bridge_driver.c 184.2 KB
Newer Older
1
/*
2
 * bridge_driver.c: core driver methods for managing network
3
 *
4
 * Copyright (C) 2006-2016 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16 17
 * Copyright (C) 2006 Daniel P. Berrange
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with this library.  If not, see
O
Osier Yang 已提交
19
 * <http://www.gnu.org/licenses/>.
20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41
 *
 * 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 <pwd.h>
#include <sys/wait.h>
#include <sys/ioctl.h>
42
#include <net/if.h>
43
#include <dirent.h>
44 45 46
#if HAVE_SYS_SYSCTL_H
# include <sys/sysctl.h>
#endif
47

48
#include "virerror.h"
49
#include "datatypes.h"
50
#include "bridge_driver.h"
51
#include "bridge_driver_platform.h"
52
#include "device_conf.h"
53
#include "driver.h"
54
#include "virbuffer.h"
55
#include "virpidfile.h"
56
#include "vircommand.h"
57
#include "viralloc.h"
58
#include "viruuid.h"
59
#include "viriptables.h"
60
#include "virlog.h"
61
#include "virdnsmasq.h"
62
#include "configmake.h"
63
#include "virnetlink.h"
64
#include "virnetdev.h"
65
#include "virnetdevip.h"
66 67
#include "virnetdevbridge.h"
#include "virnetdevtap.h"
68
#include "virnetdevvportprofile.h"
69
#include "virpci.h"
70
#include "virdbus.h"
71
#include "virfile.h"
72
#include "virstring.h"
73
#include "viraccessapicheck.h"
74
#include "network_event.h"
75
#include "virhook.h"
76
#include "virjson.h"
77

78
#define VIR_FROM_THIS VIR_FROM_NETWORK
79
#define MAX_BRIDGE_ID 256
80

81 82 83 84 85 86 87
/**
 * VIR_NETWORK_DHCP_LEASE_FILE_SIZE_MAX:
 *
 * Macro providing the upper limit on the size of leases file
 */
#define VIR_NETWORK_DHCP_LEASE_FILE_SIZE_MAX (32 * 1024 * 1024)

88 89
#define SYSCTL_PATH "/proc/sys"

90 91
VIR_LOG_INIT("network.bridge_driver");

92
static virNetworkDriverStatePtr network_driver;
93

94

95 96 97 98 99 100 101 102
static virNetworkDriverStatePtr
networkGetDriver(void)
{
    /* Maybe one day we can store @network_driver in the
     * connection object, but until then, it's just a global
     * variable which is returned. */
    return network_driver;
}
103

104 105 106

static void
networkDriverLock(virNetworkDriverStatePtr driver)
107
{
108
    virMutexLock(&driver->lock);
109
}
110 111 112 113


static void
networkDriverUnlock(virNetworkDriverStatePtr driver)
114
{
115
    virMutexUnlock(&driver->lock);
116 117
}

118

119 120 121 122
static dnsmasqCapsPtr
networkGetDnsmasqCaps(virNetworkDriverStatePtr driver)
{
    dnsmasqCapsPtr ret;
123
    networkDriverLock(driver);
124
    ret = virObjectRef(driver->dnsmasqCaps);
125
    networkDriverUnlock(driver);
126 127 128
    return ret;
}

129

130 131 132 133 134 135 136 137
static int
networkDnsmasqCapsRefresh(virNetworkDriverStatePtr driver)
{
    dnsmasqCapsPtr caps;

    if (!(caps = dnsmasqCapsNewFromBinary(DNSMASQ)))
        return -1;

138
    networkDriverLock(driver);
139 140
    virObjectUnref(driver->dnsmasqCaps);
    driver->dnsmasqCaps = caps;
141
    networkDriverUnlock(driver);
142 143 144
    return 0;
}

145

146 147 148 149 150
static int
networkStateCleanup(void);

static int
networkStartNetwork(virNetworkDriverStatePtr driver,
151
                    virNetworkObjPtr obj);
152 153 154

static int
networkShutdownNetwork(virNetworkDriverStatePtr driver,
155
                       virNetworkObjPtr obj);
156 157 158

static int
networkStartNetworkVirtual(virNetworkDriverStatePtr driver,
159
                           virNetworkObjPtr obj);
160 161 162

static int
networkShutdownNetworkVirtual(virNetworkDriverStatePtr driver,
163
                              virNetworkObjPtr obj);
164

165
static int
166
networkStartNetworkExternal(virNetworkObjPtr obj);
167

168
static int
169
networkShutdownNetworkExternal(virNetworkObjPtr obj);
170

171 172
static void
networkReloadFirewallRules(virNetworkDriverStatePtr driver);
173

174 175
static void
networkRefreshDaemons(virNetworkDriverStatePtr driver);
176

177
static int
178
networkPlugBandwidth(virNetworkObjPtr obj,
179
                     virDomainNetDefPtr iface);
180

181
static int
182
networkUnplugBandwidth(virNetworkObjPtr obj,
183
                       virDomainNetDefPtr iface);
184

185
static void
186
networkNetworkObjTaint(virNetworkObjPtr obj,
187
                       virNetworkTaintFlags taint);
188

189

190 191 192
static virNetworkObjPtr
networkObjFromNetwork(virNetworkPtr net)
{
193
    virNetworkDriverStatePtr driver = networkGetDriver();
194
    virNetworkObjPtr obj;
195 196
    char uuidstr[VIR_UUID_STRING_BUFLEN];

197 198
    obj = virNetworkObjFindByUUID(driver->networks, net->uuid);
    if (!obj) {
199 200 201 202 203 204
        virUUIDFormat(net->uuid, uuidstr);
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching uuid '%s' (%s)"),
                       uuidstr, net->name);
    }

205
    return obj;
206 207
}

208

209
static int
210
networkRunHook(virNetworkObjPtr obj,
211
               virDomainDefPtr dom,
212
               virDomainNetDefPtr iface,
213 214 215 216 217 218 219 220 221
               int op,
               int sub_op)
{
    virBuffer buf = VIR_BUFFER_INITIALIZER;
    char *xml = NULL, *net_xml = NULL, *dom_xml = NULL;
    int hookret;
    int ret = -1;

    if (virHookPresent(VIR_HOOK_DRIVER_NETWORK)) {
222 223
        if (!obj) {
            VIR_DEBUG("Not running hook as @obj is NULL");
224 225 226 227
            ret = 0;
            goto cleanup;
        }

228 229
        virBufferAddLit(&buf, "<hookData>\n");
        virBufferAdjustIndent(&buf, 2);
230
        if (iface && virDomainNetDefFormat(&buf, iface, NULL, 0) < 0)
231
            goto cleanup;
232
        if (virNetworkDefFormatBuf(&buf, obj->def, 0) < 0)
233
            goto cleanup;
234
        if (dom && virDomainDefFormatInternal(dom, NULL, 0, &buf) < 0)
235 236 237 238 239
            goto cleanup;

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

240
        if (virBufferCheckError(&buf) < 0)
241 242
            goto cleanup;

243
        xml = virBufferContentAndReset(&buf);
244
        hookret = virHookCall(VIR_HOOK_DRIVER_NETWORK, obj->def->name,
245 246 247 248 249 250 251
                              op, sub_op, NULL, xml, NULL);

        /*
         * If the script raised an error, pass it to the callee.
         */
        if (hookret < 0)
            goto cleanup;
252

253
        networkNetworkObjTaint(obj, VIR_NETWORK_TAINT_HOOK);
254 255 256
    }

    ret = 0;
257
 cleanup:
258 259 260 261 262 263 264
    virBufferFreeAndReset(&buf);
    VIR_FREE(xml);
    VIR_FREE(net_xml);
    VIR_FREE(dom_xml);
    return ret;
}

265

266
static char *
267 268
networkDnsmasqLeaseFileNameDefault(virNetworkDriverStatePtr driver,
                                   const char *netname)
269 270 271
{
    char *leasefile;

272
    ignore_value(virAsprintf(&leasefile, "%s/%s.leases",
273
                             driver->dnsmasqStateDir, netname));
274 275 276
    return leasefile;
}

277

278
static char *
279 280
networkDnsmasqLeaseFileNameCustom(virNetworkDriverStatePtr driver,
                                  const char *bridge)
281 282 283 284
{
    char *leasefile;

    ignore_value(virAsprintf(&leasefile, "%s/%s.status",
285
                             driver->dnsmasqStateDir, bridge));
286 287 288
    return leasefile;
}

289

290
static char *
291 292
networkDnsmasqConfigFileName(virNetworkDriverStatePtr driver,
                             const char *netname)
293 294 295
{
    char *conffile;

296
    ignore_value(virAsprintf(&conffile, "%s/%s.conf",
297
                             driver->dnsmasqStateDir, netname));
298 299 300
    return conffile;
}

301

302 303 304 305 306 307
static char *
networkRadvdPidfileBasename(const char *netname)
{
    /* this is simple but we want to be sure it's consistently done */
    char *pidfilebase;

308
    ignore_value(virAsprintf(&pidfilebase, "%s-radvd", netname));
309 310 311
    return pidfilebase;
}

312

313
static char *
314 315
networkRadvdConfigFileName(virNetworkDriverStatePtr driver,
                           const char *netname)
316 317 318
{
    char *configfile;

319
    ignore_value(virAsprintf(&configfile, "%s/%s-radvd.conf",
320
                             driver->radvdStateDir, netname));
321 322
    return configfile;
}
323

324

325 326
/* do needed cleanup steps and remove the network from the list */
static int
327
networkRemoveInactive(virNetworkDriverStatePtr driver,
328
                      virNetworkObjPtr obj)
329 330
{
    char *leasefile = NULL;
331
    char *customleasefile = NULL;
332
    char *radvdconfigfile = NULL;
333
    char *configfile = NULL;
334
    char *radvdpidbase = NULL;
335
    char *statusfile = NULL;
M
Michal Privoznik 已提交
336
    char *macMapFile = NULL;
337
    dnsmasqContext *dctx = NULL;
338
    virNetworkDefPtr def = virNetworkObjGetPersistentDef(obj);
339 340 341 342

    int ret = -1;

    /* remove the (possibly) existing dnsmasq and radvd files */
343
    if (!(dctx = dnsmasqContextNew(def->name,
344
                                   driver->dnsmasqStateDir))) {
345
        goto cleanup;
346
    }
347

348
    if (!(leasefile = networkDnsmasqLeaseFileNameDefault(driver, def->name)))
349 350
        goto cleanup;

351
    if (!(customleasefile = networkDnsmasqLeaseFileNameCustom(driver, def->bridge)))
352 353
        goto cleanup;

354
    if (!(radvdconfigfile = networkRadvdConfigFileName(driver, def->name)))
355
        goto cleanup;
356 357

    if (!(radvdpidbase = networkRadvdPidfileBasename(def->name)))
358
        goto cleanup;
359

360
    if (!(configfile = networkDnsmasqConfigFileName(driver, def->name)))
361
        goto cleanup;
362

363
    if (!(statusfile = virNetworkConfigFile(driver->stateDir, def->name)))
364
        goto cleanup;
365

366
    if (!(macMapFile = virMacMapFileName(driver->dnsmasqStateDir, def->bridge)))
M
Michal Privoznik 已提交
367 368
        goto cleanup;

369 370 371
    /* dnsmasq */
    dnsmasqDelete(dctx);
    unlink(leasefile);
372
    unlink(customleasefile);
373
    unlink(configfile);
374

M
Michal Privoznik 已提交
375 376 377
    /* MAC map manager */
    unlink(macMapFile);

378 379
    /* radvd */
    unlink(radvdconfigfile);
380
    virPidFileDelete(driver->pidDir, radvdpidbase);
381

382 383 384
    /* remove status file */
    unlink(statusfile);

385
    /* remove the network definition */
386
    virNetworkObjRemoveInactive(driver->networks, obj);
387 388 389

    ret = 0;

390
 cleanup:
391
    VIR_FREE(leasefile);
392
    VIR_FREE(configfile);
393
    VIR_FREE(customleasefile);
394 395
    VIR_FREE(radvdconfigfile);
    VIR_FREE(radvdpidbase);
396
    VIR_FREE(statusfile);
M
Michal Privoznik 已提交
397
    VIR_FREE(macMapFile);
398 399 400 401
    dnsmasqContextFree(dctx);
    return ret;
}

402

M
Michal Privoznik 已提交
403 404
static int
networkMacMgrAdd(virNetworkDriverStatePtr driver,
405
                 virNetworkObjPtr obj,
M
Michal Privoznik 已提交
406 407 408 409 410 411 412
                 const char *domain,
                 const virMacAddr *mac)
{
    char macStr[VIR_MAC_STRING_BUFLEN];
    char *file = NULL;
    int ret = -1;

413
    if (!obj->macmap)
M
Michal Privoznik 已提交
414 415 416 417
        return 0;

    virMacAddrFormat(mac, macStr);

418
    if (!(file = virMacMapFileName(driver->dnsmasqStateDir, obj->def->bridge)))
M
Michal Privoznik 已提交
419 420
        goto cleanup;

421
    if (virMacMapAdd(obj->macmap, domain, macStr) < 0)
M
Michal Privoznik 已提交
422 423
        goto cleanup;

424
    if (virMacMapWriteFile(obj->macmap, file) < 0)
M
Michal Privoznik 已提交
425 426 427 428 429 430 431 432
        goto cleanup;

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

433

M
Michal Privoznik 已提交
434 435
static int
networkMacMgrDel(virNetworkDriverStatePtr driver,
436
                 virNetworkObjPtr obj,
M
Michal Privoznik 已提交
437 438 439 440 441 442 443
                 const char *domain,
                 const virMacAddr *mac)
{
    char macStr[VIR_MAC_STRING_BUFLEN];
    char *file = NULL;
    int ret = -1;

444
    if (!obj->macmap)
M
Michal Privoznik 已提交
445 446 447 448
        return 0;

    virMacAddrFormat(mac, macStr);

449
    if (!(file = virMacMapFileName(driver->dnsmasqStateDir, obj->def->bridge)))
M
Michal Privoznik 已提交
450 451
        goto cleanup;

452
    if (virMacMapRemove(obj->macmap, domain, macStr) < 0)
M
Michal Privoznik 已提交
453 454
        goto cleanup;

455
    if (virMacMapWriteFile(obj->macmap, file) < 0)
M
Michal Privoznik 已提交
456 457 458 459 460 461 462 463
        goto cleanup;

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

464

465 466 467
static char *
networkBridgeDummyNicName(const char *brname)
{
468
    static const char dummyNicSuffix[] = "-nic";
469 470
    char *nicname;

471 472 473 474 475 476 477
    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.
         */
478 479 480 481 482
        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));
483
    } else {
484
        ignore_value(virAsprintf(&nicname, "%s%s", brname, dummyNicSuffix));
485
    }
486 487 488
    return nicname;
}

489

490 491
static int
networkUpdateState(virNetworkObjPtr obj,
492
                   void *opaque)
493
{
494
    virNetworkDriverStatePtr driver = opaque;
495
    dnsmasqCapsPtr dnsmasq_caps = networkGetDnsmasqCaps(driver);
496
    char *macMapFile = NULL;
497
    int ret = -1;
498

499
    virObjectLock(obj);
500
    if (!virNetworkObjIsActive(obj)) {
501 502
        ret = 0;
        goto cleanup;
503
    }
504

505 506 507 508
    switch (obj->def->forward.type) {
    case VIR_NETWORK_FORWARD_NONE:
    case VIR_NETWORK_FORWARD_NAT:
    case VIR_NETWORK_FORWARD_ROUTE:
509
    case VIR_NETWORK_FORWARD_OPEN:
510 511 512
        /* If bridge doesn't exist, then mark it inactive */
        if (!(obj->def->bridge && virNetDevExists(obj->def->bridge) == 1))
            obj->active = 0;
513

514 515
        if (!(macMapFile = virMacMapFileName(driver->dnsmasqStateDir,
                                             obj->def->bridge)))
516 517 518 519 520
            goto cleanup;

        if (!(obj->macmap = virMacMapNew(macMapFile)))
            goto cleanup;

521
        break;
522

523
    case VIR_NETWORK_FORWARD_BRIDGE:
524 525
        if (obj->def->bridge) {
            if (virNetDevExists(obj->def->bridge) != 1)
526 527
                obj->active = 0;
            break;
528
        }
529 530 531 532 533 534 535 536 537
        /* intentionally drop through to common case for all
         * macvtap networks (forward='bridge' with no bridge
         * device defined is macvtap using its 'bridge' mode)
         */
    case VIR_NETWORK_FORWARD_PRIVATE:
    case VIR_NETWORK_FORWARD_VEPA:
    case VIR_NETWORK_FORWARD_PASSTHROUGH:
        /* so far no extra checks */
        break;
538

539 540 541
    case VIR_NETWORK_FORWARD_HOSTDEV:
        /* so far no extra checks */
        break;
542
    }
543

544 545 546
    /* Try and read dnsmasq/radvd pids of active networks */
    if (obj->active && obj->def->ips && (obj->def->nips > 0)) {
        char *radvdpidbase;
547

548 549 550
        ignore_value(virPidFileReadIfAlive(driver->pidDir,
                                           obj->def->name,
                                           &obj->dnsmasqPid,
551
                                           dnsmasqCapsGetBinaryPath(dnsmasq_caps)));
M
Michal Privoznik 已提交
552

553 554 555
        radvdpidbase = networkRadvdPidfileBasename(obj->def->name);
        if (!radvdpidbase)
            goto cleanup;
556

557 558 559 560
        ignore_value(virPidFileReadIfAlive(driver->pidDir,
                                           radvdpidbase,
                                           &obj->radvdPid, RADVD));
        VIR_FREE(radvdpidbase);
561
    }
562

