qemu_domain.c 67.1 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
static void
qemuDomainDefNamespaceFree(void *nsdata)
{
    qemuDomainCmdlineDefPtr cmd = nsdata;
538
    size_t i;
539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555

    if (!cmd)
        return;

    for (i = 0; i < cmd->num_args; i++)
        VIR_FREE(cmd->args[i]);
    for (i = 0; i < cmd->num_env; i++) {
        VIR_FREE(cmd->env_name[i]);
        VIR_FREE(cmd->env_value[i]);
    }
    VIR_FREE(cmd->args);
    VIR_FREE(cmd->env_name);
    VIR_FREE(cmd->env_value);
    VIR_FREE(cmd);
}

static int
P
Philipp Hahn 已提交
556 557
qemuDomainDefNamespaceParse(xmlDocPtr xml ATTRIBUTE_UNUSED,
                            xmlNodePtr root ATTRIBUTE_UNUSED,
558 559 560 561
                            xmlXPathContextPtr ctxt,
                            void **data)
{
    qemuDomainCmdlineDefPtr cmd = NULL;
P
Philipp Hahn 已提交
562
    bool uses_qemu_ns = false;
563
    xmlNodePtr *nodes = NULL;
564 565
    int n;
    size_t i;
566

P
Philipp Hahn 已提交
567
    if (xmlXPathRegisterNs(ctxt, BAD_CAST "qemu", BAD_CAST QEMU_NAMESPACE_HREF) < 0) {
568 569 570
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Failed to register xml namespace '%s'"),
                       QEMU_NAMESPACE_HREF);
571 572 573
        return -1;
    }

574
    if (VIR_ALLOC(cmd) < 0)
575 576 577 578 579 580
        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
581
    uses_qemu_ns |= n > 0;
582 583

    if (n && VIR_ALLOC_N(cmd->args, n) < 0)
584
        goto error;
585 586 587 588

    for (i = 0; i < n; i++) {
        cmd->args[cmd->num_args] = virXMLPropString(nodes[i], "value");
        if (cmd->args[cmd->num_args] == NULL) {
589 590
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("No qemu command-line argument specified"));
591 592 593 594 595 596 597 598 599 600 601
            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
602
    uses_qemu_ns |= n > 0;
603 604

    if (n && VIR_ALLOC_N(cmd->env_name, n) < 0)
605
        goto error;
606 607

    if (n && VIR_ALLOC_N(cmd->env_value, n) < 0)
608
        goto error;
609 610 611 612 613 614

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

        tmp = virXMLPropString(nodes[i], "name");
        if (tmp == NULL) {
615 616
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("No qemu environment name specified"));
617 618 619
            goto error;
        }
        if (tmp[0] == '\0') {
620 621
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Empty qemu environment name specified"));
622 623 624
            goto error;
        }
        if (!c_isalpha(tmp[0]) && tmp[0] != '_') {
625 626
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Invalid environment name, it must begin with a letter or underscore"));
627 628 629
            goto error;
        }
        if (strspn(tmp, "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789_") != strlen(tmp)) {
630 631
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("Invalid environment name, it must contain only alphanumerics and underscore"));
632 633 634 635 636 637 638 639 640 641 642 643
            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
644 645 646 647
    if (uses_qemu_ns)
        *data = cmd;
    else
        VIR_FREE(cmd);
648 649 650 651 652 653 654 655 656 657 658 659 660 661

    return 0;

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

static int
qemuDomainDefNamespaceFormatXML(virBufferPtr buf,
                                void *nsdata)
{
    qemuDomainCmdlineDefPtr cmd = nsdata;
662
    size_t i;
663 664 665 666 667 668 669 670 671

    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++) {
672
        virBufferAsprintf(buf, "    <qemu:env name='%s'", cmd->env_name[i]);
673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688
        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 "'";
}


689 690 691 692 693 694
virDomainXMLNamespace virQEMUDriverDomainXMLNamespace = {
    .parse = qemuDomainDefNamespaceParse,
    .free = qemuDomainDefNamespaceFree,
    .format = qemuDomainDefNamespaceFormatXML,
    .href = qemuDomainDefNamespaceHref,
};
695

696

697 698 699 700 701
static int
qemuDomainDefPostParse(virDomainDefPtr def,
                       virCapsPtr caps,
                       void *opaque ATTRIBUTE_UNUSED)
{
702 703
    bool addPCIRoot = false;

704 705 706 707 708
    /* check for emulator and create a default one if needed */
    if (!def->emulator &&
        !(def->emulator = virDomainDefGetDefaultEmulator(def, caps)))
        return -1;

709 710 711 712 713 714 715 716 717 718 719 720
    /* 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;
        if (STRPREFIX(def->os.machine, "pc-q35") ||
            STREQ(def->os.machine, "q35") ||
            STREQ(def->os.machine, "isapc"))
            break;
        if (!STRPREFIX(def->os.machine, "pc-0.") &&
            !STRPREFIX(def->os.machine, "pc-1.") &&
721
            !STRPREFIX(def->os.machine, "pc-i440") &&
722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745
            !STREQ(def->os.machine, "pc") &&
            !STRPREFIX(def->os.machine, "rhel"))
            break;
        addPCIRoot = true;
        break;

    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;
    }

    if (addPCIRoot &&
        virDomainDefMaybeAddController(
            def, VIR_DOMAIN_CONTROLLER_TYPE_PCI, 0,
            VIR_DOMAIN_CONTROLLER_MODEL_PCI_ROOT) < 0)
        return -1;

746 747 748 749
    return 0;
}


750 751
static int
qemuDomainDeviceDefPostParse(virDomainDeviceDefPtr dev,
752
                             virDomainDefPtr def,
753
                             virCapsPtr caps ATTRIBUTE_UNUSED,
754
                             void *opaque)
755
{
756 757 758 759
    int ret = -1;
    virQEMUDriverPtr driver = opaque;
    virQEMUDriverConfigPtr cfg = NULL;

