qemu_process.c 182.0 KB
Newer Older
1
/*
2
 * qemu_process.c: QEMU process management
3
 *
4
 * Copyright (C) 2006-2015 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23 24 25 26 27
 *
 */

#include <config.h>

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

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

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

#define VIR_FROM_THIS VIR_FROM_QEMU

77 78
VIR_LOG_INIT("qemu.qemu_process");

79
#define ATTACH_POSTFIX ": attaching\n"
80 81 82
#define SHUTDOWN_POSTFIX ": shutting down\n"

/**
83
 * qemuProcessRemoveDomainStatus
84 85 86 87 88 89
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
90
qemuProcessRemoveDomainStatus(virQEMUDriverPtr driver,
91 92 93 94
                              virDomainObjPtr vm)
{
    char ebuf[1024];
    char *file = NULL;
95
    qemuDomainObjPrivatePtr priv = vm->privateData;
96 97
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;
98

99
    if (virAsprintf(&file, "%s/%s.xml", cfg->stateDir, vm->def->name) < 0)
100
        goto cleanup;
101 102 103 104 105 106

    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);

107 108 109
    if (priv->pidfile &&
        unlink(priv->pidfile) < 0 &&
        errno != ENOENT)
110 111 112
        VIR_WARN("Failed to remove PID file for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));

113
    ret = 0;
114
 cleanup:
115 116
    virObjectUnref(cfg);
    return ret;
117 118 119 120
}


/* XXX figure out how to remove this */
121
extern virQEMUDriverPtr qemu_driver;
122

D
Daniel P. Berrange 已提交
123 124 125 126 127 128 129
/*
 * 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
130
qemuProcessHandleAgentEOF(qemuAgentPtr agent,
D
Daniel P. Berrange 已提交
131 132 133 134 135 136
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

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

137
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
138 139

    priv = vm->privateData;
140 141 142 143 144 145 146 147 148 149 150 151

    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
152

153
    virObjectUnlock(vm);
154 155

    qemuAgentClose(agent);
156 157
    return;

158
 unlock:
159 160
    virObjectUnlock(vm);
    return;
D
Daniel P. Berrange 已提交
161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
}


/*
 * 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);

178
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
179 180 181 182 183

    priv = vm->privateData;

    priv->agentError = true;

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

static void qemuProcessHandleAgentDestroy(qemuAgentPtr agent,
                                          virDomainObjPtr vm)
{
190 191
    VIR_DEBUG("Received destroy agent=%p vm=%p", agent, vm);

192
    virObjectUnref(vm);
D
Daniel P. Berrange 已提交
193 194 195 196 197 198 199 200 201 202 203 204 205
}


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

static virDomainChrSourceDefPtr
qemuFindAgentConfig(virDomainDefPtr def)
{
    virDomainChrSourceDefPtr config = NULL;
206
    size_t i;
D
Daniel P. Berrange 已提交
207

208
    for (i = 0; i < def->nchannels; i++) {
D
Daniel P. Berrange 已提交
209 210 211 212 213
        virDomainChrDefPtr channel = def->channels[i];

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

214
        if (STREQ_NULLABLE(channel->target.name, "org.qemu.guest_agent.0")) {
D
Daniel P. Berrange 已提交
215 216 217 218 219 220 221 222 223
            config = &channel->source;
            break;
        }
    }

    return config;
}

static int
224
qemuConnectAgent(virQEMUDriverPtr driver, virDomainObjPtr vm)
D
Daniel P. Berrange 已提交
225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242
{
    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 */
243
    virObjectRef(vm);
D
Daniel P. Berrange 已提交
244 245

    ignore_value(virTimeMillisNow(&priv->agentStart));
246
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
247 248 249 250 251

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

252
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
253 254
    priv->agentStart = 0;

255 256 257 258 259 260 261 262 263 264 265
    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
266 267 268 269
    if (virSecurityManagerClearSocketLabel(driver->securityManager,
                                           vm->def) < 0) {
        VIR_ERROR(_("Failed to clear security context for agent for %s"),
                  vm->def->name);
270
        qemuAgentClose(agent);
D
Daniel P. Berrange 已提交
271 272 273 274 275 276 277 278 279 280 281 282 283
        goto cleanup;
    }


    priv->agent = agent;

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

    ret = 0;

284
 cleanup:
D
Daniel P. Berrange 已提交
285 286 287 288
    return ret;
}


289
/*
290
 * This is a callback registered with a qemuMonitorPtr instance,
291 292 293 294 295 296
 * 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,
297 298
                            virDomainObjPtr vm,
                            void *opaque)
299
{
300
    virQEMUDriverPtr driver = opaque;
301
    virObjectEventPtr event = NULL;
302
    qemuDomainObjPrivatePtr priv;
303 304 305
    int eventReason = VIR_DOMAIN_EVENT_STOPPED_SHUTDOWN;
    int stopReason = VIR_DOMAIN_SHUTOFF_SHUTDOWN;
    const char *auditReason = "shutdown";
306 307 308

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

309
    virObjectLock(vm);
310

311 312 313 314 315 316 317
    priv = vm->privateData;

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

318 319
    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Domain %p is not active, ignoring EOF", vm);
320
        goto unlock;
321 322
    }

323
    if (priv->monJSON && !priv->gotShutdown) {
324 325
        VIR_DEBUG("Monitor connection to '%s' closed without SHUTDOWN event; "
                  "assuming the domain crashed", vm->def->name);
326
        eventReason = VIR_DOMAIN_EVENT_STOPPED_FAILED;
327
        stopReason = VIR_DOMAIN_SHUTOFF_CRASHED;
328
        auditReason = "failed";
329 330
    }

331
    event = virDomainEventLifecycleNewFromObj(vm,
332
                                     VIR_DOMAIN_EVENT_STOPPED,
333
                                     eventReason);
334
    qemuProcessStop(driver, vm, stopReason, 0);
335
    virDomainAuditStop(vm, auditReason);
336

337
    if (!vm->persistent) {
338
        qemuDomainRemoveInactive(driver, vm);
339 340 341
        goto cleanup;
    }

342
 unlock:
343
    virObjectUnlock(vm);
344

345
 cleanup:
346
    if (event)
347
        qemuDomainEventQueue(driver, event);
348 349 350 351 352 353 354 355 356 357 358
}


/*
 * 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,
359 360
                              virDomainObjPtr vm,
                              void *opaque)
361
{
362
    virQEMUDriverPtr driver = opaque;
363
    virObjectEventPtr event = NULL;
364 365 366

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

367
    virObjectLock(vm);
368

369
    ((qemuDomainObjPrivatePtr) vm->privateData)->monError = true;
370 371 372 373
    event = virDomainEventControlErrorNewFromObj(vm);
    if (event)
        qemuDomainEventQueue(driver, event);

374
    virObjectUnlock(vm);
375 376 377 378 379 380 381
}


static virDomainDiskDefPtr
qemuProcessFindDomainDiskByPath(virDomainObjPtr vm,
                                const char *path)
{
382
    int idx = virDomainDiskIndexByName(vm->def, path, true);
383

384 385
    if (idx >= 0)
        return vm->def->disks[idx];
386

387 388 389
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with path %s"),
                   path);
390 391 392
    return NULL;
}

393
virDomainDiskDefPtr
394 395 396
qemuProcessFindDomainDiskByAlias(virDomainObjPtr vm,
                                 const char *alias)
{
397
    size_t i;
398 399 400 401 402 403 404 405 406 407 408 409

    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;
    }

410 411 412
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with alias %s"),
                   alias);
413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428
    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;

429
    if (!disk->src->encryption) {
430 431
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("disk %s does not have any encryption information"),
432
                       disk->src->path);
433 434
        return -1;
    }
435
    enc = disk->src->encryption;
436 437

    if (!conn) {
438 439
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("cannot find secrets without a connection"));
440 441 442 443
        goto cleanup;
    }

    if (conn->secretDriver == NULL ||
444 445
        conn->secretDriver->secretLookupByUUID == NULL ||
        conn->secretDriver->secretGetValue == NULL) {
446 447
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("secret storage not supported"));
448 449 450 451 452 453 454
        goto cleanup;
    }

    if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW ||
        enc->nsecrets != 1 ||
        enc->secrets[0]->type !=
        VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE) {
455
        virReportError(VIR_ERR_XML_ERROR,
456 457
                       _("invalid <encryption> for volume %s"),
                       virDomainDiskGetSource(disk));
458 459 460
        goto cleanup;
    }

461 462
    secret = conn->secretDriver->secretLookupByUUID(conn,
                                                    enc->secrets[0]->uuid);
463 464
    if (secret == NULL)
        goto cleanup;
465 466
    data = conn->secretDriver->secretGetValue(secret, &size, 0,
                                              VIR_SECRET_GET_VALUE_INTERNAL_CALL);
467
    virObjectUnref(secret);
468 469 470 471 472 473
    if (data == NULL)
        goto cleanup;

    if (memchr(data, '\0', size) != NULL) {
        memset(data, 0, size);
        VIR_FREE(data);
474 475
        virReportError(VIR_ERR_XML_ERROR,
                       _("format='qcow' passphrase for %s must not contain a "
476
                         "'\\0'"), virDomainDiskGetSource(disk));
477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495
        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;

496
 cleanup:
497 498 499 500 501 502 503 504 505
    return ret;
}

static int
qemuProcessFindVolumeQcowPassphrase(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virConnectPtr conn,
                                    virDomainObjPtr vm,
                                    const char *path,
                                    char **secretRet,
506 507
                                    size_t *secretLen,
                                    void *opaque ATTRIBUTE_UNUSED)
508 509 510 511
{
    virDomainDiskDefPtr disk;
    int ret = -1;

512
    virObjectLock(vm);
513 514 515 516 517 518 519
    disk = qemuProcessFindDomainDiskByPath(vm, path);

    if (!disk)
        goto cleanup;

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

520
 cleanup:
521
    virObjectUnlock(vm);
522 523 524 525 526 527
    return ret;
}


static int
qemuProcessHandleReset(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
528 529
                       virDomainObjPtr vm,
                       void *opaque)
530
{
531
    virQEMUDriverPtr driver = opaque;
532
    virObjectEventPtr event;
533
    qemuDomainObjPrivatePtr priv;
534
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
535

536
    virObjectLock(vm);
537

538
    event = virDomainEventRebootNewFromObj(vm);
539 540 541
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);
542 543 544

    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
        VIR_WARN("Failed to save status on vm %s", vm->def->name);
545

546
    virObjectUnlock(vm);
547

548
    if (event)
549 550
        qemuDomainEventQueue(driver, event);

551
    virObjectUnref(cfg);
552 553 554 555
    return 0;
}


556 557 558 559 560 561 562 563 564 565 566
/*
 * 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)
{
567
    virQEMUDriverPtr driver = qemu_driver;
568 569
    virDomainObjPtr vm = opaque;
    qemuDomainObjPrivatePtr priv = vm->privateData;
570
    virObjectEventPtr event = NULL;
571
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
572
    virDomainRunningReason reason = VIR_DOMAIN_RUNNING_BOOTED;
573
    int ret = -1, rc;
574

575
    VIR_DEBUG("vm=%p", vm);
576
    virObjectLock(vm);
577
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
578 579 580
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
581 582
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
583 584 585
        goto endjob;
    }

586
    qemuDomainObjEnterMonitor(driver, vm);
587 588 589
    rc = qemuMonitorSystemReset(priv->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0)
590 591
        goto endjob;

592
    if (rc < 0)
593 594
        goto endjob;

595 596 597
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_CRASHED)
        reason = VIR_DOMAIN_RUNNING_CRASHED;

598
    if (qemuProcessStartCPUs(driver, vm, NULL,
599
                             reason,
600
                             QEMU_ASYNC_JOB_NONE) < 0) {
601
        if (virGetLastError() == NULL)
602 603
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
604 605
        goto endjob;
    }
606
    priv->gotShutdown = false;
607
    event = virDomainEventLifecycleNewFromObj(vm,
608 609 610
                                     VIR_DOMAIN_EVENT_RESUMED,
                                     VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

611 612 613 614 615
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

616 617
    ret = 0;

618
 endjob:
619
    qemuDomainObjEndJob(driver, vm);
620

621
 cleanup:
622 623 624
    if (ret == -1)
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_FORCE));
    qemuDomObjEndAPI(&vm);
625 626
    if (event)
        qemuDomainEventQueue(driver, event);
627
    virObjectUnref(cfg);
628 629 630
}


631
void
632
qemuProcessShutdownOrReboot(virQEMUDriverPtr driver,
633
                            virDomainObjPtr vm)
634
{
635 636 637
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
638
        qemuDomainSetFakeReboot(driver, vm, false);
639
        virObjectRef(vm);
640 641 642 643 644
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
645
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
646
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
647
            virObjectUnref(vm);
648 649
        }
    } else {
650
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
651
    }
652
}
653

654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681

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;
}


682 683
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
684 685
                          virDomainObjPtr vm,
                          void *opaque)
686
{
687
    virQEMUDriverPtr driver = opaque;
688
    qemuDomainObjPrivatePtr priv;
689
    virObjectEventPtr event = NULL;
690
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
691

692 693
    VIR_DEBUG("vm=%p", vm);

694
    virObjectLock(vm);
695 696 697 698 699 700

    priv = vm->privateData;
    if (priv->gotShutdown) {
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
701 702 703 704
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
705 706 707 708 709 710 711 712
    }
    priv->gotShutdown = true;

    VIR_DEBUG("Transitioned guest %s to shutdown state",
              vm->def->name);
    virDomainObjSetState(vm,
                         VIR_DOMAIN_SHUTDOWN,
                         VIR_DOMAIN_SHUTDOWN_UNKNOWN);
713
    event = virDomainEventLifecycleNewFromObj(vm,
714 715 716
                                     VIR_DOMAIN_EVENT_SHUTDOWN,
                                     VIR_DOMAIN_EVENT_SHUTDOWN_FINISHED);

717
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
718 719 720 721
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

722 723 724
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

725 726
    qemuProcessShutdownOrReboot(driver, vm);

727
 unlock:
728
    virObjectUnlock(vm);
729
    if (event)
730
        qemuDomainEventQueue(driver, event);
731
    virObjectUnref(cfg);
732

733 734 735 736 737 738
    return 0;
}


static int
qemuProcessHandleStop(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
739 740
                      virDomainObjPtr vm,
                      void *opaque)
741
{
742
    virQEMUDriverPtr driver = opaque;
743
    virObjectEventPtr event = NULL;
744
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
745

746
    virObjectLock(vm);
J
Jiri Denemark 已提交
747
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
748
        qemuDomainObjPrivatePtr priv = vm->privateData;
749

750
        if (priv->gotShutdown) {
751 752
            VIR_DEBUG("Ignoring STOP event after SHUTDOWN");
            goto unlock;
753 754
        }

755 756
        VIR_DEBUG("Transitioned guest %s to paused state",
                  vm->def->name);
757

758 759 760
        if (priv->job.current)
            ignore_value(virTimeMillisNow(&priv->job.current->stopped));

761
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_UNKNOWN);
762
        event = virDomainEventLifecycleNewFromObj(vm,
763 764 765
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);

766 767 768 769 770
        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));

771
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
772 773 774
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
775
    }
776

777
 unlock:
778
    virObjectUnlock(vm);
779
    if (event)
780
        qemuDomainEventQueue(driver, event);
781
    virObjectUnref(cfg);
782 783 784 785 786

    return 0;
}


787 788
static int
qemuProcessHandleResume(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
789 790
                        virDomainObjPtr vm,
                        void *opaque)
791
{
792
    virQEMUDriverPtr driver = opaque;
793
    virObjectEventPtr event = NULL;
794
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
795

796
    virObjectLock(vm);
797 798 799 800 801 802 803 804 805 806 807 808 809
    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);
810
        event = virDomainEventLifecycleNewFromObj(vm,
811 812 813 814
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

        VIR_DEBUG("Using lock state '%s' on resume event", NULLSTR(priv->lockState));
815
        if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
816 817 818 819 820 821 822 823 824
                                       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);

825
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
826 827 828 829 830
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
    }

831
 unlock:
832
    virObjectUnlock(vm);
833
    if (event)
834
        qemuDomainEventQueue(driver, event);
835
    virObjectUnref(cfg);
836 837 838
    return 0;
}

839 840 841
static int
qemuProcessHandleRTCChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm,
842 843
                           long long offset,
                           void *opaque)
844
{
845
    virQEMUDriverPtr driver = opaque;
846
    virObjectEventPtr event = NULL;
847
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
848

849
    virObjectLock(vm);
850

851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867
    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;
868
        vm->def->clock.data.variable.adjustment = offset;
869

