1970 1971 1972 1973 1974 1975 1976

    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:
1977
           break;
1978
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
1979 1980 1981 1982 1983 1984 1985 1986
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
                virReportSystemError(VIR_ERR_INTERNAL_ERROR,
                                     _("Unable to run command to set OVS port data for "
                                     "interface %s"), netptr->ifname);
                goto cleanup;
            }
            break;
1987 1988 1989 1990 1991
        default:
            break;
        }
    }

1992
    ret = 0;
1993
 cleanup:
1994 1995 1996 1997
    return ret;
}


1998 1999 2000 2001 2002 2003
/* 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.
 */
2004 2005 2006 2007
static virDomainObjPtr
qemuMigrationCleanup(virDomainObjPtr vm,
                     virConnectPtr conn,
                     void *opaque)
2008
{
2009
    virQEMUDriverPtr driver = opaque;
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024
    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);

2025
    switch ((qemuMigrationJobPhase) priv->job.phase) {
2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
    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 */
        ;
    }

2056
 cleanup:
2057 2058 2059
    return vm;
}

2060

2061
/* The caller is supposed to lock the vm and start a migration job. */
2062 2063
static char
*qemuMigrationBeginPhase(virQEMUDriverPtr driver,
2064
                         virDomainObjPtr vm,
2065
                         const char *xmlin,
2066
                         const char *dname,
2067
                         char **cookieout,
2068 2069
                         int *cookieoutlen,
                         unsigned long flags)
2070 2071 2072
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
2073
    virDomainDefPtr def = NULL;
2074
    qemuDomainObjPrivatePtr priv = vm->privateData;
2075
    virCapsPtr caps = NULL;
2076
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
2077
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2078

2079
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
2080
              " cookieout=%p, cookieoutlen=%p, flags=%lx",
2081
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
2082
              cookieout, cookieoutlen, flags);
2083

2084 2085 2086
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2087 2088 2089 2090 2091 2092
    /* 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);
2093

2094
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
2095 2096
        goto cleanup;

2097 2098 2099
    if (!(flags & VIR_MIGRATE_UNSAFE) && !qemuMigrationIsSafe(vm->def))
        goto cleanup;

2100
    if (flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
2101
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DRIVE_MIRROR)) {
2102
        /* TODO support NBD for TUNNELLED migration */
2103
        if (flags & VIR_MIGRATE_TUNNELLED) {
2104 2105 2106 2107
            VIR_WARN("NBD in tunnelled migration is currently not supported");
        } else {
            cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
            priv->nbdPort = 0;
2108 2109 2110
        }
    }

2111
    if (!(mig = qemuMigrationEatCookie(driver, vm, NULL, 0, 0)))
2112 2113 2114 2115
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
                                cookieout, cookieoutlen,
2116
                                cookieFlags) < 0)
2117 2118
        goto cleanup;

L
liguang 已提交
2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140
    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;
        }
    }

2141
    if (xmlin) {
2142
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt,
M
Matthias Bolte 已提交
2143
                                            QEMU_EXPECTED_VIRT_TYPES,
2144 2145 2146
                                            VIR_DOMAIN_XML_INACTIVE)))
            goto cleanup;

2147
        if (!qemuDomainDefCheckABIStability(driver, vm->def, def))
2148 2149
            goto cleanup;

2150
        rv = qemuDomainDefFormatLive(driver, def, false, true);
2151
    } else {
2152
        rv = qemuDomainDefFormatLive(driver, vm->def, false, true);
2153
    }
2154

2155
 cleanup:
2156
    qemuMigrationCookieFree(mig);
2157
    virObjectUnref(caps);
2158
    virDomainDefFree(def);
2159 2160 2161
    return rv;
}

2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172
char *
qemuMigrationBegin(virConnectPtr conn,
                   virDomainObjPtr vm,
                   const char *xmlin,
                   const char *dname,
                   char **cookieout,
                   int *cookieoutlen,
                   unsigned long flags)
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
2173
    qemuDomainAsyncJob asyncJob;
2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184

    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;
    }

2185 2186
    qemuMigrationStoreDomainState(vm);

2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
        goto endjob;
    }

    /* Check if there is any ejected media.
     * We don't want to require them on the destination.
     */
    if (!(flags & VIR_MIGRATE_OFFLINE) &&
        qemuDomainCheckEjectableMedia(driver, vm, asyncJob) < 0)
        goto endjob;

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

    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
        /* We keep the job active across API calls until the confirm() call.
         * This prevents any other APIs being invoked while migration is taking
         * place.
         */
2210 2211
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                 qemuMigrationCleanup) < 0)
2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224
            goto endjob;
        if (qemuMigrationJobContinue(vm) == 0) {
            vm = NULL;
            virReportError(VIR_ERR_OPERATION_FAILED,
                           "%s", _("domain disappeared"));
            VIR_FREE(xml);
            if (cookieout)
                VIR_FREE(*cookieout);
        }
    } else {
        goto endjob;
    }

2225
 cleanup:
2226 2227 2228 2229
    if (vm)
        virObjectUnlock(vm);
    return xml;

2230
 endjob:
2231 2232 2233 2234
    if ((flags & VIR_MIGRATE_CHANGE_PROTECTION)) {
        if (qemuMigrationJobFinish(driver, vm) == 0)
            vm = NULL;
    } else {
E
Eric Blake 已提交
2235
        if (!qemuDomainObjEndJob(driver, vm))
2236 2237 2238 2239 2240
            vm = NULL;
    }
    goto cleanup;
}

2241

2242 2243
/* Prepare is the first step, and it runs on the destination host.
 */
2244

2245
static void
2246
qemuMigrationPrepareCleanup(virQEMUDriverPtr driver,
2247 2248 2249 2250 2251 2252 2253 2254 2255 2256
                            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));

2257 2258 2259
    virPortAllocatorRelease(driver->migrationPorts, priv->migrationPort);
    priv->migrationPort = 0;

2260 2261 2262 2263 2264
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

2265
static int
2266
qemuMigrationPrepareAny(virQEMUDriverPtr driver,
2267 2268 2269 2270 2271
                        virConnectPtr dconn,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
2272
                        virDomainDefPtr *def,
2273
                        const char *origname,
L
liguang 已提交
2274
                        virStreamPtr st,
2275 2276
                        unsigned short port,
                        bool autoPort,
2277
                        const char *listenAddress,
L
liguang 已提交
2278
                        unsigned long flags)
2279 2280
{
    virDomainObjPtr vm = NULL;
2281
    virObjectEventPtr event = NULL;
2282
    int ret = -1;
2283
    int dataFD[2] = { -1, -1 };
2284
    qemuDomainObjPrivatePtr priv = NULL;
J
Jiri Denemark 已提交
2285
    unsigned long long now;
2286
    qemuMigrationCookiePtr mig = NULL;
2287
    bool tunnel = !!st;
J
Jiri Denemark 已提交
2288
    char *xmlout = NULL;
L
liguang 已提交
2289
    unsigned int cookieFlags;
2290
    virCapsPtr caps = NULL;
J
Ján Tomko 已提交
2291
    char *migrateFrom = NULL;
2292
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2293
    bool taint_hook = false;
2294

2295
    if (virTimeMillisNow(&now) < 0)
2296 2297
        return -1;

L
liguang 已提交
2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319
    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;
        }
    }

2320 2321 2322
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2323
    if (!qemuMigrationIsAllowed(driver, NULL, *def, true, abort_on_error))
2324 2325
        goto cleanup;

J
Jiri Denemark 已提交
2326 2327 2328 2329 2330
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

2331
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
2332 2333
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
2334 2335
            goto cleanup;

2336
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
            if (!*xmlout) {
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
2351
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt,
J
Jiri Denemark 已提交
2352 2353 2354 2355 2356
                                                 QEMU_EXPECTED_VIRT_TYPES,
                                                 VIR_DOMAIN_XML_INACTIVE);
                if (!newdef)
                    goto cleanup;

2357
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
2358 2359 2360 2361
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

2362 2363
                virDomainDefFree(*def);
                *def = newdef;
2364 2365 2366 2367
                /* 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
2368 2369 2370 2371
            }
        }
    }

