qemu_migration.c 96.2 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
/*
 * qemu_migration.c: QEMU migration handling
 *
 * Copyright (C) 2006-2011 Red Hat, Inc.
 *
 * 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
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307  USA
 *
 */

#include <config.h>

#include <sys/time.h>
25 26
#include <gnutls/gnutls.h>
#include <gnutls/x509.h>
27
#include <fcntl.h>
28 29 30 31 32 33

#include "qemu_migration.h"
#include "qemu_monitor.h"
#include "qemu_domain.h"
#include "qemu_process.h"
#include "qemu_capabilities.h"
34
#include "qemu_cgroup.h"
35

36
#include "domain_audit.h"
37 38 39 40
#include "logging.h"
#include "virterror_internal.h"
#include "memory.h"
#include "util.h"
E
Eric Blake 已提交
41
#include "virfile.h"
42 43
#include "datatypes.h"
#include "fdstream.h"
44
#include "uuid.h"
45
#include "locking/domain_lock.h"
46
#include "rpc/virnetsocket.h"
47

48 49 50

#define VIR_FROM_THIS VIR_FROM_QEMU

51 52 53 54 55 56 57 58 59 60 61 62 63
VIR_ENUM_IMPL(qemuMigrationJobPhase, QEMU_MIGRATION_PHASE_LAST,
              "none",
              "perform2",
              "begin3",
              "perform3",
              "perform3_done",
              "confirm3_cancelled",
              "confirm3",
              "prepare",
              "finish2",
              "finish3",
);

64
enum qemuMigrationCookieFlags {
65
    QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS,
66
    QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE,
67 68 69 70 71 72 73

    QEMU_MIGRATION_COOKIE_FLAG_LAST
};

VIR_ENUM_DECL(qemuMigrationCookieFlag);
VIR_ENUM_IMPL(qemuMigrationCookieFlag,
              QEMU_MIGRATION_COOKIE_FLAG_LAST,
74
              "graphics", "lockstate");
75 76 77

enum qemuMigrationCookieFeatures {
    QEMU_MIGRATION_COOKIE_GRAPHICS  = (1 << QEMU_MIGRATION_COOKIE_FLAG_GRAPHICS),
78
    QEMU_MIGRATION_COOKIE_LOCKSTATE = (1 << QEMU_MIGRATION_COOKIE_FLAG_LOCKSTATE),
79 80 81 82 83 84 85 86 87 88 89 90
};

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

91 92 93
typedef struct _qemuMigrationCookie qemuMigrationCookie;
typedef qemuMigrationCookie *qemuMigrationCookiePtr;
struct _qemuMigrationCookie {
E
Eric Blake 已提交
94 95
    unsigned int flags;
    unsigned int flagsMandatory;
96 97

    /* Host properties */
98 99 100 101
    unsigned char localHostuuid[VIR_UUID_BUFLEN];
    unsigned char remoteHostuuid[VIR_UUID_BUFLEN];
    char *localHostname;
    char *remoteHostname;
102 103 104 105

    /* Guest properties */
    unsigned char uuid[VIR_UUID_BUFLEN];
    char *name;
106

107 108 109 110
    /* If (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) */
    char *lockState;
    char *lockDriver;

111 112
    /* If (flags & QEMU_MIGRATION_COOKIE_GRAPHICS) */
    qemuMigrationCookieGraphicsPtr graphics;
113 114
};

115 116 117 118 119 120 121 122 123
static void qemuMigrationCookieGraphicsFree(qemuMigrationCookieGraphicsPtr grap)
{
    if (!grap)
        return;
    VIR_FREE(grap->listen);
    VIR_FREE(grap->tlsSubject);
    VIR_FREE(grap);
}

124 125 126 127 128 129

static void qemuMigrationCookieFree(qemuMigrationCookiePtr mig)
{
    if (!mig)
        return;

130 131 132
    if (mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS)
        qemuMigrationCookieGraphicsFree(mig->graphics);

133 134
    VIR_FREE(mig->localHostname);
    VIR_FREE(mig->remoteHostname);
135
    VIR_FREE(mig->name);
136 137
    VIR_FREE(mig->lockState);
    VIR_FREE(mig->lockDriver);
138 139 140 141
    VIR_FREE(mig);
}


142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214
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)
        goto no_memory;

    if (virFileReadAll(certfile, 8192, &pemdata) < 0) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("unable to read server cert %s"), certfile);
        goto error;
    }

    ret = gnutls_x509_crt_init(&cert);
    if (ret < 0) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("cannot initialize cert object: %s"),
                        gnutls_strerror(ret));
        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) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("cannot load cert data from %s: %s"),
                        certfile, gnutls_strerror(ret));
        goto error;
    }

    subjectlen = 1024;
    if (VIR_ALLOC_N(subject, subjectlen+1) < 0)
        goto no_memory;

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

    VIR_FREE(certfile);
    VIR_FREE(pemdata);

    return subject;

no_memory:
    virReportOOMError();
error:
    VIR_FREE(certfile);
    VIR_FREE(pemdata);
    return NULL;
}


static qemuMigrationCookieGraphicsPtr
qemuMigrationCookieGraphicsAlloc(struct qemud_driver *driver,
                                 virDomainGraphicsDefPtr def)
{
    qemuMigrationCookieGraphicsPtr mig = NULL;
    const char *listenAddr;

    if (VIR_ALLOC(mig) < 0)
        goto no_memory;

    mig->type = def->type;
    if (mig->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
        mig->port = def->data.vnc.port;
215
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
216 217 218 219 220 221 222 223 224 225 226 227
        if (!listenAddr)
            listenAddr = driver->vncListen;

        if (driver->vncTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(driver->vncTLSx509certdir)))
            goto error;
    } else {
        mig->port = def->data.spice.port;
        if (driver->spiceTLS)
            mig->tlsPort = def->data.spice.tlsPort;
        else
            mig->tlsPort = -1;
228
        listenAddr = virDomainGraphicsListenGetAddress(def, 0);
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248
        if (!listenAddr)
            listenAddr = driver->spiceListen;

        if (driver->spiceTLS &&
            !(mig->tlsSubject = qemuDomainExtractTLSSubject(driver->spiceTLSx509certdir)))
            goto error;
    }
    if (!(mig->listen = strdup(listenAddr)))
        goto no_memory;

    return mig;

no_memory:
    virReportOOMError();
error:
    qemuMigrationCookieGraphicsFree(mig);
    return NULL;
}


249 250 251 252 253 254 255 256 257 258 259 260
static qemuMigrationCookiePtr
qemuMigrationCookieNew(virDomainObjPtr dom)
{
    qemuMigrationCookiePtr mig = NULL;

    if (VIR_ALLOC(mig) < 0)
        goto no_memory;

    if (!(mig->name = strdup(dom->def->name)))
        goto no_memory;
    memcpy(mig->uuid, dom->def->uuid, VIR_UUID_BUFLEN);

261
    if (!(mig->localHostname = virGetHostname(NULL)))
262
        goto error;
263
    if (virGetHostUUID(mig->localHostuuid) < 0) {
264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
        qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                        _("Unable to obtain host UUID"));
        goto error;
    }

    return mig;

no_memory:
    virReportOOMError();
error:
    qemuMigrationCookieFree(mig);
    return NULL;
}


279 280 281 282 283 284 285 286 287 288 289 290 291
static int
qemuMigrationCookieAddGraphics(qemuMigrationCookiePtr mig,
                               struct qemud_driver *driver,
                               virDomainObjPtr dom)
{
    if (mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                        _("Migration graphics data already present"));
        return -1;
    }

    if (dom->def->ngraphics == 1 &&
        (dom->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ||
292 293 294 295 296 297
         dom->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE)) {
        if (!(mig->graphics =
              qemuMigrationCookieGraphicsAlloc(driver, dom->def->graphics[0])))
            return -1;
        mig->flags |= QEMU_MIGRATION_COOKIE_GRAPHICS;
    }
298 299 300 301 302

    return 0;
}


303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337
static int
qemuMigrationCookieAddLockstate(qemuMigrationCookiePtr mig,
                                struct qemud_driver *driver,
                                virDomainObjPtr dom)
{
    qemuDomainObjPrivatePtr priv = dom->privateData;

    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                        _("Migration lockstate data already present"));
        return -1;
    }

    if (virDomainObjGetState(dom, NULL) == VIR_DOMAIN_PAUSED) {
        if (priv->lockState &&
            !(mig->lockState = strdup(priv->lockState)))
            return -1;
    } else {
        if (virDomainLockProcessInquire(driver->lockManager, dom, &mig->lockState) < 0)
            return -1;
    }

    if (!(mig->lockDriver = strdup(virLockManagerPluginGetName(driver->lockManager)))) {
        VIR_FREE(mig->lockState);
        return -1;
    }

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

    return 0;
}



338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
static void qemuMigrationCookieGraphicsXMLFormat(virBufferPtr buf,
                                                 qemuMigrationCookieGraphicsPtr grap)
{
    virBufferAsprintf(buf, "  <graphics type='%s' port='%d' listen='%s'",
                      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");
        virBufferEscapeString(buf, "    <cert info='subject' value='%s'/>\n", grap->tlsSubject);
        virBufferAddLit(buf, "  </graphics>\n");
    } else {
        virBufferAddLit(buf, "/>\n");
    }
}


356 357 358 359 360
static void qemuMigrationCookieXMLFormat(virBufferPtr buf,
                                         qemuMigrationCookiePtr mig)
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
    char hostuuidstr[VIR_UUID_STRING_BUFLEN];
361
    int i;
362 363

    virUUIDFormat(mig->uuid, uuidstr);
364
    virUUIDFormat(mig->localHostuuid, hostuuidstr);
365 366 367 368

    virBufferAsprintf(buf, "<qemu-migration>\n");
    virBufferEscapeString(buf, "  <name>%s</name>\n", mig->name);
    virBufferAsprintf(buf, "  <uuid>%s</uuid>\n", uuidstr);
369
    virBufferEscapeString(buf, "  <hostname>%s</hostname>\n", mig->localHostname);
370
    virBufferAsprintf(buf, "  <hostuuid>%s</hostuuid>\n", hostuuidstr);
371

372 373 374 375 376 377
    for (i = 0 ; i < QEMU_MIGRATION_COOKIE_FLAG_LAST ; i++) {
        if (mig->flagsMandatory & (1 << i))
            virBufferAsprintf(buf, "  <feature name='%s'/>\n",
                              qemuMigrationCookieFlagTypeToString(i));
    }

378 379
    if ((mig->flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        mig->graphics)
380 381
        qemuMigrationCookieGraphicsXMLFormat(buf, mig->graphics);

382 383 384 385 386 387 388 389 390
    if ((mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        mig->lockState) {
        virBufferAsprintf(buf, "  <lockstate driver='%s'>\n",
                          mig->lockDriver);
        virBufferAsprintf(buf, "    <leases>%s</leases>\n",
                          mig->lockState);
        virBufferAddLit(buf, "  </lockstate>\n");
    }