870 871 872 873 874
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
           VIR_WARN("unable to save domain status with RTC change");
    }

    event = virDomainEventRTCChangeNewFromObj(vm, offset);
875

876
    virObjectUnlock(vm);
877

878
    if (event)
879
        qemuDomainEventQueue(driver, event);
880
    virObjectUnref(cfg);
881 882 883 884 885 886 887
    return 0;
}


static int
qemuProcessHandleWatchdog(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
888 889
                          int action,
                          void *opaque)
890
{
891
    virQEMUDriverPtr driver = opaque;
892 893
    virObjectEventPtr watchdogEvent = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
894
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
895

896
    virObjectLock(vm);
897 898 899
    watchdogEvent = virDomainEventWatchdogNewFromObj(vm, action);

    if (action == VIR_DOMAIN_EVENT_WATCHDOG_PAUSE &&
J
Jiri Denemark 已提交
900
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
901
        qemuDomainObjPrivatePtr priv = vm->privateData;
902 903
        VIR_DEBUG("Transitioned guest %s to paused state due to watchdog", vm->def->name);

J
Jiri Denemark 已提交
904
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
905
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
906 907 908
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

909 910 911 912 913
        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));

914
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
915 916 917
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
918 919 920
    }

    if (vm->def->watchdog->action == VIR_DOMAIN_WATCHDOG_ACTION_DUMP) {
C
Chen Fan 已提交
921 922 923 924 925
        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
926 927 928
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
929
            virObjectRef(vm);
C
Chen Fan 已提交
930
            if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
931
                if (!virObjectUnref(vm))
932
                    vm = NULL;
C
Chen Fan 已提交
933
                VIR_FREE(processEvent);
W
Wen Congyang 已提交
934
            }
935
        }
936 937
    }

938
    if (vm)
939
        virObjectUnlock(vm);
940 941 942 943
    if (watchdogEvent)
        qemuDomainEventQueue(driver, watchdogEvent);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
944

945
    virObjectUnref(cfg);
946 947 948 949 950 951 952 953 954
    return 0;
}


static int
qemuProcessHandleIOError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virDomainObjPtr vm,
                         const char *diskAlias,
                         int action,
955 956
                         const char *reason,
                         void *opaque)
957
{
958
    virQEMUDriverPtr driver = opaque;
959 960 961
    virObjectEventPtr ioErrorEvent = NULL;
    virObjectEventPtr ioErrorEvent2 = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
962 963 964
    const char *srcPath;
    const char *devAlias;
    virDomainDiskDefPtr disk;
965
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
966

967
    virObjectLock(vm);
968 969 970
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
971
        srcPath = virDomainDiskGetSource(disk);
972 973 974 975 976 977 978 979 980 981
        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
982
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
983
        qemuDomainObjPrivatePtr priv = vm->privateData;
984 985
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

J
Jiri Denemark 已提交
986
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
987
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
988 989 990
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

991 992 993 994 995
        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));

996
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
997 998
            VIR_WARN("Unable to save status on vm %s after IO error", vm->def->name);
    }
999
    virObjectUnlock(vm);
1000

1001 1002 1003 1004 1005 1006
    if (ioErrorEvent)
        qemuDomainEventQueue(driver, ioErrorEvent);
    if (ioErrorEvent2)
        qemuDomainEventQueue(driver, ioErrorEvent2);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1007
    virObjectUnref(cfg);
1008 1009 1010
    return 0;
}

1011 1012 1013 1014 1015
static int
qemuProcessHandleBlockJob(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          const char *diskAlias,
                          int type,
1016 1017
                          int status,
                          void *opaque)
1018
{
1019
    virQEMUDriverPtr driver = opaque;
1020
    virObjectEventPtr event = NULL;
1021
    virObjectEventPtr event2 = NULL;
1022 1023
    const char *path;
    virDomainDiskDefPtr disk;
1024
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
E
Eric Blake 已提交
1025
    virDomainDiskDefPtr persistDisk = NULL;
1026
    bool save = false;
1027

1028
    virObjectLock(vm);
1029 1030 1031
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
1032 1033
        /* Have to generate two variants of the event for old vs. new
         * client callbacks */
E
Eric Blake 已提交
1034 1035 1036
        if (type == VIR_DOMAIN_BLOCK_JOB_TYPE_COMMIT &&
            disk->mirrorJob == VIR_DOMAIN_BLOCK_JOB_TYPE_ACTIVE_COMMIT)
            type = disk->mirrorJob;
1037
        path = virDomainDiskGetSource(disk);
1038
        event = virDomainEventBlockJobNewFromObj(vm, path, type, status);
1039 1040
        event2 = virDomainEventBlockJob2NewFromObj(vm, disk->dst, type,
                                                   status);
1041 1042 1043

        /* If we completed a block pull or commit, then update the XML
         * to match.  */
1044 1045
        switch ((virConnectDomainEventBlockJobStatus) status) {
        case VIR_DOMAIN_BLOCK_JOB_COMPLETED:
1046
            if (disk->mirrorState == VIR_DOMAIN_DISK_MIRROR_STATE_PIVOT) {
E
Eric Blake 已提交
1047 1048 1049 1050 1051 1052 1053 1054 1055 1056
                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,
1057
                                                             true) < 0) {
E
Eric Blake 已提交
1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071
                            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;
                    }
                }

1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
                /* 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
1092
            disk->mirrorJob = VIR_DOMAIN_BLOCK_JOB_TYPE_UNKNOWN;
1093 1094
            ignore_value(qemuDomainDetermineDiskChain(driver, vm, disk,
                                                      true, true));
1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105
            break;

        case VIR_DOMAIN_BLOCK_JOB_READY:
            disk->mirrorState = VIR_DOMAIN_DISK_MIRROR_STATE_READY;
            save = true;
            break;

        case VIR_DOMAIN_BLOCK_JOB_FAILED:
        case VIR_DOMAIN_BLOCK_JOB_CANCELED:
            virStorageSourceFree(disk->mirror);
            disk->mirror = NULL;
1106 1107
            disk->mirrorState = status == VIR_DOMAIN_BLOCK_JOB_FAILED ?
                VIR_DOMAIN_DISK_MIRROR_STATE_ABORT : VIR_DOMAIN_DISK_MIRROR_STATE_NONE;
1108 1109 1110 1111 1112 1113
            disk->mirrorJob = VIR_DOMAIN_BLOCK_JOB_TYPE_UNKNOWN;
            save = true;
            break;

        case VIR_DOMAIN_BLOCK_JOB_LAST:
            break;
1114
        }
1115 1116
    }

1117 1118 1119 1120
    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
1121 1122 1123 1124
        if (persistDisk && virDomainSaveConfig(cfg->configDir,
                                               vm->newDef) < 0)
            VIR_WARN("Unable to update persistent definition on vm %s "
                     "after block job", vm->def->name);
1125
    }
1126
    virObjectUnlock(vm);
1127
    virObjectUnref(cfg);
1128

1129
    if (event)
1130
        qemuDomainEventQueue(driver, event);
1131 1132
    if (event2)
        qemuDomainEventQueue(driver, event2);
1133 1134 1135

    return 0;
}
1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148

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,
1149 1150
                          const char *saslUsername,
                          void *opaque)
1151
{
1152
    virQEMUDriverPtr driver = opaque;
1153
    virObjectEventPtr event;
1154 1155 1156
    virDomainEventGraphicsAddressPtr localAddr = NULL;
    virDomainEventGraphicsAddressPtr remoteAddr = NULL;
    virDomainEventGraphicsSubjectPtr subject = NULL;
1157
    size_t i;
1158 1159

    if (VIR_ALLOC(localAddr) < 0)
1160
        goto error;
1161
    localAddr->family = localFamily;
1162 1163 1164
    if (VIR_STRDUP(localAddr->service, localService) < 0 ||
        VIR_STRDUP(localAddr->node, localNode) < 0)
        goto error;
1165 1166

    if (VIR_ALLOC(remoteAddr) < 0)
1167
        goto error;
1168
    remoteAddr->family = remoteFamily;
1169 1170 1171
    if (VIR_STRDUP(remoteAddr->service, remoteService) < 0 ||
        VIR_STRDUP(remoteAddr->node, remoteNode) < 0)
        goto error;
1172 1173

    if (VIR_ALLOC(subject) < 0)
1174
        goto error;
1175 1176
    if (x509dname) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1177
            goto error;
1178
        subject->nidentity++;
1179 1180 1181
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "x509dname") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, x509dname) < 0)
            goto error;
1182 1183 1184
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1185
            goto error;
1186
        subject->nidentity++;
1187 1188 1189
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "saslUsername") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, saslUsername) < 0)
            goto error;
1190 1191
    }

1192
    virObjectLock(vm);
1193
    event = virDomainEventGraphicsNewFromObj(vm, phase, localAddr, remoteAddr, authScheme, subject);
1194
    virObjectUnlock(vm);
1195

1196
    if (event)
1197 1198 1199 1200
        qemuDomainEventQueue(driver, event);

    return 0;

1201
 error:
1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212
    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) {
1213
        for (i = 0; i < subject->nidentity; i++) {
1214 1215 1216 1217 1218 1219 1220 1221 1222 1223
            VIR_FREE(subject->identities[i].type);
            VIR_FREE(subject->identities[i].name);
        }
        VIR_FREE(subject->identities);
        VIR_FREE(subject);
    }

    return -1;
}

1224 1225 1226 1227
static int
qemuProcessHandleTrayChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            const char *devAlias,
1228 1229
                            int reason,
                            void *opaque)
1230
{
1231
    virQEMUDriverPtr driver = opaque;
1232
    virObjectEventPtr event = NULL;
1233
    virDomainDiskDefPtr disk;
1234
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1235

1236
    virObjectLock(vm);
1237 1238 1239 1240 1241 1242
    disk = qemuProcessFindDomainDiskByAlias(vm, devAlias);

    if (disk) {
        event = virDomainEventTrayChangeNewFromObj(vm,
                                                   devAlias,
                                                   reason);
1243 1244 1245 1246 1247 1248
        /* 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;

1249
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1250 1251 1252
            VIR_WARN("Unable to save status on vm %s after tray moved event",
                     vm->def->name);
        }
1253 1254
    }

1255
    virObjectUnlock(vm);
1256
    if (event)
1257
        qemuDomainEventQueue(driver, event);
1258
    virObjectUnref(cfg);
1259 1260 1261
    return 0;
}

O
Osier Yang 已提交
1262 1263
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1264 1265
                          virDomainObjPtr vm,
                          void *opaque)
O
Osier Yang 已提交
1266
{
1267
    virQEMUDriverPtr driver = opaque;
1268 1269
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1270
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1271

1272
    virObjectLock(vm);
O
Osier Yang 已提交
1273 1274
    event = virDomainEventPMWakeupNewFromObj(vm);

1275 1276 1277 1278 1279 1280 1281 1282 1283
    /* 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);
1284
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
1285 1286 1287
                                                  VIR_DOMAIN_EVENT_STARTED,
                                                  VIR_DOMAIN_EVENT_STARTED_WAKEUP);

1288
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1289 1290 1291 1292 1293
            VIR_WARN("Unable to save status on vm %s after wakeup event",
                     vm->def->name);
        }
    }

1294
    virObjectUnlock(vm);
1295 1296 1297 1298
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1299
    virObjectUnref(cfg);
O
Osier Yang 已提交
1300 1301
    return 0;
}
1302

O
Osier Yang 已提交
1303 1304
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1305 1306
                           virDomainObjPtr vm,
                           void *opaque)
O
Osier Yang 已提交
1307
{
1308
    virQEMUDriverPtr driver = opaque;
1309 1310
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1311
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1312

1313
    virObjectLock(vm);
O
Osier Yang 已提交
1314 1315
    event = virDomainEventPMSuspendNewFromObj(vm);

1316
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1317
        qemuDomainObjPrivatePtr priv = vm->privateData;
1318 1319 1320 1321 1322
        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
1323
        lifecycleEvent =
1324
            virDomainEventLifecycleNewFromObj(vm,
J
Jiri Denemark 已提交
1325 1326
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1327

1328
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1329 1330 1331
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1332 1333 1334

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1335 1336
    }

1337
    virObjectUnlock(vm);
O
Osier Yang 已提交
1338

1339 1340 1341 1342
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1343
    virObjectUnref(cfg);
O
Osier Yang 已提交
1344 1345 1346
    return 0;
}

1347 1348 1349
static int
qemuProcessHandleBalloonChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1350 1351
                               unsigned long long actual,
                               void *opaque)
1352
{
1353
    virQEMUDriverPtr driver = opaque;
1354
    virObjectEventPtr event = NULL;
1355
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1356

1357
    virObjectLock(vm);
1358 1359 1360 1361 1362 1363
    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;

1364
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
1365 1366
        VIR_WARN("unable to save domain status with balloon change");

1367
    virObjectUnlock(vm);
1368

1369
    if (event)
1370
        qemuDomainEventQueue(driver, event);
1371
    virObjectUnref(cfg);
1372 1373 1374
    return 0;
}

1375 1376
static int
qemuProcessHandlePMSuspendDisk(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1377 1378
                               virDomainObjPtr vm,
                               void *opaque)
1379
{
1380
    virQEMUDriverPtr driver = opaque;
1381 1382
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1383
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1384

1385
    virObjectLock(vm);
1386 1387 1388 1389 1390 1391 1392 1393 1394 1395
    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 =
1396
            virDomainEventLifecycleNewFromObj(vm,
1397 1398 1399
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_DISK);

1400
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1401 1402 1403 1404 1405 1406 1407 1408
            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);
    }

1409
    virObjectUnlock(vm);
1410

1411 1412 1413 1414
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1415 1416
    virObjectUnref(cfg);

1417 1418 1419
    return 0;
}

1420

1421 1422
static int
qemuProcessHandleGuestPanic(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1423 1424
                            virDomainObjPtr vm,
                            void *opaque)
1425
{
1426
    virQEMUDriverPtr driver = opaque;
1427 1428 1429
    struct qemuProcessEvent *processEvent;

    virObjectLock(vm);
1430
    if (VIR_ALLOC(processEvent) < 0)
1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445
        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);
    }

1446
 cleanup:
1447
    if (vm)
1448
        virObjectUnlock(vm);
1449 1450 1451 1452 1453

    return 0;
}


1454
int
1455 1456
qemuProcessHandleDeviceDeleted(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1457 1458
                               const char *devAlias,
                               void *opaque)
1459
{
1460
    virQEMUDriverPtr driver = opaque;
1461 1462
    struct qemuProcessEvent *processEvent = NULL;
    char *data;
1463 1464 1465 1466 1467 1468

    virObjectLock(vm);

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

1469 1470
    if (qemuDomainSignalDeviceRemoval(vm, devAlias))
        goto cleanup;
1471

1472 1473
    if (VIR_ALLOC(processEvent) < 0)
        goto error;
1474

1475 1476 1477 1478 1479
    processEvent->eventType = QEMU_PROCESS_EVENT_DEVICE_DELETED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
    processEvent->vm = vm;
1480

1481 1482 1483 1484 1485
    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }
1486

1487
 cleanup:
1488 1489
    virObjectUnlock(vm);
    return 0;
1490 1491 1492 1493 1494
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
1495 1496 1497
}


1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538
static int
qemuProcessHandleNicRxFilterChanged(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virDomainObjPtr vm,
                                    const char *devAlias,
                                    void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    struct qemuProcessEvent *processEvent = NULL;
    char *data;

    virObjectLock(vm);

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

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

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

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

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


1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581
static int
qemuProcessHandleSerialChanged(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               const char *devAlias,
                               bool connected,
                               void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    struct qemuProcessEvent *processEvent = NULL;
    char *data;

    virObjectLock(vm);

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

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

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

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

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


1582 1583
static qemuMonitorCallbacks monitorCallbacks = {
    .eofNotify = qemuProcessHandleMonitorEOF,
1584
    .errorNotify = qemuProcessHandleMonitorError,
1585
    .diskSecretLookup = qemuProcessFindVolumeQcowPassphrase,
1586
    .domainEvent = qemuProcessHandleEvent,
1587 1588
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
1589
    .domainResume = qemuProcessHandleResume,
1590 1591 1592 1593 1594
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1595
    .domainBlockJob = qemuProcessHandleBlockJob,
1596
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1597
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1598
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1599
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1600
    .domainPMSuspendDisk = qemuProcessHandlePMSuspendDisk,
1601
    .domainGuestPanic = qemuProcessHandleGuestPanic,
1602
    .domainDeviceDeleted = qemuProcessHandleDeviceDeleted,
1603
    .domainNicRxFilterChanged = qemuProcessHandleNicRxFilterChanged,
1604
    .domainSerialChange = qemuProcessHandleSerialChanged,
1605 1606 1607
};

static int
1608 1609
qemuConnectMonitor(virQEMUDriverPtr driver, virDomainObjPtr vm, int asyncJob,
                   int logfd)
1610 1611 1612
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
1613
    qemuMonitorPtr mon = NULL;
1614

1615
    if (virSecurityManagerSetDaemonSocketLabel(driver->securityManager,
1616
                                               vm->def) < 0) {
1617 1618 1619 1620 1621 1622
        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
1623
     * deleted unitl the monitor gets its own reference. */
1624
    virObjectRef(vm);
1625

