bridge_driver.c 136.5 KB
Newer Older
1
/*
2
 * bridge_driver.c: core driver methods for managing network
3
 *
4
 * Copyright (C) 2006-2013 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 45 46
#if HAVE_SYS_SYSCTL_H
# include <sys/sysctl.h>
#endif
47

48
#include "virerror.h"
49
#include "datatypes.h"
50
#include "bridge_driver.h"
51
#include "bridge_driver_platform.h"
52
#include "network_conf.h"
53
#include "device_conf.h"
54
#include "driver.h"
55
#include "virbuffer.h"
56
#include "virpidfile.h"
57
#include "vircommand.h"
58
#include "viralloc.h"
59
#include "viruuid.h"
60
#include "viriptables.h"
61
#include "virlog.h"
62
#include "virdnsmasq.h"
63
#include "configmake.h"
64
#include "virnetdev.h"
65
#include "virpci.h"
66 67
#include "virnetdevbridge.h"
#include "virnetdevtap.h"
68
#include "virnetdevvportprofile.h"
69
#include "virdbus.h"
70
#include "virfile.h"
71
#include "virstring.h"
72
#include "viraccessapicheck.h"
73

74 75
#define VIR_FROM_THIS VIR_FROM_NETWORK

76
static void networkDriverLock(virNetworkDriverStatePtr driver)
77
{
78
    virMutexLock(&driver->lock);
79
}
80
static void networkDriverUnlock(virNetworkDriverStatePtr driver)
81
{
82
    virMutexUnlock(&driver->lock);
83 84
}

85
static int networkStateCleanup(void);
86

87
static int networkStartNetwork(virNetworkDriverStatePtr driver,
88 89
                               virNetworkObjPtr network);

90
static int networkShutdownNetwork(virNetworkDriverStatePtr driver,
91 92
                                  virNetworkObjPtr network);

93
static int networkStartNetworkVirtual(virNetworkDriverStatePtr driver,
94
                                     virNetworkObjPtr network);
95

96
static int networkShutdownNetworkVirtual(virNetworkDriverStatePtr driver,
R
Roman Bogorodskiy 已提交
97
                                         virNetworkObjPtr network);
98

99
static int networkStartNetworkExternal(virNetworkDriverStatePtr driver,
100 101
                                     virNetworkObjPtr network);

102
static int networkShutdownNetworkExternal(virNetworkDriverStatePtr driver,
103
                                        virNetworkObjPtr network);
104

105
static void networkReloadFirewallRules(virNetworkDriverStatePtr driver);
106
static void networkRefreshDaemons(virNetworkDriverStatePtr driver);
107

108 109 110 111 112
static int networkPlugBandwidth(virNetworkObjPtr net,
                                virDomainNetDefPtr iface);
static int networkUnplugBandwidth(virNetworkObjPtr net,
                                  virDomainNetDefPtr iface);

113
static virNetworkDriverStatePtr driverState = NULL;
114

115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
static virNetworkObjPtr
networkObjFromNetwork(virNetworkPtr net)
{
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
    virNetworkObjPtr network;
    char uuidstr[VIR_UUID_STRING_BUFLEN];

    networkDriverLock(driver);
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
    networkDriverUnlock(driver);

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

    return network;
}

136
static char *
137
networkDnsmasqLeaseFileNameDefault(const char *netname)
138 139 140
{
    char *leasefile;

141 142
    ignore_value(virAsprintf(&leasefile, "%s/%s.leases",
                             driverState->dnsmasqStateDir, netname));
143 144 145
    return leasefile;
}

146 147 148
networkDnsmasqLeaseFileNameFunc networkDnsmasqLeaseFileName =
    networkDnsmasqLeaseFileNameDefault;

149 150 151 152 153
static char *
networkDnsmasqConfigFileName(const char *netname)
{
    char *conffile;

154 155
    ignore_value(virAsprintf(&conffile, "%s/%s.conf",
                             driverState->dnsmasqStateDir, netname));
156 157 158
    return conffile;
}

159 160 161 162 163 164
static char *
networkRadvdPidfileBasename(const char *netname)
{
    /* this is simple but we want to be sure it's consistently done */
    char *pidfilebase;

165
    ignore_value(virAsprintf(&pidfilebase, "%s-radvd", netname));
166 167 168 169 170 171 172 173
    return pidfilebase;
}

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

174 175
    ignore_value(virAsprintf(&configfile, "%s/%s-radvd.conf",
                             driverState->radvdStateDir, netname));
176 177
    return configfile;
}
178

179 180
/* do needed cleanup steps and remove the network from the list */
static int
181
networkRemoveInactive(virNetworkDriverStatePtr driver,
182 183 184 185
                      virNetworkObjPtr net)
{
    char *leasefile = NULL;
    char *radvdconfigfile = NULL;
186
    char *configfile = NULL;
187
    char *radvdpidbase = NULL;
188
    char *statusfile = NULL;
189 190 191 192 193 194
    dnsmasqContext *dctx = NULL;
    virNetworkDefPtr def = virNetworkObjGetPersistentDef(net);

    int ret = -1;

    /* remove the (possibly) existing dnsmasq and radvd files */
195 196
    if (!(dctx = dnsmasqContextNew(def->name,
                                   driverState->dnsmasqStateDir))) {
197
        goto cleanup;
198
    }
199 200 201 202 203

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

    if (!(radvdconfigfile = networkRadvdConfigFileName(def->name)))
204
        goto cleanup;
205 206

    if (!(radvdpidbase = networkRadvdPidfileBasename(def->name)))
207
        goto cleanup;
208

209
    if (!(configfile = networkDnsmasqConfigFileName(def->name)))
210
        goto cleanup;
211

212 213
    if (!(statusfile
          = virNetworkConfigFile(driverState->stateDir, def->name)))
214
        goto cleanup;
215

216 217 218
    /* dnsmasq */
    dnsmasqDelete(dctx);
    unlink(leasefile);
219
    unlink(configfile);
220 221 222

    /* radvd */
    unlink(radvdconfigfile);
223
    virPidFileDelete(driverState->pidDir, radvdpidbase);
224

225 226 227
    /* remove status file */
    unlink(statusfile);

228 229 230 231 232 233 234
    /* remove the network definition */
    virNetworkRemoveInactive(&driver->networks, net);

    ret = 0;

cleanup:
    VIR_FREE(leasefile);
235
    VIR_FREE(configfile);
236 237
    VIR_FREE(radvdconfigfile);
    VIR_FREE(radvdpidbase);
238
    VIR_FREE(statusfile);
239 240 241 242
    dnsmasqContextFree(dctx);
    return ret;
}

243 244 245
static char *
networkBridgeDummyNicName(const char *brname)
{
246
    static const char dummyNicSuffix[] = "-nic";
247 248
    char *nicname;

249 250 251 252 253 254 255
    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.
         */
256 257 258 259 260
        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));
261
    } else {
262
        ignore_value(virAsprintf(&nicname, "%s%s", brname, dummyNicSuffix));
263
    }
264 265 266
    return nicname;
}

267
static void
268
networkFindActiveConfigs(virNetworkDriverStatePtr driver)
269
{
270
    size_t i;
271

272
    for (i = 0; i < driver->networks.count; i++) {
273 274 275 276 277 278
        virNetworkObjPtr obj = driver->networks.objs[i];

        virNetworkObjLock(obj);

        /* If bridge exists, then mark it active */
        if (obj->def->bridge &&
H
Hu Tao 已提交
279
            virNetDevExists(obj->def->bridge) == 1) {
280 281
            obj->active = 1;

282 283
            /* Try and read dnsmasq/radvd pids if any */
            if (obj->def->ips && (obj->def->nips > 0)) {
284 285
                char *radvdpidbase;

286
                ignore_value(virPidFileReadIfAlive(driverState->pidDir, obj->def->name,
287 288
                                                   &obj->dnsmasqPid,
                                                   dnsmasqCapsGetBinaryPath(driver->dnsmasqCaps)));
289

290
                if (!(radvdpidbase = networkRadvdPidfileBasename(obj->def->name)))
291
                    goto cleanup;
292
                ignore_value(virPidFileReadIfAlive(driverState->pidDir, radvdpidbase,
293
                                                   &obj->radvdPid, RADVD));
294
                VIR_FREE(radvdpidbase);
295 296 297
            }
        }

298
    cleanup:
299 300
        virNetworkObjUnlock(obj);
    }
301 302 303 304 305 306 307 308 309 310 311 312 313 314 315

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

        if (!obj->persistent && !obj->active) {
            networkRemoveInactive(driver, obj);
            continue;
        }

        virNetworkObjUnlock(obj);
        i++;
    }
316 317 318
}


319
static void
320
networkAutostartConfigs(virNetworkDriverStatePtr driver) {
321
    size_t i;
322

323
    for (i = 0; i < driver->networks.count; i++) {
324
        virNetworkObjLock(driver->networks.objs[i]);
325
        if (driver->networks.objs[i]->autostart &&
326 327
            !virNetworkObjIsActive(driver->networks.objs[i])) {
            if (networkStartNetwork(driver, driver->networks.objs[i]) < 0) {
328
            /* failed to start but already logged */
329
            }
330
        }
331
        virNetworkObjUnlock(driver->networks.objs[i]);
332 333 334
    }
}

335 336 337 338
#if HAVE_FIREWALLD
static DBusHandlerResult
firewalld_dbus_filter_bridge(DBusConnection *connection ATTRIBUTE_UNUSED,
                             DBusMessage *message, void *user_data) {
339
    virNetworkDriverStatePtr _driverState = user_data;
340 341 342 343 344 345 346

    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.");
347
        networkReloadFirewallRules(_driverState);
348 349 350 351 352 353
    }

    return DBUS_HANDLER_RESULT_NOT_YET_HANDLED;
}
#endif

354
/**
355
 * networkStateInitialize:
356 357 358 359
 *
 * Initialization function for the QEmu daemon
 */
static int
360 361 362
networkStateInitialize(bool privileged,
                       virStateInhibitCallback callback ATTRIBUTE_UNUSED,
                       void *opaque ATTRIBUTE_UNUSED)
363
{
364 365 366
    int ret = -1;
    char *configdir = NULL;
    char *rundir = NULL;
367 368 369
#ifdef HAVE_FIREWALLD
    DBusConnection *sysbus = NULL;
#endif
370 371

    if (VIR_ALLOC(driverState) < 0)
372
        goto error;
373

374 375 376 377
    if (virMutexInit(&driverState->lock) < 0) {
        VIR_FREE(driverState);
        goto error;
    }
378 379
    networkDriverLock(driverState);

380 381 382 383 384 385 386 387 388
    /* configuration/state paths are one of
     * ~/.config/libvirt/... (session/unprivileged)
     * /etc/libvirt/... && /var/(run|lib)/libvirt/... (system/privileged).
     *
     * NB: The qemu driver puts its domain state in /var/run, and I
     * think the network driver should have used /var/run too (instead
     * of /var/lib), but it's been this way for a long time, and we
     * probably shouldn't change it now.
     */
389
    if (privileged) {
390 391 392 393 394 395 396 397 398 399 400 401 402
        if (VIR_STRDUP(driverState->networkConfigDir,
                       SYSCONFDIR "/libvirt/qemu/networks") < 0 ||
            VIR_STRDUP(driverState->networkAutostartDir,
                       SYSCONFDIR "/libvirt/qemu/networks/autostart") < 0 ||
            VIR_STRDUP(driverState->stateDir,
                       LOCALSTATEDIR "/lib/libvirt/network") < 0 ||
            VIR_STRDUP(driverState->pidDir,
                       LOCALSTATEDIR "/run/libvirt/network") < 0 ||
            VIR_STRDUP(driverState->dnsmasqStateDir,
                       LOCALSTATEDIR "/lib/libvirt/dnsmasq") < 0 ||
            VIR_STRDUP(driverState->radvdStateDir,
                       LOCALSTATEDIR "/lib/libvirt/radvd") < 0)
            goto error;
403
    } else {
404 405 406
        configdir = virGetUserConfigDirectory();
        rundir = virGetUserRuntimeDirectory();
        if (!(configdir && rundir))
407
            goto error;
408

409 410 411 412 413 414 415 416 417 418 419 420
        if ((virAsprintf(&driverState->networkConfigDir,
                         "%s/qemu/networks", configdir) < 0) ||
            (virAsprintf(&driverState->networkAutostartDir,
                         "%s/qemu/networks/autostart", configdir) < 0) ||
            (virAsprintf(&driverState->stateDir,
                         "%s/network/lib", rundir) < 0) ||
            (virAsprintf(&driverState->pidDir,
                         "%s/network/run", rundir) < 0) ||
            (virAsprintf(&driverState->dnsmasqStateDir,
                         "%s/dnsmasq/lib", rundir) < 0) ||
            (virAsprintf(&driverState->radvdStateDir,
                         "%s/radvd/lib", rundir) < 0)) {
421
            goto error;
422
        }
423 424
    }

425 426
    /* if this fails now, it will be retried later with dnsmasqCapsRefresh() */
    driverState->dnsmasqCaps = dnsmasqCapsNewFromBinary(DNSMASQ);
427

428
    if (virNetworkLoadAllState(&driverState->networks,
429
                               driverState->stateDir) < 0)
430 431
        goto error;

432
    if (virNetworkLoadAllConfigs(&driverState->networks,
433
                                 driverState->networkConfigDir,
434 435 436
                                 driverState->networkAutostartDir) < 0)
        goto error;

437
    networkFindActiveConfigs(driverState);
438
    networkReloadFirewallRules(driverState);
439
    networkRefreshDaemons(driverState);
440

441 442
    networkDriverUnlock(driverState);

443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
#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,
                                   driverState, NULL);
    }
#endif

469 470 471 472 473
    ret = 0;
cleanup:
    VIR_FREE(configdir);
    VIR_FREE(rundir);
    return ret;
474

475
error:
476 477
    if (driverState)
        networkDriverUnlock(driverState);
478
    networkStateCleanup();
479
    goto cleanup;
480 481
}