563 564
    ret = 0;
 cleanup:
565
    virObjectUnlock(obj);
566
    virObjectUnref(dnsmasq_caps);
567
    VIR_FREE(macMapFile);
568 569
    return ret;
}
570

571

572
static int
573
networkAutostartConfig(virNetworkObjPtr obj,
574
                       void *opaque)
575
{
576
    virNetworkDriverStatePtr driver = opaque;
577
    int ret = -1;
578

579 580 581 582
    virObjectLock(obj);
    if (obj->autostart &&
        !virNetworkObjIsActive(obj) &&
        networkStartNetwork(driver, obj) < 0)
583 584 585 586
        goto cleanup;

    ret = 0;
 cleanup:
587
    virObjectUnlock(obj);
588
    return ret;
589 590
}

591

592 593 594
#if HAVE_FIREWALLD
static DBusHandlerResult
firewalld_dbus_filter_bridge(DBusConnection *connection ATTRIBUTE_UNUSED,
595 596
                             DBusMessage *message,
                             void *user_data)
597
{
598 599
    virNetworkDriverStatePtr driver = user_data;

600 601 602 603 604 605
    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.");
606
        networkReloadFirewallRules(driver);
607 608 609 610 611 612
    }

    return DBUS_HANDLER_RESULT_NOT_YET_HANDLED;
}
#endif

613

614
static int
615
networkMigrateStateFiles(virNetworkDriverStatePtr driver)
616 617 618 619 620 621 622 623 624 625 626 627 628
{
    /* Due to a change in location of network state xml beginning in
     * libvirt 1.2.4 (from /var/lib/libvirt/network to
     * /var/run/libvirt/network), we must check for state files in two
     * locations. Anything found in the old location must be written
     * to the new location, then erased from the old location. (Note
     * that we read/write the file rather than calling rename()
     * because the old and new state directories are likely in
     * different filesystems).
     */
    int ret = -1;
    const char *oldStateDir = LOCALSTATEDIR "/lib/libvirt/network";
    DIR *dir;
629
    int direrr;
630 631 632
    struct dirent *entry;
    char *oldPath = NULL, *newPath = NULL;
    char *contents = NULL;
J
Ján Tomko 已提交
633
    int rc;
634

J
Ján Tomko 已提交
635 636
    if ((rc = virDirOpenIfExists(&dir, oldStateDir)) <= 0)
        return rc;
637 638 639 640 641 642 643

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

644
    while ((direrr = virDirRead(dir, &entry, oldStateDir)) > 0) {
645 646 647
        if (entry->d_type != DT_UNKNOWN &&
            entry->d_type != DT_REG)
            continue;
648 649 650

        if (virAsprintf(&oldPath, "%s/%s",
                        oldStateDir, entry->d_name) < 0)
J
Ján Tomko 已提交
651
            goto cleanup;
652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668

        if (entry->d_type == DT_UNKNOWN) {
            struct stat st;

            if (lstat(oldPath, &st) < 0) {
                virReportSystemError(errno,
                                     _("failed to stat network status file '%s'"),
                                     oldPath);
                goto cleanup;
            }

            if (!S_ISREG(st.st_mode)) {
                VIR_FREE(oldPath);
                continue;
            }
        }

669
        if (virFileReadAll(oldPath, 1024*1024, &contents) < 0)
J
Ján Tomko 已提交
670
            goto cleanup;
671 672 673

        if (virAsprintf(&newPath, "%s/%s",
                        driver->stateDir, entry->d_name) < 0)
J
Ján Tomko 已提交
674
            goto cleanup;
675 676 677 678 679 680 681 682 683 684 685 686
        if (virFileWriteStr(newPath, contents, S_IRUSR | S_IWUSR) < 0) {
            virReportSystemError(errno,
                                 _("failed to write network status file '%s'"),
                                 newPath);
            goto cleanup;
        }

        unlink(oldPath);
        VIR_FREE(oldPath);
        VIR_FREE(newPath);
        VIR_FREE(contents);
    }
687
    if (direrr < 0)
J
Ján Tomko 已提交
688
        goto cleanup;
689 690 691

    ret = 0;
 cleanup:
J
Ján Tomko 已提交
692
    VIR_DIR_CLOSE(dir);
693 694 695 696 697 698
    VIR_FREE(oldPath);
    VIR_FREE(newPath);
    VIR_FREE(contents);
    return ret;
}

699

700
/**
701
 * networkStateInitialize:
702 703 704 705
 *
 * Initialization function for the QEmu daemon
 */
static int
706 707 708
networkStateInitialize(bool privileged,
                       virStateInhibitCallback callback ATTRIBUTE_UNUSED,
                       void *opaque ATTRIBUTE_UNUSED)
709
{
710 711 712
    int ret = -1;
    char *configdir = NULL;
    char *rundir = NULL;
713 714 715
#ifdef HAVE_FIREWALLD
    DBusConnection *sysbus = NULL;
#endif
716

717
    if (VIR_ALLOC(network_driver) < 0)
718
        goto error;
719

720 721
    if (virMutexInit(&network_driver->lock) < 0) {
        VIR_FREE(network_driver);
722 723
        goto error;
    }
724

725 726 727 728
    /* configuration/state paths are one of
     * ~/.config/libvirt/... (session/unprivileged)
     * /etc/libvirt/... && /var/(run|lib)/libvirt/... (system/privileged).
     */
729
    if (privileged) {
730
        if (VIR_STRDUP(network_driver->networkConfigDir,
731
                       SYSCONFDIR "/libvirt/qemu/networks") < 0 ||
732
            VIR_STRDUP(network_driver->networkAutostartDir,
733
                       SYSCONFDIR "/libvirt/qemu/networks/autostart") < 0 ||
734
            VIR_STRDUP(network_driver->stateDir,
735
                       LOCALSTATEDIR "/run/libvirt/network") < 0 ||
736
            VIR_STRDUP(network_driver->pidDir,
737
                       LOCALSTATEDIR "/run/libvirt/network") < 0 ||
738
            VIR_STRDUP(network_driver->dnsmasqStateDir,
739
                       LOCALSTATEDIR "/lib/libvirt/dnsmasq") < 0 ||
740
            VIR_STRDUP(network_driver->radvdStateDir,
741 742
                       LOCALSTATEDIR "/lib/libvirt/radvd") < 0)
            goto error;
743 744 745 746 747

        /* migration from old to new location is only applicable for
         * privileged mode - unprivileged mode directories haven't
         * changed location.
         */
748
        if (networkMigrateStateFiles(network_driver) < 0)
749
            goto error;
750
    } else {
751 752 753
        configdir = virGetUserConfigDirectory();
        rundir = virGetUserRuntimeDirectory();
        if (!(configdir && rundir))
754
            goto error;
755

756
        if ((virAsprintf(&network_driver->networkConfigDir,
757
                         "%s/qemu/networks", configdir) < 0) ||
758
            (virAsprintf(&network_driver->networkAutostartDir,
759
                         "%s/qemu/networks/autostart", configdir) < 0) ||
760
            (virAsprintf(&network_driver->stateDir,
761
                         "%s/network/lib", rundir) < 0) ||
762
            (virAsprintf(&network_driver->pidDir,
763
                         "%s/network/run", rundir) < 0) ||
764
            (virAsprintf(&network_driver->dnsmasqStateDir,
765
                         "%s/dnsmasq/lib", rundir) < 0) ||
766
            (virAsprintf(&network_driver->radvdStateDir,
767
                         "%s/radvd/lib", rundir) < 0)) {
768
            goto error;
769
        }
770 771
    }

772
    if (virFileMakePath(network_driver->stateDir) < 0) {
773 774
        virReportSystemError(errno,
                             _("cannot create directory %s"),
775
                             network_driver->stateDir);
776 777 778
        goto error;
    }

779
    /* if this fails now, it will be retried later with dnsmasqCapsRefresh() */
780
    network_driver->dnsmasqCaps = dnsmasqCapsNewFromBinary(DNSMASQ);
781

782
    if (!(network_driver->networks = virNetworkObjListNew()))
783 784
        goto error;

785 786
    if (virNetworkObjLoadAllState(network_driver->networks,
                                  network_driver->stateDir) < 0)
787 788
        goto error;

789 790 791
    if (virNetworkObjLoadAllConfigs(network_driver->networks,
                                    network_driver->networkConfigDir,
                                    network_driver->networkAutostartDir) < 0)
792 793
        goto error;

794 795 796 797
    /* Update the internal status of all allegedly active
     * networks according to external conditions on the host
     * (i.e. anything that isn't stored directly in each
     * network's state file). */
798
    virNetworkObjListForEach(network_driver->networks,
799
                             networkUpdateState,
800 801
                             network_driver);
    virNetworkObjListPrune(network_driver->networks,
802 803
                           VIR_CONNECT_LIST_NETWORKS_INACTIVE |
                           VIR_CONNECT_LIST_NETWORKS_TRANSIENT);
804 805
    networkReloadFirewallRules(network_driver);
    networkRefreshDaemons(network_driver);
806

807
    network_driver->networkEventState = virObjectEventStateNew();
808

809 810 811
#ifdef HAVE_FIREWALLD
    if (!(sysbus = virDBusGetSystemBus())) {
        VIR_WARN("DBus not available, disabling firewalld support "
812
                 "in bridge_network_driver: %s", virGetLastErrorMessage());
813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829
    } 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,
830
                                   network_driver, NULL);
831 832 833
    }
#endif

834
    ret = 0;
835
 cleanup:
836 837 838
    VIR_FREE(configdir);
    VIR_FREE(rundir);
    return ret;
839

840
 error:
841
    networkStateCleanup();
842
    goto cleanup;
843 844
}

845

846 847 848 849 850 851 852 853
/**
 * networkStateAutoStart:
 *
 * Function to AutoStart the bridge configs
 */
static void
networkStateAutoStart(void)
{
854
    if (!network_driver)
855 856
        return;

857
    virNetworkObjListForEach(network_driver->networks,
858
                             networkAutostartConfig,
859
                             network_driver);
860 861
}

862

863
/**
864
 * networkStateReload:
865 866 867 868 869
 *
 * Function to restart the QEmu daemon, it will recheck the configuration
 * files and update its state and the networking
 */
static int
870 871
networkStateReload(void)
{
872
    if (!network_driver)
873 874
        return 0;

875 876 877 878 879
    virNetworkObjLoadAllState(network_driver->networks,
                              network_driver->stateDir);
    virNetworkObjLoadAllConfigs(network_driver->networks,
                                network_driver->networkConfigDir,
                                network_driver->networkAutostartDir);
880 881 882
    networkReloadFirewallRules(network_driver);
    networkRefreshDaemons(network_driver);
    virNetworkObjListForEach(network_driver->networks,
883
                             networkAutostartConfig,
884
                             network_driver);
885 886 887 888 889
    return 0;
}


/**
890
 * networkStateCleanup:
891 892 893 894
 *
 * Shutdown the QEmu daemon, it will stop all active domains and networks
 */
static int
895 896
networkStateCleanup(void)
{
897
    if (!network_driver)
898 899
        return -1;

900
    virObjectUnref(network_driver->networkEventState);
901

902
    /* free inactive networks */
903
    virObjectUnref(network_driver->networks);
904

905 906 907 908 909 910
    VIR_FREE(network_driver->networkConfigDir);
    VIR_FREE(network_driver->networkAutostartDir);
    VIR_FREE(network_driver->stateDir);
    VIR_FREE(network_driver->pidDir);
    VIR_FREE(network_driver->dnsmasqStateDir);
    VIR_FREE(network_driver->radvdStateDir);
911

912
    virObjectUnref(network_driver->dnsmasqCaps);
913

914
    virMutexDestroy(&network_driver->lock);
915

916
    VIR_FREE(network_driver);
917 918 919 920 921

    return 0;
}


922 923 924 925 926
/* networkKillDaemon:
 *
 * kill the specified pid/name, and wait a bit to make sure it's dead.
 */
static int
927 928 929
networkKillDaemon(pid_t pid,
                  const char *daemonName,
                  const char *networkName)
930
{
931 932
    size_t i;
    int ret = -1;
933 934 935 936 937 938 939
    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.
     */
940
    for (i = 0; i < 25; i++) {
941
        int signum = 0;
942
        if (i == 0) {
943
            signum = SIGTERM;
944
        } else if (i == 15) {
945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979
            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);
980
 cleanup:
981 982 983
    return ret;
}

984

J
Ján Tomko 已提交
985 986 987
/* the following does not build a file, it builds a list
 * which is later saved into a file
 */
988
static int
G
Gene Czarcinski 已提交
989
networkBuildDnsmasqDhcpHostsList(dnsmasqContext *dctx,
990
                                 virNetworkIPDefPtr ipdef)
991
{
992
    size_t i;
G
Gene Czarcinski 已提交
993
    bool ipv6 = false;
994

G
Gene Czarcinski 已提交
995 996
    if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
        ipv6 = true;
997 998
    for (i = 0; i < ipdef->nhosts; i++) {
        virNetworkDHCPHostDefPtr host = &(ipdef->hosts[i]);
G
Gene Czarcinski 已提交
999
        if (VIR_SOCKET_ADDR_VALID(&host->ip))
1000 1001
            if (dnsmasqAddDhcpHost(dctx, host->mac, &host->ip,
                                   host->name, host->id, ipv6) < 0)
1002
                return -1;
1003
    }
1004

G
Gene Czarcinski 已提交
1005 1006 1007
    return 0;
}

1008

G
Gene Czarcinski 已提交
1009 1010 1011 1012
static int
networkBuildDnsmasqHostsList(dnsmasqContext *dctx,
                             virNetworkDNSDefPtr dnsdef)
{
1013
    size_t i, j;
G
Gene Czarcinski 已提交
1014

1015 1016
    if (dnsdef) {
        for (i = 0; i < dnsdef->nhosts; i++) {
1017
            virNetworkDNSHostDefPtr host = &(dnsdef->hosts[i]);
1018
            if (VIR_SOCKET_ADDR_VALID(&host->ip)) {
1019
                for (j = 0; j < host->nnames; j++)
1020 1021
                    if (dnsmasqAddHost(dctx, &host->ip, host->names[j]) < 0)
                        return -1;
1022 1023
            }
        }
1024 1025
    }

1026
    return 0;
1027 1028 1029
}


1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
static int
networkDnsmasqConfLocalPTRs(virBufferPtr buf,
                            virNetworkDefPtr def)
{
    virNetworkIPDefPtr ip;
    size_t i;
    char *ptr = NULL;
    int rc;

    for (i = 0; i < def->nips; i++) {
        ip = def->ips + i;

        if (ip->localPTR != VIR_TRISTATE_BOOL_YES)
            continue;

        if ((rc = virSocketAddrPTRDomain(&ip->address,
                                         virNetworkIPDefPrefix(ip),
                                         &ptr)) < 0) {
            if (rc == -2) {
                int family = VIR_SOCKET_ADDR_FAMILY(&ip->address);
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("PTR domain for %s network with prefix %u "
                                 "cannot be automatically created"),
                               (family == AF_INET) ? "IPv4" : "IPv6",
                               virNetworkIPDefPrefix(ip));
            }
            return -1;
        }

        virBufferAsprintf(buf, "local=/%s/\n", ptr);
        VIR_FREE(ptr);
    }

    return 0;
}


1067
int
1068
networkDnsmasqConfContents(virNetworkObjPtr obj,
1069 1070 1071 1072
                           const char *pidfile,
                           char **configstr,
                           dnsmasqContext *dctx,
                           dnsmasqCapsPtr caps ATTRIBUTE_UNUSED)
1073
{
1074
    virBuffer configbuf = VIR_BUFFER_INITIALIZER;
1075
    int r, ret = -1;
1076
    int nbleases = 0;
1077
    size_t i;
1078
    virNetworkDNSDefPtr dns = &obj->def->dns;
1079
    bool wantDNS = dns->enable != VIR_TRISTATE_BOOL_NO;
1080
    virNetworkIPDefPtr tmpipdef, ipdef, ipv4def, ipv6def;
G
Gene Czarcinski 已提交
1081
    bool ipv6SLAAC;
1082
    char *saddr = NULL, *eaddr = NULL;
1083

1084 1085
    *configstr = NULL;

1086
    /*
1087 1088 1089
     * All dnsmasq parameters are put into a configuration file, except the
     * command line --conf-file=parameter which specifies the location of
     * configuration file.
1090
     *
1091 1092
     * All dnsmasq conf-file parameters must be specified as "foo=bar"
     * as oppose to "--foo bar" which was acceptable on the command line.
1093
     */
1094 1095 1096 1097 1098 1099

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

1100
    /* create dnsmasq config file appropriate for this network */
1101 1102

    /* Don't forget to update networkxml2conftest :-) */
1103
    virBufferAsprintf(&configbuf,
1104 1105 1106 1107 1108 1109 1110
                      "##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"
1111
                      "strict-order\n",
1112
                      obj->def->name);
1113

1114 1115 1116 1117 1118 1119
    /* if dns is disabled, set its listening port to 0, which
     * tells dnsmasq to not listen
     */
    if (!wantDNS)
        virBufferAddLit(&configbuf, "port=0\n");

1120
    if (wantDNS && obj->def->dns.forwarders) {
1121 1122 1123 1124 1125 1126 1127 1128 1129
        /* addNoResolv should be set to true if there are any entries
         * that specify an IP address for requests, but no domain
         * qualifier (implying that all requests otherwise "unclaimed"
         * should be sent to that address). if it is still false when
         * we've looked at all entries, it means we still need the
         * host's resolv.conf for some cases.
         */
        bool addNoResolv = false;

