qemu_process.c 246.4 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
#define LIBVIRT_QEMU_PROCESSPRIV_H_ALLOW
39
#include "qemu_processpriv.h"
40
#include "qemu_alias.h"
41
#include "qemu_block.h"
42
#include "qemu_domain.h"
43
#include "qemu_domain_address.h"
44 45 46 47 48 49
#include "qemu_cgroup.h"
#include "qemu_capabilities.h"
#include "qemu_monitor.h"
#include "qemu_command.h"
#include "qemu_hostdev.h"
#include "qemu_hotplug.h"
50
#include "qemu_migration.h"
51
#include "qemu_migration_params.h"
52
#include "qemu_interface.h"
53
#include "qemu_security.h"
54
#include "qemu_extdevice.h"
55

56
#include "cpu/cpu.h"
57
#include "datatypes.h"
58
#include "virlog.h"
59
#include "virerror.h"
60
#include "viralloc.h"
61
#include "virhook.h"
E
Eric Blake 已提交
62
#include "virfile.h"
63
#include "virpidfile.h"
64
#include "virhostcpu.h"
65
#include "domain_audit.h"
66
#include "domain_nwfilter.h"
67
#include "locking/domain_lock.h"
68
#include "viruuid.h"
69
#include "virprocess.h"
70
#include "virtime.h"
A
Ansis Atteka 已提交
71
#include "virnetdevtap.h"
72
#include "virnetdevopenvswitch.h"
73
#include "virnetdevmidonet.h"
74
#include "virbitmap.h"
75
#include "viratomic.h"
76
#include "virnuma.h"
77
#include "virstring.h"
78
#include "virhostdev.h"
J
John Ferlan 已提交
79
#include "secret_util.h"
80
#include "configmake.h"
81
#include "nwfilter_conf.h"
82
#include "netdev_bandwidth_conf.h"
83
#include "virresctrl.h"
84
#include "virvsock.h"
85
#include "viridentity.h"
86 87 88

#define VIR_FROM_THIS VIR_FROM_QEMU

89 90
VIR_LOG_INIT("qemu.qemu_process");

91
/**
92
 * qemuProcessRemoveDomainStatus
93 94 95 96 97 98
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
99
qemuProcessRemoveDomainStatus(virQEMUDriverPtr driver,
100 101 102 103
                              virDomainObjPtr vm)
{
    char ebuf[1024];
    char *file = NULL;
104
    qemuDomainObjPrivatePtr priv = vm->privateData;
105 106
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;
107

108
    if (virAsprintf(&file, "%s/%s.xml", cfg->stateDir, vm->def->name) < 0)
109
        goto cleanup;
110 111 112 113 114 115

    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);

116 117 118
    if (priv->pidfile &&
        unlink(priv->pidfile) < 0 &&
        errno != ENOENT)
119 120 121
        VIR_WARN("Failed to remove PID file for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));

122
    ret = 0;
123
 cleanup:
124 125
    virObjectUnref(cfg);
    return ret;
126 127 128
}


D
Daniel P. Berrange 已提交
129 130 131 132 133 134 135
/*
 * 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
136
qemuProcessHandleAgentEOF(qemuAgentPtr agent,
D
Daniel P. Berrange 已提交
137 138 139 140 141 142
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

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

143
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
144 145

    priv = vm->privateData;
146 147 148 149 150 151 152 153 154 155 156

    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;
    }

157
    qemuAgentClose(agent);
158
    priv->agent = NULL;
159
    priv->agentError = false;
D
Daniel P. Berrange 已提交
160

161
    virObjectUnlock(vm);
162 163
    return;

164
 unlock:
165 166
    virObjectUnlock(vm);
    return;
D
Daniel P. Berrange 已提交
167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
}


/*
 * 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);

184
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
185 186 187 188 189

    priv = vm->privateData;

    priv->agentError = true;

190
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
191 192 193 194 195
}

static void qemuProcessHandleAgentDestroy(qemuAgentPtr agent,
                                          virDomainObjPtr vm)
{
196 197
    VIR_DEBUG("Received destroy agent=%p vm=%p", agent, vm);

198
    virObjectUnref(vm);
D
Daniel P. Berrange 已提交
199 200 201 202 203 204 205 206 207 208
}


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


209
int
210
qemuConnectAgent(virQEMUDriverPtr driver, virDomainObjPtr vm)
D
Daniel P. Berrange 已提交
211 212 213
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuAgentPtr agent = NULL;
214
    virDomainChrDefPtr config = qemuFindAgentConfig(vm->def);
D
Daniel P. Berrange 已提交
215 216 217 218

    if (!config)
        return 0;

219 220 221 222 223 224 225 226 227
    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;
    }

228
    if (qemuSecuritySetDaemonSocketLabel(driver->securityManager, vm->def) < 0) {
D
Daniel P. Berrange 已提交
229 230 231 232 233 234 235
        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 */
236
    virObjectRef(vm);
D
Daniel P. Berrange 已提交
237

238
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
239 240

    agent = qemuAgentOpen(vm,
241
                          config->source,
D
Daniel P. Berrange 已提交
242 243
                          &agentCallbacks);

244
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
245

246 247 248 249 250 251 252
    if (agent == NULL)
        virObjectUnref(vm);

    if (!virDomainObjIsActive(vm)) {
        qemuAgentClose(agent);
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest crashed while connecting to the guest agent"));
253
        return -1;
254 255
    }

256
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0) {
D
Daniel P. Berrange 已提交
257 258
        VIR_ERROR(_("Failed to clear security context for agent for %s"),
                  vm->def->name);
259
        qemuAgentClose(agent);
D
Daniel P. Berrange 已提交
260 261 262 263
        goto cleanup;
    }

    priv->agent = agent;
264
    if (!priv->agent)
D
Daniel P. Berrange 已提交
265 266
        VIR_INFO("Failed to connect agent for %s", vm->def->name);

267
 cleanup:
268 269 270 271 272 273 274
    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
275 276 277
}


278
/*
279
 * This is a callback registered with a qemuMonitorPtr instance,
280 281 282 283 284
 * 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
285
qemuProcessHandleMonitorEOF(qemuMonitorPtr mon,
286 287
                            virDomainObjPtr vm,
                            void *opaque)
288
{
289
    virQEMUDriverPtr driver = opaque;
290
    qemuDomainObjPrivatePtr priv;
291
    struct qemuProcessEvent *processEvent;
292

293
    virObjectLock(vm);
294

295
    VIR_DEBUG("Received EOF on %p '%s'", vm, vm->def->name);
296

297
    priv = vm->privateData;
298 299
    if (priv->beingDestroyed) {
        VIR_DEBUG("Domain is being destroyed, EOF is expected");
300
        goto cleanup;
301 302
    }

303
    if (VIR_ALLOC(processEvent) < 0)
304
        goto cleanup;
305

306
    processEvent->eventType = QEMU_PROCESS_EVENT_MONITOR_EOF;
307
    processEvent->vm = virObjectRef(vm);
308

309 310
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
311
        qemuProcessEventFree(processEvent);
312
        goto cleanup;
313
    }
314

315 316 317 318
    /* We don't want this EOF handler to be called over and over while the
     * thread is waiting for a job.
     */
    qemuMonitorUnregister(mon);
319

320 321 322 323
    /* We don't want any cleanup from EOF handler (or any other
     * thread) to enter qemu namespace. */
    qemuDomainDestroyNamespace(driver, vm);

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


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

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

345
    virObjectLock(vm);
346

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

351
    virObjectUnlock(vm);
352 353 354
}


355 356 357 358 359 360 361 362 363 364 365 366 367
/**
 * 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.
 */
368
virDomainDiskDefPtr
369 370 371
qemuProcessFindDomainDiskByAliasOrQOM(virDomainObjPtr vm,
                                      const char *alias,
                                      const char *qomid)
372
{
373
    size_t i;
374

375 376 377 378 379
    if (alias && *alias == '\0')
        alias = NULL;

    if (alias)
        alias = qemuAliasDiskDriveSkipPrefix(alias);
380 381

    for (i = 0; i < vm->def->ndisks; i++) {
382 383
        virDomainDiskDefPtr disk = vm->def->disks[i];
        qemuDomainDiskPrivatePtr diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
384

385 386
        if ((disk->info.alias && STREQ_NULLABLE(disk->info.alias, alias)) ||
            (diskPriv->qomName && STREQ_NULLABLE(diskPriv->qomName, qomid)))
387 388 389
            return disk;
    }

390
    virReportError(VIR_ERR_INTERNAL_ERROR,
391 392
                   _("no disk found with alias '%s' or id '%s'"),
                   NULLSTR(alias), NULLSTR(qomid));
393 394 395
    return NULL;
}

396

397 398
static int
qemuProcessHandleReset(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
399 400
                       virDomainObjPtr vm,
                       void *opaque)
401
{
402
    virQEMUDriverPtr driver = opaque;
403
    virObjectEventPtr event;
404
    qemuDomainObjPrivatePtr priv;
405
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
M
Michal Privoznik 已提交
406
    int ret = -1;
407

408
    virObjectLock(vm);
409

410
    event = virDomainEventRebootNewFromObj(vm);
411 412 413
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);
414

415
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
416
        VIR_WARN("Failed to save status on vm %s", vm->def->name);
417

418 419
    if (vm->def->onReboot == VIR_DOMAIN_LIFECYCLE_ACTION_DESTROY ||
        vm->def->onReboot == VIR_DOMAIN_LIFECYCLE_ACTION_PRESERVE) {
420

M
Michal Privoznik 已提交
421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436
        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);
    }
437

M
Michal Privoznik 已提交
438 439 440
    ret = 0;
 cleanup:
    virObjectUnlock(vm);
441
    virObjectEventStateQueue(driver->domainEventState, event);
442
    virObjectUnref(cfg);
M
Michal Privoznik 已提交
443
    return ret;
444 445 446
}


447 448 449 450 451 452 453 454 455 456 457 458 459
/*
 * 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;
460
    virQEMUDriverPtr driver = priv->driver;
461
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
462
    virDomainRunningReason reason = VIR_DOMAIN_RUNNING_BOOTED;
463
    int ret = -1, rc;
464

465
    VIR_DEBUG("vm=%p", vm);
466
    virObjectLock(vm);
467
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
468 469 470
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
471 472
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
473 474 475
        goto endjob;
    }

476
    qemuDomainObjEnterMonitor(driver, vm);
477 478 479
    rc = qemuMonitorSystemReset(priv->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0)
480 481
        goto endjob;

482
    if (rc < 0)
483 484
        goto endjob;

485 486 487
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_CRASHED)
        reason = VIR_DOMAIN_RUNNING_CRASHED;

488
    if (qemuProcessStartCPUs(driver, vm,
489
                             reason,
490
                             QEMU_ASYNC_JOB_NONE) < 0) {
491
        if (virGetLastErrorCode() == VIR_ERR_OK)
492 493
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
494 495 496
        goto endjob;
    }

497
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
498 499 500 501
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

502 503
    ret = 0;

504
 endjob:
505
    qemuDomainObjEndJob(driver, vm);
506

507
 cleanup:
508 509
    if (ret == -1)
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_FORCE));
M
Michal Privoznik 已提交
510
    virDomainObjEndAPI(&vm);
511
    virObjectUnref(cfg);
512 513 514
}


515
void
516
qemuProcessShutdownOrReboot(virQEMUDriverPtr driver,
517
                            virDomainObjPtr vm)
518
{
519 520 521
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
522
        qemuDomainSetFakeReboot(driver, vm, false);
523
        virObjectRef(vm);
524 525 526 527 528
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
529
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
530
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
531
            virObjectUnref(vm);
532 533
        }
    } else {
534
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
535
    }
536
}
537

538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558

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);
559
    virObjectEventStateQueue(driver->domainEventState, event);
560 561 562 563 564

    return 0;
}


565 566
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
567
                          virDomainObjPtr vm,
568
                          virTristateBool guest_initiated,
569
                          void *opaque)
570
{
571
    virQEMUDriverPtr driver = opaque;
572
    qemuDomainObjPrivatePtr priv;
573
    virObjectEventPtr event = NULL;
574
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
575
    int detail = 0;
576

577 578
    VIR_DEBUG("vm=%p", vm);

579
    virObjectLock(vm);
580 581

    priv = vm->privateData;
J
Jiri Denemark 已提交
582
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_SHUTDOWN) {
583 584 585
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
586 587 588 589
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
590 591 592 593 594 595 596
    }

    VIR_DEBUG("Transitioned guest %s to shutdown state",
              vm->def->name);
    virDomainObjSetState(vm,
                         VIR_DOMAIN_SHUTDOWN,
                         VIR_DOMAIN_SHUTDOWN_UNKNOWN);
597 598 599 600 601 602 603 604 605 606

    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;

607 608
    case VIR_TRISTATE_BOOL_ABSENT:
    case VIR_TRISTATE_BOOL_LAST:
609 610 611 612 613
    default:
        detail = VIR_DOMAIN_EVENT_SHUTDOWN_FINISHED;
        break;
    }

614
    event = virDomainEventLifecycleNewFromObj(vm,
615 616
                                              VIR_DOMAIN_EVENT_SHUTDOWN,
                                              detail);
617

618
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
619 620 621 622
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

623 624 625
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

626 627
    qemuProcessShutdownOrReboot(driver, vm);

628
 unlock:
629
    virObjectUnlock(vm);
630
    virObjectEventStateQueue(driver->domainEventState, event);
631
    virObjectUnref(cfg);
632

633 634 635 636 637 638
    return 0;
}


static int
qemuProcessHandleStop(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
639 640
                      virDomainObjPtr vm,
                      void *opaque)
641
{
642
    virQEMUDriverPtr driver = opaque;
643
    virObjectEventPtr event = NULL;
644 645
    virDomainPausedReason reason = VIR_DOMAIN_PAUSED_UNKNOWN;
    virDomainEventSuspendedDetailType detail = VIR_DOMAIN_EVENT_SUSPENDED_PAUSED;
646
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
647

648
    virObjectLock(vm);
J
Jiri Denemark 已提交
649
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
650
        qemuDomainObjPrivatePtr priv = vm->privateData;
651

652
        if (priv->job.asyncJob == QEMU_ASYNC_JOB_MIGRATION_OUT) {
653 654
            if (priv->job.current->status ==
                        QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
655 656 657 658 659 660
                reason = VIR_DOMAIN_PAUSED_POSTCOPY;
                detail = VIR_DOMAIN_EVENT_SUSPENDED_POSTCOPY;
            } else {
                reason = VIR_DOMAIN_PAUSED_MIGRATION;
                detail = VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED;
            }
661 662 663 664
        }

        VIR_DEBUG("Transitioned guest %s to paused state, reason %s",
                  vm->def->name, virDomainPausedReasonTypeToString(reason));
665

666 667 668
        if (priv->job.current)
            ignore_value(virTimeMillisNow(&priv->job.current->stopped));

669 670 671 672
        if (priv->signalStop)
            virDomainObjBroadcast(vm);

        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
673
        event = virDomainEventLifecycleNewFromObj(vm,
674 675
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  detail);
676

677 678 679 680 681
        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));

682
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
683 684 685
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
686
    }
687

688
    virObjectUnlock(vm);
689
    virObjectEventStateQueue(driver->domainEventState, event);
690
    virObjectUnref(cfg);
691 692 693 694 695

    return 0;
}


696 697
static int
qemuProcessHandleResume(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
698 699
                        virDomainObjPtr vm,
                        void *opaque)
700
{
701
    virQEMUDriverPtr driver = opaque;
702
    virObjectEventPtr event = NULL;
703
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
704 705
    qemuDomainObjPrivatePtr priv;
    virDomainRunningReason reason = VIR_DOMAIN_RUNNING_UNPAUSED;
706
    virDomainEventResumedDetailType eventDetail;
707

708
    virObjectLock(vm);
709

710 711 712 713 714 715
    priv = vm->privateData;
    if (priv->runningReason != VIR_DOMAIN_RUNNING_UNKNOWN) {
        reason = priv->runningReason;
        priv->runningReason = VIR_DOMAIN_RUNNING_UNKNOWN;
    }

J
Jiri Denemark 已提交
716
    if (virDomainObjGetState(vm, NULL) != VIR_DOMAIN_RUNNING) {
717
        eventDetail = qemuDomainRunningReasonToResumeEvent(reason);
J
Jiri Denemark 已提交
718 719
        VIR_DEBUG("Transitioned guest %s into running state, reason '%s', "
                  "event detail %d",
720 721
                  vm->def->name, virDomainRunningReasonTypeToString(reason),
                  eventDetail);
722

723
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
724
        event = virDomainEventLifecycleNewFromObj(vm,
725 726
                                                  VIR_DOMAIN_EVENT_RESUMED,
                                                  eventDetail);
727

728
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
729 730 731 732 733
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
    }

734
    virObjectUnlock(vm);
735
    virObjectEventStateQueue(driver->domainEventState, event);
736
    virObjectUnref(cfg);
737 738 739
    return 0;
}

740 741 742
static int
qemuProcessHandleRTCChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm,
743 744
                           long long offset,
                           void *opaque)
745
{
746
    virQEMUDriverPtr driver = opaque;
747
    virObjectEventPtr event = NULL;
748
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
749

750
    virObjectLock(vm);
751

752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768
    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;
769
        vm->def->clock.data.variable.adjustment = offset;
770

771
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
772 773 774 775
           VIR_WARN("unable to save domain status with RTC change");
    }

    event = virDomainEventRTCChangeNewFromObj(vm, offset);
776

777
    virObjectUnlock(vm);
778

779
    virObjectEventStateQueue(driver->domainEventState, event);
780
    virObjectUnref(cfg);
781 782 783 784 785 786 787
    return 0;
}


static int
qemuProcessHandleWatchdog(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
788 789
                          int action,
                          void *opaque)
790
{
791
    virQEMUDriverPtr driver = opaque;
792 793
    virObjectEventPtr watchdogEvent = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
794
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
795

796
    virObjectLock(vm);
797 798 799
    watchdogEvent = virDomainEventWatchdogNewFromObj(vm, action);

    if (action == VIR_DOMAIN_EVENT_WATCHDOG_PAUSE &&
J
Jiri Denemark 已提交
800
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
801
        qemuDomainObjPrivatePtr priv = vm->privateData;
802 803
        VIR_DEBUG("Transitioned guest %s to paused state due to watchdog", vm->def->name);

J
Jiri Denemark 已提交
804
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
805
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
806 807 808
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

809 810 811 812 813
        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));

814
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
815 816 817
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
818 819 820
    }

    if (vm->def->watchdog->action == VIR_DOMAIN_WATCHDOG_ACTION_DUMP) {
C
Chen Fan 已提交
821 822 823 824
        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
825 826 827
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
828
            processEvent->vm = virObjectRef(vm);
C
Chen Fan 已提交
829
            if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
830
                if (!virObjectUnref(vm))
831
                    vm = NULL;
832
                qemuProcessEventFree(processEvent);
W
Wen Congyang 已提交
833
            }
834
        }
835 836
    }

837
    if (vm)
838
        virObjectUnlock(vm);
839 840
    virObjectEventStateQueue(driver->domainEventState, watchdogEvent);
    virObjectEventStateQueue(driver->domainEventState, lifecycleEvent);
841

842
    virObjectUnref(cfg);
843 844 845 846 847 848 849 850
    return 0;
}


static int
qemuProcessHandleIOError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virDomainObjPtr vm,
                         const char *diskAlias,
851
                         const char *nodename,
852
                         int action,
853 854
                         const char *reason,
                         void *opaque)
855
{
856
    virQEMUDriverPtr driver = opaque;
857 858 859
    virObjectEventPtr ioErrorEvent = NULL;
    virObjectEventPtr ioErrorEvent2 = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
860 861 862
    const char *srcPath;
    const char *devAlias;
    virDomainDiskDefPtr disk;
863
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
864

865
    virObjectLock(vm);
866 867 868 869 870 871 872 873 874 875

    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;
876 877

    if (disk) {
878
        srcPath = virDomainDiskGetSource(disk);
879 880 881 882 883 884 885 886 887 888
        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
889
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
890
        qemuDomainObjPrivatePtr priv = vm->privateData;
891 892
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

893 894 895
        if (priv->signalIOError)
            virDomainObjBroadcast(vm);

J
Jiri Denemark 已提交
896
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
897
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
898 899 900
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

901 902 903 904 905
        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));

906
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
907 908
            VIR_WARN("Unable to save status on vm %s after IO error", vm->def->name);
    }
909
    virObjectUnlock(vm);
910

911 912 913
    virObjectEventStateQueue(driver->domainEventState, ioErrorEvent);
    virObjectEventStateQueue(driver->domainEventState, ioErrorEvent2);
    virObjectEventStateQueue(driver->domainEventState, lifecycleEvent);
914
    virObjectUnref(cfg);
915 916 917
    return 0;
}

918 919 920 921 922
static int
qemuProcessHandleBlockJob(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          const char *diskAlias,
                          int type,
923
                          int status,
924
                          const char *error,
925
                          void *opaque)
926
{
927
    virQEMUDriverPtr driver = opaque;
928
    struct qemuProcessEvent *processEvent = NULL;
929
    virDomainDiskDefPtr disk;
930
    qemuDomainDiskPrivatePtr diskPriv;
931
    char *data = NULL;
932

933
    virObjectLock(vm);
934

935 936
    VIR_DEBUG("Block job for device %s (domain: %p,%s) type %d status %d",
              diskAlias, vm, vm->def->name, type, status);
937

938
    if (!(disk = qemuProcessFindDomainDiskByAliasOrQOM(vm, diskAlias, NULL)))
939
        goto error;
940
    diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
941