391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409
    virBufferAddLit(buf, "</qemu-migration>\n");
}


static char *qemuMigrationCookieXMLFormatStr(qemuMigrationCookiePtr mig)
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;

    qemuMigrationCookieXMLFormat(&buf, mig);

    if (virBufferError(&buf)) {
        virReportOOMError();
        return NULL;
    }

    return virBufferContentAndReset(&buf);
}


410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447
static qemuMigrationCookieGraphicsPtr
qemuMigrationCookieGraphicsXMLParse(xmlXPathContextPtr ctxt)
{
    qemuMigrationCookieGraphicsPtr grap;
    char *tmp;

    if (VIR_ALLOC(grap) < 0)
        goto no_memory;

    if (!(tmp = virXPathString("string(./graphics/@type)", ctxt))) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        "%s", _("missing type attribute in migration data"));
        goto error;
    }
    if ((grap->type = virDomainGraphicsTypeFromString(tmp)) < 0) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("unknown graphics type %s"), tmp);
        VIR_FREE(tmp);
        goto error;
    }
    if (virXPathInt("string(./graphics/@port)", ctxt, &grap->port) < 0) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        "%s", _("missing port attribute in migration data"));
        goto error;
    }
    if (grap->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        if (virXPathInt("string(./graphics/@tlsPort)", ctxt, &grap->tlsPort) < 0) {
            qemuReportError(VIR_ERR_INTERNAL_ERROR,
                            "%s", _("missing tlsPort attribute in migration data"));
            goto error;
        }
    }
    if (!(grap->listen = virXPathString("string(./graphics/@listen)", ctxt))) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        "%s", _("missing listen attribute in migration data"));
        goto error;
    }
    /* Optional */
448
    grap->tlsSubject = virXPathString("string(./graphics/cert[@info='subject']/@value)", ctxt);
449 450 451 452 453 454 455 456 457 458 459 460


    return grap;

no_memory:
    virReportOOMError();
error:
    qemuMigrationCookieGraphicsFree(grap);
    return NULL;
}


461 462 463
static int
qemuMigrationCookieXMLParse(qemuMigrationCookiePtr mig,
                            xmlXPathContextPtr ctxt,
E
Eric Blake 已提交
464
                            unsigned int flags)
465 466 467
{
    char uuidstr[VIR_UUID_STRING_BUFLEN];
    char *tmp;
468 469
    xmlNodePtr *nodes = NULL;
    int i, n;
470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505

    /* 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))) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        "%s", _("missing name element in migration data"));
        goto error;
    }
    if (STRNEQ(tmp, mig->name)) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("Incoming cookie data had unexpected name %s vs %s"),
                        tmp, mig->name);
        goto error;
    }
    VIR_FREE(tmp);

    /* Extract domain uuid */
    tmp = virXPathString("string(./uuid[1])", ctxt);
    if (!tmp) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        "%s", _("missing uuid element in migration data"));
        goto error;
    }
    virUUIDFormat(mig->uuid, uuidstr);
    if (STRNEQ(tmp, uuidstr)) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("Incoming cookie data had unexpected UUID %s vs %s"),
                        tmp, uuidstr);
    }
    VIR_FREE(tmp);

    /* Check & forbid "localhost" migration */
506
    if (!(mig->remoteHostname = virXPathString("string(./hostname[1])", ctxt))) {
507 508 509 510
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        "%s", _("missing hostname element in migration data"));
        goto error;
    }
511
    if (STREQ(mig->remoteHostname, mig->localHostname)) {
512 513
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("Attempt to migrate guest to the same host %s"),
514
                        mig->remoteHostname);
515 516 517 518 519 520 521 522
        goto error;
    }

    if (!(tmp = virXPathString("string(./hostuuid[1])", ctxt))) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        "%s", _("missing hostuuid element in migration data"));
        goto error;
    }
523 524 525 526 527 528
    if (virUUIDParse(tmp, mig->remoteHostuuid) < 0) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        "%s", _("malformed hostuuid element in migration data"));
        goto error;
    }
    if (memcmp(mig->remoteHostuuid, mig->localHostuuid, VIR_UUID_BUFLEN) == 0) {
529 530 531 532 533 534 535
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("Attempt to migrate guest to the same host %s"),
                        tmp);
        goto error;
    }
    VIR_FREE(tmp);

536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567
    /* Check to ensure all mandatory features from XML are also
     * present in 'flags' */
    if ((n = virXPathNodeSet("./features", ctxt, &nodes)) < 0)
        goto error;

    for (i = 0 ; i < n ; i++) {
        int val;
        char *str = virXMLPropString(nodes[i], "name");
        if (!str) {
            qemuReportError(VIR_ERR_INTERNAL_ERROR,
                            "%s", _("missing feature name"));
            goto error;
        }

        if ((val = qemuMigrationCookieFlagTypeFromString(str)) < 0) {
            qemuReportError(VIR_ERR_INTERNAL_ERROR,
                            _("Unknown migration cookie feature %s"),
                            str);
            VIR_FREE(str);
            goto error;
        }

        if ((flags & (1 << val)) == 0) {
            qemuReportError(VIR_ERR_INTERNAL_ERROR,
                            _("Unsupported migration cookie feature %s"),
                            str);
            VIR_FREE(str);
        }
        VIR_FREE(str);
    }
    VIR_FREE(nodes);

568 569 570 571 572
    if ((flags & QEMU_MIGRATION_COOKIE_GRAPHICS) &&
        virXPathBoolean("count(./graphics) > 0", ctxt) &&
        (!(mig->graphics = qemuMigrationCookieGraphicsXMLParse(ctxt))))
        goto error;

573 574 575 576 577 578 579 580 581 582 583 584 585
    if ((flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) &&
        virXPathBoolean("count(./lockstate) > 0", ctxt)) {
        mig->lockDriver = virXPathString("string(./lockstate[1]/@driver)", ctxt);
        if (!mig->lockDriver) {
            qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                            _("Missing lock driver name in migration cookie"));
            goto error;
        }
        mig->lockState = virXPathString("string(./lockstate[1]/leases[1])", ctxt);
        if (mig->lockState && STREQ(mig->lockState, ""))
            VIR_FREE(mig->lockState);
    }

586 587 588 589
    return 0;

error:
    VIR_FREE(tmp);
590
    VIR_FREE(nodes);
591 592 593 594 595 596 597
    return -1;
}


static int
qemuMigrationCookieXMLParseStr(qemuMigrationCookiePtr mig,
                               const char *xml,
E
Eric Blake 已提交
598
                               unsigned int flags)
599 600 601
{
    xmlDocPtr doc = NULL;
    xmlXPathContextPtr ctxt = NULL;
602
    int ret = -1;
603 604 605

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

606
    if (!(doc = virXMLParseStringCtxt(xml, "qemumigration.xml", &ctxt)))
607 608 609 610 611 612 613 614 615 616 617 618 619 620
        goto cleanup;

    ret = qemuMigrationCookieXMLParse(mig, ctxt, flags);

cleanup:
    xmlXPathFreeContext(ctxt);
    xmlFreeDoc(doc);

    return ret;
}


static int
qemuMigrationBakeCookie(qemuMigrationCookiePtr mig,
621 622
                        struct qemud_driver *driver,
                        virDomainObjPtr dom,
623 624
                        char **cookieout,
                        int *cookieoutlen,
E
Eric Blake 已提交
625
                        unsigned int flags)
626
{
627 628
    if (!cookieout || !cookieoutlen)
        return 0;
629 630 631

    *cookieoutlen = 0;

632 633 634 635
    if (flags & QEMU_MIGRATION_COOKIE_GRAPHICS &&
        qemuMigrationCookieAddGraphics(mig, driver, dom) < 0)
        return -1;

636 637 638 639
    if (flags & QEMU_MIGRATION_COOKIE_LOCKSTATE &&
        qemuMigrationCookieAddLockstate(mig, driver, dom) < 0)
        return -1;

640 641 642 643 644 645 646 647 648 649 650 651
    if (!(*cookieout = qemuMigrationCookieXMLFormatStr(mig)))
        return -1;

    *cookieoutlen = strlen(*cookieout) + 1;

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

    return 0;
}


static qemuMigrationCookiePtr
652 653
qemuMigrationEatCookie(struct qemud_driver *driver,
                       virDomainObjPtr dom,
654 655
                       const char *cookiein,
                       int cookieinlen,
E
Eric Blake 已提交
656
                       unsigned int flags)
657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678
{
    qemuMigrationCookiePtr mig = NULL;

    /* Parse & validate incoming cookie (if any) */
    if (cookiein && cookieinlen &&
        cookiein[cookieinlen-1] != '\0') {
        qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                        _("Migration cookie was not NULL terminated"));
        goto error;
    }

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

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

    if (cookiein && cookieinlen &&
        qemuMigrationCookieXMLParseStr(mig,
                                       cookiein,
                                       flags) < 0)
        goto error;

679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696
    if (mig->flags & QEMU_MIGRATION_COOKIE_LOCKSTATE) {
        if (!mig->lockDriver) {
            if (virLockManagerPluginUsesState(driver->lockManager)) {
                qemuReportError(VIR_ERR_INTERNAL_ERROR,
                                _("Missing %s lock state for migration cookie"),
                                virLockManagerPluginGetName(driver->lockManager));
                goto error;
            }
        } else if (STRNEQ(mig->lockDriver,
                          virLockManagerPluginGetName(driver->lockManager))) {
            qemuReportError(VIR_ERR_INTERNAL_ERROR,
                            _("Source host lock driver %s different from target %s"),
                            mig->lockDriver,
                            virLockManagerPluginGetName(driver->lockManager));
            goto error;
        }
    }

697 698 699 700 701 702
    return mig;

error:
    qemuMigrationCookieFree(mig);
    return NULL;
}
703

704 705 706 707 708 709 710 711 712 713 714 715
/* 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. */
static bool
qemuMigrationIsAllowed(struct qemud_driver *driver, virDomainObjPtr vm,
                       virDomainDefPtr def)
716
{
717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733
    int nsnapshots;

    if (vm) {
        if (qemuProcessAutoDestroyActive(driver, vm)) {
            qemuReportError(VIR_ERR_OPERATION_INVALID,
                            "%s", _("domain is marked for auto destroy"));
            return false;
        }
        if ((nsnapshots = virDomainSnapshotObjListNum(&vm->snapshots, 0))) {
            qemuReportError(VIR_ERR_OPERATION_INVALID,
                            _("cannot migrate domain with %d snapshots"),
                            nsnapshots);
            return false;
        }

        def = vm->def;
    }
734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750
    if (def->nhostdevs > 0) {
        qemuReportError(VIR_ERR_OPERATION_INVALID,
            "%s", _("Domain with assigned host devices cannot be migrated"));
        return false;
    }

    return true;
}

/** qemuMigrationSetOffline
 * Pause domain for non-live migration.
 */
int
qemuMigrationSetOffline(struct qemud_driver *driver,
                        virDomainObjPtr vm)
{
    int ret;
751
    VIR_DEBUG("driver=%p vm=%p", driver, vm);
752 753
    ret = qemuProcessStopCPUs(driver, vm, VIR_DOMAIN_PAUSED_MIGRATION,
                              QEMU_ASYNC_JOB_MIGRATION_OUT);
754 755 756 757 758 759 760 761 762 763 764 765 766 767
    if (ret == 0) {
        virDomainEventPtr event;

