bridge_driver.c 156.5 KB
Newer Older
G
Gene Czarcinski 已提交
1

2
/*
3
 * bridge_driver.c: core driver methods for managing network
4
 *
5
 * Copyright (C) 2006-2012 Red Hat, Inc.
6 7 8 9 10 11 12 13 14 15 16 17 18
 * 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
19
 * License along with this library.  If not, see
O
Osier Yang 已提交
20
 * <http://www.gnu.org/licenses/>.
21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
 *
 * 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 <stdio.h>
#include <sys/wait.h>
#include <sys/ioctl.h>
45
#include <net/if.h>
46

47
#include "virterror_internal.h"
48
#include "datatypes.h"
49
#include "bridge_driver.h"
50
#include "network_conf.h"
51
#include "device_conf.h"
52
#include "driver.h"
53
#include "virbuffer.h"
54
#include "virpidfile.h"
55
#include "util.h"
56
#include "vircommand.h"
57 58
#include "memory.h"
#include "uuid.h"
59
#include "viriptables.h"
60
#include "logging.h"
61
#include "virdnsmasq.h"
62
#include "configmake.h"
63
#include "virnetdev.h"
64
#include "pci.h"
65 66
#include "virnetdevbridge.h"
#include "virnetdevtap.h"
67
#include "virnetdevvportprofile.h"
68
#include "virdbus.h"
69
#include "virfile.h"
70

71 72
#define NETWORK_PID_DIR LOCALSTATEDIR "/run/libvirt/network"
#define NETWORK_STATE_DIR LOCALSTATEDIR "/lib/libvirt/network"
73

74
#define DNSMASQ_STATE_DIR LOCALSTATEDIR "/lib/libvirt/dnsmasq"
75
#define RADVD_STATE_DIR LOCALSTATEDIR "/lib/libvirt/radvd"
76

77 78
#define VIR_FROM_THIS VIR_FROM_NETWORK

79 80
/* Main driver state */
struct network_driver {
81
    virMutex lock;
82

83
    virNetworkObjList networks;
84 85 86 87 88

    iptablesContext *iptables;
    char *networkConfigDir;
    char *networkAutostartDir;
    char *logDir;
89
    dnsmasqCapsPtr dnsmasqCaps;
90 91
};

92 93 94

static void networkDriverLock(struct network_driver *driver)
{
95
    virMutexLock(&driver->lock);
96 97 98
}
static void networkDriverUnlock(struct network_driver *driver)
{
99
    virMutexUnlock(&driver->lock);
100 101
}

102 103
static int networkShutdown(void);

104 105 106 107 108 109 110
static int networkStartNetwork(struct network_driver *driver,
                               virNetworkObjPtr network);

static int networkShutdownNetwork(struct network_driver *driver,
                                  virNetworkObjPtr network);

static int networkStartNetworkVirtual(struct network_driver *driver,
111
                                     virNetworkObjPtr network);
112

113 114 115 116 117 118 119
static int networkShutdownNetworkVirtual(struct network_driver *driver,
                                        virNetworkObjPtr network);

static int networkStartNetworkExternal(struct network_driver *driver,
                                     virNetworkObjPtr network);

static int networkShutdownNetworkExternal(struct network_driver *driver,
120
                                        virNetworkObjPtr network);
121

122
static void networkReloadIptablesRules(struct network_driver *driver);
123
static void networkRefreshDaemons(struct network_driver *driver);
124

125 126 127 128 129
static int networkPlugBandwidth(virNetworkObjPtr net,
                                virDomainNetDefPtr iface);
static int networkUnplugBandwidth(virNetworkObjPtr net,
                                  virDomainNetDefPtr iface);

130 131
static struct network_driver *driverState = NULL;

132
static char *
133
networkDnsmasqLeaseFileNameDefault(const char *netname)
134 135 136
{
    char *leasefile;

137 138
    ignore_value(virAsprintf(&leasefile, DNSMASQ_STATE_DIR "/%s.leases",
                             netname));
139 140 141
    return leasefile;
}

142 143 144
networkDnsmasqLeaseFileNameFunc networkDnsmasqLeaseFileName =
    networkDnsmasqLeaseFileNameDefault;

145 146 147 148 149 150 151 152 153 154
static char *
networkDnsmasqConfigFileName(const char *netname)
{
    char *conffile;

    ignore_value(virAsprintf(&conffile, DNSMASQ_STATE_DIR "/%s.conf",
                             netname));
    return conffile;
}

155 156 157 158 159 160
static char *
networkRadvdPidfileBasename(const char *netname)
{
    /* this is simple but we want to be sure it's consistently done */
    char *pidfilebase;

161
    ignore_value(virAsprintf(&pidfilebase, "%s-radvd", netname));
162 163 164 165 166 167 168 169
    return pidfilebase;
}

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

170 171
    ignore_value(virAsprintf(&configfile, RADVD_STATE_DIR "/%s-radvd.conf",
                             netname));
172 173
    return configfile;
}
174

175 176 177 178 179 180 181
/* do needed cleanup steps and remove the network from the list */
static int
networkRemoveInactive(struct network_driver *driver,
                      virNetworkObjPtr net)
{
    char *leasefile = NULL;
    char *radvdconfigfile = NULL;
182
    char *configfile = NULL;
183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
    char *radvdpidbase = NULL;
    dnsmasqContext *dctx = NULL;
    virNetworkDefPtr def = virNetworkObjGetPersistentDef(net);

    int ret = -1;

    /* remove the (possibly) existing dnsmasq and radvd files */
    if (!(dctx = dnsmasqContextNew(def->name, DNSMASQ_STATE_DIR)))
        goto cleanup;

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

    if (!(radvdconfigfile = networkRadvdConfigFileName(def->name)))
        goto no_memory;

    if (!(radvdpidbase = networkRadvdPidfileBasename(def->name)))
        goto no_memory;

202 203 204
    if (!(configfile = networkDnsmasqConfigFileName(def->name)))
        goto no_memory;

205 206 207
    /* dnsmasq */
    dnsmasqDelete(dctx);
    unlink(leasefile);
208
    unlink(configfile);
209 210 211 212 213 214 215 216 217 218 219 220

    /* radvd */
    unlink(radvdconfigfile);
    virPidFileDelete(NETWORK_PID_DIR, radvdpidbase);

    /* remove the network definition */
    virNetworkRemoveInactive(&driver->networks, net);

    ret = 0;

cleanup:
    VIR_FREE(leasefile);
221
    VIR_FREE(configfile);
222 223 224 225 226 227 228 229 230 231
    VIR_FREE(radvdconfigfile);
    VIR_FREE(radvdpidbase);
    dnsmasqContextFree(dctx);
    return ret;

no_memory:
    virReportOOMError();
    goto cleanup;
}

232 233 234
static char *
networkBridgeDummyNicName(const char *brname)
{
235
    static const char dummyNicSuffix[] = "-nic";
236 237
    char *nicname;

238 239 240 241 242 243 244
    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.
         */
245 246 247 248 249
        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));
250
    } else {
251
        ignore_value(virAsprintf(&nicname, "%s%s", brname, dummyNicSuffix));
252
    }
253 254 255
    return nicname;
}

256 257 258 259 260 261 262 263 264 265
static void
networkFindActiveConfigs(struct network_driver *driver) {
    unsigned int i;

    for (i = 0 ; i < driver->networks.count ; i++) {
        virNetworkObjPtr obj = driver->networks.objs[i];
        char *config;

        virNetworkObjLock(obj);

266
        if ((config = virNetworkConfigFile(NETWORK_STATE_DIR,
267 268 269 270 271 272 273 274 275 276 277 278
                                           obj->def->name)) == NULL) {
            virNetworkObjUnlock(obj);
            continue;
        }

        if (access(config, R_OK) < 0) {
            VIR_FREE(config);
            virNetworkObjUnlock(obj);
            continue;
        }

        /* Try and load the live config */
279 280 281
        if (virNetworkObjUpdateParseFile(config, obj) < 0)
            VIR_WARN("Unable to update config of '%s' network",
                     obj->def->name);
282 283 284 285
        VIR_FREE(config);

        /* If bridge exists, then mark it active */
        if (obj->def->bridge &&
H
Hu Tao 已提交
286
            virNetDevExists(obj->def->bridge) == 1) {
287 288
            obj->active = 1;

289 290
            /* Try and read dnsmasq/radvd pids if any */
            if (obj->def->ips && (obj->def->nips > 0)) {
291 292 293
                char *radvdpidbase;

                ignore_value(virPidFileReadIfAlive(NETWORK_PID_DIR, obj->def->name,
294 295
                                                   &obj->dnsmasqPid,
                                                   dnsmasqCapsGetBinaryPath(driver->dnsmasqCaps)));
296

297
                if (!(radvdpidbase = networkRadvdPidfileBasename(obj->def->name))) {
298
                    virReportOOMError();
299 300
                    goto cleanup;
                }
301 302
                ignore_value(virPidFileReadIfAlive(NETWORK_PID_DIR, radvdpidbase,
                                                   &obj->radvdPid, RADVD));
303
                VIR_FREE(radvdpidbase);
304 305 306
            }
        }

307
    cleanup:
308 309 310 311 312
        virNetworkObjUnlock(obj);
    }
}


313 314 315
static void
networkAutostartConfigs(struct network_driver *driver) {
    unsigned int i;
316

317
    for (i = 0 ; i < driver->networks.count ; i++) {
318
        virNetworkObjLock(driver->networks.objs[i]);
319
        if (driver->networks.objs[i]->autostart &&
320 321
            !virNetworkObjIsActive(driver->networks.objs[i])) {
            if (networkStartNetwork(driver, driver->networks.objs[i]) < 0) {
322
            /* failed to start but already logged */
323
            }
324
        }
325
        virNetworkObjUnlock(driver->networks.objs[i]);
326 327 328
    }
}

329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347
#if HAVE_FIREWALLD
static DBusHandlerResult
firewalld_dbus_filter_bridge(DBusConnection *connection ATTRIBUTE_UNUSED,
                             DBusMessage *message, void *user_data) {
    struct network_driver *_driverState = user_data;

    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.");
        networkReloadIptablesRules(_driverState);
    }

    return DBUS_HANDLER_RESULT_NOT_YET_HANDLED;
}
#endif

348 349 350 351 352 353
/**
 * networkStartup:
 *
 * Initialization function for the QEmu daemon
 */
static int
354 355 356 357
networkStartup(bool privileged,
               virStateInhibitCallback callback ATTRIBUTE_UNUSED,
               void *opaque ATTRIBUTE_UNUSED)
{
358
    char *base = NULL;
359 360 361
#ifdef HAVE_FIREWALLD
    DBusConnection *sysbus = NULL;
#endif
362 363

    if (VIR_ALLOC(driverState) < 0)
364
        goto error;
365

366 367 368 369
    if (virMutexInit(&driverState->lock) < 0) {
        VIR_FREE(driverState);
        goto error;
    }
370 371
    networkDriverLock(driverState);

372
    if (privileged) {
373
        if (virAsprintf(&driverState->logDir,
374
                        "%s/log/libvirt/qemu", LOCALSTATEDIR) == -1)
375 376
            goto out_of_memory;

377
        if ((base = strdup(SYSCONFDIR "/libvirt")) == NULL)
378 379
            goto out_of_memory;
    } else {
380
        char *userdir = virGetUserCacheDirectory();
381 382 383

        if (!userdir)
            goto error;
384

385
        if (virAsprintf(&driverState->logDir,
386
                        "%s/qemu/log", userdir) == -1) {
387
            VIR_FREE(userdir);
388
            goto out_of_memory;
389
        }
390
        VIR_FREE(userdir);
391

392
        userdir = virGetUserConfigDirectory();
393
        if (virAsprintf(&base, "%s", userdir) == -1) {
394
            VIR_FREE(userdir);
395 396
            goto out_of_memory;
        }
397
        VIR_FREE(userdir);
398 399 400 401 402
    }

    /* Configuration paths are either ~/.libvirt/qemu/... (session) or
     * /etc/libvirt/qemu/... (system).
     */
403
    if (virAsprintf(&driverState->networkConfigDir, "%s/qemu/networks", base) == -1)
404 405
        goto out_of_memory;

406 407
    if (virAsprintf(&driverState->networkAutostartDir, "%s/qemu/networks/autostart",
                    base) == -1)
408 409 410 411
        goto out_of_memory;

    VIR_FREE(base);

412
    if (!(driverState->iptables = iptablesContextNew())) {
413
        goto out_of_memory;
414 415
    }

416 417
    /* if this fails now, it will be retried later with dnsmasqCapsRefresh() */
    driverState->dnsmasqCaps = dnsmasqCapsNewFromBinary(DNSMASQ);
418

419
    if (virNetworkLoadAllConfigs(&driverState->networks,
420
                                 driverState->networkConfigDir,
421 422 423
                                 driverState->networkAutostartDir) < 0)
        goto error;

424
    networkFindActiveConfigs(driverState);
425
    networkReloadIptablesRules(driverState);
426
    networkRefreshDaemons(driverState);
427 428
    networkAutostartConfigs(driverState);

429 430
    networkDriverUnlock(driverState);

431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456
#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

457 458
    return 0;

459
out_of_memory:
460
    virReportOOMError();
461 462

error:
463 464 465
    if (driverState)
        networkDriverUnlock(driverState);

466
    VIR_FREE(base);
467
    networkShutdown();
468 469 470 471 472 473 474 475 476 477 478
    return -1;
}

/**
 * networkReload:
 *
 * Function to restart the QEmu daemon, it will recheck the configuration
 * files and update its state and the networking
 */
static int
networkReload(void) {
479 480 481
    if (!driverState)
        return 0;

482
    networkDriverLock(driverState);
483
    virNetworkLoadAllConfigs(&driverState->networks,
484 485
                             driverState->networkConfigDir,
                             driverState->networkAutostartDir);
486
    networkReloadIptablesRules(driverState);
487
    networkRefreshDaemons(driverState);
488
    networkAutostartConfigs(driverState);
489
    networkDriverUnlock(driverState);
490 491 492 493 494 495 496 497 498 499 500 501 502 503
    return 0;
}


/**
 * networkShutdown:
 *
 * Shutdown the QEmu daemon, it will stop all active domains and networks
 */
static int
networkShutdown(void) {
    if (!driverState)
        return -1;

504 505
    networkDriverLock(driverState);

506
    /* free inactive networks */
507
    virNetworkObjListFree(&driverState->networks);
508 509 510 511 512 513 514 515

    VIR_FREE(driverState->logDir);
    VIR_FREE(driverState->networkConfigDir);
    VIR_FREE(driverState->networkAutostartDir);

    if (driverState->iptables)
        iptablesContextFree(driverState->iptables);

516 517
    virObjectUnref(driverState->dnsmasqCaps);

518
    networkDriverUnlock(driverState);
519
    virMutexDestroy(&driverState->lock);
520

521 522 523 524 525 526
    VIR_FREE(driverState);

    return 0;
}


527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
/* 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)
{
    int ii, ret = -1;
    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.
     */
    for (ii = 0; ii < 25; ii++) {
        int signum = 0;
        if (ii == 0)
            signum = SIGTERM;
        else if (ii == 15) {
            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
586 587 588 589
    /* the following does not build a file, it builds a list
     * which is later saved into a file
     */

590
static int
G
Gene Czarcinski 已提交
591 592
networkBuildDnsmasqDhcpHostsList(dnsmasqContext *dctx,
                                 virNetworkIpDefPtr ipdef)
593
{
G
Gene Czarcinski 已提交
594 595
    unsigned int i;
    bool ipv6 = false;
596

G
Gene Czarcinski 已提交
597 598
    if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
        ipv6 = true;
599 600
    for (i = 0; i < ipdef->nhosts; i++) {
        virNetworkDHCPHostDefPtr host = &(ipdef->hosts[i]);
G
Gene Czarcinski 已提交
601 602
        if (VIR_SOCKET_ADDR_VALID(&host->ip))
            if (dnsmasqAddDhcpHost(dctx, host->mac, &host->ip, host->name, ipv6) < 0)
603
                return -1;
604
    }
605

G
Gene Czarcinski 已提交
606 607 608 609 610 611 612 613 614
    return 0;
}

static int
networkBuildDnsmasqHostsList(dnsmasqContext *dctx,
                             virNetworkDNSDefPtr dnsdef)
{
    unsigned int i, j;

615 616
    if (dnsdef) {
        for (i = 0; i < dnsdef->nhosts; i++) {
617
            virNetworkDNSHostDefPtr host = &(dnsdef->hosts[i]);
618
            if (VIR_SOCKET_ADDR_VALID(&host->ip)) {
619
                for (j = 0; j < host->nnames; j++)
620 621
                    if (dnsmasqAddHost(dctx, &host->ip, host->names[j]) < 0)
                        return -1;
622 623
            }
        }
624 625
    }