1626
    ignore_value(virTimeMillisNow(&priv->monStart));
1627
    virObjectUnlock(vm);
1628 1629 1630 1631

    mon = qemuMonitorOpen(vm,
                          priv->monConfig,
                          priv->monJSON,
1632 1633
                          &monitorCallbacks,
                          driver);
1634

1635 1636 1637
    if (mon)
        ignore_value(qemuMonitorSetDomainLog(mon, logfd));

1638
    virObjectLock(vm);
M
Michal Privoznik 已提交
1639
    virObjectUnref(vm);
1640
    priv->monStart = 0;
1641

M
Michal Privoznik 已提交
1642
    if (!virDomainObjIsActive(vm)) {
1643
        qemuMonitorClose(mon);
1644
        mon = NULL;
1645 1646 1647
    }
    priv->mon = mon;

1648
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0) {
1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659
        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;
    }


1660 1661
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto error;
1662 1663
    ret = qemuMonitorSetCapabilities(priv->mon);
    if (ret == 0 &&
1664 1665
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON))
        ret = virQEMUCapsProbeQMP(priv->qemuCaps, priv->mon);
1666 1667
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
1668

1669
 error:
1670 1671 1672 1673

    return ret;
}

1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687

/**
 * 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)
1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700
{
    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';

1701 1702 1703
        if (bytes == 0)
            break;

1704 1705 1706
        /* Filter out debug messages from intermediate libvirt process */
        while ((eol = strchr(filter_next, '\n'))) {
            *eol = '\0';
1707 1708 1709
            if (virLogProbablyLogMessage(filter_next) ||
                (skipchar &&
                 STRPREFIX(filter_next, "char device redirected to"))) {
1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721
                memmove(filter_next, eol + 1, off - (eol - buf));
                off -= eol + 1 - filter_next;
            } else {
                filter_next = eol + 1;
                *eol = '\n';
            }
        }
    }

    return off;
}

1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739
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;
1740 1741
    int ret = -1;

1742 1743 1744
    buf[0] = '\0';

    while (retries) {
1745
        ssize_t func_ret;
1746
        bool isdead;
1747 1748 1749

        func_ret = func(vm, buf, fd);

1750
        isdead = kill(vm->pid, 0) == -1 && errno == ESRCH;
1751

1752
        got = qemuProcessReadLog(fd, buf, buflen, got, false);
1753
        if (got < 0) {
1754 1755 1756
            virReportSystemError(errno,
                                 _("Failure while reading %s log output"),
                                 what);
1757
            goto cleanup;
1758 1759 1760
        }

        if (got == buflen-1) {
1761 1762 1763
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Out of space while reading %s log output: %s"),
                           what, buf);
1764
            goto cleanup;
1765 1766 1767
        }

        if (isdead) {
1768 1769 1770
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Process exited while reading %s log output: %s"),
                           what, buf);
1771
            goto cleanup;
1772 1773
        }

1774 1775 1776 1777
        if (func_ret <= 0) {
            ret = func_ret;
            goto cleanup;
        }
1778 1779 1780 1781 1782

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

1783 1784 1785
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("Timed out while reading %s log output: %s"),
                   what, buf);
1786

1787
 cleanup:
1788
    return ret;
1789 1790 1791
}


1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849
/*
 * 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;

1850
 cleanup:
1851 1852 1853 1854 1855
    VIR_FORCE_CLOSE(logfd);
    return ret;
}


1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868
/*
 * 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)
{
1869 1870
    static const char needle[] = "char device redirected to";
    char *tmp, *dev;
1871 1872 1873

    VIR_FREE(*path);
    /* First look for our magic string */
1874
    if (!(tmp = strstr(haystack + *offset, needle)))
1875
        return 1;
1876
    tmp += sizeof(needle);
1877 1878 1879 1880 1881 1882 1883 1884
    dev = tmp;

    /*
     * And look for first whitespace character and nul terminate
     * to mark end of the pty path
     */
    while (*tmp) {
        if (c_isspace(*tmp)) {
1885
            if (VIR_STRNDUP(*path, dev, tmp - dev) < 0)
1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903
                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;
}

1904
static int
1905 1906 1907
qemuProcessLookupPTYs(virDomainDefPtr def,
                      virQEMUCapsPtr qemuCaps,
                      virDomainChrDefPtr *devices,
1908
                      int count,
1909
                      virHashTablePtr info)
1910
{
1911
    size_t i;
1912

1913
    for (i = 0; i < count; i++) {
1914
        virDomainChrDefPtr chr = devices[i];
1915 1916
        bool chardevfmt = virQEMUCapsSupportsChardev(def, qemuCaps, chr);

1917
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
C
Cole Robinson 已提交
1918
            char id[32];
1919
            qemuMonitorChardevInfoPtr entry;
1920

C
Cole Robinson 已提交
1921
            if (snprintf(id, sizeof(id), "%s%s",
1922
                         chardevfmt ? "char" : "",
1923 1924 1925 1926
                         chr->info.alias) >= sizeof(id)) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("failed to format device alias "
                                 "for PTY retrieval"));
1927
                return -1;
1928
            }
1929

1930 1931
            entry = virHashLookup(info, id);
            if (!entry || !entry->ptyPath) {
1932 1933 1934 1935
                if (chr->source.data.file.path == NULL) {
                    /* neither the log output nor 'info chardev' had a
                     * pty path for this chardev, report an error
                     */
1936 1937
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1938 1939 1940 1941 1942 1943 1944 1945 1946 1947
                    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);
1948
            if (VIR_STRDUP(chr->source.data.file.path, entry->ptyPath) < 0)
1949 1950 1951 1952 1953 1954 1955
                return -1;
        }
    }

    return 0;
}

1956 1957
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
1958
                                     virQEMUCapsPtr qemuCaps,
1959
                                     virHashTablePtr info)
1960
{
1961
    size_t i = 0;
C
Cole Robinson 已提交
1962

1963 1964
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->serials, vm->def->nserials,
1965
                              info) < 0)
1966 1967
        return -1;

1968 1969
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->parallels, vm->def->nparallels,
1970
                              info) < 0)
1971
        return -1;
1972

1973 1974
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->channels, vm->def->nchannels,
1975
                              info) < 0)
1976
        return -1;
1977 1978 1979 1980
    /* 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];
1981

1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992
        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;
        }
    }

1993 1994
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->consoles + i, vm->def->nconsoles - i,
1995
                              info) < 0)
1996
        return -1;
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

    return 0;
}

static int
qemuProcessFindCharDevicePTYs(virDomainObjPtr vm,
                              const char *output,
                              int fd ATTRIBUTE_UNUSED)
{
    size_t offset = 0;
2007 2008
    int ret;
    size_t i;
2009 2010 2011 2012 2013 2014

    /* 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 */
2015
    for (i = 0; i < vm->def->nserials; i++) {
2016 2017 2018 2019 2020 2021 2022 2023 2024
        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 */
2025
    for (i = 0; i < vm->def->nparallels; i++) {
2026 2027 2028 2029 2030 2031 2032 2033 2034
        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 */
2035
    for (i = 0; i < vm->def->nchannels; i++) {
2036 2037 2038 2039 2040 2041 2042 2043
        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;
        }
    }

2044
    for (i = 0; i < vm->def->nconsoles; i++) {
2045
        virDomainChrDefPtr chr = vm->def->consoles[i];
2046 2047 2048 2049 2050 2051 2052
        /* 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)
2053
                return ret;
2054 2055 2056 2057 2058 2059 2060
        } 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;
            }
2061 2062 2063
        }
    }

2064 2065 2066 2067
    return 0;
}


2068
static int
2069 2070 2071 2072
qemuProcessRefreshChannelVirtioState(virQEMUDriverPtr driver,
                                     virDomainObjPtr vm,
                                     virHashTablePtr info,
                                     int booted)
2073 2074
{
    size_t i;
2075
    int agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_CHANNEL;
2076
    qemuMonitorChardevInfoPtr entry;
2077
    virObjectEventPtr event = NULL;
2078 2079
    char id[32];

2080 2081 2082
    if (booted)
        agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_DOMAIN_STARTED;

2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098
    for (i = 0; i < vm->def->nchannels; i++) {
        virDomainChrDefPtr chr = vm->def->channels[i];
        if (chr->targetType == VIR_DOMAIN_CHR_CHANNEL_TARGET_TYPE_VIRTIO) {
            if (snprintf(id, sizeof(id), "char%s",
                         chr->info.alias) >= sizeof(id)) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("failed to format device alias "
                                 "for PTY retrieval"));
                return -1;
            }

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

2099 2100 2101 2102 2103 2104
            if (entry->state != VIR_DOMAIN_CHR_DEVICE_STATE_DEFAULT &&
                STREQ_NULLABLE(chr->target.name, "org.qemu.guest_agent.0") &&
                (event = virDomainEventAgentLifecycleNewFromObj(vm, entry->state,
                                                                agentReason)))
                qemuDomainEventQueue(driver, event);

2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122
            chr->state = entry->state;
        }
    }

    return 0;
}


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

    qemuDomainObjEnterMonitor(driver, vm);
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
2123 2124
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2125 2126 2127 2128

    if (ret < 0)
        goto cleanup;

2129
    ret = qemuProcessRefreshChannelVirtioState(driver, vm, info, false);
2130 2131 2132 2133 2134 2135 2136

 cleanup:
    virHashFree(info);
    return ret;
}


2137
static int
2138
qemuProcessWaitForMonitor(virQEMUDriverPtr driver,
C
Cole Robinson 已提交
2139
                          virDomainObjPtr vm,
2140
                          int asyncJob,
2141
                          virQEMUCapsPtr qemuCaps,
C
Cole Robinson 已提交
2142
                          off_t pos)
2143
{
2144
    char *buf = NULL;
2145
    size_t buf_size = 4096; /* Plenty of space to get startup greeting */
2146
    int logfd = -1;
2147
    int ret = -1;
2148
    virHashTablePtr info = NULL;
2149
    qemuDomainObjPrivatePtr priv;
2150

2151 2152 2153
    if (pos != -1 &&
        (logfd = qemuDomainOpenLog(driver, vm, pos)) < 0)
        return -1;
2154

2155
    if (logfd != -1 && !virQEMUCapsUsedQMP(qemuCaps)) {
2156
        if (VIR_ALLOC_N(buf, buf_size) < 0)
2157
            goto closelog;
2158

2159 2160 2161 2162 2163
        if (qemuProcessReadLogOutput(vm, logfd, buf, buf_size,
                                     qemuProcessFindCharDevicePTYs,
                                     "console", 30) < 0)
            goto closelog;
    }
2164 2165

    VIR_DEBUG("Connect monitor to %p '%s'", vm, vm->def->name);
2166
    if (qemuConnectMonitor(driver, vm, asyncJob, logfd) < 0)
2167 2168 2169 2170 2171 2172
        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. */
2173
    priv = vm->privateData;
2174 2175
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
2176 2177
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
    VIR_DEBUG("qemuMonitorGetChardevInfo returned %i", ret);
2178 2179 2180
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

2181 2182 2183 2184 2185
    if (ret == 0) {
        if ((ret = qemuProcessFindCharDevicePTYsMonitor(vm, qemuCaps,
                                                        info)) < 0)
            goto cleanup;

2186 2187
        if ((ret = qemuProcessRefreshChannelVirtioState(driver, vm, info,
                                                        true)) < 0)
2188 2189
            goto cleanup;
    }
2190

2191
 cleanup:
2192
    virHashFree(info);
2193

E
Eric Blake 已提交
2194
    if (pos != -1 && kill(vm->pid, 0) == -1 && errno == ESRCH) {
2195
        int len;
2196 2197
        /* VM is dead, any other error raised in the interim is probably
         * not as important as the qemu cmdline output */
2198
        if (virQEMUCapsUsedQMP(qemuCaps)) {
2199
            if (VIR_ALLOC_N(buf, buf_size) < 0)
2200 2201
                goto closelog;
        }
2202 2203

        len = strlen(buf);
2204
        /* best effort seek - we need to reset to the original position, so that
2205 2206
         * a possible read of the fd in the monitor code doesn't influence this
         * error delivery option */
2207
        ignore_value(lseek(logfd, pos, SEEK_SET));
2208
        qemuProcessReadLog(logfd, buf + len, buf_size - len - 1, 0, true);
2209 2210 2211
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("process exited while connecting to monitor: %s"),
                       buf);
2212 2213 2214
        ret = -1;
    }

2215
 closelog:
2216
    if (VIR_CLOSE(logfd) < 0) {
2217
        char ebuf[1024];
2218
        VIR_WARN("Unable to close logfile: %s",
2219
                 virStrerror(errno, ebuf, sizeof(ebuf)));
2220 2221
    }

2222 2223
    VIR_FREE(buf);

2224 2225 2226 2227
    return ret;
}

static int
2228
qemuProcessDetectVcpuPIDs(virQEMUDriverPtr driver,
2229
                          virDomainObjPtr vm, int asyncJob)
2230 2231 2232 2233 2234
{
    pid_t *cpupids = NULL;
    int ncpupids;
    qemuDomainObjPrivatePtr priv = vm->privateData;

2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268
    /*
     * Current QEMU *can* report info about host threads mapped
     * to vCPUs, but it is not in a manner we can correctly
     * deal with. The TCG CPU emulation does have a separate vCPU
     * thread, but it runs every vCPU in that same thread. So it
     * is impossible to setup different affinity per thread.
     *
     * What's more the 'query-cpus' command returns bizarre
     * data for the threads. It gives the TCG thread for the
     * vCPU 0, but for vCPUs 1-> N, it actually replies with
     * the main process thread ID.
     *
     * The result is that when we try to set affinity for
     * vCPU 1, it will actually change the affinity of the
     * emulator thread :-( When you try to set affinity for
     * vCPUs 2, 3.... it will fail if the affinity was
     * different from vCPU 1.
     *
     * We *could* allow vcpu pinning with TCG, if we made the
     * restriction that all vCPUs had the same mask. This would
     * at least let us separate emulator from vCPUs threads, as
     * we do for KVM. It would need some changes to our cgroups
     * CPU layout though, and error reporting for the config
     * restrictions.
     *
     * Just disable CPU pinning with TCG until someone wants
     * to try to do this hard work.
     */
    if (vm->def->virtType == VIR_DOMAIN_VIRT_QEMU) {
        priv->nvcpupids = 0;
        priv->vcpupids = NULL;
        return 0;
    }

2269 2270
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2271 2272 2273
    ncpupids = qemuMonitorGetCPUInfo(priv->mon, &cpupids);
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2274 2275 2276
    /* 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 */
2277
    if (ncpupids <= 0) {
2278 2279
        virResetLastError();

2280 2281
        priv->nvcpupids = 0;
        priv->vcpupids = NULL;
2282 2283 2284 2285
        return 0;
    }

    if (ncpupids != vm->def->vcpus) {
2286 2287 2288 2289
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of vCPU pids from QEMU monitor. "
                         "got %d, wanted %d"),
                       ncpupids, vm->def->vcpus);
2290 2291 2292 2293 2294 2295 2296 2297 2298
        VIR_FREE(cpupids);
        return -1;
    }

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

2299

2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310
static int
qemuProcessDetectIOThreadPIDs(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuMonitorIOThreadsInfoPtr *iothreads = NULL;
    int niothreads = 0;
    int ret = -1;
    size_t i;

2311 2312 2313
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_OBJECT_IOTHREAD))
        return 0;

2314 2315 2316 2317
    /* Get the list of IOThreads from qemu */
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
    niothreads = qemuMonitorGetIOThreads(priv->mon, &iothreads);
2318 2319
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
J
John Ferlan 已提交
2320
    if (niothreads < 0)
2321 2322
        goto cleanup;

J
John Ferlan 已提交
2323
    /* Nothing to do */
J
Ján Tomko 已提交
2324 2325 2326 2327
    if (niothreads == 0) {
        ret = 0;
        goto cleanup;
    }
J
John Ferlan 已提交
2328

2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354
    if (niothreads != vm->def->iothreads) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of IOThread pids from QEMU monitor. "
                         "got %d, wanted %d"),
                       niothreads, vm->def->iothreads);
        goto cleanup;
    }

    if (VIR_ALLOC_N(priv->iothreadpids, niothreads) < 0)
        goto cleanup;
    priv->niothreadpids = niothreads;

    for (i = 0; i < priv->niothreadpids; i++)
        priv->iothreadpids[i] = iothreads[i]->thread_id;

    ret = 0;

 cleanup:
    if (iothreads) {
        for (i = 0; i < niothreads; i++)
            qemuMonitorIOThreadsInfoFree(iothreads[i]);
        VIR_FREE(iothreads);
    }
    return ret;
}

