qemu_migration.c 204.9 KB
Newer Older
1 2 3
/*
 * qemu_migration.c: QEMU migration handling
 *
4
 * Copyright (C) 2006-2015 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23
 *
 */

#include <config.h>

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

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

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

#define VIR_FROM_THIS VIR_FROM_QEMU

67 68
VIR_LOG_INIT("qemu.qemu_migration");

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

82
enum qemuMigrationCookieFlags {
83
    QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS,
84
    QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE,
85
    QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT,
86
    QEMU_MIGRATION_COOKIE_FLAG_NETWORK,
87
    QEMU_MIGRATION_COOKIE_FLAG_NBD,
88
    QEMU_MIGRATION_COOKIE_FLAG_STATS,
89
    QEMU_MIGRATION_COOKIE_FLAG_MEMORY_HOTPLUG,
90 91 92 93 94 95 96

    QEMU_MIGRATION_COOKIE_FLAG_LAST
};

VIR_ENUM_DECL(qemuMigrationCookieFlag);
VIR_ENUM_IMPL(qemuMigrationCookieFlag,
              QEMU_MIGRATION_COOKIE_FLAG_LAST,
97 98 99 100
              "graphics",
              "lockstate",
              "persistent",
              "network",
101
              "nbd",
102 103
              "statistics",
              "memory-hotplug");
104 105 106

enum qemuMigrationCookieFeatures {
    QEMU_MIGRATION_COOKIE_GRAPHICS  = (1 << QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS),
107
    QEMU_MIGRATION_COOKIE_LOCKSTATE = (1 << QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE),
108
    QEMU_MIGRATION_COOKIE_PERSISTENT = (1 << QEMU_MIGRATION_COOKIE_FLAG_PERSISTENT),
109
    QEMU_MIGRATION_COOKIE_NETWORK = (1 << QEMU_MIGRATION_COOKIE_FLAG_NETWORK),
110
    QEMU_MIGRATION_COOKIE_NBD = (1 << QEMU_MIGRATION_COOKIE_FLAG_NBD),
111
    QEMU_MIGRATION_COOKIE_STATS = (1 << QEMU_MIGRATION_COOKIE_FLAG_STATS),
112
    QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG = (1 << QEMU_MIGRATION_COOKIE_FLAG_MEMORY_HOTPLUG),
113 114 115 116 117 118 119 120 121 122 123 124
};

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

125 126 127 128 129 130
typedef struct _qemuMigrationCookieNetData qemuMigrationCookieNetData;
typedef qemuMigrationCookieNetData *qemuMigrationCookieNetDataPtr;
struct _qemuMigrationCookieNetData {
    int vporttype; /* enum virNetDevVPortProfile */

    /*
131
     * Array of pointers to saved data. Each VIF will have its own
132 133 134 135 136 137 138 139 140 141 142 143 144 145
     * 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;
};

146 147 148 149
typedef struct _qemuMigrationCookieNBD qemuMigrationCookieNBD;
typedef qemuMigrationCookieNBD *qemuMigrationCookieNBDPtr;
struct _qemuMigrationCookieNBD {
    int port; /* on which port does NBD server listen for incoming data */
150 151 152 153 154 155

    size_t ndisks;  /* Number of items in @disk array */
    struct {
        char *target;                   /* Disk target */
        unsigned long long capacity;    /* And its capacity */
    } *disks;
156 157
};

158 159 160
typedef struct _qemuMigrationCookie qemuMigrationCookie;
typedef qemuMigrationCookie *qemuMigrationCookiePtr;
struct _qemuMigrationCookie {
E
Eric Blake 已提交
161 162
    unsigned int flags;
    unsigned int flagsMandatory;
163 164

    /* Host properties */
165 166 167 168
    unsigned char localHostuuid[VIR_UUID_BUFLEN];
    unsigned char remoteHostuuid[VIR_UUID_BUFLEN];
    char *localHostname;
    char *remoteHostname;
169 170 171 172

    /* Guest properties */
    unsigned char uuid[VIR_UUID_BUFLEN];
    char *name;
173

174 175 176 177
    /* If (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) */
    char *lockState;
    char *lockDriver;

178 179
    /* If (flags & QEMU_MIGRATION_COOKIE_GRAPHICS) */
    qemuMigrationCookieGraphicsPtr graphics;
180 181 182

    /* If (flags & QEMU_MIGRATION_COOKIE_PERSISTENT) */
    virDomainDefPtr persistent;
183 184 185

    /* If (flags & QEMU_MIGRATION_COOKIE_NETWORK) */
    qemuMigrationCookieNetworkPtr network;
186 187 188

    /* If (flags & QEMU_MIGRATION_COOKIE_NBD) */
    qemuMigrationCookieNBDPtr nbd;
189 190 191

    /* If (flags & QEMU_MIGRATION_COOKIE_STATS) */
    qemuDomainJobInfoPtr jobInfo;
192 193
};

194 195 196 197 198 199 200 201 202
static void qemuMigrationCookieGraphicsFree(qemuMigrationCookieGraphicsPtr grap)
{
    if (!grap)
        return;
    VIR_FREE(grap->listen);
    VIR_FREE(grap->tlsSubject);
    VIR_FREE(grap);
}

203

204 205 206
static void
qemuMigrationCookieNetworkFree(qemuMigrationCookieNetworkPtr network)
{
207
    size_t i;
208 209 210 211 212 213 214 215 216 217 218 219 220

    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);
}


221 222 223 224 225 226 227 228 229 230 231 232
static void qemuMigrationCookieNBDFree(qemuMigrationCookieNBDPtr nbd)
{
    if (!nbd)
        return;

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


233 234 235 236 237
static void qemuMigrationCookieFree(qemuMigrationCookiePtr mig)
{
    if (!mig)
        return;

238 239
    qemuMigrationCookieGraphicsFree(mig->graphics);
    qemuMigrationCookieNetworkFree(mig->network);
240
    qemuMigrationCookieNBDFree(mig->nbd);
241

242 243
    VIR_FREE(mig->localHostname);
    VIR_FREE(mig->remoteHostname);
244
    VIR_FREE(mig->name);
245 246
    VIR_FREE(mig->lockState);
    VIR_FREE(mig->lockDriver);
247
    VIR_FREE(mig->jobInfo);
248 249 250 251
    VIR_FREE(mig);
}


252
#ifdef WITH_GNUTLS
253 254 255 256 257 258 259 260 261 262 263 264
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)
265
        goto error;
266 267

    if (virFileReadAll(certfile, 8192, &pemdata) < 0) {
268 269
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unable to read server cert %s"), certfile);
270 271 272 273 274
        goto error;
    }

    ret = gnutls_x509_crt_init(&cert);
    if (ret < 0) {
275 276 277
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot initialize cert object: %s"),
                       gnutls_strerror(ret));
278 279 280 281 282 283 284 285
        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) {
286 287 288
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot load cert data from %s: %s"),
                       certfile, gnutls_strerror(ret));
289 290 291 292 293
        goto error;
    }

    subjectlen = 1024;
    if (VIR_ALLOC_N(subject, subjectlen+1) < 0)
294
        goto error;
295 296 297 298 299 300 301 302 303

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

    VIR_FREE(certfile);
    VIR_FREE(pemdata);

    return subject;

304
 error:
305 306 307 308
    VIR_FREE(certfile);
    VIR_FREE(pemdata);
    return NULL;
}
309
#endif
310 311

static qemuMigrationCookieGraphicsPtr
312
qemuMigrationCookieGraphicsAlloc(virQEMUDriverPtr driver,
313 314 315 316
                                 virDomainGraphicsDefPtr def)
{
    qemuMigrationCookieGraphicsPtr mig = NULL;
    const char *listenAddr;
317
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
318 319

    if (VIR_ALLOC(mig) < 0)
320
        goto error;
321 322 323 324

    mig->type = def->type;
    if (mig->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
        mig->port = def->data.vnc.port;
325
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
326
        if (!listenAddr)
327
            listenAddr = cfg->vncListen;
328

329
#ifdef WITH_GNUTLS
330 331
        if (cfg->vncTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->vncTLSx509certdir)))
332
            goto error;
333
#endif
334 335
    } else {
        mig->port = def->data.spice.port;
336
        if (cfg->spiceTLS)
337 338 339
            mig->tlsPort = def->data.spice.tlsPort;
        else
            mig->tlsPort = -1;
340
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
341
        if (!listenAddr)
342
            listenAddr = cfg->spiceListen;
343

344
#ifdef WITH_GNUTLS
345 346
        if (cfg->spiceTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(cfg->spiceTLSx509certdir)))
347
            goto error;
348
#endif
349
    }
350 351
    if (VIR_STRDUP(mig->listen, listenAddr) < 0)
        goto error;
352

353
    virObjectUnref(cfg);
354 355
    return mig;

356
 error:
357
    qemuMigrationCookieGraphicsFree(mig);
358
    virObjectUnref(cfg);
359 360 361 362
    return NULL;
}


363
static qemuMigrationCookieNetworkPtr
364
qemuMigrationCookieNetworkAlloc(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
365 366 367
                                virDomainDefPtr def)
{
    qemuMigrationCookieNetworkPtr mig;
368
    size_t i;
369 370

    if (VIR_ALLOC(mig) < 0)
371
        goto error;
372 373 374 375

    mig->nnets = def->nnets;

    if (VIR_ALLOC_N(mig->net, def->nnets) <0)
376
        goto error;
377 378 379 380 381 382 383 384 385 386 387 388 389 390 391

    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:
392
               break;
393
            case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
394 395
                if (virNetDevOpenvswitchGetMigrateData(&mig->net[i].portdata,
                                                       netptr->ifname) != 0) {
J
Jiri Denemark 已提交
396 397 398
                        virReportError(VIR_ERR_INTERNAL_ERROR,
                                       _("Unable to run command to get OVS port data for "
                                         "interface %s"), netptr->ifname);
399 400 401
                        goto error;
                }
                break;
402 403 404 405 406 407 408
            default:
                break;
            }
        }
    }
    return mig;

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

414 415 416
static qemuMigrationCookiePtr
qemuMigrationCookieNew(virDomainObjPtr dom)
{
J
Jiri Denemark 已提交
417
    qemuDomainObjPrivatePtr priv = dom->privateData;
418
    qemuMigrationCookiePtr mig = NULL;
J
Jiri Denemark 已提交
419
    const char *name;
420 421

    if (VIR_ALLOC(mig) < 0)
422
        goto error;
423

J
Jiri Denemark 已提交
424 425 426 427
    if (priv->origname)
        name = priv->origname;
    else
        name = dom->def->name;
428 429
    if (VIR_STRDUP(mig->name, name) < 0)
        goto error;
430 431
    memcpy(mig->uuid, dom->def->uuid, VIR_UUID_BUFLEN);

432
    if (!(mig->localHostname = virGetHostname()))
433
        goto error;
434
    if (virGetHostUUID(mig->localHostuuid) < 0) {
435 436
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Unable to obtain host UUID"));
437 438 439 440 441
        goto error;
    }

    return mig;

442
 error:
443 444 445 446 447
    qemuMigrationCookieFree(mig);
    return NULL;
}


448 449
static int
qemuMigrationCookieAddGraphics(qemuMigrationCookiePtr mig,
450
                               virQEMUDriverPtr driver,
451 452
                               virDomainObjPtr dom)
{
453 454
    size_t i = 0;

455
    if (mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) {
456 457
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration graphics data already present"));
458 459 460
        return -1;
    }

461 462 463 464 465 466 467 468
    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;
       }
469
    }
470 471 472 473 474

    return 0;
}


475 476
static int
qemuMigrationCookieAddLockstate(qemuMigrationCookiePtr mig,
477
                                virQEMUDriverPtr driver,
478 479 480 481 482
                                virDomainObjPtr dom)
{
    qemuDomainObjPrivatePtr priv = dom->privateData;

    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
483 484
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration lockstate data already present"));
485 486 487 488
        return -1;
    }

    if (virDomainObjGetState(dom, NULL) == VIR_DOMAIN_PAUSED) {
489
        if (VIR_STRDUP(mig->lockState, priv->lockState) < 0)
490 491 492 493 494 495
            return -1;
    } else {
        if (virDomainLockProcessInquire(driver->lockManager, dom, &mig->lockState) < 0)
            return -1;
    }

496
    if (VIR_STRDUP(mig->lockDriver, virLockManagerPluginGetName(driver->lockManager)) < 0) {
497 498 499 500 501 502 503 504 505 506 507
        VIR_FREE(mig->lockState);
        return -1;
    }

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

    return 0;
}


508 509 510 511 512
static int
qemuMigrationCookieAddPersistent(qemuMigrationCookiePtr mig,
                                 virDomainObjPtr dom)
{
    if (mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) {
513 514
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration persistent data already present"));
515 516 517 518 519 520 521 522 523 524 525 526 527
        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;
}


528 529 530 531 532 533 534 535 536 537 538 539 540
static virDomainDefPtr
qemuMigrationCookieGetPersistent(qemuMigrationCookiePtr mig)
{
    virDomainDefPtr def = mig->persistent;

    mig->persistent = NULL;
    mig->flags &= ~QEMU_MIGRATION_COOKIE_PERSISTENT;
    mig->flagsMandatory &= ~QEMU_MIGRATION_COOKIE_PERSISTENT;

    return def;
}


541 542
static int
qemuMigrationCookieAddNetwork(qemuMigrationCookiePtr mig,
543
                              virQEMUDriverPtr driver,
544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561
                              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;
}

562

563 564
static int
qemuMigrationCookieAddNBD(qemuMigrationCookiePtr mig,
565
                          virQEMUDriverPtr driver,
566 567 568
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
569 570
    virHashTablePtr stats = NULL;
    size_t i;
571
    int ret = -1, rc;
572 573 574

    /* It is not a bug if there already is a NBD data */
    if (!mig->nbd &&
575
        VIR_ALLOC(mig->nbd) < 0)
576 577
        return -1;

578 579 580 581 582 583 584 585 586 587 588 589 590
    if (vm->def->ndisks &&
        VIR_ALLOC_N(mig->nbd->disks, vm->def->ndisks) < 0)
        return -1;
    mig->nbd->ndisks = 0;

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

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

591 592 593
            if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                               priv->job.asyncJob) < 0)
                goto cleanup;
594 595
            rc = qemuMonitorBlockStatsUpdateCapacity(priv->mon, stats, false);
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
596
                goto cleanup;
597
            if (rc < 0)
598 599 600 601 602 603 604 605 606 607 608 609 610 611
                goto cleanup;
        }

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

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

612 613 614
    mig->nbd->port = priv->nbdPort;
    mig->flags |= QEMU_MIGRATION_COOKIE_NBD;

615 616 617 618
    ret = 0;
 cleanup:
    virHashFree(stats);
    return ret;
619 620 621
}


622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640
static int
qemuMigrationCookieAddStatistics(qemuMigrationCookiePtr mig,
                                 virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (!priv->job.completed)
        return 0;

    if (!mig->jobInfo && VIR_ALLOC(mig->jobInfo) < 0)
        return -1;

    *mig->jobInfo = *priv->job.completed;
    mig->flags |= QEMU_MIGRATION_COOKIE_STATS;

    return 0;
}


641 642 643
static void qemuMigrationCookieGraphicsXMLFormat(virBufferPtr buf,
                                                 qemuMigrationCookieGraphicsPtr grap)
{
644
    virBufferAsprintf(buf, "<graphics type='%s' port='%d' listen='%s'",
645 646 647 648 649 650
                      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");
651 652 653 654
        virBufferAdjustIndent(buf, 2);
        virBufferEscapeString(buf, "<cert info='subject' value='%s'/>\n", grap->tlsSubject);
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</graphics>\n");
655 656 657 658 659 660
    } else {
        virBufferAddLit(buf, "/>\n");
    }
}


661 662 663 664
static void
qemuMigrationCookieNetworkXMLFormat(virBufferPtr buf,
                                    qemuMigrationCookieNetworkPtr optr)
{
665
    size_t i;
666 667 668 669 670 671
    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) {
672 673
                virBufferAddLit(buf, "<network>\n");
                virBufferAdjustIndent(buf, 2);
674 675
                empty = false;
            }
676
            virBufferAsprintf(buf, "<interface index='%zu' vporttype='%s'",
677 678 679
                              i, virNetDevVPortTypeToString(optr->net[i].vporttype));
            if (optr->net[i].portdata) {
                virBufferAddLit(buf, ">\n");
680 681
                virBufferAdjustIndent(buf, 2);
                virBufferEscapeString(buf, "<portdata>%s</portdata>\n",
682
                                      optr->net[i].portdata);
683 684
                virBufferAdjustIndent(buf, -2);
                virBufferAddLit(buf, "</interface>\n");
685 686 687 688 689
            } else {
                virBufferAddLit(buf, "/>\n");
            }
        }
    }
690 691 692 693
    if (!empty) {
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</network>\n");
    }
694 695 696
}


697 698 699 700 701 702 703 704 705
static void
qemuMigrationCookieStatisticsXMLFormat(virBufferPtr buf,
                                       qemuDomainJobInfoPtr jobInfo)
{
    qemuMonitorMigrationStatus *status = &jobInfo->status;

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

706 707
    virBufferAsprintf(buf, "<started>%llu</started>\n", jobInfo->started);
    virBufferAsprintf(buf, "<stopped>%llu</stopped>\n", jobInfo->stopped);
708 709 710
    virBufferAsprintf(buf, "<sent>%llu</sent>\n", jobInfo->sent);
    if (jobInfo->timeDeltaSet)
        virBufferAsprintf(buf, "<delta>%lld</delta>\n", jobInfo->timeDelta);
711

712 713 714 715 716 717 718 719 720 721
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_TIME_ELAPSED,
                      jobInfo->timeElapsed);
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_TIME_REMAINING,
                      jobInfo->timeRemaining);
    if (status->downtime_set)
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_DOWNTIME,
                          status->downtime);
722 723 724 725
    if (status->setup_time_set)
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_SETUP_TIME,
                          status->setup_time);
726 727 728 729 730 731 732 733 734 735