626
    return 0;
627 628 629
}


630 631
int
networkDnsmasqConfContents(virNetworkObjPtr network,
632 633 634 635
                           const char *pidfile,
                           char **configstr,
                           dnsmasqContext *dctx,
                           dnsmasqCapsPtr caps ATTRIBUTE_UNUSED)
636
{
637
    virBuffer configbuf = VIR_BUFFER_INITIALIZER;
638
    int r, ret = -1;
639
    int nbleases = 0;
640
    int ii;
641 642 643 644
    char *record = NULL;
    char *recordPort = NULL;
    char *recordWeight = NULL;
    char *recordPriority = NULL;
645
    virNetworkDNSDefPtr dns = &network->def->dns;
G
Gene Czarcinski 已提交
646 647
    virNetworkIpDefPtr tmpipdef, ipdef, ipv4def, ipv6def;
    bool ipv6SLAAC;
648

649 650
    *configstr = NULL;

651
    /*
652 653 654
     * All dnsmasq parameters are put into a configuration file, except the
     * command line --conf-file=parameter which specifies the location of
     * configuration file.
655
     *
656 657
     * All dnsmasq conf-file parameters must be specified as "foo=bar"
     * as oppose to "--foo bar" which was acceptable on the command line.
658
     */
659 660 661 662 663 664

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

665 666
    /* create dnsmasq config file appropriate for this network */
    virBufferAsprintf(&configbuf,
667 668 669 670 671 672 673 674 675 676 677 678
                      "##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"
                      "strict-order\n"
                      "domain-needed\n",
                      network->def->name);

    if (network->def->domain) {
679
        virBufferAsprintf(&configbuf,
680 681 682 683 684
                          "domain=%s\n"
                          "expand-hosts\n",
                          network->def->domain);
    }
    /* need to specify local even if no domain specified */
685
    virBufferAsprintf(&configbuf,
686 687
                      "local=/%s/\n",
                      network->def->domain ? network->def->domain : "");
688

689
    if (pidfile)
690
        virBufferAsprintf(&configbuf, "pid-file=%s\n", pidfile);
691

692 693 694
    /* dnsmasq will *always* listen on localhost unless told otherwise */
    virBufferAddLit(&configbuf, "except-interface=lo\n");

695 696 697 698 699 700 701 702
    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.
         */
703
        virBufferAsprintf(&configbuf,
704 705 706
                          "bind-dynamic\n"
                          "interface=%s\n",
                          network->def->bridge);
707
    } else {
708
        virBufferAddLit(&configbuf, "bind-interfaces\n");
709 710 711 712 713 714 715 716 717 718 719 720 721 722 723
        /*
         * --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
         */
        for (ii = 0;
             (tmpipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, ii));
             ii++) {
            char *ipaddr = virSocketAddrFormat(&tmpipdef->address);

            if (!ipaddr)
                goto cleanup;
724

725
            /* also part of CVE 2012-3411 - if the host's version of
726
             * dnsmasq doesn't have bind-dynamic, only allow listening on
727 728
             * private/local IP addresses (see RFC1918/RFC3484/RFC4193)
             */
729 730
            if (!dnsmasqCapsGet(caps, DNSMASQ_CAPS_BINDTODEVICE) &&
                !virSocketAddrIsPrivate(&tmpipdef->address)) {
731 732 733 734
                unsigned long version = dnsmasqCapsGetVersion(caps);

                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("Publicly routable address %s is prohibited. "
735
                                 "The version of dnsmasq on this host (%d.%d) "
736 737 738 739
                                 "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 "
740 741 742 743 744 745
                                 "(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);
746 747
                goto cleanup;
            }
748
            virBufferAsprintf(&configbuf, "listen-address=%s\n", ipaddr);
749 750 751
            VIR_FREE(ipaddr);
        }
    }
752

753 754
    /* If this is an isolated network, set the default route option
     * (3) to be empty to avoid setting a default route that's
755
     * guaranteed to not work, and set no-resolv so that no dns
756 757 758
     * 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).
759
     */
760
    if (network->def->forward.type == VIR_NETWORK_FORWARD_NONE) {
761
        virBufferAddLit(&configbuf, "dhcp-option=3\n"
762
                        "no-resolv\n");
763
    }
764

765
    for (ii = 0; ii < dns->ntxts; ii++) {
766 767 768
        virBufferAsprintf(&configbuf, "txt-record=%s,%s\n",
                          dns->txts[ii].name,
                          dns->txts[ii].value);
769
    }
770

771 772 773 774 775 776
    for (ii = 0; ii < dns->nsrvs; ii++) {
        if (dns->srvs[ii].service && dns->srvs[ii].protocol) {
            if (dns->srvs[ii].port) {
                if (virAsprintf(&recordPort, "%d", dns->srvs[ii].port) < 0) {
                    virReportOOMError();
                    goto cleanup;
777
                }
778 779 780 781 782
            }
            if (dns->srvs[ii].priority) {
                if (virAsprintf(&recordPriority, "%d", dns->srvs[ii].priority) < 0) {
                    virReportOOMError();
                    goto cleanup;
783
                }
784 785 786
            }
            if (dns->srvs[ii].weight) {
                if (virAsprintf(&recordWeight, "%d", dns->srvs[ii].weight) < 0) {
787 788 789
                    virReportOOMError();
                    goto cleanup;
                }
790
            }
791

792 793 794
            if (virAsprintf(&record, "%s.%s.%s,%s,%s,%s,%s",
                            dns->srvs[ii].service,
                            dns->srvs[ii].protocol,
795 796 797 798 799
                            dns->srvs[ii].domain ? dns->srvs[ii].domain : "",
                            dns->srvs[ii].target ? dns->srvs[ii].target : "",
                            recordPort           ? recordPort           : "",
                            recordPriority       ? recordPriority       : "",
                            recordWeight         ? recordWeight         : "") < 0) {
800 801
                virReportOOMError();
                goto cleanup;
802
            }
803

804
            virBufferAsprintf(&configbuf, "srv-host=%s\n", record);
805 806 807 808
            VIR_FREE(record);
            VIR_FREE(recordPort);
            VIR_FREE(recordWeight);
            VIR_FREE(recordPriority);
809
        }
810 811
    }

G
Gene Czarcinski 已提交
812 813 814 815 816 817 818 819
    /* Find the first dhcp for both IPv4 and IPv6 */
    for (ii = 0, ipv4def = NULL, ipv6def = NULL, ipv6SLAAC = false;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, ii));
         ii++) {
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
            if (ipdef->nranges || ipdef->nhosts) {
                if (ipv4def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
820 821
                                   _("For IPv4, multiple DHCP definitions "
                                     "cannot be specified."));
G
Gene Czarcinski 已提交
822 823 824 825 826 827 828 829 830 831 832
                    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,
833 834 835 836 837 838 839 840 841
                                   _("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
842 843 844 845
                    goto cleanup;
                }
                if (ipv6def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
846 847
                                   _("For IPv6, multiple DHCP definitions "
                                     "cannot be specified."));
G
Gene Czarcinski 已提交
848 849 850 851 852 853 854 855 856 857 858 859 860
                    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 "
861 862 863 864
                 "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
865 866 867 868 869
    }

    ipdef = ipv4def ? ipv4def : ipv6def;

    while (ipdef) {
870
        for (r = 0 ; r < ipdef->nranges ; r++) {
871
            char *saddr = virSocketAddrFormat(&ipdef->ranges[r].start);
872 873
            if (!saddr)
                goto cleanup;
874
            char *eaddr = virSocketAddrFormat(&ipdef->ranges[r].end);
875 876 877 878
            if (!eaddr) {
                VIR_FREE(saddr);
                goto cleanup;
            }
879
            virBufferAsprintf(&configbuf, "dhcp-range=%s,%s\n",
880
                              saddr, eaddr);
881
            VIR_FREE(saddr);
882
            VIR_FREE(eaddr);
883 884
            nbleases += virSocketAddrGetRange(&ipdef->ranges[r].start,
                                              &ipdef->ranges[r].end);
885
        }
886

887
        /*
888 889 890 891
         * 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)
892 893
         */
        if (!ipdef->nranges && ipdef->nhosts) {
894
            char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
895 896
            if (!bridgeaddr)
                goto cleanup;
897
            virBufferAsprintf(&configbuf, "dhcp-range=%s,static\n", bridgeaddr);
898 899
            VIR_FREE(bridgeaddr);
        }
900

G
Gene Czarcinski 已提交
901 902
        if (networkBuildDnsmasqDhcpHostsList(dctx, ipdef) < 0)
            goto cleanup;
903

G
Gene Czarcinski 已提交
904 905 906
        /* Note: the following is IPv4 only */
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
            if (ipdef->nranges || ipdef->nhosts)
907
                virBufferAddLit(&configbuf, "dhcp-no-override\n");
908

G
Gene Czarcinski 已提交
909
            if (ipdef->tftproot) {
910 911
                virBufferAddLit(&configbuf, "enable-tftp\n");
                virBufferAsprintf(&configbuf, "tftp-root=%s\n", ipdef->tftproot);
G
Gene Czarcinski 已提交
912
            }
913

G
Gene Czarcinski 已提交
914 915 916
            if (ipdef->bootfile) {
                if (VIR_SOCKET_ADDR_VALID(&ipdef->bootserver)) {
                    char *bootserver = virSocketAddrFormat(&ipdef->bootserver);
917

G
Gene Czarcinski 已提交
918 919 920 921
                    if (!bootserver) {
                        virReportOOMError();
                        goto cleanup;
                    }
922
                    virBufferAsprintf(&configbuf, "dhcp-boot=%s%s%s\n",
923
                                      ipdef->bootfile, ",,", bootserver);
G
Gene Czarcinski 已提交
924 925
                    VIR_FREE(bootserver);
                } else {
926
                    virBufferAsprintf(&configbuf, "dhcp-boot=%s\n", ipdef->bootfile);
G
Gene Czarcinski 已提交
927 928 929 930 931
                }
            }
        }
        ipdef = (ipdef == ipv6def) ? NULL : ipv6def;
    }
932

G
Gene Czarcinski 已提交
933 934 935 936 937
    if (nbleases > 0) {
        char *leasefile = networkDnsmasqLeaseFileName(network->def->name);
        if (!leasefile) {
            virReportOOMError();
            goto cleanup;
938
        }
939
        virBufferAsprintf(&configbuf, "dhcp-leasefile=%s\n", leasefile);
G
Gene Czarcinski 已提交
940
        VIR_FREE(leasefile);
941
        virBufferAsprintf(&configbuf, "dhcp-lease-max=%d\n", nbleases);
G
Gene Czarcinski 已提交
942
    }
943

G
Gene Czarcinski 已提交
944 945
    /* this is done once per interface */
    if (networkBuildDnsmasqHostsList(dctx, dns) < 0)
946
        goto cleanup;
G
Gene Czarcinski 已提交
947 948 949 950 951 952

    /* 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)
953 954
        virBufferAsprintf(&configbuf, "dhcp-hostsfile=%s\n",
                          dctx->hostsfile->path);
G
Gene Czarcinski 已提交
955

956 957
    /* Likewise, always create this file and put it on the
     * commandline, to allow for runtime additions.
G
Gene Czarcinski 已提交
958
     */
959
    virBufferAsprintf(&configbuf, "addn-hosts=%s\n",
960
                      dctx->addnhostsfile->path);
G
Gene Czarcinski 已提交
961 962 963 964

    /* Are we doing RA instead of radvd? */
    if (DNSMASQ_RA_SUPPORT(caps)) {
        if (ipv6def)
965
            virBufferAddLit(&configbuf, "enable-ra\n");
G
Gene Czarcinski 已提交
966 967
        else {
            for (ii = 0;
968 969
                 (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, ii));
                 ii++) {
G
Gene Czarcinski 已提交
970 971 972 973
                if (!(ipdef->nranges || ipdef->nhosts)) {
                    char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
                    if (!bridgeaddr)
                        goto cleanup;
974 975
                    virBufferAsprintf(&configbuf,
                                      "dhcp-range=%s,ra-only\n", bridgeaddr);
G
Gene Czarcinski 已提交
976 977
                    VIR_FREE(bridgeaddr);
                }
978
            }
979
        }
980 981
    }

982 983 984
    if (!(*configstr = virBufferContentAndReset(&configbuf)))
        goto cleanup;

985
    ret = 0;
G
Gene Czarcinski 已提交
986

987
cleanup:
988
    virBufferFreeAndReset(&configbuf);
989 990 991 992
    VIR_FREE(record);
    VIR_FREE(recordPort);
    VIR_FREE(recordWeight);
    VIR_FREE(recordPriority);
993
    return ret;
994 995
}

996 997
/* build the dnsmasq command line */
static int
998
networkBuildDhcpDaemonCommandLine(virNetworkObjPtr network, virCommandPtr *cmdout,
999 1000
                                  char *pidfile, dnsmasqContext *dctx,
                                  dnsmasqCapsPtr caps)
1001
{
1002
    virCommandPtr cmd = NULL;
G
Gene Czarcinski 已提交
1003
    int ret = -1;
1004 1005
    char *configfile = NULL;
    char *configstr = NULL;
1006 1007

    network->dnsmasqPid = -1;
1008

1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
    if (networkDnsmasqConfContents(network, pidfile, &configstr, dctx, caps) < 0)
        goto cleanup;
    if (!configstr)
        goto cleanup;

    /* construct the filename */
    if (!(configfile = networkDnsmasqConfigFileName(network->def->name))) {
        virReportOOMError();
        goto cleanup;
    }

    /* Write the file */
    if (virFileWriteStr(configfile, configstr, 0600) < 0) {
        virReportSystemError(errno,
                         _("couldn't write dnsmasq config file '%s'"),
                         configfile);
1025 1026 1027
        goto cleanup;
    }

1028 1029 1030
    cmd = virCommandNew(dnsmasqCapsGetBinaryPath(caps));
    virCommandAddArgFormat(cmd, "--conf-file=%s", configfile);

1031 1032 1033 1034 1035 1036 1037 1038 1039 1040
    if (cmdout)
        *cmdout = cmd;
    ret = 0;
cleanup:
    if (ret < 0)
        virCommandFree(cmd);
    return ret;
}

static int
1041 1042
networkStartDhcpDaemon(struct network_driver *driver,
                       virNetworkObjPtr network)
1043 1044 1045 1046
{
    virCommandPtr cmd = NULL;
    char *pidfile = NULL;
    int ret = -1;
1047
    dnsmasqContext *dctx = NULL;
1048

1049
    if (!virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, 0)) {
G
Gene Czarcinski 已提交
1050
        /* no IP addresses, so we don't need to run */
1051 1052 1053 1054
        ret = 0;
        goto cleanup;
    }

1055 1056
    if (virFileMakePath(NETWORK_PID_DIR) < 0) {
        virReportSystemError(errno,
1057 1058
                             _("cannot create directory %s"),
                             NETWORK_PID_DIR);
1059
        goto cleanup;
1060
    }
1061 1062
    if (virFileMakePath(NETWORK_STATE_DIR) < 0) {
        virReportSystemError(errno,
1063 1064
                             _("cannot create directory %s"),
                             NETWORK_STATE_DIR);
1065
        goto cleanup;
1066 1067
    }

1068
    if (!(pidfile = virPidFileBuildPath(NETWORK_PID_DIR, network->def->name))) {
1069
        virReportOOMError();
1070
        goto cleanup;
1071 1072
    }

1073 1074
    if (virFileMakePath(DNSMASQ_STATE_DIR) < 0) {
        virReportSystemError(errno,
1075 1076 1077 1078 1079
                             _("cannot create directory %s"),
                             DNSMASQ_STATE_DIR);
        goto cleanup;
    }

1080 1081 1082 1083
    dctx = dnsmasqContextNew(network->def->name, DNSMASQ_STATE_DIR);
    if (dctx == NULL)
        goto cleanup;

1084 1085 1086 1087
    dnsmasqCapsRefresh(&driver->dnsmasqCaps, false);

    ret = networkBuildDhcpDaemonCommandLine(network, &cmd, pidfile,
                                            dctx, driver->dnsmasqCaps);
1088 1089 1090 1091 1092
    if (ret < 0)
        goto cleanup;

    ret = dnsmasqSave(dctx);
    if (ret < 0)
1093
        goto cleanup;
1094

G
Guido Günther 已提交
1095 1096
    ret = virCommandRun(cmd, NULL);
    if (ret < 0) {
1097
        goto cleanup;
G
Guido Günther 已提交
1098
    }
1099 1100

    /*
1101 1102 1103 1104 1105
     * 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
1106 1107
     */

1108
    ret = virPidFileRead(NETWORK_PID_DIR, network->def->name,
1109 1110
                         &network->dnsmasqPid);
    if (ret < 0)
1111
        goto cleanup;
1112