482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497
/**
 * networkStateAutoStart:
 *
 * Function to AutoStart the bridge configs
 */
static void
networkStateAutoStart(void)
{
    if (!driverState)
        return;

    networkDriverLock(driverState);
    networkAutostartConfigs(driverState);
    networkDriverUnlock(driverState);
}

498
/**
499
 * networkStateReload:
500 501 502 503 504
 *
 * Function to restart the QEmu daemon, it will recheck the configuration
 * files and update its state and the networking
 */
static int
505
networkStateReload(void) {
506 507 508
    if (!driverState)
        return 0;

509
    networkDriverLock(driverState);
510
    virNetworkLoadAllState(&driverState->networks,
511
                           driverState->stateDir);
512
    virNetworkLoadAllConfigs(&driverState->networks,
513 514
                             driverState->networkConfigDir,
                             driverState->networkAutostartDir);
515
    networkReloadFirewallRules(driverState);
516
    networkRefreshDaemons(driverState);
517
    networkAutostartConfigs(driverState);
518
    networkDriverUnlock(driverState);
519 520 521 522 523
    return 0;
}


/**
524
 * networkStateCleanup:
525 526 527 528
 *
 * Shutdown the QEmu daemon, it will stop all active domains and networks
 */
static int
529
networkStateCleanup(void) {
530 531 532
    if (!driverState)
        return -1;

533 534
    networkDriverLock(driverState);

535
    /* free inactive networks */
536
    virNetworkObjListFree(&driverState->networks);
537 538 539

    VIR_FREE(driverState->networkConfigDir);
    VIR_FREE(driverState->networkAutostartDir);
540 541 542 543
    VIR_FREE(driverState->stateDir);
    VIR_FREE(driverState->pidDir);
    VIR_FREE(driverState->dnsmasqStateDir);
    VIR_FREE(driverState->radvdStateDir);
544

545 546
    virObjectUnref(driverState->dnsmasqCaps);

547
    networkDriverUnlock(driverState);
548
    virMutexDestroy(&driverState->lock);
549

550 551 552 553 554 555
    VIR_FREE(driverState);

    return 0;
}


556 557 558 559 560 561 562
/* 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)
{
563 564
    size_t i;
    int ret = -1;
565 566 567 568 569 570 571
    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.
     */
572
    for (i = 0; i < 25; i++) {
573
        int signum = 0;
574
        if (i == 0)
575
            signum = SIGTERM;
576
        else if (i == 15) {
577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615
            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);
cleanup:
    return ret;
}

G
Gene Czarcinski 已提交
616 617 618 619
    /* the following does not build a file, it builds a list
     * which is later saved into a file
     */

620
static int
G
Gene Czarcinski 已提交
621 622
networkBuildDnsmasqDhcpHostsList(dnsmasqContext *dctx,
                                 virNetworkIpDefPtr ipdef)
623
{
624
    size_t i;
G
Gene Czarcinski 已提交
625
    bool ipv6 = false;
626

G
Gene Czarcinski 已提交
627 628
    if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
        ipv6 = true;
629 630
    for (i = 0; i < ipdef->nhosts; i++) {
        virNetworkDHCPHostDefPtr host = &(ipdef->hosts[i]);
G
Gene Czarcinski 已提交
631
        if (VIR_SOCKET_ADDR_VALID(&host->ip))
632 633
            if (dnsmasqAddDhcpHost(dctx, host->mac, &host->ip,
                                   host->name, host->id, ipv6) < 0)
634
                return -1;
635
    }
636

G
Gene Czarcinski 已提交
637 638 639 640 641 642 643
    return 0;
}

static int
networkBuildDnsmasqHostsList(dnsmasqContext *dctx,
                             virNetworkDNSDefPtr dnsdef)
{
644
    size_t i, j;
G
Gene Czarcinski 已提交
645

646 647
    if (dnsdef) {
        for (i = 0; i < dnsdef->nhosts; i++) {
648
            virNetworkDNSHostDefPtr host = &(dnsdef->hosts[i]);
649
            if (VIR_SOCKET_ADDR_VALID(&host->ip)) {
650
                for (j = 0; j < host->nnames; j++)
651 652
                    if (dnsmasqAddHost(dctx, &host->ip, host->names[j]) < 0)
                        return -1;
653 654
            }
        }
655 656
    }

657
    return 0;
658 659 660
}


661 662
int
networkDnsmasqConfContents(virNetworkObjPtr network,
663 664 665 666
                           const char *pidfile,
                           char **configstr,
                           dnsmasqContext *dctx,
                           dnsmasqCapsPtr caps ATTRIBUTE_UNUSED)
667
{
668
    virBuffer configbuf = VIR_BUFFER_INITIALIZER;
669
    int r, ret = -1;
670
    int nbleases = 0;
671
    size_t i;
672 673 674 675
    char *record = NULL;
    char *recordPort = NULL;
    char *recordWeight = NULL;
    char *recordPriority = NULL;
676
    virNetworkDNSDefPtr dns = &network->def->dns;
G
Gene Czarcinski 已提交
677 678
    virNetworkIpDefPtr tmpipdef, ipdef, ipv4def, ipv6def;
    bool ipv6SLAAC;
679

680 681
    *configstr = NULL;

682
    /*
683 684 685
     * All dnsmasq parameters are put into a configuration file, except the
     * command line --conf-file=parameter which specifies the location of
     * configuration file.
686
     *
687 688
     * All dnsmasq conf-file parameters must be specified as "foo=bar"
     * as oppose to "--foo bar" which was acceptable on the command line.
689
     */
690 691 692 693 694 695

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

696 697
    /* create dnsmasq config file appropriate for this network */
    virBufferAsprintf(&configbuf,
698 699 700 701 702 703 704
                      "##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"
705
                      "strict-order\n",
706 707
                      network->def->name);

708 709 710
    if (!network->def->dns.forwardPlainNames)
        virBufferAddLit(&configbuf, "domain-needed\n");

711
    if (network->def->domain) {
712
        virBufferAsprintf(&configbuf,
713 714 715 716
                          "domain=%s\n"
                          "expand-hosts\n",
                          network->def->domain);
    }
717 718 719 720 721 722 723 724 725 726

    if (network->def->domain || !network->def->dns.forwardPlainNames) {
        /* need to specify local even if no domain specified, unless
         * the config says we should forward "plain" names (i.e. not
         * fully qualified, no '.' characters)
         */
        virBufferAsprintf(&configbuf,
                          "local=/%s/\n",
                          network->def->domain ? network->def->domain : "");
    }
727

728
    if (pidfile)
729
        virBufferAsprintf(&configbuf, "pid-file=%s\n", pidfile);
730

731 732 733
    /* dnsmasq will *always* listen on localhost unless told otherwise */
    virBufferAddLit(&configbuf, "except-interface=lo\n");

734 735 736 737 738 739 740 741
    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.
         */
742
        virBufferAsprintf(&configbuf,
743 744 745
                          "bind-dynamic\n"
                          "interface=%s\n",
                          network->def->bridge);
746
    } else {
747
        virBufferAddLit(&configbuf, "bind-interfaces\n");
748 749 750 751 752 753 754 755
        /*
         * --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
         */
756 757 758
        for (i = 0;
             (tmpipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, i));
             i++) {
759 760 761 762
            char *ipaddr = virSocketAddrFormat(&tmpipdef->address);

            if (!ipaddr)
                goto cleanup;
763

764
            /* also part of CVE 2012-3411 - if the host's version of
765
             * dnsmasq doesn't have bind-dynamic, only allow listening on
766 767
             * private/local IP addresses (see RFC1918/RFC3484/RFC4193)
             */
768 769
            if (!dnsmasqCapsGet(caps, DNSMASQ_CAPS_BINDTODEVICE) &&
                !virSocketAddrIsPrivate(&tmpipdef->address)) {
770 771 772 773
                unsigned long version = dnsmasqCapsGetVersion(caps);

                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("Publicly routable address %s is prohibited. "
774
                                 "The version of dnsmasq on this host (%d.%d) "
775 776 777 778
                                 "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 "
779 780 781 782 783 784
                                 "(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);
785 786
                goto cleanup;
            }
787
            virBufferAsprintf(&configbuf, "listen-address=%s\n", ipaddr);
788 789 790
            VIR_FREE(ipaddr);
        }
    }
791

792 793
    /* If this is an isolated network, set the default route option
     * (3) to be empty to avoid setting a default route that's
794
     * guaranteed to not work, and set no-resolv so that no dns
795 796 797
     * 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).
798
     */
799
    if (network->def->forward.type == VIR_NETWORK_FORWARD_NONE) {
800
        virBufferAddLit(&configbuf, "dhcp-option=3\n"
801
                        "no-resolv\n");
802
    }
803

804
    for (i = 0; i < dns->ntxts; i++) {
805
        virBufferAsprintf(&configbuf, "txt-record=%s,%s\n",
806 807
                          dns->txts[i].name,
                          dns->txts[i].value);
808
    }
809

810 811 812 813
    for (i = 0; i < dns->nsrvs; i++) {
        if (dns->srvs[i].service && dns->srvs[i].protocol) {
            if (dns->srvs[i].port &&
                virAsprintf(&recordPort, "%d", dns->srvs[i].port) < 0)
814
                goto cleanup;
815 816
            if (dns->srvs[i].priority &&
                virAsprintf(&recordPriority, "%d", dns->srvs[i].priority) < 0)
817
                goto cleanup;
818 819
            if (dns->srvs[i].weight &&
                virAsprintf(&recordWeight, "%d", dns->srvs[i].weight) < 0)
820
                goto cleanup;
821

822
            if (virAsprintf(&record, "%s.%s.%s,%s,%s,%s,%s",
823 824 825 826
                            dns->srvs[i].service,
                            dns->srvs[i].protocol,
                            dns->srvs[i].domain ? dns->srvs[i].domain : "",
                            dns->srvs[i].target ? dns->srvs[i].target : "",
827 828
                            recordPort           ? recordPort           : "",
                            recordPriority       ? recordPriority       : "",
829
                            recordWeight         ? recordWeight         : "") < 0)
830 831
                goto cleanup;

832
            virBufferAsprintf(&configbuf, "srv-host=%s\n", record);
833 834 835 836
            VIR_FREE(record);
            VIR_FREE(recordPort);
            VIR_FREE(recordWeight);
            VIR_FREE(recordPriority);
837
        }
838 839
    }

G
Gene Czarcinski 已提交
840
    /* Find the first dhcp for both IPv4 and IPv6 */
841 842 843
    for (i = 0, ipv4def = NULL, ipv6def = NULL, ipv6SLAAC = false;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, i));
         i++) {
G
Gene Czarcinski 已提交
844 845 846 847
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
            if (ipdef->nranges || ipdef->nhosts) {
                if (ipv4def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
848 849
                                   _("For IPv4, multiple DHCP definitions "
                                     "cannot be specified."));
G
Gene Czarcinski 已提交
850 851 852 853 854 855 856 857 858 859 860
                    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,
861 862 863 864 865 866 867 868 869
                                   _("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
870 871 872 873
                    goto cleanup;
                }
                if (ipv6def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
874 875
                                   _("For IPv6, multiple DHCP definitions "
                                     "cannot be specified."));
G
Gene Czarcinski 已提交
876 877 878 879 880 881 882 883 884 885 886 887 888
                    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 "
889 890 891 892
                 "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
893 894 895 896 897
    }

    ipdef = ipv4def ? ipv4def : ipv6def;

    while (ipdef) {
898
        for (r = 0; r < ipdef->nranges; r++) {
899
            char *saddr = virSocketAddrFormat(&ipdef->ranges[r].start);
900 901
            if (!saddr)
                goto cleanup;
902
            char *eaddr = virSocketAddrFormat(&ipdef->ranges[r].end);
903 904 905 906
            if (!eaddr) {
                VIR_FREE(saddr);
                goto cleanup;
            }
907
            virBufferAsprintf(&configbuf, "dhcp-range=%s,%s\n",
908
                              saddr, eaddr);
909
            VIR_FREE(saddr);
910
            VIR_FREE(eaddr);
911 912
            nbleases += virSocketAddrGetRange(&ipdef->ranges[r].start,
                                              &ipdef->ranges[r].end);
913
        }
914

915
        /*
916 917 918 919
         * 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)
920 921
         */
        if (!ipdef->nranges && ipdef->nhosts) {
922
            char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
923 924
            if (!bridgeaddr)
                goto cleanup;
925
            virBufferAsprintf(&configbuf, "dhcp-range=%s,static\n", bridgeaddr);
926 927
            VIR_FREE(bridgeaddr);
        }
928

G
Gene Czarcinski 已提交
929 930
        if (networkBuildDnsmasqDhcpHostsList(dctx, ipdef) < 0)
            goto cleanup;
931

G
Gene Czarcinski 已提交
932 933 934
        /* Note: the following is IPv4 only */
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
            if (ipdef->nranges || ipdef->nhosts)
935
                virBufferAddLit(&configbuf, "dhcp-no-override\n");
936

G
Gene Czarcinski 已提交
937
            if (ipdef->tftproot) {
938 939
                virBufferAddLit(&configbuf, "enable-tftp\n");
                virBufferAsprintf(&configbuf, "tftp-root=%s\n", ipdef->tftproot);
G
Gene Czarcinski 已提交
940
            }
941

G
Gene Czarcinski 已提交
942 943 944
            if (ipdef->bootfile) {
                if (VIR_SOCKET_ADDR_VALID(&ipdef->bootserver)) {
                    char *bootserver = virSocketAddrFormat(&ipdef->bootserver);
945

946
                    if (!bootserver)
G
Gene Czarcinski 已提交
947
                        goto cleanup;
948
                    virBufferAsprintf(&configbuf, "dhcp-boot=%s%s%s\n",
949
                                      ipdef->bootfile, ",,", bootserver);
G
Gene Czarcinski 已提交
950 951
                    VIR_FREE(bootserver);
                } else {
952
                    virBufferAsprintf(&configbuf, "dhcp-boot=%s\n", ipdef->bootfile);
G
Gene Czarcinski 已提交
953 954 955 956 957
                }
            }
        }
        ipdef = (ipdef == ipv6def) ? NULL : ipv6def;
    }
958