    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_TOTAL,
                      status->ram_total);
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_PROCESSED,
                      status->ram_transferred);
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_REMAINING,
                      status->ram_remaining);
736 737 738
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_BPS,
                      status->ram_bps);
739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760

    if (status->ram_duplicate_set) {
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_MEMORY_CONSTANT,
                          status->ram_duplicate);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_MEMORY_NORMAL,
                          status->ram_normal);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_MEMORY_NORMAL_BYTES,
                          status->ram_normal_bytes);
    }

    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_TOTAL,
                      status->disk_total);
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_PROCESSED,
                      status->disk_transferred);
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_REMAINING,
                      status->disk_remaining);
761 762 763
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_BPS,
                      status->disk_bps);
764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787

    if (status->xbzrle_set) {
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_COMPRESSION_CACHE,
                          status->xbzrle_cache_size);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_COMPRESSION_BYTES,
                          status->xbzrle_bytes);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_COMPRESSION_PAGES,
                          status->xbzrle_pages);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_COMPRESSION_CACHE_MISSES,
                          status->xbzrle_cache_miss);
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_COMPRESSION_OVERFLOW,
                          status->xbzrle_overflow);
    }

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


788
static int
789
qemuMigrationCookieXMLFormat(virQEMUDriverPtr driver,
790
                             virBufferPtr buf,
791
                             qemuMigrationCookiePtr mig)
792 793 794
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
    char hostuuidstr[VIR_UUID_STRING_BUFLEN];
795
    size_t i;
796 797

    virUUIDFormat(mig->uuid, uuidstr);
798
    virUUIDFormat(mig->localHostuuid, hostuuidstr);
799

800
    virBufferAddLit(buf, "<qemu-migration>\n");
801 802 803 804 805
    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);
806

807
    for (i = 0; i < QEMU_MIGRATION_COOKIE_FLAG_LAST; i++) {
808
        if (mig->flagsMandatory & (1 << i))
809
            virBufferAsprintf(buf, "<feature name='%s'/>\n",
810 811 812
                              qemuMigrationCookieFlagTypeToString(i));
    }

813 814
    if ((mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        mig->graphics)
815 816
        qemuMigrationCookieGraphicsXMLFormat(buf, mig->graphics);

817 818
    if ((mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        mig->lockState) {
819
        virBufferAsprintf(buf, "<lockstate driver='%s'>\n",
820
                          mig->lockDriver);
821 822
        virBufferAdjustIndent(buf, 2);
        virBufferAsprintf(buf, "<leases>%s</leases>\n",
823
                          mig->lockState);
824 825
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</lockstate>\n");
826 827
    }

828 829
    if ((mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        mig->persistent) {
830 831 832
        if (qemuDomainDefFormatBuf(driver,
                                   mig->persistent,
                                   VIR_DOMAIN_XML_INACTIVE |
833 834
                                   VIR_DOMAIN_XML_SECURE |
                                   VIR_DOMAIN_XML_MIGRATABLE,
835
                                   buf) < 0)
836
            return -1;
837 838
    }

839 840 841
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NETWORK) && mig->network)
        qemuMigrationCookieNetworkXMLFormat(buf, mig->network);

842
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NBD) && mig->nbd) {
843
        virBufferAddLit(buf, "<nbd");
844 845
        if (mig->nbd->port)
            virBufferAsprintf(buf, " port='%d'", mig->nbd->port);
846 847 848 849 850 851 852 853 854 855 856 857 858 859
        if (mig->nbd->ndisks) {
            virBufferAddLit(buf, ">\n");
            virBufferAdjustIndent(buf, 2);
            for (i = 0; i < mig->nbd->ndisks; i++) {
                virBufferEscapeString(buf, "<disk target='%s'",
                                      mig->nbd->disks[i].target);
                virBufferAsprintf(buf, " capacity='%llu'/>\n",
                                  mig->nbd->disks[i].capacity);
            }
            virBufferAdjustIndent(buf, -2);
            virBufferAddLit(buf, "</nbd>\n");
        } else {
            virBufferAddLit(buf, "/>\n");
        }
860 861
    }

862 863 864
    if (mig->flags & QEMU_MIGRATION_COOKIE_STATS && mig->jobInfo)
        qemuMigrationCookieStatisticsXMLFormat(buf, mig->jobInfo);

865
    virBufferAdjustIndent(buf, -2);
866
    virBufferAddLit(buf, "</qemu-migration>\n");
867
    return 0;
868 869 870
}


871
static char *qemuMigrationCookieXMLFormatStr(virQEMUDriverPtr driver,
872
                                             qemuMigrationCookiePtr mig)
873 874 875
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;

876
    if (qemuMigrationCookieXMLFormat(driver, &buf, mig) < 0) {
877 878 879
        virBufferFreeAndReset(&buf);
        return NULL;
    }
880

881
    if (virBufferCheckError(&buf) < 0)
882 883 884 885 886 887
        return NULL;

    return virBufferContentAndReset(&buf);
}


888 889 890 891 892 893 894
static qemuMigrationCookieGraphicsPtr
qemuMigrationCookieGraphicsXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieGraphicsPtr grap;
    char *tmp;

    if (VIR_ALLOC(grap) < 0)
895
        goto error;
896 897

    if (!(tmp = virXPathString("string(./graphics/@type)", ctxt))) {
898 899
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing type attribute in migration data"));
900 901 902
        goto error;
    }
    if ((grap->type = virDomainGraphicsTypeFromString(tmp)) < 0) {
903 904
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unknown graphics type %s"), tmp);
905 906 907
        VIR_FREE(tmp);
        goto error;
    }
E
Eric Blake 已提交
908
    VIR_FREE(tmp);
909
    if (virXPathInt("string(./graphics/@port)", ctxt, &grap->port) < 0) {
910 911
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing port attribute in migration data"));
912 913 914 915
        goto error;
    }
    if (grap->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        if (virXPathInt("string(./graphics/@tlsPort)", ctxt, &grap->tlsPort) < 0) {
916 917
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing tlsPort attribute in migration data"));
918 919 920 921
            goto error;
        }
    }
    if (!(grap->listen = virXPathString("string(./graphics/@listen)", ctxt))) {
922 923
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing listen attribute in migration data"));
924 925 926
        goto error;
    }
    /* Optional */
927
    grap->tlsSubject = virXPathString("string(./graphics/cert[@info='subject']/@value)", ctxt);
928 929 930

    return grap;

931
 error:
932 933 934 935 936
    qemuMigrationCookieGraphicsFree(grap);
    return NULL;
}


937 938 939 940
static qemuMigrationCookieNetworkPtr
qemuMigrationCookieNetworkXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieNetworkPtr optr;
941
    size_t i;
942 943 944 945 946 947
    int n;
    xmlNodePtr *interfaces = NULL;
    char *vporttype;
    xmlNodePtr save_ctxt = ctxt->node;

    if (VIR_ALLOC(optr) < 0)
948
        goto error;
949 950 951 952 953 954 955 956

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

    optr->nnets = n;
957
    if (VIR_ALLOC_N(optr->net, optr->nnets) < 0)
958
        goto error;
959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974

    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);

975
 cleanup:
976 977 978
    ctxt->node = save_ctxt;
    return optr;

979
 error:
980 981 982 983 984 985 986
    VIR_FREE(interfaces);
    qemuMigrationCookieNetworkFree(optr);
    optr = NULL;
    goto cleanup;
}


987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
static qemuMigrationCookieNBDPtr
qemuMigrationCookieNBDXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieNBDPtr ret = NULL;
    char *port = NULL, *capacity = NULL;
    size_t i;
    int n;
    xmlNodePtr *disks = NULL;
    xmlNodePtr save_ctxt = ctxt->node;

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

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

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

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

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

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

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


1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
static qemuDomainJobInfoPtr
qemuMigrationCookieStatisticsXMLParse(xmlXPathContextPtr ctxt)
{
    qemuDomainJobInfoPtr jobInfo = NULL;
    qemuMonitorMigrationStatus *status;
    xmlNodePtr save_ctxt = ctxt->node;

    if (!(ctxt->node = virXPathNode("./statistics", ctxt)))
        goto cleanup;

    if (VIR_ALLOC(jobInfo) < 0)
        goto cleanup;

    status = &jobInfo->status;
    jobInfo->type = VIR_DOMAIN_JOB_COMPLETED;

1067 1068
    virXPathULongLong("string(./started[1])", ctxt, &jobInfo->started);
    virXPathULongLong("string(./stopped[1])", ctxt, &jobInfo->stopped);
1069 1070 1071
    virXPathULongLong("string(./sent[1])", ctxt, &jobInfo->sent);
    if (virXPathLongLong("string(./delta[1])", ctxt, &jobInfo->timeDelta) == 0)
        jobInfo->timeDeltaSet = true;
1072

1073 1074 1075 1076
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_TIME_ELAPSED "[1])",
                      ctxt, &jobInfo->timeElapsed);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_TIME_REMAINING "[1])",
                      ctxt, &jobInfo->timeRemaining);
1077

1078 1079 1080
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_DOWNTIME "[1])",
                          ctxt, &status->downtime) == 0)
        status->downtime_set = true;
1081 1082 1083
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_SETUP_TIME "[1])",
                          ctxt, &status->setup_time) == 0)
        status->setup_time_set = true;
1084 1085 1086 1087 1088 1089 1090

    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_TOTAL "[1])",
                      ctxt, &status->ram_total);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_PROCESSED "[1])",
                      ctxt, &status->ram_transferred);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_REMAINING "[1])",
                      ctxt, &status->ram_remaining);
1091 1092
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_BPS "[1])",
                      ctxt, &status->ram_bps);
1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107

    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_CONSTANT "[1])",
                          ctxt, &status->ram_duplicate) == 0)
        status->ram_duplicate_set = true;
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_NORMAL "[1])",
                      ctxt, &status->ram_normal);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_NORMAL_BYTES "[1])",
                      ctxt, &status->ram_normal_bytes);

    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_TOTAL "[1])",
                      ctxt, &status->disk_total);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_PROCESSED "[1])",
                      ctxt, &status->disk_transferred);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_REMAINING "[1])",
                      ctxt, &status->disk_remaining);
1108 1109
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_BPS "[1])",
                      ctxt, &status->disk_bps);
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128

    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_COMPRESSION_CACHE "[1])",
                          ctxt, &status->xbzrle_cache_size) == 0)
        status->xbzrle_set = true;
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_COMPRESSION_BYTES "[1])",
                      ctxt, &status->xbzrle_bytes);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_COMPRESSION_PAGES "[1])",
                      ctxt, &status->xbzrle_pages);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_COMPRESSION_CACHE_MISSES "[1])",
                      ctxt, &status->xbzrle_cache_miss);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_COMPRESSION_OVERFLOW "[1])",
                      ctxt, &status->xbzrle_overflow);

 cleanup:
    ctxt->node = save_ctxt;
    return jobInfo;
}


1129 1130
static int
qemuMigrationCookieXMLParse(qemuMigrationCookiePtr mig,
1131
                            virQEMUDriverPtr driver,
1132
                            xmlDocPtr doc,
1133
                            xmlXPathContextPtr ctxt,
E
Eric Blake 已提交
1134
                            unsigned int flags)
1135 1136
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
1137
    char *tmp = NULL;
1138
    xmlNodePtr *nodes = NULL;
1139 1140
    size_t i;
    int n;
1141 1142 1143 1144
    virCapsPtr caps = NULL;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto error;
1145 1146 1147 1148 1149 1150 1151 1152

    /* 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))) {
1153 1154
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing name element in migration data"));
1155 1156 1157
        goto error;
    }
    if (STRNEQ(tmp, mig->name)) {
1158 1159 1160
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected name %s vs %s"),
                       tmp, mig->name);
1161 1162 1163 1164 1165 1166 1167
        goto error;
    }
    VIR_FREE(tmp);

    /* Extract domain uuid */
    tmp = virXPathString("string(./uuid[1])", ctxt);
    if (!tmp) {
1168 1169
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing uuid element in migration data"));
1170 1171 1172 1173
        goto error;
    }
    virUUIDFormat(mig->uuid, uuidstr);
    if (STRNEQ(tmp, uuidstr)) {
1174 1175 1176
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected UUID %s vs %s"),
                       tmp, uuidstr);
1177
        goto error;
1178 1179 1180 1181
    }
    VIR_FREE(tmp);

    /* Check & forbid "localhost" migration */
1182
    if (!(mig->remoteHostname = virXPathString("string(./hostname[1])", ctxt))) {
1183 1184
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostname element in migration data"));
1185 1186
        goto error;
    }
1187
    if (STREQ(mig->remoteHostname, mig->localHostname)) {
1188 1189 1190
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Attempt to migrate guest to the same host %s"),
                       mig->remoteHostname);
1191 1192 1193 1194
        goto error;
    }

    if (!(tmp = virXPathString("string(./hostuuid[1])", ctxt))) {
1195 1196 1197
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostuuid element in migration data"));
        goto error;
1198
    }
1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210
    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);
1211

1212 1213
    /* Check to ensure all mandatory features from XML are also
     * present in 'flags' */
1214
    if ((n = virXPathNodeSet("./feature", ctxt, &nodes)) < 0)
1215 1216
        goto error;

1217
    for (i = 0; i < n; i++) {
1218 1219 1220
        int val;
        char *str = virXMLPropString(nodes[i], "name");
        if (!str) {
1221 1222
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing feature name"));
1223 1224 1225 1226
            goto error;
        }

        if ((val = qemuMigrationCookieFlagTypeFromString(str)) < 0) {
1227 1228 1229
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unknown migration cookie feature %s"),
                           str);
1230 1231 1232 1233 1234
            VIR_FREE(str);
            goto error;
        }

        if ((flags & (1 << val)) == 0) {
1235 1236 1237
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unsupported migration cookie feature %s"),
                           str);
1238 1239 1240 1241 1242 1243
            VIR_FREE(str);
        }
        VIR_FREE(str);
    }
    VIR_FREE(nodes);

1244 1245 1246 1247 1248
    if ((flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        virXPathBoolean("count(./graphics) > 0", ctxt) &&
        (!(mig->graphics = qemuMigrationCookieGraphicsXMLParse(ctxt))))
        goto error;

1249 1250 1251 1252
    if ((flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        virXPathBoolean("count(./lockstate) > 0", ctxt)) {
        mig->lockDriver = virXPathString("string(./lockstate[1]/@driver)", ctxt);
        if (!mig->lockDriver) {
1253 1254
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Missing lock driver name in migration cookie"));
1255 1256 1257 1258 1259 1260 1261
            goto error;
        }
        mig->lockState = virXPathString("string(./lockstate[1]/leases[1])", ctxt);
        if (mig->lockState && STREQ(mig->lockState, ""))
            VIR_FREE(mig->lockState);
    }

1262 1263 1264
    if ((flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        virXPathBoolean("count(./domain) > 0", ctxt)) {
        if ((n = virXPathNodeSet("./domain", ctxt, &nodes)) > 1) {
1265 1266 1267 1268
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Too many domain elements in "
                             "migration cookie: %d"),
                           n);
1269 1270
            goto error;
        }
1271 1272
        mig->persistent = virDomainDefParseNode(doc, nodes[0],
                                                caps, driver->xmlopt,
1273 1274
                                                VIR_DOMAIN_DEF_PARSE_INACTIVE |
                                                VIR_DOMAIN_DEF_PARSE_ABI_UPDATE);
1275 1276 1277 1278 1279 1280 1281 1282
        if (!mig->persistent) {
            /* virDomainDefParseNode already reported
             * an error for us */
            goto error;
        }
        VIR_FREE(nodes);
    }

1283 1284 1285 1286 1287
    if ((flags & QEMU_MIGRATION_COOKIE_NETWORK) &&
        virXPathBoolean("count(./network) > 0", ctxt) &&
        (!(mig->network = qemuMigrationCookieNetworkXMLParse(ctxt))))
        goto error;

1288
    if (flags & QEMU_MIGRATION_COOKIE_NBD &&
1289 1290 1291
        virXPathBoolean("boolean(./nbd)", ctxt) &&
        (!(mig->nbd = qemuMigrationCookieNBDXMLParse(ctxt))))
        goto error;
1292

1293 1294 1295 1296 1297
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        virXPathBoolean("boolean(./statistics)", ctxt) &&
        (!(mig->jobInfo = qemuMigrationCookieStatisticsXMLParse(ctxt))))
        goto error;

1298
    virObjectUnref(caps);
1299 1300
    return 0;

1301
 error:
1302
    VIR_FREE(tmp);
1303
    VIR_FREE(nodes);
1304
    virObjectUnref(caps);
1305 1306 1307 1308 1309 1310
    return -1;
}


static int
qemuMigrationCookieXMLParseStr(qemuMigrationCookiePtr mig,
1311
                               virQEMUDriverPtr driver,
1312
                               const char *xml,
E
Eric Blake 已提交
1313
                               unsigned int flags)
1314 1315 1316
{
    xmlDocPtr doc = NULL;
    xmlXPathContextPtr ctxt = NULL;
1317
    int ret = -1;
1318 1319 1320

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

1321
    if (!(doc = virXMLParseStringCtxt(xml, _("(qemu_migration_cookie)"), &ctxt)))
1322 1323
        goto cleanup;

1324
    ret = qemuMigrationCookieXMLParse(mig, driver, doc, ctxt, flags);
1325

1326
 cleanup:
1327 1328 1329 1330 1331 1332 1333 1334 1335
    xmlXPathFreeContext(ctxt);
    xmlFreeDoc(doc);

    return ret;
}


static int
qemuMigrationBakeCookie(qemuMigrationCookiePtr mig,
1336
                        virQEMUDriverPtr driver,
1337
                        virDomainObjPtr dom,
1338 1339
                        char **cookieout,
                        int *cookieoutlen,
E
Eric Blake 已提交
1340
                        unsigned int flags)
1341
{
1342 1343
    if (!cookieout || !cookieoutlen)
        return 0;
1344 1345 1346

    *cookieoutlen = 0;

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

1351 1352 1353 1354
    if (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE &&
        qemuMigrationCookieAddLockstate(mig, driver, dom) < 0)
        return -1;

1355 1356 1357 1358
    if (flags & QEMU_MIGRATION_COOKIE_PERSISTENT &&
        qemuMigrationCookieAddPersistent(mig, dom) < 0)
        return -1;

1359 1360 1361 1362 1363
    if (flags & QEMU_MIGRATION_COOKIE_NETWORK &&
        qemuMigrationCookieAddNetwork(mig, driver, dom) < 0) {
        return -1;
    }