1113 1114 1115
    ret = 0;
cleanup:
    VIR_FREE(pidfile);
1116
    virCommandFree(cmd);
1117
    dnsmasqContextFree(dctx);
1118 1119 1120
    return ret;
}

1121 1122
/* networkRefreshDhcpDaemon:
 *  Update dnsmasq config files, then send a SIGHUP so that it rereads
G
Gene Czarcinski 已提交
1123 1124
 *  them.   This only works for the dhcp-hostsfile and the
 *  addn-hosts file.
1125 1126 1127
 *
 *  Returns 0 on success, -1 on failure.
 */
1128
static int
1129 1130
networkRefreshDhcpDaemon(struct network_driver *driver,
                         virNetworkObjPtr network)
1131
{
1132
    int ret = -1, ii;
G
Gene Czarcinski 已提交
1133
    virNetworkIpDefPtr ipdef, ipv4def, ipv6def;
1134
    dnsmasqContext *dctx = NULL;
1135

G
Gene Czarcinski 已提交
1136
    /* if no IP addresses specified, nothing to do */
1137
    if (!virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, 0))
G
Gene Czarcinski 已提交
1138 1139
        return 0;

1140 1141
    /* if there's no running dnsmasq, just start it */
    if (network->dnsmasqPid <= 0 || (kill(network->dnsmasqPid, 0) < 0))
1142
        return networkStartDhcpDaemon(driver, network);
1143

G
Gene Czarcinski 已提交
1144 1145 1146 1147 1148 1149 1150 1151 1152
    VIR_INFO("Refreshing dnsmasq for network %s", network->def->bridge);
    if (!(dctx = dnsmasqContextNew(network->def->name, DNSMASQ_STATE_DIR)))
        goto cleanup;

    /* 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;
1153 1154 1155
    for (ii = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, ii));
         ii++) {
G
Gene Czarcinski 已提交
1156 1157
        if (!ipv4def && (ipdef->nranges || ipdef->nhosts))
            ipv4def = ipdef;
1158
    }
1159 1160 1161
    /* If no IPv4 addresses had dhcp info, pick the first (if there were any). */
    if (!ipdef)
        ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, 0);
1162

G
Gene Czarcinski 已提交
1163 1164 1165 1166 1167 1168
    ipv6def = NULL;
    for (ii = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, ii));
         ii++) {
        if (!ipv6def && (ipdef->nranges || ipdef->nhosts))
            ipv6def = ipdef;
1169 1170
    }

G
Gene Czarcinski 已提交
1171 1172 1173 1174 1175
    if (ipv4def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv4def) < 0))
           goto cleanup;

    if (ipv6def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv6def) < 0))
           goto cleanup;
1176

G
Gene Czarcinski 已提交
1177
    if (networkBuildDnsmasqHostsList(dctx, &network->def->dns) < 0)
1178 1179 1180
       goto cleanup;

    if ((ret = dnsmasqSave(dctx)) < 0)
1181
        goto cleanup;
1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196

    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
1197 1198
networkRestartDhcpDaemon(struct network_driver *driver,
                         virNetworkObjPtr network)
1199 1200 1201 1202 1203 1204
{
    /* if there is a running dnsmasq, kill it */
    if (network->dnsmasqPid > 0) {
        networkKillDaemon(network->dnsmasqPid, "dnsmasq",
                          network->def->name);
        network->dnsmasqPid = -1;
1205
    }
1206
    /* now start dnsmasq if it should be started */
1207
    return networkStartDhcpDaemon(driver, network);
1208 1209
}

G
Gene Czarcinski 已提交
1210 1211 1212 1213 1214 1215
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";

1216 1217 1218
static int
networkRadvdConfContents(virNetworkObjPtr network, char **configstr)
{
E
Eric Blake 已提交
1219
    virBuffer configbuf = VIR_BUFFER_INITIALIZER;
1220 1221
    int ret = -1, ii;
    virNetworkIpDefPtr ipdef;
G
Gene Czarcinski 已提交
1222
    bool v6present = false, dhcp6 = false;
1223 1224

    *configstr = NULL;
1225

G
Gene Czarcinski 已提交
1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242
    /* Check if DHCPv6 is needed */
    for (ii = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, ii));
         ii++) {
        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;
    }

1243 1244 1245
    /* create radvd config file appropriate for this network;
     * IgnoreIfMissing allows radvd to start even when the bridge is down
     */
1246
    virBufferAsprintf(&configbuf, "interface %s\n"
1247 1248
                      "{\n"
                      "  AdvSendAdvert on;\n"
1249
                      "  IgnoreIfMissing on;\n"
G
Gene Czarcinski 已提交
1250 1251 1252 1253 1254
                      "  AdvManagedFlag %s;\n"
                      "%s",
                      network->def->bridge,
                      dhcp6 ? "on" : "off",
                      dhcp6 ? "\n" : radvd1);
1255 1256

    /* add a section for each IPv6 address in the config */
1257 1258 1259 1260 1261 1262 1263 1264
    for (ii = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET6, ii));
         ii++) {
        int prefix;
        char *netaddr;

        prefix = virNetworkIpDefPrefix(ipdef);
        if (prefix < 0) {
1265 1266 1267
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("bridge '%s' has an invalid prefix"),
                           network->def->bridge);
1268 1269
            goto cleanup;
        }
1270
        if (!(netaddr = virSocketAddrFormat(&ipdef->address)))
1271
            goto cleanup;
1272
        virBufferAsprintf(&configbuf,
1273
                          "  prefix %s/%d\n"
G
Gene Czarcinski 已提交
1274 1275 1276
                          "  {\n%s  };\n",
                          netaddr, prefix,
                          dhcp6 ? radvd2 : radvd3);
1277 1278 1279
        VIR_FREE(netaddr);
    }

1280 1281 1282
    /* only create the string if we found at least one IPv6 address */
    if (v6present) {
        virBufferAddLit(&configbuf, "};\n");
1283

1284 1285 1286 1287 1288 1289 1290 1291
        if (virBufferError(&configbuf)) {
            virReportOOMError();
            goto cleanup;
        }
        if (!(*configstr = virBufferContentAndReset(&configbuf))) {
            virReportOOMError();
            goto cleanup;
        }
1292
    }
1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317

    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;
1318 1319 1320 1321
        goto cleanup;
    }

    /* construct the filename */
1322
    if (!(*configFile = networkRadvdConfigFileName(network->def->name))) {
1323 1324 1325 1326
        virReportOOMError();
        goto cleanup;
    }
    /* write the file */
1327
    if (virFileWriteStr(*configFile, configStr, 0600) < 0) {
1328 1329
        virReportSystemError(errno,
                             _("couldn't write radvd config file '%s'"),
1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341
                             *configFile);
        goto cleanup;
    }

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

static int
G
Gene Czarcinski 已提交
1342 1343
networkStartRadvd(struct network_driver *driver ATTRIBUTE_UNUSED,
                        virNetworkObjPtr network)
1344 1345 1346 1347 1348 1349 1350 1351 1352
{
    char *pidfile = NULL;
    char *radvdpidbase = NULL;
    char *configfile = NULL;
    virCommandPtr cmd = NULL;
    int ret = -1;

    network->radvdPid = -1;

G
Gene Czarcinski 已提交
1353
    /* Is dnsmasq handling RA? */
1354
   if (DNSMASQ_RA_SUPPORT(driver->dnsmasqCaps)) {
G
Gene Czarcinski 已提交
1355 1356 1357 1358
        ret = 0;
        goto cleanup;
    }

1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369
    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);
1370 1371 1372
        goto cleanup;
    }

1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398
    if (virFileMakePath(NETWORK_PID_DIR) < 0) {
        virReportSystemError(errno,
                             _("cannot create directory %s"),
                             NETWORK_PID_DIR);
        goto cleanup;
    }
    if (virFileMakePath(RADVD_STATE_DIR) < 0) {
        virReportSystemError(errno,
                             _("cannot create directory %s"),
                             RADVD_STATE_DIR);
        goto cleanup;
    }

    /* construct pidfile name */
    if (!(radvdpidbase = networkRadvdPidfileBasename(network->def->name))) {
        virReportOOMError();
        goto cleanup;
    }
    if (!(pidfile = virPidFileBuildPath(NETWORK_PID_DIR, radvdpidbase))) {
        virReportOOMError();
        goto cleanup;
    }

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

1399 1400 1401 1402
    /* 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
1403
     * virPidFileRead() below will fail if we use them).
1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418
     * 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;

1419
    if (virPidFileRead(NETWORK_PID_DIR, radvdpidbase, &network->radvdPid) < 0)
1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430
        goto cleanup;

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

1431
static int
1432 1433
networkRefreshRadvd(struct network_driver *driver ATTRIBUTE_UNUSED,
                    virNetworkObjPtr network)
1434
{
G
Gene Czarcinski 已提交
1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
    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)) {
            virPidFileDelete(NETWORK_PID_DIR, radvdpidbase);
            VIR_FREE(radvdpidbase);
        }
        network->radvdPid = -1;
        return 0;
    }

1453 1454
    /* if there's no running radvd, just start it */
    if (network->radvdPid <= 0 || (kill(network->radvdPid, 0) < 0))
G
Gene Czarcinski 已提交
1455
        return networkStartRadvd(driver, network);
1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467

    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);
}

1468 1469
#if 0
/* currently unused, so it causes a build error unless we #if it out */
1470
static int
1471 1472
networkRestartRadvd(struct network_driver *driver,
                    virNetworkObjPtr network)
1473 1474 1475 1476 1477 1478 1479 1480 1481
{
    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
1482
        if ((networkKillDaemon(network->radvdPid, "radvd",
1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495
                               network->def->name) >= 0) &&
            ((radvdpidbase = networkRadvdPidfileBasename(network->def->name))
             != NULL)) {
            virPidFileDelete(NETWORK_PID_DIR, radvdpidbase);
            VIR_FREE(radvdpidbase);
        }
        network->radvdPid = -1;
    }
    /* now start radvd if it should be started */
    return networkStartRadvd(network);
}
#endif /* #if 0 */

1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510
/* SIGHUP/restart any dnsmasq or radvd daemons.
 * This should be called when libvirtd is restarted.
 */
static void
networkRefreshDaemons(struct network_driver *driver)
{
    unsigned int i;

    VIR_INFO("Refreshing network daemons");

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

        virNetworkObjLock(network);
        if (virNetworkObjIsActive(network) &&
1511 1512 1513
            ((network->def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE))) {
1514 1515 1516 1517 1518 1519
            /* 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.
             */
1520 1521
            networkRefreshDhcpDaemon(driver, network);
            networkRefreshRadvd(driver, network);
1522 1523 1524 1525 1526
        }
        virNetworkObjUnlock(network);
    }
}

1527
static int
1528
networkAddMasqueradingIptablesRules(struct network_driver *driver,
1529 1530
                                    virNetworkObjPtr network,
                                    virNetworkIpDefPtr ipdef)
1531 1532
{
    int prefix = virNetworkIpDefPrefix(ipdef);
1533
    const char *forwardIf = virNetworkDefForwardIf(network->def, 0);
1534 1535

    if (prefix < 0) {
1536 1537 1538
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Invalid prefix or netmask for '%s'"),
                       network->def->bridge);
1539 1540
        goto masqerr1;
    }
1541

1542
    /* allow forwarding packets from the bridge interface */
1543
    if (iptablesAddForwardAllowOut(driver->iptables,
1544
                                   &ipdef->address,
1545
                                   prefix,
1546
                                   network->def->bridge,
1547
                                   forwardIf) < 0) {
1548 1549 1550
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to allow forwarding from '%s'"),
                       network->def->bridge);
1551 1552 1553
        goto masqerr1;
    }

1554 1555 1556
    /* allow forwarding packets to the bridge interface if they are
     * part of an existing connection
     */
1557
    if (iptablesAddForwardAllowRelatedIn(driver->iptables,
1558
                                         &ipdef->address,
1559
                                         prefix,
1560
                                         network->def->bridge,
1561
                                         forwardIf) < 0) {
1562 1563 1564
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to allow forwarding to '%s'"),
                       network->def->bridge);
1565 1566 1567
        goto masqerr2;
    }

1568 1569 1570 1571 1572
    /*
     * Enable masquerading.
     *
     * We need to end up with 3 rules in the table in this order
     *
E
Eric Blake 已提交
1573 1574
     *  1. protocol=tcp with sport mapping restriction
     *  2. protocol=udp with sport mapping restriction
1575 1576 1577
     *  3. generic any protocol
     *
     * The sport mappings are required, because default IPtables
E
Eric Blake 已提交
1578
     * MASQUERADE maintain port numbers unchanged where possible.
1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591
     *
     * NFS can be configured to only "trust" port numbers < 1023.
     *
     * Guests using NAT thus need to be prevented from having port
     * numbers < 1023, otherwise they can bypass the NFS "security"
     * check on the source port number.
     *
     * Since we use '--insert' to add rules to the header of the
     * chain, we actually need to add them in the reverse of the
     * order just mentioned !
     */

    /* First the generic masquerade rule for other protocols */
1592
    if (iptablesAddForwardMasquerade(driver->iptables,
1593
                                     &ipdef->address,
1594
                                     prefix,
1595
                                     forwardIf,
1596
                                     NULL) < 0) {
1597 1598 1599 1600 1601
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       forwardIf ?
                       _("failed to add iptables rule to enable masquerading to %s") :
                       _("failed to add iptables rule to enable masquerading"),
                       forwardIf);
1602 1603 1604
        goto masqerr3;
    }

1605
    /* UDP with a source port restriction */
1606
    if (iptablesAddForwardMasquerade(driver->iptables,
1607
                                     &ipdef->address,
1608
                                     prefix,
1609
                                     forwardIf,
1610
                                     "udp") < 0) {
1611 1612 1613 1614 1615
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       forwardIf ?
                       _("failed to add iptables rule to enable UDP masquerading to %s") :
                       _("failed to add iptables rule to enable UDP masquerading"),
                       forwardIf);
1616 1617 1618 1619
        goto masqerr4;
    }

    /* TCP with a source port restriction */
1620
    if (iptablesAddForwardMasquerade(driver->iptables,
1621
                                     &ipdef->address,
1622
                                     prefix,
1623
                                     forwardIf,
1624
                                     "tcp") < 0) {
1625 1626 1627 1628 1629
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       forwardIf ?
                       _("failed to add iptables rule to enable TCP masquerading to %s") :
                       _("failed to add iptables rule to enable TCP masquerading"),
                       forwardIf);
1630 1631 1632
        goto masqerr5;
    }

1633
    return 0;
1634

1635 1636
 masqerr5:
    iptablesRemoveForwardMasquerade(driver->iptables,
1637
                                    &ipdef->address,
1638
                                    prefix,
1639
                                    forwardIf,
1640 1641 1642
                                    "udp");
 masqerr4:
    iptablesRemoveForwardMasquerade(driver->iptables,
1643
                                    &ipdef->address,
1644
                                    prefix,
1645
                                    forwardIf,
1646
                                    NULL);
1647 1648
 masqerr3:
    iptablesRemoveForwardAllowRelatedIn(driver->iptables,
1649
                                        &ipdef->address,
1650
                                        prefix,
1651
                                        network->def->bridge,
1652
                                        forwardIf);
1653 1654
 masqerr2:
    iptablesRemoveForwardAllowOut(driver->iptables,
1655
                                  &ipdef->address,
1656
                                  prefix,
1657
                                  network->def->bridge,
1658
                                  forwardIf);
1659
 masqerr1:
1660
    return -1;
1661 1662
}

1663 1664 1665 1666 1667 1668
static void
networkRemoveMasqueradingIptablesRules(struct network_driver *driver,
                                       virNetworkObjPtr network,
                                       virNetworkIpDefPtr ipdef)
{
    int prefix = virNetworkIpDefPrefix(ipdef);
1669
    const char *forwardIf = virNetworkDefForwardIf(network->def, 0);
1670 1671 1672 1673 1674

    if (prefix >= 0) {
        iptablesRemoveForwardMasquerade(driver->iptables,
                                        &ipdef->address,
                                        prefix,
1675
                                        forwardIf,
1676 1677 1678 1679
                                        "tcp");
        iptablesRemoveForwardMasquerade(driver->iptables,
                                        &ipdef->address,
                                        prefix,
1680
                                        forwardIf,
1681 1682 1683 1684
                                        "udp");
        iptablesRemoveForwardMasquerade(driver->iptables,
                                        &ipdef->address,
                                        prefix,
1685
                                        forwardIf,
1686 1687 1688 1689 1690 1691
                                        NULL);

        iptablesRemoveForwardAllowRelatedIn(driver->iptables,
                                            &ipdef->address,
                                            prefix,
                                            network->def->bridge,
1692
                                            forwardIf);
1693 1694 1695 1696
        iptablesRemoveForwardAllowOut(driver->iptables,
                                      &ipdef->address,
                                      prefix,
                                      network->def->bridge,
1697
                                      forwardIf);
1698 1699 1700
    }
}

