bridge_driver.c 154.3 KB
Newer Older
1
/*
2
 * bridge_driver.c: core driver methods for managing network
3
 *
4
 * Copyright (C) 2006-2014 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16 17
 * 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
18
 * License along with this library.  If not, see
O
Osier Yang 已提交
19
 * <http://www.gnu.org/licenses/>.
20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42
 *
 * Author: Daniel P. Berrange <berrange@redhat.com>
 */

#include <config.h>

#include <sys/types.h>
#include <sys/poll.h>
#include <limits.h>
#include <string.h>
#include <stdio.h>
#include <stdarg.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
#include <sys/utsname.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <signal.h>
#include <paths.h>
#include <pwd.h>
#include <sys/wait.h>
#include <sys/ioctl.h>
43
#include <net/if.h>
44
#include <dirent.h>
45 46 47
#if HAVE_SYS_SYSCTL_H
# include <sys/sysctl.h>
#endif
48

49
#include "virerror.h"
50
#include "datatypes.h"
51
#include "bridge_driver.h"
52
#include "bridge_driver_platform.h"
53
#include "network_conf.h"
54
#include "device_conf.h"
55
#include "driver.h"
56
#include "virbuffer.h"
57
#include "virpidfile.h"
58
#include "vircommand.h"
59
#include "viralloc.h"
60
#include "viruuid.h"
61
#include "viriptables.h"
62
#include "virlog.h"
63
#include "virdnsmasq.h"
64
#include "configmake.h"
65
#include "virnetdev.h"
66
#include "virpci.h"
67 68
#include "virnetdevbridge.h"
#include "virnetdevtap.h"
69
#include "virnetdevvportprofile.h"
70
#include "virdbus.h"
71
#include "virfile.h"
72
#include "virstring.h"
73
#include "viraccessapicheck.h"
74
#include "network_event.h"
75
#include "virhook.h"
76
#include "virjson.h"
77

78 79
#define VIR_FROM_THIS VIR_FROM_NETWORK

80 81 82 83 84 85 86
/**
 * VIR_NETWORK_DHCP_LEASE_FILE_SIZE_MAX:
 *
 * Macro providing the upper limit on the size of leases file
 */
#define VIR_NETWORK_DHCP_LEASE_FILE_SIZE_MAX (32 * 1024 * 1024)

87 88
VIR_LOG_INIT("network.bridge_driver");

89 90 91 92
static virNetworkDriverStatePtr driver;


static void networkDriverLock(void)
93
{
94
    virMutexLock(&driver->lock);
95
}
96
static void networkDriverUnlock(void)
97
{
98
    virMutexUnlock(&driver->lock);
99 100
}

101
static int networkStateCleanup(void);
102

103
static int networkStartNetwork(virNetworkObjPtr network);
104

105
static int networkShutdownNetwork(virNetworkObjPtr network);
106

107
static int networkStartNetworkVirtual(virNetworkObjPtr network);
108

109
static int networkShutdownNetworkVirtual(virNetworkObjPtr network);
110

111
static int networkStartNetworkExternal(virNetworkObjPtr network);
112

113
static int networkShutdownNetworkExternal(virNetworkObjPtr network);
114

115 116
static void networkReloadFirewallRules(void);
static void networkRefreshDaemons(void);
117

118 119 120 121 122
static int networkPlugBandwidth(virNetworkObjPtr net,
                                virDomainNetDefPtr iface);
static int networkUnplugBandwidth(virNetworkObjPtr net,
                                  virDomainNetDefPtr iface);

123
static void networkNetworkObjTaint(virNetworkObjPtr net,
124
                                   virNetworkTaintFlags taint);
125

126 127 128 129 130 131
static virNetworkObjPtr
networkObjFromNetwork(virNetworkPtr net)
{
    virNetworkObjPtr network;
    char uuidstr[VIR_UUID_STRING_BUFLEN];

132
    networkDriverLock();
133
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
134
    networkDriverUnlock();
135 136 137 138 139 140 141 142 143 144 145

    if (!network) {
        virUUIDFormat(net->uuid, uuidstr);
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching uuid '%s' (%s)"),
                       uuidstr, net->name);
    }

    return network;
}

146 147 148
static int
networkRunHook(virNetworkObjPtr network,
               virDomainDefPtr dom,
149
               virDomainNetDefPtr iface,
150 151 152 153 154 155 156 157 158
               int op,
               int sub_op)
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;
    char *xml = NULL, *net_xml = NULL, *dom_xml = NULL;
    int hookret;
    int ret = -1;

    if (virHookPresent(VIR_HOOK_DRIVER_NETWORK)) {
159 160 161 162 163 164
        if (!network) {
            VIR_DEBUG("Not running hook as @network is NULL");
            ret = 0;
            goto cleanup;
        }

165 166
        virBufferAddLit(&buf, "<hookData>\n");
        virBufferAdjustIndent(&buf, 2);
167 168
        if (iface && virDomainNetDefFormat(&buf, iface, 0) < 0)
            goto cleanup;
169 170 171 172 173 174 175 176
        if (virNetworkDefFormatBuf(&buf, network->def, 0) < 0)
            goto cleanup;
        if (dom && virDomainDefFormatInternal(dom, 0, &buf) < 0)
            goto cleanup;

        virBufferAdjustIndent(&buf, -2);
        virBufferAddLit(&buf, "</hookData>");

177
        if (virBufferCheckError(&buf) < 0)
178 179
            goto cleanup;

180
        xml = virBufferContentAndReset(&buf);
181 182 183 184 185 186 187 188
        hookret = virHookCall(VIR_HOOK_DRIVER_NETWORK, network->def->name,
                              op, sub_op, NULL, xml, NULL);

        /*
         * If the script raised an error, pass it to the callee.
         */
        if (hookret < 0)
            goto cleanup;
189 190

        networkNetworkObjTaint(network, VIR_NETWORK_TAINT_HOOK);
191 192 193
    }

    ret = 0;
194
 cleanup:
195 196 197 198 199 200 201
    virBufferFreeAndReset(&buf);
    VIR_FREE(xml);
    VIR_FREE(net_xml);
    VIR_FREE(dom_xml);
    return ret;
}

202
static char *
203
networkDnsmasqLeaseFileNameDefault(const char *netname)
204 205 206
{
    char *leasefile;

207
    ignore_value(virAsprintf(&leasefile, "%s/%s.leases",
208
                             driver->dnsmasqStateDir, netname));
209 210 211
    return leasefile;
}

212 213 214
networkDnsmasqLeaseFileNameFunc networkDnsmasqLeaseFileName =
    networkDnsmasqLeaseFileNameDefault;

215 216 217 218 219 220
static char *
networkDnsmasqLeaseFileNameCustom(const char *bridge)
{
    char *leasefile;

    ignore_value(virAsprintf(&leasefile, "%s/%s.status",
221
                             driver->dnsmasqStateDir, bridge));
222 223 224
    return leasefile;
}

225 226 227 228 229
static char *
networkDnsmasqConfigFileName(const char *netname)
{
    char *conffile;

230
    ignore_value(virAsprintf(&conffile, "%s/%s.conf",
231
                             driver->dnsmasqStateDir, netname));
232 233 234
    return conffile;
}

235 236 237 238 239 240
static char *
networkRadvdPidfileBasename(const char *netname)
{
    /* this is simple but we want to be sure it's consistently done */
    char *pidfilebase;

241
    ignore_value(virAsprintf(&pidfilebase, "%s-radvd", netname));
242 243 244 245 246 247 248 249
    return pidfilebase;
}

static char *
networkRadvdConfigFileName(const char *netname)
{
    char *configfile;

250
    ignore_value(virAsprintf(&configfile, "%s/%s-radvd.conf",
251
                             driver->radvdStateDir, netname));
252 253
    return configfile;
}
254

255 256
/* do needed cleanup steps and remove the network from the list */
static int
257
networkRemoveInactive(virNetworkObjPtr net)
258 259
{
    char *leasefile = NULL;
260
    char *customleasefile = NULL;
261
    char *radvdconfigfile = NULL;
262
    char *configfile = NULL;
263
    char *radvdpidbase = NULL;
264
    char *statusfile = NULL;
265 266 267 268 269 270
    dnsmasqContext *dctx = NULL;
    virNetworkDefPtr def = virNetworkObjGetPersistentDef(net);

    int ret = -1;

    /* remove the (possibly) existing dnsmasq and radvd files */
271
    if (!(dctx = dnsmasqContextNew(def->name,
272
                                   driver->dnsmasqStateDir))) {
273
        goto cleanup;
274
    }
275 276 277 278

    if (!(leasefile = networkDnsmasqLeaseFileName(def->name)))
        goto cleanup;

279 280 281
    if (!(customleasefile = networkDnsmasqLeaseFileNameCustom(def->bridge)))
        goto cleanup;

282
    if (!(radvdconfigfile = networkRadvdConfigFileName(def->name)))
283
        goto cleanup;
284 285

    if (!(radvdpidbase = networkRadvdPidfileBasename(def->name)))
286
        goto cleanup;
287

288
    if (!(configfile = networkDnsmasqConfigFileName(def->name)))
289
        goto cleanup;
290

291
    if (!(statusfile
292
          = virNetworkConfigFile(driver->stateDir, def->name)))
293
        goto cleanup;
294

295 296 297
    /* dnsmasq */
    dnsmasqDelete(dctx);
    unlink(leasefile);
298
    unlink(customleasefile);
299
    unlink(configfile);
300 301 302

    /* radvd */
    unlink(radvdconfigfile);
303
    virPidFileDelete(driver->pidDir, radvdpidbase);
304

305 306 307
    /* remove status file */
    unlink(statusfile);

308 309 310 311 312
    /* remove the network definition */
    virNetworkRemoveInactive(&driver->networks, net);

    ret = 0;

313
 cleanup:
314
    VIR_FREE(leasefile);
315
    VIR_FREE(configfile);
316
    VIR_FREE(customleasefile);
317 318
    VIR_FREE(radvdconfigfile);
    VIR_FREE(radvdpidbase);
319
    VIR_FREE(statusfile);
320 321 322 323
    dnsmasqContextFree(dctx);
    return ret;
}

324 325 326
static char *
networkBridgeDummyNicName(const char *brname)
{
327
    static const char dummyNicSuffix[] = "-nic";
328 329
    char *nicname;

330 331 332 333 334 335 336
    if (strlen(brname) + sizeof(dummyNicSuffix) > IFNAMSIZ) {
        /* because the length of an ifname is limited to IFNAMSIZ-1
         * (usually 15), and we're adding 4 more characters, we must
         * truncate the original name to 11 to fit. In order to catch
         * a possible numeric ending (eg virbr0, virbr1, etc), we grab
         * the first 8 and last 3 characters of the string.
         */
337 338 339 340 341
        ignore_value(virAsprintf(&nicname, "%.*s%s%s",
                                 /* space for last 3 chars + "-nic" + NULL */
                                 (int)(IFNAMSIZ - (3 + sizeof(dummyNicSuffix))),
                                 brname, brname + strlen(brname) - 3,
                                 dummyNicSuffix));
342
    } else {
343
        ignore_value(virAsprintf(&nicname, "%s%s", brname, dummyNicSuffix));
344
    }
345 346 347
    return nicname;
}

348 349 350
/* Update the internal status of all allegedly active networks
 * according to external conditions on the host (i.e. anything that
 * isn't stored directly in each network's state file). */
351
static void
352
networkUpdateAllState(void)
353
{
354
    size_t i;
355

356
    for (i = 0; i < driver->networks.count; i++) {
357 358
        virNetworkObjPtr obj = driver->networks.objs[i];

359
        if (!obj->active)
J
Ján Tomko 已提交
360
            continue;
361

362 363
        virNetworkObjLock(obj);

364 365 366 367 368 369 370 371
        switch (obj->def->forward.type) {
        case VIR_NETWORK_FORWARD_NONE:
        case VIR_NETWORK_FORWARD_NAT:
        case VIR_NETWORK_FORWARD_ROUTE:
            /* If bridge doesn't exist, then mark it inactive */
            if (!(obj->def->bridge && virNetDevExists(obj->def->bridge) == 1))
                obj->active = 0;
            break;
372

373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
        case VIR_NETWORK_FORWARD_BRIDGE:
            if (obj->def->bridge) {
                if (virNetDevExists(obj->def->bridge) != 1)
                    obj->active = 0;
                break;
            }
            /* intentionally drop through to common case for all
             * macvtap networks (forward='bridge' with no bridge
             * device defined is macvtap using its 'bridge' mode)
             */
        case VIR_NETWORK_FORWARD_PRIVATE:
        case VIR_NETWORK_FORWARD_VEPA:
        case VIR_NETWORK_FORWARD_PASSTHROUGH:
            /* so far no extra checks */
            break;
388

389 390 391 392
        case VIR_NETWORK_FORWARD_HOSTDEV:
            /* so far no extra checks */
            break;
        }
393

394 395 396 397
        /* Try and read dnsmasq/radvd pids of active networks */
        if (obj->active && obj->def->ips && (obj->def->nips > 0)) {
            char *radvdpidbase;

398
            ignore_value(virPidFileReadIfAlive(driver->pidDir,
399 400 401 402 403 404
                                               obj->def->name,
                                               &obj->dnsmasqPid,
                                               dnsmasqCapsGetBinaryPath(driver->dnsmasqCaps)));
            radvdpidbase = networkRadvdPidfileBasename(obj->def->name);
            if (!radvdpidbase)
                break;
405
            ignore_value(virPidFileReadIfAlive(driver->pidDir,
406 407 408
                                               radvdpidbase,
                                               &obj->radvdPid, RADVD));
            VIR_FREE(radvdpidbase);
409 410 411 412
        }

        virNetworkObjUnlock(obj);
    }
413 414 415 416 417 418 419 420

    /* remove inactive transient networks */
    i = 0;
    while (i < driver->networks.count) {
        virNetworkObjPtr obj = driver->networks.objs[i];
        virNetworkObjLock(obj);

        if (!obj->persistent && !obj->active) {
421
            networkRemoveInactive(obj);
422 423 424 425 426 427
            continue;
        }

        virNetworkObjUnlock(obj);
        i++;
    }
428 429 430
}


431
static void
432
networkAutostartConfigs(void)
433
{
434
    size_t i;
435

436
    for (i = 0; i < driver->networks.count; i++) {
437
        virNetworkObjLock(driver->networks.objs[i]);
438
        if (driver->networks.objs[i]->autostart &&
439
            !virNetworkObjIsActive(driver->networks.objs[i])) {
440
            if (networkStartNetwork(driver->networks.objs[i]) < 0) {
J
Ján Tomko 已提交
441
                /* failed to start but already logged */
442
            }
443
        }
444
        virNetworkObjUnlock(driver->networks.objs[i]);
445 446 447
    }
}

448 449 450
#if HAVE_FIREWALLD
static DBusHandlerResult
firewalld_dbus_filter_bridge(DBusConnection *connection ATTRIBUTE_UNUSED,
451
                             DBusMessage *message, void *user_data ATTRIBUTE_UNUSED)
452
{
453 454 455 456 457 458
    if (dbus_message_is_signal(message, DBUS_INTERFACE_DBUS,
                               "NameOwnerChanged") ||
        dbus_message_is_signal(message, "org.fedoraproject.FirewallD1",
                               "Reloaded"))
    {
        VIR_DEBUG("Reload in bridge_driver because of firewalld.");
459
        networkReloadFirewallRules();
460 461 462 463 464 465
    }

    return DBUS_HANDLER_RESULT_NOT_YET_HANDLED;
}
#endif

466
static int
467
networkMigrateStateFiles(void)
468 469 470 471 472 473 474 475 476 477 478 479 480
{
    /* Due to a change in location of network state xml beginning in
     * libvirt 1.2.4 (from /var/lib/libvirt/network to
     * /var/run/libvirt/network), we must check for state files in two
     * locations. Anything found in the old location must be written
     * to the new location, then erased from the old location. (Note
     * that we read/write the file rather than calling rename()
     * because the old and new state directories are likely in
     * different filesystems).
     */
    int ret = -1;
    const char *oldStateDir = LOCALSTATEDIR "/lib/libvirt/network";
    DIR *dir;
481
    int direrr;
482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
    struct dirent *entry;
    char *oldPath = NULL, *newPath = NULL;
    char *contents = NULL;

    if (!(dir = opendir(oldStateDir))) {
        if (errno == ENOENT)
            return 0;

        virReportSystemError(errno, _("failed to open directory '%s'"),
                             oldStateDir);
        return -1;
    }

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

501
    while ((direrr = virDirRead(dir, &entry, oldStateDir)) > 0) {
502 503 504 505 506 507 508 509

        if (entry->d_type != DT_REG ||
            STREQ(entry->d_name, ".") ||
            STREQ(entry->d_name, ".."))
            continue;

        if (virAsprintf(&oldPath, "%s/%s",
                        oldStateDir, entry->d_name) < 0)
J
Ján Tomko 已提交
510
            goto cleanup;
511
        if (virFileReadAll(oldPath, 1024*1024, &contents) < 0)
J
Ján Tomko 已提交
512
            goto cleanup;
513 514 515

        if (virAsprintf(&newPath, "%s/%s",
                        driver->stateDir, entry->d_name) < 0)
J
Ján Tomko 已提交
516
            goto cleanup;
517 518 519 520 521 522 523 524 525 526 527 528
        if (virFileWriteStr(newPath, contents, S_IRUSR | S_IWUSR) < 0) {
            virReportSystemError(errno,
                                 _("failed to write network status file '%s'"),
                                 newPath);
            goto cleanup;
        }

        unlink(oldPath);
        VIR_FREE(oldPath);
        VIR_FREE(newPath);
        VIR_FREE(contents);
    }
529
    if (direrr < 0)
J
Ján Tomko 已提交
530
        goto cleanup;
531 532 533 534 535 536 537 538 539 540

    ret = 0;
 cleanup:
    closedir(dir);
    VIR_FREE(oldPath);
    VIR_FREE(newPath);
    VIR_FREE(contents);
    return ret;
}

541
/**
542
 * networkStateInitialize:
543 544 545 546
 *
 * Initialization function for the QEmu daemon
 */
static int
547 548 549
networkStateInitialize(bool privileged,
                       virStateInhibitCallback callback ATTRIBUTE_UNUSED,
                       void *opaque ATTRIBUTE_UNUSED)
550
{
551 552 553
    int ret = -1;
    char *configdir = NULL;
    char *rundir = NULL;
554 555 556
#ifdef HAVE_FIREWALLD
    DBusConnection *sysbus = NULL;
#endif
557

558
    if (VIR_ALLOC(driver) < 0)
559
        goto error;
560

561 562
    if (virMutexInit(&driver->lock) < 0) {
        VIR_FREE(driver);
563 564
        goto error;
    }
565
    networkDriverLock();
566

567 568 569 570
    /* configuration/state paths are one of
     * ~/.config/libvirt/... (session/unprivileged)
     * /etc/libvirt/... && /var/(run|lib)/libvirt/... (system/privileged).
     */
571
    if (privileged) {
572
        if (VIR_STRDUP(driver->networkConfigDir,
573
                       SYSCONFDIR "/libvirt/qemu/networks") < 0 ||
574
            VIR_STRDUP(driver->networkAutostartDir,
575
                       SYSCONFDIR "/libvirt/qemu/networks/autostart") < 0 ||
576
            VIR_STRDUP(driver->stateDir,
577
                       LOCALSTATEDIR "/run/libvirt/network") < 0 ||
578
            VIR_STRDUP(driver->pidDir,
579
                       LOCALSTATEDIR "/run/libvirt/network") < 0 ||
580
            VIR_STRDUP(driver->dnsmasqStateDir,
581
                       LOCALSTATEDIR "/lib/libvirt/dnsmasq") < 0 ||
582
            VIR_STRDUP(driver->radvdStateDir,
583 584
                       LOCALSTATEDIR "/lib/libvirt/radvd") < 0)
            goto error;
585 586 587 588 589

        /* migration from old to new location is only applicable for
         * privileged mode - unprivileged mode directories haven't
         * changed location.
         */
590
        if (networkMigrateStateFiles() < 0)
591
            goto error;
592
    } else {
593 594 595
        configdir = virGetUserConfigDirectory();
        rundir = virGetUserRuntimeDirectory();
        if (!(configdir && rundir))
596
            goto error;
597

598
        if ((virAsprintf(&driver->networkConfigDir,
599
                         "%s/qemu/networks", configdir) < 0) ||
600
            (virAsprintf(&driver->networkAutostartDir,
601
                         "%s/qemu/networks/autostart", configdir) < 0) ||
602
            (virAsprintf(&driver->stateDir,
603
                         "%s/network/lib", rundir) < 0) ||
604
            (virAsprintf(&driver->pidDir,
605
                         "%s/network/run", rundir) < 0) ||
606
            (virAsprintf(&driver->dnsmasqStateDir,
607
                         "%s/dnsmasq/lib", rundir) < 0) ||
608
            (virAsprintf(&driver->radvdStateDir,
609
                         "%s/radvd/lib", rundir) < 0)) {
610
            goto error;
611
        }
612 613
    }

614
    if (virFileMakePath(driver->stateDir) < 0) {
615 616
        virReportSystemError(errno,
                             _("cannot create directory %s"),
617
                             driver->stateDir);
618 619 620
        goto error;
    }

621
    /* if this fails now, it will be retried later with dnsmasqCapsRefresh() */
622
    driver->dnsmasqCaps = dnsmasqCapsNewFromBinary(DNSMASQ);
623

