qemu_process.c 136.1 KB
Newer Older
1 2 3
/*
 * qemu_process.h: QEMU process management
 *
4
 * Copyright (C) 2006-2012 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>
30
#include <linux/capability.h>
31 32 33 34 35 36 37 38 39 40

#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"
41
#include "qemu_migration.h"
42

43
#if HAVE_NUMACTL
E
Eric Blake 已提交
44
# define NUMA_VERSION1_COMPATIBILITY 1
45 46 47
# include <numa.h>
#endif

48 49 50 51 52
#include "datatypes.h"
#include "logging.h"
#include "virterror_internal.h"
#include "memory.h"
#include "hooks.h"
E
Eric Blake 已提交
53
#include "virfile.h"
54
#include "virpidfile.h"
55 56 57 58
#include "util.h"
#include "c-ctype.h"
#include "nodeinfo.h"
#include "processinfo.h"
59
#include "domain_audit.h"
60
#include "domain_nwfilter.h"
61
#include "locking/domain_lock.h"
62
#include "network/bridge_driver.h"
63
#include "uuid.h"
64
#include "virprocess.h"
65
#include "virtime.h"
A
Ansis Atteka 已提交
66
#include "virnetdevtap.h"
67
#include "bitmap.h"
68 69 70 71

#define VIR_FROM_THIS VIR_FROM_QEMU

#define START_POSTFIX ": starting up\n"
72
#define ATTACH_POSTFIX ": attaching\n"
73 74 75 76 77 78 79 80 81 82 83 84 85 86 87
#define SHUTDOWN_POSTFIX ": shutting down\n"

/**
 * qemudRemoveDomainStatus
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
qemuProcessRemoveDomainStatus(struct qemud_driver *driver,
                              virDomainObjPtr vm)
{
    char ebuf[1024];
    char *file = NULL;
88
    qemuDomainObjPrivatePtr priv = vm->privateData;
89 90 91

    if (virAsprintf(&file, "%s/%s.xml", driver->stateDir, vm->def->name) < 0) {
        virReportOOMError();
92
        return -1;
93 94 95 96 97 98 99
    }

    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);

100 101 102
    if (priv->pidfile &&
        unlink(priv->pidfile) < 0 &&
        errno != ENOENT)
103 104 105 106 107 108 109 110 111 112
        VIR_WARN("Failed to remove PID file for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));

    return 0;
}


/* XXX figure out how to remove this */
extern struct qemud_driver *qemu_driver;

D
Daniel P. Berrange 已提交
113 114 115 116 117 118 119
/*
 * 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
120
qemuProcessHandleAgentEOF(qemuAgentPtr agent,
D
Daniel P. Berrange 已提交
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
                          virDomainObjPtr vm)
{
    struct qemud_driver *driver = qemu_driver;
    qemuDomainObjPrivatePtr priv;

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

    qemuDriverLock(driver);
    virDomainObjLock(vm);

    priv = vm->privateData;
    priv->agent = NULL;

    virDomainObjUnlock(vm);
    qemuDriverUnlock(driver);
136 137

    qemuAgentClose(agent);
D
Daniel P. Berrange 已提交
138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171
}


/*
 * 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)
{
    struct qemud_driver *driver = qemu_driver;
    qemuDomainObjPrivatePtr priv;

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

    qemuDriverLock(driver);
    virDomainObjLock(vm);

    priv = vm->privateData;

    priv->agentError = true;

    virDomainObjUnlock(vm);
    qemuDriverUnlock(driver);
}

static void qemuProcessHandleAgentDestroy(qemuAgentPtr agent,
                                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

172 173
    VIR_DEBUG("Received destroy agent=%p vm=%p", agent, vm);

D
Daniel P. Berrange 已提交
174 175 176 177
    virDomainObjLock(vm);
    priv = vm->privateData;
    if (priv->agent == agent)
        priv->agent = NULL;
178
    if (virObjectUnref(vm))
D
Daniel P. Berrange 已提交
179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
        virDomainObjUnlock(vm);
}


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;

201
        if (STREQ_NULLABLE(channel->target.name, "org.qemu.guest_agent.0")) {
D
Daniel P. Berrange 已提交
202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229
            config = &channel->source;
            break;
        }
    }

    return config;
}

static int
qemuConnectAgent(struct qemud_driver *driver, virDomainObjPtr vm)
{
    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 */
230
    virObjectRef(vm);
D
Daniel P. Berrange 已提交
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251

    ignore_value(virTimeMillisNow(&priv->agentStart));
    virDomainObjUnlock(vm);
    qemuDriverUnlock(driver);

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

    qemuDriverLock(driver);
    virDomainObjLock(vm);
    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)
252
        virObjectUnref(vm);
D
Daniel P. Berrange 已提交
253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271

    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;
}


272
/*
273
 * This is a callback registered with a qemuMonitorPtr instance,
274 275 276 277 278 279
 * 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,
280
                            virDomainObjPtr vm)
281 282 283 284
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;
    qemuDomainObjPrivatePtr priv;
285 286 287
    int eventReason = VIR_DOMAIN_EVENT_STOPPED_SHUTDOWN;
    int stopReason = VIR_DOMAIN_SHUTOFF_SHUTDOWN;
    const char *auditReason = "shutdown";
288 289 290

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

291
    qemuDriverLock(driver);
292 293
    virDomainObjLock(vm);

294 295 296 297 298 299 300
    priv = vm->privateData;

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

301 302
    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Domain %p is not active, ignoring EOF", vm);
303
        goto unlock;
304 305
    }

306
    if (priv->monJSON && !priv->gotShutdown) {
307 308
        VIR_DEBUG("Monitor connection to '%s' closed without SHUTDOWN event; "
                  "assuming the domain crashed", vm->def->name);
309
        eventReason = VIR_DOMAIN_EVENT_STOPPED_FAILED;
310
        stopReason = VIR_DOMAIN_SHUTOFF_CRASHED;
311
        auditReason = "failed";
312 313 314 315
    }

    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STOPPED,
316
                                     eventReason);
317
    qemuProcessStop(driver, vm, stopReason, 0);
318
    virDomainAuditStop(vm, auditReason);
319

320
    if (!vm->persistent) {
321
        qemuDomainRemoveInactive(driver, vm);
322 323 324 325 326
        goto cleanup;
    }

unlock:
    virDomainObjUnlock(vm);
327

328
cleanup:
329
    if (event)
330
        qemuDomainEventQueue(driver, event);
331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352
    qemuDriverUnlock(driver);
}


/*
 * 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)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;

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

    qemuDriverLock(driver);
    virDomainObjLock(vm);

353
    ((qemuDomainObjPrivatePtr) vm->privateData)->monError = true;
354 355 356 357 358
    event = virDomainEventControlErrorNewFromObj(vm);
    if (event)
        qemuDomainEventQueue(driver, event);

    virDomainObjUnlock(vm);
359
    qemuDriverUnlock(driver);
360 361 362 363 364 365 366
}


static virDomainDiskDefPtr
qemuProcessFindDomainDiskByPath(virDomainObjPtr vm,
                                const char *path)
{
367
    int i = virDomainDiskIndexByName(vm->def, path, true);
368

369 370
    if (i >= 0)
        return vm->def->disks[i];
371

372 373 374
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with path %s"),
                   path);
375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
    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;
    }

395 396 397
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with alias %s"),
                   alias);
398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414
    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) {
415 416 417
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("disk %s does not have any encryption information"),
                       disk->src);
418 419 420 421 422
        return -1;
    }
    enc = disk->encryption;

    if (!conn) {
423 424
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("cannot find secrets without a connection"));
425 426 427 428 429 430
        goto cleanup;
    }

    if (conn->secretDriver == NULL ||
        conn->secretDriver->lookupByUUID == NULL ||
        conn->secretDriver->getValue == NULL) {
431 432
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("secret storage not supported"));
433 434 435 436 437 438 439
        goto cleanup;
    }

    if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW ||
        enc->nsecrets != 1 ||
        enc->secrets[0]->type !=
        VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE) {
440 441
        virReportError(VIR_ERR_XML_ERROR,
                       _("invalid <encryption> for volume %s"), disk->src);
442 443 444 445 446 447 448
        goto cleanup;
    }

    secret = conn->secretDriver->lookupByUUID(conn,
                                              enc->secrets[0]->uuid);
    if (secret == NULL)
        goto cleanup;
449
    data = conn->secretDriver->getValue(secret, &size, 0,
450
                                        VIR_SECRET_GET_VALUE_INTERNAL_CALL);
451
    virObjectUnref(secret);
452 453 454 455 456 457
    if (data == NULL)
        goto cleanup;

    if (memchr(data, '\0', size) != NULL) {
        memset(data, 0, size);
        VIR_FREE(data);
458 459 460
        virReportError(VIR_ERR_XML_ERROR,
                       _("format='qcow' passphrase for %s must not contain a "
                         "'\\0'"), disk->src);
461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515
        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;

    virDomainObjLock(vm);
    disk = qemuProcessFindDomainDiskByPath(vm, path);

    if (!disk)
        goto cleanup;

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

cleanup:
    virDomainObjUnlock(vm);
    return ret;
}


static int
qemuProcessHandleReset(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                       virDomainObjPtr vm)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event;
516
    qemuDomainObjPrivatePtr priv;
517 518

    virDomainObjLock(vm);
519

520
    event = virDomainEventRebootNewFromObj(vm);
521 522 523 524
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);

525 526 527 528 529 530 531 532 533 534 535 536
    virDomainObjUnlock(vm);

    if (event) {
        qemuDriverLock(driver);
        qemuDomainEventQueue(driver, event);
        qemuDriverUnlock(driver);
    }

    return 0;
}


537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555
/*
 * 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)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainObjPtr vm = opaque;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainEventPtr event = NULL;
    int ret = -1;
    VIR_DEBUG("vm=%p", vm);
    qemuDriverLock(driver);
    virDomainObjLock(vm);
556
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
557 558 559
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
560 561
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
562 563 564
        goto endjob;
    }

565
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
566 567 568 569 570 571 572
    if (qemuMonitorSystemReset(priv->mon) < 0) {
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        goto endjob;
    }
    qemuDomainObjExitMonitorWithDriver(driver, vm);

    if (!virDomainObjIsActive(vm)) {
573 574
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
575 576 577 578
        goto endjob;
    }

    if (qemuProcessStartCPUs(driver, vm, NULL,
579 580
                             VIR_DOMAIN_RUNNING_BOOTED,
                             QEMU_ASYNC_JOB_NONE) < 0) {
581
        if (virGetLastError() == NULL)
582 583
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
584 585
        goto endjob;
    }
586
    priv->gotShutdown = false;
587 588 589 590 591 592 593
    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_RESUMED,
                                     VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

    ret = 0;

endjob:
594
    if (!qemuDomainObjEndJob(driver, vm))
595 596 597 598
        vm = NULL;

cleanup:
    if (vm) {
599 600 601 602
        if (ret == -1) {
            ignore_value(qemuProcessKill(driver, vm,
                                         VIR_QEMU_PROCESS_KILL_FORCE));
        }
603
        if (virObjectUnref(vm))
604 605 606 607 608 609 610 611
            virDomainObjUnlock(vm);
    }
    if (event)
        qemuDomainEventQueue(driver, event);
    qemuDriverUnlock(driver);
}


612
static void
613 614
qemuProcessShutdownOrReboot(struct qemud_driver *driver,
                            virDomainObjPtr vm)
615
{
616 617 618
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
619
        qemuDomainSetFakeReboot(driver, vm, false);
620
        virObjectRef(vm);
621 622 623 624 625
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
626
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
627 628
            ignore_value(qemuProcessKill(driver, vm,
                                         VIR_QEMU_PROCESS_KILL_NOWAIT));
629
            virObjectUnref(vm);
630 631
        }
    } else {
632
        ignore_value(qemuProcessKill(driver, vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
633
    }
634
}
635

636 637 638 639
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm)
{
640 641 642 643
    struct qemud_driver *driver = qemu_driver;
    qemuDomainObjPrivatePtr priv;
    virDomainEventPtr event = NULL;