J
Ján Tomko 已提交
2372 2373 2374 2375
    if (tunnel) {
        /* QEMU will be started with -incoming stdio
         * (which qemu_command might convert to exec:cat or fd:n)
         */
2376
        if (VIR_STRDUP(migrateFrom, "stdio") < 0)
J
Ján Tomko 已提交
2377 2378
            goto cleanup;
    } else {
2379 2380 2381 2382
        virSocketAddr listenAddressSocket;
        bool encloseAddress = false;
        bool hostIPv6Capable = false;
        bool qemuIPv6Capable = false;
J
Ján Tomko 已提交
2383 2384 2385 2386 2387
        virQEMUCapsPtr qemuCaps = NULL;
        struct addrinfo *info = NULL;
        struct addrinfo hints = { .ai_flags = AI_ADDRCONFIG,
                                  .ai_socktype = SOCK_STREAM };

2388 2389 2390 2391
        if (getaddrinfo("::", NULL, &hints, &info) == 0) {
            freeaddrinfo(info);
            hostIPv6Capable = true;
        }
J
Ján Tomko 已提交
2392
        if (!(qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
2393
                                                    (*def)->emulator)))
J
Ján Tomko 已提交
2394 2395
            goto cleanup;

2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422
        qemuIPv6Capable = virQEMUCapsGet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
        virObjectUnref(qemuCaps);

        if (listenAddress) {
            if (virSocketAddrIsNumeric(listenAddress)) {
                /* listenAddress is numeric IPv4 or IPv6 */
                if (virSocketAddrParse(&listenAddressSocket, listenAddress, AF_UNSPEC) < 0)
                    goto cleanup;

                /* address parsed successfully */
                if (VIR_SOCKET_ADDR_IS_FAMILY(&listenAddressSocket, 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"));
                        goto cleanup;
                    }
                    /* IPv6 address must be escaped in brackets on the cmd line */
                    encloseAddress = true;
                }
            } else {
                /* listenAddress is a hostname */
            }
J
Ján Tomko 已提交
2423
        } else {
2424 2425 2426 2427 2428
            /* Listen on :: instead of 0.0.0.0 if QEMU understands it
             * and there is at least one IPv6 address configured
             */
            listenAddress = qemuIPv6Capable && hostIPv6Capable ?
                encloseAddress = true, "::" : "0.0.0.0";
J
Ján Tomko 已提交
2429 2430
        }

2431 2432
        /* QEMU will be started with -incoming [<IPv6 addr>]:port,
         * -incoming <IPv4 addr>:port or -incoming <hostname>:port
J
Ján Tomko 已提交
2433
         */
2434 2435 2436 2437
        if ((encloseAddress &&
             virAsprintf(&migrateFrom, "tcp:[%s]:%d", listenAddress, port) < 0) ||
            (!encloseAddress &&
             virAsprintf(&migrateFrom, "tcp:%s:%d", listenAddress, port) < 0))
J
Ján Tomko 已提交
2438 2439 2440
            goto cleanup;
    }

2441
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
2442
                                   driver->xmlopt,
2443 2444 2445
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
2446
        goto cleanup;
2447

2448
    *def = NULL;
2449
    priv = vm->privateData;
2450 2451
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
2452

2453 2454 2455 2456 2457
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

2458
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
2459 2460
                                       QEMU_MIGRATION_COOKIE_LOCKSTATE |
                                       QEMU_MIGRATION_COOKIE_NBD)))
2461 2462
        goto cleanup;

2463
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
2464
        goto cleanup;
2465
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
2466 2467 2468 2469

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

L
liguang 已提交
2470 2471 2472
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

2473 2474
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
2475 2476
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
2477 2478 2479 2480
        goto endjob;
    }

    /* Start the QEMU daemon, with the same command-line arguments plus
2481
     * -incoming $migrateFrom
2482
     */
2483 2484
    if (qemuProcessStart(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                         migrateFrom, dataFD[0], NULL, NULL,
2485 2486
                         VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
                         VIR_QEMU_PROCESS_START_PAUSED |
2487
                         VIR_QEMU_PROCESS_START_AUTODESTROY) < 0) {
2488
        virDomainAuditStart(vm, "migrated", false);
2489 2490 2491
        goto endjob;
    }

2492 2493 2494 2495
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
2496
            goto stop;
2497
        }
2498
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
2499 2500
    }

2501 2502 2503 2504 2505
    if (flags & VIR_MIGRATE_COMPRESSED &&
        qemuMigrationSetCompression(driver, vm,
                                    QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stop;

2506 2507 2508 2509 2510 2511 2512 2513
    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");
    }
2514

2515
 done:
L
liguang 已提交
2516 2517 2518 2519 2520
    if (flags & VIR_MIGRATE_OFFLINE)
        cookieFlags = 0;
    else
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS;

2521 2522 2523
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
2524
        if (qemuMigrationStartNBDServer(driver, vm, listenAddress) < 0) {
2525 2526
            /* error already reported */
            goto endjob;
2527
        }
2528
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
2529 2530 2531 2532
    }

    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
2533 2534 2535 2536 2537 2538 2539
        /* 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");
    }

2540 2541 2542
    if (qemuDomainCleanupAdd(vm, qemuMigrationPrepareCleanup) < 0)
        goto endjob;

L
liguang 已提交
2543 2544
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
2545
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
2546 2547 2548
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
2549

2550 2551 2552 2553
    /* 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.
     */
2554
    if (!qemuMigrationJobContinue(vm)) {
2555
        vm = NULL;
2556 2557
        virReportError(VIR_ERR_OPERATION_FAILED,
                       "%s", _("domain disappeared"));
2558
        goto cleanup;
2559
    }
2560

2561 2562
    if (autoPort)
        priv->migrationPort = port;
2563
    ret = 0;
2564

2565
 cleanup:
J
Ján Tomko 已提交
2566
    VIR_FREE(migrateFrom);
J
Jiri Denemark 已提交
2567
    VIR_FREE(xmlout);
2568 2569
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
2570
    if (vm) {
2571
        if (ret < 0) {
2572 2573 2574
            virPortAllocatorRelease(driver->remotePorts, priv->nbdPort);
            priv->nbdPort = 0;
        }
2575 2576 2577 2578
        if (ret >= 0 || vm->persistent)
            virObjectUnlock(vm);
        else
            qemuDomainRemoveInactive(driver, vm);
2579
    }
2580 2581
    if (event)
        qemuDomainEventQueue(driver, event);
2582
    qemuMigrationCookieFree(mig);
2583
    virObjectUnref(caps);
2584
    return ret;
2585

2586
 stop:
2587 2588 2589
    virDomainAuditStart(vm, "migrated", false);
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, 0);

2590
 endjob:
2591
    if (!qemuMigrationJobFinish(driver, vm)) {
2592 2593 2594
        vm = NULL;
    }
    goto cleanup;
2595 2596 2597
}


2598 2599 2600 2601 2602
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
2603
qemuMigrationPrepareTunnel(virQEMUDriverPtr driver,
2604 2605 2606 2607 2608 2609
                           virConnectPtr dconn,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virStreamPtr st,
2610
                           virDomainDefPtr *def,
2611
                           const char *origname,
L
liguang 已提交
2612
                           unsigned long flags)
2613 2614 2615 2616
{
    int ret;

    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
2617
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
2618
              "origname=%s, flags=%lx",
2619
              driver, dconn, NULLSTR(cookiein), cookieinlen,
2620
              cookieout, cookieoutlen, st, *def, origname, flags);
2621

2622 2623 2624 2625 2626 2627
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

2628
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
2629
                                  cookieout, cookieoutlen, def, origname,
2630
                                  st, 0, false, NULL, flags);
2631 2632 2633 2634
    return ret;
}


2635
int
2636
qemuMigrationPrepareDirect(virQEMUDriverPtr driver,
2637
                           virConnectPtr dconn,
2638 2639 2640 2641
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
2642 2643
                           const char *uri_in,
                           char **uri_out,
2644
                           virDomainDefPtr *def,
2645
                           const char *origname,
2646
                           const char *listenAddress,
L
liguang 已提交
2647
                           unsigned long flags)
2648
{
2649 2650
    unsigned short port = 0;
    bool autoPort = true;
2651 2652
    char *hostname = NULL;
    const char *p;
J
Ján Tomko 已提交
2653
    char *uri_str = NULL;
2654
    int ret = -1;
2655
    virURIPtr uri = NULL;
2656
    bool well_formed_uri = true;
2657 2658
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
2659

2660 2661
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
2662
              "def=%p, origname=%s, listenAddress=%s, flags=%lx",
2663 2664
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
2665
              *def, origname, NULLSTR(listenAddress), flags);
2666

2667 2668
    *uri_out = NULL;

2669 2670 2671
    /* 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
2672 2673 2674
     * 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".
2675 2676 2677 2678 2679 2680
     *
     * 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) {
2681
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2682
            goto cleanup;
2683