G
Gene Czarcinski 已提交
959 960
    if (nbleases > 0) {
        char *leasefile = networkDnsmasqLeaseFileName(network->def->name);
961
        if (!leasefile)
G
Gene Czarcinski 已提交
962
            goto cleanup;
963
        virBufferAsprintf(&configbuf, "dhcp-leasefile=%s\n", leasefile);
G
Gene Czarcinski 已提交
964
        VIR_FREE(leasefile);
965
        virBufferAsprintf(&configbuf, "dhcp-lease-max=%d\n", nbleases);
G
Gene Czarcinski 已提交
966
    }
967

G
Gene Czarcinski 已提交
968 969
    /* this is done once per interface */
    if (networkBuildDnsmasqHostsList(dctx, dns) < 0)
970
        goto cleanup;
G
Gene Czarcinski 已提交
971 972 973 974 975 976

    /* 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)
977 978
        virBufferAsprintf(&configbuf, "dhcp-hostsfile=%s\n",
                          dctx->hostsfile->path);
G
Gene Czarcinski 已提交
979

980 981
    /* Likewise, always create this file and put it on the
     * commandline, to allow for runtime additions.
G
Gene Czarcinski 已提交
982
     */
983
    virBufferAsprintf(&configbuf, "addn-hosts=%s\n",
984
                      dctx->addnhostsfile->path);
G
Gene Czarcinski 已提交
985 986 987 988

    /* Are we doing RA instead of radvd? */
    if (DNSMASQ_RA_SUPPORT(caps)) {
        if (ipv6def)
989
            virBufferAddLit(&configbuf, "enable-ra\n");
G
Gene Czarcinski 已提交
990
        else {
991 992 993
            for (i = 0;
                 (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
                 i++) {
G
Gene Czarcinski 已提交
994 995 996 997
                if (!(ipdef->nranges || ipdef->nhosts)) {
                    char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
                    if (!bridgeaddr)
                        goto cleanup;
998 999
                    virBufferAsprintf(&configbuf,
                                      "dhcp-range=%s,ra-only\n", bridgeaddr);
G
Gene Czarcinski 已提交
1000 1001
                    VIR_FREE(bridgeaddr);
                }
1002
            }
1003
        }
1004 1005
    }

1006 1007 1008
    if (!(*configstr = virBufferContentAndReset(&configbuf)))
        goto cleanup;

1009
    ret = 0;
G
Gene Czarcinski 已提交
1010

1011
cleanup:
1012
    virBufferFreeAndReset(&configbuf);
1013 1014 1015 1016
    VIR_FREE(record);
    VIR_FREE(recordPort);
    VIR_FREE(recordWeight);
    VIR_FREE(recordPriority);
1017
    return ret;
1018 1019
}

1020
/* build the dnsmasq command line */
1021 1022 1023
static int ATTRIBUTE_NONNULL(2)
networkBuildDhcpDaemonCommandLine(virNetworkObjPtr network,
                                  virCommandPtr *cmdout,
1024 1025
                                  char *pidfile, dnsmasqContext *dctx,
                                  dnsmasqCapsPtr caps)
1026
{
1027
    virCommandPtr cmd = NULL;
G
Gene Czarcinski 已提交
1028
    int ret = -1;
1029 1030
    char *configfile = NULL;
    char *configstr = NULL;
1031 1032

    network->dnsmasqPid = -1;
1033

1034 1035 1036 1037 1038 1039
    if (networkDnsmasqConfContents(network, pidfile, &configstr, dctx, caps) < 0)
        goto cleanup;
    if (!configstr)
        goto cleanup;

    /* construct the filename */
1040
    if (!(configfile = networkDnsmasqConfigFileName(network->def->name)))
1041 1042 1043 1044 1045 1046 1047
        goto cleanup;

    /* Write the file */
    if (virFileWriteStr(configfile, configstr, 0600) < 0) {
        virReportSystemError(errno,
                         _("couldn't write dnsmasq config file '%s'"),
                         configfile);
1048 1049 1050
        goto cleanup;
    }

1051 1052
    cmd = virCommandNew(dnsmasqCapsGetBinaryPath(caps));
    virCommandAddArgFormat(cmd, "--conf-file=%s", configfile);
1053
    *cmdout = cmd;
1054 1055 1056 1057 1058 1059
    ret = 0;
cleanup:
    return ret;
}

static int
1060
networkStartDhcpDaemon(virNetworkDriverStatePtr driver,
1061
                       virNetworkObjPtr network)
1062 1063 1064 1065
{
    virCommandPtr cmd = NULL;
    char *pidfile = NULL;
    int ret = -1;
1066
    dnsmasqContext *dctx = NULL;
1067

1068
    if (!virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, 0)) {
G
Gene Czarcinski 已提交
1069
        /* no IP addresses, so we don't need to run */
1070 1071 1072 1073
        ret = 0;
        goto cleanup;
    }

1074
    if (virFileMakePath(driverState->pidDir) < 0) {
1075
        virReportSystemError(errno,
1076
                             _("cannot create directory %s"),
1077
                             driverState->pidDir);
1078
        goto cleanup;
1079
    }
1080
    if (virFileMakePath(driverState->stateDir) < 0) {
1081
        virReportSystemError(errno,
1082
                             _("cannot create directory %s"),
1083
                             driverState->stateDir);
1084
        goto cleanup;
1085 1086
    }

1087
    if (!(pidfile = virPidFileBuildPath(driverState->pidDir,
1088
                                        network->def->name)))
1089
        goto cleanup;
1090

1091
    if (virFileMakePath(driverState->dnsmasqStateDir) < 0) {
1092
        virReportSystemError(errno,
1093
                             _("cannot create directory %s"),
1094
                             driverState->dnsmasqStateDir);
1095 1096 1097
        goto cleanup;
    }

1098
    dctx = dnsmasqContextNew(network->def->name, driverState->dnsmasqStateDir);
1099 1100 1101
    if (dctx == NULL)
        goto cleanup;

1102 1103 1104 1105
    dnsmasqCapsRefresh(&driver->dnsmasqCaps, false);

    ret = networkBuildDhcpDaemonCommandLine(network, &cmd, pidfile,
                                            dctx, driver->dnsmasqCaps);
1106 1107 1108 1109 1110
    if (ret < 0)
        goto cleanup;

    ret = dnsmasqSave(dctx);
    if (ret < 0)
1111
        goto cleanup;
1112

G
Guido Günther 已提交
1113 1114
    ret = virCommandRun(cmd, NULL);
    if (ret < 0) {
1115
        goto cleanup;
G
Guido Günther 已提交
1116
    }
1117 1118

    /*
1119 1120 1121 1122 1123
     * 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
1124 1125
     */

1126
    ret = virPidFileRead(driverState->pidDir, network->def->name,
1127 1128
                         &network->dnsmasqPid);
    if (ret < 0)
1129
        goto cleanup;
1130

1131 1132 1133
    ret = 0;
cleanup:
    VIR_FREE(pidfile);
1134
    virCommandFree(cmd);
1135
    dnsmasqContextFree(dctx);
1136 1137 1138
    return ret;
}

1139 1140
/* networkRefreshDhcpDaemon:
 *  Update dnsmasq config files, then send a SIGHUP so that it rereads
G
Gene Czarcinski 已提交
1141 1142
 *  them.   This only works for the dhcp-hostsfile and the
 *  addn-hosts file.
1143 1144 1145
 *
 *  Returns 0 on success, -1 on failure.
 */
1146
static int
1147
networkRefreshDhcpDaemon(virNetworkDriverStatePtr driver,
1148
                         virNetworkObjPtr network)
1149
{
1150 1151
    int ret = -1;
    size_t i;
G
Gene Czarcinski 已提交
1152
    virNetworkIpDefPtr ipdef, ipv4def, ipv6def;
1153
    dnsmasqContext *dctx = NULL;
1154

G
Gene Czarcinski 已提交
1155
    /* if no IP addresses specified, nothing to do */
1156
    if (!virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, 0))
G
Gene Czarcinski 已提交
1157 1158
        return 0;

1159 1160
    /* if there's no running dnsmasq, just start it */
    if (network->dnsmasqPid <= 0 || (kill(network->dnsmasqPid, 0) < 0))
1161
        return networkStartDhcpDaemon(driver, network);
1162

G
Gene Czarcinski 已提交
1163
    VIR_INFO("Refreshing dnsmasq for network %s", network->def->bridge);
1164 1165
    if (!(dctx = dnsmasqContextNew(network->def->name,
                                   driverState->dnsmasqStateDir))) {
G
Gene Czarcinski 已提交
1166
        goto cleanup;
1167
    }
G
Gene Czarcinski 已提交
1168 1169 1170 1171 1172 1173

    /* 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;
1174 1175 1176
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, i));
         i++) {
G
Gene Czarcinski 已提交
1177 1178
        if (!ipv4def && (ipdef->nranges || ipdef->nhosts))
            ipv4def = ipdef;
1179 1180
    }

G
Gene Czarcinski 已提交
1181
    ipv6def = NULL;
1182 1183 1184
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
         i++) {
G
Gene Czarcinski 已提交
1185 1186
        if (!ipv6def && (ipdef->nranges || ipdef->nhosts))
            ipv6def = ipdef;
1187 1188
    }

G
Gene Czarcinski 已提交
1189 1190 1191 1192 1193
    if (ipv4def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv4def) < 0))
           goto cleanup;

    if (ipv6def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv6def) < 0))
           goto cleanup;
1194

G
Gene Czarcinski 已提交
1195
    if (networkBuildDnsmasqHostsList(dctx, &network->def->dns) < 0)
1196 1197 1198
       goto cleanup;

    if ((ret = dnsmasqSave(dctx)) < 0)
1199
        goto cleanup;
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214

    ret = kill(network->dnsmasqPid, SIGHUP);
cleanup:
    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
1215
networkRestartDhcpDaemon(virNetworkDriverStatePtr driver,
1216
                         virNetworkObjPtr network)
1217 1218 1219 1220 1221 1222
{
    /* if there is a running dnsmasq, kill it */
    if (network->dnsmasqPid > 0) {
        networkKillDaemon(network->dnsmasqPid, "dnsmasq",
                          network->def->name);
        network->dnsmasqPid = -1;
1223
    }
1224
    /* now start dnsmasq if it should be started */
1225
    return networkStartDhcpDaemon(driver, network);
1226 1227
}

G
Gene Czarcinski 已提交
1228 1229 1230 1231 1232 1233
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";

1234 1235 1236
static int
networkRadvdConfContents(virNetworkObjPtr network, char **configstr)
{
E
Eric Blake 已提交
1237
    virBuffer configbuf = VIR_BUFFER_INITIALIZER;
1238 1239
    int ret = -1;
    size_t i;
1240
    virNetworkIpDefPtr ipdef;
G
Gene Czarcinski 已提交
1241
    bool v6present = false, dhcp6 = false;
1242 1243

    *configstr = NULL;
1244

G
Gene Czarcinski 已提交
1245
    /* Check if DHCPv6 is needed */
1246 1247 1248
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
         i++) {
G
Gene Czarcinski 已提交
1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261
        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;
    }

1262 1263 1264
    /* create radvd config file appropriate for this network;
     * IgnoreIfMissing allows radvd to start even when the bridge is down
     */
1265
    virBufferAsprintf(&configbuf, "interface %s\n"
1266 1267
                      "{\n"
                      "  AdvSendAdvert on;\n"
1268
                      "  IgnoreIfMissing on;\n"
G
Gene Czarcinski 已提交
1269 1270 1271 1272 1273
                      "  AdvManagedFlag %s;\n"
                      "%s",
                      network->def->bridge,
                      dhcp6 ? "on" : "off",
                      dhcp6 ? "\n" : radvd1);
1274 1275

    /* add a section for each IPv6 address in the config */
1276 1277 1278
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
         i++) {
1279 1280 1281 1282 1283
        int prefix;
        char *netaddr;

        prefix = virNetworkIpDefPrefix(ipdef);
        if (prefix < 0) {
1284 1285 1286
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("bridge '%s' has an invalid prefix"),
                           network->def->bridge);
1287 1288
            goto cleanup;
        }
1289
        if (!(netaddr = virSocketAddrFormat(&ipdef->address)))
1290
            goto cleanup;
1291
        virBufferAsprintf(&configbuf,
1292
                          "  prefix %s/%d\n"
G
Gene Czarcinski 已提交
1293 1294 1295
                          "  {\n%s  };\n",
                          netaddr, prefix,
                          dhcp6 ? radvd2 : radvd3);
1296 1297 1298
        VIR_FREE(netaddr);
    }

1299 1300 1301
    /* only create the string if we found at least one IPv6 address */
    if (v6present) {
        virBufferAddLit(&configbuf, "};\n");
1302

1303 1304 1305 1306 1307 1308 1309 1310
        if (virBufferError(&configbuf)) {
            virReportOOMError();
            goto cleanup;
        }
        if (!(*configstr = virBufferContentAndReset(&configbuf))) {
            virReportOOMError();
            goto cleanup;
        }
1311
    }
1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336

    ret = 0;
cleanup:
    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;
1337 1338 1339 1340
        goto cleanup;
    }

    /* construct the filename */
1341
    if (!(*configFile = networkRadvdConfigFileName(network->def->name)))
1342 1343
        goto cleanup;
    /* write the file */
1344
    if (virFileWriteStr(*configFile, configStr, 0600) < 0) {
1345 1346
        virReportSystemError(errno,
                             _("couldn't write radvd config file '%s'"),
1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358
                             *configFile);
        goto cleanup;
    }

    ret = 0;
cleanup:
    VIR_FREE(configStr);
    VIR_FREE(myConfigFile);
    return ret;
}

static int
1359
networkStartRadvd(virNetworkDriverStatePtr driver ATTRIBUTE_UNUSED,
G
Gene Czarcinski 已提交
1360
                        virNetworkObjPtr network)
1361 1362 1363 1364 1365 1366 1367 1368 1369
{
    char *pidfile = NULL;
    char *radvdpidbase = NULL;
    char *configfile = NULL;
    virCommandPtr cmd = NULL;
    int ret = -1;

    network->radvdPid = -1;