644 645 646
    VIR_DEBUG("vm=%p", vm);

    virDomainObjLock(vm);
647 648 649 650 651 652

    priv = vm->privateData;
    if (priv->gotShutdown) {
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
653 654 655 656
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
    }
    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);

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

674 675 676
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

677 678 679
    qemuProcessShutdownOrReboot(driver, vm);

unlock:
680
    virDomainObjUnlock(vm);
681 682 683 684 685 686 687

    if (event) {
        qemuDriverLock(driver);
        qemuDomainEventQueue(driver, event);
        qemuDriverUnlock(driver);
    }

688 689 690 691 692 693 694 695 696 697 698 699
    return 0;
}


static int
qemuProcessHandleStop(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                      virDomainObjPtr vm)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;

    virDomainObjLock(vm);
J
Jiri Denemark 已提交
700
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
701
        qemuDomainObjPrivatePtr priv = vm->privateData;
702

703
        if (priv->gotShutdown) {
704 705
            VIR_DEBUG("Ignoring STOP event after SHUTDOWN");
            goto unlock;
706 707
        }

708 709
        VIR_DEBUG("Transitioned guest %s to paused state",
                  vm->def->name);
710

711
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_UNKNOWN);
712 713 714 715
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);

716 717 718 719 720
        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));

721 722 723 724
        if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
725
    }
726 727

unlock:
728 729 730 731
    virDomainObjUnlock(vm);

    if (event) {
        qemuDriverLock(driver);
732
        qemuDomainEventQueue(driver, event);
733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751
        qemuDriverUnlock(driver);
    }

    return 0;
}


static int
qemuProcessHandleRTCChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm,
                           long long offset)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event;

    virDomainObjLock(vm);
    event = virDomainEventRTCChangeNewFromObj(vm, offset);

    if (vm->def->clock.offset == VIR_DOMAIN_CLOCK_OFFSET_VARIABLE)
752
        vm->def->clock.data.variable.adjustment = offset;
753 754

    if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0)
755
        VIR_WARN("unable to save domain status with RTC change");
756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781

    virDomainObjUnlock(vm);

    if (event) {
        qemuDriverLock(driver);
        qemuDomainEventQueue(driver, event);
        qemuDriverUnlock(driver);
    }

    return 0;
}


static int
qemuProcessHandleWatchdog(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          int action)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr watchdogEvent = NULL;
    virDomainEventPtr lifecycleEvent = NULL;

    virDomainObjLock(vm);
    watchdogEvent = virDomainEventWatchdogNewFromObj(vm, action);

    if (action == VIR_DOMAIN_EVENT_WATCHDOG_PAUSE &&
J
Jiri Denemark 已提交
782
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
783
        qemuDomainObjPrivatePtr priv = vm->privateData;
784 785
        VIR_DEBUG("Transitioned guest %s to paused state due to watchdog", vm->def->name);

J
Jiri Denemark 已提交
786
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
787 788 789 790
        lifecycleEvent = virDomainEventNewFromObj(vm,
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

791 792 793 794 795
        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));

796 797 798 799
        if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
800 801 802 803 804 805 806
    }

    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
807 808 809
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
810
            virObjectRef(vm);
811
            if (virThreadPoolSendJob(driver->workerPool, 0, wdEvent) < 0) {
812
                if (!virObjectUnref(vm))
813
                    vm = NULL;
W
Wen Congyang 已提交
814 815
                VIR_FREE(wdEvent);
            }
816
        } else {
817
            virReportOOMError();
818
        }
819 820
    }

821 822
    if (vm)
        virDomainObjUnlock(vm);
823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866

    if (watchdogEvent || lifecycleEvent) {
        qemuDriverLock(driver);
        if (watchdogEvent)
            qemuDomainEventQueue(driver, watchdogEvent);
        if (lifecycleEvent)
            qemuDomainEventQueue(driver, lifecycleEvent);
        qemuDriverUnlock(driver);
    }

    return 0;
}


static int
qemuProcessHandleIOError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virDomainObjPtr vm,
                         const char *diskAlias,
                         int action,
                         const char *reason)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr ioErrorEvent = NULL;
    virDomainEventPtr ioErrorEvent2 = NULL;
    virDomainEventPtr lifecycleEvent = NULL;
    const char *srcPath;
    const char *devAlias;
    virDomainDiskDefPtr disk;

    virDomainObjLock(vm);
    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
867
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
868
        qemuDomainObjPrivatePtr priv = vm->privateData;
869 870
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

J
Jiri Denemark 已提交
871
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
872 873 874 875
        lifecycleEvent = virDomainEventNewFromObj(vm,
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

876 877 878 879 880
        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));

881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899
        if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0)
            VIR_WARN("Unable to save status on vm %s after IO error", vm->def->name);
    }
    virDomainObjUnlock(vm);

    if (ioErrorEvent || ioErrorEvent2 || lifecycleEvent) {
        qemuDriverLock(driver);
        if (ioErrorEvent)
            qemuDomainEventQueue(driver, ioErrorEvent);
        if (ioErrorEvent2)
            qemuDomainEventQueue(driver, ioErrorEvent2);
        if (lifecycleEvent)
            qemuDomainEventQueue(driver, lifecycleEvent);
        qemuDriverUnlock(driver);
    }

    return 0;
}

900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929
static int
qemuProcessHandleBlockJob(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          const char *diskAlias,
                          int type,
                          int status)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;
    const char *path;
    virDomainDiskDefPtr disk;

    virDomainObjLock(vm);
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
        path = disk->src;
        event = virDomainEventBlockJobNewFromObj(vm, path, type, status);
    }

    virDomainObjUnlock(vm);

    if (event) {
        qemuDriverLock(driver);
        qemuDomainEventQueue(driver, event);
        qemuDriverUnlock(driver);
    }

    return 0;
}
930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971

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)
{
    struct qemud_driver *driver = qemu_driver;
    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++;
972 973 974
        if (!(subject->identities[subject->nidentity-1].type = strdup("x509dname")) ||
            !(subject->identities[subject->nidentity-1].name = strdup(x509dname)))
            goto no_memory;
975 976 977 978 979
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
            goto no_memory;
        subject->nidentity++;
980 981 982
        if (!(subject->identities[subject->nidentity-1].type = strdup("saslUsername")) ||
            !(subject->identities[subject->nidentity-1].name = strdup(saslUsername)))
            goto no_memory;
983 984
    }

985
    virDomainObjLock(vm);
986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
    event = virDomainEventGraphicsNewFromObj(vm, phase, localAddr, remoteAddr, authScheme, subject);
    virDomainObjUnlock(vm);

    if (event) {
        qemuDriverLock(driver);
        qemuDomainEventQueue(driver, event);
        qemuDriverUnlock(driver);
    }

    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;
}


static void qemuProcessHandleMonitorDestroy(qemuMonitorPtr mon,
                                            virDomainObjPtr vm)
{
1025 1026 1027 1028
    qemuDomainObjPrivatePtr priv;

    virDomainObjLock(vm);
    priv = vm->privateData;
1029 1030
    if (priv->mon == mon)
        priv->mon = NULL;
1031
    if (virObjectUnref(vm))
1032
        virDomainObjUnlock(vm);
1033 1034
}

1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
static int
qemuProcessHandleTrayChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            const char *devAlias,
                            int reason)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;
    virDomainDiskDefPtr disk;

    virDomainObjLock(vm);
    disk = qemuProcessFindDomainDiskByAlias(vm, devAlias);

    if (disk) {
        event = virDomainEventTrayChangeNewFromObj(vm,
                                                   devAlias,
                                                   reason);
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
        /* 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;

        if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
            VIR_WARN("Unable to save status on vm %s after tray moved event",
                     vm->def->name);
        }
1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074
    }

    virDomainObjUnlock(vm);

    if (event) {
        qemuDriverLock(driver);
        qemuDomainEventQueue(driver, event);
        qemuDriverUnlock(driver);
    }

    return 0;
}

O
Osier Yang 已提交
1075 1076 1077 1078 1079 1080
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;
1081
    virDomainEventPtr lifecycleEvent = NULL;
O
Osier Yang 已提交
1082 1083 1084 1085

    virDomainObjLock(vm);
    event = virDomainEventPMWakeupNewFromObj(vm);

1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104
    /* 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);

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

O
Osier Yang 已提交
1105 1106
    virDomainObjUnlock(vm);

1107
    if (event || lifecycleEvent) {
O
Osier Yang 已提交
1108
        qemuDriverLock(driver);
1109 1110 1111 1112
        if (event)
            qemuDomainEventQueue(driver, event);
        if (lifecycleEvent)
            qemuDomainEventQueue(driver, lifecycleEvent);
O
Osier Yang 已提交
1113 1114 1115 1116 1117
        qemuDriverUnlock(driver);
    }

    return 0;
}
1118

O
Osier Yang 已提交
1119 1120 1121 1122 1123 1124
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;
J
Jiri Denemark 已提交
1125
    virDomainEventPtr lifecycleEvent = NULL;
O
Osier Yang 已提交
1126 1127 1128 1129

    virDomainObjLock(vm);
    event = virDomainEventPMSuspendNewFromObj(vm);

1130
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1131
        qemuDomainObjPrivatePtr priv = vm->privateData;
1132 1133 1134 1135 1136
        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
1137 1138 1139 1140
        lifecycleEvent =
            virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1141 1142 1143 1144 1145

        if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1146 1147 1148

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1149 1150
    }

O
Osier Yang 已提交
1151 1152
    virDomainObjUnlock(vm);

J
Jiri Denemark 已提交
1153
    if (event || lifecycleEvent) {
O
Osier Yang 已提交
1154
        qemuDriverLock(driver);
J
Jiri Denemark 已提交
1155 1156 1157 1158
        if (event)
            qemuDomainEventQueue(driver, event);
        if (lifecycleEvent)
            qemuDomainEventQueue(driver, lifecycleEvent);
O
Osier Yang 已提交
1159 1160 1161 1162 1163 1164
        qemuDriverUnlock(driver);
    }

    return 0;
}

1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194
static int
qemuProcessHandleBalloonChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               unsigned long long actual)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event;

    virDomainObjLock(vm);
    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;

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

    virDomainObjUnlock(vm);

    if (event) {
        qemuDriverLock(driver);
        qemuDomainEventQueue(driver, event);
        qemuDriverUnlock(driver);
    }

    return 0;
}


1195 1196 1197
static qemuMonitorCallbacks monitorCallbacks = {
    .destroy = qemuProcessHandleMonitorDestroy,
    .eofNotify = qemuProcessHandleMonitorEOF,
1198
    .errorNotify = qemuProcessHandleMonitorError,
1199 1200 1201 1202 1203 1204 1205 1206
    .diskSecretLookup = qemuProcessFindVolumeQcowPassphrase,
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1207
    .domainBlockJob = qemuProcessHandleBlockJob,
1208
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1209
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1210
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1211
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1212 1213 1214 1215 1216 1217 1218
};

static int
qemuConnectMonitor(struct qemud_driver *driver, virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
1219
    qemuMonitorPtr mon = NULL;
1220

1221
    if (virSecurityManagerSetDaemonSocketLabel(driver->securityManager,
1222
                                               vm->def) < 0) {
1223 1224 1225 1226 1227 1228 1229
        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 */
1230
    virObjectRef(vm);
1231

1232
    ignore_value(virTimeMillisNow(&priv->monStart));
1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243
    virDomainObjUnlock(vm);
    qemuDriverUnlock(driver);

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

    qemuDriverLock(driver);
    virDomainObjLock(vm);
    priv->monStart = 0;
1244

1245
    if (mon == NULL) {
1246
        virObjectUnref(vm);
1247
    } else if (!virDomainObjIsActive(vm)) {
1248
        qemuMonitorClose(mon);
1249
        mon = NULL;
1250 1251 1252
    }
    priv->mon = mon;

1253
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0) {
1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264
        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;
    }


