qemu_process.c 138.6 KB
Newer Older
1 2 3
/*
 * qemu_process.h: QEMU process management
 *
4
 * Copyright (C) 2006-2013 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>
28 29
#include <sys/time.h>
#include <sys/resource.h>
R
Roman Bogorodskiy 已提交
30 31 32 33 34 35
#if defined(__linux__)
# include <linux/capability.h>
#elif defined(__FreeBSD__)
# include <sys/param.h>
# include <sys/cpuset.h>
#endif
36 37 38 39 40 41 42 43 44 45

#include "qemu_process.h"
#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"
#include "qemu_bridge_filter.h"
46
#include "qemu_migration.h"
47 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
#include "virutil.h"
56 57
#include "c-ctype.h"
#include "nodeinfo.h"
58
#include "domain_audit.h"
59
#include "domain_nwfilter.h"
60
#include "locking/domain_lock.h"
61
#include "network/bridge_driver.h"
62
#include "viruuid.h"
63
#include "virprocess.h"
64
#include "virtime.h"
A
Ansis Atteka 已提交
65
#include "virnetdevtap.h"
66
#include "virbitmap.h"
67
#include "viratomic.h"
68
#include "virnuma.h"
69 70 71 72

#define VIR_FROM_THIS VIR_FROM_QEMU

#define START_POSTFIX ": starting up\n"
73
#define ATTACH_POSTFIX ": attaching\n"
74 75 76
#define SHUTDOWN_POSTFIX ": shutting down\n"

/**
77
 * qemuProcessRemoveDomainStatus
78 79 80 81 82 83
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
84
qemuProcessRemoveDomainStatus(virQEMUDriverPtr driver,
85 86 87 88
                              virDomainObjPtr vm)
{
    char ebuf[1024];
    char *file = NULL;
89
    qemuDomainObjPrivatePtr priv = vm->privateData;
90 91
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;
92

93
    if (virAsprintf(&file, "%s/%s.xml", cfg->stateDir, vm->def->name) < 0) {
94
        virReportOOMError();
95
        goto cleanup;
96 97 98 99 100 101 102
    }

    if (unlink(file) < 0 && errno != ENOENT && errno != ENOTDIR)
        VIR_WARN("Failed to remove domain XML for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));
    VIR_FREE(file);

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

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


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

D
Daniel P. Berrange 已提交
119 120 121 122 123 124 125
/*
 * This is a callback registered with a qemuAgentPtr instance,
 * and to be invoked when the agent console hits an end of file
 * condition, or error, thus indicating VM shutdown should be
 * performed
 */
static void
126
qemuProcessHandleAgentEOF(qemuAgentPtr agent,
D
Daniel P. Berrange 已提交
127 128 129 130 131 132
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

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

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

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

    if (!priv->agent) {
        VIR_DEBUG("Agent freed already");
        goto unlock;
    }

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

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

149
    virObjectUnlock(vm);
150 151

    qemuAgentClose(agent);
152 153 154 155 156
    return;

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


/*
 * This is invoked when there is some kind of error
 * parsing data to/from the agent. The VM can continue
 * to run, but no further agent commands will be
 * allowed
 */
static void
qemuProcessHandleAgentError(qemuAgentPtr agent ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

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

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

    priv = vm->privateData;

    priv->agentError = true;

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

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

188
    virObjectUnref(vm);
D
Daniel P. Berrange 已提交
189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
}


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

static virDomainChrSourceDefPtr
qemuFindAgentConfig(virDomainDefPtr def)
{
    virDomainChrSourceDefPtr config = NULL;
    int i;

    for (i = 0 ; i < def->nchannels ; i++) {
        virDomainChrDefPtr channel = def->channels[i];

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

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

    return config;
}

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

    if (!config)
        return 0;

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

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

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

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

248
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
249 250 251 252 253 254 255 256 257 258
    priv->agentStart = 0;

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

    if (agent == NULL)
259
        virObjectUnref(vm);
D
Daniel P. Berrange 已提交
260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278

    if (!virDomainObjIsActive(vm)) {
        qemuAgentClose(agent);
        goto cleanup;
    }
    priv->agent = agent;

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

    ret = 0;

cleanup:
    return ret;
}


279
/*
280
 * This is a callback registered with a qemuMonitorPtr instance,
281 282 283 284 285 286
 * 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,
287
                            virDomainObjPtr vm)
288
{
289
    virQEMUDriverPtr driver = qemu_driver;
290 291
    virDomainEventPtr event = NULL;
    qemuDomainObjPrivatePtr priv;
292 293 294
    int eventReason = VIR_DOMAIN_EVENT_STOPPED_SHUTDOWN;
    int stopReason = VIR_DOMAIN_SHUTOFF_SHUTDOWN;
    const char *auditReason = "shutdown";
295 296 297

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

298
    virObjectLock(vm);
299

300 301 302 303 304 305 306
    priv = vm->privateData;

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

307 308
    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Domain %p is not active, ignoring EOF", vm);
309
        goto unlock;
310 311
    }

312
    if (priv->monJSON && !priv->gotShutdown) {
313 314
        VIR_DEBUG("Monitor connection to '%s' closed without SHUTDOWN event; "
                  "assuming the domain crashed", vm->def->name);
315
        eventReason = VIR_DOMAIN_EVENT_STOPPED_FAILED;
316
        stopReason = VIR_DOMAIN_SHUTOFF_CRASHED;
317
        auditReason = "failed";
318 319 320 321
    }

    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STOPPED,
322
                                     eventReason);
323
    qemuProcessStop(driver, vm, stopReason, 0);
324
    virDomainAuditStop(vm, auditReason);
325

326
    if (!vm->persistent) {
327
        qemuDomainRemoveInactive(driver, vm);
328 329 330 331
        goto cleanup;
    }

unlock:
332
    virObjectUnlock(vm);
333

334
cleanup:
335
    if (event)
336
        qemuDomainEventQueue(driver, event);
337 338 339 340 341 342 343 344 345 346 347 348 349
}


/*
 * 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,
                              virDomainObjPtr vm)
{
350
    virQEMUDriverPtr driver = qemu_driver;
351 352 353 354
    virDomainEventPtr event = NULL;

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

355
    virObjectLock(vm);
356

357
    ((qemuDomainObjPrivatePtr) vm->privateData)->monError = true;
358 359 360 361
    event = virDomainEventControlErrorNewFromObj(vm);
    if (event)
        qemuDomainEventQueue(driver, event);

362
    virObjectUnlock(vm);
363 364 365 366 367 368 369
}


static virDomainDiskDefPtr
qemuProcessFindDomainDiskByPath(virDomainObjPtr vm,
                                const char *path)
{
370
    int i = virDomainDiskIndexByName(vm->def, path, true);
371

372 373
    if (i >= 0)
        return vm->def->disks[i];
374

375 376 377
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with path %s"),
                   path);
378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397
    return NULL;
}

static virDomainDiskDefPtr
qemuProcessFindDomainDiskByAlias(virDomainObjPtr vm,
                                 const char *alias)
{
    int i;

    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;
    }

398 399 400
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with alias %s"),
                   alias);
401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417
    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;

    if (!disk->encryption) {
418 419 420
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("disk %s does not have any encryption information"),
                       disk->src);
421 422 423 424 425
        return -1;
    }
    enc = disk->encryption;

    if (!conn) {
426 427
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("cannot find secrets without a connection"));
428 429 430 431 432 433
        goto cleanup;
    }

    if (conn->secretDriver == NULL ||
        conn->secretDriver->lookupByUUID == NULL ||
        conn->secretDriver->getValue == NULL) {
434 435
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("secret storage not supported"));
436 437 438 439 440 441 442
        goto cleanup;
    }

    if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW ||
        enc->nsecrets != 1 ||
        enc->secrets[0]->type !=
        VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE) {
443 444
        virReportError(VIR_ERR_XML_ERROR,
                       _("invalid <encryption> for volume %s"), disk->src);
445 446 447 448 449 450 451
        goto cleanup;
    }

    secret = conn->secretDriver->lookupByUUID(conn,
                                              enc->secrets[0]->uuid);
    if (secret == NULL)
        goto cleanup;
452
    data = conn->secretDriver->getValue(secret, &size, 0,
453
                                        VIR_SECRET_GET_VALUE_INTERNAL_CALL);
454
    virObjectUnref(secret);
455 456 457 458 459 460
    if (data == NULL)
        goto cleanup;

    if (memchr(data, '\0', size) != NULL) {
        memset(data, 0, size);
        VIR_FREE(data);
461 462 463
        virReportError(VIR_ERR_XML_ERROR,
                       _("format='qcow' passphrase for %s must not contain a "
                         "'\\0'"), disk->src);
464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498
        goto cleanup;
    }

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

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

    *secretRet = passphrase;
    *secretLen = size;

    ret = 0;

cleanup:
    return ret;
}

static int
qemuProcessFindVolumeQcowPassphrase(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virConnectPtr conn,
                                    virDomainObjPtr vm,
                                    const char *path,
                                    char **secretRet,
                                    size_t *secretLen)
{
    virDomainDiskDefPtr disk;
    int ret = -1;

499
    virObjectLock(vm);
500 501 502 503 504 505 506 507
    disk = qemuProcessFindDomainDiskByPath(vm, path);

    if (!disk)
        goto cleanup;

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

cleanup:
508
    virObjectUnlock(vm);
509 510 511 512 513 514 515 516
    return ret;
}


static int
qemuProcessHandleReset(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                       virDomainObjPtr vm)
{
517
    virQEMUDriverPtr driver = qemu_driver;
518
    virDomainEventPtr event;
519
    qemuDomainObjPrivatePtr priv;
520

521
    virObjectLock(vm);
522

523
    event = virDomainEventRebootNewFromObj(vm);
524 525 526 527
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);

528
    virObjectUnlock(vm);
529

530
    if (event)
531 532 533 534 535 536
        qemuDomainEventQueue(driver, event);

    return 0;
}


537 538 539 540 541 542 543 544 545 546 547
/*
 * 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)
{
548
    virQEMUDriverPtr driver = qemu_driver;
549 550 551 552 553
    virDomainObjPtr vm = opaque;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainEventPtr event = NULL;
    int ret = -1;
    VIR_DEBUG("vm=%p", vm);
554
    virObjectLock(vm);
555
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
556 557 558
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
559 560
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
561 562 563
        goto endjob;
    }

564
    qemuDomainObjEnterMonitor(driver, vm);
565
    if (qemuMonitorSystemReset(priv->mon) < 0) {
566
        qemuDomainObjExitMonitor(driver, vm);
567 568
        goto endjob;
    }
569
    qemuDomainObjExitMonitor(driver, vm);
570 571

    if (!virDomainObjIsActive(vm)) {
572 573
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
574 575 576 577
        goto endjob;
    }

    if (qemuProcessStartCPUs(driver, vm, NULL,
578 579
                             VIR_DOMAIN_RUNNING_BOOTED,
                             QEMU_ASYNC_JOB_NONE) < 0) {
580
        if (virGetLastError() == NULL)
581 582
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
583 584
        goto endjob;
    }
585
    priv->gotShutdown = false;
586 587 588 589 590 591 592
    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_RESUMED,
                                     VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

    ret = 0;

endjob:
593
    if (!qemuDomainObjEndJob(driver, vm))
594 595 596 597
        vm = NULL;

cleanup:
    if (vm) {
598
        if (ret == -1) {
599
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_FORCE));
600
        }
601
        if (virObjectUnref(vm))
602
            virObjectUnlock(vm);
603 604 605 606 607 608
    }
    if (event)
        qemuDomainEventQueue(driver, event);
}


609
static void
610
qemuProcessShutdownOrReboot(virQEMUDriverPtr driver,
611
                            virDomainObjPtr vm)
612
{
613 614 615
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
616
        qemuDomainSetFakeReboot(driver, vm, false);
617
        virObjectRef(vm);
618 619 620 621 622
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
623
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
624
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
625
            virObjectUnref(vm);
626 627
        }
    } else {
628
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
629
    }
630
}
631

632 633 634 635
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm)
{
636
    virQEMUDriverPtr driver = qemu_driver;
637 638
    qemuDomainObjPrivatePtr priv;
    virDomainEventPtr event = NULL;
639
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
640

641 642
    VIR_DEBUG("vm=%p", vm);

643
    virObjectLock(vm);
644 645 646 647 648 649

    priv = vm->privateData;
    if (priv->gotShutdown) {
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
650 651 652 653
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
654 655 656 657 658 659 660 661 662 663 664 665
    }
    priv->gotShutdown = true;

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

666
    if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0) {
667 668 669 670
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

671 672 673
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

674 675 676
    qemuProcessShutdownOrReboot(driver, vm);

unlock:
677
    virObjectUnlock(vm);
678
    if (event)
679
        qemuDomainEventQueue(driver, event);
680
    virObjectUnref(cfg);
681

682 683 684 685 686 687 688 689
    return 0;
}


static int
qemuProcessHandleStop(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                      virDomainObjPtr vm)
{
690
    virQEMUDriverPtr driver = qemu_driver;
691
    virDomainEventPtr event = NULL;
692
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
693

694
    virObjectLock(vm);
J
Jiri Denemark 已提交
695
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
696
        qemuDomainObjPrivatePtr priv = vm->privateData;
697

698
        if (priv->gotShutdown) {
699 700
            VIR_DEBUG("Ignoring STOP event after SHUTDOWN");
            goto unlock;
701 702
        }

703 704
        VIR_DEBUG("Transitioned guest %s to paused state",
                  vm->def->name);
705

706
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_UNKNOWN);
707 708 709 710
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);

711 712 713 714 715
        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));

716
        if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0) {
717 718 719
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
720
    }
721 722

unlock:
723
    virObjectUnlock(vm);
724
    if (event)
725
        qemuDomainEventQueue(driver, event);
726
    virObjectUnref(cfg);
727 728 729 730 731

    return 0;
}


732 733 734 735 736 737
static int
qemuProcessHandleResume(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                        virDomainObjPtr vm)
{
    virQEMUDriverPtr driver = qemu_driver;
    virDomainEventPtr event = NULL;
738
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
739