1130 1131
        for (i = 0; i < obj->def->dns.nfwds; i++) {
            virNetworkDNSForwarderPtr fwd = &obj->def->dns.forwarders[i];
1132 1133 1134 1135 1136 1137 1138 1139 1140 1141

            virBufferAddLit(&configbuf, "server=");
            if (fwd->domain)
                virBufferAsprintf(&configbuf, "/%s/", fwd->domain);
            if (VIR_SOCKET_ADDR_VALID(&fwd->addr)) {
                char *addr = virSocketAddrFormat(&fwd->addr);

                if (!addr)
                    goto cleanup;
                virBufferAsprintf(&configbuf, "%s\n", addr);
1142
                VIR_FREE(addr);
1143 1144
                if (!fwd->domain)
                    addNoResolv = true;
1145 1146 1147 1148
            } else {
                /* "don't forward requests for this domain" */
                virBufferAddLit(&configbuf, "#\n");
            }
1149
        }
1150 1151
        if (addNoResolv)
            virBufferAddLit(&configbuf, "no-resolv\n");
1152 1153
    }

1154 1155
    if (obj->def->domain) {
        if (obj->def->domainLocalOnly == VIR_TRISTATE_BOOL_YES) {
1156 1157
            virBufferAsprintf(&configbuf,
                              "local=/%s/\n",
1158
                              obj->def->domain);
1159
        }
1160
        virBufferAsprintf(&configbuf,
1161 1162
                          "domain=%s\n"
                          "expand-hosts\n",
1163
                          obj->def->domain);
1164
    }
1165

1166
    if (wantDNS &&
1167
        networkDnsmasqConfLocalPTRs(&configbuf, obj->def) < 0)
1168 1169
        goto cleanup;

1170
    if (wantDNS && obj->def->dns.forwardPlainNames == VIR_TRISTATE_BOOL_NO) {
1171 1172 1173 1174
        virBufferAddLit(&configbuf, "domain-needed\n");
        /* need to specify local=// whether or not a domain is
         * specified, unless the config says we should forward "plain"
         * names (i.e. not fully qualified, no '.' characters)
1175
         */
1176
        virBufferAddLit(&configbuf, "local=//\n");
1177
    }
1178

1179
    if (pidfile)
1180
        virBufferAsprintf(&configbuf, "pid-file=%s\n", pidfile);
1181

1182
    /* dnsmasq will *always* listen on localhost unless told otherwise */
P
Pavel Timofeev 已提交
1183
#ifdef __linux__
1184
    virBufferAddLit(&configbuf, "except-interface=lo\n");
P
Pavel Timofeev 已提交
1185 1186 1187 1188
#else
    /* BSD family OSes and Solaris call loopback interface as lo0 */
    virBufferAddLit(&configbuf, "except-interface=lo0\n");
#endif
1189

1190 1191 1192 1193 1194 1195 1196 1197
    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.
         */
1198
        virBufferAsprintf(&configbuf,
1199 1200
                          "bind-dynamic\n"
                          "interface=%s\n",
1201
                          obj->def->bridge);
1202
    } else {
1203
        virBufferAddLit(&configbuf, "bind-interfaces\n");
1204 1205 1206 1207
        /*
         * --interface does not actually work with dnsmasq < 2.47,
         * due to DAD for ipv6 addresses on the interface.
         *
1208
         * virCommandAddArgList(cmd, "--interface", obj->def->bridge, NULL);
1209 1210 1211
         *
         * So listen on all defined IPv[46] addresses
         */
1212
        for (i = 0;
1213
             (tmpipdef = virNetworkDefGetIPByIndex(obj->def, AF_UNSPEC, i));
1214
             i++) {
1215 1216 1217 1218
            char *ipaddr = virSocketAddrFormat(&tmpipdef->address);

            if (!ipaddr)
                goto cleanup;
1219

1220
            /* also part of CVE 2012-3411 - if the host's version of
1221
             * dnsmasq doesn't have bind-dynamic, only allow listening on
1222 1223
             * private/local IP addresses (see RFC1918/RFC3484/RFC4193)
             */
1224 1225
            if (!dnsmasqCapsGet(caps, DNSMASQ_CAPS_BINDTODEVICE) &&
                !virSocketAddrIsPrivate(&tmpipdef->address)) {
1226 1227 1228 1229
                unsigned long version = dnsmasqCapsGetVersion(caps);

                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("Publicly routable address %s is prohibited. "
1230
                                 "The version of dnsmasq on this host (%d.%d) "
1231 1232 1233 1234
                                 "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 "
1235 1236 1237 1238 1239 1240
                                 "(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);
1241
                VIR_FREE(ipaddr);
1242 1243
                goto cleanup;
            }
1244
            virBufferAsprintf(&configbuf, "listen-address=%s\n", ipaddr);
1245 1246 1247
            VIR_FREE(ipaddr);
        }
    }
1248

1249 1250
    /* If this is an isolated network, set the default route option
     * (3) to be empty to avoid setting a default route that's
1251
     * guaranteed to not work, and set no-resolv so that no dns
1252 1253 1254
     * 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).
1255 1256 1257
     * IPv6 RA always contains an implicit default route
     * via the sender's link-local address. The only thing we can do
     * is set the lifetime of this route to 0, i.e. disable it.
1258
     */
1259
    if (obj->def->forward.type == VIR_NETWORK_FORWARD_NONE) {
1260
        virBufferAddLit(&configbuf, "dhcp-option=3\n"
1261
                        "no-resolv\n");
1262 1263 1264 1265
        if (dnsmasqCapsGet(caps, DNSMASQ_CAPS_RA_PARAM)) {
            /* interface=* (any), interval=0 (default), lifetime=0 (seconds) */
            virBufferAddLit(&configbuf, "ra-param=*,0,0\n");
        }
1266
    }
1267

1268 1269 1270 1271 1272
    if (wantDNS) {
        for (i = 0; i < dns->ntxts; i++) {
            virBufferAsprintf(&configbuf, "txt-record=%s,%s\n",
                              dns->txts[i].name,
                              dns->txts[i].value);
1273
        }
1274

1275 1276 1277 1278 1279 1280 1281 1282
        for (i = 0; i < dns->nsrvs; i++) {
            /* service/protocol are required, and should have been validated
             * by the parser.
             */
            if (!dns->srvs[i].service) {
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Missing required 'service' "
                                 "attribute in SRV record of network '%s'"),
1283
                               obj->def->name);
1284 1285 1286 1287 1288 1289
                goto cleanup;
            }
            if (!dns->srvs[i].protocol) {
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("Missing required 'service' "
                                 "attribute in SRV record of network '%s'"),
1290
                               obj->def->name);
1291 1292 1293 1294 1295 1296 1297
                goto cleanup;
            }
            /* RFC2782 requires that service and protocol be preceded by
             * an underscore.
             */
            virBufferAsprintf(&configbuf, "srv-host=_%s._%s",
                              dns->srvs[i].service, dns->srvs[i].protocol);
1298

1299 1300 1301 1302 1303 1304 1305
            /* domain is optional - it defaults to the domain of this network */
            if (dns->srvs[i].domain)
                virBufferAsprintf(&configbuf, ".%s", dns->srvs[i].domain);

            /* If target is empty or ".", that means "the service is
             * decidedly not available at this domain" (RFC2782). In that
             * case, any port, priority, or weight is irrelevant.
1306
             */
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
            if (dns->srvs[i].target && STRNEQ(dns->srvs[i].target, ".")) {

                virBufferAsprintf(&configbuf, ",%s", dns->srvs[i].target);
                /* port, priority, and weight are optional, but are
                 * identified by their position in the line. If an item is
                 * unspecified, but something later in the line *is*
                 * specified, we need to give the default value for the
                 * unspecified item. (According to the dnsmasq manpage,
                 * the default for port is 1).
                 */
                if (dns->srvs[i].port ||
                    dns->srvs[i].priority || dns->srvs[i].weight)
                    virBufferAsprintf(&configbuf, ",%d",
                                      dns->srvs[i].port ? dns->srvs[i].port : 1);
                if (dns->srvs[i].priority || dns->srvs[i].weight)
                    virBufferAsprintf(&configbuf, ",%d", dns->srvs[i].priority);
                if (dns->srvs[i].weight)
                    virBufferAsprintf(&configbuf, ",%d", dns->srvs[i].weight);
            }
            virBufferAddLit(&configbuf, "\n");
1327
        }
1328 1329
    }

G
Gene Czarcinski 已提交
1330
    /* Find the first dhcp for both IPv4 and IPv6 */
1331
    for (i = 0, ipv4def = NULL, ipv6def = NULL, ipv6SLAAC = false;
1332
         (ipdef = virNetworkDefGetIPByIndex(obj->def, AF_UNSPEC, i));
1333
         i++) {
G
Gene Czarcinski 已提交
1334 1335 1336 1337
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
            if (ipdef->nranges || ipdef->nhosts) {
                if (ipv4def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
1338 1339
                                   _("For IPv4, multiple DHCP definitions "
                                     "cannot be specified."));
G
Gene Czarcinski 已提交
1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350
                    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,
1351 1352 1353 1354 1355 1356 1357 1358 1359
                                   _("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
1360 1361 1362 1363
                    goto cleanup;
                }
                if (ipv6def) {
                    virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
1364 1365
                                   _("For IPv6, multiple DHCP definitions "
                                     "cannot be specified."));
G
Gene Czarcinski 已提交
1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378
                    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 "
1379 1380 1381 1382
                 "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
1383 1384 1385 1386 1387
    }

    ipdef = ipv4def ? ipv4def : ipv6def;

    while (ipdef) {
1388 1389
        int prefix;

1390
        prefix = virNetworkIPDefPrefix(ipdef);
1391 1392 1393
        if (prefix < 0) {
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("bridge '%s' has an invalid prefix"),
1394
                           obj->def->bridge);
1395 1396
            goto cleanup;
        }
1397
        for (r = 0; r < ipdef->nranges; r++) {
1398 1399
            int thisRange;

1400 1401
            if (!(saddr = virSocketAddrFormat(&ipdef->ranges[r].start)) ||
                !(eaddr = virSocketAddrFormat(&ipdef->ranges[r].end)))
1402
                goto cleanup;
1403

1404
            virBufferAsprintf(&configbuf, "dhcp-range=%s,%s",
1405
                              saddr, eaddr);
1406
            if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
1407
                virBufferAsprintf(&configbuf, ",%d", prefix);
1408 1409
            virBufferAddLit(&configbuf, "\n");

1410
            VIR_FREE(saddr);
1411
            VIR_FREE(eaddr);
1412
            thisRange = virSocketAddrGetRange(&ipdef->ranges[r].start,
1413 1414
                                              &ipdef->ranges[r].end,
                                              &ipdef->address,
1415
                                              virNetworkIPDefPrefix(ipdef));
1416 1417 1418
            if (thisRange < 0)
                goto cleanup;
            nbleases += thisRange;
1419
        }
1420

1421
        /*
1422 1423 1424 1425
         * 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)
1426 1427
         */
        if (!ipdef->nranges && ipdef->nhosts) {
1428
            char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
1429 1430
            if (!bridgeaddr)
                goto cleanup;
1431 1432 1433
            virBufferAsprintf(&configbuf, "dhcp-range=%s,static",
                              bridgeaddr);
            if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
1434
                virBufferAsprintf(&configbuf, ",%d", prefix);
1435
            virBufferAddLit(&configbuf, "\n");
1436 1437
            VIR_FREE(bridgeaddr);
        }
1438

G
Gene Czarcinski 已提交
1439 1440
        if (networkBuildDnsmasqDhcpHostsList(dctx, ipdef) < 0)
            goto cleanup;
1441

G
Gene Czarcinski 已提交
1442 1443
        /* Note: the following is IPv4 only */
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
1444
            if (ipdef->nranges || ipdef->nhosts) {
1445
                virBufferAddLit(&configbuf, "dhcp-no-override\n");
1446 1447
                virBufferAddLit(&configbuf, "dhcp-authoritative\n");
            }
1448

G
Gene Czarcinski 已提交
1449
            if (ipdef->tftproot) {
1450 1451
                virBufferAddLit(&configbuf, "enable-tftp\n");
                virBufferAsprintf(&configbuf, "tftp-root=%s\n", ipdef->tftproot);
G
Gene Czarcinski 已提交
1452
            }
1453

G
Gene Czarcinski 已提交
1454 1455 1456
            if (ipdef->bootfile) {
                if (VIR_SOCKET_ADDR_VALID(&ipdef->bootserver)) {
                    char *bootserver = virSocketAddrFormat(&ipdef->bootserver);
1457

1458
                    if (!bootserver)
G
Gene Czarcinski 已提交
1459
                        goto cleanup;
1460
                    virBufferAsprintf(&configbuf, "dhcp-boot=%s%s%s\n",
1461
                                      ipdef->bootfile, ",,", bootserver);
G
Gene Czarcinski 已提交
1462 1463
                    VIR_FREE(bootserver);
                } else {
1464
                    virBufferAsprintf(&configbuf, "dhcp-boot=%s\n", ipdef->bootfile);
G
Gene Czarcinski 已提交
1465 1466 1467 1468 1469
                }
            }
        }
        ipdef = (ipdef == ipv6def) ? NULL : ipv6def;
    }
1470

1471
    if (nbleases > 0)
1472
        virBufferAsprintf(&configbuf, "dhcp-lease-max=%d\n", nbleases);
1473

G
Gene Czarcinski 已提交
1474 1475
    /* this is done once per interface */
    if (networkBuildDnsmasqHostsList(dctx, dns) < 0)
1476
        goto cleanup;
G
Gene Czarcinski 已提交
1477 1478 1479 1480 1481 1482

    /* 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)
1483 1484
        virBufferAsprintf(&configbuf, "dhcp-hostsfile=%s\n",
                          dctx->hostsfile->path);
G
Gene Czarcinski 已提交
1485

1486 1487
    /* Likewise, always create this file and put it on the
     * commandline, to allow for runtime additions.
G
Gene Czarcinski 已提交
1488
     */
1489 1490 1491 1492
    if (wantDNS) {
        virBufferAsprintf(&configbuf, "addn-hosts=%s\n",
                          dctx->addnhostsfile->path);
    }
G
Gene Czarcinski 已提交
1493 1494 1495

    /* Are we doing RA instead of radvd? */
    if (DNSMASQ_RA_SUPPORT(caps)) {
1496
        if (ipv6def) {
1497
            virBufferAddLit(&configbuf, "enable-ra\n");
1498
        } else {
1499
            for (i = 0;
1500
                 (ipdef = virNetworkDefGetIPByIndex(obj->def, AF_INET6, i));
1501
                 i++) {
G
Gene Czarcinski 已提交
1502 1503 1504 1505
                if (!(ipdef->nranges || ipdef->nhosts)) {
                    char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
                    if (!bridgeaddr)
                        goto cleanup;
1506 1507
                    virBufferAsprintf(&configbuf,
                                      "dhcp-range=%s,ra-only\n", bridgeaddr);
G
Gene Czarcinski 已提交
1508 1509
                    VIR_FREE(bridgeaddr);
                }
1510
            }
1511
        }
1512 1513
    }

1514 1515 1516
    if (!(*configstr = virBufferContentAndReset(&configbuf)))
        goto cleanup;

1517
    ret = 0;
G
Gene Czarcinski 已提交
1518

1519
 cleanup:
1520 1521
    VIR_FREE(saddr);
    VIR_FREE(eaddr);
1522
    virBufferFreeAndReset(&configbuf);
1523
    return ret;
1524 1525
}

1526

1527
/* build the dnsmasq command line */
1528 1529
static int ATTRIBUTE_NONNULL(3)
networkBuildDhcpDaemonCommandLine(virNetworkDriverStatePtr driver,
1530
                                  virNetworkObjPtr obj,
1531
                                  virCommandPtr *cmdout,
1532 1533
                                  char *pidfile,
                                  dnsmasqContext *dctx)
1534
{
1535
    dnsmasqCapsPtr dnsmasq_caps = networkGetDnsmasqCaps(driver);
1536
    virCommandPtr cmd = NULL;
G
Gene Czarcinski 已提交
1537
    int ret = -1;
1538 1539
    char *configfile = NULL;
    char *configstr = NULL;
1540
    char *leaseshelper_path = NULL;
1541

1542
    obj->dnsmasqPid = -1;
1543

1544
    if (networkDnsmasqConfContents(obj, pidfile, &configstr,
1545
                                   dctx, dnsmasq_caps) < 0)
1546 1547 1548 1549 1550
        goto cleanup;
    if (!configstr)
        goto cleanup;

    /* construct the filename */
1551
    if (!(configfile = networkDnsmasqConfigFileName(driver, obj->def->name)))
1552 1553 1554 1555 1556
        goto cleanup;

    /* Write the file */
    if (virFileWriteStr(configfile, configstr, 0600) < 0) {
        virReportSystemError(errno,
J
Ján Tomko 已提交
1557 1558
                             _("couldn't write dnsmasq config file '%s'"),
                             configfile);
1559 1560 1561
        goto cleanup;
    }

1562 1563
    /* This helper is used to create custom leases file for libvirt */
    if (!(leaseshelper_path = virFileFindResource("libvirt_leaseshelper",
1564
                                                  abs_topbuilddir "/src",
1565 1566 1567
                                                  LIBEXECDIR)))
        goto cleanup;

1568
    cmd = virCommandNew(dnsmasqCapsGetBinaryPath(dnsmasq_caps));
1569
    virCommandAddArgFormat(cmd, "--conf-file=%s", configfile);
1570 1571
    /* Libvirt gains full control of leases database */
    virCommandAddArgFormat(cmd, "--leasefile-ro");
1572
    virCommandAddArgFormat(cmd, "--dhcp-script=%s", leaseshelper_path);
1573
    virCommandAddEnvPair(cmd, "VIR_BRIDGE_NAME", obj->def->bridge);
1574

