qemu_process.c 243.3 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 36
#include <sys/utsname.h>

37
#include "qemu_process.h"
38
#include "qemu_processpriv.h"
39
#include "qemu_alias.h"
40
#include "qemu_block.h"
41
#include "qemu_domain.h"
42
#include "qemu_domain_address.h"
43 44 45 46 47 48
#include "qemu_cgroup.h"
#include "qemu_capabilities.h"
#include "qemu_monitor.h"
#include "qemu_command.h"
#include "qemu_hostdev.h"
#include "qemu_hotplug.h"
49
#include "qemu_migration.h"
50
#include "qemu_migration_params.h"
51
#include "qemu_interface.h"
52
#include "qemu_security.h"
53
#include "qemu_extdevice.h"
54

55
#include "cpu/cpu.h"
56
#include "datatypes.h"
57
#include "virlog.h"
58
#include "virerror.h"
59
#include "viralloc.h"
60
#include "virhook.h"
E
Eric Blake 已提交
61
#include "virfile.h"
62
#include "virpidfile.h"
63
#include "virhostcpu.h"
64
#include "domain_audit.h"
65
#include "domain_nwfilter.h"
66
#include "locking/domain_lock.h"
67
#include "viruuid.h"
68
#include "virprocess.h"
69
#include "virtime.h"
A
Ansis Atteka 已提交
70
#include "virnetdevtap.h"
71
#include "virnetdevopenvswitch.h"
72
#include "virnetdevmidonet.h"
73
#include "virbitmap.h"
74
#include "viratomic.h"
75
#include "virnuma.h"
76
#include "virstring.h"
77
#include "virhostdev.h"
J
John Ferlan 已提交
78
#include "secret_util.h"
79
#include "configmake.h"
80
#include "nwfilter_conf.h"
81
#include "netdev_bandwidth_conf.h"
82
#include "virresctrl.h"
83
#include "virvsock.h"
84 85 86

#define VIR_FROM_THIS VIR_FROM_QEMU

87 88
VIR_LOG_INIT("qemu.qemu_process");

89
/**
90
 * qemuProcessRemoveDomainStatus
91 92 93 94 95 96
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
97
qemuProcessRemoveDomainStatus(virQEMUDriverPtr driver,
98 99 100 101
                              virDomainObjPtr vm)
{
    char ebuf[1024];
    char *file = NULL;
102
    qemuDomainObjPrivatePtr priv = vm->privateData;
103 104
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;
105

106
    if (virAsprintf(&file, "%s/%s.xml", cfg->stateDir, vm->def->name) < 0)
107
        goto cleanup;
108 109 110 111 112 113

    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);

114 115 116
    if (priv->pidfile &&
        unlink(priv->pidfile) < 0 &&
        errno != ENOENT)
117 118 119
        VIR_WARN("Failed to remove PID file for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));

120
    ret = 0;
121
 cleanup:
122 123
    virObjectUnref(cfg);
    return ret;
124 125 126
}


D
Daniel P. Berrange 已提交
127 128 129 130 131 132 133
/*
 * 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
134
qemuProcessHandleAgentEOF(qemuAgentPtr agent,
D
Daniel P. Berrange 已提交
135 136 137 138 139 140
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

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

141
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
142 143

    priv = vm->privateData;
144 145 146 147 148 149 150 151 152 153 154

    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;
    }

155
    qemuAgentClose(agent);
156
    priv->agent = NULL;
157
    priv->agentError = false;
D
Daniel P. Berrange 已提交
158

159
    virObjectUnlock(vm);
160 161
    return;

162
 unlock:
163 164
    virObjectUnlock(vm);
    return;
D
Daniel P. Berrange 已提交
165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181
}


/*
 * 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);

182
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
183 184 185 186 187

    priv = vm->privateData;

    priv->agentError = true;

188
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
189 190 191 192 193
}

static void qemuProcessHandleAgentDestroy(qemuAgentPtr agent,
                                          virDomainObjPtr vm)
{
194 195
    VIR_DEBUG("Received destroy agent=%p vm=%p", agent, vm);

196
    virObjectUnref(vm);
D
Daniel P. Berrange 已提交
197 198 199 200 201 202 203 204 205 206
}


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


207
int
208
qemuConnectAgent(virQEMUDriverPtr driver, virDomainObjPtr vm)
D
Daniel P. Berrange 已提交
209 210 211
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuAgentPtr agent = NULL;
212
    virDomainChrDefPtr config = qemuFindAgentConfig(vm->def);
D
Daniel P. Berrange 已提交
213 214 215 216

    if (!config)
        return 0;

217 218 219 220 221 222 223 224 225
    if (priv->agent)
        return 0;

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_VSERPORT_CHANGE) &&
        config->state != VIR_DOMAIN_CHR_DEVICE_STATE_CONNECTED) {
        VIR_DEBUG("Deferring connecting to guest agent");
        return 0;
    }

226
    if (qemuSecuritySetDaemonSocketLabel(driver->securityManager, vm->def) < 0) {
D
Daniel P. Berrange 已提交
227 228 229 230 231 232 233
        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 */
234
    virObjectRef(vm);
D
Daniel P. Berrange 已提交
235

236
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
237 238

    agent = qemuAgentOpen(vm,
239
                          config->source,
D
Daniel P. Berrange 已提交
240 241
                          &agentCallbacks);

242
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
243

244 245 246 247 248 249 250
    if (agent == NULL)
        virObjectUnref(vm);

    if (!virDomainObjIsActive(vm)) {
        qemuAgentClose(agent);
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest crashed while connecting to the guest agent"));
251
        return -1;
252 253
    }

254
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0) {
D
Daniel P. Berrange 已提交
255 256
        VIR_ERROR(_("Failed to clear security context for agent for %s"),
                  vm->def->name);
257
        qemuAgentClose(agent);
D
Daniel P. Berrange 已提交
258 259 260 261
        goto cleanup;
    }

    priv->agent = agent;
262
    if (!priv->agent)
D
Daniel P. Berrange 已提交
263 264
        VIR_INFO("Failed to connect agent for %s", vm->def->name);

265
 cleanup:
266 267 268 269 270 271 272
    if (!priv->agent) {
        VIR_WARN("Cannot connect to QEMU guest agent for %s", vm->def->name);
        priv->agentError = true;
        virResetLastError();
    }

    return 0;
D
Daniel P. Berrange 已提交
273 274 275
}


276
/*
277
 * This is a callback registered with a qemuMonitorPtr instance,
278 279 280 281 282
 * 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
283
qemuProcessHandleMonitorEOF(qemuMonitorPtr mon,
284 285
                            virDomainObjPtr vm,
                            void *opaque)
286
{
287
    virQEMUDriverPtr driver = opaque;
288
    qemuDomainObjPrivatePtr priv;
289
    struct qemuProcessEvent *processEvent;
290

291
    virObjectLock(vm);
292

293
    VIR_DEBUG("Received EOF on %p '%s'", vm, vm->def->name);
294

295
    priv = vm->privateData;
296 297
    if (priv->beingDestroyed) {
        VIR_DEBUG("Domain is being destroyed, EOF is expected");
298
        goto cleanup;
299 300
    }

301
    if (VIR_ALLOC(processEvent) < 0)
302
        goto cleanup;
303

304
    processEvent->eventType = QEMU_PROCESS_EVENT_MONITOR_EOF;
305
    processEvent->vm = virObjectRef(vm);
306

307 308
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
309
        qemuProcessEventFree(processEvent);
310
        goto cleanup;
311
    }
312

313 314 315 316
    /* We don't want this EOF handler to be called over and over while the
     * thread is waiting for a job.
     */
    qemuMonitorUnregister(mon);
317

318 319 320 321
    /* We don't want any cleanup from EOF handler (or any other
     * thread) to enter qemu namespace. */
    qemuDomainDestroyNamespace(driver, vm);

322
 cleanup:
323
    virObjectUnlock(vm);
324 325 326 327 328 329 330 331 332 333 334
}


/*
 * 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,
335 336
                              virDomainObjPtr vm,
                              void *opaque)
337
{
338
    virQEMUDriverPtr driver = opaque;
339
    virObjectEventPtr event = NULL;
340 341 342

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

343
    virObjectLock(vm);
344

345
    ((qemuDomainObjPrivatePtr) vm->privateData)->monError = true;
346
    event = virDomainEventControlErrorNewFromObj(vm);
347
    virObjectEventStateQueue(driver->domainEventState, event);
348

349
    virObjectUnlock(vm);
350 351 352
}


353 354 355 356 357 358 359 360 361 362 363 364 365
/**
 * qemuProcessFindDomainDiskByAliasOrQOM:
 * @vm: domain object to search for the disk
 * @alias: -drive or -device alias of the disk
 * @qomid: QOM tree device name
 *
 * Looks up a disk in the domain definition of @vm which either matches the
 * -drive or -device alias used for the backend and frontend respectively or the
 * QOM name. If @alias is empty it's treated as NULL as it's a mandatory field
 * in some cases.
 *
 * Returns a disk from @vm or NULL if it could not be found.
 */
366
virDomainDiskDefPtr
367 368 369
qemuProcessFindDomainDiskByAliasOrQOM(virDomainObjPtr vm,
                                      const char *alias,
                                      const char *qomid)
370
{
371
    size_t i;
372

373 374 375 376 377
    if (alias && *alias == '\0')
        alias = NULL;

    if (alias)
        alias = qemuAliasDiskDriveSkipPrefix(alias);
378 379

    for (i = 0; i < vm->def->ndisks; i++) {
380 381
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
382

383 384
        if ((disk->info.alias && STREQ_NULLABLE(disk->info.alias, alias)) ||
            (diskPriv->qomName && STREQ_NULLABLE(diskPriv->qomName, qomid)))
385 386 387
            return disk;
    }

388
    virReportError(VIR_ERR_INTERNAL_ERROR,
389 390
                   _("no disk found with alias '%s' or id '%s'"),
                   NULLSTR(alias), NULLSTR(qomid));
391 392 393
    return NULL;
}

394

395 396
static int
qemuProcessHandleReset(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
397 398
                       virDomainObjPtr vm,
                       void *opaque)
399
{
400
    virQEMUDriverPtr driver = opaque;
401
    virObjectEventPtr event;
402
    qemuDomainObjPrivatePtr priv;
403
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
M
Michal Privoznik 已提交
404
    int ret = -1;
405

406
    virObjectLock(vm);
407

408
    event = virDomainEventRebootNewFromObj(vm);
409 410 411
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);
412

413
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
414
        VIR_WARN("Failed to save status on vm %s", vm->def->name);
415

416 417
    if (vm->def->onReboot == VIR_DOMAIN_LIFECYCLE_ACTION_DESTROY ||
        vm->def->onReboot == VIR_DOMAIN_LIFECYCLE_ACTION_PRESERVE) {
418

M
Michal Privoznik 已提交
419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434
        if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
            goto cleanup;

        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("Ignoring RESET event from inactive domain %s",
                      vm->def->name);
            goto endjob;
        }

        qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_DESTROYED,
                        QEMU_ASYNC_JOB_NONE, 0);
        virDomainAuditStop(vm, "destroyed");
        qemuDomainRemoveInactive(driver, vm);
     endjob:
        qemuDomainObjEndJob(driver, vm);
    }
435

M
Michal Privoznik 已提交
436 437 438
    ret = 0;
 cleanup:
    virObjectUnlock(vm);
439
    virObjectEventStateQueue(driver->domainEventState, event);
440
    virObjectUnref(cfg);
M
Michal Privoznik 已提交
441
    return ret;
442 443 444
}


445 446 447 448 449 450 451 452 453 454 455 456 457
/*
 * 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)
{
    virDomainObjPtr vm = opaque;
    qemuDomainObjPrivatePtr priv = vm->privateData;
458
    virQEMUDriverPtr driver = priv->driver;
459
    virObjectEventPtr event = NULL;
460
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
461
    virDomainRunningReason reason = VIR_DOMAIN_RUNNING_BOOTED;
462
    int ret = -1, rc;
463

464
    VIR_DEBUG("vm=%p", vm);
465
    virObjectLock(vm);
466
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
467 468 469
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
470 471
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
472 473 474
        goto endjob;
    }

475
    qemuDomainObjEnterMonitor(driver, vm);
476 477 478
    rc = qemuMonitorSystemReset(priv->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0)
479 480
        goto endjob;

481
    if (rc < 0)
482 483
        goto endjob;

484 485 486
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_CRASHED)
        reason = VIR_DOMAIN_RUNNING_CRASHED;

487
    if (qemuProcessStartCPUs(driver, vm,
488
                             reason,
489
                             QEMU_ASYNC_JOB_NONE) < 0) {
490
        if (virGetLastErrorCode() == VIR_ERR_OK)
491 492
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
493 494
        goto endjob;
    }
495
    priv->gotShutdown = false;
496
    event = virDomainEventLifecycleNewFromObj(vm,
497 498 499
                                     VIR_DOMAIN_EVENT_RESUMED,
                                     VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

500
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
501 502 503 504
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

505 506
    ret = 0;

507
 endjob:
508
    qemuDomainObjEndJob(driver, vm);
509

510
 cleanup:
511 512
    if (ret == -1)
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_FORCE));
M
Michal Privoznik 已提交
513
    virDomainObjEndAPI(&vm);
514
    virObjectEventStateQueue(driver->domainEventState, event);
515
    virObjectUnref(cfg);
516 517 518
}


519
void
520
qemuProcessShutdownOrReboot(virQEMUDriverPtr driver,
521
                            virDomainObjPtr vm)
522
{
523 524 525
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
526
        qemuDomainSetFakeReboot(driver, vm, false);
527
        virObjectRef(vm);
528 529 530 531 532
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
533
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
534
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
535
            virObjectUnref(vm);
536 537
        }
    } else {
538
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
539
    }
540
}
541

542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562

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);
563
    virObjectEventStateQueue(driver->domainEventState, event);
564 565 566 567 568

    return 0;
}


569 570
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
571
                          virDomainObjPtr vm,
572
                          virTristateBool guest_initiated,
573
                          void *opaque)
574
{
575
    virQEMUDriverPtr driver = opaque;
576
    qemuDomainObjPrivatePtr priv;
577
    virObjectEventPtr event = NULL;
578
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
579
    int detail = 0;
580

581 582
    VIR_DEBUG("vm=%p", vm);

583
    virObjectLock(vm);
584 585 586 587 588 589

    priv = vm->privateData;
    if (priv->gotShutdown) {
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
590 591 592 593
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
594 595 596 597 598 599 600 601
    }
    priv->gotShutdown = true;

    VIR_DEBUG("Transitioned guest %s to shutdown state",
              vm->def->name);
    virDomainObjSetState(vm,
                         VIR_DOMAIN_SHUTDOWN,
                         VIR_DOMAIN_SHUTDOWN_UNKNOWN);
602 603 604 605 606 607 608 609 610 611

    switch (guest_initiated) {
    case VIR_TRISTATE_BOOL_YES:
        detail = VIR_DOMAIN_EVENT_SHUTDOWN_GUEST;
        break;

    case VIR_TRISTATE_BOOL_NO:
        detail = VIR_DOMAIN_EVENT_SHUTDOWN_HOST;
        break;

612 613
    case VIR_TRISTATE_BOOL_ABSENT:
    case VIR_TRISTATE_BOOL_LAST:
614 615 616 617 618
    default:
        detail = VIR_DOMAIN_EVENT_SHUTDOWN_FINISHED;
        break;
    }

619
    event = virDomainEventLifecycleNewFromObj(vm,
620 621
                                              VIR_DOMAIN_EVENT_SHUTDOWN,
                                              detail);
622

623
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
624 625 626 627
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

628 629 630
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

631 632
    qemuProcessShutdownOrReboot(driver, vm);

633
 unlock:
634
    virObjectUnlock(vm);
635
    virObjectEventStateQueue(driver->domainEventState, event);
636
    virObjectUnref(cfg);
637

638 639 640 641 642 643
    return 0;
}


static int
qemuProcessHandleStop(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
644 645
                      virDomainObjPtr vm,
                      void *opaque)
646
{
647
    virQEMUDriverPtr driver = opaque;
648
    virObjectEventPtr event = NULL;
649 650
    virDomainPausedReason reason = VIR_DOMAIN_PAUSED_UNKNOWN;
    virDomainEventSuspendedDetailType detail = VIR_DOMAIN_EVENT_SUSPENDED_PAUSED;
651
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
652

653
    virObjectLock(vm);
J
Jiri Denemark 已提交
654
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
655
        qemuDomainObjPrivatePtr priv = vm->privateData;
656

657
        if (priv->gotShutdown) {
658 659
            VIR_DEBUG("Ignoring STOP event after SHUTDOWN");
            goto unlock;
660 661
        }

662
        if (priv->job.asyncJob == QEMU_ASYNC_JOB_MIGRATION_OUT) {
663 664
            if (priv->job.current->status ==
                        QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
665 666 667 668 669 670
                reason = VIR_DOMAIN_PAUSED_POSTCOPY;
                detail = VIR_DOMAIN_EVENT_SUSPENDED_POSTCOPY;
            } else {
                reason = VIR_DOMAIN_PAUSED_MIGRATION;
                detail = VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED;
            }
671 672 673 674
        }

        VIR_DEBUG("Transitioned guest %s to paused state, reason %s",
                  vm->def->name, virDomainPausedReasonTypeToString(reason));
675

676 677 678
        if (priv->job.current)
            ignore_value(virTimeMillisNow(&priv->job.current->stopped));

679 680 681 682
        if (priv->signalStop)
            virDomainObjBroadcast(vm);

        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
683
        event = virDomainEventLifecycleNewFromObj(vm,
684 685
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  detail);
686

687 688 689 690 691
        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));

692
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
693 694 695
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
696
    }
697

698
 unlock:
699
    virObjectUnlock(vm);
700
    virObjectEventStateQueue(driver->domainEventState, event);
701
    virObjectUnref(cfg);
702 703 704 705 706

    return 0;
}


707 708
static int
qemuProcessHandleResume(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
709 710
                        virDomainObjPtr vm,
                        void *opaque)
711
{
712
    virQEMUDriverPtr driver = opaque;
713
    virObjectEventPtr event = NULL;
714
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
715

716
    virObjectLock(vm);
717 718 719 720 721 722 723 724 725 726 727 728 729
    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);
730
        event = virDomainEventLifecycleNewFromObj(vm,
731 732 733
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

734
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
735 736 737 738 739
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
    }

740
 unlock:
741
    virObjectUnlock(vm);
742
    virObjectEventStateQueue(driver->domainEventState, event);
743
    virObjectUnref(cfg);
744 745 746
    return 0;
}

747 748 749
static int
qemuProcessHandleRTCChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm,
750 751
                           long long offset,
                           void *opaque)
752
{
753
    virQEMUDriverPtr driver = opaque;
754
    virObjectEventPtr event = NULL;
755
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
756

757
    virObjectLock(vm);
758

759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775
    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;
776
        vm->def->clock.data.variable.adjustment = offset;
777

778
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
779 780 781 782
           VIR_WARN("unable to save domain status with RTC change");
    }

    event = virDomainEventRTCChangeNewFromObj(vm, offset);
783

784
    virObjectUnlock(vm);
785

786
    virObjectEventStateQueue(driver->domainEventState, event);
787
    virObjectUnref(cfg);
788 789 790 791 792 793 794
    return 0;
}


static int
qemuProcessHandleWatchdog(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
795 796
                          int action,
                          void *opaque)
797
{
798
    virQEMUDriverPtr driver = opaque;
799 800
    virObjectEventPtr watchdogEvent = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
801
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
802

803
    virObjectLock(vm);
804 805 806
    watchdogEvent = virDomainEventWatchdogNewFromObj(vm, action);

    if (action == VIR_DOMAIN_EVENT_WATCHDOG_PAUSE &&
J
Jiri Denemark 已提交
807
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
808
        qemuDomainObjPrivatePtr priv = vm->privateData;
809 810
        VIR_DEBUG("Transitioned guest %s to paused state due to watchdog", vm->def->name);

J
Jiri Denemark 已提交
811
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
812
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
813 814 815
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

816 817 818 819 820
        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));

821
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
822 823 824
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
825 826 827
    }

    if (vm->def->watchdog->action == VIR_DOMAIN_WATCHDOG_ACTION_DUMP) {
C
Chen Fan 已提交
828 829 830 831
        struct qemuProcessEvent *processEvent;
        if (VIR_ALLOC(processEvent) == 0) {
            processEvent->eventType = QEMU_PROCESS_EVENT_WATCHDOG;
            processEvent->action = VIR_DOMAIN_WATCHDOG_ACTION_DUMP;
W
Wen Congyang 已提交
832 833 834
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
835
            processEvent->vm = virObjectRef(vm);
C
Chen Fan 已提交
836
            if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
837
                if (!virObjectUnref(vm))
838
                    vm = NULL;
839
                qemuProcessEventFree(processEvent);
W
Wen Congyang 已提交
840
            }
841
        }
842 843
    }

844
    if (vm)
845
        virObjectUnlock(vm);
846 847
    virObjectEventStateQueue(driver->domainEventState, watchdogEvent);
    virObjectEventStateQueue(driver->domainEventState, lifecycleEvent);
848

849
    virObjectUnref(cfg);
850 851 852 853 854 855 856 857
    return 0;
}


static int
qemuProcessHandleIOError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virDomainObjPtr vm,
                         const char *diskAlias,
858
                         const char *nodename,
859
                         int action,
860 861
                         const char *reason,
                         void *opaque)
862
{
863
    virQEMUDriverPtr driver = opaque;
864 865 866
    virObjectEventPtr ioErrorEvent = NULL;
    virObjectEventPtr ioErrorEvent2 = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
867 868 869
    const char *srcPath;
    const char *devAlias;
    virDomainDiskDefPtr disk;
870
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
871

872
    virObjectLock(vm);
873 874 875 876 877 878 879 880 881 882

    if (*diskAlias == '\0')
        diskAlias = NULL;

    if (diskAlias)
        disk = qemuProcessFindDomainDiskByAliasOrQOM(vm, diskAlias, NULL);
    else if (nodename)
        disk = qemuDomainDiskLookupByNodename(vm->def, nodename, NULL, NULL);
    else
        disk = NULL;
883 884

    if (disk) {
885
        srcPath = virDomainDiskGetSource(disk);
886 887 888 889 890 891 892 893 894 895
        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
896
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
897
        qemuDomainObjPrivatePtr priv = vm->privateData;
898 899
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

900 901 902
        if (priv->signalIOError)
            virDomainObjBroadcast(vm);

J
Jiri Denemark 已提交
903
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
904
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
905 906 907
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

908 909 910 911 912
        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));

913
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
914 915
            VIR_WARN("Unable to save status on vm %s after IO error", vm->def->name);
    }
916
    virObjectUnlock(vm);
917

918 919 920
    virObjectEventStateQueue(driver->domainEventState, ioErrorEvent);
    virObjectEventStateQueue(driver->domainEventState, ioErrorEvent2);
    virObjectEventStateQueue(driver->domainEventState, lifecycleEvent);
921
    virObjectUnref(cfg);
922 923 924
    return 0;
}

925 926 927 928 929
static int
qemuProcessHandleBlockJob(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          const char *diskAlias,
                          int type,
930
                          int status,
931
                          const char *error,
932
                          void *opaque)
933
{
934
    virQEMUDriverPtr driver = opaque;
935
    struct qemuProcessEvent *processEvent = NULL;
936
    virDomainDiskDefPtr disk;
937
    qemuDomainDiskPrivatePtr diskPriv;
938
    char *data = NULL;
939

940
    virObjectLock(vm);
941

942 943
    VIR_DEBUG("Block job for device %s (domain: %p,%s) type %d status %d",
              diskAlias, vm, vm->def->name, type, status);
944

945
    if (!(disk = qemuProcessFindDomainDiskByAliasOrQOM(vm, diskAlias, NULL)))
946
        goto error;
947
    diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
948

949
    if (diskPriv->blockJobSync) {
950
        /* We have a SYNC API waiting for this event, dispatch it back */
951 952
        diskPriv->blockJobType = type;
        diskPriv->blockJobStatus = status;
953 954
        VIR_FREE(diskPriv->blockJobError);
        ignore_value(VIR_STRDUP_QUIET(diskPriv->blockJobError, error));
955
        virDomainObjBroadcast(vm);
956 957 958 959 960 961 962 963 964
    } 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;
965
        processEvent->vm = virObjectRef(vm);
966 967
        processEvent->action = type;
        processEvent->status = status;
968

969 970 971 972
        if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
            ignore_value(virObjectUnref(vm));
            goto error;
        }
973 974
    }

975
 cleanup:
976
    virObjectUnlock(vm);
977
    return 0;
978
 error:
979
    qemuProcessEventFree(processEvent);
980
    goto cleanup;
981
}
982

983

984 985 986 987 988 989 990 991 992 993 994 995
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,
996 997
                          const char *saslUsername,
                          void *opaque)
998
{
999
    virQEMUDriverPtr driver = opaque;
1000
    virObjectEventPtr event;
1001 1002 1003
    virDomainEventGraphicsAddressPtr localAddr = NULL;
    virDomainEventGraphicsAddressPtr remoteAddr = NULL;
    virDomainEventGraphicsSubjectPtr subject = NULL;
1004
    size_t i;
1005 1006

    if (VIR_ALLOC(localAddr) < 0)
1007
        goto error;
1008
    localAddr->family = localFamily;
1009 1010 1011
    if (VIR_STRDUP(localAddr->service, localService) < 0 ||
        VIR_STRDUP(localAddr->node, localNode) < 0)
        goto error;
1012 1013

    if (VIR_ALLOC(remoteAddr) < 0)
1014
        goto error;
1015
    remoteAddr->family = remoteFamily;
1016 1017 1018
    if (VIR_STRDUP(remoteAddr->service, remoteService) < 0 ||
        VIR_STRDUP(remoteAddr->node, remoteNode) < 0)
        goto error;
1019 1020

    if (VIR_ALLOC(subject) < 0)
1021
        goto error;
1022 1023
    if (x509dname) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1024
            goto error;
1025
        subject->nidentity++;
1026 1027 1028
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "x509dname") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, x509dname) < 0)
            goto error;
1029 1030 1031
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1032
            goto error;
1033
        subject->nidentity++;
1034 1035 1036
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "saslUsername") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, saslUsername) < 0)
            goto error;
1037 1038
    }

1039
    virObjectLock(vm);
1040
    event = virDomainEventGraphicsNewFromObj(vm, phase, localAddr, remoteAddr, authScheme, subject);
1041
    virObjectUnlock(vm);
1042

1043
    virObjectEventStateQueue(driver->domainEventState, event);
1044 1045 1046

    return 0;

1047
 error:
1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058
    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) {
1059
        for (i = 0; i < subject->nidentity; i++) {
1060 1061 1062 1063 1064 1065 1066 1067 1068 1069
            VIR_FREE(subject->identities[i].type);
            VIR_FREE(subject->identities[i].name);
        }
        VIR_FREE(subject->identities);
        VIR_FREE(subject);
    }

    return -1;
}

1070 1071 1072 1073
static int
qemuProcessHandleTrayChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            const char *devAlias,
1074
                            const char *devid,
1075 1076
                            int reason,
                            void *opaque)
1077
{
1078
    virQEMUDriverPtr driver = opaque;
1079
    virObjectEventPtr event = NULL;
1080
    virDomainDiskDefPtr disk;
1081
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1082

1083
    virObjectLock(vm);
1084
    disk = qemuProcessFindDomainDiskByAliasOrQOM(vm, devAlias, devid);
1085 1086

    if (disk) {
1087
        event = virDomainEventTrayChangeNewFromObj(vm, disk->info.alias, reason);
1088 1089 1090 1091 1092 1093
        /* 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;

1094
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1095 1096 1097
            VIR_WARN("Unable to save status on vm %s after tray moved event",
                     vm->def->name);
        }
1098 1099

        virDomainObjBroadcast(vm);
1100 1101
    }

1102
    virObjectUnlock(vm);
1103
    virObjectEventStateQueue(driver->domainEventState, event);
1104
    virObjectUnref(cfg);
1105 1106 1107
    return 0;
}

O
Osier Yang 已提交
1108 1109
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1110 1111
                          virDomainObjPtr vm,
                          void *opaque)
O
Osier Yang 已提交
1112
{
1113
    virQEMUDriverPtr driver = opaque;
1114 1115
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1116
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1117

1118
    virObjectLock(vm);
O
Osier Yang 已提交
1119 1120
    event = virDomainEventPMWakeupNewFromObj(vm);

1121 1122 1123 1124 1125 1126 1127 1128 1129
    /* 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);
1130
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
1131 1132 1133
                                                  VIR_DOMAIN_EVENT_STARTED,
                                                  VIR_DOMAIN_EVENT_STARTED_WAKEUP);

1134
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1135 1136 1137 1138 1139
            VIR_WARN("Unable to save status on vm %s after wakeup event",
                     vm->def->name);
        }
    }

1140
    virObjectUnlock(vm);
1141 1142
    virObjectEventStateQueue(driver->domainEventState, event);
    virObjectEventStateQueue(driver->domainEventState, lifecycleEvent);
1143
    virObjectUnref(cfg);
O
Osier Yang 已提交
1144 1145
    return 0;
}
1146

O
Osier Yang 已提交
1147 1148
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1149 1150
                           virDomainObjPtr vm,
                           void *opaque)
O
Osier Yang 已提交
1151
{
1152
    virQEMUDriverPtr driver = opaque;
1153 1154
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1155
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1156

1157
    virObjectLock(vm);
O
Osier Yang 已提交
1158 1159
    event = virDomainEventPMSuspendNewFromObj(vm);

1160
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1161
        qemuDomainObjPrivatePtr priv = vm->privateData;
1162 1163 1164 1165 1166
        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
1167
        lifecycleEvent =
1168
            virDomainEventLifecycleNewFromObj(vm,
J
Jiri Denemark 已提交
1169 1170
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1171

1172
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1173 1174 1175
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1176 1177 1178

