qemu_driver.c 231.9 KB
Newer Older
D
Daniel P. Berrange 已提交
1 2 3
/*
 * driver.c: core driver methods for managing qemu guests
 *
4
 * Copyright (C) 2006, 2007, 2008, 2009 Red Hat, Inc.
D
Daniel P. Berrange 已提交
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
 * Copyright (C) 2006 Daniel P. Berrange
 *
 * 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
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307  USA
 *
 * Author: Daniel P. Berrange <berrange@redhat.com>
 */

24
#include <config.h>
25

D
Daniel P. Berrange 已提交
26 27
#include <sys/types.h>
#include <sys/poll.h>
28
#include <sys/time.h>
D
Daniel P. Berrange 已提交
29 30 31
#include <dirent.h>
#include <limits.h>
#include <string.h>
32
#include <stdbool.h>
D
Daniel P. Berrange 已提交
33 34 35 36 37 38
#include <stdio.h>
#include <strings.h>
#include <stdarg.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
39
#include <sys/utsname.h>
40 41 42 43
#include <sys/stat.h>
#include <fcntl.h>
#include <signal.h>
#include <paths.h>
44 45
#include <pwd.h>
#include <stdio.h>
46
#include <sys/wait.h>
47
#include <sys/ioctl.h>
48
#include <sys/un.h>
D
Daniel P. Berrange 已提交
49

50
#include "virterror_internal.h"
51
#include "logging.h"
52
#include "datatypes.h"
53 54
#include "qemu_driver.h"
#include "qemu_conf.h"
55
#include "qemu_monitor.h"
56
#include "qemu_monitor_text.h"
57
#include "qemu_bridge_filter.h"
J
Jim Meyering 已提交
58
#include "c-ctype.h"
59
#include "event.h"
60
#include "buf.h"
61
#include "util.h"
62
#include "nodeinfo.h"
63
#include "stats_linux.h"
64
#include "capabilities.h"
65
#include "memory.h"
66
#include "uuid.h"
67
#include "domain_conf.h"
68 69
#include "node_device_conf.h"
#include "pci.h"
70
#include "hostusb.h"
71
#include "processinfo.h"
72
#include "security/security_driver.h"
73
#include "cgroup.h"
C
Chris Lalancette 已提交
74
#include "libvirt_internal.h"
75

76

77 78
#define VIR_FROM_THIS VIR_FROM_QEMU

79 80 81
typedef struct _qemuDomainObjPrivate qemuDomainObjPrivate;
typedef qemuDomainObjPrivate *qemuDomainObjPrivatePtr;
struct _qemuDomainObjPrivate {
82 83 84 85 86
    virCond jobCond; /* Use in conjunction with main virDomainObjPtr lock */
    int jobActive; /* Non-zero if a job is active. Only 1 job is allowed at any time
                    * A job includes *all* monitor commands, even those just querying
                    * information, not merely actions */

87 88 89
    qemuMonitorPtr mon;
};

90 91
static int qemudShutdown(void);

92 93
static void qemuDriverLock(struct qemud_driver *driver)
{
94
    virMutexLock(&driver->lock);
95 96 97
}
static void qemuDriverUnlock(struct qemud_driver *driver)
{
98
    virMutexUnlock(&driver->lock);
99 100
}

101 102 103
static void qemuDomainEventFlush(int timer, void *opaque);
static void qemuDomainEventQueue(struct qemud_driver *driver,
                                 virDomainEventPtr event);
104

105 106
static int qemudStartVMDaemon(virConnectPtr conn,
                              struct qemud_driver *driver,
107
                              virDomainObjPtr vm,
108 109
                              const char *migrateFrom,
                              int stdin_fd);
110

111 112
static void qemudShutdownVMDaemon(virConnectPtr conn,
                                  struct qemud_driver *driver,
113
                                  virDomainObjPtr vm);
114

115
static int qemudDomainGetMaxVcpus(virDomainPtr dom);
116

117
static int qemuDetectVcpuPIDs(virConnectPtr conn,
118
                              struct qemud_driver *driver,
119
                              virDomainObjPtr vm);
120

121 122 123
static int qemuUpdateActivePciHostdevs(struct qemud_driver *driver,
                                       virDomainDefPtr def);

J
Jim Meyering 已提交
124
static struct qemud_driver *qemu_driver = NULL;
125

126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149

static void *qemuDomainObjPrivateAlloc(void)
{
    qemuDomainObjPrivatePtr priv;

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

    return priv;
}

static void qemuDomainObjPrivateFree(void *data)
{
    qemuDomainObjPrivatePtr priv = data;

    /* This should never be non-NULL if we get here, but just in case... */
    if (priv->mon) {
        VIR_ERROR0("Unexpected QEMU monitor still active during domain deletion");
        qemuMonitorClose(priv->mon);
    }
    VIR_FREE(priv);
}


150 151 152 153 154 155 156 157 158
/*
 * obj must be locked before calling, qemud_driver must NOT be locked
 *
 * This must be called by anything that will change the VM state
 * in any way, or anything that will use the QEMU monitor.
 *
 * Upon successful return, the object will have its ref count increased,
 * successful calls must be followed by EndJob eventually
 */
159 160 161 162

/* Give up waiting for mutex after 30 seconds */
#define QEMU_JOB_WAIT_TIME (1000ull * 30)

163 164 165 166
static int qemuDomainObjBeginJob(virDomainObjPtr obj) ATTRIBUTE_RETURN_CHECK;
static int qemuDomainObjBeginJob(virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
167 168 169 170 171 172 173 174 175 176
    struct timeval now;
    unsigned long long then;

    if (gettimeofday(&now, NULL) < 0) {
        virReportSystemError(NULL, errno, "%s",
                             _("cannot get time of day"));
        return -1;
    }
    then = (now.tv_sec * 1000ull) + (now.tv_usec / 1000);
    then += QEMU_JOB_WAIT_TIME;
177 178 179 180

    virDomainObjRef(obj);

    while (priv->jobActive) {
181
        if (virCondWaitUntil(&priv->jobCond, &obj->lock, then) < 0) {
182
            virDomainObjUnref(obj);
183 184 185 186 187 188
            if (errno == ETIMEDOUT)
                qemudReportError(NULL, NULL, NULL, VIR_ERR_OPERATION_TIMEOUT,
                                 "%s", _("cannot acquire state change lock"));
            else
                virReportSystemError(NULL, errno,
                                     "%s", _("cannot acquire job mutex"));
189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208
            return -1;
        }
    }
    priv->jobActive = 1;

    return 0;
}

/*
 * obj must be locked before calling, qemud_driver must be locked
 *
 * This must be called by anything that will change the VM state
 * in any way, or anything that will use the QEMU monitor.
 */
static int qemuDomainObjBeginJobWithDriver(struct qemud_driver *driver,
                                           virDomainObjPtr obj) ATTRIBUTE_RETURN_CHECK;
static int qemuDomainObjBeginJobWithDriver(struct qemud_driver *driver,
                                           virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;
209 210 211 212 213 214 215 216 217 218
    struct timeval now;
    unsigned long long then;

    if (gettimeofday(&now, NULL) < 0) {
        virReportSystemError(NULL, errno, "%s",
                             _("cannot get time of day"));
        return -1;
    }
    then = (now.tv_sec * 1000ull) + (now.tv_usec / 1000);
    then += QEMU_JOB_WAIT_TIME;
219 220 221 222 223

    virDomainObjRef(obj);
    qemuDriverUnlock(driver);

    while (priv->jobActive) {
224
        if (virCondWaitUntil(&priv->jobCond, &obj->lock, then) < 0) {
225
            virDomainObjUnref(obj);
226 227 228 229 230 231
            if (errno == ETIMEDOUT)
                qemudReportError(NULL, NULL, NULL, VIR_ERR_OPERATION_TIMEOUT,
                                 "%s", _("cannot acquire state change lock"));
            else
                virReportSystemError(NULL, errno,
                                     "%s", _("cannot acquire job mutex"));
232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268
            return -1;
        }
    }
    priv->jobActive = 1;

    virDomainObjUnlock(obj);
    qemuDriverLock(driver);
    virDomainObjLock(obj);

    return 0;
}

/*
 * obj must be locked before calling, qemud_driver does not matter
 *
 * To be called after completing the work associated with the
 * earlier  qemuDomainBeginJob() call
 */
static void qemuDomainObjEndJob(virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    priv->jobActive = 0;
    virCondSignal(&priv->jobCond);

    virDomainObjUnref(obj);
}


/*
 * obj must be locked before calling, qemud_driver must be unlocked
 *
 * To be called immediately before any QEMU monitor API call
 * Must have alrady called qemuDomainObjBeginJob().
 *
 * To be followed with qemuDomainObjExitMonitor() once complete
 */
269 270 271 272 273
static void qemuDomainObjEnterMonitor(virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    qemuMonitorLock(priv->mon);
274
    virDomainObjUnlock(obj);
275 276 277
}


278 279 280 281
/* obj must NOT be locked before calling, qemud_driver must be unlocked
 *
 * Should be paired with an earlier  qemuDomainObjEnterMonitor() call
 */
282 283 284 285 286
static void qemuDomainObjExitMonitor(virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    qemuMonitorUnlock(priv->mon);
287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320
    virDomainObjLock(obj);
}


/*
 * obj must be locked before calling, qemud_driver must be locked
 *
 * To be called immediately before any QEMU monitor API call
 * Must have alrady called qemuDomainObjBeginJob().
 *
 * To be followed with qemuDomainObjExitMonitorWithDriver() once complete
 */
static void qemuDomainObjEnterMonitorWithDriver(struct qemud_driver *driver, virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    qemuMonitorLock(priv->mon);
    virDomainObjUnlock(obj);
    qemuDriverUnlock(driver);
}


/* obj must NOT be locked before calling, qemud_driver must be unlocked,
 * and will be locked after returning
 *
 * Should be paired with an earlier  qemuDomainObjEnterMonitor() call
 */
static void qemuDomainObjExitMonitorWithDriver(struct qemud_driver *driver, virDomainObjPtr obj)
{
    qemuDomainObjPrivatePtr priv = obj->privateData;

    qemuMonitorUnlock(priv->mon);
    qemuDriverLock(driver);
    virDomainObjLock(obj);
321 322 323
}


324 325 326 327 328 329 330 331 332
static int qemuCgroupControllerActive(struct qemud_driver *driver,
                                      int controller)
{
    if (driver->cgroup == NULL)
        return 0;
    if (driver->cgroupControllers & (1 << controller))
        return 1;
    return 0;
}
333

334
static int
335
qemudLogFD(virConnectPtr conn, struct qemud_driver *driver, const char* name)
336 337 338
{
    char logfile[PATH_MAX];
    mode_t logmode;
G
Guido Günther 已提交
339
    int ret, fd = -1;
340

341 342
    if ((ret = snprintf(logfile, sizeof(logfile), "%s/%s.log",
                        driver->logDir, name))
G
Guido Günther 已提交
343
        < 0 || ret >= sizeof(logfile)) {
344
        virReportOOMError(conn);
345 346 347 348
        return -1;
    }

    logmode = O_CREAT | O_WRONLY;
349 350
    /* Only logrotate files in /var/log, so only append if running privileged */
    if (driver->privileged)
351
        logmode |= O_APPEND;
352 353 354
    else
        logmode |= O_TRUNC;

355
    if ((fd = open(logfile, logmode, S_IRUSR | S_IWUSR)) < 0) {
356 357 358
        virReportSystemError(conn, errno,
                             _("failed to create logfile %s"),
                             logfile);
359 360
        return -1;
    }
361
    if (virSetCloseExec(fd) < 0) {
362 363
        virReportSystemError(conn, errno, "%s",
                             _("Unable to set VM logfile close-on-exec flag"));
364 365 366 367 368 369 370
        close(fd);
        return -1;
    }
    return fd;
}


371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
static int
qemudLogReadFD(virConnectPtr conn, const char* logDir, const char* name, off_t pos)
{
    char logfile[PATH_MAX];
    mode_t logmode = O_RDONLY;
    int ret, fd = -1;

    if ((ret = snprintf(logfile, sizeof(logfile), "%s/%s.log", logDir, name))
        < 0 || ret >= sizeof(logfile)) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("failed to build logfile name %s/%s.log"),
                         logDir, name);
        return -1;
    }


    if ((fd = open(logfile, logmode)) < 0) {
388 389 390
        virReportSystemError(conn, errno,
                             _("failed to create logfile %s"),
                             logfile);
391 392
        return -1;
    }
393
    if (virSetCloseExec(fd) < 0) {
394 395
        virReportSystemError(conn, errno, "%s",
                             _("Unable to set VM logfile close-on-exec flag"));
396 397 398 399
        close(fd);
        return -1;
    }
    if (lseek(fd, pos, SEEK_SET) < 0) {
400 401 402
        virReportSystemError(conn, errno,
                             _("Unable to seek to %lld in %s"),
                             (long long) pos, logfile);
403 404 405 406 407 408
        close(fd);
    }
    return fd;
}


409 410 411 412 413 414 415 416 417 418 419 420
struct qemuAutostartData {
    struct qemud_driver *driver;
    virConnectPtr conn;
};
static void
qemuAutostartDomain(void *payload, const char *name ATTRIBUTE_UNUSED, void *opaque)
{
    virDomainObjPtr vm = payload;
    struct qemuAutostartData *data = opaque;

    virDomainObjLock(vm);
    if (vm->autostart &&
D
Daniel P. Berrange 已提交
421
        !virDomainObjIsActive(vm)) {
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442
        int ret;

        virResetLastError();
        ret = qemudStartVMDaemon(data->conn, data->driver, vm, NULL, -1);
        if (ret < 0) {
            virErrorPtr err = virGetLastError();
            VIR_ERROR(_("Failed to autostart VM '%s': %s\n"),
                      vm->def->name,
                      err ? err->message : "");
        } else {
            virDomainEventPtr event =
                virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_BOOTED);
            if (event)
                qemuDomainEventQueue(data->driver, event);
        }
    }
    virDomainObjUnlock(vm);
}

443 444
static void
qemudAutostartConfigs(struct qemud_driver *driver) {
445 446 447 448 449
    /* XXX: Figure out a better way todo this. The domain
     * startup code needs a connection handle in order
     * to lookup the bridge associated with a virtual
     * network
     */
450 451 452
    virConnectPtr conn = virConnectOpen(driver->privileged ?
                                        "qemu:///system" :
                                        "qemu:///session");
453
    /* Ignoring NULL conn which is mostly harmless here */
454
    struct qemuAutostartData data = { driver, conn };
455

456
    qemuDriverLock(driver);
457
    virHashForEach(driver->domains.objs, qemuAutostartDomain, &data);
458
    qemuDriverUnlock(driver);
459

460 461
    if (conn)
        virConnectClose(conn);
462 463
}

464 465 466 467 468 469 470 471 472 473 474 475 476

/**
 * qemudRemoveDomainStatus
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
qemudRemoveDomainStatus(virConnectPtr conn,
                        struct qemud_driver *driver,
                        virDomainObjPtr vm)
{
477
    char ebuf[1024];
478 479 480
    char *file = NULL;

    if (virAsprintf(&file, "%s/%s.xml", driver->stateDir, vm->def->name) < 0) {
481
        virReportOOMError(conn);
D
Daniel Veillard 已提交
482
        return(-1);
483 484
    }

485 486
    if (unlink(file) < 0 && errno != ENOENT && errno != ENOTDIR)
        VIR_WARN(_("Failed to remove domain XML for %s: %s"),
D
Daniel Veillard 已提交
487 488 489
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));
    VIR_FREE(file);

490 491 492
    if (virFileDeletePid(driver->stateDir, vm->def->name) != 0)
        VIR_WARN(_("Failed to remove PID file for %s: %s"),
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));
493

D
Daniel Veillard 已提交
494

495
    return 0;
496 497
}

498 499 500 501 502 503 504 505 506 507 508 509 510 511

/*
 * This is a callback registered with a qemuMonitorPtr  instance,
 * 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
qemuHandleMonitorEOF(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                     virDomainObjPtr vm,
                     int hasError) {
    struct qemud_driver *driver = qemu_driver;
    virDomainEventPtr event = NULL;

512
    VIR_DEBUG("Received EOF on %p '%s'", vm, vm->def->name);
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534
    virDomainObjLock(vm);

    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     hasError ?
                                     VIR_DOMAIN_EVENT_STOPPED_FAILED :
                                     VIR_DOMAIN_EVENT_STOPPED_SHUTDOWN);

    qemudShutdownVMDaemon(NULL, driver, vm);
    if (!vm->persistent)
        virDomainRemoveInactive(&driver->domains, vm);
    else
        virDomainObjUnlock(vm);

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


535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576
static virStorageEncryptionPtr
findDomainDiskEncryption(virConnectPtr conn, virDomainObjPtr vm,
                         const char *path)
{
    bool seen_volume;
    int i;

    seen_volume = false;
    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk;

        disk = vm->def->disks[i];
        if (disk->src != NULL && STREQ(disk->src, path)) {
            seen_volume = true;
            if (disk->encryption != NULL)
                return disk->encryption;
        }
    }
    if (seen_volume)
        qemudReportError(conn, NULL, NULL, VIR_ERR_INVALID_DOMAIN,
                         _("missing <encryption> for volume %s"), path);
    else
        qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("unexpected passphrase request for volume %s"),
                         path);
    return NULL;
}


static int
findVolumeQcowPassphrase(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virConnectPtr conn,
                         virDomainObjPtr vm,
                         const char *path,
                         char **secretRet,
                         size_t *secretLen)
{
    virStorageEncryptionPtr enc;
    virSecretPtr secret;
    char *passphrase;
    unsigned char *data;
    size_t size;
577 578
    int ret = -1;

579
    virDomainObjLock(vm);
580 581 582 583

    if (!conn) {
        qemudReportError(NULL, NULL, NULL, VIR_ERR_NO_SUPPORT,
                         "%s", _("cannot find secrets without a connection"));
584
        goto cleanup;
585 586 587 588 589 590 591
    }

    if (conn->secretDriver == NULL ||
        conn->secretDriver->lookupByUUID == NULL ||
        conn->secretDriver->getValue == NULL) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_NO_SUPPORT, "%s",
                         _("secret storage not supported"));
592
        goto cleanup;
593 594 595 596 597 598 599 600 601 602 603 604
    }

    enc = findDomainDiskEncryption(conn, vm, path);
    if (enc == NULL)
        return -1;

    if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW ||
        enc->nsecrets != 1 ||
        enc->secrets[0]->type !=
        VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_INVALID_DOMAIN,
                         _("invalid <encryption> for volume %s"), path);
605
        goto cleanup;
606 607 608 609 610
    }

    secret = conn->secretDriver->lookupByUUID(conn,
                                              enc->secrets[0]->uuid);
    if (secret == NULL)
611
        goto cleanup;
612 613 614 615
    data = conn->secretDriver->getValue(secret, &size,
                                        VIR_SECRET_GET_VALUE_INTERNAL_CALL);
    virUnrefSecret(secret);
    if (data == NULL)
616
        goto cleanup;
617 618 619 620 621 622 623

    if (memchr(data, '\0', size) != NULL) {
        memset(data, 0, size);
        VIR_FREE(data);
        qemudReportError(conn, NULL, NULL, VIR_ERR_INVALID_SECRET,
                         _("format='qcow' passphrase for %s must not contain a "
                           "'\\0'"), path);
624
        goto cleanup;
625 626 627 628 629 630
    }

    if (VIR_ALLOC_N(passphrase, size + 1) < 0) {
        memset(data, 0, size);
        VIR_FREE(data);
        virReportOOMError(conn);
631
        goto cleanup;
632 633 634 635 636 637 638 639 640 641
    }
    memcpy(passphrase, data, size);
    passphrase[size] = '\0';

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

    *secretRet = passphrase;
    *secretLen = size;

642 643 644
    ret = 0;

cleanup:
645 646
    virDomainObjUnlock(vm);

647
    return ret;
648 649
}

650
static int
651
qemuConnectMonitor(virDomainObjPtr vm)
652
{
653
    qemuDomainObjPrivatePtr priv = vm->privateData;
654

655
    if ((priv->mon = qemuMonitorOpen(vm, qemuHandleMonitorEOF)) == NULL) {
656
        VIR_ERROR(_("Failed to connect monitor for %s\n"), vm->def->name);
657
        return -1;
658
    }
659

660 661 662
    qemuMonitorRegisterDiskSecretLookup(priv->mon,
                                        findVolumeQcowPassphrase);

663 664
    return 0;
}
665 666 667 668

/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
669
 */
670 671
static void
qemuReconnectDomain(void *payload, const char *name ATTRIBUTE_UNUSED, void *opaque)
672
{
673 674 675 676
    virDomainObjPtr obj = payload;
    struct qemud_driver *driver = opaque;

    virDomainObjLock(obj);
677

678 679 680 681
    VIR_DEBUG("Reconnect monitor to %p '%s'", obj, obj->def->name);

    /* XXX check PID liveliness & EXE path */
    if (qemuConnectMonitor(obj) < 0)
682
        goto error;
683

684 685 686 687
    if (qemuUpdateActivePciHostdevs(driver, obj->def) < 0) {
        goto error;
    }

688 689 690 691
    if (obj->def->seclabel.type == VIR_DOMAIN_SECLABEL_DYNAMIC &&
        driver->securityDriver &&
        driver->securityDriver->domainReserveSecurityLabel &&
        driver->securityDriver->domainReserveSecurityLabel(NULL, obj) < 0)
692
        goto error;
693

694 695
    if (obj->def->id >= driver->nextvmid)
        driver->nextvmid = obj->def->id + 1;
696

697 698
    virDomainObjUnlock(obj);
    return;
699

700
error:
701 702 703 704
    /* 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 */
    qemudShutdownVMDaemon(NULL, driver, obj);
705 706 707 708
    if (!obj->persistent)
        virDomainRemoveInactive(&driver->domains, obj);
    else
        virDomainObjUnlock(obj);
709
}
710

711
/**
712
 * qemudReconnectDomains
713 714 715 716 717 718 719
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
static void
qemuReconnectDomains(struct qemud_driver *driver)
{
720
    virHashForEach(driver->domains.objs, qemuReconnectDomain, driver);
721 722
}

723 724 725 726 727 728 729 730 731 732 733 734

static int
qemudSecurityCapsInit(virSecurityDriverPtr secdrv,
                      virCapsPtr caps)
{
    const char *doi, *model;

    doi = virSecurityDriverGetDOI(secdrv);
    model = virSecurityDriverGetModel(secdrv);

    caps->host.secModel.model = strdup(model);
    if (!caps->host.secModel.model) {
735
        virReportOOMError(NULL);
736 737 738 739 740
        return -1;
    }

    caps->host.secModel.doi = strdup(doi);
    if (!caps->host.secModel.doi) {
741
        virReportOOMError(NULL);
742 743 744 745 746 747 748 749 750 751
        return -1;
    }

    VIR_DEBUG("Initialized caps for security driver \"%s\" with "
              "DOI \"%s\"", model, doi);

    return 0;
}


752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771
static int
qemudSecurityInit(struct qemud_driver *qemud_drv)
{
    int ret;
    virSecurityDriverPtr security_drv;

    ret = virSecurityDriverStartup(&security_drv,
                                   qemud_drv->securityDriverName);
    if (ret == -1) {
        VIR_ERROR0(_("Failed to start security driver"));
        return -1;
    }
    /* No security driver wanted to be enabled: just return */
    if (ret == -2) {
        VIR_INFO0(_("No security driver available"));
        return 0;
    }

    qemud_drv->securityDriver = security_drv;

772
    VIR_INFO("Initialized security driver %s", security_drv->name);
773 774 775 776 777

    /*
     * Add security policy host caps now that the security driver is
     * initialized.
     */
778 779
    return qemudSecurityCapsInit(security_drv, qemud_drv->caps);
}
780 781


782

783 784 785 786 787 788
/**
 * qemudStartup:
 *
 * Initialization function for the QEmu daemon
 */
static int
789
qemudStartup(int privileged) {
790
    char *base = NULL;
D
Daniel P. Berrange 已提交
791
    char driverConf[PATH_MAX];
792
    int rc;
793

794
    if (VIR_ALLOC(qemu_driver) < 0)
795 796
        return -1;

797
    if (virMutexInit(&qemu_driver->lock) < 0) {
798
        VIR_ERROR("%s", _("cannot initialize mutex"));
799 800 801
        VIR_FREE(qemu_driver);
        return -1;
    }
802
    qemuDriverLock(qemu_driver);
803
    qemu_driver->privileged = privileged;
804

805 806 807
    /* Don't have a dom0 so start from 1 */
    qemu_driver->nextvmid = 1;

808 809 810
    if (virDomainObjListInit(&qemu_driver->domains) < 0)
        goto out_of_memory;

811
    /* Init callback list */
812
    if (VIR_ALLOC(qemu_driver->domainEventCallbacks) < 0)
813
        goto out_of_memory;
814 815 816 817 818 819
    if (!(qemu_driver->domainEventQueue = virDomainEventQueueNew()))
        goto out_of_memory;

    if ((qemu_driver->domainEventTimer =
         virEventAddTimeout(-1, qemuDomainEventFlush, qemu_driver, NULL)) < 0)
        goto error;
820

821
    if (privileged) {
822 823
        if (virAsprintf(&qemu_driver->logDir,
                        "%s/log/libvirt/qemu", LOCAL_STATE_DIR) == -1)
824
            goto out_of_memory;
825

D
Daniel P. Berrange 已提交
826
        if ((base = strdup (SYSCONF_DIR "/libvirt")) == NULL)
827
            goto out_of_memory;
828 829

        if (virAsprintf(&qemu_driver->stateDir,
830
                      "%s/run/libvirt/qemu", LOCAL_STATE_DIR) == -1)
831
            goto out_of_memory;
832 833 834 835 836 837 838 839

        if (virAsprintf(&qemu_driver->libDir,
                      "%s/lib/libvirt/qemu", LOCAL_STATE_DIR) == -1)
            goto out_of_memory;

        if (virAsprintf(&qemu_driver->cacheDir,
                      "%s/cache/libvirt/qemu", LOCAL_STATE_DIR) == -1)
            goto out_of_memory;
840
    } else {
841
        uid_t uid = geteuid();
842 843
        char *userdir = virGetUserDirectory(NULL, uid);
        if (!userdir)
844
            goto error;
845

846
        if (virAsprintf(&qemu_driver->logDir,
847 848
                        "%s/.libvirt/qemu/log", userdir) == -1) {
            VIR_FREE(userdir);
849
            goto out_of_memory;
850
        }
851

852 853
        if (virAsprintf(&base, "%s/.libvirt", userdir) == -1) {
            VIR_FREE(userdir);
854
            goto out_of_memory;
855 856
        }
        VIR_FREE(userdir);
857 858 859

        if (virAsprintf(&qemu_driver->stateDir, "%s/qemu/run", base) == -1)
            goto out_of_memory;
860 861 862 863
        if (virAsprintf(&qemu_driver->libDir, "%s/qemu/lib", base) == -1)
            goto out_of_memory;
        if (virAsprintf(&qemu_driver->cacheDir, "%s/qemu/cache", base) == -1)
            goto out_of_memory;
864 865 866
    }

    if (virFileMakePath(qemu_driver->stateDir) < 0) {
867
        char ebuf[1024];
868 869
        VIR_ERROR(_("Failed to create state dir '%s': %s\n"),
                  qemu_driver->stateDir, virStrerror(errno, ebuf, sizeof ebuf));
870
        goto error;
871
    }
872 873 874 875 876 877 878 879 880 881 882 883
    if (virFileMakePath(qemu_driver->libDir) < 0) {
        char ebuf[1024];
        VIR_ERROR(_("Failed to create lib dir '%s': %s\n"),
                  qemu_driver->libDir, virStrerror(errno, ebuf, sizeof ebuf));
        goto error;
    }
    if (virFileMakePath(qemu_driver->cacheDir) < 0) {
        char ebuf[1024];
        VIR_ERROR(_("Failed to create cache dir '%s': %s\n"),
                  qemu_driver->cacheDir, virStrerror(errno, ebuf, sizeof ebuf));
        goto error;
    }
884 885 886 887

    /* Configuration paths are either ~/.libvirt/qemu/... (session) or
     * /etc/libvirt/qemu/... (system).
     */
D
Daniel P. Berrange 已提交
888
    if (snprintf (driverConf, sizeof(driverConf), "%s/qemu.conf", base) == -1)
889
        goto out_of_memory;
D
Daniel P. Berrange 已提交
890
    driverConf[sizeof(driverConf)-1] = '\0';
891

892
    if (virAsprintf(&qemu_driver->configDir, "%s/qemu", base) == -1)
893 894
        goto out_of_memory;

895
    if (virAsprintf(&qemu_driver->autostartDir, "%s/qemu/autostart", base) == -1)
896 897
        goto out_of_memory;

898
    VIR_FREE(base);
899

900 901 902 903 904 905 906
    rc = virCgroupForDriver("qemu", &qemu_driver->cgroup, privileged, 1);
    if (rc < 0) {
        char buf[1024];
        VIR_WARN("Unable to create cgroup for driver: %s",
                 virStrerror(-rc, buf, sizeof(buf)));
    }

907
    if ((qemu_driver->caps = qemudCapsInit(NULL)) == NULL)
908
        goto out_of_memory;
D
Daniel P. Berrange 已提交
909

910 911 912
    qemu_driver->caps->privateDataAllocFunc = qemuDomainObjPrivateAlloc;
    qemu_driver->caps->privateDataFreeFunc = qemuDomainObjPrivateFree;

913 914 915
    if ((qemu_driver->activePciHostdevs = pciDeviceListNew(NULL)) == NULL)
        goto error;

916
    if (qemudLoadDriverConfig(qemu_driver, driverConf) < 0) {
917 918 919
        goto error;
    }

920 921 922 923 924 925 926 927 928 929 930 931 932 933 934
    if (privileged) {
        if (chown(qemu_driver->libDir, qemu_driver->user, qemu_driver->group) < 0) {
            virReportSystemError(NULL, errno,
                                 _("unable to set ownership of '%s' to user %d:%d"),
                                 qemu_driver->libDir, qemu_driver->user, qemu_driver->group);
            goto error;
        }
        if (chown(qemu_driver->cacheDir, qemu_driver->user, qemu_driver->group) < 0) {
            virReportSystemError(NULL, errno,
                                 _("unable to set ownership of '%s' to %d:%d"),
                                 qemu_driver->cacheDir, qemu_driver->user, qemu_driver->group);
            goto error;
        }
    }

935
    if (qemudSecurityInit(qemu_driver) < 0) {
936
        goto error;
D
Daniel P. Berrange 已提交
937 938
    }

939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969
    /* If hugetlbfs is present, then we need to create a sub-directory within
     * it, since we can't assume the root mount point has permissions that
     * will let our spawned QEMU instances use it.
     *
     * NB the check for '/', since user may config "" to disable hugepages
     * even when mounted
     */
    if (qemu_driver->hugetlbfs_mount &&
        qemu_driver->hugetlbfs_mount[0] == '/') {
        char *mempath = NULL;
        if (virAsprintf(&mempath, "%s/libvirt/qemu", qemu_driver->hugetlbfs_mount) < 0)
            goto out_of_memory;

        if ((rc = virFileMakePath(mempath)) != 0) {
            virReportSystemError(NULL, rc,
                                 _("unable to create hugepage path %s"), mempath);
            VIR_FREE(mempath);
            goto error;
        }
        if (qemu_driver->privileged &&
            chown(mempath, qemu_driver->user, qemu_driver->group) < 0) {
            virReportSystemError(NULL, errno,
                                 _("unable to set ownership on %s to %d:%d"),
                                 mempath, qemu_driver->user, qemu_driver->group);
            VIR_FREE(mempath);
            goto error;
        }

        qemu_driver->hugepage_path = mempath;
    }

970 971 972 973 974 975 976 977 978 979 980 981
    /* Get all the running persistent or transient configs first */
    if (virDomainLoadAllConfigs(NULL,
                                qemu_driver->caps,
                                &qemu_driver->domains,
                                qemu_driver->stateDir,
                                NULL,
                                1, NULL, NULL) < 0)
        goto error;

    qemuReconnectDomains(qemu_driver);

    /* Then inactive persistent configs */
982 983 984 985
    if (virDomainLoadAllConfigs(NULL,
                                qemu_driver->caps,
                                &qemu_driver->domains,
                                qemu_driver->configDir,
986
                                qemu_driver->autostartDir,
987
                                0, NULL, NULL) < 0)
988
        goto error;
989 990
    qemuDriverUnlock(qemu_driver);

991 992
    qemudAutostartConfigs(qemu_driver);

993

994 995
    return 0;

996
out_of_memory:
997
    virReportOOMError(NULL);
998 999 1000
error:
    if (qemu_driver)
        qemuDriverUnlock(qemu_driver);
1001
    VIR_FREE(base);
1002
    qemudShutdown();
1003 1004 1005
    return -1;
}

1006 1007 1008 1009
static void qemudNotifyLoadDomain(virDomainObjPtr vm, int newVM, void *opaque)
{
    struct qemud_driver *driver = opaque;

1010 1011 1012 1013 1014 1015 1016 1017
    if (newVM) {
        virDomainEventPtr event =
            virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_DEFINED,
                                     VIR_DOMAIN_EVENT_DEFINED_ADDED);
        if (event)
            qemuDomainEventQueue(driver, event);
    }
1018 1019
}

1020 1021 1022 1023 1024 1025 1026 1027
/**
 * qemudReload:
 *
 * Function to restart the QEmu daemon, it will recheck the configuration
 * files and update its state and the networking
 */
static int
qemudReload(void) {
1028 1029 1030
    if (!qemu_driver)
        return 0;

1031
    qemuDriverLock(qemu_driver);
1032 1033 1034 1035
    virDomainLoadAllConfigs(NULL,
                            qemu_driver->caps,
                            &qemu_driver->domains,
                            qemu_driver->configDir,
1036
                            qemu_driver->autostartDir,
1037
                            0, qemudNotifyLoadDomain, qemu_driver);
1038
    qemuDriverUnlock(qemu_driver);
1039