1701
static int
1702
networkAddRoutingIptablesRules(struct network_driver *driver,
1703
                               virNetworkObjPtr network,
1704 1705
                               virNetworkIpDefPtr ipdef)
{
1706
    int prefix = virNetworkIpDefPrefix(ipdef);
1707
    const char *forwardIf = virNetworkDefForwardIf(network->def, 0);
1708 1709

    if (prefix < 0) {
1710 1711 1712
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Invalid prefix or netmask for '%s'"),
                       network->def->bridge);
1713 1714
        goto routeerr1;
    }
1715

1716
    /* allow routing packets from the bridge interface */
1717
    if (iptablesAddForwardAllowOut(driver->iptables,
1718
                                   &ipdef->address,
1719
                                   prefix,
1720
                                   network->def->bridge,
1721
                                   forwardIf) < 0) {
1722 1723 1724
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to allow routing from '%s'"),
                       network->def->bridge);
1725 1726 1727 1728
        goto routeerr1;
    }

    /* allow routing packets to the bridge interface */
1729
    if (iptablesAddForwardAllowIn(driver->iptables,
1730
                                  &ipdef->address,
1731
                                  prefix,
1732
                                  network->def->bridge,
1733
                                  forwardIf) < 0) {
1734 1735 1736
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to allow routing to '%s'"),
                       network->def->bridge);
1737 1738 1739
        goto routeerr2;
    }

1740
    return 0;
1741

1742
routeerr2:
1743
    iptablesRemoveForwardAllowOut(driver->iptables,
1744
                                  &ipdef->address,
1745
                                  prefix,
1746
                                  network->def->bridge,
1747
                                  forwardIf);
1748
routeerr1:
1749
    return -1;
1750 1751
}

1752 1753 1754 1755 1756 1757
static void
networkRemoveRoutingIptablesRules(struct network_driver *driver,
                                  virNetworkObjPtr network,
                                  virNetworkIpDefPtr ipdef)
{
    int prefix = virNetworkIpDefPrefix(ipdef);
1758
    const char *forwardIf = virNetworkDefForwardIf(network->def, 0);
1759 1760 1761 1762 1763 1764

    if (prefix >= 0) {
        iptablesRemoveForwardAllowIn(driver->iptables,
                                     &ipdef->address,
                                     prefix,
                                     network->def->bridge,
1765
                                     forwardIf);
1766 1767 1768 1769 1770

        iptablesRemoveForwardAllowOut(driver->iptables,
                                      &ipdef->address,
                                      prefix,
                                      network->def->bridge,
1771
                                      forwardIf);
1772 1773 1774
    }
}

1775 1776
/* Add all once/network rules required for IPv6.
 * If no IPv6 addresses are defined and <network ipv6='yes'> is
G
Gene Czarcinski 已提交
1777 1778
 * specified, then allow IPv6 commuinications between virtual systems.
 * If any IPv6 addresses are defined, then add the rules for regular operation.
1779
 */
1780 1781 1782 1783 1784
static int
networkAddGeneralIp6tablesRules(struct network_driver *driver,
                               virNetworkObjPtr network)
{

1785 1786
    if (!virNetworkDefGetIpByIndex(network->def, AF_INET6, 0) &&
        !network->def->ipv6nogw) {
1787
        return 0;
1788
    }
1789 1790 1791 1792 1793

    /* Catch all rules to block forwarding to/from bridges */

    if (iptablesAddForwardRejectOut(driver->iptables, AF_INET6,
                                    network->def->bridge) < 0) {
1794 1795 1796
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add ip6tables rule to block outbound traffic from '%s'"),
                       network->def->bridge);
1797 1798 1799 1800 1801
        goto err1;
    }

    if (iptablesAddForwardRejectIn(driver->iptables, AF_INET6,
                                   network->def->bridge) < 0) {
1802 1803 1804
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add ip6tables rule to block inbound traffic to '%s'"),
                       network->def->bridge);
1805 1806 1807 1808 1809 1810
        goto err2;
    }

    /* Allow traffic between guests on the same bridge */
    if (iptablesAddForwardAllowCross(driver->iptables, AF_INET6,
                                     network->def->bridge) < 0) {
1811 1812 1813
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add ip6tables rule to allow cross bridge traffic on '%s'"),
                       network->def->bridge);
1814 1815 1816
        goto err3;
    }

1817 1818 1819 1820
    /* if no IPv6 addresses are defined, we are done. */
    if (!virNetworkDefGetIpByIndex(network->def, AF_INET6, 0))
        return 0;

1821 1822 1823
    /* allow DNS over IPv6 */
    if (iptablesAddTcpInput(driver->iptables, AF_INET6,
                            network->def->bridge, 53) < 0) {
1824 1825 1826
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add ip6tables rule to allow DNS requests from '%s'"),
                       network->def->bridge);
1827 1828 1829 1830 1831
        goto err4;
    }

    if (iptablesAddUdpInput(driver->iptables, AF_INET6,
                            network->def->bridge, 53) < 0) {
1832 1833 1834
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add ip6tables rule to allow DNS requests from '%s'"),
                       network->def->bridge);
1835 1836 1837
        goto err5;
    }

G
Gene Czarcinski 已提交
1838 1839 1840 1841 1842 1843 1844 1845
    if (iptablesAddUdpInput(driver->iptables, AF_INET6,
                            network->def->bridge, 547) < 0) {
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add ip6tables rule to allow DHCP6 requests from '%s'"),
                       network->def->bridge);
        goto err6;
    }

1846 1847 1848
    return 0;

    /* unwind in reverse order from the point of failure */
G
Gene Czarcinski 已提交
1849 1850
err6:
    iptablesRemoveUdpInput(driver->iptables, AF_INET6, network->def->bridge, 53);
1851 1852 1853 1854
err5:
    iptablesRemoveTcpInput(driver->iptables, AF_INET6, network->def->bridge, 53);
err4:
    iptablesRemoveForwardAllowCross(driver->iptables, AF_INET6, network->def->bridge);
1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
err3:
    iptablesRemoveForwardRejectIn(driver->iptables, AF_INET6, network->def->bridge);
err2:
    iptablesRemoveForwardRejectOut(driver->iptables, AF_INET6, network->def->bridge);
err1:
    return -1;
}

static void
networkRemoveGeneralIp6tablesRules(struct network_driver *driver,
                                  virNetworkObjPtr network)
{
1867
    if (!virNetworkDefGetIpByIndex(network->def, AF_INET6, 0) &&
G
Gene Czarcinski 已提交
1868
        !network->def->ipv6nogw) {
1869
        return;
G
Gene Czarcinski 已提交
1870
    }
1871
    if (virNetworkDefGetIpByIndex(network->def, AF_INET6, 0)) {
G
Gene Czarcinski 已提交
1872
        iptablesRemoveUdpInput(driver->iptables, AF_INET6, network->def->bridge, 547);
1873 1874 1875
        iptablesRemoveUdpInput(driver->iptables, AF_INET6, network->def->bridge, 53);
        iptablesRemoveTcpInput(driver->iptables, AF_INET6, network->def->bridge, 53);
    }
1876

1877 1878 1879
    /* the following rules are there if no IPv6 address has been defined
     * but network->def->ipv6nogw == true
     */
1880 1881 1882 1883 1884
    iptablesRemoveForwardAllowCross(driver->iptables, AF_INET6, network->def->bridge);
    iptablesRemoveForwardRejectIn(driver->iptables, AF_INET6, network->def->bridge);
    iptablesRemoveForwardRejectOut(driver->iptables, AF_INET6, network->def->bridge);
}

1885
static int
1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899
networkAddGeneralIptablesRules(struct network_driver *driver,
                               virNetworkObjPtr network)
{
    int ii;
    virNetworkIpDefPtr ipv4def;

    /* First look for first IPv4 address that has dhcp or tftpboot defined. */
    /* We support dhcp config on 1 IPv4 interface only. */
    for (ii = 0;
         (ipv4def = virNetworkDefGetIpByIndex(network->def, AF_INET, ii));
         ii++) {
        if (ipv4def->nranges || ipv4def->nhosts || ipv4def->tftproot)
            break;
    }
1900 1901

    /* allow DHCP requests through to dnsmasq */
1902

1903 1904
    if (iptablesAddTcpInput(driver->iptables, AF_INET,
                            network->def->bridge, 67) < 0) {
1905 1906 1907
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to allow DHCP requests from '%s'"),
                       network->def->bridge);
1908 1909 1910
        goto err1;
    }

1911 1912
    if (iptablesAddUdpInput(driver->iptables, AF_INET,
                            network->def->bridge, 67) < 0) {
1913 1914 1915
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to allow DHCP requests from '%s'"),
                       network->def->bridge);
1916 1917 1918
        goto err2;
    }

1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929
    /* If we are doing local DHCP service on this network, attempt to
     * add a rule that will fixup the checksum of DHCP response
     * packets back to the guests (but report failure without
     * aborting, since not all iptables implementations support it).
     */

    if (ipv4def && (ipv4def->nranges || ipv4def->nhosts) &&
        (iptablesAddOutputFixUdpChecksum(driver->iptables,
                                         network->def->bridge, 68) < 0)) {
        VIR_WARN("Could not add rule to fixup DHCP response checksums "
                 "on network '%s'.", network->def->name);
1930
        VIR_WARN("May need to update iptables package & kernel to support CHECKSUM rule.");
1931 1932
    }

1933
    /* allow DNS requests through to dnsmasq */
1934 1935
    if (iptablesAddTcpInput(driver->iptables, AF_INET,
                            network->def->bridge, 53) < 0) {
1936 1937 1938
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to allow DNS requests from '%s'"),
                       network->def->bridge);
1939 1940 1941
        goto err3;
    }

1942 1943
    if (iptablesAddUdpInput(driver->iptables, AF_INET,
                            network->def->bridge, 53) < 0) {
1944 1945 1946
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to allow DNS requests from '%s'"),
                       network->def->bridge);
1947 1948 1949
        goto err4;
    }

1950 1951
    /* allow TFTP requests through to dnsmasq if necessary */
    if (ipv4def && ipv4def->tftproot &&
1952 1953
        iptablesAddUdpInput(driver->iptables, AF_INET,
                            network->def->bridge, 69) < 0) {
1954 1955 1956
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to allow TFTP requests from '%s'"),
                       network->def->bridge);
1957
        goto err5;
1958 1959
    }

1960 1961
    /* Catch all rules to block forwarding to/from bridges */

1962 1963
    if (iptablesAddForwardRejectOut(driver->iptables, AF_INET,
                                    network->def->bridge) < 0) {
1964 1965 1966
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to block outbound traffic from '%s'"),
                       network->def->bridge);
1967
        goto err6;
1968 1969
    }

1970 1971
    if (iptablesAddForwardRejectIn(driver->iptables, AF_INET,
                                   network->def->bridge) < 0) {
1972 1973 1974
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to block inbound traffic to '%s'"),
                       network->def->bridge);
1975
        goto err7;
1976 1977 1978
    }

    /* Allow traffic between guests on the same bridge */
1979 1980
    if (iptablesAddForwardAllowCross(driver->iptables, AF_INET,
                                     network->def->bridge) < 0) {
1981 1982 1983
        virReportError(VIR_ERR_SYSTEM_ERROR,
                       _("failed to add iptables rule to allow cross bridge traffic on '%s'"),
                       network->def->bridge);
1984
        goto err8;
1985 1986
    }

1987 1988 1989 1990 1991
    /* add IPv6 general rules, if needed */
    if (networkAddGeneralIp6tablesRules(driver, network) < 0) {
        goto err9;
    }

1992
    return 0;
1993

1994
    /* unwind in reverse order from the point of failure */
1995 1996
err9:
    iptablesRemoveForwardAllowCross(driver->iptables, AF_INET, network->def->bridge);
1997
err8:
1998
    iptablesRemoveForwardRejectIn(driver->iptables, AF_INET, network->def->bridge);
1999
err7:
2000
    iptablesRemoveForwardRejectOut(driver->iptables, AF_INET, network->def->bridge);
2001 2002
err6:
    if (ipv4def && ipv4def->tftproot) {
2003
        iptablesRemoveUdpInput(driver->iptables, AF_INET, network->def->bridge, 69);
2004
    }
2005
err5:
2006
    iptablesRemoveUdpInput(driver->iptables, AF_INET, network->def->bridge, 53);
2007
err4:
2008
    iptablesRemoveTcpInput(driver->iptables, AF_INET, network->def->bridge, 53);
2009
err3:
2010
    iptablesRemoveUdpInput(driver->iptables, AF_INET, network->def->bridge, 67);
2011
err2:
2012
    iptablesRemoveTcpInput(driver->iptables, AF_INET, network->def->bridge, 67);
2013
err1:
2014
    return -1;
2015 2016 2017
}

static void
2018 2019 2020 2021 2022
networkRemoveGeneralIptablesRules(struct network_driver *driver,
                                  virNetworkObjPtr network)
{
    int ii;
    virNetworkIpDefPtr ipv4def;
2023

2024 2025
    networkRemoveGeneralIp6tablesRules(driver, network);

2026 2027 2028 2029 2030
    for (ii = 0;
         (ipv4def = virNetworkDefGetIpByIndex(network->def, AF_INET, ii));
         ii++) {
        if (ipv4def->nranges || ipv4def->nhosts || ipv4def->tftproot)
            break;
2031
    }
2032

2033 2034 2035
    iptablesRemoveForwardAllowCross(driver->iptables, AF_INET, network->def->bridge);
    iptablesRemoveForwardRejectIn(driver->iptables, AF_INET, network->def->bridge);
    iptablesRemoveForwardRejectOut(driver->iptables, AF_INET, network->def->bridge);
2036
    if (ipv4def && ipv4def->tftproot) {
2037
        iptablesRemoveUdpInput(driver->iptables, AF_INET, network->def->bridge, 69);
2038
    }
2039 2040
    iptablesRemoveUdpInput(driver->iptables, AF_INET, network->def->bridge, 53);
    iptablesRemoveTcpInput(driver->iptables, AF_INET, network->def->bridge, 53);
2041 2042 2043 2044
    if (ipv4def && (ipv4def->nranges || ipv4def->nhosts)) {
        iptablesRemoveOutputFixUdpChecksum(driver->iptables,
                                           network->def->bridge, 68);
    }
2045 2046
    iptablesRemoveUdpInput(driver->iptables, AF_INET, network->def->bridge, 67);
    iptablesRemoveTcpInput(driver->iptables, AF_INET, network->def->bridge, 67);
2047 2048
}

2049 2050 2051 2052 2053
static int
networkAddIpSpecificIptablesRules(struct network_driver *driver,
                                  virNetworkObjPtr network,
                                  virNetworkIpDefPtr ipdef)
{
2054 2055 2056
    /* NB: in the case of IPv6, routing rules are added when the
     * forward mode is NAT. This is because IPv6 has no NAT.
     */
2057

2058
    if (network->def->forward.type == VIR_NETWORK_FORWARD_NAT) {
2059
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET))
2060
            return networkAddMasqueradingIptablesRules(driver, network, ipdef);
2061
        else if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
2062
            return networkAddRoutingIptablesRules(driver, network, ipdef);
2063
    } else if (network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE) {
2064 2065
        return networkAddRoutingIptablesRules(driver, network, ipdef);
    }
2066 2067 2068 2069 2070 2071 2072 2073
    return 0;
}

static void
networkRemoveIpSpecificIptablesRules(struct network_driver *driver,
                                     virNetworkObjPtr network,
                                     virNetworkIpDefPtr ipdef)
{
2074
    if (network->def->forward.type == VIR_NETWORK_FORWARD_NAT) {
2075
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET))
2076
            networkRemoveMasqueradingIptablesRules(driver, network, ipdef);
2077
        else if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
2078
            networkRemoveRoutingIptablesRules(driver, network, ipdef);
2079
    } else if (network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE) {
2080
        networkRemoveRoutingIptablesRules(driver, network, ipdef);
2081
    }
2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134
}

/* Add all rules for all ip addresses (and general rules) on a network */
static int
networkAddIptablesRules(struct network_driver *driver,
                        virNetworkObjPtr network)
{
    int ii;
    virNetworkIpDefPtr ipdef;

    /* Add "once per network" rules */
    if (networkAddGeneralIptablesRules(driver, network) < 0)
        return -1;

    for (ii = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, ii));
         ii++) {
        /* Add address-specific iptables rules */
        if (networkAddIpSpecificIptablesRules(driver, network, ipdef) < 0) {
            goto err;
        }
    }
    return 0;

err:
    /* The final failed call to networkAddIpSpecificIptablesRules will
     * have removed any rules it created, but we need to remove those
     * added for previous IP addresses.
     */
    while ((--ii >= 0) &&
           (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, ii))) {
        networkRemoveIpSpecificIptablesRules(driver, network, ipdef);
    }
    networkRemoveGeneralIptablesRules(driver, network);
    return -1;
}