2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694
        if (migrateHost != NULL) {
            if (virSocketAddrIsNumeric(migrateHost) &&
                virSocketAddrParse(NULL, migrateHost, AF_UNSPEC) < 0)
                goto cleanup;

           if (VIR_STRDUP(hostname, migrateHost) < 0)
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
2695 2696

        if (STRPREFIX(hostname, "localhost")) {
2697 2698 2699
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
2700 2701 2702 2703 2704
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
2705
         * compatibility with old targets. We at least make the
2706 2707 2708
         * new targets accept both syntaxes though.
         */
        /* Caller frees */
2709
        if (virAsprintf(uri_out, "tcp:%s:%d", hostname, port) < 0)
2710 2711 2712 2713 2714 2715
            goto cleanup;
    } else {
        /* Check the URI starts with "tcp:".  We will escape the
         * URI when passing it to the qemu monitor, so bad
         * characters in hostname part don't matter.
         */
J
Ján Tomko 已提交
2716
        if (!(p = STRSKIP(uri_in, "tcp:"))) {
2717 2718 2719
            virReportError(VIR_ERR_INVALID_ARG, "%s",
                           _("only tcp URIs are supported for KVM/QEMU"
                             " migrations"));
2720 2721 2722
            goto cleanup;
        }

J
Ján Tomko 已提交
2723 2724
        /* Convert uri_in to well-formed URI with // after tcp: */
        if (!(STRPREFIX(uri_in, "tcp://"))) {
2725
            well_formed_uri = false;
2726
            if (virAsprintf(&uri_str, "tcp://%s", p) < 0)
J
Ján Tomko 已提交
2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745
                goto cleanup;
        }

        uri = virURIParse(uri_str ? uri_str : uri_in);
        VIR_FREE(uri_str);

        if (uri == NULL) {
            virReportError(VIR_ERR_INVALID_ARG, _("unable to parse URI: %s"),
                           uri_in);
            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) {
2746
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
2747
                goto cleanup;
2748

2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759
            if (well_formed_uri) {
                uri->port = port;

                /* Caller frees */
                if (!(*uri_out = virURIFormat(uri)))
                    goto cleanup;
            } else {
                /* Caller frees */
                if (virAsprintf(uri_out, "%s:%d", uri_in, port) < 0)
                    goto cleanup;
            }
2760 2761

        } else {
2762 2763
            port = uri->port;
            autoPort = false;
2764 2765 2766 2767 2768 2769
        }
    }

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

2770
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
2771
                                  cookieout, cookieoutlen, def, origname,
2772
                                  NULL, port, autoPort, listenAddress, flags);
2773
 cleanup:
2774
    virURIFree(uri);
2775
    VIR_FREE(hostname);
2776
    virObjectUnref(cfg);
2777
    if (ret != 0) {
2778
        VIR_FREE(*uri_out);
2779 2780 2781
        if (autoPort)
            virPortAllocatorRelease(driver->migrationPorts, port);
    }
2782 2783 2784 2785
    return ret;
}


2786 2787 2788
virDomainDefPtr
qemuMigrationPrepareDef(virQEMUDriverPtr driver,
                        const char *dom_xml,
2789 2790
                        const char *dname,
                        char **origname)
2791 2792 2793
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
2794
    char *name = NULL;
2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810

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

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

    if (!(def = virDomainDefParseString(dom_xml, caps, driver->xmlopt,
                                        QEMU_EXPECTED_VIRT_TYPES,
                                        VIR_DOMAIN_XML_INACTIVE)))
        goto cleanup;

    if (dname) {
2811
        name = def->name;
2812 2813 2814 2815 2816 2817
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

2818
 cleanup:
2819
    virObjectUnref(caps);
2820 2821 2822 2823
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
2824 2825 2826 2827
    return def;
}


2828 2829 2830 2831 2832 2833 2834 2835 2836 2837
static int
qemuMigrationConfirmPhase(virQEMUDriverPtr driver,
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *cookiein,
                          int cookieinlen,
                          unsigned int flags,
                          int retcode)
{
    qemuMigrationCookiePtr mig;
2838
    virObjectEventPtr event = NULL;
2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871
    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);

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen, 0)))
        goto cleanup;

    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

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

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

2872
        event = virDomainEventLifecycleNewFromObj(vm,
2873 2874 2875 2876 2877 2878 2879
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
    } else {

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

2880 2881 2882 2883
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
2884 2885 2886 2887 2888 2889 2890 2891
        }

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

2892
 done:
2893 2894 2895
    qemuMigrationCookieFree(mig);
    rv = 0;

2896
 cleanup:
2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911
    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;
2912
    qemuMigrationJobPhase phase;
2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926
    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);
2927 2928
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943

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

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

2944
 cleanup:
2945 2946 2947 2948 2949 2950 2951
    if (vm)
        virObjectUnlock(vm);
    virObjectUnref(cfg);
    return ret;
}


2952 2953
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
2954
    MIGRATION_DEST_CONNECT_HOST,
2955
    MIGRATION_DEST_UNIX,
2956
    MIGRATION_DEST_FD,
2957
};
2958

2959 2960 2961 2962
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
2963

2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
            const char *name;
            int port;
        } host;

        struct {
2975
            char *file;
2976 2977
            int sock;
        } unix_socket;
2978 2979 2980 2981 2982

        struct {
            int qemu;
            int local;
        } fd;
2983 2984 2985 2986 2987 2988 2989
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
2990 2991 2992

#define TUNNEL_SEND_BUF_SIZE 65536

2993 2994 2995 2996 2997 2998 2999
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3000 3001
    int wakeupRecvFD;
    int wakeupSendFD;
3002 3003 3004
};

static void qemuMigrationIOFunc(void *arg)
3005
{
3006
    qemuMigrationIOThreadPtr data = arg;
3007 3008 3009 3010 3011 3012 3013
    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);
3014

3015
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3016
        goto abrt;
3017

3018 3019 3020
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3021
    for (;;) {
3022 3023 3024 3025 3026 3027 3028 3029 3030 3031
        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;
3032
            virReportSystemError(errno, "%s",
3033 3034
                                 _("poll failed in migration tunnel"));
            goto abrt;
3035
        }
3036 3037 3038 3039 3040 3041 3042

        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");
3043
            break;
3044
        }
3045

3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061
        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;
            }
3062 3063
        }

3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080
        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;
            }
        }
    }
3081

3082 3083
    if (virStreamFinish(data->st) < 0)
        goto error;
3084

3085 3086
    VIR_FREE(buffer);

3087 3088
    return;

3089
 abrt:
3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

3101
 error:
3102 3103
    virCopyLastError(&data->err);
    virResetLastError();
3104
    VIR_FREE(buffer);
3105 3106 3107 3108 3109 3110 3111
}


static qemuMigrationIOThreadPtr
qemuMigrationStartTunnel(virStreamPtr st,
                         int sock)
{
3112 3113
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
3114

3115 3116 3117 3118
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
3119 3120
    }

3121
    if (VIR_ALLOC(io) < 0)
3122
        goto error;
3123

3124 3125
    io->st = st;
    io->sock = sock;
3126 3127
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
3128 3129 3130 3131 3132 3133

    if (virThreadCreate(&io->thread, true,
                        qemuMigrationIOFunc,
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
3134
        goto error;
3135 3136 3137
    }

    return io;
3138

3139
 error:
3140 3141 3142 3143
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
3144 3145 3146
}

static int
3147
qemuMigrationStopTunnel(qemuMigrationIOThreadPtr io, bool error)
3148 3149
{
    int rv = -1;
3150 3151 3152 3153 3154 3155 3156 3157 3158
    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;
    }

3159 3160 3161 3162
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
3163 3164 3165 3166
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
3167 3168 3169 3170 3171 3172
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

3173
 cleanup:
3174 3175
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
3176 3177
    VIR_FREE(io);
    return rv;
3178 3179
}

3180
static int
3181
qemuMigrationConnect(virQEMUDriverPtr driver,
3182 3183 3184 3185 3186 3187 3188 3189 3190
                     virDomainObjPtr vm,
                     qemuMigrationSpecPtr spec)
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
3191
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
3192 3193 3194 3195 3196 3197 3198 3199 3200
        return -1;

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

    if (virSecurityManagerSetSocketLabel(driver->securityManager, vm->def) < 0)
        goto cleanup;
    if (virNetSocketNewConnectTCP(host, port, &sock) == 0) {
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
3201
        virObjectUnref(sock);
3202 3203 3204 3205 3206
    }
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0 ||
        spec->dest.fd.qemu == -1)
        goto cleanup;

3207 3208 3209 3210 3211 3212 3213
    /* 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;
    }