G
Gene Czarcinski 已提交
1370
    /* Is dnsmasq handling RA? */
1371
   if (DNSMASQ_RA_SUPPORT(driver->dnsmasqCaps)) {
G
Gene Czarcinski 已提交
1372 1373 1374 1375
        ret = 0;
        goto cleanup;
    }

1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
    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);
1387 1388 1389
        goto cleanup;
    }

1390
    if (virFileMakePath(driverState->pidDir) < 0) {
1391 1392
        virReportSystemError(errno,
                             _("cannot create directory %s"),
1393
                             driverState->pidDir);
1394 1395
        goto cleanup;
    }
1396
    if (virFileMakePath(driverState->radvdStateDir) < 0) {
1397 1398
        virReportSystemError(errno,
                             _("cannot create directory %s"),
1399
                             driverState->radvdStateDir);
1400 1401 1402 1403
        goto cleanup;
    }

    /* construct pidfile name */
1404
    if (!(radvdpidbase = networkRadvdPidfileBasename(network->def->name)))
1405
        goto cleanup;
1406
    if (!(pidfile = virPidFileBuildPath(driverState->pidDir, radvdpidbase)))
1407 1408 1409 1410 1411
        goto cleanup;

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

1412 1413 1414 1415
    /* 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
1416
     * virPidFileRead() below will fail if we use them).
1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431
     * 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;

1432
    if (virPidFileRead(driverState->pidDir, radvdpidbase, &network->radvdPid) < 0)
1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443
        goto cleanup;

    ret = 0;
cleanup:
    virCommandFree(cmd);
    VIR_FREE(configfile);
    VIR_FREE(radvdpidbase);
    VIR_FREE(pidfile);
    return ret;
}

1444
static int
1445
networkRefreshRadvd(virNetworkDriverStatePtr driver ATTRIBUTE_UNUSED,
1446
                    virNetworkObjPtr network)
1447
{
G
Gene Czarcinski 已提交
1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458
    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)) {
1459
            virPidFileDelete(driverState->pidDir, radvdpidbase);
G
Gene Czarcinski 已提交
1460 1461 1462 1463 1464 1465
            VIR_FREE(radvdpidbase);
        }
        network->radvdPid = -1;
        return 0;
    }

1466 1467
    /* if there's no running radvd, just start it */
    if (network->radvdPid <= 0 || (kill(network->radvdPid, 0) < 0))
G
Gene Czarcinski 已提交
1468
        return networkStartRadvd(driver, network);
1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480

    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);
}

1481 1482
#if 0
/* currently unused, so it causes a build error unless we #if it out */
1483
static int
1484
networkRestartRadvd(virNetworkDriverStatePtr driver,
1485
                    virNetworkObjPtr network)
1486 1487 1488 1489 1490 1491 1492 1493 1494
{
    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
1495
        if ((networkKillDaemon(network->radvdPid, "radvd",
1496 1497 1498
                               network->def->name) >= 0) &&
            ((radvdpidbase = networkRadvdPidfileBasename(network->def->name))
             != NULL)) {
1499
            virPidFileDelete(driverState->pidDir, radvdpidbase);
1500 1501 1502 1503 1504 1505 1506 1507 1508
            VIR_FREE(radvdpidbase);
        }
        network->radvdPid = -1;
    }
    /* now start radvd if it should be started */
    return networkStartRadvd(network);
}
#endif /* #if 0 */

1509 1510 1511 1512
/* SIGHUP/restart any dnsmasq or radvd daemons.
 * This should be called when libvirtd is restarted.
 */
static void
1513
networkRefreshDaemons(virNetworkDriverStatePtr driver)
1514
{
1515
    size_t i;
1516 1517 1518

    VIR_INFO("Refreshing network daemons");

1519
    for (i = 0; i < driver->networks.count; i++) {
1520 1521 1522 1523
        virNetworkObjPtr network = driver->networks.objs[i];

        virNetworkObjLock(network);
        if (virNetworkObjIsActive(network) &&
1524 1525 1526
            ((network->def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE))) {
1527 1528 1529 1530 1531 1532
            /* 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.
             */
1533 1534
            networkRefreshDhcpDaemon(driver, network);
            networkRefreshRadvd(driver, network);
1535 1536 1537 1538 1539
        }
        virNetworkObjUnlock(network);
    }
}

1540
static void
1541
networkReloadFirewallRules(virNetworkDriverStatePtr driver)
1542
{
1543
    size_t i;
1544

1545
    VIR_INFO("Reloading iptables rules");
1546

1547
    for (i = 0; i < driver->networks.count; i++) {
1548 1549 1550 1551
        virNetworkObjPtr network = driver->networks.objs[i];

        virNetworkObjLock(network);
        if (virNetworkObjIsActive(network) &&
1552 1553 1554
            ((network->def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE))) {
1555 1556 1557
            /* Only the three L3 network types that are configured by libvirt
             * need to have iptables rules reloaded.
             */
1558 1559
            networkRemoveFirewallRules(network);
            if (networkAddFirewallRules(network) < 0) {
1560 1561
                /* failed to add but already logged */
            }
1562
        }
1563
        virNetworkObjUnlock(network);
1564 1565 1566
    }
}

1567
/* Enable IP Forwarding. Return 0 for success, -1 for failure. */
1568
static int
1569
networkEnableIpForwarding(bool enableIPv4, bool enableIPv6)
1570
{
1571
    int ret = 0;
1572 1573 1574 1575 1576 1577 1578 1579 1580
#ifdef HAVE_SYSCTLBYNAME
    int enabled = 1;
    if (enableIPv4)
        ret = sysctlbyname("net.inet.ip.forwarding", NULL, 0,
                            &enabled, sizeof(enabled));
    if (enableIPv6 && ret == 0)
        ret = sysctlbyname("net.inet6.ip6.forwarding", NULL, 0,
                            &enabled, sizeof(enabled));
#else
1581 1582 1583 1584
    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);
1585
#endif
1586
    return ret;
1587 1588
}

1589 1590
#define SYSCTL_PATH "/proc/sys"

1591 1592
static int
networkSetIPv6Sysctls(virNetworkObjPtr network)
1593 1594 1595 1596
{
    char *field = NULL;
    int ret = -1;

1597 1598 1599 1600 1601
    if (!virNetworkDefGetIpByIndex(network->def, AF_INET6, 0)) {
        /* Only set disable_ipv6 if there are no ipv6 addresses defined for
         * the network.
         */
        if (virAsprintf(&field, SYSCTL_PATH "/net/ipv6/conf/%s/disable_ipv6",
1602
                        network->def->bridge) < 0)
1603
            goto cleanup;
1604

1605 1606 1607 1608 1609 1610
        if (access(field, W_OK) < 0 && errno == ENOENT) {
            VIR_DEBUG("ipv6 appears to already be disabled on %s",
                      network->def->bridge);
            ret = 0;
            goto cleanup;
        }
1611

1612 1613 1614 1615 1616 1617 1618
        if (virFileWriteStr(field, "1", 0) < 0) {
            virReportSystemError(errno,
                                 _("cannot write to %s to disable IPv6 on bridge %s"),
                                 field, network->def->bridge);
            goto cleanup;
        }
        VIR_FREE(field);
1619 1620
    }

1621 1622 1623 1624 1625 1626 1627 1628
    /* The rest of the ipv6 sysctl tunables should always be set,
     * whether or not we're using ipv6 on this bridge.
     */

    /* 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",
1629
                    network->def->bridge) < 0)
1630 1631
        goto cleanup;

1632
    if (virFileWriteStr(field, "0", 0) < 0) {
1633
        virReportSystemError(errno,
1634 1635 1636 1637 1638
                             _("cannot disable %s"), field);
        goto cleanup;
    }
    VIR_FREE(field);

1639 1640 1641 1642
    /* 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",
1643
                    network->def->bridge) < 0)
1644 1645
        goto cleanup;

1646
    if (virFileWriteStr(field, "0", 0) < 0) {
1647
        virReportSystemError(errno,
1648
                             _("cannot disable %s"), field);
1649 1650 1651 1652 1653 1654 1655 1656 1657
        goto cleanup;
    }

    ret = 0;
cleanup:
    VIR_FREE(field);
    return ret;
}

1658
/* add an IP address to a bridge */
1659
static int
D
Daniel P. Berrange 已提交
1660
networkAddAddrToBridge(virNetworkObjPtr network,
1661
                       virNetworkIpDefPtr ipdef)
1662
{
1663 1664 1665
    int prefix = virNetworkIpDefPrefix(ipdef);

    if (prefix < 0) {
1666 1667 1668
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("bridge '%s' has an invalid netmask or IP address"),
                       network->def->bridge);
1669 1670 1671
        return -1;
    }

1672 1673
    if (virNetDevSetIPv4Address(network->def->bridge,
                                &ipdef->address, prefix) < 0)
1674 1675 1676 1677 1678
        return -1;

    return 0;
}

1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727
/* 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,
                                    (VIR_SOCKET_ADDR_IS_FAMILY(addr,AF_INET)
                                     ? "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) &&
                virSocketAddrEqual(mask, &zero)))
            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;
}

1728
static int
1729
networkStartNetworkVirtual(virNetworkDriverStatePtr driver,
1730 1731
                          virNetworkObjPtr network)
{
1732
    size_t i;
1733
    bool v4present = false, v6present = false;
1734 1735
    virErrorPtr save_err = NULL;
    virNetworkIpDefPtr ipdef;
1736
    virNetworkRouteDefPtr routedef;
1737
    char *macTapIfName = NULL;
1738
    int tapfd = -1;
1739

1740 1741
    /* Check to see if any network IP collides with an existing route */
    if (networkCheckRouteCollision(network) < 0)
1742 1743
        return -1;

1744
    /* Create and configure the bridge device */
1745
    if (virNetDevBridgeCreate(network->def->bridge) < 0)
1746 1747
        return -1;

1748 1749 1750 1751 1752 1753 1754 1755
    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);
1756
        if (!macTapIfName)
1757
            goto err0;
1758
        /* Keep tun fd open and interface up to allow for IPv6 DAD to happen */
1759
        if (virNetDevTapCreateInBridgePort(network->def->bridge,
1760
                                           &macTapIfName, &network->def->mac,
1761
                                           NULL, &tapfd, 1, NULL, NULL,
1762 1763 1764
                                           VIR_NETDEV_TAP_CREATE_USE_MAC_FOR_BRIDGE |
                                           VIR_NETDEV_TAP_CREATE_IFUP |
                                           VIR_NETDEV_TAP_CREATE_PERSIST) < 0) {
1765 1766 1767 1768 1769
            VIR_FREE(macTapIfName);
            goto err0;
        }
    }

1770
    /* Set bridge options */
1771 1772 1773 1774

    /* delay is configured in seconds, but virNetDevBridgeSetSTPDelay
     * expects milliseconds
     */
1775
    if (virNetDevBridgeSetSTPDelay(network->def->bridge,
1776
                                   network->def->delay * 1000) < 0)
1777
        goto err1;
1778

1779
    if (virNetDevBridgeSetSTP(network->def->bridge,
1780
                              network->def->stp ? true : false) < 0)
1781
        goto err1;
1782

1783 1784 1785 1786
    /* Disable IPv6 on the bridge if there are no IPv6 addresses
     * defined, and set other IPv6 sysctl tunables appropriately.
     */
    if (networkSetIPv6Sysctls(network) < 0)
1787
        goto err1;
1788

1789
    /* Add "once per network" rules */
1790
    if (networkAddFirewallRules(network) < 0)
1791 1792
        goto err1;

1793 1794 1795
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, i));
         i++) {
1796
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET))
1797
            v4present = true;
1798
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
1799
            v6present = true;
1800

1801
        /* Add the IP address/netmask to the bridge */
D
Daniel P. Berrange 已提交
1802
        if (networkAddAddrToBridge(network, ipdef) < 0) {
1803
            goto err2;
1804
        }
1805 1806
    }

1807
    /* Bring up the bridge interface */
1808
    if (virNetDevSetOnline(network->def->bridge, 1) < 0)
1809
        goto err2;
1810

1811 1812
    for (i = 0; i < network->def->nroutes; i++) {
        routedef = &network->def->routes[i];
1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823
        /* 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 */
            }
        }
    }

1824 1825
    /* If forward.type != NONE, turn on global IP forwarding */
    if (network->def->forward.type != VIR_NETWORK_FORWARD_NONE &&
1826
        networkEnableIpForwarding(v4present, v6present) < 0) {
1827
        virReportSystemError(errno, "%s",
1828
                             _("failed to enable IP forwarding"));
1829
        goto err3;
1830 1831
    }

1832

1833
    /* start dnsmasq if there are any IP addresses (v4 or v6) */
1834 1835
    if ((v4present || v6present) &&
        networkStartDhcpDaemon(driver, network) < 0)
1836
        goto err3;
1837

1838
    /* start radvd if there are any ipv6 addresses */
G
Gene Czarcinski 已提交
1839
    if (v6present && networkStartRadvd(driver, network) < 0)
1840 1841
        goto err4;

1842 1843 1844 1845 1846 1847 1848 1849 1850
    /* 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);
    }

1851 1852
    if (virNetDevBandwidthSet(network->def->bridge,
                              network->def->bandwidth, true) < 0) {
1853 1854 1855
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot set bandwidth limits on %s"),
                       network->def->bridge);
1856 1857 1858
        goto err5;
    }

1859
    VIR_FREE(macTapIfName);
1860 1861 1862

    return 0;

1863
 err5:
1864
    virNetDevBandwidthClear(network->def->bridge);
1865

1866 1867 1868 1869
 err4:
    if (!save_err)
        save_err = virSaveLastError();

1870 1871 1872 1873 1874
    if (network->dnsmasqPid > 0) {
        kill(network->dnsmasqPid, SIGTERM);
        network->dnsmasqPid = -1;
    }

1875 1876 1877
 err3:
    if (!save_err)
        save_err = virSaveLastError();
1878
    ignore_value(virNetDevSetOnline(network->def->bridge, 0));
1879

1880 1881 1882
 err2:
    if (!save_err)
        save_err = virSaveLastError();
1883
    networkRemoveFirewallRules(network);
1884 1885

 err1:
1886 1887 1888
    if (!save_err)
        save_err = virSaveLastError();