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1179 1180
    }

1181
    virObjectUnlock(vm);
O
Osier Yang 已提交
1182

1183 1184
    virObjectEventStateQueue(driver->domainEventState, event);
    virObjectEventStateQueue(driver->domainEventState, lifecycleEvent);
1185
    virObjectUnref(cfg);
O
Osier Yang 已提交
1186 1187 1188
    return 0;
}

1189 1190 1191
static int
qemuProcessHandleBalloonChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1192 1193
                               unsigned long long actual,
                               void *opaque)
1194
{
1195
    virQEMUDriverPtr driver = opaque;
1196
    virObjectEventPtr event = NULL;
1197
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1198

1199
    virObjectLock(vm);
1200 1201 1202 1203 1204 1205
    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;

1206
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
1207 1208
        VIR_WARN("unable to save domain status with balloon change");

1209
    virObjectUnlock(vm);
1210

1211
    virObjectEventStateQueue(driver->domainEventState, event);
1212
    virObjectUnref(cfg);
1213 1214 1215
    return 0;
}

1216 1217
static int
qemuProcessHandlePMSuspendDisk(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1218 1219
                               virDomainObjPtr vm,
                               void *opaque)
1220
{
1221
    virQEMUDriverPtr driver = opaque;
1222 1223
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1224
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1225

1226
    virObjectLock(vm);
1227 1228 1229 1230 1231 1232 1233 1234 1235 1236
    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 =
1237
            virDomainEventLifecycleNewFromObj(vm,
1238 1239 1240
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_DISK);

1241
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1242 1243 1244 1245 1246 1247 1248 1249
            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);
    }

1250
    virObjectUnlock(vm);
1251

1252 1253
    virObjectEventStateQueue(driver->domainEventState, event);
    virObjectEventStateQueue(driver->domainEventState, lifecycleEvent);
1254 1255
    virObjectUnref(cfg);

1256 1257 1258
    return 0;
}

1259

1260 1261
static int
qemuProcessHandleGuestPanic(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1262
                            virDomainObjPtr vm,
1263
                            qemuMonitorEventPanicInfoPtr info,
1264
                            void *opaque)
1265
{
1266
    virQEMUDriverPtr driver = opaque;
1267 1268 1269
    struct qemuProcessEvent *processEvent;

    virObjectLock(vm);
1270
    if (VIR_ALLOC(processEvent) < 0)
1271 1272 1273 1274
        goto cleanup;

    processEvent->eventType = QEMU_PROCESS_EVENT_GUESTPANIC;
    processEvent->action = vm->def->onCrash;
1275
    processEvent->data = info;
1276 1277 1278
    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted before handling guest panic event is finished.
     */
1279 1280
    processEvent->vm = virObjectRef(vm);

1281 1282 1283
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        if (!virObjectUnref(vm))
            vm = NULL;
1284
        qemuProcessEventFree(processEvent);
1285 1286
    }

1287
 cleanup:
1288
    if (vm)
1289
        virObjectUnlock(vm);
1290 1291 1292 1293 1294

    return 0;
}


1295
int
1296 1297
qemuProcessHandleDeviceDeleted(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1298 1299
                               const char *devAlias,
                               void *opaque)
1300
{
1301
    virQEMUDriverPtr driver = opaque;
1302 1303
    struct qemuProcessEvent *processEvent = NULL;
    char *data;
1304 1305 1306 1307 1308 1309

    virObjectLock(vm);

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

1310 1311
    if (qemuDomainSignalDeviceRemoval(vm, devAlias,
                                      QEMU_DOMAIN_UNPLUGGING_DEVICE_STATUS_OK))
1312
        goto cleanup;
1313

1314 1315
    if (VIR_ALLOC(processEvent) < 0)
        goto error;
1316

1317 1318 1319 1320
    processEvent->eventType = QEMU_PROCESS_EVENT_DEVICE_DELETED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
1321
    processEvent->vm = virObjectRef(vm);
1322

1323 1324 1325 1326
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }
1327

1328
 cleanup:
1329 1330
    virObjectUnlock(vm);
    return 0;
1331
 error:
1332
    qemuProcessEventFree(processEvent);
1333
    goto cleanup;
1334 1335 1336
}


1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398
/**
 *
 * Meaning of fields reported by the event according to the ACPI standard:
 * @source:
 *  0x00 - 0xff: Notification values, as passed at the request time
 *  0x100: Operating System Shutdown Processing
 *  0x103: Ejection processing
 *  0x200: Insertion processing
 *  other values are reserved
 *
 * @status:
 *   general values
 *     0x00: success
 *     0x01: non-specific failure
 *     0x02: unrecognized notify code
 *     0x03 - 0x7f: reserved
 *     other values are specific to the notification type
 *
 *   for the 0x100 source the following additional codes are standardized
 *     0x80: OS Shutdown request denied
 *     0x81: OS Shutdown in progress
 *     0x82: OS Shutdown completed
 *     0x83: OS Graceful shutdown not supported
 *     other values are reserved
 *
 * Other fields and semantics are specific to the qemu handling of the event.
 *  - @alias may be NULL for successful unplug operations
 *  - @slotType describes the device type a bit more closely, currently the
 *    only known value is 'DIMM'
 *  - @slot describes the specific device
 *
 *  Note that qemu does not emit the event for all the documented sources or
 *  devices.
 */
static int
qemuProcessHandleAcpiOstInfo(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                             virDomainObjPtr vm,
                             const char *alias,
                             const char *slotType,
                             const char *slot,
                             unsigned int source,
                             unsigned int status,
                             void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    virObjectEventPtr event = NULL;

    virObjectLock(vm);

    VIR_DEBUG("ACPI OST info for device %s domain %p %s. "
              "slotType='%s' slot='%s' source=%u status=%u",
              NULLSTR(alias), vm, vm->def->name, slotType, slot, source, status);

    /* handle memory unplug failure */
    if (STREQ(slotType, "DIMM") && alias && status == 1) {
        qemuDomainSignalDeviceRemoval(vm, alias,
                                      QEMU_DOMAIN_UNPLUGGING_DEVICE_STATUS_GUEST_REJECTED);

        event = virDomainEventDeviceRemovalFailedNewFromObj(vm, alias);
    }

    virObjectUnlock(vm);
1399
    virObjectEventStateQueue(driver->domainEventState, event);
1400 1401 1402 1403 1404

    return 0;
}


1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438
static int
qemuProcessHandleBlockThreshold(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                virDomainObjPtr vm,
                                const char *nodename,
                                unsigned long long threshold,
                                unsigned long long excess,
                                void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    virObjectEventPtr event = NULL;
    virDomainDiskDefPtr disk;
    virStorageSourcePtr src;
    unsigned int idx;
    char *dev = NULL;
    const char *path = NULL;

    virObjectLock(vm);

    VIR_DEBUG("BLOCK_WRITE_THRESHOLD event for block node '%s' in domain %p %s:"
              "threshold '%llu' exceeded by '%llu'",
              nodename, vm, vm->def->name, threshold, excess);

    if ((disk = qemuDomainDiskLookupByNodename(vm->def, nodename, &src, &idx))) {
        if (virStorageSourceIsLocalStorage(src))
            path = src->path;

        if ((dev = qemuDomainDiskBackingStoreGetName(disk, src, idx))) {
            event = virDomainEventBlockThresholdNewFromObj(vm, dev, path,
                                                           threshold, excess);
            VIR_FREE(dev);
        }
    }

    virObjectUnlock(vm);
1439
    virObjectEventStateQueue(driver->domainEventState, event);
1440 1441 1442 1443 1444

    return 0;
}


1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466
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;
1467
    processEvent->vm = virObjectRef(vm);
1468 1469 1470 1471 1472 1473 1474 1475 1476 1477

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

 cleanup:
    virObjectUnlock(vm);
    return 0;
 error:
1478
    qemuProcessEventFree(processEvent);
1479 1480 1481 1482
    goto cleanup;
}


1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506
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;
1507
    processEvent->vm = virObjectRef(vm);
1508 1509 1510 1511 1512 1513 1514 1515 1516 1517

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

 cleanup:
    virObjectUnlock(vm);
    return 0;
 error:
1518
    qemuProcessEventFree(processEvent);
1519 1520 1521 1522
    goto cleanup;
}


1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541
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;
1542
    virDomainObjBroadcast(vm);
1543 1544 1545 1546 1547 1548 1549

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564
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;
1565
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_NONE) {
1566 1567 1568 1569
        VIR_DEBUG("got MIGRATION event without a migration job");
        goto cleanup;
    }

1570
    priv->job.current->stats.mig.status = status;
1571 1572 1573 1574 1575 1576 1577 1578
    virDomainObjBroadcast(vm);

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
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;
    }

1599
    virObjectEventStateQueue(driver->domainEventState,
1600 1601 1602 1603 1604 1605 1606 1607
                         virDomainEventMigrationIterationNewFromObj(vm, pass));

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646
static int
qemuProcessHandleDumpCompleted(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               int status,
                               qemuMonitorDumpStatsPtr stats,
                               const char *error,
                               void *opaque ATTRIBUTE_UNUSED)
{
    qemuDomainObjPrivatePtr priv;

    virObjectLock(vm);

    VIR_DEBUG("Dump completed for domain %p %s with stats=%p error='%s'",
              vm, vm->def->name, stats, NULLSTR(error));

    priv = vm->privateData;
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_NONE) {
        VIR_DEBUG("got DUMP_COMPLETED event without a dump_completed job");
        goto cleanup;
    }
    priv->job.dumpCompleted = true;
    priv->job.current->stats.dump = *stats;
    ignore_value(VIR_STRDUP_QUIET(priv->job.error, error));

    /* Force error if extracting the DUMP_COMPLETED status failed */
    if (!error && status < 0) {
        ignore_value(VIR_STRDUP_QUIET(priv->job.error, virGetLastErrorMessage()));
        priv->job.current->stats.dump.status = QEMU_MONITOR_DUMP_STATUS_FAILED;
    }

    virDomainObjBroadcast(vm);

 cleanup:
    virResetLastError();
    virObjectUnlock(vm);
    return 0;
}


1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700
static int
qemuProcessHandlePRManagerStatusChanged(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                        virDomainObjPtr vm,
                                        const char *prManager,
                                        bool connected,
                                        void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    qemuDomainObjPrivatePtr priv;
    struct qemuProcessEvent *processEvent = NULL;
    const char *managedAlias = qemuDomainGetManagedPRAlias();
    int ret = -1;

    virObjectLock(vm);

    VIR_DEBUG("pr-manager %s status changed for domain %p %s connected=%d",
              prManager, vm, vm->def->name, connected);

    if (connected) {
        /* Connect events are boring. */
        ret = 0;
        goto cleanup;
    }
    /* Disconnect events are more interesting. */

    if (STRNEQ(prManager, managedAlias)) {
        VIR_DEBUG("pr-manager %s not managed, ignoring event",
                  prManager);
        ret = 0;
        goto cleanup;
    }

    priv = vm->privateData;
    priv->prDaemonRunning = false;

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

    processEvent->eventType = QEMU_PROCESS_EVENT_PR_DISCONNECT;
    processEvent->vm = virObjectRef(vm);

    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        qemuProcessEventFree(processEvent);
        virObjectUnref(vm);
        goto cleanup;
    }

    ret = 0;
 cleanup:
    virObjectUnlock(vm);
    return ret;
}


1701 1702
static qemuMonitorCallbacks monitorCallbacks = {
    .eofNotify = qemuProcessHandleMonitorEOF,
1703
    .errorNotify = qemuProcessHandleMonitorError,
1704
    .domainEvent = qemuProcessHandleEvent,
1705 1706
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
1707
    .domainResume = qemuProcessHandleResume,
1708 1709 1710 1711 1712
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1713
    .domainBlockJob = qemuProcessHandleBlockJob,
1714
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1715
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1716
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1717
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1718
    .domainPMSuspendDisk = qemuProcessHandlePMSuspendDisk,
1719
    .domainGuestPanic = qemuProcessHandleGuestPanic,
1720
    .domainDeviceDeleted = qemuProcessHandleDeviceDeleted,
1721
    .domainNicRxFilterChanged = qemuProcessHandleNicRxFilterChanged,
1722
    .domainSerialChange = qemuProcessHandleSerialChanged,
1723
    .domainSpiceMigrated = qemuProcessHandleSpiceMigrated,
1724
    .domainMigrationStatus = qemuProcessHandleMigrationStatus,
1725
    .domainMigrationPass = qemuProcessHandleMigrationPass,
1726
    .domainAcpiOstInfo = qemuProcessHandleAcpiOstInfo,
1727
    .domainBlockThreshold = qemuProcessHandleBlockThreshold,
1728
    .domainDumpCompleted = qemuProcessHandleDumpCompleted,
1729
    .domainPRManagerStatusChanged = qemuProcessHandlePRManagerStatusChanged,
1730 1731
};

1732 1733 1734 1735 1736 1737 1738 1739 1740 1741
static void
qemuProcessMonitorReportLogError(qemuMonitorPtr mon,
                                 const char *msg,
                                 void *opaque);


static void
qemuProcessMonitorLogFree(void *opaque)
{
    qemuDomainLogContextPtr logCtxt = opaque;
1742
    virObjectUnref(logCtxt);
1743 1744
}

1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764

static int
qemuProcessInitMonitor(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
                       qemuDomainAsyncJob asyncJob)
{
    int ret;

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

    ret = qemuMonitorSetCapabilities(QEMU_DOMAIN_PRIVATE(vm)->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

    return ret;
}


1765
static int
1766
qemuConnectMonitor(virQEMUDriverPtr driver, virDomainObjPtr vm, int asyncJob,
1767
                   bool retry, qemuDomainLogContextPtr logCtxt)
1768 1769
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1770
    qemuMonitorPtr mon = NULL;
1771
    unsigned long long timeout = 0;
1772
    virDomainChrSourceDefPtr monConfig;
1773

1774
    if (qemuSecuritySetDaemonSocketLabel(driver->securityManager, vm->def) < 0) {
1775 1776
        VIR_ERROR(_("Failed to set security context for monitor for %s"),
                  vm->def->name);
1777
        return -1;
1778 1779
    }

1780 1781 1782 1783 1784 1785
    /* When using hugepages, kernel zeroes them out before
     * handing them over to qemu. This can be very time
     * consuming. Therefore, add a second to timeout for each
     * 1GiB of guest RAM. */
    timeout = vm->def->mem.total_memory / (1024 * 1024);

1786
    /* Hold an extra reference because we can't allow 'vm' to be
C
Chen Hanxiao 已提交
1787
     * deleted until the monitor gets its own reference. */
1788
    virObjectRef(vm);
1789

1790
    ignore_value(virTimeMillisNow(&priv->monStart));
1791 1792
    monConfig = priv->monConfig;
    virObjectRef(monConfig);
1793
    virObjectUnlock(vm);
1794 1795

    mon = qemuMonitorOpen(vm,
1796
                          monConfig,
1797
                          priv->monJSON,
1798
                          retry,
1799
                          timeout,
1800 1801
                          &monitorCallbacks,
                          driver);
1802

1803
    if (mon && logCtxt) {
1804
        virObjectRef(logCtxt);
1805 1806 1807 1808 1809
        qemuMonitorSetDomainLog(mon,
                                qemuProcessMonitorReportLogError,
                                logCtxt,
                                qemuProcessMonitorLogFree);
    }
1810

1811
    virObjectLock(vm);
1812
    virObjectUnref(monConfig);
M
Michal Privoznik 已提交
1813
    virObjectUnref(vm);
1814
    priv->monStart = 0;
1815

M
Michal Privoznik 已提交
1816
    if (!virDomainObjIsActive(vm)) {
1817
        qemuMonitorClose(mon);
1818
        mon = NULL;
1819 1820 1821
    }
    priv->mon = mon;

1822
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0) {
1823 1824
        VIR_ERROR(_("Failed to clear security context for monitor for %s"),
                  vm->def->name);
1825
        return -1;
1826 1827 1828 1829
    }

    if (priv->mon == NULL) {
        VIR_INFO("Failed to connect monitor for %s", vm->def->name);
1830
        return -1;
1831 1832
    }

1833 1834
    if (qemuProcessInitMonitor(driver, vm, asyncJob) < 0)
        return -1;
1835

1836
    if (qemuMigrationCapsCheck(driver, vm, asyncJob) < 0)
1837 1838 1839
        return -1;

    return 0;
1840 1841
}

1842 1843 1844

/**
 * qemuProcessReadLog: Read log file of a qemu VM
1845
 * @logCtxt: the domain log context
1846
 * @msg: pointer to buffer to store the read messages in
1847
 * @max: maximum length of the message returned in @msg
1848 1849
 *
 * Reads log of a qemu VM. Skips messages not produced by qemu or irrelevant
1850 1851 1852 1853
 * messages. If @max is not zero, @msg will contain at most @max characters
 * from the end of the log and @msg will start after a new line if possible.
 *
 * Returns 0 on success or -1 on error
1854
 */
1855
static int
1856 1857 1858
qemuProcessReadLog(qemuDomainLogContextPtr logCtxt,
                   char **msg,
                   size_t max)
1859
{
1860 1861
    char *buf;
    ssize_t got;
1862
    char *eol;
1863
    char *filter_next;
1864
    size_t skip;
1865

1866
    if ((got = qemuDomainLogContextRead(logCtxt, &buf)) < 0)
1867
        return -1;
1868

1869 1870 1871 1872 1873
    /* Filter out debug messages from intermediate libvirt process */
    filter_next = buf;
    while ((eol = strchr(filter_next, '\n'))) {
        *eol = '\0';
        if (virLogProbablyLogMessage(filter_next) ||
1874
            strstr(filter_next, "char device redirected to")) {
1875
            skip = (eol + 1) - filter_next;
1876
            memmove(filter_next, eol + 1, buf + got - eol);
1877 1878 1879 1880
            got -= skip;
        } else {
            filter_next = eol + 1;
            *eol = '\n';
1881 1882
        }
    }
1883
    filter_next = NULL; /* silence false coverity warning */
1884

1885 1886
    if (got > 0 &&
        buf[got - 1] == '\n') {
1887 1888
        buf[got - 1] = '\0';
        got--;
1889
    }
1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902

    if (max > 0 && got > max) {
        skip = got - max;

        if (buf[skip - 1] != '\n' &&
            (eol = strchr(buf + skip, '\n')) &&
            !virStringIsEmpty(eol + 1))
            skip = eol + 1 - buf;

        memmove(buf, buf + skip, got - skip + 1);
        got -= skip;
    }

1903
    ignore_value(VIR_REALLOC_N_QUIET(buf, got + 1));
1904 1905 1906
    *msg = buf;
    return 0;
}
1907 1908


1909 1910
static int
qemuProcessReportLogError(qemuDomainLogContextPtr logCtxt,
1911 1912 1913
                          const char *msgprefix)
{
    char *logmsg = NULL;
1914 1915 1916 1917 1918 1919
    size_t max;

    max = VIR_ERROR_MAX_LENGTH - 1;
    max -= strlen(msgprefix);
    /* The length of the formatting string minus two '%s' */
    max -= strlen(_("%s: %s")) - 4;
1920

1921
    if (qemuProcessReadLog(logCtxt, &logmsg, max) < 0)
1922 1923 1924
        return -1;

    virResetLastError();
1925 1926 1927 1928 1929
    if (virStringIsEmpty(logmsg))
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s", msgprefix);
    else
        virReportError(VIR_ERR_INTERNAL_ERROR, _("%s: %s"), msgprefix, logmsg);

1930 1931
    VIR_FREE(logmsg);
    return 0;
1932 1933 1934
}


1935 1936 1937 1938 1939 1940 1941 1942 1943 1944
static void
qemuProcessMonitorReportLogError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                 const char *msg,
                                 void *opaque)
{
    qemuDomainLogContextPtr logCtxt = opaque;
    qemuProcessReportLogError(logCtxt, msg);
}


1945
static int
1946
qemuProcessLookupPTYs(virDomainChrDefPtr *devices,
1947
                      int count,
1948
                      virHashTablePtr info)
1949
{
1950
    char *id = NULL;
1951
    size_t i;
1952
    int ret = -1;
1953

1954
    for (i = 0; i < count; i++) {
1955
        virDomainChrDefPtr chr = devices[i];
1956
        if (chr->source->type == VIR_DOMAIN_CHR_TYPE_PTY) {
1957
            qemuMonitorChardevInfoPtr entry;
1958

1959 1960
            VIR_FREE(id);
            if (virAsprintf(&id, "char%s", chr->info.alias) < 0)
1961 1962
                return -1;

1963 1964
            entry = virHashLookup(info, id);
            if (!entry || !entry->ptyPath) {
1965
                if (chr->source->data.file.path == NULL) {
1966 1967 1968
                    /* neither the log output nor 'info chardev' had a
                     * pty path for this chardev, report an error
                     */
1969 1970
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1971
                    goto cleanup;
1972 1973 1974 1975 1976 1977 1978 1979
                } else {
                    /* 'info chardev' had no pty path for this chardev,
                     * but the log output had, so we're fine
                     */
                    continue;
                }
            }

1980 1981
            VIR_FREE(chr->source->data.file.path);
            if (VIR_STRDUP(chr->source->data.file.path, entry->ptyPath) < 0)
1982
                goto cleanup;
1983 1984 1985
        }
    }

1986 1987 1988 1989
    ret = 0;
 cleanup:
    VIR_FREE(id);
    return ret;
1990 1991
}

1992 1993
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
1994
                                     virHashTablePtr info)
1995
{
1996
    size_t i = 0;
C
Cole Robinson 已提交
1997

1998
    if (qemuProcessLookupPTYs(vm->def->serials, vm->def->nserials, info) < 0)
1999 2000
        return -1;

2001
    if (qemuProcessLookupPTYs(vm->def->parallels, vm->def->nparallels,
2002
                              info) < 0)
2003
        return -1;
2004

2005
    if (qemuProcessLookupPTYs(vm->def->channels, vm->def->nchannels, info) < 0)
2006
        return -1;
2007 2008 2009 2010
    /* 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];
2011

2012 2013 2014 2015 2016
        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;
2017 2018
            if (virDomainChrSourceDefCopy(chr->source,
                                          ((vm->def->serials[0])->source)) < 0)
2019 2020 2021 2022
                return -1;
        }
    }

2023
    if (qemuProcessLookupPTYs(vm->def->consoles + i, vm->def->nconsoles - i,
2024
                              info) < 0)
2025
        return -1;
2026 2027 2028 2029 2030

    return 0;
}


2031
static int
2032 2033 2034 2035
qemuProcessRefreshChannelVirtioState(virQEMUDriverPtr driver,
                                     virDomainObjPtr vm,
                                     virHashTablePtr info,
                                     int booted)
2036 2037
{
    size_t i;
2038
    int agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_CHANNEL;
2039
    qemuMonitorChardevInfoPtr entry;
2040
    virObjectEventPtr event = NULL;
2041 2042
    char *id = NULL;
    int ret = -1;
2043

2044 2045 2046
    if (booted)
        agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_DOMAIN_STARTED;

2047 2048 2049
    for (i = 0; i < vm->def->nchannels; i++) {
        virDomainChrDefPtr chr = vm->def->channels[i];
        if (chr->targetType == VIR_DOMAIN_CHR_CHANNEL_TARGET_TYPE_VIRTIO) {
2050 2051 2052 2053

            VIR_FREE(id);
            if (virAsprintf(&id, "char%s", chr->info.alias) < 0)
                goto cleanup;
2054 2055 2056 2057 2058 2059

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

2060 2061 2062 2063
            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)))
2064
                virObjectEventStateQueue(driver->domainEventState, event);
2065

2066 2067 2068 2069
            chr->state = entry->state;
        }
    }

2070 2071 2072 2073
    ret = 0;
 cleanup:
    VIR_FREE(id);
    return ret;
2074 2075 2076
}


2077 2078
int
qemuRefreshVirtioChannelState(virQEMUDriverPtr driver,
2079 2080
                              virDomainObjPtr vm,
                              qemuDomainAsyncJob asyncJob)
2081 2082 2083 2084 2085
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virHashTablePtr info = NULL;
    int ret = -1;

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

2089
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
2090 2091
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2092 2093 2094 2095

    if (ret < 0)
        goto cleanup;

2096
    ret = qemuProcessRefreshChannelVirtioState(driver, vm, info, false);
2097 2098 2099 2100 2101 2102

 cleanup:
    virHashFree(info);
    return ret;
}

2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159

static int
qemuProcessRefreshPRManagerState(virDomainObjPtr vm,
                                 virHashTablePtr info)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuMonitorPRManagerInfoPtr prManagerInfo;
    const char *managedAlias = qemuDomainGetManagedPRAlias();
    int ret = -1;

    if (!(prManagerInfo = virHashLookup(info, managedAlias))) {
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("missing info on pr-manager %s"),
                       managedAlias);
        goto cleanup;
    }

    priv->prDaemonRunning = prManagerInfo->connected;

    if (!priv->prDaemonRunning &&
        qemuProcessStartManagedPRDaemon(vm) < 0)
        goto cleanup;

    ret = 0;
 cleanup:
    return ret;
}


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

    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_PR_MANAGER_HELPER) ||
        !virDomainDefHasManagedPR(vm->def))
        return 0;

    qemuDomainObjEnterMonitor(driver, vm);
    ret = qemuMonitorGetPRManagerInfo(priv->mon, &info);
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

    if (ret < 0)
        goto cleanup;

    ret = qemuProcessRefreshPRManagerState(vm, info);

 cleanup:
    virHashFree(info);
    return ret;
}


2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183
static void
qemuRefreshRTC(virQEMUDriverPtr driver,
               virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    time_t now, then;
    struct tm thenbits;
    long localOffset;
    int rv;

    if (vm->def->clock.offset != VIR_DOMAIN_CLOCK_OFFSET_VARIABLE)
        return;

    memset(&thenbits, 0, sizeof(thenbits));
    qemuDomainObjEnterMonitor(driver, vm);
    now = time(NULL);
    rv = qemuMonitorGetRTCTime(priv->mon, &thenbits);
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        rv = -1;

    if (rv < 0)
        return;

    thenbits.tm_isdst = -1;
2184
    if ((then = mktime(&thenbits)) == (time_t)-1) {
2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Unable to convert time"));
        return;
    }

    /* Thing is, @now is in local TZ but @then in UTC. */
    if (virTimeLocalOffsetFromUTC(&localOffset) < 0)
        return;

    vm->def->clock.data.variable.adjustment = then - now + localOffset;
}
2196

2197
int
2198 2199 2200 2201 2202 2203 2204 2205 2206
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 */
2207
    if (!virDomainDefHasMemballoon(vm->def)) {
2208
        vm->def->mem.cur_balloon = virDomainDefGetMemoryTotal(vm->def);
2209 2210 2211 2212 2213 2214 2215
        return 0;
    }

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

    rc = qemuMonitorGetBalloonInfo(qemuDomainGetMonitor(vm), &balloon);
2216
    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
2217 2218 2219 2220 2221 2222 2223 2224
        return -1;

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

    return 0;
}


2225
static int
2226
qemuProcessWaitForMonitor(virQEMUDriverPtr driver,
C
Cole Robinson 已提交
2227
                          virDomainObjPtr vm,
2228
                          int asyncJob,
2229
                          qemuDomainLogContextPtr logCtxt)
2230 2231
{
    int ret = -1;
2232
    virHashTablePtr info = NULL;
2233 2234 2235 2236 2237 2238
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool retry = true;

    if (priv->qemuCaps &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_CHARDEV_FD_PASS))
        retry = false;
2239

2240 2241 2242 2243
    VIR_DEBUG("Connect monitor to vm=%p name='%s' retry=%d",
              vm, vm->def->name, retry);

    if (qemuConnectMonitor(driver, vm, asyncJob, retry, logCtxt) < 0)
2244 2245 2246 2247 2248 2249
        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. */
2250 2251
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
2252 2253
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
    VIR_DEBUG("qemuMonitorGetChardevInfo returned %i", ret);
2254 2255 2256
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

2257
    if (ret == 0) {
2258
        if ((ret = qemuProcessFindCharDevicePTYsMonitor(vm, info)) < 0)
2259 2260
            goto cleanup;

2261 2262
        if ((ret = qemuProcessRefreshChannelVirtioState(driver, vm, info,
                                                        true)) < 0)
2263 2264
            goto cleanup;
    }
2265

2266
 cleanup:
2267
    virHashFree(info);
2268

2269 2270
    if (logCtxt && kill(vm->pid, 0) == -1 && errno == ESRCH) {
        qemuProcessReportLogError(logCtxt,
2271
                                  _("process exited while connecting to monitor"));
2272 2273 2274 2275 2276 2277
        ret = -1;
    }

    return ret;
}

2278

2279 2280 2281 2282 2283 2284
static int
qemuProcessDetectIOThreadPIDs(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2285
    qemuMonitorIOThreadInfoPtr *iothreads = NULL;
2286 2287 2288 2289
    int niothreads = 0;
    int ret = -1;
    size_t i;

2290 2291 2292 2293 2294
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_OBJECT_IOTHREAD)) {
        ret = 0;
        goto cleanup;
    }

2295 2296 2297 2298
    /* Get the list of IOThreads from qemu */
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
    niothreads = qemuMonitorGetIOThreads(priv->mon, &iothreads);
2299 2300
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
J
John Ferlan 已提交
2301
    if (niothreads < 0)
2302 2303
        goto cleanup;

2304
    if (niothreads != vm->def->niothreadids) {
2305 2306
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of IOThread pids from QEMU monitor. "
2307 2308
                         "got %d, wanted %zu"),
                       niothreads, vm->def->niothreadids);
2309 2310 2311
        goto cleanup;
    }

2312 2313 2314 2315 2316 2317
    /* Nothing to do */
    if (niothreads == 0) {
        ret = 0;
        goto cleanup;
    }

2318 2319 2320
    for (i = 0; i < niothreads; i++) {
        virDomainIOThreadIDDefPtr iothrid;