1040
    qemudAutostartConfigs(qemu_driver);
1041 1042

    return 0;
1043 1044
}

1045 1046 1047 1048 1049 1050 1051 1052 1053 1054
/**
 * qemudActive:
 *
 * Checks if the QEmu daemon is active, i.e. has an active domain or
 * an active network
 *
 * Returns 1 if active, 0 otherwise
 */
static int
qemudActive(void) {
1055
    int active = 0;
1056

1057 1058 1059
    if (!qemu_driver)
        return 0;

1060
    /* XXX having to iterate here is not great because it requires many locks */
1061
    qemuDriverLock(qemu_driver);
1062
    active = virDomainObjListNumOfDomains(&qemu_driver->domains, 1);
1063 1064
    qemuDriverUnlock(qemu_driver);
    return active;
1065 1066
}

1067 1068 1069 1070 1071 1072 1073
/**
 * qemudShutdown:
 *
 * Shutdown the QEmu daemon, it will stop all active domains and networks
 */
static int
qemudShutdown(void) {
1074

1075
    if (!qemu_driver)
1076
        return -1;
1077

1078
    qemuDriverLock(qemu_driver);
1079
    pciDeviceListFree(NULL, qemu_driver->activePciHostdevs);
1080 1081
    virCapabilitiesFree(qemu_driver->caps);

1082
    virDomainObjListDeinit(&qemu_driver->domains);
1083

1084
    VIR_FREE(qemu_driver->securityDriverName);
1085
    VIR_FREE(qemu_driver->logDir);
1086 1087
    VIR_FREE(qemu_driver->configDir);
    VIR_FREE(qemu_driver->autostartDir);
1088
    VIR_FREE(qemu_driver->stateDir);
1089 1090
    VIR_FREE(qemu_driver->libDir);
    VIR_FREE(qemu_driver->cacheDir);
1091
    VIR_FREE(qemu_driver->vncTLSx509certdir);
J
Jim Meyering 已提交
1092
    VIR_FREE(qemu_driver->vncListen);
1093
    VIR_FREE(qemu_driver->vncPassword);
1094
    VIR_FREE(qemu_driver->vncSASLdir);
1095
    VIR_FREE(qemu_driver->saveImageFormat);
1096 1097
    VIR_FREE(qemu_driver->hugetlbfs_mount);
    VIR_FREE(qemu_driver->hugepage_path);
D
Daniel P. Berrange 已提交
1098

1099 1100
    /* Free domain callback list */
    virDomainEventCallbackListFree(qemu_driver->domainEventCallbacks);
1101 1102 1103 1104
    virDomainEventQueueFree(qemu_driver->domainEventQueue);

    if (qemu_driver->domainEventTimer != -1)
        virEventRemoveTimeout(qemu_driver->domainEventTimer);
1105

1106 1107 1108
    if (qemu_driver->brctl)
        brShutdown(qemu_driver->brctl);

1109 1110
    virCgroupFree(&qemu_driver->cgroup);

1111
    qemuDriverUnlock(qemu_driver);
1112
    virMutexDestroy(&qemu_driver->lock);
1113
    VIR_FREE(qemu_driver);
1114 1115

    return 0;
1116 1117
}

1118 1119 1120 1121
typedef int qemuLogHandleOutput(virConnectPtr conn,
                                virDomainObjPtr vm,
                                const char *output,
                                int fd);
1122 1123 1124 1125 1126 1127 1128 1129 1130

/*
 * Returns -1 for error, 0 on success
 */
static int
qemudReadLogOutput(virConnectPtr conn,
                   virDomainObjPtr vm,
                   int fd,
                   char *buf,
G
Guido Günther 已提交
1131
                   size_t buflen,
1132
                   qemuLogHandleOutput func,
1133 1134 1135
                   const char *what,
                   int timeout)
{
1136
    int retries = (timeout*10);
1137
    int got = 0;
1138 1139 1140
    buf[0] = '\0';

    while (retries) {
1141
        ssize_t func_ret, ret;
1142
        int isdead = 0;
G
Guido Günther 已提交
1143

1144 1145
        func_ret = func(conn, vm, buf, fd);

1146 1147
        if (kill(vm->pid, 0) == -1 && errno == ESRCH)
            isdead = 1;
1148

1149 1150
        /* Any failures should be detected before we read the log, so we
         * always have something useful to report on failure. */
1151 1152
        ret = saferead(fd, buf+got, buflen-got-1);
        if (ret < 0) {
1153 1154 1155 1156 1157 1158
            virReportSystemError(conn, errno,
                                 _("Failure while reading %s log output"),
                                 what);
            return -1;
        }

1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174
        got += ret;
        buf[got] = '\0';
        if (got == buflen-1) {
            qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                             _("Out of space while reading %s log output"),
                             what);
            return -1;
        }

        if (isdead) {
            qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                             _("Process exited while reading %s log output"),
                             what);
            return -1;
        }

1175 1176
        if (func_ret <= 0)
            return func_ret;
1177 1178 1179 1180

        usleep(100*1000);
        retries--;
    }
1181 1182 1183

    qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                     _("Timed out while reading %s log output"), what);
1184 1185 1186
    return -1;
}

1187

1188 1189 1190 1191 1192 1193 1194 1195
/*
 * Look at a chunk of data from the QEMU stdout logs and try to
 * find a TTY device, as indicated by a line like
 *
 * char device redirected to /dev/pts/3
 *
 * Returns -1 for error, 0 success, 1 continue reading
 */
1196
static int
1197 1198 1199 1200
qemudExtractTTYPath(virConnectPtr conn,
                    const char *haystack,
                    size_t *offset,
                    char **path)
1201
{
1202
    static const char needle[] = "char device redirected to";
1203
    char *tmp, *dev;
1204

1205
    VIR_FREE(*path);
1206
    /* First look for our magic string */
1207 1208 1209 1210 1211
    if (!(tmp = strstr(haystack + *offset, needle))) {
        return 1;
    }
    tmp += sizeof(needle);
    dev = tmp;
1212

1213 1214 1215 1216 1217
    /*
     * And look for first whitespace character and nul terminate
     * to mark end of the pty path
     */
    while (*tmp) {
1218
        if (c_isspace(*tmp)) {
1219 1220
            *path = strndup(dev, tmp-dev);
            if (*path == NULL) {
1221
                virReportOOMError(conn);
1222 1223
                return -1;
            }
1224

1225
            /* ... now further update offset till we get EOL */
1226
            *offset = tmp - haystack;
1227 1228
            return 0;
        }
1229
        tmp++;
1230 1231 1232 1233 1234
    }

    /*
     * We found a path, but didn't find any whitespace,
     * so it must be still incomplete - we should at
1235 1236
     * least see a \n - indicate that we want to carry
     * on trying again
1237
     */
1238
    return 1;
1239 1240 1241
}

static int
1242
qemudFindCharDevicePTYs(virConnectPtr conn,
1243
                        virDomainObjPtr vm,
1244 1245
                        const char *output,
                        int fd ATTRIBUTE_UNUSED)
1246
{
1247
    size_t offset = 0;
1248
    int ret, i;
1249 1250

    /* The order in which QEMU prints out the PTY paths is
1251 1252
       the order in which it procsses its serial and parallel
       device args. This code must match that ordering.... */
1253

1254
    /* first comes the serial devices */
1255 1256
    for (i = 0 ; i < vm->def->nserials ; i++) {
        virDomainChrDefPtr chr = vm->def->serials[i];
1257
        if (chr->type == VIR_DOMAIN_CHR_TYPE_PTY) {
1258 1259
            if ((ret = qemudExtractTTYPath(conn, output, &offset,
                                           &chr->data.file.path)) != 0)
1260
                return ret;
1261 1262 1263
        }
    }

1264
    /* then the parallel devices */
1265 1266
    for (i = 0 ; i < vm->def->nparallels ; i++) {
        virDomainChrDefPtr chr = vm->def->parallels[i];
1267
        if (chr->type == VIR_DOMAIN_CHR_TYPE_PTY) {
1268 1269
            if ((ret = qemudExtractTTYPath(conn, output, &offset,
                                           &chr->data.file.path)) != 0)
1270
                return ret;
1271 1272 1273
        }
    }

1274
    return 0;
1275 1276
}

1277 1278 1279 1280
static int
qemudWaitForMonitor(virConnectPtr conn,
                    struct qemud_driver* driver,
                    virDomainObjPtr vm, off_t pos)
1281
{
1282
    char buf[4096]; /* Plenty of space to get startup greeting */
1283 1284 1285 1286 1287
    int logfd;
    int ret;

    if ((logfd = qemudLogReadFD(conn, driver->logDir, vm->def->name, pos))
        < 0)
1288
        return -1;
1289

1290 1291 1292
    ret = qemudReadLogOutput(conn, vm, logfd, buf, sizeof(buf),
                             qemudFindCharDevicePTYs,
                             "console", 3);
1293
    if (close(logfd) < 0) {
1294
        char ebuf[4096];
1295
        VIR_WARN(_("Unable to close logfile: %s\n"),
1296 1297
                 virStrerror(errno, ebuf, sizeof ebuf));
    }
1298

1299 1300 1301 1302 1303 1304
    if (ret < 0) {
        /* Unexpected end of file - inform user of QEMU log data */
        qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("unable to start guest: %s"), buf);
        return -1;
    }
1305

1306 1307
    VIR_DEBUG("Connect monitor to %p '%s'", vm, vm->def->name);
    if (qemuConnectMonitor(vm) < 0)
1308 1309 1310
        return -1;

    return 0;
1311 1312
}

1313
static int
1314
qemuDetectVcpuPIDs(virConnectPtr conn,
1315
                   struct qemud_driver *driver,
1316 1317 1318
                   virDomainObjPtr vm) {
    pid_t *cpupids = NULL;
    int ncpupids;
1319
    qemuDomainObjPrivatePtr priv = vm->privateData;
1320

1321
    if (vm->def->virtType != VIR_DOMAIN_VIRT_KVM) {
1322 1323 1324 1325 1326
        vm->nvcpupids = 1;
        if (VIR_ALLOC_N(vm->vcpupids, vm->nvcpupids) < 0) {
            virReportOOMError(conn);
            return -1;
        }
1327 1328 1329 1330
        vm->vcpupids[0] = vm->pid;
        return 0;
    }

1331
    /* What follows is now all KVM specific */
1332

1333
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
1334
    if ((ncpupids = qemuMonitorGetCPUInfo(priv->mon, &cpupids)) < 0) {
1335
        qemuDomainObjExitMonitorWithDriver(driver, vm);
1336
        return -1;
1337
    }
1338
    qemuDomainObjExitMonitorWithDriver(driver, vm);
1339

1340 1341 1342
    /* Treat failure to get VCPU<->PID mapping as non-fatal */
    if (ncpupids == 0)
        return 0;
1343

1344 1345 1346 1347 1348 1349 1350
    if (ncpupids != vm->def->vcpus) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("got wrong number of vCPU pids from QEMU monitor. got %d, wanted %d"),
                         ncpupids, (int)vm->def->vcpus);
        VIR_FREE(cpupids);
        return -1;
    }
1351

1352 1353
    vm->nvcpupids = ncpupids;
    vm->vcpupids = cpupids;
1354 1355 1356
    return 0;
}

1357 1358
static int
qemudInitCpus(virConnectPtr conn,
1359
              struct qemud_driver *driver,
D
Daniel Veillard 已提交
1360 1361
              virDomainObjPtr vm,
              const char *migrateFrom) {
1362
    int i, hostcpus, maxcpu = QEMUD_CPUMASK_LEN;
1363
    virNodeInfo nodeinfo;
1364
    qemuDomainObjPrivatePtr priv = vm->privateData;
1365 1366
    unsigned char *cpumap;
    int cpumaplen;
1367

1368
    if (nodeGetInfo(conn, &nodeinfo) < 0)
1369 1370 1371 1372
        return -1;

    /* setaffinity fails if you set bits for CPUs which
     * aren't present, so we have to limit ourselves */
1373 1374 1375
    hostcpus = VIR_NODEINFO_MAXCPUS(nodeinfo);
    if (maxcpu > hostcpus)
        maxcpu = hostcpus;
1376

1377 1378 1379 1380 1381 1382
    cpumaplen = VIR_CPU_MAPLEN(maxcpu);
    if (VIR_ALLOC_N(cpumap, cpumaplen) < 0) {
        virReportOOMError(conn);
        return -1;
    }

D
Daniel P. Berrange 已提交
1383
    if (vm->def->cpumask) {
1384 1385 1386
        /* XXX why don't we keep 'cpumask' in the libvirt cpumap
         * format to start with ?!?! */
        for (i = 0 ; i < maxcpu && i < vm->def->cpumasklen ; i++)
D
Daniel P. Berrange 已提交
1387
            if (vm->def->cpumask[i])
1388
                VIR_USE_CPU(cpumap, i);
D
Daniel P. Berrange 已提交
1389
    } else {
1390 1391 1392 1393
        /* 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 */
D
Daniel P. Berrange 已提交
1394
        for (i = 0 ; i < maxcpu ; i++)
1395
            VIR_USE_CPU(cpumap, i);
D
Daniel P. Berrange 已提交
1396
    }
1397

1398 1399
    /* The XML config only gives a per-VM affinity, so we apply
     * the same mapping to all vCPUs */
1400
    for (i = 0 ; i < vm->nvcpupids ; i++) {
1401 1402 1403
        if (virProcessInfoSetAffinity(vm->vcpupids[i],
                                      cpumap, cpumaplen, maxcpu) < 0) {
            VIR_FREE(cpumap);
1404 1405 1406
            return -1;
        }
    }
1407
    VIR_FREE(cpumap);
1408

1409
    /* XXX This resume doesn't really belong here. Move it up to caller */
D
Daniel Veillard 已提交
1410 1411
    if (migrateFrom == NULL) {
        /* Allow the CPUS to start executing */
1412
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
1413
        if (qemuMonitorStartCPUs(priv->mon, conn) < 0) {
1414 1415 1416
            if (virGetLastError() == NULL)
                qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                                 "%s", _("resume operation failed"));
1417
            qemuDomainObjExitMonitorWithDriver(driver, vm);
D
Daniel Veillard 已提交
1418 1419
            return -1;
        }
1420
        qemuDomainObjExitMonitorWithDriver(driver, vm);
1421 1422 1423 1424 1425 1426
    }

    return 0;
}


1427
static int
1428 1429
qemuInitPasswords(struct qemud_driver *driver,
                  virDomainObjPtr vm) {
1430
    int ret = 0;
1431
    qemuDomainObjPrivatePtr priv = vm->privateData;
1432

1433 1434 1435
    if ((vm->def->ngraphics == 1) &&
        vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
        (vm->def->graphics[0]->data.vnc.passwd || driver->vncPassword)) {
1436

1437
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
1438
        ret = qemuMonitorSetVNCPassword(priv->mon,
1439 1440 1441
                                        vm->def->graphics[0]->data.vnc.passwd ?
                                        vm->def->graphics[0]->data.vnc.passwd :
                                        driver->vncPassword);
1442
        qemuDomainObjExitMonitorWithDriver(driver, vm);
1443 1444
    }

1445
    return ret;
1446 1447 1448
}


1449
static int qemudNextFreeVNCPort(struct qemud_driver *driver ATTRIBUTE_UNUSED) {
1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484
    int i;

    for (i = 5900 ; i < 6000 ; i++) {
        int fd;
        int reuse = 1;
        struct sockaddr_in addr;
        addr.sin_family = AF_INET;
        addr.sin_port = htons(i);
        addr.sin_addr.s_addr = htonl(INADDR_ANY);
        fd = socket(PF_INET, SOCK_STREAM, 0);
        if (fd < 0)
            return -1;

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

        if (bind(fd, (struct sockaddr*)&addr, sizeof(addr)) == 0) {
            /* Not in use, lets grab it */
            close(fd);
            return i;
        }
        close(fd);

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

1485 1486 1487 1488 1489
static pciDeviceList *
qemuGetPciHostDeviceList(virConnectPtr conn,
                         virDomainDefPtr def)
{
    pciDeviceList *list;
1490 1491
    int i;

1492 1493
    if (!(list = pciDeviceListNew(conn)))
        return NULL;
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508

    for (i = 0 ; i < def->nhostdevs ; i++) {
        virDomainHostdevDefPtr hostdev = def->hostdevs[i];
        pciDevice *dev;

        if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
            continue;
        if (hostdev->source.subsys.type != VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI)
            continue;

        dev = pciGetDevice(conn,
                           hostdev->source.subsys.u.pci.domain,
                           hostdev->source.subsys.u.pci.bus,
                           hostdev->source.subsys.u.pci.slot,
                           hostdev->source.subsys.u.pci.function);
1509 1510 1511 1512
        if (!dev) {
            pciDeviceListFree(conn, list);
            return NULL;
        }
1513

1514
        if (pciDeviceListAdd(conn, list, dev) < 0) {
1515
            pciFreeDevice(conn, dev);
1516 1517
            pciDeviceListFree(conn, list);
            return NULL;
1518 1519
        }

1520
        pciDeviceSetManaged(dev, hostdev->managed);
1521 1522
    }

1523 1524 1525 1526
    return list;
}

static int
1527 1528 1529 1530
qemuUpdateActivePciHostdevs(struct qemud_driver *driver,
                            virDomainDefPtr def)
{
    pciDeviceList *pcidevs;
1531
    int ret = -1;
1532 1533 1534 1535 1536 1537 1538

    if (!def->nhostdevs)
        return 0;

    if (!(pcidevs = qemuGetPciHostDeviceList(NULL, def)))
        return -1;

1539 1540
    while (pciDeviceListCount(pcidevs) > 0) {
        pciDevice *dev = pciDeviceListSteal(NULL, pcidevs, 0);
1541 1542
        if (pciDeviceListAdd(NULL,
                             driver->activePciHostdevs,
1543 1544 1545
                             dev) < 0) {
            pciFreeDevice(NULL, dev);
            goto cleanup;
1546 1547 1548
        }
    }

1549 1550 1551
    ret = 0;

cleanup:
1552 1553 1554 1555 1556 1557 1558 1559
    pciDeviceListFree(NULL, pcidevs);
    return ret;
}

static int
qemuPrepareHostDevices(virConnectPtr conn,
                       struct qemud_driver *driver,
                       virDomainDefPtr def)
1560 1561 1562
{
    pciDeviceList *pcidevs;
    int i;
1563
    int ret = -1;
1564 1565 1566 1567 1568 1569 1570

    if (!def->nhostdevs)
        return 0;

    if (!(pcidevs = qemuGetPciHostDeviceList(conn, def)))
        return -1;

1571
    /* We have to use 3 loops here. *All* devices must
1572 1573
     * be detached before we reset any of them, because
     * in some cases you have to reset the whole PCI,
1574 1575
     * which impacts all devices on it. Also, all devices
     * must be reset before being marked as active.
1576 1577 1578 1579 1580 1581 1582
     */

    /* XXX validate that non-managed device isn't in use, eg
     * by checking that device is either un-bound, or bound
     * to pci-stub.ko
     */

1583 1584 1585 1586 1587 1588
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
        if (pciDeviceGetManaged(dev) &&
            pciDettachDevice(conn, dev) < 0)
            goto cleanup;
    }
1589 1590 1591

    /* Now that all the PCI hostdevs have be dettached, we can safely
     * reset them */
1592 1593 1594
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
        if (pciResetDevice(conn, dev,
1595
                           driver->activePciHostdevs) < 0)
1596 1597
            goto cleanup;
    }
1598

1599
    /* Now mark all the devices as active */
1600 1601 1602
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
        pciDeviceListSteal(NULL, pcidevs, dev);
1603 1604
        if (pciDeviceListAdd(conn,
                             driver->activePciHostdevs,
1605 1606 1607 1608
                             dev) < 0) {
            pciFreeDevice(NULL, dev);
            goto cleanup;
        }
1609 1610
    }

1611
    ret = 0;
1612

1613
cleanup:
1614
    pciDeviceListFree(conn, pcidevs);
1615
    return ret;
1616 1617
}

1618
static void
1619 1620 1621
qemuDomainReAttachHostDevices(virConnectPtr conn,
                              struct qemud_driver *driver,
                              virDomainDefPtr def)
1622
{
1623
    pciDeviceList *pcidevs;
1624 1625
    int i;

1626 1627
    if (!def->nhostdevs)
        return;
1628

1629 1630 1631 1632 1633 1634
    if (!(pcidevs = qemuGetPciHostDeviceList(conn, def))) {
        virErrorPtr err = virGetLastError();
        VIR_ERROR(_("Failed to allocate pciDeviceList: %s\n"),
                  err ? err->message : "");
        virResetError(err);
        return;
1635 1636
    }

1637 1638
    /* Again 3 loops; mark all devices as inactive before reset
     * them and reset all the devices before re-attach */
1639

1640 1641 1642 1643
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
        pciDeviceListDel(conn, driver->activePciHostdevs, dev);
    }
1644

1645 1646 1647
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
        if (pciResetDevice(conn, dev,
1648
                           driver->activePciHostdevs) < 0) {
1649
            virErrorPtr err = virGetLastError();
1650
            VIR_ERROR(_("Failed to reset PCI device: %s\n"),
1651 1652 1653
                      err ? err->message : "");
            virResetError(err);
        }
1654
    }
1655

1656 1657 1658 1659
    for (i = 0; i < pciDeviceListCount(pcidevs); i++) {
        pciDevice *dev = pciDeviceListGet(pcidevs, i);
        if (pciDeviceGetManaged(dev) &&
            pciReAttachDevice(NULL, dev) < 0) {
1660
            virErrorPtr err = virGetLastError();
1661
            VIR_ERROR(_("Failed to re-attach PCI device: %s\n"),
1662 1663 1664
                      err ? err->message : "");
            virResetError(err);
        }
1665
    }
1666

1667
    pciDeviceListFree(conn, pcidevs);
1668 1669
}

1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
static const char *const defaultDeviceACL[] = {
    "/dev/null", "/dev/full", "/dev/zero",
    "/dev/random", "/dev/urandom",
    "/dev/ptmx", "/dev/kvm", "/dev/kqemu",
    "/dev/rtc", "/dev/hpet", "/dev/net/tun",
    NULL,
};
#define DEVICE_PTY_MAJOR 136
#define DEVICE_SND_MAJOR 116

1680 1681 1682 1683 1684 1685
static int qemuSetupCgroup(virConnectPtr conn,
                           struct qemud_driver *driver,
                           virDomainObjPtr vm)
{
    virCgroupPtr cgroup = NULL;
    int rc;
1686
    unsigned int i;
1687 1688 1689 1690
    const char *const *deviceACL =
        driver->cgroupDeviceACL ?
        (const char *const *)driver->cgroupDeviceACL :
        defaultDeviceACL;
1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702

    if (driver->cgroup == NULL)
        return 0; /* Not supported, so claim success */

    rc = virCgroupForDomain(driver->cgroup, vm->def->name, &cgroup, 1);
    if (rc != 0) {
        virReportSystemError(conn, -rc,
                             _("Unable to create cgroup for %s"),
                             vm->def->name);
        goto cleanup;
    }

1703 1704
    if (qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_DEVICES)) {
        rc = virCgroupDenyAllDevices(cgroup);
1705
        if (rc != 0) {
1706 1707 1708 1709 1710
            if (rc == -EPERM) {
                VIR_WARN0("Group devices ACL is not accessible, disabling whitelisting");
                goto done;
            }

1711
            virReportSystemError(conn, -rc,
1712
                                 _("Unable to deny all devices for %s"), vm->def->name);
1713 1714 1715
            goto cleanup;
        }

1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729
        for (i = 0; i < vm->def->ndisks ; i++) {
            if (vm->def->disks[i]->type != VIR_DOMAIN_DISK_TYPE_BLOCK ||
                vm->def->disks[i]->src == NULL)
                continue;

            rc = virCgroupAllowDevicePath(cgroup,
                                          vm->def->disks[i]->src);
            if (rc != 0) {
                virReportSystemError(conn, -rc,
                                     _("Unable to allow device %s for %s"),
                                     vm->def->disks[i]->src, vm->def->name);
                goto cleanup;
            }
        }
1730

1731
        rc = virCgroupAllowDeviceMajor(cgroup, 'c', DEVICE_PTY_MAJOR);
1732 1733
        if (rc != 0) {
            virReportSystemError(conn, -rc, "%s",
1734
                                 _("unable to allow /dev/pts/ devices"));
1735 1736 1737
            goto cleanup;
        }

1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756
        if (vm->def->nsounds) {
            rc = virCgroupAllowDeviceMajor(cgroup, 'c', DEVICE_SND_MAJOR);
            if (rc != 0) {
                virReportSystemError(conn, -rc, "%s",
                                     _("unable to allow /dev/snd/ devices"));
                goto cleanup;
            }
        }

        for (i = 0; deviceACL[i] != NULL ; i++) {
            rc = virCgroupAllowDevicePath(cgroup,
                                          deviceACL[i]);
            if (rc < 0 &&
                rc != -ENOENT) {
                virReportSystemError(conn, -rc,
                                     _("unable to allow device %s"),
                                     deviceACL[i]);
                goto cleanup;
            }
1757 1758 1759 1760
        }
    }

done:
1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829
    virCgroupFree(&cgroup);
    return 0;

cleanup:
    if (cgroup) {
        virCgroupRemove(cgroup);
        virCgroupFree(&cgroup);
    }
    return -1;
}


static int qemuRemoveCgroup(virConnectPtr conn,
                            struct qemud_driver *driver,
                            virDomainObjPtr vm)
{
    virCgroupPtr cgroup;
    int rc;

    if (driver->cgroup == NULL)
        return 0; /* Not supported, so claim success */

    rc = virCgroupForDomain(driver->cgroup, vm->def->name, &cgroup, 0);
    if (rc != 0) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("Unable to find cgroup for %s\n"),
                         vm->def->name);
        return rc;
    }

    rc = virCgroupRemove(cgroup);
    virCgroupFree(&cgroup);
    return rc;
}

static int qemuAddToCgroup(struct qemud_driver *driver,
                           virDomainDefPtr def)
{
    virCgroupPtr cgroup = NULL;
    int ret = -1;
    int rc;

    if (driver->cgroup == NULL)
        return 0; /* Not supported, so claim success */

    rc = virCgroupForDomain(driver->cgroup, def->name, &cgroup, 0);
    if (rc != 0) {
        virReportSystemError(NULL, -rc,
                             _("unable to find cgroup for domain %s"),
                             def->name);
        goto cleanup;
    }

    rc = virCgroupAddTask(cgroup, getpid());
    if (rc != 0) {
        virReportSystemError(NULL, -rc,
                             _("unable to add domain %s task %d to cgroup"),
                             def->name, getpid());
        goto cleanup;
    }

    ret = 0;

cleanup:
    virCgroupFree(&cgroup);
    return ret;
}


1830 1831 1832 1833 1834 1835 1836 1837 1838
static int qemudDomainSetSecurityLabel(virConnectPtr conn, struct qemud_driver *driver, virDomainObjPtr vm)
{
    if (vm->def->seclabel.label != NULL)
        if (driver->securityDriver && driver->securityDriver->domainSetSecurityLabel)
            return driver->securityDriver->domainSetSecurityLabel(conn, driver->securityDriver,
                                                                 vm);
    return 0;
}

1839 1840

#ifdef __linux__
1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851
struct qemuFileOwner {
    uid_t uid;
    gid_t gid;
};

static int qemuDomainSetHostdevUSBOwnershipActor(virConnectPtr conn,
                                                 usbDevice *dev ATTRIBUTE_UNUSED,
                                                 const char *file, void *opaque)
{
    struct qemuFileOwner *owner = opaque;

1852 1853
    VIR_DEBUG("Setting ownership on %s to %d:%d", file, owner->uid, owner->gid);

1854 1855 1856 1857 1858 1859 1860 1861
    if (chown(file, owner->uid, owner->gid) < 0) {
        virReportSystemError(conn, errno, _("cannot set ownership on %s"), file);
        return -1;
    }

    return 0;
}

1862 1863 1864 1865
static int qemuDomainSetHostdevUSBOwnership(virConnectPtr conn,
                                            virDomainHostdevDefPtr def,
                                            uid_t uid, gid_t gid)
{
1866 1867
    struct qemuFileOwner owner = { uid, gid };
    int ret = -1;
1868 1869 1870 1871 1872 1873

    /* XXX what todo for USB devs assigned based on product/vendor ? Doom :-( */
    if (!def->source.subsys.u.usb.bus ||
        !def->source.subsys.u.usb.device)
        return 0;

1874 1875 1876 1877 1878 1879 1880 1881 1882
    usbDevice *dev = usbGetDevice(conn,
                                  def->source.subsys.u.usb.bus,
                                  def->source.subsys.u.usb.device);

    if (!dev)
        goto cleanup;

    ret = usbDeviceFileIterate(conn, dev,
                               qemuDomainSetHostdevUSBOwnershipActor, &owner);
1883

1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894
    usbFreeDevice(conn, dev);
cleanup:
    return ret;
}

static int qemuDomainSetHostdevPCIOwnershipActor(virConnectPtr conn,
                                                 pciDevice *dev ATTRIBUTE_UNUSED,
                                                 const char *file, void *opaque)
{
    struct qemuFileOwner *owner = opaque;

1895 1896
    VIR_DEBUG("Setting ownership on %s to %d:%d", file, owner->uid, owner->gid);

1897 1898
    if (chown(file, owner->uid, owner->gid) < 0) {
        virReportSystemError(conn, errno, _("cannot set ownership on %s"), file);
1899 1900 1901 1902 1903 1904 1905 1906 1907 1908
        return -1;
    }

    return 0;
}

static int qemuDomainSetHostdevPCIOwnership(virConnectPtr conn,
                                            virDomainHostdevDefPtr def,
                                            uid_t uid, gid_t gid)
{
1909
    struct qemuFileOwner owner = { uid, gid };
1910 1911
    int ret = -1;

1912 1913 1914 1915 1916
    pciDevice *dev = pciGetDevice(conn,
                                  def->source.subsys.u.pci.domain,
                                  def->source.subsys.u.pci.bus,
                                  def->source.subsys.u.pci.slot,
                                  def->source.subsys.u.pci.function);
1917

1918
    if (!dev)
1919 1920
        goto cleanup;

1921 1922
    ret = pciDeviceFileIterate(conn, dev,
                               qemuDomainSetHostdevPCIOwnershipActor, &owner);
1923

1924
    pciFreeDevice(conn, dev);
1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948
cleanup:
    return ret;
}
#endif


static int qemuDomainSetHostdevOwnership(virConnectPtr conn,
                                         virDomainHostdevDefPtr def,
                                         uid_t uid, gid_t gid)
{
    if (def->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS)
        return 0;

#ifdef __linux__
    switch (def->source.subsys.type) {
    case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB:
        return qemuDomainSetHostdevUSBOwnership(conn, def, uid, gid);

    case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI:
        return qemuDomainSetHostdevPCIOwnership(conn, def, uid, gid);

    }
    return 0;
#else
C
Cole Robinson 已提交
1949
    qemudReportError(conn, NULL, NULL, VIR_ERR_NO_SUPPORT, "%s",
1950 1951 1952 1953 1954 1955
                     _("unable to set host device ownership on this platform"));
    return -1;
#endif

}

1956 1957
static int qemuDomainSetFileOwnership(virConnectPtr conn,
                                      const char *path,
1958 1959 1960
                                      uid_t uid, gid_t gid)
{

1961
    if (!path)
1962 1963
        return 0;

1964 1965
    VIR_DEBUG("Setting ownership on %s to %d:%d", path, uid, gid);
    if (chown(path, uid, gid) < 0) {
1966
        virReportSystemError(conn, errno, _("cannot set ownership on %s"),
1967
                             path);
1968 1969 1970 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
        return -1;
    }
    return 0;
}

static int qemuDomainSetDeviceOwnership(virConnectPtr conn,
                                        struct qemud_driver *driver,
                                        virDomainDeviceDefPtr def,
                                        int restore)
{
    uid_t uid;
    gid_t gid;

    if (!driver->privileged)
        return 0;

    /* short circuit case of root:root */
    if (!driver->user && !driver->group)
        return 0;

    uid = restore ? 0 : driver->user;
    gid = restore ? 0 : driver->group;

    switch (def->type) {
    case VIR_DOMAIN_DEVICE_DISK:
        if (restore &&
            (def->data.disk->readonly || def->data.disk->shared))
            return 0;

1997
        return qemuDomainSetFileOwnership(conn, def->data.disk->src, uid, gid);
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024

    case VIR_DOMAIN_DEVICE_HOSTDEV:
        return qemuDomainSetHostdevOwnership(conn, def->data.hostdev, uid, gid);
    }

    return 0;
}

static int qemuDomainSetAllDeviceOwnership(virConnectPtr conn,
                                           struct qemud_driver *driver,
                                           virDomainDefPtr def,
                                           int restore)
{
    int i;
    uid_t uid;
    gid_t gid;

    if (!driver->privileged)
        return 0;

    /* short circuit case of root:root */
    if (!driver->user && !driver->group)
        return 0;

    uid = restore ? 0 : driver->user;
    gid = restore ? 0 : driver->group;

2025 2026 2027 2028
    if (qemuDomainSetFileOwnership(conn, def->os.kernel, uid, gid) < 0 ||
        qemuDomainSetFileOwnership(conn, def->os.initrd, uid, gid) < 0)
        return -1;

2029 2030 2031 2032 2033
    for (i = 0 ; i < def->ndisks ; i++) {
        if (restore &&
            (def->disks[i]->readonly || def->disks[i]->shared))
            continue;

2034
        if (qemuDomainSetFileOwnership(conn, def->disks[i]->src, uid, gid) < 0)
2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045
            return -1;
    }

    for (i = 0 ; i < def->nhostdevs ; i++) {
        if (qemuDomainSetHostdevOwnership(conn, def->hostdevs[i], uid, gid) < 0)
            return -1;
    }

    return 0;
}

2046 2047 2048
static virDomainPtr qemudDomainLookupByName(virConnectPtr conn,
                                            const char *name);

2049 2050 2051 2052
struct qemudHookData {
    virConnectPtr conn;
    virDomainObjPtr vm;
    struct qemud_driver *driver;
2053 2054 2055
};

static int qemudSecurityHook(void *data) {
2056 2057 2058 2059
    struct qemudHookData *h = data;

    if (qemuAddToCgroup(h->driver, h->vm->def) < 0)
        return -1;
2060

2061
    if (qemudDomainSetSecurityLabel(h->conn, h->driver, h->vm) < 0)
2062 2063 2064 2065 2066
        return -1;

    if (h->driver->privileged) {
        if (qemuDomainSetAllDeviceOwnership(h->conn, h->driver, h->vm->def, 0) < 0)
            return -1;
2067