H
Hu Tao 已提交
1889
    if (macTapIfName) {
1890
        VIR_FORCE_CLOSE(tapfd);
H
Hu Tao 已提交
1891 1892 1893
        ignore_value(virNetDevTapDelete(macTapIfName));
        VIR_FREE(macTapIfName);
    }
1894 1895

 err0:
1896 1897
    if (!save_err)
        save_err = virSaveLastError();
1898
    ignore_value(virNetDevBridgeDelete(network->def->bridge));
1899

1900 1901 1902 1903
    if (save_err) {
        virSetError(save_err);
        virFreeError(save_err);
    }
1904
    /* coverity[leaked_handle] - 'tapfd' is not leaked */
1905 1906 1907
    return -1;
}

1908
static int networkShutdownNetworkVirtual(virNetworkDriverStatePtr driver ATTRIBUTE_UNUSED,
R
Roman Bogorodskiy 已提交
1909
                                         virNetworkObjPtr network)
1910
{
1911
    virNetDevBandwidthClear(network->def->bridge);
1912

1913 1914 1915 1916 1917
    if (network->radvdPid > 0) {
        char *radvdpidbase;

        kill(network->radvdPid, SIGTERM);
        /* attempt to delete the pidfile we created */
1918
        if ((radvdpidbase = networkRadvdPidfileBasename(network->def->name))) {
1919
            virPidFileDelete(driverState->pidDir, radvdpidbase);
1920 1921 1922 1923
            VIR_FREE(radvdpidbase);
        }
    }

1924 1925 1926
    if (network->dnsmasqPid > 0)
        kill(network->dnsmasqPid, SIGTERM);

1927
    if (network->def->mac_specified) {
1928
        char *macTapIfName = networkBridgeDummyNicName(network->def->bridge);
1929
        if (macTapIfName) {
1930
            ignore_value(virNetDevTapDelete(macTapIfName));
1931 1932 1933 1934
            VIR_FREE(macTapIfName);
        }
    }

1935
    ignore_value(virNetDevSetOnline(network->def->bridge, 0));
1936

1937
    networkRemoveFirewallRules(network);
1938

1939
    ignore_value(virNetDevBridgeDelete(network->def->bridge));
1940

1941
    /* See if its still alive and really really kill it */
1942
    if (network->dnsmasqPid > 0 &&
1943
        (kill(network->dnsmasqPid, 0) == 0))
1944 1945
        kill(network->dnsmasqPid, SIGKILL);
    network->dnsmasqPid = -1;
1946 1947 1948 1949 1950 1951

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

1952 1953 1954 1955
    return 0;
}

static int
1956
networkStartNetworkExternal(virNetworkDriverStatePtr driver ATTRIBUTE_UNUSED,
1957 1958 1959
                            virNetworkObjPtr network ATTRIBUTE_UNUSED)
{
    /* put anything here that needs to be done each time a network of
1960
     * type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is started. On
1961 1962 1963 1964 1965 1966
     * failure, undo anything you've done, and return -1. On success
     * return 0.
     */
    return 0;
}

1967
static int networkShutdownNetworkExternal(virNetworkDriverStatePtr driver ATTRIBUTE_UNUSED,
1968 1969 1970
                                        virNetworkObjPtr network ATTRIBUTE_UNUSED)
{
    /* put anything here that needs to be done each time a network of
1971
     * type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is shutdown. On
1972 1973 1974 1975 1976 1977 1978
     * failure, undo anything you've done, and return -1. On success
     * return 0.
     */
    return 0;
}

static int
1979
networkStartNetwork(virNetworkDriverStatePtr driver,
1980 1981 1982 1983 1984
                    virNetworkObjPtr network)
{
    int ret = 0;

    if (virNetworkObjIsActive(network)) {
1985 1986
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("network is already active"));
1987 1988 1989
        return -1;
    }

1990 1991 1992
    if (virNetworkObjSetDefTransient(network, true) < 0)
        return -1;

1993
    switch (network->def->forward.type) {
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

    case VIR_NETWORK_FORWARD_NONE:
    case VIR_NETWORK_FORWARD_NAT:
    case VIR_NETWORK_FORWARD_ROUTE:
        ret = networkStartNetworkVirtual(driver, network);
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
    case VIR_NETWORK_FORWARD_PRIVATE:
    case VIR_NETWORK_FORWARD_VEPA:
    case VIR_NETWORK_FORWARD_PASSTHROUGH:
2005
    case VIR_NETWORK_FORWARD_HOSTDEV:
2006 2007 2008 2009
        ret = networkStartNetworkExternal(driver, network);
        break;
    }

2010 2011
    if (ret < 0) {
        virNetworkObjUnsetDefTransient(network);
2012
        return ret;
2013
    }
2014 2015 2016 2017

    /* Persist the live configuration now that anything autogenerated
     * is setup.
     */
2018 2019
    if ((ret = virNetworkSaveStatus(driverState->stateDir,
                                    network)) < 0) {
2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037
        goto error;
    }

    VIR_INFO("Starting up network '%s'", network->def->name);
    network->active = 1;

error:
    if (ret < 0) {
        virErrorPtr save_err = virSaveLastError();
        int save_errno = errno;
        networkShutdownNetwork(driver, network);
        virSetError(save_err);
        virFreeError(save_err);
        errno = save_errno;
    }
    return ret;
}

2038
static int networkShutdownNetwork(virNetworkDriverStatePtr driver,
2039 2040 2041 2042 2043 2044 2045 2046 2047 2048
                                        virNetworkObjPtr network)
{
    int ret = 0;
    char *stateFile;

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

    if (!virNetworkObjIsActive(network))
        return 0;

2049 2050
    stateFile = virNetworkConfigFile(driverState->stateDir,
                                     network->def->name);
2051 2052 2053 2054 2055 2056
    if (!stateFile)
        return -1;

    unlink(stateFile);
    VIR_FREE(stateFile);

2057
    switch (network->def->forward.type) {
2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068

    case VIR_NETWORK_FORWARD_NONE:
    case VIR_NETWORK_FORWARD_NAT:
    case VIR_NETWORK_FORWARD_ROUTE:
        ret = networkShutdownNetworkVirtual(driver, network);
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
    case VIR_NETWORK_FORWARD_PRIVATE:
    case VIR_NETWORK_FORWARD_VEPA:
    case VIR_NETWORK_FORWARD_PASSTHROUGH:
2069
    case VIR_NETWORK_FORWARD_HOSTDEV:
2070 2071 2072 2073
        ret = networkShutdownNetworkExternal(driver, network);
        break;
    }

2074
    network->active = 0;
2075
    virNetworkObjUnsetDefTransient(network);
2076
    return ret;
2077 2078 2079
}


2080 2081
static virNetworkPtr networkLookupByUUID(virConnectPtr conn,
                                         const unsigned char *uuid) {
2082
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2083 2084
    virNetworkObjPtr network;
    virNetworkPtr ret = NULL;
2085

2086
    networkDriverLock(driver);
2087
    network = virNetworkFindByUUID(&driver->networks, uuid);
2088
    networkDriverUnlock(driver);
2089
    if (!network) {
2090 2091
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2092
        goto cleanup;
2093 2094
    }

2095 2096 2097
    if (virNetworkLookupByUUIDEnsureACL(conn, network->def) < 0)
        goto cleanup;

2098 2099 2100
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

cleanup:
2101 2102
    if (network)
        virNetworkObjUnlock(network);
2103
    return ret;
2104 2105
}

2106 2107
static virNetworkPtr networkLookupByName(virConnectPtr conn,
                                         const char *name) {
2108
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2109 2110 2111
    virNetworkObjPtr network;
    virNetworkPtr ret = NULL;

2112
    networkDriverLock(driver);
2113
    network = virNetworkFindByName(&driver->networks, name);
2114
    networkDriverUnlock(driver);
2115
    if (!network) {
2116 2117
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"), name);
2118
        goto cleanup;
2119 2120
    }

2121 2122 2123
    if (virNetworkLookupByNameEnsureACL(conn, network->def) < 0)
        goto cleanup;

2124 2125 2126
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

cleanup:
2127 2128
    if (network)
        virNetworkObjUnlock(network);
2129
    return ret;
2130 2131
}

2132 2133 2134
static virDrvOpenStatus networkOpen(virConnectPtr conn,
                                    virConnectAuthPtr auth ATTRIBUTE_UNUSED,
                                    unsigned int flags)
2135 2136 2137
{
    virCheckFlags(VIR_CONNECT_RO, VIR_DRV_OPEN_ERROR);

2138 2139 2140 2141 2142 2143 2144
    if (!driverState)
        return VIR_DRV_OPEN_DECLINED;

    conn->networkPrivateData = driverState;
    return VIR_DRV_OPEN_SUCCESS;
}

2145
static int networkClose(virConnectPtr conn) {
2146 2147 2148 2149
    conn->networkPrivateData = NULL;
    return 0;
}

2150
static int networkConnectNumOfNetworks(virConnectPtr conn) {
2151 2152
    int nactive = 0;
    size_t i;
2153
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2154

2155 2156 2157
    if (virConnectNumOfNetworksEnsureACL(conn) < 0)
        return -1;

2158
    networkDriverLock(driver);
2159
    for (i = 0; i < driver->networks.count; i++) {
2160 2161 2162 2163
        virNetworkObjPtr obj = driver->networks.objs[i];
        virNetworkObjLock(obj);
        if (virConnectNumOfNetworksCheckACL(conn, obj->def) &&
            virNetworkObjIsActive(obj))
2164
            nactive++;
2165
        virNetworkObjUnlock(obj);
2166 2167
    }
    networkDriverUnlock(driver);
2168

2169 2170 2171
    return nactive;
}

2172
static int networkConnectListNetworks(virConnectPtr conn, char **const names, int nnames) {
2173
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2174 2175
    int got = 0;
    size_t i;
2176

2177 2178 2179
    if (virConnectListNetworksEnsureACL(conn) < 0)
        return -1;

2180
    networkDriverLock(driver);
2181
    for (i = 0; i < driver->networks.count && got < nnames; i++) {
2182 2183 2184 2185 2186 2187
        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);
2188 2189 2190 2191
                goto cleanup;
            }
            got++;
        }
2192
        virNetworkObjUnlock(obj);
2193
    }
2194 2195
    networkDriverUnlock(driver);

2196 2197 2198
    return got;

 cleanup:
2199
    networkDriverUnlock(driver);
2200
    for (i = 0; i < got; i++)
2201 2202 2203 2204
        VIR_FREE(names[i]);
    return -1;
}

2205
static int networkConnectNumOfDefinedNetworks(virConnectPtr conn) {
2206 2207
    int ninactive = 0;
    size_t i;
2208
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2209

2210 2211 2212
    if (virConnectNumOfDefinedNetworksEnsureACL(conn) < 0)
        return -1;

2213
    networkDriverLock(driver);
2214
    for (i = 0; i < driver->networks.count; i++) {
2215 2216 2217 2218
        virNetworkObjPtr obj = driver->networks.objs[i];
        virNetworkObjLock(obj);
        if (virConnectNumOfDefinedNetworksCheckACL(conn, obj->def) &&
            !virNetworkObjIsActive(obj))
2219
            ninactive++;
2220
        virNetworkObjUnlock(obj);
2221 2222
    }
    networkDriverUnlock(driver);
2223

2224 2225 2226
    return ninactive;
}

2227
static int networkConnectListDefinedNetworks(virConnectPtr conn, char **const names, int nnames) {
2228
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2229 2230
    int got = 0;
    size_t i;
2231

2232 2233 2234
    if (virConnectListDefinedNetworksEnsureACL(conn) < 0)
        return -1;

2235
    networkDriverLock(driver);
2236
    for (i = 0; i < driver->networks.count && got < nnames; i++) {
2237 2238 2239 2240 2241 2242
        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);
2243 2244 2245 2246
                goto cleanup;
            }
            got++;
        }
2247
        virNetworkObjUnlock(obj);
2248
    }
2249
    networkDriverUnlock(driver);
2250 2251 2252
    return got;

 cleanup:
2253
    networkDriverUnlock(driver);
2254
    for (i = 0; i < got; i++)
2255 2256 2257 2258
        VIR_FREE(names[i]);
    return -1;
}

2259
static int
2260 2261 2262
networkConnectListAllNetworks(virConnectPtr conn,
                              virNetworkPtr **nets,
                              unsigned int flags)
2263
{
2264
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2265 2266 2267 2268
    int ret = -1;

    virCheckFlags(VIR_CONNECT_LIST_NETWORKS_FILTERS_ALL, -1);

2269 2270 2271
    if (virConnectListAllNetworksEnsureACL(conn) < 0)
        goto cleanup;

2272
    networkDriverLock(driver);
2273 2274 2275
    ret = virNetworkObjListExport(conn, driver->networks, nets,
                                  virConnectListAllNetworksCheckACL,
                                  flags);
2276 2277
    networkDriverUnlock(driver);

2278
cleanup:
2279 2280
    return ret;
}
2281 2282 2283 2284 2285 2286

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

2287 2288
    if (!(obj = networkObjFromNetwork(net)))
        return ret;
2289 2290 2291 2292

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

2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305
    ret = virNetworkObjIsActive(obj);

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

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

2306 2307
    if (!(obj = networkObjFromNetwork(net)))
        return ret;
2308 2309 2310 2311

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

2312 2313 2314 2315 2316 2317 2318 2319 2320
    ret = obj->persistent;

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


2321
static int
2322
networkValidate(virNetworkDriverStatePtr driver,
2323 2324
                virNetworkDefPtr def,
                bool check_active)
2325
{
2326
    size_t i;
2327 2328
    bool vlanUsed, vlanAllowed, badVlanUse = false;
    virPortGroupDefPtr defaultPortGroup = NULL;
2329
    virNetworkIpDefPtr ipdef;
G
Gene Czarcinski 已提交
2330
    bool ipv4def = false, ipv6def = false;
2331 2332 2333 2334 2335 2336 2337 2338

