qemu_domain.c 69.0 KB
Newer Older
1 2 3
/*
 * qemu_domain.h: QEMU domain private state
 *
4
 * Copyright (C) 2006-2013 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16 17
 * Copyright (C) 2006 Daniel P. Berrange
 *
 * 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
18
 * License along with this library.  If not, see
O
Osier Yang 已提交
19
 * <http://www.gnu.org/licenses/>.
20 21 22 23 24 25 26 27
 *
 * Author: Daniel P. Berrange <berrange@redhat.com>
 */

#include <config.h>

#include "qemu_domain.h"
#include "qemu_command.h"
28
#include "qemu_capabilities.h"
29
#include "qemu_migration.h"
30
#include "viralloc.h"
31
#include "virlog.h"
32
#include "virerror.h"
33
#include "c-ctype.h"
34
#include "cpu/cpu.h"
35
#include "viruuid.h"
E
Eric Blake 已提交
36
#include "virfile.h"
37
#include "domain_event.h"
38
#include "virtime.h"
39
#include "virstoragefile.h"
40
#include "virstring.h"
41

42
#include <sys/time.h>
43
#include <fcntl.h>
44

45 46 47 48 49 50
#include <libxml/xpathInternals.h>

#define VIR_FROM_THIS VIR_FROM_QEMU

#define QEMU_NAMESPACE_HREF "http://libvirt.org/schemas/domain/qemu/1.0"

51 52 53 54 55 56
VIR_ENUM_IMPL(qemuDomainJob, QEMU_JOB_LAST,
              "none",
              "query",
              "destroy",
              "suspend",
              "modify",
57
              "abort",
58
              "migration operation",
59 60 61 62 63 64 65 66 67 68
              "none",   /* async job is never stored in job.active */
              "async nested",
);

VIR_ENUM_IMPL(qemuDomainAsyncJob, QEMU_ASYNC_JOB_LAST,
              "none",
              "migration out",
              "migration in",
              "save",
              "dump",
69
              "snapshot",
70 71
);

72

J
Jiri Denemark 已提交
73 74 75 76 77 78 79
const char *
qemuDomainAsyncJobPhaseToString(enum qemuDomainAsyncJob job,
                                int phase ATTRIBUTE_UNUSED)
{
    switch (job) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
    case QEMU_ASYNC_JOB_MIGRATION_IN:
80 81
        return qemuMigrationJobPhaseTypeToString(phase);

J
Jiri Denemark 已提交
82 83
    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
84
    case QEMU_ASYNC_JOB_SNAPSHOT:
J
Jiri Denemark 已提交
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102
    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        ; /* fall through */
    }

    return "none";
}

int
qemuDomainAsyncJobPhaseFromString(enum qemuDomainAsyncJob job,
                                  const char *phase)
{
    if (!phase)
        return 0;

    switch (job) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
    case QEMU_ASYNC_JOB_MIGRATION_IN:
103 104
        return qemuMigrationJobPhaseTypeFromString(phase);

J
Jiri Denemark 已提交
105 106
    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
107
    case QEMU_ASYNC_JOB_SNAPSHOT:
J
Jiri Denemark 已提交
108 109 110 111 112 113 114 115 116 117 118
    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        ; /* fall through */
    }

    if (STREQ(phase, "none"))
        return 0;
    else
        return -1;
}

119

120
void qemuDomainEventQueue(virQEMUDriverPtr driver,
121 122
                          virDomainEventPtr event)
{
123
    virDomainEventStateQueue(driver->domainEventState, event);
124 125 126
}


127 128 129 130 131 132 133 134
static int
qemuDomainObjInitJob(qemuDomainObjPrivatePtr priv)
{
    memset(&priv->job, 0, sizeof(priv->job));

    if (virCondInit(&priv->job.cond) < 0)
        return -1;

135
    if (virCondInit(&priv->job.asyncCond) < 0) {
136
        virCondDestroy(&priv->job.cond);
137 138 139
        return -1;
    }

140 141 142 143 144 145 146 147 148
    return 0;
}

static void
qemuDomainObjResetJob(qemuDomainObjPrivatePtr priv)
{
    struct qemuDomainJobObj *job = &priv->job;

    job->active = QEMU_JOB_NONE;
149
    job->owner = 0;
150 151 152 153 154 155 156 157
}

static void
qemuDomainObjResetAsyncJob(qemuDomainObjPrivatePtr priv)
{
    struct qemuDomainJobObj *job = &priv->job;

    job->asyncJob = QEMU_ASYNC_JOB_NONE;
158
    job->asyncOwner = 0;
J
Jiri Denemark 已提交
159
    job->phase = 0;
160
    job->mask = DEFAULT_JOB_MASK;
161
    job->start = 0;
162
    job->dump_memory_only = false;
163
    job->asyncAbort = false;
164
    memset(&job->status, 0, sizeof(job->status));
165 166 167
    memset(&job->info, 0, sizeof(job->info));
}

168 169 170 171 172 173 174 175
void
qemuDomainObjRestoreJob(virDomainObjPtr obj,
                        struct qemuDomainJobObj *job)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    memset(job, 0, sizeof(*job));
    job->active = priv->job.active;
176
    job->owner = priv->job.owner;
177
    job->asyncJob = priv->job.asyncJob;
178
    job->asyncOwner = priv->job.asyncOwner;
J
Jiri Denemark 已提交
179
    job->phase = priv->job.phase;
180 181 182 183 184

    qemuDomainObjResetJob(priv);
    qemuDomainObjResetAsyncJob(priv);
}

185 186 187 188 189
void
qemuDomainObjTransferJob(virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

190
    VIR_DEBUG("Changing job owner from %llu to %llu",
191 192 193 194
              priv->job.owner, virThreadSelfID());
    priv->job.owner = virThreadSelfID();
}

195 196 197
static void
qemuDomainObjFreeJob(qemuDomainObjPrivatePtr priv)
{
198 199
    virCondDestroy(&priv->job.cond);
    virCondDestroy(&priv->job.asyncCond);
200 201
}

202 203 204 205 206 207
static bool
qemuDomainTrackJob(enum qemuDomainJob job)
{
    return (QEMU_DOMAIN_TRACK_JOBS & JOB_MASK(job)) != 0;
}

208

209 210
static void *
qemuDomainObjPrivateAlloc(void)
211 212 213 214 215 216
{
    qemuDomainObjPrivatePtr priv;

    if (VIR_ALLOC(priv) < 0)
        return NULL;

217 218 219
    if (qemuDomainObjInitJob(priv) < 0) {
        virReportSystemError(errno, "%s",
                             _("Unable to init qemu driver mutexes"));
220
        goto error;
221
    }
222

223 224 225
    if (virCondInit(&priv->unplugFinished) < 0)
        goto error;

226
    if (!(priv->devs = virChrdevAlloc()))
227 228
        goto error;

229
    priv->migMaxBandwidth = QEMU_DOMAIN_MIG_BANDWIDTH_MAX;
230

231
    return priv;
232 233 234 235

error:
    VIR_FREE(priv);
    return NULL;
236 237
}

238 239
static void
qemuDomainObjPrivateFree(void *data)
240 241 242
{
    qemuDomainObjPrivatePtr priv = data;

243
    virObjectUnref(priv->qemuCaps);
244

245
    virCgroupFree(&priv->cgroup);
246
    qemuDomainPCIAddressSetFree(priv->pciaddrs);
247
    qemuDomainCCWAddressSetFree(priv->ccwaddrs);
248
    virDomainChrSourceDefFree(priv->monConfig);
249
    qemuDomainObjFreeJob(priv);
250
    VIR_FREE(priv->vcpupids);
251
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
252
    VIR_FREE(priv->origname);
253

254
    virCondDestroy(&priv->unplugFinished);
255
    virChrdevFree(priv->devs);
256

257 258
    /* This should never be non-NULL if we get here, but just in case... */
    if (priv->mon) {
259
        VIR_ERROR(_("Unexpected QEMU monitor still active during domain deletion"));
260 261
        qemuMonitorClose(priv->mon);
    }
D
Daniel P. Berrange 已提交
262 263 264 265
    if (priv->agent) {
        VIR_ERROR(_("Unexpected QEMU agent still active during domain deletion"));
        qemuAgentClose(priv->agent);
    }
266
    VIR_FREE(priv->cleanupCallbacks);
267 268 269 270
    VIR_FREE(priv);
}


271 272
static int
qemuDomainObjPrivateXMLFormat(virBufferPtr buf, void *data)
273 274 275
{
    qemuDomainObjPrivatePtr priv = data;
    const char *monitorpath;
276
    enum qemuDomainJob job;
277 278 279

    /* priv->monitor_chr is set only for qemu */
    if (priv->monConfig) {
280
        switch (priv->monConfig->type) {
281
        case VIR_DOMAIN_CHR_TYPE_UNIX:
282
            monitorpath = priv->monConfig->data.nix.path;
283 284 285
            break;
        default:
        case VIR_DOMAIN_CHR_TYPE_PTY:
286
            monitorpath = priv->monConfig->data.file.path;
287 288 289 290 291 292
            break;
        }

        virBufferEscapeString(buf, "  <monitor path='%s'", monitorpath);
        if (priv->monJSON)
            virBufferAddLit(buf, " json='1'");
293
        virBufferAsprintf(buf, " type='%s'/>\n",
294
                          virDomainChrTypeToString(priv->monConfig->type));
295 296 297 298
    }


    if (priv->nvcpupids) {
299
        size_t i;
300
        virBufferAddLit(buf, "  <vcpus>\n");
301
        for (i = 0; i < priv->nvcpupids; i++) {
302
            virBufferAsprintf(buf, "    <vcpu pid='%d'/>\n", priv->vcpupids[i]);
303 304 305 306
        }
        virBufferAddLit(buf, "  </vcpus>\n");
    }

307
    if (priv->qemuCaps) {
308
        size_t i;
309
        virBufferAddLit(buf, "  <qemuCaps>\n");
310
        for (i = 0; i < QEMU_CAPS_LAST; i++) {
311
            if (virQEMUCapsGet(priv->qemuCaps, i)) {
312
                virBufferAsprintf(buf, "    <flag name='%s'/>\n",
313
                                  virQEMUCapsTypeToString(i));
314 315 316 317 318
            }
        }
        virBufferAddLit(buf, "  </qemuCaps>\n");
    }

319 320 321
    if (priv->lockState)
        virBufferAsprintf(buf, "  <lockstate>%s</lockstate>\n", priv->lockState);

322 323 324 325
    job = priv->job.active;
    if (!qemuDomainTrackJob(job))
        priv->job.active = QEMU_JOB_NONE;