        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
        if (event)
            qemuDomainEventQueue(driver, event);
    }

    return ret;
}


768 769 770
static int
qemuMigrationUpdateJobStatus(struct qemud_driver *driver,
                             virDomainObjPtr vm,
771 772
                             const char *job,
                             enum qemuDomainAsyncJob asyncJob)
773
{
774
    qemuDomainObjPrivatePtr priv = vm->privateData;
775 776 777 778 779
    int ret = -1;
    int status;
    unsigned long long memProcessed;
    unsigned long long memRemaining;
    unsigned long long memTotal;
780

781 782 783
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
    if (ret < 0) {
        /* Guest already exited; nothing further to update.  */
784 785
        return -1;
    }
786 787 788 789 790 791
    ret = qemuMonitorGetMigrationStatus(priv->mon,
                                        &status,
                                        &memProcessed,
                                        &memRemaining,
                                        &memTotal);
    qemuDomainObjExitMonitorWithDriver(driver, vm);
792

793 794
    if (ret < 0 || virTimeMs(&priv->job.info.timeElapsed) < 0) {
        priv->job.info.type = VIR_DOMAIN_JOB_FAILED;
795 796
        return -1;
    }
797
    priv->job.info.timeElapsed -= priv->job.start;
798 799 800

    switch (status) {
    case QEMU_MONITOR_MIGRATION_STATUS_INACTIVE:
801
        priv->job.info.type = VIR_DOMAIN_JOB_NONE;
802 803 804 805 806
        qemuReportError(VIR_ERR_OPERATION_FAILED,
                        _("%s: %s"), job, _("is not active"));
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ACTIVE:
807 808 809
        priv->job.info.dataTotal = memTotal;
        priv->job.info.dataRemaining = memRemaining;
        priv->job.info.dataProcessed = memProcessed;
810

811 812 813
        priv->job.info.memTotal = memTotal;
        priv->job.info.memRemaining = memRemaining;
        priv->job.info.memProcessed = memProcessed;
814 815 816 817 818

        ret = 0;
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_COMPLETED:
819
        priv->job.info.type = VIR_DOMAIN_JOB_COMPLETED;
820 821 822 823
        ret = 0;
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_ERROR:
824
        priv->job.info.type = VIR_DOMAIN_JOB_FAILED;
825 826 827 828 829
        qemuReportError(VIR_ERR_OPERATION_FAILED,
                        _("%s: %s"), job, _("unexpectedly failed"));
        break;

    case QEMU_MONITOR_MIGRATION_STATUS_CANCELLED:
830
        priv->job.info.type = VIR_DOMAIN_JOB_CANCELLED;
831 832 833 834 835 836 837 838 839
        qemuReportError(VIR_ERR_OPERATION_FAILED,
                        _("%s: %s"), job, _("canceled by client"));
        break;
    }

    return ret;
}


840 841 842
static int
qemuMigrationWaitForCompletion(struct qemud_driver *driver, virDomainObjPtr vm,
                               enum qemuDomainAsyncJob asyncJob)
843
{
844
    qemuDomainObjPrivatePtr priv = vm->privateData;
845 846
    const char *job;

847 848
    switch (priv->job.asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
849 850
        job = _("migration job");
        break;
851
    case QEMU_ASYNC_JOB_SAVE:
852 853
        job = _("domain save job");
        break;
854
    case QEMU_ASYNC_JOB_DUMP:
855 856 857 858 859
        job = _("domain core dump job");
        break;
    default:
        job = _("job");
    }
860

861
    priv->job.info.type = VIR_DOMAIN_JOB_UNBOUNDED;
862

863
    while (priv->job.info.type == VIR_DOMAIN_JOB_UNBOUNDED) {
864 865 866
        /* Poll every 50ms for progress & to allow cancellation */
        struct timespec ts = { .tv_sec = 0, .tv_nsec = 50 * 1000 * 1000ull };

867
        if (qemuMigrationUpdateJobStatus(driver, vm, job, asyncJob) < 0)
868 869 870 871 872 873 874 875 876 877 878 879
            goto cleanup;

        virDomainObjUnlock(vm);
        qemuDriverUnlock(driver);

        nanosleep(&ts, NULL);

        qemuDriverLock(driver);
        virDomainObjLock(vm);
    }

cleanup:
880
    if (priv->job.info.type == VIR_DOMAIN_JOB_COMPLETED)
881 882 883
        return 0;
    else
        return -1;
884 885 886
}


887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906
static int
qemuDomainMigrateGraphicsRelocate(struct qemud_driver *driver,
                                  virDomainObjPtr vm,
                                  qemuMigrationCookiePtr cookie)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret;

    if (!cookie)
        return 0;

    if (!cookie->graphics)
        return 0;

    /* QEMU doesn't support VNC relocation yet, so
     * skip it to avoid generating an error
     */
    if (cookie->graphics->type != VIR_DOMAIN_GRAPHICS_TYPE_SPICE)
        return 0;

907 908
    ret = qemuDomainObjEnterMonitorAsync(driver, vm,
                                         QEMU_ASYNC_JOB_MIGRATION_OUT);
909 910 911 912 913 914 915 916 917
    if (ret == 0) {
        ret = qemuMonitorGraphicsRelocate(priv->mon,
                                          cookie->graphics->type,
                                          cookie->remoteHostname,
                                          cookie->graphics->port,
                                          cookie->graphics->tlsPort,
                                          cookie->graphics->tlsSubject);
        qemuDomainObjExitMonitorWithDriver(driver, vm);
    }
918 919 920 921 922

    return ret;
}


923
/* The caller is supposed to lock the vm and start a migration job. */
924 925
char *qemuMigrationBegin(struct qemud_driver *driver,
                         virDomainObjPtr vm,
926
                         const char *xmlin,
927 928 929 930 931
                         char **cookieout,
                         int *cookieoutlen)
{
    char *rv = NULL;
    qemuMigrationCookiePtr mig = NULL;
932
    virDomainDefPtr def = NULL;
933 934
    qemuDomainObjPrivatePtr priv = vm->privateData;

935 936
    VIR_DEBUG("driver=%p, vm=%p, xmlin=%s, cookieout=%p, cookieoutlen=%p",
              driver, vm, NULLSTR(xmlin), cookieout, cookieoutlen);
937

938 939 940 941 942 943
    /* 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);
944

945
    if (!qemuMigrationIsAllowed(driver, vm, NULL))
946 947
        goto cleanup;

948
    if (!(mig = qemuMigrationEatCookie(driver, vm, NULL, 0, 0)))
949 950 951 952
        goto cleanup;

    if (qemuMigrationBakeCookie(mig, driver, vm,
                                cookieout, cookieoutlen,
953
                                QEMU_MIGRATION_COOKIE_LOCKSTATE) < 0)
954 955
        goto cleanup;

956 957
    if (xmlin) {
        if (!(def = virDomainDefParseString(driver->caps, xmlin,
M
Matthias Bolte 已提交
958
                                            QEMU_EXPECTED_VIRT_TYPES,
959 960 961
                                            VIR_DOMAIN_XML_INACTIVE)))
            goto cleanup;

962
        if (!virDomainDefCheckABIStability(vm->def, def))
963 964 965 966 967 968 969 970 971 972
            goto cleanup;

        rv = qemuDomainDefFormatXML(driver, def,
                                    VIR_DOMAIN_XML_SECURE |
                                    VIR_DOMAIN_XML_UPDATE_CPU);
    } else {
        rv = qemuDomainFormatXML(driver, vm,
                                 VIR_DOMAIN_XML_SECURE |
                                 VIR_DOMAIN_XML_UPDATE_CPU);
    }
973 974 975

cleanup:
    qemuMigrationCookieFree(mig);
976
    virDomainDefFree(def);
977 978 979 980
    return rv;
}


981 982
/* Prepare is the first step, and it runs on the destination host.
 */
983 984 985 986 987 988 989 990 991 992 993 994

static int
qemuMigrationPrepareAny(struct qemud_driver *driver,
                        virConnectPtr dconn,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        const char *dname,
                        const char *dom_xml,
                        const char *migrateFrom,
                        virStreamPtr st)
995 996 997 998 999
{
    virDomainDefPtr def = NULL;
    virDomainObjPtr vm = NULL;
    virDomainEventPtr event = NULL;
    int ret = -1;
1000
    int dataFD[2] = { -1, -1 };
1001
    qemuDomainObjPrivatePtr priv = NULL;
J
Jiri Denemark 已提交
1002
    unsigned long long now;
1003
    qemuMigrationCookiePtr mig = NULL;
1004
    bool tunnel = !!st;
1005

J
Jiri Denemark 已提交
1006
    if (virTimeMs(&now) < 0)
1007 1008 1009
        return -1;

    if (!(def = virDomainDefParseString(driver->caps, dom_xml,
M
Matthias Bolte 已提交
1010
                                        QEMU_EXPECTED_VIRT_TYPES,
1011 1012 1013
                                        VIR_DOMAIN_XML_INACTIVE)))
        goto cleanup;

1014
    if (!qemuMigrationIsAllowed(driver, NULL, def))
1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036
        goto cleanup;

    /* Target domain name, maybe renamed. */
    if (dname) {
        VIR_FREE(def->name);
        def->name = strdup(dname);
        if (def->name == NULL)
            goto cleanup;
    }

    if (virDomainObjIsDuplicate(&driver->domains, def, 1) < 0)
        goto cleanup;

    if (!(vm = virDomainAssignDef(driver->caps,
                                  &driver->domains,
                                  def, true))) {
        /* virDomainAssignDef already set the error */
        goto cleanup;
    }
    def = NULL;
    priv = vm->privateData;

1037 1038
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
                                       QEMU_MIGRATION_COOKIE_LOCKSTATE)))
1039 1040
        goto cleanup;

1041
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_IN) < 0)
1042
        goto cleanup;
1043
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PREPARE);
1044 1045 1046 1047

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

1048 1049
    if (tunnel &&
        (pipe(dataFD) < 0 || virSetCloseExec(dataFD[1]) < 0)) {
1050 1051
        virReportSystemError(errno, "%s",
                             _("cannot create pipe for tunnelled migration"));
1052 1053 1054 1055
        goto endjob;
    }

    /* Start the QEMU daemon, with the same command-line arguments plus
1056
     * -incoming $migrateFrom
1057
     */
1058
    if (qemuProcessStart(dconn, driver, vm, migrateFrom, true,
1059
                         true, dataFD[0], NULL, NULL,
1060
                         VIR_VM_OP_MIGRATE_IN_START) < 0) {
1061
        virDomainAuditStart(vm, "migrated", false);
1062 1063 1064 1065 1066 1067
        /* Note that we don't set an error here because qemuProcessStart
         * should have already done that.
         */
        goto endjob;
    }

1068 1069 1070 1071 1072 1073 1074
    if (tunnel) {
        if (virFDStreamOpen(st, dataFD[1]) < 0) {
            virReportSystemError(errno, "%s",
                                 _("cannot pass pipe for tunnelled migration"));
            virDomainAuditStart(vm, "migrated", false);
            qemuProcessStop(driver, vm, 0, VIR_DOMAIN_SHUTOFF_FAILED);
            goto endjob;
1075
        }
1076
        dataFD[1] = -1; /* 'st' owns the FD now & will close it */
1077 1078
    }

