qemu_process.c 228.8 KB
Newer Older
1
/*
2
 * qemu_process.c: QEMU process management
3
 *
4
 * Copyright (C) 2006-2016 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23 24 25 26 27
 *
 */

#include <config.h>

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

#include "qemu_process.h"
36
#include "qemu_processpriv.h"
37
#include "qemu_alias.h"
38
#include "qemu_block.h"
39
#include "qemu_domain.h"
40
#include "qemu_domain_address.h"
41 42 43 44 45 46
#include "qemu_cgroup.h"
#include "qemu_capabilities.h"
#include "qemu_monitor.h"
#include "qemu_command.h"
#include "qemu_hostdev.h"
#include "qemu_hotplug.h"
47
#include "qemu_migration.h"
48
#include "qemu_interface.h"
49
#include "qemu_security.h"
50

51
#include "cpu/cpu.h"
52
#include "datatypes.h"
53
#include "virlog.h"
54
#include "virerror.h"
55
#include "viralloc.h"
56
#include "virhook.h"
E
Eric Blake 已提交
57
#include "virfile.h"
58
#include "virpidfile.h"
59
#include "virhostcpu.h"
60
#include "domain_audit.h"
61
#include "domain_nwfilter.h"
62
#include "locking/domain_lock.h"
63
#include "viruuid.h"
64
#include "virprocess.h"
65
#include "virtime.h"
A
Ansis Atteka 已提交
66
#include "virnetdevtap.h"
67
#include "virnetdevopenvswitch.h"
68
#include "virnetdevmidonet.h"
69
#include "virbitmap.h"
70
#include "viratomic.h"
71
#include "virnuma.h"
72
#include "virstring.h"
73
#include "virhostdev.h"
J
John Ferlan 已提交
74
#include "secret_util.h"
75
#include "configmake.h"
76
#include "nwfilter_conf.h"
77
#include "netdev_bandwidth_conf.h"
78
#include "virresctrl.h"
79 80 81

#define VIR_FROM_THIS VIR_FROM_QEMU

82 83
VIR_LOG_INIT("qemu.qemu_process");

84
/**
85
 * qemuProcessRemoveDomainStatus
86 87 88 89 90 91
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
92
qemuProcessRemoveDomainStatus(virQEMUDriverPtr driver,
93 94 95 96
                              virDomainObjPtr vm)
{
    char ebuf[1024];
    char *file = NULL;
97
    qemuDomainObjPrivatePtr priv = vm->privateData;
98 99
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;
100

101
    if (virAsprintf(&file, "%s/%s.xml", cfg->stateDir, vm->def->name) < 0)
102
        goto cleanup;
103 104 105 106 107 108

    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);

109 110 111
    if (priv->pidfile &&
        unlink(priv->pidfile) < 0 &&
        errno != ENOENT)
112 113 114
        VIR_WARN("Failed to remove PID file for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));

115
    ret = 0;
116
 cleanup:
117 118
    virObjectUnref(cfg);
    return ret;
119 120 121
}


D
Daniel P. Berrange 已提交
122 123 124 125 126 127 128
/*
 * 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
129
qemuProcessHandleAgentEOF(qemuAgentPtr agent,
D
Daniel P. Berrange 已提交
130 131 132 133 134 135
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

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

136
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
137 138

    priv = vm->privateData;
139 140 141 142 143 144 145 146 147 148 149

    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;
    }

150
    qemuAgentClose(agent);
151
    priv->agent = NULL;
152
    priv->agentError = false;
D
Daniel P. Berrange 已提交
153

154
    virObjectUnlock(vm);
155 156
    return;

157
 unlock:
158 159
    virObjectUnlock(vm);
    return;
D
Daniel P. Berrange 已提交
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
}


/*
 * 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);

177
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
178 179 180 181 182

    priv = vm->privateData;

    priv->agentError = true;

183
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
184 185 186 187 188
}

static void qemuProcessHandleAgentDestroy(qemuAgentPtr agent,
                                          virDomainObjPtr vm)
{
189 190
    VIR_DEBUG("Received destroy agent=%p vm=%p", agent, vm);

191
    virObjectUnref(vm);
D
Daniel P. Berrange 已提交
192 193 194 195 196 197 198 199 200 201
}


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


202
int
203
qemuConnectAgent(virQEMUDriverPtr driver, virDomainObjPtr vm)
D
Daniel P. Berrange 已提交
204 205 206
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuAgentPtr agent = NULL;
207
    virDomainChrDefPtr config = qemuFindAgentConfig(vm->def);
D
Daniel P. Berrange 已提交
208 209 210 211

    if (!config)
        return 0;

212 213 214 215 216 217 218 219 220
    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;
    }

221
    if (qemuSecuritySetDaemonSocketLabel(driver->securityManager, vm->def) < 0) {
D
Daniel P. Berrange 已提交
222 223 224 225 226 227 228
        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 */
229
    virObjectRef(vm);
D
Daniel P. Berrange 已提交
230

231
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
232 233

    agent = qemuAgentOpen(vm,
234
                          config->source,
D
Daniel P. Berrange 已提交
235 236
                          &agentCallbacks);

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

239 240 241 242 243 244 245
    if (agent == NULL)
        virObjectUnref(vm);

    if (!virDomainObjIsActive(vm)) {
        qemuAgentClose(agent);
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest crashed while connecting to the guest agent"));
246
        return -1;
247 248
    }

249
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0) {
D
Daniel P. Berrange 已提交
250 251
        VIR_ERROR(_("Failed to clear security context for agent for %s"),
                  vm->def->name);
252
        qemuAgentClose(agent);
D
Daniel P. Berrange 已提交
253 254 255 256
        goto cleanup;
    }

    priv->agent = agent;
257
    if (!priv->agent)
D
Daniel P. Berrange 已提交
258 259
        VIR_INFO("Failed to connect agent for %s", vm->def->name);

260
 cleanup:
261 262 263 264 265 266 267
    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
268 269 270
}


271
/*
272
 * This is a callback registered with a qemuMonitorPtr instance,
273 274 275 276 277
 * 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
278
qemuProcessHandleMonitorEOF(qemuMonitorPtr mon,
279 280
                            virDomainObjPtr vm,
                            void *opaque)
281
{
282
    virQEMUDriverPtr driver = opaque;
283
    qemuDomainObjPrivatePtr priv;
284
    struct qemuProcessEvent *processEvent;
285

286
    virObjectLock(vm);
287

288
    VIR_DEBUG("Received EOF on %p '%s'", vm, vm->def->name);
289

290
    priv = vm->privateData;
291 292
    if (priv->beingDestroyed) {
        VIR_DEBUG("Domain is being destroyed, EOF is expected");
293
        goto cleanup;
294 295
    }

296
    if (VIR_ALLOC(processEvent) < 0)
297
        goto cleanup;
298

299
    processEvent->eventType = QEMU_PROCESS_EVENT_MONITOR_EOF;
300
    processEvent->vm = virObjectRef(vm);
301

302 303
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
304
        qemuProcessEventFree(processEvent);
305
        goto cleanup;
306
    }
307

308 309 310 311
    /* We don't want this EOF handler to be called over and over while the
     * thread is waiting for a job.
     */
    qemuMonitorUnregister(mon);
312

313 314 315 316
    /* We don't want any cleanup from EOF handler (or any other
     * thread) to enter qemu namespace. */
    qemuDomainDestroyNamespace(driver, vm);

317
 cleanup:
318
    virObjectUnlock(vm);
319 320 321 322 323 324 325 326 327 328 329
}


/*
 * 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,
330 331
                              virDomainObjPtr vm,
                              void *opaque)
332
{
333
    virQEMUDriverPtr driver = opaque;
334
    virObjectEventPtr event = NULL;
335 336 337

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

338
    virObjectLock(vm);
339

340
    ((qemuDomainObjPrivatePtr) vm->privateData)->monError = true;
341
    event = virDomainEventControlErrorNewFromObj(vm);
342
    qemuDomainEventQueue(driver, event);
343

344
    virObjectUnlock(vm);
345 346 347
}


348
virDomainDiskDefPtr
349 350 351
qemuProcessFindDomainDiskByAlias(virDomainObjPtr vm,
                                 const char *alias)
{
352
    size_t i;
353

354
    alias = qemuAliasDiskDriveSkipPrefix(alias);
355 356 357 358 359 360 361 362 363

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

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

364 365 366
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with alias %s"),
                   alias);
367 368 369 370
    return NULL;
}

static int
371
qemuProcessGetVolumeQcowPassphrase(virDomainDiskDefPtr disk,
372 373 374
                                   char **secretRet,
                                   size_t *secretLen)
{
375
    virConnectPtr conn = NULL;
376 377 378 379 380 381
    char *passphrase;
    unsigned char *data;
    size_t size;
    int ret = -1;
    virStorageEncryptionPtr enc;

382
    if (!disk->src->encryption) {
383 384
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("disk %s does not have any encryption information"),
385
                       disk->src->path);
386 387
        return -1;
    }
388
    enc = disk->src->encryption;
389

390
    if (!(conn = virGetConnectSecret()))
391 392 393 394 395 396
        goto cleanup;

    if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW ||
        enc->nsecrets != 1 ||
        enc->secrets[0]->type !=
        VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE) {
397
        virReportError(VIR_ERR_XML_ERROR,
398 399
                       _("invalid <encryption> for volume %s"),
                       virDomainDiskGetSource(disk));
400 401 402
        goto cleanup;
    }

J
John Ferlan 已提交
403 404 405
    if (virSecretGetSecretString(conn, &enc->secrets[0]->seclookupdef,
                                 VIR_SECRET_USAGE_TYPE_VOLUME,
                                 &data, &size) < 0)
406 407 408 409 410
        goto cleanup;

    if (memchr(data, '\0', size) != NULL) {
        memset(data, 0, size);
        VIR_FREE(data);
411 412
        virReportError(VIR_ERR_XML_ERROR,
                       _("format='qcow' passphrase for %s must not contain a "
413
                         "'\\0'"), virDomainDiskGetSource(disk));
414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432
        goto cleanup;
    }

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

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

    *secretRet = passphrase;
    *secretLen = size;

    ret = 0;

433
 cleanup:
434
    virObjectUnref(conn);
435 436 437 438 439 440 441 442
    return ret;
}

static int
qemuProcessFindVolumeQcowPassphrase(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virDomainObjPtr vm,
                                    const char *path,
                                    char **secretRet,
443 444
                                    size_t *secretLen,
                                    void *opaque ATTRIBUTE_UNUSED)
445 446 447 448
{
    virDomainDiskDefPtr disk;
    int ret = -1;

449
    virObjectLock(vm);
450 451 452 453
    if (!(disk = virDomainDiskByName(vm->def, path, true))) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("no disk found with path %s"),
                       path);
454
        goto cleanup;
455
    }
456

457
    ret = qemuProcessGetVolumeQcowPassphrase(disk, secretRet, secretLen);
458

459
 cleanup:
460
    virObjectUnlock(vm);
461 462 463 464 465 466
    return ret;
}


static int
qemuProcessHandleReset(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
467 468
                       virDomainObjPtr vm,
                       void *opaque)
469
{
470
    virQEMUDriverPtr driver = opaque;
471
    virObjectEventPtr event;
472
    qemuDomainObjPrivatePtr priv;
473
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
M
Michal Privoznik 已提交
474
    int ret = -1;
475

476
    virObjectLock(vm);
477

478
    event = virDomainEventRebootNewFromObj(vm);
479 480 481
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);
482

483
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
484
        VIR_WARN("Failed to save status on vm %s", vm->def->name);
485

486 487
    if (vm->def->onReboot == VIR_DOMAIN_LIFECYCLE_ACTION_DESTROY ||
        vm->def->onReboot == VIR_DOMAIN_LIFECYCLE_ACTION_PRESERVE) {
488

M
Michal Privoznik 已提交
489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504
        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);
    }
505

M
Michal Privoznik 已提交
506 507 508 509
    ret = 0;
 cleanup:
    virObjectUnlock(vm);
    qemuDomainEventQueue(driver, event);
510
    virObjectUnref(cfg);
M
Michal Privoznik 已提交
511
    return ret;
512 513 514
}


515 516 517 518 519 520 521 522 523 524 525 526 527
/*
 * 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;
528
    virQEMUDriverPtr driver = priv->driver;
529
    virObjectEventPtr event = NULL;
530
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
531
    virDomainRunningReason reason = VIR_DOMAIN_RUNNING_BOOTED;
532
    int ret = -1, rc;
533

534
    VIR_DEBUG("vm=%p", vm);
535
    virObjectLock(vm);
536
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
537 538 539
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
540 541
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
542 543 544
        goto endjob;
    }

545
    qemuDomainObjEnterMonitor(driver, vm);
546 547 548
    rc = qemuMonitorSystemReset(priv->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0)
549 550
        goto endjob;

551
    if (rc < 0)
552 553
        goto endjob;

554 555 556
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_CRASHED)
        reason = VIR_DOMAIN_RUNNING_CRASHED;

557
    if (qemuProcessStartCPUs(driver, vm,
558
                             reason,
559
                             QEMU_ASYNC_JOB_NONE) < 0) {
560
        if (virGetLastError() == NULL)
561 562
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
563 564
        goto endjob;
    }
565
    priv->gotShutdown = false;
566
    event = virDomainEventLifecycleNewFromObj(vm,
567 568 569
                                     VIR_DOMAIN_EVENT_RESUMED,
                                     VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

570
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
571 572 573 574
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

575 576
    ret = 0;

577
 endjob:
578
    qemuDomainObjEndJob(driver, vm);
579

580
 cleanup:
581 582
    if (ret == -1)
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_FORCE));
M
Michal Privoznik 已提交
583
    virDomainObjEndAPI(&vm);
584
    qemuDomainEventQueue(driver, event);
585
    virObjectUnref(cfg);
586 587 588
}


589
void
590
qemuProcessShutdownOrReboot(virQEMUDriverPtr driver,
591
                            virDomainObjPtr vm)
592
{
593 594 595
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
596
        qemuDomainSetFakeReboot(driver, vm, false);
597
        virObjectRef(vm);
598 599 600 601 602
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
603
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
604
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
605
            virObjectUnref(vm);
606 607
        }
    } else {
608
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
609
    }
610
}
611

612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632

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);
633
    qemuDomainEventQueue(driver, event);
634 635 636 637 638

    return 0;
}


639 640
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
641
                          virDomainObjPtr vm,
642
                          virTristateBool guest_initiated,
643
                          void *opaque)
644
{
645
    virQEMUDriverPtr driver = opaque;
646
    qemuDomainObjPrivatePtr priv;
647
    virObjectEventPtr event = NULL;
648
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
649
    int detail = 0;
650

651 652
    VIR_DEBUG("vm=%p", vm);

653
    virObjectLock(vm);
654 655 656 657 658 659

    priv = vm->privateData;
    if (priv->gotShutdown) {
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
660 661 662 663
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
664 665 666 667 668 669 670 671
    }
    priv->gotShutdown = true;

    VIR_DEBUG("Transitioned guest %s to shutdown state",
              vm->def->name);
    virDomainObjSetState(vm,
                         VIR_DOMAIN_SHUTDOWN,
                         VIR_DOMAIN_SHUTDOWN_UNKNOWN);
672 673 674 675 676 677 678 679 680 681

    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;

682 683
    case VIR_TRISTATE_BOOL_ABSENT:
    case VIR_TRISTATE_BOOL_LAST:
684 685 686 687 688
    default:
        detail = VIR_DOMAIN_EVENT_SHUTDOWN_FINISHED;
        break;
    }

689
    event = virDomainEventLifecycleNewFromObj(vm,
690 691
                                              VIR_DOMAIN_EVENT_SHUTDOWN,
                                              detail);
692

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

698 699 700
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

701 702
    qemuProcessShutdownOrReboot(driver, vm);

703
 unlock:
704
    virObjectUnlock(vm);
705
    qemuDomainEventQueue(driver, event);
706
    virObjectUnref(cfg);
707

708 709 710 711 712 713
    return 0;
}


static int
qemuProcessHandleStop(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
714 715
                      virDomainObjPtr vm,
                      void *opaque)
716
{
717
    virQEMUDriverPtr driver = opaque;
718
    virObjectEventPtr event = NULL;
719 720
    virDomainPausedReason reason = VIR_DOMAIN_PAUSED_UNKNOWN;
    virDomainEventSuspendedDetailType detail = VIR_DOMAIN_EVENT_SUSPENDED_PAUSED;
721
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
722

723
    virObjectLock(vm);
J
Jiri Denemark 已提交
724
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
725
        qemuDomainObjPrivatePtr priv = vm->privateData;
726

727
        if (priv->gotShutdown) {
728 729
            VIR_DEBUG("Ignoring STOP event after SHUTDOWN");
            goto unlock;
730 731
        }

732
        if (priv->job.asyncJob == QEMU_ASYNC_JOB_MIGRATION_OUT) {
733 734
            if (priv->job.current->status ==
                        QEMU_DOMAIN_JOB_STATUS_POSTCOPY) {
735 736 737 738 739 740
                reason = VIR_DOMAIN_PAUSED_POSTCOPY;
                detail = VIR_DOMAIN_EVENT_SUSPENDED_POSTCOPY;
            } else {
                reason = VIR_DOMAIN_PAUSED_MIGRATION;
                detail = VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED;
            }
741 742 743 744
        }

        VIR_DEBUG("Transitioned guest %s to paused state, reason %s",
                  vm->def->name, virDomainPausedReasonTypeToString(reason));
745

746 747 748
        if (priv->job.current)
            ignore_value(virTimeMillisNow(&priv->job.current->stopped));

749 750 751 752
        if (priv->signalStop)
            virDomainObjBroadcast(vm);

        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
753
        event = virDomainEventLifecycleNewFromObj(vm,
754 755
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  detail);
756

757 758 759 760 761
        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));

762
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
763 764 765
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
766
    }
767

768
 unlock:
769
    virObjectUnlock(vm);
770
    qemuDomainEventQueue(driver, event);
771
    virObjectUnref(cfg);
772 773 774 775 776

    return 0;
}


777 778
static int
qemuProcessHandleResume(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
779 780
                        virDomainObjPtr vm,
                        void *opaque)
781
{
782
    virQEMUDriverPtr driver = opaque;
783
    virObjectEventPtr event = NULL;
784
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
785

786
    virObjectLock(vm);
787 788 789 790 791 792 793 794 795 796 797 798 799
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
        qemuDomainObjPrivatePtr priv = vm->privateData;

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

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

        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_UNPAUSED);
800
        event = virDomainEventLifecycleNewFromObj(vm,
801 802 803
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

804
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
805 806 807 808 809
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
    }

810
 unlock:
811
    virObjectUnlock(vm);
812
    qemuDomainEventQueue(driver, event);
813
    virObjectUnref(cfg);
814 815 816
    return 0;
}

817 818 819
static int
qemuProcessHandleRTCChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm,
820 821
                           long long offset,
                           void *opaque)
822
{
823
    virQEMUDriverPtr driver = opaque;
824
    virObjectEventPtr event = NULL;
825
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
826

827
    virObjectLock(vm);
828

829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845
    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;
846
        vm->def->clock.data.variable.adjustment = offset;
847

848
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
849 850 851 852
           VIR_WARN("unable to save domain status with RTC change");
    }

    event = virDomainEventRTCChangeNewFromObj(vm, offset);
853

854
    virObjectUnlock(vm);
855

856
    qemuDomainEventQueue(driver, event);
857
    virObjectUnref(cfg);
858 859 860 861 862 863 864
    return 0;
}


static int
qemuProcessHandleWatchdog(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
865 866
                          int action,
                          void *opaque)
867
{
868
    virQEMUDriverPtr driver = opaque;
869 870
    virObjectEventPtr watchdogEvent = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
871
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
872

873
    virObjectLock(vm);
874 875 876
    watchdogEvent = virDomainEventWatchdogNewFromObj(vm, action);

    if (action == VIR_DOMAIN_EVENT_WATCHDOG_PAUSE &&
J
Jiri Denemark 已提交
877
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
878
        qemuDomainObjPrivatePtr priv = vm->privateData;
879 880
        VIR_DEBUG("Transitioned guest %s to paused state due to watchdog", vm->def->name);

J
Jiri Denemark 已提交
881
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
882
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
883 884 885
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

886 887 888 889 890
        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));

891
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
892 893 894
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
895 896 897
    }

    if (vm->def->watchdog->action == VIR_DOMAIN_WATCHDOG_ACTION_DUMP) {
C
Chen Fan 已提交
898 899 900 901
        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
902 903 904
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
905
            processEvent->vm = virObjectRef(vm);
C
Chen Fan 已提交
906
            if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
907
                if (!virObjectUnref(vm))
908
                    vm = NULL;
909
                qemuProcessEventFree(processEvent);
W
Wen Congyang 已提交
910
            }
911
        }
912 913
    }

914
    if (vm)
915
        virObjectUnlock(vm);
916 917
    qemuDomainEventQueue(driver, watchdogEvent);
    qemuDomainEventQueue(driver, lifecycleEvent);
918

