You need to sign in or sign up before continuing.
qemu_migration.c 204.1 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
static int
qemuMigrationCookieAddNetwork(qemuMigrationCookiePtr mig,
530
                              virQEMUDriverPtr driver,
531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548
                              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;
}

549

550 551
static int
qemuMigrationCookieAddNBD(qemuMigrationCookiePtr mig,
552
                          virQEMUDriverPtr driver,
553 554 555
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
556 557
    virHashTablePtr stats = NULL;
    size_t i;
558
    int ret = -1, rc;
559 560 561

    /* It is not a bug if there already is a NBD data */
    if (!mig->nbd &&
562
        VIR_ALLOC(mig->nbd) < 0)
563 564
        return -1;

565 566 567 568 569 570 571 572 573 574 575 576 577
    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;

578 579 580
            if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                               priv->job.asyncJob) < 0)
                goto cleanup;
581 582
            rc = qemuMonitorBlockStatsUpdateCapacity(priv->mon, stats, false);
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
583
                goto cleanup;
584
            if (rc < 0)
585 586 587 588 589 590 591 592 593 594 595 596 597 598
                goto cleanup;
        }

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

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

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

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


609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627
static int
qemuMigrationCookieAddStatistics(qemuMigrationCookiePtr mig,
                                 virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

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

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

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

    return 0;
}


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


648 649 650 651
static void
qemuMigrationCookieNetworkXMLFormat(virBufferPtr buf,
                                    qemuMigrationCookieNetworkPtr optr)
{
652
    size_t i;
653 654 655 656 657 658
    bool empty = true;

    for (i = 0; i < optr->nnets; i++) {
        /* If optr->net[i].vporttype is not set, there is nothing to transfer */
        if (optr->net[i].vporttype != VIR_NETDEV_VPORT_PROFILE_NONE) {
            if (empty) {
659 660
                virBufferAddLit(buf, "<network>\n");
                virBufferAdjustIndent(buf, 2);
661 662
                empty = false;
            }
663
            virBufferAsprintf(buf, "<interface index='%zu' vporttype='%s'",
664 665 666
                              i, virNetDevVPortTypeToString(optr->net[i].vporttype));
            if (optr->net[i].portdata) {
                virBufferAddLit(buf, ">\n");
667 668
                virBufferAdjustIndent(buf, 2);
                virBufferEscapeString(buf, "<portdata>%s</portdata>\n",
669
                                      optr->net[i].portdata);
670 671
                virBufferAdjustIndent(buf, -2);
                virBufferAddLit(buf, "</interface>\n");
672 673 674 675 676
            } else {
                virBufferAddLit(buf, "/>\n");
            }
        }
    }
677 678 679 680
    if (!empty) {
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</network>\n");
    }
681 682 683
}


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

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

693 694
    virBufferAsprintf(buf, "<started>%llu</started>\n", jobInfo->started);
    virBufferAsprintf(buf, "<stopped>%llu</stopped>\n", jobInfo->stopped);
695 696 697
    virBufferAsprintf(buf, "<sent>%llu</sent>\n", jobInfo->sent);
    if (jobInfo->timeDeltaSet)
        virBufferAsprintf(buf, "<delta>%lld</delta>\n", jobInfo->timeDelta);
698

699 700 701 702 703 704 705 706 707 708
    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$s>\n",
                          VIR_DOMAIN_JOB_DOWNTIME,
                          status->downtime);
709 710 711 712
    if (status->setup_time_set)
        virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                          VIR_DOMAIN_JOB_SETUP_TIME,
                          status->setup_time);
713 714 715 716 717 718 719 720 721 722

    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$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_REMAINING,
                      status->ram_remaining);
723 724 725
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_MEMORY_BPS,
                      status->ram_bps);
726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747

    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$s>\n",
                      VIR_DOMAIN_JOB_DISK_REMAINING,
                      status->disk_remaining);
748 749 750
    virBufferAsprintf(buf, "<%1$s>%2$llu</%1$s>\n",
                      VIR_DOMAIN_JOB_DISK_BPS,
                      status->disk_bps);
751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774

    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");
}


775
static int
776
qemuMigrationCookieXMLFormat(virQEMUDriverPtr driver,
777
                             virBufferPtr buf,
778
                             qemuMigrationCookiePtr mig)
779 780 781
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
    char hostuuidstr[VIR_UUID_STRING_BUFLEN];
782
    size_t i;
783 784

    virUUIDFormat(mig->uuid, uuidstr);
785
    virUUIDFormat(mig->localHostuuid, hostuuidstr);
786

787
    virBufferAddLit(buf, "<qemu-migration>\n");
788 789 790 791 792
    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);
793

794
    for (i = 0; i < QEMU_MIGRATION_COOKIE_FLAG_LAST; i++) {
795
        if (mig->flagsMandatory & (1 << i))
796
            virBufferAsprintf(buf, "<feature name='%s'/>\n",
797 798 799
                              qemuMigrationCookieFlagTypeToString(i));
    }

800 801
    if ((mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        mig->graphics)
802 803
        qemuMigrationCookieGraphicsXMLFormat(buf, mig->graphics);

804 805
    if ((mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        mig->lockState) {
806
        virBufferAsprintf(buf, "<lockstate driver='%s'>\n",
807
                          mig->lockDriver);
808 809
        virBufferAdjustIndent(buf, 2);
        virBufferAsprintf(buf, "<leases>%s</leases>\n",
810
                          mig->lockState);
811 812
        virBufferAdjustIndent(buf, -2);
        virBufferAddLit(buf, "</lockstate>\n");
813 814
    }

815 816
    if ((mig->flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        mig->persistent) {
817 818 819
        if (qemuDomainDefFormatBuf(driver,
                                   mig->persistent,
                                   VIR_DOMAIN_XML_INACTIVE |
820 821
                                   VIR_DOMAIN_XML_SECURE |
                                   VIR_DOMAIN_XML_MIGRATABLE,
822
                                   buf) < 0)
823
            return -1;
824 825
    }

826 827 828
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NETWORK) && mig->network)
        qemuMigrationCookieNetworkXMLFormat(buf, mig->network);

829
    if ((mig->flags & QEMU_MIGRATION_COOKIE_NBD) && mig->nbd) {
830
        virBufferAddLit(buf, "<nbd");
831 832
        if (mig->nbd->port)
            virBufferAsprintf(buf, " port='%d'", mig->nbd->port);
833 834 835 836 837 838 839 840 841 842 843 844 845 846
        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");
        }
847 848
    }

849 850 851
    if (mig->flags & QEMU_MIGRATION_COOKIE_STATS && mig->jobInfo)
        qemuMigrationCookieStatisticsXMLFormat(buf, mig->jobInfo);

852
    virBufferAdjustIndent(buf, -2);
853
    virBufferAddLit(buf, "</qemu-migration>\n");
854
    return 0;
855 856 857
}


858
static char *qemuMigrationCookieXMLFormatStr(virQEMUDriverPtr driver,
859
                                             qemuMigrationCookiePtr mig)
860 861 862
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;

863
    if (qemuMigrationCookieXMLFormat(driver, &buf, mig) < 0) {
864 865 866
        virBufferFreeAndReset(&buf);
        return NULL;
    }
867

868
    if (virBufferCheckError(&buf) < 0)
869 870 871 872 873 874
        return NULL;

    return virBufferContentAndReset(&buf);
}


875 876 877 878 879 880 881
static qemuMigrationCookieGraphicsPtr
qemuMigrationCookieGraphicsXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieGraphicsPtr grap;
    char *tmp;

    if (VIR_ALLOC(grap) < 0)
882
        goto error;
883 884

    if (!(tmp = virXPathString("string(./graphics/@type)", ctxt))) {
885 886
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing type attribute in migration data"));
887 888 889
        goto error;
    }
    if ((grap->type = virDomainGraphicsTypeFromString(tmp)) < 0) {
890 891
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unknown graphics type %s"), tmp);
892 893 894
        VIR_FREE(tmp);
        goto error;
    }
E
Eric Blake 已提交
895
    VIR_FREE(tmp);
896
    if (virXPathInt("string(./graphics/@port)", ctxt, &grap->port) < 0) {
897 898
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing port attribute in migration data"));
899 900 901 902
        goto error;
    }
    if (grap->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        if (virXPathInt("string(./graphics/@tlsPort)", ctxt, &grap->tlsPort) < 0) {
903 904
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing tlsPort attribute in migration data"));
905 906 907 908
            goto error;
        }
    }
    if (!(grap->listen = virXPathString("string(./graphics/@listen)", ctxt))) {
909 910
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing listen attribute in migration data"));
911 912 913
        goto error;
    }
    /* Optional */
914
    grap->tlsSubject = virXPathString("string(./graphics/cert[@info='subject']/@value)", ctxt);
915 916 917

    return grap;

918
 error:
919 920 921 922 923
    qemuMigrationCookieGraphicsFree(grap);
    return NULL;
}


924 925 926 927
static qemuMigrationCookieNetworkPtr
qemuMigrationCookieNetworkXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieNetworkPtr optr;
928
    size_t i;
929 930 931 932 933 934
    int n;
    xmlNodePtr *interfaces = NULL;
    char *vporttype;
    xmlNodePtr save_ctxt = ctxt->node;

    if (VIR_ALLOC(optr) < 0)
935
        goto error;
936 937 938 939 940 941 942 943

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

    optr->nnets = n;
944
    if (VIR_ALLOC_N(optr->net, optr->nnets) < 0)
945
        goto error;
946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961

    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);

962
 cleanup:
963 964 965
    ctxt->node = save_ctxt;
    return optr;

966
 error:
967 968 969 970 971 972 973
    VIR_FREE(interfaces);
    qemuMigrationCookieNetworkFree(optr);
    optr = NULL;
    goto cleanup;
}


974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
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;
}


1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
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;

1054 1055
    virXPathULongLong("string(./started[1])", ctxt, &jobInfo->started);
    virXPathULongLong("string(./stopped[1])", ctxt, &jobInfo->stopped);
1056 1057 1058
    virXPathULongLong("string(./sent[1])", ctxt, &jobInfo->sent);
    if (virXPathLongLong("string(./delta[1])", ctxt, &jobInfo->timeDelta) == 0)
        jobInfo->timeDeltaSet = true;
1059

1060 1061 1062 1063
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_TIME_ELAPSED "[1])",
                      ctxt, &jobInfo->timeElapsed);
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_TIME_REMAINING "[1])",
                      ctxt, &jobInfo->timeRemaining);
1064

1065 1066 1067
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_DOWNTIME "[1])",
                          ctxt, &status->downtime) == 0)
        status->downtime_set = true;
1068 1069 1070
    if (virXPathULongLong("string(./" VIR_DOMAIN_JOB_SETUP_TIME "[1])",
                          ctxt, &status->setup_time) == 0)
        status->setup_time_set = true;
1071 1072 1073 1074 1075 1076 1077

    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);
1078 1079
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_MEMORY_BPS "[1])",
                      ctxt, &status->ram_bps);
1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094

    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);
1095 1096
    virXPathULongLong("string(./" VIR_DOMAIN_JOB_DISK_BPS "[1])",
                      ctxt, &status->disk_bps);
1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115

    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;
}


1116 1117
static int
qemuMigrationCookieXMLParse(qemuMigrationCookiePtr mig,
1118
                            virQEMUDriverPtr driver,
1119
                            xmlDocPtr doc,
1120
                            xmlXPathContextPtr ctxt,
E
Eric Blake 已提交
1121
                            unsigned int flags)
1122 1123
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
1124
    char *tmp = NULL;
1125
    xmlNodePtr *nodes = NULL;
1126 1127
    size_t i;
    int n;
1128 1129 1130 1131
    virCapsPtr caps = NULL;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto error;
1132 1133 1134 1135 1136 1137 1138 1139

    /* 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))) {
1140 1141
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing name element in migration data"));
1142 1143 1144
        goto error;
    }
    if (STRNEQ(tmp, mig->name)) {
1145 1146 1147
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected name %s vs %s"),
                       tmp, mig->name);
1148 1149 1150 1151 1152 1153 1154
        goto error;
    }
    VIR_FREE(tmp);

    /* Extract domain uuid */
    tmp = virXPathString("string(./uuid[1])", ctxt);
    if (!tmp) {
1155 1156
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing uuid element in migration data"));
1157 1158 1159 1160
        goto error;
    }
    virUUIDFormat(mig->uuid, uuidstr);
    if (STRNEQ(tmp, uuidstr)) {
1161 1162 1163
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Incoming cookie data had unexpected UUID %s vs %s"),
                       tmp, uuidstr);
1164
        goto error;
1165 1166 1167 1168
    }
    VIR_FREE(tmp);

    /* Check & forbid "localhost" migration */
1169
    if (!(mig->remoteHostname = virXPathString("string(./hostname[1])", ctxt))) {
1170 1171
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostname element in migration data"));
1172 1173
        goto error;
    }
1174
    if (STREQ(mig->remoteHostname, mig->localHostname)) {
1175 1176 1177
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Attempt to migrate guest to the same host %s"),
                       mig->remoteHostname);
1178 1179 1180 1181
        goto error;
    }

    if (!(tmp = virXPathString("string(./hostuuid[1])", ctxt))) {
1182 1183 1184
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("missing hostuuid element in migration data"));
        goto error;
1185
    }
1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
    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);
1198

1199 1200
    /* Check to ensure all mandatory features from XML are also
     * present in 'flags' */
1201
    if ((n = virXPathNodeSet("./feature", ctxt, &nodes)) < 0)
1202 1203
        goto error;

1204
    for (i = 0; i < n; i++) {
1205 1206 1207
        int val;
        char *str = virXMLPropString(nodes[i], "name");
        if (!str) {
1208 1209
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("missing feature name"));
1210 1211 1212 1213
            goto error;
        }

        if ((val = qemuMigrationCookieFlagTypeFromString(str)) < 0) {
1214 1215 1216
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unknown migration cookie feature %s"),
                           str);
1217 1218 1219 1220 1221
            VIR_FREE(str);
            goto error;
        }

        if ((flags & (1 << val)) == 0) {
1222 1223 1224
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unsupported migration cookie feature %s"),
                           str);
1225 1226 1227 1228 1229 1230
            VIR_FREE(str);
        }
        VIR_FREE(str);
    }
    VIR_FREE(nodes);

1231 1232 1233 1234 1235
    if ((flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        virXPathBoolean("count(./graphics) > 0", ctxt) &&
        (!(mig->graphics = qemuMigrationCookieGraphicsXMLParse(ctxt))))
        goto error;

1236 1237 1238 1239
    if ((flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        virXPathBoolean("count(./lockstate) > 0", ctxt)) {
        mig->lockDriver = virXPathString("string(./lockstate[1]/@driver)", ctxt);
        if (!mig->lockDriver) {
1240 1241
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Missing lock driver name in migration cookie"));
1242 1243 1244 1245 1246 1247 1248
            goto error;
        }
        mig->lockState = virXPathString("string(./lockstate[1]/leases[1])", ctxt);
        if (mig->lockState && STREQ(mig->lockState, ""))
            VIR_FREE(mig->lockState);
    }

1249 1250 1251
    if ((flags & QEMU_MIGRATION_COOKIE_PERSISTENT) &&
        virXPathBoolean("count(./domain) > 0", ctxt)) {
        if ((n = virXPathNodeSet("./domain", ctxt, &nodes)) > 1) {
1252 1253 1254 1255
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Too many domain elements in "
                             "migration cookie: %d"),
                           n);
1256 1257
            goto error;
        }
1258 1259
        mig->persistent = virDomainDefParseNode(doc, nodes[0],
                                                caps, driver->xmlopt,
1260
                                                VIR_DOMAIN_DEF_PARSE_INACTIVE);
1261 1262 1263 1264 1265 1266 1267 1268
        if (!mig->persistent) {
            /* virDomainDefParseNode already reported
             * an error for us */
            goto error;
        }
        VIR_FREE(nodes);
    }

1269 1270 1271 1272 1273
    if ((flags & QEMU_MIGRATION_COOKIE_NETWORK) &&
        virXPathBoolean("count(./network) > 0", ctxt) &&
        (!(mig->network = qemuMigrationCookieNetworkXMLParse(ctxt))))
        goto error;

1274
    if (flags & QEMU_MIGRATION_COOKIE_NBD &&
1275 1276 1277
        virXPathBoolean("boolean(./nbd)", ctxt) &&
        (!(mig->nbd = qemuMigrationCookieNBDXMLParse(ctxt))))
        goto error;
1278

1279 1280 1281 1282 1283
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        virXPathBoolean("boolean(./statistics)", ctxt) &&
        (!(mig->jobInfo = qemuMigrationCookieStatisticsXMLParse(ctxt))))
        goto error;

