bridge_driver.c 141.6 KB
Newer Older
1
/*
2
 * bridge_driver.c: core driver methods for managing network
3
 *
4
 * Copyright (C) 2006-2014 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16 17
 * Copyright (C) 2006 Daniel P. Berrange
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with this library.  If not, see
O
Osier Yang 已提交
19
 * <http://www.gnu.org/licenses/>.
20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42
 *
 * Author: Daniel P. Berrange <berrange@redhat.com>
 */

#include <config.h>

#include <sys/types.h>
#include <sys/poll.h>
#include <limits.h>
#include <string.h>
#include <stdio.h>
#include <stdarg.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
#include <sys/utsname.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <signal.h>
#include <paths.h>
#include <pwd.h>
#include <sys/wait.h>
#include <sys/ioctl.h>
43
#include <net/if.h>
44 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
#include "network_event.h"
74

75 76
#define VIR_FROM_THIS VIR_FROM_NETWORK

77
static void networkDriverLock(virNetworkDriverStatePtr driver)
78
{
79
    virMutexLock(&driver->lock);
80
}
81
static void networkDriverUnlock(virNetworkDriverStatePtr driver)
82
{
83
    virMutexUnlock(&driver->lock);
84 85
}

86
static int networkStateCleanup(void);
87

88
static int networkStartNetwork(virNetworkDriverStatePtr driver,
89 90
                               virNetworkObjPtr network);

91
static int networkShutdownNetwork(virNetworkDriverStatePtr driver,
92 93
                                  virNetworkObjPtr network);

94
static int networkStartNetworkVirtual(virNetworkDriverStatePtr driver,
95
                                     virNetworkObjPtr network);
96

97
static int networkShutdownNetworkVirtual(virNetworkDriverStatePtr driver,
R
Roman Bogorodskiy 已提交
98
                                         virNetworkObjPtr network);
99

100
static int networkStartNetworkExternal(virNetworkDriverStatePtr driver,
101 102
                                     virNetworkObjPtr network);

103
static int networkShutdownNetworkExternal(virNetworkDriverStatePtr driver,
104
                                        virNetworkObjPtr network);
105

106
static void networkReloadFirewallRules(virNetworkDriverStatePtr driver);
107
static void networkRefreshDaemons(virNetworkDriverStatePtr driver);
108

109 110 111 112 113
static int networkPlugBandwidth(virNetworkObjPtr net,
                                virDomainNetDefPtr iface);
static int networkUnplugBandwidth(virNetworkObjPtr net,
                                  virDomainNetDefPtr iface);

114
static virNetworkDriverStatePtr driverState = NULL;
115

116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136
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;
}

137
static char *
138
networkDnsmasqLeaseFileNameDefault(const char *netname)
139 140 141
{
    char *leasefile;

142 143
    ignore_value(virAsprintf(&leasefile, "%s/%s.leases",
                             driverState->dnsmasqStateDir, netname));
144 145 146
    return leasefile;
}

147 148 149
networkDnsmasqLeaseFileNameFunc networkDnsmasqLeaseFileName =
    networkDnsmasqLeaseFileNameDefault;

150 151 152 153 154
static char *
networkDnsmasqConfigFileName(const char *netname)
{
    char *conffile;

155 156
    ignore_value(virAsprintf(&conffile, "%s/%s.conf",
                             driverState->dnsmasqStateDir, netname));
157 158 159
    return conffile;
}

160 161 162 163 164 165
static char *
networkRadvdPidfileBasename(const char *netname)
{
    /* this is simple but we want to be sure it's consistently done */
    char *pidfilebase;

166
    ignore_value(virAsprintf(&pidfilebase, "%s-radvd", netname));
167 168 169 170 171 172 173 174
    return pidfilebase;
}

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

175 176
    ignore_value(virAsprintf(&configfile, "%s/%s-radvd.conf",
                             driverState->radvdStateDir, netname));
177 178
    return configfile;
}
179

180 181
/* do needed cleanup steps and remove the network from the list */
static int
182
networkRemoveInactive(virNetworkDriverStatePtr driver,
183 184 185 186
                      virNetworkObjPtr net)
{
    char *leasefile = NULL;
    char *radvdconfigfile = NULL;
187
    char *configfile = NULL;
188
    char *radvdpidbase = NULL;
189
    char *statusfile = NULL;
190 191 192 193 194 195
    dnsmasqContext *dctx = NULL;
    virNetworkDefPtr def = virNetworkObjGetPersistentDef(net);

    int ret = -1;

    /* remove the (possibly) existing dnsmasq and radvd files */
196 197
    if (!(dctx = dnsmasqContextNew(def->name,
                                   driverState->dnsmasqStateDir))) {
198
        goto cleanup;
199
    }
200 201 202 203 204

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

    if (!(radvdconfigfile = networkRadvdConfigFileName(def->name)))
205
        goto cleanup;
206 207

    if (!(radvdpidbase = networkRadvdPidfileBasename(def->name)))
208
        goto cleanup;
209

210
    if (!(configfile = networkDnsmasqConfigFileName(def->name)))
211
        goto cleanup;
212

213 214
    if (!(statusfile
          = virNetworkConfigFile(driverState->stateDir, def->name)))
215
        goto cleanup;
216

217 218 219
    /* dnsmasq */
    dnsmasqDelete(dctx);
    unlink(leasefile);
220
    unlink(configfile);
221 222 223

    /* radvd */
    unlink(radvdconfigfile);
224
    virPidFileDelete(driverState->pidDir, radvdpidbase);
225

226 227 228
    /* remove status file */
    unlink(statusfile);

229 230 231 232 233 234 235
    /* remove the network definition */
    virNetworkRemoveInactive(&driver->networks, net);

    ret = 0;

cleanup:
    VIR_FREE(leasefile);
236
    VIR_FREE(configfile);
237 238
    VIR_FREE(radvdconfigfile);
    VIR_FREE(radvdpidbase);
239
    VIR_FREE(statusfile);
240 241 242 243
    dnsmasqContextFree(dctx);
    return ret;
}

244 245 246
static char *
networkBridgeDummyNicName(const char *brname)
{
247
    static const char dummyNicSuffix[] = "-nic";
248 249
    char *nicname;

250 251 252 253 254 255 256
    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.
         */
257 258 259 260 261
        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));
262
    } else {
263
        ignore_value(virAsprintf(&nicname, "%s%s", brname, dummyNicSuffix));
264
    }
265 266 267
    return nicname;
}

268
static void
269
networkFindActiveConfigs(virNetworkDriverStatePtr driver)
270
{
271
    size_t i;
272

273
    for (i = 0; i < driver->networks.count; i++) {
274 275 276 277 278 279
        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
280
            virNetDevExists(obj->def->bridge) == 1) {
281 282
            obj->active = 1;

283 284
            /* Try and read dnsmasq/radvd pids if any */
            if (obj->def->ips && (obj->def->nips > 0)) {
285 286
                char *radvdpidbase;

287
                ignore_value(virPidFileReadIfAlive(driverState->pidDir, obj->def->name,
288 289
                                                   &obj->dnsmasqPid,
                                                   dnsmasqCapsGetBinaryPath(driver->dnsmasqCaps)));
290

291
                if (!(radvdpidbase = networkRadvdPidfileBasename(obj->def->name)))
292
                    goto cleanup;
293
                ignore_value(virPidFileReadIfAlive(driverState->pidDir, radvdpidbase,
294
                                                   &obj->radvdPid, RADVD));
295
                VIR_FREE(radvdpidbase);
296 297 298
            }
        }

299
    cleanup:
300 301
        virNetworkObjUnlock(obj);
    }
302 303 304 305 306 307 308 309 310 311 312 313 314 315 316

    /* 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++;
    }
317 318 319
}


320
static void
321
networkAutostartConfigs(virNetworkDriverStatePtr driver) {
322
    size_t i;
323

324
    for (i = 0; i < driver->networks.count; i++) {
325
        virNetworkObjLock(driver->networks.objs[i]);
326
        if (driver->networks.objs[i]->autostart &&
327 328
            !virNetworkObjIsActive(driver->networks.objs[i])) {
            if (networkStartNetwork(driver, driver->networks.objs[i]) < 0) {
329
            /* failed to start but already logged */
330
            }
331
        }
332
        virNetworkObjUnlock(driver->networks.objs[i]);
333 334 335
    }
}

336 337 338 339
#if HAVE_FIREWALLD
static DBusHandlerResult
firewalld_dbus_filter_bridge(DBusConnection *connection ATTRIBUTE_UNUSED,
                             DBusMessage *message, void *user_data) {
340
    virNetworkDriverStatePtr _driverState = user_data;
341 342 343 344 345 346 347

    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.");
348
        networkReloadFirewallRules(_driverState);
349 350 351 352 353 354
    }

    return DBUS_HANDLER_RESULT_NOT_YET_HANDLED;
}
#endif

355
/**
356
 * networkStateInitialize:
357 358 359 360
 *
 * Initialization function for the QEmu daemon
 */
static int
361 362 363
networkStateInitialize(bool privileged,
                       virStateInhibitCallback callback ATTRIBUTE_UNUSED,
                       void *opaque ATTRIBUTE_UNUSED)
364
{
365 366 367
    int ret = -1;
    char *configdir = NULL;
    char *rundir = NULL;
368 369 370
#ifdef HAVE_FIREWALLD
    DBusConnection *sysbus = NULL;
#endif
371 372

    if (VIR_ALLOC(driverState) < 0)
373
        goto error;
374

375 376 377 378
    if (virMutexInit(&driverState->lock) < 0) {
        VIR_FREE(driverState);
        goto error;
    }
379 380
    networkDriverLock(driverState);

381 382 383 384 385 386 387 388 389
    /* 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.
     */
390
    if (privileged) {
391 392 393 394 395 396 397 398 399 400 401 402 403
        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;
404
    } else {
405 406 407
        configdir = virGetUserConfigDirectory();
        rundir = virGetUserRuntimeDirectory();
        if (!(configdir && rundir))
408
            goto error;
409

410 411 412 413 414 415 416 417 418 419 420 421
        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)) {
422
            goto error;
423
        }
424 425
    }

426 427
    /* if this fails now, it will be retried later with dnsmasqCapsRefresh() */
    driverState->dnsmasqCaps = dnsmasqCapsNewFromBinary(DNSMASQ);
428

429
    if (virNetworkLoadAllState(&driverState->networks,
430
                               driverState->stateDir) < 0)
431 432
        goto error;

433
    if (virNetworkLoadAllConfigs(&driverState->networks,
434
                                 driverState->networkConfigDir,
435 436 437
                                 driverState->networkAutostartDir) < 0)
        goto error;

438
    networkFindActiveConfigs(driverState);
439
    networkReloadFirewallRules(driverState);
440
    networkRefreshDaemons(driverState);
441

442 443
    driverState->networkEventState = virObjectEventStateNew();

444 445
    networkDriverUnlock(driverState);

446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471
#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

472 473 474 475 476
    ret = 0;
cleanup:
    VIR_FREE(configdir);
    VIR_FREE(rundir);
    return ret;
477

478
error:
479 480
    if (driverState)
        networkDriverUnlock(driverState);
481
    networkStateCleanup();
482
    goto cleanup;
483 484
}

485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
/**
 * networkStateAutoStart:
 *
 * Function to AutoStart the bridge configs
 */
static void
networkStateAutoStart(void)
{
    if (!driverState)
        return;

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

501
/**
502
 * networkStateReload:
503 504 505 506 507
 *
 * Function to restart the QEmu daemon, it will recheck the configuration
 * files and update its state and the networking
 */
static int
508
networkStateReload(void) {
509 510 511
    if (!driverState)
        return 0;

512
    networkDriverLock(driverState);
513
    virNetworkLoadAllState(&driverState->networks,
514
                           driverState->stateDir);
515
    virNetworkLoadAllConfigs(&driverState->networks,
516 517
                             driverState->networkConfigDir,
                             driverState->networkAutostartDir);
518
    networkReloadFirewallRules(driverState);
519
    networkRefreshDaemons(driverState);
520
    networkAutostartConfigs(driverState);
521
    networkDriverUnlock(driverState);
522 523 524 525 526
    return 0;
}


/**
527
 * networkStateCleanup:
528 529 530 531
 *
 * Shutdown the QEmu daemon, it will stop all active domains and networks
 */
static int
532
networkStateCleanup(void) {
533 534 535
    if (!driverState)
        return -1;

536 537
    networkDriverLock(driverState);

538 539
    virObjectEventStateFree(driverState->networkEventState);

540
    /* free inactive networks */
541
    virNetworkObjListFree(&driverState->networks);
542 543 544

    VIR_FREE(driverState->networkConfigDir);
    VIR_FREE(driverState->networkAutostartDir);
545 546 547 548
    VIR_FREE(driverState->stateDir);
    VIR_FREE(driverState->pidDir);
    VIR_FREE(driverState->dnsmasqStateDir);
    VIR_FREE(driverState->radvdStateDir);
549

550 551
    virObjectUnref(driverState->dnsmasqCaps);

552
    networkDriverUnlock(driverState);
553
    virMutexDestroy(&driverState->lock);
554

555 556 557 558 559 560
    VIR_FREE(driverState);

    return 0;
}


561 562 563 564 565 566 567
/* 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)
{
568 569
    size_t i;
    int ret = -1;
570 571 572 573 574 575 576
    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.
     */
577
    for (i = 0; i < 25; i++) {
578
        int signum = 0;
579
        if (i == 0)
580
            signum = SIGTERM;
581
        else if (i == 15) {
582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
            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
621 622 623 624
    /* the following does not build a file, it builds a list
     * which is later saved into a file
     */