740
    virObjectLock(vm);
741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758
    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);
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

        VIR_DEBUG("Using lock state '%s' on resume event", NULLSTR(priv->lockState));
759
        if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
760 761 762 763 764 765 766 767 768
                                       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);

769
        if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0) {
770 771 772 773 774 775
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
    }

unlock:
776
    virObjectUnlock(vm);
777
    if (event)
778
        qemuDomainEventQueue(driver, event);
779
    virObjectUnref(cfg);
780 781 782 783
    return 0;
}


784 785 786 787 788
static int
qemuProcessHandleRTCChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm,
                           long long offset)
{
789
    virQEMUDriverPtr driver = qemu_driver;
790
    virDomainEventPtr event = NULL;
791
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
792

793
    virObjectLock(vm);
794 795 796
    event = virDomainEventRTCChangeNewFromObj(vm, offset);

    if (vm->def->clock.offset == VIR_DOMAIN_CLOCK_OFFSET_VARIABLE)
797
        vm->def->clock.data.variable.adjustment = offset;
798

799
    if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0)
800
        VIR_WARN("unable to save domain status with RTC change");
801

802
    virObjectUnlock(vm);
803

804
    if (event)
805
        qemuDomainEventQueue(driver, event);
806
    virObjectUnref(cfg);
807 808 809 810 811 812 813 814 815
    return 0;
}


static int
qemuProcessHandleWatchdog(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          int action)
{
816
    virQEMUDriverPtr driver = qemu_driver;
817 818
    virDomainEventPtr watchdogEvent = NULL;
    virDomainEventPtr lifecycleEvent = NULL;
819
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
820

821
    virObjectLock(vm);
822 823 824
    watchdogEvent = virDomainEventWatchdogNewFromObj(vm, action);

    if (action == VIR_DOMAIN_EVENT_WATCHDOG_PAUSE &&
J
Jiri Denemark 已提交
825
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
826
        qemuDomainObjPrivatePtr priv = vm->privateData;
827 828
        VIR_DEBUG("Transitioned guest %s to paused state due to watchdog", vm->def->name);

J
Jiri Denemark 已提交
829
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
830 831 832 833
        lifecycleEvent = virDomainEventNewFromObj(vm,
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

834 835 836 837 838
        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));

839
        if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0) {
840 841 842
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
843 844 845 846 847 848 849
    }

    if (vm->def->watchdog->action == VIR_DOMAIN_WATCHDOG_ACTION_DUMP) {
        struct qemuDomainWatchdogEvent *wdEvent;
        if (VIR_ALLOC(wdEvent) == 0) {
            wdEvent->action = VIR_DOMAIN_WATCHDOG_ACTION_DUMP;
            wdEvent->vm = vm;
W
Wen Congyang 已提交
850 851 852
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
853
            virObjectRef(vm);
854
            if (virThreadPoolSendJob(driver->workerPool, 0, wdEvent) < 0) {
855
                if (!virObjectUnref(vm))
856
                    vm = NULL;
W
Wen Congyang 已提交
857 858
                VIR_FREE(wdEvent);
            }
859
        } else {
860
            virReportOOMError();
861
        }
862 863
    }

864
    if (vm)
865
        virObjectUnlock(vm);
866 867 868 869
    if (watchdogEvent)
        qemuDomainEventQueue(driver, watchdogEvent);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
870

871
    virObjectUnref(cfg);
872 873 874 875 876 877 878 879 880 881 882
    return 0;
}


static int
qemuProcessHandleIOError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virDomainObjPtr vm,
                         const char *diskAlias,
                         int action,
                         const char *reason)
{
883
    virQEMUDriverPtr driver = qemu_driver;
884 885 886 887 888 889
    virDomainEventPtr ioErrorEvent = NULL;
    virDomainEventPtr ioErrorEvent2 = NULL;
    virDomainEventPtr lifecycleEvent = NULL;
    const char *srcPath;
    const char *devAlias;
    virDomainDiskDefPtr disk;
890
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
891

892
    virObjectLock(vm);
893 894 895 896 897 898 899 900 901 902 903 904 905 906
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
        srcPath = disk->src;
        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
907
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
908
        qemuDomainObjPrivatePtr priv = vm->privateData;
909 910
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

J
Jiri Denemark 已提交
911
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
912 913 914 915
        lifecycleEvent = virDomainEventNewFromObj(vm,
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

916 917 918 919 920
        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));

921
        if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0)
922 923
            VIR_WARN("Unable to save status on vm %s after IO error", vm->def->name);
    }
924
    virObjectUnlock(vm);
925

926 927 928 929 930 931
    if (ioErrorEvent)
        qemuDomainEventQueue(driver, ioErrorEvent);
    if (ioErrorEvent2)
        qemuDomainEventQueue(driver, ioErrorEvent2);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
932
    virObjectUnref(cfg);
933 934 935
    return 0;
}

936 937 938 939 940 941 942
static int
qemuProcessHandleBlockJob(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          const char *diskAlias,
                          int type,
                          int status)
{
943
    virQEMUDriverPtr driver = qemu_driver;
944 945 946 947
    virDomainEventPtr event = NULL;
    const char *path;
    virDomainDiskDefPtr disk;

948
    virObjectLock(vm);
949 950 951 952 953
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
        path = disk->src;
        event = virDomainEventBlockJobNewFromObj(vm, path, type, status);
954 955 956 957 958 959 960 961 962
        /* XXX If we completed a block pull or commit, then recompute
         * the cached backing chain to match.  Better would be storing
         * the chain ourselves rather than reprobing, but this
         * requires modifying domain_conf and our XML to fully track
         * the chain across libvirtd restarts.  For that matter, if
         * qemu gains support for committing the active layer, we have
         * to update disk->src.  */
        if ((type == VIR_DOMAIN_BLOCK_JOB_TYPE_PULL ||
             type == VIR_DOMAIN_BLOCK_JOB_TYPE_COMMIT) &&
963 964
            status == VIR_DOMAIN_BLOCK_JOB_COMPLETED)
            qemuDomainDetermineDiskChain(driver, disk, true);
965 966 967
        if (disk->mirror && type == VIR_DOMAIN_BLOCK_JOB_TYPE_COPY &&
            status == VIR_DOMAIN_BLOCK_JOB_READY)
            disk->mirroring = true;
968 969 970
        if (disk->mirror && type == VIR_DOMAIN_BLOCK_JOB_TYPE_COPY &&
            status == VIR_DOMAIN_BLOCK_JOB_FAILED)
            VIR_FREE(disk->mirror);
971 972
    }

973
    virObjectUnlock(vm);
974

975
    if (event)
976 977 978 979
        qemuDomainEventQueue(driver, event);

    return 0;
}
980 981 982 983 984 985 986 987 988 989 990 991 992 993 994

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,
                          const char *saslUsername)
{
995
    virQEMUDriverPtr driver = qemu_driver;
996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
    virDomainEventPtr event;
    virDomainEventGraphicsAddressPtr localAddr = NULL;
    virDomainEventGraphicsAddressPtr remoteAddr = NULL;
    virDomainEventGraphicsSubjectPtr subject = NULL;
    int i;

    if (VIR_ALLOC(localAddr) < 0)
        goto no_memory;
    localAddr->family = localFamily;
    if (!(localAddr->service = strdup(localService)) ||
        !(localAddr->node = strdup(localNode)))
        goto no_memory;

    if (VIR_ALLOC(remoteAddr) < 0)
        goto no_memory;
    remoteAddr->family = remoteFamily;
    if (!(remoteAddr->service = strdup(remoteService)) ||
        !(remoteAddr->node = strdup(remoteNode)))
        goto no_memory;

    if (VIR_ALLOC(subject) < 0)
        goto no_memory;
    if (x509dname) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
            goto no_memory;
        subject->nidentity++;
1022 1023 1024
        if (!(subject->identities[subject->nidentity-1].type = strdup("x509dname")) ||
            !(subject->identities[subject->nidentity-1].name = strdup(x509dname)))
            goto no_memory;
1025 1026 1027 1028 1029
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
            goto no_memory;
        subject->nidentity++;
1030 1031 1032
        if (!(subject->identities[subject->nidentity-1].type = strdup("saslUsername")) ||
            !(subject->identities[subject->nidentity-1].name = strdup(saslUsername)))
            goto no_memory;
1033 1034
    }

1035
    virObjectLock(vm);
1036
    event = virDomainEventGraphicsNewFromObj(vm, phase, localAddr, remoteAddr, authScheme, subject);
1037
    virObjectUnlock(vm);
1038

1039
    if (event)
1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068
        qemuDomainEventQueue(driver, event);

    return 0;

no_memory:
    virReportOOMError();
    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) {
        for (i = 0 ; i < subject->nidentity ; i++) {
            VIR_FREE(subject->identities[i].type);
            VIR_FREE(subject->identities[i].name);
        }
        VIR_FREE(subject->identities);
        VIR_FREE(subject);
    }

    return -1;
}


1069
static void qemuProcessHandleMonitorDestroy(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1070 1071
                                            virDomainObjPtr vm)
{
1072
    virObjectUnref(vm);
1073 1074
}

1075 1076 1077 1078 1079 1080
static int
qemuProcessHandleTrayChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            const char *devAlias,
                            int reason)
{
1081
    virQEMUDriverPtr driver = qemu_driver;
1082 1083
    virDomainEventPtr event = NULL;
    virDomainDiskDefPtr disk;
1084
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1085

1086
    virObjectLock(vm);
1087 1088 1089 1090 1091 1092
    disk = qemuProcessFindDomainDiskByAlias(vm, devAlias);

    if (disk) {
        event = virDomainEventTrayChangeNewFromObj(vm,
                                                   devAlias,
                                                   reason);
1093 1094 1095 1096 1097 1098
        /* 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;

1099
        if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0) {
1100 1101 1102
            VIR_WARN("Unable to save status on vm %s after tray moved event",
                     vm->def->name);
        }
1103 1104
    }

1105
    virObjectUnlock(vm);
1106
    if (event)
1107
        qemuDomainEventQueue(driver, event);
1108
    virObjectUnref(cfg);
1109 1110 1111
    return 0;
}

O
Osier Yang 已提交
1112 1113 1114 1115
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm)
{
1116
    virQEMUDriverPtr driver = qemu_driver;
O
Osier Yang 已提交
1117
    virDomainEventPtr event = NULL;
1118
    virDomainEventPtr lifecycleEvent = NULL;
1119
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1120

1121
    virObjectLock(vm);
O
Osier Yang 已提交
1122 1123
    event = virDomainEventPMWakeupNewFromObj(vm);

1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136
    /* 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);
        lifecycleEvent = virDomainEventNewFromObj(vm,
                                                  VIR_DOMAIN_EVENT_STARTED,
                                                  VIR_DOMAIN_EVENT_STARTED_WAKEUP);

1137
        if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0) {
1138 1139 1140 1141 1142
            VIR_WARN("Unable to save status on vm %s after wakeup event",
                     vm->def->name);
        }
    }

1143
    virObjectUnlock(vm);
1144 1145 1146 1147
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1148
    virObjectUnref(cfg);
O
Osier Yang 已提交
1149 1150
    return 0;
}
1151

O
Osier Yang 已提交
1152 1153 1154 1155
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm)
{
1156
    virQEMUDriverPtr driver = qemu_driver;
O
Osier Yang 已提交
1157
    virDomainEventPtr event = NULL;
J
Jiri Denemark 已提交
1158
    virDomainEventPtr lifecycleEvent = NULL;
1159
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1160

1161
    virObjectLock(vm);
O
Osier Yang 已提交
1162 1163
    event = virDomainEventPMSuspendNewFromObj(vm);

1164
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1165
        qemuDomainObjPrivatePtr priv = vm->privateData;
1166 1167 1168 1169 1170
        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
1171 1172 1173 1174
        lifecycleEvent =
            virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1175

1176
        if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0) {
1177 1178 1179
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1180 1181 1182

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1183 1184
    }

1185
    virObjectUnlock(vm);
O
Osier Yang 已提交
1186

1187 1188 1189 1190
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1191
    virObjectUnref(cfg);
O
Osier Yang 已提交
1192 1193 1194
    return 0;
}

1195 1196 1197 1198 1199
static int
qemuProcessHandleBalloonChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               unsigned long long actual)
{
1200
    virQEMUDriverPtr driver = qemu_driver;
1201
    virDomainEventPtr event = NULL;
1202
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1203

1204
    virObjectLock(vm);
1205 1206 1207 1208 1209 1210
    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;

1211
    if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0)
1212 1213
        VIR_WARN("unable to save domain status with balloon change");

1214
    virObjectUnlock(vm);
1215

1216
    if (event)
1217
        qemuDomainEventQueue(driver, event);
1218
    virObjectUnref(cfg);
1219 1220 1221
    return 0;
}

1222 1223 1224 1225
static int
qemuProcessHandlePMSuspendDisk(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm)
{
1226
    virQEMUDriverPtr driver = qemu_driver;
1227 1228
    virDomainEventPtr event = NULL;
    virDomainEventPtr lifecycleEvent = NULL;
1229
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1230

1231
    virObjectLock(vm);
1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245
    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 =
            virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_DISK);

1246
        if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0) {
1247 1248 1249 1250 1251 1252 1253 1254
            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);
    }

1255
    virObjectUnlock(vm);
1256

1257 1258 1259 1260
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1261 1262
    virObjectUnref(cfg);