1364 1365 1366 1367
    if ((flags & QEMU_MIGRATION_COOKIE_NBD) &&
        qemuMigrationCookieAddNBD(mig, driver, dom) < 0)
        return -1;

1368 1369 1370 1371
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        qemuMigrationCookieAddStatistics(mig, dom) < 0)
        return -1;

1372 1373 1374
    if (flags & QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG)
        mig->flagsMandatory |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

1375
    if (!(*cookieout = qemuMigrationCookieXMLFormatStr(driver, mig)))
1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
        return -1;

    *cookieoutlen = strlen(*cookieout) + 1;

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

    return 0;
}


static qemuMigrationCookiePtr
1387
qemuMigrationEatCookie(virQEMUDriverPtr driver,
1388
                       virDomainObjPtr dom,
1389 1390
                       const char *cookiein,
                       int cookieinlen,
E
Eric Blake 已提交
1391
                       unsigned int flags)
1392 1393 1394 1395 1396 1397
{
    qemuMigrationCookiePtr mig = NULL;

    /* Parse & validate incoming cookie (if any) */
    if (cookiein && cookieinlen &&
        cookiein[cookieinlen-1] != '\0') {
1398 1399
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration cookie was not NULL terminated"));
1400 1401 1402 1403 1404 1405 1406 1407 1408 1409
        goto error;
    }

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

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

    if (cookiein && cookieinlen &&
        qemuMigrationCookieXMLParseStr(mig,
1410
                                       driver,
1411 1412 1413 1414
                                       cookiein,
                                       flags) < 0)
        goto error;

1415 1416 1417 1418 1419 1420 1421 1422
    if (flags & QEMU_MIGRATION_COOKIE_PERSISTENT &&
        mig->persistent &&
        STRNEQ(dom->def->name, mig->persistent->name)) {
        VIR_FREE(mig->persistent->name);
        if (VIR_STRDUP(mig->persistent->name, dom->def->name) < 0)
            goto error;
    }

1423 1424 1425
    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
        if (!mig->lockDriver) {
            if (virLockManagerPluginUsesState(driver->lockManager)) {
1426 1427 1428
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Missing %s lock state for migration cookie"),
                               virLockManagerPluginGetName(driver->lockManager));
1429 1430 1431 1432
                goto error;
            }
        } else if (STRNEQ(mig->lockDriver,
                          virLockManagerPluginGetName(driver->lockManager))) {
1433 1434 1435 1436
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Source host lock driver %s different from target %s"),
                           mig->lockDriver,
                           virLockManagerPluginGetName(driver->lockManager));
1437 1438 1439 1440
            goto error;
        }
    }

1441 1442
    return mig;

1443
 error:
1444 1445 1446
    qemuMigrationCookieFree(mig);
    return NULL;
}
1447

1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494
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;
}

1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548

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

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

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

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

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

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

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

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

1549 1550 1551 1552 1553
    case VIR_STORAGE_TYPE_NETWORK:
        VIR_DEBUG("Skipping creation of network disk '%s'",
                  disk->dst);
        return 0;

1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601
    case VIR_STORAGE_TYPE_BLOCK:
    case VIR_STORAGE_TYPE_DIR:
    case VIR_STORAGE_TYPE_NONE:
    case VIR_STORAGE_TYPE_LAST:
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot precreate storage for disk type '%s'"),
                       virStorageTypeToString(disk->src->type));
        goto cleanup;
        break;
    }

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

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

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

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

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

1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
static bool
qemuMigrateDisk(virDomainDiskDef const *disk,
                size_t nmigrate_disks, const char **migrate_disks)
{
    size_t i;

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

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

1623 1624 1625 1626 1627

static int
qemuMigrationPrecreateStorage(virConnectPtr conn,
                              virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                              virDomainObjPtr vm,
1628 1629
                              qemuMigrationCookieNBDPtr nbd,
                              size_t nmigrate_disks,
1630 1631
                              const char **migrate_disks,
                              bool incremental)
1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642
{
    int ret = -1;
    size_t i = 0;

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

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

1643
        VIR_DEBUG("Looking up disk target '%s' (capacity=%llu)",
1644 1645
                  nbd->disks[i].target, nbd->disks[i].capacity);

1646 1647
        if (!(disk = virDomainDiskByName(vm->def, nbd->disks[i].target,
                                         false))) {
1648 1649 1650 1651 1652 1653 1654 1655
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unable to find disk by target: %s"),
                           nbd->disks[i].target);
            goto cleanup;
        }

        diskSrcPath = virDomainDiskGetSource(disk);

1656 1657
        /* Skip disks we don't want to migrate and already existing disks. */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks) ||
1658 1659 1660 1661
            (diskSrcPath && virFileExists(diskSrcPath))) {
            continue;
        }

1662 1663 1664 1665 1666 1667 1668
        if (incremental) {
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                           _("pre-creation of storage targets for incremental "
                             "storage migration is not supported"));
            goto cleanup;
        }

1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680
        VIR_DEBUG("Proceeding with disk source %s", NULLSTR(diskSrcPath));

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

    ret = 0;
 cleanup:
    return ret;
}


1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694
/**
 * 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
1695
                            virDomainObjPtr vm,
1696 1697 1698
                            const char *listenAddr,
                            size_t nmigrate_disks,
                            const char **migrate_disks)
1699 1700 1701 1702 1703 1704 1705 1706 1707 1708
{
    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];

1709 1710
        /* check whether disk should be migrated */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks))
1711 1712 1713 1714
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1715
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1716 1717 1718 1719 1720 1721 1722
            goto cleanup;

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

        if (!port &&
1723
            ((virPortAllocatorAcquire(driver->migrationPorts, &port) < 0) ||
1724
             (qemuMonitorNBDServerStart(priv->mon, listenAddr, port) < 0))) {
1725
            goto exit_monitor;
1726 1727
        }

1728 1729 1730
        if (qemuMonitorNBDServerAdd(priv->mon, diskAlias, true) < 0)
            goto exit_monitor;
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
1731 1732 1733 1734 1735 1736
            goto cleanup;
    }

    priv->nbdPort = port;
    ret = 0;

1737
 cleanup:
1738
    VIR_FREE(diskAlias);
1739
    if (ret < 0)
1740
        virPortAllocatorRelease(driver->migrationPorts, port);
1741
    return ret;
1742 1743 1744 1745

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;
1746 1747
}

1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774

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

    if (!mig->nbd)
        return 0;

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

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

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


/**
1775
 * qemuMigrationDriveMirrorReady:
1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787
 * @driver: qemu driver
 * @vm: domain
 *
 * Check the status of all drive-mirrors started by
 * qemuMigrationDriveMirror. Any pending block job events
 * for the mirrored disks will be processed.
 *
 * Returns 1 if all mirrors are "ready",
 *         0 if some mirrors are still performing initial sync,
 *        -1 on error.
 */
static int
1788
qemuMigrationDriveMirrorReady(virQEMUDriverPtr driver,
1789 1790 1791
                              virDomainObjPtr vm)
{
    size_t i;
1792 1793
    size_t notReady = 0;
    int status;
1794 1795 1796

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
1797
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1798

1799
        if (!diskPriv->migrating)
1800 1801
            continue;

1802 1803
        status = qemuBlockJobUpdate(driver, vm, disk);
        if (status == VIR_DOMAIN_BLOCK_JOB_FAILED) {
1804 1805 1806 1807 1808
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("migration of disk %s failed"),
                           disk->dst);
            return -1;
        }
1809 1810 1811

        if (disk->mirrorState != VIR_DOMAIN_DISK_MIRROR_STATE_READY)
            notReady++;
1812 1813
    }

1814 1815 1816 1817 1818 1819 1820
    if (notReady) {
        VIR_DEBUG("Waiting for %zu disk mirrors to get ready", notReady);
        return 0;
    } else {
        VIR_DEBUG("All disk mirrors are ready");
        return 1;
    }
1821 1822 1823
}


1824 1825 1826 1827
/*
 * If @check is true, the function will report an error and return a different
 * code in case a block job fails. This way we can properly abort migration in
 * case some block jobs failed once all memory has already been transferred.
1828
 *
1829 1830 1831 1832
 * Returns 1 if all mirrors are gone,
 *         0 if some mirrors are still active,
 *         -1 some mirrors failed but some are still active,
 *         -2 all mirrors are gone but some of them failed.
1833 1834
 */
static int
1835
qemuMigrationDriveMirrorCancelled(virQEMUDriverPtr driver,
1836
                                  virDomainObjPtr vm,
1837
                                  bool check)
1838
{
1839 1840 1841 1842
    size_t i;
    size_t active = 0;
    int status;
    bool failed = false;
1843

1844 1845 1846
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1847

1848 1849
        if (!diskPriv->migrating)
            continue;
1850

1851 1852 1853 1854
        status = qemuBlockJobUpdate(driver, vm, disk);
        switch (status) {
        case VIR_DOMAIN_BLOCK_JOB_FAILED:
            if (check) {
1855
                virReportError(VIR_ERR_OPERATION_FAILED,
1856
                               _("migration of disk %s failed"),
1857
                               disk->dst);
1858
                failed = true;
1859
            }
1860 1861 1862 1863 1864 1865
            /* fallthrough */
        case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
            qemuBlockJobSyncEnd(driver, vm, disk);
            diskPriv->migrating = false;
            break;
1866

1867 1868
        default:
            active++;
1869
        }
1870
    }
1871

1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887
    if (failed) {
        if (active) {
            VIR_DEBUG("Some disk mirrors failed; still waiting for %zu "
                      "disk mirrors to finish", active);
            return -1;
        } else {
            VIR_DEBUG("All disk mirrors are gone; some of them failed");
            return -2;
        }
    } else {
        if (active) {
            VIR_DEBUG("Waiting for %zu disk mirrors to finish", active);
            return 0;
        } else {
            VIR_DEBUG("All disk mirrors are gone");
            return 1;
1888
        }
1889
    }
1890
}
1891 1892


1893 1894 1895 1896 1897 1898 1899 1900 1901
/*
 * Returns 0 on success,
 *         1 when job is already completed or it failed and failNoJob is false,
 *         -1 on error or when job failed and failNoJob is true.
 */
static int
qemuMigrationCancelOneDriveMirror(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  virDomainDiskDefPtr disk,
1902 1903
                                  bool failNoJob,
                                  qemuDomainAsyncJob asyncJob)
1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char *diskAlias = NULL;
    int ret = -1;
    int status;
    int rv;

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

    case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
        return 1;
    }

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

1931
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
1932 1933 1934 1935 1936 1937 1938 1939
        goto cleanup;

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

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

    ret = 0;
1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950

 cleanup:
    VIR_FREE(diskAlias);
    return ret;
}


/**
 * qemuMigrationCancelDriveMirror:
 * @driver: qemu driver
 * @vm: domain
1951
 * @check: if true report an error when some of the mirrors fails
1952 1953 1954 1955 1956 1957 1958 1959 1960
 *
 * Cancel all drive-mirrors started by qemuMigrationDriveMirror.
 * Any pending block job events for the affected disks will be
 * processed.
 *
 * Returns 0 on success, -1 otherwise.
 */
static int
qemuMigrationCancelDriveMirror(virQEMUDriverPtr driver,
1961
                               virDomainObjPtr vm,
1962
                               bool check,
1963 1964
                               qemuDomainAsyncJob asyncJob,
                               virConnectPtr dconn)
1965
{
1966
    virErrorPtr err = NULL;
1967
    int ret = -1;
1968
    size_t i;
1969 1970 1971 1972
    int rv;
    bool failed = false;

    VIR_DEBUG("Cancelling drive mirrors for domain %s", vm->def->name);
1973 1974 1975

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
1976
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1977

1978
        if (!diskPriv->migrating)
1979 1980
            continue;

1981 1982
        rv = qemuMigrationCancelOneDriveMirror(driver, vm, disk,
                                               check, asyncJob);
1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
        if (rv != 0) {
            if (rv < 0) {
                if (!err)
                    err = virSaveLastError();
                failed = true;
            }
            qemuBlockJobSyncEnd(driver, vm, disk);
            diskPriv->migrating = false;
        }
    }

    while ((rv = qemuMigrationDriveMirrorCancelled(driver, vm, check)) != 1) {
1995 1996 1997 1998 1999 2000 2001
        if (check && !failed &&
            dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            failed = true;
        }

2002 2003 2004 2005
        if (rv < 0) {
            failed = true;
            if (rv == -2)
                break;
2006
        }
2007

2008 2009 2010 2011 2012
        if (failed && !err)
            err = virSaveLastError();

        if (virDomainObjWait(vm) < 0)
            goto cleanup;
2013 2014
    }

2015 2016 2017
    ret = failed ? -1 : 0;

 cleanup:
2018 2019 2020 2021 2022
    if (err) {
        virSetError(err);
        virFreeError(err);
    }
    return ret;
2023 2024 2025
}


2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
/**
 * 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
2037 2038 2039 2040 2041
 * simultaneously to both source and destination. On success,
 * update @migrate_flags so we don't tell 'migrate' command
 * to do the very same operation. On failure, the caller is
 * expected to call qemuMigrationCancelDriveMirror to stop all
 * running mirrors.
2042 2043 2044 2045 2046 2047 2048 2049 2050 2051
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
qemuMigrationDriveMirror(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMigrationCookiePtr mig,
                         const char *host,
                         unsigned long speed,
2052 2053
                         unsigned int *migrate_flags,
                         size_t nmigrate_disks,
2054 2055
                         const char **migrate_disks,
                         virConnectPtr dconn)
2056 2057 2058 2059
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int port;
2060
    size_t i;
2061 2062
    char *diskAlias = NULL;
    char *nbd_dest = NULL;
2063
    char *hoststr = NULL;
2064
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
2065
    int rv;
2066
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2067 2068

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
2069 2070 2071 2072 2073

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

2074 2075
    /* escape literal IPv6 address */
    if (strchr(host, ':')) {
2076
        if (virAsprintf(&hoststr, "[%s]", host) < 0)
2077
            goto cleanup;
2078
    } else if (VIR_STRDUP(hoststr, host) < 0) {
2079
        goto cleanup;
2080 2081
    }

2082 2083 2084 2085 2086
    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];
2087
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
2088
        int mon_ret;
2089

2090 2091
        /* check whether disk should be migrated */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks))
2092 2093 2094 2095 2096
            continue;

        if ((virAsprintf(&diskAlias, "%s%s",
                         QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0) ||
            (virAsprintf(&nbd_dest, "nbd:%s:%d:exportname=%s",
2097
                         hoststr, port, diskAlias) < 0))
2098 2099
            goto cleanup;

2100
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
2101
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
2102 2103
            goto cleanup;

2104
        qemuBlockJobSyncBegin(disk);
2105
        /* Force "raw" format for NBD export */
2106
        mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest,
2107
                                         "raw", speed, 0, 0, mirror_flags);
2108 2109
        VIR_FREE(diskAlias);
        VIR_FREE(nbd_dest);
2110

2111
        if (qemuDomainObjExitMonitor(driver, vm) < 0 || mon_ret < 0) {
2112
            qemuBlockJobSyncEnd(driver, vm, disk);
2113 2114
            goto cleanup;
        }
2115
        diskPriv->migrating = true;
2116 2117 2118 2119 2120

        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto cleanup;
        }
2121
    }
2122

2123 2124 2125
    while ((rv = qemuMigrationDriveMirrorReady(driver, vm)) != 1) {
        if (rv < 0)
            goto cleanup;
2126

2127
        if (priv->job.abortJob) {
2128 2129 2130 2131 2132
            priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
            goto cleanup;
2133
        }
2134

2135 2136 2137 2138 2139 2140
        if (dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            goto cleanup;
        }

2141
        if (virDomainObjWait(vm) < 0)
2142
            goto cleanup;
2143 2144
    }

2145
    /* Okay, all disks are ready. Modify migrate_flags */
2146 2147 2148 2149
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

2150
 cleanup:
2151
    virObjectUnref(cfg);
2152 2153
    VIR_FREE(diskAlias);
    VIR_FREE(nbd_dest);
2154
    VIR_FREE(hoststr);
2155 2156
    return ret;
}
2157

2158

2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186
/**
 * qemuMigrationIsAllowedHostdev:
 * @def: domain definition
 *
 * Checks that @def does not contain any host devices unsupported accross
 * migrations. Returns true if the vm is allowed to migrate.
 */
static bool
qemuMigrationIsAllowedHostdev(const virDomainDef *def)
{
    size_t i;

    /* Migration with USB host devices is allowed, all other devices are
     * forbidden. */
    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) {
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("domain has assigned non-USB host devices"));
            return false;
        }
    }

    return true;
}


2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200
/**
 * qemuMigrationIsAllowed:
 * @driver: qemu driver struct
 * @vm: domain object
 * @remote: migration is remote
 * @flags: migration flags (see struct virDomainMigrateFlags)
 *
 * Validates that the configuration of @vm can be migrated in various
 * situations. If @remote is true, the migration happens to remote host. @flags
 * is used to check various special migration types according to the request.
 *
 * Returns true if migration is supported. Reports libvirt error and returns
 * false otherwise.
 */
2201
bool
2202 2203 2204
qemuMigrationIsAllowed(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
                       bool remote,
2205
                       unsigned int flags)
2206
{
2207
    int nsnapshots;
2208
    int pauseReason;
2209
    size_t i;
2210

2211 2212 2213 2214 2215
    if (qemuProcessAutoDestroyActive(driver, vm)) {
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is marked for auto destroy"));
        return false;
    }
2216

2217 2218 2219 2220 2221
    /* perform these checks only when migrating to remote hosts */
    if (remote) {
        nsnapshots = virDomainSnapshotObjListNum(vm->snapshots, NULL, 0);
        if (nsnapshots < 0)
            return false;
2222

2223 2224 2225 2226 2227
        if (nsnapshots > 0) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("cannot migrate domain with %d snapshots"),
                           nsnapshots);
            return false;
2228
        }
2229

2230
        /* cancel migration if disk I/O error is emitted while migrating */
2231
        if (flags & VIR_MIGRATE_ABORT_ON_ERROR &&
2232 2233
            virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
            pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
E
Eric Blake 已提交
2234
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
2235
                           _("cannot migrate domain with I/O error"));
E
Eric Blake 已提交
2236 2237
            return false;
        }
2238 2239