624 625
    if (virNetworkLoadAllState(&driver->networks,
                               driver->stateDir) < 0)
626 627
        goto error;

628 629 630
    if (virNetworkLoadAllConfigs(&driver->networks,
                                 driver->networkConfigDir,
                                 driver->networkAutostartDir) < 0)
631 632
        goto error;

633 634 635
    networkUpdateAllState();
    networkReloadFirewallRules();
    networkRefreshDaemons();
636

637
    driver->networkEventState = virObjectEventStateNew();
638

639
    networkDriverUnlock();
640

641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662
#ifdef HAVE_FIREWALLD
    if (!(sysbus = virDBusGetSystemBus())) {
        virErrorPtr err = virGetLastError();
        VIR_WARN("DBus not available, disabling firewalld support "
                 "in bridge_driver: %s", err->message);
    } else {
        /* add matches for
         * NameOwnerChanged on org.freedesktop.DBus for firewalld start/stop
         * Reloaded on org.fedoraproject.FirewallD1 for firewalld reload
         */
        dbus_bus_add_match(sysbus,
                           "type='signal'"
                           ",interface='"DBUS_INTERFACE_DBUS"'"
                           ",member='NameOwnerChanged'"
                           ",arg0='org.fedoraproject.FirewallD1'",
                           NULL);
        dbus_bus_add_match(sysbus,
                           "type='signal'"
                           ",interface='org.fedoraproject.FirewallD1'"
                           ",member='Reloaded'",
                           NULL);
        dbus_connection_add_filter(sysbus, firewalld_dbus_filter_bridge,
663
                                   NULL, NULL);
664 665 666
    }
#endif

667
    ret = 0;
668
 cleanup:
669 670 671
    VIR_FREE(configdir);
    VIR_FREE(rundir);
    return ret;
672

673
 error:
674 675
    if (driver)
        networkDriverUnlock();
676
    networkStateCleanup();
677
    goto cleanup;
678 679
}

680 681 682 683 684 685 686 687
/**
 * networkStateAutoStart:
 *
 * Function to AutoStart the bridge configs
 */
static void
networkStateAutoStart(void)
{
688
    if (!driver)
689 690
        return;

691 692 693
    networkDriverLock();
    networkAutostartConfigs();
    networkDriverUnlock();
694 695
}

696
/**
697
 * networkStateReload:
698 699 700 701 702
 *
 * Function to restart the QEmu daemon, it will recheck the configuration
 * files and update its state and the networking
 */
static int
703 704
networkStateReload(void)
{
705
    if (!driver)
706 707
        return 0;

708 709 710 711 712 713 714 715 716 717
    networkDriverLock();
    virNetworkLoadAllState(&driver->networks,
                           driver->stateDir);
    virNetworkLoadAllConfigs(&driver->networks,
                             driver->networkConfigDir,
                             driver->networkAutostartDir);
    networkReloadFirewallRules();
    networkRefreshDaemons();
    networkAutostartConfigs();
    networkDriverUnlock();
718 719 720 721 722
    return 0;
}


/**
723
 * networkStateCleanup:
724 725 726 727
 *
 * Shutdown the QEmu daemon, it will stop all active domains and networks
 */
static int
728 729
networkStateCleanup(void)
{
730
    if (!driver)
731 732
        return -1;

733
    networkDriverLock();
734

735
    virObjectEventStateFree(driver->networkEventState);
736

737
    /* free inactive networks */
738
    virNetworkObjListFree(&driver->networks);
739

740 741 742 743 744 745
    VIR_FREE(driver->networkConfigDir);
    VIR_FREE(driver->networkAutostartDir);
    VIR_FREE(driver->stateDir);
    VIR_FREE(driver->pidDir);
    VIR_FREE(driver->dnsmasqStateDir);
    VIR_FREE(driver->radvdStateDir);
746

747
    virObjectUnref(driver->dnsmasqCaps);
748

749 750
    networkDriverUnlock();
    virMutexDestroy(&driver->lock);
751

752
    VIR_FREE(driver);
753 754 755 756 757

    return 0;
}


758 759 760 761 762 763 764
/* networkKillDaemon:
 *
 * kill the specified pid/name, and wait a bit to make sure it's dead.
 */
static int
networkKillDaemon(pid_t pid, const char *daemonName, const char *networkName)
{
765 766
    size_t i;
    int ret = -1;
767 768 769 770 771 772 773
    const char *signame = "TERM";

    /* send SIGTERM, then wait up to 3 seconds for the process to
     * disappear, send SIGKILL, then wait for up to another 2
     * seconds. If that fails, log a warning and continue, hoping
     * for the best.
     */
774
    for (i = 0; i < 25; i++) {
775
        int signum = 0;
776
        if (i == 0) {
777
            signum = SIGTERM;
778
        } else if (i == 15) {
779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813
            signum = SIGKILL;
            signame = "KILL";
        }
        if (kill(pid, signum) < 0) {
            if (errno == ESRCH) {
                ret = 0;
            } else {
                char ebuf[1024];
                VIR_WARN("Failed to terminate %s process %d "
                         "for network '%s' with SIG%s: %s",
                         daemonName, pid, networkName, signame,
                         virStrerror(errno, ebuf, sizeof(ebuf)));
            }
            goto cleanup;
        }
        /* NB: since networks have no reference count like
         * domains, there is no safe way to unlock the network
         * object temporarily, and so we can't follow the
         * procedure used by the qemu driver of 1) unlock driver
         * 2) sleep, 3) add ref to object 4) unlock object, 5)
         * re-lock driver, 6) re-lock object. We may need to add
         * that functionality eventually, but for now this
         * function is rarely used and, at worst, leaving the
         * network driver locked during this loop of sleeps will
         * have the effect of holding up any other thread trying
         * to make modifications to a network for up to 5 seconds;
         * since modifications to networks are much less common
         * than modifications to domains, this seems a reasonable
         * tradeoff in exchange for less code disruption.
         */
        usleep(20 * 1000);
    }
    VIR_WARN("Timed out waiting after SIG%s to %s process %d "
             "(network '%s')",
             signame, daemonName, pid, networkName);
814
 cleanup:
815 816 817
    return ret;
}

J
Ján Tomko 已提交
818 819 820
/* the following does not build a file, it builds a list
 * which is later saved into a file
 */
G
Gene Czarcinski 已提交
821

822
static int
G
Gene Czarcinski 已提交
823 824
networkBuildDnsmasqDhcpHostsList(dnsmasqContext *dctx,
                                 virNetworkIpDefPtr ipdef)
825
{
826
    size_t i;
G
Gene Czarcinski 已提交
827
    bool ipv6 = false;
828

G
Gene Czarcinski 已提交
829 830
    if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
        ipv6 = true;
831 832
    for (i = 0; i < ipdef->nhosts; i++) {
        virNetworkDHCPHostDefPtr host = &(ipdef->hosts[i]);
G
Gene Czarcinski 已提交
833
        if (VIR_SOCKET_ADDR_VALID(&host->ip))
834 835
            if (dnsmasqAddDhcpHost(dctx, host->mac, &host->ip,
                                   host->name, host->id, ipv6) < 0)
836
                return -1;
837
    }
838

G
Gene Czarcinski 已提交
839 840 841 842 843 844 845
    return 0;
}

static int
networkBuildDnsmasqHostsList(dnsmasqContext *dctx,
                             virNetworkDNSDefPtr dnsdef)
{
846
    size_t i, j;
G
Gene Czarcinski 已提交
847

848 849
    if (dnsdef) {
        for (i = 0; i < dnsdef->nhosts; i++) {
850
            virNetworkDNSHostDefPtr host = &(dnsdef->hosts[i]);
851
            if (VIR_SOCKET_ADDR_VALID(&host->ip)) {
852
                for (j = 0; j < host->nnames; j++)
853 854
                    if (dnsmasqAddHost(dctx, &host->ip, host->names[j]) < 0)
                        return -1;
855 856
            }
        }
857 858
    }

859
    return 0;
860 861 862
}


863 864
int
networkDnsmasqConfContents(virNetworkObjPtr network,
865 866 867 868
                           const char *pidfile,
                           char **configstr,
                           dnsmasqContext *dctx,
                           dnsmasqCapsPtr caps ATTRIBUTE_UNUSED)
869
{
870
    virBuffer configbuf = VIR_BUFFER_INITIALIZER;
871
    int r, ret = -1;
872
    int nbleases = 0;
873
    size_t i;
874
    virNetworkDNSDefPtr dns = &network->def->dns;
G
Gene Czarcinski 已提交
875 876
    virNetworkIpDefPtr tmpipdef, ipdef, ipv4def, ipv6def;
    bool ipv6SLAAC;
877

878 879
    *configstr = NULL;

880
    /*
881 882 883
     * All dnsmasq parameters are put into a configuration file, except the
     * command line --conf-file=parameter which specifies the location of
     * configuration file.
884
     *
885 886
     * All dnsmasq conf-file parameters must be specified as "foo=bar"
     * as oppose to "--foo bar" which was acceptable on the command line.
887
     */
888 889 890 891 892 893

    /*
     * Needed to ensure dnsmasq uses same algorithm for processing
     * multiple namedriver entries in /etc/resolv.conf as GLibC.
     */

894 895
    /* create dnsmasq config file appropriate for this network */
    virBufferAsprintf(&configbuf,
896 897 898 899 900 901 902
                      "##WARNING:  THIS IS AN AUTO-GENERATED FILE. "
                      "CHANGES TO IT ARE LIKELY TO BE\n"
                      "##OVERWRITTEN AND LOST.  Changes to this "
                      "configuration should be made using:\n"
                      "##    virsh net-edit %s\n"
                      "## or other application using the libvirt API.\n"
                      "##\n## dnsmasq conf file created by libvirt\n"
903
                      "strict-order\n",
904 905
                      network->def->name);

906 907 908 909
    if (network->def->dns.forwarders) {
        virBufferAddLit(&configbuf, "no-resolv\n");
        for (i = 0; i < network->def->dns.nfwds; i++) {
            virBufferAsprintf(&configbuf, "server=%s\n",
J
Ján Tomko 已提交
910
                              network->def->dns.forwarders[i]);
911 912 913
        }
    }

914
    if (network->def->domain) {
915
        virBufferAsprintf(&configbuf,
916 917 918 919
                          "domain=%s\n"
                          "expand-hosts\n",
                          network->def->domain);
    }
920

J
Ján Tomko 已提交
921
    if (network->def->dns.forwardPlainNames == VIR_TRISTATE_BOOL_NO) {
922 923 924 925
        virBufferAddLit(&configbuf, "domain-needed\n");
        /* need to specify local=// whether or not a domain is
         * specified, unless the config says we should forward "plain"
         * names (i.e. not fully qualified, no '.' characters)
926
         */
927
        virBufferAddLit(&configbuf, "local=//\n");
928
    }
929

930
    if (pidfile)
931
        virBufferAsprintf(&configbuf, "pid-file=%s\n", pidfile);
932

933 934 935
    /* dnsmasq will *always* listen on localhost unless told otherwise */
    virBufferAddLit(&configbuf, "except-interface=lo\n");

936 937 938 939 940 941 942 943
    if (dnsmasqCapsGet(caps, DNSMASQ_CAPS_BIND_DYNAMIC)) {
        /* using --bind-dynamic with only --interface (no
         * --listen-address) prevents dnsmasq from responding to dns
         * queries that arrive on some interface other than our bridge
         * interface (in other words, requests originating somewhere
         * other than one of the virtual guests connected directly to
         * this network). This was added in response to CVE 2012-3411.
         */
944
        virBufferAsprintf(&configbuf,
945 946 947
                          "bind-dynamic\n"
                          "interface=%s\n",
                          network->def->bridge);
948
    } else {
949
        virBufferAddLit(&configbuf, "bind-interfaces\n");
950 951 952 953 954 955 956 957
        /*
         * --interface does not actually work with dnsmasq < 2.47,
         * due to DAD for ipv6 addresses on the interface.
         *
         * virCommandAddArgList(cmd, "--interface", network->def->bridge, NULL);
         *
         * So listen on all defined IPv[46] addresses
         */
958 959 960
        for (i = 0;
             (tmpipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, i));
             i++) {
961 962 963 964
            char *ipaddr = virSocketAddrFormat(&tmpipdef->address);

            if (!ipaddr)
                goto cleanup;
965

966
            /* also part of CVE 2012-3411 - if the host's version of
967
             * dnsmasq doesn't have bind-dynamic, only allow listening on
968 969
             * private/local IP addresses (see RFC1918/RFC3484/RFC4193)
             */
970 971
            if (!dnsmasqCapsGet(caps, DNSMASQ_CAPS_BINDTODEVICE) &&
                !virSocketAddrIsPrivate(&tmpipdef->address)) {
972 973 974 975
                unsigned long version = dnsmasqCapsGetVersion(caps);

                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("Publicly routable address %s is prohibited. "
976
                                 "The version of dnsmasq on this host (%d.%d) "
977 978 979 980
                                 "doesn't support the bind-dynamic option or "
                                 "use SO_BINDTODEVICE on listening sockets, "
                                 "one of which is required for safe operation "
                                 "on a publicly routable subnet "
981 982 983 984 985 986
                                 "(see CVE-2012-3411). You must either "
                                 "upgrade dnsmasq, or use a private/local "
                                 "subnet range for this network "
                                 "(as described in RFC1918/RFC3484/RFC4193)."),
                               ipaddr, (int)version / 1000000,
                               (int)(version % 1000000) / 1000);
987
                VIR_FREE(ipaddr);
988 989
                goto cleanup;
            }
990
            virBufferAsprintf(&configbuf, "listen-address=%s\n", ipaddr);
991 992 993
            VIR_FREE(ipaddr);
        }
    }
994

995 996
    /* If this is an isolated network, set the default route option
     * (3) to be empty to avoid setting a default route that's
997
     * guaranteed to not work, and set no-resolv so that no dns
998 999 1000
     * requests are forwarded on to the dns server listed in the
     * host's /etc/resolv.conf (since this could be used as a channel
     * to build a connection to the outside).
1001
     */
1002
    if (network->def->forward.type == VIR_NETWORK_FORWARD_NONE) {
1003
        virBufferAddLit(&configbuf, "dhcp-option=3\n"
1004
                        "no-resolv\n");
1005
    }
1006

1007
    for (i = 0; i < dns->ntxts; i++) {
1008
        virBufferAsprintf(&configbuf, "txt-record=%s,%s\n",
1009 1010
                          dns->txts[i].name,
                          dns->txts[i].value);
1011
    }
1012

1013
    for (i = 0; i < dns->nsrvs; i++) {
1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
        /* service/protocol are required, and should have been validated
         * by the parser.
         */
        if (!dns->srvs[i].service) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Missing required 'service' "
                             "attribute in SRV record of network '%s'"),
                           network->def->name);
            goto cleanup;
        }
        if (!dns->srvs[i].protocol) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Missing required 'service' "
                             "attribute in SRV record of network '%s'"),
                           network->def->name);
            goto cleanup;
        }
        /* RFC2782 requires that service and protocol be preceded by
         * an underscore.
         */
        virBufferAsprintf(&configbuf, "srv-host=_%s._%s",
                          dns->srvs[i].service, dns->srvs[i].protocol);
1036

1037 1038 1039
        /* domain is optional - it defaults to the domain of this network */
        if (dns->srvs[i].domain)
            virBufferAsprintf(&configbuf, ".%s", dns->srvs[i].domain);
1040

1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
        /* If target is empty or ".", that means "the service is
         * decidedly not available at this domain" (RFC2782). In that
         * case, any port, priority, or weight is irrelevant.
         */
        if (dns->srvs[i].target && STRNEQ(dns->srvs[i].target, ".")) {

            virBufferAsprintf(&configbuf, ",%s", dns->srvs[i].target);
            /* port, priority, and weight are optional, but are
             * identified by their position in the line. If an item is
             * unspecified, but something later in the line *is*
             * specified, we need to give the default value for the
             * unspecified item. (According to the dnsmasq manpage,
             * the default for port is 1).
             */
            if (dns->srvs[i].port ||
                dns->srvs[i].priority || dns->srvs[i].weight)
                virBufferAsprintf(&configbuf, ",%d",
                                  dns->srvs[i].port ? dns->srvs[i].port : 1);
            if (dns->srvs[i].priority || dns->srvs[i].weight)
                virBufferAsprintf(&configbuf, ",%d", dns->srvs[i].priority);
            if (dns->srvs[i].weight)
                virBufferAsprintf(&configbuf, ",%d", dns->srvs[i].weight);
1063
        }
1064
        virBufferAddLit(&configbuf, "\n");
1065 1066
    }

G
Gene Czarcinski 已提交
1067
    /* Find the first dhcp for both IPv4 and IPv6 */
1068 1069 1070
    for (i = 0, ipv4def = NULL, ipv6def = NULL, ipv6SLAAC = false;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, i));
         i++) {
G
Gene Czarcinski 已提交
1071 1072 1073 1074
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
            if (ipdef->nranges || ipdef->nhosts) {
                if (ipv4def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
1075 1076
                                   _("For IPv4, multiple DHCP definitions "
                                     "cannot be specified."));
G
Gene Czarcinski 已提交
1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087
                    goto cleanup;
                } else {
                    ipv4def = ipdef;
                }
            }
        }
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6)) {
            if (ipdef->nranges || ipdef->nhosts) {
                if (!DNSMASQ_DHCPv6_SUPPORT(caps)) {
                    unsigned long version = dnsmasqCapsGetVersion(caps);
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
1088 1089 1090 1091 1092 1093 1094 1095 1096
                                   _("The version of dnsmasq on this host "
                                     "(%d.%d) doesn't adequately support "
                                     "IPv6 dhcp range or dhcp host "
                                     "specification. Version %d.%d or later "
                                     "is required."),
                                   (int)version / 1000000,
                                   (int)(version % 1000000) / 1000,
                                   DNSMASQ_DHCPv6_MAJOR_REQD,
                                   DNSMASQ_DHCPv6_MINOR_REQD);
G
Gene Czarcinski 已提交
1097 1098 1099 1100
                    goto cleanup;
                }
                if (ipv6def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
1101 1102
                                   _("For IPv6, multiple DHCP definitions "
                                     "cannot be specified."));
G
Gene Czarcinski 已提交
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115
                    goto cleanup;
                } else {
                    ipv6def = ipdef;
                }
            } else {
                ipv6SLAAC = true;
            }
        }
    }

    if (ipv6def && ipv6SLAAC) {
        VIR_WARN("For IPv6, when DHCP is specified for one address, then "
                 "state-full Router Advertising will occur.  The additional "
1116 1117 1118 1119
                 "IPv6 addresses specified require manually configured guest "
                 "network to work properly since both state-full (DHCP) "
                 "and state-less (SLAAC) addressing are not supported "
                 "on the same network interface.");
G
Gene Czarcinski 已提交
1120 1121 1122 1123 1124
    }

    ipdef = ipv4def ? ipv4def : ipv6def;

    while (ipdef) {
1125
        for (r = 0; r < ipdef->nranges; r++) {
1126
            char *saddr = virSocketAddrFormat(&ipdef->ranges[r].start);
1127 1128
            if (!saddr)
                goto cleanup;
1129
            char *eaddr = virSocketAddrFormat(&ipdef->ranges[r].end);
1130 1131 1132 1133
            if (!eaddr) {
                VIR_FREE(saddr);
                goto cleanup;
            }
1134
            virBufferAsprintf(&configbuf, "dhcp-range=%s,%s\n",
1135
                              saddr, eaddr);
1136
            VIR_FREE(saddr);
1137
            VIR_FREE(eaddr);
1138 1139
            nbleases += virSocketAddrGetRange(&ipdef->ranges[r].start,
                                              &ipdef->ranges[r].end);
1140
        }
1141

1142
        /*
1143 1144 1145 1146
         * For static-only DHCP, i.e. with no range but at least one
         * host element, we have to add a special --dhcp-range option
         * to enable the service in dnsmasq. (this is for dhcp-hosts=
         * support)
1147 1148
         */
        if (!ipdef->nranges && ipdef->nhosts) {
1149
            char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
1150 1151
            if (!bridgeaddr)
                goto cleanup;
1152
            virBufferAsprintf(&configbuf, "dhcp-range=%s,static\n", bridgeaddr);
1153 1154
            VIR_FREE(bridgeaddr);
        }
1155

G
Gene Czarcinski 已提交
1156 1157
        if (networkBuildDnsmasqDhcpHostsList(dctx, ipdef) < 0)
            goto cleanup;
1158

G
Gene Czarcinski 已提交
1159 1160 1161
        /* Note: the following is IPv4 only */
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
            if (ipdef->nranges || ipdef->nhosts)
1162
                virBufferAddLit(&configbuf, "dhcp-no-override\n");
1163

G
Gene Czarcinski 已提交
1164
            if (ipdef->tftproot) {
1165 1166
                virBufferAddLit(&configbuf, "enable-tftp\n");
                virBufferAsprintf(&configbuf, "tftp-root=%s\n", ipdef->tftproot);
G
Gene Czarcinski 已提交
1167
            }
1168

G
Gene Czarcinski 已提交
1169 1170 1171
            if (ipdef->bootfile) {
                if (VIR_SOCKET_ADDR_VALID(&ipdef->bootserver)) {
                    char *bootserver = virSocketAddrFormat(&ipdef->bootserver);
1172

