qemu_process.c 135.6 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
                          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;
132 133
    if (priv->agent == agent)
        priv->agent = NULL;
D
Daniel P. Berrange 已提交
134 135 136

    virDomainObjUnlock(vm);
    qemuDriverUnlock(driver);
137 138

    qemuAgentClose(agent);
D
Daniel P. Berrange 已提交
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
}


/*
 * 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)
{
171 172
    VIR_DEBUG("Received destroy agent=%p vm=%p", agent, vm);

173
    virObjectUnref(vm);
D
Daniel P. Berrange 已提交
174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
}


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;

195
        if (STREQ_NULLABLE(channel->target.name, "org.qemu.guest_agent.0")) {
D
Daniel P. Berrange 已提交
196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
            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 */
224
    virObjectRef(vm);
D
Daniel P. Berrange 已提交
225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245

    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)
246
        virObjectUnref(vm);
D
Daniel P. Berrange 已提交
247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265

    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;
}


266
/*
267
 * This is a callback registered with a qemuMonitorPtr instance,
268 269 270 271 272 273
 * 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,
274
                            virDomainObjPtr vm)
275 276 277 278
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;
    qemuDomainObjPrivatePtr priv;
279 280 281
    int eventReason = VIR_DOMAIN_EVENT_STOPPED_SHUTDOWN;
    int stopReason = VIR_DOMAIN_SHUTOFF_SHUTDOWN;
    const char *auditReason = "shutdown";
282 283 284

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

285
    qemuDriverLock(driver);
286 287
    virDomainObjLock(vm);

288 289 290 291 292 293 294
    priv = vm->privateData;

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

295 296
    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Domain %p is not active, ignoring EOF", vm);
297
        goto unlock;
298 299
    }

300
    if (priv->monJSON && !priv->gotShutdown) {
301 302
        VIR_DEBUG("Monitor connection to '%s' closed without SHUTDOWN event; "
                  "assuming the domain crashed", vm->def->name);
303
        eventReason = VIR_DOMAIN_EVENT_STOPPED_FAILED;
304
        stopReason = VIR_DOMAIN_SHUTOFF_CRASHED;
305
        auditReason = "failed";
306 307 308 309
    }

    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STOPPED,
310
                                     eventReason);
311
    qemuProcessStop(driver, vm, stopReason, 0);
312
    virDomainAuditStop(vm, auditReason);
313

314
    if (!vm->persistent) {
315
        qemuDomainRemoveInactive(driver, vm);
316 317 318 319 320
        goto cleanup;
    }

unlock:
    virDomainObjUnlock(vm);
321

322
cleanup:
323
    if (event)
324
        qemuDomainEventQueue(driver, event);
325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346
    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);

347
    ((qemuDomainObjPrivatePtr) vm->privateData)->monError = true;
348 349 350 351 352
    event = virDomainEventControlErrorNewFromObj(vm);
    if (event)
        qemuDomainEventQueue(driver, event);

    virDomainObjUnlock(vm);
353
    qemuDriverUnlock(driver);
354 355 356 357 358 359 360
}


static virDomainDiskDefPtr
qemuProcessFindDomainDiskByPath(virDomainObjPtr vm,
                                const char *path)
{
361
    int i = virDomainDiskIndexByName(vm->def, path, true);
362

363 364
    if (i >= 0)
        return vm->def->disks[i];
365

366 367 368
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with path %s"),
                   path);
369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388
    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;
    }

389 390 391
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with alias %s"),
                   alias);
392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408
    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) {
409 410 411
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("disk %s does not have any encryption information"),
                       disk->src);
412 413 414 415 416
        return -1;
    }
    enc = disk->encryption;

    if (!conn) {
417 418
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("cannot find secrets without a connection"));
419 420 421 422 423 424
        goto cleanup;
    }

    if (conn->secretDriver == NULL ||
        conn->secretDriver->lookupByUUID == NULL ||
        conn->secretDriver->getValue == NULL) {
425 426
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("secret storage not supported"));
427 428 429 430 431 432 433
        goto cleanup;
    }

    if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW ||
        enc->nsecrets != 1 ||
        enc->secrets[0]->type !=
        VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE) {
434 435
        virReportError(VIR_ERR_XML_ERROR,
                       _("invalid <encryption> for volume %s"), disk->src);
436 437 438 439 440 441 442
        goto cleanup;
    }

    secret = conn->secretDriver->lookupByUUID(conn,
                                              enc->secrets[0]->uuid);
    if (secret == NULL)
        goto cleanup;
443
    data = conn->secretDriver->getValue(secret, &size, 0,
444
                                        VIR_SECRET_GET_VALUE_INTERNAL_CALL);
445
    virObjectUnref(secret);
446 447 448 449 450 451
    if (data == NULL)
        goto cleanup;

    if (memchr(data, '\0', size) != NULL) {
        memset(data, 0, size);
        VIR_FREE(data);
452 453 454
        virReportError(VIR_ERR_XML_ERROR,
                       _("format='qcow' passphrase for %s must not contain a "
                         "'\\0'"), disk->src);
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509
        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;
510
    qemuDomainObjPrivatePtr priv;
511 512

    virDomainObjLock(vm);
513

514
    event = virDomainEventRebootNewFromObj(vm);
515 516 517 518
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);

519 520 521 522 523 524 525 526 527 528 529 530
    virDomainObjUnlock(vm);

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

    return 0;
}


531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
/*
 * 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);
550
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
551 552 553
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
554 555
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
556 557 558
        goto endjob;
    }

559
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
560 561 562 563 564 565 566
    if (qemuMonitorSystemReset(priv->mon) < 0) {
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        goto endjob;
    }
    qemuDomainObjExitMonitorWithDriver(driver, vm);

    if (!virDomainObjIsActive(vm)) {
567 568
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
569 570 571 572
        goto endjob;
    }

    if (qemuProcessStartCPUs(driver, vm, NULL,
573 574
                             VIR_DOMAIN_RUNNING_BOOTED,
                             QEMU_ASYNC_JOB_NONE) < 0) {
575
        if (virGetLastError() == NULL)
576 577
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
578 579
        goto endjob;
    }
580
    priv->gotShutdown = false;
581 582 583 584 585 586 587
    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_RESUMED,
                                     VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

    ret = 0;

endjob:
588
    if (!qemuDomainObjEndJob(driver, vm))
589 590 591 592
        vm = NULL;

cleanup:
    if (vm) {
593 594 595 596
        if (ret == -1) {
            ignore_value(qemuProcessKill(driver, vm,
                                         VIR_QEMU_PROCESS_KILL_FORCE));
        }
597
        if (virObjectUnref(vm))
598 599 600 601 602 603 604 605
            virDomainObjUnlock(vm);
    }
    if (event)
        qemuDomainEventQueue(driver, event);
    qemuDriverUnlock(driver);
}


606
static void
607 608
qemuProcessShutdownOrReboot(struct qemud_driver *driver,
                            virDomainObjPtr vm)
609
{
610 611 612
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
613
        qemuDomainSetFakeReboot(driver, vm, false);
614
        virObjectRef(vm);
615 616 617 618 619
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
620
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
621 622
            ignore_value(qemuProcessKill(driver, vm,
                                         VIR_QEMU_PROCESS_KILL_NOWAIT));
623
            virObjectUnref(vm);
624 625
        }
    } else {
626
        ignore_value(qemuProcessKill(driver, vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
627
    }
628
}
629

630 631 632 633
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm)
{
634 635 636 637
    struct qemud_driver *driver = qemu_driver;
    qemuDomainObjPrivatePtr priv;
    virDomainEventPtr event = NULL;

638 639 640
    VIR_DEBUG("vm=%p", vm);

    virDomainObjLock(vm);
641 642 643 644 645 646

    priv = vm->privateData;
    if (priv->gotShutdown) {
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
647 648 649 650
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667
    }
    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);
    }

668 669 670
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

671 672 673
    qemuProcessShutdownOrReboot(driver, vm);

unlock:
674
    virDomainObjUnlock(vm);
675 676 677 678 679 680 681

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

682 683 684 685 686 687 688 689 690 691 692 693
    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
694
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
695
        qemuDomainObjPrivatePtr priv = vm->privateData;
696

697
        if (priv->gotShutdown) {
698 699
            VIR_DEBUG("Ignoring STOP event after SHUTDOWN");
            goto unlock;
700 701
        }

702 703
        VIR_DEBUG("Transitioned guest %s to paused state",
                  vm->def->name);
704

705
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_UNKNOWN);
706 707 708 709
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);

710 711 712 713 714
        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));

715 716 717 718
        if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
719
    }
720 721

unlock:
722 723 724 725
    virDomainObjUnlock(vm);

    if (event) {
        qemuDriverLock(driver);
726
        qemuDomainEventQueue(driver, event);
727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745
        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)
746
        vm->def->clock.data.variable.adjustment = offset;
747 748

    if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0)
749
        VIR_WARN("unable to save domain status with RTC change");
750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775

    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
776
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
777
        qemuDomainObjPrivatePtr priv = vm->privateData;
778 779
        VIR_DEBUG("Transitioned guest %s to paused state due to watchdog", vm->def->name);

J
Jiri Denemark 已提交
780
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
781 782 783 784
        lifecycleEvent = virDomainEventNewFromObj(vm,
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