2321 2322
        if (!(iothrid = virDomainIOThreadIDFind(vm->def,
                                                iothreads[i]->iothread_id))) {
2323
            virReportError(VIR_ERR_INTERNAL_ERROR,
2324 2325
                           _("iothread %d not found"),
                           iothreads[i]->iothread_id);
2326 2327 2328 2329
            goto cleanup;
        }
        iothrid->thread_id = iothreads[i]->thread_id;
    }
2330 2331 2332 2333 2334 2335

    ret = 0;

 cleanup:
    if (iothreads) {
        for (i = 0; i < niothreads; i++)
2336
            VIR_FREE(iothreads[i]);
2337 2338 2339 2340 2341
        VIR_FREE(iothreads);
    }
    return ret;
}

2342 2343 2344 2345

/*
 * To be run between fork/exec of QEMU only
 */
2346
#if defined(HAVE_SCHED_GETAFFINITY) || defined(HAVE_BSD_CPU_AFFINITY)
2347
static int
2348
qemuProcessInitCpuAffinity(virDomainObjPtr vm)
2349 2350 2351 2352
{
    int ret = -1;
    virBitmapPtr cpumap = NULL;
    virBitmapPtr cpumapToSet = NULL;
2353
    virBitmapPtr hostcpumap = NULL;
2354
    qemuDomainObjPrivatePtr priv = vm->privateData;
2355

2356 2357 2358 2359 2360 2361
    if (!vm->pid) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Cannot setup CPU affinity until process is started"));
        return -1;
    }

2362 2363
    if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) {
        VIR_DEBUG("Set CPU affinity with advisory nodeset from numad");
2364
        cpumapToSet = priv->autoCpuset;
2365
    } else {
2366
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
2367
        if (vm->def->cpumask) {
H
Hu Tao 已提交
2368
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
2369 2370 2371 2372 2373
        } 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 */
2374 2375
            int hostcpus;

2376 2377 2378 2379 2380 2381 2382 2383 2384 2385
            if (virHostCPUHasBitmap()) {
                hostcpumap = virHostCPUGetOnlineBitmap();
                cpumap = virProcessGetAffinity(vm->pid);
            }

            if (hostcpumap && cpumap && virBitmapEqual(hostcpumap, cpumap)) {
                /* we're using all available CPUs, no reason to set
                 * mask. If libvirtd is running without explicit
                 * affinity, we can use hotplugged CPUs for this VM */
                ret = 0;
2386
                goto cleanup;
2387 2388 2389 2390 2391
            } else {
                /* setaffinity fails if you set bits for CPUs which
                 * aren't present, so we have to limit ourselves */
                if ((hostcpus = virHostCPUGetCount()) < 0)
                    goto cleanup;
2392

2393 2394
                if (hostcpus > QEMUD_CPUMASK_LEN)
                    hostcpus = QEMUD_CPUMASK_LEN;
2395

2396 2397 2398
                virBitmapFree(cpumap);
                if (!(cpumap = virBitmapNew(hostcpus)))
                    goto cleanup;
2399

2400
                virBitmapSetAll(cpumap);
2401

2402 2403
                cpumapToSet = cpumap;
            }
O
Osier Yang 已提交
2404
        }
2405 2406
    }

2407
    if (virProcessSetAffinity(vm->pid, cpumapToSet) < 0)
2408
        goto cleanup;
2409

2410 2411
    ret = 0;

2412
 cleanup:
2413
    virBitmapFree(cpumap);
2414
    virBitmapFree(hostcpumap);
2415
    return ret;
2416
}
2417 2418 2419 2420 2421 2422 2423
#else /* !defined(HAVE_SCHED_GETAFFINITY) && !defined(HAVE_BSD_CPU_AFFINITY) */
static int
qemuProcessInitCpuAffinity(virDomainObjPtr vm ATTRIBUTE_UNUSED)
{
    return 0;
}
#endif /* !defined(HAVE_SCHED_GETAFFINITY) && !defined(HAVE_BSD_CPU_AFFINITY) */
2424

2425 2426
/* set link states to down on interfaces at qemu start */
static int
2427 2428 2429
qemuProcessSetLinkStates(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob)
2430 2431 2432
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2433
    size_t i;
2434 2435 2436 2437 2438
    int ret = -1;
    int rv;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2439 2440 2441

    for (i = 0; i < def->nnets; i++) {
        if (def->nets[i]->linkstate == VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN) {
2442 2443 2444
            if (!def->nets[i]->info.alias) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("missing alias for network device"));
2445
                goto cleanup;
2446 2447
            }

2448 2449
            VIR_DEBUG("Setting link state: %s", def->nets[i]->info.alias);

2450 2451 2452 2453
            rv = qemuMonitorSetLink(priv->mon,
                                    def->nets[i]->info.alias,
                                    VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (rv < 0) {
2454
                virReportError(VIR_ERR_OPERATION_FAILED,
2455 2456 2457
                               _("Couldn't set link state on interface: %s"),
                               def->nets[i]->info.alias);
                goto cleanup;
2458 2459 2460 2461
            }
        }
    }

2462 2463 2464 2465 2466
    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2467 2468 2469
    return ret;
}

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
/**
 * qemuProcessSetupPid:
 *
 * This function sets resource properities (affinity, cgroups,
 * scheduler) for any PID associated with a domain.  It should be used
 * to set up emulator PIDs as well as vCPU and I/O thread pids to
 * ensure they are all handled the same way.
 *
 * Returns 0 on success, -1 on error.
 */
static int
qemuProcessSetupPid(virDomainObjPtr vm,
                    pid_t pid,
                    virCgroupThreadName nameval,
                    int id,
                    virBitmapPtr cpumask,
                    unsigned long long period,
                    long long quota,
                    virDomainThreadSchedParamPtr sched)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainNumatuneMemMode mem_mode;
    virCgroupPtr cgroup = NULL;
    virBitmapPtr use_cpumask;
    char *mem_mask = NULL;
    int ret = -1;

    if ((period || quota) &&
        !virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPU)) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("cgroup cpu is required for scheduler tuning"));
        goto cleanup;
    }

    /* Infer which cpumask shall be used. */
    if (cpumask)
        use_cpumask = cpumask;
    else if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO)
        use_cpumask = priv->autoCpuset;
    else
        use_cpumask = vm->def->cpumask;

    /*
     * If CPU cgroup controller is not initialized here, then we need
     * neither period nor quota settings.  And if CPUSET controller is
     * not initialized either, then there's nothing to do anyway.
     */
    if (virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPU) ||
        virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPUSET)) {

        if (virDomainNumatuneGetMode(vm->def->numa, -1, &mem_mode) == 0 &&
            mem_mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT &&
            virDomainNumatuneMaybeFormatNodeset(vm->def->numa,
                                                priv->autoNodeset,
                                                &mem_mask, -1) < 0)
            goto cleanup;

        if (virCgroupNewThread(priv->cgroup, nameval, id, true, &cgroup) < 0)
            goto cleanup;

        if (virCgroupHasController(priv->cgroup, VIR_CGROUP_CONTROLLER_CPUSET)) {
            if (use_cpumask &&
                qemuSetupCgroupCpusetCpus(cgroup, use_cpumask) < 0)
                goto cleanup;

            /*
             * Don't setup cpuset.mems for the emulator, they need to
             * be set up after initialization in order for kvm
             * allocations to succeed.
             */
            if (nameval != VIR_CGROUP_THREAD_EMULATOR &&
                mem_mask && virCgroupSetCpusetMems(cgroup, mem_mask) < 0)
                goto cleanup;

        }

        if ((period || quota) &&
            qemuSetupCgroupVcpuBW(cgroup, period, quota) < 0)
            goto cleanup;

        /* Move the thread to the sub dir */
        if (virCgroupAddTask(cgroup, pid) < 0)
            goto cleanup;

    }

    /* Setup legacy affinity. */
    if (use_cpumask && virProcessSetAffinity(pid, use_cpumask) < 0)
        goto cleanup;

    /* Set scheduler type and priority. */
    if (sched &&
        virProcessSetScheduler(pid, sched->policy, sched->priority) < 0)
        goto cleanup;

    ret = 0;
 cleanup:
    VIR_FREE(mem_mask);
    if (cgroup) {
        if (ret < 0)
            virCgroupRemove(cgroup);
2572
        virCgroupFree(&cgroup);
2573 2574 2575 2576 2577 2578
    }

    return ret;
}


2579
static int
2580
qemuProcessSetupEmulator(virDomainObjPtr vm)
2581
{
2582 2583 2584 2585 2586
    return qemuProcessSetupPid(vm, vm->pid, VIR_CGROUP_THREAD_EMULATOR,
                               0, vm->def->cputune.emulatorpin,
                               vm->def->cputune.emulator_period,
                               vm->def->cputune.emulator_quota,
                               NULL);
2587 2588
}

2589

2590 2591 2592 2593 2594 2595
static int
qemuProcessResctrlCreate(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    int ret = -1;
    size_t i = 0;
2596
    virCapsPtr caps = NULL;
2597 2598
    qemuDomainObjPrivatePtr priv = vm->privateData;

B
Bing Niu 已提交
2599
    if (!vm->def->nresctrls)
2600 2601 2602 2603 2604
        return 0;

    /* Force capability refresh since resctrl info can change
     * XXX: move cache info into virresctrl so caps are not needed */
    caps = virQEMUDriverGetCapabilities(driver, true);
2605 2606 2607
    if (!caps)
        return -1;

B
Bing Niu 已提交
2608
    for (i = 0; i < vm->def->nresctrls; i++) {
2609
        if (virResctrlAllocCreate(caps->host.resctrl,
B
Bing Niu 已提交
2610
                                  vm->def->resctrls[i]->alloc,
2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621
                                  priv->machineName) < 0)
            goto cleanup;
    }

    ret = 0;
 cleanup:
    virObjectUnref(caps);
    return ret;
}


2622 2623 2624 2625 2626 2627 2628 2629 2630 2631
static char *
qemuProcessBuildPRHelperPidfilePath(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    const char *prdAlias = qemuDomainGetManagedPRAlias();

    return virPidFileBuildPath(priv->libDir, prdAlias);
}


2632
void
2633
qemuProcessKillManagedPRDaemon(virDomainObjPtr vm)
2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    char *pidfile;

    if (!priv->prDaemonRunning)
        return;

    if (!(pidfile = qemuProcessBuildPRHelperPidfilePath(vm))) {
        VIR_WARN("Unable to construct pr-helper pidfile path");
        return;
    }

    virErrorPreserveLast(&orig_err);
    if (virPidFileForceCleanupPath(pidfile) < 0) {
        VIR_WARN("Unable to kill pr-helper process");
    } else {
        if (unlink(pidfile) < 0 &&
            errno != ENOENT) {
            virReportSystemError(errno,
                                 _("Unable to remove stale pidfile %s"),
                                 pidfile);
        } else {
            priv->prDaemonRunning = false;
        }
    }
    virErrorRestore(&orig_err);

    VIR_FREE(pidfile);
}


static int
qemuProcessStartPRDaemonHook(void *opaque)
{
    virDomainObjPtr vm = opaque;
    size_t i, nfds = 0;
    int *fds = NULL;
    int ret = -1;

2674 2675 2676
    if (qemuDomainNamespaceEnabled(vm, QEMU_DOMAIN_NS_MOUNT)) {
        if (virProcessGetNamespaces(vm->pid, &nfds, &fds) < 0)
            return ret;
2677

2678 2679 2680 2681
        if (nfds > 0 &&
            virProcessSetNamespaces(nfds, fds) < 0)
            goto cleanup;
    }
2682 2683 2684 2685 2686 2687 2688 2689 2690 2691

    ret = 0;
 cleanup:
    for (i = 0; i < nfds; i++)
        VIR_FORCE_CLOSE(fds[i]);
    VIR_FREE(fds);
    return ret;
}


2692
int
2693
qemuProcessStartManagedPRDaemon(virDomainObjPtr vm)
2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virQEMUDriverPtr driver = priv->driver;
    virQEMUDriverConfigPtr cfg;
    int errfd = -1;
    char *pidfile = NULL;
    int pidfd = -1;
    char *socketPath = NULL;
    pid_t cpid = -1;
    virCommandPtr cmd = NULL;
    virTimeBackOffVar timebackoff;
    const unsigned long long timeout = 500000; /* ms */
    int ret = -1;

    cfg = virQEMUDriverGetConfig(driver);

    if (!virFileIsExecutable(cfg->prHelperName)) {
        virReportSystemError(errno, _("'%s' is not a suitable pr helper"),
                             cfg->prHelperName);
        goto cleanup;
    }

    if (!(pidfile = qemuProcessBuildPRHelperPidfilePath(vm)))
        goto cleanup;

    /* Just try to acquire. Dummy pid will be replaced later */
    if ((pidfd = virPidFileAcquirePath(pidfile, false, -1)) < 0)
        goto cleanup;

2723
    if (!(socketPath = qemuDomainGetManagedPRSocketPath(priv)))
2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792
        goto cleanup;

    /* Remove stale socket */
    if (unlink(socketPath) < 0 &&
        errno != ENOENT) {
        virReportSystemError(errno,
                             _("Unable to remove stale socket path: %s"),
                             socketPath);
        goto cleanup;
    }

    if (!(cmd = virCommandNewArgList(cfg->prHelperName,
                                     "-k", socketPath,
                                     "-f", pidfile,
                                     NULL)))
        goto cleanup;

    virCommandDaemonize(cmd);
    /* We want our virCommand to write child PID into the pidfile
     * so that we can read it even before exec(). */
    virCommandSetPidFile(cmd, pidfile);
    virCommandSetErrorFD(cmd, &errfd);

    /* Place the process into the same namespace and cgroup as
     * qemu (so that it shares the same view of the system). */
    virCommandSetPreExecHook(cmd, qemuProcessStartPRDaemonHook, vm);

    if (virCommandRun(cmd, NULL) < 0)
        goto cleanup;

    if (virPidFileReadPath(pidfile, &cpid) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("pr helper %s didn't show up"),
                       cfg->prHelperName);
        goto cleanup;
    }

    if (virTimeBackOffStart(&timebackoff, 1, timeout) < 0)
        goto cleanup;
    while (virTimeBackOffWait(&timebackoff)) {
        char errbuf[1024] = { 0 };

        if (virFileExists(socketPath))
            break;

        if (virProcessKill(cpid, 0) == 0)
            continue;

        if (saferead(errfd, errbuf, sizeof(errbuf) - 1) < 0) {
            virReportSystemError(errno,
                                 _("pr helper %s died unexpectedly"),
                                 cfg->prHelperName);
        } else {
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("pr helper died and reported: %s"), errbuf);
        }
        goto cleanup;
    }

    if (!virFileExists(socketPath)) {
        virReportError(VIR_ERR_OPERATION_TIMEOUT, "%s",
                       _("pr helper socked did not show up"));
        goto cleanup;
    }

    if (priv->cgroup &&
        virCgroupAddMachineTask(priv->cgroup, cpid) < 0)
        goto cleanup;

2793
    if (qemuSecurityDomainSetPathLabel(driver, vm, socketPath, true) < 0)
2794 2795 2796
        goto cleanup;

    priv->prDaemonRunning = true;
2797
    ret = 0;
2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815
 cleanup:
    if (ret < 0) {
        virCommandAbort(cmd);
        if (cpid >= 0)
            virProcessKillPainfully(cpid, true);
        if (pidfile)
            unlink(pidfile);
    }
    virCommandFree(cmd);
    VIR_FREE(socketPath);
    VIR_FORCE_CLOSE(pidfd);
    VIR_FREE(pidfile);
    VIR_FORCE_CLOSE(errfd);
    virObjectUnref(cfg);
    return ret;
}


2816
static int
2817
qemuProcessInitPasswords(virQEMUDriverPtr driver,
2818 2819
                         virDomainObjPtr vm,
                         int asyncJob)
2820 2821
{
    int ret = 0;
2822
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2823
    size_t i;
2824

2825
    for (i = 0; i < vm->def->ngraphics; ++i) {
2826 2827
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
2828 2829
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_VNC,
2830
                                                    &graphics->data.vnc.auth,
2831 2832
                                                    cfg->vncPassword,
                                                    asyncJob);
2833
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
2834 2835
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_SPICE,
2836
                                                    &graphics->data.spice.auth,
2837 2838
                                                    cfg->spicePassword,
                                                    asyncJob);
2839 2840
        }

2841 2842 2843
        if (ret < 0)
            goto cleanup;
    }
2844

2845
 cleanup:
2846
    virObjectUnref(cfg);
2847 2848 2849 2850 2851 2852 2853 2854 2855 2856
    return ret;
}


static int
qemuProcessPrepareChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
    int fd;
2857
    if (dev->source->type != VIR_DOMAIN_CHR_TYPE_FILE)
2858 2859
        return 0;

2860
    if ((fd = open(dev->source->data.file.path,
2861 2862 2863
                   O_CREAT | O_APPEND, S_IRUSR|S_IWUSR)) < 0) {
        virReportSystemError(errno,
                             _("Unable to pre-create chardev file '%s'"),
2864
                             dev->source->data.file.path);
2865 2866 2867 2868 2869 2870 2871 2872 2873
        return -1;
    }

    VIR_FORCE_CLOSE(fd);

    return 0;
}


2874 2875 2876 2877 2878
static int
qemuProcessCleanupChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
2879 2880 2881 2882
    if (dev->source->type == VIR_DOMAIN_CHR_TYPE_UNIX &&
        dev->source->data.nix.listen &&
        dev->source->data.nix.path)
        unlink(dev->source->data.nix.path);
2883 2884 2885 2886 2887

    return 0;
}


2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923
/**
 * 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) {
2924
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VGAMEM) &&
2925 2926
                    qemuMonitorUpdateVideoMemorySize(priv->mon, video,
                                                     "qxl-vga") < 0)
2927
                        goto error;
2928

2929
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VRAM64) &&
2930 2931 2932
                    qemuMonitorUpdateVideoVram64Size(priv->mon, video,
                                                     "qxl-vga") < 0)
                    goto error;
2933
            } else {
2934 2935 2936 2937 2938 2939 2940 2941
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VGAMEM) &&
                    qemuMonitorUpdateVideoMemorySize(priv->mon, video,
                                                     "qxl") < 0)
                        goto error;

                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VRAM64) &&
                    qemuMonitorUpdateVideoVram64Size(priv->mon, video,
                                                     "qxl") < 0)
2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960
                        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;
        }

    }

2961 2962
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2963 2964

    cfg = virQEMUDriverGetConfig(driver);
2965
    ret = virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps);
2966 2967 2968 2969 2970
    virObjectUnref(cfg);

    return ret;

 error:
2971
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
2972 2973 2974 2975
    return -1;
}


2976 2977
struct qemuProcessHookData {
    virDomainObjPtr vm;
2978
    virQEMUDriverPtr driver;
2979
    virQEMUDriverConfigPtr cfg;
2980 2981 2982 2983 2984
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
2985
    qemuDomainObjPrivatePtr priv = h->vm->privateData;
2986
    int ret = -1;
2987
    int fd;
2988 2989 2990
    virBitmapPtr nodeset = NULL;
    virDomainNumatuneMemMode mode;

2991 2992 2993 2994
    /* This method cannot use any mutexes, which are not
     * protected across fork()
     */

2995
    qemuSecurityPostFork(h->driver->securityManager);
2996 2997 2998 2999 3000

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

    VIR_DEBUG("Obtaining domain lock");
3001 3002 3003 3004 3005 3006 3007
    /*
     * 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.
     */
3008
    if (qemuSecuritySetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
3009
        goto cleanup;
3010
    if (virDomainLockProcessStart(h->driver->lockManager,
3011
                                  h->cfg->uri,
3012
                                  h->vm,
J
Ján Tomko 已提交
3013
                                  /* QEMU is always paused initially */
3014 3015
                                  true,
                                  &fd) < 0)
3016
        goto cleanup;
3017
    if (qemuSecurityClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
3018
        goto cleanup;
3019

3020
    if (qemuDomainBuildNamespace(h->cfg, h->driver->securityManager, h->vm) < 0)
3021 3022
        goto cleanup;

3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033
    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);
3034

3035 3036 3037
            if (virNumaSetupMemoryPolicy(mode, nodeset) < 0)
                goto cleanup;
        }
3038
    }
3039

3040 3041
    ret = 0;

3042
 cleanup:
3043
    virObjectUnref(h->cfg);
3044 3045
    VIR_DEBUG("Hook complete ret=%d", ret);
    return ret;
3046 3047 3048
}

int
3049 3050
qemuProcessPrepareMonitorChr(virDomainChrSourceDefPtr monConfig,
                             const char *domainDir)
3051 3052 3053 3054
{
    monConfig->type = VIR_DOMAIN_CHR_TYPE_UNIX;
    monConfig->data.nix.listen = true;

3055 3056
    if (virAsprintf(&monConfig->data.nix.path, "%s/monitor.sock",
                    domainDir) < 0)
3057 3058
        return -1;
    return 0;
3059 3060 3061
}


3062
/*
3063 3064
 * Precondition: vm must be locked, and a job must be active.
 * This method will call {Enter,Exit}Monitor
3065
 */
E
Eric Blake 已提交
3066
int
3067
qemuProcessStartCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
3068
                     virDomainRunningReason reason,
3069
                     qemuDomainAsyncJob asyncJob)
3070
{
3071
    int ret = -1;
3072
    qemuDomainObjPrivatePtr priv = vm->privateData;
3073
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3074

3075
    /* Bring up netdevs before starting CPUs */
3076
    if (qemuInterfaceStartDevices(vm->def) < 0)
3077 3078
       goto cleanup;

3079
    VIR_DEBUG("Using lock state '%s'", NULLSTR(priv->lockState));
3080
    if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
3081
                                   vm, priv->lockState) < 0) {
3082 3083 3084 3085
        /* 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
         */
3086
        goto cleanup;
3087 3088 3089
    }
    VIR_FREE(priv->lockState);

3090 3091
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto release;
J
Jiri Denemark 已提交
3092

3093
    ret = qemuMonitorStartCPUs(priv->mon);
3094 3095
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3096 3097 3098 3099 3100

    if (ret < 0)
        goto release;

    virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
3101

3102
 cleanup:
3103
    virObjectUnref(cfg);
3104
    return ret;
3105 3106 3107 3108 3109 3110

 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;
3111 3112 3113
}


3114 3115
int qemuProcessStopCPUs(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
3116
                        virDomainPausedReason reason,
3117
                        qemuDomainAsyncJob asyncJob)
3118
{
3119
    int ret = -1;
3120 3121
    qemuDomainObjPrivatePtr priv = vm->privateData;

3122
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
3123

3124 3125
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
J
Jiri Denemark 已提交
3126

3127
    ret = qemuMonitorStopCPUs(priv->mon);
3128 3129
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3130 3131 3132 3133

    if (ret < 0)
        goto cleanup;

3134 3135 3136
    /* de-activate netdevs after stopping CPUs */
    ignore_value(qemuInterfaceStopDevices(vm->def));

3137 3138 3139
    if (priv->job.current)
        ignore_value(virTimeMillisNow(&priv->job.current->stopped));

3140 3141 3142 3143
    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
3144

3145
 cleanup:
3146 3147 3148 3149 3150
    return ret;
}



3151
static void
3152 3153
qemuProcessNotifyNets(virDomainDefPtr def)
{
3154
    size_t i;
3155

3156 3157
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
3158 3159 3160 3161 3162 3163 3164 3165
        /* 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));

3166
        virDomainNetNotifyActualDevice(def, net);
3167 3168 3169
    }
}

3170
static int
3171
qemuProcessFiltersInstantiate(virDomainDefPtr def, bool ignoreExists)
3172
{
3173
    size_t i;
3174

3175
    for (i = 0; i < def->nnets; i++) {
3176 3177
        virDomainNetDefPtr net = def->nets[i];
        if ((net->filter) && (net->ifname)) {
3178
            if (virDomainConfNWFilterInstantiate(def->name, def->uuid, net, ignoreExists) < 0)
J
Ján Tomko 已提交
3179
                return 1;
3180 3181 3182
        }
    }

J
Ján Tomko 已提交
3183
    return 0;
3184 3185
}

3186
static int
3187
qemuProcessUpdateState(virQEMUDriverPtr driver, virDomainObjPtr vm)
3188 3189 3190
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
3191
    virDomainPausedReason reason;
3192
    virDomainState newState = VIR_DOMAIN_NOSTATE;
3193
    int oldReason;
3194
    int newReason;
3195
    bool running;
3196
    char *msg = NULL;
3197 3198
    int ret;

3199
    qemuDomainObjEnterMonitor(driver, vm);
3200
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
3201 3202
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
3203

3204
    if (ret < 0)
3205 3206
        return -1;

3207
    state = virDomainObjGetState(vm, &oldReason);
3208

3209 3210 3211 3212 3213 3214 3215 3216
    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) {
3217 3218
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
3219
        ignore_value(VIR_STRDUP_QUIET(msg, "was unpaused"));
3220
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
3221 3222 3223
        if (reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN) {
            newState = VIR_DOMAIN_SHUTDOWN;
            newReason = VIR_DOMAIN_SHUTDOWN_UNKNOWN;
3224
            ignore_value(VIR_STRDUP_QUIET(msg, "shutdown"));
3225
        } else if (reason == VIR_DOMAIN_PAUSED_CRASHED) {
3226 3227 3228
            newState = VIR_DOMAIN_CRASHED;
            newReason = VIR_DOMAIN_CRASHED_PANICKED;
            ignore_value(VIR_STRDUP_QUIET(msg, "crashed"));
3229 3230 3231
        } else {
            newState = VIR_DOMAIN_PAUSED;
            newReason = reason;
S
Stefan Berger 已提交
3232 3233
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
3234 3235 3236 3237 3238 3239 3240
        }
    }

    if (newState != VIR_DOMAIN_NOSTATE) {
        VIR_DEBUG("Domain %s %s while its monitor was disconnected;"
                  " changing state to %s (%s)",
                  vm->def->name,
3241
                  NULLSTR(msg),
3242 3243 3244 3245
                  virDomainStateTypeToString(newState),
                  virDomainStateReasonToString(newState, newReason));
        VIR_FREE(msg);
        virDomainObjSetState(vm, newState, newReason);
3246 3247 3248 3249 3250
    }

    return 0;
}

3251
static int
3252 3253
qemuProcessRecoverMigrationIn(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
3254
                              const qemuDomainJobObj *job,
3255
                              virDomainState state,
3256
                              int reason)
3257
{
3258 3259 3260 3261 3262
    bool postcopy = (state == VIR_DOMAIN_PAUSED &&
                     reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED) ||
                    (state == VIR_DOMAIN_RUNNING &&
                     reason == VIR_DOMAIN_RUNNING_POSTCOPY);

3263
    switch ((qemuMigrationJobPhase) job->phase) {
3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278
    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:
        /* N/A for incoming migration */
        break;

    case QEMU_MIGRATION_PHASE_PREPARE:
        VIR_DEBUG("Killing unfinished incoming migration for domain %s",
                  vm->def->name);
        return -1;
3279

3280 3281 3282 3283 3284
    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);
3285
        if (qemuProcessStartCPUs(driver, vm,
3286 3287 3288 3289 3290 3291 3292 3293 3294
                                 VIR_DOMAIN_RUNNING_UNPAUSED,
                                 QEMU_ASYNC_JOB_NONE) < 0) {
            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 unless
3295 3296
         * we already started guest CPUs or we were in post-copy mode */
        if (postcopy) {
3297
            qemuMigrationAnyPostcopyFailed(driver, vm);
3298
        } else if (state != VIR_DOMAIN_RUNNING) {
3299
            VIR_DEBUG("Killing migrated domain %s", vm->def->name);
3300
            return -1;
3301 3302 3303
        }
        break;
    }
3304

3305 3306
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_NONE,
                             job->migParams, job->apiFlags);
3307 3308
    return 0;
}
3309

3310 3311 3312
static int
qemuProcessRecoverMigrationOut(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
3313
                               const qemuDomainJobObj *job,
3314
                               virDomainState state,
3315 3316
                               int reason,
                               unsigned int *stopFlags)
3317
{
3318 3319 3320
    bool postcopy = state == VIR_DOMAIN_PAUSED &&
                    (reason == VIR_DOMAIN_PAUSED_POSTCOPY ||
                     reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
3321
    bool resume = false;
3322

3323
    switch ((qemuMigrationJobPhase) job->phase) {
3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339
    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:
        /* N/A for outgoing migration */
        break;

    case QEMU_MIGRATION_PHASE_BEGIN3:
        /* nothing happened 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
3340 3341 3342 3343
         * domain; however we can only do that before migration enters
         * post-copy mode
         */
        if (postcopy) {
3344
            qemuMigrationAnyPostcopyFailed(driver, vm);
3345 3346 3347
        } else {
            VIR_DEBUG("Cancelling unfinished migration of domain %s",
                      vm->def->name);
3348
            if (qemuMigrationSrcCancel(driver, vm) < 0) {
3349 3350 3351
                VIR_WARN("Could not cancel ongoing migration of domain %s",
                         vm->def->name);
            }
3352
            resume = true;
3353
        }
3354
        break;
3355

3356 3357
    case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
        /* migration finished but we didn't have a chance to get the result
3358 3359
         * of Finish3 step; third party needs to check what to do next; in
         * post-copy mode we can use PAUSED_POSTCOPY_FAILED state for this
3360
         */
3361
        if (postcopy)
3362
            qemuMigrationAnyPostcopyFailed(driver, vm);
3363
        break;
3364

3365
    case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
3366 3367 3368 3369 3370
        /* Finish3 failed, we need to resume the domain, but once we enter
         * post-copy mode there's no way back, so let's just mark the domain
         * as broken in that case
         */
        if (postcopy) {
3371
            qemuMigrationAnyPostcopyFailed(driver, vm);
3372 3373 3374
        } else {
            VIR_DEBUG("Resuming domain %s after failed migration",
                      vm->def->name);
3375
            resume = true;
3376 3377
        }
        break;
3378

3379 3380
    case QEMU_MIGRATION_PHASE_CONFIRM3:
        /* migration completed, we need to kill the domain here */
3381
        *stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
3382 3383
        return -1;
    }
3384