1284
    virObjectUnref(caps);
1285 1286
    return 0;

1287
 error:
1288
    VIR_FREE(tmp);
1289
    VIR_FREE(nodes);
1290
    virObjectUnref(caps);
1291 1292 1293 1294 1295 1296
    return -1;
}


static int
qemuMigrationCookieXMLParseStr(qemuMigrationCookiePtr mig,
1297
                               virQEMUDriverPtr driver,
1298
                               const char *xml,
E
Eric Blake 已提交
1299
                               unsigned int flags)
1300 1301 1302
{
    xmlDocPtr doc = NULL;
    xmlXPathContextPtr ctxt = NULL;
1303
    int ret = -1;
1304 1305 1306

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

1307
    if (!(doc = virXMLParseStringCtxt(xml, _("(qemu_migration_cookie)"), &ctxt)))
1308 1309
        goto cleanup;

1310
    ret = qemuMigrationCookieXMLParse(mig, driver, doc, ctxt, flags);
1311

1312
 cleanup:
1313 1314 1315 1316 1317 1318 1319 1320 1321
    xmlXPathFreeContext(ctxt);
    xmlFreeDoc(doc);

    return ret;
}


static int
qemuMigrationBakeCookie(qemuMigrationCookiePtr mig,
1322
                        virQEMUDriverPtr driver,
1323
                        virDomainObjPtr dom,
1324 1325
                        char **cookieout,
                        int *cookieoutlen,
E
Eric Blake 已提交
1326
                        unsigned int flags)
1327
{
1328 1329
    if (!cookieout || !cookieoutlen)
        return 0;
1330 1331 1332

    *cookieoutlen = 0;

1333 1334 1335 1336
    if (flags & QEMU_MIGRATION_COOKIE_GRAPHICS &&
        qemuMigrationCookieAddGraphics(mig, driver, dom) < 0)
        return -1;

1337 1338 1339 1340
    if (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE &&
        qemuMigrationCookieAddLockstate(mig, driver, dom) < 0)
        return -1;

1341 1342 1343 1344
    if (flags & QEMU_MIGRATION_COOKIE_PERSISTENT &&
        qemuMigrationCookieAddPersistent(mig, dom) < 0)
        return -1;

1345 1346 1347 1348 1349
    if (flags & QEMU_MIGRATION_COOKIE_NETWORK &&
        qemuMigrationCookieAddNetwork(mig, driver, dom) < 0) {
        return -1;
    }

1350 1351 1352 1353
    if ((flags & QEMU_MIGRATION_COOKIE_NBD) &&
        qemuMigrationCookieAddNBD(mig, driver, dom) < 0)
        return -1;

1354 1355 1356 1357
    if (flags & QEMU_MIGRATION_COOKIE_STATS &&
        qemuMigrationCookieAddStatistics(mig, dom) < 0)
        return -1;

1358 1359 1360
    if (flags & QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG)
        mig->flagsMandatory |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

1361
    if (!(*cookieout = qemuMigrationCookieXMLFormatStr(driver, mig)))
1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372
        return -1;

    *cookieoutlen = strlen(*cookieout) + 1;

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

    return 0;
}


static qemuMigrationCookiePtr
1373
qemuMigrationEatCookie(virQEMUDriverPtr driver,
1374
                       virDomainObjPtr dom,
1375 1376
                       const char *cookiein,
                       int cookieinlen,
E
Eric Blake 已提交
1377
                       unsigned int flags)
1378 1379 1380 1381 1382 1383
{
    qemuMigrationCookiePtr mig = NULL;

    /* Parse & validate incoming cookie (if any) */
    if (cookiein && cookieinlen &&
        cookiein[cookieinlen-1] != '\0') {
1384 1385
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Migration cookie was not NULL terminated"));
1386 1387 1388 1389 1390 1391 1392 1393 1394 1395
        goto error;
    }

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

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

    if (cookiein && cookieinlen &&
        qemuMigrationCookieXMLParseStr(mig,
1396
                                       driver,
1397 1398 1399 1400
                                       cookiein,
                                       flags) < 0)
        goto error;

1401 1402 1403 1404 1405 1406 1407 1408
    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;
    }

1409 1410 1411
    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
        if (!mig->lockDriver) {
            if (virLockManagerPluginUsesState(driver->lockManager)) {
1412 1413 1414
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Missing %s lock state for migration cookie"),
                               virLockManagerPluginGetName(driver->lockManager));
1415 1416 1417 1418
                goto error;
            }
        } else if (STRNEQ(mig->lockDriver,
                          virLockManagerPluginGetName(driver->lockManager))) {
1419 1420 1421 1422
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Source host lock driver %s different from target %s"),
                           mig->lockDriver,
                           virLockManagerPluginGetName(driver->lockManager));
1423 1424 1425 1426
            goto error;
        }
    }

1427 1428
    return mig;

1429
 error:
1430 1431 1432
    qemuMigrationCookieFree(mig);
    return NULL;
}
1433

1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480
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;
}

1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534

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;

1535 1536 1537 1538 1539
    case VIR_STORAGE_TYPE_NETWORK:
        VIR_DEBUG("Skipping creation of network disk '%s'",
                  disk->dst);
        return 0;

1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
    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;
}

1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
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);
}

1609 1610 1611 1612 1613

static int
qemuMigrationPrecreateStorage(virConnectPtr conn,
                              virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
                              virDomainObjPtr vm,
1614 1615
                              qemuMigrationCookieNBDPtr nbd,
                              size_t nmigrate_disks,
1616 1617
                              const char **migrate_disks,
                              bool incremental)
1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628
{
    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;

1629
        VIR_DEBUG("Looking up disk target '%s' (capacity=%llu)",
1630 1631
                  nbd->disks[i].target, nbd->disks[i].capacity);

1632 1633
        if (!(disk = virDomainDiskByName(vm->def, nbd->disks[i].target,
                                         false))) {
1634 1635 1636 1637 1638 1639 1640 1641
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unable to find disk by target: %s"),
                           nbd->disks[i].target);
            goto cleanup;
        }

        diskSrcPath = virDomainDiskGetSource(disk);

1642 1643
        /* Skip disks we don't want to migrate and already existing disks. */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks) ||
1644 1645 1646 1647
            (diskSrcPath && virFileExists(diskSrcPath))) {
            continue;
        }

1648 1649 1650 1651 1652 1653 1654
        if (incremental) {
            virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
                           _("pre-creation of storage targets for incremental "
                             "storage migration is not supported"));
            goto cleanup;
        }

1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666
        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;
}


1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680
/**
 * 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
1681
                            virDomainObjPtr vm,
1682 1683 1684
                            const char *listenAddr,
                            size_t nmigrate_disks,
                            const char **migrate_disks)
1685 1686 1687 1688 1689 1690 1691 1692 1693 1694
{
    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];

1695 1696
        /* check whether disk should be migrated */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks))
1697 1698 1699 1700
            continue;

        VIR_FREE(diskAlias);
        if (virAsprintf(&diskAlias, "%s%s",
1701
                        QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0)
1702 1703 1704 1705 1706 1707 1708
            goto cleanup;

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

        if (!port &&
1709
            ((virPortAllocatorAcquire(driver->migrationPorts, &port) < 0) ||
1710
             (qemuMonitorNBDServerStart(priv->mon, listenAddr, port) < 0))) {
1711
            goto exit_monitor;
1712 1713
        }

1714 1715 1716
        if (qemuMonitorNBDServerAdd(priv->mon, diskAlias, true) < 0)
            goto exit_monitor;
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
1717 1718 1719 1720 1721 1722
            goto cleanup;
    }

    priv->nbdPort = port;
    ret = 0;

1723
 cleanup:
1724
    VIR_FREE(diskAlias);
1725
    if (ret < 0)
1726
        virPortAllocatorRelease(driver->migrationPorts, port);
1727
    return ret;
1728 1729 1730 1731

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;
1732 1733
}

1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760

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;
}


/**
1761
 * qemuMigrationDriveMirrorReady:
1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773
 * @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
1774
qemuMigrationDriveMirrorReady(virQEMUDriverPtr driver,
1775 1776 1777
                              virDomainObjPtr vm)
{
    size_t i;
1778 1779
    size_t notReady = 0;
    int status;
1780 1781 1782

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
1783
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1784

1785
        if (!diskPriv->migrating)
1786 1787
            continue;

1788 1789
        status = qemuBlockJobUpdate(driver, vm, disk);
        if (status == VIR_DOMAIN_BLOCK_JOB_FAILED) {
1790 1791 1792 1793 1794
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("migration of disk %s failed"),
                           disk->dst);
            return -1;
        }
1795 1796 1797

        if (disk->mirrorState != VIR_DOMAIN_DISK_MIRROR_STATE_READY)
            notReady++;
1798 1799
    }

1800 1801 1802 1803 1804 1805 1806
    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;
    }
1807 1808 1809
}


1810 1811 1812 1813
/*
 * 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.
1814
 *
1815 1816 1817 1818
 * 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.
1819 1820
 */
static int
1821
qemuMigrationDriveMirrorCancelled(virQEMUDriverPtr driver,
1822
                                  virDomainObjPtr vm,
1823
                                  bool check)
1824
{
1825 1826 1827 1828
    size_t i;
    size_t active = 0;
    int status;
    bool failed = false;
1829

1830 1831 1832
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1833

1834 1835
        if (!diskPriv->migrating)
            continue;
1836

1837 1838 1839 1840
        status = qemuBlockJobUpdate(driver, vm, disk);
        switch (status) {
        case VIR_DOMAIN_BLOCK_JOB_FAILED:
            if (check) {
1841
                virReportError(VIR_ERR_OPERATION_FAILED,
1842
                               _("migration of disk %s failed"),
1843
                               disk->dst);
1844
                failed = true;
1845
            }
1846 1847 1848 1849 1850 1851
            /* fallthrough */
        case VIR_DOMAIN_BLOCK_JOB_CANCELED:
        case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
            qemuBlockJobSyncEnd(driver, vm, disk);
            diskPriv->migrating = false;
            break;
1852

1853 1854
        default:
            active++;
1855
        }
1856
    }
1857

1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
    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;
1874
        }
1875
    }
1876
}
1877 1878


1879 1880 1881 1882 1883 1884 1885 1886 1887
/*
 * 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,
1888 1889
                                  bool failNoJob,
                                  qemuDomainAsyncJob asyncJob)
1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916
{
    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;

1917
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
1918 1919 1920 1921 1922 1923 1924 1925
        goto cleanup;

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

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

    ret = 0;
1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936

 cleanup:
    VIR_FREE(diskAlias);
    return ret;
}


/**
 * qemuMigrationCancelDriveMirror:
 * @driver: qemu driver
 * @vm: domain
1937
 * @check: if true report an error when some of the mirrors fails
1938 1939 1940 1941 1942 1943 1944 1945 1946
 *
 * 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,
1947
                               virDomainObjPtr vm,
1948
                               bool check,
1949 1950
                               qemuDomainAsyncJob asyncJob,
                               virConnectPtr dconn)
1951
{
1952
    virErrorPtr err = NULL;
1953
    int ret = -1;
1954
    size_t i;
1955 1956 1957 1958
    int rv;
    bool failed = false;

    VIR_DEBUG("Cancelling drive mirrors for domain %s", vm->def->name);
1959 1960 1961

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
1962
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1963

1964
        if (!diskPriv->migrating)
1965 1966
            continue;

1967 1968
        rv = qemuMigrationCancelOneDriveMirror(driver, vm, disk,
                                               check, asyncJob);
1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
        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) {
1981 1982 1983 1984 1985 1986 1987
        if (check && !failed &&
            dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            failed = true;
        }

1988 1989 1990 1991
        if (rv < 0) {
            failed = true;
            if (rv == -2)
                break;
1992
        }
1993

1994 1995 1996 1997 1998
        if (failed && !err)
            err = virSaveLastError();

        if (virDomainObjWait(vm) < 0)
            goto cleanup;
1999 2000
    }

2001 2002 2003
    ret = failed ? -1 : 0;

 cleanup:
2004 2005 2006 2007 2008
    if (err) {
        virSetError(err);
        virFreeError(err);
    }
    return ret;
2009 2010 2011
}


2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
/**
 * 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
2023 2024 2025 2026 2027
 * 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.
2028 2029 2030 2031 2032 2033 2034 2035 2036 2037
 *
 * Returns 0 on success (@migrate_flags updated),
 *        -1 otherwise.
 */
static int
qemuMigrationDriveMirror(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuMigrationCookiePtr mig,
                         const char *host,
                         unsigned long speed,
2038 2039
                         unsigned int *migrate_flags,
                         size_t nmigrate_disks,
2040 2041
                         const char **migrate_disks,
                         virConnectPtr dconn)
2042 2043 2044 2045
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    int port;
2046
    size_t i;
2047 2048
    char *diskAlias = NULL;
    char *nbd_dest = NULL;
2049
    char *hoststr = NULL;
2050
    unsigned int mirror_flags = VIR_DOMAIN_BLOCK_REBASE_REUSE_EXT;
2051
    int rv;
2052
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2053 2054

    VIR_DEBUG("Starting drive mirrors for domain %s", vm->def->name);
2055 2056 2057 2058 2059

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

2060 2061
    /* escape literal IPv6 address */
    if (strchr(host, ':')) {
2062
        if (virAsprintf(&hoststr, "[%s]", host) < 0)
2063
            goto cleanup;
2064
    } else if (VIR_STRDUP(hoststr, host) < 0) {
2065
        goto cleanup;
2066 2067
    }

2068 2069 2070 2071 2072
    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];
2073
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
2074
        int mon_ret;
2075

2076 2077
        /* check whether disk should be migrated */
        if (!qemuMigrateDisk(disk, nmigrate_disks, migrate_disks))
2078 2079 2080 2081 2082
            continue;

        if ((virAsprintf(&diskAlias, "%s%s",
                         QEMU_DRIVE_HOST_PREFIX, disk->info.alias) < 0) ||
            (virAsprintf(&nbd_dest, "nbd:%s:%d:exportname=%s",
2083
                         hoststr, port, diskAlias) < 0))
2084 2085
            goto cleanup;

2086
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
2087
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
2088 2089
            goto cleanup;

2090
        qemuBlockJobSyncBegin(disk);
2091
        /* Force "raw" format for NBD export */
2092
        mon_ret = qemuMonitorDriveMirror(priv->mon, diskAlias, nbd_dest,
2093
                                         "raw", speed, 0, 0, mirror_flags);
2094 2095
        VIR_FREE(diskAlias);
        VIR_FREE(nbd_dest);
2096

2097
        if (qemuDomainObjExitMonitor(driver, vm) < 0 || mon_ret < 0) {
2098
            qemuBlockJobSyncEnd(driver, vm, disk);
2099 2100
            goto cleanup;
        }
2101
        diskPriv->migrating = true;
2102 2103 2104 2105 2106

        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto cleanup;
        }
2107
    }
2108

2109 2110 2111
    while ((rv = qemuMigrationDriveMirrorReady(driver, vm)) != 1) {
        if (rv < 0)
            goto cleanup;
2112

2113
        if (priv->job.abortJob) {
2114 2115 2116 2117 2118
            priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
            virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                           qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                           _("canceled by client"));
            goto cleanup;
2119
        }
2120

2121 2122 2123 2124 2125 2126
        if (dconn && virConnectIsAlive(dconn) <= 0) {
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("Lost connection to destination host"));
            goto cleanup;
        }

2127
        if (virDomainObjWait(vm) < 0)
2128
            goto cleanup;
2129 2130
    }

2131
    /* Okay, all disks are ready. Modify migrate_flags */
2132 2133 2134 2135
    *migrate_flags &= ~(QEMU_MONITOR_MIGRATE_NON_SHARED_DISK |
                        QEMU_MONITOR_MIGRATE_NON_SHARED_INC);
    ret = 0;

2136
 cleanup:
2137
    virObjectUnref(cfg);
2138 2139
    VIR_FREE(diskAlias);
    VIR_FREE(nbd_dest);
2140
    VIR_FREE(hoststr);
2141 2142
    return ret;
}
2143