    }
2240

2241 2242 2243 2244 2245 2246 2247
    if (qemuDomainHasBlockjob(vm, false)) {
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("domain has an active block job"));
        return false;
    }

    if (!qemuMigrationIsAllowedHostdev(vm->def))
2248
        return false;
2249

2250 2251 2252
    if (vm->def->cpu && vm->def->cpu->mode != VIR_CPU_MODE_HOST_PASSTHROUGH) {
        for (i = 0; i < vm->def->cpu->nfeatures; i++) {
            virCPUFeatureDefPtr feature = &vm->def->cpu->features[i];
J
Ján Tomko 已提交
2253

2254 2255
            if (feature->policy != VIR_CPU_FEATURE_REQUIRE)
                continue;
J
Ján Tomko 已提交
2256

2257 2258 2259 2260 2261 2262 2263
            /* QEMU blocks migration and save with invariant TSC enabled */
            if (STREQ(feature->name, "invtsc")) {
                virReportError(VIR_ERR_OPERATION_INVALID,
                               _("domain has CPU feature: %s"),
                               feature->name);
                return false;
            }
J
Ján Tomko 已提交
2264 2265 2266
        }
    }

2267
    /* Verify that memory device config can be transferred reliably */
2268 2269
    for (i = 0; i < vm->def->nmems; i++) {
        virDomainMemoryDefPtr mem = vm->def->mems[i];
2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280

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

            return false;
        }
    }

2281 2282 2283
    return true;
}

2284
static bool
2285 2286 2287
qemuMigrationIsSafe(virDomainDefPtr def,
                    size_t nmigrate_disks,
                    const char **migrate_disks)
2288
{
2289
    size_t i;
2290

2291
    for (i = 0; i < def->ndisks; i++) {
2292
        virDomainDiskDefPtr disk = def->disks[i];
2293
        const char *src = virDomainDiskGetSource(disk);
2294

2295 2296
        /* Our code elsewhere guarantees shared disks are either readonly (in
         * which case cache mode doesn't matter) or used with cache=none */
2297
        if (qemuMigrateDisk(disk, nmigrate_disks, migrate_disks) &&
2298
            disk->cachemode != VIR_DOMAIN_DISK_CACHE_DISABLE) {
2299
            int rc;
2300

E
Eric Blake 已提交
2301
            if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
2302
                if ((rc = virFileIsSharedFS(src)) < 0)
2303 2304
                    return false;
                else if (rc == 0)
2305
                    continue;
2306
                if ((rc = virStorageFileIsClusterFS(src)) < 0)
2307
                    return false;
2308 2309
                else if (rc == 1)
                    continue;
2310 2311
            } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                       disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
2312
                continue;
2313
            }
2314

2315 2316 2317
            virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                           _("Migration may lead to data corruption if disks"
                             " use cache != none"));
2318 2319 2320 2321 2322 2323 2324
            return false;
        }
    }

    return true;
}

2325 2326 2327 2328
/** qemuMigrationSetOffline
 * Pause domain for non-live migration.
 */
int
2329
qemuMigrationSetOffline(virQEMUDriverPtr driver,
2330 2331 2332
                        virDomainObjPtr vm)
{
    int ret;
2333
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
2334 2335
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
2336
    if (ret == 0) {
2337
        virObjectEventPtr event;
2338

2339
        event = virDomainEventLifecycleNewFromObj(vm,
2340 2341
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
2342
        qemuDomainEventQueue(driver, event);
2343 2344 2345 2346 2347 2348
    }

    return ret;
}


2349 2350 2351
static int
qemuMigrationSetCompression(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
2352
                            bool state,
2353
                            qemuDomainAsyncJob job)
2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366
{
    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;
2367 2368 2369
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385
    } 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,
2386 2387
                QEMU_MONITOR_MIGRATION_CAPS_XBZRLE,
                state);
2388

2389
 cleanup:
2390 2391
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2392 2393 2394
    return ret;
}

2395 2396 2397
static int
qemuMigrationSetAutoConverge(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
2398
                             bool state,
2399
                             qemuDomainAsyncJob job)
2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412
{
    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;
2413 2414 2415
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2416 2417 2418 2419 2420 2421 2422 2423 2424 2425
    } 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,
2426 2427
                QEMU_MONITOR_MIGRATION_CAPS_AUTO_CONVERGE,
                state);
2428

2429
 cleanup:
2430 2431
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2432 2433 2434 2435
    return ret;
}


2436 2437 2438
static int
qemuMigrationSetPinAll(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
2439
                       bool state,
2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453
                       qemuDomainAsyncJob job)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret;

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

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

    if (ret < 0) {
        goto cleanup;
2454 2455 2456
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472
    } else if (ret == 0) {
        if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                           _("rdma pinning migration is not supported by "
                             "target QEMU binary"));
        } else {
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                           _("rdma pinning migration is not supported by "
                             "source QEMU binary"));
        }
        ret = -1;
        goto cleanup;
    }

    ret = qemuMonitorSetMigrationCapability(
                priv->mon,
2473 2474
                QEMU_MONITOR_MIGRATION_CAPS_RDMA_PIN_ALL,
                state);
2475 2476

 cleanup:
2477 2478
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2479 2480 2481
    return ret;
}

2482
static int
2483
qemuMigrationWaitForSpice(virDomainObjPtr vm)
2484 2485 2486
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool wait_for_spice = false;
2487
    size_t i = 0;
2488

2489 2490 2491 2492 2493 2494 2495
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION))
        return 0;

    for (i = 0; i < vm->def->ngraphics; i++) {
        if (vm->def->graphics[i]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            wait_for_spice = true;
            break;
2496 2497
        }
    }
2498 2499 2500 2501

    if (!wait_for_spice)
        return 0;

2502 2503
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
2504 2505 2506 2507
            return -1;
    }
    return 0;
}
2508

2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560

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

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

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

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

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


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

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

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

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

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


2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578
static const char *
qemuMigrationJobName(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
        return _("migration job");
    case QEMU_ASYNC_JOB_SAVE:
        return _("domain save job");
    case QEMU_ASYNC_JOB_DUMP:
        return _("domain core dump job");
    default:
        return _("job");
    }
}


2579
static int
2580
qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver,
2581
                             virDomainObjPtr vm,
2582
                             qemuDomainAsyncJob asyncJob)
2583
{
2584
    qemuDomainObjPrivatePtr priv = vm->privateData;
2585 2586
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    qemuDomainJobInfo newInfo = *jobInfo;
2587

2588 2589
    if (qemuMigrationFetchJobStatus(driver, vm, asyncJob, &newInfo) < 0)
        return -1;
2590

2591 2592 2593
    *jobInfo = newInfo;
    return 0;
}
2594

2595

2596 2597 2598
static int
qemuMigrationCheckJobStatus(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
2599 2600
                            qemuDomainAsyncJob asyncJob,
                            bool updateJobStats)
2601 2602 2603
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
2604

2605 2606 2607 2608 2609
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);

    if (events)
        qemuMigrationUpdateJobType(jobInfo);
    else if (qemuMigrationUpdateJobStatus(driver, vm, asyncJob) < 0)
2610
        return -1;
2611

2612 2613
    switch (jobInfo->type) {
    case VIR_DOMAIN_JOB_NONE:
2614 2615
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("is not active"));
2616
        return -1;
2617

2618
    case VIR_DOMAIN_JOB_FAILED:
2619 2620
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("unexpectedly failed"));
2621
        return -1;
2622

2623
    case VIR_DOMAIN_JOB_CANCELLED:
2624 2625
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("canceled by client"));
2626 2627
        return -1;

2628 2629
    case VIR_DOMAIN_JOB_COMPLETED:
        /* Fetch statistics of a completed migration */
2630
        if (events && updateJobStats &&
2631 2632 2633 2634
            qemuMigrationUpdateJobStatus(driver, vm, asyncJob) < 0)
            return -1;
        break;

2635 2636 2637
    case VIR_DOMAIN_JOB_BOUNDED:
    case VIR_DOMAIN_JOB_UNBOUNDED:
    case VIR_DOMAIN_JOB_LAST:
2638 2639
        break;
    }
2640
    return 0;
2641 2642 2643
}


2644 2645 2646
enum qemuMigrationCompletedFlags {
    QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR = (1 << 0),
    QEMU_MIGRATION_COMPLETED_CHECK_STORAGE  = (1 << 1),
2647
    QEMU_MIGRATION_COMPLETED_UPDATE_STATS   = (1 << 2),
2648 2649
};

2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660
/**
 * Returns 1 if migration completed successfully,
 *         0 if the domain is still being migrated,
 *         -1 migration failed,
 *         -2 something else failed, we need to cancel migration.
 */
static int
qemuMigrationCompleted(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
                       qemuDomainAsyncJob asyncJob,
                       virConnectPtr dconn,
2661
                       unsigned int flags)
2662 2663 2664 2665
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    int pauseReason;
2666
    bool updateStats = !!(flags & QEMU_MIGRATION_COMPLETED_UPDATE_STATS);
2667

2668
    if (qemuMigrationCheckJobStatus(driver, vm, asyncJob, updateStats) < 0)
2669 2670
        goto error;

2671 2672
    if (flags & QEMU_MIGRATION_COMPLETED_CHECK_STORAGE &&
        qemuMigrationDriveMirrorReady(driver, vm) < 0)
2673 2674
        goto error;

2675
    if (flags & QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR &&
2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707
        virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
        pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("failed due to I/O error"));
        goto error;
    }

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

    if (jobInfo->type == VIR_DOMAIN_JOB_COMPLETED)
        return 1;
    else
        return 0;

 error:
    if (jobInfo->type == VIR_DOMAIN_JOB_UNBOUNDED) {
        /* The migration was aborted by us rather than QEMU itself. */
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
        return -2;
    } else if (jobInfo->type == VIR_DOMAIN_JOB_COMPLETED) {
        jobInfo->type = VIR_DOMAIN_JOB_FAILED;
        return -1;
    } else {
        return -1;
    }
}


2708 2709 2710
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
2711
static int
J
Jiri Denemark 已提交
2712 2713
qemuMigrationWaitForCompletion(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
2714
                               qemuDomainAsyncJob asyncJob,
J
Jiri Denemark 已提交
2715
                               virConnectPtr dconn,
2716 2717
                               bool abort_on_error,
                               bool storage)
2718
{
2719
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
2720
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
2721
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
2722
    unsigned int flags = QEMU_MIGRATION_COMPLETED_UPDATE_STATS;
2723
    int rv;
2724

2725 2726 2727 2728 2729
    if (abort_on_error)
        flags |= QEMU_MIGRATION_COMPLETED_ABORT_ON_ERROR;
    if (storage)
        flags |= QEMU_MIGRATION_COMPLETED_CHECK_STORAGE;

J
Jiri Denemark 已提交
2730
    jobInfo->type = VIR_DOMAIN_JOB_UNBOUNDED;
2731 2732
    while ((rv = qemuMigrationCompleted(driver, vm, asyncJob,
                                        dconn, flags)) != 1) {
2733 2734
        if (rv < 0)
            return rv;
2735

2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748
        if (events) {
            if (virDomainObjWait(vm) < 0) {
                jobInfo->type = VIR_DOMAIN_JOB_FAILED;
                return -2;
            }
        } else {
            /* Poll every 50ms for progress & to allow cancellation */
            struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };

            virObjectUnlock(vm);
            nanosleep(&ts, NULL);
            virObjectLock(vm);
        }
2749 2750
    }

2751 2752 2753 2754 2755
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
    if (VIR_ALLOC(priv->job.completed) == 0)
        *priv->job.completed = *jobInfo;

2756
    return 0;
2757 2758 2759
}


2760
static int
2761
qemuDomainMigrateGraphicsRelocate(virQEMUDriverPtr driver,
2762
                                  virDomainObjPtr vm,
2763 2764
                                  qemuMigrationCookiePtr cookie,
                                  const char *graphicsuri)
2765 2766
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2767 2768
    int ret = -1;
    const char *listenAddress = NULL;
2769
    virSocketAddr addr;
2770 2771 2772 2773 2774
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
2775

2776
    if (!cookie || (!cookie->graphics && !graphicsuri))
2777 2778
        return 0;

2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797
    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) {
2798
        size_t i;
2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825

        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;
            }
        }
    }
2826 2827 2828 2829

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
2830 2831 2832 2833
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
2834

2835 2836 2837 2838
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
2839 2840
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
2841
    }
2842

2843
 cleanup:
2844
    virURIFree(uri);
2845 2846 2847 2848
    return ret;
}


2849
static int
2850
qemuDomainMigrateOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
2851 2852 2853
                             virDomainObjPtr vm,
                             qemuMigrationCookiePtr cookie)
{
2854 2855
    virDomainNetDefPtr netptr;
    int ret = -1;
2856
    size_t i;
2857 2858 2859 2860 2861 2862 2863 2864

    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:
2865
           break;
2866
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
2867 2868
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
2869 2870 2871
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
2872 2873 2874
                goto cleanup;
            }
            break;
2875 2876 2877 2878 2879
        default:
            break;
        }
    }

2880
    ret = 0;
2881
 cleanup:
2882 2883 2884 2885
    return ret;
}


2886 2887 2888 2889 2890 2891
/* 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.
 */
2892 2893 2894 2895
static virDomainObjPtr
qemuMigrationCleanup(virDomainObjPtr vm,
                     virConnectPtr conn,
                     void *opaque)
2896
{
2897
    virQEMUDriverPtr driver = opaque;
2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912
    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);

2913
    switch ((qemuMigrationJobPhase) priv->job.phase) {
2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943
    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 */
        ;
    }

2944
 cleanup:
2945 2946 2947
    return vm;
}

2948

2949
/* The caller is supposed to lock the vm and start a migration job. */
2950 2951 2952 2953 2954 2955 2956
static char *
qemuMigrationBeginPhase(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dname,
                        char **cookieout,
                        int *cookieoutlen,
2957 2958
                        size_t nmigrate_disks,
                        const char **migrate_disks,
2959
                        unsigned long flags)
2960 2961 2962
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
2963
    virDomainDefPtr def = NULL;
2964
    qemuDomainObjPrivatePtr priv = vm->privateData;
2965
    virCapsPtr caps = NULL;
2966
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
2967

2968
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
2969 2970
              " cookieout=%p, cookieoutlen=%p,"
              " nmigrate_disks=%zu, migrate_disks=%p, flags=%lx",
2971
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
2972 2973
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
2974

2975 2976 2977
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2978 2979 2980 2981 2982 2983
    /* 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);
2984

2985
    if (!qemuMigrationIsAllowed(driver, vm, true, flags))
2986 2987
        goto cleanup;

2988 2989
    if (!(flags & VIR_MIGRATE_UNSAFE) &&
        !qemuMigrationIsSafe(vm->def, nmigrate_disks, migrate_disks))
2990 2991
        goto cleanup;

2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035
    if (flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC)) {
        bool has_drive_mirror =  virQEMUCapsGet(priv->qemuCaps,
                                                QEMU_CAPS_DRIVE_MIRROR);

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

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

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

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

        if (has_drive_mirror) {
            /* TODO support NBD for TUNNELLED migration */
            if (flags & VIR_MIGRATE_TUNNELLED) {
                VIR_WARN("NBD in tunnelled migration is currently not supported");
            } else {
                cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
                priv->nbdPort = 0;
            }
3036 3037 3038
        }
    }

3039
    if (virDomainDefHasMemoryHotplug(vm->def) ||
3040
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
3041
         vm->newDef && virDomainDefHasMemoryHotplug(vm->newDef)))
3042 3043
        cookieFlags |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

3044
    if (!(mig = qemuMigrationEatCookie(driver, vm, NULL, 0, 0)))
3045 3046 3047 3048
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
                                cookieout, cookieoutlen,
3049
                                cookieFlags) < 0)
3050 3051
        goto cleanup;

L
liguang 已提交
3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073
    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;
        }
    }

3074
    if (xmlin) {
3075
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt,
3076
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE)))
3077 3078
            goto cleanup;

3079
        if (!qemuDomainDefCheckABIStability(driver, vm->def, def))
3080 3081
            goto cleanup;

3082
        rv = qemuDomainDefFormatLive(driver, def, false, true);
3083
    } else {
3084
        rv = qemuDomainDefFormatLive(driver, vm->def, false, true);
3085
    }
3086

3087
 cleanup:
3088
    qemuMigrationCookieFree(mig);
3089
    virObjectUnref(caps);
3090
    virDomainDefFree(def);
3091 3092 3093
    return rv;
}

3094 3095 3096 3097 3098 3099 3100
char *
qemuMigrationBegin(virConnectPtr conn,
                   virDomainObjPtr vm,
                   const char *xmlin,
                   const char *dname,
                   char **cookieout,
                   int *cookieoutlen,
3101 3102
                   size_t nmigrate_disks,
                   const char **migrate_disks,
3103 3104 3105 3106
                   unsigned long flags)
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
3107
    qemuDomainAsyncJob asyncJob;
3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118

    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;
    }

3119 3120
    qemuMigrationStoreDomainState(vm);

3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
    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,
3136
                                        nmigrate_disks, migrate_disks, flags)))
3137 3138 3139 3140 3141 3142 3143
        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.
         */
3144
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
3145 3146
                                 qemuMigrationCleanup) < 0) {
            VIR_FREE(xml);
3147
            goto endjob;
3148
        }
3149
        qemuMigrationJobContinue(vm);
3150 3151 3152 3153
    } else {
        goto endjob;
    }

3154
 cleanup:
M
Michal Privoznik 已提交
3155
    virDomainObjEndAPI(&vm);
3156 3157
    return xml;

3158
 endjob:
3159 3160 3161 3162
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
3163 3164 3165
    goto cleanup;
}

3166

3167 3168
/* Prepare is the first step, and it runs on the destination host.
 */
3169

3170
static void
3171
qemuMigrationPrepareCleanup(virQEMUDriverPtr driver,
3172 3173 3174 3175 3176 3177 3178 3179 3180 3181
                            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));

3182 3183 3184
    virPortAllocatorRelease(driver->migrationPorts, priv->migrationPort);
    priv->migrationPort = 0;