1575
    *cmdout = cmd;
1576
    ret = 0;
1577
 cleanup:
1578
    virObjectUnref(dnsmasq_caps);
1579 1580
    VIR_FREE(configfile);
    VIR_FREE(configstr);
1581
    VIR_FREE(leaseshelper_path);
1582 1583 1584
    return ret;
}

1585

1586
static int
1587
networkStartDhcpDaemon(virNetworkDriverStatePtr driver,
1588
                       virNetworkObjPtr obj)
1589
{
1590 1591 1592
    virNetworkIPDefPtr ipdef;
    size_t i;
    bool needDnsmasq = false;
1593 1594 1595
    virCommandPtr cmd = NULL;
    char *pidfile = NULL;
    int ret = -1;
1596
    dnsmasqContext *dctx = NULL;
1597

1598
    /* see if there are any IP addresses that need a dhcp server */
1599
    i = 0;
1600
    while ((ipdef = virNetworkDefGetIPByIndex(obj->def, AF_UNSPEC, i))) {
1601
        i++;
1602 1603 1604 1605
        if (ipdef->nranges || ipdef->nhosts)
            needDnsmasq = true;
    }

1606 1607 1608 1609 1610 1611
    if (i == 0) {
        /* no IP addresses at all, so we don't need to run */
        ret = 0;
        goto cleanup;
    }

1612
    if (!needDnsmasq && obj->def->dns.enable == VIR_TRISTATE_BOOL_NO) {
1613
        /* no DHCP services needed, and user disabled DNS service */
1614 1615 1616 1617
        ret = 0;
        goto cleanup;
    }

1618
    if (virFileMakePath(driver->pidDir) < 0) {
1619
        virReportSystemError(errno,
1620
                             _("cannot create directory %s"),
1621
                             driver->pidDir);
1622
        goto cleanup;
1623 1624
    }

1625
    if (!(pidfile = virPidFileBuildPath(driver->pidDir,
1626
                                        obj->def->name)))
1627
        goto cleanup;
1628

1629
    if (virFileMakePath(driver->dnsmasqStateDir) < 0) {
1630
        virReportSystemError(errno,
1631
                             _("cannot create directory %s"),
1632
                             driver->dnsmasqStateDir);
1633 1634 1635
        goto cleanup;
    }

1636
    dctx = dnsmasqContextNew(obj->def->name, driver->dnsmasqStateDir);
1637 1638 1639
    if (dctx == NULL)
        goto cleanup;

1640
    if (networkDnsmasqCapsRefresh(driver) < 0)
1641
        goto cleanup;
1642

1643
    ret = networkBuildDhcpDaemonCommandLine(driver, obj, &cmd, pidfile, dctx);
1644 1645 1646 1647 1648
    if (ret < 0)
        goto cleanup;

    ret = dnsmasqSave(dctx);
    if (ret < 0)
1649
        goto cleanup;
1650

G
Guido Günther 已提交
1651
    ret = virCommandRun(cmd, NULL);
1652
    if (ret < 0)
1653 1654 1655
        goto cleanup;

    /*
1656 1657 1658 1659 1660
     * 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
1661 1662
     */

1663
    ret = virPidFileRead(driver->pidDir, obj->def->name, &obj->dnsmasqPid);
1664
    if (ret < 0)
1665
        goto cleanup;
1666

1667
    ret = 0;
1668
 cleanup:
1669
    VIR_FREE(pidfile);
1670
    virCommandFree(cmd);
1671
    dnsmasqContextFree(dctx);
1672 1673 1674
    return ret;
}

1675

1676 1677
/* networkRefreshDhcpDaemon:
 *  Update dnsmasq config files, then send a SIGHUP so that it rereads
G
Gene Czarcinski 已提交
1678 1679
 *  them.   This only works for the dhcp-hostsfile and the
 *  addn-hosts file.
1680 1681 1682
 *
 *  Returns 0 on success, -1 on failure.
 */
1683
static int
1684
networkRefreshDhcpDaemon(virNetworkDriverStatePtr driver,
1685
                         virNetworkObjPtr obj)
1686
{
1687 1688
    int ret = -1;
    size_t i;
1689
    virNetworkIPDefPtr ipdef, ipv4def, ipv6def;
1690
    dnsmasqContext *dctx = NULL;
1691

G
Gene Czarcinski 已提交
1692
    /* if no IP addresses specified, nothing to do */
1693
    if (!virNetworkDefGetIPByIndex(obj->def, AF_UNSPEC, 0))
G
Gene Czarcinski 已提交
1694 1695
        return 0;

1696
    /* if there's no running dnsmasq, just start it */
1697 1698
    if (obj->dnsmasqPid <= 0 || (kill(obj->dnsmasqPid, 0) < 0))
        return networkStartDhcpDaemon(driver, obj);
1699

1700 1701
    VIR_INFO("Refreshing dnsmasq for network %s", obj->def->bridge);
    if (!(dctx = dnsmasqContextNew(obj->def->name,
1702
                                   driver->dnsmasqStateDir))) {
G
Gene Czarcinski 已提交
1703
        goto cleanup;
1704
    }
G
Gene Czarcinski 已提交
1705 1706 1707 1708 1709 1710

    /* 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;
1711
    for (i = 0;
1712
         (ipdef = virNetworkDefGetIPByIndex(obj->def, AF_INET, i));
1713
         i++) {
G
Gene Czarcinski 已提交
1714 1715
        if (!ipv4def && (ipdef->nranges || ipdef->nhosts))
            ipv4def = ipdef;
1716 1717
    }

G
Gene Czarcinski 已提交
1718
    ipv6def = NULL;
1719
    for (i = 0;
1720
         (ipdef = virNetworkDefGetIPByIndex(obj->def, AF_INET6, i));
1721
         i++) {
G
Gene Czarcinski 已提交
1722 1723
        if (!ipv6def && (ipdef->nranges || ipdef->nhosts))
            ipv6def = ipdef;
1724 1725
    }

G
Gene Czarcinski 已提交
1726
    if (ipv4def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv4def) < 0))
J
Ján Tomko 已提交
1727
        goto cleanup;
G
Gene Czarcinski 已提交
1728 1729

    if (ipv6def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv6def) < 0))
J
Ján Tomko 已提交
1730
        goto cleanup;
1731

1732
    if (networkBuildDnsmasqHostsList(dctx, &obj->def->dns) < 0)
J
Ján Tomko 已提交
1733
        goto cleanup;
1734 1735

    if ((ret = dnsmasqSave(dctx)) < 0)
1736
        goto cleanup;
1737

1738
    ret = kill(obj->dnsmasqPid, SIGHUP);
1739
 cleanup:
1740 1741 1742 1743
    dnsmasqContextFree(dctx);
    return ret;
}

1744

1745 1746 1747 1748 1749 1750 1751 1752
/* 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
1753
networkRestartDhcpDaemon(virNetworkDriverStatePtr driver,
1754
                         virNetworkObjPtr obj)
1755 1756
{
    /* if there is a running dnsmasq, kill it */
1757 1758 1759
    if (obj->dnsmasqPid > 0) {
        networkKillDaemon(obj->dnsmasqPid, "dnsmasq", obj->def->name);
        obj->dnsmasqPid = -1;
1760
    }
1761
    /* now start dnsmasq if it should be started */
1762
    return networkStartDhcpDaemon(driver, obj);
1763 1764
}

1765

G
Gene Czarcinski 已提交
1766 1767 1768 1769 1770 1771
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";

1772
static int
1773
networkRadvdConfContents(virNetworkObjPtr obj,
1774
                         char **configstr)
1775
{
E
Eric Blake 已提交
1776
    virBuffer configbuf = VIR_BUFFER_INITIALIZER;
1777 1778
    int ret = -1;
    size_t i;
1779
    virNetworkIPDefPtr ipdef;
G
Gene Czarcinski 已提交
1780
    bool v6present = false, dhcp6 = false;
1781 1782

    *configstr = NULL;
1783

G
Gene Czarcinski 已提交
1784
    /* Check if DHCPv6 is needed */
1785
    for (i = 0;
1786
         (ipdef = virNetworkDefGetIPByIndex(obj->def, AF_INET6, i));
1787
         i++) {
G
Gene Czarcinski 已提交
1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800
        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;
    }

1801 1802 1803
    /* create radvd config file appropriate for this network;
     * IgnoreIfMissing allows radvd to start even when the bridge is down
     */
1804
    virBufferAsprintf(&configbuf, "interface %s\n"
1805 1806
                      "{\n"
                      "  AdvSendAdvert on;\n"
1807
                      "  IgnoreIfMissing on;\n"
G
Gene Czarcinski 已提交
1808 1809
                      "  AdvManagedFlag %s;\n"
                      "%s",
1810
                      obj->def->bridge,
G
Gene Czarcinski 已提交
1811 1812
                      dhcp6 ? "on" : "off",
                      dhcp6 ? "\n" : radvd1);
1813 1814

    /* add a section for each IPv6 address in the config */
1815
    for (i = 0;
1816
         (ipdef = virNetworkDefGetIPByIndex(obj->def, AF_INET6, i));
1817
         i++) {
1818 1819 1820
        int prefix;
        char *netaddr;

1821
        prefix = virNetworkIPDefPrefix(ipdef);
1822
        if (prefix < 0) {
1823 1824
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("bridge '%s' has an invalid prefix"),
1825
                           obj->def->bridge);
1826 1827
            goto cleanup;
        }
1828
        if (!(netaddr = virSocketAddrFormat(&ipdef->address)))
1829
            goto cleanup;
1830
        virBufferAsprintf(&configbuf,
1831
                          "  prefix %s/%d\n"
G
Gene Czarcinski 已提交
1832 1833 1834
                          "  {\n%s  };\n",
                          netaddr, prefix,
                          dhcp6 ? radvd2 : radvd3);
1835 1836 1837
        VIR_FREE(netaddr);
    }

1838
    virBufferAddLit(&configbuf, "};\n");
1839

1840
    if (virBufferCheckError(&configbuf) < 0)
1841
        goto cleanup;
1842

1843 1844
    *configstr = virBufferContentAndReset(&configbuf);

1845
    ret = 0;
1846
 cleanup:
1847 1848 1849 1850
    virBufferFreeAndReset(&configbuf);
    return ret;
}

1851

1852
/* write file and return its name (which must be freed by caller) */
1853
static int
1854
networkRadvdConfWrite(virNetworkDriverStatePtr driver,
1855
                      virNetworkObjPtr obj,
1856
                      char **configFile)
1857 1858 1859 1860 1861 1862 1863 1864 1865 1866
{
    int ret = -1;
    char *configStr = NULL;
    char *myConfigFile = NULL;

    if (!configFile)
        configFile = &myConfigFile;

    *configFile = NULL;

1867
    if (networkRadvdConfContents(obj, &configStr) < 0)
1868 1869 1870 1871
        goto cleanup;

    if (!configStr) {
        ret = 0;
1872 1873 1874 1875
        goto cleanup;
    }

    /* construct the filename */
1876
    if (!(*configFile = networkRadvdConfigFileName(driver, obj->def->name)))
1877 1878
        goto cleanup;
    /* write the file */
1879
    if (virFileWriteStr(*configFile, configStr, 0600) < 0) {
1880 1881
        virReportSystemError(errno,
                             _("couldn't write radvd config file '%s'"),
1882 1883 1884 1885 1886
                             *configFile);
        goto cleanup;
    }

    ret = 0;
1887
 cleanup:
1888 1889 1890 1891 1892
    VIR_FREE(configStr);
    VIR_FREE(myConfigFile);
    return ret;
}

1893

1894
static int
1895
networkStartRadvd(virNetworkDriverStatePtr driver,
1896
                  virNetworkObjPtr obj)
1897
{
1898
    dnsmasqCapsPtr dnsmasq_caps = networkGetDnsmasqCaps(driver);
1899 1900 1901 1902 1903 1904
    char *pidfile = NULL;
    char *radvdpidbase = NULL;
    char *configfile = NULL;
    virCommandPtr cmd = NULL;
    int ret = -1;

1905
    obj->radvdPid = -1;
1906

G
Gene Czarcinski 已提交
1907
    /* Is dnsmasq handling RA? */
1908
    if (DNSMASQ_RA_SUPPORT(dnsmasq_caps)) {
G
Gene Czarcinski 已提交
1909 1910 1911 1912
        ret = 0;
        goto cleanup;
    }

1913
    if (!virNetworkDefGetIPByIndex(obj->def, AF_INET6, 0)) {
1914 1915 1916 1917 1918 1919 1920 1921 1922 1923
        /* 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);
1924 1925 1926
        goto cleanup;
    }

1927
    if (virFileMakePath(driver->pidDir) < 0) {
1928 1929
        virReportSystemError(errno,
                             _("cannot create directory %s"),
1930
                             driver->pidDir);
1931 1932
        goto cleanup;
    }
1933
    if (virFileMakePath(driver->radvdStateDir) < 0) {
1934 1935
        virReportSystemError(errno,
                             _("cannot create directory %s"),
1936
                             driver->radvdStateDir);
1937 1938 1939 1940
        goto cleanup;
    }

    /* construct pidfile name */
1941
    if (!(radvdpidbase = networkRadvdPidfileBasename(obj->def->name)))
1942
        goto cleanup;
1943
    if (!(pidfile = virPidFileBuildPath(driver->pidDir, radvdpidbase)))
1944 1945
        goto cleanup;

1946
    if (networkRadvdConfWrite(driver, obj, &configfile) < 0)
1947 1948
        goto cleanup;

1949 1950 1951 1952
    /* 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
1953
     * virPidFileRead() below will fail if we use them).
1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968
     * 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;

1969
    if (virPidFileRead(driver->pidDir, radvdpidbase, &obj->radvdPid) < 0)
1970 1971 1972
        goto cleanup;

    ret = 0;
1973
 cleanup:
1974
    virObjectUnref(dnsmasq_caps);
1975 1976 1977 1978 1979 1980 1981
    virCommandFree(cmd);
    VIR_FREE(configfile);
    VIR_FREE(radvdpidbase);
    VIR_FREE(pidfile);
    return ret;
}

1982

1983
static int
1984
networkRefreshRadvd(virNetworkDriverStatePtr driver,
1985
                    virNetworkObjPtr obj)
1986
{
1987
    dnsmasqCapsPtr dnsmasq_caps = networkGetDnsmasqCaps(driver);
G
Gene Czarcinski 已提交
1988 1989 1990
    char *radvdpidbase;

    /* Is dnsmasq handling RA? */
1991 1992
    if (DNSMASQ_RA_SUPPORT(dnsmasq_caps)) {
        virObjectUnref(dnsmasq_caps);
1993
        if (obj->radvdPid <= 0)
G
Gene Czarcinski 已提交
1994 1995
            return 0;
        /* radvd should not be running but in case it is */
1996 1997
        if ((networkKillDaemon(obj->radvdPid, "radvd", obj->def->name) >= 0) &&
            ((radvdpidbase = networkRadvdPidfileBasename(obj->def->name))
G
Gene Czarcinski 已提交
1998
             != NULL)) {
1999
            virPidFileDelete(driver->pidDir, radvdpidbase);
G
Gene Czarcinski 已提交
2000 2001
            VIR_FREE(radvdpidbase);
        }
2002
        obj->radvdPid = -1;
G
Gene Czarcinski 已提交
2003 2004
        return 0;
    }
2005
    virObjectUnref(dnsmasq_caps);
G
Gene Czarcinski 已提交
2006

2007
    /* if there's no running radvd, just start it */
2008 2009
    if (obj->radvdPid <= 0 || (kill(obj->radvdPid, 0) < 0))
        return networkStartRadvd(driver, obj);
2010

2011
    if (!virNetworkDefGetIPByIndex(obj->def, AF_INET6, 0)) {
2012 2013 2014 2015
        /* no IPv6 addresses, so we don't need to run radvd */
        return 0;
    }

2016
    if (networkRadvdConfWrite(driver, obj, NULL) < 0)
2017 2018
        return -1;

2019
    return kill(obj->radvdPid, SIGHUP);
2020 2021
}

2022

2023 2024
#if 0
/* currently unused, so it causes a build error unless we #if it out */
2025
static int
2026
networkRestartRadvd(virNetworkObjPtr obj)
2027 2028 2029 2030
{
    char *radvdpidbase;

    /* if there is a running radvd, kill it */
2031
    if (obj->radvdPid > 0) {
2032 2033 2034 2035
        /* 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).
         */
2036 2037 2038
        if ((networkKillDaemon(obj->radvdPid, "radvd",
                               obj->def->name) >= 0) &&
            ((radvdpidbase = networkRadvdPidfileBasename(obj->def->name))
2039
             != NULL)) {
2040
            virPidFileDelete(driver->pidDir, radvdpidbase);
2041 2042
            VIR_FREE(radvdpidbase);
        }
2043
        obj->radvdPid = -1;
2044 2045
    }
    /* now start radvd if it should be started */
2046
    return networkStartRadvd(obj);
2047 2048 2049
}
#endif /* #if 0 */

2050

2051
static int
2052
networkRefreshDaemonsHelper(virNetworkObjPtr obj,
2053
                            void *opaque)
2054
{
2055
    virNetworkDriverStatePtr driver = opaque;
2056