/* Remove all rules for all ip addresses (and general rules) on a network */
static void
networkRemoveIptablesRules(struct network_driver *driver,
                           virNetworkObjPtr network)
{
    int ii;
    virNetworkIpDefPtr ipdef;

    for (ii = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, ii));
         ii++) {
        networkRemoveIpSpecificIptablesRules(driver, network, ipdef);
    }
    networkRemoveGeneralIptablesRules(driver, network);
}

2135 2136 2137 2138 2139
static void
networkReloadIptablesRules(struct network_driver *driver)
{
    unsigned int i;

2140
    VIR_INFO("Reloading iptables rules");
2141 2142

    for (i = 0 ; i < driver->networks.count ; i++) {
2143 2144 2145 2146
        virNetworkObjPtr network = driver->networks.objs[i];

        virNetworkObjLock(network);
        if (virNetworkObjIsActive(network) &&
2147 2148 2149
            ((network->def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
             (network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE))) {
2150 2151 2152 2153 2154
            /* Only the three L3 network types that are configured by libvirt
             * need to have iptables rules reloaded.
             */
            networkRemoveIptablesRules(driver, network);
            if (networkAddIptablesRules(driver, network) < 0) {
2155 2156
                /* failed to add but already logged */
            }
2157
        }
2158
        virNetworkObjUnlock(network);
2159 2160 2161
    }
}

2162
/* Enable IP Forwarding. Return 0 for success, -1 for failure. */
2163
static int
2164
networkEnableIpForwarding(bool enableIPv4, bool enableIPv6)
2165
{
2166 2167 2168 2169 2170 2171
    int ret = 0;
    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);
    return ret;
2172 2173
}

2174 2175
#define SYSCTL_PATH "/proc/sys"

2176 2177
static int
networkSetIPv6Sysctls(virNetworkObjPtr network)
2178 2179 2180 2181
{
    char *field = NULL;
    int ret = -1;

2182 2183 2184 2185 2186 2187 2188 2189 2190
    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",
                        network->def->bridge) < 0) {
            virReportOOMError();
            goto cleanup;
        }
2191

2192 2193 2194 2195 2196 2197
        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;
        }
2198

2199 2200 2201 2202 2203 2204 2205
        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);
2206 2207
    }

2208 2209 2210 2211 2212 2213 2214 2215 2216
    /* 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",
                    network->def->bridge) < 0) {
2217
        virReportOOMError();
2218 2219 2220
        goto cleanup;
    }

2221
    if (virFileWriteStr(field, "0", 0) < 0) {
2222
        virReportSystemError(errno,
2223 2224 2225 2226 2227
                             _("cannot disable %s"), field);
        goto cleanup;
    }
    VIR_FREE(field);

2228 2229 2230 2231 2232
    /* 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",
                    network->def->bridge) < 0) {
2233
        virReportOOMError();
2234 2235 2236
        goto cleanup;
    }

2237
    if (virFileWriteStr(field, "0", 0) < 0) {
2238
        virReportSystemError(errno,
2239
                             _("cannot disable %s"), field);
2240 2241 2242 2243 2244 2245 2246 2247 2248
        goto cleanup;
    }

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

2249 2250 2251 2252 2253 2254
#define PROC_NET_ROUTE "/proc/net/route"

/* XXX: This function can be a lot more exhaustive, there are certainly
 *      other scenarios where we can ruin host network connectivity.
 * XXX: Using a proper library is preferred over parsing /proc
 */
2255 2256
static int
networkCheckRouteCollision(virNetworkObjPtr network)
2257
{
2258
    int ret = 0, len;
2259 2260 2261 2262 2263
    char *cur, *buf = NULL;
    enum {MAX_ROUTE_SIZE = 1024*64};

    /* Read whole routing table into memory */
    if ((len = virFileReadAll(PROC_NET_ROUTE, MAX_ROUTE_SIZE, &buf)) < 0)
2264
        goto out;
2265 2266 2267 2268 2269 2270 2271

    /* Dropping the last character shouldn't hurt */
    if (len > 0)
        buf[len-1] = '\0';

    VIR_DEBUG("%s output:\n%s", PROC_NET_ROUTE, buf);

2272
    if (!STRPREFIX(buf, "Iface"))
2273 2274 2275 2276 2277 2278 2279 2280 2281 2282
        goto out;

    /* First line is just headings, skip it */
    cur = strchr(buf, '\n');
    if (cur)
        cur++;

    while (cur) {
        char iface[17], dest[128], mask[128];
        unsigned int addr_val, mask_val;
2283 2284
        virNetworkIpDefPtr ipdef;
        int num, ii;
2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312

        /* NUL-terminate the line, so sscanf doesn't go beyond a newline.  */
        char *nl = strchr(cur, '\n');
        if (nl) {
            *nl++ = '\0';
        }

        num = sscanf(cur, "%16s %127s %*s %*s %*s %*s %*s %127s",
                     iface, dest, mask);
        cur = nl;

        if (num != 3) {
            VIR_DEBUG("Failed to parse %s", PROC_NET_ROUTE);
            continue;
        }

        if (virStrToLong_ui(dest, NULL, 16, &addr_val) < 0) {
            VIR_DEBUG("Failed to convert network address %s to uint", dest);
            continue;
        }

        if (virStrToLong_ui(mask, NULL, 16, &mask_val) < 0) {
            VIR_DEBUG("Failed to convert network mask %s to uint", mask);
            continue;
        }

        addr_val &= mask_val;

2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330
        for (ii = 0;
             (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, ii));
             ii++) {

            unsigned int net_dest;
            virSocketAddr netmask;

            if (virNetworkIpDefNetmask(ipdef, &netmask) < 0) {
                VIR_WARN("Failed to get netmask of '%s'",
                         network->def->bridge);
                continue;
            }

            net_dest = (ipdef->address.data.inet4.sin_addr.s_addr &
                        netmask.data.inet4.sin_addr.s_addr);

            if ((net_dest == addr_val) &&
                (netmask.data.inet4.sin_addr.s_addr == mask_val)) {
2331 2332 2333
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Network is already in use by interface %s"),
                               iface);
2334 2335 2336
                ret = -1;
                goto out;
            }
2337 2338 2339 2340 2341 2342 2343 2344
        }
    }

out:
    VIR_FREE(buf);
    return ret;
}

2345
static int
D
Daniel P. Berrange 已提交
2346
networkAddAddrToBridge(virNetworkObjPtr network,
2347
                       virNetworkIpDefPtr ipdef)
2348
{
2349 2350 2351
    int prefix = virNetworkIpDefPrefix(ipdef);

    if (prefix < 0) {
2352 2353 2354
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("bridge '%s' has an invalid netmask or IP address"),
                       network->def->bridge);
2355 2356 2357
        return -1;
    }

2358 2359
    if (virNetDevSetIPv4Address(network->def->bridge,
                                &ipdef->address, prefix) < 0)
2360 2361 2362 2363 2364 2365
        return -1;

    return 0;
}

static int
2366
networkStartNetworkVirtual(struct network_driver *driver,
2367 2368
                          virNetworkObjPtr network)
{
2369
    int ii;
2370
    bool v4present = false, v6present = false;
2371 2372
    virErrorPtr save_err = NULL;
    virNetworkIpDefPtr ipdef;
2373
    char *macTapIfName = NULL;
2374
    int tapfd = -1;
2375

2376 2377
    /* Check to see if any network IP collides with an existing route */
    if (networkCheckRouteCollision(network) < 0)
2378 2379
        return -1;

2380
    /* Create and configure the bridge device */
2381
    if (virNetDevBridgeCreate(network->def->bridge) < 0)
2382 2383
        return -1;

2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395
    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);
        if (!macTapIfName) {
            virReportOOMError();
            goto err0;
        }
2396
        /* Keep tun fd open and interface up to allow for IPv6 DAD to happen */
2397
        if (virNetDevTapCreateInBridgePort(network->def->bridge,
2398
                                           &macTapIfName, &network->def->mac,
2399 2400 2401 2402
                                           NULL, &tapfd, NULL, NULL,
                                           VIR_NETDEV_TAP_CREATE_USE_MAC_FOR_BRIDGE |
                                           VIR_NETDEV_TAP_CREATE_IFUP |
                                           VIR_NETDEV_TAP_CREATE_PERSIST) < 0) {
2403 2404 2405 2406 2407
            VIR_FREE(macTapIfName);
            goto err0;
        }
    }

2408
    /* Set bridge options */
2409 2410 2411 2412

    /* delay is configured in seconds, but virNetDevBridgeSetSTPDelay
     * expects milliseconds
     */
2413
    if (virNetDevBridgeSetSTPDelay(network->def->bridge,
2414
                                   network->def->delay * 1000) < 0)
2415
        goto err1;
2416

2417
    if (virNetDevBridgeSetSTP(network->def->bridge,
2418
                              network->def->stp ? true : false) < 0)
2419
        goto err1;
2420

2421 2422 2423 2424
    /* Disable IPv6 on the bridge if there are no IPv6 addresses
     * defined, and set other IPv6 sysctl tunables appropriately.
     */
    if (networkSetIPv6Sysctls(network) < 0)
2425
        goto err1;
2426

2427 2428 2429 2430 2431 2432 2433
    /* Add "once per network" rules */
    if (networkAddIptablesRules(driver, network) < 0)
        goto err1;

    for (ii = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_UNSPEC, ii));
         ii++) {
2434
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET))
2435
            v4present = true;
2436
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
2437
            v6present = true;
2438

2439
        /* Add the IP address/netmask to the bridge */
D
Daniel P. Berrange 已提交
2440
        if (networkAddAddrToBridge(network, ipdef) < 0) {
2441
            goto err2;
2442
        }
2443 2444
    }

2445
    /* Bring up the bridge interface */
2446
    if (virNetDevSetOnline(network->def->bridge, 1) < 0)
2447
        goto err2;
2448

2449 2450
    /* If forward.type != NONE, turn on global IP forwarding */
    if (network->def->forward.type != VIR_NETWORK_FORWARD_NONE &&
2451
        networkEnableIpForwarding(v4present, v6present) < 0) {
2452
        virReportSystemError(errno, "%s",
2453
                             _("failed to enable IP forwarding"));
2454
        goto err3;
2455 2456
    }

2457

2458
    /* start dnsmasq if there are any IP addresses (v4 or v6) */
2459 2460
    if ((v4present || v6present) &&
        networkStartDhcpDaemon(driver, network) < 0)
2461
        goto err3;
2462

2463
    /* start radvd if there are any ipv6 addresses */
G
Gene Czarcinski 已提交
2464
    if (v6present && networkStartRadvd(driver, network) < 0)
2465 2466
        goto err4;

2467 2468 2469 2470 2471 2472 2473 2474 2475
    /* 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);
    }

2476 2477
    if (virNetDevBandwidthSet(network->def->bridge,
                              network->def->bandwidth, true) < 0) {
2478 2479 2480
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("cannot set bandwidth limits on %s"),
                       network->def->bridge);
2481 2482 2483
        goto err5;
    }

2484
    VIR_FREE(macTapIfName);
2485 2486 2487

    return 0;

2488
 err5:
2489
    virNetDevBandwidthClear(network->def->bridge);
2490

2491 2492 2493 2494
 err4:
    if (!save_err)
        save_err = virSaveLastError();

2495 2496 2497 2498 2499
    if (network->dnsmasqPid > 0) {
        kill(network->dnsmasqPid, SIGTERM);
        network->dnsmasqPid = -1;
    }

2500 2501 2502
 err3:
    if (!save_err)
        save_err = virSaveLastError();
2503
    ignore_value(virNetDevSetOnline(network->def->bridge, 0));
2504

2505 2506 2507 2508 2509 2510
 err2:
    if (!save_err)
        save_err = virSaveLastError();
    networkRemoveIptablesRules(driver, network);

 err1:
2511 2512 2513
    if (!save_err)
        save_err = virSaveLastError();

H
Hu Tao 已提交
2514
    if (macTapIfName) {
2515
        VIR_FORCE_CLOSE(tapfd);
H
Hu Tao 已提交
2516 2517 2518
        ignore_value(virNetDevTapDelete(macTapIfName));
        VIR_FREE(macTapIfName);
    }
2519 2520

 err0:
2521 2522
    if (!save_err)
        save_err = virSaveLastError();
2523
    ignore_value(virNetDevBridgeDelete(network->def->bridge));
2524

2525 2526 2527 2528
    if (save_err) {
        virSetError(save_err);
        virFreeError(save_err);
    }
2529 2530 2531
    return -1;
}

2532
static int networkShutdownNetworkVirtual(struct network_driver *driver,
2533 2534
                                        virNetworkObjPtr network)
{
2535
    virNetDevBandwidthClear(network->def->bridge);
2536

2537 2538 2539 2540 2541 2542 2543 2544
    if (network->radvdPid > 0) {
        char *radvdpidbase;

        kill(network->radvdPid, SIGTERM);
        /* attempt to delete the pidfile we created */
        if (!(radvdpidbase = networkRadvdPidfileBasename(network->def->name))) {
            virReportOOMError();
        } else {
2545
            virPidFileDelete(NETWORK_PID_DIR, radvdpidbase);
2546 2547 2548 2549
            VIR_FREE(radvdpidbase);
        }
    }

2550 2551 2552
    if (network->dnsmasqPid > 0)
        kill(network->dnsmasqPid, SIGTERM);

2553
    if (network->def->mac_specified) {
2554
        char *macTapIfName = networkBridgeDummyNicName(network->def->bridge);
2555 2556 2557
        if (!macTapIfName) {
            virReportOOMError();
        } else {
2558
            ignore_value(virNetDevTapDelete(macTapIfName));
2559 2560 2561 2562
            VIR_FREE(macTapIfName);
        }
    }

2563
    ignore_value(virNetDevSetOnline(network->def->bridge, 0));
2564

2565 2566
    networkRemoveIptablesRules(driver, network);

2567
    ignore_value(virNetDevBridgeDelete(network->def->bridge));
2568

2569
    /* See if its still alive and really really kill it */
2570
    if (network->dnsmasqPid > 0 &&
2571
        (kill(network->dnsmasqPid, 0) == 0))
2572 2573
        kill(network->dnsmasqPid, SIGKILL);
    network->dnsmasqPid = -1;
2574 2575 2576 2577 2578 2579

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

2580 2581 2582 2583 2584 2585 2586 2587
    return 0;
}

static int
networkStartNetworkExternal(struct network_driver *driver ATTRIBUTE_UNUSED,
                            virNetworkObjPtr network ATTRIBUTE_UNUSED)
{
    /* put anything here that needs to be done each time a network of
2588
     * type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is started. On
2589 2590 2591 2592 2593 2594 2595 2596 2597 2598
     * failure, undo anything you've done, and return -1. On success
     * return 0.
     */
    return 0;
}

static int networkShutdownNetworkExternal(struct network_driver *driver ATTRIBUTE_UNUSED,
                                        virNetworkObjPtr network ATTRIBUTE_UNUSED)
{
    /* put anything here that needs to be done each time a network of
2599
     * type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is shutdown. On
2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612
     * failure, undo anything you've done, and return -1. On success
     * return 0.
     */
    return 0;
}

static int
networkStartNetwork(struct network_driver *driver,
                    virNetworkObjPtr network)
{
    int ret = 0;

    if (virNetworkObjIsActive(network)) {
2613 2614
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("network is already active"));
2615 2616 2617
        return -1;
    }

2618 2619 2620
    if (virNetworkObjSetDefTransient(network, true) < 0)
        return -1;

2621
    switch (network->def->forward.type) {
2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632

    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:
2633
    case VIR_NETWORK_FORWARD_HOSTDEV:
2634 2635 2636 2637
        ret = networkStartNetworkExternal(driver, network);
        break;
    }

2638 2639
    if (ret < 0) {
        virNetworkObjUnsetDefTransient(network);
2640
        return ret;
2641
    }
2642 2643 2644 2645

    /* Persist the live configuration now that anything autogenerated
     * is setup.
     */
2646
    if ((ret = virNetworkSaveStatus(NETWORK_STATE_DIR, network)) < 0) {
2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682
        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;
}

static int networkShutdownNetwork(struct network_driver *driver,
                                        virNetworkObjPtr network)
{
    int ret = 0;
    char *stateFile;

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

    if (!virNetworkObjIsActive(network))
        return 0;

    stateFile = virNetworkConfigFile(NETWORK_STATE_DIR, network->def->name);
    if (!stateFile)
        return -1;

    unlink(stateFile);
    VIR_FREE(stateFile);

