qemu_process.c 181.9 KB
Newer Older
1
/*
2
 * qemu_process.c: QEMU process management
3
 *
4
 * Copyright (C) 2006-2015 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23 24 25 26 27
 *
 */

#include <config.h>

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

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

47
#include "cpu/cpu.h"
48
#include "datatypes.h"
49
#include "virlog.h"
50
#include "virerror.h"
51
#include "viralloc.h"
52
#include "virhook.h"
E
Eric Blake 已提交
53
#include "virfile.h"
54
#include "virpidfile.h"
55
#include "nodeinfo.h"
56
#include "domain_audit.h"
57
#include "domain_nwfilter.h"
58
#include "locking/domain_lock.h"
59
#include "network/bridge_driver.h"
60
#include "viruuid.h"
61
#include "virprocess.h"
62
#include "virtime.h"
A
Ansis Atteka 已提交
63
#include "virnetdevtap.h"
64
#include "virnetdevopenvswitch.h"
65
#include "virnetdevmidonet.h"
66
#include "virbitmap.h"
67
#include "viratomic.h"
68
#include "virnuma.h"
69
#include "virstring.h"
70
#include "virhostdev.h"
71
#include "storage/storage_driver.h"
72
#include "configmake.h"
73
#include "nwfilter_conf.h"
74
#include "netdev_bandwidth_conf.h"
75 76 77

#define VIR_FROM_THIS VIR_FROM_QEMU

78 79
VIR_LOG_INIT("qemu.qemu_process");

80
#define ATTACH_POSTFIX ": attaching\n"
81 82 83
#define SHUTDOWN_POSTFIX ": shutting down\n"

/**
84
 * qemuProcessRemoveDomainStatus
85 86 87 88 89 90
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
91
qemuProcessRemoveDomainStatus(virQEMUDriverPtr driver,
92 93 94 95
                              virDomainObjPtr vm)
{
    char ebuf[1024];
    char *file = NULL;
96
    qemuDomainObjPrivatePtr priv = vm->privateData;
97 98
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;
99

100
    if (virAsprintf(&file, "%s/%s.xml", cfg->stateDir, vm->def->name) < 0)
101
        goto cleanup;
102 103 104 105 106 107

    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);

108 109 110
    if (priv->pidfile &&
        unlink(priv->pidfile) < 0 &&
        errno != ENOENT)
111 112 113
        VIR_WARN("Failed to remove PID file for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));

114
    ret = 0;
115
 cleanup:
116 117
    virObjectUnref(cfg);
    return ret;
118 119 120 121
}


/* XXX figure out how to remove this */
122
extern virQEMUDriverPtr qemu_driver;
123

D
Daniel P. Berrange 已提交
124 125 126 127 128 129 130
/*
 * 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
131
qemuProcessHandleAgentEOF(qemuAgentPtr agent,
D
Daniel P. Berrange 已提交
132 133 134 135 136 137
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

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

138
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
139 140

    priv = vm->privateData;
141 142 143 144 145 146 147 148 149 150 151 152

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

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

    priv->agent = NULL;
D
Daniel P. Berrange 已提交
153

154
    virObjectUnlock(vm);
155 156

    qemuAgentClose(agent);
157 158
    return;

159
 unlock:
160 161
    virObjectUnlock(vm);
    return;
D
Daniel P. Berrange 已提交
162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178
}


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

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

179
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
180 181 182 183 184

    priv = vm->privateData;

    priv->agentError = true;

185
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
186 187 188 189 190
}

static void qemuProcessHandleAgentDestroy(qemuAgentPtr agent,
                                          virDomainObjPtr vm)
{
191 192
    VIR_DEBUG("Received destroy agent=%p vm=%p", agent, vm);

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


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


204
int
205
qemuConnectAgent(virQEMUDriverPtr driver, virDomainObjPtr vm)
D
Daniel P. Berrange 已提交
206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
{
    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

    ignore_value(virTimeMillisNow(&priv->agentStart));
227
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
228 229 230 231 232

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

233
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
234 235
    priv->agentStart = 0;

236 237 238 239 240 241 242 243 244 245 246
    if (agent == NULL)
        virObjectUnref(vm);

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

D
Daniel P. Berrange 已提交
247 248 249 250
    if (virSecurityManagerClearSocketLabel(driver->securityManager,
                                           vm->def) < 0) {
        VIR_ERROR(_("Failed to clear security context for agent for %s"),
                  vm->def->name);
251
        qemuAgentClose(agent);
D
Daniel P. Berrange 已提交
252 253 254 255 256 257 258 259 260 261 262 263 264
        goto cleanup;
    }


    priv->agent = agent;

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

    ret = 0;

265
 cleanup:
D
Daniel P. Berrange 已提交
266 267 268 269
    return ret;
}


270
/*
271
 * This is a callback registered with a qemuMonitorPtr instance,
272 273 274 275 276 277
 * 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,
278 279
                            virDomainObjPtr vm,
                            void *opaque)
280
{
281
    virQEMUDriverPtr driver = opaque;
282
    virObjectEventPtr event = NULL;
283
    qemuDomainObjPrivatePtr priv;
284 285 286
    int eventReason = VIR_DOMAIN_EVENT_STOPPED_SHUTDOWN;
    int stopReason = VIR_DOMAIN_SHUTOFF_SHUTDOWN;
    const char *auditReason = "shutdown";
287
    unsigned int stopFlags = 0;
288 289 290

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

291
    virObjectLock(vm);
292

293 294 295 296
    priv = vm->privateData;

    if (priv->beingDestroyed) {
        VIR_DEBUG("Domain is being destroyed, EOF is expected");
297
        goto cleanup;
298 299
    }

300 301
    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Domain %p is not active, ignoring EOF", vm);
302
        goto cleanup;
303 304
    }

305
    if (priv->monJSON && !priv->gotShutdown) {
306 307
        VIR_DEBUG("Monitor connection to '%s' closed without SHUTDOWN event; "
                  "assuming the domain crashed", vm->def->name);
308
        eventReason = VIR_DOMAIN_EVENT_STOPPED_FAILED;
309
        stopReason = VIR_DOMAIN_SHUTOFF_CRASHED;
310
        auditReason = "failed";
311 312
    }

313 314
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN) {
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
315 316
        qemuMigrationErrorSave(driver, vm->def->name,
                               qemuMonitorLastError(priv->mon));
317
    }
318

319
    event = virDomainEventLifecycleNewFromObj(vm,
320
                                     VIR_DOMAIN_EVENT_STOPPED,
321
                                     eventReason);
322
    qemuProcessStop(driver, vm, stopReason, stopFlags);
323
    virDomainAuditStop(vm, auditReason);
324

325
    qemuDomainRemoveInactive(driver, vm);
326

327
 cleanup:
328
    virObjectUnlock(vm);
329
    qemuDomainEventQueue(driver, event);
330 331 332 333 334 335 336 337 338 339 340
}


/*
 * 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,
341 342
                              virDomainObjPtr vm,
                              void *opaque)
343
{
344
    virQEMUDriverPtr driver = opaque;
345
    virObjectEventPtr event = NULL;
346 347 348

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

349
    virObjectLock(vm);
350

351
    ((qemuDomainObjPrivatePtr) vm->privateData)->monError = true;
352
    event = virDomainEventControlErrorNewFromObj(vm);
353
    qemuDomainEventQueue(driver, event);
354

355
    virObjectUnlock(vm);
356 357 358
}


359
virDomainDiskDefPtr
360 361 362
qemuProcessFindDomainDiskByAlias(virDomainObjPtr vm,
                                 const char *alias)
{
363
    size_t i;
364 365 366 367 368 369 370 371 372 373 374 375

    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;
    }

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

395
    if (!disk->src->encryption) {
396 397
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("disk %s does not have any encryption information"),
398
                       disk->src->path);
399 400
        return -1;
    }
401
    enc = disk->src->encryption;
402 403

    if (!conn) {
404 405
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("cannot find secrets without a connection"));
406 407 408 409
        goto cleanup;
    }

    if (conn->secretDriver == NULL ||
410 411
        conn->secretDriver->secretLookupByUUID == NULL ||
        conn->secretDriver->secretGetValue == NULL) {
412 413
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("secret storage not supported"));
414 415 416 417 418 419 420
        goto cleanup;
    }

    if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW ||
        enc->nsecrets != 1 ||
        enc->secrets[0]->type !=
        VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE) {
421
        virReportError(VIR_ERR_XML_ERROR,
422 423
                       _("invalid <encryption> for volume %s"),
                       virDomainDiskGetSource(disk));
424 425 426
        goto cleanup;
    }

427 428
    secret = conn->secretDriver->secretLookupByUUID(conn,
                                                    enc->secrets[0]->uuid);
429 430
    if (secret == NULL)
        goto cleanup;
431 432
    data = conn->secretDriver->secretGetValue(secret, &size, 0,
                                              VIR_SECRET_GET_VALUE_INTERNAL_CALL);
433
    virObjectUnref(secret);
434 435 436 437 438 439
    if (data == NULL)
        goto cleanup;

    if (memchr(data, '\0', size) != NULL) {
        memset(data, 0, size);
        VIR_FREE(data);
440 441
        virReportError(VIR_ERR_XML_ERROR,
                       _("format='qcow' passphrase for %s must not contain a "
442
                         "'\\0'"), virDomainDiskGetSource(disk));
443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461
        goto cleanup;
    }

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

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

    *secretRet = passphrase;
    *secretLen = size;

    ret = 0;

462
 cleanup:
463 464 465 466 467 468 469 470 471
    return ret;
}

static int
qemuProcessFindVolumeQcowPassphrase(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virConnectPtr conn,
                                    virDomainObjPtr vm,
                                    const char *path,
                                    char **secretRet,
472 473
                                    size_t *secretLen,
                                    void *opaque ATTRIBUTE_UNUSED)
474 475 476 477
{
    virDomainDiskDefPtr disk;
    int ret = -1;

478
    virObjectLock(vm);
479 480 481 482
    if (!(disk = virDomainDiskByName(vm->def, path, true))) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("no disk found with path %s"),
                       path);
483
        goto cleanup;
484
    }
485 486 487

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

488
 cleanup:
489
    virObjectUnlock(vm);
490 491 492 493 494 495
    return ret;
}


static int
qemuProcessHandleReset(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
496 497
                       virDomainObjPtr vm,
                       void *opaque)
498
{
499
    virQEMUDriverPtr driver = opaque;
500
    virObjectEventPtr event;
501
    qemuDomainObjPrivatePtr priv;
502
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
503

504
    virObjectLock(vm);
505

506
    event = virDomainEventRebootNewFromObj(vm);
507 508 509
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);
510 511 512

    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
        VIR_WARN("Failed to save status on vm %s", vm->def->name);
513

514
    virObjectUnlock(vm);
515

516
    qemuDomainEventQueue(driver, event);
517

518
    virObjectUnref(cfg);
519 520 521 522
    return 0;
}


523 524 525 526 527 528 529 530 531 532 533
/*
 * 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)
{
534
    virQEMUDriverPtr driver = qemu_driver;
535 536
    virDomainObjPtr vm = opaque;
    qemuDomainObjPrivatePtr priv = vm->privateData;
537
    virObjectEventPtr event = NULL;
538
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
539
    virDomainRunningReason reason = VIR_DOMAIN_RUNNING_BOOTED;
540
    int ret = -1, rc;
541

542
    VIR_DEBUG("vm=%p", vm);
543
    virObjectLock(vm);
544
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
545 546 547
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
548 549
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
550 551 552
        goto endjob;
    }

553
    qemuDomainObjEnterMonitor(driver, vm);
554 555 556
    rc = qemuMonitorSystemReset(priv->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0)
557 558
        goto endjob;

559
    if (rc < 0)
560 561
        goto endjob;

562 563 564
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_CRASHED)
        reason = VIR_DOMAIN_RUNNING_CRASHED;

565
    if (qemuProcessStartCPUs(driver, vm, NULL,
566
                             reason,
567
                             QEMU_ASYNC_JOB_NONE) < 0) {
568
        if (virGetLastError() == NULL)
569 570
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
571 572
        goto endjob;
    }
573
    priv->gotShutdown = false;
574
    event = virDomainEventLifecycleNewFromObj(vm,
575 576 577
                                     VIR_DOMAIN_EVENT_RESUMED,
                                     VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

578 579 580 581 582
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

583 584
    ret = 0;

585
 endjob:
586
    qemuDomainObjEndJob(driver, vm);
587

588
 cleanup:
589 590
    if (ret == -1)
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_FORCE));
M
Michal Privoznik 已提交
591
    virDomainObjEndAPI(&vm);
592
    qemuDomainEventQueue(driver, event);
593
    virObjectUnref(cfg);
594 595 596
}


597
void
598
qemuProcessShutdownOrReboot(virQEMUDriverPtr driver,
599
                            virDomainObjPtr vm)
600
{
601 602 603
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
604
        qemuDomainSetFakeReboot(driver, vm, false);
605
        virObjectRef(vm);
606 607 608 609 610
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
611
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
612
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
613
            virObjectUnref(vm);
614 615
        }
    } else {
616
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
617
    }
618
}
619

620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640

static int
qemuProcessHandleEvent(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                       virDomainObjPtr vm,
                       const char *eventName,
                       long long seconds,
                       unsigned int micros,
                       const char *details,
                       void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    virObjectEventPtr event = NULL;

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

    virObjectLock(vm);
    event = virDomainQemuMonitorEventNew(vm->def->id, vm->def->name,
                                         vm->def->uuid, eventName,
                                         seconds, micros, details);

    virObjectUnlock(vm);
641
    qemuDomainEventQueue(driver, event);
642 643 644 645 646

    return 0;
}


647 648
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
649 650
                          virDomainObjPtr vm,
                          void *opaque)
651
{
652
    virQEMUDriverPtr driver = opaque;
653
    qemuDomainObjPrivatePtr priv;
654
    virObjectEventPtr event = NULL;
655
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
656

657 658
    VIR_DEBUG("vm=%p", vm);

659
    virObjectLock(vm);
660 661 662 663 664 665

    priv = vm->privateData;
    if (priv->gotShutdown) {
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
666 667 668 669
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
670 671 672 673 674 675 676 677
    }
    priv->gotShutdown = true;

    VIR_DEBUG("Transitioned guest %s to shutdown state",
              vm->def->name);
    virDomainObjSetState(vm,
                         VIR_DOMAIN_SHUTDOWN,
                         VIR_DOMAIN_SHUTDOWN_UNKNOWN);
678
    event = virDomainEventLifecycleNewFromObj(vm,
679 680 681
                                     VIR_DOMAIN_EVENT_SHUTDOWN,
                                     VIR_DOMAIN_EVENT_SHUTDOWN_FINISHED);

682
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
683 684 685 686
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

687 688 689
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

690 691
    qemuProcessShutdownOrReboot(driver, vm);

692
 unlock:
693
    virObjectUnlock(vm);
694
    qemuDomainEventQueue(driver, event);
695
    virObjectUnref(cfg);
696

697 698 699 700 701 702
    return 0;
}


static int
qemuProcessHandleStop(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
703 704
                      virDomainObjPtr vm,
                      void *opaque)
705
{
706
    virQEMUDriverPtr driver = opaque;
707
    virObjectEventPtr event = NULL;
708
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
709

710
    virObjectLock(vm);
J
Jiri Denemark 已提交
711
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
712
        qemuDomainObjPrivatePtr priv = vm->privateData;
713

714
        if (priv->gotShutdown) {
715 716
            VIR_DEBUG("Ignoring STOP event after SHUTDOWN");
            goto unlock;
717 718
        }

719 720
        VIR_DEBUG("Transitioned guest %s to paused state",
                  vm->def->name);
721

722 723 724
        if (priv->job.current)
            ignore_value(virTimeMillisNow(&priv->job.current->stopped));

725
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_UNKNOWN);
726
        event = virDomainEventLifecycleNewFromObj(vm,
727 728 729
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);

730 731 732 733 734
        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));

735
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
736 737 738
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
739
    }
740

741
 unlock:
742
    virObjectUnlock(vm);
743
    qemuDomainEventQueue(driver, event);
744
    virObjectUnref(cfg);
745 746 747 748 749

    return 0;
}


750 751
static int
qemuProcessHandleResume(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
752 753
                        virDomainObjPtr vm,
                        void *opaque)
754
{
755
    virQEMUDriverPtr driver = opaque;
756
    virObjectEventPtr event = NULL;
757
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
758

759
    virObjectLock(vm);
760 761 762 763 764 765 766 767 768 769 770 771 772
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
        qemuDomainObjPrivatePtr priv = vm->privateData;

        if (priv->gotShutdown) {
            VIR_DEBUG("Ignoring RESUME event after SHUTDOWN");
            goto unlock;
        }

        VIR_DEBUG("Transitioned guest %s out of paused into resumed state",
                  vm->def->name);

        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_UNPAUSED);
773
        event = virDomainEventLifecycleNewFromObj(vm,
774 775 776 777
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

        VIR_DEBUG("Using lock state '%s' on resume event", NULLSTR(priv->lockState));
778
        if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
779 780 781 782 783 784 785 786 787
                                       vm, priv->lockState) < 0) {
            /* Don't free priv->lockState on error, because we need
             * to make sure we have state still present if the user
             * tries to resume again
             */
            goto unlock;
        }
        VIR_FREE(priv->lockState);

788
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
789 790 791 792 793
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
    }

794
 unlock:
795
    virObjectUnlock(vm);
796
    qemuDomainEventQueue(driver, event);
797
    virObjectUnref(cfg);
798 799 800
    return 0;
}

801 802 803
static int
qemuProcessHandleRTCChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm,
804 805
                           long long offset,
                           void *opaque)
806
{
807
    virQEMUDriverPtr driver = opaque;
808
    virObjectEventPtr event = NULL;
809
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
810

811
    virObjectLock(vm);
812

813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829
    if (vm->def->clock.offset == VIR_DOMAIN_CLOCK_OFFSET_VARIABLE) {
        /* when a basedate is manually given on the qemu commandline
         * rather than simply "-rtc base=utc", the offset sent by qemu
         * in this event is *not* the new offset from UTC, but is
         * instead the new offset from the *original basedate* +
         * uptime. For example, if the original offset was 3600 and
         * the guest clock has been advanced by 10 seconds, qemu will
         * send "10" in the event - this means that the new offset
         * from UTC is 3610, *not* 10. If the guest clock is advanced
         * by another 10 seconds, qemu will now send "20" - i.e. each
         * event is the sum of the most recent change and all previous
         * changes since the domain was started. Fortunately, we have
         * saved the initial offset in "adjustment0", so to arrive at
         * the proper new "adjustment", we just add the most recent
         * offset to adjustment0.
         */
        offset += vm->def->clock.data.variable.adjustment0;
830
        vm->def->clock.data.variable.adjustment = offset;
831

832 833 834 835 836
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
           VIR_WARN("unable to save domain status with RTC change");
    }

    event = virDomainEventRTCChangeNewFromObj(vm, offset);
837

838
    virObjectUnlock(vm);
839

840
    qemuDomainEventQueue(driver, event);
841
    virObjectUnref(cfg);
842 843 844 845 846 847 848
    return 0;
}


static int
qemuProcessHandleWatchdog(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
849 850
                          int action,
                          void *opaque)
851
{
852
    virQEMUDriverPtr driver = opaque;
853 854
    virObjectEventPtr watchdogEvent = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
855
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
856

857
    virObjectLock(vm);
858 859 860
    watchdogEvent = virDomainEventWatchdogNewFromObj(vm, action);

    if (action == VIR_DOMAIN_EVENT_WATCHDOG_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 watchdog", vm->def->name);