1263 1264 1265
    return 0;
}

1266

1267 1268 1269
static qemuMonitorCallbacks monitorCallbacks = {
    .destroy = qemuProcessHandleMonitorDestroy,
    .eofNotify = qemuProcessHandleMonitorEOF,
1270
    .errorNotify = qemuProcessHandleMonitorError,
1271 1272 1273
    .diskSecretLookup = qemuProcessFindVolumeQcowPassphrase,
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
1274
    .domainResume = qemuProcessHandleResume,
1275 1276 1277 1278 1279
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1280
    .domainBlockJob = qemuProcessHandleBlockJob,
1281
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1282
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1283
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1284
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1285
    .domainPMSuspendDisk = qemuProcessHandlePMSuspendDisk,
1286 1287 1288
};

static int
1289
qemuConnectMonitor(virQEMUDriverPtr driver, virDomainObjPtr vm)
1290 1291 1292
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
1293
    qemuMonitorPtr mon = NULL;
1294

1295
    if (virSecurityManagerSetDaemonSocketLabel(driver->securityManager,
1296
                                               vm->def) < 0) {
1297 1298 1299 1300 1301 1302 1303
        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
     * deleted while the monitor is active */
1304
    virObjectRef(vm);
1305

1306
    ignore_value(virTimeMillisNow(&priv->monStart));
1307
    virObjectUnlock(vm);
1308 1309 1310 1311 1312 1313

    mon = qemuMonitorOpen(vm,
                          priv->monConfig,
                          priv->monJSON,
                          &monitorCallbacks);

1314
    virObjectLock(vm);
1315
    priv->monStart = 0;
1316

1317
    if (mon == NULL) {
1318
        virObjectUnref(vm);
1319
    } else if (!virDomainObjIsActive(vm)) {
1320
        qemuMonitorClose(mon);
1321
        mon = NULL;
1322 1323 1324
    }
    priv->mon = mon;

1325
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0) {
1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336
        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;
    }


1337
    qemuDomainObjEnterMonitor(driver, vm);
1338 1339
    ret = qemuMonitorSetCapabilities(priv->mon);
    if (ret == 0 &&
1340 1341
        virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON))
        ret = virQEMUCapsProbeQMP(priv->qemuCaps, priv->mon);
1342
    qemuDomainObjExitMonitor(driver, vm);
1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366

error:

    return ret;
}

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;
1367 1368 1369
    int ret = -1;
    char *filter_next = buf;

1370 1371 1372
    buf[0] = '\0';

    while (retries) {
1373
        ssize_t func_ret, bytes;
1374
        int isdead = 0;
1375
        char *eol;
1376 1377 1378 1379 1380 1381 1382 1383

        func_ret = func(vm, buf, fd);

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

        /* Any failures should be detected before we read the log, so we
         * always have something useful to report on failure. */
1384 1385
        bytes = saferead(fd, buf+got, buflen-got-1);
        if (bytes < 0) {
1386 1387 1388
            virReportSystemError(errno,
                                 _("Failure while reading %s log output"),
                                 what);
1389
            goto cleanup;
1390 1391
        }

1392
        got += bytes;
1393
        buf[got] = '\0';
1394 1395 1396 1397

        /* Filter out debug messages from intermediate libvirt process */
        while ((eol = strchr(filter_next, '\n'))) {
            *eol = '\0';
1398
            if (virLogProbablyLogMessage(filter_next)) {
1399 1400 1401 1402 1403 1404 1405 1406
                memmove(filter_next, eol + 1, got - (eol - buf));
                got -= eol + 1 - filter_next;
            } else {
                filter_next = eol + 1;
                *eol = '\n';
            }
        }

1407
        if (got == buflen-1) {
1408 1409 1410
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Out of space while reading %s log output: %s"),
                           what, buf);
1411
            goto cleanup;
1412 1413 1414
        }

        if (isdead) {
1415 1416 1417
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Process exited while reading %s log output: %s"),
                           what, buf);
1418
            goto cleanup;
1419 1420
        }

1421 1422 1423 1424
        if (func_ret <= 0) {
            ret = func_ret;
            goto cleanup;
        }
1425 1426 1427 1428 1429

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

1430 1431 1432
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("Timed out while reading %s log output: %s"),
                   what, buf);
1433 1434 1435

cleanup:
    return ret;
1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451
}


/*
 * 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)
{
1452 1453
    static const char needle[] = "char device redirected to";
    char *tmp, *dev;
1454 1455 1456

    VIR_FREE(*path);
    /* First look for our magic string */
1457
    if (!(tmp = strstr(haystack + *offset, needle))) {
1458 1459
        return 1;
    }
1460
    tmp += sizeof(needle);
1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490
    dev = tmp;

    /*
     * And look for first whitespace character and nul terminate
     * to mark end of the pty path
     */
    while (*tmp) {
        if (c_isspace(*tmp)) {
            *path = strndup(dev, tmp-dev);
            if (*path == NULL) {
                virReportOOMError();
                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;
}

1491 1492 1493
static int
qemuProcessLookupPTYs(virDomainChrDefPtr *devices,
                      int count,
C
Cole Robinson 已提交
1494 1495
                      virHashTablePtr paths,
                      bool chardevfmt)
1496 1497
{
    int i;
C
Cole Robinson 已提交
1498
    const char *prefix = chardevfmt ? "char" : "";
1499 1500 1501 1502

    for (i = 0 ; i < count ; i++) {
        virDomainChrDefPtr chr = devices[i];
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
C
Cole Robinson 已提交
1503
            char id[32];
1504 1505
            const char *path;

C
Cole Robinson 已提交
1506 1507
            if (snprintf(id, sizeof(id), "%s%s",
                         prefix, chr->info.alias) >= sizeof(id))
1508 1509 1510 1511 1512 1513 1514 1515
                return -1;

            path = (const char *) virHashLookup(paths, id);
            if (path == NULL) {
                if (chr->source.data.file.path == NULL) {
                    /* neither the log output nor 'info chardev' had a
                     * pty path for this chardev, report an error
                     */
1516 1517
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539
                    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);
            chr->source.data.file.path = strdup(path);

            if (chr->source.data.file.path == NULL) {
                virReportOOMError();
                return -1;
            }
        }
    }

    return 0;
}

1540 1541
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
1542
                                     virQEMUCapsPtr qemuCaps,
1543 1544
                                     virHashTablePtr paths)
{
1545
    bool chardevfmt = virQEMUCapsGet(qemuCaps, QEMU_CAPS_CHARDEV);
1546
    int i = 0;
C
Cole Robinson 已提交
1547

1548
    if (qemuProcessLookupPTYs(vm->def->serials, vm->def->nserials,
C
Cole Robinson 已提交
1549
                              paths, chardevfmt) < 0)
1550 1551 1552
        return -1;

    if (qemuProcessLookupPTYs(vm->def->parallels, vm->def->nparallels,
C
Cole Robinson 已提交
1553
                              paths, chardevfmt) < 0)
1554
        return -1;
1555

1556
    if (qemuProcessLookupPTYs(vm->def->channels, vm->def->nchannels,
C
Cole Robinson 已提交
1557
                              paths, chardevfmt) < 0)
1558
        return -1;
1559 1560 1561 1562
    /* 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];
1563

1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575
        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;
        }
    }

    if (qemuProcessLookupPTYs(vm->def->consoles + i, vm->def->nconsoles - i,
1576
                              paths, chardevfmt) < 0)
1577
        return -1;
1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623

    return 0;
}

static int
qemuProcessFindCharDevicePTYs(virDomainObjPtr vm,
                              const char *output,
                              int fd ATTRIBUTE_UNUSED)
{
    size_t offset = 0;
    int ret, i;

    /* 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 */
    for (i = 0 ; i < vm->def->nserials ; i++) {
        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 */
    for (i = 0 ; i < vm->def->nparallels ; i++) {
        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 */
    for (i = 0 ; i < vm->def->nchannels ; i++) {
        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;
        }
    }

1624 1625
    for (i = 0 ; i < vm->def->nconsoles ; i++) {
        virDomainChrDefPtr chr = vm->def->consoles[i];
1626 1627 1628 1629 1630 1631 1632
        /* 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)
1633
                return ret;
1634 1635 1636 1637 1638 1639 1640
        } 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;
            }
1641 1642 1643
        }
    }

1644 1645 1646
    return 0;
}

1647
static void qemuProcessFreePtyPath(void *payload, const void *name ATTRIBUTE_UNUSED)
1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665
{
    VIR_FREE(payload);
}

static void
qemuProcessReadLogFD(int logfd, char *buf, int maxlen, int off)
{
    int ret;
    char *tmpbuf = buf + off;

    ret = saferead(logfd, tmpbuf, maxlen - off - 1);
    if (ret < 0) {
        ret = 0;
    }

    tmpbuf[ret] = '\0';
}

1666

1667
static int
1668
qemuProcessWaitForMonitor(virQEMUDriverPtr driver,
C
Cole Robinson 已提交
1669
                          virDomainObjPtr vm,
1670
                          virQEMUCapsPtr qemuCaps,
C
Cole Robinson 已提交
1671
                          off_t pos)
1672
{
1673
    char *buf = NULL;
1674
    size_t buf_size = 4096; /* Plenty of space to get startup greeting */
1675
    int logfd = -1;
1676 1677
    int ret = -1;
    virHashTablePtr paths = NULL;
1678
    qemuDomainObjPrivatePtr priv;
1679

1680
    if (!virQEMUCapsUsedQMP(qemuCaps) && pos != -1) {
1681 1682
        if ((logfd = qemuDomainOpenLog(driver, vm, pos)) < 0)
            return -1;
1683

1684 1685
        if (VIR_ALLOC_N(buf, buf_size) < 0) {
            virReportOOMError();
1686
            goto closelog;
1687
        }
1688

1689 1690 1691 1692 1693
        if (qemuProcessReadLogOutput(vm, logfd, buf, buf_size,
                                     qemuProcessFindCharDevicePTYs,
                                     "console", 30) < 0)
            goto closelog;
    }
1694 1695 1696 1697 1698 1699 1700 1701 1702 1703

    VIR_DEBUG("Connect monitor to %p '%s'", vm, vm->def->name);
    if (qemuConnectMonitor(driver, vm) < 0) {
        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. */
1704
    paths = virHashCreate(0, qemuProcessFreePtyPath);
1705
    if (paths == NULL)
1706 1707
        goto cleanup;

1708
    priv = vm->privateData;
1709
    qemuDomainObjEnterMonitor(driver, vm);
1710
    ret = qemuMonitorGetPtyPaths(priv->mon, paths);
1711
    qemuDomainObjExitMonitor(driver, vm);
1712 1713 1714

    VIR_DEBUG("qemuMonitorGetPtyPaths returned %i", ret);
    if (ret == 0)
1715
        ret = qemuProcessFindCharDevicePTYsMonitor(vm, qemuCaps, paths);
1716 1717

cleanup:
1718
    virHashFree(paths);
1719

E
Eric Blake 已提交
1720
    if (pos != -1 && kill(vm->pid, 0) == -1 && errno == ESRCH) {
1721 1722
        /* VM is dead, any other error raised in the interim is probably
         * not as important as the qemu cmdline output */
1723
        if (virQEMUCapsUsedQMP(qemuCaps)) {
1724 1725 1726 1727 1728 1729 1730 1731
            if ((logfd = qemuDomainOpenLog(driver, vm, pos)) < 0)
                return -1;

            if (VIR_ALLOC_N(buf, buf_size) < 0) {
                virReportOOMError();
                goto closelog;
            }
        }
1732
        qemuProcessReadLogFD(logfd, buf, buf_size, strlen(buf));
1733 1734 1735
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("process exited while connecting to monitor: %s"),
                       buf);
1736 1737 1738
        ret = -1;
    }

1739
closelog:
1740
    if (VIR_CLOSE(logfd) < 0) {
1741
        char ebuf[1024];
1742
        VIR_WARN("Unable to close logfile: %s",
1743
                 virStrerror(errno, ebuf, sizeof(ebuf)));
1744 1745
    }

1746 1747
    VIR_FREE(buf);

1748 1749 1750 1751
    return ret;
}

static int
1752
qemuProcessDetectVcpuPIDs(virQEMUDriverPtr driver,
1753 1754 1755 1756 1757 1758
                          virDomainObjPtr vm)
{
    pid_t *cpupids = NULL;
    int ncpupids;
    qemuDomainObjPrivatePtr priv = vm->privateData;

1759
    qemuDomainObjEnterMonitor(driver, vm);
1760 1761 1762 1763
    /* failure to get the VCPU<-> PID mapping or to execute the query
     * command will not be treated fatal as some versions of qemu don't
     * support this command */
    if ((ncpupids = qemuMonitorGetCPUInfo(priv->mon, &cpupids)) <= 0) {
1764
        qemuDomainObjExitMonitor(driver, vm);
1765 1766
        virResetLastError();

1767 1768 1769 1770 1771 1772 1773 1774
        priv->nvcpupids = 1;
        if (VIR_ALLOC_N(priv->vcpupids, priv->nvcpupids) < 0) {
            virReportOOMError();
            return -1;
        }
        priv->vcpupids[0] = vm->pid;
        return 0;
    }
1775
    qemuDomainObjExitMonitor(driver, vm);
1776 1777

    if (ncpupids != vm->def->vcpus) {
1778 1779 1780 1781
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of vCPU pids from QEMU monitor. "
                         "got %d, wanted %d"),
                       ncpupids, vm->def->vcpus);
1782 1783 1784 1785 1786 1787 1788 1789 1790
        VIR_FREE(cpupids);
        return -1;
    }

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

1791

1792 1793 1794
/* Helper to prepare cpumap for affinity setting, convert
 * NUMA nodeset into cpuset if @nodemask is not NULL, otherwise
 * just return a new allocated bitmap.
1795
 */
1796
virBitmapPtr
1797
qemuPrepareCpumap(virQEMUDriverPtr driver,
1798
                  virBitmapPtr nodemask)
1799 1800
{
    int i, hostcpus, maxcpu = QEMUD_CPUMASK_LEN;
1801
    virBitmapPtr cpumap = NULL;
1802
    virCapsPtr caps = NULL;
1803 1804 1805

    /* setaffinity fails if you set bits for CPUs which
     * aren't present, so we have to limit ourselves */
1806 1807 1808
    if ((hostcpus = nodeGetCPUCount()) < 0)
        return NULL;