326
    if (priv->job.active || priv->job.asyncJob) {
J
Jiri Denemark 已提交
327
        virBufferAsprintf(buf, "  <job type='%s' async='%s'",
328 329
                          qemuDomainJobTypeToString(priv->job.active),
                          qemuDomainAsyncJobTypeToString(priv->job.asyncJob));
J
Jiri Denemark 已提交
330 331 332 333 334 335
        if (priv->job.phase) {
            virBufferAsprintf(buf, " phase='%s'",
                              qemuDomainAsyncJobPhaseToString(
                                    priv->job.asyncJob, priv->job.phase));
        }
        virBufferAddLit(buf, "/>\n");
336
    }
337
    priv->job.active = job;
338

339
    if (priv->fakeReboot)
340
        virBufferAddLit(buf, "  <fakereboot/>\n");
341

342 343 344 345 346 347 348 349 350 351
    if (priv->qemuDevices && *priv->qemuDevices) {
        char **tmp = priv->qemuDevices;
        virBufferAddLit(buf, "  <devices>\n");
        while (*tmp) {
            virBufferAsprintf(buf, "    <device alias='%s'/>\n", *tmp);
            tmp++;
        }
        virBufferAddLit(buf, "  </devices>\n");
    }

352 353 354
    return 0;
}

355 356
static int
qemuDomainObjPrivateXMLParse(xmlXPathContextPtr ctxt, void *data)
357 358 359 360
{
    qemuDomainObjPrivatePtr priv = data;
    char *monitorpath;
    char *tmp;
361 362
    int n;
    size_t i;
363
    xmlNodePtr *nodes = NULL;
364
    virQEMUCapsPtr qemuCaps = NULL;
365

366
    if (VIR_ALLOC(priv->monConfig) < 0)
367 368 369 370
        goto error;

    if (!(monitorpath =
          virXPathString("string(./monitor[1]/@path)", ctxt))) {
371 372
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("no monitor path"));
373 374 375 376 377
        goto error;
    }

    tmp = virXPathString("string(./monitor[1]/@type)", ctxt);
    if (tmp)
378
        priv->monConfig->type = virDomainChrTypeFromString(tmp);
379
    else
380
        priv->monConfig->type = VIR_DOMAIN_CHR_TYPE_PTY;
381 382
    VIR_FREE(tmp);

E
Eric Blake 已提交
383 384
    priv->monJSON = virXPathBoolean("count(./monitor[@json = '1']) > 0",
                                    ctxt) > 0;
385

386
    switch (priv->monConfig->type) {
387
    case VIR_DOMAIN_CHR_TYPE_PTY:
388
        priv->monConfig->data.file.path = monitorpath;
389 390
        break;
    case VIR_DOMAIN_CHR_TYPE_UNIX:
391
        priv->monConfig->data.nix.path = monitorpath;
392 393 394
        break;
    default:
        VIR_FREE(monitorpath);
395 396 397
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unsupported monitor type '%s'"),
                       virDomainChrTypeToString(priv->monConfig->type));
398 399 400 401 402 403 404 405
        goto error;
    }

    n = virXPathNodeSet("./vcpus/vcpu", ctxt, &nodes);
    if (n < 0)
        goto error;
    if (n) {
        priv->nvcpupids = n;
406
        if (VIR_REALLOC_N(priv->vcpupids, priv->nvcpupids) < 0)
407 408
            goto error;

409
        for (i = 0; i < n; i++) {
410 411 412 413 414 415 416 417 418 419 420 421 422
            char *pidstr = virXMLPropString(nodes[i], "pid");
            if (!pidstr)
                goto error;

            if (virStrToLong_i(pidstr, NULL, 10, &(priv->vcpupids[i])) < 0) {
                VIR_FREE(pidstr);
                goto error;
            }
            VIR_FREE(pidstr);
        }
        VIR_FREE(nodes);
    }

423
    if ((n = virXPathNodeSet("./qemuCaps/flag", ctxt, &nodes)) < 0) {
424 425
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("failed to parse qemu capabilities flags"));
426 427 428
        goto error;
    }
    if (n > 0) {
429
        if (!(qemuCaps = virQEMUCapsNew()))
430 431
            goto error;

432
        for (i = 0; i < n; i++) {
433 434
            char *str = virXMLPropString(nodes[i], "name");
            if (str) {
435
                int flag = virQEMUCapsTypeFromString(str);
436
                if (flag < 0) {
437 438
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("Unknown qemu capabilities flag %s"), str);
439
                    VIR_FREE(str);
440 441
                    goto error;
                }
442
                VIR_FREE(str);
443
                virQEMUCapsSet(qemuCaps, flag);
444 445 446
            }
        }

447
        priv->qemuCaps = qemuCaps;
448 449 450
    }
    VIR_FREE(nodes);

451
    priv->lockState = virXPathString("string(./lockstate)", ctxt);
452

453 454 455 456
    if ((tmp = virXPathString("string(./job[1]/@type)", ctxt))) {
        int type;

        if ((type = qemuDomainJobTypeFromString(tmp)) < 0) {
457 458
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unknown job type %s"), tmp);
459 460 461 462 463 464 465 466 467 468 469
            VIR_FREE(tmp);
            goto error;
        }
        VIR_FREE(tmp);
        priv->job.active = type;
    }

    if ((tmp = virXPathString("string(./job[1]/@async)", ctxt))) {
        int async;

        if ((async = qemuDomainAsyncJobTypeFromString(tmp)) < 0) {
470 471
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Unknown async job type %s"), tmp);
472 473 474 475 476
            VIR_FREE(tmp);
            goto error;
        }
        VIR_FREE(tmp);
        priv->job.asyncJob = async;
J
Jiri Denemark 已提交
477 478 479 480

        if ((tmp = virXPathString("string(./job[1]/@phase)", ctxt))) {
            priv->job.phase = qemuDomainAsyncJobPhaseFromString(async, tmp);
            if (priv->job.phase < 0) {
481 482
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Unknown job phase %s"), tmp);
J
Jiri Denemark 已提交
483 484 485 486 487
                VIR_FREE(tmp);
                goto error;
            }
            VIR_FREE(tmp);
        }
488 489
    }

490 491
    priv->fakeReboot = virXPathBoolean("boolean(./fakereboot)", ctxt) == 1;

492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512
    if ((n = virXPathNodeSet("./devices/device", ctxt, &nodes)) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("failed to parse qemu device list"));
        goto error;
    }
    if (n > 0) {
        /* NULL-terminated list */
        if (VIR_ALLOC_N(priv->qemuDevices, n + 1) < 0)
            goto error;

        for (i = 0; i < n; i++) {
            priv->qemuDevices[i] = virXMLPropString(nodes[i], "alias");
            if (!priv->qemuDevices[i]) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("failed to parse qemu device list"));
                goto error;
            }
        }
    }
    VIR_FREE(nodes);

513 514 515
    return 0;

error:
516
    virDomainChrSourceDefFree(priv->monConfig);
517 518
    priv->monConfig = NULL;
    VIR_FREE(nodes);
519 520
    virStringFreeList(priv->qemuDevices);
    priv->qemuDevices = NULL;
521
    virObjectUnref(qemuCaps);
522 523 524 525
    return -1;
}


526 527 528 529 530 531 532 533
virDomainXMLPrivateDataCallbacks virQEMUDriverPrivateDataCallbacks = {
    .alloc = qemuDomainObjPrivateAlloc,
    .free = qemuDomainObjPrivateFree,
    .parse = qemuDomainObjPrivateXMLParse,
    .format = qemuDomainObjPrivateXMLFormat,
};


534 535 536 537 538
static void
qemuDomainDefNamespaceFree(void *nsdata)
{
    qemuDomainCmdlineDefPtr cmd = nsdata;

539
    qemuDomainCmdlineDefFree(cmd);
540 541 542
}

static int
P
Philipp Hahn 已提交
543 544
qemuDomainDefNamespaceParse(xmlDocPtr xml ATTRIBUTE_UNUSED,
                            xmlNodePtr root ATTRIBUTE_UNUSED,
545 546 547 548
                            xmlXPathContextPtr ctxt,
                            void **data)
{
    qemuDomainCmdlineDefPtr cmd = NULL;
P
Philipp Hahn 已提交
549
    bool uses_qemu_ns = false;
550
    xmlNodePtr *nodes = NULL;
551 552
    int n;
    size_t i;
553

P
Philipp Hahn 已提交
554
    if (xmlXPathRegisterNs(ctxt, BAD_CAST "qemu", BAD_CAST QEMU_NAMESPACE_HREF) < 0) {
555 556 557
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Failed to register xml namespace '%s'"),
                       QEMU_NAMESPACE_HREF);
558 559 560
        return -1;
    }

561
    if (VIR_ALLOC(cmd) < 0)
562 563 564 565 566 567
        return -1;

    /* first handle the extra command-line arguments */
    n = virXPathNodeSet("./qemu:commandline/qemu:arg", ctxt, &nodes);
    if (n < 0)
        goto error;
P
Philipp Hahn 已提交
568
    uses_qemu_ns |= n > 0;
569 570

    if (n && VIR_ALLOC_N(cmd->args, n) < 0)
571
        goto error;
572 573 574 575

    for (i = 0; i < n; i++) {
        cmd->args[cmd->num_args] = virXMLPropString(nodes[i], "value");
        if (cmd->args[cmd->num_args] == NULL) {
576 577
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("No qemu command-line argument specified"));
578 579 580 581 582 583 584 585 586 587 588
            goto error;
        }
        cmd->num_args++;
    }

    VIR_FREE(nodes);

    /* now handle the extra environment variables */
    n = virXPathNodeSet("./qemu:commandline/qemu:env", ctxt, &nodes);
    if (n < 0)
        goto error;
P
Philipp Hahn 已提交
589
    uses_qemu_ns |= n > 0;
590 591

    if (n && VIR_ALLOC_N(cmd->env_name, n) < 0)
592
        goto error;
593 594

    if (n && VIR_ALLOC_N(cmd->env_value, n) < 0)
595
        goto error;
596 597 598 599 600 601

    for (i = 0; i < n; i++) {
        char *tmp;

        tmp = virXMLPropString(nodes[i], "name");
        if (tmp == NULL) {
602 603
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("No qemu environment name specified"));
604 605 606
            goto error;
        }
        if (tmp[0] == '\0') {
607 608
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Empty qemu environment name specified"));
609 610 611
            goto error;
        }
        if (!c_isalpha(tmp[0]) && tmp[0] != '_') {
612 613
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Invalid environment name, it must begin with a letter or underscore"));
614 615 616
            goto error;
        }
        if (strspn(tmp, "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789_") != strlen(tmp)) {
617 618
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Invalid environment name, it must contain only alphanumerics and underscore"));
619 620 621 622 623 624 625 626 627 628 629 630
            goto error;
        }

        cmd->env_name[cmd->num_env] = tmp;

        cmd->env_value[cmd->num_env] = virXMLPropString(nodes[i], "value");
        /* a NULL value for command is allowed, since it might be empty */
        cmd->num_env++;
    }