785 786 787 788 789
        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));

790 791 792 793
        if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
794 795 796 797 798 799 800
    }

    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
801 802 803
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
804
            virObjectRef(vm);
805
            if (virThreadPoolSendJob(driver->workerPool, 0, wdEvent) < 0) {
806
                if (!virObjectUnref(vm))
807
                    vm = NULL;
W
Wen Congyang 已提交
808 809
                VIR_FREE(wdEvent);
            }
810
        } else {
811
            virReportOOMError();
812
        }
813 814
    }

815 816
    if (vm)
        virDomainObjUnlock(vm);
817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860

    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
861
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
862
        qemuDomainObjPrivatePtr priv = vm->privateData;
863 864
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

J
Jiri Denemark 已提交
865
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
866 867 868 869
        lifecycleEvent = virDomainEventNewFromObj(vm,
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

870 871 872 873 874
        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));

875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893
        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;
}

894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923
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;
}
924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965

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++;
966 967 968
        if (!(subject->identities[subject->nidentity-1].type = strdup("x509dname")) ||
            !(subject->identities[subject->nidentity-1].name = strdup(x509dname)))
            goto no_memory;
969 970 971 972 973
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
            goto no_memory;
        subject->nidentity++;
974 975 976
        if (!(subject->identities[subject->nidentity-1].type = strdup("saslUsername")) ||
            !(subject->identities[subject->nidentity-1].name = strdup(saslUsername)))
            goto no_memory;
977 978
    }

979
    virDomainObjLock(vm);
980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
    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;
}


1016
static void qemuProcessHandleMonitorDestroy(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1017 1018
                                            virDomainObjPtr vm)
{
1019
    virObjectUnref(vm);
1020 1021
}

1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
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);
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
        /* 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);
        }
1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
    }

    virDomainObjUnlock(vm);

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

    return 0;
}

O
Osier Yang 已提交
1062 1063 1064 1065 1066 1067
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;
1068
    virDomainEventPtr lifecycleEvent = NULL;
O
Osier Yang 已提交
1069 1070 1071 1072

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

1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
    /* 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
1092 1093
    virDomainObjUnlock(vm);

1094
    if (event || lifecycleEvent) {
O
Osier Yang 已提交
1095
        qemuDriverLock(driver);
1096 1097 1098 1099
        if (event)
            qemuDomainEventQueue(driver, event);
        if (lifecycleEvent)
            qemuDomainEventQueue(driver, lifecycleEvent);
O
Osier Yang 已提交
1100 1101 1102 1103 1104
        qemuDriverUnlock(driver);
    }

    return 0;
}
1105

O
Osier Yang 已提交
1106 1107 1108 1109 1110 1111
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm)
{
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;
J
Jiri Denemark 已提交
1112
    virDomainEventPtr lifecycleEvent = NULL;
O
Osier Yang 已提交
1113 1114 1115 1116

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

1117
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1118
        qemuDomainObjPrivatePtr priv = vm->privateData;
1119 1120 1121 1122 1123
        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
1124 1125 1126 1127
        lifecycleEvent =
            virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1128 1129 1130 1131 1132

        if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1133 1134 1135

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1136 1137
    }

O
Osier Yang 已提交
1138 1139
    virDomainObjUnlock(vm);

J
Jiri Denemark 已提交
1140
    if (event || lifecycleEvent) {
O
Osier Yang 已提交
1141
        qemuDriverLock(driver);
J
Jiri Denemark 已提交
1142 1143 1144 1145
        if (event)
            qemuDomainEventQueue(driver, event);
        if (lifecycleEvent)
            qemuDomainEventQueue(driver, lifecycleEvent);
O
Osier Yang 已提交
1146 1147 1148 1149 1150 1151
        qemuDriverUnlock(driver);
    }

    return 0;
}

1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181
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;
}


1182 1183 1184
static qemuMonitorCallbacks monitorCallbacks = {
    .destroy = qemuProcessHandleMonitorDestroy,
    .eofNotify = qemuProcessHandleMonitorEOF,
1185
    .errorNotify = qemuProcessHandleMonitorError,
1186 1187 1188 1189 1190 1191 1192 1193
    .diskSecretLookup = qemuProcessFindVolumeQcowPassphrase,
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1194
    .domainBlockJob = qemuProcessHandleBlockJob,
1195
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1196
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1197
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1198
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1199 1200 1201 1202 1203 1204 1205
};

static int
qemuConnectMonitor(struct qemud_driver *driver, virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
1206
    qemuMonitorPtr mon = NULL;
1207

1208
    if (virSecurityManagerSetDaemonSocketLabel(driver->securityManager,
1209
                                               vm->def) < 0) {
1210 1211 1212 1213 1214 1215 1216
        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 */
1217
    virObjectRef(vm);
1218

1219
    ignore_value(virTimeMillisNow(&priv->monStart));
1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230
    virDomainObjUnlock(vm);
    qemuDriverUnlock(driver);

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

    qemuDriverLock(driver);
    virDomainObjLock(vm);
    priv->monStart = 0;
1231

1232
    if (mon == NULL) {
1233
        virObjectUnref(vm);
1234
    } else if (!virDomainObjIsActive(vm)) {
1235
        qemuMonitorClose(mon);
1236
        mon = NULL;
1237 1238 1239
    }
    priv->mon = mon;

1240
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0) {
1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
        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;
    }


1252
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
1253 1254 1255 1256
    ret = qemuMonitorSetCapabilities(priv->mon);
    if (ret == 0 &&
        qemuCapsGet(priv->caps, QEMU_CAPS_MONITOR_JSON))
        ret = qemuCapsProbeQMP(priv->caps, priv->mon);
1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281
    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;
1282 1283 1284 1285
    char *debug = NULL;
    int ret = -1;
    char *filter_next = buf;

1286 1287
    buf[0] = '\0';

1288 1289 1290 1291 1292 1293 1294
    /* 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;
    }

1295
    while (retries) {
1296
        ssize_t func_ret, bytes;
1297
        int isdead = 0;
1298
        char *eol;
1299 1300 1301 1302 1303 1304 1305 1306

        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. */
1307 1308
        bytes = saferead(fd, buf+got, buflen-got-1);
        if (bytes < 0) {
1309 1310 1311
            virReportSystemError(errno,
                                 _("Failure while reading %s log output"),
                                 what);
1312
            goto cleanup;
1313 1314
        }

1315
        got += bytes;
1316
        buf[got] = '\0';
1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329

        /* 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';
            }
        }

1330
        if (got == buflen-1) {
1331 1332 1333
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Out of space while reading %s log output: %s"),
                           what, buf);
1334
            goto cleanup;
1335 1336 1337
        }

        if (isdead) {
1338 1339 1340
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Process exited while reading %s log output: %s"),
                           what, buf);
1341
            goto cleanup;
1342 1343
        }

1344 1345 1346 1347
        if (func_ret <= 0) {
            ret = func_ret;
            goto cleanup;
        }
1348 1349 1350 1351 1352

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

1353 1354 1355
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("Timed out while reading %s log output: %s"),
                   what, buf);
1356 1357 1358 1359

cleanup:
    VIR_FREE(debug);
    return ret;
1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 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
}


/*
 * 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;
}

1415 1416 1417
static int
qemuProcessLookupPTYs(virDomainChrDefPtr *devices,
                      int count,
C
Cole Robinson 已提交
1418 1419
                      virHashTablePtr paths,
                      bool chardevfmt)
1420 1421
{
    int i;
C
Cole Robinson 已提交
1422
    const char *prefix = chardevfmt ? "char" : "";
1423 1424 1425 1426

    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
1427
            char id[32];
1428 1429
            const char *path;

C
Cole Robinson 已提交
1430 1431
            if (snprintf(id, sizeof(id), "%s%s",
                         prefix, chr->info.alias) >= sizeof(id))
1432 1433 1434 1435 1436 1437 1438 1439
                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
                     */
1440 1441
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463
                    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;
}

1464 1465
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
1466
                                     qemuCapsPtr caps,
1467 1468
                                     virHashTablePtr paths)
{
1469
    bool chardevfmt = qemuCapsGet(caps, QEMU_CAPS_CHARDEV);
C
Cole Robinson 已提交
1470

1471
    if (qemuProcessLookupPTYs(vm->def->serials, vm->def->nserials,
C
Cole Robinson 已提交
1472
                              paths, chardevfmt) < 0)
1473 1474 1475
        return -1;

    if (qemuProcessLookupPTYs(vm->def->parallels, vm->def->nparallels,
C
Cole Robinson 已提交
1476
                              paths, chardevfmt) < 0)
1477
        return -1;
1478

1479
    if (qemuProcessLookupPTYs(vm->def->channels, vm->def->nchannels,
C
Cole Robinson 已提交
1480
                              paths, chardevfmt) < 0)
1481 1482
        return -1;

1483 1484
    if (qemuProcessLookupPTYs(vm->def->consoles, vm->def->nconsoles,
                              paths, chardevfmt) < 0)
1485
        return -1;
1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531

    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;
        }
    }

1532 1533
    for (i = 0 ; i < vm->def->nconsoles ; i++) {
        virDomainChrDefPtr chr = vm->def->consoles[i];
1534 1535 1536 1537 1538 1539 1540
        /* 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)
1541
                return ret;
1542 1543 1544 1545 1546 1547 1548
        } 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;
            }
1549 1550 1551
        }
    }