3214 3215
    ret = 0;

3216
 cleanup:
3217 3218 3219 3220 3221 3222
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

3223
static int
3224
qemuMigrationRun(virQEMUDriverPtr driver,
3225 3226 3227 3228 3229 3230 3231
                 virDomainObjPtr vm,
                 const char *cookiein,
                 int cookieinlen,
                 char **cookieout,
                 int *cookieoutlen,
                 unsigned long flags,
                 unsigned long resource,
3232
                 qemuMigrationSpecPtr spec,
3233 3234
                 virConnectPtr dconn,
                 const char *graphicsuri)
3235
{
3236
    int ret = -1;
3237 3238
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3239
    qemuMigrationCookiePtr mig = NULL;
3240
    qemuMigrationIOThreadPtr iothread = NULL;
3241
    int fd = -1;
3242
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
3243
    virErrorPtr orig_err = NULL;
3244
    unsigned int cookieFlags = 0;
3245
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3246
    int rc;
3247 3248 3249

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
3250
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s",
3251 3252
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
3253 3254
              spec, spec->destType, spec->fwdType, dconn,
              NULLSTR(graphicsuri));
3255

3256 3257 3258 3259 3260 3261 3262 3263 3264 3265
    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;
    }

3266 3267
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
3268 3269 3270 3271
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
3272 3273 3274
        return -1;
    }

3275 3276 3277
    mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                 cookieFlags | QEMU_MIGRATION_COOKIE_GRAPHICS);
    if (!mig)
3278 3279
        goto cleanup;

3280
    if (qemuDomainMigrateGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
3281 3282
        VIR_WARN("unable to provide data for graphics client relocation");

3283 3284 3285 3286 3287 3288 3289
    /* this will update migrate_flags on success */
    if (qemuMigrationDriveMirror(driver, vm, mig, spec->dest.host.name,
                                 migrate_speed, &migrate_flags) < 0) {
        /* error reported by helper func */
        goto cleanup;
    }

3290
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
3291 3292 3293 3294 3295 3296
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

3297 3298 3299 3300 3301
    if (flags & VIR_MIGRATE_COMPRESSED &&
        qemuMigrationSetCompression(driver, vm,
                                    QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3302 3303 3304 3305 3306
    if (flags & VIR_MIGRATE_AUTO_CONVERGE &&
        qemuMigrationSetAutoConverge(driver, vm,
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

3307 3308
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
3309 3310
        goto cleanup;

3311 3312 3313 3314
    if (priv->job.asyncAbort) {
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
         * priv->job.asyncAbort will not change */
3315
        qemuDomainObjExitMonitor(driver, vm);
3316 3317 3318 3319 3320 3321
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
        goto cleanup;
    }

3322
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0) {
3323
        qemuDomainObjExitMonitor(driver, vm);
3324 3325
        goto cleanup;
    }
3326

3327 3328
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
3329
        qemuMigrationConnect(driver, vm, spec) < 0) {
3330
        qemuDomainObjExitMonitor(driver, vm);
3331
        goto cleanup;
3332
    }
3333

3334 3335 3336 3337 3338 3339 3340
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
        ret = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
                                       spec->dest.host.name,
                                       spec->dest.host.port);
        break;

3341 3342 3343 3344
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

3345
    case MIGRATION_DEST_UNIX:
3346
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX)) {
3347 3348 3349 3350 3351 3352 3353 3354 3355
            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;
3356 3357

    case MIGRATION_DEST_FD:
3358
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3359
            fd = spec->dest.fd.local;
3360 3361
            spec->dest.fd.local = -1;
        }
3362 3363 3364 3365
        ret = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                     spec->dest.fd.qemu);
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
3366
    }
3367
    qemuDomainObjExitMonitor(driver, vm);
3368
    if (ret < 0)
3369 3370
        goto cleanup;
    ret = -1;
3371 3372

    if (!virDomainObjIsActive(vm)) {
3373 3374
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
3375 3376 3377 3378 3379 3380
        goto cleanup;
    }

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

3381 3382 3383 3384 3385 3386 3387 3388
    if (spec->destType == MIGRATION_DEST_UNIX) {
        /* It is also possible that the migrate didn't fail initially, but
         * rather failed later on.  Check its status before waiting for a
         * connection from qemu which may never be initiated.
         */
        if (qemuMigrationUpdateJobStatus(driver, vm, _("migration job"),
                                         QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            goto cancel;
3389

3390 3391 3392 3393 3394 3395 3396
        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;
        }
3397 3398
    }

3399 3400
    if (spec->fwdType != MIGRATION_FWD_DIRECT &&
        !(iothread = qemuMigrationStartTunnel(spec->fwd.stream, fd)))
3401 3402
        goto cancel;

3403 3404 3405 3406 3407 3408
    rc = qemuMigrationWaitForCompletion(driver, vm,
                                        QEMU_ASYNC_JOB_MIGRATION_OUT,
                                        dconn, abort_on_error);
    if (rc == -2)
        goto cancel;
    else if (rc == -1)
3409
        goto cleanup;
3410

3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423
    /* When migration completed, QEMU will have paused the
     * CPUs for us, but unless we're using the JSON monitor
     * we won't have been notified of this, so might still
     * think we're running. For v2 protocol this doesn't
     * matter because we'll kill the VM soon, but for v3
     * this is important because we stay paused until the
     * confirm3 step, but need to release the lock state
     */
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

3424
    ret = 0;
3425

3426
 cleanup:
3427 3428 3429
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

3430 3431 3432
    /* cancel any outstanding NBD jobs */
    qemuMigrationCancelDriveMirror(mig, driver, vm);

3433
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
3434
        if (iothread && qemuMigrationStopTunnel(iothread, ret < 0) < 0)
3435
            ret = -1;
3436
        VIR_FORCE_CLOSE(fd);
3437
    }
3438

3439 3440 3441
    cookieFlags |= QEMU_MIGRATION_COOKIE_NETWORK;
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookieFlags |= QEMU_MIGRATION_COOKIE_PERSISTENT;
3442
    if (ret == 0 &&
3443 3444
        qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
3445
        VIR_WARN("Unable to encode migration cookie");
3446
    }
3447

3448 3449
    qemuMigrationCookieFree(mig);

3450 3451 3452 3453 3454
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

3455 3456
    return ret;

3457
 cancel:
3458 3459
    orig_err = virSaveLastError();

3460
    if (virDomainObjIsActive(vm)) {
3461 3462
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
3463
            qemuMonitorMigrateCancel(priv->mon);
3464
            qemuDomainObjExitMonitor(driver, vm);
3465
        }
3466
    }
3467 3468 3469 3470 3471 3472
    goto cleanup;
}

/* Perform migration using QEMU's native TCP migrate support,
 * not encrypted obviously
 */
3473
static int doNativeMigrate(virQEMUDriverPtr driver,
3474 3475 3476 3477 3478 3479 3480
                           virDomainObjPtr vm,
                           const char *uri,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
3481
                           unsigned long resource,
3482 3483
                           virConnectPtr dconn,
                           const char *graphicsuri)
3484
{
3485
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
3486
    virURIPtr uribits = NULL;
3487
    int ret = -1;
3488 3489 3490
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
3491 3492
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "graphicsuri=%s",
3493
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
3494 3495
              cookieout, cookieoutlen, flags, resource,
              NULLSTR(graphicsuri));
3496 3497

    if (STRPREFIX(uri, "tcp:") && !STRPREFIX(uri, "tcp://")) {
3498
        char *tmp;
3499
        /* HACK: source host generates bogus URIs, so fix them up */
3500
        if (virAsprintf(&tmp, "tcp://%s", uri + strlen("tcp:")) < 0)
3501
            return -1;
M
Martin Kletzander 已提交
3502
        uribits = virURIParse(tmp);
3503
        VIR_FREE(tmp);
3504
    } else {
M
Martin Kletzander 已提交
3505
        uribits = virURIParse(uri);
3506
    }
3507
    if (!uribits)
3508 3509
        return -1;

3510
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD))
3511 3512
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
3513
        spec.destType = MIGRATION_DEST_HOST;
3514 3515 3516
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
3517

3518
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
3519 3520
                           cookieoutlen, flags, resource, &spec, dconn,
                           graphicsuri);
3521 3522 3523 3524