1265
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
1266
    ret = qemuMonitorSetCapabilities(priv->mon, priv->caps);
1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291
    qemuDomainObjExitMonitorWithDriver(driver, vm);

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;
1292 1293 1294 1295
    char *debug = NULL;
    int ret = -1;
    char *filter_next = buf;

1296 1297
    buf[0] = '\0';

1298 1299 1300 1301 1302 1303 1304
    /* This relies on log message format generated by virLogFormatString() and
     * might need to be modified when message format changes. */
    if (virAsprintf(&debug, ": %d: debug : ", vm->pid) < 0) {
        virReportOOMError();
        return -1;
    }

1305
    while (retries) {
1306
        ssize_t func_ret, bytes;
1307
        int isdead = 0;
1308
        char *eol;
1309 1310 1311 1312 1313 1314 1315 1316

        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. */
1317 1318
        bytes = saferead(fd, buf+got, buflen-got-1);
        if (bytes < 0) {
1319 1320 1321
            virReportSystemError(errno,
                                 _("Failure while reading %s log output"),
                                 what);
1322
            goto cleanup;
1323 1324
        }

1325
        got += bytes;
1326
        buf[got] = '\0';
1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339

        /* Filter out debug messages from intermediate libvirt process */
        while ((eol = strchr(filter_next, '\n'))) {
            *eol = '\0';
            if (strstr(filter_next, debug)) {
                memmove(filter_next, eol + 1, got - (eol - buf));
                got -= eol + 1 - filter_next;
            } else {
                filter_next = eol + 1;
                *eol = '\n';
            }
        }

1340
        if (got == buflen-1) {
1341 1342 1343
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Out of space while reading %s log output: %s"),
                           what, buf);
1344
            goto cleanup;
1345 1346 1347
        }

        if (isdead) {
1348 1349 1350
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Process exited while reading %s log output: %s"),
                           what, buf);
1351
            goto cleanup;
1352 1353
        }

1354 1355 1356 1357
        if (func_ret <= 0) {
            ret = func_ret;
            goto cleanup;
        }
1358 1359 1360 1361 1362

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

1363 1364 1365
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("Timed out while reading %s log output: %s"),
                   what, buf);
1366 1367 1368 1369

cleanup:
    VIR_FREE(debug);
    return ret;
1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424
}


/*
 * 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)
{
    static const char needle[] = "char device redirected to";
    char *tmp, *dev;

    VIR_FREE(*path);
    /* First look for our magic string */
    if (!(tmp = strstr(haystack + *offset, needle))) {
        return 1;
    }
    tmp += sizeof(needle);
    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;
}

1425 1426 1427
static int
qemuProcessLookupPTYs(virDomainChrDefPtr *devices,
                      int count,
C
Cole Robinson 已提交
1428 1429
                      virHashTablePtr paths,
                      bool chardevfmt)
1430 1431
{
    int i;
C
Cole Robinson 已提交
1432
    const char *prefix = chardevfmt ? "char" : "";
1433 1434 1435 1436

    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
1437
            char id[32];
1438 1439
            const char *path;

C
Cole Robinson 已提交
1440 1441
            if (snprintf(id, sizeof(id), "%s%s",
                         prefix, chr->info.alias) >= sizeof(id))
1442 1443 1444 1445 1446 1447 1448 1449
                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
                     */
1450 1451
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473
                    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;
}

1474 1475
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
1476
                                     qemuCapsPtr caps,
1477 1478
                                     virHashTablePtr paths)
{
1479
    bool chardevfmt = qemuCapsGet(caps, QEMU_CAPS_CHARDEV);
C
Cole Robinson 已提交
1480

1481
    if (qemuProcessLookupPTYs(vm->def->serials, vm->def->nserials,
C
Cole Robinson 已提交
1482
                              paths, chardevfmt) < 0)
1483 1484 1485
        return -1;

    if (qemuProcessLookupPTYs(vm->def->parallels, vm->def->nparallels,
C
Cole Robinson 已提交
1486
                              paths, chardevfmt) < 0)
1487
        return -1;
1488

1489
    if (qemuProcessLookupPTYs(vm->def->channels, vm->def->nchannels,
C
Cole Robinson 已提交
1490
                              paths, chardevfmt) < 0)
1491 1492
        return -1;

1493 1494
    if (qemuProcessLookupPTYs(vm->def->consoles, vm->def->nconsoles,
                              paths, chardevfmt) < 0)
1495
        return -1;
1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541

    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;
        }
    }

1542 1543
    for (i = 0 ; i < vm->def->nconsoles ; i++) {
        virDomainChrDefPtr chr = vm->def->consoles[i];
1544 1545 1546 1547 1548 1549 1550
        /* 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)
1551
                return ret;
1552 1553 1554 1555 1556 1557 1558
        } 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;
            }
1559 1560 1561
        }
    }

1562 1563 1564
    return 0;
}

1565
static void qemuProcessFreePtyPath(void *payload, const void *name ATTRIBUTE_UNUSED)
1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583
{
    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';
}

1584

1585 1586
static int
qemuProcessWaitForMonitor(struct qemud_driver* driver,
C
Cole Robinson 已提交
1587
                          virDomainObjPtr vm,
1588
                          qemuCapsPtr caps,
C
Cole Robinson 已提交
1589
                          off_t pos)
1590
{
1591
    char *buf = NULL;
1592
    size_t buf_size = 4096; /* Plenty of space to get startup greeting */
1593
    int logfd = -1;
1594 1595
    int ret = -1;
    virHashTablePtr paths = NULL;
1596
    qemuDomainObjPrivatePtr priv;
1597

1598 1599 1600
    if (pos != -1) {
        if ((logfd = qemuDomainOpenLog(driver, vm, pos)) < 0)
            return -1;
1601

1602 1603
        if (VIR_ALLOC_N(buf, buf_size) < 0) {
            virReportOOMError();
1604
            goto closelog;
1605
        }
1606

1607 1608 1609 1610 1611
        if (qemuProcessReadLogOutput(vm, logfd, buf, buf_size,
                                     qemuProcessFindCharDevicePTYs,
                                     "console", 30) < 0)
            goto closelog;
    }
1612 1613 1614 1615 1616 1617 1618 1619 1620 1621

    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. */
1622
    paths = virHashCreate(0, qemuProcessFreePtyPath);
1623
    if (paths == NULL)
1624 1625
        goto cleanup;

1626
    priv = vm->privateData;
1627
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
1628 1629 1630 1631 1632
    ret = qemuMonitorGetPtyPaths(priv->mon, paths);
    qemuDomainObjExitMonitorWithDriver(driver, vm);

    VIR_DEBUG("qemuMonitorGetPtyPaths returned %i", ret);
    if (ret == 0)
1633
        ret = qemuProcessFindCharDevicePTYsMonitor(vm, caps, paths);
1634 1635

cleanup:
1636
    virHashFree(paths);
1637

E
Eric Blake 已提交
1638
    if (pos != -1 && kill(vm->pid, 0) == -1 && errno == ESRCH) {
1639 1640
        /* VM is dead, any other error raised in the interim is probably
         * not as important as the qemu cmdline output */
1641
        qemuProcessReadLogFD(logfd, buf, buf_size, strlen(buf));
1642 1643 1644
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("process exited while connecting to monitor: %s"),
                       buf);
1645 1646 1647
        ret = -1;
    }

1648
closelog:
1649
    if (VIR_CLOSE(logfd) < 0) {
1650
        char ebuf[1024];
1651
        VIR_WARN("Unable to close logfile: %s",
1652
                 virStrerror(errno, ebuf, sizeof(ebuf)));
1653 1654
    }

1655 1656
    VIR_FREE(buf);

1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667
    return ret;
}

static int
qemuProcessDetectVcpuPIDs(struct qemud_driver *driver,
                          virDomainObjPtr vm)
{
    pid_t *cpupids = NULL;
    int ncpupids;
    qemuDomainObjPrivatePtr priv = vm->privateData;

1668 1669 1670 1671 1672 1673 1674 1675
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
    /* 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) {
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        virResetLastError();

1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
        priv->nvcpupids = 1;
        if (VIR_ALLOC_N(priv->vcpupids, priv->nvcpupids) < 0) {
            virReportOOMError();
            return -1;
        }
        priv->vcpupids[0] = vm->pid;
        return 0;
    }
    qemuDomainObjExitMonitorWithDriver(driver, vm);

    if (ncpupids != vm->def->vcpus) {
1687 1688 1689 1690
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of vCPU pids from QEMU monitor. "
                         "got %d, wanted %d"),
                       ncpupids, vm->def->vcpus);
1691 1692 1693 1694 1695 1696 1697 1698 1699
        VIR_FREE(cpupids);
        return -1;
    }

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

1700 1701 1702 1703 1704 1705 1706

/*
 * Set NUMA memory policy for qemu process, to be run between
 * fork/exec of QEMU only.
 */
#if HAVE_NUMACTL
static int
1707
qemuProcessInitNumaMemoryPolicy(virDomainObjPtr vm,
1708
                                virBitmapPtr nodemask)
1709
{
E
Eric Blake 已提交
1710
    nodemask_t mask;
1711 1712 1713 1714 1715
    int mode = -1;
    int node = -1;
    int ret = -1;
    int i = 0;
    int maxnode = 0;
E
Eric Blake 已提交
1716
    bool warned = false;
1717
    virDomainNumatuneDef numatune = vm->def->numatune;
1718
    virBitmapPtr tmp_nodemask = NULL;
1719

1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730
    if (numatune.memory.placement_mode ==
        VIR_DOMAIN_NUMATUNE_MEM_PLACEMENT_MODE_STATIC) {
        if (!numatune.memory.nodemask)
            return 0;
        VIR_DEBUG("Set NUMA memory policy with specified nodeset");
        tmp_nodemask = numatune.memory.nodemask;
    } else if (numatune.memory.placement_mode ==
               VIR_DOMAIN_NUMATUNE_MEM_PLACEMENT_MODE_AUTO) {
        VIR_DEBUG("Set NUMA memory policy with advisory nodeset from numad");
        tmp_nodemask = nodemask;
    } else {
1731
        return 0;
1732
    }
1733 1734

    if (numa_available() < 0) {
1735 1736
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("Host kernel is not aware of NUMA."));
1737 1738 1739
        return -1;
    }

E
Eric Blake 已提交
1740
    maxnode = numa_max_node() + 1;
1741
    /* Convert nodemask to NUMA bitmask. */
E
Eric Blake 已提交
1742
    nodemask_zero(&mask);
1743 1744 1745 1746 1747 1748
    i = -1;
    while ((i = virBitmapNextSetBit(tmp_nodemask, i)) >= 0) {
        if (i > NUMA_NUM_NODES) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Host cannot support NUMA node %d"), i);
            return -1;
1749
        }
1750 1751 1752 1753 1754 1755
        if (i > maxnode && !warned) {
            VIR_WARN("nodeset is out of range, there is only %d NUMA "
                     "nodes on host", maxnode);
            warned = true;
        }
        nodemask_set(&mask, i);
1756 1757
    }