2355 2356 2357
/* Helper to prepare cpumap for affinity setting, convert
 * NUMA nodeset into cpuset if @nodemask is not NULL, otherwise
 * just return a new allocated bitmap.
2358
 */
2359
virBitmapPtr
2360
qemuPrepareCpumap(virQEMUDriverPtr driver,
2361
                  virBitmapPtr nodemask)
2362
{
2363 2364
    size_t i;
    int hostcpus, maxcpu = QEMUD_CPUMASK_LEN;
2365
    virBitmapPtr cpumap = NULL;
2366
    virCapsPtr caps = NULL;
2367 2368 2369

    /* setaffinity fails if you set bits for CPUs which
     * aren't present, so we have to limit ourselves */
2370 2371 2372
    if ((hostcpus = nodeGetCPUCount()) < 0)
        return NULL;

2373 2374 2375
    if (maxcpu > hostcpus)
        maxcpu = hostcpus;

2376
    if (!(cpumap = virBitmapNew(maxcpu)))
2377
        return NULL;
2378

2379
    if (nodemask) {
2380 2381 2382 2383 2384 2385 2386
        if (!(caps = virQEMUDriverGetCapabilities(driver, false))) {
            virBitmapFree(cpumap);
            cpumap = NULL;
            goto cleanup;
        }

        for (i = 0; i < caps->host.nnumaCell; i++) {
2387
            size_t j;
2388
            int cur_ncpus = caps->host.numaCell[i]->ncpus;
2389
            bool result;
2390
            if (virBitmapGetBit(nodemask, caps->host.numaCell[i]->num, &result) < 0) {
2391
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
E
Eric Blake 已提交
2392
                               _("Failed to convert nodeset to cpuset"));
2393
                virBitmapFree(cpumap);
2394 2395
                cpumap = NULL;
                goto cleanup;
2396
            }
2397
            if (result) {
2398
                for (j = 0; j < cur_ncpus; j++)
2399
                    ignore_value(virBitmapSetBit(cpumap,
2400
                                                 caps->host.numaCell[i]->cpus[j].id));
2401
            }
O
Osier Yang 已提交
2402
        }
2403 2404
    }

2405
 cleanup:
2406
    virObjectUnref(caps);
2407 2408 2409 2410 2411 2412 2413
    return cpumap;
}

/*
 * To be run between fork/exec of QEMU only
 */
static int
2414
qemuProcessInitCpuAffinity(virQEMUDriverPtr driver,
2415
                           virDomainObjPtr vm)
2416 2417 2418 2419
{
    int ret = -1;
    virBitmapPtr cpumap = NULL;
    virBitmapPtr cpumapToSet = NULL;
2420
    qemuDomainObjPrivatePtr priv = vm->privateData;
2421

2422 2423 2424 2425 2426 2427
    if (!vm->pid) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Cannot setup CPU affinity until process is started"));
        return -1;
    }

2428
    if (!(cpumap = qemuPrepareCpumap(driver, priv->autoNodeset)))
2429 2430 2431 2432 2433
        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;
2434
    } else {
2435
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
2436
        if (vm->def->cpumask) {
H
Hu Tao 已提交
2437
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
2438
        } else {
2439
            cpumapToSet = cpumap;
O
Osier Yang 已提交
2440 2441 2442 2443
            /* 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 */
2444
            virBitmapSetAll(cpumap);
O
Osier Yang 已提交
2445
        }
2446 2447
    }

2448
    if (virProcessSetAffinity(vm->pid, cpumapToSet) < 0)
2449
        goto cleanup;
2450

2451 2452
    ret = 0;

2453
 cleanup:
2454
    virBitmapFree(cpumap);
2455
    return ret;
2456 2457
}

2458 2459 2460 2461 2462 2463
/* set link states to down on interfaces at qemu start */
static int
qemuProcessSetLinkStates(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2464
    size_t i;
2465 2466 2467 2468 2469 2470
    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);

2471
            if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NETDEV)) {
2472
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
2473
                               _("Setting of link state is not supported by this qemu"));
2474 2475 2476 2477 2478 2479 2480
                return -1;
            }

            ret = qemuMonitorSetLink(priv->mon,
                                     def->nets[i]->info.alias,
                                     VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (ret != 0) {
2481
                virReportError(VIR_ERR_OPERATION_FAILED,
2482 2483 2484 2485 2486 2487 2488 2489 2490 2491
                               _("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. */
2492
static int
2493
qemuProcessSetVcpuAffinities(virDomainObjPtr vm)
2494 2495 2496
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2497 2498
    virDomainVcpuPinDefPtr pininfo;
    int n;
E
Eric Blake 已提交
2499
    int ret = -1;
2500 2501
    VIR_DEBUG("Setting affinity on CPUs nvcpupin=%zu nvcpus=%d nvcpupids=%d",
              def->cputune.nvcpupin, def->vcpus, priv->nvcpupids);
2502 2503 2504 2505
    if (!def->cputune.nvcpupin)
        return 0;

    if (priv->vcpupids == NULL) {
2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517
        /* If any CPU has custom affinity that differs from the
         * VM default affinity, we must reject it
         */
        for (n = 0; n < def->vcpus; n++) {
            if (!virBitmapEqual(def->cpumask,
                                def->cputune.vcpupin[n]->cpumask)) {
                virReportError(VIR_ERR_OPERATION_INVALID,
                               "%s", _("cpu affinity is not supported"));
                return -1;
            }
        }
        return 0;
2518 2519
    }

2520 2521 2522 2523 2524 2525
    for (n = 0; n < def->vcpus; n++) {
        /* set affinity only for existing vcpus */
        if (!(pininfo = virDomainVcpuPinFindByVcpu(def->cputune.vcpupin,
                                                   def->cputune.nvcpupin,
                                                   n)))
            continue;
2526

2527 2528
        if (virProcessSetAffinity(priv->vcpupids[n],
                                  pininfo->cpumask) < 0) {
E
Eric Blake 已提交
2529
            goto cleanup;
2530 2531 2532
        }
    }

E
Eric Blake 已提交
2533
    ret = 0;
2534
 cleanup:
E
Eric Blake 已提交
2535
    return ret;
2536
}
2537

2538
/* Set CPU affinities for emulator threads. */
2539
static int
2540
qemuProcessSetEmulatorAffinity(virDomainObjPtr vm)
2541
{
2542
    virBitmapPtr cpumask;
2543 2544 2545
    virDomainDefPtr def = vm->def;
    int ret = -1;

O
Osier Yang 已提交
2546
    if (def->cputune.emulatorpin)
2547
        cpumask = def->cputune.emulatorpin->cpumask;
O
Osier Yang 已提交
2548
    else if (def->cpumask)
2549
        cpumask = def->cpumask;
O
Osier Yang 已提交
2550 2551
    else
        return 0;
2552

2553
    ret = virProcessSetAffinity(vm->pid, cpumask);
2554 2555 2556
    return ret;
}

2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579
/* Set CPU affinities for IOThreads threads. */
static int
qemuProcessSetIOThreadsAffinity(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
    virDomainVcpuPinDefPtr pininfo;
    size_t i;
    int ret = -1;

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

    if (priv->iothreadpids == NULL) {
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("IOThread affinity is not supported"));
        return -1;
    }

    for (i = 0; i < def->iothreads; i++) {
        /* set affinity only for existing vcpus */
        if (!(pininfo = virDomainVcpuPinFindByVcpu(def->cputune.iothreadspin,
                                                   def->cputune.niothreadspin,
2580
                                                   i + 1)))
2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591
            continue;

        if (virProcessSetAffinity(priv->iothreadpids[i], pininfo->cpumask) < 0)
            goto cleanup;
    }
    ret = 0;

 cleanup:
    return ret;
}

2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616
/* Set Scheduler parameters for vCPU or I/O threads. */
int
qemuProcessSetSchedParams(int id,
                          pid_t pid,
                          size_t nsp,
                          virDomainThreadSchedParamPtr sp)
{
    bool val = false;
    size_t i = 0;
    virDomainThreadSchedParamPtr s = NULL;

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

    if (!s)
        return 0;

2617
    return virProcessSetScheduler(pid, s->policy, s->priority);
2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642
}

static int
qemuProcessSetSchedulers(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    size_t i = 0;

    for (i = 0; i < priv->nvcpupids; i++) {
        if (qemuProcessSetSchedParams(i, priv->vcpupids[i],
                                      vm->def->cputune.nvcpusched,
                                      vm->def->cputune.vcpusched) < 0)
            return -1;
    }

    for (i = 0; i < priv->niothreadpids; i++) {
        if (qemuProcessSetSchedParams(i + 1, priv->iothreadpids[i],
                                      vm->def->cputune.niothreadsched,
                                      vm->def->cputune.iothreadsched) < 0)
            return -1;
    }

    return 0;
}

2643 2644
static int
qemuProcessInitPasswords(virConnectPtr conn,
2645
                         virQEMUDriverPtr driver,
2646 2647
                         virDomainObjPtr vm,
                         int asyncJob)
2648 2649 2650
{
    int ret = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
2651
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2652
    size_t i;
2653 2654
    char *alias = NULL;
    char *secret = NULL;
2655

2656
    for (i = 0; i < vm->def->ngraphics; ++i) {
2657 2658
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
2659 2660
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_VNC,
2661
                                                    &graphics->data.vnc.auth,
2662 2663
                                                    cfg->vncPassword,
                                                    asyncJob);
2664
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
2665 2666
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_SPICE,
2667
                                                    &graphics->data.spice.auth,
2668 2669
                                                    cfg->spicePassword,
                                                    asyncJob);
2670 2671
        }

2672 2673 2674
        if (ret < 0)
            goto cleanup;
    }
2675

2676
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
2677
        for (i = 0; i < vm->def->ndisks; i++) {
2678 2679
            size_t secretLen;

2680
            if (!vm->def->disks[i]->src->encryption ||
2681
                !virDomainDiskGetSource(vm->def->disks[i]))
2682 2683
                continue;

2684
            VIR_FREE(secret);
2685 2686 2687 2688 2689
            if (qemuProcessGetVolumeQcowPassphrase(conn,
                                                   vm->def->disks[i],
                                                   &secret, &secretLen) < 0)
                goto cleanup;

2690 2691 2692 2693
            VIR_FREE(alias);
            if (VIR_STRDUP(alias, vm->def->disks[i]->info.alias) < 0)
                goto cleanup;
            if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
2694
                goto cleanup;
E
Eric Blake 已提交
2695
            ret = qemuMonitorSetDrivePassphrase(priv->mon, alias, secret);
2696 2697
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                ret = -1;
2698 2699 2700 2701 2702
            if (ret < 0)
                goto cleanup;
        }
    }

2703
 cleanup:
2704 2705
    VIR_FREE(alias);
    VIR_FREE(secret);
2706
    virObjectUnref(cfg);
2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749
    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)
{
2750
    bool found = false;
2751
    size_t i;
2752 2753 2754

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

2755
    for (i = 0; i < naddrs; i++) {
2756 2757 2758
        VIR_DEBUG("Maybe %x:%x", addrs[i].vendor, addrs[i].product);
        if (addrs[i].vendor == vendor &&
            addrs[i].product == product) {
2759
            VIR_DEBUG("Match %zu", i);
2760
            found = true;
2761 2762 2763
            break;
        }
    }
2764
    if (!found)
2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964
        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;
2965
    size_t i;
2966 2967 2968 2969 2970

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

2971
    for (i = 0; i < vm->def->ndisks; i++) {
2972 2973 2974 2975 2976 2977
        if (qemuProcessGetPCIDiskVendorProduct(vm->def->disks[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->disks[i]->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2978 2979 2980
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for VirtIO disk %s"),
                           vm->def->disks[i]->dst);
2981 2982 2983 2984
            return -1;
        }
    }

2985
    for (i = 0; i < vm->def->nnets; i++) {
2986 2987 2988 2989 2990 2991
        if (qemuProcessGetPCINetVendorProduct(vm->def->nets[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->nets[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2992 2993 2994
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for %s NIC"),
                           vm->def->nets[i]->model);
2995 2996 2997 2998
            return -1;
        }
    }

2999
    for (i = 0; i < vm->def->ncontrollers; i++) {
3000 3001 3002 3003 3004 3005
        if (qemuProcessGetPCIControllerVendorProduct(vm->def->controllers[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->controllers[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
3006 3007 3008
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for controller %s"),
                           virDomainControllerTypeToString(vm->def->controllers[i]->type));
3009 3010 3011 3012
            return -1;
        }
    }

3013
    for (i = 0; i < vm->def->nvideos; i++) {
3014 3015 3016 3017 3018 3019
        if (qemuProcessGetPCIVideoVendorProduct(vm->def->videos[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->videos[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
3020 3021 3022
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for video adapter %s"),
                           virDomainVideoTypeToString(vm->def->videos[i]->type));
3023 3024 3025 3026
            return -1;
        }
    }

3027
    for (i = 0; i < vm->def->nsounds; i++) {
3028 3029 3030 3031 3032 3033
        if (qemuProcessGetPCISoundVendorProduct(vm->def->sounds[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->sounds[i]->info),
                                    vendor, product,
                                     addrs,  naddrs) < 0) {
3034 3035 3036
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for sound adapter %s"),
                           virDomainSoundModelTypeToString(vm->def->sounds[i]->model));
3037 3038 3039 3040 3041 3042 3043 3044 3045 3046
            return -1;
        }
    }


    if (vm->def->watchdog &&
        qemuProcessGetPCIWatchdogVendorProduct(vm->def->watchdog, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->watchdog->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
3047 3048 3049
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for watchdog %s"),
                           virDomainWatchdogModelTypeToString(vm->def->watchdog->model));
3050 3051 3052 3053 3054 3055 3056 3057 3058
            return -1;
        }
    }

    if (vm->def->memballoon &&
        qemuProcessGetPCIMemballoonVendorProduct(vm->def->memballoon, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->memballoon->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
3059 3060 3061
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for balloon %s"),
                           virDomainMemballoonModelTypeToString(vm->def->memballoon->model));
3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078
            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
3079
qemuProcessInitPCIAddresses(virQEMUDriverPtr driver,
3080 3081
                            virDomainObjPtr vm,
                            int asyncJob)
3082 3083 3084
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int naddrs;
3085
    int ret = -1;
3086 3087
    qemuMonitorPCIAddress *addrs = NULL;

3088 3089
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
3090 3091
    naddrs = qemuMonitorGetAllPCIAddresses(priv->mon,
                                           &addrs);
3092 3093
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
3094

3095 3096
    if (naddrs > 0)
        ret = qemuProcessDetectPCIAddresses(vm, addrs, naddrs);
3097

3098
 cleanup:
3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127
    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;
}


3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140
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;
}


3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205
/**
 * Loads and update video memory size for video devices according to QEMU
 * process as the QEMU will silently update the values that we pass to QEMU
 * through command line.  We need to load these updated values and store them
 * into the status XML.
 *
 * We will fail if for some reason the values cannot be loaded from QEMU because
 * its mandatory to get the correct video memory size to status XML to not break
 * migration.
 */
static int
qemuProcessUpdateVideoRamSize(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    int ret = -1;
    ssize_t i;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainVideoDefPtr video = NULL;
    virQEMUDriverConfigPtr cfg = NULL;

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

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

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

    }

3206 3207
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
3208 3209 3210 3211 3212 3213 3214 3215

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

    return ret;

 error:
3216
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
3217 3218 3219 3220
    return -1;
}


3221 3222 3223
struct qemuProcessHookData {
    virConnectPtr conn;
    virDomainObjPtr vm;
3224
    virQEMUDriverPtr driver;
3225
    virQEMUDriverConfigPtr cfg;
3226 3227 3228 3229 3230
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
3231
    qemuDomainObjPrivatePtr priv = h->vm->privateData;
3232
    int ret = -1;
3233
    int fd;
3234 3235 3236
    virBitmapPtr nodeset = NULL;
    virDomainNumatuneMemMode mode;

3237 3238 3239 3240 3241
    /* This method cannot use any mutexes, which are not
     * protected across fork()
     */

    virSecurityManagerPostFork(h->driver->securityManager);
3242 3243 3244 3245 3246

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

    VIR_DEBUG("Obtaining domain lock");
3247 3248 3249 3250 3251 3252 3253
    /*
     * 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.
     */
3254
    if (virSecurityManagerSetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
3255
        goto cleanup;
3256
    if (virDomainLockProcessStart(h->driver->lockManager,
3257
                                  h->cfg->uri,
3258
                                  h->vm,
J
Ján Tomko 已提交
3259
                                  /* QEMU is always paused initially */
3260 3261
                                  true,
                                  &fd) < 0)
3262
        goto cleanup;
3263
    if (virSecurityManagerClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
3264
        goto cleanup;
3265

3266 3267
    mode = virDomainNumatuneGetMode(h->vm->def->numa, -1);
    nodeset = virDomainNumatuneGetNodeset(h->vm->def->numa,
3268
                                          priv->autoNodeset, -1);
3269 3270

    if (virNumaSetupMemoryPolicy(mode, nodeset) < 0)
3271
        goto cleanup;
3272

3273 3274
    ret = 0;