919
    virObjectUnref(cfg);
920 921 922 923 924 925 926 927 928
    return 0;
}


static int
qemuProcessHandleIOError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virDomainObjPtr vm,
                         const char *diskAlias,
                         int action,
929 930
                         const char *reason,
                         void *opaque)
931
{
932
    virQEMUDriverPtr driver = opaque;
933 934 935
    virObjectEventPtr ioErrorEvent = NULL;
    virObjectEventPtr ioErrorEvent2 = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
936 937 938
    const char *srcPath;
    const char *devAlias;
    virDomainDiskDefPtr disk;
939
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
940

941
    virObjectLock(vm);
942 943 944
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
945
        srcPath = virDomainDiskGetSource(disk);
946 947 948 949 950 951 952 953 954 955
        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
956
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
957
        qemuDomainObjPrivatePtr priv = vm->privateData;
958 959
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

960 961 962
        if (priv->signalIOError)
            virDomainObjBroadcast(vm);

J
Jiri Denemark 已提交
963
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
964
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
965 966 967
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

968 969 970 971 972
        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));

973
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
974 975
            VIR_WARN("Unable to save status on vm %s after IO error", vm->def->name);
    }
976
    virObjectUnlock(vm);
977

978 979 980
    qemuDomainEventQueue(driver, ioErrorEvent);
    qemuDomainEventQueue(driver, ioErrorEvent2);
    qemuDomainEventQueue(driver, lifecycleEvent);
981
    virObjectUnref(cfg);
982 983 984
    return 0;
}

985 986 987 988 989
static int
qemuProcessHandleBlockJob(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          const char *diskAlias,
                          int type,
990
                          int status,
991
                          const char *error,
992
                          void *opaque)
993
{
994
    virQEMUDriverPtr driver = opaque;
995
    struct qemuProcessEvent *processEvent = NULL;
996
    virDomainDiskDefPtr disk;
997
    qemuDomainDiskPrivatePtr diskPriv;
998
    char *data = NULL;
999

1000
    virObjectLock(vm);
1001

1002 1003
    VIR_DEBUG("Block job for device %s (domain: %p,%s) type %d status %d",
              diskAlias, vm, vm->def->name, type, status);
1004

1005
    if (!(disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias)))
1006
        goto error;
1007
    diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1008

1009
    if (diskPriv->blockJobSync) {
1010
        /* We have a SYNC API waiting for this event, dispatch it back */
1011 1012
        diskPriv->blockJobType = type;
        diskPriv->blockJobStatus = status;
1013 1014
        VIR_FREE(diskPriv->blockJobError);
        ignore_value(VIR_STRDUP_QUIET(diskPriv->blockJobError, error));
1015
        virDomainObjBroadcast(vm);
1016 1017 1018 1019 1020 1021 1022 1023 1024
    } 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;
1025
        processEvent->vm = virObjectRef(vm);
1026 1027
        processEvent->action = type;
        processEvent->status = status;
1028

1029 1030 1031 1032
        if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
            ignore_value(virObjectUnref(vm));
            goto error;
        }
1033 1034
    }

1035
 cleanup:
1036
    virObjectUnlock(vm);
1037
    return 0;
1038
 error:
1039
    qemuProcessEventFree(processEvent);
1040
    goto cleanup;
1041
}
1042

1043

1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055
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,
1056 1057
                          const char *saslUsername,
                          void *opaque)
1058
{
1059
    virQEMUDriverPtr driver = opaque;
1060
    virObjectEventPtr event;
1061 1062 1063
    virDomainEventGraphicsAddressPtr localAddr = NULL;
    virDomainEventGraphicsAddressPtr remoteAddr = NULL;
    virDomainEventGraphicsSubjectPtr subject = NULL;
1064
    size_t i;
1065 1066

    if (VIR_ALLOC(localAddr) < 0)
1067
        goto error;
1068
    localAddr->family = localFamily;
1069 1070 1071
    if (VIR_STRDUP(localAddr->service, localService) < 0 ||
        VIR_STRDUP(localAddr->node, localNode) < 0)
        goto error;
1072 1073

    if (VIR_ALLOC(remoteAddr) < 0)
1074
        goto error;
1075
    remoteAddr->family = remoteFamily;
1076 1077 1078
    if (VIR_STRDUP(remoteAddr->service, remoteService) < 0 ||
        VIR_STRDUP(remoteAddr->node, remoteNode) < 0)
        goto error;
1079 1080

    if (VIR_ALLOC(subject) < 0)
1081
        goto error;
1082 1083
    if (x509dname) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1084
            goto error;
1085
        subject->nidentity++;
1086 1087 1088
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "x509dname") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, x509dname) < 0)
            goto error;
1089 1090 1091
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1092
            goto error;
1093
        subject->nidentity++;
1094 1095 1096
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "saslUsername") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, saslUsername) < 0)
            goto error;
1097 1098
    }

1099
    virObjectLock(vm);
1100
    event = virDomainEventGraphicsNewFromObj(vm, phase, localAddr, remoteAddr, authScheme, subject);
1101
    virObjectUnlock(vm);
1102

1103
    qemuDomainEventQueue(driver, event);
1104 1105 1106

    return 0;

1107
 error:
1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
    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) {
1119
        for (i = 0; i < subject->nidentity; i++) {
1120 1121 1122 1123 1124 1125 1126 1127 1128 1129
            VIR_FREE(subject->identities[i].type);
            VIR_FREE(subject->identities[i].name);
        }
        VIR_FREE(subject->identities);
        VIR_FREE(subject);
    }

    return -1;
}

1130 1131 1132 1133
static int
qemuProcessHandleTrayChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            const char *devAlias,
1134 1135
                            int reason,
                            void *opaque)
1136
{
1137
    virQEMUDriverPtr driver = opaque;
1138
    virObjectEventPtr event = NULL;
1139
    virDomainDiskDefPtr disk;
1140
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1141

1142
    virObjectLock(vm);
1143 1144 1145 1146 1147 1148
    disk = qemuProcessFindDomainDiskByAlias(vm, devAlias);

    if (disk) {
        event = virDomainEventTrayChangeNewFromObj(vm,
                                                   devAlias,
                                                   reason);
1149 1150 1151 1152 1153 1154
        /* 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;

1155
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1156 1157 1158
            VIR_WARN("Unable to save status on vm %s after tray moved event",
                     vm->def->name);
        }
1159 1160

        virDomainObjBroadcast(vm);
1161 1162
    }

1163
    virObjectUnlock(vm);
1164
    qemuDomainEventQueue(driver, event);
1165
    virObjectUnref(cfg);
1166 1167 1168
    return 0;
}

O
Osier Yang 已提交
1169 1170
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1171 1172
                          virDomainObjPtr vm,
                          void *opaque)
O
Osier Yang 已提交
1173
{
1174
    virQEMUDriverPtr driver = opaque;
1175 1176
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1177
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1178

1179
    virObjectLock(vm);
O
Osier Yang 已提交
1180 1181
    event = virDomainEventPMWakeupNewFromObj(vm);

1182 1183 1184 1185 1186 1187 1188 1189 1190
    /* 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);
1191
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
1192 1193 1194
                                                  VIR_DOMAIN_EVENT_STARTED,
                                                  VIR_DOMAIN_EVENT_STARTED_WAKEUP);

1195
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1196 1197 1198 1199 1200
            VIR_WARN("Unable to save status on vm %s after wakeup event",
                     vm->def->name);
        }
    }

1201
    virObjectUnlock(vm);
1202 1203
    qemuDomainEventQueue(driver, event);
    qemuDomainEventQueue(driver, lifecycleEvent);
1204
    virObjectUnref(cfg);
O
Osier Yang 已提交
1205 1206
    return 0;
}
1207

O
Osier Yang 已提交
1208 1209
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1210 1211
                           virDomainObjPtr vm,
                           void *opaque)
O
Osier Yang 已提交
1212
{
1213
    virQEMUDriverPtr driver = opaque;
1214 1215
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1216
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1217

1218
    virObjectLock(vm);
O
Osier Yang 已提交
1219 1220
    event = virDomainEventPMSuspendNewFromObj(vm);

1221
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1222
        qemuDomainObjPrivatePtr priv = vm->privateData;
1223 1224 1225 1226 1227
        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
1228
        lifecycleEvent =
1229
            virDomainEventLifecycleNewFromObj(vm,
J
Jiri Denemark 已提交
1230 1231
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1232

1233
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1234 1235 1236
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1237 1238 1239

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1240 1241
    }

1242
    virObjectUnlock(vm);
O
Osier Yang 已提交
1243

1244 1245
    qemuDomainEventQueue(driver, event);
    qemuDomainEventQueue(driver, lifecycleEvent);
1246
    virObjectUnref(cfg);
O
Osier Yang 已提交
1247 1248 1249
    return 0;
}

1250 1251 1252
static int
qemuProcessHandleBalloonChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1253 1254
                               unsigned long long actual,
                               void *opaque)
1255
{
1256
    virQEMUDriverPtr driver = opaque;
1257
    virObjectEventPtr event = NULL;
1258
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1259

1260
    virObjectLock(vm);
1261 1262 1263 1264 1265 1266
    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;

1267
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
1268 1269
        VIR_WARN("unable to save domain status with balloon change");

1270
    virObjectUnlock(vm);
1271

1272
    qemuDomainEventQueue(driver, event);
1273
    virObjectUnref(cfg);
1274 1275 1276
    return 0;
}

1277 1278
static int
qemuProcessHandlePMSuspendDisk(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1279 1280
                               virDomainObjPtr vm,
                               void *opaque)
1281
{
1282
    virQEMUDriverPtr driver = opaque;
1283 1284
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1285
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1286

1287
    virObjectLock(vm);
1288 1289 1290 1291 1292 1293 1294 1295 1296 1297
    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 =
1298
            virDomainEventLifecycleNewFromObj(vm,
1299 1300 1301
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_DISK);

1302
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0) {
1303 1304 1305 1306 1307 1308 1309 1310
            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);
    }

1311
    virObjectUnlock(vm);
1312

1313 1314
    qemuDomainEventQueue(driver, event);
    qemuDomainEventQueue(driver, lifecycleEvent);
1315 1316
    virObjectUnref(cfg);

1317 1318 1319
    return 0;
}

1320

1321 1322
static int
qemuProcessHandleGuestPanic(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1323
                            virDomainObjPtr vm,
1324
                            qemuMonitorEventPanicInfoPtr info,
1325
                            void *opaque)
1326
{
1327
    virQEMUDriverPtr driver = opaque;
1328 1329 1330
    struct qemuProcessEvent *processEvent;

    virObjectLock(vm);
1331
    if (VIR_ALLOC(processEvent) < 0)
1332 1333 1334 1335
        goto cleanup;

    processEvent->eventType = QEMU_PROCESS_EVENT_GUESTPANIC;
    processEvent->action = vm->def->onCrash;
1336
    processEvent->data = info;
1337 1338 1339
    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted before handling guest panic event is finished.
     */
1340 1341
    processEvent->vm = virObjectRef(vm);

1342 1343 1344
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        if (!virObjectUnref(vm))
            vm = NULL;
1345
        qemuProcessEventFree(processEvent);
1346 1347
    }

1348
 cleanup:
1349
    if (vm)
1350
        virObjectUnlock(vm);
1351 1352 1353 1354 1355

    return 0;
}


1356
int
1357 1358
qemuProcessHandleDeviceDeleted(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1359 1360
                               const char *devAlias,
                               void *opaque)
1361
{
1362
    virQEMUDriverPtr driver = opaque;
1363 1364
    struct qemuProcessEvent *processEvent = NULL;
    char *data;
1365 1366 1367 1368 1369 1370

    virObjectLock(vm);

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

1371 1372
    if (qemuDomainSignalDeviceRemoval(vm, devAlias,
                                      QEMU_DOMAIN_UNPLUGGING_DEVICE_STATUS_OK))
1373
        goto cleanup;
1374

1375 1376
    if (VIR_ALLOC(processEvent) < 0)
        goto error;
1377

1378 1379 1380 1381
    processEvent->eventType = QEMU_PROCESS_EVENT_DEVICE_DELETED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
1382
    processEvent->vm = virObjectRef(vm);
1383

1384 1385 1386 1387
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }
1388

1389
 cleanup:
1390 1391
    virObjectUnlock(vm);
    return 0;
1392
 error:
1393
    qemuProcessEventFree(processEvent);
1394
    goto cleanup;
1395 1396 1397
}


1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465
/**
 *
 * 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);
    qemuDomainEventQueue(driver, event);

    return 0;
}


1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
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);
    qemuDomainEventQueue(driver, event);

    return 0;
}


1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
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;
1528
    processEvent->vm = virObjectRef(vm);
1529 1530 1531 1532 1533 1534 1535 1536 1537 1538

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

 cleanup:
    virObjectUnlock(vm);
    return 0;
 error:
1539
    qemuProcessEventFree(processEvent);
1540 1541 1542 1543
    goto cleanup;
}


1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567
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;
1568
    processEvent->vm = virObjectRef(vm);
1569 1570 1571 1572 1573 1574 1575 1576 1577 1578

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

 cleanup:
    virObjectUnlock(vm);
    return 0;
 error:
1579
    qemuProcessEventFree(processEvent);
1580 1581 1582 1583
    goto cleanup;
}


1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602
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;
1603
    virDomainObjBroadcast(vm);
1604 1605 1606 1607 1608 1609 1610

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625
static int
qemuProcessHandleMigrationStatus(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                 virDomainObjPtr vm,
                                 int status,
                                 void *opaque ATTRIBUTE_UNUSED)
{
    qemuDomainObjPrivatePtr priv;

    virObjectLock(vm);

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

    priv = vm->privateData;
1626
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_NONE) {
1627 1628 1629 1630
        VIR_DEBUG("got MIGRATION event without a migration job");
        goto cleanup;
    }

1631
    priv->job.current->stats.mig.status = status;
1632 1633 1634 1635 1636 1637 1638 1639
    virDomainObjBroadcast(vm);

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668
static int
qemuProcessHandleMigrationPass(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               int pass,
                               void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    qemuDomainObjPrivatePtr priv;

    virObjectLock(vm);

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

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

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

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


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
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;
}


1708 1709
static qemuMonitorCallbacks monitorCallbacks = {
    .eofNotify = qemuProcessHandleMonitorEOF,
1710
    .errorNotify = qemuProcessHandleMonitorError,
1711
    .diskSecretLookup = qemuProcessFindVolumeQcowPassphrase,
1712
    .domainEvent = qemuProcessHandleEvent,
1713 1714
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
1715
    .domainResume = qemuProcessHandleResume,
1716 1717 1718 1719 1720
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1721
    .domainBlockJob = qemuProcessHandleBlockJob,
1722
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1723
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1724
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1725
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1726
    .domainPMSuspendDisk = qemuProcessHandlePMSuspendDisk,
1727
    .domainGuestPanic = qemuProcessHandleGuestPanic,
1728
    .domainDeviceDeleted = qemuProcessHandleDeviceDeleted,
1729
    .domainNicRxFilterChanged = qemuProcessHandleNicRxFilterChanged,
1730
    .domainSerialChange = qemuProcessHandleSerialChanged,
1731
    .domainSpiceMigrated = qemuProcessHandleSpiceMigrated,
1732
    .domainMigrationStatus = qemuProcessHandleMigrationStatus,
1733
    .domainMigrationPass = qemuProcessHandleMigrationPass,
1734
    .domainAcpiOstInfo = qemuProcessHandleAcpiOstInfo,
1735
    .domainBlockThreshold = qemuProcessHandleBlockThreshold,
1736
    .domainDumpCompleted = qemuProcessHandleDumpCompleted,
1737 1738
};

1739 1740 1741 1742 1743 1744 1745 1746 1747 1748
static void
qemuProcessMonitorReportLogError(qemuMonitorPtr mon,
                                 const char *msg,
                                 void *opaque);


static void
qemuProcessMonitorLogFree(void *opaque)
{
    qemuDomainLogContextPtr logCtxt = opaque;
1749
    virObjectUnref(logCtxt);
1750 1751
}

1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771

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;
}


1772
static int
1773
qemuConnectMonitor(virQEMUDriverPtr driver, virDomainObjPtr vm, int asyncJob,
1774
                   qemuDomainLogContextPtr logCtxt)
1775 1776
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
1777
    qemuMonitorPtr mon = NULL;
1778
    unsigned long long timeout = 0;
1779
    virDomainChrSourceDefPtr monConfig;
1780

1781
    if (qemuSecuritySetDaemonSocketLabel(driver->securityManager, vm->def) < 0) {
1782 1783
        VIR_ERROR(_("Failed to set security context for monitor for %s"),
                  vm->def->name);
1784
        return -1;
1785 1786
    }

1787 1788 1789 1790 1791 1792
    /* 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);

1793
    /* Hold an extra reference because we can't allow 'vm' to be
C
Chen Hanxiao 已提交
1794
     * deleted until the monitor gets its own reference. */
1795
    virObjectRef(vm);
1796

1797
    ignore_value(virTimeMillisNow(&priv->monStart));
1798 1799
    monConfig = priv->monConfig;
    virObjectRef(monConfig);
1800
    virObjectUnlock(vm);
1801 1802

    mon = qemuMonitorOpen(vm,
1803
                          monConfig,
1804
                          priv->monJSON,
1805
                          timeout,
1806 1807
                          &monitorCallbacks,
                          driver);
1808

1809
    if (mon && logCtxt) {
1810
        virObjectRef(logCtxt);
1811 1812 1813 1814 1815
        qemuMonitorSetDomainLog(mon,
                                qemuProcessMonitorReportLogError,
                                logCtxt,
                                qemuProcessMonitorLogFree);
    }
1816

1817
    virObjectLock(vm);
1818
    virObjectUnref(monConfig);
M
Michal Privoznik 已提交
1819
    virObjectUnref(vm);
1820
    priv->monStart = 0;
1821

M
Michal Privoznik 已提交
1822
    if (!virDomainObjIsActive(vm)) {
1823
        qemuMonitorClose(mon);
1824
        mon = NULL;
1825 1826 1827
    }
    priv->mon = mon;

1828
    if (qemuSecurityClearSocketLabel(driver->securityManager, vm->def) < 0) {
1829 1830
        VIR_ERROR(_("Failed to clear security context for monitor for %s"),
                  vm->def->name);
1831
        return -1;
1832 1833 1834 1835
    }

    if (priv->mon == NULL) {
        VIR_INFO("Failed to connect monitor for %s", vm->def->name);
1836
        return -1;
1837 1838
    }

1839 1840
    if (qemuProcessInitMonitor(driver, vm, asyncJob) < 0)
        return -1;
1841

1842
    if (qemuDomainCheckMigrationCapabilities(driver, vm, asyncJob) < 0)
1843 1844 1845
        return -1;

    return 0;
1846 1847
}

1848 1849 1850

/**
 * qemuProcessReadLog: Read log file of a qemu VM
1851
 * @logCtxt: the domain log context
1852
 * @msg: pointer to buffer to store the read messages in
1853
 * @max: maximum length of the message returned in @msg
1854 1855
 *
 * Reads log of a qemu VM. Skips messages not produced by qemu or irrelevant
1856 1857 1858 1859
 * 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
1860
 */
1861
static int
1862 1863 1864
qemuProcessReadLog(qemuDomainLogContextPtr logCtxt,
                   char **msg,
                   size_t max)
1865
{
1866 1867
    char *buf;
    ssize_t got;
1868
    char *eol;
1869
    char *filter_next;
1870
    size_t skip;
1871

1872
    if ((got = qemuDomainLogContextRead(logCtxt, &buf)) < 0)
1873
        return -1;
1874

1875 1876 1877 1878 1879
    /* Filter out debug messages from intermediate libvirt process */
    filter_next = buf;
    while ((eol = strchr(filter_next, '\n'))) {
        *eol = '\0';
        if (virLogProbablyLogMessage(filter_next) ||
1880
            strstr(filter_next, "char device redirected to")) {
1881
            skip = (eol + 1) - filter_next;
1882
            memmove(filter_next, eol + 1, buf + got - eol);
1883 1884 1885 1886
            got -= skip;
        } else {
            filter_next = eol + 1;
            *eol = '\n';
1887 1888
        }
    }
1889
    filter_next = NULL; /* silence false coverity warning */
1890

1891 1892
    if (got > 0 &&
        buf[got - 1] == '\n') {
1893 1894
        buf[got - 1] = '\0';
        got--;
1895
    }
1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908

    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;
    }

1909
    ignore_value(VIR_REALLOC_N_QUIET(buf, got + 1));
1910 1911 1912
    *msg = buf;
    return 0;
}
1913 1914


1915 1916
static int
qemuProcessReportLogError(qemuDomainLogContextPtr logCtxt,
1917 1918 1919
                          const char *msgprefix)
{
    char *logmsg = NULL;
1920 1921 1922 1923 1924 1925
    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;
1926

1927
    if (qemuProcessReadLog(logCtxt, &logmsg, max) < 0)
1928 1929 1930
        return -1;

    virResetLastError();
1931 1932 1933 1934 1935
    if (virStringIsEmpty(logmsg))
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s", msgprefix);
    else
        virReportError(VIR_ERR_INTERNAL_ERROR, _("%s: %s"), msgprefix, logmsg);

