qemu_process.c 158.9 KB
Newer Older
1
/*
2
 * qemu_process.c: QEMU process management
3
 *
4
 * Copyright (C) 2006-2014 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23 24 25 26 27
 *
 */

#include <config.h>

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

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

46
#include "cpu/cpu.h"
47
#include "datatypes.h"
48
#include "virlog.h"
49
#include "virerror.h"
50
#include "viralloc.h"
51
#include "virhook.h"
E
Eric Blake 已提交
52
#include "virfile.h"
53
#include "virpidfile.h"
54 55
#include "c-ctype.h"
#include "nodeinfo.h"
56
#include "domain_audit.h"
57
#include "domain_nwfilter.h"
58
#include "locking/domain_lock.h"
59
#include "network/bridge_driver.h"
60
#include "viruuid.h"
61
#include "virprocess.h"
62
#include "virtime.h"
A
Ansis Atteka 已提交
63
#include "virnetdevtap.h"
64
#include "virbitmap.h"
65
#include "viratomic.h"
66
#include "virnuma.h"
67
#include "virstring.h"
68
#include "virhostdev.h"
69 70 71

#define VIR_FROM_THIS VIR_FROM_QEMU

72 73
VIR_LOG_INIT("qemu.qemu_process");

74
#define START_POSTFIX ": starting up\n"
75
#define ATTACH_POSTFIX ": attaching\n"
76 77 78
#define SHUTDOWN_POSTFIX ": shutting down\n"

/**
79
 * qemuProcessRemoveDomainStatus
80 81 82 83 84 85
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
86
qemuProcessRemoveDomainStatus(virQEMUDriverPtr driver,
87 88 89 90
                              virDomainObjPtr vm)
{
    char ebuf[1024];
    char *file = NULL;
91
    qemuDomainObjPrivatePtr priv = vm->privateData;
92 93
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;
94

95
    if (virAsprintf(&file, "%s/%s.xml", cfg->stateDir, vm->def->name) < 0)
96
        goto cleanup;
97 98 99 100 101 102

    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);

103 104 105
    if (priv->pidfile &&
        unlink(priv->pidfile) < 0 &&
        errno != ENOENT)
106 107 108
        VIR_WARN("Failed to remove PID file for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));

109
    ret = 0;
110
 cleanup:
111 112
    virObjectUnref(cfg);
    return ret;
113 114 115 116
}


/* XXX figure out how to remove this */
117
extern virQEMUDriverPtr qemu_driver;
118

D
Daniel P. Berrange 已提交
119 120 121 122 123 124 125
/*
 * 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
126
qemuProcessHandleAgentEOF(qemuAgentPtr agent,
D
Daniel P. Berrange 已提交
127 128 129 130 131 132
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

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

133
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
134 135

    priv = vm->privateData;
136 137 138 139 140 141 142 143 144 145 146 147

    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;
    }

    priv->agent = NULL;
D
Daniel P. Berrange 已提交
148

149
    virObjectUnlock(vm);
150 151

    qemuAgentClose(agent);
152 153
    return;

154
 unlock:
155 156
    virObjectUnlock(vm);
    return;
D
Daniel P. Berrange 已提交
157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173
}


/*
 * 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);

174
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
175 176 177 178 179

    priv = vm->privateData;

    priv->agentError = true;

180
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
181 182 183 184 185
}

static void qemuProcessHandleAgentDestroy(qemuAgentPtr agent,
                                          virDomainObjPtr vm)
{
186 187
    VIR_DEBUG("Received destroy agent=%p vm=%p", agent, vm);

188
    virObjectUnref(vm);
D
Daniel P. Berrange 已提交
189 190 191 192 193 194 195 196 197 198 199 200 201
}


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

static virDomainChrSourceDefPtr
qemuFindAgentConfig(virDomainDefPtr def)
{
    virDomainChrSourceDefPtr config = NULL;
202
    size_t i;
D
Daniel P. Berrange 已提交
203

204
    for (i = 0; i < def->nchannels; i++) {
D
Daniel P. Berrange 已提交
205 206 207 208 209
        virDomainChrDefPtr channel = def->channels[i];

        if (channel->targetType != VIR_DOMAIN_CHR_CHANNEL_TARGET_TYPE_VIRTIO)
            continue;

210
        if (STREQ_NULLABLE(channel->target.name, "org.qemu.guest_agent.0")) {
D
Daniel P. Berrange 已提交
211 212 213 214 215 216 217 218 219
            config = &channel->source;
            break;
        }
    }

    return config;
}

static int
220
qemuConnectAgent(virQEMUDriverPtr driver, virDomainObjPtr vm)
D
Daniel P. Berrange 已提交
221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    qemuAgentPtr agent = NULL;
    virDomainChrSourceDefPtr config = qemuFindAgentConfig(vm->def);

    if (!config)
        return 0;

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

    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted while the agent is active */
239
    virObjectRef(vm);
D
Daniel P. Berrange 已提交
240 241

    ignore_value(virTimeMillisNow(&priv->agentStart));
242
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
243 244 245 246 247

    agent = qemuAgentOpen(vm,
                          config,
                          &agentCallbacks);

248
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
249 250
    priv->agentStart = 0;

251 252 253 254 255 256 257 258 259 260 261
    if (agent == NULL)
        virObjectUnref(vm);

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

D
Daniel P. Berrange 已提交
262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
    if (virSecurityManagerClearSocketLabel(driver->securityManager,
                                           vm->def) < 0) {
        VIR_ERROR(_("Failed to clear security context for agent for %s"),
                  vm->def->name);
        goto cleanup;
    }


    priv->agent = agent;

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

    ret = 0;

279
 cleanup:
D
Daniel P. Berrange 已提交
280 281 282 283
    return ret;
}


284
/*
285
 * This is a callback registered with a qemuMonitorPtr instance,
286 287 288 289 290 291
 * and to be invoked when the monitor console hits an end of file
 * condition, or error, thus indicating VM shutdown should be
 * performed
 */
static void
qemuProcessHandleMonitorEOF(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
292 293
                            virDomainObjPtr vm,
                            void *opaque)
294
{
295
    virQEMUDriverPtr driver = opaque;
296
    virObjectEventPtr event = NULL;
297
    qemuDomainObjPrivatePtr priv;
298 299 300
    int eventReason = VIR_DOMAIN_EVENT_STOPPED_SHUTDOWN;
    int stopReason = VIR_DOMAIN_SHUTOFF_SHUTDOWN;
    const char *auditReason = "shutdown";
301 302 303

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

304
    virObjectLock(vm);
305

306 307 308 309 310 311 312
    priv = vm->privateData;

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

313 314
    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Domain %p is not active, ignoring EOF", vm);
315
        goto unlock;
316 317
    }

318
    if (priv->monJSON && !priv->gotShutdown) {
319 320
        VIR_DEBUG("Monitor connection to '%s' closed without SHUTDOWN event; "
                  "assuming the domain crashed", vm->def->name);
321
        eventReason = VIR_DOMAIN_EVENT_STOPPED_FAILED;
322
        stopReason = VIR_DOMAIN_SHUTOFF_CRASHED;
323
        auditReason = "failed";
324 325
    }

326
    event = virDomainEventLifecycleNewFromObj(vm,
327
                                     VIR_DOMAIN_EVENT_STOPPED,
328
                                     eventReason);
329
    qemuProcessStop(driver, vm, stopReason, 0);
330
    virDomainAuditStop(vm, auditReason);
331

332
    if (!vm->persistent) {
333
        qemuDomainRemoveInactive(driver, vm);
334 335 336
        goto cleanup;
    }

337
 unlock:
338
    virObjectUnlock(vm);
339

340
 cleanup:
341
    if (event)
342
        qemuDomainEventQueue(driver, event);
343 344 345 346 347 348 349 350 351 352 353
}


/*
 * 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,
354 355
                              virDomainObjPtr vm,
                              void *opaque)
356
{
357
    virQEMUDriverPtr driver = opaque;
358
    virObjectEventPtr event = NULL;
359 360 361

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

362
    virObjectLock(vm);
363

364
    ((qemuDomainObjPrivatePtr) vm->privateData)->monError = true;
365 366 367 368
    event = virDomainEventControlErrorNewFromObj(vm);
    if (event)
        qemuDomainEventQueue(driver, event);

369
    virObjectUnlock(vm);
370 371 372 373 374 375 376
}


static virDomainDiskDefPtr
qemuProcessFindDomainDiskByPath(virDomainObjPtr vm,
                                const char *path)
{
377
    int idx = virDomainDiskIndexByName(vm->def, path, true);
378

379 380
    if (idx >= 0)
        return vm->def->disks[idx];
381

382 383 384
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with path %s"),
                   path);
385 386 387 388 389 390 391
    return NULL;
}

static virDomainDiskDefPtr
qemuProcessFindDomainDiskByAlias(virDomainObjPtr vm,
                                 const char *alias)
{
392
    size_t i;
393 394 395 396 397 398 399 400 401 402 403 404

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

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

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

405 406 407
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with alias %s"),
                   alias);
408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423
    return NULL;
}

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

424
    if (!disk->src->encryption) {
425 426
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("disk %s does not have any encryption information"),
427
                       disk->src->path);
428 429
        return -1;
    }
430
    enc = disk->src->encryption;
431 432

    if (!conn) {
433 434
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("cannot find secrets without a connection"));
435 436 437 438
        goto cleanup;
    }

    if (conn->secretDriver == NULL ||
439 440
        conn->secretDriver->secretLookupByUUID == NULL ||
        conn->secretDriver->secretGetValue == NULL) {
441 442
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("secret storage not supported"));
443 444 445 446 447 448 449
        goto cleanup;
    }

    if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW ||
        enc->nsecrets != 1 ||
        enc->secrets[0]->type !=
        VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE) {
450
        virReportError(VIR_ERR_XML_ERROR,
451 452
                       _("invalid <encryption> for volume %s"),
                       virDomainDiskGetSource(disk));
453 454 455
        goto cleanup;
    }

456 457
    secret = conn->secretDriver->secretLookupByUUID(conn,
                                                    enc->secrets[0]->uuid);
458 459
    if (secret == NULL)
        goto cleanup;
460 461
    data = conn->secretDriver->secretGetValue(secret, &size, 0,
                                              VIR_SECRET_GET_VALUE_INTERNAL_CALL);
462
    virObjectUnref(secret);
463 464 465 466 467 468
    if (data == NULL)
        goto cleanup;

    if (memchr(data, '\0', size) != NULL) {
        memset(data, 0, size);
        VIR_FREE(data);
469 470
        virReportError(VIR_ERR_XML_ERROR,
                       _("format='qcow' passphrase for %s must not contain a "
471
                         "'\\0'"), virDomainDiskGetSource(disk));
472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490
        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;

491
 cleanup:
492 493 494 495 496 497 498 499 500
    return ret;
}

static int
qemuProcessFindVolumeQcowPassphrase(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virConnectPtr conn,
                                    virDomainObjPtr vm,
                                    const char *path,
                                    char **secretRet,
501 502
                                    size_t *secretLen,
                                    void *opaque ATTRIBUTE_UNUSED)
503 504 505 506
{
    virDomainDiskDefPtr disk;
    int ret = -1;

507
    virObjectLock(vm);
508 509 510 511 512 513 514
    disk = qemuProcessFindDomainDiskByPath(vm, path);

    if (!disk)
        goto cleanup;

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

515
 cleanup:
516
    virObjectUnlock(vm);
517 518 519 520 521 522
    return ret;
}


static int
qemuProcessHandleReset(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
523 524
                       virDomainObjPtr vm,
                       void *opaque)
525
{
526
    virQEMUDriverPtr driver = opaque;
527
    virObjectEventPtr event;
528
    qemuDomainObjPrivatePtr priv;
529

530
    virObjectLock(vm);
531

532
    event = virDomainEventRebootNewFromObj(vm);
533 534 535 536
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);

537
    virObjectUnlock(vm);
538

539
    if (event)
540 541 542 543 544 545
        qemuDomainEventQueue(driver, event);

    return 0;
}


546 547 548 549 550 551 552 553 554 555 556
/*
 * 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)
{
557
    virQEMUDriverPtr driver = qemu_driver;
558 559
    virDomainObjPtr vm = opaque;
    qemuDomainObjPrivatePtr priv = vm->privateData;
560
    virObjectEventPtr event = NULL;
561
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
562
    virDomainRunningReason reason = VIR_DOMAIN_RUNNING_BOOTED;
563 564
    int ret = -1;
    VIR_DEBUG("vm=%p", vm);
565
    virObjectLock(vm);
566
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
567 568 569
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
570 571
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
572 573 574
        goto endjob;
    }

575
    qemuDomainObjEnterMonitor(driver, vm);
576
    if (qemuMonitorSystemReset(priv->mon) < 0) {
577
        qemuDomainObjExitMonitor(driver, vm);
578 579
        goto endjob;
    }
580
    qemuDomainObjExitMonitor(driver, vm);
581 582

    if (!virDomainObjIsActive(vm)) {
583 584
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
585 586 587
        goto endjob;
    }

588 589 590
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_CRASHED)
        reason = VIR_DOMAIN_RUNNING_CRASHED;

591
    if (qemuProcessStartCPUs(driver, vm, NULL,
592
                             reason,
593
                             QEMU_ASYNC_JOB_NONE) < 0) {
594
        if (virGetLastError() == NULL)
595 596
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
597 598
        goto endjob;
    }
599
    priv->gotShutdown = false;
600
    event = virDomainEventLifecycleNewFromObj(vm,
601 602 603
                                     VIR_DOMAIN_EVENT_RESUMED,
                                     VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

604 605 606 607 608
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

609 610
    ret = 0;

611
 endjob:
612
    if (!qemuDomainObjEndJob(driver, vm))
613 614
        vm = NULL;

615
 cleanup:
616
    if (vm) {
617
        if (ret == -1) {
618
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_FORCE));
619
        }
620
        if (virObjectUnref(vm))
621
            virObjectUnlock(vm);
622 623 624
    }
    if (event)
        qemuDomainEventQueue(driver, event);
625
    virObjectUnref(cfg);
626 627 628
}


629
void
630
qemuProcessShutdownOrReboot(virQEMUDriverPtr driver,
631
                            virDomainObjPtr vm)
632
{
633 634 635
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
636
        qemuDomainSetFakeReboot(driver, vm, false);
637
        virObjectRef(vm);
638 639 640 641 642
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
643
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
644
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
645
            virObjectUnref(vm);
646 647
        }
    } else {
648
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
649
    }
650
}
651

652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679

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);
    if (event)
        qemuDomainEventQueue(driver, event);

    return 0;
}


680 681
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
682 683
                          virDomainObjPtr vm,
                          void *opaque)
684
{
685
    virQEMUDriverPtr driver = opaque;
686
    qemuDomainObjPrivatePtr priv;
687
    virObjectEventPtr event = NULL;
688
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
689

690 691
    VIR_DEBUG("vm=%p", vm);

692
    virObjectLock(vm);
693 694 695 696 697 698

    priv = vm->privateData;
    if (priv->gotShutdown) {
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
699 700 701 702
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
703 704 705 706 707 708 709 710
    }
    priv->gotShutdown = true;

    VIR_DEBUG("Transitioned guest %s to shutdown state",
              vm->def->name);
    virDomainObjSetState(vm,
                         VIR_DOMAIN_SHUTDOWN,
                         VIR_DOMAIN_SHUTDOWN_UNKNOWN);
711
    event = virDomainEventLifecycleNewFromObj(vm,
712 713 714
                                     VIR_DOMAIN_EVENT_SHUTDOWN,
                                     VIR_DOMAIN_EVENT_SHUTDOWN_FINISHED);

715
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
716 717 718 719
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

720 721 722
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

723 724
    qemuProcessShutdownOrReboot(driver, vm);

725
 unlock:
726
    virObjectUnlock(vm);
727
    if (event)
728
        qemuDomainEventQueue(driver, event);
729
    virObjectUnref(cfg);
730

731 732 733 734 735 736
    return 0;
}


static int
qemuProcessHandleStop(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
737 738
                      virDomainObjPtr vm,
                      void *opaque)
739
{
740
    virQEMUDriverPtr driver = opaque;
741
    virObjectEventPtr event = NULL;
742
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
743

744
    virObjectLock(vm);
J
Jiri Denemark 已提交
745
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
746
        qemuDomainObjPrivatePtr priv = vm->privateData;
747

748
        if (priv->gotShutdown) {
749 750
            VIR_DEBUG("Ignoring STOP event after SHUTDOWN");
            goto unlock;
751 752
        }

753 754
        VIR_DEBUG("Transitioned guest %s to paused state",
                  vm->def->name);
755

756
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_UNKNOWN);
757
        event = virDomainEventLifecycleNewFromObj(vm,
758 759 760
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);

761 762 763 764 765
        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));

766
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
767 768 769
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
770
    }
771

772
 unlock:
773
    virObjectUnlock(vm);
774
    if (event)
775
        qemuDomainEventQueue(driver, event);
776
    virObjectUnref(cfg);
777 778 779 780 781

    return 0;
}


782 783
static int
qemuProcessHandleResume(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
784 785
                        virDomainObjPtr vm,
                        void *opaque)
786
{
787
    virQEMUDriverPtr driver = opaque;
788
    virObjectEventPtr event = NULL;
789
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
790

791
    virObjectLock(vm);
792 793 794 795 796 797 798 799 800 801 802 803 804
    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);
805
        event = virDomainEventLifecycleNewFromObj(vm,
806 807 808 809
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

        VIR_DEBUG("Using lock state '%s' on resume event", NULLSTR(priv->lockState));
810
        if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
811 812 813 814 815 816 817 818 819
                                       vm, priv->lockState) < 0) {
            /* Don't free priv->lockState on error, because we need
             * to make sure we have state still present if the user
             * tries to resume again
             */
            goto unlock;
        }
        VIR_FREE(priv->lockState);