625
static int
G
Gene Czarcinski 已提交
626 627
networkBuildDnsmasqDhcpHostsList(dnsmasqContext *dctx,
                                 virNetworkIpDefPtr ipdef)
628
{
629
    size_t i;
G
Gene Czarcinski 已提交
630
    bool ipv6 = false;
631

G
Gene Czarcinski 已提交
632 633
    if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
        ipv6 = true;
634 635
    for (i = 0; i < ipdef->nhosts; i++) {
        virNetworkDHCPHostDefPtr host = &(ipdef->hosts[i]);
G
Gene Czarcinski 已提交
636
        if (VIR_SOCKET_ADDR_VALID(&host->ip))
637 638
            if (dnsmasqAddDhcpHost(dctx, host->mac, &host->ip,
                                   host->name, host->id, ipv6) < 0)
639
                return -1;
640
    }
641

G
Gene Czarcinski 已提交
642 643 644 645 646 647 648
    return 0;
}

static int
networkBuildDnsmasqHostsList(dnsmasqContext *dctx,
                             virNetworkDNSDefPtr dnsdef)
{
649
    size_t i, j;
G
Gene Czarcinski 已提交
650

651 652
    if (dnsdef) {
        for (i = 0; i < dnsdef->nhosts; i++) {
653
            virNetworkDNSHostDefPtr host = &(dnsdef->hosts[i]);
654
            if (VIR_SOCKET_ADDR_VALID(&host->ip)) {
655
                for (j = 0; j < host->nnames; j++)
656 657
                    if (dnsmasqAddHost(dctx, &host->ip, host->names[j]) < 0)
                        return -1;
658 659
            }
        }
660 661
    }

662
    return 0;
663 664 665
}


666 667
int
networkDnsmasqConfContents(virNetworkObjPtr network,
668 669 670 671
                           const char *pidfile,
                           char **configstr,
                           dnsmasqContext *dctx,
                           dnsmasqCapsPtr caps ATTRIBUTE_UNUSED)
672
{
673
    virBuffer configbuf = VIR_BUFFER_INITIALIZER;
674
    int r, ret = -1;
675
    int nbleases = 0;
676
    size_t i;
677 678 679 680
    char *record = NULL;
    char *recordPort = NULL;
    char *recordWeight = NULL;
    char *recordPriority = NULL;
681
    virNetworkDNSDefPtr dns = &network->def->dns;
G
Gene Czarcinski 已提交
682 683
    virNetworkIpDefPtr tmpipdef, ipdef, ipv4def, ipv6def;
    bool ipv6SLAAC;
684

685 686
    *configstr = NULL;

687
    /*
688 689 690
     * All dnsmasq parameters are put into a configuration file, except the
     * command line --conf-file=parameter which specifies the location of
     * configuration file.
691
     *
692 693
     * All dnsmasq conf-file parameters must be specified as "foo=bar"
     * as oppose to "--foo bar" which was acceptable on the command line.
694
     */
695 696 697 698 699 700

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

701 702
    /* create dnsmasq config file appropriate for this network */
    virBufferAsprintf(&configbuf,
703 704 705 706 707 708 709
                      "##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"
710
                      "strict-order\n",
711 712
                      network->def->name);

713 714 715
    if (!network->def->dns.forwardPlainNames)
        virBufferAddLit(&configbuf, "domain-needed\n");

716 717 718 719 720 721 722 723
    if (network->def->dns.forwarders) {
        virBufferAddLit(&configbuf, "no-resolv\n");
        for (i = 0; i < network->def->dns.nfwds; i++) {
            virBufferAsprintf(&configbuf, "server=%s\n",
                               network->def->dns.forwarders[i]);
        }
    }

724
    if (network->def->domain) {
725
        virBufferAsprintf(&configbuf,
726 727 728 729
                          "domain=%s\n"
                          "expand-hosts\n",
                          network->def->domain);
    }
730 731 732 733 734 735 736 737 738 739

    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 : "");
    }
740

741
    if (pidfile)
742
        virBufferAsprintf(&configbuf, "pid-file=%s\n", pidfile);
743

744 745 746
    /* dnsmasq will *always* listen on localhost unless told otherwise */
    virBufferAddLit(&configbuf, "except-interface=lo\n");

747 748 749 750 751 752 753 754
    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.
         */
755
        virBufferAsprintf(&configbuf,
756 757 758
                          "bind-dynamic\n"
                          "interface=%s\n",
                          network->def->bridge);
759
    } else {
760
        virBufferAddLit(&configbuf, "bind-interfaces\n");
761 762 763 764 765 766 767 768
        /*
         * --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
         */
769 770 771
        for (i = 0;
             (tmpipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, i));
             i++) {
772 773 774 775
            char *ipaddr = virSocketAddrFormat(&tmpipdef->address);

            if (!ipaddr)
                goto cleanup;
776

777
            /* also part of CVE 2012-3411 - if the host's version of
778
             * dnsmasq doesn't have bind-dynamic, only allow listening on
779 780
             * private/local IP addresses (see RFC1918/RFC3484/RFC4193)
             */
781 782
            if (!dnsmasqCapsGet(caps, DNSMASQ_CAPS_BINDTODEVICE) &&
                !virSocketAddrIsPrivate(&tmpipdef->address)) {
783 784 785 786
                unsigned long version = dnsmasqCapsGetVersion(caps);

                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("Publicly routable address %s is prohibited. "
787
                                 "The version of dnsmasq on this host (%d.%d) "
788 789 790 791
                                 "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 "
792 793 794 795 796 797
                                 "(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);
798 799
                goto cleanup;
            }
800
            virBufferAsprintf(&configbuf, "listen-address=%s\n", ipaddr);
801 802 803
            VIR_FREE(ipaddr);
        }
    }
804

805 806
    /* If this is an isolated network, set the default route option
     * (3) to be empty to avoid setting a default route that's
807
     * guaranteed to not work, and set no-resolv so that no dns
808 809 810
     * 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).
811
     */
812
    if (network->def->forward.type == VIR_NETWORK_FORWARD_NONE) {
813
        virBufferAddLit(&configbuf, "dhcp-option=3\n"
814
                        "no-resolv\n");
815
    }
816

817
    for (i = 0; i < dns->ntxts; i++) {
818
        virBufferAsprintf(&configbuf, "txt-record=%s,%s\n",
819 820
                          dns->txts[i].name,
                          dns->txts[i].value);
821
    }
822

823 824 825 826
    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)
827
                goto cleanup;
828 829
            if (dns->srvs[i].priority &&
                virAsprintf(&recordPriority, "%d", dns->srvs[i].priority) < 0)
830
                goto cleanup;
831 832
            if (dns->srvs[i].weight &&
                virAsprintf(&recordWeight, "%d", dns->srvs[i].weight) < 0)
833
                goto cleanup;
834

835
            if (virAsprintf(&record, "%s.%s.%s,%s,%s,%s,%s",
836 837 838 839
                            dns->srvs[i].service,
                            dns->srvs[i].protocol,
                            dns->srvs[i].domain ? dns->srvs[i].domain : "",
                            dns->srvs[i].target ? dns->srvs[i].target : "",
840 841
                            recordPort           ? recordPort           : "",
                            recordPriority       ? recordPriority       : "",
842
                            recordWeight         ? recordWeight         : "") < 0)
843 844
                goto cleanup;

845
            virBufferAsprintf(&configbuf, "srv-host=%s\n", record);
846 847 848 849
            VIR_FREE(record);
            VIR_FREE(recordPort);
            VIR_FREE(recordWeight);
            VIR_FREE(recordPriority);
850
        }
851 852
    }

G
Gene Czarcinski 已提交
853
    /* Find the first dhcp for both IPv4 and IPv6 */
854 855 856
    for (i = 0, ipv4def = NULL, ipv6def = NULL, ipv6SLAAC = false;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, i));
         i++) {
G
Gene Czarcinski 已提交
857 858 859 860
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
            if (ipdef->nranges || ipdef->nhosts) {
                if (ipv4def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
861 862
                                   _("For IPv4, multiple DHCP definitions "
                                     "cannot be specified."));
G
Gene Czarcinski 已提交
863 864 865 866 867 868 869 870 871 872 873
                    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,
874 875 876 877 878 879 880 881 882
                                   _("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
883 884 885 886
                    goto cleanup;
                }
                if (ipv6def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
887 888
                                   _("For IPv6, multiple DHCP definitions "
                                     "cannot be specified."));
G
Gene Czarcinski 已提交
889 890 891 892 893 894 895 896 897 898 899 900 901
                    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 "
902 903 904 905
                 "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
906 907 908 909 910
    }

    ipdef = ipv4def ? ipv4def : ipv6def;

    while (ipdef) {
911
        for (r = 0; r < ipdef->nranges; r++) {
912
            char *saddr = virSocketAddrFormat(&ipdef->ranges[r].start);
913 914
            if (!saddr)
                goto cleanup;
915
            char *eaddr = virSocketAddrFormat(&ipdef->ranges[r].end);
916 917 918 919
            if (!eaddr) {
                VIR_FREE(saddr);
                goto cleanup;
            }
920
            virBufferAsprintf(&configbuf, "dhcp-range=%s,%s\n",
921
                              saddr, eaddr);
922
            VIR_FREE(saddr);
923
            VIR_FREE(eaddr);
924 925
            nbleases += virSocketAddrGetRange(&ipdef->ranges[r].start,
                                              &ipdef->ranges[r].end);
926
        }
927

928
        /*
929 930 931 932
         * 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)
933 934
         */
        if (!ipdef->nranges && ipdef->nhosts) {
935
            char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
936 937
            if (!bridgeaddr)
                goto cleanup;
938
            virBufferAsprintf(&configbuf, "dhcp-range=%s,static\n", bridgeaddr);
939 940
            VIR_FREE(bridgeaddr);
        }
941

G
Gene Czarcinski 已提交
942 943
        if (networkBuildDnsmasqDhcpHostsList(dctx, ipdef) < 0)
            goto cleanup;
944

G
Gene Czarcinski 已提交
945 946 947
        /* Note: the following is IPv4 only */
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
            if (ipdef->nranges || ipdef->nhosts)
948
                virBufferAddLit(&configbuf, "dhcp-no-override\n");
949

G
Gene Czarcinski 已提交
950
            if (ipdef->tftproot) {
951 952
                virBufferAddLit(&configbuf, "enable-tftp\n");
                virBufferAsprintf(&configbuf, "tftp-root=%s\n", ipdef->tftproot);
G
Gene Czarcinski 已提交
953
            }
954

G
Gene Czarcinski 已提交
955 956 957
            if (ipdef->bootfile) {
                if (VIR_SOCKET_ADDR_VALID(&ipdef->bootserver)) {
                    char *bootserver = virSocketAddrFormat(&ipdef->bootserver);
958

959
                    if (!bootserver)
G
Gene Czarcinski 已提交
960
                        goto cleanup;
961
                    virBufferAsprintf(&configbuf, "dhcp-boot=%s%s%s\n",
962
                                      ipdef->bootfile, ",,", bootserver);
G
Gene Czarcinski 已提交
963 964
                    VIR_FREE(bootserver);
                } else {
965
                    virBufferAsprintf(&configbuf, "dhcp-boot=%s\n", ipdef->bootfile);
G
Gene Czarcinski 已提交
966 967 968 969 970
                }
            }
        }
        ipdef = (ipdef == ipv6def) ? NULL : ipv6def;
    }
971

G
Gene Czarcinski 已提交
972 973
    if (nbleases > 0) {
        char *leasefile = networkDnsmasqLeaseFileName(network->def->name);
974
        if (!leasefile)
G
Gene Czarcinski 已提交
975
            goto cleanup;
976
        virBufferAsprintf(&configbuf, "dhcp-leasefile=%s\n", leasefile);
G
Gene Czarcinski 已提交
977
        VIR_FREE(leasefile);
978
        virBufferAsprintf(&configbuf, "dhcp-lease-max=%d\n", nbleases);
G
Gene Czarcinski 已提交
979
    }
980

G
Gene Czarcinski 已提交
981 982
    /* this is done once per interface */
    if (networkBuildDnsmasqHostsList(dctx, dns) < 0)
983
        goto cleanup;
G
Gene Czarcinski 已提交
984 985 986 987 988 989

    /* 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)
990 991
        virBufferAsprintf(&configbuf, "dhcp-hostsfile=%s\n",
                          dctx->hostsfile->path);
G
Gene Czarcinski 已提交
992

993 994
    /* Likewise, always create this file and put it on the
     * commandline, to allow for runtime additions.
G
Gene Czarcinski 已提交
995
     */
996
    virBufferAsprintf(&configbuf, "addn-hosts=%s\n",
997
                      dctx->addnhostsfile->path);
G
Gene Czarcinski 已提交
998 999 1000 1001

    /* Are we doing RA instead of radvd? */
    if (DNSMASQ_RA_SUPPORT(caps)) {
        if (ipv6def)
1002
            virBufferAddLit(&configbuf, "enable-ra\n");
G
Gene Czarcinski 已提交
1003
        else {
1004 1005 1006
            for (i = 0;
                 (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
                 i++) {
G
Gene Czarcinski 已提交
1007 1008 1009 1010
                if (!(ipdef->nranges || ipdef->nhosts)) {
                    char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
                    if (!bridgeaddr)
                        goto cleanup;
1011 1012
                    virBufferAsprintf(&configbuf,
                                      "dhcp-range=%s,ra-only\n", bridgeaddr);
G
Gene Czarcinski 已提交
1013 1014
                    VIR_FREE(bridgeaddr);
                }
1015
            }
1016
        }
1017 1018
    }