3385 3386 3387 3388 3389 3390 3391
    if (resume) {
        /* 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)) {
3392
            if (qemuProcessStartCPUs(driver, vm,
3393 3394 3395 3396
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
                VIR_WARN("Could not resume domain %s", vm->def->name);
            }
3397 3398
        }
    }
3399

3400 3401
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_NONE,
                             job->migParams, job->apiFlags);
3402 3403 3404
    return 0;
}

3405
static int
3406
qemuProcessRecoverJob(virQEMUDriverPtr driver,
3407
                      virDomainObjPtr vm,
3408
                      const qemuDomainJobObj *job,
3409
                      unsigned int *stopFlags)
3410
{
3411
    qemuDomainObjPrivatePtr priv = vm->privateData;
3412 3413 3414 3415 3416 3417 3418
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
3419
        if (qemuProcessRecoverMigrationOut(driver, vm, job,
3420
                                           state, reason, stopFlags) < 0)
3421 3422 3423
            return -1;
        break;

3424
    case QEMU_ASYNC_JOB_MIGRATION_IN:
3425
        if (qemuProcessRecoverMigrationIn(driver, vm, job,
3426
                                          state, reason) < 0)
3427
            return -1;
3428 3429 3430 3431
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
3432
    case QEMU_ASYNC_JOB_SNAPSHOT:
3433
        qemuDomainObjEnterMonitor(driver, vm);
3434
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
3435 3436
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
3437
        /* resume the domain but only if it was paused as a result of
3438 3439 3440 3441 3442 3443 3444 3445 3446
         * 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 &&
3447 3448
               (reason == VIR_DOMAIN_PAUSED_SNAPSHOT ||
                reason == VIR_DOMAIN_PAUSED_MIGRATION)) ||
3449
              reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
3450
             if (qemuProcessStartCPUs(driver, vm,
3451 3452 3453 3454
                                      VIR_DOMAIN_RUNNING_UNPAUSED,
                                      QEMU_ASYNC_JOB_NONE) < 0) {
                 VIR_WARN("Could not resume domain '%s' after migration to file",
                          vm->def->name);
3455 3456 3457 3458
            }
        }
        break;

3459 3460 3461 3462
    case QEMU_ASYNC_JOB_START:
        /* Already handled in VIR_DOMAIN_PAUSED_STARTING_UP check. */
        break;

3463 3464 3465 3466 3467 3468 3469 3470
    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        break;
    }

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

3471 3472 3473 3474
    /* 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.
     */
3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494
    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;

3495
    case QEMU_JOB_MIGRATION_OP:
3496
    case QEMU_JOB_ABORT:
3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507
    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;
}

3508 3509 3510 3511 3512 3513
static int
qemuProcessUpdateDevices(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDeviceDef dev;
3514
    const char **qemuDevices;
3515 3516 3517 3518 3519 3520 3521 3522 3523
    char **old;
    char **tmp;
    int ret = -1;

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

    old = priv->qemuDevices;
    priv->qemuDevices = NULL;
3524
    if (qemuDomainUpdateDeviceList(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
3525 3526
        goto cleanup;

3527
    qemuDevices = (const char **)priv->qemuDevices;
3528 3529
    if ((tmp = old)) {
        while (*tmp) {
3530
            if (!virStringListHasString(qemuDevices, *tmp) &&
3531 3532 3533 3534
                virDomainDefFindDevice(vm->def, *tmp, &dev, false) == 0 &&
                qemuDomainRemoveDevice(driver, vm, &dev) < 0) {
                goto cleanup;
            }
3535 3536 3537 3538 3539
            tmp++;
        }
    }
    ret = 0;

3540
 cleanup:
3541
    virStringListFree(old);
3542 3543 3544
    return ret;
}

3545 3546 3547 3548 3549 3550 3551
static int
qemuDomainPerfRestart(virDomainObjPtr vm)
{
    size_t i;
    virDomainDefPtr def = vm->def;
    qemuDomainObjPrivatePtr priv = vm->privateData;

3552
    if (!(priv->perf = virPerfNew()))
3553 3554 3555
        return -1;

    for (i = 0; i < VIR_PERF_EVENT_LAST; i++) {
3556 3557
        if (def->perf.events[i] &&
            def->perf.events[i] == VIR_TRISTATE_BOOL_YES) {
3558 3559 3560

            /* Failure to re-enable the perf event should not be fatal */
            if (virPerfEventEnable(priv->perf, i, vm->pid) < 0)
3561
                def->perf.events[i] = VIR_TRISTATE_BOOL_NO;
3562 3563 3564 3565 3566 3567
        }
    }

    return 0;
}

3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590

static void
qemuProcessReconnectCheckMemAliasOrderMismatch(virDomainObjPtr vm)
{
    size_t i;
    int aliasidx;
    virDomainDefPtr def = vm->def;
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (!virDomainDefHasMemoryHotplug(def) || def->nmems == 0)
        return;

    for (i = 0; i < def->nmems; i++) {
        aliasidx = qemuDomainDeviceAliasIndex(&def->mems[i]->info, "dimm");

        if (def->mems[i]->info.addr.dimm.slot != aliasidx) {
            priv->memAliasOrderMismatch = true;
            break;
        }
    }
}


3591
static bool
3592 3593
qemuProcessNeedHugepagesPath(virDomainDefPtr def,
                             virDomainMemoryDefPtr mem)
3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612
{
    const long system_pagesize = virGetSystemPageSizeKB();
    size_t i;

    if (def->mem.source == VIR_DOMAIN_MEMORY_SOURCE_FILE)
        return true;

    for (i = 0; i < def->mem.nhugepages; i++) {
        if (def->mem.hugepages[i].size != system_pagesize)
            return true;
    }

    for (i = 0; i < def->nmems; i++) {
        if (def->mems[i]->model == VIR_DOMAIN_MEMORY_MODEL_DIMM &&
            def->mems[i]->pagesize &&
            def->mems[i]->pagesize != system_pagesize)
            return true;
    }

3613 3614 3615 3616 3617 3618
    if (mem &&
        mem->model == VIR_DOMAIN_MEMORY_MODEL_DIMM &&
        mem->pagesize &&
        mem->pagesize != system_pagesize)
        return true;

3619 3620 3621 3622
    return false;
}


3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652
static bool
qemuProcessNeedMemoryBackingPath(virDomainDefPtr def,
                                 virDomainMemoryDefPtr mem)
{
    size_t i;
    size_t numaNodes;

    if (def->mem.source == VIR_DOMAIN_MEMORY_SOURCE_FILE ||
        def->mem.access != VIR_DOMAIN_MEMORY_ACCESS_DEFAULT)
        return true;

    numaNodes = virDomainNumaGetNodeCount(def->numa);
    for (i = 0; i < numaNodes; i++) {
        if (virDomainNumaGetNodeMemoryAccessMode(def->numa, i)
            != VIR_DOMAIN_MEMORY_ACCESS_DEFAULT)
            return true;
    }

    if (mem &&
        mem->model == VIR_DOMAIN_MEMORY_MODEL_DIMM &&
        (mem->access != VIR_DOMAIN_MEMORY_ACCESS_DEFAULT ||
         (mem->targetNode >= 0 &&
          virDomainNumaGetNodeMemoryAccessMode(def->numa, mem->targetNode)
          != VIR_DOMAIN_MEMORY_ACCESS_DEFAULT)))
        return true;

    return false;
}


3653 3654
static int
qemuProcessBuildDestroyMemoryPathsImpl(virQEMUDriverPtr driver,
3655
                                       virDomainObjPtr vm,
3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669
                                       const char *path,
                                       bool build)
{
    if (build) {
        if (virFileExists(path))
            return 0;

        if (virFileMakePathWithMode(path, 0700) < 0) {
            virReportSystemError(errno,
                                 _("Unable to create %s"),
                                 path);
            return -1;
        }

3670
        if (qemuSecurityDomainSetPathLabel(driver, vm, path, true) < 0)
3671 3672
            return -1;
    } else {
M
Michal Privoznik 已提交
3673 3674
        if (virFileDeleteTree(path) < 0)
            return -1;
3675 3676 3677 3678 3679 3680
    }

    return 0;
}


3681
int
3682 3683 3684 3685
qemuProcessBuildDestroyMemoryPaths(virQEMUDriverPtr driver,
                                   virDomainObjPtr vm,
                                   virDomainMemoryDefPtr mem,
                                   bool build)
3686 3687
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3688
    char *path = NULL;
3689
    size_t i;
3690 3691
    bool shouldBuildHP = false;
    bool shouldBuildMB = false;
3692 3693
    int ret = -1;

3694 3695 3696 3697
    if (build) {
        shouldBuildHP = qemuProcessNeedHugepagesPath(vm->def, mem);
        shouldBuildMB = qemuProcessNeedMemoryBackingPath(vm->def, mem);
    }
3698

3699
    if (!build || shouldBuildHP) {
3700
        for (i = 0; i < cfg->nhugetlbfs; i++) {
3701
            path = qemuGetDomainHugepagePath(vm->def, &cfg->hugetlbfs[i]);
3702

3703
            if (!path)
3704 3705
                goto cleanup;

3706
            if (qemuProcessBuildDestroyMemoryPathsImpl(driver, vm,
3707
                                                       path, build) < 0)
3708
                goto cleanup;
3709

3710
            VIR_FREE(path);
3711 3712 3713
        }
    }

3714 3715 3716 3717
    if (!build || shouldBuildMB) {
        if (qemuGetMemoryBackingDomainPath(vm->def, cfg, &path) < 0)
            goto cleanup;

3718
        if (qemuProcessBuildDestroyMemoryPathsImpl(driver, vm,
3719 3720 3721 3722 3723 3724
                                                   path, build) < 0)
            goto cleanup;

        VIR_FREE(path);
    }

3725 3726
    ret = 0;
 cleanup:
3727
    VIR_FREE(path);
3728 3729 3730 3731 3732
    virObjectUnref(cfg);
    return ret;
}


3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758
int
qemuProcessDestroyMemoryBackingPath(virQEMUDriverPtr driver,
                                    virDomainObjPtr vm,
                                    virDomainMemoryDefPtr mem)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    char *path = NULL;
    int ret = -1;

    if (qemuGetMemoryBackingPath(vm->def, cfg, mem->info.alias, &path) < 0)
        goto cleanup;

    if (unlink(path) < 0 &&
        errno != ENOENT) {
        virReportSystemError(errno, _("Unable to remove %s"), path);
        goto cleanup;
    }

    ret = 0;
 cleanup:
    VIR_FREE(path);
    virObjectUnref(cfg);
    return ret;
}


3759 3760 3761 3762
static int
qemuProcessVNCAllocatePorts(virQEMUDriverPtr driver,
                            virDomainGraphicsDefPtr graphics,
                            bool allocate)
3763
{
3764
    unsigned short port;
3765

3766 3767 3768
    if (!allocate) {
        if (graphics->data.vnc.autoport)
            graphics->data.vnc.port = 5900;
3769

3770 3771
        return 0;
    }
3772

3773 3774 3775 3776 3777
    if (graphics->data.vnc.autoport) {
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
            return -1;
        graphics->data.vnc.port = port;
    }
3778

3779 3780 3781 3782 3783 3784
    if (graphics->data.vnc.websocket == -1) {
        if (virPortAllocatorAcquire(driver->webSocketPorts, &port) < 0)
            return -1;
        graphics->data.vnc.websocket = port;
        graphics->data.vnc.websocketGenerated = true;
    }
3785

3786 3787
    return 0;
}
J
John Ferlan 已提交
3788

3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799
static int
qemuProcessSPICEAllocatePorts(virQEMUDriverPtr driver,
                              virDomainGraphicsDefPtr graphics,
                              bool allocate)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    unsigned short port = 0;
    unsigned short tlsPort;
    size_t i;
    int defaultMode = graphics->data.spice.defaultMode;
    int ret = -1;
3800

3801 3802
    bool needTLSPort = false;
    bool needPort = false;
3803

3804 3805 3806 3807 3808 3809 3810
    if (graphics->data.spice.autoport) {
        /* check if tlsPort or port need allocation */
        for (i = 0; i < VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_LAST; i++) {
            switch (graphics->data.spice.channels[i]) {
            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
                needTLSPort = true;
                break;
3811

3812 3813 3814
            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
                needPort = true;
                break;
3815

3816 3817 3818 3819 3820 3821 3822 3823 3824
            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
                /* default mode will be used */
                break;
            }
        }
        switch (defaultMode) {
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
            needTLSPort = true;
            break;
3825

3826 3827 3828
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
            needPort = true;
            break;
3829

3830 3831 3832 3833 3834 3835 3836
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
            if (cfg->spiceTLS)
                needTLSPort = true;
            needPort = true;
            break;
        }
    }
3837

3838 3839 3840
    if (!allocate) {
        if (needPort || graphics->data.spice.port == -1)
            graphics->data.spice.port = 5901;
3841

3842 3843
        if (needTLSPort || graphics->data.spice.tlsPort == -1)
            graphics->data.spice.tlsPort = 5902;
3844

3845 3846
        ret = 0;
        goto cleanup;
3847 3848
    }

3849 3850 3851
    if (needPort || graphics->data.spice.port == -1) {
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
            goto cleanup;
3852

3853 3854 3855 3856
        graphics->data.spice.port = port;

        if (!graphics->data.spice.autoport)
            graphics->data.spice.portReserved = true;
3857 3858
    }

3859 3860 3861 3862 3863 3864 3865
    if (needTLSPort || graphics->data.spice.tlsPort == -1) {
        if (!cfg->spiceTLS) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Auto allocation of spice TLS port requested "
                             "but spice TLS is disabled in qemu.conf"));
            goto cleanup;
        }
3866

3867 3868
        if (virPortAllocatorAcquire(driver->remotePorts, &tlsPort) < 0)
            goto cleanup;
3869

3870
        graphics->data.spice.tlsPort = tlsPort;
3871

3872 3873
        if (!graphics->data.spice.autoport)
            graphics->data.spice.tlsPortReserved = true;
3874
    }
3875

3876
    ret = 0;
3877

3878 3879 3880 3881
 cleanup:
    virObjectUnref(cfg);
    return ret;
}
3882

3883

3884 3885 3886 3887 3888
static int
qemuValidateCpuCount(virDomainDefPtr def,
                     virQEMUCapsPtr qemuCaps)
{
    unsigned int maxCpus = virQEMUCapsGetMachineMaxCpus(qemuCaps, def->os.machine);
3889

3890 3891 3892 3893 3894
    if (virDomainDefGetVcpus(def) == 0) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Domain requires at least 1 vCPU"));
        return -1;
    }
3895

3896 3897 3898 3899 3900
    if (maxCpus > 0 && virDomainDefGetVcpusMax(def) > maxCpus) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Maximum CPUs greater than specified machine type limit"));
        return -1;
    }
3901

3902 3903
    return 0;
}
3904

3905

3906 3907 3908 3909 3910 3911 3912
static int
qemuProcessVerifyHypervFeatures(virDomainDefPtr def,
                                virCPUDataPtr cpu)
{
    char *cpuFeature;
    size_t i;
    int rc;
3913

3914 3915 3916 3917
    for (i = 0; i < VIR_DOMAIN_HYPERV_LAST; i++) {
        /* always supported string property */
        if (i == VIR_DOMAIN_HYPERV_VENDOR_ID)
            continue;
3918

3919 3920
        if (def->hyperv_features[i] != VIR_TRISTATE_SWITCH_ON)
            continue;
3921

3922 3923 3924
        if (virAsprintf(&cpuFeature, "__kvm_hv_%s",
                        virDomainHypervTypeToString(i)) < 0)
            return -1;
3925

3926 3927
        rc = virCPUDataCheckFeature(cpu, cpuFeature);
        VIR_FREE(cpuFeature);
3928

3929 3930 3931 3932
        if (rc < 0)
            return -1;
        else if (rc == 1)
            continue;
3933

3934 3935 3936 3937 3938 3939 3940
        switch ((virDomainHyperv) i) {
        case VIR_DOMAIN_HYPERV_RELAXED:
        case VIR_DOMAIN_HYPERV_VAPIC:
        case VIR_DOMAIN_HYPERV_SPINLOCKS:
            VIR_WARN("host doesn't support hyperv '%s' feature",
                     virDomainHypervTypeToString(i));
            break;
3941

3942 3943 3944 3945 3946
        case VIR_DOMAIN_HYPERV_VPINDEX:
        case VIR_DOMAIN_HYPERV_RUNTIME:
        case VIR_DOMAIN_HYPERV_SYNIC:
        case VIR_DOMAIN_HYPERV_STIMER:
        case VIR_DOMAIN_HYPERV_RESET:
3947
        case VIR_DOMAIN_HYPERV_FREQUENCIES:
3948
        case VIR_DOMAIN_HYPERV_REENLIGHTENMENT:
3949
        case VIR_DOMAIN_HYPERV_TLBFLUSH:
3950 3951 3952 3953
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("host doesn't support hyperv '%s' feature"),
                           virDomainHypervTypeToString(i));
            return -1;
3954

3955 3956 3957 3958
        /* coverity[dead_error_begin] */
        case VIR_DOMAIN_HYPERV_VENDOR_ID:
        case VIR_DOMAIN_HYPERV_LAST:
            break;
3959
        }
3960
    }
3961 3962

    return 0;
3963 3964
}

3965

3966
static int
3967 3968
qemuProcessVerifyKVMFeatures(virDomainDefPtr def,
                             virCPUDataPtr cpu)
3969
{
3970
    int rc = 0;
3971

3972
    if (def->features[VIR_DOMAIN_FEATURE_PVSPINLOCK] != VIR_TRISTATE_SWITCH_ON)
3973 3974
        return 0;

3975 3976 3977 3978 3979 3980
    rc = virCPUDataCheckFeature(cpu, VIR_CPU_x86_KVM_PV_UNHALT);

    if (rc <= 0) {
        if (rc == 0)
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("host doesn't support paravirtual spinlocks"));
3981
        return -1;
3982
    }
3983

3984 3985
    return 0;
}
3986 3987


3988 3989 3990 3991 3992
static int
qemuProcessVerifyCPUFeatures(virDomainDefPtr def,
                             virCPUDataPtr cpu)
{
    int rc;
S
Stefan Berger 已提交
3993

3994
    rc = virCPUCheckFeature(def->os.arch, def->cpu, "invtsc");
3995

3996
    if (rc < 0) {
3997
        return -1;
3998 3999 4000 4001 4002 4003 4004 4005 4006
    } else if (rc == 1) {
        rc = virCPUDataCheckFeature(cpu, "invtsc");
        if (rc <= 0) {
            if (rc == 0) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("host doesn't support invariant TSC"));
            }
            return -1;
        }
4007
    }
4008

4009
    return 0;
4010 4011 4012
}


4013
static int
4014 4015 4016 4017 4018
qemuProcessFetchGuestCPU(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob,
                         virCPUDataPtr *enabled,
                         virCPUDataPtr *disabled)
4019
{
4020 4021 4022 4023
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCPUDataPtr dataEnabled = NULL;
    virCPUDataPtr dataDisabled = NULL;
    int rc;
4024

4025 4026
    *enabled = NULL;
    *disabled = NULL;
4027

4028
    if (!ARCH_IS_X86(vm->def->os.arch))
4029 4030
        return 0;

4031 4032
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto error;
4033

4034 4035
    rc = qemuMonitorGetGuestCPU(priv->mon, vm->def->os.arch,
                                &dataEnabled, &dataDisabled);
4036

4037 4038 4039 4040 4041 4042 4043 4044
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto error;

    if (rc == -1)
        goto error;

    *enabled = dataEnabled;
    *disabled = dataDisabled;
4045
    return 0;
4046 4047 4048 4049 4050

 error:
    virCPUDataFree(dataEnabled);
    virCPUDataFree(dataDisabled);
    return -1;
4051
}
4052

4053

4054
static int
4055 4056
qemuProcessVerifyCPU(virDomainObjPtr vm,
                     virCPUDataPtr cpu)
4057
{
4058
    virDomainDefPtr def = vm->def;
4059

4060 4061
    if (!cpu)
        return 0;
J
Jiri Denemark 已提交
4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078

    if (qemuProcessVerifyKVMFeatures(def, cpu) < 0 ||
        qemuProcessVerifyHypervFeatures(def, cpu) < 0)
        return -1;

    if (!def->cpu ||
        (def->cpu->mode == VIR_CPU_MODE_CUSTOM &&
         !def->cpu->model))
        return 0;

    if (qemuProcessVerifyCPUFeatures(def, cpu) < 0)
        return -1;

    return 0;
}


4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120
static int
qemuProcessUpdateLiveGuestCPU(virDomainObjPtr vm,
                              virCPUDataPtr enabled,
                              virCPUDataPtr disabled)
{
    virDomainDefPtr def = vm->def;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCPUDefPtr orig = NULL;
    int rc;
    int ret = -1;

    if (!enabled)
        return 0;

    if (!def->cpu ||
        (def->cpu->mode == VIR_CPU_MODE_CUSTOM &&
         !def->cpu->model))
        return 0;

    if (!(orig = virCPUDefCopy(def->cpu)))
        goto cleanup;

    if ((rc = virCPUUpdateLive(def->os.arch, def->cpu, enabled, disabled)) < 0) {
        goto cleanup;
    } else if (rc == 0) {
        /* Store the original CPU in priv if QEMU changed it and we didn't
         * get the original CPU via migration, restore, or snapshot revert.
         */
        if (!priv->origCPU && !virCPUDefIsEqual(def->cpu, orig, false))
            VIR_STEAL_PTR(priv->origCPU, orig);

        def->cpu->check = VIR_CPU_CHECK_FULL;
    }

    ret = 0;

 cleanup:
    virCPUDefFree(orig);
    return ret;
}


4121
static int
4122
qemuProcessUpdateAndVerifyCPU(virQEMUDriverPtr driver,
4123 4124
                              virDomainObjPtr vm,
                              qemuDomainAsyncJob asyncJob)
4125
{
4126
    virCPUDataPtr cpu = NULL;
4127
    virCPUDataPtr disabled = NULL;
4128
    int ret = -1;
4129

4130 4131
    if (qemuProcessFetchGuestCPU(driver, vm, asyncJob, &cpu, &disabled) < 0)
        goto cleanup;
4132

J
Jiri Denemark 已提交
4133 4134
    if (qemuProcessVerifyCPU(vm, cpu) < 0)
        goto cleanup;
J
Ján Tomko 已提交
4135

4136 4137
    if (qemuProcessUpdateLiveGuestCPU(vm, cpu, disabled) < 0)
        goto cleanup;
4138

4139
    ret = 0;
4140

4141
 cleanup:
4142
    virCPUDataFree(cpu);
4143
    virCPUDataFree(disabled);
4144 4145 4146 4147
    return ret;
}


4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171
static virDomainCapsCPUModelsPtr
qemuProcessFetchCPUDefinitions(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
                               qemuDomainAsyncJob asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainCapsCPUModelsPtr models = NULL;

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

    models = virQEMUCapsFetchCPUDefinitions(priv->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto error;

    return models;

 error:
    virObjectUnref(models);
    return NULL;
}


4172 4173 4174 4175 4176 4177 4178
static int
qemuProcessUpdateCPU(virQEMUDriverPtr driver,
                     virDomainObjPtr vm,
                     qemuDomainAsyncJob asyncJob)
{
    virCPUDataPtr cpu = NULL;
    virCPUDataPtr disabled = NULL;
4179
    virDomainCapsCPUModelsPtr models = NULL;
4180 4181
    int ret = -1;

4182 4183 4184 4185 4186
    /* The host CPU model comes from host caps rather than QEMU caps so
     * fallback must be allowed no matter what the user specified in the XML.
     */
    vm->def->cpu->fallback = VIR_CPU_FALLBACK_ALLOW;

4187 4188 4189 4190 4191 4192
    if (qemuProcessFetchGuestCPU(driver, vm, asyncJob, &cpu, &disabled) < 0)
        goto cleanup;

    if (qemuProcessUpdateLiveGuestCPU(vm, cpu, disabled) < 0)
        goto cleanup;

4193 4194 4195 4196
    if (!(models = qemuProcessFetchCPUDefinitions(driver, vm, asyncJob)) ||
        virCPUTranslate(vm->def->os.arch, vm->def->cpu, models) < 0)
        goto cleanup;

4197 4198 4199 4200 4201
    ret = 0;

 cleanup:
    virCPUDataFree(cpu);
    virCPUDataFree(disabled);
4202
    virObjectUnref(models);
4203 4204 4205 4206
    return ret;
}


4207 4208
static int
qemuPrepareNVRAM(virQEMUDriverConfigPtr cfg,
4209
                 virDomainObjPtr vm)
4210 4211 4212 4213
{
    int ret = -1;
    int srcFD = -1;
    int dstFD = -1;
4214
    virDomainLoaderDefPtr loader = vm->def->os.loader;
4215
    bool created = false;
4216 4217
    const char *master_nvram_path;
    ssize_t r;
4218

4219
    if (!loader || !loader->nvram || virFileExists(loader->nvram))
4220 4221
        return 0;

4222 4223 4224
    master_nvram_path = loader->templt;
    if (!loader->templt) {
        size_t i;
4225 4226 4227
        for (i = 0; i < cfg->nfirmwares; i++) {
            if (STREQ(cfg->firmwares[i]->name, loader->path)) {
                master_nvram_path = cfg->firmwares[i]->nvram;
4228
                break;
4229 4230
            }
        }
4231
    }
4232

4233 4234 4235 4236 4237 4238
    if (!master_nvram_path) {
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("unable to find any master var store for "
                         "loader: %s"), loader->path);
        goto cleanup;
    }
4239

4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256
    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;
4257

4258 4259
    do {
        char buf[1024];
4260

4261
        if ((r = saferead(srcFD, buf, sizeof(buf))) < 0) {
4262
            virReportSystemError(errno,
4263
                                 _("Unable to read from file '%s'"),
4264 4265 4266
                                 master_nvram_path);
            goto cleanup;
        }
4267 4268

        if (safewrite(dstFD, buf, r) < 0) {
4269
            virReportSystemError(errno,
4270
                                 _("Unable to write to file '%s'"),
4271 4272 4273
                                 loader->nvram);
            goto cleanup;
        }
4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286
    } 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;
4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303
    }

    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);
    }

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


4304 4305 4306
static void
qemuLogOperation(virDomainObjPtr vm,
                 const char *msg,
4307 4308
                 virCommandPtr cmd,
                 qemuDomainLogContextPtr logCtxt)
4309 4310 4311 4312 4313
{
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int qemuVersion = virQEMUCapsGetVersion(priv->qemuCaps);
    const char *package = virQEMUCapsGetPackage(priv->qemuCaps);
4314
    char *hostname = virGetHostname();
4315 4316 4317
    struct utsname uts;

    uname(&uts);
4318 4319

    if ((timestamp = virTimeStringNow()) == NULL)
4320
        goto cleanup;
4321

4322
    if (qemuDomainLogContextWrite(logCtxt,
4323
                                  "%s: %s %s, qemu version: %d.%d.%d%s, kernel: %s, hostname: %s\n",
4324 4325 4326 4327
                                  timestamp, msg, VIR_LOG_VERSION_STRING,
                                  (qemuVersion / 1000000) % 1000,
                                  (qemuVersion / 1000) % 1000,
                                  qemuVersion % 1000,
4328
                                  package ? package : "",
4329
                                  uts.release,
4330
                                  hostname ? hostname : "") < 0)
4331
        goto cleanup;
4332

4333 4334 4335 4336 4337
    if (cmd) {
        char *args = virCommandToString(cmd);
        qemuDomainLogContextWrite(logCtxt, "%s\n", args);
        VIR_FREE(args);
    }
4338 4339

 cleanup:
4340
    VIR_FREE(hostname);
4341 4342 4343
    VIR_FREE(timestamp);
}

4344 4345 4346 4347 4348 4349 4350

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

4351
    VIR_FREE(inc->address);
4352
    VIR_FREE(inc->launchURI);
4353
    VIR_FREE(inc->deferredURI);
4354 4355 4356 4357 4358 4359 4360 4361
    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.
4362 4363 4364
 *
 * The caller is responsible for closing @fd, calling
 * qemuProcessIncomingDefFree will NOT close it.
4365 4366 4367
 */
qemuProcessIncomingDefPtr
qemuProcessIncomingDefNew(virQEMUCapsPtr qemuCaps,
4368
                          const char *listenAddress,
4369 4370 4371 4372 4373 4374
                          const char *migrateFrom,
                          int fd,
                          const char *path)
{
    qemuProcessIncomingDefPtr inc = NULL;

4375
    if (qemuMigrationDstCheckProtocol(qemuCaps, migrateFrom) < 0)
4376 4377 4378 4379 4380
        return NULL;

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

4381 4382 4383
    if (VIR_STRDUP(inc->address, listenAddress) < 0)
        goto error;

4384
    inc->launchURI = qemuMigrationDstGetURI(migrateFrom, fd);
4385 4386 4387
    if (!inc->launchURI)
        goto error;

4388 4389 4390 4391 4392 4393
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_INCOMING_DEFER)) {
        inc->deferredURI = inc->launchURI;
        if (VIR_STRDUP(inc->launchURI, "defer") < 0)
            goto error;
    }

4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404
    inc->fd = fd;
    inc->path = path;

    return inc;

 error:
    qemuProcessIncomingDefFree(inc);
    return NULL;
}


4405 4406 4407 4408 4409 4410 4411 4412
/*
 * 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,
4413
                    virDomainObjPtr vm,
4414 4415
                    virDomainJobOperation operation,
                    unsigned long apiFlags)
4416
{
4417
    if (qemuDomainObjBeginAsyncJob(driver, vm, QEMU_ASYNC_JOB_START,
4418
                                   operation, apiFlags) < 0)
4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433
        return -1;

    qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
    return 0;
}


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


4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456
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;
}