3185 3186 3187 3188 3189
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

3190
static int
3191
qemuMigrationPrepareAny(virQEMUDriverPtr driver,
3192 3193 3194 3195 3196
                        virConnectPtr dconn,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
3197
                        virDomainDefPtr *def,
3198
                        const char *origname,
L
liguang 已提交
3199
                        virStreamPtr st,
3200
                        const char *protocol,
3201 3202
                        unsigned short port,
                        bool autoPort,
3203
                        const char *listenAddress,
3204 3205
                        size_t nmigrate_disks,
                        const char **migrate_disks,
L
liguang 已提交
3206
                        unsigned long flags)
3207 3208
{
    virDomainObjPtr vm = NULL;
3209
    virObjectEventPtr event = NULL;
3210
    int ret = -1;
3211
    int dataFD[2] = { -1, -1 };
3212
    qemuDomainObjPrivatePtr priv = NULL;
3213
    qemuMigrationCookiePtr mig = NULL;
3214
    bool tunnel = !!st;
J
Jiri Denemark 已提交
3215
    char *xmlout = NULL;
L
liguang 已提交
3216
    unsigned int cookieFlags;
3217
    virCapsPtr caps = NULL;
J
Ján Tomko 已提交
3218
    char *migrateFrom = NULL;
3219
    bool taint_hook = false;
3220

3221 3222
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244
    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;
        }
    }

3245 3246 3247
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

3248
    if (!qemuMigrationIsAllowedHostdev(*def))
3249 3250
        goto cleanup;

J
Jiri Denemark 已提交
3251 3252 3253 3254 3255
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

3256
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
3257 3258
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
3259 3260
            goto cleanup;

3261
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
3262 3263 3264 3265 3266 3267 3268
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
3269
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
3270 3271 3272 3273 3274 3275
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
3276
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt,
3277
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE);
J
Jiri Denemark 已提交
3278 3279 3280
                if (!newdef)
                    goto cleanup;

3281
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
3282 3283 3284 3285
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

3286 3287
                virDomainDefFree(*def);
                *def = newdef;
3288 3289 3290 3291
                /* 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
3292 3293 3294 3295
            }
        }
    }

J
Ján Tomko 已提交
3296 3297 3298 3299
    if (tunnel) {
        /* QEMU will be started with -incoming stdio
         * (which qemu_command might convert to exec:cat or fd:n)
         */
3300
        if (VIR_STRDUP(migrateFrom, "stdio") < 0)
J
Ján Tomko 已提交
3301 3302
            goto cleanup;
    } else {
3303 3304 3305
        bool encloseAddress = false;
        bool hostIPv6Capable = false;
        bool qemuIPv6Capable = false;
J
Ján Tomko 已提交
3306 3307 3308 3309
        virQEMUCapsPtr qemuCaps = NULL;
        struct addrinfo *info = NULL;
        struct addrinfo hints = { .ai_flags = AI_ADDRCONFIG,
                                  .ai_socktype = SOCK_STREAM };
3310
        const char *incFormat;
J
Ján Tomko 已提交
3311

3312 3313 3314 3315
        if (getaddrinfo("::", NULL, &hints, &info) == 0) {
            freeaddrinfo(info);
            hostIPv6Capable = true;
        }
J
Ján Tomko 已提交
3316
        if (!(qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
3317 3318
                                                    (*def)->emulator,
                                                    (*def)->os.machine)))
J
Ján Tomko 已提交
3319 3320
            goto cleanup;

3321 3322 3323 3324
        qemuIPv6Capable = virQEMUCapsGet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
        virObjectUnref(qemuCaps);

        if (listenAddress) {
3325 3326 3327 3328 3329 3330 3331 3332 3333
            if (virSocketAddrNumericFamily(listenAddress) == AF_INET6) {
                if (!qemuIPv6Capable) {
                    virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                                   _("qemu isn't capable of IPv6"));
                    goto cleanup;
                }
                if (!hostIPv6Capable) {
                    virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                                   _("host isn't capable of IPv6"));
3334 3335
                    goto cleanup;
                }
3336 3337
                /* IPv6 address must be escaped in brackets on the cmd line */
                encloseAddress = true;
3338
            } else {
3339
                /* listenAddress is a hostname or IPv4 */
3340
            }
3341
        } else if (qemuIPv6Capable && hostIPv6Capable) {
3342 3343 3344
            /* Listen on :: instead of 0.0.0.0 if QEMU understands it
             * and there is at least one IPv6 address configured
             */
3345 3346 3347 3348
            listenAddress = "::";
            encloseAddress = true;
        } else {
            listenAddress = "0.0.0.0";
J
Ján Tomko 已提交
3349 3350
        }

3351 3352 3353 3354
        /* QEMU will be started with
         *   -incoming protocol:[<IPv6 addr>]:port,
         *   -incoming protocol:<IPv4 addr>:port, or
         *   -incoming protocol:<hostname>:port
J
Ján Tomko 已提交
3355
         */
3356 3357 3358 3359 3360 3361
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";
        if (virAsprintf(&migrateFrom, incFormat,
                        protocol, listenAddress, port) < 0)
J
Ján Tomko 已提交
3362 3363 3364
            goto cleanup;
    }

3365
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
3366
                                   driver->xmlopt,
3367 3368 3369
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
3370
        goto cleanup;
3371

3372
    virObjectRef(vm);
3373
    *def = NULL;
3374
    priv = vm->privateData;
3375 3376
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
3377

3378 3379 3380 3381 3382
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

3383
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
3384 3385
                                       QEMU_MIGRATION_COOKIE_LOCKSTATE |
                                       QEMU_MIGRATION_COOKIE_NBD)))
3386 3387
        goto cleanup;

3388 3389
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
3390 3391 3392 3393 3394 3395
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

3396
    if (qemuMigrationPrecreateStorage(dconn, driver, vm, mig->nbd,
3397 3398
                                      nmigrate_disks, migrate_disks,
                                      !!(flags & VIR_MIGRATE_NON_SHARED_INC)) < 0)
3399 3400
        goto cleanup;

3401
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
3402
        goto cleanup;
3403
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
3404 3405 3406 3407

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

L
liguang 已提交
3408 3409 3410
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

3411 3412
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
3413 3414
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
3415 3416 3417 3418
        goto endjob;
    }

    /* Start the QEMU daemon, with the same command-line arguments plus
3419
     * -incoming $migrateFrom
3420
     */
3421 3422
    if (qemuProcessStart(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                         migrateFrom, dataFD[0], NULL, NULL,
3423 3424
                         VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
                         VIR_QEMU_PROCESS_START_PAUSED |
3425
                         VIR_QEMU_PROCESS_START_AUTODESTROY) < 0) {
3426
        virDomainAuditStart(vm, "migrated", false);
3427 3428 3429
        goto endjob;
    }

3430 3431 3432 3433
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
3434
            goto stop;
3435
        }
3436
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
3437 3438
    }

3439 3440
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
3441 3442 3443
                                    QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stop;

3444
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
3445 3446 3447 3448
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
        goto stop;
    }

3449 3450 3451
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
                               QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
3452 3453
        goto stop;

3454 3455 3456 3457 3458 3459 3460 3461
    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");
    }
3462

3463
 done:
L
liguang 已提交
3464 3465 3466 3467 3468
    if (flags & VIR_MIGRATE_OFFLINE)
        cookieFlags = 0;
    else
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS;

3469 3470 3471
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
3472 3473
        if (qemuMigrationStartNBDServer(driver, vm, listenAddress,
                                        nmigrate_disks, migrate_disks) < 0) {
3474 3475
            /* error already reported */
            goto endjob;
3476
        }
3477
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
3478 3479 3480 3481
    }

    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
3482 3483 3484 3485 3486 3487 3488
        /* 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");
    }

3489 3490 3491
    if (qemuDomainCleanupAdd(vm, qemuMigrationPrepareCleanup) < 0)
        goto endjob;

L
liguang 已提交
3492 3493
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
3494
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
3495 3496 3497
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
3498

3499 3500 3501 3502
    /* 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.
     */
3503
    qemuMigrationJobContinue(vm);
3504

3505 3506
    if (autoPort)
        priv->migrationPort = port;
3507
    ret = 0;
3508

3509
 cleanup:
J
Ján Tomko 已提交
3510
    VIR_FREE(migrateFrom);
J
Jiri Denemark 已提交
3511
    VIR_FREE(xmlout);
3512 3513
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
3514 3515 3516
    if (ret < 0 && priv) {
        /* priv is set right after vm is added to the list of domains
         * and there is no 'goto cleanup;' in the middle of those */
3517
        VIR_FREE(priv->origname);
3518 3519
        virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
        priv->nbdPort = 0;
3520
        qemuDomainRemoveInactive(driver, vm);
3521
    }
M
Michal Privoznik 已提交
3522
    virDomainObjEndAPI(&vm);
3523
    qemuDomainEventQueue(driver, event);
3524
    qemuMigrationCookieFree(mig);
3525
    virObjectUnref(caps);
3526
    virNWFilterUnlockFilterUpdates();
3527
    return ret;
3528

3529
 stop:
3530
    virDomainAuditStart(vm, "migrated", false);
3531 3532
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                    VIR_QEMU_PROCESS_STOP_MIGRATED);
3533

3534
 endjob:
3535
    qemuMigrationJobFinish(driver, vm);
3536
    goto cleanup;
3537 3538 3539
}


3540 3541 3542 3543 3544
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
3545
qemuMigrationPrepareTunnel(virQEMUDriverPtr driver,
3546 3547 3548 3549 3550 3551
                           virConnectPtr dconn,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virStreamPtr st,
3552
                           virDomainDefPtr *def,
3553
                           const char *origname,
L
liguang 已提交
3554
                           unsigned long flags)
3555 3556 3557 3558
{
    int ret;

    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
3559
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
3560
              "origname=%s, flags=%lx",
3561
              driver, dconn, NULLSTR(cookiein), cookieinlen,
3562
              cookieout, cookieoutlen, st, *def, origname, flags);
3563

3564 3565 3566 3567 3568 3569
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

3570
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3571
                                  cookieout, cookieoutlen, def, origname,
3572
                                  st, NULL, 0, false, NULL, 0, NULL, flags);
3573 3574 3575 3576
    return ret;
}


3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599
static virURIPtr
qemuMigrationParseURI(const char *uri, bool *wellFormed)
{
    char *tmp = NULL;
    virURIPtr parsed;

    /* For compatibility reasons tcp://... URIs are sent as tcp:...
     * We need to transform them to a well-formed URI before parsing. */
    if (STRPREFIX(uri, "tcp:") && !STRPREFIX(uri + 4, "//")) {
        if (virAsprintf(&tmp, "tcp://%s", uri + 4) < 0)
            return NULL;
        uri = tmp;
    }

    parsed = virURIParse(uri);
    if (parsed && wellFormed)
        *wellFormed = !tmp;
    VIR_FREE(tmp);

    return parsed;
}


3600
int
3601
qemuMigrationPrepareDirect(virQEMUDriverPtr driver,
3602
                           virConnectPtr dconn,
3603 3604 3605 3606
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
3607 3608
                           const char *uri_in,
                           char **uri_out,
3609
                           virDomainDefPtr *def,
3610
                           const char *origname,
3611
                           const char *listenAddress,
3612 3613
                           size_t nmigrate_disks,
                           const char **migrate_disks,
L
liguang 已提交
3614
                           unsigned long flags)
3615
{
3616 3617
    unsigned short port = 0;
    bool autoPort = true;
3618 3619
    char *hostname = NULL;
    int ret = -1;
3620
    virURIPtr uri = NULL;
3621 3622
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
3623

3624 3625
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
3626 3627
              "def=%p, origname=%s, listenAddress=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p, flags=%lx",
3628 3629
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
3630 3631
              *def, origname, NULLSTR(listenAddress),
              nmigrate_disks, migrate_disks, flags);
3632

3633 3634
    *uri_out = NULL;

3635 3636 3637
    /* 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
3638 3639 3640
     * 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".
3641 3642 3643 3644 3645 3646
     *
     * 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) {
3647 3648 3649
        bool encloseAddress = false;
        const char *incFormat;

3650
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3651
            goto cleanup;
3652

3653
        if (migrateHost != NULL) {
3654 3655
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
3656

3657
            if (VIR_STRDUP(hostname, migrateHost) < 0)
3658 3659 3660 3661 3662
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
3663 3664

        if (STRPREFIX(hostname, "localhost")) {
3665 3666 3667
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
3668 3669 3670 3671 3672
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
3673
         * compatibility with old targets. We at least make the
3674 3675
         * new targets accept both syntaxes though.
         */
3676 3677 3678 3679 3680 3681
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
3682 3683
            goto cleanup;
    } else {
3684
        bool well_formed_uri;
J
Ján Tomko 已提交
3685

3686 3687
        if (!(uri = qemuMigrationParseURI(uri_in, &well_formed_uri)))
            goto cleanup;
J
Ján Tomko 已提交
3688

3689 3690 3691 3692 3693 3694 3695
        if (uri->scheme == NULL) {
            virReportError(VIR_ERR_INVALID_ARG,
                           _("missing scheme in migration URI: %s"),
                           uri_in);
            goto cleanup;
        }

M
Michael R. Hines 已提交
3696 3697
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
3698 3699 3700
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
3701 3702 3703 3704 3705 3706 3707 3708 3709 3710
            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) {
3711
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3712
                goto cleanup;
3713

3714
            /* Send well-formed URI only if uri_in was well-formed */
3715 3716 3717 3718 3719 3720 3721 3722
            if (well_formed_uri) {
                uri->port = port;
                if (!(*uri_out = virURIFormat(uri)))
                    goto cleanup;
            } else {
                if (virAsprintf(uri_out, "%s:%d", uri_in, port) < 0)
                    goto cleanup;
            }
3723
        } else {
3724 3725
            port = uri->port;
            autoPort = false;
3726 3727 3728 3729 3730 3731
        }
    }

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

3732
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3733
                                  cookieout, cookieoutlen, def, origname,
3734
                                  NULL, uri ? uri->scheme : "tcp",
3735 3736
                                  port, autoPort, listenAddress,
                                  nmigrate_disks, migrate_disks, flags);
3737
 cleanup:
3738
    virURIFree(uri);
3739
    VIR_FREE(hostname);
3740
    virObjectUnref(cfg);
3741
    if (ret != 0) {
3742
        VIR_FREE(*uri_out);
3743 3744 3745
        if (autoPort)
            virPortAllocatorRelease(driver->migrationPorts, port);
    }
3746 3747 3748 3749
    return ret;
}


3750 3751 3752
virDomainDefPtr
qemuMigrationPrepareDef(virQEMUDriverPtr driver,
                        const char *dom_xml,
3753 3754
                        const char *dname,
                        char **origname)
3755 3756 3757
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
3758
    char *name = NULL;
3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769

    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,
3770
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE)))
3771 3772 3773
        goto cleanup;

    if (dname) {
3774
        name = def->name;
3775 3776 3777 3778 3779 3780
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

3781
 cleanup:
3782
    virObjectUnref(caps);
3783 3784 3785 3786
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
3787 3788 3789 3790
    return def;
}


3791 3792 3793 3794 3795 3796 3797 3798 3799 3800
static int
qemuMigrationConfirmPhase(virQEMUDriverPtr driver,
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *cookiein,
                          int cookieinlen,
                          unsigned int flags,
                          int retcode)
{
    qemuMigrationCookiePtr mig;
3801
    virObjectEventPtr event = NULL;
3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816
    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);

3817 3818
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                       QEMU_MIGRATION_COOKIE_STATS)))
3819 3820
        goto cleanup;

3821 3822 3823
    /* Update total times with the values sent by the destination daemon */
    if (mig->jobInfo) {
        qemuDomainObjPrivatePtr priv = vm->privateData;
3824 3825 3826
        VIR_FREE(priv->job.completed);
        priv->job.completed = mig->jobInfo;
        mig->jobInfo = NULL;
3827 3828
    }

3829 3830 3831 3832 3833 3834 3835 3836 3837
    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 */
3838
        qemuMigrationWaitForSpice(vm);
3839 3840 3841 3842 3843

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

3844
        event = virDomainEventLifecycleNewFromObj(vm,
3845 3846 3847
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
    } else {
3848
        virErrorPtr orig_err = virSaveLastError();
3849 3850

        /* cancel any outstanding NBD jobs */
3851
        qemuMigrationCancelDriveMirror(driver, vm, false,
3852
                                       QEMU_ASYNC_JOB_MIGRATION_OUT, NULL);
3853 3854 3855

        virSetError(orig_err);
        virFreeError(orig_err);
3856

3857 3858 3859 3860
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
3861 3862
        }

3863
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
3864 3865 3866
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
    }

3867
 done:
3868 3869 3870
    qemuMigrationCookieFree(mig);
    rv = 0;

3871
 cleanup:
3872
    qemuDomainEventQueue(driver, event);
3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885
    virObjectUnref(cfg);
    return rv;
}

int
qemuMigrationConfirm(virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *cookiein,
                     int cookieinlen,
                     unsigned int flags,
                     int cancelled)
{
    virQEMUDriverPtr driver = conn->privateData;
3886
    qemuMigrationJobPhase phase;
3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900
    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);
3901 3902
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
3903 3904 3905 3906 3907

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

3908
    qemuMigrationJobFinish(driver, vm);
3909
    if (!virDomainObjIsActive(vm)) {
3910 3911 3912 3913 3914
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
        qemuDomainRemoveInactive(driver, vm);
    }

3915
 cleanup:
M
Michal Privoznik 已提交
3916
    virDomainObjEndAPI(&vm);
3917 3918 3919 3920 3921
    virObjectUnref(cfg);
    return ret;
}


3922 3923
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3924
    MIGRATION_DEST_CONNECT_HOST,
3925
    MIGRATION_DEST_UNIX,
3926
    MIGRATION_DEST_FD,
3927
};
3928

3929 3930 3931 3932
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3933

3934 3935 3936 3937 3938 3939
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3940
            const char *protocol;
3941 3942 3943 3944 3945
            const char *name;
            int port;
        } host;

        struct {
3946
            char *file;
3947 3948
            int sock;
        } unix_socket;
3949 3950 3951 3952 3953

        struct {
            int qemu;
            int local;
        } fd;
3954 3955 3956 3957 3958 3959 3960
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3961 3962 3963