    VIR_FREE(nodes);

P
Philipp Hahn 已提交
631 632 633 634
    if (uses_qemu_ns)
        *data = cmd;
    else
        VIR_FREE(cmd);
635 636 637 638 639 640 641 642 643 644 645 646 647 648

    return 0;

error:
    VIR_FREE(nodes);
    qemuDomainDefNamespaceFree(cmd);
    return -1;
}

static int
qemuDomainDefNamespaceFormatXML(virBufferPtr buf,
                                void *nsdata)
{
    qemuDomainCmdlineDefPtr cmd = nsdata;
649
    size_t i;
650 651 652 653 654 655 656 657 658

    if (!cmd->num_args && !cmd->num_env)
        return 0;

    virBufferAddLit(buf, "  <qemu:commandline>\n");
    for (i = 0; i < cmd->num_args; i++)
        virBufferEscapeString(buf, "    <qemu:arg value='%s'/>\n",
                              cmd->args[i]);
    for (i = 0; i < cmd->num_env; i++) {
659
        virBufferAsprintf(buf, "    <qemu:env name='%s'", cmd->env_name[i]);
660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675
        if (cmd->env_value[i])
            virBufferEscapeString(buf, " value='%s'", cmd->env_value[i]);
        virBufferAddLit(buf, "/>\n");
    }
    virBufferAddLit(buf, "  </qemu:commandline>\n");

    return 0;
}

static const char *
qemuDomainDefNamespaceHref(void)
{
    return "xmlns:qemu='" QEMU_NAMESPACE_HREF "'";
}


676 677 678 679 680 681
virDomainXMLNamespace virQEMUDriverDomainXMLNamespace = {
    .parse = qemuDomainDefNamespaceParse,
    .free = qemuDomainDefNamespaceFree,
    .format = qemuDomainDefNamespaceFormatXML,
    .href = qemuDomainDefNamespaceHref,
};
682

683

684 685 686 687 688
static int
qemuDomainDefPostParse(virDomainDefPtr def,
                       virCapsPtr caps,
                       void *opaque ATTRIBUTE_UNUSED)
{
689
    bool addDefaultUSB = true;
690
    bool addImplicitSATA = false;
691
    bool addPCIRoot = false;
L
Laine Stump 已提交
692
    bool addPCIeRoot = false;
693
    bool addDefaultMemballoon = true;
694

695 696 697 698 699
    /* check for emulator and create a default one if needed */
    if (!def->emulator &&
        !(def->emulator = virDomainDefGetDefaultEmulator(def, caps)))
        return -1;

700 701 702 703 704 705
    /* Add implicit PCI root controller if the machine has one */
    switch (def->os.arch) {
    case VIR_ARCH_I686:
    case VIR_ARCH_X86_64:
        if (!def->os.machine)
            break;
L
Laine Stump 已提交
706
        if (STREQ(def->os.machine, "isapc")) {
707
            addDefaultUSB = false;
708
            break;
709
        }
L
Laine Stump 已提交
710 711 712 713
        if (STRPREFIX(def->os.machine, "pc-q35") ||
            STREQ(def->os.machine, "q35")) {
           addPCIeRoot = true;
           addDefaultUSB = false;
714
           addImplicitSATA = true;
L
Laine Stump 已提交
715 716
           break;
        }
717 718
        if (!STRPREFIX(def->os.machine, "pc-0.") &&
            !STRPREFIX(def->os.machine, "pc-1.") &&
719
            !STRPREFIX(def->os.machine, "pc-i440") &&
720 721 722 723 724 725
            !STREQ(def->os.machine, "pc") &&
            !STRPREFIX(def->os.machine, "rhel"))
            break;
        addPCIRoot = true;
        break;

726 727
    case VIR_ARCH_ARMV7L:
       addDefaultUSB = false;
728
       addDefaultMemballoon = false;
729 730
       break;

731 732 733 734 735 736 737 738 739 740 741 742
    case VIR_ARCH_ALPHA:
    case VIR_ARCH_PPC:
    case VIR_ARCH_PPC64:
    case VIR_ARCH_PPCEMB:
    case VIR_ARCH_SH4:
    case VIR_ARCH_SH4EB:
        addPCIRoot = true;
        break;
    default:
        break;
    }

743 744 745 746 747
    if (addDefaultUSB &&
        virDomainDefMaybeAddController(
            def, VIR_DOMAIN_CONTROLLER_TYPE_USB, 0, -1) < 0)
        return -1;

748 749 750 751 752
    if (addImplicitSATA &&
        virDomainDefMaybeAddController(
            def, VIR_DOMAIN_CONTROLLER_TYPE_SATA, 0, -1) < 0)
        return -1;

753 754 755 756 757 758
    if (addPCIRoot &&
        virDomainDefMaybeAddController(
            def, VIR_DOMAIN_CONTROLLER_TYPE_PCI, 0,
            VIR_DOMAIN_CONTROLLER_MODEL_PCI_ROOT) < 0)
        return -1;

759 760 761 762 763 764 765 766 767 768 769 770 771 772
    /* When a machine has a pcie-root, make sure that there is always
     * a dmi-to-pci-bridge controller added as bus 1, and a pci-bridge
     * as bus 2, so that standard PCI devices can be connected
     */
    if (addPCIeRoot) {
        if (virDomainDefMaybeAddController(
                def, VIR_DOMAIN_CONTROLLER_TYPE_PCI, 0,
                VIR_DOMAIN_CONTROLLER_MODEL_PCIE_ROOT) < 0 ||
            virDomainDefMaybeAddController(
                def, VIR_DOMAIN_CONTROLLER_TYPE_PCI, 1,
                VIR_DOMAIN_CONTROLLER_MODEL_DMI_TO_PCI_BRIDGE) < 0 ||
            virDomainDefMaybeAddController(
                def, VIR_DOMAIN_CONTROLLER_TYPE_PCI, 2,
                VIR_DOMAIN_CONTROLLER_MODEL_PCI_BRIDGE) < 0) {
L
Laine Stump 已提交
773
        return -1;
774 775
        }
    }
776

777
    if (addDefaultMemballoon && !def->memballoon) {
778 779 780 781 782 783 784 785
        virDomainMemballoonDefPtr memballoon;
        if (VIR_ALLOC(memballoon) < 0)
            return -1;

        memballoon->model = VIR_DOMAIN_MEMBALLOON_MODEL_VIRTIO;
        def->memballoon = memballoon;
    }

786 787 788
    return 0;
}

789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805
static const char *
qemuDomainDefaultNetModel(virDomainDefPtr def) {
    if (def->os.arch == VIR_ARCH_S390 ||
        def->os.arch == VIR_ARCH_S390X)
        return "virtio";

    if (def->os.arch == VIR_ARCH_ARMV7L) {
        if (STREQ(def->os.machine, "versatilepb"))
            return "smc91c111";

        /* Incomplete. vexpress (and a few others) use this, but not all
         * arm boards */
        return "lan9118";
    }

    return "rtl8139";
}
806

807 808
static int
qemuDomainDeviceDefPostParse(virDomainDeviceDefPtr dev,
809
                             virDomainDefPtr def,
810
                             virCapsPtr caps ATTRIBUTE_UNUSED,
811
                             void *opaque)
812
{
813 814 815 816
    int ret = -1;
    virQEMUDriverPtr driver = opaque;
    virQEMUDriverConfigPtr cfg = NULL;

817
    if (dev->type == VIR_DOMAIN_DEVICE_NET &&
818 819
        dev->data.net->type != VIR_DOMAIN_NET_TYPE_HOSTDEV &&
        !dev->data.net->model) {
820
        if (VIR_STRDUP(dev->data.net->model,
821
                       qemuDomainDefaultNetModel(def)) < 0)
822
            goto cleanup;
823
    }
824

825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845
    /* set default disk types and drivers */
    if (dev->type == VIR_DOMAIN_DEVICE_DISK) {
        virDomainDiskDefPtr disk = dev->data.disk;

        /* both of these require data from the driver config */
        if (driver && (cfg = virQEMUDriverGetConfig(driver))) {
            /* assign default storage format and driver according to config */
            if (cfg->allowDiskFormatProbing) {
                /* default disk format for drives */
                if (disk->format == VIR_STORAGE_FILE_NONE &&
                    (disk->type == VIR_DOMAIN_DISK_TYPE_FILE ||
                     disk->type == VIR_DOMAIN_DISK_TYPE_BLOCK))
                    disk->format = VIR_STORAGE_FILE_AUTO;

                 /* default disk format for mirrored drive */
                if (disk->mirror &&
                    disk->mirrorFormat == VIR_STORAGE_FILE_NONE)
                    disk->mirrorFormat = VIR_STORAGE_FILE_AUTO;
            } else {
                /* default driver if probing is forbidden */
                if (!disk->driverName &&
846 847
                    VIR_STRDUP(disk->driverName, "qemu") < 0)
                        goto cleanup;
848 849 850 851 852 853 854 855 856 857 858 859

                /* default disk format for drives */
                if (disk->format == VIR_STORAGE_FILE_NONE &&
                    (disk->type == VIR_DOMAIN_DISK_TYPE_FILE ||
                     disk->type == VIR_DOMAIN_DISK_TYPE_BLOCK))
                    disk->format = VIR_STORAGE_FILE_RAW;

                 /* default disk format for mirrored drive */
                if (disk->mirror &&
                    disk->mirrorFormat == VIR_STORAGE_FILE_NONE)
                    disk->mirrorFormat = VIR_STORAGE_FILE_RAW;
            }
860 861 862
        }
    }

863 864 865 866 867 868 869
    /* set the default console type for S390 arches */
    if (dev->type == VIR_DOMAIN_DEVICE_CHR &&
        dev->data.chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_CONSOLE &&
        dev->data.chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_NONE &&
        (def->os.arch == VIR_ARCH_S390 || def->os.arch == VIR_ARCH_S390X))
        dev->data.chr->targetType = VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_VIRTIO;

870 871 872 873 874 875 876 877
    /* set the default USB model to none for s390 unless an address is found */
    if (dev->type == VIR_DOMAIN_DEVICE_CONTROLLER &&
        dev->data.controller->type == VIR_DOMAIN_CONTROLLER_TYPE_USB &&
        dev->data.controller->model == -1 &&
        dev->data.controller->info.type == VIR_DOMAIN_DEVICE_ADDRESS_TYPE_NONE &&
        (def->os.arch == VIR_ARCH_S390 || def->os.arch == VIR_ARCH_S390X))
        dev->data.controller->model = VIR_DOMAIN_CONTROLLER_MODEL_USB_NONE;