1936 1937
    VIR_FREE(logmsg);
    return 0;
1938 1939 1940
}


1941 1942 1943 1944 1945 1946 1947 1948 1949 1950
static void
qemuProcessMonitorReportLogError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                 const char *msg,
                                 void *opaque)
{
    qemuDomainLogContextPtr logCtxt = opaque;
    qemuProcessReportLogError(logCtxt, msg);
}


1951
static int
1952
qemuProcessLookupPTYs(virDomainChrDefPtr *devices,
1953
                      int count,
1954
                      virHashTablePtr info)
1955
{
1956
    char *id = NULL;
1957
    size_t i;
1958
    int ret = -1;
1959

1960
    for (i = 0; i < count; i++) {
1961
        virDomainChrDefPtr chr = devices[i];
1962
        if (chr->source->type == VIR_DOMAIN_CHR_TYPE_PTY) {
1963
            qemuMonitorChardevInfoPtr entry;
1964

1965 1966
            VIR_FREE(id);
            if (virAsprintf(&id, "char%s", chr->info.alias) < 0)
1967 1968
                return -1;

1969 1970
            entry = virHashLookup(info, id);
            if (!entry || !entry->ptyPath) {
1971
                if (chr->source->data.file.path == NULL) {
1972 1973 1974
                    /* neither the log output nor 'info chardev' had a
                     * pty path for this chardev, report an error
                     */
1975 1976
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1977
                    goto cleanup;
1978 1979 1980 1981 1982 1983 1984 1985
                } else {
                    /* 'info chardev' had no pty path for this chardev,
                     * but the log output had, so we're fine
                     */
                    continue;
                }
            }

1986 1987
            VIR_FREE(chr->source->data.file.path);
            if (VIR_STRDUP(chr->source->data.file.path, entry->ptyPath) < 0)
1988
                goto cleanup;
1989 1990 1991
        }
    }

1992 1993 1994 1995
    ret = 0;
 cleanup:
    VIR_FREE(id);
    return ret;
1996 1997
}

1998 1999
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
2000
                                     virHashTablePtr info)
2001
{
2002
    size_t i = 0;
C
Cole Robinson 已提交
2003

2004
    if (qemuProcessLookupPTYs(vm->def->serials, vm->def->nserials, info) < 0)
2005 2006
        return -1;

2007
    if (qemuProcessLookupPTYs(vm->def->parallels, vm->def->nparallels,
2008
                              info) < 0)
2009
        return -1;
2010

2011
    if (qemuProcessLookupPTYs(vm->def->channels, vm->def->nchannels, info) < 0)
2012
        return -1;
2013 2014 2015 2016
    /* 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];
2017

2018 2019 2020 2021 2022
        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;
2023 2024
            if (virDomainChrSourceDefCopy(chr->source,
                                          ((vm->def->serials[0])->source)) < 0)
2025 2026 2027 2028
                return -1;
        }
    }

2029
    if (qemuProcessLookupPTYs(vm->def->consoles + i, vm->def->nconsoles - i,
2030
                              info) < 0)
2031
        return -1;
2032 2033 2034 2035 2036

    return 0;
}


2037
static int
2038 2039 2040 2041
qemuProcessRefreshChannelVirtioState(virQEMUDriverPtr driver,
                                     virDomainObjPtr vm,
                                     virHashTablePtr info,
                                     int booted)
2042 2043
{
    size_t i;
2044
    int agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_CHANNEL;
2045
    qemuMonitorChardevInfoPtr entry;
2046
    virObjectEventPtr event = NULL;
2047 2048
    char *id = NULL;
    int ret = -1;
2049

2050 2051 2052
    if (booted)
        agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_DOMAIN_STARTED;

2053 2054 2055
    for (i = 0; i < vm->def->nchannels; i++) {
        virDomainChrDefPtr chr = vm->def->channels[i];
        if (chr->targetType == VIR_DOMAIN_CHR_CHANNEL_TARGET_TYPE_VIRTIO) {
2056 2057 2058 2059

            VIR_FREE(id);
            if (virAsprintf(&id, "char%s", chr->info.alias) < 0)
                goto cleanup;
2060 2061 2062 2063 2064 2065

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

2066 2067 2068 2069 2070 2071
            if (entry->state != VIR_DOMAIN_CHR_DEVICE_STATE_DEFAULT &&
                STREQ_NULLABLE(chr->target.name, "org.qemu.guest_agent.0") &&
                (event = virDomainEventAgentLifecycleNewFromObj(vm, entry->state,
                                                                agentReason)))
                qemuDomainEventQueue(driver, event);

2072 2073 2074 2075
            chr->state = entry->state;
        }
    }

2076 2077 2078 2079
    ret = 0;
 cleanup:
    VIR_FREE(id);
    return ret;
2080 2081 2082
}


2083 2084
int
qemuRefreshVirtioChannelState(virQEMUDriverPtr driver,
2085 2086
                              virDomainObjPtr vm,
                              qemuDomainAsyncJob asyncJob)
2087 2088 2089 2090 2091
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virHashTablePtr info = NULL;
    int ret = -1;

2092 2093 2094
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;

2095
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
2096 2097
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2098 2099 2100 2101

    if (ret < 0)
        goto cleanup;

2102
    ret = qemuProcessRefreshChannelVirtioState(driver, vm, info, false);
2103 2104 2105 2106 2107 2108

 cleanup:
    virHashFree(info);
    return ret;
}

2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144
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;
    if ((then = mktime(&thenbits)) == (time_t) -1) {
        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;
}
2145

2146
int
2147 2148 2149 2150 2151 2152 2153 2154 2155
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 */
2156
    if (!virDomainDefHasMemballoon(vm->def)) {
2157
        vm->def->mem.cur_balloon = virDomainDefGetMemoryTotal(vm->def);
2158 2159 2160 2161 2162 2163 2164
        return 0;
    }

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

    rc = qemuMonitorGetBalloonInfo(qemuDomainGetMonitor(vm), &balloon);
2165
    if (qemuDomainObjExitMonitor(driver, vm) < 0 || rc < 0)
2166 2167 2168 2169 2170 2171 2172 2173
        return -1;

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

    return 0;
}


2174
static int
2175
qemuProcessWaitForMonitor(virQEMUDriverPtr driver,
C
Cole Robinson 已提交
2176
                          virDomainObjPtr vm,
2177
                          int asyncJob,
2178
                          qemuDomainLogContextPtr logCtxt)
2179 2180
{
    int ret = -1;
2181
    virHashTablePtr info = NULL;
2182
    qemuDomainObjPrivatePtr priv;
2183 2184

    VIR_DEBUG("Connect monitor to %p '%s'", vm, vm->def->name);
2185
    if (qemuConnectMonitor(driver, vm, asyncJob, logCtxt) < 0)
2186 2187 2188 2189 2190 2191
        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. */
2192
    priv = vm->privateData;
2193 2194
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
2195 2196
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
    VIR_DEBUG("qemuMonitorGetChardevInfo returned %i", ret);
2197 2198 2199
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

2200
    if (ret == 0) {
2201
        if ((ret = qemuProcessFindCharDevicePTYsMonitor(vm, info)) < 0)
2202 2203
            goto cleanup;

2204 2205
        if ((ret = qemuProcessRefreshChannelVirtioState(driver, vm, info,
                                                        true)) < 0)
2206 2207
            goto cleanup;
    }
2208

2209
 cleanup:
2210
    virHashFree(info);
2211

2212 2213
    if (logCtxt && kill(vm->pid, 0) == -1 && errno == ESRCH) {
        qemuProcessReportLogError(logCtxt,
2214
                                  _("process exited while connecting to monitor"));
2215 2216 2217 2218 2219 2220
        ret = -1;
    }

    return ret;
}

2221

2222 2223 2224 2225 2226 2227
static int
qemuProcessDetectIOThreadPIDs(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2228
    qemuMonitorIOThreadInfoPtr *iothreads = NULL;
2229 2230 2231 2232
    int niothreads = 0;
    int ret = -1;
    size_t i;

2233 2234 2235 2236 2237
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_OBJECT_IOTHREAD)) {
        ret = 0;
        goto cleanup;
    }

2238 2239 2240 2241
    /* Get the list of IOThreads from qemu */
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
    niothreads = qemuMonitorGetIOThreads(priv->mon, &iothreads);
2242 2243
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
J
John Ferlan 已提交
2244
    if (niothreads < 0)
2245 2246
        goto cleanup;

2247
    if (niothreads != vm->def->niothreadids) {
2248 2249
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of IOThread pids from QEMU monitor. "
2250 2251
                         "got %d, wanted %zu"),
                       niothreads, vm->def->niothreadids);
2252 2253 2254
        goto cleanup;
    }

2255 2256 2257 2258 2259 2260
    /* Nothing to do */
    if (niothreads == 0) {
        ret = 0;
        goto cleanup;
    }

2261 2262 2263
    for (i = 0; i < niothreads; i++) {
        virDomainIOThreadIDDefPtr iothrid;

2264 2265
        if (!(iothrid = virDomainIOThreadIDFind(vm->def,
                                                iothreads[i]->iothread_id))) {
2266
            virReportError(VIR_ERR_INTERNAL_ERROR,
2267 2268
                           _("iothread %d not found"),
                           iothreads[i]->iothread_id);
2269 2270 2271 2272
            goto cleanup;
        }
        iothrid->thread_id = iothreads[i]->thread_id;
    }
2273 2274 2275 2276 2277 2278

    ret = 0;

 cleanup:
    if (iothreads) {
        for (i = 0; i < niothreads; i++)
2279
            VIR_FREE(iothreads[i]);
2280 2281 2282 2283 2284
        VIR_FREE(iothreads);
    }
    return ret;
}

2285 2286 2287 2288 2289

/*
 * To be run between fork/exec of QEMU only
 */
static int
2290
qemuProcessInitCpuAffinity(virDomainObjPtr vm)
2291 2292 2293 2294
{
    int ret = -1;
    virBitmapPtr cpumap = NULL;
    virBitmapPtr cpumapToSet = NULL;
2295
    virBitmapPtr hostcpumap = NULL;
2296
    qemuDomainObjPrivatePtr priv = vm->privateData;
2297

2298 2299 2300 2301 2302 2303
    if (!vm->pid) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Cannot setup CPU affinity until process is started"));
        return -1;
    }

2304 2305
    if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) {
        VIR_DEBUG("Set CPU affinity with advisory nodeset from numad");
2306
        cpumapToSet = priv->autoCpuset;
2307
    } else {
2308
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
2309
        if (vm->def->cpumask) {
H
Hu Tao 已提交
2310
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
2311 2312 2313 2314 2315
        } 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 */
2316 2317
            int hostcpus;

2318 2319 2320 2321 2322 2323 2324 2325 2326 2327
            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;
2328
                goto cleanup;
2329 2330 2331 2332 2333
            } 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;
2334

2335 2336
                if (hostcpus > QEMUD_CPUMASK_LEN)
                    hostcpus = QEMUD_CPUMASK_LEN;
2337

2338 2339 2340
                virBitmapFree(cpumap);
                if (!(cpumap = virBitmapNew(hostcpus)))
                    goto cleanup;
2341

2342
                virBitmapSetAll(cpumap);
2343

2344 2345
                cpumapToSet = cpumap;
            }
O
Osier Yang 已提交
2346
        }
2347 2348
    }

2349
    if (virProcessSetAffinity(vm->pid, cpumapToSet) < 0)
2350
        goto cleanup;
2351

2352 2353
    ret = 0;

2354
 cleanup:
2355
    virBitmapFree(cpumap);
2356
    virBitmapFree(hostcpumap);
2357
    return ret;
2358 2359
}

2360 2361
/* set link states to down on interfaces at qemu start */
static int
2362 2363 2364
qemuProcessSetLinkStates(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob)
2365 2366 2367
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2368
    size_t i;
2369 2370 2371 2372 2373
    int ret = -1;
    int rv;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2374 2375 2376

    for (i = 0; i < def->nnets; i++) {
        if (def->nets[i]->linkstate == VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN) {
2377 2378 2379
            if (!def->nets[i]->info.alias) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("missing alias for network device"));
2380
                goto cleanup;
2381 2382
            }

2383 2384
            VIR_DEBUG("Setting link state: %s", def->nets[i]->info.alias);

2385 2386 2387 2388
            rv = qemuMonitorSetLink(priv->mon,
                                    def->nets[i]->info.alias,
                                    VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (rv < 0) {
2389
                virReportError(VIR_ERR_OPERATION_FAILED,
2390 2391 2392
                               _("Couldn't set link state on interface: %s"),
                               def->nets[i]->info.alias);
                goto cleanup;
2393 2394 2395 2396
            }
        }
    }

2397 2398 2399 2400 2401
    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2402 2403 2404
    return ret;
}

2405

2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513
/**
 * qemuProcessSetupPid:
 *
 * This function sets resource properities (affinity, cgroups,
 * scheduler) for any PID associated with a domain.  It should be used
 * to set up emulator PIDs as well as vCPU and I/O thread pids to
 * ensure they are all handled the same way.
 *
 * Returns 0 on success, -1 on error.
 */
static int
qemuProcessSetupPid(virDomainObjPtr vm,
                    pid_t pid,
                    virCgroupThreadName nameval,
                    int id,
                    virBitmapPtr cpumask,
                    unsigned long long period,
                    long long quota,
                    virDomainThreadSchedParamPtr sched)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainNumatuneMemMode mem_mode;
    virCgroupPtr cgroup = NULL;
    virBitmapPtr use_cpumask;
    char *mem_mask = NULL;
    int ret = -1;

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

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

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

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

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

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

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

        }

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

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

    }

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

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

    ret = 0;
 cleanup:
    VIR_FREE(mem_mask);
    if (cgroup) {
        if (ret < 0)
            virCgroupRemove(cgroup);
        virCgroupFree(&cgroup);
    }

    return ret;
}


2514
static int
2515
qemuProcessSetupEmulator(virDomainObjPtr vm)
2516
{
2517 2518 2519 2520 2521
    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);
2522 2523
}

2524

2525 2526 2527 2528 2529 2530
static int
qemuProcessResctrlCreate(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    int ret = -1;
    size_t i = 0;
2531
    virCapsPtr caps = NULL;
2532 2533
    qemuDomainObjPrivatePtr priv = vm->privateData;

2534 2535 2536 2537 2538 2539
    if (!vm->def->ncachetunes)
        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);
2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556
    if (!caps)
        return -1;

    for (i = 0; i < vm->def->ncachetunes; i++) {
        if (virResctrlAllocCreate(caps->host.resctrl,
                                  vm->def->cachetunes[i]->alloc,
                                  priv->machineName) < 0)
            goto cleanup;
    }

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


2557
static int
2558
qemuProcessInitPasswords(virQEMUDriverPtr driver,
2559 2560
                         virDomainObjPtr vm,
                         int asyncJob)
2561 2562 2563
{
    int ret = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
2564
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2565
    size_t i;
2566 2567
    char *alias = NULL;
    char *secret = NULL;
2568

2569
    for (i = 0; i < vm->def->ngraphics; ++i) {
2570 2571
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
2572 2573
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_VNC,
2574
                                                    &graphics->data.vnc.auth,
2575 2576
                                                    cfg->vncPassword,
                                                    asyncJob);
2577
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
2578 2579
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_SPICE,
2580
                                                    &graphics->data.spice.auth,
2581 2582
                                                    cfg->spicePassword,
                                                    asyncJob);
2583 2584
        }

2585 2586 2587
        if (ret < 0)
            goto cleanup;
    }
2588

2589 2590
    for (i = 0; i < vm->def->ndisks; i++) {
        size_t secretLen;
2591

2592 2593 2594
        if (!vm->def->disks[i]->src->encryption ||
            !virDomainDiskGetSource(vm->def->disks[i]))
            continue;
2595

2596 2597 2598 2599 2600 2601
        if (vm->def->disks[i]->src->encryption->format !=
            VIR_STORAGE_ENCRYPTION_FORMAT_DEFAULT &&
            vm->def->disks[i]->src->encryption->format !=
            VIR_STORAGE_ENCRYPTION_FORMAT_QCOW)
            continue;

2602
        VIR_FREE(secret);
2603
        if (qemuProcessGetVolumeQcowPassphrase(vm->def->disks[i],
2604 2605
                                               &secret, &secretLen) < 0)
            goto cleanup;
2606

2607
        VIR_FREE(alias);
2608
        if (!(alias = qemuAliasFromDisk(vm->def->disks[i])))
2609 2610 2611 2612 2613 2614 2615 2616
            goto cleanup;
        if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
            goto cleanup;
        ret = qemuMonitorSetDrivePassphrase(priv->mon, alias, secret);
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            ret = -1;
        if (ret < 0)
            goto cleanup;
2617 2618
    }

2619
 cleanup:
2620 2621
    VIR_FREE(alias);
    VIR_FREE(secret);
2622
    virObjectUnref(cfg);
2623 2624 2625 2626 2627 2628 2629 2630 2631 2632
    return ret;
}


static int
qemuProcessPrepareChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
    int fd;
2633
    if (dev->source->type != VIR_DOMAIN_CHR_TYPE_FILE)
2634 2635
        return 0;

2636
    if ((fd = open(dev->source->data.file.path,
2637 2638 2639
                   O_CREAT | O_APPEND, S_IRUSR|S_IWUSR)) < 0) {
        virReportSystemError(errno,
                             _("Unable to pre-create chardev file '%s'"),
2640
                             dev->source->data.file.path);
2641 2642 2643 2644 2645 2646 2647 2648 2649
        return -1;
    }

    VIR_FORCE_CLOSE(fd);

    return 0;
}


2650 2651 2652 2653 2654
static int
qemuProcessCleanupChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
2655 2656 2657 2658
    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);
2659 2660 2661 2662 2663

    return 0;
}


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
/**
 * 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) {
2700
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VGAMEM) &&
2701 2702
                    qemuMonitorUpdateVideoMemorySize(priv->mon, video,
                                                     "qxl-vga") < 0)
2703
                        goto error;
2704

2705
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VRAM64) &&
2706 2707 2708
                    qemuMonitorUpdateVideoVram64Size(priv->mon, video,
                                                     "qxl-vga") < 0)
                    goto error;
2709
            } else {
2710 2711 2712 2713 2714 2715 2716 2717
                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)
2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736
                        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;
        }

    }

2737 2738
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2739 2740

    cfg = virQEMUDriverGetConfig(driver);
2741
    ret = virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps);
2742 2743 2744 2745 2746
    virObjectUnref(cfg);

    return ret;

 error:
2747
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
2748 2749 2750 2751
    return -1;
}


2752 2753
struct qemuProcessHookData {
    virDomainObjPtr vm;
2754
    virQEMUDriverPtr driver;
2755
    virQEMUDriverConfigPtr cfg;
2756 2757 2758 2759 2760
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
2761
    qemuDomainObjPrivatePtr priv = h->vm->privateData;
2762
    int ret = -1;
2763
    int fd;
2764 2765 2766
    virBitmapPtr nodeset = NULL;
    virDomainNumatuneMemMode mode;

2767 2768 2769 2770
    /* This method cannot use any mutexes, which are not
     * protected across fork()
     */

2771
    qemuSecurityPostFork(h->driver->securityManager);
2772 2773 2774 2775 2776

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

    VIR_DEBUG("Obtaining domain lock");
2777 2778 2779 2780 2781 2782 2783
    /*
     * 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.
     */
2784
    if (qemuSecuritySetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2785
        goto cleanup;
2786
    if (virDomainLockProcessStart(h->driver->lockManager,
2787
                                  h->cfg->uri,
2788
                                  h->vm,
J
Ján Tomko 已提交
2789
                                  /* QEMU is always paused initially */
2790 2791
                                  true,
                                  &fd) < 0)
2792
        goto cleanup;
2793
    if (qemuSecurityClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2794
        goto cleanup;
2795

2796
    if (qemuDomainBuildNamespace(h->cfg, h->driver->securityManager, h->vm) < 0)
2797 2798
        goto cleanup;

2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809
    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);
2810

2811 2812 2813
            if (virNumaSetupMemoryPolicy(mode, nodeset) < 0)
                goto cleanup;
        }
2814
    }
2815

2816 2817
    ret = 0;

2818
 cleanup:
2819
    virObjectUnref(h->cfg);
2820 2821
    VIR_DEBUG("Hook complete ret=%d", ret);
    return ret;
2822 2823 2824
}

int
2825 2826
qemuProcessPrepareMonitorChr(virDomainChrSourceDefPtr monConfig,
                             const char *domainDir)
2827 2828 2829 2830
{
    monConfig->type = VIR_DOMAIN_CHR_TYPE_UNIX;
    monConfig->data.nix.listen = true;

2831 2832
    if (virAsprintf(&monConfig->data.nix.path, "%s/monitor.sock",
                    domainDir) < 0)
2833 2834
        return -1;
    return 0;
2835 2836 2837
}