J
Jiri Denemark 已提交
865
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
866
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
867 868 869
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

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
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
876 877 878
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
879 880 881
    }

    if (vm->def->watchdog->action == VIR_DOMAIN_WATCHDOG_ACTION_DUMP) {
C
Chen Fan 已提交
882 883 884 885 886
        struct qemuProcessEvent *processEvent;
        if (VIR_ALLOC(processEvent) == 0) {
            processEvent->eventType = QEMU_PROCESS_EVENT_WATCHDOG;
            processEvent->action = VIR_DOMAIN_WATCHDOG_ACTION_DUMP;
            processEvent->vm = vm;
W
Wen Congyang 已提交
887 888 889
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
890
            virObjectRef(vm);
C
Chen Fan 已提交
891
            if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
892
                if (!virObjectUnref(vm))
893
                    vm = NULL;
C
Chen Fan 已提交
894
                VIR_FREE(processEvent);
W
Wen Congyang 已提交
895
            }
896
        }
897 898
    }

899
    if (vm)
900
        virObjectUnlock(vm);
901 902
    qemuDomainEventQueue(driver, watchdogEvent);
    qemuDomainEventQueue(driver, lifecycleEvent);
903

904
    virObjectUnref(cfg);
905 906 907 908 909 910 911 912 913
    return 0;
}


static int
qemuProcessHandleIOError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virDomainObjPtr vm,
                         const char *diskAlias,
                         int action,
914 915
                         const char *reason,
                         void *opaque)
916
{
917
    virQEMUDriverPtr driver = opaque;
918 919 920
    virObjectEventPtr ioErrorEvent = NULL;
    virObjectEventPtr ioErrorEvent2 = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
921 922 923
    const char *srcPath;
    const char *devAlias;
    virDomainDiskDefPtr disk;
924
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
925

926
    virObjectLock(vm);
927 928 929
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
930
        srcPath = virDomainDiskGetSource(disk);
931 932 933 934 935 936 937 938 939 940
        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
941
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
942
        qemuDomainObjPrivatePtr priv = vm->privateData;
943 944
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

945 946 947
        if (priv->signalIOError)
            virDomainObjBroadcast(vm);

J
Jiri Denemark 已提交
948
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
949
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
950 951 952
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

953 954 955 956 957
        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));

958
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
959 960
            VIR_WARN("Unable to save status on vm %s after IO error", vm->def->name);
    }
961
    virObjectUnlock(vm);
962

963 964 965
    qemuDomainEventQueue(driver, ioErrorEvent);
    qemuDomainEventQueue(driver, ioErrorEvent2);
    qemuDomainEventQueue(driver, lifecycleEvent);
966
    virObjectUnref(cfg);
967 968 969
    return 0;
}

970 971 972 973 974
static int
qemuProcessHandleBlockJob(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          const char *diskAlias,
                          int type,
975 976
                          int status,
                          void *opaque)
977
{
978
    virQEMUDriverPtr driver = opaque;
979
    struct qemuProcessEvent *processEvent = NULL;
980
    virDomainDiskDefPtr disk;
981
    qemuDomainDiskPrivatePtr diskPriv;
982
    char *data = NULL;
983

984
    virObjectLock(vm);
985

986 987
    VIR_DEBUG("Block job for device %s (domain: %p,%s) type %d status %d",
              diskAlias, vm, vm->def->name, type, status);
988

989
    if (!(disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias)))
990
        goto error;
991
    diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
992

993
    if (diskPriv->blockJobSync) {
994
        /* We have a SYNC API waiting for this event, dispatch it back */
995 996
        diskPriv->blockJobType = type;
        diskPriv->blockJobStatus = status;
997
        virDomainObjBroadcast(vm);
998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009
    } else {
        /* there is no waiting SYNC API, dispatch the update to a thread */
        if (VIR_ALLOC(processEvent) < 0)
            goto error;

        processEvent->eventType = QEMU_PROCESS_EVENT_BLOCK_JOB;
        if (VIR_STRDUP(data, diskAlias) < 0)
            goto error;
        processEvent->data = data;
        processEvent->vm = vm;
        processEvent->action = type;
        processEvent->status = status;
1010

1011 1012 1013 1014 1015
        virObjectRef(vm);
        if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
            ignore_value(virObjectUnref(vm));
            goto error;
        }
1016 1017
    }

1018
 cleanup:
1019
    virObjectUnlock(vm);
1020
    return 0;
1021 1022 1023 1024 1025
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
1026
}
1027

1028

1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040
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,
1041 1042
                          const char *saslUsername,
                          void *opaque)
1043
{
1044
    virQEMUDriverPtr driver = opaque;
1045
    virObjectEventPtr event;
1046 1047 1048
    virDomainEventGraphicsAddressPtr localAddr = NULL;
    virDomainEventGraphicsAddressPtr remoteAddr = NULL;
    virDomainEventGraphicsSubjectPtr subject = NULL;
1049
    size_t i;
1050 1051

    if (VIR_ALLOC(localAddr) < 0)
1052
        goto error;
1053
    localAddr->family = localFamily;
1054 1055 1056
    if (VIR_STRDUP(localAddr->service, localService) < 0 ||
        VIR_STRDUP(localAddr->node, localNode) < 0)
        goto error;
1057 1058

    if (VIR_ALLOC(remoteAddr) < 0)
1059
        goto error;
1060
    remoteAddr->family = remoteFamily;
1061 1062 1063
    if (VIR_STRDUP(remoteAddr->service, remoteService) < 0 ||
        VIR_STRDUP(remoteAddr->node, remoteNode) < 0)
        goto error;
1064 1065

    if (VIR_ALLOC(subject) < 0)
1066
        goto error;
1067 1068
    if (x509dname) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1069
            goto error;
1070
        subject->nidentity++;
1071 1072 1073
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "x509dname") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, x509dname) < 0)
            goto error;
1074 1075 1076
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1077
            goto error;
1078
        subject->nidentity++;
1079 1080 1081
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "saslUsername") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, saslUsername) < 0)
            goto error;
1082 1083
    }

1084
    virObjectLock(vm);
1085
    event = virDomainEventGraphicsNewFromObj(vm, phase, localAddr, remoteAddr, authScheme, subject);
1086
    virObjectUnlock(vm);
1087

1088
    qemuDomainEventQueue(driver, event);
1089 1090 1091

    return 0;

1092
 error:
1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103
    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) {
1104
        for (i = 0; i < subject->nidentity; i++) {
1105 1106 1107 1108 1109 1110 1111 1112 1113 1114
            VIR_FREE(subject->identities[i].type);
            VIR_FREE(subject->identities[i].name);
        }
        VIR_FREE(subject->identities);
        VIR_FREE(subject);
    }

    return -1;
}

1115 1116 1117 1118
static int
qemuProcessHandleTrayChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            const char *devAlias,
1119 1120
                            int reason,
                            void *opaque)
1121
{
1122
    virQEMUDriverPtr driver = opaque;
1123
    virObjectEventPtr event = NULL;
1124
    virDomainDiskDefPtr disk;
1125
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1126

1127
    virObjectLock(vm);
1128 1129 1130 1131 1132 1133
    disk = qemuProcessFindDomainDiskByAlias(vm, devAlias);

    if (disk) {
        event = virDomainEventTrayChangeNewFromObj(vm,
                                                   devAlias,
                                                   reason);
1134 1135 1136 1137 1138 1139
        /* 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;

1140
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1141 1142 1143
            VIR_WARN("Unable to save status on vm %s after tray moved event",
                     vm->def->name);
        }
1144 1145

        virDomainObjBroadcast(vm);
1146 1147
    }

1148
    virObjectUnlock(vm);
1149
    qemuDomainEventQueue(driver, event);
1150
    virObjectUnref(cfg);
1151 1152 1153
    return 0;
}

O
Osier Yang 已提交
1154 1155
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1156 1157
                          virDomainObjPtr vm,
                          void *opaque)
O
Osier Yang 已提交
1158
{
1159
    virQEMUDriverPtr driver = opaque;
1160 1161
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1162
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1163

1164
    virObjectLock(vm);
O
Osier Yang 已提交
1165 1166
    event = virDomainEventPMWakeupNewFromObj(vm);

1167 1168 1169 1170 1171 1172 1173 1174 1175
    /* 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);
1176
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
1177 1178 1179
                                                  VIR_DOMAIN_EVENT_STARTED,
                                                  VIR_DOMAIN_EVENT_STARTED_WAKEUP);

1180
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1181 1182 1183 1184 1185
            VIR_WARN("Unable to save status on vm %s after wakeup event",
                     vm->def->name);
        }
    }

1186
    virObjectUnlock(vm);
1187 1188
    qemuDomainEventQueue(driver, event);
    qemuDomainEventQueue(driver, lifecycleEvent);
1189
    virObjectUnref(cfg);
O
Osier Yang 已提交
1190 1191
    return 0;
}
1192

O
Osier Yang 已提交
1193 1194
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1195 1196
                           virDomainObjPtr vm,
                           void *opaque)
O
Osier Yang 已提交
1197
{
1198
    virQEMUDriverPtr driver = opaque;
1199 1200
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1201
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1202

1203
    virObjectLock(vm);
O
Osier Yang 已提交
1204 1205
    event = virDomainEventPMSuspendNewFromObj(vm);

1206
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1207
        qemuDomainObjPrivatePtr priv = vm->privateData;
1208 1209 1210 1211 1212
        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
1213
        lifecycleEvent =
1214
            virDomainEventLifecycleNewFromObj(vm,
J
Jiri Denemark 已提交
1215 1216
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1217

1218
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1219 1220 1221
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1222 1223 1224

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1225 1226
    }

1227
    virObjectUnlock(vm);
O
Osier Yang 已提交
1228

1229 1230
    qemuDomainEventQueue(driver, event);
    qemuDomainEventQueue(driver, lifecycleEvent);
1231
    virObjectUnref(cfg);
O
Osier Yang 已提交
1232 1233 1234
    return 0;
}

1235 1236 1237
static int
qemuProcessHandleBalloonChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1238 1239
                               unsigned long long actual,
                               void *opaque)
1240
{
1241
    virQEMUDriverPtr driver = opaque;
1242
    virObjectEventPtr event = NULL;
1243
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1244

1245
    virObjectLock(vm);
1246 1247 1248 1249 1250 1251
    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;

1252
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
1253 1254
        VIR_WARN("unable to save domain status with balloon change");

1255
    virObjectUnlock(vm);
1256

1257
    qemuDomainEventQueue(driver, event);
1258
    virObjectUnref(cfg);
1259 1260 1261
    return 0;
}

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

1272
    virObjectLock(vm);
1273 1274 1275 1276 1277 1278 1279 1280 1281 1282
    event = virDomainEventPMSuspendDiskNewFromObj(vm);

    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        qemuDomainObjPrivatePtr priv = vm->privateData;
        VIR_DEBUG("Transitioned guest %s to pmsuspended state due to "
                  "QMP suspend_disk event", vm->def->name);

        virDomainObjSetState(vm, VIR_DOMAIN_PMSUSPENDED,
                             VIR_DOMAIN_PMSUSPENDED_UNKNOWN);
        lifecycleEvent =
1283
            virDomainEventLifecycleNewFromObj(vm,
1284 1285 1286
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_DISK);

1287
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1288 1289 1290 1291 1292 1293 1294 1295
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }

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

1296
    virObjectUnlock(vm);
1297

1298 1299
    qemuDomainEventQueue(driver, event);
    qemuDomainEventQueue(driver, lifecycleEvent);
1300 1301
    virObjectUnref(cfg);

1302 1303 1304
    return 0;
}

1305

1306 1307
static int
qemuProcessHandleGuestPanic(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1308 1309
                            virDomainObjPtr vm,
                            void *opaque)
1310
{
1311
    virQEMUDriverPtr driver = opaque;
1312 1313 1314
    struct qemuProcessEvent *processEvent;

    virObjectLock(vm);
1315
    if (VIR_ALLOC(processEvent) < 0)
1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330
        goto cleanup;

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

1331
 cleanup:
1332
    if (vm)
1333
        virObjectUnlock(vm);
1334 1335 1336 1337 1338

    return 0;
}


1339
int
1340 1341
qemuProcessHandleDeviceDeleted(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1342 1343
                               const char *devAlias,
                               void *opaque)
1344
{
1345
    virQEMUDriverPtr driver = opaque;
1346 1347
    struct qemuProcessEvent *processEvent = NULL;
    char *data;
1348 1349 1350 1351 1352 1353

    virObjectLock(vm);

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

1354 1355
    if (qemuDomainSignalDeviceRemoval(vm, devAlias))
        goto cleanup;
1356

1357 1358
    if (VIR_ALLOC(processEvent) < 0)
        goto error;
1359

1360 1361 1362 1363 1364
    processEvent->eventType = QEMU_PROCESS_EVENT_DEVICE_DELETED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
    processEvent->vm = vm;
1365

1366 1367 1368 1369 1370
    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }
1371

1372
 cleanup:
1373 1374
    virObjectUnlock(vm);
    return 0;
1375 1376 1377 1378 1379
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
1380 1381 1382
}


1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423
static int
qemuProcessHandleNicRxFilterChanged(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virDomainObjPtr vm,
                                    const char *devAlias,
                                    void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    struct qemuProcessEvent *processEvent = NULL;
    char *data;

    virObjectLock(vm);

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

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

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

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

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


1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466
static int
qemuProcessHandleSerialChanged(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               const char *devAlias,
                               bool connected,
                               void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    struct qemuProcessEvent *processEvent = NULL;
    char *data;

    virObjectLock(vm);

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

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

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

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

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


1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485
static int
qemuProcessHandleSpiceMigrated(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               void *opaque ATTRIBUTE_UNUSED)
{
    qemuDomainObjPrivatePtr priv;

    virObjectLock(vm);

    VIR_DEBUG("Spice migration completed for domain %p %s",
              vm, vm->def->name);

    priv = vm->privateData;
    if (priv->job.asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT) {
        VIR_DEBUG("got SPICE_MIGRATE_COMPLETED event without a migration job");
        goto cleanup;
    }

    priv->job.spiceMigrated = true;
1486
    virDomainObjBroadcast(vm);
1487 1488 1489 1490 1491 1492 1493

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508
static int
qemuProcessHandleMigrationStatus(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                 virDomainObjPtr vm,
                                 int status,
                                 void *opaque ATTRIBUTE_UNUSED)
{
    qemuDomainObjPrivatePtr priv;

    virObjectLock(vm);

    VIR_DEBUG("Migration of domain %p %s changed state to %s",
              vm, vm->def->name,
              qemuMonitorMigrationStatusTypeToString(status));

    priv = vm->privateData;
1509
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_NONE) {
1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522
        VIR_DEBUG("got MIGRATION event without a migration job");
        goto cleanup;
    }

    priv->job.current->status.status = status;
    virDomainObjBroadcast(vm);

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1523 1524
static qemuMonitorCallbacks monitorCallbacks = {
    .eofNotify = qemuProcessHandleMonitorEOF,
1525
    .errorNotify = qemuProcessHandleMonitorError,
1526
    .diskSecretLookup = qemuProcessFindVolumeQcowPassphrase,
1527
    .domainEvent = qemuProcessHandleEvent,
1528 1529
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
1530
    .domainResume = qemuProcessHandleResume,
1531 1532 1533 1534 1535
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1536
    .domainBlockJob = qemuProcessHandleBlockJob,
1537
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1538
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1539
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1540
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1541
    .domainPMSuspendDisk = qemuProcessHandlePMSuspendDisk,
1542
    .domainGuestPanic = qemuProcessHandleGuestPanic,
1543
    .domainDeviceDeleted = qemuProcessHandleDeviceDeleted,
1544
    .domainNicRxFilterChanged = qemuProcessHandleNicRxFilterChanged,
1545
    .domainSerialChange = qemuProcessHandleSerialChanged,
1546
    .domainSpiceMigrated = qemuProcessHandleSpiceMigrated,
1547
    .domainMigrationStatus = qemuProcessHandleMigrationStatus,
1548 1549 1550
};

static int
1551
qemuConnectMonitor(virQEMUDriverPtr driver, virDomainObjPtr vm, int asyncJob,
1552
                   int logfd, off_t pos)
1553 1554 1555
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
1556
    qemuMonitorPtr mon = NULL;
1557

1558
    if (virSecurityManagerSetDaemonSocketLabel(driver->securityManager,
1559
                                               vm->def) < 0) {
1560 1561
        VIR_ERROR(_("Failed to set security context for monitor for %s"),
                  vm->def->name);
1562
        return -1;
1563 1564 1565
    }

    /* Hold an extra reference because we can't allow 'vm' to be
M
Michal Privoznik 已提交
1566
     * deleted unitl the monitor gets its own reference. */
1567
    virObjectRef(vm);
1568

1569
    ignore_value(virTimeMillisNow(&priv->monStart));
1570
    virObjectUnlock(vm);
1571 1572 1573 1574

    mon = qemuMonitorOpen(vm,
                          priv->monConfig,
                          priv->monJSON,
1575 1576
                          &monitorCallbacks,
                          driver);
1577

1578 1579
    if (mon && logfd != -1 && pos != -1)
        ignore_value(qemuMonitorSetDomainLog(mon, logfd, pos));
1580

1581
    virObjectLock(vm);
M
Michal Privoznik 已提交
1582
    virObjectUnref(vm);
1583
    priv->monStart = 0;
1584

M
Michal Privoznik 已提交
1585
    if (!virDomainObjIsActive(vm)) {
1586
        qemuMonitorClose(mon);
1587
        mon = NULL;
1588 1589 1590
    }
    priv->mon = mon;

1591
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0) {
1592 1593
        VIR_ERROR(_("Failed to clear security context for monitor for %s"),
                  vm->def->name);
1594
        return -1;
1595 1596 1597 1598
    }

    if (priv->mon == NULL) {
        VIR_INFO("Failed to connect monitor for %s", vm->def->name);
1599
        return -1;
1600 1601 1602
    }


1603
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
1604
        return -1;
1605

1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619
    if (qemuMonitorSetCapabilities(priv->mon) < 0)
        goto cleanup;

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON) &&
        virQEMUCapsProbeQMP(priv->qemuCaps, priv->mon) < 0)
        goto cleanup;

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT) &&
        qemuMonitorSetMigrationCapability(priv->mon,
                                          QEMU_MONITOR_MIGRATION_CAPS_EVENTS,
                                          true) < 0) {
        VIR_DEBUG("Cannot enable migration events; clearing capability");
        virQEMUCapsClear(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
    }
1620

1621 1622 1623 1624 1625
    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
1626 1627 1628
    return ret;
}

1629 1630 1631 1632 1633

/**
 * qemuProcessReadLog: Read log file of a qemu VM
 * @fd: File descriptor of the log file
 * @off: Offset to start reading from
1634
 * @msg: pointer to buffer to store the read messages in
1635 1636
 *
 * Reads log of a qemu VM. Skips messages not produced by qemu or irrelevant
1637
 * messages. Returns returns 0 on success or -1 on error
1638
 */
1639 1640
static int
qemuProcessReadLog(int fd, off_t offset, char **msg)
1641
{
1642 1643 1644
    char *buf;
    size_t buflen = 1024 * 128;
    ssize_t got;
1645
    char *eol;
1646
    char *filter_next;
1647

1648 1649
    /* Best effort jump to start of messages */
    ignore_value(lseek(fd, offset, SEEK_SET));
1650