3275
 cleanup:
3276
    virObjectUnref(h->cfg);
3277 3278
    VIR_DEBUG("Hook complete ret=%d", ret);
    return ret;
3279 3280 3281
}

int
3282
qemuProcessPrepareMonitorChr(virQEMUDriverConfigPtr cfg,
3283 3284 3285 3286 3287 3288
                             virDomainChrSourceDefPtr monConfig,
                             const char *vm)
{
    monConfig->type = VIR_DOMAIN_CHR_TYPE_UNIX;
    monConfig->data.nix.listen = true;

3289 3290 3291 3292
    if (virAsprintf(&monConfig->data.nix.path, "%s/%s.monitor",
                    cfg->libDir, vm) < 0)
        return -1;
    return 0;
3293 3294 3295
}


3296
/*
3297 3298
 * Precondition: vm must be locked, and a job must be active.
 * This method will call {Enter,Exit}Monitor
3299
 */
E
Eric Blake 已提交
3300
int
3301
qemuProcessStartCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
3302
                     virConnectPtr conn, virDomainRunningReason reason,
3303
                     qemuDomainAsyncJob asyncJob)
3304
{
3305
    int ret = -1;
3306
    qemuDomainObjPrivatePtr priv = vm->privateData;
3307
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3308

3309
    /* Bring up netdevs before starting CPUs */
3310
    if (qemuInterfaceStartDevices(vm->def) < 0)
3311 3312
       goto cleanup;

3313
    VIR_DEBUG("Using lock state '%s'", NULLSTR(priv->lockState));
3314
    if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
3315
                                   vm, priv->lockState) < 0) {
3316 3317 3318 3319
        /* 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
         */
3320
        goto cleanup;
3321 3322 3323
    }
    VIR_FREE(priv->lockState);

3324 3325
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto release;
J
Jiri Denemark 已提交
3326

3327
    ret = qemuMonitorStartCPUs(priv->mon, conn);
3328 3329
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3330 3331 3332 3333 3334

    if (ret < 0)
        goto release;

    virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
3335

3336
 cleanup:
3337
    virObjectUnref(cfg);
3338
    return ret;
3339 3340 3341 3342 3343 3344

 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;
3345 3346 3347
}


3348 3349
int qemuProcessStopCPUs(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
3350
                        virDomainPausedReason reason,
3351
                        qemuDomainAsyncJob asyncJob)
3352
{
3353
    int ret = -1;
3354 3355
    qemuDomainObjPrivatePtr priv = vm->privateData;

3356
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
3357

3358 3359
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
J
Jiri Denemark 已提交
3360

3361
    ret = qemuMonitorStopCPUs(priv->mon);
3362 3363
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3364 3365 3366 3367

    if (ret < 0)
        goto cleanup;

3368 3369 3370
    /* de-activate netdevs after stopping CPUs */
    ignore_value(qemuInterfaceStopDevices(vm->def));

3371 3372 3373
    if (priv->job.current)
        ignore_value(virTimeMillisNow(&priv->job.current->stopped));

3374 3375 3376 3377
    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
3378

3379
 cleanup:
3380 3381 3382 3383 3384
    return ret;
}



3385 3386 3387
static int
qemuProcessNotifyNets(virDomainDefPtr def)
{
3388
    size_t i;
3389

3390 3391
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
3392
        if (networkNotifyActualDevice(def, net) < 0)
3393 3394 3395 3396 3397
            return -1;
    }
    return 0;
}

3398
static int
3399
qemuProcessFiltersInstantiate(virDomainDefPtr def)
3400
{
3401
    size_t i;
3402

3403
    for (i = 0; i < def->nnets; i++) {
3404 3405
        virDomainNetDefPtr net = def->nets[i];
        if ((net->filter) && (net->ifname)) {
3406
            if (virDomainConfNWFilterInstantiate(def->uuid, net) < 0)
J
Ján Tomko 已提交
3407
                return 1;
3408 3409 3410
        }
    }

J
Ján Tomko 已提交
3411
    return 0;
3412 3413
}

3414
static int
3415
qemuProcessUpdateState(virQEMUDriverPtr driver, virDomainObjPtr vm)
3416 3417 3418
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
3419
    virDomainPausedReason reason;
3420 3421
    virDomainState newState = VIR_DOMAIN_NOSTATE;
    int newReason;
3422
    bool running;
3423
    char *msg = NULL;
3424 3425
    int ret;

3426
    qemuDomainObjEnterMonitor(driver, vm);
3427
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
3428 3429
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
3430

3431
    if (ret < 0)
3432 3433 3434 3435 3436
        return -1;

    state = virDomainObjGetState(vm, NULL);

    if (state == VIR_DOMAIN_PAUSED && running) {
3437 3438
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
3439
        ignore_value(VIR_STRDUP_QUIET(msg, "was unpaused"));
3440
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
3441 3442 3443
        if (reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN) {
            newState = VIR_DOMAIN_SHUTDOWN;
            newReason = VIR_DOMAIN_SHUTDOWN_UNKNOWN;
3444
            ignore_value(VIR_STRDUP_QUIET(msg, "shutdown"));
3445
        } else if (reason == VIR_DOMAIN_PAUSED_CRASHED) {
3446 3447 3448
            newState = VIR_DOMAIN_CRASHED;
            newReason = VIR_DOMAIN_CRASHED_PANICKED;
            ignore_value(VIR_STRDUP_QUIET(msg, "crashed"));
3449 3450 3451
        } else {
            newState = VIR_DOMAIN_PAUSED;
            newReason = reason;
S
Stefan Berger 已提交
3452 3453
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
3454
        }
3455
    } else if (state == VIR_DOMAIN_SHUTOFF && running) {
3456 3457
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_BOOTED;
3458
        ignore_value(VIR_STRDUP_QUIET(msg, "finished booting"));
3459 3460 3461 3462 3463 3464
    }

    if (newState != VIR_DOMAIN_NOSTATE) {
        VIR_DEBUG("Domain %s %s while its monitor was disconnected;"
                  " changing state to %s (%s)",
                  vm->def->name,
3465
                  NULLSTR(msg),
3466 3467 3468 3469
                  virDomainStateTypeToString(newState),
                  virDomainStateReasonToString(newState, newReason));
        VIR_FREE(msg);
        virDomainObjSetState(vm, newState, newReason);
3470 3471 3472 3473 3474
    }

    return 0;
}

3475
static int
3476
qemuProcessRecoverMigration(virQEMUDriverPtr driver,
3477 3478
                            virDomainObjPtr vm,
                            virConnectPtr conn,
3479 3480
                            qemuDomainAsyncJob job,
                            qemuMigrationJobPhase phase,
3481 3482 3483
                            virDomainState state,
                            int reason)
{
3484 3485
    qemuDomainObjPrivatePtr priv = vm->privateData;

3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508
    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,
3509 3510
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540
                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);
3541
            qemuDomainObjEnterMonitor(driver, vm);
3542
            ignore_value(qemuMonitorMigrateCancel(priv->mon));
3543 3544
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                return -1;
3545 3546 3547 3548 3549 3550
            /* 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,
3551 3552
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571
                    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,
3572 3573
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587
                    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;
}

3588
static int
3589
qemuProcessRecoverJob(virQEMUDriverPtr driver,
3590 3591 3592 3593
                      virDomainObjPtr vm,
                      virConnectPtr conn,
                      const struct qemuDomainJobObj *job)
{
3594
    qemuDomainObjPrivatePtr priv = vm->privateData;
3595 3596 3597 3598 3599 3600 3601 3602
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
    case QEMU_ASYNC_JOB_MIGRATION_IN:
3603 3604 3605
        if (qemuProcessRecoverMigration(driver, vm, conn, job->asyncJob,
                                        job->phase, state, reason) < 0)
            return -1;
3606 3607 3608 3609
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
3610
    case QEMU_ASYNC_JOB_SNAPSHOT:
3611
        qemuDomainObjEnterMonitor(driver, vm);
3612
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
3613 3614
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
3615
        /* resume the domain but only if it was paused as a result of
3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631
         * 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);
3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643
            }
        }
        break;

    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        break;
    }

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

3644 3645 3646 3647
    /* 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.
     */
3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667
    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;

3668
    case QEMU_JOB_MIGRATION_OP:
3669
    case QEMU_JOB_ABORT:
3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680
    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;
}

3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695
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;
3696
    if (qemuDomainUpdateDeviceList(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
3697 3698 3699 3700 3701
        goto cleanup;

    if ((tmp = old)) {
        while (*tmp) {
            if (!virStringArrayHasString(priv->qemuDevices, *tmp) &&
3702 3703 3704 3705
                virDomainDefFindDevice(vm->def, *tmp, &dev, false) == 0 &&
                qemuDomainRemoveDevice(driver, vm, &dev) < 0) {
                goto cleanup;
            }
3706 3707 3708 3709 3710
            tmp++;
        }
    }
    ret = 0;

3711
 cleanup:
3712 3713 3714 3715
    virStringFreeList(old);
    return ret;
}

3716 3717
struct qemuProcessReconnectData {
    virConnectPtr conn;
3718
    virQEMUDriverPtr driver;
3719
    virDomainObjPtr obj;
3720 3721 3722 3723
};
/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
S
Stefan Berger 已提交
3724 3725 3726 3727
 *
 * 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.
3728
 *
3729
 * This function also inherits a locked and ref'd domain object.
3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741
 *
 * This function needs to:
 * 1. Enter job
 * 1. just before monitor reconnect do lightweight MonitorEnter
 *    (increase VM refcount and unlock VM)
 * 2. reconnect to monitor
 * 3. do lightweight MonitorExit (lock VM)
 * 4. continue reconnect process
 * 5. EndJob
 *
 * We can't do normal MonitorEnter & MonitorExit because these two lock the
 * monitor lock, which does not exists in this early phase.
3742 3743
 */
static void
3744
qemuProcessReconnect(void *opaque)
3745 3746
{
    struct qemuProcessReconnectData *data = opaque;
3747
    virQEMUDriverPtr driver = data->driver;
3748
    virDomainObjPtr obj = data->obj;
3749 3750
    qemuDomainObjPrivatePtr priv;
    virConnectPtr conn = data->conn;
3751
    struct qemuDomainJobObj oldjob;
3752 3753
    int state;
    int reason;
3754
    virQEMUDriverConfigPtr cfg;
3755
    size_t i;
3756
    int ret;
3757

3758 3759
    VIR_FREE(data);

3760
    qemuDomainObjRestoreJob(obj, &oldjob);
3761

3762 3763 3764
    cfg = virQEMUDriverGetConfig(driver);
    priv = obj->privateData;

3765 3766 3767 3768 3769
    /* XXX If we ever gonna change pid file pattern, come up with
     * some intelligence here to deal with old paths. */
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, obj->def->name)))
        goto killvm;

3770 3771 3772 3773
    if (qemuDomainObjBeginJob(driver, obj, QEMU_JOB_MODIFY) < 0)
        goto killvm;

    virNWFilterReadLockFilterUpdates();
3774 3775 3776 3777

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

    /* XXX check PID liveliness & EXE path */
3778
    if (qemuConnectMonitor(driver, obj, QEMU_ASYNC_JOB_NONE, -1) < 0)
3779 3780
        goto error;

D
Daniel P. Berrange 已提交
3781
    /* Failure to connect to agent shouldn't be fatal */
3782 3783 3784 3785
    if ((ret = qemuConnectAgent(driver, obj)) < 0) {
        if (ret == -2)
            goto error;

D
Daniel P. Berrange 已提交
3786 3787 3788 3789 3790 3791
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 obj->def->name);
        virResetLastError();
        priv->agentError = true;
    }

3792
    if (qemuUpdateActivePCIHostdevs(driver, obj->def) < 0)
3793 3794
        goto error;

3795
    if (qemuUpdateActiveUSBHostdevs(driver, obj->def) < 0)
3796 3797
        goto error;

3798
    if (qemuUpdateActiveSCSIHostdevs(driver, obj->def) < 0)
3799 3800
        goto error;

3801
    if (qemuConnectCgroup(driver, obj) < 0)
3802 3803
        goto error;

3804
    /* XXX: Need to change as long as lock is introduced for
3805
     * qemu_driver->sharedDevices.
3806 3807
     */
    for (i = 0; i < obj->def->ndisks; i++) {
3808 3809
        virDomainDeviceDef dev;

3810
        if (virStorageTranslateDiskSourcePool(conn, obj->def->disks[i]) < 0)
3811
            goto error;
3812

3813 3814 3815 3816 3817 3818
        /* XXX we should be able to restore all data from XML in the future.
         * This should be the only place that calls qemuDomainDetermineDiskChain
         * with @report_broken == false to guarantee best-effort domain
         * reconnect */
        if (qemuDomainDetermineDiskChain(driver, obj, obj->def->disks[i],
                                         true, false) < 0)
3819 3820
            goto error;

3821 3822 3823
        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = obj->def->disks[i];
        if (qemuAddSharedDevice(driver, &dev, obj->def->name) < 0)
3824 3825 3826
            goto error;
    }

3827 3828 3829
    if (qemuProcessUpdateState(driver, obj) < 0)
        goto error;

3830 3831
    state = virDomainObjGetState(obj, &reason);
    if (state == VIR_DOMAIN_SHUTOFF) {
3832 3833 3834 3835 3836
        VIR_DEBUG("Domain '%s' wasn't fully started yet, killing it",
                  obj->def->name);
        goto error;
    }

3837 3838 3839
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
3840 3841
    if (!priv->qemuCaps &&
        !(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
3842 3843
                                                      obj->def->emulator,
                                                      obj->def->os.machine)))
3844 3845
        goto error;

3846
    /* In case the domain shutdown while we were not running,
3847
     * we need to finish the shutdown process. And we need to do it after
3848
     * we have virQEMUCaps filled in.
3849
     */
3850 3851 3852 3853 3854 3855
    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);
3856 3857
        qemuDomainObjEndJob(driver, obj);
        goto cleanup;
3858 3859
    }

3860 3861
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE))
        if ((qemuDomainAssignAddresses(obj->def, priv->qemuCaps, obj)) < 0)
3862
            goto error;
3863

3864
    if (virSecurityManagerReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
3865 3866
        goto error;

3867 3868 3869
    if (qemuProcessNotifyNets(obj->def) < 0)
        goto error;

3870
    if (qemuProcessFiltersInstantiate(obj->def))
3871 3872
        goto error;

3873
    if (qemuDomainCheckEjectableMedia(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
3874 3875
        goto error;

3876 3877 3878
    if (qemuProcessReconnectRefreshChannelVirtioState(driver, obj) < 0)
        goto error;

3879 3880 3881
    if (qemuProcessRecoverJob(driver, obj, conn, &oldjob) < 0)
        goto error;

3882 3883 3884
    if (qemuProcessUpdateDevices(driver, obj) < 0)
        goto error;

3885
    /* update domain state XML with possibly updated state in virDomainObj */
3886
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, obj) < 0)
3887 3888
        goto error;

3889 3890
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3891
        char *xml = qemuDomainDefFormatXML(driver, obj->def, 0);
3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905
        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;
    }

3906
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
3907 3908
        driver->inhibitCallback(true, driver->inhibitOpaque);

3909 3910
    qemuDomainObjEndJob(driver, obj);
    goto cleanup;
3911

3912
 error:
3913
    qemuDomainObjEndJob(driver, obj);
3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928
 killvm:
    if (virDomainObjIsActive(obj)) {
        /* We can't get the monitor back, so must kill the VM
         * to remove danger of it ending up running twice if
         * user tries to start it again later
         */
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NO_SHUTDOWN)) {
            /* If we couldn't get the monitor and qemu supports
             * no-shutdown, we can safely say that the domain
             * crashed ... */
            state = VIR_DOMAIN_SHUTOFF_CRASHED;
        } else {
            /* ... but if it doesn't we can't say what the state
             * really is and FAILED means "failed to start" */
            state = VIR_DOMAIN_SHUTOFF_UNKNOWN;
3929
        }
3930
        qemuProcessStop(driver, obj, state, 0);
3931
    }
3932

3933 3934
    if (!obj->persistent)
        qemuDomainRemoveInactive(driver, obj);
3935

3936 3937
 cleanup:
    qemuDomObjEndAPI(&obj);
3938
    virObjectUnref(conn);
3939
    virObjectUnref(cfg);
3940
    virNWFilterUnlockFilterUpdates();
3941 3942
}

3943 3944
static int
qemuProcessReconnectHelper(virDomainObjPtr obj,
3945 3946 3947 3948 3949 3950
                           void *opaque)
{
    virThread thread;
    struct qemuProcessReconnectData *src = opaque;
    struct qemuProcessReconnectData *data;

3951
    /* If the VM was inactive, we don't need to reconnect */
3952 3953 3954
    if (!obj->pid)
        return 0;