1079 1080 1081 1082 1083 1084 1085 1086
    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");
    }
1087

1088 1089
    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen,
                                QEMU_MIGRATION_COOKIE_GRAPHICS) < 0) {
1090 1091 1092 1093 1094 1095 1096
        /* 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");
    }

1097 1098 1099 1100
    virDomainAuditStart(vm, "migrated", true);
    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STARTED,
                                     VIR_DOMAIN_EVENT_STARTED_MIGRATED);
1101

1102 1103 1104 1105 1106
    /* 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.
     */
    if (qemuMigrationJobContinue(vm) == 0) {
1107
        vm = NULL;
1108 1109 1110
        qemuReportError(VIR_ERR_OPERATION_FAILED,
                        "%s", _("domain disappeared"));
        goto cleanup;
1111
    }
1112

1113
    ret = 0;
1114 1115 1116

cleanup:
    virDomainDefFree(def);
1117 1118
    VIR_FORCE_CLOSE(dataFD[0]);
    VIR_FORCE_CLOSE(dataFD[1]);
1119 1120 1121 1122
    if (vm)
        virDomainObjUnlock(vm);
    if (event)
        qemuDomainEventQueue(driver, event);
1123
    qemuMigrationCookieFree(mig);
1124
    return ret;
1125 1126 1127 1128 1129 1130 1131 1132 1133

endjob:
    if (qemuMigrationJobFinish(driver, vm) == 0) {
        vm = NULL;
    } else if (!vm->persistent) {
        virDomainRemoveInactive(&driver->domains, vm);
        vm = NULL;
    }
    goto cleanup;
1134 1135 1136
}


1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168
/*
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
int
qemuMigrationPrepareTunnel(struct qemud_driver *driver,
                           virConnectPtr dconn,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           virStreamPtr st,
                           const char *dname,
                           const char *dom_xml)
{
    int ret;

    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, st=%p, dname=%s, dom_xml=%s",
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, st, NULLSTR(dname), dom_xml);

    /* QEMU will be started with -incoming stdio (which qemu_command might
     * convert to exec:cat or fd:n)
     */
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
                                  cookieout, cookieoutlen, dname, dom_xml,
                                  "stdio", st);
    return ret;
}


1169 1170 1171
int
qemuMigrationPrepareDirect(struct qemud_driver *driver,
                           virConnectPtr dconn,
1172 1173 1174 1175
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186
                           const char *uri_in,
                           char **uri_out,
                           const char *dname,
                           const char *dom_xml)
{
    static int port = 0;
    int this_port;
    char *hostname = NULL;
    char migrateFrom [64];
    const char *p;
    int ret = -1;
J
Jiri Denemark 已提交
1187

1188 1189 1190 1191 1192 1193
    VIR_DEBUG("driver=%p, dconn=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, uri_in=%s, uri_out=%p, "
              "dname=%s, dom_xml=%s",
              driver, dconn, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, NULLSTR(uri_in), uri_out,
              NULLSTR(dname), dom_xml);
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214

    /* 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
     * from our pool of port numbers and return a URI of
     * "tcp://ourhostname:port".
     *
     * 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) {
        this_port = QEMUD_MIGRATION_FIRST_PORT + port++;
        if (port == QEMUD_MIGRATION_NUM_PORTS) port = 0;

        /* Get hostname */
        if ((hostname = virGetHostname(NULL)) == NULL)
            goto cleanup;

        if (STRPREFIX(hostname, "localhost")) {
            qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
1215 1216
                            _("hostname on destination resolved to localhost,"
                              " but migration requires an FQDN"));
1217 1218 1219 1220 1221 1222 1223 1224 1225
            goto cleanup;
        }

        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
         * compatability with old targets. We at least make the
         * new targets accept both syntaxes though.
         */
        /* Caller frees */
1226
        if (virAsprintf(uri_out, "tcp:%s:%d", hostname, this_port) < 0) {
1227 1228 1229 1230 1231 1232 1233 1234 1235
            virReportOOMError();
            goto cleanup;
        }
    } else {
        /* Check the URI starts with "tcp:".  We will escape the
         * URI when passing it to the qemu monitor, so bad
         * characters in hostname part don't matter.
         */
        if (!STRPREFIX (uri_in, "tcp:")) {
1236 1237 1238
            qemuReportError(VIR_ERR_INVALID_ARG, "%s",
                            _("only tcp URIs are supported for KVM/QEMU"
                              " migrations"));
1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
            goto cleanup;
        }

        /* Get the port number. */
        p = strrchr (uri_in, ':');
        if (p == strchr(uri_in, ':')) {
            /* Generate a port */
            this_port = QEMUD_MIGRATION_FIRST_PORT + port++;
            if (port == QEMUD_MIGRATION_NUM_PORTS)
                port = 0;

            /* Caller frees */
            if (virAsprintf(uri_out, "%s:%d", uri_in, this_port) < 0) {
                virReportOOMError();
                goto cleanup;
            }

        } else {
            p++; /* definitely has a ':' in it, see above */
            this_port = virParseNumber (&p);
            if (this_port == -1 || p-uri_in != strlen (uri_in)) {
                qemuReportError(VIR_ERR_INVALID_ARG,
                                "%s", _("URI ended with incorrect ':port'"));
                goto cleanup;
            }
        }
    }

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

1270 1271
    /* QEMU will be started with -incoming tcp:0.0.0.0:port */
    snprintf(migrateFrom, sizeof (migrateFrom), "tcp:0.0.0.0:%d", this_port);
1272

1273 1274 1275
    ret = qemuMigrationPrepareAny(driver, dconn, cookiein, cookieinlen,
                                  cookieout, cookieoutlen, dname, dom_xml,
                                  migrateFrom, NULL);
1276 1277 1278 1279 1280 1281 1282 1283
cleanup:
    VIR_FREE(hostname);
    if (ret != 0)
        VIR_FREE(*uri_out);
    return ret;
}


1284 1285 1286
enum qemuMigrationDestinationType {
    MIGRATION_DEST_HOST,
    MIGRATION_DEST_UNIX,
1287
    MIGRATION_DEST_FD,
1288
};
1289

1290 1291 1292 1293
enum qemuMigrationForwardType {
    MIGRATION_FWD_DIRECT,
    MIGRATION_FWD_STREAM,
};
1294

1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305
typedef struct _qemuMigrationSpec qemuMigrationSpec;
typedef qemuMigrationSpec *qemuMigrationSpecPtr;
struct _qemuMigrationSpec {
    enum qemuMigrationDestinationType destType;
    union {
        struct {
            const char *name;
            int port;
        } host;

        struct {
1306
            char *file;
1307 1308
            int sock;
        } unix_socket;
1309 1310 1311 1312 1313

        struct {
            int qemu;
            int local;
        } fd;
1314 1315 1316 1317 1318 1319 1320
    } dest;

    enum qemuMigrationForwardType fwdType;
    union {
        virStreamPtr stream;
    } fwd;
};
1321 1322 1323

#define TUNNEL_SEND_BUF_SIZE 65536

1324 1325 1326 1327 1328 1329 1330 1331 1332 1333
typedef struct _qemuMigrationIOThread qemuMigrationIOThread;
typedef qemuMigrationIOThread *qemuMigrationIOThreadPtr;
struct _qemuMigrationIOThread {
    virThread thread;
    virStreamPtr st;
    int sock;
    virError err;
};

static void qemuMigrationIOFunc(void *arg)
1334
{
1335
    qemuMigrationIOThreadPtr data = arg;
1336 1337 1338 1339 1340
    char *buffer;
    int nbytes = TUNNEL_SEND_BUF_SIZE;

    if (VIR_ALLOC_N(buffer, TUNNEL_SEND_BUF_SIZE) < 0) {
        virReportOOMError();
1341 1342
        virStreamAbort(data->st);
        goto error;
1343 1344 1345
    }

    for (;;) {
1346
        nbytes = saferead(data->sock, buffer, TUNNEL_SEND_BUF_SIZE);
1347 1348 1349
        if (nbytes < 0) {
            virReportSystemError(errno, "%s",
                                 _("tunnelled migration failed to read from qemu"));
1350
            virStreamAbort(data->st);
1351
            VIR_FREE(buffer);
1352
            goto error;
1353 1354 1355 1356 1357
        }
        else if (nbytes == 0)
            /* EOF; get out of here */
            break;

1358
        if (virStreamSend(data->st, buffer, nbytes) < 0) {
1359
            VIR_FREE(buffer);
1360
            goto error;
1361 1362 1363 1364 1365
        }
    }

    VIR_FREE(buffer);

1366 1367
    if (virStreamFinish(data->st) < 0)
        goto error;
1368

1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420
    return;

error:
    virCopyLastError(&data->err);
    virResetLastError();
}


static qemuMigrationIOThreadPtr
qemuMigrationStartTunnel(virStreamPtr st,
                         int sock)
{
    qemuMigrationIOThreadPtr io;

    if (VIR_ALLOC(io) < 0) {
        virReportOOMError();
        return NULL;
    }

    io->st = st;
    io->sock = sock;

    if (virThreadCreate(&io->thread, true,
                        qemuMigrationIOFunc,
                        io) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to create migration thread"));
        VIR_FREE(io);
        return NULL;
    }

    return io;
}

static int
qemuMigrationStopTunnel(qemuMigrationIOThreadPtr io)
{
    int rv = -1;
    virThreadJoin(&io->thread);

    /* Forward error from the IO thread, to this thread */
    if (io->err.code != VIR_ERR_OK) {
        virSetError(&io->err);
        virResetError(&io->err);
        goto cleanup;
    }

    rv = 0;

cleanup:
    VIR_FREE(io);
    return rv;
1421 1422
}

1423 1424 1425 1426 1427 1428 1429 1430 1431 1432
static int
qemuMigrationRun(struct qemud_driver *driver,
                 virDomainObjPtr vm,
                 const char *cookiein,
                 int cookieinlen,
                 char **cookieout,
                 int *cookieoutlen,
                 unsigned long flags,
                 unsigned long resource,
                 qemuMigrationSpecPtr spec)
1433
{
1434
    int ret = -1;
1435 1436
    unsigned int migrate_flags = QEMU_MONITOR_MIGRATE_BACKGROUND;
    qemuDomainObjPrivatePtr priv = vm->privateData;
1437
    qemuMigrationCookiePtr mig = NULL;
1438
    qemuMigrationIOThreadPtr iothread = NULL;
1439 1440 1441 1442 1443 1444 1445 1446
    int fd = -1;

    VIR_DEBUG("driver=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu, "
              "spec=%p (dest=%d, fwd=%d)",
              driver, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource,
              spec, spec->destType, spec->fwdType);
1447

1448 1449 1450
    if (virLockManagerPluginUsesState(driver->lockManager) &&
        !cookieout) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
1451 1452
                        _("Migration with lock driver %s requires"
                          " cookie support"),
1453 1454 1455 1456 1457
                        virLockManagerPluginGetName(driver->lockManager));
        return -1;
    }