1552 1553 1554
    return 0;
}

1555
static void qemuProcessFreePtyPath(void *payload, const void *name ATTRIBUTE_UNUSED)
1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573
{
    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';
}

1574

1575 1576
static int
qemuProcessWaitForMonitor(struct qemud_driver* driver,
C
Cole Robinson 已提交
1577
                          virDomainObjPtr vm,
1578
                          qemuCapsPtr caps,
C
Cole Robinson 已提交
1579
                          off_t pos)
1580
{
1581
    char *buf = NULL;
1582
    size_t buf_size = 4096; /* Plenty of space to get startup greeting */
1583
    int logfd = -1;
1584 1585
    int ret = -1;
    virHashTablePtr paths = NULL;
1586
    qemuDomainObjPrivatePtr priv;
1587

1588 1589 1590
    if (pos != -1) {
        if ((logfd = qemuDomainOpenLog(driver, vm, pos)) < 0)
            return -1;
1591

1592 1593
        if (VIR_ALLOC_N(buf, buf_size) < 0) {
            virReportOOMError();
1594
            goto closelog;
1595
        }
1596

1597 1598 1599 1600 1601
        if (qemuProcessReadLogOutput(vm, logfd, buf, buf_size,
                                     qemuProcessFindCharDevicePTYs,
                                     "console", 30) < 0)
            goto closelog;
    }
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611

    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. */
1612
    paths = virHashCreate(0, qemuProcessFreePtyPath);
1613
    if (paths == NULL)
1614 1615
        goto cleanup;

1616
    priv = vm->privateData;
1617
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
1618 1619 1620 1621 1622
    ret = qemuMonitorGetPtyPaths(priv->mon, paths);
    qemuDomainObjExitMonitorWithDriver(driver, vm);

    VIR_DEBUG("qemuMonitorGetPtyPaths returned %i", ret);
    if (ret == 0)
1623
        ret = qemuProcessFindCharDevicePTYsMonitor(vm, caps, paths);
1624 1625

cleanup:
1626
    virHashFree(paths);
1627

E
Eric Blake 已提交
1628
    if (pos != -1 && kill(vm->pid, 0) == -1 && errno == ESRCH) {
1629 1630
        /* VM is dead, any other error raised in the interim is probably
         * not as important as the qemu cmdline output */
1631
        qemuProcessReadLogFD(logfd, buf, buf_size, strlen(buf));
1632 1633 1634
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("process exited while connecting to monitor: %s"),
                       buf);
1635 1636 1637
        ret = -1;
    }

1638
closelog:
1639
    if (VIR_CLOSE(logfd) < 0) {
1640
        char ebuf[1024];
1641
        VIR_WARN("Unable to close logfile: %s",
1642
                 virStrerror(errno, ebuf, sizeof(ebuf)));
1643 1644
    }

1645 1646
    VIR_FREE(buf);

1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657
    return ret;
}

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

1658 1659 1660 1661 1662 1663 1664 1665
    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();

1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676
        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) {
1677 1678 1679 1680
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of vCPU pids from QEMU monitor. "
                         "got %d, wanted %d"),
                       ncpupids, vm->def->vcpus);
1681 1682 1683 1684 1685 1686 1687 1688 1689
        VIR_FREE(cpupids);
        return -1;
    }

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

1690 1691 1692 1693 1694 1695 1696

/*
 * Set NUMA memory policy for qemu process, to be run between
 * fork/exec of QEMU only.
 */
#if HAVE_NUMACTL
static int
1697
qemuProcessInitNumaMemoryPolicy(virDomainObjPtr vm,
1698
                                virBitmapPtr nodemask)
1699
{
E
Eric Blake 已提交
1700
    nodemask_t mask;
1701 1702 1703 1704 1705
    int mode = -1;
    int node = -1;
    int ret = -1;
    int i = 0;
    int maxnode = 0;
E
Eric Blake 已提交
1706
    bool warned = false;
1707
    virDomainNumatuneDef numatune = vm->def->numatune;
1708
    virBitmapPtr tmp_nodemask = NULL;
1709

1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720
    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 {
1721
        return 0;
1722
    }
1723 1724

    if (numa_available() < 0) {
1725 1726
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("Host kernel is not aware of NUMA."));
1727 1728 1729
        return -1;
    }

E
Eric Blake 已提交
1730
    maxnode = numa_max_node() + 1;
1731
    /* Convert nodemask to NUMA bitmask. */
E
Eric Blake 已提交
1732
    nodemask_zero(&mask);
1733 1734 1735 1736 1737 1738
    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;
1739
        }
1740 1741 1742 1743 1744 1745
        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);
1746 1747
    }

1748
    mode = numatune.memory.mode;
1749 1750 1751

    if (mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT) {
        numa_set_bind_policy(1);
E
Eric Blake 已提交
1752
        numa_set_membind(&mask);
1753 1754 1755
        numa_set_bind_policy(0);
    } else if (mode == VIR_DOMAIN_NUMATUNE_MEM_PREFERRED) {
        int nnodes = 0;
E
Eric Blake 已提交
1756 1757
        for (i = 0; i < NUMA_NUM_NODES; i++) {
            if (nodemask_isset(&mask, i)) {
1758 1759 1760 1761 1762 1763
                node = i;
                nnodes++;
            }
        }

        if (nnodes != 1) {
1764 1765 1766
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("NUMA memory tuning in 'preferred' mode "
                                   "only supports single node"));
1767 1768 1769 1770 1771 1772
            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
1773
        numa_set_interleave_mask(&mask);
1774 1775 1776 1777
    } else {
        /* XXX: Shouldn't go here, as we already do checking when
         * parsing domain XML.
         */
1778 1779
        virReportError(VIR_ERR_XML_ERROR,
                       "%s", _("Invalid mode for memory NUMA tuning."));
1780 1781 1782 1783 1784 1785 1786 1787 1788 1789
        goto cleanup;
    }

    ret = 0;

cleanup:
    return ret;
}
#else
static int
E
Eric Blake 已提交
1790
qemuProcessInitNumaMemoryPolicy(virDomainObjPtr vm,
1791
                                virBitmapPtr nodemask ATTRIBUTE_UNUSED)
1792 1793
{
    if (vm->def->numatune.memory.nodemask) {
1794 1795
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("libvirt is compiled without NUMA tuning support"));
1796 1797 1798 1799 1800 1801 1802 1803

        return -1;
    }

    return 0;
}
#endif

1804
#if HAVE_NUMAD
O
Osier Yang 已提交
1805 1806 1807 1808 1809 1810
static char *
qemuGetNumadAdvice(virDomainDefPtr def)
{
    virCommandPtr cmd = NULL;
    char *output = NULL;

1811 1812
    cmd = virCommandNewArgList(NUMAD, "-w", NULL);
    virCommandAddArgFormat(cmd, "%d:%llu", def->vcpus,
1813
                           VIR_DIV_UP(def->mem.cur_balloon, 1024));
O
Osier Yang 已提交
1814 1815 1816 1817

    virCommandSetOutputBuffer(cmd, &output);

    if (virCommandRun(cmd, NULL) < 0)
1818 1819 1820
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Failed to query numad for the "
                         "advisory nodeset"));
O
Osier Yang 已提交
1821 1822 1823 1824 1825 1826 1827 1828

    virCommandFree(cmd);
    return output;
}
#else
static char *
qemuGetNumadAdvice(virDomainDefPtr def ATTRIBUTE_UNUSED)
{
1829 1830
    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                   _("numad is not available on this host"));
O
Osier Yang 已提交
1831 1832 1833 1834
    return NULL;
}
#endif

1835 1836 1837 1838
/*
 * To be run between fork/exec of QEMU only
 */
static int
O
Osier Yang 已提交
1839
qemuProcessInitCpuAffinity(struct qemud_driver *driver,
1840
                           virDomainObjPtr vm,
1841
                           virBitmapPtr nodemask)
1842
{
1843
    int ret = -1;
1844 1845
    int i, hostcpus, maxcpu = QEMUD_CPUMASK_LEN;
    virNodeInfo nodeinfo;
H
Hu Tao 已提交
1846
    virBitmapPtr cpumap, cpumapToSet;
1847

1848
    VIR_DEBUG("Setting CPU affinity");
1849 1850 1851 1852 1853 1854 1855 1856 1857 1858

    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;

1859 1860
    cpumap = virBitmapNew(maxcpu);
    if (!cpumap) {
1861 1862 1863 1864
        virReportOOMError();
        return -1;
    }

H
Hu Tao 已提交
1865 1866
    cpumapToSet = cpumap;

O
Osier Yang 已提交
1867
    if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) {
1868
        VIR_DEBUG("Set CPU affinity with advisory nodeset from numad");
1869 1870 1871 1872
        /* 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;
1873 1874 1875 1876
            bool result;
            if (virBitmapGetBit(nodemask, i, &result) < 0)
                goto cleanup;
            if (result) {
1877
                for (j = 0; j < cur_ncpus; j++)
1878 1879
                    ignore_value(virBitmapSetBit(cpumap,
                                                 driver->caps->host.numaCell[i]->cpus[j]));
1880
            }
O
Osier Yang 已提交
1881
        }
1882
    } else {
1883
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
1884
        if (vm->def->cpumask) {
H
Hu Tao 已提交
1885
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
1886 1887 1888 1889 1890
        } 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 */
1891
            virBitmapSetAll(cpumap);
O
Osier Yang 已提交
1892
        }
1893 1894 1895 1896 1897 1898
    }