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

3525
    virURIFree(uribits);
3526 3527 3528 3529 3530

    return ret;
}


3531
static int doTunnelMigrate(virQEMUDriverPtr driver,
3532 3533 3534 3535 3536 3537 3538
                           virDomainObjPtr vm,
                           virStreamPtr st,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
3539
                           unsigned long resource,
3540 3541
                           virConnectPtr dconn,
                           const char *graphicsuri)
3542 3543
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
3544
    virNetSocketPtr sock = NULL;
3545 3546
    int ret = -1;
    qemuMigrationSpec spec;
3547
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3548 3549

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
3550 3551
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "graphicsuri=%s",
3552
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
3553 3554
              cookieout, cookieoutlen, flags, resource,
              NULLSTR(graphicsuri));
3555

3556 3557 3558
    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)) {
3559 3560
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Source qemu is too old to support tunnelled migration"));
3561
        virObjectUnref(cfg);
3562
        return -1;
3563 3564 3565 3566 3567
    }

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

3568
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD)) {
3569 3570 3571 3572 3573 3574
        int fds[2];

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

3575
        if (pipe2(fds, O_CLOEXEC) == 0) {
3576 3577 3578 3579
            spec.dest.fd.qemu = fds[1];
            spec.dest.fd.local = fds[0];
        }
        if (spec.dest.fd.qemu == -1 ||
3580
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
3581
                                              spec.dest.fd.qemu) < 0) {
3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592
            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",
3593
                        cfg->libDir, vm->def->name) < 0)
3594 3595 3596
            goto cleanup;

        if (virNetSocketNewListenUNIX(spec.dest.unix_socket.file, 0700,
3597
                                      cfg->user, cfg->group,
3598 3599 3600 3601 3602 3603 3604
                                      &sock) < 0 ||
            virNetSocketListen(sock, 1) < 0)
            goto cleanup;

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

3605
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
3606 3607
                           cookieoutlen, flags, resource, &spec, dconn,
                           graphicsuri);
3608

3609
 cleanup:
3610 3611 3612 3613
    if (spec.destType == MIGRATION_DEST_FD) {
        VIR_FORCE_CLOSE(spec.dest.fd.qemu);
        VIR_FORCE_CLOSE(spec.dest.fd.local);
    } else {
3614
        virObjectUnref(sock);
3615 3616
        VIR_FREE(spec.dest.unix_socket.file);
    }
3617

3618
    virObjectUnref(cfg);
3619 3620 3621 3622
    return ret;
}


3623 3624 3625 3626
/* 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 */
3627
static int doPeer2PeerMigrate2(virQEMUDriverPtr driver,
3628
                               virConnectPtr sconn ATTRIBUTE_UNUSED,
3629 3630
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
3631
                               const char *dconnuri,
3632 3633 3634
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
3635 3636 3637
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
3638
    char *cookie = NULL;
3639 3640 3641
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
3642
    bool cancelled;
3643
    virStreamPtr st = NULL;
3644
    unsigned long destflags;
3645

3646
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
3647
              "flags=%lx, dname=%s, resource=%lu",
3648 3649
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
3650

3651 3652 3653 3654 3655
    /* 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,
3656 3657
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
3658 3659 3660 3661 3662
        return -1;

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

3663 3664
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
3665

3666 3667 3668 3669 3670 3671 3672 3673 3674 3675
    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;

3676
        qemuDomainObjEnterRemote(vm);
3677
        ret = dconn->driver->domainMigratePrepareTunnel
3678
            (dconn, st, destflags, dname, resource, dom_xml);
3679
        qemuDomainObjExitRemote(vm);
3680
    } else {
3681
        qemuDomainObjEnterRemote(vm);
3682 3683
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
3684
             destflags, dname, resource, dom_xml);
3685
        qemuDomainObjExitRemote(vm);
3686 3687 3688
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
3689 3690 3691
        goto cleanup;

    /* the domain may have shutdown or crashed while we had the locks dropped
3692
     * in qemuDomainObjEnterRemote, so check again
3693 3694
     */
    if (!virDomainObjIsActive(vm)) {
3695 3696
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
3697 3698 3699
        goto cleanup;
    }

3700 3701
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
3702 3703
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
3704
        cancelled = true;
3705
        orig_err = virSaveLastError();
3706
        goto finish;
3707 3708
    }

3709 3710 3711 3712
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
3713
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
3714
    if (flags & VIR_MIGRATE_TUNNELLED)
3715 3716
        ret = doTunnelMigrate(driver, vm, st,
                              NULL, 0, NULL, NULL,
3717
                              flags, resource, dconn, NULL);
3718 3719 3720 3721
    else
        ret = doNativeMigrate(driver, vm, uri_out,
                              cookie, cookielen,
                              NULL, NULL, /* No out cookie with v2 migration */
3722
                              flags, resource, dconn, NULL);
3723 3724 3725 3726

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
3727

3728 3729 3730
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
3731
    cancelled = ret < 0;
3732

3733
 finish:
3734 3735 3736 3737
    /* 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.
     */
3738
    dname = dname ? dname : vm->def->name;
3739
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
3740
    qemuDomainObjEnterRemote(vm);
3741
    ddomain = dconn->driver->domainMigrateFinish2
3742
        (dconn, dname, cookie, cookielen,
3743
         uri_out ? uri_out : dconnuri, destflags, cancelled);
3744
    qemuDomainObjExitRemote(vm);
3745 3746
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
3747

3748
 cleanup:
3749
    if (ddomain) {
3750
        virObjectUnref(ddomain);
3751 3752 3753 3754
        ret = 0;
    } else {
        ret = -1;
    }
3755

3756
    virObjectUnref(st);
3757 3758 3759 3760 3761 3762

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
3763
    VIR_FREE(cookie);
3764 3765

    return ret;
3766 3767 3768
}


3769 3770 3771 3772
/* 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 */
3773 3774 3775 3776 3777 3778 3779 3780 3781
static int
doPeer2PeerMigrate3(virQEMUDriverPtr driver,
                    virConnectPtr sconn,
                    virConnectPtr dconn,
                    const char *dconnuri,
                    virDomainObjPtr vm,
                    const char *xmlin,
                    const char *dname,
                    const char *uri,
3782
                    const char *graphicsuri,
3783
                    const char *listenAddress,
3784 3785 3786
                    unsigned long long bandwidth,
                    bool useParams,
                    unsigned long flags)
3787 3788 3789 3790 3791 3792 3793 3794 3795 3796
{
    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;
3797
    bool cancelled = true;
3798
    virStreamPtr st = NULL;
3799
    unsigned long destflags;
3800 3801 3802 3803 3804
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
3805 3806
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
              "bandwidth=%llu, useParams=%d, flags=%lx",
3807
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
3808 3809
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
              NULLSTR(listenAddress), bandwidth, useParams, flags);
3810

3811 3812 3813 3814 3815
    /* 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.  */

3816 3817
    dom_xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
                                      &cookieout, &cookieoutlen, flags);
3818 3819 3820
    if (!dom_xml)
        goto cleanup;

3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840
    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;
3841 3842 3843 3844 3845 3846

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
3847 3848 3849 3850 3851
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
3852 3853
    }

3854 3855 3856
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

3857 3858
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
3859

3860 3861 3862 3863 3864 3865 3866 3867 3868
    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;

3869
        qemuDomainObjEnterRemote(vm);
3870 3871 3872 3873 3874 3875 3876 3877 3878
        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);
        }
3879
        qemuDomainObjExitRemote(vm);
3880
    } else {
3881
        qemuDomainObjEnterRemote(vm);
3882 3883 3884 3885 3886 3887 3888 3889 3890
        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);
        }
3891
        qemuDomainObjExitRemote(vm);
3892 3893 3894 3895 3896
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

L
liguang 已提交
3897 3898 3899 3900
    if (flags & VIR_MIGRATE_OFFLINE) {
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
3901
        cancelled = false;
L
liguang 已提交
3902 3903 3904
        goto finish;
    }

3905 3906 3907 3908
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
3909 3910
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
3911
            goto finish;
3912
        }
3913
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
3914 3915
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
3916
        orig_err = virSaveLastError();
3917 3918 3919 3920 3921 3922 3923 3924
        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.
     */
3925
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
3926
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
3927 3928 3929 3930 3931
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
3932
    if (flags & VIR_MIGRATE_TUNNELLED) {
3933 3934 3935
        ret = doTunnelMigrate(driver, vm, st,
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
3936
                              flags, bandwidth, dconn, graphicsuri);
3937 3938
    } else {
        ret = doNativeMigrate(driver, vm, uri,
3939 3940
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
3941
                              flags, bandwidth, dconn, graphicsuri);
3942
    }
3943 3944

    /* Perform failed. Make sure Finish doesn't overwrite the error */
3945
    if (ret < 0) {
3946
        orig_err = virSaveLastError();
3947 3948 3949 3950
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
3951 3952 3953 3954

