qemu_process.c 182.3 KB
Newer Older
1
/*
2
 * qemu_process.c: QEMU process management
3
 *
4
 * Copyright (C) 2006-2016 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23 24 25 26 27
 *
 */

#include <config.h>

#include <fcntl.h>
#include <unistd.h>
#include <signal.h>
#include <sys/stat.h>
R
Roman Bogorodskiy 已提交
28 29 30 31 32 33
#if defined(__linux__)
# include <linux/capability.h>
#elif defined(__FreeBSD__)
# include <sys/param.h>
# include <sys/cpuset.h>
#endif
34 35

#include "qemu_process.h"
36
#include "qemu_processpriv.h"
37 38 39 40 41 42 43
#include "qemu_domain.h"
#include "qemu_cgroup.h"
#include "qemu_capabilities.h"
#include "qemu_monitor.h"
#include "qemu_command.h"
#include "qemu_hostdev.h"
#include "qemu_hotplug.h"
44
#include "qemu_migration.h"
45
#include "qemu_interface.h"
46

47
#include "cpu/cpu.h"
48
#include "datatypes.h"
49
#include "virlog.h"
50
#include "virerror.h"
51
#include "viralloc.h"
52
#include "virhook.h"
E
Eric Blake 已提交
53
#include "virfile.h"
54
#include "virpidfile.h"
55
#include "nodeinfo.h"
56
#include "domain_audit.h"
57
#include "domain_nwfilter.h"
58
#include "locking/domain_lock.h"
59
#include "network/bridge_driver.h"
60
#include "viruuid.h"
61
#include "virprocess.h"
62
#include "virtime.h"
A
Ansis Atteka 已提交
63
#include "virnetdevtap.h"
64
#include "virnetdevopenvswitch.h"
65
#include "virnetdevmidonet.h"
66
#include "virbitmap.h"
67
#include "viratomic.h"
68
#include "virnuma.h"
69
#include "virstring.h"
70
#include "virhostdev.h"
71
#include "storage/storage_driver.h"
72
#include "configmake.h"
73
#include "nwfilter_conf.h"
74
#include "netdev_bandwidth_conf.h"
75 76 77

#define VIR_FROM_THIS VIR_FROM_QEMU

78 79
VIR_LOG_INIT("qemu.qemu_process");

80
/**
81
 * qemuProcessRemoveDomainStatus
82 83 84 85 86 87
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
88
qemuProcessRemoveDomainStatus(virQEMUDriverPtr driver,
89 90 91 92
                              virDomainObjPtr vm)
{
    char ebuf[1024];
    char *file = NULL;
93
    qemuDomainObjPrivatePtr priv = vm->privateData;
94 95
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;
96

97
    if (virAsprintf(&file, "%s/%s.xml", cfg->stateDir, vm->def->name) < 0)
98
        goto cleanup;
99 100 101 102 103 104

    if (unlink(file) < 0 && errno != ENOENT && errno != ENOTDIR)
        VIR_WARN("Failed to remove domain XML for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));
    VIR_FREE(file);

105 106 107
    if (priv->pidfile &&
        unlink(priv->pidfile) < 0 &&
        errno != ENOENT)
108 109 110
        VIR_WARN("Failed to remove PID file for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));

111
    ret = 0;
112
 cleanup:
113 114
    virObjectUnref(cfg);
    return ret;
115 116 117 118
}


/* XXX figure out how to remove this */
119
extern virQEMUDriverPtr qemu_driver;
120

D
Daniel P. Berrange 已提交
121 122 123 124 125 126 127
/*
 * This is a callback registered with a qemuAgentPtr instance,
 * and to be invoked when the agent console hits an end of file
 * condition, or error, thus indicating VM shutdown should be
 * performed
 */
static void
128
qemuProcessHandleAgentEOF(qemuAgentPtr agent,
D
Daniel P. Berrange 已提交
129 130 131 132 133 134
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

    VIR_DEBUG("Received EOF from agent on %p '%s'", vm, vm->def->name);

135
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
136 137

    priv = vm->privateData;
138 139 140 141 142 143 144 145 146 147 148

    if (!priv->agent) {
        VIR_DEBUG("Agent freed already");
        goto unlock;
    }

    if (priv->beingDestroyed) {
        VIR_DEBUG("Domain is being destroyed, agent EOF is expected");
        goto unlock;
    }

149
    qemuAgentClose(agent);
150
    priv->agent = NULL;
D
Daniel P. Berrange 已提交
151

152
    virObjectUnlock(vm);
153 154
    return;

155
 unlock:
156 157
    virObjectUnlock(vm);
    return;
D
Daniel P. Berrange 已提交
158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
}


/*
 * This is invoked when there is some kind of error
 * parsing data to/from the agent. The VM can continue
 * to run, but no further agent commands will be
 * allowed
 */
static void
qemuProcessHandleAgentError(qemuAgentPtr agent ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

    VIR_DEBUG("Received error from agent on %p '%s'", vm, vm->def->name);

175
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
176 177 178 179 180

    priv = vm->privateData;

    priv->agentError = true;

181
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
182 183 184 185 186
}

static void qemuProcessHandleAgentDestroy(qemuAgentPtr agent,
                                          virDomainObjPtr vm)
{
187 188
    VIR_DEBUG("Received destroy agent=%p vm=%p", agent, vm);

189
    virObjectUnref(vm);
D
Daniel P. Berrange 已提交
190 191 192 193 194 195 196 197 198 199
}


static qemuAgentCallbacks agentCallbacks = {
    .destroy = qemuProcessHandleAgentDestroy,
    .eofNotify = qemuProcessHandleAgentEOF,
    .errorNotify = qemuProcessHandleAgentError,
};


200
int
201
qemuConnectAgent(virQEMUDriverPtr driver, virDomainObjPtr vm)
D
Daniel P. Berrange 已提交
202 203 204 205
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    qemuAgentPtr agent = NULL;
206
    virDomainChrDefPtr config = qemuFindAgentConfig(vm->def);
D
Daniel P. Berrange 已提交
207 208 209 210 211 212 213 214 215 216 217 218 219

    if (!config)
        return 0;

    if (virSecurityManagerSetDaemonSocketLabel(driver->securityManager,
                                               vm->def) < 0) {
        VIR_ERROR(_("Failed to set security context for agent for %s"),
                  vm->def->name);
        goto cleanup;
    }

    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted while the agent is active */
220
    virObjectRef(vm);
D
Daniel P. Berrange 已提交
221 222

    ignore_value(virTimeMillisNow(&priv->agentStart));
223
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
224 225

    agent = qemuAgentOpen(vm,
226
                          &config->source,
D
Daniel P. Berrange 已提交
227 228
                          &agentCallbacks);

229
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
230 231
    priv->agentStart = 0;

232 233 234 235 236 237 238 239 240 241 242
    if (agent == NULL)
        virObjectUnref(vm);

    if (!virDomainObjIsActive(vm)) {
        qemuAgentClose(agent);
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest crashed while connecting to the guest agent"));
        ret = -2;
        goto cleanup;
    }

D
Daniel P. Berrange 已提交
243 244 245 246
    if (virSecurityManagerClearSocketLabel(driver->securityManager,
                                           vm->def) < 0) {
        VIR_ERROR(_("Failed to clear security context for agent for %s"),
                  vm->def->name);
247
        qemuAgentClose(agent);
D
Daniel P. Berrange 已提交
248 249 250 251 252 253 254 255 256 257 258 259 260
        goto cleanup;
    }


    priv->agent = agent;

    if (priv->agent == NULL) {
        VIR_INFO("Failed to connect agent for %s", vm->def->name);
        goto cleanup;
    }

    ret = 0;

261
 cleanup:
D
Daniel P. Berrange 已提交
262 263 264 265
    return ret;
}


266
/*
267
 * This is a callback registered with a qemuMonitorPtr instance,
268 269 270 271 272 273
 * and to be invoked when the monitor console hits an end of file
 * condition, or error, thus indicating VM shutdown should be
 * performed
 */
static void
qemuProcessHandleMonitorEOF(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
274 275
                            virDomainObjPtr vm,
                            void *opaque)
276
{
277
    virQEMUDriverPtr driver = opaque;
278
    virObjectEventPtr event = NULL;
279
    qemuDomainObjPrivatePtr priv;
280 281 282
    int eventReason = VIR_DOMAIN_EVENT_STOPPED_SHUTDOWN;
    int stopReason = VIR_DOMAIN_SHUTOFF_SHUTDOWN;
    const char *auditReason = "shutdown";
283
    unsigned int stopFlags = 0;
284 285 286

    VIR_DEBUG("Received EOF on %p '%s'", vm, vm->def->name);

287
    virObjectLock(vm);
288

289 290 291 292
    priv = vm->privateData;

    if (priv->beingDestroyed) {
        VIR_DEBUG("Domain is being destroyed, EOF is expected");
293
        goto cleanup;
294 295
    }

296 297
    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Domain %p is not active, ignoring EOF", vm);
298
        goto cleanup;
299 300
    }

301
    if (priv->monJSON && !priv->gotShutdown) {
302 303
        VIR_DEBUG("Monitor connection to '%s' closed without SHUTDOWN event; "
                  "assuming the domain crashed", vm->def->name);
304
        eventReason = VIR_DOMAIN_EVENT_STOPPED_FAILED;
305
        stopReason = VIR_DOMAIN_SHUTOFF_CRASHED;
306
        auditReason = "failed";
307 308
    }

309 310
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN) {
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
311 312
        qemuMigrationErrorSave(driver, vm->def->name,
                               qemuMonitorLastError(priv->mon));
313
    }
314

315
    event = virDomainEventLifecycleNewFromObj(vm,
316
                                     VIR_DOMAIN_EVENT_STOPPED,
317
                                     eventReason);
318
    qemuProcessStop(driver, vm, stopReason, stopFlags);
319
    virDomainAuditStop(vm, auditReason);
320

321
    qemuDomainRemoveInactive(driver, vm);
322

323
 cleanup:
324
    virObjectUnlock(vm);
325
    qemuDomainEventQueue(driver, event);
326 327 328 329 330 331 332 333 334 335 336
}


/*
 * This is invoked when there is some kind of error
 * parsing data to/from the monitor. The VM can continue
 * to run, but no further monitor commands will be
 * allowed
 */
static void
qemuProcessHandleMonitorError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
337 338
                              virDomainObjPtr vm,
                              void *opaque)
339
{
340
    virQEMUDriverPtr driver = opaque;
341
    virObjectEventPtr event = NULL;
342 343 344

    VIR_DEBUG("Received error on %p '%s'", vm, vm->def->name);

345
    virObjectLock(vm);
346

347
    ((qemuDomainObjPrivatePtr) vm->privateData)->monError = true;
348
    event = virDomainEventControlErrorNewFromObj(vm);
349
    qemuDomainEventQueue(driver, event);
350

351
    virObjectUnlock(vm);
352 353 354
}


355
virDomainDiskDefPtr
356 357 358
qemuProcessFindDomainDiskByAlias(virDomainObjPtr vm,
                                 const char *alias)
{
359
    size_t i;
360 361 362 363 364 365 366 367 368 369 370 371

    if (STRPREFIX(alias, QEMU_DRIVE_HOST_PREFIX))
        alias += strlen(QEMU_DRIVE_HOST_PREFIX);

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk;

        disk = vm->def->disks[i];
        if (disk->info.alias != NULL && STREQ(disk->info.alias, alias))
            return disk;
    }

372 373 374
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with alias %s"),
                   alias);
375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390
    return NULL;
}

static int
qemuProcessGetVolumeQcowPassphrase(virConnectPtr conn,
                                   virDomainDiskDefPtr disk,
                                   char **secretRet,
                                   size_t *secretLen)
{
    virSecretPtr secret;
    char *passphrase;
    unsigned char *data;
    size_t size;
    int ret = -1;
    virStorageEncryptionPtr enc;

391
    if (!disk->src->encryption) {
392 393
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("disk %s does not have any encryption information"),
394
                       disk->src->path);
395 396
        return -1;
    }
397
    enc = disk->src->encryption;
398 399

    if (!conn) {
400 401
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("cannot find secrets without a connection"));
402 403 404 405
        goto cleanup;
    }

    if (conn->secretDriver == NULL ||
406 407
        conn->secretDriver->secretLookupByUUID == NULL ||
        conn->secretDriver->secretGetValue == NULL) {
408 409
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("secret storage not supported"));
410 411 412 413 414 415 416
        goto cleanup;
    }

    if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW ||
        enc->nsecrets != 1 ||
        enc->secrets[0]->type !=
        VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE) {
417
        virReportError(VIR_ERR_XML_ERROR,
418 419
                       _("invalid <encryption> for volume %s"),
                       virDomainDiskGetSource(disk));
420 421 422
        goto cleanup;
    }

423 424
    secret = conn->secretDriver->secretLookupByUUID(conn,
                                                    enc->secrets[0]->uuid);
425 426
    if (secret == NULL)
        goto cleanup;
427 428
    data = conn->secretDriver->secretGetValue(secret, &size, 0,
                                              VIR_SECRET_GET_VALUE_INTERNAL_CALL);
429
    virObjectUnref(secret);
430 431 432 433 434 435
    if (data == NULL)
        goto cleanup;

    if (memchr(data, '\0', size) != NULL) {
        memset(data, 0, size);
        VIR_FREE(data);
436 437
        virReportError(VIR_ERR_XML_ERROR,
                       _("format='qcow' passphrase for %s must not contain a "
438
                         "'\\0'"), virDomainDiskGetSource(disk));
439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457
        goto cleanup;
    }

    if (VIR_ALLOC_N(passphrase, size + 1) < 0) {
        memset(data, 0, size);
        VIR_FREE(data);
        goto cleanup;
    }
    memcpy(passphrase, data, size);
    passphrase[size] = '\0';

    memset(data, 0, size);
    VIR_FREE(data);

    *secretRet = passphrase;
    *secretLen = size;

    ret = 0;

458
 cleanup:
459 460 461 462 463 464 465 466 467
    return ret;
}

static int
qemuProcessFindVolumeQcowPassphrase(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virConnectPtr conn,
                                    virDomainObjPtr vm,
                                    const char *path,
                                    char **secretRet,
468 469
                                    size_t *secretLen,
                                    void *opaque ATTRIBUTE_UNUSED)
470 471 472 473
{
    virDomainDiskDefPtr disk;
    int ret = -1;

474
    virObjectLock(vm);
475 476 477 478
    if (!(disk = virDomainDiskByName(vm->def, path, true))) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("no disk found with path %s"),
                       path);
479
        goto cleanup;
480
    }
481 482 483

    ret = qemuProcessGetVolumeQcowPassphrase(conn, disk, secretRet, secretLen);

484
 cleanup:
485
    virObjectUnlock(vm);
486 487 488 489 490 491
    return ret;
}


static int
qemuProcessHandleReset(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
492 493
                       virDomainObjPtr vm,
                       void *opaque)
494
{
495
    virQEMUDriverPtr driver = opaque;
496
    virObjectEventPtr event;
497
    qemuDomainObjPrivatePtr priv;
498
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
499

500
    virObjectLock(vm);
501

502
    event = virDomainEventRebootNewFromObj(vm);
503 504 505
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);
506 507 508

    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
        VIR_WARN("Failed to save status on vm %s", vm->def->name);
509

510
    virObjectUnlock(vm);
511

512
    qemuDomainEventQueue(driver, event);
513

514
    virObjectUnref(cfg);
515 516 517 518
    return 0;
}


519 520 521 522 523 524 525 526 527 528 529
/*
 * Since we have the '-no-shutdown' flag set, the
 * QEMU process will currently have guest OS shutdown
 * and the CPUS stopped. To fake the reboot, we thus
 * want todo a reset of the virtual hardware, followed
 * by restart of the CPUs. This should result in the
 * guest OS booting up again
 */
static void
qemuProcessFakeReboot(void *opaque)
{
530
    virQEMUDriverPtr driver = qemu_driver;
531 532
    virDomainObjPtr vm = opaque;
    qemuDomainObjPrivatePtr priv = vm->privateData;
533
    virObjectEventPtr event = NULL;
534
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
535
    virDomainRunningReason reason = VIR_DOMAIN_RUNNING_BOOTED;
536
    int ret = -1, rc;
537

538
    VIR_DEBUG("vm=%p", vm);
539
    virObjectLock(vm);
540
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
541 542 543
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
544 545
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
546 547 548
        goto endjob;
    }

549
    qemuDomainObjEnterMonitor(driver, vm);
550 551 552
    rc = qemuMonitorSystemReset(priv->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0)
553 554
        goto endjob;

555
    if (rc < 0)
556 557
        goto endjob;

558 559 560
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_CRASHED)
        reason = VIR_DOMAIN_RUNNING_CRASHED;

561
    if (qemuProcessStartCPUs(driver, vm, NULL,
562
                             reason,
563
                             QEMU_ASYNC_JOB_NONE) < 0) {
564
        if (virGetLastError() == NULL)
565 566
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
567 568
        goto endjob;
    }
569
    priv->gotShutdown = false;
570
    event = virDomainEventLifecycleNewFromObj(vm,
571 572 573
                                     VIR_DOMAIN_EVENT_RESUMED,
                                     VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

574 575 576 577 578
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

579 580
    ret = 0;

581
 endjob:
582
    qemuDomainObjEndJob(driver, vm);
583

584
 cleanup:
585 586
    if (ret == -1)
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_FORCE));
M
Michal Privoznik 已提交
587
    virDomainObjEndAPI(&vm);
588
    qemuDomainEventQueue(driver, event);
589
    virObjectUnref(cfg);
590 591 592
}


593
void
594
qemuProcessShutdownOrReboot(virQEMUDriverPtr driver,
595
                            virDomainObjPtr vm)
596
{
597 598 599
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
600
        qemuDomainSetFakeReboot(driver, vm, false);
601
        virObjectRef(vm);
602 603 604 605 606
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
607
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
608
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
609
            virObjectUnref(vm);
610 611
        }
    } else {
612
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
613
    }
614
}
615

616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636

static int
qemuProcessHandleEvent(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                       virDomainObjPtr vm,
                       const char *eventName,
                       long long seconds,
                       unsigned int micros,
                       const char *details,
                       void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    virObjectEventPtr event = NULL;

    VIR_DEBUG("vm=%p", vm);

    virObjectLock(vm);
    event = virDomainQemuMonitorEventNew(vm->def->id, vm->def->name,
                                         vm->def->uuid, eventName,
                                         seconds, micros, details);

    virObjectUnlock(vm);
637
    qemuDomainEventQueue(driver, event);
638 639 640 641 642

    return 0;
}


643 644
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
645 646
                          virDomainObjPtr vm,
                          void *opaque)
647
{
648
    virQEMUDriverPtr driver = opaque;
649
    qemuDomainObjPrivatePtr priv;
650
    virObjectEventPtr event = NULL;
651
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
652

653 654
    VIR_DEBUG("vm=%p", vm);

655
    virObjectLock(vm);
656 657 658 659 660 661

    priv = vm->privateData;
    if (priv->gotShutdown) {
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
662 663 664 665
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
666 667 668 669 670 671 672 673
    }
    priv->gotShutdown = true;

    VIR_DEBUG("Transitioned guest %s to shutdown state",
              vm->def->name);
    virDomainObjSetState(vm,
                         VIR_DOMAIN_SHUTDOWN,
                         VIR_DOMAIN_SHUTDOWN_UNKNOWN);
674
    event = virDomainEventLifecycleNewFromObj(vm,
675 676 677
                                     VIR_DOMAIN_EVENT_SHUTDOWN,
                                     VIR_DOMAIN_EVENT_SHUTDOWN_FINISHED);

678
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
679 680 681 682
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

683 684 685
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

686 687
    qemuProcessShutdownOrReboot(driver, vm);

688
 unlock:
689
    virObjectUnlock(vm);
690
    qemuDomainEventQueue(driver, event);
691
    virObjectUnref(cfg);
692

693 694 695 696 697 698
    return 0;
}


static int
qemuProcessHandleStop(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
699 700
                      virDomainObjPtr vm,
                      void *opaque)
701
{
702
    virQEMUDriverPtr driver = opaque;
703
    virObjectEventPtr event = NULL;
704
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
705

706
    virObjectLock(vm);
J
Jiri Denemark 已提交
707
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
708
        qemuDomainObjPrivatePtr priv = vm->privateData;
709

710
        if (priv->gotShutdown) {
711 712
            VIR_DEBUG("Ignoring STOP event after SHUTDOWN");
            goto unlock;
713 714
        }

715 716
        VIR_DEBUG("Transitioned guest %s to paused state",
                  vm->def->name);
717

718 719 720
        if (priv->job.current)
            ignore_value(virTimeMillisNow(&priv->job.current->stopped));

721
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_UNKNOWN);
722
        event = virDomainEventLifecycleNewFromObj(vm,
723 724 725
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);

726 727 728 729 730
        VIR_FREE(priv->lockState);
        if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
            VIR_WARN("Unable to release lease on %s", vm->def->name);
        VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));

731
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
732 733 734
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
735
    }
736

737
 unlock:
738
    virObjectUnlock(vm);
739
    qemuDomainEventQueue(driver, event);
740
    virObjectUnref(cfg);
741 742 743 744 745

    return 0;
}


746 747
static int
qemuProcessHandleResume(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
748 749
                        virDomainObjPtr vm,
                        void *opaque)
750
{
751
    virQEMUDriverPtr driver = opaque;
752
    virObjectEventPtr event = NULL;
753
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
754

755
    virObjectLock(vm);
756 757 758 759 760 761 762 763 764 765 766 767 768
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
        qemuDomainObjPrivatePtr priv = vm->privateData;

        if (priv->gotShutdown) {
            VIR_DEBUG("Ignoring RESUME event after SHUTDOWN");
            goto unlock;
        }

        VIR_DEBUG("Transitioned guest %s out of paused into resumed state",
                  vm->def->name);

        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_UNPAUSED);
769
        event = virDomainEventLifecycleNewFromObj(vm,
770 771 772 773
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

        VIR_DEBUG("Using lock state '%s' on resume event", NULLSTR(priv->lockState));
774
        if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
775 776 777 778 779 780 781 782 783
                                       vm, priv->lockState) < 0) {
            /* Don't free priv->lockState on error, because we need
             * to make sure we have state still present if the user
             * tries to resume again
             */
            goto unlock;
        }
        VIR_FREE(priv->lockState);

784
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
785 786 787 788 789
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
    }

790
 unlock:
791
    virObjectUnlock(vm);
792
    qemuDomainEventQueue(driver, event);
793
    virObjectUnref(cfg);
794 795 796
    return 0;
}

797 798 799
static int
qemuProcessHandleRTCChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm,
800 801
                           long long offset,
                           void *opaque)
802
{
803
    virQEMUDriverPtr driver = opaque;
804
    virObjectEventPtr event = NULL;
805
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
806

807
    virObjectLock(vm);
808

809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825
    if (vm->def->clock.offset == VIR_DOMAIN_CLOCK_OFFSET_VARIABLE) {
        /* when a basedate is manually given on the qemu commandline
         * rather than simply "-rtc base=utc", the offset sent by qemu
         * in this event is *not* the new offset from UTC, but is
         * instead the new offset from the *original basedate* +
         * uptime. For example, if the original offset was 3600 and
         * the guest clock has been advanced by 10 seconds, qemu will
         * send "10" in the event - this means that the new offset
         * from UTC is 3610, *not* 10. If the guest clock is advanced
         * by another 10 seconds, qemu will now send "20" - i.e. each
         * event is the sum of the most recent change and all previous
         * changes since the domain was started. Fortunately, we have
         * saved the initial offset in "adjustment0", so to arrive at
         * the proper new "adjustment", we just add the most recent
         * offset to adjustment0.
         */
        offset += vm->def->clock.data.variable.adjustment0;
826
        vm->def->clock.data.variable.adjustment = offset;
827