942
    if (diskPriv->blockJobSync) {
943
        /* We have a SYNC API waiting for this event, dispatch it back */
944 945
        diskPriv->blockJobType = type;
        diskPriv->blockJobStatus = status;
946 947
        VIR_FREE(diskPriv->blockJobError);
        ignore_value(VIR_STRDUP_QUIET(diskPriv->blockJobError, error));
948
        virDomainObjBroadcast(vm);
949 950 951 952 953 954 955 956 957
    } 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;
958
        processEvent->vm = virObjectRef(vm);
959 960
        processEvent->action = type;
        processEvent->status = status;
961

962 963 964 965
        if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
            ignore_value(virObjectUnref(vm));
            goto error;
        }
966 967
    }

968
 cleanup:
969
    virObjectUnlock(vm);
970
    return 0;
971
 error:
972
    qemuProcessEventFree(processEvent);
973
    goto cleanup;
974
}
975

976

977 978 979 980 981 982 983 984 985 986 987 988
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,
989 990
                          const char *saslUsername,
                          void *opaque)
991
{
992
    virQEMUDriverPtr driver = opaque;
993
    virObjectEventPtr event;
994 995 996
    virDomainEventGraphicsAddressPtr localAddr = NULL;
    virDomainEventGraphicsAddressPtr remoteAddr = NULL;
    virDomainEventGraphicsSubjectPtr subject = NULL;
997
    size_t i;
998 999

    if (VIR_ALLOC(localAddr) < 0)
1000
        goto error;
1001
    localAddr->family = localFamily;
1002 1003 1004
    if (VIR_STRDUP(localAddr->service, localService) < 0 ||
        VIR_STRDUP(localAddr->node, localNode) < 0)
        goto error;
1005 1006

    if (VIR_ALLOC(remoteAddr) < 0)
1007
        goto error;
1008
    remoteAddr->family = remoteFamily;
1009 1010 1011
    if (VIR_STRDUP(remoteAddr->service, remoteService) < 0 ||
        VIR_STRDUP(remoteAddr->node, remoteNode) < 0)
        goto error;
1012 1013

    if (VIR_ALLOC(subject) < 0)
1014
        goto error;
1015 1016
    if (x509dname) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1017
            goto error;
1018
        subject->nidentity++;
1019 1020 1021
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "x509dname") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, x509dname) < 0)
            goto error;
1022 1023 1024
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1025
            goto error;
1026
        subject->nidentity++;
1027 1028 1029
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "saslUsername") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, saslUsername) < 0)
            goto error;
1030 1031
    }

1032
    virObjectLock(vm);
1033
    event = virDomainEventGraphicsNewFromObj(vm, phase, localAddr, remoteAddr, authScheme, subject);
1034
    virObjectUnlock(vm);
1035

1036
    virObjectEventStateQueue(driver->domainEventState, event);
1037 1038 1039

    return 0;

1040
 error:
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
    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) {
1052
        for (i = 0; i < subject->nidentity; i++) {
1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
            VIR_FREE(subject->identities[i].type);
            VIR_FREE(subject->identities[i].name);
        }
        VIR_FREE(subject->identities);
        VIR_FREE(subject);
    }

    return -1;
}

1063 1064 1065 1066
static int
qemuProcessHandleTrayChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            const char *devAlias,
1067
                            const char *devid,
1068 1069
                            int reason,
                            void *opaque)
1070
{
1071
    virQEMUDriverPtr driver = opaque;
1072
    virObjectEventPtr event = NULL;
1073
    virDomainDiskDefPtr disk;
1074
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1075

1076
    virObjectLock(vm);
1077
    disk = qemuProcessFindDomainDiskByAliasOrQOM(vm, devAlias, devid);
1078 1079

    if (disk) {
1080
        event = virDomainEventTrayChangeNewFromObj(vm, disk->info.alias, reason);
1081 1082 1083 1084 1085 1086
        /* 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;

1087
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1088 1089 1090
            VIR_WARN("Unable to save status on vm %s after tray moved event",
                     vm->def->name);
        }
1091 1092

        virDomainObjBroadcast(vm);
1093 1094
    }

1095
    virObjectUnlock(vm);
1096
    virObjectEventStateQueue(driver->domainEventState, event);
1097
    virObjectUnref(cfg);
1098 1099 1100
    return 0;
}

O
Osier Yang 已提交
1101 1102
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1103 1104
                          virDomainObjPtr vm,
                          void *opaque)
O
Osier Yang 已提交
1105
{
1106
    virQEMUDriverPtr driver = opaque;
1107 1108
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1109
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1110

1111
    virObjectLock(vm);
O
Osier Yang 已提交
1112 1113
    event = virDomainEventPMWakeupNewFromObj(vm);

1114 1115 1116 1117 1118 1119 1120 1121 1122
    /* 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);
1123
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
1124 1125 1126
                                                  VIR_DOMAIN_EVENT_STARTED,
                                                  VIR_DOMAIN_EVENT_STARTED_WAKEUP);

1127
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1128 1129 1130 1131 1132
            VIR_WARN("Unable to save status on vm %s after wakeup event",
                     vm->def->name);
        }
    }

1133
    virObjectUnlock(vm);
1134 1135
    virObjectEventStateQueue(driver->domainEventState, event);
    virObjectEventStateQueue(driver->domainEventState, lifecycleEvent);
1136
    virObjectUnref(cfg);
O
Osier Yang 已提交
1137 1138
    return 0;
}
1139

O
Osier Yang 已提交
1140 1141
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1142 1143
                           virDomainObjPtr vm,
                           void *opaque)
O
Osier Yang 已提交
1144
{
1145
    virQEMUDriverPtr driver = opaque;
1146 1147
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1148
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1149

1150
    virObjectLock(vm);
O
Osier Yang 已提交
1151 1152
    event = virDomainEventPMSuspendNewFromObj(vm);

1153
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1154
        qemuDomainObjPrivatePtr priv = vm->privateData;
1155 1156 1157 1158 1159
        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
1160
        lifecycleEvent =
1161
            virDomainEventLifecycleNewFromObj(vm,
J
Jiri Denemark 已提交
1162 1163
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1164

1165
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1166 1167 1168
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1169 1170 1171

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1172 1173
    }

1174
    virObjectUnlock(vm);
O
Osier Yang 已提交
1175

1176 1177
    virObjectEventStateQueue(driver->domainEventState, event);
    virObjectEventStateQueue(driver->domainEventState, lifecycleEvent);
1178
    virObjectUnref(cfg);
O
Osier Yang 已提交
1179 1180 1181
    return 0;
}

1182 1183 1184
static int
qemuProcessHandleBalloonChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1185 1186
                               unsigned long long actual,
                               void *opaque)
1187
{
1188
    virQEMUDriverPtr driver = opaque;
1189
    virObjectEventPtr event = NULL;
1190
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1191

1192
    virObjectLock(vm);
1193 1194 1195 1196 1197 1198
    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;

1199
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
1200 1201
        VIR_WARN("unable to save domain status with balloon change");

1202
    virObjectUnlock(vm);
1203

1204
    virObjectEventStateQueue(driver->domainEventState, event);
1205
    virObjectUnref(cfg);
1206 1207 1208
    return 0;
}

1209 1210
static int
qemuProcessHandlePMSuspendDisk(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1211 1212
                               virDomainObjPtr vm,
                               void *opaque)
1213
{
1214
    virQEMUDriverPtr driver = opaque;
1215 1216
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1217
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1218

1219
    virObjectLock(vm);
1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
    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 =
1230
            virDomainEventLifecycleNewFromObj(vm,
1231 1232 1233
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_DISK);

1234
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1235 1236 1237 1238 1239 1240 1241 1242
            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);
    }

1243
    virObjectUnlock(vm);
1244

1245 1246
    virObjectEventStateQueue(driver->domainEventState, event);
    virObjectEventStateQueue(driver->domainEventState, lifecycleEvent);
1247 1248
    virObjectUnref(cfg);

1249 1250 1251
    return 0;
}

1252

1253 1254
static int
qemuProcessHandleGuestPanic(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1255
                            virDomainObjPtr vm,
1256
                            qemuMonitorEventPanicInfoPtr info,
1257
                            void *opaque)
1258
{
1259
    virQEMUDriverPtr driver = opaque;
1260 1261 1262
    struct qemuProcessEvent *processEvent;

    virObjectLock(vm);
1263
    if (VIR_ALLOC(processEvent) < 0)
1264 1265 1266 1267
        goto cleanup;

    processEvent->eventType = QEMU_PROCESS_EVENT_GUESTPANIC;
    processEvent->action = vm->def->onCrash;
1268
    processEvent->data = info;
1269 1270 1271
    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted before handling guest panic event is finished.
     */
1272 1273
    processEvent->vm = virObjectRef(vm);

1274 1275 1276
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        if (!virObjectUnref(vm))
            vm = NULL;
1277
        qemuProcessEventFree(processEvent);
1278 1279
    }

1280
 cleanup:
1281
    if (vm)
1282
        virObjectUnlock(vm);
1283 1284 1285 1286 1287

    return 0;
}


1288
int
1289 1290
qemuProcessHandleDeviceDeleted(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1291 1292
                               const char *devAlias,
                               void *opaque)
1293
{
1294
    virQEMUDriverPtr driver = opaque;
1295 1296
    struct qemuProcessEvent *processEvent = NULL;
    char *data;
1297 1298 1299 1300 1301 1302

    virObjectLock(vm);

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

1303 1304
    if (qemuDomainSignalDeviceRemoval(vm, devAlias,
                                      QEMU_DOMAIN_UNPLUGGING_DEVICE_STATUS_OK))
1305
        goto cleanup;
1306

1307 1308
    if (VIR_ALLOC(processEvent) < 0)
        goto error;
1309

1310 1311 1312 1313
    processEvent->eventType = QEMU_PROCESS_EVENT_DEVICE_DELETED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
1314
    processEvent->vm = virObjectRef(vm);
1315

1316 1317 1318 1319
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }
1320

1321
 cleanup:
1322 1323
    virObjectUnlock(vm);
    return 0;
1324
 error:
1325
    qemuProcessEventFree(processEvent);
1326
    goto cleanup;
1327 1328 1329
}


1330 1331 1332 1333 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
/**
 *
 * 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);
1392
    virObjectEventStateQueue(driver->domainEventState, event);
1393 1394 1395 1396 1397

    return 0;
}


1398 1399 1400 1401 1402 1403 1404 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
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);
1432
    virObjectEventStateQueue(driver->domainEventState, event);
1433 1434 1435 1436 1437

    return 0;
}


1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459
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;
1460
    processEvent->vm = virObjectRef(vm);
1461 1462 1463 1464 1465 1466 1467 1468 1469 1470

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

 cleanup:
    virObjectUnlock(vm);
    return 0;
 error:
1471
    qemuProcessEventFree(processEvent);
1472 1473 1474 1475
    goto cleanup;
}


1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499
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;
1500
    processEvent->vm = virObjectRef(vm);
1501 1502 1503 1504 1505 1506 1507 1508 1509 1510

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

 cleanup:
    virObjectUnlock(vm);
    return 0;
 error:
1511
    qemuProcessEventFree(processEvent);
1512 1513 1514 1515
    goto cleanup;
}


1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534
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;
1535
    virDomainObjBroadcast(vm);
1536 1537 1538 1539 1540 1541 1542

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1543 1544 1545 1546
static int
qemuProcessHandleMigrationStatus(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                 virDomainObjPtr vm,
                                 int status,
1547
                                 void *opaque)
1548 1549
{
    qemuDomainObjPrivatePtr priv;
1550 1551 1552 1553
    virQEMUDriverPtr driver = opaque;
    virObjectEventPtr event = NULL;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int reason;
1554 1555 1556 1557 1558 1559 1560 1561

    virObjectLock(vm);

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

    priv = vm->privateData;
1562
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_NONE) {
1563 1564 1565 1566
        VIR_DEBUG("got MIGRATION event without a migration job");
        goto cleanup;
    }

1567
    priv->job.current->stats.mig.status = status;
1568 1569
    virDomainObjBroadcast(vm);

1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
    if (status == QEMU_MONITOR_MIGRATION_STATUS_POSTCOPY &&
        virDomainObjGetState(vm, &reason) == VIR_DOMAIN_PAUSED &&
        reason == VIR_DOMAIN_PAUSED_MIGRATION) {
        VIR_DEBUG("Correcting paused state reason for domain %s to %s",
                  vm->def->name,
                  virDomainPausedReasonTypeToString(VIR_DOMAIN_PAUSED_POSTCOPY));

        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_POSTCOPY);
        event = virDomainEventLifecycleNewFromObj(vm,
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_POSTCOPY);

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

1588 1589
 cleanup:
    virObjectUnlock(vm);
1590 1591
    virObjectEventStateQueue(driver->domainEventState, event);
    virObjectUnref(cfg);
1592 1593 1594 1595
    return 0;
}


1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
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;
    }

1616
    virObjectEventStateQueue(driver->domainEventState,
1617 1618 1619 1620 1621 1622 1623 1624
                         virDomainEventMigrationIterationNewFromObj(vm, pass));

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663
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;
}


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 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717
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;
}


1718 1719
static qemuMonitorCallbacks monitorCallbacks = {
    .eofNotify = qemuProcessHandleMonitorEOF,
1720
    .errorNotify = qemuProcessHandleMonitorError,
1721
    .domainEvent = qemuProcessHandleEvent,
1722 1723
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
1724
    .domainResume = qemuProcessHandleResume,
1725 1726 1727 1728 1729
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1730
    .domainBlockJob = qemuProcessHandleBlockJob,
1731
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1732
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1733
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1734
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1735
    .domainPMSuspendDisk = qemuProcessHandlePMSuspendDisk,
1736
    .domainGuestPanic = qemuProcessHandleGuestPanic,
1737
    .domainDeviceDeleted = qemuProcessHandleDeviceDeleted,
1738
    .domainNicRxFilterChanged = qemuProcessHandleNicRxFilterChanged,
1739
    .domainSerialChange = qemuProcessHandleSerialChanged,
1740
    .domainSpiceMigrated = qemuProcessHandleSpiceMigrated,
1741
    .domainMigrationStatus = qemuProcessHandleMigrationStatus,
1742
    .domainMigrationPass = qemuProcessHandleMigrationPass,
1743
    .domainAcpiOstInfo = qemuProcessHandleAcpiOstInfo,
1744
    .domainBlockThreshold = qemuProcessHandleBlockThreshold,
1745
    .domainDumpCompleted = qemuProcessHandleDumpCompleted,
1746
    .domainPRManagerStatusChanged = qemuProcessHandlePRManagerStatusChanged,
1747 1748
};

1749 1750 1751 1752 1753 1754 1755 1756 1757 1758
static void
qemuProcessMonitorReportLogError(qemuMonitorPtr mon,
                                 const char *msg,
                                 void *opaque);


static void
qemuProcessMonitorLogFree(void *opaque)
{
    qemuDomainLogContextPtr logCtxt = opaque;
1759
    virObjectUnref(logCtxt);
1760 1761
}

1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781

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;
}


1782
static int
1783
qemuConnectMonitor(virQEMUDriverPtr driver, virDomainObjPtr vm, int asyncJob,
1784
                   bool retry, qemuDomainLogContextPtr logCtxt)
1785 1786
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1787
    qemuMonitorPtr mon = NULL;
1788
    unsigned long long timeout = 0;
1789
    virDomainChrSourceDefPtr monConfig;
1790

1791
    if (qemuSecuritySetDaemonSocketLabel(driver->securityManager, vm->def) < 0) {
1792 1793
        VIR_ERROR(_("Failed to set security context for monitor for %s"),
                  vm->def->name);
1794
        return -1;
1795 1796
    }

1797 1798 1799 1800 1801 1802
    /* 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);

1803
    /* Hold an extra reference because we can't allow 'vm' to be
C
Chen Hanxiao 已提交
1804
     * deleted until the monitor gets its own reference. */
1805
    virObjectRef(vm);
1806

1807
    ignore_value(virTimeMillisNow(&priv->monStart));
1808 1809
    monConfig = priv->monConfig;
    virObjectRef(monConfig);
1810
    virObjectUnlock(vm);
1811 1812

    mon = qemuMonitorOpen(vm,
1813
                          monConfig,
1814
                          priv->monJSON,
1815
                          retry,
1816
                          timeout,
1817 1818
                          &monitorCallbacks,
                          driver);
1819

1820
    if (mon && logCtxt) {
1821
        virObjectRef(logCtxt);
1822 1823 1824 1825 1826
        qemuMonitorSetDomainLog(mon,
                                qemuProcessMonitorReportLogError,
                                logCtxt,
                                qemuProcessMonitorLogFree);
    }
1827

1828
    virObjectLock(vm);
1829
    virObjectUnref(monConfig);
M
Michal Privoznik 已提交
1830
    virObjectUnref(vm);
1831
    priv->monStart = 0;
1832

M
Michal Privoznik 已提交
1833
    if (!virDomainObjIsActive(vm)) {
1834
        qemuMonitorClose(mon);
1835
        mon = NULL;
1836 1837 1838
    }
    priv->mon = mon;

1839
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0) {
1840 1841
        VIR_ERROR(_("Failed to clear security context for monitor for %s"),
                  vm->def->name);
1842
        return -1;
1843 1844 1845 1846
    }

    if (priv->mon == NULL) {
        VIR_INFO("Failed to connect monitor for %s", vm->def->name);
1847
        return -1;
1848 1849
    }

1850 1851
    if (qemuProcessInitMonitor(driver, vm, asyncJob) < 0)
        return -1;
1852

1853
    if (qemuMigrationCapsCheck(driver, vm, asyncJob) < 0)
1854 1855 1856
        return -1;

    return 0;
1857 1858
}

1859 1860 1861

/**
 * qemuProcessReadLog: Read log file of a qemu VM
1862
 * @logCtxt: the domain log context
1863
 * @msg: pointer to buffer to store the read messages in
1864
 * @max: maximum length of the message returned in @msg
1865 1866
 *
 * Reads log of a qemu VM. Skips messages not produced by qemu or irrelevant
1867 1868 1869 1870
 * 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
1871
 */
1872
static int
1873 1874 1875
qemuProcessReadLog(qemuDomainLogContextPtr logCtxt,
                   char **msg,
                   size_t max)
1876
{
1877 1878
    char *buf;
    ssize_t got;
1879
    char *eol;
1880
    char *filter_next;
1881
    size_t skip;
1882

1883
    if ((got = qemuDomainLogContextRead(logCtxt, &buf)) < 0)
1884
        return -1;
1885

1886 1887 1888 1889 1890
    /* Filter out debug messages from intermediate libvirt process */
    filter_next = buf;
    while ((eol = strchr(filter_next, '\n'))) {
        *eol = '\0';
        if (virLogProbablyLogMessage(filter_next) ||
1891
            strstr(filter_next, "char device redirected to")) {
1892
            skip = (eol + 1) - filter_next;
1893
            memmove(filter_next, eol + 1, buf + got - eol);
1894 1895 1896 1897
            got -= skip;
        } else {
            filter_next = eol + 1;
            *eol = '\n';
1898 1899
        }
    }
1900
    filter_next = NULL; /* silence false coverity warning */
1901

1902 1903
    if (got > 0 &&
        buf[got - 1] == '\n') {
1904 1905
        buf[got - 1] = '\0';
        got--;
1906
    }
1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919

    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;
    }

1920
    ignore_value(VIR_REALLOC_N_QUIET(buf, got + 1));
1921 1922 1923
    *msg = buf;
    return 0;
}
1924 1925


1926 1927
static int
qemuProcessReportLogError(qemuDomainLogContextPtr logCtxt,
1928 1929 1930
                          const char *msgprefix)
{
    char *logmsg = NULL;
1931 1932 1933 1934 1935 1936
    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;
1937

1938
    if (qemuProcessReadLog(logCtxt, &logmsg, max) < 0)
1939 1940 1941
        return -1;

    virResetLastError();
1942 1943 1944 1945 1946
    if (virStringIsEmpty(logmsg))
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s", msgprefix);
    else
        virReportError(VIR_ERR_INTERNAL_ERROR, _("%s: %s"), msgprefix, logmsg);

1947 1948
    VIR_FREE(logmsg);
    return 0;
1949 1950 1951
}


1952 1953 1954 1955 1956 1957 1958 1959 1960 1961
static void
qemuProcessMonitorReportLogError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                 const char *msg,
                                 void *opaque)
{
    qemuDomainLogContextPtr logCtxt = opaque;
    qemuProcessReportLogError(logCtxt, msg);
}


1962
static int
1963
qemuProcessLookupPTYs(virDomainChrDefPtr *devices,
1964
                      int count,
1965
                      virHashTablePtr info)
1966
{
1967
    char *id = NULL;
1968
    size_t i;
1969
    int ret = -1;
1970

1971
    for (i = 0; i < count; i++) {
1972
        virDomainChrDefPtr chr = devices[i];
1973
        if (chr->source->type == VIR_DOMAIN_CHR_TYPE_PTY) {
1974
            qemuMonitorChardevInfoPtr entry;
1975

1976 1977
            VIR_FREE(id);
            if (virAsprintf(&id, "char%s", chr->info.alias) < 0)
1978 1979
                return -1;

1980 1981
            entry = virHashLookup(info, id);
            if (!entry || !entry->ptyPath) {
1982
                if (chr->source->data.file.path == NULL) {
1983 1984 1985
                    /* neither the log output nor 'info chardev' had a
                     * pty path for this chardev, report an error
                     */
1986 1987
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1988
                    goto cleanup;
1989 1990 1991 1992 1993 1994 1995 1996
                } else {
                    /* 'info chardev' had no pty path for this chardev,
                     * but the log output had, so we're fine
                     */
                    continue;
                }
            }