2068 2069
        DEBUG("Dropping privileges of VM to %d:%d", h->driver->user, h->driver->group);

2070 2071 2072 2073 2074
        if (h->driver->group) {
            if (setregid(h->driver->group, h->driver->group) < 0) {
                virReportSystemError(NULL, errno,
                                     _("cannot change to '%d' group"),
                                     h->driver->group);
2075
                return -1;
2076
            }
2077
        }
2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088
        if (h->driver->user) {
            if (setreuid(h->driver->user, h->driver->user) < 0) {
                virReportSystemError(NULL, errno,
                                     _("cannot change to '%d' user"),
                                     h->driver->user);
                return -1;
            }
        }
    }

    return 0;
2089 2090
}

2091 2092 2093 2094 2095 2096
static int
qemuPrepareMonitorChr(virConnectPtr conn,
                      struct qemud_driver *driver,
                      virDomainChrDefPtr monitor_chr,
                      const char *vm)
{
2097 2098
    monitor_chr->targetType = VIR_DOMAIN_CHR_TARGET_TYPE_MONITOR;

2099 2100 2101 2102
    monitor_chr->type = VIR_DOMAIN_CHR_TYPE_UNIX;
    monitor_chr->data.nix.listen = 1;

    if (virAsprintf(&monitor_chr->data.nix.path, "%s/%s.monitor",
2103
                    driver->libDir, vm) < 0) {
2104 2105 2106 2107 2108 2109 2110
        virReportOOMError(conn);
        return -1;
    }

    return 0;
}

2111 2112
static int qemudStartVMDaemon(virConnectPtr conn,
                              struct qemud_driver *driver,
2113
                              virDomainObjPtr vm,
2114 2115
                              const char *migrateFrom,
                              int stdin_fd) {
2116
    const char **argv = NULL, **tmp;
2117
    const char **progenv = NULL;
2118
    int i, ret;
2119
    struct stat sb;
2120 2121
    int *tapfds = NULL;
    int ntapfds = 0;
2122
    unsigned int qemuCmdFlags;
2123
    fd_set keepfd;
2124
    const char *emulator;
G
Guido Günther 已提交
2125
    pid_t child;
2126
    int pos = -1;
2127
    char ebuf[1024];
2128
    char *pidfile = NULL;
2129
    int logfile;
2130
    qemuDomainObjPrivatePtr priv = vm->privateData;
2131

2132
    struct qemudHookData hookData;
2133 2134 2135 2136
    hookData.conn = conn;
    hookData.vm = vm;
    hookData.driver = driver;

2137
    FD_ZERO(&keepfd);
2138

D
Daniel P. Berrange 已提交
2139
    if (virDomainObjIsActive(vm)) {
2140
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_INVALID,
2141
                         "%s", _("VM is already active"));
2142 2143 2144
        return -1;
    }

2145 2146 2147 2148 2149 2150 2151 2152
    /* If you are using a SecurityDriver with dynamic labelling,
       then generate a security label for isolation */
    if (vm->def->seclabel.type == VIR_DOMAIN_SECLABEL_DYNAMIC &&
        driver->securityDriver &&
        driver->securityDriver->domainGenSecurityLabel &&
        driver->securityDriver->domainGenSecurityLabel(conn, vm) < 0)
        return -1;

2153 2154 2155
    /* Ensure no historical cgroup for this VM is lieing around bogus settings */
    qemuRemoveCgroup(conn, driver, vm);

2156 2157 2158
    if ((vm->def->ngraphics == 1) &&
        vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
        vm->def->graphics[0]->data.vnc.autoport) {
2159
        int port = qemudNextFreeVNCPort(driver);
2160
        if (port < 0) {
2161
            qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
2162
                             "%s", _("Unable to find an unused VNC port"));
2163
            goto cleanup;
2164
        }
2165
        vm->def->graphics[0]->data.vnc.port = port;
2166
    }
2167

2168
    if (virFileMakePath(driver->logDir) < 0) {
2169 2170 2171
        virReportSystemError(conn, errno,
                             _("cannot create log directory %s"),
                             driver->logDir);
2172
        goto cleanup;
2173 2174
    }

2175
    if ((logfile = qemudLogFD(conn, driver, vm->def->name)) < 0)
2176
        goto cleanup;
2177

2178 2179
    emulator = vm->def->emulator;

2180 2181 2182 2183
    /* Make sure the binary we are about to try exec'ing exists.
     * Technically we could catch the exec() failure, but that's
     * in a sub-process so its hard to feed back a useful error
     */
2184
    if (stat(emulator, &sb) < 0) {
2185 2186 2187
        virReportSystemError(conn, errno,
                             _("Cannot find QEMU binary %s"),
                             emulator);
2188
        goto cleanup;
2189 2190
    }

2191
    if (qemudExtractVersionInfo(emulator,
2192
                                NULL,
2193
                                &qemuCmdFlags) < 0)
2194
        goto cleanup;
2195

2196 2197 2198
    if (qemuSetupCgroup(conn, driver, vm) < 0)
        goto cleanup;

2199
    if (qemuPrepareHostDevices(conn, driver, vm->def) < 0)
2200
        goto cleanup;
2201

2202 2203 2204 2205 2206
    if (VIR_ALLOC(vm->monitor_chr) < 0) {
        virReportOOMError(conn);
        goto cleanup;
    }

2207 2208
    if (qemuPrepareMonitorChr(conn, driver, vm->monitor_chr, vm->def->name) < 0)
        goto cleanup;
2209

D
Daniel P. Berrange 已提交
2210 2211 2212 2213 2214 2215 2216
    if ((ret = virFileDeletePid(driver->stateDir, vm->def->name)) != 0) {
        virReportSystemError(conn, ret,
                             _("Cannot remove stale PID file for %s"),
                             vm->def->name);
        goto cleanup;
    }

2217 2218 2219 2220 2221
    if (!(pidfile = virFilePid(driver->stateDir, vm->def->name))) {
        virReportSystemError(conn, errno,
                             "%s", _("Failed to build pidfile path."));
        goto cleanup;
    }
D
Daniel P. Berrange 已提交
2222

2223
    vm->def->id = driver->nextvmid++;
2224
    if (qemudBuildCommandLine(conn, driver, vm->def, vm->monitor_chr,
2225
                              qemuCmdFlags, &argv, &progenv,
2226 2227
                              &tapfds, &ntapfds, migrateFrom) < 0)
        goto cleanup;
2228

2229 2230
    tmp = progenv;
    while (*tmp) {
2231
        if (safewrite(logfile, *tmp, strlen(*tmp)) < 0)
2232
            VIR_WARN(_("Unable to write envv to logfile: %s\n"),
2233
                     virStrerror(errno, ebuf, sizeof ebuf));
2234
        if (safewrite(logfile, " ", 1) < 0)
2235
            VIR_WARN(_("Unable to write envv to logfile: %s\n"),
2236
                     virStrerror(errno, ebuf, sizeof ebuf));
2237 2238
        tmp++;
    }
2239 2240
    tmp = argv;
    while (*tmp) {
2241
        if (safewrite(logfile, *tmp, strlen(*tmp)) < 0)
2242
            VIR_WARN(_("Unable to write argv to logfile: %s\n"),
2243
                     virStrerror(errno, ebuf, sizeof ebuf));
2244
        if (safewrite(logfile, " ", 1) < 0)
2245
            VIR_WARN(_("Unable to write argv to logfile: %s\n"),
2246
                     virStrerror(errno, ebuf, sizeof ebuf));
2247 2248
        tmp++;
    }
2249
    if (safewrite(logfile, "\n", 1) < 0)
2250
        VIR_WARN(_("Unable to write argv to logfile: %s\n"),
2251
                 virStrerror(errno, ebuf, sizeof ebuf));
2252

2253
    if ((pos = lseek(logfile, 0, SEEK_END)) < 0)
2254
        VIR_WARN(_("Unable to seek to end of logfile: %s\n"),
2255
                 virStrerror(errno, ebuf, sizeof ebuf));
2256

2257 2258 2259
    for (i = 0 ; i < ntapfds ; i++)
        FD_SET(tapfds[i], &keepfd);

2260
    ret = virExecDaemonize(conn, argv, progenv, &keepfd, &child,
2261
                           stdin_fd, &logfile, &logfile,
2262
                           VIR_EXEC_NONBLOCK | VIR_EXEC_CLEAR_CAPS,
2263 2264 2265
                           qemudSecurityHook, &hookData,
                           pidfile);
    VIR_FREE(pidfile);
G
Guido Günther 已提交
2266 2267 2268

    /* wait for qemu process to to show up */
    if (ret == 0) {
2269
        if (virFileReadPid(driver->stateDir, vm->def->name, &vm->pid)) {
2270
            qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
2271
                             _("Domain %s didn't show up\n"), vm->def->name);
2272
            ret = -1;
G
Guido Günther 已提交
2273
        }
2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284
    } else if (ret == -2) {
        /* The virExec process that launches the daemon failed. Pending on
         * when it failed (we can't determine for sure), there may be
         * extra info in the domain log (if the hook failed for example).
         *
         * Pretend like things succeeded, and let 'WaitForMonitor' report
         * the log contents for us.
         */
        vm->pid = child;
        ret = 0;
    }
2285 2286

    vm->state = migrateFrom ? VIR_DOMAIN_PAUSED : VIR_DOMAIN_RUNNING;
2287

2288
    for (i = 0 ; argv[i] ; i++)
2289 2290
        VIR_FREE(argv[i]);
    VIR_FREE(argv);
2291

2292 2293 2294 2295
    for (i = 0 ; progenv[i] ; i++)
        VIR_FREE(progenv[i]);
    VIR_FREE(progenv);

2296 2297 2298
    if (tapfds) {
        for (i = 0 ; i < ntapfds ; i++) {
            close(tapfds[i]);
2299
        }
2300
        VIR_FREE(tapfds);
2301 2302
    }

2303
    if (ret == -1) /* The VM failed to start */
2304 2305
        goto cleanup;

2306 2307 2308
    if (qemudWaitForMonitor(conn, driver, vm, pos) < 0)
        goto abort;

2309
    if (qemuDetectVcpuPIDs(conn, driver, vm) < 0)
2310 2311
        goto abort;

2312
    if (qemudInitCpus(conn, driver, vm, migrateFrom) < 0)
2313 2314 2315 2316 2317
        goto abort;

    if (qemuInitPasswords(driver, vm) < 0)
        goto abort;

2318
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
2319
    if (qemuMonitorSetBalloon(priv->mon, vm->def->memory) < 0) {
2320
        qemuDomainObjExitMonitorWithDriver(driver, vm);
2321
        goto abort;
2322
    }
2323
    qemuDomainObjExitMonitorWithDriver(driver, vm);
2324

2325 2326
    if (virDomainSaveStatus(conn, driver->stateDir, vm) < 0)
        goto abort;
2327

2328
    return 0;
2329 2330

cleanup:
2331 2332 2333
    /* We jump here if we failed to start the VM for any reason
     * XXX investigate if we can kill this block and safely call
     * qemudShutdownVMDaemon even though no PID is running */
2334 2335 2336 2337 2338
    if (vm->def->seclabel.type == VIR_DOMAIN_SECLABEL_DYNAMIC) {
        VIR_FREE(vm->def->seclabel.model);
        VIR_FREE(vm->def->seclabel.label);
        VIR_FREE(vm->def->seclabel.imagelabel);
    }
2339
    qemuRemoveCgroup(conn, driver, vm);
2340 2341 2342 2343
    if ((vm->def->ngraphics == 1) &&
        vm->def->graphics[0]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
        vm->def->graphics[0]->data.vnc.autoport)
        vm->def->graphics[0]->data.vnc.port = -1;
2344 2345
    if (logfile != -1)
        close(logfile);
2346 2347
    vm->def->id = -1;
    return -1;
2348 2349 2350 2351 2352 2353 2354 2355 2356 2357

abort:
    /* We jump here if we failed to initialize the now running VM
     * killing it off and pretend we never started it */
    qemudShutdownVMDaemon(conn, driver, vm);

    if (logfile != -1)
        close(logfile);

    return -1;
2358 2359 2360
}


2361
static void qemudShutdownVMDaemon(virConnectPtr conn,
2362 2363
                                  struct qemud_driver *driver,
                                  virDomainObjPtr vm) {
D
Daniel P. Berrange 已提交
2364
    int ret;
2365
    int retries = 0;
2366
    qemuDomainObjPrivatePtr priv = vm->privateData;
D
Daniel P. Berrange 已提交
2367

D
Daniel P. Berrange 已提交
2368
    if (!virDomainObjIsActive(vm))
2369
        return;
2370

2371
    VIR_DEBUG("Shutting down VM '%s'", vm->def->name);
2372

2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388
    if (driver->macFilter) {
        int i;
        virDomainDefPtr def = vm->def;
        for (i = 0 ; i < def->nnets ; i++) {
            virDomainNetDefPtr net = def->nets[i];
            if (net->ifname == NULL)
                continue;
            if ((errno = networkDisallowMacOnPort(conn, driver, net->ifname,
                                                  net->mac))) {
                virReportSystemError(conn, errno,
             _("failed to remove ebtables rule to allow MAC address on  '%s'"),
                                     net->ifname);
            }
        }
    }

G
Guido Günther 已提交
2389 2390
    if (virKillProcess(vm->pid, 0) == 0 &&
        virKillProcess(vm->pid, SIGTERM) < 0)
2391 2392 2393
        virReportSystemError(conn, errno,
                             _("Failed to send SIGTERM to %s (%d)"),
                             vm->def->name, vm->pid);
2394

2395 2396 2397
    if (priv->mon) {
        qemuMonitorClose(priv->mon);
        priv->mon = NULL;
2398
    }
2399

2400 2401 2402 2403 2404 2405 2406
    if (vm->monitor_chr) {
        if (vm->monitor_chr->type == VIR_DOMAIN_CHR_TYPE_UNIX)
            unlink(vm->monitor_chr->data.nix.path);
        virDomainChrDefFree(vm->monitor_chr);
        vm->monitor_chr = NULL;
    }

G
Guido Günther 已提交
2407 2408
    /* shut it off for sure */
    virKillProcess(vm->pid, SIGKILL);
2409

2410 2411 2412 2413
    /* Reset Security Labels */
    if (driver->securityDriver)
        driver->securityDriver->domainRestoreSecurityLabel(conn, vm);

2414 2415 2416 2417 2418 2419 2420
    /* Clear out dynamically assigned labels */
    if (vm->def->seclabel.type == VIR_DOMAIN_SECLABEL_DYNAMIC) {
        VIR_FREE(vm->def->seclabel.model);
        VIR_FREE(vm->def->seclabel.label);
        VIR_FREE(vm->def->seclabel.imagelabel);
    }

2421 2422 2423 2424
    if (qemuDomainSetAllDeviceOwnership(conn, driver, vm->def, 1) < 0)
        VIR_WARN("Failed to restore all device ownership for %s",
                 vm->def->name);

2425
    qemuDomainReAttachHostDevices(conn, driver, vm->def);
2426

2427 2428 2429 2430 2431 2432 2433 2434 2435 2436
retry:
    if ((ret = qemuRemoveCgroup(conn, driver, vm)) < 0) {
        if (ret == -EBUSY && (retries++ < 5)) {
            usleep(200*1000);
            goto retry;
        }
        VIR_WARN("Failed to remove cgroup for %s",
                 vm->def->name);
    }

2437
    qemudRemoveDomainStatus(conn, driver, vm);
D
Daniel P. Berrange 已提交
2438

2439
    vm->pid = -1;
2440
    vm->def->id = -1;
2441
    vm->state = VIR_DOMAIN_SHUTOFF;
2442
    VIR_FREE(vm->vcpupids);
2443
    vm->nvcpupids = 0;
2444 2445

    if (vm->newDef) {
2446
        virDomainDefFree(vm->def);
2447
        vm->def = vm->newDef;
2448
        vm->def->id = -1;
2449 2450 2451 2452 2453
        vm->newDef = NULL;
    }
}


2454
static virDrvOpenStatus qemudOpen(virConnectPtr conn,
2455
                                  virConnectAuthPtr auth ATTRIBUTE_UNUSED,
2456
                                  int flags ATTRIBUTE_UNUSED) {
2457
    if (conn->uri == NULL) {
2458 2459 2460
        if (qemu_driver == NULL)
            return VIR_DRV_OPEN_DECLINED;

2461
        conn->uri = xmlParseURI(qemu_driver->privileged ?
2462 2463
                                "qemu:///system" :
                                "qemu:///session");
2464
        if (!conn->uri) {
2465
            virReportOOMError(conn);
2466 2467
            return VIR_DRV_OPEN_ERROR;
        }
2468 2469 2470 2471 2472 2473 2474 2475 2476 2477
    } else {
        /* If URI isn't 'qemu' its definitely not for us */
        if (conn->uri->scheme == NULL ||
            STRNEQ(conn->uri->scheme, "qemu"))
            return VIR_DRV_OPEN_DECLINED;

        /* Allow remote driver to deal with URIs with hostname server */
        if (conn->uri->server != NULL)
            return VIR_DRV_OPEN_DECLINED;

2478 2479 2480 2481 2482 2483
        if (qemu_driver == NULL) {
            qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR, "%s",
                             _("qemu state driver is not active"));
            return VIR_DRV_OPEN_ERROR;
        }

2484
        if (qemu_driver->privileged) {
2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499
            if (STRNEQ (conn->uri->path, "/system") &&
                STRNEQ (conn->uri->path, "/session")) {
                qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                                 _("unexpected QEMU URI path '%s', try qemu:///system"),
                                 conn->uri->path);
                return VIR_DRV_OPEN_ERROR;
            }
        } else {
            if (STRNEQ (conn->uri->path, "/session")) {
                qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                                 _("unexpected QEMU URI path '%s', try qemu:///session"),
                                 conn->uri->path);
                return VIR_DRV_OPEN_ERROR;
            }
        }
2500 2501 2502 2503 2504 2505 2506
    }
    conn->privateData = qemu_driver;

    return VIR_DRV_OPEN_SUCCESS;
}

static int qemudClose(virConnectPtr conn) {
2507
    struct qemud_driver *driver = conn->privateData;
2508 2509

    /* Get rid of callbacks registered for this conn */
2510
    qemuDriverLock(driver);
2511
    virDomainEventCallbackListRemoveConn(conn, driver->domainEventCallbacks);
2512
    qemuDriverUnlock(driver);
2513 2514 2515 2516 2517 2518

    conn->privateData = NULL;

    return 0;
}

D
Daniel Veillard 已提交
2519 2520 2521 2522 2523
/* Which features are supported by this driver? */
static int
qemudSupportsFeature (virConnectPtr conn ATTRIBUTE_UNUSED, int feature)
{
    switch (feature) {
2524 2525 2526 2527 2528
    case VIR_DRV_FEATURE_MIGRATION_V2:
    case VIR_DRV_FEATURE_MIGRATION_P2P:
        return 1;
    default:
        return 0;
D
Daniel Veillard 已提交
2529 2530 2531
    }
}

2532
static const char *qemudGetType(virConnectPtr conn ATTRIBUTE_UNUSED) {
2533
    return "QEMU";
2534 2535
}

2536

2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549
static int qemuIsSecure(virConnectPtr conn ATTRIBUTE_UNUSED)
{
    /* Trivially secure, since always inside the daemon */
    return 1;
}

static int qemuIsEncrypted(virConnectPtr conn ATTRIBUTE_UNUSED)
{
    /* Not encrypted, but remote driver takes care of that */
    return 0;
}


2550 2551 2552 2553
static int kvmGetMaxVCPUs(void) {
    int maxvcpus = 1;

    int r, fd;
2554

2555 2556
    fd = open(KVM_DEVICE, O_RDONLY);
    if (fd < 0) {
2557 2558
        virReportSystemError(NULL, errno, _("Unable to open %s"), KVM_DEVICE);
        return -1;
2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569
    }

    r = ioctl(fd, KVM_CHECK_EXTENSION, KVM_CAP_NR_VCPUS);
    if (r > 0)
        maxvcpus = r;

    close(fd);
    return maxvcpus;
}


2570
static int qemudGetMaxVCPUs(virConnectPtr conn, const char *type) {
2571 2572 2573
    if (!type)
        return 16;

2574
    if (STRCASEEQ(type, "qemu"))
2575 2576
        return 16;

2577
    if (STRCASEEQ(type, "kvm"))
2578
        return kvmGetMaxVCPUs();
2579

2580
    if (STRCASEEQ(type, "kqemu"))
2581
        return 1;
2582 2583 2584

    qemudReportError(conn, NULL, NULL, VIR_ERR_INVALID_ARG,
                     _("unknown type '%s'"), type);
2585 2586 2587
    return -1;
}

2588

2589
static char *qemudGetCapabilities(virConnectPtr conn) {
2590
    struct qemud_driver *driver = conn->privateData;
2591
    virCapsPtr caps;
2592
    char *xml = NULL;
2593

2594
    qemuDriverLock(driver);
2595
    if ((caps = qemudCapsInit(qemu_driver->caps)) == NULL) {
2596 2597 2598 2599
        virReportOOMError(conn);
        goto cleanup;
    }

2600 2601 2602
    caps->privateDataAllocFunc = qemuDomainObjPrivateAlloc;
    caps->privateDataFreeFunc = qemuDomainObjPrivateFree;

2603 2604 2605 2606 2607 2608 2609
    if (qemu_driver->securityDriver &&
        qemudSecurityCapsInit(qemu_driver->securityDriver, caps) < 0) {
        virCapabilitiesFree(caps);
        virReportOOMError(conn);
        goto cleanup;
    }

2610
    virCapabilitiesFree(qemu_driver->caps);
2611 2612 2613
    qemu_driver->caps = caps;

    if ((xml = virCapabilitiesFormatXML(driver->caps)) == NULL)
2614
        virReportOOMError(conn);
2615 2616

cleanup:
2617
    qemuDriverUnlock(driver);
2618

2619
    return xml;
2620 2621 2622
}


2623
static int qemudGetProcessInfo(unsigned long long *cpuTime, int *lastCpu, int pid, int tid) {
D
Daniel P. Berrange 已提交
2624 2625
    char proc[PATH_MAX];
    FILE *pidinfo;
2626
    unsigned long long usertime, systime;
2627 2628
    int cpu;
    int ret;
D
Daniel P. Berrange 已提交
2629

2630 2631 2632 2633 2634 2635
    if (tid)
        ret = snprintf(proc, sizeof(proc), "/proc/%d/task/%d/stat", pid, tid);
    else
        ret = snprintf(proc, sizeof(proc), "/proc/%d/stat", pid);
    if (ret >= (int)sizeof(proc)) {
        errno = E2BIG;
D
Daniel P. Berrange 已提交
2636 2637 2638 2639
        return -1;
    }

    if (!(pidinfo = fopen(proc, "r"))) {
2640
        /*printf("cannot read pid info");*/
D
Daniel P. Berrange 已提交
2641
        /* VM probably shut down, so fake 0 */
2642 2643 2644 2645
        if (cpuTime)
            *cpuTime = 0;
        if (lastCpu)
            *lastCpu = 0;
D
Daniel P. Berrange 已提交
2646 2647 2648
        return 0;
    }

2649 2650 2651 2652 2653 2654 2655 2656 2657 2658
    /* See 'man proc' for information about what all these fields are. We're
     * only interested in a very few of them */
    if (fscanf(pidinfo,
               /* pid -> stime */
               "%*d %*s %*c %*d %*d %*d %*d %*d %*u %*u %*u %*u %*u %llu %llu"
               /* cutime -> endcode */
               "%*d %*d %*d %*d %*d %*u %*u %*d %*u %*u %*u %*u"
               /* startstack -> processor */
               "%*u %*u %*u %*u %*u %*u %*u %*u %*u %*u %*d %d",
               &usertime, &systime, &cpu) != 3) {
2659
        fclose(pidinfo);
2660 2661
        VIR_WARN0("cannot parse process status data");
        errno = -EINVAL;
D
Daniel P. Berrange 已提交
2662 2663 2664 2665 2666 2667 2668 2669
        return -1;
    }

    /* We got jiffies
     * We want nanoseconds
     * _SC_CLK_TCK is jiffies per second
     * So calulate thus....
     */
2670 2671 2672 2673 2674
    if (cpuTime)
        *cpuTime = 1000ull * 1000ull * 1000ull * (usertime + systime) / (unsigned long long)sysconf(_SC_CLK_TCK);
    if (lastCpu)
        *lastCpu = cpu;

D
Daniel P. Berrange 已提交
2675

2676 2677
    VIR_DEBUG("Got status for %d/%d user=%llu sys=%llu cpu=%d",
              pid, tid, usertime, systime, cpu);
D
Daniel P. Berrange 已提交
2678 2679 2680 2681 2682 2683 2684

    fclose(pidinfo);

    return 0;
}


2685
static virDomainPtr qemudDomainLookupByID(virConnectPtr conn,
2686
                                          int id) {
2687 2688 2689 2690
    struct qemud_driver *driver = conn->privateData;
    virDomainObjPtr vm;
    virDomainPtr dom = NULL;

2691
    qemuDriverLock(driver);
2692
    vm  = virDomainFindByID(&driver->domains, id);
2693
    qemuDriverUnlock(driver);
2694 2695

    if (!vm) {
2696 2697
        qemudReportError(conn, NULL, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching id %d"), id);
2698
        goto cleanup;
2699 2700
    }

2701
    dom = virGetDomain(conn, vm->def->name, vm->def->uuid);
2702
    if (dom) dom->id = vm->def->id;
2703 2704

cleanup:
2705 2706
    if (vm)
        virDomainObjUnlock(vm);
2707 2708
    return dom;
}
2709

2710
static virDomainPtr qemudDomainLookupByUUID(virConnectPtr conn,
2711
                                            const unsigned char *uuid) {
2712 2713 2714
    struct qemud_driver *driver = conn->privateData;
    virDomainObjPtr vm;
    virDomainPtr dom = NULL;
2715

2716
    qemuDriverLock(driver);
2717
    vm = virDomainFindByUUID(&driver->domains, uuid);
2718 2719
    qemuDriverUnlock(driver);

2720
    if (!vm) {
2721 2722 2723
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(uuid, uuidstr);
        qemudReportError(conn, NULL, NULL, VIR_ERR_NO_DOMAIN,
2724
                         _("no domain with matching uuid '%s'"), uuidstr);
2725
        goto cleanup;
2726 2727
    }

2728
    dom = virGetDomain(conn, vm->def->name, vm->def->uuid);
2729
    if (dom) dom->id = vm->def->id;
2730 2731

cleanup:
2732 2733
    if (vm)
        virDomainObjUnlock(vm);
2734 2735
    return dom;
}
2736

2737
static virDomainPtr qemudDomainLookupByName(virConnectPtr conn,
2738
                                            const char *name) {
2739 2740 2741
    struct qemud_driver *driver = conn->privateData;
    virDomainObjPtr vm;
    virDomainPtr dom = NULL;
2742

2743
    qemuDriverLock(driver);
2744
    vm = virDomainFindByName(&driver->domains, name);
2745 2746
    qemuDriverUnlock(driver);

2747
    if (!vm) {
2748 2749
        qemudReportError(conn, NULL, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching name '%s'"), name);
2750
        goto cleanup;
2751 2752
    }

2753
    dom = virGetDomain(conn, vm->def->name, vm->def->uuid);
2754
    if (dom) dom->id = vm->def->id;
2755 2756

cleanup:
2757 2758
    if (vm)
        virDomainObjUnlock(vm);
2759 2760 2761
    return dom;
}

2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805

static int qemuDomainIsActive(virDomainPtr dom)
{
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr obj;
    int ret = -1;

    qemuDriverLock(driver);
    obj = virDomainFindByUUID(&driver->domains, dom->uuid);
    qemuDriverUnlock(driver);
    if (!obj) {
        qemudReportError(dom->conn, NULL, NULL, VIR_ERR_NO_DOMAIN, NULL);
        goto cleanup;
    }
    ret = virDomainObjIsActive(obj);

cleanup:
    if (obj)
        virDomainObjUnlock(obj);
    return ret;
}

static int qemuDomainIsPersistent(virDomainPtr dom)
{
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr obj;
    int ret = -1;

    qemuDriverLock(driver);
    obj = virDomainFindByUUID(&driver->domains, dom->uuid);
    qemuDriverUnlock(driver);
    if (!obj) {
        qemudReportError(dom->conn, NULL, NULL, VIR_ERR_NO_DOMAIN, NULL);
        goto cleanup;
    }
    ret = obj->persistent;

cleanup:
    if (obj)
        virDomainObjUnlock(obj);
    return ret;
}


2806
static int qemudGetVersion(virConnectPtr conn, unsigned long *version) {
2807 2808 2809
    struct qemud_driver *driver = conn->privateData;
    int ret = -1;

2810
    qemuDriverLock(driver);
2811
    if (qemudExtractVersion(conn, driver) < 0)
2812
        goto cleanup;
2813

2814
    *version = qemu_driver->qemuVersion;
2815 2816 2817
    ret = 0;

cleanup:
2818
    qemuDriverUnlock(driver);
2819
    return ret;
D
Daniel P. Berrange 已提交
2820 2821
}

2822
static int qemudListDomains(virConnectPtr conn, int *ids, int nids) {
2823
    struct qemud_driver *driver = conn->privateData;
2824
    int n;
2825

2826
    qemuDriverLock(driver);
2827
    n = virDomainObjListGetActiveIDs(&driver->domains, ids, nids);
2828
    qemuDriverUnlock(driver);
2829

2830
    return n;
D
Daniel P. Berrange 已提交
2831
}
2832

2833
static int qemudNumDomains(virConnectPtr conn) {
2834
    struct qemud_driver *driver = conn->privateData;
2835
    int n;
2836

2837
    qemuDriverLock(driver);
2838
    n = virDomainObjListNumOfDomains(&driver->domains, 1);
2839
    qemuDriverUnlock(driver);
2840

2841
    return n;
D
Daniel P. Berrange 已提交
2842
}
2843

2844
static virDomainPtr qemudDomainCreate(virConnectPtr conn, const char *xml,
2845
                                      unsigned int flags ATTRIBUTE_UNUSED) {
2846
    struct qemud_driver *driver = conn->privateData;
2847
    virDomainDefPtr def;
2848
    virDomainObjPtr vm = NULL;
2849
    virDomainPtr dom = NULL;
2850
    virDomainEventPtr event = NULL;
D
Daniel P. Berrange 已提交
2851

2852
    qemuDriverLock(driver);
2853 2854
    if (!(def = virDomainDefParseString(conn, driver->caps, xml,
                                        VIR_DOMAIN_XML_INACTIVE)))
2855
        goto cleanup;
2856

2857 2858 2859
    if (virSecurityDriverVerify(conn, def) < 0)
        goto cleanup;

2860 2861
    if (virDomainObjIsDuplicate(&driver->domains, def, 1) < 0)
        goto cleanup;
2862

2863
    if (!(vm = virDomainAssignDef(conn,
2864
                                  driver->caps,
2865
                                  &driver->domains,
2866 2867 2868 2869
                                  def)))
        goto cleanup;

    def = NULL;
D
Daniel P. Berrange 已提交
2870

2871 2872 2873
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup; /* XXXX free the 'vm' we created ? */

2874
    if (qemudStartVMDaemon(conn, driver, vm, NULL, -1) < 0) {
2875
        qemuDomainObjEndJob(vm);
2876 2877
        virDomainRemoveInactive(&driver->domains,
                                vm);
2878
        vm = NULL;
2879
        goto endjob;
D
Daniel P. Berrange 已提交
2880
    }
2881 2882 2883 2884

    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STARTED,
                                     VIR_DOMAIN_EVENT_STARTED_BOOTED);
D
Daniel P. Berrange 已提交
2885

2886
    dom = virGetDomain(conn, vm->def->name, vm->def->uuid);
2887
    if (dom) dom->id = vm->def->id;
2888

2889 2890 2891 2892
endjob:
    if (vm)
        qemuDomainObjEndJob(vm);

2893 2894
cleanup:
    virDomainDefFree(def);
2895 2896
    if (vm)
        virDomainObjUnlock(vm);
2897 2898
    if (event)
        qemuDomainEventQueue(driver, event);
2899
    qemuDriverUnlock(driver);
2900
    return dom;
D
Daniel P. Berrange 已提交
2901 2902 2903
}


2904
static int qemudDomainSuspend(virDomainPtr dom) {
2905 2906 2907
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int ret = -1;
2908
    virDomainEventPtr event = NULL;
2909

2910
    qemuDriverLock(driver);
2911
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
2912

D
Daniel P. Berrange 已提交
2913
    if (!vm) {
2914 2915 2916 2917
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
2918
        goto cleanup;
D
Daniel P. Berrange 已提交
2919
    }
2920 2921 2922
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
2923
    if (!virDomainObjIsActive(vm)) {
2924
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
2925
                         "%s", _("domain is not running"));
2926
        goto endjob;
D
Daniel P. Berrange 已提交
2927
    }
2928
    if (vm->state != VIR_DOMAIN_PAUSED) {
2929
        qemuDomainObjPrivatePtr priv = vm->privateData;
2930
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
2931
        if (qemuMonitorStopCPUs(priv->mon) < 0) {
2932 2933
            qemuDomainObjExitMonitorWithDriver(driver, vm);
            goto endjob;
2934
        }
2935
        qemuDomainObjExitMonitorWithDriver(driver, vm);
2936
        vm->state = VIR_DOMAIN_PAUSED;
2937 2938 2939
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);
D
Daniel P. Berrange 已提交
2940
    }
2941
    if (virDomainSaveStatus(dom->conn, driver->stateDir, vm) < 0)
2942
        goto endjob;
2943 2944
    ret = 0;

2945 2946 2947
endjob:
    qemuDomainObjEndJob(vm);

2948
cleanup:
2949 2950
    if (vm)
        virDomainObjUnlock(vm);
2951

2952
    if (event)
2953
        qemuDomainEventQueue(driver, event);
2954
    qemuDriverUnlock(driver);
2955
    return ret;
D
Daniel P. Berrange 已提交
2956 2957 2958
}