2144

2145 2146 2147 2148 2149 2150 2151 2152 2153
/* Validate whether the domain is safe to migrate.  If vm is NULL,
 * then this is being run in the v2 Prepare stage on the destination
 * (where we only have the target xml); if vm is provided, then this
 * is being run in either v2 Perform or v3 Begin (where we also have
 * access to all of the domain's metadata, such as whether it is
 * marked autodestroy or has snapshots).  While it would be nice to
 * assume that checking on source is sufficient to prevent ever
 * talking to the destination in the first place, we are stuck with
 * the fact that older servers did not do checks on the source. */
2154
bool
2155
qemuMigrationIsAllowed(virQEMUDriverPtr driver, virDomainObjPtr vm,
2156
                       virDomainDefPtr def, bool remote, bool abort_on_error)
2157
{
2158
    int nsnapshots;
2159
    int pauseReason;
2160
    size_t i;
2161 2162 2163

    if (vm) {
        if (qemuProcessAutoDestroyActive(driver, vm)) {
2164 2165
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("domain is marked for auto destroy"));
2166 2167
            return false;
        }
2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180

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

            if (nsnapshots > 0) {
                virReportError(VIR_ERR_OPERATION_INVALID,
                               _("cannot migrate domain with %d snapshots"),
                               nsnapshots);
                return false;
            }
2181 2182

            /* cancel migration if disk I/O error is emitted while migrating */
2183 2184
            if (abort_on_error &&
                virDomainObjGetState(vm, &pauseReason) == VIR_DOMAIN_PAUSED &&
2185 2186 2187 2188 2189 2190
                pauseReason == VIR_DOMAIN_PAUSED_IOERROR) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                               _("cannot migrate domain with I/O error"));
                return false;
            }

2191
        }
2192

2193
        if (qemuDomainHasBlockjob(vm, false)) {
E
Eric Blake 已提交
2194
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
2195
                           _("domain has an active block job"));
E
Eric Blake 已提交
2196 2197
            return false;
        }
2198 2199 2200

        def = vm->def;
    }
2201 2202

    /* Migration with USB host devices is allowed, all other devices are
2203
     * forbidden. */
2204 2205 2206 2207
    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) {
2208 2209 2210
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("domain has assigned non-USB host devices"));
            return false;
2211 2212
        }
    }
2213

2214 2215 2216
    if (def->cpu && def->cpu->mode != VIR_CPU_MODE_HOST_PASSTHROUGH) {
        for (i = 0; i < def->cpu->nfeatures; i++) {
            virCPUFeatureDefPtr feature = &def->cpu->features[i];
J
Ján Tomko 已提交
2217

2218 2219
            if (feature->policy != VIR_CPU_FEATURE_REQUIRE)
                continue;
J
Ján Tomko 已提交
2220

2221 2222 2223 2224 2225 2226 2227
            /* 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
2228 2229 2230
        }
    }

2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244
    /* Verify that memory device config can be transferred reliably */
    for (i = 0; i < def->nmems; i++) {
        virDomainMemoryDefPtr mem = def->mems[i];

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

            return false;
        }
    }

2245 2246 2247
    return true;
}

2248
static bool
2249 2250 2251
qemuMigrationIsSafe(virDomainDefPtr def,
                    size_t nmigrate_disks,
                    const char **migrate_disks)
2252
{
2253
    size_t i;
2254

2255
    for (i = 0; i < def->ndisks; i++) {
2256
        virDomainDiskDefPtr disk = def->disks[i];
2257
        const char *src = virDomainDiskGetSource(disk);
2258

2259 2260
        /* Our code elsewhere guarantees shared disks are either readonly (in
         * which case cache mode doesn't matter) or used with cache=none */
2261
        if (qemuMigrateDisk(disk, nmigrate_disks, migrate_disks) &&
2262
            disk->cachemode != VIR_DOMAIN_DISK_CACHE_DISABLE) {
2263
            int rc;
2264

E
Eric Blake 已提交
2265
            if (virDomainDiskGetType(disk) == VIR_STORAGE_TYPE_FILE) {
2266
                if ((rc = virFileIsSharedFS(src)) < 0)
2267 2268
                    return false;
                else if (rc == 0)
2269
                    continue;
2270
                if ((rc = virStorageFileIsClusterFS(src)) < 0)
2271
                    return false;
2272 2273
                else if (rc == 1)
                    continue;
2274 2275
            } else if (disk->src->type == VIR_STORAGE_TYPE_NETWORK &&
                       disk->src->protocol == VIR_STORAGE_NET_PROTOCOL_RBD) {
2276
                continue;
2277
            }
2278

2279 2280 2281
            virReportError(VIR_ERR_MIGRATE_UNSAFE, "%s",
                           _("Migration may lead to data corruption if disks"
                             " use cache != none"));
2282 2283 2284 2285 2286 2287 2288
            return false;
        }
    }

    return true;
}

2289 2290 2291 2292
/** qemuMigrationSetOffline
 * Pause domain for non-live migration.
 */
int
2293
qemuMigrationSetOffline(virQEMUDriverPtr driver,
2294 2295 2296
                        virDomainObjPtr vm)
{
    int ret;
2297
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
2298 2299
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
2300
    if (ret == 0) {
2301
        virObjectEventPtr event;
2302

2303
        event = virDomainEventLifecycleNewFromObj(vm,
2304 2305 2306 2307 2308 2309 2310 2311 2312 2313
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
        if (event)
            qemuDomainEventQueue(driver, event);
    }

    return ret;
}


2314 2315 2316
static int
qemuMigrationSetCompression(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
2317
                            bool state,
2318
                            qemuDomainAsyncJob job)
2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331
{
    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;
2332 2333 2334
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350
    } 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,
2351 2352
                QEMU_MONITOR_MIGRATION_CAPS_XBZRLE,
                state);
2353

2354
 cleanup:
2355 2356
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2357 2358 2359
    return ret;
}

2360 2361 2362
static int
qemuMigrationSetAutoConverge(virQEMUDriverPtr driver,
                             virDomainObjPtr vm,
2363
                             bool state,
2364
                             qemuDomainAsyncJob job)
2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377
{
    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;
2378 2379 2380
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2381 2382 2383 2384 2385 2386 2387 2388 2389 2390
    } 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,
2391 2392
                QEMU_MONITOR_MIGRATION_CAPS_AUTO_CONVERGE,
                state);
2393

2394
 cleanup:
2395 2396
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2397 2398 2399 2400
    return ret;
}


2401 2402 2403
static int
qemuMigrationSetPinAll(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
2404
                       bool state,
2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418
                       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;
2419 2420 2421
    } else if (ret == 0 && !state) {
        /* Unsupported but we want it off anyway */
        goto cleanup;
2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437
    } 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,
2438 2439
                QEMU_MONITOR_MIGRATION_CAPS_RDMA_PIN_ALL,
                state);
2440 2441

 cleanup:
2442 2443
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2444 2445 2446
    return ret;
}

2447
static int
2448
qemuMigrationWaitForSpice(virDomainObjPtr vm)
2449 2450 2451
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool wait_for_spice = false;
2452
    size_t i = 0;
2453

2454 2455 2456 2457 2458 2459 2460
    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;
2461 2462
        }
    }
2463 2464 2465 2466

    if (!wait_for_spice)
        return 0;

2467 2468
    while (!priv->job.spiceMigrated && !priv->job.abortJob) {
        if (virDomainObjWait(vm) < 0)
2469 2470 2471 2472
            return -1;
    }
    return 0;
}
2473

2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525

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);
}


2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543
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");
    }
}


2544
static int
2545
qemuMigrationUpdateJobStatus(virQEMUDriverPtr driver,
2546
                             virDomainObjPtr vm,
2547
                             qemuDomainAsyncJob asyncJob)
2548
{
2549
    qemuDomainObjPrivatePtr priv = vm->privateData;
2550 2551
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    qemuDomainJobInfo newInfo = *jobInfo;
2552

2553 2554
    if (qemuMigrationFetchJobStatus(driver, vm, asyncJob, &newInfo) < 0)
        return -1;
2555

2556 2557 2558
    *jobInfo = newInfo;
    return 0;
}
2559

2560

2561 2562 2563 2564 2565 2566 2567
static int
qemuMigrationCheckJobStatus(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
                            qemuDomainAsyncJob asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
2568

2569 2570 2571 2572 2573
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);

    if (events)
        qemuMigrationUpdateJobType(jobInfo);
    else if (qemuMigrationUpdateJobStatus(driver, vm, asyncJob) < 0)
2574
        return -1;
2575

2576 2577
    switch (jobInfo->type) {
    case VIR_DOMAIN_JOB_NONE:
2578 2579
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("is not active"));
2580
        return -1;
2581

2582
    case VIR_DOMAIN_JOB_FAILED:
2583 2584
        virReportError(VIR_ERR_OPERATION_FAILED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("unexpectedly failed"));
2585
        return -1;
2586

2587
    case VIR_DOMAIN_JOB_CANCELLED:
2588 2589
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuMigrationJobName(vm), _("canceled by client"));
2590 2591
        return -1;

2592 2593 2594 2595 2596 2597 2598
    case VIR_DOMAIN_JOB_COMPLETED:
        /* Fetch statistics of a completed migration */
        if (events &&
            qemuMigrationUpdateJobStatus(driver, vm, asyncJob) < 0)
            return -1;
        break;

2599 2600 2601
    case VIR_DOMAIN_JOB_BOUNDED:
    case VIR_DOMAIN_JOB_UNBOUNDED:
    case VIR_DOMAIN_JOB_LAST:
2602 2603
        break;
    }
2604
    return 0;
2605 2606 2607
}


2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664
/**
 * 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,
                       bool abort_on_error,
                       bool storage)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
    int pauseReason;

    if (qemuMigrationCheckJobStatus(driver, vm, asyncJob) < 0)
        goto error;

    if (storage && qemuMigrationDriveMirrorReady(driver, vm) < 0)
        goto error;

    if (abort_on_error &&
        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;
    }
}


2665 2666 2667
/* Returns 0 on success, -2 when migration needs to be cancelled, or -1 when
 * QEMU reports failed migration.
 */
2668
static int
J
Jiri Denemark 已提交
2669 2670
qemuMigrationWaitForCompletion(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
2671
                               qemuDomainAsyncJob asyncJob,
J
Jiri Denemark 已提交
2672
                               virConnectPtr dconn,
2673 2674
                               bool abort_on_error,
                               bool storage)
2675
{
2676
    qemuDomainObjPrivatePtr priv = vm->privateData;
J
Jiri Denemark 已提交
2677
    qemuDomainJobInfoPtr jobInfo = priv->job.current;
2678
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
2679
    int rv;
2680

J
Jiri Denemark 已提交
2681
    jobInfo->type = VIR_DOMAIN_JOB_UNBOUNDED;
2682 2683 2684 2685
    while ((rv = qemuMigrationCompleted(driver, vm, asyncJob, dconn,
                                        abort_on_error, storage)) != 1) {
        if (rv < 0)
            return rv;
2686

2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699
        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);
        }
2700 2701
    }

2702 2703 2704 2705 2706
    qemuDomainJobInfoUpdateDowntime(jobInfo);
    VIR_FREE(priv->job.completed);
    if (VIR_ALLOC(priv->job.completed) == 0)
        *priv->job.completed = *jobInfo;

2707
    return 0;
2708 2709 2710
}


2711
static int
2712
qemuDomainMigrateGraphicsRelocate(virQEMUDriverPtr driver,
2713
                                  virDomainObjPtr vm,
2714 2715
                                  qemuMigrationCookiePtr cookie,
                                  const char *graphicsuri)
2716 2717
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2718 2719
    int ret = -1;
    const char *listenAddress = NULL;
2720
    virSocketAddr addr;
2721 2722 2723 2724 2725
    virURIPtr uri = NULL;
    int type = -1;
    int port = -1;
    int tlsPort = -1;
    const char *tlsSubject = NULL;
2726

2727
    if (!cookie || (!cookie->graphics && !graphicsuri))
2728 2729
        return 0;

2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748
    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) {
2749
        size_t i;
2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776

        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;
            }
        }
    }
2777 2778 2779 2780

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
2781 2782 2783 2784
    if (type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        ret = 0;
        goto cleanup;
    }
2785

2786 2787 2788 2789
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon, type, listenAddress,
                                          port, tlsPort, tlsSubject);
2790 2791
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
2792
    }
2793

2794
 cleanup:
2795
    virURIFree(uri);
2796 2797 2798 2799
    return ret;
}


2800
static int
2801
qemuDomainMigrateOPDRelocate(virQEMUDriverPtr driver ATTRIBUTE_UNUSED,
2802 2803 2804
                             virDomainObjPtr vm,
                             qemuMigrationCookiePtr cookie)
{
2805 2806
    virDomainNetDefPtr netptr;
    int ret = -1;
2807
    size_t i;
2808 2809 2810 2811 2812 2813 2814 2815

    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:
2816
           break;
2817
        case VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH:
2818 2819
            if (virNetDevOpenvswitchSetMigrateData(cookie->network->net[i].portdata,
                                                   netptr->ifname) != 0) {
J
Jiri Denemark 已提交
2820 2821 2822
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unable to run command to set OVS port data for "
                                 "interface %s"), netptr->ifname);
2823 2824 2825
                goto cleanup;
            }
            break;
2826 2827 2828 2829 2830
        default:
            break;
        }
    }

2831
    ret = 0;
2832
 cleanup:
2833 2834 2835 2836
    return ret;
}


2837 2838 2839 2840 2841 2842
/* 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.
 */
2843 2844 2845 2846
static virDomainObjPtr
qemuMigrationCleanup(virDomainObjPtr vm,
                     virConnectPtr conn,
                     void *opaque)
2847
{
2848
    virQEMUDriverPtr driver = opaque;
2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863
    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);

2864
    switch ((qemuMigrationJobPhase) priv->job.phase) {
2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894
    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 */
        ;
    }

2895
 cleanup:
2896 2897 2898
    return vm;
}

2899

2900
/* The caller is supposed to lock the vm and start a migration job. */
2901 2902 2903 2904 2905 2906 2907
static char *
qemuMigrationBeginPhase(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dname,
                        char **cookieout,
                        int *cookieoutlen,
2908 2909
                        size_t nmigrate_disks,
                        const char **migrate_disks,
2910
                        unsigned long flags)
2911 2912 2913
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
2914
    virDomainDefPtr def = NULL;
2915
    qemuDomainObjPrivatePtr priv = vm->privateData;
2916
    virCapsPtr caps = NULL;
2917
    unsigned int cookieFlags = QEMU_MIGRATION_COOKIE_LOCKSTATE;
2918
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
2919

2920
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, dname=%s,"
2921 2922
              " cookieout=%p, cookieoutlen=%p,"
              " nmigrate_disks=%zu, migrate_disks=%p, flags=%lx",
2923
              driver, vm, NULLSTR(xmlin), NULLSTR(dname),
2924 2925
              cookieout, cookieoutlen, nmigrate_disks,
              migrate_disks, flags);
2926

2927 2928 2929
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

2930 2931 2932 2933 2934 2935
    /* 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);
2936

2937
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
2938 2939
        goto cleanup;

2940 2941
    if (!(flags & VIR_MIGRATE_UNSAFE) &&
        !qemuMigrationIsSafe(vm->def, nmigrate_disks, migrate_disks))
2942 2943
        goto cleanup;

2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987
    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;
            }
2988 2989 2990
        }
    }

2991
    if (vm->def->mem.max_memory ||
2992 2993
        ((flags & VIR_MIGRATE_PERSIST_DEST) &&
         vm->newDef &&
2994 2995 2996
         vm->newDef->mem.max_memory))
        cookieFlags |= QEMU_MIGRATION_COOKIE_MEMORY_HOTPLUG;

2997
    if (!(mig = qemuMigrationEatCookie(driver, vm, NULL, 0, 0)))
2998 2999 3000 3001
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
                                cookieout, cookieoutlen,
3002
                                cookieFlags) < 0)
3003 3004
        goto cleanup;

L
liguang 已提交
3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026
    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;
        }
    }

3027
    if (xmlin) {
3028
        if (!(def = virDomainDefParseString(xmlin, caps, driver->xmlopt,
3029
                                            VIR_DOMAIN_DEF_PARSE_INACTIVE)))
3030 3031
            goto cleanup;

3032
        if (!qemuDomainDefCheckABIStability(driver, vm->def, def))
3033 3034
            goto cleanup;

3035
        rv = qemuDomainDefFormatLive(driver, def, false, true);
3036
    } else {
3037
        rv = qemuDomainDefFormatLive(driver, vm->def, false, true);
3038
    }
3039

3040
 cleanup:
3041
    qemuMigrationCookieFree(mig);
3042
    virObjectUnref(caps);
3043
    virDomainDefFree(def);
3044 3045 3046
    return rv;
}

3047 3048 3049 3050 3051 3052 3053
char *
qemuMigrationBegin(virConnectPtr conn,
                   virDomainObjPtr vm,
                   const char *xmlin,
                   const char *dname,
                   char **cookieout,
                   int *cookieoutlen,
3054 3055
                   size_t nmigrate_disks,
                   const char **migrate_disks,
3056 3057 3058 3059
                   unsigned long flags)
{
    virQEMUDriverPtr driver = conn->privateData;
    char *xml = NULL;
3060
    qemuDomainAsyncJob asyncJob;
3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071