2838
/*
2839 2840
 * Precondition: vm must be locked, and a job must be active.
 * This method will call {Enter,Exit}Monitor
2841
 */
E
Eric Blake 已提交
2842
int
2843
qemuProcessStartCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
2844
                     virDomainRunningReason reason,
2845
                     qemuDomainAsyncJob asyncJob)
2846
{
2847
    int ret = -1;
2848
    qemuDomainObjPrivatePtr priv = vm->privateData;
2849
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2850

2851
    /* Bring up netdevs before starting CPUs */
2852
    if (qemuInterfaceStartDevices(vm->def) < 0)
2853 2854
       goto cleanup;

2855
    VIR_DEBUG("Using lock state '%s'", NULLSTR(priv->lockState));
2856
    if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
2857
                                   vm, priv->lockState) < 0) {
2858 2859 2860 2861
        /* 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
         */
2862
        goto cleanup;
2863 2864 2865
    }
    VIR_FREE(priv->lockState);

2866 2867
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto release;
J
Jiri Denemark 已提交
2868

2869
    ret = qemuMonitorStartCPUs(priv->mon);
2870 2871
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2872 2873 2874 2875 2876

    if (ret < 0)
        goto release;

    virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
2877

2878
 cleanup:
2879
    virObjectUnref(cfg);
2880
    return ret;
2881 2882 2883 2884 2885 2886

 release:
    if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
        VIR_WARN("Unable to release lease on %s", vm->def->name);
    VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));
    goto cleanup;
2887 2888 2889
}


2890 2891
int qemuProcessStopCPUs(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
2892
                        virDomainPausedReason reason,
2893
                        qemuDomainAsyncJob asyncJob)
2894
{
2895
    int ret = -1;
2896 2897
    qemuDomainObjPrivatePtr priv = vm->privateData;

2898
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
2899

2900 2901
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
J
Jiri Denemark 已提交
2902

2903
    ret = qemuMonitorStopCPUs(priv->mon);
2904 2905
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2906 2907 2908 2909

    if (ret < 0)
        goto cleanup;

2910 2911 2912
    /* de-activate netdevs after stopping CPUs */
    ignore_value(qemuInterfaceStopDevices(vm->def));

2913 2914 2915
    if (priv->job.current)
        ignore_value(virTimeMillisNow(&priv->job.current->stopped));

2916 2917 2918 2919
    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
2920

2921
 cleanup:
2922 2923 2924 2925 2926
    return ret;
}



2927
static void
2928 2929
qemuProcessNotifyNets(virDomainDefPtr def)
{
2930
    size_t i;
2931

2932 2933
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
2934 2935 2936 2937 2938 2939 2940 2941
        /* 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));

2942
        virDomainNetNotifyActualDevice(def, net);
2943 2944 2945
    }
}

2946
static int
2947
qemuProcessFiltersInstantiate(virDomainDefPtr def)
2948
{
2949
    size_t i;
2950

2951
    for (i = 0; i < def->nnets; i++) {
2952 2953
        virDomainNetDefPtr net = def->nets[i];
        if ((net->filter) && (net->ifname)) {
2954
            if (virDomainConfNWFilterInstantiate(def->uuid, net) < 0)
J
Ján Tomko 已提交
2955
                return 1;
2956 2957 2958
        }
    }

J
Ján Tomko 已提交
2959
    return 0;
2960 2961
}

2962
static int
2963
qemuProcessUpdateState(virQEMUDriverPtr driver, virDomainObjPtr vm)
2964 2965 2966
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
2967
    virDomainPausedReason reason;
2968
    virDomainState newState = VIR_DOMAIN_NOSTATE;
2969
    int oldReason;
2970
    int newReason;
2971
    bool running;
2972
    char *msg = NULL;
2973 2974
    int ret;

2975
    qemuDomainObjEnterMonitor(driver, vm);
2976
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
2977 2978
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2979

2980
    if (ret < 0)
2981 2982
        return -1;

2983
    state = virDomainObjGetState(vm, &oldReason);
2984

2985 2986 2987 2988 2989 2990 2991 2992
    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) {
2993 2994
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
2995
        ignore_value(VIR_STRDUP_QUIET(msg, "was unpaused"));
2996
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
2997 2998 2999
        if (reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN) {
            newState = VIR_DOMAIN_SHUTDOWN;
            newReason = VIR_DOMAIN_SHUTDOWN_UNKNOWN;
3000
            ignore_value(VIR_STRDUP_QUIET(msg, "shutdown"));
3001
        } else if (reason == VIR_DOMAIN_PAUSED_CRASHED) {
3002 3003 3004
            newState = VIR_DOMAIN_CRASHED;
            newReason = VIR_DOMAIN_CRASHED_PANICKED;
            ignore_value(VIR_STRDUP_QUIET(msg, "crashed"));
3005 3006 3007
        } else {
            newState = VIR_DOMAIN_PAUSED;
            newReason = reason;
S
Stefan Berger 已提交
3008 3009
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
3010 3011 3012 3013 3014 3015 3016
        }
    }

    if (newState != VIR_DOMAIN_NOSTATE) {
        VIR_DEBUG("Domain %s %s while its monitor was disconnected;"
                  " changing state to %s (%s)",
                  vm->def->name,
3017
                  NULLSTR(msg),
3018 3019 3020 3021
                  virDomainStateTypeToString(newState),
                  virDomainStateReasonToString(newState, newReason));
        VIR_FREE(msg);
        virDomainObjSetState(vm, newState, newReason);
3022 3023 3024 3025 3026
    }

    return 0;
}

3027
static int
3028 3029 3030 3031
qemuProcessRecoverMigrationIn(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              qemuMigrationJobPhase phase,
                              virDomainState state,
3032
                              int reason)
3033
{
3034 3035 3036 3037 3038
    bool postcopy = (state == VIR_DOMAIN_PAUSED &&
                     reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED) ||
                    (state == VIR_DOMAIN_RUNNING &&
                     reason == VIR_DOMAIN_RUNNING_POSTCOPY);

3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054
    switch (phase) {
    case QEMU_MIGRATION_PHASE_NONE:
    case QEMU_MIGRATION_PHASE_PERFORM2:
    case QEMU_MIGRATION_PHASE_BEGIN3:
    case QEMU_MIGRATION_PHASE_PERFORM3:
    case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
    case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
    case QEMU_MIGRATION_PHASE_CONFIRM3:
    case QEMU_MIGRATION_PHASE_LAST:
        /* 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;
3055

3056 3057 3058 3059 3060
    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);
3061
        if (qemuProcessStartCPUs(driver, vm,
3062 3063 3064 3065 3066 3067 3068 3069 3070
                                 VIR_DOMAIN_RUNNING_UNPAUSED,
                                 QEMU_ASYNC_JOB_NONE) < 0) {
            VIR_WARN("Could not resume domain %s", vm->def->name);
        }
        break;

    case QEMU_MIGRATION_PHASE_FINISH3:
        /* migration finished, we started resuming the domain but didn't
         * confirm success or failure yet; killing it seems safest unless
3071 3072
         * we already started guest CPUs or we were in post-copy mode */
        if (postcopy) {
3073
            qemuMigrationAnyPostcopyFailed(driver, vm);
3074
        } else if (state != VIR_DOMAIN_RUNNING) {
3075
            VIR_DEBUG("Killing migrated domain %s", vm->def->name);
3076
            return -1;
3077 3078 3079
        }
        break;
    }
3080

3081
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_NONE);
3082 3083
    return 0;
}
3084

3085 3086 3087 3088 3089
static int
qemuProcessRecoverMigrationOut(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
                               qemuMigrationJobPhase phase,
                               virDomainState state,
3090 3091
                               int reason,
                               unsigned int *stopFlags)
3092
{
3093 3094 3095
    bool postcopy = state == VIR_DOMAIN_PAUSED &&
                    (reason == VIR_DOMAIN_PAUSED_POSTCOPY ||
                     reason == VIR_DOMAIN_PAUSED_POSTCOPY_FAILED);
3096
    bool resume = false;
3097

3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114
    switch (phase) {
    case QEMU_MIGRATION_PHASE_NONE:
    case QEMU_MIGRATION_PHASE_PREPARE:
    case QEMU_MIGRATION_PHASE_FINISH2:
    case QEMU_MIGRATION_PHASE_FINISH3:
    case QEMU_MIGRATION_PHASE_LAST:
        /* 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
3115 3116 3117 3118
         * domain; however we can only do that before migration enters
         * post-copy mode
         */
        if (postcopy) {
3119
            qemuMigrationAnyPostcopyFailed(driver, vm);
3120 3121 3122
        } else {
            VIR_DEBUG("Cancelling unfinished migration of domain %s",
                      vm->def->name);
3123
            if (qemuMigrationSrcCancel(driver, vm) < 0) {
3124 3125 3126
                VIR_WARN("Could not cancel ongoing migration of domain %s",
                         vm->def->name);
            }
3127
            resume = true;
3128
        }
3129
        break;
3130

3131 3132
    case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
        /* migration finished but we didn't have a chance to get the result
3133 3134
         * 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
3135
         */
3136
        if (postcopy)
3137
            qemuMigrationAnyPostcopyFailed(driver, vm);
3138
        break;
3139

3140
    case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
3141 3142 3143 3144 3145
        /* 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) {
3146
            qemuMigrationAnyPostcopyFailed(driver, vm);
3147 3148 3149
        } else {
            VIR_DEBUG("Resuming domain %s after failed migration",
                      vm->def->name);
3150
            resume = true;
3151 3152
        }
        break;
3153

3154 3155
    case QEMU_MIGRATION_PHASE_CONFIRM3:
        /* migration completed, we need to kill the domain here */
3156
        *stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
3157 3158
        return -1;
    }
3159

3160 3161 3162 3163 3164 3165 3166
    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)) {
3167
            if (qemuProcessStartCPUs(driver, vm,
3168 3169 3170 3171
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
                VIR_WARN("Could not resume domain %s", vm->def->name);
            }
3172 3173
        }
    }
3174

3175
    qemuMigrationParamsReset(driver, vm, QEMU_ASYNC_JOB_NONE);
3176 3177 3178
    return 0;
}

3179
static int
3180
qemuProcessRecoverJob(virQEMUDriverPtr driver,
3181
                      virDomainObjPtr vm,
3182 3183
                      const struct qemuDomainJobObj *job,
                      unsigned int *stopFlags)
3184
{
3185
    qemuDomainObjPrivatePtr priv = vm->privateData;
3186 3187 3188 3189 3190 3191 3192
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
3193
        if (qemuProcessRecoverMigrationOut(driver, vm, job->phase,
3194
                                           state, reason, stopFlags) < 0)
3195 3196 3197
            return -1;
        break;

3198
    case QEMU_ASYNC_JOB_MIGRATION_IN:
3199
        if (qemuProcessRecoverMigrationIn(driver, vm, job->phase,
3200
                                          state, reason) < 0)
3201
            return -1;
3202 3203 3204 3205
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
3206
    case QEMU_ASYNC_JOB_SNAPSHOT:
3207
        qemuDomainObjEnterMonitor(driver, vm);
3208
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
3209 3210
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
3211
        /* resume the domain but only if it was paused as a result of
3212 3213 3214 3215 3216 3217 3218 3219 3220
         * 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 &&
3221 3222
               (reason == VIR_DOMAIN_PAUSED_SNAPSHOT ||
                reason == VIR_DOMAIN_PAUSED_MIGRATION)) ||
3223
              reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
3224
             if (qemuProcessStartCPUs(driver, vm,
3225 3226 3227 3228
                                      VIR_DOMAIN_RUNNING_UNPAUSED,
                                      QEMU_ASYNC_JOB_NONE) < 0) {
                 VIR_WARN("Could not resume domain '%s' after migration to file",
                          vm->def->name);
3229 3230 3231 3232
            }
        }
        break;

3233 3234 3235 3236
    case QEMU_ASYNC_JOB_START:
        /* Already handled in VIR_DOMAIN_PAUSED_STARTING_UP check. */
        break;

3237 3238 3239 3240 3241 3242 3243 3244
    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        break;
    }

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

3245 3246 3247 3248
    /* 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.
     */
3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268
    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;

3269
    case QEMU_JOB_MIGRATION_OP:
3270
    case QEMU_JOB_ABORT:
3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281
    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;
}

3282 3283 3284 3285 3286 3287
static int
qemuProcessUpdateDevices(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDeviceDef dev;
3288
    const char **qemuDevices;
3289 3290 3291 3292 3293 3294 3295 3296 3297
    char **old;
    char **tmp;
    int ret = -1;

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

    old = priv->qemuDevices;
    priv->qemuDevices = NULL;
3298
    if (qemuDomainUpdateDeviceList(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
3299 3300
        goto cleanup;

3301
    qemuDevices = (const char **) priv->qemuDevices;
3302 3303
    if ((tmp = old)) {
        while (*tmp) {
3304
            if (!virStringListHasString(qemuDevices, *tmp) &&
3305 3306 3307 3308
                virDomainDefFindDevice(vm->def, *tmp, &dev, false) == 0 &&
                qemuDomainRemoveDevice(driver, vm, &dev) < 0) {
                goto cleanup;
            }
3309 3310 3311 3312 3313
            tmp++;
        }
    }
    ret = 0;

3314
 cleanup:
3315
    virStringListFree(old);
3316 3317 3318
    return ret;
}

3319 3320 3321 3322 3323 3324 3325
static int
qemuDomainPerfRestart(virDomainObjPtr vm)
{
    size_t i;
    virDomainDefPtr def = vm->def;
    qemuDomainObjPrivatePtr priv = vm->privateData;

3326
    if (!(priv->perf = virPerfNew()))
3327 3328 3329
        return -1;

    for (i = 0; i < VIR_PERF_EVENT_LAST; i++) {
3330 3331
        if (def->perf.events[i] &&
            def->perf.events[i] == VIR_TRISTATE_BOOL_YES) {
3332 3333 3334

            /* Failure to re-enable the perf event should not be fatal */
            if (virPerfEventEnable(priv->perf, i, vm->pid) < 0)
3335
                def->perf.events[i] = VIR_TRISTATE_BOOL_NO;
3336 3337 3338 3339 3340 3341
        }
    }

    return 0;
}

3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364

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;
        }
    }
}


3365
static bool
3366 3367
qemuProcessNeedHugepagesPath(virDomainDefPtr def,
                             virDomainMemoryDefPtr mem)
3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386
{
    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;
    }

3387 3388 3389 3390 3391 3392
    if (mem &&
        mem->model == VIR_DOMAIN_MEMORY_MODEL_DIMM &&
        mem->pagesize &&
        mem->pagesize != system_pagesize)
        return true;

3393 3394 3395 3396
    return false;
}


3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426
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;
}


3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444
static int
qemuProcessBuildDestroyMemoryPathsImpl(virQEMUDriverPtr driver,
                                       virDomainDefPtr def,
                                       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;
        }

        if (qemuSecurityDomainSetPathLabel(driver->securityManager,
3445
                                           def, path, true) < 0) {
3446 3447 3448 3449 3450
            virReportError(VIR_ERR_INTERNAL_ERROR,
                            _("Unable to label %s"), path);
            return -1;
        }
    } else {
M
Michal Privoznik 已提交
3451 3452
        if (virFileDeleteTree(path) < 0)
            return -1;
3453 3454 3455 3456 3457 3458
    }

    return 0;
}


3459
int
3460 3461 3462 3463
qemuProcessBuildDestroyMemoryPaths(virQEMUDriverPtr driver,
                                   virDomainObjPtr vm,
                                   virDomainMemoryDefPtr mem,
                                   bool build)
3464 3465
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3466
    char *path = NULL;
3467
    size_t i;
3468 3469
    bool shouldBuildHP = false;
    bool shouldBuildMB = false;
3470 3471
    int ret = -1;

3472 3473 3474 3475
    if (build) {
        shouldBuildHP = qemuProcessNeedHugepagesPath(vm->def, mem);
        shouldBuildMB = qemuProcessNeedMemoryBackingPath(vm->def, mem);
    }
3476

3477
    if (!build || shouldBuildHP) {
3478
        for (i = 0; i < cfg->nhugetlbfs; i++) {
3479
            path = qemuGetDomainHugepagePath(vm->def, &cfg->hugetlbfs[i]);
3480

3481
            if (!path)
3482 3483
                goto cleanup;

3484
            if (qemuProcessBuildDestroyMemoryPathsImpl(driver, vm->def,
3485
                                                       path, build) < 0)
3486
                goto cleanup;
3487

3488
            VIR_FREE(path);
3489 3490 3491
        }
    }

3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502
    if (!build || shouldBuildMB) {
        if (qemuGetMemoryBackingDomainPath(vm->def, cfg, &path) < 0)
            goto cleanup;

        if (qemuProcessBuildDestroyMemoryPathsImpl(driver, vm->def,
                                                   path, build) < 0)
            goto cleanup;

        VIR_FREE(path);
    }

3503 3504
    ret = 0;
 cleanup:
3505
    VIR_FREE(path);
3506 3507 3508 3509 3510
    virObjectUnref(cfg);
    return ret;
}


3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536
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;
}


3537 3538 3539 3540
static int
qemuProcessVNCAllocatePorts(virQEMUDriverPtr driver,
                            virDomainGraphicsDefPtr graphics,
                            bool allocate)
3541
{
3542
    unsigned short port;
3543

3544 3545 3546
    if (!allocate) {
        if (graphics->data.vnc.autoport)
            graphics->data.vnc.port = 5900;
3547

3548 3549
        return 0;
    }
3550

3551 3552 3553 3554 3555
    if (graphics->data.vnc.autoport) {
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
            return -1;
        graphics->data.vnc.port = port;
    }
3556

3557 3558 3559 3560 3561 3562
    if (graphics->data.vnc.websocket == -1) {
        if (virPortAllocatorAcquire(driver->webSocketPorts, &port) < 0)
            return -1;
        graphics->data.vnc.websocket = port;
        graphics->data.vnc.websocketGenerated = true;
    }
3563

3564 3565
    return 0;
}
J
John Ferlan 已提交
3566

3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577
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;
3578

3579 3580
    bool needTLSPort = false;
    bool needPort = false;
3581

3582 3583 3584 3585 3586 3587 3588
    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;
3589

3590 3591 3592
            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
                needPort = true;
                break;
3593

3594 3595 3596 3597 3598 3599 3600 3601 3602
            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;
3603

3604 3605 3606
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
            needPort = true;
            break;
3607

3608 3609 3610 3611 3612 3613 3614
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
            if (cfg->spiceTLS)
                needTLSPort = true;
            needPort = true;
            break;
        }
    }
3615

3616 3617 3618
    if (!allocate) {
        if (needPort || graphics->data.spice.port == -1)
            graphics->data.spice.port = 5901;
3619

3620 3621
        if (needTLSPort || graphics->data.spice.tlsPort == -1)
            graphics->data.spice.tlsPort = 5902;
3622

3623 3624
        ret = 0;
        goto cleanup;
3625 3626
    }

3627 3628 3629
    if (needPort || graphics->data.spice.port == -1) {
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
            goto cleanup;
3630

3631 3632 3633 3634
        graphics->data.spice.port = port;

        if (!graphics->data.spice.autoport)
            graphics->data.spice.portReserved = true;
3635 3636
    }

3637 3638 3639 3640 3641 3642 3643
    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;
        }
3644

3645 3646
        if (virPortAllocatorAcquire(driver->remotePorts, &tlsPort) < 0)
            goto cleanup;
3647

3648
        graphics->data.spice.tlsPort = tlsPort;
3649

3650 3651
        if (!graphics->data.spice.autoport)
            graphics->data.spice.tlsPortReserved = true;
3652
    }
3653

3654
    ret = 0;
3655

3656 3657 3658 3659
 cleanup:
    virObjectUnref(cfg);
    return ret;
}
3660

3661

3662 3663 3664 3665 3666
static int
qemuValidateCpuCount(virDomainDefPtr def,
                     virQEMUCapsPtr qemuCaps)
{
    unsigned int maxCpus = virQEMUCapsGetMachineMaxCpus(qemuCaps, def->os.machine);
3667

3668 3669 3670 3671 3672
    if (virDomainDefGetVcpus(def) == 0) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Domain requires at least 1 vCPU"));
        return -1;
    }
3673

3674 3675 3676 3677 3678
    if (maxCpus > 0 && virDomainDefGetVcpusMax(def) > maxCpus) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Maximum CPUs greater than specified machine type limit"));
        return -1;
    }
3679

3680 3681
    return 0;
}
3682

3683

3684 3685 3686 3687 3688 3689 3690
static int
qemuProcessVerifyHypervFeatures(virDomainDefPtr def,
                                virCPUDataPtr cpu)
{
    char *cpuFeature;
    size_t i;
    int rc;
3691

3692 3693 3694 3695
    for (i = 0; i < VIR_DOMAIN_HYPERV_LAST; i++) {
        /* always supported string property */
        if (i == VIR_DOMAIN_HYPERV_VENDOR_ID)
            continue;
3696

3697 3698
        if (def->hyperv_features[i] != VIR_TRISTATE_SWITCH_ON)
            continue;
3699