    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen,
1458 1459 1460 1461 1462 1463
                                       QEMU_MIGRATION_COOKIE_GRAPHICS)))
        goto cleanup;

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

1464
    /* Before EnterMonitor, since qemuMigrationSetOffline already does that */
1465 1466 1467 1468 1469 1470
    if (!(flags & VIR_MIGRATE_LIVE) &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

1471 1472
    if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
1473 1474
        goto cleanup;

1475 1476 1477 1478 1479
    if (resource > 0 &&
        qemuMonitorSetMigrationSpeed(priv->mon, resource) < 0) {
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        goto cleanup;
    }
1480

1481
    if (flags & VIR_MIGRATE_NON_SHARED_DISK)
1482 1483
        migrate_flags |= QEMU_MONITOR_MIGRATE_NON_SHARED_DISK;

1484
    if (flags & VIR_MIGRATE_NON_SHARED_INC)
1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
        migrate_flags |= QEMU_MONITOR_MIGRATE_NON_SHARED_INC;

    switch (spec->destType) {
    case MIGRATION_DEST_HOST:
        ret = qemuMonitorMigrateToHost(priv->mon, migrate_flags,
                                       spec->dest.host.name,
                                       spec->dest.host.port);
        break;

    case MIGRATION_DEST_UNIX:
        if (qemuCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX)) {
            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;
1505 1506 1507 1508 1509 1510 1511 1512

    case MIGRATION_DEST_FD:
        if (spec->fwdType != MIGRATION_FWD_DIRECT)
            fd = spec->dest.fd.local;
        ret = qemuMonitorMigrateToFd(priv->mon, migrate_flags,
                                     spec->dest.fd.qemu);
        VIR_FORCE_CLOSE(spec->dest.fd.qemu);
        break;
1513 1514
    }
    qemuDomainObjExitMonitorWithDriver(driver, vm);
1515
    if (ret < 0)
1516 1517
        goto cleanup;
    ret = -1;
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527

    if (!virDomainObjIsActive(vm)) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                        _("guest unexpectedly quit"));
        goto cleanup;
    }

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

1528 1529 1530 1531 1532 1533 1534 1535
    if (spec->destType == MIGRATION_DEST_UNIX) {
        /* It is also possible that the migrate didn't fail initially, but
         * rather failed later on.  Check its status before waiting for a
         * connection from qemu which may never be initiated.
         */
        if (qemuMigrationUpdateJobStatus(driver, vm, _("migration job"),
                                         QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
            goto cancel;
1536

1537 1538 1539 1540 1541 1542 1543
        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;
        }
1544 1545
    }

1546 1547
    if (spec->fwdType != MIGRATION_FWD_DIRECT &&
        !(iothread = qemuMigrationStartTunnel(spec->fwd.stream, fd)))
1548 1549
        goto cancel;

1550 1551 1552
    if (qemuMigrationWaitForCompletion(driver, vm,
                                       QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
        goto cleanup;
1553

1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566
    /* When migration completed, QEMU will have paused the
     * CPUs for us, but unless we're using the JSON monitor
     * we won't have been notified of this, so might still
     * think we're running. For v2 protocol this doesn't
     * matter because we'll kill the VM soon, but for v3
     * this is important because we stay paused until the
     * confirm3 step, but need to release the lock state
     */
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        if (qemuMigrationSetOffline(driver, vm) < 0)
            goto cleanup;
    }

1567
    ret = 0;
1568

1569 1570 1571 1572 1573 1574 1575
cleanup:
    if (spec->fwdType != MIGRATION_FWD_DIRECT) {
        /* Close now to ensure the IO thread quits & is joinable */
        VIR_FORCE_CLOSE(fd);
        if (iothread && qemuMigrationStopTunnel(iothread) < 0)
            ret = -1;
    }
1576

1577 1578 1579 1580
    if (ret == 0 &&
        qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen, 0) < 0)
        VIR_WARN("Unable to encode migration cookie");

1581 1582 1583 1584
    qemuMigrationCookieFree(mig);

    return ret;

1585
cancel:
1586
    if (virDomainObjIsActive(vm)) {
1587 1588
        if (qemuDomainObjEnterMonitorAsync(driver, vm,
                                           QEMU_ASYNC_JOB_MIGRATION_OUT) == 0) {
1589 1590 1591
            qemuMonitorMigrateCancel(priv->mon);
            qemuDomainObjExitMonitorWithDriver(driver, vm);
        }
1592
    }
1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
    goto cleanup;
}

/* Perform migration using QEMU's native TCP migrate support,
 * not encrypted obviously
 */
static int doNativeMigrate(struct qemud_driver *driver,
                           virDomainObjPtr vm,
                           const char *uri,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
                           unsigned long resource)
{
1609
    qemuDomainObjPrivatePtr priv = vm->privateData;
1610
    xmlURIPtr uribits = NULL;
1611
    int ret = -1;
1612
    qemuMigrationSpec spec;
1613
    char *tmp = NULL;
1614 1615 1616 1617 1618 1619 1620 1621

    VIR_DEBUG("driver=%p, vm=%p, uri=%s, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu",
              driver, vm, uri, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource);

    if (STRPREFIX(uri, "tcp:") && !STRPREFIX(uri, "tcp://")) {
        /* HACK: source host generates bogus URIs, so fix them up */
1622
        if (virAsprintf(&tmp, "tcp://%s", uri + strlen("tcp:")) < 0) {
1623 1624 1625
            virReportOOMError();
            return -1;
        }
1626 1627
        uribits = xmlParseURI(tmp);
        VIR_FREE(tmp);
1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638
    } else {
        uribits = xmlParseURI(uri);
    }
    if (!uribits) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("cannot parse URI %s"), uri);
        return -1;
    }

    spec.fwdType = MIGRATION_FWD_DIRECT;

1639 1640 1641 1642 1643 1644 1645 1646 1647 1648
    if (qemuCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD)) {
        virNetSocketPtr sock;

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

        if (virAsprintf(&tmp, "%d", uribits->port) < 0) {
            virReportOOMError();
            goto cleanup;
        }
1649 1650
        if (virSecurityManagerSetSocketLabel(driver->securityManager, vm) < 0)
            goto cleanup;
1651 1652 1653 1654
        if (virNetSocketNewConnectTCP(uribits->server, tmp, &sock) == 0) {
            spec.dest.fd.qemu = virNetSocketDupFD(sock, true);
            virNetSocketFree(sock);
        }
1655 1656
        if (virSecurityManagerClearSocketLabel(driver->securityManager, vm) < 0 ||
            spec.dest.fd.qemu == -1)
1657 1658 1659 1660 1661 1662 1663
            goto cleanup;
    } else {
        spec.destType = MIGRATION_DEST_HOST;
        spec.dest.host.name = uribits->server;
        spec.dest.host.port = uribits->port;
    }

1664 1665
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
                           cookieoutlen, flags, resource, &spec);
1666 1667 1668 1669 1670 1671

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

    VIR_FREE(tmp);
1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
    xmlFreeURI(uribits);

    return ret;
}


static int doTunnelMigrate(struct qemud_driver *driver,
                           virDomainObjPtr vm,
                           virStreamPtr st,
                           const char *cookiein,
                           int cookieinlen,
                           char **cookieout,
                           int *cookieoutlen,
                           unsigned long flags,
                           unsigned long resource)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1689
    virNetSocketPtr sock = NULL;
1690 1691 1692 1693 1694 1695 1696 1697
    int ret = -1;
    qemuMigrationSpec spec;

    VIR_DEBUG("driver=%p, vm=%p, st=%p, cookiein=%s, cookieinlen=%d, "
              "cookieout=%p, cookieoutlen=%p, flags=%lx, resource=%lu",
              driver, vm, st, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, resource);

1698 1699
    if (!qemuCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
        !qemuCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX) &&
1700
        !qemuCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_EXEC)) {
1701 1702 1703
        qemuReportError(VIR_ERR_OPERATION_FAILED, "%s",
                        _("Source qemu is too old to support tunnelled migration"));
        return -1;
1704 1705 1706 1707 1708
    }

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

1709 1710 1711 1712 1713 1714 1715
    if (qemuCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD)) {
        int fds[2];

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

1716
        if (pipe2(fds, O_CLOEXEC) == 0) {
1717 1718 1719 1720
            spec.dest.fd.qemu = fds[1];
            spec.dest.fd.local = fds[0];
        }
        if (spec.dest.fd.qemu == -1 ||
1721 1722
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm,
                                              spec.dest.fd.qemu) < 0) {
1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747
            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",
                        driver->libDir, vm->def->name) < 0) {
            virReportOOMError();
            goto cleanup;
        }

        if (virNetSocketNewListenUNIX(spec.dest.unix_socket.file, 0700,
                                      driver->user, driver->group,
                                      &sock) < 0 ||
            virNetSocketListen(sock, 1) < 0)
            goto cleanup;

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

1748 1749
    ret = qemuMigrationRun(driver, vm, cookiein, cookieinlen, cookieout,
                           cookieoutlen, flags, resource, &spec);
1750

1751
cleanup:
1752 1753 1754 1755 1756 1757 1758
    if (spec.destType == MIGRATION_DEST_FD) {
        VIR_FORCE_CLOSE(spec.dest.fd.qemu);
        VIR_FORCE_CLOSE(spec.dest.fd.local);
    } else {
        virNetSocketFree(sock);
        VIR_FREE(spec.dest.unix_socket.file);
    }
1759 1760 1761 1762 1763

    return ret;
}


1764 1765 1766 1767 1768 1769
/* 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 */
static int doPeer2PeerMigrate2(struct qemud_driver *driver,
                               virConnectPtr sconn,
1770 1771
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
1772
                               const char *dconnuri,
1773 1774 1775
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
1776 1777 1778
{
    virDomainPtr ddomain = NULL;
    char *uri_out = NULL;
1779
    char *cookie = NULL;
1780 1781 1782 1783 1784
    char *dom_xml = NULL;
    int cookielen = 0, ret;
    virErrorPtr orig_err = NULL;
    int cancelled;
    virStreamPtr st = NULL;
1785
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, dconnuri=%s, "
1786
              "flags=%lx, dname=%s, resource=%lu",
1787 1788
              driver, sconn, dconn, vm, NULLSTR(dconnuri),
              flags, NULLSTR(dname), resource);
1789

1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824
    /* 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,
                                        VIR_DOMAIN_XML_SECURE |
                                        VIR_DOMAIN_XML_UPDATE_CPU)))
        return -1;

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

    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;

        qemuDomainObjEnterRemoteWithDriver(driver, vm);
        ret = dconn->driver->domainMigratePrepareTunnel
            (dconn, st, flags, dname, resource, dom_xml);
        qemuDomainObjExitRemoteWithDriver(driver, vm);
    } else {
        qemuDomainObjEnterRemoteWithDriver(driver, vm);
        ret = dconn->driver->domainMigratePrepare2
            (dconn, &cookie, &cookielen, NULL, &uri_out,
             flags, dname, resource, dom_xml);
        qemuDomainObjExitRemoteWithDriver(driver, vm);
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835
        goto cleanup;