760
    if (dev->type == VIR_DOMAIN_DEVICE_NET &&
761 762
        dev->data.net->type != VIR_DOMAIN_NET_TYPE_HOSTDEV &&
        !dev->data.net->model) {
763 764 765 766
        if (VIR_STRDUP(dev->data.net->model,
                       def->os.arch == VIR_ARCH_S390 ||
                       def->os.arch == VIR_ARCH_S390X ? "virtio" : "rtl8139") < 0)
            goto cleanup;
767
    }
768

769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789
    /* 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 &&
790 791
                    VIR_STRDUP(disk->driverName, "qemu") < 0)
                        goto cleanup;
792 793 794 795 796 797 798 799 800 801 802 803

                /* 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;
            }
804 805 806
        }
    }

807 808 809 810 811 812 813
    /* 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;

814 815 816 817 818 819 820 821
    /* 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;

822 823 824 825 826 827 828 829 830 831 832
    /* 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,
833
                        dev->data.chr->target.name) < 0)
834
            goto cleanup;
835 836 837
        dev->data.chr->source.data.nix.listen = true;
    }

838 839 840 841 842
    ret = 0;

cleanup:
    virObjectUnref(cfg);
    return ret;
843 844 845 846 847
}


virDomainDefParserConfig virQEMUDriverDomainDefParserConfig = {
    .devicesPostParseCallback = qemuDomainDeviceDefPostParse,
848
    .domainPostParseCallback = qemuDomainDefPostParse,
849 850 851
};


852
static void
853
qemuDomainObjSaveJob(virQEMUDriverPtr driver, virDomainObjPtr obj)
854
{
855 856 857
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);

    if (virDomainObjIsActive(obj)) {
858
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, obj) < 0)
859
            VIR_WARN("Failed to save status on vm %s", obj->def->name);
860
    }
861

862
    virObjectUnref(cfg);
863 864
}

J
Jiri Denemark 已提交
865
void
866
qemuDomainObjSetJobPhase(virQEMUDriverPtr driver,
J
Jiri Denemark 已提交
867 868 869 870
                         virDomainObjPtr obj,
                         int phase)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
871
    unsigned long long me = virThreadSelfID();
J
Jiri Denemark 已提交
872 873 874 875

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

876 877 878 879 880
    VIR_DEBUG("Setting '%s' phase to '%s'",
              qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
              qemuDomainAsyncJobPhaseToString(priv->job.asyncJob, phase));

    if (priv->job.asyncOwner && me != priv->job.asyncOwner) {
881
        VIR_WARN("'%s' async job is owned by thread %llu",
882 883 884 885
                 qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                 priv->job.asyncOwner);
    }

J
Jiri Denemark 已提交
886
    priv->job.phase = phase;
887
    priv->job.asyncOwner = me;
J
Jiri Denemark 已提交
888 889 890
    qemuDomainObjSaveJob(driver, obj);
}

891
void
892 893
qemuDomainObjSetAsyncJobMask(virDomainObjPtr obj,
                             unsigned long long allowedJobs)
894 895 896
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

897 898 899 900 901 902 903
    if (!priv->job.asyncJob)
        return;

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

void
904
qemuDomainObjDiscardAsyncJob(virQEMUDriverPtr driver, virDomainObjPtr obj)
905 906 907 908 909 910
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    if (priv->job.active == QEMU_JOB_ASYNC_NESTED)
        qemuDomainObjResetJob(priv);
    qemuDomainObjResetAsyncJob(priv);
911
    qemuDomainObjSaveJob(driver, obj);
912 913
}

914 915 916 917 918 919 920 921 922
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()) {
923
        VIR_WARN("'%s' async job is owned by thread %llu",
924 925 926 927 928 929
                 qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
                 priv->job.asyncOwner);
    }
    priv->job.asyncOwner = 0;
}

930
static bool
931
qemuDomainNestedJobAllowed(qemuDomainObjPrivatePtr priv, enum qemuDomainJob job)
932 933
{
    return !priv->job.asyncJob || (priv->job.mask & JOB_MASK(job)) != 0;
934 935
}

936 937 938 939 940 941
bool
qemuDomainJobAllowed(qemuDomainObjPrivatePtr priv, enum qemuDomainJob job)
{
    return !priv->job.active && qemuDomainNestedJobAllowed(priv, job);
}

942 943 944
/* Give up waiting for mutex after 30 seconds */
#define QEMU_JOB_WAIT_TIME (1000ull * 30)

945
/*
946
 * obj must be locked before calling
947
 */
948
static int ATTRIBUTE_NONNULL(1)
949
qemuDomainObjBeginJobInternal(virQEMUDriverPtr driver,
950 951 952
                              virDomainObjPtr obj,
                              enum qemuDomainJob job,
                              enum qemuDomainAsyncJob asyncJob)
953 954
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
J
Jiri Denemark 已提交
955
    unsigned long long now;
956
    unsigned long long then;
957
    bool nested = job == QEMU_JOB_ASYNC_NESTED;
958
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
959

960 961
    priv->jobs_queued++;

962 963
    if (virTimeMillisNow(&now) < 0) {
        virObjectUnref(cfg);
964
        return -1;
965 966
    }

J
Jiri Denemark 已提交
967
    then = now + QEMU_JOB_WAIT_TIME;
968

969
    virObjectRef(obj);
970

971
retry:
972 973
    if (cfg->maxQueuedJobs &&
        priv->jobs_queued > cfg->maxQueuedJobs) {
974 975 976
        goto error;
    }

977
    while (!nested && !qemuDomainNestedJobAllowed(priv, job)) {
978
        if (virCondWaitUntil(&priv->job.asyncCond, &obj->parent.lock, then) < 0)
979 980 981
            goto error;
    }

982
    while (priv->job.active) {
983
        if (virCondWaitUntil(&priv->job.cond, &obj->parent.lock, then) < 0)
984
            goto error;
985
    }
986 987 988

    /* No job is active but a new async job could have been started while obj
     * was unlocked, so we need to recheck it. */
989
    if (!nested && !qemuDomainNestedJobAllowed(priv, job))
990 991
        goto retry;