1019 1020 1021
    if (!(*configstr = virBufferContentAndReset(&configbuf)))
        goto cleanup;

1022
    ret = 0;
G
Gene Czarcinski 已提交
1023

1024
cleanup:
1025
    virBufferFreeAndReset(&configbuf);
1026 1027 1028 1029
    VIR_FREE(record);
    VIR_FREE(recordPort);
    VIR_FREE(recordWeight);
    VIR_FREE(recordPriority);
1030
    return ret;
1031 1032
}

1033
/* build the dnsmasq command line */
1034 1035 1036
static int ATTRIBUTE_NONNULL(2)
networkBuildDhcpDaemonCommandLine(virNetworkObjPtr network,
                                  virCommandPtr *cmdout,
1037 1038
                                  char *pidfile, dnsmasqContext *dctx,
                                  dnsmasqCapsPtr caps)
1039
{
1040
    virCommandPtr cmd = NULL;
G
Gene Czarcinski 已提交
1041
    int ret = -1;
1042 1043
    char *configfile = NULL;
    char *configstr = NULL;
1044 1045

    network->dnsmasqPid = -1;
1046

1047 1048 1049 1050 1051 1052
    if (networkDnsmasqConfContents(network, pidfile, &configstr, dctx, caps) < 0)
        goto cleanup;
    if (!configstr)
        goto cleanup;

    /* construct the filename */
1053
    if (!(configfile = networkDnsmasqConfigFileName(network->def->name)))
1054 1055 1056 1057 1058 1059 1060
        goto cleanup;

    /* Write the file */
    if (virFileWriteStr(configfile, configstr, 0600) < 0) {
        virReportSystemError(errno,
                         _("couldn't write dnsmasq config file '%s'"),
                         configfile);
1061 1062 1063
        goto cleanup;
    }

1064 1065
    cmd = virCommandNew(dnsmasqCapsGetBinaryPath(caps));
    virCommandAddArgFormat(cmd, "--conf-file=%s", configfile);
1066
    *cmdout = cmd;
1067 1068
    ret = 0;
cleanup:
1069 1070
    VIR_FREE(configfile);
    VIR_FREE(configstr);
1071 1072 1073 1074
    return ret;
}

static int
1075
networkStartDhcpDaemon(virNetworkDriverStatePtr driver,
1076
                       virNetworkObjPtr network)
1077 1078 1079 1080
{
    virCommandPtr cmd = NULL;
    char *pidfile = NULL;
    int ret = -1;
1081
    dnsmasqContext *dctx = NULL;
1082

1083
    if (!virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, 0)) {
G
Gene Czarcinski 已提交
1084
        /* no IP addresses, so we don't need to run */
1085 1086 1087 1088
        ret = 0;
        goto cleanup;
    }

1089
    if (virFileMakePath(driverState->pidDir) < 0) {
1090
        virReportSystemError(errno,
1091
                             _("cannot create directory %s"),
1092
                             driverState->pidDir);
1093
        goto cleanup;
1094
    }
1095
    if (virFileMakePath(driverState->stateDir) < 0) {
1096
        virReportSystemError(errno,
1097
                             _("cannot create directory %s"),
1098
                             driverState->stateDir);
1099
        goto cleanup;
1100 1101
    }

1102
    if (!(pidfile = virPidFileBuildPath(driverState->pidDir,
1103
                                        network->def->name)))
1104
        goto cleanup;
1105

1106
    if (virFileMakePath(driverState->dnsmasqStateDir) < 0) {
1107
        virReportSystemError(errno,
1108
                             _("cannot create directory %s"),
1109
                             driverState->dnsmasqStateDir);
1110 1111 1112
        goto cleanup;
    }

1113
    dctx = dnsmasqContextNew(network->def->name, driverState->dnsmasqStateDir);
1114 1115 1116
    if (dctx == NULL)
        goto cleanup;

1117 1118
    if (dnsmasqCapsRefresh(&driver->dnsmasqCaps, NULL) < 0)
        goto cleanup;
1119 1120 1121

    ret = networkBuildDhcpDaemonCommandLine(network, &cmd, pidfile,
                                            dctx, driver->dnsmasqCaps);
1122 1123 1124 1125 1126
    if (ret < 0)
        goto cleanup;

    ret = dnsmasqSave(dctx);
    if (ret < 0)
1127
        goto cleanup;
1128

G
Guido Günther 已提交
1129 1130
    ret = virCommandRun(cmd, NULL);
    if (ret < 0) {
1131
        goto cleanup;
G
Guido Günther 已提交
1132
    }
1133 1134

    /*
1135 1136 1137 1138 1139
     * 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
1140 1141
     */

1142
    ret = virPidFileRead(driverState->pidDir, network->def->name,
1143 1144
                         &network->dnsmasqPid);
    if (ret < 0)
1145
        goto cleanup;
1146

1147 1148 1149
    ret = 0;
cleanup:
    VIR_FREE(pidfile);
1150
    virCommandFree(cmd);
1151
    dnsmasqContextFree(dctx);
1152 1153 1154
    return ret;
}

1155 1156
/* networkRefreshDhcpDaemon:
 *  Update dnsmasq config files, then send a SIGHUP so that it rereads
G
Gene Czarcinski 已提交
1157 1158
 *  them.   This only works for the dhcp-hostsfile and the
 *  addn-hosts file.
1159 1160 1161
 *
 *  Returns 0 on success, -1 on failure.
 */
1162
static int
1163
networkRefreshDhcpDaemon(virNetworkDriverStatePtr driver,
1164
                         virNetworkObjPtr network)
1165
{
1166 1167
    int ret = -1;
    size_t i;
G
Gene Czarcinski 已提交
1168
    virNetworkIpDefPtr ipdef, ipv4def, ipv6def;
1169
    dnsmasqContext *dctx = NULL;
1170

G
Gene Czarcinski 已提交
1171
    /* if no IP addresses specified, nothing to do */
1172
    if (!virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, 0))
G
Gene Czarcinski 已提交
1173 1174
        return 0;

1175 1176
    /* if there's no running dnsmasq, just start it */
    if (network->dnsmasqPid <= 0 || (kill(network->dnsmasqPid, 0) < 0))
1177
        return networkStartDhcpDaemon(driver, network);
1178

G
Gene Czarcinski 已提交
1179
    VIR_INFO("Refreshing dnsmasq for network %s", network->def->bridge);
1180 1181
    if (!(dctx = dnsmasqContextNew(network->def->name,
                                   driverState->dnsmasqStateDir))) {
G
Gene Czarcinski 已提交
1182
        goto cleanup;
1183
    }
G
Gene Czarcinski 已提交
1184 1185 1186 1187 1188 1189

    /* 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;
1190 1191 1192
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, i));
         i++) {
G
Gene Czarcinski 已提交
1193 1194
        if (!ipv4def && (ipdef->nranges || ipdef->nhosts))
            ipv4def = ipdef;
1195 1196
    }

G
Gene Czarcinski 已提交
1197
    ipv6def = NULL;
1198 1199 1200
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
         i++) {
G
Gene Czarcinski 已提交
1201 1202
        if (!ipv6def && (ipdef->nranges || ipdef->nhosts))
            ipv6def = ipdef;
1203 1204
    }

G
Gene Czarcinski 已提交
1205 1206 1207 1208 1209
    if (ipv4def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv4def) < 0))
           goto cleanup;

    if (ipv6def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv6def) < 0))
           goto cleanup;
1210

G
Gene Czarcinski 已提交
1211
    if (networkBuildDnsmasqHostsList(dctx, &network->def->dns) < 0)
1212 1213 1214
       goto cleanup;

    if ((ret = dnsmasqSave(dctx)) < 0)
1215
        goto cleanup;
1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230

    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
1231
networkRestartDhcpDaemon(virNetworkDriverStatePtr driver,
1232
                         virNetworkObjPtr network)
1233 1234 1235 1236 1237 1238
{
    /* if there is a running dnsmasq, kill it */
    if (network->dnsmasqPid > 0) {
        networkKillDaemon(network->dnsmasqPid, "dnsmasq",
                          network->def->name);
        network->dnsmasqPid = -1;
1239
    }
1240
    /* now start dnsmasq if it should be started */
1241
    return networkStartDhcpDaemon(driver, network);
1242 1243
}

G
Gene Czarcinski 已提交
1244 1245 1246 1247 1248 1249
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";

1250 1251 1252
static int
networkRadvdConfContents(virNetworkObjPtr network, char **configstr)
{
E
Eric Blake 已提交
1253
    virBuffer configbuf = VIR_BUFFER_INITIALIZER;
1254 1255
    int ret = -1;
    size_t i;
1256
    virNetworkIpDefPtr ipdef;
G
Gene Czarcinski 已提交
1257
    bool v6present = false, dhcp6 = false;
1258 1259

    *configstr = NULL;
1260

G
Gene Czarcinski 已提交
1261
    /* Check if DHCPv6 is needed */
1262 1263 1264
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
         i++) {
G
Gene Czarcinski 已提交
1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277
        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;
    }

1278 1279 1280
    /* create radvd config file appropriate for this network;
     * IgnoreIfMissing allows radvd to start even when the bridge is down
     */
1281
    virBufferAsprintf(&configbuf, "interface %s\n"
1282 1283
                      "{\n"
                      "  AdvSendAdvert on;\n"
1284
                      "  IgnoreIfMissing on;\n"
G
Gene Czarcinski 已提交
1285 1286 1287 1288 1289
                      "  AdvManagedFlag %s;\n"
                      "%s",
                      network->def->bridge,
                      dhcp6 ? "on" : "off",
                      dhcp6 ? "\n" : radvd1);
1290 1291

    /* add a section for each IPv6 address in the config */
1292 1293 1294
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, i));
         i++) {
1295 1296 1297 1298 1299
        int prefix;
        char *netaddr;

        prefix = virNetworkIpDefPrefix(ipdef);
        if (prefix < 0) {
1300 1301 1302
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("bridge '%s' has an invalid prefix"),
                           network->def->bridge);
1303 1304
            goto cleanup;
        }
1305
        if (!(netaddr = virSocketAddrFormat(&ipdef->address)))
1306
            goto cleanup;
1307
        virBufferAsprintf(&configbuf,
1308
                          "  prefix %s/%d\n"
G
Gene Czarcinski 已提交
1309 1310 1311
                          "  {\n%s  };\n",
                          netaddr, prefix,
                          dhcp6 ? radvd2 : radvd3);
1312 1313 1314
        VIR_FREE(netaddr);
    }

1315 1316 1317
    /* only create the string if we found at least one IPv6 address */
    if (v6present) {
        virBufferAddLit(&configbuf, "};\n");
1318

1319 1320 1321 1322 1323 1324 1325 1326
        if (virBufferError(&configbuf)) {
            virReportOOMError();
            goto cleanup;
        }
        if (!(*configstr = virBufferContentAndReset(&configbuf))) {
            virReportOOMError();
            goto cleanup;
        }
1327
    }
1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352

    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;
1353 1354 1355 1356
        goto cleanup;
    }

    /* construct the filename */
1357
    if (!(*configFile = networkRadvdConfigFileName(network->def->name)))
1358 1359
        goto cleanup;
    /* write the file */
1360
    if (virFileWriteStr(*configFile, configStr, 0600) < 0) {
1361 1362
        virReportSystemError(errno,
                             _("couldn't write radvd config file '%s'"),
1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374
                             *configFile);
        goto cleanup;
    }

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

static int
1375
networkStartRadvd(virNetworkDriverStatePtr driver ATTRIBUTE_UNUSED,
G
Gene Czarcinski 已提交
1376
                        virNetworkObjPtr network)
1377 1378 1379 1380 1381 1382 1383 1384 1385
{
    char *pidfile = NULL;
    char *radvdpidbase = NULL;
    char *configfile = NULL;
    virCommandPtr cmd = NULL;
    int ret = -1;

    network->radvdPid = -1;

G
Gene Czarcinski 已提交
1386
    /* Is dnsmasq handling RA? */
1387
   if (DNSMASQ_RA_SUPPORT(driver->dnsmasqCaps)) {
G
Gene Czarcinski 已提交
1388 1389 1390 1391
        ret = 0;
        goto cleanup;
    }

1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402
    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);
1403 1404 1405
        goto cleanup;
    }

1406
    if (virFileMakePath(driverState->pidDir) < 0) {
1407 1408
        virReportSystemError(errno,
                             _("cannot create directory %s"),
1409
                             driverState->pidDir);
1410 1411
        goto cleanup;
    }
1412
    if (virFileMakePath(driverState->radvdStateDir) < 0) {
1413 1414
        virReportSystemError(errno,
                             _("cannot create directory %s"),
1415
                             driverState->radvdStateDir);
1416 1417 1418 1419
        goto cleanup;
    }

    /* construct pidfile name */
1420
    if (!(radvdpidbase = networkRadvdPidfileBasename(network->def->name)))
1421
        goto cleanup;
1422
    if (!(pidfile = virPidFileBuildPath(driverState->pidDir, radvdpidbase)))
1423 1424 1425 1426 1427
        goto cleanup;

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

1428 1429 1430 1431
    /* 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
1432
     * virPidFileRead() below will fail if we use them).
1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447
     * 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;

1448
    if (virPidFileRead(driverState->pidDir, radvdpidbase, &network->radvdPid) < 0)
1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459
        goto cleanup;