2959
static int qemudDomainResume(virDomainPtr dom) {
2960 2961 2962
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int ret = -1;
2963
    virDomainEventPtr event = NULL;
2964

2965
    qemuDriverLock(driver);
2966
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
2967

D
Daniel P. Berrange 已提交
2968
    if (!vm) {
2969 2970 2971 2972
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
2973
        goto cleanup;
D
Daniel P. Berrange 已提交
2974
    }
2975 2976 2977 2978

    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
2979
    if (!virDomainObjIsActive(vm)) {
2980
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
2981
                         "%s", _("domain is not running"));
2982
        goto endjob;
D
Daniel P. Berrange 已提交
2983
    }
2984
    if (vm->state == VIR_DOMAIN_PAUSED) {
2985
        qemuDomainObjPrivatePtr priv = vm->privateData;
2986
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
2987
        if (qemuMonitorStartCPUs(priv->mon, dom->conn) < 0) {
2988
            qemuDomainObjExitMonitorWithDriver(driver, vm);
2989 2990 2991
            if (virGetLastError() == NULL)
                qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
                                 "%s", _("resume operation failed"));
2992
            goto endjob;
2993
        }
2994
        qemuDomainObjExitMonitorWithDriver(driver, vm);
2995
        vm->state = VIR_DOMAIN_RUNNING;
2996 2997 2998
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);
D
Daniel P. Berrange 已提交
2999
    }
3000
    if (virDomainSaveStatus(dom->conn, driver->stateDir, vm) < 0)
3001
        goto endjob;
3002 3003
    ret = 0;

3004 3005 3006
endjob:
    qemuDomainObjEndJob(vm);

3007
cleanup:
3008 3009
    if (vm)
        virDomainObjUnlock(vm);
3010
    if (event)
3011
        qemuDomainEventQueue(driver, event);
3012
    qemuDriverUnlock(driver);
3013
    return ret;
D
Daniel P. Berrange 已提交
3014 3015 3016
}


3017
static int qemudDomainShutdown(virDomainPtr dom) {
3018 3019 3020
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int ret = -1;
3021

3022
    qemuDriverLock(driver);
3023
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3024 3025
    qemuDriverUnlock(driver);

3026
    if (!vm) {
3027 3028 3029 3030
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
3031
        goto cleanup;
3032 3033
    }

3034 3035 3036
    if (qemuDomainObjBeginJob(vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
3037
    if (!virDomainObjIsActive(vm)) {
3038 3039
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
                         "%s", _("domain is not running"));
3040
        goto endjob;
3041 3042
    }

3043
    qemuDomainObjPrivatePtr priv = vm->privateData;
3044 3045 3046
    qemuDomainObjEnterMonitor(vm);
    ret = qemuMonitorSystemPowerdown(priv->mon);
    qemuDomainObjExitMonitor(vm);
3047

3048 3049 3050
endjob:
    qemuDomainObjEndJob(vm);

3051
cleanup:
3052 3053
    if (vm)
        virDomainObjUnlock(vm);
3054
    return ret;
3055 3056 3057
}


3058
static int qemudDomainDestroy(virDomainPtr dom) {
3059 3060 3061
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int ret = -1;
3062
    virDomainEventPtr event = NULL;
3063

3064
    qemuDriverLock(driver);
3065
    vm  = virDomainFindByUUID(&driver->domains, dom->uuid);
D
Daniel P. Berrange 已提交
3066
    if (!vm) {
3067 3068 3069 3070
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
3071
        goto cleanup;
D
Daniel P. Berrange 已提交
3072
    }
3073 3074 3075 3076

    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
3077
    if (!virDomainObjIsActive(vm)) {
3078
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
3079
                         "%s", _("domain is not running"));
3080
        goto endjob;
3081
    }
3082

3083
    qemudShutdownVMDaemon(dom->conn, driver, vm);
3084 3085 3086
    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_DESTROYED);
3087
    if (!vm->persistent) {
3088
        qemuDomainObjEndJob(vm);
3089 3090
        virDomainRemoveInactive(&driver->domains,
                                vm);
3091 3092
        vm = NULL;
    }
3093 3094
    ret = 0;

3095 3096 3097 3098
endjob:
    if (vm)
        qemuDomainObjEndJob(vm);

3099
cleanup:
3100 3101
    if (vm)
        virDomainObjUnlock(vm);
3102 3103
    if (event)
        qemuDomainEventQueue(driver, event);
3104
    qemuDriverUnlock(driver);
3105
    return ret;
D
Daniel P. Berrange 已提交
3106 3107 3108
}


3109
static char *qemudDomainGetOSType(virDomainPtr dom) {
3110 3111 3112
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    char *type = NULL;
3113

3114
    qemuDriverLock(driver);
3115
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3116
    qemuDriverUnlock(driver);
3117
    if (!vm) {
3118 3119 3120 3121
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
3122
        goto cleanup;
3123 3124
    }

3125
    if (!(type = strdup(vm->def->os.type)))
3126
        virReportOOMError(dom->conn);
3127 3128

cleanup:
3129 3130
    if (vm)
        virDomainObjUnlock(vm);
3131 3132 3133
    return type;
}

3134 3135
/* Returns max memory in kb, 0 if error */
static unsigned long qemudDomainGetMaxMemory(virDomainPtr dom) {
3136 3137 3138
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    unsigned long ret = 0;
3139

3140
    qemuDriverLock(driver);
3141
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3142 3143
    qemuDriverUnlock(driver);

3144
    if (!vm) {
3145 3146
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
3147
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
3148
                         _("no domain with matching uuid '%s'"), uuidstr);
3149
        goto cleanup;
3150 3151
    }

3152 3153 3154
    ret = vm->def->maxmem;

cleanup:
3155 3156
    if (vm)
        virDomainObjUnlock(vm);
3157
    return ret;
3158 3159 3160
}

static int qemudDomainSetMaxMemory(virDomainPtr dom, unsigned long newmax) {
3161 3162 3163
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int ret = -1;
3164

3165
    qemuDriverLock(driver);
3166
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3167 3168
    qemuDriverUnlock(driver);

3169
    if (!vm) {
3170 3171
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
3172
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
3173
                         _("no domain with matching uuid '%s'"), uuidstr);
3174
        goto cleanup;
3175 3176 3177 3178
    }

    if (newmax < vm->def->memory) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INVALID_ARG,
3179
                         "%s", _("cannot set max memory lower than current memory"));
3180
        goto cleanup;;
3181 3182 3183
    }

    vm->def->maxmem = newmax;
3184 3185 3186
    ret = 0;

cleanup:
3187 3188
    if (vm)
        virDomainObjUnlock(vm);
3189
    return ret;
3190 3191
}

3192

3193
static int qemudDomainSetMemory(virDomainPtr dom, unsigned long newmem) {
3194 3195 3196
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int ret = -1;
3197

3198
    qemuDriverLock(driver);
3199
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3200
    qemuDriverUnlock(driver);
3201
    if (!vm) {
3202 3203
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
3204
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
3205
                         _("no domain with matching uuid '%s'"), uuidstr);
3206
        goto cleanup;
3207 3208 3209 3210
    }

    if (newmem > vm->def->maxmem) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INVALID_ARG,
3211
                         "%s", _("cannot set memory higher than max memory"));
3212
        goto cleanup;
3213 3214
    }

3215 3216 3217
    if (qemuDomainObjBeginJob(vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
3218
    if (virDomainObjIsActive(vm)) {
3219
        qemuDomainObjPrivatePtr priv = vm->privateData;
3220
        qemuDomainObjEnterMonitor(vm);
3221
        int r = qemuMonitorSetBalloon(priv->mon, newmem);
3222
        qemuDomainObjExitMonitor(vm);
3223
        if (r < 0)
3224
            goto endjob;
3225 3226 3227

        /* Lack of balloon support is a fatal error */
        if (r == 0) {
3228 3229
            qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                             "%s", _("cannot set memory of an active domain"));
3230
            goto endjob;
3231
        }
3232 3233 3234
    } else {
        vm->def->memory = newmem;
    }
3235
    ret = 0;
3236

3237 3238 3239
endjob:
    qemuDomainObjEndJob(vm);

3240
cleanup:
3241 3242
    if (vm)
        virDomainObjUnlock(vm);
3243
    return ret;
3244 3245
}

3246
static int qemudDomainGetInfo(virDomainPtr dom,
3247
                              virDomainInfoPtr info) {
3248 3249 3250
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int ret = -1;
3251 3252
    int err;
    unsigned long balloon;
3253

3254
    qemuDriverLock(driver);
3255
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3256
    qemuDriverUnlock(driver);
D
Daniel P. Berrange 已提交
3257
    if (!vm) {
3258 3259 3260 3261
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
3262
        goto cleanup;
D
Daniel P. Berrange 已提交
3263 3264
    }

3265
    info->state = vm->state;
D
Daniel P. Berrange 已提交
3266

D
Daniel P. Berrange 已提交
3267
    if (!virDomainObjIsActive(vm)) {
3268
        info->cpuTime = 0;
D
Daniel P. Berrange 已提交
3269
    } else {
3270
        if (qemudGetProcessInfo(&(info->cpuTime), NULL, vm->pid, 0) < 0) {
3271
            qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED, ("cannot read cputime for domain"));
3272
            goto cleanup;
D
Daniel P. Berrange 已提交
3273 3274 3275
        }
    }

3276
    info->maxMem = vm->def->maxmem;
3277

D
Daniel P. Berrange 已提交
3278
    if (virDomainObjIsActive(vm)) {
3279
        qemuDomainObjPrivatePtr priv = vm->privateData;
3280 3281 3282
        if (!priv->jobActive) {
            if (qemuDomainObjBeginJob(vm) < 0)
                goto cleanup;
3283

3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301
            qemuDomainObjEnterMonitor(vm);
            err = qemuMonitorGetBalloonInfo(priv->mon, &balloon);
            qemuDomainObjExitMonitor(vm);
            if (err < 0) {
                qemuDomainObjEndJob(vm);
                goto cleanup;
            }

            if (err == 0)
                /* Balloon not supported, so maxmem is always the allocation */
                info->memory = vm->def->maxmem;
            else
                info->memory = balloon;

            qemuDomainObjEndJob(vm);
        } else {
            info->memory = vm->def->memory;
        }
3302 3303 3304 3305
    } else {
        info->memory = vm->def->memory;
    }

3306
    info->nrVirtCpu = vm->def->vcpus;
3307 3308 3309
    ret = 0;

cleanup:
3310 3311
    if (vm)
        virDomainObjUnlock(vm);
3312
    return ret;
D
Daniel P. Berrange 已提交
3313 3314 3315
}


3316
#define QEMUD_SAVE_MAGIC "LibvirtQemudSave"
3317 3318 3319
#define QEMUD_SAVE_VERSION 2

enum qemud_save_formats {
3320 3321 3322
    QEMUD_SAVE_FORMAT_RAW = 0,
    QEMUD_SAVE_FORMAT_GZIP = 1,
    QEMUD_SAVE_FORMAT_BZIP2 = 2,
3323 3324
    /*
     * Deprecated by xz and never used as part of a release
3325
     * QEMUD_SAVE_FORMAT_LZMA
3326 3327
     */
    QEMUD_SAVE_FORMAT_XZ = 3,
3328
    QEMUD_SAVE_FORMAT_LZOP = 4,
3329 3330 3331
    /* Note: add new members only at the end.
       These values are used in the on-disk format.
       Do not change or re-use numbers. */
3332 3333

    QEMUD_SAVE_FORMAT_LAST
3334
};
3335

3336 3337 3338 3339 3340
VIR_ENUM_DECL(qemudSaveCompression)
VIR_ENUM_IMPL(qemudSaveCompression, QEMUD_SAVE_FORMAT_LAST,
              "raw",
              "gzip",
              "bzip2",
3341 3342
              "xz",
              "lzop")
3343

3344 3345 3346 3347 3348
struct qemud_save_header {
    char magic[sizeof(QEMUD_SAVE_MAGIC)-1];
    int version;
    int xml_len;
    int was_running;
3349 3350
    int compressed;
    int unused[15];
3351 3352
};

3353
static int qemudDomainSave(virDomainPtr dom,
3354 3355
                           const char *path)
{
3356
    struct qemud_driver *driver = dom->conn->privateData;
3357
    virDomainObjPtr vm = NULL;
3358 3359
    int fd = -1;
    char *xml = NULL;
3360
    struct qemud_save_header header;
3361
    int ret = -1;
3362
    int rc;
3363
    virDomainEventPtr event = NULL;
3364 3365 3366 3367 3368

    memset(&header, 0, sizeof(header));
    memcpy(header.magic, QEMUD_SAVE_MAGIC, sizeof(header.magic));
    header.version = QEMUD_SAVE_VERSION;

3369
    qemuDriverLock(driver);
3370 3371 3372
    if (driver->saveImageFormat == NULL)
        header.compressed = QEMUD_SAVE_FORMAT_RAW;
    else {
3373 3374 3375 3376 3377 3378
        header.compressed =
            qemudSaveCompressionTypeFromString(driver->saveImageFormat);
        if (header.compressed < 0) {
            qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
                             "%s", _("Invalid save image format specified "
                                     "in configuration file"));
3379
            goto cleanup;
3380
        }
3381 3382
    }

3383
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3384

D
Daniel P. Berrange 已提交
3385
    if (!vm) {
3386 3387 3388 3389
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
3390
        goto cleanup;
D
Daniel P. Berrange 已提交
3391
    }
3392

3393 3394 3395
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
3396
    if (!virDomainObjIsActive(vm)) {
3397
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
3398
                         "%s", _("domain is not running"));
3399
        goto endjob;
D
Daniel P. Berrange 已提交
3400
    }
3401 3402 3403

    /* Pause */
    if (vm->state == VIR_DOMAIN_RUNNING) {
3404
        qemuDomainObjPrivatePtr priv = vm->privateData;
3405
        header.was_running = 1;
3406
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
3407
        if (qemuMonitorStopCPUs(priv->mon) < 0) {
3408 3409
            qemuDomainObjExitMonitorWithDriver(driver, vm);
            goto endjob;
3410
        }
3411
        qemuDomainObjExitMonitorWithDriver(driver, vm);
3412
        vm->state = VIR_DOMAIN_PAUSED;
3413 3414 3415
    }

    /* Get XML for the domain */
3416
    xml = virDomainDefFormat(dom->conn, vm->def, VIR_DOMAIN_XML_SECURE);
3417 3418
    if (!xml) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
3419
                         "%s", _("failed to get domain xml"));
3420
        goto endjob;
3421 3422 3423 3424 3425 3426
    }
    header.xml_len = strlen(xml) + 1;

    /* Write header to file, followed by XML */
    if ((fd = open(path, O_CREAT|O_TRUNC|O_WRONLY, S_IRUSR|S_IWUSR)) < 0) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
3427
                         _("failed to create '%s'"), path);
3428
        goto endjob;
3429 3430 3431 3432
    }

    if (safewrite(fd, &header, sizeof(header)) != sizeof(header)) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
3433
                         "%s", _("failed to write save header"));
3434
        goto endjob;
3435 3436 3437 3438
    }

    if (safewrite(fd, xml, header.xml_len) != header.xml_len) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
3439
                         "%s", _("failed to write xml"));
3440
        goto endjob;
3441 3442
    }

3443
    if (close(fd) < 0) {
3444 3445 3446
        virReportSystemError(dom->conn, errno,
                             _("unable to save file %s"),
                             path);
3447
        goto endjob;
3448 3449
    }
    fd = -1;
3450

3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463
    if (driver->privileged &&
        chown(path, driver->user, driver->group) < 0) {
        virReportSystemError(NULL, errno,
                             _("unable to set ownership of '%s' to user %d:%d"),
                             path, driver->user, driver->group);
        goto endjob;
    }

    if (driver->securityDriver &&
        driver->securityDriver->domainSetSavedStateLabel &&
        driver->securityDriver->domainSetSavedStateLabel(dom->conn, vm, path) == -1)
        goto endjob;

3464 3465
    if (header.compressed == QEMUD_SAVE_FORMAT_RAW) {
        const char *args[] = { "cat", NULL };
3466
        qemuDomainObjPrivatePtr priv = vm->privateData;
3467
        qemuDomainObjEnterMonitor(vm);
3468
        rc = qemuMonitorMigrateToCommand(priv->mon, 0, args, path);
3469
        qemuDomainObjExitMonitor(vm);
3470
    } else {
3471
        const char *prog = qemudSaveCompressionTypeToString(header.compressed);
3472
        qemuDomainObjPrivatePtr priv = vm->privateData;
3473 3474 3475 3476 3477
        const char *args[] = {
            prog,
            "-c",
            NULL
        };
3478
        qemuDomainObjEnterMonitor(vm);
3479
        rc = qemuMonitorMigrateToCommand(priv->mon, 0, args, path);
3480
        qemuDomainObjExitMonitor(vm);
3481 3482
    }

3483
    if (rc < 0)
3484
        goto endjob;
3485

3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500
    if (driver->privileged &&
        chown(path, 0, 0) < 0) {
        virReportSystemError(NULL, errno,
                             _("unable to set ownership of '%s' to user %d:%d"),
                             path, 0, 0);
        goto endjob;
    }

    if (driver->securityDriver &&
        driver->securityDriver->domainRestoreSavedStateLabel &&
        driver->securityDriver->domainRestoreSavedStateLabel(dom->conn, path) == -1)
        goto endjob;

    ret = 0;

3501 3502
    /* Shut it down */
    qemudShutdownVMDaemon(dom->conn, driver, vm);
3503 3504 3505
    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_SAVED);
3506
    if (!vm->persistent) {
3507
        qemuDomainObjEndJob(vm);
3508 3509
        virDomainRemoveInactive(&driver->domains,
                                vm);
3510 3511
        vm = NULL;
    }
3512

3513 3514 3515 3516
endjob:
    if (vm)
        qemuDomainObjEndJob(vm);

3517 3518 3519 3520 3521 3522
cleanup:
    if (fd != -1)
        close(fd);
    VIR_FREE(xml);
    if (ret != 0)
        unlink(path);
3523 3524
    if (vm)
        virDomainObjUnlock(vm);
3525 3526
    if (event)
        qemuDomainEventQueue(driver, event);
3527
    qemuDriverUnlock(driver);
3528
    return ret;
D
Daniel P. Berrange 已提交
3529 3530 3531
}


P
Paolo Bonzini 已提交
3532 3533 3534 3535 3536 3537 3538
static int qemudDomainCoreDump(virDomainPtr dom,
                               const char *path,
                               int flags ATTRIBUTE_UNUSED) {
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int resume = 0, paused = 0;
    int ret = -1;
3539 3540 3541 3542
    const char *args[] = {
        "cat",
        NULL,
    };
P
Paolo Bonzini 已提交
3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555

    qemuDriverLock(driver);
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
    qemuDriverUnlock(driver);

    if (!vm) {
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
        goto cleanup;
    }

3556 3557 3558
    if (qemuDomainObjBeginJob(vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
3559
    if (!virDomainObjIsActive(vm)) {
P
Paolo Bonzini 已提交
3560 3561
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
                         "%s", _("domain is not running"));
3562
        goto endjob;
P
Paolo Bonzini 已提交
3563 3564 3565 3566 3567 3568 3569
    }

    /* Migrate will always stop the VM, so once we support live dumping
       the resume condition will stay the same, independent of whether
       the stop command is issued.  */
    resume = (vm->state == VIR_DOMAIN_RUNNING);

3570 3571
    qemuDomainObjPrivatePtr priv = vm->privateData;

P
Paolo Bonzini 已提交
3572 3573
    /* Pause domain for non-live dump */
    if (vm->state == VIR_DOMAIN_RUNNING) {
3574 3575 3576
        qemuDomainObjEnterMonitor(vm);
        if (qemuMonitorStopCPUs(priv->mon) < 0) {
            qemuDomainObjExitMonitor(vm);
3577
            goto endjob;
3578 3579
        }
        qemuDomainObjExitMonitor(vm);
P
Paolo Bonzini 已提交
3580 3581 3582
        paused = 1;
    }

3583
    qemuDomainObjEnterMonitor(vm);
3584
    ret = qemuMonitorMigrateToCommand(priv->mon, 0, args, path);
3585
    qemuDomainObjExitMonitor(vm);
P
Paolo Bonzini 已提交
3586 3587 3588 3589 3590 3591
    paused = 1;

    /* Since the monitor is always attached to a pty for libvirt, it
       will support synchronous operations so we always get here after
       the migration is complete.  */
    if (resume && paused) {
3592
        qemuDomainObjEnterMonitor(vm);
3593
        if (qemuMonitorStartCPUs(priv->mon, dom->conn) < 0) {
3594 3595 3596
            if (virGetLastError() == NULL)
                qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
                                 "%s", _("resuming after dump failed"));
P
Paolo Bonzini 已提交
3597
        }
3598
        qemuDomainObjExitMonitor(vm);
P
Paolo Bonzini 已提交
3599
    }
3600 3601 3602 3603 3604

endjob:
    qemuDomainObjEndJob(vm);

cleanup:
P
Paolo Bonzini 已提交
3605 3606 3607 3608 3609 3610
    if (vm)
        virDomainObjUnlock(vm);
    return ret;
}


3611
static int qemudDomainSetVcpus(virDomainPtr dom, unsigned int nvcpus) {
3612 3613
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
3614
    int max;
3615
    int ret = -1;
3616
    const char *type;
3617

3618
    qemuDriverLock(driver);
3619
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3620 3621
    qemuDriverUnlock(driver);

3622
    if (!vm) {
3623 3624
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
3625
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
3626
                         _("no domain with matching uuid '%s'"), uuidstr);
3627
        goto cleanup;
3628 3629
    }

3630 3631 3632
    if (qemuDomainObjBeginJob(vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
3633
    if (virDomainObjIsActive(vm)) {
3634
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID, "%s",
3635
                         _("cannot change vcpu count of an active domain"));
3636
        goto endjob;
3637 3638
    }

3639 3640 3641 3642
    if (!(type = virDomainVirtTypeToString(vm->def->virtType))) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("unknown virt type in domain definition '%d'"),
                         vm->def->virtType);
3643
        goto endjob;
3644 3645 3646
    }

    if ((max = qemudGetMaxVCPUs(dom->conn, type)) < 0) {
3647 3648
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR, "%s",
                         _("could not determine max vcpus for the domain"));
3649
        goto endjob;
3650 3651 3652 3653 3654 3655
    }

    if (nvcpus > max) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INVALID_ARG,
                         _("requested vcpus is greater than max allowable"
                           " vcpus for the domain: %d > %d"), nvcpus, max);
3656
        goto endjob;
3657 3658 3659
    }

    vm->def->vcpus = nvcpus;
3660 3661
    ret = 0;

3662 3663 3664
endjob:
    qemuDomainObjEndJob(vm);

3665
cleanup:
3666 3667
    if (vm)
        virDomainObjUnlock(vm);
3668
    return ret;
3669 3670
}

3671 3672 3673 3674 3675 3676

static int
qemudDomainPinVcpu(virDomainPtr dom,
                   unsigned int vcpu,
                   unsigned char *cpumap,
                   int maplen) {
3677 3678
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
3679
    int maxcpu, hostcpus;
3680
    virNodeInfo nodeinfo;
3681
    int ret = -1;
3682

3683
    qemuDriverLock(driver);
3684
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3685 3686
    qemuDriverUnlock(driver);

3687 3688 3689 3690 3691 3692 3693 3694
    if (!vm) {
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
        goto cleanup;
    }

D
Daniel P. Berrange 已提交
3695
    if (!virDomainObjIsActive(vm)) {
3696
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
3697
                         "%s",_("cannot pin vcpus on an inactive domain"));
3698
        goto cleanup;
3699 3700 3701 3702 3703 3704
    }

    if (vcpu > (vm->nvcpupids-1)) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INVALID_ARG,
                         _("vcpu number out of range %d > %d"),
                         vcpu, vm->nvcpupids);
3705
        goto cleanup;
3706 3707
    }

3708
    if (nodeGetInfo(dom->conn, &nodeinfo) < 0)
3709
        goto cleanup;
3710

3711
    hostcpus = VIR_NODEINFO_MAXCPUS(nodeinfo);
3712
    maxcpu = maplen * 8;
3713 3714
    if (maxcpu > hostcpus)
        maxcpu = hostcpus;
3715 3716

    if (vm->vcpupids != NULL) {
3717 3718
        if (virProcessInfoSetAffinity(vm->vcpupids[vcpu],
                                      cpumap, maplen, maxcpu) < 0)
3719
            goto cleanup;
3720 3721 3722
    } else {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                         "%s", _("cpu affinity is not supported"));
3723
        goto cleanup;
3724
    }
3725
    ret = 0;
3726

3727
cleanup:
3728 3729
    if (vm)
        virDomainObjUnlock(vm);
3730
    return ret;
3731 3732 3733 3734 3735 3736 3737 3738
}

static int
qemudDomainGetVcpus(virDomainPtr dom,
                    virVcpuInfoPtr info,
                    int maxinfo,
                    unsigned char *cpumaps,
                    int maplen) {
3739 3740
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
3741
    virNodeInfo nodeinfo;
3742
    int i, v, maxcpu, hostcpus;
3743
    int ret = -1;
3744

3745
    qemuDriverLock(driver);
3746
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3747 3748
    qemuDriverUnlock(driver);

3749 3750 3751 3752 3753 3754 3755 3756
    if (!vm) {
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
        goto cleanup;
    }

D
Daniel P. Berrange 已提交
3757
    if (!virDomainObjIsActive(vm)) {
3758
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
3759 3760
                         "%s",
                         _("cannot list vcpu pinning for an inactive domain"));
3761
        goto cleanup;
3762 3763
    }

3764
    if (nodeGetInfo(dom->conn, &nodeinfo) < 0)
3765
        goto cleanup;
3766

3767
    hostcpus = VIR_NODEINFO_MAXCPUS(nodeinfo);
3768
    maxcpu = maplen * 8;
3769 3770
    if (maxcpu > hostcpus)
        maxcpu = hostcpus;
3771 3772 3773 3774 3775

    /* Clamp to actual number of vcpus */
    if (maxinfo > vm->nvcpupids)
        maxinfo = vm->nvcpupids;

3776 3777 3778 3779 3780 3781
    if (maxinfo >= 1) {
        if (info != NULL) {
            memset(info, 0, sizeof(*info) * maxinfo);
            for (i = 0 ; i < maxinfo ; i++) {
                info[i].number = i;
                info[i].state = VIR_VCPU_RUNNING;
3782 3783 3784 3785 3786 3787 3788 3789 3790 3791

                if (vm->vcpupids != NULL &&
                    qemudGetProcessInfo(&(info[i].cpuTime),
                                        &(info[i].cpu),
                                        vm->pid,
                                        vm->vcpupids[i]) < 0) {
                    virReportSystemError(dom->conn, errno, "%s",
                                         _("cannot get vCPU placement & pCPU time"));
                    goto cleanup;
                }
3792
            }
3793 3794
        }

3795 3796 3797 3798 3799 3800
        if (cpumaps != NULL) {
            memset(cpumaps, 0, maplen * maxinfo);
            if (vm->vcpupids != NULL) {
                for (v = 0 ; v < maxinfo ; v++) {
                    unsigned char *cpumap = VIR_GET_CPUMAP(cpumaps, maplen, v);

3801 3802
                    if (virProcessInfoGetAffinity(vm->vcpupids[v],
                                                  cpumap, maplen, maxcpu) < 0)
3803
                        goto cleanup;
3804
                }
3805 3806 3807 3808
            } else {
                qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                                 "%s", _("cpu affinity is not available"));
                goto cleanup;
3809 3810 3811
            }
        }
    }
3812
    ret = maxinfo;
3813

3814
cleanup:
3815 3816
    if (vm)
        virDomainObjUnlock(vm);
3817
    return ret;
3818 3819 3820
}


3821
static int qemudDomainGetMaxVcpus(virDomainPtr dom) {
3822 3823
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
3824
    const char *type;
3825
    int ret = -1;
3826

3827
    qemuDriverLock(driver);
3828
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
3829 3830
    qemuDriverUnlock(driver);

3831
    if (!vm) {
3832 3833
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
3834
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
3835
                         _("no domain with matching uuid '%s'"), uuidstr);
3836
        goto cleanup;
3837 3838
    }

3839
    if (!(type = virDomainVirtTypeToString(vm->def->virtType))) {
3840 3841 3842
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("unknown virt type in domain definition '%d'"),
                         vm->def->virtType);
3843
        goto cleanup;
3844 3845
    }

3846
    ret = qemudGetMaxVCPUs(dom->conn, type);
3847

3848
cleanup:
3849 3850
    if (vm)
        virDomainObjUnlock(vm);
3851 3852 3853
    return ret;
}

3854 3855 3856 3857 3858 3859 3860 3861 3862 3863
static int qemudDomainGetSecurityLabel(virDomainPtr dom, virSecurityLabelPtr seclabel)
{
    struct qemud_driver *driver = (struct qemud_driver *)dom->conn->privateData;
    virDomainObjPtr vm;
    const char *type;
    int ret = -1;

    qemuDriverLock(driver);
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);

3864 3865
    memset(seclabel, 0, sizeof(*seclabel));

3866 3867 3868
    if (!vm) {
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
3869
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894
                         _("no domain with matching uuid '%s'"), uuidstr);
        goto cleanup;
    }

    if (!(type = virDomainVirtTypeToString(vm->def->virtType))) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("unknown virt type in domain definition '%d'"),
                         vm->def->virtType);
        goto cleanup;
    }

    /*
     * Theoretically, the pid can be replaced during this operation and
     * return the label of a different process.  If atomicity is needed,
     * further validation will be required.
     *
     * Comment from Dan Berrange:
     *
     *   Well the PID as stored in the virDomainObjPtr can't be changed
     *   because you've got a locked object.  The OS level PID could have
     *   exited, though and in extreme circumstances have cycled through all
     *   PIDs back to ours. We could sanity check that our PID still exists
     *   after reading the label, by checking that our FD connecting to the
     *   QEMU monitor hasn't seen SIGHUP/ERR on poll().
     */
D
Daniel P. Berrange 已提交
3895
    if (virDomainObjIsActive(vm)) {
3896 3897 3898
        if (driver->securityDriver && driver->securityDriver->domainGetSecurityLabel) {
            if (driver->securityDriver->domainGetSecurityLabel(dom->conn, vm, seclabel) == -1) {
                qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
3899
                                 "%s", _("Failed to get security label"));
3900 3901 3902 3903 3904 3905 3906 3907 3908 3909
                goto cleanup;
            }
        }
    }

    ret = 0;

cleanup:
    if (vm)
        virDomainObjUnlock(vm);
3910
    qemuDriverUnlock(driver);
3911 3912 3913
    return ret;
}

3914 3915
static int qemudNodeGetSecurityModel(virConnectPtr conn,
                                     virSecurityModelPtr secmodel)
3916 3917 3918
{
    struct qemud_driver *driver = (struct qemud_driver *)conn->privateData;
    char *p;
3919
    int ret = 0;
3920

3921 3922
    qemuDriverLock(driver);
    if (!driver->securityDriver) {
3923
        memset(secmodel, 0, sizeof (*secmodel));
3924 3925
        goto cleanup;
    }
3926

3927 3928 3929 3930 3931
    p = driver->caps->host.secModel.model;
    if (strlen(p) >= VIR_SECURITY_MODEL_BUFLEN-1) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("security model string exceeds max %d bytes"),
                         VIR_SECURITY_MODEL_BUFLEN-1);
3932 3933
        ret = -1;
        goto cleanup;
3934 3935 3936 3937 3938 3939 3940 3941
    }
    strcpy(secmodel->model, p);

    p = driver->caps->host.secModel.doi;
    if (strlen(p) >= VIR_SECURITY_DOI_BUFLEN-1) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("security DOI string exceeds max %d bytes"),
                         VIR_SECURITY_DOI_BUFLEN-1);
3942 3943
        ret = -1;
        goto cleanup;
3944 3945
    }
    strcpy(secmodel->doi, p);
3946 3947 3948 3949

cleanup:
    qemuDriverUnlock(driver);
    return ret;
3950 3951 3952
}

/* TODO: check seclabel restore */
3953
static int qemudDomainRestore(virConnectPtr conn,
3954 3955 3956
                              const char *path) {
    struct qemud_driver *driver = conn->privateData;
    virDomainDefPtr def = NULL;
3957
    virDomainObjPtr vm = NULL;
3958 3959 3960
    int fd = -1;
    int ret = -1;
    char *xml = NULL;
3961
    struct qemud_save_header header;
3962
    virDomainEventPtr event = NULL;
3963 3964 3965
    int intermediatefd = -1;
    pid_t intermediate_pid = -1;
    int childstat;
3966

3967
    qemuDriverLock(driver);
3968 3969 3970
    /* Verify the header and read the XML */
    if ((fd = open(path, O_RDONLY)) < 0) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
3971
                         "%s", _("cannot read domain image"));
3972
        goto cleanup;
3973 3974 3975 3976
    }

    if (saferead(fd, &header, sizeof(header)) != sizeof(header)) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
3977
                         "%s", _("failed to read qemu header"));
3978
        goto cleanup;
3979 3980 3981 3982
    }

    if (memcmp(header.magic, QEMUD_SAVE_MAGIC, sizeof(header.magic)) != 0) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
3983
                         "%s", _("image magic is incorrect"));
3984
        goto cleanup;
3985 3986 3987 3988
    }

    if (header.version > QEMUD_SAVE_VERSION) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
3989
                         _("image version is not supported (%d > %d)"),
3990
                         header.version, QEMUD_SAVE_VERSION);
3991
        goto cleanup;
3992 3993
    }

3994
    if (VIR_ALLOC_N(xml, header.xml_len) < 0) {
3995
        virReportOOMError(conn);
3996
        goto cleanup;
3997 3998 3999 4000
    }

    if (saferead(fd, xml, header.xml_len) != header.xml_len) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
4001
                         "%s", _("failed to read XML"));
4002
        goto cleanup;
4003 4004 4005
    }

    /* Create a domain from this XML */
4006 4007
    if (!(def = virDomainDefParseString(conn, driver->caps, xml,
                                        VIR_DOMAIN_XML_INACTIVE))) {
4008
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
4009
                         "%s", _("failed to parse XML"));
4010
        goto cleanup;
4011 4012
    }