    /* 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
     */
2339 2340 2341
    if (def->forward.type == VIR_NETWORK_FORWARD_NONE ||
        def->forward.type == VIR_NETWORK_FORWARD_NAT ||
        def->forward.type == VIR_NETWORK_FORWARD_ROUTE) {
2342 2343 2344 2345 2346

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

        virNetworkSetBridgeMacAddr(def);
2347 2348 2349 2350 2351 2352 2353 2354 2355
    } else {
        /* They are also the only types that currently support setting
         * an IP address for the host-side device (bridge)
         */
        if (virNetworkDefGetIpByIndex(def, AF_UNSPEC, 0)) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <ip> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
2356
                           virNetworkForwardTypeToString(def->forward.type));
2357 2358
            return -1;
        }
2359
        if (def->dns.ntxts || def->dns.nhosts || def->dns.nsrvs) {
2360 2361 2362 2363
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <dns> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
2364
                           virNetworkForwardTypeToString(def->forward.type));
2365 2366 2367 2368 2369 2370 2371
            return -1;
        }
        if (def->domain) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <domain> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
2372
                           virNetworkForwardTypeToString(def->forward.type));
2373 2374
            return -1;
        }
2375 2376
    }

G
Gene Czarcinski 已提交
2377 2378 2379
    /* We only support dhcp on one IPv4 address and
     * on one IPv6 address per defined network
     */
2380 2381 2382
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(def, AF_UNSPEC, i));
         i++) {
G
Gene Czarcinski 已提交
2383 2384 2385 2386 2387
        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 -- "
2388 2389
                                 "dhcp is supported only for a "
                                 "single IPv4 address on each network"));
G
Gene Czarcinski 已提交
2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406
                    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;
                }
2407 2408 2409
            }
        }
    }
2410 2411 2412 2413 2414 2415

    /* 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.
     */

2416
    vlanAllowed = ((def->forward.type == VIR_NETWORK_FORWARD_BRIDGE &&
2417
                   def->virtPortProfile &&
2418 2419 2420
                   def->virtPortProfile->virtPortType
                    == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) ||
                   def->forward.type == VIR_NETWORK_FORWARD_HOSTDEV);
2421 2422

    vlanUsed = def->vlan.nTags > 0;
2423 2424
    for (i = 0; i < def->nPortGroups; i++) {
        if (vlanUsed || def->portGroups[i].vlan.nTags > 0) {
2425 2426 2427 2428 2429
            /* 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.
             */
2430
            if (def->portGroups[i].virtPortProfile) {
2431
                if (def->forward.type != VIR_NETWORK_FORWARD_BRIDGE ||
2432
                    def->portGroups[i].virtPortProfile->virtPortType
2433 2434 2435 2436 2437 2438 2439
                    != VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
                    badVlanUse = true;
                }
            } else if (!vlanAllowed) {
                /* virtualport taken from base network definition */
                badVlanUse = true;
            }
2440
        }
2441
        if (def->portGroups[i].isDefault) {
2442 2443 2444 2445 2446
            if (defaultPortGroup) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("network '%s' has multiple default "
                                 "<portgroup> elements (%s and %s), "
                                 "but only one default is allowed"),
2447
                               def->name, defaultPortGroup->name,
2448
                               def->portGroups[i].name);
2449
                return -1;
2450
            }
2451
            defaultPortGroup = &def->portGroups[i];
2452
        }
2453
    }
2454 2455 2456 2457 2458 2459 2460
    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.
         */
2461 2462 2463 2464 2465 2466 2467 2468 2469
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       _("<vlan> element specified for network %s, "
                         "whose type doesn't support vlan configuration"),
                       def->name);
        return -1;
    }
    return 0;
}

2470
static virNetworkPtr networkCreateXML(virConnectPtr conn, const char *xml) {
2471
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2472
    virNetworkDefPtr def;
2473
    virNetworkObjPtr network = NULL;
2474
    virNetworkPtr ret = NULL;
2475

2476 2477
    networkDriverLock(driver);

2478
    if (!(def = virNetworkDefParseString(xml)))
2479
        goto cleanup;
2480

2481 2482 2483
    if (virNetworkCreateXMLEnsureACL(conn, def) < 0)
        goto cleanup;

2484
    if (networkValidate(driver, def, true) < 0)
2485 2486
       goto cleanup;

2487 2488 2489 2490
    /* NB: "live" is false because this transient network hasn't yet
     * been started
     */
    if (!(network = virNetworkAssignDef(&driver->networks, def, false)))
2491 2492
        goto cleanup;
    def = NULL;
2493

2494
    if (networkStartNetwork(driver, network) < 0) {
2495 2496
        virNetworkRemoveInactive(&driver->networks,
                                 network);
2497
        network = NULL;
2498
        goto cleanup;
2499 2500
    }

2501
    VIR_INFO("Creating network '%s'", network->def->name);
2502 2503 2504 2505
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

cleanup:
    virNetworkDefFree(def);
2506 2507 2508
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
2509
    return ret;
2510 2511
}

2512
static virNetworkPtr networkDefineXML(virConnectPtr conn, const char *xml) {
2513
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2514
    virNetworkDefPtr def = NULL;
2515
    bool freeDef = true;
2516
    virNetworkObjPtr network = NULL;
2517
    virNetworkPtr ret = NULL;
2518

2519 2520
    networkDriverLock(driver);

2521
    if (!(def = virNetworkDefParseString(xml)))
2522
        goto cleanup;
2523

2524 2525 2526
    if (virNetworkDefineXMLEnsureACL(conn, def) < 0)
        goto cleanup;

2527
    if (networkValidate(driver, def, false) < 0)
2528 2529
       goto cleanup;

2530 2531 2532 2533 2534 2535 2536 2537
    if ((network = virNetworkFindByName(&driver->networks, def->name))) {
        network->persistent = 1;
        if (virNetworkObjAssignDef(network, def, false) < 0)
            goto cleanup;
    } else {
        if (!(network = virNetworkAssignDef(&driver->networks, def, false)))
            goto cleanup;
    }
2538

2539 2540 2541
    /* define makes the network persistent - always */
    network->persistent = 1;

2542 2543
    /* def was asigned */
    freeDef = false;
2544 2545

    if (virNetworkSaveConfig(driver->networkConfigDir, def) < 0) {
2546 2547 2548 2549 2550 2551 2552 2553
        if (!virNetworkObjIsActive(network)) {
            virNetworkRemoveInactive(&driver->networks, network);
            network = NULL;
            goto cleanup;
        }
        network->persistent = 0;
        virNetworkDefFree(network->newDef);
        network->newDef = NULL;
2554 2555 2556
        goto cleanup;
    }

2557 2558
    VIR_INFO("Defining network '%s'", def->name);
    ret = virGetNetwork(conn, def->name, def->uuid);
2559 2560

cleanup:
2561 2562
    if (freeDef)
       virNetworkDefFree(def);
2563 2564 2565
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
2566
    return ret;
2567 2568
}

2569 2570
static int
networkUndefine(virNetworkPtr net) {
2571
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
2572
    virNetworkObjPtr network;
2573
    int ret = -1;
2574
    bool active = false;
2575

2576 2577
    networkDriverLock(driver);

2578
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
2579
    if (!network) {
2580 2581
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2582
        goto cleanup;
2583 2584
    }

2585 2586 2587
    if (virNetworkUndefineEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2588 2589
    if (virNetworkObjIsActive(network))
        active = true;
2590

2591
    if (virNetworkDeleteConfig(driver->networkConfigDir,
2592 2593
                               driver->networkAutostartDir,
                               network) < 0)
2594
        goto cleanup;
2595

2596 2597
    /* make the network transient */
    network->persistent = 0;
2598
    network->autostart = 0;
2599 2600 2601
    virNetworkDefFree(network->newDef);
    network->newDef = NULL;

2602
    VIR_INFO("Undefining network '%s'", network->def->name);
2603 2604 2605 2606 2607
    if (!active) {
        if (networkRemoveInactive(driver, network) < 0) {
            network = NULL;
            goto cleanup;
        }
2608
        network = NULL;
2609 2610
    }

2611
    ret = 0;
2612

2613
cleanup:
2614 2615 2616
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
2617
    return ret;
2618 2619
}

2620 2621 2622 2623 2624 2625 2626 2627
static int
networkUpdate(virNetworkPtr net,
              unsigned int command,
              unsigned int section,
              int parentIndex,
              const char *xml,
              unsigned int flags)
{
2628
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
2629
    virNetworkObjPtr network = NULL;
2630 2631
    int isActive, ret = -1;
    size_t i;
2632 2633 2634
    virNetworkIpDefPtr ipdef;
    bool oldDhcpActive = false;

2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648

    virCheckFlags(VIR_NETWORK_UPDATE_AFFECT_LIVE |
                  VIR_NETWORK_UPDATE_AFFECT_CONFIG,
                  -1);

    networkDriverLock(driver);

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

2649 2650 2651
    if (virNetworkUpdateEnsureACL(net->conn, network->def, flags) < 0)
        goto cleanup;

2652
    /* see if we are listening for dhcp pre-modification */
2653 2654 2655
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, i));
         i++) {
2656 2657 2658 2659 2660 2661
        if (ipdef->nranges || ipdef->nhosts) {
            oldDhcpActive = true;
            break;
        }
    }

2662 2663 2664 2665
    /* VIR_NETWORK_UPDATE_AFFECT_CURRENT means "change LIVE if network
     * is active, else change CONFIG
    */
    isActive = virNetworkObjIsActive(network);
2666 2667
    if ((flags & (VIR_NETWORK_UPDATE_AFFECT_LIVE |
                  VIR_NETWORK_UPDATE_AFFECT_CONFIG)) ==
2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699
        VIR_NETWORK_UPDATE_AFFECT_CURRENT) {
        if (isActive)
            flags |= VIR_NETWORK_UPDATE_AFFECT_LIVE;
        else
            flags |= VIR_NETWORK_UPDATE_AFFECT_CONFIG;
    }

    /* update the network config in memory/on disk */
    if (virNetworkObjUpdate(network, command, section, parentIndex, xml, flags) < 0)
        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.
             */
2700
            if (networkRestartDhcpDaemon(driver, network) < 0)
2701 2702
                goto cleanup;

2703 2704 2705 2706 2707 2708 2709 2710
        } 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;

2711 2712 2713
            for (i = 0;
                 (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, i));
                 i++) {
2714 2715 2716 2717 2718 2719 2720
                if (ipdef->nranges || ipdef->nhosts) {
                    newDhcpActive = true;
                    break;
                }
            }

            if ((newDhcpActive != oldDhcpActive &&
2721 2722
                 networkRestartDhcpDaemon(driver, network) < 0) ||
                networkRefreshDhcpDaemon(driver, network) < 0) {
2723 2724 2725 2726
                goto cleanup;
            }

        } else if (section == VIR_NETWORK_SECTION_DNS_HOST ||
2727 2728 2729 2730 2731 2732
                   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.
             */
2733
            if (networkRefreshDhcpDaemon(driver, network) < 0)
2734 2735 2736 2737 2738 2739 2740 2741
                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.
             */
2742
            if (networkRefreshRadvd(driver, network) < 0)
2743 2744 2745
                goto cleanup;
        }

2746 2747 2748
        if ((section == VIR_NETWORK_SECTION_IP ||
             section == VIR_NETWORK_SECTION_FORWARD ||
             section == VIR_NETWORK_SECTION_FORWARD_INTERFACE) &&
2749 2750 2751
           (network->def->forward.type == VIR_NETWORK_FORWARD_NONE ||
            network->def->forward.type == VIR_NETWORK_FORWARD_NAT ||
            network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE)) {
2752
            /* these could affect the iptables rules */
2753 2754
            networkRemoveFirewallRules(network);
            if (networkAddFirewallRules(network) < 0)
2755 2756 2757 2758 2759
                goto cleanup;

        }

        /* save current network state to disk */
2760 2761
        if ((ret = virNetworkSaveStatus(driverState->stateDir,
                                        network)) < 0) {
2762
            goto cleanup;
2763
        }
2764 2765 2766 2767 2768 2769 2770 2771 2772
    }
    ret = 0;
cleanup:
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
    return ret;
}

2773
static int networkCreate(virNetworkPtr net) {
2774
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
2775 2776
    virNetworkObjPtr network;
    int ret = -1;
2777

2778
    networkDriverLock(driver);
2779
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
2780

2781
    if (!network) {
2782 2783
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2784
        goto cleanup;
2785 2786
    }

2787 2788 2789
    if (virNetworkCreateEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2790
    ret = networkStartNetwork(driver, network);
2791 2792

cleanup:
2793 2794
    if (network)
        virNetworkObjUnlock(network);
2795
    networkDriverUnlock(driver);
2796
    return ret;
2797 2798 2799
}

static int networkDestroy(virNetworkPtr net) {
2800
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
2801 2802
    virNetworkObjPtr network;
    int ret = -1;
2803

2804
    networkDriverLock(driver);
2805
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
2806

2807
    if (!network) {
2808 2809
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2810
        goto cleanup;
2811 2812
    }

2813 2814 2815
    if (virNetworkDestroyEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
2816
    if (!virNetworkObjIsActive(network)) {
2817 2818
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("network is not active"));
2819 2820 2821
        goto cleanup;
    }

2822 2823 2824
    if ((ret = networkShutdownNetwork(driver, network)) < 0)
        goto cleanup;

2825
    if (!network->persistent) {
2826 2827 2828 2829 2830
        if (networkRemoveInactive(driver, network) < 0) {
            network = NULL;
            ret = -1;
            goto cleanup;
        }
2831 2832
        network = NULL;
    }
2833

2834
cleanup:
2835 2836
    if (network)
        virNetworkObjUnlock(network);
2837
    networkDriverUnlock(driver);
2838 2839 2840
    return ret;
}

2841
static char *networkGetXMLDesc(virNetworkPtr net,
2842
                               unsigned int flags)
2843
{
2844
    virNetworkObjPtr network;
2845
    virNetworkDefPtr def;
2846
    char *ret = NULL;
2847

2848
    virCheckFlags(VIR_NETWORK_XML_INACTIVE, NULL);
2849

2850 2851
    if (!(network = networkObjFromNetwork(net)))
        return ret;
2852