828 829 830 831 832
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
           VIR_WARN("unable to save domain status with RTC change");
    }

    event = virDomainEventRTCChangeNewFromObj(vm, offset);
833

834
    virObjectUnlock(vm);
835

836
    qemuDomainEventQueue(driver, event);
837
    virObjectUnref(cfg);
838 839 840 841 842 843 844
    return 0;
}


static int
qemuProcessHandleWatchdog(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
845 846
                          int action,
                          void *opaque)
847
{
848
    virQEMUDriverPtr driver = opaque;
849 850
    virObjectEventPtr watchdogEvent = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
851
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
852

853
    virObjectLock(vm);
854 855 856
    watchdogEvent = virDomainEventWatchdogNewFromObj(vm, action);

    if (action == VIR_DOMAIN_EVENT_WATCHDOG_PAUSE &&
J
Jiri Denemark 已提交
857
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
858
        qemuDomainObjPrivatePtr priv = vm->privateData;
859 860
        VIR_DEBUG("Transitioned guest %s to paused state due to watchdog", vm->def->name);

J
Jiri Denemark 已提交
861
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
862
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
863 864 865
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

866 867 868 869 870
        VIR_FREE(priv->lockState);
        if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
            VIR_WARN("Unable to release lease on %s", vm->def->name);
        VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));

871
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
872 873 874
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
875 876 877
    }

    if (vm->def->watchdog->action == VIR_DOMAIN_WATCHDOG_ACTION_DUMP) {
C
Chen Fan 已提交
878 879 880 881 882
        struct qemuProcessEvent *processEvent;
        if (VIR_ALLOC(processEvent) == 0) {
            processEvent->eventType = QEMU_PROCESS_EVENT_WATCHDOG;
            processEvent->action = VIR_DOMAIN_WATCHDOG_ACTION_DUMP;
            processEvent->vm = vm;
W
Wen Congyang 已提交
883 884 885
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
886
            virObjectRef(vm);
C
Chen Fan 已提交
887
            if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
888
                if (!virObjectUnref(vm))
889
                    vm = NULL;
C
Chen Fan 已提交
890
                VIR_FREE(processEvent);
W
Wen Congyang 已提交
891
            }
892
        }
893 894
    }

895
    if (vm)
896
        virObjectUnlock(vm);
897 898
    qemuDomainEventQueue(driver, watchdogEvent);
    qemuDomainEventQueue(driver, lifecycleEvent);
899

900
    virObjectUnref(cfg);
901 902 903 904 905 906 907 908 909
    return 0;
}


static int
qemuProcessHandleIOError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virDomainObjPtr vm,
                         const char *diskAlias,
                         int action,
910 911
                         const char *reason,
                         void *opaque)
912
{
913
    virQEMUDriverPtr driver = opaque;
914 915 916
    virObjectEventPtr ioErrorEvent = NULL;
    virObjectEventPtr ioErrorEvent2 = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
917 918 919
    const char *srcPath;
    const char *devAlias;
    virDomainDiskDefPtr disk;
920
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
921

922
    virObjectLock(vm);
923 924 925
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
926
        srcPath = virDomainDiskGetSource(disk);
927 928 929 930 931 932 933 934 935 936
        devAlias = disk->info.alias;
    } else {
        srcPath = "";
        devAlias = "";
    }

    ioErrorEvent = virDomainEventIOErrorNewFromObj(vm, srcPath, devAlias, action);
    ioErrorEvent2 = virDomainEventIOErrorReasonNewFromObj(vm, srcPath, devAlias, action, reason);

    if (action == VIR_DOMAIN_EVENT_IO_ERROR_PAUSE &&
J
Jiri Denemark 已提交
937
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
938
        qemuDomainObjPrivatePtr priv = vm->privateData;
939 940
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

941 942 943
        if (priv->signalIOError)
            virDomainObjBroadcast(vm);

J
Jiri Denemark 已提交
944
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
945
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
946 947 948
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

949 950 951 952 953
        VIR_FREE(priv->lockState);
        if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
            VIR_WARN("Unable to release lease on %s", vm->def->name);
        VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));

954
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
955 956
            VIR_WARN("Unable to save status on vm %s after IO error", vm->def->name);
    }
957
    virObjectUnlock(vm);
958

959 960 961
    qemuDomainEventQueue(driver, ioErrorEvent);
    qemuDomainEventQueue(driver, ioErrorEvent2);
    qemuDomainEventQueue(driver, lifecycleEvent);
962
    virObjectUnref(cfg);
963 964 965
    return 0;
}

966 967 968 969 970
static int
qemuProcessHandleBlockJob(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          const char *diskAlias,
                          int type,
971 972
                          int status,
                          void *opaque)
973
{
974
    virQEMUDriverPtr driver = opaque;
975
    struct qemuProcessEvent *processEvent = NULL;
976
    virDomainDiskDefPtr disk;
977
    qemuDomainDiskPrivatePtr diskPriv;
978
    char *data = NULL;
979

980
    virObjectLock(vm);
981

982 983
    VIR_DEBUG("Block job for device %s (domain: %p,%s) type %d status %d",
              diskAlias, vm, vm->def->name, type, status);
984

985
    if (!(disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias)))
986
        goto error;
987
    diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
988

989
    if (diskPriv->blockJobSync) {
990
        /* We have a SYNC API waiting for this event, dispatch it back */
991 992
        diskPriv->blockJobType = type;
        diskPriv->blockJobStatus = status;
993
        virDomainObjBroadcast(vm);
994 995 996 997 998 999 1000 1001 1002 1003 1004 1005
    } else {
        /* there is no waiting SYNC API, dispatch the update to a thread */
        if (VIR_ALLOC(processEvent) < 0)
            goto error;

        processEvent->eventType = QEMU_PROCESS_EVENT_BLOCK_JOB;
        if (VIR_STRDUP(data, diskAlias) < 0)
            goto error;
        processEvent->data = data;
        processEvent->vm = vm;
        processEvent->action = type;
        processEvent->status = status;
1006

1007 1008 1009 1010 1011
        virObjectRef(vm);
        if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
            ignore_value(virObjectUnref(vm));
            goto error;
        }
1012 1013
    }

1014
 cleanup:
1015
    virObjectUnlock(vm);
1016
    return 0;
1017 1018 1019 1020 1021
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
1022
}
1023

1024

1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036
static int
qemuProcessHandleGraphics(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          int phase,
                          int localFamily,
                          const char *localNode,
                          const char *localService,
                          int remoteFamily,
                          const char *remoteNode,
                          const char *remoteService,
                          const char *authScheme,
                          const char *x509dname,
1037 1038
                          const char *saslUsername,
                          void *opaque)
1039
{
1040
    virQEMUDriverPtr driver = opaque;
1041
    virObjectEventPtr event;
1042 1043 1044
    virDomainEventGraphicsAddressPtr localAddr = NULL;
    virDomainEventGraphicsAddressPtr remoteAddr = NULL;
    virDomainEventGraphicsSubjectPtr subject = NULL;
1045
    size_t i;
1046 1047

    if (VIR_ALLOC(localAddr) < 0)
1048
        goto error;
1049
    localAddr->family = localFamily;
1050 1051 1052
    if (VIR_STRDUP(localAddr->service, localService) < 0 ||
        VIR_STRDUP(localAddr->node, localNode) < 0)
        goto error;
1053 1054

    if (VIR_ALLOC(remoteAddr) < 0)
1055
        goto error;
1056
    remoteAddr->family = remoteFamily;
1057 1058 1059
    if (VIR_STRDUP(remoteAddr->service, remoteService) < 0 ||
        VIR_STRDUP(remoteAddr->node, remoteNode) < 0)
        goto error;
1060 1061

    if (VIR_ALLOC(subject) < 0)
1062
        goto error;
1063 1064
    if (x509dname) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1065
            goto error;
1066
        subject->nidentity++;
1067 1068 1069
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "x509dname") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, x509dname) < 0)
            goto error;
1070 1071 1072
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1073
            goto error;
1074
        subject->nidentity++;
1075 1076 1077
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "saslUsername") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, saslUsername) < 0)
            goto error;
1078 1079
    }

1080
    virObjectLock(vm);
1081
    event = virDomainEventGraphicsNewFromObj(vm, phase, localAddr, remoteAddr, authScheme, subject);
1082
    virObjectUnlock(vm);
1083

1084
    qemuDomainEventQueue(driver, event);
1085 1086 1087

    return 0;

1088
 error:
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099
    if (localAddr) {
        VIR_FREE(localAddr->service);
        VIR_FREE(localAddr->node);
        VIR_FREE(localAddr);
    }
    if (remoteAddr) {
        VIR_FREE(remoteAddr->service);
        VIR_FREE(remoteAddr->node);
        VIR_FREE(remoteAddr);
    }
    if (subject) {
1100
        for (i = 0; i < subject->nidentity; i++) {
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110
            VIR_FREE(subject->identities[i].type);
            VIR_FREE(subject->identities[i].name);
        }
        VIR_FREE(subject->identities);
        VIR_FREE(subject);
    }

    return -1;
}

1111 1112 1113 1114
static int
qemuProcessHandleTrayChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            const char *devAlias,
1115 1116
                            int reason,
                            void *opaque)
1117
{
1118
    virQEMUDriverPtr driver = opaque;
1119
    virObjectEventPtr event = NULL;
1120
    virDomainDiskDefPtr disk;
1121
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1122

1123
    virObjectLock(vm);
1124 1125 1126 1127 1128 1129
    disk = qemuProcessFindDomainDiskByAlias(vm, devAlias);

    if (disk) {
        event = virDomainEventTrayChangeNewFromObj(vm,
                                                   devAlias,
                                                   reason);
1130 1131 1132 1133 1134 1135
        /* Update disk tray status */
        if (reason == VIR_DOMAIN_EVENT_TRAY_CHANGE_OPEN)
            disk->tray_status = VIR_DOMAIN_DISK_TRAY_OPEN;
        else if (reason == VIR_DOMAIN_EVENT_TRAY_CHANGE_CLOSE)
            disk->tray_status = VIR_DOMAIN_DISK_TRAY_CLOSED;

1136
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1137 1138 1139
            VIR_WARN("Unable to save status on vm %s after tray moved event",
                     vm->def->name);
        }
1140 1141

        virDomainObjBroadcast(vm);
1142 1143
    }

1144
    virObjectUnlock(vm);
1145
    qemuDomainEventQueue(driver, event);
1146
    virObjectUnref(cfg);
1147 1148 1149
    return 0;
}

O
Osier Yang 已提交
1150 1151
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1152 1153
                          virDomainObjPtr vm,
                          void *opaque)
O
Osier Yang 已提交
1154
{
1155
    virQEMUDriverPtr driver = opaque;
1156 1157
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1158
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1159

1160
    virObjectLock(vm);
O
Osier Yang 已提交
1161 1162
    event = virDomainEventPMWakeupNewFromObj(vm);

1163 1164 1165 1166 1167 1168 1169 1170 1171
    /* Don't set domain status back to running if it wasn't paused
     * from guest side, otherwise it can just cause confusion.
     */
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PMSUSPENDED) {
        VIR_DEBUG("Transitioned guest %s from pmsuspended to running "
                  "state due to QMP wakeup event", vm->def->name);

        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                             VIR_DOMAIN_RUNNING_WAKEUP);
1172
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
1173 1174 1175
                                                  VIR_DOMAIN_EVENT_STARTED,
                                                  VIR_DOMAIN_EVENT_STARTED_WAKEUP);

1176
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1177 1178 1179 1180 1181
            VIR_WARN("Unable to save status on vm %s after wakeup event",
                     vm->def->name);
        }
    }

1182
    virObjectUnlock(vm);
1183 1184
    qemuDomainEventQueue(driver, event);
    qemuDomainEventQueue(driver, lifecycleEvent);
1185
    virObjectUnref(cfg);
O
Osier Yang 已提交
1186 1187
    return 0;
}
1188

O
Osier Yang 已提交
1189 1190
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1191 1192
                           virDomainObjPtr vm,
                           void *opaque)
O
Osier Yang 已提交
1193
{
1194
    virQEMUDriverPtr driver = opaque;
1195 1196
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1197
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1198

1199
    virObjectLock(vm);
O
Osier Yang 已提交
1200 1201
    event = virDomainEventPMSuspendNewFromObj(vm);

1202
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1203
        qemuDomainObjPrivatePtr priv = vm->privateData;
1204 1205 1206 1207 1208
        VIR_DEBUG("Transitioned guest %s to pmsuspended state due to "
                  "QMP suspend event", vm->def->name);

        virDomainObjSetState(vm, VIR_DOMAIN_PMSUSPENDED,
                             VIR_DOMAIN_PMSUSPENDED_UNKNOWN);
J
Jiri Denemark 已提交
1209
        lifecycleEvent =
1210
            virDomainEventLifecycleNewFromObj(vm,
J
Jiri Denemark 已提交
1211 1212
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1213

1214
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1215 1216 1217
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1218 1219 1220

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1221 1222
    }

1223
    virObjectUnlock(vm);
O
Osier Yang 已提交
1224

1225 1226
    qemuDomainEventQueue(driver, event);
    qemuDomainEventQueue(driver, lifecycleEvent);
1227
    virObjectUnref(cfg);
O
Osier Yang 已提交
1228 1229 1230
    return 0;
}

1231 1232 1233
static int
qemuProcessHandleBalloonChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1234 1235
                               unsigned long long actual,
                               void *opaque)
1236
{
1237
    virQEMUDriverPtr driver = opaque;
1238
    virObjectEventPtr event = NULL;
1239
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1240

1241
    virObjectLock(vm);
1242 1243 1244 1245 1246 1247
    event = virDomainEventBalloonChangeNewFromObj(vm, actual);

    VIR_DEBUG("Updating balloon from %lld to %lld kb",
              vm->def->mem.cur_balloon, actual);
    vm->def->mem.cur_balloon = actual;

1248
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
1249 1250
        VIR_WARN("unable to save domain status with balloon change");

1251
    virObjectUnlock(vm);
1252

1253
    qemuDomainEventQueue(driver, event);
1254
    virObjectUnref(cfg);
1255 1256 1257
    return 0;
}

1258 1259
static int
qemuProcessHandlePMSuspendDisk(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1260 1261
                               virDomainObjPtr vm,
                               void *opaque)
1262
{
1263
    virQEMUDriverPtr driver = opaque;
1264 1265
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1266
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1267

1268
    virObjectLock(vm);
1269 1270 1271 1272 1273 1274 1275 1276 1277 1278
    event = virDomainEventPMSuspendDiskNewFromObj(vm);

    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        qemuDomainObjPrivatePtr priv = vm->privateData;
        VIR_DEBUG("Transitioned guest %s to pmsuspended state due to "
                  "QMP suspend_disk event", vm->def->name);

        virDomainObjSetState(vm, VIR_DOMAIN_PMSUSPENDED,
                             VIR_DOMAIN_PMSUSPENDED_UNKNOWN);
        lifecycleEvent =
1279
            virDomainEventLifecycleNewFromObj(vm,
1280 1281 1282
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_DISK);

1283
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1284 1285 1286 1287 1288 1289 1290 1291
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
    }

1292
    virObjectUnlock(vm);
1293

1294 1295
    qemuDomainEventQueue(driver, event);
    qemuDomainEventQueue(driver, lifecycleEvent);
1296 1297
    virObjectUnref(cfg);

1298 1299 1300
    return 0;
}

1301

1302 1303
static int
qemuProcessHandleGuestPanic(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1304 1305
                            virDomainObjPtr vm,
                            void *opaque)
1306
{
1307
    virQEMUDriverPtr driver = opaque;
1308 1309 1310
    struct qemuProcessEvent *processEvent;

    virObjectLock(vm);
1311
    if (VIR_ALLOC(processEvent) < 0)
1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
        goto cleanup;

    processEvent->eventType = QEMU_PROCESS_EVENT_GUESTPANIC;
    processEvent->action = vm->def->onCrash;
    processEvent->vm = vm;
    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted before handling guest panic event is finished.
     */
    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        if (!virObjectUnref(vm))
            vm = NULL;
        VIR_FREE(processEvent);
    }

1327
 cleanup:
1328
    if (vm)
1329
        virObjectUnlock(vm);
1330 1331 1332 1333 1334

    return 0;
}


1335
int
1336 1337
qemuProcessHandleDeviceDeleted(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1338 1339
                               const char *devAlias,
                               void *opaque)
1340
{
1341
    virQEMUDriverPtr driver = opaque;
1342 1343
    struct qemuProcessEvent *processEvent = NULL;
    char *data;
1344 1345 1346 1347 1348 1349

    virObjectLock(vm);

    VIR_DEBUG("Device %s removed from domain %p %s",
              devAlias, vm, vm->def->name);

1350 1351
    if (qemuDomainSignalDeviceRemoval(vm, devAlias))
        goto cleanup;
1352

1353 1354
    if (VIR_ALLOC(processEvent) < 0)
        goto error;
1355

1356 1357 1358 1359 1360
    processEvent->eventType = QEMU_PROCESS_EVENT_DEVICE_DELETED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
    processEvent->vm = vm;
1361

1362 1363 1364 1365 1366
    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }
1367

1368
 cleanup:
1369 1370
    virObjectUnlock(vm);
    return 0;
1371 1372 1373 1374 1375
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
1376 1377 1378
}


1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
static int
qemuProcessHandleNicRxFilterChanged(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virDomainObjPtr vm,
                                    const char *devAlias,
                                    void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    struct qemuProcessEvent *processEvent = NULL;
    char *data;

    virObjectLock(vm);

    VIR_DEBUG("Device %s RX Filter changed in domain %p %s",
              devAlias, vm, vm->def->name);

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

    processEvent->eventType = QEMU_PROCESS_EVENT_NIC_RX_FILTER_CHANGED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
    processEvent->vm = vm;

    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }

 cleanup:
    virObjectUnlock(vm);
    return 0;
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
}


1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462
static int
qemuProcessHandleSerialChanged(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               const char *devAlias,
                               bool connected,
                               void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    struct qemuProcessEvent *processEvent = NULL;
    char *data;

    virObjectLock(vm);

    VIR_DEBUG("Serial port %s state changed to '%d' in domain %p %s",
              devAlias, connected, vm, vm->def->name);

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

    processEvent->eventType = QEMU_PROCESS_EVENT_SERIAL_CHANGED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
    processEvent->action = connected;
    processEvent->vm = vm;

    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }

 cleanup:
    virObjectUnlock(vm);
    return 0;
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
}


1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481
static int
qemuProcessHandleSpiceMigrated(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               void *opaque ATTRIBUTE_UNUSED)
{
    qemuDomainObjPrivatePtr priv;

    virObjectLock(vm);

    VIR_DEBUG("Spice migration completed for domain %p %s",
              vm, vm->def->name);

    priv = vm->privateData;
    if (priv->job.asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT) {
        VIR_DEBUG("got SPICE_MIGRATE_COMPLETED event without a migration job");
        goto cleanup;
    }

    priv->job.spiceMigrated = true;
1482
    virDomainObjBroadcast(vm);
1483 1484 1485 1486 1487 1488 1489

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
static int
qemuProcessHandleMigrationStatus(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                 virDomainObjPtr vm,
                                 int status,
                                 void *opaque ATTRIBUTE_UNUSED)
{
    qemuDomainObjPrivatePtr priv;

    virObjectLock(vm);

    VIR_DEBUG("Migration of domain %p %s changed state to %s",
              vm, vm->def->name,
              qemuMonitorMigrationStatusTypeToString(status));

    priv = vm->privateData;
1505
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_NONE) {
1506 1507 1508 1509
        VIR_DEBUG("got MIGRATION event without a migration job");
        goto cleanup;
    }

1510
    priv->job.current->stats.status = status;
1511 1512 1513 1514 1515 1516 1517 1518
    virDomainObjBroadcast(vm);

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547
static int
qemuProcessHandleMigrationPass(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               int pass,
                               void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    qemuDomainObjPrivatePtr priv;

    virObjectLock(vm);

    VIR_DEBUG("Migrating domain %p %s, iteration %d",
              vm, vm->def->name, pass);

    priv = vm->privateData;
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_NONE) {
        VIR_DEBUG("got MIGRATION_PASS event without a migration job");
        goto cleanup;
    }

    qemuDomainEventQueue(driver,
                         virDomainEventMigrationIterationNewFromObj(vm, pass));

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1548 1549
static qemuMonitorCallbacks monitorCallbacks = {
    .eofNotify = qemuProcessHandleMonitorEOF,
1550
    .errorNotify = qemuProcessHandleMonitorError,
1551
    .diskSecretLookup = qemuProcessFindVolumeQcowPassphrase,
1552
    .domainEvent = qemuProcessHandleEvent,
1553 1554
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
1555
    .domainResume = qemuProcessHandleResume,
1556 1557 1558 1559 1560
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1561
    .domainBlockJob = qemuProcessHandleBlockJob,
1562
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1563
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1564
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1565
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1566
    .domainPMSuspendDisk = qemuProcessHandlePMSuspendDisk,
1567
    .domainGuestPanic = qemuProcessHandleGuestPanic,
1568
    .domainDeviceDeleted = qemuProcessHandleDeviceDeleted,
1569
    .domainNicRxFilterChanged = qemuProcessHandleNicRxFilterChanged,
1570
    .domainSerialChange = qemuProcessHandleSerialChanged,
1571
    .domainSpiceMigrated = qemuProcessHandleSpiceMigrated,
1572
    .domainMigrationStatus = qemuProcessHandleMigrationStatus,
1573
    .domainMigrationPass = qemuProcessHandleMigrationPass,
1574 1575
};

1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588
static void
qemuProcessMonitorReportLogError(qemuMonitorPtr mon,
                                 const char *msg,
                                 void *opaque);


static void
qemuProcessMonitorLogFree(void *opaque)
{
    qemuDomainLogContextPtr logCtxt = opaque;
    qemuDomainLogContextFree(logCtxt);
}

1589
static int
1590
qemuConnectMonitor(virQEMUDriverPtr driver, virDomainObjPtr vm, int asyncJob,
1591
                   qemuDomainLogContextPtr logCtxt)
1592 1593 1594
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
1595
    qemuMonitorPtr mon = NULL;
1596

1597
    if (virSecurityManagerSetDaemonSocketLabel(driver->securityManager,
1598
                                               vm->def) < 0) {
1599 1600
        VIR_ERROR(_("Failed to set security context for monitor for %s"),
                  vm->def->name);
1601
        return -1;
1602 1603 1604
    }

    /* Hold an extra reference because we can't allow 'vm' to be
M
Michal Privoznik 已提交
1605
     * deleted unitl the monitor gets its own reference. */
1606
    virObjectRef(vm);
1607

1608
    ignore_value(virTimeMillisNow(&priv->monStart));
1609
    virObjectUnlock(vm);
1610 1611 1612 1613

    mon = qemuMonitorOpen(vm,
                          priv->monConfig,
                          priv->monJSON,
1614 1615
                          &monitorCallbacks,
                          driver);
1616

1617 1618 1619 1620 1621 1622 1623
    if (mon && logCtxt) {
        qemuDomainLogContextRef(logCtxt);
        qemuMonitorSetDomainLog(mon,
                                qemuProcessMonitorReportLogError,
                                logCtxt,
                                qemuProcessMonitorLogFree);
    }
1624

1625
    virObjectLock(vm);
M
Michal Privoznik 已提交
1626
    virObjectUnref(vm);
1627
    priv->monStart = 0;
1628