#define TUNNEL_SEND_BUF_SIZE 65536

3964 3965 3966 3967 3968 3969 3970
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3971 3972
    int wakeupRecvFD;
    int wakeupSendFD;
3973 3974 3975
};

static void qemuMigrationIOFunc(void *arg)
3976
{
3977
    qemuMigrationIOThreadPtr data = arg;
3978 3979 3980 3981 3982 3983 3984
    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);
3985

3986
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3987
        goto abrt;
3988

3989 3990 3991
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3992
    for (;;) {
3993 3994 3995 3996 3997 3998 3999 4000 4001 4002
        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;
4003
            virReportSystemError(errno, "%s",
4004 4005
                                 _("poll failed in migration tunnel"));
            goto abrt;
4006
        }
4007 4008 4009 4010 4011 4012 4013

        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");
4014
            break;
4015
        }
4016

4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032
        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;
            }
4033 4034
        }

4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051
        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;
            }
        }
    }
4052

4053 4054
    if (virStreamFinish(data->st) < 0)
        goto error;
4055

4056 4057
    VIR_FREE(buffer);

4058 4059
    return;

4060
 abrt:
4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

4072
 error:
4073 4074
    virCopyLastError(&data->err);
    virResetLastError();
4075
    VIR_FREE(buffer);
4076 4077 4078 4079 4080 4081 4082
}


static qemuMigrationIOThreadPtr
qemuMigrationStartTunnel(virStreamPtr st,
                         int sock)
{
4083 4084
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
4085

4086 4087 4088 4089
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
4090 4091
    }

4092
    if (VIR_ALLOC(io) < 0)
4093
        goto error;
4094

4095 4096
    io->st = st;
    io->sock = sock;
4097 4098
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
4099 4100 4101 4102 4103 4104

    if (virThreadCreate(&io->thread, true,
                        qemuMigrationIOFunc,
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
4105
        goto error;
4106 4107 4108
    }

    return io;
4109

4110
 error:
4111 4112 4113 4114
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
4115 4116 4117
}

static int
4118
qemuMigrationStopTunnel(qemuMigrationIOThreadPtr io, bool error)
4119 4120
{
    int rv = -1;
4121 4122 4123 4124 4125 4126 4127 4128 4129
    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;
    }

4130 4131 4132 4133
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
4134 4135 4136 4137
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
4138 4139 4140 4141 4142 4143
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

4144
 cleanup:
4145 4146
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
4147 4148
    VIR_FREE(io);
    return rv;
4149 4150
}

4151
static int
4152
qemuMigrationConnect(virQEMUDriverPtr driver,
4153 4154 4155 4156 4157 4158 4159 4160 4161
                     virDomainObjPtr vm,
                     qemuMigrationSpecPtr spec)
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
4162
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
4163 4164 4165 4166 4167 4168 4169
        return -1;

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

    if (virSecurityManagerSetSocketLabel(driver->securityManager, vm->def) < 0)
        goto cleanup;
4170 4171 4172
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
4173
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
4174
        virObjectUnref(sock);
4175 4176 4177 4178 4179
    }
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0 ||
        spec->dest.fd.qemu == -1)
        goto cleanup;

4180 4181 4182 4183 4184 4185 4186
    /* 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;
    }

4187 4188
    ret = 0;

4189
 cleanup:
4190 4191 4192 4193 4194 4195
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

4196
static int
4197
qemuMigrationRun(virQEMUDriverPtr driver,
4198 4199 4200 4201 4202 4203 4204
                 virDomainObjPtr vm,
                 const char *cookiein,
                 int cookieinlen,
                 char **cookieout,
                 int *cookieoutlen,
                 unsigned long flags,
                 unsigned long resource,
4205
                 qemuMigrationSpecPtr spec,
4206
                 virConnectPtr dconn,
4207 4208 4209
                 const char *graphicsuri,
                 size_t nmigrate_disks,
                 const char **migrate_disks)
4210
{
4211
    int ret = -1;
4212 4213
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
4214
    qemuMigrationCookiePtr mig = NULL;
4215
    qemuMigrationIOThreadPtr iothread = NULL;
4216
    int fd = -1;
4217
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
4218
    virErrorPtr orig_err = NULL;
4219
    unsigned int cookieFlags = 0;
4220
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
4221
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
4222
    int rc;
4223 4224 4225

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
4226 4227
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
4228 4229
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
4230
              spec, spec->destType, spec->fwdType, dconn,
4231
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4232

4233 4234 4235 4236 4237 4238 4239 4240 4241 4242
    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;
    }

4243 4244
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
4245 4246 4247 4248
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
4249 4250 4251
        return -1;
    }

4252 4253 4254
    if (events)
        priv->signalIOError = abort_on_error;

4255 4256 4257
    mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                 cookieFlags | QEMU_MIGRATION_COOKIE_GRAPHICS);
    if (!mig)
4258 4259
        goto cleanup;

4260
    if (qemuDomainMigrateGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
4261 4262
        VIR_WARN("unable to provide data for graphics client relocation");

4263 4264 4265 4266 4267 4268 4269
    if (migrate_flags & (QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                         QEMU_MONITOR_MIGRATE_NON_SHARED_INC)) {
        if (mig->nbd) {
            /* This will update migrate_flags on success */
            if (qemuMigrationDriveMirror(driver, vm, mig,
                                         spec->dest.host.name,
                                         migrate_speed,
4270 4271
                                         &migrate_flags,
                                         nmigrate_disks,
4272 4273
                                         migrate_disks,
                                         dconn) < 0) {
4274 4275 4276 4277 4278 4279 4280 4281
                goto cleanup;
            }
        } else {
            /* Destination doesn't support NBD server.
             * Fall back to previous implementation. */
            VIR_DEBUG("Destination doesn't support NBD server "
                      "Falling back to previous implementation.");
        }
4282 4283
    }

4284
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
4285 4286 4287 4288 4289 4290
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

4291 4292
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
4293 4294 4295
                                    QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

4296 4297
    if (qemuMigrationSetAutoConverge(driver, vm,
                                     flags & VIR_MIGRATE_AUTO_CONVERGE,
4298 4299 4300
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

4301 4302
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
4303 4304 4305
                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

4306 4307
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4308 4309
        goto cleanup;

4310
    if (priv->job.abortJob) {
4311 4312
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
4313
         * priv->job.abortJob will not change */
4314
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
4315
        priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
4316 4317 4318 4319 4320 4321
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
        goto cleanup;
    }

4322 4323
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
4324

4325 4326
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
4327
        qemuMigrationConnect(driver, vm, spec) < 0) {
4328
        goto exit_monitor;
4329
    }
4330

4331 4332
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
4333 4334
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
4335
            goto exit_monitor;
M
Michael R. Hines 已提交
4336
        }
4337
        ret = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
4338
                                       spec->dest.host.protocol,
4339 4340 4341 4342
                                       spec->dest.host.name,
                                       spec->dest.host.port);
        break;

4343 4344 4345 4346
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

4347
    case MIGRATION_DEST_UNIX:
4348
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX)) {
4349 4350 4351 4352 4353 4354 4355 4356 4357
            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;
4358 4359

    case MIGRATION_DEST_FD:
4360
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
4361
            fd = spec->dest.fd.local;
4362 4363
            spec->dest.fd.local = -1;
        }
4364 4365 4366 4367
        ret = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                     spec->dest.fd.qemu);
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
4368
    }
4369 4370
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
4371
    if (ret < 0)
4372 4373
        goto cleanup;
    ret = -1;
4374 4375 4376 4377

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

4378 4379 4380 4381 4382
    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.
         */
4383
        if (qemuMigrationCheckJobStatus(driver, vm,
4384 4385
                                        QEMU_ASYNC_JOB_MIGRATION_OUT,
                                        false) < 0)
4386
            goto cancel;
4387

4388 4389 4390 4391 4392 4393 4394
        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;
        }
4395 4396
    }

4397 4398
    if (spec->fwdType != MIGRATION_FWD_DIRECT &&
        !(iothread = qemuMigrationStartTunnel(spec->fwd.stream, fd)))
4399 4400
        goto cancel;

4401 4402
    rc = qemuMigrationWaitForCompletion(driver, vm,
                                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4403
                                        dconn, abort_on_error, !!mig->nbd);
4404 4405 4406
    if (rc == -2)
        goto cancel;
    else if (rc == -1)
4407
        goto cleanup;
4408

4409 4410 4411 4412 4413 4414 4415 4416 4417
    /* 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) {
4418 4419
        if (qemuMigrationSetOffline(driver, vm) < 0) {
            priv->job.current->type = VIR_DOMAIN_JOB_FAILED;
4420
            goto cleanup;
4421
        }
4422 4423
    }

4424
    ret = 0;
4425

4426
 cleanup:
4427 4428 4429
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

4430
    /* cancel any outstanding NBD jobs */
4431
    if (mig && mig->nbd) {
4432
        if (qemuMigrationCancelDriveMirror(driver, vm, ret == 0,
4433 4434
                                           QEMU_ASYNC_JOB_MIGRATION_OUT,
                                           dconn) < 0)
4435 4436
            ret = -1;
    }
4437

4438
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
4439
        if (iothread && qemuMigrationStopTunnel(iothread, ret < 0) < 0)
4440 4441
            ret = -1;
    }
4442
    VIR_FORCE_CLOSE(fd);
4443

4444
    if (priv->job.completed) {
4445
        qemuDomainJobInfoUpdateTime(priv->job.completed);
4446
        qemuDomainJobInfoUpdateDowntime(priv->job.completed);
4447
        ignore_value(virTimeMillisNow(&priv->job.completed->sent));
4448
    }
4449

4450 4451 4452
    if (priv->job.current->type == VIR_DOMAIN_JOB_UNBOUNDED)
        priv->job.current->type = VIR_DOMAIN_JOB_FAILED;

4453 4454
    cookieFlags |= QEMU_MIGRATION_COOKIE_NETWORK |
                   QEMU_MIGRATION_COOKIE_STATS;
4455 4456
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookieFlags |= QEMU_MIGRATION_COOKIE_PERSISTENT;
4457
    if (ret == 0 &&
4458 4459
        qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
4460
        VIR_WARN("Unable to encode migration cookie");
4461
    }
4462

4463 4464
    qemuMigrationCookieFree(mig);

4465 4466 4467
    if (events)
        priv->signalIOError = false;

4468 4469 4470 4471 4472
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4473 4474
    return ret;

4475 4476 4477 4478
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;

4479
 cancel:
4480 4481
    orig_err = virSaveLastError();

4482
    if (virDomainObjIsActive(vm)) {
4483 4484
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
4485
            qemuMonitorMigrateCancel(priv->mon);
4486
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
4487
        }
4488
    }
4489 4490 4491
    goto cleanup;
}

4492
/* Perform migration using QEMU's native migrate support,
4493 4494
 * not encrypted obviously
 */
4495
static int doNativeMigrate(virQEMUDriverPtr driver,
4496 4497 4498 4499 4500 4501 4502
                           virDomainObjPtr vm,
                           const char *uri,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4503
                           unsigned long resource,
4504
                           virConnectPtr dconn,
4505 4506 4507
                           const char *graphicsuri,
                           size_t nmigrate_disks,
                           const char **migrate_disks)
4508
{
4509
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
4510
    virURIPtr uribits = NULL;
4511
    int ret = -1;
4512 4513 4514
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
4515
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
4516
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
4517
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
4518
              cookieout, cookieoutlen, flags, resource,
4519
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4520

4521
    if (!(uribits = qemuMigrationParseURI(uri, NULL)))
4522 4523
        return -1;

4524 4525 4526 4527 4528 4529 4530
    if (uribits->scheme == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("missing scheme in migration URI: %s"),
                       uri);
        goto cleanup;
    }

M
Michael R. Hines 已提交
4531 4532 4533 4534 4535
    if (STREQ(uribits->scheme, "rdma")) {
        if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_RDMA)) {
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                           _("outgoing RDMA migration is not supported "
                             "with this QEMU binary"));
4536
            goto cleanup;
M
Michael R. Hines 已提交
4537
        }
4538
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
4539 4540 4541
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
4542
            goto cleanup;
M
Michael R. Hines 已提交
4543 4544 4545 4546 4547
        }
    }

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
        STRNEQ(uribits->scheme, "rdma"))
4548 4549
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
4550
        spec.destType = MIGRATION_DEST_HOST;
4551
    spec.dest.host.protocol = uribits->scheme;
4552 4553 4554
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
4555

4556
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4557
                           cookieoutlen, flags, resource, &spec, dconn,
4558
                           graphicsuri, nmigrate_disks, migrate_disks);
4559 4560 4561 4562

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

4563
 cleanup:
4564
    virURIFree(uribits);
4565 4566 4567 4568 4569

    return ret;
}


4570
static int doTunnelMigrate(virQEMUDriverPtr driver,
4571 4572 4573 4574 4575 4576 4577
                           virDomainObjPtr vm,
                           virStreamPtr st,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4578
                           unsigned long resource,
4579
                           virConnectPtr dconn,
4580 4581 4582
                           const char *graphicsuri,
                           size_t nmigrate_disks,
                           const char **migrate_disks)
4583 4584
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
4585
    virNetSocketPtr sock = NULL;
4586 4587
    int ret = -1;
    qemuMigrationSpec spec;
4588
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4589 4590

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
4591
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
4592
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
4593
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
4594
              cookieout, cookieoutlen, flags, resource,
4595
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4596

4597 4598 4599
    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)) {
4600 4601
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Source qemu is too old to support tunnelled migration"));
4602
        virObjectUnref(cfg);
4603
        return -1;
4604 4605 4606 4607 4608
    }

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

4609
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD)) {
4610 4611 4612 4613 4614 4615
        int fds[2];

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

4616
        if (pipe2(fds, O_CLOEXEC) == 0) {
4617 4618 4619 4620
            spec.dest.fd.qemu = fds[1];
            spec.dest.fd.local = fds[0];
        }
        if (spec.dest.fd.qemu == -1 ||
4621
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
4622
                                              spec.dest.fd.qemu) < 0) {
4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633
            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",
4634
                        cfg->libDir, vm->def->name) < 0)
4635 4636 4637
            goto cleanup;

        if (virNetSocketNewListenUNIX(spec.dest.unix_socket.file, 0700,
4638
                                      cfg->user, cfg->group,
4639 4640 4641 4642 4643 4644 4645
                                      &sock) < 0 ||
            virNetSocketListen(sock, 1) < 0)
            goto cleanup;

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

4646
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4647
                           cookieoutlen, flags, resource, &spec, dconn,
4648
                           graphicsuri, nmigrate_disks, migrate_disks);
4649

4650
 cleanup:
4651 4652 4653 4654
    if (spec.destType == MIGRATION_DEST_FD) {
        VIR_FORCE_CLOSE(spec.dest.fd.qemu);
        VIR_FORCE_CLOSE(spec.dest.fd.local);
    } else {
4655
        virObjectUnref(sock);
4656 4657
        VIR_FREE(spec.dest.unix_socket.file);
    }
4658

4659
    virObjectUnref(cfg);
4660 4661 4662 4663
    return ret;
}


4664 4665 4666 4667
/* 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 */
4668
static int doPeer2PeerMigrate2(virQEMUDriverPtr driver,
4669
                               virConnectPtr sconn ATTRIBUTE_UNUSED,
4670 4671
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
4672
                               const char *dconnuri,
4673 4674 4675
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
4676 4677 4678
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
4679
    char *cookie = NULL;
4680 4681 4682
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
4683
    bool cancelled;
4684
    virStreamPtr st = NULL;
4685
    unsigned long destflags;
4686

4687
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
4688
              "flags=%lx, dname=%s, resource=%lu",
4689 4690
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
4691

4692 4693 4694 4695 4696
    /* 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,
4697 4698
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
4699 4700 4701 4702 4703
        return -1;

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

4704 4705
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4706

4707 4708 4709 4710 4711 4712 4713 4714 4715 4716
    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;

4717
        qemuDomainObjEnterRemote(vm);
4718
        ret = dconn->driver->domainMigratePrepareTunnel
4719
            (dconn, st, destflags, dname, resource, dom_xml);
4720
        qemuDomainObjExitRemote(vm);
4721
    } else {
4722
        qemuDomainObjEnterRemote(vm);
4723 4724
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
4725
             destflags, dname, resource, dom_xml);
4726
        qemuDomainObjExitRemote(vm);
4727 4728 4729
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
4730 4731 4732
        goto cleanup;

    /* the domain may have shutdown or crashed while we had the locks dropped
4733
     * in qemuDomainObjEnterRemote, so check again
4734 4735
     */
    if (!virDomainObjIsActive(vm)) {
4736 4737
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4738 4739 4740
        goto cleanup;
    }

4741 4742
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
4743 4744
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
4745
        cancelled = true;
4746
        orig_err = virSaveLastError();
4747
        goto finish;
4748 4749
    }

4750 4751 4752 4753
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
4754
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4755
    if (flags & VIR_MIGRATE_TUNNELLED)
4756 4757
        ret = doTunnelMigrate(driver, vm, st,
                              NULL, 0, NULL, NULL,
4758 4759
                              flags, resource, dconn,
                              NULL, 0, NULL);
4760 4761 4762 4763
    else
        ret = doNativeMigrate(driver, vm, uri_out,
                              cookie, cookielen,
                              NULL, NULL, /* No out cookie with v2 migration */
4764
                              flags, resource, dconn, NULL, 0, NULL);
4765 4766 4767 4768

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4769

4770 4771 4772
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4773
    cancelled = ret < 0;
4774

4775
 finish:
4776 4777 4778 4779
    /* 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.
     */
4780
    dname = dname ? dname : vm->def->name;
4781
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4782
    qemuDomainObjEnterRemote(vm);
4783
    ddomain = dconn->driver->domainMigrateFinish2
4784
        (dconn, dname, cookie, cookielen,
4785
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4786
    qemuDomainObjExitRemote(vm);
4787 4788
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4789

4790
 cleanup:
4791
    if (ddomain) {
4792
        virObjectUnref(ddomain);
4793 4794 4795 4796
        ret = 0;
    } else {
        ret = -1;
    }
4797