820
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
821 822 823 824 825
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
    }

826
 unlock:
827
    virObjectUnlock(vm);
828
    if (event)
829
        qemuDomainEventQueue(driver, event);
830
    virObjectUnref(cfg);
831 832 833
    return 0;
}

834 835 836
static int
qemuProcessHandleRTCChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm,
837 838
                           long long offset,
                           void *opaque)
839
{
840
    virQEMUDriverPtr driver = opaque;
841
    virObjectEventPtr event = NULL;
842
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
843

844
    virObjectLock(vm);
845

846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862
    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;
863
        vm->def->clock.data.variable.adjustment = offset;
864

865 866 867 868 869
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
           VIR_WARN("unable to save domain status with RTC change");
    }

    event = virDomainEventRTCChangeNewFromObj(vm, offset);
870

871
    virObjectUnlock(vm);
872

873
    if (event)
874
        qemuDomainEventQueue(driver, event);
875
    virObjectUnref(cfg);
876 877 878 879 880 881 882
    return 0;
}


static int
qemuProcessHandleWatchdog(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
883 884
                          int action,
                          void *opaque)
885
{
886
    virQEMUDriverPtr driver = opaque;
887 888
    virObjectEventPtr watchdogEvent = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
889
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
890

891
    virObjectLock(vm);
892 893 894
    watchdogEvent = virDomainEventWatchdogNewFromObj(vm, action);

    if (action == VIR_DOMAIN_EVENT_WATCHDOG_PAUSE &&
J
Jiri Denemark 已提交
895
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
896
        qemuDomainObjPrivatePtr priv = vm->privateData;
897 898
        VIR_DEBUG("Transitioned guest %s to paused state due to watchdog", vm->def->name);

J
Jiri Denemark 已提交
899
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
900
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
901 902 903
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

904 905 906 907 908
        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));

909
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
910 911 912
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
913 914 915
    }

    if (vm->def->watchdog->action == VIR_DOMAIN_WATCHDOG_ACTION_DUMP) {
C
Chen Fan 已提交
916 917 918 919 920
        struct qemuProcessEvent *processEvent;
        if (VIR_ALLOC(processEvent) == 0) {
            processEvent->eventType = QEMU_PROCESS_EVENT_WATCHDOG;
            processEvent->action = VIR_DOMAIN_WATCHDOG_ACTION_DUMP;
            processEvent->vm = vm;
W
Wen Congyang 已提交
921 922 923
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
924
            virObjectRef(vm);
C
Chen Fan 已提交
925
            if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
926
                if (!virObjectUnref(vm))
927
                    vm = NULL;
C
Chen Fan 已提交
928
                VIR_FREE(processEvent);
W
Wen Congyang 已提交
929
            }
930
        }
931 932
    }

933
    if (vm)
934
        virObjectUnlock(vm);
935 936 937 938
    if (watchdogEvent)
        qemuDomainEventQueue(driver, watchdogEvent);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
939

940
    virObjectUnref(cfg);
941 942 943 944 945 946 947 948 949
    return 0;
}


static int
qemuProcessHandleIOError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virDomainObjPtr vm,
                         const char *diskAlias,
                         int action,
950 951
                         const char *reason,
                         void *opaque)
952
{
953
    virQEMUDriverPtr driver = opaque;
954 955 956
    virObjectEventPtr ioErrorEvent = NULL;
    virObjectEventPtr ioErrorEvent2 = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
957 958 959
    const char *srcPath;
    const char *devAlias;
    virDomainDiskDefPtr disk;
960
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
961

962
    virObjectLock(vm);
963 964 965
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
966
        srcPath = virDomainDiskGetSource(disk);
967 968 969 970 971 972 973 974 975 976
        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
977
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
978
        qemuDomainObjPrivatePtr priv = vm->privateData;
979 980
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

J
Jiri Denemark 已提交
981
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
982
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
983 984 985
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

986 987 988 989 990
        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));

991
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
992 993
            VIR_WARN("Unable to save status on vm %s after IO error", vm->def->name);
    }
994
    virObjectUnlock(vm);
995

996 997 998 999 1000 1001
    if (ioErrorEvent)
        qemuDomainEventQueue(driver, ioErrorEvent);
    if (ioErrorEvent2)
        qemuDomainEventQueue(driver, ioErrorEvent2);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1002
    virObjectUnref(cfg);
1003 1004 1005
    return 0;
}

1006 1007 1008 1009 1010
static int
qemuProcessHandleBlockJob(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          const char *diskAlias,
                          int type,
1011 1012
                          int status,
                          void *opaque)
1013
{
1014
    virQEMUDriverPtr driver = opaque;
1015
    virObjectEventPtr event = NULL;
1016
    virObjectEventPtr event2 = NULL;
1017 1018
    const char *path;
    virDomainDiskDefPtr disk;
1019
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
E
Eric Blake 已提交
1020
    virDomainDiskDefPtr persistDisk = NULL;
1021
    bool save = false;
1022

1023
    virObjectLock(vm);
1024 1025 1026
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
1027 1028
        /* Have to generate two variants of the event for old vs. new
         * client callbacks */
E
Eric Blake 已提交
1029 1030 1031
        if (type == VIR_DOMAIN_BLOCK_JOB_TYPE_COMMIT &&
            disk->mirrorJob == VIR_DOMAIN_BLOCK_JOB_TYPE_ACTIVE_COMMIT)
            type = disk->mirrorJob;
1032
        path = virDomainDiskGetSource(disk);
1033
        event = virDomainEventBlockJobNewFromObj(vm, path, type, status);
1034 1035
        event2 = virDomainEventBlockJob2NewFromObj(vm, disk->dst, type,
                                                   status);
1036 1037 1038 1039 1040

        /* If we completed a block pull or commit, then update the XML
         * to match.  */
        if (status == VIR_DOMAIN_BLOCK_JOB_COMPLETED) {
            if (disk->mirrorState == VIR_DOMAIN_DISK_MIRROR_STATE_PIVOT) {
E
Eric Blake 已提交
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065
                if (vm->newDef) {
                    int indx = virDomainDiskIndexByName(vm->newDef, disk->dst,
                                                        false);
                    virStorageSourcePtr copy = NULL;

                    if (indx >= 0) {
                        persistDisk = vm->newDef->disks[indx];
                        copy = virStorageSourceCopy(disk->mirror, false);
                        if (virStorageSourceInitChainElement(copy,
                                                             persistDisk->src,
                                                             false) < 0) {
                            VIR_WARN("Unable to update persistent definition "
                                     "on vm %s after block job",
                                     vm->def->name);
                            virStorageSourceFree(copy);
                            copy = NULL;
                            persistDisk = NULL;
                        }
                    }
                    if (copy) {
                        virStorageSourceFree(persistDisk->src);
                        persistDisk->src = copy;
                    }
                }

1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
                /* XXX We want to revoke security labels and disk
                 * lease, as well as audit that revocation, before
                 * dropping the original source.  But it gets tricky
                 * if both source and mirror share common backing
                 * files (we want to only revoke the non-shared
                 * portion of the chain); so for now, we leak the
                 * access to the original.  */
                virStorageSourceFree(disk->src);
                disk->src = disk->mirror;
            } else {
                virStorageSourceFree(disk->mirror);
            }

            /* Recompute the cached backing chain to match our
             * updates.  Better would be storing the chain ourselves
             * rather than reprobing, but we haven't quite completed
             * that conversion to use our XML tracking. */
            disk->mirror = NULL;
            save = disk->mirrorState != VIR_DOMAIN_DISK_MIRROR_STATE_NONE;
            disk->mirrorState = VIR_DOMAIN_DISK_MIRROR_STATE_NONE;
E
Eric Blake 已提交
1086
            disk->mirrorJob = VIR_DOMAIN_BLOCK_JOB_TYPE_UNKNOWN;
1087
            qemuDomainDetermineDiskChain(driver, vm, disk, true);
E
Eric Blake 已提交
1088 1089 1090
        } else if (disk->mirror &&
                   (type == VIR_DOMAIN_BLOCK_JOB_TYPE_COPY ||
                    type == VIR_DOMAIN_BLOCK_JOB_TYPE_ACTIVE_COMMIT)) {
1091
            if (status == VIR_DOMAIN_BLOCK_JOB_READY) {
1092
                disk->mirrorState = VIR_DOMAIN_DISK_MIRROR_STATE_READY;
1093
                save = true;
1094 1095 1096
            } else if (status == VIR_DOMAIN_BLOCK_JOB_FAILED) {
                virStorageSourceFree(disk->mirror);
                disk->mirror = NULL;
1097
                disk->mirrorState = VIR_DOMAIN_DISK_MIRROR_STATE_NONE;
E
Eric Blake 已提交
1098
                disk->mirrorJob = VIR_DOMAIN_BLOCK_JOB_TYPE_UNKNOWN;
1099
                save = true;
1100 1101
            }
        }
1102 1103
    }

1104 1105 1106 1107
    if (save) {
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
            VIR_WARN("Unable to save status on vm %s after block job",
                     vm->def->name);
E
Eric Blake 已提交
1108 1109 1110 1111
        if (persistDisk && virDomainSaveConfig(cfg->configDir,
                                               vm->newDef) < 0)
            VIR_WARN("Unable to update persistent definition on vm %s "
                     "after block job", vm->def->name);
1112
    }
1113
    virObjectUnlock(vm);
1114
    virObjectUnref(cfg);
1115

1116
    if (event)
1117
        qemuDomainEventQueue(driver, event);
1118 1119
    if (event2)
        qemuDomainEventQueue(driver, event2);
1120 1121 1122

    return 0;
}
1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135

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,
1136 1137
                          const char *saslUsername,
                          void *opaque)
1138
{
1139
    virQEMUDriverPtr driver = opaque;
1140
    virObjectEventPtr event;
1141 1142 1143
    virDomainEventGraphicsAddressPtr localAddr = NULL;
    virDomainEventGraphicsAddressPtr remoteAddr = NULL;
    virDomainEventGraphicsSubjectPtr subject = NULL;
1144
    size_t i;
1145 1146

    if (VIR_ALLOC(localAddr) < 0)
1147
        goto error;
1148
    localAddr->family = localFamily;
1149 1150 1151
    if (VIR_STRDUP(localAddr->service, localService) < 0 ||
        VIR_STRDUP(localAddr->node, localNode) < 0)
        goto error;
1152 1153

    if (VIR_ALLOC(remoteAddr) < 0)
1154
        goto error;
1155
    remoteAddr->family = remoteFamily;
1156 1157 1158
    if (VIR_STRDUP(remoteAddr->service, remoteService) < 0 ||
        VIR_STRDUP(remoteAddr->node, remoteNode) < 0)
        goto error;
1159 1160

    if (VIR_ALLOC(subject) < 0)
1161
        goto error;
1162 1163
    if (x509dname) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1164
            goto error;
1165
        subject->nidentity++;
1166 1167 1168
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "x509dname") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, x509dname) < 0)
            goto error;
1169 1170 1171
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1172
            goto error;
1173
        subject->nidentity++;
1174 1175 1176
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "saslUsername") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, saslUsername) < 0)
            goto error;
1177 1178
    }

1179
    virObjectLock(vm);
1180
    event = virDomainEventGraphicsNewFromObj(vm, phase, localAddr, remoteAddr, authScheme, subject);
1181
    virObjectUnlock(vm);
1182

1183
    if (event)
1184 1185 1186 1187
        qemuDomainEventQueue(driver, event);

    return 0;

1188
 error:
1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
    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) {
1200
        for (i = 0; i < subject->nidentity; i++) {
1201 1202 1203 1204 1205 1206 1207 1208 1209 1210
            VIR_FREE(subject->identities[i].type);
            VIR_FREE(subject->identities[i].name);
        }
        VIR_FREE(subject->identities);
        VIR_FREE(subject);
    }

    return -1;
}

1211 1212 1213 1214
static int
qemuProcessHandleTrayChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            const char *devAlias,
1215 1216
                            int reason,
                            void *opaque)
1217
{
1218
    virQEMUDriverPtr driver = opaque;
1219
    virObjectEventPtr event = NULL;
1220
    virDomainDiskDefPtr disk;
1221
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1222

1223
    virObjectLock(vm);
1224 1225 1226 1227 1228 1229
    disk = qemuProcessFindDomainDiskByAlias(vm, devAlias);

    if (disk) {
        event = virDomainEventTrayChangeNewFromObj(vm,
                                                   devAlias,
                                                   reason);
1230 1231 1232 1233 1234 1235
        /* 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;

1236
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1237 1238 1239
            VIR_WARN("Unable to save status on vm %s after tray moved event",
                     vm->def->name);
        }
1240 1241
    }

1242
    virObjectUnlock(vm);
1243
    if (event)
1244
        qemuDomainEventQueue(driver, event);
1245
    virObjectUnref(cfg);
1246 1247 1248
    return 0;
}

O
Osier Yang 已提交
1249 1250
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1251 1252
                          virDomainObjPtr vm,
                          void *opaque)
O
Osier Yang 已提交
1253
{
1254
    virQEMUDriverPtr driver = opaque;
1255 1256
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1257
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1258

1259
    virObjectLock(vm);
O
Osier Yang 已提交
1260 1261
    event = virDomainEventPMWakeupNewFromObj(vm);

1262 1263 1264 1265 1266 1267 1268 1269 1270
    /* 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);
1271
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
1272 1273 1274
                                                  VIR_DOMAIN_EVENT_STARTED,
                                                  VIR_DOMAIN_EVENT_STARTED_WAKEUP);

1275
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1276 1277 1278 1279 1280
            VIR_WARN("Unable to save status on vm %s after wakeup event",
                     vm->def->name);
        }
    }

1281
    virObjectUnlock(vm);
1282 1283 1284 1285
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1286
    virObjectUnref(cfg);
O
Osier Yang 已提交
1287 1288
    return 0;
}
1289

O
Osier Yang 已提交
1290 1291
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1292 1293
                           virDomainObjPtr vm,
                           void *opaque)
O
Osier Yang 已提交
1294
{
1295
    virQEMUDriverPtr driver = opaque;
1296 1297
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1298
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1299

1300
    virObjectLock(vm);
O
Osier Yang 已提交
1301 1302
    event = virDomainEventPMSuspendNewFromObj(vm);

1303
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1304
        qemuDomainObjPrivatePtr priv = vm->privateData;
1305 1306 1307 1308 1309
        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
1310
        lifecycleEvent =
1311
            virDomainEventLifecycleNewFromObj(vm,
J
Jiri Denemark 已提交
1312 1313
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1314

1315
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1316 1317 1318
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1319 1320 1321

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1322 1323
    }

1324
    virObjectUnlock(vm);
O
Osier Yang 已提交
1325

1326 1327 1328 1329
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1330
    virObjectUnref(cfg);
O
Osier Yang 已提交
1331 1332 1333
    return 0;
}

1334 1335 1336
static int
qemuProcessHandleBalloonChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1337 1338
                               unsigned long long actual,
                               void *opaque)
1339
{
1340
    virQEMUDriverPtr driver = opaque;
1341
    virObjectEventPtr event = NULL;
1342
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1343

1344
    virObjectLock(vm);
1345 1346 1347 1348 1349 1350
    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;

1351
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
1352 1353
        VIR_WARN("unable to save domain status with balloon change");

1354
    virObjectUnlock(vm);
1355

1356
    if (event)
1357
        qemuDomainEventQueue(driver, event);
1358
    virObjectUnref(cfg);
1359 1360 1361
    return 0;
}

1362 1363
static int
qemuProcessHandlePMSuspendDisk(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1364 1365
                               virDomainObjPtr vm,
                               void *opaque)
1366
{
1367
    virQEMUDriverPtr driver = opaque;
1368 1369
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1370
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1371

1372
    virObjectLock(vm);
1373 1374 1375 1376 1377 1378 1379 1380 1381 1382
    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 =
1383
            virDomainEventLifecycleNewFromObj(vm,
1384 1385 1386
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_DISK);

1387
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1388 1389 1390 1391 1392 1393 1394 1395
            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);
    }

1396
    virObjectUnlock(vm);
1397

1398 1399 1400 1401
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1402 1403
    virObjectUnref(cfg);

1404 1405 1406
    return 0;
}

1407

1408 1409
static int
qemuProcessHandleGuestPanic(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1410 1411
                            virDomainObjPtr vm,
                            void *opaque)
1412
{
1413
    virQEMUDriverPtr driver = opaque;
1414 1415 1416
    struct qemuProcessEvent *processEvent;

    virObjectLock(vm);
1417
    if (VIR_ALLOC(processEvent) < 0)
1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432
        goto cleanup;

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

1433
 cleanup:
1434 1435 1436 1437 1438 1439 1440
    if (vm)
       virObjectUnlock(vm);

    return 0;
}


1441
int
1442 1443
qemuProcessHandleDeviceDeleted(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1444 1445
                               const char *devAlias,
                               void *opaque)
1446
{
1447
    virQEMUDriverPtr driver = opaque;
1448 1449
    struct qemuProcessEvent *processEvent = NULL;
    char *data;
1450 1451 1452 1453 1454 1455

    virObjectLock(vm);

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

1456 1457
    if (qemuDomainSignalDeviceRemoval(vm, devAlias))
        goto cleanup;
1458

1459 1460
    if (VIR_ALLOC(processEvent) < 0)
        goto error;
1461

1462 1463 1464 1465 1466
    processEvent->eventType = QEMU_PROCESS_EVENT_DEVICE_DELETED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
    processEvent->vm = vm;
1467

1468 1469 1470 1471 1472
    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }
1473

1474
 cleanup:
1475 1476
    virObjectUnlock(vm);
    return 0;
1477 1478 1479 1480 1481
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
1482 1483 1484
}