878 879 880 881 882 883 884 885 886 887 888
    /* auto generate unix socket path */
    if (dev->type == VIR_DOMAIN_DEVICE_CHR &&
        dev->data.chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_CHANNEL &&
        dev->data.chr->targetType == VIR_DOMAIN_CHR_CHANNEL_TARGET_TYPE_VIRTIO &&
        dev->data.chr->source.type == VIR_DOMAIN_CHR_TYPE_UNIX &&
        !dev->data.chr->source.data.nix.path &&
        (driver && (cfg = virQEMUDriverGetConfig(driver)))) {

        if (virAsprintf(&dev->data.chr->source.data.nix.path,
                        "%s/channel/target/%s.%s",
                        cfg->libDir, def->name,
889
                        dev->data.chr->target.name) < 0)
890
            goto cleanup;
891 892 893
        dev->data.chr->source.data.nix.listen = true;
    }

894 895 896 897 898
    ret = 0;

cleanup:
    virObjectUnref(cfg);
    return ret;
899 900 901 902 903
}


virDomainDefParserConfig virQEMUDriverDomainDefParserConfig = {
    .devicesPostParseCallback = qemuDomainDeviceDefPostParse,
904
    .domainPostParseCallback = qemuDomainDefPostParse,
905 906 907
};


908
static void
909
qemuDomainObjSaveJob(virQEMUDriverPtr driver, virDomainObjPtr obj)
910
{
911 912 913
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);

    if (virDomainObjIsActive(obj)) {
914
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, obj) < 0)
915
            VIR_WARN("Failed to save status on vm %s", obj->def->name);
916
    }
917

918
    virObjectUnref(cfg);
919 920
}

J
Jiri Denemark 已提交
921
void
922
qemuDomainObjSetJobPhase(virQEMUDriverPtr driver,
J
Jiri Denemark 已提交
923 924 925 926
                         virDomainObjPtr obj,
                         int phase)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
927
    unsigned long long me = virThreadSelfID();
J
Jiri Denemark 已提交
928 929 930 931

    if (!priv->job.asyncJob)
        return;

932 933 934 935 936
    VIR_DEBUG("Setting '%s' phase to '%s'",
              qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
              qemuDomainAsyncJobPhaseToString(priv->job.asyncJob, phase));

    if (priv->job.asyncOwner && me != priv->job.asyncOwner) {
937
        VIR_WARN("'%s' async job is owned by thread %llu",
938 939 940 941
                 qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                 priv->job.asyncOwner);
    }

J
Jiri Denemark 已提交
942
    priv->job.phase = phase;
943
    priv->job.asyncOwner = me;
J
Jiri Denemark 已提交
944 945 946
    qemuDomainObjSaveJob(driver, obj);
}

947
void
948 949
qemuDomainObjSetAsyncJobMask(virDomainObjPtr obj,
                             unsigned long long allowedJobs)
950 951 952
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

953 954 955 956 957 958 959
    if (!priv->job.asyncJob)
        return;

    priv->job.mask = allowedJobs | JOB_MASK(QEMU_JOB_DESTROY);
}

void
960
qemuDomainObjDiscardAsyncJob(virQEMUDriverPtr driver, virDomainObjPtr obj)
961 962 963 964 965 966
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    if (priv->job.active == QEMU_JOB_ASYNC_NESTED)
        qemuDomainObjResetJob(priv);
    qemuDomainObjResetAsyncJob(priv);
967
    qemuDomainObjSaveJob(driver, obj);
968 969
}

970 971 972 973 974 975 976 977 978
void
qemuDomainObjReleaseAsyncJob(virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    VIR_DEBUG("Releasing ownership of '%s' async job",
              qemuDomainAsyncJobTypeToString(priv->job.asyncJob));

    if (priv->job.asyncOwner != virThreadSelfID()) {
979
        VIR_WARN("'%s' async job is owned by thread %llu",
980 981 982 983 984 985
                 qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                 priv->job.asyncOwner);
    }
    priv->job.asyncOwner = 0;
}

986
static bool
987
qemuDomainNestedJobAllowed(qemuDomainObjPrivatePtr priv, enum qemuDomainJob job)
988 989
{
    return !priv->job.asyncJob || (priv->job.mask & JOB_MASK(job)) != 0;
990 991
}

992 993 994 995 996 997
bool
qemuDomainJobAllowed(qemuDomainObjPrivatePtr priv, enum qemuDomainJob job)
{
    return !priv->job.active && qemuDomainNestedJobAllowed(priv, job);
}

998 999 1000
/* Give up waiting for mutex after 30 seconds */
#define QEMU_JOB_WAIT_TIME (1000ull * 30)

1001
/*
1002
 * obj must be locked before calling
1003
 */
1004
static int ATTRIBUTE_NONNULL(1)
1005
qemuDomainObjBeginJobInternal(virQEMUDriverPtr driver,
1006 1007 1008
                              virDomainObjPtr obj,
                              enum qemuDomainJob job,
                              enum qemuDomainAsyncJob asyncJob)
1009 1010
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
J
Jiri Denemark 已提交
1011
    unsigned long long now;
1012
    unsigned long long then;
1013
    bool nested = job == QEMU_JOB_ASYNC_NESTED;
1014
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1015

1016 1017
    priv->jobs_queued++;

1018 1019
    if (virTimeMillisNow(&now) < 0) {
        virObjectUnref(cfg);
1020
        return -1;
1021 1022
    }

J
Jiri Denemark 已提交
1023
    then = now + QEMU_JOB_WAIT_TIME;
1024

1025
    virObjectRef(obj);
1026

1027
retry:
1028 1029
    if (cfg->maxQueuedJobs &&
        priv->jobs_queued > cfg->maxQueuedJobs) {
1030 1031 1032
        goto error;
    }

1033
    while (!nested && !qemuDomainNestedJobAllowed(priv, job)) {
1034
        if (virCondWaitUntil(&priv->job.asyncCond, &obj->parent.lock, then) < 0)
1035 1036 1037
            goto error;
    }

1038
    while (priv->job.active) {
1039
        if (virCondWaitUntil(&priv->job.cond, &obj->parent.lock, then) < 0)
1040
            goto error;
1041
    }
1042 1043 1044

    /* No job is active but a new async job could have been started while obj
     * was unlocked, so we need to recheck it. */
1045
    if (!nested && !qemuDomainNestedJobAllowed(priv, job))
1046 1047
        goto retry;

1048
    qemuDomainObjResetJob(priv);
1049 1050

    if (job != QEMU_JOB_ASYNC) {
1051 1052 1053
        VIR_DEBUG("Starting job: %s (async=%s)",
                   qemuDomainJobTypeToString(job),
                   qemuDomainAsyncJobTypeToString(priv->job.asyncJob));
1054
        priv->job.active = job;
1055
        priv->job.owner = virThreadSelfID();
1056
    } else {
1057 1058
        VIR_DEBUG("Starting async job: %s",
                  qemuDomainAsyncJobTypeToString(asyncJob));
1059 1060
        qemuDomainObjResetAsyncJob(priv);
        priv->job.asyncJob = asyncJob;
1061
        priv->job.asyncOwner = virThreadSelfID();
1062 1063
        priv->job.start = now;
    }
1064

1065 1066
    if (qemuDomainTrackJob(job))
        qemuDomainObjSaveJob(driver, obj);
1067

1068
    virObjectUnref(cfg);
1069
    return 0;
1070 1071

error:
1072
    VIR_WARN("Cannot start job (%s, %s) for domain %s;"
1073
             " current job is (%s, %s) owned by (%llu, %llu)",
1074 1075 1076 1077 1078 1079 1080
             qemuDomainJobTypeToString(job),
             qemuDomainAsyncJobTypeToString(asyncJob),
             obj->def->name,
             qemuDomainJobTypeToString(priv->job.active),
             qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
             priv->job.owner, priv->job.asyncOwner);

1081
    if (errno == ETIMEDOUT)
1082 1083
        virReportError(VIR_ERR_OPERATION_TIMEOUT,
                       "%s", _("cannot acquire state change lock"));
1084 1085
    else if (cfg->maxQueuedJobs &&
             priv->jobs_queued > cfg->maxQueuedJobs)
1086 1087 1088
        virReportError(VIR_ERR_OPERATION_FAILED,
                       "%s", _("cannot acquire state change lock "
                               "due to max_queued limit"));
1089 1090 1091
    else
        virReportSystemError(errno,
                             "%s", _("cannot acquire job mutex"));
1092
    priv->jobs_queued--;
1093
    virObjectUnref(obj);
1094
    virObjectUnref(cfg);
1095
    return -1;
1096 1097 1098
}

/*
1099
 * obj must be locked before calling
1100 1101 1102 1103 1104 1105 1106
 *
 * This must be called by anything that will change the VM state
 * in any way, or anything that will use the QEMU monitor.
 *
 * Upon successful return, the object will have its ref count increased,
 * successful calls must be followed by EndJob eventually
 */
1107
int qemuDomainObjBeginJob(virQEMUDriverPtr driver,
1108 1109
                          virDomainObjPtr obj,
                          enum qemuDomainJob job)
1110
{
1111
    return qemuDomainObjBeginJobInternal(driver, obj, job,
1112 1113 1114
                                         QEMU_ASYNC_JOB_NONE);
}

1115
int qemuDomainObjBeginAsyncJob(virQEMUDriverPtr driver,
1116
                               virDomainObjPtr obj,
1117
                               enum qemuDomainAsyncJob asyncJob)
1118
{
1119
    return qemuDomainObjBeginJobInternal(driver, obj, QEMU_JOB_ASYNC,
1120
                                         asyncJob);
1121 1122
}

1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136
int
qemuDomainObjBeginNestedJob(virQEMUDriverPtr driver,
                            virDomainObjPtr obj,
                            enum qemuDomainAsyncJob asyncJob)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    if (asyncJob != priv->job.asyncJob) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("unexpected async job %d"), asyncJob);
        return -1;
    }

    if (priv->job.asyncOwner != virThreadSelfID()) {
1137
        VIR_WARN("This thread doesn't seem to be the async job owner: %llu",
1138 1139 1140 1141 1142 1143 1144 1145
                 priv->job.asyncOwner);
    }

    return qemuDomainObjBeginJobInternal(driver, obj,
                                         QEMU_JOB_ASYNC_NESTED,
                                         QEMU_ASYNC_JOB_NONE);
}

1146

1147
/*
1148
 * obj must be locked before calling
1149 1150 1151 1152
 *
 * To be called after completing the work associated with the
 * earlier qemuDomainBeginJob() call
 *
1153 1154
 * Returns true if @obj was still referenced, false if it was
 * disposed of.
1155
 */
1156
bool qemuDomainObjEndJob(virQEMUDriverPtr driver, virDomainObjPtr obj)
1157 1158
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
1159
    enum qemuDomainJob job = priv->job.active;
1160