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

1460
static int
1461
networkRefreshRadvd(virNetworkDriverStatePtr driver ATTRIBUTE_UNUSED,
1462
                    virNetworkObjPtr network)
1463
{
G
Gene Czarcinski 已提交
1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474
    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)) {
1475
            virPidFileDelete(driverState->pidDir, radvdpidbase);
G
Gene Czarcinski 已提交
1476 1477 1478 1479 1480 1481
            VIR_FREE(radvdpidbase);
        }
        network->radvdPid = -1;
        return 0;
    }

1482 1483
    /* if there's no running radvd, just start it */
    if (network->radvdPid <= 0 || (kill(network->radvdPid, 0) < 0))
G
Gene Czarcinski 已提交
1484
        return networkStartRadvd(driver, network);
1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496

    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);
}

1497 1498
#if 0
/* currently unused, so it causes a build error unless we #if it out */
1499
static int
1500
networkRestartRadvd(virNetworkDriverStatePtr driver,
1501
                    virNetworkObjPtr network)
1502 1503 1504 1505 1506 1507 1508 1509 1510
{
    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
1511
        if ((networkKillDaemon(network->radvdPid, "radvd",
1512 1513 1514
                               network->def->name) >= 0) &&
            ((radvdpidbase = networkRadvdPidfileBasename(network->def->name))
             != NULL)) {
1515
            virPidFileDelete(driverState->pidDir, radvdpidbase);
1516 1517 1518 1519 1520 1521 1522 1523 1524
            VIR_FREE(radvdpidbase);
        }
        network->radvdPid = -1;
    }
    /* now start radvd if it should be started */
    return networkStartRadvd(network);
}
#endif /* #if 0 */

1525 1526 1527 1528
/* SIGHUP/restart any dnsmasq or radvd daemons.
 * This should be called when libvirtd is restarted.
 */
static void
1529
networkRefreshDaemons(virNetworkDriverStatePtr driver)
1530
{
1531
    size_t i;
1532 1533 1534

    VIR_INFO("Refreshing network daemons");

1535
    for (i = 0; i < driver->networks.count; i++) {
1536 1537 1538 1539
        virNetworkObjPtr network = driver->networks.objs[i];

        virNetworkObjLock(network);
        if (virNetworkObjIsActive(network) &&
1540 1541 1542
            ((network->def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE))) {
1543 1544 1545 1546 1547 1548
            /* 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.
             */
1549 1550
            networkRefreshDhcpDaemon(driver, network);
            networkRefreshRadvd(driver, network);
1551 1552 1553 1554 1555
        }
        virNetworkObjUnlock(network);
    }
}

1556
static void
1557
networkReloadFirewallRules(virNetworkDriverStatePtr driver)
1558
{
1559
    size_t i;
1560

1561
    VIR_INFO("Reloading iptables rules");
1562

1563
    for (i = 0; i < driver->networks.count; i++) {
1564 1565 1566 1567
        virNetworkObjPtr network = driver->networks.objs[i];

        virNetworkObjLock(network);
        if (virNetworkObjIsActive(network) &&
1568 1569 1570
            ((network->def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE))) {
1571 1572 1573
            /* Only the three L3 network types that are configured by libvirt
             * need to have iptables rules reloaded.
             */
1574 1575
            networkRemoveFirewallRules(network);
            if (networkAddFirewallRules(network) < 0) {
1576 1577
                /* failed to add but already logged */
            }
1578
        }
1579
        virNetworkObjUnlock(network);
1580 1581 1582
    }
}

1583
/* Enable IP Forwarding. Return 0 for success, -1 for failure. */
1584
static int
1585
networkEnableIpForwarding(bool enableIPv4, bool enableIPv6)
1586
{
1587
    int ret = 0;
1588 1589 1590 1591 1592 1593 1594 1595 1596
#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
1597 1598 1599 1600
    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);
1601
#endif
1602
    return ret;
1603 1604
}

1605 1606
#define SYSCTL_PATH "/proc/sys"

1607 1608
static int
networkSetIPv6Sysctls(virNetworkObjPtr network)
1609 1610 1611 1612
{
    char *field = NULL;
    int ret = -1;

1613 1614 1615 1616 1617
    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",
1618
                        network->def->bridge) < 0)
1619
            goto cleanup;
1620

1621 1622 1623 1624 1625 1626
        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;
        }
1627

1628 1629 1630 1631 1632 1633 1634
        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);
1635 1636
    }

1637 1638 1639 1640 1641 1642 1643 1644
    /* 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",
1645
                    network->def->bridge) < 0)
1646 1647
        goto cleanup;

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

1655 1656 1657 1658
    /* 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",
1659
                    network->def->bridge) < 0)
1660 1661
        goto cleanup;

1662
    if (virFileWriteStr(field, "0", 0) < 0) {
1663
        virReportSystemError(errno,
1664
                             _("cannot disable %s"), field);
1665 1666 1667 1668 1669 1670 1671 1672 1673
        goto cleanup;
    }

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

1674
/* add an IP address to a bridge */
1675
static int
D
Daniel P. Berrange 已提交
1676
networkAddAddrToBridge(virNetworkObjPtr network,
1677
                       virNetworkIpDefPtr ipdef)
1678
{
1679 1680 1681
    int prefix = virNetworkIpDefPrefix(ipdef);

    if (prefix < 0) {
1682 1683 1684
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("bridge '%s' has an invalid netmask or IP address"),
                       network->def->bridge);
1685 1686 1687
        return -1;
    }

1688 1689
    if (virNetDevSetIPv4Address(network->def->bridge,
                                &ipdef->address, prefix) < 0)
1690 1691 1692 1693 1694
        return -1;

    return 0;
}

1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707
/* 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,
1708
                                    (VIR_SOCKET_ADDR_IS_FAMILY(addr, AF_INET)
1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743
                                     ? "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;
}

1744
static int
1745
networkStartNetworkVirtual(virNetworkDriverStatePtr driver,
1746 1747
                          virNetworkObjPtr network)
{
1748
    size_t i;
1749
    bool v4present = false, v6present = false;
1750 1751
    virErrorPtr save_err = NULL;
    virNetworkIpDefPtr ipdef;
1752
    virNetworkRouteDefPtr routedef;
1753
    char *macTapIfName = NULL;
1754
    int tapfd = -1;
1755

1756 1757
    /* Check to see if any network IP collides with an existing route */
    if (networkCheckRouteCollision(network) < 0)
1758 1759
        return -1;

1760
    /* Create and configure the bridge device */
1761
    if (virNetDevBridgeCreate(network->def->bridge) < 0)
1762 1763
        return -1;

1764 1765 1766 1767 1768 1769 1770 1771
    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);
1772
        if (!macTapIfName)
1773
            goto err0;
1774
        /* Keep tun fd open and interface up to allow for IPv6 DAD to happen */
1775
        if (virNetDevTapCreateInBridgePort(network->def->bridge,
1776
                                           &macTapIfName, &network->def->mac,
1777
                                           NULL, &tapfd, 1, NULL, NULL,
1778 1779 1780
                                           VIR_NETDEV_TAP_CREATE_USE_MAC_FOR_BRIDGE |
                                           VIR_NETDEV_TAP_CREATE_IFUP |
                                           VIR_NETDEV_TAP_CREATE_PERSIST) < 0) {
1781 1782 1783 1784 1785
            VIR_FREE(macTapIfName);
            goto err0;
        }
    }

1786
    /* Set bridge options */
1787 1788 1789 1790

    /* delay is configured in seconds, but virNetDevBridgeSetSTPDelay
     * expects milliseconds
     */
1791
    if (virNetDevBridgeSetSTPDelay(network->def->bridge,
1792
                                   network->def->delay * 1000) < 0)
1793
        goto err1;
1794

1795
    if (virNetDevBridgeSetSTP(network->def->bridge,
1796
                              network->def->stp ? true : false) < 0)
1797
        goto err1;
1798

1799 1800 1801 1802
    /* Disable IPv6 on the bridge if there are no IPv6 addresses
     * defined, and set other IPv6 sysctl tunables appropriately.
     */
    if (networkSetIPv6Sysctls(network) < 0)
1803
        goto err1;
1804

1805
    /* Add "once per network" rules */
1806
    if (networkAddFirewallRules(network) < 0)
1807 1808
        goto err1;

1809 1810 1811
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, i));
         i++) {
1812
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET))
1813
            v4present = true;
1814
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
1815
            v6present = true;
1816

1817
        /* Add the IP address/netmask to the bridge */
D
Daniel P. Berrange 已提交
1818
        if (networkAddAddrToBridge(network, ipdef) < 0) {
1819
            goto err2;
1820
        }
1821 1822
    }

1823
    /* Bring up the bridge interface */
1824
    if (virNetDevSetOnline(network->def->bridge, 1) < 0)
1825
        goto err2;
1826

1827 1828
    for (i = 0; i < network->def->nroutes; i++) {
        routedef = &network->def->routes[i];
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839
        /* 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 */
            }
        }
    }

1840 1841
    /* If forward.type != NONE, turn on global IP forwarding */
    if (network->def->forward.type != VIR_NETWORK_FORWARD_NONE &&
1842
        networkEnableIpForwarding(v4present, v6present) < 0) {
1843
        virReportSystemError(errno, "%s",
1844
                             _("failed to enable IP forwarding"));
1845
        goto err3;
1846 1847
    }

1848

1849
    /* start dnsmasq if there are any IP addresses (v4 or v6) */
1850 1851
    if ((v4present || v6present) &&
        networkStartDhcpDaemon(driver, network) < 0)
1852
        goto err3;
1853

1854
    /* start radvd if there are any ipv6 addresses */
G
Gene Czarcinski 已提交
1855
    if (v6present && networkStartRadvd(driver, network) < 0)
1856 1857
        goto err4;

1858 1859 1860 1861 1862 1863 1864 1865 1866
    /* 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);
    }

1867 1868
    if (virNetDevBandwidthSet(network->def->bridge,
                              network->def->bandwidth, true) < 0) {
1869 1870 1871
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot set bandwidth limits on %s"),
                       network->def->bridge);
1872 1873 1874
        goto err5;
    }

1875
    VIR_FREE(macTapIfName);
1876 1877 1878

    return 0;

1879
 err5:
1880
    virNetDevBandwidthClear(network->def->bridge);
1881

1882 1883 1884 1885
 err4:
    if (!save_err)
        save_err = virSaveLastError();

1886 1887 1888 1889 1890
    if (network->dnsmasqPid > 0) {
        kill(network->dnsmasqPid, SIGTERM);
        network->dnsmasqPid = -1;
    }

1891 1892 1893
 err3:
    if (!save_err)
        save_err = virSaveLastError();
1894
    ignore_value(virNetDevSetOnline(network->def->bridge, 0));
1895

1896 1897 1898
 err2:
    if (!save_err)
        save_err = virSaveLastError();
1899
    networkRemoveFirewallRules(network);
1900 1901

 err1:
1902 1903 1904
    if (!save_err)
        save_err = virSaveLastError();

H
Hu Tao 已提交
1905
    if (macTapIfName) {
1906
        VIR_FORCE_CLOSE(tapfd);
H
Hu Tao 已提交
1907 1908 1909
        ignore_value(virNetDevTapDelete(macTapIfName));
        VIR_FREE(macTapIfName);
    }
1910 1911

 err0:
1912 1913
    if (!save_err)
        save_err = virSaveLastError();
1914
    ignore_value(virNetDevBridgeDelete(network->def->bridge));
1915

1916 1917 1918 1919
    if (save_err) {
        virSetError(save_err);
        virFreeError(save_err);
    }
1920
    /* coverity[leaked_handle] - 'tapfd' is not leaked */
1921 1922 1923
    return -1;
}

1924
static int networkShutdownNetworkVirtual(virNetworkDriverStatePtr driver ATTRIBUTE_UNUSED,
R
Roman Bogorodskiy 已提交
1925
                                         virNetworkObjPtr network)
1926
{
1927
    virNetDevBandwidthClear(network->def->bridge);
1928

1929 1930 1931 1932 1933
    if (network->radvdPid > 0) {
        char *radvdpidbase;

        kill(network->radvdPid, SIGTERM);
        /* attempt to delete the pidfile we created */
1934
        if ((radvdpidbase = networkRadvdPidfileBasename(network->def->name))) {
1935
            virPidFileDelete(driverState->pidDir, radvdpidbase);
1936 1937 1938 1939
            VIR_FREE(radvdpidbase);
        }
    }

1940 1941 1942
    if (network->dnsmasqPid > 0)
        kill(network->dnsmasqPid, SIGTERM);

1943
    if (network->def->mac_specified) {
1944
        char *macTapIfName = networkBridgeDummyNicName(network->def->bridge);
1945
        if (macTapIfName) {
1946
            ignore_value(virNetDevTapDelete(macTapIfName));
1947 1948 1949 1950
            VIR_FREE(macTapIfName);
        }
    }

1951
    ignore_value(virNetDevSetOnline(network->def->bridge, 0));
1952

1953
    networkRemoveFirewallRules(network);
1954

1955
    ignore_value(virNetDevBridgeDelete(network->def->bridge));
1956

1957
    /* See if its still alive and really really kill it */
1958
    if (network->dnsmasqPid > 0 &&
1959
        (kill(network->dnsmasqPid, 0) == 0))
1960 1961
        kill(network->dnsmasqPid, SIGKILL);
    network->dnsmasqPid = -1;
1962 1963 1964 1965 1966 1967

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

1968 1969 1970 1971
    return 0;
}

static int
1972
networkStartNetworkExternal(virNetworkDriverStatePtr driver ATTRIBUTE_UNUSED,
1973 1974 1975
                            virNetworkObjPtr network ATTRIBUTE_UNUSED)
{
    /* put anything here that needs to be done each time a network of
1976
     * type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is started. On
1977 1978 1979 1980 1981 1982
     * failure, undo anything you've done, and return -1. On success
     * return 0.
     */
    return 0;
}