1758
    mode = numatune.memory.mode;
1759 1760 1761

    if (mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT) {
        numa_set_bind_policy(1);
E
Eric Blake 已提交
1762
        numa_set_membind(&mask);
1763 1764 1765
        numa_set_bind_policy(0);
    } else if (mode == VIR_DOMAIN_NUMATUNE_MEM_PREFERRED) {
        int nnodes = 0;
E
Eric Blake 已提交
1766 1767
        for (i = 0; i < NUMA_NUM_NODES; i++) {
            if (nodemask_isset(&mask, i)) {
1768 1769 1770 1771 1772 1773
                node = i;
                nnodes++;
            }
        }

        if (nnodes != 1) {
1774 1775 1776
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("NUMA memory tuning in 'preferred' mode "
                                   "only supports single node"));
1777 1778 1779 1780 1781 1782
            goto cleanup;
        }

        numa_set_bind_policy(0);
        numa_set_preferred(node);
    } else if (mode == VIR_DOMAIN_NUMATUNE_MEM_INTERLEAVE) {
E
Eric Blake 已提交
1783
        numa_set_interleave_mask(&mask);
1784 1785 1786 1787
    } else {
        /* XXX: Shouldn't go here, as we already do checking when
         * parsing domain XML.
         */
1788 1789
        virReportError(VIR_ERR_XML_ERROR,
                       "%s", _("Invalid mode for memory NUMA tuning."));
1790 1791 1792 1793 1794 1795 1796 1797 1798 1799
        goto cleanup;
    }

    ret = 0;

cleanup:
    return ret;
}
#else
static int
E
Eric Blake 已提交
1800
qemuProcessInitNumaMemoryPolicy(virDomainObjPtr vm,
1801
                                virBitmapPtr nodemask ATTRIBUTE_UNUSED)
1802 1803
{
    if (vm->def->numatune.memory.nodemask) {
1804 1805
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("libvirt is compiled without NUMA tuning support"));
1806 1807 1808 1809 1810 1811 1812 1813

        return -1;
    }

    return 0;
}
#endif

1814
#if HAVE_NUMAD
O
Osier Yang 已提交
1815 1816 1817 1818 1819 1820
static char *
qemuGetNumadAdvice(virDomainDefPtr def)
{
    virCommandPtr cmd = NULL;
    char *output = NULL;

1821 1822
    cmd = virCommandNewArgList(NUMAD, "-w", NULL);
    virCommandAddArgFormat(cmd, "%d:%llu", def->vcpus,
1823
                           VIR_DIV_UP(def->mem.cur_balloon, 1024));
O
Osier Yang 已提交
1824 1825 1826 1827

    virCommandSetOutputBuffer(cmd, &output);

    if (virCommandRun(cmd, NULL) < 0)
1828 1829 1830
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Failed to query numad for the "
                         "advisory nodeset"));
O
Osier Yang 已提交
1831 1832 1833 1834 1835 1836 1837 1838

    virCommandFree(cmd);
    return output;
}
#else
static char *
qemuGetNumadAdvice(virDomainDefPtr def ATTRIBUTE_UNUSED)
{
1839 1840
    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                   _("numad is not available on this host"));
O
Osier Yang 已提交
1841 1842 1843 1844
    return NULL;
}
#endif

1845 1846 1847 1848
/*
 * To be run between fork/exec of QEMU only
 */
static int
O
Osier Yang 已提交
1849
qemuProcessInitCpuAffinity(struct qemud_driver *driver,
1850
                           virDomainObjPtr vm,
1851
                           virBitmapPtr nodemask)
1852
{
1853
    int ret = -1;
1854 1855
    int i, hostcpus, maxcpu = QEMUD_CPUMASK_LEN;
    virNodeInfo nodeinfo;
H
Hu Tao 已提交
1856
    virBitmapPtr cpumap, cpumapToSet;
1857

1858
    VIR_DEBUG("Setting CPU affinity");
1859 1860 1861 1862 1863 1864 1865 1866 1867 1868

    if (nodeGetInfo(NULL, &nodeinfo) < 0)
        return -1;

    /* setaffinity fails if you set bits for CPUs which
     * aren't present, so we have to limit ourselves */
    hostcpus = VIR_NODEINFO_MAXCPUS(nodeinfo);
    if (maxcpu > hostcpus)
        maxcpu = hostcpus;

1869 1870
    cpumap = virBitmapNew(maxcpu);
    if (!cpumap) {
1871 1872 1873 1874
        virReportOOMError();
        return -1;
    }

H
Hu Tao 已提交
1875 1876
    cpumapToSet = cpumap;

O
Osier Yang 已提交
1877
    if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) {
1878
        VIR_DEBUG("Set CPU affinity with advisory nodeset from numad");
1879 1880 1881 1882
        /* numad returns the NUMA node list, convert it to cpumap */
        for (i = 0; i < driver->caps->host.nnumaCell; i++) {
            int j;
            int cur_ncpus = driver->caps->host.numaCell[i]->ncpus;
1883 1884 1885 1886
            bool result;
            if (virBitmapGetBit(nodemask, i, &result) < 0)
                goto cleanup;
            if (result) {
1887
                for (j = 0; j < cur_ncpus; j++)
1888 1889
                    ignore_value(virBitmapSetBit(cpumap,
                                                 driver->caps->host.numaCell[i]->cpus[j]));
1890
            }
O
Osier Yang 已提交
1891
        }
1892
    } else {
1893
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
1894
        if (vm->def->cpumask) {
H
Hu Tao 已提交
1895
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
1896 1897 1898 1899 1900
        } else {
            /* You may think this is redundant, but we can't assume libvirtd
             * itself is running on all pCPUs, so we need to explicitly set
             * the spawned QEMU instance to all pCPUs if no map is given in
             * its config file */
1901
            virBitmapSetAll(cpumap);
O
Osier Yang 已提交
1902
        }
1903 1904 1905 1906 1907 1908
    }

    /* 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
     */
H
Hu Tao 已提交
1909
    if (virProcessInfoSetAffinity(0 /* Self */, cpumapToSet) < 0)
1910
        goto cleanup;
1911

1912 1913 1914
    ret = 0;

cleanup:
1915
    virBitmapFree(cpumap);
1916
    return ret;
1917 1918
}

1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
/* 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);

1932
            if (!qemuCapsGet(priv->caps, QEMU_CAPS_NETDEV)) {
1933 1934
                virReportError(VIR_ERR_NO_SUPPORT, "%s",
                               _("Setting of link state is not supported by this qemu"));
1935 1936 1937 1938 1939 1940 1941
                return -1;
            }

            ret = qemuMonitorSetLink(priv->mon,
                                     def->nets[i]->info.alias,
                                     VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (ret != 0) {
1942
                virReportError(VIR_ERR_OPERATION_FAILED,
1943 1944 1945 1946 1947 1948 1949 1950 1951 1952
                               _("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. */
1953 1954 1955 1956 1957 1958 1959
static int
qemuProcessSetVcpuAffinites(virConnectPtr conn,
                            virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
    virNodeInfo nodeinfo;
1960
    int vcpu, n;
E
Eric Blake 已提交
1961
    int ret = -1;
1962 1963 1964 1965 1966 1967 1968 1969 1970

    if (virNodeGetInfo(conn, &nodeinfo) != 0) {
        return  -1;
    }

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

    if (priv->vcpupids == NULL) {
1971 1972
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("cpu affinity is not supported"));
1973 1974 1975
        return -1;
    }

1976 1977
    for (n = 0; n < def->cputune.nvcpupin; n++) {
        vcpu = def->cputune.vcpupin[n]->vcpuid;
1978

1979 1980
        if (virProcessInfoSetAffinity(priv->vcpupids[vcpu],
                                      def->cputune.vcpupin[n]->cpumask) < 0) {
E
Eric Blake 已提交
1981
            goto cleanup;
1982 1983 1984
        }
    }

E
Eric Blake 已提交
1985 1986 1987
    ret = 0;
cleanup:
    return ret;
1988
}
1989

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
/* Set CPU affinities for emulator threads if emulatorpin xml provided. */
static int
qemuProcessSetEmulatorAffinites(virConnectPtr conn,
                                virDomainObjPtr vm)
{
    virDomainDefPtr def = vm->def;
    virNodeInfo nodeinfo;
    int ret = -1;

    if (virNodeGetInfo(conn, &nodeinfo) != 0)
        return -1;

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

2005 2006
    if (virProcessInfoSetAffinity(vm->pid,
                                  def->cputune.emulatorpin->cpumask) < 0) {
2007 2008 2009 2010 2011 2012 2013 2014
        goto cleanup;
    }

    ret = 0;
cleanup:
    return ret;
}

2015 2016 2017
static int
qemuProcessInitPasswords(virConnectPtr conn,
                         struct qemud_driver *driver,
2018
                         virDomainObjPtr vm)
2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039
{
    int ret = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (vm->def->ngraphics == 1) {
        if (vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_VNC,
                                                    &vm->def->graphics[0]->data.vnc.auth,
                                                    driver->vncPassword);
        } else if (vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_SPICE,
                                                    &vm->def->graphics[0]->data.spice.auth,
                                                    driver->spicePassword);
        }
    }

    if (ret < 0)
        goto cleanup;

2040
    if (qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE)) {
2041 2042 2043 2044 2045
        int i;

        for (i = 0 ; i < vm->def->ndisks ; i++) {
            char *secret;
            size_t secretLen;
E
Eric Blake 已提交
2046
            const char *alias;
2047 2048 2049 2050 2051 2052 2053 2054 2055 2056

            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
2057
            alias = vm->def->disks[i]->info.alias;
2058
            qemuDomainObjEnterMonitorWithDriver(driver, vm);
E
Eric Blake 已提交
2059
            ret = qemuMonitorSetDrivePassphrase(priv->mon, alias, secret);
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 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339
            VIR_FREE(secret);
            qemuDomainObjExitMonitorWithDriver(driver, vm);
            if (ret < 0)
                goto cleanup;
        }
    }

cleanup:
    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) {
2340 2341 2342
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for VirtIO disk %s"),
                           vm->def->disks[i]->dst);
2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353
            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) {
2354 2355 2356
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for %s NIC"),
                           vm->def->nets[i]->model);
2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367
            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) {
2368 2369 2370
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for controller %s"),
                           virDomainControllerTypeToString(vm->def->controllers[i]->type));
2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381
            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) {
2382 2383 2384
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for video adapter %s"),
                           virDomainVideoTypeToString(vm->def->videos[i]->type));
2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395
            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) {
2396 2397 2398
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for sound adapter %s"),
                           virDomainSoundModelTypeToString(vm->def->sounds[i]->model));
2399 2400 2401 2402 2403 2404 2405 2406 2407 2408
            return -1;
        }
    }


    if (vm->def->watchdog &&
        qemuProcessGetPCIWatchdogVendorProduct(vm->def->watchdog, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->watchdog->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2409 2410 2411
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for watchdog %s"),
                           virDomainWatchdogModelTypeToString(vm->def->watchdog->model));
2412 2413 2414 2415 2416 2417 2418 2419 2420
            return -1;
        }
    }

    if (vm->def->memballoon &&
        qemuProcessGetPCIMemballoonVendorProduct(vm->def->memballoon, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->memballoon->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2421 2422 2423
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for balloon %s"),
                           virDomainMemballoonModelTypeToString(vm->def->memballoon->model));
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
            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
qemuProcessInitPCIAddresses(struct qemud_driver *driver,
                            virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int naddrs;
    int ret;
    qemuMonitorPCIAddress *addrs = NULL;

2449
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466
    naddrs = qemuMonitorGetAllPCIAddresses(priv->mon,
                                           &addrs);
    qemuDomainObjExitMonitorWithDriver(driver, vm);

    ret = qemuProcessDetectPCIAddresses(vm, addrs, naddrs);

    VIR_FREE(addrs);

    return ret;
}


static int qemuProcessNextFreePort(struct qemud_driver *driver,
                                   int startPort)
{
    int i;

2467
    for (i = startPort ; i < driver->remotePortMax; i++) {
2468 2469 2470 2471 2472
        int fd;
        int reuse = 1;
        struct sockaddr_in addr;
        bool used = false;

2473
        if (virBitmapGetBit(driver->reservedRemotePorts,
2474 2475
                            i - driver->remotePortMin, &used) < 0)
            VIR_DEBUG("virBitmapGetBit failed on bit %d", i - driver->remotePortMin);
2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495

        if (used)
            continue;

        addr.sin_family = AF_INET;
        addr.sin_port = htons(i);
        addr.sin_addr.s_addr = htonl(INADDR_ANY);
        fd = socket(PF_INET, SOCK_STREAM, 0);
        if (fd < 0)
            return -1;

        if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (void*)&reuse, sizeof(reuse)) < 0) {
            VIR_FORCE_CLOSE(fd);
            break;
        }

        if (bind(fd, (struct sockaddr*)&addr, sizeof(addr)) == 0) {
            /* Not in use, lets grab it */
            VIR_FORCE_CLOSE(fd);
            /* Add port to bitmap of reserved ports */
2496
            if (virBitmapSetBit(driver->reservedRemotePorts,
2497
                                i - driver->remotePortMin) < 0) {
2498
                VIR_DEBUG("virBitmapSetBit failed on bit %d",
2499
                          i - driver->remotePortMin);
2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519
            }
            return i;
        }
        VIR_FORCE_CLOSE(fd);

        if (errno == EADDRINUSE) {
            /* In use, try next */
            continue;
        }
        /* Some other bad failure, get out.. */
        break;
    }
    return -1;
}


static void
qemuProcessReturnPort(struct qemud_driver *driver,
                      int port)
{
2520
    if (port < driver->remotePortMin)
2521 2522
        return;