1997 1998
            VIR_FREE(chr->source->data.file.path);
            if (VIR_STRDUP(chr->source->data.file.path, entry->ptyPath) < 0)
1999
                goto cleanup;
2000 2001 2002
        }
    }

2003 2004 2005 2006
    ret = 0;
 cleanup:
    VIR_FREE(id);
    return ret;
2007 2008
}

2009 2010
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
2011
                                     virHashTablePtr info)
2012
{
2013
    size_t i = 0;
C
Cole Robinson 已提交
2014

2015
    if (qemuProcessLookupPTYs(vm->def->serials, vm->def->nserials, info) < 0)
2016 2017
        return -1;

2018
    if (qemuProcessLookupPTYs(vm->def->parallels, vm->def->nparallels,
2019
                              info) < 0)
2020
        return -1;
2021

2022
    if (qemuProcessLookupPTYs(vm->def->channels, vm->def->nchannels, info) < 0)
2023
        return -1;
2024 2025 2026 2027
    /* 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];
2028

2029 2030 2031 2032 2033
        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;
2034 2035
            if (virDomainChrSourceDefCopy(chr->source,
                                          ((vm->def->serials[0])->source)) < 0)
2036 2037 2038 2039
                return -1;
        }
    }

2040
    if (qemuProcessLookupPTYs(vm->def->consoles + i, vm->def->nconsoles - i,
2041
                              info) < 0)
2042
        return -1;
2043 2044 2045 2046 2047

    return 0;
}


2048
static int
2049 2050 2051 2052
qemuProcessRefreshChannelVirtioState(virQEMUDriverPtr driver,
                                     virDomainObjPtr vm,
                                     virHashTablePtr info,
                                     int booted)
2053 2054
{
    size_t i;
2055
    int agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_CHANNEL;
2056
    qemuMonitorChardevInfoPtr entry;
2057
    virObjectEventPtr event = NULL;
2058 2059
    char *id = NULL;
    int ret = -1;
2060

2061 2062 2063
    if (booted)
        agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_DOMAIN_STARTED;

2064 2065 2066
    for (i = 0; i < vm->def->nchannels; i++) {
        virDomainChrDefPtr chr = vm->def->channels[i];
        if (chr->targetType == VIR_DOMAIN_CHR_CHANNEL_TARGET_TYPE_VIRTIO) {
2067 2068 2069 2070

            VIR_FREE(id);
            if (virAsprintf(&id, "char%s", chr->info.alias) < 0)
                goto cleanup;
2071 2072 2073 2074 2075 2076

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

2077 2078 2079 2080
            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)))
2081
                virObjectEventStateQueue(driver->domainEventState, event);
2082

2083 2084 2085 2086
            chr->state = entry->state;
        }
    }

2087 2088 2089 2090
    ret = 0;
 cleanup:
    VIR_FREE(id);
    return ret;
2091 2092 2093
}


2094 2095
int
qemuRefreshVirtioChannelState(virQEMUDriverPtr driver,
2096 2097
                              virDomainObjPtr vm,
                              qemuDomainAsyncJob asyncJob)
2098 2099 2100 2101 2102
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virHashTablePtr info = NULL;
    int ret = -1;

2103 2104 2105
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;

2106
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
2107 2108
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2109 2110 2111 2112

    if (ret < 0)
        goto cleanup;

2113
    ret = qemuProcessRefreshChannelVirtioState(driver, vm, info, false);
2114 2115 2116 2117 2118 2119

 cleanup:
    virHashFree(info);
    return ret;
}

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 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176

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;
}


2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200
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;
2201
    if ((then = mktime(&thenbits)) == (time_t)-1) {
2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212
        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;
}
2213

2214
int
2215 2216 2217 2218 2219 2220 2221 2222 2223
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 */
2224
    if (!virDomainDefHasMemballoon(vm->def)) {
2225
        vm->def->mem.cur_balloon = virDomainDefGetMemoryTotal(vm->def);
2226 2227 2228 2229 2230 2231 2232
        return 0;
    }

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

    rc = qemuMonitorGetBalloonInfo(qemuDomainGetMonitor(vm), &balloon);
2233
    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
2234 2235 2236 2237 2238 2239 2240 2241
        return -1;

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

    return 0;
}


2242
static int
2243
qemuProcessWaitForMonitor(virQEMUDriverPtr driver,
C
Cole Robinson 已提交
2244
                          virDomainObjPtr vm,
2245
                          int asyncJob,
2246
                          qemuDomainLogContextPtr logCtxt)
2247 2248
{
    int ret = -1;
2249
    virHashTablePtr info = NULL;
2250 2251 2252 2253 2254 2255
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool retry = true;

    if (priv->qemuCaps &&
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_CHARDEV_FD_PASS))
        retry = false;
2256

2257 2258 2259 2260
    VIR_DEBUG("Connect monitor to vm=%p name='%s' retry=%d",
              vm, vm->def->name, retry);

    if (qemuConnectMonitor(driver, vm, asyncJob, retry, logCtxt) < 0)
2261 2262 2263 2264 2265 2266
        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. */
2267 2268
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
2269 2270
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
    VIR_DEBUG("qemuMonitorGetChardevInfo returned %i", ret);
2271 2272 2273
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

2274
    if (ret == 0) {
2275
        if ((ret = qemuProcessFindCharDevicePTYsMonitor(vm, info)) < 0)
2276 2277
            goto cleanup;

2278 2279
        if ((ret = qemuProcessRefreshChannelVirtioState(driver, vm, info,
                                                        true)) < 0)
2280 2281
            goto cleanup;
    }
2282

2283
 cleanup:
2284
    virHashFree(info);
2285

2286 2287
    if (logCtxt && kill(vm->pid, 0) == -1 && errno == ESRCH) {
        qemuProcessReportLogError(logCtxt,
2288
                                  _("process exited while connecting to monitor"));
2289 2290 2291 2292 2293 2294
        ret = -1;
    }

    return ret;
}

2295

2296 2297 2298 2299 2300 2301
static int
qemuProcessDetectIOThreadPIDs(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2302
    qemuMonitorIOThreadInfoPtr *iothreads = NULL;
2303 2304 2305 2306
    int niothreads = 0;
    int ret = -1;
    size_t i;

2307 2308 2309 2310 2311
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_OBJECT_IOTHREAD)) {
        ret = 0;
        goto cleanup;
    }

2312 2313 2314 2315
    /* Get the list of IOThreads from qemu */
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
    niothreads = qemuMonitorGetIOThreads(priv->mon, &iothreads);
2316 2317
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
J
John Ferlan 已提交
2318
    if (niothreads < 0)
2319 2320
        goto cleanup;

2321
    if (niothreads != vm->def->niothreadids) {
2322 2323
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of IOThread pids from QEMU monitor. "
2324 2325
                         "got %d, wanted %zu"),
                       niothreads, vm->def->niothreadids);
2326 2327 2328
        goto cleanup;
    }

2329 2330 2331 2332 2333 2334
    /* Nothing to do */
    if (niothreads == 0) {
        ret = 0;
        goto cleanup;
    }

2335 2336 2337
    for (i = 0; i < niothreads; i++) {
        virDomainIOThreadIDDefPtr iothrid;

2338 2339
        if (!(iothrid = virDomainIOThreadIDFind(vm->def,
                                                iothreads[i]->iothread_id))) {
2340
            virReportError(VIR_ERR_INTERNAL_ERROR,
2341 2342
                           _("iothread %d not found"),
                           iothreads[i]->iothread_id);
2343 2344 2345 2346
            goto cleanup;
        }
        iothrid->thread_id = iothreads[i]->thread_id;
    }
2347 2348 2349 2350 2351 2352

    ret = 0;

 cleanup:
    if (iothreads) {
        for (i = 0; i < niothreads; i++)
2353
            VIR_FREE(iothreads[i]);
2354 2355 2356 2357 2358
        VIR_FREE(iothreads);
    }
    return ret;
}

2359 2360 2361 2362

/*
 * To be run between fork/exec of QEMU only
 */
2363
#if defined(HAVE_SCHED_GETAFFINITY) || defined(HAVE_BSD_CPU_AFFINITY)
2364
static int
2365
qemuProcessInitCpuAffinity(virDomainObjPtr vm)
2366 2367 2368 2369
{
    int ret = -1;
    virBitmapPtr cpumap = NULL;
    virBitmapPtr cpumapToSet = NULL;
2370
    virBitmapPtr hostcpumap = NULL;
2371
    qemuDomainObjPrivatePtr priv = vm->privateData;
2372

2373 2374 2375 2376 2377 2378
    if (!vm->pid) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Cannot setup CPU affinity until process is started"));
        return -1;
    }

2379 2380
    if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) {
        VIR_DEBUG("Set CPU affinity with advisory nodeset from numad");
2381
        cpumapToSet = priv->autoCpuset;
2382
    } else {
2383
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
2384
        if (vm->def->cpumask) {
H
Hu Tao 已提交
2385
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
2386 2387 2388 2389 2390
        } 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 */
2391 2392
            int hostcpus;

2393 2394 2395 2396 2397 2398 2399 2400 2401 2402
            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;
2403
                goto cleanup;
2404 2405 2406 2407 2408
            } 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;
2409

2410 2411
                if (hostcpus > QEMUD_CPUMASK_LEN)
                    hostcpus = QEMUD_CPUMASK_LEN;
2412

2413 2414 2415
                virBitmapFree(cpumap);
                if (!(cpumap = virBitmapNew(hostcpus)))
                    goto cleanup;
2416

2417
                virBitmapSetAll(cpumap);
2418

2419 2420
                cpumapToSet = cpumap;
            }
O
Osier Yang 已提交
2421
        }
2422 2423
    }

2424
    if (virProcessSetAffinity(vm->pid, cpumapToSet) < 0)
2425
        goto cleanup;
2426

2427 2428
    ret = 0;

2429
 cleanup:
2430
    virBitmapFree(cpumap);
2431
    virBitmapFree(hostcpumap);
2432
    return ret;
2433
}
2434 2435 2436 2437 2438 2439 2440
#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) */
2441

2442 2443
/* set link states to down on interfaces at qemu start */
static int
2444 2445 2446
qemuProcessSetLinkStates(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob)
2447 2448 2449
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2450
    size_t i;
2451 2452 2453 2454 2455
    int ret = -1;
    int rv;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2456 2457 2458

    for (i = 0; i < def->nnets; i++) {
        if (def->nets[i]->linkstate == VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN) {
2459 2460 2461
            if (!def->nets[i]->info.alias) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("missing alias for network device"));
2462
                goto cleanup;
2463 2464
            }

2465 2466
            VIR_DEBUG("Setting link state: %s", def->nets[i]->info.alias);

2467 2468 2469 2470
            rv = qemuMonitorSetLink(priv->mon,
                                    def->nets[i]->info.alias,
                                    VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (rv < 0) {
2471
                virReportError(VIR_ERR_OPERATION_FAILED,
2472 2473 2474
                               _("Couldn't set link state on interface: %s"),
                               def->nets[i]->info.alias);
                goto cleanup;
2475 2476 2477 2478
            }
        }
    }

2479 2480 2481 2482 2483
    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2484 2485 2486
    return ret;
}

2487

2488 2489 2490
/**
 * qemuProcessSetupPid:
 *
2491
 * This function sets resource properties (affinity, cgroups,
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
 * 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 */
2569
        if (virCgroupAddThread(cgroup, pid) < 0)
2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588
            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);
2589
        virCgroupFree(&cgroup);
2590 2591 2592 2593 2594 2595
    }

    return ret;
}


2596
static int
2597
qemuProcessSetupEmulator(virDomainObjPtr vm)
2598
{
2599 2600 2601 2602 2603
    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);
2604 2605
}

2606

2607 2608 2609 2610 2611 2612
static int
qemuProcessResctrlCreate(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    int ret = -1;
    size_t i = 0;
2613
    virCapsPtr caps = NULL;
2614 2615
    qemuDomainObjPrivatePtr priv = vm->privateData;

B
Bing Niu 已提交
2616
    if (!vm->def->nresctrls)
2617 2618 2619 2620 2621
        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);
2622 2623 2624
    if (!caps)
        return -1;

B
Bing Niu 已提交
2625
    for (i = 0; i < vm->def->nresctrls; i++) {
2626
        size_t j = 0;
2627
        if (virResctrlAllocCreate(caps->host.resctrl,
B
Bing Niu 已提交
2628
                                  vm->def->resctrls[i]->alloc,
2629 2630
                                  priv->machineName) < 0)
            goto cleanup;
2631 2632 2633 2634 2635 2636 2637 2638 2639

        for (j = 0; j < vm->def->resctrls[i]->nmonitors; j++) {
            virDomainResctrlMonDefPtr mon = NULL;

            mon = vm->def->resctrls[i]->monitors[j];
            if (virResctrlMonitorCreate(mon->instance,
                                        priv->machineName) < 0)
                goto cleanup;
        }
2640 2641 2642 2643 2644 2645 2646 2647 2648
    }

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


2649 2650 2651 2652 2653 2654 2655 2656 2657 2658
static char *
qemuProcessBuildPRHelperPidfilePath(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    const char *prdAlias = qemuDomainGetManagedPRAlias();

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


2659
void
2660
qemuProcessKillManagedPRDaemon(virDomainObjPtr vm)
2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700
{
    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;

2701 2702 2703
    if (qemuDomainNamespaceEnabled(vm, QEMU_DOMAIN_NS_MOUNT)) {
        if (virProcessGetNamespaces(vm->pid, &nfds, &fds) < 0)
            return ret;
2704

2705 2706 2707 2708
        if (nfds > 0 &&
            virProcessSetNamespaces(nfds, fds) < 0)
            goto cleanup;
    }
2709 2710 2711 2712 2713 2714 2715 2716 2717 2718

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


2719
int
2720
qemuProcessStartManagedPRDaemon(virDomainObjPtr vm)
2721 2722 2723 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
{
    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;

2750
    if (!(socketPath = qemuDomainGetManagedPRSocketPath(priv)))
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 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816
        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 &&
2817
        virCgroupAddMachineProcess(priv->cgroup, cpid) < 0)
2818 2819
        goto cleanup;

2820
    if (qemuSecurityDomainSetPathLabel(driver, vm, socketPath, true) < 0)
2821 2822 2823
        goto cleanup;

    priv->prDaemonRunning = true;
2824
    ret = 0;
2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842
 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;
}


2843
static int
2844
qemuProcessInitPasswords(virQEMUDriverPtr driver,
2845 2846
                         virDomainObjPtr vm,
                         int asyncJob)
2847 2848
{
    int ret = 0;
2849
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2850
    size_t i;
2851

2852
    for (i = 0; i < vm->def->ngraphics; ++i) {
2853 2854
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
2855 2856
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_VNC,
2857
                                                    &graphics->data.vnc.auth,
2858 2859
                                                    cfg->vncPassword,
                                                    asyncJob);
2860
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
2861 2862
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_SPICE,
2863
                                                    &graphics->data.spice.auth,
2864 2865
                                                    cfg->spicePassword,
                                                    asyncJob);
2866 2867
        }

2868 2869 2870
        if (ret < 0)
            goto cleanup;
    }
2871

2872
 cleanup:
2873
    virObjectUnref(cfg);
2874 2875 2876 2877 2878 2879 2880 2881 2882 2883
    return ret;
}


static int
qemuProcessPrepareChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
    int fd;
2884
    if (dev->source->type != VIR_DOMAIN_CHR_TYPE_FILE)
2885 2886
        return 0;

2887
    if ((fd = open(dev->source->data.file.path,
2888 2889 2890
                   O_CREAT | O_APPEND, S_IRUSR|S_IWUSR)) < 0) {
        virReportSystemError(errno,
                             _("Unable to pre-create chardev file '%s'"),
2891
                             dev->source->data.file.path);
2892 2893 2894 2895 2896 2897 2898 2899 2900
        return -1;
    }

    VIR_FORCE_CLOSE(fd);

    return 0;
}


2901 2902 2903 2904 2905
static int
qemuProcessCleanupChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
2906 2907 2908 2909
    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);
2910 2911 2912 2913 2914

    return 0;
}


2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950
/**
 * 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) {
2951
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VGAMEM) &&
2952 2953
                    qemuMonitorUpdateVideoMemorySize(priv->mon, video,
                                                     "qxl-vga") < 0)
2954
                        goto error;
2955

2956
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VRAM64) &&
2957 2958 2959
                    qemuMonitorUpdateVideoVram64Size(priv->mon, video,
                                                     "qxl-vga") < 0)
                    goto error;
2960
            } else {
2961 2962 2963 2964 2965 2966 2967 2968
                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)
2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987
                        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;
        }

    }

2988 2989
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2990 2991

    cfg = virQEMUDriverGetConfig(driver);
2992
    ret = virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps);
2993 2994 2995 2996 2997
    virObjectUnref(cfg);

    return ret;

 error:
2998
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
2999 3000 3001 3002
    return -1;
}


3003 3004
struct qemuProcessHookData {
    virDomainObjPtr vm;
3005
    virQEMUDriverPtr driver;
3006
    virQEMUDriverConfigPtr cfg;
3007 3008 3009 3010 3011
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
3012
    qemuDomainObjPrivatePtr priv = h->vm->privateData;
3013
    int ret = -1;
3014
    int fd;
3015 3016 3017
    virBitmapPtr nodeset = NULL;
    virDomainNumatuneMemMode mode;

3018 3019 3020 3021
    /* This method cannot use any mutexes, which are not
     * protected across fork()
     */

3022
    qemuSecurityPostFork(h->driver->securityManager);
3023 3024 3025 3026 3027

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

    VIR_DEBUG("Obtaining domain lock");
3028 3029 3030 3031 3032 3033 3034
    /*
     * 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.
     */
3035
    if (qemuSecuritySetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
3036
        goto cleanup;
3037
    if (virDomainLockProcessStart(h->driver->lockManager,
3038
                                  h->cfg->uri,
3039
                                  h->vm,
J
Ján Tomko 已提交
3040
                                  /* QEMU is always paused initially */
3041 3042
                                  true,
                                  &fd) < 0)
3043
        goto cleanup;
3044
    if (qemuSecurityClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
3045
        goto cleanup;
3046

3047
    if (qemuDomainBuildNamespace(h->cfg, h->driver->securityManager, h->vm) < 0)
3048 3049
        goto cleanup;

3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060
    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);
3061

3062 3063 3064
            if (virNumaSetupMemoryPolicy(mode, nodeset) < 0)
                goto cleanup;
        }
3065
    }
3066

3067 3068
    ret = 0;

3069
 cleanup:
3070
    virObjectUnref(h->cfg);
3071 3072
    VIR_DEBUG("Hook complete ret=%d", ret);
    return ret;
3073 3074 3075
}

int
3076 3077
qemuProcessPrepareMonitorChr(virDomainChrSourceDefPtr monConfig,
                             const char *domainDir)
3078 3079 3080 3081
{
    monConfig->type = VIR_DOMAIN_CHR_TYPE_UNIX;
    monConfig->data.nix.listen = true;

3082 3083
    if (virAsprintf(&monConfig->data.nix.path, "%s/monitor.sock",
                    domainDir) < 0)
3084 3085
        return -1;
    return 0;
3086 3087 3088
}


3089
/*
3090 3091
 * Precondition: vm must be locked, and a job must be active.
 * This method will call {Enter,Exit}Monitor
3092
 */
E
Eric Blake 已提交
3093
int
3094
qemuProcessStartCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
3095
                     virDomainRunningReason reason,
3096
                     qemuDomainAsyncJob asyncJob)
3097
{
3098
    int ret = -1;
3099
    qemuDomainObjPrivatePtr priv = vm->privateData;
3100
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3101

3102
    /* Bring up netdevs before starting CPUs */
3103
    if (qemuInterfaceStartDevices(vm->def) < 0)
3104 3105
       goto cleanup;

3106
    VIR_DEBUG("Using lock state '%s'", NULLSTR(priv->lockState));
3107
    if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
3108
                                   vm, priv->lockState) < 0) {
3109 3110 3111 3112
        /* 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
         */
3113
        goto cleanup;
3114 3115 3116
    }
    VIR_FREE(priv->lockState);

3117 3118
    priv->runningReason = reason;

3119 3120
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto release;
J
Jiri Denemark 已提交
3121

3122
    ret = qemuMonitorStartCPUs(priv->mon);
3123 3124
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3125 3126 3127 3128

    if (ret < 0)
        goto release;

3129 3130 3131 3132
    /* The RESUME event handler will change the domain state with the reason
     * saved in priv->runningReason and it will also emit corresponding domain
     * lifecycle event.
     */
3133

3134
 cleanup:
3135
    virObjectUnref(cfg);
3136
    return ret;
3137 3138

 release:
3139
    priv->runningReason = VIR_DOMAIN_RUNNING_UNKNOWN;
3140 3141 3142 3143
    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;
3144 3145 3146
}


3147 3148
int qemuProcessStopCPUs(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
3149
                        virDomainPausedReason reason,
3150
                        qemuDomainAsyncJob asyncJob)
3151
{
3152
    int ret = -1;
3153 3154
    qemuDomainObjPrivatePtr priv = vm->privateData;

3155
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
3156

3157 3158
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
J
Jiri Denemark 已提交
3159

3160
    ret = qemuMonitorStopCPUs(priv->mon);
3161 3162
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3163 3164 3165 3166

    if (ret < 0)
        goto cleanup;