4798
    virObjectUnref(st);
4799 4800 4801 4802 4803 4804

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4805
    VIR_FREE(cookie);
4806 4807

    return ret;
4808 4809 4810
}


4811 4812 4813 4814
/* 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 */
4815 4816 4817 4818 4819 4820 4821 4822 4823
static int
doPeer2PeerMigrate3(virQEMUDriverPtr driver,
                    virConnectPtr sconn,
                    virConnectPtr dconn,
                    const char *dconnuri,
                    virDomainObjPtr vm,
                    const char *xmlin,
                    const char *dname,
                    const char *uri,
4824
                    const char *graphicsuri,
4825
                    const char *listenAddress,
4826 4827
                    size_t nmigrate_disks,
                    const char **migrate_disks,
4828 4829 4830
                    unsigned long long bandwidth,
                    bool useParams,
                    unsigned long flags)
4831 4832 4833 4834 4835 4836 4837 4838 4839 4840
{
    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;
4841
    bool cancelled = true;
4842
    virStreamPtr st = NULL;
4843
    unsigned long destflags;
4844 4845 4846
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4847
    size_t i;
4848 4849

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4850
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4851 4852
              "nmigrate_disks=%zu, migrate_disks=%p, bandwidth=%llu, "
              "useParams=%d, flags=%lx",
4853
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4854
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4855 4856
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks,
              bandwidth, useParams, flags);
4857

4858 4859 4860 4861 4862
    /* 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.  */

4863
    dom_xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
4864 4865
                                      &cookieout, &cookieoutlen,
                                      nmigrate_disks, migrate_disks, flags);
4866 4867 4868
    if (!dom_xml)
        goto cleanup;

4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888
    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;
4889 4890 4891 4892 4893 4894

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4895 4896 4897 4898 4899
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4900 4901 4902 4903 4904
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4905 4906
    }

4907 4908 4909
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4910 4911
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4912

4913 4914 4915 4916 4917 4918 4919 4920 4921
    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;

4922
        qemuDomainObjEnterRemote(vm);
4923 4924 4925 4926 4927 4928 4929 4930 4931
        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);
        }
4932
        qemuDomainObjExitRemote(vm);
4933
    } else {
4934
        qemuDomainObjEnterRemote(vm);
4935 4936 4937 4938 4939 4940 4941 4942 4943
        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);
        }
4944
        qemuDomainObjExitRemote(vm);
4945 4946 4947 4948 4949
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

L
liguang 已提交
4950 4951 4952 4953
    if (flags & VIR_MIGRATE_OFFLINE) {
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4954
        cancelled = false;
L
liguang 已提交
4955 4956 4957
        goto finish;
    }

4958 4959 4960 4961
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4962 4963
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4964
            goto finish;
4965
        }
4966
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4967 4968
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4969
        orig_err = virSaveLastError();
4970 4971 4972 4973 4974 4975 4976 4977
        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.
     */
4978
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4979
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4980 4981 4982 4983 4984
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4985
    if (flags & VIR_MIGRATE_TUNNELLED) {
4986 4987 4988
        ret = doTunnelMigrate(driver, vm, st,
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4989 4990
                              flags, bandwidth, dconn, graphicsuri,
                              nmigrate_disks, migrate_disks);
4991 4992
    } else {
        ret = doNativeMigrate(driver, vm, uri,
4993 4994
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4995 4996
                              flags, bandwidth, dconn, graphicsuri,
                              nmigrate_disks, migrate_disks);
4997
    }
4998 4999

    /* Perform failed. Make sure Finish doesn't overwrite the error */
5000
    if (ret < 0) {
5001
        orig_err = virSaveLastError();
5002 5003 5004 5005
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
5006 5007 5008 5009

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
5010
    cancelled = ret < 0;
5011

5012
 finish:
5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024
    /*
     * 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;
5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047

    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);
    }
5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067

    if (cancelled) {
        if (ddomain) {
            VIR_ERROR(_("finish step ignored that migration was cancelled"));
        } else {
            /* If Finish reported a useful error, use it instead of the
             * original "migration unexpectedly failed" error.
             *
             * This is ugly but we can't do better with the APIs we have. We
             * only replace the error if Finish was called with cancelled == 1
             * and reported a real error (old libvirt would report an error
             * from RPC instead of MIGRATE_FINISH_OK), which only happens when
             * the domain died on destination. To further reduce a possibility
             * of false positives we also check that Perform returned
             * VIR_ERR_OPERATION_FAILED.
             */
            if (orig_err &&
                orig_err->domain == VIR_FROM_QEMU &&
                orig_err->code == VIR_ERR_OPERATION_FAILED) {
                virErrorPtr err = virGetLastError();
5068 5069
                if (err &&
                    err->domain == VIR_FROM_QEMU &&
5070 5071 5072 5073 5074 5075 5076
                    err->code != VIR_ERR_MIGRATE_FINISH_OK) {
                    virFreeError(orig_err);
                    orig_err = NULL;
                }
            }
        }
    }
5077

5078 5079 5080 5081 5082 5083 5084
    /* 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.
5085
     */
5086
    cancelled = ddomain == NULL;
5087

5088 5089 5090 5091 5092 5093
    /* 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();

5094 5095 5096 5097
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
5098
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
5099 5100 5101 5102 5103
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
5104 5105 5106
    ret = qemuMigrationConfirmPhase(driver, sconn, vm,
                                    cookiein, cookieinlen,
                                    flags, cancelled);
5107 5108 5109 5110
    /* 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.
     */
5111 5112 5113
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
5114 5115 5116

 cleanup:
    if (ddomain) {
5117
        virObjectUnref(ddomain);
5118 5119 5120 5121 5122
        ret = 0;
    } else {
        ret = -1;
    }

5123
    virObjectUnref(st);
5124 5125 5126 5127 5128 5129 5130 5131

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
5132
    virTypedParamsFree(params, nparams);
5133 5134 5135 5136
    return ret;
}


5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148
static void
qemuMigrationConnectionClosed(virConnectPtr conn,
                              int reason,
                              void *opaque)
{
    virDomainObjPtr vm = opaque;

    VIR_DEBUG("conn=%p, reason=%d, vm=%s", conn, reason, vm->def->name);
    virDomainObjBroadcast(vm);
}


5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


5161
static int doPeer2PeerMigrate(virQEMUDriverPtr driver,
5162
                              virConnectPtr sconn,
5163
                              virDomainObjPtr vm,
5164
                              const char *xmlin,
5165
                              const char *dconnuri,
5166
                              const char *uri,
5167
                              const char *graphicsuri,
5168
                              const char *listenAddress,
5169 5170
                              size_t nmigrate_disks,
                              const char **migrate_disks,
5171 5172
                              unsigned long flags,
                              const char *dname,
5173 5174
                              unsigned long resource,
                              bool *v3proto)
5175 5176 5177 5178
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
5179
    virErrorPtr orig_err = NULL;
5180
    bool offline = false;
5181
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5182
    bool useParams;
5183

5184 5185 5186
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, uri=%s, "
              "graphicsuri=%s, listenAddress=%s, nmigrate_disks=%zu, "
              "migrate_disks=%p, flags=%lx, dname=%s, resource=%lu",
5187
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
5188
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
5189
              nmigrate_disks, migrate_disks, flags, NULLSTR(dname), resource);
5190

5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204
    if (flags & VIR_MIGRATE_TUNNELLED && uri) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("migration URI is not supported by tunnelled "
                         "migration"));
        goto cleanup;
    }

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

5205 5206 5207 5208
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

5209
    qemuDomainObjEnterRemote(vm);
5210
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
5211
    qemuDomainObjExitRemote(vm);
5212
    if (dconn == NULL) {
5213
        virReportError(VIR_ERR_OPERATION_FAILED,
5214 5215
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
5216
        virObjectUnref(cfg);
5217 5218 5219
        return -1;
    }

5220 5221
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
5222 5223
        goto cleanup;

5224 5225 5226 5227 5228
    if (virConnectRegisterCloseCallback(dconn, qemuMigrationConnectionClosed,
                                        vm, NULL) < 0) {
        goto cleanup;
    }

5229
    qemuDomainObjEnterRemote(vm);
5230 5231
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
5232
        /* v3proto reflects whether the caller used Perform3, but with
5233
         * p2p migrate, regardless of whether Perform2 or Perform3
5234 5235 5236 5237
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
5238 5239
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
L
liguang 已提交
5240 5241 5242
    if (flags & VIR_MIGRATE_OFFLINE)
        offline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                           VIR_DRV_FEATURE_MIGRATION_OFFLINE);
5243
    qemuDomainObjExitRemote(vm);
5244

5245
    if (!p2p) {
5246 5247
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
5248 5249 5250
        goto cleanup;
    }

5251 5252
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
5253
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
5254 5255 5256 5257 5258 5259
        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
5260 5261 5262 5263 5264 5265 5266
    if (flags & VIR_MIGRATE_OFFLINE && !offline) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

5267
    /* domain may have been stopped while we were talking to remote daemon */
L
liguang 已提交
5268
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
5269 5270
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
5271 5272 5273
        goto cleanup;
    }

5274 5275 5276 5277 5278 5279 5280
    /* 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;

5281 5282
    if (*v3proto) {
        ret = doPeer2PeerMigrate3(driver, sconn, dconn, dconnuri, vm, xmlin,
5283
                                  dname, uri, graphicsuri, listenAddress,
5284 5285
                                  nmigrate_disks, migrate_disks, resource,
                                  useParams, flags);
5286
    } else {
5287
        ret = doPeer2PeerMigrate2(driver, sconn, dconn, vm,
5288
                                  dconnuri, flags, dname, resource);
5289
    }
5290

5291
 cleanup:
5292
    orig_err = virSaveLastError();
5293
    qemuDomainObjEnterRemote(vm);
5294
    virConnectUnregisterCloseCallback(dconn, qemuMigrationConnectionClosed);
5295
    virObjectUnref(dconn);
5296
    qemuDomainObjExitRemote(vm);
5297 5298 5299 5300
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5301
    virObjectUnref(cfg);
5302 5303 5304 5305
    return ret;
}


5306 5307 5308 5309 5310 5311
/*
 * 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
5312
qemuMigrationPerformJob(virQEMUDriverPtr driver,
5313 5314 5315 5316 5317
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dconnuri,
                        const char *uri,
5318
                        const char *graphicsuri,
5319
                        const char *listenAddress,
5320 5321
                        size_t nmigrate_disks,
                        const char **migrate_disks,
5322
                        const char *cookiein,
5323 5324 5325 5326 5327 5328 5329
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
5330
{
5331
    virObjectEventPtr event = NULL;
5332
    int ret = -1;
5333
    virErrorPtr orig_err = NULL;
5334
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5335

5336
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
5337 5338
        goto cleanup;

L
liguang 已提交
5339
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
5340 5341
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
5342 5343 5344
        goto endjob;
    }

5345
    if (!qemuMigrationIsAllowed(driver, vm, true, flags))
5346
        goto endjob;
5347

5348 5349
    if (!(flags & VIR_MIGRATE_UNSAFE) &&
        !qemuMigrationIsSafe(vm->def, nmigrate_disks, migrate_disks))
5350
        goto endjob;
5351

5352
    qemuMigrationStoreDomainState(vm);
5353 5354

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
5355
        ret = doPeer2PeerMigrate(driver, conn, vm, xmlin,
5356
                                 dconnuri, uri, graphicsuri, listenAddress,
5357
                                 nmigrate_disks, migrate_disks,
5358
                                 flags, dname, resource, &v3proto);
5359
    } else {
5360 5361 5362
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
        ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                              cookieout, cookieoutlen,
5363
                              flags, resource, NULL, NULL, 0, NULL);
5364
    }
5365 5366
    if (ret < 0)
        goto endjob;
5367

5368 5369 5370 5371
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
5372
    if (!v3proto) {
5373 5374
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
5375
        virDomainAuditStop(vm, "migrated");
5376
        event = virDomainEventLifecycleNewFromObj(vm,
5377 5378
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
5379 5380
    }

5381
 endjob:
5382 5383 5384
    if (ret < 0)
        orig_err = virSaveLastError();

5385
    if (qemuMigrationRestoreDomainState(conn, vm)) {
5386
        event = virDomainEventLifecycleNewFromObj(vm,
5387 5388 5389
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
5390

5391
    qemuMigrationJobFinish(driver, vm);
5392
    if (!virDomainObjIsActive(vm) && ret == 0) {
5393
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
5394
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
5395
        qemuDomainRemoveInactive(driver, vm);
5396 5397
    }

5398 5399 5400 5401 5402
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

5403
 cleanup:
M
Michal Privoznik 已提交
5404
    virDomainObjEndAPI(&vm);
5405
    qemuDomainEventQueue(driver, event);
5406
    virObjectUnref(cfg);
5407 5408 5409 5410 5411 5412 5413
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
5414
qemuMigrationPerformPhase(virQEMUDriverPtr driver,
5415 5416 5417
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *uri,
5418
                          const char *graphicsuri,
5419 5420
                          size_t nmigrate_disks,
                          const char **migrate_disks,
5421 5422 5423 5424 5425 5426 5427
                          const char *cookiein,
                          int cookieinlen,
                          char **cookieout,
                          int *cookieoutlen,
                          unsigned long flags,
                          unsigned long resource)
{
5428
    virObjectEventPtr event = NULL;
5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439
    int ret = -1;

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

    qemuMigrationJobStartPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
5440 5441
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
5442 5443 5444

    ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                          cookieout, cookieoutlen,
5445 5446
                          flags, resource, NULL, graphicsuri,
                          nmigrate_disks, migrate_disks);
5447

5448 5449 5450 5451 5452
    if (ret < 0) {
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
5453
        }
5454
        goto endjob;
5455
    }
5456 5457 5458

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

5459 5460
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                             qemuMigrationCleanup) < 0)
5461 5462
        goto endjob;

5463
 endjob:
5464
    if (ret < 0)
5465
        qemuMigrationJobFinish(driver, vm);
5466
    else
5467
        qemuMigrationJobContinue(vm);
5468
    if (!virDomainObjIsActive(vm))
5469
        qemuDomainRemoveInactive(driver, vm);
5470

5471
 cleanup:
M
Michal Privoznik 已提交
5472
    virDomainObjEndAPI(&vm);
5473
    qemuDomainEventQueue(driver, event);
5474 5475 5476
    return ret;
}

5477
int
5478
qemuMigrationPerform(virQEMUDriverPtr driver,
5479 5480 5481 5482 5483
                     virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *xmlin,
                     const char *dconnuri,
                     const char *uri,
5484
                     const char *graphicsuri,
5485
                     const char *listenAddress,
5486 5487
                     size_t nmigrate_disks,
                     const char **migrate_disks,
5488 5489 5490 5491 5492 5493 5494 5495 5496 5497
                     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, "
5498 5499
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p, "
5500 5501
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
              "flags=%lx, dname=%s, resource=%lu, v3proto=%d",
5502
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
5503
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
5504 5505
              nmigrate_disks, migrate_disks, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, NULLSTR(dname), resource, v3proto);
5506 5507 5508

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
5509 5510
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
5511 5512 5513 5514
            return -1;
        }

        return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri, uri,
5515
                                       graphicsuri, listenAddress,
5516
                                       nmigrate_disks, migrate_disks,
5517
                                       cookiein, cookieinlen,
5518 5519
                                       cookieout, cookieoutlen,
                                       flags, dname, resource, v3proto);
5520 5521
    } else {
        if (dconnuri) {
5522 5523
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
5524 5525 5526 5527 5528
            return -1;
        }

        if (v3proto) {
            return qemuMigrationPerformPhase(driver, conn, vm, uri,
5529
                                             graphicsuri,
5530
                                             nmigrate_disks, migrate_disks,
5531 5532
                                             cookiein, cookieinlen,
                                             cookieout, cookieoutlen,
5533
                                             flags, resource);
5534 5535
        } else {
            return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri,
5536
                                           uri, graphicsuri, listenAddress,
5537
                                           nmigrate_disks, migrate_disks,
5538
                                           cookiein, cookieinlen,
5539 5540 5541 5542 5543
                                           cookieout, cookieoutlen, flags,
                                           dname, resource, v3proto);
        }
    }
}
5544

5545
static int
5546 5547
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def)
{
5548
    size_t i;
5549 5550 5551 5552 5553
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
5554
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5555
            if (virNetDevVPortProfileAssociate(net->ifname,
5556
                                               virDomainNetGetActualVirtPortProfile(net),
5557
                                               &net->mac,
5558
                                               virDomainNetGetActualDirectDev(net),
5559
                                               -1,
5560
                                               def->uuid,
5561 5562
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
5563 5564 5565
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
5566
                goto err_exit;
5567
            }
5568
            last_good_net = i;
5569
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
5570

5571
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
5572 5573 5574 5575
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
5576 5577 5578
        }
    }

5579
    return 0;
5580

5581
 err_exit:
5582
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
5583
        net = def->nets[i];
5584
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5585
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
5586
                                                           virDomainNetGetActualVirtPortProfile(net),
5587
                                                           &net->mac,
5588
                                                           virDomainNetGetActualDirectDev(net),
5589
                                                           -1,
5590
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
5591 5592
        }
    }
5593
    return -1;
5594 5595 5596
}


J
Jiri Denemark 已提交
5597 5598 5599
static int
qemuMigrationPersist(virQEMUDriverPtr driver,
                     virDomainObjPtr vm,
5600 5601
                     qemuMigrationCookiePtr mig,
                     bool ignoreSaveError)
J
Jiri Denemark 已提交
5602 5603 5604 5605
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    virDomainDefPtr vmdef;
5606 5607
    virDomainDefPtr oldDef = NULL;
    unsigned int oldPersist = vm->persistent;
J
Jiri Denemark 已提交
5608 5609 5610 5611 5612 5613 5614
    virObjectEventPtr event;
    int ret = -1;

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

    vm->persistent = 1;
5615 5616
    oldDef = vm->newDef;
    vm->newDef = qemuMigrationCookieGetPersistent(mig);
J
Jiri Denemark 已提交
5617

5618 5619
    if (!(vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm)))
        goto error;
J
Jiri Denemark 已提交
5620

5621 5622
    if (virDomainSaveConfig(cfg->configDir, vmdef) < 0 && !ignoreSaveError)
        goto error;
J
Jiri Denemark 已提交
5623 5624 5625