4013 4014
    if (virDomainObjIsDuplicate(&driver->domains, def, 1) < 0)
        goto cleanup;
4015

4016
    if (!(vm = virDomainAssignDef(conn,
4017
                                  driver->caps,
4018 4019
                                  &driver->domains,
                                  def))) {
4020
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
4021
                         "%s", _("failed to assign new VM"));
4022
        goto cleanup;
4023
    }
4024
    def = NULL;
4025

4026 4027 4028
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

4029 4030
    if (header.version == 2) {
        const char *intermediate_argv[3] = { NULL, "-dc", NULL };
4031 4032
        const char *prog = qemudSaveCompressionTypeToString(header.compressed);
        if (prog == NULL) {
4033
            qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
4034
                             _("Invalid compressed save format %d"),
4035
                             header.compressed);
4036
            goto endjob;
4037
        }
4038

4039
        if (header.compressed != QEMUD_SAVE_FORMAT_RAW) {
4040
            intermediate_argv[0] = prog;
4041 4042 4043 4044 4045 4046 4047
            intermediatefd = fd;
            fd = -1;
            if (virExec(conn, intermediate_argv, NULL, NULL,
                        &intermediate_pid, intermediatefd, &fd, NULL, 0) < 0) {
                qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                                 _("Failed to start decompression binary %s"),
                                 intermediate_argv[0]);
4048
                goto endjob;
4049 4050 4051
            }
        }
    }
4052
    /* Set the migration source and start it up. */
4053
    ret = qemudStartVMDaemon(conn, driver, vm, "stdio", fd);
4054 4055 4056 4057 4058 4059 4060
    if (intermediate_pid != -1) {
        /* Wait for intermediate process to exit */
        while (waitpid(intermediate_pid, &childstat, 0) == -1 &&
               errno == EINTR);
    }
    if (intermediatefd != -1)
        close(intermediatefd);
4061
    close(fd);
4062
    fd = -1;
4063
    if (ret < 0) {
4064
        if (!vm->persistent) {
4065
            qemuDomainObjEndJob(vm);
4066 4067
            virDomainRemoveInactive(&driver->domains,
                                    vm);
4068 4069
            vm = NULL;
        }
4070
        goto endjob;
4071 4072
    }

4073 4074 4075
    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STARTED,
                                     VIR_DOMAIN_EVENT_STARTED_RESTORED);
4076

4077 4078
    /* If it was running before, resume it now. */
    if (header.was_running) {
4079
        qemuDomainObjPrivatePtr priv = vm->privateData;
4080
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
4081
        if (qemuMonitorStartCPUs(priv->mon, conn) < 0) {
4082 4083 4084
            if (virGetLastError() == NULL)
                qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                                 "%s", _("failed to resume domain"));
4085 4086
            qemuDomainObjExitMonitorWithDriver(driver,vm);
            goto endjob;
4087
        }
4088
        qemuDomainObjExitMonitorWithDriver(driver, vm);
4089
        vm->state = VIR_DOMAIN_RUNNING;
4090
        virDomainSaveStatus(conn, driver->stateDir, vm);
4091
    }
4092
    ret = 0;
4093

4094 4095 4096 4097
endjob:
    if (vm)
        qemuDomainObjEndJob(vm);

4098 4099 4100 4101 4102
cleanup:
    virDomainDefFree(def);
    VIR_FREE(xml);
    if (fd != -1)
        close(fd);
4103 4104
    if (vm)
        virDomainObjUnlock(vm);
4105 4106
    if (event)
        qemuDomainEventQueue(driver, event);
4107
    qemuDriverUnlock(driver);
4108
    return ret;
D
Daniel P. Berrange 已提交
4109 4110 4111
}


4112
static char *qemudDomainDumpXML(virDomainPtr dom,
4113
                                int flags) {
4114 4115 4116
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    char *ret = NULL;
4117 4118
    unsigned long balloon;
    int err;
4119

4120
    qemuDriverLock(driver);
4121
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
4122 4123
    qemuDriverUnlock(driver);

D
Daniel P. Berrange 已提交
4124
    if (!vm) {
4125 4126 4127 4128
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
4129
        goto cleanup;
D
Daniel P. Berrange 已提交
4130 4131
    }

4132
    /* Refresh current memory based on balloon info */
D
Daniel P. Berrange 已提交
4133
    if (virDomainObjIsActive(vm)) {
4134
        qemuDomainObjPrivatePtr priv = vm->privateData;
4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150
        /* Don't delay if someone's using the monitor, just use
         * existing most recent data instead */
        if (!priv->jobActive) {
            if (qemuDomainObjBeginJob(vm) < 0)
                goto cleanup;

            qemuDomainObjEnterMonitor(vm);
            err = qemuMonitorGetBalloonInfo(priv->mon, &balloon);
            qemuDomainObjExitMonitor(vm);
            qemuDomainObjEndJob(vm);
            if (err < 0)
                goto cleanup;
            if (err > 0)
                vm->def->memory = balloon;
            /* err == 0 indicates no balloon support, so ignore it */
        }
4151
    }
4152

4153 4154 4155 4156 4157 4158
    ret = virDomainDefFormat(dom->conn,
                             (flags & VIR_DOMAIN_XML_INACTIVE) && vm->newDef ?
                             vm->newDef : vm->def,
                             flags);

cleanup:
4159 4160
    if (vm)
        virDomainObjUnlock(vm);
4161
    return ret;
D
Daniel P. Berrange 已提交
4162 4163 4164
}


4165 4166 4167 4168
static char *qemuDomainXMLFromNative(virConnectPtr conn,
                                     const char *format,
                                     const char *config,
                                     unsigned int flags ATTRIBUTE_UNUSED) {
4169
    struct qemud_driver *driver = conn->privateData;
4170 4171 4172 4173 4174 4175 4176 4177 4178
    virDomainDefPtr def = NULL;
    char *xml = NULL;

    if (STRNEQ(format, QEMU_CONFIG_FORMAT_ARGV)) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_INVALID_ARG,
                         _("unsupported config type %s"), format);
        goto cleanup;
    }

4179
    qemuDriverLock(driver);
4180
    def = qemuParseCommandLineString(conn, driver->caps, config);
4181
    qemuDriverUnlock(driver);
4182 4183 4184 4185 4186 4187 4188 4189 4190 4191
    if (!def)
        goto cleanup;

    xml = virDomainDefFormat(conn, def, VIR_DOMAIN_XML_INACTIVE);

cleanup:
    virDomainDefFree(def);
    return xml;
}

4192 4193 4194 4195 4196 4197
static char *qemuDomainXMLToNative(virConnectPtr conn,
                                   const char *format,
                                   const char *xmlData,
                                   unsigned int flags ATTRIBUTE_UNUSED) {
    struct qemud_driver *driver = conn->privateData;
    virDomainDefPtr def = NULL;
4198
    virDomainChrDef monitor_chr;
4199 4200 4201 4202 4203 4204 4205 4206 4207 4208
    const char *emulator;
    unsigned int qemuCmdFlags;
    struct stat sb;
    const char **retargv = NULL;
    const char **retenv = NULL;
    const char **tmp;
    virBuffer buf = VIR_BUFFER_INITIALIZER;
    char *ret = NULL;
    int i;

4209 4210
    qemuDriverLock(driver);

4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 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
    if (STRNEQ(format, QEMU_CONFIG_FORMAT_ARGV)) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_INVALID_ARG,
                         _("unsupported config type %s"), format);
        goto cleanup;
    }

    def = virDomainDefParseString(conn, driver->caps, xmlData, 0);
    if (!def)
        goto cleanup;

    /* Since we're just exporting args, we can't do bridge/network
     * setups, since libvirt will normally create TAP devices
     * directly. We convert those configs into generic 'ethernet'
     * config and assume the user has suitable 'ifup-qemu' scripts
     */
    for (i = 0 ; i < def->nnets ; i++) {
        virDomainNetDefPtr net = def->nets[i];
        if (net->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
            VIR_FREE(net->data.network.name);

            memset(net, 0, sizeof *net);

            net->type = VIR_DOMAIN_NET_TYPE_ETHERNET;
            net->data.ethernet.dev = NULL;
            net->data.ethernet.script = NULL;
            net->data.ethernet.ipaddr = NULL;
        } else if (net->type == VIR_DOMAIN_NET_TYPE_BRIDGE) {
            char *brname = net->data.bridge.brname;
            char *script = net->data.bridge.script;
            char *ipaddr = net->data.bridge.ipaddr;

            memset(net, 0, sizeof *net);

            net->type = VIR_DOMAIN_NET_TYPE_ETHERNET;
            net->data.ethernet.dev = brname;
            net->data.ethernet.script = script;
            net->data.ethernet.ipaddr = ipaddr;
        }
    }
    for (i = 0 ; i < def->ngraphics ; i++) {
        if (def->graphics[i]->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
            def->graphics[i]->data.vnc.autoport)
            def->graphics[i]->data.vnc.port = 5900;
    }
    emulator = def->emulator;

    /* Make sure the binary we are about to try exec'ing exists.
     * Technically we could catch the exec() failure, but that's
     * in a sub-process so its hard to feed back a useful error
     */
    if (stat(emulator, &sb) < 0) {
        virReportSystemError(conn, errno,
                             _("Cannot find QEMU binary %s"),
                             emulator);
        goto cleanup;
    }

    if (qemudExtractVersionInfo(emulator,
                                NULL,
                                &qemuCmdFlags) < 0) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("Cannot determine QEMU argv syntax %s"),
                         emulator);
        goto cleanup;
    }

4277 4278
    if (qemuPrepareMonitorChr(conn, driver, &monitor_chr, def->name) < 0)
        goto cleanup;
4279 4280

    if (qemudBuildCommandLine(conn, driver, def,
4281
                              &monitor_chr, qemuCmdFlags,
4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306
                              &retargv, &retenv,
                              NULL, NULL, /* Don't want it to create TAP devices */
                              NULL) < 0) {
        goto cleanup;
    }

    tmp = retenv;
    while (*tmp) {
        virBufferAdd(&buf, *tmp, strlen(*tmp));
        virBufferAddLit(&buf, " ");
        tmp++;
    }
    tmp = retargv;
    while (*tmp) {
        virBufferAdd(&buf, *tmp, strlen(*tmp));
        virBufferAddLit(&buf, " ");
        tmp++;
    }

    if (virBufferError(&buf))
        goto cleanup;

    ret = virBufferContentAndReset(&buf);

cleanup:
4307
    qemuDriverUnlock(driver);
4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320
    for (tmp = retargv ; tmp && *tmp ; tmp++)
        VIR_FREE(*tmp);
    VIR_FREE(retargv);

    for (tmp = retenv ; tmp && *tmp ; tmp++)
        VIR_FREE(*tmp);
    VIR_FREE(retenv);

    virDomainDefFree(def);
    return ret;
}


4321
static int qemudListDefinedDomains(virConnectPtr conn,
4322
                            char **const names, int nnames) {
4323
    struct qemud_driver *driver = conn->privateData;
4324
    int n;
4325

4326
    qemuDriverLock(driver);
4327
    n = virDomainObjListGetInactiveNames(&driver->domains, names, nnames);
4328
    qemuDriverUnlock(driver);
4329
    return n;
D
Daniel P. Berrange 已提交
4330 4331
}

4332
static int qemudNumDefinedDomains(virConnectPtr conn) {
4333
    struct qemud_driver *driver = conn->privateData;
4334
    int n;
4335

4336
    qemuDriverLock(driver);
4337
    n = virDomainObjListNumOfDomains(&driver->domains, 0);
4338
    qemuDriverUnlock(driver);
4339

4340
    return n;
D
Daniel P. Berrange 已提交
4341 4342 4343
}


4344
static int qemudDomainStart(virDomainPtr dom) {
4345 4346 4347
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int ret = -1;
4348
    virDomainEventPtr event = NULL;
4349

4350
    qemuDriverLock(driver);
4351
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
4352

4353
    if (!vm) {
4354 4355 4356 4357
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
4358
        goto cleanup;
4359 4360
    }

4361 4362 4363 4364 4365 4366 4367 4368 4369
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

    if (virDomainObjIsActive(vm)) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
                         "%s", _("domain is already running"));
        goto endjob;
    }

4370
    ret = qemudStartVMDaemon(dom->conn, driver, vm, NULL, -1);
4371
    if (ret != -1)
4372 4373 4374
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_STARTED,
                                         VIR_DOMAIN_EVENT_STARTED_BOOTED);
4375

4376 4377 4378
endjob:
    qemuDomainObjEndJob(vm);

4379
cleanup:
4380 4381
    if (vm)
        virDomainObjUnlock(vm);
4382
    if (event)
4383
        qemuDomainEventQueue(driver, event);
4384
    qemuDriverUnlock(driver);
4385
    return ret;
D
Daniel P. Berrange 已提交
4386 4387
}

4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402
static int
qemudCanonicalizeMachineFromInfo(virDomainDefPtr def,
                                 virCapsGuestDomainInfoPtr info,
                                 char **canonical)
{
    int i;

    *canonical = NULL;

    for (i = 0; i < info->nmachines; i++) {
        virCapsGuestMachinePtr machine = info->machines[i];

        if (!machine->canonical)
            continue;

4403
        if (STRNEQ(def->os.machine, machine->name))
4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416
            continue;

        if (!(*canonical = strdup(machine->canonical))) {
            virReportOOMError(NULL);
            return -1;
        }

        break;
    }

    return 0;
}

4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431
static int
qemudCanonicalizeMachineDirect(virDomainDefPtr def, char **canonical)
{
    virCapsGuestMachinePtr *machines = NULL;
    int i, nmachines = 0;

    if (qemudProbeMachineTypes(def->emulator, &machines, &nmachines) < 0) {
        virReportOOMError(NULL);
        return -1;
    }

    for (i = 0; i < nmachines; i++) {
        if (!machines[i]->canonical)
            continue;

4432
        if (STRNEQ(def->os.machine, machines[i]->name))
4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444
            continue;

        *canonical = machines[i]->canonical;
        machines[i]->canonical = NULL;
        break;
    }

    virCapabilitiesFreeMachines(machines, nmachines);

    return 0;
}

4445 4446
int
qemudCanonicalizeMachine(struct qemud_driver *driver, virDomainDefPtr def)
4447 4448 4449 4450 4451 4452
{
    char *canonical = NULL;
    int i;

    for (i = 0; i < driver->caps->nguests; i++) {
        virCapsGuestPtr guest = driver->caps->guests[i];
4453
        virCapsGuestDomainInfoPtr info;
4454 4455 4456
        int j;

        for (j = 0; j < guest->arch.ndomains; j++) {
4457
            info = &guest->arch.domains[j]->info;
4458

4459 4460 4461 4462 4463 4464 4465 4466 4467
            if (!info->emulator || !STREQ(info->emulator, def->emulator))
                continue;

            if (!info->nmachines)
                info = &guest->arch.defaultInfo;

            if (qemudCanonicalizeMachineFromInfo(def, info, &canonical) < 0)
                return -1;
            goto out;
4468 4469
        }

4470 4471 4472 4473
        info = &guest->arch.defaultInfo;

        if (info->emulator && STREQ(info->emulator, def->emulator)) {
            if (qemudCanonicalizeMachineFromInfo(def, info, &canonical) < 0)
4474 4475 4476 4477
                return -1;
            goto out;
        }
    }
4478 4479 4480 4481

    if (qemudCanonicalizeMachineDirect(def, &canonical) < 0)
        return -1;

4482 4483 4484 4485 4486 4487 4488
out:
    if (canonical) {
        VIR_FREE(def->os.machine);
        def->os.machine = canonical;
    }
    return 0;
}
D
Daniel P. Berrange 已提交
4489

4490
static virDomainPtr qemudDomainDefine(virConnectPtr conn, const char *xml) {
4491
    struct qemud_driver *driver = conn->privateData;
4492
    virDomainDefPtr def;
4493
    virDomainObjPtr vm = NULL;
4494
    virDomainPtr dom = NULL;
4495
    virDomainEventPtr event = NULL;
4496
    int dupVM;
4497

4498
    qemuDriverLock(driver);
4499 4500
    if (!(def = virDomainDefParseString(conn, driver->caps, xml,
                                        VIR_DOMAIN_XML_INACTIVE)))
4501
        goto cleanup;
4502

4503 4504 4505
    if (virSecurityDriverVerify(conn, def) < 0)
        goto cleanup;

4506 4507
    if ((dupVM = virDomainObjIsDuplicate(&driver->domains, def, 0)) < 0)
        goto cleanup;
4508

4509
    if (qemudCanonicalizeMachine(driver, def) < 0)
4510 4511
        goto cleanup;

4512
    if (!(vm = virDomainAssignDef(conn,
4513
                                  driver->caps,
4514 4515
                                  &driver->domains,
                                  def))) {
4516
        goto cleanup;
4517
    }
4518
    def = NULL;
4519
    vm->persistent = 1;
4520

4521 4522
    if (virDomainSaveConfig(conn,
                            driver->configDir,
4523
                            vm->newDef ? vm->newDef : vm->def) < 0) {
4524 4525
        virDomainRemoveInactive(&driver->domains,
                                vm);
4526
        vm = NULL;
4527
        goto cleanup;
4528 4529
    }

4530 4531
    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_DEFINED,
4532
                                     !dupVM ?
4533 4534
                                     VIR_DOMAIN_EVENT_DEFINED_ADDED :
                                     VIR_DOMAIN_EVENT_DEFINED_UPDATED);
4535

4536
    dom = virGetDomain(conn, vm->def->name, vm->def->uuid);
4537
    if (dom) dom->id = vm->def->id;
4538 4539

cleanup:
4540
    virDomainDefFree(def);
4541 4542
    if (vm)
        virDomainObjUnlock(vm);
4543 4544
    if (event)
        qemuDomainEventQueue(driver, event);
4545
    qemuDriverUnlock(driver);
4546
    return dom;
D
Daniel P. Berrange 已提交
4547 4548
}

4549
static int qemudDomainUndefine(virDomainPtr dom) {
4550 4551
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
4552
    virDomainEventPtr event = NULL;
4553
    int ret = -1;
D
Daniel P. Berrange 已提交
4554

4555
    qemuDriverLock(driver);
4556
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
4557

D
Daniel P. Berrange 已提交
4558
    if (!vm) {
4559 4560 4561 4562
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
4563
        goto cleanup;
D
Daniel P. Berrange 已提交
4564 4565
    }

D
Daniel P. Berrange 已提交
4566
    if (virDomainObjIsActive(vm)) {
4567
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
4568
                         "%s", _("cannot delete active domain"));
4569
        goto cleanup;
D
Daniel P. Berrange 已提交
4570 4571
    }

4572 4573 4574
    if (!vm->persistent) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("cannot undefine transient domain"));
4575
        goto cleanup;
4576 4577 4578
    }

    if (virDomainDeleteConfig(dom->conn, driver->configDir, driver->autostartDir, vm) < 0)
4579
        goto cleanup;
D
Daniel P. Berrange 已提交
4580

4581 4582 4583
    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_UNDEFINED,
                                     VIR_DOMAIN_EVENT_UNDEFINED_REMOVED);
4584

4585 4586
    virDomainRemoveInactive(&driver->domains,
                            vm);
4587
    vm = NULL;
4588
    ret = 0;
D
Daniel P. Berrange 已提交
4589

4590
cleanup:
4591 4592
    if (vm)
        virDomainObjUnlock(vm);
4593 4594
    if (event)
        qemuDomainEventQueue(driver, event);
4595
    qemuDriverUnlock(driver);
4596
    return ret;
D
Daniel P. Berrange 已提交
4597 4598
}

4599
/* Return the disks name for use in monitor commands */
4600
static char *qemudDiskDeviceName(const virConnectPtr conn,
4601
                                 const virDomainDiskDefPtr disk) {
4602 4603 4604 4605 4606 4607

    int busid, devid;
    int ret;
    char *devname;

    if (virDiskNameToBusDeviceIndex(disk, &busid, &devid) < 0) {
4608
        qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
4609 4610 4611 4612 4613 4614 4615
                         _("cannot convert disk '%s' to bus/device index"),
                         disk->dst);
        return NULL;
    }

    switch (disk->bus) {
        case VIR_DOMAIN_DISK_BUS_IDE:
4616
            if (disk->device== VIR_DOMAIN_DISK_DEVICE_DISK)
4617
                ret = virAsprintf(&devname, "ide%d-hd%d", busid, devid);
4618
            else
4619
                ret = virAsprintf(&devname, "ide%d-cd%d", busid, devid);
4620 4621
            break;
        case VIR_DOMAIN_DISK_BUS_SCSI:
4622
            if (disk->device == VIR_DOMAIN_DISK_DEVICE_DISK)
4623
                ret = virAsprintf(&devname, "scsi%d-hd%d", busid, devid);
4624
            else
4625
                ret = virAsprintf(&devname, "scsi%d-cd%d", busid, devid);
4626 4627
            break;
        case VIR_DOMAIN_DISK_BUS_FDC:
4628
            ret = virAsprintf(&devname, "floppy%d", devid);
4629 4630
            break;
        case VIR_DOMAIN_DISK_BUS_VIRTIO:
4631
            ret = virAsprintf(&devname, "virtio%d", devid);
4632 4633
            break;
        default:
4634
            qemudReportError(conn, NULL, NULL, VIR_ERR_NO_SUPPORT,
4635 4636 4637 4638 4639 4640
                             _("Unsupported disk name mapping for bus '%s'"),
                             virDomainDiskBusTypeToString(disk->bus));
            return NULL;
    }

    if (ret == -1) {
4641
        virReportOOMError(conn);
4642 4643 4644 4645 4646 4647
        return NULL;
    }

    return devname;
}

4648
static int qemudDomainChangeEjectableMedia(virConnectPtr conn,
4649
                                           struct qemud_driver *driver,
4650
                                           virDomainObjPtr vm,
4651 4652
                                           virDomainDeviceDefPtr dev,
                                           unsigned int qemuCmdFlags)
4653
{
4654
    virDomainDiskDefPtr origdisk = NULL, newdisk;
4655
    char *devname = NULL;
4656
    int i;
4657
    int ret;
4658

4659
    origdisk = NULL;
4660
    newdisk = dev->data.disk;
4661 4662 4663 4664
    for (i = 0 ; i < vm->def->ndisks ; i++) {
        if (vm->def->disks[i]->bus == newdisk->bus &&
            STREQ(vm->def->disks[i]->dst, newdisk->dst)) {
            origdisk = vm->def->disks[i];
4665
            break;
4666
        }
4667 4668 4669
    }

    if (!origdisk) {
4670
        qemudReportError(conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
4671 4672 4673 4674 4675 4676 4677
                         _("No device with bus '%s' and target '%s'"),
                         virDomainDiskBusTypeToString(newdisk->bus),
                         newdisk->dst);
        return -1;
    }

    if (qemuCmdFlags & QEMUD_CMD_FLAG_DRIVE) {
4678
        if (!(devname = qemudDiskDeviceName(conn, newdisk)))
4679 4680 4681 4682 4683 4684 4685 4686 4687
            return -1;
    } else {
        /* Back compat for no -drive option */
        if (newdisk->device == VIR_DOMAIN_DISK_DEVICE_FLOPPY)
            devname = strdup(newdisk->dst);
        else if (newdisk->device == VIR_DOMAIN_DISK_DEVICE_CDROM &&
                 STREQ(newdisk->dst, "hdc"))
            devname = strdup("cdrom");
        else {
4688
            qemudReportError(conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
4689 4690 4691 4692 4693 4694 4695 4696
                             _("Emulator version does not support removable "
                               "media for device '%s' and target '%s'"),
                               virDomainDiskDeviceTypeToString(newdisk->device),
                               newdisk->dst);
            return -1;
        }

        if (!devname) {
4697
            virReportOOMError(conn);
4698 4699 4700
            return -1;
        }
    }
4701

4702
    qemuDomainObjPrivatePtr priv = vm->privateData;
4703
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
4704
    if (newdisk->src) {
4705
        ret = qemuMonitorChangeMedia(priv->mon, devname, newdisk->src);
4706
    } else {
4707
        ret = qemuMonitorEjectMedia(priv->mon, devname);
4708
    }
4709
    qemuDomainObjExitMonitorWithDriver(driver, vm);
4710

4711 4712 4713 4714 4715
    if (ret == 0) {
        VIR_FREE(origdisk->src);
        origdisk->src = newdisk->src;
        newdisk->src = NULL;
        origdisk->type = newdisk->type;
4716
    }
4717
    VIR_FREE(devname);
4718

4719
    return ret;
4720 4721
}

4722

4723
static int qemudDomainAttachPciDiskDevice(virConnectPtr conn,
4724
                                          struct qemud_driver *driver,
4725 4726
                                          virDomainObjPtr vm,
                                          virDomainDeviceDefPtr dev)
4727
{
4728
    int i, ret;
4729
    const char* type = virDomainDiskBusTypeToString(dev->data.disk->bus);
4730
    qemuDomainObjPrivatePtr priv = vm->privateData;
4731 4732 4733

    for (i = 0 ; i < vm->def->ndisks ; i++) {
        if (STREQ(vm->def->disks[i]->dst, dev->data.disk->dst)) {
4734
            qemudReportError(conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
4735 4736 4737 4738 4739 4740
                           _("target %s already exists"), dev->data.disk->dst);
            return -1;
        }
    }

    if (VIR_REALLOC_N(vm->def->disks, vm->def->ndisks+1) < 0) {
4741
        virReportOOMError(conn);
4742 4743 4744
        return -1;
    }

4745
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
4746 4747 4748 4749 4750 4751
    ret = qemuMonitorAddPCIDisk(priv->mon,
                                dev->data.disk->src,
                                type,
                                &dev->data.disk->pci_addr.domain,
                                &dev->data.disk->pci_addr.bus,
                                &dev->data.disk->pci_addr.slot);
4752
    qemuDomainObjExitMonitorWithDriver(driver, vm);
4753

4754 4755
    if (ret == 0)
        virDomainDiskInsertPreAlloced(vm->def, dev->data.disk);
4756

4757
    return ret;
4758
}
4759

4760
static int qemudDomainAttachUsbMassstorageDevice(virConnectPtr conn,
4761
                                                 struct qemud_driver *driver,
4762 4763
                                                 virDomainObjPtr vm,
                                                 virDomainDeviceDefPtr dev)
4764
{
4765
    qemuDomainObjPrivatePtr priv = vm->privateData;
4766
    int i, ret;
4767

4768 4769
    for (i = 0 ; i < vm->def->ndisks ; i++) {
        if (STREQ(vm->def->disks[i]->dst, dev->data.disk->dst)) {
4770
            qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
4771 4772 4773 4774 4775
                           _("target %s already exists"), dev->data.disk->dst);
            return -1;
        }
    }

4776 4777 4778
    if (!dev->data.disk->src) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("disk source path is missing"));
4779 4780 4781
        return -1;
    }

4782
    if (VIR_REALLOC_N(vm->def->disks, vm->def->ndisks+1) < 0) {
4783
        virReportOOMError(conn);
4784 4785 4786
        return -1;
    }

4787
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
4788
    ret = qemuMonitorAddUSBDisk(priv->mon, dev->data.disk->src);
4789
    qemuDomainObjExitMonitorWithDriver(driver, vm);
4790

4791 4792
    if (ret == 0)
        virDomainDiskInsertPreAlloced(vm->def, dev->data.disk);
4793

4794
    return ret;
4795 4796
}

M
Mark McLoughlin 已提交
4797
static int qemudDomainAttachNetDevice(virConnectPtr conn,
4798
                                      struct qemud_driver *driver,
M
Mark McLoughlin 已提交
4799 4800 4801 4802 4803
                                      virDomainObjPtr vm,
                                      virDomainDeviceDefPtr dev,
                                      unsigned int qemuCmdFlags)
{
    virDomainNetDefPtr net = dev->data.net;
4804
    qemuDomainObjPrivatePtr priv = vm->privateData;
4805
    char *tapfd_name = NULL;
4806
    int i, tapfd = -1;
4807
    char *nicstr = NULL;
4808
    char *netstr = NULL;
4809
    int ret = -1;
M
Mark McLoughlin 已提交
4810 4811 4812 4813 4814 4815 4816 4817 4818

    if (!(qemuCmdFlags & QEMUD_CMD_FLAG_HOST_NET_ADD)) {
        qemudReportError(conn, dom, NULL, VIR_ERR_NO_SUPPORT, "%s",
                         _("installed qemu version does not support host_net_add"));
        return -1;
    }

    if (net->type == VIR_DOMAIN_NET_TYPE_BRIDGE ||
        net->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
4819 4820 4821 4822 4823 4824 4825 4826 4827 4828
        if (vm->monitor_chr->type != VIR_DOMAIN_CHR_TYPE_UNIX) {
            qemudReportError(conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                             _("network device type '%s' cannot be attached: "
                               "qemu is not using a unix socket monitor"),
                             virDomainNetTypeToString(net->type));
            return -1;
        }

        if ((tapfd = qemudNetworkIfaceConnect(conn, driver, net, qemuCmdFlags)) < 0)
            return -1;
M
Mark McLoughlin 已提交
4829 4830
    }

4831 4832
    if (VIR_REALLOC_N(vm->def->nets, vm->def->nnets+1) < 0)
        goto no_memory;
M
Mark McLoughlin 已提交
4833 4834

    if ((qemuCmdFlags & QEMUD_CMD_FLAG_NET_NAME) &&
4835 4836
        qemuAssignNetNames(vm->def, net) < 0)
        goto no_memory;
M
Mark McLoughlin 已提交
4837 4838 4839 4840 4841 4842 4843 4844 4845

    /* Choose a vlan value greater than all other values since
     * older versions did not store the value in the state file.
     */
    net->vlan = vm->def->nnets;
    for (i = 0; i < vm->def->nnets; i++)
        if (vm->def->nets[i]->vlan >= net->vlan)
            net->vlan = vm->def->nets[i]->vlan;

4846 4847 4848 4849
    if (tapfd != -1) {
        if (virAsprintf(&tapfd_name, "fd-%s", net->hostnet_name) < 0)
            goto no_memory;

4850
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
4851
        if (qemuMonitorSendFileHandle(priv->mon, tapfd_name, tapfd) < 0) {
4852
            qemuDomainObjExitMonitorWithDriver(driver, vm);
4853
            goto cleanup;
4854
        }
4855
        qemuDomainObjExitMonitorWithDriver(driver, vm);
4856 4857
    }

4858 4859
    if (qemuBuildHostNetStr(conn, net, ' ',
                            net->vlan, tapfd_name, &netstr) < 0)
4860
        goto try_tapfd_close;
M
Mark McLoughlin 已提交
4861

4862
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
4863
    if (qemuMonitorAddHostNetwork(priv->mon, netstr) < 0) {
4864
        qemuDomainObjExitMonitorWithDriver(driver, vm);
4865
        goto try_tapfd_close;
4866
    }
4867
    qemuDomainObjExitMonitorWithDriver(driver, vm);
M
Mark McLoughlin 已提交
4868

4869 4870 4871
    if (tapfd != -1)
        close(tapfd);
    tapfd = -1;
M
Mark McLoughlin 已提交
4872

4873
    if (qemuBuildNicStr(conn, net, NULL, net->vlan, &nicstr) < 0)
4874
        goto try_remove;
M
Mark McLoughlin 已提交
4875

4876
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
4877
    if (qemuMonitorAddPCINetwork(priv->mon, nicstr,
4878 4879
                                 &net->pci_addr.domain,
                                 &net->pci_addr.bus,
4880
                                 &net->pci_addr.slot) < 0) {
4881
        qemuDomainObjExitMonitorWithDriver(driver, vm);
4882
        goto try_remove;
4883
    }
4884
    qemuDomainObjExitMonitorWithDriver(driver, vm);
4885

4886
    ret = 0;
M
Mark McLoughlin 已提交
4887 4888 4889

    vm->def->nets[vm->def->nnets++] = net;

4890 4891 4892 4893 4894 4895
cleanup:
    VIR_FREE(nicstr);
    VIR_FREE(netstr);
    VIR_FREE(tapfd_name);
    if (tapfd != -1)
        close(tapfd);
4896

4897
    return ret;
4898

4899 4900
try_remove:
    if (!net->hostnet_name || net->vlan == 0)
4901
        VIR_WARN0(_("Unable to remove network backend\n"));
4902
    else {
4903
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
4904 4905 4906
        if (qemuMonitorRemoveHostNetwork(priv->mon, net->vlan, net->hostnet_name) < 0)
            VIR_WARN(_("Failed to remove network backend for vlan %d, net %s"),
                     net->vlan, net->hostnet_name);
4907
        qemuDomainObjExitMonitorWithDriver(driver, vm);
4908
    }
4909
    goto cleanup;
4910

4911
try_tapfd_close:
4912
    if (tapfd_name) {
4913
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
4914 4915
        if (qemuMonitorCloseFileHandle(priv->mon, tapfd_name) < 0)
            VIR_WARN(_("Failed to close tapfd with '%s'\n"), tapfd_name);
4916
        qemuDomainObjExitMonitorWithDriver(driver, vm);
4917
    }
4918

4919 4920
    goto cleanup;

4921 4922
no_memory:
    virReportOOMError(conn);
4923
    goto cleanup;
M
Mark McLoughlin 已提交
4924 4925
}

4926
static int qemudDomainAttachHostPciDevice(virConnectPtr conn,
4927
                                          struct qemud_driver *driver,
4928 4929 4930
                                          virDomainObjPtr vm,
                                          virDomainDeviceDefPtr dev)
{
4931
    qemuDomainObjPrivatePtr priv = vm->privateData;
4932
    virDomainHostdevDefPtr hostdev = dev->data.hostdev;
4933
    pciDevice *pci;
4934
    int ret;
4935 4936 4937 4938 4939 4940

    if (VIR_REALLOC_N(vm->def->hostdevs, vm->def->nhostdevs+1) < 0) {
        virReportOOMError(conn);
        return -1;
    }

4941 4942 4943 4944 4945 4946 4947
    pci = pciGetDevice(conn,
                       hostdev->source.subsys.u.pci.domain,
                       hostdev->source.subsys.u.pci.bus,
                       hostdev->source.subsys.u.pci.slot,
                       hostdev->source.subsys.u.pci.function);
    if (!dev)
        return -1;
4948