2683
    switch (network->def->forward.type) {
2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694

    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:
2695
    case VIR_NETWORK_FORWARD_HOSTDEV:
2696 2697 2698 2699
        ret = networkShutdownNetworkExternal(driver, network);
        break;
    }

2700
    network->active = 0;
2701
    virNetworkObjUnsetDefTransient(network);
2702
    return ret;
2703 2704 2705
}


2706 2707 2708 2709 2710
static virNetworkPtr networkLookupByUUID(virConnectPtr conn,
                                         const unsigned char *uuid) {
    struct network_driver *driver = conn->networkPrivateData;
    virNetworkObjPtr network;
    virNetworkPtr ret = NULL;
2711

2712
    networkDriverLock(driver);
2713
    network = virNetworkFindByUUID(&driver->networks, uuid);
2714
    networkDriverUnlock(driver);
2715
    if (!network) {
2716 2717
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
2718
        goto cleanup;
2719 2720
    }

2721 2722 2723
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

cleanup:
2724 2725
    if (network)
        virNetworkObjUnlock(network);
2726
    return ret;
2727 2728
}

2729 2730 2731 2732 2733 2734
static virNetworkPtr networkLookupByName(virConnectPtr conn,
                                         const char *name) {
    struct network_driver *driver = conn->networkPrivateData;
    virNetworkObjPtr network;
    virNetworkPtr ret = NULL;

2735
    networkDriverLock(driver);
2736
    network = virNetworkFindByName(&driver->networks, name);
2737
    networkDriverUnlock(driver);
2738
    if (!network) {
2739 2740
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"), name);
2741
        goto cleanup;
2742 2743
    }

2744 2745 2746
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

cleanup:
2747 2748
    if (network)
        virNetworkObjUnlock(network);
2749
    return ret;
2750 2751 2752 2753
}

static virDrvOpenStatus networkOpenNetwork(virConnectPtr conn,
                                           virConnectAuthPtr auth ATTRIBUTE_UNUSED,
2754 2755 2756 2757
                                           unsigned int flags)
{
    virCheckFlags(VIR_CONNECT_RO, VIR_DRV_OPEN_ERROR);

2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770
    if (!driverState)
        return VIR_DRV_OPEN_DECLINED;

    conn->networkPrivateData = driverState;
    return VIR_DRV_OPEN_SUCCESS;
}

static int networkCloseNetwork(virConnectPtr conn) {
    conn->networkPrivateData = NULL;
    return 0;
}

static int networkNumNetworks(virConnectPtr conn) {
2771
    int nactive = 0, i;
2772
    struct network_driver *driver = conn->networkPrivateData;
2773

2774 2775 2776
    networkDriverLock(driver);
    for (i = 0 ; i < driver->networks.count ; i++) {
        virNetworkObjLock(driver->networks.objs[i]);
D
Daniel P. Berrange 已提交
2777
        if (virNetworkObjIsActive(driver->networks.objs[i]))
2778
            nactive++;
2779 2780 2781
        virNetworkObjUnlock(driver->networks.objs[i]);
    }
    networkDriverUnlock(driver);
2782

2783 2784 2785 2786
    return nactive;
}

static int networkListNetworks(virConnectPtr conn, char **const names, int nnames) {
2787
    struct network_driver *driver = conn->networkPrivateData;
2788
    int got = 0, i;
2789

2790
    networkDriverLock(driver);
2791
    for (i = 0 ; i < driver->networks.count && got < nnames ; i++) {
2792
        virNetworkObjLock(driver->networks.objs[i]);
D
Daniel P. Berrange 已提交
2793
        if (virNetworkObjIsActive(driver->networks.objs[i])) {
2794
            if (!(names[got] = strdup(driver->networks.objs[i]->def->name))) {
2795
                virNetworkObjUnlock(driver->networks.objs[i]);
2796
                virReportOOMError();
2797 2798 2799 2800
                goto cleanup;
            }
            got++;
        }
2801
        virNetworkObjUnlock(driver->networks.objs[i]);
2802
    }
2803 2804
    networkDriverUnlock(driver);

2805 2806 2807
    return got;

 cleanup:
2808
    networkDriverUnlock(driver);
2809 2810 2811 2812 2813 2814
    for (i = 0 ; i < got ; i++)
        VIR_FREE(names[i]);
    return -1;
}

static int networkNumDefinedNetworks(virConnectPtr conn) {
2815
    int ninactive = 0, i;
2816
    struct network_driver *driver = conn->networkPrivateData;
2817

2818 2819 2820
    networkDriverLock(driver);
    for (i = 0 ; i < driver->networks.count ; i++) {
        virNetworkObjLock(driver->networks.objs[i]);
D
Daniel P. Berrange 已提交
2821
        if (!virNetworkObjIsActive(driver->networks.objs[i]))
2822
            ninactive++;
2823 2824 2825
        virNetworkObjUnlock(driver->networks.objs[i]);
    }
    networkDriverUnlock(driver);
2826

2827 2828 2829 2830
    return ninactive;
}

static int networkListDefinedNetworks(virConnectPtr conn, char **const names, int nnames) {
2831
    struct network_driver *driver = conn->networkPrivateData;
2832
    int got = 0, i;
2833

2834
    networkDriverLock(driver);
2835
    for (i = 0 ; i < driver->networks.count && got < nnames ; i++) {
2836
        virNetworkObjLock(driver->networks.objs[i]);
D
Daniel P. Berrange 已提交
2837
        if (!virNetworkObjIsActive(driver->networks.objs[i])) {
2838
            if (!(names[got] = strdup(driver->networks.objs[i]->def->name))) {
2839
                virNetworkObjUnlock(driver->networks.objs[i]);
2840
                virReportOOMError();
2841 2842 2843 2844
                goto cleanup;
            }
            got++;
        }
2845
        virNetworkObjUnlock(driver->networks.objs[i]);
2846
    }
2847
    networkDriverUnlock(driver);
2848 2849 2850
    return got;

 cleanup:
2851
    networkDriverUnlock(driver);
2852 2853 2854 2855 2856
    for (i = 0 ; i < got ; i++)
        VIR_FREE(names[i]);
    return -1;
}

2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872
static int
networkListAllNetworks(virConnectPtr conn,
                       virNetworkPtr **nets,
                       unsigned int flags)
{
    struct network_driver *driver = conn->networkPrivateData;
    int ret = -1;

    virCheckFlags(VIR_CONNECT_LIST_NETWORKS_FILTERS_ALL, -1);

    networkDriverLock(driver);
    ret = virNetworkList(conn, driver->networks, nets, flags);
    networkDriverUnlock(driver);

    return ret;
}
2873 2874 2875

static int networkIsActive(virNetworkPtr net)
{
2876
    struct network_driver *driver = net->conn->networkPrivateData;
2877 2878 2879 2880 2881 2882 2883
    virNetworkObjPtr obj;
    int ret = -1;

    networkDriverLock(driver);
    obj = virNetworkFindByUUID(&driver->networks, net->uuid);
    networkDriverUnlock(driver);
    if (!obj) {
2884
        virReportError(VIR_ERR_NO_NETWORK, NULL);
2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896
        goto cleanup;
    }
    ret = virNetworkObjIsActive(obj);

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

static int networkIsPersistent(virNetworkPtr net)
{
2897
    struct network_driver *driver = net->conn->networkPrivateData;
2898 2899 2900 2901 2902 2903 2904
    virNetworkObjPtr obj;
    int ret = -1;

    networkDriverLock(driver);
    obj = virNetworkFindByUUID(&driver->networks, net->uuid);
    networkDriverUnlock(driver);
    if (!obj) {
2905
        virReportError(VIR_ERR_NO_NETWORK, NULL);
2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916
        goto cleanup;
    }
    ret = obj->persistent;

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


2917
static int
2918 2919 2920
networkValidate(struct network_driver *driver,
                virNetworkDefPtr def,
                bool check_active)
2921 2922
{
    int ii;
2923 2924
    bool vlanUsed, vlanAllowed, badVlanUse = false;
    virPortGroupDefPtr defaultPortGroup = NULL;
2925
    virNetworkIpDefPtr ipdef;
G
Gene Czarcinski 已提交
2926
    bool ipv4def = false, ipv6def = false;
2927 2928 2929 2930 2931 2932 2933 2934

    /* 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
     */
2935 2936 2937
    if (def->forward.type == VIR_NETWORK_FORWARD_NONE ||
        def->forward.type == VIR_NETWORK_FORWARD_NAT ||
        def->forward.type == VIR_NETWORK_FORWARD_ROUTE) {
2938 2939 2940 2941 2942

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

        virNetworkSetBridgeMacAddr(def);
2943 2944 2945 2946 2947 2948 2949 2950 2951
    } 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,
2952
                           virNetworkForwardTypeToString(def->forward.type));
2953 2954
            return -1;
        }
2955
        if (def->dns.ntxts || def->dns.nhosts || def->dns.nsrvs) {
2956 2957 2958 2959
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <dns> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
2960
                           virNetworkForwardTypeToString(def->forward.type));
2961 2962 2963 2964 2965 2966 2967
            return -1;
        }
        if (def->domain) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <domain> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
2968
                           virNetworkForwardTypeToString(def->forward.type));
2969 2970
            return -1;
        }
2971 2972
    }

G
Gene Czarcinski 已提交
2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983
    /* We only support dhcp on one IPv4 address and
     * on one IPv6 address per defined network
     */
    for (ii = 0;
         (ipdef = virNetworkDefGetIpByIndex(def, AF_UNSPEC, ii));
         ii++) {
        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 -- "
2984 2985
                                 "dhcp is supported only for a "
                                 "single IPv4 address on each network"));
G
Gene Czarcinski 已提交
2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002
                    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;
                }
3003 3004 3005
            }
        }
    }
3006 3007 3008 3009 3010 3011

    /* 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.
     */

3012
    vlanAllowed = (def->forward.type == VIR_NETWORK_FORWARD_BRIDGE &&
3013 3014 3015 3016
                   def->virtPortProfile &&
                   def->virtPortProfile->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH);

    vlanUsed = def->vlan.nTags > 0;
3017 3018 3019 3020 3021 3022 3023 3024
    for (ii = 0; ii < def->nPortGroups; ii++) {
        if (vlanUsed || def->portGroups[ii].vlan.nTags > 0) {
            /* 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.
             */
            if (def->portGroups[ii].virtPortProfile) {
3025
                if (def->forward.type != VIR_NETWORK_FORWARD_BRIDGE ||
3026 3027 3028 3029 3030 3031 3032 3033
                    def->portGroups[ii].virtPortProfile->virtPortType
                    != VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
                    badVlanUse = true;
                }
            } else if (!vlanAllowed) {
                /* virtualport taken from base network definition */
                badVlanUse = true;
            }
3034
        }
3035 3036 3037 3038 3039 3040
        if (def->portGroups[ii].isDefault) {
            if (defaultPortGroup) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("network '%s' has multiple default "
                                 "<portgroup> elements (%s and %s), "
                                 "but only one default is allowed"),
3041
                               def->name, defaultPortGroup->name,
3042
                               def->portGroups[ii].name);
3043
                return -1;
3044
            }
3045
            defaultPortGroup = &def->portGroups[ii];
3046
        }
3047
    }
3048 3049 3050 3051 3052 3053 3054
    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.
         */
3055 3056 3057 3058 3059 3060 3061 3062 3063
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       _("<vlan> element specified for network %s, "
                         "whose type doesn't support vlan configuration"),
                       def->name);
        return -1;
    }
    return 0;
}

3064
static virNetworkPtr networkCreate(virConnectPtr conn, const char *xml) {
3065
    struct network_driver *driver = conn->networkPrivateData;
3066
    virNetworkDefPtr def;
3067
    virNetworkObjPtr network = NULL;
3068
    virNetworkPtr ret = NULL;
3069

3070 3071
    networkDriverLock(driver);

3072
    if (!(def = virNetworkDefParseString(xml)))
3073
        goto cleanup;
3074

3075
    if (networkValidate(driver, def, true) < 0)
3076 3077
       goto cleanup;

3078 3079 3080 3081
    /* NB: "live" is false because this transient network hasn't yet
     * been started
     */
    if (!(network = virNetworkAssignDef(&driver->networks, def, false)))
3082 3083
        goto cleanup;
    def = NULL;
3084

3085
    if (networkStartNetwork(driver, network) < 0) {
3086 3087
        virNetworkRemoveInactive(&driver->networks,
                                 network);
3088
        network = NULL;
3089
        goto cleanup;
3090 3091
    }

3092
    VIR_INFO("Creating network '%s'", network->def->name);
3093 3094 3095 3096
    ret = virGetNetwork(conn, network->def->name, network->def->uuid);

cleanup:
    virNetworkDefFree(def);
3097 3098 3099
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
3100
    return ret;
3101 3102 3103
}

static virNetworkPtr networkDefine(virConnectPtr conn, const char *xml) {
3104
    struct network_driver *driver = conn->networkPrivateData;
3105
    virNetworkDefPtr def = NULL;
3106
    bool freeDef = true;
3107
    virNetworkObjPtr network = NULL;
3108
    virNetworkPtr ret = NULL;
3109

3110 3111
    networkDriverLock(driver);

3112
    if (!(def = virNetworkDefParseString(xml)))
3113
        goto cleanup;
3114

3115
    if (networkValidate(driver, def, false) < 0)
3116 3117
       goto cleanup;

3118 3119 3120 3121 3122 3123 3124 3125
    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;
    }
3126

3127 3128
    /* def was asigned */
    freeDef = false;
3129 3130 3131 3132 3133 3134 3135

    if (virNetworkSaveConfig(driver->networkConfigDir, def) < 0) {
        virNetworkRemoveInactive(&driver->networks, network);
        network = NULL;
        goto cleanup;
    }

3136 3137
    VIR_INFO("Defining network '%s'", def->name);
    ret = virGetNetwork(conn, def->name, def->uuid);
3138 3139

cleanup:
3140 3141
    if (freeDef)
       virNetworkDefFree(def);
3142 3143 3144
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
3145
    return ret;
3146 3147
}

3148 3149
static int
networkUndefine(virNetworkPtr net) {
3150
    struct network_driver *driver = net->conn->networkPrivateData;
3151
    virNetworkObjPtr network;
3152
    int ret = -1;
3153
    bool active = false;
3154

3155 3156
    networkDriverLock(driver);

3157
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3158
    if (!network) {
3159 3160
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
3161
        goto cleanup;
3162 3163
    }

3164 3165
    if (virNetworkObjIsActive(network))
        active = true;
3166

3167
    if (virNetworkDeleteConfig(driver->networkConfigDir,
3168 3169
                               driver->networkAutostartDir,
                               network) < 0)
3170
        goto cleanup;
3171

3172 3173 3174 3175 3176
    /* make the network transient */
    network->persistent = 0;
    virNetworkDefFree(network->newDef);
    network->newDef = NULL;

3177
    VIR_INFO("Undefining network '%s'", network->def->name);
3178 3179 3180 3181 3182
    if (!active) {
        if (networkRemoveInactive(driver, network) < 0) {
            network = NULL;
            goto cleanup;
        }
3183
        network = NULL;
3184 3185
    }

3186
    ret = 0;
3187

3188
cleanup:
3189 3190 3191
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
3192
    return ret;
3193 3194
}

3195 3196 3197 3198 3199 3200 3201 3202 3203 3204
static int
networkUpdate(virNetworkPtr net,
              unsigned int command,
              unsigned int section,
              int parentIndex,
              const char *xml,
              unsigned int flags)
{
    struct network_driver *driver = net->conn->networkPrivateData;
    virNetworkObjPtr network = NULL;
3205 3206 3207 3208
    int isActive, ret = -1, ii;
    virNetworkIpDefPtr ipdef;
    bool oldDhcpActive = false;

3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222

    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;
    }

3223 3224 3225 3226 3227 3228 3229 3230 3231 3232
    /* see if we are listening for dhcp pre-modification */
    for (ii = 0;
         (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, ii));
         ii++) {
        if (ipdef->nranges || ipdef->nhosts) {
            oldDhcpActive = true;
            break;
        }
    }

3233 3234 3235 3236
    /* VIR_NETWORK_UPDATE_AFFECT_CURRENT means "change LIVE if network
     * is active, else change CONFIG
    */
    isActive = virNetworkObjIsActive(network);
3237 3238
    if ((flags & (VIR_NETWORK_UPDATE_AFFECT_LIVE |
                  VIR_NETWORK_UPDATE_AFFECT_CONFIG)) ==
3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270
        VIR_NETWORK_UPDATE_AFFECT_CURRENT) {
        if (isActive)
            flags |= VIR_NETWORK_UPDATE_AFFECT_LIVE;
        else
            flags |= VIR_NETWORK_UPDATE_AFFECT_CONFIG;
    }