M
Michal Privoznik 已提交
1629
    if (!virDomainObjIsActive(vm)) {
1630
        qemuMonitorClose(mon);
1631
        mon = NULL;
1632 1633 1634
    }
    priv->mon = mon;

1635
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0) {
1636 1637
        VIR_ERROR(_("Failed to clear security context for monitor for %s"),
                  vm->def->name);
1638
        return -1;
1639 1640 1641 1642
    }

    if (priv->mon == NULL) {
        VIR_INFO("Failed to connect monitor for %s", vm->def->name);
1643
        return -1;
1644 1645 1646
    }


1647
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
1648
        return -1;
1649

1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663
    if (qemuMonitorSetCapabilities(priv->mon) < 0)
        goto cleanup;

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON) &&
        virQEMUCapsProbeQMP(priv->qemuCaps, priv->mon) < 0)
        goto cleanup;

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT) &&
        qemuMonitorSetMigrationCapability(priv->mon,
                                          QEMU_MONITOR_MIGRATION_CAPS_EVENTS,
                                          true) < 0) {
        VIR_DEBUG("Cannot enable migration events; clearing capability");
        virQEMUCapsClear(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
    }
1664

1665 1666 1667 1668 1669
    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
1670 1671 1672
    return ret;
}

1673 1674 1675

/**
 * qemuProcessReadLog: Read log file of a qemu VM
1676
 * @logCtxt: the domain log context
1677
 * @msg: pointer to buffer to store the read messages in
1678 1679
 *
 * Reads log of a qemu VM. Skips messages not produced by qemu or irrelevant
1680
 * messages. Returns returns 0 on success or -1 on error
1681
 */
1682
static int
1683
qemuProcessReadLog(qemuDomainLogContextPtr logCtxt, char **msg)
1684
{
1685 1686
    char *buf;
    ssize_t got;
1687
    char *eol;
1688
    char *filter_next;
1689

1690
    if ((got = qemuDomainLogContextRead(logCtxt, &buf)) < 0)
1691
        return -1;
1692

1693 1694 1695 1696 1697 1698 1699
    /* Filter out debug messages from intermediate libvirt process */
    filter_next = buf;
    while ((eol = strchr(filter_next, '\n'))) {
        *eol = '\0';
        if (virLogProbablyLogMessage(filter_next) ||
            STRPREFIX(filter_next, "char device redirected to")) {
            size_t skip = (eol + 1) - filter_next;
1700
            memmove(filter_next, eol + 1, buf + got - eol);
1701 1702 1703 1704
            got -= skip;
        } else {
            filter_next = eol + 1;
            *eol = '\n';
1705 1706
        }
    }
1707
    filter_next = NULL; /* silence false coverity warning */
1708

1709 1710
    if (got > 0 &&
        buf[got - 1] == '\n') {
1711 1712
        buf[got - 1] = '\0';
        got--;
1713
    }
1714
    ignore_value(VIR_REALLOC_N_QUIET(buf, got + 1));
1715 1716 1717
    *msg = buf;
    return 0;
}
1718 1719


1720 1721
static int
qemuProcessReportLogError(qemuDomainLogContextPtr logCtxt,
1722 1723 1724 1725
                          const char *msgprefix)
{
    char *logmsg = NULL;

1726
    if (qemuProcessReadLog(logCtxt, &logmsg) < 0)
1727 1728 1729 1730 1731 1732 1733
        return -1;

    virResetLastError();
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("%s: %s"), msgprefix, logmsg);
    VIR_FREE(logmsg);
    return 0;
1734 1735 1736
}


1737 1738 1739 1740 1741 1742 1743 1744 1745 1746
static void
qemuProcessMonitorReportLogError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                 const char *msg,
                                 void *opaque)
{
    qemuDomainLogContextPtr logCtxt = opaque;
    qemuProcessReportLogError(logCtxt, msg);
}


1747
static int
1748 1749 1750
qemuProcessLookupPTYs(virDomainDefPtr def,
                      virQEMUCapsPtr qemuCaps,
                      virDomainChrDefPtr *devices,
1751
                      int count,
1752
                      virHashTablePtr info)
1753
{
1754
    size_t i;
1755

1756
    for (i = 0; i < count; i++) {
1757
        virDomainChrDefPtr chr = devices[i];
1758 1759
        bool chardevfmt = virQEMUCapsSupportsChardev(def, qemuCaps, chr);

1760
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
C
Cole Robinson 已提交
1761
            char id[32];
1762
            qemuMonitorChardevInfoPtr entry;
1763

C
Cole Robinson 已提交
1764
            if (snprintf(id, sizeof(id), "%s%s",
1765
                         chardevfmt ? "char" : "",
1766 1767 1768 1769
                         chr->info.alias) >= sizeof(id)) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("failed to format device alias "
                                 "for PTY retrieval"));
1770
                return -1;
1771
            }
1772

1773 1774
            entry = virHashLookup(info, id);
            if (!entry || !entry->ptyPath) {
1775 1776 1777 1778
                if (chr->source.data.file.path == NULL) {
                    /* neither the log output nor 'info chardev' had a
                     * pty path for this chardev, report an error
                     */
1779 1780
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1781 1782 1783 1784 1785 1786 1787 1788 1789 1790
                    return -1;
                } else {
                    /* 'info chardev' had no pty path for this chardev,
                     * but the log output had, so we're fine
                     */
                    continue;
                }
            }

            VIR_FREE(chr->source.data.file.path);
1791
            if (VIR_STRDUP(chr->source.data.file.path, entry->ptyPath) < 0)
1792 1793 1794 1795 1796 1797 1798
                return -1;
        }
    }

    return 0;
}

1799 1800
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
1801
                                     virQEMUCapsPtr qemuCaps,
1802
                                     virHashTablePtr info)
1803
{
1804
    size_t i = 0;
C
Cole Robinson 已提交
1805

1806 1807
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->serials, vm->def->nserials,
1808
                              info) < 0)
1809 1810
        return -1;

1811 1812
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->parallels, vm->def->nparallels,
1813
                              info) < 0)
1814
        return -1;
1815

1816 1817
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->channels, vm->def->nchannels,
1818
                              info) < 0)
1819
        return -1;
1820 1821 1822 1823
    /* For historical reasons, console[0] can be just an alias
     * for serial[0]. That's why we need to update it as well. */
    if (vm->def->nconsoles) {
        virDomainChrDefPtr chr = vm->def->consoles[0];
1824

1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835
        if (vm->def->nserials &&
            chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_CONSOLE &&
            chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_SERIAL) {
            /* yes, the first console is just an alias for serials[0] */
            i = 1;
            if (virDomainChrSourceDefCopy(&chr->source,
                                          &((vm->def->serials[0])->source)) < 0)
                return -1;
        }
    }

1836 1837
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->consoles + i, vm->def->nconsoles - i,
1838
                              info) < 0)
1839
        return -1;
1840 1841 1842 1843 1844

    return 0;
}


1845
static int
1846 1847 1848 1849
qemuProcessRefreshChannelVirtioState(virQEMUDriverPtr driver,
                                     virDomainObjPtr vm,
                                     virHashTablePtr info,
                                     int booted)
1850 1851
{
    size_t i;
1852
    int agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_CHANNEL;
1853
    qemuMonitorChardevInfoPtr entry;
1854
    virObjectEventPtr event = NULL;
1855 1856
    char id[32];

1857 1858 1859
    if (booted)
        agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_DOMAIN_STARTED;

1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875
    for (i = 0; i < vm->def->nchannels; i++) {
        virDomainChrDefPtr chr = vm->def->channels[i];
        if (chr->targetType == VIR_DOMAIN_CHR_CHANNEL_TARGET_TYPE_VIRTIO) {
            if (snprintf(id, sizeof(id), "char%s",
                         chr->info.alias) >= sizeof(id)) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("failed to format device alias "
                                 "for PTY retrieval"));
                return -1;
            }

            /* port state not reported */
            if (!(entry = virHashLookup(info, id)) ||
                !entry->state)
                continue;

1876 1877 1878 1879 1880 1881
            if (entry->state != VIR_DOMAIN_CHR_DEVICE_STATE_DEFAULT &&
                STREQ_NULLABLE(chr->target.name, "org.qemu.guest_agent.0") &&
                (event = virDomainEventAgentLifecycleNewFromObj(vm, entry->state,
                                                                agentReason)))
                qemuDomainEventQueue(driver, event);

1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899
            chr->state = entry->state;
        }
    }

    return 0;
}


static int
qemuProcessReconnectRefreshChannelVirtioState(virQEMUDriverPtr driver,
                                              virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virHashTablePtr info = NULL;
    int ret = -1;

    qemuDomainObjEnterMonitor(driver, vm);
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
1900 1901
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
1902 1903 1904 1905

    if (ret < 0)
        goto cleanup;

1906
    ret = qemuProcessRefreshChannelVirtioState(driver, vm, info, false);
1907 1908 1909 1910 1911 1912 1913

 cleanup:
    virHashFree(info);
    return ret;
}


1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
static int
qemuProcessRefreshBalloonState(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
                               int asyncJob)
{
    unsigned long long balloon;
    int rc;

    /* if no ballooning is available, the current size equals to the current
     * full memory size */
    if (!vm->def->memballoon ||
        vm->def->memballoon->model == VIR_DOMAIN_MEMBALLOON_MODEL_NONE) {
        vm->def->mem.cur_balloon = virDomainDefGetMemoryActual(vm->def);
        return 0;
    }

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

    rc = qemuMonitorGetBalloonInfo(qemuDomainGetMonitor(vm), &balloon);
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        rc = -1;

    if (rc < 0)
        return -1;

    vm->def->mem.cur_balloon = balloon;

    return 0;
}


1946
static int
1947
qemuProcessWaitForMonitor(virQEMUDriverPtr driver,
C
Cole Robinson 已提交
1948
                          virDomainObjPtr vm,
1949
                          int asyncJob,
1950
                          virQEMUCapsPtr qemuCaps,
1951
                          qemuDomainLogContextPtr logCtxt)
1952 1953
{
    int ret = -1;
1954
    virHashTablePtr info = NULL;
1955
    qemuDomainObjPrivatePtr priv;
1956 1957

    VIR_DEBUG("Connect monitor to %p '%s'", vm, vm->def->name);
1958
    if (qemuConnectMonitor(driver, vm, asyncJob, logCtxt) < 0)
1959 1960 1961 1962 1963 1964
        goto cleanup;

    /* Try to get the pty path mappings again via the monitor. This is much more
     * reliable if it's available.
     * Note that the monitor itself can be on a pty, so we still need to try the
     * log output method. */
1965
    priv = vm->privateData;
1966 1967
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
1968 1969
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
    VIR_DEBUG("qemuMonitorGetChardevInfo returned %i", ret);
1970 1971 1972
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

1973 1974 1975 1976 1977
    if (ret == 0) {
        if ((ret = qemuProcessFindCharDevicePTYsMonitor(vm, qemuCaps,
                                                        info)) < 0)
            goto cleanup;

1978 1979
        if ((ret = qemuProcessRefreshChannelVirtioState(driver, vm, info,
                                                        true)) < 0)
1980 1981
            goto cleanup;
    }
1982

1983
 cleanup:
1984
    virHashFree(info);
1985

1986 1987
    if (logCtxt && kill(vm->pid, 0) == -1 && errno == ESRCH) {
        qemuProcessReportLogError(logCtxt,
1988
                                  _("process exited while connecting to monitor"));
1989 1990 1991 1992 1993 1994
        ret = -1;
    }

    return ret;
}

1995

1996 1997 1998 1999 2000 2001
static int
qemuProcessDetectIOThreadPIDs(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2002
    qemuMonitorIOThreadInfoPtr *iothreads = NULL;
2003 2004 2005 2006
    int niothreads = 0;
    int ret = -1;
    size_t i;

2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_OBJECT_IOTHREAD)) {
        /* The following check is because at one time a domain could
         * define iothreadids and start the domain - only failing the
         * capability check when attempting to add a disk. Because the
         * iothreads and [n]iothreadids were left untouched other code
         * assumed it could use the ->thread_id value to make thread_id
         * based adjustments (e.g. pinning, scheduling) which while
         * succeeding would execute on the calling thread.
         */
        if (vm->def->niothreadids) {
            for (i = 0; i < vm->def->niothreadids; i++) {
                /* Check if the domain had defined any iothreadid elements
                 * and supply a VIR_INFO indicating that it's being removed.
                 */
                if (!vm->def->iothreadids[i]->autofill)
                    VIR_INFO("IOThreads not supported, remove iothread id '%u'",
                             vm->def->iothreadids[i]->iothread_id);
                virDomainIOThreadIDDefFree(vm->def->iothreadids[i]);
            }
            /* Remove any trace */
            VIR_FREE(vm->def->iothreadids);
            vm->def->niothreadids = 0;
            vm->def->iothreads = 0;
        }
2031
        return 0;
2032
    }
2033

2034 2035 2036 2037
    /* Get the list of IOThreads from qemu */
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
    niothreads = qemuMonitorGetIOThreads(priv->mon, &iothreads);
2038 2039
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
J
John Ferlan 已提交
2040
    if (niothreads < 0)
2041 2042
        goto cleanup;

2043
    if (niothreads != vm->def->niothreadids) {
2044 2045
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of IOThread pids from QEMU monitor. "
2046 2047
                         "got %d, wanted %zu"),
                       niothreads, vm->def->niothreadids);
2048 2049 2050
        goto cleanup;
    }

2051 2052 2053 2054 2055 2056
    /* Nothing to do */
    if (niothreads == 0) {
        ret = 0;
        goto cleanup;
    }

2057 2058 2059
    for (i = 0; i < niothreads; i++) {
        virDomainIOThreadIDDefPtr iothrid;

2060 2061
        if (!(iothrid = virDomainIOThreadIDFind(vm->def,
                                                iothreads[i]->iothread_id))) {
2062
            virReportError(VIR_ERR_INTERNAL_ERROR,
2063 2064
                           _("iothread %d not found"),
                           iothreads[i]->iothread_id);
2065 2066 2067 2068
            goto cleanup;
        }
        iothrid->thread_id = iothreads[i]->thread_id;
    }
2069 2070 2071 2072 2073 2074

    ret = 0;

 cleanup:
    if (iothreads) {
        for (i = 0; i < niothreads; i++)
2075
            VIR_FREE(iothreads[i]);
2076 2077 2078 2079 2080
        VIR_FREE(iothreads);
    }
    return ret;
}

2081 2082 2083 2084 2085

/*
 * To be run between fork/exec of QEMU only
 */
static int
2086
qemuProcessInitCpuAffinity(virDomainObjPtr vm)
2087 2088 2089 2090
{
    int ret = -1;
    virBitmapPtr cpumap = NULL;
    virBitmapPtr cpumapToSet = NULL;
2091
    qemuDomainObjPrivatePtr priv = vm->privateData;
2092

2093 2094 2095 2096 2097 2098
    if (!vm->pid) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Cannot setup CPU affinity until process is started"));
        return -1;
    }

2099 2100
    if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) {
        VIR_DEBUG("Set CPU affinity with advisory nodeset from numad");
2101
        cpumapToSet = priv->autoCpuset;
2102
    } else {
2103
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
2104
        if (vm->def->cpumask) {
H
Hu Tao 已提交
2105
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
2106 2107 2108 2109 2110
        } else {
            /* You may think this is redundant, but we can't assume libvirtd
             * itself is running on all pCPUs, so we need to explicitly set
             * the spawned QEMU instance to all pCPUs if no map is given in
             * its config file */
2111 2112 2113 2114
            int hostcpus;

            /* setaffinity fails if you set bits for CPUs which
             * aren't present, so we have to limit ourselves */
2115
            if ((hostcpus = nodeGetCPUCount(NULL)) < 0)
2116 2117 2118 2119 2120 2121 2122 2123
                goto cleanup;

            if (hostcpus > QEMUD_CPUMASK_LEN)
                hostcpus = QEMUD_CPUMASK_LEN;

            if (!(cpumap = virBitmapNew(hostcpus)))
                goto cleanup;

2124
            virBitmapSetAll(cpumap);
2125 2126

            cpumapToSet = cpumap;
O
Osier Yang 已提交
2127
        }
2128 2129
    }

2130
    if (virProcessSetAffinity(vm->pid, cpumapToSet) < 0)
2131
        goto cleanup;
2132

2133 2134
    ret = 0;

2135
 cleanup:
2136
    virBitmapFree(cpumap);
2137
    return ret;
2138 2139
}

2140 2141
/* set link states to down on interfaces at qemu start */
static int
2142 2143 2144
qemuProcessSetLinkStates(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob)
2145 2146 2147
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2148
    size_t i;
2149 2150 2151 2152 2153
    int ret = -1;
    int rv;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2154 2155 2156

    for (i = 0; i < def->nnets; i++) {
        if (def->nets[i]->linkstate == VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN) {
2157 2158 2159
            if (!def->nets[i]->info.alias) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("missing alias for network device"));
2160
                goto cleanup;
2161 2162
            }

2163 2164
            VIR_DEBUG("Setting link state: %s", def->nets[i]->info.alias);

2165
            if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NETDEV)) {
2166
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
2167
                               _("Setting of link state is not supported by this qemu"));
2168
                goto cleanup;
2169 2170
            }

2171 2172 2173 2174
            rv = qemuMonitorSetLink(priv->mon,
                                    def->nets[i]->info.alias,
                                    VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (rv < 0) {
2175
                virReportError(VIR_ERR_OPERATION_FAILED,
2176 2177 2178
                               _("Couldn't set link state on interface: %s"),
                               def->nets[i]->info.alias);
                goto cleanup;
2179 2180 2181 2182
            }
        }
    }

2183 2184 2185 2186 2187
    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2188 2189 2190 2191
    return ret;
}

/* Set CPU affinities for vcpus if vcpupin xml provided. */
2192
static int
2193
qemuProcessSetVcpuAffinities(virDomainObjPtr vm)
2194 2195
{
    virDomainDefPtr def = vm->def;
2196
    virDomainPinDefPtr pininfo;
2197
    int n;
E
Eric Blake 已提交
2198
    int ret = -1;
2199 2200 2201
    VIR_DEBUG("Setting affinity on CPUs nvcpupin=%zu nvcpus=%d hasVcpupids=%d",
              def->cputune.nvcpupin, virDomainDefGetVcpus(def),
              qemuDomainHasVcpuPids(vm));
2202 2203 2204
    if (!def->cputune.nvcpupin)
        return 0;

2205
    if (!qemuDomainHasVcpuPids(vm)) {
2206 2207 2208
        /* If any CPU has custom affinity that differs from the
         * VM default affinity, we must reject it
         */
2209
        for (n = 0; n < def->cputune.nvcpupin; n++) {
2210 2211 2212 2213 2214 2215 2216 2217
            if (!virBitmapEqual(def->cpumask,
                                def->cputune.vcpupin[n]->cpumask)) {
                virReportError(VIR_ERR_OPERATION_INVALID,
                               "%s", _("cpu affinity is not supported"));
                return -1;
            }
        }
        return 0;
2218 2219
    }

2220
    for (n = 0; n < virDomainDefGetVcpus(def); n++) {
2221
        /* set affinity only for existing vcpus */
2222 2223 2224
        if (!(pininfo = virDomainPinFind(def->cputune.vcpupin,
                                         def->cputune.nvcpupin,
                                         n)))
2225
            continue;
2226

2227
        if (virProcessSetAffinity(qemuDomainGetVcpuPid(vm, n),
2228
                                  pininfo->cpumask) < 0) {
E
Eric Blake 已提交
2229
            goto cleanup;
2230 2231 2232
        }
    }

E
Eric Blake 已提交
2233
    ret = 0;
2234
 cleanup:
E
Eric Blake 已提交
2235
    return ret;
2236
}
2237

2238
/* Set CPU affinities for emulator threads. */
2239
static int
2240
qemuProcessSetEmulatorAffinity(virDomainObjPtr vm)
2241
{
2242
    virBitmapPtr cpumask;
2243 2244 2245
    virDomainDefPtr def = vm->def;
    int ret = -1;

O
Osier Yang 已提交
2246
    if (def->cputune.emulatorpin)
2247
        cpumask = def->cputune.emulatorpin;
O
Osier Yang 已提交
2248
    else if (def->cpumask)
2249
        cpumask = def->cpumask;
O
Osier Yang 已提交
2250 2251
    else
        return 0;
2252

2253
    ret = virProcessSetAffinity(vm->pid, cpumask);
2254 2255 2256
    return ret;
}

2257 2258 2259 2260 2261 2262 2263 2264
/* Set CPU affinities for IOThreads threads. */
static int
qemuProcessSetIOThreadsAffinity(virDomainObjPtr vm)
{
    virDomainDefPtr def = vm->def;
    size_t i;
    int ret = -1;

2265 2266
    for (i = 0; i < def->niothreadids; i++) {
        /* set affinity only for existing iothreads */
2267
        if (!def->iothreadids[i]->cpumask)
2268 2269
            continue;

2270
        if (virProcessSetAffinity(def->iothreadids[i]->thread_id,
2271
                                  def->iothreadids[i]->cpumask) < 0)
2272 2273 2274 2275 2276 2277 2278 2279
            goto cleanup;
    }
    ret = 0;

 cleanup:
    return ret;
}

2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304
/* Set Scheduler parameters for vCPU or I/O threads. */
int
qemuProcessSetSchedParams(int id,
                          pid_t pid,
                          size_t nsp,
                          virDomainThreadSchedParamPtr sp)
{
    bool val = false;
    size_t i = 0;
    virDomainThreadSchedParamPtr s = NULL;

    for (i = 0; i < nsp; i++) {
        if (virBitmapGetBit(sp[i].ids, id, &val) < 0) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Cannot get bit from bitmap"));
        }
        if (val) {
            s = &sp[i];
            break;
        }
    }

    if (!s)
        return 0;

2305
    return virProcessSetScheduler(pid, s->policy, s->priority);
2306 2307 2308 2309 2310 2311 2312
}

static int
qemuProcessSetSchedulers(virDomainObjPtr vm)
{
    size_t i = 0;

2313 2314 2315 2316 2317 2318
    for (i = 0; i < virDomainDefGetVcpusMax(vm->def); i++) {
        virDomainVcpuInfoPtr vcpu = virDomainDefGetVcpu(vm->def, i);

        if (!vcpu->online)
            continue;

2319
        if (qemuProcessSetSchedParams(i, qemuDomainGetVcpuPid(vm, i),
2320 2321 2322 2323 2324
                                      vm->def->cputune.nvcpusched,
                                      vm->def->cputune.vcpusched) < 0)
            return -1;
    }

2325 2326 2327
    for (i = 0; i < vm->def->niothreadids; i++) {
        if (qemuProcessSetSchedParams(vm->def->iothreadids[i]->iothread_id,
                                      vm->def->iothreadids[i]->thread_id,
2328 2329 2330 2331 2332 2333 2334 2335
                                      vm->def->cputune.niothreadsched,
                                      vm->def->cputune.iothreadsched) < 0)
            return -1;
    }

    return 0;
}

2336 2337
static int
qemuProcessInitPasswords(virConnectPtr conn,
2338
                         virQEMUDriverPtr driver,
2339 2340
                         virDomainObjPtr vm,
                         int asyncJob)
2341 2342 2343
{
    int ret = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
2344
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2345
    size_t i;
2346 2347
    char *alias = NULL;
    char *secret = NULL;
2348