3167 3168 3169
    /* de-activate netdevs after stopping CPUs */
    ignore_value(qemuInterfaceStopDevices(vm->def));

3170 3171 3172
    if (priv->job.current)
        ignore_value(virTimeMillisNow(&priv->job.current->stopped));

3173 3174 3175 3176
    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
3177

3178
 cleanup:
3179 3180 3181 3182 3183
    return ret;
}



3184
static void
3185 3186
qemuProcessNotifyNets(virDomainDefPtr def)
{
3187
    size_t i;
3188

3189 3190
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
3191 3192 3193 3194 3195 3196 3197 3198
        /* 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));

3199
        virDomainNetNotifyActualDevice(def, net);
3200 3201 3202
    }
}

3203 3204 3205 3206 3207 3208 3209 3210 3211
/* Attempt to instantiate the filters. Ignore failures because it's
 * possible that someone deleted a filter binding and the associated
 * filter while the guest was running and we don't want that action
 * to cause failure to keep the guest running during the reconnection
 * processing. Nor do we necessarily want other failures to do the
 * same. We'll just log the error conditions other than of course
 * ignoreExists possibility (e.g. the true flag) */
static void
qemuProcessFiltersInstantiate(virDomainDefPtr def)
3212
{
3213
    size_t i;
3214

3215
    for (i = 0; i < def->nnets; i++) {
3216 3217
        virDomainNetDefPtr net = def->nets[i];
        if ((net->filter) && (net->ifname)) {
3218 3219 3220 3221 3222 3223
            if (virDomainConfNWFilterInstantiate(def->name, def->uuid, net,
                                                 true) < 0) {
                VIR_WARN("filter '%s' instantiation for '%s' failed '%s'",
                         net->filter, net->ifname, virGetLastErrorMessage());
                virResetLastError();
            }
3224 3225 3226 3227
        }
    }
}

3228
static int
3229
qemuProcessUpdateState(virQEMUDriverPtr driver, virDomainObjPtr vm)
3230 3231 3232
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
3233
    virDomainPausedReason reason;
3234
    virDomainState newState = VIR_DOMAIN_NOSTATE;
3235
    int oldReason;
3236
    int newReason;
3237
    bool running;
3238
    char *msg = NULL;
3239 3240
    int ret;

3241
    qemuDomainObjEnterMonitor(driver, vm);
3242
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
3243 3244
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
3245

3246
    if (ret < 0)
3247 3248
        return -1;

3249
    state = virDomainObjGetState(vm, &oldReason);
3250

3251 3252 3253 3254 3255 3256 3257 3258
    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) {
3259 3260
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
3261
        ignore_value(VIR_STRDUP_QUIET(msg, "was unpaused"));
3262
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
3263 3264 3265
        if (reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN) {
            newState = VIR_DOMAIN_SHUTDOWN;
            newReason = VIR_DOMAIN_SHUTDOWN_UNKNOWN;
3266
            ignore_value(VIR_STRDUP_QUIET(msg, "shutdown"));
3267
        } else if (reason == VIR_DOMAIN_PAUSED_CRASHED) {
3268 3269 3270
            newState = VIR_DOMAIN_CRASHED;
            newReason = VIR_DOMAIN_CRASHED_PANICKED;
            ignore_value(VIR_STRDUP_QUIET(msg, "crashed"));
3271 3272 3273
        } else {
            newState = VIR_DOMAIN_PAUSED;
            newReason = reason;
S
Stefan Berger 已提交
3274 3275
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
3276 3277 3278 3279 3280 3281 3282
        }
    }

    if (newState != VIR_DOMAIN_NOSTATE) {
        VIR_DEBUG("Domain %s %s while its monitor was disconnected;"
                  " changing state to %s (%s)",
                  vm->def->name,
3283
                  NULLSTR(msg),
3284 3285 3286 3287
                  virDomainStateTypeToString(newState),
                  virDomainStateReasonToString(newState, newReason));
        VIR_FREE(msg);
        virDomainObjSetState(vm, newState, newReason);
3288 3289 3290 3291 3292
    }

    return 0;
}

3293
static int
3294 3295
qemuProcessRecoverMigrationIn(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
3296
                              const qemuDomainJobObj *job,
3297
                              virDomainState state,
3298
                              int reason)
3299
{
3300 3301 3302 3303 3304
    bool postcopy = (state == VIR_DOMAIN_PAUSED &&
                     reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED) ||
                    (state == VIR_DOMAIN_RUNNING &&
                     reason == VIR_DOMAIN_RUNNING_POSTCOPY);

3305
    switch ((qemuMigrationJobPhase) job->phase) {
3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320
    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;
3321

3322 3323 3324 3325 3326
    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);
3327
        if (qemuProcessStartCPUs(driver, vm,
3328
                                 VIR_DOMAIN_RUNNING_MIGRATED,
3329 3330 3331 3332 3333 3334 3335 3336
                                 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
3337 3338
         * we already started guest CPUs or we were in post-copy mode */
        if (postcopy) {
3339
            qemuMigrationAnyPostcopyFailed(driver, vm);
3340
        } else if (state != VIR_DOMAIN_RUNNING) {
3341
            VIR_DEBUG("Killing migrated domain %s", vm->def->name);
3342
            return -1;
3343 3344 3345
        }
        break;
    }
3346

3347 3348
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_NONE,
                             job->migParams, job->apiFlags);
3349 3350
    return 0;
}
3351

3352 3353 3354
static int
qemuProcessRecoverMigrationOut(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
3355
                               const qemuDomainJobObj *job,
3356
                               virDomainState state,
3357 3358
                               int reason,
                               unsigned int *stopFlags)
3359
{
3360 3361 3362
    bool postcopy = state == VIR_DOMAIN_PAUSED &&
                    (reason == VIR_DOMAIN_PAUSED_POSTCOPY ||
                     reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
3363
    bool resume = false;
3364

3365
    switch ((qemuMigrationJobPhase) job->phase) {
3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381
    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
3382 3383 3384 3385
         * domain; however we can only do that before migration enters
         * post-copy mode
         */
        if (postcopy) {
3386
            qemuMigrationAnyPostcopyFailed(driver, vm);
3387 3388 3389
        } else {
            VIR_DEBUG("Cancelling unfinished migration of domain %s",
                      vm->def->name);
3390
            if (qemuMigrationSrcCancel(driver, vm) < 0) {
3391 3392 3393
                VIR_WARN("Could not cancel ongoing migration of domain %s",
                         vm->def->name);
            }
3394
            resume = true;
3395
        }
3396
        break;
3397

3398 3399
    case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
        /* migration finished but we didn't have a chance to get the result
3400 3401
         * 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
3402
         */
3403
        if (postcopy)
3404
            qemuMigrationAnyPostcopyFailed(driver, vm);
3405
        break;
3406

3407
    case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
3408 3409 3410 3411 3412
        /* 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) {
3413
            qemuMigrationAnyPostcopyFailed(driver, vm);
3414 3415 3416
        } else {
            VIR_DEBUG("Resuming domain %s after failed migration",
                      vm->def->name);
3417
            resume = true;
3418 3419
        }
        break;
3420

3421 3422
    case QEMU_MIGRATION_PHASE_CONFIRM3:
        /* migration completed, we need to kill the domain here */
3423
        *stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
3424 3425
        return -1;
    }
3426

3427 3428 3429 3430 3431 3432 3433
    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)) {
3434
            if (qemuProcessStartCPUs(driver, vm,
3435
                                     VIR_DOMAIN_RUNNING_MIGRATION_CANCELED,
3436 3437 3438
                                     QEMU_ASYNC_JOB_NONE) < 0) {
                VIR_WARN("Could not resume domain %s", vm->def->name);
            }
3439 3440
        }
    }
3441

3442 3443
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_NONE,
                             job->migParams, job->apiFlags);
3444 3445 3446
    return 0;
}

3447
static int
3448
qemuProcessRecoverJob(virQEMUDriverPtr driver,
3449
                      virDomainObjPtr vm,
3450
                      const qemuDomainJobObj *job,
3451
                      unsigned int *stopFlags)
3452
{
3453
    qemuDomainObjPrivatePtr priv = vm->privateData;
3454 3455 3456 3457 3458 3459 3460
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
3461
        if (qemuProcessRecoverMigrationOut(driver, vm, job,
3462
                                           state, reason, stopFlags) < 0)
3463 3464 3465
            return -1;
        break;

3466
    case QEMU_ASYNC_JOB_MIGRATION_IN:
3467
        if (qemuProcessRecoverMigrationIn(driver, vm, job,
3468
                                          state, reason) < 0)
3469
            return -1;
3470 3471 3472 3473
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
3474
    case QEMU_ASYNC_JOB_SNAPSHOT:
3475
        qemuDomainObjEnterMonitor(driver, vm);
3476
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
3477 3478
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
3479
        /* resume the domain but only if it was paused as a result of
3480 3481 3482 3483 3484 3485 3486 3487 3488
         * 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 &&
3489 3490
               (reason == VIR_DOMAIN_PAUSED_SNAPSHOT ||
                reason == VIR_DOMAIN_PAUSED_MIGRATION)) ||
3491
              reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
3492
             if (qemuProcessStartCPUs(driver, vm,
3493
                                      VIR_DOMAIN_RUNNING_SAVE_CANCELED,
3494 3495 3496
                                      QEMU_ASYNC_JOB_NONE) < 0) {
                 VIR_WARN("Could not resume domain '%s' after migration to file",
                          vm->def->name);
3497 3498 3499 3500
            }
        }
        break;

3501 3502 3503 3504
    case QEMU_ASYNC_JOB_START:
        /* Already handled in VIR_DOMAIN_PAUSED_STARTING_UP check. */
        break;

3505 3506 3507 3508 3509 3510 3511 3512
    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        break;
    }

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

3513 3514 3515 3516
    /* 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.
     */
3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536
    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;

3537
    case QEMU_JOB_MIGRATION_OP:
3538
    case QEMU_JOB_ABORT:
3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549
    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;
}

3550 3551 3552 3553 3554 3555
static int
qemuProcessUpdateDevices(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDeviceDef dev;
3556
    const char **qemuDevices;
3557 3558 3559 3560 3561 3562 3563 3564 3565
    char **old;
    char **tmp;
    int ret = -1;

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

    old = priv->qemuDevices;
    priv->qemuDevices = NULL;
3566
    if (qemuDomainUpdateDeviceList(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
3567 3568
        goto cleanup;

3569
    qemuDevices = (const char **)priv->qemuDevices;
3570 3571
    if ((tmp = old)) {
        while (*tmp) {
3572
            if (!virStringListHasString(qemuDevices, *tmp) &&
3573 3574 3575 3576
                virDomainDefFindDevice(vm->def, *tmp, &dev, false) == 0 &&
                qemuDomainRemoveDevice(driver, vm, &dev) < 0) {
                goto cleanup;
            }
3577 3578 3579 3580 3581
            tmp++;
        }
    }
    ret = 0;

3582
 cleanup:
3583
    virStringListFree(old);
3584 3585 3586
    return ret;
}

3587 3588 3589 3590 3591 3592 3593
static int
qemuDomainPerfRestart(virDomainObjPtr vm)
{
    size_t i;
    virDomainDefPtr def = vm->def;
    qemuDomainObjPrivatePtr priv = vm->privateData;

3594
    if (!(priv->perf = virPerfNew()))
3595 3596 3597
        return -1;

    for (i = 0; i < VIR_PERF_EVENT_LAST; i++) {
3598 3599
        if (def->perf.events[i] &&
            def->perf.events[i] == VIR_TRISTATE_BOOL_YES) {
3600 3601 3602

            /* Failure to re-enable the perf event should not be fatal */
            if (virPerfEventEnable(priv->perf, i, vm->pid) < 0)
3603
                def->perf.events[i] = VIR_TRISTATE_BOOL_NO;
3604 3605 3606 3607 3608 3609
        }
    }

    return 0;
}

3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632

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;
        }
    }
}


3633
static bool
3634 3635
qemuProcessNeedHugepagesPath(virDomainDefPtr def,
                             virDomainMemoryDefPtr mem)
3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654
{
    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;
    }

3655 3656 3657 3658 3659 3660
    if (mem &&
        mem->model == VIR_DOMAIN_MEMORY_MODEL_DIMM &&
        mem->pagesize &&
        mem->pagesize != system_pagesize)
        return true;

3661 3662 3663 3664
    return false;
}


3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694
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;
}


3695 3696
static int
qemuProcessBuildDestroyMemoryPathsImpl(virQEMUDriverPtr driver,
3697
                                       virDomainObjPtr vm,
3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711
                                       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;
        }

3712
        if (qemuSecurityDomainSetPathLabel(driver, vm, path, true) < 0)
3713 3714
            return -1;
    } else {
M
Michal Privoznik 已提交
3715 3716
        if (virFileDeleteTree(path) < 0)
            return -1;
3717 3718 3719 3720 3721 3722
    }

    return 0;
}


3723
int
3724 3725 3726 3727
qemuProcessBuildDestroyMemoryPaths(virQEMUDriverPtr driver,
                                   virDomainObjPtr vm,
                                   virDomainMemoryDefPtr mem,
                                   bool build)
3728 3729
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3730
    char *path = NULL;
3731
    size_t i;
3732 3733
    bool shouldBuildHP = false;
    bool shouldBuildMB = false;
3734 3735
    int ret = -1;

3736 3737 3738 3739
    if (build) {
        shouldBuildHP = qemuProcessNeedHugepagesPath(vm->def, mem);
        shouldBuildMB = qemuProcessNeedMemoryBackingPath(vm->def, mem);
    }
3740

3741
    if (!build || shouldBuildHP) {
3742
        for (i = 0; i < cfg->nhugetlbfs; i++) {
3743
            path = qemuGetDomainHugepagePath(vm->def, &cfg->hugetlbfs[i]);
3744

3745
            if (!path)
3746 3747
                goto cleanup;

3748
            if (qemuProcessBuildDestroyMemoryPathsImpl(driver, vm,
3749
                                                       path, build) < 0)
3750
                goto cleanup;
3751

3752
            VIR_FREE(path);
3753 3754 3755
        }
    }

3756 3757 3758 3759
    if (!build || shouldBuildMB) {
        if (qemuGetMemoryBackingDomainPath(vm->def, cfg, &path) < 0)
            goto cleanup;

3760
        if (qemuProcessBuildDestroyMemoryPathsImpl(driver, vm,
3761 3762 3763 3764 3765 3766
                                                   path, build) < 0)
            goto cleanup;

        VIR_FREE(path);
    }

3767 3768
    ret = 0;
 cleanup:
3769
    VIR_FREE(path);
3770 3771 3772 3773 3774
    virObjectUnref(cfg);
    return ret;
}


3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800
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;
}


3801 3802 3803 3804
static int
qemuProcessVNCAllocatePorts(virQEMUDriverPtr driver,
                            virDomainGraphicsDefPtr graphics,
                            bool allocate)
3805
{
3806
    unsigned short port;
3807

3808 3809 3810
    if (!allocate) {
        if (graphics->data.vnc.autoport)
            graphics->data.vnc.port = 5900;
3811

3812 3813
        return 0;
    }
3814

3815 3816 3817 3818 3819
    if (graphics->data.vnc.autoport) {
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
            return -1;
        graphics->data.vnc.port = port;
    }
3820

3821 3822 3823 3824 3825 3826
    if (graphics->data.vnc.websocket == -1) {
        if (virPortAllocatorAcquire(driver->webSocketPorts, &port) < 0)
            return -1;
        graphics->data.vnc.websocket = port;
        graphics->data.vnc.websocketGenerated = true;
    }
3827

3828 3829
    return 0;
}
J
John Ferlan 已提交
3830

3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841
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;
3842

3843 3844
    bool needTLSPort = false;
    bool needPort = false;
3845

3846 3847 3848 3849 3850 3851 3852
    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;
3853

3854 3855 3856
            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
                needPort = true;
                break;
3857

3858 3859 3860 3861 3862 3863 3864 3865 3866
            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;
3867

3868 3869 3870
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
            needPort = true;
            break;
3871

3872 3873 3874 3875 3876 3877 3878
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
            if (cfg->spiceTLS)
                needTLSPort = true;
            needPort = true;
            break;
        }
    }
3879

3880 3881 3882
    if (!allocate) {
        if (needPort || graphics->data.spice.port == -1)
            graphics->data.spice.port = 5901;
3883

3884 3885
        if (needTLSPort || graphics->data.spice.tlsPort == -1)
            graphics->data.spice.tlsPort = 5902;
3886

3887 3888
        ret = 0;
        goto cleanup;
3889 3890
    }

3891 3892 3893
    if (needPort || graphics->data.spice.port == -1) {
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
            goto cleanup;
3894

3895 3896 3897 3898
        graphics->data.spice.port = port;

        if (!graphics->data.spice.autoport)
            graphics->data.spice.portReserved = true;
3899 3900
    }

3901 3902 3903 3904 3905 3906 3907
    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;
        }
3908

3909 3910
        if (virPortAllocatorAcquire(driver->remotePorts, &tlsPort) < 0)
            goto cleanup;
3911

3912
        graphics->data.spice.tlsPort = tlsPort;
3913

3914 3915
        if (!graphics->data.spice.autoport)
            graphics->data.spice.tlsPortReserved = true;
3916
    }
3917

3918
    ret = 0;
3919

3920 3921 3922 3923
 cleanup:
    virObjectUnref(cfg);
    return ret;
}
3924

3925

3926 3927 3928 3929 3930 3931 3932
static int
qemuProcessVerifyHypervFeatures(virDomainDefPtr def,
                                virCPUDataPtr cpu)
{
    char *cpuFeature;
    size_t i;
    int rc;
3933

3934 3935 3936 3937
    for (i = 0; i < VIR_DOMAIN_HYPERV_LAST; i++) {
        /* always supported string property */
        if (i == VIR_DOMAIN_HYPERV_VENDOR_ID)
            continue;
3938

3939 3940
        if (def->hyperv_features[i] != VIR_TRISTATE_SWITCH_ON)
            continue;
3941

3942 3943 3944
        if (virAsprintf(&cpuFeature, "__kvm_hv_%s",
                        virDomainHypervTypeToString(i)) < 0)
            return -1;
3945

3946 3947
        rc = virCPUDataCheckFeature(cpu, cpuFeature);
        VIR_FREE(cpuFeature);
3948

3949 3950 3951 3952
        if (rc < 0)
            return -1;
        else if (rc == 1)
            continue;
3953

3954 3955 3956 3957 3958 3959 3960
        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;
3961

3962 3963 3964 3965 3966
        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:
3967
        case VIR_DOMAIN_HYPERV_FREQUENCIES:
3968
        case VIR_DOMAIN_HYPERV_REENLIGHTENMENT:
3969
        case VIR_DOMAIN_HYPERV_TLBFLUSH:
3970
        case VIR_DOMAIN_HYPERV_IPI:
3971
        case VIR_DOMAIN_HYPERV_EVMCS:
3972 3973 3974 3975
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("host doesn't support hyperv '%s' feature"),
                           virDomainHypervTypeToString(i));
            return -1;
3976

3977 3978 3979 3980
        /* coverity[dead_error_begin] */
        case VIR_DOMAIN_HYPERV_VENDOR_ID:
        case VIR_DOMAIN_HYPERV_LAST:
            break;
3981
        }
3982
    }
3983 3984

    return 0;
3985 3986
}

3987

3988
static int
3989 3990
qemuProcessVerifyKVMFeatures(virDomainDefPtr def,
                             virCPUDataPtr cpu)
3991
{
3992
    int rc = 0;
3993

3994
    if (def->features[VIR_DOMAIN_FEATURE_PVSPINLOCK] != VIR_TRISTATE_SWITCH_ON)
3995 3996
        return 0;

3997 3998 3999 4000 4001 4002
    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"));
4003
        return -1;
4004
    }
4005

4006 4007
    return 0;
}
4008 4009


4010 4011 4012 4013 4014
static int
qemuProcessVerifyCPUFeatures(virDomainDefPtr def,
                             virCPUDataPtr cpu)
{
    int rc;
S
Stefan Berger 已提交
4015

4016
    rc = virCPUCheckFeature(def->os.arch, def->cpu, "invtsc");
4017

4018
    if (rc < 0) {
4019
        return -1;
4020 4021 4022 4023 4024 4025 4026 4027 4028
    } 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;
        }
4029
    }
4030

4031
    return 0;
4032 4033 4034
}


4035
static int
4036 4037 4038 4039 4040
qemuProcessFetchGuestCPU(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob,
                         virCPUDataPtr *enabled,
                         virCPUDataPtr *disabled)
4041
{
4042 4043 4044 4045
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCPUDataPtr dataEnabled = NULL;
    virCPUDataPtr dataDisabled = NULL;
    int rc;
4046

4047 4048
    *enabled = NULL;
    *disabled = NULL;
4049

4050
    if (!ARCH_IS_X86(vm->def->os.arch))
4051 4052
        return 0;

4053 4054
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto error;
4055

4056 4057
    rc = qemuMonitorGetGuestCPU(priv->mon, vm->def->os.arch,
                                &dataEnabled, &dataDisabled);
4058

4059 4060 4061 4062 4063 4064 4065 4066
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto error;

    if (rc == -1)
        goto error;

    *enabled = dataEnabled;
    *disabled = dataDisabled;
4067
    return 0;
4068 4069 4070 4071 4072

 error:
    virCPUDataFree(dataEnabled);
    virCPUDataFree(dataDisabled);
    return -1;
4073
}
4074

4075

4076
static int
4077 4078
qemuProcessVerifyCPU(virDomainObjPtr vm,
                     virCPUDataPtr cpu)
4079
{
4080
    virDomainDefPtr def = vm->def;
4081