1173
                    if (!bootserver)
G
Gene Czarcinski 已提交
1174
                        goto cleanup;
1175
                    virBufferAsprintf(&configbuf, "dhcp-boot=%s%s%s\n",
1176
                                      ipdef->bootfile, ",,", bootserver);
G
Gene Czarcinski 已提交
1177 1178
                    VIR_FREE(bootserver);
                } else {
1179
                    virBufferAsprintf(&configbuf, "dhcp-boot=%s\n", ipdef->bootfile);
G
Gene Czarcinski 已提交
1180 1181 1182 1183 1184
                }
            }
        }
        ipdef = (ipdef == ipv6def) ? NULL : ipv6def;
    }
1185

G
Gene Czarcinski 已提交
1186 1187
    if (nbleases > 0) {
        char *leasefile = networkDnsmasqLeaseFileName(network->def->name);
1188
        if (!leasefile)
G
Gene Czarcinski 已提交
1189
            goto cleanup;
1190
        virBufferAsprintf(&configbuf, "dhcp-leasefile=%s\n", leasefile);
G
Gene Czarcinski 已提交
1191
        VIR_FREE(leasefile);
1192
        virBufferAsprintf(&configbuf, "dhcp-lease-max=%d\n", nbleases);
G
Gene Czarcinski 已提交
1193
    }
1194

G
Gene Czarcinski 已提交
1195 1196
    /* this is done once per interface */
    if (networkBuildDnsmasqHostsList(dctx, dns) < 0)
1197
        goto cleanup;
G
Gene Czarcinski 已提交
1198 1199 1200 1201 1202 1203

    /* Even if there are currently no static hosts, if we're
     * listening for DHCP, we should write a 0-length hosts
     * file to allow for runtime additions.
     */
    if (ipv4def || ipv6def)
1204 1205
        virBufferAsprintf(&configbuf, "dhcp-hostsfile=%s\n",
                          dctx->hostsfile->path);
G
Gene Czarcinski 已提交
1206

1207 1208
    /* Likewise, always create this file and put it on the
     * commandline, to allow for runtime additions.
G
Gene Czarcinski 已提交
1209
     */
1210
    virBufferAsprintf(&configbuf, "addn-hosts=%s\n",
1211
                      dctx->addnhostsfile->path);
G
Gene Czarcinski 已提交
1212 1213 1214

    /* Are we doing RA instead of radvd? */
    if (DNSMASQ_RA_SUPPORT(caps)) {
1215
        if (ipv6def) {
1216
            virBufferAddLit(&configbuf, "enable-ra\n");
1217
        } else {
1218 1219 1220
            for (i = 0;
                 (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
                 i++) {
G
Gene Czarcinski 已提交
1221 1222 1223 1224
                if (!(ipdef->nranges || ipdef->nhosts)) {
                    char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
                    if (!bridgeaddr)
                        goto cleanup;
1225 1226
                    virBufferAsprintf(&configbuf,
                                      "dhcp-range=%s,ra-only\n", bridgeaddr);
G
Gene Czarcinski 已提交
1227 1228
                    VIR_FREE(bridgeaddr);
                }
1229
            }
1230
        }
1231 1232
    }

1233 1234 1235
    if (!(*configstr = virBufferContentAndReset(&configbuf)))
        goto cleanup;

1236
    ret = 0;
G
Gene Czarcinski 已提交
1237

1238
 cleanup:
1239
    virBufferFreeAndReset(&configbuf);
1240
    return ret;
1241 1242
}

1243
/* build the dnsmasq command line */
1244 1245 1246
static int ATTRIBUTE_NONNULL(2)
networkBuildDhcpDaemonCommandLine(virNetworkObjPtr network,
                                  virCommandPtr *cmdout,
1247 1248
                                  char *pidfile, dnsmasqContext *dctx,
                                  dnsmasqCapsPtr caps)
1249
{
1250
    virCommandPtr cmd = NULL;
G
Gene Czarcinski 已提交
1251
    int ret = -1;
1252 1253
    char *configfile = NULL;
    char *configstr = NULL;
1254
    char *leaseshelper_path = NULL;
1255 1256

    network->dnsmasqPid = -1;
1257

1258 1259 1260 1261 1262 1263
    if (networkDnsmasqConfContents(network, pidfile, &configstr, dctx, caps) < 0)
        goto cleanup;
    if (!configstr)
        goto cleanup;

    /* construct the filename */
1264
    if (!(configfile = networkDnsmasqConfigFileName(network->def->name)))
1265 1266 1267 1268 1269
        goto cleanup;

    /* Write the file */
    if (virFileWriteStr(configfile, configstr, 0600) < 0) {
        virReportSystemError(errno,
J
Ján Tomko 已提交
1270 1271
                             _("couldn't write dnsmasq config file '%s'"),
                             configfile);
1272 1273 1274
        goto cleanup;
    }

1275 1276 1277 1278 1279 1280
    /* This helper is used to create custom leases file for libvirt */
    if (!(leaseshelper_path = virFileFindResource("libvirt_leaseshelper",
                                                  "src",
                                                  LIBEXECDIR)))
        goto cleanup;

1281 1282
    cmd = virCommandNew(dnsmasqCapsGetBinaryPath(caps));
    virCommandAddArgFormat(cmd, "--conf-file=%s", configfile);
1283 1284
    virCommandAddArgFormat(cmd, "--dhcp-script=%s", leaseshelper_path);

1285
    *cmdout = cmd;
1286
    ret = 0;
1287
 cleanup:
1288 1289
    VIR_FREE(configfile);
    VIR_FREE(configstr);
1290
    VIR_FREE(leaseshelper_path);
1291 1292 1293 1294
    return ret;
}

static int
1295
networkStartDhcpDaemon(virNetworkObjPtr network)
1296 1297 1298 1299
{
    virCommandPtr cmd = NULL;
    char *pidfile = NULL;
    int ret = -1;
1300
    dnsmasqContext *dctx = NULL;
1301

1302
    if (!virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, 0)) {
G
Gene Czarcinski 已提交
1303
        /* no IP addresses, so we don't need to run */
1304 1305 1306 1307
        ret = 0;
        goto cleanup;
    }

1308
    if (virFileMakePath(driver->pidDir) < 0) {
1309
        virReportSystemError(errno,
1310
                             _("cannot create directory %s"),
1311
                             driver->pidDir);
1312
        goto cleanup;
1313 1314
    }

1315
    if (!(pidfile = virPidFileBuildPath(driver->pidDir,
1316
                                        network->def->name)))
1317
        goto cleanup;
1318

1319
    if (virFileMakePath(driver->dnsmasqStateDir) < 0) {
1320
        virReportSystemError(errno,
1321
                             _("cannot create directory %s"),
1322
                             driver->dnsmasqStateDir);
1323 1324 1325
        goto cleanup;
    }

1326
    dctx = dnsmasqContextNew(network->def->name, driver->dnsmasqStateDir);
1327 1328 1329
    if (dctx == NULL)
        goto cleanup;

1330 1331
    if (dnsmasqCapsRefresh(&driver->dnsmasqCaps, NULL) < 0)
        goto cleanup;
1332 1333 1334

    ret = networkBuildDhcpDaemonCommandLine(network, &cmd, pidfile,
                                            dctx, driver->dnsmasqCaps);
1335 1336 1337 1338 1339
    if (ret < 0)
        goto cleanup;

    ret = dnsmasqSave(dctx);
    if (ret < 0)
1340
        goto cleanup;
1341

G
Guido Günther 已提交
1342
    ret = virCommandRun(cmd, NULL);
1343
    if (ret < 0)
1344 1345 1346
        goto cleanup;

    /*
1347 1348 1349 1350 1351
     * There really is no race here - when dnsmasq daemonizes, its
     * leader process stays around until its child has actually
     * written its pidfile. So by time virCommandRun exits it has
     * waitpid'd and guaranteed the proess has started and written a
     * pid
1352 1353
     */

1354
    ret = virPidFileRead(driver->pidDir, network->def->name,
1355 1356
                         &network->dnsmasqPid);
    if (ret < 0)
1357
        goto cleanup;
1358

1359
    ret = 0;
1360
 cleanup:
1361
    VIR_FREE(pidfile);
1362
    virCommandFree(cmd);
1363
    dnsmasqContextFree(dctx);
1364 1365 1366
    return ret;
}

1367 1368
/* networkRefreshDhcpDaemon:
 *  Update dnsmasq config files, then send a SIGHUP so that it rereads
G
Gene Czarcinski 已提交
1369 1370
 *  them.   This only works for the dhcp-hostsfile and the
 *  addn-hosts file.
1371 1372 1373
 *
 *  Returns 0 on success, -1 on failure.
 */
1374
static int
1375
networkRefreshDhcpDaemon(virNetworkObjPtr network)
1376
{
1377 1378
    int ret = -1;
    size_t i;
G
Gene Czarcinski 已提交
1379
    virNetworkIpDefPtr ipdef, ipv4def, ipv6def;
1380
    dnsmasqContext *dctx = NULL;
1381

G
Gene Czarcinski 已提交
1382
    /* if no IP addresses specified, nothing to do */
1383
    if (!virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, 0))
G
Gene Czarcinski 已提交
1384 1385
        return 0;

1386 1387
    /* if there's no running dnsmasq, just start it */
    if (network->dnsmasqPid <= 0 || (kill(network->dnsmasqPid, 0) < 0))
1388
        return networkStartDhcpDaemon(network);
1389

G
Gene Czarcinski 已提交
1390
    VIR_INFO("Refreshing dnsmasq for network %s", network->def->bridge);
1391
    if (!(dctx = dnsmasqContextNew(network->def->name,
1392
                                   driver->dnsmasqStateDir))) {
G
Gene Czarcinski 已提交
1393
        goto cleanup;
1394
    }
G
Gene Czarcinski 已提交
1395 1396 1397 1398 1399 1400

    /* Look for first IPv4 address that has dhcp defined.
     * We only support dhcp-host config on one IPv4 subnetwork
     * and on one IPv6 subnetwork.
     */
    ipv4def = NULL;
1401 1402 1403
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, i));
         i++) {
G
Gene Czarcinski 已提交
1404 1405
        if (!ipv4def && (ipdef->nranges || ipdef->nhosts))
            ipv4def = ipdef;
1406 1407
    }

G
Gene Czarcinski 已提交
1408
    ipv6def = NULL;
1409 1410 1411
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
         i++) {
G
Gene Czarcinski 已提交
1412 1413
        if (!ipv6def && (ipdef->nranges || ipdef->nhosts))
            ipv6def = ipdef;
1414 1415
    }

G
Gene Czarcinski 已提交
1416
    if (ipv4def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv4def) < 0))
J
Ján Tomko 已提交
1417
        goto cleanup;
G
Gene Czarcinski 已提交
1418 1419

    if (ipv6def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv6def) < 0))
J
Ján Tomko 已提交
1420
        goto cleanup;
1421

G
Gene Czarcinski 已提交
1422
    if (networkBuildDnsmasqHostsList(dctx, &network->def->dns) < 0)
J
Ján Tomko 已提交
1423
        goto cleanup;
1424 1425

    if ((ret = dnsmasqSave(dctx)) < 0)
1426
        goto cleanup;
1427 1428

    ret = kill(network->dnsmasqPid, SIGHUP);
1429
 cleanup:
1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441
    dnsmasqContextFree(dctx);
    return ret;
}

/* networkRestartDhcpDaemon:
 *
 * kill and restart dnsmasq, in order to update any config that is on
 * the dnsmasq commandline (and any placed in separate config files).
 *
 *  Returns 0 on success, -1 on failure.
 */
static int
1442
networkRestartDhcpDaemon(virNetworkObjPtr network)
1443 1444 1445 1446 1447 1448
{
    /* if there is a running dnsmasq, kill it */
    if (network->dnsmasqPid > 0) {
        networkKillDaemon(network->dnsmasqPid, "dnsmasq",
                          network->def->name);
        network->dnsmasqPid = -1;
1449
    }
1450
    /* now start dnsmasq if it should be started */
1451
    return networkStartDhcpDaemon(network);
1452 1453
}

G
Gene Czarcinski 已提交
1454 1455 1456 1457 1458 1459
static char radvd1[] = "  AdvOtherConfigFlag off;\n\n";
static char radvd2[] = "    AdvAutonomous off;\n";
static char radvd3[] = "    AdvOnLink on;\n"
                       "    AdvAutonomous on;\n"
                       "    AdvRouterAddr off;\n";

1460 1461 1462
static int
networkRadvdConfContents(virNetworkObjPtr network, char **configstr)
{
E
Eric Blake 已提交
1463
    virBuffer configbuf = VIR_BUFFER_INITIALIZER;
1464 1465
    int ret = -1;
    size_t i;
1466
    virNetworkIpDefPtr ipdef;
G
Gene Czarcinski 已提交
1467
    bool v6present = false, dhcp6 = false;
1468 1469

    *configstr = NULL;
1470

G
Gene Czarcinski 已提交
1471
    /* Check if DHCPv6 is needed */
1472 1473 1474
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
         i++) {
G
Gene Czarcinski 已提交
1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487
        v6present = true;
        if (ipdef->nranges || ipdef->nhosts) {
            dhcp6 = true;
            break;
        }
    }

    /* If there are no IPv6 addresses, then we are done */
    if (!v6present) {
        ret = 0;
        goto cleanup;
    }

1488 1489 1490
    /* create radvd config file appropriate for this network;
     * IgnoreIfMissing allows radvd to start even when the bridge is down
     */
1491
    virBufferAsprintf(&configbuf, "interface %s\n"
1492 1493
                      "{\n"
                      "  AdvSendAdvert on;\n"
1494
                      "  IgnoreIfMissing on;\n"
G
Gene Czarcinski 已提交
1495 1496 1497 1498 1499
                      "  AdvManagedFlag %s;\n"
                      "%s",
                      network->def->bridge,
                      dhcp6 ? "on" : "off",
                      dhcp6 ? "\n" : radvd1);
1500 1501

    /* add a section for each IPv6 address in the config */
1502 1503 1504
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
         i++) {
1505 1506 1507 1508 1509
        int prefix;
        char *netaddr;

        prefix = virNetworkIpDefPrefix(ipdef);
        if (prefix < 0) {
1510 1511 1512
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("bridge '%s' has an invalid prefix"),
                           network->def->bridge);
1513 1514
            goto cleanup;
        }
1515
        if (!(netaddr = virSocketAddrFormat(&ipdef->address)))
1516
            goto cleanup;
1517
        virBufferAsprintf(&configbuf,
1518
                          "  prefix %s/%d\n"
G
Gene Czarcinski 已提交
1519 1520 1521
                          "  {\n%s  };\n",
                          netaddr, prefix,
                          dhcp6 ? radvd2 : radvd3);
1522 1523 1524
        VIR_FREE(netaddr);
    }

1525
    virBufferAddLit(&configbuf, "};\n");
1526

1527
    if (virBufferCheckError(&configbuf) < 0)
1528
        goto cleanup;
1529

1530 1531
    *configstr = virBufferContentAndReset(&configbuf);

1532
    ret = 0;
1533
 cleanup:
1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555
    virBufferFreeAndReset(&configbuf);
    return ret;
}

/* write file and return it's name (which must be freed by caller) */
static int
networkRadvdConfWrite(virNetworkObjPtr network, char **configFile)
{
    int ret = -1;
    char *configStr = NULL;
    char *myConfigFile = NULL;

    if (!configFile)
        configFile = &myConfigFile;

    *configFile = NULL;

    if (networkRadvdConfContents(network, &configStr) < 0)
        goto cleanup;

    if (!configStr) {
        ret = 0;
1556 1557 1558 1559
        goto cleanup;
    }

    /* construct the filename */
1560
    if (!(*configFile = networkRadvdConfigFileName(network->def->name)))
1561 1562
        goto cleanup;
    /* write the file */
1563
    if (virFileWriteStr(*configFile, configStr, 0600) < 0) {
1564 1565
        virReportSystemError(errno,
                             _("couldn't write radvd config file '%s'"),
1566 1567 1568 1569 1570
                             *configFile);
        goto cleanup;
    }

    ret = 0;
1571
 cleanup:
1572 1573 1574 1575 1576 1577
    VIR_FREE(configStr);
    VIR_FREE(myConfigFile);
    return ret;
}

static int
1578
networkStartRadvd(virNetworkObjPtr network)
1579 1580 1581 1582 1583 1584 1585 1586 1587
{
    char *pidfile = NULL;
    char *radvdpidbase = NULL;
    char *configfile = NULL;
    virCommandPtr cmd = NULL;
    int ret = -1;

    network->radvdPid = -1;

G
Gene Czarcinski 已提交
1588
    /* Is dnsmasq handling RA? */
J
Ján Tomko 已提交
1589
    if (DNSMASQ_RA_SUPPORT(driver->dnsmasqCaps)) {
G
Gene Czarcinski 已提交
1590 1591 1592 1593
        ret = 0;
        goto cleanup;
    }

1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604
    if (!virNetworkDefGetIpByIndex(network->def, AF_INET6, 0)) {
        /* no IPv6 addresses, so we don't need to run radvd */
        ret = 0;
        goto cleanup;
    }

    if (!virFileIsExecutable(RADVD)) {
        virReportSystemError(errno,
                             _("Cannot find %s - "
                               "Possibly the package isn't installed"),
                             RADVD);
1605 1606 1607
        goto cleanup;
    }

1608
    if (virFileMakePath(driver->pidDir) < 0) {
1609 1610
        virReportSystemError(errno,
                             _("cannot create directory %s"),
1611
                             driver->pidDir);
1612 1613
        goto cleanup;
    }
1614
    if (virFileMakePath(driver->radvdStateDir) < 0) {
1615 1616
        virReportSystemError(errno,
                             _("cannot create directory %s"),
1617
                             driver->radvdStateDir);
1618 1619 1620 1621
        goto cleanup;
    }

    /* construct pidfile name */
1622
    if (!(radvdpidbase = networkRadvdPidfileBasename(network->def->name)))
1623
        goto cleanup;
1624
    if (!(pidfile = virPidFileBuildPath(driver->pidDir, radvdpidbase)))
1625 1626 1627 1628 1629
        goto cleanup;

    if (networkRadvdConfWrite(network, &configfile) < 0)
        goto cleanup;

1630 1631 1632 1633
    /* prevent radvd from daemonizing itself with "--debug 1", and use
     * a dummy pidfile name - virCommand will create the pidfile we
     * want to use (this is necessary because radvd's internal
     * daemonization and pidfile creation causes a race, and the
1634
     * virPidFileRead() below will fail if we use them).
1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649
     * Unfortunately, it isn't possible to tell radvd to not create
     * its own pidfile, so we just let it do so, with a slightly
     * different name. Unused, but harmless.
     */
    cmd = virCommandNewArgList(RADVD, "--debug", "1",
                               "--config", configfile,
                               "--pidfile", NULL);
    virCommandAddArgFormat(cmd, "%s-bin", pidfile);

    virCommandSetPidFile(cmd, pidfile);
    virCommandDaemonize(cmd);

    if (virCommandRun(cmd, NULL) < 0)
        goto cleanup;

1650
    if (virPidFileRead(driver->pidDir, radvdpidbase, &network->radvdPid) < 0)
1651 1652 1653
        goto cleanup;

    ret = 0;
1654
 cleanup:
1655 1656 1657 1658 1659 1660 1661
    virCommandFree(cmd);
    VIR_FREE(configfile);
    VIR_FREE(radvdpidbase);
    VIR_FREE(pidfile);
    return ret;
}

1662
static int
1663
networkRefreshRadvd(virNetworkObjPtr network)
1664
{
G
Gene Czarcinski 已提交
1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675
    char *radvdpidbase;

    /* Is dnsmasq handling RA? */
    if (DNSMASQ_RA_SUPPORT(driver->dnsmasqCaps)) {
        if (network->radvdPid <= 0)
            return 0;
        /* radvd should not be running but in case it is */
        if ((networkKillDaemon(network->radvdPid, "radvd",
                               network->def->name) >= 0) &&
            ((radvdpidbase = networkRadvdPidfileBasename(network->def->name))
             != NULL)) {
1676
            virPidFileDelete(driver->pidDir, radvdpidbase);
G
Gene Czarcinski 已提交
1677 1678 1679 1680 1681 1682
            VIR_FREE(radvdpidbase);
        }
        network->radvdPid = -1;
        return 0;
    }

1683 1684
    /* if there's no running radvd, just start it */
    if (network->radvdPid <= 0 || (kill(network->radvdPid, 0) < 0))
1685
        return networkStartRadvd(network);
1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697

    if (!virNetworkDefGetIpByIndex(network->def, AF_INET6, 0)) {
        /* no IPv6 addresses, so we don't need to run radvd */
        return 0;
    }

    if (networkRadvdConfWrite(network, NULL) < 0)
        return -1;

    return kill(network->radvdPid, SIGHUP);
}

1698 1699
#if 0
/* currently unused, so it causes a build error unless we #if it out */
1700
static int
1701
networkRestartRadvd(virNetworkObjPtr network)
1702 1703 1704 1705 1706 1707 1708 1709 1710
{
    char *radvdpidbase;

    /* if there is a running radvd, kill it */
    if (network->radvdPid > 0) {
        /* essentially ignore errors from the following two functions,
         * since there's really no better recovery to be done than to
         * just push ahead (and that may be exactly what's needed).
         */
G
Gene Czarcinski 已提交
1711
        if ((networkKillDaemon(network->radvdPid, "radvd",
1712 1713 1714
                               network->def->name) >= 0) &&
            ((radvdpidbase = networkRadvdPidfileBasename(network->def->name))
             != NULL)) {
1715
            virPidFileDelete(driver->pidDir, radvdpidbase);
1716 1717 1718 1719 1720 1721 1722 1723 1724
            VIR_FREE(radvdpidbase);
        }
        network->radvdPid = -1;
    }
    /* now start radvd if it should be started */
    return networkStartRadvd(network);
}
#endif /* #if 0 */