2349
    for (i = 0; i < vm->def->ngraphics; ++i) {
2350 2351
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
2352 2353
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_VNC,
2354
                                                    &graphics->data.vnc.auth,
2355 2356
                                                    cfg->vncPassword,
                                                    asyncJob);
2357
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
2358 2359
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_SPICE,
2360
                                                    &graphics->data.spice.auth,
2361 2362
                                                    cfg->spicePassword,
                                                    asyncJob);
2363 2364
        }

2365 2366 2367
        if (ret < 0)
            goto cleanup;
    }
2368

2369
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
2370
        for (i = 0; i < vm->def->ndisks; i++) {
2371 2372
            size_t secretLen;

2373
            if (!vm->def->disks[i]->src->encryption ||
2374
                !virDomainDiskGetSource(vm->def->disks[i]))
2375 2376
                continue;

2377
            VIR_FREE(secret);
2378 2379 2380 2381 2382
            if (qemuProcessGetVolumeQcowPassphrase(conn,
                                                   vm->def->disks[i],
                                                   &secret, &secretLen) < 0)
                goto cleanup;

2383 2384 2385 2386
            VIR_FREE(alias);
            if (VIR_STRDUP(alias, vm->def->disks[i]->info.alias) < 0)
                goto cleanup;
            if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
2387
                goto cleanup;
E
Eric Blake 已提交
2388
            ret = qemuMonitorSetDrivePassphrase(priv->mon, alias, secret);
2389 2390
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                ret = -1;
2391 2392 2393 2394 2395
            if (ret < 0)
                goto cleanup;
        }
    }

2396
 cleanup:
2397 2398
    VIR_FREE(alias);
    VIR_FREE(secret);
2399
    virObjectUnref(cfg);
2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442
    return ret;
}


#define QEMU_PCI_VENDOR_INTEL     0x8086
#define QEMU_PCI_VENDOR_LSI_LOGIC 0x1000
#define QEMU_PCI_VENDOR_REDHAT    0x1af4
#define QEMU_PCI_VENDOR_CIRRUS    0x1013
#define QEMU_PCI_VENDOR_REALTEK   0x10ec
#define QEMU_PCI_VENDOR_AMD       0x1022
#define QEMU_PCI_VENDOR_ENSONIQ   0x1274
#define QEMU_PCI_VENDOR_VMWARE    0x15ad
#define QEMU_PCI_VENDOR_QEMU      0x1234

#define QEMU_PCI_PRODUCT_DISK_VIRTIO 0x1001

#define QEMU_PCI_PRODUCT_BALLOON_VIRTIO 0x1002

#define QEMU_PCI_PRODUCT_NIC_NE2K     0x8029
#define QEMU_PCI_PRODUCT_NIC_PCNET    0x2000
#define QEMU_PCI_PRODUCT_NIC_RTL8139  0x8139
#define QEMU_PCI_PRODUCT_NIC_E1000    0x100E
#define QEMU_PCI_PRODUCT_NIC_VIRTIO   0x1000

#define QEMU_PCI_PRODUCT_VGA_CIRRUS 0x00b8
#define QEMU_PCI_PRODUCT_VGA_VMWARE 0x0405
#define QEMU_PCI_PRODUCT_VGA_STDVGA 0x1111

#define QEMU_PCI_PRODUCT_AUDIO_AC97    0x2415
#define QEMU_PCI_PRODUCT_AUDIO_ES1370  0x5000

#define QEMU_PCI_PRODUCT_CONTROLLER_PIIX 0x7010
#define QEMU_PCI_PRODUCT_CONTROLLER_LSI  0x0012

#define QEMU_PCI_PRODUCT_WATCHDOG_I63000ESB 0x25ab

static int
qemuProcessAssignNextPCIAddress(virDomainDeviceInfo *info,
                                int vendor,
                                int product,
                                qemuMonitorPCIAddress *addrs,
                                int naddrs)
{
2443
    bool found = false;
2444
    size_t i;
2445 2446 2447

    VIR_DEBUG("Look for %x:%x out of %d", vendor, product, naddrs);

2448
    for (i = 0; i < naddrs; i++) {
2449 2450 2451
        VIR_DEBUG("Maybe %x:%x", addrs[i].vendor, addrs[i].product);
        if (addrs[i].vendor == vendor &&
            addrs[i].product == product) {
2452
            VIR_DEBUG("Match %zu", i);
2453
            found = true;
2454 2455 2456
            break;
        }
    }
2457
    if (!found)
2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657
        return -1;

    /* Blank it out so this device isn't matched again */
    addrs[i].vendor = 0;
    addrs[i].product = 0;

    if (info->type == VIR_DOMAIN_DEVICE_ADDRESS_TYPE_NONE)
        info->type = VIR_DOMAIN_DEVICE_ADDRESS_TYPE_PCI;

    if (info->type == VIR_DOMAIN_DEVICE_ADDRESS_TYPE_PCI) {
        info->addr.pci.domain = addrs[i].addr.domain;
        info->addr.pci.bus = addrs[i].addr.bus;
        info->addr.pci.slot = addrs[i].addr.slot;
        info->addr.pci.function = addrs[i].addr.function;
    }

    return 0;
}

static int
qemuProcessGetPCIDiskVendorProduct(virDomainDiskDefPtr def,
                                   unsigned *vendor,
                                   unsigned *product)
{
    switch (def->bus) {
    case VIR_DOMAIN_DISK_BUS_VIRTIO:
        *vendor = QEMU_PCI_VENDOR_REDHAT;
        *product = QEMU_PCI_PRODUCT_DISK_VIRTIO;
        break;

    default:
        return -1;
    }

    return 0;
}

static int
qemuProcessGetPCINetVendorProduct(virDomainNetDefPtr def,
                                  unsigned *vendor,
                                  unsigned *product)
{
    if (!def->model)
        return -1;

    if (STREQ(def->model, "ne2k_pci")) {
        *vendor = QEMU_PCI_VENDOR_REALTEK;
        *product = QEMU_PCI_PRODUCT_NIC_NE2K;
    } else if (STREQ(def->model, "pcnet")) {
        *vendor = QEMU_PCI_VENDOR_AMD;
        *product = QEMU_PCI_PRODUCT_NIC_PCNET;
    } else if (STREQ(def->model, "rtl8139")) {
        *vendor = QEMU_PCI_VENDOR_REALTEK;
        *product = QEMU_PCI_PRODUCT_NIC_RTL8139;
    } else if (STREQ(def->model, "e1000")) {
        *vendor = QEMU_PCI_VENDOR_INTEL;
        *product = QEMU_PCI_PRODUCT_NIC_E1000;
    } else if (STREQ(def->model, "virtio")) {
        *vendor = QEMU_PCI_VENDOR_REDHAT;
        *product = QEMU_PCI_PRODUCT_NIC_VIRTIO;
    } else {
        VIR_INFO("Unexpected NIC model %s, cannot get PCI address",
                 def->model);
        return -1;
    }
    return 0;
}

static int
qemuProcessGetPCIControllerVendorProduct(virDomainControllerDefPtr def,
                                         unsigned *vendor,
                                         unsigned *product)
{
    switch (def->type) {
    case VIR_DOMAIN_CONTROLLER_TYPE_SCSI:
        *vendor = QEMU_PCI_VENDOR_LSI_LOGIC;
        *product = QEMU_PCI_PRODUCT_CONTROLLER_LSI;
        break;

    case VIR_DOMAIN_CONTROLLER_TYPE_FDC:
        /* XXX we could put in the ISA bridge address, but
           that's not technically the FDC's address */
        return -1;

    case VIR_DOMAIN_CONTROLLER_TYPE_IDE:
        *vendor = QEMU_PCI_VENDOR_INTEL;
        *product = QEMU_PCI_PRODUCT_CONTROLLER_PIIX;
        break;

    default:
        VIR_INFO("Unexpected controller type %s, cannot get PCI address",
                 virDomainControllerTypeToString(def->type));
        return -1;
    }

    return 0;
}

static int
qemuProcessGetPCIVideoVendorProduct(virDomainVideoDefPtr def,
                                    unsigned *vendor,
                                    unsigned *product)
{
    switch (def->type) {
    case VIR_DOMAIN_VIDEO_TYPE_CIRRUS:
        *vendor = QEMU_PCI_VENDOR_CIRRUS;
        *product = QEMU_PCI_PRODUCT_VGA_CIRRUS;
        break;

    case VIR_DOMAIN_VIDEO_TYPE_VGA:
        *vendor = QEMU_PCI_VENDOR_QEMU;
        *product = QEMU_PCI_PRODUCT_VGA_STDVGA;
        break;

    case VIR_DOMAIN_VIDEO_TYPE_VMVGA:
        *vendor = QEMU_PCI_VENDOR_VMWARE;
        *product = QEMU_PCI_PRODUCT_VGA_VMWARE;
        break;

    default:
        return -1;
    }
    return 0;
}

static int
qemuProcessGetPCISoundVendorProduct(virDomainSoundDefPtr def,
                                    unsigned *vendor,
                                    unsigned *product)
{
    switch (def->model) {
    case VIR_DOMAIN_SOUND_MODEL_ES1370:
        *vendor = QEMU_PCI_VENDOR_ENSONIQ;
        *product = QEMU_PCI_PRODUCT_AUDIO_ES1370;
        break;

    case VIR_DOMAIN_SOUND_MODEL_AC97:
        *vendor = QEMU_PCI_VENDOR_INTEL;
        *product = QEMU_PCI_PRODUCT_AUDIO_AC97;
        break;

    default:
        return -1;
    }

    return 0;
}

static int
qemuProcessGetPCIWatchdogVendorProduct(virDomainWatchdogDefPtr def,
                                       unsigned *vendor,
                                       unsigned *product)
{
    switch (def->model) {
    case VIR_DOMAIN_WATCHDOG_MODEL_I6300ESB:
        *vendor = QEMU_PCI_VENDOR_INTEL;
        *product = QEMU_PCI_PRODUCT_WATCHDOG_I63000ESB;
        break;

    default:
        return -1;
    }

    return 0;
}


static int
qemuProcessGetPCIMemballoonVendorProduct(virDomainMemballoonDefPtr def,
                                         unsigned *vendor,
                                         unsigned *product)
{
    switch (def->model) {
    case VIR_DOMAIN_MEMBALLOON_MODEL_VIRTIO:
        *vendor = QEMU_PCI_VENDOR_REDHAT;
        *product = QEMU_PCI_PRODUCT_BALLOON_VIRTIO;
        break;

    default:
        return -1;
    }

    return 0;
}


/*
 * This entire method assumes that PCI devices in 'info pci'
 * match ordering of devices specified on the command line
 * wrt to devices of matching vendor+product
 *
 * XXXX this might not be a valid assumption if we assign
 * some static addrs on CLI. Have to check that...
 */
static int
qemuProcessDetectPCIAddresses(virDomainObjPtr vm,
                              qemuMonitorPCIAddress *addrs,
                              int naddrs)
{
    unsigned int vendor = 0, product = 0;
2658
    size_t i;
2659 2660 2661 2662 2663

    /* XXX should all these vendor/product IDs be kept in the
     * actual device data structure instead ?
     */

2664
    for (i = 0; i < vm->def->ndisks; i++) {
2665 2666 2667 2668 2669 2670
        if (qemuProcessGetPCIDiskVendorProduct(vm->def->disks[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->disks[i]->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2671 2672 2673
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for VirtIO disk %s"),
                           vm->def->disks[i]->dst);
2674 2675 2676 2677
            return -1;
        }
    }

2678
    for (i = 0; i < vm->def->nnets; i++) {
2679 2680 2681 2682 2683 2684
        if (qemuProcessGetPCINetVendorProduct(vm->def->nets[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->nets[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2685 2686 2687
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for %s NIC"),
                           vm->def->nets[i]->model);
2688 2689 2690 2691
            return -1;
        }
    }

2692
    for (i = 0; i < vm->def->ncontrollers; i++) {
2693 2694 2695 2696 2697 2698
        if (qemuProcessGetPCIControllerVendorProduct(vm->def->controllers[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->controllers[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2699 2700 2701
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for controller %s"),
                           virDomainControllerTypeToString(vm->def->controllers[i]->type));
2702 2703 2704 2705
            return -1;
        }
    }

2706
    for (i = 0; i < vm->def->nvideos; i++) {
2707 2708 2709 2710 2711 2712
        if (qemuProcessGetPCIVideoVendorProduct(vm->def->videos[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->videos[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2713 2714 2715
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for video adapter %s"),
                           virDomainVideoTypeToString(vm->def->videos[i]->type));
2716 2717 2718 2719
            return -1;
        }
    }

2720
    for (i = 0; i < vm->def->nsounds; i++) {
2721 2722 2723 2724 2725 2726
        if (qemuProcessGetPCISoundVendorProduct(vm->def->sounds[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->sounds[i]->info),
                                    vendor, product,
                                     addrs,  naddrs) < 0) {
2727 2728 2729
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for sound adapter %s"),
                           virDomainSoundModelTypeToString(vm->def->sounds[i]->model));
2730 2731 2732 2733 2734 2735 2736 2737 2738 2739
            return -1;
        }
    }


    if (vm->def->watchdog &&
        qemuProcessGetPCIWatchdogVendorProduct(vm->def->watchdog, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->watchdog->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2740 2741 2742
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for watchdog %s"),
                           virDomainWatchdogModelTypeToString(vm->def->watchdog->model));
2743 2744 2745 2746 2747 2748 2749 2750 2751
            return -1;
        }
    }

    if (vm->def->memballoon &&
        qemuProcessGetPCIMemballoonVendorProduct(vm->def->memballoon, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->memballoon->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2752 2753 2754
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for balloon %s"),
                           virDomainMemballoonModelTypeToString(vm->def->memballoon->model));
2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771
            return -1;
        }
    }

    /* XXX console (virtio) */


    /* ... and now things we don't have in our xml */

    /* XXX USB controller ? */

    /* XXX what about other PCI devices (ie bridges) */

    return 0;
}

static int
2772
qemuProcessInitPCIAddresses(virQEMUDriverPtr driver,
2773 2774
                            virDomainObjPtr vm,
                            int asyncJob)
2775 2776 2777
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int naddrs;
2778
    int ret = -1;
2779 2780
    qemuMonitorPCIAddress *addrs = NULL;

2781 2782
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2783 2784
    naddrs = qemuMonitorGetAllPCIAddresses(priv->mon,
                                           &addrs);
2785 2786
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
2787

2788 2789
    if (naddrs > 0)
        ret = qemuProcessDetectPCIAddresses(vm, addrs, naddrs);
2790

2791
 cleanup:
2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820
    VIR_FREE(addrs);

    return ret;
}


static int
qemuProcessPrepareChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
    int fd;
    if (dev->source.type != VIR_DOMAIN_CHR_TYPE_FILE)
        return 0;

    if ((fd = open(dev->source.data.file.path,
                   O_CREAT | O_APPEND, S_IRUSR|S_IWUSR)) < 0) {
        virReportSystemError(errno,
                             _("Unable to pre-create chardev file '%s'"),
                             dev->source.data.file.path);
        return -1;
    }

    VIR_FORCE_CLOSE(fd);

    return 0;
}


2821 2822 2823 2824 2825 2826
static int
qemuProcessCleanupChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
    if (dev->source.type == VIR_DOMAIN_CHR_TYPE_UNIX &&
2827 2828
        dev->source.data.nix.listen &&
        dev->source.data.nix.path)
2829 2830 2831 2832 2833 2834
        unlink(dev->source.data.nix.path);

    return 0;
}


2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899
/**
 * Loads and update video memory size for video devices according to QEMU
 * process as the QEMU will silently update the values that we pass to QEMU
 * through command line.  We need to load these updated values and store them
 * into the status XML.
 *
 * We will fail if for some reason the values cannot be loaded from QEMU because
 * its mandatory to get the correct video memory size to status XML to not break
 * migration.
 */
static int
qemuProcessUpdateVideoRamSize(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    int ret = -1;
    ssize_t i;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainVideoDefPtr video = NULL;
    virQEMUDriverConfigPtr cfg = NULL;

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

    for (i = 0; i < vm->def->nvideos; i++) {
        video = vm->def->videos[i];

        switch (video->type) {
        case VIR_DOMAIN_VIDEO_TYPE_VGA:
            if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_VGA_VGAMEM)) {
                if (qemuMonitorUpdateVideoMemorySize(priv->mon, video, "VGA") < 0)
                    goto error;
            }
            break;
        case VIR_DOMAIN_VIDEO_TYPE_QXL:
            if (i == 0) {
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VGA_VGAMEM)) {
                    if (qemuMonitorUpdateVideoMemorySize(priv->mon, video,
                                                         "qxl-vga") < 0)
                        goto error;
                }
            } else {
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VGAMEM)) {
                    if (qemuMonitorUpdateVideoMemorySize(priv->mon, video,
                                                         "qxl") < 0)
                        goto error;
                }
            }
            break;
        case VIR_DOMAIN_VIDEO_TYPE_VMVGA:
            if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_VMWARE_SVGA_VGAMEM)) {
                if (qemuMonitorUpdateVideoMemorySize(priv->mon, video,
                                                     "vmware-svga") < 0)
                    goto error;
            }
            break;
        case VIR_DOMAIN_VIDEO_TYPE_CIRRUS:
        case VIR_DOMAIN_VIDEO_TYPE_XEN:
        case VIR_DOMAIN_VIDEO_TYPE_VBOX:
        case VIR_DOMAIN_VIDEO_TYPE_LAST:
            break;
        }

    }

2900 2901
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2902 2903 2904 2905 2906 2907 2908 2909

    cfg = virQEMUDriverGetConfig(driver);
    ret = virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm);
    virObjectUnref(cfg);

    return ret;

 error:
2910
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
2911 2912 2913 2914
    return -1;
}


2915 2916 2917
struct qemuProcessHookData {
    virConnectPtr conn;
    virDomainObjPtr vm;
2918
    virQEMUDriverPtr driver;
2919
    virQEMUDriverConfigPtr cfg;
2920 2921 2922 2923 2924
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
2925
    qemuDomainObjPrivatePtr priv = h->vm->privateData;
2926
    int ret = -1;
2927
    int fd;
2928 2929 2930
    virBitmapPtr nodeset = NULL;
    virDomainNumatuneMemMode mode;

2931 2932 2933 2934 2935
    /* This method cannot use any mutexes, which are not
     * protected across fork()
     */

    virSecurityManagerPostFork(h->driver->securityManager);
2936 2937 2938 2939 2940

    /* Some later calls want pid present */
    h->vm->pid = getpid();

    VIR_DEBUG("Obtaining domain lock");
2941 2942 2943 2944 2945 2946 2947
    /*
     * Since we're going to leak the returned FD to QEMU,
     * we need to make sure it gets a sensible label.
     * This mildly sucks, because there could be other
     * sockets the lock driver opens that we don't want
     * labelled. So far we're ok though.
     */
2948
    if (virSecurityManagerSetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2949
        goto cleanup;
2950
    if (virDomainLockProcessStart(h->driver->lockManager,
2951
                                  h->cfg->uri,
2952
                                  h->vm,
J
Ján Tomko 已提交
2953
                                  /* QEMU is always paused initially */
2954 2955
                                  true,
                                  &fd) < 0)
2956
        goto cleanup;
2957
    if (virSecurityManagerClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2958
        goto cleanup;
2959

2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970
    if (virDomainNumatuneGetMode(h->vm->def->numa, -1, &mode) == 0) {
        if (mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT &&
            h->cfg->cgroupControllers & (1 << VIR_CGROUP_CONTROLLER_CPUSET) &&
            virCgroupControllerAvailable(VIR_CGROUP_CONTROLLER_CPUSET)) {
            /* Use virNuma* API iff necessary. Once set and child is exec()-ed,
             * there's no way for us to change it. Rely on cgroups (if available
             * and enabled in the config) rather than virNuma*. */
            VIR_DEBUG("Relying on CGroups for memory binding");
        } else {
            nodeset = virDomainNumatuneGetNodeset(h->vm->def->numa,
                                                  priv->autoNodeset, -1);
2971

2972 2973 2974
            if (virNumaSetupMemoryPolicy(mode, nodeset) < 0)
                goto cleanup;
        }
2975
    }
2976

2977 2978
    ret = 0;

2979
 cleanup:
2980
    virObjectUnref(h->cfg);
2981 2982
    VIR_DEBUG("Hook complete ret=%d", ret);
    return ret;
2983 2984 2985
}

int
2986
qemuProcessPrepareMonitorChr(virQEMUDriverConfigPtr cfg,
2987 2988 2989 2990 2991 2992
                             virDomainChrSourceDefPtr monConfig,
                             const char *vm)
{
    monConfig->type = VIR_DOMAIN_CHR_TYPE_UNIX;
    monConfig->data.nix.listen = true;

2993
    if (virAsprintf(&monConfig->data.nix.path, "%s/domain-%s/monitor.sock",
2994 2995 2996
                    cfg->libDir, vm) < 0)
        return -1;
    return 0;
2997 2998 2999
}


3000
/*
3001 3002
 * Precondition: vm must be locked, and a job must be active.
 * This method will call {Enter,Exit}Monitor
3003
 */
E
Eric Blake 已提交
3004
int
3005
qemuProcessStartCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
3006
                     virConnectPtr conn, virDomainRunningReason reason,
3007
                     qemuDomainAsyncJob asyncJob)
3008
{
3009
    int ret = -1;
3010
    qemuDomainObjPrivatePtr priv = vm->privateData;
3011
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3012

3013
    /* Bring up netdevs before starting CPUs */
3014
    if (qemuInterfaceStartDevices(vm->def) < 0)
3015 3016
       goto cleanup;

3017
    VIR_DEBUG("Using lock state '%s'", NULLSTR(priv->lockState));
3018
    if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
3019
                                   vm, priv->lockState) < 0) {
3020 3021 3022 3023
        /* Don't free priv->lockState on error, because we need
         * to make sure we have state still present if the user
         * tries to resume again
         */
3024
        goto cleanup;
3025 3026 3027
    }
    VIR_FREE(priv->lockState);

3028 3029
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto release;
J
Jiri Denemark 已提交
3030

3031
    ret = qemuMonitorStartCPUs(priv->mon, conn);
3032 3033
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3034 3035 3036 3037 3038

    if (ret < 0)
        goto release;

    virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
3039

3040
 cleanup:
3041
    virObjectUnref(cfg);
3042
    return ret;
3043 3044 3045 3046 3047 3048

 release:
    if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
        VIR_WARN("Unable to release lease on %s", vm->def->name);
    VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));
    goto cleanup;
3049 3050 3051
}


3052 3053
int qemuProcessStopCPUs(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
3054
                        virDomainPausedReason reason,
3055
                        qemuDomainAsyncJob asyncJob)
3056
{
3057
    int ret = -1;
3058 3059
    qemuDomainObjPrivatePtr priv = vm->privateData;

3060
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
3061

3062 3063
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
J
Jiri Denemark 已提交
3064

3065
    ret = qemuMonitorStopCPUs(priv->mon);
3066 3067
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3068 3069 3070 3071

    if (ret < 0)
        goto cleanup;

3072 3073 3074
    /* de-activate netdevs after stopping CPUs */
    ignore_value(qemuInterfaceStopDevices(vm->def));

3075 3076 3077
    if (priv->job.current)
        ignore_value(virTimeMillisNow(&priv->job.current->stopped));

3078 3079 3080 3081
    virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
    if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
        VIR_WARN("Unable to release lease on %s", vm->def->name);
    VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));
J
Jiri Denemark 已提交
3082

3083
 cleanup:
3084 3085 3086 3087 3088
    return ret;
}



3089 3090 3091
static int
qemuProcessNotifyNets(virDomainDefPtr def)
{
3092
    size_t i;
3093