992
    qemuDomainObjResetJob(priv);
993 994

    if (job != QEMU_JOB_ASYNC) {
995 996 997
        VIR_DEBUG("Starting job: %s (async=%s)",
                   qemuDomainJobTypeToString(job),
                   qemuDomainAsyncJobTypeToString(priv->job.asyncJob));
998
        priv->job.active = job;
999
        priv->job.owner = virThreadSelfID();
1000
    } else {
1001 1002
        VIR_DEBUG("Starting async job: %s",
                  qemuDomainAsyncJobTypeToString(asyncJob));
1003 1004
        qemuDomainObjResetAsyncJob(priv);
        priv->job.asyncJob = asyncJob;
1005
        priv->job.asyncOwner = virThreadSelfID();
1006 1007
        priv->job.start = now;
    }
1008

1009 1010
    if (qemuDomainTrackJob(job))
        qemuDomainObjSaveJob(driver, obj);
1011

1012
    virObjectUnref(cfg);
1013
    return 0;
1014 1015

error:
1016
    VIR_WARN("Cannot start job (%s, %s) for domain %s;"
1017
             " current job is (%s, %s) owned by (%llu, %llu)",
1018 1019 1020 1021 1022 1023 1024
             qemuDomainJobTypeToString(job),
             qemuDomainAsyncJobTypeToString(asyncJob),
             obj->def->name,
             qemuDomainJobTypeToString(priv->job.active),
             qemuDomainAsyncJobTypeToString(priv->job.asyncJob),
             priv->job.owner, priv->job.asyncOwner);

1025
    if (errno == ETIMEDOUT)
1026 1027
        virReportError(VIR_ERR_OPERATION_TIMEOUT,
                       "%s", _("cannot acquire state change lock"));
1028 1029
    else if (cfg->maxQueuedJobs &&
             priv->jobs_queued > cfg->maxQueuedJobs)
1030 1031 1032
        virReportError(VIR_ERR_OPERATION_FAILED,
                       "%s", _("cannot acquire state change lock "
                               "due to max_queued limit"));
1033 1034 1035
    else
        virReportSystemError(errno,
                             "%s", _("cannot acquire job mutex"));
1036
    priv->jobs_queued--;
1037
    virObjectUnref(obj);
1038
    virObjectUnref(cfg);
1039
    return -1;
1040 1041 1042
}

/*
1043
 * obj must be locked before calling
1044 1045 1046 1047 1048 1049 1050
 *
 * 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
 */
1051
int qemuDomainObjBeginJob(virQEMUDriverPtr driver,
1052 1053
                          virDomainObjPtr obj,
                          enum qemuDomainJob job)
1054
{
1055
    return qemuDomainObjBeginJobInternal(driver, obj, job,
1056 1057 1058
                                         QEMU_ASYNC_JOB_NONE);
}

1059
int qemuDomainObjBeginAsyncJob(virQEMUDriverPtr driver,
1060
                               virDomainObjPtr obj,
1061
                               enum qemuDomainAsyncJob asyncJob)
1062
{
1063
    return qemuDomainObjBeginJobInternal(driver, obj, QEMU_JOB_ASYNC,
1064
                                         asyncJob);
1065 1066
}

1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
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()) {
1081
        VIR_WARN("This thread doesn't seem to be the async job owner: %llu",
1082 1083 1084 1085 1086 1087 1088 1089
                 priv->job.asyncOwner);
    }

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

1090

1091
/*
1092
 * obj must be locked before calling
1093 1094 1095 1096
 *
 * To be called after completing the work associated with the
 * earlier qemuDomainBeginJob() call
 *
1097 1098
 * Returns true if @obj was still referenced, false if it was
 * disposed of.
1099
 */
1100
bool qemuDomainObjEndJob(virQEMUDriverPtr driver, virDomainObjPtr obj)
1101 1102
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
1103
    enum qemuDomainJob job = priv->job.active;
1104

1105 1106
    priv->jobs_queued--;

1107
    VIR_DEBUG("Stopping job: %s (async=%s)",
1108
              qemuDomainJobTypeToString(job),
1109 1110
              qemuDomainAsyncJobTypeToString(priv->job.asyncJob));

1111
    qemuDomainObjResetJob(priv);
1112 1113
    if (qemuDomainTrackJob(job))
        qemuDomainObjSaveJob(driver, obj);
1114
    virCondSignal(&priv->job.cond);
1115

1116
    return virObjectUnref(obj);
1117 1118
}

1119
bool
1120
qemuDomainObjEndAsyncJob(virQEMUDriverPtr driver, virDomainObjPtr obj)
1121 1122
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
1123

1124 1125
    priv->jobs_queued--;

1126 1127 1128
    VIR_DEBUG("Stopping async job: %s",
              qemuDomainAsyncJobTypeToString(priv->job.asyncJob));

1129
    qemuDomainObjResetAsyncJob(priv);
1130
    qemuDomainObjSaveJob(driver, obj);
1131 1132
    virCondBroadcast(&priv->job.asyncCond);

1133
    return virObjectUnref(obj);
1134 1135
}

1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146
void
qemuDomainObjAbortAsyncJob(virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

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

    priv->job.asyncAbort = true;
}

1147 1148 1149 1150 1151 1152 1153 1154 1155
/*
 * 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
 */
1156
static int
1157
qemuDomainObjEnterMonitorInternal(virQEMUDriverPtr driver,
1158 1159
                                  virDomainObjPtr obj,
                                  enum qemuDomainAsyncJob asyncJob)
1160 1161 1162
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

1163
    if (asyncJob != QEMU_ASYNC_JOB_NONE) {
1164
        if (qemuDomainObjBeginNestedJob(driver, obj, asyncJob) < 0)
1165 1166
            return -1;
        if (!virDomainObjIsActive(obj)) {
1167 1168
            virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                           _("domain is no longer running"));
1169 1170
            /* Still referenced by the containing async job.  */
            ignore_value(qemuDomainObjEndJob(driver, obj));
1171 1172
            return -1;
        }
1173 1174 1175
    } 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");
1176 1177
    }