1651 1652
    if (VIR_ALLOC_N(buf, buflen) < 0)
        return -1;
1653

1654 1655 1656 1657 1658 1659
    got = saferead(fd, buf, buflen - 1);
    if (got <= 0) {
        VIR_FREE(buf);
        virReportSystemError(errno, "%s",
                             _("Unable to read from log file"));
        return -1;
1660 1661
    }

1662
    buf[got] = '\0';
1663

1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675
    /* Filter out debug messages from intermediate libvirt process */
    filter_next = buf;
    while ((eol = strchr(filter_next, '\n'))) {
        *eol = '\0';
        if (virLogProbablyLogMessage(filter_next) ||
            STRPREFIX(filter_next, "char device redirected to")) {
            size_t skip = (eol + 1) - filter_next;
            memmove(filter_next, eol + 1, (got - skip) + 1);
            got -= skip;
        } else {
            filter_next = eol + 1;
            *eol = '\n';
1676 1677
        }
    }
1678
    filter_next = NULL; /* silence false coverity warning */
1679

1680 1681 1682
    if (buf[got - 1] == '\n') {
        buf[got - 1] = '\0';
        got--;
1683
    }
1684 1685 1686 1687
    VIR_SHRINK_N(buf, buflen, buflen - got - 1);
    *msg = buf;
    return 0;
}
1688 1689


1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704
int
qemuProcessReportLogError(int logfd,
                          off_t offset,
                          const char *msgprefix)
{
    char *logmsg = NULL;

    if (qemuProcessReadLog(logfd, offset, &logmsg) < 0)
        return -1;

    virResetLastError();
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("%s: %s"), msgprefix, logmsg);
    VIR_FREE(logmsg);
    return 0;
1705 1706 1707
}


1708
static int
1709 1710 1711
qemuProcessLookupPTYs(virDomainDefPtr def,
                      virQEMUCapsPtr qemuCaps,
                      virDomainChrDefPtr *devices,
1712
                      int count,
1713
                      virHashTablePtr info)
1714
{
1715
    size_t i;
1716

1717
    for (i = 0; i < count; i++) {
1718
        virDomainChrDefPtr chr = devices[i];
1719 1720
        bool chardevfmt = virQEMUCapsSupportsChardev(def, qemuCaps, chr);

1721
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
C
Cole Robinson 已提交
1722
            char id[32];
1723
            qemuMonitorChardevInfoPtr entry;
1724

C
Cole Robinson 已提交
1725
            if (snprintf(id, sizeof(id), "%s%s",
1726
                         chardevfmt ? "char" : "",
1727 1728 1729 1730
                         chr->info.alias) >= sizeof(id)) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("failed to format device alias "
                                 "for PTY retrieval"));
1731
                return -1;
1732
            }
1733

1734 1735
            entry = virHashLookup(info, id);
            if (!entry || !entry->ptyPath) {
1736 1737 1738 1739
                if (chr->source.data.file.path == NULL) {
                    /* neither the log output nor 'info chardev' had a
                     * pty path for this chardev, report an error
                     */
1740 1741
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1742 1743 1744 1745 1746 1747 1748 1749 1750 1751
                    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);
1752
            if (VIR_STRDUP(chr->source.data.file.path, entry->ptyPath) < 0)
1753 1754 1755 1756 1757 1758 1759
                return -1;
        }
    }

    return 0;
}

1760 1761
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
1762
                                     virQEMUCapsPtr qemuCaps,
1763
                                     virHashTablePtr info)
1764
{
1765
    size_t i = 0;
C
Cole Robinson 已提交
1766

1767 1768
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->serials, vm->def->nserials,
1769
                              info) < 0)
1770 1771
        return -1;

1772 1773
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->parallels, vm->def->nparallels,
1774
                              info) < 0)
1775
        return -1;
1776

1777 1778
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->channels, vm->def->nchannels,
1779
                              info) < 0)
1780
        return -1;
1781 1782 1783 1784
    /* For historical reasons, console[0] can be just an alias
     * for serial[0]. That's why we need to update it as well. */
    if (vm->def->nconsoles) {
        virDomainChrDefPtr chr = vm->def->consoles[0];
1785

1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796
        if (vm->def->nserials &&
            chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_CONSOLE &&
            chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_SERIAL) {
            /* yes, the first console is just an alias for serials[0] */
            i = 1;
            if (virDomainChrSourceDefCopy(&chr->source,
                                          &((vm->def->serials[0])->source)) < 0)
                return -1;
        }
    }

1797 1798
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->consoles + i, vm->def->nconsoles - i,
1799
                              info) < 0)
1800
        return -1;
1801 1802 1803 1804 1805

    return 0;
}


1806
static int
1807 1808 1809 1810
qemuProcessRefreshChannelVirtioState(virQEMUDriverPtr driver,
                                     virDomainObjPtr vm,
                                     virHashTablePtr info,
                                     int booted)
1811 1812
{
    size_t i;
1813
    int agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_CHANNEL;
1814
    qemuMonitorChardevInfoPtr entry;
1815
    virObjectEventPtr event = NULL;
1816 1817
    char id[32];

1818 1819 1820
    if (booted)
        agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_DOMAIN_STARTED;

1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836
    for (i = 0; i < vm->def->nchannels; i++) {
        virDomainChrDefPtr chr = vm->def->channels[i];
        if (chr->targetType == VIR_DOMAIN_CHR_CHANNEL_TARGET_TYPE_VIRTIO) {
            if (snprintf(id, sizeof(id), "char%s",
                         chr->info.alias) >= sizeof(id)) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("failed to format device alias "
                                 "for PTY retrieval"));
                return -1;
            }

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

1837 1838 1839 1840 1841 1842
            if (entry->state != VIR_DOMAIN_CHR_DEVICE_STATE_DEFAULT &&
                STREQ_NULLABLE(chr->target.name, "org.qemu.guest_agent.0") &&
                (event = virDomainEventAgentLifecycleNewFromObj(vm, entry->state,
                                                                agentReason)))
                qemuDomainEventQueue(driver, event);

1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860
            chr->state = entry->state;
        }
    }

    return 0;
}


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

    qemuDomainObjEnterMonitor(driver, vm);
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
1861 1862
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
1863 1864 1865 1866

    if (ret < 0)
        goto cleanup;

1867
    ret = qemuProcessRefreshChannelVirtioState(driver, vm, info, false);
1868 1869 1870 1871 1872 1873 1874

 cleanup:
    virHashFree(info);
    return ret;
}


1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906
static int
qemuProcessRefreshBalloonState(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
                               int asyncJob)
{
    unsigned long long balloon;
    int rc;

    /* if no ballooning is available, the current size equals to the current
     * full memory size */
    if (!vm->def->memballoon ||
        vm->def->memballoon->model == VIR_DOMAIN_MEMBALLOON_MODEL_NONE) {
        vm->def->mem.cur_balloon = virDomainDefGetMemoryActual(vm->def);
        return 0;
    }

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

    rc = qemuMonitorGetBalloonInfo(qemuDomainGetMonitor(vm), &balloon);
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        rc = -1;

    if (rc < 0)
        return -1;

    vm->def->mem.cur_balloon = balloon;

    return 0;
}


1907
static int
1908
qemuProcessWaitForMonitor(virQEMUDriverPtr driver,
C
Cole Robinson 已提交
1909
                          virDomainObjPtr vm,
1910
                          int asyncJob,
1911
                          virQEMUCapsPtr qemuCaps,
1912
                          qemuDomainLogContextPtr logCtxt)
1913 1914
{
    int ret = -1;
1915
    virHashTablePtr info = NULL;
1916
    qemuDomainObjPrivatePtr priv;
1917
    int logfd = -1;
1918
    off_t pos = -1;
1919

1920 1921 1922 1923
    if (logCtxt) {
        logfd = qemuDomainLogContextGetReadFD(logCtxt);
        pos = qemuDomainLogContextGetPosition(logCtxt);
    }
1924 1925

    VIR_DEBUG("Connect monitor to %p '%s'", vm, vm->def->name);
1926
    if (qemuConnectMonitor(driver, vm, asyncJob, logfd, pos) < 0)
1927 1928 1929 1930 1931 1932
        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. */
1933
    priv = vm->privateData;
1934 1935
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
1936 1937
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
    VIR_DEBUG("qemuMonitorGetChardevInfo returned %i", ret);
1938 1939 1940
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

1941 1942 1943 1944 1945
    if (ret == 0) {
        if ((ret = qemuProcessFindCharDevicePTYsMonitor(vm, qemuCaps,
                                                        info)) < 0)
            goto cleanup;

1946 1947
        if ((ret = qemuProcessRefreshChannelVirtioState(driver, vm, info,
                                                        true)) < 0)
1948 1949
            goto cleanup;
    }
1950

1951
 cleanup:
1952
    virHashFree(info);
1953

1954 1955 1956
    if (pos != (off_t)-1 && kill(vm->pid, 0) == -1 && errno == ESRCH) {
        qemuProcessReportLogError(logfd, pos,
                                  _("process exited while connecting to monitor"));
1957 1958 1959 1960 1961 1962 1963
        ret = -1;
    }

    return ret;
}

static int
1964
qemuProcessDetectVcpuPIDs(virQEMUDriverPtr driver,
1965
                          virDomainObjPtr vm, int asyncJob)
1966 1967 1968 1969 1970
{
    pid_t *cpupids = NULL;
    int ncpupids;
    qemuDomainObjPrivatePtr priv = vm->privateData;

1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
    /*
     * Current QEMU *can* report info about host threads mapped
     * to vCPUs, but it is not in a manner we can correctly
     * deal with. The TCG CPU emulation does have a separate vCPU
     * thread, but it runs every vCPU in that same thread. So it
     * is impossible to setup different affinity per thread.
     *
     * What's more the 'query-cpus' command returns bizarre
     * data for the threads. It gives the TCG thread for the
     * vCPU 0, but for vCPUs 1-> N, it actually replies with
     * the main process thread ID.
     *
     * The result is that when we try to set affinity for
     * vCPU 1, it will actually change the affinity of the
     * emulator thread :-( When you try to set affinity for
     * vCPUs 2, 3.... it will fail if the affinity was
     * different from vCPU 1.
     *
     * We *could* allow vcpu pinning with TCG, if we made the
     * restriction that all vCPUs had the same mask. This would
     * at least let us separate emulator from vCPUs threads, as
     * we do for KVM. It would need some changes to our cgroups
     * CPU layout though, and error reporting for the config
     * restrictions.
     *
     * Just disable CPU pinning with TCG until someone wants
     * to try to do this hard work.
     */
    if (vm->def->virtType == VIR_DOMAIN_VIRT_QEMU) {
        priv->nvcpupids = 0;
        priv->vcpupids = NULL;
        return 0;
    }

2005 2006
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2007 2008 2009
    ncpupids = qemuMonitorGetCPUInfo(priv->mon, &cpupids);
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2010 2011 2012
    /* 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 */
2013
    if (ncpupids <= 0) {
2014 2015
        virResetLastError();

2016 2017
        priv->nvcpupids = 0;
        priv->vcpupids = NULL;
2018 2019 2020 2021
        return 0;
    }

    if (ncpupids != vm->def->vcpus) {
2022 2023 2024 2025
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of vCPU pids from QEMU monitor. "
                         "got %d, wanted %d"),
                       ncpupids, vm->def->vcpus);
2026 2027 2028 2029 2030 2031 2032 2033 2034
        VIR_FREE(cpupids);
        return -1;
    }

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

2035

2036 2037 2038 2039 2040 2041
static int
qemuProcessDetectIOThreadPIDs(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2042
    qemuMonitorIOThreadInfoPtr *iothreads = NULL;
2043 2044 2045 2046
    int niothreads = 0;
    int ret = -1;
    size_t i;

2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_OBJECT_IOTHREAD)) {
        /* The following check is because at one time a domain could
         * define iothreadids and start the domain - only failing the
         * capability check when attempting to add a disk. Because the
         * iothreads and [n]iothreadids were left untouched other code
         * assumed it could use the ->thread_id value to make thread_id
         * based adjustments (e.g. pinning, scheduling) which while
         * succeeding would execute on the calling thread.
         */
        if (vm->def->niothreadids) {
            for (i = 0; i < vm->def->niothreadids; i++) {
                /* Check if the domain had defined any iothreadid elements
                 * and supply a VIR_INFO indicating that it's being removed.
                 */
                if (!vm->def->iothreadids[i]->autofill)
                    VIR_INFO("IOThreads not supported, remove iothread id '%u'",
                             vm->def->iothreadids[i]->iothread_id);
                virDomainIOThreadIDDefFree(vm->def->iothreadids[i]);
            }
            /* Remove any trace */
            VIR_FREE(vm->def->iothreadids);
            vm->def->niothreadids = 0;
            vm->def->iothreads = 0;
        }
2071
        return 0;
2072
    }
2073

2074 2075 2076 2077
    /* Get the list of IOThreads from qemu */
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
    niothreads = qemuMonitorGetIOThreads(priv->mon, &iothreads);
2078 2079
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
J
John Ferlan 已提交
2080
    if (niothreads < 0)
2081 2082
        goto cleanup;

2083
    if (niothreads != vm->def->niothreadids) {
2084 2085
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of IOThread pids from QEMU monitor. "
2086 2087
                         "got %d, wanted %zu"),
                       niothreads, vm->def->niothreadids);
2088 2089 2090
        goto cleanup;
    }

2091 2092 2093 2094 2095 2096
    /* Nothing to do */
    if (niothreads == 0) {
        ret = 0;
        goto cleanup;
    }

2097 2098 2099
    for (i = 0; i < niothreads; i++) {
        virDomainIOThreadIDDefPtr iothrid;

2100 2101
        if (!(iothrid = virDomainIOThreadIDFind(vm->def,
                                                iothreads[i]->iothread_id))) {
2102
            virReportError(VIR_ERR_INTERNAL_ERROR,
2103 2104
                           _("iothread %d not found"),
                           iothreads[i]->iothread_id);
2105 2106 2107 2108
            goto cleanup;
        }
        iothrid->thread_id = iothreads[i]->thread_id;
    }
2109 2110 2111 2112 2113 2114

    ret = 0;

 cleanup:
    if (iothreads) {
        for (i = 0; i < niothreads; i++)
2115
            VIR_FREE(iothreads[i]);
2116 2117 2118 2119 2120
        VIR_FREE(iothreads);
    }
    return ret;
}

2121 2122 2123 2124 2125

/*
 * To be run between fork/exec of QEMU only
 */
static int
2126
qemuProcessInitCpuAffinity(virDomainObjPtr vm)
2127 2128 2129 2130
{
    int ret = -1;
    virBitmapPtr cpumap = NULL;
    virBitmapPtr cpumapToSet = NULL;
2131
    qemuDomainObjPrivatePtr priv = vm->privateData;
2132

2133 2134 2135 2136 2137 2138
    if (!vm->pid) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Cannot setup CPU affinity until process is started"));
        return -1;
    }

2139 2140
    if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) {
        VIR_DEBUG("Set CPU affinity with advisory nodeset from numad");
2141
        cpumapToSet = priv->autoCpuset;
2142
    } else {
2143
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
2144
        if (vm->def->cpumask) {
H
Hu Tao 已提交
2145
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
2146 2147 2148 2149 2150
        } 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 */
2151 2152 2153 2154
            int hostcpus;

            /* setaffinity fails if you set bits for CPUs which
             * aren't present, so we have to limit ourselves */
2155
            if ((hostcpus = nodeGetCPUCount(NULL)) < 0)
2156 2157 2158 2159 2160 2161 2162 2163
                goto cleanup;

            if (hostcpus > QEMUD_CPUMASK_LEN)
                hostcpus = QEMUD_CPUMASK_LEN;

            if (!(cpumap = virBitmapNew(hostcpus)))
                goto cleanup;

2164
            virBitmapSetAll(cpumap);
2165 2166

            cpumapToSet = cpumap;
O
Osier Yang 已提交
2167
        }
2168 2169
    }

2170
    if (virProcessSetAffinity(vm->pid, cpumapToSet) < 0)
2171
        goto cleanup;
2172

2173 2174
    ret = 0;

2175
 cleanup:
2176
    virBitmapFree(cpumap);
2177
    return ret;
2178 2179
}

2180 2181
/* set link states to down on interfaces at qemu start */
static int
2182 2183 2184
qemuProcessSetLinkStates(virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob)
2185 2186 2187
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2188
    size_t i;
2189 2190 2191 2192 2193
    int ret = -1;
    int rv;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2194 2195 2196

    for (i = 0; i < def->nnets; i++) {
        if (def->nets[i]->linkstate == VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN) {
2197 2198 2199
            if (!def->nets[i]->info.alias) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("missing alias for network device"));
2200
                goto cleanup;
2201 2202
            }

2203 2204
            VIR_DEBUG("Setting link state: %s", def->nets[i]->info.alias);

2205
            if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NETDEV)) {
2206
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
2207
                               _("Setting of link state is not supported by this qemu"));
2208
                goto cleanup;
2209 2210
            }

2211 2212 2213 2214
            rv = qemuMonitorSetLink(priv->mon,
                                    def->nets[i]->info.alias,
                                    VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (rv < 0) {
2215
                virReportError(VIR_ERR_OPERATION_FAILED,
2216 2217 2218
                               _("Couldn't set link state on interface: %s"),
                               def->nets[i]->info.alias);
                goto cleanup;
2219 2220 2221 2222
            }
        }
    }

2223 2224 2225 2226 2227
    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2228 2229 2230 2231
    return ret;
}

/* Set CPU affinities for vcpus if vcpupin xml provided. */
2232
static int
2233
qemuProcessSetVcpuAffinities(virDomainObjPtr vm)
2234 2235 2236
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2237
    virDomainPinDefPtr pininfo;
2238
    int n;
E
Eric Blake 已提交
2239
    int ret = -1;
2240 2241
    VIR_DEBUG("Setting affinity on CPUs nvcpupin=%zu nvcpus=%d nvcpupids=%d",
              def->cputune.nvcpupin, def->vcpus, priv->nvcpupids);
2242 2243 2244 2245
    if (!def->cputune.nvcpupin)
        return 0;

    if (priv->vcpupids == NULL) {
2246 2247 2248
        /* If any CPU has custom affinity that differs from the
         * VM default affinity, we must reject it
         */
2249
        for (n = 0; n < def->cputune.nvcpupin; n++) {
2250 2251 2252 2253 2254 2255 2256 2257
            if (!virBitmapEqual(def->cpumask,
                                def->cputune.vcpupin[n]->cpumask)) {
                virReportError(VIR_ERR_OPERATION_INVALID,
                               "%s", _("cpu affinity is not supported"));
                return -1;
            }
        }
        return 0;
2258 2259
    }

2260 2261
    for (n = 0; n < def->vcpus; n++) {
        /* set affinity only for existing vcpus */
2262 2263 2264
        if (!(pininfo = virDomainPinFind(def->cputune.vcpupin,
                                         def->cputune.nvcpupin,
                                         n)))
2265
            continue;
2266

2267 2268
        if (virProcessSetAffinity(priv->vcpupids[n],
                                  pininfo->cpumask) < 0) {
E
Eric Blake 已提交
2269
            goto cleanup;
2270 2271 2272
        }
    }

E
Eric Blake 已提交
2273
    ret = 0;
2274
 cleanup:
E
Eric Blake 已提交
2275
    return ret;
2276
}
2277

2278
/* Set CPU affinities for emulator threads. */
2279
static int
2280
qemuProcessSetEmulatorAffinity(virDomainObjPtr vm)
2281
{
2282
    virBitmapPtr cpumask;
2283 2284 2285
    virDomainDefPtr def = vm->def;
    int ret = -1;