3094 3095
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
3096 3097 3098 3099 3100 3101 3102 3103
        /* keep others from trying to use the macvtap device name, but
         * don't return error if this happens, since that causes the
         * domain to be unceremoniously killed, which would be *very*
         * impolite.
         */
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT)
           ignore_value(virNetDevMacVLanReserveName(net->ifname, false));

3104
        if (networkNotifyActualDevice(def, net) < 0)
3105 3106 3107 3108 3109
            return -1;
    }
    return 0;
}

3110
static int
3111
qemuProcessFiltersInstantiate(virDomainDefPtr def)
3112
{
3113
    size_t i;
3114

3115
    for (i = 0; i < def->nnets; i++) {
3116 3117
        virDomainNetDefPtr net = def->nets[i];
        if ((net->filter) && (net->ifname)) {
3118
            if (virDomainConfNWFilterInstantiate(def->uuid, net) < 0)
J
Ján Tomko 已提交
3119
                return 1;
3120 3121 3122
        }
    }

J
Ján Tomko 已提交
3123
    return 0;
3124 3125
}

3126
static int
3127
qemuProcessUpdateState(virQEMUDriverPtr driver, virDomainObjPtr vm)
3128 3129 3130
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
3131
    virDomainPausedReason reason;
3132
    virDomainState newState = VIR_DOMAIN_NOSTATE;
3133
    int oldReason;
3134
    int newReason;
3135
    bool running;
3136
    char *msg = NULL;
3137 3138
    int ret;

3139
    qemuDomainObjEnterMonitor(driver, vm);
3140
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
3141 3142
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
3143

3144
    if (ret < 0)
3145 3146
        return -1;

3147
    state = virDomainObjGetState(vm, &oldReason);
3148

3149 3150 3151 3152 3153 3154 3155 3156
    if (running &&
        (state == VIR_DOMAIN_SHUTOFF ||
         (state == VIR_DOMAIN_PAUSED &&
          oldReason == VIR_DOMAIN_PAUSED_STARTING_UP))) {
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_BOOTED;
        ignore_value(VIR_STRDUP_QUIET(msg, "finished booting"));
    } else if (state == VIR_DOMAIN_PAUSED && running) {
3157 3158
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
3159
        ignore_value(VIR_STRDUP_QUIET(msg, "was unpaused"));
3160
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
3161 3162 3163
        if (reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN) {
            newState = VIR_DOMAIN_SHUTDOWN;
            newReason = VIR_DOMAIN_SHUTDOWN_UNKNOWN;
3164
            ignore_value(VIR_STRDUP_QUIET(msg, "shutdown"));
3165
        } else if (reason == VIR_DOMAIN_PAUSED_CRASHED) {
3166 3167 3168
            newState = VIR_DOMAIN_CRASHED;
            newReason = VIR_DOMAIN_CRASHED_PANICKED;
            ignore_value(VIR_STRDUP_QUIET(msg, "crashed"));
3169 3170 3171
        } else {
            newState = VIR_DOMAIN_PAUSED;
            newReason = reason;
S
Stefan Berger 已提交
3172 3173
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
3174 3175 3176 3177 3178 3179 3180
        }
    }

    if (newState != VIR_DOMAIN_NOSTATE) {
        VIR_DEBUG("Domain %s %s while its monitor was disconnected;"
                  " changing state to %s (%s)",
                  vm->def->name,
3181
                  NULLSTR(msg),
3182 3183 3184 3185
                  virDomainStateTypeToString(newState),
                  virDomainStateReasonToString(newState, newReason));
        VIR_FREE(msg);
        virDomainObjSetState(vm, newState, newReason);
3186 3187 3188 3189 3190
    }

    return 0;
}

3191
static int
3192
qemuProcessRecoverMigration(virQEMUDriverPtr driver,
3193 3194
                            virDomainObjPtr vm,
                            virConnectPtr conn,
3195 3196
                            qemuDomainAsyncJob job,
                            qemuMigrationJobPhase phase,
3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222
                            virDomainState state,
                            int reason)
{
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
        switch (phase) {
        case QEMU_MIGRATION_PHASE_NONE:
        case QEMU_MIGRATION_PHASE_PERFORM2:
        case QEMU_MIGRATION_PHASE_BEGIN3:
        case QEMU_MIGRATION_PHASE_PERFORM3:
        case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
        case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
        case QEMU_MIGRATION_PHASE_CONFIRM3:
        case QEMU_MIGRATION_PHASE_LAST:
            break;

        case QEMU_MIGRATION_PHASE_PREPARE:
            VIR_DEBUG("Killing unfinished incoming migration for domain %s",
                      vm->def->name);
            return -1;

        case QEMU_MIGRATION_PHASE_FINISH2:
            /* source domain is already killed so let's just resume the domain
             * and hope we are all set */
            VIR_DEBUG("Incoming migration finished, resuming domain %s",
                      vm->def->name);
            if (qemuProcessStartCPUs(driver, vm, conn,
3223 3224
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252
                VIR_WARN("Could not resume domain %s", vm->def->name);
            }
            break;

        case QEMU_MIGRATION_PHASE_FINISH3:
            /* migration finished, we started resuming the domain but didn't
             * confirm success or failure yet; killing it seems safest */
            VIR_DEBUG("Killing migrated domain %s", vm->def->name);
            return -1;
        }
    } else if (job == QEMU_ASYNC_JOB_MIGRATION_OUT) {
        switch (phase) {
        case QEMU_MIGRATION_PHASE_NONE:
        case QEMU_MIGRATION_PHASE_PREPARE:
        case QEMU_MIGRATION_PHASE_FINISH2:
        case QEMU_MIGRATION_PHASE_FINISH3:
        case QEMU_MIGRATION_PHASE_LAST:
            break;

        case QEMU_MIGRATION_PHASE_BEGIN3:
            /* nothing happen so far, just forget we were about to migrate the
             * domain */
            break;

        case QEMU_MIGRATION_PHASE_PERFORM2:
        case QEMU_MIGRATION_PHASE_PERFORM3:
            /* migration is still in progress, let's cancel it and resume the
             * domain */
3253
            if (qemuMigrationCancel(driver, vm) < 0)
3254
                return -1;
3255 3256 3257 3258 3259 3260
            /* resume the domain but only if it was paused as a result of
             * migration */
            if (state == VIR_DOMAIN_PAUSED &&
                (reason == VIR_DOMAIN_PAUSED_MIGRATION ||
                 reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
                if (qemuProcessStartCPUs(driver, vm, conn,
3261 3262
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281
                    VIR_WARN("Could not resume domain %s", vm->def->name);
                }
            }
            break;

        case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
            /* migration finished but we didn't have a chance to get the result
             * of Finish3 step; third party needs to check what to do next
             */
            break;

        case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
            /* Finish3 failed, we need to resume the domain */
            VIR_DEBUG("Resuming domain %s after failed migration",
                      vm->def->name);
            if (state == VIR_DOMAIN_PAUSED &&
                (reason == VIR_DOMAIN_PAUSED_MIGRATION ||
                 reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
                if (qemuProcessStartCPUs(driver, vm, conn,
3282 3283
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297
                    VIR_WARN("Could not resume domain %s", vm->def->name);
                }
            }
            break;

        case QEMU_MIGRATION_PHASE_CONFIRM3:
            /* migration completed, we need to kill the domain here */
            return -1;
        }
    }

    return 0;
}

3298
static int
3299
qemuProcessRecoverJob(virQEMUDriverPtr driver,
3300 3301 3302 3303
                      virDomainObjPtr vm,
                      virConnectPtr conn,
                      const struct qemuDomainJobObj *job)
{
3304
    qemuDomainObjPrivatePtr priv = vm->privateData;
3305 3306 3307 3308 3309 3310 3311 3312
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
    case QEMU_ASYNC_JOB_MIGRATION_IN:
3313 3314 3315
        if (qemuProcessRecoverMigration(driver, vm, conn, job->asyncJob,
                                        job->phase, state, reason) < 0)
            return -1;
3316 3317 3318 3319
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
3320
    case QEMU_ASYNC_JOB_SNAPSHOT:
3321
        qemuDomainObjEnterMonitor(driver, vm);
3322
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
3323 3324
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
3325
        /* resume the domain but only if it was paused as a result of
3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341
         * running a migration-to-file operation.  Although we are
         * recovering an async job, this function is run at startup
         * and must resume things using sync monitor connections.  */
         if (state == VIR_DOMAIN_PAUSED &&
             ((job->asyncJob == QEMU_ASYNC_JOB_DUMP &&
               reason == VIR_DOMAIN_PAUSED_DUMP) ||
              (job->asyncJob == QEMU_ASYNC_JOB_SAVE &&
               reason == VIR_DOMAIN_PAUSED_SAVE) ||
              (job->asyncJob == QEMU_ASYNC_JOB_SNAPSHOT &&
               reason == VIR_DOMAIN_PAUSED_SNAPSHOT) ||
              reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
             if (qemuProcessStartCPUs(driver, vm, conn,
                                      VIR_DOMAIN_RUNNING_UNPAUSED,
                                      QEMU_ASYNC_JOB_NONE) < 0) {
                 VIR_WARN("Could not resume domain '%s' after migration to file",
                          vm->def->name);
3342 3343 3344 3345
            }
        }
        break;

3346 3347 3348 3349
    case QEMU_ASYNC_JOB_START:
        /* Already handled in VIR_DOMAIN_PAUSED_STARTING_UP check. */
        break;

3350 3351 3352 3353 3354 3355 3356 3357
    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        break;
    }

    if (!virDomainObjIsActive(vm))
        return -1;

3358 3359 3360 3361
    /* In case any special handling is added for job type that has been ignored
     * before, QEMU_DOMAIN_TRACK_JOBS (from qemu_domain.h) needs to be updated
     * for the job to be properly tracked in domain state XML.
     */
3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381
    switch (job->active) {
    case QEMU_JOB_QUERY:
        /* harmless */
        break;

    case QEMU_JOB_DESTROY:
        VIR_DEBUG("Domain %s should have already been destroyed",
                  vm->def->name);
        return -1;

    case QEMU_JOB_SUSPEND:
        /* mostly harmless */
        break;

    case QEMU_JOB_MODIFY:
        /* XXX depending on the command we may be in an inconsistent state and
         * we should probably fall back to "monitor error" state and refuse to
         */
        break;

3382
    case QEMU_JOB_MIGRATION_OP:
3383
    case QEMU_JOB_ABORT:
3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394
    case QEMU_JOB_ASYNC:
    case QEMU_JOB_ASYNC_NESTED:
        /* async job was already handled above */
    case QEMU_JOB_NONE:
    case QEMU_JOB_LAST:
        break;
    }

    return 0;
}

3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409
static int
qemuProcessUpdateDevices(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDeviceDef dev;
    char **old;
    char **tmp;
    int ret = -1;

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

    old = priv->qemuDevices;
    priv->qemuDevices = NULL;
3410
    if (qemuDomainUpdateDeviceList(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
3411 3412 3413 3414 3415
        goto cleanup;

    if ((tmp = old)) {
        while (*tmp) {
            if (!virStringArrayHasString(priv->qemuDevices, *tmp) &&
3416 3417 3418 3419
                virDomainDefFindDevice(vm->def, *tmp, &dev, false) == 0 &&
                qemuDomainRemoveDevice(driver, vm, &dev) < 0) {
                goto cleanup;
            }
3420 3421 3422 3423 3424
            tmp++;
        }
    }
    ret = 0;

3425
 cleanup:
3426 3427 3428 3429
    virStringFreeList(old);
    return ret;
}

3430 3431
struct qemuProcessReconnectData {
    virConnectPtr conn;
3432
    virQEMUDriverPtr driver;
3433
    virDomainObjPtr obj;
3434 3435 3436 3437
};
/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
S
Stefan Berger 已提交
3438 3439 3440 3441
 *
 * We own the virConnectPtr we are passed here - whoever started
 * this thread function has increased the reference counter to it
 * so that we now have to close it.
3442
 *
3443
 * This function also inherits a locked and ref'd domain object.
3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455
 *
 * This function needs to:
 * 1. Enter job
 * 1. just before monitor reconnect do lightweight MonitorEnter
 *    (increase VM refcount and unlock VM)
 * 2. reconnect to monitor
 * 3. do lightweight MonitorExit (lock VM)
 * 4. continue reconnect process
 * 5. EndJob
 *
 * We can't do normal MonitorEnter & MonitorExit because these two lock the
 * monitor lock, which does not exists in this early phase.
3456 3457
 */
static void
3458
qemuProcessReconnect(void *opaque)
3459 3460
{
    struct qemuProcessReconnectData *data = opaque;
3461
    virQEMUDriverPtr driver = data->driver;
3462
    virDomainObjPtr obj = data->obj;
3463 3464
    qemuDomainObjPrivatePtr priv;
    virConnectPtr conn = data->conn;
3465
    struct qemuDomainJobObj oldjob;
3466 3467
    int state;
    int reason;
3468
    virQEMUDriverConfigPtr cfg;
3469
    size_t i;
3470
    int ret;
3471
    unsigned int stopFlags = 0;
3472

3473 3474
    VIR_FREE(data);

3475
    qemuDomainObjRestoreJob(obj, &oldjob);
3476 3477
    if (oldjob.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
3478

3479 3480 3481
    cfg = virQEMUDriverGetConfig(driver);
    priv = obj->privateData;

3482 3483 3484 3485 3486
    /* XXX If we ever gonna change pid file pattern, come up with
     * some intelligence here to deal with old paths. */
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, obj->def->name)))
        goto killvm;

3487 3488 3489 3490
    if (qemuDomainObjBeginJob(driver, obj, QEMU_JOB_MODIFY) < 0)
        goto killvm;

    virNWFilterReadLockFilterUpdates();
3491 3492 3493 3494

    VIR_DEBUG("Reconnect monitor to %p '%s'", obj, obj->def->name);

    /* XXX check PID liveliness & EXE path */
3495
    if (qemuConnectMonitor(driver, obj, QEMU_ASYNC_JOB_NONE, NULL) < 0)
3496 3497
        goto error;

D
Daniel P. Berrange 已提交
3498
    /* Failure to connect to agent shouldn't be fatal */
3499 3500 3501 3502
    if ((ret = qemuConnectAgent(driver, obj)) < 0) {
        if (ret == -2)
            goto error;

D
Daniel P. Berrange 已提交
3503 3504 3505 3506 3507 3508
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 obj->def->name);
        virResetLastError();
        priv->agentError = true;
    }

3509
    if (qemuHostdevUpdateActiveDomainDevices(driver, obj->def) < 0)
3510 3511
        goto error;

3512
    if (qemuConnectCgroup(driver, obj) < 0)
3513 3514
        goto error;

3515
    /* XXX: Need to change as long as lock is introduced for
3516
     * qemu_driver->sharedDevices.
3517 3518
     */
    for (i = 0; i < obj->def->ndisks; i++) {
3519 3520
        virDomainDeviceDef dev;

3521
        if (virStorageTranslateDiskSourcePool(conn, obj->def->disks[i]) < 0)
3522
            goto error;
3523

3524 3525 3526 3527 3528 3529
        /* XXX we should be able to restore all data from XML in the future.
         * This should be the only place that calls qemuDomainDetermineDiskChain
         * with @report_broken == false to guarantee best-effort domain
         * reconnect */
        if (qemuDomainDetermineDiskChain(driver, obj, obj->def->disks[i],
                                         true, false) < 0)
3530 3531
            goto error;

3532 3533 3534
        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = obj->def->disks[i];
        if (qemuAddSharedDevice(driver, &dev, obj->def->name) < 0)
3535 3536 3537
            goto error;
    }

3538 3539 3540
    if (qemuProcessUpdateState(driver, obj) < 0)
        goto error;

3541
    state = virDomainObjGetState(obj, &reason);
3542 3543 3544
    if (state == VIR_DOMAIN_SHUTOFF ||
        (state == VIR_DOMAIN_PAUSED &&
         reason == VIR_DOMAIN_PAUSED_STARTING_UP)) {
3545 3546 3547 3548 3549
        VIR_DEBUG("Domain '%s' wasn't fully started yet, killing it",
                  obj->def->name);
        goto error;
    }

3550 3551 3552
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
3553 3554
    if (!priv->qemuCaps &&
        !(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
3555 3556
                                                      obj->def->emulator,
                                                      obj->def->os.machine)))
3557 3558
        goto error;

3559
    /* In case the domain shutdown while we were not running,
3560
     * we need to finish the shutdown process. And we need to do it after
3561
     * we have virQEMUCaps filled in.
3562
     */
3563 3564 3565 3566 3567 3568
    if (state == VIR_DOMAIN_SHUTDOWN ||
        (state == VIR_DOMAIN_PAUSED &&
         reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN)) {
        VIR_DEBUG("Finishing shutdown sequence for domain %s",
                  obj->def->name);
        qemuProcessShutdownOrReboot(driver, obj);
3569 3570
        qemuDomainObjEndJob(driver, obj);
        goto cleanup;
3571 3572
    }

3573 3574
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE))
        if ((qemuDomainAssignAddresses(obj->def, priv->qemuCaps, obj)) < 0)
3575
            goto error;
3576

3577 3578 3579 3580 3581 3582
    /* if domain requests security driver we haven't loaded, report error, but
     * do not kill the domain
     */
    ignore_value(virSecurityManagerCheckAllLabel(driver->securityManager,
                                                 obj->def));

3583
    if (virSecurityManagerReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
3584 3585
        goto error;

3586 3587 3588
    if (qemuProcessNotifyNets(obj->def) < 0)
        goto error;

3589
    if (qemuProcessFiltersInstantiate(obj->def))
3590 3591
        goto error;

3592
    if (qemuDomainCheckEjectableMedia(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
3593 3594
        goto error;

3595 3596 3597
    if (qemuProcessReconnectRefreshChannelVirtioState(driver, obj) < 0)
        goto error;

3598 3599 3600
    if (qemuProcessRefreshBalloonState(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
        goto error;

3601 3602 3603
    if (qemuProcessRecoverJob(driver, obj, conn, &oldjob) < 0)
        goto error;

3604 3605 3606
    if (qemuProcessUpdateDevices(driver, obj) < 0)
        goto error;

3607
    /* update domain state XML with possibly updated state in virDomainObj */
3608
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, obj) < 0)
3609 3610
        goto error;

3611 3612
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3613
        char *xml = qemuDomainDefFormatXML(driver, obj->def, 0);
3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, obj->def->name,
                              VIR_HOOK_QEMU_OP_RECONNECT, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
        VIR_FREE(xml);

        /*
         * If the script raised an error abort the launch
         */
        if (hookret < 0)
            goto error;
    }

3628
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
3629 3630
        driver->inhibitCallback(true, driver->inhibitOpaque);

3631 3632
    qemuDomainObjEndJob(driver, obj);
    goto cleanup;
3633

3634
 error:
3635
    qemuDomainObjEndJob(driver, obj);
3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650
 killvm:
    if (virDomainObjIsActive(obj)) {
        /* We can't get the monitor back, so must kill the VM
         * to remove danger of it ending up running twice if
         * user tries to start it again later
         */
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NO_SHUTDOWN)) {
            /* If we couldn't get the monitor and qemu supports
             * no-shutdown, we can safely say that the domain
             * crashed ... */
            state = VIR_DOMAIN_SHUTOFF_CRASHED;
        } else {
            /* ... but if it doesn't we can't say what the state
             * really is and FAILED means "failed to start" */
            state = VIR_DOMAIN_SHUTOFF_UNKNOWN;
3651
        }
3652
        qemuProcessStop(driver, obj, state, stopFlags);
3653
    }
3654

3655
    qemuDomainRemoveInactive(driver, obj);
3656

3657
 cleanup:
M
Michal Privoznik 已提交
3658
    virDomainObjEndAPI(&obj);
3659
    virObjectUnref(conn);
3660
    virObjectUnref(cfg);
3661
    virNWFilterUnlockFilterUpdates();
3662 3663
}

3664 3665
static int
qemuProcessReconnectHelper(virDomainObjPtr obj,
3666 3667 3668 3669 3670 3671
                           void *opaque)
{
    virThread thread;
    struct qemuProcessReconnectData *src = opaque;
    struct qemuProcessReconnectData *data;

3672
    /* If the VM was inactive, we don't need to reconnect */
3673 3674 3675
    if (!obj->pid)
        return 0;

3676
    if (VIR_ALLOC(data) < 0)
3677
        return -1;
3678 3679

    memcpy(data, src, sizeof(*data));
3680
    data->obj = obj;
3681

3682 3683
    /* this lock and reference will be eventually transferred to the thread
     * that handles the reconnect */
3684
    virObjectLock(obj);
3685
    virObjectRef(obj);
3686

3687 3688 3689
    /* Since we close the connection later on, we have to make sure that the
     * threads we start see a valid connection throughout their lifetime. We
     * simply increase the reference counter here.
S
Stefan Berger 已提交
3690
     */
3691
    virObjectRef(data->conn);
S
Stefan Berger 已提交
3692

3693
    if (virThreadCreate(&thread, false, qemuProcessReconnect, data) < 0) {
3694 3695 3696
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not create thread. QEMU initialization "
                         "might be incomplete"));
3697 3698
       /* We can't spawn a thread and thus connect to monitor. Kill qemu. */
        qemuProcessStop(src->driver, obj, VIR_DOMAIN_SHUTOFF_FAILED, 0);
3699
        qemuDomainRemoveInactive(src->driver, obj);
3700

M
Michal Privoznik 已提交
3701
        virDomainObjEndAPI(&obj);
3702 3703 3704 3705
        virObjectUnref(data->conn);
        VIR_FREE(data);
        return -1;
    }
3706

3707
    return 0;
3708 3709 3710 3711 3712 3713 3714 3715 3716
}

/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
3717
qemuProcessReconnectAll(virConnectPtr conn, virQEMUDriverPtr driver)
3718
{
3719
    struct qemuProcessReconnectData data = {.conn = conn, .driver = driver};
3720
    virDomainObjListForEach(driver->domains, qemuProcessReconnectHelper, &data);
3721 3722
}

3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745
static int
qemuProcessVNCAllocatePorts(virQEMUDriverPtr driver,
                            virDomainGraphicsDefPtr graphics)
{
    unsigned short port;

    if (graphics->data.vnc.socket)
        return 0;

    if (graphics->data.vnc.autoport) {
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
            return -1;
        graphics->data.vnc.port = port;
    }

    if (graphics->data.vnc.websocket == -1) {
        if (virPortAllocatorAcquire(driver->webSocketPorts, &port) < 0)
            return -1;
        graphics->data.vnc.websocket = port;
    }

    return 0;
}
3746

3747
int
3748 3749
qemuProcessSPICEAllocatePorts(virQEMUDriverPtr driver,
                              virQEMUDriverConfigPtr cfg,
3750 3751
                              virDomainGraphicsDefPtr graphics,
                              bool allocate)
3752 3753 3754
{
    unsigned short port = 0;
    unsigned short tlsPort;
3755
    size_t i;
3756 3757 3758 3759 3760 3761 3762
    int defaultMode = graphics->data.spice.defaultMode;

    bool needTLSPort = false;
    bool needPort = false;

    if (graphics->data.spice.autoport) {
        /* check if tlsPort or port need allocation */
3763
        for (i = 0; i < VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_LAST; i++) {
3764 3765 3766 3767 3768 3769 3770 3771 3772 3773
            switch (graphics->data.spice.channels[i]) {
            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
                needTLSPort = true;
                break;

            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
                needPort = true;
                break;

            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
3774
                /* default mode will be used */
3775 3776 3777
                break;
            }
        }
3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792
        switch (defaultMode) {
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
            needTLSPort = true;
            break;

        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
            needPort = true;
            break;