1725 1726 1727 1728
/* SIGHUP/restart any dnsmasq or radvd daemons.
 * This should be called when libvirtd is restarted.
 */
static void
1729
networkRefreshDaemons(void)
1730
{
1731
    size_t i;
1732 1733 1734

    VIR_INFO("Refreshing network daemons");

1735
    for (i = 0; i < driver->networks.count; i++) {
1736 1737 1738 1739
        virNetworkObjPtr network = driver->networks.objs[i];

        virNetworkObjLock(network);
        if (virNetworkObjIsActive(network) &&
1740 1741 1742
            ((network->def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE))) {
1743 1744 1745 1746 1747 1748
            /* Only the three L3 network types that are configured by
             * libvirt will have a dnsmasq or radvd daemon associated
             * with them.  Here we send a SIGHUP to an existing
             * dnsmasq and/or radvd, or restart them if they've
             * disappeared.
             */
1749 1750
            networkRefreshDhcpDaemon(network);
            networkRefreshRadvd(network);
1751 1752 1753 1754 1755
        }
        virNetworkObjUnlock(network);
    }
}

1756
static void
1757
networkReloadFirewallRules(void)
1758
{
1759
    size_t i;
1760

1761
    VIR_INFO("Reloading iptables rules");
1762

1763
    for (i = 0; i < driver->networks.count; i++) {
1764 1765 1766 1767
        virNetworkObjPtr network = driver->networks.objs[i];

        virNetworkObjLock(network);
        if (virNetworkObjIsActive(network) &&
1768 1769 1770
            ((network->def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE))) {
1771 1772 1773
            /* Only the three L3 network types that are configured by libvirt
             * need to have iptables rules reloaded.
             */
1774 1775
            networkRemoveFirewallRules(network->def);
            if (networkAddFirewallRules(network->def) < 0) {
1776 1777
                /* failed to add but already logged */
            }
1778
        }
1779
        virNetworkObjUnlock(network);
1780 1781 1782
    }
}

1783
/* Enable IP Forwarding. Return 0 for success, -1 for failure. */
1784
static int
1785
networkEnableIpForwarding(bool enableIPv4, bool enableIPv6)
1786
{
1787
    int ret = 0;
1788 1789 1790 1791
#ifdef HAVE_SYSCTLBYNAME
    int enabled = 1;
    if (enableIPv4)
        ret = sysctlbyname("net.inet.ip.forwarding", NULL, 0,
J
Ján Tomko 已提交
1792
                           &enabled, sizeof(enabled));
1793 1794
    if (enableIPv6 && ret == 0)
        ret = sysctlbyname("net.inet6.ip6.forwarding", NULL, 0,
J
Ján Tomko 已提交
1795
                           &enabled, sizeof(enabled));
1796
#else
1797 1798 1799 1800
    if (enableIPv4)
        ret = virFileWriteStr("/proc/sys/net/ipv4/ip_forward", "1\n", 0);
    if (enableIPv6 && ret == 0)
        ret = virFileWriteStr("/proc/sys/net/ipv6/conf/all/forwarding", "1\n", 0);
1801
#endif
1802
    return ret;
1803 1804
}

1805 1806
#define SYSCTL_PATH "/proc/sys"

1807 1808
static int
networkSetIPv6Sysctls(virNetworkObjPtr network)
1809 1810 1811
{
    char *field = NULL;
    int ret = -1;
1812
    bool enableIPv6 =  !!virNetworkDefGetIpByIndex(network->def, AF_INET6, 0);
1813

1814 1815 1816 1817 1818 1819 1820
    /* set disable_ipv6 if there are no ipv6 addresses defined for the
     * network. But also unset it if there *are* ipv6 addresses, as we
     * can't be sure of its default value.
     */
    if (virAsprintf(&field, SYSCTL_PATH "/net/ipv6/conf/%s/disable_ipv6",
                    network->def->bridge) < 0)
       goto cleanup;
1821

1822 1823
    if (access(field, W_OK) < 0 && errno == ENOENT) {
        if (!enableIPv6)
1824 1825
            VIR_DEBUG("ipv6 appears to already be disabled on %s",
                      network->def->bridge);
1826 1827 1828
        ret = 0;
        goto cleanup;
    }
1829

1830 1831 1832 1833 1834
    if (virFileWriteStr(field, enableIPv6 ? "0" : "1", 0) < 0) {
        virReportSystemError(errno,
                             _("cannot write to %s to enable/disable IPv6 "
                               "on bridge %s"), field, network->def->bridge);
        goto cleanup;
1835
    }
1836
    VIR_FREE(field);
1837

1838 1839
    /* The rest of the ipv6 sysctl tunables should always be set the
     * same, whether or not we're using ipv6 on this bridge.
1840 1841 1842 1843 1844 1845
     */

    /* Prevent guests from hijacking the host network by sending out
     * their own router advertisements.
     */
    if (virAsprintf(&field, SYSCTL_PATH "/net/ipv6/conf/%s/accept_ra",
1846
                    network->def->bridge) < 0)
1847 1848
        goto cleanup;

1849
    if (virFileWriteStr(field, "0", 0) < 0) {
1850
        virReportSystemError(errno,
1851 1852 1853 1854 1855
                             _("cannot disable %s"), field);
        goto cleanup;
    }
    VIR_FREE(field);

1856 1857 1858 1859
    /* All interfaces used as a gateway (which is what this is, by
     * definition), must always have autoconf=0.
     */
    if (virAsprintf(&field, SYSCTL_PATH "/net/ipv6/conf/%s/autoconf",
1860
                    network->def->bridge) < 0)
1861 1862
        goto cleanup;

1863
    if (virFileWriteStr(field, "0", 0) < 0) {
1864
        virReportSystemError(errno,
1865
                             _("cannot disable %s"), field);
1866 1867 1868 1869
        goto cleanup;
    }

    ret = 0;
1870
 cleanup:
1871 1872 1873 1874
    VIR_FREE(field);
    return ret;
}

1875
/* add an IP address to a bridge */
1876
static int
D
Daniel P. Berrange 已提交
1877
networkAddAddrToBridge(virNetworkObjPtr network,
1878
                       virNetworkIpDefPtr ipdef)
1879
{
1880 1881 1882
    int prefix = virNetworkIpDefPrefix(ipdef);

    if (prefix < 0) {
1883 1884 1885
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("bridge '%s' has an invalid netmask or IP address"),
                       network->def->bridge);
1886 1887 1888
        return -1;
    }

1889 1890
    if (virNetDevSetIPv4Address(network->def->bridge,
                                &ipdef->address, prefix) < 0)
1891 1892 1893 1894 1895
        return -1;

    return 0;
}

1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908
/* add an IP (static) route to a bridge */
static int
networkAddRouteToBridge(virNetworkObjPtr network,
                        virNetworkRouteDefPtr routedef)
{
    int prefix = 0;
    unsigned int metric;
    virSocketAddrPtr addr = &routedef->address;
    virSocketAddrPtr mask = &routedef->netmask;
    virSocketAddr zero;

    /* this creates an all-0 address of the appropriate family */
    ignore_value(virSocketAddrParse(&zero,
1909
                                    (VIR_SOCKET_ADDR_IS_FAMILY(addr, AF_INET)
1910 1911 1912 1913 1914 1915 1916
                                     ? "0.0.0.0" : "::"),
                                    VIR_SOCKET_ADDR_FAMILY(addr)));

    if (virSocketAddrEqual(addr, &zero)) {
        if (routedef->has_prefix && routedef->prefix == 0)
            prefix = 0;
        else if ((VIR_SOCKET_ADDR_IS_FAMILY(mask, AF_INET) &&
J
Ján Tomko 已提交
1917
                  virSocketAddrEqual(mask, &zero)))
1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944
            prefix = 0;
        else
            prefix = virSocketAddrGetIpPrefix(addr, mask, routedef->prefix);
    } else {
        prefix = virSocketAddrGetIpPrefix(addr, mask, routedef->prefix);
    }

    if (prefix < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("network '%s' has an invalid netmask "
                         "or IP address in route definition"),
                       network->def->name);
        return -1;
    }

    if (routedef->has_metric && routedef->metric > 0)
        metric = routedef->metric;
    else
        metric = 1;

    if (virNetDevAddRoute(network->def->bridge, &routedef->address,
                          prefix, &routedef->gateway, metric) < 0) {
        return -1;
    }
    return 0;
}

1945
static int
1946
networkStartNetworkVirtual(virNetworkObjPtr network)
1947
{
1948
    size_t i;
1949
    bool v4present = false, v6present = false;
1950 1951
    virErrorPtr save_err = NULL;
    virNetworkIpDefPtr ipdef;
1952
    virNetworkRouteDefPtr routedef;
1953
    char *macTapIfName = NULL;
1954
    int tapfd = -1;
1955

1956
    /* Check to see if any network IP collides with an existing route */
1957
    if (networkCheckRouteCollision(network->def) < 0)
1958 1959
        return -1;

1960
    /* Create and configure the bridge device */
1961
    if (virNetDevBridgeCreate(network->def->bridge) < 0)
1962 1963
        return -1;

1964 1965 1966 1967 1968 1969 1970 1971
    if (network->def->mac_specified) {
        /* To set a mac for the bridge, we need to define a dummy tap
         * device, set its mac, then attach it to the bridge. As long
         * as its mac address is lower than any other interface that
         * gets attached, the bridge will always maintain this mac
         * address.
         */
        macTapIfName = networkBridgeDummyNicName(network->def->bridge);
1972
        if (!macTapIfName)
1973
            goto err0;
1974
        /* Keep tun fd open and interface up to allow for IPv6 DAD to happen */
1975
        if (virNetDevTapCreateInBridgePort(network->def->bridge,
1976
                                           &macTapIfName, &network->def->mac,
1977
                                           NULL, NULL, &tapfd, 1, NULL, NULL,
1978 1979 1980
                                           VIR_NETDEV_TAP_CREATE_USE_MAC_FOR_BRIDGE |
                                           VIR_NETDEV_TAP_CREATE_IFUP |
                                           VIR_NETDEV_TAP_CREATE_PERSIST) < 0) {
1981 1982 1983 1984 1985
            VIR_FREE(macTapIfName);
            goto err0;
        }
    }

1986
    /* Set bridge options */
1987 1988 1989 1990

    /* delay is configured in seconds, but virNetDevBridgeSetSTPDelay
     * expects milliseconds
     */
1991
    if (virNetDevBridgeSetSTPDelay(network->def->bridge,
1992
                                   network->def->delay * 1000) < 0)
1993
        goto err1;
1994

1995
    if (virNetDevBridgeSetSTP(network->def->bridge,
1996
                              network->def->stp ? true : false) < 0)
1997
        goto err1;
1998

1999 2000 2001 2002
    /* Disable IPv6 on the bridge if there are no IPv6 addresses
     * defined, and set other IPv6 sysctl tunables appropriately.
     */
    if (networkSetIPv6Sysctls(network) < 0)
2003
        goto err1;
2004

2005
    /* Add "once per network" rules */
2006
    if (networkAddFirewallRules(network->def) < 0)
2007 2008
        goto err1;

2009 2010 2011
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, i));
         i++) {
2012
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET))
2013
            v4present = true;
2014
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
2015
            v6present = true;
2016

2017
        /* Add the IP address/netmask to the bridge */
2018
        if (networkAddAddrToBridge(network, ipdef) < 0)
2019
            goto err2;
2020 2021
    }

2022
    /* Bring up the bridge interface */
2023
    if (virNetDevSetOnline(network->def->bridge, 1) < 0)
2024
        goto err2;
2025

2026 2027
    for (i = 0; i < network->def->nroutes; i++) {
        routedef = &network->def->routes[i];
2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038
        /* Add the IP route to the bridge */
        /* ignore errors, error msg will be generated */
        /* but libvirt will not know and net-destroy will work. */
        if (VIR_SOCKET_ADDR_VALID(&routedef->gateway)) {
            if (networkAddRouteToBridge(network, routedef) < 0) {
                /* an error occurred adding the static route */
                continue; /* for now, do nothing */
            }
        }
    }

2039 2040
    /* If forward.type != NONE, turn on global IP forwarding */
    if (network->def->forward.type != VIR_NETWORK_FORWARD_NONE &&
2041
        networkEnableIpForwarding(v4present, v6present) < 0) {
2042
        virReportSystemError(errno, "%s",
2043
                             _("failed to enable IP forwarding"));
2044
        goto err3;
2045 2046
    }

2047

2048
    /* start dnsmasq if there are any IP addresses (v4 or v6) */
2049
    if ((v4present || v6present) &&
2050
        networkStartDhcpDaemon(network) < 0)
2051
        goto err3;
2052

2053
    /* start radvd if there are any ipv6 addresses */
2054
    if (v6present && networkStartRadvd(network) < 0)
2055 2056
        goto err4;

2057 2058 2059 2060 2061 2062 2063 2064 2065
    /* DAD has happened (dnsmasq waits for it), dnsmasq is now bound to the
     * bridge's IPv6 address, so we can now set the dummy tun down.
     */
    if (tapfd >= 0) {
        if (virNetDevSetOnline(macTapIfName, false) < 0)
            goto err4;
        VIR_FORCE_CLOSE(tapfd);
    }

2066
    if (virNetDevBandwidthSet(network->def->bridge,
2067
                              network->def->bandwidth, true) < 0)
2068 2069
        goto err5;

2070
    VIR_FREE(macTapIfName);
2071 2072 2073

    return 0;

2074
 err5:
2075
    virNetDevBandwidthClear(network->def->bridge);
2076

2077 2078 2079 2080
 err4:
    if (!save_err)
        save_err = virSaveLastError();

2081 2082 2083 2084 2085
    if (network->dnsmasqPid > 0) {
        kill(network->dnsmasqPid, SIGTERM);
        network->dnsmasqPid = -1;
    }

2086 2087 2088
 err3:
    if (!save_err)
        save_err = virSaveLastError();
2089
    ignore_value(virNetDevSetOnline(network->def->bridge, 0));
2090

2091 2092 2093
 err2:
    if (!save_err)
        save_err = virSaveLastError();
2094
    networkRemoveFirewallRules(network->def);
2095 2096

 err1:
2097 2098 2099
    if (!save_err)
        save_err = virSaveLastError();

H
Hu Tao 已提交
2100
    if (macTapIfName) {
2101
        VIR_FORCE_CLOSE(tapfd);
2102
        ignore_value(virNetDevTapDelete(macTapIfName, NULL));
H
Hu Tao 已提交
2103 2104
        VIR_FREE(macTapIfName);
    }
2105 2106

 err0:
2107 2108
    if (!save_err)
        save_err = virSaveLastError();
2109
    ignore_value(virNetDevBridgeDelete(network->def->bridge));
2110

2111 2112 2113 2114
    if (save_err) {
        virSetError(save_err);
        virFreeError(save_err);
    }
2115
    /* coverity[leaked_handle] - 'tapfd' is not leaked */
2116 2117 2118
    return -1;
}

2119
static int networkShutdownNetworkVirtual(virNetworkObjPtr network)
2120
{
2121
    virNetDevBandwidthClear(network->def->bridge);
2122

2123 2124 2125 2126 2127
    if (network->radvdPid > 0) {
        char *radvdpidbase;

        kill(network->radvdPid, SIGTERM);
        /* attempt to delete the pidfile we created */
2128
        if ((radvdpidbase = networkRadvdPidfileBasename(network->def->name))) {
2129
            virPidFileDelete(driver->pidDir, radvdpidbase);
2130 2131 2132 2133
            VIR_FREE(radvdpidbase);
        }
    }

2134 2135 2136
    if (network->dnsmasqPid > 0)
        kill(network->dnsmasqPid, SIGTERM);

2137
    if (network->def->mac_specified) {
2138
        char *macTapIfName = networkBridgeDummyNicName(network->def->bridge);
2139
        if (macTapIfName) {
2140
            ignore_value(virNetDevTapDelete(macTapIfName, NULL));
2141 2142 2143 2144
            VIR_FREE(macTapIfName);
        }
    }

2145
    ignore_value(virNetDevSetOnline(network->def->bridge, 0));
2146

2147
    networkRemoveFirewallRules(network->def);
2148

2149
    ignore_value(virNetDevBridgeDelete(network->def->bridge));
2150

2151
    /* See if its still alive and really really kill it */
2152
    if (network->dnsmasqPid > 0 &&
2153
        (kill(network->dnsmasqPid, 0) == 0))
2154 2155
        kill(network->dnsmasqPid, SIGKILL);
    network->dnsmasqPid = -1;
2156 2157 2158 2159 2160 2161

    if (network->radvdPid > 0 &&
        (kill(network->radvdPid, 0) == 0))
        kill(network->radvdPid, SIGKILL);
    network->radvdPid = -1;

2162 2163 2164
    return 0;
}

2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180

/* networkCreateInterfacePool:
 * @netdef: the original NetDef from the network
 *
 * Creates an implicit interface pool of VF's when a PF dev is given
 */
static int
networkCreateInterfacePool(virNetworkDefPtr netdef)
{
    size_t numVirtFns = 0;
    char **vfNames = NULL;
    virPCIDeviceAddressPtr *virtFns;

    int ret = -1;
    size_t i;

2181 2182 2183
    if (netdef->forward.npfs == 0 || netdef->forward.nifs > 0)
       return 0;

2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271
    if ((virNetDevGetVirtualFunctions(netdef->forward.pfs->dev,
                                      &vfNames, &virtFns, &numVirtFns)) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Could not get Virtual functions on %s"),
                       netdef->forward.pfs->dev);
        goto cleanup;
    }

    if (VIR_ALLOC_N(netdef->forward.ifs, numVirtFns) < 0)
        goto cleanup;

    for (i = 0; i < numVirtFns; i++) {
        virPCIDeviceAddressPtr thisVirtFn = virtFns[i];
        const char *thisName = vfNames[i];
        virNetworkForwardIfDefPtr thisIf
            = &netdef->forward.ifs[netdef->forward.nifs];

        switch (netdef->forward.type) {
        case VIR_NETWORK_FORWARD_BRIDGE:
        case VIR_NETWORK_FORWARD_PRIVATE:
        case VIR_NETWORK_FORWARD_VEPA:
        case VIR_NETWORK_FORWARD_PASSTHROUGH:
            if (thisName) {
                if (VIR_STRDUP(thisIf->device.dev, thisName) < 0)
                    goto cleanup;
                thisIf->type = VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV;
                netdef->forward.nifs++;
            } else {
                VIR_WARN("VF %zu of SRIOV PF %s couldn't be added to the "
                         "interface pool because it isn't bound "
                         "to a network driver - possibly in use elsewhere",
                         i, netdef->forward.pfs->dev);
            }
            break;

        case VIR_NETWORK_FORWARD_HOSTDEV:
            /* VF's are always PCI devices */
            thisIf->type = VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI;
            thisIf->device.pci.domain = thisVirtFn->domain;
            thisIf->device.pci.bus = thisVirtFn->bus;
            thisIf->device.pci.slot = thisVirtFn->slot;
            thisIf->device.pci.function = thisVirtFn->function;
            netdef->forward.nifs++;
            break;

        case VIR_NETWORK_FORWARD_NONE:
        case VIR_NETWORK_FORWARD_NAT:
        case VIR_NETWORK_FORWARD_ROUTE:
        case VIR_NETWORK_FORWARD_LAST:
            /* by definition these will never be encountered here */
            break;
        }
    }

    if (netdef->forward.nifs == 0) {
        /* If we don't get at least one interface in the pool, declare
         * failure
         */
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("No usable Vf's present on SRIOV PF %s"),
                       netdef->forward.pfs->dev);
        goto cleanup;
    }

    ret = 0;
 cleanup:
    if (ret < 0) {
        /* free all the entries made before error */
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV)
                VIR_FREE(netdef->forward.ifs[i].device.dev);
        }
        netdef->forward.nifs = 0;
    }
    if (netdef->forward.nifs == 0)
        VIR_FREE(netdef->forward.ifs);

    for (i = 0; i < numVirtFns; i++) {
        VIR_FREE(vfNames[i]);
        VIR_FREE(virtFns[i]);
    }
    VIR_FREE(vfNames);
    VIR_FREE(virtFns);
    return ret;
}


2272
static int
2273
networkStartNetworkExternal(virNetworkObjPtr network)
2274 2275
{
    /* put anything here that needs to be done each time a network of
2276
     * type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is started. On
2277 2278 2279
     * failure, undo anything you've done, and return -1. On success
     * return 0.
     */
2280
    return networkCreateInterfacePool(network->def);
2281 2282
}

2283
static int networkShutdownNetworkExternal(virNetworkObjPtr network ATTRIBUTE_UNUSED)
2284 2285
{
    /* put anything here that needs to be done each time a network of
2286
     * type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is shutdown. On
2287 2288 2289 2290 2291 2292 2293
     * failure, undo anything you've done, and return -1. On success
     * return 0.
     */
    return 0;
}

static int
2294
networkStartNetwork(virNetworkObjPtr network)
2295
{
2296 2297 2298
    int ret = -1;

    VIR_DEBUG("driver=%p, network=%p", driver, network);
2299 2300

    if (virNetworkObjIsActive(network)) {
2301 2302
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("network is already active"));
2303
        return ret;
2304 2305
    }

2306 2307 2308
    VIR_DEBUG("Beginning network startup process");

    VIR_DEBUG("Setting current network def as transient");
2309
    if (virNetworkObjSetDefTransient(network, true) < 0)
2310
        goto cleanup;
2311