    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;
    }

3072 3073
    qemuMigrationStoreDomainState(vm);

3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088
    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,
3089
                                        nmigrate_disks, migrate_disks, flags)))
3090 3091 3092 3093 3094 3095 3096
        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.
         */
3097
        if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
3098 3099
                                 qemuMigrationCleanup) < 0) {
            VIR_FREE(xml);
3100
            goto endjob;
3101
        }
3102
        qemuMigrationJobContinue(vm);
3103 3104 3105 3106
    } else {
        goto endjob;
    }

3107
 cleanup:
M
Michal Privoznik 已提交
3108
    virDomainObjEndAPI(&vm);
3109 3110
    return xml;

3111
 endjob:
3112 3113 3114 3115
    if (flags & VIR_MIGRATE_CHANGE_PROTECTION)
        qemuMigrationJobFinish(driver, vm);
    else
        qemuDomainObjEndJob(driver, vm);
3116 3117 3118
    goto cleanup;
}

3119

3120 3121
/* Prepare is the first step, and it runs on the destination host.
 */
3122

3123
static void
3124
qemuMigrationPrepareCleanup(virQEMUDriverPtr driver,
3125 3126 3127 3128 3129 3130 3131 3132 3133 3134
                            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));

3135 3136 3137
    virPortAllocatorRelease(driver->migrationPorts, priv->migrationPort);
    priv->migrationPort = 0;

3138 3139 3140 3141 3142
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
        return;
    qemuDomainObjDiscardAsyncJob(driver, vm);
}

3143
static int
3144
qemuMigrationPrepareAny(virQEMUDriverPtr driver,
3145 3146 3147 3148 3149
                        virConnectPtr dconn,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
3150
                        virDomainDefPtr *def,
3151
                        const char *origname,
L
liguang 已提交
3152
                        virStreamPtr st,
3153
                        const char *protocol,
3154 3155
                        unsigned short port,
                        bool autoPort,
3156
                        const char *listenAddress,
3157 3158
                        size_t nmigrate_disks,
                        const char **migrate_disks,
L
liguang 已提交
3159
                        unsigned long flags)
3160 3161
{
    virDomainObjPtr vm = NULL;
3162
    virObjectEventPtr event = NULL;
3163
    int ret = -1;
3164
    int dataFD[2] = { -1, -1 };
3165
    qemuDomainObjPrivatePtr priv = NULL;
3166
    qemuMigrationCookiePtr mig = NULL;
3167
    bool tunnel = !!st;
J
Jiri Denemark 已提交
3168
    char *xmlout = NULL;
L
liguang 已提交
3169
    unsigned int cookieFlags;
3170
    virCapsPtr caps = NULL;
J
Ján Tomko 已提交
3171
    char *migrateFrom = NULL;
3172
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
3173
    bool taint_hook = false;
3174

3175 3176
    virNWFilterReadLockFilterUpdates();

L
liguang 已提交
3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198
    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;
        }
    }

3199 3200 3201
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

3202
    if (!qemuMigrationIsAllowed(driver, NULL, *def, true, abort_on_error))
3203 3204
        goto cleanup;

J
Jiri Denemark 已提交
3205 3206 3207 3208 3209
    /* Let migration hook filter domain XML */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml;
        int hookret;

3210
        if (!(xml = qemuDomainDefFormatXML(driver, *def,
3211 3212
                                           VIR_DOMAIN_XML_SECURE |
                                           VIR_DOMAIN_XML_MIGRATABLE)))
J
Jiri Denemark 已提交
3213 3214
            goto cleanup;

3215
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, (*def)->name,
J
Jiri Denemark 已提交
3216 3217 3218 3219 3220 3221 3222
                              VIR_HOOK_QEMU_OP_MIGRATE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, &xmlout);
        VIR_FREE(xml);

        if (hookret < 0) {
            goto cleanup;
        } else if (hookret == 0) {
3223
            if (virStringIsEmpty(xmlout)) {
J
Jiri Denemark 已提交
3224 3225 3226 3227 3228 3229
                VIR_DEBUG("Migrate hook filter returned nothing; using the"
                          " original XML");
            } else {
                virDomainDefPtr newdef;

                VIR_DEBUG("Using hook-filtered domain XML: %s", xmlout);
3230
                newdef = virDomainDefParseString(xmlout, caps, driver->xmlopt,
3231
                                                 VIR_DOMAIN_DEF_PARSE_INACTIVE);
J
Jiri Denemark 已提交
3232 3233 3234
                if (!newdef)
                    goto cleanup;

3235
                if (!qemuDomainDefCheckABIStability(driver, *def, newdef)) {
J
Jiri Denemark 已提交
3236 3237 3238 3239
                    virDomainDefFree(newdef);
                    goto cleanup;
                }

3240 3241
                virDomainDefFree(*def);
                *def = newdef;
3242 3243 3244 3245
                /* 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
3246 3247 3248 3249
            }
        }
    }

J
Ján Tomko 已提交
3250 3251 3252 3253
    if (tunnel) {
        /* QEMU will be started with -incoming stdio
         * (which qemu_command might convert to exec:cat or fd:n)
         */
3254
        if (VIR_STRDUP(migrateFrom, "stdio") < 0)
J
Ján Tomko 已提交
3255 3256
            goto cleanup;
    } else {
3257 3258 3259
        bool encloseAddress = false;
        bool hostIPv6Capable = false;
        bool qemuIPv6Capable = false;
J
Ján Tomko 已提交
3260 3261 3262 3263
        virQEMUCapsPtr qemuCaps = NULL;
        struct addrinfo *info = NULL;
        struct addrinfo hints = { .ai_flags = AI_ADDRCONFIG,
                                  .ai_socktype = SOCK_STREAM };
3264
        const char *incFormat;
J
Ján Tomko 已提交
3265

3266 3267 3268 3269
        if (getaddrinfo("::", NULL, &hints, &info) == 0) {
            freeaddrinfo(info);
            hostIPv6Capable = true;
        }
J
Ján Tomko 已提交
3270
        if (!(qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
3271 3272
                                                    (*def)->emulator,
                                                    (*def)->os.machine)))
J
Ján Tomko 已提交
3273 3274
            goto cleanup;

3275 3276 3277 3278
        qemuIPv6Capable = virQEMUCapsGet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
        virObjectUnref(qemuCaps);

        if (listenAddress) {
3279 3280 3281 3282 3283 3284 3285 3286 3287
            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"));
3288 3289
                    goto cleanup;
                }
3290 3291
                /* IPv6 address must be escaped in brackets on the cmd line */
                encloseAddress = true;
3292
            } else {
3293
                /* listenAddress is a hostname or IPv4 */
3294
            }
3295
        } else if (qemuIPv6Capable && hostIPv6Capable) {
3296 3297 3298
            /* Listen on :: instead of 0.0.0.0 if QEMU understands it
             * and there is at least one IPv6 address configured
             */
3299 3300 3301 3302
            listenAddress = "::";
            encloseAddress = true;
        } else {
            listenAddress = "0.0.0.0";
J
Ján Tomko 已提交
3303 3304
        }

3305 3306 3307 3308
        /* 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
3309
         */
3310 3311 3312 3313 3314 3315
        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
3316 3317 3318
            goto cleanup;
    }

3319
    if (!(vm = virDomainObjListAdd(driver->domains, *def,
3320
                                   driver->xmlopt,
3321 3322 3323
                                   VIR_DOMAIN_OBJ_LIST_ADD_LIVE |
                                   VIR_DOMAIN_OBJ_LIST_ADD_CHECK_LIVE,
                                   NULL)))
3324
        goto cleanup;
3325

3326
    virObjectRef(vm);
3327
    *def = NULL;
3328
    priv = vm->privateData;
3329 3330
    if (VIR_STRDUP(priv->origname, origname) < 0)
        goto cleanup;
3331

3332 3333 3334 3335 3336
    if (taint_hook) {
        /* Domain XML has been altered by a hook script. */
        priv->hookRun = true;
    }

3337
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
3338 3339
                                       QEMU_MIGRATION_COOKIE_LOCKSTATE |
                                       QEMU_MIGRATION_COOKIE_NBD)))
3340 3341
        goto cleanup;

3342 3343
    if (STREQ_NULLABLE(protocol, "rdma") &&
        !virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
3344 3345 3346 3347 3348 3349
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("cannot start RDMA migration with no memory hard "
                         "limit set"));
        goto cleanup;
    }

3350
    if (qemuMigrationPrecreateStorage(dconn, driver, vm, mig->nbd,
3351 3352
                                      nmigrate_disks, migrate_disks,
                                      !!(flags & VIR_MIGRATE_NON_SHARED_INC)) < 0)
3353 3354
        goto cleanup;

3355
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
3356
        goto cleanup;
3357
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
3358 3359 3360 3361

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

L
liguang 已提交
3362 3363 3364
    if (flags & VIR_MIGRATE_OFFLINE)
        goto done;

3365 3366
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
3367 3368
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
3369 3370 3371 3372
        goto endjob;
    }

    /* Start the QEMU daemon, with the same command-line arguments plus
3373
     * -incoming $migrateFrom
3374
     */
3375 3376
    if (qemuProcessStart(dconn, driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN,
                         migrateFrom, dataFD[0], NULL, NULL,
3377 3378
                         VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_START,
                         VIR_QEMU_PROCESS_START_PAUSED |
3379
                         VIR_QEMU_PROCESS_START_AUTODESTROY) < 0) {
3380
        virDomainAuditStart(vm, "migrated", false);
3381 3382 3383
        goto endjob;
    }

3384 3385 3386 3387
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
3388
            goto stop;
3389
        }
3390
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
3391 3392
    }

3393 3394
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
3395 3396 3397
                                    QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
        goto stop;

3398
    if (STREQ_NULLABLE(protocol, "rdma") &&
M
Michael R. Hines 已提交
3399 3400 3401 3402
        virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
        goto stop;
    }

3403 3404 3405
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
                               QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
3406 3407
        goto stop;

3408 3409 3410 3411 3412 3413 3414 3415
    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");
    }
3416

3417
 done:
L
liguang 已提交
3418 3419 3420 3421 3422
    if (flags & VIR_MIGRATE_OFFLINE)
        cookieFlags = 0;
    else
        cookieFlags = QEMU_MIGRATION_COOKIE_GRAPHICS;

3423 3424 3425
    if (mig->nbd &&
        flags & (VIR_MIGRATE_NON_SHARED_DISK | VIR_MIGRATE_NON_SHARED_INC) &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NBD_SERVER)) {
3426 3427
        if (qemuMigrationStartNBDServer(driver, vm, listenAddress,
                                        nmigrate_disks, migrate_disks) < 0) {
3428 3429
            /* error already reported */
            goto endjob;
3430
        }
3431
        cookieFlags |= QEMU_MIGRATION_COOKIE_NBD;
3432 3433 3434 3435
    }

    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
3436 3437 3438 3439 3440 3441 3442
        /* 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");
    }

3443 3444 3445
    if (qemuDomainCleanupAdd(vm, qemuMigrationPrepareCleanup) < 0)
        goto endjob;

L
liguang 已提交
3446 3447
    if (!(flags & VIR_MIGRATE_OFFLINE)) {
        virDomainAuditStart(vm, "migrated", true);
3448
        event = virDomainEventLifecycleNewFromObj(vm,
L
liguang 已提交
3449 3450 3451
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    }
3452

3453 3454 3455 3456
    /* 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.
     */
3457
    qemuMigrationJobContinue(vm);
3458

3459 3460
    if (autoPort)
        priv->migrationPort = port;
3461
    ret = 0;
3462

3463
 cleanup:
J
Ján Tomko 已提交
3464
    VIR_FREE(migrateFrom);
J
Jiri Denemark 已提交
3465
    VIR_FREE(xmlout);
3466 3467
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
3468 3469 3470
    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 */
3471
        VIR_FREE(priv->origname);
3472 3473
        virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
        priv->nbdPort = 0;
3474 3475
        if (!vm->persistent)
            qemuDomainRemoveInactive(driver, vm);
3476
    }
M
Michal Privoznik 已提交
3477
    virDomainObjEndAPI(&vm);
3478 3479
    if (event)
        qemuDomainEventQueue(driver, event);
3480
    qemuMigrationCookieFree(mig);
3481
    virObjectUnref(caps);
3482
    virNWFilterUnlockFilterUpdates();
3483
    return ret;
3484

3485
 stop:
3486
    virDomainAuditStart(vm, "migrated", false);
3487 3488
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                    VIR_QEMU_PROCESS_STOP_MIGRATED);
3489

3490
 endjob:
3491
    qemuMigrationJobFinish(driver, vm);
3492
    goto cleanup;
3493 3494 3495
}


3496 3497 3498 3499 3500
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
3501
qemuMigrationPrepareTunnel(virQEMUDriverPtr driver,
3502 3503 3504 3505 3506 3507
                           virConnectPtr dconn,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virStreamPtr st,
3508
                           virDomainDefPtr *def,
3509
                           const char *origname,
L
liguang 已提交
3510
                           unsigned long flags)
3511 3512 3513 3514
{
    int ret;

    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
3515
              "cookieout=%p, cookieoutlen=%p, st=%p, def=%p, "
3516
              "origname=%s, flags=%lx",
3517
              driver, dconn, NULLSTR(cookiein), cookieinlen,
3518
              cookieout, cookieoutlen, st, *def, origname, flags);
3519

3520 3521 3522 3523 3524 3525
    if (st == NULL) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("tunnelled migration requested but NULL stream passed"));
        return -1;
    }

3526
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3527
                                  cookieout, cookieoutlen, def, origname,
3528
                                  st, NULL, 0, false, NULL, 0, NULL, flags);
3529 3530 3531 3532
    return ret;
}


3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555
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;
}


3556
int
3557
qemuMigrationPrepareDirect(virQEMUDriverPtr driver,
3558
                           virConnectPtr dconn,
3559 3560 3561 3562
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
3563 3564
                           const char *uri_in,
                           char **uri_out,
3565
                           virDomainDefPtr *def,
3566
                           const char *origname,
3567
                           const char *listenAddress,
3568 3569
                           size_t nmigrate_disks,
                           const char **migrate_disks,
L
liguang 已提交
3570
                           unsigned long flags)
3571
{
3572 3573
    unsigned short port = 0;
    bool autoPort = true;
3574 3575
    char *hostname = NULL;
    int ret = -1;
3576
    virURIPtr uri = NULL;
3577 3578
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    const char *migrateHost = cfg->migrateHost;
J
Jiri Denemark 已提交
3579

3580 3581
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
3582 3583
              "def=%p, origname=%s, listenAddress=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p, flags=%lx",
3584 3585
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
3586 3587
              *def, origname, NULLSTR(listenAddress),
              nmigrate_disks, migrate_disks, flags);
3588

3589 3590
    *uri_out = NULL;

3591 3592 3593
    /* 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
3594 3595 3596
     * 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".
3597 3598 3599 3600 3601 3602
     *
     * 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) {
3603 3604 3605
        bool encloseAddress = false;
        const char *incFormat;

3606
        if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3607
            goto cleanup;
3608

3609
        if (migrateHost != NULL) {
3610 3611
            if (virSocketAddrNumericFamily(migrateHost) == AF_INET6)
                encloseAddress = true;
3612

3613
            if (VIR_STRDUP(hostname, migrateHost) < 0)
3614 3615 3616 3617 3618
                goto cleanup;
        } else {
            if ((hostname = virGetHostname()) == NULL)
                goto cleanup;
        }
3619 3620

        if (STRPREFIX(hostname, "localhost")) {
3621 3622 3623
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("hostname on destination resolved to localhost,"
                             " but migration requires an FQDN"));
3624 3625 3626 3627 3628
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
3629
         * compatibility with old targets. We at least make the
3630 3631
         * new targets accept both syntaxes though.
         */
3632 3633 3634 3635 3636 3637
        if (encloseAddress)
            incFormat = "%s:[%s]:%d";
        else
            incFormat = "%s:%s:%d";

        if (virAsprintf(uri_out, incFormat, "tcp", hostname, port) < 0)
3638 3639
            goto cleanup;
    } else {
3640
        bool well_formed_uri;
J
Ján Tomko 已提交
3641

3642 3643
        if (!(uri = qemuMigrationParseURI(uri_in, &well_formed_uri)))
            goto cleanup;
J
Ján Tomko 已提交
3644