3700 3701 3702
        if (virAsprintf(&cpuFeature, "__kvm_hv_%s",
                        virDomainHypervTypeToString(i)) < 0)
            return -1;
3703

3704 3705
        rc = virCPUDataCheckFeature(cpu, cpuFeature);
        VIR_FREE(cpuFeature);
3706

3707 3708 3709 3710
        if (rc < 0)
            return -1;
        else if (rc == 1)
            continue;
3711

3712 3713 3714 3715 3716 3717 3718
        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;
3719

3720 3721 3722 3723 3724 3725 3726 3727 3728
        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:
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("host doesn't support hyperv '%s' feature"),
                           virDomainHypervTypeToString(i));
            return -1;
3729

3730 3731 3732 3733
        /* coverity[dead_error_begin] */
        case VIR_DOMAIN_HYPERV_VENDOR_ID:
        case VIR_DOMAIN_HYPERV_LAST:
            break;
3734
        }
3735
    }
3736 3737

    return 0;
3738 3739
}

3740

3741
static int
3742 3743
qemuProcessVerifyKVMFeatures(virDomainDefPtr def,
                             virCPUDataPtr cpu)
3744
{
3745
    int rc = 0;
3746

3747
    if (def->features[VIR_DOMAIN_FEATURE_PVSPINLOCK] != VIR_TRISTATE_SWITCH_ON)
3748 3749
        return 0;

3750 3751 3752 3753 3754 3755
    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"));
3756
        return -1;
3757
    }
3758

3759 3760
    return 0;
}
3761 3762


3763 3764 3765 3766 3767
static int
qemuProcessVerifyCPUFeatures(virDomainDefPtr def,
                             virCPUDataPtr cpu)
{
    int rc;
S
Stefan Berger 已提交
3768

3769
    rc = virCPUCheckFeature(def->os.arch, def->cpu, "invtsc");
3770

3771
    if (rc < 0) {
3772
        return -1;
3773 3774 3775 3776 3777 3778 3779 3780 3781
    } 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;
        }
3782
    }
3783

3784
    return 0;
3785 3786 3787
}


3788
static int
3789 3790 3791 3792 3793
qemuProcessFetchGuestCPU(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob,
                         virCPUDataPtr *enabled,
                         virCPUDataPtr *disabled)
3794
{
3795 3796 3797 3798
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCPUDataPtr dataEnabled = NULL;
    virCPUDataPtr dataDisabled = NULL;
    int rc;
3799

3800 3801
    *enabled = NULL;
    *disabled = NULL;
3802

3803
    if (!ARCH_IS_X86(vm->def->os.arch))
3804 3805
        return 0;

3806 3807
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto error;
3808

3809 3810
    rc = qemuMonitorGetGuestCPU(priv->mon, vm->def->os.arch,
                                &dataEnabled, &dataDisabled);
3811

3812 3813 3814 3815 3816 3817 3818 3819
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto error;

    if (rc == -1)
        goto error;

    *enabled = dataEnabled;
    *disabled = dataDisabled;
3820
    return 0;
3821 3822 3823 3824 3825

 error:
    virCPUDataFree(dataEnabled);
    virCPUDataFree(dataDisabled);
    return -1;
3826
}
3827

3828

3829
static int
3830 3831
qemuProcessVerifyCPU(virDomainObjPtr vm,
                     virCPUDataPtr cpu)
3832
{
3833
    virDomainDefPtr def = vm->def;
3834

3835 3836
    if (!cpu)
        return 0;
J
Jiri Denemark 已提交
3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853

    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;
}


3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895
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;
}


3896
static int
3897
qemuProcessUpdateAndVerifyCPU(virQEMUDriverPtr driver,
3898 3899
                              virDomainObjPtr vm,
                              qemuDomainAsyncJob asyncJob)
3900
{
3901
    virCPUDataPtr cpu = NULL;
3902
    virCPUDataPtr disabled = NULL;
3903
    int ret = -1;
3904

3905 3906
    if (qemuProcessFetchGuestCPU(driver, vm, asyncJob, &cpu, &disabled) < 0)
        goto cleanup;
3907

J
Jiri Denemark 已提交
3908 3909
    if (qemuProcessVerifyCPU(vm, cpu) < 0)
        goto cleanup;
J
Ján Tomko 已提交
3910

3911 3912
    if (qemuProcessUpdateLiveGuestCPU(vm, cpu, disabled) < 0)
        goto cleanup;
3913

3914
    ret = 0;
3915

3916
 cleanup:
3917
    virCPUDataFree(cpu);
3918
    virCPUDataFree(disabled);
3919 3920 3921 3922
    return ret;
}


3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946
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;
}


3947 3948 3949 3950 3951 3952 3953
static int
qemuProcessUpdateCPU(virQEMUDriverPtr driver,
                     virDomainObjPtr vm,
                     qemuDomainAsyncJob asyncJob)
{
    virCPUDataPtr cpu = NULL;
    virCPUDataPtr disabled = NULL;
3954
    virDomainCapsCPUModelsPtr models = NULL;
3955 3956
    int ret = -1;

3957 3958 3959 3960 3961
    /* 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;

3962 3963 3964 3965 3966 3967
    if (qemuProcessFetchGuestCPU(driver, vm, asyncJob, &cpu, &disabled) < 0)
        goto cleanup;

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

3968 3969 3970 3971
    if (!(models = qemuProcessFetchCPUDefinitions(driver, vm, asyncJob)) ||
        virCPUTranslate(vm->def->os.arch, vm->def->cpu, models) < 0)
        goto cleanup;

3972 3973 3974 3975 3976
    ret = 0;

 cleanup:
    virCPUDataFree(cpu);
    virCPUDataFree(disabled);
3977
    virObjectUnref(models);
3978 3979 3980 3981
    return ret;
}


3982 3983
static int
qemuPrepareNVRAM(virQEMUDriverConfigPtr cfg,
3984
                 virDomainObjPtr vm)
3985 3986 3987 3988
{
    int ret = -1;
    int srcFD = -1;
    int dstFD = -1;
3989
    virDomainLoaderDefPtr loader = vm->def->os.loader;
3990
    bool created = false;
3991 3992
    const char *master_nvram_path;
    ssize_t r;
3993

3994
    if (!loader || !loader->nvram || virFileExists(loader->nvram))
3995 3996
        return 0;

3997 3998 3999
    master_nvram_path = loader->templt;
    if (!loader->templt) {
        size_t i;
4000 4001 4002
        for (i = 0; i < cfg->nfirmwares; i++) {
            if (STREQ(cfg->firmwares[i]->name, loader->path)) {
                master_nvram_path = cfg->firmwares[i]->nvram;
4003
                break;
4004 4005
            }
        }
4006
    }
4007

4008 4009 4010 4011 4012 4013
    if (!master_nvram_path) {
        virReportError(VIR_ERR_OPERATION_FAILED,
                       _("unable to find any master var store for "
                         "loader: %s"), loader->path);
        goto cleanup;
    }
4014

4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031
    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;
4032

4033 4034
    do {
        char buf[1024];
4035

4036
        if ((r = saferead(srcFD, buf, sizeof(buf))) < 0) {
4037
            virReportSystemError(errno,
4038
                                 _("Unable to read from file '%s'"),
4039 4040 4041
                                 master_nvram_path);
            goto cleanup;
        }
4042 4043

        if (safewrite(dstFD, buf, r) < 0) {
4044
            virReportSystemError(errno,
4045
                                 _("Unable to write to file '%s'"),
4046 4047 4048
                                 loader->nvram);
            goto cleanup;
        }
4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061
    } 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;
4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078
    }

    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;
}


4079 4080 4081
static void
qemuLogOperation(virDomainObjPtr vm,
                 const char *msg,
4082 4083
                 virCommandPtr cmd,
                 qemuDomainLogContextPtr logCtxt)
4084 4085 4086 4087 4088
{
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int qemuVersion = virQEMUCapsGetVersion(priv->qemuCaps);
    const char *package = virQEMUCapsGetPackage(priv->qemuCaps);
4089
    char *hostname = virGetHostname();
4090 4091

    if ((timestamp = virTimeStringNow()) == NULL)
4092
        goto cleanup;
4093

4094 4095
    if (qemuDomainLogContextWrite(logCtxt,
                                  "%s: %s %s, qemu version: %d.%d.%d%s, hostname: %s\n",
4096 4097 4098 4099
                                  timestamp, msg, VIR_LOG_VERSION_STRING,
                                  (qemuVersion / 1000000) % 1000,
                                  (qemuVersion / 1000) % 1000,
                                  qemuVersion % 1000,
4100 4101
                                  package ? package : "",
                                  hostname ? hostname : "") < 0)
4102
        goto cleanup;
4103

4104 4105 4106 4107 4108
    if (cmd) {
        char *args = virCommandToString(cmd);
        qemuDomainLogContextWrite(logCtxt, "%s\n", args);
        VIR_FREE(args);
    }
4109 4110

 cleanup:
4111
    VIR_FREE(hostname);
4112 4113 4114
    VIR_FREE(timestamp);
}

4115 4116 4117 4118 4119 4120 4121

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

4122
    VIR_FREE(inc->address);
4123
    VIR_FREE(inc->launchURI);
4124
    VIR_FREE(inc->deferredURI);
4125 4126 4127 4128 4129 4130 4131 4132
    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.
4133 4134 4135
 *
 * The caller is responsible for closing @fd, calling
 * qemuProcessIncomingDefFree will NOT close it.
4136 4137 4138
 */
qemuProcessIncomingDefPtr
qemuProcessIncomingDefNew(virQEMUCapsPtr qemuCaps,
4139
                          const char *listenAddress,
4140 4141 4142 4143 4144 4145
                          const char *migrateFrom,
                          int fd,
                          const char *path)
{
    qemuProcessIncomingDefPtr inc = NULL;

4146
    if (qemuMigrationDstCheckProtocol(qemuCaps, migrateFrom) < 0)
4147 4148 4149 4150 4151
        return NULL;

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

4152 4153 4154
    if (VIR_STRDUP(inc->address, listenAddress) < 0)
        goto error;

4155
    inc->launchURI = qemuMigrationDstGetURI(migrateFrom, fd);
4156 4157 4158
    if (!inc->launchURI)
        goto error;

4159 4160 4161 4162 4163 4164
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_INCOMING_DEFER)) {
        inc->deferredURI = inc->launchURI;
        if (VIR_STRDUP(inc->launchURI, "defer") < 0)
            goto error;
    }

4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175
    inc->fd = fd;
    inc->path = path;

    return inc;

 error:
    qemuProcessIncomingDefFree(inc);
    return NULL;
}


4176 4177 4178 4179 4180 4181 4182 4183
/*
 * 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,
4184 4185
                    virDomainObjPtr vm,
                    virDomainJobOperation operation)
4186
{
4187 4188
    if (qemuDomainObjBeginAsyncJob(driver, vm, QEMU_ASYNC_JOB_START,
                                   operation) < 0)
4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203
        return -1;

    qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
    return 0;
}


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


4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226
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;
}


4227
static int
4228
qemuProcessGraphicsReservePorts(virDomainGraphicsDefPtr graphics,
4229
                                bool reconnect)
4230
{
4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241
    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;

4242 4243
    switch (graphics->type) {
    case VIR_DOMAIN_GRAPHICS_TYPE_VNC:
4244 4245
        if (!graphics->data.vnc.autoport ||
            reconnect) {
4246
            if (virPortAllocatorSetUsed(graphics->data.vnc.port) < 0)
4247 4248 4249
                return -1;
            graphics->data.vnc.portReserved = true;
        }
4250
        if (graphics->data.vnc.websocket > 0 &&
4251
            virPortAllocatorSetUsed(graphics->data.vnc.websocket) < 0)
4252
            return -1;
4253 4254 4255
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SPICE:
4256
        if (graphics->data.spice.autoport && !reconnect)
4257
            return 0;
4258

4259
        if (graphics->data.spice.port > 0) {
4260
            if (virPortAllocatorSetUsed(graphics->data.spice.port) < 0)
4261
                return -1;
4262 4263
            graphics->data.spice.portReserved = true;
        }
4264

4265
        if (graphics->data.spice.tlsPort > 0) {
4266
            if (virPortAllocatorSetUsed(graphics->data.spice.tlsPort) < 0)
4267 4268
                return -1;
            graphics->data.spice.tlsPortReserved = true;
4269
        }
4270 4271 4272 4273 4274 4275 4276
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SDL:
    case VIR_DOMAIN_GRAPHICS_TYPE_RDP:
    case VIR_DOMAIN_GRAPHICS_TYPE_DESKTOP:
    case VIR_DOMAIN_GRAPHICS_TYPE_LAST:
        break;
4277 4278
    }

4279 4280 4281 4282
    return 0;
}


4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319
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:
    case VIR_DOMAIN_GRAPHICS_TYPE_LAST:
        break;
    }

    return 0;
}

4320
static int
4321
qemuProcessGetNetworkAddress(const char *netname,
4322 4323
                             char **netaddr)
{
4324
    virConnectPtr conn = NULL;
4325 4326 4327 4328 4329 4330 4331 4332 4333 4334
    int ret = -1;
    virNetworkPtr net;
    virNetworkDefPtr netdef = NULL;
    virNetworkIPDefPtr ipdef;
    virSocketAddr addr;
    virSocketAddrPtr addrptr = NULL;
    char *dev_name = NULL;
    char *xml = NULL;

    *netaddr = NULL;
4335 4336 4337 4338

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

4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404
    net = virNetworkLookupByName(conn, netname);
    if (!net)
        goto cleanup;

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

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

    switch (netdef->forward.type) {
    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;
    }

    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);
4405
    virObjectUnref(conn);
4406 4407 4408 4409
    VIR_FREE(xml);
    return ret;
}

4410

4411
static int
4412
qemuProcessGraphicsSetupNetworkAddress(virDomainGraphicsListenDefPtr glisten,
4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423
                                       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;
    }

4424
    rc = qemuProcessGetNetworkAddress(glisten->network, &glisten->address);
4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437
    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;
}


4438
static int
4439
qemuProcessGraphicsSetupListen(virQEMUDriverPtr driver,
4440 4441
                               virDomainGraphicsDefPtr graphics,
                               virDomainObjPtr vm)
4442
{
4443
    qemuDomainObjPrivatePtr priv = vm->privateData;
4444
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4445
    const char *type = virDomainGraphicsTypeToString(graphics->type);
4446
    char *listenAddr = NULL;
4447
    bool useSocket = false;
4448
    size_t i;
4449
    int ret = -1;
4450 4451 4452

    switch (graphics->type) {
    case VIR_DOMAIN_GRAPHICS_TYPE_VNC:
4453
        useSocket = cfg->vncAutoUnixSocket;
4454 4455 4456 4457
        listenAddr = cfg->vncListen;
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SPICE:
4458
        useSocket = cfg->spiceAutoUnixSocket;
4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473
        listenAddr = cfg->spiceListen;
        break;

    case VIR_DOMAIN_GRAPHICS_TYPE_SDL:
    case VIR_DOMAIN_GRAPHICS_TYPE_RDP:
    case VIR_DOMAIN_GRAPHICS_TYPE_DESKTOP:
    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:
4474 4475 4476 4477 4478
            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) {
4479 4480
                    memset(glisten, 0, sizeof(virDomainGraphicsListenDef));
                    if (virAsprintf(&glisten->socket, "%s/%s.sock",
4481
                                    priv->libDir, type) < 0)
4482
                        goto cleanup;
4483 4484
                    glisten->fromConfig = true;
                    glisten->type = VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_SOCKET;
4485 4486
                } else if (listenAddr) {
                    if (VIR_STRDUP(glisten->address, listenAddr) < 0)
4487
                        goto cleanup;
4488 4489 4490
                    glisten->fromConfig = true;
                }
            }
4491 4492 4493 4494 4495 4496
            break;

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

4497
            if (qemuProcessGraphicsSetupNetworkAddress(glisten,
4498
                                                       listenAddr) < 0)
4499
                goto cleanup;
4500 4501
            break;

4502 4503 4504 4505
        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_SOCKET:
            if (!glisten->socket) {
                if (virAsprintf(&glisten->socket, "%s/%s.sock",
                                priv->libDir, type) < 0)
4506
                    goto cleanup;
4507 4508 4509 4510
                glisten->autoGenerated = true;
            }
            break;

4511 4512 4513 4514 4515 4516
        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_NONE:
        case VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_LAST:
            break;
        }
    }

4517 4518 4519 4520 4521
    ret = 0;

 cleanup:
    virObjectUnref(cfg);
    return ret;
4522 4523 4524
}


4525
static int
4526
qemuProcessSetupGraphics(virQEMUDriverPtr driver,
4527 4528
                         virDomainObjPtr vm,
                         unsigned int flags)
4529
{
4530
    virDomainGraphicsDefPtr graphics;
4531
    bool allocate = !(flags & VIR_QEMU_PROCESS_START_PRETEND);
4532
    size_t i;
4533 4534
    int ret = -1;

4535 4536 4537
    for (i = 0; i < vm->def->ngraphics; i++) {
        graphics = vm->def->graphics[i];

4538
        if (qemuProcessGraphicsSetupListen(driver, graphics, vm) < 0)
4539 4540 4541
            goto cleanup;
    }

4542 4543 4544 4545
    if (allocate) {
        for (i = 0; i < vm->def->ngraphics; i++) {
            graphics = vm->def->graphics[i];

4546
            if (qemuProcessGraphicsReservePorts(graphics, false) < 0)
4547 4548 4549
                goto cleanup;
        }
    }
4550

4551
    for (i = 0; i < vm->def->ngraphics; ++i) {
4552
        graphics = vm->def->graphics[i];
4553

4554 4555
        if (qemuProcessGraphicsAllocatePorts(driver, graphics, allocate) < 0)
            goto cleanup;
4556 4557 4558 4559 4560 4561 4562 4563 4564
    }

    ret = 0;

 cleanup:
    return ret;
}


4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597
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++) {
4598
            if (!virHostdevIsSCSIDevice(vm->def->hostdevs[i]))
4599 4600
                continue;

4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626
            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;
}


4627 4628 4629 4630 4631 4632 4633 4634 4635
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;

4636
    if (!virDomainDefHasMemballoon(vm->def))
4637 4638 4639
        return 0;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
4640
        return -1;
4641

4642 4643 4644
    if (vm->def->memballoon->period)
        qemuMonitorSetMemoryStatsPeriod(priv->mon, vm->def->memballoon,
                                        vm->def->memballoon->period);
4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656
    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
4657 4658 4659
static int
qemuProcessMakeDir(virQEMUDriverPtr driver,
                   virDomainObjPtr vm,
4660
                   const char *path)
J
Jiri Denemark 已提交
4661 4662 4663 4664 4665 4666 4667 4668
{
    int ret = -1;

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

4669
    if (qemuSecurityDomainSetPathLabel(driver->securityManager,
4670
                                       vm->def, path, true) < 0)
J
Jiri Denemark 已提交
4671 4672 4673 4674 4675 4676 4677 4678 4679
        goto cleanup;

    ret = 0;

 cleanup:
    return ret;
}


4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715
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) ==
4716
                VIR_DOMAIN_MEMORY_ACCESS_SHARED) {
4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728
                shmem = true;
                break;
            }
        }
    }

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

4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756 4757 4758 4759 4760

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:
        case VIR_DOMAIN_GRAPHICS_TYPE_LAST:
            break;
        }
    }

    return 0;
}


4761 4762 4763 4764 4765 4766 4767 4768 4769 4770
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];

4771 4772 4773 4774 4775 4776 4777 4778 4779
        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 &&
4780 4781 4782 4783
             !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))) {
4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805
            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;
}


4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836
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;
}


4837
static int
4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856
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;
}


4857 4858 4859 4860 4861 4862 4863
static int
qemuProcessStartValidateDisks(virDomainObjPtr vm,
                              virQEMUCapsPtr qemuCaps)
{
    size_t i;

    for (i = 0; i < vm->def->ndisks; i++) {
4864 4865
        virDomainDiskDefPtr disk = vm->def->disks[i];
        virStorageSourcePtr src = disk->src;
4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877

        /* 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;
        }
4878 4879 4880 4881 4882 4883 4884 4885

        /* 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;
        }
4886 4887 4888 4889 4890 4891
    }

    return 0;
}


4892
static int
4893 4894
qemuProcessStartValidateXML(virQEMUDriverPtr driver,
                            virDomainObjPtr vm,
4895
                            virQEMUCapsPtr qemuCaps,
4896
                            virCapsPtr caps,
4897
                            unsigned int flags)
4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909
{
    /* The bits we validate here are XML configs that we previously
     * accepted. We reject them at VM startup time rather than parse
     * time so that pre-existing VMs aren't rejected and dropped from
     * the VM list when libvirt is updated.
     *
     * If back compat isn't a concern, XML validation should probably
     * be done at parse time.
     */
    if (qemuValidateCpuCount(vm->def, qemuCaps) < 0)
        return -1;