1485 1486
static qemuMonitorCallbacks monitorCallbacks = {
    .eofNotify = qemuProcessHandleMonitorEOF,
1487
    .errorNotify = qemuProcessHandleMonitorError,
1488
    .diskSecretLookup = qemuProcessFindVolumeQcowPassphrase,
1489
    .domainEvent = qemuProcessHandleEvent,
1490 1491
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
1492
    .domainResume = qemuProcessHandleResume,
1493 1494 1495 1496 1497
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1498
    .domainBlockJob = qemuProcessHandleBlockJob,
1499
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1500
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1501
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1502
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1503
    .domainPMSuspendDisk = qemuProcessHandlePMSuspendDisk,
1504
    .domainGuestPanic = qemuProcessHandleGuestPanic,
1505
    .domainDeviceDeleted = qemuProcessHandleDeviceDeleted,
1506 1507 1508
};

static int
1509 1510
qemuConnectMonitor(virQEMUDriverPtr driver, virDomainObjPtr vm, int asyncJob,
                   int logfd)
1511 1512 1513
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
1514
    qemuMonitorPtr mon = NULL;
1515

1516
    if (virSecurityManagerSetDaemonSocketLabel(driver->securityManager,
1517
                                               vm->def) < 0) {
1518 1519 1520 1521 1522 1523
        VIR_ERROR(_("Failed to set security context for monitor for %s"),
                  vm->def->name);
        goto error;
    }

    /* Hold an extra reference because we can't allow 'vm' to be
M
Michal Privoznik 已提交
1524
     * deleted unitl the monitor gets its own reference. */
1525
    virObjectRef(vm);
1526

1527
    ignore_value(virTimeMillisNow(&priv->monStart));
1528
    virObjectUnlock(vm);
1529 1530 1531 1532

    mon = qemuMonitorOpen(vm,
                          priv->monConfig,
                          priv->monJSON,
1533 1534
                          &monitorCallbacks,
                          driver);
1535

1536 1537 1538
    if (mon)
        ignore_value(qemuMonitorSetDomainLog(mon, logfd));

1539
    virObjectLock(vm);
M
Michal Privoznik 已提交
1540
    virObjectUnref(vm);
1541
    priv->monStart = 0;
1542

M
Michal Privoznik 已提交
1543
    if (!virDomainObjIsActive(vm)) {
1544
        qemuMonitorClose(mon);
1545
        mon = NULL;
1546 1547 1548
    }
    priv->mon = mon;

1549
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0) {
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560
        VIR_ERROR(_("Failed to clear security context for monitor for %s"),
                  vm->def->name);
        goto error;
    }

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


1561 1562
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto error;
1563 1564
    ret = qemuMonitorSetCapabilities(priv->mon);
    if (ret == 0 &&
1565 1566
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON))
        ret = virQEMUCapsProbeQMP(priv->qemuCaps, priv->mon);
1567
    qemuDomainObjExitMonitor(driver, vm);
1568

1569
 error:
1570 1571 1572 1573

    return ret;
}

1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587

/**
 * qemuProcessReadLog: Read log file of a qemu VM
 * @fd: File descriptor of the log file
 * @buf: buffer to store the read messages
 * @buflen: allocated space available in @buf
 * @off: Offset to start reading from
 * @skipchar: Skip messages about created character devices
 *
 * Reads log of a qemu VM. Skips messages not produced by qemu or irrelevant
 * messages. Returns length of the message stored in @buf, or -1 on error.
 */
int
qemuProcessReadLog(int fd, char *buf, int buflen, int off, bool skipchar)
1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
{
    char *filter_next = buf;
    ssize_t bytes;
    char *eol;

    while (off < buflen - 1) {
        bytes = saferead(fd, buf + off, buflen - off - 1);
        if (bytes < 0)
            return -1;

        off += bytes;
        buf[off] = '\0';

1601 1602 1603
        if (bytes == 0)
            break;

1604 1605 1606
        /* Filter out debug messages from intermediate libvirt process */
        while ((eol = strchr(filter_next, '\n'))) {
            *eol = '\0';
1607 1608 1609
            if (virLogProbablyLogMessage(filter_next) ||
                (skipchar &&
                 STRPREFIX(filter_next, "char device redirected to"))) {
1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
                memmove(filter_next, eol + 1, off - (eol - buf));
                off -= eol + 1 - filter_next;
            } else {
                filter_next = eol + 1;
                *eol = '\n';
            }
        }
    }

    return off;
}

1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
typedef int qemuProcessLogHandleOutput(virDomainObjPtr vm,
                                       const char *output,
                                       int fd);

/*
 * Returns -1 for error, 0 on success
 */
static int
qemuProcessReadLogOutput(virDomainObjPtr vm,
                         int fd,
                         char *buf,
                         size_t buflen,
                         qemuProcessLogHandleOutput func,
                         const char *what,
                         int timeout)
{
    int retries = (timeout*10);
    int got = 0;
1640 1641
    int ret = -1;

1642 1643 1644
    buf[0] = '\0';

    while (retries) {
1645
        ssize_t func_ret;
1646
        bool isdead;
1647 1648 1649

        func_ret = func(vm, buf, fd);

1650
        isdead = kill(vm->pid, 0) == -1 && errno == ESRCH;
1651

1652
        got = qemuProcessReadLog(fd, buf, buflen, got, false);
1653
        if (got < 0) {
1654 1655 1656
            virReportSystemError(errno,
                                 _("Failure while reading %s log output"),
                                 what);
1657
            goto cleanup;
1658 1659 1660
        }

        if (got == buflen-1) {
1661 1662 1663
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Out of space while reading %s log output: %s"),
                           what, buf);
1664
            goto cleanup;
1665 1666 1667
        }

        if (isdead) {
1668 1669 1670
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Process exited while reading %s log output: %s"),
                           what, buf);
1671
            goto cleanup;
1672 1673
        }

1674 1675 1676 1677
        if (func_ret <= 0) {
            ret = func_ret;
            goto cleanup;
        }
1678 1679 1680 1681 1682

        usleep(100*1000);
        retries--;
    }

1683 1684 1685
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("Timed out while reading %s log output: %s"),
                   what, buf);
1686

1687
 cleanup:
1688
    return ret;
1689 1690 1691
}


1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749
/*
 * Read domain log and probably overwrite error if there's one in
 * the domain log file. This function exists to cover the small
 * window between fork() and exec() during which child may fail
 * by libvirt's hand, e.g. placing onto a NUMA node failed.
 */
static int
qemuProcessReadChildErrors(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           off_t originalOff)
{
    int ret = -1;
    int logfd;
    off_t off = 0;
    ssize_t bytes;
    char buf[1024] = {0};
    char *eol, *filter_next = buf;

    if ((logfd = qemuDomainOpenLog(driver, vm, originalOff)) < 0)
        goto cleanup;

    while (off < sizeof(buf) - 1) {
        bytes = saferead(logfd, buf + off, sizeof(buf) - off - 1);
        if (bytes < 0) {
            VIR_WARN("unable to read from log file: %s",
                     virStrerror(errno, buf, sizeof(buf)));
            goto cleanup;
        }

        off += bytes;
        buf[off] = '\0';

        if (bytes == 0)
            break;

        while ((eol = strchr(filter_next, '\n'))) {
            *eol = '\0';
            if (STRPREFIX(filter_next, "libvirt: ")) {
                filter_next = eol + 1;
                *eol = '\n';
                break;
            } else {
                memmove(filter_next, eol + 1, off - (eol - buf));
                off -= eol + 1 - filter_next;
            }
        }
    }

    if (off > 0) {
        /* Found an error in the log. Report it */
        virResetLastError();
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Process exited prior to exec: %s"),
                       buf);
    }

    ret = 0;

1750
 cleanup:
1751 1752 1753 1754 1755
    VIR_FORCE_CLOSE(logfd);
    return ret;
}


1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768
/*
 * Look at a chunk of data from the QEMU stdout logs and try to
 * find a TTY device, as indicated by a line like
 *
 * char device redirected to /dev/pts/3
 *
 * Returns -1 for error, 0 success, 1 continue reading
 */
static int
qemuProcessExtractTTYPath(const char *haystack,
                          size_t *offset,
                          char **path)
{
1769 1770
    static const char needle[] = "char device redirected to";
    char *tmp, *dev;
1771 1772 1773

    VIR_FREE(*path);
    /* First look for our magic string */
1774
    if (!(tmp = strstr(haystack + *offset, needle))) {
1775 1776
        return 1;
    }
1777
    tmp += sizeof(needle);
1778 1779 1780 1781 1782 1783 1784 1785
    dev = tmp;

    /*
     * And look for first whitespace character and nul terminate
     * to mark end of the pty path
     */
    while (*tmp) {
        if (c_isspace(*tmp)) {
1786
            if (VIR_STRNDUP(*path, dev, tmp - dev) < 0)
1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804
                return -1;

            /* ... now further update offset till we get EOL */
            *offset = tmp - haystack;
            return 0;
        }
        tmp++;
    }

    /*
     * We found a path, but didn't find any whitespace,
     * so it must be still incomplete - we should at
     * least see a \n - indicate that we want to carry
     * on trying again
     */
    return 1;
}

1805
static int
1806 1807 1808
qemuProcessLookupPTYs(virDomainDefPtr def,
                      virQEMUCapsPtr qemuCaps,
                      virDomainChrDefPtr *devices,
1809
                      int count,
1810
                      virHashTablePtr paths)
1811
{
1812
    size_t i;
1813

1814
    for (i = 0; i < count; i++) {
1815
        virDomainChrDefPtr chr = devices[i];
1816 1817
        bool chardevfmt = virQEMUCapsSupportsChardev(def, qemuCaps, chr);

1818
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
C
Cole Robinson 已提交
1819
            char id[32];
1820 1821
            const char *path;

C
Cole Robinson 已提交
1822
            if (snprintf(id, sizeof(id), "%s%s",
1823 1824
                         chardevfmt ? "char" : "",
                         chr->info.alias) >= sizeof(id))
1825 1826 1827 1828 1829 1830 1831 1832
                return -1;

            path = (const char *) virHashLookup(paths, id);
            if (path == NULL) {
                if (chr->source.data.file.path == NULL) {
                    /* neither the log output nor 'info chardev' had a
                     * pty path for this chardev, report an error
                     */
1833 1834
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1835 1836 1837 1838 1839 1840 1841 1842 1843 1844
                    return -1;
                } else {
                    /* 'info chardev' had no pty path for this chardev,
                     * but the log output had, so we're fine
                     */
                    continue;
                }
            }

            VIR_FREE(chr->source.data.file.path);
1845
            if (VIR_STRDUP(chr->source.data.file.path, path) < 0)
1846 1847 1848 1849 1850 1851 1852
                return -1;
        }
    }

    return 0;
}

1853 1854
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
1855
                                     virQEMUCapsPtr qemuCaps,
1856 1857
                                     virHashTablePtr paths)
{
1858
    size_t i = 0;
C
Cole Robinson 已提交
1859

1860 1861 1862
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->serials, vm->def->nserials,
                              paths) < 0)
1863 1864
        return -1;

1865 1866 1867
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->parallels, vm->def->nparallels,
                              paths) < 0)
1868
        return -1;
1869

1870 1871 1872
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->channels, vm->def->nchannels,
                              paths) < 0)
1873
        return -1;
1874 1875 1876 1877
    /* 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];
1878

1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889
        if (vm->def->nserials &&
            chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_CONSOLE &&
            chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_SERIAL) {
            /* yes, the first console is just an alias for serials[0] */
            i = 1;
            if (virDomainChrSourceDefCopy(&chr->source,
                                          &((vm->def->serials[0])->source)) < 0)
                return -1;
        }
    }

1890 1891 1892
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->consoles + i, vm->def->nconsoles - i,
                              paths) < 0)
1893
        return -1;
1894 1895 1896 1897 1898 1899 1900 1901 1902 1903

    return 0;
}

static int
qemuProcessFindCharDevicePTYs(virDomainObjPtr vm,
                              const char *output,
                              int fd ATTRIBUTE_UNUSED)
{
    size_t offset = 0;
1904 1905
    int ret;
    size_t i;
1906 1907 1908 1909 1910 1911

    /* The order in which QEMU prints out the PTY paths is
       the order in which it procsses its serial and parallel
       device args. This code must match that ordering.... */

    /* first comes the serial devices */
1912
    for (i = 0; i < vm->def->nserials; i++) {
1913 1914 1915 1916 1917 1918 1919 1920 1921
        virDomainChrDefPtr chr = vm->def->serials[i];
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
            if ((ret = qemuProcessExtractTTYPath(output, &offset,
                                                 &chr->source.data.file.path)) != 0)
                return ret;
        }
    }

    /* then the parallel devices */
1922
    for (i = 0; i < vm->def->nparallels; i++) {
1923 1924 1925 1926 1927 1928 1929 1930 1931
        virDomainChrDefPtr chr = vm->def->parallels[i];
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
            if ((ret = qemuProcessExtractTTYPath(output, &offset,
                                                 &chr->source.data.file.path)) != 0)
                return ret;
        }
    }

    /* then the channel devices */
1932
    for (i = 0; i < vm->def->nchannels; i++) {
1933 1934 1935 1936 1937 1938 1939 1940
        virDomainChrDefPtr chr = vm->def->channels[i];
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
            if ((ret = qemuProcessExtractTTYPath(output, &offset,
                                                 &chr->source.data.file.path)) != 0)
                return ret;
        }
    }

1941
    for (i = 0; i < vm->def->nconsoles; i++) {
1942
        virDomainChrDefPtr chr = vm->def->consoles[i];
1943 1944 1945 1946 1947 1948 1949
        /* For historical reasons, console[0] can be just an alias
         * for serial[0]; That's why we need to update it as well */
        if (i == 0 && vm->def->nserials &&
            chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_CONSOLE &&
            chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_SERIAL) {
            if ((ret = virDomainChrSourceDefCopy(&chr->source,
                                                 &((vm->def->serials[0])->source))) != 0)
1950
                return ret;
1951 1952 1953 1954 1955 1956 1957
        } else {
            if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY &&
                chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_VIRTIO) {
                if ((ret = qemuProcessExtractTTYPath(output, &offset,
                                                     &chr->source.data.file.path)) != 0)
                    return ret;
            }
1958 1959 1960
        }
    }

1961 1962 1963 1964 1965
    return 0;
}


static int
1966
qemuProcessWaitForMonitor(virQEMUDriverPtr driver,
C
Cole Robinson 已提交
1967
                          virDomainObjPtr vm,
1968
                          int asyncJob,
1969
                          virQEMUCapsPtr qemuCaps,
C
Cole Robinson 已提交
1970
                          off_t pos)
1971
{
1972
    char *buf = NULL;
1973
    size_t buf_size = 4096; /* Plenty of space to get startup greeting */
1974
    int logfd = -1;
1975 1976
    int ret = -1;
    virHashTablePtr paths = NULL;
1977
    qemuDomainObjPrivatePtr priv;
1978

1979 1980 1981
    if (pos != -1 &&
        (logfd = qemuDomainOpenLog(driver, vm, pos)) < 0)
        return -1;
1982

1983
    if (logfd != -1 && !virQEMUCapsUsedQMP(qemuCaps)) {
1984
        if (VIR_ALLOC_N(buf, buf_size) < 0)
1985
            goto closelog;
1986

1987 1988 1989 1990 1991
        if (qemuProcessReadLogOutput(vm, logfd, buf, buf_size,
                                     qemuProcessFindCharDevicePTYs,
                                     "console", 30) < 0)
            goto closelog;
    }
1992 1993

    VIR_DEBUG("Connect monitor to %p '%s'", vm, vm->def->name);
1994
    if (qemuConnectMonitor(driver, vm, asyncJob, logfd) < 0)
1995 1996 1997 1998 1999 2000
        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. */
E
Eric Blake 已提交
2001
    paths = virHashCreate(0, virHashValueFree);
2002
    if (paths == NULL)
2003 2004
        goto cleanup;

2005
    priv = vm->privateData;
2006 2007
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
2008
    ret = qemuMonitorGetPtyPaths(priv->mon, paths);
2009
    qemuDomainObjExitMonitor(driver, vm);
2010 2011 2012

    VIR_DEBUG("qemuMonitorGetPtyPaths returned %i", ret);
    if (ret == 0)
2013
        ret = qemuProcessFindCharDevicePTYsMonitor(vm, qemuCaps, paths);
2014

2015
 cleanup:
2016
    virHashFree(paths);
2017

E
Eric Blake 已提交
2018
    if (pos != -1 && kill(vm->pid, 0) == -1 && errno == ESRCH) {
2019
        int len;
2020 2021
        /* VM is dead, any other error raised in the interim is probably
         * not as important as the qemu cmdline output */
2022
        if (virQEMUCapsUsedQMP(qemuCaps)) {
2023
            if (VIR_ALLOC_N(buf, buf_size) < 0)
2024 2025
                goto closelog;
        }
2026 2027

        len = strlen(buf);
2028
        /* best effort seek - we need to reset to the original position, so that
2029 2030
         * a possible read of the fd in the monitor code doesn't influence this
         * error delivery option */
2031
        ignore_value(lseek(logfd, pos, SEEK_SET));
2032
        qemuProcessReadLog(logfd, buf + len, buf_size - len - 1, 0, true);
2033 2034 2035
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("process exited while connecting to monitor: %s"),
                       buf);
2036 2037 2038
        ret = -1;
    }

2039
 closelog:
2040
    if (VIR_CLOSE(logfd) < 0) {
2041
        char ebuf[1024];
2042
        VIR_WARN("Unable to close logfile: %s",
2043
                 virStrerror(errno, ebuf, sizeof(ebuf)));
2044 2045
    }