2312 2313
    /* Run an early hook to set-up missing devices.
     * If the script raised an error abort the launch. */
2314
    if (networkRunHook(network, NULL, NULL,
2315 2316 2317 2318
                       VIR_HOOK_NETWORK_OP_START,
                       VIR_HOOK_SUBOP_BEGIN) < 0)
        goto cleanup;

2319
    switch (network->def->forward.type) {
2320 2321 2322 2323

    case VIR_NETWORK_FORWARD_NONE:
    case VIR_NETWORK_FORWARD_NAT:
    case VIR_NETWORK_FORWARD_ROUTE:
2324
        if (networkStartNetworkVirtual(network) < 0)
2325
            goto cleanup;
2326 2327 2328 2329 2330 2331
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
    case VIR_NETWORK_FORWARD_PRIVATE:
    case VIR_NETWORK_FORWARD_VEPA:
    case VIR_NETWORK_FORWARD_PASSTHROUGH:
2332
    case VIR_NETWORK_FORWARD_HOSTDEV:
2333
        if (networkStartNetworkExternal(network) < 0)
2334
            goto cleanup;
2335 2336 2337
        break;
    }

2338
    /* finally we can call the 'started' hook script if any */
2339
    if (networkRunHook(network, NULL, NULL,
2340 2341 2342 2343
                       VIR_HOOK_NETWORK_OP_STARTED,
                       VIR_HOOK_SUBOP_BEGIN) < 0)
        goto cleanup;

2344 2345 2346
    /* Persist the live configuration now that anything autogenerated
     * is setup.
     */
2347
    VIR_DEBUG("Writing network status to disk");
2348
    if (virNetworkSaveStatus(driver->stateDir, network) < 0)
2349
        goto cleanup;
2350 2351

    network->active = 1;
2352 2353
    VIR_INFO("Network '%s' started up", network->def->name);
    ret = 0;
2354

2355
 cleanup:
2356
    if (ret < 0) {
2357
        virNetworkObjUnsetDefTransient(network);
2358 2359
        virErrorPtr save_err = virSaveLastError();
        int save_errno = errno;
2360
        networkShutdownNetwork(network);
2361 2362 2363 2364 2365 2366 2367
        virSetError(save_err);
        virFreeError(save_err);
        errno = save_errno;
    }
    return ret;
}

2368
static int networkShutdownNetwork(virNetworkObjPtr network)
2369 2370 2371 2372 2373 2374 2375 2376 2377
{
    int ret = 0;
    char *stateFile;

    VIR_INFO("Shutting down network '%s'", network->def->name);

    if (!virNetworkObjIsActive(network))
        return 0;

2378
    stateFile = virNetworkConfigFile(driver->stateDir,
2379
                                     network->def->name);
2380 2381 2382 2383 2384 2385
    if (!stateFile)
        return -1;

    unlink(stateFile);
    VIR_FREE(stateFile);

2386
    switch (network->def->forward.type) {
2387 2388 2389 2390

    case VIR_NETWORK_FORWARD_NONE:
    case VIR_NETWORK_FORWARD_NAT:
    case VIR_NETWORK_FORWARD_ROUTE:
2391
        ret = networkShutdownNetworkVirtual(network);
2392 2393 2394 2395 2396 2397
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
    case VIR_NETWORK_FORWARD_PRIVATE:
    case VIR_NETWORK_FORWARD_VEPA:
    case VIR_NETWORK_FORWARD_PASSTHROUGH:
2398
    case VIR_NETWORK_FORWARD_HOSTDEV:
2399
        ret = networkShutdownNetworkExternal(network);
2400 2401 2402
        break;
    }

2403
    /* now that we know it's stopped call the hook if present */
2404
    networkRunHook(network, NULL, NULL, VIR_HOOK_NETWORK_OP_STOPPED,
2405 2406
                   VIR_HOOK_SUBOP_END);

2407
    network->active = 0;
2408
    virNetworkObjUnsetDefTransient(network);
2409
    return ret;
2410 2411 2412
}


2413
static virNetworkPtr networkLookupByUUID(virConnectPtr conn,
2414 2415
                                         const unsigned char *uuid)
{
2416 2417
    virNetworkObjPtr network;
    virNetworkPtr ret = NULL;
2418

2419
    networkDriverLock();
2420
    network = virNetworkFindByUUID(&driver->networks, uuid);
2421
    networkDriverUnlock();
2422
    if (!network) {
2423 2424
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2425
        goto cleanup;
2426 2427
    }

2428 2429 2430
    if (virNetworkLookupByUUIDEnsureACL(conn, network->def) < 0)
        goto cleanup;

2431 2432
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

2433
 cleanup:
2434 2435
    if (network)
        virNetworkObjUnlock(network);
2436
    return ret;
2437 2438
}

2439
static virNetworkPtr networkLookupByName(virConnectPtr conn,
2440 2441
                                         const char *name)
{
2442 2443 2444
    virNetworkObjPtr network;
    virNetworkPtr ret = NULL;

2445
    networkDriverLock();
2446
    network = virNetworkFindByName(&driver->networks, name);
2447
    networkDriverUnlock();
2448
    if (!network) {
2449 2450
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"), name);
2451
        goto cleanup;
2452 2453
    }

2454 2455 2456
    if (virNetworkLookupByNameEnsureACL(conn, network->def) < 0)
        goto cleanup;

2457 2458
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

2459
 cleanup:
2460 2461
    if (network)
        virNetworkObjUnlock(network);
2462
    return ret;
2463 2464
}

2465
static virDrvOpenStatus networkOpen(virConnectPtr conn ATTRIBUTE_UNUSED,
2466 2467
                                    virConnectAuthPtr auth ATTRIBUTE_UNUSED,
                                    unsigned int flags)
2468 2469 2470
{
    virCheckFlags(VIR_CONNECT_RO, VIR_DRV_OPEN_ERROR);

2471
    if (!driver)
2472 2473 2474 2475 2476
        return VIR_DRV_OPEN_DECLINED;

    return VIR_DRV_OPEN_SUCCESS;
}

2477
static int networkClose(virConnectPtr conn ATTRIBUTE_UNUSED)
2478
{
2479 2480 2481
    return 0;
}

2482 2483
static int networkConnectNumOfNetworks(virConnectPtr conn)
{
2484 2485
    int nactive = 0;
    size_t i;
2486

2487 2488 2489
    if (virConnectNumOfNetworksEnsureACL(conn) < 0)
        return -1;

2490
    networkDriverLock();
2491
    for (i = 0; i < driver->networks.count; i++) {
2492 2493 2494 2495
        virNetworkObjPtr obj = driver->networks.objs[i];
        virNetworkObjLock(obj);
        if (virConnectNumOfNetworksCheckACL(conn, obj->def) &&
            virNetworkObjIsActive(obj))
2496
            nactive++;
2497
        virNetworkObjUnlock(obj);
2498
    }
2499
    networkDriverUnlock();
2500

2501 2502 2503
    return nactive;
}

2504
static int networkConnectListNetworks(virConnectPtr conn, char **const names, int nnames) {
2505 2506
    int got = 0;
    size_t i;
2507

2508 2509 2510
    if (virConnectListNetworksEnsureACL(conn) < 0)
        return -1;

2511
    networkDriverLock();
2512
    for (i = 0; i < driver->networks.count && got < nnames; i++) {
2513 2514 2515 2516 2517 2518
        virNetworkObjPtr obj = driver->networks.objs[i];
        virNetworkObjLock(obj);
        if (virConnectListNetworksCheckACL(conn, obj->def) &&
            virNetworkObjIsActive(obj)) {
            if (VIR_STRDUP(names[got], obj->def->name) < 0) {
                virNetworkObjUnlock(obj);
2519 2520 2521 2522
                goto cleanup;
            }
            got++;
        }
2523
        virNetworkObjUnlock(obj);
2524
    }
2525
    networkDriverUnlock();
2526

2527 2528 2529
    return got;

 cleanup:
2530
    networkDriverUnlock();
2531
    for (i = 0; i < got; i++)
2532 2533 2534 2535
        VIR_FREE(names[i]);
    return -1;
}

2536 2537
static int networkConnectNumOfDefinedNetworks(virConnectPtr conn)
{
2538 2539
    int ninactive = 0;
    size_t i;
2540

2541 2542 2543
    if (virConnectNumOfDefinedNetworksEnsureACL(conn) < 0)
        return -1;

2544
    networkDriverLock();
2545
    for (i = 0; i < driver->networks.count; i++) {
2546 2547 2548 2549
        virNetworkObjPtr obj = driver->networks.objs[i];
        virNetworkObjLock(obj);
        if (virConnectNumOfDefinedNetworksCheckACL(conn, obj->def) &&
            !virNetworkObjIsActive(obj))
2550
            ninactive++;
2551
        virNetworkObjUnlock(obj);
2552
    }
2553
    networkDriverUnlock();
2554

2555 2556 2557
    return ninactive;
}

2558
static int networkConnectListDefinedNetworks(virConnectPtr conn, char **const names, int nnames) {
2559 2560
    int got = 0;
    size_t i;
2561

2562 2563 2564
    if (virConnectListDefinedNetworksEnsureACL(conn) < 0)
        return -1;

2565
    networkDriverLock();
2566
    for (i = 0; i < driver->networks.count && got < nnames; i++) {
2567 2568 2569 2570 2571 2572
        virNetworkObjPtr obj = driver->networks.objs[i];
        virNetworkObjLock(obj);
        if (virConnectListDefinedNetworksCheckACL(conn, obj->def) &&
            !virNetworkObjIsActive(obj)) {
            if (VIR_STRDUP(names[got], obj->def->name) < 0) {
                virNetworkObjUnlock(obj);
2573 2574 2575 2576
                goto cleanup;
            }
            got++;
        }
2577
        virNetworkObjUnlock(obj);
2578
    }
2579
    networkDriverUnlock();
2580 2581 2582
    return got;

 cleanup:
2583
    networkDriverUnlock();
2584
    for (i = 0; i < got; i++)
2585 2586 2587 2588
        VIR_FREE(names[i]);
    return -1;
}

2589
static int
2590 2591 2592
networkConnectListAllNetworks(virConnectPtr conn,
                              virNetworkPtr **nets,
                              unsigned int flags)
2593 2594 2595 2596 2597
{
    int ret = -1;

    virCheckFlags(VIR_CONNECT_LIST_NETWORKS_FILTERS_ALL, -1);

2598 2599 2600
    if (virConnectListAllNetworksEnsureACL(conn) < 0)
        goto cleanup;

2601
    networkDriverLock();
2602 2603 2604
    ret = virNetworkObjListExport(conn, driver->networks, nets,
                                  virConnectListAllNetworksCheckACL,
                                  flags);
2605
    networkDriverUnlock();
2606

2607
 cleanup:
2608 2609
    return ret;
}
2610

2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624
static int
networkConnectNetworkEventRegisterAny(virConnectPtr conn,
                                      virNetworkPtr net,
                                      int eventID,
                                      virConnectNetworkEventGenericCallback callback,
                                      void *opaque,
                                      virFreeCallback freecb)
{
    int ret = -1;

    if (virConnectNetworkEventRegisterAnyEnsureACL(conn) < 0)
        goto cleanup;

    if (virNetworkEventStateRegisterID(conn, driver->networkEventState,
2625
                                       net, eventID, callback,
2626 2627 2628
                                       opaque, freecb, &ret) < 0)
        ret = -1;

2629
 cleanup:
2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641
    return ret;
}

static int
networkConnectNetworkEventDeregisterAny(virConnectPtr conn,
                                        int callbackID)
{
    int ret = -1;

    if (virConnectNetworkEventDeregisterAnyEnsureACL(conn) < 0)
        goto cleanup;

2642 2643 2644 2645 2646 2647
    if (virObjectEventStateDeregisterID(conn,
                                        driver->networkEventState,
                                        callbackID) < 0)
        goto cleanup;

    ret = 0;
2648

2649
 cleanup:
2650 2651 2652
    return ret;
}

2653 2654 2655 2656 2657
static int networkIsActive(virNetworkPtr net)
{
    virNetworkObjPtr obj;
    int ret = -1;

2658 2659
    if (!(obj = networkObjFromNetwork(net)))
        return ret;
2660 2661 2662 2663

    if (virNetworkIsActiveEnsureACL(net->conn, obj->def) < 0)
        goto cleanup;

2664 2665
    ret = virNetworkObjIsActive(obj);

2666
 cleanup:
2667 2668 2669 2670 2671 2672 2673 2674 2675 2676
    if (obj)
        virNetworkObjUnlock(obj);
    return ret;
}

static int networkIsPersistent(virNetworkPtr net)
{
    virNetworkObjPtr obj;
    int ret = -1;

2677 2678
    if (!(obj = networkObjFromNetwork(net)))
        return ret;
2679 2680 2681 2682

    if (virNetworkIsPersistentEnsureACL(net->conn, obj->def) < 0)
        goto cleanup;

2683 2684
    ret = obj->persistent;

2685
 cleanup:
2686 2687 2688 2689 2690 2691
    if (obj)
        virNetworkObjUnlock(obj);
    return ret;
}


2692
static int
2693
networkValidate(virNetworkDefPtr def,
2694
                bool check_active)
2695
{
2696
    size_t i;
2697 2698
    bool vlanUsed, vlanAllowed, badVlanUse = false;
    virPortGroupDefPtr defaultPortGroup = NULL;
2699
    virNetworkIpDefPtr ipdef;
G
Gene Czarcinski 已提交
2700
    bool ipv4def = false, ipv6def = false;
2701 2702 2703 2704 2705 2706 2707 2708

    /* check for duplicate networks */
    if (virNetworkObjIsDuplicate(&driver->networks, def, check_active) < 0)
        return -1;

    /* Only the three L3 network types that are configured by libvirt
     * need to have a bridge device name / mac address provided
     */
2709 2710 2711
    if (def->forward.type == VIR_NETWORK_FORWARD_NONE ||
        def->forward.type == VIR_NETWORK_FORWARD_NAT ||
        def->forward.type == VIR_NETWORK_FORWARD_ROUTE) {
2712 2713 2714 2715 2716

        if (virNetworkSetBridgeName(&driver->networks, def, 1))
            return -1;

        virNetworkSetBridgeMacAddr(def);
2717 2718
    } else {
        /* They are also the only types that currently support setting
2719 2720
         * a MAC or IP address for the host-side device (bridge), DNS
         * configuration, or network-wide bandwidth limits.
2721
         */
2722 2723 2724 2725 2726 2727 2728 2729
        if (def->mac_specified) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <mac> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
                           virNetworkForwardTypeToString(def->forward.type));
            return -1;
        }
2730 2731 2732 2733 2734
        if (virNetworkDefGetIpByIndex(def, AF_UNSPEC, 0)) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <ip> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
2735
                           virNetworkForwardTypeToString(def->forward.type));
2736 2737
            return -1;
        }
2738
        if (def->dns.ntxts || def->dns.nhosts || def->dns.nsrvs) {
2739 2740 2741 2742
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <dns> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
2743
                           virNetworkForwardTypeToString(def->forward.type));
2744 2745 2746 2747 2748 2749 2750
            return -1;
        }
        if (def->domain) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <domain> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
2751
                           virNetworkForwardTypeToString(def->forward.type));
2752 2753
            return -1;
        }
2754 2755 2756 2757 2758 2759 2760 2761
        if (def->bandwidth) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported network-wide <bandwidth> element "
                             "in network %s with forward mode='%s'"),
                           def->name,
                           virNetworkForwardTypeToString(def->forward.type));
            return -1;
        }
2762 2763
    }

G
Gene Czarcinski 已提交
2764 2765 2766
    /* We only support dhcp on one IPv4 address and
     * on one IPv6 address per defined network
     */
2767 2768 2769
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(def, AF_UNSPEC, i));
         i++) {
G
Gene Czarcinski 已提交
2770 2771 2772 2773 2774
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
            if (ipdef->nranges || ipdef->nhosts) {
                if (ipv4def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("Multiple IPv4 dhcp sections found -- "
2775 2776
                                 "dhcp is supported only for a "
                                 "single IPv4 address on each network"));
G
Gene Czarcinski 已提交
2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793
                    return -1;
                } else {
                    ipv4def = true;
                }
            }
        }
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6)) {
            if (ipdef->nranges || ipdef->nhosts) {
                if (ipv6def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("Multiple IPv6 dhcp sections found -- "
                                 "dhcp is supported only for a "
                                 "single IPv6 address on each network"));
                    return -1;
                } else {
                    ipv6def = true;
                }
2794 2795 2796
            }
        }
    }
2797 2798 2799 2800 2801 2802

    /* The only type of networks that currently support transparent
     * vlan configuration are those using hostdev sr-iov devices from
     * a pool, and those using an Open vSwitch bridge.
     */

2803
    vlanAllowed = ((def->forward.type == VIR_NETWORK_FORWARD_BRIDGE &&
J
Ján Tomko 已提交
2804 2805
                    def->virtPortProfile &&
                    def->virtPortProfile->virtPortType
2806 2807
                    == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) ||
                   def->forward.type == VIR_NETWORK_FORWARD_HOSTDEV);
2808 2809

    vlanUsed = def->vlan.nTags > 0;
2810 2811
    for (i = 0; i < def->nPortGroups; i++) {
        if (vlanUsed || def->portGroups[i].vlan.nTags > 0) {
2812 2813 2814 2815 2816
            /* anyone using this portgroup will get a vlan tag. Verify
             * that they will also be using an openvswitch connection,
             * as that is the only type of network that currently
             * supports a vlan tag.
             */
2817
            if (def->portGroups[i].virtPortProfile) {
2818
                if (def->forward.type != VIR_NETWORK_FORWARD_BRIDGE ||
2819
                    def->portGroups[i].virtPortProfile->virtPortType
2820 2821 2822 2823 2824 2825 2826
                    != VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
                    badVlanUse = true;
                }
            } else if (!vlanAllowed) {
                /* virtualport taken from base network definition */
                badVlanUse = true;
            }
2827
        }
2828
        if (def->portGroups[i].isDefault) {
2829 2830 2831 2832 2833
            if (defaultPortGroup) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("network '%s' has multiple default "
                                 "<portgroup> elements (%s and %s), "
                                 "but only one default is allowed"),
2834
                               def->name, defaultPortGroup->name,
2835
                               def->portGroups[i].name);
2836
                return -1;
2837
            }
2838
            defaultPortGroup = &def->portGroups[i];
2839
        }
2840
    }
2841 2842 2843 2844 2845 2846 2847
    if (badVlanUse ||
        (vlanUsed && !vlanAllowed && !defaultPortGroup)) {
        /* NB: if defaultPortGroup is set, we don't directly look at
         * vlanUsed && !vlanAllowed, because the network will never be
         * used without having a portgroup added in, so all necessary
         * checks were done in the loop above.
         */
2848 2849 2850 2851 2852 2853 2854 2855 2856
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       _("<vlan> element specified for network %s, "
                         "whose type doesn't support vlan configuration"),
                       def->name);
        return -1;
    }
    return 0;
}

2857 2858
static virNetworkPtr networkCreateXML(virConnectPtr conn, const char *xml)
{
2859
    virNetworkDefPtr def;
2860
    virNetworkObjPtr network = NULL;
2861
    virNetworkPtr ret = NULL;
2862
    virObjectEventPtr event = NULL;
2863

2864
    networkDriverLock();
2865

2866
    if (!(def = virNetworkDefParseString(xml)))
2867
        goto cleanup;
2868

2869 2870 2871
    if (virNetworkCreateXMLEnsureACL(conn, def) < 0)
        goto cleanup;

2872
    if (networkValidate(def, true) < 0)
J
Ján Tomko 已提交
2873
        goto cleanup;
2874

2875 2876 2877
    /* NB: even though this transient network hasn't yet been started,
     * we assign the def with live = true in anticipation that it will
     * be started momentarily.
2878
     */
2879
    if (!(network = virNetworkAssignDef(&driver->networks, def, true)))
2880 2881
        goto cleanup;
    def = NULL;
2882

2883
    if (networkStartNetwork(network) < 0) {
2884 2885
        virNetworkRemoveInactive(&driver->networks,
                                 network);
2886
        network = NULL;
2887
        goto cleanup;
2888 2889
    }

2890 2891
    event = virNetworkEventLifecycleNew(network->def->name,
                                        network->def->uuid,
2892 2893
                                        VIR_NETWORK_EVENT_STARTED,
                                        0);
2894

2895
    VIR_INFO("Creating network '%s'", network->def->name);
2896 2897
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

2898
 cleanup:
2899
    virNetworkDefFree(def);
2900 2901
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
2902 2903
    if (network)
        virNetworkObjUnlock(network);
2904
    networkDriverUnlock();
2905
    return ret;
2906 2907
}

2908 2909
static virNetworkPtr networkDefineXML(virConnectPtr conn, const char *xml)
{
2910
    virNetworkDefPtr def = NULL;
2911
    bool freeDef = true;
2912
    virNetworkObjPtr network = NULL;
2913
    virNetworkPtr ret = NULL;
2914
    virObjectEventPtr event = NULL;
2915

2916
    networkDriverLock();
2917

2918
    if (!(def = virNetworkDefParseString(xml)))
2919
        goto cleanup;
2920

2921 2922 2923
    if (virNetworkDefineXMLEnsureACL(conn, def) < 0)
        goto cleanup;

2924
    if (networkValidate(def, false) < 0)
J
Ján Tomko 已提交
2925
        goto cleanup;
2926