    /* the domain may have shutdown or crashed while we had the locks dropped
     * in qemuDomainObjEnterRemoteWithDriver, so check again
     */
    if (!virDomainObjIsActive(vm)) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                        _("guest unexpectedly quit"));
        goto cleanup;
    }

1836 1837 1838
    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
1839
                        _("domainMigratePrepare2 did not set uri"));
1840 1841
        cancelled = 1;
        goto finish;
1842 1843
    }

1844 1845 1846 1847
    /* Perform the migration.  The driver isn't supposed to return
     * until the migration is complete.
     */
    VIR_DEBUG("Perform %p", sconn);
1848
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
1849
    if (flags & VIR_MIGRATE_TUNNELLED)
1850 1851 1852
        ret = doTunnelMigrate(driver, vm, st,
                              NULL, 0, NULL, NULL,
                              flags, resource);
1853 1854 1855 1856
    else
        ret = doNativeMigrate(driver, vm, uri_out,
                              cookie, cookielen,
                              NULL, NULL, /* No out cookie with v2 migration */
1857
                              flags, resource);
1858 1859 1860 1861

    /* Perform failed. Make sure Finish doesn't overwrite the error */
    if (ret < 0)
        orig_err = virSaveLastError();
1862

1863 1864 1865 1866
    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
    cancelled = ret < 0 ? 1 : 0;
1867 1868

finish:
1869 1870 1871 1872
    /* 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.
     */
1873
    dname = dname ? dname : vm->def->name;
1874
    VIR_DEBUG("Finish2 %p ret=%d", dconn, ret);
1875 1876
    qemuDomainObjEnterRemoteWithDriver(driver, vm);
    ddomain = dconn->driver->domainMigrateFinish2
1877
        (dconn, dname, cookie, cookielen,
1878
         uri_out ? uri_out : dconnuri, flags, cancelled);
1879 1880
    qemuDomainObjExitRemoteWithDriver(driver, vm);

1881 1882
cleanup:
    if (ddomain) {
1883
        virUnrefDomain(ddomain);
1884 1885 1886 1887
        ret = 0;
    } else {
        ret = -1;
    }
1888

1889 1890 1891 1892 1893 1894 1895 1896
    if (st)
        virUnrefStream(st);

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
1897
    VIR_FREE(cookie);
1898 1899

    return ret;
1900 1901 1902
}


1903 1904 1905 1906 1907 1908 1909 1910
/* 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 */
static int doPeer2PeerMigrate3(struct qemud_driver *driver,
                               virConnectPtr sconn,
                               virConnectPtr dconn,
                               virDomainObjPtr vm,
1911
                               const char *xmlin,
1912
                               const char *dconnuri,
1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928
                               const char *uri,
                               unsigned long flags,
                               const char *dname,
                               unsigned long resource)
{
    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;
    int cancelled;
    virStreamPtr st = NULL;
1929
    VIR_DEBUG("driver=%p, sconn=%p, dconn=%p, vm=%p, xmlin=%s, "
1930
              "dconnuri=%s, uri=%s, flags=%lx, dname=%s, resource=%lu",
1931 1932 1933
              driver, sconn, dconn, vm, NULLSTR(xmlin),
              NULLSTR(dconnuri), NULLSTR(uri), flags,
              NULLSTR(dname), resource);
1934

1935 1936 1937 1938 1939
    /* 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.  */

1940
    dom_xml = qemuMigrationBegin(driver, vm, xmlin,
1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966
                                 &cookieout, &cookieoutlen);
    if (!dom_xml)
        goto cleanup;

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

    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;

        qemuDomainObjEnterRemoteWithDriver(driver, vm);
        ret = dconn->driver->domainMigratePrepareTunnel3
            (dconn, st, cookiein, cookieinlen,
             &cookieout, &cookieoutlen,
             flags, dname, resource, dom_xml);
        qemuDomainObjExitRemoteWithDriver(driver, vm);
    } else {
        qemuDomainObjEnterRemoteWithDriver(driver, vm);
        ret = dconn->driver->domainMigratePrepare3
            (dconn, cookiein, cookieinlen, &cookieout, &cookieoutlen,
1967
             uri, &uri_out, flags, dname, resource, dom_xml);
1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
        qemuDomainObjExitRemoteWithDriver(driver, vm);
    }
    VIR_FREE(dom_xml);
    if (ret == -1)
        goto cleanup;

    if (!(flags & VIR_MIGRATE_TUNNELLED) &&
        (uri_out == NULL)) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR,
                        _("domainMigratePrepare3 did not set uri"));
        cancelled = 1;
        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.
     */
1987
    VIR_DEBUG("Perform3 %p uri=%s uri_out=%s", sconn, uri, uri_out);
1988
    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3);
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
    if (flags & VIR_MIGRATE_TUNNELLED)
        ret = doTunnelMigrate(driver, vm, st,
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
                              flags, resource);
    else
        ret = doNativeMigrate(driver, vm, uri_out,
                              cookiein, cookieinlen,
                              &cookieout, &cookieoutlen,
2003
                              flags, resource);
2004 2005

    /* Perform failed. Make sure Finish doesn't overwrite the error */
2006
    if (ret < 0) {
2007
        orig_err = virSaveLastError();
2008 2009 2010 2011
    } else {
        qemuMigrationJobSetPhase(driver, vm,
                                 QEMU_MIGRATION_PHASE_PERFORM3_DONE);
    }
2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032

    /* If Perform returns < 0, then we need to cancel the VM
     * startup on the destination
     */
    cancelled = ret < 0 ? 1 : 0;

finish:
    /*
     * 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;
    dname = dname ? dname : vm->def->name;
    qemuDomainObjEnterRemoteWithDriver(driver, vm);
2033
    ddomain = dconn->driver->domainMigrateFinish3
2034
        (dconn, dname, cookiein, cookieinlen, &cookieout, &cookieoutlen,
2035
         dconnuri, uri_out ? uri_out : uri, flags, cancelled);
2036 2037
    qemuDomainObjExitRemoteWithDriver(driver, vm);

2038 2039 2040 2041 2042 2043 2044
    /* 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.
2045
     */
2046
    cancelled = ddomain == NULL ? 1 : 0;
2047

2048 2049 2050 2051 2052 2053
    /* 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();

2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065
    /*
     * If cancelled, then src VM will be restarted, else
     * it will be killed
     */
    VIR_DEBUG("Confirm3 %p ret=%d vm=%p", sconn, ret, vm);
    VIR_FREE(cookiein);
    cookiein = cookieout;
    cookieinlen = cookieoutlen;
    cookieout = NULL;
    cookieoutlen = 0;
    ret = qemuMigrationConfirm(driver, sconn, vm,
                               cookiein, cookieinlen,
2066
                               flags, cancelled);
2067 2068 2069 2070
    /* 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.
     */
2071 2072 2073
    if (ret < 0)
        VIR_WARN("Guest %s probably left in 'paused' state on source",
                 vm->def->name);
2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097

 cleanup:
    if (ddomain) {
        virUnrefDomain(ddomain);
        ret = 0;
    } else {
        ret = -1;
    }

    if (st)
        virUnrefStream(st);

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
    VIR_FREE(uri_out);
    VIR_FREE(cookiein);
    VIR_FREE(cookieout);

    return ret;
}


2098
static int doPeer2PeerMigrate(struct qemud_driver *driver,
2099
                              virConnectPtr sconn,
2100
                              virDomainObjPtr vm,
2101
                              const char *xmlin,
2102
                              const char *dconnuri,
2103 2104 2105
                              const char *uri,
                              unsigned long flags,
                              const char *dname,
2106 2107
                              unsigned long resource,
                              bool *v3proto)
2108 2109 2110 2111
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    bool p2p;
2112 2113
    virErrorPtr orig_err = NULL;

2114
    VIR_DEBUG("driver=%p, sconn=%p, vm=%p, xmlin=%s, dconnuri=%s, "
2115
              "uri=%s, flags=%lx, dname=%s, resource=%lu",
2116 2117
              driver, sconn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
              NULLSTR(uri), flags, NULLSTR(dname), resource);
2118 2119 2120 2121 2122 2123

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

    qemuDomainObjEnterRemoteWithDriver(driver, vm);
2124
    dconn = virConnectOpen(dconnuri);
2125 2126 2127
    qemuDomainObjExitRemoteWithDriver(driver, vm);
    if (dconn == NULL) {
        qemuReportError(VIR_ERR_OPERATION_FAILED,
2128
                        _("Failed to connect to remote libvirt URI %s"), dconnuri);
2129 2130 2131 2132 2133 2134
        return -1;
    }

    qemuDomainObjEnterRemoteWithDriver(driver, vm);
    p2p = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                   VIR_DRV_FEATURE_MIGRATION_P2P);
2135
        /* v3proto reflects whether the caller used Perform3, but with
2136
         * p2p migrate, regardless of whether Perform2 or Perform3
2137 2138 2139 2140
         * were used, we decide protocol based on what target supports
         */
    *v3proto = VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
                                        VIR_DRV_FEATURE_MIGRATION_V3);
2141
    qemuDomainObjExitRemoteWithDriver(driver, vm);
2142

2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155
    if (!p2p) {
        qemuReportError(VIR_ERR_OPERATION_FAILED, "%s",
                        _("Destination libvirt does not support peer-to-peer migration protocol"));
        goto cleanup;
    }

    /* domain may have been stopped while we were talking to remote daemon */
    if (!virDomainObjIsActive(vm)) {
        qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                        _("guest unexpectedly quit"));
        goto cleanup;
    }

2156 2157 2158 2159 2160 2161 2162
    /* 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;

2163
    if (*v3proto)
2164
        ret = doPeer2PeerMigrate3(driver, sconn, dconn, vm, xmlin,
2165
                                  dconnuri, uri, flags, dname, resource);
2166 2167
    else
        ret = doPeer2PeerMigrate2(driver, sconn, dconn, vm,
2168
                                  dconnuri, flags, dname, resource);
2169 2170

cleanup:
2171
    orig_err = virSaveLastError();
2172
    qemuDomainObjEnterRemoteWithDriver(driver, vm);
2173
    virConnectClose(dconn);
2174
    qemuDomainObjExitRemoteWithDriver(driver, vm);
2175 2176 2177 2178
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
2179 2180 2181 2182 2183

    return ret;
}


2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203
/*
 * 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
qemuMigrationPerformJob(struct qemud_driver *driver,
                        virConnectPtr conn,
                        virDomainObjPtr vm,
                        const char *xmlin,
                        const char *dconnuri,
                        const char *uri,
                        const char *cookiein,
                        int cookieinlen,
                        char **cookieout,
                        int *cookieoutlen,
                        unsigned long flags,
                        const char *dname,
                        unsigned long resource,
                        bool v3proto)
2204 2205 2206 2207 2208
{
    virDomainEventPtr event = NULL;
    int ret = -1;
    int resume = 0;

2209
    if (qemuMigrationJobStart(driver, vm, QEMU_ASYNC_JOB_MIGRATION_OUT) < 0)
2210 2211 2212 2213 2214 2215 2216 2217
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
        qemuReportError(VIR_ERR_OPERATION_INVALID,
                        "%s", _("domain is not running"));
        goto endjob;
    }