O
Osier Yang 已提交
2286
    if (def->cputune.emulatorpin)
2287
        cpumask = def->cputune.emulatorpin;
O
Osier Yang 已提交
2288
    else if (def->cpumask)
2289
        cpumask = def->cpumask;
O
Osier Yang 已提交
2290 2291
    else
        return 0;
2292

2293
    ret = virProcessSetAffinity(vm->pid, cpumask);
2294 2295 2296
    return ret;
}

2297 2298 2299 2300 2301 2302 2303 2304
/* Set CPU affinities for IOThreads threads. */
static int
qemuProcessSetIOThreadsAffinity(virDomainObjPtr vm)
{
    virDomainDefPtr def = vm->def;
    size_t i;
    int ret = -1;

2305 2306
    for (i = 0; i < def->niothreadids; i++) {
        /* set affinity only for existing iothreads */
2307
        if (!def->iothreadids[i]->cpumask)
2308 2309
            continue;

2310
        if (virProcessSetAffinity(def->iothreadids[i]->thread_id,
2311
                                  def->iothreadids[i]->cpumask) < 0)
2312 2313 2314 2315 2316 2317 2318 2319
            goto cleanup;
    }
    ret = 0;

 cleanup:
    return ret;
}

2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344
/* Set Scheduler parameters for vCPU or I/O threads. */
int
qemuProcessSetSchedParams(int id,
                          pid_t pid,
                          size_t nsp,
                          virDomainThreadSchedParamPtr sp)
{
    bool val = false;
    size_t i = 0;
    virDomainThreadSchedParamPtr s = NULL;

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

    if (!s)
        return 0;

2345
    return virProcessSetScheduler(pid, s->policy, s->priority);
2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360
}

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

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

2361 2362 2363
    for (i = 0; i < vm->def->niothreadids; i++) {
        if (qemuProcessSetSchedParams(vm->def->iothreadids[i]->iothread_id,
                                      vm->def->iothreadids[i]->thread_id,
2364 2365 2366 2367 2368 2369 2370 2371
                                      vm->def->cputune.niothreadsched,
                                      vm->def->cputune.iothreadsched) < 0)
            return -1;
    }

    return 0;
}

2372 2373
static int
qemuProcessInitPasswords(virConnectPtr conn,
2374
                         virQEMUDriverPtr driver,
2375 2376
                         virDomainObjPtr vm,
                         int asyncJob)
2377 2378 2379
{
    int ret = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
2380
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2381
    size_t i;
2382 2383
    char *alias = NULL;
    char *secret = NULL;
2384

2385
    for (i = 0; i < vm->def->ngraphics; ++i) {
2386 2387
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
2388 2389
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_VNC,
2390
                                                    &graphics->data.vnc.auth,
2391 2392
                                                    cfg->vncPassword,
                                                    asyncJob);
2393
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
2394 2395
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_SPICE,
2396
                                                    &graphics->data.spice.auth,
2397 2398
                                                    cfg->spicePassword,
                                                    asyncJob);
2399 2400
        }

2401 2402 2403
        if (ret < 0)
            goto cleanup;
    }
2404

2405
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
2406
        for (i = 0; i < vm->def->ndisks; i++) {
2407 2408
            size_t secretLen;

2409
            if (!vm->def->disks[i]->src->encryption ||
2410
                !virDomainDiskGetSource(vm->def->disks[i]))
2411 2412
                continue;

2413
            VIR_FREE(secret);
2414 2415 2416 2417 2418
            if (qemuProcessGetVolumeQcowPassphrase(conn,
                                                   vm->def->disks[i],
                                                   &secret, &secretLen) < 0)
                goto cleanup;

2419 2420 2421 2422
            VIR_FREE(alias);
            if (VIR_STRDUP(alias, vm->def->disks[i]->info.alias) < 0)
                goto cleanup;
            if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
2423
                goto cleanup;
E
Eric Blake 已提交
2424
            ret = qemuMonitorSetDrivePassphrase(priv->mon, alias, secret);
2425 2426
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                ret = -1;
2427 2428 2429 2430 2431
            if (ret < 0)
                goto cleanup;
        }
    }

2432
 cleanup:
2433 2434
    VIR_FREE(alias);
    VIR_FREE(secret);
2435
    virObjectUnref(cfg);
2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478
    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)
{
2479
    bool found = false;
2480
    size_t i;
2481 2482 2483

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

2484
    for (i = 0; i < naddrs; i++) {
2485 2486 2487
        VIR_DEBUG("Maybe %x:%x", addrs[i].vendor, addrs[i].product);
        if (addrs[i].vendor == vendor &&
            addrs[i].product == product) {
2488
            VIR_DEBUG("Match %zu", i);
2489
            found = true;
2490 2491 2492
            break;
        }
    }
2493
    if (!found)
2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693
        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;
2694
    size_t i;
2695 2696 2697 2698 2699

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

2700
    for (i = 0; i < vm->def->ndisks; i++) {
2701 2702 2703 2704 2705 2706
        if (qemuProcessGetPCIDiskVendorProduct(vm->def->disks[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->disks[i]->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2707 2708 2709
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for VirtIO disk %s"),
                           vm->def->disks[i]->dst);
2710 2711 2712 2713
            return -1;
        }
    }

2714
    for (i = 0; i < vm->def->nnets; i++) {
2715 2716 2717 2718 2719 2720
        if (qemuProcessGetPCINetVendorProduct(vm->def->nets[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->nets[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2721 2722 2723
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for %s NIC"),
                           vm->def->nets[i]->model);
2724 2725 2726 2727
            return -1;
        }
    }

2728
    for (i = 0; i < vm->def->ncontrollers; i++) {
2729 2730 2731 2732 2733 2734
        if (qemuProcessGetPCIControllerVendorProduct(vm->def->controllers[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->controllers[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2735 2736 2737
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for controller %s"),
                           virDomainControllerTypeToString(vm->def->controllers[i]->type));
2738 2739 2740 2741
            return -1;
        }
    }

2742
    for (i = 0; i < vm->def->nvideos; i++) {
2743 2744 2745 2746 2747 2748
        if (qemuProcessGetPCIVideoVendorProduct(vm->def->videos[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->videos[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2749 2750 2751
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for video adapter %s"),
                           virDomainVideoTypeToString(vm->def->videos[i]->type));
2752 2753 2754 2755
            return -1;
        }
    }

2756
    for (i = 0; i < vm->def->nsounds; i++) {
2757 2758 2759 2760 2761 2762
        if (qemuProcessGetPCISoundVendorProduct(vm->def->sounds[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->sounds[i]->info),
                                    vendor, product,
                                     addrs,  naddrs) < 0) {
2763 2764 2765
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for sound adapter %s"),
                           virDomainSoundModelTypeToString(vm->def->sounds[i]->model));
2766 2767 2768 2769 2770 2771 2772 2773 2774 2775
            return -1;
        }
    }


    if (vm->def->watchdog &&
        qemuProcessGetPCIWatchdogVendorProduct(vm->def->watchdog, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->watchdog->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2776 2777 2778
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for watchdog %s"),
                           virDomainWatchdogModelTypeToString(vm->def->watchdog->model));
2779 2780 2781 2782 2783 2784 2785 2786 2787
            return -1;
        }
    }

    if (vm->def->memballoon &&
        qemuProcessGetPCIMemballoonVendorProduct(vm->def->memballoon, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->memballoon->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2788 2789 2790
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for balloon %s"),
                           virDomainMemballoonModelTypeToString(vm->def->memballoon->model));
2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807
            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
2808
qemuProcessInitPCIAddresses(virQEMUDriverPtr driver,
2809 2810
                            virDomainObjPtr vm,
                            int asyncJob)
2811 2812 2813
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int naddrs;
2814
    int ret = -1;
2815 2816
    qemuMonitorPCIAddress *addrs = NULL;

2817 2818
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2819 2820
    naddrs = qemuMonitorGetAllPCIAddresses(priv->mon,
                                           &addrs);
2821 2822
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
2823

2824 2825
    if (naddrs > 0)
        ret = qemuProcessDetectPCIAddresses(vm, addrs, naddrs);
2826

2827
 cleanup:
2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856
    VIR_FREE(addrs);

    return ret;
}


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

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

    VIR_FORCE_CLOSE(fd);

    return 0;
}


2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869
static int
qemuProcessCleanupChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
    if (dev->source.type == VIR_DOMAIN_CHR_TYPE_UNIX &&
        dev->source.data.nix.listen)
        unlink(dev->source.data.nix.path);

    return 0;
}


2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934
/**
 * Loads and update video memory size for video devices according to QEMU
 * process as the QEMU will silently update the values that we pass to QEMU
 * through command line.  We need to load these updated values and store them
 * into the status XML.
 *
 * We will fail if for some reason the values cannot be loaded from QEMU because
 * its mandatory to get the correct video memory size to status XML to not break
 * migration.
 */
static int
qemuProcessUpdateVideoRamSize(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    int ret = -1;
    ssize_t i;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainVideoDefPtr video = NULL;
    virQEMUDriverConfigPtr cfg = NULL;

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

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

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

    }

2935 2936
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2937 2938 2939 2940 2941 2942 2943 2944

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

    return ret;

 error:
2945
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
2946 2947 2948 2949
    return -1;
}


2950 2951 2952
struct qemuProcessHookData {
    virConnectPtr conn;
    virDomainObjPtr vm;
2953
    virQEMUDriverPtr driver;
2954
    virQEMUDriverConfigPtr cfg;
2955 2956 2957 2958 2959
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
2960
    qemuDomainObjPrivatePtr priv = h->vm->privateData;
2961
    int ret = -1;
2962
    int fd;
2963 2964 2965
    virBitmapPtr nodeset = NULL;
    virDomainNumatuneMemMode mode;

2966 2967 2968 2969 2970
    /* This method cannot use any mutexes, which are not
     * protected across fork()
     */

    virSecurityManagerPostFork(h->driver->securityManager);
2971 2972 2973 2974 2975

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

    VIR_DEBUG("Obtaining domain lock");
2976 2977 2978 2979 2980 2981 2982
    /*
     * 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.
     */
2983
    if (virSecurityManagerSetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2984
        goto cleanup;
2985
    if (virDomainLockProcessStart(h->driver->lockManager,
2986
                                  h->cfg->uri,
2987
                                  h->vm,
J
Ján Tomko 已提交
2988
                                  /* QEMU is always paused initially */
2989 2990
                                  true,
                                  &fd) < 0)
2991
        goto cleanup;
2992
    if (virSecurityManagerClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
2993
        goto cleanup;
2994

2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005
    if (virDomainNumatuneGetMode(h->vm->def->numa, -1, &mode) == 0) {
        if (mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT &&
            h->cfg->cgroupControllers & (1 << VIR_CGROUP_CONTROLLER_CPUSET) &&
            virCgroupControllerAvailable(VIR_CGROUP_CONTROLLER_CPUSET)) {
            /* Use virNuma* API iff necessary. Once set and child is exec()-ed,
             * there's no way for us to change it. Rely on cgroups (if available
             * and enabled in the config) rather than virNuma*. */
            VIR_DEBUG("Relying on CGroups for memory binding");
        } else {
            nodeset = virDomainNumatuneGetNodeset(h->vm->def->numa,
                                                  priv->autoNodeset, -1);
3006

3007 3008 3009
            if (virNumaSetupMemoryPolicy(mode, nodeset) < 0)
                goto cleanup;
        }
3010
    }
3011

3012 3013
    ret = 0;

3014
 cleanup:
3015
    virObjectUnref(h->cfg);
3016 3017
    VIR_DEBUG("Hook complete ret=%d", ret);
    return ret;
3018 3019 3020
}

int
3021
qemuProcessPrepareMonitorChr(virQEMUDriverConfigPtr cfg,
3022 3023 3024 3025 3026 3027
                             virDomainChrSourceDefPtr monConfig,
                             const char *vm)
{
    monConfig->type = VIR_DOMAIN_CHR_TYPE_UNIX;
    monConfig->data.nix.listen = true;

3028
    if (virAsprintf(&monConfig->data.nix.path, "%s/domain-%s/monitor.sock",
3029 3030 3031
                    cfg->libDir, vm) < 0)
        return -1;
    return 0;
3032 3033 3034
}


3035
/*
3036 3037
 * Precondition: vm must be locked, and a job must be active.
 * This method will call {Enter,Exit}Monitor
3038
 */
E
Eric Blake 已提交
3039
int
3040
qemuProcessStartCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
3041
                     virConnectPtr conn, virDomainRunningReason reason,
3042
                     qemuDomainAsyncJob asyncJob)
3043
{
3044
    int ret = -1;
3045
    qemuDomainObjPrivatePtr priv = vm->privateData;
3046
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3047

3048
    /* Bring up netdevs before starting CPUs */
3049
    if (qemuInterfaceStartDevices(vm->def) < 0)
3050 3051
       goto cleanup;

3052
    VIR_DEBUG("Using lock state '%s'", NULLSTR(priv->lockState));
3053
    if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
3054
                                   vm, priv->lockState) < 0) {
3055 3056 3057 3058
        /* 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
         */
3059
        goto cleanup;
3060 3061 3062
    }
    VIR_FREE(priv->lockState);

3063 3064
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto release;
J
Jiri Denemark 已提交
3065

3066
    ret = qemuMonitorStartCPUs(priv->mon, conn);
3067 3068
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3069 3070 3071 3072 3073

    if (ret < 0)
        goto release;

    virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
3074

3075
 cleanup:
3076
    virObjectUnref(cfg);
3077
    return ret;
3078 3079 3080 3081 3082 3083

 release:
    if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
        VIR_WARN("Unable to release lease on %s", vm->def->name);
    VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));
    goto cleanup;
3084 3085 3086
}


3087 3088
int qemuProcessStopCPUs(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
3089
                        virDomainPausedReason reason,
3090
                        qemuDomainAsyncJob asyncJob)
3091
{
3092
    int ret = -1;
3093 3094
    qemuDomainObjPrivatePtr priv = vm->privateData;

3095
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
3096

3097 3098
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
J
Jiri Denemark 已提交
3099

3100
    ret = qemuMonitorStopCPUs(priv->mon);
3101 3102
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3103 3104 3105 3106

    if (ret < 0)
        goto cleanup;

3107 3108 3109
    /* de-activate netdevs after stopping CPUs */
    ignore_value(qemuInterfaceStopDevices(vm->def));

3110 3111 3112
    if (priv->job.current)
        ignore_value(virTimeMillisNow(&priv->job.current->stopped));

3113 3114 3115 3116
    virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
    if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
        VIR_WARN("Unable to release lease on %s", vm->def->name);
    VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));
J
Jiri Denemark 已提交
3117

3118
 cleanup:
3119 3120 3121 3122 3123
    return ret;
}



3124 3125 3126
static int
qemuProcessNotifyNets(virDomainDefPtr def)
{
3127
    size_t i;
3128

3129 3130
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
3131
        if (networkNotifyActualDevice(def, net) < 0)
3132 3133 3134 3135 3136
            return -1;
    }
    return 0;
}

3137
static int
3138
qemuProcessFiltersInstantiate(virDomainDefPtr def)
3139
{
3140
    size_t i;
3141

3142
    for (i = 0; i < def->nnets; i++) {
3143 3144
        virDomainNetDefPtr net = def->nets[i];
        if ((net->filter) && (net->ifname)) {
3145
            if (virDomainConfNWFilterInstantiate(def->uuid, net) < 0)
J
Ján Tomko 已提交
3146
                return 1;
3147 3148 3149
        }
    }

J
Ján Tomko 已提交
3150
    return 0;
3151 3152
}

3153
static int
3154
qemuProcessUpdateState(virQEMUDriverPtr driver, virDomainObjPtr vm)
3155 3156 3157
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
3158
    virDomainPausedReason reason;
3159
    virDomainState newState = VIR_DOMAIN_NOSTATE;
3160
    int oldReason;
3161
    int newReason;
3162
    bool running;
3163
    char *msg = NULL;
3164 3165
    int ret;

3166
    qemuDomainObjEnterMonitor(driver, vm);
3167
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
3168 3169
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
3170

3171
    if (ret < 0)
3172 3173
        return -1;

3174
    state = virDomainObjGetState(vm, &oldReason);
3175

3176 3177 3178 3179 3180 3181 3182 3183
    if (running &&
        (state == VIR_DOMAIN_SHUTOFF ||
         (state == VIR_DOMAIN_PAUSED &&
          oldReason == VIR_DOMAIN_PAUSED_STARTING_UP))) {
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_BOOTED;
        ignore_value(VIR_STRDUP_QUIET(msg, "finished booting"));
    } else if (state == VIR_DOMAIN_PAUSED && running) {
3184 3185
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
3186
        ignore_value(VIR_STRDUP_QUIET(msg, "was unpaused"));
3187
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
3188 3189 3190
        if (reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN) {
            newState = VIR_DOMAIN_SHUTDOWN;
            newReason = VIR_DOMAIN_SHUTDOWN_UNKNOWN;
3191
            ignore_value(VIR_STRDUP_QUIET(msg, "shutdown"));
3192
        } else if (reason == VIR_DOMAIN_PAUSED_CRASHED) {
3193 3194 3195
            newState = VIR_DOMAIN_CRASHED;
            newReason = VIR_DOMAIN_CRASHED_PANICKED;
            ignore_value(VIR_STRDUP_QUIET(msg, "crashed"));
3196 3197 3198
        } else {
            newState = VIR_DOMAIN_PAUSED;
            newReason = reason;
S
Stefan Berger 已提交
3199 3200
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
3201 3202 3203 3204 3205 3206 3207
        }
    }

    if (newState != VIR_DOMAIN_NOSTATE) {
        VIR_DEBUG("Domain %s %s while its monitor was disconnected;"
                  " changing state to %s (%s)",
                  vm->def->name,
3208
                  NULLSTR(msg),
3209 3210 3211 3212
                  virDomainStateTypeToString(newState),
                  virDomainStateReasonToString(newState, newReason));
        VIR_FREE(msg);
        virDomainObjSetState(vm, newState, newReason);
3213 3214 3215 3216 3217
    }

    return 0;
}

3218
static int
3219
qemuProcessRecoverMigration(virQEMUDriverPtr driver,
3220 3221
                            virDomainObjPtr vm,
                            virConnectPtr conn,
3222 3223
                            qemuDomainAsyncJob job,
                            qemuMigrationJobPhase phase,
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
                            virDomainState state,
                            int reason)
{
    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,
3250 3251
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279
                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 */
3280
            if (qemuMigrationCancel(driver, vm) < 0)
3281
                return -1;
3282 3283 3284 3285 3286 3287
            /* 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,
3288 3289
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308
                    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,
3309 3310
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324
                    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;
}