2057 2058 2059 2060 2061 2062
    virObjectLock(obj);
    if (virNetworkObjIsActive(obj) &&
        ((obj->def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
         (obj->def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
         (obj->def->forward.type == VIR_NETWORK_FORWARD_ROUTE) ||
         (obj->def->forward.type == VIR_NETWORK_FORWARD_OPEN))) {
2063 2064 2065 2066 2067 2068
        /* 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.
         */
2069 2070
        networkRefreshDhcpDaemon(driver, obj);
        networkRefreshRadvd(driver, obj);
2071
    }
2072
    virObjectUnlock(obj);
2073 2074 2075
    return 0;
}

2076

2077 2078 2079 2080
/* SIGHUP/restart any dnsmasq or radvd daemons.
 * This should be called when libvirtd is restarted.
 */
static void
2081
networkRefreshDaemons(virNetworkDriverStatePtr driver)
2082 2083
{
    VIR_INFO("Refreshing network daemons");
2084 2085
    virNetworkObjListForEach(driver->networks,
                             networkRefreshDaemonsHelper,
2086
                             driver);
2087
}
2088

2089

2090
static int
2091
networkReloadFirewallRulesHelper(virNetworkObjPtr obj,
2092 2093 2094
                                 void *opaque ATTRIBUTE_UNUSED)
{

2095 2096 2097 2098 2099
    virObjectLock(obj);
    if (virNetworkObjIsActive(obj) &&
        ((obj->def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
         (obj->def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
         (obj->def->forward.type == VIR_NETWORK_FORWARD_ROUTE))) {
2100 2101 2102 2103
        /* Only three of the L3 network types that are configured by
         * libvirt need to have iptables rules reloaded. The 4th L3
         * network type, forward='open', doesn't need this because it
         * has no iptables rules.
2104
         */
2105 2106
        networkRemoveFirewallRules(obj->def);
        if (networkAddFirewallRules(obj->def) < 0) {
2107
            /* failed to add but already logged */
2108 2109
        }
    }
2110
    virObjectUnlock(obj);
2111
    return 0;
2112 2113
}

2114

2115
static void
2116
networkReloadFirewallRules(virNetworkDriverStatePtr driver)
2117
{
2118
    VIR_INFO("Reloading iptables rules");
2119 2120 2121
    virNetworkObjListForEach(driver->networks,
                             networkReloadFirewallRulesHelper,
                             NULL);
2122 2123
}

2124

2125
/* Enable IP Forwarding. Return 0 for success, -1 for failure. */
2126
static int
2127 2128
networkEnableIPForwarding(bool enableIPv4,
                          bool enableIPv6)
2129
{
2130
    int ret = 0;
2131 2132 2133 2134
#ifdef HAVE_SYSCTLBYNAME
    int enabled = 1;
    if (enableIPv4)
        ret = sysctlbyname("net.inet.ip.forwarding", NULL, 0,
J
Ján Tomko 已提交
2135
                           &enabled, sizeof(enabled));
2136 2137
    if (enableIPv6 && ret == 0)
        ret = sysctlbyname("net.inet6.ip6.forwarding", NULL, 0,
J
Ján Tomko 已提交
2138
                           &enabled, sizeof(enabled));
2139
#else
2140
    if (enableIPv4)
2141
        ret = virFileWriteStr(SYSCTL_PATH "/net/ipv4/ip_forward", "1\n", 0);
2142
    if (enableIPv6 && ret == 0)
2143 2144
        ret = virFileWriteStr(SYSCTL_PATH "/net/ipv6/conf/all/forwarding", "1\n", 0);

2145
#endif
2146
    return ret;
2147 2148
}

2149

2150
static int
2151
networkSetIPv6Sysctls(virNetworkObjPtr obj)
2152 2153 2154
{
    char *field = NULL;
    int ret = -1;
2155
    bool enableIPv6 =  !!virNetworkDefGetIPByIndex(obj->def, AF_INET6, 0);
2156

2157 2158 2159 2160 2161
    /* set disable_ipv6 if there are no ipv6 addresses defined for the
     * network. But also unset it if there *are* ipv6 addresses, as we
     * can't be sure of its default value.
     */
    if (virAsprintf(&field, SYSCTL_PATH "/net/ipv6/conf/%s/disable_ipv6",
2162
                    obj->def->bridge) < 0)
2163
       goto cleanup;
2164

2165 2166
    if (access(field, W_OK) < 0 && errno == ENOENT) {
        if (!enableIPv6)
2167
            VIR_DEBUG("ipv6 appears to already be disabled on %s",
2168
                      obj->def->bridge);
2169 2170 2171
        ret = 0;
        goto cleanup;
    }
2172

2173 2174 2175
    if (virFileWriteStr(field, enableIPv6 ? "0" : "1", 0) < 0) {
        virReportSystemError(errno,
                             _("cannot write to %s to enable/disable IPv6 "
2176
                               "on bridge %s"), field, obj->def->bridge);
2177
        goto cleanup;
2178
    }
2179
    VIR_FREE(field);
2180

2181 2182
    /* The rest of the ipv6 sysctl tunables should always be set the
     * same, whether or not we're using ipv6 on this bridge.
2183 2184 2185 2186 2187 2188
     */

    /* 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",
2189
                    obj->def->bridge) < 0)
2190 2191
        goto cleanup;

2192
    if (virFileWriteStr(field, "0", 0) < 0) {
2193
        virReportSystemError(errno,
2194 2195 2196 2197 2198
                             _("cannot disable %s"), field);
        goto cleanup;
    }
    VIR_FREE(field);

2199 2200 2201 2202
    /* 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",
2203
                    obj->def->bridge) < 0)
2204 2205
        goto cleanup;

2206
    if (virFileWriteStr(field, "0", 0) < 0) {
2207
        virReportSystemError(errno,
2208
                             _("cannot disable %s"), field);
2209 2210 2211 2212
        goto cleanup;
    }

    ret = 0;
2213
 cleanup:
2214 2215 2216 2217
    VIR_FREE(field);
    return ret;
}

2218

2219
/* add an IP address to a bridge */
2220
static int
2221
networkAddAddrToBridge(virNetworkObjPtr obj,
2222
                       virNetworkIPDefPtr ipdef)
2223
{
2224
    int prefix = virNetworkIPDefPrefix(ipdef);
2225 2226

    if (prefix < 0) {
2227 2228
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("bridge '%s' has an invalid netmask or IP address"),
2229
                       obj->def->bridge);
2230 2231 2232
        return -1;
    }

2233
    if (virNetDevIPAddrAdd(obj->def->bridge,
2234
                           &ipdef->address, NULL, prefix) < 0)
2235 2236 2237 2238 2239
        return -1;

    return 0;
}

2240 2241

static int
2242
networkStartHandleMACTableManagerMode(virNetworkObjPtr obj,
2243 2244
                                      const char *macTapIfName)
{
2245
    const char *brname = obj->def->bridge;
2246 2247

    if (brname &&
2248
        obj->def->macTableManager
2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262
        == VIR_NETWORK_BRIDGE_MAC_TABLE_MANAGER_LIBVIRT) {
        if (virNetDevBridgeSetVlanFiltering(brname, true) < 0)
            return -1;
        if (macTapIfName) {
            if (virNetDevBridgePortSetLearning(brname, macTapIfName, false) < 0)
                return -1;
            if (virNetDevBridgePortSetUnicastFlood(brname, macTapIfName, false) < 0)
                return -1;
        }
    }
    return 0;
}


2263 2264
/* add an IP (static) route to a bridge */
static int
2265
networkAddRouteToBridge(virNetworkObjPtr obj,
2266
                        virNetDevIPRoutePtr routedef)
2267
{
2268 2269 2270 2271
    int prefix = virNetDevIPRouteGetPrefix(routedef);
    unsigned int metric = virNetDevIPRouteGetMetric(routedef);
    virSocketAddrPtr addr = virNetDevIPRouteGetAddress(routedef);
    virSocketAddrPtr gateway = virNetDevIPRouteGetGateway(routedef);
2272 2273 2274 2275 2276

    if (prefix < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("network '%s' has an invalid netmask "
                         "or IP address in route definition"),
2277
                       obj->def->name);
2278 2279 2280
        return -1;
    }

2281
    if (virNetDevIPRouteAdd(obj->def->bridge, addr,
2282
                            prefix, gateway, metric) < 0) {
2283 2284 2285 2286 2287
        return -1;
    }
    return 0;
}

2288
static int
2289
networkWaitDadFinish(virNetworkObjPtr obj)
2290
{
2291
    virNetworkIPDefPtr ipdef;
2292 2293 2294 2295
    virSocketAddrPtr *addrs = NULL, addr = NULL;
    size_t naddrs = 0;
    int ret = -1;

2296
    VIR_DEBUG("Begin waiting for IPv6 DAD on network %s", obj->def->name);
2297

2298
    while ((ipdef = virNetworkDefGetIPByIndex(obj->def, AF_INET6, naddrs))) {
2299 2300 2301 2302 2303
        addr = &ipdef->address;
        if (VIR_APPEND_ELEMENT_COPY(addrs, naddrs, addr) < 0)
            goto cleanup;
    }

2304
    ret = (naddrs == 0) ? 0 : virNetDevIPWaitDadFinish(addrs, naddrs);
2305 2306 2307 2308

 cleanup:
    VIR_FREE(addrs);
    VIR_DEBUG("Finished waiting for IPv6 DAD on network %s with status %d",
2309
              obj->def->name, ret);
2310 2311 2312
    return ret;
}

2313

2314
static int
2315
networkStartNetworkVirtual(virNetworkDriverStatePtr driver,
2316
                           virNetworkObjPtr obj)
2317
{
2318
    size_t i;
2319
    bool v4present = false, v6present = false;
2320
    virErrorPtr save_err = NULL;
2321
    virNetworkIPDefPtr ipdef;
2322
    virNetDevIPRoutePtr routedef;
2323
    char *macTapIfName = NULL;
M
Michal Privoznik 已提交
2324
    char *macMapFile = NULL;
2325
    int tapfd = -1;
2326

2327
    /* Check to see if any network IP collides with an existing route */
2328
    if (networkCheckRouteCollision(obj->def) < 0)
2329 2330
        return -1;

2331
    /* Create and configure the bridge device */
2332
    if (!obj->def->bridge) {
2333 2334 2335 2336 2337 2338 2339 2340 2341 2342
        /* bridge name can only be empty if the config files were
         * edited directly. Otherwise networkValidate() (called after
         * parsing the XML from networkCreateXML() and
         * networkDefine()) guarantees we will have a valid bridge
         * name before this point. Since hand editing of the config
         * files is explicitly prohibited we can, with clear
         * conscience, log an error and fail at this point.
         */
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("network '%s' has no bridge name defined"),
2343
                       obj->def->name);
2344 2345
        return -1;
    }
2346
    if (virNetDevBridgeCreate(obj->def->bridge) < 0)
2347 2348
        return -1;

2349
    if (obj->def->mac_specified) {
2350 2351 2352 2353 2354 2355
        /* 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.
         */
2356
        macTapIfName = networkBridgeDummyNicName(obj->def->bridge);
2357
        if (!macTapIfName)
2358
            goto err0;
2359
        /* Keep tun fd open and interface up to allow for IPv6 DAD to happen */
2360 2361
        if (virNetDevTapCreateInBridgePort(obj->def->bridge,
                                           &macTapIfName, &obj->def->mac,
2362
                                           NULL, NULL, &tapfd, 1, NULL, NULL,
2363
                                           NULL, obj->def->mtu, NULL,
2364 2365 2366
                                           VIR_NETDEV_TAP_CREATE_USE_MAC_FOR_BRIDGE |
                                           VIR_NETDEV_TAP_CREATE_IFUP |
                                           VIR_NETDEV_TAP_CREATE_PERSIST) < 0) {
2367 2368 2369 2370 2371
            VIR_FREE(macTapIfName);
            goto err0;
        }
    }

2372 2373
    if (!(macMapFile = virMacMapFileName(driver->dnsmasqStateDir,
                                         obj->def->bridge)) ||
2374
        !(obj->macmap = virMacMapNew(macMapFile)))
M
Michal Privoznik 已提交
2375 2376
        goto err1;

2377
    /* Set bridge options */
2378 2379 2380 2381

    /* delay is configured in seconds, but virNetDevBridgeSetSTPDelay
     * expects milliseconds
     */
2382 2383
    if (virNetDevBridgeSetSTPDelay(obj->def->bridge,
                                   obj->def->delay * 1000) < 0)
2384
        goto err1;
2385

2386 2387
    if (virNetDevBridgeSetSTP(obj->def->bridge,
                              obj->def->stp ? true : false) < 0)
2388
        goto err1;
2389

2390 2391 2392
    /* Disable IPv6 on the bridge if there are no IPv6 addresses
     * defined, and set other IPv6 sysctl tunables appropriately.
     */
2393
    if (networkSetIPv6Sysctls(obj) < 0)
2394
        goto err1;
2395

2396
    /* Add "once per network" rules */
2397 2398
    if (obj->def->forward.type != VIR_NETWORK_FORWARD_OPEN &&
        networkAddFirewallRules(obj->def) < 0)
2399 2400
        goto err1;

2401
    for (i = 0;
2402
         (ipdef = virNetworkDefGetIPByIndex(obj->def, AF_UNSPEC, i));
2403
         i++) {
2404
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET))
2405
            v4present = true;
2406
        if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
2407
            v6present = true;
2408

2409
        /* Add the IP address/netmask to the bridge */
2410
        if (networkAddAddrToBridge(obj, ipdef) < 0)
2411
            goto err2;
2412 2413
    }

2414
    if (networkStartHandleMACTableManagerMode(obj, macTapIfName) < 0)
2415 2416
        goto err2;

2417
    /* Bring up the bridge interface */
2418
    if (virNetDevSetOnline(obj->def->bridge, 1) < 0)
2419
        goto err2;
2420

2421
    for (i = 0; i < obj->def->nroutes; i++) {
2422 2423
        virSocketAddrPtr gateway = NULL;

2424
        routedef = obj->def->routes[i];
2425
        gateway = virNetDevIPRouteGetGateway(routedef);
2426

2427 2428 2429
        /* Add the IP route to the bridge */
        /* ignore errors, error msg will be generated */
        /* but libvirt will not know and net-destroy will work. */
2430
        if (VIR_SOCKET_ADDR_VALID(gateway)) {
2431
            if (networkAddRouteToBridge(obj, routedef) < 0) {
2432 2433 2434 2435 2436 2437
                /* an error occurred adding the static route */
                continue; /* for now, do nothing */
            }
        }
    }

2438
    /* If forward.type != NONE, turn on global IP forwarding */
2439
    if (obj->def->forward.type != VIR_NETWORK_FORWARD_NONE) {
2440
        if (v6present && !virNetDevIPCheckIPv6Forwarding())
2441 2442 2443 2444 2445 2446 2447
            goto err3; /* Precise error message already provided */

        if (networkEnableIPForwarding(v4present, v6present) < 0) {
            virReportSystemError(errno, "%s",
                                 _("failed to enable IP forwarding"));
            goto err3;
        }
2448 2449
    }

2450

2451
    /* start dnsmasq if there are any IP addresses (v4 or v6) */
2452
    if ((v4present || v6present) &&
2453
        networkStartDhcpDaemon(driver, obj) < 0)
2454
        goto err3;
2455

2456
    /* start radvd if there are any ipv6 addresses */
2457
    if (v6present && networkStartRadvd(driver, obj) < 0)
2458 2459
        goto err4;

2460 2461 2462
    /* dnsmasq does not wait for DAD to complete before daemonizing,
     * so we need to wait for it ourselves.
     */
2463
    if (v6present && networkWaitDadFinish(obj) < 0)
2464 2465 2466 2467
        goto err4;

    /* DAD has finished, dnsmasq is now bound to the
     * bridge's IPv6 address, so we can set the dummy tun down.
2468 2469 2470 2471 2472 2473 2474
     */
    if (tapfd >= 0) {
        if (virNetDevSetOnline(macTapIfName, false) < 0)
            goto err4;
        VIR_FORCE_CLOSE(tapfd);
    }

2475
    if (virNetDevBandwidthSet(obj->def->bridge, obj->def->bandwidth, true) < 0)
2476 2477
        goto err5;

2478
    VIR_FREE(macTapIfName);
M
Michal Privoznik 已提交
2479
    VIR_FREE(macMapFile);
2480 2481 2482

    return 0;

2483
 err5:
2484 2485
    if (obj->def->bandwidth)
       virNetDevBandwidthClear(obj->def->bridge);
2486

2487 2488 2489 2490
 err4:
    if (!save_err)
        save_err = virSaveLastError();

2491 2492 2493
    if (obj->dnsmasqPid > 0) {
        kill(obj->dnsmasqPid, SIGTERM);
        obj->dnsmasqPid = -1;
2494 2495
    }

2496 2497 2498
 err3:
    if (!save_err)
        save_err = virSaveLastError();
2499
    ignore_value(virNetDevSetOnline(obj->def->bridge, 0));
2500

2501 2502 2503
 err2:
    if (!save_err)
        save_err = virSaveLastError();
2504 2505
    if (obj->def->forward.type != VIR_NETWORK_FORWARD_OPEN)
        networkRemoveFirewallRules(obj->def);
2506 2507

 err1:
2508 2509 2510
    if (!save_err)
        save_err = virSaveLastError();

H
Hu Tao 已提交
2511
    if (macTapIfName) {
2512
        VIR_FORCE_CLOSE(tapfd);
2513
        ignore_value(virNetDevTapDelete(macTapIfName, NULL));
H
Hu Tao 已提交
2514 2515
        VIR_FREE(macTapIfName);
    }
M
Michal Privoznik 已提交
2516
    VIR_FREE(macMapFile);
2517 2518

 err0:
2519 2520
    if (!save_err)
        save_err = virSaveLastError();
2521
    ignore_value(virNetDevBridgeDelete(obj->def->bridge));
2522