1809 1810 1811
    if (maxcpu > hostcpus)
        maxcpu = hostcpus;

1812
    if (!(cpumap = virBitmapNew(maxcpu))) {
1813
        virReportOOMError();
1814
        return NULL;
1815 1816
    }

1817
    if (nodemask) {
1818 1819 1820 1821 1822 1823 1824
        if (!(caps = virQEMUDriverGetCapabilities(driver, false))) {
            virBitmapFree(cpumap);
            cpumap = NULL;
            goto cleanup;
        }

        for (i = 0; i < caps->host.nnumaCell; i++) {
1825
            int j;
1826
            int cur_ncpus = caps->host.numaCell[i]->ncpus;
1827
            bool result;
1828 1829
            if (virBitmapGetBit(nodemask, i, &result) < 0) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
E
Eric Blake 已提交
1830
                               _("Failed to convert nodeset to cpuset"));
1831
                virBitmapFree(cpumap);
1832 1833
                cpumap = NULL;
                goto cleanup;
1834
            }
1835
            if (result) {
1836
                for (j = 0; j < cur_ncpus; j++)
1837
                    ignore_value(virBitmapSetBit(cpumap,
1838
                                                 caps->host.numaCell[i]->cpus[j].id));
1839
            }
O
Osier Yang 已提交
1840
        }
1841 1842
    }

1843 1844
cleanup:
    virObjectUnref(caps);
1845 1846 1847 1848 1849 1850 1851
    return cpumap;
}

/*
 * To be run between fork/exec of QEMU only
 */
static int
1852
qemuProcessInitCpuAffinity(virQEMUDriverPtr driver,
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865
                           virDomainObjPtr vm,
                           virBitmapPtr nodemask)
{
    int ret = -1;
    virBitmapPtr cpumap = NULL;
    virBitmapPtr cpumapToSet = NULL;

    if (!(cpumap = qemuPrepareCpumap(driver, nodemask)))
        return -1;

    if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) {
        VIR_DEBUG("Set CPU affinity with advisory nodeset from numad");
        cpumapToSet = cpumap;
1866
    } else {
1867
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
1868
        if (vm->def->cpumask) {
H
Hu Tao 已提交
1869
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
1870
        } else {
1871
            cpumapToSet = cpumap;
O
Osier Yang 已提交
1872 1873 1874 1875
            /* 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 */
1876
            virBitmapSetAll(cpumap);
O
Osier Yang 已提交
1877
        }
1878 1879 1880 1881 1882 1883
    }

    /* We are pressuming we are running between fork/exec of QEMU
     * so use '0' to indicate our own process ID. No threads are
     * running at this point
     */
1884
    if (virProcessSetAffinity(0 /* Self */, cpumapToSet) < 0)
1885
        goto cleanup;
1886

1887 1888 1889
    ret = 0;

cleanup:
1890
    virBitmapFree(cpumap);
1891
    return ret;
1892 1893
}

1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906
/* set link states to down on interfaces at qemu start */
static int
qemuProcessSetLinkStates(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
    int i;
    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);

1907
            if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NETDEV)) {
1908
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
1909
                               _("Setting of link state is not supported by this qemu"));
1910 1911 1912 1913 1914 1915 1916
                return -1;
            }

            ret = qemuMonitorSetLink(priv->mon,
                                     def->nets[i]->info.alias,
                                     VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (ret != 0) {
1917
                virReportError(VIR_ERR_OPERATION_FAILED,
1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
                               _("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. */
1928
static int
1929
qemuProcessSetVcpuAffinities(virConnectPtr conn ATTRIBUTE_UNUSED,
1930 1931 1932 1933
                            virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
1934
    int vcpu, n;
E
Eric Blake 已提交
1935
    int ret = -1;
1936 1937 1938 1939 1940

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

    if (priv->vcpupids == NULL) {
1941 1942
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("cpu affinity is not supported"));
1943 1944 1945
        return -1;
    }

1946 1947
    for (n = 0; n < def->cputune.nvcpupin; n++) {
        vcpu = def->cputune.vcpupin[n]->vcpuid;
1948

1949 1950
        if (virProcessSetAffinity(priv->vcpupids[vcpu],
                                  def->cputune.vcpupin[n]->cpumask) < 0) {
E
Eric Blake 已提交
1951
            goto cleanup;
1952 1953 1954
        }
    }

E
Eric Blake 已提交
1955 1956 1957
    ret = 0;
cleanup:
    return ret;
1958
}
1959

1960
/* Set CPU affinities for emulator threads. */
1961
static int
1962
qemuProcessSetEmulatorAffinities(virConnectPtr conn ATTRIBUTE_UNUSED,
1963 1964
                                virDomainObjPtr vm)
{
1965
    virBitmapPtr cpumask;
1966 1967 1968
    virDomainDefPtr def = vm->def;
    int ret = -1;

O
Osier Yang 已提交
1969
    if (def->cputune.emulatorpin)
1970
        cpumask = def->cputune.emulatorpin->cpumask;
O
Osier Yang 已提交
1971
    else if (def->cpumask)
1972
        cpumask = def->cpumask;
O
Osier Yang 已提交
1973 1974
    else
        return 0;
1975

1976
    ret = virProcessSetAffinity(vm->pid, cpumask);
1977 1978 1979
    return ret;
}

1980 1981
static int
qemuProcessInitPasswords(virConnectPtr conn,
1982
                         virQEMUDriverPtr driver,
1983
                         virDomainObjPtr vm)
1984 1985 1986
{
    int ret = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
1987
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
E
Eric Blake 已提交
1988
    int i;
1989

E
Eric Blake 已提交
1990
    for (i = 0 ; i < vm->def->ngraphics; ++i) {
1991 1992
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
1993 1994
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_VNC,
1995
                                                    &graphics->data.vnc.auth,
1996
                                                    cfg->vncPassword);
1997
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
1998 1999
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_SPICE,
2000
                                                    &graphics->data.spice.auth,
2001
                                                    cfg->spicePassword);
2002 2003 2004 2005 2006 2007
        }
    }

    if (ret < 0)
        goto cleanup;

2008
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
2009 2010 2011
        for (i = 0 ; i < vm->def->ndisks ; i++) {
            char *secret;
            size_t secretLen;
E
Eric Blake 已提交
2012
            const char *alias;
2013 2014 2015 2016 2017 2018 2019 2020 2021 2022

            if (!vm->def->disks[i]->encryption ||
                !vm->def->disks[i]->src)
                continue;

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

E
Eric Blake 已提交
2023
            alias = vm->def->disks[i]->info.alias;
2024
            qemuDomainObjEnterMonitor(driver, vm);
E
Eric Blake 已提交
2025
            ret = qemuMonitorSetDrivePassphrase(priv->mon, alias, secret);
2026
            VIR_FREE(secret);
2027
            qemuDomainObjExitMonitor(driver, vm);
2028 2029 2030 2031 2032 2033
            if (ret < 0)
                goto cleanup;
        }
    }

cleanup:
2034
    virObjectUnref(cfg);
2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 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 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306
    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)
{
    int found = 0;
    int i;

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

    for (i = 0 ; (i < naddrs) && !found; i++) {
        VIR_DEBUG("Maybe %x:%x", addrs[i].vendor, addrs[i].product);
        if (addrs[i].vendor == vendor &&
            addrs[i].product == product) {
            VIR_DEBUG("Match %d", i);
            found = 1;
            break;
        }
    }
    if (!found) {
        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;
    int i;

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

    for (i = 0 ; i < vm->def->ndisks ; i++) {
        if (qemuProcessGetPCIDiskVendorProduct(vm->def->disks[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->disks[i]->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2307 2308 2309
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for VirtIO disk %s"),
                           vm->def->disks[i]->dst);
2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320
            return -1;
        }
    }

    for (i = 0 ; i < vm->def->nnets ; i++) {
        if (qemuProcessGetPCINetVendorProduct(vm->def->nets[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->nets[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2321 2322 2323
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for %s NIC"),
                           vm->def->nets[i]->model);
2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334
            return -1;
        }
    }

    for (i = 0 ; i < vm->def->ncontrollers ; i++) {
        if (qemuProcessGetPCIControllerVendorProduct(vm->def->controllers[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->controllers[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2335 2336 2337
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for controller %s"),
                           virDomainControllerTypeToString(vm->def->controllers[i]->type));
2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348
            return -1;
        }
    }

    for (i = 0 ; i < vm->def->nvideos ; i++) {
        if (qemuProcessGetPCIVideoVendorProduct(vm->def->videos[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->videos[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2349 2350 2351
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for video adapter %s"),
                           virDomainVideoTypeToString(vm->def->videos[i]->type));
2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362
            return -1;
        }
    }

    for (i = 0 ; i < vm->def->nsounds ; i++) {
        if (qemuProcessGetPCISoundVendorProduct(vm->def->sounds[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->sounds[i]->info),
                                    vendor, product,
                                     addrs,  naddrs) < 0) {
2363 2364 2365
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for sound adapter %s"),
                           virDomainSoundModelTypeToString(vm->def->sounds[i]->model));
2366 2367 2368 2369 2370 2371 2372 2373 2374 2375
            return -1;
        }
    }


    if (vm->def->watchdog &&
        qemuProcessGetPCIWatchdogVendorProduct(vm->def->watchdog, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->watchdog->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2376 2377 2378
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for watchdog %s"),
                           virDomainWatchdogModelTypeToString(vm->def->watchdog->model));
2379 2380 2381 2382 2383 2384 2385 2386 2387
            return -1;
        }
    }

    if (vm->def->memballoon &&
        qemuProcessGetPCIMemballoonVendorProduct(vm->def->memballoon, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->memballoon->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2388 2389 2390
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for balloon %s"),
                           virDomainMemballoonModelTypeToString(vm->def->memballoon->model));
2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407
            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
2408
qemuProcessInitPCIAddresses(virQEMUDriverPtr driver,
2409 2410 2411 2412 2413 2414 2415
                            virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int naddrs;
    int ret;
    qemuMonitorPCIAddress *addrs = NULL;

2416
    qemuDomainObjEnterMonitor(driver, vm);
2417 2418
    naddrs = qemuMonitorGetAllPCIAddresses(priv->mon,
                                           &addrs);
2419
    qemuDomainObjExitMonitor(driver, vm);
2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451

    ret = qemuProcessDetectPCIAddresses(vm, addrs, naddrs);

    VIR_FREE(addrs);

    return ret;
}


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

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

    VIR_FORCE_CLOSE(fd);

    return 0;
}


2452
static int
2453
qemuProcessLimits(virQEMUDriverConfigPtr cfg)
2454
{
2455
    struct rlimit rlim;
2456

2457 2458
    if (cfg->maxProcesses > 0) {
        rlim.rlim_cur = rlim.rlim_max = cfg->maxProcesses;
2459 2460 2461
        if (setrlimit(RLIMIT_NPROC, &rlim) < 0) {
            virReportSystemError(errno,
                                 _("cannot limit number of processes to %d"),
2462
                                 cfg->maxProcesses);
2463 2464 2465 2466
            return -1;
        }
    }

2467
    if (cfg->maxFiles > 0) {
2468 2469
        /* Max number of opened files is one greater than
         * actual limit. See man setrlimit */
2470
        rlim.rlim_cur = rlim.rlim_max = cfg->maxFiles + 1;
2471 2472 2473
        if (setrlimit(RLIMIT_NOFILE, &rlim) < 0) {
            virReportSystemError(errno,
                                 _("cannot set max opened files to %d"),
2474
                                 cfg->maxFiles);
2475 2476 2477 2478
            return -1;
        }
    }

2479 2480 2481 2482
    return 0;
}


2483 2484 2485
struct qemuProcessHookData {
    virConnectPtr conn;
    virDomainObjPtr vm;
2486
    virQEMUDriverPtr driver;
2487
    virBitmapPtr nodemask;
2488
    virQEMUDriverConfigPtr cfg;
2489 2490 2491 2492 2493
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
2494
    int ret = -1;
2495
    int fd;
2496 2497 2498 2499 2500
    /* This method cannot use any mutexes, which are not
     * protected across fork()
     */

    virSecurityManagerPostFork(h->driver->securityManager);
2501 2502 2503 2504 2505

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

    VIR_DEBUG("Obtaining domain lock");
2506 2507 2508 2509 2510 2511 2512
    /*
     * 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.
     */
2513
    if (virSecurityManagerSetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2514
        goto cleanup;
2515
    if (virDomainLockProcessStart(h->driver->lockManager,
2516
                                  h->cfg->uri,
2517
                                  h->vm,
J
Ján Tomko 已提交
2518
                                  /* QEMU is always paused initially */
2519 2520
                                  true,
                                  &fd) < 0)
2521
        goto cleanup;
2522
    if (virSecurityManagerClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2523
        goto cleanup;
2524

2525
    if (qemuProcessLimits(h->cfg) < 0)
2526
        goto cleanup;
2527

2528 2529 2530
    /* This must take place before exec(), so that all QEMU
     * memory allocation is on the correct NUMA node
     */
2531
    VIR_DEBUG("Moving process to cgroup");
2532
    if (qemuAddToCgroup(h->driver, h->vm->def) < 0)
2533
        goto cleanup;
2534 2535 2536

    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
2537 2538
    if (!h->vm->def->cputune.emulatorpin &&
        qemuProcessInitCpuAffinity(h->driver, h->vm, h->nodemask) < 0)
2539
        goto cleanup;
2540

G
Gao feng 已提交
2541
    if (virNumaSetupMemoryPolicy(h->vm->def->numatune, h->nodemask) < 0)
2542
        goto cleanup;
2543