1161 1162
    priv->jobs_queued--;

1163
    VIR_DEBUG("Stopping job: %s (async=%s)",
1164
              qemuDomainJobTypeToString(job),
1165 1166
              qemuDomainAsyncJobTypeToString(priv->job.asyncJob));

1167
    qemuDomainObjResetJob(priv);
1168 1169
    if (qemuDomainTrackJob(job))
        qemuDomainObjSaveJob(driver, obj);
1170
    virCondSignal(&priv->job.cond);
1171

1172
    return virObjectUnref(obj);
1173 1174
}

1175
bool
1176
qemuDomainObjEndAsyncJob(virQEMUDriverPtr driver, virDomainObjPtr obj)
1177 1178
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
1179

1180 1181
    priv->jobs_queued--;

1182 1183 1184
    VIR_DEBUG("Stopping async job: %s",
              qemuDomainAsyncJobTypeToString(priv->job.asyncJob));

1185
    qemuDomainObjResetAsyncJob(priv);
1186
    qemuDomainObjSaveJob(driver, obj);
1187 1188
    virCondBroadcast(&priv->job.asyncCond);

1189
    return virObjectUnref(obj);
1190 1191
}

1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
void
qemuDomainObjAbortAsyncJob(virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    VIR_DEBUG("Requesting abort of async job: %s",
              qemuDomainAsyncJobTypeToString(priv->job.asyncJob));

    priv->job.asyncAbort = true;
}

1203 1204 1205 1206 1207 1208 1209 1210 1211
/*
 * obj must be locked before calling
 *
 * To be called immediately before any QEMU monitor API call
 * Must have already either called qemuDomainObjBeginJob() and checked
 * that the VM is still active; may not be used for nested async jobs.
 *
 * To be followed with qemuDomainObjExitMonitor() once complete
 */
1212
static int
1213
qemuDomainObjEnterMonitorInternal(virQEMUDriverPtr driver,
1214 1215
                                  virDomainObjPtr obj,
                                  enum qemuDomainAsyncJob asyncJob)
1216 1217 1218
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

1219
    if (asyncJob != QEMU_ASYNC_JOB_NONE) {
1220
        if (qemuDomainObjBeginNestedJob(driver, obj, asyncJob) < 0)
1221 1222
            return -1;
        if (!virDomainObjIsActive(obj)) {
1223 1224
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("domain is no longer running"));
1225 1226
            /* Still referenced by the containing async job.  */
            ignore_value(qemuDomainObjEndJob(driver, obj));
1227 1228
            return -1;
        }
1229 1230 1231
    } else if (priv->job.asyncOwner == virThreadSelfID()) {
        VIR_WARN("This thread seems to be the async job owner; entering"
                 " monitor without asking for a nested job is dangerous");
1232 1233
    }

1234
    virObjectLock(priv->mon);
1235
    virObjectRef(priv->mon);
1236
    ignore_value(virTimeMillisNow(&priv->monStart));
1237
    virObjectUnlock(obj);
1238 1239

    return 0;
1240 1241
}

1242
static void ATTRIBUTE_NONNULL(1)
1243
qemuDomainObjExitMonitorInternal(virQEMUDriverPtr driver,
1244
                                 virDomainObjPtr obj)
1245 1246
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
1247
    bool hasRefs;
1248

1249
    hasRefs = virObjectUnref(priv->mon);
1250

1251
    if (hasRefs)
1252
        virObjectUnlock(priv->mon);
1253

1254
    virObjectLock(obj);
1255

1256
    priv->monStart = 0;
1257
    if (!hasRefs)
1258
        priv->mon = NULL;
1259

1260 1261 1262 1263 1264
    if (priv->job.active == QEMU_JOB_ASYNC_NESTED) {
        qemuDomainObjResetJob(priv);
        qemuDomainObjSaveJob(driver, obj);
        virCondSignal(&priv->job.cond);

1265
        virObjectUnref(obj);
1266
    }
1267 1268
}

1269
void qemuDomainObjEnterMonitor(virQEMUDriverPtr driver,
1270
                               virDomainObjPtr obj)
1271
{
1272
    ignore_value(qemuDomainObjEnterMonitorInternal(driver, obj,
1273
                                                   QEMU_ASYNC_JOB_NONE));
1274 1275
}

1276
/* obj must NOT be locked before calling
1277 1278 1279
 *
 * Should be paired with an earlier qemuDomainObjEnterMonitor() call
 */
1280
void qemuDomainObjExitMonitor(virQEMUDriverPtr driver,
1281
                              virDomainObjPtr obj)
1282
{
1283
    qemuDomainObjExitMonitorInternal(driver, obj);
1284
}
1285 1286

/*
1287
 * obj must be locked before calling
1288 1289
 *
 * To be called immediately before any QEMU monitor API call.
1290
 * Must have already either called qemuDomainObjBeginJob()
1291 1292 1293 1294 1295
 * and checked that the VM is still active, with asyncJob of
 * QEMU_ASYNC_JOB_NONE; or already called qemuDomainObjBeginAsyncJob,
 * with the same asyncJob.
 *
 * Returns 0 if job was started, in which case this must be followed with
1296
 * qemuDomainObjExitMonitor(); or -1 if the job could not be
1297 1298 1299
 * started (probably because the vm exited in the meantime).
 */
int
1300
qemuDomainObjEnterMonitorAsync(virQEMUDriverPtr driver,
1301 1302
                               virDomainObjPtr obj,
                               enum qemuDomainAsyncJob asyncJob)
1303
{
1304
    return qemuDomainObjEnterMonitorInternal(driver, obj, asyncJob);
1305 1306
}

D
Daniel P. Berrange 已提交
1307 1308


1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319
/*
 * obj must be locked before calling
 *
 * To be called immediately before any QEMU agent API call.
 * Must have already called qemuDomainObjBeginJob() and checked
 * that the VM is still active.
 *
 * To be followed with qemuDomainObjExitAgent() once complete
 */
void
qemuDomainObjEnterAgent(virDomainObjPtr obj)
D
Daniel P. Berrange 已提交
1320 1321 1322
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

1323
    virObjectLock(priv->agent);
1324
    virObjectRef(priv->agent);
D
Daniel P. Berrange 已提交
1325
    ignore_value(virTimeMillisNow(&priv->agentStart));
1326
    virObjectUnlock(obj);
D
Daniel P. Berrange 已提交
1327 1328
}

1329 1330 1331 1332 1333 1334 1335

/* obj must NOT be locked before calling
 *
 * Should be paired with an earlier qemuDomainObjEnterAgent() call
 */
void
qemuDomainObjExitAgent(virDomainObjPtr obj)
D
Daniel P. Berrange 已提交
1336 1337
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
1338
    bool hasRefs;
D
Daniel P. Berrange 已提交
1339

1340
    hasRefs = virObjectUnref(priv->agent);
D
Daniel P. Berrange 已提交
1341

1342
    if (hasRefs)
1343
        virObjectUnlock(priv->agent);
D
Daniel P. Berrange 已提交
1344

1345
    virObjectLock(obj);
D
Daniel P. Berrange 已提交
1346 1347

    priv->agentStart = 0;
1348
    if (!hasRefs)
D
Daniel P. Berrange 已提交
1349 1350 1351
        priv->agent = NULL;
}

1352
void qemuDomainObjEnterRemote(virDomainObjPtr obj)
1353
{
1354
    virObjectRef(obj);
1355
    virObjectUnlock(obj);
1356 1357
}

1358
void qemuDomainObjExitRemote(virDomainObjPtr obj)
1359
{
1360
    virObjectLock(obj);
1361
    virObjectUnref(obj);
1362
}
1363 1364


1365 1366 1367 1368 1369 1370 1371 1372
virDomainDefPtr
qemuDomainDefCopy(virQEMUDriverPtr driver,
                  virDomainDefPtr src,
                  unsigned int flags)
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;
    virDomainDefPtr ret = NULL;
    virCapsPtr caps = NULL;
1373
    char *xml = NULL;
1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393

    if (qemuDomainDefFormatBuf(driver, src, flags, &buf) < 0)
        goto cleanup;

    xml = virBufferContentAndReset(&buf);

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

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

cleanup:
    VIR_FREE(xml);
    virObjectUnref(caps);
    return ret;
}

1394
int
1395
qemuDomainDefFormatBuf(virQEMUDriverPtr driver,
1396 1397 1398
                       virDomainDefPtr def,
                       unsigned int flags,
                       virBuffer *buf)
1399
{
1400
    int ret = -1;
1401
    virCPUDefPtr cpu = NULL;
1402
    virCPUDefPtr def_cpu = def->cpu;
1403 1404
    virDomainControllerDefPtr *controllers = NULL;
    int ncontrollers = 0;
1405 1406 1407 1408
    virCapsPtr caps = NULL;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;
1409 1410

    /* Update guest CPU requirements according to host CPU */
1411 1412 1413
    if ((flags & VIR_DOMAIN_XML_UPDATE_CPU) &&
        def_cpu &&
        (def_cpu->mode != VIR_CPU_MODE_CUSTOM || def_cpu->model)) {
1414 1415
        if (!caps->host.cpu ||
            !caps->host.cpu->model) {
1416 1417
            virReportError(VIR_ERR_OPERATION_FAILED,
                           "%s", _("cannot get host CPU capabilities"));
1418 1419 1420
            goto cleanup;
        }

1421
        if (!(cpu = virCPUDefCopy(def_cpu)) ||
1422
            cpuUpdate(cpu, caps->host.cpu) < 0)
1423 1424 1425 1426
            goto cleanup;
        def->cpu = cpu;
    }

1427
    if ((flags & VIR_DOMAIN_XML_MIGRATABLE)) {
1428
        size_t i;
1429
        int toremove = 0;
1430
        virDomainControllerDefPtr usb = NULL, pci = NULL;
1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448

        /* If only the default USB controller is present, we can remove it
         * and make the XML compatible with older versions of libvirt which
         * didn't support USB controllers in the XML but always added the
         * default one to qemu anyway.
         */
        for (i = 0; i < def->ncontrollers; i++) {
            if (def->controllers[i]->type == VIR_DOMAIN_CONTROLLER_TYPE_USB) {
                if (usb) {
                    usb = NULL;
                    break;
                }
                usb = def->controllers[i];
            }
        }
        if (usb && usb->idx == 0 && usb->model == -1) {
            VIR_DEBUG("Removing default USB controller from domain '%s'"
                      " for migration compatibility", def->name);
1449
            toremove++;
1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467
        } else {
            usb = NULL;
        }