4949
    if ((hostdev->managed && pciDettachDevice(conn, pci) < 0) ||
4950 4951 4952 4953 4954 4955
        pciResetDevice(conn, pci, driver->activePciHostdevs) < 0) {
        pciFreeDevice(conn, pci);
        return -1;
    }

    if (pciDeviceListAdd(conn, driver->activePciHostdevs, pci) < 0) {
4956
        pciFreeDevice(conn, pci);
4957
        return -1;
4958 4959
    }

4960
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
4961 4962 4963 4964 4965 4966 4967 4968
    ret = qemuMonitorAddPCIHostDevice(priv->mon,
                                      hostdev->source.subsys.u.pci.domain,
                                      hostdev->source.subsys.u.pci.bus,
                                      hostdev->source.subsys.u.pci.slot,
                                      hostdev->source.subsys.u.pci.function,
                                      &hostdev->source.subsys.u.pci.guest_addr.domain,
                                      &hostdev->source.subsys.u.pci.guest_addr.bus,
                                      &hostdev->source.subsys.u.pci.guest_addr.slot);
4969
    qemuDomainObjExitMonitorWithDriver(driver, vm);
4970
    if (ret < 0)
4971
        goto error;
4972 4973 4974 4975

    vm->def->hostdevs[vm->def->nhostdevs++] = hostdev;

    return 0;
4976 4977 4978 4979 4980

error:
    pciDeviceListDel(conn, driver->activePciHostdevs, pci);

    return -1;
4981 4982
}

M
Mark McLoughlin 已提交
4983
static int qemudDomainAttachHostUsbDevice(virConnectPtr conn,
4984
                                          struct qemud_driver *driver,
M
Mark McLoughlin 已提交
4985 4986
                                          virDomainObjPtr vm,
                                          virDomainDeviceDefPtr dev)
4987 4988
{
    int ret;
4989
    qemuDomainObjPrivatePtr priv = vm->privateData;
4990

4991
    if (VIR_REALLOC_N(vm->def->hostdevs, vm->def->nhostdevs+1) < 0) {
4992
        virReportOOMError(conn);
4993 4994
        return -1;
    }
4995

4996
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
4997
    if (dev->data.hostdev->source.subsys.u.usb.vendor) {
4998
        ret = qemuMonitorAddUSBDeviceMatch(priv->mon,
4999 5000
                                           dev->data.hostdev->source.subsys.u.usb.vendor,
                                           dev->data.hostdev->source.subsys.u.usb.product);
5001
    } else {
5002
        ret = qemuMonitorAddUSBDeviceExact(priv->mon,
5003 5004
                                           dev->data.hostdev->source.subsys.u.usb.bus,
                                           dev->data.hostdev->source.subsys.u.usb.device);
5005
    }
5006
    qemuDomainObjExitMonitorWithDriver(driver, vm);
5007

5008 5009
    if (ret != -1)
        vm->def->hostdevs[vm->def->nhostdevs++] = dev->data.hostdev;
5010

5011
    return ret;
5012 5013
}

M
Mark McLoughlin 已提交
5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029
static int qemudDomainAttachHostDevice(virConnectPtr conn,
                                       struct qemud_driver *driver,
                                       virDomainObjPtr vm,
                                       virDomainDeviceDefPtr dev)
{
    virDomainHostdevDefPtr hostdev = dev->data.hostdev;

    if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) {
        qemudReportError(conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                         _("hostdev mode '%s' not supported"),
                         virDomainHostdevModeTypeToString(hostdev->mode));
        return -1;
    }

    if (qemuDomainSetDeviceOwnership(conn, driver, dev, 0) < 0)
        return -1;
5030 5031 5032
    if (driver->securityDriver &&
        driver->securityDriver->domainSetSecurityHostdevLabel(conn, vm, dev->data.hostdev) < 0)
        return -1;
M
Mark McLoughlin 已提交
5033 5034

    switch (hostdev->source.subsys.type) {
5035
    case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI:
5036
        return qemudDomainAttachHostPciDevice(conn, driver, vm, dev);
M
Mark McLoughlin 已提交
5037
    case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_USB:
5038
        return qemudDomainAttachHostUsbDevice(conn, driver, vm, dev);
M
Mark McLoughlin 已提交
5039 5040 5041 5042 5043 5044 5045 5046
    default:
        qemudReportError(conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                         _("hostdev subsys type '%s' not supported"),
                         virDomainHostdevSubsysTypeToString(hostdev->source.subsys.type));
        return -1;
    }
}

5047 5048
static int qemudDomainAttachDevice(virDomainPtr dom,
                                   const char *xml) {
5049 5050 5051
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    virDomainDeviceDefPtr dev = NULL;
5052
    unsigned int qemuCmdFlags;
5053
    virCgroupPtr cgroup = NULL;
5054
    int ret = -1;
5055

5056
    qemuDriverLock(driver);
5057
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
5058
    if (!vm) {
5059 5060 5061 5062
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
5063
        goto cleanup;
5064 5065
    }

5066 5067 5068
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
5069
    if (!virDomainObjIsActive(vm)) {
5070
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
5071
                         "%s", _("cannot attach device on inactive domain"));
5072
        goto endjob;
5073 5074
    }

5075 5076
    dev = virDomainDeviceDefParse(dom->conn, driver->caps, vm->def, xml,
                                  VIR_DOMAIN_XML_INACTIVE);
5077
    if (dev == NULL)
5078
        goto endjob;
5079

5080 5081 5082
    if (qemudExtractVersionInfo(vm->def->emulator,
                                NULL,
                                &qemuCmdFlags) < 0)
5083
        goto endjob;
5084

5085
    if (dev->type == VIR_DOMAIN_DEVICE_DISK) {
5086
        if (qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_DEVICES)) {
5087 5088 5089 5090
            if (virCgroupForDomain(driver->cgroup, vm->def->name, &cgroup, 0) !=0 ) {
                qemudReportError(dom->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                                 _("Unable to find cgroup for %s\n"),
                                 vm->def->name);
5091
                goto endjob;
5092 5093 5094 5095 5096 5097 5098 5099
            }
            if (dev->data.disk->src != NULL &&
                dev->data.disk->type == VIR_DOMAIN_DISK_TYPE_BLOCK &&
                virCgroupAllowDevicePath(cgroup,
                                         dev->data.disk->src) < 0) {
                qemudReportError(dom->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                                 _("unable to allow device %s"),
                                 dev->data.disk->src);
5100
                goto endjob;
5101 5102 5103
            }
        }

5104
        switch (dev->data.disk->device) {
5105 5106
        case VIR_DOMAIN_DISK_DEVICE_CDROM:
        case VIR_DOMAIN_DISK_DEVICE_FLOPPY:
5107 5108
            if (driver->securityDriver)
                driver->securityDriver->domainSetSecurityImageLabel(dom->conn, vm, dev->data.disk);
5109 5110

            if (qemuDomainSetDeviceOwnership(dom->conn, driver, dev, 0) < 0)
5111
                goto endjob;
5112

5113
            ret = qemudDomainChangeEjectableMedia(dom->conn, driver, vm, dev, qemuCmdFlags);
5114
            break;
5115

5116
        case VIR_DOMAIN_DISK_DEVICE_DISK:
5117 5118
            if (driver->securityDriver)
                driver->securityDriver->domainSetSecurityImageLabel(dom->conn, vm, dev->data.disk);
5119 5120

            if (qemuDomainSetDeviceOwnership(dom->conn, driver, dev, 0) < 0)
5121
                goto endjob;
5122

5123
            if (dev->data.disk->bus == VIR_DOMAIN_DISK_BUS_USB) {
5124
                ret = qemudDomainAttachUsbMassstorageDevice(dom->conn, driver, vm, dev);
5125 5126
            } else if (dev->data.disk->bus == VIR_DOMAIN_DISK_BUS_SCSI ||
                       dev->data.disk->bus == VIR_DOMAIN_DISK_BUS_VIRTIO) {
5127
                ret = qemudDomainAttachPciDiskDevice(dom->conn, driver, vm, dev);
5128 5129 5130 5131
            } else {
                qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                                 _("disk bus '%s' cannot be hotplugged."),
                                 virDomainDiskBusTypeToString(dev->data.disk->bus));
5132
                /* fallthrough */
5133 5134
            }
            break;
5135

5136 5137
        default:
            qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
5138 5139
                             _("disk device type '%s' cannot be hotplugged"),
                             virDomainDiskDeviceTypeToString(dev->data.disk->device));
5140 5141 5142 5143 5144
            /* Fallthrough */
        }
        if (ret != 0) {
            virCgroupDenyDevicePath(cgroup,
                                    dev->data.disk->src);
5145
        }
M
Mark McLoughlin 已提交
5146
    } else if (dev->type == VIR_DOMAIN_DEVICE_NET) {
5147
        ret = qemudDomainAttachNetDevice(dom->conn, driver, vm, dev, qemuCmdFlags);
M
Mark McLoughlin 已提交
5148 5149
    } else if (dev->type == VIR_DOMAIN_DEVICE_HOSTDEV) {
        ret = qemudDomainAttachHostDevice(dom->conn, driver, vm, dev);
5150
    } else {
5151
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
5152 5153
                         _("device type '%s' cannot be attached"),
                         virDomainDeviceTypeToString(dev->type));
5154
        goto endjob;
5155 5156
    }

5157
    if (!ret && virDomainSaveStatus(dom->conn, driver->stateDir, vm) < 0)
5158 5159
        ret = -1;

5160 5161 5162
endjob:
    qemuDomainObjEndJob(vm);

5163
cleanup:
5164 5165 5166
    if (cgroup)
        virCgroupFree(&cgroup);

5167
    if (ret < 0 && dev != NULL) {
5168 5169
        if (qemuDomainSetDeviceOwnership(dom->conn, driver, dev, 1) < 0)
            VIR_WARN0("Fail to restore disk device ownership");
G
Guido Günther 已提交
5170
        virDomainDeviceDefFree(dev);
5171
    }
5172 5173
    if (vm)
        virDomainObjUnlock(vm);
5174
    qemuDriverUnlock(driver);
5175 5176 5177
    return ret;
}

5178
static int qemudDomainDetachPciDiskDevice(virConnectPtr conn,
5179 5180 5181
                                          struct qemud_driver *driver,
                                          virDomainObjPtr vm,
                                          virDomainDeviceDefPtr dev)
5182 5183 5184
{
    int i, ret = -1;
    virDomainDiskDefPtr detach = NULL;
5185
    qemuDomainObjPrivatePtr priv = vm->privateData;
5186 5187 5188 5189 5190 5191 5192 5193 5194

    for (i = 0 ; i < vm->def->ndisks ; i++) {
        if (STREQ(vm->def->disks[i]->dst, dev->data.disk->dst)) {
            detach = vm->def->disks[i];
            break;
        }
    }

    if (!detach) {
5195
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
5196
                         _("disk %s not found"), dev->data.disk->dst);
5197
        goto cleanup;
5198 5199
    }

5200
    if (!virDiskHasValidPciAddr(detach)) {
5201
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
5202 5203
                         _("disk %s cannot be detached - no PCI address for device"),
                           detach->dst);
5204
        goto cleanup;
5205 5206
    }

5207
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
5208
    if (qemuMonitorRemovePCIDevice(priv->mon,
5209 5210
                                   detach->pci_addr.domain,
                                   detach->pci_addr.bus,
5211 5212
                                   detach->pci_addr.slot) < 0) {
        qemuDomainObjExitMonitor(vm);
5213
        goto cleanup;
5214
    }
5215
    qemuDomainObjExitMonitorWithDriver(driver, vm);
5216

5217 5218 5219 5220 5221 5222 5223 5224 5225 5226
    if (vm->def->ndisks > 1) {
        memmove(vm->def->disks + i,
                vm->def->disks + i + 1,
                sizeof(*vm->def->disks) *
                (vm->def->ndisks - (i + 1)));
        vm->def->ndisks--;
        if (VIR_REALLOC_N(vm->def->disks, vm->def->ndisks) < 0) {
            /* ignore, harmless */
        }
    } else {
5227
        VIR_FREE(vm->def->disks);
5228
        vm->def->ndisks = 0;
5229
    }
5230
    virDomainDiskDefFree(detach);
5231

5232
    ret = 0;
5233 5234

cleanup:
5235 5236 5237
    return ret;
}

5238 5239
static int
qemudDomainDetachNetDevice(virConnectPtr conn,
5240
                           struct qemud_driver *driver,
5241 5242 5243 5244 5245
                           virDomainObjPtr vm,
                           virDomainDeviceDefPtr dev)
{
    int i, ret = -1;
    virDomainNetDefPtr detach = NULL;
5246
    qemuDomainObjPrivatePtr priv = vm->privateData;
5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271

    for (i = 0 ; i < vm->def->nnets ; i++) {
        virDomainNetDefPtr net = vm->def->nets[i];

        if (!memcmp(net->mac, dev->data.net->mac,  sizeof(net->mac))) {
            detach = net;
            break;
        }
    }

    if (!detach) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                         _("network device %02x:%02x:%02x:%02x:%02x:%02x not found"),
                         dev->data.net->mac[0], dev->data.net->mac[1],
                         dev->data.net->mac[2], dev->data.net->mac[3],
                         dev->data.net->mac[4], dev->data.net->mac[5]);
        goto cleanup;
    }

    if (!virNetHasValidPciAddr(detach) || detach->vlan < 0 || !detach->hostnet_name) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                         "%s", _("network device cannot be detached - device state missing"));
        goto cleanup;
    }

5272
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
5273
    if (qemuMonitorRemovePCIDevice(priv->mon,
5274 5275
                                   detach->pci_addr.domain,
                                   detach->pci_addr.bus,
5276
                                   detach->pci_addr.slot) < 0) {
5277
        qemuDomainObjExitMonitorWithDriver(driver, vm);
5278
        goto cleanup;
5279
    }
5280

5281
    if (qemuMonitorRemoveHostNetwork(priv->mon, detach->vlan, detach->hostnet_name) < 0) {
5282
        qemuDomainObjExitMonitorWithDriver(driver, vm);
5283
        goto cleanup;
5284
    }
5285
    qemuDomainObjExitMonitorWithDriver(driver, vm);
5286

5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297
    if ((driver->macFilter) && (detach->ifname != NULL)) {
        if ((errno = networkDisallowMacOnPort(conn,
                                              driver,
                                              detach->ifname,
                                              detach->mac))) {
            virReportSystemError(conn, errno,
             _("failed to remove ebtables rule on  '%s'"),
                                 detach->ifname);
        }
    }

5298 5299 5300 5301 5302 5303 5304 5305 5306 5307
    if (vm->def->nnets > 1) {
        memmove(vm->def->nets + i,
                vm->def->nets + i + 1,
                sizeof(*vm->def->nets) *
                (vm->def->nnets - (i + 1)));
        vm->def->nnets--;
        if (VIR_REALLOC_N(vm->def->nets, vm->def->nnets) < 0) {
            /* ignore, harmless */
        }
    } else {
5308
        VIR_FREE(vm->def->nets);
5309
        vm->def->nnets = 0;
5310
    }
5311
    virDomainNetDefFree(detach);
5312

5313 5314 5315 5316 5317 5318
    ret = 0;

cleanup:
    return ret;
}

5319
static int qemudDomainDetachHostPciDevice(virConnectPtr conn,
5320
                                          struct qemud_driver *driver,
5321 5322 5323
                                          virDomainObjPtr vm,
                                          virDomainDeviceDefPtr dev)
{
5324
    virDomainHostdevDefPtr detach = NULL;
5325
    qemuDomainObjPrivatePtr priv = vm->privateData;
5326
    int i, ret;
5327
    pciDevice *pci;
5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359

    for (i = 0 ; i < vm->def->nhostdevs ; i++) {
        unsigned domain   = vm->def->hostdevs[i]->source.subsys.u.pci.domain;
        unsigned bus      = vm->def->hostdevs[i]->source.subsys.u.pci.bus;
        unsigned slot     = vm->def->hostdevs[i]->source.subsys.u.pci.slot;
        unsigned function = vm->def->hostdevs[i]->source.subsys.u.pci.function;

        if (dev->data.hostdev->source.subsys.u.pci.domain   == domain &&
            dev->data.hostdev->source.subsys.u.pci.bus      == bus &&
            dev->data.hostdev->source.subsys.u.pci.slot     == slot &&
            dev->data.hostdev->source.subsys.u.pci.function == function) {
            detach = vm->def->hostdevs[i];
            break;
        }
    }

    if (!detach) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                         _("host pci device %.4x:%.2x:%.2x.%.1x not found"),
                         dev->data.hostdev->source.subsys.u.pci.domain,
                         dev->data.hostdev->source.subsys.u.pci.bus,
                         dev->data.hostdev->source.subsys.u.pci.slot,
                         dev->data.hostdev->source.subsys.u.pci.function);
        return -1;
    }

    if (!virHostdevHasValidGuestAddr(detach)) {
        qemudReportError(conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                         "%s", _("hostdev cannot be detached - device state missing"));
        return -1;
    }

5360
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
5361
    if (qemuMonitorRemovePCIDevice(priv->mon,
5362 5363
                                   detach->source.subsys.u.pci.guest_addr.domain,
                                   detach->source.subsys.u.pci.guest_addr.bus,
5364
                                   detach->source.subsys.u.pci.guest_addr.slot) < 0) {
5365
        qemuDomainObjExitMonitorWithDriver(driver, vm);
5366
        return -1;
5367
    }
5368
    qemuDomainObjExitMonitorWithDriver(driver, vm);
5369 5370 5371

    ret = 0;

5372 5373 5374 5375 5376 5377 5378 5379
    pci = pciGetDevice(conn,
                       detach->source.subsys.u.pci.domain,
                       detach->source.subsys.u.pci.bus,
                       detach->source.subsys.u.pci.slot,
                       detach->source.subsys.u.pci.function);
    if (!pci)
        ret = -1;
    else {
5380 5381
        pciDeviceListDel(conn, driver->activePciHostdevs, pci);
        if (pciResetDevice(conn, pci, driver->activePciHostdevs) < 0)
5382 5383
            ret = -1;
        if (detach->managed && pciReAttachDevice(conn, pci) < 0)
5384
            ret = -1;
5385
        pciFreeDevice(conn, pci);
5386 5387
    }

5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399
    if (vm->def->nhostdevs > 1) {
        memmove(vm->def->hostdevs + i,
                vm->def->hostdevs + i + 1,
                sizeof(*vm->def->hostdevs) *
                (vm->def->nhostdevs - (i + 1)));
        vm->def->nhostdevs--;
        if (VIR_REALLOC_N(vm->def->hostdevs, vm->def->nhostdevs) < 0) {
            /* ignore, harmless */
        }
    } else {
        VIR_FREE(vm->def->hostdevs);
        vm->def->nhostdevs = 0;
5400
    }
5401
    virDomainHostdevDefFree(detach);
5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422

    return ret;
}

static int qemudDomainDetachHostDevice(virConnectPtr conn,
                                       struct qemud_driver *driver,
                                       virDomainObjPtr vm,
                                       virDomainDeviceDefPtr dev)
{
    virDomainHostdevDefPtr hostdev = dev->data.hostdev;
    int ret;

    if (hostdev->mode != VIR_DOMAIN_HOSTDEV_MODE_SUBSYS) {
        qemudReportError(conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                         _("hostdev mode '%s' not supported"),
                         virDomainHostdevModeTypeToString(hostdev->mode));
        return -1;
    }

    switch (hostdev->source.subsys.type) {
    case VIR_DOMAIN_HOSTDEV_SUBSYS_TYPE_PCI:
5423
        ret = qemudDomainDetachHostPciDevice(conn, driver, vm, dev);
5424
        break;
5425 5426 5427 5428 5429 5430 5431
    default:
        qemudReportError(conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                         _("hostdev subsys type '%s' not supported"),
                         virDomainHostdevSubsysTypeToString(hostdev->source.subsys.type));
        return -1;
    }

5432 5433 5434 5435
    if (driver->securityDriver &&
        driver->securityDriver->domainSetSecurityHostdevLabel(conn, vm, dev->data.hostdev) < 0)
        VIR_WARN0("Failed to restore device labelling");

5436
    if (qemuDomainSetDeviceOwnership(conn, driver, dev, 1) < 0)
5437
        VIR_WARN0("Failed to restore device ownership");
5438 5439 5440 5441

    return ret;
}

5442 5443
static int qemudDomainDetachDevice(virDomainPtr dom,
                                   const char *xml) {
5444 5445 5446 5447
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    virDomainDeviceDefPtr dev = NULL;
    int ret = -1;
5448

5449
    qemuDriverLock(driver);
5450
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
5451
    if (!vm) {
5452 5453 5454 5455
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
5456
        goto cleanup;
5457 5458
    }

5459 5460 5461
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
5462
    if (!virDomainObjIsActive(vm)) {
5463
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
G
Guido Günther 已提交
5464
                         "%s", _("cannot detach device on inactive domain"));
5465
        goto endjob;
5466 5467
    }

5468 5469
    dev = virDomainDeviceDefParse(dom->conn, driver->caps, vm->def, xml,
                                  VIR_DOMAIN_XML_INACTIVE);
5470
    if (dev == NULL)
5471
        goto endjob;
5472

5473 5474 5475 5476

    if (dev->type == VIR_DOMAIN_DEVICE_DISK &&
        dev->data.disk->device == VIR_DOMAIN_DISK_DEVICE_DISK &&
        (dev->data.disk->bus == VIR_DOMAIN_DISK_BUS_SCSI ||
5477
         dev->data.disk->bus == VIR_DOMAIN_DISK_BUS_VIRTIO)) {
5478
        ret = qemudDomainDetachPciDiskDevice(dom->conn, driver, vm, dev);
5479
        if (driver->securityDriver)
5480
            driver->securityDriver->domainRestoreSecurityImageLabel(dom->conn, vm, dev->data.disk);
5481 5482
        if (qemuDomainSetDeviceOwnership(dom->conn, driver, dev, 1) < 0)
            VIR_WARN0("Fail to restore disk device ownership");
5483
    } else if (dev->type == VIR_DOMAIN_DEVICE_NET) {
5484
        ret = qemudDomainDetachNetDevice(dom->conn, driver, vm, dev);
5485 5486
    } else if (dev->type == VIR_DOMAIN_DEVICE_HOSTDEV) {
        ret = qemudDomainDetachHostDevice(dom->conn, driver, vm, dev);
5487
    } else
5488 5489 5490
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                         "%s", _("only SCSI or virtio disk device can be detached dynamically"));

5491
    if (!ret && virDomainSaveStatus(dom->conn, driver->stateDir, vm) < 0)
5492 5493
        ret = -1;

5494 5495 5496
endjob:
    qemuDomainObjEndJob(vm);

5497 5498
cleanup:
    virDomainDeviceDefFree(dev);
5499 5500
    if (vm)
        virDomainObjUnlock(vm);
5501
    qemuDriverUnlock(driver);
5502 5503 5504
    return ret;
}

5505
static int qemudDomainGetAutostart(virDomainPtr dom,
5506
                                   int *autostart) {
5507 5508 5509
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int ret = -1;
5510

5511
    qemuDriverLock(driver);
5512
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
5513 5514
    qemuDriverUnlock(driver);

5515
    if (!vm) {
5516 5517 5518 5519
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
5520
        goto cleanup;
5521 5522 5523
    }

    *autostart = vm->autostart;
5524
    ret = 0;
5525

5526
cleanup:
5527 5528
    if (vm)
        virDomainObjUnlock(vm);
5529
    return ret;
5530 5531
}

5532
static int qemudDomainSetAutostart(virDomainPtr dom,
5533
                                   int autostart) {
5534 5535
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
5536 5537
    char *configFile = NULL, *autostartLink = NULL;
    int ret = -1;
5538

5539
    qemuDriverLock(driver);
5540
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
5541

5542
    if (!vm) {
5543 5544 5545 5546
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
5547
        goto cleanup;
5548 5549
    }

5550 5551 5552
    if (!vm->persistent) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("cannot set autostart for transient domain"));
5553
        goto cleanup;
5554 5555
    }

5556 5557
    autostart = (autostart != 0);

5558 5559 5560 5561 5562
    if (vm->autostart != autostart) {
        if ((configFile = virDomainConfigFile(dom->conn, driver->configDir, vm->def->name)) == NULL)
            goto cleanup;
        if ((autostartLink = virDomainConfigFile(dom->conn, driver->autostartDir, vm->def->name)) == NULL)
            goto cleanup;
5563

5564 5565
        if (autostart) {
            int err;
5566

5567
            if ((err = virFileMakePath(driver->autostartDir))) {
5568 5569 5570
                virReportSystemError(dom->conn, err,
                                     _("cannot create autostart directory %s"),
                                     driver->autostartDir);
5571 5572
                goto cleanup;
            }
5573

5574
            if (symlink(configFile, autostartLink) < 0) {
5575 5576 5577
                virReportSystemError(dom->conn, errno,
                                     _("Failed to create symlink '%s to '%s'"),
                                     autostartLink, configFile);
5578 5579 5580 5581
                goto cleanup;
            }
        } else {
            if (unlink(autostartLink) < 0 && errno != ENOENT && errno != ENOTDIR) {
5582 5583 5584
                virReportSystemError(dom->conn, errno,
                                     _("Failed to delete symlink '%s'"),
                                     autostartLink);
5585 5586
                goto cleanup;
            }
5587 5588
        }

5589
        vm->autostart = autostart;
5590
    }
5591
    ret = 0;
5592

5593 5594 5595
cleanup:
    VIR_FREE(configFile);
    VIR_FREE(autostartLink);
5596 5597
    if (vm)
        virDomainObjUnlock(vm);
5598
    qemuDriverUnlock(driver);
5599
    return ret;
5600 5601
}

5602 5603 5604 5605 5606

static char *qemuGetSchedulerType(virDomainPtr dom,
                                  int *nparams)
{
    struct qemud_driver *driver = dom->conn->privateData;
5607
    char *ret = NULL;
5608

5609
    qemuDriverLock(driver);
5610
    if (!qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_CPU)) {
5611 5612
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                         __FUNCTION__);
5613
        goto cleanup;
5614 5615 5616 5617 5618 5619 5620 5621
    }

    if (nparams)
        *nparams = 1;

    ret = strdup("posix");
    if (!ret)
        virReportOOMError(dom->conn);
5622 5623 5624

cleanup:
    qemuDriverUnlock(driver);
5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637
    return ret;
}

static int qemuSetSchedulerParameters(virDomainPtr dom,
                                      virSchedParameterPtr params,
                                      int nparams)
{
    struct qemud_driver *driver = dom->conn->privateData;
    int i;
    virCgroupPtr group = NULL;
    virDomainObjPtr vm = NULL;
    int ret = -1;

5638
    qemuDriverLock(driver);
5639
    if (!qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_CPU)) {
5640 5641
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                         __FUNCTION__);
5642
        goto cleanup;
5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664
    }

    vm = virDomainFindByUUID(&driver->domains, dom->uuid);

    if (vm == NULL) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("No such domain %s"), dom->uuid);
        goto cleanup;
    }

    if (virCgroupForDomain(driver->cgroup, vm->def->name, &group, 0) != 0) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("cannot find cgroup for domain %s"), vm->def->name);
        goto cleanup;
    }

    for (i = 0; i < nparams; i++) {
        virSchedParameterPtr param = &params[i];

        if (STREQ(param->field, "cpu_shares")) {
            int rc;
            if (param->type != VIR_DOMAIN_SCHED_FIELD_ULLONG) {
C
Chris Lalancette 已提交
5665
                qemudReportError(dom->conn, dom, NULL, VIR_ERR_INVALID_ARG, "%s",
5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687
                                 _("invalid type for cpu_shares tunable, expected a 'ullong'"));
                goto cleanup;
            }

            rc = virCgroupSetCpuShares(group, params[i].value.ul);
            if (rc != 0) {
                virReportSystemError(dom->conn, -rc, "%s",
                                     _("unable to set cpu shares tunable"));
                goto cleanup;
            }
        } else {
            qemudReportError(dom->conn, domain, NULL, VIR_ERR_INVALID_ARG,
                             _("Invalid parameter `%s'"), param->field);
            goto cleanup;
        }
    }
    ret = 0;

cleanup:
    virCgroupFree(&group);
    if (vm)
        virDomainObjUnlock(vm);
5688
    qemuDriverUnlock(driver);
5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702
    return ret;
}

static int qemuGetSchedulerParameters(virDomainPtr dom,
                                      virSchedParameterPtr params,
                                      int *nparams)
{
    struct qemud_driver *driver = dom->conn->privateData;
    virCgroupPtr group = NULL;
    virDomainObjPtr vm = NULL;
    unsigned long long val;
    int ret = -1;
    int rc;

5703
    qemuDriverLock(driver);
5704
    if (!qemuCgroupControllerActive(driver, VIR_CGROUP_CONTROLLER_CPU)) {
5705 5706
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                         __FUNCTION__);
5707
        goto cleanup;
5708 5709 5710 5711 5712
    }

    if ((*nparams) != 1) {
        qemudReportError(dom->conn, domain, NULL, VIR_ERR_INVALID_ARG,
                         "%s", _("Invalid parameter count"));
5713
        goto cleanup;
5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735 5736 5737
    }

    vm = virDomainFindByUUID(&driver->domains, dom->uuid);

    if (vm == NULL) {
        qemudReportError(dom->conn, domain, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("No such domain %s"), dom->uuid);
        goto cleanup;
    }

    if (virCgroupForDomain(driver->cgroup, vm->def->name, &group, 0) != 0) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("cannot find cgroup for domain %s"), vm->def->name);
        goto cleanup;
    }

    rc = virCgroupGetCpuShares(group, &val);
    if (rc != 0) {
        virReportSystemError(dom->conn, -rc, "%s",
                             _("unable to get cpu shares tunable"));
        goto cleanup;
    }
    params[0].value.ul = val;
    params[0].type = VIR_DOMAIN_SCHED_FIELD_ULLONG;
C
Chris Lalancette 已提交
5738 5739 5740 5741 5742
    if (virStrcpyStatic(params[0].field, "cpu_shares") == NULL) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("Field cpu_shares too long for destination"));
        goto cleanup;
    }
5743 5744 5745 5746 5747 5748 5749

    ret = 0;

cleanup:
    virCgroupFree(&group);
    if (vm)
        virDomainObjUnlock(vm);
5750
    qemuDriverUnlock(driver);
5751 5752 5753 5754
    return ret;
}


5755 5756 5757 5758 5759 5760 5761 5762 5763
/* This uses the 'info blockstats' monitor command which was
 * integrated into both qemu & kvm in late 2007.  If the command is
 * not supported we detect this and return the appropriate error.
 */
static int
qemudDomainBlockStats (virDomainPtr dom,
                       const char *path,
                       struct _virDomainBlockStats *stats)
{
5764
    struct qemud_driver *driver = dom->conn->privateData;
5765 5766
    const char *qemu_dev_name = NULL;
    int i, ret = -1;
5767
    virDomainObjPtr vm;
5768
    virDomainDiskDefPtr disk = NULL;
5769

5770
    qemuDriverLock(driver);
5771
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
5772
    qemuDriverUnlock(driver);
5773
    if (!vm) {
5774 5775 5776 5777
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
5778
        goto cleanup;
5779
    }
5780 5781 5782 5783

    if (qemuDomainObjBeginJob(vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
5784
    if (!virDomainObjIsActive (vm)) {
5785
        qemudReportError (dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
5786
                          "%s", _("domain is not running"));
5787
        goto endjob;
5788 5789
    }

5790 5791 5792 5793 5794 5795 5796 5797
    for (i = 0 ; i < vm->def->ndisks ; i++) {
        if (STREQ(path, vm->def->disks[i]->dst)) {
            disk = vm->def->disks[i];
            break;
        }
    }

    if (!disk) {
5798 5799
        qemudReportError (dom->conn, dom, NULL, VIR_ERR_INVALID_ARG,
                          _("invalid path: %s"), path);
5800
        goto endjob;
5801 5802
    }

5803
    qemu_dev_name = qemudDiskDeviceName(dom->conn, disk);
5804
    if (!qemu_dev_name)
5805
        goto endjob;
5806

5807
    qemuDomainObjPrivatePtr priv = vm->privateData;
5808 5809 5810 5811 5812 5813 5814 5815 5816
    qemuDomainObjEnterMonitor(vm);
    ret = qemuMonitorGetBlockStatsInfo(priv->mon,
                                       qemu_dev_name,
                                       &stats->rd_req,
                                       &stats->rd_bytes,
                                       &stats->wr_req,
                                       &stats->wr_bytes,
                                       &stats->errs);
    qemuDomainObjExitMonitor(vm);
5817

5818 5819 5820
endjob:
    qemuDomainObjEndJob(vm);

5821
cleanup:
5822
    VIR_FREE(qemu_dev_name);
5823 5824
    if (vm)
        virDomainObjUnlock(vm);
5825
    return ret;
5826 5827
}

5828
#ifdef __linux__
5829 5830 5831 5832 5833
static int
qemudDomainInterfaceStats (virDomainPtr dom,
                           const char *path,
                           struct _virDomainInterfaceStats *stats)
{
5834 5835
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
5836
    int i;
5837
    int ret = -1;
5838

5839
    qemuDriverLock(driver);
5840
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
5841 5842
    qemuDriverUnlock(driver);

5843
    if (!vm) {
5844 5845 5846 5847
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
5848
        goto cleanup;
5849 5850
    }

D
Daniel P. Berrange 已提交
5851
    if (!virDomainObjIsActive(vm)) {
5852
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
5853
                         "%s", _("domain is not running"));
5854
        goto cleanup;
5855 5856 5857
    }

    /* Check the path is one of the domain's network interfaces. */
5858 5859
    for (i = 0 ; i < vm->def->nnets ; i++) {
        if (vm->def->nets[i]->ifname &&
5860 5861 5862 5863
            STREQ (vm->def->nets[i]->ifname, path)) {
            ret = 0;
            break;
        }
5864 5865
    }

5866 5867 5868 5869 5870
    if (ret == 0)
        ret = linuxDomainInterfaceStats (dom->conn, path, stats);
    else
        qemudReportError (dom->conn, dom, NULL, VIR_ERR_INVALID_ARG,
                          _("invalid path, '%s' is not a known interface"), path);
5871