2544 2545 2546
    ret = 0;

cleanup:
2547
    virObjectUnref(h->cfg);
2548 2549
    VIR_DEBUG("Hook complete ret=%d", ret);
    return ret;
2550 2551 2552
}

int
2553
qemuProcessPrepareMonitorChr(virQEMUDriverConfigPtr cfg,
2554 2555 2556 2557 2558 2559 2560
                             virDomainChrSourceDefPtr monConfig,
                             const char *vm)
{
    monConfig->type = VIR_DOMAIN_CHR_TYPE_UNIX;
    monConfig->data.nix.listen = true;

    if (virAsprintf(&monConfig->data.nix.path, "%s/%s.monitor",
2561
                    cfg->libDir, vm) < 0) {
2562 2563 2564 2565 2566 2567 2568 2569
        virReportOOMError();
        return -1;
    }

    return 0;
}


2570
/*
2571 2572
 * Precondition: vm must be locked, and a job must be active.
 * This method will call {Enter,Exit}Monitor
2573
 */
E
Eric Blake 已提交
2574
int
2575
qemuProcessStartCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
2576 2577
                     virConnectPtr conn, virDomainRunningReason reason,
                     enum qemuDomainAsyncJob asyncJob)
2578
{
2579
    int ret = -1;
2580
    qemuDomainObjPrivatePtr priv = vm->privateData;
2581
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2582

2583
    VIR_DEBUG("Using lock state '%s'", NULLSTR(priv->lockState));
2584
    if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
2585
                                   vm, priv->lockState) < 0) {
2586 2587 2588 2589
        /* 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
         */
2590
        goto cleanup;
2591 2592 2593
    }
    VIR_FREE(priv->lockState);

2594 2595 2596
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
    if (ret == 0) {
        ret = qemuMonitorStartCPUs(priv->mon, conn);
2597
        qemuDomainObjExitMonitor(driver, vm);
2598
    }
J
Jiri Denemark 已提交
2599

2600
    if (ret == 0) {
J
Jiri Denemark 已提交
2601
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
2602 2603 2604 2605 2606
    } else {
        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));
    }
2607

2608 2609
cleanup:
    virObjectUnref(cfg);
2610 2611 2612 2613
    return ret;
}


2614
int qemuProcessStopCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
2615 2616
                        virDomainPausedReason reason,
                        enum qemuDomainAsyncJob asyncJob)
2617 2618 2619 2620
{
    int ret;
    qemuDomainObjPrivatePtr priv = vm->privateData;

2621
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
2622

2623
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
2624 2625
    if (ret == 0) {
        ret = qemuMonitorStopCPUs(priv->mon);
2626
        qemuDomainObjExitMonitor(driver, vm);
2627
    }
J
Jiri Denemark 已提交
2628

2629
    if (ret == 0) {
2630
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
2631 2632 2633 2634
        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
2635

2636 2637 2638 2639 2640
    return ret;
}



2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653
static int
qemuProcessNotifyNets(virDomainDefPtr def)
{
    int ii;

    for (ii = 0 ; ii < def->nnets ; ii++) {
        virDomainNetDefPtr net = def->nets[ii];
        if (networkNotifyActualDevice(net) < 0)
            return -1;
    }
    return 0;
}

2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666
static int
qemuProcessFiltersInstantiate(virConnectPtr conn,
                              virDomainDefPtr def)
{
    int err = 0;
    int i;

    if (!conn)
        return 1;

    for (i = 0 ; i < def->nnets ; i++) {
        virDomainNetDefPtr net = def->nets[i];
        if ((net->filter) && (net->ifname)) {
2667
           if (virDomainConfNWFilterInstantiate(conn, def->uuid, net) < 0) {
2668 2669 2670 2671 2672 2673 2674 2675 2676
                err = 1;
                break;
            }
        }
    }

    return err;
}

2677
static int
2678
qemuProcessUpdateState(virQEMUDriverPtr driver, virDomainObjPtr vm)
2679 2680 2681
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
2682
    virDomainPausedReason reason;
2683 2684
    virDomainState newState = VIR_DOMAIN_NOSTATE;
    int newReason;
2685
    bool running;
2686
    char *msg = NULL;
2687 2688
    int ret;

2689
    qemuDomainObjEnterMonitor(driver, vm);
2690
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
2691
    qemuDomainObjExitMonitor(driver, vm);
2692 2693 2694 2695 2696 2697 2698

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

    state = virDomainObjGetState(vm, NULL);

    if (state == VIR_DOMAIN_PAUSED && running) {
2699 2700 2701
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
        msg = strdup("was unpaused");
2702
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
2703 2704 2705 2706 2707 2708 2709
        if (reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN) {
            newState = VIR_DOMAIN_SHUTDOWN;
            newReason = VIR_DOMAIN_SHUTDOWN_UNKNOWN;
            msg = strdup("shutdown");
        } else {
            newState = VIR_DOMAIN_PAUSED;
            newReason = reason;
S
Stefan Berger 已提交
2710 2711
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
2712
        }
2713
    } else if (state == VIR_DOMAIN_SHUTOFF && running) {
2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_BOOTED;
        msg = strdup("finished booting");
    }

    if (newState != VIR_DOMAIN_NOSTATE) {
        if (!msg) {
            virReportOOMError();
            return -1;
        }

        VIR_DEBUG("Domain %s %s while its monitor was disconnected;"
                  " changing state to %s (%s)",
                  vm->def->name,
                  msg,
                  virDomainStateTypeToString(newState),
                  virDomainStateReasonToString(newState, newReason));
        VIR_FREE(msg);
        virDomainObjSetState(vm, newState, newReason);
2733 2734 2735 2736 2737
    }

    return 0;
}

2738
static int
2739
qemuProcessRecoverMigration(virQEMUDriverPtr driver,
2740 2741 2742 2743 2744 2745 2746
                            virDomainObjPtr vm,
                            virConnectPtr conn,
                            enum qemuDomainAsyncJob job,
                            enum qemuMigrationJobPhase phase,
                            virDomainState state,
                            int reason)
{
2747 2748
    qemuDomainObjPrivatePtr priv = vm->privateData;

2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771
    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,
2772 2773
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
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
                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);
2804
            qemuDomainObjEnterMonitor(driver, vm);
2805 2806
            ignore_value(qemuMonitorMigrateCancel(priv->mon));
            qemuDomainObjExitMonitor(driver, vm);
2807 2808 2809 2810 2811 2812
            /* 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,
2813 2814
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833
                    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,
2834 2835
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849
                    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;
}

2850
static int
2851
qemuProcessRecoverJob(virQEMUDriverPtr driver,
2852 2853 2854 2855
                      virDomainObjPtr vm,
                      virConnectPtr conn,
                      const struct qemuDomainJobObj *job)
{
2856
    qemuDomainObjPrivatePtr priv = vm->privateData;
2857 2858 2859 2860 2861 2862 2863 2864
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
    case QEMU_ASYNC_JOB_MIGRATION_IN:
2865 2866 2867
        if (qemuProcessRecoverMigration(driver, vm, conn, job->asyncJob,
                                        job->phase, state, reason) < 0)
            return -1;
2868 2869 2870 2871
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
2872
    case QEMU_ASYNC_JOB_SNAPSHOT:
2873
        qemuDomainObjEnterMonitor(driver, vm);
2874 2875
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
        qemuDomainObjExitMonitor(driver, vm);
2876
        /* resume the domain but only if it was paused as a result of
2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892
         * 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);
2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904
            }
        }
        break;

    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        break;
    }

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

2905 2906 2907 2908
    /* 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.
     */
2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928
    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;

2929
    case QEMU_JOB_MIGRATION_OP:
2930
    case QEMU_JOB_ABORT:
2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941
    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;
}

2942 2943
struct qemuProcessReconnectData {
    virConnectPtr conn;
2944
    virQEMUDriverPtr driver;
2945 2946
    void *payload;
    struct qemuDomainJobObj oldjob;
2947 2948 2949 2950
};
/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
S
Stefan Berger 已提交
2951 2952 2953 2954
 *
 * 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.
2955 2956
 */
static void
2957
qemuProcessReconnect(void *opaque)
2958 2959
{
    struct qemuProcessReconnectData *data = opaque;
2960
    virQEMUDriverPtr driver = data->driver;
2961
    virDomainObjPtr obj = data->payload;
2962 2963
    qemuDomainObjPrivatePtr priv;
    virConnectPtr conn = data->conn;
2964
    struct qemuDomainJobObj oldjob;
2965 2966
    int state;
    int reason;
2967
    virQEMUDriverConfigPtr cfg;
2968
    size_t i;
2969

2970 2971 2972 2973
    memcpy(&oldjob, &data->oldjob, sizeof(oldjob));

    VIR_FREE(data);

2974
    virObjectLock(obj);
2975

2976
    cfg = virQEMUDriverGetConfig(driver);
2977 2978 2979 2980
    VIR_DEBUG("Reconnect monitor to %p '%s'", obj, obj->def->name);

    priv = obj->privateData;

2981 2982
    /* Job was started by the caller for us */
    qemuDomainObjTransferJob(obj);
2983

2984 2985
    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted if qemuConnectMonitor() failed */
2986
    virObjectRef(obj);
2987 2988 2989 2990 2991

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

D
Daniel P. Berrange 已提交
2992 2993 2994 2995 2996 2997 2998 2999
    /* Failure to connect to agent shouldn't be fatal */
    if (qemuConnectAgent(driver, obj) < 0) {
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 obj->def->name);
        virResetLastError();
        priv->agentError = true;
    }

3000 3001 3002 3003
    if (qemuUpdateActivePciHostdevs(driver, obj->def) < 0) {
        goto error;
    }

3004 3005 3006
    if (qemuUpdateActiveUsbHostdevs(driver, obj->def) < 0)
        goto error;

3007 3008 3009 3010 3011 3012 3013 3014 3015
    /* XXX: Need to change as long as lock is introduced for
     * qemu_driver->sharedDisks.
     */
    for (i = 0; i < obj->def->ndisks; i++) {
        if (qemuAddSharedDisk(driver, obj->def->disks[i],
                              obj->def->name) < 0)
            goto error;
    }

3016 3017 3018
    if (qemuProcessUpdateState(driver, obj) < 0)
        goto error;

3019 3020
    state = virDomainObjGetState(obj, &reason);
    if (state == VIR_DOMAIN_SHUTOFF) {
3021 3022 3023 3024 3025
        VIR_DEBUG("Domain '%s' wasn't fully started yet, killing it",
                  obj->def->name);
        goto error;
    }

3026 3027 3028
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
3029 3030 3031
    if (!priv->qemuCaps &&
        !(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
                                                      obj->def->emulator)))
3032 3033
        goto error;

3034
    /* In case the domain shutdown while we were not running,
3035
     * we need to finish the shutdown process. And we need to do it after
3036
     * we have virQEMUCaps filled in.
3037
     */
3038 3039 3040 3041 3042 3043
    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);
3044 3045 3046
        goto endjob;
    }

3047 3048
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE))
        if ((qemuDomainAssignAddresses(obj->def, priv->qemuCaps, obj)) < 0)
3049
            goto error;
3050

3051
    if (virSecurityManagerReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
3052 3053
        goto error;

3054 3055 3056
    if (qemuProcessNotifyNets(obj->def) < 0)
        goto error;

3057 3058 3059
    if (qemuProcessFiltersInstantiate(conn, obj->def))
        goto error;

3060
    if (qemuDomainCheckEjectableMedia(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
3061 3062
        goto error;

3063 3064 3065
    if (qemuProcessRecoverJob(driver, obj, conn, &oldjob) < 0)
        goto error;

3066
    /* update domain state XML with possibly updated state in virDomainObj */
3067
    if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, obj) < 0)
3068 3069
        goto error;

3070 3071
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3072
        char *xml = qemuDomainDefFormatXML(driver, obj->def, 0);
3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086
        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;
    }

3087
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
3088 3089
        driver->inhibitCallback(true, driver->inhibitOpaque);

3090
endjob:
3091
    if (!qemuDomainObjEndJob(driver, obj))
3092 3093
        obj = NULL;

3094
    if (obj && virObjectUnref(obj))
3095
        virObjectUnlock(obj);
3096

S
Stefan Berger 已提交
3097
    virConnectClose(conn);
3098
    virObjectUnref(cfg);
S
Stefan Berger 已提交
3099

3100 3101 3102
    return;

error:
3103
    if (!qemuDomainObjEndJob(driver, obj))
3104 3105
        obj = NULL;

3106 3107
    if (obj) {
        if (!virDomainObjIsActive(obj)) {
3108
            if (virObjectUnref(obj))
3109
                virObjectUnlock(obj);
3110 3111
            return;
        }
3112

3113
        if (virObjectUnref(obj)) {
3114
            /* We can't get the monitor back, so must kill the VM
3115 3116 3117
             * to remove danger of it ending up running twice if
             * user tries to start it again later
             */
3118
            if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NO_SHUTDOWN)) {
3119 3120 3121 3122 3123 3124 3125 3126 3127
                /* 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;
            }
3128
            qemuProcessStop(driver, obj, state, 0);
3129
            if (!obj->persistent)
3130
                qemuDomainRemoveInactive(driver, obj);
3131
            else
3132
                virObjectUnlock(obj);
3133
        }
3134
    }
S
Stefan Berger 已提交
3135
    virConnectClose(conn);
3136
    virObjectUnref(cfg);
3137 3138
}

3139 3140
static int
qemuProcessReconnectHelper(virDomainObjPtr obj,
3141 3142 3143 3144 3145 3146 3147 3148
                           void *opaque)
{
    virThread thread;
    struct qemuProcessReconnectData *src = opaque;
    struct qemuProcessReconnectData *data;

    if (VIR_ALLOC(data) < 0) {
        virReportOOMError();
3149
        return -1;
3150 3151 3152
    }

    memcpy(data, src, sizeof(*data));
3153
    data->payload = obj;
3154

3155
    /*
3156 3157
     * We create a separate thread to run qemuProcessReconnect in it.
     * However, qemuProcessReconnect needs to:
3158
     * 1. just before monitor reconnect do lightweight MonitorEnter
3159
     *    (increase VM refcount, unlock VM & driver)
3160 3161 3162 3163
     * 2. reconnect to monitor
     * 3. do lightweight MonitorExit (lock VM)
     * 4. continue reconnect process
     * 5. EndJob
3164 3165 3166 3167 3168 3169
     *
     * NB, we can't do normal MonitorEnter & MonitorExit because
     * these two lock the monitor lock, which does not exists in
     * this early phase.
     */