1983
static int networkShutdownNetworkExternal(virNetworkDriverStatePtr driver ATTRIBUTE_UNUSED,
1984 1985 1986
                                        virNetworkObjPtr network ATTRIBUTE_UNUSED)
{
    /* put anything here that needs to be done each time a network of
1987
     * type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is shutdown. On
1988 1989 1990 1991 1992 1993 1994
     * failure, undo anything you've done, and return -1. On success
     * return 0.
     */
    return 0;
}

static int
1995
networkStartNetwork(virNetworkDriverStatePtr driver,
1996 1997 1998 1999 2000
                    virNetworkObjPtr network)
{
    int ret = 0;

    if (virNetworkObjIsActive(network)) {
2001 2002
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("network is already active"));
2003 2004 2005
        return -1;
    }

2006 2007 2008
    if (virNetworkObjSetDefTransient(network, true) < 0)
        return -1;

2009
    switch (network->def->forward.type) {
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

    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:
2021
    case VIR_NETWORK_FORWARD_HOSTDEV:
2022 2023 2024 2025
        ret = networkStartNetworkExternal(driver, network);
        break;
    }

2026 2027
    if (ret < 0) {
        virNetworkObjUnsetDefTransient(network);
2028
        return ret;
2029
    }
2030 2031 2032 2033

    /* Persist the live configuration now that anything autogenerated
     * is setup.
     */
2034 2035
    if ((ret = virNetworkSaveStatus(driverState->stateDir,
                                    network)) < 0) {
2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053
        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;
}

2054
static int networkShutdownNetwork(virNetworkDriverStatePtr driver,
2055 2056 2057 2058 2059 2060 2061 2062 2063 2064
                                        virNetworkObjPtr network)
{
    int ret = 0;
    char *stateFile;

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

    if (!virNetworkObjIsActive(network))
        return 0;

2065 2066
    stateFile = virNetworkConfigFile(driverState->stateDir,
                                     network->def->name);
2067 2068 2069 2070 2071 2072
    if (!stateFile)
        return -1;

    unlink(stateFile);
    VIR_FREE(stateFile);

2073
    switch (network->def->forward.type) {
2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084

    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:
2085
    case VIR_NETWORK_FORWARD_HOSTDEV:
2086 2087 2088 2089
        ret = networkShutdownNetworkExternal(driver, network);
        break;
    }

2090
    network->active = 0;
2091
    virNetworkObjUnsetDefTransient(network);
2092
    return ret;
2093 2094 2095
}


2096 2097
static virNetworkPtr networkLookupByUUID(virConnectPtr conn,
                                         const unsigned char *uuid) {
2098
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2099 2100
    virNetworkObjPtr network;
    virNetworkPtr ret = NULL;
2101

2102
    networkDriverLock(driver);
2103
    network = virNetworkFindByUUID(&driver->networks, uuid);
2104
    networkDriverUnlock(driver);
2105
    if (!network) {
2106 2107
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2108
        goto cleanup;
2109 2110
    }

2111 2112 2113
    if (virNetworkLookupByUUIDEnsureACL(conn, network->def) < 0)
        goto cleanup;

2114 2115 2116
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

cleanup:
2117 2118
    if (network)
        virNetworkObjUnlock(network);
2119
    return ret;
2120 2121
}

2122 2123
static virNetworkPtr networkLookupByName(virConnectPtr conn,
                                         const char *name) {
2124
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2125 2126 2127
    virNetworkObjPtr network;
    virNetworkPtr ret = NULL;

2128
    networkDriverLock(driver);
2129
    network = virNetworkFindByName(&driver->networks, name);
2130
    networkDriverUnlock(driver);
2131
    if (!network) {
2132 2133
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"), name);
2134
        goto cleanup;
2135 2136
    }

2137 2138 2139
    if (virNetworkLookupByNameEnsureACL(conn, network->def) < 0)
        goto cleanup;

2140 2141 2142
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

cleanup:
2143 2144
    if (network)
        virNetworkObjUnlock(network);
2145
    return ret;
2146 2147
}

2148 2149 2150
static virDrvOpenStatus networkOpen(virConnectPtr conn,
                                    virConnectAuthPtr auth ATTRIBUTE_UNUSED,
                                    unsigned int flags)
2151 2152 2153
{
    virCheckFlags(VIR_CONNECT_RO, VIR_DRV_OPEN_ERROR);

2154 2155 2156 2157 2158 2159 2160
    if (!driverState)
        return VIR_DRV_OPEN_DECLINED;

    conn->networkPrivateData = driverState;
    return VIR_DRV_OPEN_SUCCESS;
}

2161
static int networkClose(virConnectPtr conn) {
2162 2163 2164 2165
    conn->networkPrivateData = NULL;
    return 0;
}

2166
static int networkConnectNumOfNetworks(virConnectPtr conn) {
2167 2168
    int nactive = 0;
    size_t i;
2169
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2170

2171 2172 2173
    if (virConnectNumOfNetworksEnsureACL(conn) < 0)
        return -1;

2174
    networkDriverLock(driver);
2175
    for (i = 0; i < driver->networks.count; i++) {
2176 2177 2178 2179
        virNetworkObjPtr obj = driver->networks.objs[i];
        virNetworkObjLock(obj);
        if (virConnectNumOfNetworksCheckACL(conn, obj->def) &&
            virNetworkObjIsActive(obj))
2180
            nactive++;
2181
        virNetworkObjUnlock(obj);
2182 2183
    }
    networkDriverUnlock(driver);
2184

2185 2186 2187
    return nactive;
}

2188
static int networkConnectListNetworks(virConnectPtr conn, char **const names, int nnames) {
2189
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2190 2191
    int got = 0;
    size_t i;
2192

2193 2194 2195
    if (virConnectListNetworksEnsureACL(conn) < 0)
        return -1;

2196
    networkDriverLock(driver);
2197
    for (i = 0; i < driver->networks.count && got < nnames; i++) {
2198 2199 2200 2201 2202 2203
        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);
2204 2205 2206 2207
                goto cleanup;
            }
            got++;
        }
2208
        virNetworkObjUnlock(obj);
2209
    }
2210 2211
    networkDriverUnlock(driver);

2212 2213 2214
    return got;

 cleanup:
2215
    networkDriverUnlock(driver);
2216
    for (i = 0; i < got; i++)
2217 2218 2219 2220
        VIR_FREE(names[i]);
    return -1;
}

2221
static int networkConnectNumOfDefinedNetworks(virConnectPtr conn) {
2222 2223
    int ninactive = 0;
    size_t i;
2224
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2225

2226 2227 2228
    if (virConnectNumOfDefinedNetworksEnsureACL(conn) < 0)
        return -1;

2229
    networkDriverLock(driver);
2230
    for (i = 0; i < driver->networks.count; i++) {
2231 2232 2233 2234
        virNetworkObjPtr obj = driver->networks.objs[i];
        virNetworkObjLock(obj);
        if (virConnectNumOfDefinedNetworksCheckACL(conn, obj->def) &&
            !virNetworkObjIsActive(obj))
2235
            ninactive++;
2236
        virNetworkObjUnlock(obj);
2237 2238
    }
    networkDriverUnlock(driver);
2239

2240 2241 2242
    return ninactive;
}

2243
static int networkConnectListDefinedNetworks(virConnectPtr conn, char **const names, int nnames) {
2244
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2245 2246
    int got = 0;
    size_t i;
2247

2248 2249 2250
    if (virConnectListDefinedNetworksEnsureACL(conn) < 0)
        return -1;

2251
    networkDriverLock(driver);
2252
    for (i = 0; i < driver->networks.count && got < nnames; i++) {
2253 2254 2255 2256 2257 2258
        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);
2259 2260 2261 2262
                goto cleanup;
            }
            got++;
        }
2263
        virNetworkObjUnlock(obj);
2264
    }
2265
    networkDriverUnlock(driver);
2266 2267 2268
    return got;

 cleanup:
2269
    networkDriverUnlock(driver);
2270
    for (i = 0; i < got; i++)
2271 2272 2273 2274
        VIR_FREE(names[i]);
    return -1;
}

2275
static int
2276 2277 2278
networkConnectListAllNetworks(virConnectPtr conn,
                              virNetworkPtr **nets,
                              unsigned int flags)
2279
{
2280
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2281 2282 2283 2284
    int ret = -1;

    virCheckFlags(VIR_CONNECT_LIST_NETWORKS_FILTERS_ALL, -1);

2285 2286 2287
    if (virConnectListAllNetworksEnsureACL(conn) < 0)
        goto cleanup;

2288
    networkDriverLock(driver);
2289 2290 2291
    ret = virNetworkObjListExport(conn, driver->networks, nets,
                                  virConnectListAllNetworksCheckACL,
                                  flags);
2292 2293
    networkDriverUnlock(driver);

2294
cleanup:
2295 2296
    return ret;
}
2297

2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312
static int
networkConnectNetworkEventRegisterAny(virConnectPtr conn,
                                      virNetworkPtr net,
                                      int eventID,
                                      virConnectNetworkEventGenericCallback callback,
                                      void *opaque,
                                      virFreeCallback freecb)
{
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
    int ret = -1;

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

    if (virNetworkEventStateRegisterID(conn, driver->networkEventState,
2313
                                       virConnectNetworkEventRegisterAnyCheckACL,
2314
                                       net, eventID, callback,
2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331
                                       opaque, freecb, &ret) < 0)
        ret = -1;

cleanup:
    return ret;
}

static int
networkConnectNetworkEventDeregisterAny(virConnectPtr conn,
                                        int callbackID)
{
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
    int ret = -1;

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

2332 2333 2334 2335 2336 2337
    if (virObjectEventStateDeregisterID(conn,
                                        driver->networkEventState,
                                        callbackID) < 0)
        goto cleanup;

    ret = 0;
2338 2339 2340 2341 2342

cleanup:
    return ret;
}

2343 2344 2345 2346 2347
static int networkIsActive(virNetworkPtr net)
{
    virNetworkObjPtr obj;
    int ret = -1;

2348 2349
    if (!(obj = networkObjFromNetwork(net)))
        return ret;
2350 2351 2352 2353

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

2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366
    ret = virNetworkObjIsActive(obj);

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

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

2367 2368
    if (!(obj = networkObjFromNetwork(net)))
        return ret;
2369 2370 2371 2372

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

2373 2374 2375 2376 2377 2378 2379 2380 2381
    ret = obj->persistent;

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


2382
static int
2383
networkValidate(virNetworkDriverStatePtr driver,
2384 2385
                virNetworkDefPtr def,
                bool check_active)
2386
{
2387
    size_t i;
2388 2389
    bool vlanUsed, vlanAllowed, badVlanUse = false;
    virPortGroupDefPtr defaultPortGroup = NULL;
2390
    virNetworkIpDefPtr ipdef;
G
Gene Czarcinski 已提交
2391
    bool ipv4def = false, ipv6def = false;
2392 2393 2394 2395 2396 2397 2398 2399

    /* 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
     */
2400 2401 2402
    if (def->forward.type == VIR_NETWORK_FORWARD_NONE ||
        def->forward.type == VIR_NETWORK_FORWARD_NAT ||
        def->forward.type == VIR_NETWORK_FORWARD_ROUTE) {
2403 2404 2405 2406 2407

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

        virNetworkSetBridgeMacAddr(def);
2408 2409 2410 2411 2412 2413 2414 2415 2416
    } 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,
2417
                           virNetworkForwardTypeToString(def->forward.type));
2418 2419
            return -1;
        }
2420
        if (def->dns.ntxts || def->dns.nhosts || def->dns.nsrvs) {
2421 2422 2423 2424
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <dns> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
2425
                           virNetworkForwardTypeToString(def->forward.type));
2426 2427 2428 2429 2430 2431 2432
            return -1;
        }
        if (def->domain) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <domain> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
2433
                           virNetworkForwardTypeToString(def->forward.type));
2434 2435
            return -1;
        }
2436 2437
    }

G
Gene Czarcinski 已提交
2438 2439 2440
    /* We only support dhcp on one IPv4 address and
     * on one IPv6 address per defined network
     */
2441 2442 2443
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(def, AF_UNSPEC, i));
         i++) {
G
Gene Czarcinski 已提交
2444 2445 2446 2447 2448
        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 -- "
2449 2450
                                 "dhcp is supported only for a "
                                 "single IPv4 address on each network"));
G
Gene Czarcinski 已提交
2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467
                    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;
                }
2468 2469 2470
            }
        }
    }
2471 2472 2473 2474 2475 2476

    /* 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.
     */

2477
    vlanAllowed = ((def->forward.type == VIR_NETWORK_FORWARD_BRIDGE &&
2478
                   def->virtPortProfile &&
2479 2480 2481
                   def->virtPortProfile->virtPortType
                    == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) ||
                   def->forward.type == VIR_NETWORK_FORWARD_HOSTDEV);
2482 2483

    vlanUsed = def->vlan.nTags > 0;
2484 2485
    for (i = 0; i < def->nPortGroups; i++) {
        if (vlanUsed || def->portGroups[i].vlan.nTags > 0) {
2486 2487 2488 2489 2490
            /* 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.
             */
2491
            if (def->portGroups[i].virtPortProfile) {
2492
                if (def->forward.type != VIR_NETWORK_FORWARD_BRIDGE ||
2493
                    def->portGroups[i].virtPortProfile->virtPortType
2494 2495 2496 2497 2498 2499 2500
                    != VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
                    badVlanUse = true;
                }
            } else if (!vlanAllowed) {
                /* virtualport taken from base network definition */
                badVlanUse = true;
            }
2501
        }
2502
        if (def->portGroups[i].isDefault) {
2503 2504 2505 2506 2507
            if (defaultPortGroup) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("network '%s' has multiple default "
                                 "<portgroup> elements (%s and %s), "
                                 "but only one default is allowed"),
2508
                               def->name, defaultPortGroup->name,
2509
                               def->portGroups[i].name);
2510
                return -1;
2511
            }
2512
            defaultPortGroup = &def->portGroups[i];
2513
        }
2514
    }
2515 2516 2517 2518 2519 2520 2521
    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.
         */
2522 2523 2524 2525 2526 2527 2528 2529 2530
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       _("<vlan> element specified for network %s, "
                         "whose type doesn't support vlan configuration"),
                       def->name);
        return -1;
    }
    return 0;
}