2927
    if (!(network = virNetworkAssignDef(&driver->networks, def, false)))
J
Ján Tomko 已提交
2928
        goto cleanup;
2929

2930
    /* def was assigned to network object */
2931
    freeDef = false;
2932 2933

    if (virNetworkSaveConfig(driver->networkConfigDir, def) < 0) {
2934 2935 2936 2937 2938
        if (!virNetworkObjIsActive(network)) {
            virNetworkRemoveInactive(&driver->networks, network);
            network = NULL;
            goto cleanup;
        }
2939 2940 2941 2942 2943
        /* if network was active already, just undo new persistent
         * definition by making it transient.
         * XXX - this isn't necessarily the correct thing to do.
         */
        virNetworkObjAssignDef(network, NULL, false);
2944 2945 2946
        goto cleanup;
    }

2947
    event = virNetworkEventLifecycleNew(def->name, def->uuid,
2948 2949
                                        VIR_NETWORK_EVENT_DEFINED,
                                        0);
2950

2951 2952
    VIR_INFO("Defining network '%s'", def->name);
    ret = virGetNetwork(conn, def->name, def->uuid);
2953

2954
 cleanup:
2955 2956
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
2957
    if (freeDef)
J
Ján Tomko 已提交
2958
        virNetworkDefFree(def);
2959 2960
    if (network)
        virNetworkObjUnlock(network);
2961
    networkDriverUnlock();
2962
    return ret;
2963 2964
}

2965
static int
2966 2967
networkUndefine(virNetworkPtr net)
{
2968
    virNetworkObjPtr network;
2969
    int ret = -1;
2970
    bool active = false;
2971
    virObjectEventPtr event = NULL;
2972

2973
    networkDriverLock();
2974

2975
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
2976
    if (!network) {
2977 2978
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2979
        goto cleanup;
2980 2981
    }

2982 2983 2984
    if (virNetworkUndefineEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2985 2986
    if (virNetworkObjIsActive(network))
        active = true;
2987

2988
    /* remove autostart link */
2989
    if (virNetworkDeleteConfig(driver->networkConfigDir,
2990 2991
                               driver->networkAutostartDir,
                               network) < 0)
2992
        goto cleanup;
2993
    network->autostart = 0;
2994

2995 2996
    event = virNetworkEventLifecycleNew(network->def->name,
                                        network->def->uuid,
2997 2998
                                        VIR_NETWORK_EVENT_UNDEFINED,
                                        0);
2999

3000
    VIR_INFO("Undefining network '%s'", network->def->name);
3001
    if (!active) {
3002
        if (networkRemoveInactive(network) < 0) {
3003 3004 3005
            network = NULL;
            goto cleanup;
        }
3006
        network = NULL;
3007 3008 3009 3010 3011 3012
    } else {

        /* if the network still exists, it was active, and we need to make
         * it transient (by deleting the persistent def)
         */
        virNetworkObjAssignDef(network, NULL, false);
3013 3014
    }

3015
    ret = 0;
3016

3017
 cleanup:
3018 3019
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
3020 3021
    if (network)
        virNetworkObjUnlock(network);
3022
    networkDriverUnlock();
3023
    return ret;
3024 3025
}

3026 3027 3028 3029 3030 3031 3032 3033 3034
static int
networkUpdate(virNetworkPtr net,
              unsigned int command,
              unsigned int section,
              int parentIndex,
              const char *xml,
              unsigned int flags)
{
    virNetworkObjPtr network = NULL;
3035 3036
    int isActive, ret = -1;
    size_t i;
3037 3038
    virNetworkIpDefPtr ipdef;
    bool oldDhcpActive = false;
3039
    bool needFirewallRefresh = false;
3040

3041 3042 3043 3044 3045

    virCheckFlags(VIR_NETWORK_UPDATE_AFFECT_LIVE |
                  VIR_NETWORK_UPDATE_AFFECT_CONFIG,
                  -1);

3046
    networkDriverLock();
3047 3048 3049 3050 3051 3052 3053 3054

    network = virNetworkFindByUUID(&driver->networks, net->uuid);
    if (!network) {
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
        goto cleanup;
    }

3055 3056 3057
    if (virNetworkUpdateEnsureACL(net->conn, network->def, flags) < 0)
        goto cleanup;

3058
    /* see if we are listening for dhcp pre-modification */
3059 3060 3061
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, i));
         i++) {
3062 3063 3064 3065 3066 3067
        if (ipdef->nranges || ipdef->nhosts) {
            oldDhcpActive = true;
            break;
        }
    }

3068 3069
    /* VIR_NETWORK_UPDATE_AFFECT_CURRENT means "change LIVE if network
     * is active, else change CONFIG
J
Ján Tomko 已提交
3070
     */
3071
    isActive = virNetworkObjIsActive(network);
3072 3073
    if ((flags & (VIR_NETWORK_UPDATE_AFFECT_LIVE |
                  VIR_NETWORK_UPDATE_AFFECT_CONFIG)) ==
3074 3075 3076 3077 3078 3079 3080
        VIR_NETWORK_UPDATE_AFFECT_CURRENT) {
        if (isActive)
            flags |= VIR_NETWORK_UPDATE_AFFECT_LIVE;
        else
            flags |= VIR_NETWORK_UPDATE_AFFECT_CONFIG;
    }

3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097
    if (isActive && (flags & VIR_NETWORK_UPDATE_AFFECT_LIVE)) {
        /* Take care of anything that must be done before updating the
         * live NetworkDef.
         */
        if (network->def->forward.type == VIR_NETWORK_FORWARD_NONE ||
            network->def->forward.type == VIR_NETWORK_FORWARD_NAT ||
            network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE) {
            switch (section) {
            case VIR_NETWORK_SECTION_FORWARD:
            case VIR_NETWORK_SECTION_FORWARD_INTERFACE:
            case VIR_NETWORK_SECTION_IP:
            case VIR_NETWORK_SECTION_IP_DHCP_RANGE:
            case VIR_NETWORK_SECTION_IP_DHCP_HOST:
                /* these could affect the firewall rules, so remove the
                 * old rules (and remember to load new ones after the
                 * update).
                 */
3098
                networkRemoveFirewallRules(network->def);
3099 3100 3101 3102 3103 3104 3105 3106
                needFirewallRefresh = true;
                break;
            default:
                break;
            }
        }
    }

3107
    /* update the network config in memory/on disk */
3108 3109
    if (virNetworkObjUpdate(network, command, section, parentIndex, xml, flags) < 0) {
        if (needFirewallRefresh)
3110
            ignore_value(networkAddFirewallRules(network->def));
3111 3112 3113
        goto cleanup;
    }

3114
    if (needFirewallRefresh && networkAddFirewallRules(network->def) < 0)
3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137
        goto cleanup;

    if (flags & VIR_NETWORK_UPDATE_AFFECT_CONFIG) {
        /* save updated persistent config to disk */
        if (virNetworkSaveConfig(driver->networkConfigDir,
                                 virNetworkObjGetPersistentDef(network)) < 0) {
            goto cleanup;
        }
    }

    if (isActive && (flags & VIR_NETWORK_UPDATE_AFFECT_LIVE)) {
        /* rewrite dnsmasq host files, restart dnsmasq, update iptables
         * rules, etc, according to which section was modified. Note that
         * some sections require multiple actions, so a single switch
         * statement is inadequate.
         */
        if (section == VIR_NETWORK_SECTION_BRIDGE ||
            section == VIR_NETWORK_SECTION_DOMAIN ||
            section == VIR_NETWORK_SECTION_IP ||
            section == VIR_NETWORK_SECTION_IP_DHCP_RANGE) {
            /* these sections all change things on the dnsmasq commandline,
             * so we need to kill and restart dnsmasq.
             */
3138
            if (networkRestartDhcpDaemon(network) < 0)
3139 3140
                goto cleanup;

3141 3142 3143 3144 3145 3146 3147 3148
        } else if (section == VIR_NETWORK_SECTION_IP_DHCP_HOST) {
            /* if we previously weren't listening for dhcp and now we
             * are (or vice-versa) then we need to do a restart,
             * otherwise we just need to do a refresh (redo the config
             * files and send SIGHUP)
             */
            bool newDhcpActive = false;

3149 3150 3151
            for (i = 0;
                 (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, i));
                 i++) {
3152 3153 3154 3155 3156 3157 3158
                if (ipdef->nranges || ipdef->nhosts) {
                    newDhcpActive = true;
                    break;
                }
            }

            if ((newDhcpActive != oldDhcpActive &&
3159 3160
                 networkRestartDhcpDaemon(network) < 0) ||
                networkRefreshDhcpDaemon(network) < 0) {
3161 3162 3163 3164
                goto cleanup;
            }

        } else if (section == VIR_NETWORK_SECTION_DNS_HOST ||
3165 3166 3167 3168 3169 3170
                   section == VIR_NETWORK_SECTION_DNS_TXT ||
                   section == VIR_NETWORK_SECTION_DNS_SRV) {
            /* these sections only change things in config files, so we
             * can just update the config files and send SIGHUP to
             * dnsmasq.
             */
3171
            if (networkRefreshDhcpDaemon(network) < 0)
3172 3173 3174 3175 3176 3177 3178 3179
                goto cleanup;

        }

        if (section == VIR_NETWORK_SECTION_IP) {
            /* only a change in IP addresses will affect radvd, and all of radvd's
             * config is stored in the conf file which will be re-read with a SIGHUP.
             */
3180
            if (networkRefreshRadvd(network) < 0)
3181 3182 3183 3184
                goto cleanup;
        }

        /* save current network state to disk */
3185
        if ((ret = virNetworkSaveStatus(driver->stateDir,
3186
                                        network)) < 0) {
3187
            goto cleanup;
3188
        }
3189 3190
    }
    ret = 0;
3191
 cleanup:
3192 3193
    if (network)
        virNetworkObjUnlock(network);
3194
    networkDriverUnlock();
3195 3196 3197
    return ret;
}

3198 3199
static int networkCreate(virNetworkPtr net)
{
3200 3201
    virNetworkObjPtr network;
    int ret = -1;
3202
    virObjectEventPtr event = NULL;
3203

3204
    networkDriverLock();
3205
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3206

3207
    if (!network) {
3208 3209
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
3210
        goto cleanup;
3211 3212
    }

3213 3214 3215
    if (virNetworkCreateEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

3216
    if ((ret = networkStartNetwork(network)) < 0)
3217
        goto cleanup;
3218

3219 3220
    event = virNetworkEventLifecycleNew(network->def->name,
                                        network->def->uuid,
3221 3222
                                        VIR_NETWORK_EVENT_STARTED,
                                        0);
3223

3224
 cleanup:
3225 3226
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
3227 3228
    if (network)
        virNetworkObjUnlock(network);
3229
    networkDriverUnlock();
3230
    return ret;
3231 3232
}

3233 3234
static int networkDestroy(virNetworkPtr net)
{
3235 3236
    virNetworkObjPtr network;
    int ret = -1;
3237
    virObjectEventPtr event = NULL;
3238

3239
    networkDriverLock();
3240
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3241

3242
    if (!network) {
3243 3244
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
3245
        goto cleanup;
3246 3247
    }

3248 3249 3250
    if (virNetworkDestroyEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
3251
    if (!virNetworkObjIsActive(network)) {
3252
        virReportError(VIR_ERR_OPERATION_INVALID,
3253 3254
                       _("network '%s' is not active"),
                       network->def->name);
3255 3256 3257
        goto cleanup;
    }

3258
    if ((ret = networkShutdownNetwork(network)) < 0)
3259 3260
        goto cleanup;

3261 3262
    event = virNetworkEventLifecycleNew(network->def->name,
                                        network->def->uuid,
3263 3264
                                        VIR_NETWORK_EVENT_STOPPED,
                                        0);
3265

3266
    if (!network->persistent) {
3267
        if (networkRemoveInactive(network) < 0) {
3268 3269 3270 3271
            network = NULL;
            ret = -1;
            goto cleanup;
        }
3272 3273
        network = NULL;
    }
3274

3275
 cleanup:
3276 3277
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
3278 3279
    if (network)
        virNetworkObjUnlock(network);
3280
    networkDriverUnlock();
3281 3282 3283
    return ret;
}

3284
static char *networkGetXMLDesc(virNetworkPtr net,
3285
                               unsigned int flags)
3286
{
3287
    virNetworkObjPtr network;
3288
    virNetworkDefPtr def;
3289
    char *ret = NULL;
3290

3291
    virCheckFlags(VIR_NETWORK_XML_INACTIVE, NULL);
3292

3293 3294
    if (!(network = networkObjFromNetwork(net)))
        return ret;
3295

3296 3297 3298
    if (virNetworkGetXMLDescEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

3299 3300 3301 3302 3303 3304
    if ((flags & VIR_NETWORK_XML_INACTIVE) && network->newDef)
        def = network->newDef;
    else
        def = network->def;

    ret = virNetworkDefFormat(def, flags);
3305

3306
 cleanup:
3307 3308
    if (network)
        virNetworkObjUnlock(network);
3309
    return ret;
3310 3311 3312
}

static char *networkGetBridgeName(virNetworkPtr net) {
3313 3314 3315
    virNetworkObjPtr network;
    char *bridge = NULL;

3316 3317
    if (!(network = networkObjFromNetwork(net)))
        return bridge;
3318

3319 3320 3321
    if (virNetworkGetBridgeNameEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

3322
    if (!(network->def->bridge)) {
3323 3324 3325
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("network '%s' does not have a bridge name."),
                       network->def->name);
3326 3327 3328
        goto cleanup;
    }

3329
    ignore_value(VIR_STRDUP(bridge, network->def->bridge));
3330

3331
 cleanup:
3332 3333
    if (network)
        virNetworkObjUnlock(network);
3334 3335 3336 3337
    return bridge;
}

static int networkGetAutostart(virNetworkPtr net,
J
Ján Tomko 已提交
3338
                               int *autostart)
3339
{
3340 3341
    virNetworkObjPtr network;
    int ret = -1;
3342

3343 3344
    if (!(network = networkObjFromNetwork(net)))
        return ret;
3345

3346 3347 3348
    if (virNetworkGetAutostartEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

3349
    *autostart = network->autostart;
3350
    ret = 0;
3351

3352
 cleanup:
3353 3354
    if (network)
        virNetworkObjUnlock(network);
3355
    return ret;
3356 3357 3358
}

static int networkSetAutostart(virNetworkPtr net,
3359 3360
                               int autostart)
{
3361
    virNetworkObjPtr network;
3362
    char *configFile = NULL, *autostartLink = NULL;
3363
    int ret = -1;
3364

3365
    networkDriverLock();
3366
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3367

3368
    if (!network) {
3369 3370
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
3371
        goto cleanup;
3372 3373
    }

3374 3375 3376
    if (virNetworkSetAutostartEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

3377
    if (!network->persistent) {
3378 3379
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("cannot set autostart for transient network"));
3380 3381 3382
        goto cleanup;
    }

3383 3384
    autostart = (autostart != 0);

3385
    if (network->autostart != autostart) {
3386
        if ((configFile = virNetworkConfigFile(driver->networkConfigDir, network->def->name)) == NULL)
3387
            goto cleanup;
3388
        if ((autostartLink = virNetworkConfigFile(driver->networkAutostartDir, network->def->name)) == NULL)
3389 3390
            goto cleanup;

3391
        if (autostart) {
3392
            if (virFileMakePath(driver->networkAutostartDir) < 0) {
3393
                virReportSystemError(errno,
3394 3395
                                     _("cannot create autostart directory '%s'"),
                                     driver->networkAutostartDir);
3396 3397
                goto cleanup;
            }
3398

3399
            if (symlink(configFile, autostartLink) < 0) {
3400
                virReportSystemError(errno,
3401
                                     _("Failed to create symlink '%s' to '%s'"),
3402
                                     autostartLink, configFile);
3403 3404 3405
                goto cleanup;
            }
        } else {
3406
            if (unlink(autostartLink) < 0 && errno != ENOENT && errno != ENOTDIR) {
3407
                virReportSystemError(errno,
3408
                                     _("Failed to delete symlink '%s'"),
3409
                                     autostartLink);
3410 3411
                goto cleanup;
            }
3412 3413
        }

3414
        network->autostart = autostart;
3415
    }
3416
    ret = 0;
3417

3418
 cleanup:
3419 3420
    VIR_FREE(configFile);
    VIR_FREE(autostartLink);
3421 3422
    if (network)
        virNetworkObjUnlock(network);
3423
    networkDriverUnlock();
3424
    return ret;
3425 3426
}

3427
static int
3428 3429 3430 3431
networkGetDHCPLeases(virNetworkPtr network,
                     const char *mac,
                     virNetworkDHCPLeasePtr **leases,
                     unsigned int flags)
3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450
{
    size_t i, j;
    size_t nleases = 0;
    int rv = -1;
    int size = 0;
    int custom_lease_file_len = 0;
    bool need_results = !!leases;
    long long currtime = 0;
    long long expirytime_tmp = -1;
    bool ipv6 = false;
    char *lease_entries = NULL;
    char *custom_lease_file = NULL;
    const char *ip_tmp = NULL;
    const char *mac_tmp = NULL;
    virJSONValuePtr lease_tmp = NULL;
    virJSONValuePtr leases_array = NULL;
    virNetworkIpDefPtr ipdef_tmp = NULL;
    virNetworkDHCPLeasePtr lease = NULL;
    virNetworkDHCPLeasePtr *leases_ret = NULL;
3451 3452 3453 3454 3455 3456 3457 3458 3459
    virNetworkObjPtr obj;

    virCheckFlags(0, -1);

    if (!(obj = networkObjFromNetwork(network)))
        return -1;

    if (virNetworkGetDHCPLeasesEnsureACL(network->conn, obj->def) < 0)
        goto cleanup;
3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505

    /* Retrieve custom leases file location */
    custom_lease_file = networkDnsmasqLeaseFileNameCustom(obj->def->bridge);

    /* Read entire contents */
    if ((custom_lease_file_len = virFileReadAll(custom_lease_file,
                                                VIR_NETWORK_DHCP_LEASE_FILE_SIZE_MAX,
                                                &lease_entries)) < 0) {
        /* Even though src/network/leaseshelper.c guarantees the existence of
         * leases file (even if no leases are present), and the control reaches
         * here, instead of reporting error, return 0 leases */
        rv = 0;
        goto error;
    }

    if (custom_lease_file_len) {
        if (!(leases_array = virJSONValueFromString(lease_entries))) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("invalid json in file: %s"), custom_lease_file);
            goto error;
        }

        if ((size = virJSONValueArraySize(leases_array)) < 0) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("couldn't fetch array of leases"));
            goto error;
        }
    }

    currtime = (long long) time(NULL);

    for (i = 0; i < size; i++) {
        if (!(lease_tmp = virJSONValueArrayGet(leases_array, i))) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("failed to parse json"));
            goto error;
        }

        if (!(mac_tmp = virJSONValueObjectGetString(lease_tmp, "mac-address"))) {
            /* leaseshelper program guarantees that lease will be stored only if
             * mac-address is known otherwise not */
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("found lease without mac-address"));
            goto error;
        }

3506
        if (mac && virMacAddrCompare(mac, mac_tmp))
3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556
            continue;

        if (virJSONValueObjectGetNumberLong(lease_tmp, "expiry-time", &expirytime_tmp) < 0) {
            /* A lease cannot be present without expiry-time */
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("found lease without expiry-time"));
            goto error;
        }

        /* Do not report expired lease */
        if (expirytime_tmp < currtime)
            continue;

        if (need_results) {
            if (VIR_ALLOC(lease) < 0)
                goto error;

            lease->expirytime = expirytime_tmp;

            if (!(ip_tmp = virJSONValueObjectGetString(lease_tmp, "ip-address"))) {
                /* A lease without ip-address makes no sense */
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("found lease without ip-address"));
                goto error;
            }

            /* Unlike IPv4, IPv6 uses ':' instead of '.' as separator */
            ipv6 = strchr(ip_tmp, ':') ? true : false;
            lease->type = ipv6 ? VIR_IP_ADDR_TYPE_IPV6 : VIR_IP_ADDR_TYPE_IPV4;

            /* Obtain prefix */
            for (j = 0; j < obj->def->nips; j++) {
                ipdef_tmp = &obj->def->ips[j];

                if (ipv6 && VIR_SOCKET_ADDR_IS_FAMILY(&ipdef_tmp->address,
                                                      AF_INET6)) {
                    lease->prefix = ipdef_tmp->prefix;
                    break;
                }
                if (!ipv6 && VIR_SOCKET_ADDR_IS_FAMILY(&ipdef_tmp->address,
                                                      AF_INET)) {
                    lease->prefix = virSocketAddrGetIpPrefix(&ipdef_tmp->address,
                                                             &ipdef_tmp->netmask,
                                                             ipdef_tmp->prefix);
                    break;
                }
            }

            if ((VIR_STRDUP(lease->mac, mac_tmp) < 0) ||
                (VIR_STRDUP(lease->ipaddr, ip_tmp) < 0) ||
3557
                (VIR_STRDUP(lease->iface, obj->def->bridge) < 0))
3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595
                goto error;

            /* Fields that can be NULL */
            if ((VIR_STRDUP(lease->iaid,
                            virJSONValueObjectGetString(lease_tmp, "iaid")) < 0) ||
                (VIR_STRDUP(lease->clientid,
                            virJSONValueObjectGetString(lease_tmp, "client-id")) < 0) ||
                (VIR_STRDUP(lease->hostname,
                            virJSONValueObjectGetString(lease_tmp, "hostname")) < 0))
                goto error;

            if (VIR_INSERT_ELEMENT(leases_ret, nleases, nleases, lease) < 0)
                goto error;