3645 3646 3647 3648 3649 3650 3651
        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
3652 3653
        if (STRNEQ(uri->scheme, "tcp") &&
            STRNEQ(uri->scheme, "rdma")) {
3654 3655 3656
            virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED,
                           _("unsupported scheme %s in migration URI %s"),
                           uri->scheme, uri_in);
J
Ján Tomko 已提交
3657 3658 3659 3660 3661 3662 3663 3664 3665 3666
            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) {
3667
            if (virPortAllocatorAcquire(driver->migrationPorts, &port) < 0)
3668
                goto cleanup;
3669

3670
            /* Send well-formed URI only if uri_in was well-formed */
3671 3672 3673 3674 3675 3676 3677 3678
            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;
            }
3679
        } else {
3680 3681
            port = uri->port;
            autoPort = false;
3682 3683 3684 3685 3686 3687
        }
    }

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

3688
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
3689
                                  cookieout, cookieoutlen, def, origname,
3690
                                  NULL, uri ? uri->scheme : "tcp",
3691 3692
                                  port, autoPort, listenAddress,
                                  nmigrate_disks, migrate_disks, flags);
3693
 cleanup:
3694
    virURIFree(uri);
3695
    VIR_FREE(hostname);
3696
    virObjectUnref(cfg);
3697
    if (ret != 0) {
3698
        VIR_FREE(*uri_out);
3699 3700 3701
        if (autoPort)
            virPortAllocatorRelease(driver->migrationPorts, port);
    }
3702 3703 3704 3705
    return ret;
}


3706 3707 3708
virDomainDefPtr
qemuMigrationPrepareDef(virQEMUDriverPtr driver,
                        const char *dom_xml,
3709 3710
                        const char *dname,
                        char **origname)
3711 3712 3713
{
    virCapsPtr caps = NULL;
    virDomainDefPtr def;
3714
    char *name = NULL;
3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725

    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,
3726
                                        VIR_DOMAIN_DEF_PARSE_INACTIVE)))
3727 3728 3729
        goto cleanup;

    if (dname) {
3730
        name = def->name;
3731 3732 3733 3734 3735 3736
        if (VIR_STRDUP(def->name, dname) < 0) {
            virDomainDefFree(def);
            def = NULL;
        }
    }

3737
 cleanup:
3738
    virObjectUnref(caps);
3739 3740 3741 3742
    if (def && origname)
        *origname = name;
    else
        VIR_FREE(name);
3743 3744 3745 3746
    return def;
}


3747 3748 3749 3750 3751 3752 3753 3754 3755 3756
static int
qemuMigrationConfirmPhase(virQEMUDriverPtr driver,
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *cookiein,
                          int cookieinlen,
                          unsigned int flags,
                          int retcode)
{
    qemuMigrationCookiePtr mig;
3757
    virObjectEventPtr event = NULL;
3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772
    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);

3773 3774
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                       QEMU_MIGRATION_COOKIE_STATS)))
3775 3776
        goto cleanup;

3777 3778 3779
    /* Update total times with the values sent by the destination daemon */
    if (mig->jobInfo) {
        qemuDomainObjPrivatePtr priv = vm->privateData;
3780 3781 3782
        VIR_FREE(priv->job.completed);
        priv->job.completed = mig->jobInfo;
        mig->jobInfo = NULL;
3783 3784
    }

3785 3786 3787 3788 3789 3790 3791 3792 3793
    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 */
3794
        qemuMigrationWaitForSpice(vm);
3795 3796 3797 3798 3799

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

3800
        event = virDomainEventLifecycleNewFromObj(vm,
3801 3802 3803
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
    } else {
3804
        virErrorPtr orig_err = virSaveLastError();
3805 3806

        /* cancel any outstanding NBD jobs */
3807
        qemuMigrationCancelDriveMirror(driver, vm, false,
3808
                                       QEMU_ASYNC_JOB_MIGRATION_OUT, NULL);
3809 3810 3811

        virSetError(orig_err);
        virFreeError(orig_err);
3812

3813 3814 3815 3816
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
3817 3818 3819 3820 3821 3822 3823 3824
        }

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

3825
 done:
3826 3827 3828
    qemuMigrationCookieFree(mig);
    rv = 0;

3829
 cleanup:
3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844
    if (event)
        qemuDomainEventQueue(driver, event);
    virObjectUnref(cfg);
    return rv;
}

int
qemuMigrationConfirm(virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *cookiein,
                     int cookieinlen,
                     unsigned int flags,
                     int cancelled)
{
    virQEMUDriverPtr driver = conn->privateData;
3845
    qemuMigrationJobPhase phase;
3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859
    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);
3860 3861
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
3862 3863 3864 3865 3866

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

3867 3868 3869
    qemuMigrationJobFinish(driver, vm);
    if (!virDomainObjIsActive(vm) &&
        (!vm->persistent || (flags & VIR_MIGRATE_UNDEFINE_SOURCE))) {
3870 3871 3872 3873 3874
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
        qemuDomainRemoveInactive(driver, vm);
    }

3875
 cleanup:
M
Michal Privoznik 已提交
3876
    virDomainObjEndAPI(&vm);
3877 3878 3879 3880 3881
    virObjectUnref(cfg);
    return ret;
}


3882 3883
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
3884
    MIGRATION_DEST_CONNECT_HOST,
3885
    MIGRATION_DEST_UNIX,
3886
    MIGRATION_DEST_FD,
3887
};
3888

3889 3890 3891 3892
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
3893

3894 3895 3896 3897 3898 3899
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
3900
            const char *protocol;
3901 3902 3903 3904 3905
            const char *name;
            int port;
        } host;

        struct {
3906
            char *file;
3907 3908
            int sock;
        } unix_socket;
3909 3910 3911 3912 3913

        struct {
            int qemu;
            int local;
        } fd;
3914 3915 3916 3917 3918 3919 3920
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
3921 3922 3923

#define TUNNEL_SEND_BUF_SIZE 65536

3924 3925 3926 3927 3928 3929 3930
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
3931 3932
    int wakeupRecvFD;
    int wakeupSendFD;
3933 3934 3935
};

static void qemuMigrationIOFunc(void *arg)
3936
{
3937
    qemuMigrationIOThreadPtr data = arg;
3938 3939 3940 3941 3942 3943 3944
    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);
3945

3946
    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0)
3947
        goto abrt;
3948

3949 3950 3951
    fds[0].fd = data->sock;
    fds[1].fd = data->wakeupRecvFD;

3952
    for (;;) {
3953 3954 3955 3956 3957 3958 3959 3960 3961 3962
        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;
3963
            virReportSystemError(errno, "%s",
3964 3965
                                 _("poll failed in migration tunnel"));
            goto abrt;
3966
        }
3967 3968 3969 3970 3971 3972 3973

        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");
3974
            break;
3975
        }
3976

3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992
        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;
            }
3993 3994
        }

3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011
        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;
            }
        }
    }
4012

4013 4014
    if (virStreamFinish(data->st) < 0)
        goto error;
4015

4016 4017
    VIR_FREE(buffer);

4018 4019
    return;

4020
 abrt:
4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031
    err = virSaveLastError();
    if (err && err->code == VIR_ERR_OK) {
        virFreeError(err);
        err = NULL;
    }
    virStreamAbort(data->st);
    if (err) {
        virSetError(err);
        virFreeError(err);
    }

4032
 error:
4033 4034
    virCopyLastError(&data->err);
    virResetLastError();
4035
    VIR_FREE(buffer);
4036 4037 4038 4039 4040 4041 4042
}


static qemuMigrationIOThreadPtr
qemuMigrationStartTunnel(virStreamPtr st,
                         int sock)
{
4043 4044
    qemuMigrationIOThreadPtr io = NULL;
    int wakeupFD[2] = { -1, -1 };
4045

4046 4047 4048 4049
    if (pipe2(wakeupFD, O_CLOEXEC) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to make pipe"));
        goto error;
4050 4051
    }

4052
    if (VIR_ALLOC(io) < 0)
4053
        goto error;
4054

4055 4056
    io->st = st;
    io->sock = sock;
4057 4058
    io->wakeupRecvFD = wakeupFD[0];
    io->wakeupSendFD = wakeupFD[1];
4059 4060 4061 4062 4063 4064

    if (virThreadCreate(&io->thread, true,
                        qemuMigrationIOFunc,
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
4065
        goto error;
4066 4067 4068
    }

    return io;
4069

4070
 error:
4071 4072 4073 4074
    VIR_FORCE_CLOSE(wakeupFD[0]);
    VIR_FORCE_CLOSE(wakeupFD[1]);
    VIR_FREE(io);
    return NULL;
4075 4076 4077
}

static int
4078
qemuMigrationStopTunnel(qemuMigrationIOThreadPtr io, bool error)
4079 4080
{
    int rv = -1;
4081 4082 4083 4084 4085 4086 4087 4088 4089
    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;
    }

4090 4091 4092 4093
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
4094 4095 4096 4097
        if (error)
            rv = 0;
        else
            virSetError(&io->err);
4098 4099 4100 4101 4102 4103
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

4104
 cleanup:
4105 4106
    VIR_FORCE_CLOSE(io->wakeupSendFD);
    VIR_FORCE_CLOSE(io->wakeupRecvFD);
4107 4108
    VIR_FREE(io);
    return rv;
4109 4110
}

4111
static int
4112
qemuMigrationConnect(virQEMUDriverPtr driver,
4113 4114 4115 4116 4117 4118 4119 4120 4121
                     virDomainObjPtr vm,
                     qemuMigrationSpecPtr spec)
{
    virNetSocketPtr sock;
    const char *host;
    char *port = NULL;
    int ret = -1;

    host = spec->dest.host.name;
4122
    if (virAsprintf(&port, "%d", spec->dest.host.port) < 0)
4123 4124 4125 4126 4127 4128 4129
        return -1;

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

    if (virSecurityManagerSetSocketLabel(driver->securityManager, vm->def) < 0)
        goto cleanup;
4130 4131 4132
    if (virNetSocketNewConnectTCP(host, port,
                                  AF_UNSPEC,
                                  &sock) == 0) {
4133
        spec->dest.fd.qemu = virNetSocketDupFD(sock, true);
4134
        virObjectUnref(sock);
4135 4136 4137 4138 4139
    }
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0 ||
        spec->dest.fd.qemu == -1)
        goto cleanup;

4140 4141 4142 4143 4144 4145 4146
    /* 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;
    }

4147 4148
    ret = 0;

4149
 cleanup:
4150 4151 4152 4153 4154 4155
    VIR_FREE(port);
    if (ret < 0)
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
    return ret;
}

4156
static int
4157
qemuMigrationRun(virQEMUDriverPtr driver,
4158 4159 4160 4161 4162 4163 4164
                 virDomainObjPtr vm,
                 const char *cookiein,
                 int cookieinlen,
                 char **cookieout,
                 int *cookieoutlen,
                 unsigned long flags,
                 unsigned long resource,
4165
                 qemuMigrationSpecPtr spec,
4166
                 virConnectPtr dconn,
4167 4168 4169
                 const char *graphicsuri,
                 size_t nmigrate_disks,
                 const char **migrate_disks)
4170
{
4171
    int ret = -1;
4172 4173
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
4174
    qemuMigrationCookiePtr mig = NULL;
4175
    qemuMigrationIOThreadPtr iothread = NULL;
4176
    int fd = -1;
4177
    unsigned long migrate_speed = resource ? resource : priv->migMaxBandwidth;
4178
    virErrorPtr orig_err = NULL;
4179
    unsigned int cookieFlags = 0;
4180
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
4181
    bool events = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
4182
    int rc;
4183 4184 4185

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
4186 4187
              "spec=%p (dest=%d, fwd=%d), dconn=%p, graphicsuri=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p",
4188 4189
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
4190
              spec, spec->destType, spec->fwdType, dconn,
4191
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4192

4193 4194 4195 4196 4197 4198 4199 4200 4201 4202
    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;
    }

4203 4204
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
4205 4206 4207 4208
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Migration with lock driver %s requires"
                         " cookie support"),
                       virLockManagerPluginGetName(driver->lockManager));
4209 4210 4211
        return -1;
    }

4212 4213 4214
    if (events)
        priv->signalIOError = abort_on_error;

4215 4216 4217
    mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                 cookieFlags | QEMU_MIGRATION_COOKIE_GRAPHICS);
    if (!mig)
4218 4219
        goto cleanup;

4220
    if (qemuDomainMigrateGraphicsRelocate(driver, vm, mig, graphicsuri) < 0)
4221 4222
        VIR_WARN("unable to provide data for graphics client relocation");

4223 4224 4225 4226 4227 4228 4229
    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,
4230 4231
                                         &migrate_flags,
                                         nmigrate_disks,
4232 4233
                                         migrate_disks,
                                         dconn) < 0) {
4234 4235 4236 4237 4238 4239 4240 4241
                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.");
        }
4242 4243
    }

4244
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
4245 4246 4247 4248 4249 4250
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

4251 4252
    if (qemuMigrationSetCompression(driver, vm,
                                    flags & VIR_MIGRATE_COMPRESSED,
4253 4254 4255
                                    QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

4256 4257
    if (qemuMigrationSetAutoConverge(driver, vm,
                                     flags & VIR_MIGRATE_AUTO_CONVERGE,
4258 4259 4260
                                     QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

4261 4262
    if (qemuMigrationSetPinAll(driver, vm,
                               flags & VIR_MIGRATE_RDMA_PIN_ALL,
4263 4264 4265
                               QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;

4266 4267
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4268 4269
        goto cleanup;

4270
    if (priv->job.abortJob) {
4271 4272
        /* explicitly do this *after* we entered the monitor,
         * as this is a critical section so we are guaranteed
4273
         * priv->job.abortJob will not change */
4274
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
4275
        priv->job.current->type = VIR_DOMAIN_JOB_CANCELLED;
4276 4277 4278 4279 4280 4281
        virReportError(VIR_ERR_OPERATION_ABORTED, _("%s: %s"),
                       qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                       _("canceled by client"));
        goto cleanup;
    }

4282 4283
    if (qemuMonitorSetMigrationSpeed(priv->mon, migrate_speed) < 0)
        goto exit_monitor;
4284

4285 4286
    /* connect to the destination qemu if needed */
    if (spec->destType == MIGRATION_DEST_CONNECT_HOST &&
4287
        qemuMigrationConnect(driver, vm, spec) < 0) {
4288
        goto exit_monitor;
4289
    }
4290

4291 4292
    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
M
Michael R. Hines 已提交
4293 4294
        if (STREQ(spec->dest.host.protocol, "rdma") &&
            virProcessSetMaxMemLock(vm->pid, vm->def->mem.hard_limit << 10) < 0) {
4295
            goto exit_monitor;
M
Michael R. Hines 已提交
4296
        }
4297
        ret = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
4298
                                       spec->dest.host.protocol,
4299 4300 4301 4302
                                       spec->dest.host.name,
                                       spec->dest.host.port);
        break;

4303 4304 4305 4306
    case MIGRATION_DEST_CONNECT_HOST:
        /* handled above and transformed into MIGRATION_DEST_FD */
        break;

4307
    case MIGRATION_DEST_UNIX:
4308
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX)) {
4309 4310 4311 4312 4313 4314 4315 4316 4317
            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;
4318 4319

    case MIGRATION_DEST_FD:
4320
        if (spec->fwdType != MIGRATION_FWD_DIRECT) {
4321
            fd = spec->dest.fd.local;
4322 4323
            spec->dest.fd.local = -1;
        }
4324 4325 4326 4327
        ret = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                     spec->dest.fd.qemu);
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
4328
    }
4329 4330
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
4331
    if (ret < 0)
4332 4333
        goto cleanup;
    ret = -1;
4334 4335 4336 4337

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

4338 4339 4340 4341 4342
    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.
         */
4343
        if (qemuMigrationCheckJobStatus(driver, vm,
4344
                                        QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
4345
            goto cancel;
4346

4347 4348 4349 4350 4351 4352 4353
        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;
        }
4354 4355
    }

4356 4357
    if (spec->fwdType != MIGRATION_FWD_DIRECT &&
        !(iothread = qemuMigrationStartTunnel(spec->fwd.stream, fd)))
4358 4359
        goto cancel;

4360 4361
    rc = qemuMigrationWaitForCompletion(driver, vm,
                                        QEMU_ASYNC_JOB_MIGRATION_OUT,
4362
                                        dconn, abort_on_error, !!mig->nbd);
4363 4364 4365
    if (rc == -2)
        goto cancel;
    else if (rc == -1)
4366
        goto cleanup;
4367

4368 4369 4370 4371 4372 4373 4374 4375 4376
    /* 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) {
4377 4378
        if (qemuMigrationSetOffline(driver, vm) < 0) {
            priv->job.current->type = VIR_DOMAIN_JOB_FAILED;
4379
            goto cleanup;
4380
        }
4381 4382
    }