2531
static virNetworkPtr networkCreateXML(virConnectPtr conn, const char *xml) {
2532
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2533
    virNetworkDefPtr def;
2534
    virNetworkObjPtr network = NULL;
2535
    virNetworkPtr ret = NULL;
2536
    virObjectEventPtr event = NULL;
2537

2538 2539
    networkDriverLock(driver);

2540
    if (!(def = virNetworkDefParseString(xml)))
2541
        goto cleanup;
2542

2543 2544 2545
    if (virNetworkCreateXMLEnsureACL(conn, def) < 0)
        goto cleanup;

2546
    if (networkValidate(driver, def, true) < 0)
2547 2548
       goto cleanup;

2549 2550 2551 2552
    /* NB: "live" is false because this transient network hasn't yet
     * been started
     */
    if (!(network = virNetworkAssignDef(&driver->networks, def, false)))
2553 2554
        goto cleanup;
    def = NULL;
2555

2556
    if (networkStartNetwork(driver, network) < 0) {
2557 2558
        virNetworkRemoveInactive(&driver->networks,
                                 network);
2559
        network = NULL;
2560
        goto cleanup;
2561 2562
    }

2563 2564
    event = virNetworkEventLifecycleNew(network->def->name,
                                        network->def->uuid,
2565 2566
                                        VIR_NETWORK_EVENT_STARTED,
                                        0);
2567

2568
    VIR_INFO("Creating network '%s'", network->def->name);
2569 2570 2571 2572
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

cleanup:
    virNetworkDefFree(def);
2573 2574
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
2575 2576 2577
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
2578
    return ret;
2579 2580
}

2581
static virNetworkPtr networkDefineXML(virConnectPtr conn, const char *xml) {
2582
    virNetworkDriverStatePtr driver = conn->networkPrivateData;
2583
    virNetworkDefPtr def = NULL;
2584
    bool freeDef = true;
2585
    virNetworkObjPtr network = NULL;
2586
    virNetworkPtr ret = NULL;
2587
    virObjectEventPtr event = NULL;
2588

2589 2590
    networkDriverLock(driver);

2591
    if (!(def = virNetworkDefParseString(xml)))
2592
        goto cleanup;
2593

2594 2595 2596
    if (virNetworkDefineXMLEnsureACL(conn, def) < 0)
        goto cleanup;

2597
    if (networkValidate(driver, def, false) < 0)
2598 2599
       goto cleanup;

2600 2601 2602 2603 2604 2605 2606 2607
    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;
    }
2608

2609 2610 2611
    /* define makes the network persistent - always */
    network->persistent = 1;

2612 2613
    /* def was asigned */
    freeDef = false;
2614 2615

    if (virNetworkSaveConfig(driver->networkConfigDir, def) < 0) {
2616 2617 2618 2619 2620 2621 2622 2623
        if (!virNetworkObjIsActive(network)) {
            virNetworkRemoveInactive(&driver->networks, network);
            network = NULL;
            goto cleanup;
        }
        network->persistent = 0;
        virNetworkDefFree(network->newDef);
        network->newDef = NULL;
2624 2625 2626
        goto cleanup;
    }

2627
    event = virNetworkEventLifecycleNew(def->name, def->uuid,
2628 2629
                                        VIR_NETWORK_EVENT_DEFINED,
                                        0);
2630

2631 2632
    VIR_INFO("Defining network '%s'", def->name);
    ret = virGetNetwork(conn, def->name, def->uuid);
2633 2634

cleanup:
2635 2636
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
2637 2638
    if (freeDef)
       virNetworkDefFree(def);
2639 2640 2641
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
2642
    return ret;
2643 2644
}

2645 2646
static int
networkUndefine(virNetworkPtr net) {
2647
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
2648
    virNetworkObjPtr network;
2649
    int ret = -1;
2650
    bool active = false;
2651
    virObjectEventPtr event = NULL;
2652

2653 2654
    networkDriverLock(driver);

2655
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
2656
    if (!network) {
2657 2658
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2659
        goto cleanup;
2660 2661
    }

2662 2663 2664
    if (virNetworkUndefineEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2665 2666
    if (virNetworkObjIsActive(network))
        active = true;
2667

2668
    if (virNetworkDeleteConfig(driver->networkConfigDir,
2669 2670
                               driver->networkAutostartDir,
                               network) < 0)
2671
        goto cleanup;
2672

2673 2674
    /* make the network transient */
    network->persistent = 0;
2675
    network->autostart = 0;
2676 2677 2678
    virNetworkDefFree(network->newDef);
    network->newDef = NULL;

2679 2680
    event = virNetworkEventLifecycleNew(network->def->name,
                                        network->def->uuid,
2681 2682
                                        VIR_NETWORK_EVENT_UNDEFINED,
                                        0);
2683

2684
    VIR_INFO("Undefining network '%s'", network->def->name);
2685 2686 2687 2688 2689
    if (!active) {
        if (networkRemoveInactive(driver, network) < 0) {
            network = NULL;
            goto cleanup;
        }
2690
        network = NULL;
2691 2692
    }

2693
    ret = 0;
2694

2695
cleanup:
2696 2697
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
2698 2699 2700
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
2701
    return ret;
2702 2703
}

2704 2705 2706 2707 2708 2709 2710 2711
static int
networkUpdate(virNetworkPtr net,
              unsigned int command,
              unsigned int section,
              int parentIndex,
              const char *xml,
              unsigned int flags)
{
2712
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
2713
    virNetworkObjPtr network = NULL;
2714 2715
    int isActive, ret = -1;
    size_t i;
2716 2717
    virNetworkIpDefPtr ipdef;
    bool oldDhcpActive = false;
2718
    bool needFirewallRefresh = false;
2719

2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733

    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;
    }

2734 2735 2736
    if (virNetworkUpdateEnsureACL(net->conn, network->def, flags) < 0)
        goto cleanup;

2737
    /* see if we are listening for dhcp pre-modification */
2738 2739 2740
    for (i = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, i));
         i++) {
2741 2742 2743 2744 2745 2746
        if (ipdef->nranges || ipdef->nhosts) {
            oldDhcpActive = true;
            break;
        }
    }

2747 2748 2749 2750
    /* VIR_NETWORK_UPDATE_AFFECT_CURRENT means "change LIVE if network
     * is active, else change CONFIG
    */
    isActive = virNetworkObjIsActive(network);
2751 2752
    if ((flags & (VIR_NETWORK_UPDATE_AFFECT_LIVE |
                  VIR_NETWORK_UPDATE_AFFECT_CONFIG)) ==
2753 2754 2755 2756 2757 2758 2759
        VIR_NETWORK_UPDATE_AFFECT_CURRENT) {
        if (isActive)
            flags |= VIR_NETWORK_UPDATE_AFFECT_LIVE;
        else
            flags |= VIR_NETWORK_UPDATE_AFFECT_CONFIG;
    }

2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785
    if (isActive && (flags & VIR_NETWORK_UPDATE_AFFECT_LIVE)) {
        /* Take care of anything that must be done before updating the
         * live NetworkDef.
         */
        if (network->def->forward.type == VIR_NETWORK_FORWARD_NONE ||
            network->def->forward.type == VIR_NETWORK_FORWARD_NAT ||
            network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE) {
            switch (section) {
            case VIR_NETWORK_SECTION_FORWARD:
            case VIR_NETWORK_SECTION_FORWARD_INTERFACE:
            case VIR_NETWORK_SECTION_IP:
            case VIR_NETWORK_SECTION_IP_DHCP_RANGE:
            case VIR_NETWORK_SECTION_IP_DHCP_HOST:
                /* these could affect the firewall rules, so remove the
                 * old rules (and remember to load new ones after the
                 * update).
                 */
                networkRemoveFirewallRules(network);
                needFirewallRefresh = true;
                break;
            default:
                break;
            }
        }
    }

2786
    /* update the network config in memory/on disk */
2787 2788 2789 2790 2791 2792 2793
    if (virNetworkObjUpdate(network, command, section, parentIndex, xml, flags) < 0) {
        if (needFirewallRefresh)
            ignore_value(networkAddFirewallRules(network));
        goto cleanup;
    }

    if (needFirewallRefresh && networkAddFirewallRules(network) < 0)
2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816
        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.
             */
2817
            if (networkRestartDhcpDaemon(driver, network) < 0)
2818 2819
                goto cleanup;

2820 2821 2822 2823 2824 2825 2826 2827
        } 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;

2828 2829 2830
            for (i = 0;
                 (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, i));
                 i++) {
2831 2832 2833 2834 2835 2836 2837
                if (ipdef->nranges || ipdef->nhosts) {
                    newDhcpActive = true;
                    break;
                }
            }

            if ((newDhcpActive != oldDhcpActive &&
2838 2839
                 networkRestartDhcpDaemon(driver, network) < 0) ||
                networkRefreshDhcpDaemon(driver, network) < 0) {
2840 2841 2842 2843
                goto cleanup;
            }

        } else if (section == VIR_NETWORK_SECTION_DNS_HOST ||
2844 2845 2846 2847 2848 2849
                   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.
             */
2850
            if (networkRefreshDhcpDaemon(driver, network) < 0)
2851 2852 2853 2854 2855 2856 2857 2858
                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.
             */
2859
            if (networkRefreshRadvd(driver, network) < 0)
2860 2861 2862 2863
                goto cleanup;
        }

        /* save current network state to disk */
2864 2865
        if ((ret = virNetworkSaveStatus(driverState->stateDir,
                                        network)) < 0) {
2866
            goto cleanup;
2867
        }
2868 2869 2870 2871 2872 2873 2874 2875 2876
    }
    ret = 0;
cleanup:
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
    return ret;
}

2877
static int networkCreate(virNetworkPtr net) {
2878
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
2879 2880
    virNetworkObjPtr network;
    int ret = -1;
2881
    virObjectEventPtr event = NULL;
2882

2883
    networkDriverLock(driver);
2884
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
2885

2886
    if (!network) {
2887 2888
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2889
        goto cleanup;
2890 2891
    }

2892 2893 2894
    if (virNetworkCreateEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2895
    ret = networkStartNetwork(driver, network);
2896

2897 2898
    event = virNetworkEventLifecycleNew(network->def->name,
                                        network->def->uuid,
2899 2900
                                        VIR_NETWORK_EVENT_STARTED,
                                        0);
2901

2902
cleanup:
2903 2904
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
2905 2906
    if (network)
        virNetworkObjUnlock(network);
2907
    networkDriverUnlock(driver);
2908
    return ret;
2909 2910 2911
}

static int networkDestroy(virNetworkPtr net) {
2912
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
2913 2914
    virNetworkObjPtr network;
    int ret = -1;
2915
    virObjectEventPtr event = NULL;
2916

2917
    networkDriverLock(driver);
2918
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
2919

2920
    if (!network) {
2921 2922
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2923
        goto cleanup;
2924 2925
    }

2926 2927 2928
    if (virNetworkDestroyEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

D
Daniel P. Berrange 已提交
2929
    if (!virNetworkObjIsActive(network)) {
2930 2931
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("network is not active"));
2932 2933 2934
        goto cleanup;
    }

2935 2936 2937
    if ((ret = networkShutdownNetwork(driver, network)) < 0)
        goto cleanup;

2938 2939
    event = virNetworkEventLifecycleNew(network->def->name,
                                        network->def->uuid,
2940 2941
                                        VIR_NETWORK_EVENT_STOPPED,
                                        0);
2942

2943
    if (!network->persistent) {
2944 2945 2946 2947 2948
        if (networkRemoveInactive(driver, network) < 0) {
            network = NULL;
            ret = -1;
            goto cleanup;
        }
2949 2950
        network = NULL;
    }
2951

2952
cleanup:
2953 2954
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
2955 2956
    if (network)
        virNetworkObjUnlock(network);
2957
    networkDriverUnlock(driver);
2958 2959 2960
    return ret;
}

2961
static char *networkGetXMLDesc(virNetworkPtr net,
2962
                               unsigned int flags)
2963
{
2964
    virNetworkObjPtr network;
2965
    virNetworkDefPtr def;
2966
    char *ret = NULL;
2967

2968
    virCheckFlags(VIR_NETWORK_XML_INACTIVE, NULL);
2969

2970 2971
    if (!(network = networkObjFromNetwork(net)))
        return ret;
2972

2973 2974 2975
    if (virNetworkGetXMLDescEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2976 2977 2978 2979 2980 2981
    if ((flags & VIR_NETWORK_XML_INACTIVE) && network->newDef)
        def = network->newDef;
    else
        def = network->def;

    ret = virNetworkDefFormat(def, flags);
2982 2983

cleanup:
2984 2985
    if (network)
        virNetworkObjUnlock(network);
2986
    return ret;
2987 2988 2989
}

static char *networkGetBridgeName(virNetworkPtr net) {
2990 2991 2992
    virNetworkObjPtr network;
    char *bridge = NULL;