3170
    virObjectLock(obj);
3171 3172 3173

    qemuDomainObjRestoreJob(obj, &data->oldjob);

3174
    if (qemuDomainObjBeginJob(src->driver, obj, QEMU_JOB_MODIFY) < 0)
3175 3176
        goto error;

S
Stefan Berger 已提交
3177 3178 3179 3180 3181 3182
    /* 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.
     */
    virConnectRef(data->conn);

3183
    if (virThreadCreate(&thread, true, qemuProcessReconnect, data) < 0) {
S
Stefan Berger 已提交
3184 3185 3186

        virConnectClose(data->conn);

3187 3188 3189
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not create thread. QEMU initialization "
                         "might be incomplete"));
3190
        if (!qemuDomainObjEndJob(src->driver, obj)) {
3191
            obj = NULL;
3192
        } else if (virObjectUnref(obj)) {
3193 3194
           /* We can't spawn a thread and thus connect to monitor.
            * Kill qemu */
3195
            qemuProcessStop(src->driver, obj, VIR_DOMAIN_SHUTOFF_FAILED, 0);
3196
            if (!obj->persistent)
3197
                qemuDomainRemoveInactive(src->driver, obj);
3198
            else
3199
                virObjectUnlock(obj);
3200 3201 3202 3203
        }
        goto error;
    }

3204
    virObjectUnlock(obj);
3205

3206
    return 0;
3207 3208 3209

error:
    VIR_FREE(data);
3210
    return -1;
3211 3212 3213 3214 3215 3216 3217 3218 3219
}

/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
3220
qemuProcessReconnectAll(virConnectPtr conn, virQEMUDriverPtr driver)
3221
{
3222
    struct qemuProcessReconnectData data = {.conn = conn, .driver = driver};
3223
    virDomainObjListForEach(driver->domains, qemuProcessReconnectHelper, &data);
3224 3225
}

3226 3227 3228
int
qemuSetUnprivSGIO(virDomainDiskDefPtr disk)
{
3229
    char *sysfs_path = NULL;
3230
    int val = -1;
3231
    int ret = 0;
3232

3233 3234 3235 3236
    /* "sgio" is only valid for block disk; cdrom
     * and floopy disk can have empty source.
     */
    if (disk->type != VIR_DOMAIN_DISK_TYPE_BLOCK ||
3237
        disk->device != VIR_DOMAIN_DISK_DEVICE_LUN ||
3238 3239 3240
        !disk->src)
        return 0;

3241 3242 3243
    sysfs_path = virGetUnprivSGIOSysfsPath(disk->src, NULL);
    if (sysfs_path == NULL)
        return -1;
3244

3245 3246
    /* By default, filter the SG_IO commands, i.e. set unpriv_sgio to 0.  */
    val = (disk->sgio == VIR_DOMAIN_DISK_SGIO_UNFILTERED);
3247

3248 3249 3250 3251 3252 3253 3254
    /* Do not do anything if unpriv_sgio is not supported by the kernel and the
     * whitelist is enabled.  But if requesting unfiltered access, always call
     * virSetDeviceUnprivSGIO, to report an error for unsupported unpriv_sgio.
     */
    if ((virFileExists(sysfs_path) || val == 1) &&
        virSetDeviceUnprivSGIO(disk->src, NULL, val) < 0)
        ret = -1;
3255

3256 3257
    VIR_FREE(sysfs_path);
    return ret;
3258 3259
}

3260
int qemuProcessStart(virConnectPtr conn,
3261
                     virQEMUDriverPtr driver,
3262 3263 3264 3265
                     virDomainObjPtr vm,
                     const char *migrateFrom,
                     int stdin_fd,
                     const char *stdin_path,
3266
                     virDomainSnapshotObjPtr snapshot,
3267 3268
                     enum virNetDevVPortProfileOp vmop,
                     unsigned int flags)
3269 3270 3271 3272 3273 3274 3275 3276 3277
{
    int ret;
    off_t pos = -1;
    char ebuf[1024];
    int logfile = -1;
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCommandPtr cmd = NULL;
    struct qemuProcessHookData hookData;
E
Eric Blake 已提交
3278
    unsigned long cur_balloon;
3279
    int i;
3280
    char *nodeset = NULL;
3281
    virBitmapPtr nodemask = NULL;
3282
    unsigned int stop_flags;
3283
    virQEMUDriverConfigPtr cfg;
3284
    virCapsPtr caps = NULL;
3285

3286 3287 3288 3289
    /* Okay, these are just internal flags,
     * but doesn't hurt to check */
    virCheckFlags(VIR_QEMU_PROCESS_START_COLD |
                  VIR_QEMU_PROCESS_START_PAUSED |
3290
                  VIR_QEMU_PROCESS_START_AUTODESTROY, -1);
3291

3292 3293
    cfg = virQEMUDriverGetConfig(driver);

3294 3295 3296 3297 3298 3299
    /* 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;

3300 3301 3302
    hookData.conn = conn;
    hookData.vm = vm;
    hookData.driver = driver;
3303
    hookData.cfg = virObjectRef(cfg);
3304

3305
    VIR_DEBUG("Beginning VM startup process");
3306 3307

    if (virDomainObjIsActive(vm)) {
3308 3309
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
3310
        virObjectUnref(cfg);
3311 3312 3313
        return -1;
    }

3314 3315 3316
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

3317 3318 3319 3320
    /* 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
     */
3321
    VIR_DEBUG("Setting current domain def as transient");
3322
    if (virDomainObjSetDefTransient(caps, driver->xmlconf, vm, true) < 0)
3323 3324
        goto cleanup;

3325
    vm->def->id = qemuDriverAllocateID(driver);
3326
    qemuDomainSetFakeReboot(driver, vm, false);
3327
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, VIR_DOMAIN_SHUTOFF_UNKNOWN);
3328

3329
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
3330 3331
        driver->inhibitCallback(true, driver->inhibitOpaque);

3332
    /* Run an early hook to set-up missing devices */
3333
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3334
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
3335 3336 3337
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
3338 3339
                              VIR_HOOK_QEMU_OP_PREPARE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
3340 3341 3342 3343 3344 3345 3346 3347 3348
        VIR_FREE(xml);

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

3349
    /* Must be run before security labelling */
3350
    VIR_DEBUG("Preparing host devices");
3351
    if (qemuPrepareHostDevices(driver, vm->def, !migrateFrom) < 0)
3352 3353
        goto cleanup;

3354
    VIR_DEBUG("Preparing chr devices");
3355 3356 3357 3358 3359 3360 3361 3362
    if (virDomainChrDefForeach(vm->def,
                               true,
                               qemuProcessPrepareChardevDevice,
                               NULL) < 0)
        goto cleanup;

    /* If you are using a SecurityDriver with dynamic labelling,
       then generate a security label for isolation */
3363
    VIR_DEBUG("Generating domain security label (if required)");
3364
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0) {
3365
        virDomainAuditSecurityLabel(vm, false);
3366 3367
        goto cleanup;
    }
3368
    virDomainAuditSecurityLabel(vm, true);
3369

3370
    if (cfg->hugepagePath && vm->def->mem.hugepage_backed) {
3371
        if (virSecurityManagerSetHugepages(driver->securityManager,
3372
                                           vm->def, cfg->hugepagePath) < 0) {
3373 3374 3375 3376 3377 3378
            virReportError(VIR_ERR_INTERNAL_ERROR,
                    "%s", _("Unable to set huge path in security driver"));
            goto cleanup;
        }
    }

3379 3380
    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
3381
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
3382 3383
    qemuRemoveCgroup(driver, vm, 1);

3384 3385 3386 3387 3388
    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.socket &&
            graphics->data.vnc.autoport) {
3389 3390
            unsigned short port;
            if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
3391
                goto cleanup;
3392 3393
            graphics->data.vnc.port = port;
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
3394
            unsigned short port = 0;
3395 3396
            if (graphics->data.spice.autoport ||
                graphics->data.spice.port == -1) {
3397 3398
                if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
                    goto cleanup;
M
Michal Privoznik 已提交
3399

3400
                if (port == 0) {
3401
                    virReportError(VIR_ERR_INTERNAL_ERROR,
3402
                                   "%s", _("Unable to find an unused port for SPICE"));
M
Michal Privoznik 已提交
3403 3404 3405
                    goto cleanup;
                }

3406
                graphics->data.spice.port = port;
3407
            }
3408
            if (cfg->spiceTLS &&
3409 3410
                (graphics->data.spice.autoport ||
                 graphics->data.spice.tlsPort == -1)) {
3411 3412 3413 3414 3415
                unsigned short tlsPort;
                if (virPortAllocatorAcquire(driver->remotePorts, &tlsPort) < 0)
                    goto cleanup;

                if (tlsPort == 0) {
3416
                    virReportError(VIR_ERR_INTERNAL_ERROR,
3417
                                   "%s", _("Unable to find an unused port for SPICE TLS"));
3418
                    virPortAllocatorRelease(driver->remotePorts, port);
3419 3420 3421
                    goto cleanup;
                }

3422
                graphics->data.spice.tlsPort = tlsPort;
M
Michal Privoznik 已提交
3423
            }
3424
        }
3425

3426 3427
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ||
            graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
3428 3429 3430 3431 3432 3433
            if (graphics->nListens == 0) {
                if (VIR_EXPAND_N(graphics->listens, graphics->nListens, 1) < 0) {
                    virReportOOMError();
                    goto cleanup;
                }
                graphics->listens[0].type = VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_ADDRESS;
3434
                if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC)
3435
                    graphics->listens[0].address = strdup(cfg->vncListen);
3436
                else
3437
                    graphics->listens[0].address = strdup(cfg->spiceListen);
3438 3439 3440 3441 3442 3443 3444
                if (!graphics->listens[0].address) {
                    VIR_SHRINK_N(graphics->listens, graphics->nListens, 1);
                    virReportOOMError();
                    goto cleanup;
                }
            }
        }
3445 3446
    }

3447
    if (virFileMakePath(cfg->logDir) < 0) {
3448 3449
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
3450
                             cfg->logDir);
3451 3452 3453
        goto cleanup;
    }

3454
    VIR_DEBUG("Creating domain log file");
3455
    if ((logfile = qemuDomainCreateLog(driver, vm, false)) < 0)
3456 3457
        goto cleanup;

3458 3459 3460
    if (vm->def->virtType == VIR_DOMAIN_VIRT_KVM) {
        VIR_DEBUG("Checking for KVM availability");
        if (access("/dev/kvm", F_OK) != 0) {
3461 3462 3463 3464
            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."));
3465 3466 3467 3468
            goto cleanup;
        }
    }

3469
    VIR_DEBUG("Determining emulator version");
3470 3471 3472
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
                                                      vm->def->emulator)))
3473 3474
        goto cleanup;

3475
    if (qemuAssignDeviceAliases(vm->def, priv->qemuCaps) < 0)
3476 3477 3478
        goto cleanup;

    VIR_DEBUG("Checking for CDROM and floppy presence");
3479 3480 3481 3482
    if (qemuDomainCheckDiskPresence(driver, vm,
                                    flags & VIR_QEMU_PROCESS_START_COLD) < 0)
        goto cleanup;

3483 3484 3485 3486 3487
    for (i = 0; i < vm->def->ndisks ; i++) {
        if (qemuDomainDetermineDiskChain(driver, vm->def->disks[i],
                                         false) < 0)
            goto cleanup;
    }
3488

3489 3490 3491 3492 3493 3494
    /* Get the advisory nodeset from numad if 'placement' of
     * either <vcpu> or <numatune> is 'auto'.
     */
    if ((vm->def->placement_mode ==
         VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) ||
        (vm->def->numatune.memory.placement_mode ==
G
Gao feng 已提交
3495
         VIR_NUMA_TUNE_MEM_PLACEMENT_MODE_AUTO)) {
3496 3497
        nodeset = virNumaGetAutoPlacementAdvice(vm->def->vcpus,
                                                vm->def->mem.cur_balloon);
3498 3499 3500 3501 3502
        if (!nodeset)
            goto cleanup;

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

3503 3504
        if (virBitmapParse(nodeset, 0, &nodemask,
                           VIR_DOMAIN_CPUMASK_LEN) < 0)
3505 3506 3507 3508
            goto cleanup;
    }
    hookData.nodemask = nodemask;

3509
    VIR_DEBUG("Setting up domain cgroup (if required)");
3510
    if (qemuSetupCgroup(driver, vm, nodemask) < 0)
3511 3512 3513 3514 3515 3516 3517
        goto cleanup;

    if (VIR_ALLOC(priv->monConfig) < 0) {
        virReportOOMError();
        goto cleanup;
    }

3518
    VIR_DEBUG("Preparing monitor state");
3519
    if (qemuProcessPrepareMonitorChr(cfg, priv->monConfig, vm->def->name) < 0)
3520 3521
        goto cleanup;

3522
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON))
3523 3524 3525 3526
        priv->monJSON = 1;
    else
        priv->monJSON = 0;

3527 3528
    priv->monError = false;
    priv->monStart = 0;
3529 3530
    priv->gotShutdown = false;

3531
    VIR_FREE(priv->pidfile);
3532
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, vm->def->name))) {
3533 3534
        virReportSystemError(errno,
                             "%s", _("Failed to build pidfile path."));
3535 3536 3537
        goto cleanup;
    }

3538 3539
    if (unlink(priv->pidfile) < 0 &&
        errno != ENOENT) {
3540
        virReportSystemError(errno,
3541 3542
                             _("Cannot remove stale PID file %s"),
                             priv->pidfile);
3543 3544 3545 3546 3547 3548 3549 3550 3551 3552
        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
     * we also need to populate the PCi address set cache for later
     * use in hotplug
     */
3553
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
3554
        VIR_DEBUG("Assigning domain PCI addresses");
3555
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
3556
            goto cleanup;
3557 3558
    }