4383
    ret = 0;
4384

4385
 cleanup:
4386 4387 4388
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

4389
    /* cancel any outstanding NBD jobs */
4390
    if (mig && mig->nbd) {
4391
        if (qemuMigrationCancelDriveMirror(driver, vm, ret == 0,
4392 4393
                                           QEMU_ASYNC_JOB_MIGRATION_OUT,
                                           dconn) < 0)
4394 4395
            ret = -1;
    }
4396

4397
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
4398
        if (iothread && qemuMigrationStopTunnel(iothread, ret < 0) < 0)
4399 4400
            ret = -1;
    }
4401
    VIR_FORCE_CLOSE(fd);
4402

4403
    if (priv->job.completed) {
4404
        qemuDomainJobInfoUpdateTime(priv->job.completed);
4405
        qemuDomainJobInfoUpdateDowntime(priv->job.completed);
4406
        ignore_value(virTimeMillisNow(&priv->job.completed->sent));
4407
    }
4408

4409 4410 4411
    if (priv->job.current->type == VIR_DOMAIN_JOB_UNBOUNDED)
        priv->job.current->type = VIR_DOMAIN_JOB_FAILED;

4412 4413
    cookieFlags |= QEMU_MIGRATION_COOKIE_NETWORK |
                   QEMU_MIGRATION_COOKIE_STATS;
4414 4415
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookieFlags |= QEMU_MIGRATION_COOKIE_PERSISTENT;
4416
    if (ret == 0 &&
4417 4418
        qemuMigrationBakeCookie(mig, driver, vm, cookieout,
                                cookieoutlen, cookieFlags) < 0) {
4419
        VIR_WARN("Unable to encode migration cookie");
4420
    }
4421

4422 4423
    qemuMigrationCookieFree(mig);

4424 4425 4426
    if (events)
        priv->signalIOError = false;

4427 4428 4429 4430 4431
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

4432 4433
    return ret;

4434 4435 4436 4437
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;

4438
 cancel:
4439 4440
    orig_err = virSaveLastError();

4441
    if (virDomainObjIsActive(vm)) {
4442 4443
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
4444
            qemuMonitorMigrateCancel(priv->mon);
4445
            ignore_value(qemuDomainObjExitMonitor(driver, vm));
4446
        }
4447
    }
4448 4449 4450
    goto cleanup;
}

4451
/* Perform migration using QEMU's native migrate support,
4452 4453
 * not encrypted obviously
 */
4454
static int doNativeMigrate(virQEMUDriverPtr driver,
4455 4456 4457 4458 4459 4460 4461
                           virDomainObjPtr vm,
                           const char *uri,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4462
                           unsigned long resource,
4463
                           virConnectPtr dconn,
4464 4465 4466
                           const char *graphicsuri,
                           size_t nmigrate_disks,
                           const char **migrate_disks)
4467
{
4468
    qemuDomainObjPrivatePtr priv = vm->privateData;
M
Martin Kletzander 已提交
4469
    virURIPtr uribits = NULL;
4470
    int ret = -1;
4471 4472 4473
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
4474
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
4475
              "graphicsuri=%s, nmigrate_disks=%zu migrate_disks=%p",
4476
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
4477
              cookieout, cookieoutlen, flags, resource,
4478
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4479

4480
    if (!(uribits = qemuMigrationParseURI(uri, NULL)))
4481 4482
        return -1;

4483 4484 4485 4486 4487 4488 4489
    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
4490 4491 4492 4493 4494
    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"));
4495
            goto cleanup;
M
Michael R. Hines 已提交
4496
        }
4497
        if (!virMemoryLimitIsSet(vm->def->mem.hard_limit)) {
M
Michael R. Hines 已提交
4498 4499 4500
            virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                           _("cannot start RDMA migration with no memory hard "
                             "limit set"));
4501
            goto cleanup;
M
Michael R. Hines 已提交
4502 4503 4504 4505 4506
        }
    }

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
        STRNEQ(uribits->scheme, "rdma"))
4507 4508
        spec.destType = MIGRATION_DEST_CONNECT_HOST;
    else
4509
        spec.destType = MIGRATION_DEST_HOST;
4510
    spec.dest.host.protocol = uribits->scheme;
4511 4512 4513
    spec.dest.host.name = uribits->server;
    spec.dest.host.port = uribits->port;
    spec.fwdType = MIGRATION_FWD_DIRECT;
4514

4515
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4516
                           cookieoutlen, flags, resource, &spec, dconn,
4517
                           graphicsuri, nmigrate_disks, migrate_disks);
4518 4519 4520 4521

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

4522
 cleanup:
4523
    virURIFree(uribits);
4524 4525 4526 4527 4528

    return ret;
}


4529
static int doTunnelMigrate(virQEMUDriverPtr driver,
4530 4531 4532 4533 4534 4535 4536
                           virDomainObjPtr vm,
                           virStreamPtr st,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
4537
                           unsigned long resource,
4538
                           virConnectPtr dconn,
4539 4540 4541
                           const char *graphicsuri,
                           size_t nmigrate_disks,
                           const char **migrate_disks)
4542 4543
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
4544
    virNetSocketPtr sock = NULL;
4545 4546
    int ret = -1;
    qemuMigrationSpec spec;
4547
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4548 4549

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
4550
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
4551
              "graphicsuri=%s, nmigrate_disks=%zu, migrate_disks=%p",
4552
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
4553
              cookieout, cookieoutlen, flags, resource,
4554
              NULLSTR(graphicsuri), nmigrate_disks, migrate_disks);
4555

4556 4557 4558
    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)) {
4559 4560
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Source qemu is too old to support tunnelled migration"));
4561
        virObjectUnref(cfg);
4562
        return -1;
4563 4564 4565 4566 4567
    }

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

4568
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD)) {
4569 4570 4571 4572 4573 4574
        int fds[2];

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

4575
        if (pipe2(fds, O_CLOEXEC) == 0) {
4576 4577 4578 4579
            spec.dest.fd.qemu = fds[1];
            spec.dest.fd.local = fds[0];
        }
        if (spec.dest.fd.qemu == -1 ||
4580
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
4581
                                              spec.dest.fd.qemu) < 0) {
4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592
            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",
4593
                        cfg->libDir, vm->def->name) < 0)
4594 4595 4596
            goto cleanup;

        if (virNetSocketNewListenUNIX(spec.dest.unix_socket.file, 0700,
4597
                                      cfg->user, cfg->group,
4598 4599 4600 4601 4602 4603 4604
                                      &sock) < 0 ||
            virNetSocketListen(sock, 1) < 0)
            goto cleanup;

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

4605
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
4606
                           cookieoutlen, flags, resource, &spec, dconn,
4607
                           graphicsuri, nmigrate_disks, migrate_disks);
4608

4609
 cleanup:
4610 4611 4612 4613
    if (spec.destType == MIGRATION_DEST_FD) {
        VIR_FORCE_CLOSE(spec.dest.fd.qemu);
        VIR_FORCE_CLOSE(spec.dest.fd.local);
    } else {
4614
        virObjectUnref(sock);
4615 4616
        VIR_FREE(spec.dest.unix_socket.file);
    }
4617

4618
    virObjectUnref(cfg);
4619 4620 4621 4622
    return ret;
}


4623 4624 4625 4626
/* 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 */
4627
static int doPeer2PeerMigrate2(virQEMUDriverPtr driver,
4628
                               virConnectPtr sconn ATTRIBUTE_UNUSED,
4629 4630
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
4631
                               const char *dconnuri,
4632 4633 4634
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
4635 4636 4637
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
4638
    char *cookie = NULL;
4639 4640 4641
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
4642
    bool cancelled;
4643
    virStreamPtr st = NULL;
4644
    unsigned long destflags;
4645

4646
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
4647
              "flags=%lx, dname=%s, resource=%lu",
4648 4649
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
4650

4651 4652 4653 4654 4655
    /* 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,
4656 4657
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS |
                                        VIR_DOMAIN_XML_MIGRATABLE)))
4658 4659 4660 4661 4662
        return -1;

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

4663 4664
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4665

4666 4667 4668 4669 4670 4671 4672 4673 4674 4675
    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;

4676
        qemuDomainObjEnterRemote(vm);
4677
        ret = dconn->driver->domainMigratePrepareTunnel
4678
            (dconn, st, destflags, dname, resource, dom_xml);
4679
        qemuDomainObjExitRemote(vm);
4680
    } else {
4681
        qemuDomainObjEnterRemote(vm);
4682 4683
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
4684
             destflags, dname, resource, dom_xml);
4685
        qemuDomainObjExitRemote(vm);
4686 4687 4688
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
4689 4690 4691
        goto cleanup;

    /* the domain may have shutdown or crashed while we had the locks dropped
4692
     * in qemuDomainObjEnterRemote, so check again
4693 4694
     */
    if (!virDomainObjIsActive(vm)) {
4695 4696
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
4697 4698 4699
        goto cleanup;
    }

4700 4701
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
4702 4703
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare2 did not set uri"));
4704
        cancelled = true;
4705
        orig_err = virSaveLastError();
4706
        goto finish;
4707 4708
    }

4709 4710 4711 4712
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
4713
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
4714
    if (flags & VIR_MIGRATE_TUNNELLED)
4715 4716
        ret = doTunnelMigrate(driver, vm, st,
                              NULL, 0, NULL, NULL,
4717 4718
                              flags, resource, dconn,
                              NULL, 0, NULL);
4719 4720 4721 4722
    else
        ret = doNativeMigrate(driver, vm, uri_out,
                              cookie, cookielen,
                              NULL, NULL, /* No out cookie with v2 migration */
4723
                              flags, resource, dconn, NULL, 0, NULL);
4724 4725 4726 4727

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
4728

4729 4730 4731
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4732
    cancelled = ret < 0;
4733

4734
 finish:
4735 4736 4737 4738
    /* 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.
     */
4739
    dname = dname ? dname : vm->def->name;
4740
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
4741
    qemuDomainObjEnterRemote(vm);
4742
    ddomain = dconn->driver->domainMigrateFinish2
4743
        (dconn, dname, cookie, cookielen,
4744
         uri_out ? uri_out : dconnuri, destflags, cancelled);
4745
    qemuDomainObjExitRemote(vm);
4746 4747
    if (cancelled && ddomain)
        VIR_ERROR(_("finish step ignored that migration was cancelled"));
4748

4749
 cleanup:
4750
    if (ddomain) {
4751
        virObjectUnref(ddomain);
4752 4753 4754 4755
        ret = 0;
    } else {
        ret = -1;
    }
4756

4757
    virObjectUnref(st);
4758 4759 4760 4761 4762 4763

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
4764
    VIR_FREE(cookie);
4765 4766

    return ret;
4767 4768 4769
}


4770 4771 4772 4773
/* 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 */
4774 4775 4776 4777 4778 4779 4780 4781 4782
static int
doPeer2PeerMigrate3(virQEMUDriverPtr driver,
                    virConnectPtr sconn,
                    virConnectPtr dconn,
                    const char *dconnuri,
                    virDomainObjPtr vm,
                    const char *xmlin,
                    const char *dname,
                    const char *uri,
4783
                    const char *graphicsuri,
4784
                    const char *listenAddress,
4785 4786
                    size_t nmigrate_disks,
                    const char **migrate_disks,
4787 4788 4789
                    unsigned long long bandwidth,
                    bool useParams,
                    unsigned long flags)
4790 4791 4792 4793 4794 4795 4796 4797 4798 4799
{
    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;
4800
    bool cancelled = true;
4801
    virStreamPtr st = NULL;
4802
    unsigned long destflags;
4803 4804 4805
    virTypedParameterPtr params = NULL;
    int nparams = 0;
    int maxparams = 0;
4806
    size_t i;
4807 4808

    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, dconnuri=%s, vm=%p, xmlin=%s, "
4809
              "dname=%s, uri=%s, graphicsuri=%s, listenAddress=%s, "
4810 4811
              "nmigrate_disks=%zu, migrate_disks=%p, bandwidth=%llu, "
              "useParams=%d, flags=%lx",
4812
              driver, sconn, dconn, NULLSTR(dconnuri), vm, NULLSTR(xmlin),
4813
              NULLSTR(dname), NULLSTR(uri), NULLSTR(graphicsuri),
4814 4815
              NULLSTR(listenAddress), nmigrate_disks, migrate_disks,
              bandwidth, useParams, flags);
4816

4817 4818 4819 4820 4821
    /* 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.  */

4822
    dom_xml = qemuMigrationBeginPhase(driver, vm, xmlin, dname,
4823 4824
                                      &cookieout, &cookieoutlen,
                                      nmigrate_disks, migrate_disks, flags);
4825 4826 4827
    if (!dom_xml)
        goto cleanup;

4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847
    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;
4848 4849 4850 4851 4852 4853

        if (graphicsuri &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_GRAPHICS_URI,
                                    graphicsuri) < 0)
            goto cleanup;
4854 4855 4856 4857 4858
        if (listenAddress &&
            virTypedParamsAddString(&params, &nparams, &maxparams,
                                    VIR_MIGRATE_PARAM_LISTEN_ADDRESS,
                                    listenAddress) < 0)
            goto cleanup;
4859 4860 4861 4862 4863
        for (i = 0; i < nmigrate_disks; i++)
            if (virTypedParamsAddString(&params, &nparams, &maxparams,
                                        VIR_MIGRATE_PARAM_MIGRATE_DISKS,
                                        migrate_disks[i]) < 0)
                goto cleanup;
4864 4865
    }

4866 4867 4868
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED)
        flags |= VIR_MIGRATE_PAUSED;

4869 4870
    destflags = flags & ~(VIR_MIGRATE_ABORT_ON_ERROR |
                          VIR_MIGRATE_AUTO_CONVERGE);
4871

4872 4873 4874 4875 4876 4877 4878 4879 4880
    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;

4881
        qemuDomainObjEnterRemote(vm);
4882 4883 4884 4885 4886 4887 4888 4889 4890
        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);
        }
4891
        qemuDomainObjExitRemote(vm);
4892
    } else {
4893
        qemuDomainObjEnterRemote(vm);
4894 4895 4896 4897 4898 4899 4900 4901 4902
        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);
        }
4903
        qemuDomainObjExitRemote(vm);
4904 4905 4906 4907 4908
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

L
liguang 已提交
4909 4910 4911 4912
    if (flags & VIR_MIGRATE_OFFLINE) {
        VIR_DEBUG("Offline migration, skipping Perform phase");
        VIR_FREE(cookieout);
        cookieoutlen = 0;
4913
        cancelled = false;
L
liguang 已提交
4914 4915 4916
        goto finish;
    }

4917 4918 4919 4920
    if (uri_out) {
        uri = uri_out;
        if (useParams &&
            virTypedParamsReplaceString(&params, &nparams,
4921 4922
                                        VIR_MIGRATE_PARAM_URI, uri_out) < 0) {
            orig_err = virSaveLastError();
4923
            goto finish;
4924
        }
4925
    } else if (!uri && !(flags & VIR_MIGRATE_TUNNELLED)) {
4926 4927
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("domainMigratePrepare3 did not set uri"));
4928
        orig_err = virSaveLastError();
4929 4930 4931 4932 4933 4934 4935 4936
        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.
     */
4937
    VIR_DEBUG("Perform3 %p uri=%s", sconn, NULLSTR(uri));
4938
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
4939 4940 4941 4942 4943
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
4944
    if (flags & VIR_MIGRATE_TUNNELLED) {
4945 4946 4947
        ret = doTunnelMigrate(driver, vm, st,
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4948 4949
                              flags, bandwidth, dconn, graphicsuri,
                              nmigrate_disks, migrate_disks);
4950 4951
    } else {
        ret = doNativeMigrate(driver, vm, uri,
4952 4953
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
4954 4955
                              flags, bandwidth, dconn, graphicsuri,
                              nmigrate_disks, migrate_disks);
4956
    }
4957 4958

    /* Perform failed. Make sure Finish doesn't overwrite the error */
4959
    if (ret < 0) {
4960
        orig_err = virSaveLastError();
4961 4962 4963 4964
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
4965 4966 4967 4968

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
4969
    cancelled = ret < 0;
4970

4971
 finish:
4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983
    /*
     * 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;
4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006

    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);
    }
5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026

    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();
5027 5028
                if (err &&
                    err->domain == VIR_FROM_QEMU &&
5029 5030 5031 5032 5033 5034 5035
                    err->code != VIR_ERR_MIGRATE_FINISH_OK) {
                    virFreeError(orig_err);
                    orig_err = NULL;
                }
            }
        }
    }
5036

5037 5038 5039 5040 5041 5042 5043
    /* 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.
5044
     */
5045
    cancelled = ddomain == NULL;
5046