        /* Remove the default PCI controller if there is only one present
         * and its model is pci-root */
        for (i = 0; i < def->ncontrollers; i++) {
            if (def->controllers[i]->type == VIR_DOMAIN_CONTROLLER_TYPE_PCI) {
                if (pci) {
                    pci = NULL;
                    break;
                }
                pci = def->controllers[i];
            }
        }

        if (pci && pci->idx == 0 &&
            pci->model == VIR_DOMAIN_CONTROLLER_MODEL_PCI_ROOT) {
L
Laine Stump 已提交
1468
            VIR_DEBUG("Removing default pci-root from domain '%s'"
1469
                      " for migration compatibility", def->name);
1470
            toremove++;
1471 1472 1473 1474
        } else {
            pci = NULL;
        }

1475
        if (toremove) {
1476 1477
            controllers = def->controllers;
            ncontrollers = def->ncontrollers;
1478
            if (VIR_ALLOC_N(def->controllers, ncontrollers - toremove) < 0) {
1479 1480 1481 1482 1483 1484
                controllers = NULL;
                goto cleanup;
            }

            def->ncontrollers = 0;
            for (i = 0; i < ncontrollers; i++) {
1485
                if (controllers[i] != usb && controllers[i] != pci)
1486 1487 1488
                    def->controllers[def->ncontrollers++] = controllers[i];
            }
        }
1489 1490


1491 1492
    }

1493
    ret = virDomainDefFormatInternal(def, flags, buf);
1494 1495 1496 1497

cleanup:
    def->cpu = def_cpu;
    virCPUDefFree(cpu);
1498 1499 1500 1501 1502
    if (controllers) {
        VIR_FREE(def->controllers);
        def->controllers = controllers;
        def->ncontrollers = ncontrollers;
    }
1503
    virObjectUnref(caps);
1504 1505
    return ret;
}
1506

1507
char *qemuDomainDefFormatXML(virQEMUDriverPtr driver,
1508
                             virDomainDefPtr def,
1509
                             unsigned int flags)
1510 1511 1512
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;

1513
    if (qemuDomainDefFormatBuf(driver, def, flags, &buf) < 0) {
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526
        virBufferFreeAndReset(&buf);
        return NULL;
    }

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

    return virBufferContentAndReset(&buf);
}

1527
char *qemuDomainFormatXML(virQEMUDriverPtr driver,
1528
                          virDomainObjPtr vm,
1529
                          unsigned int flags)
1530 1531 1532 1533 1534 1535 1536 1537
{
    virDomainDefPtr def;

    if ((flags & VIR_DOMAIN_XML_INACTIVE) && vm->newDef)
        def = vm->newDef;
    else
        def = vm->def;

1538
    return qemuDomainDefFormatXML(driver, def, flags);
1539 1540
}

1541
char *
1542
qemuDomainDefFormatLive(virQEMUDriverPtr driver,
1543
                        virDomainDefPtr def,
1544 1545
                        bool inactive,
                        bool compatible)
1546 1547 1548 1549 1550
{
    unsigned int flags = QEMU_DOMAIN_FORMAT_LIVE_FLAGS;

    if (inactive)
        flags |= VIR_DOMAIN_XML_INACTIVE;
1551 1552
    if (compatible)
        flags |= VIR_DOMAIN_XML_MIGRATABLE;
1553

1554
    return qemuDomainDefFormatXML(driver, def, flags);
1555 1556
}

1557

1558
void qemuDomainObjTaint(virQEMUDriverPtr driver,
1559
                        virDomainObjPtr obj,
1560 1561
                        enum virDomainTaintFlags taint,
                        int logFD)
1562
{
1563 1564
    virErrorPtr orig_err = NULL;

1565 1566 1567 1568 1569 1570 1571 1572 1573
    if (virDomainObjTaint(obj, taint)) {
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(obj->def->uuid, uuidstr);

        VIR_WARN("Domain id=%d name='%s' uuid=%s is tainted: %s",
                 obj->def->id,
                 obj->def->name,
                 uuidstr,
                 virDomainTaintTypeToString(taint));
1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587

        /* We don't care about errors logging taint info, so
         * preserve original error, and clear any error that
         * is raised */
        orig_err = virSaveLastError();
        if (qemuDomainAppendLog(driver, obj, logFD,
                                "Domain id=%d is tainted: %s\n",
                                obj->def->id,
                                virDomainTaintTypeToString(taint)) < 0)
            virResetLastError();
        if (orig_err) {
            virSetError(orig_err);
            virFreeError(orig_err);
        }
1588 1589 1590 1591
    }
}


1592
void qemuDomainObjCheckTaint(virQEMUDriverPtr driver,
1593 1594
                             virDomainObjPtr obj,
                             int logFD)
1595
{
1596
    size_t i;
1597
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1598

1599 1600 1601 1602
    if (cfg->privileged &&
        (!cfg->clearEmulatorCapabilities ||
         cfg->user == 0 ||
         cfg->group == 0))
1603
        qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_HIGH_PRIVILEGES, logFD);
1604 1605 1606 1607

    if (obj->def->namespaceData) {
        qemuDomainCmdlineDefPtr qemucmd = obj->def->namespaceData;
        if (qemucmd->num_args || qemucmd->num_env)
1608
            qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_CUSTOM_ARGV, logFD);
1609 1610
    }

1611 1612 1613
    if (obj->def->cpu && obj->def->cpu->mode == VIR_CPU_MODE_HOST_PASSTHROUGH)
        qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_HOST_CPU, logFD);

1614
    for (i = 0; i < obj->def->ndisks; i++)
1615
        qemuDomainObjCheckDiskTaint(driver, obj, obj->def->disks[i], logFD);
1616

1617
    for (i = 0; i < obj->def->nnets; i++)
1618
        qemuDomainObjCheckNetTaint(driver, obj, obj->def->nets[i], logFD);
1619 1620

    virObjectUnref(cfg);
1621 1622 1623
}


1624
void qemuDomainObjCheckDiskTaint(virQEMUDriverPtr driver,
1625
                                 virDomainObjPtr obj,
1626 1627
                                 virDomainDiskDefPtr disk,
                                 int logFD)
1628
{
1629 1630
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);

1631
    if ((!disk->format || disk->format == VIR_STORAGE_FILE_AUTO) &&
1632
        cfg->allowDiskFormatProbing)
1633
        qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_DISK_PROBING, logFD);
1634

1635
    if (disk->rawio == 1)
1636
        qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_HIGH_PRIVILEGES, logFD);
1637 1638

    virObjectUnref(cfg);
1639 1640 1641
}


1642
void qemuDomainObjCheckNetTaint(virQEMUDriverPtr driver,
1643
                                virDomainObjPtr obj,
1644 1645
                                virDomainNetDefPtr net,
                                int logFD)
1646
{
1647 1648 1649 1650 1651 1652
    /* script is only useful for NET_TYPE_ETHERNET (qemu) and
     * NET_TYPE_BRIDGE (xen), but could be (incorrectly) specified for
     * any interface type. In any case, it's adding user sauce into
     * the soup, so it should taint the domain.
     */
    if (net->script != NULL)
1653
        qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_SHELL_SCRIPTS, logFD);
1654
}
1655 1656 1657


static int
1658
qemuDomainOpenLogHelper(virQEMUDriverConfigPtr cfg,
1659
                        virDomainObjPtr vm,
E
Eric Blake 已提交
1660
                        int oflags,
1661 1662 1663 1664
                        mode_t mode)
{
    char *logfile;
    int fd = -1;
1665
    bool trunc = false;
1666

1667
    if (virAsprintf(&logfile, "%s/%s.log", cfg->logDir, vm->def->name) < 0)
1668 1669
        return -1;

1670 1671 1672 1673 1674 1675 1676 1677 1678
    /* To make SELinux happy we always need to open in append mode.
     * So we fake O_TRUNC by calling ftruncate after open instead
     */
    if (oflags & O_TRUNC) {
        oflags &= ~O_TRUNC;
        oflags |= O_APPEND;
        trunc = true;
    }

E
Eric Blake 已提交
1679
    if ((fd = open(logfile, oflags, mode)) < 0) {
1680 1681 1682 1683 1684 1685 1686
        virReportSystemError(errno, _("failed to create logfile %s"),
                             logfile);
        goto cleanup;
    }
    if (virSetCloseExec(fd) < 0) {
        virReportSystemError(errno, _("failed to set close-on-exec flag on %s"),
                             logfile);
1687 1688 1689 1690 1691 1692 1693
        VIR_FORCE_CLOSE(fd);
        goto cleanup;
    }
    if (trunc &&
        ftruncate(fd, 0) < 0) {
        virReportSystemError(errno, _("failed to truncate %s"),
                             logfile);
1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704
        VIR_FORCE_CLOSE(fd);
        goto cleanup;
    }

cleanup:
    VIR_FREE(logfile);
    return fd;
}


int
1705
qemuDomainCreateLog(virQEMUDriverPtr driver, virDomainObjPtr vm,
E
Eric Blake 已提交
1706
                    bool append)
1707
{
1708
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
E
Eric Blake 已提交
1709
    int oflags;
1710
    int ret;
1711

E
Eric Blake 已提交
1712
    oflags = O_CREAT | O_WRONLY;
1713
    /* Only logrotate files in /var/log, so only append if running privileged */
1714
    if (cfg->privileged || append)
E
Eric Blake 已提交
1715
        oflags |= O_APPEND;
1716
    else
E
Eric Blake 已提交
1717
        oflags |= O_TRUNC;
1718

1719 1720 1721
    ret = qemuDomainOpenLogHelper(cfg, vm, oflags, S_IRUSR | S_IWUSR);
    virObjectUnref(cfg);
    return ret;
1722 1723 1724 1725
}


int
1726
qemuDomainOpenLog(virQEMUDriverPtr driver, virDomainObjPtr vm, off_t pos)
1727
{
1728
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1729 1730 1731 1732
    int fd;
    off_t off;
    int whence;

1733 1734 1735
    fd = qemuDomainOpenLogHelper(cfg, vm, O_RDONLY, 0);
    virObjectUnref(cfg);
    if (fd < 0)
1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761
        return -1;

    if (pos < 0) {
        off = 0;
        whence = SEEK_END;
    } else {
        off = pos;
        whence = SEEK_SET;
    }

    if (lseek(fd, off, whence) < 0) {
        if (whence == SEEK_END)
            virReportSystemError(errno,
                                 _("unable to seek to end of log for %s"),
                                 vm->def->name);
        else
            virReportSystemError(errno,
                                 _("unable to seek to %lld from start for %s"),
                                 (long long)off, vm->def->name);
        VIR_FORCE_CLOSE(fd);
    }

    return fd;
}


1762
int qemuDomainAppendLog(virQEMUDriverPtr driver,
1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777
                        virDomainObjPtr obj,
                        int logFD,
                        const char *fmt, ...)
{
    int fd = logFD;
    va_list argptr;
    char *message = NULL;
    int ret = -1;

    va_start(argptr, fmt);

    if ((fd == -1) &&
        (fd = qemuDomainCreateLog(driver, obj, true)) < 0)
        goto cleanup;