2853 2854 2855
    if (virNetworkGetXMLDescEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2856 2857 2858 2859 2860 2861
    if ((flags & VIR_NETWORK_XML_INACTIVE) && network->newDef)
        def = network->newDef;
    else
        def = network->def;

    ret = virNetworkDefFormat(def, flags);
2862 2863

cleanup:
2864 2865
    if (network)
        virNetworkObjUnlock(network);
2866
    return ret;
2867 2868 2869
}

static char *networkGetBridgeName(virNetworkPtr net) {
2870 2871 2872
    virNetworkObjPtr network;
    char *bridge = NULL;

2873 2874
    if (!(network = networkObjFromNetwork(net)))
        return bridge;
2875

2876 2877 2878
    if (virNetworkGetBridgeNameEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2879
    if (!(network->def->bridge)) {
2880 2881 2882
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("network '%s' does not have a bridge name."),
                       network->def->name);
2883 2884 2885
        goto cleanup;
    }

2886
    ignore_value(VIR_STRDUP(bridge, network->def->bridge));
2887 2888

cleanup:
2889 2890
    if (network)
        virNetworkObjUnlock(network);
2891 2892 2893 2894 2895
    return bridge;
}

static int networkGetAutostart(virNetworkPtr net,
                             int *autostart) {
2896 2897
    virNetworkObjPtr network;
    int ret = -1;
2898

2899 2900
    if (!(network = networkObjFromNetwork(net)))
        return ret;
2901

2902 2903 2904
    if (virNetworkGetAutostartEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2905
    *autostart = network->autostart;
2906
    ret = 0;
2907

2908
cleanup:
2909 2910
    if (network)
        virNetworkObjUnlock(network);
2911
    return ret;
2912 2913 2914
}

static int networkSetAutostart(virNetworkPtr net,
2915
                               int autostart) {
2916
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
2917
    virNetworkObjPtr network;
2918
    char *configFile = NULL, *autostartLink = NULL;
2919
    int ret = -1;
2920

2921
    networkDriverLock(driver);
2922
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
2923

2924
    if (!network) {
2925 2926
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2927
        goto cleanup;
2928 2929
    }

2930 2931 2932
    if (virNetworkSetAutostartEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2933
    if (!network->persistent) {
2934 2935
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("cannot set autostart for transient network"));
2936 2937 2938
        goto cleanup;
    }

2939 2940
    autostart = (autostart != 0);

2941
    if (network->autostart != autostart) {
2942
        if ((configFile = virNetworkConfigFile(driver->networkConfigDir, network->def->name)) == NULL)
2943
            goto cleanup;
2944
        if ((autostartLink = virNetworkConfigFile(driver->networkAutostartDir, network->def->name)) == NULL)
2945 2946
            goto cleanup;

2947
        if (autostart) {
2948
            if (virFileMakePath(driver->networkAutostartDir) < 0) {
2949
                virReportSystemError(errno,
2950 2951
                                     _("cannot create autostart directory '%s'"),
                                     driver->networkAutostartDir);
2952 2953
                goto cleanup;
            }
2954

2955
            if (symlink(configFile, autostartLink) < 0) {
2956
                virReportSystemError(errno,
2957
                                     _("Failed to create symlink '%s' to '%s'"),
2958
                                     autostartLink, configFile);
2959 2960 2961
                goto cleanup;
            }
        } else {
2962
            if (unlink(autostartLink) < 0 && errno != ENOENT && errno != ENOTDIR) {
2963
                virReportSystemError(errno,
2964
                                     _("Failed to delete symlink '%s'"),
2965
                                     autostartLink);
2966 2967
                goto cleanup;
            }
2968 2969
        }

2970
        network->autostart = autostart;
2971
    }
2972
    ret = 0;
2973

2974
cleanup:
2975 2976
    VIR_FREE(configFile);
    VIR_FREE(autostartLink);
2977 2978
    if (network)
        virNetworkObjUnlock(network);
2979
    networkDriverUnlock(driver);
2980
    return ret;
2981 2982 2983 2984 2985
}


static virNetworkDriver networkDriver = {
    "Network",
2986 2987 2988 2989 2990 2991 2992
    .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 */
2993 2994
    .networkLookupByUUID = networkLookupByUUID, /* 0.2.0 */
    .networkLookupByName = networkLookupByName, /* 0.2.0 */
2995 2996
    .networkCreateXML = networkCreateXML, /* 0.2.0 */
    .networkDefineXML = networkDefineXML, /* 0.2.0 */
2997
    .networkUndefine = networkUndefine, /* 0.2.0 */
2998
    .networkUpdate = networkUpdate, /* 0.10.2 */
2999
    .networkCreate = networkCreate, /* 0.2.0 */
3000 3001 3002 3003 3004 3005 3006
    .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 */
3007 3008 3009
};

static virStateDriver networkStateDriver = {
3010
    .name = "Network",
3011
    .stateInitialize  = networkStateInitialize,
3012
    .stateAutoStart  = networkStateAutoStart,
3013 3014
    .stateCleanup = networkStateCleanup,
    .stateReload = networkStateReload,
3015 3016 3017 3018 3019 3020 3021
};

int networkRegister(void) {
    virRegisterNetworkDriver(&networkDriver);
    virRegisterStateDriver(&networkStateDriver);
    return 0;
}
3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032

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

/* 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.
 */

3033 3034 3035 3036 3037 3038 3039 3040 3041
/* 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) {
    unsigned int num_virt_fns = 0;
    char **vfname = NULL;
3042
    virPCIDeviceAddressPtr *virt_fns;
3043 3044
    int ret = -1;
    size_t i;
3045

3046
    if ((virNetDevGetVirtualFunctions(netdef->forward.pfs->dev,
3047
                                      &vfname, &virt_fns, &num_virt_fns)) < 0) {
3048 3049
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Could not get Virtual functions on %s"),
3050
                       netdef->forward.pfs->dev);
3051 3052 3053 3054 3055 3056
        goto finish;
    }

    if (num_virt_fns == 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("No Vf's present on SRIOV PF %s"),
3057
                       netdef->forward.pfs->dev);
3058 3059 3060
       goto finish;
    }

3061
    if (VIR_ALLOC_N(netdef->forward.ifs, num_virt_fns) < 0)
3062 3063
        goto finish;

3064
    netdef->forward.nifs = num_virt_fns;
3065

3066
    for (i = 0; i < netdef->forward.nifs; i++) {
3067 3068 3069 3070
        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)) {
3071 3072 3073
            netdef->forward.ifs[i].type = VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV;
            if (vfname[i]) {
                if (VIR_STRDUP(netdef->forward.ifs[i].device.dev, vfname[i]) < 0)
3074
                    goto finish;
3075
            } else {
3076
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
3077
                               _("Direct mode types require interface names"));
3078 3079
                goto finish;
            }
3080
        }
3081
        else if (netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
3082
            /* VF's are always PCI devices */
3083 3084 3085 3086 3087
            netdef->forward.ifs[i].type = VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI;
            netdef->forward.ifs[i].device.pci.domain = virt_fns[i]->domain;
            netdef->forward.ifs[i].device.pci.bus = virt_fns[i]->bus;
            netdef->forward.ifs[i].device.pci.slot = virt_fns[i]->slot;
            netdef->forward.ifs[i].device.pci.function = virt_fns[i]->function;
3088
        }
3089 3090 3091 3092
    }

    ret = 0;
finish:
3093 3094 3095
    for (i = 0; i < num_virt_fns; i++) {
        VIR_FREE(vfname[i]);
        VIR_FREE(virt_fns[i]);
3096
    }
3097
    VIR_FREE(vfname);
3098
    VIR_FREE(virt_fns);
3099 3100 3101
    return ret;
}

3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115
/* networkAllocateActualDevice:
 * @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
networkAllocateActualDevice(virDomainNetDefPtr iface)
{
3116
    virNetworkDriverStatePtr driver = driverState;
3117 3118 3119
    enum virDomainNetType actualType = iface->type;
    virNetworkObjPtr network = NULL;
    virNetworkDefPtr netdef = NULL;
3120
    virNetDevBandwidthPtr bandwidth = NULL;
3121 3122 3123
    virPortGroupDefPtr portgroup = NULL;
    virNetDevVPortProfilePtr virtport = iface->virtPortProfile;
    virNetDevVlanPtr vlan = NULL;
3124
    virNetworkForwardIfDefPtr dev = NULL;
3125
    size_t i;
3126 3127 3128
    int ret = -1;

    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
3129
        goto validate;
3130 3131 3132 3133 3134 3135 3136 3137

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

    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
    networkDriverUnlock(driver);
    if (!network) {
3138 3139 3140
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
3141
        goto error;
3142 3143
    }
    netdef = network->def;
3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154

    /* 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.
    */
    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.
     */
3155 3156 3157 3158 3159 3160 3161

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

    if (bandwidth) {
3162 3163
        if (!iface->data.network.actual &&
            VIR_ALLOC(iface->data.network.actual) < 0)
3164
            goto error;
3165
        if (virNetDevBandwidthCopy(&iface->data.network.actual->bandwidth,
3166
                                   bandwidth) < 0)
3167
            goto error;
3168 3169
    }

3170 3171 3172 3173 3174 3175 3176 3177 3178
    /* 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;

    if (vlan) {
3179 3180
        if (!iface->data.network.actual &&
            VIR_ALLOC(iface->data.network.actual) < 0)
3181 3182 3183 3184 3185
            goto error;
        if (virNetDevVlanCopy(&iface->data.network.actual->vlan, vlan) < 0)
           goto error;
    }

3186 3187 3188
    if ((netdef->forward.type == VIR_NETWORK_FORWARD_NONE) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_NAT) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE)) {
3189 3190 3191 3192 3193 3194
        /* for these forward types, the actual net type really *is*
         *NETWORK; we just keep the info from the portgroup in
         * iface->data.network.actual
        */
        if (iface->data.network.actual)
            iface->data.network.actual->type = VIR_DOMAIN_NET_TYPE_NETWORK;
3195 3196 3197 3198

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

3199
    } else if ((netdef->forward.type == VIR_NETWORK_FORWARD_BRIDGE) &&
3200
               netdef->bridge) {
3201 3202 3203 3204 3205

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

3206 3207
        if (!iface->data.network.actual &&
            VIR_ALLOC(iface->data.network.actual) < 0)
3208
            goto error;
3209

3210
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_BRIDGE;
3211 3212
        if (VIR_STRDUP(iface->data.network.actual->data.bridge.brname,
                       netdef->bridge) < 0)
3213
            goto error;
3214

3215 3216 3217 3218 3219 3220 3221 3222
        /* 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) {
3223
            goto error;
3224 3225 3226 3227 3228 3229 3230 3231 3232 3233
        }
        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);
3234
                goto error;
3235 3236 3237
            }
        }

3238
    } else if (netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
3239

3240 3241
        virDomainHostdevSubsysPciBackendType backend;

3242 3243
        if (!iface->data.network.actual &&
            VIR_ALLOC(iface->data.network.actual) < 0)
3244 3245 3246
            goto error;

        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_HOSTDEV;
3247
        if (netdef->forward.npfs > 0 && netdef->forward.nifs <= 0 &&
3248 3249 3250 3251 3252
            networkCreateInterfacePool(netdef) < 0) {
            goto error;
        }

        /* pick first dev with 0 connections */
3253 3254 3255
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].connections == 0) {
                dev = &netdef->forward.ifs[i];
3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269
                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;
3270
        iface->data.network.actual->data.hostdev.def.managed = netdef->forward.managed ? 1 : 0;
3271
        iface->data.network.actual->data.hostdev.def.source.subsys.type = dev->type;
3272
        iface->data.network.actual->data.hostdev.def.source.subsys.u.pci.addr = dev->device.pci;
3273

3274 3275 3276
        switch (netdef->forward.driverName)
        {
        case VIR_NETWORK_FORWARD_DRIVER_NAME_DEFAULT:
3277
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_DEFAULT;
3278 3279
            break;
        case VIR_NETWORK_FORWARD_DRIVER_NAME_KVM:
3280
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_KVM;
3281 3282
            break;
        case VIR_NETWORK_FORWARD_DRIVER_NAME_VFIO:
3283
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_VFIO;
3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294
            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;

3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319
        /* 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;
            }
        }

3320 3321 3322 3323
    } 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)) {
3324 3325 3326 3327 3328

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

3329 3330
        if (!iface->data.network.actual &&
            VIR_ALLOC(iface->data.network.actual) < 0)
3331
            goto error;
3332 3333

        /* Set type=direct and appropriate <source mode='xxx'/> */
3334
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_DIRECT;
3335
        switch (netdef->forward.type) {
3336
        case VIR_NETWORK_FORWARD_BRIDGE:
3337
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_BRIDGE;
3338 3339
            break;
        case VIR_NETWORK_FORWARD_PRIVATE:
3340
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PRIVATE;
3341 3342
            break;
        case VIR_NETWORK_FORWARD_VEPA:
3343
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_VEPA;
3344 3345
            break;
        case VIR_NETWORK_FORWARD_PASSTHROUGH:
3346
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PASSTHRU;
3347 3348 3349
            break;
        }

3350 3351 3352 3353 3354 3355 3356 3357
        /* 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) {
3358
            goto error;
3359
        }
3360
        virtport = iface->data.network.actual->virtPortProfile;
3361
        if (virtport) {
3362 3363 3364 3365 3366 3367 3368 3369
            /* 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);
3370
                goto error;
3371 3372
            }
        }
3373

3374 3375 3376
        /* 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).
         */
3377
        if ((netdef->forward.nifs <= 0) && (netdef->forward.npfs <= 0)) {
3378 3379 3380 3381
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' uses a direct mode, but "
                             "has no forward dev and no interface pool"),
                           netdef->name);
3382
            goto error;
3383 3384 3385
        } else {
            /* pick an interface from the pool */

3386
            if (netdef->forward.npfs > 0 && netdef->forward.nifs == 0 &&
3387 3388 3389 3390
                networkCreateInterfacePool(netdef) < 0) {
                goto error;
            }