3325
static int
3326
qemuProcessRecoverJob(virQEMUDriverPtr driver,
3327 3328 3329 3330
                      virDomainObjPtr vm,
                      virConnectPtr conn,
                      const struct qemuDomainJobObj *job)
{
3331
    qemuDomainObjPrivatePtr priv = vm->privateData;
3332 3333 3334 3335 3336 3337 3338 3339
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
    case QEMU_ASYNC_JOB_MIGRATION_IN:
3340 3341 3342
        if (qemuProcessRecoverMigration(driver, vm, conn, job->asyncJob,
                                        job->phase, state, reason) < 0)
            return -1;
3343 3344 3345 3346
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
3347
    case QEMU_ASYNC_JOB_SNAPSHOT:
3348
        qemuDomainObjEnterMonitor(driver, vm);
3349
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
3350 3351
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
3352
        /* resume the domain but only if it was paused as a result of
3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368
         * running a migration-to-file operation.  Although we are
         * recovering an async job, this function is run at startup
         * and must resume things using sync monitor connections.  */
         if (state == VIR_DOMAIN_PAUSED &&
             ((job->asyncJob == QEMU_ASYNC_JOB_DUMP &&
               reason == VIR_DOMAIN_PAUSED_DUMP) ||
              (job->asyncJob == QEMU_ASYNC_JOB_SAVE &&
               reason == VIR_DOMAIN_PAUSED_SAVE) ||
              (job->asyncJob == QEMU_ASYNC_JOB_SNAPSHOT &&
               reason == VIR_DOMAIN_PAUSED_SNAPSHOT) ||
              reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
             if (qemuProcessStartCPUs(driver, vm, conn,
                                      VIR_DOMAIN_RUNNING_UNPAUSED,
                                      QEMU_ASYNC_JOB_NONE) < 0) {
                 VIR_WARN("Could not resume domain '%s' after migration to file",
                          vm->def->name);
3369 3370 3371 3372
            }
        }
        break;

3373 3374 3375 3376
    case QEMU_ASYNC_JOB_START:
        /* Already handled in VIR_DOMAIN_PAUSED_STARTING_UP check. */
        break;

3377 3378 3379 3380 3381 3382 3383 3384
    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        break;
    }

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

3385 3386 3387 3388
    /* 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.
     */
3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408
    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;

3409
    case QEMU_JOB_MIGRATION_OP:
3410
    case QEMU_JOB_ABORT:
3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421
    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;
}

3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436
static int
qemuProcessUpdateDevices(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDeviceDef dev;
    char **old;
    char **tmp;
    int ret = -1;

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

    old = priv->qemuDevices;
    priv->qemuDevices = NULL;
3437
    if (qemuDomainUpdateDeviceList(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
3438 3439 3440 3441 3442
        goto cleanup;

    if ((tmp = old)) {
        while (*tmp) {
            if (!virStringArrayHasString(priv->qemuDevices, *tmp) &&
3443 3444 3445 3446
                virDomainDefFindDevice(vm->def, *tmp, &dev, false) == 0 &&
                qemuDomainRemoveDevice(driver, vm, &dev) < 0) {
                goto cleanup;
            }
3447 3448 3449 3450 3451
            tmp++;
        }
    }
    ret = 0;

3452
 cleanup:
3453 3454 3455 3456
    virStringFreeList(old);
    return ret;
}

3457 3458
struct qemuProcessReconnectData {
    virConnectPtr conn;
3459
    virQEMUDriverPtr driver;
3460
    virDomainObjPtr obj;
3461 3462 3463 3464
};
/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
S
Stefan Berger 已提交
3465 3466 3467 3468
 *
 * 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.
3469
 *
3470
 * This function also inherits a locked and ref'd domain object.
3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482
 *
 * This function needs to:
 * 1. Enter job
 * 1. just before monitor reconnect do lightweight MonitorEnter
 *    (increase VM refcount and unlock VM)
 * 2. reconnect to monitor
 * 3. do lightweight MonitorExit (lock VM)
 * 4. continue reconnect process
 * 5. EndJob
 *
 * We can't do normal MonitorEnter & MonitorExit because these two lock the
 * monitor lock, which does not exists in this early phase.
3483 3484
 */
static void
3485
qemuProcessReconnect(void *opaque)
3486 3487
{
    struct qemuProcessReconnectData *data = opaque;
3488
    virQEMUDriverPtr driver = data->driver;
3489
    virDomainObjPtr obj = data->obj;
3490 3491
    qemuDomainObjPrivatePtr priv;
    virConnectPtr conn = data->conn;
3492
    struct qemuDomainJobObj oldjob;
3493 3494
    int state;
    int reason;
3495
    virQEMUDriverConfigPtr cfg;
3496
    size_t i;
3497
    int ret;
3498
    unsigned int stopFlags = 0;
3499

3500 3501
    VIR_FREE(data);

3502
    qemuDomainObjRestoreJob(obj, &oldjob);
3503 3504
    if (oldjob.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
3505

3506 3507 3508
    cfg = virQEMUDriverGetConfig(driver);
    priv = obj->privateData;

3509 3510 3511 3512 3513
    /* XXX If we ever gonna change pid file pattern, come up with
     * some intelligence here to deal with old paths. */
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, obj->def->name)))
        goto killvm;

3514 3515 3516 3517
    if (qemuDomainObjBeginJob(driver, obj, QEMU_JOB_MODIFY) < 0)
        goto killvm;

    virNWFilterReadLockFilterUpdates();
3518 3519 3520 3521

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

    /* XXX check PID liveliness & EXE path */
3522
    if (qemuConnectMonitor(driver, obj, QEMU_ASYNC_JOB_NONE, -1, -1) < 0)
3523 3524
        goto error;

D
Daniel P. Berrange 已提交
3525
    /* Failure to connect to agent shouldn't be fatal */
3526 3527 3528 3529
    if ((ret = qemuConnectAgent(driver, obj)) < 0) {
        if (ret == -2)
            goto error;

D
Daniel P. Berrange 已提交
3530 3531 3532 3533 3534 3535
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 obj->def->name);
        virResetLastError();
        priv->agentError = true;
    }

3536
    if (qemuHostdevUpdateActiveDomainDevices(driver, obj->def) < 0)
3537 3538
        goto error;

3539
    if (qemuConnectCgroup(driver, obj) < 0)
3540 3541
        goto error;

3542
    /* XXX: Need to change as long as lock is introduced for
3543
     * qemu_driver->sharedDevices.
3544 3545
     */
    for (i = 0; i < obj->def->ndisks; i++) {
3546 3547
        virDomainDeviceDef dev;

3548
        if (virStorageTranslateDiskSourcePool(conn, obj->def->disks[i]) < 0)
3549
            goto error;
3550

3551 3552 3553 3554 3555 3556
        /* XXX we should be able to restore all data from XML in the future.
         * This should be the only place that calls qemuDomainDetermineDiskChain
         * with @report_broken == false to guarantee best-effort domain
         * reconnect */
        if (qemuDomainDetermineDiskChain(driver, obj, obj->def->disks[i],
                                         true, false) < 0)
3557 3558
            goto error;

3559 3560 3561
        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = obj->def->disks[i];
        if (qemuAddSharedDevice(driver, &dev, obj->def->name) < 0)
3562 3563 3564
            goto error;
    }

3565 3566 3567
    if (qemuProcessUpdateState(driver, obj) < 0)
        goto error;

3568
    state = virDomainObjGetState(obj, &reason);
3569 3570 3571
    if (state == VIR_DOMAIN_SHUTOFF ||
        (state == VIR_DOMAIN_PAUSED &&
         reason == VIR_DOMAIN_PAUSED_STARTING_UP)) {
3572 3573 3574 3575 3576
        VIR_DEBUG("Domain '%s' wasn't fully started yet, killing it",
                  obj->def->name);
        goto error;
    }

3577 3578 3579
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
3580 3581
    if (!priv->qemuCaps &&
        !(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
3582 3583
                                                      obj->def->emulator,
                                                      obj->def->os.machine)))
3584 3585
        goto error;

3586
    /* In case the domain shutdown while we were not running,
3587
     * we need to finish the shutdown process. And we need to do it after
3588
     * we have virQEMUCaps filled in.
3589
     */
3590 3591 3592 3593 3594 3595
    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);
3596 3597
        qemuDomainObjEndJob(driver, obj);
        goto cleanup;
3598 3599
    }

3600 3601
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE))
        if ((qemuDomainAssignAddresses(obj->def, priv->qemuCaps, obj)) < 0)
3602
            goto error;
3603

3604 3605 3606 3607 3608 3609
    /* if domain requests security driver we haven't loaded, report error, but
     * do not kill the domain
     */
    ignore_value(virSecurityManagerCheckAllLabel(driver->securityManager,
                                                 obj->def));

3610
    if (virSecurityManagerReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
3611 3612
        goto error;

3613 3614 3615
    if (qemuProcessNotifyNets(obj->def) < 0)
        goto error;

3616
    if (qemuProcessFiltersInstantiate(obj->def))
3617 3618
        goto error;

3619
    if (qemuDomainCheckEjectableMedia(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
3620 3621
        goto error;

3622 3623 3624
    if (qemuProcessReconnectRefreshChannelVirtioState(driver, obj) < 0)
        goto error;

3625 3626 3627
    if (qemuProcessRefreshBalloonState(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
        goto error;

3628 3629 3630
    if (qemuProcessRecoverJob(driver, obj, conn, &oldjob) < 0)
        goto error;

3631 3632 3633
    if (qemuProcessUpdateDevices(driver, obj) < 0)
        goto error;

3634
    /* update domain state XML with possibly updated state in virDomainObj */
3635
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, obj) < 0)
3636 3637
        goto error;

3638 3639
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3640
        char *xml = qemuDomainDefFormatXML(driver, obj->def, 0);
3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654
        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;
    }

3655
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
3656 3657
        driver->inhibitCallback(true, driver->inhibitOpaque);

3658 3659
    qemuDomainObjEndJob(driver, obj);
    goto cleanup;
3660

3661
 error:
3662
    qemuDomainObjEndJob(driver, obj);
3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677
 killvm:
    if (virDomainObjIsActive(obj)) {
        /* We can't get the monitor back, so must kill the VM
         * to remove danger of it ending up running twice if
         * user tries to start it again later
         */
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NO_SHUTDOWN)) {
            /* If we couldn't get the monitor and qemu supports
             * no-shutdown, we can safely say that the domain
             * crashed ... */
            state = VIR_DOMAIN_SHUTOFF_CRASHED;
        } else {
            /* ... but if it doesn't we can't say what the state
             * really is and FAILED means "failed to start" */
            state = VIR_DOMAIN_SHUTOFF_UNKNOWN;
3678
        }
3679
        qemuProcessStop(driver, obj, state, stopFlags);
3680
    }
3681

3682
    qemuDomainRemoveInactive(driver, obj);
3683

3684
 cleanup:
M
Michal Privoznik 已提交
3685
    virDomainObjEndAPI(&obj);
3686
    virObjectUnref(conn);
3687
    virObjectUnref(cfg);
3688
    virNWFilterUnlockFilterUpdates();
3689 3690
}

3691 3692
static int
qemuProcessReconnectHelper(virDomainObjPtr obj,
3693 3694 3695 3696 3697 3698
                           void *opaque)
{
    virThread thread;
    struct qemuProcessReconnectData *src = opaque;
    struct qemuProcessReconnectData *data;

3699
    /* If the VM was inactive, we don't need to reconnect */
3700 3701 3702
    if (!obj->pid)
        return 0;

3703
    if (VIR_ALLOC(data) < 0)
3704
        return -1;
3705 3706

    memcpy(data, src, sizeof(*data));
3707
    data->obj = obj;
3708

3709 3710
    /* this lock and reference will be eventually transferred to the thread
     * that handles the reconnect */
3711
    virObjectLock(obj);
3712
    virObjectRef(obj);
3713

3714 3715 3716
    /* Since we close the connection later on, we have to make sure that the
     * threads we start see a valid connection throughout their lifetime. We
     * simply increase the reference counter here.
S
Stefan Berger 已提交
3717
     */
3718
    virObjectRef(data->conn);
S
Stefan Berger 已提交
3719

3720
    if (virThreadCreate(&thread, false, qemuProcessReconnect, data) < 0) {
3721 3722 3723
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not create thread. QEMU initialization "
                         "might be incomplete"));
3724 3725
       /* We can't spawn a thread and thus connect to monitor. Kill qemu. */
        qemuProcessStop(src->driver, obj, VIR_DOMAIN_SHUTOFF_FAILED, 0);
3726
        qemuDomainRemoveInactive(src->driver, obj);
3727

M
Michal Privoznik 已提交
3728
        virDomainObjEndAPI(&obj);
3729 3730 3731 3732
        virObjectUnref(data->conn);
        VIR_FREE(data);
        return -1;
    }
3733

3734
    return 0;
3735 3736 3737 3738 3739 3740 3741 3742 3743
}

/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
3744
qemuProcessReconnectAll(virConnectPtr conn, virQEMUDriverPtr driver)
3745
{
3746
    struct qemuProcessReconnectData data = {.conn = conn, .driver = driver};
3747
    virDomainObjListForEach(driver->domains, qemuProcessReconnectHelper, &data);
3748 3749
}

3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772
static int
qemuProcessVNCAllocatePorts(virQEMUDriverPtr driver,
                            virDomainGraphicsDefPtr graphics)
{
    unsigned short port;

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

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

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

    return 0;
}
3773

3774
int
3775 3776
qemuProcessSPICEAllocatePorts(virQEMUDriverPtr driver,
                              virQEMUDriverConfigPtr cfg,
3777 3778
                              virDomainGraphicsDefPtr graphics,
                              bool allocate)
3779 3780 3781
{
    unsigned short port = 0;
    unsigned short tlsPort;
3782
    size_t i;
3783 3784 3785 3786 3787 3788 3789
    int defaultMode = graphics->data.spice.defaultMode;

    bool needTLSPort = false;
    bool needPort = false;

    if (graphics->data.spice.autoport) {
        /* check if tlsPort or port need allocation */
3790
        for (i = 0; i < VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_LAST; i++) {
3791 3792 3793 3794 3795 3796 3797 3798 3799 3800
            switch (graphics->data.spice.channels[i]) {
            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
                needTLSPort = true;
                break;

            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
                needPort = true;
                break;

            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
3801
                /* default mode will be used */
3802 3803 3804
                break;
            }
        }
3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819
        switch (defaultMode) {
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
            needTLSPort = true;
            break;

        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
            needPort = true;
            break;

        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
            if (cfg->spiceTLS)
                needTLSPort = true;
            needPort = true;
            break;
        }
3820 3821
    }

3822 3823 3824 3825 3826 3827 3828 3829 3830 3831
    if (!allocate) {
        if (needPort || graphics->data.spice.port == -1)
            graphics->data.spice.port = 5901;

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

        return 0;
    }

3832
    if (needPort || graphics->data.spice.port == -1) {
3833
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
3834
            goto error;
3835 3836 3837 3838

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

3839 3840
    if (needTLSPort || graphics->data.spice.tlsPort == -1) {
        if (!cfg->spiceTLS) {
3841 3842 3843 3844
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Auto allocation of spice TLS port requested "
                             "but spice TLS is disabled in qemu.conf"));
            goto error;
3845
        }
3846 3847 3848 3849 3850

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

        graphics->data.spice.tlsPort = tlsPort;
3851 3852
    }

3853
    return 0;
3854

3855
 error:
3856
    virPortAllocatorRelease(driver->remotePorts, port);
3857
    return -1;
3858 3859 3860
}


3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878
static bool
qemuValidateCpuMax(virDomainDefPtr def, virQEMUCapsPtr qemuCaps)
{
    unsigned int maxCpus;

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

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

    return true;
}

3879 3880

static bool
3881 3882 3883
qemuProcessVerifyGuestCPU(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          int asyncJob)
3884 3885 3886 3887 3888
{
    virDomainDefPtr def = vm->def;
    virArch arch = def->os.arch;
    virCPUDataPtr guestcpu = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
3889
    int rc;
3890
    bool ret = false;
J
Ján Tomko 已提交
3891
    size_t i;
3892

3893 3894 3895 3896 3897
    /* no features are passed to QEMU with -cpu host
     * so it makes no sense to verify them */
    if (def->cpu && def->cpu->mode == VIR_CPU_MODE_HOST_PASSTHROUGH)
        return true;

3898 3899 3900
    switch (arch) {
    case VIR_ARCH_I686:
    case VIR_ARCH_X86_64:
3901 3902
        if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
            return false;
3903
        rc = qemuMonitorGetGuestCPU(priv->mon, arch, &guestcpu);
3904 3905
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return false;
3906

3907 3908 3909 3910
        if (rc < 0) {
            if (rc == -2)
                break;

3911
            goto cleanup;
3912
        }
3913

J
Ján Tomko 已提交
3914
        if (def->features[VIR_DOMAIN_FEATURE_PVSPINLOCK] == VIR_TRISTATE_SWITCH_ON) {
3915 3916 3917 3918 3919 3920
            if (!cpuHasFeature(guestcpu, VIR_CPU_x86_KVM_PV_UNHALT)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("host doesn't support paravirtual spinlocks"));
                goto cleanup;
            }
        }
J
Ján Tomko 已提交
3921

3922
        for (i = 0; def->cpu && i < def->cpu->nfeatures; i++) {
J
Ján Tomko 已提交
3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934
            virCPUFeatureDefPtr feature = &def->cpu->features[i];

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

            if (STREQ(feature->name, "invtsc") &&
                !cpuHasFeature(guestcpu, feature->name)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("host doesn't support invariant TSC"));
                goto cleanup;
            }
        }
3935 3936 3937 3938 3939 3940 3941 3942
        break;

    default:
        break;
    }

    ret = true;

3943
 cleanup:
3944 3945 3946 3947 3948
    cpuDataFree(guestcpu);
    return ret;
}


3949 3950
static int
qemuPrepareNVRAM(virQEMUDriverConfigPtr cfg,
3951
                 virDomainObjPtr vm,
3952 3953 3954 3955 3956
                 bool migrated)
{
    int ret = -1;
    int srcFD = -1;
    int dstFD = -1;
3957
    virDomainLoaderDefPtr loader = vm->def->os.loader;
3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982
    bool generated = false;
    bool created = false;

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

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

    /* Autogenerate nvram path if needed.*/
    if (!loader->nvram) {
        if (virAsprintf(&loader->nvram,
3983 3984
                        "%s/%s_VARS.fd",
                        cfg->nvramDir, vm->def->name) < 0)
3985 3986 3987
            goto cleanup;

        generated = true;
3988

3989 3990
        if (vm->persistent &&
            virDomainSaveConfig(cfg->configDir, vm->def) < 0)
3991
            goto cleanup;
3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081
    }

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

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

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

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

        do {
            char buf[1024];

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

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

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

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

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


4082 4083 4084
static void
qemuLogOperation(virDomainObjPtr vm,
                 const char *msg,
4085 4086
                 virCommandPtr cmd,
                 qemuDomainLogContextPtr logCtxt)
4087 4088 4089 4090 4091 4092 4093
{
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int qemuVersion = virQEMUCapsGetVersion(priv->qemuCaps);
    const char *package = virQEMUCapsGetPackage(priv->qemuCaps);

    if ((timestamp = virTimeStringNow()) == NULL)
4094
        goto cleanup;
4095

4096 4097 4098 4099 4100 4101 4102
    if (qemuDomainLogContextWrite(logCtxt, "%s: %s %s, qemu version: %d.%d.%d%s\n",
                                  timestamp, msg, VIR_LOG_VERSION_STRING,
                                  (qemuVersion / 1000000) % 1000,
                                  (qemuVersion / 1000) % 1000,
                                  qemuVersion % 1000,
                                  package ? package : "") < 0)
        goto cleanup;
4103

4104 4105 4106 4107 4108
    if (cmd) {
        char *args = virCommandToString(cmd);
        qemuDomainLogContextWrite(logCtxt, "%s\n", args);
        VIR_FREE(args);
    }
4109 4110 4111 4112 4113

 cleanup:
    VIR_FREE(timestamp);
}

4114 4115 4116 4117 4118 4119 4120 4121

void
qemuProcessIncomingDefFree(qemuProcessIncomingDefPtr inc)
{
    if (!inc)
        return;