1778
    if (virVasprintf(&message, fmt, argptr) < 0)
1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793
        goto cleanup;
    if (safewrite(fd, message, strlen(message)) < 0) {
        virReportSystemError(errno, _("Unable to write to domain logfile %s"),
                             obj->def->name);
        goto cleanup;
    }

    ret = 0;

cleanup:
    va_end(argptr);

    if (fd != logFD)
        VIR_FORCE_CLOSE(fd);

O
Osier Yang 已提交
1794
    VIR_FREE(message);
1795 1796
    return ret;
}
1797 1798 1799

/* Locate an appropriate 'qemu-img' binary.  */
const char *
1800
qemuFindQemuImgBinary(virQEMUDriverPtr driver)
1801
{
1802 1803 1804
    if (!driver->qemuImgBinary)
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("unable to find kvm-img or qemu-img"));
1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821

    return driver->qemuImgBinary;
}

int
qemuDomainSnapshotWriteMetadata(virDomainObjPtr vm,
                                virDomainSnapshotObjPtr snapshot,
                                char *snapshotDir)
{
    char *newxml = NULL;
    int ret = -1;
    char *snapDir = NULL;
    char *snapFile = NULL;
    char uuidstr[VIR_UUID_STRING_BUFLEN];

    virUUIDFormat(vm->def->uuid, uuidstr);
    newxml = virDomainSnapshotDefFormat(uuidstr, snapshot->def,
1822 1823
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS, 1);
    if (newxml == NULL)
1824 1825
        return -1;

1826
    if (virAsprintf(&snapDir, "%s/%s", snapshotDir, vm->def->name) < 0)
1827 1828 1829 1830 1831 1832 1833
        goto cleanup;
    if (virFileMakePath(snapDir) < 0) {
        virReportSystemError(errno, _("cannot create snapshot directory '%s'"),
                             snapDir);
        goto cleanup;
    }

1834
    if (virAsprintf(&snapFile, "%s/%s.xml", snapDir, snapshot->def->name) < 0)
1835 1836
        goto cleanup;

J
Ján Tomko 已提交
1837
    ret = virXMLSaveFile(snapFile, NULL, "snapshot-edit", newxml);
1838 1839 1840 1841 1842 1843 1844 1845 1846 1847

cleanup:
    VIR_FREE(snapFile);
    VIR_FREE(snapDir);
    VIR_FREE(newxml);
    return ret;
}

/* The domain is expected to be locked and inactive. Return -1 on normal
 * failure, 1 if we skipped a disk due to try_all.  */
1848
static int
1849
qemuDomainSnapshotForEachQcow2Raw(virQEMUDriverPtr driver,
1850 1851 1852 1853 1854
                                  virDomainDefPtr def,
                                  const char *name,
                                  const char *op,
                                  bool try_all,
                                  int ndisks)
1855 1856
{
    const char *qemuimgarg[] = { NULL, "snapshot", NULL, NULL, NULL, NULL };
1857
    size_t i;
1858 1859 1860 1861 1862 1863 1864 1865 1866
    bool skipped = false;

    qemuimgarg[0] = qemuFindQemuImgBinary(driver);
    if (qemuimgarg[0] == NULL) {
        /* qemuFindQemuImgBinary set the error */
        return -1;
    }

    qemuimgarg[2] = op;
1867
    qemuimgarg[3] = name;
1868

1869
    for (i = 0; i < ndisks; i++) {
1870
        /* FIXME: we also need to handle LVM here */
1871
        if (def->disks[i]->device == VIR_DOMAIN_DISK_DEVICE_DISK) {
1872 1873
            if (def->disks[i]->format > 0 &&
                def->disks[i]->format != VIR_STORAGE_FILE_QCOW2) {
1874 1875 1876 1877 1878
                if (try_all) {
                    /* Continue on even in the face of error, since other
                     * disks in this VM may have the same snapshot name.
                     */
                    VIR_WARN("skipping snapshot action on %s",
1879
                             def->disks[i]->dst);
1880 1881
                    skipped = true;
                    continue;
1882 1883 1884 1885 1886
                } else if (STREQ(op, "-c") && i) {
                    /* We must roll back partial creation by deleting
                     * all earlier snapshots.  */
                    qemuDomainSnapshotForEachQcow2Raw(driver, def, name,
                                                      "-d", false, i);
1887
                }
1888 1889 1890 1891
                virReportError(VIR_ERR_OPERATION_INVALID,
                               _("Disk device '%s' does not support"
                                 " snapshotting"),
                               def->disks[i]->dst);
1892 1893 1894
                return -1;
            }

1895
            qemuimgarg[4] = def->disks[i]->src;
1896 1897 1898 1899

            if (virRun(qemuimgarg, NULL) < 0) {
                if (try_all) {
                    VIR_WARN("skipping snapshot action on %s",
1900
                             def->disks[i]->dst);
1901 1902
                    skipped = true;
                    continue;
1903 1904 1905 1906 1907
                } else if (STREQ(op, "-c") && i) {
                    /* We must roll back partial creation by deleting
                     * all earlier snapshots.  */
                    qemuDomainSnapshotForEachQcow2Raw(driver, def, name,
                                                      "-d", false, i);
1908 1909 1910 1911 1912 1913 1914 1915 1916
                }
                return -1;
            }
        }
    }

    return skipped ? 1 : 0;
}

1917 1918 1919
/* The domain is expected to be locked and inactive. Return -1 on normal
 * failure, 1 if we skipped a disk due to try_all.  */
int
1920
qemuDomainSnapshotForEachQcow2(virQEMUDriverPtr driver,
1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936
                               virDomainObjPtr vm,
                               virDomainSnapshotObjPtr snap,
                               const char *op,
                               bool try_all)
{
    /* Prefer action on the disks in use at the time the snapshot was
     * created; but fall back to current definition if dealing with a
     * snapshot created prior to libvirt 0.9.5.  */
    virDomainDefPtr def = snap->def->dom;

    if (!def)
        def = vm->def;
    return qemuDomainSnapshotForEachQcow2Raw(driver, def, snap->def->name,
                                             op, try_all, def->ndisks);
}

1937 1938
/* Discard one snapshot (or its metadata), without reparenting any children.  */
int
1939
qemuDomainSnapshotDiscard(virQEMUDriverPtr driver,
1940 1941 1942 1943 1944 1945 1946 1947 1948
                          virDomainObjPtr vm,
                          virDomainSnapshotObjPtr snap,
                          bool update_current,
                          bool metadata_only)
{
    char *snapFile = NULL;
    int ret = -1;
    qemuDomainObjPrivatePtr priv;
    virDomainSnapshotObjPtr parentsnap = NULL;
1949
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1950 1951 1952 1953 1954 1955 1956 1957 1958

    if (!metadata_only) {
        if (!virDomainObjIsActive(vm)) {
            /* Ignore any skipped disks */
            if (qemuDomainSnapshotForEachQcow2(driver, vm, snap, "-d",
                                               true) < 0)
                goto cleanup;
        } else {
            priv = vm->privateData;
1959
            qemuDomainObjEnterMonitor(driver, vm);
1960 1961
            /* we continue on even in the face of error */
            qemuMonitorDeleteSnapshot(priv->mon, snap->def->name);
1962
            qemuDomainObjExitMonitor(driver, vm);
1963 1964 1965
        }
    }

1966
    if (virAsprintf(&snapFile, "%s/%s/%s.xml", cfg->snapshotDir,
1967
                    vm->def->name, snap->def->name) < 0)
1968 1969 1970 1971
        goto cleanup;

    if (snap == vm->current_snapshot) {
        if (update_current && snap->def->parent) {
1972
            parentsnap = virDomainSnapshotFindByName(vm->snapshots,
1973 1974 1975 1976 1977 1978 1979
                                                     snap->def->parent);
            if (!parentsnap) {
                VIR_WARN("missing parent snapshot matching name '%s'",
                         snap->def->parent);
            } else {
                parentsnap->def->current = true;
                if (qemuDomainSnapshotWriteMetadata(vm, parentsnap,
1980
                                                    cfg->snapshotDir) < 0) {
1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992
                    VIR_WARN("failed to set parent snapshot '%s' as current",
                             snap->def->parent);
                    parentsnap->def->current = false;
                    parentsnap = NULL;
                }
            }
        }
        vm->current_snapshot = parentsnap;
    }

    if (unlink(snapFile) < 0)
        VIR_WARN("Failed to unlink %s", snapFile);
1993
    virDomainSnapshotObjListRemove(vm->snapshots, snap);
1994 1995 1996 1997 1998

    ret = 0;

cleanup:
    VIR_FREE(snapFile);
1999
    virObjectUnref(cfg);
2000 2001 2002 2003 2004 2005 2006 2007 2008
    return ret;
}

/* Hash iterator callback to discard multiple snapshots.  */
void qemuDomainSnapshotDiscardAll(void *payload,
                                  const void *name ATTRIBUTE_UNUSED,
                                  void *data)
{
    virDomainSnapshotObjPtr snap = payload;
2009
    virQEMUSnapRemovePtr curr = data;
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
    int err;

    if (snap->def->current)
        curr->current = true;
    err = qemuDomainSnapshotDiscard(curr->driver, curr->vm, snap, false,
                                    curr->metadata_only);
    if (err && !curr->err)
        curr->err = err;
}

int
2021
qemuDomainSnapshotDiscardAllMetadata(virQEMUDriverPtr driver,
2022 2023
                                     virDomainObjPtr vm)
{
2024
    virQEMUSnapRemove rem;
2025 2026 2027 2028 2029

    rem.driver = driver;
    rem.vm = vm;
    rem.metadata_only = true;
    rem.err = 0;
2030 2031
    virDomainSnapshotForEach(vm->snapshots, qemuDomainSnapshotDiscardAll,
                             &rem);
2032 2033 2034 2035 2036

    return rem.err;
}

/*
2037
 * The caller must hold a lock the vm and there must
2038 2039 2040
 * be no remaining references to vm.
 */
void
2041
qemuDomainRemoveInactive(virQEMUDriverPtr driver,
2042 2043
                         virDomainObjPtr vm)
{
2044
    char *snapDir;
2045
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2046