3955
    if (VIR_ALLOC(data) < 0)
3956
        return -1;
3957 3958

    memcpy(data, src, sizeof(*data));
3959
    data->obj = obj;
3960

3961 3962
    /* this lock and reference will be eventually transferred to the thread
     * that handles the reconnect */
3963
    virObjectLock(obj);
3964
    virObjectRef(obj);
3965

3966 3967 3968
    /* Since we close the connection later on, we have to make sure that the
     * threads we start see a valid connection throughout their lifetime. We
     * simply increase the reference counter here.
S
Stefan Berger 已提交
3969
     */
3970
    virObjectRef(data->conn);
S
Stefan Berger 已提交
3971

3972
    if (virThreadCreate(&thread, false, qemuProcessReconnect, data) < 0) {
3973 3974 3975
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not create thread. QEMU initialization "
                         "might be incomplete"));
3976 3977 3978 3979
       /* We can't spawn a thread and thus connect to monitor. Kill qemu. */
        qemuProcessStop(src->driver, obj, VIR_DOMAIN_SHUTOFF_FAILED, 0);
        if (!obj->persistent)
            qemuDomainRemoveInactive(src->driver, obj);
3980

3981
        qemuDomObjEndAPI(&obj);
3982 3983 3984 3985
        virObjectUnref(data->conn);
        VIR_FREE(data);
        return -1;
    }
3986

3987
    return 0;
3988 3989 3990 3991 3992 3993 3994 3995 3996
}

/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
3997
qemuProcessReconnectAll(virConnectPtr conn, virQEMUDriverPtr driver)
3998
{
3999
    struct qemuProcessReconnectData data = {.conn = conn, .driver = driver};
4000
    virDomainObjListForEach(driver->domains, qemuProcessReconnectHelper, &data);
4001 4002
}

4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025
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;
}
4026

4027
int
4028 4029
qemuProcessSPICEAllocatePorts(virQEMUDriverPtr driver,
                              virQEMUDriverConfigPtr cfg,
4030 4031
                              virDomainGraphicsDefPtr graphics,
                              bool allocate)
4032 4033 4034
{
    unsigned short port = 0;
    unsigned short tlsPort;
4035
    size_t i;
4036 4037 4038 4039 4040 4041 4042
    int defaultMode = graphics->data.spice.defaultMode;

    bool needTLSPort = false;
    bool needPort = false;

    if (graphics->data.spice.autoport) {
        /* check if tlsPort or port need allocation */
4043
        for (i = 0; i < VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_LAST; i++) {
4044 4045 4046 4047 4048 4049 4050 4051 4052 4053
            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:
4054
                /* default mode will be used */
4055 4056 4057
                break;
            }
        }
4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072
        switch (defaultMode) {
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
            needTLSPort = true;
            break;

        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
            needPort = true;
            break;

        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
            if (cfg->spiceTLS)
                needTLSPort = true;
            needPort = true;
            break;
        }
4073 4074
    }

4075 4076 4077 4078 4079 4080 4081 4082 4083 4084
    if (!allocate) {
        if (needPort || graphics->data.spice.port == -1)
            graphics->data.spice.port = 5901;

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

        return 0;
    }

4085
    if (needPort || graphics->data.spice.port == -1) {
4086
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
4087
            goto error;
4088 4089 4090 4091

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

4092 4093
    if (needTLSPort || graphics->data.spice.tlsPort == -1) {
        if (!cfg->spiceTLS) {
4094 4095 4096 4097
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Auto allocation of spice TLS port requested "
                             "but spice TLS is disabled in qemu.conf"));
            goto error;
4098
        }
4099 4100 4101 4102 4103

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

        graphics->data.spice.tlsPort = tlsPort;
4104 4105
    }

4106
    return 0;
4107

4108
 error:
4109
    virPortAllocatorRelease(driver->remotePorts, port);
4110
    return -1;
4111 4112 4113
}


4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131
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;
}

4132 4133

static bool
4134 4135 4136
qemuProcessVerifyGuestCPU(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          int asyncJob)
4137 4138 4139 4140 4141
{
    virDomainDefPtr def = vm->def;
    virArch arch = def->os.arch;
    virCPUDataPtr guestcpu = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
4142
    int rc;
4143
    bool ret = false;
J
Ján Tomko 已提交
4144
    size_t i;
4145

4146 4147 4148 4149 4150
    /* no features are passed to QEMU with -cpu host
     * so it makes no sense to verify them */
    if (def->cpu && def->cpu->mode == VIR_CPU_MODE_HOST_PASSTHROUGH)
        return true;

4151 4152 4153
    switch (arch) {
    case VIR_ARCH_I686:
    case VIR_ARCH_X86_64:
4154 4155
        if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
            return false;
4156
        rc = qemuMonitorGetGuestCPU(priv->mon, arch, &guestcpu);
4157 4158
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return false;
4159

4160 4161 4162 4163
        if (rc < 0) {
            if (rc == -2)
                break;

4164
            goto cleanup;
4165
        }
4166

J
Ján Tomko 已提交
4167
        if (def->features[VIR_DOMAIN_FEATURE_PVSPINLOCK] == VIR_TRISTATE_SWITCH_ON) {
4168 4169 4170 4171 4172 4173
            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
4174

4175
        for (i = 0; def->cpu && i < def->cpu->nfeatures; i++) {
J
Ján Tomko 已提交
4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187
            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;
            }
        }
4188 4189 4190 4191 4192 4193 4194 4195
        break;

    default:
        break;
    }

    ret = true;

4196
 cleanup:
4197 4198 4199 4200 4201
    cpuDataFree(guestcpu);
    return ret;
}


4202 4203
static int
qemuPrepareNVRAM(virQEMUDriverConfigPtr cfg,
4204
                 virDomainObjPtr vm,
4205 4206 4207 4208 4209
                 bool migrated)
{
    int ret = -1;
    int srcFD = -1;
    int dstFD = -1;
4210
    virDomainLoaderDefPtr loader = vm->def->os.loader;
4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236
    bool generated = false;
    bool created = false;

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

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

    /* Autogenerate nvram path if needed.*/
    if (!loader->nvram) {
        if (virAsprintf(&loader->nvram,
                        "%s/lib/libvirt/qemu/nvram/%s_VARS.fd",
4237
                        LOCALSTATEDIR, vm->def->name) < 0)
4238 4239 4240
            goto cleanup;

        generated = true;
4241

4242 4243
        if (vm->persistent &&
            virDomainSaveConfig(cfg->configDir, vm->def) < 0)
4244
            goto cleanup;
4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334
    }

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

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

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

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

        do {
            char buf[1024];

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

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

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

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

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


4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373
static void
qemuLogOperation(virDomainObjPtr vm,
                 const char *msg,
                 int logfd,
                 virCommandPtr cmd)
{
    char *timestamp;
    char *logline;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int qemuVersion = virQEMUCapsGetVersion(priv->qemuCaps);
    const char *package = virQEMUCapsGetPackage(priv->qemuCaps);
    char ebuf[1024];

    if ((timestamp = virTimeStringNow()) == NULL)
        goto error;

    if (virAsprintf(&logline, "%s: %s %s, qemu version: %d.%d.%d%s\n",
                    timestamp, msg, VIR_LOG_VERSION_STRING,
                    (qemuVersion / 1000000) % 1000, (qemuVersion / 1000) % 1000, qemuVersion % 1000,
                    package ? package : "") < 0)
        goto error;

    if (safewrite(logfd, logline, strlen(logline)) < 0)
        goto error;

    if (cmd)
        virCommandWriteArgLog(cmd, logfd);

 cleanup:
    VIR_FREE(timestamp);
    VIR_FREE(logline);
    return;

 error:
    VIR_WARN("Unable to write banner to logfile: %s",
             virStrerror(errno, ebuf, sizeof(ebuf)));
    goto cleanup;
}

4374
int qemuProcessStart(virConnectPtr conn,
4375
                     virQEMUDriverPtr driver,
4376
                     virDomainObjPtr vm,
4377
                     int asyncJob,
4378 4379 4380
                     const char *migrateFrom,
                     int stdin_fd,
                     const char *stdin_path,
4381
                     virDomainSnapshotObjPtr snapshot,
4382
                     virNetDevVPortProfileOp vmop,
4383
                     unsigned int flags)
4384 4385 4386 4387 4388 4389 4390 4391
{
    int ret;
    off_t pos = -1;
    char ebuf[1024];
    int logfile = -1;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCommandPtr cmd = NULL;
    struct qemuProcessHookData hookData;
E
Eric Blake 已提交
4392
    unsigned long cur_balloon;
4393
    unsigned int period = 0;
4394
    size_t i;
4395
    bool rawio_set = false;
4396
    char *nodeset = NULL;
4397
    virBitmapPtr nodemask = NULL;
4398
    unsigned int stop_flags;
4399
    virQEMUDriverConfigPtr cfg;
4400
    virCapsPtr caps = NULL;
4401
    unsigned int hostdev_flags = 0;
4402 4403
    size_t nnicindexes = 0;
    int *nicindexes = NULL;
4404

4405 4406 4407 4408
    VIR_DEBUG("vm=%p name=%s id=%d pid=%llu",
              vm, vm->def->name, vm->def->id,
              (unsigned long long)vm->pid);

4409 4410 4411 4412
    /* Okay, these are just internal flags,
     * but doesn't hurt to check */
    virCheckFlags(VIR_QEMU_PROCESS_START_COLD |
                  VIR_QEMU_PROCESS_START_PAUSED |
4413
                  VIR_QEMU_PROCESS_START_AUTODESTROY, -1);
4414

4415 4416
    cfg = virQEMUDriverGetConfig(driver);

4417 4418 4419 4420 4421
    /* 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;
4422 4423 4424 4425
    /* 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;
4426

4427 4428 4429
    hookData.conn = conn;
    hookData.vm = vm;
    hookData.driver = driver;
4430 4431
    /* We don't increase cfg's reference counter here. */
    hookData.cfg = cfg;
4432

4433
    VIR_DEBUG("Beginning VM startup process");
4434 4435

    if (virDomainObjIsActive(vm)) {
4436 4437
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
4438
        virObjectUnref(cfg);
4439 4440 4441
        return -1;
    }

4442 4443 4444
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

4445 4446 4447 4448
    /* Some things, paths, ... are generated here and we want them to persist.
     * Fill them in prior to setting the domain def as transient. */
    VIR_DEBUG("Generating paths");

4449
    if (qemuPrepareNVRAM(cfg, vm, migrateFrom) < 0)
4450 4451
        goto cleanup;

4452 4453 4454 4455
    /* 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
     */
4456
    VIR_DEBUG("Setting current domain def as transient");
4457
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm, true) < 0)
4458 4459
        goto cleanup;

4460
    vm->def->id = qemuDriverAllocateID(driver);
4461
    qemuDomainSetFakeReboot(driver, vm, false);
4462
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, VIR_DOMAIN_SHUTOFF_UNKNOWN);
4463

4464
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
4465 4466
        driver->inhibitCallback(true, driver->inhibitOpaque);

4467
    /* Run an early hook to set-up missing devices */
4468
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4469
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4470 4471 4472
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4473 4474
                              VIR_HOOK_QEMU_OP_PREPARE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
4475 4476 4477 4478 4479 4480 4481 4482 4483
        VIR_FREE(xml);

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

4484 4485 4486
    VIR_DEBUG("Determining emulator version");
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
4487 4488
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
4489 4490
        goto cleanup;

4491 4492 4493 4494 4495 4496 4497 4498
    /* 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;

4499
    /* Must be run before security labelling */
4500
    VIR_DEBUG("Preparing host devices");
4501 4502
    if (!cfg->relaxedACS)
        hostdev_flags |= VIR_HOSTDEV_STRICT_ACS_CHECK;
4503 4504
    if (!migrateFrom)
        hostdev_flags |= VIR_HOSTDEV_COLD_BOOT;
4505
    if (qemuPrepareHostDevices(driver, vm->def, priv->qemuCaps,
4506
                               hostdev_flags) < 0)
4507 4508
        goto cleanup;

4509
    VIR_DEBUG("Preparing chr devices");
4510 4511 4512 4513 4514 4515
    if (virDomainChrDefForeach(vm->def,
                               true,
                               qemuProcessPrepareChardevDevice,
                               NULL) < 0)
        goto cleanup;

4516 4517 4518 4519
    VIR_DEBUG("Checking domain and device security labels");
    if (virSecurityManagerCheckAllLabel(driver->securityManager, vm->def) < 0)
        goto cleanup;

4520 4521
    /* If you are using a SecurityDriver with dynamic labelling,
       then generate a security label for isolation */
4522
    VIR_DEBUG("Generating domain security label (if required)");
4523
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0) {
4524
        virDomainAuditSecurityLabel(vm, false);
4525 4526
        goto cleanup;
    }
4527
    virDomainAuditSecurityLabel(vm, true);
4528

4529
    if (vm->def->mem.nhugepages) {
4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543
        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);
4544 4545 4546
        }
    }

4547 4548
    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
4549
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
4550
    qemuRemoveCgroup(driver, vm);
4551

4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586
    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;
            }
        }
    }

4587
    for (i = 0; i < vm->def->ngraphics; ++i) {
4588
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
4589 4590
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (qemuProcessVNCAllocatePorts(driver, graphics) < 0)
4591
                goto cleanup;
4592
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
4593
            if (qemuProcessSPICEAllocatePorts(driver, cfg, graphics, true) < 0)
4594
                goto cleanup;
4595
        }
4596

4597 4598
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ||
            graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
4599
            if (graphics->nListens == 0) {
4600
                if (VIR_EXPAND_N(graphics->listens, graphics->nListens, 1) < 0)
4601 4602
                    goto cleanup;
                graphics->listens[0].type = VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_ADDRESS;
4603 4604 4605
                if (VIR_STRDUP(graphics->listens[0].address,
                               graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ?
                               cfg->vncListen : cfg->spiceListen) < 0) {
4606 4607 4608
                    VIR_SHRINK_N(graphics->listens, graphics->nListens, 1);
                    goto cleanup;
                }
4609
                graphics->listens[0].fromConfig = true;
4610 4611 4612 4613 4614
            } else if (graphics->nListens > 1) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("QEMU does not support multiple listen "
                                 "addresses for one graphics device."));
                goto cleanup;
4615 4616
            }
        }
4617 4618
    }

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

4626
    VIR_DEBUG("Creating domain log file");
4627
    if ((logfile = qemuDomainCreateLog(driver, vm, false)) < 0)
4628 4629
        goto cleanup;

4630 4631
    if (vm->def->virtType == VIR_DOMAIN_VIRT_KVM) {
        VIR_DEBUG("Checking for KVM availability");
4632
        if (!virFileExists("/dev/kvm")) {
4633 4634 4635 4636
            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."));
4637 4638 4639 4640
            goto cleanup;
        }
    }

4641 4642 4643
    if (!qemuValidateCpuMax(vm->def, priv->qemuCaps))
        goto cleanup;

4644
    if (qemuAssignDeviceAliases(vm->def, priv->qemuCaps) < 0)
4645 4646
        goto cleanup;

4647 4648 4649
    /* Get the advisory nodeset from numad if 'placement' of
     * either <vcpu> or <numatune> is 'auto'.
     */
4650
    if (virDomainDefNeedsPlacementAdvice(vm->def)) {
4651
        nodeset = virNumaGetAutoPlacementAdvice(vm->def->vcpus,
4652
                                                vm->def->mem.max_balloon);
4653 4654 4655 4656 4657
        if (!nodeset)
            goto cleanup;

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

4658
        if (virBitmapParse(nodeset, 0, &nodemask, VIR_DOMAIN_CPUMASK_LEN) < 0)
4659 4660
            goto cleanup;
    }
4661
    priv->autoNodeset = nodemask;
4662

4663 4664 4665 4666
    /* "volume" type disk's source must be translated before
     * cgroup and security setting.
     */
    for (i = 0; i < vm->def->ndisks; i++) {
4667
        if (virStorageTranslateDiskSourcePool(conn, vm->def->disks[i]) < 0)
4668 4669 4670
            goto cleanup;
    }

4671 4672 4673 4674
    if (qemuDomainCheckDiskPresence(driver, vm,
                                    flags & VIR_QEMU_PROCESS_START_COLD) < 0)
        goto cleanup;

4675 4676 4677 4678 4679 4680 4681
    if (vm->def->mem.min_guarantee) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Parameter 'min_guarantee' "
                         "not supported by QEMU."));
        goto cleanup;
    }

4682
    if (VIR_ALLOC(priv->monConfig) < 0)
4683 4684
        goto cleanup;

4685
    VIR_DEBUG("Preparing monitor state");
4686
    if (qemuProcessPrepareMonitorChr(cfg, priv->monConfig, vm->def->name) < 0)
4687 4688
        goto cleanup;

E
Eric Blake 已提交
4689
    priv->monJSON = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON);
4690 4691
    priv->monError = false;
    priv->monStart = 0;
4692 4693
    priv->gotShutdown = false;