2046 2047
    VIR_FREE(buf);

2048 2049 2050 2051
    return ret;
}

static int
2052
qemuProcessDetectVcpuPIDs(virQEMUDriverPtr driver,
2053
                          virDomainObjPtr vm, int asyncJob)
2054 2055 2056 2057 2058
{
    pid_t *cpupids = NULL;
    int ncpupids;
    qemuDomainObjPrivatePtr priv = vm->privateData;

2059 2060
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2061 2062 2063 2064
    /* failure to get the VCPU<-> PID mapping or to execute the query
     * command will not be treated fatal as some versions of qemu don't
     * support this command */
    if ((ncpupids = qemuMonitorGetCPUInfo(priv->mon, &cpupids)) <= 0) {
2065
        qemuDomainObjExitMonitor(driver, vm);
2066 2067
        virResetLastError();

2068
        priv->nvcpupids = 1;
2069
        if (VIR_ALLOC_N(priv->vcpupids, priv->nvcpupids) < 0)
2070 2071 2072 2073
            return -1;
        priv->vcpupids[0] = vm->pid;
        return 0;
    }
2074
    qemuDomainObjExitMonitor(driver, vm);
2075 2076

    if (ncpupids != vm->def->vcpus) {
2077 2078 2079 2080
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of vCPU pids from QEMU monitor. "
                         "got %d, wanted %d"),
                       ncpupids, vm->def->vcpus);
2081 2082 2083 2084 2085 2086 2087 2088 2089
        VIR_FREE(cpupids);
        return -1;
    }

    priv->nvcpupids = ncpupids;
    priv->vcpupids = cpupids;
    return 0;
}

2090

2091 2092 2093
/* Helper to prepare cpumap for affinity setting, convert
 * NUMA nodeset into cpuset if @nodemask is not NULL, otherwise
 * just return a new allocated bitmap.
2094
 */
2095
virBitmapPtr
2096
qemuPrepareCpumap(virQEMUDriverPtr driver,
2097
                  virBitmapPtr nodemask)
2098
{
2099 2100
    size_t i;
    int hostcpus, maxcpu = QEMUD_CPUMASK_LEN;
2101
    virBitmapPtr cpumap = NULL;
2102
    virCapsPtr caps = NULL;
2103 2104 2105

    /* setaffinity fails if you set bits for CPUs which
     * aren't present, so we have to limit ourselves */
2106 2107 2108
    if ((hostcpus = nodeGetCPUCount()) < 0)
        return NULL;

2109 2110 2111
    if (maxcpu > hostcpus)
        maxcpu = hostcpus;

2112
    if (!(cpumap = virBitmapNew(maxcpu)))
2113
        return NULL;
2114

2115
    if (nodemask) {
2116 2117 2118 2119 2120 2121 2122
        if (!(caps = virQEMUDriverGetCapabilities(driver, false))) {
            virBitmapFree(cpumap);
            cpumap = NULL;
            goto cleanup;
        }

        for (i = 0; i < caps->host.nnumaCell; i++) {
2123
            size_t j;
2124
            int cur_ncpus = caps->host.numaCell[i]->ncpus;
2125
            bool result;
2126
            if (virBitmapGetBit(nodemask, caps->host.numaCell[i]->num, &result) < 0) {
2127
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
E
Eric Blake 已提交
2128
                               _("Failed to convert nodeset to cpuset"));
2129
                virBitmapFree(cpumap);
2130 2131
                cpumap = NULL;
                goto cleanup;
2132
            }
2133
            if (result) {
2134
                for (j = 0; j < cur_ncpus; j++)
2135
                    ignore_value(virBitmapSetBit(cpumap,
2136
                                                 caps->host.numaCell[i]->cpus[j].id));
2137
            }
O
Osier Yang 已提交
2138
        }
2139 2140
    }

2141
 cleanup:
2142
    virObjectUnref(caps);
2143 2144 2145 2146 2147 2148 2149
    return cpumap;
}

/*
 * To be run between fork/exec of QEMU only
 */
static int
2150
qemuProcessInitCpuAffinity(virQEMUDriverPtr driver,
2151 2152 2153 2154 2155 2156 2157
                           virDomainObjPtr vm,
                           virBitmapPtr nodemask)
{
    int ret = -1;
    virBitmapPtr cpumap = NULL;
    virBitmapPtr cpumapToSet = NULL;

2158 2159 2160 2161 2162 2163
    if (!vm->pid) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Cannot setup CPU affinity until process is started"));
        return -1;
    }

2164 2165 2166 2167 2168 2169
    if (!(cpumap = qemuPrepareCpumap(driver, nodemask)))
        return -1;

    if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) {
        VIR_DEBUG("Set CPU affinity with advisory nodeset from numad");
        cpumapToSet = cpumap;
2170
    } else {
2171
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
2172
        if (vm->def->cpumask) {
H
Hu Tao 已提交
2173
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
2174
        } else {
2175
            cpumapToSet = cpumap;
O
Osier Yang 已提交
2176 2177 2178 2179
            /* 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 */
2180
            virBitmapSetAll(cpumap);
O
Osier Yang 已提交
2181
        }
2182 2183
    }

2184
    if (virProcessSetAffinity(vm->pid, cpumapToSet) < 0)
2185
        goto cleanup;
2186

2187 2188
    ret = 0;

2189
 cleanup:
2190
    virBitmapFree(cpumap);
2191
    return ret;
2192 2193
}

2194 2195 2196 2197 2198 2199
/* set link states to down on interfaces at qemu start */
static int
qemuProcessSetLinkStates(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2200
    size_t i;
2201 2202 2203 2204 2205 2206
    int ret = 0;

    for (i = 0; i < def->nnets; i++) {
        if (def->nets[i]->linkstate == VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN) {
            VIR_DEBUG("Setting link state: %s", def->nets[i]->info.alias);

2207
            if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NETDEV)) {
2208
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
2209
                               _("Setting of link state is not supported by this qemu"));
2210 2211 2212 2213 2214 2215 2216
                return -1;
            }

            ret = qemuMonitorSetLink(priv->mon,
                                     def->nets[i]->info.alias,
                                     VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (ret != 0) {
2217
                virReportError(VIR_ERR_OPERATION_FAILED,
2218 2219 2220 2221 2222 2223 2224 2225 2226 2227
                               _("Couldn't set link state on interface: %s"), def->nets[i]->info.alias);
                break;
            }
        }
    }

    return ret;
}

/* Set CPU affinities for vcpus if vcpupin xml provided. */
2228
static int
2229
qemuProcessSetVcpuAffinities(virDomainObjPtr vm)
2230 2231 2232
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2233
    int vcpu, n;
E
Eric Blake 已提交
2234
    int ret = -1;
2235 2236 2237 2238 2239

    if (!def->cputune.nvcpupin)
        return 0;

    if (priv->vcpupids == NULL) {
2240 2241
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("cpu affinity is not supported"));
2242 2243 2244
        return -1;
    }

2245 2246
    for (n = 0; n < def->cputune.nvcpupin; n++) {
        vcpu = def->cputune.vcpupin[n]->vcpuid;
2247

2248 2249
        if (virProcessSetAffinity(priv->vcpupids[vcpu],
                                  def->cputune.vcpupin[n]->cpumask) < 0) {
E
Eric Blake 已提交
2250
            goto cleanup;
2251 2252 2253
        }
    }

E
Eric Blake 已提交
2254
    ret = 0;
2255
 cleanup:
E
Eric Blake 已提交
2256
    return ret;
2257
}
2258

2259
/* Set CPU affinities for emulator threads. */
2260
static int
2261
qemuProcessSetEmulatorAffinities(virConnectPtr conn ATTRIBUTE_UNUSED,
2262 2263
                                virDomainObjPtr vm)
{
2264
    virBitmapPtr cpumask;
2265 2266 2267
    virDomainDefPtr def = vm->def;
    int ret = -1;

O
Osier Yang 已提交
2268
    if (def->cputune.emulatorpin)
2269
        cpumask = def->cputune.emulatorpin->cpumask;
O
Osier Yang 已提交
2270
    else if (def->cpumask)
2271
        cpumask = def->cpumask;
O
Osier Yang 已提交
2272 2273
    else
        return 0;
2274

2275
    ret = virProcessSetAffinity(vm->pid, cpumask);
2276 2277 2278
    return ret;
}

2279 2280
static int
qemuProcessInitPasswords(virConnectPtr conn,
2281
                         virQEMUDriverPtr driver,
2282 2283
                         virDomainObjPtr vm,
                         int asyncJob)
2284 2285 2286
{
    int ret = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
2287
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2288
    size_t i;
2289

2290
    for (i = 0; i < vm->def->ngraphics; ++i) {
2291 2292
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
2293 2294
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_VNC,
2295
                                                    &graphics->data.vnc.auth,
2296 2297
                                                    cfg->vncPassword,
                                                    asyncJob);
2298
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
2299 2300
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_SPICE,
2301
                                                    &graphics->data.spice.auth,
2302 2303
                                                    cfg->spicePassword,
                                                    asyncJob);
2304 2305
        }

2306 2307 2308
        if (ret < 0)
            goto cleanup;
    }
2309

2310
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
2311
        for (i = 0; i < vm->def->ndisks; i++) {
2312 2313
            char *secret;
            size_t secretLen;
E
Eric Blake 已提交
2314
            const char *alias;
2315

2316
            if (!vm->def->disks[i]->src->encryption ||
2317
                !virDomainDiskGetSource(vm->def->disks[i]))
2318 2319 2320 2321 2322 2323 2324
                continue;

            if (qemuProcessGetVolumeQcowPassphrase(conn,
                                                   vm->def->disks[i],
                                                   &secret, &secretLen) < 0)
                goto cleanup;

E
Eric Blake 已提交
2325
            alias = vm->def->disks[i]->info.alias;
2326 2327
            if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0) {
                VIR_FREE(secret);
2328
                goto cleanup;
2329
            }
E
Eric Blake 已提交
2330
            ret = qemuMonitorSetDrivePassphrase(priv->mon, alias, secret);
2331
            VIR_FREE(secret);
2332
            qemuDomainObjExitMonitor(driver, vm);
2333 2334 2335 2336 2337
            if (ret < 0)
                goto cleanup;
        }
    }

2338
 cleanup:
2339
    virObjectUnref(cfg);
2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382
    return ret;
}


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

#define QEMU_PCI_PRODUCT_DISK_VIRTIO 0x1001

#define QEMU_PCI_PRODUCT_BALLOON_VIRTIO 0x1002

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

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

#define QEMU_PCI_PRODUCT_AUDIO_AC97    0x2415
#define QEMU_PCI_PRODUCT_AUDIO_ES1370  0x5000

#define QEMU_PCI_PRODUCT_CONTROLLER_PIIX 0x7010
#define QEMU_PCI_PRODUCT_CONTROLLER_LSI  0x0012

#define QEMU_PCI_PRODUCT_WATCHDOG_I63000ESB 0x25ab

static int
qemuProcessAssignNextPCIAddress(virDomainDeviceInfo *info,
                                int vendor,
                                int product,
                                qemuMonitorPCIAddress *addrs,
                                int naddrs)
{
2383
    bool found = false;
2384
    size_t i;
2385 2386 2387

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

2388
    for (i = 0; i < naddrs; i++) {
2389 2390 2391
        VIR_DEBUG("Maybe %x:%x", addrs[i].vendor, addrs[i].product);
        if (addrs[i].vendor == vendor &&
            addrs[i].product == product) {
2392
            VIR_DEBUG("Match %zu", i);
2393
            found = true;
2394 2395 2396
            break;
        }
    }
2397
    if (!found)
2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 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 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597
        return -1;

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

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

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

    return 0;
}

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

    default:
        return -1;
    }

    return 0;
}

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

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

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

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

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

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

    return 0;
}

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

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

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

    default:
        return -1;
    }
    return 0;
}

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

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

    default:
        return -1;
    }

    return 0;
}

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

    default:
        return -1;
    }

    return 0;
}


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

    default:
        return -1;
    }

    return 0;
}


/*
 * This entire method assumes that PCI devices in 'info pci'
 * match ordering of devices specified on the command line
 * wrt to devices of matching vendor+product
 *
 * XXXX this might not be a valid assumption if we assign
 * some static addrs on CLI. Have to check that...
 */
static int
qemuProcessDetectPCIAddresses(virDomainObjPtr vm,
                              qemuMonitorPCIAddress *addrs,
                              int naddrs)
{
    unsigned int vendor = 0, product = 0;
2598
    size_t i;
2599 2600 2601 2602 2603

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

2604
    for (i = 0; i < vm->def->ndisks; i++) {
2605 2606 2607 2608 2609 2610
        if (qemuProcessGetPCIDiskVendorProduct(vm->def->disks[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->disks[i]->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2611 2612 2613
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for VirtIO disk %s"),
                           vm->def->disks[i]->dst);
2614 2615 2616 2617
            return -1;
        }
    }

2618
    for (i = 0; i < vm->def->nnets; i++) {
2619 2620 2621 2622 2623 2624
        if (qemuProcessGetPCINetVendorProduct(vm->def->nets[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->nets[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2625 2626 2627
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for %s NIC"),
                           vm->def->nets[i]->model);
2628 2629 2630 2631
            return -1;
        }
    }

2632
    for (i = 0; i < vm->def->ncontrollers; i++) {
2633 2634 2635 2636 2637 2638
        if (qemuProcessGetPCIControllerVendorProduct(vm->def->controllers[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->controllers[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2639 2640 2641
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for controller %s"),
                           virDomainControllerTypeToString(vm->def->controllers[i]->type));
2642 2643 2644 2645
            return -1;
        }
    }

2646
    for (i = 0; i < vm->def->nvideos; i++) {
2647 2648 2649 2650 2651 2652
        if (qemuProcessGetPCIVideoVendorProduct(vm->def->videos[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->videos[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2653 2654 2655
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for video adapter %s"),
                           virDomainVideoTypeToString(vm->def->videos[i]->type));
2656 2657 2658 2659
            return -1;
        }
    }

2660
    for (i = 0; i < vm->def->nsounds; i++) {
2661 2662 2663 2664 2665 2666
        if (qemuProcessGetPCISoundVendorProduct(vm->def->sounds[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->sounds[i]->info),
                                    vendor, product,
                                     addrs,  naddrs) < 0) {
2667 2668 2669
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for sound adapter %s"),
                           virDomainSoundModelTypeToString(vm->def->sounds[i]->model));
2670 2671 2672 2673 2674 2675 2676 2677 2678 2679
            return -1;
        }
    }


    if (vm->def->watchdog &&
        qemuProcessGetPCIWatchdogVendorProduct(vm->def->watchdog, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->watchdog->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2680 2681 2682
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for watchdog %s"),
                           virDomainWatchdogModelTypeToString(vm->def->watchdog->model));
2683 2684 2685 2686 2687 2688 2689 2690 2691
            return -1;
        }
    }

    if (vm->def->memballoon &&
        qemuProcessGetPCIMemballoonVendorProduct(vm->def->memballoon, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->memballoon->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2692 2693 2694
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for balloon %s"),
                           virDomainMemballoonModelTypeToString(vm->def->memballoon->model));
2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711
            return -1;
        }
    }

    /* XXX console (virtio) */


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

    /* XXX USB controller ? */

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

    return 0;
}

static int
2712
qemuProcessInitPCIAddresses(virQEMUDriverPtr driver,
2713 2714
                            virDomainObjPtr vm,
                            int asyncJob)
2715 2716 2717 2718 2719 2720
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int naddrs;
    int ret;
    qemuMonitorPCIAddress *addrs = NULL;

2721 2722
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2723 2724
    naddrs = qemuMonitorGetAllPCIAddresses(priv->mon,
                                           &addrs);
2725
    qemuDomainObjExitMonitor(driver, vm);
2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757

    ret = qemuProcessDetectPCIAddresses(vm, addrs, naddrs);

    VIR_FREE(addrs);

    return ret;
}


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

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

    VIR_FORCE_CLOSE(fd);

    return 0;
}


2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770
static int
qemuProcessCleanupChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
    if (dev->source.type == VIR_DOMAIN_CHR_TYPE_UNIX &&
        dev->source.data.nix.listen)
        unlink(dev->source.data.nix.path);

    return 0;
}


2771 2772 2773
struct qemuProcessHookData {
    virConnectPtr conn;
    virDomainObjPtr vm;
2774
    virQEMUDriverPtr driver;
2775
    virBitmapPtr nodemask;
2776
    virQEMUDriverConfigPtr cfg;
2777 2778 2779 2780 2781
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
2782
    int ret = -1;
2783
    int fd;
2784 2785 2786 2787 2788
    /* This method cannot use any mutexes, which are not
     * protected across fork()
     */

    virSecurityManagerPostFork(h->driver->securityManager);
2789 2790 2791 2792 2793

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

    VIR_DEBUG("Obtaining domain lock");
2794 2795 2796 2797 2798 2799 2800
    /*
     * 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.
     */
2801
    if (virSecurityManagerSetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2802
        goto cleanup;
2803
    if (virDomainLockProcessStart(h->driver->lockManager,
2804
                                  h->cfg->uri,
2805
                                  h->vm,
J
Ján Tomko 已提交
2806
                                  /* QEMU is always paused initially */
2807 2808
                                  true,
                                  &fd) < 0)
2809
        goto cleanup;
2810
    if (virSecurityManagerClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2811
        goto cleanup;
2812

G
Gao feng 已提交
2813
    if (virNumaSetupMemoryPolicy(h->vm->def->numatune, h->nodemask) < 0)
2814
        goto cleanup;
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
qemuProcessPrepareMonitorChr(virQEMUDriverConfigPtr cfg,
2826 2827 2828 2829 2830 2831
                             virDomainChrSourceDefPtr monConfig,
                             const char *vm)
{
    monConfig->type = VIR_DOMAIN_CHR_TYPE_UNIX;
    monConfig->data.nix.listen = true;