    /* 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
1899
    if (virProcessInfoSetAffinity(0 /* Self */, cpumapToSet) < 0)
1900
        goto cleanup;
1901

1902 1903 1904
    ret = 0;

cleanup:
1905
    virBitmapFree(cpumap);
1906
    return ret;
1907 1908
}

1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921
/* 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);

1922
            if (!qemuCapsGet(priv->caps, QEMU_CAPS_NETDEV)) {
1923 1924
                virReportError(VIR_ERR_NO_SUPPORT, "%s",
                               _("Setting of link state is not supported by this qemu"));
1925 1926 1927 1928 1929 1930 1931
                return -1;
            }

            ret = qemuMonitorSetLink(priv->mon,
                                     def->nets[i]->info.alias,
                                     VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (ret != 0) {
1932
                virReportError(VIR_ERR_OPERATION_FAILED,
1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
                               _("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. */
1943 1944 1945 1946 1947 1948 1949
static int
qemuProcessSetVcpuAffinites(virConnectPtr conn,
                            virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
    virNodeInfo nodeinfo;
1950
    int vcpu, n;
E
Eric Blake 已提交
1951
    int ret = -1;
1952 1953 1954 1955 1956 1957 1958 1959 1960

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

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

    if (priv->vcpupids == NULL) {
1961 1962
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("cpu affinity is not supported"));
1963 1964 1965
        return -1;
    }

1966 1967
    for (n = 0; n < def->cputune.nvcpupin; n++) {
        vcpu = def->cputune.vcpupin[n]->vcpuid;
1968

1969 1970
        if (virProcessInfoSetAffinity(priv->vcpupids[vcpu],
                                      def->cputune.vcpupin[n]->cpumask) < 0) {
E
Eric Blake 已提交
1971
            goto cleanup;
1972 1973 1974
        }
    }

E
Eric Blake 已提交
1975 1976 1977
    ret = 0;
cleanup:
    return ret;
1978
}
1979

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
/* 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;

1995 1996
    if (virProcessInfoSetAffinity(vm->pid,
                                  def->cputune.emulatorpin->cpumask) < 0) {
1997 1998 1999 2000 2001 2002 2003 2004
        goto cleanup;
    }

    ret = 0;
cleanup:
    return ret;
}

2005 2006 2007
static int
qemuProcessInitPasswords(virConnectPtr conn,
                         struct qemud_driver *driver,
2008
                         virDomainObjPtr vm)
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029
{
    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;

2030
    if (qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE)) {
2031 2032 2033 2034 2035
        int i;

        for (i = 0 ; i < vm->def->ndisks ; i++) {
            char *secret;
            size_t secretLen;
E
Eric Blake 已提交
2036
            const char *alias;
2037 2038 2039 2040 2041 2042 2043 2044 2045 2046

            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
2047
            alias = vm->def->disks[i]->info.alias;
2048
            qemuDomainObjEnterMonitorWithDriver(driver, vm);
E
Eric Blake 已提交
2049
            ret = qemuMonitorSetDrivePassphrase(priv->mon, alias, secret);
2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329
            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) {
2330 2331 2332
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for VirtIO disk %s"),
                           vm->def->disks[i]->dst);
2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343
            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) {
2344 2345 2346
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for %s NIC"),
                           vm->def->nets[i]->model);
2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357
            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) {
2358 2359 2360
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for controller %s"),
                           virDomainControllerTypeToString(vm->def->controllers[i]->type));
2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371
            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) {
2372 2373 2374
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for video adapter %s"),
                           virDomainVideoTypeToString(vm->def->videos[i]->type));
2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385
            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) {
2386 2387 2388
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for sound adapter %s"),
                           virDomainSoundModelTypeToString(vm->def->sounds[i]->model));
2389 2390 2391 2392 2393 2394 2395 2396 2397 2398
            return -1;
        }
    }


    if (vm->def->watchdog &&
        qemuProcessGetPCIWatchdogVendorProduct(vm->def->watchdog, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->watchdog->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2399 2400 2401
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for watchdog %s"),
                           virDomainWatchdogModelTypeToString(vm->def->watchdog->model));
2402 2403 2404 2405 2406 2407 2408 2409 2410
            return -1;
        }
    }

    if (vm->def->memballoon &&
        qemuProcessGetPCIMemballoonVendorProduct(vm->def->memballoon, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->memballoon->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2411 2412 2413
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for balloon %s"),
                           virDomainMemballoonModelTypeToString(vm->def->memballoon->model));
2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438
            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;

2439
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456
    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;

2457
    for (i = startPort ; i < driver->remotePortMax; i++) {
2458 2459 2460 2461 2462
        int fd;
        int reuse = 1;
        struct sockaddr_in addr;
        bool used = false;

2463
        if (virBitmapGetBit(driver->reservedRemotePorts,
2464 2465
                            i - driver->remotePortMin, &used) < 0)
            VIR_DEBUG("virBitmapGetBit failed on bit %d", i - driver->remotePortMin);
2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485

        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 */
2486
            if (virBitmapSetBit(driver->reservedRemotePorts,
2487
                                i - driver->remotePortMin) < 0) {
2488
                VIR_DEBUG("virBitmapSetBit failed on bit %d",
2489
                          i - driver->remotePortMin);
2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509
            }
            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)
{
2510
    if (port < driver->remotePortMin)
2511 2512
        return;

2513
    if (virBitmapClearBit(driver->reservedRemotePorts,
2514
                          port - driver->remotePortMin) < 0)
2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541
        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;
}


2542 2543 2544
static int
qemuProcessLimits(struct qemud_driver *driver)
{
2545
    struct rlimit rlim;
2546

2547
    if (driver->maxProcesses > 0) {
2548 2549 2550 2551 2552 2553 2554 2555 2556
        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;
        }
    }

2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568
    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;
        }
    }

2569 2570 2571 2572
    return 0;
}


2573 2574 2575 2576
struct qemuProcessHookData {
    virConnectPtr conn;
    virDomainObjPtr vm;
    struct qemud_driver *driver;
2577
    virBitmapPtr nodemask;
2578 2579 2580 2581 2582
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
2583
    int ret = -1;
2584
    int fd;
2585 2586 2587 2588 2589

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

    VIR_DEBUG("Obtaining domain lock");
2590 2591 2592 2593 2594 2595 2596
    /*
     * 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.
     */
2597
    if (virSecurityManagerSetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2598
        goto cleanup;
2599 2600 2601
    if (virDomainLockProcessStart(h->driver->lockManager,
                                  h->vm,
                                  /* QEMU is always pased initially */
2602 2603
                                  true,
                                  &fd) < 0)
2604
        goto cleanup;
2605
    if (virSecurityManagerClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2606
        goto cleanup;
2607

2608
    if (qemuProcessLimits(h->driver) < 0)
2609
        goto cleanup;
2610

2611 2612 2613
    /* This must take place before exec(), so that all QEMU
     * memory allocation is on the correct NUMA node
     */
2614
    VIR_DEBUG("Moving process to cgroup");
2615
    if (qemuAddToCgroup(h->driver, h->vm->def) < 0)
2616
        goto cleanup;
2617 2618 2619

    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
2620
    if (qemuProcessInitCpuAffinity(h->driver, h->vm, h->nodemask) < 0)
2621
        goto cleanup;
2622

2623
    if (qemuProcessInitNumaMemoryPolicy(h->vm, h->nodemask) < 0)
2624
        goto cleanup;
2625

2626
    VIR_DEBUG("Setting up security labelling");
2627
    if (virSecurityManagerSetProcessLabel(h->driver->securityManager, h->vm->def) < 0)
2628
        goto cleanup;
2629

2630 2631 2632 2633 2634
    ret = 0;

cleanup:
    VIR_DEBUG("Hook complete ret=%d", ret);
    return ret;
2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654
}

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;
}


2655 2656 2657 2658 2659
/*
 * 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
2660 2661
int
qemuProcessStartCPUs(struct qemud_driver *driver, virDomainObjPtr vm,
2662 2663
                     virConnectPtr conn, virDomainRunningReason reason,
                     enum qemuDomainAsyncJob asyncJob)
2664 2665 2666 2667
{
    int ret;
    qemuDomainObjPrivatePtr priv = vm->privateData;

2668 2669 2670 2671 2672 2673 2674 2675 2676 2677
    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);

2678 2679 2680 2681 2682
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
    if (ret == 0) {
        ret = qemuMonitorStartCPUs(priv->mon, conn);
        qemuDomainObjExitMonitorWithDriver(driver, vm);
    }
J
Jiri Denemark 已提交
2683

2684
    if (ret == 0) {
J
Jiri Denemark 已提交
2685
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
2686 2687 2688 2689 2690
    } 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));
    }
2691 2692 2693 2694 2695

    return ret;
}


J
Jiri Denemark 已提交
2696
int qemuProcessStopCPUs(struct qemud_driver *driver, virDomainObjPtr vm,
2697 2698
                        virDomainPausedReason reason,
                        enum qemuDomainAsyncJob asyncJob)
2699 2700 2701 2702
{
    int ret;
    qemuDomainObjPrivatePtr priv = vm->privateData;