4457
static int
4458
qemuProcessGraphicsReservePorts(virDomainGraphicsDefPtr graphics,
4459
                                bool reconnect)
4460
{
4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471
    virDomainGraphicsListenDefPtr glisten;

    if (graphics->nListens <= 0)
        return 0;

    glisten = &graphics->listens[0];

    if (glisten->type != VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_ADDRESS &&
        glisten->type != VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_NETWORK)
        return 0;

4472 4473
    switch (graphics->type) {
    case VIR_DOMAIN_GRAPHICS_TYPE_VNC:
4474 4475
        if (!graphics->data.vnc.autoport ||
            reconnect) {
4476
            if (virPortAllocatorSetUsed(graphics->data.vnc.port) < 0)
4477 4478 4479
                return -1;
            graphics->data.vnc.portReserved = true;
        }
4480
        if (graphics->data.vnc.websocket > 0 &&
4481
            virPortAllocatorSetUsed(graphics->data.vnc.websocket) < 0)
4482
            return -1;
4483 4484 4485
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SPICE:
4486
        if (graphics->data.spice.autoport && !reconnect)
4487
            return 0;
4488

4489
        if (graphics->data.spice.port > 0) {
4490
            if (virPortAllocatorSetUsed(graphics->data.spice.port) < 0)
4491
                return -1;
4492 4493
            graphics->data.spice.portReserved = true;
        }
4494

4495
        if (graphics->data.spice.tlsPort > 0) {
4496
            if (virPortAllocatorSetUsed(graphics->data.spice.tlsPort) < 0)
4497 4498
                return -1;
            graphics->data.spice.tlsPortReserved = true;
4499
        }
4500 4501 4502 4503 4504
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SDL:
    case VIR_DOMAIN_GRAPHICS_TYPE_RDP:
    case VIR_DOMAIN_GRAPHICS_TYPE_DESKTOP:
4505
    case VIR_DOMAIN_GRAPHICS_TYPE_EGL_HEADLESS:
4506 4507
    case VIR_DOMAIN_GRAPHICS_TYPE_LAST:
        break;
4508 4509
    }

4510 4511 4512 4513
    return 0;
}


4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543
static int
qemuProcessGraphicsAllocatePorts(virQEMUDriverPtr driver,
                                 virDomainGraphicsDefPtr graphics,
                                 bool allocate)
{
    virDomainGraphicsListenDefPtr glisten;

    if (graphics->nListens <= 0)
        return 0;

    glisten = &graphics->listens[0];

    if (glisten->type != VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_ADDRESS &&
        glisten->type != VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_NETWORK)
        return 0;

    switch (graphics->type) {
    case VIR_DOMAIN_GRAPHICS_TYPE_VNC:
        if (qemuProcessVNCAllocatePorts(driver, graphics, allocate) < 0)
            return -1;
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SPICE:
        if (qemuProcessSPICEAllocatePorts(driver, graphics, allocate) < 0)
            return -1;
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SDL:
    case VIR_DOMAIN_GRAPHICS_TYPE_RDP:
    case VIR_DOMAIN_GRAPHICS_TYPE_DESKTOP:
4544
    case VIR_DOMAIN_GRAPHICS_TYPE_EGL_HEADLESS:
4545 4546 4547 4548 4549 4550 4551
    case VIR_DOMAIN_GRAPHICS_TYPE_LAST:
        break;
    }

    return 0;
}

4552
static int
4553
qemuProcessGetNetworkAddress(const char *netname,
4554 4555
                             char **netaddr)
{
4556
    virConnectPtr conn = NULL;
4557 4558 4559 4560 4561 4562 4563 4564 4565 4566
    int ret = -1;
    virNetworkPtr net;
    virNetworkDefPtr netdef = NULL;
    virNetworkIPDefPtr ipdef;
    virSocketAddr addr;
    virSocketAddrPtr addrptr = NULL;
    char *dev_name = NULL;
    char *xml = NULL;

    *netaddr = NULL;
4567 4568 4569 4570

    if (!(conn = virGetConnectNetwork()))
        return -1;

4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582
    net = virNetworkLookupByName(conn, netname);
    if (!net)
        goto cleanup;

    xml = virNetworkGetXMLDesc(net, 0);
    if (!xml)
        goto cleanup;

    netdef = virNetworkDefParseString(xml);
    if (!netdef)
        goto cleanup;

4583
    switch ((virNetworkForwardType) netdef->forward.type) {
4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619
    case VIR_NETWORK_FORWARD_NONE:
    case VIR_NETWORK_FORWARD_NAT:
    case VIR_NETWORK_FORWARD_ROUTE:
    case VIR_NETWORK_FORWARD_OPEN:
        ipdef = virNetworkDefGetIPByIndex(netdef, AF_UNSPEC, 0);
        if (!ipdef) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' doesn't have an IP address"),
                           netdef->name);
            goto cleanup;
        }
        addrptr = &ipdef->address;
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
        if ((dev_name = netdef->bridge))
            break;
        /*
         * fall through if netdef->bridge wasn't set, since that is
         * macvtap bridge mode network.
         */
        ATTRIBUTE_FALLTHROUGH;

    case VIR_NETWORK_FORWARD_PRIVATE:
    case VIR_NETWORK_FORWARD_VEPA:
    case VIR_NETWORK_FORWARD_PASSTHROUGH:
        if ((netdef->forward.nifs > 0) && netdef->forward.ifs)
            dev_name = netdef->forward.ifs[0].device.dev;

        if (!dev_name) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' has no associated interface or bridge"),
                           netdef->name);
            goto cleanup;
        }
        break;
4620 4621 4622 4623 4624 4625 4626 4627

    case VIR_NETWORK_FORWARD_HOSTDEV:
        break;

    case VIR_NETWORK_FORWARD_LAST:
    default:
        virReportEnumRangeError(virNetworkForwardType, netdef->forward.type);
        goto cleanup;
4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644
    }

    if (dev_name) {
        if (virNetDevIPAddrGet(dev_name, &addr) < 0)
            goto cleanup;
        addrptr = &addr;
    }

    if (!(addrptr &&
          (*netaddr = virSocketAddrFormat(addrptr)))) {
        goto cleanup;
    }

    ret = 0;
 cleanup:
    virNetworkDefFree(netdef);
    virObjectUnref(net);
4645
    virObjectUnref(conn);
4646 4647 4648 4649
    VIR_FREE(xml);
    return ret;
}

4650

4651
static int
4652
qemuProcessGraphicsSetupNetworkAddress(virDomainGraphicsListenDefPtr glisten,
4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663
                                       const char *listenAddr)
{
    int rc;

    /* TODO: reject configuration without network specified for network listen */
    if (!glisten->network) {
        if (VIR_STRDUP(glisten->address, listenAddr) < 0)
            return -1;
        return 0;
    }

4664
    rc = qemuProcessGetNetworkAddress(glisten->network, &glisten->address);
4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677
    if (rc <= -2) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("network-based listen isn't possible, "
                         "network driver isn't present"));
        return -1;
    }
    if (rc < 0)
        return -1;

    return 0;
}


4678
static int
4679
qemuProcessGraphicsSetupListen(virQEMUDriverPtr driver,
4680 4681
                               virDomainGraphicsDefPtr graphics,
                               virDomainObjPtr vm)
4682
{
4683
    qemuDomainObjPrivatePtr priv = vm->privateData;
4684
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4685
    const char *type = virDomainGraphicsTypeToString(graphics->type);
4686
    char *listenAddr = NULL;
4687
    bool useSocket = false;
4688
    size_t i;
4689
    int ret = -1;
4690 4691 4692

    switch (graphics->type) {
    case VIR_DOMAIN_GRAPHICS_TYPE_VNC:
4693
        useSocket = cfg->vncAutoUnixSocket;
4694 4695 4696 4697
        listenAddr = cfg->vncListen;
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SPICE:
4698
        useSocket = cfg->spiceAutoUnixSocket;
4699 4700 4701 4702 4703 4704
        listenAddr = cfg->spiceListen;
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SDL:
    case VIR_DOMAIN_GRAPHICS_TYPE_RDP:
    case VIR_DOMAIN_GRAPHICS_TYPE_DESKTOP:
4705
    case VIR_DOMAIN_GRAPHICS_TYPE_EGL_HEADLESS:
4706 4707 4708 4709 4710 4711 4712 4713 4714
    case VIR_DOMAIN_GRAPHICS_TYPE_LAST:
        break;
    }

    for (i = 0; i < graphics->nListens; i++) {
        virDomainGraphicsListenDefPtr glisten = &graphics->listens[i];

        switch (glisten->type) {
        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_ADDRESS:
4715 4716 4717 4718 4719
            if (!glisten->address) {
                /* If there is no address specified and qemu.conf has
                 * *_auto_unix_socket set we should use unix socket as
                 * default instead of tcp listen. */
                if (useSocket) {
4720 4721
                    memset(glisten, 0, sizeof(virDomainGraphicsListenDef));
                    if (virAsprintf(&glisten->socket, "%s/%s.sock",
4722
                                    priv->libDir, type) < 0)
4723
                        goto cleanup;
4724 4725
                    glisten->fromConfig = true;
                    glisten->type = VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_SOCKET;
4726 4727
                } else if (listenAddr) {
                    if (VIR_STRDUP(glisten->address, listenAddr) < 0)
4728
                        goto cleanup;
4729 4730 4731
                    glisten->fromConfig = true;
                }
            }
4732 4733 4734 4735 4736 4737
            break;

        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_NETWORK:
            if (glisten->address || !listenAddr)
                continue;

4738
            if (qemuProcessGraphicsSetupNetworkAddress(glisten,
4739
                                                       listenAddr) < 0)
4740
                goto cleanup;
4741 4742
            break;

4743 4744 4745 4746
        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_SOCKET:
            if (!glisten->socket) {
                if (virAsprintf(&glisten->socket, "%s/%s.sock",
                                priv->libDir, type) < 0)
4747
                    goto cleanup;
4748 4749 4750 4751
                glisten->autoGenerated = true;
            }
            break;

4752 4753 4754 4755 4756 4757
        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_NONE:
        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_LAST:
            break;
        }
    }

4758 4759 4760 4761 4762
    ret = 0;

 cleanup:
    virObjectUnref(cfg);
    return ret;
4763 4764 4765
}


4766
static int
4767
qemuProcessSetupGraphics(virQEMUDriverPtr driver,
4768 4769
                         virDomainObjPtr vm,
                         unsigned int flags)
4770
{
4771
    virDomainGraphicsDefPtr graphics;
4772
    bool allocate = !(flags & VIR_QEMU_PROCESS_START_PRETEND);
4773
    size_t i;
4774 4775
    int ret = -1;

4776 4777 4778
    for (i = 0; i < vm->def->ngraphics; i++) {
        graphics = vm->def->graphics[i];

4779
        if (qemuProcessGraphicsSetupListen(driver, graphics, vm) < 0)
4780 4781 4782
            goto cleanup;
    }

4783 4784 4785 4786
    if (allocate) {
        for (i = 0; i < vm->def->ngraphics; i++) {
            graphics = vm->def->graphics[i];

4787
            if (qemuProcessGraphicsReservePorts(graphics, false) < 0)
4788 4789 4790
                goto cleanup;
        }
    }
4791

4792
    for (i = 0; i < vm->def->ngraphics; ++i) {
4793
        graphics = vm->def->graphics[i];
4794

4795 4796
        if (qemuProcessGraphicsAllocatePorts(driver, graphics, allocate) < 0)
            goto cleanup;
4797 4798 4799 4800 4801 4802 4803 4804 4805
    }

    ret = 0;

 cleanup:
    return ret;
}


4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838
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++) {
4839
            if (!virHostdevIsSCSIDevice(vm->def->hostdevs[i]))
4840 4841
                continue;

4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867
            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;
}


4868 4869 4870 4871 4872 4873 4874 4875 4876
static int
qemuProcessSetupBalloon(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuDomainAsyncJob asyncJob)
{
    unsigned long long balloon = vm->def->mem.cur_balloon;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;

4877
    if (!virDomainDefHasMemballoon(vm->def))
4878 4879 4880
        return 0;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
4881
        return -1;
4882

4883 4884 4885
    if (vm->def->memballoon->period)
        qemuMonitorSetMemoryStatsPeriod(priv->mon, vm->def->memballoon,
                                        vm->def->memballoon->period);
4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897
    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
4898 4899 4900
static int
qemuProcessMakeDir(virQEMUDriverPtr driver,
                   virDomainObjPtr vm,
4901
                   const char *path)
J
Jiri Denemark 已提交
4902 4903 4904 4905 4906 4907 4908 4909
{
    int ret = -1;

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

4910
    if (qemuSecurityDomainSetPathLabel(driver, vm, path, true) < 0)
J
Jiri Denemark 已提交
4911 4912 4913 4914 4915 4916 4917 4918 4919
        goto cleanup;

    ret = 0;

 cleanup:
    return ret;
}


4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955
static void
qemuProcessStartWarnShmem(virDomainObjPtr vm)
{
    size_t i;
    bool check_shmem = false;
    bool shmem = vm->def->nshmems;

    /*
     * 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)
        return;

    /*
     * 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) ==
4956
                VIR_DOMAIN_MEMORY_ACCESS_SHARED) {
4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968
                shmem = true;
                break;
            }
        }
    }

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

4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991

static int
qemuProcessStartValidateGraphics(virDomainObjPtr vm)
{
    size_t i;

    for (i = 0; i < vm->def->ngraphics; i++) {
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];

        switch (graphics->type) {
        case VIR_DOMAIN_GRAPHICS_TYPE_VNC:
        case VIR_DOMAIN_GRAPHICS_TYPE_SPICE:
            if (graphics->nListens > 1) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("QEMU does not support multiple listens for "
                                 "one graphics device."));
                return -1;
            }
            break;

        case VIR_DOMAIN_GRAPHICS_TYPE_SDL:
        case VIR_DOMAIN_GRAPHICS_TYPE_RDP:
        case VIR_DOMAIN_GRAPHICS_TYPE_DESKTOP:
4992
        case VIR_DOMAIN_GRAPHICS_TYPE_EGL_HEADLESS:
4993 4994 4995 4996 4997 4998 4999 5000 5001
        case VIR_DOMAIN_GRAPHICS_TYPE_LAST:
            break;
        }
    }

    return 0;
}


5002 5003 5004 5005 5006 5007 5008 5009 5010 5011
static int
qemuProcessStartValidateVideo(virDomainObjPtr vm,
                              virQEMUCapsPtr qemuCaps)
{
    size_t i;
    virDomainVideoDefPtr video;

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

5012 5013 5014 5015 5016 5017 5018 5019 5020
        if ((video->type == VIR_DOMAIN_VIDEO_TYPE_VGA &&
             !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_VGA)) ||
            (video->type == VIR_DOMAIN_VIDEO_TYPE_CIRRUS &&
             !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_CIRRUS_VGA)) ||
            (video->type == VIR_DOMAIN_VIDEO_TYPE_VMVGA &&
             !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_VMWARE_SVGA)) ||
            (video->type == VIR_DOMAIN_VIDEO_TYPE_QXL &&
             !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_QXL)) ||
            (video->type == VIR_DOMAIN_VIDEO_TYPE_VIRTIO &&
5021 5022 5023 5024
             !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_VIRTIO_GPU)) ||
            (video->type == VIR_DOMAIN_VIDEO_TYPE_VIRTIO &&
             video->info.type == VIR_DOMAIN_DEVICE_ADDRESS_TYPE_CCW &&
             !virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_VIRTIO_GPU_CCW))) {
5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("this QEMU does not support '%s' video device"),
                           virDomainVideoTypeToString(video->type));
            return -1;
        }

        if (video->accel) {
            if (video->accel->accel3d == VIR_TRISTATE_SWITCH_ON &&
                (video->type != VIR_DOMAIN_VIDEO_TYPE_VIRTIO ||
                 !virQEMUCapsGet(qemuCaps, QEMU_CAPS_VIRTIO_GPU_VIRGL))) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("%s 3d acceleration is not supported"),
                               virDomainVideoTypeToString(video->type));
                return -1;
            }
        }
    }

    return 0;
}


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
static int
qemuProcessStartValidateIOThreads(virDomainObjPtr vm,
                                  virQEMUCapsPtr qemuCaps)
{
    size_t i;

    if (vm->def->niothreadids > 0 &&
        !virQEMUCapsGet(qemuCaps, QEMU_CAPS_OBJECT_IOTHREAD)) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("IOThreads not supported for this QEMU"));
        return -1;
    }

    for (i = 0; i < vm->def->ncontrollers; i++) {
        virDomainControllerDefPtr cont = vm->def->controllers[i];

        if (cont->type == VIR_DOMAIN_CONTROLLER_TYPE_SCSI &&
            cont->model == VIR_DOMAIN_CONTROLLER_MODEL_SCSI_VIRTIO_SCSI &&
            cont->iothread > 0 &&
            !virQEMUCapsGet(qemuCaps, QEMU_CAPS_VIRTIO_SCSI_IOTHREAD)) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("IOThreads for virtio-scsi not supported for "
                             "this QEMU"));
            return -1;
        }
    }

    return 0;
}


5078
static int
5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097
qemuProcessStartValidateShmem(virDomainObjPtr vm)
{
    size_t i;

    for (i = 0; i < vm->def->nshmems; i++) {
        virDomainShmemDefPtr shmem = vm->def->shmems[i];

        if (strchr(shmem->name, '/')) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("shmem name '%s' must not contain '/'"),
                           shmem->name);
            return -1;
        }
    }

    return 0;
}


5098 5099 5100 5101 5102 5103 5104
static int
qemuProcessStartValidateDisks(virDomainObjPtr vm,
                              virQEMUCapsPtr qemuCaps)
{
    size_t i;

    for (i = 0; i < vm->def->ndisks; i++) {
5105 5106
        virDomainDiskDefPtr disk = vm->def->disks[i];
        virStorageSourcePtr src = disk->src;
5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118

        /* This is a best effort check as we can only check if the command
         * option exists, but we cannot determine whether the running QEMU
         * was build with '--enable-vxhs'. */
        if (src->type == VIR_STORAGE_TYPE_NETWORK &&
            src->protocol == VIR_STORAGE_NET_PROTOCOL_VXHS &&
            !virQEMUCapsGet(qemuCaps, QEMU_CAPS_VXHS)) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("VxHS protocol is not supported with this "
                             "QEMU binary"));
            return -1;
        }
5119 5120 5121 5122 5123 5124 5125 5126

        /* PowerPC pseries based VMs do not support floppy device */
        if (disk->device == VIR_DOMAIN_DISK_DEVICE_FLOPPY &&
            qemuDomainIsPSeries(vm->def)) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("PowerPC pseries machines do not support floppy device"));
            return -1;
        }
5127 5128 5129 5130 5131 5132
    }

    return 0;
}


5133
static int
5134 5135
qemuProcessStartValidateXML(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
5136
                            virQEMUCapsPtr qemuCaps,
5137
                            virCapsPtr caps,
5138
                            unsigned int flags)
5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150
{
    /* The bits we validate here are XML configs that we previously
     * accepted. We reject them at VM startup time rather than parse
     * time so that pre-existing VMs aren't rejected and dropped from
     * the VM list when libvirt is updated.
     *
     * If back compat isn't a concern, XML validation should probably
     * be done at parse time.
     */
    if (qemuValidateCpuCount(vm->def, qemuCaps) < 0)
        return -1;

5151 5152 5153
    /* checks below should not be executed when starting a qemu process for a
     * VM that was running before (migration, snapshots, save). It's more
     * important to start such VM than keep the configuration clean */
5154 5155 5156
    if ((flags & VIR_QEMU_PROCESS_START_NEW) &&
        virDomainDefValidate(vm->def, caps, 0, driver->xmlopt) < 0)
        return -1;
5157 5158 5159

    return 0;
}
5160

5161

5162 5163 5164 5165 5166 5167
/**
 * qemuProcessStartValidate:
 * @vm: domain object
 * @qemuCaps: emulator capabilities
 * @migration: restoration of existing state
 *
5168 5169 5170 5171 5172
 * This function aggregates checks done prior to start of a VM.
 *
 * Flag VIR_QEMU_PROCESS_START_PRETEND tells, that we don't want to actually
 * start the domain but create a valid qemu command.  If some code shouldn't be
 * executed in this case, make sure to check this flag.
5173
 */
5174
static int
5175 5176
qemuProcessStartValidate(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
5177
                         virQEMUCapsPtr qemuCaps,
5178
                         virCapsPtr caps,
5179
                         unsigned int flags)
5180
{
5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194
    if (!(flags & VIR_QEMU_PROCESS_START_PRETEND)) {
        if (vm->def->virtType == VIR_DOMAIN_VIRT_KVM) {
            VIR_DEBUG("Checking for KVM availability");
            if (!virFileExists("/dev/kvm")) {
                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."));
                return -1;
            }
        }

        VIR_DEBUG("Checking domain and device security labels");
5195
        if (qemuSecurityCheckAllLabel(driver->securityManager, vm->def) < 0)
5196 5197 5198 5199
            return -1;

    }

5200
    if (qemuProcessStartValidateXML(driver, vm, qemuCaps, caps, flags) < 0)
5201 5202
        return -1;

5203 5204 5205
    if (qemuProcessStartValidateGraphics(vm) < 0)
        return -1;

5206 5207 5208
    if (qemuProcessStartValidateVideo(vm, qemuCaps) < 0)
        return -1;

5209 5210 5211
    if (qemuProcessStartValidateIOThreads(vm, qemuCaps) < 0)
        return -1;

5212 5213 5214
    if (qemuProcessStartValidateShmem(vm) < 0)
        return -1;

5215 5216 5217 5218
    if (vm->def->cpu &&
        virCPUValidateFeatures(vm->def->os.arch, vm->def->cpu) < 0)
        return -1;

5219 5220 5221
    if (qemuProcessStartValidateDisks(vm, qemuCaps) < 0)
        return -1;

5222 5223
    VIR_DEBUG("Checking for any possible (non-fatal) issues");

5224
    qemuProcessStartWarnShmem(vm);
5225

5226
    return 0;
5227 5228 5229
}


J
Jiri Denemark 已提交
5230 5231 5232 5233 5234 5235
/**
 * qemuProcessInit:
 *
 * Prepares the domain up to the point when priv->qemuCaps is initialized. The
 * function calls qemuProcessStop when needed.
 *
5236 5237 5238 5239
 * Flag VIR_QEMU_PROCESS_START_PRETEND tells, that we don't want to actually
 * start the domain but create a valid qemu command.  If some code shouldn't be
 * executed in this case, make sure to check this flag.
 *
J
Jiri Denemark 已提交
5240 5241 5242 5243 5244
 * Returns 0 on success, -1 on error.
 */
int
qemuProcessInit(virQEMUDriverPtr driver,
                virDomainObjPtr vm,
5245
                virCPUDefPtr updatedCPU,
5246
                qemuDomainAsyncJob asyncJob,
5247
                bool migration,
5248
                unsigned int flags)
J
Jiri Denemark 已提交
5249 5250 5251 5252 5253
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int stopFlags;
5254
    virCPUDefPtr origCPU = NULL;
J
Jiri Denemark 已提交
5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268
    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)))
5269 5270
        goto cleanup;

5271 5272 5273 5274 5275 5276 5277 5278 5279
    /* in case when the post parse callback failed we need to re-run it on the
     * old config prior we start the VM */
    if (vm->def->postParseFailed) {
        VIR_DEBUG("re-running the post parse callback");

        if (virDomainDefPostParse(vm->def, caps, 0, driver->xmlopt, NULL) < 0)
            goto cleanup;
    }

5280 5281
    VIR_DEBUG("Determining emulator version");
    virObjectUnref(priv->qemuCaps);
5282
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
5283 5284 5285
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
        goto cleanup;
J
Jiri Denemark 已提交
5286

5287 5288 5289
    if (flags & VIR_QEMU_PROCESS_START_STANDALONE)
        virQEMUCapsClear(priv->qemuCaps, QEMU_CAPS_CHARDEV_FD_PASS);

5290 5291 5292
    if (qemuDomainUpdateCPU(vm, updatedCPU, &origCPU) < 0)
        goto cleanup;

5293
    if (qemuProcessStartValidate(driver, vm, priv->qemuCaps, caps, flags) < 0)
5294 5295
        goto cleanup;

J
Jiri Denemark 已提交
5296 5297 5298 5299 5300
    /* 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");
5301
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm) < 0)
5302
        goto cleanup;
J
Jiri Denemark 已提交
5303

5304
    if (flags & VIR_QEMU_PROCESS_START_PRETEND) {
5305 5306 5307 5308
        if (qemuDomainSetPrivatePaths(driver, vm) < 0) {
            virDomainObjRemoveTransientDef(vm);
            goto cleanup;
        }
5309
    } else {
5310 5311 5312
        vm->def->id = qemuDriverAllocateID(driver);
        qemuDomainSetFakeReboot(driver, vm, false);
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_STARTING_UP);
J
Jiri Denemark 已提交
5313

5314 5315
        if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
            driver->inhibitCallback(true, driver->inhibitOpaque);
J
Jiri Denemark 已提交
5316

5317 5318 5319 5320 5321
        /* 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;
J
Jiri Denemark 已提交
5322

5323 5324
        if (qemuDomainSetPrivatePaths(driver, vm) < 0)
            goto stop;
5325 5326

        VIR_STEAL_PTR(priv->origCPU, origCPU);
5327
    }
5328

J
Jiri Denemark 已提交
5329 5330 5331
    ret = 0;

 cleanup:
5332
    virCPUDefFree(origCPU);
J
Jiri Denemark 已提交
5333 5334 5335 5336 5337 5338 5339 5340
    virObjectUnref(cfg);
    virObjectUnref(caps);
    return ret;

 stop:
    stopFlags = VIR_QEMU_PROCESS_STOP_NO_RELABEL;
    if (migration)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
5341
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, asyncJob, stopFlags);
J
Jiri Denemark 已提交
5342 5343 5344 5345
    goto cleanup;
}


5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356
/**
 * qemuProcessNetworkPrepareDevices
 */
static int
qemuProcessNetworkPrepareDevices(virDomainDefPtr def)
{
    int ret = -1;
    size_t i;

    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
5357
        virDomainNetType actualType;
5358 5359 5360 5361 5362

        /* If appropriate, grab a physical device from the configured
         * network's pool of devices, or resolve bridge device name
         * to the one defined in the network definition.
         */
5363
        if (virDomainNetAllocateActualDevice(def, net) < 0)
5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398
            goto cleanup;

        actualType = virDomainNetGetActualType(net);
        if (actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV &&
            net->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
            /* Each type='hostdev' network device must also have a
             * corresponding entry in the hostdevs array. For netdevs
             * that are hardcoded as type='hostdev', this is already
             * done by the parser, but for those allocated from a
             * network / determined at runtime, we need to do it
             * separately.
             */
            virDomainHostdevDefPtr hostdev = virDomainNetGetActualHostdev(net);
            virDomainHostdevSubsysPCIPtr pcisrc = &hostdev->source.subsys.u.pci;

            if (virDomainHostdevFind(def, hostdev, NULL) >= 0) {
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("PCI device %04x:%02x:%02x.%x "
                                 "allocated from network %s is already "
                                 "in use by domain %s"),
                               pcisrc->addr.domain, pcisrc->addr.bus,
                               pcisrc->addr.slot, pcisrc->addr.function,
                               net->data.network.name, def->name);
                goto cleanup;
            }
            if (virDomainHostdevInsert(def, hostdev) < 0)
                goto cleanup;
        }
    }
    ret = 0;
 cleanup:
    return ret;
}


5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414
/**
 * qemuProcessSetupVcpu:
 * @vm: domain object
 * @vcpuid: id of VCPU to set defaults
 *
 * This function sets resource properties (cgroups, affinity, scheduler) for a
 * vCPU. This function expects that the vCPU is online and the vCPU pids were
 * correctly detected at the point when it's called.
 *
 * Returns 0 on success, -1 on error.
 */
int
qemuProcessSetupVcpu(virDomainObjPtr vm,
                     unsigned int vcpuid)
{
    pid_t vcpupid = qemuDomainGetVcpuPid(vm, vcpuid);
5415
    virDomainVcpuDefPtr vcpu = virDomainDefGetVcpu(vm->def, vcpuid);
5416 5417 5418 5419 5420 5421 5422 5423 5424
    size_t i = 0;

    if (qemuProcessSetupPid(vm, vcpupid, VIR_CGROUP_THREAD_VCPU,
                            vcpuid, vcpu->cpumask,
                            vm->def->cputune.period,
                            vm->def->cputune.quota,
                            &vcpu->sched) < 0)
        return -1;

B
Bing Niu 已提交
5425 5426
    for (i = 0; i < vm->def->nresctrls; i++) {
        virDomainResctrlDefPtr ct = vm->def->resctrls[i];
5427 5428 5429 5430 5431 5432 5433

        if (virBitmapIsBitSet(ct->vcpus, vcpuid)) {
            if (virResctrlAllocAddPID(ct->alloc, vcpupid) < 0)
                return -1;
            break;
        }
    }
5434

5435
    return 0;
5436 5437 5438 5439 5440 5441
}


static int
qemuProcessSetupVcpus(virDomainObjPtr vm)
{
5442
    virDomainVcpuDefPtr vcpu;
5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487
    unsigned int maxvcpus = virDomainDefGetVcpusMax(vm->def);
    size_t i;

    if ((vm->def->cputune.period || vm->def->cputune.quota) &&
        !virCgroupHasController(((qemuDomainObjPrivatePtr) vm->privateData)->cgroup,
                                VIR_CGROUP_CONTROLLER_CPU)) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("cgroup cpu is required for scheduler tuning"));
        return -1;
    }

    if (!qemuDomainHasVcpuPids(vm)) {
        /* If any CPU has custom affinity that differs from the
         * VM default affinity, we must reject it */
        for (i = 0; i < maxvcpus; i++) {
            vcpu = virDomainDefGetVcpu(vm->def, i);

            if (!vcpu->online)
                continue;

            if (vcpu->cpumask &&
                !virBitmapEqual(vm->def->cpumask, vcpu->cpumask)) {
                virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                                _("cpu affinity is not supported"));
                return -1;
            }
        }

        return 0;
    }

    for (i = 0; i < maxvcpus; i++) {
        vcpu = virDomainDefGetVcpu(vm->def, i);

        if (!vcpu->online)
            continue;

        if (qemuProcessSetupVcpu(vm, i) < 0)
            return -1;
    }

    return 0;
}


5488 5489 5490 5491 5492
int
qemuProcessSetupIOThread(virDomainObjPtr vm,
                         virDomainIOThreadIDDefPtr iothread)
{