4694
    VIR_FREE(priv->pidfile);
4695
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, vm->def->name))) {
4696 4697
        virReportSystemError(errno,
                             "%s", _("Failed to build pidfile path."));
4698 4699 4700
        goto cleanup;
    }

4701 4702
    if (unlink(priv->pidfile) < 0 &&
        errno != ENOENT) {
4703
        virReportSystemError(errno,
4704 4705
                             _("Cannot remove stale PID file %s"),
                             priv->pidfile);
4706 4707 4708 4709 4710 4711 4712
        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
4713
     * we also need to populate the PCI address set cache for later
4714 4715
     * use in hotplug
     */
4716
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4717
        VIR_DEBUG("Assigning domain PCI addresses");
4718
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
4719
            goto cleanup;
4720 4721
    }

4722
    VIR_DEBUG("Building emulator command line");
4723
    if (!(cmd = qemuBuildCommandLine(conn, driver, vm->def, priv->monConfig,
E
Eric Blake 已提交
4724
                                     priv->monJSON, priv->qemuCaps,
4725
                                     migrateFrom, stdin_fd, snapshot, vmop,
4726
                                     &buildCommandLineCallbacks, false,
4727
                                     qemuCheckFips(),
4728 4729
                                     nodemask,
                                     &nnicindexes, &nicindexes)))
4730 4731 4732 4733
        goto cleanup;

    /* now that we know it is about to start call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4734
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4735 4736 4737
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4738 4739
                              VIR_HOOK_QEMU_OP_START, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
4740 4741 4742 4743 4744 4745 4746 4747 4748
        VIR_FREE(xml);

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

4749
    qemuLogOperation(vm, "starting up", logfile, cmd);
4750

4751
    qemuDomainObjCheckTaint(driver, vm, logfile);
4752

4753 4754
    if ((pos = lseek(logfile, 0, SEEK_END)) < 0)
        VIR_WARN("Unable to seek to end of logfile: %s",
4755
                 virStrerror(errno, ebuf, sizeof(ebuf)));
4756 4757

    VIR_DEBUG("Clear emulator capabilities: %d",
4758 4759
              cfg->clearEmulatorCapabilities);
    if (cfg->clearEmulatorCapabilities)
4760 4761
        virCommandClearCaps(cmd);

4762 4763
    /* in case a certain disk is desirous of CAP_SYS_RAWIO, add this */
    for (i = 0; i < vm->def->ndisks; i++) {
4764
        virDomainDeviceDef dev;
4765 4766
        virDomainDiskDefPtr disk = vm->def->disks[i];

4767
        if (vm->def->disks[i]->rawio == VIR_TRISTATE_BOOL_YES) {
R
Roman Bogorodskiy 已提交
4768
#ifdef CAP_SYS_RAWIO
4769
            virCommandAllowCap(cmd, CAP_SYS_RAWIO);
4770
            rawio_set = true;
R
Roman Bogorodskiy 已提交
4771 4772 4773
#else
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Raw I/O is not supported on this platform"));
J
John Ferlan 已提交
4774
            goto cleanup;
R
Roman Bogorodskiy 已提交
4775
#endif
J
John Ferlan 已提交
4776
        }
4777

4778 4779 4780
        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        if (qemuAddSharedDevice(driver, &dev, vm->def->name) < 0)
4781
            goto cleanup;
4782

4783
        if (qemuSetUnprivSGIO(&dev) < 0)
4784
            goto cleanup;
4785 4786
    }

4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804
    /* If rawio not already set, check hostdevs as well */
    if (!rawio_set) {
        for (i = 0; i < vm->def->nhostdevs; i++) {
            virDomainHostdevSubsysSCSIPtr scsisrc =
                &vm->def->hostdevs[i]->source.subsys.u.scsi;
            if (scsisrc->rawio == VIR_TRISTATE_BOOL_YES) {
#ifdef CAP_SYS_RAWIO
                virCommandAllowCap(cmd, CAP_SYS_RAWIO);
                break;
#else
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("Raw I/O is not supported on this platform"));
                goto cleanup;
#endif
            }
        }
    }

4805
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);
4806 4807
    virCommandSetMaxProcesses(cmd, cfg->maxProcesses);
    virCommandSetMaxFiles(cmd, cfg->maxFiles);
4808
    virCommandSetUmask(cmd, 0x002);
4809

4810 4811 4812 4813 4814 4815
    VIR_DEBUG("Setting up security labelling");
    if (virSecurityManagerSetChildProcessLabel(driver->securityManager,
                                               vm->def, cmd) < 0) {
        goto cleanup;
    }

4816 4817 4818
    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
4819
    virCommandSetPidFile(cmd, priv->pidfile);
4820
    virCommandDaemonize(cmd);
4821
    virCommandRequireHandshake(cmd);
4822

4823 4824
    if (virSecurityManagerPreFork(driver->securityManager) < 0)
        goto cleanup;
4825
    ret = virCommandRun(cmd, NULL);
4826
    virSecurityManagerPostFork(driver->securityManager);
4827

E
Eric Blake 已提交
4828
    /* wait for qemu process to show up */
4829
    if (ret == 0) {
4830
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
4831 4832
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
4833 4834
            ret = -1;
        }
4835 4836 4837 4838 4839
        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);
4840 4841
    }

4842
    VIR_DEBUG("Writing early domain status to disk");
4843
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
4844 4845
        goto cleanup;

4846 4847
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
4848 4849
        /* Read errors from child that occurred between fork and exec. */
        qemuProcessReadChildErrors(driver, vm, pos);
4850 4851 4852
        goto cleanup;
    }

4853
    VIR_DEBUG("Setting up domain cgroup (if required)");
4854
    if (qemuSetupCgroup(driver, vm, nnicindexes, nicindexes) < 0)
4855 4856 4857 4858 4859
        goto cleanup;

    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
    if (!vm->def->cputune.emulatorpin &&
4860
        qemuProcessInitCpuAffinity(driver, vm) < 0)
4861 4862
        goto cleanup;

4863 4864
    VIR_DEBUG("Setting domain security labels");
    if (virSecurityManagerSetAllLabel(driver->securityManager,
4865
                                      vm->def, stdin_path) < 0)
4866 4867
        goto cleanup;

4868 4869 4870 4871 4872 4873
    /* 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;

4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887
    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) &&
4888
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def, stdin_fd) < 0)
4889 4890 4891 4892
            goto cleanup;
    }

    VIR_DEBUG("Labelling done, completing handshake to child");
4893
    if (virCommandHandshakeNotify(cmd) < 0)
4894 4895 4896
        goto cleanup;
    VIR_DEBUG("Handshake complete, child running");

4897
    if (migrateFrom)
4898
        flags |= VIR_QEMU_PROCESS_START_PAUSED;
4899 4900 4901 4902 4903 4904 4905

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

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

4906
    VIR_DEBUG("Setting cgroup for emulator (if required)");
4907
    if (qemuSetupCgroupForEmulator(driver, vm) < 0)
4908 4909 4910 4911 4912 4913
        goto cleanup;

    VIR_DEBUG("Setting affinity of emulator threads");
    if (qemuProcessSetEmulatorAffinity(vm) < 0)
        goto cleanup;

4914
    VIR_DEBUG("Waiting for monitor to show up");
4915
    if (qemuProcessWaitForMonitor(driver, vm, asyncJob, priv->qemuCaps, pos) < 0)
4916 4917
        goto cleanup;

D
Daniel P. Berrange 已提交
4918
    /* Failure to connect to agent shouldn't be fatal */
4919 4920 4921 4922
    if ((ret = qemuConnectAgent(driver, vm)) < 0) {
        if (ret == -2)
            goto cleanup;

D
Daniel P. Berrange 已提交
4923 4924 4925 4926 4927 4928
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

4929
    VIR_DEBUG("Detecting if required emulator features are present");
4930
    if (!qemuProcessVerifyGuestCPU(driver, vm, asyncJob))
4931 4932
        goto cleanup;

4933
    VIR_DEBUG("Setting up post-init cgroup restrictions");
4934
    if (qemuSetupCpusetMems(vm) < 0)
4935 4936
        goto cleanup;

4937
    VIR_DEBUG("Detecting VCPU PIDs");
4938
    if (qemuProcessDetectVcpuPIDs(driver, vm, asyncJob) < 0)
4939 4940
        goto cleanup;

4941 4942 4943 4944
    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, asyncJob) < 0)
        goto cleanup;

4945
    VIR_DEBUG("Setting cgroup for each VCPU (if required)");
4946
    if (qemuSetupCgroupForVcpu(vm) < 0)
4947 4948
        goto cleanup;

4949 4950 4951 4952
    VIR_DEBUG("Setting cgroup for each IOThread (if required)");
    if (qemuSetupCgroupForIOThreads(vm) < 0)
        goto cleanup;

4953
    VIR_DEBUG("Setting VCPU affinities");
4954
    if (qemuProcessSetVcpuAffinities(vm) < 0)
4955 4956
        goto cleanup;

4957 4958 4959 4960
    VIR_DEBUG("Setting affinity of IOThread threads");
    if (qemuProcessSetIOThreadsAffinity(vm) < 0)
        goto cleanup;

4961 4962 4963 4964
    VIR_DEBUG("Setting scheduler parameters");
    if (qemuProcessSetSchedulers(vm) < 0)
        goto cleanup;

4965
    VIR_DEBUG("Setting any required VM passwords");
4966
    if (qemuProcessInitPasswords(conn, driver, vm, asyncJob) < 0)
4967 4968 4969 4970
        goto cleanup;

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
4971
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4972
        VIR_DEBUG("Determining domain device PCI addresses");
4973
        if (qemuProcessInitPCIAddresses(driver, vm, asyncJob) < 0)
4974 4975 4976
            goto cleanup;
    }

4977 4978 4979 4980
    /* set default link states */
    /* qemu doesn't support setting this on the command line, so
     * enter the monitor */
    VIR_DEBUG("Setting network link states");
4981 4982
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
4983 4984 4985
    if (qemuProcessSetLinkStates(vm) < 0)
        goto exit_monitor;
    if (qemuDomainObjExitMonitor(driver, vm))
4986 4987
        goto cleanup;

4988
    VIR_DEBUG("Fetching list of active devices");
4989
    if (qemuDomainUpdateDeviceList(driver, vm, asyncJob) < 0)
4990 4991
        goto cleanup;

4992 4993 4994 4995
    /* 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.  */
4996
    VIR_DEBUG("Setting initial memory amount");
E
Eric Blake 已提交
4997
    cur_balloon = vm->def->mem.cur_balloon;
4998
    if (cur_balloon != vm->def->mem.cur_balloon) {
4999 5000 5001
        virReportError(VIR_ERR_OVERFLOW,
                       _("unable to set balloon to %lld"),
                       vm->def->mem.cur_balloon);
5002 5003
        goto cleanup;
    }
5004 5005
    if (vm->def->memballoon && vm->def->memballoon->period)
        period = vm->def->memballoon->period;
5006 5007
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
5008 5009
    if (period)
        qemuMonitorSetMemoryStatsPeriod(priv->mon, period);
5010 5011
    if (qemuMonitorSetBalloon(priv->mon, cur_balloon) < 0)
        goto exit_monitor;
5012 5013
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
5014

5015 5016 5017 5018
    VIR_DEBUG("Detecting actual memory size for video device");
    if (qemuProcessUpdateVideoRamSize(driver, vm, asyncJob) < 0)
        goto cleanup;

5019
    if (!(flags & VIR_QEMU_PROCESS_START_PAUSED)) {
5020
        VIR_DEBUG("Starting domain CPUs");
5021
        /* Allow the CPUS to start executing */
J
Jiri Denemark 已提交
5022
        if (qemuProcessStartCPUs(driver, vm, conn,
5023 5024
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 QEMU_ASYNC_JOB_NONE) < 0) {
5025
            if (virGetLastError() == NULL)
5026 5027
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("resume operation failed"));
5028 5029
            goto cleanup;
        }
J
Jiri Denemark 已提交
5030 5031 5032 5033 5034
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             migrateFrom ?
                             VIR_DOMAIN_PAUSED_MIGRATION :
                             VIR_DOMAIN_PAUSED_USER);
5035 5036
    }

5037
    if (flags & VIR_QEMU_PROCESS_START_AUTODESTROY &&
5038 5039 5040
        qemuProcessAutoDestroyAdd(driver, vm, conn) < 0)
        goto cleanup;

5041
    VIR_DEBUG("Writing domain status to disk");
5042
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
5043 5044
        goto cleanup;

5045 5046
    /* finally we can call the 'started' hook script if any */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5047
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061
        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;
    }

5062 5063 5064 5065
    /* Keep watching qemu log for errors during incoming migration, otherwise
     * unset reporting errors from qemu log. */
    if (!migrateFrom)
        qemuMonitorSetDomainLog(priv->mon, -1);
5066

5067 5068
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);
5069
    virObjectUnref(cfg);
5070
    virObjectUnref(caps);
5071
    VIR_FREE(nicindexes);
5072 5073 5074

    return 0;

5075
 cleanup:
5076 5077 5078
    /* 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 */
5079
    VIR_FREE(nodeset);
5080 5081
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);
5082 5083
    if (priv->mon)
        qemuMonitorSetDomainLog(priv->mon, -1);
5084
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, stop_flags);
5085
    virObjectUnref(cfg);
5086
    virObjectUnref(caps);
5087
    VIR_FREE(nicindexes);
5088 5089

    return -1;
5090 5091 5092 5093

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;
5094 5095 5096
}


5097
int
5098
qemuProcessKill(virDomainObjPtr vm, unsigned int flags)
5099
{
5100
    int ret;
5101

5102 5103 5104
    VIR_DEBUG("vm=%p name=%s pid=%llu flags=%x",
              vm, vm->def->name,
              (unsigned long long)vm->pid, flags);
5105

5106 5107 5108 5109 5110
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
5111 5112
    }

5113
    if (flags & VIR_QEMU_PROCESS_KILL_NOWAIT) {
5114 5115 5116 5117 5118
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
5119

5120 5121
    ret = virProcessKillPainfully(vm->pid,
                                  !!(flags & VIR_QEMU_PROCESS_KILL_FORCE));
5122

5123
    return ret;
5124 5125 5126
}


5127
void qemuProcessStop(virQEMUDriverPtr driver,
5128
                     virDomainObjPtr vm,
5129 5130
                     virDomainShutoffReason reason,
                     unsigned int flags)
5131 5132 5133 5134 5135 5136
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
5137
    virNetDevVPortProfilePtr vport = NULL;
5138
    size_t i;
5139 5140 5141
    int logfile = -1;
    char *timestamp;
    char ebuf[1024];
5142
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5143

5144 5145 5146
    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);
5147 5148 5149

    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("VM '%s' not active", vm->def->name);
5150
        virObjectUnref(cfg);
5151 5152 5153
        return;
    }

5154 5155 5156 5157
    /* This method is routinely used in clean up paths. Disable error
     * reporting so we don't squash a legit error. */
    orig_err = virSaveLastError();

5158
    /*
5159 5160
     * We may unlock the vm in qemuProcessKill(), and another thread
     * can lock the vm, and then call qemuProcessStop(). So we should
5161 5162 5163 5164
     * set vm->def->id to -1 here to avoid qemuProcessStop() to be called twice.
     */
    vm->def->id = -1;

5165
    if (virAtomicIntDecAndTest(&driver->nactive) && driver->inhibitCallback)
5166 5167
        driver->inhibitCallback(false, driver->inhibitOpaque);

5168
    if ((logfile = qemuDomainCreateLog(driver, vm, true)) < 0) {
5169 5170 5171
        /* 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",
5172
                  virStrerror(errno, ebuf, sizeof(ebuf)));
5173
    } else {
5174
        if ((timestamp = virTimeStringNow()) != NULL) {
5175 5176 5177 5178
            if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
                safewrite(logfile, SHUTDOWN_POSTFIX,
                          strlen(SHUTDOWN_POSTFIX)) < 0) {
                VIR_WARN("Unable to write timestamp to logfile: %s",
5179
                         virStrerror(errno, ebuf, sizeof(ebuf)));
5180 5181 5182 5183 5184 5185 5186
            }

            VIR_FREE(timestamp);
        }

        if (VIR_CLOSE(logfile) < 0)
             VIR_WARN("Unable to close logfile: %s",
5187
                      virStrerror(errno, ebuf, sizeof(ebuf)));
5188 5189
    }

5190 5191 5192
    /* Clear network bandwidth */
    virDomainClearNetBandwidth(vm);

5193 5194
    virDomainConfVMNWFilterTeardown(vm);

5195
    if (cfg->macFilter) {
5196
        def = vm->def;
5197
        for (i = 0; i < def->nnets; i++) {
5198 5199 5200
            virDomainNetDefPtr net = def->nets[i];
            if (net->ifname == NULL)
                continue;
5201 5202 5203
            ignore_value(ebtablesRemoveForwardAllowIn(driver->ebtables,
                                                      net->ifname,
                                                      &net->mac));
5204 5205 5206
        }
    }