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
3955
    cancelled = ret < 0;
3956

3957
 finish:
3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969
    /*
     * 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;
3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992

    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);
    }
3993 3994
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
3995

3996 3997 3998 3999 4000 4001 4002
    /* 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.
4003
     */
4004
    cancelled = ddomain == NULL;
4005

4006 4007 4008 4009 4010 4011
    /* 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();

4012 4013 4014 4015
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
4016
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
4017 4018 4019 4020 4021
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4022 4023 4024
    ret = qemuMigrationConfirmPhase(driver, sconn, vm,
                                    cookiein, cookieinlen,
                                    flags, cancelled);
4025 4026 4027 4028
    /* 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.
     */
4029 4030 4031
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
4032 4033 4034

 cleanup:
    if (ddomain) {
4035
        virObjectUnref(ddomain);
4036 4037 4038 4039 4040
        ret = 0;
    } else {
        ret = -1;
    }

4041
    virObjectUnref(st);
4042 4043 4044 4045 4046 4047 4048 4049

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
4050
    virTypedParamsFree(params, nparams);
4051 4052 4053 4054
    return ret;
}


4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


4067
static int doPeer2PeerMigrate(virQEMUDriverPtr driver,
4068
                              virConnectPtr sconn,
4069
                              virDomainObjPtr vm,
4070
                              const char *xmlin,
4071
                              const char *dconnuri,
4072
                              const char *uri,
4073
                              const char *graphicsuri,
4074
                              const char *listenAddress,
4075 4076
                              unsigned long flags,
                              const char *dname,
4077 4078
                              unsigned long resource,
                              bool *v3proto)
4079 4080 4081 4082
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
4083
    virErrorPtr orig_err = NULL;
4084
    bool offline = false;
4085
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4086
    bool useParams;
4087

4088
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
4089 4090
              "uri=%s, graphicsuri=%s, listenAddress=%s, flags=%lx, "
              "dname=%s, resource=%lu",
4091
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4092 4093
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
              flags, NULLSTR(dname), resource);
4094 4095 4096 4097 4098

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

4099
    qemuDomainObjEnterRemote(vm);
4100
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
4101
    qemuDomainObjExitRemote(vm);
4102
    if (dconn == NULL) {
4103
        virReportError(VIR_ERR_OPERATION_FAILED,
4104 4105
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
4106
        virObjectUnref(cfg);
4107 4108 4109
        return -1;
    }

4110 4111
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
4112 4113
        goto cleanup;

4114
    qemuDomainObjEnterRemote(vm);
4115 4116
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
4117
        /* v3proto reflects whether the caller used Perform3, but with
4118
         * p2p migrate, regardless of whether Perform2 or Perform3
4119 4120 4121 4122
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
4123 4124
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
L
liguang 已提交
4125 4126 4127
    if (flags & VIR_MIGRATE_OFFLINE)
        offline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                           VIR_DRV_FEATURE_MIGRATION_OFFLINE);
4128
    qemuDomainObjExitRemote(vm);
4129

4130
    if (!p2p) {
4131 4132
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
4133 4134 4135
        goto cleanup;
    }

4136 4137
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
4138
    if (!useParams && graphicsuri) {
4139 4140 4141 4142 4143 4144
        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
4145 4146 4147 4148 4149 4150 4151
    if (flags & VIR_MIGRATE_OFFLINE && !offline) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

4152
    /* domain may have been stopped while we were talking to remote daemon */
L
liguang 已提交
4153
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4154 4155
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4156 4157 4158
        goto cleanup;
    }

4159 4160 4161 4162 4163 4164 4165
    /* 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;

4166 4167
    if (*v3proto) {
        ret = doPeer2PeerMigrate3(driver, sconn, dconn, dconnuri, vm, xmlin,
4168 4169
                                  dname, uri, graphicsuri, listenAddress,
                                  resource, useParams, flags);
4170
    } else {
4171
        ret = doPeer2PeerMigrate2(driver, sconn, dconn, vm,
4172
                                  dconnuri, flags, dname, resource);
4173
    }
4174

4175
 cleanup:
4176
    orig_err = virSaveLastError();
4177
    qemuDomainObjEnterRemote(vm);
4178
    virObjectUnref(dconn);
4179
    qemuDomainObjExitRemote(vm);
4180 4181 4182 4183
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4184
    virObjectUnref(cfg);
4185 4186 4187 4188
    return ret;
}


4189 4190 4191 4192 4193 4194
/*
 * 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
4195
qemuMigrationPerformJob(virQEMUDriverPtr driver,
4196 4197 4198 4199 4200
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dconnuri,
                        const char *uri,
4201
                        const char *graphicsuri,
4202
                        const char *listenAddress,
4203
                        const char *cookiein,
4204 4205 4206 4207 4208 4209 4210
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
4211
{
4212
    virObjectEventPtr event = NULL;
4213
    int ret = -1;
4214
    virErrorPtr orig_err = NULL;
4215
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4216
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
4217

4218
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4219 4220
        goto cleanup;

L
liguang 已提交
4221
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4222 4223
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
4224 4225 4226
        goto endjob;
    }

4227
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
4228
        goto endjob;
4229

4230
    if (!(flags & VIR_MIGRATE_UNSAFE) && !qemuMigrationIsSafe(vm->def))
4231
        goto endjob;
4232

4233
    qemuMigrationStoreDomainState(vm);
4234 4235

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
4236
        ret = doPeer2PeerMigrate(driver, conn, vm, xmlin,
4237 4238
                                 dconnuri, uri, graphicsuri, listenAddress,
                                 flags, dname, resource, &v3proto);
4239
    } else {
4240 4241 4242
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
        ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                              cookieout, cookieoutlen,
4243
                              flags, resource, NULL, NULL);
4244
    }
4245 4246
    if (ret < 0)
        goto endjob;
4247

4248 4249 4250 4251
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
4252
    if (!v3proto) {
4253 4254
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4255
        virDomainAuditStop(vm, "migrated");
4256
        event = virDomainEventLifecycleNewFromObj(vm,
4257 4258
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
4259 4260
    }

4261
 endjob:
4262 4263 4264
    if (ret < 0)
        orig_err = virSaveLastError();

4265
    if (qemuMigrationRestoreDomainState(conn, vm)) {
4266
        event = virDomainEventLifecycleNewFromObj(vm,
4267 4268 4269
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
4270

4271
    if (!qemuMigrationJobFinish(driver, vm)) {
4272 4273 4274 4275 4276
        vm = NULL;
    } else if (!virDomainObjIsActive(vm) &&
               (!vm->persistent ||
                (ret == 0 && (flags & VIR_MIGRATE_UNDEFINE_SOURCE)))) {
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
4277
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
4278
        qemuDomainRemoveInactive(driver, vm);
4279 4280 4281
        vm = NULL;
    }

4282 4283 4284 4285 4286
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4287
 cleanup:
4288
    if (vm)
4289
        virObjectUnlock(vm);
4290 4291
    if (event)
        qemuDomainEventQueue(driver, event);
4292
    virObjectUnref(cfg);
4293 4294 4295 4296 4297 4298 4299
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
4300
qemuMigrationPerformPhase(virQEMUDriverPtr driver,
4301 4302 4303
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *uri,
4304
                          const char *graphicsuri,
4305 4306 4307 4308 4309 4310 4311
                          const char *cookiein,
                          int cookieinlen,
                          char **cookieout,
                          int *cookieoutlen,
                          unsigned long flags,
                          unsigned long resource)
{
4312
    virObjectEventPtr event = NULL;
4313
    int ret = -1;
4314
    bool hasrefs;
4315 4316 4317 4318 4319 4320 4321 4322 4323 4324

    /* 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);
4325 4326
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
4327 4328 4329

    ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                          cookieout, cookieoutlen,
4330
                          flags, resource, NULL, graphicsuri);
4331

4332 4333 4334 4335 4336
    if (ret < 0) {
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
4337
        }
4338
        goto endjob;
4339
    }
4340 4341 4342

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

4343 4344
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                             qemuMigrationCleanup) < 0)
4345 4346
        goto endjob;

4347
 endjob:
4348
    if (ret < 0)
4349
        hasrefs = qemuMigrationJobFinish(driver, vm);
4350
    else
4351 4352
        hasrefs = qemuMigrationJobContinue(vm);
    if (!hasrefs) {
4353 4354
        vm = NULL;
    } else if (!virDomainObjIsActive(vm) && !vm->persistent) {
4355
        qemuDomainRemoveInactive(driver, vm);
4356
        vm = NULL;
4357
    }
4358