4082 4083
    if (!cpu)
        return 0;
J
Jiri Denemark 已提交
4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100

    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;
}


4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142
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;
}


4143
static int
4144
qemuProcessUpdateAndVerifyCPU(virQEMUDriverPtr driver,
4145 4146
                              virDomainObjPtr vm,
                              qemuDomainAsyncJob asyncJob)
4147
{
4148
    virCPUDataPtr cpu = NULL;
4149
    virCPUDataPtr disabled = NULL;
4150
    int ret = -1;
4151

4152 4153
    if (qemuProcessFetchGuestCPU(driver, vm, asyncJob, &cpu, &disabled) < 0)
        goto cleanup;
4154

J
Jiri Denemark 已提交
4155 4156
    if (qemuProcessVerifyCPU(vm, cpu) < 0)
        goto cleanup;
J
Ján Tomko 已提交
4157

4158 4159
    if (qemuProcessUpdateLiveGuestCPU(vm, cpu, disabled) < 0)
        goto cleanup;
4160

4161
    ret = 0;
4162

4163
 cleanup:
4164
    virCPUDataFree(cpu);
4165
    virCPUDataFree(disabled);
4166 4167 4168 4169
    return ret;
}


4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193
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;
}


4194 4195 4196 4197 4198 4199 4200
static int
qemuProcessUpdateCPU(virQEMUDriverPtr driver,
                     virDomainObjPtr vm,
                     qemuDomainAsyncJob asyncJob)
{
    virCPUDataPtr cpu = NULL;
    virCPUDataPtr disabled = NULL;
4201
    virDomainCapsCPUModelsPtr models = NULL;
4202 4203
    int ret = -1;

4204 4205 4206 4207 4208
    /* 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;

4209 4210 4211 4212 4213 4214
    if (qemuProcessFetchGuestCPU(driver, vm, asyncJob, &cpu, &disabled) < 0)
        goto cleanup;

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

4215 4216 4217 4218
    if (!(models = qemuProcessFetchCPUDefinitions(driver, vm, asyncJob)) ||
        virCPUTranslate(vm->def->os.arch, vm->def->cpu, models) < 0)
        goto cleanup;

4219 4220 4221 4222 4223
    ret = 0;

 cleanup:
    virCPUDataFree(cpu);
    virCPUDataFree(disabled);
4224
    virObjectUnref(models);
4225 4226 4227 4228
    return ret;
}


4229 4230
static int
qemuPrepareNVRAM(virQEMUDriverConfigPtr cfg,
4231
                 virDomainObjPtr vm)
4232 4233 4234 4235
{
    int ret = -1;
    int srcFD = -1;
    int dstFD = -1;
4236
    virDomainLoaderDefPtr loader = vm->def->os.loader;
4237
    bool created = false;
4238 4239
    const char *master_nvram_path;
    ssize_t r;
4240

4241
    if (!loader || !loader->nvram || virFileExists(loader->nvram))
4242 4243
        return 0;

4244 4245 4246
    master_nvram_path = loader->templt;
    if (!loader->templt) {
        size_t i;
4247 4248 4249
        for (i = 0; i < cfg->nfirmwares; i++) {
            if (STREQ(cfg->firmwares[i]->name, loader->path)) {
                master_nvram_path = cfg->firmwares[i]->nvram;
4250
                break;
4251 4252
            }
        }
4253
    }
4254

4255 4256 4257 4258 4259 4260
    if (!master_nvram_path) {
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("unable to find any master var store for "
                         "loader: %s"), loader->path);
        goto cleanup;
    }
4261

4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278
    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;
4279

4280 4281
    do {
        char buf[1024];
4282

4283
        if ((r = saferead(srcFD, buf, sizeof(buf))) < 0) {
4284
            virReportSystemError(errno,
4285
                                 _("Unable to read from file '%s'"),
4286 4287 4288
                                 master_nvram_path);
            goto cleanup;
        }
4289 4290

        if (safewrite(dstFD, buf, r) < 0) {
4291
            virReportSystemError(errno,
4292
                                 _("Unable to write to file '%s'"),
4293 4294 4295
                                 loader->nvram);
            goto cleanup;
        }
4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308
    } 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;
4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325
    }

    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;
}


4326 4327 4328
static void
qemuLogOperation(virDomainObjPtr vm,
                 const char *msg,
4329 4330
                 virCommandPtr cmd,
                 qemuDomainLogContextPtr logCtxt)
4331 4332 4333 4334 4335
{
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int qemuVersion = virQEMUCapsGetVersion(priv->qemuCaps);
    const char *package = virQEMUCapsGetPackage(priv->qemuCaps);
4336
    char *hostname = virGetHostname();
4337 4338 4339
    struct utsname uts;

    uname(&uts);
4340 4341

    if ((timestamp = virTimeStringNow()) == NULL)
4342
        goto cleanup;
4343

4344
    if (qemuDomainLogContextWrite(logCtxt,
4345
                                  "%s: %s %s, qemu version: %d.%d.%d%s, kernel: %s, hostname: %s\n",
4346 4347 4348 4349
                                  timestamp, msg, VIR_LOG_VERSION_STRING,
                                  (qemuVersion / 1000000) % 1000,
                                  (qemuVersion / 1000) % 1000,
                                  qemuVersion % 1000,
4350
                                  package ? package : "",
4351
                                  uts.release,
4352
                                  hostname ? hostname : "") < 0)
4353
        goto cleanup;
4354

4355 4356 4357 4358 4359
    if (cmd) {
        char *args = virCommandToString(cmd);
        qemuDomainLogContextWrite(logCtxt, "%s\n", args);
        VIR_FREE(args);
    }
4360 4361

 cleanup:
4362
    VIR_FREE(hostname);
4363 4364 4365
    VIR_FREE(timestamp);
}

4366 4367 4368 4369 4370 4371 4372

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

4373
    VIR_FREE(inc->address);
4374
    VIR_FREE(inc->launchURI);
4375
    VIR_FREE(inc->deferredURI);
4376 4377 4378 4379 4380 4381 4382 4383
    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.
4384 4385 4386
 *
 * The caller is responsible for closing @fd, calling
 * qemuProcessIncomingDefFree will NOT close it.
4387 4388 4389
 */
qemuProcessIncomingDefPtr
qemuProcessIncomingDefNew(virQEMUCapsPtr qemuCaps,
4390
                          const char *listenAddress,
4391 4392 4393 4394 4395 4396
                          const char *migrateFrom,
                          int fd,
                          const char *path)
{
    qemuProcessIncomingDefPtr inc = NULL;

4397
    if (qemuMigrationDstCheckProtocol(qemuCaps, migrateFrom) < 0)
4398 4399 4400 4401 4402
        return NULL;

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

4403 4404 4405
    if (VIR_STRDUP(inc->address, listenAddress) < 0)
        goto error;

4406
    inc->launchURI = qemuMigrationDstGetURI(migrateFrom, fd);
4407 4408 4409
    if (!inc->launchURI)
        goto error;

4410 4411 4412 4413 4414 4415
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_INCOMING_DEFER)) {
        inc->deferredURI = inc->launchURI;
        if (VIR_STRDUP(inc->launchURI, "defer") < 0)
            goto error;
    }

4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426
    inc->fd = fd;
    inc->path = path;

    return inc;

 error:
    qemuProcessIncomingDefFree(inc);
    return NULL;
}


4427 4428 4429 4430 4431 4432 4433 4434
/*
 * 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,
4435
                    virDomainObjPtr vm,
4436 4437
                    virDomainJobOperation operation,
                    unsigned long apiFlags)
4438
{
4439
    if (qemuDomainObjBeginAsyncJob(driver, vm, QEMU_ASYNC_JOB_START,
4440
                                   operation, apiFlags) < 0)
4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455
        return -1;

    qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
    return 0;
}


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


4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478
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;
}


4479
static int
4480
qemuProcessGraphicsReservePorts(virDomainGraphicsDefPtr graphics,
4481
                                bool reconnect)
4482
{
4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493
    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;

4494 4495
    switch (graphics->type) {
    case VIR_DOMAIN_GRAPHICS_TYPE_VNC:
4496 4497
        if (!graphics->data.vnc.autoport ||
            reconnect) {
4498
            if (virPortAllocatorSetUsed(graphics->data.vnc.port) < 0)
4499 4500 4501
                return -1;
            graphics->data.vnc.portReserved = true;
        }
4502
        if (graphics->data.vnc.websocket > 0 &&
4503
            virPortAllocatorSetUsed(graphics->data.vnc.websocket) < 0)
4504
            return -1;
4505 4506 4507
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SPICE:
4508
        if (graphics->data.spice.autoport && !reconnect)
4509
            return 0;
4510

4511
        if (graphics->data.spice.port > 0) {
4512
            if (virPortAllocatorSetUsed(graphics->data.spice.port) < 0)
4513
                return -1;
4514 4515
            graphics->data.spice.portReserved = true;
        }
4516

4517
        if (graphics->data.spice.tlsPort > 0) {
4518
            if (virPortAllocatorSetUsed(graphics->data.spice.tlsPort) < 0)
4519 4520
                return -1;
            graphics->data.spice.tlsPortReserved = true;
4521
        }
4522 4523 4524 4525 4526
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SDL:
    case VIR_DOMAIN_GRAPHICS_TYPE_RDP:
    case VIR_DOMAIN_GRAPHICS_TYPE_DESKTOP:
4527
    case VIR_DOMAIN_GRAPHICS_TYPE_EGL_HEADLESS:
4528 4529
    case VIR_DOMAIN_GRAPHICS_TYPE_LAST:
        break;
4530 4531
    }

4532 4533 4534 4535
    return 0;
}


4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565
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:
4566
    case VIR_DOMAIN_GRAPHICS_TYPE_EGL_HEADLESS:
4567 4568 4569 4570 4571 4572 4573
    case VIR_DOMAIN_GRAPHICS_TYPE_LAST:
        break;
    }

    return 0;
}

4574
static int
4575
qemuProcessGetNetworkAddress(const char *netname,
4576 4577
                             char **netaddr)
{
4578
    virConnectPtr conn = NULL;
4579 4580 4581 4582 4583 4584 4585 4586 4587 4588
    int ret = -1;
    virNetworkPtr net;
    virNetworkDefPtr netdef = NULL;
    virNetworkIPDefPtr ipdef;
    virSocketAddr addr;
    virSocketAddrPtr addrptr = NULL;
    char *dev_name = NULL;
    char *xml = NULL;

    *netaddr = NULL;
4589 4590 4591 4592

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

4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604
    net = virNetworkLookupByName(conn, netname);
    if (!net)
        goto cleanup;

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

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

4605
    switch ((virNetworkForwardType) netdef->forward.type) {
4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641
    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;
4642 4643 4644 4645 4646 4647 4648 4649

    case VIR_NETWORK_FORWARD_HOSTDEV:
        break;

    case VIR_NETWORK_FORWARD_LAST:
    default:
        virReportEnumRangeError(virNetworkForwardType, netdef->forward.type);
        goto cleanup;
4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666
    }

    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);
4667
    virObjectUnref(conn);
4668 4669 4670 4671
    VIR_FREE(xml);
    return ret;
}

4672

4673
static int
4674
qemuProcessGraphicsSetupNetworkAddress(virDomainGraphicsListenDefPtr glisten,
4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685
                                       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;
    }

4686
    rc = qemuProcessGetNetworkAddress(glisten->network, &glisten->address);
4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699
    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;
}


4700
static int
4701
qemuProcessGraphicsSetupListen(virQEMUDriverPtr driver,
4702 4703
                               virDomainGraphicsDefPtr graphics,
                               virDomainObjPtr vm)
4704
{
4705
    qemuDomainObjPrivatePtr priv = vm->privateData;
4706
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4707
    const char *type = virDomainGraphicsTypeToString(graphics->type);
4708
    char *listenAddr = NULL;
4709
    bool useSocket = false;
4710
    size_t i;
4711
    int ret = -1;
4712 4713 4714

    switch (graphics->type) {
    case VIR_DOMAIN_GRAPHICS_TYPE_VNC:
4715
        useSocket = cfg->vncAutoUnixSocket;
4716 4717 4718 4719
        listenAddr = cfg->vncListen;
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SPICE:
4720
        useSocket = cfg->spiceAutoUnixSocket;
4721 4722 4723 4724 4725 4726
        listenAddr = cfg->spiceListen;
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SDL:
    case VIR_DOMAIN_GRAPHICS_TYPE_RDP:
    case VIR_DOMAIN_GRAPHICS_TYPE_DESKTOP:
4727
    case VIR_DOMAIN_GRAPHICS_TYPE_EGL_HEADLESS:
4728 4729 4730 4731 4732 4733 4734 4735 4736
    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:
4737 4738 4739 4740 4741
            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) {
4742 4743
                    memset(glisten, 0, sizeof(virDomainGraphicsListenDef));
                    if (virAsprintf(&glisten->socket, "%s/%s.sock",
4744
                                    priv->libDir, type) < 0)
4745
                        goto cleanup;
4746 4747
                    glisten->fromConfig = true;
                    glisten->type = VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_SOCKET;
4748 4749
                } else if (listenAddr) {
                    if (VIR_STRDUP(glisten->address, listenAddr) < 0)
4750
                        goto cleanup;
4751 4752 4753
                    glisten->fromConfig = true;
                }
            }
4754 4755 4756 4757 4758 4759
            break;

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

4760
            if (qemuProcessGraphicsSetupNetworkAddress(glisten,
4761
                                                       listenAddr) < 0)
4762
                goto cleanup;
4763 4764
            break;

4765 4766 4767 4768
        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_SOCKET:
            if (!glisten->socket) {
                if (virAsprintf(&glisten->socket, "%s/%s.sock",
                                priv->libDir, type) < 0)
4769
                    goto cleanup;
4770 4771 4772 4773
                glisten->autoGenerated = true;
            }
            break;

4774 4775 4776 4777 4778 4779
        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_NONE:
        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_LAST:
            break;
        }
    }

4780 4781 4782 4783 4784
    ret = 0;

 cleanup:
    virObjectUnref(cfg);
    return ret;
4785 4786 4787
}


4788 4789 4790 4791
static int
qemuProcessGraphicsSetupRenderNode(virDomainGraphicsDefPtr graphics,
                                   virQEMUCapsPtr qemuCaps)
{
4792 4793
    char **rendernode = NULL;

4794 4795 4796 4797
    if (!virDomainGraphicsNeedsAutoRenderNode(graphics))
        return 0;

    /* Don't bother picking a DRM node if QEMU doesn't support it. */
4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808
    if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
        if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_SPICE_RENDERNODE))
            return 0;

        rendernode = &graphics->data.spice.rendernode;
    } else {
        if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_EGL_HEADLESS_RENDERNODE))
            return 0;

        rendernode = &graphics->data.egl_headless.rendernode;
    }
4809

4810
    if (!(*rendernode = virHostGetDRMRenderNode()))
4811 4812 4813 4814 4815 4816
        return -1;

    return 0;
}


4817
static int
4818
qemuProcessSetupGraphics(virQEMUDriverPtr driver,
4819
                         virDomainObjPtr vm,
4820
                         virQEMUCapsPtr qemuCaps,
4821
                         unsigned int flags)
4822
{
4823
    virDomainGraphicsDefPtr graphics;
4824
    bool allocate = !(flags & VIR_QEMU_PROCESS_START_PRETEND);
4825
    size_t i;
4826 4827
    int ret = -1;

4828 4829 4830
    for (i = 0; i < vm->def->ngraphics; i++) {
        graphics = vm->def->graphics[i];

4831 4832 4833
        if (qemuProcessGraphicsSetupRenderNode(graphics, qemuCaps) < 0)
            goto cleanup;

4834
        if (qemuProcessGraphicsSetupListen(driver, graphics, vm) < 0)
4835 4836 4837
            goto cleanup;
    }

4838 4839 4840 4841
    if (allocate) {
        for (i = 0; i < vm->def->ngraphics; i++) {
            graphics = vm->def->graphics[i];

4842
            if (qemuProcessGraphicsReservePorts(graphics, false) < 0)
4843 4844 4845
                goto cleanup;
        }
    }
4846

4847
    for (i = 0; i < vm->def->ngraphics; ++i) {
4848
        graphics = vm->def->graphics[i];
4849

4850 4851
        if (qemuProcessGraphicsAllocatePorts(driver, graphics, allocate) < 0)
            goto cleanup;
4852 4853 4854 4855 4856 4857 4858 4859 4860
    }

    ret = 0;

 cleanup:
    return ret;
}


4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893
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++) {
4894
            if (!virHostdevIsSCSIDevice(vm->def->hostdevs[i]))
4895 4896
                continue;

4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922
            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;
}


4923 4924 4925 4926 4927 4928 4929 4930 4931
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;

4932
    if (!virDomainDefHasMemballoon(vm->def))
4933 4934 4935
        return 0;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
4936
        return -1;
4937

4938 4939 4940
    if (vm->def->memballoon->period)
        qemuMonitorSetMemoryStatsPeriod(priv->mon, vm->def->memballoon,
                                        vm->def->memballoon->period);
4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952
    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
4953 4954 4955
static int
qemuProcessMakeDir(virQEMUDriverPtr driver,
                   virDomainObjPtr vm,
4956
                   const char *path)
J
Jiri Denemark 已提交
4957 4958 4959 4960 4961 4962 4963 4964
{
    int ret = -1;

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

4965
    if (qemuSecurityDomainSetPathLabel(driver, vm, path, true) < 0)
J
Jiri Denemark 已提交
4966 4967 4968 4969 4970 4971 4972 4973 4974
        goto cleanup;

    ret = 0;

 cleanup:
    return ret;
}


4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010
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) ==
5011
                VIR_DOMAIN_MEMORY_ACCESS_SHARED) {
5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023
                shmem = true;
                break;
            }
        }
    }

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

5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046

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:
5047
        case VIR_DOMAIN_GRAPHICS_TYPE_EGL_HEADLESS:
5048 5049 5050 5051 5052 5053 5054 5055 5056
        case VIR_DOMAIN_GRAPHICS_TYPE_LAST:
            break;
        }
    }

    return 0;
}


5057 5058 5059 5060 5061 5062 5063 5064 5065 5066
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];

5067 5068 5069 5070 5071 5072 5073 5074 5075
        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 &&
5076 5077 5078 5079
             !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))) {
5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101
            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;
}


5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132
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;
}


5133
static int
5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152
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;
}


5153 5154 5155 5156 5157 5158 5159
static int
qemuProcessStartValidateDisks(virDomainObjPtr vm,
                              virQEMUCapsPtr qemuCaps)
{
    size_t i;

    for (i = 0; i < vm->def->ndisks; i++) {
5160 5161
        virDomainDiskDefPtr disk = vm->def->disks[i];
        virStorageSourcePtr src = disk->src;
5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173

        /* 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;
        }
5174 5175 5176 5177 5178 5179 5180 5181

        /* 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;
        }
5182 5183 5184 5185 5186 5187
    }

    return 0;
}


5188 5189 5190 5191 5192 5193
/**
 * qemuProcessStartValidate:
 * @vm: domain object
 * @qemuCaps: emulator capabilities
 * @migration: restoration of existing state
 *
5194 5195 5196 5197 5198
 * 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.
5199
 */
5200
static int
5201 5202
qemuProcessStartValidate(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
5203
                         virQEMUCapsPtr qemuCaps,
5204
                         virCapsPtr caps,
5205
                         unsigned int flags)
5206
{
5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220
    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");
5221
        if (qemuSecurityCheckAllLabel(driver->securityManager, vm->def) < 0)
5222 5223 5224 5225
            return -1;

    }

5226 5227 5228 5229 5230
    /* 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 */
    if ((flags & VIR_QEMU_PROCESS_START_NEW) &&
        virDomainDefValidate(vm->def, caps, 0, driver->xmlopt) < 0)
5231 5232
        return -1;

5233 5234 5235
    if (qemuProcessStartValidateGraphics(vm) < 0)
        return -1;

5236 5237 5238
    if (qemuProcessStartValidateVideo(vm, qemuCaps) < 0)
        return -1;

5239 5240 5241
    if (qemuProcessStartValidateIOThreads(vm, qemuCaps) < 0)
        return -1;

5242 5243 5244
    if (qemuProcessStartValidateShmem(vm) < 0)
        return -1;

5245 5246 5247 5248
    if (vm->def->cpu &&
        virCPUValidateFeatures(vm->def->os.arch, vm->def->cpu) < 0)
        return -1;

5249 5250 5251
    if (qemuProcessStartValidateDisks(vm, qemuCaps) < 0)
        return -1;

5252 5253
    VIR_DEBUG("Checking for any possible (non-fatal) issues");

5254
    qemuProcessStartWarnShmem(vm);
5255

5256
    return 0;
5257 5258 5259
}


J
Jiri Denemark 已提交
5260 5261 5262 5263 5264 5265
/**
 * qemuProcessInit:
 *
 * Prepares the domain up to the point when priv->qemuCaps is initialized. The
 * function calls qemuProcessStop when needed.
 *
5266 5267 5268 5269
 * 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
5270 5271 5272 5273 5274
 * Returns 0 on success, -1 on error.
 */
int
qemuProcessInit(virQEMUDriverPtr driver,
                virDomainObjPtr vm,
5275
                virCPUDefPtr updatedCPU,
5276
                qemuDomainAsyncJob asyncJob,
5277
                bool migration,
5278
                unsigned int flags)
J
Jiri Denemark 已提交
5279 5280 5281 5282
{
    virCapsPtr caps = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int stopFlags;
5283
    virCPUDefPtr origCPU = NULL;
J
Jiri Denemark 已提交
5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297
    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)))
5298 5299
        goto cleanup;