2832 2833 2834 2835
    if (virAsprintf(&monConfig->data.nix.path, "%s/%s.monitor",
                    cfg->libDir, vm) < 0)
        return -1;
    return 0;
2836 2837 2838
}


2839
/*
2840 2841
 * Precondition: vm must be locked, and a job must be active.
 * This method will call {Enter,Exit}Monitor
2842
 */
E
Eric Blake 已提交
2843
int
2844
qemuProcessStartCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
2845
                     virConnectPtr conn, virDomainRunningReason reason,
2846
                     qemuDomainAsyncJob asyncJob)
2847
{
2848
    int ret = -1;
2849
    qemuDomainObjPrivatePtr priv = vm->privateData;
2850
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2851

2852
    VIR_DEBUG("Using lock state '%s'", NULLSTR(priv->lockState));
2853
    if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
2854
                                   vm, priv->lockState) < 0) {
2855 2856 2857 2858
        /* 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
         */
2859
        goto cleanup;
2860 2861 2862
    }
    VIR_FREE(priv->lockState);

2863 2864
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto release;
J
Jiri Denemark 已提交
2865

2866 2867 2868 2869 2870 2871 2872
    ret = qemuMonitorStartCPUs(priv->mon, conn);
    qemuDomainObjExitMonitor(driver, vm);

    if (ret < 0)
        goto release;

    virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
2873

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

 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;
2883 2884 2885
}


2886
int qemuProcessStopCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
2887
                        virDomainPausedReason reason,
2888
                        qemuDomainAsyncJob asyncJob)
2889
{
2890
    int ret = -1;
2891 2892
    qemuDomainObjPrivatePtr priv = vm->privateData;

2893
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
2894

2895 2896
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
J
Jiri Denemark 已提交
2897

2898 2899 2900 2901 2902 2903 2904 2905 2906 2907
    ret = qemuMonitorStopCPUs(priv->mon);
    qemuDomainObjExitMonitor(driver, vm);

    if (ret < 0)
        goto cleanup;

    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
2908

2909
 cleanup:
2910 2911 2912 2913 2914
    return ret;
}



2915 2916 2917
static int
qemuProcessNotifyNets(virDomainDefPtr def)
{
2918
    size_t i;
2919

2920 2921
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
2922
        if (networkNotifyActualDevice(def, net) < 0)
2923 2924 2925 2926 2927
            return -1;
    }
    return 0;
}

2928 2929 2930 2931
static int
qemuProcessFiltersInstantiate(virConnectPtr conn,
                              virDomainDefPtr def)
{
2932
    size_t i;
2933 2934 2935 2936

    if (!conn)
        return 1;

2937
    for (i = 0; i < def->nnets; i++) {
2938 2939
        virDomainNetDefPtr net = def->nets[i];
        if ((net->filter) && (net->ifname)) {
J
Ján Tomko 已提交
2940 2941
            if (virDomainConfNWFilterInstantiate(conn, def->uuid, net) < 0)
                return 1;
2942 2943 2944
        }
    }

J
Ján Tomko 已提交
2945
    return 0;
2946 2947
}

2948
static int
2949
qemuProcessUpdateState(virQEMUDriverPtr driver, virDomainObjPtr vm)
2950 2951 2952
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
2953
    virDomainPausedReason reason;
2954 2955
    virDomainState newState = VIR_DOMAIN_NOSTATE;
    int newReason;
2956
    bool running;
2957
    char *msg = NULL;
2958 2959
    int ret;

2960
    qemuDomainObjEnterMonitor(driver, vm);
2961
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
2962
    qemuDomainObjExitMonitor(driver, vm);
2963 2964 2965 2966 2967 2968 2969

    if (ret < 0 || !virDomainObjIsActive(vm))
        return -1;

    state = virDomainObjGetState(vm, NULL);

    if (state == VIR_DOMAIN_PAUSED && running) {
2970 2971
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
2972
        ignore_value(VIR_STRDUP_QUIET(msg, "was unpaused"));
2973
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
2974 2975 2976
        if (reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN) {
            newState = VIR_DOMAIN_SHUTDOWN;
            newReason = VIR_DOMAIN_SHUTDOWN_UNKNOWN;
2977
            ignore_value(VIR_STRDUP_QUIET(msg, "shutdown"));
2978
        } else if (reason == VIR_DOMAIN_PAUSED_CRASHED) {
2979 2980 2981
            newState = VIR_DOMAIN_CRASHED;
            newReason = VIR_DOMAIN_CRASHED_PANICKED;
            ignore_value(VIR_STRDUP_QUIET(msg, "crashed"));
2982 2983 2984
        } else {
            newState = VIR_DOMAIN_PAUSED;
            newReason = reason;
S
Stefan Berger 已提交
2985 2986
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
2987
        }
2988
    } else if (state == VIR_DOMAIN_SHUTOFF && running) {
2989 2990
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_BOOTED;
2991
        ignore_value(VIR_STRDUP_QUIET(msg, "finished booting"));
2992 2993 2994 2995 2996 2997
    }

    if (newState != VIR_DOMAIN_NOSTATE) {
        VIR_DEBUG("Domain %s %s while its monitor was disconnected;"
                  " changing state to %s (%s)",
                  vm->def->name,
2998
                  NULLSTR(msg),
2999 3000 3001 3002
                  virDomainStateTypeToString(newState),
                  virDomainStateReasonToString(newState, newReason));
        VIR_FREE(msg);
        virDomainObjSetState(vm, newState, newReason);
3003 3004 3005 3006 3007
    }

    return 0;
}

3008
static int
3009
qemuProcessRecoverMigration(virQEMUDriverPtr driver,
3010 3011
                            virDomainObjPtr vm,
                            virConnectPtr conn,
3012 3013
                            qemuDomainAsyncJob job,
                            qemuMigrationJobPhase phase,
3014 3015 3016
                            virDomainState state,
                            int reason)
{
3017 3018
    qemuDomainObjPrivatePtr priv = vm->privateData;

3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
        switch (phase) {
        case QEMU_MIGRATION_PHASE_NONE:
        case QEMU_MIGRATION_PHASE_PERFORM2:
        case QEMU_MIGRATION_PHASE_BEGIN3:
        case QEMU_MIGRATION_PHASE_PERFORM3:
        case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
        case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
        case QEMU_MIGRATION_PHASE_CONFIRM3:
        case QEMU_MIGRATION_PHASE_LAST:
            break;

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

        case QEMU_MIGRATION_PHASE_FINISH2:
            /* source domain is already killed so let's just resume the domain
             * and hope we are all set */
            VIR_DEBUG("Incoming migration finished, resuming domain %s",
                      vm->def->name);
            if (qemuProcessStartCPUs(driver, vm, conn,
3042 3043
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073
                VIR_WARN("Could not resume domain %s", vm->def->name);
            }
            break;

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

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

        case QEMU_MIGRATION_PHASE_PERFORM2:
        case QEMU_MIGRATION_PHASE_PERFORM3:
            /* migration is still in progress, let's cancel it and resume the
             * domain */
            VIR_DEBUG("Canceling unfinished outgoing migration of domain %s",
                      vm->def->name);
3074
            qemuDomainObjEnterMonitor(driver, vm);
3075 3076
            ignore_value(qemuMonitorMigrateCancel(priv->mon));
            qemuDomainObjExitMonitor(driver, vm);
3077 3078 3079 3080 3081 3082
            /* resume the domain but only if it was paused as a result of
             * migration */
            if (state == VIR_DOMAIN_PAUSED &&
                (reason == VIR_DOMAIN_PAUSED_MIGRATION ||
                 reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
                if (qemuProcessStartCPUs(driver, vm, conn,
3083 3084
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103
                    VIR_WARN("Could not resume domain %s", vm->def->name);
                }
            }
            break;

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

        case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
            /* Finish3 failed, we need to resume the domain */
            VIR_DEBUG("Resuming domain %s after failed migration",
                      vm->def->name);
            if (state == VIR_DOMAIN_PAUSED &&
                (reason == VIR_DOMAIN_PAUSED_MIGRATION ||
                 reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
                if (qemuProcessStartCPUs(driver, vm, conn,
3104 3105
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119
                    VIR_WARN("Could not resume domain %s", vm->def->name);
                }
            }
            break;

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

    return 0;
}

3120
static int
3121
qemuProcessRecoverJob(virQEMUDriverPtr driver,
3122 3123 3124 3125
                      virDomainObjPtr vm,
                      virConnectPtr conn,
                      const struct qemuDomainJobObj *job)
{
3126
    qemuDomainObjPrivatePtr priv = vm->privateData;
3127 3128 3129 3130 3131 3132 3133 3134
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
    case QEMU_ASYNC_JOB_MIGRATION_IN:
3135 3136 3137
        if (qemuProcessRecoverMigration(driver, vm, conn, job->asyncJob,
                                        job->phase, state, reason) < 0)
            return -1;
3138 3139 3140 3141
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
3142
    case QEMU_ASYNC_JOB_SNAPSHOT:
3143
        qemuDomainObjEnterMonitor(driver, vm);
3144 3145
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
        qemuDomainObjExitMonitor(driver, vm);
3146
        /* resume the domain but only if it was paused as a result of
3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162
         * running a migration-to-file operation.  Although we are
         * recovering an async job, this function is run at startup
         * and must resume things using sync monitor connections.  */
         if (state == VIR_DOMAIN_PAUSED &&
             ((job->asyncJob == QEMU_ASYNC_JOB_DUMP &&
               reason == VIR_DOMAIN_PAUSED_DUMP) ||
              (job->asyncJob == QEMU_ASYNC_JOB_SAVE &&
               reason == VIR_DOMAIN_PAUSED_SAVE) ||
              (job->asyncJob == QEMU_ASYNC_JOB_SNAPSHOT &&
               reason == VIR_DOMAIN_PAUSED_SNAPSHOT) ||
              reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
             if (qemuProcessStartCPUs(driver, vm, conn,
                                      VIR_DOMAIN_RUNNING_UNPAUSED,
                                      QEMU_ASYNC_JOB_NONE) < 0) {
                 VIR_WARN("Could not resume domain '%s' after migration to file",
                          vm->def->name);
3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174
            }
        }
        break;

    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        break;
    }

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

3175 3176 3177 3178
    /* 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.
     */
3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198
    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;

3199
    case QEMU_JOB_MIGRATION_OP:
3200
    case QEMU_JOB_ABORT:
3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211
    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;
}

3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226
static int
qemuProcessUpdateDevices(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDeviceDef dev;
    char **old;
    char **tmp;
    int ret = -1;

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

    old = priv->qemuDevices;
    priv->qemuDevices = NULL;
3227
    if (qemuDomainUpdateDeviceList(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239
        goto cleanup;

    if ((tmp = old)) {
        while (*tmp) {
            if (!virStringArrayHasString(priv->qemuDevices, *tmp) &&
                virDomainDefFindDevice(vm->def, *tmp, &dev, false) == 0)
                qemuDomainRemoveDevice(driver, vm, &dev);
            tmp++;
        }
    }
    ret = 0;

3240
 cleanup:
3241 3242 3243 3244
    virStringFreeList(old);
    return ret;
}

3245 3246
struct qemuProcessReconnectData {
    virConnectPtr conn;
3247
    virQEMUDriverPtr driver;
3248 3249
    void *payload;
    struct qemuDomainJobObj oldjob;
3250 3251 3252 3253
};
/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
S
Stefan Berger 已提交
3254 3255 3256 3257
 *
 * We own the virConnectPtr we are passed here - whoever started
 * this thread function has increased the reference counter to it
 * so that we now have to close it.
3258 3259
 */
static void
3260
qemuProcessReconnect(void *opaque)
3261 3262
{
    struct qemuProcessReconnectData *data = opaque;
3263
    virQEMUDriverPtr driver = data->driver;
3264
    virDomainObjPtr obj = data->payload;
3265 3266
    qemuDomainObjPrivatePtr priv;
    virConnectPtr conn = data->conn;
3267
    struct qemuDomainJobObj oldjob;
3268 3269
    int state;
    int reason;
3270
    virQEMUDriverConfigPtr cfg;
3271
    size_t i;
3272
    int ret;
3273

3274 3275 3276 3277
    memcpy(&oldjob, &data->oldjob, sizeof(oldjob));

    VIR_FREE(data);

3278
    virObjectLock(obj);
3279

3280
    cfg = virQEMUDriverGetConfig(driver);
3281 3282 3283 3284
    VIR_DEBUG("Reconnect monitor to %p '%s'", obj, obj->def->name);

    priv = obj->privateData;

3285 3286
    /* Job was started by the caller for us */
    qemuDomainObjTransferJob(obj);
3287

3288 3289
    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted if qemuConnectMonitor() failed */
3290
    virObjectRef(obj);
3291 3292

    /* XXX check PID liveliness & EXE path */
3293
    if (qemuConnectMonitor(driver, obj, QEMU_ASYNC_JOB_NONE, -1) < 0)
3294 3295
        goto error;

D
Daniel P. Berrange 已提交
3296
    /* Failure to connect to agent shouldn't be fatal */
3297 3298 3299 3300
    if ((ret = qemuConnectAgent(driver, obj)) < 0) {
        if (ret == -2)
            goto error;

D
Daniel P. Berrange 已提交
3301 3302 3303 3304 3305 3306
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 obj->def->name);
        virResetLastError();
        priv->agentError = true;
    }

3307
    if (qemuUpdateActivePCIHostdevs(driver, obj->def) < 0) {
3308 3309 3310
        goto error;
    }

3311
    if (qemuUpdateActiveUSBHostdevs(driver, obj->def) < 0)
3312 3313
        goto error;

3314
    if (qemuUpdateActiveSCSIHostdevs(driver, obj->def) < 0)
3315 3316
        goto error;

3317
    if (qemuConnectCgroup(driver, obj) < 0)
3318 3319
        goto error;

3320
    /* XXX: Need to change as long as lock is introduced for
3321
     * qemu_driver->sharedDevices.
3322 3323
     */
    for (i = 0; i < obj->def->ndisks; i++) {
3324 3325
        virDomainDeviceDef dev;

3326 3327
        if (qemuTranslateDiskSourcePool(conn, obj->def->disks[i]) < 0)
            goto error;
3328

3329 3330 3331 3332 3333
        /* XXX we should be able to restore all data from XML in the future */
        if (qemuDomainDetermineDiskChain(driver, obj,
                                         obj->def->disks[i], true) < 0)
            goto error;

3334 3335 3336
        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = obj->def->disks[i];
        if (qemuAddSharedDevice(driver, &dev, obj->def->name) < 0)
3337 3338 3339
            goto error;
    }

3340 3341 3342
    if (qemuProcessUpdateState(driver, obj) < 0)
        goto error;

3343 3344
    state = virDomainObjGetState(obj, &reason);
    if (state == VIR_DOMAIN_SHUTOFF) {
3345 3346 3347 3348 3349
        VIR_DEBUG("Domain '%s' wasn't fully started yet, killing it",
                  obj->def->name);
        goto error;
    }

3350 3351 3352
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
3353 3354 3355
    if (!priv->qemuCaps &&
        !(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
                                                      obj->def->emulator)))
3356 3357
        goto error;

3358
    /* In case the domain shutdown while we were not running,
3359
     * we need to finish the shutdown process. And we need to do it after
3360
     * we have virQEMUCaps filled in.
3361
     */
3362 3363 3364 3365 3366 3367
    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);
3368 3369 3370
        goto endjob;
    }

3371 3372
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE))
        if ((qemuDomainAssignAddresses(obj->def, priv->qemuCaps, obj)) < 0)
3373
            goto error;
3374

3375
    if (virSecurityManagerReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
3376 3377
        goto error;

3378 3379 3380
    if (qemuProcessNotifyNets(obj->def) < 0)
        goto error;

3381 3382 3383
    if (qemuProcessFiltersInstantiate(conn, obj->def))
        goto error;

3384
    if (qemuDomainCheckEjectableMedia(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
3385 3386
        goto error;

3387 3388 3389
    if (qemuProcessRecoverJob(driver, obj, conn, &oldjob) < 0)
        goto error;

3390 3391 3392
    if (qemuProcessUpdateDevices(driver, obj) < 0)
        goto error;

3393
    /* update domain state XML with possibly updated state in virDomainObj */
3394
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, obj) < 0)
3395 3396
        goto error;

3397 3398
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3399
        char *xml = qemuDomainDefFormatXML(driver, obj->def, 0);
3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413
        int hookret;

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

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

3414
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
3415 3416
        driver->inhibitCallback(true, driver->inhibitOpaque);

3417
 endjob:
3418
    if (!qemuDomainObjEndJob(driver, obj))
3419 3420
        obj = NULL;

3421
    if (obj && virObjectUnref(obj))
3422
        virObjectUnlock(obj);
3423

3424
    virObjectUnref(conn);
3425
    virObjectUnref(cfg);
S
Stefan Berger 已提交
3426

3427 3428
    return;

3429
 error:
3430
    if (!qemuDomainObjEndJob(driver, obj))
3431 3432
        obj = NULL;

3433 3434
    if (obj) {
        if (!virDomainObjIsActive(obj)) {
3435
            if (virObjectUnref(obj))
3436
                virObjectUnlock(obj);
3437
        } else if (virObjectUnref(obj)) {
3438
            /* We can't get the monitor back, so must kill the VM
3439 3440 3441
             * to remove danger of it ending up running twice if
             * user tries to start it again later
             */
3442
            if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NO_SHUTDOWN)) {
3443 3444 3445 3446 3447 3448 3449 3450 3451
                /* 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;
            }
3452
            qemuProcessStop(driver, obj, state, 0);
3453
            if (!obj->persistent)
3454
                qemuDomainRemoveInactive(driver, obj);
3455
            else
3456
                virObjectUnlock(obj);
3457
        }
3458
    }
3459
    virObjectUnref(conn);
3460
    virObjectUnref(cfg);
3461 3462
}