        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
            if (cfg->spiceTLS)
                needTLSPort = true;
            needPort = true;
            break;
        }
3793 3794
    }

3795 3796 3797 3798 3799 3800 3801 3802 3803 3804
    if (!allocate) {
        if (needPort || graphics->data.spice.port == -1)
            graphics->data.spice.port = 5901;

        if (needTLSPort || graphics->data.spice.tlsPort == -1)
            graphics->data.spice.tlsPort = 5902;

        return 0;
    }

3805
    if (needPort || graphics->data.spice.port == -1) {
3806
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
3807
            goto error;
3808 3809 3810 3811

        graphics->data.spice.port = port;
    }

3812 3813
    if (needTLSPort || graphics->data.spice.tlsPort == -1) {
        if (!cfg->spiceTLS) {
3814 3815 3816 3817
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Auto allocation of spice TLS port requested "
                             "but spice TLS is disabled in qemu.conf"));
            goto error;
3818
        }
3819 3820 3821 3822 3823

        if (virPortAllocatorAcquire(driver->remotePorts, &tlsPort) < 0)
            goto error;

        graphics->data.spice.tlsPort = tlsPort;
3824 3825
    }

3826
    return 0;
3827

3828
 error:
3829
    virPortAllocatorRelease(driver->remotePorts, port);
3830
    return -1;
3831 3832 3833
}


3834 3835 3836
static int
qemuValidateCpuCount(virDomainDefPtr def,
                     virQEMUCapsPtr qemuCaps)
3837
{
3838
    unsigned int maxCpus = virQEMUCapsGetMachineMaxCpus(qemuCaps, def->os.machine);
3839

3840 3841 3842 3843 3844 3845
    if (virDomainDefGetVcpus(def) == 0) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Domain requires at least 1 vCPU"));
        return -1;
    }

3846 3847 3848 3849
    if (maxCpus > 0 && virDomainDefGetVcpusMax(def) > maxCpus) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Maximum CPUs greater than specified machine type limit"));
        return -1;
3850 3851
    }

3852
    return 0;
3853 3854
}

3855 3856

static bool
3857 3858 3859
qemuProcessVerifyGuestCPU(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          int asyncJob)
3860 3861 3862 3863 3864
{
    virDomainDefPtr def = vm->def;
    virArch arch = def->os.arch;
    virCPUDataPtr guestcpu = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3865
    int rc;
3866
    bool ret = false;
J
Ján Tomko 已提交
3867
    size_t i;
3868

3869 3870 3871 3872 3873
    /* no features are passed to QEMU with -cpu host
     * so it makes no sense to verify them */
    if (def->cpu && def->cpu->mode == VIR_CPU_MODE_HOST_PASSTHROUGH)
        return true;

3874 3875 3876
    switch (arch) {
    case VIR_ARCH_I686:
    case VIR_ARCH_X86_64:
3877 3878
        if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
            return false;
3879
        rc = qemuMonitorGetGuestCPU(priv->mon, arch, &guestcpu);
3880 3881
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return false;
3882

3883 3884 3885 3886
        if (rc < 0) {
            if (rc == -2)
                break;

3887
            goto cleanup;
3888
        }
3889

J
Ján Tomko 已提交
3890
        if (def->features[VIR_DOMAIN_FEATURE_PVSPINLOCK] == VIR_TRISTATE_SWITCH_ON) {
3891 3892 3893 3894 3895 3896
            if (!cpuHasFeature(guestcpu, VIR_CPU_x86_KVM_PV_UNHALT)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("host doesn't support paravirtual spinlocks"));
                goto cleanup;
            }
        }
J
Ján Tomko 已提交
3897

3898
        for (i = 0; def->cpu && i < def->cpu->nfeatures; i++) {
J
Ján Tomko 已提交
3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910
            virCPUFeatureDefPtr feature = &def->cpu->features[i];

            if (feature->policy != VIR_CPU_FEATURE_REQUIRE)
                continue;

            if (STREQ(feature->name, "invtsc") &&
                !cpuHasFeature(guestcpu, feature->name)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("host doesn't support invariant TSC"));
                goto cleanup;
            }
        }
3911 3912 3913 3914 3915 3916 3917 3918
        break;

    default:
        break;
    }

    ret = true;

3919
 cleanup:
3920 3921 3922 3923 3924
    cpuDataFree(guestcpu);
    return ret;
}


3925 3926
static int
qemuPrepareNVRAM(virQEMUDriverConfigPtr cfg,
3927
                 virDomainObjPtr vm,
3928 3929 3930 3931 3932
                 bool migrated)
{
    int ret = -1;
    int srcFD = -1;
    int dstFD = -1;
3933
    virDomainLoaderDefPtr loader = vm->def->os.loader;
3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958
    bool generated = false;
    bool created = false;

    /* Unless domain has RO loader of pflash type, we have
     * nothing to do here.  If the loader is RW then it's not
     * using split code and vars feature, so no nvram file needs
     * to be created. */
    if (!loader || loader->type != VIR_DOMAIN_LOADER_TYPE_PFLASH ||
        loader->readonly != VIR_TRISTATE_SWITCH_ON)
        return 0;

    /* If the nvram path is configured already, there's nothing
     * we need to do. Unless we are starting the destination side
     * of migration in which case nvram is configured in the
     * domain XML but the file doesn't exist yet. Moreover, after
     * the migration is completed, qemu will invoke a
     * synchronization write into the nvram file so we don't have
     * to take care about transmitting the real data on the other
     * side. */
    if (loader->nvram && !migrated)
        return 0;

    /* Autogenerate nvram path if needed.*/
    if (!loader->nvram) {
        if (virAsprintf(&loader->nvram,
3959 3960
                        "%s/%s_VARS.fd",
                        cfg->nvramDir, vm->def->name) < 0)
3961 3962 3963
            goto cleanup;

        generated = true;
3964

3965 3966
        if (vm->persistent &&
            virDomainSaveConfig(cfg->configDir, vm->def) < 0)
3967
            goto cleanup;
3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057
    }

    if (!virFileExists(loader->nvram)) {
        const char *master_nvram_path = loader->templt;
        ssize_t r;

        if (!loader->templt) {
            size_t i;
            for (i = 0; i < cfg->nloader; i++) {
                if (STREQ(cfg->loader[i], loader->path)) {
                    master_nvram_path = cfg->nvram[i];
                    break;
                }
            }
        }

        if (!master_nvram_path) {
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("unable to find any master var store for "
                             "loader: %s"), loader->path);
            goto cleanup;
        }

        if ((srcFD = virFileOpenAs(master_nvram_path, O_RDONLY,
                                   0, -1, -1, 0)) < 0) {
            virReportSystemError(-srcFD,
                                 _("Failed to open file '%s'"),
                                 master_nvram_path);
            goto cleanup;
        }
        if ((dstFD = virFileOpenAs(loader->nvram,
                                   O_WRONLY | O_CREAT | O_EXCL,
                                   S_IRUSR | S_IWUSR,
                                   cfg->user, cfg->group, 0)) < 0) {
            virReportSystemError(-dstFD,
                                 _("Failed to create file '%s'"),
                                 loader->nvram);
            goto cleanup;
        }
        created = true;

        do {
            char buf[1024];

            if ((r = saferead(srcFD, buf, sizeof(buf))) < 0) {
                virReportSystemError(errno,
                                     _("Unable to read from file '%s'"),
                                     master_nvram_path);
                goto cleanup;
            }

            if (safewrite(dstFD, buf, r) < 0) {
                virReportSystemError(errno,
                                     _("Unable to write to file '%s'"),
                                     loader->nvram);
                goto cleanup;
            }
        } while (r);

        if (VIR_CLOSE(srcFD) < 0) {
            virReportSystemError(errno,
                                 _("Unable to close file '%s'"),
                                 master_nvram_path);
            goto cleanup;
        }
        if (VIR_CLOSE(dstFD) < 0) {
            virReportSystemError(errno,
                                 _("Unable to close file '%s'"),
                                 loader->nvram);
            goto cleanup;
        }
    }

    ret = 0;
 cleanup:
    /* We successfully generated the nvram path, but failed to
     * copy the file content. Roll back. */
    if (ret < 0) {
        if (created)
            unlink(loader->nvram);
        if (generated)
            VIR_FREE(loader->nvram);
    }

    VIR_FORCE_CLOSE(srcFD);
    VIR_FORCE_CLOSE(dstFD);
    return ret;
}


4058 4059 4060
static void
qemuLogOperation(virDomainObjPtr vm,
                 const char *msg,
4061 4062
                 virCommandPtr cmd,
                 qemuDomainLogContextPtr logCtxt)
4063 4064 4065 4066 4067
{
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int qemuVersion = virQEMUCapsGetVersion(priv->qemuCaps);
    const char *package = virQEMUCapsGetPackage(priv->qemuCaps);
4068
    char *hostname = virGetHostname();
4069 4070

    if ((timestamp = virTimeStringNow()) == NULL)
4071
        goto cleanup;
4072

4073 4074
    if (qemuDomainLogContextWrite(logCtxt,
                                  "%s: %s %s, qemu version: %d.%d.%d%s, hostname: %s\n",
4075 4076 4077 4078
                                  timestamp, msg, VIR_LOG_VERSION_STRING,
                                  (qemuVersion / 1000000) % 1000,
                                  (qemuVersion / 1000) % 1000,
                                  qemuVersion % 1000,
4079 4080
                                  package ? package : "",
                                  hostname ? hostname : "") < 0)
4081
        goto cleanup;
4082

4083 4084 4085 4086 4087
    if (cmd) {
        char *args = virCommandToString(cmd);
        qemuDomainLogContextWrite(logCtxt, "%s\n", args);
        VIR_FREE(args);
    }
4088 4089

 cleanup:
4090
    VIR_FREE(hostname);
4091 4092 4093
    VIR_FREE(timestamp);
}

4094 4095 4096 4097 4098 4099 4100

void
qemuProcessIncomingDefFree(qemuProcessIncomingDefPtr inc)
{
    if (!inc)
        return;

4101
    VIR_FREE(inc->address);
4102
    VIR_FREE(inc->launchURI);
4103
    VIR_FREE(inc->deferredURI);
4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114
    VIR_FREE(inc);
}


/*
 * This function does not copy @path, the caller is responsible for keeping
 * the @path pointer valid during the lifetime of the allocated
 * qemuProcessIncomingDef structure.
 */
qemuProcessIncomingDefPtr
qemuProcessIncomingDefNew(virQEMUCapsPtr qemuCaps,
4115
                          const char *listenAddress,
4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127
                          const char *migrateFrom,
                          int fd,
                          const char *path)
{
    qemuProcessIncomingDefPtr inc = NULL;

    if (qemuMigrationCheckIncoming(qemuCaps, migrateFrom) < 0)
        return NULL;

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

4128 4129 4130
    if (VIR_STRDUP(inc->address, listenAddress) < 0)
        goto error;

4131 4132 4133 4134
    inc->launchURI = qemuMigrationIncomingURI(migrateFrom, fd);
    if (!inc->launchURI)
        goto error;

4135 4136 4137 4138 4139 4140
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_INCOMING_DEFER)) {
        inc->deferredURI = inc->launchURI;
        if (VIR_STRDUP(inc->launchURI, "defer") < 0)
            goto error;
    }

4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151
    inc->fd = fd;
    inc->path = path;

    return inc;

 error:
    qemuProcessIncomingDefFree(inc);
    return NULL;
}


4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181
/*
 * This function starts a new QEMU_ASYNC_JOB_START async job. The user is
 * responsible for calling qemuProcessEndJob to stop this job and for passing
 * QEMU_ASYNC_JOB_START as @asyncJob argument to any function requiring this
 * parameter between qemuProcessBeginJob and qemuProcessEndJob.
 */
int
qemuProcessBeginJob(virQEMUDriverPtr driver,
                    virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (qemuDomainObjBeginAsyncJob(driver, vm, QEMU_ASYNC_JOB_START) < 0)
        return -1;

    qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
    priv->job.current->type = VIR_DOMAIN_JOB_UNBOUNDED;

    return 0;
}


void
qemuProcessEndJob(virQEMUDriverPtr driver,
                  virDomainObjPtr vm)
{
    qemuDomainObjEndAsyncJob(driver, vm);
}


4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204
static int
qemuProcessStartHook(virQEMUDriverPtr driver,
                     virDomainObjPtr vm,
                     virHookQemuOpType op,
                     virHookSubopType subop)
{
    char *xml;
    int ret;

    if (!virHookPresent(VIR_HOOK_DRIVER_QEMU))
        return 0;

    if (!(xml = qemuDomainDefFormatXML(driver, vm->def, 0)))
        return -1;

    ret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name, op, subop,
                      NULL, xml, NULL);
    VIR_FREE(xml);

    return ret;
}


4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282
static int
qemuProcessSetupGraphics(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    size_t i;
    int ret = -1;

    for (i = 0; i < vm->def->ngraphics; ++i) {
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
            !graphics->data.vnc.autoport) {
            if (virPortAllocatorSetUsed(driver->remotePorts,
                                        graphics->data.vnc.port,
                                        true) < 0)
                goto cleanup;
            graphics->data.vnc.portReserved = true;

        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE &&
                   !graphics->data.spice.autoport) {
            if (graphics->data.spice.port > 0) {
                if (virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.port,
                                            true) < 0)
                    goto cleanup;
                graphics->data.spice.portReserved = true;
            }

            if (graphics->data.spice.tlsPort > 0) {
                if (virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.tlsPort,
                                            true) < 0)
                    goto cleanup;
                graphics->data.spice.tlsPortReserved = true;
            }
        }
    }

    for (i = 0; i < vm->def->ngraphics; ++i) {
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (qemuProcessVNCAllocatePorts(driver, graphics) < 0)
                goto cleanup;
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (qemuProcessSPICEAllocatePorts(driver, cfg, graphics, true) < 0)
                goto cleanup;
        }

        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ||
            graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (graphics->nListens == 0) {
                if (VIR_EXPAND_N(graphics->listens, graphics->nListens, 1) < 0)
                    goto cleanup;
                graphics->listens[0].type = VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_ADDRESS;
                if (VIR_STRDUP(graphics->listens[0].address,
                               graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ?
                               cfg->vncListen : cfg->spiceListen) < 0) {
                    VIR_SHRINK_N(graphics->listens, graphics->nListens, 1);
                    goto cleanup;
                }
                graphics->listens[0].fromConfig = true;
            } else if (graphics->nListens > 1) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("QEMU does not support multiple listen "
                                 "addresses for one graphics device."));
                goto cleanup;
            }
        }
    }

    ret = 0;

 cleanup:
    virObjectUnref(cfg);
    return ret;
}


4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341
static int
qemuProcessSetupRawIO(virQEMUDriverPtr driver,
                      virDomainObjPtr vm,
                      virCommandPtr cmd ATTRIBUTE_UNUSED)
{
    bool rawio = false;
    size_t i;
    int ret = -1;

    /* in case a certain disk is desirous of CAP_SYS_RAWIO, add this */
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDeviceDef dev;
        virDomainDiskDefPtr disk = vm->def->disks[i];

        if (disk->rawio == VIR_TRISTATE_BOOL_YES) {
            rawio = true;
#ifndef CAP_SYS_RAWIO
            break;
#endif
        }

        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        if (qemuAddSharedDevice(driver, &dev, vm->def->name) < 0)
            goto cleanup;

        if (qemuSetUnprivSGIO(&dev) < 0)
            goto cleanup;
    }

    /* If rawio not already set, check hostdevs as well */
    if (!rawio) {
        for (i = 0; i < vm->def->nhostdevs; i++) {
            virDomainHostdevSubsysSCSIPtr scsisrc =
                &vm->def->hostdevs[i]->source.subsys.u.scsi;
            if (scsisrc->rawio == VIR_TRISTATE_BOOL_YES) {
                rawio = true;
                break;
            }
        }
    }

    ret = 0;

 cleanup:
    if (rawio) {
#ifdef CAP_SYS_RAWIO
        if (ret == 0)
            virCommandAllowCap(cmd, CAP_SYS_RAWIO);
#else
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Raw I/O is not supported on this platform"));
        ret = -1;
#endif
    }
    return ret;
}


4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374
static int
qemuProcessSetupBalloon(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuDomainAsyncJob asyncJob)
{
    unsigned long long balloon = vm->def->mem.cur_balloon;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int period;
    int ret = -1;

    if (!vm->def->memballoon ||
        vm->def->memballoon->model == VIR_DOMAIN_MEMBALLOON_MODEL_NONE)
        return 0;

    period = vm->def->memballoon->period;

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

    if (period)
        qemuMonitorSetMemoryStatsPeriod(priv->mon, period);
    if (qemuMonitorSetBalloon(priv->mon, balloon) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
    return ret;
}


J
Jiri Denemark 已提交
4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402
static int
qemuProcessMakeDir(virQEMUDriverPtr driver,
                   virDomainObjPtr vm,
                   const char *parentDir)
{
    char *path = NULL;
    int ret = -1;

    if (virAsprintf(&path, "%s/domain-%s", parentDir, vm->def->name) < 0)
        goto cleanup;

    if (virFileMakePathWithMode(path, 0750) < 0) {
        virReportSystemError(errno, _("Cannot create directory '%s'"), path);
        goto cleanup;
    }

    if (virSecurityManagerDomainSetDirLabel(driver->securityManager,
                                            vm->def, path) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    VIR_FREE(path);
    return ret;
}


J
Jiri Denemark 已提交
4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486
/**
 * qemuProcessInit:
 *
 * Prepares the domain up to the point when priv->qemuCaps is initialized. The
 * function calls qemuProcessStop when needed.
 *
 * Returns 0 on success, -1 on error.
 */
int
qemuProcessInit(virQEMUDriverPtr driver,
                virDomainObjPtr vm,
                bool migration)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int stopFlags;
    int ret = -1;

    VIR_DEBUG("vm=%p name=%s id=%d migration=%d",
              vm, vm->def->name, vm->def->id, migration);

    VIR_DEBUG("Beginning VM startup process");

    if (virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("VM is already active"));
        goto cleanup;
    }

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

    /* Some things, paths, ... are generated here and we want them to persist.
     * Fill them in prior to setting the domain def as transient. */
    VIR_DEBUG("Generating paths");

    if (qemuPrepareNVRAM(cfg, vm, migration) < 0)
        goto stop;

    /* Do this upfront, so any part of the startup process can add
     * runtime state to vm->def that won't be persisted. This let's us
     * report implicit runtime defaults in the XML, like vnc listen/socket
     */
    VIR_DEBUG("Setting current domain def as transient");
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm, true) < 0)
        goto stop;

    vm->def->id = qemuDriverAllocateID(driver);
    qemuDomainSetFakeReboot(driver, vm, false);
    virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_STARTING_UP);

    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
        driver->inhibitCallback(true, driver->inhibitOpaque);

    /* Run an early hook to set-up missing devices */
    if (qemuProcessStartHook(driver, vm,
                             VIR_HOOK_QEMU_OP_PREPARE,
                             VIR_HOOK_SUBOP_BEGIN) < 0)
        goto stop;

    VIR_DEBUG("Determining emulator version");
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
        goto stop;

    ret = 0;

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

 stop:
    stopFlags = VIR_QEMU_PROCESS_STOP_NO_RELABEL;
    if (migration)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, stopFlags);
    goto cleanup;
}


J
Jiri Denemark 已提交
4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508
/**
 * qemuProcessLaunch:
 *
 * Launch a new QEMU process with stopped virtual CPUs.
 *
 * The caller is supposed to call qemuProcessStop with appropriate
 * flags in case of failure.
 *
 * Returns 0 on success,
 *        -1 on error which happened before devices were labeled and thus
 *           there is no need to restore them,
 *        -2 on error requesting security labels to be restored.
 */
int
qemuProcessLaunch(virConnectPtr conn,
                  virQEMUDriverPtr driver,
                  virDomainObjPtr vm,
                  qemuDomainAsyncJob asyncJob,
                  qemuProcessIncomingDefPtr incoming,
                  virDomainSnapshotObjPtr snapshot,
                  virNetDevVPortProfileOp vmop,
                  unsigned int flags)
4509
{
4510
    int ret = -1;
4511
    int rv;
4512
    int logfile = -1;
4513
    qemuDomainLogContextPtr logCtxt = NULL;
4514 4515 4516
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCommandPtr cmd = NULL;
    struct qemuProcessHookData hookData;
4517
    size_t i;
4518
    char *nodeset = NULL;
J
Jiri Denemark 已提交
4519
    virQEMUDriverConfigPtr cfg;
4520
    virCapsPtr caps = NULL;
4521
    unsigned int hostdev_flags = 0;
4522 4523
    size_t nnicindexes = 0;
    int *nicindexes = NULL;
4524
    bool check_shmem = false;
4525

J
Jiri Denemark 已提交
4526 4527 4528 4529 4530 4531 4532 4533 4534 4535
    VIR_DEBUG("vm=%p name=%s id=%d asyncJob=%d "
              "incoming.launchURI=%s incoming.deferredURI=%s "
              "incoming.fd=%d incoming.path=%s "
              "snapshot=%p vmop=%d flags=0x%x",
              vm, vm->def->name, vm->def->id, asyncJob,
              NULLSTR(incoming ? incoming->launchURI : NULL),
              NULLSTR(incoming ? incoming->deferredURI : NULL),
              incoming ? incoming->fd : -1,
              NULLSTR(incoming ? incoming->path : NULL),
              snapshot, vmop, flags);
4536

4537 4538 4539 4540
    /* Okay, these are just internal flags,
     * but doesn't hurt to check */
    virCheckFlags(VIR_QEMU_PROCESS_START_COLD |
                  VIR_QEMU_PROCESS_START_PAUSED |
4541
                  VIR_QEMU_PROCESS_START_AUTODESTROY, -1);
4542

4543 4544
    cfg = virQEMUDriverGetConfig(driver);

4545 4546 4547
    hookData.conn = conn;
    hookData.vm = vm;
    hookData.driver = driver;
4548 4549
    /* We don't increase cfg's reference counter here. */
    hookData.cfg = cfg;
4550

4551
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
J
Jiri Denemark 已提交
4552
        goto cleanup;
4553

4554 4555 4556 4557 4558 4559
    /* network devices must be "prepared" before hostdevs, because
     * setting up a network device might create a new hostdev that
     * will need to be setup.
     */
    VIR_DEBUG("Preparing network devices");
    if (qemuNetworkPrepareDevices(vm->def) < 0)
J
Jiri Denemark 已提交
4560
        goto cleanup;
4561

4562
    /* Must be run before security labelling */
4563
    VIR_DEBUG("Preparing host devices");
4564 4565
    if (!cfg->relaxedACS)
        hostdev_flags |= VIR_HOSTDEV_STRICT_ACS_CHECK;
J
Jiri Denemark 已提交
4566
    if (!incoming)
4567
        hostdev_flags |= VIR_HOSTDEV_COLD_BOOT;
4568 4569
    if (qemuHostdevPrepareDomainDevices(driver, vm->def, priv->qemuCaps,
                                        hostdev_flags) < 0)
J
Jiri Denemark 已提交
4570
        goto cleanup;
4571

4572
    VIR_DEBUG("Preparing chr devices");
4573 4574 4575 4576
    if (virDomainChrDefForeach(vm->def,
                               true,
                               qemuProcessPrepareChardevDevice,
                               NULL) < 0)
J
Jiri Denemark 已提交
4577
        goto cleanup;
4578

4579 4580
    VIR_DEBUG("Checking domain and device security labels");
    if (virSecurityManagerCheckAllLabel(driver->securityManager, vm->def) < 0)
J
Jiri Denemark 已提交
4581
        goto cleanup;
4582