1178
    virObjectLock(priv->mon);
1179
    virObjectRef(priv->mon);
1180
    ignore_value(virTimeMillisNow(&priv->monStart));
1181
    virObjectUnlock(obj);
1182 1183

    return 0;
1184 1185
}

1186
static void ATTRIBUTE_NONNULL(1)
1187
qemuDomainObjExitMonitorInternal(virQEMUDriverPtr driver,
1188
                                 virDomainObjPtr obj)
1189 1190
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
1191
    bool hasRefs;
1192

1193
    hasRefs = virObjectUnref(priv->mon);
1194

1195
    if (hasRefs)
1196
        virObjectUnlock(priv->mon);
1197

1198
    virObjectLock(obj);
1199

1200
    priv->monStart = 0;
1201
    if (!hasRefs)
1202
        priv->mon = NULL;
1203

1204 1205 1206 1207 1208
    if (priv->job.active == QEMU_JOB_ASYNC_NESTED) {
        qemuDomainObjResetJob(priv);
        qemuDomainObjSaveJob(driver, obj);
        virCondSignal(&priv->job.cond);

1209
        virObjectUnref(obj);
1210
    }
1211 1212
}

1213
void qemuDomainObjEnterMonitor(virQEMUDriverPtr driver,
1214
                               virDomainObjPtr obj)
1215
{
1216
    ignore_value(qemuDomainObjEnterMonitorInternal(driver, obj,
1217
                                                   QEMU_ASYNC_JOB_NONE));
1218 1219
}

1220
/* obj must NOT be locked before calling
1221 1222 1223
 *
 * Should be paired with an earlier qemuDomainObjEnterMonitor() call
 */
1224
void qemuDomainObjExitMonitor(virQEMUDriverPtr driver,
1225
                              virDomainObjPtr obj)
1226
{
1227
    qemuDomainObjExitMonitorInternal(driver, obj);
1228
}
1229 1230

/*
1231
 * obj must be locked before calling
1232 1233
 *
 * To be called immediately before any QEMU monitor API call.
1234
 * Must have already either called qemuDomainObjBeginJob()
1235 1236 1237 1238 1239
 * 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
1240
 * qemuDomainObjExitMonitor(); or -1 if the job could not be
1241 1242 1243
 * started (probably because the vm exited in the meantime).
 */
int
1244
qemuDomainObjEnterMonitorAsync(virQEMUDriverPtr driver,
1245 1246
                               virDomainObjPtr obj,
                               enum qemuDomainAsyncJob asyncJob)
1247
{
1248
    return qemuDomainObjEnterMonitorInternal(driver, obj, asyncJob);
1249 1250
}

D
Daniel P. Berrange 已提交
1251 1252


1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263
/*
 * 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
1264 1265 1266
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

1267
    virObjectLock(priv->agent);
1268
    virObjectRef(priv->agent);
D
Daniel P. Berrange 已提交
1269
    ignore_value(virTimeMillisNow(&priv->agentStart));
1270
    virObjectUnlock(obj);
D
Daniel P. Berrange 已提交
1271 1272
}

1273 1274 1275 1276 1277 1278 1279

/* 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
1280 1281
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
1282
    bool hasRefs;
D
Daniel P. Berrange 已提交
1283

1284
    hasRefs = virObjectUnref(priv->agent);
D
Daniel P. Berrange 已提交
1285

1286
    if (hasRefs)
1287
        virObjectUnlock(priv->agent);
D
Daniel P. Berrange 已提交
1288

1289
    virObjectLock(obj);
D
Daniel P. Berrange 已提交
1290 1291

    priv->agentStart = 0;
1292
    if (!hasRefs)
D
Daniel P. Berrange 已提交
1293 1294 1295
        priv->agent = NULL;
}

1296
void qemuDomainObjEnterRemote(virDomainObjPtr obj)
1297
{
1298
    virObjectRef(obj);
1299
    virObjectUnlock(obj);
1300 1301
}

1302
void qemuDomainObjExitRemote(virDomainObjPtr obj)
1303
{
1304
    virObjectLock(obj);
1305
    virObjectUnref(obj);
1306
}
1307 1308


1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337
virDomainDefPtr
qemuDomainDefCopy(virQEMUDriverPtr driver,
                  virDomainDefPtr src,
                  unsigned int flags)
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;
    virDomainDefPtr ret = NULL;
    virCapsPtr caps = NULL;
    const char *xml = NULL;

    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;
}

1338
int
1339
qemuDomainDefFormatBuf(virQEMUDriverPtr driver,
1340 1341 1342
                       virDomainDefPtr def,
                       unsigned int flags,
                       virBuffer *buf)
1343
{
1344
    int ret = -1;
1345
    virCPUDefPtr cpu = NULL;
1346
    virCPUDefPtr def_cpu = def->cpu;
1347 1348
    virDomainControllerDefPtr *controllers = NULL;
    int ncontrollers = 0;
1349 1350 1351 1352
    virCapsPtr caps = NULL;

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;
1353 1354

    /* Update guest CPU requirements according to host CPU */
1355 1356 1357
    if ((flags & VIR_DOMAIN_XML_UPDATE_CPU) &&
        def_cpu &&
        (def_cpu->mode != VIR_CPU_MODE_CUSTOM || def_cpu->model)) {
1358 1359
        if (!caps->host.cpu ||
            !caps->host.cpu->model) {
1360 1361
            virReportError(VIR_ERR_OPERATION_FAILED,
                           "%s", _("cannot get host CPU capabilities"));
1362 1363 1364
            goto cleanup;
        }

1365
        if (!(cpu = virCPUDefCopy(def_cpu)) ||
1366
            cpuUpdate(cpu, caps->host.cpu) < 0)
1367 1368 1369 1370
            goto cleanup;
        def->cpu = cpu;
    }