3559
    VIR_DEBUG("Building emulator command line");
3560
    if (!(cmd = qemuBuildCommandLine(conn, driver, vm->def, priv->monConfig,
3561
                                     priv->monJSON != 0, priv->qemuCaps,
3562
                                     migrateFrom, stdin_fd, snapshot, vmop)))
3563 3564 3565 3566
        goto cleanup;

    /* now that we know it is about to start call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3567
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
3568 3569 3570
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
3571 3572
                              VIR_HOOK_QEMU_OP_START, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
3573 3574 3575 3576 3577 3578 3579 3580 3581
        VIR_FREE(xml);

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

3582
    if ((timestamp = virTimeStringNow()) == NULL) {
3583 3584 3585 3586 3587 3588
        virReportOOMError();
        goto cleanup;
    } else {
        if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
            safewrite(logfile, START_POSTFIX, strlen(START_POSTFIX)) < 0) {
            VIR_WARN("Unable to write timestamp to logfile: %s",
3589
                     virStrerror(errno, ebuf, sizeof(ebuf)));
3590 3591 3592 3593 3594 3595 3596
        }

        VIR_FREE(timestamp);
    }

    virCommandWriteArgLog(cmd, logfile);

3597
    qemuDomainObjCheckTaint(driver, vm, logfile);
3598

3599 3600
    if ((pos = lseek(logfile, 0, SEEK_END)) < 0)
        VIR_WARN("Unable to seek to end of logfile: %s",
3601
                 virStrerror(errno, ebuf, sizeof(ebuf)));
3602 3603

    VIR_DEBUG("Clear emulator capabilities: %d",
3604 3605
              cfg->clearEmulatorCapabilities);
    if (cfg->clearEmulatorCapabilities)
3606 3607
        virCommandClearCaps(cmd);

3608 3609
    /* in case a certain disk is desirous of CAP_SYS_RAWIO, add this */
    for (i = 0; i < vm->def->ndisks; i++) {
3610 3611
        virDomainDiskDefPtr disk = vm->def->disks[i];

3612
        if (vm->def->disks[i]->rawio == 1)
R
Roman Bogorodskiy 已提交
3613
#ifdef CAP_SYS_RAWIO
3614
            virCommandAllowCap(cmd, CAP_SYS_RAWIO);
R
Roman Bogorodskiy 已提交
3615 3616 3617 3618
#else
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Raw I/O is not supported on this platform"));
#endif
3619

3620
        if (qemuAddSharedDisk(driver, disk, vm->def->name) < 0)
3621
            goto cleanup;
3622

3623 3624
        if (qemuSetUnprivSGIO(disk) < 0)
            goto cleanup;
3625 3626
    }

3627 3628
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);

3629 3630 3631 3632 3633 3634
    VIR_DEBUG("Setting up security labelling");
    if (virSecurityManagerSetChildProcessLabel(driver->securityManager,
                                               vm->def, cmd) < 0) {
        goto cleanup;
    }

3635 3636 3637
    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
3638
    virCommandSetPidFile(cmd, priv->pidfile);
3639
    virCommandDaemonize(cmd);
3640
    virCommandRequireHandshake(cmd);
3641

3642
    virSecurityManagerPreFork(driver->securityManager);
3643
    ret = virCommandRun(cmd, NULL);
3644
    virSecurityManagerPostFork(driver->securityManager);
3645

E
Eric Blake 已提交
3646
    /* wait for qemu process to show up */
3647
    if (ret == 0) {
3648
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
3649 3650
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
3651 3652 3653 3654 3655 3656 3657 3658 3659
            ret = -1;
        }
#if 0
    } else if (ret == -2) {
        /*
         * XXX this is bogus. It isn't safe to set vm->pid = child
         * because the child no longer exists.
         */

3660
        /* The virCommand process that launches the daemon failed. Pending on
3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671
         * when it failed (we can't determine for sure), there may be
         * extra info in the domain log (if the hook failed for example).
         *
         * Pretend like things succeeded, and let 'WaitForMonitor' report
         * the log contents for us.
         */
        vm->pid = child;
        ret = 0;
#endif
    }

3672
    VIR_DEBUG("Writing early domain status to disk");
3673
    if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0) {
3674 3675 3676
        goto cleanup;
    }

3677 3678 3679 3680 3681 3682 3683
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
        goto cleanup;
    }

    VIR_DEBUG("Setting domain security labels");
    if (virSecurityManagerSetAllLabel(driver->securityManager,
3684
                                      vm->def, stdin_path) < 0)
3685 3686
        goto cleanup;

3687 3688 3689 3690 3691 3692
    /* 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;

3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706
    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) &&
3707
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def, stdin_fd) < 0)
3708 3709 3710 3711 3712 3713 3714 3715 3716
            goto cleanup;
    }

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

3717
    if (migrateFrom)
3718
        flags |= VIR_QEMU_PROCESS_START_PAUSED;
3719 3720 3721 3722 3723 3724 3725

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

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

3726
    VIR_DEBUG("Waiting for monitor to show up");
3727
    if (qemuProcessWaitForMonitor(driver, vm, priv->qemuCaps, pos) < 0)
3728 3729
        goto cleanup;

D
Daniel P. Berrange 已提交
3730 3731 3732 3733 3734 3735 3736 3737
    /* Failure to connect to agent shouldn't be fatal */
    if (qemuConnectAgent(driver, vm) < 0) {
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

3738
    VIR_DEBUG("Detecting VCPU PIDs");
3739 3740 3741
    if (qemuProcessDetectVcpuPIDs(driver, vm) < 0)
        goto cleanup;

3742
    VIR_DEBUG("Setting cgroup for each VCPU (if required)");
3743 3744 3745
    if (qemuSetupCgroupForVcpu(driver, vm) < 0)
        goto cleanup;

3746
    VIR_DEBUG("Setting cgroup for emulator (if required)");
3747
    if (qemuSetupCgroupForEmulator(driver, vm, nodemask) < 0)
3748 3749
        goto cleanup;

3750
    VIR_DEBUG("Setting VCPU affinities");
3751
    if (qemuProcessSetVcpuAffinities(conn, vm) < 0)
3752 3753
        goto cleanup;

3754
    VIR_DEBUG("Setting affinity of emulator threads");
3755
    if (qemuProcessSetEmulatorAffinities(conn, vm) < 0)
3756 3757
        goto cleanup;

3758
    VIR_DEBUG("Setting any required VM passwords");
3759
    if (qemuProcessInitPasswords(conn, driver, vm) < 0)
3760 3761 3762 3763
        goto cleanup;

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
3764
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
3765
        VIR_DEBUG("Determining domain device PCI addresses");
3766 3767 3768 3769
        if (qemuProcessInitPCIAddresses(driver, vm) < 0)
            goto cleanup;
    }

3770 3771 3772 3773
    /* set default link states */
    /* qemu doesn't support setting this on the command line, so
     * enter the monitor */
    VIR_DEBUG("Setting network link states");
3774
    qemuDomainObjEnterMonitor(driver, vm);
3775
    if (qemuProcessSetLinkStates(vm) < 0) {
3776
        qemuDomainObjExitMonitor(driver, vm);
3777 3778 3779
        goto cleanup;
    }

3780
    qemuDomainObjExitMonitor(driver, vm);
3781

3782 3783 3784 3785
    /* 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.  */
3786
    VIR_DEBUG("Setting initial memory amount");
E
Eric Blake 已提交
3787
    cur_balloon = vm->def->mem.cur_balloon;
3788
    if (cur_balloon != vm->def->mem.cur_balloon) {
3789 3790 3791
        virReportError(VIR_ERR_OVERFLOW,
                       _("unable to set balloon to %lld"),
                       vm->def->mem.cur_balloon);
3792 3793
        goto cleanup;
    }
3794
    qemuDomainObjEnterMonitor(driver, vm);
E
Eric Blake 已提交
3795
    if (qemuMonitorSetBalloon(priv->mon, cur_balloon) < 0) {
3796
        qemuDomainObjExitMonitor(driver, vm);
3797 3798
        goto cleanup;
    }
3799
    qemuDomainObjExitMonitor(driver, vm);
3800

3801
    if (!(flags & VIR_QEMU_PROCESS_START_PAUSED)) {
3802
        VIR_DEBUG("Starting domain CPUs");
3803
        /* Allow the CPUS to start executing */
J
Jiri Denemark 已提交
3804
        if (qemuProcessStartCPUs(driver, vm, conn,
3805 3806
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 QEMU_ASYNC_JOB_NONE) < 0) {
3807
            if (virGetLastError() == NULL)
3808 3809
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("resume operation failed"));
3810 3811
            goto cleanup;
        }
J
Jiri Denemark 已提交
3812 3813 3814 3815 3816
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             migrateFrom ?
                             VIR_DOMAIN_PAUSED_MIGRATION :
                             VIR_DOMAIN_PAUSED_USER);
3817 3818
    }

3819
    if (flags & VIR_QEMU_PROCESS_START_AUTODESTROY &&
3820 3821 3822
        qemuProcessAutoDestroyAdd(driver, vm, conn) < 0)
        goto cleanup;

3823
    VIR_DEBUG("Writing domain status to disk");
3824
    if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0)
3825 3826
        goto cleanup;

3827 3828
    /* finally we can call the 'started' hook script if any */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3829
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843
        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;
    }

3844 3845
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);
3846
    virObjectUnref(cfg);
3847
    virObjectUnref(caps);
3848 3849 3850 3851 3852 3853 3854

    return 0;

cleanup:
    /* 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 */
3855
    VIR_FREE(nodeset);
3856
    virBitmapFree(nodemask);
3857 3858
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);
3859
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, stop_flags);
3860
    virObjectUnref(cfg);
3861
    virObjectUnref(caps);
3862 3863 3864 3865 3866

    return -1;
}


3867
int
3868
qemuProcessKill(virDomainObjPtr vm, unsigned int flags)
3869
{
3870
    int ret;
3871 3872 3873

    VIR_DEBUG("vm=%s pid=%d flags=%x",
              vm->def->name, vm->pid, flags);
3874

3875 3876 3877 3878 3879
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
3880 3881
    }

3882 3883 3884 3885 3886 3887
    if ((flags & VIR_QEMU_PROCESS_KILL_NOWAIT)) {
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
3888

3889 3890
    ret = virProcessKillPainfully(vm->pid,
                                  !!(flags & VIR_QEMU_PROCESS_KILL_FORCE));
3891

3892
    return ret;
3893 3894 3895
}


3896
void qemuProcessStop(virQEMUDriverPtr driver,
3897
                     virDomainObjPtr vm,
3898 3899
                     virDomainShutoffReason reason,
                     unsigned int flags)
3900 3901 3902 3903 3904 3905
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
3906
    virNetDevVPortProfilePtr vport = NULL;
3907 3908 3909 3910
    int i;
    int logfile = -1;
    char *timestamp;
    char ebuf[1024];
3911
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3912

3913 3914
    VIR_DEBUG("Shutting down VM '%s' pid=%d flags=%x",
              vm->def->name, vm->pid, flags);
3915 3916 3917

    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("VM '%s' not active", vm->def->name);
3918
        virObjectUnref(cfg);
3919 3920 3921
        return;
    }

3922
    /*
3923 3924
     * We may unlock the vm in qemuProcessKill(), and another thread
     * can lock the vm, and then call qemuProcessStop(). So we should
3925 3926 3927 3928
     * set vm->def->id to -1 here to avoid qemuProcessStop() to be called twice.
     */
    vm->def->id = -1;

3929
    if (virAtomicIntDecAndTest(&driver->nactive) && driver->inhibitCallback)
3930 3931
        driver->inhibitCallback(false, driver->inhibitOpaque);

3932
    if ((logfile = qemuDomainCreateLog(driver, vm, true)) < 0) {
3933 3934 3935
        /* 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",
3936
                  virStrerror(errno, ebuf, sizeof(ebuf)));
3937
    } else {
3938
        if ((timestamp = virTimeStringNow()) == NULL) {
3939 3940 3941 3942 3943 3944
            virReportOOMError();
        } else {
            if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
                safewrite(logfile, SHUTDOWN_POSTFIX,
                          strlen(SHUTDOWN_POSTFIX)) < 0) {
                VIR_WARN("Unable to write timestamp to logfile: %s",
3945
                         virStrerror(errno, ebuf, sizeof(ebuf)));
3946 3947 3948 3949 3950 3951 3952
            }

            VIR_FREE(timestamp);
        }

        if (VIR_CLOSE(logfile) < 0)
             VIR_WARN("Unable to close logfile: %s",
3953
                      virStrerror(errno, ebuf, sizeof(ebuf)));
3954 3955 3956 3957 3958 3959 3960 3961
    }

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

    virDomainConfVMNWFilterTeardown(vm);

3962
    if (cfg->macFilter) {
3963 3964 3965 3966 3967 3968
        def = vm->def;
        for (i = 0 ; i < def->nnets ; i++) {
            virDomainNetDefPtr net = def->nets[i];
            if (net->ifname == NULL)
                continue;
            if ((errno = networkDisallowMacOnPort(driver, net->ifname,
3969
                                                  &net->mac))) {
3970
                virReportSystemError(errno,
3971
             _("failed to remove ebtables rule to allow MAC address on '%s'"),
3972 3973 3974 3975 3976
                                     net->ifname);
            }
        }
    }

3977 3978 3979 3980 3981
    if (priv->nbdPort) {
        virPortAllocatorRelease(driver->remotePorts, priv->nbdPort);
        priv->nbdPort = 0;
    }

D
Daniel P. Berrange 已提交
3982 3983 3984 3985 3986 3987
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
        priv->agentError = false;
    }

3988
    if (priv->mon) {
3989
        qemuMonitorClose(priv->mon);
3990 3991
        priv->mon = NULL;
    }
3992 3993 3994 3995 3996 3997 3998 3999 4000

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

    /* shut it off for sure */
4001 4002 4003
    ignore_value(qemuProcessKill(vm,
                                 VIR_QEMU_PROCESS_KILL_FORCE|
                                 VIR_QEMU_PROCESS_KILL_NOCHECK));
4004