2703
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
2704

2705
    ret = qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob);
2706 2707 2708 2709
    if (ret == 0) {
        ret = qemuMonitorStopCPUs(priv->mon);
        qemuDomainObjExitMonitorWithDriver(driver, vm);
    }
J
Jiri Denemark 已提交
2710

2711
    if (ret == 0) {
2712
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
2713 2714 2715 2716
        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
2717

2718 2719 2720 2721 2722
    return ret;
}



2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735
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;
}

2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748
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)) {
2749
           if (virDomainConfNWFilterInstantiate(conn, def->uuid, net) < 0) {
2750 2751 2752 2753 2754 2755 2756 2757 2758
                err = 1;
                break;
            }
        }
    }

    return err;
}

2759 2760 2761 2762 2763
static int
qemuProcessUpdateState(struct qemud_driver *driver, virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
2764
    virDomainPausedReason reason;
2765 2766
    virDomainState newState = VIR_DOMAIN_NOSTATE;
    int newReason;
2767
    bool running;
2768
    char *msg = NULL;
2769 2770
    int ret;

2771
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
2772
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
2773 2774 2775 2776 2777 2778 2779 2780
    qemuDomainObjExitMonitorWithDriver(driver, vm);

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

    state = virDomainObjGetState(vm, NULL);

    if (state == VIR_DOMAIN_PAUSED && running) {
2781 2782 2783
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
        msg = strdup("was unpaused");
2784
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
2785 2786 2787 2788 2789 2790 2791
        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
2792 2793
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
2794
        }
2795
    } else if (state == VIR_DOMAIN_SHUTOFF && running) {
2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814
        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);
2815 2816 2817 2818 2819
    }

    return 0;
}

2820 2821 2822 2823 2824 2825 2826 2827 2828
static int
qemuProcessRecoverMigration(struct qemud_driver *driver,
                            virDomainObjPtr vm,
                            virConnectPtr conn,
                            enum qemuDomainAsyncJob job,
                            enum qemuMigrationJobPhase phase,
                            virDomainState state,
                            int reason)
{
2829 2830
    qemuDomainObjPrivatePtr priv = vm->privateData;

2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853
    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,
2854 2855
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885
                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);
2886
            qemuDomainObjEnterMonitor(driver, vm);
2887 2888
            ignore_value(qemuMonitorMigrateCancel(priv->mon));
            qemuDomainObjExitMonitor(driver, vm);
2889 2890 2891 2892 2893 2894
            /* 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,
2895 2896
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915
                    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,
2916 2917
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931
                    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;
}

2932 2933 2934 2935 2936 2937
static int
qemuProcessRecoverJob(struct qemud_driver *driver,
                      virDomainObjPtr vm,
                      virConnectPtr conn,
                      const struct qemuDomainJobObj *job)
{
2938
    qemuDomainObjPrivatePtr priv = vm->privateData;
2939 2940 2941 2942 2943 2944 2945 2946
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
    case QEMU_ASYNC_JOB_MIGRATION_IN:
2947 2948 2949
        if (qemuProcessRecoverMigration(driver, vm, conn, job->asyncJob,
                                        job->phase, state, reason) < 0)
            return -1;
2950 2951 2952 2953
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
2954
        qemuDomainObjEnterMonitor(driver, vm);
2955 2956
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
        qemuDomainObjExitMonitor(driver, vm);
2957
        /* resume the domain but only if it was paused as a result of
2958 2959 2960
         * 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.  */
2961 2962 2963 2964 2965 2966 2967
        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,
2968 2969
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982
                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;

2983 2984 2985 2986
    /* 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.
     */
2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006
    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;

3007
    case QEMU_JOB_MIGRATION_OP:
3008
    case QEMU_JOB_ABORT:
3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019
    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;
}

3020 3021 3022
struct qemuProcessReconnectData {
    virConnectPtr conn;
    struct qemud_driver *driver;
3023 3024
    void *payload;
    struct qemuDomainJobObj oldjob;
3025 3026 3027 3028
};
/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
S
Stefan Berger 已提交
3029 3030 3031 3032
 *
 * 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.
3033 3034
 */
static void
3035
qemuProcessReconnect(void *opaque)
3036 3037 3038
{
    struct qemuProcessReconnectData *data = opaque;
    struct qemud_driver *driver = data->driver;
3039
    virDomainObjPtr obj = data->payload;
3040 3041
    qemuDomainObjPrivatePtr priv;
    virConnectPtr conn = data->conn;
3042
    struct qemuDomainJobObj oldjob;
3043 3044
    int state;
    int reason;
3045

3046 3047 3048 3049 3050
    memcpy(&oldjob, &data->oldjob, sizeof(oldjob));

    VIR_FREE(data);

    qemuDriverLock(driver);
3051 3052
    virDomainObjLock(obj);

3053

3054 3055 3056 3057
    VIR_DEBUG("Reconnect monitor to %p '%s'", obj, obj->def->name);

    priv = obj->privateData;

3058 3059
    /* Job was started by the caller for us */
    qemuDomainObjTransferJob(obj);
3060

3061 3062
    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted if qemuConnectMonitor() failed */
3063
    virObjectRef(obj);
3064 3065 3066 3067 3068

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

D
Daniel P. Berrange 已提交
3069 3070 3071 3072 3073 3074 3075 3076
    /* 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;
    }

3077 3078 3079 3080
    if (qemuUpdateActivePciHostdevs(driver, obj->def) < 0) {
        goto error;
    }

3081 3082 3083
    if (qemuUpdateActiveUsbHostdevs(driver, obj->def) < 0)
        goto error;

3084 3085 3086
    if (qemuProcessUpdateState(driver, obj) < 0)
        goto error;

3087 3088
    state = virDomainObjGetState(obj, &reason);
    if (state == VIR_DOMAIN_SHUTOFF) {
3089 3090 3091 3092 3093
        VIR_DEBUG("Domain '%s' wasn't fully started yet, killing it",
                  obj->def->name);
        goto error;
    }

3094 3095 3096
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
3097
    if (!priv->caps &&
3098 3099
        !(priv->caps = qemuCapsCacheLookupCopy(driver->capsCache,
                                               obj->def->emulator)))
3100 3101
        goto error;

3102
    /* In case the domain shutdown while we were not running,
3103 3104 3105
     * we need to finish the shutdown process. And we need to do it after
     * we have qemuCaps filled in.
     */
3106 3107 3108 3109 3110 3111
    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);
3112 3113 3114
        goto endjob;
    }

3115 3116
    if (qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE))
        if ((qemuDomainAssignAddresses(obj->def, priv->caps, obj)) < 0)
3117
            goto error;
3118

3119
    if (virSecurityManagerReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
3120 3121
        goto error;

3122 3123 3124
    if (qemuProcessNotifyNets(obj->def) < 0)
        goto error;

3125 3126 3127
    if (qemuProcessFiltersInstantiate(conn, obj->def))
        goto error;

3128
    if (qemuDomainCheckEjectableMedia(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
3129 3130
        goto error;

3131 3132 3133
    if (qemuProcessRecoverJob(driver, obj, conn, &oldjob) < 0)
        goto error;

3134 3135 3136 3137
    /* update domain state XML with possibly updated state in virDomainObj */
    if (virDomainSaveStatus(driver->caps, driver->stateDir, obj) < 0)
        goto error;

3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154
    /* 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;
    }

3155 3156 3157
    if (obj->def->id >= driver->nextvmid)
        driver->nextvmid = obj->def->id + 1;

3158
endjob:
3159
    if (!qemuDomainObjEndJob(driver, obj))
3160 3161
        obj = NULL;

3162
    if (obj && virObjectUnref(obj))
3163 3164
        virDomainObjUnlock(obj);

3165 3166
    qemuDriverUnlock(driver);

S
Stefan Berger 已提交
3167 3168
    virConnectClose(conn);

3169 3170 3171
    return;

error:
3172
    if (!qemuDomainObjEndJob(driver, obj))
3173 3174
        obj = NULL;

3175 3176
    if (obj) {
        if (!virDomainObjIsActive(obj)) {
3177
            if (virObjectUnref(obj))
3178 3179 3180 3181
                virDomainObjUnlock(obj);
            qemuDriverUnlock(driver);
            return;
        }
3182

3183
        if (virObjectUnref(obj)) {
3184
            /* We can't get the monitor back, so must kill the VM
3185 3186 3187
             * to remove danger of it ending up running twice if
             * user tries to start it again later
             */
3188
            if (qemuCapsGet(priv->caps, QEMU_CAPS_NO_SHUTDOWN)) {
3189 3190 3191 3192 3193 3194 3195 3196 3197
                /* 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;
            }
3198
            qemuProcessStop(driver, obj, state, 0);
3199
            if (!obj->persistent)
3200
                qemuDomainRemoveInactive(driver, obj);
3201 3202 3203
            else
                virDomainObjUnlock(obj);
        }
3204
    }
3205
    qemuDriverUnlock(driver);
S
Stefan Berger 已提交
3206 3207

    virConnectClose(conn);
3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 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
}

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
3258 3259 3260 3261 3262 3263
    /* 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);