4359
 cleanup:
4360
    if (vm)
4361
        virObjectUnlock(vm);
4362 4363 4364 4365 4366
    if (event)
        qemuDomainEventQueue(driver, event);
    return ret;
}

4367
int
4368
qemuMigrationPerform(virQEMUDriverPtr driver,
4369 4370 4371 4372 4373
                     virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *xmlin,
                     const char *dconnuri,
                     const char *uri,
4374
                     const char *graphicsuri,
4375
                     const char *listenAddress,
4376 4377 4378 4379 4380 4381 4382 4383 4384 4385
                     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, "
4386
              "uri=%s, graphicsuri=%s, listenAddress=%s"
4387 4388
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
              "flags=%lx, dname=%s, resource=%lu, v3proto=%d",
4389
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
4390
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
4391 4392
              NULLSTR(cookiein), cookieinlen, cookieout, cookieoutlen,
              flags, NULLSTR(dname), resource, v3proto);
4393 4394 4395

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
4396 4397
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
4398 4399 4400 4401
            return -1;
        }

        return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri, uri,
4402 4403
                                       graphicsuri, listenAddress,
                                       cookiein, cookieinlen,
4404 4405
                                       cookieout, cookieoutlen,
                                       flags, dname, resource, v3proto);
4406 4407
    } else {
        if (dconnuri) {
4408 4409
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
4410 4411 4412 4413 4414
            return -1;
        }

        if (v3proto) {
            return qemuMigrationPerformPhase(driver, conn, vm, uri,
4415
                                             graphicsuri,
4416 4417
                                             cookiein, cookieinlen,
                                             cookieout, cookieoutlen,
4418
                                             flags, resource);
4419 4420
        } else {
            return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri,
4421
                                           uri, graphicsuri, listenAddress,
4422
                                           cookiein, cookieinlen,
4423 4424 4425 4426 4427
                                           cookieout, cookieoutlen, flags,
                                           dname, resource, v3proto);
        }
    }
}
4428

4429
static int
4430 4431
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def)
{
4432
    size_t i;
4433 4434 4435 4436 4437
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
4438
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4439
            if (virNetDevVPortProfileAssociate(net->ifname,
4440
                                               virDomainNetGetActualVirtPortProfile(net),
4441
                                               &net->mac,
4442
                                               virDomainNetGetActualDirectDev(net),
4443
                                               -1,
4444
                                               def->uuid,
4445 4446
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
4447 4448 4449
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
4450
                goto err_exit;
4451 4452
            }
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
4453

4454
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
4455 4456 4457 4458
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
4459 4460 4461 4462
        }
        last_good_net = i;
    }

4463
    return 0;
4464

4465
 err_exit:
4466
    for (i = 0; last_good_net != -1 && i < last_good_net; i++) {
4467
        net = def->nets[i];
4468
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4469
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
4470
                                                           virDomainNetGetActualVirtPortProfile(net),
4471
                                                           &net->mac,
4472
                                                           virDomainNetGetActualDirectDev(net),
4473
                                                           -1,
4474
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
4475 4476
        }
    }
4477
    return -1;
4478 4479 4480 4481
}


virDomainPtr
4482
qemuMigrationFinish(virQEMUDriverPtr driver,
4483 4484
                    virConnectPtr dconn,
                    virDomainObjPtr vm,
4485 4486 4487 4488
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
4489
                    unsigned long flags,
4490 4491
                    int retcode,
                    bool v3proto)
4492 4493
{
    virDomainPtr dom = NULL;
4494
    virObjectEventPtr event = NULL;
4495
    bool newVM = true;
4496
    qemuMigrationCookiePtr mig = NULL;
4497
    virErrorPtr orig_err = NULL;
4498
    int cookie_flags = 0;
J
Jiri Denemark 已提交
4499
    qemuDomainObjPrivatePtr priv = vm->privateData;
4500
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4501
    virCapsPtr caps = NULL;
4502
    unsigned short port;
4503

4504
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
4505
              "cookieout=%p, cookieoutlen=%p, flags=%lx, retcode=%d",
4506 4507
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
4508

4509 4510 4511
    port = priv->migrationPort;
    priv->migrationPort = 0;

4512 4513 4514
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

4515
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
4516 4517
        goto cleanup;

4518 4519 4520
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
4521

4522 4523
    qemuDomainCleanupRemove(vm, qemuMigrationPrepareCleanup);

4524
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK;
4525 4526 4527 4528 4529
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookie_flags |= QEMU_MIGRATION_COOKIE_PERSISTENT;

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein,
                                       cookieinlen, cookie_flags)))
4530
        goto endjob;
4531 4532 4533 4534 4535

    /* 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
4536
        if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
4537 4538
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("guest unexpectedly quit"));
4539
            goto endjob;
4540 4541
        }

L
liguang 已提交
4542 4543 4544 4545 4546
        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");
4547
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
4548 4549 4550 4551 4552 4553 4554
                                                 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");
4555
        }
4556

4557 4558
        qemuMigrationStopNBDServer(driver, vm, mig);

4559
        if (flags & VIR_MIGRATE_PERSIST_DEST) {
4560
            virDomainDefPtr vmdef;
4561
            if (vm->persistent)
4562
                newVM = false;
4563
            vm->persistent = 1;
4564 4565 4566
            if (mig->persistent)
                vm->newDef = vmdef = mig->persistent;
            else
4567
                vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm);
4568
            if (!vmdef || virDomainSaveConfig(cfg->configDir, vmdef) < 0) {
4569 4570 4571 4572 4573 4574 4575 4576 4577
                /* 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.
                 */
4578 4579

                /*
4580 4581
                 * However, in v3 protocol, the source VM is still available
                 * to restart during confirm() step, so we kill it off now.
4582 4583
                 */
                if (v3proto) {
L
liguang 已提交
4584 4585 4586 4587 4588
                    if (!(flags & VIR_MIGRATE_OFFLINE)) {
                        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                        VIR_QEMU_PROCESS_STOP_MIGRATED);
                        virDomainAuditStop(vm, "failed");
                    }
4589 4590
                    if (newVM)
                        vm->persistent = 0;
4591
                }
A
Alex Jia 已提交
4592
                if (!vmdef)
4593 4594
                    virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                                   _("can't get vmdef"));
4595 4596 4597
                goto endjob;
            }

4598
            event = virDomainEventLifecycleNewFromObj(vm,
4599 4600 4601 4602 4603 4604 4605 4606 4607
                                             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
4608
        if (!(flags & VIR_MIGRATE_PAUSED) && !(flags & VIR_MIGRATE_OFFLINE)) {
4609 4610 4611 4612
            /* 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
4613
            if (qemuProcessStartCPUs(driver, vm, dconn,
4614 4615
                                     VIR_DOMAIN_RUNNING_MIGRATED,
                                     QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
4616
                if (virGetLastError() == NULL)
4617 4618
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   "%s", _("resume operation failed"));
4619 4620 4621 4622 4623
                /* Need to save the current error, in case shutting
                 * down the process overwrites it
                 */
                orig_err = virSaveLastError();

4624 4625 4626 4627 4628 4629 4630 4631 4632
                /*
                 * 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) {
4633 4634
                    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                    VIR_QEMU_PROCESS_STOP_MIGRATED);
4635
                    virDomainAuditStop(vm, "failed");
4636
                    event = virDomainEventLifecycleNewFromObj(vm,
4637 4638 4639
                                                     VIR_DOMAIN_EVENT_STOPPED,
                                                     VIR_DOMAIN_EVENT_STOPPED_FAILED);
                }
4640 4641 4642 4643
                goto endjob;
            }
        }

4644
        dom = virGetDomain(dconn, vm->def->name, vm->def->uuid);
4645

L
liguang 已提交
4646
        if (!(flags & VIR_MIGRATE_OFFLINE)) {
4647
            event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
4648 4649 4650 4651 4652 4653 4654
                                             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);
4655
                event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
4656 4657 4658
                                                 VIR_DOMAIN_EVENT_SUSPENDED,
                                                 VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);
            }
4659
        }
L
liguang 已提交
4660 4661

        if (virDomainObjIsActive(vm) &&
4662
            virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
4663 4664 4665
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto endjob;
        }
4666 4667 4668

        /* Guest is successfully running, so cancel previous auto destroy */
        qemuProcessAutoDestroyRemove(driver, vm);
L
liguang 已提交
4669
    } else if (!(flags & VIR_MIGRATE_OFFLINE)) {
4670 4671
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
4672
        virDomainAuditStop(vm, "failed");
4673
        event = virDomainEventLifecycleNewFromObj(vm,
4674 4675 4676 4677
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_FAILED);
    }