2523
    if (virBitmapClearBit(driver->reservedRemotePorts,
2524
                          port - driver->remotePortMin) < 0)
2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551
        VIR_DEBUG("Could not mark port %d as unused", port);
}


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;
}


2552 2553 2554
static int
qemuProcessLimits(struct qemud_driver *driver)
{
2555
    struct rlimit rlim;
2556

2557
    if (driver->maxProcesses > 0) {
2558 2559 2560 2561 2562 2563 2564 2565 2566
        rlim.rlim_cur = rlim.rlim_max = driver->maxProcesses;
        if (setrlimit(RLIMIT_NPROC, &rlim) < 0) {
            virReportSystemError(errno,
                                 _("cannot limit number of processes to %d"),
                                 driver->maxProcesses);
            return -1;
        }
    }

2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578
    if (driver->maxFiles > 0) {
        /* Max number of opened files is one greater than
         * actual limit. See man setrlimit */
        rlim.rlim_cur = rlim.rlim_max = driver->maxFiles + 1;
        if (setrlimit(RLIMIT_NOFILE, &rlim) < 0) {
            virReportSystemError(errno,
                                 _("cannot set max opened files to %d"),
                                 driver->maxFiles);
            return -1;
        }
    }

2579 2580 2581 2582
    return 0;
}


2583 2584 2585 2586
struct qemuProcessHookData {
    virConnectPtr conn;
    virDomainObjPtr vm;
    struct qemud_driver *driver;
2587
    virBitmapPtr nodemask;
2588 2589 2590 2591 2592
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
2593
    int ret = -1;
2594
    int fd;
2595 2596 2597 2598 2599

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

    VIR_DEBUG("Obtaining domain lock");
2600 2601 2602 2603 2604 2605 2606
    /*
     * 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.
     */
2607
    if (virSecurityManagerSetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2608
        goto cleanup;
2609 2610 2611
    if (virDomainLockProcessStart(h->driver->lockManager,
                                  h->vm,
                                  /* QEMU is always pased initially */
2612 2613
                                  true,
                                  &fd) < 0)
2614
        goto cleanup;
2615
    if (virSecurityManagerClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2616
        goto cleanup;
2617

2618
    if (qemuProcessLimits(h->driver) < 0)
2619
        goto cleanup;
2620

2621 2622 2623
    /* This must take place before exec(), so that all QEMU
     * memory allocation is on the correct NUMA node
     */
2624
    VIR_DEBUG("Moving process to cgroup");
2625
    if (qemuAddToCgroup(h->driver, h->vm->def) < 0)
2626
        goto cleanup;
2627 2628 2629

    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
2630
    if (qemuProcessInitCpuAffinity(h->driver, h->vm, h->nodemask) < 0)
2631
        goto cleanup;
2632

2633
    if (qemuProcessInitNumaMemoryPolicy(h->vm, h->nodemask) < 0)
2634
        goto cleanup;
2635

2636
    VIR_DEBUG("Setting up security labelling");
2637
    if (virSecurityManagerSetProcessLabel(h->driver->securityManager, h->vm->def) < 0)
2638
        goto cleanup;
2639

2640 2641 2642 2643 2644
    ret = 0;

cleanup:
    VIR_DEBUG("Hook complete ret=%d", ret);
    return ret;
2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664
}

int
qemuProcessPrepareMonitorChr(struct qemud_driver *driver,
                             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",
                    driver->libDir, vm) < 0) {
        virReportOOMError();
        return -1;
    }

    return 0;
}


2665 2666 2667 2668 2669
/*
 * Precondition: Both driver and vm must be locked,
 * and a job must be active. This method will call
 * {Enter,Exit}MonitorWithDriver
 */
E
Eric Blake 已提交
2670 2671
int
qemuProcessStartCPUs(struct qemud_driver *driver, virDomainObjPtr vm,
2672 2673
                     virConnectPtr conn, virDomainRunningReason reason,
                     enum qemuDomainAsyncJob asyncJob)
2674 2675 2676 2677
{
    int ret;
    qemuDomainObjPrivatePtr priv = vm->privateData;

2678 2679 2680 2681 2682 2683 2684 2685 2686 2687
    VIR_DEBUG("Using lock state '%s'", NULLSTR(priv->lockState));
    if (virDomainLockProcessResume(driver->lockManager, 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
         */
        return -1;
    }
    VIR_FREE(priv->lockState);

2688 2689 2690 2691 2692
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
    if (ret == 0) {
        ret = qemuMonitorStartCPUs(priv->mon, conn);
        qemuDomainObjExitMonitorWithDriver(driver, vm);
    }
J
Jiri Denemark 已提交
2693

2694
    if (ret == 0) {
J
Jiri Denemark 已提交
2695
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
2696 2697 2698 2699 2700
    } 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));
    }
2701 2702 2703 2704 2705

    return ret;
}


J
Jiri Denemark 已提交
2706
int qemuProcessStopCPUs(struct qemud_driver *driver, virDomainObjPtr vm,
2707 2708
                        virDomainPausedReason reason,
                        enum qemuDomainAsyncJob asyncJob)
2709 2710 2711 2712
{
    int ret;
    qemuDomainObjPrivatePtr priv = vm->privateData;

2713
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
2714

2715
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
2716 2717 2718 2719
    if (ret == 0) {
        ret = qemuMonitorStopCPUs(priv->mon);
        qemuDomainObjExitMonitorWithDriver(driver, vm);
    }
J
Jiri Denemark 已提交
2720

2721
    if (ret == 0) {
2722
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
2723 2724 2725 2726
        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
2727

2728 2729 2730 2731 2732
    return ret;
}



2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745
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;
}

2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758
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)) {
2759
           if (virDomainConfNWFilterInstantiate(conn, def->uuid, net) < 0) {
2760 2761 2762 2763 2764 2765 2766 2767 2768
                err = 1;
                break;
            }
        }
    }

    return err;
}

2769 2770 2771 2772 2773
static int
qemuProcessUpdateState(struct qemud_driver *driver, virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
2774
    virDomainPausedReason reason;
2775 2776
    virDomainState newState = VIR_DOMAIN_NOSTATE;
    int newReason;
2777
    bool running;
2778
    char *msg = NULL;
2779 2780
    int ret;

2781
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
2782
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
2783 2784 2785 2786 2787 2788 2789 2790
    qemuDomainObjExitMonitorWithDriver(driver, vm);

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

    state = virDomainObjGetState(vm, NULL);

    if (state == VIR_DOMAIN_PAUSED && running) {
2791 2792 2793
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
        msg = strdup("was unpaused");
2794
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
2795 2796 2797 2798 2799 2800 2801
        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
2802 2803
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
2804
        }
2805
    } else if (state == VIR_DOMAIN_SHUTOFF && running) {
2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824
        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);
2825 2826 2827 2828 2829
    }

    return 0;
}

2830 2831 2832 2833 2834 2835 2836 2837 2838
static int
qemuProcessRecoverMigration(struct qemud_driver *driver,
                            virDomainObjPtr vm,
                            virConnectPtr conn,
                            enum qemuDomainAsyncJob job,
                            enum qemuMigrationJobPhase phase,
                            virDomainState state,
                            int reason)
{
2839 2840
    qemuDomainObjPrivatePtr priv = vm->privateData;

2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863
    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,
2864 2865
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895
                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);
2896
            qemuDomainObjEnterMonitor(driver, vm);
2897 2898
            ignore_value(qemuMonitorMigrateCancel(priv->mon));
            qemuDomainObjExitMonitor(driver, vm);
2899 2900 2901 2902 2903 2904
            /* 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,
2905 2906
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925
                    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,
2926 2927
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941
                    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;
}

2942 2943 2944 2945 2946 2947
static int
qemuProcessRecoverJob(struct qemud_driver *driver,
                      virDomainObjPtr vm,
                      virConnectPtr conn,
                      const struct qemuDomainJobObj *job)
{
2948
    qemuDomainObjPrivatePtr priv = vm->privateData;
2949 2950 2951 2952 2953 2954 2955 2956
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
    case QEMU_ASYNC_JOB_MIGRATION_IN:
2957 2958 2959
        if (qemuProcessRecoverMigration(driver, vm, conn, job->asyncJob,
                                        job->phase, state, reason) < 0)
            return -1;
2960 2961 2962 2963
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
2964
        qemuDomainObjEnterMonitor(driver, vm);
2965 2966
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
        qemuDomainObjExitMonitor(driver, vm);
2967
        /* resume the domain but only if it was paused as a result of
2968 2969 2970
         * running save/dump operation.  Although we are recovering an
         * async job, this function is run at startup and must resume
         * things using sync monitor connections.  */
2971 2972 2973 2974 2975 2976 2977
        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) ||
             reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
            if (qemuProcessStartCPUs(driver, vm, conn,
2978 2979
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992
                VIR_WARN("Could not resume domain %s after", vm->def->name);
            }
        }
        break;

    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        break;
    }

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

2993 2994 2995 2996
    /* 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.
     */
2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016
    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;

3017
    case QEMU_JOB_MIGRATION_OP:
3018
    case QEMU_JOB_ABORT:
3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029
    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;
}

3030 3031 3032
struct qemuProcessReconnectData {
    virConnectPtr conn;
    struct qemud_driver *driver;
3033 3034
    void *payload;
    struct qemuDomainJobObj oldjob;
3035 3036 3037 3038
};
/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
S
Stefan Berger 已提交
3039 3040 3041 3042
 *
 * 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.
3043 3044
 */
static void
3045
qemuProcessReconnect(void *opaque)
3046 3047 3048
{
    struct qemuProcessReconnectData *data = opaque;
    struct qemud_driver *driver = data->driver;
3049
    virDomainObjPtr obj = data->payload;
3050 3051
    qemuDomainObjPrivatePtr priv;
    virConnectPtr conn = data->conn;
3052
    struct qemuDomainJobObj oldjob;
3053 3054
    int state;
    int reason;
3055

3056 3057 3058 3059 3060
    memcpy(&oldjob, &data->oldjob, sizeof(oldjob));

    VIR_FREE(data);

    qemuDriverLock(driver);
3061 3062
    virDomainObjLock(obj);

3063

3064 3065 3066 3067
    VIR_DEBUG("Reconnect monitor to %p '%s'", obj, obj->def->name);

    priv = obj->privateData;

3068 3069
    /* Job was started by the caller for us */
    qemuDomainObjTransferJob(obj);
3070

3071 3072
    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted if qemuConnectMonitor() failed */
3073
    virObjectRef(obj);
3074 3075 3076 3077 3078

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

D
Daniel P. Berrange 已提交
3079 3080 3081 3082 3083 3084 3085 3086
    /* 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;
    }

3087 3088 3089 3090
    if (qemuUpdateActivePciHostdevs(driver, obj->def) < 0) {
        goto error;
    }

3091 3092 3093
    if (qemuUpdateActiveUsbHostdevs(driver, obj->def) < 0)
        goto error;

3094 3095 3096
    if (qemuProcessUpdateState(driver, obj) < 0)
        goto error;

3097 3098
    state = virDomainObjGetState(obj, &reason);
    if (state == VIR_DOMAIN_SHUTOFF) {
3099 3100 3101 3102 3103
        VIR_DEBUG("Domain '%s' wasn't fully started yet, killing it",
                  obj->def->name);
        goto error;
    }