4005 4006
    qemuDomainCleanupRun(driver, vm);

4007
    /* Stop autodestroy in case guest is restarted */
4008
    qemuProcessAutoDestroyRemove(driver, vm);
4009

4010 4011
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4012
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4013 4014 4015

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4016 4017
                    VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
4018 4019 4020
        VIR_FREE(xml);
    }

4021 4022 4023 4024 4025
    /* Reset Security Labels unless caller don't want us to */
    if (!(flags & VIR_QEMU_PROCESS_STOP_NO_RELABEL))
        virSecurityManagerRestoreAllLabel(driver->securityManager,
                                          vm->def,
                                          flags & VIR_QEMU_PROCESS_STOP_MIGRATED);
4026
    virSecurityManagerReleaseLabel(driver->securityManager, vm->def);
4027

4028 4029
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk = vm->def->disks[i];
4030
        ignore_value(qemuRemoveSharedDisk(driver, disk, vm->def->name));
4031 4032
    }

4033
    /* Clear out dynamically assigned labels */
4034 4035 4036 4037 4038
    for (i = 0; i < vm->def->nseclabels; i++) {
        if (vm->def->seclabels[i]->type == VIR_DOMAIN_SECLABEL_DYNAMIC) {
            VIR_FREE(vm->def->seclabels[i]->label);
        }
        VIR_FREE(vm->def->seclabels[i]->imagelabel);
4039 4040 4041 4042 4043 4044 4045
    }

    virDomainDefClearDeviceAliases(vm->def);
    if (!priv->persistentAddrs) {
        virDomainDefClearPCIAddresses(vm->def);
        qemuDomainPCIAddressSetFree(priv->pciaddrs);
        priv->pciaddrs = NULL;
4046 4047 4048
        virDomainDefClearCCWAddresses(vm->def);
        qemuDomainCCWAddressSetFree(priv->ccwaddrs);
        priv->ccwaddrs = NULL;
4049 4050 4051 4052 4053 4054 4055
    }

    qemuDomainReAttachHostDevices(driver, vm->def);

    def = vm->def;
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
4056
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4057
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
4058
                             net->ifname, &net->mac,
4059 4060
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
4061
                             virDomainNetGetActualVirtPortProfile(net),
4062
                             cfg->stateDir));
4063 4064
            VIR_FREE(net->ifname);
        }
4065 4066 4067
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
A
Ansis Atteka 已提交
4068 4069 4070 4071 4072 4073
        vport = virDomainNetGetActualVirtPortProfile(net);
        if (vport && vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH)
            ignore_value(virNetDevOpenvswitchRemovePort(
                                       virDomainNetGetActualBridgeName(net),
                                       net->ifname));

4074 4075
        networkReleaseActualDevice(net);
    }
4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088

retry:
    if ((ret = qemuRemoveCgroup(driver, vm, 0)) < 0) {
        if (ret == -EBUSY && (retries++ < 5)) {
            usleep(200*1000);
            goto retry;
        }
        VIR_WARN("Failed to remove cgroup for %s",
                 vm->def->name);
    }

    qemuProcessRemoveDomainStatus(driver, vm);

4089 4090
    /* Remove VNC and Spice ports from port reservation bitmap, but only if
       they were reserved by the driver (autoport=yes)
4091
    */
4092 4093 4094 4095
    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) {
4096 4097
            ignore_value(virPortAllocatorRelease(driver->remotePorts,
                                                 graphics->data.vnc.port));
4098 4099 4100
        }
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE &&
            graphics->data.spice.autoport) {
4101 4102 4103 4104
            ignore_value(virPortAllocatorRelease(driver->remotePorts,
                                                 graphics->data.spice.port));
            ignore_value(virPortAllocatorRelease(driver->remotePorts,
                                                 graphics->data.spice.tlsPort));
4105
        }
4106 4107
    }

4108
    vm->taint = 0;
4109
    vm->pid = -1;
J
Jiri Denemark 已提交
4110
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
4111 4112
    VIR_FREE(priv->vcpupids);
    priv->nvcpupids = 0;
4113 4114
    virObjectUnref(priv->qemuCaps);
    priv->qemuCaps = NULL;
4115
    VIR_FREE(priv->pidfile);
4116

4117
    /* The "release" hook cleans up additional resources */
4118
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4119
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4120 4121 4122

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4123 4124
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
4125 4126 4127
        VIR_FREE(xml);
    }

4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138
    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);
    }
4139
    virObjectUnref(cfg);
4140
}
4141 4142


4143
int qemuProcessAttach(virConnectPtr conn ATTRIBUTE_UNUSED,
4144
                      virQEMUDriverPtr driver,
4145
                      virDomainObjPtr vm,
4146
                      pid_t pid,
4147 4148 4149 4150
                      const char *pidfile,
                      virDomainChrSourceDefPtr monConfig,
                      bool monJSON)
{
4151
    size_t i;
4152 4153 4154 4155 4156
    char ebuf[1024];
    int logfile = -1;
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool running = true;
4157
    virDomainPausedReason reason;
4158
    virSecurityLabelPtr seclabel = NULL;
4159 4160 4161
    virSecurityLabelDefPtr seclabeldef = NULL;
    virSecurityManagerPtr* sec_managers = NULL;
    const char *model;
4162
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
4163
    virCapsPtr caps = NULL;
4164 4165 4166 4167

    VIR_DEBUG("Beginning VM attach process");

    if (virDomainObjIsActive(vm)) {
4168 4169
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
4170
        virObjectUnref(cfg);
4171 4172 4173
        return -1;
    }

4174 4175 4176
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

4177 4178 4179 4180 4181
    /* 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");
4182
    if (virDomainObjSetDefTransient(caps, driver->xmlconf, vm, true) < 0)
4183 4184
        goto cleanup;

4185
    vm->def->id = qemuDriverAllocateID(driver);
4186

4187
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
4188 4189
        driver->inhibitCallback(true, driver->inhibitOpaque);

4190
    if (virFileMakePath(cfg->logDir) < 0) {
4191 4192
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
4193
                             cfg->logDir);
4194 4195 4196 4197 4198 4199 4200 4201 4202
        goto cleanup;
    }

    VIR_FREE(priv->pidfile);
    if (pidfile &&
        !(priv->pidfile = strdup(pidfile)))
        goto no_memory;

    VIR_DEBUG("Detect security driver config");
4203 4204
    sec_managers = virSecurityManagerGetNested(driver->securityManager);
    if (sec_managers == NULL) {
4205
        goto cleanup;
4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227
    }

    for (i = 0; sec_managers[i]; i++) {
        model = virSecurityManagerGetModel(sec_managers[i]);
        seclabeldef = virDomainDefGetSecurityLabelDef(vm->def, model);
        if (seclabeldef == NULL) {
            goto cleanup;
        }
        seclabeldef->type = VIR_DOMAIN_SECLABEL_STATIC;
        if (VIR_ALLOC(seclabel) < 0)
            goto no_memory;
        if (virSecurityManagerGetProcessLabel(driver->securityManager,
                                              vm->def, vm->pid, seclabel) < 0)
            goto cleanup;

        if (!(seclabeldef->model = strdup(model)))
            goto no_memory;

        if (!(seclabeldef->label = strdup(seclabel->label)))
            goto no_memory;
        VIR_FREE(seclabel);
    }
4228 4229 4230 4231 4232 4233

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

    VIR_DEBUG("Determining emulator version");
4234 4235 4236
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
                                                      vm->def->emulator)))
4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252
        goto cleanup;

    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
     * we also need to populate the PCi address set cache for later
     * use in hotplug
     */
4253
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4254
        VIR_DEBUG("Assigning domain PCI addresses");
4255
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
4256
            goto cleanup;
4257 4258
    }

4259
    if ((timestamp = virTimeStringNow()) == NULL) {
4260 4261 4262 4263 4264 4265
        virReportOOMError();
        goto cleanup;
    } 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",
4266
                     virStrerror(errno, ebuf, sizeof(ebuf)));
4267 4268 4269 4270 4271 4272 4273 4274 4275 4276
        }

        VIR_FREE(timestamp);
    }

    qemuDomainObjTaint(driver, vm, VIR_DOMAIN_TAINT_EXTERNAL_LAUNCH, logfile);

    vm->pid = pid;

    VIR_DEBUG("Waiting for monitor to show up");
4277
    if (qemuProcessWaitForMonitor(driver, vm, priv->qemuCaps, -1) < 0)
4278 4279
        goto cleanup;

D
Daniel P. Berrange 已提交
4280 4281 4282 4283 4284 4285 4286 4287
    /* Failure to connect to agent shouldn't be fatal */
    if (qemuConnectAgent(driver, vm) < 0) {
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

4288 4289 4290 4291 4292 4293
    VIR_DEBUG("Detecting VCPU PIDs");
    if (qemuProcessDetectVcpuPIDs(driver, vm) < 0)
        goto cleanup;

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
4294
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4295 4296 4297 4298 4299 4300
        VIR_DEBUG("Determining domain device PCI addresses");
        if (qemuProcessInitPCIAddresses(driver, vm) < 0)
            goto cleanup;
    }

    VIR_DEBUG("Getting initial memory amount");
4301
    qemuDomainObjEnterMonitor(driver, vm);
4302
    if (qemuMonitorGetBalloonInfo(priv->mon, &vm->def->mem.cur_balloon) < 0) {
4303
        qemuDomainObjExitMonitor(driver, vm);
4304 4305
        goto cleanup;
    }
4306
    if (qemuMonitorGetStatus(priv->mon, &running, &reason) < 0) {
4307
        qemuDomainObjExitMonitor(driver, vm);
4308 4309 4310
        goto cleanup;
    }
    if (qemuMonitorGetVirtType(priv->mon, &vm->def->virtType) < 0) {
4311
        qemuDomainObjExitMonitor(driver, vm);
4312 4313
        goto cleanup;
    }
4314
    qemuDomainObjExitMonitor(driver, vm);
4315 4316 4317 4318 4319 4320 4321 4322

    if (!virDomainObjIsActive(vm))
        goto cleanup;

    if (running)
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                             VIR_DOMAIN_RUNNING_UNPAUSED);
    else
4323
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
4324 4325

    VIR_DEBUG("Writing domain status to disk");
4326
    if (virDomainSaveStatus(driver->xmlconf, cfg->stateDir, vm) < 0)
4327 4328
        goto cleanup;

4329 4330
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4331
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345
        int hookret;

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

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

4346 4347
    VIR_FORCE_CLOSE(logfile);
    VIR_FREE(seclabel);
4348
    VIR_FREE(sec_managers);
4349
    virObjectUnref(cfg);
4350
    virObjectUnref(caps);
4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361

    return 0;

no_memory:
    virReportOOMError();
cleanup:
    /* 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 */
    VIR_FORCE_CLOSE(logfile);
    VIR_FREE(seclabel);
4362
    VIR_FREE(sec_managers);
4363
    virDomainChrSourceDefFree(monConfig);
4364
    virObjectUnref(cfg);
4365
    virObjectUnref(caps);
4366 4367 4368 4369
    return -1;
}


4370
static virDomainObjPtr
4371
qemuProcessAutoDestroy(virQEMUDriverPtr driver,
4372 4373
                       virDomainObjPtr dom,
                       virConnectPtr conn)
4374
{
4375
    qemuDomainObjPrivatePtr priv = dom->privateData;
4376 4377
    virDomainEventPtr event = NULL;

4378
    VIR_DEBUG("vm=%s, conn=%p", dom->def->name, conn);
4379

4380 4381
    if (priv->job.asyncJob) {
        VIR_DEBUG("vm=%s has long-term job active, cancelling",
4382
                  dom->def->name);
4383
        qemuDomainObjDiscardAsyncJob(driver, dom);
4384 4385
    }

4386 4387
    if (qemuDomainObjBeginJob(driver, dom,
                              QEMU_JOB_DESTROY) < 0)
4388 4389 4390
        goto cleanup;

    VIR_DEBUG("Killing domain");
4391

4392 4393
    qemuProcessStop(driver, dom, VIR_DOMAIN_SHUTOFF_DESTROYED,
                    VIR_QEMU_PROCESS_STOP_MIGRATED);
4394

4395
    virDomainAuditStop(dom, "destroyed");
4396 4397 4398
    event = virDomainEventNewFromObj(dom,
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_DESTROYED);
4399

4400
    if (!qemuDomainObjEndJob(driver, dom))
4401
        dom = NULL;
4402
    if (dom && !dom->persistent) {
4403
        qemuDomainRemoveInactive(driver, dom);
4404 4405
        dom = NULL;
    }
4406
    if (event)
4407
        qemuDomainEventQueue(driver, event);
4408

4409 4410
cleanup:
    return dom;
4411 4412
}

4413
int qemuProcessAutoDestroyAdd(virQEMUDriverPtr driver,
4414 4415 4416
                              virDomainObjPtr vm,
                              virConnectPtr conn)
{
4417
    VIR_DEBUG("vm=%s, conn=%p", vm->def->name, conn);
4418 4419
    return virQEMUCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                    qemuProcessAutoDestroy);
4420 4421
}

4422
int qemuProcessAutoDestroyRemove(virQEMUDriverPtr driver,
4423 4424
                                 virDomainObjPtr vm)
{
4425
    VIR_DEBUG("vm=%s", vm->def->name);
4426 4427
    return virQEMUCloseCallbacksUnset(driver->closeCallbacks, vm,
                                      qemuProcessAutoDestroy);
4428
}
4429

4430
bool qemuProcessAutoDestroyActive(virQEMUDriverPtr driver,
4431 4432
                                  virDomainObjPtr vm)
{
4433
    virQEMUCloseCallback cb;
4434
    VIR_DEBUG("vm=%s", vm->def->name);
4435
    cb = virQEMUCloseCallbacksGet(driver->closeCallbacks, vm, NULL);
4436
    return cb == qemuProcessAutoDestroy;
4437
}