5493 5494
    return qemuProcessSetupPid(vm, iothread->thread_id,
                               VIR_CGROUP_THREAD_IOTHREAD,
5495
                               iothread->iothread_id,
5496
                               iothread->cpumask,
5497 5498
                               vm->def->cputune.iothread_period,
                               vm->def->cputune.iothread_quota,
5499
                               &iothread->sched);
5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518
}


static int
qemuProcessSetupIOThreads(virDomainObjPtr vm)
{
    size_t i;

    for (i = 0; i < vm->def->niothreadids; i++) {
        virDomainIOThreadIDDefPtr info = vm->def->iothreadids[i];

        if (qemuProcessSetupIOThread(vm, info) < 0)
            return -1;
    }

    return 0;
}


5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530
static int
qemuProcessValidateHotpluggableVcpus(virDomainDefPtr def)
{
    virDomainVcpuDefPtr vcpu;
    virDomainVcpuDefPtr subvcpu;
    qemuDomainVcpuPrivatePtr vcpupriv;
    unsigned int maxvcpus = virDomainDefGetVcpusMax(def);
    size_t i = 0;
    size_t j;
    virBitmapPtr ordermap = NULL;
    int ret = -1;

5531
    if (!(ordermap = virBitmapNew(maxvcpus + 1)))
5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547
        goto cleanup;

    /* validate:
     * - all hotpluggable entities to be hotplugged have the correct data
     * - vcpus belonging to a hotpluggable entity share configuration
     * - order of the hotpluggable entities is unique
     */
    for (i = 0; i < maxvcpus; i++) {
        vcpu = virDomainDefGetVcpu(def, i);
        vcpupriv = QEMU_DOMAIN_VCPU_PRIVATE(vcpu);

        /* skip over hotpluggable entities  */
        if (vcpupriv->vcpus == 0)
            continue;

        if (vcpu->order != 0) {
5548
            if (virBitmapIsBitSet(ordermap, vcpu->order)) {
5549
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
5550
                               _("duplicate vcpu order '%u'"), vcpu->order);
5551 5552 5553
                goto cleanup;
            }

5554 5555 5556 5557 5558 5559
            if (virBitmapSetBit(ordermap, vcpu->order)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("vcpu order '%u' exceeds vcpu count"),
                               vcpu->order);
                goto cleanup;
            }
5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575
        }

        for (j = i + 1; j < (i + vcpupriv->vcpus); j++) {
            subvcpu = virDomainDefGetVcpu(def, j);
            if (subvcpu->hotpluggable != vcpu->hotpluggable ||
                subvcpu->online != vcpu->online ||
                subvcpu->order != vcpu->order) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("vcpus '%zu' and '%zu' are in the same hotplug "
                                 "group but differ in configuration"), i, j);
                goto cleanup;
            }
        }

        if (vcpu->online && vcpu->hotpluggable == VIR_TRISTATE_BOOL_YES) {
            if ((vcpupriv->socket_id == -1 && vcpupriv->core_id == -1 &&
5576
                 vcpupriv->thread_id == -1 && vcpupriv->node_id == -1) ||
5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626
                !vcpupriv->type) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("vcpu '%zu' is missing hotplug data"), i);
                goto cleanup;
            }
        }
    }

    ret = 0;
 cleanup:
    virBitmapFree(ordermap);
    return ret;
}


static int
qemuDomainHasHotpluggableStartupVcpus(virDomainDefPtr def)
{
    size_t maxvcpus = virDomainDefGetVcpusMax(def);
    virDomainVcpuDefPtr vcpu;
    size_t i;

    for (i = 0; i < maxvcpus; i++) {
        vcpu = virDomainDefGetVcpu(def, i);

        if (vcpu->online && vcpu->hotpluggable == VIR_TRISTATE_BOOL_YES)
            return true;
    }

    return false;
}


static int
qemuProcessVcpusSortOrder(const void *a,
                          const void *b)
{
    virDomainVcpuDefPtr vcpua = *((virDomainVcpuDefPtr *)a);
    virDomainVcpuDefPtr vcpub = *((virDomainVcpuDefPtr *)b);

    return vcpua->order - vcpub->order;
}


static int
qemuProcessSetupHotpluggableVcpus(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm,
                                  qemuDomainAsyncJob asyncJob)
{
    unsigned int maxvcpus = virDomainDefGetVcpusMax(vm->def);
5627 5628
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuCgroupEmulatorAllNodesDataPtr emulatorCgroup = NULL;
5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660
    virDomainVcpuDefPtr vcpu;
    qemuDomainVcpuPrivatePtr vcpupriv;
    virJSONValuePtr vcpuprops = NULL;
    size_t i;
    int ret = -1;
    int rc;

    virDomainVcpuDefPtr *bootHotplug = NULL;
    size_t nbootHotplug = 0;

    for (i = 0; i < maxvcpus; i++) {
        vcpu = virDomainDefGetVcpu(vm->def, i);
        vcpupriv = QEMU_DOMAIN_VCPU_PRIVATE(vcpu);

        if (vcpu->hotpluggable == VIR_TRISTATE_BOOL_YES && vcpu->online &&
            vcpupriv->vcpus != 0) {
            if (virAsprintf(&vcpupriv->alias, "vcpu%zu", i) < 0)
                goto cleanup;

            if (VIR_APPEND_ELEMENT(bootHotplug, nbootHotplug, vcpu) < 0)
                goto cleanup;
        }
    }

    if (nbootHotplug == 0) {
        ret = 0;
        goto cleanup;
    }

    qsort(bootHotplug, nbootHotplug, sizeof(*bootHotplug),
          qemuProcessVcpusSortOrder);

5661 5662 5663
    if (qemuCgroupEmulatorAllNodesAllow(priv->cgroup, &emulatorCgroup) < 0)
        goto cleanup;

5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687
    for (i = 0; i < nbootHotplug; i++) {
        vcpu = bootHotplug[i];

        if (!(vcpuprops = qemuBuildHotpluggableCPUProps(vcpu)))
            goto cleanup;

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

        rc = qemuMonitorAddDeviceArgs(qemuDomainGetMonitor(vm), vcpuprops);
        vcpuprops = NULL;

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

        if (rc < 0)
            goto cleanup;

        virJSONValueFree(vcpuprops);
    }

    ret = 0;

 cleanup:
5688
    qemuCgroupEmulatorAllNodesRestore(emulatorCgroup);
5689 5690 5691 5692 5693 5694
    VIR_FREE(bootHotplug);
    virJSONValueFree(vcpuprops);
    return ret;
}


5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733
static int
qemuProcessUpdateGuestCPU(virDomainDefPtr def,
                          virQEMUCapsPtr qemuCaps,
                          virCapsPtr caps,
                          unsigned int flags)
{
    int ret = -1;

    if (!def->cpu)
        return 0;

    /* nothing to do if only topology part of CPU def is used */
    if (def->cpu->mode == VIR_CPU_MODE_CUSTOM && !def->cpu->model)
        return 0;

    /* Old libvirt added host CPU model to host-model CPUs for migrations,
     * while new libvirt just turns host-model into custom mode. We need
     * to fix the mode to maintain backward compatibility and to avoid
     * the CPU model to be replaced in virCPUUpdate.
     */
    if (!(flags & VIR_QEMU_PROCESS_START_NEW) &&
        ARCH_IS_X86(def->os.arch) &&
        def->cpu->mode == VIR_CPU_MODE_HOST_MODEL &&
        def->cpu->model) {
        def->cpu->mode = VIR_CPU_MODE_CUSTOM;
    }

    if (!virQEMUCapsIsCPUModeSupported(qemuCaps, caps, def->virtType,
                                       def->cpu->mode)) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       _("CPU mode '%s' for %s %s domain on %s host is not "
                         "supported by hypervisor"),
                       virCPUModeTypeToString(def->cpu->mode),
                       virArchToString(def->os.arch),
                       virDomainVirtTypeToString(def->virtType),
                       virArchToString(caps->host.arch));
        return -1;
    }

5734 5735 5736
    if (virCPUConvertLegacy(caps->host.arch, def->cpu) < 0)
        return -1;

5737 5738 5739 5740
    /* nothing to update for host-passthrough */
    if (def->cpu->mode == VIR_CPU_MODE_HOST_PASSTHROUGH)
        return 0;

5741 5742
    if (def->cpu->check == VIR_CPU_CHECK_PARTIAL &&
        virCPUCompare(caps->host.arch,
5743
                      virQEMUCapsGetHostModel(qemuCaps, def->virtType,
5744
                                              VIR_QEMU_CAPS_HOST_CPU_FULL),
5745 5746
                      def->cpu, true) < 0)
        return -1;
5747 5748

    if (virCPUUpdate(def->os.arch, def->cpu,
5749
                     virQEMUCapsGetHostModel(qemuCaps, def->virtType,
5750
                                             VIR_QEMU_CAPS_HOST_CPU_MIGRATABLE)) < 0)
5751 5752
        goto cleanup;

5753 5754
    if (virCPUTranslate(def->os.arch, def->cpu,
                        virQEMUCapsGetCPUDefinitions(qemuCaps, def->virtType)) < 0)
5755 5756 5757 5758 5759 5760 5761 5762 5763 5764
        goto cleanup;

    def->cpu->fallback = VIR_CPU_FALLBACK_FORBID;
    ret = 0;

 cleanup:
    return ret;
}


5765 5766 5767 5768 5769 5770
static int
qemuProcessPrepareDomainNUMAPlacement(virDomainObjPtr vm,
                                      virCapsPtr caps)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char *nodeset = NULL;
5771 5772
    virBitmapPtr numadNodeset = NULL;
    virBitmapPtr hostMemoryNodeset = NULL;
5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785 5786
    int ret = -1;

    /* Get the advisory nodeset from numad if 'placement' of
     * either <vcpu> or <numatune> is 'auto'.
     */
    if (!virDomainDefNeedsPlacementAdvice(vm->def))
        return 0;

    nodeset = virNumaGetAutoPlacementAdvice(virDomainDefGetVcpus(vm->def),
                                            virDomainDefGetMemoryTotal(vm->def));

    if (!nodeset)
        goto cleanup;

5787 5788 5789
    if (!(hostMemoryNodeset = virNumaGetHostMemoryNodeset()))
        goto cleanup;

5790 5791
    VIR_DEBUG("Nodeset returned from numad: %s", nodeset);

5792
    if (virBitmapParse(nodeset, &numadNodeset, VIR_DOMAIN_CPUMASK_LEN) < 0)
5793 5794
        goto cleanup;

5795 5796 5797 5798
    /* numad may return a nodeset that only contains cpus but cgroups don't play
     * well with that. Set the autoCpuset from all cpus from that nodeset, but
     * assign autoNodeset only with nodes containing memory. */
    if (!(priv->autoCpuset = virCapabilitiesGetCpusForNodemask(caps, numadNodeset)))
5799 5800
        goto cleanup;

5801 5802 5803 5804
    virBitmapIntersect(numadNodeset, hostMemoryNodeset);

    VIR_STEAL_PTR(priv->autoNodeset, numadNodeset);

5805 5806 5807 5808
    ret = 0;

 cleanup:
    VIR_FREE(nodeset);
5809 5810
    virBitmapFree(numadNodeset);
    virBitmapFree(hostMemoryNodeset);
5811 5812 5813 5814
    return ret;
}


5815
static int
5816
qemuProcessPrepareDomainStorage(virQEMUDriverPtr driver,
5817
                                virDomainObjPtr vm,
5818
                                qemuDomainObjPrivatePtr priv,
5819
                                virQEMUDriverConfigPtr cfg,
5820 5821 5822 5823 5824 5825 5826 5827 5828
                                unsigned int flags)
{
    size_t i;
    bool cold_boot = flags & VIR_QEMU_PROCESS_START_COLD;

    for (i = vm->def->ndisks; i > 0; i--) {
        size_t idx = i - 1;
        virDomainDiskDefPtr disk = vm->def->disks[idx];

5829
        if (virDomainDiskTranslateSourcePool(disk) < 0) {
5830 5831 5832 5833 5834 5835
            if (qemuDomainCheckDiskStartupPolicy(driver, vm, idx, cold_boot) < 0)
                return -1;

            /* disk source was dropped */
            continue;
        }
5836

5837
        if (qemuDomainPrepareDiskSource(disk, priv, cfg) < 0)
5838
            return -1;
5839 5840 5841 5842 5843 5844
    }

    return 0;
}


5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862 5863 5864
static void
qemuProcessPrepareAllowReboot(virDomainObjPtr vm)
{
    virDomainDefPtr def = vm->def;
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->allowReboot != VIR_TRISTATE_BOOL_ABSENT)
        return;

    if (def->onReboot == VIR_DOMAIN_LIFECYCLE_ACTION_DESTROY &&
        def->onPoweroff == VIR_DOMAIN_LIFECYCLE_ACTION_DESTROY &&
        (def->onCrash == VIR_DOMAIN_LIFECYCLE_ACTION_DESTROY ||
         def->onCrash == VIR_DOMAIN_LIFECYCLE_ACTION_COREDUMP_DESTROY)) {
        priv->allowReboot = VIR_TRISTATE_BOOL_NO;
    } else {
        priv->allowReboot = VIR_TRISTATE_BOOL_YES;
    }
}


5865
/**
5866 5867 5868 5869
 * qemuProcessPrepareDomain:
 * @driver: qemu driver
 * @vm: domain object
 * @flags: qemuProcessStartFlags
5870 5871 5872 5873 5874 5875 5876 5877 5878 5879 5880
 *
 * This function groups all code that modifies only live XML of a domain which
 * is about to start and it's the only place to do those modifications.
 *
 * Flag VIR_QEMU_PROCESS_START_PRETEND tells, that we don't want to actually
 * start the domain but create a valid qemu command.  If some code shouldn't be
 * executed in this case, make sure to check this flag.
 *
 * TODO: move all XML modification from qemuBuildCommandLine into this function
 */
int
5881
qemuProcessPrepareDomain(virQEMUDriverPtr driver,
5882 5883 5884 5885 5886 5887
                         virDomainObjPtr vm,
                         unsigned int flags)
{
    int ret = -1;
    size_t i;
    qemuDomainObjPrivatePtr priv = vm->privateData;
5888
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5889 5890 5891 5892 5893
    virCapsPtr caps;

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

5894 5895 5896 5897
    priv->machineName = qemuDomainGetMachineName(vm);
    if (!priv->machineName)
        goto cleanup;

5898 5899 5900 5901
    if (!(flags & VIR_QEMU_PROCESS_START_PRETEND)) {
        /* If you are using a SecurityDriver with dynamic labelling,
           then generate a security label for isolation */
        VIR_DEBUG("Generating domain security label (if required)");
5902
        if (qemuSecurityGenLabel(driver->securityManager, vm->def) < 0) {
5903 5904 5905 5906 5907
            virDomainAuditSecurityLabel(vm, false);
            goto cleanup;
        }
        virDomainAuditSecurityLabel(vm, true);

5908 5909
        if (qemuProcessPrepareDomainNUMAPlacement(vm, caps) < 0)
            goto cleanup;
5910 5911
    }

5912 5913 5914 5915 5916 5917 5918
    /* Whether we should use virtlogd as stdio handler for character
     * devices source backend. */
    if (cfg->stdioLogD &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_CHARDEV_FILE_APPEND)) {
        priv->chardevStdioLogd = true;
    }

5919 5920
    qemuProcessPrepareAllowReboot(vm);

5921 5922 5923 5924 5925 5926 5927 5928 5929
    /* clear the 'blockdev' capability for VMs which have disks that need
     * -drive or which have floppies where we can't reliably get the QOM path */
    for (i = 0; i < vm->def->ndisks; i++) {
        if (qemuDiskBusNeedsDriveArg(vm->def->disks[i]->bus)) {
            virQEMUCapsClear(priv->qemuCaps, QEMU_CAPS_BLOCKDEV);
            break;
        }
    }

5930 5931 5932 5933 5934 5935 5936
    /*
     * 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
     * we also need to populate the PCI address set cache for later
     * use in hotplug
     */
5937
    VIR_DEBUG("Assigning domain PCI addresses");
5938 5939
    if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, driver, vm,
                                   !!(flags & VIR_QEMU_PROCESS_START_NEW))) < 0) {
5940
        goto cleanup;
5941
    }
5942

5943 5944 5945
    if (qemuAssignDeviceAliases(vm->def, priv->qemuCaps) < 0)
        goto cleanup;

5946
    VIR_DEBUG("Setting graphics devices");
5947
    if (qemuProcessSetupGraphics(driver, vm, flags) < 0)
5948 5949
        goto cleanup;

5950 5951 5952 5953
    VIR_DEBUG("Create domain masterKey");
    if (qemuDomainMasterKeyCreate(vm) < 0)
        goto cleanup;

5954
    VIR_DEBUG("Setting up storage");
5955
    if (qemuProcessPrepareDomainStorage(driver, vm, priv, cfg, flags) < 0)
5956 5957
        goto cleanup;

5958
    VIR_DEBUG("Prepare chardev source backends for TLS");
5959
    qemuDomainPrepareChardevSource(vm->def, cfg);
5960

5961
    VIR_DEBUG("Add secrets to hostdevs and chardevs");
5962
    if (qemuDomainSecretPrepare(driver, vm) < 0)
5963 5964
        goto cleanup;

5965 5966 5967 5968 5969 5970
    for (i = 0; i < vm->def->nchannels; i++) {
        if (qemuDomainPrepareChannel(vm->def->channels[i],
                                     priv->channelTargetDir) < 0)
            goto cleanup;
    }

5971
    if (!(priv->monConfig = virDomainChrSourceDefNew(NULL)))
5972 5973 5974 5975 5976 5977
        goto cleanup;

    VIR_DEBUG("Preparing monitor state");
    if (qemuProcessPrepareMonitorChr(priv->monConfig, priv->libDir) < 0)
        goto cleanup;

5978
    priv->monJSON = true;
5979 5980 5981 5982
    priv->monError = false;
    priv->monStart = 0;
    priv->gotShutdown = false;

5983 5984 5985 5986
    VIR_DEBUG("Updating guest CPU definition");
    if (qemuProcessUpdateGuestCPU(vm->def, priv->qemuCaps, caps, flags) < 0)
        goto cleanup;

5987 5988 5989 5990 5991
    for (i = 0; i < vm->def->nshmems; i++) {
        if (qemuDomainPrepareShmemChardev(vm->def->shmems[i]) < 0)
            goto cleanup;
    }

5992 5993 5994
    ret = 0;
 cleanup:
    virObjectUnref(caps);
5995
    virObjectUnref(cfg);
5996 5997 5998 5999
    return ret;
}


6000
static int
6001 6002 6003
qemuProcessSEVCreateFile(const char *configDir,
                         const char *name,
                         const char *data)
6004 6005
{
    char *configFile;
6006
    int ret = -1;
6007 6008 6009 6010 6011 6012 6013

    if (!(configFile = virFileBuildPath(configDir, name, ".base64")))
        return -1;

    if (virFileRewriteStr(configFile, S_IRUSR | S_IWUSR, data) < 0) {
        virReportSystemError(errno, _("failed to write data to config '%s'"),
                             configFile);
6014
        goto cleanup;
6015 6016
    }

6017
    ret = 0;
6018
 cleanup:
6019
    VIR_FREE(configFile);
6020
    return ret;
6021 6022 6023 6024
}


static int
J
Ján Tomko 已提交
6025
qemuProcessPrepareSEVGuestInput(virDomainObjPtr vm)
6026 6027 6028 6029
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
    virQEMUCapsPtr qemuCaps = priv->qemuCaps;
6030
    virDomainSEVDefPtr sev = def->sev;
6031 6032 6033 6034

    if (!sev)
        return 0;

6035
    VIR_DEBUG("Preparing SEV guest");
6036 6037 6038 6039 6040 6041 6042 6043 6044

    if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_SEV_GUEST)) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                        _("Domain %s asked for 'sev' launch but this "
                          "QEMU does not support SEV feature"), vm->def->name);
        return -1;
    }

    if (sev->dh_cert) {
6045
        if (qemuProcessSEVCreateFile(priv->libDir, "dh_cert", sev->dh_cert) < 0)
6046 6047 6048 6049
            return -1;
    }

    if (sev->session) {
6050
        if (qemuProcessSEVCreateFile(priv->libDir, "session", sev->session) < 0)
6051 6052 6053 6054 6055 6056 6057
            return -1;
    }

    return 0;
}


6058 6059 6060 6061 6062
static int
qemuProcessPrepareHostStorage(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              unsigned int flags)
{
6063
    qemuDomainObjPrivatePtr priv = vm->privateData;
6064 6065
    size_t i;
    bool cold_boot = flags & VIR_QEMU_PROCESS_START_COLD;
6066
    bool blockdev = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV);
6067 6068 6069 6070 6071 6072 6073 6074

    for (i = vm->def->ndisks; i > 0; i--) {
        size_t idx = i - 1;
        virDomainDiskDefPtr disk = vm->def->disks[idx];

        if (virStorageSourceIsEmpty(disk->src))
            continue;

6075 6076 6077
        /* backing chain needs to be redetected if we aren't using blockdev */
        if (!blockdev)
            virStorageSourceBackingStoreClear(disk->src);
6078 6079

        if (qemuDomainDetermineDiskChain(driver, vm, disk, true) >= 0)
6080 6081 6082 6083 6084 6085 6086 6087 6088 6089 6090 6091
            continue;

        if (qemuDomainCheckDiskStartupPolicy(driver, vm, idx, cold_boot) >= 0)
            continue;

        return -1;
    }

    return 0;
}


J
Ján Tomko 已提交
6092
int
6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121
qemuProcessOpenVhostVsock(virDomainVsockDefPtr vsock)
{
    qemuDomainVsockPrivatePtr priv = (qemuDomainVsockPrivatePtr)vsock->privateData;
    const char *vsock_path = "/dev/vhost-vsock";
    int fd;

    if ((fd = open(vsock_path, O_RDWR)) < 0) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       "%s", _("unable to open vhost-vsock device"));
        return -1;
    }

    if (vsock->auto_cid == VIR_TRISTATE_BOOL_YES) {
        if (virVsockAcquireGuestCid(fd, &vsock->guest_cid) < 0)
            goto error;
    } else {
        if (virVsockSetGuestCid(fd, vsock->guest_cid) < 0)
            goto error;
    }

    priv->vhostfd = fd;
    return 0;

 error:
    VIR_FORCE_CLOSE(fd);
    return -1;
}


J
Jiri Denemark 已提交
6122
/**
6123 6124 6125
 * qemuProcessPrepareHost:
 * @driver: qemu driver
 * @vm: domain object
6126
 * @flags: qemuProcessStartFlags
J
Jiri Denemark 已提交
6127
 *
6128 6129 6130
 * This function groups all code that modifies host system (which also may
 * update live XML) to prepare environment for a domain which is about to start
 * and it's the only place to do those modifications.
J
Jiri Denemark 已提交
6131
 *
6132
 * TODO: move all host modification from qemuBuildCommandLine into this function
J
Jiri Denemark 已提交
6133 6134
 */
int
6135 6136
qemuProcessPrepareHost(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
6137
                       unsigned int flags)
6138
{
6139
    int ret = -1;
6140
    unsigned int hostdev_flags = 0;
6141 6142
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
6143

6144
    if (qemuPrepareNVRAM(cfg, vm) < 0)
J
Jiri Denemark 已提交
6145
        goto cleanup;
6146

6147 6148 6149 6150
    if (vm->def->vsock) {
        if (qemuProcessOpenVhostVsock(vm->def->vsock) < 0)
            goto cleanup;
    }
6151 6152 6153 6154 6155
    /* 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");
6156
    if (qemuProcessNetworkPrepareDevices(vm->def) < 0)
J
Jiri Denemark 已提交
6157
        goto cleanup;
6158

6159
    /* Must be run before security labelling */
6160
    VIR_DEBUG("Preparing host devices");
6161 6162
    if (!cfg->relaxedACS)
        hostdev_flags |= VIR_HOSTDEV_STRICT_ACS_CHECK;
6163
    if (flags & VIR_QEMU_PROCESS_START_NEW)
6164
        hostdev_flags |= VIR_HOSTDEV_COLD_BOOT;
6165 6166
    if (qemuHostdevPrepareDomainDevices(driver, vm->def, priv->qemuCaps,
                                        hostdev_flags) < 0)
J
Jiri Denemark 已提交
6167
        goto cleanup;
6168

6169
    VIR_DEBUG("Preparing chr devices");
6170 6171 6172 6173
    if (virDomainChrDefForeach(vm->def,
                               true,
                               qemuProcessPrepareChardevDevice,
                               NULL) < 0)
J
Jiri Denemark 已提交
6174
        goto cleanup;
6175

6176
    if (qemuProcessBuildDestroyMemoryPaths(driver, vm, NULL, true) < 0)
6177
        goto cleanup;
6178

6179 6180
    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
6181
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
6182
    qemuRemoveCgroup(vm);
6183

6184
    if (virFileMakePath(cfg->logDir) < 0) {
6185 6186
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
6187
                             cfg->logDir);
J
Jiri Denemark 已提交
6188
        goto cleanup;
6189 6190
    }

6191 6192 6193 6194 6195 6196 6197 6198 6199 6200 6201 6202 6203 6204 6205 6206 6207 6208 6209 6210 6211 6212 6213
    VIR_FREE(priv->pidfile);
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, vm->def->name))) {
        virReportSystemError(errno,
                             "%s", _("Failed to build pidfile path."));
        goto cleanup;
    }

    if (unlink(priv->pidfile) < 0 &&
        errno != ENOENT) {
        virReportSystemError(errno,
                             _("Cannot remove stale PID file %s"),
                             priv->pidfile);
        goto cleanup;
    }

    /*
     * Create all per-domain directories in order to make sure domain
     * with any possible seclabels can access it.
     */
    if (qemuProcessMakeDir(driver, vm, priv->libDir) < 0 ||
        qemuProcessMakeDir(driver, vm, priv->channelTargetDir) < 0)
        goto cleanup;

6214 6215
    VIR_DEBUG("Write domain masterKey");
    if (qemuDomainWriteMasterKeyFile(driver, vm) < 0)
J
John Ferlan 已提交
6216 6217
        goto cleanup;

6218 6219 6220 6221
    VIR_DEBUG("Preparing disks (host)");
    if (qemuProcessPrepareHostStorage(driver, vm, flags) < 0)
        goto cleanup;

6222 6223 6224 6225
    VIR_DEBUG("Preparing external devices");
    if (qemuExtDevicesPrepareHost(driver, vm->def) < 0)
        goto cleanup;

J
Ján Tomko 已提交
6226
    if (qemuProcessPrepareSEVGuestInput(vm) < 0)
6227 6228
        goto cleanup;

6229 6230 6231 6232 6233 6234 6235
    ret = 0;
 cleanup:
    virObjectUnref(cfg);
    return ret;
}


6236 6237 6238 6239 6240 6241 6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256 6257 6258 6259 6260 6261 6262 6263 6264 6265 6266 6267 6268
/**
 * qemuProcessGenID:
 * @vm: Pointer to domain object
 * @flags: qemuProcessStartFlags
 *
 * If this domain is requesting to use genid, then update the GUID
 * value if the VIR_QEMU_PROCESS_START_GEN_VMID flag is set. This
 * flag is set on specific paths during domain start processing when
 * there is the possibility that the VM is potentially re-executing
 * something that has already been executed before.
 */
static int
qemuProcessGenID(virDomainObjPtr vm,
                 unsigned int flags)
{
    if (!vm->def->genidRequested)
        return 0;

    /* If we are coming from a path where we must provide a new gen id
     * value regardless of whether it was previously generated or provided,
     * then generate a new GUID value before we build the command line. */
    if (flags & VIR_QEMU_PROCESS_START_GEN_VMID) {
        if (virUUIDGenerate(vm->def->genid) < 0) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("failed to regenerate genid"));
            return -1;
        }
    }

    return 0;
}


6269 6270 6271 6272 6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 6285 6286 6287 6288 6289 6290 6291 6292 6293 6294 6295 6296 6297 6298 6299 6300 6301 6302 6303 6304 6305 6306 6307 6308 6309 6310 6311 6312 6313 6314
/**
 * qemuProcessSetupDiskThrottlingBlockdev:
 *
 * Sets up disk trottling for -blockdev via block_set_io_throttle monitor
 * command. This hack should be replaced by proper use of the 'throttle'
 * blockdev driver in qemu once it will support changing of the throttle group.
 */
static int
qemuProcessSetupDiskThrottlingBlockdev(virQEMUDriverPtr driver,
                                       virDomainObjPtr vm,
                                       qemuDomainAsyncJob asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    size_t i;
    int ret = -1;

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

    VIR_DEBUG("Setting up disk throttling for -blockdev via block_set_io_throttle");

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

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

        if (!qemuDiskConfigBlkdeviotuneEnabled(disk))
            continue;

        if (qemuMonitorSetBlockIoThrottle(qemuDomainGetMonitor(vm), NULL,
                                          diskPriv->qomName, &disk->blkdeviotune,
                                          true, true, true) < 0)
            goto cleanup;
    }

    ret = 0;

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


6315 6316 6317 6318 6319 6320 6321 6322 6323 6324 6325 6326 6327 6328 6329 6330 6331 6332 6333 6334 6335 6336 6337 6338 6339 6340 6341 6342 6343 6344 6345 6346 6347 6348
/**
 * 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)
{
    int ret = -1;
    int rv;
    int logfile = -1;
    qemuDomainLogContextPtr logCtxt = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCommandPtr cmd = NULL;
    struct qemuProcessHookData hookData;
    virQEMUDriverConfigPtr cfg;
    virCapsPtr caps = NULL;
    size_t nnicindexes = 0;
    int *nicindexes = NULL;
Q
Qiaowei Ren 已提交
6349
    size_t i;
6350