2523 2524 2525 2526
    if (save_err) {
        virSetError(save_err);
        virFreeError(save_err);
    }
2527
    /* coverity[leaked_handle] - 'tapfd' is not leaked */
2528 2529 2530
    return -1;
}

2531

2532 2533
static int
networkShutdownNetworkVirtual(virNetworkDriverStatePtr driver,
2534
                              virNetworkObjPtr obj)
2535
{
2536 2537
    if (obj->def->bandwidth)
        virNetDevBandwidthClear(obj->def->bridge);
2538

2539 2540
    if (!virObjectUnref(obj->macmap))
        obj->macmap = NULL;
M
Michal Privoznik 已提交
2541

2542
    if (obj->radvdPid > 0) {
2543 2544
        char *radvdpidbase;

2545
        kill(obj->radvdPid, SIGTERM);
2546
        /* attempt to delete the pidfile we created */
2547
        if ((radvdpidbase = networkRadvdPidfileBasename(obj->def->name))) {
2548
            virPidFileDelete(driver->pidDir, radvdpidbase);
2549 2550 2551 2552
            VIR_FREE(radvdpidbase);
        }
    }

2553 2554
    if (obj->dnsmasqPid > 0)
        kill(obj->dnsmasqPid, SIGTERM);
2555

2556 2557
    if (obj->def->mac_specified) {
        char *macTapIfName = networkBridgeDummyNicName(obj->def->bridge);
2558
        if (macTapIfName) {
2559
            ignore_value(virNetDevTapDelete(macTapIfName, NULL));
2560 2561 2562 2563
            VIR_FREE(macTapIfName);
        }
    }

2564
    ignore_value(virNetDevSetOnline(obj->def->bridge, 0));
2565

2566 2567
    if (obj->def->forward.type != VIR_NETWORK_FORWARD_OPEN)
        networkRemoveFirewallRules(obj->def);
2568

2569
    ignore_value(virNetDevBridgeDelete(obj->def->bridge));
2570

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

2577 2578 2579 2580
    if (obj->radvdPid > 0 &&
        (kill(obj->radvdPid, 0) == 0))
        kill(obj->radvdPid, SIGKILL);
    obj->radvdPid = -1;
2581

2582 2583 2584
    return 0;
}

2585

2586
static int
2587
networkStartNetworkBridge(virNetworkObjPtr obj)
2588 2589 2590 2591 2592
{
    /* put anything here that needs to be done each time a network of
     * type BRIDGE, is started. On failure, undo anything you've done,
     * and return -1. On success return 0.
     */
2593
    return networkStartHandleMACTableManagerMode(obj, NULL);
2594 2595
}

2596

2597
static int
2598
networkShutdownNetworkBridge(virNetworkObjPtr obj ATTRIBUTE_UNUSED)
2599 2600 2601 2602 2603 2604 2605 2606 2607
{
    /* put anything here that needs to be done each time a network of
     * type BRIDGE is shutdown. On failure, undo anything you've done,
     * and return -1. On success return 0.
     */
    return 0;
}


2608 2609 2610 2611 2612 2613 2614 2615 2616
/* networkCreateInterfacePool:
 * @netdef: the original NetDef from the network
 *
 * Creates an implicit interface pool of VF's when a PF dev is given
 */
static int
networkCreateInterfacePool(virNetworkDefPtr netdef)
{
    size_t numVirtFns = 0;
2617
    unsigned int maxVirtFns = 0;
2618 2619 2620 2621 2622 2623
    char **vfNames = NULL;
    virPCIDeviceAddressPtr *virtFns;

    int ret = -1;
    size_t i;

2624 2625 2626
    if (netdef->forward.npfs == 0 || netdef->forward.nifs > 0)
       return 0;

2627 2628
    if ((virNetDevGetVirtualFunctions(netdef->forward.pfs->dev, &vfNames,
                                      &virtFns, &numVirtFns, &maxVirtFns)) < 0) {
2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 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
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Could not get Virtual functions on %s"),
                       netdef->forward.pfs->dev);
        goto cleanup;
    }

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

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

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

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

        case VIR_NETWORK_FORWARD_NONE:
        case VIR_NETWORK_FORWARD_NAT:
        case VIR_NETWORK_FORWARD_ROUTE:
2675
        case VIR_NETWORK_FORWARD_OPEN:
2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715
        case VIR_NETWORK_FORWARD_LAST:
            /* by definition these will never be encountered here */
            break;
        }
    }

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

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

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


2716
static int
2717
networkStartNetworkExternal(virNetworkObjPtr obj)
2718 2719
{
    /* put anything here that needs to be done each time a network of
2720
     * type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is started. On
2721 2722 2723
     * failure, undo anything you've done, and return -1. On success
     * return 0.
     */
2724
    return networkCreateInterfacePool(obj->def);
2725 2726
}

2727 2728

static int
2729
networkShutdownNetworkExternal(virNetworkObjPtr obj ATTRIBUTE_UNUSED)
2730 2731
{
    /* put anything here that needs to be done each time a network of
2732
     * type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is shutdown. On
2733 2734 2735 2736 2737 2738
     * failure, undo anything you've done, and return -1. On success
     * return 0.
     */
    return 0;
}

2739

2740
static int
2741
networkStartNetwork(virNetworkDriverStatePtr driver,
2742
                    virNetworkObjPtr obj)
2743
{
2744 2745
    int ret = -1;

2746
    VIR_DEBUG("driver=%p, network=%p", driver, obj);
2747

2748
    if (virNetworkObjIsActive(obj)) {
2749 2750
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("network is already active"));
2751
        return ret;
2752 2753
    }

2754 2755 2756
    VIR_DEBUG("Beginning network startup process");

    VIR_DEBUG("Setting current network def as transient");
2757
    if (virNetworkObjSetDefTransient(obj, true) < 0)
2758
        goto cleanup;
2759

2760 2761
    /* Run an early hook to set-up missing devices.
     * If the script raised an error abort the launch. */
2762
    if (networkRunHook(obj, NULL, NULL,
2763 2764 2765 2766
                       VIR_HOOK_NETWORK_OP_START,
                       VIR_HOOK_SUBOP_BEGIN) < 0)
        goto cleanup;

2767
    switch (obj->def->forward.type) {
2768 2769 2770 2771

    case VIR_NETWORK_FORWARD_NONE:
    case VIR_NETWORK_FORWARD_NAT:
    case VIR_NETWORK_FORWARD_ROUTE:
2772
    case VIR_NETWORK_FORWARD_OPEN:
2773
        if (networkStartNetworkVirtual(driver, obj) < 0)
2774
            goto cleanup;
2775 2776 2777
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
2778 2779
        if (obj->def->bridge) {
            if (networkStartNetworkBridge(obj) < 0)
2780 2781 2782 2783 2784 2785 2786
                goto cleanup;
            break;
        }
        /* intentionally fall through to the macvtap/direct case for
         * VIR_NETWORK_FORWARD_BRIDGE with no bridge device defined
         * (since that is macvtap bridge mode).
         */
2787 2788
        ATTRIBUTE_FALLTHROUGH;

2789 2790 2791
    case VIR_NETWORK_FORWARD_PRIVATE:
    case VIR_NETWORK_FORWARD_VEPA:
    case VIR_NETWORK_FORWARD_PASSTHROUGH:
2792
    case VIR_NETWORK_FORWARD_HOSTDEV:
2793
        if (networkStartNetworkExternal(obj) < 0)
2794
            goto cleanup;
2795 2796 2797
        break;
    }

2798
    /* finally we can call the 'started' hook script if any */
2799
    if (networkRunHook(obj, NULL, NULL,
2800 2801 2802 2803
                       VIR_HOOK_NETWORK_OP_STARTED,
                       VIR_HOOK_SUBOP_BEGIN) < 0)
        goto cleanup;

2804 2805 2806
    /* Persist the live configuration now that anything autogenerated
     * is setup.
     */
2807
    VIR_DEBUG("Writing network status to disk");
2808
    if (virNetworkObjSaveStatus(driver->stateDir, obj) < 0)
2809
        goto cleanup;
2810

2811 2812
    obj->active = 1;
    VIR_INFO("Network '%s' started up", obj->def->name);
2813
    ret = 0;
2814

2815
 cleanup:
2816
    if (ret < 0) {
2817
        virNetworkObjUnsetDefTransient(obj);
2818 2819
        virErrorPtr save_err = virSaveLastError();
        int save_errno = errno;
2820
        networkShutdownNetwork(driver, obj);
2821 2822 2823 2824 2825 2826 2827
        virSetError(save_err);
        virFreeError(save_err);
        errno = save_errno;
    }
    return ret;
}

2828

2829 2830
static int
networkShutdownNetwork(virNetworkDriverStatePtr driver,
2831
                       virNetworkObjPtr obj)
2832 2833 2834 2835
{
    int ret = 0;
    char *stateFile;

2836
    VIR_INFO("Shutting down network '%s'", obj->def->name);
2837

2838
    if (!virNetworkObjIsActive(obj))
2839 2840
        return 0;

2841
    stateFile = virNetworkConfigFile(driver->stateDir,
2842
                                     obj->def->name);
2843 2844 2845 2846 2847 2848
    if (!stateFile)
        return -1;

    unlink(stateFile);
    VIR_FREE(stateFile);

2849
    switch (obj->def->forward.type) {
2850 2851 2852 2853

    case VIR_NETWORK_FORWARD_NONE:
    case VIR_NETWORK_FORWARD_NAT:
    case VIR_NETWORK_FORWARD_ROUTE:
2854
    case VIR_NETWORK_FORWARD_OPEN:
2855
        ret = networkShutdownNetworkVirtual(driver, obj);
2856 2857 2858
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
2859 2860
        if (obj->def->bridge) {
            ret = networkShutdownNetworkBridge(obj);
2861 2862 2863 2864 2865 2866
            break;
        }
        /* intentionally fall through to the macvtap/direct case for
         * VIR_NETWORK_FORWARD_BRIDGE with no bridge device defined
         * (since that is macvtap bridge mode).
         */
2867 2868
        ATTRIBUTE_FALLTHROUGH;

2869 2870 2871
    case VIR_NETWORK_FORWARD_PRIVATE:
    case VIR_NETWORK_FORWARD_VEPA:
    case VIR_NETWORK_FORWARD_PASSTHROUGH:
2872
    case VIR_NETWORK_FORWARD_HOSTDEV:
2873
        ret = networkShutdownNetworkExternal(obj);
2874 2875 2876
        break;
    }

2877
    /* now that we know it's stopped call the hook if present */
2878
    networkRunHook(obj, NULL, NULL, VIR_HOOK_NETWORK_OP_STOPPED,
2879 2880
                   VIR_HOOK_SUBOP_END);

2881 2882
    obj->active = 0;
    virNetworkObjUnsetDefTransient(obj);
2883
    return ret;
2884 2885 2886
}


2887 2888 2889
static virNetworkPtr
networkLookupByUUID(virConnectPtr conn,
                    const unsigned char *uuid)
2890
{
2891
    virNetworkDriverStatePtr driver = networkGetDriver();
2892 2893
    virNetworkObjPtr obj;
    virNetworkPtr net = NULL;
2894

2895 2896
    obj = virNetworkObjFindByUUID(driver->networks, uuid);
    if (!obj) {
2897 2898
        char uuidstr[VIR_UUID_STRING_BUFLEN];
        virUUIDFormat(uuid, uuidstr);
2899
        virReportError(VIR_ERR_NO_NETWORK,
2900 2901
                       _("no network with matching uuid '%s'"),
                       uuidstr);
2902
        goto cleanup;
2903 2904
    }

2905
    if (virNetworkLookupByUUIDEnsureACL(conn, obj->def) < 0)
2906 2907
        goto cleanup;

2908
    net = virGetNetwork(conn, obj->def->name, obj->def->uuid);
2909

2910
 cleanup:
2911 2912
    virNetworkObjEndAPI(&obj);
    return net;
2913 2914
}

2915 2916 2917 2918

static virNetworkPtr
networkLookupByName(virConnectPtr conn,
                    const char *name)
2919
{
2920
    virNetworkDriverStatePtr driver = networkGetDriver();
2921 2922
    virNetworkObjPtr obj;
    virNetworkPtr net = NULL;
2923

2924 2925
    obj = virNetworkObjFindByName(driver->networks, name);
    if (!obj) {
2926 2927
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"), name);
2928
        goto cleanup;
2929 2930
    }

2931
    if (virNetworkLookupByNameEnsureACL(conn, obj->def) < 0)
2932 2933
        goto cleanup;

2934
    net = virGetNetwork(conn, obj->def->name, obj->def->uuid);
2935

2936
 cleanup:
2937 2938
    virNetworkObjEndAPI(&obj);
    return net;
2939 2940
}

2941 2942 2943

static int
networkConnectNumOfNetworks(virConnectPtr conn)
2944
{
2945
    virNetworkDriverStatePtr driver = networkGetDriver();
2946
    int nactive;
2947

2948 2949 2950
    if (virConnectNumOfNetworksEnsureACL(conn) < 0)
        return -1;

2951 2952 2953 2954
    nactive = virNetworkObjListNumOfNetworks(driver->networks,
                                             true,
                                             virConnectNumOfNetworksCheckACL,
                                             conn);
2955

2956 2957 2958
    return nactive;
}

2959 2960 2961 2962 2963

static int
networkConnectListNetworks(virConnectPtr conn,
                           char **const names,
                           int nnames)
2964 2965
{
    virNetworkDriverStatePtr driver = networkGetDriver();
2966
    int got = 0;
2967

2968 2969 2970
    if (virConnectListNetworksEnsureACL(conn) < 0)
        return -1;

2971 2972 2973 2974
    got = virNetworkObjListGetNames(driver->networks,
                                    true, names, nnames,
                                    virConnectListNetworksCheckACL,
                                    conn);
2975

2976 2977 2978
    return got;
}

2979 2980 2981

static int
networkConnectNumOfDefinedNetworks(virConnectPtr conn)
2982
{
2983
    virNetworkDriverStatePtr driver = networkGetDriver();
2984
    int ninactive = 0;
2985

2986 2987 2988
    if (virConnectNumOfDefinedNetworksEnsureACL(conn) < 0)
        return -1;

2989 2990 2991 2992
    ninactive = virNetworkObjListNumOfNetworks(driver->networks,
                                               false,
                                               virConnectNumOfDefinedNetworksCheckACL,
                                               conn);
2993

2994 2995 2996
    return ninactive;
}

2997 2998 2999 3000 3001

static int
networkConnectListDefinedNetworks(virConnectPtr conn,
                                  char **const names,
                                  int nnames)
3002 3003
{
    virNetworkDriverStatePtr driver = networkGetDriver();
3004
    int got = 0;
3005

3006 3007 3008
    if (virConnectListDefinedNetworksEnsureACL(conn) < 0)
        return -1;

3009 3010 3011 3012
    got = virNetworkObjListGetNames(driver->networks,
                                    false, names, nnames,
                                    virConnectListDefinedNetworksCheckACL,
                                    conn);
3013 3014 3015
    return got;
}

3016

3017
static int
3018 3019 3020
networkConnectListAllNetworks(virConnectPtr conn,
                              virNetworkPtr **nets,
                              unsigned int flags)
3021
{
3022
    virNetworkDriverStatePtr driver = networkGetDriver();
3023 3024 3025 3026
    int ret = -1;

    virCheckFlags(VIR_CONNECT_LIST_NETWORKS_FILTERS_ALL, -1);

3027 3028 3029
    if (virConnectListAllNetworksEnsureACL(conn) < 0)
        goto cleanup;

3030
    ret = virNetworkObjListExport(conn, driver->networks, nets,
3031 3032
                                  virConnectListAllNetworksCheckACL,
                                  flags);
3033

3034
 cleanup:
3035 3036
    return ret;
}
3037

3038

3039 3040 3041 3042 3043 3044 3045 3046
static int
networkConnectNetworkEventRegisterAny(virConnectPtr conn,
                                      virNetworkPtr net,
                                      int eventID,
                                      virConnectNetworkEventGenericCallback callback,
                                      void *opaque,
                                      virFreeCallback freecb)
{
3047
    virNetworkDriverStatePtr driver = networkGetDriver();
3048 3049 3050 3051 3052 3053
    int ret = -1;

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

    if (virNetworkEventStateRegisterID(conn, driver->networkEventState,
3054
                                       net, eventID, callback,
3055 3056 3057
                                       opaque, freecb, &ret) < 0)
        ret = -1;

3058
 cleanup:
3059 3060 3061
    return ret;
}

3062

3063 3064 3065 3066
static int
networkConnectNetworkEventDeregisterAny(virConnectPtr conn,
                                        int callbackID)
{
3067
    virNetworkDriverStatePtr driver = networkGetDriver();
3068 3069 3070 3071 3072
    int ret = -1;

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

3073 3074
    if (virObjectEventStateDeregisterID(conn,
                                        driver->networkEventState,
3075
                                        callbackID, true) < 0)
3076 3077 3078
        goto cleanup;

    ret = 0;
3079

3080
 cleanup:
3081 3082 3083
    return ret;
}

3084 3085 3086

static int
networkIsActive(virNetworkPtr net)
3087 3088 3089 3090
{
    virNetworkObjPtr obj;
    int ret = -1;

3091 3092
    if (!(obj = networkObjFromNetwork(net)))
        return ret;
3093 3094 3095 3096

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

3097 3098
    ret = virNetworkObjIsActive(obj);

3099
 cleanup:
3100
    virNetworkObjEndAPI(&obj);
3101 3102 3103
    return ret;
}

3104 3105 3106

static int
networkIsPersistent(virNetworkPtr net)
3107 3108 3109 3110
{
    virNetworkObjPtr obj;
    int ret = -1;