5872
cleanup:
5873 5874
    if (vm)
        virDomainObjUnlock(vm);
5875 5876
    return ret;
}
5877
#else
5878 5879 5880 5881
static int
qemudDomainInterfaceStats (virDomainPtr dom,
                           const char *path ATTRIBUTE_UNUSED,
                           struct _virDomainInterfaceStats *stats ATTRIBUTE_UNUSED)
5882 5883 5884 5885
    qemudReportError (dom->conn, dom, NULL, VIR_ERR_NO_SUPPORT,
                      "%s", __FUNCTION__);
    return -1;
}
5886
#endif
5887

5888 5889 5890 5891 5892 5893 5894
static int
qemudDomainBlockPeek (virDomainPtr dom,
                      const char *path,
                      unsigned long long offset, size_t size,
                      void *buffer,
                      unsigned int flags ATTRIBUTE_UNUSED)
{
5895 5896 5897
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
    int fd = -1, ret = -1, i;
5898

5899
    qemuDriverLock(driver);
5900
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
5901 5902
    qemuDriverUnlock(driver);

5903
    if (!vm) {
5904 5905 5906 5907
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
5908
        goto cleanup;
5909 5910 5911 5912
    }

    if (!path || path[0] == '\0') {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INVALID_ARG,
J
Jim Meyering 已提交
5913
                         "%s", _("NULL or empty path"));
5914
        goto cleanup;
5915 5916 5917
    }

    /* Check the path belongs to this domain. */
5918 5919
    for (i = 0 ; i < vm->def->ndisks ; i++) {
        if (vm->def->disks[i]->src != NULL &&
5920 5921 5922 5923
            STREQ (vm->def->disks[i]->src, path)) {
            ret = 0;
            break;
        }
5924 5925
    }

5926 5927 5928 5929 5930
    if (ret == 0) {
        ret = -1;
        /* The path is correct, now try to open it and get its size. */
        fd = open (path, O_RDONLY);
        if (fd == -1) {
5931 5932
            virReportSystemError (dom->conn, errno,
                                  _("%s: failed to open"), path);
5933 5934
            goto cleanup;
        }
5935

5936 5937 5938 5939 5940 5941
        /* Seek and read. */
        /* NB. Because we configure with AC_SYS_LARGEFILE, off_t should
         * be 64 bits on all platforms.
         */
        if (lseek (fd, offset, SEEK_SET) == (off_t) -1 ||
            saferead (fd, buffer, size) == (ssize_t) -1) {
5942 5943
            virReportSystemError (dom->conn, errno,
                                  _("%s: failed to seek or read"), path);
5944 5945 5946 5947 5948 5949 5950
            goto cleanup;
        }

        ret = 0;
    } else {
        qemudReportError (dom->conn, dom, NULL, VIR_ERR_INVALID_ARG,
                          "%s", _("invalid path"));
5951 5952
    }

5953 5954 5955
cleanup:
    if (fd >= 0)
        close (fd);
5956 5957
    if (vm)
        virDomainObjUnlock(vm);
5958 5959 5960
    return ret;
}

R
Richard W.M. Jones 已提交
5961 5962 5963 5964 5965 5966
static int
qemudDomainMemoryPeek (virDomainPtr dom,
                       unsigned long long offset, size_t size,
                       void *buffer,
                       unsigned int flags)
{
5967 5968
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
5969
    char *tmp = NULL;
R
Richard W.M. Jones 已提交
5970 5971
    int fd = -1, ret = -1;

5972
    qemuDriverLock(driver);
5973
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
5974
    qemuDriverUnlock(driver);
R
Richard W.M. Jones 已提交
5975 5976

    if (!vm) {
5977 5978 5979 5980
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
5981 5982 5983
        goto cleanup;
    }

5984
    if (flags != VIR_MEMORY_VIRTUAL && flags != VIR_MEMORY_PHYSICAL) {
5985
        qemudReportError (dom->conn, dom, NULL, VIR_ERR_INVALID_ARG,
5986
                     "%s", _("flags parameter must be VIR_MEMORY_VIRTUAL or VIR_MEMORY_PHYSICAL"));
5987
        goto cleanup;
R
Richard W.M. Jones 已提交
5988 5989
    }

5990 5991 5992
    if (qemuDomainObjBeginJob(vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
5993
    if (!virDomainObjIsActive(vm)) {
5994
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
R
Richard W.M. Jones 已提交
5995
                         "%s", _("domain is not running"));
5996
        goto endjob;
R
Richard W.M. Jones 已提交
5997 5998
    }

5999 6000
    if (virAsprintf(&tmp, driver->cacheDir,  "/qemu.mem.XXXXXX") < 0) {
        virReportOOMError(dom->conn);
6001
        goto endjob;
6002 6003
    }

R
Richard W.M. Jones 已提交
6004 6005
    /* Create a temporary filename. */
    if ((fd = mkstemp (tmp)) == -1) {
6006 6007
        virReportSystemError (dom->conn, errno,
                              _("mkstemp(\"%s\") failed"), tmp);
6008
        goto endjob;
R
Richard W.M. Jones 已提交
6009 6010
    }

6011
    qemuDomainObjPrivatePtr priv = vm->privateData;
6012
    qemuDomainObjEnterMonitor(vm);
6013
    if (flags == VIR_MEMORY_VIRTUAL) {
6014 6015
        if (qemuMonitorSaveVirtualMemory(priv->mon, offset, size, tmp) < 0) {
            qemuDomainObjExitMonitor(vm);
6016
            goto endjob;
6017
        }
6018
    } else {
6019 6020
        if (qemuMonitorSavePhysicalMemory(priv->mon, offset, size, tmp) < 0) {
            qemuDomainObjExitMonitor(vm);
6021
            goto endjob;
6022
        }
R
Richard W.M. Jones 已提交
6023
    }
6024
    qemuDomainObjExitMonitor(vm);
R
Richard W.M. Jones 已提交
6025 6026 6027

    /* Read the memory file into buffer. */
    if (saferead (fd, buffer, size) == (ssize_t) -1) {
6028 6029 6030
        virReportSystemError (dom->conn, errno,
                              _("failed to read temporary file "
                                "created with template %s"), tmp);
6031
        goto endjob;
R
Richard W.M. Jones 已提交
6032 6033 6034
    }

    ret = 0;
6035

6036 6037 6038
endjob:
    qemuDomainObjEndJob(vm);

6039
cleanup:
6040
    VIR_FREE(tmp);
R
Richard W.M. Jones 已提交
6041 6042
    if (fd >= 0) close (fd);
    unlink (tmp);
6043 6044
    if (vm)
        virDomainObjUnlock(vm);
R
Richard W.M. Jones 已提交
6045 6046 6047
    return ret;
}

6048

6049 6050
static int
qemudDomainEventRegister (virConnectPtr conn,
6051
                          virConnectDomainEventCallback callback,
6052 6053
                          void *opaque,
                          virFreeCallback freecb)
6054
{
6055 6056 6057
    struct qemud_driver *driver = conn->privateData;
    int ret;

6058
    qemuDriverLock(driver);
6059 6060
    ret = virDomainEventCallbackListAdd(conn, driver->domainEventCallbacks,
                                        callback, opaque, freecb);
6061
    qemuDriverUnlock(driver);
6062

6063
    return ret;
6064 6065 6066 6067
}

static int
qemudDomainEventDeregister (virConnectPtr conn,
6068
                            virConnectDomainEventCallback callback)
6069
{
6070 6071 6072
    struct qemud_driver *driver = conn->privateData;
    int ret;

6073
    qemuDriverLock(driver);
6074 6075 6076 6077 6078 6079
    if (driver->domainEventDispatching)
        ret = virDomainEventCallbackListMarkDelete(conn, driver->domainEventCallbacks,
                                                   callback);
    else
        ret = virDomainEventCallbackListRemove(conn, driver->domainEventCallbacks,
                                               callback);
6080
    qemuDriverUnlock(driver);
6081

6082
    return ret;
6083 6084
}

6085 6086 6087 6088 6089
static void qemuDomainEventDispatchFunc(virConnectPtr conn,
                                        virDomainEventPtr event,
                                        virConnectDomainEventCallback cb,
                                        void *cbopaque,
                                        void *opaque)
6090
{
6091
    struct qemud_driver *driver = opaque;
6092

6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121 6122 6123 6124 6125 6126 6127 6128 6129 6130 6131 6132 6133 6134 6135 6136
    /* Drop the lock whle dispatching, for sake of re-entrancy */
    qemuDriverUnlock(driver);
    virDomainEventDispatchDefaultFunc(conn, event, cb, cbopaque, NULL);
    qemuDriverLock(driver);
}

static void qemuDomainEventFlush(int timer ATTRIBUTE_UNUSED, void *opaque)
{
    struct qemud_driver *driver = opaque;
    virDomainEventQueue tempQueue;

    qemuDriverLock(driver);

    driver->domainEventDispatching = 1;

    /* Copy the queue, so we're reentrant safe */
    tempQueue.count = driver->domainEventQueue->count;
    tempQueue.events = driver->domainEventQueue->events;
    driver->domainEventQueue->count = 0;
    driver->domainEventQueue->events = NULL;

    virEventUpdateTimeout(driver->domainEventTimer, -1);
    virDomainEventQueueDispatch(&tempQueue,
                                driver->domainEventCallbacks,
                                qemuDomainEventDispatchFunc,
                                driver);

    /* Purge any deleted callbacks */
    virDomainEventCallbackListPurgeMarked(driver->domainEventCallbacks);

    driver->domainEventDispatching = 0;
    qemuDriverUnlock(driver);
}


/* driver must be locked before calling */
static void qemuDomainEventQueue(struct qemud_driver *driver,
                                 virDomainEventPtr event)
{
    if (virDomainEventQueuePush(driver->domainEventQueue,
                                event) < 0)
        virDomainEventFree(event);
    if (qemu_driver->domainEventQueue->count == 1)
        virEventUpdateTimeout(driver->domainEventTimer, 0);
6137 6138
}

D
Daniel Veillard 已提交
6139 6140
/* Migration support. */

C
Chris Lalancette 已提交
6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172 6173 6174 6175 6176 6177 6178 6179 6180 6181 6182 6183 6184 6185 6186 6187 6188 6189 6190 6191 6192 6193 6194 6195 6196 6197 6198 6199 6200 6201 6202 6203 6204 6205 6206 6207 6208 6209 6210 6211 6212 6213 6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231 6232 6233 6234 6235 6236 6237 6238 6239 6240 6241 6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256 6257 6258 6259 6260 6261 6262 6263 6264 6265 6266 6267 6268 6269 6270 6271 6272 6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 6285 6286 6287 6288 6289 6290 6291 6292 6293 6294 6295 6296 6297 6298 6299 6300 6301 6302 6303 6304 6305 6306 6307 6308 6309 6310 6311 6312 6313
/* Tunnelled migration stream support */
struct qemuStreamMigFile {
    int fd;

    int watch;
    unsigned int cbRemoved;
    unsigned int dispatching;
    virStreamEventCallback cb;
    void *opaque;
    virFreeCallback ff;
};

static int qemuStreamMigRemoveCallback(virStreamPtr stream)
{
    struct qemud_driver *driver = stream->conn->privateData;
    struct qemuStreamMigFile *qemust = stream->privateData;
    int ret = -1;

    if (!qemust) {
        qemudReportError(stream->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("stream is not open"));
        return -1;
    }

    qemuDriverLock(driver);
    if (qemust->watch == 0) {
        qemudReportError(stream->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("stream does not have a callback registered"));
        goto cleanup;
    }

    virEventRemoveHandle(qemust->watch);
    if (qemust->dispatching)
        qemust->cbRemoved = 1;
    else if (qemust->ff)
        (qemust->ff)(qemust->opaque);

    qemust->watch = 0;
    qemust->ff = NULL;
    qemust->cb = NULL;
    qemust->opaque = NULL;

    ret = 0;

cleanup:
    qemuDriverUnlock(driver);
    return ret;
}

static int qemuStreamMigUpdateCallback(virStreamPtr stream, int events)
{
    struct qemud_driver *driver = stream->conn->privateData;
    struct qemuStreamMigFile *qemust = stream->privateData;
    int ret = -1;

    if (!qemust) {
        qemudReportError(stream->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("stream is not open"));
        return -1;
    }

    qemuDriverLock(driver);
    if (qemust->watch == 0) {
        qemudReportError(stream->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("stream does not have a callback registered"));
        goto cleanup;
    }

    virEventUpdateHandle(qemust->watch, events);

    ret = 0;

cleanup:
    qemuDriverUnlock(driver);
    return ret;
}

static void qemuStreamMigEvent(int watch ATTRIBUTE_UNUSED,
                               int fd ATTRIBUTE_UNUSED,
                               int events,
                               void *opaque)
{
    virStreamPtr stream = opaque;
    struct qemud_driver *driver = stream->conn->privateData;
    struct qemuStreamMigFile *qemust = stream->privateData;
    virStreamEventCallback cb;
    void *cbopaque;
    virFreeCallback ff;

    qemuDriverLock(driver);
    if (!qemust || !qemust->cb) {
        qemuDriverUnlock(driver);
        return;
    }

    cb = qemust->cb;
    cbopaque = qemust->opaque;
    ff = qemust->ff;
    qemust->dispatching = 1;
    qemuDriverUnlock(driver);

    cb(stream, events, cbopaque);

    qemuDriverLock(driver);
    qemust->dispatching = 0;
    if (qemust->cbRemoved && ff)
        (ff)(cbopaque);
    qemuDriverUnlock(driver);
}

static int
qemuStreamMigAddCallback(virStreamPtr st,
                         int events,
                         virStreamEventCallback cb,
                         void *opaque,
                         virFreeCallback ff)
{
    struct qemud_driver *driver = st->conn->privateData;
    struct qemuStreamMigFile *qemust = st->privateData;
    int ret = -1;

    if (!qemust) {
        qemudReportError(st->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("stream is not open"));
        return -1;
    }

    qemuDriverLock(driver);
    if (qemust->watch != 0) {
        qemudReportError(st->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("stream already has a callback registered"));
        goto cleanup;
    }

    if ((qemust->watch = virEventAddHandle(qemust->fd,
                                           events,
                                           qemuStreamMigEvent,
                                           st,
                                           NULL)) < 0) {
        qemudReportError(st->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("cannot register file watch on stream"));
        goto cleanup;
    }

    qemust->cbRemoved = 0;
    qemust->cb = cb;
    qemust->opaque = opaque;
    qemust->ff = ff;
    virStreamRef(st);

    ret = 0;

cleanup:
    qemuDriverUnlock(driver);
    return ret;
}

static void qemuStreamMigFree(struct qemuStreamMigFile *qemust)
{
    if (qemust->fd != -1)
        close(qemust->fd);
    VIR_FREE(qemust);
}

static struct qemuStreamMigFile *qemuStreamMigOpen(virStreamPtr st,
                                                   const char *unixfile)
{
    struct qemuStreamMigFile *qemust = NULL;
    struct sockaddr_un sa_qemu;
    int i = 0;
    int timeout = 3;
    int ret;

6314 6315
    if (VIR_ALLOC(qemust) < 0) {
        virReportOOMError(st->conn);
C
Chris Lalancette 已提交
6316
        return NULL;
6317
    }
C
Chris Lalancette 已提交
6318 6319 6320 6321 6322 6323 6324 6325 6326 6327 6328 6329 6330 6331 6332 6333 6334 6335 6336 6337 6338 6339 6340 6341 6342 6343 6344 6345 6346 6347 6348 6349 6350 6351 6352 6353 6354 6355 6356 6357 6358 6359 6360 6361 6362 6363 6364 6365 6366 6367 6368 6369 6370 6371 6372 6373 6374 6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404 6405 6406 6407 6408 6409 6410 6411 6412 6413 6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428 6429 6430 6431 6432 6433 6434 6435 6436 6437 6438 6439 6440 6441 6442 6443 6444 6445 6446 6447 6448 6449 6450 6451 6452 6453 6454 6455 6456 6457 6458 6459 6460 6461 6462 6463 6464

    qemust->fd = socket(AF_UNIX, SOCK_STREAM, 0);
    if (qemust->fd < 0)
        goto cleanup;

    memset(&sa_qemu, 0, sizeof(sa_qemu));
    sa_qemu.sun_family = AF_UNIX;
    if (virStrcpy(sa_qemu.sun_path, unixfile, sizeof(sa_qemu.sun_path)) == NULL)
        goto cleanup;

    do {
        ret = connect(qemust->fd, (struct sockaddr *)&sa_qemu, sizeof(sa_qemu));
        if (ret == 0)
            break;

        if (errno == ENOENT || errno == ECONNREFUSED) {
            /* ENOENT       : Socket may not have shown up yet
             * ECONNREFUSED : Leftover socket hasn't been removed yet */
            continue;
        }

        goto cleanup;
    } while ((++i <= timeout*5) && (usleep(.2 * 1000000) <= 0));

    if ((st->flags & VIR_STREAM_NONBLOCK) && virSetNonBlock(qemust->fd) < 0)
        goto cleanup;

    return qemust;

cleanup:
    qemuStreamMigFree(qemust);
    return NULL;
}

static int
qemuStreamMigClose(virStreamPtr st)
{
    struct qemud_driver *driver = st->conn->privateData;
    struct qemuStreamMigFile *qemust = st->privateData;

    if (!qemust)
        return 0;

    qemuDriverLock(driver);

    qemuStreamMigFree(qemust);

    st->privateData = NULL;

    qemuDriverUnlock(driver);

    return 0;
}

static int qemuStreamMigWrite(virStreamPtr st, const char *bytes, size_t nbytes)
{
    struct qemud_driver *driver = st->conn->privateData;
    struct qemuStreamMigFile *qemust = st->privateData;
    int ret;

    if (!qemust) {
        qemudReportError(st->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("stream is not open"));
        return -1;
    }

    qemuDriverLock(driver);

retry:
    ret = write(qemust->fd, bytes, nbytes);
    if (ret < 0) {
        if (errno == EAGAIN || errno == EWOULDBLOCK) {
            ret = -2;
        } else if (errno == EINTR) {
            goto retry;
        } else {
            ret = -1;
            virReportSystemError(st->conn, errno, "%s",
                                 _("cannot write to stream"));
        }
    }

    qemuDriverUnlock(driver);
    return ret;
}

static virStreamDriver qemuStreamMigDrv = {
    .streamSend = qemuStreamMigWrite,
    .streamFinish = qemuStreamMigClose,
    .streamAbort = qemuStreamMigClose,
    .streamAddCallback = qemuStreamMigAddCallback,
    .streamUpdateCallback = qemuStreamMigUpdateCallback,
    .streamRemoveCallback = qemuStreamMigRemoveCallback
};

/* Prepare is the first step, and it runs on the destination host.
 *
 * This version starts an empty VM listening on a localhost TCP port, and
 * sets up the corresponding virStream to handle the incoming data.
 */
static int
qemudDomainMigratePrepareTunnel(virConnectPtr dconn,
                                virStreamPtr st,
                                unsigned long flags,
                                const char *dname,
                                unsigned long resource ATTRIBUTE_UNUSED,
                                const char *dom_xml)
{
    struct qemud_driver *driver = dconn->privateData;
    virDomainDefPtr def = NULL;
    virDomainObjPtr vm = NULL;
    char *migrateFrom;
    virDomainEventPtr event = NULL;
    int ret = -1;
    int internalret;
    char *unixfile = NULL;
    unsigned int qemuCmdFlags;
    struct qemuStreamMigFile *qemust = NULL;

    qemuDriverLock(driver);
    if (!dom_xml) {
        qemudReportError(dconn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("no domain XML passed"));
        goto cleanup;
    }
    if (!(flags & VIR_MIGRATE_TUNNELLED)) {
        qemudReportError(dconn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("PrepareTunnel called but no TUNNELLED flag set"));
        goto cleanup;
    }
    if (st == NULL) {
        qemudReportError(dconn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("tunnelled migration requested but NULL stream passed"));
        goto cleanup;
    }

    /* Parse the domain XML. */
    if (!(def = virDomainDefParseString(dconn, driver->caps, dom_xml,
                                        VIR_DOMAIN_XML_INACTIVE))) {
        qemudReportError(dconn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                         "%s", _("failed to parse XML"));
        goto cleanup;
    }

    /* Target domain name, maybe renamed. */
    dname = dname ? dname : def->name;

6465 6466
    if (virDomainObjIsDuplicate(&driver->domains, def, 1) < 0)
        goto cleanup;
C
Chris Lalancette 已提交
6467 6468

    if (!(vm = virDomainAssignDef(dconn,
6469
                                  driver->caps,
C
Chris Lalancette 已提交
6470 6471 6472 6473 6474 6475 6476 6477
                                  &driver->domains,
                                  def))) {
        qemudReportError(dconn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                         "%s", _("failed to assign new VM"));
        goto cleanup;
    }
    def = NULL;

6478 6479 6480
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

C
Chris Lalancette 已提交
6481 6482 6483 6484 6485 6486
    /* Domain starts inactive, even if the domain XML had an id field. */
    vm->def->id = -1;

    if (virAsprintf(&unixfile, "%s/qemu.tunnelmigrate.dest.%s",
                    driver->stateDir, vm->def->name) < 0) {
        virReportOOMError (dconn);
6487
        goto endjob;
C
Chris Lalancette 已提交
6488 6489 6490 6491 6492 6493 6494 6495
    }
    unlink(unixfile);

    /* check that this qemu version supports the interactive exec */
    if (qemudExtractVersionInfo(vm->def->emulator, NULL, &qemuCmdFlags) < 0) {
        qemudReportError(dconn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("Cannot determine QEMU argv syntax %s"),
                         vm->def->emulator);
6496
        goto endjob;
C
Chris Lalancette 已提交
6497 6498 6499 6500 6501 6502 6503 6504
    }
    if (qemuCmdFlags & QEMUD_CMD_FLAG_MIGRATE_QEMU_UNIX)
        internalret = virAsprintf(&migrateFrom, "unix:%s", unixfile);
    else if (qemuCmdFlags & QEMUD_CMD_FLAG_MIGRATE_QEMU_EXEC)
        internalret = virAsprintf(&migrateFrom, "exec:nc -U -l %s", unixfile);
    else {
        qemudReportError(dconn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                         "%s", _("Destination qemu is too old to support tunnelled migration"));
6505
        goto endjob;
C
Chris Lalancette 已提交
6506 6507 6508
    }
    if (internalret < 0) {
        virReportOOMError(dconn);
6509
        goto endjob;
C
Chris Lalancette 已提交
6510 6511 6512 6513 6514 6515 6516 6517 6518 6519 6520 6521 6522 6523
    }
    /* Start the QEMU daemon, with the same command-line arguments plus
     * -incoming unix:/path/to/file or exec:nc -U /path/to/file
     */
    internalret = qemudStartVMDaemon(dconn, driver, vm, migrateFrom, -1);
    VIR_FREE(migrateFrom);
    if (internalret < 0) {
        /* Note that we don't set an error here because qemudStartVMDaemon
         * should have already done that.
         */
        if (!vm->persistent) {
            virDomainRemoveInactive(&driver->domains, vm);
            vm = NULL;
        }
6524
        goto endjob;
C
Chris Lalancette 已提交
6525 6526 6527 6528 6529
    }

    qemust = qemuStreamMigOpen(st, unixfile);
    if (qemust == NULL) {
        qemudShutdownVMDaemon(NULL, driver, vm);
6530
        if (!vm->persistent) {
6531
            qemuDomainObjEndJob(vm);
6532 6533 6534
            virDomainRemoveInactive(&driver->domains, vm);
            vm = NULL;
        }
C
Chris Lalancette 已提交
6535 6536 6537
        virReportSystemError(dconn, errno,
                             _("cannot open unix socket '%s' for tunnelled migration"),
                             unixfile);
6538
        goto endjob;
C
Chris Lalancette 已提交
6539 6540 6541 6542 6543 6544 6545 6546 6547 6548
    }

    st->driver = &qemuStreamMigDrv;
    st->privateData = qemust;

    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STARTED,
                                     VIR_DOMAIN_EVENT_STARTED_MIGRATED);
    ret = 0;

6549 6550 6551 6552
endjob:
    if (vm)
        qemuDomainObjEndJob(vm);

C
Chris Lalancette 已提交
6553 6554 6555 6556 6557 6558 6559 6560 6561 6562 6563 6564
cleanup:
    virDomainDefFree(def);
    unlink(unixfile);
    VIR_FREE(unixfile);
    if (vm)
        virDomainObjUnlock(vm);
    if (event)
        qemuDomainEventQueue(driver, event);
    qemuDriverUnlock(driver);
    return ret;
}

D
Daniel Veillard 已提交
6565 6566 6567 6568 6569 6570 6571 6572 6573 6574
/* Prepare is the first step, and it runs on the destination host.
 *
 * This starts an empty VM listening on a TCP port.
 */
static int
qemudDomainMigratePrepare2 (virConnectPtr dconn,
                            char **cookie ATTRIBUTE_UNUSED,
                            int *cookielen ATTRIBUTE_UNUSED,
                            const char *uri_in,
                            char **uri_out,
C
Chris Lalancette 已提交
6575
                            unsigned long flags,
D
Daniel Veillard 已提交
6576 6577 6578 6579 6580
                            const char *dname,
                            unsigned long resource ATTRIBUTE_UNUSED,
                            const char *dom_xml)
{
    static int port = 0;
6581 6582
    struct qemud_driver *driver = dconn->privateData;
    virDomainDefPtr def = NULL;
D
Daniel Veillard 已提交
6583 6584
    virDomainObjPtr vm = NULL;
    int this_port;
6585
    char *hostname;
D
Daniel Veillard 已提交
6586 6587
    char migrateFrom [64];
    const char *p;
6588
    virDomainEventPtr event = NULL;
6589
    int ret = -1;
6590
    int internalret;
6591 6592

    *uri_out = NULL;
D
Daniel Veillard 已提交
6593

6594
    qemuDriverLock(driver);
C
Chris Lalancette 已提交
6595 6596 6597 6598 6599 6600 6601 6602 6603
    if (flags & VIR_MIGRATE_TUNNELLED) {
        /* this is a logical error; we never should have gotten here with
         * VIR_MIGRATE_TUNNELLED set
         */
        qemudReportError(dconn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         "%s", _("Tunnelled migration requested but invalid RPC method called"));
        goto cleanup;
    }

D
Daniel Veillard 已提交
6604 6605 6606
    if (!dom_xml) {
        qemudReportError (dconn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                          "%s", _("no domain XML passed"));
6607
        goto cleanup;
D
Daniel Veillard 已提交
6608 6609 6610 6611 6612 6613 6614 6615 6616 6617 6618 6619 6620 6621 6622 6623 6624
    }

    /* The URI passed in may be NULL or a string "tcp://somehostname:port".
     *
     * If the URI passed in is NULL then we allocate a port number
     * from our pool of port numbers and return a URI of
     * "tcp://ourhostname:port".
     *
     * If the URI passed in is not NULL then we try to parse out the
     * port number and use that (note that the hostname is assumed
     * to be a correct hostname which refers to the target machine).
     */
    if (uri_in == NULL) {
        this_port = QEMUD_MIGRATION_FIRST_PORT + port++;
        if (port == QEMUD_MIGRATION_NUM_PORTS) port = 0;

        /* Get hostname */
6625
        if ((hostname = virGetHostname(dconn)) == NULL)
6626
            goto cleanup;
D
Daniel Veillard 已提交
6627

6628 6629 6630 6631 6632
        /* XXX this really should have been a properly well-formed
         * URI, but we can't add in tcp:// now without breaking
         * compatability with old targets. We at least make the
         * new targets accept both syntaxes though.
         */
D
Daniel Veillard 已提交
6633
        /* Caller frees */
6634 6635 6636
        internalret = virAsprintf(uri_out, "tcp:%s:%d", hostname, this_port);
        VIR_FREE(hostname);
        if (internalret < 0) {
6637
            virReportOOMError (dconn);
6638
            goto cleanup;
D
Daniel Veillard 已提交
6639 6640 6641 6642 6643 6644
        }
    } else {
        /* Check the URI starts with "tcp:".  We will escape the
         * URI when passing it to the qemu monitor, so bad
         * characters in hostname part don't matter.
         */
6645
        if (!STRPREFIX (uri_in, "tcp:")) {
D
Daniel Veillard 已提交
6646
            qemudReportError (dconn, NULL, NULL, VIR_ERR_INVALID_ARG,
6647
                  "%s", _("only tcp URIs are supported for KVM/QEMU migrations"));
6648
            goto cleanup;
D
Daniel Veillard 已提交
6649 6650 6651 6652
        }

        /* Get the port number. */
        p = strrchr (uri_in, ':');
6653 6654 6655 6656 6657 6658 6659 6660 6661 6662 6663 6664 6665 6666 6667 6668 6669 6670 6671 6672
        if (p == strchr(uri_in, ':')) {
            /* Generate a port */
            this_port = QEMUD_MIGRATION_FIRST_PORT + port++;
            if (port == QEMUD_MIGRATION_NUM_PORTS)
                port = 0;

            /* Caller frees */
            if (virAsprintf(uri_out, "%s:%d", uri_in, this_port) < 0) {
                virReportOOMError (dconn);
                goto cleanup;
            }

        } else {
            p++; /* definitely has a ':' in it, see above */
            this_port = virParseNumber (&p);
            if (this_port == -1 || p-uri_in != strlen (uri_in)) {
                qemudReportError (dconn, NULL, NULL, VIR_ERR_INVALID_ARG,
                                  "%s", _("URI ended with incorrect ':port'"));
                goto cleanup;
            }
D
Daniel Veillard 已提交
6673 6674 6675
        }
    }

6676 6677 6678
    if (uri_out && *uri_out)
        VIR_DEBUG("Generated uri_out=%s", *uri_out);

D
Daniel Veillard 已提交
6679
    /* Parse the domain XML. */
6680 6681
    if (!(def = virDomainDefParseString(dconn, driver->caps, dom_xml,
                                        VIR_DOMAIN_XML_INACTIVE))) {
D
Daniel Veillard 已提交
6682 6683
        qemudReportError (dconn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                          "%s", _("failed to parse XML"));
6684
        goto cleanup;
D
Daniel Veillard 已提交
6685 6686 6687 6688 6689
    }

    /* Target domain name, maybe renamed. */
    dname = dname ? dname : def->name;

6690 6691
    if (virDomainObjIsDuplicate(&driver->domains, def, 1) < 0)
        goto cleanup;
D
Daniel Veillard 已提交
6692 6693

    if (!(vm = virDomainAssignDef(dconn,
6694
                                  driver->caps,
D
Daniel Veillard 已提交
6695 6696 6697 6698
                                  &driver->domains,
                                  def))) {
        qemudReportError (dconn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                          "%s", _("failed to assign new VM"));
6699
        goto cleanup;
D
Daniel Veillard 已提交
6700
    }
6701
    def = NULL;
D
Daniel Veillard 已提交
6702

6703 6704 6705
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

D
Daniel Veillard 已提交
6706 6707 6708 6709 6710 6711 6712
    /* Domain starts inactive, even if the domain XML had an id field. */
    vm->def->id = -1;

    /* Start the QEMU daemon, with the same command-line arguments plus
     * -incoming tcp:0.0.0.0:port
     */
    snprintf (migrateFrom, sizeof (migrateFrom), "tcp:0.0.0.0:%d", this_port);
6713
    if (qemudStartVMDaemon (dconn, driver, vm, migrateFrom, -1) < 0) {
6714 6715 6716
        /* Note that we don't set an error here because qemudStartVMDaemon
         * should have already done that.
         */
6717
        if (!vm->persistent) {
6718
            qemuDomainObjEndJob(vm);
D
Daniel Veillard 已提交
6719
            virDomainRemoveInactive(&driver->domains, vm);
6720 6721
            vm = NULL;
        }
6722
        goto endjob;
D
Daniel Veillard 已提交
6723
    }
6724 6725 6726 6727

    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STARTED,
                                     VIR_DOMAIN_EVENT_STARTED_MIGRATED);
6728
    ret = 0;
D
Daniel Veillard 已提交
6729

6730 6731 6732 6733
endjob:
    if (vm)
        qemuDomainObjEndJob(vm);

6734 6735 6736 6737 6738
cleanup:
    virDomainDefFree(def);
    if (ret != 0) {
        VIR_FREE(*uri_out);
    }
6739 6740
    if (vm)
        virDomainObjUnlock(vm);
6741 6742
    if (event)
        qemuDomainEventQueue(driver, event);
6743
    qemuDriverUnlock(driver);
6744
    return ret;
C
Chris Lalancette 已提交
6745 6746 6747

}

6748 6749 6750 6751 6752

/* Perform migration using QEMU's native TCP migrate support,
 * not encrypted obviously
 */
static int doNativeMigrate(virDomainPtr dom,
6753
                           struct qemud_driver *driver,
6754 6755 6756 6757 6758 6759 6760
                           virDomainObjPtr vm,
                           const char *uri,
                           unsigned long flags ATTRIBUTE_UNUSED,
                           const char *dname ATTRIBUTE_UNUSED,
                           unsigned long resource)
{
    int ret = -1;
6761
    xmlURIPtr uribits = NULL;
6762 6763
    int status;
    unsigned long long transferred, remaining, total;
6764
    qemuDomainObjPrivatePtr priv = vm->privateData;
6765 6766 6767 6768 6769 6770 6771 6772 6773 6774 6775 6776 6777 6778 6779 6780 6781 6782 6783 6784

    /* Issue the migrate command. */
    if (STRPREFIX(uri, "tcp:") && !STRPREFIX(uri, "tcp://")) {
        /* HACK: source host generates bogus URIs, so fix them up */
        char *tmpuri;
        if (virAsprintf(&tmpuri, "tcp://%s", uri + strlen("tcp:")) < 0) {
            virReportOOMError(dom->conn);
            goto cleanup;
        }
        uribits = xmlParseURI(tmpuri);
        VIR_FREE(tmpuri);
    } else {
        uribits = xmlParseURI(uri);
    }
    if (!uribits) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("cannot parse URI %s"), uri);
        goto cleanup;
    }

6785
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
6786
    if (resource > 0 &&
6787
        qemuMonitorSetMigrationSpeed(priv->mon, resource) < 0) {
6788
        qemuDomainObjExitMonitorWithDriver(driver, vm);
6789
        goto cleanup;
6790
    }
6791