5300 5301 5302 5303 5304 5305 5306 5307 5308
    /* 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;
    }

5309
    VIR_DEBUG("Determining emulator version");
5310
    if (qemuDomainUpdateQEMUCaps(vm, driver->qemuCapsCache) < 0)
5311
        goto cleanup;
J
Jiri Denemark 已提交
5312

5313 5314 5315
    if (flags & VIR_QEMU_PROCESS_START_STANDALONE)
        virQEMUCapsClear(priv->qemuCaps, QEMU_CAPS_CHARDEV_FD_PASS);

5316 5317 5318
    if (qemuDomainUpdateCPU(vm, updatedCPU, &origCPU) < 0)
        goto cleanup;

5319
    if (qemuProcessStartValidate(driver, vm, priv->qemuCaps, caps, flags) < 0)
5320 5321
        goto cleanup;

J
Jiri Denemark 已提交
5322 5323 5324 5325 5326
    /* 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");
5327
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm) < 0)
5328
        goto cleanup;
J
Jiri Denemark 已提交
5329

5330
    if (flags & VIR_QEMU_PROCESS_START_PRETEND) {
5331 5332 5333 5334
        if (qemuDomainSetPrivatePaths(driver, vm) < 0) {
            virDomainObjRemoveTransientDef(vm);
            goto cleanup;
        }
5335
    } else {
5336 5337 5338
        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
5339

5340 5341
        if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
            driver->inhibitCallback(true, driver->inhibitOpaque);
J
Jiri Denemark 已提交
5342

5343 5344 5345 5346 5347
        /* 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
5348

5349 5350
        if (qemuDomainSetPrivatePaths(driver, vm) < 0)
            goto stop;
5351 5352

        VIR_STEAL_PTR(priv->origCPU, origCPU);
5353
    }
5354

J
Jiri Denemark 已提交
5355 5356 5357
    ret = 0;

 cleanup:
5358
    virCPUDefFree(origCPU);
J
Jiri Denemark 已提交
5359 5360 5361 5362 5363 5364 5365
    virObjectUnref(caps);
    return ret;

 stop:
    stopFlags = VIR_QEMU_PROCESS_STOP_NO_RELABEL;
    if (migration)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
5366
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, asyncJob, stopFlags);
J
Jiri Denemark 已提交
5367 5368 5369 5370
    goto cleanup;
}


5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381
/**
 * qemuProcessNetworkPrepareDevices
 */
static int
qemuProcessNetworkPrepareDevices(virDomainDefPtr def)
{
    int ret = -1;
    size_t i;

    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
5382
        virDomainNetType actualType;
5383 5384 5385 5386 5387

        /* 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.
         */
5388
        if (virDomainNetAllocateActualDevice(def, net) < 0)
5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423
            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;
}


5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439
/**
 * 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);
5440
    virDomainVcpuDefPtr vcpu = virDomainDefGetVcpu(vm->def, vcpuid);
5441
    virDomainResctrlMonDefPtr mon = NULL;
5442 5443 5444 5445 5446 5447 5448 5449 5450
    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
5451
    for (i = 0; i < vm->def->nresctrls; i++) {
5452
        size_t j = 0;
B
Bing Niu 已提交
5453
        virDomainResctrlDefPtr ct = vm->def->resctrls[i];
5454 5455 5456 5457

        if (virBitmapIsBitSet(ct->vcpus, vcpuid)) {
            if (virResctrlAllocAddPID(ct->alloc, vcpupid) < 0)
                return -1;
5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471

            for (j = 0; j < ct->nmonitors; j++) {
                mon = ct->monitors[j];

                if (virBitmapEqual(ct->vcpus, mon->vcpus))
                    continue;

                if (virBitmapIsBitSet(mon->vcpus, vcpuid)) {
                    if (virResctrlMonitorAddPID(mon->instance, vcpupid) < 0)
                        return -1;
                    break;
                }
            }

5472 5473 5474
            break;
        }
    }
5475

5476
    return 0;
5477 5478 5479 5480 5481 5482
}


static int
qemuProcessSetupVcpus(virDomainObjPtr vm)
{
5483
    virDomainVcpuDefPtr vcpu;
5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528
    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;
}


5529 5530 5531 5532
int
qemuProcessSetupIOThread(virDomainObjPtr vm,
                         virDomainIOThreadIDDefPtr iothread)
{
5533 5534
    return qemuProcessSetupPid(vm, iothread->thread_id,
                               VIR_CGROUP_THREAD_IOTHREAD,
5535
                               iothread->iothread_id,
5536
                               iothread->cpumask,
5537 5538
                               vm->def->cputune.iothread_period,
                               vm->def->cputune.iothread_quota,
5539
                               &iothread->sched);
5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558
}


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;
}


5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570
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;

5571
    if (!(ordermap = virBitmapNew(maxvcpus + 1)))
5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587
        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) {
5588
            if (virBitmapIsBitSet(ordermap, vcpu->order)) {
5589
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
5590
                               _("duplicate vcpu order '%u'"), vcpu->order);
5591 5592 5593
                goto cleanup;
            }

5594 5595 5596 5597 5598 5599
            if (virBitmapSetBit(ordermap, vcpu->order)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("vcpu order '%u' exceeds vcpu count"),
                               vcpu->order);
                goto cleanup;
            }
5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615
        }

        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 &&
5616
                 vcpupriv->thread_id == -1 && vcpupriv->node_id == -1) ||
5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 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 5661 5662 5663 5664 5665 5666
                !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);
5667 5668
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuCgroupEmulatorAllNodesDataPtr emulatorCgroup = NULL;
5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700
    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);

5701 5702 5703
    if (qemuCgroupEmulatorAllNodesAllow(priv->cgroup, &emulatorCgroup) < 0)
        goto cleanup;

5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727
    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:
5728
    qemuCgroupEmulatorAllNodesRestore(emulatorCgroup);
5729 5730 5731 5732 5733 5734
    VIR_FREE(bootHotplug);
    virJSONValueFree(vcpuprops);
    return ret;
}


5735 5736 5737 5738 5739 5740 5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773
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;
    }

5774 5775 5776
    if (virCPUConvertLegacy(caps->host.arch, def->cpu) < 0)
        return -1;

5777 5778 5779 5780
    /* nothing to update for host-passthrough */
    if (def->cpu->mode == VIR_CPU_MODE_HOST_PASSTHROUGH)
        return 0;

5781 5782
    if (def->cpu->check == VIR_CPU_CHECK_PARTIAL &&
        virCPUCompare(caps->host.arch,
5783
                      virQEMUCapsGetHostModel(qemuCaps, def->virtType,
5784
                                              VIR_QEMU_CAPS_HOST_CPU_FULL),
5785 5786
                      def->cpu, true) < 0)
        return -1;
5787 5788

    if (virCPUUpdate(def->os.arch, def->cpu,
5789
                     virQEMUCapsGetHostModel(qemuCaps, def->virtType,
5790
                                             VIR_QEMU_CAPS_HOST_CPU_MIGRATABLE)) < 0)
5791 5792
        goto cleanup;

5793 5794
    if (virCPUTranslate(def->os.arch, def->cpu,
                        virQEMUCapsGetCPUDefinitions(qemuCaps, def->virtType)) < 0)
5795 5796 5797 5798 5799 5800 5801 5802 5803 5804
        goto cleanup;

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

 cleanup:
    return ret;
}


5805 5806 5807 5808 5809 5810
static int
qemuProcessPrepareDomainNUMAPlacement(virDomainObjPtr vm,
                                      virCapsPtr caps)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char *nodeset = NULL;
5811 5812
    virBitmapPtr numadNodeset = NULL;
    virBitmapPtr hostMemoryNodeset = NULL;
5813 5814 5815 5816 5817 5818 5819 5820 5821 5822 5823 5824 5825 5826
    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;

5827 5828 5829
    if (!(hostMemoryNodeset = virNumaGetHostMemoryNodeset()))
        goto cleanup;

5830 5831
    VIR_DEBUG("Nodeset returned from numad: %s", nodeset);

5832
    if (virBitmapParse(nodeset, &numadNodeset, VIR_DOMAIN_CPUMASK_LEN) < 0)
5833 5834
        goto cleanup;

5835 5836 5837 5838
    /* 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)))
5839 5840
        goto cleanup;

5841 5842 5843 5844
    virBitmapIntersect(numadNodeset, hostMemoryNodeset);

    VIR_STEAL_PTR(priv->autoNodeset, numadNodeset);

5845 5846 5847 5848
    ret = 0;

 cleanup:
    VIR_FREE(nodeset);
5849 5850
    virBitmapFree(numadNodeset);
    virBitmapFree(hostMemoryNodeset);
5851 5852 5853 5854
    return ret;
}


5855
static int
5856
qemuProcessPrepareDomainStorage(virQEMUDriverPtr driver,
5857
                                virDomainObjPtr vm,
5858
                                qemuDomainObjPrivatePtr priv,
5859
                                virQEMUDriverConfigPtr cfg,
5860 5861 5862 5863 5864 5865 5866 5867 5868
                                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];

5869
        if (virDomainDiskTranslateSourcePool(disk) < 0) {
5870 5871 5872 5873 5874 5875
            if (qemuDomainCheckDiskStartupPolicy(driver, vm, idx, cold_boot) < 0)
                return -1;

            /* disk source was dropped */
            continue;
        }
5876

5877
        if (qemuDomainPrepareDiskSource(disk, priv, cfg) < 0)
5878
            return -1;
5879 5880 5881 5882 5883 5884
    }

    return 0;
}


5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904
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;
    }
}


5905
/**
5906 5907 5908 5909
 * qemuProcessPrepareDomain:
 * @driver: qemu driver
 * @vm: domain object
 * @flags: qemuProcessStartFlags
5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920
 *
 * 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
5921
qemuProcessPrepareDomain(virQEMUDriverPtr driver,
5922 5923 5924 5925 5926 5927
                         virDomainObjPtr vm,
                         unsigned int flags)
{
    int ret = -1;
    size_t i;
    qemuDomainObjPrivatePtr priv = vm->privateData;
5928
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5929 5930 5931 5932 5933
    virCapsPtr caps;

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

5934 5935 5936 5937
    priv->machineName = qemuDomainGetMachineName(vm);
    if (!priv->machineName)
        goto cleanup;

5938 5939 5940 5941
    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)");
5942
        if (qemuSecurityGenLabel(driver->securityManager, vm->def) < 0) {
5943 5944 5945 5946 5947
            virDomainAuditSecurityLabel(vm, false);
            goto cleanup;
        }
        virDomainAuditSecurityLabel(vm, true);

5948 5949
        if (qemuProcessPrepareDomainNUMAPlacement(vm, caps) < 0)
            goto cleanup;
5950 5951
    }

5952 5953 5954 5955 5956 5957 5958
    /* 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;
    }

5959 5960 5961
    /* Track if this domain remembers original owner */
    priv->rememberOwner = cfg->rememberOwner;

5962 5963
    qemuProcessPrepareAllowReboot(vm);

5964 5965 5966 5967 5968 5969 5970 5971 5972
    /* 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;
        }
    }

5973 5974 5975 5976 5977 5978 5979
    /*
     * 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
     */
5980
    VIR_DEBUG("Assigning domain PCI addresses");
5981 5982
    if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, driver, vm,
                                   !!(flags & VIR_QEMU_PROCESS_START_NEW))) < 0) {
5983
        goto cleanup;
5984
    }
5985

5986 5987 5988
    if (qemuAssignDeviceAliases(vm->def, priv->qemuCaps) < 0)
        goto cleanup;

5989
    VIR_DEBUG("Setting graphics devices");
5990
    if (qemuProcessSetupGraphics(driver, vm, priv->qemuCaps, flags) < 0)
5991 5992
        goto cleanup;

5993 5994 5995 5996
    VIR_DEBUG("Create domain masterKey");
    if (qemuDomainMasterKeyCreate(vm) < 0)
        goto cleanup;

5997
    VIR_DEBUG("Setting up storage");
5998
    if (qemuProcessPrepareDomainStorage(driver, vm, priv, cfg, flags) < 0)
5999 6000
        goto cleanup;

6001
    VIR_DEBUG("Prepare chardev source backends for TLS");
6002
    qemuDomainPrepareChardevSource(vm->def, cfg);
6003

6004
    VIR_DEBUG("Add secrets to hostdevs and chardevs");
6005
    if (qemuDomainSecretPrepare(driver, vm) < 0)
6006 6007
        goto cleanup;

6008 6009 6010 6011 6012 6013
    for (i = 0; i < vm->def->nchannels; i++) {
        if (qemuDomainPrepareChannel(vm->def->channels[i],
                                     priv->channelTargetDir) < 0)
            goto cleanup;
    }

6014
    if (!(priv->monConfig = virDomainChrSourceDefNew(driver->xmlopt)))
6015 6016 6017 6018 6019 6020
        goto cleanup;

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

6021
    priv->monJSON = true;
6022 6023
    priv->monError = false;
    priv->monStart = 0;
6024
    priv->runningReason = VIR_DOMAIN_RUNNING_UNKNOWN;
6025

6026 6027 6028 6029
    VIR_DEBUG("Updating guest CPU definition");
    if (qemuProcessUpdateGuestCPU(vm->def, priv->qemuCaps, caps, flags) < 0)
        goto cleanup;

6030 6031 6032 6033 6034
    for (i = 0; i < vm->def->nshmems; i++) {
        if (qemuDomainPrepareShmemChardev(vm->def->shmems[i]) < 0)
            goto cleanup;
    }

6035 6036 6037
    ret = 0;
 cleanup:
    virObjectUnref(caps);
6038
    virObjectUnref(cfg);
6039 6040 6041 6042
    return ret;
}


6043
static int
6044 6045 6046
qemuProcessSEVCreateFile(const char *configDir,
                         const char *name,
                         const char *data)
6047 6048
{
    char *configFile;
6049
    int ret = -1;
6050 6051 6052 6053 6054 6055 6056

    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);
6057
        goto cleanup;
6058 6059
    }

6060
    ret = 0;
6061
 cleanup:
6062
    VIR_FREE(configFile);
6063
    return ret;
6064 6065 6066 6067
}


static int
J
Ján Tomko 已提交
6068
qemuProcessPrepareSEVGuestInput(virDomainObjPtr vm)
6069 6070 6071 6072
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
    virQEMUCapsPtr qemuCaps = priv->qemuCaps;
6073
    virDomainSEVDefPtr sev = def->sev;
6074 6075 6076 6077

    if (!sev)
        return 0;

6078
    VIR_DEBUG("Preparing SEV guest");
6079 6080 6081 6082 6083 6084 6085 6086 6087

    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) {
6088
        if (qemuProcessSEVCreateFile(priv->libDir, "dh_cert", sev->dh_cert) < 0)
6089 6090 6091 6092
            return -1;
    }

    if (sev->session) {
6093
        if (qemuProcessSEVCreateFile(priv->libDir, "session", sev->session) < 0)
6094 6095 6096 6097 6098 6099 6100
            return -1;
    }

    return 0;
}


6101 6102 6103 6104 6105
static int
qemuProcessPrepareHostStorage(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              unsigned int flags)
{
6106
    qemuDomainObjPrivatePtr priv = vm->privateData;
6107 6108
    size_t i;
    bool cold_boot = flags & VIR_QEMU_PROCESS_START_COLD;
6109
    bool blockdev = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV);
6110 6111 6112 6113 6114 6115 6116 6117

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

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

6118 6119 6120
        /* backing chain needs to be redetected if we aren't using blockdev */
        if (!blockdev)
            virStorageSourceBackingStoreClear(disk->src);
6121 6122

        if (qemuDomainDetermineDiskChain(driver, vm, disk, true) >= 0)
6123 6124 6125 6126 6127 6128 6129 6130 6131 6132 6133 6134
            continue;

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

        return -1;
    }

    return 0;
}


J
Ján Tomko 已提交
6135
int
6136 6137 6138 6139 6140 6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163 6164
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
6165
/**
6166 6167 6168
 * qemuProcessPrepareHost:
 * @driver: qemu driver
 * @vm: domain object
6169
 * @flags: qemuProcessStartFlags
J
Jiri Denemark 已提交
6170
 *
6171 6172 6173
 * 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
6174
 *
6175
 * TODO: move all host modification from qemuBuildCommandLine into this function
J
Jiri Denemark 已提交
6176 6177
 */
int
6178 6179
qemuProcessPrepareHost(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
6180
                       unsigned int flags)
6181
{
6182
    int ret = -1;
6183
    unsigned int hostdev_flags = 0;
6184 6185
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
6186

6187
    if (qemuPrepareNVRAM(cfg, vm) < 0)
J
Jiri Denemark 已提交
6188
        goto cleanup;
6189

6190 6191 6192 6193
    if (vm->def->vsock) {
        if (qemuProcessOpenVhostVsock(vm->def->vsock) < 0)
            goto cleanup;
    }
6194 6195 6196 6197 6198
    /* 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");
6199
    if (qemuProcessNetworkPrepareDevices(vm->def) < 0)
J
Jiri Denemark 已提交
6200
        goto cleanup;
6201

6202
    /* Must be run before security labelling */
6203
    VIR_DEBUG("Preparing host devices");
6204 6205
    if (!cfg->relaxedACS)
        hostdev_flags |= VIR_HOSTDEV_STRICT_ACS_CHECK;
6206
    if (flags & VIR_QEMU_PROCESS_START_NEW)
6207
        hostdev_flags |= VIR_HOSTDEV_COLD_BOOT;
6208 6209
    if (qemuHostdevPrepareDomainDevices(driver, vm->def, priv->qemuCaps,
                                        hostdev_flags) < 0)
J
Jiri Denemark 已提交
6210
        goto cleanup;
6211

6212
    VIR_DEBUG("Preparing chr devices");
6213 6214 6215 6216
    if (virDomainChrDefForeach(vm->def,
                               true,
                               qemuProcessPrepareChardevDevice,
                               NULL) < 0)
J
Jiri Denemark 已提交
6217
        goto cleanup;
6218

6219
    if (qemuProcessBuildDestroyMemoryPaths(driver, vm, NULL, true) < 0)
6220
        goto cleanup;
6221

6222 6223
    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
6224
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
6225
    qemuRemoveCgroup(vm);
6226

6227
    if (virFileMakePath(cfg->logDir) < 0) {
6228 6229
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
6230
                             cfg->logDir);
J
Jiri Denemark 已提交
6231
        goto cleanup;
6232 6233
    }

6234 6235 6236 6237 6238 6239 6240 6241 6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256
    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;

6257 6258
    VIR_DEBUG("Write domain masterKey");
    if (qemuDomainWriteMasterKeyFile(driver, vm) < 0)
J
John Ferlan 已提交
6259 6260
        goto cleanup;

6261 6262 6263 6264
    VIR_DEBUG("Preparing disks (host)");
    if (qemuProcessPrepareHostStorage(driver, vm, flags) < 0)
        goto cleanup;

6265 6266 6267 6268
    VIR_DEBUG("Preparing external devices");
    if (qemuExtDevicesPrepareHost(driver, vm->def) < 0)
        goto cleanup;

J
Ján Tomko 已提交
6269
    if (qemuProcessPrepareSEVGuestInput(vm) < 0)
6270 6271
        goto cleanup;

6272 6273 6274 6275 6276 6277 6278
    ret = 0;
 cleanup:
    virObjectUnref(cfg);
    return ret;
}


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
/**
 * 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;
}


6312 6313 6314 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 6349 6350 6351 6352 6353 6354 6355 6356 6357
/**
 * 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;
}


6358 6359 6360 6361 6362 6363 6364 6365 6366 6367 6368 6369 6370 6371 6372 6373 6374 6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391
/**
 * 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
6392
    size_t i;
6393

6394
    VIR_DEBUG("conn=%p driver=%p vm=%p name=%s if=%d asyncJob=%d "
6395 6396 6397
              "incoming.launchURI=%s incoming.deferredURI=%s "
              "incoming.fd=%d incoming.path=%s "
              "snapshot=%p vmop=%d flags=0x%x",
6398
              conn, driver, vm, vm->def->name, vm->def->id, asyncJob,
6399 6400 6401 6402 6403 6404 6405 6406 6407 6408
              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 |
6409
                  VIR_QEMU_PROCESS_START_AUTODESTROY |
6410 6411
                  VIR_QEMU_PROCESS_START_NEW |
                  VIR_QEMU_PROCESS_START_GEN_VMID, -1);
6412 6413 6414

    cfg = virQEMUDriverGetConfig(driver);

6415 6416 6417 6418 6419 6420
    if ((flags & VIR_QEMU_PROCESS_START_AUTODESTROY) && !conn) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Domain autodestroy requires a connection handle"));
        return -1;
    }

6421 6422 6423 6424 6425 6426 6427 6428
    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;

6429
    VIR_DEBUG("Creating domain log file");
6430 6431
    if (!(logCtxt = qemuDomainLogContextNew(driver, vm,
                                            QEMU_DOMAIN_LOG_CONTEXT_MODE_START)))
J
Jiri Denemark 已提交
6432
        goto cleanup;
6433
    logfile = qemuDomainLogContextGetWriteFD(logCtxt);
6434

6435 6436 6437
    if (qemuProcessGenID(vm, flags) < 0)
        goto cleanup;

6438
    if (qemuExtDevicesStart(driver, vm, logCtxt) < 0)
6439 6440
        goto cleanup;