3111 3112
    if (!(obj = networkObjFromNetwork(net)))
        return ret;
3113 3114 3115 3116

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

3117 3118
    ret = obj->persistent;

3119
 cleanup:
3120
    virNetworkObjEndAPI(&obj);
3121 3122 3123 3124
    return ret;
}


3125 3126
/*
 * networkFindUnusedBridgeName() - try to find a bridge name that is
3127 3128 3129
 * unused by the currently configured libvirt networks, as well as by
 * the host system itself (possibly created by someone/something other
 * than libvirt). Set this network's name to that new name.
3130 3131 3132 3133 3134 3135 3136 3137
 */
static int
networkFindUnusedBridgeName(virNetworkObjListPtr nets,
                            virNetworkDefPtr def)
{

    int ret = -1, id = 0;
    char *newname = NULL;
3138 3139 3140 3141 3142
    const char *templ = "virbr%d";
    const char *p;

    if (def->bridge &&
        (p = strchr(def->bridge, '%')) == strrchr(def->bridge, '%') &&
3143
        p && p[1] == 'd')
3144
        templ = def->bridge;
3145 3146 3147 3148

    do {
        if (virAsprintf(&newname, templ, id) < 0)
            goto cleanup;
3149 3150 3151 3152 3153
        /* check if this name is used in another libvirt network or
         * there is an existing device with that name. ignore errors
         * from virNetDevExists(), just in case it isn't implemented
         * on this platform (probably impossible).
         */
3154
        if (!(virNetworkObjBridgeInUse(nets, newname, def->name) ||
3155
              virNetDevExists(newname) == 1)) {
3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188
            VIR_FREE(def->bridge); /*could contain template */
            def->bridge = newname;
            ret = 0;
            goto cleanup;
        }
        VIR_FREE(newname);
    } while (++id <= MAX_BRIDGE_ID);

    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("Bridge generation exceeded max id %d"),
                   MAX_BRIDGE_ID);
    ret = 0;
 cleanup:
    if (ret < 0)
        VIR_FREE(newname);
    return ret;
}


/*
 * networkValidateBridgeName() - if no bridge name is set, or if the
 * bridge name contains a %d (indicating that this is a template for
 * the actual name) try to set an appropriate bridge name.  If a
 * bridge name *is* set, make sure it doesn't conflict with any other
 * network's bridge name.
 */
static int
networkBridgeNameValidate(virNetworkObjListPtr nets,
                          virNetworkDefPtr def)
{
    int ret = -1;

    if (def->bridge && !strstr(def->bridge, "%d")) {
3189
        if (virNetworkObjBridgeInUse(nets, def->bridge, def->name)) {
3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("bridge name '%s' already in use."),
                           def->bridge);
            goto cleanup;
        }
    } else {
        /* Allocate a bridge name */
        if (networkFindUnusedBridgeName(nets, def) < 0)
            goto cleanup;
    }

    ret = 0;
 cleanup:
    return ret;
}


3207
static int
3208
networkValidate(virNetworkDriverStatePtr driver,
3209
                virNetworkDefPtr def)
3210
{
3211
    size_t i, j;
3212 3213
    bool vlanUsed, vlanAllowed, badVlanUse = false;
    virPortGroupDefPtr defaultPortGroup = NULL;
3214
    virNetworkIPDefPtr ipdef;
G
Gene Czarcinski 已提交
3215
    bool ipv4def = false, ipv6def = false;
3216
    bool bandwidthAllowed = true;
3217
    bool usesInterface = false, usesAddress = false;
3218

3219 3220 3221
    if (virXMLCheckIllegalChars("name", def->name, "\n") < 0)
        return -1;

3222 3223 3224
    /* Only the three L3 network types that are configured by libvirt
     * need to have a bridge device name / mac address provided
     */
3225 3226
    if (def->forward.type == VIR_NETWORK_FORWARD_NONE ||
        def->forward.type == VIR_NETWORK_FORWARD_NAT ||
3227 3228
        def->forward.type == VIR_NETWORK_FORWARD_ROUTE ||
        def->forward.type == VIR_NETWORK_FORWARD_OPEN) {
3229

3230 3231 3232 3233
        /* if no bridge name was given in the config, find a name
         * unused by any other libvirt networks and assign it.
         */
        if (networkBridgeNameValidate(driver->networks, def) < 0)
3234 3235 3236
            return -1;

        virNetworkSetBridgeMacAddr(def);
3237 3238
    } else {
        /* They are also the only types that currently support setting
3239 3240
         * a MAC or IP address for the host-side device (bridge), DNS
         * configuration, or network-wide bandwidth limits.
3241
         */
3242 3243 3244 3245 3246 3247 3248 3249
        if (def->mac_specified) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <mac> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
                           virNetworkForwardTypeToString(def->forward.type));
            return -1;
        }
3250
        if (virNetworkDefGetIPByIndex(def, AF_UNSPEC, 0)) {
3251 3252 3253 3254
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <ip> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
3255
                           virNetworkForwardTypeToString(def->forward.type));
3256 3257
            return -1;
        }
3258
        if (def->dns.ntxts || def->dns.nhosts || def->dns.nsrvs) {
3259 3260 3261 3262
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <dns> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
3263
                           virNetworkForwardTypeToString(def->forward.type));
3264 3265 3266 3267 3268 3269 3270
            return -1;
        }
        if (def->domain) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <domain> element in network %s "
                             "with forward mode='%s'"),
                           def->name,
3271
                           virNetworkForwardTypeToString(def->forward.type));
3272 3273
            return -1;
        }
3274 3275 3276 3277 3278 3279 3280 3281
        if (def->bandwidth) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported network-wide <bandwidth> element "
                             "in network %s with forward mode='%s'"),
                           def->name,
                           virNetworkForwardTypeToString(def->forward.type));
            return -1;
        }
3282
        bandwidthAllowed = false;
3283 3284
    }

3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297
    /* we support configs with a single PF defined:
     *   <pf dev='eth0'/>
     * or with a list of netdev names:
     *   <interface dev='eth9'/>
     * OR a list of PCI addresses
     *   <address type='pci' domain='0' bus='4' slot='0' function='1'/>
     * but not any combination of those.
     *
     * Since <interface> and <address> are for some strange reason
     * stored in the same array, we need to cycle through it and check
     * the type of each.
     */
    for (i = 0; i < def->forward.nifs; i++) {
3298 3299 3300 3301
        virNetworkForwardIfDefPtr iface = &def->forward.ifs[i];
        char *sysfs_path = NULL;

        switch ((virNetworkForwardHostdevDeviceType) iface->type) {
3302 3303
        case VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV:
            usesInterface = true;
3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314

            if (def->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("hostdev network '%s' lists '%s' "
                                 "in the device pool, but hostdev "
                                 "networks require all devices to "
                                 "be listed by PCI address, not "
                                 "network device name"),
                               def->name, iface->device.dev);
                return -1;
            }
3315
            break;
3316 3317

        case VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI: {
3318
            usesAddress = true;
3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343

            if (def->forward.type != VIR_NETWORK_FORWARD_HOSTDEV) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("network '%s' has forward mode '%s' "
                                 " but lists a device by PCI address "
                                 "in the device pool. This is only "
                                 "supported for networks with forward "
                                 "mode 'hostdev'"),
                               def->name,
                               virNetworkForwardTypeToString(def->forward.type));
                return -1;
            }

            if (virPCIDeviceAddressGetSysfsFile(&iface->device.pci, &sysfs_path) < 0)
                return -1;

            if (!virPCIIsVirtualFunction(sysfs_path)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("device '%s' in network '%s' is not "
                                 "an SR-IOV Virtual Function"),
                               sysfs_path, def->name);
                VIR_FREE(sysfs_path);
                return -1;
            }
            VIR_FREE(sysfs_path);
3344
            break;
3345 3346
        }

3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359
        case VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NONE:
        case VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_LAST:
            break;
        }
    }
    if ((def->forward.npfs > 0) + usesInterface + usesAddress > 1) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       _("<address>, <interface>, and <pf> elements of "
                         "<forward> in network %s are mutually exclusive"),
                       def->name);
        return -1;
    }

G
Gene Czarcinski 已提交
3360 3361 3362
    /* We only support dhcp on one IPv4 address and
     * on one IPv6 address per defined network
     */
3363
    for (i = 0;
3364
         (ipdef = virNetworkDefGetIPByIndex(def, AF_UNSPEC, i));
3365
         i++) {
G
Gene Czarcinski 已提交
3366 3367 3368 3369 3370
        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 -- "
3371 3372
                                 "dhcp is supported only for a "
                                 "single IPv4 address on each network"));
G
Gene Czarcinski 已提交
3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389
                    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;
                }
3390 3391 3392
            }
        }
    }
3393 3394 3395 3396 3397 3398

    /* 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.
     */

3399 3400 3401
    vlanAllowed = (def->forward.type == VIR_NETWORK_FORWARD_HOSTDEV ||
                   def->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH ||
                   (def->forward.type == VIR_NETWORK_FORWARD_BRIDGE &&
J
Ján Tomko 已提交
3402 3403
                    def->virtPortProfile &&
                    def->virtPortProfile->virtPortType
3404
                    == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH));
3405 3406

    vlanUsed = def->vlan.nTags > 0;
3407 3408
    for (i = 0; i < def->nPortGroups; i++) {
        if (vlanUsed || def->portGroups[i].vlan.nTags > 0) {
3409 3410 3411 3412 3413
            /* 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.
             */
3414
            if (def->portGroups[i].virtPortProfile) {
3415
                if (def->forward.type != VIR_NETWORK_FORWARD_BRIDGE ||
3416
                    def->portGroups[i].virtPortProfile->virtPortType
3417 3418 3419 3420 3421 3422 3423
                    != VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
                    badVlanUse = true;
                }
            } else if (!vlanAllowed) {
                /* virtualport taken from base network definition */
                badVlanUse = true;
            }
3424
        }
3425
        if (def->portGroups[i].isDefault) {
3426 3427 3428 3429 3430
            if (defaultPortGroup) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("network '%s' has multiple default "
                                 "<portgroup> elements (%s and %s), "
                                 "but only one default is allowed"),
3431
                               def->name, defaultPortGroup->name,
3432
                               def->portGroups[i].name);
3433
                return -1;
3434
            }
3435
            defaultPortGroup = &def->portGroups[i];
3436
        }
3437 3438 3439 3440 3441 3442 3443 3444 3445
        for (j = i + 1; j < def->nPortGroups; j++) {
            if (STREQ(def->portGroups[i].name, def->portGroups[j].name)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("multiple <portgroup> elements with the "
                                 "same name (%s) in network '%s'"),
                               def->portGroups[i].name, def->name);
                return -1;
            }
        }
3446 3447 3448 3449 3450 3451 3452 3453
        if (def->portGroups[i].bandwidth && !bandwidthAllowed) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                           _("Unsupported <bandwidth> element in network '%s' "
                             "in portgroup '%s' with forward mode='%s'"),
                           def->name, def->portGroups[i].name,
                           virNetworkForwardTypeToString(def->forward.type));
            return -1;
        }
3454
    }
3455 3456 3457 3458 3459 3460 3461
    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.
         */
3462 3463 3464 3465 3466 3467
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       _("<vlan> element specified for network %s, "
                         "whose type doesn't support vlan configuration"),
                       def->name);
        return -1;
    }
3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481

    if (def->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
        for (i = 0; i < def->nPortGroups; i++) {
            if (def->portGroups[i].bandwidth) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                               _("unsupported <bandwidth> element "
                                 "in <portgroup name='%s'> of "
                                 "network '%s' with forward mode='%s'"),
                               def->portGroups[i].name, def->name,
                               virNetworkForwardTypeToString(def->forward.type));
                return -1;
            }
        }
    }
3482 3483 3484
    return 0;
}

3485 3486 3487 3488

static virNetworkPtr
networkCreateXML(virConnectPtr conn,
                 const char *xml)
3489
{
3490
    virNetworkDriverStatePtr driver = networkGetDriver();
3491
    virNetworkDefPtr def;
3492 3493
    virNetworkObjPtr obj = NULL;
    virNetworkPtr net = NULL;
3494
    virObjectEventPtr event = NULL;
3495

3496
    if (!(def = virNetworkDefParseString(xml)))
3497
        goto cleanup;
3498

3499 3500 3501
    if (virNetworkCreateXMLEnsureACL(conn, def) < 0)
        goto cleanup;

3502
    if (networkValidate(driver, def) < 0)
J
Ján Tomko 已提交
3503
        goto cleanup;
3504

3505 3506 3507
    /* NB: even though this transient network hasn't yet been started,
     * we assign the def with live = true in anticipation that it will
     * be started momentarily.
3508
     */
3509 3510 3511
    if (!(obj = virNetworkObjAssignDef(driver->networks, def,
                                       VIR_NETWORK_OBJ_LIST_ADD_LIVE |
                                       VIR_NETWORK_OBJ_LIST_ADD_CHECK_LIVE)))
3512 3513
        goto cleanup;
    def = NULL;
3514

3515 3516
    if (networkStartNetwork(driver, obj) < 0) {
        virNetworkObjRemoveInactive(driver->networks, obj);
3517
        goto cleanup;
3518 3519
    }

3520 3521
    event = virNetworkEventLifecycleNew(obj->def->name,
                                        obj->def->uuid,
3522 3523
                                        VIR_NETWORK_EVENT_STARTED,
                                        0);
3524

3525 3526
    VIR_INFO("Creating network '%s'", obj->def->name);
    net = virGetNetwork(conn, obj->def->name, obj->def->uuid);
3527

3528
 cleanup:
3529
    virNetworkDefFree(def);
3530 3531
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
3532 3533
    virNetworkObjEndAPI(&obj);
    return net;
3534 3535
}

3536 3537 3538 3539

static virNetworkPtr
networkDefineXML(virConnectPtr conn,
                 const char *xml)
3540
{
3541
    virNetworkDriverStatePtr driver = networkGetDriver();
3542
    virNetworkDefPtr def = NULL;
3543
    bool freeDef = true;
3544 3545
    virNetworkObjPtr obj = NULL;
    virNetworkPtr net = NULL;
3546
    virObjectEventPtr event = NULL;
3547

3548
    if (!(def = virNetworkDefParseString(xml)))
3549
        goto cleanup;
3550

3551 3552 3553
    if (virNetworkDefineXMLEnsureACL(conn, def) < 0)
        goto cleanup;

3554
    if (networkValidate(driver, def) < 0)
J
Ján Tomko 已提交
3555
        goto cleanup;
3556

3557
    if (!(obj = virNetworkObjAssignDef(driver->networks, def, 0)))
J
Ján Tomko 已提交
3558
        goto cleanup;
3559

3560
    /* def was assigned to network object */
3561
    freeDef = false;
3562 3563

    if (virNetworkSaveConfig(driver->networkConfigDir, def) < 0) {
3564 3565
        if (!virNetworkObjIsActive(obj)) {
            virNetworkObjRemoveInactive(driver->networks, obj);
3566 3567
            goto cleanup;
        }
3568 3569 3570 3571
        /* if network was active already, just undo new persistent
         * definition by making it transient.
         * XXX - this isn't necessarily the correct thing to do.
         */
3572
        virNetworkObjUpdateAssignDef(obj, NULL, false);
3573 3574 3575
        goto cleanup;
    }

3576
    event = virNetworkEventLifecycleNew(def->name, def->uuid,
3577 3578
                                        VIR_NETWORK_EVENT_DEFINED,
                                        0);
3579

3580
    VIR_INFO("Defining network '%s'", def->name);
3581
    net = virGetNetwork(conn, def->name, def->uuid);
3582

3583
 cleanup:
3584 3585
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
3586
    if (freeDef)
J
Ján Tomko 已提交
3587
        virNetworkDefFree(def);
3588 3589
    virNetworkObjEndAPI(&obj);
    return net;
3590 3591
}

3592

3593
static int
3594 3595
networkUndefine(virNetworkPtr net)
{
3596
    virNetworkDriverStatePtr driver = networkGetDriver();
3597
    virNetworkObjPtr obj;
3598
    int ret = -1;
3599
    bool active = false;
3600
    virObjectEventPtr event = NULL;
3601

3602
    if (!(obj = networkObjFromNetwork(net)))
3603
        goto cleanup;
3604

3605
    if (virNetworkUndefineEnsureACL(net->conn, obj->def) < 0)
3606 3607
        goto cleanup;

3608
    if (virNetworkObjIsActive(obj))
3609
        active = true;
3610

3611
    if (!obj->persistent) {
3612 3613 3614 3615 3616
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("can't undefine transient network"));
        goto cleanup;
    }

3617
    /* remove autostart link */
3618 3619
    if (virNetworkObjDeleteConfig(driver->networkConfigDir,
                                  driver->networkAutostartDir,
3620
                                  obj) < 0)
3621
        goto cleanup;
3622

3623 3624
    event = virNetworkEventLifecycleNew(obj->def->name,
                                        obj->def->uuid,
3625 3626
                                        VIR_NETWORK_EVENT_UNDEFINED,
                                        0);
3627

3628
    VIR_INFO("Undefining network '%s'", obj->def->name);
3629
    if (!active) {
3630
        if (networkRemoveInactive(driver, obj) < 0)
3631
            goto cleanup;
3632 3633 3634 3635 3636
    } else {

        /* if the network still exists, it was active, and we need to make
         * it transient (by deleting the persistent def)
         */
3637
        virNetworkObjUpdateAssignDef(obj, NULL, false);
3638 3639
    }

3640
    ret = 0;
3641

3642
 cleanup:
3643 3644
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
3645
    virNetworkObjEndAPI(&obj);
3646
    return ret;
3647 3648
}