2993 2994
    if (!(network = networkObjFromNetwork(net)))
        return bridge;
2995

2996 2997 2998
    if (virNetworkGetBridgeNameEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

2999
    if (!(network->def->bridge)) {
3000 3001 3002
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("network '%s' does not have a bridge name."),
                       network->def->name);
3003 3004 3005
        goto cleanup;
    }

3006
    ignore_value(VIR_STRDUP(bridge, network->def->bridge));
3007 3008

cleanup:
3009 3010
    if (network)
        virNetworkObjUnlock(network);
3011 3012 3013 3014 3015
    return bridge;
}

static int networkGetAutostart(virNetworkPtr net,
                             int *autostart) {
3016 3017
    virNetworkObjPtr network;
    int ret = -1;
3018

3019 3020
    if (!(network = networkObjFromNetwork(net)))
        return ret;
3021

3022 3023 3024
    if (virNetworkGetAutostartEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

3025
    *autostart = network->autostart;
3026
    ret = 0;
3027

3028
cleanup:
3029 3030
    if (network)
        virNetworkObjUnlock(network);
3031
    return ret;
3032 3033 3034
}

static int networkSetAutostart(virNetworkPtr net,
3035
                               int autostart) {
3036
    virNetworkDriverStatePtr driver = net->conn->networkPrivateData;
3037
    virNetworkObjPtr network;
3038
    char *configFile = NULL, *autostartLink = NULL;
3039
    int ret = -1;
3040

3041
    networkDriverLock(driver);
3042
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3043

3044
    if (!network) {
3045 3046
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
3047
        goto cleanup;
3048 3049
    }

3050 3051 3052
    if (virNetworkSetAutostartEnsureACL(net->conn, network->def) < 0)
        goto cleanup;

3053
    if (!network->persistent) {
3054 3055
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("cannot set autostart for transient network"));
3056 3057 3058
        goto cleanup;
    }

3059 3060
    autostart = (autostart != 0);

3061
    if (network->autostart != autostart) {
3062
        if ((configFile = virNetworkConfigFile(driver->networkConfigDir, network->def->name)) == NULL)
3063
            goto cleanup;
3064
        if ((autostartLink = virNetworkConfigFile(driver->networkAutostartDir, network->def->name)) == NULL)
3065 3066
            goto cleanup;

3067
        if (autostart) {
3068
            if (virFileMakePath(driver->networkAutostartDir) < 0) {
3069
                virReportSystemError(errno,
3070 3071
                                     _("cannot create autostart directory '%s'"),
                                     driver->networkAutostartDir);
3072 3073
                goto cleanup;
            }
3074

3075
            if (symlink(configFile, autostartLink) < 0) {
3076
                virReportSystemError(errno,
3077
                                     _("Failed to create symlink '%s' to '%s'"),
3078
                                     autostartLink, configFile);
3079 3080 3081
                goto cleanup;
            }
        } else {
3082
            if (unlink(autostartLink) < 0 && errno != ENOENT && errno != ENOTDIR) {
3083
                virReportSystemError(errno,
3084
                                     _("Failed to delete symlink '%s'"),
3085
                                     autostartLink);
3086 3087
                goto cleanup;
            }
3088 3089
        }

3090
        network->autostart = autostart;
3091
    }
3092
    ret = 0;
3093

3094
cleanup:
3095 3096
    VIR_FREE(configFile);
    VIR_FREE(autostartLink);
3097 3098
    if (network)
        virNetworkObjUnlock(network);
3099
    networkDriverUnlock(driver);
3100
    return ret;
3101 3102 3103 3104 3105
}


static virNetworkDriver networkDriver = {
    "Network",
3106 3107 3108 3109 3110 3111 3112
    .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 */
3113 3114
    .connectNetworkEventRegisterAny = networkConnectNetworkEventRegisterAny, /* 1.2.1 */
    .connectNetworkEventDeregisterAny = networkConnectNetworkEventDeregisterAny, /* 1.2.1 */
3115 3116
    .networkLookupByUUID = networkLookupByUUID, /* 0.2.0 */
    .networkLookupByName = networkLookupByName, /* 0.2.0 */
3117 3118
    .networkCreateXML = networkCreateXML, /* 0.2.0 */
    .networkDefineXML = networkDefineXML, /* 0.2.0 */
3119
    .networkUndefine = networkUndefine, /* 0.2.0 */
3120
    .networkUpdate = networkUpdate, /* 0.10.2 */
3121
    .networkCreate = networkCreate, /* 0.2.0 */
3122 3123 3124 3125 3126 3127 3128
    .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 */
3129 3130 3131
};

static virStateDriver networkStateDriver = {
3132
    .name = "Network",
3133
    .stateInitialize  = networkStateInitialize,
3134
    .stateAutoStart  = networkStateAutoStart,
3135 3136
    .stateCleanup = networkStateCleanup,
    .stateReload = networkStateReload,
3137 3138 3139
};

int networkRegister(void) {
3140 3141
    if (virRegisterNetworkDriver(&networkDriver) < 0)
        return -1;
3142 3143 3144
    virRegisterStateDriver(&networkStateDriver);
    return 0;
}
3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155

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

/* 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.
 */

3156 3157 3158 3159 3160 3161 3162
/* 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) {
3163
    size_t num_virt_fns = 0;
3164
    char **vfname = NULL;
3165
    virPCIDeviceAddressPtr *virt_fns;
3166 3167
    int ret = -1;
    size_t i;
3168

3169
    if ((virNetDevGetVirtualFunctions(netdef->forward.pfs->dev,
3170
                                      &vfname, &virt_fns, &num_virt_fns)) < 0) {
3171 3172
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Could not get Virtual functions on %s"),
3173
                       netdef->forward.pfs->dev);
3174 3175 3176 3177 3178 3179
        goto finish;
    }

    if (num_virt_fns == 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("No Vf's present on SRIOV PF %s"),
3180
                       netdef->forward.pfs->dev);
3181 3182 3183
       goto finish;
    }

3184
    if (VIR_ALLOC_N(netdef->forward.ifs, num_virt_fns) < 0)
3185 3186
        goto finish;

3187
    netdef->forward.nifs = num_virt_fns;
3188

3189
    for (i = 0; i < netdef->forward.nifs; i++) {
3190 3191 3192 3193
        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)) {
3194 3195 3196
            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)
3197
                    goto finish;
3198
            } else {
3199
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
3200
                               _("Direct mode types require interface names"));
3201 3202
                goto finish;
            }
3203
        }
3204
        else if (netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
3205
            /* VF's are always PCI devices */
3206 3207 3208 3209 3210
            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;
3211
        }
3212 3213 3214 3215
    }

    ret = 0;
finish:
3216 3217 3218
    for (i = 0; i < num_virt_fns; i++) {
        VIR_FREE(vfname[i]);
        VIR_FREE(virt_fns[i]);
3219
    }
3220
    VIR_FREE(vfname);
3221
    VIR_FREE(virt_fns);
3222 3223 3224
    return ret;
}

3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238
/* 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)
{
3239
    virNetworkDriverStatePtr driver = driverState;
3240 3241 3242
    enum virDomainNetType actualType = iface->type;
    virNetworkObjPtr network = NULL;
    virNetworkDefPtr netdef = NULL;
3243
    virNetDevBandwidthPtr bandwidth = NULL;
3244 3245 3246
    virPortGroupDefPtr portgroup = NULL;
    virNetDevVPortProfilePtr virtport = iface->virtPortProfile;
    virNetDevVlanPtr vlan = NULL;
3247
    virNetworkForwardIfDefPtr dev = NULL;
3248
    size_t i;
3249 3250 3251
    int ret = -1;

    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
3252
        goto validate;
3253 3254 3255 3256 3257 3258 3259 3260

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

    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
    networkDriverUnlock(driver);
    if (!network) {
3261 3262 3263
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
3264
        goto error;
3265 3266
    }
    netdef = network->def;
3267

3268 3269 3270
    if (VIR_ALLOC(iface->data.network.actual) < 0)
        goto error;

3271 3272 3273 3274 3275 3276 3277 3278 3279 3280
    /* 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.
     */
3281 3282 3283 3284 3285 3286

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

3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308
    /* For bridgeless networks we must take into account any QoS set to them.
     * This is, however, a bit tricky. With bridged network there are two
     * points where QoS can be set: domain's interface and the bridge. The
     * limits on those differ (in general) in which case the lower limit gets
     * into effect. For example, if domain's interface average is 10Mbps but
     * the network's is just 1Mbps, the domain will not be able to send data
     * any faster than 1Mbps. However, on bridgeless networks we can't just set
     * those two points and let kernel do its job since we have only single
     * point. Therefore, we must combine the QoS settings from both domain's
     * interface and network and choose the minimal value from pairs.
     */
    if (netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE ||
        netdef->forward.type == VIR_NETWORK_FORWARD_VEPA ||
        netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH ||
        netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
        if (virNetDevBandwidthMinimal(&iface->data.network.actual->bandwidth,
                                      bandwidth,
                                      netdef->bandwidth) < 0)
            goto error;
    } else if (bandwidth &&
               virNetDevBandwidthCopy(&iface->data.network.actual->bandwidth,
                                      bandwidth) < 0) {
3309
        goto error;
3310
    }
3311

3312 3313 3314 3315 3316 3317 3318 3319
    /* 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;

3320 3321
    if (vlan && virNetDevVlanCopy(&iface->data.network.actual->vlan, vlan) < 0)
        goto error;
3322

3323 3324 3325
    if ((netdef->forward.type == VIR_NETWORK_FORWARD_NONE) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_NAT) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE)) {
3326 3327 3328 3329
        /* for these forward types, the actual net type really *is*
         *NETWORK; we just keep the info from the portgroup in
         * iface->data.network.actual
        */
3330
        iface->data.network.actual->type = VIR_DOMAIN_NET_TYPE_NETWORK;
3331 3332 3333 3334

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

3335
    } else if ((netdef->forward.type == VIR_NETWORK_FORWARD_BRIDGE) &&
3336
               netdef->bridge) {
3337 3338 3339 3340 3341

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

3342
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_BRIDGE;
3343 3344
        if (VIR_STRDUP(iface->data.network.actual->data.bridge.brname,
                       netdef->bridge) < 0)
3345
            goto error;
3346

3347 3348 3349 3350 3351 3352 3353 3354
        /* 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) {
3355
            goto error;
3356 3357 3358 3359 3360 3361 3362 3363 3364 3365
        }
        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);
3366
                goto error;
3367 3368 3369
            }
        }

3370
    } else if (netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
3371

3372 3373
        virDomainHostdevSubsysPciBackendType backend;

3374
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_HOSTDEV;
3375
        if (netdef->forward.npfs > 0 && netdef->forward.nifs <= 0 &&
3376 3377 3378 3379 3380
            networkCreateInterfacePool(netdef) < 0) {
            goto error;
        }

        /* pick first dev with 0 connections */
3381 3382 3383
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].connections == 0) {
                dev = &netdef->forward.ifs[i];
3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397
                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;
3398
        iface->data.network.actual->data.hostdev.def.managed = netdef->forward.managed ? 1 : 0;
3399
        iface->data.network.actual->data.hostdev.def.source.subsys.type = dev->type;
3400
        iface->data.network.actual->data.hostdev.def.source.subsys.u.pci.addr = dev->device.pci;
3401

3402 3403 3404
        switch (netdef->forward.driverName)
        {
        case VIR_NETWORK_FORWARD_DRIVER_NAME_DEFAULT:
3405
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_DEFAULT;
3406 3407
            break;
        case VIR_NETWORK_FORWARD_DRIVER_NAME_KVM:
3408
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_KVM;
3409 3410
            break;
        case VIR_NETWORK_FORWARD_DRIVER_NAME_VFIO:
3411
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_VFIO;
3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422
            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;

3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447
        /* 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;
            }
        }

3448 3449 3450 3451
    } 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)) {
3452 3453 3454 3455 3456 3457

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

        /* Set type=direct and appropriate <source mode='xxx'/> */
3458
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_DIRECT;
3459
        switch (netdef->forward.type) {
3460
        case VIR_NETWORK_FORWARD_BRIDGE:
3461
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_BRIDGE;
3462 3463
            break;
        case VIR_NETWORK_FORWARD_PRIVATE:
3464
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PRIVATE;
3465 3466
            break;
        case VIR_NETWORK_FORWARD_VEPA:
3467
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_VEPA;
3468 3469
            break;
        case VIR_NETWORK_FORWARD_PASSTHROUGH:
3470
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PASSTHRU;
3471 3472 3473
            break;
        }

3474 3475 3476 3477 3478 3479 3480 3481
        /* 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) {
3482
            goto error;
3483
        }
3484
        virtport = iface->data.network.actual->virtPortProfile;
3485
        if (virtport) {
3486 3487 3488 3489 3490 3491 3492 3493
            /* 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);
3494
                goto error;
3495 3496
            }
        }
3497

3498 3499 3500
        /* 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).
         */
3501
        if ((netdef->forward.nifs <= 0) && (netdef->forward.npfs <= 0)) {
3502 3503 3504 3505
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' uses a direct mode, but "
                             "has no forward dev and no interface pool"),
                           netdef->name);
3506
            goto error;
3507 3508 3509
        } else {
            /* pick an interface from the pool */

3510
            if (netdef->forward.npfs > 0 && netdef->forward.nifs == 0 &&
3511 3512 3513 3514
                networkCreateInterfacePool(netdef) < 0) {
                goto error;
            }