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

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

    if (isActive && (flags & VIR_NETWORK_UPDATE_AFFECT_LIVE)) {
        /* rewrite dnsmasq host files, restart dnsmasq, update iptables
         * rules, etc, according to which section was modified. Note that
         * some sections require multiple actions, so a single switch
         * statement is inadequate.
         */
        if (section == VIR_NETWORK_SECTION_BRIDGE ||
            section == VIR_NETWORK_SECTION_DOMAIN ||
            section == VIR_NETWORK_SECTION_IP ||
            section == VIR_NETWORK_SECTION_IP_DHCP_RANGE) {
            /* these sections all change things on the dnsmasq commandline,
             * so we need to kill and restart dnsmasq.
             */
3271
            if (networkRestartDhcpDaemon(driver, network) < 0)
3272 3273
                goto cleanup;

3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291
        } 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;

            for (ii = 0;
                 (ipdef = virNetworkDefGetIpByIndex(network->def, AF_INET, ii));
                 ii++) {
                if (ipdef->nranges || ipdef->nhosts) {
                    newDhcpActive = true;
                    break;
                }
            }

            if ((newDhcpActive != oldDhcpActive &&
3292 3293
                 networkRestartDhcpDaemon(driver, network) < 0) ||
                networkRefreshDhcpDaemon(driver, network) < 0) {
3294 3295 3296 3297
                goto cleanup;
            }

        } else if (section == VIR_NETWORK_SECTION_DNS_HOST ||
3298 3299 3300 3301 3302 3303
                   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.
             */
3304
            if (networkRefreshDhcpDaemon(driver, network) < 0)
3305 3306 3307 3308 3309 3310 3311 3312
                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.
             */
3313
            if (networkRefreshRadvd(driver, network) < 0)
3314 3315 3316
                goto cleanup;
        }

3317 3318 3319
        if ((section == VIR_NETWORK_SECTION_IP ||
             section == VIR_NETWORK_SECTION_FORWARD ||
             section == VIR_NETWORK_SECTION_FORWARD_INTERFACE) &&
3320 3321 3322
           (network->def->forward.type == VIR_NETWORK_FORWARD_NONE ||
            network->def->forward.type == VIR_NETWORK_FORWARD_NAT ||
            network->def->forward.type == VIR_NETWORK_FORWARD_ROUTE)) {
3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341
            /* these could affect the iptables rules */
            networkRemoveIptablesRules(driver, network);
            if (networkAddIptablesRules(driver, network) < 0)
                goto cleanup;

        }

        /* save current network state to disk */
        if ((ret = virNetworkSaveStatus(NETWORK_STATE_DIR, network)) < 0)
            goto cleanup;
    }
    ret = 0;
cleanup:
    if (network)
        virNetworkObjUnlock(network);
    networkDriverUnlock(driver);
    return ret;
}

3342
static int networkStart(virNetworkPtr net) {
3343 3344 3345
    struct network_driver *driver = net->conn->networkPrivateData;
    virNetworkObjPtr network;
    int ret = -1;
3346

3347
    networkDriverLock(driver);
3348
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3349

3350
    if (!network) {
3351 3352
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
3353
        goto cleanup;
3354 3355
    }

3356
    ret = networkStartNetwork(driver, network);
3357 3358

cleanup:
3359 3360
    if (network)
        virNetworkObjUnlock(network);
3361
    networkDriverUnlock(driver);
3362
    return ret;
3363 3364 3365
}

static int networkDestroy(virNetworkPtr net) {
3366 3367 3368
    struct network_driver *driver = net->conn->networkPrivateData;
    virNetworkObjPtr network;
    int ret = -1;
3369

3370
    networkDriverLock(driver);
3371
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3372

3373
    if (!network) {
3374 3375
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
3376
        goto cleanup;
3377 3378
    }

D
Daniel P. Berrange 已提交
3379
    if (!virNetworkObjIsActive(network)) {
3380 3381
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("network is not active"));
3382 3383 3384
        goto cleanup;
    }

3385 3386 3387
    if ((ret = networkShutdownNetwork(driver, network)) < 0)
        goto cleanup;

3388
    if (!network->persistent) {
3389 3390 3391 3392 3393
        if (networkRemoveInactive(driver, network) < 0) {
            network = NULL;
            ret = -1;
            goto cleanup;
        }
3394 3395
        network = NULL;
    }
3396

3397
cleanup:
3398 3399
    if (network)
        virNetworkObjUnlock(network);
3400
    networkDriverUnlock(driver);
3401 3402 3403
    return ret;
}

3404
static char *networkGetXMLDesc(virNetworkPtr net,
3405
                               unsigned int flags)
3406
{
3407 3408
    struct network_driver *driver = net->conn->networkPrivateData;
    virNetworkObjPtr network;
3409
    virNetworkDefPtr def;
3410
    char *ret = NULL;
3411

3412
    virCheckFlags(VIR_NETWORK_XML_INACTIVE, NULL);
3413

3414
    networkDriverLock(driver);
3415
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3416 3417
    networkDriverUnlock(driver);

3418
    if (!network) {
3419 3420
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
3421
        goto cleanup;
3422 3423
    }

3424 3425 3426 3427 3428 3429
    if ((flags & VIR_NETWORK_XML_INACTIVE) && network->newDef)
        def = network->newDef;
    else
        def = network->def;

    ret = virNetworkDefFormat(def, flags);
3430 3431

cleanup:
3432 3433
    if (network)
        virNetworkObjUnlock(network);
3434
    return ret;
3435 3436 3437
}

static char *networkGetBridgeName(virNetworkPtr net) {
3438 3439 3440 3441
    struct network_driver *driver = net->conn->networkPrivateData;
    virNetworkObjPtr network;
    char *bridge = NULL;

3442
    networkDriverLock(driver);
3443
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3444 3445
    networkDriverUnlock(driver);

3446
    if (!network) {
3447 3448
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching id"));
3449
        goto cleanup;
3450 3451
    }

3452
    if (!(network->def->bridge)) {
3453 3454 3455
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("network '%s' does not have a bridge name."),
                       network->def->name);
3456 3457 3458
        goto cleanup;
    }

3459
    bridge = strdup(network->def->bridge);
3460
    if (!bridge)
3461
        virReportOOMError();
3462 3463

cleanup:
3464 3465
    if (network)
        virNetworkObjUnlock(network);
3466 3467 3468 3469 3470
    return bridge;
}

static int networkGetAutostart(virNetworkPtr net,
                             int *autostart) {
3471 3472 3473
    struct network_driver *driver = net->conn->networkPrivateData;
    virNetworkObjPtr network;
    int ret = -1;
3474

3475
    networkDriverLock(driver);
3476
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3477
    networkDriverUnlock(driver);
3478
    if (!network) {
3479 3480
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
3481
        goto cleanup;
3482 3483 3484
    }

    *autostart = network->autostart;
3485
    ret = 0;
3486

3487
cleanup:
3488 3489
    if (network)
        virNetworkObjUnlock(network);
3490
    return ret;
3491 3492 3493
}

static int networkSetAutostart(virNetworkPtr net,
3494
                               int autostart) {
3495 3496
    struct network_driver *driver = net->conn->networkPrivateData;
    virNetworkObjPtr network;
3497
    char *configFile = NULL, *autostartLink = NULL;
3498
    int ret = -1;
3499

3500
    networkDriverLock(driver);
3501
    network = virNetworkFindByUUID(&driver->networks, net->uuid);
3502

3503
    if (!network) {
3504 3505
        virReportError(VIR_ERR_NO_NETWORK,
                       "%s", _("no network with matching uuid"));
3506
        goto cleanup;
3507 3508
    }

3509
    if (!network->persistent) {
3510 3511
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("cannot set autostart for transient network"));
3512 3513 3514
        goto cleanup;
    }

3515 3516
    autostart = (autostart != 0);

3517
    if (network->autostart != autostart) {
3518
        if ((configFile = virNetworkConfigFile(driver->networkConfigDir, network->def->name)) == NULL)
3519
            goto cleanup;
3520
        if ((autostartLink = virNetworkConfigFile(driver->networkAutostartDir, network->def->name)) == NULL)
3521 3522
            goto cleanup;

3523
        if (autostart) {
3524
            if (virFileMakePath(driver->networkAutostartDir) < 0) {
3525
                virReportSystemError(errno,
3526 3527
                                     _("cannot create autostart directory '%s'"),
                                     driver->networkAutostartDir);
3528 3529
                goto cleanup;
            }
3530

3531
            if (symlink(configFile, autostartLink) < 0) {
3532
                virReportSystemError(errno,
3533
                                     _("Failed to create symlink '%s' to '%s'"),
3534
                                     autostartLink, configFile);
3535 3536 3537
                goto cleanup;
            }
        } else {
3538
            if (unlink(autostartLink) < 0 && errno != ENOENT && errno != ENOTDIR) {
3539
                virReportSystemError(errno,
3540
                                     _("Failed to delete symlink '%s'"),
3541
                                     autostartLink);
3542 3543
                goto cleanup;
            }
3544 3545
        }

3546
        network->autostart = autostart;
3547
    }
3548
    ret = 0;
3549

3550
cleanup:
3551 3552
    VIR_FREE(configFile);
    VIR_FREE(autostartLink);
3553 3554
    if (network)
        virNetworkObjUnlock(network);
3555
    networkDriverUnlock(driver);
3556
    return ret;
3557 3558 3559 3560 3561
}


static virNetworkDriver networkDriver = {
    "Network",
3562 3563 3564 3565 3566 3567
    .open = networkOpenNetwork, /* 0.2.0 */
    .close = networkCloseNetwork, /* 0.2.0 */
    .numOfNetworks = networkNumNetworks, /* 0.2.0 */
    .listNetworks = networkListNetworks, /* 0.2.0 */
    .numOfDefinedNetworks = networkNumDefinedNetworks, /* 0.2.0 */
    .listDefinedNetworks = networkListDefinedNetworks, /* 0.2.0 */
3568
    .listAllNetworks = networkListAllNetworks, /* 0.10.2 */
3569 3570 3571 3572 3573
    .networkLookupByUUID = networkLookupByUUID, /* 0.2.0 */
    .networkLookupByName = networkLookupByName, /* 0.2.0 */
    .networkCreateXML = networkCreate, /* 0.2.0 */
    .networkDefineXML = networkDefine, /* 0.2.0 */
    .networkUndefine = networkUndefine, /* 0.2.0 */
3574
    .networkUpdate = networkUpdate, /* 0.10.2 */
3575 3576 3577 3578 3579 3580 3581 3582
    .networkCreate = networkStart, /* 0.2.0 */
    .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 */
3583 3584 3585
};

static virStateDriver networkStateDriver = {
3586 3587 3588 3589
    .name = "Network",
    .initialize  = networkStartup,
    .cleanup = networkShutdown,
    .reload = networkReload,
3590 3591 3592 3593 3594 3595 3596
};

int networkRegister(void) {
    virRegisterNetworkDriver(&networkDriver);
    virRegisterStateDriver(&networkStateDriver);
    return 0;
}
3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607

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

/* 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.
 */

3608 3609 3610 3611 3612 3613 3614 3615 3616
/* networkCreateInterfacePool:
 * @netdef: the original NetDef from the network
 *
 * Creates an implicit interface pool of VF's when a PF dev is given
 */
static int
networkCreateInterfacePool(virNetworkDefPtr netdef) {
    unsigned int num_virt_fns = 0;
    char **vfname = NULL;
3617
    struct pci_config_address **virt_fns;
3618 3619
    int ret = -1, ii = 0;

3620
    if ((virNetDevGetVirtualFunctions(netdef->forward.pfs->dev,
3621
                                      &vfname, &virt_fns, &num_virt_fns)) < 0) {
3622 3623
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Could not get Virtual functions on %s"),
3624
                       netdef->forward.pfs->dev);
3625 3626 3627 3628 3629 3630
        goto finish;
    }

    if (num_virt_fns == 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("No Vf's present on SRIOV PF %s"),
3631
                       netdef->forward.pfs->dev);
3632 3633 3634
       goto finish;
    }

3635
    if ((VIR_ALLOC_N(netdef->forward.ifs, num_virt_fns)) < 0) {
3636 3637 3638 3639
        virReportOOMError();
        goto finish;
    }

3640
    netdef->forward.nifs = num_virt_fns;
3641

3642 3643 3644 3645 3646 3647
    for (ii = 0; ii < netdef->forward.nifs; ii++) {
        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)) {
            netdef->forward.ifs[ii].type = VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV;
3648
            if (vfname[ii]) {
3649 3650
                netdef->forward.ifs[ii].device.dev = strdup(vfname[ii]);
                if (!netdef->forward.ifs[ii].device.dev) {
3651 3652 3653 3654 3655 3656
                    virReportOOMError();
                    goto finish;
                }
            }
            else {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
3657
                               _("Direct mode types require interface names"));
3658 3659
                goto finish;
            }
3660
        }
3661
        else if (netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
3662
            /* VF's are always PCI devices */
3663 3664 3665 3666 3667
            netdef->forward.ifs[ii].type = VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI;
            netdef->forward.ifs[ii].device.pci.domain = virt_fns[ii]->domain;
            netdef->forward.ifs[ii].device.pci.bus = virt_fns[ii]->bus;
            netdef->forward.ifs[ii].device.pci.slot = virt_fns[ii]->slot;
            netdef->forward.ifs[ii].device.pci.function = virt_fns[ii]->function;
3668
        }
3669 3670 3671 3672
    }

    ret = 0;
finish:
3673
    for (ii = 0; ii < num_virt_fns; ii++) {
3674
        VIR_FREE(vfname[ii]);
3675 3676
        VIR_FREE(virt_fns[ii]);
    }
3677
    VIR_FREE(vfname);
3678
    VIR_FREE(virt_fns);
3679 3680 3681
    return ret;
}

3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696
/* 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)
{
    struct network_driver *driver = driverState;
3697 3698 3699
    enum virDomainNetType actualType = iface->type;
    virNetworkObjPtr network = NULL;
    virNetworkDefPtr netdef = NULL;
3700
    virNetDevBandwidthPtr bandwidth = NULL;
3701 3702 3703
    virPortGroupDefPtr portgroup = NULL;
    virNetDevVPortProfilePtr virtport = iface->virtPortProfile;
    virNetDevVlanPtr vlan = NULL;
3704
    virNetworkForwardIfDefPtr dev = NULL;
3705
    int ii;
3706 3707
    int ret = -1;

3708 3709 3710 3711
    /* it's handy to have this initialized if we skip directly to validate */
    if (iface->vlan.nTags > 0)
        vlan = &iface->vlan;

3712
    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
3713
        goto validate;
3714 3715 3716 3717 3718 3719 3720 3721

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

    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
    networkDriverUnlock(driver);
    if (!network) {
3722 3723 3724
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
3725
        goto error;
3726 3727
    }
    netdef = network->def;
3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738

    /* 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.
     */
3739 3740 3741 3742 3743 3744 3745

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

    if (bandwidth) {
3746 3747 3748
        if (!iface->data.network.actual
            && (VIR_ALLOC(iface->data.network.actual) < 0)) {
            virReportOOMError();
3749
            goto error;
3750 3751
        }

3752
        if (virNetDevBandwidthCopy(&iface->data.network.actual->bandwidth,
3753
                                   bandwidth) < 0)
3754
            goto error;
3755 3756
    }

3757 3758 3759
    if ((netdef->forward.type == VIR_NETWORK_FORWARD_NONE) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_NAT) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE)) {
3760 3761 3762 3763 3764 3765
        /* for these forward types, the actual net type really *is*
         *NETWORK; we just keep the info from the portgroup in
         * iface->data.network.actual
        */
        if (iface->data.network.actual)
            iface->data.network.actual->type = VIR_DOMAIN_NET_TYPE_NETWORK;
3766 3767 3768 3769

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

3770
    } else if ((netdef->forward.type == VIR_NETWORK_FORWARD_BRIDGE) &&
3771
               netdef->bridge) {
3772 3773 3774 3775 3776

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

3777 3778
        if (!iface->data.network.actual
            && (VIR_ALLOC(iface->data.network.actual) < 0)) {
3779
            virReportOOMError();
3780
            goto error;
3781 3782
        }

3783
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_BRIDGE;
3784 3785 3786
        iface->data.network.actual->data.bridge.brname = strdup(netdef->bridge);
        if (!iface->data.network.actual->data.bridge.brname) {
            virReportOOMError();
3787
            goto error;
3788 3789
        }

3790 3791 3792 3793 3794 3795 3796 3797
        /* 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) {
3798
            goto error;
3799 3800 3801 3802 3803 3804 3805 3806 3807 3808
        }
        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);
3809
                goto error;
3810 3811 3812
            }
        }