4583 4584
    /* If you are using a SecurityDriver with dynamic labelling,
       then generate a security label for isolation */
4585
    VIR_DEBUG("Generating domain security label (if required)");
4586
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0) {
4587
        virDomainAuditSecurityLabel(vm, false);
J
Jiri Denemark 已提交
4588
        goto cleanup;
4589
    }
4590
    virDomainAuditSecurityLabel(vm, true);
4591

4592
    if (vm->def->mem.nhugepages) {
4593 4594 4595 4596
        for (i = 0; i < cfg->nhugetlbfs; i++) {
            char *hugepagePath = qemuGetHugepagePath(&cfg->hugetlbfs[i]);

            if (!hugepagePath)
J
Jiri Denemark 已提交
4597
                goto cleanup;
4598 4599 4600 4601 4602 4603

            if (virSecurityManagerSetHugepages(driver->securityManager,
                                               vm->def, hugepagePath) < 0) {
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("Unable to set huge path in security driver"));
                VIR_FREE(hugepagePath);
J
Jiri Denemark 已提交
4604
                goto cleanup;
4605 4606
            }
            VIR_FREE(hugepagePath);
4607 4608 4609
        }
    }

4610 4611
    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
4612
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
4613
    qemuRemoveCgroup(driver, vm);
4614

4615 4616
    VIR_DEBUG("Setting up ports for graphics");
    if (qemuProcessSetupGraphics(driver, vm) < 0)
J
Jiri Denemark 已提交
4617
        goto cleanup;
4618

4619
    if (virFileMakePath(cfg->logDir) < 0) {
4620 4621
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
4622
                             cfg->logDir);
J
Jiri Denemark 已提交
4623
        goto cleanup;
4624 4625
    }

4626
    VIR_DEBUG("Creating domain log file");
4627 4628
    if (!(logCtxt = qemuDomainLogContextNew(driver, vm,
                                            QEMU_DOMAIN_LOG_CONTEXT_MODE_START)))
J
Jiri Denemark 已提交
4629
        goto cleanup;
4630
    logfile = qemuDomainLogContextGetWriteFD(logCtxt);
4631

4632 4633
    if (vm->def->virtType == VIR_DOMAIN_VIRT_KVM) {
        VIR_DEBUG("Checking for KVM availability");
4634
        if (!virFileExists("/dev/kvm")) {
4635 4636 4637 4638
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Domain requires KVM, but it is not available. "
                             "Check that virtualization is enabled in the host BIOS, "
                             "and host configuration is setup to load the kvm modules."));
J
Jiri Denemark 已提交
4639
            goto cleanup;
4640 4641 4642
        }
    }

4643
    if (qemuValidateCpuCount(vm->def, priv->qemuCaps) < 0)
J
Jiri Denemark 已提交
4644
        goto cleanup;
4645

4646
    if (qemuAssignDeviceAliases(vm->def, priv->qemuCaps) < 0)
J
Jiri Denemark 已提交
4647
        goto cleanup;
4648

4649 4650 4651
    /* Get the advisory nodeset from numad if 'placement' of
     * either <vcpu> or <numatune> is 'auto'.
     */
4652
    if (virDomainDefNeedsPlacementAdvice(vm->def)) {
4653
        nodeset = virNumaGetAutoPlacementAdvice(virDomainDefGetVcpus(vm->def),
4654
                                                virDomainDefGetMemoryActual(vm->def));
4655
        if (!nodeset)
J
Jiri Denemark 已提交
4656
            goto cleanup;
4657 4658 4659

        VIR_DEBUG("Nodeset returned from numad: %s", nodeset);

4660 4661
        if (virBitmapParse(nodeset, 0, &priv->autoNodeset,
                           VIR_DOMAIN_CPUMASK_LEN) < 0)
J
Jiri Denemark 已提交
4662
            goto cleanup;
4663 4664 4665

        if (!(priv->autoCpuset = virCapabilitiesGetCpusForNodemask(caps,
                                                                   priv->autoNodeset)))
J
Jiri Denemark 已提交
4666
            goto cleanup;
4667 4668
    }

J
Jiri Denemark 已提交
4669
    if (!incoming && !snapshot &&
4670
        virDomainDefCheckDuplicateDiskInfo(vm->def) < 0)
J
Jiri Denemark 已提交
4671
        goto cleanup;
4672

4673 4674 4675 4676
    /* "volume" type disk's source must be translated before
     * cgroup and security setting.
     */
    for (i = 0; i < vm->def->ndisks; i++) {
4677
        if (virStorageTranslateDiskSourcePool(conn, vm->def->disks[i]) < 0)
J
Jiri Denemark 已提交
4678
            goto cleanup;
4679 4680
    }

4681 4682
    if (qemuDomainCheckDiskPresence(driver, vm,
                                    flags & VIR_QEMU_PROCESS_START_COLD) < 0)
J
Jiri Denemark 已提交
4683
        goto cleanup;
4684

4685 4686 4687 4688
    if (vm->def->mem.min_guarantee) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Parameter 'min_guarantee' "
                         "not supported by QEMU."));
J
Jiri Denemark 已提交
4689
        goto cleanup;
4690 4691
    }

4692
    if (VIR_ALLOC(priv->monConfig) < 0)
J
Jiri Denemark 已提交
4693
        goto cleanup;
4694

4695
    VIR_DEBUG("Preparing monitor state");
4696
    if (qemuProcessPrepareMonitorChr(cfg, priv->monConfig, vm->def->name) < 0)
J
Jiri Denemark 已提交
4697
        goto cleanup;
4698

E
Eric Blake 已提交
4699
    priv->monJSON = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON);
4700 4701
    priv->monError = false;
    priv->monStart = 0;
4702 4703
    priv->gotShutdown = false;

4704
    VIR_FREE(priv->pidfile);
4705
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, vm->def->name))) {
4706 4707
        virReportSystemError(errno,
                             "%s", _("Failed to build pidfile path."));
J
Jiri Denemark 已提交
4708
        goto cleanup;
4709 4710
    }

4711 4712
    if (unlink(priv->pidfile) < 0 &&
        errno != ENOENT) {
4713
        virReportSystemError(errno,
4714 4715
                             _("Cannot remove stale PID file %s"),
                             priv->pidfile);
J
Jiri Denemark 已提交
4716
        goto cleanup;
4717 4718 4719 4720 4721 4722
    }

    /*
     * Normally PCI addresses are assigned in the virDomainCreate
     * or virDomainDefine methods. We might still need to assign
     * some here to cope with the question of upgrades. Regardless
M
Martin Kletzander 已提交
4723
     * we also need to populate the PCI address set cache for later
4724 4725
     * use in hotplug
     */
4726
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4727
        VIR_DEBUG("Assigning domain PCI addresses");
4728
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
J
Jiri Denemark 已提交
4729
            goto cleanup;
4730 4731
    }

4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762
    VIR_DEBUG("Checking for any possible (non-fatal) issues");

    /*
     * For vhost-user to work, the domain has to have some type of
     * shared memory configured.  We're not the proper ones to judge
     * whether shared hugepages or shm are enough and will be in the
     * future, so we'll just warn in case neither is configured.
     * Moreover failing would give the false illusion that libvirt is
     * really checking that everything works before running the domain
     * and not only we are unable to do that, but it's also not our
     * aim to do so.
     */
    for (i = 0; i < vm->def->nnets; i++) {
        if (virDomainNetGetActualType(vm->def->nets[i]) ==
                                      VIR_DOMAIN_NET_TYPE_VHOSTUSER) {
            check_shmem = true;
            break;
        }
    }

    if (check_shmem) {
        bool shmem = vm->def->nshmems;

        /*
         * This check is by no means complete.  We merely check
         * whether there are *some* hugepages enabled and *some* NUMA
         * nodes with shared memory access.
         */
        if (!shmem && vm->def->mem.nhugepages) {
            for (i = 0; i < virDomainNumaGetNodeCount(vm->def->numa); i++) {
                if (virDomainNumaGetNodeMemoryAccessMode(vm->def->numa, i) ==
4763
                    VIR_NUMA_MEM_ACCESS_SHARED) {
4764
                    shmem = true;
4765 4766
                    break;
                }
4767 4768 4769 4770 4771 4772 4773 4774 4775
            }
        }

        if (!shmem) {
            VIR_WARN("Detected vhost-user interface without any shared memory, "
                     "the interface might not be operational");
        }
    }

4776
    VIR_DEBUG("Building emulator command line");
4777
    if (!(cmd = qemuBuildCommandLine(conn, driver, vm->def, priv->monConfig,
E
Eric Blake 已提交
4778
                                     priv->monJSON, priv->qemuCaps,
4779 4780
                                     incoming ? incoming->launchURI : NULL,
                                     snapshot, vmop,
4781
                                     &buildCommandLineCallbacks, false,
4782
                                     qemuCheckFips(),
4783
                                     priv->autoNodeset,
4784
                                     &nnicindexes, &nicindexes)))
J
Jiri Denemark 已提交
4785
        goto cleanup;
4786

4787 4788
    if (incoming && incoming->fd != -1)
        virCommandPassFD(cmd, incoming->fd, 0);
4789 4790 4791 4792 4793

    /*
     * Create all per-domain directories in order to make sure domain
     * with any possible seclabels can access it.
     */
J
Jiri Denemark 已提交
4794 4795
    if (qemuProcessMakeDir(driver, vm, cfg->libDir) < 0 ||
        qemuProcessMakeDir(driver, vm, cfg->channelTargetDir) < 0)
J
Jiri Denemark 已提交
4796
        goto cleanup;
4797

4798
    /* now that we know it is about to start call the hook if present */
4799 4800 4801
    if (qemuProcessStartHook(driver, vm,
                             VIR_HOOK_QEMU_OP_START,
                             VIR_HOOK_SUBOP_BEGIN) < 0)
J
Jiri Denemark 已提交
4802
        goto cleanup;
4803

4804
    qemuLogOperation(vm, "starting up", cmd, logCtxt);
4805

4806
    qemuDomainObjCheckTaint(driver, vm, logCtxt);
4807

4808
    qemuDomainLogContextMarkPosition(logCtxt);
4809 4810

    VIR_DEBUG("Clear emulator capabilities: %d",
4811 4812
              cfg->clearEmulatorCapabilities);
    if (cfg->clearEmulatorCapabilities)
4813 4814
        virCommandClearCaps(cmd);

4815 4816
    VIR_DEBUG("Setting up raw IO");
    if (qemuProcessSetupRawIO(driver, vm, cmd) < 0)
J
Jiri Denemark 已提交
4817
        goto cleanup;
4818

4819
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);
4820 4821
    virCommandSetMaxProcesses(cmd, cfg->maxProcesses);
    virCommandSetMaxFiles(cmd, cfg->maxFiles);
4822
    virCommandSetUmask(cmd, 0x002);
4823

4824 4825
    VIR_DEBUG("Setting up security labelling");
    if (virSecurityManagerSetChildProcessLabel(driver->securityManager,
4826
                                               vm->def, cmd) < 0)
J
Jiri Denemark 已提交
4827
        goto cleanup;
4828

4829 4830 4831
    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
4832
    virCommandSetPidFile(cmd, priv->pidfile);
4833
    virCommandDaemonize(cmd);
4834
    virCommandRequireHandshake(cmd);
4835

4836
    if (virSecurityManagerPreFork(driver->securityManager) < 0)
J
Jiri Denemark 已提交
4837
        goto cleanup;
4838
    rv = virCommandRun(cmd, NULL);
4839
    virSecurityManagerPostFork(driver->securityManager);
4840

E
Eric Blake 已提交
4841
    /* wait for qemu process to show up */
4842
    if (rv == 0) {
4843
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
4844 4845
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
4846
            rv = -1;
4847
        }
4848 4849 4850 4851 4852
        VIR_DEBUG("QEMU vm=%p name=%s running with pid=%llu",
                  vm, vm->def->name, (unsigned long long)vm->pid);
    } else {
        VIR_DEBUG("QEMU vm=%p name=%s failed to spawn",
                  vm, vm->def->name);
4853 4854
    }

4855
    VIR_DEBUG("Writing early domain status to disk");
4856
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
J
Jiri Denemark 已提交
4857
        goto cleanup;
4858

4859 4860
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
4861
        /* Read errors from child that occurred between fork and exec. */
4862 4863
        qemuProcessReportLogError(logCtxt,
                                  _("Process exited prior to exec"));
J
Jiri Denemark 已提交
4864
        goto cleanup;
4865 4866
    }

4867
    VIR_DEBUG("Setting up domain cgroup (if required)");
4868
    if (qemuSetupCgroup(driver, vm, nnicindexes, nicindexes) < 0)
J
Jiri Denemark 已提交
4869
        goto cleanup;
4870

4871 4872 4873 4874 4875 4876
    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
    if (!vm->def->cputune.emulatorpin &&
        qemuProcessInitCpuAffinity(vm) < 0)
        goto cleanup;

4877 4878
    VIR_DEBUG("Setting domain security labels");
    if (virSecurityManagerSetAllLabel(driver->securityManager,
J
Jiri Denemark 已提交
4879 4880 4881
                                      vm->def,
                                      incoming ? incoming->path : NULL) < 0)
        goto cleanup;
4882

4883
    /* Security manager labeled all devices, therefore
J
Jiri Denemark 已提交
4884 4885 4886 4887
     * if any operation from now on fails, we need to ask the caller to
     * restore labels.
     */
    ret = -2;
4888

J
Jiri Denemark 已提交
4889
    if (incoming && incoming->fd != -1) {
4890 4891 4892 4893 4894 4895 4896
        /* if there's an fd to migrate from, and it's a pipe, put the
         * proper security label on it
         */
        struct stat stdin_sb;

        VIR_DEBUG("setting security label on pipe used for migration");

J
Jiri Denemark 已提交
4897
        if (fstat(incoming->fd, &stdin_sb) < 0) {
4898
            virReportSystemError(errno,
J
Jiri Denemark 已提交
4899 4900
                                 _("cannot stat fd %d"), incoming->fd);
            goto cleanup;
4901 4902
        }
        if (S_ISFIFO(stdin_sb.st_mode) &&
J
Jiri Denemark 已提交
4903 4904 4905
            virSecurityManagerSetImageFDLabel(driver->securityManager,
                                              vm->def, incoming->fd) < 0)
            goto cleanup;
4906 4907 4908
    }

    VIR_DEBUG("Labelling done, completing handshake to child");
4909
    if (virCommandHandshakeNotify(cmd) < 0)
J
Jiri Denemark 已提交
4910
        goto cleanup;
4911 4912
    VIR_DEBUG("Handshake complete, child running");

4913
    if (rv == -1) /* The VM failed to start; tear filters before taps */
4914 4915
        virDomainConfVMNWFilterTeardown(vm);

4916
    if (rv == -1) /* The VM failed to start */
J
Jiri Denemark 已提交
4917
        goto cleanup;
4918

4919
    VIR_DEBUG("Setting cgroup for emulator (if required)");
4920
    if (qemuSetupCgroupForEmulator(vm) < 0)
J
Jiri Denemark 已提交
4921
        goto cleanup;
4922 4923 4924

    VIR_DEBUG("Setting affinity of emulator threads");
    if (qemuProcessSetEmulatorAffinity(vm) < 0)
J
Jiri Denemark 已提交
4925
        goto cleanup;
4926

4927
    VIR_DEBUG("Waiting for monitor to show up");
4928
    if (qemuProcessWaitForMonitor(driver, vm, asyncJob, priv->qemuCaps, logCtxt) < 0)
J
Jiri Denemark 已提交
4929
        goto cleanup;
4930

D
Daniel P. Berrange 已提交
4931
    /* Failure to connect to agent shouldn't be fatal */
4932 4933
    if ((rv = qemuConnectAgent(driver, vm)) < 0) {
        if (rv == -2)
J
Jiri Denemark 已提交
4934
            goto cleanup;
4935

D
Daniel P. Berrange 已提交
4936 4937 4938 4939 4940 4941
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

4942
    VIR_DEBUG("Detecting if required emulator features are present");
4943
    if (!qemuProcessVerifyGuestCPU(driver, vm, asyncJob))
J
Jiri Denemark 已提交
4944
        goto cleanup;
4945

4946
    VIR_DEBUG("Setting up post-init cgroup restrictions");
4947
    if (qemuSetupCpusetMems(vm) < 0)
J
Jiri Denemark 已提交
4948
        goto cleanup;
4949

4950
    VIR_DEBUG("Detecting VCPU PIDs");
4951
    if (qemuDomainDetectVcpuPids(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4952
        goto cleanup;
4953

4954 4955
    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4956
        goto cleanup;
4957

4958
    VIR_DEBUG("Setting cgroup for each VCPU (if required)");
4959
    if (qemuSetupCgroupForVcpu(vm) < 0)
J
Jiri Denemark 已提交
4960
        goto cleanup;
4961

4962 4963
    VIR_DEBUG("Setting cgroup for each IOThread (if required)");
    if (qemuSetupCgroupForIOThreads(vm) < 0)
J
Jiri Denemark 已提交
4964
        goto cleanup;
4965

4966
    VIR_DEBUG("Setting VCPU affinities");
4967
    if (qemuProcessSetVcpuAffinities(vm) < 0)
J
Jiri Denemark 已提交
4968
        goto cleanup;
4969

4970 4971
    VIR_DEBUG("Setting affinity of IOThread threads");
    if (qemuProcessSetIOThreadsAffinity(vm) < 0)
J
Jiri Denemark 已提交
4972
        goto cleanup;
4973

4974 4975
    VIR_DEBUG("Setting scheduler parameters");
    if (qemuProcessSetSchedulers(vm) < 0)
J
Jiri Denemark 已提交
4976
        goto cleanup;
4977

4978
    VIR_DEBUG("Setting any required VM passwords");
4979
    if (qemuProcessInitPasswords(conn, driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4980
        goto cleanup;
4981 4982 4983

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
4984
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4985
        VIR_DEBUG("Determining domain device PCI addresses");
4986
        if (qemuProcessInitPCIAddresses(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4987
            goto cleanup;
4988 4989
    }

4990 4991 4992 4993
    /* set default link states */
    /* qemu doesn't support setting this on the command line, so
     * enter the monitor */
    VIR_DEBUG("Setting network link states");
4994
    if (qemuProcessSetLinkStates(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4995
        goto cleanup;
4996

4997
    VIR_DEBUG("Fetching list of active devices");
4998
    if (qemuDomainUpdateDeviceList(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4999
        goto cleanup;
5000

5001 5002
    VIR_DEBUG("Updating info of memory devices");
    if (qemuDomainUpdateMemoryDeviceInfo(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
5003
        goto cleanup;
5004

5005
    VIR_DEBUG("Setting initial memory amount");
5006
    if (qemuProcessSetupBalloon(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
5007
        goto cleanup;
5008

5009
    /* Since CPUs were not started yet, the balloon could not return the memory
5010 5011
     * to the host and thus cur_balloon needs to be updated so that GetXMLdesc
     * and friends return the correct size in case they can't grab the job */
5012
    if (!incoming && !snapshot &&
5013
        qemuProcessRefreshBalloonState(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
5014
        goto cleanup;
5015

5016 5017
    VIR_DEBUG("Detecting actual memory size for video device");
    if (qemuProcessUpdateVideoRamSize(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
5018 5019 5020 5021 5022 5023 5024 5025 5026 5027
        goto cleanup;

    if (flags & VIR_QEMU_PROCESS_START_AUTODESTROY &&
        qemuProcessAutoDestroyAdd(driver, vm, conn) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    virCommandFree(cmd);
5028
    qemuDomainLogContextFree(logCtxt);
J
Jiri Denemark 已提交
5029 5030 5031 5032 5033 5034 5035 5036
    virObjectUnref(cfg);
    virObjectUnref(caps);
    VIR_FREE(nicindexes);
    VIR_FREE(nodeset);
    return ret;
}


5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083
/**
 * qemuProcessFinishStartup:
 *
 * Finish starting a new domain.
 */
int
qemuProcessFinishStartup(virConnectPtr conn,
                         virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob,
                         bool startCPUs,
                         virDomainPausedReason pausedReason)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;

    if (startCPUs) {
        VIR_DEBUG("Starting domain CPUs");
        if (qemuProcessStartCPUs(driver, vm, conn,
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 asyncJob) < 0) {
            if (!virGetLastError())
                virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                               _("resume operation failed"));
            goto cleanup;
        }
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, pausedReason);
    }

    VIR_DEBUG("Writing domain status to disk");
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
        goto cleanup;

    if (qemuProcessStartHook(driver, vm,
                             VIR_HOOK_QEMU_OP_STARTED,
                             VIR_HOOK_SUBOP_BEGIN) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    virObjectUnref(cfg);
    return ret;
}


J
Jiri Denemark 已提交
5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118
int
qemuProcessStart(virConnectPtr conn,
                 virQEMUDriverPtr driver,
                 virDomainObjPtr vm,
                 qemuDomainAsyncJob asyncJob,
                 const char *migrateFrom,
                 int migrateFd,
                 const char *migratePath,
                 virDomainSnapshotObjPtr snapshot,
                 virNetDevVPortProfileOp vmop,
                 unsigned int flags)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuProcessIncomingDefPtr incoming = NULL;
    unsigned int stopFlags;
    bool relabel = false;
    int ret = -1;
    int rv;

    VIR_DEBUG("conn=%p driver=%p vm=%p name=%s id=%d asyncJob=%s "
              "migrateFrom=%s migrateFd=%d migratePath=%s "
              "snapshot=%p vmop=%d flags=0x%x",
              conn, driver, vm, vm->def->name, vm->def->id,
              qemuDomainAsyncJobTypeToString(asyncJob),
              NULLSTR(migrateFrom), migrateFd, NULLSTR(migratePath),
              snapshot, vmop, flags);

    virCheckFlagsGoto(VIR_QEMU_PROCESS_START_COLD |
                      VIR_QEMU_PROCESS_START_PAUSED |
                      VIR_QEMU_PROCESS_START_AUTODESTROY, cleanup);

    if (qemuProcessInit(driver, vm, !!migrateFrom) < 0)
        goto cleanup;

    if (migrateFrom) {
5119
        incoming = qemuProcessIncomingDefNew(priv->qemuCaps, NULL, migrateFrom,
J
Jiri Denemark 已提交
5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131
                                             migrateFd, migratePath);
        if (!incoming)
            goto stop;
    }

    if ((rv = qemuProcessLaunch(conn, driver, vm, asyncJob, incoming,
                                snapshot, vmop, flags)) < 0) {
        if (rv == -1)
            relabel = true;
        goto stop;
    }
    relabel = true;
5132

5133 5134 5135
    if (incoming &&
        incoming->deferredURI &&
        qemuMigrationRunIncoming(driver, vm, incoming->deferredURI, asyncJob) < 0)
J
Jiri Denemark 已提交
5136
        goto stop;
5137

5138 5139 5140 5141 5142
    if (qemuProcessFinishStartup(conn, driver, vm, asyncJob,
                                 !(flags & VIR_QEMU_PROCESS_START_PAUSED),
                                 incoming ?
                                 VIR_DOMAIN_PAUSED_MIGRATION :
                                 VIR_DOMAIN_PAUSED_USER) < 0)
J
Jiri Denemark 已提交
5143
        goto stop;
5144

5145 5146
    /* Keep watching qemu log for errors during incoming migration, otherwise
     * unset reporting errors from qemu log. */
5147
    if (!incoming)
5148
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
5149

5150 5151 5152
    ret = 0;

 cleanup:
5153
    qemuProcessIncomingDefFree(incoming);
5154
    return ret;
5155