3515 3516 3517 3518 3519
            /* 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.
3520
             */
3521 3522
            if ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
                ((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
3523 3524 3525
                 iface->data.network.actual->virtPortProfile &&
                 (iface->data.network.actual->virtPortProfile->virtPortType
                  == VIR_NETDEV_VPORT_PROFILE_8021QBH))) {
3526

3527
                /* pick first dev with 0 connections */
3528 3529 3530
                for (i = 0; i < netdef->forward.nifs; i++) {
                    if (netdef->forward.ifs[i].connections == 0) {
                        dev = &netdef->forward.ifs[i];
3531 3532 3533 3534 3535
                        break;
                    }
                }
            } else {
                /* pick least used dev */
3536
                dev = &netdef->forward.ifs[0];
3537 3538 3539
                for (i = 1; i < netdef->forward.nifs; i++) {
                    if (netdef->forward.ifs[i].connections < dev->connections)
                        dev = &netdef->forward.ifs[i];
3540 3541 3542 3543
                }
            }
            /* dev points at the physical device we want to use */
            if (!dev) {
3544 3545 3546 3547
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("network '%s' requires exclusive access "
                                 "to interfaces, but none are available"),
                               netdef->name);
3548
                goto error;
3549
            }
3550 3551
            if (VIR_STRDUP(iface->data.network.actual->data.direct.linkdev,
                           dev->device.dev) < 0)
3552
                goto error;
3553 3554 3555
        }
    }

3556
    if (virNetDevVPortProfileCheckComplete(virtport, true) < 0)
3557
        goto error;
3558

3559 3560 3561 3562 3563 3564
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.
     */

3565
    if (virDomainNetGetActualVlan(iface)) {
3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590
        /* 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;
        }
    }

3591 3592
    if (dev) {
        /* we are now assured of success, so mark the allocation */
3593
        dev->connections++;
3594 3595 3596 3597 3598 3599 3600 3601 3602
        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);
        }
3603
    }
3604

3605 3606 3607 3608 3609
    if (netdef) {
        netdef->connections++;
        VIR_DEBUG("Using network %s, %d connections",
                  netdef->name, netdef->connections);
    }
3610
    ret = 0;
3611

3612 3613 3614
cleanup:
    if (network)
        virNetworkObjUnlock(network);
3615 3616 3617 3618
    return ret;

error:
    if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
3619 3620 3621
        virDomainActualNetDefFree(iface->data.network.actual);
        iface->data.network.actual = NULL;
    }
3622
    goto cleanup;
3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637
}

/* 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)
{
3638
    virNetworkDriverStatePtr driver = driverState;
3639
    enum virDomainNetType actualType = virDomainNetGetActualType(iface);
3640 3641
    virNetworkObjPtr network;
    virNetworkDefPtr netdef;
3642
    virNetworkForwardIfDefPtr dev = NULL;
3643 3644
    size_t i;
    int ret = -1;
3645 3646 3647 3648 3649 3650 3651 3652

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

    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
    networkDriverUnlock(driver);
    if (!network) {
3653 3654 3655
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
3656 3657 3658 3659 3660
        goto error;
    }
    netdef = network->def;

    if (!iface->data.network.actual ||
3661 3662
        (actualType != VIR_DOMAIN_NET_TYPE_DIRECT &&
         actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV)) {
3663 3664
        VIR_DEBUG("Nothing to claim from network %s", iface->data.network.name);
        goto success;
3665 3666
    }

3667
    if (netdef->forward.npfs > 0 && netdef->forward.nifs == 0 &&
3668
        networkCreateInterfacePool(netdef) < 0) {
3669
        goto error;
3670
    }
3671
    if (netdef->forward.nifs == 0) {
3672
        virReportError(VIR_ERR_INTERNAL_ERROR,
3673 3674
                       _("network '%s' uses a direct or hostdev mode, "
                         "but has no forward dev and no interface pool"),
3675
                       netdef->name);
3676
        goto error;
3677
    }
3678

3679 3680
    if (actualType == VIR_DOMAIN_NET_TYPE_DIRECT) {
        const char *actualDev;
3681

3682 3683 3684 3685 3686 3687 3688 3689 3690
        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 */
3691 3692
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
3693
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
3694 3695
                STREQ(actualDev, netdef->forward.ifs[i].device.dev)) {
                dev = &netdef->forward.ifs[i];
3696 3697 3698 3699 3700
                break;
            }
        }
        /* dev points at the physical device we want to use */
        if (!dev) {
3701
            virReportError(VIR_ERR_INTERNAL_ERROR,
3702 3703
                           _("network '%s' doesn't have dev='%s' "
                             "in use by domain"),
3704
                           netdef->name, actualDev);
3705
            goto error;
3706 3707
        }

3708
        /* PASSTHROUGH mode and PRIVATE Mode + 802.1Qbh both require
3709 3710
         * exclusive access to a device, so current connections count
         * must be 0 in those cases.
3711
         */
3712
        if ((dev->connections > 0) &&
3713 3714
            ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
             ((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
3715 3716
              iface->data.network.actual->virtPortProfile &&
              (iface->data.network.actual->virtPortProfile->virtPortType
3717
               == VIR_NETDEV_VPORT_PROFILE_8021QBH)))) {
3718
            virReportError(VIR_ERR_INTERNAL_ERROR,
3719 3720
                           _("network '%s' claims dev='%s' is already in "
                             "use by a different domain"),
3721
                           netdef->name, actualDev);
3722
            goto error;
3723
        }
3724

3725
        /* we are now assured of success, so mark the allocation */
3726
        dev->connections++;
3727
        VIR_DEBUG("Using physical device %s, connections %d",
3728
                  dev->device.dev, dev->connections);
3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741

    }  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 */
3742 3743
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
3744
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
3745
                virDevicePCIAddressEqual(&hostdev->source.subsys.u.pci.addr,
3746 3747
                                         &netdef->forward.ifs[i].device.pci)) {
                dev = &netdef->forward.ifs[i];
3748 3749 3750 3751 3752 3753 3754 3755 3756
                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,
3757 3758 3759 3760
                           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);
3761 3762 3763 3764 3765 3766 3767 3768
                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) &&
3769
            netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785
            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);
3786 3787
    }

3788
success:
3789 3790 3791
    netdef->connections++;
    VIR_DEBUG("Using network %s, %d connections",
              netdef->name, netdef->connections);
3792 3793 3794 3795 3796
    ret = 0;
cleanup:
    if (network)
        virNetworkObjUnlock(network);
    return ret;
3797 3798 3799

error:
    goto cleanup;
3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815
}


/* 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)
{
3816
    virNetworkDriverStatePtr driver = driverState;
3817
    enum virDomainNetType actualType = virDomainNetGetActualType(iface);
3818
    virNetworkObjPtr network;
3819
    virNetworkDefPtr netdef;
3820
    virNetworkForwardIfDefPtr dev = NULL;
3821 3822
    size_t i;
    int ret = -1;
3823 3824 3825 3826 3827 3828 3829 3830

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

    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
    networkDriverUnlock(driver);
    if (!network) {
3831 3832 3833
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
3834 3835 3836 3837
        goto error;
    }
    netdef = network->def;

3838 3839 3840 3841 3842 3843
    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;

3844 3845 3846
    if ((!iface->data.network.actual) ||
        ((actualType != VIR_DOMAIN_NET_TYPE_DIRECT) &&
         (actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV))) {
3847 3848
        VIR_DEBUG("Nothing to release to network %s", iface->data.network.name);
        goto success;
3849 3850
    }

3851
    if (netdef->forward.nifs == 0) {
3852
        virReportError(VIR_ERR_INTERNAL_ERROR,
3853
                       _("network '%s' uses a direct/hostdev mode, but "
3854 3855
                         "has no forward dev and no interface pool"),
                       netdef->name);
3856
        goto error;
3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868
    }

    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;
        }
3869

3870 3871
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
3872
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
3873 3874
                STREQ(actualDev, netdef->forward.ifs[i].device.dev)) {
                dev = &netdef->forward.ifs[i];
3875 3876 3877
                break;
            }
        }
3878

3879
        if (!dev) {
3880
            virReportError(VIR_ERR_INTERNAL_ERROR,
3881 3882
                           _("network '%s' doesn't have dev='%s' "
                             "in use by domain"),
3883
                           netdef->name, actualDev);
3884
            goto error;
3885 3886
        }

3887
        dev->connections--;
3888
        VIR_DEBUG("Releasing physical device %s, connections %d",
3889
                  dev->device.dev, dev->connections);
3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900

    } 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;
        }

3901 3902
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
3903
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
3904
                virDevicePCIAddressEqual(&hostdev->source.subsys.u.pci.addr,
3905 3906
                                         &netdef->forward.ifs[i].device.pci)) {
                dev = &netdef->forward.ifs[i];
3907 3908 3909 3910 3911 3912 3913 3914 3915
                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,
3916 3917 3918 3919
                           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);
3920 3921 3922 3923 3924 3925 3926 3927 3928
                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);
   }
3929

3930
success:
3931 3932
    if (iface->data.network.actual)
        netdef->connections--;
3933 3934
    VIR_DEBUG("Releasing network %s, %d connections",
              netdef->name, netdef->connections);
3935 3936 3937 3938
    ret = 0;
cleanup:
    if (network)
        virNetworkObjUnlock(network);
3939 3940 3941 3942
    if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
        virDomainActualNetDefFree(iface->data.network.actual);
        iface->data.network.actual = NULL;
    }
3943
    return ret;
3944 3945 3946

error:
    goto cleanup;
3947
}
3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971

/*
 * 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;
3972
    virNetworkDriverStatePtr driver = driverState;
3973
    virNetworkObjPtr network;
3974 3975 3976 3977
    virNetworkDefPtr netdef;
    virNetworkIpDefPtr ipdef;
    virSocketAddr addr;
    virSocketAddrPtr addrptr = NULL;
3978
    char *dev_name = NULL;
3979 3980 3981 3982 3983 3984

    *netaddr = NULL;
    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, netname);
    networkDriverUnlock(driver);
    if (!network) {
3985 3986 3987
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       netname);
3988
        goto error;
3989 3990 3991
    }
    netdef = network->def;

3992
    switch (netdef->forward.type) {
3993 3994 3995 3996 3997 3998
    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) {
3999 4000 4001
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' doesn't have an IPv4 address"),
                           netdef->name);
4002 4003 4004 4005 4006 4007
            break;
        }
        addrptr = &ipdef->address;
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
4008
        if ((dev_name = netdef->bridge))
4009 4010 4011 4012 4013 4014 4015 4016
            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:
4017 4018
        if ((netdef->forward.nifs > 0) && netdef->forward.ifs)
            dev_name = netdef->forward.ifs[0].device.dev;
4019

4020
        if (!dev_name) {
4021 4022 4023
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' has no associated interface or bridge"),
                           netdef->name);
4024 4025 4026 4027
        }
        break;
    }

4028
    if (dev_name) {
4029
        if (virNetDevGetIPv4Address(dev_name, &addr) < 0)
4030
            goto error;
4031
        addrptr = &addr;
4032 4033
    }

4034 4035 4036
    if (!(addrptr &&
          (*netaddr = virSocketAddrFormat(addrptr)))) {
        goto error;
4037 4038
    }

4039
    ret = 0;
4040 4041 4042 4043
cleanup:
    if (network)
        virNetworkObjUnlock(network);
    return ret;
4044 4045 4046

error:
    goto cleanup;
4047
}
4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065

/**
 * 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;
4066
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
4067 4068 4069 4070
    unsigned long long tmp_floor_sum = net->floor_sum;
    unsigned long long tmp_new_rate = 0;
    char ifmac[VIR_MAC_STRING_BUFLEN];

4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081
    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;
    }

4082
    if (!ifaceBand || !ifaceBand->in || !ifaceBand->in->floor ||
4083 4084
        !netBand || !netBand->in) {
        /* no QoS required, claim success */
4085
        return 1;
4086
    }
4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155

    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];
4156
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184

    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;
    }

4185 4186
    plug_ret = virNetDevBandwidthPlug(net->def->bridge, net->def->bandwidth,
                                      &iface->mac, ifaceBand, class_id);
4187 4188 4189 4190 4191 4192 4193 4194
    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 */
4195
    net->floor_sum += ifaceBand->in->floor;
4196
    /* update status file */
4197
    if (virNetworkSaveStatus(driverState->stateDir, net) < 0) {
4198
        ignore_value(virBitmapClearBit(net->class_id, class_id));
4199
        net->floor_sum -= ifaceBand->in->floor;
4200 4201 4202 4203
        iface->data.network.actual->class_id = 0;
        ignore_value(virNetDevBandwidthUnplug(net->def->bridge, class_id));
        goto cleanup;
    }
4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222
    /* 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;
4223
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
4224 4225 4226

    if (iface->data.network.actual &&
        iface->data.network.actual->class_id) {
4227 4228 4229 4230 4231
        if (!net->def->bandwidth || !net->def->bandwidth->in) {
            VIR_WARN("Network %s has no bandwidth but unplug requested",
                     net->def->name);
            goto cleanup;
        }
4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242
        /* 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 */
4243
        net->floor_sum -= ifaceBand->in->floor;
4244 4245 4246 4247
        /* return class ID */
        ignore_value(virBitmapClearBit(net->class_id,
                                       iface->data.network.actual->class_id));
        /* update status file */
4248
        if (virNetworkSaveStatus(driverState->stateDir, net) < 0) {
4249
            net->floor_sum += ifaceBand->in->floor;
4250 4251 4252 4253
            ignore_value(virBitmapSetBit(net->class_id,
                                         iface->data.network.actual->class_id));
            goto cleanup;
        }
4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266
        /* 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;
}