2047 2048 2049 2050 2051
    /* Remove any snapshot metadata prior to removing the domain */
    if (qemuDomainSnapshotDiscardAllMetadata(driver, vm) < 0) {
        VIR_WARN("unable to remove all snapshots for domain %s",
                 vm->def->name);
    }
2052
    else if (virAsprintf(&snapDir, "%s/%s", cfg->snapshotDir,
2053 2054
                         vm->def->name) < 0) {
        VIR_WARN("unable to remove snapshot directory %s/%s",
2055
                 cfg->snapshotDir, vm->def->name);
2056 2057 2058 2059 2060
    } else {
        if (rmdir(snapDir) < 0 && errno != ENOENT)
            VIR_WARN("unable to remove snapshot directory %s", snapDir);
        VIR_FREE(snapDir);
    }
2061
    virDomainObjListRemove(driver->domains, vm);
2062
    virObjectUnref(cfg);
2063
}
2064 2065

void
2066
qemuDomainSetFakeReboot(virQEMUDriverPtr driver,
2067 2068 2069 2070
                        virDomainObjPtr vm,
                        bool value)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2071
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2072 2073

    if (priv->fakeReboot == value)
2074
        goto cleanup;
2075 2076 2077

    priv->fakeReboot = value;

2078
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
2079
        VIR_WARN("Failed to save status on vm %s", vm->def->name);
2080

2081
cleanup:
2082
    virObjectUnref(cfg);
2083
}
M
Michal Privoznik 已提交
2084

2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125
static int
qemuDomainCheckRemoveOptionalDisk(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  virDomainDiskDefPtr disk)
{
    char uuid[VIR_UUID_STRING_BUFLEN];
    virDomainEventPtr event = NULL;
    virDomainDiskDefPtr del_disk = NULL;

    virUUIDFormat(vm->def->uuid, uuid);

    VIR_DEBUG("Dropping disk '%s' on domain '%s' (UUID '%s') "
              "due to inaccessible source '%s'",
              disk->dst, vm->def->name, uuid, disk->src);

    if (disk->device == VIR_DOMAIN_DISK_DEVICE_CDROM ||
        disk->device == VIR_DOMAIN_DISK_DEVICE_FLOPPY) {

        event = virDomainEventDiskChangeNewFromObj(vm, disk->src, NULL,
                                                   disk->info.alias,
                                                   VIR_DOMAIN_EVENT_DISK_CHANGE_MISSING_ON_START);
        VIR_FREE(disk->src);
    } else {
        event = virDomainEventDiskChangeNewFromObj(vm, disk->src, NULL,
                                                   disk->info.alias,
                                                   VIR_DOMAIN_EVENT_DISK_DROP_MISSING_ON_START);

        if (!(del_disk = virDomainDiskRemoveByName(vm->def, disk->src))) {
            virReportError(VIR_ERR_INVALID_ARG,
                           _("no source device %s"), disk->src);
            return -1;
        }
        virDomainDiskDefFree(del_disk);
    }

    if (event)
        qemuDomainEventQueue(driver, event);

    return 0;
}

2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144
static int
qemuDomainCheckDiskStartupPolicy(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 virDomainDiskDefPtr disk,
                                 bool cold_boot)
{
    char uuid[VIR_UUID_STRING_BUFLEN];
    int startupPolicy = disk->startupPolicy;

    virUUIDFormat(vm->def->uuid, uuid);

    switch ((enum virDomainStartupPolicy) startupPolicy) {
        case VIR_DOMAIN_STARTUP_POLICY_OPTIONAL:
            break;

        case VIR_DOMAIN_STARTUP_POLICY_MANDATORY:
            goto error;

        case VIR_DOMAIN_STARTUP_POLICY_REQUISITE:
2145
            if (cold_boot)
2146 2147 2148 2149 2150 2151 2152 2153 2154
                goto error;
            break;

        case VIR_DOMAIN_STARTUP_POLICY_DEFAULT:
        case VIR_DOMAIN_STARTUP_POLICY_LAST:
            /* this should never happen */
            break;
    }

2155 2156
    if (qemuDomainCheckRemoveOptionalDisk(driver, vm, disk) < 0)
        goto error;
2157 2158 2159 2160 2161 2162 2163

    return 0;

error:
    return -1;
}

M
Michal Privoznik 已提交
2164
int
2165
qemuDomainCheckDiskPresence(virQEMUDriverPtr driver,
M
Michal Privoznik 已提交
2166
                            virDomainObjPtr vm,
2167
                            bool cold_boot)
M
Michal Privoznik 已提交
2168 2169
{
    int ret = -1;
2170
    size_t i;
M
Michal Privoznik 已提交
2171 2172
    virDomainDiskDefPtr disk;

2173
    VIR_DEBUG("Checking for disk presence");
2174 2175
    for (i = vm->def->ndisks; i > 0; i--) {
        disk = vm->def->disks[i - 1];
M
Michal Privoznik 已提交
2176

2177
        if (!disk->src)
M
Michal Privoznik 已提交
2178 2179
            continue;

2180 2181
        if (qemuDomainDetermineDiskChain(driver, disk, false) >= 0 &&
            qemuDiskChainCheckBroken(disk) >= 0)
M
Michal Privoznik 已提交
2182
            continue;
2183

2184 2185 2186 2187 2188
        if (disk->startupPolicy &&
            qemuDomainCheckDiskStartupPolicy(driver, vm, disk,
                                             cold_boot) >= 0) {
            virResetLastError();
            continue;
M
Michal Privoznik 已提交
2189 2190
        }

2191
        goto error;
M
Michal Privoznik 已提交
2192 2193 2194 2195
    }

    ret = 0;

2196
error:
M
Michal Privoznik 已提交
2197 2198
    return ret;
}
2199 2200 2201 2202 2203 2204 2205 2206 2207 2208

/*
 * The vm must be locked when any of the following cleanup functions is
 * called.
 */
int
qemuDomainCleanupAdd(virDomainObjPtr vm,
                     qemuDomainCleanupCallback cb)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2209
    size_t i;
2210 2211 2212 2213 2214 2215 2216 2217 2218 2219

    VIR_DEBUG("vm=%s, cb=%p", vm->def->name, cb);

    for (i = 0; i < priv->ncleanupCallbacks; i++) {
        if (priv->cleanupCallbacks[i] == cb)
            return 0;
    }

    if (VIR_RESIZE_N(priv->cleanupCallbacks,
                     priv->ncleanupCallbacks_max,
2220
                     priv->ncleanupCallbacks, 1) < 0)
2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231
        return -1;

    priv->cleanupCallbacks[priv->ncleanupCallbacks++] = cb;
    return 0;
}

void
qemuDomainCleanupRemove(virDomainObjPtr vm,
                        qemuDomainCleanupCallback cb)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2232
    size_t i;
2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250

    VIR_DEBUG("vm=%s, cb=%p", vm->def->name, cb);

    for (i = 0; i < priv->ncleanupCallbacks; i++) {
        if (priv->cleanupCallbacks[i] == cb) {
            memmove(priv->cleanupCallbacks + i,
                    priv->cleanupCallbacks + i + 1,
                    priv->ncleanupCallbacks - i - 1);
            priv->ncleanupCallbacks--;
        }
    }

    VIR_SHRINK_N(priv->cleanupCallbacks,
                 priv->ncleanupCallbacks_max,
                 priv->ncleanupCallbacks_max - priv->ncleanupCallbacks);
}

void
2251
qemuDomainCleanupRun(virQEMUDriverPtr driver,
2252 2253 2254
                     virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2255
    size_t i;
2256 2257 2258 2259

    VIR_DEBUG("driver=%p, vm=%s", driver, vm->def->name);

    /* run cleanup callbacks in reverse order */
2260 2261
    for (i = 0; i < priv->ncleanupCallbacks; i++) {
        if (priv->cleanupCallbacks[priv->ncleanupCallbacks - (i + 1)])
2262 2263 2264 2265 2266 2267 2268
            priv->cleanupCallbacks[i](driver, vm);
    }

    VIR_FREE(priv->cleanupCallbacks);
    priv->ncleanupCallbacks = 0;
    priv->ncleanupCallbacks_max = 0;
}
2269

2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291
int
qemuDiskChainCheckBroken(virDomainDiskDefPtr disk)
{
    char *brokenFile = NULL;

    if (!disk->src || !disk->backingChain)
        return 0;

    if (virStorageFileChainGetBroken(disk->backingChain, &brokenFile) < 0)
        return -1;

    if (brokenFile) {
        virReportError(VIR_ERR_INVALID_ARG,
                       _("Backing file '%s' of image '%s' is missing."),
                       brokenFile, disk->src);
        VIR_FREE(brokenFile);
        return -1;
    }

    return 0;
}

2292
int
2293
qemuDomainDetermineDiskChain(virQEMUDriverPtr driver,
2294 2295 2296
                             virDomainDiskDefPtr disk,
                             bool force)
{
2297 2298
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = 0;
2299

2300 2301 2302
    if (!disk->src ||
        disk->type == VIR_DOMAIN_DISK_TYPE_NETWORK ||
        disk->type == VIR_DOMAIN_DISK_TYPE_VOLUME)
2303
        goto cleanup;
2304 2305 2306 2307 2308 2309

    if (disk->backingChain) {
        if (force) {
            virStorageFileFreeMetadata(disk->backingChain);
            disk->backingChain = NULL;
        } else {
2310
            goto cleanup;
2311 2312 2313
        }
    }
    disk->backingChain = virStorageFileGetMetadata(disk->src, disk->format,
2314 2315
                                                   cfg->user, cfg->group,
                                                   cfg->allowDiskFormatProbing);
2316
    if (!disk->backingChain)
2317 2318 2319 2320 2321
        ret = -1;

cleanup:
    virObjectUnref(cfg);
    return ret;
2322
}
2323

2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344
int
qemuDomainUpdateDeviceList(virQEMUDriverPtr driver,
                           virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char **aliases;

    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE_DEL_EVENT))
        return 0;

    qemuDomainObjEnterMonitor(driver, vm);
    if (qemuMonitorGetDeviceAliases(priv->mon, &aliases) < 0) {
        qemuDomainObjExitMonitor(driver, vm);
        return -1;
    }
    qemuDomainObjExitMonitor(driver, vm);

    virStringFreeList(priv->qemuDevices);
    priv->qemuDevices = aliases;
    return 0;
}
2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366

bool
qemuDomainDefCheckABIStability(virQEMUDriverPtr driver,
                               virDomainDefPtr src,
                               virDomainDefPtr dst)
{
    virDomainDefPtr migratableDefSrc = NULL;
    virDomainDefPtr migratableDefDst = NULL;
    const int flags = VIR_DOMAIN_XML_SECURE | VIR_DOMAIN_XML_UPDATE_CPU | VIR_DOMAIN_XML_MIGRATABLE;
    bool ret = false;

    if (!(migratableDefSrc = qemuDomainDefCopy(driver, src, flags)) ||
        !(migratableDefDst = qemuDomainDefCopy(driver, dst, flags)))
        goto cleanup;

    ret = virDomainDefCheckABIStability(migratableDefSrc, migratableDefDst);

cleanup:
    virDomainDefFree(migratableDefSrc);
    virDomainDefFree(migratableDefDst);
    return ret;
}