3813
    } else if (netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
3814 3815 3816 3817 3818 3819 3820 3821

        if (!iface->data.network.actual
            && (VIR_ALLOC(iface->data.network.actual) < 0)) {
            virReportOOMError();
            goto error;
        }

        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_HOSTDEV;
3822
        if (netdef->forward.npfs > 0 && netdef->forward.nifs <= 0 &&
3823 3824 3825 3826 3827
            networkCreateInterfacePool(netdef) < 0) {
            goto error;
        }

        /* pick first dev with 0 connections */
3828 3829 3830
        for (ii = 0; ii < netdef->forward.nifs; ii++) {
            if (netdef->forward.ifs[ii].connections == 0) {
                dev = &netdef->forward.ifs[ii];
3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844
                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;
3845
        iface->data.network.actual->data.hostdev.def.managed = netdef->forward.managed ? 1 : 0;
3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873
        iface->data.network.actual->data.hostdev.def.source.subsys.type = dev->type;
        iface->data.network.actual->data.hostdev.def.source.subsys.u.pci = dev->device.pci;

        /* 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;
            }
        }

3874 3875 3876 3877
    } 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)) {
3878 3879 3880 3881 3882

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

3883 3884
        if (!iface->data.network.actual
            && (VIR_ALLOC(iface->data.network.actual) < 0)) {
3885
            virReportOOMError();
3886
            goto error;
3887 3888 3889
        }

        /* Set type=direct and appropriate <source mode='xxx'/> */
3890
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_DIRECT;
3891
        switch (netdef->forward.type) {
3892
        case VIR_NETWORK_FORWARD_BRIDGE:
3893
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_BRIDGE;
3894 3895
            break;
        case VIR_NETWORK_FORWARD_PRIVATE:
3896
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PRIVATE;
3897 3898
            break;
        case VIR_NETWORK_FORWARD_VEPA:
3899
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_VEPA;
3900 3901
            break;
        case VIR_NETWORK_FORWARD_PASSTHROUGH:
3902
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PASSTHRU;
3903 3904 3905
            break;
        }

3906 3907 3908 3909 3910 3911 3912 3913
        /* 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) {
3914
            goto error;
3915
        }
3916
        virtport = iface->data.network.actual->virtPortProfile;
3917
        if (virtport) {
3918 3919 3920 3921 3922 3923 3924 3925
            /* 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);
3926
                goto error;
3927 3928
            }
        }
3929

3930 3931 3932
        /* 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).
         */
3933
        if ((netdef->forward.nifs <= 0) && (netdef->forward.npfs <= 0)) {
3934 3935 3936 3937
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' uses a direct mode, but "
                             "has no forward dev and no interface pool"),
                           netdef->name);
3938
            goto error;
3939 3940 3941
        } else {
            /* pick an interface from the pool */

3942
            if (netdef->forward.npfs > 0 && netdef->forward.nifs == 0 &&
3943 3944 3945 3946
                networkCreateInterfacePool(netdef) < 0) {
                goto error;
            }

3947 3948 3949 3950 3951
            /* 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.
3952
             */
3953 3954
            if ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
                ((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
3955 3956 3957
                 iface->data.network.actual->virtPortProfile &&
                 (iface->data.network.actual->virtPortProfile->virtPortType
                  == VIR_NETDEV_VPORT_PROFILE_8021QBH))) {
3958

3959
                /* pick first dev with 0 connections */
3960 3961 3962
                for (ii = 0; ii < netdef->forward.nifs; ii++) {
                    if (netdef->forward.ifs[ii].connections == 0) {
                        dev = &netdef->forward.ifs[ii];
3963 3964 3965 3966 3967
                        break;
                    }
                }
            } else {
                /* pick least used dev */
3968 3969 3970 3971
                dev = &netdef->forward.ifs[0];
                for (ii = 1; ii < netdef->forward.nifs; ii++) {
                    if (netdef->forward.ifs[ii].connections < dev->connections)
                        dev = &netdef->forward.ifs[ii];
3972 3973 3974 3975
                }
            }
            /* dev points at the physical device we want to use */
            if (!dev) {
3976 3977 3978 3979
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("network '%s' requires exclusive access "
                                 "to interfaces, but none are available"),
                               netdef->name);
3980
                goto error;
3981
            }
3982
            iface->data.network.actual->data.direct.linkdev = strdup(dev->device.dev);
3983 3984
            if (!iface->data.network.actual->data.direct.linkdev) {
                virReportOOMError();
3985
                goto error;
3986 3987 3988 3989
            }
        }
    }

3990
    if (virNetDevVPortProfileCheckComplete(virtport, true) < 0)
3991
        goto error;
3992

3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035
    /* 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 && netdef->vlan.nTags > 0)
        vlan = &netdef->vlan;

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

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.
     */

    if (vlan) {
        /* 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;
        }
    }

4036 4037
    if (dev) {
        /* we are now assured of success, so mark the allocation */
4038
        dev->connections++;
4039 4040 4041 4042 4043 4044 4045 4046 4047
        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);
        }
4048
    }
4049

4050 4051 4052 4053 4054
    if (netdef) {
        netdef->connections++;
        VIR_DEBUG("Using network %s, %d connections",
                  netdef->name, netdef->connections);
    }
4055
    ret = 0;
4056

4057 4058 4059
cleanup:
    if (network)
        virNetworkObjUnlock(network);
4060 4061 4062 4063
    return ret;

error:
    if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
4064 4065 4066
        virDomainActualNetDefFree(iface->data.network.actual);
        iface->data.network.actual = NULL;
    }
4067
    goto cleanup;
4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083
}

/* 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)
{
    struct network_driver *driver = driverState;
4084
    enum virDomainNetType actualType = virDomainNetGetActualType(iface);
4085 4086
    virNetworkObjPtr network;
    virNetworkDefPtr netdef;
4087 4088
    virNetworkForwardIfDefPtr dev = NULL;
    int ii, ret = -1;
4089 4090 4091 4092 4093 4094 4095 4096

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

    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
    networkDriverUnlock(driver);
    if (!network) {
4097 4098 4099
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
4100 4101 4102 4103 4104
        goto error;
    }
    netdef = network->def;

    if (!iface->data.network.actual ||
4105 4106
        (actualType != VIR_DOMAIN_NET_TYPE_DIRECT &&
         actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV)) {
4107 4108
        VIR_DEBUG("Nothing to claim from network %s", iface->data.network.name);
        goto success;
4109 4110
    }

4111
    if (netdef->forward.npfs > 0 && netdef->forward.nifs == 0 &&
4112
        networkCreateInterfacePool(netdef) < 0) {
4113
        goto error;
4114
    }
4115
    if (netdef->forward.nifs == 0) {
4116
        virReportError(VIR_ERR_INTERNAL_ERROR,
4117 4118
                       _("network '%s' uses a direct or hostdev mode, "
                         "but has no forward dev and no interface pool"),
4119
                       netdef->name);
4120
        goto error;
4121
    }
4122

4123 4124
    if (actualType == VIR_DOMAIN_NET_TYPE_DIRECT) {
        const char *actualDev;
4125

4126 4127 4128 4129 4130 4131 4132 4133 4134
        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 */
4135 4136
        for (ii = 0; ii < netdef->forward.nifs; ii++) {
            if (netdef->forward.ifs[ii].type
4137
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
4138 4139
                STREQ(actualDev, netdef->forward.ifs[ii].device.dev)) {
                dev = &netdef->forward.ifs[ii];
4140 4141 4142 4143 4144
                break;
            }
        }
        /* dev points at the physical device we want to use */
        if (!dev) {
4145
            virReportError(VIR_ERR_INTERNAL_ERROR,
4146 4147
                           _("network '%s' doesn't have dev='%s' "
                             "in use by domain"),
4148
                           netdef->name, actualDev);
4149
            goto error;
4150 4151
        }

4152
        /* PASSTHROUGH mode and PRIVATE Mode + 802.1Qbh both require
4153 4154
         * exclusive access to a device, so current connections count
         * must be 0 in those cases.
4155
         */
4156
        if ((dev->connections > 0) &&
4157 4158
            ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
             ((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
4159 4160
              iface->data.network.actual->virtPortProfile &&
              (iface->data.network.actual->virtPortProfile->virtPortType
4161
               == VIR_NETDEV_VPORT_PROFILE_8021QBH)))) {
4162
            virReportError(VIR_ERR_INTERNAL_ERROR,
4163 4164
                           _("network '%s' claims dev='%s' is already in "
                             "use by a different domain"),
4165
                           netdef->name, actualDev);
4166
            goto error;
4167
        }
4168

4169
        /* we are now assured of success, so mark the allocation */
4170
        dev->connections++;
4171
        VIR_DEBUG("Using physical device %s, connections %d",
4172
                  dev->device.dev, dev->connections);
4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185

    }  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 */
4186 4187
        for (ii = 0; ii < netdef->forward.nifs; ii++) {
            if (netdef->forward.ifs[ii].type
4188
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
4189
                virDevicePCIAddressEqual(&hostdev->source.subsys.u.pci,
4190 4191
                                         &netdef->forward.ifs[ii].device.pci)) {
                dev = &netdef->forward.ifs[ii];
4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212
                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,
                           hostdev->source.subsys.u.pci.domain,
                           hostdev->source.subsys.u.pci.bus,
                           hostdev->source.subsys.u.pci.slot,
                           hostdev->source.subsys.u.pci.function);
                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) &&
4213
            netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229
            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);
4230 4231
    }

4232
success:
4233 4234 4235
    netdef->connections++;
    VIR_DEBUG("Using network %s, %d connections",
              netdef->name, netdef->connections);
4236 4237 4238 4239 4240
    ret = 0;
cleanup:
    if (network)
        virNetworkObjUnlock(network);
    return ret;
4241 4242 4243

error:
    goto cleanup;
4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260
}


/* 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)
{
    struct network_driver *driver = driverState;
4261
    enum virDomainNetType actualType = virDomainNetGetActualType(iface);
4262
    virNetworkObjPtr network;
4263
    virNetworkDefPtr netdef;
4264 4265
    virNetworkForwardIfDefPtr dev = NULL;
    int ii, ret = -1;
4266 4267 4268 4269 4270 4271 4272 4273

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

    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, iface->data.network.name);
    networkDriverUnlock(driver);
    if (!network) {
4274 4275 4276
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
4277 4278 4279 4280
        goto error;
    }
    netdef = network->def;

4281 4282 4283 4284 4285 4286
    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;

4287 4288 4289
    if ((!iface->data.network.actual) ||
        ((actualType != VIR_DOMAIN_NET_TYPE_DIRECT) &&
         (actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV))) {
4290 4291
        VIR_DEBUG("Nothing to release to network %s", iface->data.network.name);
        goto success;
4292 4293
    }

4294
    if (netdef->forward.nifs == 0) {
4295
        virReportError(VIR_ERR_INTERNAL_ERROR,
4296
                       _("network '%s' uses a direct/hostdev mode, but "
4297 4298
                         "has no forward dev and no interface pool"),
                       netdef->name);
4299
        goto error;
4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311
    }

    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;
        }
4312

4313 4314
        for (ii = 0; ii < netdef->forward.nifs; ii++) {
            if (netdef->forward.ifs[ii].type
4315
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
4316 4317
                STREQ(actualDev, netdef->forward.ifs[ii].device.dev)) {
                dev = &netdef->forward.ifs[ii];
4318 4319 4320
                break;
            }
        }
4321

4322
        if (!dev) {
4323
            virReportError(VIR_ERR_INTERNAL_ERROR,
4324 4325
                           _("network '%s' doesn't have dev='%s' "
                             "in use by domain"),
4326
                           netdef->name, actualDev);
4327
            goto error;
4328 4329
        }

4330
        dev->connections--;
4331
        VIR_DEBUG("Releasing physical device %s, connections %d",
4332
                  dev->device.dev, dev->connections);
4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343

    } 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;
        }

4344 4345
        for (ii = 0; ii < netdef->forward.nifs; ii++) {
            if (netdef->forward.ifs[ii].type
4346
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
4347
                virDevicePCIAddressEqual(&hostdev->source.subsys.u.pci,
4348 4349
                                          &netdef->forward.ifs[ii].device.pci)) {
                dev = &netdef->forward.ifs[ii];
4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371
                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,
                           hostdev->source.subsys.u.pci.domain,
                           hostdev->source.subsys.u.pci.bus,
                           hostdev->source.subsys.u.pci.slot,
                           hostdev->source.subsys.u.pci.function);
                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);
   }
4372

4373
success:
4374 4375 4376
    netdef->connections--;
    VIR_DEBUG("Releasing network %s, %d connections",
              netdef->name, netdef->connections);
4377 4378 4379 4380
    ret = 0;
cleanup:
    if (network)
        virNetworkObjUnlock(network);
4381 4382 4383 4384
    if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
        virDomainActualNetDefFree(iface->data.network.actual);
        iface->data.network.actual = NULL;
    }
4385
    return ret;
4386 4387 4388

error:
    goto cleanup;
4389
}
4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414

/*
 * 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;
    struct network_driver *driver = driverState;
4415
    virNetworkObjPtr network;
4416 4417 4418 4419
    virNetworkDefPtr netdef;
    virNetworkIpDefPtr ipdef;
    virSocketAddr addr;
    virSocketAddrPtr addrptr = NULL;
4420
    char *dev_name = NULL;
4421 4422 4423 4424 4425 4426

    *netaddr = NULL;
    networkDriverLock(driver);
    network = virNetworkFindByName(&driver->networks, netname);
    networkDriverUnlock(driver);
    if (!network) {
4427 4428 4429
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       netname);
4430
        goto error;
4431 4432 4433
    }
    netdef = network->def;

4434
    switch (netdef->forward.type) {
4435 4436 4437 4438 4439 4440
    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) {
4441 4442 4443
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' doesn't have an IPv4 address"),
                           netdef->name);
4444 4445 4446 4447 4448 4449
            break;
        }
        addrptr = &ipdef->address;
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
4450
        if ((dev_name = netdef->bridge))
4451 4452 4453 4454 4455 4456 4457 4458
            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:
4459 4460
        if ((netdef->forward.nifs > 0) && netdef->forward.ifs)
            dev_name = netdef->forward.ifs[0].device.dev;
4461

4462
        if (!dev_name) {
4463 4464 4465
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' has no associated interface or bridge"),
                           netdef->name);
4466 4467 4468 4469
        }
        break;
    }

4470
    if (dev_name) {
4471
        if (virNetDevGetIPv4Address(dev_name, &addr) < 0)
4472
            goto error;
4473
        addrptr = &addr;
4474 4475
    }

4476 4477 4478
    if (!(addrptr &&
          (*netaddr = virSocketAddrFormat(addrptr)))) {
        goto error;
4479 4480
    }

4481
    ret = 0;
4482 4483 4484 4485
cleanup:
    if (network)
        virNetworkObjUnlock(network);
    return ret;
4486 4487 4488

error:
    goto cleanup;
4489
}
4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628

/**
 * 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;
    virNetDevBandwidthPtr ifaceBand = iface->bandwidth;
    unsigned long long tmp_floor_sum = net->floor_sum;
    unsigned long long tmp_new_rate = 0;
    char ifmac[VIR_MAC_STRING_BUFLEN];

    if (!ifaceBand || !ifaceBand->in || !ifaceBand->in->floor ||
        !netBand || !netBand->in)
        return 1;

    virMacAddrFormat(&iface->mac, ifmac);

    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];

    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;
    }

    plug_ret = virNetDevBandwidthPlug(net->def->bridge,
                                      net->def->bandwidth,
                                      &iface->mac,
                                      iface->bandwidth,
                                      class_id);
    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 */
    net->floor_sum += iface->bandwidth->in->floor;
4629 4630 4631 4632 4633 4634 4635 4636
    /* update status file */
    if (virNetworkSaveStatus(NETWORK_STATE_DIR, net) < 0) {
        ignore_value(virBitmapClearBit(net->class_id, class_id));
        net->floor_sum -= iface->bandwidth->in->floor;
        iface->data.network.actual->class_id = 0;
        ignore_value(virNetDevBandwidthUnplug(net->def->bridge, class_id));
        goto cleanup;
    }
4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670
    /* 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;

    if (iface->data.network.actual &&
        iface->data.network.actual->class_id) {
        /* 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 */
        net->floor_sum -= iface->bandwidth->in->floor;
4671 4672 4673 4674 4675 4676 4677 4678 4679 4680
        /* return class ID */
        ignore_value(virBitmapClearBit(net->class_id,
                                       iface->data.network.actual->class_id));
        /* update status file */
        if (virNetworkSaveStatus(NETWORK_STATE_DIR, net) < 0) {
            net->floor_sum += iface->bandwidth->in->floor;
            ignore_value(virBitmapSetBit(net->class_id,
                                         iface->data.network.actual->class_id));
            goto cleanup;
        }
4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693
        /* 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;
}