5047 5048 5049 5050 5051 5052
    /* 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();

5053 5054 5055 5056
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
5057
    VIR_DEBUG("Confirm3 %p cancelled=%d vm=%p", sconn, cancelled, vm);
5058 5059 5060 5061 5062
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
5063 5064 5065
    ret = qemuMigrationConfirmPhase(driver, sconn, vm,
                                    cookiein, cookieinlen,
                                    flags, cancelled);
5066 5067 5068 5069
    /* 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.
     */
5070 5071 5072
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
5073 5074 5075

 cleanup:
    if (ddomain) {
5076
        virObjectUnref(ddomain);
5077 5078 5079 5080 5081
        ret = 0;
    } else {
        ret = -1;
    }

5082
    virObjectUnref(st);
5083 5084 5085 5086 5087 5088 5089 5090

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);
5091
    virTypedParamsFree(params, nparams);
5092 5093 5094 5095
    return ret;
}


5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107
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);
}


5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119
static int virConnectCredType[] = {
    VIR_CRED_AUTHNAME,
    VIR_CRED_PASSPHRASE,
};


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


5120
static int doPeer2PeerMigrate(virQEMUDriverPtr driver,
5121
                              virConnectPtr sconn,
5122
                              virDomainObjPtr vm,
5123
                              const char *xmlin,
5124
                              const char *dconnuri,
5125
                              const char *uri,
5126
                              const char *graphicsuri,
5127
                              const char *listenAddress,
5128 5129
                              size_t nmigrate_disks,
                              const char **migrate_disks,
5130 5131
                              unsigned long flags,
                              const char *dname,
5132 5133
                              unsigned long resource,
                              bool *v3proto)
5134 5135 5136 5137
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
5138
    virErrorPtr orig_err = NULL;
5139
    bool offline = false;
5140
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5141
    bool useParams;
5142

5143 5144 5145
    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",
5146
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
5147
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
5148
              nmigrate_disks, migrate_disks, flags, NULLSTR(dname), resource);
5149

5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163
    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;
    }

5164 5165 5166 5167
    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

5168
    qemuDomainObjEnterRemote(vm);
5169
    dconn = virConnectOpenAuth(dconnuri, &virConnectAuthConfig, 0);
5170
    qemuDomainObjExitRemote(vm);
5171
    if (dconn == NULL) {
5172
        virReportError(VIR_ERR_OPERATION_FAILED,
5173 5174
                       _("Failed to connect to remote libvirt URI %s: %s"),
                       dconnuri, virGetLastErrorMessage());
5175
        virObjectUnref(cfg);
5176 5177 5178
        return -1;
    }

5179 5180
    if (virConnectSetKeepAlive(dconn, cfg->keepAliveInterval,
                               cfg->keepAliveCount) < 0)
5181 5182
        goto cleanup;

5183 5184 5185 5186 5187
    if (virConnectRegisterCloseCallback(dconn, qemuMigrationConnectionClosed,
                                        vm, NULL) < 0) {
        goto cleanup;
    }

5188
    qemuDomainObjEnterRemote(vm);
5189 5190
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
5191
        /* v3proto reflects whether the caller used Perform3, but with
5192
         * p2p migrate, regardless of whether Perform2 or Perform3
5193 5194 5195 5196
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
5197 5198
    useParams = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                         VIR_DRV_FEATURE_MIGRATION_PARAMS);
L
liguang 已提交
5199 5200 5201
    if (flags & VIR_MIGRATE_OFFLINE)
        offline = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                           VIR_DRV_FEATURE_MIGRATION_OFFLINE);
5202
    qemuDomainObjExitRemote(vm);
5203

5204
    if (!p2p) {
5205 5206
        virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                       _("Destination libvirt does not support peer-to-peer migration protocol"));
5207 5208 5209
        goto cleanup;
    }

5210 5211
    /* Only xmlin, dname, uri, and bandwidth parameters can be used with
     * old-style APIs. */
5212
    if (!useParams && (graphicsuri || listenAddress || nmigrate_disks)) {
5213 5214 5215 5216 5217 5218
        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
5219 5220 5221 5222 5223 5224 5225
    if (flags & VIR_MIGRATE_OFFLINE && !offline) {
        virReportError(VIR_ERR_ARGUMENT_UNSUPPORTED, "%s",
                       _("offline migration is not supported by "
                         "the destination host"));
        goto cleanup;
    }

5226
    /* domain may have been stopped while we were talking to remote daemon */
L
liguang 已提交
5227
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
5228 5229
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
5230 5231 5232
        goto cleanup;
    }

5233 5234 5235 5236 5237 5238 5239
    /* 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;

5240 5241
    if (*v3proto) {
        ret = doPeer2PeerMigrate3(driver, sconn, dconn, dconnuri, vm, xmlin,
5242
                                  dname, uri, graphicsuri, listenAddress,
5243 5244
                                  nmigrate_disks, migrate_disks, resource,
                                  useParams, flags);
5245
    } else {
5246
        ret = doPeer2PeerMigrate2(driver, sconn, dconn, vm,
5247
                                  dconnuri, flags, dname, resource);
5248
    }
5249

5250
 cleanup:
5251
    orig_err = virSaveLastError();
5252
    qemuDomainObjEnterRemote(vm);
5253
    virConnectUnregisterCloseCallback(dconn, qemuMigrationConnectionClosed);
5254
    virObjectUnref(dconn);
5255
    qemuDomainObjExitRemote(vm);
5256 5257 5258 5259
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5260
    virObjectUnref(cfg);
5261 5262 5263 5264
    return ret;
}


5265 5266 5267 5268 5269 5270
/*
 * 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
5271
qemuMigrationPerformJob(virQEMUDriverPtr driver,
5272 5273 5274 5275 5276
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dconnuri,
                        const char *uri,
5277
                        const char *graphicsuri,
5278
                        const char *listenAddress,
5279 5280
                        size_t nmigrate_disks,
                        const char **migrate_disks,
5281
                        const char *cookiein,
5282 5283 5284 5285 5286 5287 5288
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
5289
{
5290
    virObjectEventPtr event = NULL;
5291
    int ret = -1;
5292
    virErrorPtr orig_err = NULL;
5293
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5294
    bool abort_on_error = !!(flags & VIR_MIGRATE_ABORT_ON_ERROR);
5295

5296
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
5297 5298
        goto cleanup;

L
liguang 已提交
5299
    if (!virDomainObjIsActive(vm) && !(flags & VIR_MIGRATE_OFFLINE)) {
5300 5301
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("domain is not running"));
5302 5303 5304
        goto endjob;
    }

5305
    if (!qemuMigrationIsAllowed(driver, vm, NULL, true, abort_on_error))
5306
        goto endjob;
5307

5308 5309
    if (!(flags & VIR_MIGRATE_UNSAFE) &&
        !qemuMigrationIsSafe(vm->def, nmigrate_disks, migrate_disks))
5310
        goto endjob;
5311

5312
    qemuMigrationStoreDomainState(vm);
5313 5314

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
5315
        ret = doPeer2PeerMigrate(driver, conn, vm, xmlin,
5316
                                 dconnuri, uri, graphicsuri, listenAddress,
5317
                                 nmigrate_disks, migrate_disks,
5318
                                 flags, dname, resource, &v3proto);
5319
    } else {
5320 5321 5322
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
        ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                              cookieout, cookieoutlen,
5323
                              flags, resource, NULL, NULL, 0, NULL);
5324
    }
5325 5326
    if (ret < 0)
        goto endjob;
5327

5328 5329 5330 5331
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
5332
    if (!v3proto) {
5333 5334
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_MIGRATED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
5335
        virDomainAuditStop(vm, "migrated");
5336
        event = virDomainEventLifecycleNewFromObj(vm,
5337 5338
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
5339 5340
    }

5341
 endjob:
5342 5343 5344
    if (ret < 0)
        orig_err = virSaveLastError();

5345
    if (qemuMigrationRestoreDomainState(conn, vm)) {
5346
        event = virDomainEventLifecycleNewFromObj(vm,
5347 5348 5349
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
5350

5351 5352 5353 5354
    qemuMigrationJobFinish(driver, vm);
    if (!virDomainObjIsActive(vm) &&
        (!vm->persistent ||
         (ret == 0 && (flags & VIR_MIGRATE_UNDEFINE_SOURCE)))) {
5355
        if (flags & VIR_MIGRATE_UNDEFINE_SOURCE)
5356
            virDomainDeleteConfig(cfg->configDir, cfg->autostartDir, vm);
5357
        qemuDomainRemoveInactive(driver, vm);
5358 5359
    }

5360 5361 5362 5363 5364
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }

5365
 cleanup:
M
Michal Privoznik 已提交
5366
    virDomainObjEndAPI(&vm);
5367 5368
    if (event)
        qemuDomainEventQueue(driver, event);
5369
    virObjectUnref(cfg);
5370 5371 5372 5373 5374 5375 5376
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
5377
qemuMigrationPerformPhase(virQEMUDriverPtr driver,
5378 5379 5380
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *uri,
5381
                          const char *graphicsuri,
5382 5383
                          size_t nmigrate_disks,
                          const char **migrate_disks,
5384 5385 5386 5387 5388 5389 5390
                          const char *cookiein,
                          int cookieinlen,
                          char **cookieout,
                          int *cookieoutlen,
                          unsigned long flags,
                          unsigned long resource)
{
5391
    virObjectEventPtr event = NULL;
5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402
    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);
5403 5404
    virCloseCallbacksUnset(driver->closeCallbacks, vm,
                           qemuMigrationCleanup);
5405 5406 5407

    ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                          cookieout, cookieoutlen,
5408 5409
                          flags, resource, NULL, graphicsuri,
                          nmigrate_disks, migrate_disks);
5410

5411 5412 5413 5414 5415
    if (ret < 0) {
        if (qemuMigrationRestoreDomainState(conn, vm)) {
            event = virDomainEventLifecycleNewFromObj(vm,
                                                      VIR_DOMAIN_EVENT_RESUMED,
                                                      VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
5416
        }
5417
        goto endjob;
5418
    }
5419 5420 5421

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

5422 5423
    if (virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                             qemuMigrationCleanup) < 0)
5424 5425
        goto endjob;

5426
 endjob:
5427
    if (ret < 0)
5428
        qemuMigrationJobFinish(driver, vm);
5429
    else
5430 5431
        qemuMigrationJobContinue(vm);
    if (!virDomainObjIsActive(vm) && !vm->persistent)
5432
        qemuDomainRemoveInactive(driver, vm);
5433

5434
 cleanup:
M
Michal Privoznik 已提交
5435
    virDomainObjEndAPI(&vm);
5436 5437 5438 5439 5440
    if (event)
        qemuDomainEventQueue(driver, event);
    return ret;
}

5441
int
5442
qemuMigrationPerform(virQEMUDriverPtr driver,
5443 5444 5445 5446 5447
                     virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *xmlin,
                     const char *dconnuri,
                     const char *uri,
5448
                     const char *graphicsuri,
5449
                     const char *listenAddress,
5450 5451
                     size_t nmigrate_disks,
                     const char **migrate_disks,
5452 5453 5454 5455 5456 5457 5458 5459 5460 5461
                     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, "
5462 5463
              "uri=%s, graphicsuri=%s, listenAddress=%s, "
              "nmigrate_disks=%zu, migrate_disks=%p, "
5464 5465
              "cookiein=%s, cookieinlen=%d, cookieout=%p, cookieoutlen=%p, "
              "flags=%lx, dname=%s, resource=%lu, v3proto=%d",
5466
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
5467
              NULLSTR(uri), NULLSTR(graphicsuri), NULLSTR(listenAddress),
5468 5469
              nmigrate_disks, migrate_disks, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, NULLSTR(dname), resource, v3proto);
5470 5471 5472

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
5473 5474
            virReportError(VIR_ERR_OPERATION_INVALID,
                           "%s", _("received unexpected cookie with P2P migration"));
5475 5476 5477 5478
            return -1;
        }

        return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri, uri,
5479
                                       graphicsuri, listenAddress,
5480
                                       nmigrate_disks, migrate_disks,
5481
                                       cookiein, cookieinlen,
5482 5483
                                       cookieout, cookieoutlen,
                                       flags, dname, resource, v3proto);
5484 5485
    } else {
        if (dconnuri) {
5486 5487
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
5488 5489 5490 5491 5492
            return -1;
        }

        if (v3proto) {
            return qemuMigrationPerformPhase(driver, conn, vm, uri,
5493
                                             graphicsuri,
5494
                                             nmigrate_disks, migrate_disks,
5495 5496
                                             cookiein, cookieinlen,
                                             cookieout, cookieoutlen,
5497
                                             flags, resource);
5498 5499
        } else {
            return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri,
5500
                                           uri, graphicsuri, listenAddress,
5501
                                           nmigrate_disks, migrate_disks,
5502
                                           cookiein, cookieinlen,
5503 5504 5505 5506 5507
                                           cookieout, cookieoutlen, flags,
                                           dname, resource, v3proto);
        }
    }
}
5508

5509
static int
5510 5511
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def)
{
5512
    size_t i;
5513 5514 5515 5516 5517
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
5518
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5519
            if (virNetDevVPortProfileAssociate(net->ifname,
5520
                                               virDomainNetGetActualVirtPortProfile(net),
5521
                                               &net->mac,
5522
                                               virDomainNetGetActualDirectDev(net),
5523
                                               -1,
5524
                                               def->uuid,
5525 5526
                                               VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH,
                                               false) < 0) {
5527 5528 5529
                virReportError(VIR_ERR_OPERATION_FAILED,
                               _("Port profile Associate failed for %s"),
                               net->ifname);
5530
                goto err_exit;
5531
            }
5532
            last_good_net = i;
5533
            VIR_DEBUG("Port profile Associate succeeded for %s", net->ifname);
5534

5535
            if (virNetDevMacVLanVPortProfileRegisterCallback(net->ifname, &net->mac,
5536 5537 5538 5539
                                                             virDomainNetGetActualDirectDev(net), def->uuid,
                                                             virDomainNetGetActualVirtPortProfile(net),
                                                             VIR_NETDEV_VPORT_PROFILE_OP_CREATE))
                goto err_exit;
5540 5541 5542
        }
    }

5543
    return 0;
5544

5545
 err_exit:
5546
    for (i = 0; last_good_net != -1 && i <= last_good_net; i++) {
5547
        net = def->nets[i];
5548
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
5549
            ignore_value(virNetDevVPortProfileDisassociate(net->ifname,
5550
                                                           virDomainNetGetActualVirtPortProfile(net),
5551
                                                           &net->mac,
5552
                                                           virDomainNetGetActualDirectDev(net),
5553
                                                           -1,
5554
                                                           VIR_NETDEV_VPORT_PROFILE_OP_MIGRATE_IN_FINISH));
5555 5556
        }
    }
5557
    return -1;
5558 5559 5560
}


J
Jiri Denemark 已提交
5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608
static int
qemuMigrationPersist(virQEMUDriverPtr driver,
                     virDomainObjPtr vm,
                     qemuMigrationCookiePtr mig)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    virDomainDefPtr vmdef;
    virObjectEventPtr event;
    bool newVM;
    int ret = -1;

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

    newVM = !vm->persistent;
    vm->persistent = 1;

    if (mig->persistent)
        vm->newDef = mig->persistent;

    vmdef = virDomainObjGetPersistentDef(caps, driver->xmlopt, vm);
    if (!vmdef) {
        if (newVM)
            vm->persistent = 0;
        goto cleanup;
    }

    if (virDomainSaveConfig(cfg->configDir, vmdef) < 0)
        goto cleanup;

    event = virDomainEventLifecycleNewFromObj(vm,
                                              VIR_DOMAIN_EVENT_DEFINED,
                                              newVM ?
                                              VIR_DOMAIN_EVENT_DEFINED_ADDED :
                                              VIR_DOMAIN_EVENT_DEFINED_UPDATED);
    if (event)
        qemuDomainEventQueue(driver, event);

    ret = 0;

 cleanup:
    virObjectUnref(caps);
    virObjectUnref(cfg);
    return ret;
}


5609
virDomainPtr
5610
qemuMigrationFinish(virQEMUDriverPtr driver,
5611 5612
                    virConnectPtr dconn,
                    virDomainObjPtr vm,
5613 5614 5615 5616
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
5617
                    unsigned long flags,
5618 5619
                    int retcode,
                    bool v3proto)
5620 5621
{
    virDomainPtr dom = NULL;
5622
    virObjectEventPtr event = NULL;
5623
    qemuMigrationCookiePtr mig = NULL;
5624
    virErrorPtr orig_err = NULL;
5625
    int cookie_flags = 0;
J
Jiri Denemark 已提交
5626
    qemuDomainObjPrivatePtr priv = vm->privateData;
5627
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5628
    unsigned short port;
5629