        } else {
            nleases++;
        }

        VIR_FREE(lease);
    }

    if (need_results && mac && !leases_ret) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("no lease with matching MAC address: %s"), mac);
        goto error;
    }

    if (leases_ret) {
        /* NULL terminated array */
        ignore_value(VIR_REALLOC_N(leases_ret, nleases + 1));
        *leases = leases_ret;
        leases_ret = NULL;
    }

    rv = nleases;

 cleanup:
    VIR_FREE(lease);
3596
    VIR_FREE(lease_entries);
3597 3598
    VIR_FREE(custom_lease_file);
    virJSONValueFree(leases_array);
3599 3600 3601 3602

    if (obj)
        virNetworkObjUnlock(obj);

3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613
    return rv;

 error:
    if (leases_ret) {
        for (i = 0; i < nleases; i++)
            virNetworkDHCPLeaseFree(leases_ret[i]);
        VIR_FREE(leases_ret);
    }
    goto cleanup;
}

3614 3615 3616

static virNetworkDriver networkDriver = {
    "Network",
3617 3618 3619 3620 3621 3622 3623
    .networkOpen = networkOpen, /* 0.2.0 */
    .networkClose = networkClose, /* 0.2.0 */
    .connectNumOfNetworks = networkConnectNumOfNetworks, /* 0.2.0 */
    .connectListNetworks = networkConnectListNetworks, /* 0.2.0 */
    .connectNumOfDefinedNetworks = networkConnectNumOfDefinedNetworks, /* 0.2.0 */
    .connectListDefinedNetworks = networkConnectListDefinedNetworks, /* 0.2.0 */
    .connectListAllNetworks = networkConnectListAllNetworks, /* 0.10.2 */
3624 3625
    .connectNetworkEventRegisterAny = networkConnectNetworkEventRegisterAny, /* 1.2.1 */
    .connectNetworkEventDeregisterAny = networkConnectNetworkEventDeregisterAny, /* 1.2.1 */
3626 3627
    .networkLookupByUUID = networkLookupByUUID, /* 0.2.0 */
    .networkLookupByName = networkLookupByName, /* 0.2.0 */
3628 3629
    .networkCreateXML = networkCreateXML, /* 0.2.0 */
    .networkDefineXML = networkDefineXML, /* 0.2.0 */
3630
    .networkUndefine = networkUndefine, /* 0.2.0 */
3631
    .networkUpdate = networkUpdate, /* 0.10.2 */
3632
    .networkCreate = networkCreate, /* 0.2.0 */
3633 3634 3635 3636 3637 3638 3639
    .networkDestroy = networkDestroy, /* 0.2.0 */
    .networkGetXMLDesc = networkGetXMLDesc, /* 0.2.0 */
    .networkGetBridgeName = networkGetBridgeName, /* 0.2.0 */
    .networkGetAutostart = networkGetAutostart, /* 0.2.1 */
    .networkSetAutostart = networkSetAutostart, /* 0.2.1 */
    .networkIsActive = networkIsActive, /* 0.7.3 */
    .networkIsPersistent = networkIsPersistent, /* 0.7.3 */
3640
    .networkGetDHCPLeases = networkGetDHCPLeases, /* 1.2.6 */
3641 3642 3643
};

static virStateDriver networkStateDriver = {
3644
    .name = "Network",
3645
    .stateInitialize  = networkStateInitialize,
3646
    .stateAutoStart  = networkStateAutoStart,
3647 3648
    .stateCleanup = networkStateCleanup,
    .stateReload = networkStateReload,
3649 3650
};

3651 3652
int networkRegister(void)
{
3653 3654
    if (virRegisterNetworkDriver(&networkDriver) < 0)
        return -1;
3655 3656
    if (virRegisterStateDriver(&networkStateDriver) < 0)
        return -1;
3657 3658
    return 0;
}
3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670

/********************************************************/

/* Private API to deal with logical switch capabilities.
 * These functions are exported so that other parts of libvirt can
 * call them, but are not part of the public API and not in the
 * driver's function table. If we ever have more than one network
 * driver, we will need to present these functions via a second
 * "backend" function table.
 */

/* networkAllocateActualDevice:
3671
 * @dom: domain definition that @iface belongs to
3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682
 * @iface: the original NetDef from the domain
 *
 * Looks up the network reference by iface, allocates a physical
 * device from that network (if appropriate), and returns with the
 * virDomainActualNetDef filled in accordingly. If there are no
 * changes to be made in the netdef, then just leave the actualdef
 * empty.
 *
 * Returns 0 on success, -1 on failure.
 */
int
3683 3684
networkAllocateActualDevice(virDomainDefPtr dom,
                            virDomainNetDefPtr iface)
3685
{
3686
    virDomainNetType actualType = iface->type;
3687 3688
    virNetworkObjPtr network = NULL;
    virNetworkDefPtr netdef = NULL;
3689
    virNetDevBandwidthPtr bandwidth = NULL;
3690 3691 3692
    virPortGroupDefPtr portgroup = NULL;
    virNetDevVPortProfilePtr virtport = iface->virtPortProfile;
    virNetDevVlanPtr vlan = NULL;
3693
    virNetworkForwardIfDefPtr dev = NULL;
3694
    size_t i;
3695 3696 3697
    int ret = -1;

    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
3698
        goto validate;
3699 3700 3701 3702

    virDomainActualNetDefFree(iface->data.network.actual);
    iface->data.network.actual = NULL;

3703
    networkDriverLock();
3704
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
3705
    networkDriverUnlock();
3706
    if (!network) {
3707 3708 3709
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
3710
        goto error;
3711 3712
    }
    netdef = network->def;
3713

3714 3715 3716 3717 3718 3719 3720
    if (!virNetworkObjIsActive(network)) {
        virReportError(VIR_ERR_OPERATION_INVALID,
                       _("network '%s' is not active"),
                       netdef->name);
        goto error;
    }

3721 3722 3723
    if (VIR_ALLOC(iface->data.network.actual) < 0)
        goto error;

3724 3725 3726
    /* portgroup can be present for any type of network, in particular
     * for bandwidth information, so we need to check for that and
     * fill it in appropriately for all forward types.
J
Ján Tomko 已提交
3727
     */
3728 3729 3730 3731 3732 3733
    portgroup = virPortGroupFindByName(netdef, iface->data.network.portgroup);

    /* If there is already interface-specific bandwidth, just use that
     * (already in NetDef). Otherwise, if there is bandwidth info in
     * the portgroup, fill that into the ActualDef.
     */
3734 3735 3736 3737 3738 3739

    if (iface->bandwidth)
        bandwidth = iface->bandwidth;
    else if (portgroup && portgroup->bandwidth)
        bandwidth = portgroup->bandwidth;

3740 3741
    if (bandwidth && virNetDevBandwidthCopy(&iface->data.network.actual->bandwidth,
                                            bandwidth) < 0)
3742
        goto error;
3743

3744 3745 3746 3747 3748 3749 3750 3751
    /* copy appropriate vlan info to actualNet */
    if (iface->vlan.nTags > 0)
        vlan = &iface->vlan;
    else if (portgroup && portgroup->vlan.nTags > 0)
        vlan = &portgroup->vlan;
    else if (netdef->vlan.nTags > 0)
        vlan = &netdef->vlan;

3752 3753
    if (vlan && virNetDevVlanCopy(&iface->data.network.actual->vlan, vlan) < 0)
        goto error;
3754

3755 3756 3757 3758 3759 3760 3761 3762 3763 3764
    if (iface->trustGuestRxFilters)
       iface->data.network.actual->trustGuestRxFilters
          = iface->trustGuestRxFilters;
    else if (portgroup && portgroup->trustGuestRxFilters)
       iface->data.network.actual->trustGuestRxFilters
          = portgroup->trustGuestRxFilters;
    else if (netdef->trustGuestRxFilters)
       iface->data.network.actual->trustGuestRxFilters
          = netdef->trustGuestRxFilters;

3765 3766 3767
    if ((netdef->forward.type == VIR_NETWORK_FORWARD_NONE) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_NAT) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE)) {
3768 3769 3770
        /* for these forward types, the actual net type really *is*
         *NETWORK; we just keep the info from the portgroup in
         * iface->data.network.actual
J
Ján Tomko 已提交
3771
         */
3772
        iface->data.network.actual->type = VIR_DOMAIN_NET_TYPE_NETWORK;
3773 3774 3775 3776

        if (networkPlugBandwidth(network, iface) < 0)
            goto error;

3777
    } else if ((netdef->forward.type == VIR_NETWORK_FORWARD_BRIDGE) &&
3778
               netdef->bridge) {
3779 3780 3781 3782 3783

        /* <forward type='bridge'/> <bridge name='xxx'/>
         * is VIR_DOMAIN_NET_TYPE_BRIDGE
         */

3784
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_BRIDGE;
3785 3786
        if (VIR_STRDUP(iface->data.network.actual->data.bridge.brname,
                       netdef->bridge) < 0)
3787
            goto error;
3788

3789 3790 3791 3792 3793 3794 3795 3796
        /* merge virtualports from interface, network, and portgroup to
         * arrive at actual virtualport to use
         */
        if (virNetDevVPortProfileMerge3(&iface->data.network.actual->virtPortProfile,
                                        iface->virtPortProfile,
                                        netdef->virtPortProfile,
                                        portgroup
                                        ? portgroup->virtPortProfile : NULL) < 0) {
3797
            goto error;
3798 3799 3800 3801 3802 3803 3804 3805 3806 3807
        }
        virtport = iface->data.network.actual->virtPortProfile;
        if (virtport) {
            /* only type='openvswitch' is allowed for bridges */
            if (virtport->virtPortType != VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("<virtualport type='%s'> not supported for network "
                                 "'%s' which uses a bridge device"),
                               virNetDevVPortTypeToString(virtport->virtPortType),
                               netdef->name);
3808
                goto error;
3809 3810 3811
            }
        }

3812
    } else if (netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
3813

3814
        virDomainHostdevSubsysPCIBackendType backend;
3815

3816
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_HOSTDEV;
3817
        if (networkCreateInterfacePool(netdef) < 0)
3818 3819 3820
            goto error;

        /* pick first dev with 0 connections */
3821 3822 3823
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].connections == 0) {
                dev = &netdef->forward.ifs[i];
3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837
                break;
            }
        }
        if (!dev) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' requires exclusive access "
                             "to interfaces, but none are available"),
                           netdef->name);
            goto error;
        }
        iface->data.network.actual->data.hostdev.def.parent.type = VIR_DOMAIN_DEVICE_NET;
        iface->data.network.actual->data.hostdev.def.parent.data.net = iface;
        iface->data.network.actual->data.hostdev.def.info = &iface->info;
        iface->data.network.actual->data.hostdev.def.mode = VIR_DOMAIN_HOSTDEV_MODE_SUBSYS;
3838
        iface->data.network.actual->data.hostdev.def.managed = netdef->forward.managed ? 1 : 0;
3839
        iface->data.network.actual->data.hostdev.def.source.subsys.type = dev->type;
3840
        iface->data.network.actual->data.hostdev.def.source.subsys.u.pci.addr = dev->device.pci;
3841

E
Eric Blake 已提交
3842
        switch (netdef->forward.driverName) {
3843
        case VIR_NETWORK_FORWARD_DRIVER_NAME_DEFAULT:
3844
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_DEFAULT;
3845 3846
            break;
        case VIR_NETWORK_FORWARD_DRIVER_NAME_KVM:
3847
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_KVM;
3848 3849
            break;
        case VIR_NETWORK_FORWARD_DRIVER_NAME_VFIO:
3850
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_VFIO;
3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861
            break;
        default:
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unrecognized driver name value %d "
                             " in network '%s'"),
                           netdef->forward.driverName, netdef->name);
            goto error;
        }
        iface->data.network.actual->data.hostdev.def.source.subsys.u.pci.backend
            = backend;

3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886
        /* merge virtualports from interface, network, and portgroup to
         * arrive at actual virtualport to use
         */
        if (virNetDevVPortProfileMerge3(&iface->data.network.actual->virtPortProfile,
                                        iface->virtPortProfile,
                                        netdef->virtPortProfile,
                                        portgroup
                                        ? portgroup->virtPortProfile : NULL) < 0) {
            goto error;
        }
        virtport = iface->data.network.actual->virtPortProfile;
        if (virtport) {
            /* make sure type is supported for hostdev connections */
            if (virtport->virtPortType != VIR_NETDEV_VPORT_PROFILE_8021QBG &&
                virtport->virtPortType != VIR_NETDEV_VPORT_PROFILE_8021QBH) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("<virtualport type='%s'> not supported for network "
                                 "'%s' which uses an SR-IOV Virtual Function "
                                 "via PCI passthrough"),
                               virNetDevVPortTypeToString(virtport->virtPortType),
                               netdef->name);
                goto error;
            }
        }

3887 3888 3889 3890
    } else if ((netdef->forward.type == VIR_NETWORK_FORWARD_BRIDGE) ||
               (netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) ||
               (netdef->forward.type == VIR_NETWORK_FORWARD_VEPA) ||
               (netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH)) {
3891 3892 3893 3894 3895 3896

        /* <forward type='bridge|private|vepa|passthrough'> are all
         * VIR_DOMAIN_NET_TYPE_DIRECT.
         */

        /* Set type=direct and appropriate <source mode='xxx'/> */
3897
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_DIRECT;
3898
        switch (netdef->forward.type) {
3899
        case VIR_NETWORK_FORWARD_BRIDGE:
3900
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_BRIDGE;
3901 3902
            break;
        case VIR_NETWORK_FORWARD_PRIVATE:
3903
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PRIVATE;
3904 3905
            break;
        case VIR_NETWORK_FORWARD_VEPA:
3906
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_VEPA;
3907 3908
            break;
        case VIR_NETWORK_FORWARD_PASSTHROUGH:
3909
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PASSTHRU;
3910 3911 3912
            break;
        }

3913 3914 3915 3916 3917 3918 3919 3920
        /* merge virtualports from interface, network, and portgroup to
         * arrive at actual virtualport to use
         */
        if (virNetDevVPortProfileMerge3(&iface->data.network.actual->virtPortProfile,
                                        iface->virtPortProfile,
                                        netdef->virtPortProfile,
                                        portgroup
                                        ? portgroup->virtPortProfile : NULL) < 0) {
3921
            goto error;
3922
        }
3923
        virtport = iface->data.network.actual->virtPortProfile;
3924
        if (virtport) {
3925 3926 3927 3928 3929 3930 3931 3932
            /* make sure type is supported for macvtap connections */
            if (virtport->virtPortType != VIR_NETDEV_VPORT_PROFILE_8021QBG &&
                virtport->virtPortType != VIR_NETDEV_VPORT_PROFILE_8021QBH) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("<virtualport type='%s'> not supported for network "
                                 "'%s' which uses a macvtap device"),
                               virNetDevVPortTypeToString(virtport->virtPortType),
                               netdef->name);
3933
                goto error;
3934 3935
            }
        }
3936

3937 3938 3939
        /* If there is only a single device, just return it (caller will detect
         * any error if exclusive use is required but could not be acquired).
         */
3940
        if ((netdef->forward.nifs <= 0) && (netdef->forward.npfs <= 0)) {
3941 3942 3943 3944
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' uses a direct mode, but "
                             "has no forward dev and no interface pool"),
                           netdef->name);
3945
            goto error;
3946 3947 3948
        } else {
            /* pick an interface from the pool */

3949
            if (networkCreateInterfacePool(netdef) < 0)
3950 3951
                goto error;

3952 3953 3954 3955 3956
            /* PASSTHROUGH mode, and PRIVATE Mode + 802.1Qbh both
             * require exclusive access to a device, so current
             * connections count must be 0.  Other modes can share, so
             * just search for the one with the lowest number of
             * connections.
3957
             */
3958 3959
            if ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
                ((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
3960 3961 3962
                 iface->data.network.actual->virtPortProfile &&
                 (iface->data.network.actual->virtPortProfile->virtPortType
                  == VIR_NETDEV_VPORT_PROFILE_8021QBH))) {
3963

3964
                /* pick first dev with 0 connections */
3965 3966 3967
                for (i = 0; i < netdef->forward.nifs; i++) {
                    if (netdef->forward.ifs[i].connections == 0) {
                        dev = &netdef->forward.ifs[i];
3968 3969 3970 3971 3972
                        break;
                    }
                }
            } else {
                /* pick least used dev */
3973
                dev = &netdef->forward.ifs[0];
3974 3975 3976
                for (i = 1; i < netdef->forward.nifs; i++) {
                    if (netdef->forward.ifs[i].connections < dev->connections)
                        dev = &netdef->forward.ifs[i];
3977 3978 3979 3980
                }
            }
            /* dev points at the physical device we want to use */
            if (!dev) {
3981 3982 3983 3984
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("network '%s' requires exclusive access "
                                 "to interfaces, but none are available"),
                               netdef->name);
3985
                goto error;
3986
            }
3987 3988
            if (VIR_STRDUP(iface->data.network.actual->data.direct.linkdev,
                           dev->device.dev) < 0)
3989
                goto error;
3990 3991 3992
        }
    }

3993
    if (virNetDevVPortProfileCheckComplete(virtport, true) < 0)
3994
        goto error;
3995

3996
 validate:
3997 3998 3999 4000 4001
    /* make sure that everything now specified for the device is
     * actually supported on this type of network. NB: network,
     * netdev, and iface->data.network.actual may all be NULL.
     */

4002
    if (virDomainNetGetActualVlan(iface)) {
4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031
        /* vlan configuration via libvirt is only supported for
         * PCI Passthrough SR-IOV devices and openvswitch bridges.
         * otherwise log an error and fail
         */
        if (!(actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV ||
              (actualType == VIR_DOMAIN_NET_TYPE_BRIDGE &&
               virtport && virtport->virtPortType
               == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH))) {
            if (netdef) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("an interface connecting to network '%s' "
                                 "is requesting a vlan tag, but that is not "
                                 "supported for this type of network"),
                               netdef->name);
            } else {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("an interface of type '%s' "
                                 "is requesting a vlan tag, but that is not "
                                 "supported for this type of connection"),
                               virDomainNetTypeToString(iface->type));
            }
            goto error;
        }
    }

    if (netdef) {
        netdef->connections++;
        VIR_DEBUG("Using network %s, %d connections",
                  netdef->name, netdef->connections);
4032

4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051
        if (dev) {
            /* mark the allocation */
            dev->connections++;
            if (actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV) {
                VIR_DEBUG("Using physical device %s, %d connections",
                          dev->device.dev, dev->connections);
            } else {
                VIR_DEBUG("Using physical device %04x:%02x:%02x.%x, connections %d",
                          dev->device.pci.domain, dev->device.pci.bus,
                          dev->device.pci.slot, dev->device.pci.function,
                          dev->connections);
            }
        }

        /* finally we can call the 'plugged' hook script if any */
        if (networkRunHook(network, dom, iface,
                           VIR_HOOK_NETWORK_OP_IFACE_PLUGGED,
                           VIR_HOOK_SUBOP_BEGIN) < 0) {
            /* adjust for failure */
4052
            netdef->connections--;
4053 4054 4055 4056
            if (dev)
                dev->connections--;
            goto error;
        }
4057 4058
    }

4059
    ret = 0;
4060

4061
 cleanup:
4062 4063
    if (network)
        virNetworkObjUnlock(network);
4064 4065
    return ret;

4066
 error:
4067
    if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
4068 4069 4070
        virDomainActualNetDefFree(iface->data.network.actual);
        iface->data.network.actual = NULL;
    }
4071
    goto cleanup;
4072 4073 4074
}

/* networkNotifyActualDevice:
4075
 * @dom: domain definition that @iface belongs to
4076 4077 4078 4079 4080 4081 4082 4083 4084 4085
 * @iface:  the domain's NetDef with an "actual" device already filled in.
 *
 * Called to notify the network driver when libvirtd is restarted and
 * finds an already running domain. If appropriate it will force an
 * allocation of the actual->direct.linkdev to get everything back in
 * order.
 *
 * Returns 0 on success, -1 on failure.
 */
int
4086 4087
networkNotifyActualDevice(virDomainDefPtr dom,
                          virDomainNetDefPtr iface)
4088
{
4089
    virDomainNetType actualType = virDomainNetGetActualType(iface);
4090 4091
    virNetworkObjPtr network;
    virNetworkDefPtr netdef;
4092
    virNetworkForwardIfDefPtr dev = NULL;
4093 4094
    size_t i;
    int ret = -1;
4095 4096

    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
J
Ján Tomko 已提交
4097
        return 0;
4098

4099
    networkDriverLock();
4100
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
4101
    networkDriverUnlock();
4102
    if (!network) {
4103 4104 4105
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
4106 4107 4108 4109 4110
        goto error;
    }
    netdef = network->def;

    if (!iface->data.network.actual ||
4111 4112
        (actualType != VIR_DOMAIN_NET_TYPE_DIRECT &&
         actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV)) {
4113 4114
        VIR_DEBUG("Nothing to claim from network %s", iface->data.network.name);
        goto success;
4115 4116
    }

4117
    if (networkCreateInterfacePool(netdef) < 0)
4118
        goto error;
4119

4120
    if (netdef->forward.nifs == 0) {
4121
        virReportError(VIR_ERR_INTERNAL_ERROR,
4122 4123
                       _("network '%s' uses a direct or hostdev mode, "
                         "but has no forward dev and no interface pool"),
4124
                       netdef->name);
4125
        goto error;
4126
    }
4127