3104 3105 3106
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
3107
    if (!priv->caps &&
3108
        qemuCapsExtractVersionInfo(obj->def->emulator, obj->def->os.arch,
3109
                                   false,
3110
                                   NULL,
3111
                                   &priv->caps) < 0)
3112 3113
        goto error;

3114
    /* In case the domain shutdown while we were not running,
3115 3116 3117
     * we need to finish the shutdown process. And we need to do it after
     * we have qemuCaps filled in.
     */
3118 3119 3120 3121 3122 3123
    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);
3124 3125 3126
        goto endjob;
    }

3127 3128
    if (qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE))
        if ((qemuDomainAssignAddresses(obj->def, priv->caps, obj)) < 0)
3129
            goto error;
3130

3131
    if (virSecurityManagerReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
3132 3133
        goto error;

3134 3135 3136
    if (qemuProcessNotifyNets(obj->def) < 0)
        goto error;

3137 3138 3139
    if (qemuProcessFiltersInstantiate(conn, obj->def))
        goto error;

3140
    if (qemuDomainCheckEjectableMedia(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
3141 3142
        goto error;

3143 3144 3145
    if (qemuProcessRecoverJob(driver, obj, conn, &oldjob) < 0)
        goto error;

3146 3147 3148 3149
    /* update domain state XML with possibly updated state in virDomainObj */
    if (virDomainSaveStatus(driver->caps, driver->stateDir, obj) < 0)
        goto error;

3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml = qemuDomainDefFormatXML(driver, obj->def, 0, false);
        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;
    }

3167 3168 3169
    if (obj->def->id >= driver->nextvmid)
        driver->nextvmid = obj->def->id + 1;

3170
endjob:
3171
    if (!qemuDomainObjEndJob(driver, obj))
3172 3173
        obj = NULL;

3174
    if (obj && virObjectUnref(obj))
3175 3176
        virDomainObjUnlock(obj);

3177 3178
    qemuDriverUnlock(driver);

S
Stefan Berger 已提交
3179 3180
    virConnectClose(conn);

3181 3182 3183
    return;

error:
3184
    if (!qemuDomainObjEndJob(driver, obj))
3185 3186
        obj = NULL;

3187 3188
    if (obj) {
        if (!virDomainObjIsActive(obj)) {
3189
            if (virObjectUnref(obj))
3190 3191 3192 3193
                virDomainObjUnlock(obj);
            qemuDriverUnlock(driver);
            return;
        }
3194

3195
        if (virObjectUnref(obj)) {
3196
            /* We can't get the monitor back, so must kill the VM
3197 3198 3199
             * to remove danger of it ending up running twice if
             * user tries to start it again later
             */
3200
            if (qemuCapsGet(priv->caps, QEMU_CAPS_NO_SHUTDOWN)) {
3201 3202 3203 3204 3205 3206 3207 3208 3209
                /* 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;
            }
3210
            qemuProcessStop(driver, obj, state, 0);
3211
            if (!obj->persistent)
3212
                qemuDomainRemoveInactive(driver, obj);
3213 3214 3215
            else
                virDomainObjUnlock(obj);
        }
3216
    }
3217
    qemuDriverUnlock(driver);
S
Stefan Berger 已提交
3218 3219

    virConnectClose(conn);
3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269
}

static void
qemuProcessReconnectHelper(void *payload,
                           const void *name ATTRIBUTE_UNUSED,
                           void *opaque)
{
    virThread thread;
    struct qemuProcessReconnectData *src = opaque;
    struct qemuProcessReconnectData *data;
    virDomainObjPtr obj = payload;

    if (VIR_ALLOC(data) < 0) {
        virReportOOMError();
        return;
    }

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

    /* This iterator is called with driver being locked.
     * We create a separate thread to run qemuProcessReconnect in it.
     * However, qemuProcessReconnect needs to:
     * 1. lock driver
     * 2. just before monitor reconnect do lightweight MonitorEnter
     *    (increase VM refcount, unlock VM & driver)
     * 3. reconnect to monitor
     * 4. do lightweight MonitorExit (lock driver & VM)
     * 5. continue reconnect process
     * 6. EndJob
     * 7. unlock driver
     *
     * It is necessary to NOT hold driver lock for the entire run
     * of reconnect, otherwise we will get blocked if there is
     * unresponsive qemu.
     * However, iterating over hash table MUST be done on locked
     * driver.
     *
     * NB, we can't do normal MonitorEnter & MonitorExit because
     * these two lock the monitor lock, which does not exists in
     * this early phase.
     */

    virDomainObjLock(obj);

    qemuDomainObjRestoreJob(obj, &data->oldjob);

    if (qemuDomainObjBeginJobWithDriver(src->driver, obj, QEMU_JOB_MODIFY) < 0)
        goto error;

S
Stefan Berger 已提交
3270 3271 3272 3273 3274 3275
    /* 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);

3276
    if (virThreadCreate(&thread, true, qemuProcessReconnect, data) < 0) {
S
Stefan Berger 已提交
3277 3278 3279

        virConnectClose(data->conn);

3280 3281 3282
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not create thread. QEMU initialization "
                         "might be incomplete"));
3283
        if (!qemuDomainObjEndJob(src->driver, obj)) {
3284
            obj = NULL;
3285
        } else if (virObjectUnref(obj)) {
3286 3287
           /* We can't spawn a thread and thus connect to monitor.
            * Kill qemu */
3288
            qemuProcessStop(src->driver, obj, VIR_DOMAIN_SHUTOFF_FAILED, 0);
3289
            if (!obj->persistent)
3290
                qemuDomainRemoveInactive(src->driver, obj);
3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302
            else
                virDomainObjUnlock(obj);
        }
        goto error;
    }

    virDomainObjUnlock(obj);

    return;

error:
    VIR_FREE(data);
3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313
}

/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
qemuProcessReconnectAll(virConnectPtr conn, struct qemud_driver *driver)
{
3314
    struct qemuProcessReconnectData data = {.conn = conn, .driver = driver};
3315
    virHashForEach(driver->domains.objs, qemuProcessReconnectHelper, &data);
3316 3317 3318 3319 3320 3321 3322 3323
}

int qemuProcessStart(virConnectPtr conn,
                     struct qemud_driver *driver,
                     virDomainObjPtr vm,
                     const char *migrateFrom,
                     int stdin_fd,
                     const char *stdin_path,
3324
                     virDomainSnapshotObjPtr snapshot,
3325 3326
                     enum virNetDevVPortProfileOp vmop,
                     unsigned int flags)
3327 3328 3329 3330 3331 3332 3333 3334 3335
{
    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
3336
    unsigned long cur_balloon;
3337
    int i;
3338
    char *nodeset = NULL;
3339
    virBitmapPtr nodemask = NULL;
3340
    unsigned int stop_flags;
3341

3342 3343 3344 3345 3346 3347
    /* Okay, these are just internal flags,
     * but doesn't hurt to check */
    virCheckFlags(VIR_QEMU_PROCESS_START_COLD |
                  VIR_QEMU_PROCESS_START_PAUSED |
                  VIR_QEMU_PROCESS_START_AUTODESROY, -1);

3348 3349 3350 3351 3352 3353
    /* 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;

3354 3355 3356 3357
    hookData.conn = conn;
    hookData.vm = vm;
    hookData.driver = driver;

3358
    VIR_DEBUG("Beginning VM startup process");
3359 3360

    if (virDomainObjIsActive(vm)) {
3361 3362
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
3363 3364 3365 3366 3367 3368 3369
        return -1;
    }

    /* 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
     */
3370
    VIR_DEBUG("Setting current domain def as transient");
3371 3372 3373 3374
    if (virDomainObjSetDefTransient(driver->caps, vm, true) < 0)
        goto cleanup;

    vm->def->id = driver->nextvmid++;
3375
    qemuDomainSetFakeReboot(driver, vm, false);
3376
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, VIR_DOMAIN_SHUTOFF_UNKNOWN);
3377

3378
    /* Run an early hook to set-up missing devices */
3379
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3380
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0, false);
3381 3382 3383
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
3384 3385
                              VIR_HOOK_QEMU_OP_PREPARE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
3386 3387 3388 3389 3390 3391 3392 3393 3394
        VIR_FREE(xml);

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

3395
    /* Must be run before security labelling */
3396
    VIR_DEBUG("Preparing host devices");
3397 3398 3399
    if (qemuPrepareHostDevices(driver, vm->def) < 0)
        goto cleanup;

3400
    VIR_DEBUG("Preparing chr devices");
3401 3402 3403 3404 3405 3406 3407 3408
    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 */
3409
    VIR_DEBUG("Generating domain security label (if required)");
3410
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0) {
3411
        virDomainAuditSecurityLabel(vm, false);
3412 3413
        goto cleanup;
    }
3414
    virDomainAuditSecurityLabel(vm, true);
3415 3416 3417

    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
3418
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
3419 3420 3421 3422 3423 3424
    qemuRemoveCgroup(driver, vm, 1);

    if (vm->def->ngraphics == 1) {
        if (vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
            !vm->def->graphics[0]->data.vnc.socket &&
            vm->def->graphics[0]->data.vnc.autoport) {
3425
            int port = qemuProcessNextFreePort(driver, driver->remotePortMin);
3426
            if (port < 0) {
3427
                virReportError(VIR_ERR_INTERNAL_ERROR,
3428
                               "%s", _("Unable to find an unused port for VNC"));
3429 3430 3431
                goto cleanup;
            }
            vm->def->graphics[0]->data.vnc.port = port;
M
Michal Privoznik 已提交
3432 3433 3434 3435
        } else if (vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            int port = -1;
            if (vm->def->graphics[0]->data.spice.autoport ||
                vm->def->graphics[0]->data.spice.port == -1) {
3436
                port = qemuProcessNextFreePort(driver, driver->remotePortMin);
M
Michal Privoznik 已提交
3437 3438

                if (port < 0) {
3439
                    virReportError(VIR_ERR_INTERNAL_ERROR,
3440
                                   "%s", _("Unable to find an unused port for SPICE"));
M
Michal Privoznik 已提交
3441 3442 3443 3444
                    goto cleanup;
                }

                vm->def->graphics[0]->data.spice.port = port;
3445
            }
M
Michal Privoznik 已提交
3446 3447 3448 3449 3450
            if (driver->spiceTLS &&
                (vm->def->graphics[0]->data.spice.autoport ||
                 vm->def->graphics[0]->data.spice.tlsPort == -1)) {
                int tlsPort = qemuProcessNextFreePort(driver,
                                                      vm->def->graphics[0]->data.spice.port + 1);
3451
                if (tlsPort < 0) {
3452
                    virReportError(VIR_ERR_INTERNAL_ERROR,
3453
                                   "%s", _("Unable to find an unused port for SPICE TLS"));
3454 3455 3456 3457
                    qemuProcessReturnPort(driver, port);
                    goto cleanup;
                }

3458
                vm->def->graphics[0]->data.spice.tlsPort = tlsPort;
M
Michal Privoznik 已提交
3459
            }
3460
        }
3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481

        if (vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ||
            vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            virDomainGraphicsDefPtr graphics = vm->def->graphics[0];
            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;
                if (vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC)
                    graphics->listens[0].address = strdup(driver->vncListen);
                else
                    graphics->listens[0].address = strdup(driver->spiceListen);
                if (!graphics->listens[0].address) {
                    VIR_SHRINK_N(graphics->listens, graphics->nListens, 1);
                    virReportOOMError();
                    goto cleanup;
                }
            }
        }
3482 3483
    }

3484
    if (virFileMakePath(driver->logDir) < 0) {
3485 3486 3487 3488 3489 3490
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
                             driver->logDir);
        goto cleanup;
    }

3491
    VIR_DEBUG("Creating domain log file");
3492
    if ((logfile = qemuDomainCreateLog(driver, vm, false)) < 0)
3493 3494
        goto cleanup;

3495 3496 3497
    if (vm->def->virtType == VIR_DOMAIN_VIRT_KVM) {
        VIR_DEBUG("Checking for KVM availability");
        if (access("/dev/kvm", F_OK) != 0) {
3498 3499 3500 3501
            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."));
3502 3503 3504 3505
            goto cleanup;
        }
    }