3264
    if (virThreadCreate(&thread, true, qemuProcessReconnect, data) < 0) {
S
Stefan Berger 已提交
3265 3266 3267

        virConnectClose(data->conn);

3268 3269 3270
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not create thread. QEMU initialization "
                         "might be incomplete"));
3271
        if (!qemuDomainObjEndJob(src->driver, obj)) {
3272
            obj = NULL;
3273
        } else if (virObjectUnref(obj)) {
3274 3275
           /* We can't spawn a thread and thus connect to monitor.
            * Kill qemu */
3276
            qemuProcessStop(src->driver, obj, VIR_DOMAIN_SHUTOFF_FAILED, 0);
3277
            if (!obj->persistent)
3278
                qemuDomainRemoveInactive(src->driver, obj);
3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290
            else
                virDomainObjUnlock(obj);
        }
        goto error;
    }

    virDomainObjUnlock(obj);

    return;

error:
    VIR_FREE(data);
3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301
}

/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
qemuProcessReconnectAll(virConnectPtr conn, struct qemud_driver *driver)
{
3302
    struct qemuProcessReconnectData data = {.conn = conn, .driver = driver};
3303
    virHashForEach(driver->domains.objs, qemuProcessReconnectHelper, &data);
3304 3305 3306 3307 3308 3309 3310 3311
}

int qemuProcessStart(virConnectPtr conn,
                     struct qemud_driver *driver,
                     virDomainObjPtr vm,
                     const char *migrateFrom,
                     int stdin_fd,
                     const char *stdin_path,
3312
                     virDomainSnapshotObjPtr snapshot,
3313 3314
                     enum virNetDevVPortProfileOp vmop,
                     unsigned int flags)
3315 3316 3317 3318 3319 3320 3321 3322 3323
{
    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
3324
    unsigned long cur_balloon;
3325
    int i;
3326
    char *nodeset = NULL;
3327
    virBitmapPtr nodemask = NULL;
3328
    unsigned int stop_flags;
3329

3330 3331 3332 3333 3334 3335
    /* 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);

3336 3337 3338 3339 3340 3341
    /* 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;

3342 3343 3344 3345
    hookData.conn = conn;
    hookData.vm = vm;
    hookData.driver = driver;

3346
    VIR_DEBUG("Beginning VM startup process");
3347 3348

    if (virDomainObjIsActive(vm)) {
3349 3350
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
3351 3352 3353 3354 3355 3356 3357
        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
     */
3358
    VIR_DEBUG("Setting current domain def as transient");
3359 3360 3361 3362
    if (virDomainObjSetDefTransient(driver->caps, vm, true) < 0)
        goto cleanup;

    vm->def->id = driver->nextvmid++;
3363
    qemuDomainSetFakeReboot(driver, vm, false);
3364
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, VIR_DOMAIN_SHUTOFF_UNKNOWN);
3365

3366
    /* Run an early hook to set-up missing devices */
3367
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3368
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0, false);
3369 3370 3371
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
3372 3373
                              VIR_HOOK_QEMU_OP_PREPARE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
3374 3375 3376 3377 3378 3379 3380 3381 3382
        VIR_FREE(xml);

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

3383
    /* Must be run before security labelling */
3384
    VIR_DEBUG("Preparing host devices");
3385 3386 3387
    if (qemuPrepareHostDevices(driver, vm->def) < 0)
        goto cleanup;

3388
    VIR_DEBUG("Preparing chr devices");
3389 3390 3391 3392 3393 3394 3395 3396
    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 */
3397
    VIR_DEBUG("Generating domain security label (if required)");
3398
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0) {
3399
        virDomainAuditSecurityLabel(vm, false);
3400 3401
        goto cleanup;
    }
3402
    virDomainAuditSecurityLabel(vm, true);
3403 3404 3405

    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
3406
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
3407 3408 3409 3410 3411 3412
    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) {
3413
            int port = qemuProcessNextFreePort(driver, driver->remotePortMin);
3414
            if (port < 0) {
3415
                virReportError(VIR_ERR_INTERNAL_ERROR,
3416
                               "%s", _("Unable to find an unused port for VNC"));
3417 3418 3419
                goto cleanup;
            }
            vm->def->graphics[0]->data.vnc.port = port;
M
Michal Privoznik 已提交
3420 3421 3422 3423
        } 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) {
3424
                port = qemuProcessNextFreePort(driver, driver->remotePortMin);
M
Michal Privoznik 已提交
3425 3426

                if (port < 0) {
3427
                    virReportError(VIR_ERR_INTERNAL_ERROR,
3428
                                   "%s", _("Unable to find an unused port for SPICE"));
M
Michal Privoznik 已提交
3429 3430 3431 3432
                    goto cleanup;
                }

                vm->def->graphics[0]->data.spice.port = port;
3433
            }
M
Michal Privoznik 已提交
3434 3435 3436 3437 3438
            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);
3439
                if (tlsPort < 0) {
3440
                    virReportError(VIR_ERR_INTERNAL_ERROR,
3441
                                   "%s", _("Unable to find an unused port for SPICE TLS"));
3442 3443 3444 3445
                    qemuProcessReturnPort(driver, port);
                    goto cleanup;
                }

3446
                vm->def->graphics[0]->data.spice.tlsPort = tlsPort;
M
Michal Privoznik 已提交
3447
            }
3448
        }
3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469

        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;
                }
            }
        }
3470 3471
    }

3472
    if (virFileMakePath(driver->logDir) < 0) {
3473 3474 3475 3476 3477 3478
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
                             driver->logDir);
        goto cleanup;
    }

3479
    VIR_DEBUG("Creating domain log file");
3480
    if ((logfile = qemuDomainCreateLog(driver, vm, false)) < 0)
3481 3482
        goto cleanup;

3483 3484 3485
    if (vm->def->virtType == VIR_DOMAIN_VIRT_KVM) {
        VIR_DEBUG("Checking for KVM availability");
        if (access("/dev/kvm", F_OK) != 0) {
3486 3487 3488 3489
            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."));
3490 3491 3492 3493
            goto cleanup;
        }
    }

3494
    VIR_DEBUG("Determining emulator version");
3495
    virObjectUnref(priv->caps);
3496 3497
    if (!(priv->caps = qemuCapsCacheLookupCopy(driver->capsCache,
                                               vm->def->emulator)))
3498 3499
        goto cleanup;

3500
    if (qemuAssignDeviceAliases(vm->def, priv->caps) < 0)
3501 3502 3503
        goto cleanup;

    VIR_DEBUG("Checking for CDROM and floppy presence");
3504 3505
    if (qemuDomainCheckDiskPresence(driver, vm,
                                    flags & VIR_QEMU_PROCESS_START_COLD) < 0)
3506 3507
        goto cleanup;

3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520
    /* 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);

3521 3522
        if (virBitmapParse(nodeset, 0, &nodemask,
                           VIR_DOMAIN_CPUMASK_LEN) < 0)
3523 3524 3525 3526
            goto cleanup;
    }
    hookData.nodemask = nodemask;

3527
    VIR_DEBUG("Setting up domain cgroup (if required)");
3528
    if (qemuSetupCgroup(driver, vm, nodemask) < 0)
3529 3530 3531 3532 3533 3534 3535
        goto cleanup;

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

3536
    VIR_DEBUG("Preparing monitor state");
3537 3538 3539
    if (qemuProcessPrepareMonitorChr(driver, priv->monConfig, vm->def->name) < 0)
        goto cleanup;

3540
    if (qemuCapsGet(priv->caps, QEMU_CAPS_MONITOR_JSON))
3541 3542 3543 3544
        priv->monJSON = 1;
    else
        priv->monJSON = 0;

3545 3546
    priv->monError = false;
    priv->monStart = 0;
3547 3548
    priv->gotShutdown = false;

3549
    VIR_FREE(priv->pidfile);
3550
    if (!(priv->pidfile = virPidFileBuildPath(driver->stateDir, vm->def->name))) {
3551 3552
        virReportSystemError(errno,
                             "%s", _("Failed to build pidfile path."));
3553 3554 3555
        goto cleanup;
    }

3556 3557
    if (unlink(priv->pidfile) < 0 &&
        errno != ENOENT) {
3558
        virReportSystemError(errno,
3559 3560
                             _("Cannot remove stale PID file %s"),
                             priv->pidfile);
3561 3562 3563 3564 3565 3566 3567 3568 3569 3570
        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
     */
3571
    if (qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE)) {
3572
        VIR_DEBUG("Assigning domain PCI addresses");
3573
        if ((qemuDomainAssignAddresses(vm->def, priv->caps, vm)) < 0)
3574
            goto cleanup;
3575 3576
    }

3577
    VIR_DEBUG("Building emulator command line");
3578
    if (!(cmd = qemuBuildCommandLine(conn, driver, vm->def, priv->monConfig,
3579
                                     priv->monJSON != 0, priv->caps,
3580
                                     migrateFrom, stdin_fd, snapshot, vmop)))
3581 3582 3583 3584
        goto cleanup;

    /* now that we know it is about to start call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3585
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0, false);
3586 3587 3588
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
3589 3590
                              VIR_HOOK_QEMU_OP_START, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
3591 3592 3593 3594 3595 3596 3597 3598 3599
        VIR_FREE(xml);