J
Jiri Denemark 已提交
5156 5157 5158 5159 5160 5161
 stop:
    stopFlags = 0;
    if (!relabel)
        stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
    if (migrateFrom)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
5162
    if (priv->mon)
5163
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
J
Jiri Denemark 已提交
5164
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, stopFlags);
5165
    goto cleanup;
5166 5167 5168
}


5169
int
5170
qemuProcessKill(virDomainObjPtr vm, unsigned int flags)
5171
{
5172
    int ret;
5173

5174 5175 5176
    VIR_DEBUG("vm=%p name=%s pid=%llu flags=%x",
              vm, vm->def->name,
              (unsigned long long)vm->pid, flags);
5177

5178 5179 5180 5181 5182
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
5183 5184
    }

5185
    if (flags & VIR_QEMU_PROCESS_KILL_NOWAIT) {
5186 5187 5188 5189 5190
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
5191

5192 5193
    ret = virProcessKillPainfully(vm->pid,
                                  !!(flags & VIR_QEMU_PROCESS_KILL_FORCE));
5194

5195
    return ret;
5196 5197 5198
}


5199
void qemuProcessStop(virQEMUDriverPtr driver,
5200
                     virDomainObjPtr vm,
5201 5202
                     virDomainShutoffReason reason,
                     unsigned int flags)
5203 5204 5205 5206 5207 5208
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
5209
    virNetDevVPortProfilePtr vport = NULL;
5210
    size_t i;
5211
    char *timestamp;
5212
    char *tmppath = NULL;
5213
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5214
    qemuDomainLogContextPtr logCtxt = NULL;
5215

5216 5217 5218
    VIR_DEBUG("Shutting down vm=%p name=%s id=%d pid=%llu flags=%x",
              vm, vm->def->name, vm->def->id,
              (unsigned long long)vm->pid, flags);
5219 5220 5221

    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("VM '%s' not active", vm->def->name);
5222
        virObjectUnref(cfg);
5223 5224 5225
        return;
    }

5226 5227 5228 5229
    /* This method is routinely used in clean up paths. Disable error
     * reporting so we don't squash a legit error. */
    orig_err = virSaveLastError();

5230
    /*
5231 5232
     * We may unlock the vm in qemuProcessKill(), and another thread
     * can lock the vm, and then call qemuProcessStop(). So we should
5233 5234 5235 5236
     * set vm->def->id to -1 here to avoid qemuProcessStop() to be called twice.
     */
    vm->def->id = -1;

5237
    if (virAtomicIntDecAndTest(&driver->nactive) && driver->inhibitCallback)
5238 5239
        driver->inhibitCallback(false, driver->inhibitOpaque);

5240 5241
    /* Wake up anything waiting on domain condition */
    virDomainObjBroadcast(vm);
5242

5243 5244
    if ((logCtxt = qemuDomainLogContextNew(driver, vm,
                                           QEMU_DOMAIN_LOG_CONTEXT_MODE_STOP))) {
5245
        if ((timestamp = virTimeStringNow()) != NULL) {
5246
            qemuDomainLogContextWrite(logCtxt, "%s: shutting down\n", timestamp);
5247 5248
            VIR_FREE(timestamp);
        }
5249
        qemuDomainLogContextFree(logCtxt);
5250 5251
    }

5252 5253 5254
    /* Clear network bandwidth */
    virDomainClearNetBandwidth(vm);

5255 5256
    virDomainConfVMNWFilterTeardown(vm);

5257
    if (cfg->macFilter) {
5258
        def = vm->def;
5259
        for (i = 0; i < def->nnets; i++) {
5260 5261 5262
            virDomainNetDefPtr net = def->nets[i];
            if (net->ifname == NULL)
                continue;
5263 5264 5265
            ignore_value(ebtablesRemoveForwardAllowIn(driver->ebtables,
                                                      net->ifname,
                                                      &net->mac));
5266 5267 5268
        }
    }

5269
    virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
5270
    priv->nbdPort = 0;
5271

D
Daniel P. Berrange 已提交
5272 5273 5274 5275 5276 5277
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
        priv->agentError = false;
    }

5278
    if (priv->mon) {
5279
        qemuMonitorClose(priv->mon);
5280 5281
        priv->mon = NULL;
    }
5282 5283 5284 5285 5286 5287 5288 5289

    if (priv->monConfig) {
        if (priv->monConfig->type == VIR_DOMAIN_CHR_TYPE_UNIX)
            unlink(priv->monConfig->data.nix.path);
        virDomainChrSourceDefFree(priv->monConfig);
        priv->monConfig = NULL;
    }

5290 5291
    ignore_value(virAsprintf(&tmppath, "%s/domain-%s",
                             cfg->libDir, vm->def->name));
5292
    virFileDeleteTree(tmppath);
5293 5294 5295 5296
    VIR_FREE(tmppath);

    ignore_value(virAsprintf(&tmppath, "%s/domain-%s",
                             cfg->channelTargetDir, vm->def->name));
5297
    virFileDeleteTree(tmppath);
5298 5299
    VIR_FREE(tmppath);

5300 5301 5302 5303 5304 5305
    ignore_value(virDomainChrDefForeach(vm->def,
                                        false,
                                        qemuProcessCleanupChardevDevice,
                                        NULL));


5306
    /* shut it off for sure */
5307 5308 5309
    ignore_value(qemuProcessKill(vm,
                                 VIR_QEMU_PROCESS_KILL_FORCE|
                                 VIR_QEMU_PROCESS_KILL_NOCHECK));
5310

5311 5312
    qemuDomainCleanupRun(driver, vm);

5313
    /* Stop autodestroy in case guest is restarted */
5314
    qemuProcessAutoDestroyRemove(driver, vm);
5315

5316 5317
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5318
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5319 5320

        /* we can't stop the operation even if the script raised an error */
5321 5322 5323
        ignore_value(virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
                                 VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                                 NULL, xml, NULL));
5324 5325 5326
        VIR_FREE(xml);
    }

5327 5328 5329 5330
    /* Reset Security Labels unless caller don't want us to */
    if (!(flags & VIR_QEMU_PROCESS_STOP_NO_RELABEL))
        virSecurityManagerRestoreAllLabel(driver->securityManager,
                                          vm->def,
5331
                                          !!(flags & VIR_QEMU_PROCESS_STOP_MIGRATED));
5332
    virSecurityManagerReleaseLabel(driver->securityManager, vm->def);
5333

5334
    for (i = 0; i < vm->def->ndisks; i++) {
5335
        virDomainDeviceDef dev;
5336
        virDomainDiskDefPtr disk = vm->def->disks[i];
5337 5338 5339 5340

        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        ignore_value(qemuRemoveSharedDevice(driver, &dev, vm->def->name));
5341 5342
    }

5343
    /* Clear out dynamically assigned labels */
5344
    for (i = 0; i < vm->def->nseclabels; i++) {
5345
        if (vm->def->seclabels[i]->type == VIR_DOMAIN_SECLABEL_DYNAMIC)
5346 5347
            VIR_FREE(vm->def->seclabels[i]->label);
        VIR_FREE(vm->def->seclabels[i]->imagelabel);
5348 5349
    }

5350 5351 5352
    virStringFreeList(priv->qemuDevices);
    priv->qemuDevices = NULL;

5353 5354 5355
    virDomainDefClearDeviceAliases(vm->def);
    if (!priv->persistentAddrs) {
        virDomainDefClearPCIAddresses(vm->def);
5356
        virDomainPCIAddressSetFree(priv->pciaddrs);
5357
        priv->pciaddrs = NULL;
5358
        virDomainDefClearCCWAddresses(vm->def);
J
Ján Tomko 已提交
5359
        virDomainCCWAddressSetFree(priv->ccwaddrs);
5360
        priv->ccwaddrs = NULL;
5361 5362
        virDomainVirtioSerialAddrSetFree(priv->vioserialaddrs);
        priv->vioserialaddrs = NULL;
5363 5364
    }

5365
    qemuHostdevReAttachDomainDevices(driver, vm->def);
5366 5367 5368 5369

    def = vm->def;
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
5370 5371 5372 5373
        vport = virDomainNetGetActualVirtPortProfile(net);

        switch (virDomainNetGetActualType(net)) {
        case VIR_DOMAIN_NET_TYPE_DIRECT:
5374
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
5375
                             net->ifname, &net->mac,
5376 5377
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
5378
                             virDomainNetGetActualVirtPortProfile(net),
5379
                             cfg->stateDir));
5380 5381 5382 5383 5384
            break;
        case VIR_DOMAIN_NET_TYPE_BRIDGE:
        case VIR_DOMAIN_NET_TYPE_NETWORK:
#ifdef VIR_NETDEV_TAP_REQUIRE_MANUAL_CLEANUP
            if (!(vport && vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH))
5385
                ignore_value(virNetDevTapDelete(net->ifname, net->backend.tap));
5386 5387
#endif
            break;
5388
        }
5389 5390 5391
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
5392 5393 5394 5395 5396 5397 5398 5399 5400
        if (vport) {
            if (vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_MIDONET) {
                ignore_value(virNetDevMidonetUnbindPort(vport));
            } else if (vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
                ignore_value(virNetDevOpenvswitchRemovePort(
                                 virDomainNetGetActualBridgeName(net),
                                 net->ifname));
            }
        }
A
Ansis Atteka 已提交
5401

5402 5403
        /* kick the device out of the hostdev list too */
        virDomainNetRemoveHostdev(def, net);
5404
        networkReleaseActualDevice(vm->def, net);
5405
    }
5406

5407
 retry:
5408
    if ((ret = qemuRemoveCgroup(driver, vm)) < 0) {
5409 5410 5411 5412 5413 5414 5415
        if (ret == -EBUSY && (retries++ < 5)) {
            usleep(200*1000);
            goto retry;
        }
        VIR_WARN("Failed to remove cgroup for %s",
                 vm->def->name);
    }
5416
    virCgroupFree(&priv->cgroup);
5417 5418 5419

    qemuProcessRemoveDomainStatus(driver, vm);

5420 5421
    /* Remove VNC and Spice ports from port reservation bitmap, but only if
       they were reserved by the driver (autoport=yes)
5422
    */
5423
    for (i = 0; i < vm->def->ngraphics; ++i) {
5424
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
5425 5426
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (graphics->data.vnc.autoport) {
5427 5428
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.vnc.port);
5429
            } else if (graphics->data.vnc.portReserved) {
5430 5431 5432 5433 5434
                virPortAllocatorSetUsed(driver->remotePorts,
                                        graphics->data.spice.port,
                                        false);
                graphics->data.vnc.portReserved = false;
            }
5435 5436
            virPortAllocatorRelease(driver->webSocketPorts,
                                    graphics->data.vnc.websocket);
5437
        }
5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (graphics->data.spice.autoport) {
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.spice.port);
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.spice.tlsPort);
            } else {
                if (graphics->data.spice.portReserved) {
                    virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.port,
                                            false);
                    graphics->data.spice.portReserved = false;
                }

                if (graphics->data.spice.tlsPortReserved) {
                    virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.tlsPort,
                                            false);
                    graphics->data.spice.tlsPortReserved = false;
                }
            }
5459
        }
5460 5461
    }

5462
    vm->taint = 0;
5463
    vm->pid = -1;
J
Jiri Denemark 已提交
5464
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
5465 5466
    VIR_FREE(priv->vcpupids);
    priv->nvcpupids = 0;
5467 5468
    for (i = 0; i < vm->def->niothreadids; i++)
        vm->def->iothreadids[i]->thread_id = 0;
5469 5470
    virObjectUnref(priv->qemuCaps);
    priv->qemuCaps = NULL;
5471
    VIR_FREE(priv->pidfile);
5472

5473
    /* The "release" hook cleans up additional resources */
5474
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5475
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5476 5477 5478

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
5479 5480
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
5481 5482 5483
        VIR_FREE(xml);
    }

5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494
    if (vm->newDef) {
        virDomainDefFree(vm->def);
        vm->def = vm->newDef;
        vm->def->id = -1;
        vm->newDef = NULL;
    }

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5495
    virObjectUnref(cfg);
5496
}
5497 5498


5499
int qemuProcessAttach(virConnectPtr conn ATTRIBUTE_UNUSED,
5500
                      virQEMUDriverPtr driver,
5501
                      virDomainObjPtr vm,
5502
                      pid_t pid,
5503 5504 5505 5506
                      const char *pidfile,
                      virDomainChrSourceDefPtr monConfig,
                      bool monJSON)
{
5507
    size_t i;
5508
    qemuDomainLogContextPtr logCtxt = NULL;
5509 5510 5511
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool running = true;
5512
    virDomainPausedReason reason;
5513
    virSecurityLabelPtr seclabel = NULL;
5514
    virSecurityLabelDefPtr seclabeldef = NULL;
5515
    bool seclabelgen = false;
5516 5517
    virSecurityManagerPtr* sec_managers = NULL;
    const char *model;
5518
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5519
    virCapsPtr caps = NULL;
5520
    bool active = false;
5521
    int ret;
5522 5523 5524 5525

    VIR_DEBUG("Beginning VM attach process");

    if (virDomainObjIsActive(vm)) {
5526 5527
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
5528
        virObjectUnref(cfg);
5529 5530 5531
        return -1;
    }

5532
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
5533
        goto error;
5534

5535 5536 5537 5538 5539
    /* Do this upfront, so any part of the startup process can add
     * runtime state to vm->def that won't be persisted. This let's us
     * report implicit runtime defaults in the XML, like vnc listen/socket
     */
    VIR_DEBUG("Setting current domain def as transient");
5540
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm, true) < 0)
5541
        goto error;
5542

5543
    vm->def->id = qemuDriverAllocateID(driver);
5544

5545
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
5546
        driver->inhibitCallback(true, driver->inhibitOpaque);
5547
    active = true;
5548

5549
    if (virFileMakePath(cfg->logDir) < 0) {
5550 5551
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
5552
                             cfg->logDir);
5553
        goto error;
5554 5555 5556
    }

    VIR_FREE(priv->pidfile);
5557
    if (VIR_STRDUP(priv->pidfile, pidfile) < 0)
5558
        goto error;
5559

5560 5561
    vm->pid = pid;

5562
    VIR_DEBUG("Detect security driver config");
5563
    sec_managers = virSecurityManagerGetNested(driver->securityManager);
5564 5565
    if (sec_managers == NULL)
        goto error;
5566 5567

    for (i = 0; sec_managers[i]; i++) {
5568
        seclabelgen = false;
5569 5570
        model = virSecurityManagerGetModel(sec_managers[i]);
        seclabeldef = virDomainDefGetSecurityLabelDef(vm->def, model);
5571
        if (seclabeldef == NULL) {
5572
            if (!(seclabeldef = virSecurityLabelDefNew(model)))
5573 5574 5575
                goto error;
            seclabelgen = true;
        }
5576 5577
        seclabeldef->type = VIR_DOMAIN_SECLABEL_STATIC;
        if (VIR_ALLOC(seclabel) < 0)
5578
            goto error;
5579
        if (virSecurityManagerGetProcessLabel(sec_managers[i],
5580
                                              vm->def, vm->pid, seclabel) < 0)
5581
            goto error;
5582

5583
        if (VIR_STRDUP(seclabeldef->model, model) < 0)
5584
            goto error;
5585

5586
        if (VIR_STRDUP(seclabeldef->label, seclabel->label) < 0)
5587
            goto error;
5588
        VIR_FREE(seclabel);
5589 5590 5591 5592 5593 5594

        if (seclabelgen) {
            if (VIR_APPEND_ELEMENT(vm->def->seclabels, vm->def->nseclabels, seclabeldef) < 0)
                goto error;
            seclabelgen = false;
        }
5595
    }
5596

5597 5598
    if (virSecurityManagerCheckAllLabel(driver->securityManager, vm->def) < 0)
        goto error;
5599 5600 5601
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0)
        goto error;

5602
    VIR_DEBUG("Creating domain log file");
5603 5604
    if (!(logCtxt = qemuDomainLogContextNew(driver, vm,
                                            QEMU_DOMAIN_LOG_CONTEXT_MODE_ATTACH)))
5605
        goto error;
5606 5607

    VIR_DEBUG("Determining emulator version");
5608 5609
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
5610 5611
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
5612
        goto error;
5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624

    VIR_DEBUG("Preparing monitor state");
    priv->monConfig = monConfig;
    monConfig = NULL;
    priv->monJSON = monJSON;

    priv->gotShutdown = false;

    /*
     * Normally PCI addresses are assigned in the virDomainCreate
     * or virDomainDefine methods. We might still need to assign
     * some here to cope with the question of upgrades. Regardless
M
Martin Kletzander 已提交
5625
     * we also need to populate the PCI address set cache for later
5626 5627
     * use in hotplug
     */
5628
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
5629
        VIR_DEBUG("Assigning domain PCI addresses");
5630
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
5631
            goto error;
5632 5633
    }

5634
    if ((timestamp = virTimeStringNow()) == NULL)
5635
        goto error;
5636

5637
    qemuDomainLogContextWrite(logCtxt, "%s: attaching\n", timestamp);
5638
    VIR_FREE(timestamp);
5639

5640
    qemuDomainObjTaint(driver, vm, VIR_DOMAIN_TAINT_EXTERNAL_LAUNCH, logCtxt);
5641 5642

    VIR_DEBUG("Waiting for monitor to show up");
5643
    if (qemuProcessWaitForMonitor(driver, vm, QEMU_ASYNC_JOB_NONE, priv->qemuCaps, NULL) < 0)
5644
        goto error;
5645

D
Daniel P. Berrange 已提交
5646
    /* Failure to connect to agent shouldn't be fatal */
5647 5648 5649 5650
    if ((ret = qemuConnectAgent(driver, vm)) < 0) {
        if (ret == -2)
            goto error;

D
Daniel P. Berrange 已提交
5651 5652 5653 5654 5655 5656
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

5657
    VIR_DEBUG("Detecting VCPU PIDs");
5658
    if (qemuDomainDetectVcpuPids(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5659 5660 5661 5662
        goto error;

    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5663
        goto error;
5664 5665 5666

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
5667
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
5668
        VIR_DEBUG("Determining domain device PCI addresses");
5669
        if (qemuProcessInitPCIAddresses(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5670
            goto error;
5671 5672 5673
    }

    VIR_DEBUG("Getting initial memory amount");
5674
    qemuDomainObjEnterMonitor(driver, vm);
5675 5676 5677 5678 5679 5680 5681
    if (qemuMonitorGetBalloonInfo(priv->mon, &vm->def->mem.cur_balloon) < 0)
        goto exit_monitor;
    if (qemuMonitorGetStatus(priv->mon, &running, &reason) < 0)
        goto exit_monitor;
    if (qemuMonitorGetVirtType(priv->mon, &vm->def->virtType) < 0)
        goto exit_monitor;
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5682
        goto error;
5683

5684
    if (running) {
5685 5686
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                             VIR_DOMAIN_RUNNING_UNPAUSED);
5687 5688 5689
        if (vm->def->memballoon &&
            vm->def->memballoon->model == VIR_DOMAIN_MEMBALLOON_MODEL_VIRTIO &&
            vm->def->memballoon->period) {
5690 5691 5692
            qemuDomainObjEnterMonitor(driver, vm);
            qemuMonitorSetMemoryStatsPeriod(priv->mon,
                                            vm->def->memballoon->period);
5693 5694
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                goto error;
5695 5696
        }
    } else {
5697
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
5698
    }
5699 5700

    VIR_DEBUG("Writing domain status to disk");
5701
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
5702
        goto error;
5703

5704 5705
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5706
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
                              VIR_HOOK_QEMU_OP_ATTACH, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
        VIR_FREE(xml);

        /*
         * If the script raised an error abort the launch
         */
        if (hookret < 0)
5718
            goto error;
5719 5720
    }

5721
    qemuDomainLogContextFree(logCtxt);
5722
    VIR_FREE(seclabel);
5723
    VIR_FREE(sec_managers);
5724
    virObjectUnref(cfg);
5725
    virObjectUnref(caps);
5726 5727 5728

    return 0;

5729 5730
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
5731
 error:
5732 5733 5734 5735 5736 5737
    /* We jump here if we failed to attach to the VM for any reason.
     * Leave the domain running, but pretend we never attempted to
     * attach to it.  */
    if (active && virAtomicIntDecAndTest(&driver->nactive) &&
        driver->inhibitCallback)
        driver->inhibitCallback(false, driver->inhibitOpaque);
5738
    qemuDomainLogContextFree(logCtxt);
5739
    VIR_FREE(seclabel);
5740
    VIR_FREE(sec_managers);
5741 5742
    if (seclabelgen)
        virSecurityLabelDefFree(seclabeldef);
5743
    virDomainChrSourceDefFree(monConfig);
5744
    virObjectUnref(cfg);
5745
    virObjectUnref(caps);
5746 5747 5748 5749
    return -1;
}


5750
static virDomainObjPtr
5751 5752 5753
qemuProcessAutoDestroy(virDomainObjPtr dom,
                       virConnectPtr conn,
                       void *opaque)
5754
{
5755
    virQEMUDriverPtr driver = opaque;
5756
    qemuDomainObjPrivatePtr priv = dom->privateData;
5757
    virObjectEventPtr event = NULL;
5758
    unsigned int stopFlags = 0;
5759

5760
    VIR_DEBUG("vm=%s, conn=%p", dom->def->name, conn);
5761

5762 5763
    virObjectRef(dom);

5764 5765 5766
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;

5767 5768
    if (priv->job.asyncJob) {
        VIR_DEBUG("vm=%s has long-term job active, cancelling",
5769
                  dom->def->name);
5770
        qemuDomainObjDiscardAsyncJob(driver, dom);
5771 5772
    }

5773 5774
    if (qemuDomainObjBeginJob(driver, dom,
                              QEMU_JOB_DESTROY) < 0)
5775 5776 5777
        goto cleanup;

    VIR_DEBUG("Killing domain");
5778

5779
    qemuProcessStop(driver, dom, VIR_DOMAIN_SHUTOFF_DESTROYED, stopFlags);
5780

5781
    virDomainAuditStop(dom, "destroyed");
5782
    event = virDomainEventLifecycleNewFromObj(dom,
5783 5784
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_DESTROYED);
5785

5786 5787
    qemuDomainObjEndJob(driver, dom);

5788
    qemuDomainRemoveInactive(driver, dom);
5789

5790
    qemuDomainEventQueue(driver, event);
5791

5792
 cleanup:
5793
    virDomainObjEndAPI(&dom);
5794
    return dom;
5795 5796
}

5797
int qemuProcessAutoDestroyAdd(virQEMUDriverPtr driver,
5798 5799 5800
                              virDomainObjPtr vm,
                              virConnectPtr conn)
{
5801
    VIR_DEBUG("vm=%s, conn=%p", vm->def->name, conn);
5802 5803
    return virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                qemuProcessAutoDestroy);
5804 5805
}

5806
int qemuProcessAutoDestroyRemove(virQEMUDriverPtr driver,
5807 5808
                                 virDomainObjPtr vm)
{
5809
    int ret;
5810
    VIR_DEBUG("vm=%s", vm->def->name);
5811 5812 5813
    ret = virCloseCallbacksUnset(driver->closeCallbacks, vm,
                                 qemuProcessAutoDestroy);
    return ret;
5814
}
5815

5816
bool qemuProcessAutoDestroyActive(virQEMUDriverPtr driver,
5817 5818
                                  virDomainObjPtr vm)
{
5819
    virCloseCallback cb;
5820
    VIR_DEBUG("vm=%s", vm->def->name);
5821
    cb = virCloseCallbacksGet(driver->closeCallbacks, vm, NULL);
5822
    return cb == qemuProcessAutoDestroy;
5823
}