    VIR_FREE(inc->launchURI);
4122
    VIR_FREE(inc->deferredURI);
4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149
    VIR_FREE(inc);
}


/*
 * This function does not copy @path, the caller is responsible for keeping
 * the @path pointer valid during the lifetime of the allocated
 * qemuProcessIncomingDef structure.
 */
qemuProcessIncomingDefPtr
qemuProcessIncomingDefNew(virQEMUCapsPtr qemuCaps,
                          const char *migrateFrom,
                          int fd,
                          const char *path)
{
    qemuProcessIncomingDefPtr inc = NULL;

    if (qemuMigrationCheckIncoming(qemuCaps, migrateFrom) < 0)
        return NULL;

    if (VIR_ALLOC(inc) < 0)
        return NULL;

    inc->launchURI = qemuMigrationIncomingURI(migrateFrom, fd);
    if (!inc->launchURI)
        goto error;

4150 4151 4152 4153 4154 4155
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_INCOMING_DEFER)) {
        inc->deferredURI = inc->launchURI;
        if (VIR_STRDUP(inc->launchURI, "defer") < 0)
            goto error;
    }

4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166
    inc->fd = fd;
    inc->path = path;

    return inc;

 error:
    qemuProcessIncomingDefFree(inc);
    return NULL;
}


4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196
/*
 * This function starts a new QEMU_ASYNC_JOB_START async job. The user is
 * responsible for calling qemuProcessEndJob to stop this job and for passing
 * QEMU_ASYNC_JOB_START as @asyncJob argument to any function requiring this
 * parameter between qemuProcessBeginJob and qemuProcessEndJob.
 */
int
qemuProcessBeginJob(virQEMUDriverPtr driver,
                    virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (qemuDomainObjBeginAsyncJob(driver, vm, QEMU_ASYNC_JOB_START) < 0)
        return -1;

    qemuDomainObjSetAsyncJobMask(vm, QEMU_JOB_NONE);
    priv->job.current->type = VIR_DOMAIN_JOB_UNBOUNDED;

    return 0;
}


void
qemuProcessEndJob(virQEMUDriverPtr driver,
                  virDomainObjPtr vm)
{
    qemuDomainObjEndAsyncJob(driver, vm);
}


4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219
static int
qemuProcessStartHook(virQEMUDriverPtr driver,
                     virDomainObjPtr vm,
                     virHookQemuOpType op,
                     virHookSubopType subop)
{
    char *xml;
    int ret;

    if (!virHookPresent(VIR_HOOK_DRIVER_QEMU))
        return 0;

    if (!(xml = qemuDomainDefFormatXML(driver, vm->def, 0)))
        return -1;

    ret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name, op, subop,
                      NULL, xml, NULL);
    VIR_FREE(xml);

    return ret;
}


4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297
static int
qemuProcessSetupGraphics(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    size_t i;
    int ret = -1;

    for (i = 0; i < vm->def->ngraphics; ++i) {
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
            !graphics->data.vnc.autoport) {
            if (virPortAllocatorSetUsed(driver->remotePorts,
                                        graphics->data.vnc.port,
                                        true) < 0)
                goto cleanup;
            graphics->data.vnc.portReserved = true;

        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE &&
                   !graphics->data.spice.autoport) {
            if (graphics->data.spice.port > 0) {
                if (virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.port,
                                            true) < 0)
                    goto cleanup;
                graphics->data.spice.portReserved = true;
            }

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

    for (i = 0; i < vm->def->ngraphics; ++i) {
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (qemuProcessVNCAllocatePorts(driver, graphics) < 0)
                goto cleanup;
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (qemuProcessSPICEAllocatePorts(driver, cfg, graphics, true) < 0)
                goto cleanup;
        }

        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ||
            graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (graphics->nListens == 0) {
                if (VIR_EXPAND_N(graphics->listens, graphics->nListens, 1) < 0)
                    goto cleanup;
                graphics->listens[0].type = VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_ADDRESS;
                if (VIR_STRDUP(graphics->listens[0].address,
                               graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ?
                               cfg->vncListen : cfg->spiceListen) < 0) {
                    VIR_SHRINK_N(graphics->listens, graphics->nListens, 1);
                    goto cleanup;
                }
                graphics->listens[0].fromConfig = true;
            } else if (graphics->nListens > 1) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("QEMU does not support multiple listen "
                                 "addresses for one graphics device."));
                goto cleanup;
            }
        }
    }

    ret = 0;

 cleanup:
    virObjectUnref(cfg);
    return ret;
}


4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356
static int
qemuProcessSetupRawIO(virQEMUDriverPtr driver,
                      virDomainObjPtr vm,
                      virCommandPtr cmd ATTRIBUTE_UNUSED)
{
    bool rawio = false;
    size_t i;
    int ret = -1;

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

        if (disk->rawio == VIR_TRISTATE_BOOL_YES) {
            rawio = true;
#ifndef CAP_SYS_RAWIO
            break;
#endif
        }

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

        if (qemuSetUnprivSGIO(&dev) < 0)
            goto cleanup;
    }

    /* If rawio not already set, check hostdevs as well */
    if (!rawio) {
        for (i = 0; i < vm->def->nhostdevs; i++) {
            virDomainHostdevSubsysSCSIPtr scsisrc =
                &vm->def->hostdevs[i]->source.subsys.u.scsi;
            if (scsisrc->rawio == VIR_TRISTATE_BOOL_YES) {
                rawio = true;
                break;
            }
        }
    }

    ret = 0;

 cleanup:
    if (rawio) {
#ifdef CAP_SYS_RAWIO
        if (ret == 0)
            virCommandAllowCap(cmd, CAP_SYS_RAWIO);
#else
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Raw I/O is not supported on this platform"));
        ret = -1;
#endif
    }
    return ret;
}


4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389
static int
qemuProcessSetupBalloon(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
                        qemuDomainAsyncJob asyncJob)
{
    unsigned long long balloon = vm->def->mem.cur_balloon;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int period;
    int ret = -1;

    if (!vm->def->memballoon ||
        vm->def->memballoon->model == VIR_DOMAIN_MEMBALLOON_MODEL_NONE)
        return 0;

    period = vm->def->memballoon->period;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;

    if (period)
        qemuMonitorSetMemoryStatsPeriod(priv->mon, period);
    if (qemuMonitorSetBalloon(priv->mon, balloon) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
    return ret;
}


J
Jiri Denemark 已提交
4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417
static int
qemuProcessMakeDir(virQEMUDriverPtr driver,
                   virDomainObjPtr vm,
                   const char *parentDir)
{
    char *path = NULL;
    int ret = -1;

    if (virAsprintf(&path, "%s/domain-%s", parentDir, vm->def->name) < 0)
        goto cleanup;

    if (virFileMakePathWithMode(path, 0750) < 0) {
        virReportSystemError(errno, _("Cannot create directory '%s'"), path);
        goto cleanup;
    }

    if (virSecurityManagerDomainSetDirLabel(driver->securityManager,
                                            vm->def, path) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    VIR_FREE(path);
    return ret;
}


J
Jiri Denemark 已提交
4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501
/**
 * qemuProcessInit:
 *
 * Prepares the domain up to the point when priv->qemuCaps is initialized. The
 * function calls qemuProcessStop when needed.
 *
 * Returns 0 on success, -1 on error.
 */
int
qemuProcessInit(virQEMUDriverPtr driver,
                virDomainObjPtr vm,
                bool migration)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    virCapsPtr caps = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int stopFlags;
    int ret = -1;

    VIR_DEBUG("vm=%p name=%s id=%d migration=%d",
              vm, vm->def->name, vm->def->id, migration);

    VIR_DEBUG("Beginning VM startup process");

    if (virDomainObjIsActive(vm)) {
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("VM is already active"));
        goto cleanup;
    }

    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto stop;

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

    if (qemuPrepareNVRAM(cfg, vm, migration) < 0)
        goto stop;

    /* 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(caps, driver->xmlopt, vm, true) < 0)
        goto stop;

    vm->def->id = qemuDriverAllocateID(driver);
    qemuDomainSetFakeReboot(driver, vm, false);
    virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_STARTING_UP);

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

    /* Run an early hook to set-up missing devices */
    if (qemuProcessStartHook(driver, vm,
                             VIR_HOOK_QEMU_OP_PREPARE,
                             VIR_HOOK_SUBOP_BEGIN) < 0)
        goto stop;

    VIR_DEBUG("Determining emulator version");
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
        goto stop;

    ret = 0;

 cleanup:
    virObjectUnref(cfg);
    virObjectUnref(caps);
    return ret;

 stop:
    stopFlags = VIR_QEMU_PROCESS_STOP_NO_RELABEL;
    if (migration)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, stopFlags);
    goto cleanup;
}


J
Jiri Denemark 已提交
4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523
/**
 * qemuProcessLaunch:
 *
 * Launch a new QEMU process with stopped virtual CPUs.
 *
 * The caller is supposed to call qemuProcessStop with appropriate
 * flags in case of failure.
 *
 * Returns 0 on success,
 *        -1 on error which happened before devices were labeled and thus
 *           there is no need to restore them,
 *        -2 on error requesting security labels to be restored.
 */
int
qemuProcessLaunch(virConnectPtr conn,
                  virQEMUDriverPtr driver,
                  virDomainObjPtr vm,
                  qemuDomainAsyncJob asyncJob,
                  qemuProcessIncomingDefPtr incoming,
                  virDomainSnapshotObjPtr snapshot,
                  virNetDevVPortProfileOp vmop,
                  unsigned int flags)
4524
{
4525
    int ret = -1;
4526
    int rv;
4527
    int logfile = -1;
4528
    qemuDomainLogContextPtr logCtxt = NULL;
4529 4530 4531
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCommandPtr cmd = NULL;
    struct qemuProcessHookData hookData;
4532
    size_t i;
4533
    char *nodeset = NULL;
J
Jiri Denemark 已提交
4534
    virQEMUDriverConfigPtr cfg;
4535
    virCapsPtr caps = NULL;
4536
    unsigned int hostdev_flags = 0;
4537 4538
    size_t nnicindexes = 0;
    int *nicindexes = NULL;
4539

J
Jiri Denemark 已提交
4540 4541 4542 4543 4544 4545 4546 4547 4548 4549
    VIR_DEBUG("vm=%p name=%s id=%d asyncJob=%d "
              "incoming.launchURI=%s incoming.deferredURI=%s "
              "incoming.fd=%d incoming.path=%s "
              "snapshot=%p vmop=%d flags=0x%x",
              vm, vm->def->name, vm->def->id, asyncJob,
              NULLSTR(incoming ? incoming->launchURI : NULL),
              NULLSTR(incoming ? incoming->deferredURI : NULL),
              incoming ? incoming->fd : -1,
              NULLSTR(incoming ? incoming->path : NULL),
              snapshot, vmop, flags);
4550

4551 4552 4553 4554
    /* Okay, these are just internal flags,
     * but doesn't hurt to check */
    virCheckFlags(VIR_QEMU_PROCESS_START_COLD |
                  VIR_QEMU_PROCESS_START_PAUSED |
4555
                  VIR_QEMU_PROCESS_START_AUTODESTROY, -1);
4556

4557 4558
    cfg = virQEMUDriverGetConfig(driver);

4559 4560 4561
    hookData.conn = conn;
    hookData.vm = vm;
    hookData.driver = driver;
4562 4563
    /* We don't increase cfg's reference counter here. */
    hookData.cfg = cfg;
4564

4565
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
J
Jiri Denemark 已提交
4566
        goto cleanup;
4567

4568 4569 4570 4571 4572 4573
    /* network devices must be "prepared" before hostdevs, because
     * setting up a network device might create a new hostdev that
     * will need to be setup.
     */
    VIR_DEBUG("Preparing network devices");
    if (qemuNetworkPrepareDevices(vm->def) < 0)
J
Jiri Denemark 已提交
4574
        goto cleanup;
4575

4576
    /* Must be run before security labelling */
4577
    VIR_DEBUG("Preparing host devices");
4578 4579
    if (!cfg->relaxedACS)
        hostdev_flags |= VIR_HOSTDEV_STRICT_ACS_CHECK;
J
Jiri Denemark 已提交
4580
    if (!incoming)
4581
        hostdev_flags |= VIR_HOSTDEV_COLD_BOOT;
4582 4583
    if (qemuHostdevPrepareDomainDevices(driver, vm->def, priv->qemuCaps,
                                        hostdev_flags) < 0)
J
Jiri Denemark 已提交
4584
        goto cleanup;
4585

4586
    VIR_DEBUG("Preparing chr devices");
4587 4588 4589 4590
    if (virDomainChrDefForeach(vm->def,
                               true,
                               qemuProcessPrepareChardevDevice,
                               NULL) < 0)
J
Jiri Denemark 已提交
4591
        goto cleanup;
4592

4593 4594
    VIR_DEBUG("Checking domain and device security labels");
    if (virSecurityManagerCheckAllLabel(driver->securityManager, vm->def) < 0)
J
Jiri Denemark 已提交
4595
        goto cleanup;
4596

4597 4598
    /* If you are using a SecurityDriver with dynamic labelling,
       then generate a security label for isolation */
4599
    VIR_DEBUG("Generating domain security label (if required)");
4600
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0) {
4601
        virDomainAuditSecurityLabel(vm, false);
J
Jiri Denemark 已提交
4602
        goto cleanup;
4603
    }
4604
    virDomainAuditSecurityLabel(vm, true);
4605

4606
    if (vm->def->mem.nhugepages) {
4607 4608 4609 4610
        for (i = 0; i < cfg->nhugetlbfs; i++) {
            char *hugepagePath = qemuGetHugepagePath(&cfg->hugetlbfs[i]);

            if (!hugepagePath)
J
Jiri Denemark 已提交
4611
                goto cleanup;
4612 4613 4614 4615 4616 4617

            if (virSecurityManagerSetHugepages(driver->securityManager,
                                               vm->def, hugepagePath) < 0) {
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("Unable to set huge path in security driver"));
                VIR_FREE(hugepagePath);
J
Jiri Denemark 已提交
4618
                goto cleanup;
4619 4620
            }
            VIR_FREE(hugepagePath);
4621 4622 4623
        }
    }

4624 4625
    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
4626
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
4627
    qemuRemoveCgroup(driver, vm);
4628

4629 4630
    VIR_DEBUG("Setting up ports for graphics");
    if (qemuProcessSetupGraphics(driver, vm) < 0)
J
Jiri Denemark 已提交
4631
        goto cleanup;
4632

4633
    if (virFileMakePath(cfg->logDir) < 0) {
4634 4635
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
4636
                             cfg->logDir);
J
Jiri Denemark 已提交
4637
        goto cleanup;
4638 4639
    }

4640
    VIR_DEBUG("Creating domain log file");
4641 4642
    if (!(logCtxt = qemuDomainLogContextNew(driver, vm,
                                            QEMU_DOMAIN_LOG_CONTEXT_MODE_START)))
J
Jiri Denemark 已提交
4643
        goto cleanup;
4644
    logfile = qemuDomainLogContextGetWriteFD(logCtxt);
4645

4646 4647
    if (vm->def->virtType == VIR_DOMAIN_VIRT_KVM) {
        VIR_DEBUG("Checking for KVM availability");
4648
        if (!virFileExists("/dev/kvm")) {
4649 4650 4651 4652
            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."));
J
Jiri Denemark 已提交
4653
            goto cleanup;
4654 4655 4656
        }
    }

4657
    if (!qemuValidateCpuMax(vm->def, priv->qemuCaps))
J
Jiri Denemark 已提交
4658
        goto cleanup;
4659

4660
    if (qemuAssignDeviceAliases(vm->def, priv->qemuCaps) < 0)
J
Jiri Denemark 已提交
4661
        goto cleanup;
4662

4663 4664 4665
    /* Get the advisory nodeset from numad if 'placement' of
     * either <vcpu> or <numatune> is 'auto'.
     */
4666
    if (virDomainDefNeedsPlacementAdvice(vm->def)) {
4667
        nodeset = virNumaGetAutoPlacementAdvice(vm->def->vcpus,
4668
                                                virDomainDefGetMemoryActual(vm->def));
4669
        if (!nodeset)
J
Jiri Denemark 已提交
4670
            goto cleanup;
4671 4672 4673

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

4674 4675
        if (virBitmapParse(nodeset, 0, &priv->autoNodeset,
                           VIR_DOMAIN_CPUMASK_LEN) < 0)
J
Jiri Denemark 已提交
4676
            goto cleanup;
4677 4678 4679

        if (!(priv->autoCpuset = virCapabilitiesGetCpusForNodemask(caps,
                                                                   priv->autoNodeset)))
J
Jiri Denemark 已提交
4680
            goto cleanup;
4681 4682
    }

J
Jiri Denemark 已提交
4683
    if (!incoming && !snapshot &&
4684
        virDomainDefCheckDuplicateDiskInfo(vm->def) < 0)
J
Jiri Denemark 已提交
4685
        goto cleanup;
4686

4687 4688 4689 4690
    /* "volume" type disk's source must be translated before
     * cgroup and security setting.
     */
    for (i = 0; i < vm->def->ndisks; i++) {
4691
        if (virStorageTranslateDiskSourcePool(conn, vm->def->disks[i]) < 0)
J
Jiri Denemark 已提交
4692
            goto cleanup;
4693 4694
    }

4695 4696
    if (qemuDomainCheckDiskPresence(driver, vm,
                                    flags & VIR_QEMU_PROCESS_START_COLD) < 0)
J
Jiri Denemark 已提交
4697
        goto cleanup;
4698

4699 4700 4701 4702
    if (vm->def->mem.min_guarantee) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Parameter 'min_guarantee' "
                         "not supported by QEMU."));
J
Jiri Denemark 已提交
4703
        goto cleanup;
4704 4705
    }

4706
    if (VIR_ALLOC(priv->monConfig) < 0)
J
Jiri Denemark 已提交
4707
        goto cleanup;
4708

4709
    VIR_DEBUG("Preparing monitor state");
4710
    if (qemuProcessPrepareMonitorChr(cfg, priv->monConfig, vm->def->name) < 0)
J
Jiri Denemark 已提交
4711
        goto cleanup;
4712

E
Eric Blake 已提交
4713
    priv->monJSON = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON);
4714 4715
    priv->monError = false;
    priv->monStart = 0;
4716 4717
    priv->gotShutdown = false;

4718
    VIR_FREE(priv->pidfile);
4719
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, vm->def->name))) {
4720 4721
        virReportSystemError(errno,
                             "%s", _("Failed to build pidfile path."));
J
Jiri Denemark 已提交
4722
        goto cleanup;
4723 4724
    }

4725 4726
    if (unlink(priv->pidfile) < 0 &&
        errno != ENOENT) {
4727
        virReportSystemError(errno,
4728 4729
                             _("Cannot remove stale PID file %s"),
                             priv->pidfile);
J
Jiri Denemark 已提交
4730
        goto cleanup;
4731 4732 4733 4734 4735 4736
    }

    /*
     * Normally PCI addresses are assigned in the virDomainCreate
     * or virDomainDefine methods. We might still need to assign
     * some here to cope with the question of upgrades. Regardless
M
Martin Kletzander 已提交
4737
     * we also need to populate the PCI address set cache for later
4738 4739
     * use in hotplug
     */
4740
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4741
        VIR_DEBUG("Assigning domain PCI addresses");
4742
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
J
Jiri Denemark 已提交
4743
            goto cleanup;
4744 4745
    }