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

3600
    if ((timestamp = virTimeStringNow()) == NULL) {
3601 3602 3603 3604 3605 3606
        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",
3607
                     virStrerror(errno, ebuf, sizeof(ebuf)));
3608 3609 3610 3611 3612 3613 3614
        }

        VIR_FREE(timestamp);
    }

    virCommandWriteArgLog(cmd, logfile);

3615
    qemuDomainObjCheckTaint(driver, vm, logfile);
3616

3617 3618
    if ((pos = lseek(logfile, 0, SEEK_END)) < 0)
        VIR_WARN("Unable to seek to end of logfile: %s",
3619
                 virStrerror(errno, ebuf, sizeof(ebuf)));
3620 3621 3622 3623 3624 3625

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

3626 3627 3628 3629 3630 3631
    /* 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);
    }

3632 3633 3634 3635 3636
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);

    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
3637
    virCommandSetPidFile(cmd, priv->pidfile);
3638
    virCommandDaemonize(cmd);
3639
    virCommandRequireHandshake(cmd);
3640 3641 3642

    ret = virCommandRun(cmd, NULL);

E
Eric Blake 已提交
3643
    /* wait for qemu process to show up */
3644
    if (ret == 0) {
3645
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
3646 3647
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
3648 3649 3650 3651 3652 3653 3654 3655 3656
            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.
         */

3657
        /* The virCommand process that launches the daemon failed. Pending on
3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668
         * 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
    }

3669 3670 3671 3672 3673
    VIR_DEBUG("Writing early domain status to disk");
    if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0) {
        goto cleanup;
    }

3674 3675 3676 3677 3678 3679 3680
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
        goto cleanup;
    }

    VIR_DEBUG("Setting domain security labels");
    if (virSecurityManagerSetAllLabel(driver->securityManager,
3681
                                      vm->def, stdin_path) < 0)
3682 3683
        goto cleanup;

3684 3685 3686 3687 3688 3689
    /* 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;

3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703
    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) &&
3704
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def, stdin_fd) < 0)
3705 3706 3707 3708 3709 3710 3711 3712 3713
            goto cleanup;
    }

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

3714
    if (migrateFrom)
3715
        flags |= VIR_QEMU_PROCESS_START_PAUSED;
3716 3717 3718 3719 3720 3721 3722

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

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

3723
    VIR_DEBUG("Waiting for monitor to show up");
3724
    if (qemuProcessWaitForMonitor(driver, vm, priv->caps, pos) < 0)
3725 3726
        goto cleanup;

D
Daniel P. Berrange 已提交
3727 3728 3729 3730 3731 3732 3733 3734
    /* 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;
    }

3735
    VIR_DEBUG("Detecting VCPU PIDs");
3736 3737 3738
    if (qemuProcessDetectVcpuPIDs(driver, vm) < 0)
        goto cleanup;

3739
    VIR_DEBUG("Setting cgroup for each VCPU (if required)");
3740 3741 3742
    if (qemuSetupCgroupForVcpu(driver, vm) < 0)
        goto cleanup;

3743 3744 3745 3746
    VIR_DEBUG("Setting cgroup for emulator (if required)");
    if (qemuSetupCgroupForEmulator(driver, vm) < 0)
        goto cleanup;

3747
    VIR_DEBUG("Setting VCPU affinities");
3748 3749 3750
    if (qemuProcessSetVcpuAffinites(conn, vm) < 0)
        goto cleanup;

3751 3752 3753 3754
    VIR_DEBUG("Setting affinity of emulator threads");
    if (qemuProcessSetEmulatorAffinites(conn, vm) < 0)
        goto cleanup;

3755
    VIR_DEBUG("Setting any required VM passwords");
3756
    if (qemuProcessInitPasswords(conn, driver, vm) < 0)
3757 3758 3759 3760
        goto cleanup;

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
3761
    if (!qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE)) {
3762
        VIR_DEBUG("Determining domain device PCI addresses");
3763 3764 3765 3766
        if (qemuProcessInitPCIAddresses(driver, vm) < 0)
            goto cleanup;
    }

3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778
    /* 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);

3779 3780 3781 3782
    /* 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.  */
3783
    VIR_DEBUG("Setting initial memory amount");
E
Eric Blake 已提交
3784
    cur_balloon = vm->def->mem.cur_balloon;
3785
    if (cur_balloon != vm->def->mem.cur_balloon) {
3786 3787 3788
        virReportError(VIR_ERR_OVERFLOW,
                       _("unable to set balloon to %lld"),
                       vm->def->mem.cur_balloon);
3789 3790
        goto cleanup;
    }
3791
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
E
Eric Blake 已提交
3792
    if (qemuMonitorSetBalloon(priv->mon, cur_balloon) < 0) {
3793 3794 3795 3796 3797
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        goto cleanup;
    }
    qemuDomainObjExitMonitorWithDriver(driver, vm);

3798
    if (!(flags & VIR_QEMU_PROCESS_START_PAUSED)) {
3799
        VIR_DEBUG("Starting domain CPUs");
3800
        /* Allow the CPUS to start executing */
J
Jiri Denemark 已提交
3801
        if (qemuProcessStartCPUs(driver, vm, conn,
3802 3803
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 QEMU_ASYNC_JOB_NONE) < 0) {
3804
            if (virGetLastError() == NULL)
3805 3806
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("resume operation failed"));
3807 3808
            goto cleanup;
        }
J
Jiri Denemark 已提交
3809 3810 3811 3812 3813
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             migrateFrom ?
                             VIR_DOMAIN_PAUSED_MIGRATION :
                             VIR_DOMAIN_PAUSED_USER);
3814 3815
    }

3816
    if (flags & VIR_QEMU_PROCESS_START_AUTODESROY &&
3817 3818 3819
        qemuProcessAutoDestroyAdd(driver, vm, conn) < 0)
        goto cleanup;

3820
    VIR_DEBUG("Writing domain status to disk");
3821 3822 3823
    if (virDomainSaveStatus(driver->caps, driver->stateDir, vm) < 0)
        goto cleanup;

3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840
    /* 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;
    }

3841 3842 3843 3844 3845 3846 3847 3848 3849
    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 */
3850
    VIR_FREE(nodeset);
3851
    virBitmapFree(nodemask);
3852 3853
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);
3854
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, stop_flags);
3855 3856 3857 3858 3859

    return -1;
}


3860 3861 3862
int
qemuProcessKill(struct qemud_driver *driver,
                virDomainObjPtr vm, unsigned int flags)
3863
{
3864
    int ret;
3865 3866 3867

    VIR_DEBUG("vm=%s pid=%d flags=%x",
              vm->def->name, vm->pid, flags);
3868

3869 3870 3871 3872 3873
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
3874 3875
    }

3876 3877 3878 3879 3880 3881
    if ((flags & VIR_QEMU_PROCESS_KILL_NOWAIT)) {
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
3882

3883 3884
    if (driver)
        qemuDriverUnlock(driver);
3885

3886 3887
    ret = virProcessKillPainfully(vm->pid,
                                  !!(flags & VIR_QEMU_PROCESS_KILL_FORCE));
3888

3889
    if (driver) {
3890
        virObjectRef(vm);
3891 3892 3893
        virDomainObjUnlock(vm);
        qemuDriverLock(driver);
        virDomainObjLock(vm);
3894
        virObjectUnref(vm);
3895
    }
3896

3897
    return ret;
3898 3899 3900
}


3901 3902
void qemuProcessStop(struct qemud_driver *driver,
                     virDomainObjPtr vm,
3903 3904
                     virDomainShutoffReason reason,
                     unsigned int flags)
3905 3906 3907 3908 3909 3910
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
3911
    virNetDevVPortProfilePtr vport = NULL;
3912 3913 3914 3915 3916
    int i;
    int logfile = -1;
    char *timestamp;
    char ebuf[1024];

3917 3918
    VIR_DEBUG("Shutting down VM '%s' pid=%d flags=%x",
              vm->def->name, vm->pid, flags);
3919 3920 3921 3922 3923 3924

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

3925 3926 3927 3928 3929 3930 3931
    /*
     * 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;

3932
    if ((logfile = qemuDomainCreateLog(driver, vm, true)) < 0) {
3933 3934 3935
        /* To not break the normal domain shutdown process, skip the
         * timestamp log writing if failed on opening log file. */
        VIR_WARN("Unable to open logfile: %s",
3936
                  virStrerror(errno, ebuf, sizeof(ebuf)));
3937
    } else {
3938
        if ((timestamp = virTimeStringNow()) == NULL) {
3939 3940 3941 3942 3943 3944
            virReportOOMError();
        } else {
            if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
                safewrite(logfile, SHUTDOWN_POSTFIX,
                          strlen(SHUTDOWN_POSTFIX)) < 0) {
                VIR_WARN("Unable to write timestamp to logfile: %s",
3945
                         virStrerror(errno, ebuf, sizeof(ebuf)));
3946 3947 3948 3949 3950 3951 3952
            }

            VIR_FREE(timestamp);
        }

        if (VIR_CLOSE(logfile) < 0)
             VIR_WARN("Unable to close logfile: %s",
3953
                      virStrerror(errno, ebuf, sizeof(ebuf)));
3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968
    }