5207
    virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
5208
    priv->nbdPort = 0;
5209

D
Daniel P. Berrange 已提交
5210 5211 5212 5213 5214 5215
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
        priv->agentError = false;
    }

5216
    if (priv->mon) {
5217
        qemuMonitorClose(priv->mon);
5218 5219
        priv->mon = NULL;
    }
5220 5221 5222 5223 5224 5225 5226 5227

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

5228 5229 5230 5231 5232 5233
    ignore_value(virDomainChrDefForeach(vm->def,
                                        false,
                                        qemuProcessCleanupChardevDevice,
                                        NULL));


5234
    /* shut it off for sure */
5235 5236 5237
    ignore_value(qemuProcessKill(vm,
                                 VIR_QEMU_PROCESS_KILL_FORCE|
                                 VIR_QEMU_PROCESS_KILL_NOCHECK));
5238

5239 5240
    qemuDomainCleanupRun(driver, vm);

5241
    /* Stop autodestroy in case guest is restarted */
5242
    qemuProcessAutoDestroyRemove(driver, vm);
5243

5244 5245
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5246
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5247 5248 5249

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
5250 5251
                    VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
5252 5253 5254
        VIR_FREE(xml);
    }

5255 5256 5257 5258
    /* Reset Security Labels unless caller don't want us to */
    if (!(flags & VIR_QEMU_PROCESS_STOP_NO_RELABEL))
        virSecurityManagerRestoreAllLabel(driver->securityManager,
                                          vm->def,
5259
                                          !!(flags & VIR_QEMU_PROCESS_STOP_MIGRATED));
5260
    virSecurityManagerReleaseLabel(driver->securityManager, vm->def);
5261

5262
    for (i = 0; i < vm->def->ndisks; i++) {
5263
        virDomainDeviceDef dev;
5264
        virDomainDiskDefPtr disk = vm->def->disks[i];
5265 5266 5267 5268

        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        ignore_value(qemuRemoveSharedDevice(driver, &dev, vm->def->name));
5269 5270
    }

5271
    /* Clear out dynamically assigned labels */
5272
    for (i = 0; i < vm->def->nseclabels; i++) {
5273
        if (vm->def->seclabels[i]->type == VIR_DOMAIN_SECLABEL_DYNAMIC)
5274 5275
            VIR_FREE(vm->def->seclabels[i]->label);
        VIR_FREE(vm->def->seclabels[i]->imagelabel);
5276 5277
    }

5278 5279 5280
    virStringFreeList(priv->qemuDevices);
    priv->qemuDevices = NULL;

5281 5282 5283
    virDomainDefClearDeviceAliases(vm->def);
    if (!priv->persistentAddrs) {
        virDomainDefClearPCIAddresses(vm->def);
5284
        virDomainPCIAddressSetFree(priv->pciaddrs);
5285
        priv->pciaddrs = NULL;
5286
        virDomainDefClearCCWAddresses(vm->def);
J
Ján Tomko 已提交
5287
        virDomainCCWAddressSetFree(priv->ccwaddrs);
5288
        priv->ccwaddrs = NULL;
5289 5290 5291 5292 5293 5294 5295
    }

    qemuDomainReAttachHostDevices(driver, vm->def);

    def = vm->def;
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
5296 5297 5298 5299
        vport = virDomainNetGetActualVirtPortProfile(net);

        switch (virDomainNetGetActualType(net)) {
        case VIR_DOMAIN_NET_TYPE_DIRECT:
5300
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
5301
                             net->ifname, &net->mac,
5302 5303
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
5304
                             virDomainNetGetActualVirtPortProfile(net),
5305
                             cfg->stateDir));
5306
            VIR_FREE(net->ifname);
5307 5308 5309 5310 5311
            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))
5312
                ignore_value(virNetDevTapDelete(net->ifname, net->backend.tap));
5313 5314
#endif
            break;
5315
        }
5316 5317 5318
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
A
Ansis Atteka 已提交
5319 5320 5321 5322 5323
        if (vport && vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH)
            ignore_value(virNetDevOpenvswitchRemovePort(
                                       virDomainNetGetActualBridgeName(net),
                                       net->ifname));

5324 5325
        /* kick the device out of the hostdev list too */
        virDomainNetRemoveHostdev(def, net);
5326
        networkReleaseActualDevice(vm->def, net);
5327
    }
5328

5329
 retry:
5330
    if ((ret = qemuRemoveCgroup(driver, vm)) < 0) {
5331 5332 5333 5334 5335 5336 5337
        if (ret == -EBUSY && (retries++ < 5)) {
            usleep(200*1000);
            goto retry;
        }
        VIR_WARN("Failed to remove cgroup for %s",
                 vm->def->name);
    }
5338
    virCgroupFree(&priv->cgroup);
5339 5340 5341

    qemuProcessRemoveDomainStatus(driver, vm);

5342 5343
    /* Remove VNC and Spice ports from port reservation bitmap, but only if
       they were reserved by the driver (autoport=yes)
5344
    */
5345
    for (i = 0; i < vm->def->ngraphics; ++i) {
5346
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
5347 5348
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (graphics->data.vnc.autoport) {
5349 5350
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.vnc.port);
5351
            } else if (graphics->data.vnc.portReserved) {
5352 5353 5354 5355 5356
                virPortAllocatorSetUsed(driver->remotePorts,
                                        graphics->data.spice.port,
                                        false);
                graphics->data.vnc.portReserved = false;
            }
5357 5358
            virPortAllocatorRelease(driver->webSocketPorts,
                                    graphics->data.vnc.websocket);
5359
        }
5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380
        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;
                }
            }
5381
        }
5382 5383
    }

5384
    vm->taint = 0;
5385
    vm->pid = -1;
J
Jiri Denemark 已提交
5386
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
5387 5388
    VIR_FREE(priv->vcpupids);
    priv->nvcpupids = 0;
5389 5390
    VIR_FREE(priv->iothreadpids);
    priv->niothreadpids = 0;
5391 5392
    virObjectUnref(priv->qemuCaps);
    priv->qemuCaps = NULL;
5393
    VIR_FREE(priv->pidfile);
5394

5395
    /* The "release" hook cleans up additional resources */
5396
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5397
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5398 5399 5400

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
5401 5402
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
5403 5404 5405
        VIR_FREE(xml);
    }

5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416
    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);
    }
5417
    virObjectUnref(cfg);
5418
}
5419 5420


5421
int qemuProcessAttach(virConnectPtr conn ATTRIBUTE_UNUSED,
5422
                      virQEMUDriverPtr driver,
5423
                      virDomainObjPtr vm,
5424
                      pid_t pid,
5425 5426 5427 5428
                      const char *pidfile,
                      virDomainChrSourceDefPtr monConfig,
                      bool monJSON)
{
5429
    size_t i;
5430 5431 5432 5433 5434
    char ebuf[1024];
    int logfile = -1;
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool running = true;
5435
    virDomainPausedReason reason;
5436
    virSecurityLabelPtr seclabel = NULL;
5437
    virSecurityLabelDefPtr seclabeldef = NULL;
5438
    bool seclabelgen = false;
5439 5440
    virSecurityManagerPtr* sec_managers = NULL;
    const char *model;
5441
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5442
    virCapsPtr caps = NULL;
5443
    bool active = false;
5444
    int ret;
5445 5446 5447 5448

    VIR_DEBUG("Beginning VM attach process");

    if (virDomainObjIsActive(vm)) {
5449 5450
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
5451
        virObjectUnref(cfg);
5452 5453 5454
        return -1;
    }

5455
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
5456
        goto error;
5457

5458 5459 5460 5461 5462
    /* 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");
5463
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm, true) < 0)
5464
        goto error;
5465

5466
    vm->def->id = qemuDriverAllocateID(driver);
5467

5468
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
5469
        driver->inhibitCallback(true, driver->inhibitOpaque);
5470
    active = true;
5471

5472
    if (virFileMakePath(cfg->logDir) < 0) {
5473 5474
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
5475
                             cfg->logDir);
5476
        goto error;
5477 5478 5479
    }

    VIR_FREE(priv->pidfile);
5480
    if (VIR_STRDUP(priv->pidfile, pidfile) < 0)
5481
        goto error;
5482

5483 5484
    vm->pid = pid;

5485
    VIR_DEBUG("Detect security driver config");
5486
    sec_managers = virSecurityManagerGetNested(driver->securityManager);
5487 5488
    if (sec_managers == NULL)
        goto error;
5489 5490

    for (i = 0; sec_managers[i]; i++) {
5491
        seclabelgen = false;
5492 5493
        model = virSecurityManagerGetModel(sec_managers[i]);
        seclabeldef = virDomainDefGetSecurityLabelDef(vm->def, model);
5494
        if (seclabeldef == NULL) {
5495
            if (!(seclabeldef = virSecurityLabelDefNew(model)))
5496 5497 5498
                goto error;
            seclabelgen = true;
        }
5499 5500
        seclabeldef->type = VIR_DOMAIN_SECLABEL_STATIC;
        if (VIR_ALLOC(seclabel) < 0)
5501
            goto error;
5502
        if (virSecurityManagerGetProcessLabel(sec_managers[i],
5503
                                              vm->def, vm->pid, seclabel) < 0)
5504
            goto error;
5505

5506
        if (VIR_STRDUP(seclabeldef->model, model) < 0)
5507
            goto error;
5508

5509
        if (VIR_STRDUP(seclabeldef->label, seclabel->label) < 0)
5510
            goto error;
5511
        VIR_FREE(seclabel);
5512 5513 5514 5515 5516 5517

        if (seclabelgen) {
            if (VIR_APPEND_ELEMENT(vm->def->seclabels, vm->def->nseclabels, seclabeldef) < 0)
                goto error;
            seclabelgen = false;
        }
5518
    }
5519

5520 5521
    if (virSecurityManagerCheckAllLabel(driver->securityManager, vm->def) < 0)
        goto error;
5522 5523 5524
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0)
        goto error;

5525 5526
    VIR_DEBUG("Creating domain log file");
    if ((logfile = qemuDomainCreateLog(driver, vm, false)) < 0)
5527
        goto error;
5528 5529

    VIR_DEBUG("Determining emulator version");
5530 5531
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
5532 5533
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
5534
        goto error;
5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546

    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
5547
     * we also need to populate the PCI address set cache for later
5548 5549
     * use in hotplug
     */
5550
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
5551
        VIR_DEBUG("Assigning domain PCI addresses");
5552
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
5553
            goto error;
5554 5555
    }

5556
    if ((timestamp = virTimeStringNow()) == NULL) {
5557
        goto error;
5558 5559 5560 5561
    } 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",
5562
                     virStrerror(errno, ebuf, sizeof(ebuf)));
5563 5564 5565 5566 5567 5568 5569 5570
        }

        VIR_FREE(timestamp);
    }

    qemuDomainObjTaint(driver, vm, VIR_DOMAIN_TAINT_EXTERNAL_LAUNCH, logfile);

    VIR_DEBUG("Waiting for monitor to show up");
5571
    if (qemuProcessWaitForMonitor(driver, vm, QEMU_ASYNC_JOB_NONE, priv->qemuCaps, -1) < 0)
5572
        goto error;
5573

D
Daniel P. Berrange 已提交
5574
    /* Failure to connect to agent shouldn't be fatal */
5575 5576 5577 5578
    if ((ret = qemuConnectAgent(driver, vm)) < 0) {
        if (ret == -2)
            goto error;

D
Daniel P. Berrange 已提交
5579 5580 5581 5582 5583 5584
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

5585
    VIR_DEBUG("Detecting VCPU PIDs");
5586
    if (qemuProcessDetectVcpuPIDs(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5587 5588 5589 5590
        goto error;

    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5591
        goto error;
5592 5593 5594

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
5595
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
5596
        VIR_DEBUG("Determining domain device PCI addresses");
5597
        if (qemuProcessInitPCIAddresses(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5598
            goto error;
5599 5600 5601
    }

    VIR_DEBUG("Getting initial memory amount");
5602
    qemuDomainObjEnterMonitor(driver, vm);
5603 5604 5605 5606 5607 5608 5609
    if (qemuMonitorGetBalloonInfo(priv->mon, &vm->def->mem.cur_balloon) < 0)
        goto exit_monitor;
    if (qemuMonitorGetStatus(priv->mon, &running, &reason) < 0)
        goto exit_monitor;
    if (qemuMonitorGetVirtType(priv->mon, &vm->def->virtType) < 0)
        goto exit_monitor;
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5610
        goto error;
5611

5612
    if (running) {
5613 5614
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                             VIR_DOMAIN_RUNNING_UNPAUSED);
5615 5616 5617 5618
        if (vm->def->memballoon && vm->def->memballoon->period) {
            qemuDomainObjEnterMonitor(driver, vm);
            qemuMonitorSetMemoryStatsPeriod(priv->mon,
                                            vm->def->memballoon->period);
5619 5620
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                goto error;
5621 5622
        }
    } else {
5623
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
5624
    }
5625 5626

    VIR_DEBUG("Writing domain status to disk");
5627
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
5628
        goto error;
5629

5630 5631
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5632
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643
        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)
5644
            goto error;
5645 5646
    }

5647 5648
    VIR_FORCE_CLOSE(logfile);
    VIR_FREE(seclabel);
5649
    VIR_FREE(sec_managers);
5650
    virObjectUnref(cfg);
5651
    virObjectUnref(caps);
5652 5653 5654

    return 0;

5655 5656
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
5657
 error:
5658 5659 5660 5661 5662 5663
    /* 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);
5664 5665
    VIR_FORCE_CLOSE(logfile);
    VIR_FREE(seclabel);
5666
    VIR_FREE(sec_managers);
5667 5668
    if (seclabelgen)
        virSecurityLabelDefFree(seclabeldef);
5669
    virDomainChrSourceDefFree(monConfig);
5670
    virObjectUnref(cfg);
5671
    virObjectUnref(caps);
5672 5673 5674 5675
    return -1;
}


5676
static virDomainObjPtr
5677 5678 5679
qemuProcessAutoDestroy(virDomainObjPtr dom,
                       virConnectPtr conn,
                       void *opaque)
5680
{
5681
    virQEMUDriverPtr driver = opaque;
5682
    qemuDomainObjPrivatePtr priv = dom->privateData;
5683
    virObjectEventPtr event = NULL;
5684

5685
    VIR_DEBUG("vm=%s, conn=%p", dom->def->name, conn);
5686

5687 5688
    if (priv->job.asyncJob) {
        VIR_DEBUG("vm=%s has long-term job active, cancelling",
5689
                  dom->def->name);
5690
        qemuDomainObjDiscardAsyncJob(driver, dom);
5691 5692
    }

5693 5694
    if (qemuDomainObjBeginJob(driver, dom,
                              QEMU_JOB_DESTROY) < 0)
5695 5696 5697
        goto cleanup;

    VIR_DEBUG("Killing domain");
5698

5699 5700
    qemuProcessStop(driver, dom, VIR_DOMAIN_SHUTOFF_DESTROYED,
                    VIR_QEMU_PROCESS_STOP_MIGRATED);
5701

5702
    virDomainAuditStop(dom, "destroyed");
5703
    event = virDomainEventLifecycleNewFromObj(dom,
5704 5705
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_DESTROYED);
5706

5707 5708 5709
    qemuDomainObjEndJob(driver, dom);

    if (!dom->persistent)
5710
        qemuDomainRemoveInactive(driver, dom);
5711

5712
    if (event)
5713
        qemuDomainEventQueue(driver, event);
5714

5715
 cleanup:
5716
    return dom;
5717 5718
}

5719
int qemuProcessAutoDestroyAdd(virQEMUDriverPtr driver,
5720 5721 5722
                              virDomainObjPtr vm,
                              virConnectPtr conn)
{
5723
    VIR_DEBUG("vm=%s, conn=%p", vm->def->name, conn);
5724 5725
    return virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                qemuProcessAutoDestroy);
5726 5727
}

5728
int qemuProcessAutoDestroyRemove(virQEMUDriverPtr driver,
5729 5730
                                 virDomainObjPtr vm)
{
5731
    int ret;
5732
    VIR_DEBUG("vm=%s", vm->def->name);
5733 5734 5735
    ret = virCloseCallbacksUnset(driver->closeCallbacks, vm,
                                 qemuProcessAutoDestroy);
    return ret;
5736
}
5737

5738
bool qemuProcessAutoDestroyActive(virQEMUDriverPtr driver,
5739 5740
                                  virDomainObjPtr vm)
{
5741
    virCloseCallback cb;
5742
    VIR_DEBUG("vm=%s", vm->def->name);
5743
    cb = virCloseCallbacksGet(driver->closeCallbacks, vm, NULL);
5744
    return cb == qemuProcessAutoDestroy;
5745
}