5630
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
5631
              "cookieout=%p, cookieoutlen=%p, flags=%lx, retcode=%d",
5632 5633
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
5634

5635 5636 5637
    port = priv->migrationPort;
    priv->migrationPort = 0;

5638 5639
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN)) {
        qemuMigrationErrorReport(driver, vm->def->name);
5640
        goto cleanup;
5641
    }
5642

5643 5644 5645
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
5646

5647
    qemuDomainCleanupRemove(vm, qemuMigrationPrepareCleanup);
5648
    VIR_FREE(priv->job.completed);
5649

5650
    cookie_flags = QEMU_MIGRATION_COOKIE_NETWORK |
5651 5652
                   QEMU_MIGRATION_COOKIE_STATS |
                   QEMU_MIGRATION_COOKIE_NBD;
5653 5654 5655 5656 5657
    if (flags & VIR_MIGRATE_PERSIST_DEST)
        cookie_flags |= QEMU_MIGRATION_COOKIE_PERSISTENT;

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein,
                                       cookieinlen, cookie_flags)))
5658
        goto endjob;
5659 5660 5661 5662

    /* Did the migration go as planned?  If yes, return the domain
     * object, but if no, clean up the empty qemu process.
     */
5663 5664 5665 5666 5667 5668 5669 5670
    if (flags & VIR_MIGRATE_OFFLINE) {
        if (retcode != 0 ||
            qemuMigrationPersist(driver, vm, mig) < 0)
            goto endjob;

        dom = virGetDomain(dconn, vm->def->name, vm->def->uuid);
    } else if (retcode == 0) {
        if (!virDomainObjIsActive(vm)) {
5671 5672
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("guest unexpectedly quit"));
5673
            qemuMigrationErrorReport(driver, vm->def->name);
5674
            goto endjob;
5675 5676
        }

5677
        if (mig->jobInfo) {
5678 5679
            qemuDomainJobInfoPtr jobInfo = mig->jobInfo;
            priv->job.completed = jobInfo;
5680
            mig->jobInfo = NULL;
5681 5682 5683 5684
            if (jobInfo->sent && virTimeMillisNow(&jobInfo->received) == 0) {
                jobInfo->timeDelta = jobInfo->received - jobInfo->sent;
                jobInfo->timeDeltaSet = true;
            }
5685 5686
        }

5687 5688 5689 5690 5691 5692 5693 5694
        if (qemuMigrationVPAssociatePortProfiles(vm->def) < 0) {
            qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                            VIR_QEMU_PROCESS_STOP_MIGRATED);
            virDomainAuditStop(vm, "failed");
            event = virDomainEventLifecycleNewFromObj(vm,
                                             VIR_DOMAIN_EVENT_STOPPED,
                                             VIR_DOMAIN_EVENT_STOPPED_FAILED);
            goto endjob;
5695
        }
5696 5697
        if (mig->network && qemuDomainMigrateOPDRelocate(driver, vm, mig) < 0)
            VIR_WARN("unable to provide network data for relocation");
5698

5699 5700
        if (qemuMigrationStopNBDServer(driver, vm, mig) < 0)
            goto endjob;
5701

5702
        if (flags & VIR_MIGRATE_PERSIST_DEST) {
J
Jiri Denemark 已提交
5703
            if (qemuMigrationPersist(driver, vm, mig) < 0) {
5704 5705 5706 5707 5708 5709 5710 5711 5712
                /* Hmpf.  Migration was successful, but making it persistent
                 * was not.  If we report successful, then when this domain
                 * shuts down, management tools are in for a surprise.  On the
                 * other hand, if we report failure, then the management tools
                 * might try to restart the domain on the source side, even
                 * though the domain is actually running on the destination.
                 * Return a NULL dom pointer, and hope that this is a rare
                 * situation and management tools are smart.
                 */
5713 5714

                /*
5715 5716
                 * However, in v3 protocol, the source VM is still available
                 * to restart during confirm() step, so we kill it off now.
5717 5718
                 */
                if (v3proto) {
5719 5720 5721
                    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                    VIR_QEMU_PROCESS_STOP_MIGRATED);
                    virDomainAuditStop(vm, "failed");
5722
                }
5723 5724 5725 5726
                goto endjob;
            }
        }

5727
        if (!(flags & VIR_MIGRATE_PAUSED)) {
5728 5729 5730 5731
            /* 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
5732
            if (qemuProcessStartCPUs(driver, vm, dconn,
5733 5734
                                     VIR_DOMAIN_RUNNING_MIGRATED,
                                     QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
5735
                if (virGetLastError() == NULL)
5736 5737
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   "%s", _("resume operation failed"));
5738 5739 5740 5741 5742
                /* Need to save the current error, in case shutting
                 * down the process overwrites it
                 */
                orig_err = virSaveLastError();

5743 5744 5745 5746 5747 5748 5749 5750 5751
                /*
                 * In v3 protocol, the source VM is still available to
                 * restart during confirm() step, so we kill it off
                 * now.
                 * In v2 protocol, the source is dead, so we leave
                 * target in paused state, in case admin can fix
                 * things up
                 */
                if (v3proto) {
5752 5753
                    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                                    VIR_QEMU_PROCESS_STOP_MIGRATED);
5754
                    virDomainAuditStop(vm, "failed");
5755
                    event = virDomainEventLifecycleNewFromObj(vm,
5756 5757 5758
                                                     VIR_DOMAIN_EVENT_STOPPED,
                                                     VIR_DOMAIN_EVENT_STOPPED_FAILED);
                }
5759 5760
                goto endjob;
            }
5761 5762 5763 5764
            if (priv->job.completed) {
                qemuDomainJobInfoUpdateTime(priv->job.completed);
                qemuDomainJobInfoUpdateDowntime(priv->job.completed);
            }
5765 5766
        }

5767
        dom = virGetDomain(dconn, vm->def->name, vm->def->uuid);
5768

5769 5770 5771 5772 5773 5774 5775 5776
        event = virDomainEventLifecycleNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
            virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                                 VIR_DOMAIN_PAUSED_USER);
            if (event)
                qemuDomainEventQueue(driver, event);
5777
            event = virDomainEventLifecycleNewFromObj(vm,
5778 5779
                                             VIR_DOMAIN_EVENT_SUSPENDED,
                                             VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);
5780
        }
L
liguang 已提交
5781 5782

        if (virDomainObjIsActive(vm) &&
5783
            virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
5784 5785 5786
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto endjob;
        }
5787 5788 5789

        /* Guest is successfully running, so cancel previous auto destroy */
        qemuProcessAutoDestroyRemove(driver, vm);
5790
    } else {
5791 5792 5793 5794 5795 5796 5797 5798 5799
        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);

5800 5801
        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED,
                        VIR_QEMU_PROCESS_STOP_MIGRATED);
5802
        virDomainAuditStop(vm, "failed");
5803
        event = virDomainEventLifecycleNewFromObj(vm,
5804 5805 5806 5807
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_FAILED);
    }

5808 5809 5810
 endjob:
    if (dom &&
        qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen,
5811
                                QEMU_MIGRATION_COOKIE_STATS) < 0)
5812 5813
        VIR_WARN("Unable to encode migration cookie");

5814 5815
    qemuMigrationJobFinish(driver, vm);
    if (!vm->persistent && !virDomainObjIsActive(vm))
5816
        qemuDomainRemoveInactive(driver, vm);
5817

5818
 cleanup:
5819
    virPortAllocatorRelease(driver->migrationPorts, port);
5820 5821 5822
    if (priv->mon)
        qemuMonitorSetDomainLog(priv->mon, -1);
    VIR_FREE(priv->origname);
M
Michal Privoznik 已提交
5823
    virDomainObjEndAPI(&vm);
5824 5825
    if (event)
        qemuDomainEventQueue(driver, event);
5826
    qemuMigrationCookieFree(mig);
5827 5828 5829 5830
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5831
    virObjectUnref(cfg);
5832 5833 5834 5835 5836 5837

    /* 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);
5838 5839
    return dom;
}
5840

5841

5842
/* Helper function called while vm is active.  */
5843
int
5844
qemuMigrationToFile(virQEMUDriverPtr driver, virDomainObjPtr vm,
5845 5846
                    int fd, off_t offset, const char *path,
                    const char *compressor,
E
Eric Blake 已提交
5847
                    bool bypassSecurityDriver,
5848
                    qemuDomainAsyncJob asyncJob)
5849 5850 5851
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int rc;
5852
    int ret = -1;
5853
    bool restoreLabel = false;
5854 5855
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
5856
    unsigned long saveMigBandwidth = priv->migMaxBandwidth;
5857
    char *errbuf = NULL;
5858
    virErrorPtr orig_err = NULL;
5859 5860 5861 5862 5863

    /* 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,
5864 5865
                                     QEMU_DOMAIN_MIG_BANDWIDTH_MAX);
        priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
5866 5867
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
5868
    }
5869

5870 5871 5872 5873 5874 5875 5876
    if (!virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
        /* nothing to tear down */
        return -1;
    }

5877
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5878
        (!compressor || pipe(pipeFD) == 0)) {
5879
        /* All right! We can use fd migration, which means that qemu
5880 5881 5882
         * 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.  */
5883
        if (virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def,
5884
                                              compressor ? pipeFD[1] : fd) < 0)
5885
            goto cleanup;
5886 5887 5888
        bypassSecurityDriver = true;
    } else {
        /* Phooey - we have to fall back on exec migration, where qemu
E
Eric Blake 已提交
5889 5890
         * has to popen() the file by name, and block devices have to be
         * given cgroup ACL permission.  We might also stumble on
5891 5892
         * a race present in some qemu versions where it does a wait()
         * that botches pclose.  */
5893 5894
        if (virCgroupHasController(priv->cgroup,
                                   VIR_CGROUP_CONTROLLER_DEVICES)) {
5895 5896 5897 5898
            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
5899
                /* path was not a device, no further need for cgroup */
5900
            } else if (rv < 0) {
5901 5902
                goto cleanup;
            }
5903
        }
5904 5905
        if ((!bypassSecurityDriver) &&
            virSecurityManagerSetSavedStateLabel(driver->securityManager,
5906
                                                 vm->def, path) < 0)
5907
            goto cleanup;
5908
        restoreLabel = true;
5909 5910
    }

5911
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
5912 5913
        goto cleanup;

5914 5915 5916
    if (!compressor) {
        const char *args[] = { "cat", NULL };

5917
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933
            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
        };
5934 5935 5936 5937
        if (pipeFD[0] != -1) {
            cmd = virCommandNewArgs(args);
            virCommandSetInputFD(cmd, pipeFD[0]);
            virCommandSetOutputFD(cmd, &fd);
5938 5939
            virCommandSetErrorBuffer(cmd, &errbuf);
            virCommandDoAsyncIO(cmd);
5940 5941 5942
            if (virSetCloseExec(pipeFD[1]) < 0) {
                virReportSystemError(errno, "%s",
                                     _("Unable to set cloexec flag"));
5943
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5944 5945 5946
                goto cleanup;
            }
            if (virCommandRunAsync(cmd, NULL) < 0) {
5947
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5948 5949 5950 5951 5952 5953 5954
                goto cleanup;
            }
            rc = qemuMonitorMigrateToFd(priv->mon,
                                        QEMU_MONITOR_MIGRATE_BACKGROUND,
                                        pipeFD[1]);
            if (VIR_CLOSE(pipeFD[0]) < 0 ||
                VIR_CLOSE(pipeFD[1]) < 0)
5955
                VIR_WARN("failed to close intermediate pipe");
5956 5957 5958 5959 5960
        } else {
            rc = qemuMonitorMigrateToFile(priv->mon,
                                          QEMU_MONITOR_MIGRATE_BACKGROUND,
                                          args, path, offset);
        }
5961
    }
5962
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5963
        goto cleanup;
5964 5965 5966
    if (rc < 0)
        goto cleanup;

5967 5968
    rc = qemuMigrationWaitForCompletion(driver, vm, asyncJob,
                                        NULL, false, false);
5969

5970 5971 5972 5973
    if (rc < 0) {
        if (rc == -2) {
            orig_err = virSaveLastError();
            virCommandAbort(cmd);
5974 5975
            if (virDomainObjIsActive(vm) &&
                qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5976
                qemuMonitorMigrateCancel(priv->mon);
5977
                ignore_value(qemuDomainObjExitMonitor(driver, vm));
5978 5979
            }
        }
5980
        goto cleanup;
5981
    }
5982

5983 5984 5985
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

5986 5987
    ret = 0;

5988
 cleanup:
5989 5990 5991
    if (ret < 0 && !orig_err)
        orig_err = virSaveLastError();

5992
    /* Restore max migration bandwidth */
5993 5994
    if (virDomainObjIsActive(vm) &&
        qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
5995 5996
        qemuMonitorSetMigrationSpeed(priv->mon, saveMigBandwidth);
        priv->migMaxBandwidth = saveMigBandwidth;
5997
        ignore_value(qemuDomainObjExitMonitor(driver, vm));
5998 5999
    }

6000 6001
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
6002 6003 6004 6005 6006
    if (cmd) {
        VIR_DEBUG("Compression binary stderr: %s", NULLSTR(errbuf));
        VIR_FREE(errbuf);
        virCommandFree(cmd);
    }
6007 6008
    if (restoreLabel && (!bypassSecurityDriver) &&
        virSecurityManagerRestoreSavedStateLabel(driver->securityManager,
6009
                                                 vm->def, path) < 0)
6010 6011
        VIR_WARN("failed to restore save state label on %s", path);

6012 6013
    if (virCgroupHasController(priv->cgroup,
                               VIR_CGROUP_CONTROLLER_DEVICES)) {
6014 6015 6016
        int rv = virCgroupDenyDevicePath(priv->cgroup, path,
                                         VIR_CGROUP_DEVICE_RWM);
        virDomainAuditCgroupPath(vm, priv->cgroup, "deny", path, "rwm", rv == 0);
6017
    }
6018 6019 6020 6021 6022 6023

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

6024 6025
    return ret;
}
6026

6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079

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,
6080
                                       QEMU_ASYNC_JOB_NONE, NULL) < 0)
6081 6082 6083 6084 6085 6086 6087 6088 6089 6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104
        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;
}


6105
int
6106
qemuMigrationJobStart(virQEMUDriverPtr driver,
6107
                      virDomainObjPtr vm,
6108
                      qemuDomainAsyncJob job)
6109 6110 6111
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

6112
    if (qemuDomainObjBeginAsyncJob(driver, vm, job) < 0)
6113 6114
        return -1;

6115
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
6116
        qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
6117
    } else {
6118 6119 6120
        qemuDomainObjSetAsyncJobMask(vm, (QEMU_JOB_DEFAULT_MASK |
                                          JOB_MASK(QEMU_JOB_SUSPEND) |
                                          JOB_MASK(QEMU_JOB_MIGRATION_OP)));
6121
    }
6122

J
Jiri Denemark 已提交
6123
    priv->job.current->type = VIR_DOMAIN_JOB_UNBOUNDED;
6124 6125 6126 6127 6128

    return 0;
}

void
6129
qemuMigrationJobSetPhase(virQEMUDriverPtr driver,
6130
                         virDomainObjPtr vm,
6131
                         qemuMigrationJobPhase phase)
6132 6133 6134 6135 6136 6137 6138 6139 6140 6141 6142 6143 6144 6145
{
    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
6146
qemuMigrationJobStartPhase(virQEMUDriverPtr driver,
6147
                           virDomainObjPtr vm,
6148
                           qemuMigrationJobPhase phase)
6149 6150 6151 6152
{
    qemuMigrationJobSetPhase(driver, vm, phase);
}

6153
void
6154 6155
qemuMigrationJobContinue(virDomainObjPtr vm)
{
6156
    qemuDomainObjReleaseAsyncJob(vm);
6157 6158 6159 6160
}

bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
6161
                         qemuDomainAsyncJob job)
6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172
{
    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");

6173
        virReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
6174 6175 6176 6177 6178
        return false;
    }
    return true;
}

6179
void
6180
qemuMigrationJobFinish(virQEMUDriverPtr driver, virDomainObjPtr vm)
6181
{
6182
    qemuDomainObjEndAsyncJob(driver, vm);
6183
}
6184 6185 6186 6187 6188 6189 6190 6191 6192 6193 6194 6195 6196 6197 6198 6199 6200 6201 6202 6203 6204 6205 6206 6207 6208 6209 6210 6211 6212 6213 6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231 6232 6233 6234 6235 6236 6237


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);
}