4128 4129
    if (actualType == VIR_DOMAIN_NET_TYPE_DIRECT) {
        const char *actualDev;
4130

4131 4132 4133 4134 4135 4136 4137 4138 4139
        actualDev = virDomainNetGetActualDirectDev(iface);
        if (!actualDev) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("the interface uses a direct mode, "
                             "but has no source dev"));
            goto error;
        }

        /* find the matching interface and increment its connections */
4140 4141
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
4142
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
4143 4144
                STREQ(actualDev, netdef->forward.ifs[i].device.dev)) {
                dev = &netdef->forward.ifs[i];
4145 4146 4147 4148 4149
                break;
            }
        }
        /* dev points at the physical device we want to use */
        if (!dev) {
4150
            virReportError(VIR_ERR_INTERNAL_ERROR,
4151 4152
                           _("network '%s' doesn't have dev='%s' "
                             "in use by domain"),
4153
                           netdef->name, actualDev);
4154
            goto error;
4155 4156
        }

4157
        /* PASSTHROUGH mode and PRIVATE Mode + 802.1Qbh both require
4158 4159
         * exclusive access to a device, so current connections count
         * must be 0 in those cases.
4160
         */
4161
        if ((dev->connections > 0) &&
4162 4163
            ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
             ((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
4164 4165
              iface->data.network.actual->virtPortProfile &&
              (iface->data.network.actual->virtPortProfile->virtPortType
4166
               == VIR_NETDEV_VPORT_PROFILE_8021QBH)))) {
4167
            virReportError(VIR_ERR_INTERNAL_ERROR,
4168 4169
                           _("network '%s' claims dev='%s' is already in "
                             "use by a different domain"),
4170
                           netdef->name, actualDev);
4171
            goto error;
4172
        }
4173

4174
        /* we are now assured of success, so mark the allocation */
4175
        dev->connections++;
4176
        VIR_DEBUG("Using physical device %s, connections %d",
4177
                  dev->device.dev, dev->connections);
4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190

    }  else /* if (actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV) */ {
        virDomainHostdevDefPtr hostdev;

        hostdev = virDomainNetGetActualHostdev(iface);
        if (!hostdev) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("the interface uses a hostdev mode, "
                             "but has no hostdev"));
            goto error;
        }

        /* find the matching interface and increment its connections */
4191 4192
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
4193
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
4194
                virDevicePCIAddressEqual(&hostdev->source.subsys.u.pci.addr,
4195 4196
                                         &netdef->forward.ifs[i].device.pci)) {
                dev = &netdef->forward.ifs[i];
4197 4198 4199 4200 4201 4202 4203 4204 4205
                break;
            }
        }
        /* dev points at the physical device we want to use */
        if (!dev) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' doesn't have "
                             "PCI device %04x:%02x:%02x.%x in use by domain"),
                           netdef->name,
4206 4207 4208 4209
                           hostdev->source.subsys.u.pci.addr.domain,
                           hostdev->source.subsys.u.pci.addr.bus,
                           hostdev->source.subsys.u.pci.addr.slot,
                           hostdev->source.subsys.u.pci.addr.function);
J
Ján Tomko 已提交
4210
            goto error;
4211 4212 4213 4214 4215 4216 4217
        }

        /* PASSTHROUGH mode, PRIVATE Mode + 802.1Qbh, and hostdev (PCI
         * passthrough) all require exclusive access to a device, so
         * current connections count must be 0 in those cases.
         */
        if ((dev->connections > 0) &&
4218
            netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' claims the PCI device at "
                             "domain=%d bus=%d slot=%d function=%d "
                             "is already in use by a different domain"),
                           netdef->name,
                           dev->device.pci.domain, dev->device.pci.bus,
                           dev->device.pci.slot, dev->device.pci.function);
            goto error;
        }

        /* we are now assured of success, so mark the allocation */
        dev->connections++;
        VIR_DEBUG("Using physical device %04x:%02x:%02x.%x, connections %d",
                  dev->device.pci.domain, dev->device.pci.bus,
                  dev->device.pci.slot, dev->device.pci.function,
                  dev->connections);
4235 4236
    }

4237
 success:
4238 4239 4240
    netdef->connections++;
    VIR_DEBUG("Using network %s, %d connections",
              netdef->name, netdef->connections);
4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251

    /* finally we can call the 'plugged' hook script if any */
    if (networkRunHook(network, dom, iface, VIR_HOOK_NETWORK_OP_IFACE_PLUGGED,
                       VIR_HOOK_SUBOP_BEGIN) < 0) {
        /* adjust for failure */
        if (dev)
            dev->connections--;
        netdef->connections--;
        goto error;
    }

4252
    ret = 0;
4253
 cleanup:
4254 4255 4256
    if (network)
        virNetworkObjUnlock(network);
    return ret;
4257

4258
 error:
4259
    goto cleanup;
4260 4261 4262 4263
}


/* networkReleaseActualDevice:
4264
 * @dom: domain definition that @iface belongs to
4265 4266 4267 4268 4269 4270 4271 4272 4273 4274
 * @iface:  a domain's NetDef (interface definition)
 *
 * Given a domain <interface> element that previously had its <actual>
 * element filled in (and possibly a physical device allocated to it),
 * free up the physical device for use by someone else, and free the
 * virDomainActualNetDef.
 *
 * Returns 0 on success, -1 on failure.
 */
int
4275 4276
networkReleaseActualDevice(virDomainDefPtr dom,
                           virDomainNetDefPtr iface)
4277
{
4278
    virDomainNetType actualType = virDomainNetGetActualType(iface);
4279
    virNetworkObjPtr network;
4280
    virNetworkDefPtr netdef;
4281
    virNetworkForwardIfDefPtr dev = NULL;
4282 4283
    size_t i;
    int ret = -1;
4284 4285

    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
J
Ján Tomko 已提交
4286
        return 0;
4287

4288
    networkDriverLock();
4289
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
4290
    networkDriverUnlock();
4291
    if (!network) {
4292 4293 4294
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
4295 4296 4297 4298
        goto error;
    }
    netdef = network->def;

4299 4300
    if (iface->data.network.actual &&
        (netdef->forward.type == VIR_NETWORK_FORWARD_NONE ||
4301 4302 4303 4304 4305
         netdef->forward.type == VIR_NETWORK_FORWARD_NAT ||
         netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE) &&
        networkUnplugBandwidth(network, iface) < 0)
        goto error;

4306 4307 4308
    if ((!iface->data.network.actual) ||
        ((actualType != VIR_DOMAIN_NET_TYPE_DIRECT) &&
         (actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV))) {
4309 4310
        VIR_DEBUG("Nothing to release to network %s", iface->data.network.name);
        goto success;
4311 4312
    }

4313
    if (netdef->forward.nifs == 0) {
4314
        virReportError(VIR_ERR_INTERNAL_ERROR,
4315
                       _("network '%s' uses a direct/hostdev mode, but "
4316 4317
                         "has no forward dev and no interface pool"),
                       netdef->name);
4318
        goto error;
4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330
    }

    if (actualType == VIR_DOMAIN_NET_TYPE_DIRECT) {
        const char *actualDev;

        actualDev = virDomainNetGetActualDirectDev(iface);
        if (!actualDev) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("the interface uses a direct mode, "
                             "but has no source dev"));
            goto error;
        }
4331

4332 4333
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
4334
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
4335 4336
                STREQ(actualDev, netdef->forward.ifs[i].device.dev)) {
                dev = &netdef->forward.ifs[i];
4337 4338 4339
                break;
            }
        }
4340

4341
        if (!dev) {
4342
            virReportError(VIR_ERR_INTERNAL_ERROR,
4343 4344
                           _("network '%s' doesn't have dev='%s' "
                             "in use by domain"),
4345
                           netdef->name, actualDev);
4346
            goto error;
4347 4348
        }

4349
        dev->connections--;
4350
        VIR_DEBUG("Releasing physical device %s, connections %d",
4351
                  dev->device.dev, dev->connections);
4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362

    } else /* if (actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV) */ {
        virDomainHostdevDefPtr hostdev;

        hostdev = virDomainNetGetActualHostdev(iface);
        if (!hostdev) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("the interface uses a hostdev mode, but has no hostdev"));
            goto error;
        }

4363 4364
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
4365
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
4366
                virDevicePCIAddressEqual(&hostdev->source.subsys.u.pci.addr,
4367 4368
                                         &netdef->forward.ifs[i].device.pci)) {
                dev = &netdef->forward.ifs[i];
4369 4370 4371 4372 4373 4374 4375 4376 4377
                break;
            }
        }

        if (!dev) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' doesn't have "
                             "PCI device %04x:%02x:%02x.%x in use by domain"),
                           netdef->name,
4378 4379 4380 4381
                           hostdev->source.subsys.u.pci.addr.domain,
                           hostdev->source.subsys.u.pci.addr.bus,
                           hostdev->source.subsys.u.pci.addr.slot,
                           hostdev->source.subsys.u.pci.addr.function);
J
Ján Tomko 已提交
4382
            goto error;
4383 4384 4385 4386 4387 4388 4389
        }

        dev->connections--;
        VIR_DEBUG("Releasing physical device %04x:%02x:%02x.%x, connections %d",
                  dev->device.pci.domain, dev->device.pci.bus,
                  dev->device.pci.slot, dev->device.pci.function,
                  dev->connections);
J
Ján Tomko 已提交
4390
    }
4391

4392
 success:
4393
    if (iface->data.network.actual) {
4394
        netdef->connections--;
4395 4396
        VIR_DEBUG("Releasing network %s, %d connections",
                  netdef->name, netdef->connections);
4397

4398 4399 4400 4401
        /* finally we can call the 'unplugged' hook script if any */
        networkRunHook(network, dom, iface, VIR_HOOK_NETWORK_OP_IFACE_UNPLUGGED,
                       VIR_HOOK_SUBOP_BEGIN);
    }
4402
    ret = 0;
4403
 cleanup:
4404 4405
    if (network)
        virNetworkObjUnlock(network);
4406 4407 4408 4409
    if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
        virDomainActualNetDefFree(iface->data.network.actual);
        iface->data.network.actual = NULL;
    }
4410
    return ret;
4411

4412
 error:
4413
    goto cleanup;
4414
}
4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438

/*
 * networkGetNetworkAddress:
 * @netname: the name of a network
 * @netaddr: string representation of IP address for that network.
 *
 * Attempt to return an IP (v4) address associated with the named
 * network. If a libvirt virtual network, that will be provided in the
 * configuration. For host bridge and direct (macvtap) networks, we
 * must do an ioctl to learn the address.
 *
 * Note: This function returns the 1st IPv4 address it finds. It might
 * be useful if it was more flexible, but the current use (getting a
 * listen address for qemu's vnc/spice graphics server) can only use a
 * single address anyway.
 *
 * Returns 0 on success, and puts a string (which must be free'd by
 * the caller) into *netaddr. Returns -1 on failure or -2 if
 * completely unsupported.
 */
int
networkGetNetworkAddress(const char *netname, char **netaddr)
{
    int ret = -1;
4439
    virNetworkObjPtr network;
4440 4441 4442 4443
    virNetworkDefPtr netdef;
    virNetworkIpDefPtr ipdef;
    virSocketAddr addr;
    virSocketAddrPtr addrptr = NULL;
4444
    char *dev_name = NULL;
4445 4446

    *netaddr = NULL;
4447
    networkDriverLock();
4448
    network = virNetworkFindByName(&driver->networks, netname);
4449
    networkDriverUnlock();
4450
    if (!network) {
4451 4452 4453
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       netname);
4454
        goto error;
4455 4456 4457
    }
    netdef = network->def;

4458
    switch (netdef->forward.type) {
4459 4460 4461 4462 4463 4464
    case VIR_NETWORK_FORWARD_NONE:
    case VIR_NETWORK_FORWARD_NAT:
    case VIR_NETWORK_FORWARD_ROUTE:
        /* if there's an ipv4def, get it's address */
        ipdef = virNetworkDefGetIpByIndex(netdef, AF_INET, 0);
        if (!ipdef) {
4465 4466 4467
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' doesn't have an IPv4 address"),
                           netdef->name);
4468 4469 4470 4471 4472 4473
            break;
        }
        addrptr = &ipdef->address;
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
4474
        if ((dev_name = netdef->bridge))
4475 4476 4477 4478 4479 4480 4481 4482
            break;
        /*
         * fall through if netdef->bridge wasn't set, since this is
         * also a direct-mode interface.
         */
    case VIR_NETWORK_FORWARD_PRIVATE:
    case VIR_NETWORK_FORWARD_VEPA:
    case VIR_NETWORK_FORWARD_PASSTHROUGH:
4483 4484
        if ((netdef->forward.nifs > 0) && netdef->forward.ifs)
            dev_name = netdef->forward.ifs[0].device.dev;
4485

4486
        if (!dev_name) {
4487 4488 4489
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' has no associated interface or bridge"),
                           netdef->name);
4490 4491 4492 4493
        }
        break;
    }

4494
    if (dev_name) {
4495
        if (virNetDevGetIPv4Address(dev_name, &addr) < 0)
4496
            goto error;
4497
        addrptr = &addr;
4498 4499
    }

4500 4501 4502
    if (!(addrptr &&
          (*netaddr = virSocketAddrFormat(addrptr)))) {
        goto error;
4503 4504
    }

4505
    ret = 0;
4506
 cleanup:
4507 4508 4509
    if (network)
        virNetworkObjUnlock(network);
    return ret;
4510

4511
 error:
4512
    goto cleanup;
4513
}
4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531

/**
 * networkCheckBandwidth:
 * @net: network QoS
 * @iface: interface QoS
 * @new_rate: new rate for non guaranteed class
 *
 * Returns: -1 if plugging would overcommit network QoS
 *           0 if plugging is safe (@new_rate updated)
 *           1 if no QoS is set (@new_rate untouched)
 */
static int
networkCheckBandwidth(virNetworkObjPtr net,
                      virDomainNetDefPtr iface,
                      unsigned long long *new_rate)
{
    int ret = -1;
    virNetDevBandwidthPtr netBand = net->def->bandwidth;
4532
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
4533 4534 4535 4536
    unsigned long long tmp_floor_sum = net->floor_sum;
    unsigned long long tmp_new_rate = 0;
    char ifmac[VIR_MAC_STRING_BUFLEN];

4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547
    virMacAddrFormat(&iface->mac, ifmac);

    if (ifaceBand && ifaceBand->in && ifaceBand->in->floor &&
        !(netBand && netBand->in)) {
        virReportError(VIR_ERR_OPERATION_UNSUPPORTED,
                       _("Invalid use of 'floor' on interface with MAC "
                         "address %s - network '%s' has no inbound QoS set"),
                       ifmac, net->def->name);
        return -1;
    }

4548
    if (!ifaceBand || !ifaceBand->in || !ifaceBand->in->floor ||
4549 4550
        !netBand || !netBand->in) {
        /* no QoS required, claim success */
4551
        return 1;
4552
    }
4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583

    tmp_new_rate = netBand->in->average;
    tmp_floor_sum += ifaceBand->in->floor;

    /* check against peak */
    if (netBand->in->peak) {
        tmp_new_rate = netBand->in->peak;
        if (tmp_floor_sum > netBand->in->peak) {
            virReportError(VIR_ERR_OPERATION_INVALID,
                           _("Cannot plug '%s' interface into '%s' because it "
                             "would overcommit 'peak' on network '%s'"),
                           ifmac,
                           net->def->bridge,
                           net->def->name);
            goto cleanup;
        }
    } else if (tmp_floor_sum > netBand->in->average) {
        /* tmp_floor_sum can be between 'average' and 'peak' iff 'peak' is set.
         * Otherwise, tmp_floor_sum must be below 'average'. */
        virReportError(VIR_ERR_OPERATION_INVALID,
                       _("Cannot plug '%s' interface into '%s' because it "
                         "would overcommit 'average' on network '%s'"),
                       ifmac,
                       net->def->bridge,
                       net->def->name);
        goto cleanup;
    }

    *new_rate = tmp_new_rate;
    ret = 0;

4584
 cleanup:
4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621
    return ret;
}

/**
 * networkNextClassID:
 * @net: network object
 *
 * Find next free class ID. @net is supposed
 * to be locked already. If there is a free ID,
 * it is marked as used and returned.
 *
 * Returns next free class ID or -1 if none is available.
 */
static ssize_t
networkNextClassID(virNetworkObjPtr net)
{
    size_t ret = 0;
    bool is_set = false;

    while (virBitmapGetBit(net->class_id, ret, &is_set) == 0 && is_set)
        ret++;

    if (is_set || virBitmapSetBit(net->class_id, ret) < 0)
        return -1;

    return ret;
}

static int
networkPlugBandwidth(virNetworkObjPtr net,
                     virDomainNetDefPtr iface)
{
    int ret = -1;
    int plug_ret;
    unsigned long long new_rate = 0;
    ssize_t class_id = 0;
    char ifmac[VIR_MAC_STRING_BUFLEN];
4622
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650

    if ((plug_ret = networkCheckBandwidth(net, iface, &new_rate)) < 0) {
        /* helper reported error */
        goto cleanup;
    }

    if (plug_ret > 0) {
        /* no QoS needs to be set; claim success */
        ret = 0;
        goto cleanup;
    }

    virMacAddrFormat(&iface->mac, ifmac);
    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK ||
        !iface->data.network.actual) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Cannot set bandwidth on interface '%s' of type %d"),
                       ifmac, iface->type);
        goto cleanup;
    }

    /* generate new class_id */
    if ((class_id = networkNextClassID(net)) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not generate next class ID"));
        goto cleanup;
    }

4651 4652
    plug_ret = virNetDevBandwidthPlug(net->def->bridge, net->def->bandwidth,
                                      &iface->mac, ifaceBand, class_id);
4653 4654 4655 4656 4657 4658 4659 4660
    if (plug_ret < 0) {
        ignore_value(virNetDevBandwidthUnplug(net->def->bridge, class_id));
        goto cleanup;
    }

    /* QoS was set, generate new class ID */
    iface->data.network.actual->class_id = class_id;
    /* update sum of 'floor'-s of attached NICs */
4661
    net->floor_sum += ifaceBand->in->floor;
4662
    /* update status file */
4663
    if (virNetworkSaveStatus(driver->stateDir, net) < 0) {
4664
        ignore_value(virBitmapClearBit(net->class_id, class_id));
4665
        net->floor_sum -= ifaceBand->in->floor;
4666 4667 4668 4669
        iface->data.network.actual->class_id = 0;
        ignore_value(virNetDevBandwidthUnplug(net->def->bridge, class_id));
        goto cleanup;
    }
4670 4671 4672 4673 4674 4675 4676 4677 4678
    /* update rate for non guaranteed NICs */
    new_rate -= net->floor_sum;
    if (virNetDevBandwidthUpdateRate(net->def->bridge, "1:2",
                                     net->def->bandwidth, new_rate) < 0)
        VIR_WARN("Unable to update rate for 1:2 class on %s bridge",
                 net->def->bridge);

    ret = 0;

4679
 cleanup:
4680 4681 4682 4683 4684 4685 4686 4687 4688
    return ret;
}

static int
networkUnplugBandwidth(virNetworkObjPtr net,
                       virDomainNetDefPtr iface)
{
    int ret = 0;
    unsigned long long new_rate;
4689
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
4690 4691 4692

    if (iface->data.network.actual &&
        iface->data.network.actual->class_id) {
4693 4694 4695 4696 4697
        if (!net->def->bandwidth || !net->def->bandwidth->in) {
            VIR_WARN("Network %s has no bandwidth but unplug requested",
                     net->def->name);
            goto cleanup;
        }
4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708
        /* we must remove class from bridge */
        new_rate = net->def->bandwidth->in->average;

        if (net->def->bandwidth->in->peak > 0)
            new_rate = net->def->bandwidth->in->peak;

        ret = virNetDevBandwidthUnplug(net->def->bridge,
                                       iface->data.network.actual->class_id);
        if (ret < 0)
            goto cleanup;
        /* update sum of 'floor'-s of attached NICs */
4709
        net->floor_sum -= ifaceBand->in->floor;
4710 4711 4712 4713
        /* return class ID */
        ignore_value(virBitmapClearBit(net->class_id,
                                       iface->data.network.actual->class_id));
        /* update status file */
4714
        if (virNetworkSaveStatus(driver->stateDir, net) < 0) {
4715
            net->floor_sum += ifaceBand->in->floor;
4716 4717 4718 4719
            ignore_value(virBitmapSetBit(net->class_id,
                                         iface->data.network.actual->class_id));
            goto cleanup;
        }
4720 4721 4722 4723 4724 4725 4726 4727 4728 4729
        /* update rate for non guaranteed NICs */
        new_rate -= net->floor_sum;
        if (virNetDevBandwidthUpdateRate(net->def->bridge, "1:2",
                                         net->def->bandwidth, new_rate) < 0)
            VIR_WARN("Unable to update rate for 1:2 class on %s bridge",
                     net->def->bridge);
        /* no class is associated any longer */
        iface->data.network.actual->class_id = 0;
    }

4730
 cleanup:
4731 4732
    return ret;
}
4733 4734 4735

static void
networkNetworkObjTaint(virNetworkObjPtr net,
4736
                       virNetworkTaintFlags taint)
4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747
{
    if (virNetworkObjTaint(net, taint)) {
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(net->def->uuid, uuidstr);

        VIR_WARN("Network name='%s' uuid=%s is tainted: %s",
                 net->def->name,
                 uuidstr,
                 virNetworkTaintTypeToString(taint));
    }
}