6351
    VIR_DEBUG("conn=%p driver=%p vm=%p name=%s if=%d asyncJob=%d "
6352 6353 6354
              "incoming.launchURI=%s incoming.deferredURI=%s "
              "incoming.fd=%d incoming.path=%s "
              "snapshot=%p vmop=%d flags=0x%x",
6355
              conn, driver, vm, vm->def->name, vm->def->id, asyncJob,
6356 6357 6358 6359 6360 6361 6362 6363 6364 6365
              NULLSTR(incoming ? incoming->launchURI : NULL),
              NULLSTR(incoming ? incoming->deferredURI : NULL),
              incoming ? incoming->fd : -1,
              NULLSTR(incoming ? incoming->path : NULL),
              snapshot, vmop, flags);

    /* Okay, these are just internal flags,
     * but doesn't hurt to check */
    virCheckFlags(VIR_QEMU_PROCESS_START_COLD |
                  VIR_QEMU_PROCESS_START_PAUSED |
6366
                  VIR_QEMU_PROCESS_START_AUTODESTROY |
6367 6368
                  VIR_QEMU_PROCESS_START_NEW |
                  VIR_QEMU_PROCESS_START_GEN_VMID, -1);
6369 6370 6371

    cfg = virQEMUDriverGetConfig(driver);

6372 6373 6374 6375 6376 6377
    if ((flags & VIR_QEMU_PROCESS_START_AUTODESTROY) && !conn) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Domain autodestroy requires a connection handle"));
        return -1;
    }

6378 6379 6380 6381 6382 6383 6384 6385
    hookData.vm = vm;
    hookData.driver = driver;
    /* We don't increase cfg's reference counter here. */
    hookData.cfg = cfg;

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

6386
    VIR_DEBUG("Creating domain log file");
6387 6388
    if (!(logCtxt = qemuDomainLogContextNew(driver, vm,
                                            QEMU_DOMAIN_LOG_CONTEXT_MODE_START)))
J
Jiri Denemark 已提交
6389
        goto cleanup;
6390
    logfile = qemuDomainLogContextGetWriteFD(logCtxt);
6391

6392 6393 6394
    if (qemuProcessGenID(vm, flags) < 0)
        goto cleanup;

6395 6396 6397
    if (qemuExtDevicesStart(driver, vm->def, logCtxt) < 0)
        goto cleanup;

6398
    VIR_DEBUG("Building emulator command line");
6399
    if (!(cmd = qemuBuildCommandLine(driver,
6400
                                     qemuDomainLogContextGetManager(logCtxt),
6401
                                     driver->securityManager,
6402
                                     vm,
6403 6404
                                     incoming ? incoming->launchURI : NULL,
                                     snapshot, vmop,
J
Ján Tomko 已提交
6405
                                     false,
6406
                                     qemuCheckFips(),
6407
                                     &nnicindexes, &nicindexes)))
J
Jiri Denemark 已提交
6408
        goto cleanup;
6409

6410 6411
    if (incoming && incoming->fd != -1)
        virCommandPassFD(cmd, incoming->fd, 0);
6412

6413
    /* now that we know it is about to start call the hook if present */
6414 6415 6416
    if (qemuProcessStartHook(driver, vm,
                             VIR_HOOK_QEMU_OP_START,
                             VIR_HOOK_SUBOP_BEGIN) < 0)
J
Jiri Denemark 已提交
6417
        goto cleanup;
6418

6419
    qemuLogOperation(vm, "starting up", cmd, logCtxt);
6420

6421
    qemuDomainObjCheckTaint(driver, vm, logCtxt);
6422

6423
    qemuDomainLogContextMarkPosition(logCtxt);
6424

6425 6426 6427 6428 6429
    VIR_DEBUG("Building mount namespace");

    if (qemuDomainCreateNamespace(driver, vm) < 0)
        goto cleanup;

6430
    VIR_DEBUG("Clear emulator capabilities: %d",
6431 6432
              cfg->clearEmulatorCapabilities);
    if (cfg->clearEmulatorCapabilities)
6433 6434
        virCommandClearCaps(cmd);

6435 6436
    VIR_DEBUG("Setting up raw IO");
    if (qemuProcessSetupRawIO(driver, vm, cmd) < 0)
J
Jiri Denemark 已提交
6437
        goto cleanup;
6438

6439
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);
6440 6441
    virCommandSetMaxProcesses(cmd, cfg->maxProcesses);
    virCommandSetMaxFiles(cmd, cfg->maxFiles);
6442
    virCommandSetMaxCoreSize(cmd, cfg->maxCore);
6443
    virCommandSetUmask(cmd, 0x002);
6444

6445
    VIR_DEBUG("Setting up security labelling");
6446 6447
    if (qemuSecuritySetChildProcessLabel(driver->securityManager,
                                         vm->def, cmd) < 0)
J
Jiri Denemark 已提交
6448
        goto cleanup;
6449

6450 6451 6452
    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
6453
    virCommandSetPidFile(cmd, priv->pidfile);
6454
    virCommandDaemonize(cmd);
6455
    virCommandRequireHandshake(cmd);
6456

6457
    if (qemuSecurityPreFork(driver->securityManager) < 0)
J
Jiri Denemark 已提交
6458
        goto cleanup;
6459
    rv = virCommandRun(cmd, NULL);
6460
    qemuSecurityPostFork(driver->securityManager);
6461

E
Eric Blake 已提交
6462
    /* wait for qemu process to show up */
6463
    if (rv == 0) {
6464
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
6465 6466
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
6467
            rv = -1;
6468
        }
M
Michal Privoznik 已提交
6469
        VIR_DEBUG("QEMU vm=%p name=%s running with pid=%lld",
6470
                  vm, vm->def->name, (long long)vm->pid);
6471 6472 6473
    } else {
        VIR_DEBUG("QEMU vm=%p name=%s failed to spawn",
                  vm, vm->def->name);
6474 6475
    }

6476
    VIR_DEBUG("Writing early domain status to disk");
6477
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
J
Jiri Denemark 已提交
6478
        goto cleanup;
6479

6480 6481
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
6482
        /* Read errors from child that occurred between fork and exec. */
6483 6484
        qemuProcessReportLogError(logCtxt,
                                  _("Process exited prior to exec"));
J
Jiri Denemark 已提交
6485
        goto cleanup;
6486 6487
    }

6488
    VIR_DEBUG("Setting up domain cgroup (if required)");
6489
    if (qemuSetupCgroup(vm, nnicindexes, nicindexes) < 0)
J
Jiri Denemark 已提交
6490
        goto cleanup;
6491

6492 6493 6494 6495
    if (!(priv->perf = virPerfNew()))
        goto cleanup;

    for (i = 0; i < VIR_PERF_EVENT_LAST; i++) {
6496
        if (vm->def->perf.events[i] == VIR_TRISTATE_BOOL_YES &&
6497 6498
            virPerfEventEnable(priv->perf, i, vm->pid) < 0)
            goto cleanup;
Q
Qiaowei Ren 已提交
6499
    }
6500

6501 6502 6503 6504 6505 6506
    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
    if (!vm->def->cputune.emulatorpin &&
        qemuProcessInitCpuAffinity(vm) < 0)
        goto cleanup;

6507 6508 6509 6510
    VIR_DEBUG("Setting emulator tuning/settings");
    if (qemuProcessSetupEmulator(vm) < 0)
        goto cleanup;

6511 6512 6513 6514
    VIR_DEBUG("Setting cgroup for external devices (if required)");
    if (qemuSetupCgroupForExtDevices(vm, driver) < 0)
        goto cleanup;

6515
    VIR_DEBUG("Setting up resctrl");
6516 6517 6518
    if (qemuProcessResctrlCreate(driver, vm) < 0)
        goto cleanup;

6519
    VIR_DEBUG("Setting up managed PR daemon");
6520 6521
    if (virDomainDefHasManagedPR(vm->def) &&
        qemuProcessStartManagedPRDaemon(vm) < 0)
6522 6523
        goto cleanup;

6524
    VIR_DEBUG("Setting domain security labels");
6525 6526 6527 6528
    if (qemuSecuritySetAllLabel(driver,
                                vm,
                                incoming ? incoming->path : NULL) < 0)
        goto cleanup;
6529

6530
    /* Security manager labeled all devices, therefore
J
Jiri Denemark 已提交
6531 6532 6533 6534
     * if any operation from now on fails, we need to ask the caller to
     * restore labels.
     */
    ret = -2;
6535

J
Jiri Denemark 已提交
6536
    if (incoming && incoming->fd != -1) {
6537 6538 6539 6540 6541 6542 6543
        /* 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
6544
        if (fstat(incoming->fd, &stdin_sb) < 0) {
6545
            virReportSystemError(errno,
J
Jiri Denemark 已提交
6546 6547
                                 _("cannot stat fd %d"), incoming->fd);
            goto cleanup;
6548 6549
        }
        if (S_ISFIFO(stdin_sb.st_mode) &&
6550 6551
            qemuSecuritySetImageFDLabel(driver->securityManager,
                                        vm->def, incoming->fd) < 0)
J
Jiri Denemark 已提交
6552
            goto cleanup;
6553 6554 6555
    }

    VIR_DEBUG("Labelling done, completing handshake to child");
6556
    if (virCommandHandshakeNotify(cmd) < 0)
J
Jiri Denemark 已提交
6557
        goto cleanup;
6558 6559
    VIR_DEBUG("Handshake complete, child running");

6560
    if (rv == -1) /* The VM failed to start; tear filters before taps */
6561 6562
        virDomainConfVMNWFilterTeardown(vm);

6563
    if (rv == -1) /* The VM failed to start */
J
Jiri Denemark 已提交
6564
        goto cleanup;
6565

6566
    VIR_DEBUG("Waiting for monitor to show up");
6567
    if (qemuProcessWaitForMonitor(driver, vm, asyncJob, logCtxt) < 0)
J
Jiri Denemark 已提交
6568
        goto cleanup;
6569

6570 6571
    if (qemuConnectAgent(driver, vm) < 0)
        goto cleanup;
D
Daniel P. Berrange 已提交
6572

6573
    VIR_DEBUG("Verifying and updating provided guest CPU");
6574
    if (qemuProcessUpdateAndVerifyCPU(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6575
        goto cleanup;
6576

6577
    VIR_DEBUG("Setting up post-init cgroup restrictions");
6578
    if (qemuSetupCpusetMems(vm) < 0)
J
Jiri Denemark 已提交
6579
        goto cleanup;
6580

6581 6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592
    VIR_DEBUG("setting up hotpluggable cpus");
    if (qemuDomainHasHotpluggableStartupVcpus(vm->def)) {
        if (qemuDomainRefreshVcpuInfo(driver, vm, asyncJob, false) < 0)
            goto cleanup;

        if (qemuProcessValidateHotpluggableVcpus(vm->def) < 0)
            goto cleanup;

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

6593
    VIR_DEBUG("Refreshing VCPU info");
6594
    if (qemuDomainRefreshVcpuInfo(driver, vm, asyncJob, false) < 0)
J
Jiri Denemark 已提交
6595
        goto cleanup;
6596

6597 6598 6599
    if (qemuDomainValidateVcpuInfo(vm) < 0)
        goto cleanup;

6600 6601
    qemuDomainVcpuPersistOrder(vm->def);

6602 6603
    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6604
        goto cleanup;
6605

6606 6607 6608 6609
    VIR_DEBUG("Setting global CPU cgroup (if required)");
    if (qemuSetupGlobalCpuCgroup(vm) < 0)
        goto cleanup;

6610 6611
    VIR_DEBUG("Setting vCPU tuning/settings");
    if (qemuProcessSetupVcpus(vm) < 0)
J
Jiri Denemark 已提交
6612
        goto cleanup;
6613

6614 6615
    VIR_DEBUG("Setting IOThread tuning/settings");
    if (qemuProcessSetupIOThreads(vm) < 0)
J
Jiri Denemark 已提交
6616
        goto cleanup;
6617

6618
    VIR_DEBUG("Setting any required VM passwords");
6619
    if (qemuProcessInitPasswords(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6620
        goto cleanup;
6621

6622 6623 6624 6625
    /* set default link states */
    /* qemu doesn't support setting this on the command line, so
     * enter the monitor */
    VIR_DEBUG("Setting network link states");
6626
    if (qemuProcessSetLinkStates(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6627
        goto cleanup;
6628

6629
    VIR_DEBUG("Setting initial memory amount");
6630
    if (qemuProcessSetupBalloon(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6631
        goto cleanup;
6632

6633 6634 6635
    if (qemuProcessSetupDiskThrottlingBlockdev(driver, vm, asyncJob) < 0)
        goto cleanup;

6636
    /* Since CPUs were not started yet, the balloon could not return the memory
6637 6638
     * 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 */
6639
    if (!incoming && !snapshot &&
6640
        qemuProcessRefreshBalloonState(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6641
        goto cleanup;
6642

J
Jiri Denemark 已提交
6643 6644 6645 6646 6647 6648 6649
    if (flags & VIR_QEMU_PROCESS_START_AUTODESTROY &&
        qemuProcessAutoDestroyAdd(driver, vm, conn) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
6650 6651
    if (ret < 0)
        qemuExtDevicesStop(driver, vm->def);
6652
    qemuDomainSecretDestroy(vm);
J
Jiri Denemark 已提交
6653
    virCommandFree(cmd);
6654
    virObjectUnref(logCtxt);
J
Jiri Denemark 已提交
6655 6656 6657 6658 6659 6660 6661
    virObjectUnref(cfg);
    virObjectUnref(caps);
    VIR_FREE(nicindexes);
    return ret;
}


6662 6663 6664 6665 6666 6667 6668 6669 6670 6671
/**
 * qemuProcessRefreshState:
 * @driver: qemu driver data
 * @vm: domain to refresh
 * @asyncJob: async job type
 *
 * This function gathers calls to refresh qemu state after startup. This
 * function is called after a deferred migration finishes so that we can update
 * state influenced by the migration stream.
 */
6672
int
6673 6674 6675 6676 6677 6678 6679 6680 6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693 6694 6695 6696
qemuProcessRefreshState(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuDomainAsyncJob asyncJob)
{
    VIR_DEBUG("Fetching list of active devices");
    if (qemuDomainUpdateDeviceList(driver, vm, asyncJob) < 0)
        return -1;

    VIR_DEBUG("Updating info of memory devices");
    if (qemuDomainUpdateMemoryDeviceInfo(driver, vm, asyncJob) < 0)
        return -1;

    VIR_DEBUG("Detecting actual memory size for video device");
    if (qemuProcessUpdateVideoRamSize(driver, vm, asyncJob) < 0)
        return -1;

    VIR_DEBUG("Updating disk data");
    if (qemuProcessRefreshDisks(driver, vm, asyncJob) < 0)
        return -1;

    return 0;
}


6697 6698 6699 6700 6701 6702
/**
 * qemuProcessFinishStartup:
 *
 * Finish starting a new domain.
 */
int
6703
qemuProcessFinishStartup(virQEMUDriverPtr driver,
6704 6705 6706 6707 6708 6709 6710 6711 6712 6713
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob,
                         bool startCPUs,
                         virDomainPausedReason pausedReason)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;

    if (startCPUs) {
        VIR_DEBUG("Starting domain CPUs");
6714
        if (qemuProcessStartCPUs(driver, vm,
6715 6716
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 asyncJob) < 0) {
6717
            if (virGetLastErrorCode() == VIR_ERR_OK)
6718 6719 6720 6721 6722 6723 6724 6725 6726
                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");
6727
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
6728 6729 6730 6731 6732 6733 6734 6735 6736 6737 6738 6739 6740 6741 6742
        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
6743 6744 6745 6746
int
qemuProcessStart(virConnectPtr conn,
                 virQEMUDriverPtr driver,
                 virDomainObjPtr vm,
6747
                 virCPUDefPtr updatedCPU,
J
Jiri Denemark 已提交
6748 6749 6750 6751 6752 6753 6754 6755 6756 6757 6758 6759 6760 6761 6762 6763 6764 6765 6766 6767 6768 6769 6770 6771 6772
                 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 |
6773 6774
                      VIR_QEMU_PROCESS_START_AUTODESTROY |
                      VIR_QEMU_PROCESS_START_GEN_VMID, cleanup);
J
Jiri Denemark 已提交
6775

6776 6777 6778
    if (!migrateFrom && !snapshot)
        flags |= VIR_QEMU_PROCESS_START_NEW;

6779 6780
    if (qemuProcessInit(driver, vm, updatedCPU,
                        asyncJob, !!migrateFrom, flags) < 0)
J
Jiri Denemark 已提交
6781 6782 6783
        goto cleanup;

    if (migrateFrom) {
6784
        incoming = qemuProcessIncomingDefNew(priv->qemuCaps, NULL, migrateFrom,
J
Jiri Denemark 已提交
6785 6786 6787 6788 6789
                                             migrateFd, migratePath);
        if (!incoming)
            goto stop;
    }

6790
    if (qemuProcessPrepareDomain(driver, vm, flags) < 0)
6791 6792
        goto stop;

6793
    if (qemuProcessPrepareHost(driver, vm, flags) < 0)
6794 6795
        goto stop;

J
Jiri Denemark 已提交
6796 6797
    if ((rv = qemuProcessLaunch(conn, driver, vm, asyncJob, incoming,
                                snapshot, vmop, flags)) < 0) {
6798
        if (rv == -2)
J
Jiri Denemark 已提交
6799 6800 6801 6802
            relabel = true;
        goto stop;
    }
    relabel = true;
6803

6804 6805
    if (incoming &&
        incoming->deferredURI &&
6806
        qemuMigrationDstRun(driver, vm, incoming->deferredURI, asyncJob) < 0)
J
Jiri Denemark 已提交
6807
        goto stop;
6808

6809
    if (qemuProcessFinishStartup(driver, vm, asyncJob,
6810 6811 6812 6813
                                 !(flags & VIR_QEMU_PROCESS_START_PAUSED),
                                 incoming ?
                                 VIR_DOMAIN_PAUSED_MIGRATION :
                                 VIR_DOMAIN_PAUSED_USER) < 0)
J
Jiri Denemark 已提交
6814
        goto stop;
6815

6816 6817 6818
    if (!incoming) {
        /* Keep watching qemu log for errors during incoming migration, otherwise
         * unset reporting errors from qemu log. */
6819
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
6820

6821 6822 6823 6824 6825 6826
        /* Refresh state of devices from qemu. During migration this needs to
         * happen after the state information is fully transferred. */
        if (qemuProcessRefreshState(driver, vm, asyncJob) < 0)
            goto stop;
    }

6827 6828 6829
    ret = 0;

 cleanup:
6830
    qemuProcessIncomingDefFree(incoming);
6831
    return ret;
6832

J
Jiri Denemark 已提交
6833 6834 6835 6836 6837 6838
 stop:
    stopFlags = 0;
    if (!relabel)
        stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
    if (migrateFrom)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
6839
    if (priv->mon)
6840
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
6841
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, asyncJob, stopFlags);
6842
    goto cleanup;
6843 6844 6845
}


6846
virCommandPtr
6847
qemuProcessCreatePretendCmd(virQEMUDriverPtr driver,
6848 6849
                            virDomainObjPtr vm,
                            const char *migrateURI,
6850
                            bool enableFips,
6851 6852 6853 6854 6855 6856 6857 6858 6859 6860
                            bool standalone,
                            unsigned int flags)
{
    virCommandPtr cmd = NULL;

    virCheckFlagsGoto(VIR_QEMU_PROCESS_START_COLD |
                      VIR_QEMU_PROCESS_START_PAUSED |
                      VIR_QEMU_PROCESS_START_AUTODESTROY, cleanup);

    flags |= VIR_QEMU_PROCESS_START_PRETEND;
6861
    flags |= VIR_QEMU_PROCESS_START_NEW;
6862 6863
    if (standalone)
        flags |= VIR_QEMU_PROCESS_START_STANDALONE;
6864

6865 6866
    if (qemuProcessInit(driver, vm, NULL, QEMU_ASYNC_JOB_NONE,
                        !!migrateURI, flags) < 0)
6867 6868
        goto cleanup;

6869
    if (qemuProcessPrepareDomain(driver, vm, flags) < 0)
6870 6871 6872
        goto cleanup;

    VIR_DEBUG("Building emulator command line");
6873
    cmd = qemuBuildCommandLine(driver,
6874
                               NULL,
6875
                               driver->securityManager,
6876
                               vm,
6877 6878 6879 6880
                               migrateURI,
                               NULL,
                               VIR_NETDEV_VPORT_PROFILE_OP_NO_OP,
                               standalone,
6881
                               enableFips,
6882
                               NULL,
6883
                               NULL);
6884 6885 6886 6887 6888 6889

 cleanup:
    return cmd;
}


6890
int
6891
qemuProcessKill(virDomainObjPtr vm, unsigned int flags)
6892
{
6893
    int ret;
6894

6895
    VIR_DEBUG("vm=%p name=%s pid=%lld flags=0x%x",
6896
              vm, vm->def->name,
6897
              (long long)vm->pid, flags);
6898

6899 6900 6901 6902 6903
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
6904 6905
    }

6906
    if (flags & VIR_QEMU_PROCESS_KILL_NOWAIT) {
6907 6908 6909 6910 6911
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
6912

6913 6914 6915 6916 6917
    /* Request an extra delay of two seconds per current nhostdevs
     * to be safe against stalls by the kernel freeing up the resources */
    ret = virProcessKillPainfullyDelay(vm->pid,
                                       !!(flags & VIR_QEMU_PROCESS_KILL_FORCE),
                                       vm->def->nhostdevs * 2);
6918

6919
    return ret;
6920 6921 6922
}


6923 6924 6925 6926 6927 6928 6929 6930 6931 6932 6933 6934 6935 6936 6937 6938 6939 6940 6941 6942 6943 6944 6945 6946 6947 6948 6949 6950 6951 6952 6953 6954 6955 6956 6957 6958 6959 6960 6961
/**
 * qemuProcessBeginStopJob:
 *
 * Stop all current jobs by killing the domain and start a new one for
 * qemuProcessStop.
 */
int
qemuProcessBeginStopJob(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuDomainJob job,
                        bool forceKill)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    unsigned int killFlags = forceKill ? VIR_QEMU_PROCESS_KILL_FORCE : 0;
    int ret = -1;

    /* We need to prevent monitor EOF callback from doing our work (and
     * sending misleading events) while the vm is unlocked inside
     * BeginJob/ProcessKill API
     */
    priv->beingDestroyed = true;

    if (qemuProcessKill(vm, killFlags) < 0)
        goto cleanup;

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

    if (qemuDomainObjBeginJob(driver, vm, job) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    priv->beingDestroyed = false;
    return ret;
}


6962
void qemuProcessStop(virQEMUDriverPtr driver,
6963
                     virDomainObjPtr vm,
6964
                     virDomainShutoffReason reason,
6965
                     qemuDomainAsyncJob asyncJob,
6966
                     unsigned int flags)
6967 6968 6969 6970 6971 6972
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
6973
    virNetDevVPortProfilePtr vport = NULL;
6974
    size_t i;
6975
    char *timestamp;
6976
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
6977

M
Michal Privoznik 已提交
6978
    VIR_DEBUG("Shutting down vm=%p name=%s id=%d pid=%lld, "
6979
              "reason=%s, asyncJob=%s, flags=0x%x",
6980
              vm, vm->def->name, vm->def->id,
6981
              (long long)vm->pid,
6982 6983 6984
              virDomainShutoffReasonTypeToString(reason),
              qemuDomainAsyncJobTypeToString(asyncJob),
              flags);
6985

6986 6987 6988 6989
    /* This method is routinely used in clean up paths. Disable error
     * reporting so we don't squash a legit error. */
    orig_err = virSaveLastError();

6990 6991 6992 6993 6994 6995 6996 6997 6998 6999 7000 7001 7002 7003 7004
    if (asyncJob != QEMU_ASYNC_JOB_NONE) {
        if (qemuDomainObjBeginNestedJob(driver, vm, asyncJob) < 0)
            goto cleanup;
    } else if (priv->job.asyncJob != QEMU_ASYNC_JOB_NONE &&
               priv->job.asyncOwner == virThreadSelfID() &&
               priv->job.active != QEMU_JOB_ASYNC_NESTED) {
        VIR_WARN("qemuProcessStop called without a nested job (async=%s)",
                 qemuDomainAsyncJobTypeToString(asyncJob));
    }

    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("VM '%s' not active", vm->def->name);
        goto endjob;
    }

7005
    qemuProcessBuildDestroyMemoryPaths(driver, vm, NULL, false);
7006

7007 7008
    vm->def->id = -1;

7009
    if (virAtomicIntDecAndTest(&driver->nactive) && driver->inhibitCallback)
7010 7011
        driver->inhibitCallback(false, driver->inhibitOpaque);

7012 7013
    /* Wake up anything waiting on domain condition */
    virDomainObjBroadcast(vm);
7014

7015
    if ((timestamp = virTimeStringNow()) != NULL) {
7016 7017 7018
        qemuDomainLogAppendMessage(driver, vm, "%s: shutting down, reason=%s\n",
                                   timestamp,
                                   virDomainShutoffReasonTypeToString(reason));
7019
        VIR_FREE(timestamp);
7020 7021
    }

7022 7023 7024
    /* Clear network bandwidth */
    virDomainClearNetBandwidth(vm);

7025 7026
    virDomainConfVMNWFilterTeardown(vm);

7027
    if (cfg->macFilter) {
7028
        def = vm->def;
7029
        for (i = 0; i < def->nnets; i++) {
7030 7031 7032
            virDomainNetDefPtr net = def->nets[i];
            if (net->ifname == NULL)
                continue;
7033 7034 7035
            ignore_value(ebtablesRemoveForwardAllowIn(driver->ebtables,
                                                      net->ifname,
                                                      &net->mac));
7036 7037 7038
        }
    }

7039
    virPortAllocatorRelease(priv->nbdPort);
7040
    priv->nbdPort = 0;
7041

D
Daniel P. Berrange 已提交
7042 7043 7044 7045
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
    }
7046
    priv->agentError = false;
D
Daniel P. Berrange 已提交
7047

7048
    if (priv->mon) {
7049
        qemuMonitorClose(priv->mon);
7050 7051
        priv->mon = NULL;
    }
7052 7053 7054 7055 7056 7057 7058 7059

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

J
John Ferlan 已提交
7060 7061 7062
    /* Remove the master key */
    qemuDomainMasterKeyRemove(priv);

7063
    /* Do this before we delete the tree and remove pidfile. */
7064
    qemuProcessKillManagedPRDaemon(vm);
7065

7066 7067
    virFileDeleteTree(priv->libDir);
    virFileDeleteTree(priv->channelTargetDir);
7068

7069 7070 7071 7072 7073 7074
    ignore_value(virDomainChrDefForeach(vm->def,
                                        false,
                                        qemuProcessCleanupChardevDevice,
                                        NULL));


7075
    /* shut it off for sure */
7076 7077 7078
    ignore_value(qemuProcessKill(vm,
                                 VIR_QEMU_PROCESS_KILL_FORCE|
                                 VIR_QEMU_PROCESS_KILL_NOCHECK));
7079

7080 7081
    qemuDomainCleanupRun(driver, vm);

7082 7083
    qemuExtDevicesStop(driver, vm->def);

7084
    /* Stop autodestroy in case guest is restarted */
7085
    qemuProcessAutoDestroyRemove(driver, vm);
7086

7087 7088
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
7089
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
7090 7091

        /* we can't stop the operation even if the script raised an error */
7092 7093 7094
        ignore_value(virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
                                 VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                                 NULL, xml, NULL));
7095 7096 7097
        VIR_FREE(xml);
    }

7098 7099
    /* Reset Security Labels unless caller don't want us to */
    if (!(flags & VIR_QEMU_PROCESS_STOP_NO_RELABEL))
7100 7101 7102
        qemuSecurityRestoreAllLabel(driver, vm,
                                    !!(flags & VIR_QEMU_PROCESS_STOP_MIGRATED));

7103
    qemuSecurityReleaseLabel(driver->securityManager, vm->def);
7104

7105
    for (i = 0; i < vm->def->ndisks; i++) {
7106
        virDomainDeviceDef dev;
7107
        virDomainDiskDefPtr disk = vm->def->disks[i];
7108 7109 7110 7111

        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        ignore_value(qemuRemoveSharedDevice(driver, &dev, vm->def->name));
7112 7113
    }

7114
    /* Clear out dynamically assigned labels */
7115
    for (i = 0; i < vm->def->nseclabels; i++) {
7116
        if (vm->def->seclabels[i]->type == VIR_DOMAIN_SECLABEL_DYNAMIC)
7117 7118
            VIR_FREE(vm->def->seclabels[i]->label);
        VIR_FREE(vm->def->seclabels[i]->imagelabel);
7119 7120
    }

7121
    qemuHostdevReAttachDomainDevices(driver, vm->def);
7122 7123 7124 7125

    def = vm->def;
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
7126 7127 7128 7129
        vport = virDomainNetGetActualVirtPortProfile(net);

        switch (virDomainNetGetActualType(net)) {
        case VIR_DOMAIN_NET_TYPE_DIRECT:
7130
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
7131
                             net->ifname, &net->mac,
7132 7133
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
7134
                             virDomainNetGetActualVirtPortProfile(net),
7135
                             cfg->stateDir));
7136
            break;
7137 7138 7139 7140 7141 7142
        case VIR_DOMAIN_NET_TYPE_ETHERNET:
            if (net->ifname) {
                ignore_value(virNetDevTapDelete(net->ifname, net->backend.tap));
                VIR_FREE(net->ifname);
            }
            break;
7143 7144 7145 7146
        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))
7147
                ignore_value(virNetDevTapDelete(net->ifname, net->backend.tap));
7148 7149
#endif
            break;
7150 7151 7152 7153 7154 7155 7156 7157 7158 7159 7160
        case VIR_DOMAIN_NET_TYPE_USER:
        case VIR_DOMAIN_NET_TYPE_VHOSTUSER:
        case VIR_DOMAIN_NET_TYPE_SERVER:
        case VIR_DOMAIN_NET_TYPE_CLIENT:
        case VIR_DOMAIN_NET_TYPE_MCAST:
        case VIR_DOMAIN_NET_TYPE_INTERNAL:
        case VIR_DOMAIN_NET_TYPE_HOSTDEV:
        case VIR_DOMAIN_NET_TYPE_UDP:
        case VIR_DOMAIN_NET_TYPE_LAST:
            /* No special cleanup procedure for these types. */
            break;
7161
        }
7162 7163 7164
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
7165 7166 7167 7168 7169 7170 7171 7172 7173
        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
7174