2218 2219
    if (!qemuMigrationIsAllowed(driver, vm, NULL))
        goto cleanup;
2220

J
Jiri Denemark 已提交
2221
    resume = virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING;
2222 2223

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
2224 2225 2226
        ret = doPeer2PeerMigrate(driver, conn, vm, xmlin,
                                 dconnuri, uri, flags, dname,
                                 resource, &v3proto);
2227
    } else {
2228 2229 2230
        qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM2);
        ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                              cookieout, cookieoutlen,
2231
                              flags, resource);
2232
    }
2233 2234
    if (ret < 0)
        goto endjob;
2235

2236 2237 2238 2239
    /*
     * In v3 protocol, the source VM is not killed off until the
     * confirm step.
     */
2240
    if (!v3proto) {
2241
        qemuProcessStop(driver, vm, 1, VIR_DOMAIN_SHUTOFF_MIGRATED);
2242
        virDomainAuditStop(vm, "migrated");
2243 2244 2245
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
2246
    }
2247
    resume = 0;
2248 2249

endjob:
J
Jiri Denemark 已提交
2250
    if (resume && virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
2251
        /* we got here through some sort of failure; start the domain again */
J
Jiri Denemark 已提交
2252
        if (qemuProcessStartCPUs(driver, vm, conn,
2253 2254
                                 VIR_DOMAIN_RUNNING_MIGRATION_CANCELED,
                                 QEMU_ASYNC_JOB_MIGRATION_OUT) < 0) {
2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266
            /* 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);
        }

        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319

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

cleanup:
    if (vm)
        virDomainObjUnlock(vm);
    if (event)
        qemuDomainEventQueue(driver, event);
    return ret;
}

/*
 * This implements perform phase of v3 migration protocol.
 */
static int
qemuMigrationPerformPhase(struct qemud_driver *driver,
                          virConnectPtr conn,
                          virDomainObjPtr vm,
                          const char *uri,
                          const char *cookiein,
                          int cookieinlen,
                          char **cookieout,
                          int *cookieoutlen,
                          unsigned long flags,
                          unsigned long resource)
{
    virDomainEventPtr event = NULL;
    int ret = -1;
    bool resume;
    int refs;

    /* 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);

    resume = virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING;
    ret = doNativeMigrate(driver, vm, uri, cookiein, cookieinlen,
                          cookieout, cookieoutlen,
2320
                          flags, resource);
2321 2322 2323 2324 2325

    if (ret < 0 && resume &&
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
        /* we got here through some sort of failure; start the domain again */
        if (qemuProcessStartCPUs(driver, vm, conn,
2326 2327
                                 VIR_DOMAIN_RUNNING_MIGRATION_CANCELED,
                                 QEMU_ASYNC_JOB_MIGRATION_OUT) < 0) {
2328 2329 2330 2331 2332 2333
            /* 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);
2334
        }
2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355

        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }

    if (ret < 0)
        goto endjob;

    qemuMigrationJobSetPhase(driver, vm, QEMU_MIGRATION_PHASE_PERFORM3_DONE);

endjob:
    if (ret < 0)
        refs = qemuMigrationJobFinish(driver, vm);
    else
        refs = qemuMigrationJobContinue(vm);
    if (refs == 0) {
        vm = NULL;
    } else if (!virDomainObjIsActive(vm) && !vm->persistent) {
        virDomainRemoveInactive(&driver->domains, vm);
        vm = NULL;
2356
    }
2357 2358 2359 2360 2361 2362 2363 2364 2365

cleanup:
    if (vm)
        virDomainObjUnlock(vm);
    if (event)
        qemuDomainEventQueue(driver, event);
    return ret;
}

2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411
int
qemuMigrationPerform(struct qemud_driver *driver,
                     virConnectPtr conn,
                     virDomainObjPtr vm,
                     const char *xmlin,
                     const char *dconnuri,
                     const char *uri,
                     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, "
              "uri=%s, cookiein=%s, cookieinlen=%d, cookieout=%p, "
              "cookieoutlen=%p, flags=%lx, dname=%s, resource=%lu, v3proto=%d",
              driver, conn, vm, NULLSTR(xmlin), NULLSTR(dconnuri),
              NULLSTR(uri), NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, NULLSTR(dname),
              resource, v3proto);

    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
        if (cookieinlen) {
            qemuReportError(VIR_ERR_OPERATION_INVALID,
                            "%s", _("received unexpected cookie with P2P migration"));
            return -1;
        }

        return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri, uri,
                                       cookiein, cookieinlen, cookieout,
                                       cookieoutlen, flags, dname, resource,
                                       v3proto);
    } else {
        if (dconnuri) {
            qemuReportError(VIR_ERR_INTERNAL_ERROR,
                            "%s", _("Unexpected dconnuri parameter with non-peer2peer migration"));
            return -1;
        }

        if (v3proto) {
            return qemuMigrationPerformPhase(driver, conn, vm, uri,
                                             cookiein, cookieinlen,
                                             cookieout, cookieoutlen,
2412
                                             flags, resource);
2413 2414 2415 2416 2417 2418 2419 2420
        } else {
            return qemuMigrationPerformJob(driver, conn, vm, xmlin, dconnuri,
                                           uri, cookiein, cookieinlen,
                                           cookieout, cookieoutlen, flags,
                                           dname, resource, v3proto);
        }
    }
}
2421 2422 2423 2424 2425 2426 2427 2428 2429 2430

#if WITH_MACVTAP
static void
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def) {
    int i;
    int last_good_net = -1;
    virDomainNetDefPtr net;

    for (i = 0; i < def->nnets; i++) {
        net = def->nets[i];
2431
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
2432 2433
            if (vpAssociatePortProfileId(net->ifname,
                                         net->mac,
2434 2435
                                         virDomainNetGetActualDirectDev(net),
                                         virDomainNetGetActualDirectVirtPortProfile(net),
2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447
                                         def->uuid,
                                         VIR_VM_OP_MIGRATE_IN_FINISH) != 0)
                goto err_exit;
        }
        last_good_net = i;
    }

    return;

err_exit:
    for (i = 0; i < last_good_net; i++) {
        net = def->nets[i];
2448
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
2449 2450
            vpDisassociatePortProfileId(net->ifname,
                                        net->mac,
2451 2452
                                        virDomainNetGetActualDirectDev(net),
                                        virDomainNetGetActualDirectVirtPortProfile(net),
2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466
                                        VIR_VM_OP_MIGRATE_IN_FINISH);
        }
    }
}
#else /* !WITH_MACVTAP */
static void
qemuMigrationVPAssociatePortProfiles(virDomainDefPtr def ATTRIBUTE_UNUSED) { }
#endif /* WITH_MACVTAP */


virDomainPtr
qemuMigrationFinish(struct qemud_driver *driver,
                    virConnectPtr dconn,
                    virDomainObjPtr vm,
2467 2468 2469 2470
                    const char *cookiein,
                    int cookieinlen,
                    char **cookieout,
                    int *cookieoutlen,
2471
                    unsigned long flags,
2472 2473
                    int retcode,
                    bool v3proto)
2474 2475 2476 2477
{
    virDomainPtr dom = NULL;
    virDomainEventPtr event = NULL;
    int newVM = 1;
2478
    qemuMigrationCookiePtr mig = NULL;
2479 2480
    virErrorPtr orig_err = NULL;

2481
    VIR_DEBUG("driver=%p, dconn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
2482
              "cookieout=%p, cookieoutlen=%p, flags=%lx, retcode=%d",
2483 2484
              driver, dconn, vm, NULLSTR(cookiein), cookieinlen,
              cookieout, cookieoutlen, flags, retcode);
2485

2486
    if (!qemuMigrationJobIsActive(vm, QEMU_ASYNC_JOB_MIGRATION_IN))
2487 2488
        goto cleanup;

2489 2490 2491
    qemuMigrationJobStartPhase(driver, vm,
                               v3proto ? QEMU_MIGRATION_PHASE_FINISH3
                                       : QEMU_MIGRATION_PHASE_FINISH2);
2492

2493 2494
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen, 0)))
        goto endjob;
2495 2496 2497 2498 2499 2500 2501 2502

    /* Did the migration go as planned?  If yes, return the domain
     * object, but if no, clean up the empty qemu process.
     */
    if (retcode == 0) {
        if (!virDomainObjIsActive(vm)) {
            qemuReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                            _("guest unexpectedly quit"));
2503
            goto endjob;
2504 2505 2506 2507 2508
        }

        qemuMigrationVPAssociatePortProfiles(vm->def);

        if (flags & VIR_MIGRATE_PERSIST_DEST) {
2509
            virDomainDefPtr vmdef;
2510 2511 2512
            if (vm->persistent)
                newVM = 0;
            vm->persistent = 1;
2513 2514
            vmdef = virDomainObjGetPersistentDef(driver->caps, vm);
            if (virDomainSaveConfig(driver->configDir, vmdef) < 0) {
2515 2516 2517 2518 2519 2520 2521 2522 2523
                /* 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.
                 */
2524 2525

                /*
2526 2527
                 * However, in v3 protocol, the source VM is still available
                 * to restart during confirm() step, so we kill it off now.
2528 2529 2530
                 */
                if (v3proto) {
                    qemuProcessStop(driver, vm, 1, VIR_DOMAIN_SHUTOFF_FAILED);
2531
                    virDomainAuditStop(vm, "failed");
2532 2533
                    if (newVM)
                        vm->persistent = 0;
2534
                }
2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552
                goto endjob;
            }

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

        if (!(flags & VIR_MIGRATE_PAUSED)) {
            /* 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
2553
            if (qemuProcessStartCPUs(driver, vm, dconn,
2554 2555
                                     VIR_DOMAIN_RUNNING_MIGRATED,
                                     QEMU_ASYNC_JOB_MIGRATION_IN) < 0) {
2556 2557 2558 2559 2560 2561 2562 2563
                if (virGetLastError() == NULL)
                    qemuReportError(VIR_ERR_INTERNAL_ERROR,
                                    "%s", _("resume operation failed"));
                /* Need to save the current error, in case shutting
                 * down the process overwrites it
                 */
                orig_err = virSaveLastError();

2564 2565 2566 2567 2568 2569 2570 2571 2572 2573
                /*
                 * 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) {
                    qemuProcessStop(driver, vm, 1, VIR_DOMAIN_SHUTOFF_FAILED);
2574
                    virDomainAuditStop(vm, "failed");
2575 2576 2577 2578
                    event = virDomainEventNewFromObj(vm,
                                                     VIR_DOMAIN_EVENT_STOPPED,
                                                     VIR_DOMAIN_EVENT_STOPPED_FAILED);
                }
2579 2580 2581 2582
                goto endjob;
            }
        }

2583 2584
        dom = virGetDomain (dconn, vm->def->name, vm->def->uuid);

2585 2586 2587
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
J
Jiri Denemark 已提交
2588 2589
        if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
            virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_USER);
2590 2591
            if (event)
                qemuDomainEventQueue(driver, event);
2592 2593 2594 2595 2596 2597 2598 2599
            event = virDomainEventNewFromObj(vm,
                                             VIR_DOMAIN_EVENT_SUSPENDED,
                                             VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);
        }
        if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
            goto endjob;
        }
2600 2601 2602