1371
    if ((flags & VIR_DOMAIN_XML_MIGRATABLE)) {
1372
        size_t i;
1373
        int toremove = 0;
1374
        virDomainControllerDefPtr usb = NULL, pci = NULL;
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392

        /* 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);
1393
            toremove++;
1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413
        } 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) {
            VIR_DEBUG("Removing default 'pci-root' from domain '%s'"
                      " for migration compatibility", def->name);
1414
            toremove++;
1415 1416 1417 1418
        } else {
            pci = NULL;
        }

1419
        if (toremove) {
1420 1421
            controllers = def->controllers;
            ncontrollers = def->ncontrollers;
1422
            if (VIR_ALLOC_N(def->controllers, ncontrollers - toremove) < 0) {
1423 1424 1425 1426 1427 1428
                controllers = NULL;
                goto cleanup;
            }

            def->ncontrollers = 0;
            for (i = 0; i < ncontrollers; i++) {
1429
                if (controllers[i] != usb && controllers[i] != pci)
1430 1431 1432
                    def->controllers[def->ncontrollers++] = controllers[i];
            }
        }
1433 1434


1435 1436
    }

1437
    ret = virDomainDefFormatInternal(def, flags, buf);
1438 1439 1440 1441

cleanup:
    def->cpu = def_cpu;
    virCPUDefFree(cpu);
1442 1443 1444 1445 1446
    if (controllers) {
        VIR_FREE(def->controllers);
        def->controllers = controllers;
        def->ncontrollers = ncontrollers;
    }
1447
    virObjectUnref(caps);
1448 1449
    return ret;
}
1450

1451
char *qemuDomainDefFormatXML(virQEMUDriverPtr driver,
1452
                             virDomainDefPtr def,
1453
                             unsigned int flags)
1454 1455 1456
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;

1457
    if (qemuDomainDefFormatBuf(driver, def, flags, &buf) < 0) {
1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470
        virBufferFreeAndReset(&buf);
        return NULL;
    }

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

    return virBufferContentAndReset(&buf);
}

1471
char *qemuDomainFormatXML(virQEMUDriverPtr driver,
1472
                          virDomainObjPtr vm,
1473
                          unsigned int flags)
1474 1475 1476 1477 1478 1479 1480 1481
{
    virDomainDefPtr def;

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

1482
    return qemuDomainDefFormatXML(driver, def, flags);
1483 1484
}

1485
char *
1486
qemuDomainDefFormatLive(virQEMUDriverPtr driver,
1487
                        virDomainDefPtr def,
1488 1489
                        bool inactive,
                        bool compatible)
1490 1491 1492 1493 1494
{
    unsigned int flags = QEMU_DOMAIN_FORMAT_LIVE_FLAGS;

    if (inactive)
        flags |= VIR_DOMAIN_XML_INACTIVE;
1495 1496
    if (compatible)
        flags |= VIR_DOMAIN_XML_MIGRATABLE;
1497

1498
    return qemuDomainDefFormatXML(driver, def, flags);
1499 1500
}

1501

1502
void qemuDomainObjTaint(virQEMUDriverPtr driver,
1503
                        virDomainObjPtr obj,
1504 1505
                        enum virDomainTaintFlags taint,
                        int logFD)
1506
{
1507 1508
    virErrorPtr orig_err = NULL;

1509 1510 1511 1512 1513 1514 1515 1516 1517
    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));
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531

        /* 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);
        }
1532 1533 1534 1535
    }
}


1536
void qemuDomainObjCheckTaint(virQEMUDriverPtr driver,
1537 1538
                             virDomainObjPtr obj,
                             int logFD)
1539
{
1540
    size_t i;
1541
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1542

1543 1544 1545 1546
    if (cfg->privileged &&
        (!cfg->clearEmulatorCapabilities ||
         cfg->user == 0 ||
         cfg->group == 0))
1547
        qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_HIGH_PRIVILEGES, logFD);
1548 1549 1550 1551

    if (obj->def->namespaceData) {
        qemuDomainCmdlineDefPtr qemucmd = obj->def->namespaceData;
        if (qemucmd->num_args || qemucmd->num_env)
1552
            qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_CUSTOM_ARGV, logFD);
1553 1554
    }

1555 1556 1557
    if (obj->def->cpu && obj->def->cpu->mode == VIR_CPU_MODE_HOST_PASSTHROUGH)
        qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_HOST_CPU, logFD);

1558
    for (i = 0; i < obj->def->ndisks; i++)
1559
        qemuDomainObjCheckDiskTaint(driver, obj, obj->def->disks[i], logFD);
1560

1561
    for (i = 0; i < obj->def->nnets; i++)
1562
        qemuDomainObjCheckNetTaint(driver, obj, obj->def->nets[i], logFD);
1563 1564

    virObjectUnref(cfg);
1565 1566 1567
}


1568
void qemuDomainObjCheckDiskTaint(virQEMUDriverPtr driver,
1569
                                 virDomainObjPtr obj,
1570 1571
                                 virDomainDiskDefPtr disk,
                                 int logFD)
1572
{
1573 1574
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);

1575
    if ((!disk->format || disk->format == VIR_STORAGE_FILE_AUTO) &&
1576
        cfg->allowDiskFormatProbing)
1577
        qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_DISK_PROBING, logFD);
1578

1579
    if (disk->rawio == 1)
1580
        qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_HIGH_PRIVILEGES, logFD);
1581 1582

    virObjectUnref(cfg);
1583 1584 1585
}


1586
void qemuDomainObjCheckNetTaint(virQEMUDriverPtr driver,
1587
                                virDomainObjPtr obj,
1588 1589
                                virDomainNetDefPtr net,
                                int logFD)
1590
{
1591 1592 1593 1594 1595 1596
    /* 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)
1597
        qemuDomainObjTaint(driver, obj, VIR_DOMAIN_TAINT_SHELL_SCRIPTS, logFD);
1598
}
1599 1600 1601


static int
1602
qemuDomainOpenLogHelper(virQEMUDriverConfigPtr cfg,
1603
                        virDomainObjPtr vm,
E
Eric Blake 已提交
1604
                        int oflags,
1605 1606 1607 1608
                        mode_t mode)
{
    char *logfile;
    int fd = -1;
1609
    bool trunc = false;
1610