3463 3464
static int
qemuProcessReconnectHelper(virDomainObjPtr obj,
3465 3466 3467 3468 3469 3470
                           void *opaque)
{
    virThread thread;
    struct qemuProcessReconnectData *src = opaque;
    struct qemuProcessReconnectData *data;

3471 3472 3473
    if (!obj->pid)
        return 0;

3474
    if (VIR_ALLOC(data) < 0)
3475
        return -1;
3476 3477

    memcpy(data, src, sizeof(*data));
3478
    data->payload = obj;
3479

3480
    /*
3481 3482
     * We create a separate thread to run qemuProcessReconnect in it.
     * However, qemuProcessReconnect needs to:
3483
     * 1. just before monitor reconnect do lightweight MonitorEnter
3484
     *    (increase VM refcount, unlock VM & driver)
3485 3486 3487 3488
     * 2. reconnect to monitor
     * 3. do lightweight MonitorExit (lock VM)
     * 4. continue reconnect process
     * 5. EndJob
3489 3490 3491 3492 3493 3494
     *
     * NB, we can't do normal MonitorEnter & MonitorExit because
     * these two lock the monitor lock, which does not exists in
     * this early phase.
     */

3495
    virObjectLock(obj);
3496 3497 3498

    qemuDomainObjRestoreJob(obj, &data->oldjob);

3499
    if (qemuDomainObjBeginJob(src->driver, obj, QEMU_JOB_MODIFY) < 0)
3500 3501
        goto error;

S
Stefan Berger 已提交
3502 3503 3504 3505
    /* Since we close the connection later on, we have to make sure
     * that the threads we start see a valid connection throughout their
     * lifetime. We simply increase the reference counter here.
     */
3506
    virObjectRef(data->conn);
S
Stefan Berger 已提交
3507

3508
    if (virThreadCreate(&thread, false, qemuProcessReconnect, data) < 0) {
S
Stefan Berger 已提交
3509

3510
        virObjectUnref(data->conn);
S
Stefan Berger 已提交
3511

3512 3513 3514
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not create thread. QEMU initialization "
                         "might be incomplete"));
3515
        if (!qemuDomainObjEndJob(src->driver, obj)) {
3516
            obj = NULL;
3517
        } else if (virObjectUnref(obj)) {
3518 3519
           /* We can't spawn a thread and thus connect to monitor.
            * Kill qemu */
3520
            qemuProcessStop(src->driver, obj, VIR_DOMAIN_SHUTOFF_FAILED, 0);
3521
            if (!obj->persistent)
3522
                qemuDomainRemoveInactive(src->driver, obj);
3523
            else
3524
                virObjectUnlock(obj);
3525 3526 3527 3528
        }
        goto error;
    }

3529
    virObjectUnlock(obj);
3530

3531
    return 0;
3532

3533
 error:
3534
    VIR_FREE(data);
3535
    return -1;
3536 3537 3538 3539 3540 3541 3542 3543 3544
}

/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
3545
qemuProcessReconnectAll(virConnectPtr conn, virQEMUDriverPtr driver)
3546
{
3547
    struct qemuProcessReconnectData data = {.conn = conn, .driver = driver};
3548
    virDomainObjListForEach(driver->domains, qemuProcessReconnectHelper, &data);
3549 3550
}

3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573
static int
qemuProcessVNCAllocatePorts(virQEMUDriverPtr driver,
                            virDomainGraphicsDefPtr graphics)
{
    unsigned short port;

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

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

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

    return 0;
}
3574 3575 3576 3577 3578 3579 3580 3581

static int
qemuProcessSPICEAllocatePorts(virQEMUDriverPtr driver,
                              virQEMUDriverConfigPtr cfg,
                              virDomainGraphicsDefPtr graphics)
{
    unsigned short port = 0;
    unsigned short tlsPort;
3582
    size_t i;
3583 3584 3585 3586 3587 3588 3589
    int defaultMode = graphics->data.spice.defaultMode;

    bool needTLSPort = false;
    bool needPort = false;

    if (graphics->data.spice.autoport) {
        /* check if tlsPort or port need allocation */
3590
        for (i = 0; i < VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_LAST; i++) {
3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604
            switch (graphics->data.spice.channels[i]) {
            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
                needTLSPort = true;
                break;

            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
                needPort = true;
                break;

            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
                switch (defaultMode) {
                case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
                    needTLSPort = true;
                    break;
3605

3606 3607 3608 3609 3610
                case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
                    needPort = true;
                    break;

                case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
3611 3612
                    if (cfg->spiceTLS)
                        needTLSPort = true;
3613 3614 3615 3616 3617 3618 3619 3620 3621
                    needPort = true;
                    break;
                }
                break;
            }
        }
    }

    if (needPort || graphics->data.spice.port == -1) {
3622
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
3623
            goto error;
3624 3625 3626 3627

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

3628 3629
    if (needTLSPort || graphics->data.spice.tlsPort == -1) {
        if (!cfg->spiceTLS) {
3630 3631 3632 3633
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Auto allocation of spice TLS port requested "
                             "but spice TLS is disabled in qemu.conf"));
            goto error;
3634
        }
3635 3636 3637 3638 3639

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

        graphics->data.spice.tlsPort = tlsPort;
3640 3641
    }

3642
    return 0;
3643

3644
 error:
3645
    virPortAllocatorRelease(driver->remotePorts, port);
3646
    return -1;
3647 3648 3649
}


3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667
static bool
qemuValidateCpuMax(virDomainDefPtr def, virQEMUCapsPtr qemuCaps)
{
    unsigned int maxCpus;

    maxCpus = virQEMUCapsGetMachineMaxCpus(qemuCaps, def->os.machine);
    if (!maxCpus)
        return true;

    if (def->maxvcpus > maxCpus) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       "%s", _("Maximum CPUs greater than specified machine type limit"));
        return false;
    }

    return true;
}

3668 3669

static bool
3670 3671 3672
qemuProcessVerifyGuestCPU(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          int asyncJob)
3673 3674 3675 3676 3677
{
    virDomainDefPtr def = vm->def;
    virArch arch = def->os.arch;
    virCPUDataPtr guestcpu = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3678
    int rc;
3679
    bool ret = false;
J
Ján Tomko 已提交
3680
    size_t i;
3681 3682 3683 3684

    switch (arch) {
    case VIR_ARCH_I686:
    case VIR_ARCH_X86_64:
3685 3686
        if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
            return false;
3687
        rc = qemuMonitorGetGuestCPU(priv->mon, arch, &guestcpu);
3688 3689
        qemuDomainObjExitMonitor(driver, vm);

3690 3691 3692 3693
        if (rc < 0) {
            if (rc == -2)
                break;

3694
            goto cleanup;
3695
        }
3696

J
Ján Tomko 已提交
3697
        if (def->features[VIR_DOMAIN_FEATURE_PVSPINLOCK] == VIR_TRISTATE_SWITCH_ON) {
3698 3699 3700 3701 3702 3703
            if (!cpuHasFeature(guestcpu, VIR_CPU_x86_KVM_PV_UNHALT)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("host doesn't support paravirtual spinlocks"));
                goto cleanup;
            }
        }
J
Ján Tomko 已提交
3704

3705
        for (i = 0; def->cpu && i < def->cpu->nfeatures; i++) {
J
Ján Tomko 已提交
3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717
            virCPUFeatureDefPtr feature = &def->cpu->features[i];

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

            if (STREQ(feature->name, "invtsc") &&
                !cpuHasFeature(guestcpu, feature->name)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("host doesn't support invariant TSC"));
                goto cleanup;
            }
        }
3718 3719 3720 3721 3722 3723 3724 3725
        break;

    default:
        break;
    }

    ret = true;

3726
 cleanup:
3727 3728 3729 3730 3731
    cpuDataFree(guestcpu);
    return ret;
}


3732
int qemuProcessStart(virConnectPtr conn,
3733
                     virQEMUDriverPtr driver,
3734
                     virDomainObjPtr vm,
3735
                     int asyncJob,
3736 3737 3738
                     const char *migrateFrom,
                     int stdin_fd,
                     const char *stdin_path,
3739
                     virDomainSnapshotObjPtr snapshot,
3740
                     virNetDevVPortProfileOp vmop,
3741
                     unsigned int flags)
3742 3743 3744 3745 3746 3747 3748 3749 3750
{
    int ret;
    off_t pos = -1;
    char ebuf[1024];
    int logfile = -1;
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCommandPtr cmd = NULL;
    struct qemuProcessHookData hookData;
E
Eric Blake 已提交
3751
    unsigned long cur_balloon;
3752
    size_t i;
3753
    char *nodeset = NULL;
3754
    virBitmapPtr nodemask = NULL;
3755
    unsigned int stop_flags;
3756
    virQEMUDriverConfigPtr cfg;
3757
    virCapsPtr caps = NULL;
3758
    unsigned int hostdev_flags = 0;
3759

3760 3761 3762 3763
    VIR_DEBUG("vm=%p name=%s id=%d pid=%llu",
              vm, vm->def->name, vm->def->id,
              (unsigned long long)vm->pid);

3764 3765 3766 3767
    /* Okay, these are just internal flags,
     * but doesn't hurt to check */
    virCheckFlags(VIR_QEMU_PROCESS_START_COLD |
                  VIR_QEMU_PROCESS_START_PAUSED |
3768
                  VIR_QEMU_PROCESS_START_AUTODESTROY, -1);
3769

3770 3771
    cfg = virQEMUDriverGetConfig(driver);

3772 3773 3774 3775 3776
    /* From now on until domain security labeling is done:
     * if any operation fails and we goto cleanup, we must not
     * restore any security label as we would overwrite labels
     * we did not set. */
    stop_flags = VIR_QEMU_PROCESS_STOP_NO_RELABEL;
3777 3778 3779 3780
    /* If we fail while doing incoming migration, then we must not
     * relabel, as the source is still using the files.  */
    if (migrateFrom)
        stop_flags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
3781

3782 3783 3784
    hookData.conn = conn;
    hookData.vm = vm;
    hookData.driver = driver;
3785 3786
    /* We don't increase cfg's reference counter here. */
    hookData.cfg = cfg;
3787

3788
    VIR_DEBUG("Beginning VM startup process");
3789 3790

    if (virDomainObjIsActive(vm)) {
3791 3792
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
3793
        virObjectUnref(cfg);
3794 3795 3796
        return -1;
    }

3797 3798 3799
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

3800 3801 3802 3803
    /* 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
     */
3804
    VIR_DEBUG("Setting current domain def as transient");
3805
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm, true) < 0)
3806 3807
        goto cleanup;

3808
    vm->def->id = qemuDriverAllocateID(driver);
3809
    qemuDomainSetFakeReboot(driver, vm, false);
3810
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, VIR_DOMAIN_SHUTOFF_UNKNOWN);
3811

3812
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
3813 3814
        driver->inhibitCallback(true, driver->inhibitOpaque);

3815
    /* Run an early hook to set-up missing devices */
3816
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3817
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
3818 3819 3820
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
3821 3822
                              VIR_HOOK_QEMU_OP_PREPARE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
3823 3824 3825 3826 3827 3828 3829 3830 3831
        VIR_FREE(xml);

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

3832 3833 3834 3835 3836 3837
    VIR_DEBUG("Determining emulator version");
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
                                                      vm->def->emulator)))
        goto cleanup;

3838 3839 3840 3841 3842 3843 3844 3845
    /* network devices must be "prepared" before hostdevs, because
     * setting up a network device might create a new hostdev that
     * will need to be setup.
     */
    VIR_DEBUG("Preparing network devices");
    if (qemuNetworkPrepareDevices(vm->def) < 0)
       goto cleanup;

3846
    /* Must be run before security labelling */
3847
    VIR_DEBUG("Preparing host devices");
3848 3849
    if (!cfg->relaxedACS)
        hostdev_flags |= VIR_HOSTDEV_STRICT_ACS_CHECK;
3850 3851
    if (!migrateFrom)
        hostdev_flags |= VIR_HOSTDEV_COLD_BOOT;
3852
    if (qemuPrepareHostDevices(driver, vm->def, priv->qemuCaps,
3853
                               hostdev_flags) < 0)
3854 3855
        goto cleanup;

3856
    VIR_DEBUG("Preparing chr devices");
3857 3858 3859 3860 3861 3862 3863 3864
    if (virDomainChrDefForeach(vm->def,
                               true,
                               qemuProcessPrepareChardevDevice,
                               NULL) < 0)
        goto cleanup;

    /* If you are using a SecurityDriver with dynamic labelling,
       then generate a security label for isolation */
3865
    VIR_DEBUG("Generating domain security label (if required)");
3866
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0) {
3867
        virDomainAuditSecurityLabel(vm, false);
3868 3869
        goto cleanup;
    }
3870
    virDomainAuditSecurityLabel(vm, true);
3871

3872
    if (vm->def->mem.nhugepages) {
3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886
        for (i = 0; i < cfg->nhugetlbfs; i++) {
            char *hugepagePath = qemuGetHugepagePath(&cfg->hugetlbfs[i]);

            if (!hugepagePath)
                goto cleanup;

            if (virSecurityManagerSetHugepages(driver->securityManager,
                                               vm->def, hugepagePath) < 0) {
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("Unable to set huge path in security driver"));
                VIR_FREE(hugepagePath);
                goto cleanup;
            }
            VIR_FREE(hugepagePath);
3887 3888 3889
        }
    }

3890 3891
    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
3892
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
3893
    qemuRemoveCgroup(vm);
3894

3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929
    for (i = 0; i < vm->def->ngraphics; ++i) {
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
            !graphics->data.vnc.autoport) {
            if (virPortAllocatorSetUsed(driver->remotePorts,
                                        graphics->data.vnc.port,
                                        true) < 0) {
                goto cleanup;
            }

            graphics->data.vnc.portReserved = true;

        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE &&
                   !graphics->data.spice.autoport) {

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

                graphics->data.spice.portReserved = true;
            }

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

                graphics->data.spice.tlsPortReserved = true;
            }
        }
    }

3930
    for (i = 0; i < vm->def->ngraphics; ++i) {
3931
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
3932 3933
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (qemuProcessVNCAllocatePorts(driver, graphics) < 0)
3934
                goto cleanup;
3935
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
3936 3937
            if (qemuProcessSPICEAllocatePorts(driver, cfg, graphics) < 0)
                goto cleanup;
3938
        }
3939

3940 3941
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ||
            graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
3942
            if (graphics->nListens == 0) {
3943
                if (VIR_EXPAND_N(graphics->listens, graphics->nListens, 1) < 0)
3944 3945
                    goto cleanup;
                graphics->listens[0].type = VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_ADDRESS;
3946 3947 3948
                if (VIR_STRDUP(graphics->listens[0].address,
                               graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ?
                               cfg->vncListen : cfg->spiceListen) < 0) {
3949 3950 3951
                    VIR_SHRINK_N(graphics->listens, graphics->nListens, 1);
                    goto cleanup;
                }
3952
                graphics->listens[0].fromConfig = true;
3953 3954
            }
        }
3955 3956
    }

3957
    if (virFileMakePath(cfg->logDir) < 0) {
3958 3959
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
3960
                             cfg->logDir);
3961 3962 3963
        goto cleanup;
    }

3964
    VIR_DEBUG("Creating domain log file");
3965
    if ((logfile = qemuDomainCreateLog(driver, vm, false)) < 0)
3966 3967
        goto cleanup;

3968 3969
    if (vm->def->virtType == VIR_DOMAIN_VIRT_KVM) {
        VIR_DEBUG("Checking for KVM availability");
3970
        if (!virFileExists("/dev/kvm")) {
3971 3972 3973 3974
            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."));
3975 3976 3977 3978
            goto cleanup;
        }
    }

3979 3980 3981
    if (!qemuValidateCpuMax(vm->def, priv->qemuCaps))
        goto cleanup;

3982
    if (qemuAssignDeviceAliases(vm->def, priv->qemuCaps) < 0)
3983 3984
        goto cleanup;

3985 3986 3987
    /* Get the advisory nodeset from numad if 'placement' of
     * either <vcpu> or <numatune> is 'auto'.
     */
3988
    if (virDomainDefNeedsPlacementAdvice(vm->def)) {
3989
        nodeset = virNumaGetAutoPlacementAdvice(vm->def->vcpus,
3990
                                                vm->def->mem.max_balloon);
3991 3992 3993 3994 3995
        if (!nodeset)
            goto cleanup;

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

3996
        if (virBitmapParse(nodeset, 0, &nodemask, VIR_DOMAIN_CPUMASK_LEN) < 0)
3997 3998 3999 4000
            goto cleanup;
    }
    hookData.nodemask = nodemask;

4001 4002 4003 4004 4005 4006 4007 4008
    /* "volume" type disk's source must be translated before
     * cgroup and security setting.
     */
    for (i = 0; i < vm->def->ndisks; i++) {
        if (qemuTranslateDiskSourcePool(conn, vm->def->disks[i]) < 0)
            goto cleanup;
    }

4009 4010 4011 4012
    if (qemuDomainCheckDiskPresence(driver, vm,
                                    flags & VIR_QEMU_PROCESS_START_COLD) < 0)
        goto cleanup;

4013
    if (VIR_ALLOC(priv->monConfig) < 0)
4014 4015
        goto cleanup;

4016
    VIR_DEBUG("Preparing monitor state");
4017
    if (qemuProcessPrepareMonitorChr(cfg, priv->monConfig, vm->def->name) < 0)
4018 4019
        goto cleanup;