        /* Guest is successfully running, so cancel previous auto destroy */
        qemuProcessAutoDestroyRemove(driver, vm);
2603
    } else {
J
Jiri Denemark 已提交
2604
        qemuProcessStop(driver, vm, 1, VIR_DOMAIN_SHUTOFF_FAILED);
2605
        virDomainAuditStop(vm, "failed");
2606 2607 2608 2609 2610
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_FAILED);
    }

2611 2612 2613
    if (qemuMigrationBakeCookie(mig, driver, vm, cookieout, cookieoutlen, 0) < 0)
        VIR_WARN("Unable to encode migration cookie");

2614
endjob:
E
Eric Blake 已提交
2615 2616 2617 2618 2619
    if (qemuMigrationJobFinish(driver, vm) == 0) {
        vm = NULL;
    } else if (!vm->persistent && !virDomainObjIsActive(vm)) {
        virDomainRemoveInactive(&driver->domains, vm);
        vm = NULL;
2620
    }
2621 2622 2623 2624 2625 2626

cleanup:
    if (vm)
        virDomainObjUnlock(vm);
    if (event)
        qemuDomainEventQueue(driver, event);
2627
    qemuMigrationCookieFree(mig);
2628 2629 2630 2631
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
2632 2633
    return dom;
}
2634

2635 2636 2637 2638 2639 2640

int qemuMigrationConfirm(struct qemud_driver *driver,
                         virConnectPtr conn,
                         virDomainObjPtr vm,
                         const char *cookiein,
                         int cookieinlen,
E
Eric Blake 已提交
2641
                         unsigned int flags,
2642
                         int retcode)
2643 2644 2645 2646
{
    qemuMigrationCookiePtr mig;
    virDomainEventPtr event = NULL;
    int rv = -1;
2647
    VIR_DEBUG("driver=%p, conn=%p, vm=%p, cookiein=%s, cookieinlen=%d, "
2648
              "flags=%x, retcode=%d",
2649 2650
              driver, conn, vm, NULLSTR(cookiein), cookieinlen,
              flags, retcode);
2651

2652
    virCheckFlags(QEMU_MIGRATION_FLAGS, -1);
E
Eric Blake 已提交
2653

2654 2655 2656 2657 2658
    qemuMigrationJobSetPhase(driver, vm,
                             retcode == 0
                             ? QEMU_MIGRATION_PHASE_CONFIRM3
                             : QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED);

2659
    if (!(mig = qemuMigrationEatCookie(driver, vm, cookiein, cookieinlen, 0)))
2660 2661 2662 2663 2664 2665 2666
        return -1;

    /* Did the migration go as planned?  If yes, kill off the
     * domain object, but if no, resume CPUs
     */
    if (retcode == 0) {
        qemuProcessStop(driver, vm, 1, VIR_DOMAIN_SHUTOFF_MIGRATED);
2667
        virDomainAuditStop(vm, "migrated");
2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678

        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
    } else {

        /* 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
         */
        if (qemuProcessStartCPUs(driver, vm, conn,
2679 2680
                                 VIR_DOMAIN_RUNNING_MIGRATED,
                                 QEMU_ASYNC_JOB_MIGRATION_OUT) < 0) {
2681 2682 2683
            if (virGetLastError() == NULL)
                qemuReportError(VIR_ERR_INTERNAL_ERROR,
                                "%s", _("resume operation failed"));
2684
            goto cleanup;
2685 2686 2687 2688 2689 2690 2691
        }

        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
        if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
            VIR_WARN("Failed to save status on vm %s", vm->def->name);
2692
            goto cleanup;
2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705
        }
    }

    qemuMigrationCookieFree(mig);
    rv = 0;

cleanup:
    if (event)
        qemuDomainEventQueue(driver, event);
    return rv;
}


2706 2707 2708 2709 2710
/* Helper function called while driver lock is held and vm is active.  */
int
qemuMigrationToFile(struct qemud_driver *driver, virDomainObjPtr vm,
                    int fd, off_t offset, const char *path,
                    const char *compressor,
E
Eric Blake 已提交
2711
                    bool bypassSecurityDriver,
2712
                    enum qemuDomainAsyncJob asyncJob)
2713 2714 2715 2716 2717 2718
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCgroupPtr cgroup = NULL;
    int ret = -1;
    int rc;
    bool restoreLabel = false;
2719 2720
    virCommandPtr cmd = NULL;
    int pipeFD[2] = { -1, -1 };
2721

2722
    if (qemuCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
2723
        (!compressor || pipe(pipeFD) == 0)) {
2724
        /* All right! We can use fd migration, which means that qemu
2725 2726 2727
         * 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.  */
2728 2729
        if (virSecurityManagerSetImageFDLabel(driver->securityManager, vm,
                                              compressor ? pipeFD[1] : fd) < 0)
2730
            goto cleanup;
2731 2732 2733
        bypassSecurityDriver = true;
    } else {
        /* Phooey - we have to fall back on exec migration, where qemu
E
Eric Blake 已提交
2734 2735
         * has to popen() the file by name, and block devices have to be
         * given cgroup ACL permission.  We might also stumble on
2736 2737
         * a race present in some qemu versions where it does a wait()
         * that botches pclose.  */
E
Eric Blake 已提交
2738
        if (qemuCgroupControllerActive(driver,
2739 2740 2741 2742 2743 2744 2745 2746 2747 2748
                                       VIR_CGROUP_CONTROLLER_DEVICES)) {
            if (virCgroupForDomain(driver->cgroup, vm->def->name,
                                   &cgroup, 0) != 0) {
                qemuReportError(VIR_ERR_INTERNAL_ERROR,
                                _("Unable to find cgroup for %s"),
                                vm->def->name);
                goto cleanup;
            }
            rc = virCgroupAllowDevicePath(cgroup, path,
                                          VIR_CGROUP_DEVICE_RW);
2749
            virDomainAuditCgroupPath(vm, cgroup, "allow", path, "rw", rc);
E
Eric Blake 已提交
2750 2751 2752 2753
            if (rc == 1) {
                /* path was not a device, no further need for cgroup */
                virCgroupFree(&cgroup);
            } else if (rc < 0) {
2754 2755 2756 2757 2758
                virReportSystemError(-rc,
                                     _("Unable to allow device %s for %s"),
                                     path, vm->def->name);
                goto cleanup;
            }
2759
        }
2760 2761 2762
        if ((!bypassSecurityDriver) &&
            virSecurityManagerSetSavedStateLabel(driver->securityManager,
                                                 vm, path) < 0)
2763
            goto cleanup;
2764
        restoreLabel = true;
2765 2766
    }

2767
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
2768 2769
        goto cleanup;

2770 2771 2772
    if (!compressor) {
        const char *args[] = { "cat", NULL };

2773
        if (qemuCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD) &&
2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789
            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
        };
2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808
        if (pipeFD[0] != -1) {
            cmd = virCommandNewArgs(args);
            virCommandSetInputFD(cmd, pipeFD[0]);
            virCommandSetOutputFD(cmd, &fd);
            if (virSetCloseExec(pipeFD[1]) < 0) {
                virReportSystemError(errno, "%s",
                                     _("Unable to set cloexec flag"));
                qemuDomainObjExitMonitorWithDriver(driver, vm);
                goto cleanup;
            }
            if (virCommandRunAsync(cmd, NULL) < 0) {
                qemuDomainObjExitMonitorWithDriver(driver, vm);
                goto cleanup;
            }
            rc = qemuMonitorMigrateToFd(priv->mon,
                                        QEMU_MONITOR_MIGRATE_BACKGROUND,
                                        pipeFD[1]);
            if (VIR_CLOSE(pipeFD[0]) < 0 ||
                VIR_CLOSE(pipeFD[1]) < 0)
2809
                VIR_WARN("failed to close intermediate pipe");
2810 2811 2812 2813 2814
        } else {
            rc = qemuMonitorMigrateToFile(priv->mon,
                                          QEMU_MONITOR_MIGRATE_BACKGROUND,
                                          args, path, offset);
        }
2815
    }
2816
    qemuDomainObjExitMonitorWithDriver(driver, vm);
2817 2818 2819 2820

    if (rc < 0)
        goto cleanup;

2821
    rc = qemuMigrationWaitForCompletion(driver, vm, asyncJob);
2822 2823 2824 2825

    if (rc < 0)
        goto cleanup;

2826 2827 2828
    if (cmd && virCommandWait(cmd, NULL) < 0)
        goto cleanup;

2829 2830 2831
    ret = 0;

cleanup:
2832 2833 2834
    VIR_FORCE_CLOSE(pipeFD[0]);
    VIR_FORCE_CLOSE(pipeFD[1]);
    virCommandFree(cmd);
2835 2836 2837 2838 2839 2840 2841 2842
    if (restoreLabel && (!bypassSecurityDriver) &&
        virSecurityManagerRestoreSavedStateLabel(driver->securityManager,
                                                 vm, path) < 0)
        VIR_WARN("failed to restore save state label on %s", path);

    if (cgroup != NULL) {
        rc = virCgroupDenyDevicePath(cgroup, path,
                                     VIR_CGROUP_DEVICE_RWM);
2843
        virDomainAuditCgroupPath(vm, cgroup, "deny", path, "rwm", rc);
2844 2845 2846 2847 2848 2849 2850
        if (rc < 0)
            VIR_WARN("Unable to deny device %s for %s %d",
                     path, vm->def->name, rc);
        virCgroupFree(&cgroup);
    }
    return ret;
}
2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861

int
qemuMigrationJobStart(struct qemud_driver *driver,
                      virDomainObjPtr vm,
                      enum qemuDomainAsyncJob job)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

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

2862
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
2863
        qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
2864 2865
    } else {
        qemuDomainObjSetAsyncJobMask(vm, DEFAULT_JOB_MASK |
2866
                                     JOB_MASK(QEMU_JOB_SUSPEND) |
2867 2868
                                     JOB_MASK(QEMU_JOB_MIGRATION_OP));
    }
2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931

    priv->job.info.type = VIR_DOMAIN_JOB_UNBOUNDED;

    return 0;
}

void
qemuMigrationJobSetPhase(struct qemud_driver *driver,
                         virDomainObjPtr vm,
                         enum qemuMigrationJobPhase phase)
{
    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
qemuMigrationJobStartPhase(struct qemud_driver *driver,
                           virDomainObjPtr vm,
                           enum qemuMigrationJobPhase phase)
{
    virDomainObjRef(vm);
    qemuMigrationJobSetPhase(driver, vm, phase);
}

int
qemuMigrationJobContinue(virDomainObjPtr vm)
{
    return virDomainObjUnref(vm);
}

bool
qemuMigrationJobIsActive(virDomainObjPtr vm,
                         enum qemuDomainAsyncJob job)
{
    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");

        qemuReportError(VIR_ERR_OPERATION_INVALID, msg, vm->def->name);
        return false;
    }
    return true;
}

int
qemuMigrationJobFinish(struct qemud_driver *driver, virDomainObjPtr vm)
{
    return qemuDomainObjEndAsyncJob(driver, vm);
}