1611
    if (virAsprintf(&logfile, "%s/%s.log", cfg->logDir, vm->def->name) < 0)
1612 1613
        return -1;

1614 1615 1616 1617 1618 1619 1620 1621 1622
    /* 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
1623
    if ((fd = open(logfile, oflags, mode)) < 0) {
1624 1625 1626 1627 1628 1629 1630
        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);
1631 1632 1633 1634 1635 1636 1637
        VIR_FORCE_CLOSE(fd);
        goto cleanup;
    }
    if (trunc &&
        ftruncate(fd, 0) < 0) {
        virReportSystemError(errno, _("failed to truncate %s"),
                             logfile);
1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648
        VIR_FORCE_CLOSE(fd);
        goto cleanup;
    }

cleanup:
    VIR_FREE(logfile);
    return fd;
}


int
1649
qemuDomainCreateLog(virQEMUDriverPtr driver, virDomainObjPtr vm,
E
Eric Blake 已提交
1650
                    bool append)
1651
{
1652
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
E
Eric Blake 已提交
1653
    int oflags;
1654
    int ret;
1655

E
Eric Blake 已提交
1656
    oflags = O_CREAT | O_WRONLY;
1657
    /* Only logrotate files in /var/log, so only append if running privileged */
1658
    if (cfg->privileged || append)
E
Eric Blake 已提交
1659
        oflags |= O_APPEND;
1660
    else
E
Eric Blake 已提交
1661
        oflags |= O_TRUNC;
1662

1663 1664 1665
    ret = qemuDomainOpenLogHelper(cfg, vm, oflags, S_IRUSR | S_IWUSR);
    virObjectUnref(cfg);
    return ret;
1666 1667 1668 1669
}


int
1670
qemuDomainOpenLog(virQEMUDriverPtr driver, virDomainObjPtr vm, off_t pos)
1671
{
1672
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1673 1674 1675 1676
    int fd;
    off_t off;
    int whence;

1677 1678 1679
    fd = qemuDomainOpenLogHelper(cfg, vm, O_RDONLY, 0);
    virObjectUnref(cfg);
    if (fd < 0)
1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705
        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;
}


1706
int qemuDomainAppendLog(virQEMUDriverPtr driver,
1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721
                        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;

1722
    if (virVasprintf(&message, fmt, argptr) < 0)
1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737
        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
1738
    VIR_FREE(message);
1739 1740
    return ret;
}
1741 1742 1743

/* Locate an appropriate 'qemu-img' binary.  */
const char *
1744
qemuFindQemuImgBinary(virQEMUDriverPtr driver)
1745
{
1746 1747 1748
    if (!driver->qemuImgBinary)
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("unable to find kvm-img or qemu-img"));
1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765

    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,
1766 1767
                                        QEMU_DOMAIN_FORMAT_LIVE_FLAGS, 1);
    if (newxml == NULL)
1768 1769
        return -1;

1770
    if (virAsprintf(&snapDir, "%s/%s", snapshotDir, vm->def->name) < 0)
1771 1772 1773 1774 1775 1776 1777
        goto cleanup;
    if (virFileMakePath(snapDir) < 0) {
        virReportSystemError(errno, _("cannot create snapshot directory '%s'"),
                             snapDir);
        goto cleanup;
    }

1778
    if (virAsprintf(&snapFile, "%s/%s.xml", snapDir, snapshot->def->name) < 0)
1779 1780
        goto cleanup;

J
Ján Tomko 已提交
1781
    ret = virXMLSaveFile(snapFile, NULL, "snapshot-edit", newxml);
1782 1783 1784 1785 1786 1787 1788 1789 1790 1791

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.  */
1792
static int
1793
qemuDomainSnapshotForEachQcow2Raw(virQEMUDriverPtr driver,
1794 1795 1796 1797 1798
                                  virDomainDefPtr def,
                                  const char *name,
                                  const char *op,
                                  bool try_all,
                                  int ndisks)
1799 1800
{
    const char *qemuimgarg[] = { NULL, "snapshot", NULL, NULL, NULL, NULL };
1801
    size_t i;
1802 1803 1804 1805 1806 1807 1808 1809 1810
    bool skipped = false;

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

    qemuimgarg[2] = op;
1811
    qemuimgarg[3] = name;
1812

1813
    for (i = 0; i < ndisks; i++) {
1814
        /* FIXME: we also need to handle LVM here */
1815
        if (def->disks[i]->device == VIR_DOMAIN_DISK_DEVICE_DISK) {
1816 1817
            if (def->disks[i]->format > 0 &&
                def->disks[i]->format != VIR_STORAGE_FILE_QCOW2) {
1818 1819 1820 1821 1822
                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",
1823
                             def->disks[i]->dst);
1824 1825
                    skipped = true;
                    continue;
1826 1827 1828 1829 1830
                } else if (STREQ(op, "-c") && i) {
                    /* We must roll back partial creation by deleting
                     * all earlier snapshots.  */
                    qemuDomainSnapshotForEachQcow2Raw(driver, def, name,
                                                      "-d", false, i);
1831
                }
1832 1833 1834 1835
                virReportError(VIR_ERR_OPERATION_INVALID,
                               _("Disk device '%s' does not support"
                                 " snapshotting"),
                               def->disks[i]->dst);
1836 1837 1838
                return -1;
            }

1839
            qemuimgarg[4] = def->disks[i]->src;
1840 1841 1842 1843

            if (virRun(qemuimgarg, NULL) < 0) {
                if (try_all) {
                    VIR_WARN("skipping snapshot action on %s",
1844
                             def->disks[i]->dst);
1845 1846
                    skipped = true;
                    continue;
1847 1848 1849 1850 1851
                } else if (STREQ(op, "-c") && i) {
                    /* We must roll back partial creation by deleting
                     * all earlier snapshots.  */
                    qemuDomainSnapshotForEachQcow2Raw(driver, def, name,
                                                      "-d", false, i);
1852 1853 1854 1855 1856 1857 1858 1859 1860
                }
                return -1;
            }
        }
    }

    return skipped ? 1 : 0;
}

1861 1862 1863
/* 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
1864
qemuDomainSnapshotForEachQcow2(virQEMUDriverPtr driver,
1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880
                               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);
}