3391 3392 3393 3394 3395
            /* 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.
3396
             */
3397 3398
            if ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
                ((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
3399 3400 3401
                 iface->data.network.actual->virtPortProfile &&
                 (iface->data.network.actual->virtPortProfile->virtPortType
                  == VIR_NETDEV_VPORT_PROFILE_8021QBH))) {
3402

3403
                /* pick first dev with 0 connections */
3404 3405 3406
                for (i = 0; i < netdef->forward.nifs; i++) {
                    if (netdef->forward.ifs[i].connections == 0) {
                        dev = &netdef->forward.ifs[i];
3407 3408 3409 3410 3411
                        break;
                    }
                }
            } else {
                /* pick least used dev */
3412
                dev = &netdef->forward.ifs[0];
3413 3414 3415
                for (i = 1; i < netdef->forward.nifs; i++) {
                    if (netdef->forward.ifs[i].connections < dev->connections)
                        dev = &netdef->forward.ifs[i];
3416 3417 3418 3419
                }
            }
            /* dev points at the physical device we want to use */
            if (!dev) {
3420 3421 3422 3423
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("network '%s' requires exclusive access "
                                 "to interfaces, but none are available"),
                               netdef->name);
3424
                goto error;
3425
            }
3426 3427
            if (VIR_STRDUP(iface->data.network.actual->data.direct.linkdev,
                           dev->device.dev) < 0)
3428
                goto error;
3429 3430 3431
        }
    }

3432
    if (virNetDevVPortProfileCheckComplete(virtport, true) < 0)
3433
        goto error;
3434

3435 3436 3437 3438 3439 3440
validate:
    /* 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.
     */

3441
    if (virDomainNetGetActualVlan(iface)) {
3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466
        /* 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;
        }
    }

3467 3468
    if (dev) {
        /* we are now assured of success, so mark the allocation */
3469
        dev->connections++;
3470 3471 3472 3473 3474 3475 3476 3477 3478
        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);
        }
3479
    }
3480

3481 3482 3483 3484 3485
    if (netdef) {
        netdef->connections++;
        VIR_DEBUG("Using network %s, %d connections",
                  netdef->name, netdef->connections);
    }
3486
    ret = 0;
3487

3488 3489 3490
cleanup:
    if (network)
        virNetworkObjUnlock(network);
3491 3492 3493 3494
    return ret;

error:
    if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
3495 3496 3497
        virDomainActualNetDefFree(iface->data.network.actual);
        iface->data.network.actual = NULL;
    }
3498
    goto cleanup;
3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513
}

/* networkNotifyActualDevice:
 * @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
networkNotifyActualDevice(virDomainNetDefPtr iface)
{
3514
    virNetworkDriverStatePtr driver = driverState;
3515
    enum virDomainNetType actualType = virDomainNetGetActualType(iface);
3516 3517
    virNetworkObjPtr network;
    virNetworkDefPtr netdef;
3518
    virNetworkForwardIfDefPtr dev = NULL;
3519 3520
    size_t i;
    int ret = -1;
3521 3522 3523 3524 3525 3526 3527 3528

    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
       return 0;

    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
    networkDriverUnlock(driver);
    if (!network) {
3529 3530 3531
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
3532 3533 3534 3535 3536
        goto error;
    }
    netdef = network->def;

    if (!iface->data.network.actual ||
3537 3538
        (actualType != VIR_DOMAIN_NET_TYPE_DIRECT &&
         actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV)) {
3539 3540
        VIR_DEBUG("Nothing to claim from network %s", iface->data.network.name);
        goto success;
3541 3542
    }

3543
    if (netdef->forward.npfs > 0 && netdef->forward.nifs == 0 &&
3544
        networkCreateInterfacePool(netdef) < 0) {
3545
        goto error;
3546
    }
3547
    if (netdef->forward.nifs == 0) {
3548
        virReportError(VIR_ERR_INTERNAL_ERROR,
3549 3550
                       _("network '%s' uses a direct or hostdev mode, "
                         "but has no forward dev and no interface pool"),
3551
                       netdef->name);
3552
        goto error;
3553
    }
3554

3555 3556
    if (actualType == VIR_DOMAIN_NET_TYPE_DIRECT) {
        const char *actualDev;
3557

3558 3559 3560 3561 3562 3563 3564 3565 3566
        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 */
3567 3568
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
3569
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
3570 3571
                STREQ(actualDev, netdef->forward.ifs[i].device.dev)) {
                dev = &netdef->forward.ifs[i];
3572 3573 3574 3575 3576
                break;
            }
        }
        /* dev points at the physical device we want to use */
        if (!dev) {
3577
            virReportError(VIR_ERR_INTERNAL_ERROR,
3578 3579
                           _("network '%s' doesn't have dev='%s' "
                             "in use by domain"),
3580
                           netdef->name, actualDev);
3581
            goto error;
3582 3583
        }

3584
        /* PASSTHROUGH mode and PRIVATE Mode + 802.1Qbh both require
3585 3586
         * exclusive access to a device, so current connections count
         * must be 0 in those cases.
3587
         */
3588
        if ((dev->connections > 0) &&
3589 3590
            ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
             ((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
3591 3592
              iface->data.network.actual->virtPortProfile &&
              (iface->data.network.actual->virtPortProfile->virtPortType
3593
               == VIR_NETDEV_VPORT_PROFILE_8021QBH)))) {
3594
            virReportError(VIR_ERR_INTERNAL_ERROR,
3595 3596
                           _("network '%s' claims dev='%s' is already in "
                             "use by a different domain"),
3597
                           netdef->name, actualDev);
3598
            goto error;
3599
        }
3600

3601
        /* we are now assured of success, so mark the allocation */
3602
        dev->connections++;
3603
        VIR_DEBUG("Using physical device %s, connections %d",
3604
                  dev->device.dev, dev->connections);
3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617

    }  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 */
3618 3619
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
3620
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
3621
                virDevicePCIAddressEqual(&hostdev->source.subsys.u.pci.addr,
3622 3623
                                         &netdef->forward.ifs[i].device.pci)) {
                dev = &netdef->forward.ifs[i];
3624 3625 3626 3627 3628 3629 3630 3631 3632
                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,
3633 3634 3635 3636
                           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);
3637 3638 3639 3640 3641 3642 3643 3644
                goto error;
        }

        /* 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) &&
3645
            netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661
            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);
3662 3663
    }

3664
success:
3665 3666 3667
    netdef->connections++;
    VIR_DEBUG("Using network %s, %d connections",
              netdef->name, netdef->connections);
3668 3669 3670 3671 3672
    ret = 0;
cleanup:
    if (network)
        virNetworkObjUnlock(network);
    return ret;
3673 3674 3675

error:
    goto cleanup;
3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691
}


/* networkReleaseActualDevice:
 * @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
networkReleaseActualDevice(virDomainNetDefPtr iface)
{
3692
    virNetworkDriverStatePtr driver = driverState;
3693
    enum virDomainNetType actualType = virDomainNetGetActualType(iface);
3694
    virNetworkObjPtr network;
3695
    virNetworkDefPtr netdef;
3696
    virNetworkForwardIfDefPtr dev = NULL;
3697 3698
    size_t i;
    int ret = -1;
3699 3700 3701 3702 3703 3704 3705 3706

    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
       return 0;

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

3714 3715 3716 3717 3718 3719
    if ((netdef->forward.type == VIR_NETWORK_FORWARD_NONE ||
         netdef->forward.type == VIR_NETWORK_FORWARD_NAT ||
         netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE) &&
        networkUnplugBandwidth(network, iface) < 0)
        goto error;

3720 3721 3722
    if ((!iface->data.network.actual) ||
        ((actualType != VIR_DOMAIN_NET_TYPE_DIRECT) &&
         (actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV))) {
3723 3724
        VIR_DEBUG("Nothing to release to network %s", iface->data.network.name);
        goto success;
3725 3726
    }

3727
    if (netdef->forward.nifs == 0) {
3728
        virReportError(VIR_ERR_INTERNAL_ERROR,
3729
                       _("network '%s' uses a direct/hostdev mode, but "
3730 3731
                         "has no forward dev and no interface pool"),
                       netdef->name);
3732
        goto error;
3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744
    }

    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;
        }
3745

3746 3747
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
3748
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
3749 3750
                STREQ(actualDev, netdef->forward.ifs[i].device.dev)) {
                dev = &netdef->forward.ifs[i];
3751 3752 3753
                break;
            }
        }
3754

3755
        if (!dev) {
3756
            virReportError(VIR_ERR_INTERNAL_ERROR,
3757 3758
                           _("network '%s' doesn't have dev='%s' "
                             "in use by domain"),
3759
                           netdef->name, actualDev);
3760
            goto error;
3761 3762
        }

3763
        dev->connections--;
3764
        VIR_DEBUG("Releasing physical device %s, connections %d",
3765
                  dev->device.dev, dev->connections);
3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776

    } 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;
        }

3777 3778
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
3779
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
3780
                virDevicePCIAddressEqual(&hostdev->source.subsys.u.pci.addr,
3781 3782
                                         &netdef->forward.ifs[i].device.pci)) {
                dev = &netdef->forward.ifs[i];
3783 3784 3785 3786 3787 3788 3789 3790 3791
                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,
3792 3793 3794 3795
                           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);
3796 3797 3798 3799 3800 3801 3802 3803 3804
                goto error;
        }

        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);
   }
3805

3806
success:
3807 3808 3809
    netdef->connections--;
    VIR_DEBUG("Releasing network %s, %d connections",
              netdef->name, netdef->connections);
3810 3811 3812 3813
    ret = 0;
cleanup:
    if (network)
        virNetworkObjUnlock(network);
3814 3815 3816 3817
    if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
        virDomainActualNetDefFree(iface->data.network.actual);
        iface->data.network.actual = NULL;
    }
3818
    return ret;
3819 3820 3821

error:
    goto cleanup;
3822
}
3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846

/*
 * 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;
3847
    virNetworkDriverStatePtr driver = driverState;
3848
    virNetworkObjPtr network;
3849 3850 3851 3852
    virNetworkDefPtr netdef;
    virNetworkIpDefPtr ipdef;
    virSocketAddr addr;
    virSocketAddrPtr addrptr = NULL;
3853
    char *dev_name = NULL;
3854 3855 3856 3857 3858 3859

    *netaddr = NULL;
    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, netname);
    networkDriverUnlock(driver);
    if (!network) {
3860 3861 3862
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       netname);
3863
        goto error;
3864 3865 3866
    }
    netdef = network->def;

3867
    switch (netdef->forward.type) {
3868 3869 3870 3871 3872 3873
    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) {
3874 3875 3876
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' doesn't have an IPv4 address"),
                           netdef->name);
3877 3878 3879 3880 3881 3882
            break;
        }
        addrptr = &ipdef->address;
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
3883
        if ((dev_name = netdef->bridge))
3884 3885 3886 3887 3888 3889 3890 3891
            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:
3892 3893
        if ((netdef->forward.nifs > 0) && netdef->forward.ifs)
            dev_name = netdef->forward.ifs[0].device.dev;
3894

3895
        if (!dev_name) {
3896 3897 3898
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' has no associated interface or bridge"),
                           netdef->name);
3899 3900 3901 3902
        }
        break;
    }

3903
    if (dev_name) {
3904
        if (virNetDevGetIPv4Address(dev_name, &addr) < 0)
3905
            goto error;
3906
        addrptr = &addr;
3907 3908
    }

3909 3910 3911
    if (!(addrptr &&
          (*netaddr = virSocketAddrFormat(addrptr)))) {
        goto error;
3912 3913
    }

3914
    ret = 0;
3915 3916 3917 3918
cleanup:
    if (network)
        virNetworkObjUnlock(network);
    return ret;
3919 3920 3921

error:
    goto cleanup;
3922
}
3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940

/**
 * 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;
3941
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
3942 3943 3944 3945
    unsigned long long tmp_floor_sum = net->floor_sum;
    unsigned long long tmp_new_rate = 0;
    char ifmac[VIR_MAC_STRING_BUFLEN];

3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956
    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;
    }

3957
    if (!ifaceBand || !ifaceBand->in || !ifaceBand->in->floor ||
3958 3959
        !netBand || !netBand->in) {
        /* no QoS required, claim success */
3960
        return 1;
3961
    }
3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030

    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;

cleanup:
    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];
4031
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059

    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;
    }

4060 4061
    plug_ret = virNetDevBandwidthPlug(net->def->bridge, net->def->bandwidth,
                                      &iface->mac, ifaceBand, class_id);
4062 4063 4064 4065 4066 4067 4068 4069
    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 */
4070
    net->floor_sum += ifaceBand->in->floor;
4071
    /* update status file */
4072
    if (virNetworkSaveStatus(driverState->stateDir, net) < 0) {
4073
        ignore_value(virBitmapClearBit(net->class_id, class_id));
4074
        net->floor_sum -= ifaceBand->in->floor;
4075 4076 4077 4078
        iface->data.network.actual->class_id = 0;
        ignore_value(virNetDevBandwidthUnplug(net->def->bridge, class_id));
        goto cleanup;
    }
4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097
    /* 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;

cleanup:
    return ret;
}

static int
networkUnplugBandwidth(virNetworkObjPtr net,
                       virDomainNetDefPtr iface)
{
    int ret = 0;
    unsigned long long new_rate;
4098
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
4099 4100 4101

    if (iface->data.network.actual &&
        iface->data.network.actual->class_id) {
4102 4103 4104 4105 4106
        if (!net->def->bandwidth || !net->def->bandwidth->in) {
            VIR_WARN("Network %s has no bandwidth but unplug requested",
                     net->def->name);
            goto cleanup;
        }
4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117
        /* 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 */
4118
        net->floor_sum -= ifaceBand->in->floor;
4119 4120 4121 4122
        /* return class ID */
        ignore_value(virBitmapClearBit(net->class_id,
                                       iface->data.network.actual->class_id));
        /* update status file */
4123
        if (virNetworkSaveStatus(driverState->stateDir, net) < 0) {
4124
            net->floor_sum += ifaceBand->in->floor;
4125 4126 4127 4128
            ignore_value(virBitmapSetBit(net->class_id,
                                         iface->data.network.actual->class_id));
            goto cleanup;
        }
4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141
        /* 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;
    }

cleanup:
    return ret;
}