7175 7176
        /* kick the device out of the hostdev list too */
        virDomainNetRemoveHostdev(def, net);
7177
        virDomainNetReleaseActualDevice(vm->def, net);
7178
    }
7179

7180
 retry:
7181
    if ((ret = qemuRemoveCgroup(vm)) < 0) {
7182 7183 7184 7185 7186 7187 7188
        if (ret == -EBUSY && (retries++ < 5)) {
            usleep(200*1000);
            goto retry;
        }
        VIR_WARN("Failed to remove cgroup for %s",
                 vm->def->name);
    }
7189

7190 7191 7192
    /* Remove resctrl allocation after cgroups are cleaned up which makes it
     * kind of safer (although removing the allocation should work even with
     * pids in tasks file */
B
Bing Niu 已提交
7193 7194
    for (i = 0; i < vm->def->nresctrls; i++)
        virResctrlAllocRemove(vm->def->resctrls[i]->alloc);
7195

7196 7197
    qemuProcessRemoveDomainStatus(driver, vm);

7198 7199
    /* Remove VNC and Spice ports from port reservation bitmap, but only if
       they were reserved by the driver (autoport=yes)
7200
    */
7201
    for (i = 0; i < vm->def->ngraphics; ++i) {
7202
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
7203 7204
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (graphics->data.vnc.autoport) {
7205
                virPortAllocatorRelease(graphics->data.vnc.port);
7206
            } else if (graphics->data.vnc.portReserved) {
7207
                virPortAllocatorRelease(graphics->data.vnc.port);
7208 7209
                graphics->data.vnc.portReserved = false;
            }
7210
            if (graphics->data.vnc.websocketGenerated) {
7211
                virPortAllocatorRelease(graphics->data.vnc.websocket);
7212 7213 7214
                graphics->data.vnc.websocketGenerated = false;
                graphics->data.vnc.websocket = -1;
            } else if (graphics->data.vnc.websocket) {
7215
                virPortAllocatorRelease(graphics->data.vnc.websocket);
7216
            }
7217
        }
7218 7219
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (graphics->data.spice.autoport) {
7220 7221
                virPortAllocatorRelease(graphics->data.spice.port);
                virPortAllocatorRelease(graphics->data.spice.tlsPort);
7222 7223
            } else {
                if (graphics->data.spice.portReserved) {
7224
                    virPortAllocatorRelease(graphics->data.spice.port);
7225 7226 7227 7228
                    graphics->data.spice.portReserved = false;
                }

                if (graphics->data.spice.tlsPortReserved) {
7229
                    virPortAllocatorRelease(graphics->data.spice.tlsPort);
7230 7231 7232
                    graphics->data.spice.tlsPortReserved = false;
                }
            }
7233
        }
7234 7235
    }

7236
    vm->taint = 0;
7237
    vm->pid = -1;
J
Jiri Denemark 已提交
7238
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
7239 7240
    for (i = 0; i < vm->def->niothreadids; i++)
        vm->def->iothreadids[i]->thread_id = 0;
7241

7242 7243
    /* clear all private data entries which are no longer needed */
    qemuDomainObjPrivateDataClear(priv);
7244

7245 7246 7247
    /* reset node name allocator */
    qemuDomainStorageIdReset(priv);

7248
    /* The "release" hook cleans up additional resources */
7249
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
7250
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
7251 7252 7253

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
7254 7255
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
7256 7257 7258
        VIR_FREE(xml);
    }

7259
    virDomainObjRemoveTransientDef(vm);
7260

7261 7262 7263 7264 7265
 endjob:
    if (asyncJob != QEMU_ASYNC_JOB_NONE)
        qemuDomainObjEndJob(driver, vm);

 cleanup:
7266 7267 7268 7269
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
7270
    virObjectUnref(cfg);
7271
}
7272 7273


7274 7275 7276 7277 7278 7279 7280 7281 7282 7283 7284 7285 7286 7287 7288 7289 7290 7291 7292 7293 7294 7295
int qemuProcessAttach(virConnectPtr conn ATTRIBUTE_UNUSED,
                      virQEMUDriverPtr driver,
                      virDomainObjPtr vm,
                      pid_t pid,
                      const char *pidfile,
                      virDomainChrSourceDefPtr monConfig,
                      bool monJSON)
{
    size_t i;
    qemuDomainLogContextPtr logCtxt = NULL;
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool running = true;
    virDomainPausedReason reason;
    virSecurityLabelPtr seclabel = NULL;
    virSecurityLabelDefPtr seclabeldef = NULL;
    bool seclabelgen = false;
    virSecurityManagerPtr* sec_managers = NULL;
    const char *model;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    bool active = false;
7296
    virDomainVirtType virtType;
7297 7298 7299 7300 7301 7302 7303 7304 7305 7306 7307 7308 7309 7310 7311 7312 7313 7314 7315 7316 7317 7318 7319 7320 7321 7322 7323 7324 7325 7326 7327 7328 7329 7330 7331 7332 7333 7334 7335 7336 7337 7338 7339 7340 7341 7342 7343 7344 7345 7346 7347 7348 7349 7350 7351 7352 7353 7354 7355 7356 7357 7358 7359 7360 7361 7362 7363 7364 7365 7366 7367 7368 7369 7370 7371 7372 7373 7374 7375 7376 7377 7378 7379 7380 7381 7382 7383 7384 7385 7386

    VIR_DEBUG("Beginning VM attach process");

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

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

    /* 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) < 0)
        goto error;

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

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

    if (virFileMakePath(cfg->logDir) < 0) {
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
                             cfg->logDir);
        goto error;
    }

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

    vm->pid = pid;

    VIR_DEBUG("Detect security driver config");
    sec_managers = qemuSecurityGetNested(driver->securityManager);
    if (sec_managers == NULL)
        goto error;

    for (i = 0; sec_managers[i]; i++) {
        seclabelgen = false;
        model = qemuSecurityGetModel(sec_managers[i]);
        seclabeldef = virDomainDefGetSecurityLabelDef(vm->def, model);
        if (seclabeldef == NULL) {
            if (!(seclabeldef = virSecurityLabelDefNew(model)))
                goto error;
            seclabelgen = true;
        }
        seclabeldef->type = VIR_DOMAIN_SECLABEL_STATIC;
        if (VIR_ALLOC(seclabel) < 0)
            goto error;
        if (qemuSecurityGetProcessLabel(sec_managers[i], vm->def,
                                        vm->pid, seclabel) < 0)
            goto error;

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

        if (VIR_STRDUP(seclabeldef->label, seclabel->label) < 0)
            goto error;
        VIR_FREE(seclabel);

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

    if (qemuSecurityCheckAllLabel(driver->securityManager, vm->def) < 0)
        goto error;
    if (qemuSecurityGenLabel(driver->securityManager, vm->def) < 0)
        goto error;

    if (qemuDomainPerfRestart(vm) < 0)
        goto error;

    VIR_DEBUG("Creating domain log file");
    if (!(logCtxt = qemuDomainLogContextNew(driver, vm,
                                            QEMU_DOMAIN_LOG_CONTEXT_MODE_ATTACH)))
        goto error;

    VIR_DEBUG("Determining emulator version");
    virObjectUnref(priv->qemuCaps);
7387
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
7388 7389 7390 7391 7392 7393 7394 7395 7396 7397 7398
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
        goto error;

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

    priv->gotShutdown = false;

7399 7400 7401 7402
    /* Attaching to running QEMU so we need to detect whether it was started
     * with -no-reboot. */
    qemuProcessPrepareAllowReboot(vm);

7403 7404 7405 7406 7407 7408 7409 7410 7411 7412 7413 7414 7415 7416 7417 7418 7419 7420 7421 7422 7423 7424 7425 7426 7427 7428 7429 7430 7431 7432 7433 7434 7435 7436 7437 7438 7439 7440 7441 7442 7443 7444 7445 7446 7447
    /*
     * 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
     * we also need to populate the PCI address set cache for later
     * use in hotplug
     */
    VIR_DEBUG("Assigning domain PCI addresses");
    if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps,
                                   driver, vm, false)) < 0) {
        goto error;
    }

    if ((timestamp = virTimeStringNow()) == NULL)
        goto error;

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

    qemuDomainObjTaint(driver, vm, VIR_DOMAIN_TAINT_EXTERNAL_LAUNCH, logCtxt);

    VIR_DEBUG("Waiting for monitor to show up");
    if (qemuProcessWaitForMonitor(driver, vm, QEMU_ASYNC_JOB_NONE, NULL) < 0)
        goto error;

    if (qemuConnectAgent(driver, vm) < 0)
        goto error;

    VIR_DEBUG("Detecting VCPU PIDs");
    if (qemuDomainRefreshVcpuInfo(driver, vm, QEMU_ASYNC_JOB_NONE, false) < 0)
        goto error;

    if (qemuDomainValidateVcpuInfo(vm) < 0)
        goto error;

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

    VIR_DEBUG("Getting initial memory amount");
    qemuDomainObjEnterMonitor(driver, vm);
    if (qemuMonitorGetBalloonInfo(priv->mon, &vm->def->mem.cur_balloon) < 0)
        goto exit_monitor;
    if (qemuMonitorGetStatus(priv->mon, &running, &reason) < 0)
        goto exit_monitor;
7448
    if (qemuMonitorGetVirtType(priv->mon, &virtType) < 0)
7449
        goto exit_monitor;
7450
    vm->def->virtType = virtType;
7451 7452 7453 7454 7455 7456 7457 7458 7459 7460 7461 7462 7463 7464 7465 7466 7467 7468 7469 7470 7471 7472 7473 7474 7475 7476 7477 7478 7479 7480 7481 7482 7483 7484 7485 7486 7487 7488 7489 7490 7491 7492 7493 7494 7495 7496 7497 7498 7499 7500 7501 7502 7503 7504 7505 7506 7507 7508 7509 7510 7511 7512 7513 7514 7515 7516 7517 7518 7519 7520 7521 7522
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto error;

    if (running) {
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                             VIR_DOMAIN_RUNNING_UNPAUSED);
        if (vm->def->memballoon &&
            vm->def->memballoon->model == VIR_DOMAIN_MEMBALLOON_MODEL_VIRTIO &&
            vm->def->memballoon->period) {
            qemuDomainObjEnterMonitor(driver, vm);
            qemuMonitorSetMemoryStatsPeriod(priv->mon, vm->def->memballoon,
                                            vm->def->memballoon->period);
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                goto error;
        }
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
    }

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

    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
        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)
            goto error;
    }

    virObjectUnref(logCtxt);
    VIR_FREE(seclabel);
    VIR_FREE(sec_managers);
    virObjectUnref(cfg);
    virObjectUnref(caps);

    return 0;

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
 error:
    /* 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);

    qemuMonitorClose(priv->mon);
    priv->mon = NULL;
    virObjectUnref(logCtxt);
    VIR_FREE(seclabel);
    VIR_FREE(sec_managers);
    if (seclabelgen)
        virSecurityLabelDefFree(seclabeldef);
    virDomainChrSourceDefFree(monConfig);
    virObjectUnref(cfg);
    virObjectUnref(caps);
    return -1;
}


7523
static void
7524 7525 7526 7527 7528 7529 7530 7531 7532 7533 7534 7535 7536 7537 7538 7539 7540 7541 7542 7543 7544 7545 7546
qemuProcessAutoDestroy(virDomainObjPtr dom,
                       virConnectPtr conn,
                       void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    qemuDomainObjPrivatePtr priv = dom->privateData;
    virObjectEventPtr event = NULL;
    unsigned int stopFlags = 0;

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

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

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

    VIR_DEBUG("Killing domain");

    if (qemuProcessBeginStopJob(driver, dom, QEMU_JOB_DESTROY, true) < 0)
7547
        return;
7548 7549 7550 7551 7552 7553 7554 7555 7556 7557 7558

    qemuProcessStop(driver, dom, VIR_DOMAIN_SHUTOFF_DESTROYED,
                    QEMU_ASYNC_JOB_NONE, stopFlags);

    virDomainAuditStop(dom, "destroyed");
    event = virDomainEventLifecycleNewFromObj(dom,
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_DESTROYED);

    qemuDomainRemoveInactive(driver, dom);

7559 7560
    qemuDomainObjEndJob(driver, dom);

7561
    virObjectEventStateQueue(driver->domainEventState, event);
7562 7563 7564 7565 7566 7567 7568 7569 7570 7571 7572 7573 7574 7575 7576 7577 7578 7579 7580 7581 7582 7583 7584 7585 7586 7587 7588 7589 7590 7591 7592 7593 7594 7595 7596 7597 7598
}

int qemuProcessAutoDestroyAdd(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              virConnectPtr conn)
{
    VIR_DEBUG("vm=%s, conn=%p", vm->def->name, conn);
    return virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                qemuProcessAutoDestroy);
}

int qemuProcessAutoDestroyRemove(virQEMUDriverPtr driver,
                                 virDomainObjPtr vm)
{
    int ret;
    VIR_DEBUG("vm=%s", vm->def->name);
    ret = virCloseCallbacksUnset(driver->closeCallbacks, vm,
                                 qemuProcessAutoDestroy);
    return ret;
}

bool qemuProcessAutoDestroyActive(virQEMUDriverPtr driver,
                                  virDomainObjPtr vm)
{
    virCloseCallback cb;
    VIR_DEBUG("vm=%s", vm->def->name);
    cb = virCloseCallbacksGet(driver->closeCallbacks, vm, NULL);
    return cb == qemuProcessAutoDestroy;
}


int
qemuProcessRefreshDisks(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuDomainAsyncJob asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
7599
    bool blockdev = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV);
7600 7601 7602 7603 7604 7605 7606 7607 7608 7609 7610 7611 7612 7613 7614 7615 7616
    virHashTablePtr table = NULL;
    int ret = -1;
    size_t i;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) == 0) {
        table = qemuMonitorGetBlockInfo(priv->mon);
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            goto cleanup;
    }

    if (!table)
        goto cleanup;

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskpriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
        struct qemuDomainDiskInfo *info;
7617 7618 7619 7620
        const char *entryname = disk->info.alias;

        if (blockdev)
            entryname = diskpriv->qomName;
7621

7622
        if (!(info = virHashLookup(table, entryname)))
7623 7624 7625 7626 7627 7628 7629 7630 7631 7632 7633 7634 7635 7636 7637 7638 7639 7640 7641 7642 7643 7644 7645 7646 7647 7648 7649
            continue;

        if (info->removable) {
            if (info->empty)
                virDomainDiskEmptySource(disk);

            if (info->tray) {
                if (info->tray_open)
                    disk->tray_status = VIR_DOMAIN_DISK_TRAY_OPEN;
                else
                    disk->tray_status = VIR_DOMAIN_DISK_TRAY_CLOSED;
            }
        }

        /* fill in additional data */
        diskpriv->removable = info->removable;
        diskpriv->tray = info->tray;
    }

    ret = 0;

 cleanup:
    virHashFree(table);
    return ret;
}


7650 7651 7652 7653 7654
static int
qemuProcessRefreshCPU(virQEMUDriverPtr driver,
                      virDomainObjPtr vm)
{
    virCapsPtr caps = virQEMUDriverGetCapabilities(driver, false);
7655
    qemuDomainObjPrivatePtr priv = vm->privateData;
7656
    virCPUDefPtr host = NULL;
7657
    virCPUDefPtr cpu = NULL;
7658 7659 7660 7661 7662 7663 7664 7665 7666 7667 7668 7669 7670 7671 7672 7673 7674 7675 7676 7677 7678
    int ret = -1;

    if (!caps)
        return -1;

    if (!virQEMUCapsGuestIsNative(caps->host.arch, vm->def->os.arch) ||
        !caps->host.cpu ||
        !vm->def->cpu) {
        ret = 0;
        goto cleanup;
    }

    /* If the domain with a host-model CPU was started by an old libvirt
     * (< 2.3) which didn't replace the CPU with a custom one, let's do it now
     * since the rest of our code does not really expect a host-model CPU in a
     * running domain.
     */
    if (vm->def->cpu->mode == VIR_CPU_MODE_HOST_MODEL) {
        if (!(host = virCPUCopyMigratable(caps->host.cpu->arch, caps->host.cpu)))
            goto cleanup;

7679 7680 7681 7682 7683 7684 7685
        if (!(cpu = virCPUDefCopyWithoutModel(host)) ||
            virCPUDefCopyModelFilter(cpu, host, false,
                                     virQEMUCapsCPUFilterFeatures,
                                     &caps->host.cpu->arch) < 0)
            goto cleanup;

        if (virCPUUpdate(vm->def->os.arch, vm->def->cpu, cpu) < 0)
7686 7687 7688 7689
            goto cleanup;

        if (qemuProcessUpdateCPU(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
            goto cleanup;
7690 7691 7692 7693 7694 7695 7696 7697
    } else if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QUERY_CPU_MODEL_EXPANSION)) {
        /* We only try to fix CPUs when the libvirt/QEMU combo used to start
         * the domain did not know about query-cpu-model-expansion in which
         * case the host-model is known to not contain features which QEMU
         * doesn't know about.
         */
        if (qemuDomainFixupCPUs(vm, &priv->origCPU) < 0)
            goto cleanup;
7698 7699 7700 7701 7702
    }

    ret = 0;

 cleanup:
7703
    virCPUDefFree(cpu);
7704 7705 7706 7707 7708 7709
    virCPUDefFree(host);
    virObjectUnref(caps);
    return ret;
}


7710 7711 7712 7713 7714 7715 7716 7717 7718 7719 7720 7721 7722 7723 7724 7725 7726 7727 7728 7729 7730 7731 7732 7733
struct qemuProcessReconnectData {
    virQEMUDriverPtr driver;
    virDomainObjPtr obj;
};
/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
 *
 * This function also inherits a locked and ref'd domain object.
 *
 * 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.
 */
static void
qemuProcessReconnect(void *opaque)
7734
{
7735 7736 7737 7738
    struct qemuProcessReconnectData *data = opaque;
    virQEMUDriverPtr driver = data->driver;
    virDomainObjPtr obj = data->obj;
    qemuDomainObjPrivatePtr priv;
7739
    qemuDomainJobObj oldjob;
7740 7741 7742
    int state;
    int reason;
    virQEMUDriverConfigPtr cfg;
7743
    size_t i;
7744 7745
    unsigned int stopFlags = 0;
    bool jobStarted = false;
7746
    virCapsPtr caps = NULL;
7747
    bool retry = true;
7748

7749
    VIR_FREE(data);
7750

7751 7752 7753 7754 7755 7756
    qemuDomainObjRestoreJob(obj, &oldjob);
    if (oldjob.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;

    cfg = virQEMUDriverGetConfig(driver);
    priv = obj->privateData;
7757

7758
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
7759
        goto error;
7760

7761
    if (qemuDomainObjBeginJob(driver, obj, QEMU_JOB_MODIFY) < 0)
7762
        goto error;
7763
    jobStarted = true;
7764

7765 7766 7767 7768
    /* 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 error;
7769

7770 7771 7772
    /* Restore the masterKey */
    if (qemuDomainMasterKeyReadFile(priv) < 0)
        goto error;
7773

7774 7775 7776 7777
    /* If we are connecting to a guest started by old libvirt there is no
     * allowReboot in status XML and we need to initialize it. */
    qemuProcessPrepareAllowReboot(obj);

7778 7779 7780 7781 7782 7783
    if (priv->qemuCaps &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_CHARDEV_FD_PASS))
        retry = false;

    VIR_DEBUG("Reconnect monitor to def=%p name='%s' retry=%d",
              obj, obj->def->name, retry);
7784 7785

    /* XXX check PID liveliness & EXE path */
7786
    if (qemuConnectMonitor(driver, obj, QEMU_ASYNC_JOB_NONE, retry, NULL) < 0)
7787
        goto error;
7788

7789
    if (qemuHostdevUpdateActiveDomainDevices(driver, obj->def) < 0)
7790
        goto error;
7791

7792 7793 7794 7795
    priv->machineName = qemuDomainGetMachineName(obj);
    if (!priv->machineName)
        goto error;

7796
    if (qemuConnectCgroup(obj) < 0)
7797
        goto error;
7798

7799
    if (qemuDomainPerfRestart(obj) < 0)
7800
        goto error;
7801

7802 7803 7804 7805
    /* XXX: Need to change as long as lock is introduced for
     * qemu_driver->sharedDevices.
     */
    for (i = 0; i < obj->def->ndisks; i++) {
7806
        virDomainDiskDefPtr disk = obj->def->disks[i];
7807
        virDomainDeviceDef dev;
7808

7809
        if (virDomainDiskTranslateSourcePool(disk) < 0)
7810
            goto error;
7811

7812
        /* backing chains need to be refreshed only if they could change */
7813 7814
        if (priv->reconnectBlockjobs != VIR_TRISTATE_BOOL_NO &&
            !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV)) {
7815 7816 7817
            /* This should be the only place that calls
             * qemuDomainDetermineDiskChain with @report_broken == false
             * to guarantee best-effort domain reconnect */
7818 7819
            virStorageSourceBackingStoreClear(disk->src);
            if (qemuDomainDetermineDiskChain(driver, obj, disk, false) < 0)
7820 7821 7822 7823
                goto error;
        } else {
            VIR_DEBUG("skipping backing chain detection for '%s'", disk->dst);
        }
7824

7825
        dev.type = VIR_DOMAIN_DEVICE_DISK;
7826
        dev.data.disk = disk;
7827
        if (qemuAddSharedDevice(driver, &dev, obj->def->name) < 0)
7828 7829 7830 7831
            goto error;
    }

    for (i = 0; i < obj->def->ngraphics; i++) {
7832
        if (qemuProcessGraphicsReservePorts(obj->def->graphics[i], true) < 0)
7833
            goto error;
7834
    }
7835

7836
    if (qemuProcessUpdateState(driver, obj) < 0)
7837 7838
        goto error;

7839 7840 7841 7842 7843 7844
    state = virDomainObjGetState(obj, &reason);
    if (state == VIR_DOMAIN_SHUTOFF ||
        (state == VIR_DOMAIN_PAUSED &&
         reason == VIR_DOMAIN_PAUSED_STARTING_UP)) {
        VIR_DEBUG("Domain '%s' wasn't fully started yet, killing it",
                  obj->def->name);
7845
        goto error;
7846
    }
7847

7848 7849 7850 7851
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
    if (!priv->qemuCaps &&
7852
        !(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
7853 7854
                                                      obj->def->emulator,
                                                      obj->def->os.machine)))
7855
        goto error;
7856

7857 7858 7859 7860 7861 7862 7863 7864 7865 7866 7867 7868
    /* In case the domain shutdown while we were not running,
     * we need to finish the shutdown process. And we need to do it after
     * we have virQEMUCaps filled in.
     */
    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);
        goto cleanup;
    }
7869

7870
    if (qemuProcessBuildDestroyMemoryPaths(driver, obj, NULL, true) < 0)
7871
        goto error;
7872

7873 7874
    if ((qemuDomainAssignAddresses(obj->def, priv->qemuCaps,
                                   driver, obj, false)) < 0) {
7875
        goto error;
7876
    }
7877

7878 7879 7880 7881 7882 7883
    /* if domain requests security driver we haven't loaded, report error, but
     * do not kill the domain
     */
    ignore_value(qemuSecurityCheckAllLabel(driver->securityManager,
                                           obj->def));

7884 7885
    if (qemuProcessRefreshCPU(driver, obj) < 0)
        goto error;
7886

7887
    if (qemuDomainRefreshVcpuInfo(driver, obj, QEMU_ASYNC_JOB_NONE, true) < 0)
7888
        goto error;
7889

7890
    qemuDomainVcpuPersistOrder(obj->def);
7891

7892 7893
    if (qemuSecurityReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
        goto error;
7894

7895 7896
    qemuProcessNotifyNets(obj->def);

7897
    if (qemuProcessFiltersInstantiate(obj->def, true))
7898
        goto error;
7899

7900
    if (qemuProcessRefreshDisks(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
7901
        goto error;
D
Daniel P. Berrange 已提交
7902

7903 7904
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV) &&
        qemuBlockNodeNamesDetect(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
7905 7906
        goto error;

7907
    if (qemuRefreshVirtioChannelState(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
7908 7909
        goto error;

7910 7911 7912 7913
    /* If querying of guest's RTC failed, report error, but do not kill the domain. */
    qemuRefreshRTC(driver, obj);

    if (qemuProcessRefreshBalloonState(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
7914
        goto error;
7915

7916
    if (qemuProcessRecoverJob(driver, obj, &oldjob, &stopFlags) < 0)
7917
        goto error;
7918

7919 7920
    if (qemuProcessUpdateDevices(driver, obj) < 0)
        goto error;
7921

7922 7923 7924
    if (qemuRefreshPRManagerState(driver, obj) < 0)
        goto error;

7925 7926 7927 7928 7929
    qemuProcessReconnectCheckMemAliasOrderMismatch(obj);

    if (qemuConnectAgent(driver, obj) < 0)
        goto error;

B
Bing Niu 已提交
7930 7931
    for (i = 0; i < obj->def->nresctrls; i++) {
        if (virResctrlAllocDeterminePath(obj->def->resctrls[i]->alloc,
7932 7933 7934 7935
                                         priv->machineName) < 0)
            goto error;
    }

7936 7937
    /* update domain state XML with possibly updated state in virDomainObj */
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, obj, driver->caps) < 0)
7938
        goto error;
7939

7940 7941
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
7942
        char *xml = qemuDomainDefFormatXML(driver, obj->def, 0);
7943 7944
        int hookret;

7945 7946
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, obj->def->name,
                              VIR_HOOK_QEMU_OP_RECONNECT, VIR_HOOK_SUBOP_BEGIN,
7947 7948 7949 7950 7951 7952 7953
                              NULL, xml, NULL);
        VIR_FREE(xml);

        /*
         * If the script raised an error abort the launch
         */
        if (hookret < 0)
7954
            goto error;
7955 7956
    }

7957 7958
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
        driver->inhibitCallback(true, driver->inhibitOpaque);
7959

7960
 cleanup:
7961 7962 7963
    if (jobStarted) {
        if (!virDomainObjIsActive(obj))
            qemuDomainRemoveInactive(driver, obj);
7964
        qemuDomainObjEndJob(driver, obj);
7965 7966 7967 7968
    } else {
        if (!virDomainObjIsActive(obj))
            qemuDomainRemoveInactiveJob(driver, obj);
    }
7969
    virDomainObjEndAPI(&obj);
7970
    virObjectUnref(cfg);
7971
    virObjectUnref(caps);
7972 7973
    virNWFilterUnlockFilterUpdates();
    return;
7974

7975 7976 7977 7978 7979
 error:
    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
7980 7981 7982 7983
         * If we couldn't get the monitor since QEMU supports
         * no-shutdown, we can safely say that the domain
         * crashed ... */
        state = VIR_DOMAIN_SHUTOFF_CRASHED;
7984 7985 7986 7987 7988
        /* If BeginJob failed, we jumped here without a job, let's hope another
         * thread didn't have a chance to start playing with the domain yet
         * (it's all we can do anyway).
         */
        qemuProcessStop(driver, obj, state, QEMU_ASYNC_JOB_NONE, stopFlags);
7989
    }
7990
    goto cleanup;
7991
}
7992

7993 7994 7995
static int
qemuProcessReconnectHelper(virDomainObjPtr obj,
                           void *opaque)
7996
{
7997 7998 7999
    virThread thread;
    struct qemuProcessReconnectData *src = opaque;
    struct qemuProcessReconnectData *data;
8000

8001 8002 8003
    /* If the VM was inactive, we don't need to reconnect */
    if (!obj->pid)
        return 0;
8004

8005 8006
    if (VIR_ALLOC(data) < 0)
        return -1;
8007

8008 8009
    memcpy(data, src, sizeof(*data));
    data->obj = obj;
8010

8011 8012
    virNWFilterReadLockFilterUpdates();

8013 8014 8015 8016
    /* this lock and reference will be eventually transferred to the thread
     * that handles the reconnect */
    virObjectLock(obj);
    virObjectRef(obj);
8017

8018 8019 8020 8021 8022 8023 8024 8025 8026 8027 8028
    if (virThreadCreate(&thread, false, qemuProcessReconnect, data) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not create thread. QEMU initialization "
                         "might be incomplete"));
        /* We can't spawn a thread and thus connect to monitor. Kill qemu.
         * It's safe to call qemuProcessStop without a job here since there
         * is no thread that could be doing anything else with the same domain
         * object.
         */
        qemuProcessStop(src->driver, obj, VIR_DOMAIN_SHUTOFF_FAILED,
                        QEMU_ASYNC_JOB_NONE, 0);
8029
        qemuDomainRemoveInactiveJob(src->driver, obj);
8030

8031
        virDomainObjEndAPI(&obj);
8032
        virNWFilterUnlockFilterUpdates();
8033 8034
        VIR_FREE(data);
        return -1;
8035 8036
    }

8037 8038
    return 0;
}
8039

8040 8041 8042 8043 8044 8045 8046
/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
8047
qemuProcessReconnectAll(virQEMUDriverPtr driver)
8048
{
8049
    struct qemuProcessReconnectData data = {.driver = driver};
8050
    virDomainObjListForEach(driver->domains, qemuProcessReconnectHelper, &data);
8051
}