1881 1882
/* Discard one snapshot (or its metadata), without reparenting any children.  */
int
1883
qemuDomainSnapshotDiscard(virQEMUDriverPtr driver,
1884 1885 1886 1887 1888 1889 1890 1891 1892
                          virDomainObjPtr vm,
                          virDomainSnapshotObjPtr snap,
                          bool update_current,
                          bool metadata_only)
{
    char *snapFile = NULL;
    int ret = -1;
    qemuDomainObjPrivatePtr priv;
    virDomainSnapshotObjPtr parentsnap = NULL;
1893
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1894 1895 1896 1897 1898 1899 1900 1901 1902

    if (!metadata_only) {
        if (!virDomainObjIsActive(vm)) {
            /* Ignore any skipped disks */
            if (qemuDomainSnapshotForEachQcow2(driver, vm, snap, "-d",
                                               true) < 0)
                goto cleanup;
        } else {
            priv = vm->privateData;
1903
            qemuDomainObjEnterMonitor(driver, vm);
1904 1905
            /* we continue on even in the face of error */
            qemuMonitorDeleteSnapshot(priv->mon, snap->def->name);
1906
            qemuDomainObjExitMonitor(driver, vm);
1907 1908 1909
        }
    }

1910
    if (virAsprintf(&snapFile, "%s/%s/%s.xml", cfg->snapshotDir,
1911
                    vm->def->name, snap->def->name) < 0)
1912 1913 1914 1915
        goto cleanup;

    if (snap == vm->current_snapshot) {
        if (update_current && snap->def->parent) {
1916
            parentsnap = virDomainSnapshotFindByName(vm->snapshots,
1917 1918 1919 1920 1921 1922 1923
                                                     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,
1924
                                                    cfg->snapshotDir) < 0) {
1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936
                    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);
1937
    virDomainSnapshotObjListRemove(vm->snapshots, snap);
1938 1939 1940 1941 1942

    ret = 0;

cleanup:
    VIR_FREE(snapFile);
1943
    virObjectUnref(cfg);
1944 1945 1946 1947 1948 1949 1950 1951 1952
    return ret;
}

/* Hash iterator callback to discard multiple snapshots.  */
void qemuDomainSnapshotDiscardAll(void *payload,
                                  const void *name ATTRIBUTE_UNUSED,
                                  void *data)
{
    virDomainSnapshotObjPtr snap = payload;
1953
    virQEMUSnapRemovePtr curr = data;
1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964
    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
1965
qemuDomainSnapshotDiscardAllMetadata(virQEMUDriverPtr driver,
1966 1967
                                     virDomainObjPtr vm)
{
1968
    virQEMUSnapRemove rem;
1969 1970 1971 1972 1973

    rem.driver = driver;
    rem.vm = vm;
    rem.metadata_only = true;
    rem.err = 0;
1974 1975
    virDomainSnapshotForEach(vm->snapshots, qemuDomainSnapshotDiscardAll,
                             &rem);
1976 1977 1978 1979 1980

    return rem.err;
}

/*
1981
 * The caller must hold a lock the vm and there must
1982 1983 1984
 * be no remaining references to vm.
 */
void
1985
qemuDomainRemoveInactive(virQEMUDriverPtr driver,
1986 1987
                         virDomainObjPtr vm)
{
1988
    char *snapDir;
1989
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1990

1991 1992 1993 1994 1995
    /* 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);
    }
1996
    else if (virAsprintf(&snapDir, "%s/%s", cfg->snapshotDir,
1997 1998
                         vm->def->name) < 0) {
        VIR_WARN("unable to remove snapshot directory %s/%s",
1999
                 cfg->snapshotDir, vm->def->name);
2000 2001 2002 2003 2004
    } else {
        if (rmdir(snapDir) < 0 && errno != ENOENT)
            VIR_WARN("unable to remove snapshot directory %s", snapDir);
        VIR_FREE(snapDir);
    }
2005
    virDomainObjListRemove(driver->domains, vm);
2006
    virObjectUnref(cfg);
2007
}
2008 2009

void
2010
qemuDomainSetFakeReboot(virQEMUDriverPtr driver,
2011 2012 2013 2014
                        virDomainObjPtr vm,
                        bool value)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2015
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2016 2017

    if (priv->fakeReboot == value)
2018
        goto cleanup;
2019 2020 2021

    priv->fakeReboot = value;

2022
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
2023
        VIR_WARN("Failed to save status on vm %s", vm->def->name);
2024

2025
cleanup:
2026
    virObjectUnref(cfg);
2027
}
M
Michal Privoznik 已提交
2028

2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048
static int
qemuDomainCheckDiskStartupPolicy(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm,
                                 virDomainDiskDefPtr disk,
                                 bool cold_boot)
{
    char uuid[VIR_UUID_STRING_BUFLEN];
    virDomainEventPtr event = NULL;
    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:
2049
            if (cold_boot)
2050 2051 2052 2053 2054 2055 2056 2057 2058
                goto error;
            break;

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

2059
    virResetLastError();
2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076
    VIR_DEBUG("Dropping disk '%s' on domain '%s' (UUID '%s') "
              "due to inaccessible source '%s'",
              disk->dst, vm->def->name, uuid, disk->src);

    event = virDomainEventDiskChangeNewFromObj(vm, disk->src, NULL, disk->info.alias,
                                               VIR_DOMAIN_EVENT_DISK_CHANGE_MISSING_ON_START);
    if (event)
        qemuDomainEventQueue(driver, event);