4678 4679 4680
    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen, 0) < 0)
        VIR_WARN("Unable to encode migration cookie");

4681
 endjob:
E
Eric Blake 已提交
4682 4683 4684
    if (qemuMigrationJobFinish(driver, vm) == 0) {
        vm = NULL;
    } else if (!vm->persistent && !virDomainObjIsActive(vm)) {
4685
        qemuDomainRemoveInactive(driver, vm);
E
Eric Blake 已提交
4686
        vm = NULL;
4687
    }
4688

4689
 cleanup:
4690
    virPortAllocatorRelease(driver->migrationPorts, port);
J
Jiri Denemark 已提交
4691
    if (vm) {
4692 4693
        if (priv->mon)
            qemuMonitorSetDomainLog(priv->mon, -1);
J
Jiri Denemark 已提交
4694
        VIR_FREE(priv->origname);
4695
        virObjectUnlock(vm);
J
Jiri Denemark 已提交
4696
    }
4697 4698
    if (event)
        qemuDomainEventQueue(driver, event);
4699
    qemuMigrationCookieFree(mig);
4700 4701 4702 4703
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4704
    virObjectUnref(caps);
4705
    virObjectUnref(cfg);
4706 4707
    return dom;
}
4708

4709

4710
/* Helper function called while vm is active.  */
4711
int
4712
qemuMigrationToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
4713 4714
                    int fd, off_t offset, const char *path,
                    const char *compressor,
E
Eric Blake 已提交
4715
                    bool bypassSecurityDriver,
4716
                    qemuDomainAsyncJob asyncJob)
4717 4718 4719
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
4720
    int ret = -1;
4721
    bool restoreLabel = false;
4722 4723
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
4724
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
4725
    char *errbuf = NULL;
4726
    virErrorPtr orig_err = NULL;
4727 4728 4729 4730 4731

    /* 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,
4732 4733
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
4734
        qemuDomainObjExitMonitor(driver, vm);
4735
    }
4736

4737 4738 4739 4740 4741 4742 4743
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

4744
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
4745
        (!compressor || pipe(pipeFD) == 0)) {
4746
        /* All right! We can use fd migration, which means that qemu
4747 4748 4749
         * 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.  */
4750
        if (virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
4751
                                              compressor ? pipeFD[1] : fd) < 0)
4752
            goto cleanup;
4753 4754 4755
        bypassSecurityDriver = true;
    } else {
        /* Phooey - we have to fall back on exec migration, where qemu
E
Eric Blake 已提交
4756 4757
         * has to popen() the file by name, and block devices have to be
         * given cgroup ACL permission.  We might also stumble on
4758 4759
         * a race present in some qemu versions where it does a wait()
         * that botches pclose.  */
4760 4761
        if (virCgroupHasController(priv->cgroup,
                                   VIR_CGROUP_CONTROLLER_DEVICES)) {
4762 4763 4764 4765
            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
4766
                /* path was not a device, no further need for cgroup */
4767
            } else if (rv < 0) {
4768 4769
                goto cleanup;
            }
4770
        }
4771 4772
        if ((!bypassSecurityDriver) &&
            virSecurityManagerSetSavedStateLabel(driver->securityManager,
4773
                                                 vm->def, path) < 0)
4774
            goto cleanup;
4775
        restoreLabel = true;
4776 4777
    }

4778
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
4779 4780
        goto cleanup;

4781 4782 4783
    if (!compressor) {
        const char *args[] = { "cat", NULL };

4784
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800
            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
        };
4801 4802 4803 4804
        if (pipeFD[0] != -1) {
            cmd = virCommandNewArgs(args);
            virCommandSetInputFD(cmd, pipeFD[0]);
            virCommandSetOutputFD(cmd, &fd);
4805 4806
            virCommandSetErrorBuffer(cmd, &errbuf);
            virCommandDoAsyncIO(cmd);
4807 4808 4809
            if (virSetCloseExec(pipeFD[1]) < 0) {
                virReportSystemError(errno, "%s",
                                     _("Unable to set cloexec flag"));
4810
                qemuDomainObjExitMonitor(driver, vm);
4811 4812 4813
                goto cleanup;
            }
            if (virCommandRunAsync(cmd, NULL) < 0) {
4814
                qemuDomainObjExitMonitor(driver, vm);
4815 4816 4817 4818 4819 4820 4821
                goto cleanup;
            }
            rc = qemuMonitorMigrateToFd(priv->mon,
                                        QEMU_MONITOR_MIGRATE_BACKGROUND,
                                        pipeFD[1]);
            if (VIR_CLOSE(pipeFD[0]) < 0 ||
                VIR_CLOSE(pipeFD[1]) < 0)
4822
                VIR_WARN("failed to close intermediate pipe");
4823 4824 4825 4826 4827
        } else {
            rc = qemuMonitorMigrateToFile(priv->mon,
                                          QEMU_MONITOR_MIGRATE_BACKGROUND,
                                          args, path, offset);
        }
4828
    }
4829
    qemuDomainObjExitMonitor(driver, vm);
4830

4831 4832 4833 4834 4835 4836
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        goto cleanup;
    }

4837 4838 4839
    if (rc < 0)
        goto cleanup;

4840
    rc = qemuMigrationWaitForCompletion(driver, vm, asyncJob, NULL, false);
4841

4842 4843 4844 4845
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
4846 4847
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
4848 4849 4850 4851
                qemuMonitorMigrateCancel(priv->mon);
                qemuDomainObjExitMonitor(driver, vm);
            }
        }
4852
        goto cleanup;
4853
    }
4854

4855 4856 4857
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

4858 4859
    ret = 0;

4860
 cleanup:
4861 4862 4863
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

4864
    /* Restore max migration bandwidth */
4865 4866
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
4867 4868
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
4869
        qemuDomainObjExitMonitor(driver, vm);
4870 4871
    }

4872 4873
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
4874 4875 4876 4877 4878
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
4879 4880
    if (restoreLabel && (!bypassSecurityDriver) &&
        virSecurityManagerRestoreSavedStateLabel(driver->securityManager,
4881
                                                 vm->def, path) < 0)
4882 4883
        VIR_WARN("failed to restore save state label on %s", path);

4884 4885
    if (virCgroupHasController(priv->cgroup,
                               VIR_CGROUP_CONTROLLER_DEVICES)) {
4886 4887 4888
        int rv = virCgroupDenyDevicePath(priv->cgroup, path,
                                         VIR_CGROUP_DEVICE_RWM);
        virDomainAuditCgroupPath(vm, priv->cgroup, "deny", path, "rwm", rv == 0);
4889
    }
4890 4891 4892 4893 4894 4895

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

4896 4897
    return ret;
}
4898 4899

int
4900
qemuMigrationJobStart(virQEMUDriverPtr driver,
4901
                      virDomainObjPtr vm,
4902
                      qemuDomainAsyncJob job)
4903 4904 4905
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

4906
    if (qemuDomainObjBeginAsyncJob(driver, vm, job) < 0)
4907 4908
        return -1;

4909
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
4910
        qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
4911
    } else {
4912 4913 4914
        qemuDomainObjSetAsyncJobMask(vm, (QEMU_JOB_DEFAULT_MASK |
                                          JOB_MASK(QEMU_JOB_SUSPEND) |
                                          JOB_MASK(QEMU_JOB_MIGRATION_OP)));
4915
    }
4916 4917 4918 4919 4920 4921 4922

    priv->job.info.type = VIR_DOMAIN_JOB_UNBOUNDED;

    return 0;
}

void
4923
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
4924
                         virDomainObjPtr vm,
4925
                         qemuMigrationJobPhase phase)
4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939
{
    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
4940
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
4941
                           virDomainObjPtr vm,
4942
                           qemuMigrationJobPhase phase)
4943
{
4944
    virObjectRef(vm);
4945 4946 4947
    qemuMigrationJobSetPhase(driver, vm, phase);
}

4948
bool
4949 4950
qemuMigrationJobContinue(virDomainObjPtr vm)
{
4951
    qemuDomainObjReleaseAsyncJob(vm);
4952
    return virObjectUnref(vm);
4953 4954 4955 4956
}

bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
4957
                         qemuDomainAsyncJob job)
4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968
{
    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");

4969
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
4970 4971 4972 4973 4974
        return false;
    }
    return true;
}

4975
bool
4976
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
4977 4978 4979
{
    return qemuDomainObjEndAsyncJob(driver, vm);
}