E
Eric Blake 已提交
4020
    priv->monJSON = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON);
4021 4022
    priv->monError = false;
    priv->monStart = 0;
4023 4024
    priv->gotShutdown = false;

4025
    VIR_FREE(priv->pidfile);
4026
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, vm->def->name))) {
4027 4028
        virReportSystemError(errno,
                             "%s", _("Failed to build pidfile path."));
4029 4030 4031
        goto cleanup;
    }

4032 4033
    if (unlink(priv->pidfile) < 0 &&
        errno != ENOENT) {
4034
        virReportSystemError(errno,
4035 4036
                             _("Cannot remove stale PID file %s"),
                             priv->pidfile);
4037 4038 4039 4040 4041 4042 4043
        goto cleanup;
    }

    /*
     * Normally PCI addresses are assigned in the virDomainCreate
     * or virDomainDefine methods. We might still need to assign
     * some here to cope with the question of upgrades. Regardless
M
Martin Kletzander 已提交
4044
     * we also need to populate the PCI address set cache for later
4045 4046
     * use in hotplug
     */
4047
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4048
        VIR_DEBUG("Assigning domain PCI addresses");
4049
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
4050
            goto cleanup;
4051 4052
    }

4053
    VIR_DEBUG("Building emulator command line");
4054
    if (!(cmd = qemuBuildCommandLine(conn, driver, vm->def, priv->monConfig,
E
Eric Blake 已提交
4055
                                     priv->monJSON, priv->qemuCaps,
4056
                                     migrateFrom, stdin_fd, snapshot, vmop,
C
Cole Robinson 已提交
4057
                                     &buildCommandLineCallbacks, false)))
4058 4059 4060 4061
        goto cleanup;

    /* now that we know it is about to start call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4062
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4063 4064 4065
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4066 4067
                              VIR_HOOK_QEMU_OP_START, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
4068 4069 4070 4071 4072 4073 4074 4075 4076
        VIR_FREE(xml);

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

4077
    if ((timestamp = virTimeStringNow()) == NULL) {
4078 4079 4080 4081 4082
        goto cleanup;
    } else {
        if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
            safewrite(logfile, START_POSTFIX, strlen(START_POSTFIX)) < 0) {
            VIR_WARN("Unable to write timestamp to logfile: %s",
4083
                     virStrerror(errno, ebuf, sizeof(ebuf)));
4084 4085 4086 4087 4088 4089 4090
        }

        VIR_FREE(timestamp);
    }

    virCommandWriteArgLog(cmd, logfile);

4091
    qemuDomainObjCheckTaint(driver, vm, logfile);
4092

4093 4094
    if ((pos = lseek(logfile, 0, SEEK_END)) < 0)
        VIR_WARN("Unable to seek to end of logfile: %s",
4095
                 virStrerror(errno, ebuf, sizeof(ebuf)));
4096 4097

    VIR_DEBUG("Clear emulator capabilities: %d",
4098 4099
              cfg->clearEmulatorCapabilities);
    if (cfg->clearEmulatorCapabilities)
4100 4101
        virCommandClearCaps(cmd);

4102 4103
    /* in case a certain disk is desirous of CAP_SYS_RAWIO, add this */
    for (i = 0; i < vm->def->ndisks; i++) {
4104
        virDomainDeviceDef dev;
4105 4106
        virDomainDiskDefPtr disk = vm->def->disks[i];

4107
        if (vm->def->disks[i]->rawio == 1)
R
Roman Bogorodskiy 已提交
4108
#ifdef CAP_SYS_RAWIO
4109
            virCommandAllowCap(cmd, CAP_SYS_RAWIO);
R
Roman Bogorodskiy 已提交
4110 4111 4112 4113
#else
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Raw I/O is not supported on this platform"));
#endif
4114

4115 4116 4117
        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        if (qemuAddSharedDevice(driver, &dev, vm->def->name) < 0)
4118
            goto cleanup;
4119

4120
        if (qemuSetUnprivSGIO(&dev) < 0)
4121
            goto cleanup;
4122 4123
    }

4124
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);
4125 4126
    virCommandSetMaxProcesses(cmd, cfg->maxProcesses);
    virCommandSetMaxFiles(cmd, cfg->maxFiles);
4127

4128 4129 4130 4131 4132 4133
    VIR_DEBUG("Setting up security labelling");
    if (virSecurityManagerSetChildProcessLabel(driver->securityManager,
                                               vm->def, cmd) < 0) {
        goto cleanup;
    }

4134 4135 4136
    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
4137
    virCommandSetPidFile(cmd, priv->pidfile);
4138
    virCommandDaemonize(cmd);
4139
    virCommandRequireHandshake(cmd);
4140

4141 4142
    if (virSecurityManagerPreFork(driver->securityManager) < 0)
        goto cleanup;
4143
    ret = virCommandRun(cmd, NULL);
4144
    virSecurityManagerPostFork(driver->securityManager);
4145

E
Eric Blake 已提交
4146
    /* wait for qemu process to show up */
4147
    if (ret == 0) {
4148
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
4149 4150
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
4151 4152
            ret = -1;
        }
4153 4154 4155 4156 4157
        VIR_DEBUG("QEMU vm=%p name=%s running with pid=%llu",
                  vm, vm->def->name, (unsigned long long)vm->pid);
    } else {
        VIR_DEBUG("QEMU vm=%p name=%s failed to spawn",
                  vm, vm->def->name);
4158 4159
    }

4160
    VIR_DEBUG("Writing early domain status to disk");
4161
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
4162 4163 4164
        goto cleanup;
    }

4165 4166
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
4167 4168
        /* Read errors from child that occurred between fork and exec. */
        qemuProcessReadChildErrors(driver, vm, pos);
4169 4170 4171
        goto cleanup;
    }

4172 4173 4174 4175 4176 4177 4178 4179 4180 4181
    VIR_DEBUG("Setting up domain cgroup (if required)");
    if (qemuSetupCgroup(driver, vm, nodemask) < 0)
        goto cleanup;

    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
    if (!vm->def->cputune.emulatorpin &&
        qemuProcessInitCpuAffinity(driver, vm, nodemask) < 0)
        goto cleanup;

4182 4183
    VIR_DEBUG("Setting domain security labels");
    if (virSecurityManagerSetAllLabel(driver->securityManager,
4184
                                      vm->def, stdin_path) < 0)
4185 4186
        goto cleanup;

4187 4188 4189 4190 4191 4192
    /* Security manager labeled all devices, therefore
     * if any operation from now on fails and we goto cleanup,
     * where virSecurityManagerRestoreAllLabel() is called
     * (hidden under qemuProcessStop) we need to restore labels. */
    stop_flags &= ~VIR_QEMU_PROCESS_STOP_NO_RELABEL;

4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206
    if (stdin_fd != -1) {
        /* 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");

        if (fstat(stdin_fd, &stdin_sb) < 0) {
            virReportSystemError(errno,
                                 _("cannot stat fd %d"), stdin_fd);
            goto cleanup;
        }
        if (S_ISFIFO(stdin_sb.st_mode) &&
4207
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def, stdin_fd) < 0)
4208 4209 4210 4211 4212 4213 4214 4215 4216
            goto cleanup;
    }

    VIR_DEBUG("Labelling done, completing handshake to child");
    if (virCommandHandshakeNotify(cmd) < 0) {
        goto cleanup;
    }
    VIR_DEBUG("Handshake complete, child running");

4217
    if (migrateFrom)
4218
        flags |= VIR_QEMU_PROCESS_START_PAUSED;
4219 4220 4221 4222 4223 4224 4225

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

    if (ret == -1) /* The VM failed to start */
        goto cleanup;

4226
    VIR_DEBUG("Waiting for monitor to show up");
4227
    if (qemuProcessWaitForMonitor(driver, vm, asyncJob, priv->qemuCaps, pos) < 0)
4228 4229
        goto cleanup;

D
Daniel P. Berrange 已提交
4230
    /* Failure to connect to agent shouldn't be fatal */
4231 4232 4233 4234
    if ((ret = qemuConnectAgent(driver, vm)) < 0) {
        if (ret == -2)
            goto cleanup;

D
Daniel P. Berrange 已提交
4235 4236 4237 4238 4239 4240
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

4241
    VIR_DEBUG("Detecting if required emulator features are present");
4242
    if (!qemuProcessVerifyGuestCPU(driver, vm, asyncJob))
4243 4244
        goto cleanup;

4245 4246 4247 4248
    VIR_DEBUG("Setting up post-init cgroup restrictions");
    if (qemuSetupCgroupPostInit(vm, nodemask) < 0)
        goto cleanup;

4249
    VIR_DEBUG("Detecting VCPU PIDs");
4250
    if (qemuProcessDetectVcpuPIDs(driver, vm, asyncJob) < 0)
4251 4252
        goto cleanup;

4253
    VIR_DEBUG("Setting cgroup for each VCPU (if required)");
4254
    if (qemuSetupCgroupForVcpu(vm) < 0)
4255 4256
        goto cleanup;

4257
    VIR_DEBUG("Setting cgroup for emulator (if required)");
4258
    if (qemuSetupCgroupForEmulator(driver, vm, nodemask) < 0)
4259 4260
        goto cleanup;

4261
    VIR_DEBUG("Setting VCPU affinities");
4262
    if (qemuProcessSetVcpuAffinities(vm) < 0)
4263 4264
        goto cleanup;

4265
    VIR_DEBUG("Setting affinity of emulator threads");
4266
    if (qemuProcessSetEmulatorAffinities(conn, vm) < 0)
4267 4268
        goto cleanup;

4269
    VIR_DEBUG("Setting any required VM passwords");
4270
    if (qemuProcessInitPasswords(conn, driver, vm, asyncJob) < 0)
4271 4272 4273 4274
        goto cleanup;

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
4275
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4276
        VIR_DEBUG("Determining domain device PCI addresses");
4277
        if (qemuProcessInitPCIAddresses(driver, vm, asyncJob) < 0)
4278 4279 4280
            goto cleanup;
    }

4281 4282 4283 4284
    /* set default link states */
    /* qemu doesn't support setting this on the command line, so
     * enter the monitor */
    VIR_DEBUG("Setting network link states");
4285 4286
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
4287
    if (qemuProcessSetLinkStates(vm) < 0) {
4288
        qemuDomainObjExitMonitor(driver, vm);
4289 4290 4291
        goto cleanup;
    }

4292
    qemuDomainObjExitMonitor(driver, vm);
4293

4294
    VIR_DEBUG("Fetching list of active devices");
4295
    if (qemuDomainUpdateDeviceList(driver, vm, asyncJob) < 0)
4296 4297
        goto cleanup;

4298 4299 4300 4301
    /* Technically, qemuProcessStart can be called from inside
     * QEMU_ASYNC_JOB_MIGRATION_IN, but we are okay treating this like
     * a sync job since no other job can call into the domain until
     * migration completes.  */
4302
    VIR_DEBUG("Setting initial memory amount");
E
Eric Blake 已提交
4303
    cur_balloon = vm->def->mem.cur_balloon;
4304
    if (cur_balloon != vm->def->mem.cur_balloon) {
4305 4306 4307
        virReportError(VIR_ERR_OVERFLOW,
                       _("unable to set balloon to %lld"),
                       vm->def->mem.cur_balloon);
4308 4309
        goto cleanup;
    }
4310 4311
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
4312 4313
    if (vm->def->memballoon && vm->def->memballoon->period)
        qemuMonitorSetMemoryStatsPeriod(priv->mon, vm->def->memballoon->period);
E
Eric Blake 已提交
4314
    if (qemuMonitorSetBalloon(priv->mon, cur_balloon) < 0) {
4315
        qemuDomainObjExitMonitor(driver, vm);
4316 4317
        goto cleanup;
    }
4318
    qemuDomainObjExitMonitor(driver, vm);
4319

4320
    if (!(flags & VIR_QEMU_PROCESS_START_PAUSED)) {
4321
        VIR_DEBUG("Starting domain CPUs");
4322
        /* Allow the CPUS to start executing */
J
Jiri Denemark 已提交
4323
        if (qemuProcessStartCPUs(driver, vm, conn,
4324 4325
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 QEMU_ASYNC_JOB_NONE) < 0) {
4326
            if (virGetLastError() == NULL)
4327 4328
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("resume operation failed"));
4329 4330
            goto cleanup;
        }
J
Jiri Denemark 已提交
4331 4332 4333 4334 4335
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             migrateFrom ?
                             VIR_DOMAIN_PAUSED_MIGRATION :
                             VIR_DOMAIN_PAUSED_USER);
4336 4337
    }

4338
    if (flags & VIR_QEMU_PROCESS_START_AUTODESTROY &&
4339 4340 4341
        qemuProcessAutoDestroyAdd(driver, vm, conn) < 0)
        goto cleanup;

4342
    VIR_DEBUG("Writing domain status to disk");
4343
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
4344 4345
        goto cleanup;

4346 4347
    /* finally we can call the 'started' hook script if any */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4348
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362
        int hookret;

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

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

4363 4364 4365
    /* unset reporting errors from qemu log */
    qemuMonitorSetDomainLog(priv->mon, -1);

4366 4367
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);
4368
    virObjectUnref(cfg);
4369
    virObjectUnref(caps);
4370 4371 4372

    return 0;

4373
 cleanup:
4374 4375 4376
    /* We jump here if we failed to start the VM for any reason, or
     * if we failed to initialize the now running VM. kill it off and
     * pretend we never started it */
4377
    VIR_FREE(nodeset);
4378
    virBitmapFree(nodemask);
4379 4380
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);
4381 4382
    if (priv->mon)
        qemuMonitorSetDomainLog(priv->mon, -1);
4383
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, stop_flags);
4384
    virObjectUnref(cfg);
4385
    virObjectUnref(caps);
4386 4387 4388 4389 4390

    return -1;
}


4391
int
4392
qemuProcessKill(virDomainObjPtr vm, unsigned int flags)
4393
{
4394
    int ret;
4395

4396 4397 4398
    VIR_DEBUG("vm=%p name=%s pid=%llu flags=%x",
              vm, vm->def->name,
              (unsigned long long)vm->pid, flags);
4399

4400 4401 4402 4403 4404
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
4405 4406
    }

4407
    if (flags & VIR_QEMU_PROCESS_KILL_NOWAIT) {
4408 4409 4410 4411 4412
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
4413

4414 4415
    ret = virProcessKillPainfully(vm->pid,
                                  !!(flags & VIR_QEMU_PROCESS_KILL_FORCE));
4416

4417
    return ret;
4418 4419 4420
}


4421
void qemuProcessStop(virQEMUDriverPtr driver,
4422
                     virDomainObjPtr vm,
4423 4424
                     virDomainShutoffReason reason,
                     unsigned int flags)
4425 4426 4427 4428 4429 4430
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
4431
    virNetDevVPortProfilePtr vport = NULL;
4432
    size_t i;
4433 4434 4435
    int logfile = -1;
    char *timestamp;
    char ebuf[1024];
4436
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4437

4438 4439 4440
    VIR_DEBUG("Shutting down vm=%p name=%s id=%d pid=%llu flags=%x",
              vm, vm->def->name, vm->def->id,
              (unsigned long long)vm->pid, flags);
4441 4442 4443

    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("VM '%s' not active", vm->def->name);
4444
        virObjectUnref(cfg);
4445 4446 4447
        return;
    }

4448 4449 4450 4451
    /* This method is routinely used in clean up paths. Disable error
     * reporting so we don't squash a legit error. */
    orig_err = virSaveLastError();

4452
    /*
4453 4454
     * We may unlock the vm in qemuProcessKill(), and another thread
     * can lock the vm, and then call qemuProcessStop(). So we should
4455 4456 4457 4458
     * set vm->def->id to -1 here to avoid qemuProcessStop() to be called twice.
     */
    vm->def->id = -1;

4459
    if (virAtomicIntDecAndTest(&driver->nactive) && driver->inhibitCallback)
4460 4461
        driver->inhibitCallback(false, driver->inhibitOpaque);

4462
    if ((logfile = qemuDomainCreateLog(driver, vm, true)) < 0) {
4463 4464 4465
        /* To not break the normal domain shutdown process, skip the
         * timestamp log writing if failed on opening log file. */
        VIR_WARN("Unable to open logfile: %s",
4466
                  virStrerror(errno, ebuf, sizeof(ebuf)));
4467
    } else {
4468
        if ((timestamp = virTimeStringNow()) != NULL) {
4469 4470 4471 4472
            if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
                safewrite(logfile, SHUTDOWN_POSTFIX,
                          strlen(SHUTDOWN_POSTFIX)) < 0) {
                VIR_WARN("Unable to write timestamp to logfile: %s",
4473
                         virStrerror(errno, ebuf, sizeof(ebuf)));
4474 4475 4476 4477 4478 4479 4480
            }

            VIR_FREE(timestamp);
        }

        if (VIR_CLOSE(logfile) < 0)
             VIR_WARN("Unable to close logfile: %s",
4481
                      virStrerror(errno, ebuf, sizeof(ebuf)));
4482 4483 4484 4485
    }

    virDomainConfVMNWFilterTeardown(vm);

4486
    if (cfg->macFilter) {
4487
        def = vm->def;
4488
        for (i = 0; i < def->nnets; i++) {
4489 4490 4491
            virDomainNetDefPtr net = def->nets[i];
            if (net->ifname == NULL)
                continue;
4492 4493 4494
            ignore_value(ebtablesRemoveForwardAllowIn(driver->ebtables,
                                                      net->ifname,
                                                      &net->mac));
4495 4496 4497
        }
    }

4498 4499
    virPortAllocatorRelease(driver->remotePorts, priv->nbdPort);
    priv->nbdPort = 0;
4500

D
Daniel P. Berrange 已提交
4501 4502 4503 4504 4505 4506
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
        priv->agentError = false;
    }