3506
    VIR_DEBUG("Determining emulator version");
3507 3508
    virObjectUnref(priv->caps);
    priv->caps = NULL;
3509
    if (qemuCapsExtractVersionInfo(vm->def->emulator, vm->def->os.arch,
3510
                                   true,
3511
                                   NULL,
3512
                                   &priv->caps) < 0)
3513 3514
        goto cleanup;

3515
    if (qemuAssignDeviceAliases(vm->def, priv->caps) < 0)
3516 3517 3518
        goto cleanup;

    VIR_DEBUG("Checking for CDROM and floppy presence");
3519 3520
    if (qemuDomainCheckDiskPresence(driver, vm,
                                    flags & VIR_QEMU_PROCESS_START_COLD) < 0)
3521 3522
        goto cleanup;

3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535
    /* 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 ==
         VIR_DOMAIN_NUMATUNE_MEM_PLACEMENT_MODE_AUTO)) {
        nodeset = qemuGetNumadAdvice(vm->def);
        if (!nodeset)
            goto cleanup;

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

3536 3537
        if (virBitmapParse(nodeset, 0, &nodemask,
                           VIR_DOMAIN_CPUMASK_LEN) < 0)
3538 3539 3540 3541
            goto cleanup;
    }
    hookData.nodemask = nodemask;

3542
    VIR_DEBUG("Setting up domain cgroup (if required)");
3543
    if (qemuSetupCgroup(driver, vm, nodemask) < 0)
3544 3545 3546 3547 3548 3549 3550
        goto cleanup;

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

3551
    VIR_DEBUG("Preparing monitor state");
3552 3553 3554
    if (qemuProcessPrepareMonitorChr(driver, priv->monConfig, vm->def->name) < 0)
        goto cleanup;

3555
    if (qemuCapsGet(priv->caps, QEMU_CAPS_MONITOR_JSON))
3556 3557 3558 3559
        priv->monJSON = 1;
    else
        priv->monJSON = 0;

3560 3561
    priv->monError = false;
    priv->monStart = 0;
3562 3563
    priv->gotShutdown = false;

3564
    VIR_FREE(priv->pidfile);
3565
    if (!(priv->pidfile = virPidFileBuildPath(driver->stateDir, vm->def->name))) {
3566 3567
        virReportSystemError(errno,
                             "%s", _("Failed to build pidfile path."));
3568 3569 3570
        goto cleanup;
    }

3571 3572
    if (unlink(priv->pidfile) < 0 &&
        errno != ENOENT) {
3573
        virReportSystemError(errno,
3574 3575
                             _("Cannot remove stale PID file %s"),
                             priv->pidfile);
3576 3577 3578 3579 3580 3581 3582 3583 3584 3585
        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
     */
3586
    if (qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE)) {
3587
        VIR_DEBUG("Assigning domain PCI addresses");
3588
        if ((qemuDomainAssignAddresses(vm->def, priv->caps, vm)) < 0)
3589
            goto cleanup;
3590 3591
    }

3592
    VIR_DEBUG("Building emulator command line");
3593
    if (!(cmd = qemuBuildCommandLine(conn, driver, vm->def, priv->monConfig,
3594
                                     priv->monJSON != 0, priv->caps,
3595
                                     migrateFrom, stdin_fd, snapshot, vmop)))
3596 3597 3598 3599
        goto cleanup;

    /* now that we know it is about to start call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3600
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0, false);
3601 3602 3603
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
3604 3605
                              VIR_HOOK_QEMU_OP_START, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
3606 3607 3608 3609 3610 3611 3612 3613 3614
        VIR_FREE(xml);

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

3615
    if ((timestamp = virTimeStringNow()) == NULL) {
3616 3617 3618 3619 3620 3621
        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",
3622
                     virStrerror(errno, ebuf, sizeof(ebuf)));
3623 3624 3625 3626 3627 3628 3629
        }

        VIR_FREE(timestamp);
    }

    virCommandWriteArgLog(cmd, logfile);

3630
    qemuDomainObjCheckTaint(driver, vm, logfile);
3631

3632 3633
    if ((pos = lseek(logfile, 0, SEEK_END)) < 0)
        VIR_WARN("Unable to seek to end of logfile: %s",
3634
                 virStrerror(errno, ebuf, sizeof(ebuf)));
3635 3636 3637 3638 3639 3640

    VIR_DEBUG("Clear emulator capabilities: %d",
              driver->clearEmulatorCapabilities);
    if (driver->clearEmulatorCapabilities)
        virCommandClearCaps(cmd);

3641 3642 3643 3644 3645 3646
    /* in case a certain disk is desirous of CAP_SYS_RAWIO, add this */
    for (i = 0; i < vm->def->ndisks; i++) {
        if (vm->def->disks[i]->rawio == 1)
            virCommandAllowCap(cmd, CAP_SYS_RAWIO);
    }

3647 3648 3649 3650 3651
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);

    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
3652
    virCommandSetPidFile(cmd, priv->pidfile);
3653
    virCommandDaemonize(cmd);
3654
    virCommandRequireHandshake(cmd);
3655 3656 3657

    ret = virCommandRun(cmd, NULL);

E
Eric Blake 已提交
3658
    /* wait for qemu process to show up */
3659
    if (ret == 0) {
3660
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
3661 3662
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
3663 3664 3665 3666 3667 3668 3669 3670 3671
            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.
         */

3672
        /* The virCommand process that launches the daemon failed. Pending on
3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683
         * 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
    }

3684 3685 3686 3687 3688
    VIR_DEBUG("Writing early domain status to disk");
    if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
        goto cleanup;
    }

3689 3690 3691 3692 3693 3694 3695
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
        goto cleanup;
    }

    VIR_DEBUG("Setting domain security labels");
    if (virSecurityManagerSetAllLabel(driver->securityManager,
3696
                                      vm->def, stdin_path) < 0)
3697 3698
        goto cleanup;

3699 3700 3701 3702 3703 3704
    /* 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;

3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718
    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) &&
3719
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def, stdin_fd) < 0)
3720 3721 3722 3723 3724 3725 3726 3727 3728
            goto cleanup;
    }

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

3729
    if (migrateFrom)
3730
        flags |= VIR_QEMU_PROCESS_START_PAUSED;
3731 3732 3733 3734 3735 3736 3737

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

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

3738
    VIR_DEBUG("Waiting for monitor to show up");
3739
    if (qemuProcessWaitForMonitor(driver, vm, priv->caps, pos) < 0)
3740 3741
        goto cleanup;

D
Daniel P. Berrange 已提交
3742 3743 3744 3745 3746 3747 3748 3749
    /* 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;
    }

3750
    VIR_DEBUG("Detecting VCPU PIDs");
3751 3752 3753
    if (qemuProcessDetectVcpuPIDs(driver, vm) < 0)
        goto cleanup;

3754
    VIR_DEBUG("Setting cgroup for each VCPU (if required)");
3755 3756 3757
    if (qemuSetupCgroupForVcpu(driver, vm) < 0)
        goto cleanup;

3758 3759 3760 3761
    VIR_DEBUG("Setting cgroup for emulator (if required)");
    if (qemuSetupCgroupForEmulator(driver, vm) < 0)
        goto cleanup;

3762
    VIR_DEBUG("Setting VCPU affinities");
3763 3764 3765
    if (qemuProcessSetVcpuAffinites(conn, vm) < 0)
        goto cleanup;

3766 3767 3768 3769
    VIR_DEBUG("Setting affinity of emulator threads");
    if (qemuProcessSetEmulatorAffinites(conn, vm) < 0)
        goto cleanup;

3770
    VIR_DEBUG("Setting any required VM passwords");
3771
    if (qemuProcessInitPasswords(conn, driver, vm) < 0)
3772 3773 3774 3775
        goto cleanup;

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
3776
    if (!qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE)) {
3777
        VIR_DEBUG("Determining domain device PCI addresses");
3778 3779 3780 3781
        if (qemuProcessInitPCIAddresses(driver, vm) < 0)
            goto cleanup;
    }

3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793
    /* set default link states */
    /* qemu doesn't support setting this on the command line, so
     * enter the monitor */
    VIR_DEBUG("Setting network link states");
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
    if (qemuProcessSetLinkStates(vm) < 0) {
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        goto cleanup;
    }

    qemuDomainObjExitMonitorWithDriver(driver, vm);

3794 3795 3796 3797
    /* 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.  */
3798
    VIR_DEBUG("Setting initial memory amount");
E
Eric Blake 已提交
3799
    cur_balloon = vm->def->mem.cur_balloon;
3800
    if (cur_balloon != vm->def->mem.cur_balloon) {
3801 3802 3803
        virReportError(VIR_ERR_OVERFLOW,
                       _("unable to set balloon to %lld"),
                       vm->def->mem.cur_balloon);
3804 3805
        goto cleanup;
    }
3806
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
E
Eric Blake 已提交
3807
    if (qemuMonitorSetBalloon(priv->mon, cur_balloon) < 0) {
3808 3809 3810 3811 3812
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        goto cleanup;
    }
    qemuDomainObjExitMonitorWithDriver(driver, vm);

3813
    if (!(flags & VIR_QEMU_PROCESS_START_PAUSED)) {
3814
        VIR_DEBUG("Starting domain CPUs");
3815
        /* Allow the CPUS to start executing */
J
Jiri Denemark 已提交
3816
        if (qemuProcessStartCPUs(driver, vm, conn,
3817 3818
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 QEMU_ASYNC_JOB_NONE) < 0) {
3819
            if (virGetLastError() == NULL)
3820 3821
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("resume operation failed"));
3822 3823
            goto cleanup;
        }
J
Jiri Denemark 已提交
3824 3825 3826 3827 3828
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             migrateFrom ?
                             VIR_DOMAIN_PAUSED_MIGRATION :
                             VIR_DOMAIN_PAUSED_USER);
3829 3830
    }

3831
    if (flags & VIR_QEMU_PROCESS_START_AUTODESROY &&
3832 3833 3834
        qemuProcessAutoDestroyAdd(driver, vm, conn) < 0)
        goto cleanup;

3835
    VIR_DEBUG("Writing domain status to disk");
3836 3837 3838
    if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0)
        goto cleanup;

3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855
    /* finally we can call the 'started' hook script if any */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0, false);
        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;
    }

3856 3857 3858 3859 3860 3861 3862 3863 3864
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);

    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 */
3865
    VIR_FREE(nodeset);
3866
    virBitmapFree(nodemask);
3867 3868
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);
3869
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, stop_flags);
3870 3871 3872 3873 3874

    return -1;
}


3875 3876 3877
int
qemuProcessKill(struct qemud_driver *driver,
                virDomainObjPtr vm, unsigned int flags)
3878
{
3879
    int ret;
3880 3881 3882

    VIR_DEBUG("vm=%s pid=%d flags=%x",
              vm->def->name, vm->pid, flags);
3883

3884 3885 3886 3887 3888
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
3889 3890
    }

3891 3892 3893 3894 3895 3896
    if ((flags & VIR_QEMU_PROCESS_KILL_NOWAIT)) {
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
3897

3898 3899
    if (driver)
        qemuDriverUnlock(driver);
3900

3901 3902
    ret = virProcessKillPainfully(vm->pid,
                                  !!(flags & VIR_QEMU_PROCESS_KILL_FORCE));
3903

3904
    if (driver) {
3905
        virObjectRef(vm);
3906 3907 3908
        virDomainObjUnlock(vm);
        qemuDriverLock(driver);
        virDomainObjLock(vm);
3909
        virObjectUnref(vm);
3910
    }
3911

3912
    return ret;
3913 3914 3915
}


3916 3917
void qemuProcessStop(struct qemud_driver *driver,
                     virDomainObjPtr vm,
3918 3919
                     virDomainShutoffReason reason,
                     unsigned int flags)
3920 3921 3922 3923 3924 3925
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
3926
    virNetDevVPortProfilePtr vport = NULL;
3927 3928 3929 3930 3931
    int i;
    int logfile = -1;
    char *timestamp;
    char ebuf[1024];