    VIR_FREE(disk->src);

    return 0;

error:
    return -1;
}

M
Michal Privoznik 已提交
2077
int
2078
qemuDomainCheckDiskPresence(virQEMUDriverPtr driver,
M
Michal Privoznik 已提交
2079
                            virDomainObjPtr vm,
2080
                            bool cold_boot)
M
Michal Privoznik 已提交
2081 2082
{
    int ret = -1;
2083
    size_t i;
M
Michal Privoznik 已提交
2084 2085
    virDomainDiskDefPtr disk;

2086
    VIR_DEBUG("Checking for disk presence");
M
Michal Privoznik 已提交
2087 2088 2089
    for (i = 0; i < vm->def->ndisks; i++) {
        disk = vm->def->disks[i];

2090
        if (!disk->src)
M
Michal Privoznik 已提交
2091 2092
            continue;

2093 2094
        if (qemuDomainDetermineDiskChain(driver, disk, false) >= 0 &&
            qemuDiskChainCheckBroken(disk) >= 0)
M
Michal Privoznik 已提交
2095
            continue;
2096 2097 2098 2099 2100

        if (disk->startupPolicy) {
            if (qemuDomainCheckDiskStartupPolicy(driver, vm, disk,
                                                 cold_boot) >= 0)
                continue;
M
Michal Privoznik 已提交
2101 2102
        }

2103
        goto error;
M
Michal Privoznik 已提交
2104 2105 2106 2107
    }

    ret = 0;

2108
error:
M
Michal Privoznik 已提交
2109 2110
    return ret;
}
2111 2112 2113 2114 2115 2116 2117 2118 2119 2120

/*
 * The vm must be locked when any of the following cleanup functions is
 * called.
 */
int
qemuDomainCleanupAdd(virDomainObjPtr vm,
                     qemuDomainCleanupCallback cb)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2121
    size_t i;
2122 2123 2124 2125 2126 2127 2128 2129 2130 2131

    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,
2132
                     priv->ncleanupCallbacks, 1) < 0)
2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143
        return -1;

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

void
qemuDomainCleanupRemove(virDomainObjPtr vm,
                        qemuDomainCleanupCallback cb)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2144
    size_t i;
2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162

    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
2163
qemuDomainCleanupRun(virQEMUDriverPtr driver,
2164 2165 2166
                     virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2167
    size_t i;
2168 2169 2170 2171

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

    /* run cleanup callbacks in reverse order */
2172 2173
    for (i = 0; i < priv->ncleanupCallbacks; i++) {
        if (priv->cleanupCallbacks[priv->ncleanupCallbacks - (i + 1)])
2174 2175 2176 2177 2178 2179 2180
            priv->cleanupCallbacks[i](driver, vm);
    }

    VIR_FREE(priv->cleanupCallbacks);
    priv->ncleanupCallbacks = 0;
    priv->ncleanupCallbacks_max = 0;
}
2181

2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203
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;
}

2204
int
2205
qemuDomainDetermineDiskChain(virQEMUDriverPtr driver,
2206 2207 2208
                             virDomainDiskDefPtr disk,
                             bool force)
{
2209 2210
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = 0;
2211

2212 2213 2214
    if (!disk->src ||
        disk->type == VIR_DOMAIN_DISK_TYPE_NETWORK ||
        disk->type == VIR_DOMAIN_DISK_TYPE_VOLUME)
2215
        goto cleanup;
2216 2217 2218 2219 2220 2221

    if (disk->backingChain) {
        if (force) {
            virStorageFileFreeMetadata(disk->backingChain);
            disk->backingChain = NULL;
        } else {
2222
            goto cleanup;
2223 2224 2225
        }
    }
    disk->backingChain = virStorageFileGetMetadata(disk->src, disk->format,
2226 2227
                                                   cfg->user, cfg->group,
                                                   cfg->allowDiskFormatProbing);
2228
    if (!disk->backingChain)
2229 2230 2231 2232 2233
        ret = -1;

cleanup:
    virObjectUnref(cfg);
    return ret;
2234
}
2235 2236 2237 2238 2239 2240


unsigned long long
qemuDomainMemoryLimit(virDomainDefPtr def)
{
    unsigned long long mem;
2241
    size_t i;
2242 2243 2244 2245 2246 2247 2248 2249 2250

    if (def->mem.hard_limit) {
        mem = def->mem.hard_limit;
    } else {
        /* If there is no hard_limit set, compute a reasonable one to avoid
         * system thrashing caused by exploited qemu.  A 'reasonable
         * limit' has been chosen:
         *     (1 + k) * (domain memory + total video memory) + (32MB for
         *     cache per each disk) + F
2251
         * where k = 0.5 and F = 400MB.  The cache for disks is important as
2252
         * kernel cache on the host side counts into the RSS limit.
2253 2254 2255
         * Moreover, VFIO requires some amount for IO space. Alex Williamson
         * suggested adding 1GiB for IO space just to be safe (some finer
         * tuning might be nice, though).
2256 2257 2258 2259 2260 2261 2262 2263 2264 2265
         *
         * Technically, the disk cache does not have to be included in
         * RLIMIT_MEMLOCK but it doesn't hurt as it's just an upper limit and
         * it makes this function and its usage simpler.
         */
        mem = def->mem.max_balloon;
        for (i = 0; i < def->nvideos; i++)
            mem += def->videos[i]->vram;
        mem *= 1.5;
        mem += def->ndisks * 32768;
2266
        mem += 409600;
2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278

        for (i = 0; i < def->nhostdevs; i++) {
            virDomainHostdevDefPtr hostdev = def->hostdevs[i];
            if (hostdev->mode == VIR_DOMAIN_HOSTDEV_MODE_SUBSYS &&
                hostdev->source.subsys.type ==
                    VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI &&
                hostdev->source.subsys.u.pci.backend ==
                    VIR_DOMAIN_HOSTDEV_PCI_BACKEND_VFIO) {
                mem += 1024 * 1024;
                break;
            }
        }
2279 2280 2281 2282
    }

    return mem;
}
2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305


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;
}