4507
    if (priv->mon) {
4508
        qemuMonitorClose(priv->mon);
4509 4510
        priv->mon = NULL;
    }
4511 4512 4513 4514 4515 4516 4517 4518

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

4519 4520 4521 4522 4523 4524
    ignore_value(virDomainChrDefForeach(vm->def,
                                        false,
                                        qemuProcessCleanupChardevDevice,
                                        NULL));


4525
    /* shut it off for sure */
4526 4527 4528
    ignore_value(qemuProcessKill(vm,
                                 VIR_QEMU_PROCESS_KILL_FORCE|
                                 VIR_QEMU_PROCESS_KILL_NOCHECK));
4529

4530 4531
    qemuDomainCleanupRun(driver, vm);

4532
    /* Stop autodestroy in case guest is restarted */
4533
    qemuProcessAutoDestroyRemove(driver, vm);
4534

4535 4536
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4537
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4538 4539 4540

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4541 4542
                    VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
4543 4544 4545
        VIR_FREE(xml);
    }

4546 4547 4548 4549
    /* Reset Security Labels unless caller don't want us to */
    if (!(flags & VIR_QEMU_PROCESS_STOP_NO_RELABEL))
        virSecurityManagerRestoreAllLabel(driver->securityManager,
                                          vm->def,
4550
                                          !!(flags & VIR_QEMU_PROCESS_STOP_MIGRATED));
4551
    virSecurityManagerReleaseLabel(driver->securityManager, vm->def);
4552

4553
    for (i = 0; i < vm->def->ndisks; i++) {
4554
        virDomainDeviceDef dev;
4555
        virDomainDiskDefPtr disk = vm->def->disks[i];
4556 4557 4558 4559

        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        ignore_value(qemuRemoveSharedDevice(driver, &dev, vm->def->name));
4560 4561
    }

4562
    /* Clear out dynamically assigned labels */
4563 4564 4565 4566 4567
    for (i = 0; i < vm->def->nseclabels; i++) {
        if (vm->def->seclabels[i]->type == VIR_DOMAIN_SECLABEL_DYNAMIC) {
            VIR_FREE(vm->def->seclabels[i]->label);
        }
        VIR_FREE(vm->def->seclabels[i]->imagelabel);
4568 4569
    }

4570 4571 4572
    virStringFreeList(priv->qemuDevices);
    priv->qemuDevices = NULL;

4573 4574 4575
    virDomainDefClearDeviceAliases(vm->def);
    if (!priv->persistentAddrs) {
        virDomainDefClearPCIAddresses(vm->def);
4576
        virDomainPCIAddressSetFree(priv->pciaddrs);
4577
        priv->pciaddrs = NULL;
4578
        virDomainDefClearCCWAddresses(vm->def);
J
Ján Tomko 已提交
4579
        virDomainCCWAddressSetFree(priv->ccwaddrs);
4580
        priv->ccwaddrs = NULL;
4581 4582 4583 4584 4585 4586 4587
    }

    qemuDomainReAttachHostDevices(driver, vm->def);

    def = vm->def;
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
4588 4589 4590 4591
        vport = virDomainNetGetActualVirtPortProfile(net);

        switch (virDomainNetGetActualType(net)) {
        case VIR_DOMAIN_NET_TYPE_DIRECT:
4592
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
4593
                             net->ifname, &net->mac,
4594 4595
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
4596
                             virDomainNetGetActualVirtPortProfile(net),
4597
                             cfg->stateDir));
4598
            VIR_FREE(net->ifname);
4599 4600 4601 4602 4603 4604 4605 4606
            break;
        case VIR_DOMAIN_NET_TYPE_BRIDGE:
        case VIR_DOMAIN_NET_TYPE_NETWORK:
#ifdef VIR_NETDEV_TAP_REQUIRE_MANUAL_CLEANUP
            if (!(vport && vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH))
                ignore_value(virNetDevTapDelete(net->ifname));
#endif
            break;
4607
        }
4608 4609 4610
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
A
Ansis Atteka 已提交
4611 4612 4613 4614 4615
        if (vport && vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH)
            ignore_value(virNetDevOpenvswitchRemovePort(
                                       virDomainNetGetActualBridgeName(net),
                                       net->ifname));

4616 4617
        /* kick the device out of the hostdev list too */
        virDomainNetRemoveHostdev(def, net);
4618
        networkReleaseActualDevice(vm->def, net);
4619
    }
4620

4621
 retry:
4622
    if ((ret = qemuRemoveCgroup(vm)) < 0) {
4623 4624 4625 4626 4627 4628 4629
        if (ret == -EBUSY && (retries++ < 5)) {
            usleep(200*1000);
            goto retry;
        }
        VIR_WARN("Failed to remove cgroup for %s",
                 vm->def->name);
    }
4630
    virCgroupFree(&priv->cgroup);
4631 4632 4633

    qemuProcessRemoveDomainStatus(driver, vm);

4634 4635
    /* Remove VNC and Spice ports from port reservation bitmap, but only if
       they were reserved by the driver (autoport=yes)
4636
    */
4637
    for (i = 0; i < vm->def->ngraphics; ++i) {
4638
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
4639 4640
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (graphics->data.vnc.autoport) {
4641 4642
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.vnc.port);
4643
            }
4644 4645 4646 4647 4648 4649
            else if (graphics->data.vnc.portReserved) {
                virPortAllocatorSetUsed(driver->remotePorts,
                                        graphics->data.spice.port,
                                        false);
                graphics->data.vnc.portReserved = false;
            }
4650 4651
            virPortAllocatorRelease(driver->webSocketPorts,
                                    graphics->data.vnc.websocket);
4652
        }
4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (graphics->data.spice.autoport) {
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.spice.port);
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.spice.tlsPort);
            } else {
                if (graphics->data.spice.portReserved) {
                    virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.port,
                                            false);
                    graphics->data.spice.portReserved = false;
                }

                if (graphics->data.spice.tlsPortReserved) {
                    virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.tlsPort,
                                            false);
                    graphics->data.spice.tlsPortReserved = false;
                }
            }
4674
        }
4675 4676
    }

4677
    vm->taint = 0;
4678
    vm->pid = -1;
J
Jiri Denemark 已提交
4679
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
4680 4681
    VIR_FREE(priv->vcpupids);
    priv->nvcpupids = 0;
4682 4683
    virObjectUnref(priv->qemuCaps);
    priv->qemuCaps = NULL;
4684
    VIR_FREE(priv->pidfile);
4685

4686
    /* The "release" hook cleans up additional resources */
4687
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4688
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4689 4690 4691

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4692 4693
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
4694 4695 4696
        VIR_FREE(xml);
    }

4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707
    if (vm->newDef) {
        virDomainDefFree(vm->def);
        vm->def = vm->newDef;
        vm->def->id = -1;
        vm->newDef = NULL;
    }

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
4708
    virObjectUnref(cfg);
4709
}
4710 4711


4712
int qemuProcessAttach(virConnectPtr conn ATTRIBUTE_UNUSED,
4713
                      virQEMUDriverPtr driver,
4714
                      virDomainObjPtr vm,
4715
                      pid_t pid,
4716 4717 4718 4719
                      const char *pidfile,
                      virDomainChrSourceDefPtr monConfig,
                      bool monJSON)
{
4720
    size_t i;
4721 4722 4723 4724 4725
    char ebuf[1024];
    int logfile = -1;
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool running = true;
4726
    virDomainPausedReason reason;
4727
    virSecurityLabelPtr seclabel = NULL;
4728
    virSecurityLabelDefPtr seclabeldef = NULL;
4729
    bool seclabelgen = false;
4730 4731
    virSecurityManagerPtr* sec_managers = NULL;
    const char *model;
4732
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4733
    virCapsPtr caps = NULL;
4734
    bool active = false;
4735
    int ret;
4736 4737 4738 4739

    VIR_DEBUG("Beginning VM attach process");

    if (virDomainObjIsActive(vm)) {
4740 4741
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
4742
        virObjectUnref(cfg);
4743 4744 4745
        return -1;
    }

4746
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
4747
        goto error;
4748

4749 4750 4751 4752 4753
    /* 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");
4754
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm, true) < 0)
4755
        goto error;
4756

4757
    vm->def->id = qemuDriverAllocateID(driver);
4758

4759
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
4760
        driver->inhibitCallback(true, driver->inhibitOpaque);
4761
    active = true;
4762

4763
    if (virFileMakePath(cfg->logDir) < 0) {
4764 4765
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
4766
                             cfg->logDir);
4767
        goto error;
4768 4769 4770
    }

    VIR_FREE(priv->pidfile);
4771
    if (VIR_STRDUP(priv->pidfile, pidfile) < 0)
4772
        goto error;
4773 4774

    VIR_DEBUG("Detect security driver config");
4775
    sec_managers = virSecurityManagerGetNested(driver->securityManager);
4776 4777
    if (sec_managers == NULL)
        goto error;
4778 4779

    for (i = 0; sec_managers[i]; i++) {
4780
        seclabelgen = false;
4781 4782
        model = virSecurityManagerGetModel(sec_managers[i]);
        seclabeldef = virDomainDefGetSecurityLabelDef(vm->def, model);
4783
        if (seclabeldef == NULL) {
4784
            if (!(seclabeldef = virSecurityLabelDefNew(model)))
4785 4786 4787
                goto error;
            seclabelgen = true;
        }
4788 4789
        seclabeldef->type = VIR_DOMAIN_SECLABEL_STATIC;
        if (VIR_ALLOC(seclabel) < 0)
4790
            goto error;
4791 4792
        if (virSecurityManagerGetProcessLabel(driver->securityManager,
                                              vm->def, vm->pid, seclabel) < 0)
4793
            goto error;
4794

4795
        if (VIR_STRDUP(seclabeldef->model, model) < 0)
4796
            goto error;
4797

4798
        if (VIR_STRDUP(seclabeldef->label, seclabel->label) < 0)
4799
            goto error;
4800
        VIR_FREE(seclabel);
4801 4802 4803 4804 4805 4806

        if (seclabelgen) {
            if (VIR_APPEND_ELEMENT(vm->def->seclabels, vm->def->nseclabels, seclabeldef) < 0)
                goto error;
            seclabelgen = false;
        }
4807
    }
4808 4809 4810

    VIR_DEBUG("Creating domain log file");
    if ((logfile = qemuDomainCreateLog(driver, vm, false)) < 0)
4811
        goto error;
4812 4813

    VIR_DEBUG("Determining emulator version");
4814 4815 4816
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
                                                      vm->def->emulator)))
4817
        goto error;
4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829

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

    priv->gotShutdown = false;

    /*
     * Normally PCI addresses are assigned in the virDomainCreate
     * or virDomainDefine methods. We might still need to assign
     * some here to cope with the question of upgrades. Regardless
M
Martin Kletzander 已提交
4830
     * we also need to populate the PCI address set cache for later
4831 4832
     * use in hotplug
     */
4833
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4834
        VIR_DEBUG("Assigning domain PCI addresses");
4835
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
4836
            goto error;
4837 4838
    }

4839
    if ((timestamp = virTimeStringNow()) == NULL) {
4840
        goto error;
4841 4842 4843 4844
    } else {
        if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
            safewrite(logfile, ATTACH_POSTFIX, strlen(ATTACH_POSTFIX)) < 0) {
            VIR_WARN("Unable to write timestamp to logfile: %s",
4845
                     virStrerror(errno, ebuf, sizeof(ebuf)));
4846 4847 4848 4849 4850 4851 4852 4853 4854 4855
        }

        VIR_FREE(timestamp);
    }

    qemuDomainObjTaint(driver, vm, VIR_DOMAIN_TAINT_EXTERNAL_LAUNCH, logfile);

    vm->pid = pid;

    VIR_DEBUG("Waiting for monitor to show up");
4856
    if (qemuProcessWaitForMonitor(driver, vm, QEMU_ASYNC_JOB_NONE, priv->qemuCaps, -1) < 0)
4857
        goto error;
4858

D
Daniel P. Berrange 已提交
4859
    /* Failure to connect to agent shouldn't be fatal */
4860 4861 4862 4863
    if ((ret = qemuConnectAgent(driver, vm)) < 0) {
        if (ret == -2)
            goto error;

D
Daniel P. Berrange 已提交
4864 4865 4866 4867 4868 4869
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

4870
    VIR_DEBUG("Detecting VCPU PIDs");
4871
    if (qemuProcessDetectVcpuPIDs(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
4872
        goto error;
4873 4874 4875

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
4876
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4877
        VIR_DEBUG("Determining domain device PCI addresses");
4878
        if (qemuProcessInitPCIAddresses(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
4879
            goto error;
4880 4881 4882
    }

    VIR_DEBUG("Getting initial memory amount");
4883
    qemuDomainObjEnterMonitor(driver, vm);
4884
    if (qemuMonitorGetBalloonInfo(priv->mon, &vm->def->mem.cur_balloon) < 0) {
4885
        qemuDomainObjExitMonitor(driver, vm);
4886
        goto error;
4887
    }
4888
    if (qemuMonitorGetStatus(priv->mon, &running, &reason) < 0) {
4889
        qemuDomainObjExitMonitor(driver, vm);
4890
        goto error;
4891 4892
    }
    if (qemuMonitorGetVirtType(priv->mon, &vm->def->virtType) < 0) {
4893
        qemuDomainObjExitMonitor(driver, vm);
4894
        goto error;
4895
    }
4896
    qemuDomainObjExitMonitor(driver, vm);
4897 4898

    if (!virDomainObjIsActive(vm))
4899
        goto error;
4900

4901
    if (running) {
4902 4903
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                             VIR_DOMAIN_RUNNING_UNPAUSED);
4904 4905 4906 4907 4908 4909 4910
        if (vm->def->memballoon && vm->def->memballoon->period) {
            qemuDomainObjEnterMonitor(driver, vm);
            qemuMonitorSetMemoryStatsPeriod(priv->mon,
                                            vm->def->memballoon->period);
            qemuDomainObjExitMonitor(driver, vm);
        }
    } else {
4911
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
4912
    }
4913 4914

    VIR_DEBUG("Writing domain status to disk");
4915
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
4916
        goto error;
4917

4918 4919
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4920
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931
        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)
4932
            goto error;
4933 4934
    }

4935 4936
    VIR_FORCE_CLOSE(logfile);
    VIR_FREE(seclabel);
4937
    VIR_FREE(sec_managers);
4938
    virObjectUnref(cfg);
4939
    virObjectUnref(caps);
4940 4941 4942

    return 0;

4943
 error:
4944 4945 4946 4947 4948 4949
    /* 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);
4950 4951
    VIR_FORCE_CLOSE(logfile);
    VIR_FREE(seclabel);
4952
    VIR_FREE(sec_managers);
4953 4954
    if (seclabelgen)
        virSecurityLabelDefFree(seclabeldef);
4955
    virDomainChrSourceDefFree(monConfig);
4956
    virObjectUnref(cfg);
4957
    virObjectUnref(caps);
4958 4959 4960 4961
    return -1;
}


4962
static virDomainObjPtr
4963 4964 4965
qemuProcessAutoDestroy(virDomainObjPtr dom,
                       virConnectPtr conn,
                       void *opaque)
4966
{
4967
    virQEMUDriverPtr driver = opaque;
4968
    qemuDomainObjPrivatePtr priv = dom->privateData;
4969
    virObjectEventPtr event = NULL;
4970

4971
    VIR_DEBUG("vm=%s, conn=%p", dom->def->name, conn);
4972

4973 4974
    if (priv->job.asyncJob) {
        VIR_DEBUG("vm=%s has long-term job active, cancelling",
4975
                  dom->def->name);
4976
        qemuDomainObjDiscardAsyncJob(driver, dom);
4977 4978
    }

4979 4980
    if (qemuDomainObjBeginJob(driver, dom,
                              QEMU_JOB_DESTROY) < 0)
4981 4982 4983
        goto cleanup;

    VIR_DEBUG("Killing domain");
4984

4985 4986
    qemuProcessStop(driver, dom, VIR_DOMAIN_SHUTOFF_DESTROYED,
                    VIR_QEMU_PROCESS_STOP_MIGRATED);
4987

4988
    virDomainAuditStop(dom, "destroyed");
4989
    event = virDomainEventLifecycleNewFromObj(dom,
4990 4991
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_DESTROYED);
4992

4993
    if (!qemuDomainObjEndJob(driver, dom))
4994
        dom = NULL;
4995
    if (dom && !dom->persistent) {
4996
        qemuDomainRemoveInactive(driver, dom);
4997 4998
        dom = NULL;
    }
4999
    if (event)
5000
        qemuDomainEventQueue(driver, event);
5001

5002
 cleanup:
5003
    return dom;
5004 5005
}

5006
int qemuProcessAutoDestroyAdd(virQEMUDriverPtr driver,
5007 5008 5009
                              virDomainObjPtr vm,
                              virConnectPtr conn)
{
5010
    VIR_DEBUG("vm=%s, conn=%p", vm->def->name, conn);
5011 5012
    return virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                qemuProcessAutoDestroy);
5013 5014
}

5015
int qemuProcessAutoDestroyRemove(virQEMUDriverPtr driver,
5016 5017
                                 virDomainObjPtr vm)
{
5018
    VIR_DEBUG("vm=%s", vm->def->name);
5019 5020
    return virCloseCallbacksUnset(driver->closeCallbacks, vm,
                                  qemuProcessAutoDestroy);
5021
}
5022

5023
bool qemuProcessAutoDestroyActive(virQEMUDriverPtr driver,
5024 5025
                                  virDomainObjPtr vm)
{
5026
    virCloseCallback cb;
5027
    VIR_DEBUG("vm=%s", vm->def->name);
5028
    cb = virCloseCallbacksGet(driver->closeCallbacks, vm, NULL);
5029
    return cb == qemuProcessAutoDestroy;
5030
}