4910 4911 4912
    /* 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 */
4913 4914 4915
    if ((flags & VIR_QEMU_PROCESS_START_NEW) &&
        virDomainDefValidate(vm->def, caps, 0, driver->xmlopt) < 0)
        return -1;
4916 4917 4918

    return 0;
}
4919

4920

4921 4922 4923 4924 4925 4926
/**
 * qemuProcessStartValidate:
 * @vm: domain object
 * @qemuCaps: emulator capabilities
 * @migration: restoration of existing state
 *
4927 4928 4929 4930 4931
 * 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.
4932
 */
4933
static int
4934 4935
qemuProcessStartValidate(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
4936
                         virQEMUCapsPtr qemuCaps,
4937
                         virCapsPtr caps,
4938
                         unsigned int flags)
4939
{
4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953
    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");
4954
        if (qemuSecurityCheckAllLabel(driver->securityManager, vm->def) < 0)
4955 4956 4957 4958
            return -1;

    }

4959
    if (qemuProcessStartValidateXML(driver, vm, qemuCaps, caps, flags) < 0)
4960 4961
        return -1;

4962 4963 4964
    if (qemuProcessStartValidateGraphics(vm) < 0)
        return -1;

4965 4966 4967
    if (qemuProcessStartValidateVideo(vm, qemuCaps) < 0)
        return -1;

4968 4969 4970
    if (qemuProcessStartValidateIOThreads(vm, qemuCaps) < 0)
        return -1;

4971 4972 4973
    if (qemuProcessStartValidateShmem(vm) < 0)
        return -1;

4974 4975 4976 4977
    if (vm->def->cpu &&
        virCPUValidateFeatures(vm->def->os.arch, vm->def->cpu) < 0)
        return -1;

4978 4979 4980
    if (qemuProcessStartValidateDisks(vm, qemuCaps) < 0)
        return -1;

4981 4982
    VIR_DEBUG("Checking for any possible (non-fatal) issues");

4983
    qemuProcessStartWarnShmem(vm);
4984

4985
    return 0;
4986 4987 4988
}


J
Jiri Denemark 已提交
4989 4990 4991 4992 4993 4994
/**
 * qemuProcessInit:
 *
 * Prepares the domain up to the point when priv->qemuCaps is initialized. The
 * function calls qemuProcessStop when needed.
 *
4995 4996 4997 4998
 * 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
4999 5000 5001 5002 5003
 * Returns 0 on success, -1 on error.
 */
int
qemuProcessInit(virQEMUDriverPtr driver,
                virDomainObjPtr vm,
5004
                virCPUDefPtr updatedCPU,
5005
                qemuDomainAsyncJob asyncJob,
5006
                bool migration,
5007
                unsigned int flags)
J
Jiri Denemark 已提交
5008 5009 5010 5011 5012
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int stopFlags;
5013
    virCPUDefPtr origCPU = NULL;
J
Jiri Denemark 已提交
5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027
    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)))
5028 5029
        goto cleanup;

5030 5031 5032 5033 5034 5035 5036 5037 5038
    /* 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;
    }

5039 5040
    VIR_DEBUG("Determining emulator version");
    virObjectUnref(priv->qemuCaps);
5041
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
5042 5043 5044
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
        goto cleanup;
J
Jiri Denemark 已提交
5045

5046 5047 5048
    if (qemuDomainUpdateCPU(vm, updatedCPU, &origCPU) < 0)
        goto cleanup;

5049
    if (qemuProcessStartValidate(driver, vm, priv->qemuCaps, caps, flags) < 0)
5050 5051
        goto cleanup;

J
Jiri Denemark 已提交
5052 5053 5054 5055 5056
    /* 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");
5057
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm) < 0)
5058
        goto cleanup;
J
Jiri Denemark 已提交
5059

5060
    if (flags & VIR_QEMU_PROCESS_START_PRETEND) {
5061 5062 5063 5064
        if (qemuDomainSetPrivatePaths(driver, vm) < 0) {
            virDomainObjRemoveTransientDef(vm);
            goto cleanup;
        }
5065
    } else {
5066 5067 5068
        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
5069

5070 5071
        if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
            driver->inhibitCallback(true, driver->inhibitOpaque);
J
Jiri Denemark 已提交
5072

5073 5074 5075 5076 5077
        /* 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
5078

5079 5080
        if (qemuDomainSetPrivatePaths(driver, vm) < 0)
            goto stop;
5081 5082

        VIR_STEAL_PTR(priv->origCPU, origCPU);
5083
    }
5084

J
Jiri Denemark 已提交
5085 5086 5087
    ret = 0;

 cleanup:
5088
    virCPUDefFree(origCPU);
J
Jiri Denemark 已提交
5089 5090 5091 5092 5093 5094 5095 5096
    virObjectUnref(cfg);
    virObjectUnref(caps);
    return ret;

 stop:
    stopFlags = VIR_QEMU_PROCESS_STOP_NO_RELABEL;
    if (migration)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
5097
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, asyncJob, stopFlags);
J
Jiri Denemark 已提交
5098 5099 5100 5101
    goto cleanup;
}


5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112
/**
 * qemuProcessNetworkPrepareDevices
 */
static int
qemuProcessNetworkPrepareDevices(virDomainDefPtr def)
{
    int ret = -1;
    size_t i;

    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
5113
        virDomainNetType actualType;
5114 5115 5116 5117 5118

        /* 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.
         */
5119
        if (virDomainNetAllocateActualDevice(def, net) < 0)
5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154
            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;
}


5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170
/**
 * 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);
5171
    virDomainVcpuDefPtr vcpu = virDomainDefGetVcpu(vm->def, vcpuid);
5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189
    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;

    for (i = 0; i < vm->def->ncachetunes; i++) {
        virDomainCachetuneDefPtr ct = vm->def->cachetunes[i];

        if (virBitmapIsBitSet(ct->vcpus, vcpuid)) {
            if (virResctrlAllocAddPID(ct->alloc, vcpupid) < 0)
                return -1;
            break;
        }
    }
5190

5191
    return 0;
5192 5193 5194 5195 5196 5197
}


static int
qemuProcessSetupVcpus(virDomainObjPtr vm)
{
5198
    virDomainVcpuDefPtr vcpu;
5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243
    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;
}


5244 5245 5246 5247 5248
int
qemuProcessSetupIOThread(virDomainObjPtr vm,
                         virDomainIOThreadIDDefPtr iothread)
{

5249 5250
    return qemuProcessSetupPid(vm, iothread->thread_id,
                               VIR_CGROUP_THREAD_IOTHREAD,
5251
                               iothread->iothread_id,
5252
                               iothread->cpumask,
5253 5254
                               vm->def->cputune.iothread_period,
                               vm->def->cputune.iothread_quota,
5255
                               &iothread->sched);
5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274
}


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;
}


5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286
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;

5287
    if (!(ordermap = virBitmapNew(maxvcpus + 1)))
5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303
        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) {
5304
            if (virBitmapIsBitSet(ordermap, vcpu->order)) {
5305
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
5306
                               _("duplicate vcpu order '%u'"), vcpu->order);
5307 5308 5309
                goto cleanup;
            }

5310 5311 5312 5313 5314 5315
            if (virBitmapSetBit(ordermap, vcpu->order)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("vcpu order '%u' exceeds vcpu count"),
                               vcpu->order);
                goto cleanup;
            }
5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331
        }

        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 &&
5332
                 vcpupriv->thread_id == -1 && vcpupriv->node_id == -1) ||
5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382
                !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);
5383 5384
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuCgroupEmulatorAllNodesDataPtr emulatorCgroup = NULL;
5385 5386 5387 5388 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
    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);

5417 5418 5419
    if (qemuCgroupEmulatorAllNodesAllow(priv->cgroup, &emulatorCgroup) < 0)
        goto cleanup;

5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443
    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:
5444
    qemuCgroupEmulatorAllNodesRestore(emulatorCgroup);
5445 5446 5447 5448 5449 5450
    VIR_FREE(bootHotplug);
    virJSONValueFree(vcpuprops);
    return ret;
}


5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489
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;
    }

5490 5491 5492
    if (virCPUConvertLegacy(caps->host.arch, def->cpu) < 0)
        return -1;

5493 5494 5495 5496
    /* nothing to update for host-passthrough */
    if (def->cpu->mode == VIR_CPU_MODE_HOST_PASSTHROUGH)
        return 0;

5497 5498
    if (def->cpu->check == VIR_CPU_CHECK_PARTIAL &&
        virCPUCompare(caps->host.arch,
5499
                      virQEMUCapsGetHostModel(qemuCaps, def->virtType,
5500
                                              VIR_QEMU_CAPS_HOST_CPU_FULL),
5501 5502
                      def->cpu, true) < 0)
        return -1;
5503 5504

    if (virCPUUpdate(def->os.arch, def->cpu,
5505
                     virQEMUCapsGetHostModel(qemuCaps, def->virtType,
5506
                                             VIR_QEMU_CAPS_HOST_CPU_MIGRATABLE)) < 0)
5507 5508
        goto cleanup;

5509 5510
    if (virCPUTranslate(def->os.arch, def->cpu,
                        virQEMUCapsGetCPUDefinitions(qemuCaps, def->virtType)) < 0)
5511 5512 5513 5514 5515 5516 5517 5518 5519 5520
        goto cleanup;

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

 cleanup:
    return ret;
}


5521 5522 5523 5524 5525 5526
static int
qemuProcessPrepareDomainNUMAPlacement(virDomainObjPtr vm,
                                      virCapsPtr caps)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    char *nodeset = NULL;
5527 5528
    virBitmapPtr numadNodeset = NULL;
    virBitmapPtr hostMemoryNodeset = NULL;
5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542
    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;

5543 5544 5545
    if (!(hostMemoryNodeset = virNumaGetHostMemoryNodeset()))
        goto cleanup;

5546 5547
    VIR_DEBUG("Nodeset returned from numad: %s", nodeset);

5548
    if (virBitmapParse(nodeset, &numadNodeset, VIR_DOMAIN_CPUMASK_LEN) < 0)
5549 5550
        goto cleanup;

5551 5552 5553 5554
    /* 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)))
5555 5556
        goto cleanup;

5557 5558 5559 5560
    virBitmapIntersect(numadNodeset, hostMemoryNodeset);

    VIR_STEAL_PTR(priv->autoNodeset, numadNodeset);

5561 5562 5563 5564
    ret = 0;

 cleanup:
    VIR_FREE(nodeset);
5565 5566
    virBitmapFree(numadNodeset);
    virBitmapFree(hostMemoryNodeset);
5567 5568 5569 5570
    return ret;
}


5571
static int
5572
qemuProcessPrepareDomainStorage(virQEMUDriverPtr driver,
5573
                                virDomainObjPtr vm,
5574
                                qemuDomainObjPrivatePtr priv,
5575
                                virQEMUDriverConfigPtr cfg,
5576 5577 5578 5579 5580 5581 5582 5583 5584
                                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];

5585
        if (virDomainDiskTranslateSourcePool(disk) < 0) {
5586 5587 5588 5589 5590 5591
            if (qemuDomainCheckDiskStartupPolicy(driver, vm, idx, cold_boot) < 0)
                return -1;

            /* disk source was dropped */
            continue;
        }
5592

5593
        if (qemuDomainPrepareDiskSource(disk, priv, cfg) < 0)
5594
            return -1;
5595 5596 5597 5598 5599 5600
    }

    return 0;
}


5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620
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;
    }
}


5621
/**
5622 5623 5624 5625
 * qemuProcessPrepareDomain:
 * @driver: qemu driver
 * @vm: domain object
 * @flags: qemuProcessStartFlags
5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636
 *
 * 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
5637
qemuProcessPrepareDomain(virQEMUDriverPtr driver,
5638 5639 5640 5641 5642 5643
                         virDomainObjPtr vm,
                         unsigned int flags)
{
    int ret = -1;
    size_t i;
    qemuDomainObjPrivatePtr priv = vm->privateData;
5644
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5645 5646 5647 5648 5649
    virCapsPtr caps;

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

5650 5651 5652 5653
    priv->machineName = qemuDomainGetMachineName(vm);
    if (!priv->machineName)
        goto cleanup;

5654 5655 5656 5657
    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)");
5658
        if (qemuSecurityGenLabel(driver->securityManager, vm->def) < 0) {
5659 5660 5661 5662 5663
            virDomainAuditSecurityLabel(vm, false);
            goto cleanup;
        }
        virDomainAuditSecurityLabel(vm, true);

5664 5665
        if (qemuProcessPrepareDomainNUMAPlacement(vm, caps) < 0)
            goto cleanup;
5666 5667
    }

5668 5669 5670 5671 5672 5673 5674
    /* 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;
    }

5675 5676
    qemuProcessPrepareAllowReboot(vm);

5677 5678 5679 5680 5681 5682 5683
    /*
     * 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
     */
5684
    VIR_DEBUG("Assigning domain PCI addresses");
5685 5686
    if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, driver, vm,
                                   !!(flags & VIR_QEMU_PROCESS_START_NEW))) < 0) {
5687
        goto cleanup;
5688
    }
5689

5690 5691 5692
    if (qemuAssignDeviceAliases(vm->def, priv->qemuCaps) < 0)
        goto cleanup;

5693
    VIR_DEBUG("Setting graphics devices");
5694
    if (qemuProcessSetupGraphics(driver, vm, flags) < 0)
5695 5696
        goto cleanup;

5697 5698 5699 5700
    VIR_DEBUG("Create domain masterKey");
    if (qemuDomainMasterKeyCreate(vm) < 0)
        goto cleanup;

5701
    VIR_DEBUG("Setting up storage");
5702
    if (qemuProcessPrepareDomainStorage(driver, vm, priv, cfg, flags) < 0)
5703 5704
        goto cleanup;

5705
    VIR_DEBUG("Prepare chardev source backends for TLS");
5706
    qemuDomainPrepareChardevSource(vm->def, cfg);
5707

5708
    VIR_DEBUG("Add secrets to hostdevs and chardevs");
5709
    if (qemuDomainSecretPrepare(driver, vm) < 0)
5710 5711
        goto cleanup;

5712 5713 5714 5715 5716 5717
    for (i = 0; i < vm->def->nchannels; i++) {
        if (qemuDomainPrepareChannel(vm->def->channels[i],
                                     priv->channelTargetDir) < 0)
            goto cleanup;
    }

5718
    if (!(priv->monConfig = virDomainChrSourceDefNew(NULL)))
5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729
        goto cleanup;

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

    priv->monJSON = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON);
    priv->monError = false;
    priv->monStart = 0;
    priv->gotShutdown = false;

5730 5731 5732 5733
    VIR_DEBUG("Updating guest CPU definition");
    if (qemuProcessUpdateGuestCPU(vm->def, priv->qemuCaps, caps, flags) < 0)
        goto cleanup;

5734 5735 5736 5737 5738
    for (i = 0; i < vm->def->nshmems; i++) {
        if (qemuDomainPrepareShmemChardev(vm->def->shmems[i]) < 0)
            goto cleanup;
    }

5739 5740 5741
    ret = 0;
 cleanup:
    virObjectUnref(caps);
5742
    virObjectUnref(cfg);
5743 5744 5745 5746
    return ret;
}


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 5774
static int
qemuProcessPrepareHostStorage(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              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];

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

        if (qemuDomainDetermineDiskChain(driver, vm, disk, true, true) >= 0)
            continue;

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

        return -1;
    }

    return 0;
}


J
Jiri Denemark 已提交
5775
/**
5776 5777 5778
 * qemuProcessPrepareHost:
 * @driver: qemu driver
 * @vm: domain object
5779
 * @flags: qemuProcessStartFlags
J
Jiri Denemark 已提交
5780
 *
5781 5782 5783
 * 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
5784
 *
5785
 * TODO: move all host modification from qemuBuildCommandLine into this function
J
Jiri Denemark 已提交
5786 5787
 */
int
5788 5789
qemuProcessPrepareHost(virQEMUDriverPtr driver,
                       virDomainObjPtr vm,
5790
                       unsigned int flags)
5791
{
5792
    int ret = -1;
5793
    unsigned int hostdev_flags = 0;
5794 5795
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5796

5797
    if (qemuPrepareNVRAM(cfg, vm) < 0)
J
Jiri Denemark 已提交
5798
        goto cleanup;
5799

5800 5801 5802 5803 5804
    /* 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");
5805
    if (qemuProcessNetworkPrepareDevices(vm->def) < 0)
J
Jiri Denemark 已提交
5806
        goto cleanup;
5807

5808
    /* Must be run before security labelling */
5809
    VIR_DEBUG("Preparing host devices");
5810 5811
    if (!cfg->relaxedACS)
        hostdev_flags |= VIR_HOSTDEV_STRICT_ACS_CHECK;
5812
    if (flags & VIR_QEMU_PROCESS_START_NEW)
5813
        hostdev_flags |= VIR_HOSTDEV_COLD_BOOT;
5814 5815
    if (qemuHostdevPrepareDomainDevices(driver, vm->def, priv->qemuCaps,
                                        hostdev_flags) < 0)
J
Jiri Denemark 已提交
5816
        goto cleanup;
5817

5818
    VIR_DEBUG("Preparing chr devices");
5819 5820 5821 5822
    if (virDomainChrDefForeach(vm->def,
                               true,
                               qemuProcessPrepareChardevDevice,
                               NULL) < 0)
J
Jiri Denemark 已提交
5823
        goto cleanup;
5824

5825
    if (qemuProcessBuildDestroyMemoryPaths(driver, vm, NULL, true) < 0)
5826
        goto cleanup;
5827

5828 5829
    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
5830
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
5831
    qemuRemoveCgroup(vm);
5832

5833
    if (virFileMakePath(cfg->logDir) < 0) {
5834 5835
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
5836
                             cfg->logDir);
J
Jiri Denemark 已提交
5837
        goto cleanup;
5838 5839
    }

5840 5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858 5859 5860 5861 5862
    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;

5863 5864
    VIR_DEBUG("Write domain masterKey");
    if (qemuDomainWriteMasterKeyFile(driver, vm) < 0)
J
John Ferlan 已提交
5865 5866
        goto cleanup;

5867 5868 5869 5870
    VIR_DEBUG("Preparing disks (host)");
    if (qemuProcessPrepareHostStorage(driver, vm, flags) < 0)
        goto cleanup;

5871 5872 5873 5874 5875 5876 5877 5878 5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911
    ret = 0;
 cleanup:
    virObjectUnref(cfg);
    return ret;
}


/**
 * 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
5912
    size_t i;
5913

5914
    VIR_DEBUG("conn=%p driver=%p vm=%p name=%s if=%d asyncJob=%d "
5915 5916 5917
              "incoming.launchURI=%s incoming.deferredURI=%s "
              "incoming.fd=%d incoming.path=%s "
              "snapshot=%p vmop=%d flags=0x%x",
5918
              conn, driver, vm, vm->def->name, vm->def->id, asyncJob,
5919 5920 5921 5922 5923 5924 5925 5926 5927 5928
              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 |
5929 5930
                  VIR_QEMU_PROCESS_START_AUTODESTROY |
                  VIR_QEMU_PROCESS_START_NEW, -1);
5931 5932 5933

    cfg = virQEMUDriverGetConfig(driver);

5934 5935 5936 5937 5938 5939
    if ((flags & VIR_QEMU_PROCESS_START_AUTODESTROY) && !conn) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Domain autodestroy requires a connection handle"));
        return -1;
    }

5940 5941 5942 5943 5944 5945 5946 5947
    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;

5948
    VIR_DEBUG("Creating domain log file");
5949 5950
    if (!(logCtxt = qemuDomainLogContextNew(driver, vm,
                                            QEMU_DOMAIN_LOG_CONTEXT_MODE_START)))
J
Jiri Denemark 已提交
5951
        goto cleanup;
5952
    logfile = qemuDomainLogContextGetWriteFD(logCtxt);
5953

5954
    VIR_DEBUG("Building emulator command line");
5955
    if (!(cmd = qemuBuildCommandLine(driver,
5956
                                     qemuDomainLogContextGetManager(logCtxt),
5957
                                     vm,
5958 5959
                                     incoming ? incoming->launchURI : NULL,
                                     snapshot, vmop,
J
Ján Tomko 已提交
5960
                                     false,
5961
                                     qemuCheckFips(),
5962
                                     &nnicindexes, &nicindexes)))
J
Jiri Denemark 已提交
5963
        goto cleanup;
5964

5965 5966
    if (incoming && incoming->fd != -1)
        virCommandPassFD(cmd, incoming->fd, 0);
5967

5968
    /* now that we know it is about to start call the hook if present */
5969 5970 5971
    if (qemuProcessStartHook(driver, vm,
                             VIR_HOOK_QEMU_OP_START,
                             VIR_HOOK_SUBOP_BEGIN) < 0)
J
Jiri Denemark 已提交
5972
        goto cleanup;
5973

5974
    qemuLogOperation(vm, "starting up", cmd, logCtxt);
5975

5976
    qemuDomainObjCheckTaint(driver, vm, logCtxt);
5977

5978
    qemuDomainLogContextMarkPosition(logCtxt);
5979

5980 5981 5982 5983 5984
    VIR_DEBUG("Building mount namespace");

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

5985
    VIR_DEBUG("Clear emulator capabilities: %d",
5986 5987
              cfg->clearEmulatorCapabilities);
    if (cfg->clearEmulatorCapabilities)
5988 5989
        virCommandClearCaps(cmd);