3932 3933
    VIR_DEBUG("Shutting down VM '%s' pid=%d flags=%x",
              vm->def->name, vm->pid, flags);
3934 3935 3936 3937 3938 3939

    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("VM '%s' not active", vm->def->name);
        return;
    }

3940 3941 3942 3943 3944 3945 3946
    /*
     * We may unlock the driver and vm in qemuProcessKill(), and another thread
     * can lock driver and vm, and then call qemuProcessStop(). So we should
     * set vm->def->id to -1 here to avoid qemuProcessStop() to be called twice.
     */
    vm->def->id = -1;

3947
    if ((logfile = qemuDomainCreateLog(driver, vm, true)) < 0) {
3948 3949 3950
        /* 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",
3951
                  virStrerror(errno, ebuf, sizeof(ebuf)));
3952
    } else {
3953
        if ((timestamp = virTimeStringNow()) == NULL) {
3954 3955 3956 3957 3958 3959
            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",
3960
                         virStrerror(errno, ebuf, sizeof(ebuf)));
3961 3962 3963 3964 3965 3966 3967
            }

            VIR_FREE(timestamp);
        }

        if (VIR_CLOSE(logfile) < 0)
             VIR_WARN("Unable to close logfile: %s",
3968
                      virStrerror(errno, ebuf, sizeof(ebuf)));
3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983
    }

    /* 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);

    if (driver->macFilter) {
        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,
3984
                                                  &net->mac))) {
3985
                virReportSystemError(errno,
3986
             _("failed to remove ebtables rule to allow MAC address on '%s'"),
3987 3988 3989 3990 3991
                                     net->ifname);
            }
        }
    }

D
Daniel P. Berrange 已提交
3992 3993 3994 3995 3996 3997
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
        priv->agentError = false;
    }

3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008
    if (priv->mon)
        qemuMonitorClose(priv->mon);

    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 */
4009 4010
    ignore_value(qemuProcessKill(driver, vm, VIR_QEMU_PROCESS_KILL_FORCE|
                                             VIR_QEMU_PROCESS_KILL_NOCHECK));
4011

4012 4013
    qemuDomainCleanupRun(driver, vm);

4014 4015 4016
    /* Stop autodestroy in case guest is restarted */
    qemuProcessAutoDestroyRemove(driver, vm);

4017 4018
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4019
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0, false);
4020 4021 4022

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4023 4024
                    VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
4025 4026 4027
        VIR_FREE(xml);
    }

4028 4029 4030 4031 4032
    /* 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);
4033
    virSecurityManagerReleaseLabel(driver->securityManager, vm->def);
4034 4035

    /* Clear out dynamically assigned labels */
4036 4037 4038 4039 4040
    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);
4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054
    }

    virDomainDefClearDeviceAliases(vm->def);
    if (!priv->persistentAddrs) {
        virDomainDefClearPCIAddresses(vm->def);
        qemuDomainPCIAddressSetFree(priv->pciaddrs);
        priv->pciaddrs = NULL;
    }

    qemuDomainReAttachHostDevices(driver, vm->def);

    def = vm->def;
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
4055
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4056
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
4057
                             net->ifname, &net->mac,
4058 4059
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
4060
                             virDomainNetGetActualVirtPortProfile(net),
4061
                             driver->stateDir));
4062 4063
            VIR_FREE(net->ifname);
        }
4064 4065 4066
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
A
Ansis Atteka 已提交
4067 4068 4069 4070 4071 4072
        vport = virDomainNetGetActualVirtPortProfile(net);
        if (vport && vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH)
            ignore_value(virNetDevOpenvswitchRemovePort(
                                       virDomainNetGetActualBridgeName(net),
                                       net->ifname));

4073 4074
        networkReleaseActualDevice(net);
    }
4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102

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);

    /* Remove VNC port from port reservation bitmap, but only if it was
       reserved by the driver (autoport=yes)
    */
    if ((vm->def->ngraphics == 1) &&
        vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
        vm->def->graphics[0]->data.vnc.autoport) {
        qemuProcessReturnPort(driver, vm->def->graphics[0]->data.vnc.port);
    }
    if ((vm->def->ngraphics == 1) &&
        vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE &&
        vm->def->graphics[0]->data.spice.autoport) {
        qemuProcessReturnPort(driver, vm->def->graphics[0]->data.spice.port);
        qemuProcessReturnPort(driver, vm->def->graphics[0]->data.spice.tlsPort);
    }

4103
    vm->taint = 0;
4104
    vm->pid = -1;
J
Jiri Denemark 已提交
4105
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
4106 4107
    VIR_FREE(priv->vcpupids);
    priv->nvcpupids = 0;
4108 4109
    virObjectUnref(priv->caps);
    priv->caps = NULL;
4110
    VIR_FREE(priv->pidfile);
4111

4112
    /* The "release" hook cleans up additional resources */
4113
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4114
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0, false);
4115 4116 4117

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4118 4119
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
4120 4121 4122
        VIR_FREE(xml);
    }

4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134
    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);
    }
}
4135 4136


4137 4138 4139
int qemuProcessAttach(virConnectPtr conn ATTRIBUTE_UNUSED,
                      struct qemud_driver *driver,
                      virDomainObjPtr vm,
4140
                      pid_t pid,
4141 4142 4143 4144
                      const char *pidfile,
                      virDomainChrSourceDefPtr monConfig,
                      bool monJSON)
{
4145
    size_t i;
4146 4147 4148 4149 4150
    char ebuf[1024];
    int logfile = -1;
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool running = true;
4151
    virDomainPausedReason reason;
4152
    virSecurityLabelPtr seclabel = NULL;
4153 4154 4155
    virSecurityLabelDefPtr seclabeldef = NULL;
    virSecurityManagerPtr* sec_managers = NULL;
    const char *model;
4156 4157 4158 4159

    VIR_DEBUG("Beginning VM attach process");

    if (virDomainObjIsActive(vm)) {
4160 4161
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187
        return -1;
    }

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

    vm->def->id = driver->nextvmid++;

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

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

    VIR_DEBUG("Detect security driver config");
4188 4189
    sec_managers = virSecurityManagerGetNested(driver->securityManager);
    if (sec_managers == NULL) {
4190
        goto cleanup;
4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212
    }

    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);
    }
4213 4214 4215 4216 4217 4218

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

    VIR_DEBUG("Determining emulator version");
4219 4220
    virObjectUnref(priv->caps);
    priv->caps = NULL;
4221 4222
    if (qemuCapsExtractVersionInfo(vm->def->emulator,
                                   vm->def->os.arch,
4223
                                   false,
4224
                                   NULL,
4225
                                   &priv->caps) < 0)
4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241
        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
     */
4242
    if (qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE)) {
4243
        VIR_DEBUG("Assigning domain PCI addresses");
4244
        if ((qemuDomainAssignAddresses(vm->def, priv->caps, vm)) < 0)
4245
            goto cleanup;
4246 4247
    }

4248
    if ((timestamp = virTimeStringNow()) == NULL) {
4249 4250 4251 4252 4253 4254
        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",
4255
                     virStrerror(errno, ebuf, sizeof(ebuf)));
4256 4257 4258 4259 4260 4261 4262 4263 4264 4265
        }

        VIR_FREE(timestamp);
    }

    qemuDomainObjTaint(driver, vm, VIR_DOMAIN_TAINT_EXTERNAL_LAUNCH, logfile);

    vm->pid = pid;

    VIR_DEBUG("Waiting for monitor to show up");
4266
    if (qemuProcessWaitForMonitor(driver, vm, priv->caps, -1) < 0)
4267 4268
        goto cleanup;

D
Daniel P. Berrange 已提交
4269 4270 4271 4272 4273 4274 4275 4276
    /* 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;
    }

4277 4278 4279 4280 4281 4282
    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 */
4283
    if (!qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE)) {
4284 4285 4286 4287 4288 4289
        VIR_DEBUG("Determining domain device PCI addresses");
        if (qemuProcessInitPCIAddresses(driver, vm) < 0)
            goto cleanup;
    }

    VIR_DEBUG("Getting initial memory amount");
4290
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
4291 4292 4293 4294
    if (qemuMonitorGetBalloonInfo(priv->mon, &vm->def->mem.cur_balloon) < 0) {
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        goto cleanup;
    }
4295
    if (qemuMonitorGetStatus(priv->mon, &running, &reason) < 0) {
4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        goto cleanup;
    }
    if (qemuMonitorGetVirtType(priv->mon, &vm->def->virtType) < 0) {
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        goto cleanup;
    }
    qemuDomainObjExitMonitorWithDriver(driver, vm);

    if (!virDomainObjIsActive(vm))
        goto cleanup;

    if (running)
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                             VIR_DOMAIN_RUNNING_UNPAUSED);
    else
4312
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
4313 4314 4315 4316 4317

    VIR_DEBUG("Writing domain status to disk");
    if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0)
        goto cleanup;

4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0, false);
        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;
    }

4335 4336
    VIR_FORCE_CLOSE(logfile);
    VIR_FREE(seclabel);
4337
    VIR_FREE(sec_managers);
4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348

    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);
4349
    VIR_FREE(sec_managers);
4350 4351 4352 4353 4354
    virDomainChrSourceDefFree(monConfig);
    return -1;
}


4355 4356 4357 4358
static virDomainObjPtr
qemuProcessAutoDestroy(struct qemud_driver *driver,
                       virDomainObjPtr dom,
                       virConnectPtr conn)
4359
{
4360
    qemuDomainObjPrivatePtr priv = dom->privateData;
4361 4362
    virDomainEventPtr event = NULL;

4363
    VIR_DEBUG("vm=%s, conn=%p", dom->def->name, conn);
4364

4365 4366
    if (priv->job.asyncJob) {
        VIR_DEBUG("vm=%s has long-term job active, cancelling",
4367
                  dom->def->name);
4368
        qemuDomainObjDiscardAsyncJob(driver, dom);
4369 4370
    }

4371
    if (qemuDomainObjBeginJobWithDriver(driver, dom,
4372
                                        QEMU_JOB_DESTROY) < 0)
4373 4374 4375
        goto cleanup;

    VIR_DEBUG("Killing domain");
4376 4377
    qemuProcessStop(driver, dom, VIR_DOMAIN_SHUTOFF_DESTROYED,
                    VIR_QEMU_PROCESS_STOP_MIGRATED);
4378
    virDomainAuditStop(dom, "destroyed");
4379 4380 4381
    event = virDomainEventNewFromObj(dom,
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_DESTROYED);
4382

4383
    if (!qemuDomainObjEndJob(driver, dom))
4384 4385
        dom = NULL;
    if (dom && !dom->persistent)
4386
        qemuDomainRemoveInactive(driver, dom);
4387
    if (event)
4388
        qemuDomainEventQueue(driver, event);
4389

4390 4391
cleanup:
    return dom;
4392 4393 4394 4395 4396 4397
}

int qemuProcessAutoDestroyAdd(struct qemud_driver *driver,
                              virDomainObjPtr vm,
                              virConnectPtr conn)
{
4398 4399 4400
    VIR_DEBUG("vm=%s, conn=%p", vm->def->name, conn);
    return qemuDriverCloseCallbackSet(driver, vm, conn,
                                      qemuProcessAutoDestroy);
4401 4402 4403 4404 4405
}

int qemuProcessAutoDestroyRemove(struct qemud_driver *driver,
                                 virDomainObjPtr vm)
{
4406 4407
    VIR_DEBUG("vm=%s", vm->def->name);
    return qemuDriverCloseCallbackUnset(driver, vm, qemuProcessAutoDestroy);
4408
}
4409 4410 4411 4412

bool qemuProcessAutoDestroyActive(struct qemud_driver *driver,
                                  virDomainObjPtr vm)
{
4413 4414 4415 4416
    qemuDriverCloseCallback cb;
    VIR_DEBUG("vm=%s", vm->def->name);
    cb = qemuDriverCloseCallbackGet(driver, vm, NULL);
    return cb == qemuProcessAutoDestroy;
4417
}