    /* 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,
3969
                                                  &net->mac))) {
3970
                virReportSystemError(errno,
3971
             _("failed to remove ebtables rule to allow MAC address on '%s'"),
3972 3973 3974 3975 3976
                                     net->ifname);
            }
        }
    }

D
Daniel P. Berrange 已提交
3977 3978 3979 3980 3981 3982
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
        priv->agentError = false;
    }

3983
    if (priv->mon) {
3984
        qemuMonitorClose(priv->mon);
3985 3986
        priv->mon = NULL;
    }
3987 3988 3989 3990 3991 3992 3993 3994 3995

    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 */
3996 3997
    ignore_value(qemuProcessKill(driver, vm, VIR_QEMU_PROCESS_KILL_FORCE|
                                             VIR_QEMU_PROCESS_KILL_NOCHECK));
3998

3999 4000
    qemuDomainCleanupRun(driver, vm);

4001 4002 4003
    /* Stop autodestroy in case guest is restarted */
    qemuProcessAutoDestroyRemove(driver, vm);

4004 4005
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4006
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0, false);
4007 4008 4009

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4010 4011
                    VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
4012 4013 4014
        VIR_FREE(xml);
    }

4015 4016 4017 4018 4019
    /* 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);
4020
    virSecurityManagerReleaseLabel(driver->securityManager, vm->def);
4021 4022

    /* Clear out dynamically assigned labels */
4023 4024 4025 4026 4027
    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);
4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041
    }

    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];
4042
        if (virDomainNetGetActualType(net) == VIR_DOMAIN_NET_TYPE_DIRECT) {
4043
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
4044
                             net->ifname, &net->mac,
4045 4046
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
4047
                             virDomainNetGetActualVirtPortProfile(net),
4048
                             driver->stateDir));
4049 4050
            VIR_FREE(net->ifname);
        }
4051 4052 4053
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
A
Ansis Atteka 已提交
4054 4055 4056 4057 4058 4059
        vport = virDomainNetGetActualVirtPortProfile(net);
        if (vport && vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH)
            ignore_value(virNetDevOpenvswitchRemovePort(
                                       virDomainNetGetActualBridgeName(net),
                                       net->ifname));

4060 4061
        networkReleaseActualDevice(net);
    }
4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089

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);
    }

4090
    vm->taint = 0;
4091
    vm->pid = -1;
J
Jiri Denemark 已提交
4092
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
4093 4094
    VIR_FREE(priv->vcpupids);
    priv->nvcpupids = 0;
4095 4096
    virObjectUnref(priv->caps);
    priv->caps = NULL;
4097
    VIR_FREE(priv->pidfile);
4098

4099
    /* The "release" hook cleans up additional resources */
4100
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4101
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0, false);
4102 4103 4104

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4105 4106
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
4107 4108 4109
        VIR_FREE(xml);
    }

4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121
    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);
    }
}
4122 4123


4124 4125 4126
int qemuProcessAttach(virConnectPtr conn ATTRIBUTE_UNUSED,
                      struct qemud_driver *driver,
                      virDomainObjPtr vm,
4127
                      pid_t pid,
4128 4129 4130 4131
                      const char *pidfile,
                      virDomainChrSourceDefPtr monConfig,
                      bool monJSON)
{
4132
    size_t i;
4133 4134 4135 4136 4137
    char ebuf[1024];
    int logfile = -1;
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool running = true;
4138
    virDomainPausedReason reason;
4139
    virSecurityLabelPtr seclabel = NULL;
4140 4141 4142
    virSecurityLabelDefPtr seclabeldef = NULL;
    virSecurityManagerPtr* sec_managers = NULL;
    const char *model;
4143 4144 4145 4146

    VIR_DEBUG("Beginning VM attach process");

    if (virDomainObjIsActive(vm)) {
4147 4148
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174
        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");
4175 4176
    sec_managers = virSecurityManagerGetNested(driver->securityManager);
    if (sec_managers == NULL) {
4177
        goto cleanup;
4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199
    }

    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);
    }
4200 4201 4202 4203 4204 4205

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

    VIR_DEBUG("Determining emulator version");
4206
    virObjectUnref(priv->caps);
4207 4208
    if (!(priv->caps = qemuCapsCacheLookupCopy(driver->capsCache,
                                               vm->def->emulator)))
4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224
        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
     */
4225
    if (qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE)) {
4226
        VIR_DEBUG("Assigning domain PCI addresses");
4227
        if ((qemuDomainAssignAddresses(vm->def, priv->caps, vm)) < 0)
4228
            goto cleanup;
4229 4230
    }

4231
    if ((timestamp = virTimeStringNow()) == NULL) {
4232 4233 4234 4235 4236 4237
        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",
4238
                     virStrerror(errno, ebuf, sizeof(ebuf)));
4239 4240 4241 4242 4243 4244 4245 4246 4247 4248
        }

        VIR_FREE(timestamp);
    }

    qemuDomainObjTaint(driver, vm, VIR_DOMAIN_TAINT_EXTERNAL_LAUNCH, logfile);

    vm->pid = pid;

    VIR_DEBUG("Waiting for monitor to show up");
4249
    if (qemuProcessWaitForMonitor(driver, vm, priv->caps, -1) < 0)
4250 4251
        goto cleanup;

D
Daniel P. Berrange 已提交
4252 4253 4254 4255 4256 4257 4258 4259
    /* 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;
    }

4260 4261 4262 4263 4264 4265
    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 */
4266
    if (!qemuCapsGet(priv->caps, QEMU_CAPS_DEVICE)) {
4267 4268 4269 4270 4271 4272
        VIR_DEBUG("Determining domain device PCI addresses");
        if (qemuProcessInitPCIAddresses(driver, vm) < 0)
            goto cleanup;
    }

    VIR_DEBUG("Getting initial memory amount");
4273
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
4274 4275 4276 4277
    if (qemuMonitorGetBalloonInfo(priv->mon, &vm->def->mem.cur_balloon) < 0) {
        qemuDomainObjExitMonitorWithDriver(driver, vm);
        goto cleanup;
    }
4278
    if (qemuMonitorGetStatus(priv->mon, &running, &reason) < 0) {
4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294
        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
4295
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
4296 4297 4298 4299 4300

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

4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317
    /* 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;
    }

4318 4319
    VIR_FORCE_CLOSE(logfile);
    VIR_FREE(seclabel);
4320
    VIR_FREE(sec_managers);
4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331

    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);
4332
    VIR_FREE(sec_managers);
4333 4334 4335 4336 4337
    virDomainChrSourceDefFree(monConfig);
    return -1;
}


4338 4339 4340 4341
static virDomainObjPtr
qemuProcessAutoDestroy(struct qemud_driver *driver,
                       virDomainObjPtr dom,
                       virConnectPtr conn)
4342
{
4343
    qemuDomainObjPrivatePtr priv = dom->privateData;
4344 4345
    virDomainEventPtr event = NULL;

4346
    VIR_DEBUG("vm=%s, conn=%p", dom->def->name, conn);
4347

4348 4349
    if (priv->job.asyncJob) {
        VIR_DEBUG("vm=%s has long-term job active, cancelling",
4350
                  dom->def->name);
4351
        qemuDomainObjDiscardAsyncJob(driver, dom);
4352 4353
    }

4354
    if (qemuDomainObjBeginJobWithDriver(driver, dom,
4355
                                        QEMU_JOB_DESTROY) < 0)
4356 4357 4358
        goto cleanup;

    VIR_DEBUG("Killing domain");
4359 4360
    qemuProcessStop(driver, dom, VIR_DOMAIN_SHUTOFF_DESTROYED,
                    VIR_QEMU_PROCESS_STOP_MIGRATED);
4361
    virDomainAuditStop(dom, "destroyed");
4362 4363 4364
    event = virDomainEventNewFromObj(dom,
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_DESTROYED);
4365

4366
    if (!qemuDomainObjEndJob(driver, dom))
4367 4368
        dom = NULL;
    if (dom && !dom->persistent)
4369
        qemuDomainRemoveInactive(driver, dom);
4370
    if (event)
4371
        qemuDomainEventQueue(driver, event);
4372

4373 4374
cleanup:
    return dom;
4375 4376 4377 4378 4379 4380
}

int qemuProcessAutoDestroyAdd(struct qemud_driver *driver,
                              virDomainObjPtr vm,
                              virConnectPtr conn)
{
4381 4382 4383
    VIR_DEBUG("vm=%s, conn=%p", vm->def->name, conn);
    return qemuDriverCloseCallbackSet(driver, vm, conn,
                                      qemuProcessAutoDestroy);
4384 4385 4386 4387 4388
}

int qemuProcessAutoDestroyRemove(struct qemud_driver *driver,
                                 virDomainObjPtr vm)
{
4389 4390
    VIR_DEBUG("vm=%s", vm->def->name);
    return qemuDriverCloseCallbackUnset(driver, vm, qemuProcessAutoDestroy);
4391
}
4392 4393 4394 4395

bool qemuProcessAutoDestroyActive(struct qemud_driver *driver,
                                  virDomainObjPtr vm)
{
4396 4397 4398 4399
    qemuDriverCloseCallback cb;
    VIR_DEBUG("vm=%s", vm->def->name);
    cb = qemuDriverCloseCallbackGet(driver, vm, NULL);
    return cb == qemuProcessAutoDestroy;
4400
}