6441
    VIR_DEBUG("Building emulator command line");
6442
    if (!(cmd = qemuBuildCommandLine(driver,
6443
                                     qemuDomainLogContextGetManager(logCtxt),
6444
                                     driver->securityManager,
6445
                                     vm,
6446 6447
                                     incoming ? incoming->launchURI : NULL,
                                     snapshot, vmop,
J
Ján Tomko 已提交
6448
                                     false,
6449
                                     qemuCheckFips(),
6450
                                     &nnicindexes, &nicindexes)))
J
Jiri Denemark 已提交
6451
        goto cleanup;
6452

6453 6454
    if (incoming && incoming->fd != -1)
        virCommandPassFD(cmd, incoming->fd, 0);
6455

6456
    /* now that we know it is about to start call the hook if present */
6457 6458 6459
    if (qemuProcessStartHook(driver, vm,
                             VIR_HOOK_QEMU_OP_START,
                             VIR_HOOK_SUBOP_BEGIN) < 0)
J
Jiri Denemark 已提交
6460
        goto cleanup;
6461

6462
    qemuLogOperation(vm, "starting up", cmd, logCtxt);
6463

6464
    qemuDomainObjCheckTaint(driver, vm, logCtxt);
6465

6466
    qemuDomainLogContextMarkPosition(logCtxt);
6467

6468 6469 6470 6471 6472
    VIR_DEBUG("Building mount namespace");

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

6473
    VIR_DEBUG("Clear emulator capabilities: %d",
6474 6475
              cfg->clearEmulatorCapabilities);
    if (cfg->clearEmulatorCapabilities)
6476 6477
        virCommandClearCaps(cmd);

6478 6479
    VIR_DEBUG("Setting up raw IO");
    if (qemuProcessSetupRawIO(driver, vm, cmd) < 0)
J
Jiri Denemark 已提交
6480
        goto cleanup;
6481

6482
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);
6483 6484
    virCommandSetMaxProcesses(cmd, cfg->maxProcesses);
    virCommandSetMaxFiles(cmd, cfg->maxFiles);
6485
    virCommandSetMaxCoreSize(cmd, cfg->maxCore);
6486
    virCommandSetUmask(cmd, 0x002);
6487

6488
    VIR_DEBUG("Setting up security labelling");
6489 6490
    if (qemuSecuritySetChildProcessLabel(driver->securityManager,
                                         vm->def, cmd) < 0)
J
Jiri Denemark 已提交
6491
        goto cleanup;
6492

6493 6494 6495
    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
6496
    virCommandSetPidFile(cmd, priv->pidfile);
6497
    virCommandDaemonize(cmd);
6498
    virCommandRequireHandshake(cmd);
6499

6500
    if (qemuSecurityPreFork(driver->securityManager) < 0)
J
Jiri Denemark 已提交
6501
        goto cleanup;
6502
    rv = virCommandRun(cmd, NULL);
6503
    qemuSecurityPostFork(driver->securityManager);
6504

E
Eric Blake 已提交
6505
    /* wait for qemu process to show up */
6506
    if (rv == 0) {
6507
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
6508 6509
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
6510
            rv = -1;
6511
        }
M
Michal Privoznik 已提交
6512
        VIR_DEBUG("QEMU vm=%p name=%s running with pid=%lld",
6513
                  vm, vm->def->name, (long long)vm->pid);
6514 6515 6516
    } else {
        VIR_DEBUG("QEMU vm=%p name=%s failed to spawn",
                  vm, vm->def->name);
6517 6518
    }

6519
    VIR_DEBUG("Writing early domain status to disk");
6520
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
J
Jiri Denemark 已提交
6521
        goto cleanup;
6522

6523 6524
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
6525
        /* Read errors from child that occurred between fork and exec. */
6526 6527
        qemuProcessReportLogError(logCtxt,
                                  _("Process exited prior to exec"));
J
Jiri Denemark 已提交
6528
        goto cleanup;
6529 6530
    }

6531
    VIR_DEBUG("Setting up domain cgroup (if required)");
6532
    if (qemuSetupCgroup(vm, nnicindexes, nicindexes) < 0)
J
Jiri Denemark 已提交
6533
        goto cleanup;
6534

6535 6536 6537 6538
    if (!(priv->perf = virPerfNew()))
        goto cleanup;

    for (i = 0; i < VIR_PERF_EVENT_LAST; i++) {
6539
        if (vm->def->perf.events[i] == VIR_TRISTATE_BOOL_YES &&
6540 6541
            virPerfEventEnable(priv->perf, i, vm->pid) < 0)
            goto cleanup;
Q
Qiaowei Ren 已提交
6542
    }
6543

6544 6545 6546 6547 6548 6549
    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
    if (!vm->def->cputune.emulatorpin &&
        qemuProcessInitCpuAffinity(vm) < 0)
        goto cleanup;

6550 6551 6552 6553
    VIR_DEBUG("Setting emulator tuning/settings");
    if (qemuProcessSetupEmulator(vm) < 0)
        goto cleanup;

6554 6555 6556 6557
    VIR_DEBUG("Setting cgroup for external devices (if required)");
    if (qemuSetupCgroupForExtDevices(vm, driver) < 0)
        goto cleanup;

6558
    VIR_DEBUG("Setting up resctrl");
6559 6560 6561
    if (qemuProcessResctrlCreate(driver, vm) < 0)
        goto cleanup;

6562
    VIR_DEBUG("Setting up managed PR daemon");
6563 6564
    if (virDomainDefHasManagedPR(vm->def) &&
        qemuProcessStartManagedPRDaemon(vm) < 0)
6565 6566
        goto cleanup;

6567
    VIR_DEBUG("Setting domain security labels");
6568 6569 6570 6571
    if (qemuSecuritySetAllLabel(driver,
                                vm,
                                incoming ? incoming->path : NULL) < 0)
        goto cleanup;
6572

6573
    /* Security manager labeled all devices, therefore
J
Jiri Denemark 已提交
6574 6575 6576 6577
     * if any operation from now on fails, we need to ask the caller to
     * restore labels.
     */
    ret = -2;
6578

J
Jiri Denemark 已提交
6579
    if (incoming && incoming->fd != -1) {
6580 6581 6582 6583 6584 6585 6586
        /* 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
6587
        if (fstat(incoming->fd, &stdin_sb) < 0) {
6588
            virReportSystemError(errno,
J
Jiri Denemark 已提交
6589 6590
                                 _("cannot stat fd %d"), incoming->fd);
            goto cleanup;
6591 6592
        }
        if (S_ISFIFO(stdin_sb.st_mode) &&
6593 6594
            qemuSecuritySetImageFDLabel(driver->securityManager,
                                        vm->def, incoming->fd) < 0)
J
Jiri Denemark 已提交
6595
            goto cleanup;
6596 6597 6598
    }

    VIR_DEBUG("Labelling done, completing handshake to child");
6599
    if (virCommandHandshakeNotify(cmd) < 0)
J
Jiri Denemark 已提交
6600
        goto cleanup;
6601 6602
    VIR_DEBUG("Handshake complete, child running");

6603
    if (rv == -1) /* The VM failed to start; tear filters before taps */
6604 6605
        virDomainConfVMNWFilterTeardown(vm);

6606
    if (rv == -1) /* The VM failed to start */
J
Jiri Denemark 已提交
6607
        goto cleanup;
6608

6609
    VIR_DEBUG("Waiting for monitor to show up");
6610
    if (qemuProcessWaitForMonitor(driver, vm, asyncJob, logCtxt) < 0)
J
Jiri Denemark 已提交
6611
        goto cleanup;
6612

6613 6614
    if (qemuConnectAgent(driver, vm) < 0)
        goto cleanup;
D
Daniel P. Berrange 已提交
6615

6616
    VIR_DEBUG("Verifying and updating provided guest CPU");
6617
    if (qemuProcessUpdateAndVerifyCPU(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6618
        goto cleanup;
6619

6620
    VIR_DEBUG("Setting up post-init cgroup restrictions");
6621
    if (qemuSetupCpusetMems(vm) < 0)
J
Jiri Denemark 已提交
6622
        goto cleanup;
6623

6624 6625 6626 6627 6628 6629 6630 6631 6632 6633 6634 6635
    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;
    }

6636
    VIR_DEBUG("Refreshing VCPU info");
6637
    if (qemuDomainRefreshVcpuInfo(driver, vm, asyncJob, false) < 0)
J
Jiri Denemark 已提交
6638
        goto cleanup;
6639

6640 6641 6642
    if (qemuDomainValidateVcpuInfo(vm) < 0)
        goto cleanup;

6643 6644
    qemuDomainVcpuPersistOrder(vm->def);

6645 6646
    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6647
        goto cleanup;
6648

6649 6650 6651 6652
    VIR_DEBUG("Setting global CPU cgroup (if required)");
    if (qemuSetupGlobalCpuCgroup(vm) < 0)
        goto cleanup;

6653 6654
    VIR_DEBUG("Setting vCPU tuning/settings");
    if (qemuProcessSetupVcpus(vm) < 0)
J
Jiri Denemark 已提交
6655
        goto cleanup;
6656

6657 6658
    VIR_DEBUG("Setting IOThread tuning/settings");
    if (qemuProcessSetupIOThreads(vm) < 0)
J
Jiri Denemark 已提交
6659
        goto cleanup;
6660

6661
    VIR_DEBUG("Setting any required VM passwords");
6662
    if (qemuProcessInitPasswords(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6663
        goto cleanup;
6664

6665 6666 6667 6668
    /* set default link states */
    /* qemu doesn't support setting this on the command line, so
     * enter the monitor */
    VIR_DEBUG("Setting network link states");
6669
    if (qemuProcessSetLinkStates(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6670
        goto cleanup;
6671

6672
    VIR_DEBUG("Setting initial memory amount");
6673
    if (qemuProcessSetupBalloon(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6674
        goto cleanup;
6675

6676 6677 6678
    if (qemuProcessSetupDiskThrottlingBlockdev(driver, vm, asyncJob) < 0)
        goto cleanup;

6679
    /* Since CPUs were not started yet, the balloon could not return the memory
6680 6681
     * 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 */
6682
    if (!incoming && !snapshot &&
6683
        qemuProcessRefreshBalloonState(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6684
        goto cleanup;
6685

J
Jiri Denemark 已提交
6686 6687 6688 6689 6690 6691 6692
    if (flags & VIR_QEMU_PROCESS_START_AUTODESTROY &&
        qemuProcessAutoDestroyAdd(driver, vm, conn) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
6693
    if (ret < 0)
6694
        qemuExtDevicesStop(driver, vm);
6695
    qemuDomainSecretDestroy(vm);
J
Jiri Denemark 已提交
6696
    virCommandFree(cmd);
6697
    virObjectUnref(logCtxt);
J
Jiri Denemark 已提交
6698 6699 6700 6701 6702 6703 6704
    virObjectUnref(cfg);
    virObjectUnref(caps);
    VIR_FREE(nicindexes);
    return ret;
}


6705 6706 6707 6708 6709 6710 6711 6712 6713 6714
/**
 * 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.
 */
6715
int
6716 6717 6718 6719 6720 6721 6722 6723 6724 6725 6726 6727 6728 6729 6730 6731 6732 6733 6734 6735 6736 6737 6738 6739
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;
}


6740 6741 6742 6743 6744 6745
/**
 * qemuProcessFinishStartup:
 *
 * Finish starting a new domain.
 */
int
6746
qemuProcessFinishStartup(virQEMUDriverPtr driver,
6747 6748 6749 6750 6751 6752 6753 6754 6755 6756
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob,
                         bool startCPUs,
                         virDomainPausedReason pausedReason)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;

    if (startCPUs) {
        VIR_DEBUG("Starting domain CPUs");
6757
        if (qemuProcessStartCPUs(driver, vm,
6758 6759
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 asyncJob) < 0) {
6760
            if (virGetLastErrorCode() == VIR_ERR_OK)
6761 6762 6763 6764 6765 6766 6767 6768 6769
                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");
6770
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
6771 6772 6773 6774 6775 6776 6777 6778 6779 6780 6781 6782 6783 6784 6785
        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
6786 6787 6788 6789
int
qemuProcessStart(virConnectPtr conn,
                 virQEMUDriverPtr driver,
                 virDomainObjPtr vm,
6790
                 virCPUDefPtr updatedCPU,
J
Jiri Denemark 已提交
6791 6792 6793 6794 6795 6796 6797 6798 6799 6800 6801 6802 6803 6804 6805 6806 6807 6808 6809 6810 6811 6812 6813 6814 6815
                 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 |
6816 6817
                      VIR_QEMU_PROCESS_START_AUTODESTROY |
                      VIR_QEMU_PROCESS_START_GEN_VMID, cleanup);
J
Jiri Denemark 已提交
6818

6819 6820 6821
    if (!migrateFrom && !snapshot)
        flags |= VIR_QEMU_PROCESS_START_NEW;

6822 6823
    if (qemuProcessInit(driver, vm, updatedCPU,
                        asyncJob, !!migrateFrom, flags) < 0)
J
Jiri Denemark 已提交
6824 6825 6826
        goto cleanup;

    if (migrateFrom) {
6827
        incoming = qemuProcessIncomingDefNew(priv->qemuCaps, NULL, migrateFrom,
J
Jiri Denemark 已提交
6828 6829 6830 6831 6832
                                             migrateFd, migratePath);
        if (!incoming)
            goto stop;
    }

6833
    if (qemuProcessPrepareDomain(driver, vm, flags) < 0)
6834 6835
        goto stop;

6836
    if (qemuProcessPrepareHost(driver, vm, flags) < 0)
6837 6838
        goto stop;

J
Jiri Denemark 已提交
6839 6840
    if ((rv = qemuProcessLaunch(conn, driver, vm, asyncJob, incoming,
                                snapshot, vmop, flags)) < 0) {
6841
        if (rv == -2)
J
Jiri Denemark 已提交
6842 6843 6844 6845
            relabel = true;
        goto stop;
    }
    relabel = true;
6846

6847 6848
    if (incoming &&
        incoming->deferredURI &&
6849
        qemuMigrationDstRun(driver, vm, incoming->deferredURI, asyncJob) < 0)
J
Jiri Denemark 已提交
6850
        goto stop;
6851

6852
    if (qemuProcessFinishStartup(driver, vm, asyncJob,
6853 6854 6855 6856
                                 !(flags & VIR_QEMU_PROCESS_START_PAUSED),
                                 incoming ?
                                 VIR_DOMAIN_PAUSED_MIGRATION :
                                 VIR_DOMAIN_PAUSED_USER) < 0)
J
Jiri Denemark 已提交
6857
        goto stop;
6858

6859 6860 6861
    if (!incoming) {
        /* Keep watching qemu log for errors during incoming migration, otherwise
         * unset reporting errors from qemu log. */
6862
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
6863

6864 6865 6866 6867 6868 6869
        /* 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;
    }

6870 6871 6872
    ret = 0;

 cleanup:
6873
    qemuProcessIncomingDefFree(incoming);
6874
    return ret;
6875

J
Jiri Denemark 已提交
6876 6877 6878 6879 6880 6881
 stop:
    stopFlags = 0;
    if (!relabel)
        stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
    if (migrateFrom)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
6882
    if (priv->mon)
6883
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
6884
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, asyncJob, stopFlags);
6885
    goto cleanup;
6886 6887 6888
}


6889
virCommandPtr
6890
qemuProcessCreatePretendCmd(virQEMUDriverPtr driver,
6891 6892
                            virDomainObjPtr vm,
                            const char *migrateURI,
6893
                            bool enableFips,
6894 6895 6896 6897 6898 6899 6900 6901 6902 6903
                            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;
6904
    flags |= VIR_QEMU_PROCESS_START_NEW;
6905 6906
    if (standalone)
        flags |= VIR_QEMU_PROCESS_START_STANDALONE;
6907

6908 6909
    if (qemuProcessInit(driver, vm, NULL, QEMU_ASYNC_JOB_NONE,
                        !!migrateURI, flags) < 0)
6910 6911
        goto cleanup;

6912
    if (qemuProcessPrepareDomain(driver, vm, flags) < 0)
6913 6914 6915
        goto cleanup;

    VIR_DEBUG("Building emulator command line");
6916
    cmd = qemuBuildCommandLine(driver,
6917
                               NULL,
6918
                               driver->securityManager,
6919
                               vm,
6920 6921 6922 6923
                               migrateURI,
                               NULL,
                               VIR_NETDEV_VPORT_PROFILE_OP_NO_OP,
                               standalone,
6924
                               enableFips,
6925
                               NULL,
6926
                               NULL);
6927 6928 6929 6930 6931 6932

 cleanup:
    return cmd;
}


6933
int
6934
qemuProcessKill(virDomainObjPtr vm, unsigned int flags)
6935
{
6936
    int ret;
6937

6938
    VIR_DEBUG("vm=%p name=%s pid=%lld flags=0x%x",
6939
              vm, vm->def->name,
6940
              (long long)vm->pid, flags);
6941

6942 6943 6944 6945 6946
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
6947 6948
    }

6949
    if (flags & VIR_QEMU_PROCESS_KILL_NOWAIT) {
6950 6951 6952 6953 6954
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
6955

6956 6957 6958 6959 6960
    /* 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);
6961

6962
    return ret;
6963 6964 6965
}


6966 6967 6968 6969 6970 6971 6972 6973 6974 6975 6976 6977 6978 6979 6980 6981 6982 6983 6984 6985 6986 6987 6988 6989 6990 6991 6992 6993 6994 6995 6996 6997 6998 6999 7000 7001 7002 7003 7004
/**
 * 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;
}


7005
void qemuProcessStop(virQEMUDriverPtr driver,
7006
                     virDomainObjPtr vm,
7007
                     virDomainShutoffReason reason,
7008
                     qemuDomainAsyncJob asyncJob,
7009
                     unsigned int flags)
7010 7011 7012 7013 7014 7015
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
7016
    virNetDevVPortProfilePtr vport = NULL;
7017
    size_t i;
7018
    char *timestamp;
7019
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
7020

M
Michal Privoznik 已提交
7021
    VIR_DEBUG("Shutting down vm=%p name=%s id=%d pid=%lld, "
7022
              "reason=%s, asyncJob=%s, flags=0x%x",
7023
              vm, vm->def->name, vm->def->id,
7024
              (long long)vm->pid,
7025 7026 7027
              virDomainShutoffReasonTypeToString(reason),
              qemuDomainAsyncJobTypeToString(asyncJob),
              flags);
7028

7029 7030 7031 7032
    /* This method is routinely used in clean up paths. Disable error
     * reporting so we don't squash a legit error. */
    orig_err = virSaveLastError();

7033 7034 7035 7036 7037 7038 7039 7040 7041 7042 7043 7044 7045 7046 7047
    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;
    }

7048
    qemuProcessBuildDestroyMemoryPaths(driver, vm, NULL, false);
7049

7050 7051
    vm->def->id = -1;

7052
    if (virAtomicIntDecAndTest(&driver->nactive) && driver->inhibitCallback)
7053 7054
        driver->inhibitCallback(false, driver->inhibitOpaque);

7055 7056
    /* Wake up anything waiting on domain condition */
    virDomainObjBroadcast(vm);
7057

7058
    if ((timestamp = virTimeStringNow()) != NULL) {
7059 7060 7061
        qemuDomainLogAppendMessage(driver, vm, "%s: shutting down, reason=%s\n",
                                   timestamp,
                                   virDomainShutoffReasonTypeToString(reason));
7062
        VIR_FREE(timestamp);
7063 7064
    }

7065 7066 7067
    /* Clear network bandwidth */
    virDomainClearNetBandwidth(vm);

7068 7069
    virDomainConfVMNWFilterTeardown(vm);

7070
    if (cfg->macFilter) {
7071
        def = vm->def;
7072
        for (i = 0; i < def->nnets; i++) {
7073 7074 7075
            virDomainNetDefPtr net = def->nets[i];
            if (net->ifname == NULL)
                continue;
7076 7077 7078
            ignore_value(ebtablesRemoveForwardAllowIn(driver->ebtables,
                                                      net->ifname,
                                                      &net->mac));
7079 7080 7081
        }
    }

7082
    virPortAllocatorRelease(priv->nbdPort);
7083
    priv->nbdPort = 0;
7084

D
Daniel P. Berrange 已提交
7085 7086 7087 7088
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
    }
7089
    priv->agentError = false;
D
Daniel P. Berrange 已提交
7090

7091
    if (priv->mon) {
7092
        qemuMonitorClose(priv->mon);
7093 7094
        priv->mon = NULL;
    }
7095 7096 7097 7098 7099 7100 7101 7102

    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
7103 7104 7105
    /* Remove the master key */
    qemuDomainMasterKeyRemove(priv);

7106
    /* Do this before we delete the tree and remove pidfile. */
7107
    qemuProcessKillManagedPRDaemon(vm);
7108

7109 7110
    virFileDeleteTree(priv->libDir);
    virFileDeleteTree(priv->channelTargetDir);
7111

7112 7113 7114 7115 7116 7117
    ignore_value(virDomainChrDefForeach(vm->def,
                                        false,
                                        qemuProcessCleanupChardevDevice,
                                        NULL));


7118
    /* shut it off for sure */
7119 7120 7121
    ignore_value(qemuProcessKill(vm,
                                 VIR_QEMU_PROCESS_KILL_FORCE|
                                 VIR_QEMU_PROCESS_KILL_NOCHECK));
7122

7123 7124
    qemuDomainCleanupRun(driver, vm);

7125
    qemuExtDevicesStop(driver, vm);
7126

7127
    /* Stop autodestroy in case guest is restarted */
7128
    qemuProcessAutoDestroyRemove(driver, vm);
7129

7130 7131
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
7132
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
7133 7134