5990 5991
    VIR_DEBUG("Setting up raw IO");
    if (qemuProcessSetupRawIO(driver, vm, cmd) < 0)
J
Jiri Denemark 已提交
5992
        goto cleanup;
5993

5994
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);
5995 5996
    virCommandSetMaxProcesses(cmd, cfg->maxProcesses);
    virCommandSetMaxFiles(cmd, cfg->maxFiles);
5997
    virCommandSetMaxCoreSize(cmd, cfg->maxCore);
5998
    virCommandSetUmask(cmd, 0x002);
5999

6000
    VIR_DEBUG("Setting up security labelling");
6001 6002
    if (qemuSecuritySetChildProcessLabel(driver->securityManager,
                                         vm->def, cmd) < 0)
J
Jiri Denemark 已提交
6003
        goto cleanup;
6004

6005 6006 6007
    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
6008
    virCommandSetPidFile(cmd, priv->pidfile);
6009
    virCommandDaemonize(cmd);
6010
    virCommandRequireHandshake(cmd);
6011

6012
    if (qemuSecurityPreFork(driver->securityManager) < 0)
J
Jiri Denemark 已提交
6013
        goto cleanup;
6014
    rv = virCommandRun(cmd, NULL);
6015
    qemuSecurityPostFork(driver->securityManager);
6016

E
Eric Blake 已提交
6017
    /* wait for qemu process to show up */
6018
    if (rv == 0) {
6019
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
6020 6021
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
6022
            rv = -1;
6023
        }
M
Michal Privoznik 已提交
6024 6025
        VIR_DEBUG("QEMU vm=%p name=%s running with pid=%lld",
                  vm, vm->def->name, (long long) vm->pid);
6026 6027 6028
    } else {
        VIR_DEBUG("QEMU vm=%p name=%s failed to spawn",
                  vm, vm->def->name);
6029 6030
    }

6031
    VIR_DEBUG("Writing early domain status to disk");
6032
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
J
Jiri Denemark 已提交
6033
        goto cleanup;
6034

6035 6036
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
6037
        /* Read errors from child that occurred between fork and exec. */
6038 6039
        qemuProcessReportLogError(logCtxt,
                                  _("Process exited prior to exec"));
J
Jiri Denemark 已提交
6040
        goto cleanup;
6041 6042
    }

6043
    VIR_DEBUG("Setting up domain cgroup (if required)");
6044
    if (qemuSetupCgroup(vm, nnicindexes, nicindexes) < 0)
J
Jiri Denemark 已提交
6045
        goto cleanup;
6046

6047 6048 6049 6050
    if (!(priv->perf = virPerfNew()))
        goto cleanup;

    for (i = 0; i < VIR_PERF_EVENT_LAST; i++) {
6051
        if (vm->def->perf.events[i] == VIR_TRISTATE_BOOL_YES &&
6052 6053
            virPerfEventEnable(priv->perf, i, vm->pid) < 0)
            goto cleanup;
Q
Qiaowei Ren 已提交
6054
    }
6055

6056 6057 6058 6059 6060 6061
    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
    if (!vm->def->cputune.emulatorpin &&
        qemuProcessInitCpuAffinity(vm) < 0)
        goto cleanup;

6062 6063 6064 6065
    VIR_DEBUG("Setting emulator tuning/settings");
    if (qemuProcessSetupEmulator(vm) < 0)
        goto cleanup;

6066
    VIR_DEBUG("Setting up resctrl");
6067 6068 6069
    if (qemuProcessResctrlCreate(driver, vm) < 0)
        goto cleanup;

6070
    VIR_DEBUG("Setting domain security labels");
6071 6072 6073 6074
    if (qemuSecuritySetAllLabel(driver,
                                vm,
                                incoming ? incoming->path : NULL) < 0)
        goto cleanup;
6075

6076
    /* Security manager labeled all devices, therefore
J
Jiri Denemark 已提交
6077 6078 6079 6080
     * if any operation from now on fails, we need to ask the caller to
     * restore labels.
     */
    ret = -2;
6081

J
Jiri Denemark 已提交
6082
    if (incoming && incoming->fd != -1) {
6083 6084 6085 6086 6087 6088 6089
        /* 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
6090
        if (fstat(incoming->fd, &stdin_sb) < 0) {
6091
            virReportSystemError(errno,
J
Jiri Denemark 已提交
6092 6093
                                 _("cannot stat fd %d"), incoming->fd);
            goto cleanup;
6094 6095
        }
        if (S_ISFIFO(stdin_sb.st_mode) &&
6096 6097
            qemuSecuritySetImageFDLabel(driver->securityManager,
                                        vm->def, incoming->fd) < 0)
J
Jiri Denemark 已提交
6098
            goto cleanup;
6099 6100 6101
    }

    VIR_DEBUG("Labelling done, completing handshake to child");
6102
    if (virCommandHandshakeNotify(cmd) < 0)
J
Jiri Denemark 已提交
6103
        goto cleanup;
6104 6105
    VIR_DEBUG("Handshake complete, child running");

6106
    if (rv == -1) /* The VM failed to start; tear filters before taps */
6107 6108
        virDomainConfVMNWFilterTeardown(vm);

6109
    if (rv == -1) /* The VM failed to start */
J
Jiri Denemark 已提交
6110
        goto cleanup;
6111

6112
    VIR_DEBUG("Waiting for monitor to show up");
6113
    if (qemuProcessWaitForMonitor(driver, vm, asyncJob, logCtxt) < 0)
J
Jiri Denemark 已提交
6114
        goto cleanup;
6115

6116 6117
    if (qemuConnectAgent(driver, vm) < 0)
        goto cleanup;
D
Daniel P. Berrange 已提交
6118

6119
    VIR_DEBUG("Verifying and updating provided guest CPU");
6120
    if (qemuProcessUpdateAndVerifyCPU(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6121
        goto cleanup;
6122

6123
    VIR_DEBUG("Setting up post-init cgroup restrictions");
6124
    if (qemuSetupCpusetMems(vm) < 0)
J
Jiri Denemark 已提交
6125
        goto cleanup;
6126

6127 6128 6129 6130 6131 6132 6133 6134 6135 6136 6137 6138
    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;
    }

6139
    VIR_DEBUG("Refreshing VCPU info");
6140
    if (qemuDomainRefreshVcpuInfo(driver, vm, asyncJob, false) < 0)
J
Jiri Denemark 已提交
6141
        goto cleanup;
6142

6143 6144 6145
    if (qemuDomainValidateVcpuInfo(vm) < 0)
        goto cleanup;

6146 6147
    qemuDomainVcpuPersistOrder(vm->def);

6148 6149
    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6150
        goto cleanup;
6151

6152 6153 6154 6155
    VIR_DEBUG("Setting global CPU cgroup (if required)");
    if (qemuSetupGlobalCpuCgroup(vm) < 0)
        goto cleanup;

6156 6157
    VIR_DEBUG("Setting vCPU tuning/settings");
    if (qemuProcessSetupVcpus(vm) < 0)
J
Jiri Denemark 已提交
6158
        goto cleanup;
6159

6160 6161
    VIR_DEBUG("Setting IOThread tuning/settings");
    if (qemuProcessSetupIOThreads(vm) < 0)
J
Jiri Denemark 已提交
6162
        goto cleanup;
6163

6164
    VIR_DEBUG("Setting any required VM passwords");
6165
    if (qemuProcessInitPasswords(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6166
        goto cleanup;
6167

6168 6169 6170 6171
    /* set default link states */
    /* qemu doesn't support setting this on the command line, so
     * enter the monitor */
    VIR_DEBUG("Setting network link states");
6172
    if (qemuProcessSetLinkStates(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6173
        goto cleanup;
6174

6175
    VIR_DEBUG("Setting initial memory amount");
6176
    if (qemuProcessSetupBalloon(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6177
        goto cleanup;
6178

6179
    /* Since CPUs were not started yet, the balloon could not return the memory
6180 6181
     * 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 */
6182
    if (!incoming && !snapshot &&
6183
        qemuProcessRefreshBalloonState(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
6184
        goto cleanup;
6185

J
Jiri Denemark 已提交
6186 6187 6188 6189 6190 6191 6192
    if (flags & VIR_QEMU_PROCESS_START_AUTODESTROY &&
        qemuProcessAutoDestroyAdd(driver, vm, conn) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
6193
    qemuDomainSecretDestroy(vm);
J
Jiri Denemark 已提交
6194
    virCommandFree(cmd);
6195
    virObjectUnref(logCtxt);
J
Jiri Denemark 已提交
6196 6197 6198 6199 6200 6201 6202
    virObjectUnref(cfg);
    virObjectUnref(caps);
    VIR_FREE(nicindexes);
    return ret;
}


6203 6204 6205 6206 6207 6208 6209 6210 6211 6212
/**
 * 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.
 */
6213
int
6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231 6232 6233 6234 6235 6236 6237
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;
}


6238 6239 6240 6241 6242 6243
/**
 * qemuProcessFinishStartup:
 *
 * Finish starting a new domain.
 */
int
6244
qemuProcessFinishStartup(virQEMUDriverPtr driver,
6245 6246 6247 6248 6249 6250 6251 6252 6253 6254
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob,
                         bool startCPUs,
                         virDomainPausedReason pausedReason)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;

    if (startCPUs) {
        VIR_DEBUG("Starting domain CPUs");
6255
        if (qemuProcessStartCPUs(driver, vm,
6256 6257 6258 6259 6260 6261 6262 6263 6264 6265 6266 6267
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 asyncJob) < 0) {
            if (!virGetLastError())
                virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                               _("resume operation failed"));
            goto cleanup;
        }
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, pausedReason);
    }

    VIR_DEBUG("Writing domain status to disk");
6268
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm, driver->caps) < 0)
6269 6270 6271 6272 6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283
        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
6284 6285 6286 6287
int
qemuProcessStart(virConnectPtr conn,
                 virQEMUDriverPtr driver,
                 virDomainObjPtr vm,
6288
                 virCPUDefPtr updatedCPU,
J
Jiri Denemark 已提交
6289 6290 6291 6292 6293 6294 6295 6296 6297 6298 6299 6300 6301 6302 6303 6304 6305 6306 6307 6308 6309 6310 6311 6312 6313 6314 6315
                 qemuDomainAsyncJob asyncJob,
                 const char *migrateFrom,
                 int migrateFd,
                 const char *migratePath,
                 virDomainSnapshotObjPtr snapshot,
                 virNetDevVPortProfileOp vmop,
                 unsigned int flags)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuProcessIncomingDefPtr incoming = NULL;
    unsigned int stopFlags;
    bool relabel = false;
    int ret = -1;
    int rv;

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

    virCheckFlagsGoto(VIR_QEMU_PROCESS_START_COLD |
                      VIR_QEMU_PROCESS_START_PAUSED |
                      VIR_QEMU_PROCESS_START_AUTODESTROY, cleanup);

6316 6317 6318
    if (!migrateFrom && !snapshot)
        flags |= VIR_QEMU_PROCESS_START_NEW;

6319 6320
    if (qemuProcessInit(driver, vm, updatedCPU,
                        asyncJob, !!migrateFrom, flags) < 0)
J
Jiri Denemark 已提交
6321 6322 6323
        goto cleanup;

    if (migrateFrom) {
6324
        incoming = qemuProcessIncomingDefNew(priv->qemuCaps, NULL, migrateFrom,
J
Jiri Denemark 已提交
6325 6326 6327 6328 6329
                                             migrateFd, migratePath);
        if (!incoming)
            goto stop;
    }

6330
    if (qemuProcessPrepareDomain(driver, vm, flags) < 0)
6331 6332
        goto stop;

6333
    if (qemuProcessPrepareHost(driver, vm, flags) < 0)
6334 6335
        goto stop;

J
Jiri Denemark 已提交
6336 6337
    if ((rv = qemuProcessLaunch(conn, driver, vm, asyncJob, incoming,
                                snapshot, vmop, flags)) < 0) {
6338
        if (rv == -2)
J
Jiri Denemark 已提交
6339 6340 6341 6342
            relabel = true;
        goto stop;
    }
    relabel = true;
6343

6344 6345
    if (incoming &&
        incoming->deferredURI &&
6346
        qemuMigrationDstRun(driver, vm, incoming->deferredURI, asyncJob) < 0)
J
Jiri Denemark 已提交
6347
        goto stop;
6348

6349
    if (qemuProcessFinishStartup(driver, vm, asyncJob,
6350 6351 6352 6353
                                 !(flags & VIR_QEMU_PROCESS_START_PAUSED),
                                 incoming ?
                                 VIR_DOMAIN_PAUSED_MIGRATION :
                                 VIR_DOMAIN_PAUSED_USER) < 0)
J
Jiri Denemark 已提交
6354
        goto stop;
6355

6356 6357 6358
    if (!incoming) {
        /* Keep watching qemu log for errors during incoming migration, otherwise
         * unset reporting errors from qemu log. */
6359
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
6360

6361 6362 6363 6364 6365 6366
        /* 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;
    }

6367 6368 6369
    ret = 0;

 cleanup:
6370
    qemuProcessIncomingDefFree(incoming);
6371
    return ret;
6372

J
Jiri Denemark 已提交
6373 6374 6375 6376 6377 6378
 stop:
    stopFlags = 0;
    if (!relabel)
        stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
    if (migrateFrom)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
6379
    if (priv->mon)
6380
        qemuMonitorSetDomainLog(priv->mon, NULL, NULL, NULL);
6381
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, asyncJob, stopFlags);
6382
    goto cleanup;
6383 6384 6385
}


6386
virCommandPtr
6387
qemuProcessCreatePretendCmd(virQEMUDriverPtr driver,
6388 6389
                            virDomainObjPtr vm,
                            const char *migrateURI,
6390
                            bool enableFips,
6391 6392 6393 6394 6395 6396 6397 6398 6399 6400
                            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;
6401
    flags |= VIR_QEMU_PROCESS_START_NEW;
6402

6403 6404
    if (qemuProcessInit(driver, vm, NULL, QEMU_ASYNC_JOB_NONE,
                        !!migrateURI, flags) < 0)
6405 6406
        goto cleanup;

6407
    if (qemuProcessPrepareDomain(driver, vm, flags) < 0)
6408 6409 6410
        goto cleanup;

    VIR_DEBUG("Building emulator command line");
6411
    cmd = qemuBuildCommandLine(driver,
6412
                               NULL,
6413
                               vm,
6414 6415 6416 6417
                               migrateURI,
                               NULL,
                               VIR_NETDEV_VPORT_PROFILE_OP_NO_OP,
                               standalone,
6418
                               enableFips,
6419
                               NULL,
6420
                               NULL);
6421 6422 6423 6424 6425 6426

 cleanup:
    return cmd;
}


6427
int
6428
qemuProcessKill(virDomainObjPtr vm, unsigned int flags)
6429
{
6430
    int ret;
6431

6432
    VIR_DEBUG("vm=%p name=%s pid=%lld flags=0x%x",
6433
              vm, vm->def->name,
M
Michal Privoznik 已提交
6434
              (long long) vm->pid, flags);
6435

6436 6437 6438 6439 6440
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
6441 6442
    }

6443
    if (flags & VIR_QEMU_PROCESS_KILL_NOWAIT) {
6444 6445 6446 6447 6448
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
6449

6450 6451
    ret = virProcessKillPainfully(vm->pid,
                                  !!(flags & VIR_QEMU_PROCESS_KILL_FORCE));
6452

6453
    return ret;
6454 6455 6456
}


6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467 6468 6469 6470 6471 6472 6473 6474 6475 6476 6477 6478 6479 6480 6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493 6494 6495
/**
 * 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;
}


6496
void qemuProcessStop(virQEMUDriverPtr driver,
6497
                     virDomainObjPtr vm,
6498
                     virDomainShutoffReason reason,
6499
                     qemuDomainAsyncJob asyncJob,
6500
                     unsigned int flags)
6501 6502 6503 6504 6505 6506
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
6507
    virNetDevVPortProfilePtr vport = NULL;
6508
    size_t i;
6509
    char *timestamp;
6510
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
6511

M
Michal Privoznik 已提交
6512
    VIR_DEBUG("Shutting down vm=%p name=%s id=%d pid=%lld, "
6513
              "reason=%s, asyncJob=%s, flags=0x%x",
6514
              vm, vm->def->name, vm->def->id,
M
Michal Privoznik 已提交
6515
              (long long) vm->pid,
6516 6517 6518
              virDomainShutoffReasonTypeToString(reason),
              qemuDomainAsyncJobTypeToString(asyncJob),
              flags);
6519

6520 6521 6522 6523
    /* This method is routinely used in clean up paths. Disable error
     * reporting so we don't squash a legit error. */
    orig_err = virSaveLastError();

6524 6525 6526 6527 6528 6529 6530 6531 6532 6533 6534 6535 6536 6537 6538
    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;
    }

6539
    qemuProcessBuildDestroyMemoryPaths(driver, vm, NULL, false);
6540

6541 6542
    vm->def->id = -1;

6543
    if (virAtomicIntDecAndTest(&driver->nactive) && driver->inhibitCallback)
6544 6545
        driver->inhibitCallback(false, driver->inhibitOpaque);

6546 6547
    /* Wake up anything waiting on domain condition */
    virDomainObjBroadcast(vm);
6548

6549
    if ((timestamp = virTimeStringNow()) != NULL) {
6550 6551 6552
        qemuDomainLogAppendMessage(driver, vm, "%s: shutting down, reason=%s\n",
                                   timestamp,
                                   virDomainShutoffReasonTypeToString(reason));
6553
        VIR_FREE(timestamp);
6554 6555
    }

6556 6557 6558
    /* Clear network bandwidth */
    virDomainClearNetBandwidth(vm);

6559 6560
    virDomainConfVMNWFilterTeardown(vm);

6561
    if (cfg->macFilter) {
6562
        def = vm->def;
6563
        for (i = 0; i < def->nnets; i++) {
6564 6565 6566
            virDomainNetDefPtr net = def->nets[i];
            if (net->ifname == NULL)
                continue;
6567 6568 6569
            ignore_value(ebtablesRemoveForwardAllowIn(driver->ebtables,
                                                      net->ifname,
                                                      &net->mac));
6570 6571 6572
        }
    }

6573
    virPortAllocatorRelease(priv->nbdPort);
6574
    priv->nbdPort = 0;
6575

D
Daniel P. Berrange 已提交
6576 6577 6578 6579
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
    }
6580
    priv->agentError = false;
D
Daniel P. Berrange 已提交
6581

6582
    if (priv->mon) {
6583
        qemuMonitorClose(priv->mon);
6584 6585
        priv->mon = NULL;
    }
6586 6587 6588 6589 6590 6591 6592 6593

    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
6594 6595 6596
    /* Remove the master key */
    qemuDomainMasterKeyRemove(priv);

6597 6598
    virFileDeleteTree(priv->libDir);
    virFileDeleteTree(priv->channelTargetDir);
6599

6600 6601 6602 6603 6604 6605
    ignore_value(virDomainChrDefForeach(vm->def,
                                        false,
                                        qemuProcessCleanupChardevDevice,
                                        NULL));


6606
    /* shut it off for sure */
6607 6608 6609
    ignore_value(qemuProcessKill(vm,
                                 VIR_QEMU_PROCESS_KILL_FORCE|
                                 VIR_QEMU_PROCESS_KILL_NOCHECK));
6610

6611 6612
    qemuDomainCleanupRun(driver, vm);

6613
    /* Stop autodestroy in case guest is restarted */
6614
    qemuProcessAutoDestroyRemove(driver, vm);
6615

6616 6617
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
6618
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
6619 6620

        /* we can't stop the operation even if the script raised an error */
6621 6622 6623
        ignore_value(virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
                                 VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                                 NULL, xml, NULL));
6624 6625 6626
        VIR_FREE(xml);
    }

6627 6628
    /* Reset Security Labels unless caller don't want us to */
    if (!(flags & VIR_QEMU_PROCESS_STOP_NO_RELABEL))
6629 6630 6631
        qemuSecurityRestoreAllLabel(driver, vm,
                                    !!(flags & VIR_QEMU_PROCESS_STOP_MIGRATED));

6632
    qemuSecurityReleaseLabel(driver->securityManager, vm->def);
6633

6634
    for (i = 0; i < vm->def->ndisks; i++) {
6635
        virDomainDeviceDef dev;
6636
        virDomainDiskDefPtr disk = vm->def->disks[i];
6637 6638 6639 6640

        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        ignore_value(qemuRemoveSharedDevice(driver, &dev, vm->def->name));
6641 6642
    }

6643
    /* Clear out dynamically assigned labels */
6644
    for (i = 0; i < vm->def->nseclabels; i++) {
6645
        if (vm->def->seclabels[i]->type == VIR_DOMAIN_SECLABEL_DYNAMIC)
6646 6647
            VIR_FREE(vm->def->seclabels[i]->label);
        VIR_FREE(vm->def->seclabels[i]->imagelabel);
6648 6649
    }