4746
    VIR_DEBUG("Building emulator command line");
4747
    if (!(cmd = qemuBuildCommandLine(conn, driver, vm->def, priv->monConfig,
E
Eric Blake 已提交
4748
                                     priv->monJSON, priv->qemuCaps,
4749 4750
                                     incoming ? incoming->launchURI : NULL,
                                     snapshot, vmop,
4751
                                     &buildCommandLineCallbacks, false,
4752
                                     qemuCheckFips(),
4753
                                     priv->autoNodeset,
4754
                                     &nnicindexes, &nicindexes)))
J
Jiri Denemark 已提交
4755
        goto cleanup;
4756

4757 4758
    if (incoming && incoming->fd != -1)
        virCommandPassFD(cmd, incoming->fd, 0);
4759 4760 4761 4762 4763

    /*
     * Create all per-domain directories in order to make sure domain
     * with any possible seclabels can access it.
     */
J
Jiri Denemark 已提交
4764 4765
    if (qemuProcessMakeDir(driver, vm, cfg->libDir) < 0 ||
        qemuProcessMakeDir(driver, vm, cfg->channelTargetDir) < 0)
J
Jiri Denemark 已提交
4766
        goto cleanup;
4767

4768
    /* now that we know it is about to start call the hook if present */
4769 4770 4771
    if (qemuProcessStartHook(driver, vm,
                             VIR_HOOK_QEMU_OP_START,
                             VIR_HOOK_SUBOP_BEGIN) < 0)
J
Jiri Denemark 已提交
4772
        goto cleanup;
4773

4774
    qemuLogOperation(vm, "starting up", cmd, logCtxt);
4775

4776
    qemuDomainObjCheckTaint(driver, vm, logCtxt);
4777

4778
    qemuDomainLogContextMarkPosition(logCtxt);
4779 4780

    VIR_DEBUG("Clear emulator capabilities: %d",
4781 4782
              cfg->clearEmulatorCapabilities);
    if (cfg->clearEmulatorCapabilities)
4783 4784
        virCommandClearCaps(cmd);

4785 4786
    VIR_DEBUG("Setting up raw IO");
    if (qemuProcessSetupRawIO(driver, vm, cmd) < 0)
J
Jiri Denemark 已提交
4787
        goto cleanup;
4788

4789
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);
4790 4791
    virCommandSetMaxProcesses(cmd, cfg->maxProcesses);
    virCommandSetMaxFiles(cmd, cfg->maxFiles);
4792
    virCommandSetUmask(cmd, 0x002);
4793

4794 4795
    VIR_DEBUG("Setting up security labelling");
    if (virSecurityManagerSetChildProcessLabel(driver->securityManager,
4796
                                               vm->def, cmd) < 0)
J
Jiri Denemark 已提交
4797
        goto cleanup;
4798

4799 4800 4801
    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
4802
    virCommandSetPidFile(cmd, priv->pidfile);
4803
    virCommandDaemonize(cmd);
4804
    virCommandRequireHandshake(cmd);
4805

4806
    if (virSecurityManagerPreFork(driver->securityManager) < 0)
J
Jiri Denemark 已提交
4807
        goto cleanup;
4808
    rv = virCommandRun(cmd, NULL);
4809
    virSecurityManagerPostFork(driver->securityManager);
4810

E
Eric Blake 已提交
4811
    /* wait for qemu process to show up */
4812
    if (rv == 0) {
4813
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
4814 4815
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
4816
            rv = -1;
4817
        }
4818 4819 4820 4821 4822
        VIR_DEBUG("QEMU vm=%p name=%s running with pid=%llu",
                  vm, vm->def->name, (unsigned long long)vm->pid);
    } else {
        VIR_DEBUG("QEMU vm=%p name=%s failed to spawn",
                  vm, vm->def->name);
4823 4824
    }

4825
    VIR_DEBUG("Writing early domain status to disk");
4826
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
J
Jiri Denemark 已提交
4827
        goto cleanup;
4828

4829 4830
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
4831
        /* Read errors from child that occurred between fork and exec. */
4832 4833
        int logfd = qemuDomainLogContextGetReadFD(logCtxt);
        off_t pos = qemuDomainLogContextGetPosition(logCtxt);
4834 4835 4836 4837
        if (logfd >= 0) {
            qemuProcessReportLogError(logfd, pos,
                                      _("Process exited prior to exec"));
        }
J
Jiri Denemark 已提交
4838
        goto cleanup;
4839 4840
    }

4841
    VIR_DEBUG("Setting up domain cgroup (if required)");
4842
    if (qemuSetupCgroup(driver, vm, nnicindexes, nicindexes) < 0)
J
Jiri Denemark 已提交
4843
        goto cleanup;
4844 4845 4846 4847

    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
    if (!vm->def->cputune.emulatorpin &&
4848
        qemuProcessInitCpuAffinity(vm) < 0)
J
Jiri Denemark 已提交
4849
        goto cleanup;
4850

4851 4852
    VIR_DEBUG("Setting domain security labels");
    if (virSecurityManagerSetAllLabel(driver->securityManager,
J
Jiri Denemark 已提交
4853 4854 4855
                                      vm->def,
                                      incoming ? incoming->path : NULL) < 0)
        goto cleanup;
4856

4857
    /* Security manager labeled all devices, therefore
J
Jiri Denemark 已提交
4858 4859 4860 4861
     * if any operation from now on fails, we need to ask the caller to
     * restore labels.
     */
    ret = -2;
4862

J
Jiri Denemark 已提交
4863
    if (incoming && incoming->fd != -1) {
4864 4865 4866 4867 4868 4869 4870
        /* 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");

J
Jiri Denemark 已提交
4871
        if (fstat(incoming->fd, &stdin_sb) < 0) {
4872
            virReportSystemError(errno,
J
Jiri Denemark 已提交
4873 4874
                                 _("cannot stat fd %d"), incoming->fd);
            goto cleanup;
4875 4876
        }
        if (S_ISFIFO(stdin_sb.st_mode) &&
J
Jiri Denemark 已提交
4877 4878 4879
            virSecurityManagerSetImageFDLabel(driver->securityManager,
                                              vm->def, incoming->fd) < 0)
            goto cleanup;
4880 4881 4882
    }

    VIR_DEBUG("Labelling done, completing handshake to child");
4883
    if (virCommandHandshakeNotify(cmd) < 0)
J
Jiri Denemark 已提交
4884
        goto cleanup;
4885 4886
    VIR_DEBUG("Handshake complete, child running");

4887
    if (rv == -1) /* The VM failed to start; tear filters before taps */
4888 4889
        virDomainConfVMNWFilterTeardown(vm);

4890
    if (rv == -1) /* The VM failed to start */
J
Jiri Denemark 已提交
4891
        goto cleanup;
4892

4893
    VIR_DEBUG("Setting cgroup for emulator (if required)");
4894
    if (qemuSetupCgroupForEmulator(vm) < 0)
J
Jiri Denemark 已提交
4895
        goto cleanup;
4896 4897 4898

    VIR_DEBUG("Setting affinity of emulator threads");
    if (qemuProcessSetEmulatorAffinity(vm) < 0)
J
Jiri Denemark 已提交
4899
        goto cleanup;
4900

4901
    VIR_DEBUG("Waiting for monitor to show up");
4902
    if (qemuProcessWaitForMonitor(driver, vm, asyncJob, priv->qemuCaps, logCtxt) < 0)
J
Jiri Denemark 已提交
4903
        goto cleanup;
4904

D
Daniel P. Berrange 已提交
4905
    /* Failure to connect to agent shouldn't be fatal */
4906 4907
    if ((rv = qemuConnectAgent(driver, vm)) < 0) {
        if (rv == -2)
J
Jiri Denemark 已提交
4908
            goto cleanup;
4909

D
Daniel P. Berrange 已提交
4910 4911 4912 4913 4914 4915
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

4916
    VIR_DEBUG("Detecting if required emulator features are present");
4917
    if (!qemuProcessVerifyGuestCPU(driver, vm, asyncJob))
J
Jiri Denemark 已提交
4918
        goto cleanup;
4919

4920
    VIR_DEBUG("Setting up post-init cgroup restrictions");
4921
    if (qemuSetupCpusetMems(vm) < 0)
J
Jiri Denemark 已提交
4922
        goto cleanup;
4923

4924
    VIR_DEBUG("Detecting VCPU PIDs");
4925
    if (qemuProcessDetectVcpuPIDs(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4926
        goto cleanup;
4927

4928 4929
    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4930
        goto cleanup;
4931

4932
    VIR_DEBUG("Setting cgroup for each VCPU (if required)");
4933
    if (qemuSetupCgroupForVcpu(vm) < 0)
J
Jiri Denemark 已提交
4934
        goto cleanup;
4935

4936 4937
    VIR_DEBUG("Setting cgroup for each IOThread (if required)");
    if (qemuSetupCgroupForIOThreads(vm) < 0)
J
Jiri Denemark 已提交
4938
        goto cleanup;
4939

4940
    VIR_DEBUG("Setting VCPU affinities");
4941
    if (qemuProcessSetVcpuAffinities(vm) < 0)
J
Jiri Denemark 已提交
4942
        goto cleanup;
4943

4944 4945
    VIR_DEBUG("Setting affinity of IOThread threads");
    if (qemuProcessSetIOThreadsAffinity(vm) < 0)
J
Jiri Denemark 已提交
4946
        goto cleanup;
4947

4948 4949
    VIR_DEBUG("Setting scheduler parameters");
    if (qemuProcessSetSchedulers(vm) < 0)
J
Jiri Denemark 已提交
4950
        goto cleanup;
4951

4952
    VIR_DEBUG("Setting any required VM passwords");
4953
    if (qemuProcessInitPasswords(conn, driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4954
        goto cleanup;
4955 4956 4957

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
4958
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4959
        VIR_DEBUG("Determining domain device PCI addresses");
4960
        if (qemuProcessInitPCIAddresses(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4961
            goto cleanup;
4962 4963
    }

4964 4965 4966 4967
    /* set default link states */
    /* qemu doesn't support setting this on the command line, so
     * enter the monitor */
    VIR_DEBUG("Setting network link states");
4968
    if (qemuProcessSetLinkStates(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4969
        goto cleanup;
4970

4971
    VIR_DEBUG("Fetching list of active devices");
4972
    if (qemuDomainUpdateDeviceList(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4973
        goto cleanup;
4974

4975 4976
    VIR_DEBUG("Updating info of memory devices");
    if (qemuDomainUpdateMemoryDeviceInfo(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4977
        goto cleanup;
4978

4979
    VIR_DEBUG("Setting initial memory amount");
4980
    if (qemuProcessSetupBalloon(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4981
        goto cleanup;
4982

4983
    /* Since CPUs were not started yet, the balloon could not return the memory
4984 4985
     * to the host and thus cur_balloon needs to be updated so that GetXMLdesc
     * and friends return the correct size in case they can't grab the job */
4986
    if (!incoming && !snapshot &&
4987
        qemuProcessRefreshBalloonState(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4988
        goto cleanup;
4989

4990 4991
    VIR_DEBUG("Detecting actual memory size for video device");
    if (qemuProcessUpdateVideoRamSize(driver, vm, asyncJob) < 0)
J
Jiri Denemark 已提交
4992 4993 4994 4995 4996 4997 4998 4999 5000 5001
        goto cleanup;

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

    ret = 0;

 cleanup:
    virCommandFree(cmd);
5002
    qemuDomainLogContextFree(logCtxt);
J
Jiri Denemark 已提交
5003 5004 5005 5006 5007 5008 5009 5010
    virObjectUnref(cfg);
    virObjectUnref(caps);
    VIR_FREE(nicindexes);
    VIR_FREE(nodeset);
    return ret;
}


5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057
/**
 * qemuProcessFinishStartup:
 *
 * Finish starting a new domain.
 */
int
qemuProcessFinishStartup(virConnectPtr conn,
                         virQEMUDriverPtr driver,
                         virDomainObjPtr vm,
                         qemuDomainAsyncJob asyncJob,
                         bool startCPUs,
                         virDomainPausedReason pausedReason)
{
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;

    if (startCPUs) {
        VIR_DEBUG("Starting domain CPUs");
        if (qemuProcessStartCPUs(driver, vm, conn,
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 asyncJob) < 0) {
            if (!virGetLastError())
                virReportError(VIR_ERR_OPERATION_FAILED, "%s",
                               _("resume operation failed"));
            goto cleanup;
        }
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, pausedReason);
    }

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

    if (qemuProcessStartHook(driver, vm,
                             VIR_HOOK_QEMU_OP_STARTED,
                             VIR_HOOK_SUBOP_BEGIN) < 0)
        goto cleanup;

    ret = 0;

 cleanup:
    virObjectUnref(cfg);
    return ret;
}


J
Jiri Denemark 已提交
5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105
int
qemuProcessStart(virConnectPtr conn,
                 virQEMUDriverPtr driver,
                 virDomainObjPtr vm,
                 qemuDomainAsyncJob asyncJob,
                 const char *migrateFrom,
                 int migrateFd,
                 const char *migratePath,
                 virDomainSnapshotObjPtr snapshot,
                 virNetDevVPortProfileOp vmop,
                 unsigned int flags)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    qemuProcessIncomingDefPtr incoming = NULL;
    unsigned int stopFlags;
    bool relabel = false;
    int ret = -1;
    int rv;

    VIR_DEBUG("conn=%p driver=%p vm=%p name=%s id=%d asyncJob=%s "
              "migrateFrom=%s migrateFd=%d migratePath=%s "
              "snapshot=%p vmop=%d flags=0x%x",
              conn, driver, vm, vm->def->name, vm->def->id,
              qemuDomainAsyncJobTypeToString(asyncJob),
              NULLSTR(migrateFrom), migrateFd, NULLSTR(migratePath),
              snapshot, vmop, flags);

    virCheckFlagsGoto(VIR_QEMU_PROCESS_START_COLD |
                      VIR_QEMU_PROCESS_START_PAUSED |
                      VIR_QEMU_PROCESS_START_AUTODESTROY, cleanup);

    if (qemuProcessInit(driver, vm, !!migrateFrom) < 0)
        goto cleanup;

    if (migrateFrom) {
        incoming = qemuProcessIncomingDefNew(priv->qemuCaps, migrateFrom,
                                             migrateFd, migratePath);
        if (!incoming)
            goto stop;
    }

    if ((rv = qemuProcessLaunch(conn, driver, vm, asyncJob, incoming,
                                snapshot, vmop, flags)) < 0) {
        if (rv == -1)
            relabel = true;
        goto stop;
    }
    relabel = true;
5106

5107 5108 5109
    if (incoming &&
        incoming->deferredURI &&
        qemuMigrationRunIncoming(driver, vm, incoming->deferredURI, asyncJob) < 0)
J
Jiri Denemark 已提交
5110
        goto stop;
5111

5112 5113 5114 5115 5116
    if (qemuProcessFinishStartup(conn, driver, vm, asyncJob,
                                 !(flags & VIR_QEMU_PROCESS_START_PAUSED),
                                 incoming ?
                                 VIR_DOMAIN_PAUSED_MIGRATION :
                                 VIR_DOMAIN_PAUSED_USER) < 0)
J
Jiri Denemark 已提交
5117
        goto stop;
5118

5119 5120
    /* Keep watching qemu log for errors during incoming migration, otherwise
     * unset reporting errors from qemu log. */
5121
    if (!incoming)
5122
        qemuMonitorSetDomainLog(priv->mon, -1, -1);
5123

5124 5125 5126
    ret = 0;

 cleanup:
5127
    qemuProcessIncomingDefFree(incoming);
5128
    return ret;
5129

J
Jiri Denemark 已提交
5130 5131 5132 5133 5134 5135
 stop:
    stopFlags = 0;
    if (!relabel)
        stopFlags |= VIR_QEMU_PROCESS_STOP_NO_RELABEL;
    if (migrateFrom)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
5136 5137
    if (priv->mon)
        qemuMonitorSetDomainLog(priv->mon, -1, -1);
J
Jiri Denemark 已提交
5138
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, stopFlags);
5139
    goto cleanup;
5140 5141 5142
}


5143
int
5144
qemuProcessKill(virDomainObjPtr vm, unsigned int flags)
5145
{
5146
    int ret;
5147

5148 5149 5150
    VIR_DEBUG("vm=%p name=%s pid=%llu flags=%x",
              vm, vm->def->name,
              (unsigned long long)vm->pid, flags);
5151

5152 5153 5154 5155 5156
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
5157 5158
    }

5159
    if (flags & VIR_QEMU_PROCESS_KILL_NOWAIT) {
5160 5161 5162 5163 5164
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
5165

5166 5167
    ret = virProcessKillPainfully(vm->pid,
                                  !!(flags & VIR_QEMU_PROCESS_KILL_FORCE));
5168

5169
    return ret;
5170 5171 5172
}


5173
void qemuProcessStop(virQEMUDriverPtr driver,
5174
                     virDomainObjPtr vm,
5175 5176
                     virDomainShutoffReason reason,
                     unsigned int flags)
5177 5178 5179 5180 5181 5182
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
5183
    virNetDevVPortProfilePtr vport = NULL;
5184
    size_t i;
5185
    char *timestamp;
5186
    char *tmppath = NULL;
5187
    char ebuf[1024];
5188
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5189
    qemuDomainLogContextPtr logCtxt = NULL;
5190

5191 5192 5193
    VIR_DEBUG("Shutting down vm=%p name=%s id=%d pid=%llu flags=%x",
              vm, vm->def->name, vm->def->id,
              (unsigned long long)vm->pid, flags);
5194 5195 5196

    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("VM '%s' not active", vm->def->name);
5197
        virObjectUnref(cfg);
5198 5199 5200
        return;
    }

5201 5202 5203 5204
    /* This method is routinely used in clean up paths. Disable error
     * reporting so we don't squash a legit error. */
    orig_err = virSaveLastError();

5205
    /*
5206 5207
     * We may unlock the vm in qemuProcessKill(), and another thread
     * can lock the vm, and then call qemuProcessStop(). So we should
5208 5209 5210 5211
     * set vm->def->id to -1 here to avoid qemuProcessStop() to be called twice.
     */
    vm->def->id = -1;

5212
    if (virAtomicIntDecAndTest(&driver->nactive) && driver->inhibitCallback)
5213 5214
        driver->inhibitCallback(false, driver->inhibitOpaque);

5215 5216
    /* Wake up anything waiting on domain condition */
    virDomainObjBroadcast(vm);
5217

5218 5219 5220
    if ((logCtxt = qemuDomainLogContextNew(driver, vm,
                                           QEMU_DOMAIN_LOG_CONTEXT_MODE_STOP))) {
        int logfile = qemuDomainLogContextGetWriteFD(logCtxt);
5221
        if ((timestamp = virTimeStringNow()) != NULL) {
5222 5223 5224 5225
            if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
                safewrite(logfile, SHUTDOWN_POSTFIX,
                          strlen(SHUTDOWN_POSTFIX)) < 0) {
                VIR_WARN("Unable to write timestamp to logfile: %s",
5226
                         virStrerror(errno, ebuf, sizeof(ebuf)));
5227 5228 5229 5230
            }

            VIR_FREE(timestamp);
        }
5231
        qemuDomainLogContextFree(logCtxt);
5232 5233
    }

5234 5235 5236
    /* Clear network bandwidth */
    virDomainClearNetBandwidth(vm);

5237 5238
    virDomainConfVMNWFilterTeardown(vm);

5239
    if (cfg->macFilter) {
5240
        def = vm->def;
5241
        for (i = 0; i < def->nnets; i++) {
5242 5243 5244
            virDomainNetDefPtr net = def->nets[i];
            if (net->ifname == NULL)
                continue;
5245 5246 5247
            ignore_value(ebtablesRemoveForwardAllowIn(driver->ebtables,
                                                      net->ifname,
                                                      &net->mac));
5248 5249 5250
        }
    }