6792
    if (qemuMonitorMigrateToHost(priv->mon, 0, uribits->server, uribits->port) < 0) {
6793
        qemuDomainObjExitMonitorWithDriver(driver, vm);
6794
        goto cleanup;
6795
    }
6796 6797 6798 6799

    /* it is also possible that the migrate didn't fail initially, but
     * rather failed later on.  Check the output of "info migrate"
     */
6800 6801
    if (qemuMonitorGetMigrationStatus(priv->mon,
                                      &status,
6802 6803 6804
                                      &transferred,
                                      &remaining,
                                      &total) < 0) {
6805
        qemuDomainObjExitMonitorWithDriver(driver, vm);
6806 6807
        goto cleanup;
    }
6808
    qemuDomainObjExitMonitorWithDriver(driver, vm);
6809 6810 6811 6812 6813 6814 6815 6816 6817 6818 6819 6820 6821 6822 6823

    if (status != QEMU_MONITOR_MIGRATION_STATUS_COMPLETED) {
        qemudReportError (dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
                          "%s", _("migrate did not successfully complete"));
        goto cleanup;
    }

    ret = 0;

cleanup:
    xmlFreeURI(uribits);
    return ret;
}


6824 6825 6826 6827 6828 6829 6830 6831 6832 6833 6834 6835 6836 6837 6838 6839 6840 6841 6842 6843 6844
static int doTunnelSendAll(virDomainPtr dom,
                           virStreamPtr st,
                           int sock)
{
    char buffer[65536];
    int nbytes = sizeof(buffer);

    /* XXX should honour the 'resource' parameter here */
    for (;;) {
        nbytes = saferead(sock, buffer, nbytes);
        if (nbytes < 0) {
            virStreamAbort(st);
            virReportSystemError(dom->conn, errno, "%s",
                                 _("tunnelled migration failed to read from qemu"));
            return -1;
        }
        else if (nbytes == 0)
            /* EOF; get out of here */
            break;

        if (virStreamSend(st, buffer, nbytes) < 0) {
C
Chris Lalancette 已提交
6845
            qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED, "%s",
6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857
                             _("Failed to write migration data to remote libvirtd"));
            return -1;
        }
    }

    if (virStreamFinish(st) < 0)
        /* virStreamFinish set the error for us */
        return -1;

    return 0;
}

C
Chris Lalancette 已提交
6858
static int doTunnelMigrate(virDomainPtr dom,
6859
                           struct qemud_driver *driver,
6860
                           virConnectPtr dconn,
C
Chris Lalancette 已提交
6861
                           virDomainObjPtr vm,
6862
                           const char *dom_xml,
C
Chris Lalancette 已提交
6863 6864 6865 6866 6867
                           const char *uri,
                           unsigned long flags,
                           const char *dname,
                           unsigned long resource)
{
6868
    qemuDomainObjPrivatePtr priv = vm->privateData;
6869 6870
    int client_sock = -1;
    int qemu_sock = -1;
C
Chris Lalancette 已提交
6871 6872
    struct sockaddr_un sa_qemu, sa_client;
    socklen_t addrlen;
6873
    virDomainPtr ddomain = NULL;
C
Chris Lalancette 已提交
6874
    int retval = -1;
6875
    virStreamPtr st = NULL;
C
Chris Lalancette 已提交
6876 6877 6878 6879 6880 6881
    char *unixfile = NULL;
    int internalret;
    unsigned int qemuCmdFlags;
    int status;
    unsigned long long transferred, remaining, total;

6882 6883 6884 6885 6886 6887 6888 6889
    /*
     * The order of operations is important here to avoid touching
     * the source VM until we are very sure we can successfully
     * start the migration operation.
     *
     *   1. setup local support infrastructure (eg sockets)
     *   2. setup destination fully
     *   3. start migration on source
C
Chris Lalancette 已提交
6890 6891
     */

6892

6893
    /* Stage 1. setup local support infrastructure */
C
Chris Lalancette 已提交
6894 6895 6896 6897

    if (virAsprintf(&unixfile, "%s/qemu.tunnelmigrate.src.%s",
                    driver->stateDir, vm->def->name) < 0) {
        virReportOOMError(dom->conn);
6898
        goto cleanup;
C
Chris Lalancette 已提交
6899 6900 6901 6902 6903 6904
    }

    qemu_sock = socket(AF_UNIX, SOCK_STREAM, 0);
    if (qemu_sock < 0) {
        virReportSystemError(dom->conn, errno, "%s",
                             _("cannot open tunnelled migration socket"));
6905
        goto cleanup;
C
Chris Lalancette 已提交
6906 6907 6908 6909 6910 6911 6912 6913
    }
    memset(&sa_qemu, 0, sizeof(sa_qemu));
    sa_qemu.sun_family = AF_UNIX;
    if (virStrcpy(sa_qemu.sun_path, unixfile,
                  sizeof(sa_qemu.sun_path)) == NULL) {
        qemudReportError(dom->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("Unix socket '%s' too big for destination"),
                         unixfile);
6914
        goto cleanup;
C
Chris Lalancette 已提交
6915 6916 6917 6918 6919 6920
    }
    unlink(unixfile);
    if (bind(qemu_sock, (struct sockaddr *)&sa_qemu, sizeof(sa_qemu)) < 0) {
        virReportSystemError(dom->conn, errno,
                             _("Cannot bind to unix socket '%s' for tunnelled migration"),
                             unixfile);
6921
        goto cleanup;
C
Chris Lalancette 已提交
6922 6923 6924 6925 6926
    }
    if (listen(qemu_sock, 1) < 0) {
        virReportSystemError(dom->conn, errno,
                             _("Cannot listen on unix socket '%s' for tunnelled migration"),
                             unixfile);
6927
        goto cleanup;
C
Chris Lalancette 已提交
6928 6929 6930 6931 6932 6933 6934
    }

    /* check that this qemu version supports the unix migration */
    if (qemudExtractVersionInfo(vm->def->emulator, NULL, &qemuCmdFlags) < 0) {
        qemudReportError(dom->conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                         _("Cannot extract Qemu version from '%s'"),
                         vm->def->emulator);
6935 6936 6937 6938 6939 6940 6941 6942
        goto cleanup;
    }

    if (!(qemuCmdFlags & QEMUD_CMD_FLAG_MIGRATE_QEMU_UNIX) &&
        !(qemuCmdFlags & QEMUD_CMD_FLAG_MIGRATE_QEMU_EXEC)) {
        qemudReportError(dom->conn, NULL, NULL, VIR_ERR_OPERATION_FAILED,
                         "%s", _("Source qemu is too old to support tunnelled migration"));
        goto cleanup;
C
Chris Lalancette 已提交
6943
    }
6944 6945 6946 6947 6948 6949 6950 6951 6952 6953 6954 6955 6956 6957 6958 6959 6960 6961 6962 6963 6964


    /* Stage 2. setup destination fully
     *
     * Once stage 2 has completed successfully, we *must* call finish
     * to cleanup the target whether we succeed or fail
     */
    st = virStreamNew(dconn, 0);
    if (st == NULL)
        /* virStreamNew only fails on OOM, and it reports the error itself */
        goto cleanup;

    internalret = dconn->driver->domainMigratePrepareTunnel(dconn, st,
                                                            flags, dname,
                                                            resource, dom_xml);

    if (internalret < 0)
        /* domainMigratePrepareTunnel sets the error for us */
        goto cleanup;

    /*   3. start migration on source */
6965
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
C
Chris Lalancette 已提交
6966
    if (qemuCmdFlags & QEMUD_CMD_FLAG_MIGRATE_QEMU_UNIX)
6967
        internalret = qemuMonitorMigrateToUnix(priv->mon, 1, unixfile);
C
Chris Lalancette 已提交
6968 6969
    else if (qemuCmdFlags & QEMUD_CMD_FLAG_MIGRATE_QEMU_EXEC) {
        const char *args[] = { "nc", "-U", unixfile, NULL };
6970
        internalret = qemuMonitorMigrateToCommand(priv->mon, 1, args, "/dev/null");
6971 6972
    } else {
        internalret = -1;
C
Chris Lalancette 已提交
6973
    }
6974
    qemuDomainObjExitMonitorWithDriver(driver, vm);
C
Chris Lalancette 已提交
6975 6976 6977
    if (internalret < 0) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
                         "%s", _("tunnelled migration monitor command failed"));
6978
        goto finish;
C
Chris Lalancette 已提交
6979 6980
    }

6981 6982 6983
    /* From this point onwards we *must* call cancel to abort the
     * migration on source if anything goes wrong */

C
Chris Lalancette 已提交
6984 6985 6986
    /* it is also possible that the migrate didn't fail initially, but
     * rather failed later on.  Check the output of "info migrate"
     */
6987
    qemuDomainObjEnterMonitorWithDriver(driver, vm);
6988 6989
    if (qemuMonitorGetMigrationStatus(priv->mon,
                                      &status,
C
Chris Lalancette 已提交
6990 6991 6992
                                      &transferred,
                                      &remaining,
                                      &total) < 0) {
6993
        qemuDomainObjExitMonitorWithDriver(driver, vm);
6994
        goto cancel;
C
Chris Lalancette 已提交
6995
    }
6996
    qemuDomainObjExitMonitorWithDriver(driver, vm);
C
Chris Lalancette 已提交
6997 6998 6999 7000

    if (status == QEMU_MONITOR_MIGRATION_STATUS_ERROR) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
                         "%s",_("migrate failed"));
7001
        goto cancel;
C
Chris Lalancette 已提交
7002 7003 7004 7005 7006 7007 7008 7009
    }

    addrlen = sizeof(sa_client);
    while ((client_sock = accept(qemu_sock, (struct sockaddr *)&sa_client, &addrlen)) < 0) {
        if (errno == EAGAIN || errno == EINTR)
            continue;
        virReportSystemError(dom->conn, errno, "%s",
                             _("tunnelled migration failed to accept from qemu"));
7010
        goto cancel;
C
Chris Lalancette 已提交
7011 7012
    }

7013
    retval = doTunnelSendAll(dom, st, client_sock);
7014

7015
cancel:
7016
    if (retval != 0) {
7017
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
7018
        qemuMonitorMigrateCancel(priv->mon);
7019
        qemuDomainObjExitMonitorWithDriver(driver, vm);
7020
    }
C
Chris Lalancette 已提交
7021

7022
finish:
C
Chris Lalancette 已提交
7023 7024 7025
    dname = dname ? dname : dom->name;
    ddomain = dconn->driver->domainMigrateFinish2
        (dconn, dname, NULL, 0, uri, flags, retval);
7026 7027 7028 7029 7030 7031 7032

cleanup:
    if (client_sock != -1)
        close(client_sock);
    if (qemu_sock != -1)
        close(qemu_sock);

C
Chris Lalancette 已提交
7033 7034 7035
    if (ddomain)
        virUnrefDomain(ddomain);

7036 7037 7038 7039
    if (unixfile) {
        unlink(unixfile);
        VIR_FREE(unixfile);
    }
C
Chris Lalancette 已提交
7040

7041 7042 7043
    if (st)
        /* don't call virStreamFree(), because that resets any pending errors */
        virUnrefStream(st);
7044 7045 7046 7047
    return retval;
}


7048 7049 7050 7051
/* This is essentially a simplified re-impl of
 * virDomainMigrateVersion2 from libvirt.c, but running in source
 * libvirtd context, instead of client app context */
static int doNonTunnelMigrate(virDomainPtr dom,
7052
                              struct qemud_driver *driver,
7053 7054 7055 7056 7057 7058 7059 7060 7061 7062 7063 7064 7065 7066 7067 7068 7069 7070 7071 7072 7073 7074 7075 7076 7077
                              virConnectPtr dconn,
                              virDomainObjPtr vm,
                              const char *dom_xml,
                              const char *uri ATTRIBUTE_UNUSED,
                              unsigned long flags,
                              const char *dname,
                              unsigned long resource)
{
    virDomainPtr ddomain = NULL;
    int retval = -1;
    char *uri_out = NULL;

    /* NB we don't pass 'uri' into this, since that's the libvirtd
     * URI in this context - so we let dest pick it */
    if (dconn->driver->domainMigratePrepare2(dconn,
                                             NULL, /* cookie */
                                             0, /* cookielen */
                                             NULL, /* uri */
                                             &uri_out,
                                             flags, dname,
                                             resource, dom_xml) < 0)
        /* domainMigratePrepare2 sets the error for us */
        goto cleanup;

    if (uri_out == NULL) {
C
Chris Lalancette 已提交
7078
        qemudReportError(NULL, NULL, NULL, VIR_ERR_INTERNAL_ERROR, "%s",
7079 7080 7081
                         _("domainMigratePrepare2 did not set uri"));
    }

7082
    if (doNativeMigrate(dom, driver, vm, uri_out, flags, dname, resource) < 0)
7083 7084 7085 7086 7087 7088 7089 7090 7091 7092 7093 7094 7095 7096 7097 7098 7099
        goto finish;

    retval = 0;

finish:
    dname = dname ? dname : dom->name;
    ddomain = dconn->driver->domainMigrateFinish2
        (dconn, dname, NULL, 0, uri_out, flags, retval);

    if (ddomain)
        virUnrefDomain(ddomain);

cleanup:
    return retval;
}


7100
static int doPeer2PeerMigrate(virDomainPtr dom,
7101
                              struct qemud_driver *driver,
7102 7103 7104 7105 7106 7107 7108 7109 7110 7111 7112 7113 7114 7115 7116 7117 7118 7119 7120 7121 7122
                              virDomainObjPtr vm,
                              const char *uri,
                              unsigned long flags,
                              const char *dname,
                              unsigned long resource)
{
    int ret = -1;
    virConnectPtr dconn = NULL;
    char *dom_xml;

    /* the order of operations is important here; we make sure the
     * destination side is completely setup before we touch the source
     */

    dconn = virConnectOpen(uri);
    if (dconn == NULL) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
                         _("Failed to connect to remote libvirt URI %s"), uri);
        return -1;
    }
    if (!VIR_DRV_SUPPORTS_FEATURE(dconn->driver, dconn,
7123
                                  VIR_DRV_FEATURE_MIGRATION_P2P)) {
7124
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED, "%s",
7125
                         _("Destination libvirt does not support peer-to-peer migration protocol"));
7126 7127 7128 7129 7130 7131 7132 7133 7134 7135
        goto cleanup;
    }

    dom_xml = virDomainDefFormat(dom->conn, vm->def, VIR_DOMAIN_XML_SECURE);
    if (!dom_xml) {
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_OPERATION_FAILED,
                         "%s", _("failed to get domain xml"));
        goto cleanup;
    }

7136
    if (flags & VIR_MIGRATE_TUNNELLED)
7137
        ret = doTunnelMigrate(dom, driver, dconn, vm, dom_xml, uri, flags, dname, resource);
7138
    else
7139
        ret = doNonTunnelMigrate(dom, driver, dconn, vm, dom_xml, uri, flags, dname, resource);
7140 7141 7142

cleanup:
    VIR_FREE(dom_xml);
C
Chris Lalancette 已提交
7143 7144 7145
    /* don't call virConnectClose(), because that resets any pending errors */
    virUnrefConnect(dconn);

7146
    return ret;
D
Daniel Veillard 已提交
7147 7148
}

7149

D
Daniel Veillard 已提交
7150 7151 7152 7153 7154 7155
/* Perform is the second step, and it runs on the source host. */
static int
qemudDomainMigratePerform (virDomainPtr dom,
                           const char *cookie ATTRIBUTE_UNUSED,
                           int cookielen ATTRIBUTE_UNUSED,
                           const char *uri,
7156
                           unsigned long flags,
7157
                           const char *dname,
D
Daniel Veillard 已提交
7158 7159
                           unsigned long resource)
{
7160 7161
    struct qemud_driver *driver = dom->conn->privateData;
    virDomainObjPtr vm;
7162
    virDomainEventPtr event = NULL;
7163
    int ret = -1;
7164
    int paused = 0;
D
Daniel Veillard 已提交
7165

7166
    qemuDriverLock(driver);
7167
    vm = virDomainFindByUUID(&driver->domains, dom->uuid);
D
Daniel Veillard 已提交
7168
    if (!vm) {
7169 7170 7171 7172
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(dom->uuid, uuidstr);
        qemudReportError(dom->conn, dom, NULL, VIR_ERR_NO_DOMAIN,
                         _("no domain with matching uuid '%s'"), uuidstr);
7173
        goto cleanup;
D
Daniel Veillard 已提交
7174 7175
    }

7176 7177 7178
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
7179
    if (!virDomainObjIsActive(vm)) {
7180
        qemudReportError (dom->conn, dom, NULL, VIR_ERR_OPERATION_INVALID,
D
Daniel Veillard 已提交
7181
                          "%s", _("domain is not running"));
7182
        goto endjob;
D
Daniel Veillard 已提交
7183 7184
    }

7185
    if (!(flags & VIR_MIGRATE_LIVE)) {
7186
        qemuDomainObjPrivatePtr priv = vm->privateData;
7187
        /* Pause domain for non-live migration */
7188
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
7189
        if (qemuMonitorStopCPUs(priv->mon) < 0) {
7190 7191
            qemuDomainObjExitMonitorWithDriver(driver, vm);
            goto endjob;
7192
        }
7193
        qemuDomainObjExitMonitorWithDriver(driver, vm);
7194
        paused = 1;
7195

7196 7197 7198 7199 7200 7201
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_MIGRATED);
        if (event)
            qemuDomainEventQueue(driver, event);
        event = NULL;
7202 7203
    }

7204
    if ((flags & (VIR_MIGRATE_TUNNELLED | VIR_MIGRATE_PEER2PEER))) {
7205
        if (doPeer2PeerMigrate(dom, driver, vm, uri, flags, dname, resource) < 0)
7206
            /* doPeer2PeerMigrate already set the error, so just get out */
7207
            goto endjob;
7208
    } else {
7209 7210
        if (doNativeMigrate(dom, driver, vm, uri, flags, dname, resource) < 0)
            goto endjob;
7211 7212
    }

D
Daniel Veillard 已提交
7213 7214
    /* Clean up the source domain. */
    qemudShutdownVMDaemon (dom->conn, driver, vm);
7215
    paused = 0;
7216 7217 7218 7219

    event = virDomainEventNewFromObj(vm,
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_MIGRATED);
C
Chris Lalancette 已提交
7220 7221
    if (!vm->persistent || (flags & VIR_MIGRATE_UNDEFINE_SOURCE)) {
        virDomainDeleteConfig(dom->conn, driver->configDir, driver->autostartDir, vm);
7222
        qemuDomainObjEndJob(vm);
D
Daniel Veillard 已提交
7223
        virDomainRemoveInactive(&driver->domains, vm);
7224 7225
        vm = NULL;
    }
7226
    ret = 0;
D
Daniel Veillard 已提交
7227

7228
endjob:
7229
    if (paused) {
7230
        qemuDomainObjPrivatePtr priv = vm->privateData;
7231
        /* we got here through some sort of failure; start the domain again */
7232
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
7233
        if (qemuMonitorStartCPUs(priv->mon, dom->conn) < 0) {
7234 7235 7236 7237
            /* Hm, we already know we are in error here.  We don't want to
             * overwrite the previous error, though, so we just throw something
             * to the logs and hope for the best
             */
7238 7239
            VIR_ERROR(_("Failed to resume guest %s after failure\n"),
                      vm->def->name);
7240
        }
7241
        qemuDomainObjExitMonitorWithDriver(driver, vm);
7242 7243 7244 7245 7246

        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
    }
7247 7248
    if (vm)
        qemuDomainObjEndJob(vm);
7249

7250
cleanup:
7251 7252
    if (vm)
        virDomainObjUnlock(vm);
7253 7254
    if (event)
        qemuDomainEventQueue(driver, event);
7255
    qemuDriverUnlock(driver);
7256
    return ret;
D
Daniel Veillard 已提交
7257 7258 7259 7260 7261 7262 7263 7264 7265
}

/* Finish is the third and final step, and it runs on the destination host. */
static virDomainPtr
qemudDomainMigrateFinish2 (virConnectPtr dconn,
                           const char *dname,
                           const char *cookie ATTRIBUTE_UNUSED,
                           int cookielen ATTRIBUTE_UNUSED,
                           const char *uri ATTRIBUTE_UNUSED,
C
Chris Lalancette 已提交
7266
                           unsigned long flags,
D
Daniel Veillard 已提交
7267 7268
                           int retcode)
{
7269 7270 7271
    struct qemud_driver *driver = dconn->privateData;
    virDomainObjPtr vm;
    virDomainPtr dom = NULL;
7272
    virDomainEventPtr event = NULL;
C
Chris Lalancette 已提交
7273
    int newVM = 1;
D
Daniel Veillard 已提交
7274

7275
    qemuDriverLock(driver);
7276
    vm = virDomainFindByName(&driver->domains, dname);
D
Daniel Veillard 已提交
7277
    if (!vm) {
7278 7279
        qemudReportError (dconn, NULL, NULL, VIR_ERR_NO_DOMAIN,
                          _("no domain with matching name '%s'"), dname);
7280
        goto cleanup;
D
Daniel Veillard 已提交
7281 7282
    }

7283 7284 7285
    if (qemuDomainObjBeginJobWithDriver(driver, vm) < 0)
        goto cleanup;

D
Daniel Veillard 已提交
7286 7287 7288 7289
    /* Did the migration go as planned?  If yes, return the domain
     * object, but if no, clean up the empty qemu process.
     */
    if (retcode == 0) {
C
Chris Lalancette 已提交
7290 7291 7292 7293 7294 7295 7296 7297 7298 7299 7300 7301 7302 7303 7304 7305
        if (flags & VIR_MIGRATE_PERSIST_DEST) {
            if (vm->persistent)
                newVM = 0;
            vm->persistent = 1;

            if (virDomainSaveConfig(dconn, driver->configDir, vm->def) < 0) {
                /* Hmpf.  Migration was successful, but making it persistent
                 * was not.  If we report successful, then when this domain
                 * shuts down, management tools are in for a surprise.  On the
                 * other hand, if we report failure, then the management tools
                 * might try to restart the domain on the source side, even
                 * though the domain is actually running on the destination.
                 * Return a NULL dom pointer, and hope that this is a rare
                 * situation and management tools are smart.
                 */
                vm = NULL;
7306
                goto endjob;
C
Chris Lalancette 已提交
7307 7308 7309 7310 7311 7312 7313 7314 7315 7316 7317
            }

            event = virDomainEventNewFromObj(vm,
                                             VIR_DOMAIN_EVENT_DEFINED,
                                             newVM ?
                                             VIR_DOMAIN_EVENT_DEFINED_ADDED :
                                             VIR_DOMAIN_EVENT_DEFINED_UPDATED);
            if (event)
                qemuDomainEventQueue(driver, event);

        }
7318
        qemuDomainObjPrivatePtr priv = vm->privateData;
D
Daniel Veillard 已提交
7319
        dom = virGetDomain (dconn, vm->def->name, vm->def->uuid);
7320 7321 7322 7323 7324

        /* run 'cont' on the destination, which allows migration on qemu
         * >= 0.10.6 to work properly.  This isn't strictly necessary on
         * older qemu's, but it also doesn't hurt anything there
         */
7325
        qemuDomainObjEnterMonitorWithDriver(driver, vm);
7326
        if (qemuMonitorStartCPUs(priv->mon, dconn) < 0) {
7327 7328 7329
            if (virGetLastError() == NULL)
                qemudReportError(dconn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,
                                 "%s", _("resume operation failed"));
7330 7331
            qemuDomainObjExitMonitorWithDriver(driver, vm);
            goto endjob;
7332
        }
7333
        qemuDomainObjExitMonitorWithDriver(driver, vm);
7334

D
Daniel Veillard 已提交
7335
        vm->state = VIR_DOMAIN_RUNNING;
7336 7337 7338
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_MIGRATED);
D
Daniel Veillard 已提交
7339
        virDomainSaveStatus(dconn, driver->stateDir, vm);
D
Daniel Veillard 已提交
7340 7341
    } else {
        qemudShutdownVMDaemon (dconn, driver, vm);
7342 7343 7344
        event = virDomainEventNewFromObj(vm,
                                         VIR_DOMAIN_EVENT_STOPPED,
                                         VIR_DOMAIN_EVENT_STOPPED_FAILED);
7345
        if (!vm->persistent) {
7346
            qemuDomainObjEndJob(vm);
D
Daniel Veillard 已提交
7347
            virDomainRemoveInactive(&driver->domains, vm);
7348 7349
            vm = NULL;
        }
D
Daniel Veillard 已提交
7350
    }
7351

7352 7353 7354 7355
endjob:
    if (vm)
        qemuDomainObjEndJob(vm);

7356
cleanup:
7357 7358
    if (vm)
        virDomainObjUnlock(vm);
7359 7360
    if (event)
        qemuDomainEventQueue(driver, event);
7361
    qemuDriverUnlock(driver);
7362
    return dom;
D
Daniel Veillard 已提交
7363 7364
}

7365 7366 7367 7368 7369 7370 7371 7372 7373 7374 7375 7376 7377 7378 7379 7380
static int
qemudNodeDeviceGetPciInfo (virNodeDevicePtr dev,
                           unsigned *domain,
                           unsigned *bus,
                           unsigned *slot,
                           unsigned *function)
{
    virNodeDeviceDefPtr def = NULL;
    virNodeDevCapsDefPtr cap;
    char *xml = NULL;
    int ret = -1;

    xml = virNodeDeviceGetXMLDesc(dev, 0);
    if (!xml)
        goto out;

7381
    def = virNodeDeviceDefParseString(dev->conn, xml, EXISTING_DEVICE);
7382 7383 7384 7385 7386 7387 7388 7389 7390 7391 7392 7393 7394 7395 7396 7397 7398 7399 7400 7401 7402 7403 7404 7405 7406 7407 7408 7409 7410 7411 7412 7413 7414 7415 7416 7417 7418 7419 7420 7421 7422 7423 7424 7425 7426 7427 7428 7429 7430 7431 7432 7433 7434 7435 7436 7437 7438 7439 7440 7441 7442 7443 7444 7445 7446 7447 7448 7449 7450 7451 7452 7453 7454 7455 7456 7457 7458 7459
    if (!def)
        goto out;

    cap = def->caps;
    while (cap) {
        if (cap->type == VIR_NODE_DEV_CAP_PCI_DEV) {
            *domain   = cap->data.pci_dev.domain;
            *bus      = cap->data.pci_dev.bus;
            *slot     = cap->data.pci_dev.slot;
            *function = cap->data.pci_dev.function;
            break;
        }

        cap = cap->next;
    }

    if (!cap) {
        qemudReportError(dev->conn, NULL, NULL, VIR_ERR_INVALID_ARG,
                         _("device %s is not a PCI device"), dev->name);
        goto out;
    }

    ret = 0;
out:
    virNodeDeviceDefFree(def);
    VIR_FREE(xml);
    return ret;
}

static int
qemudNodeDeviceDettach (virNodeDevicePtr dev)
{
    pciDevice *pci;
    unsigned domain, bus, slot, function;
    int ret = -1;

    if (qemudNodeDeviceGetPciInfo(dev, &domain, &bus, &slot, &function) < 0)
        return -1;

    pci = pciGetDevice(dev->conn, domain, bus, slot, function);
    if (!pci)
        return -1;

    if (pciDettachDevice(dev->conn, pci) < 0)
        goto out;

    ret = 0;
out:
    pciFreeDevice(dev->conn, pci);
    return ret;
}

static int
qemudNodeDeviceReAttach (virNodeDevicePtr dev)
{
    pciDevice *pci;
    unsigned domain, bus, slot, function;
    int ret = -1;

    if (qemudNodeDeviceGetPciInfo(dev, &domain, &bus, &slot, &function) < 0)
        return -1;

    pci = pciGetDevice(dev->conn, domain, bus, slot, function);
    if (!pci)
        return -1;

    if (pciReAttachDevice(dev->conn, pci) < 0)
        goto out;

    ret = 0;
out:
    pciFreeDevice(dev->conn, pci);
    return ret;
}

static int
qemudNodeDeviceReset (virNodeDevicePtr dev)
{
7460
    struct qemud_driver *driver = dev->conn->privateData;
7461 7462 7463 7464 7465 7466 7467 7468 7469 7470 7471
    pciDevice *pci;
    unsigned domain, bus, slot, function;
    int ret = -1;

    if (qemudNodeDeviceGetPciInfo(dev, &domain, &bus, &slot, &function) < 0)
        return -1;

    pci = pciGetDevice(dev->conn, domain, bus, slot, function);
    if (!pci)
        return -1;

7472 7473 7474
    qemuDriverLock(driver);

    if (pciResetDevice(dev->conn, pci, driver->activePciHostdevs) < 0)
7475 7476 7477 7478
        goto out;

    ret = 0;
out:
7479
    qemuDriverUnlock(driver);
7480 7481 7482 7483
    pciFreeDevice(dev->conn, pci);
    return ret;
}

7484 7485 7486 7487 7488
static virDriver qemuDriver = {
    VIR_DRV_QEMU,
    "QEMU",
    qemudOpen, /* open */
    qemudClose, /* close */
D
Daniel Veillard 已提交
7489
    qemudSupportsFeature, /* supports_feature */
7490 7491
    qemudGetType, /* type */
    qemudGetVersion, /* version */
7492
    NULL, /* libvirtVersion (impl. in libvirt.c) */
7493
    virGetHostname, /* getHostname */
7494
    qemudGetMaxVCPUs, /* getMaxVcpus */
7495
    nodeGetInfo, /* nodeGetInfo */
7496 7497 7498
    qemudGetCapabilities, /* getCapabilities */
    qemudListDomains, /* listDomains */
    qemudNumDomains, /* numOfDomains */
7499
    qemudDomainCreate, /* domainCreateXML */
7500 7501 7502 7503 7504
    qemudDomainLookupByID, /* domainLookupByID */
    qemudDomainLookupByUUID, /* domainLookupByUUID */
    qemudDomainLookupByName, /* domainLookupByName */
    qemudDomainSuspend, /* domainSuspend */
    qemudDomainResume, /* domainResume */
7505
    qemudDomainShutdown, /* domainShutdown */
7506 7507 7508
    NULL, /* domainReboot */
    qemudDomainDestroy, /* domainDestroy */
    qemudDomainGetOSType, /* domainGetOSType */
7509 7510 7511
    qemudDomainGetMaxMemory, /* domainGetMaxMemory */
    qemudDomainSetMaxMemory, /* domainSetMaxMemory */
    qemudDomainSetMemory, /* domainSetMemory */
7512 7513 7514
    qemudDomainGetInfo, /* domainGetInfo */
    qemudDomainSave, /* domainSave */
    qemudDomainRestore, /* domainRestore */
P
Paolo Bonzini 已提交
7515
    qemudDomainCoreDump, /* domainCoreDump */
7516
    qemudDomainSetVcpus, /* domainSetVcpus */
7517 7518
    qemudDomainPinVcpu, /* domainPinVcpu */
    qemudDomainGetVcpus, /* domainGetVcpus */
7519
    qemudDomainGetMaxVcpus, /* domainGetMaxVcpus */
7520 7521
    qemudDomainGetSecurityLabel, /* domainGetSecurityLabel */
    qemudNodeGetSecurityModel, /* nodeGetSecurityModel */
7522
    qemudDomainDumpXML, /* domainDumpXML */
7523
    qemuDomainXMLFromNative, /* domainXmlFromNative */
7524
    qemuDomainXMLToNative, /* domainXMLToNative */
7525 7526
    qemudListDefinedDomains, /* listDefinedDomains */
    qemudNumDefinedDomains, /* numOfDefinedDomains */
7527 7528 7529
    qemudDomainStart, /* domainCreate */
    qemudDomainDefine, /* domainDefineXML */
    qemudDomainUndefine, /* domainUndefine */
7530
    qemudDomainAttachDevice, /* domainAttachDevice */
7531
    qemudDomainDetachDevice, /* domainDetachDevice */
7532 7533
    qemudDomainGetAutostart, /* domainGetAutostart */
    qemudDomainSetAutostart, /* domainSetAutostart */
7534 7535 7536
    qemuGetSchedulerType, /* domainGetSchedulerType */
    qemuGetSchedulerParameters, /* domainGetSchedulerParameters */
    qemuSetSchedulerParameters, /* domainSetSchedulerParameters */
D
Daniel Veillard 已提交
7537 7538
    NULL, /* domainMigratePrepare (v1) */
    qemudDomainMigratePerform, /* domainMigratePerform */
7539
    NULL, /* domainMigrateFinish */
7540
    qemudDomainBlockStats, /* domainBlockStats */
7541
    qemudDomainInterfaceStats, /* domainInterfaceStats */
7542
    qemudDomainBlockPeek, /* domainBlockPeek */
R
Richard W.M. Jones 已提交
7543
    qemudDomainMemoryPeek, /* domainMemoryPeek */
7544 7545
    nodeGetCellsFreeMemory, /* nodeGetCellsFreeMemory */
    nodeGetFreeMemory,  /* getFreeMemory */
7546 7547
    qemudDomainEventRegister, /* domainEventRegister */
    qemudDomainEventDeregister, /* domainEventDeregister */
D
Daniel Veillard 已提交
7548 7549
    qemudDomainMigratePrepare2, /* domainMigratePrepare2 */
    qemudDomainMigrateFinish2, /* domainMigrateFinish2 */
7550 7551 7552
    qemudNodeDeviceDettach, /* nodeDeviceDettach */
    qemudNodeDeviceReAttach, /* nodeDeviceReAttach */
    qemudNodeDeviceReset, /* nodeDeviceReset */
C
Chris Lalancette 已提交
7553
    qemudDomainMigratePrepareTunnel, /* domainMigratePrepareTunnel */
7554 7555 7556 7557
    qemuIsEncrypted,
    qemuIsSecure,
    qemuDomainIsActive,
    qemuDomainIsPersistent,
7558 7559 7560
};


7561
static virStateDriver qemuStateDriver = {
7562
    .name = "QEMU",
7563 7564 7565 7566
    .initialize = qemudStartup,
    .cleanup = qemudShutdown,
    .reload = qemudReload,
    .active = qemudActive,
7567
};
7568

7569
int qemuRegister(void) {
7570 7571 7572 7573
    virRegisterDriver(&qemuDriver);
    virRegisterStateDriver(&qemuStateDriver);
    return 0;
}