6650
    qemuHostdevReAttachDomainDevices(driver, vm->def);
6651 6652 6653 6654

    def = vm->def;
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
6655 6656 6657 6658
        vport = virDomainNetGetActualVirtPortProfile(net);

        switch (virDomainNetGetActualType(net)) {
        case VIR_DOMAIN_NET_TYPE_DIRECT:
6659
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
6660
                             net->ifname, &net->mac,
6661 6662
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
6663
                             virDomainNetGetActualVirtPortProfile(net),
6664
                             cfg->stateDir));
6665
            break;
6666 6667 6668 6669 6670 6671
        case VIR_DOMAIN_NET_TYPE_ETHERNET:
            if (net->ifname) {
                ignore_value(virNetDevTapDelete(net->ifname, net->backend.tap));
                VIR_FREE(net->ifname);
            }
            break;
6672 6673 6674 6675
        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))
6676
                ignore_value(virNetDevTapDelete(net->ifname, net->backend.tap));
6677 6678
#endif
            break;
6679 6680 6681 6682 6683 6684 6685 6686 6687 6688 6689
        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;
6690
        }
6691 6692 6693
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
6694 6695 6696 6697 6698 6699 6700 6701 6702
        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
6703

6704 6705
        /* kick the device out of the hostdev list too */
        virDomainNetRemoveHostdev(def, net);
6706
        virDomainNetReleaseActualDevice(vm->def, net);
6707
    }
6708

6709
 retry:
6710
    if ((ret = qemuRemoveCgroup(vm)) < 0) {
6711 6712 6713 6714 6715 6716 6717
        if (ret == -EBUSY && (retries++ < 5)) {
            usleep(200*1000);
            goto retry;
        }
        VIR_WARN("Failed to remove cgroup for %s",
                 vm->def->name);
    }
6718

6719 6720 6721 6722 6723 6724
    /* 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 */
    for (i = 0; i < vm->def->ncachetunes; i++)
        virResctrlAllocRemove(vm->def->cachetunes[i]->alloc);

6725 6726
    qemuProcessRemoveDomainStatus(driver, vm);

6727 6728
    /* Remove VNC and Spice ports from port reservation bitmap, but only if
       they were reserved by the driver (autoport=yes)
6729
    */
6730
    for (i = 0; i < vm->def->ngraphics; ++i) {
6731
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
6732 6733
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (graphics->data.vnc.autoport) {
6734
                virPortAllocatorRelease(graphics->data.vnc.port);
6735
            } else if (graphics->data.vnc.portReserved) {
6736
                virPortAllocatorRelease(graphics->data.spice.port);
6737 6738
                graphics->data.vnc.portReserved = false;
            }
6739
            if (graphics->data.vnc.websocketGenerated) {
6740
                virPortAllocatorRelease(graphics->data.vnc.websocket);
6741 6742 6743
                graphics->data.vnc.websocketGenerated = false;
                graphics->data.vnc.websocket = -1;
            } else if (graphics->data.vnc.websocket) {
6744
                virPortAllocatorRelease(graphics->data.vnc.websocket);
6745
            }
6746
        }
6747 6748
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (graphics->data.spice.autoport) {
6749 6750
                virPortAllocatorRelease(graphics->data.spice.port);
                virPortAllocatorRelease(graphics->data.spice.tlsPort);
6751 6752
            } else {
                if (graphics->data.spice.portReserved) {
6753
                    virPortAllocatorRelease(graphics->data.spice.port);
6754 6755 6756 6757
                    graphics->data.spice.portReserved = false;
                }

                if (graphics->data.spice.tlsPortReserved) {
6758
                    virPortAllocatorRelease(graphics->data.spice.tlsPort);
6759 6760 6761
                    graphics->data.spice.tlsPortReserved = false;
                }
            }
6762
        }
6763 6764
    }

6765
    vm->taint = 0;
6766
    vm->pid = -1;
J
Jiri Denemark 已提交
6767
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
6768 6769
    for (i = 0; i < vm->def->niothreadids; i++)
        vm->def->iothreadids[i]->thread_id = 0;
6770

6771 6772
    /* clear all private data entries which are no longer needed */
    qemuDomainObjPrivateDataClear(priv);
6773

6774
    /* The "release" hook cleans up additional resources */
6775
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
6776
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
6777 6778 6779

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
6780 6781
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
6782 6783 6784
        VIR_FREE(xml);
    }

6785
    virDomainObjRemoveTransientDef(vm);
6786

6787 6788 6789 6790 6791
 endjob:
    if (asyncJob != QEMU_ASYNC_JOB_NONE)
        qemuDomainObjEndJob(driver, vm);

 cleanup:
6792 6793 6794 6795
    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
6796
    virObjectUnref(cfg);
6797
}
6798 6799


6800 6801 6802 6803 6804 6805 6806 6807 6808 6809 6810 6811 6812 6813 6814 6815 6816 6817 6818 6819 6820 6821 6822 6823 6824 6825 6826 6827 6828 6829 6830 6831 6832 6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843 6844 6845 6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858 6859 6860 6861 6862 6863 6864 6865 6866 6867 6868 6869 6870 6871 6872 6873 6874 6875 6876 6877 6878 6879 6880 6881 6882 6883 6884 6885 6886 6887 6888 6889 6890 6891 6892 6893 6894 6895 6896 6897 6898 6899 6900 6901 6902 6903 6904 6905 6906 6907 6908 6909 6910 6911
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;

    VIR_DEBUG("Beginning VM attach process");

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

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

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

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

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

    if (virFileMakePath(cfg->logDir) < 0) {
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
                             cfg->logDir);
        goto error;
    }

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

    vm->pid = pid;

    VIR_DEBUG("Detect security driver config");
    sec_managers = qemuSecurityGetNested(driver->securityManager);
    if (sec_managers == NULL)
        goto error;

    for (i = 0; sec_managers[i]; i++) {
        seclabelgen = false;
        model = qemuSecurityGetModel(sec_managers[i]);
        seclabeldef = virDomainDefGetSecurityLabelDef(vm->def, model);
        if (seclabeldef == NULL) {
            if (!(seclabeldef = virSecurityLabelDefNew(model)))
                goto error;
            seclabelgen = true;
        }
        seclabeldef->type = VIR_DOMAIN_SECLABEL_STATIC;
        if (VIR_ALLOC(seclabel) < 0)
            goto error;
        if (qemuSecurityGetProcessLabel(sec_managers[i], vm->def,
                                        vm->pid, seclabel) < 0)
            goto error;

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

        if (VIR_STRDUP(seclabeldef->label, seclabel->label) < 0)
            goto error;
        VIR_FREE(seclabel);

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

    if (qemuSecurityCheckAllLabel(driver->securityManager, vm->def) < 0)
        goto error;
    if (qemuSecurityGenLabel(driver->securityManager, vm->def) < 0)
        goto error;

    if (qemuDomainPerfRestart(vm) < 0)
        goto error;

    VIR_DEBUG("Creating domain log file");
    if (!(logCtxt = qemuDomainLogContextNew(driver, vm,
                                            QEMU_DOMAIN_LOG_CONTEXT_MODE_ATTACH)))
        goto error;

    VIR_DEBUG("Determining emulator version");
    virObjectUnref(priv->qemuCaps);
6912
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
6913 6914 6915 6916 6917 6918 6919 6920 6921 6922 6923
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
        goto error;

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

    priv->gotShutdown = false;

6924 6925 6926 6927
    /* Attaching to running QEMU so we need to detect whether it was started
     * with -no-reboot. */
    qemuProcessPrepareAllowReboot(vm);

6928 6929 6930 6931 6932 6933 6934 6935 6936 6937 6938 6939 6940 6941 6942 6943 6944 6945 6946 6947 6948 6949 6950 6951 6952 6953 6954 6955 6956 6957 6958 6959 6960 6961 6962 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 7005 7006 7007 7008 7009 7010 7011 7012 7013 7014 7015 7016 7017 7018 7019 7020 7021 7022 7023 7024 7025 7026 7027 7028 7029 7030 7031 7032 7033 7034 7035 7036 7037 7038 7039 7040 7041 7042 7043 7044 7045 7046
    /*
     * 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;
    if (qemuMonitorGetVirtType(priv->mon, &vm->def->virtType) < 0)
        goto exit_monitor;
    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;
}


7047
static void
7048 7049 7050 7051 7052 7053 7054 7055 7056 7057 7058 7059 7060 7061 7062 7063 7064 7065 7066 7067 7068 7069 7070
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)
7071
        return;
7072 7073 7074 7075 7076 7077 7078 7079 7080 7081 7082

    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);

7083 7084
    qemuDomainObjEndJob(driver, dom);

7085 7086 7087 7088 7089 7090 7091 7092 7093 7094 7095 7096 7097 7098 7099 7100 7101 7102 7103 7104 7105 7106 7107 7108 7109 7110 7111 7112 7113 7114 7115 7116 7117 7118 7119 7120 7121 7122 7123 7124 7125 7126 7127 7128 7129 7130 7131 7132 7133 7134 7135 7136 7137 7138 7139 7140 7141 7142 7143 7144 7145 7146 7147 7148 7149 7150 7151 7152 7153 7154 7155 7156 7157 7158 7159 7160 7161 7162 7163 7164 7165 7166 7167 7168
    qemuDomainEventQueue(driver, event);
}

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;
    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;

        if (!(info = virHashLookup(table, disk->info.alias)))
            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;
}


7169 7170 7171 7172 7173
static int
qemuProcessRefreshCPU(virQEMUDriverPtr driver,
                      virDomainObjPtr vm)
{
    virCapsPtr caps = virQEMUDriverGetCapabilities(driver, false);
7174
    qemuDomainObjPrivatePtr priv = vm->privateData;
7175
    virCPUDefPtr host = NULL;
7176
    virCPUDefPtr cpu = NULL;
7177 7178 7179 7180 7181 7182 7183 7184 7185 7186 7187 7188 7189 7190 7191 7192 7193 7194 7195 7196 7197
    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;

7198 7199 7200 7201 7202 7203 7204
        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)
7205 7206 7207 7208
            goto cleanup;

        if (qemuProcessUpdateCPU(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
            goto cleanup;
7209 7210 7211 7212 7213 7214 7215 7216
    } 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;
7217 7218 7219 7220 7221
    }

    ret = 0;

 cleanup:
7222
    virCPUDefFree(cpu);
7223 7224 7225 7226 7227 7228
    virCPUDefFree(host);
    virObjectUnref(caps);
    return ret;
}


7229 7230 7231 7232 7233 7234 7235 7236 7237 7238 7239 7240 7241 7242 7243 7244 7245 7246 7247 7248 7249 7250 7251 7252
struct qemuProcessReconnectData {
    virQEMUDriverPtr driver;
    virDomainObjPtr obj;
};
/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
 *
 * This function also inherits a locked and ref'd domain object.
 *
 * This function needs to:
 * 1. Enter job
 * 1. just before monitor reconnect do lightweight MonitorEnter
 *    (increase VM refcount and unlock VM)
 * 2. reconnect to monitor
 * 3. do lightweight MonitorExit (lock VM)
 * 4. continue reconnect process
 * 5. EndJob
 *
 * We can't do normal MonitorEnter & MonitorExit because these two lock the
 * monitor lock, which does not exists in this early phase.
 */
static void
qemuProcessReconnect(void *opaque)
7253
{
7254 7255 7256 7257 7258 7259 7260 7261
    struct qemuProcessReconnectData *data = opaque;
    virQEMUDriverPtr driver = data->driver;
    virDomainObjPtr obj = data->obj;
    qemuDomainObjPrivatePtr priv;
    struct qemuDomainJobObj oldjob;
    int state;
    int reason;
    virQEMUDriverConfigPtr cfg;
7262
    size_t i;
7263 7264
    unsigned int stopFlags = 0;
    bool jobStarted = false;
7265
    virCapsPtr caps = NULL;
7266

7267
    VIR_FREE(data);
7268

7269 7270 7271 7272 7273 7274
    qemuDomainObjRestoreJob(obj, &oldjob);
    if (oldjob.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;

    cfg = virQEMUDriverGetConfig(driver);
    priv = obj->privateData;
7275

7276
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
7277
        goto error;
7278

7279
    if (qemuDomainObjBeginJob(driver, obj, QEMU_JOB_MODIFY) < 0)
7280
        goto error;
7281
    jobStarted = true;
7282

7283 7284 7285 7286
    /* 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;
7287

7288 7289 7290
    /* Restore the masterKey */
    if (qemuDomainMasterKeyReadFile(priv) < 0)
        goto error;
7291

7292 7293 7294 7295
    /* 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);

7296 7297 7298 7299
    VIR_DEBUG("Reconnect monitor to %p '%s'", obj, obj->def->name);

    /* XXX check PID liveliness & EXE path */
    if (qemuConnectMonitor(driver, obj, QEMU_ASYNC_JOB_NONE, NULL) < 0)
7300
        goto error;
7301

7302
    if (qemuHostdevUpdateActiveDomainDevices(driver, obj->def) < 0)
7303
        goto error;
7304

7305 7306 7307 7308
    priv->machineName = qemuDomainGetMachineName(obj);
    if (!priv->machineName)
        goto error;

7309
    if (qemuConnectCgroup(obj) < 0)
7310
        goto error;
7311

7312
    if (qemuDomainPerfRestart(obj) < 0)
7313
        goto error;
7314

7315 7316 7317 7318
    /* XXX: Need to change as long as lock is introduced for
     * qemu_driver->sharedDevices.
     */
    for (i = 0; i < obj->def->ndisks; i++) {
7319
        virDomainDiskDefPtr disk = obj->def->disks[i];
7320
        virDomainDeviceDef dev;
7321

7322
        if (virDomainDiskTranslateSourcePool(disk) < 0)
7323
            goto error;
7324

7325 7326 7327 7328 7329 7330 7331 7332 7333 7334
        /* backing chains need to be refreshed only if they could change */
        if (priv->reconnectBlockjobs != VIR_TRISTATE_BOOL_NO) {
            /* This should be the only place that calls
             * qemuDomainDetermineDiskChain with @report_broken == false
             * to guarantee best-effort domain reconnect */
            if (qemuDomainDetermineDiskChain(driver, obj, disk, true, false) < 0)
                goto error;
        } else {
            VIR_DEBUG("skipping backing chain detection for '%s'", disk->dst);
        }
7335

7336
        dev.type = VIR_DOMAIN_DEVICE_DISK;
7337
        dev.data.disk = disk;
7338
        if (qemuAddSharedDevice(driver, &dev, obj->def->name) < 0)
7339 7340 7341 7342
            goto error;
    }

    for (i = 0; i < obj->def->ngraphics; i++) {
7343
        if (qemuProcessGraphicsReservePorts(obj->def->graphics[i], true) < 0)
7344
            goto error;
7345
    }
7346

7347
    if (qemuProcessUpdateState(driver, obj) < 0)
7348 7349
        goto error;

7350 7351 7352 7353 7354 7355
    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);
7356
        goto error;
7357
    }
7358

7359 7360 7361 7362
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
    if (!priv->qemuCaps &&
7363
        !(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
7364 7365
                                                      obj->def->emulator,
                                                      obj->def->os.machine)))
7366
        goto error;
7367

7368 7369 7370 7371 7372 7373 7374 7375 7376 7377 7378 7379
    /* 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;
    }
7380

7381
    if (qemuProcessBuildDestroyMemoryPaths(driver, obj, NULL, true) < 0)
7382
        goto error;
7383

7384 7385
    if ((qemuDomainAssignAddresses(obj->def, priv->qemuCaps,
                                   driver, obj, false)) < 0) {
7386
        goto error;
7387
    }
7388

7389 7390 7391 7392 7393 7394
    /* if domain requests security driver we haven't loaded, report error, but
     * do not kill the domain
     */
    ignore_value(qemuSecurityCheckAllLabel(driver->securityManager,
                                           obj->def));

7395 7396
    if (qemuProcessRefreshCPU(driver, obj) < 0)
        goto error;
7397

7398
    if (qemuDomainRefreshVcpuInfo(driver, obj, QEMU_ASYNC_JOB_NONE, true) < 0)
7399
        goto error;
7400

7401
    qemuDomainVcpuPersistOrder(obj->def);
7402

7403 7404
    if (qemuSecurityReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
        goto error;
7405

7406 7407 7408
    qemuProcessNotifyNets(obj->def);

    if (qemuProcessFiltersInstantiate(obj->def))
7409
        goto error;
7410

7411
    if (qemuProcessRefreshDisks(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
7412
        goto error;
D
Daniel P. Berrange 已提交
7413

7414
    if (qemuBlockNodeNamesDetect(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
7415 7416
        goto error;

7417
    if (qemuRefreshVirtioChannelState(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
7418 7419
        goto error;

7420 7421 7422 7423
    /* 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)
7424
        goto error;
7425

7426
    if (qemuProcessRecoverJob(driver, obj, &oldjob, &stopFlags) < 0)
7427
        goto error;
7428

7429 7430
    if (qemuProcessUpdateDevices(driver, obj) < 0)
        goto error;
7431

7432 7433 7434 7435 7436
    qemuProcessReconnectCheckMemAliasOrderMismatch(obj);

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

7437 7438 7439 7440 7441 7442
    for (i = 0; i < obj->def->ncachetunes; i++) {
        if (virResctrlAllocDeterminePath(obj->def->cachetunes[i]->alloc,
                                         priv->machineName) < 0)
            goto error;
    }

7443 7444
    /* update domain state XML with possibly updated state in virDomainObj */
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, obj, driver->caps) < 0)
7445
        goto error;
7446

7447 7448
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
7449
        char *xml = qemuDomainDefFormatXML(driver, obj->def, 0);
7450 7451
        int hookret;

7452 7453
        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, obj->def->name,
                              VIR_HOOK_QEMU_OP_RECONNECT, VIR_HOOK_SUBOP_BEGIN,
7454 7455 7456 7457 7458 7459 7460
                              NULL, xml, NULL);
        VIR_FREE(xml);

        /*
         * If the script raised an error abort the launch
         */
        if (hookret < 0)
7461
            goto error;
7462 7463
    }

7464 7465
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
        driver->inhibitCallback(true, driver->inhibitOpaque);
7466

7467
 cleanup:
7468 7469 7470
    if (jobStarted) {
        if (!virDomainObjIsActive(obj))
            qemuDomainRemoveInactive(driver, obj);
7471
        qemuDomainObjEndJob(driver, obj);
7472 7473 7474 7475
    } else {
        if (!virDomainObjIsActive(obj))
            qemuDomainRemoveInactiveJob(driver, obj);
    }
7476
    virDomainObjEndAPI(&obj);
7477
    virObjectUnref(cfg);
7478
    virObjectUnref(caps);
7479 7480
    virNWFilterUnlockFilterUpdates();
    return;
7481

7482 7483 7484 7485 7486 7487 7488 7489 7490 7491 7492 7493 7494 7495 7496 7497 7498 7499 7500 7501 7502
 error:
    if (virDomainObjIsActive(obj)) {
        /* We can't get the monitor back, so must kill the VM
         * to remove danger of it ending up running twice if
         * user tries to start it again later
         */
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NO_SHUTDOWN)) {
            /* If we couldn't get the monitor and qemu supports
             * no-shutdown, we can safely say that the domain
             * crashed ... */
            state = VIR_DOMAIN_SHUTOFF_CRASHED;
        } else {
            /* ... but if it doesn't we can't say what the state
             * really is and FAILED means "failed to start" */
            state = VIR_DOMAIN_SHUTOFF_UNKNOWN;
        }
        /* 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);
7503
    }
7504
    goto cleanup;
7505
}
7506

7507 7508 7509
static int
qemuProcessReconnectHelper(virDomainObjPtr obj,
                           void *opaque)
7510
{
7511 7512 7513
    virThread thread;
    struct qemuProcessReconnectData *src = opaque;
    struct qemuProcessReconnectData *data;
7514

7515 7516 7517
    /* If the VM was inactive, we don't need to reconnect */
    if (!obj->pid)
        return 0;
7518

7519 7520
    if (VIR_ALLOC(data) < 0)
        return -1;
7521

7522 7523
    memcpy(data, src, sizeof(*data));
    data->obj = obj;
7524

7525 7526
    virNWFilterReadLockFilterUpdates();

7527 7528 7529 7530
    /* this lock and reference will be eventually transferred to the thread
     * that handles the reconnect */
    virObjectLock(obj);
    virObjectRef(obj);
7531

7532 7533 7534 7535 7536 7537 7538 7539 7540 7541 7542
    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);
7543
        qemuDomainRemoveInactiveJob(src->driver, obj);
7544

7545
        virDomainObjEndAPI(&obj);
7546
        virNWFilterUnlockFilterUpdates();
7547 7548
        VIR_FREE(data);
        return -1;
7549 7550
    }

7551 7552
    return 0;
}
7553

7554 7555 7556 7557 7558 7559 7560
/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
7561
qemuProcessReconnectAll(virQEMUDriverPtr driver)
7562
{
7563
    struct qemuProcessReconnectData data = {.driver = driver};
7564
    virDomainObjListForEach(driver->domains, qemuProcessReconnectHelper, &data);
7565
}