    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_DEFINED,
5626 5627 5628
                                              oldPersist ?
                                              VIR_DOMAIN_EVENT_DEFINED_UPDATED :
                                              VIR_DOMAIN_EVENT_DEFINED_ADDED);
5629
    qemuDomainEventQueue(driver, event);
J
Jiri Denemark 已提交
5630 5631 5632 5633

    ret = 0;

 cleanup:
5634
    virDomainDefFree(oldDef);
J
Jiri Denemark 已提交
5635 5636 5637
    virObjectUnref(caps);
    virObjectUnref(cfg);
    return ret;
5638 5639 5640 5641 5642 5643 5644

 error:
    virDomainDefFree(vm->newDef);
    vm->persistent = oldPersist;
    vm->newDef = oldDef;
    oldDef = NULL;
    goto cleanup;
J
Jiri Denemark 已提交
5645 5646 5647
}


5648
virDomainPtr
5649
qemuMigrationFinish(virQEMUDriverPtr driver,
5650 5651
                    virConnectPtr dconn,
                    virDomainObjPtr vm,
5652 5653 5654 5655
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
5656
                    unsigned long flags,
5657 5658
                    int retcode,
                    bool v3proto)
5659 5660
{
    virDomainPtr dom = NULL;
5661
    qemuMigrationCookiePtr mig = NULL;
5662
    virErrorPtr orig_err = NULL;
5663
    int cookie_flags = 0;
J
Jiri Denemark 已提交
5664
    qemuDomainObjPrivatePtr priv = vm->privateData;
5665
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5666
    unsigned short port;
5667

5668
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
5669
              "cookieout=%p, cookieoutlen=%p, flags=%lx, retcode=%d",
5670 5671
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
5672

5673 5674 5675
    port = priv->migrationPort;
    priv->migrationPort = 0;

5676 5677
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN)) {
        qemuMigrationErrorReport(driver, vm->def->name);
5678
        goto cleanup;
5679
    }
5680

5681 5682 5683
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
5684

5685
    qemuDomainCleanupRemove(vm, qemuMigrationPrepareCleanup);
5686
    VIR_FREE(priv->job.completed);
5687

5688
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
5689 5690
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
5691 5692 5693 5694 5695
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookie_flags |= QEMU_MIGRATION_COOKIE_PERSISTENT;

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein,
                                       cookieinlen, cookie_flags)))
5696
        goto endjob;
5697 5698 5699 5700

    /* Did the migration go as planned?  If yes, return the domain
     * object, but if no, clean up the empty qemu process.
     */
5701 5702
    if (flags & VIR_MIGRATE_OFFLINE) {
        if (retcode != 0 ||
5703
            qemuMigrationPersist(driver, vm, mig, false) < 0)
5704 5705 5706 5707
            goto endjob;

        dom = virGetDomain(dconn, vm->def->name, vm->def->uuid);
    } else if (retcode == 0) {
5708 5709 5710 5711
        unsigned long long timeReceived = 0;

        ignore_value(virTimeMillisNow(&timeReceived));

5712
        if (!virDomainObjIsActive(vm)) {
5713 5714
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("guest unexpectedly quit"));
5715
            qemuMigrationErrorReport(driver, vm->def->name);
5716
            goto endjob;
5717 5718
        }

5719
        if (qemuMigrationVPAssociatePortProfiles(vm->def) < 0)
5720
            goto endjob;
5721

5722 5723
        if (mig->network && qemuDomainMigrateOPDRelocate(driver, vm, mig) < 0)
            VIR_WARN("unable to provide network data for relocation");
5724

5725 5726
        if (qemuMigrationStopNBDServer(driver, vm, mig) < 0)
            goto endjob;
5727

5728
        if (flags & VIR_MIGRATE_PERSIST_DEST) {
5729
            if (qemuMigrationPersist(driver, vm, mig, !v3proto) < 0) {
5730 5731 5732 5733 5734 5735
                /* 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.
5736 5737 5738
                 * Pretend success and hope that this is a rare situation and
                 * management tools are smart.
                 *
5739 5740
                 * However, in v3 protocol, the source VM is still available
                 * to restart during confirm() step, so we kill it off now.
5741
                 */
5742 5743
                if (v3proto)
                    goto endjob;
5744 5745 5746
            }
        }

5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767
        /* We need to wait for QEMU to process all data sent by the source
         * before starting guest CPUs.
         */
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT)) {
            int rv;
            VIR_DEBUG("Waiting for migration to complete");
            while ((rv = qemuMigrationCompleted(driver, vm,
                                                QEMU_ASYNC_JOB_MIGRATION_IN,
                                                NULL, 0)) != 1) {
                if (rv < 0 || virDomainObjWait(vm) < 0) {
                    /* There's not much we can do for v2 protocol since the
                     * original domain on the source host is already gone.
                     */
                    if (v3proto)
                        goto endjob;
                    else
                        break;
                }
            }
        }

5768
        if (!(flags & VIR_MIGRATE_PAUSED)) {
5769 5770 5771 5772
            /* 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
5773
            if (qemuProcessStartCPUs(driver, vm, dconn,
5774 5775
                                     VIR_DOMAIN_RUNNING_MIGRATED,
                                     QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5776
                if (virGetLastError() == NULL)
5777 5778
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   "%s", _("resume operation failed"));
5779 5780 5781 5782 5783
                /* Need to save the current error, in case shutting
                 * down the process overwrites it
                 */
                orig_err = virSaveLastError();

5784 5785 5786 5787 5788 5789 5790 5791
                /*
                 * 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
                 */
5792 5793
                if (v3proto)
                    goto endjob;
5794
            }
5795 5796
        }

5797 5798 5799 5800 5801 5802 5803 5804 5805 5806
        if (mig->jobInfo) {
            qemuDomainJobInfoPtr jobInfo = mig->jobInfo;
            priv->job.completed = jobInfo;
            mig->jobInfo = NULL;

            if (jobInfo->sent && timeReceived) {
                jobInfo->timeDelta = timeReceived - jobInfo->sent;
                jobInfo->received = timeReceived;
                jobInfo->timeDeltaSet = true;
            }
5807 5808
            qemuDomainJobInfoUpdateTime(priv->job.completed);
            qemuDomainJobInfoUpdateDowntime(priv->job.completed);
5809 5810
        }

5811
        dom = virGetDomain(dconn, vm->def->name, vm->def->uuid);
5812

5813 5814 5815 5816
        qemuDomainEventQueue(driver,
                             virDomainEventLifecycleNewFromObj(vm,
                                    VIR_DOMAIN_EVENT_RESUMED,
                                    VIR_DOMAIN_EVENT_RESUMED_MIGRATED));
5817 5818 5819
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
            virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                                 VIR_DOMAIN_PAUSED_USER);
5820 5821 5822 5823
            qemuDomainEventQueue(driver,
                                 virDomainEventLifecycleNewFromObj(vm,
                                        VIR_DOMAIN_EVENT_SUSPENDED,
                                        VIR_DOMAIN_EVENT_SUSPENDED_PAUSED));
5824
        }
L
liguang 已提交
5825 5826

        if (virDomainObjIsActive(vm) &&
5827
            virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
5828
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
5829 5830 5831

        /* Guest is successfully running, so cancel previous auto destroy */
        qemuProcessAutoDestroyRemove(driver, vm);
5832
    } else {
5833 5834 5835 5836 5837 5838 5839 5840
        qemuDomainJobInfo info;

        /* Check for a possible error on the monitor in case Finish was called
         * earlier than monitor EOF handler got a chance to process the error
         */
        qemuMigrationFetchJobStatus(driver, vm,
                                    QEMU_ASYNC_JOB_MIGRATION_IN,
                                    &info);
5841
    }
5842

5843
 endjob:
5844
    if (!dom &&
5845 5846
        !(flags & VIR_MIGRATE_OFFLINE) &&
        virDomainObjIsActive(vm)) {
5847 5848
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
5849
        virDomainAuditStop(vm, "failed");
5850 5851 5852 5853
        qemuDomainEventQueue(driver,
                             virDomainEventLifecycleNewFromObj(vm,
                                    VIR_DOMAIN_EVENT_STOPPED,
                                    VIR_DOMAIN_EVENT_STOPPED_FAILED));
5854 5855
    }

5856 5857
    if (dom &&
        qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen,
5858
                                QEMU_MIGRATION_COOKIE_STATS) < 0)
5859 5860
        VIR_WARN("Unable to encode migration cookie");

5861
    qemuMigrationJobFinish(driver, vm);
5862
    if (!virDomainObjIsActive(vm))
5863
        qemuDomainRemoveInactive(driver, vm);
5864

5865
 cleanup:
5866
    virPortAllocatorRelease(driver->migrationPorts, port);
5867 5868 5869
    if (priv->mon)
        qemuMonitorSetDomainLog(priv->mon, -1);
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5870
    virDomainObjEndAPI(&vm);
5871 5872 5873 5874
    if (mig) {
        virDomainDefFree(qemuMigrationCookieGetPersistent(mig));
        qemuMigrationCookieFree(mig);
    }
5875 5876 5877 5878
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5879
    virObjectUnref(cfg);
5880 5881 5882 5883 5884 5885

    /* Set a special error if Finish is expected to return NULL as a result of
     * successful call with retcode != 0
     */
    if (retcode != 0 && !dom && !virGetLastError())
        virReportError(VIR_ERR_MIGRATE_FINISH_OK, NULL);
5886 5887
    return dom;
}
5888

5889

5890
/* Helper function called while vm is active.  */
5891
int
5892
qemuMigrationToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
5893 5894
                    int fd, off_t offset, const char *path,
                    const char *compressor,
E
Eric Blake 已提交
5895
                    bool bypassSecurityDriver,
5896
                    qemuDomainAsyncJob asyncJob)
5897 5898 5899
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5900
    int ret = -1;
5901
    bool restoreLabel = false;
5902 5903
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5904
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5905
    char *errbuf = NULL;
5906
    virErrorPtr orig_err = NULL;
5907 5908 5909 5910 5911

    /* 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,
5912 5913
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5914 5915
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5916
    }
5917

5918 5919 5920 5921 5922 5923 5924
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5925
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5926
        (!compressor || pipe(pipeFD) == 0)) {
5927
        /* All right! We can use fd migration, which means that qemu
5928 5929 5930
         * 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.  */
5931
        if (virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
5932
                                              compressor ? pipeFD[1] : fd) < 0)
5933
            goto cleanup;
5934 5935 5936
        bypassSecurityDriver = true;
    } else {
        /* Phooey - we have to fall back on exec migration, where qemu
E
Eric Blake 已提交
5937 5938
         * has to popen() the file by name, and block devices have to be
         * given cgroup ACL permission.  We might also stumble on
5939 5940
         * a race present in some qemu versions where it does a wait()
         * that botches pclose.  */
5941 5942
        if (virCgroupHasController(priv->cgroup,
                                   VIR_CGROUP_CONTROLLER_DEVICES)) {
5943 5944 5945 5946
            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
5947
                /* path was not a device, no further need for cgroup */
5948
            } else if (rv < 0) {
5949 5950
                goto cleanup;
            }
5951
        }
5952 5953
        if ((!bypassSecurityDriver) &&
            virSecurityManagerSetSavedStateLabel(driver->securityManager,
5954
                                                 vm->def, path) < 0)
5955
            goto cleanup;
5956
        restoreLabel = true;
5957 5958
    }

5959
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5960 5961
        goto cleanup;

5962 5963 5964
    if (!compressor) {
        const char *args[] = { "cat", NULL };

5965
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977 5978 5979 5980 5981
            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
        };
5982 5983 5984 5985
        if (pipeFD[0] != -1) {
            cmd = virCommandNewArgs(args);
            virCommandSetInputFD(cmd, pipeFD[0]);
            virCommandSetOutputFD(cmd, &fd);
5986 5987
            virCommandSetErrorBuffer(cmd, &errbuf);
            virCommandDoAsyncIO(cmd);
5988 5989 5990
            if (virSetCloseExec(pipeFD[1]) < 0) {
                virReportSystemError(errno, "%s",
                                     _("Unable to set cloexec flag"));
5991
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5992 5993 5994
                goto cleanup;
            }
            if (virCommandRunAsync(cmd, NULL) < 0) {
5995
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5996 5997 5998 5999 6000 6001 6002
                goto cleanup;
            }
            rc = qemuMonitorMigrateToFd(priv->mon,
                                        QEMU_MONITOR_MIGRATE_BACKGROUND,
                                        pipeFD[1]);
            if (VIR_CLOSE(pipeFD[0]) < 0 ||
                VIR_CLOSE(pipeFD[1]) < 0)
6003
                VIR_WARN("failed to close intermediate pipe");
6004 6005 6006 6007 6008
        } else {
            rc = qemuMonitorMigrateToFile(priv->mon,
                                          QEMU_MONITOR_MIGRATE_BACKGROUND,
                                          args, path, offset);
        }
6009
    }
6010
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
6011
        goto cleanup;
6012 6013 6014
    if (rc < 0)
        goto cleanup;

6015 6016
    rc = qemuMigrationWaitForCompletion(driver, vm, asyncJob,
                                        NULL, false, false);
6017

6018 6019 6020 6021
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
6022 6023
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
6024
                qemuMonitorMigrateCancel(priv->mon);
6025
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
6026 6027
            }
        }
6028
        goto cleanup;
6029
    }
6030

6031 6032 6033
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

6034 6035
    ret = 0;

6036
 cleanup:
6037 6038 6039
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

6040
    /* Restore max migration bandwidth */
6041 6042
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
6043 6044
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
6045
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
6046 6047
    }

6048 6049
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
6050 6051 6052 6053 6054
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
6055 6056
    if (restoreLabel && (!bypassSecurityDriver) &&
        virSecurityManagerRestoreSavedStateLabel(driver->securityManager,
6057
                                                 vm->def, path) < 0)
6058 6059
        VIR_WARN("failed to restore save state label on %s", path);

6060 6061
    if (virCgroupHasController(priv->cgroup,
                               VIR_CGROUP_CONTROLLER_DEVICES)) {
6062 6063 6064
        int rv = virCgroupDenyDevicePath(priv->cgroup, path,
                                         VIR_CGROUP_DEVICE_RWM);
        virDomainAuditCgroupPath(vm, priv->cgroup, "deny", path, "rwm", rv == 0);
6065
    }
6066 6067 6068 6069 6070 6071

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

6072 6073
    return ret;
}
6074

6075 6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088 6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121 6122 6123 6124 6125 6126 6127

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

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

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

    qemuDomainObjEnterMonitor(driver, vm);

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

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

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

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

        if (!diskPriv->migrating)
            continue;

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

    if (qemuMigrationCancelDriveMirror(driver, vm, false,
6128
                                       QEMU_ASYNC_JOB_NONE, NULL) < 0)
6129 6130 6131 6132 6133 6134 6135 6136 6137 6138 6139 6140 6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152
        goto endsyncjob;

    ret = 0;

 cleanup:
    virHashFree(blockJobs);
    return ret;

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

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


6153
int
6154
qemuMigrationJobStart(virQEMUDriverPtr driver,
6155
                      virDomainObjPtr vm,
6156
                      qemuDomainAsyncJob job)
6157 6158 6159
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

6160
    if (qemuDomainObjBeginAsyncJob(driver, vm, job) < 0)
6161 6162
        return -1;

6163
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
6164
        qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
6165
    } else {
6166 6167 6168
        qemuDomainObjSetAsyncJobMask(vm, (QEMU_JOB_DEFAULT_MASK |
                                          JOB_MASK(QEMU_JOB_SUSPEND) |
                                          JOB_MASK(QEMU_JOB_MIGRATION_OP)));
6169
    }
6170

J
Jiri Denemark 已提交
6171
    priv->job.current->type = VIR_DOMAIN_JOB_UNBOUNDED;
6172 6173 6174 6175 6176

    return 0;
}

void
6177
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
6178
                         virDomainObjPtr vm,
6179
                         qemuMigrationJobPhase phase)
6180 6181 6182 6183 6184 6185 6186 6187 6188 6189 6190 6191 6192 6193
{
    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
6194
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
6195
                           virDomainObjPtr vm,
6196
                           qemuMigrationJobPhase phase)
6197 6198 6199 6200
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

6201
void
6202 6203
qemuMigrationJobContinue(virDomainObjPtr vm)
{
6204
    qemuDomainObjReleaseAsyncJob(vm);
6205 6206 6207 6208
}

bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
6209
                         qemuDomainAsyncJob job)
6210 6211 6212 6213 6214 6215 6216 6217 6218 6219 6220
{
    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");

6221
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
6222 6223 6224 6225 6226
        return false;
    }
    return true;
}

6227
void
6228
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
6229
{
6230
    qemuDomainObjEndAsyncJob(driver, vm);
6231
}
6232 6233 6234 6235 6236 6237 6238 6239 6240 6241 6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256 6257 6258 6259 6260 6261 6262 6263 6264 6265 6266 6267 6268 6269 6270 6271 6272 6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 6285


static void
qemuMigrationErrorFree(void *data,
                       const void *name ATTRIBUTE_UNUSED)
{
    virErrorPtr err = data;
    virFreeError(err);
}

int
qemuMigrationErrorInit(virQEMUDriverPtr driver)
{
    driver->migrationErrors = virHashAtomicNew(64, qemuMigrationErrorFree);
    if (driver->migrationErrors)
        return 0;
    else
        return -1;
}

/**
 * This function consumes @err; the caller should consider the @err pointer
 * invalid after calling this function.
 */
void
qemuMigrationErrorSave(virQEMUDriverPtr driver,
                       const char *name,
                       virErrorPtr err)
{
    if (!err)
        return;

    VIR_DEBUG("Saving incoming migration error for domain %s: %s",
              name, err->message);
    if (virHashAtomicUpdate(driver->migrationErrors, name, err) < 0) {
        VIR_WARN("Failed to save migration error for domain '%s'", name);
        virFreeError(err);
    }
}

void
qemuMigrationErrorReport(virQEMUDriverPtr driver,
                         const char *name)
{
    virErrorPtr err;

    if (!(err = virHashAtomicSteal(driver->migrationErrors, name)))
        return;

    VIR_DEBUG("Restoring saved incoming migration error for domain %s: %s",
              name, err->message);
    virSetError(err);
    virFreeError(err);
}