        /* we can't stop the operation even if the script raised an error */
7135 7136 7137
        ignore_value(virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
                                 VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                                 NULL, xml, NULL));
7138 7139 7140
        VIR_FREE(xml);
    }

7141 7142
    /* Reset Security Labels unless caller don't want us to */
    if (!(flags & VIR_QEMU_PROCESS_STOP_NO_RELABEL))
7143 7144 7145
        qemuSecurityRestoreAllLabel(driver, vm,
                                    !!(flags & VIR_QEMU_PROCESS_STOP_MIGRATED));

7146
    qemuSecurityReleaseLabel(driver->securityManager, vm->def);
7147

7148
    for (i = 0; i < vm->def->ndisks; i++) {
7149
        virDomainDeviceDef dev;
7150
        virDomainDiskDefPtr disk = vm->def->disks[i];
7151 7152 7153 7154

        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        ignore_value(qemuRemoveSharedDevice(driver, &dev, vm->def->name));
7155 7156
    }

7157
    /* Clear out dynamically assigned labels */
7158
    for (i = 0; i < vm->def->nseclabels; i++) {
7159
        if (vm->def->seclabels[i]->type == VIR_DOMAIN_SECLABEL_DYNAMIC)
7160 7161
            VIR_FREE(vm->def->seclabels[i]->label);
        VIR_FREE(vm->def->seclabels[i]->imagelabel);
7162 7163
    }

7164
    qemuHostdevReAttachDomainDevices(driver, vm->def);
7165 7166 7167 7168

    def = vm->def;
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
7169 7170 7171 7172
        vport = virDomainNetGetActualVirtPortProfile(net);

        switch (virDomainNetGetActualType(net)) {
        case VIR_DOMAIN_NET_TYPE_DIRECT:
7173
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
7174
                             net->ifname, &net->mac,
7175 7176
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
7177
                             virDomainNetGetActualVirtPortProfile(net),
7178
                             cfg->stateDir));
7179
            break;
7180 7181 7182 7183 7184 7185
        case VIR_DOMAIN_NET_TYPE_ETHERNET:
            if (net->ifname) {
                ignore_value(virNetDevTapDelete(net->ifname, net->backend.tap));
                VIR_FREE(net->ifname);
            }
            break;
7186 7187 7188 7189
        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))
7190
                ignore_value(virNetDevTapDelete(net->ifname, net->backend.tap));
7191 7192
#endif
            break;
7193 7194 7195 7196 7197 7198 7199 7200 7201 7202 7203
        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;
7204
        }
7205 7206 7207
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
7208 7209 7210 7211 7212 7213 7214 7215 7216
        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
7217

7218 7219
        /* kick the device out of the hostdev list too */
        virDomainNetRemoveHostdev(def, net);
7220
        virDomainNetReleaseActualDevice(vm->def, net);
7221
    }
7222

7223
 retry:
7224
    if ((ret = qemuRemoveCgroup(vm)) < 0) {
7225 7226 7227 7228 7229 7230 7231
        if (ret == -EBUSY && (retries++ < 5)) {
            usleep(200*1000);
            goto retry;
        }
        VIR_WARN("Failed to remove cgroup for %s",
                 vm->def->name);
    }
7232

7233 7234 7235
    /* 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 */
7236 7237 7238 7239 7240 7241 7242 7243 7244 7245
    for (i = 0; i < vm->def->nresctrls; i++) {
        size_t j = 0;

        for (j = 0; j < vm->def->resctrls[i]->nmonitors; j++) {
            virDomainResctrlMonDefPtr mon = NULL;

            mon = vm->def->resctrls[i]->monitors[j];
            virResctrlMonitorRemove(mon->instance);
        }

B
Bing Niu 已提交
7246
        virResctrlAllocRemove(vm->def->resctrls[i]->alloc);
7247
    }
7248

7249 7250
    qemuProcessRemoveDomainStatus(driver, vm);

7251 7252
    /* Remove VNC and Spice ports from port reservation bitmap, but only if
       they were reserved by the driver (autoport=yes)
7253
    */
7254
    for (i = 0; i < vm->def->ngraphics; ++i) {
7255
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
7256 7257
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (graphics->data.vnc.autoport) {
7258
                virPortAllocatorRelease(graphics->data.vnc.port);
7259
            } else if (graphics->data.vnc.portReserved) {
7260
                virPortAllocatorRelease(graphics->data.vnc.port);
7261 7262
                graphics->data.vnc.portReserved = false;
            }
7263
            if (graphics->data.vnc.websocketGenerated) {
7264
                virPortAllocatorRelease(graphics->data.vnc.websocket);
7265 7266 7267
                graphics->data.vnc.websocketGenerated = false;
                graphics->data.vnc.websocket = -1;
            } else if (graphics->data.vnc.websocket) {
7268
                virPortAllocatorRelease(graphics->data.vnc.websocket);
7269
            }
7270
        }
7271 7272
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (graphics->data.spice.autoport) {
7273 7274
                virPortAllocatorRelease(graphics->data.spice.port);
                virPortAllocatorRelease(graphics->data.spice.tlsPort);
7275 7276
            } else {
                if (graphics->data.spice.portReserved) {
7277
                    virPortAllocatorRelease(graphics->data.spice.port);
7278 7279 7280 7281
                    graphics->data.spice.portReserved = false;
                }

                if (graphics->data.spice.tlsPortReserved) {
7282
                    virPortAllocatorRelease(graphics->data.spice.tlsPort);
7283 7284 7285
                    graphics->data.spice.tlsPortReserved = false;
                }
            }
7286
        }
7287 7288
    }

7289
    vm->taint = 0;
7290
    vm->pid = -1;
J
Jiri Denemark 已提交
7291
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
7292 7293
    for (i = 0; i < vm->def->niothreadids; i++)
        vm->def->iothreadids[i]->thread_id = 0;
7294

7295 7296
    /* clear all private data entries which are no longer needed */
    qemuDomainObjPrivateDataClear(priv);
7297

7298 7299 7300
    /* reset node name allocator */
    qemuDomainStorageIdReset(priv);

7301
    /* The "release" hook cleans up additional resources */
7302
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
7303
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
7304 7305 7306

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
7307 7308
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
7309 7310 7311
        VIR_FREE(xml);
    }

7312
    virDomainObjRemoveTransientDef(vm);
7313

7314 7315 7316 7317 7318
 endjob:
    if (asyncJob != QEMU_ASYNC_JOB_NONE)
        qemuDomainObjEndJob(driver, vm);

 cleanup:
7319 7320 7321 7322
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
7323
    virObjectUnref(cfg);
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
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;
7349
    virDomainVirtType virtType;
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 7387 7388 7389 7390 7391 7392 7393 7394 7395 7396 7397 7398 7399 7400 7401 7402 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

    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");
7439
    if (qemuDomainUpdateQEMUCaps(vm, driver->qemuCapsCache) < 0)
7440 7441 7442 7443 7444 7445 7446
        goto error;

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

7447 7448 7449 7450
    /* Attaching to running QEMU so we need to detect whether it was started
     * with -no-reboot. */
    qemuProcessPrepareAllowReboot(vm);

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
    /*
     * 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;
7496
    if (qemuMonitorGetVirtType(priv->mon, &virtType) < 0)
7497
        goto exit_monitor;
7498
    vm->def->virtType = virtType;
7499 7500 7501 7502 7503 7504 7505 7506 7507 7508 7509 7510 7511 7512 7513 7514 7515 7516 7517 7518 7519 7520 7521 7522 7523 7524 7525 7526 7527 7528 7529 7530 7531 7532 7533 7534 7535 7536 7537 7538 7539 7540 7541 7542 7543 7544 7545 7546 7547 7548 7549 7550 7551 7552 7553 7554 7555 7556 7557 7558 7559 7560 7561 7562 7563 7564 7565 7566 7567 7568 7569 7570
    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;
}


7571
static void
7572 7573 7574 7575 7576 7577 7578 7579 7580 7581 7582 7583 7584 7585 7586 7587 7588 7589 7590 7591 7592 7593 7594
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)
7595
        return;
7596 7597 7598 7599 7600 7601 7602 7603 7604 7605 7606

    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);

7607 7608
    qemuDomainObjEndJob(driver, dom);

7609
    virObjectEventStateQueue(driver->domainEventState, event);
7610 7611 7612 7613 7614 7615 7616 7617 7618 7619 7620 7621 7622 7623 7624 7625 7626 7627 7628 7629 7630 7631 7632 7633 7634 7635 7636 7637 7638 7639 7640 7641 7642 7643 7644 7645 7646
}

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;
7647
    bool blockdev = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV);
7648 7649 7650 7651 7652 7653 7654 7655 7656 7657 7658 7659 7660 7661 7662 7663 7664
    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;
7665 7666 7667 7668
        const char *entryname = disk->info.alias;

        if (blockdev)
            entryname = diskpriv->qomName;
7669

7670
        if (!(info = virHashLookup(table, entryname)))
7671 7672 7673 7674 7675 7676 7677 7678 7679 7680 7681 7682 7683 7684 7685 7686 7687 7688 7689 7690 7691 7692 7693 7694 7695 7696 7697
            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;
}


7698 7699 7700 7701 7702
static int
qemuProcessRefreshCPU(virQEMUDriverPtr driver,
                      virDomainObjPtr vm)
{
    virCapsPtr caps = virQEMUDriverGetCapabilities(driver, false);
7703
    qemuDomainObjPrivatePtr priv = vm->privateData;
7704
    virCPUDefPtr host = NULL;
7705
    virCPUDefPtr cpu = NULL;
7706 7707 7708 7709 7710 7711 7712 7713 7714 7715 7716 7717 7718 7719 7720 7721 7722 7723 7724 7725 7726
    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;

7727 7728 7729 7730 7731 7732 7733
        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)
7734 7735 7736 7737
            goto cleanup;

        if (qemuProcessUpdateCPU(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
            goto cleanup;
7738 7739 7740 7741 7742 7743 7744 7745
    } 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;
7746 7747 7748 7749 7750
    }

    ret = 0;

 cleanup:
7751
    virCPUDefFree(cpu);
7752 7753 7754 7755 7756 7757
    virCPUDefFree(host);
    virObjectUnref(caps);
    return ret;
}


7758 7759 7760
struct qemuProcessReconnectData {
    virQEMUDriverPtr driver;
    virDomainObjPtr obj;
7761
    virIdentityPtr identity;
7762 7763 7764 7765 7766 7767 7768 7769 7770 7771 7772 7773 7774 7775 7776 7777 7778 7779 7780 7781 7782
};
/*
 * 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)
7783
{
7784 7785 7786 7787
    struct qemuProcessReconnectData *data = opaque;
    virQEMUDriverPtr driver = data->driver;
    virDomainObjPtr obj = data->obj;
    qemuDomainObjPrivatePtr priv;
7788
    qemuDomainJobObj oldjob;
7789 7790 7791
    int state;
    int reason;
    virQEMUDriverConfigPtr cfg;
7792
    size_t i;
7793 7794
    unsigned int stopFlags = 0;
    bool jobStarted = false;
7795
    virCapsPtr caps = NULL;
7796
    bool retry = true;
7797
    bool tryMonReconn = false;
7798

7799 7800
    virIdentitySetCurrent(data->identity);
    virObjectUnref(data->identity);
7801
    VIR_FREE(data);
7802

7803 7804 7805 7806 7807 7808
    qemuDomainObjRestoreJob(obj, &oldjob);
    if (oldjob.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;

    cfg = virQEMUDriverGetConfig(driver);
    priv = obj->privateData;
7809

7810
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
7811
        goto error;
7812

7813
    if (qemuDomainObjBeginJob(driver, obj, QEMU_JOB_MODIFY) < 0)
7814
        goto error;
7815
    jobStarted = true;
7816

7817 7818 7819 7820
    /* 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;
7821

7822 7823 7824
    /* Restore the masterKey */
    if (qemuDomainMasterKeyReadFile(priv) < 0)
        goto error;
7825

7826 7827 7828 7829
    /* 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);

7830 7831 7832
    if (qemuHostdevUpdateActiveDomainDevices(driver, obj->def) < 0)
        goto error;

7833 7834 7835 7836 7837 7838
    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);
7839

7840 7841
    tryMonReconn = true;

7842
    /* XXX check PID liveliness & EXE path */
7843
    if (qemuConnectMonitor(driver, obj, QEMU_ASYNC_JOB_NONE, retry, NULL) < 0)
7844
        goto error;
7845

7846 7847 7848 7849
    priv->machineName = qemuDomainGetMachineName(obj);
    if (!priv->machineName)
        goto error;

7850
    if (qemuConnectCgroup(obj) < 0)
7851
        goto error;
7852

7853
    if (qemuDomainPerfRestart(obj) < 0)
7854
        goto error;
7855

7856 7857 7858 7859
    /* XXX: Need to change as long as lock is introduced for
     * qemu_driver->sharedDevices.
     */
    for (i = 0; i < obj->def->ndisks; i++) {
7860
        virDomainDiskDefPtr disk = obj->def->disks[i];
7861
        virDomainDeviceDef dev;
7862

7863
        if (virDomainDiskTranslateSourcePool(disk) < 0)
7864
            goto error;
7865

7866
        /* backing chains need to be refreshed only if they could change */
7867 7868
        if (priv->reconnectBlockjobs != VIR_TRISTATE_BOOL_NO &&
            !virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV)) {
7869 7870 7871
            /* This should be the only place that calls
             * qemuDomainDetermineDiskChain with @report_broken == false
             * to guarantee best-effort domain reconnect */
7872 7873
            virStorageSourceBackingStoreClear(disk->src);
            if (qemuDomainDetermineDiskChain(driver, obj, disk, false) < 0)
7874 7875 7876 7877
                goto error;
        } else {
            VIR_DEBUG("skipping backing chain detection for '%s'", disk->dst);
        }
7878

7879
        dev.type = VIR_DOMAIN_DEVICE_DISK;
7880
        dev.data.disk = disk;
7881
        if (qemuAddSharedDevice(driver, &dev, obj->def->name) < 0)
7882 7883 7884 7885
            goto error;
    }

    for (i = 0; i < obj->def->ngraphics; i++) {
7886
        if (qemuProcessGraphicsReservePorts(obj->def->graphics[i], true) < 0)
7887
            goto error;
7888
    }
7889

7890
    if (qemuProcessUpdateState(driver, obj) < 0)
7891 7892
        goto error;

7893 7894 7895 7896 7897 7898
    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);
7899
        goto error;
7900
    }
7901

7902 7903 7904 7905
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
    if (!priv->qemuCaps &&
7906
        (qemuDomainUpdateQEMUCaps(obj, driver->qemuCapsCache) < 0))
7907
        goto error;
7908

7909 7910 7911 7912 7913 7914 7915 7916 7917 7918 7919 7920
    /* 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;
    }
7921

7922
    if (qemuProcessBuildDestroyMemoryPaths(driver, obj, NULL, true) < 0)
7923
        goto error;
7924

7925 7926
    if ((qemuDomainAssignAddresses(obj->def, priv->qemuCaps,
                                   driver, obj, false)) < 0) {
7927
        goto error;
7928
    }
7929

7930 7931 7932 7933 7934 7935
    /* if domain requests security driver we haven't loaded, report error, but
     * do not kill the domain
     */
    ignore_value(qemuSecurityCheckAllLabel(driver->securityManager,
                                           obj->def));

7936 7937
    if (qemuProcessRefreshCPU(driver, obj) < 0)
        goto error;
7938

7939
    if (qemuDomainRefreshVcpuInfo(driver, obj, QEMU_ASYNC_JOB_NONE, true) < 0)
7940
        goto error;
7941

7942
    qemuDomainVcpuPersistOrder(obj->def);
7943

7944 7945 7946
    if (qemuProcessDetectIOThreadPIDs(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
        goto error;

7947 7948
    if (qemuSecurityReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
        goto error;
7949

7950 7951
    qemuProcessNotifyNets(obj->def);

7952
    qemuProcessFiltersInstantiate(obj->def);
7953

7954
    if (qemuProcessRefreshDisks(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
7955
        goto error;
D
Daniel P. Berrange 已提交
7956

7957 7958
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV) &&
        qemuBlockNodeNamesDetect(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
7959 7960
        goto error;

7961
    if (qemuRefreshVirtioChannelState(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
7962 7963
        goto error;

7964 7965 7966 7967
    /* 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)
7968
        goto error;
7969

7970
    if (qemuProcessRecoverJob(driver, obj, &oldjob, &stopFlags) < 0)
7971
        goto error;
7972

7973 7974
    if (qemuProcessUpdateDevices(driver, obj) < 0)
        goto error;
7975

7976 7977 7978
    if (qemuRefreshPRManagerState(driver, obj) < 0)
        goto error;

7979 7980 7981 7982 7983
    qemuProcessReconnectCheckMemAliasOrderMismatch(obj);

    if (qemuConnectAgent(driver, obj) < 0)
        goto error;

B
Bing Niu 已提交
7984
    for (i = 0; i < obj->def->nresctrls; i++) {
7985 7986
        size_t j = 0;

B
Bing Niu 已提交
7987
        if (virResctrlAllocDeterminePath(obj->def->resctrls[i]->alloc,
7988 7989
                                         priv->machineName) < 0)
            goto error;
7990 7991 7992 7993 7994 7995 7996 7997 7998

        for (j = 0; j < obj->def->resctrls[i]->nmonitors; j++) {
            virDomainResctrlMonDefPtr mon = NULL;

            mon = obj->def->resctrls[i]->monitors[j];
            if (virResctrlMonitorDeterminePath(mon->instance,
                                               priv->machineName) < 0)
                goto error;
        }
7999 8000
    }

8001 8002
    /* update domain state XML with possibly updated state in virDomainObj */
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, obj, driver->caps) < 0)
8003
        goto error;
8004

8005 8006
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
8007
        char *xml = qemuDomainDefFormatXML(driver, obj->def, 0);
8008 8009
        int hookret;

8010 8011
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, obj->def->name,
                              VIR_HOOK_QEMU_OP_RECONNECT, VIR_HOOK_SUBOP_BEGIN,
8012 8013 8014 8015 8016 8017 8018
                              NULL, xml, NULL);
        VIR_FREE(xml);

        /*
         * If the script raised an error abort the launch
         */
        if (hookret < 0)
8019
            goto error;
8020 8021
    }

8022 8023
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
        driver->inhibitCallback(true, driver->inhibitOpaque);
8024

8025
 cleanup:
8026 8027 8028
    if (jobStarted) {
        if (!virDomainObjIsActive(obj))
            qemuDomainRemoveInactive(driver, obj);
8029
        qemuDomainObjEndJob(driver, obj);
8030 8031 8032 8033
    } else {
        if (!virDomainObjIsActive(obj))
            qemuDomainRemoveInactiveJob(driver, obj);
    }
8034
    virDomainObjEndAPI(&obj);
8035
    virObjectUnref(cfg);
8036
    virObjectUnref(caps);
8037
    virNWFilterUnlockFilterUpdates();
8038
    virIdentitySetCurrent(NULL);
8039
    return;
8040

8041 8042 8043 8044
 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
J
John Ferlan 已提交
8045 8046 8047 8048
         * user tries to start it again later.
         *
         * If we cannot get to the monitor when the QEMU command
         * line used -no-shutdown, then we can safely say that the
8049 8050 8051
         * domain crashed; otherwise, if the monitor was started,
         * then we can blame ourselves, else we failed before the
         * monitor started so we don't really know. */
8052 8053
        if (!priv->mon && tryMonReconn &&
            qemuDomainIsUsingNoShutdown(priv))
J
John Ferlan 已提交
8054
            state = VIR_DOMAIN_SHUTOFF_CRASHED;
8055 8056
        else if (priv->mon)
            state = VIR_DOMAIN_SHUTOFF_DAEMON;
J
John Ferlan 已提交
8057 8058 8059
        else
            state = VIR_DOMAIN_SHUTOFF_UNKNOWN;

8060 8061 8062 8063 8064
        /* 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);
8065
    }
8066
    goto cleanup;
8067
}
8068

8069 8070 8071
static int
qemuProcessReconnectHelper(virDomainObjPtr obj,
                           void *opaque)
8072
{
8073 8074 8075
    virThread thread;
    struct qemuProcessReconnectData *src = opaque;
    struct qemuProcessReconnectData *data;
8076

8077 8078 8079
    /* If the VM was inactive, we don't need to reconnect */
    if (!obj->pid)
        return 0;
8080

8081 8082
    if (VIR_ALLOC(data) < 0)
        return -1;
8083

8084 8085
    memcpy(data, src, sizeof(*data));
    data->obj = obj;
8086
    data->identity = virIdentityGetCurrent();
8087

8088 8089
    virNWFilterReadLockFilterUpdates();

8090 8091 8092 8093
    /* this lock and reference will be eventually transferred to the thread
     * that handles the reconnect */
    virObjectLock(obj);
    virObjectRef(obj);
8094

8095 8096 8097 8098 8099 8100 8101 8102 8103 8104 8105
    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);
8106
        qemuDomainRemoveInactiveJobLocked(src->driver, obj);
8107

8108
        virDomainObjEndAPI(&obj);
8109
        virNWFilterUnlockFilterUpdates();
8110
        virObjectUnref(data->identity);
8111 8112
        VIR_FREE(data);
        return -1;
8113 8114
    }

8115 8116
    return 0;
}
8117

8118 8119 8120 8121 8122 8123 8124
/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
8125
qemuProcessReconnectAll(virQEMUDriverPtr driver)
8126
{
8127
    struct qemuProcessReconnectData data = {.driver = driver};
8128
    virDomainObjListForEach(driver->domains, qemuProcessReconnectHelper, &data);
8129
}