5251
    virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
5252
    priv->nbdPort = 0;
5253

D
Daniel P. Berrange 已提交
5254 5255 5256 5257 5258 5259
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
        priv->agentError = false;
    }

5260
    if (priv->mon) {
5261
        qemuMonitorClose(priv->mon);
5262 5263
        priv->mon = NULL;
    }
5264 5265 5266 5267 5268 5269 5270 5271

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

5272 5273
    ignore_value(virAsprintf(&tmppath, "%s/domain-%s",
                             cfg->libDir, vm->def->name));
5274
    virFileDeleteTree(tmppath);
5275 5276 5277 5278
    VIR_FREE(tmppath);

    ignore_value(virAsprintf(&tmppath, "%s/domain-%s",
                             cfg->channelTargetDir, vm->def->name));
5279
    virFileDeleteTree(tmppath);
5280 5281
    VIR_FREE(tmppath);

5282 5283 5284 5285 5286 5287
    ignore_value(virDomainChrDefForeach(vm->def,
                                        false,
                                        qemuProcessCleanupChardevDevice,
                                        NULL));


5288
    /* shut it off for sure */
5289 5290 5291
    ignore_value(qemuProcessKill(vm,
                                 VIR_QEMU_PROCESS_KILL_FORCE|
                                 VIR_QEMU_PROCESS_KILL_NOCHECK));
5292

5293 5294
    qemuDomainCleanupRun(driver, vm);

5295
    /* Stop autodestroy in case guest is restarted */
5296
    qemuProcessAutoDestroyRemove(driver, vm);
5297

5298 5299
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5300
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5301 5302

        /* we can't stop the operation even if the script raised an error */
5303 5304 5305
        ignore_value(virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
                                 VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                                 NULL, xml, NULL));
5306 5307 5308
        VIR_FREE(xml);
    }

5309 5310 5311 5312
    /* Reset Security Labels unless caller don't want us to */
    if (!(flags & VIR_QEMU_PROCESS_STOP_NO_RELABEL))
        virSecurityManagerRestoreAllLabel(driver->securityManager,
                                          vm->def,
5313
                                          !!(flags & VIR_QEMU_PROCESS_STOP_MIGRATED));
5314
    virSecurityManagerReleaseLabel(driver->securityManager, vm->def);
5315

5316
    for (i = 0; i < vm->def->ndisks; i++) {
5317
        virDomainDeviceDef dev;
5318
        virDomainDiskDefPtr disk = vm->def->disks[i];
5319 5320 5321 5322

        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        ignore_value(qemuRemoveSharedDevice(driver, &dev, vm->def->name));
5323 5324
    }

5325
    /* Clear out dynamically assigned labels */
5326
    for (i = 0; i < vm->def->nseclabels; i++) {
5327
        if (vm->def->seclabels[i]->type == VIR_DOMAIN_SECLABEL_DYNAMIC)
5328 5329
            VIR_FREE(vm->def->seclabels[i]->label);
        VIR_FREE(vm->def->seclabels[i]->imagelabel);
5330 5331
    }

5332 5333 5334
    virStringFreeList(priv->qemuDevices);
    priv->qemuDevices = NULL;

5335 5336 5337
    virDomainDefClearDeviceAliases(vm->def);
    if (!priv->persistentAddrs) {
        virDomainDefClearPCIAddresses(vm->def);
5338
        virDomainPCIAddressSetFree(priv->pciaddrs);
5339
        priv->pciaddrs = NULL;
5340
        virDomainDefClearCCWAddresses(vm->def);
J
Ján Tomko 已提交
5341
        virDomainCCWAddressSetFree(priv->ccwaddrs);
5342
        priv->ccwaddrs = NULL;
5343 5344
        virDomainVirtioSerialAddrSetFree(priv->vioserialaddrs);
        priv->vioserialaddrs = NULL;
5345 5346
    }

5347
    qemuHostdevReAttachDomainDevices(driver, vm->def);
5348 5349 5350 5351

    def = vm->def;
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
5352 5353 5354 5355
        vport = virDomainNetGetActualVirtPortProfile(net);

        switch (virDomainNetGetActualType(net)) {
        case VIR_DOMAIN_NET_TYPE_DIRECT:
5356
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
5357
                             net->ifname, &net->mac,
5358 5359
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
5360
                             virDomainNetGetActualVirtPortProfile(net),
5361
                             cfg->stateDir));
5362 5363 5364 5365 5366
            break;
        case VIR_DOMAIN_NET_TYPE_BRIDGE:
        case VIR_DOMAIN_NET_TYPE_NETWORK:
#ifdef VIR_NETDEV_TAP_REQUIRE_MANUAL_CLEANUP
            if (!(vport && vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH))
5367
                ignore_value(virNetDevTapDelete(net->ifname, net->backend.tap));
5368 5369
#endif
            break;
5370
        }
5371 5372 5373
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
5374 5375 5376 5377 5378 5379 5380 5381 5382
        if (vport) {
            if (vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_MIDONET) {
                ignore_value(virNetDevMidonetUnbindPort(vport));
            } else if (vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
                ignore_value(virNetDevOpenvswitchRemovePort(
                                 virDomainNetGetActualBridgeName(net),
                                 net->ifname));
            }
        }
A
Ansis Atteka 已提交
5383

5384 5385
        /* kick the device out of the hostdev list too */
        virDomainNetRemoveHostdev(def, net);
5386
        networkReleaseActualDevice(vm->def, net);
5387
    }
5388

5389
 retry:
5390
    if ((ret = qemuRemoveCgroup(driver, vm)) < 0) {
5391 5392 5393 5394 5395 5396 5397
        if (ret == -EBUSY && (retries++ < 5)) {
            usleep(200*1000);
            goto retry;
        }
        VIR_WARN("Failed to remove cgroup for %s",
                 vm->def->name);
    }
5398
    virCgroupFree(&priv->cgroup);
5399 5400 5401

    qemuProcessRemoveDomainStatus(driver, vm);

5402 5403
    /* Remove VNC and Spice ports from port reservation bitmap, but only if
       they were reserved by the driver (autoport=yes)
5404
    */
5405
    for (i = 0; i < vm->def->ngraphics; ++i) {
5406
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
5407 5408
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (graphics->data.vnc.autoport) {
5409 5410
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.vnc.port);
5411
            } else if (graphics->data.vnc.portReserved) {
5412 5413 5414 5415 5416
                virPortAllocatorSetUsed(driver->remotePorts,
                                        graphics->data.spice.port,
                                        false);
                graphics->data.vnc.portReserved = false;
            }
5417 5418
            virPortAllocatorRelease(driver->webSocketPorts,
                                    graphics->data.vnc.websocket);
5419
        }
5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (graphics->data.spice.autoport) {
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.spice.port);
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.spice.tlsPort);
            } else {
                if (graphics->data.spice.portReserved) {
                    virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.port,
                                            false);
                    graphics->data.spice.portReserved = false;
                }

                if (graphics->data.spice.tlsPortReserved) {
                    virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.tlsPort,
                                            false);
                    graphics->data.spice.tlsPortReserved = false;
                }
            }
5441
        }
5442 5443
    }

5444
    vm->taint = 0;
5445
    vm->pid = -1;
J
Jiri Denemark 已提交
5446
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
5447 5448
    VIR_FREE(priv->vcpupids);
    priv->nvcpupids = 0;
5449 5450
    for (i = 0; i < vm->def->niothreadids; i++)
        vm->def->iothreadids[i]->thread_id = 0;
5451 5452
    virObjectUnref(priv->qemuCaps);
    priv->qemuCaps = NULL;
5453
    VIR_FREE(priv->pidfile);
5454

5455
    /* The "release" hook cleans up additional resources */
5456
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5457
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5458 5459 5460

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
5461 5462
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
5463 5464 5465
        VIR_FREE(xml);
    }

5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476
    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);
    }
5477
    virObjectUnref(cfg);
5478
}
5479 5480


5481
int qemuProcessAttach(virConnectPtr conn ATTRIBUTE_UNUSED,
5482
                      virQEMUDriverPtr driver,
5483
                      virDomainObjPtr vm,
5484
                      pid_t pid,
5485 5486 5487 5488
                      const char *pidfile,
                      virDomainChrSourceDefPtr monConfig,
                      bool monJSON)
{
5489
    size_t i;
5490 5491
    char ebuf[1024];
    int logfile = -1;
5492
    qemuDomainLogContextPtr logCtxt = NULL;
5493 5494 5495
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool running = true;
5496
    virDomainPausedReason reason;
5497
    virSecurityLabelPtr seclabel = NULL;
5498
    virSecurityLabelDefPtr seclabeldef = NULL;
5499
    bool seclabelgen = false;
5500 5501
    virSecurityManagerPtr* sec_managers = NULL;
    const char *model;
5502
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5503
    virCapsPtr caps = NULL;
5504
    bool active = false;
5505
    int ret;
5506 5507 5508 5509

    VIR_DEBUG("Beginning VM attach process");

    if (virDomainObjIsActive(vm)) {
5510 5511
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
5512
        virObjectUnref(cfg);
5513 5514 5515
        return -1;
    }

5516
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
5517
        goto error;
5518

5519 5520 5521 5522 5523
    /* 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");
5524
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm, true) < 0)
5525
        goto error;
5526

5527
    vm->def->id = qemuDriverAllocateID(driver);
5528

5529
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
5530
        driver->inhibitCallback(true, driver->inhibitOpaque);
5531
    active = true;
5532

5533
    if (virFileMakePath(cfg->logDir) < 0) {
5534 5535
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
5536
                             cfg->logDir);
5537
        goto error;
5538 5539 5540
    }

    VIR_FREE(priv->pidfile);
5541
    if (VIR_STRDUP(priv->pidfile, pidfile) < 0)
5542
        goto error;
5543

5544 5545
    vm->pid = pid;

5546
    VIR_DEBUG("Detect security driver config");
5547
    sec_managers = virSecurityManagerGetNested(driver->securityManager);
5548 5549
    if (sec_managers == NULL)
        goto error;
5550 5551

    for (i = 0; sec_managers[i]; i++) {
5552
        seclabelgen = false;
5553 5554
        model = virSecurityManagerGetModel(sec_managers[i]);
        seclabeldef = virDomainDefGetSecurityLabelDef(vm->def, model);
5555
        if (seclabeldef == NULL) {
5556
            if (!(seclabeldef = virSecurityLabelDefNew(model)))
5557 5558 5559
                goto error;
            seclabelgen = true;
        }
5560 5561
        seclabeldef->type = VIR_DOMAIN_SECLABEL_STATIC;
        if (VIR_ALLOC(seclabel) < 0)
5562
            goto error;
5563
        if (virSecurityManagerGetProcessLabel(sec_managers[i],
5564
                                              vm->def, vm->pid, seclabel) < 0)
5565
            goto error;
5566

5567
        if (VIR_STRDUP(seclabeldef->model, model) < 0)
5568
            goto error;
5569

5570
        if (VIR_STRDUP(seclabeldef->label, seclabel->label) < 0)
5571
            goto error;
5572
        VIR_FREE(seclabel);
5573 5574 5575 5576 5577 5578

        if (seclabelgen) {
            if (VIR_APPEND_ELEMENT(vm->def->seclabels, vm->def->nseclabels, seclabeldef) < 0)
                goto error;
            seclabelgen = false;
        }
5579
    }
5580

5581 5582
    if (virSecurityManagerCheckAllLabel(driver->securityManager, vm->def) < 0)
        goto error;
5583 5584 5585
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0)
        goto error;

5586
    VIR_DEBUG("Creating domain log file");
5587 5588
    if (!(logCtxt = qemuDomainLogContextNew(driver, vm,
                                            QEMU_DOMAIN_LOG_CONTEXT_MODE_ATTACH)))
5589
        goto error;
5590
    logfile = qemuDomainLogContextGetWriteFD(logCtxt);
5591 5592

    VIR_DEBUG("Determining emulator version");
5593 5594
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
5595 5596
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
5597
        goto error;
5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609

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

    priv->gotShutdown = false;

    /*
     * Normally PCI addresses are assigned in the virDomainCreate
     * or virDomainDefine methods. We might still need to assign
     * some here to cope with the question of upgrades. Regardless
M
Martin Kletzander 已提交
5610
     * we also need to populate the PCI address set cache for later
5611 5612
     * use in hotplug
     */
5613
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
5614
        VIR_DEBUG("Assigning domain PCI addresses");
5615
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
5616
            goto error;
5617 5618
    }

5619
    if ((timestamp = virTimeStringNow()) == NULL)
5620
        goto error;
5621

5622 5623 5624 5625
    if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
        safewrite(logfile, ATTACH_POSTFIX, strlen(ATTACH_POSTFIX)) < 0) {
        VIR_WARN("Unable to write timestamp to logfile: %s",
                 virStrerror(errno, ebuf, sizeof(ebuf)));
5626
    }
5627
    VIR_FREE(timestamp);
5628

5629
    qemuDomainObjTaint(driver, vm, VIR_DOMAIN_TAINT_EXTERNAL_LAUNCH, logCtxt);
5630 5631

    VIR_DEBUG("Waiting for monitor to show up");
5632
    if (qemuProcessWaitForMonitor(driver, vm, QEMU_ASYNC_JOB_NONE, priv->qemuCaps, NULL) < 0)
5633
        goto error;
5634

D
Daniel P. Berrange 已提交
5635
    /* Failure to connect to agent shouldn't be fatal */
5636 5637 5638 5639
    if ((ret = qemuConnectAgent(driver, vm)) < 0) {
        if (ret == -2)
            goto error;

D
Daniel P. Berrange 已提交
5640 5641 5642 5643 5644 5645
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

5646
    VIR_DEBUG("Detecting VCPU PIDs");
5647
    if (qemuProcessDetectVcpuPIDs(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5648 5649 5650 5651
        goto error;

    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5652
        goto error;
5653 5654 5655

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
5656
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
5657
        VIR_DEBUG("Determining domain device PCI addresses");
5658
        if (qemuProcessInitPCIAddresses(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5659
            goto error;
5660 5661 5662
    }

    VIR_DEBUG("Getting initial memory amount");
5663
    qemuDomainObjEnterMonitor(driver, vm);
5664 5665 5666 5667 5668 5669 5670
    if (qemuMonitorGetBalloonInfo(priv->mon, &vm->def->mem.cur_balloon) < 0)
        goto exit_monitor;
    if (qemuMonitorGetStatus(priv->mon, &running, &reason) < 0)
        goto exit_monitor;
    if (qemuMonitorGetVirtType(priv->mon, &vm->def->virtType) < 0)
        goto exit_monitor;
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5671
        goto error;
5672

5673
    if (running) {
5674 5675
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                             VIR_DOMAIN_RUNNING_UNPAUSED);
5676 5677 5678
        if (vm->def->memballoon &&
            vm->def->memballoon->model == VIR_DOMAIN_MEMBALLOON_MODEL_VIRTIO &&
            vm->def->memballoon->period) {
5679 5680 5681
            qemuDomainObjEnterMonitor(driver, vm);
            qemuMonitorSetMemoryStatsPeriod(priv->mon,
                                            vm->def->memballoon->period);
5682 5683
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                goto error;
5684 5685
        }
    } else {
5686
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
5687
    }
5688 5689

    VIR_DEBUG("Writing domain status to disk");
5690
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
5691
        goto error;
5692

5693 5694
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5695
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706
        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)
5707
            goto error;
5708 5709
    }

5710
    qemuDomainLogContextFree(logCtxt);
5711
    VIR_FREE(seclabel);
5712
    VIR_FREE(sec_managers);
5713
    virObjectUnref(cfg);
5714
    virObjectUnref(caps);
5715 5716 5717

    return 0;

5718 5719
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
5720
 error:
5721 5722 5723 5724 5725 5726
    /* We jump here if we failed to attach to the VM for any reason.
     * Leave the domain running, but pretend we never attempted to
     * attach to it.  */
    if (active && virAtomicIntDecAndTest(&driver->nactive) &&
        driver->inhibitCallback)
        driver->inhibitCallback(false, driver->inhibitOpaque);
5727
    qemuDomainLogContextFree(logCtxt);
5728
    VIR_FREE(seclabel);
5729
    VIR_FREE(sec_managers);
5730 5731
    if (seclabelgen)
        virSecurityLabelDefFree(seclabeldef);
5732
    virDomainChrSourceDefFree(monConfig);
5733
    virObjectUnref(cfg);
5734
    virObjectUnref(caps);
5735 5736 5737 5738
    return -1;
}


5739
static virDomainObjPtr
5740 5741 5742
qemuProcessAutoDestroy(virDomainObjPtr dom,
                       virConnectPtr conn,
                       void *opaque)
5743
{
5744
    virQEMUDriverPtr driver = opaque;
5745
    qemuDomainObjPrivatePtr priv = dom->privateData;
5746
    virObjectEventPtr event = NULL;
5747
    unsigned int stopFlags = 0;
5748

5749
    VIR_DEBUG("vm=%s, conn=%p", dom->def->name, conn);
5750

5751 5752
    virObjectRef(dom);

5753 5754 5755
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;

5756 5757
    if (priv->job.asyncJob) {
        VIR_DEBUG("vm=%s has long-term job active, cancelling",
5758
                  dom->def->name);
5759
        qemuDomainObjDiscardAsyncJob(driver, dom);
5760 5761
    }

5762 5763
    if (qemuDomainObjBeginJob(driver, dom,
                              QEMU_JOB_DESTROY) < 0)
5764 5765 5766
        goto cleanup;

    VIR_DEBUG("Killing domain");
5767

5768
    qemuProcessStop(driver, dom, VIR_DOMAIN_SHUTOFF_DESTROYED, stopFlags);
5769

5770
    virDomainAuditStop(dom, "destroyed");
5771
    event = virDomainEventLifecycleNewFromObj(dom,
5772 5773
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_DESTROYED);
5774

5775 5776
    qemuDomainObjEndJob(driver, dom);

5777
    qemuDomainRemoveInactive(driver, dom);
5778

5779
    qemuDomainEventQueue(driver, event);
5780

5781
 cleanup:
5782
    virDomainObjEndAPI(&dom);
5783
    return dom;
5784 5785
}

5786
int qemuProcessAutoDestroyAdd(virQEMUDriverPtr driver,
5787 5788 5789
                              virDomainObjPtr vm,
                              virConnectPtr conn)
{
5790
    VIR_DEBUG("vm=%s, conn=%p", vm->def->name, conn);
5791 5792
    return virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                qemuProcessAutoDestroy);
5793 5794
}

5795
int qemuProcessAutoDestroyRemove(virQEMUDriverPtr driver,
5796 5797
                                 virDomainObjPtr vm)
{
5798
    int ret;
5799
    VIR_DEBUG("vm=%s", vm->def->name);
5800 5801 5802
    ret = virCloseCallbacksUnset(driver->closeCallbacks, vm,
                                 qemuProcessAutoDestroy);
    return ret;
5803
}
5804

5805
bool qemuProcessAutoDestroyActive(virQEMUDriverPtr driver,
5806 5807
                                  virDomainObjPtr vm)
{
5808
    virCloseCallback cb;
5809
    VIR_DEBUG("vm=%s", vm->def->name);
5810
    cb = virCloseCallbacksGet(driver->closeCallbacks, vm, NULL);
5811
    return cb == qemuProcessAutoDestroy;
5812
}