3649

3650 3651 3652 3653 3654 3655 3656 3657
static int
networkUpdate(virNetworkPtr net,
              unsigned int command,
              unsigned int section,
              int parentIndex,
              const char *xml,
              unsigned int flags)
{
3658
    virNetworkDriverStatePtr driver = networkGetDriver();
3659
    virNetworkObjPtr obj = NULL;
3660 3661
    int isActive, ret = -1;
    size_t i;
3662
    virNetworkIPDefPtr ipdef;
3663
    bool oldDhcpActive = false;
3664
    bool needFirewallRefresh = false;
3665

3666 3667 3668 3669 3670

    virCheckFlags(VIR_NETWORK_UPDATE_AFFECT_LIVE |
                  VIR_NETWORK_UPDATE_AFFECT_CONFIG,
                  -1);

3671
    if (!(obj = networkObjFromNetwork(net)))
3672 3673
        goto cleanup;

3674
    if (virNetworkUpdateEnsureACL(net->conn, obj->def, flags) < 0)
3675 3676
        goto cleanup;

3677
    /* see if we are listening for dhcp pre-modification */
3678
    for (i = 0;
3679
         (ipdef = virNetworkDefGetIPByIndex(obj->def, AF_INET, i));
3680
         i++) {
3681 3682 3683 3684 3685 3686
        if (ipdef->nranges || ipdef->nhosts) {
            oldDhcpActive = true;
            break;
        }
    }

3687 3688
    /* VIR_NETWORK_UPDATE_AFFECT_CURRENT means "change LIVE if network
     * is active, else change CONFIG
J
Ján Tomko 已提交
3689
     */
3690
    isActive = virNetworkObjIsActive(obj);
3691 3692
    if ((flags & (VIR_NETWORK_UPDATE_AFFECT_LIVE |
                  VIR_NETWORK_UPDATE_AFFECT_CONFIG)) ==
3693 3694 3695 3696 3697 3698 3699
        VIR_NETWORK_UPDATE_AFFECT_CURRENT) {
        if (isActive)
            flags |= VIR_NETWORK_UPDATE_AFFECT_LIVE;
        else
            flags |= VIR_NETWORK_UPDATE_AFFECT_CONFIG;
    }

3700 3701 3702 3703
    if (isActive && (flags & VIR_NETWORK_UPDATE_AFFECT_LIVE)) {
        /* Take care of anything that must be done before updating the
         * live NetworkDef.
         */
3704 3705 3706
        if (obj->def->forward.type == VIR_NETWORK_FORWARD_NONE ||
            obj->def->forward.type == VIR_NETWORK_FORWARD_NAT ||
            obj->def->forward.type == VIR_NETWORK_FORWARD_ROUTE) {
3707 3708 3709 3710 3711 3712 3713 3714 3715 3716
            switch (section) {
            case VIR_NETWORK_SECTION_FORWARD:
            case VIR_NETWORK_SECTION_FORWARD_INTERFACE:
            case VIR_NETWORK_SECTION_IP:
            case VIR_NETWORK_SECTION_IP_DHCP_RANGE:
            case VIR_NETWORK_SECTION_IP_DHCP_HOST:
                /* these could affect the firewall rules, so remove the
                 * old rules (and remember to load new ones after the
                 * update).
                 */
3717 3718
                if (obj->def->forward.type != VIR_NETWORK_FORWARD_OPEN) {
                    networkRemoveFirewallRules(obj->def);
3719 3720
                    needFirewallRefresh = true;
                }
3721 3722 3723 3724 3725 3726 3727
                break;
            default:
                break;
            }
        }
    }

3728
    /* update the network config in memory/on disk */
3729
    if (virNetworkObjUpdate(obj, command, section, parentIndex, xml, flags) < 0) {
3730
        if (needFirewallRefresh)
3731
            ignore_value(networkAddFirewallRules(obj->def));
3732 3733 3734
        goto cleanup;
    }

3735
    if (needFirewallRefresh && networkAddFirewallRules(obj->def) < 0)
3736 3737 3738 3739 3740
        goto cleanup;

    if (flags & VIR_NETWORK_UPDATE_AFFECT_CONFIG) {
        /* save updated persistent config to disk */
        if (virNetworkSaveConfig(driver->networkConfigDir,
3741
                                 virNetworkObjGetPersistentDef(obj)) < 0) {
3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754
            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 ||
3755 3756 3757 3758 3759 3760 3761 3762
            section == VIR_NETWORK_SECTION_IP_DHCP_RANGE ||
            section == VIR_NETWORK_SECTION_DNS_TXT ||
            section == VIR_NETWORK_SECTION_DNS_SRV) {
            /* these sections all change things on the dnsmasq
             * commandline (i.e. in the .conf file), so we need to
             * kill and restart dnsmasq, because dnsmasq sets its uid
             * to "nobody" after it starts, and is unable to re-read
             * the conf file (owned by root, mode 600)
3763
             */
3764
            if (networkRestartDhcpDaemon(driver, obj) < 0)
3765 3766
                goto cleanup;

3767 3768 3769 3770 3771 3772 3773 3774
        } 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;

3775
            for (i = 0;
3776
                 (ipdef = virNetworkDefGetIPByIndex(obj->def, AF_INET, i));
3777
                 i++) {
3778 3779 3780 3781 3782 3783 3784
                if (ipdef->nranges || ipdef->nhosts) {
                    newDhcpActive = true;
                    break;
                }
            }

            if ((newDhcpActive != oldDhcpActive &&
3785 3786
                 networkRestartDhcpDaemon(driver, obj) < 0) ||
                networkRefreshDhcpDaemon(driver, obj) < 0) {
3787 3788 3789
                goto cleanup;
            }

3790 3791 3792 3793
        } else if (section == VIR_NETWORK_SECTION_DNS_HOST) {
            /* this section only changes data in an external file
             * (not the .conf file) so we can just update the config
             * files and send SIGHUP to dnsmasq.
3794
             */
3795
            if (networkRefreshDhcpDaemon(driver, obj) < 0)
3796 3797 3798 3799 3800 3801 3802 3803
                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.
             */
3804
            if (networkRefreshRadvd(driver, obj) < 0)
3805 3806 3807 3808
                goto cleanup;
        }

        /* save current network state to disk */
3809
        if ((ret = virNetworkObjSaveStatus(driver->stateDir, obj)) < 0)
3810 3811
            goto cleanup;
    }
3812 3813

    /* call the 'updated' network hook script */
3814
    if (networkRunHook(obj, NULL, NULL, VIR_HOOK_NETWORK_OP_UPDATED,
3815 3816 3817
                       VIR_HOOK_SUBOP_BEGIN) < 0)
        goto cleanup;

3818
    ret = 0;
3819
 cleanup:
3820
    virNetworkObjEndAPI(&obj);
3821 3822 3823
    return ret;
}

3824 3825 3826

static int
networkCreate(virNetworkPtr net)
3827
{
3828
    virNetworkDriverStatePtr driver = networkGetDriver();
3829
    virNetworkObjPtr obj;
3830
    int ret = -1;
3831
    virObjectEventPtr event = NULL;
3832

3833
    if (!(obj = networkObjFromNetwork(net)))
3834
        goto cleanup;
3835

3836
    if (virNetworkCreateEnsureACL(net->conn, obj->def) < 0)
3837 3838
        goto cleanup;

3839
    if ((ret = networkStartNetwork(driver, obj)) < 0)
3840
        goto cleanup;
3841

3842 3843
    event = virNetworkEventLifecycleNew(obj->def->name,
                                        obj->def->uuid,
3844 3845
                                        VIR_NETWORK_EVENT_STARTED,
                                        0);
3846

3847
 cleanup:
3848 3849
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
3850
    virNetworkObjEndAPI(&obj);
3851
    return ret;
3852 3853
}

3854 3855 3856

static int
networkDestroy(virNetworkPtr net)
3857
{
3858
    virNetworkDriverStatePtr driver = networkGetDriver();
3859
    virNetworkObjPtr obj;
3860
    int ret = -1;
3861
    virObjectEventPtr event = NULL;
3862

3863
    if (!(obj = networkObjFromNetwork(net)))
3864
        goto cleanup;
3865

3866
    if (virNetworkDestroyEnsureACL(net->conn, obj->def) < 0)
3867 3868
        goto cleanup;

3869
    if (!virNetworkObjIsActive(obj)) {
3870
        virReportError(VIR_ERR_OPERATION_INVALID,
3871
                       _("network '%s' is not active"),
3872
                       obj->def->name);
3873 3874 3875
        goto cleanup;
    }

3876
    if ((ret = networkShutdownNetwork(driver, obj)) < 0)
3877 3878
        goto cleanup;

3879 3880
    event = virNetworkEventLifecycleNew(obj->def->name,
                                        obj->def->uuid,
3881 3882
                                        VIR_NETWORK_EVENT_STOPPED,
                                        0);
3883

3884
    if (!obj->persistent && networkRemoveInactive(driver, obj) < 0) {
3885 3886
        ret = -1;
        goto cleanup;
3887
    }
3888

3889
 cleanup:
3890 3891
    if (event)
        virObjectEventStateQueue(driver->networkEventState, event);
3892
    virNetworkObjEndAPI(&obj);
3893 3894 3895
    return ret;
}

3896 3897 3898 3899

static char *
networkGetXMLDesc(virNetworkPtr net,
                  unsigned int flags)
3900
{
3901
    virNetworkObjPtr obj;
3902
    virNetworkDefPtr def;
3903
    char *ret = NULL;
3904

3905
    virCheckFlags(VIR_NETWORK_XML_INACTIVE, NULL);
3906

3907
    if (!(obj = networkObjFromNetwork(net)))
3908
        return ret;
3909

3910
    if (virNetworkGetXMLDescEnsureACL(net->conn, obj->def) < 0)
3911 3912
        goto cleanup;

3913 3914
    if ((flags & VIR_NETWORK_XML_INACTIVE) && obj->newDef)
        def = obj->newDef;
3915
    else
3916
        def = obj->def;
3917 3918

    ret = virNetworkDefFormat(def, flags);
3919

3920
 cleanup:
3921
    virNetworkObjEndAPI(&obj);
3922
    return ret;
3923 3924
}

3925 3926 3927 3928

static char *
networkGetBridgeName(virNetworkPtr net)
{
3929
    virNetworkObjPtr obj;
3930 3931
    char *bridge = NULL;

3932
    if (!(obj = networkObjFromNetwork(net)))
3933
        return bridge;
3934

3935
    if (virNetworkGetBridgeNameEnsureACL(net->conn, obj->def) < 0)
3936 3937
        goto cleanup;

3938
    if (!(obj->def->bridge)) {
3939 3940
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("network '%s' does not have a bridge name."),
3941
                       obj->def->name);
3942 3943 3944
        goto cleanup;
    }

3945
    ignore_value(VIR_STRDUP(bridge, obj->def->bridge));
3946

3947
 cleanup:
3948
    virNetworkObjEndAPI(&obj);
3949 3950 3951
    return bridge;
}

3952 3953 3954 3955

static int
networkGetAutostart(virNetworkPtr net,
                    int *autostart)
3956
{
3957
    virNetworkObjPtr obj;
3958
    int ret = -1;
3959

3960
    if (!(obj = networkObjFromNetwork(net)))
3961
        return ret;
3962

3963
    if (virNetworkGetAutostartEnsureACL(net->conn, obj->def) < 0)
3964 3965
        goto cleanup;

3966
    *autostart = obj->autostart;
3967
    ret = 0;
3968

3969
 cleanup:
3970
    virNetworkObjEndAPI(&obj);
3971
    return ret;
3972 3973
}

3974 3975 3976 3977

static int
networkSetAutostart(virNetworkPtr net,
                    int autostart)
3978
{
3979
    virNetworkDriverStatePtr driver = networkGetDriver();
3980
    virNetworkObjPtr obj;
3981
    char *configFile = NULL, *autostartLink = NULL;
3982
    int ret = -1;
3983

3984

3985
    if (!(obj = networkObjFromNetwork(net)))
3986
        goto cleanup;
3987

3988
    if (virNetworkSetAutostartEnsureACL(net->conn, obj->def) < 0)
3989 3990
        goto cleanup;

3991
    if (!obj->persistent) {
3992 3993
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("cannot set autostart for transient network"));
3994 3995 3996
        goto cleanup;
    }

3997 3998
    autostart = (autostart != 0);

3999 4000
    if (obj->autostart != autostart) {
        if ((configFile = virNetworkConfigFile(driver->networkConfigDir, obj->def->name)) == NULL)
4001
            goto cleanup;
4002
        if ((autostartLink = virNetworkConfigFile(driver->networkAutostartDir, obj->def->name)) == NULL)
4003 4004
            goto cleanup;

4005
        if (autostart) {
4006
            if (virFileMakePath(driver->networkAutostartDir) < 0) {
4007
                virReportSystemError(errno,
4008 4009
                                     _("cannot create autostart directory '%s'"),
                                     driver->networkAutostartDir);
4010 4011
                goto cleanup;
            }
4012

4013
            if (symlink(configFile, autostartLink) < 0) {
4014
                virReportSystemError(errno,
4015
                                     _("Failed to create symlink '%s' to '%s'"),
4016
                                     autostartLink, configFile);
4017 4018 4019
                goto cleanup;
            }
        } else {
4020
            if (unlink(autostartLink) < 0 && errno != ENOENT && errno != ENOTDIR) {
4021
                virReportSystemError(errno,
4022
                                     _("Failed to delete symlink '%s'"),
4023
                                     autostartLink);
4024 4025
                goto cleanup;
            }
4026 4027
        }

4028
        obj->autostart = autostart;
4029
    }
4030
    ret = 0;
4031

4032
 cleanup:
4033 4034
    VIR_FREE(configFile);
    VIR_FREE(autostartLink);
4035
    virNetworkObjEndAPI(&obj);
4036
    return ret;
4037 4038
}

4039

4040
static int
4041
networkGetDHCPLeases(virNetworkPtr net,
4042 4043 4044
                     const char *mac,
                     virNetworkDHCPLeasePtr **leases,
                     unsigned int flags)
4045
{
4046
    virNetworkDriverStatePtr driver = networkGetDriver();
4047 4048 4049
    size_t i, j;
    size_t nleases = 0;
    int rv = -1;
4050
    ssize_t size = 0;
4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061
    int custom_lease_file_len = 0;
    bool need_results = !!leases;
    long long currtime = 0;
    long long expirytime_tmp = -1;
    bool ipv6 = false;
    char *lease_entries = NULL;
    char *custom_lease_file = NULL;
    const char *ip_tmp = NULL;
    const char *mac_tmp = NULL;
    virJSONValuePtr lease_tmp = NULL;
    virJSONValuePtr leases_array = NULL;
4062
    virNetworkIPDefPtr ipdef_tmp = NULL;
4063 4064
    virNetworkDHCPLeasePtr lease = NULL;
    virNetworkDHCPLeasePtr *leases_ret = NULL;
4065
    virNetworkObjPtr obj;
4066
    virMacAddr mac_addr;
4067 4068 4069

    virCheckFlags(0, -1);

4070 4071 4072 4073 4074 4075
    /* only to check if the MAC is valid */
    if (mac && virMacAddrParse(mac, &mac_addr) < 0) {
        virReportError(VIR_ERR_INVALID_MAC, "%s", mac);
        return -1;
    }

4076
    if (!(obj = networkObjFromNetwork(net)))
4077 4078
        return -1;

4079
    if (virNetworkGetDHCPLeasesEnsureACL(net->conn, obj->def) < 0)
4080
        goto cleanup;
4081 4082

    /* Retrieve custom leases file location */
4083
    custom_lease_file = networkDnsmasqLeaseFileNameCustom(driver, obj->def->bridge);
4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126

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

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

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

    currtime = (long long) time(NULL);

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

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

4127
        if (mac && virMacAddrCompare(mac, mac_tmp))
4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168
            continue;

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

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

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

            lease->expirytime = expirytime_tmp;

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

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

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

                if (ipv6 && VIR_SOCKET_ADDR_IS_FAMILY(&ipdef_tmp->address,
                                                      AF_INET6)) {
                    lease->prefix = ipdef_tmp->prefix;
                    break;
                }
                if (!ipv6 && VIR_SOCKET_ADDR_IS_FAMILY(&ipdef_tmp->address,
                                                      AF_INET)) {
4169
                    lease->prefix = virSocketAddrGetIPPrefix(&ipdef_tmp->address,
4170 4171 4172 4173 4174 4175 4176 4177
                                                             &ipdef_tmp->netmask,
                                                             ipdef_tmp->prefix);
                    break;
                }
            }

            if ((VIR_STRDUP(lease->mac, mac_tmp) < 0) ||
                (VIR_STRDUP(lease->ipaddr, ip_tmp) < 0) ||
4178
                (VIR_STRDUP(lease->iface, obj->def->bridge) < 0))
4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210
                goto error;

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

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

        } else {
            nleases++;
        }

        VIR_FREE(lease);
    }

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

    rv = nleases;

 cleanup:
    VIR_FREE(lease);
4211
    VIR_FREE(lease_entries);
4212 4213
    VIR_FREE(custom_lease_file);
    virJSONValueFree(leases_array);
4214

4215
    virNetworkObjEndAPI(&obj);
4216

4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227
    return rv;

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

4228 4229

static virNetworkDriver networkDriver = {
4230
    .name = "bridge",
4231 4232 4233 4234 4235
    .connectNumOfNetworks = networkConnectNumOfNetworks, /* 0.2.0 */
    .connectListNetworks = networkConnectListNetworks, /* 0.2.0 */
    .connectNumOfDefinedNetworks = networkConnectNumOfDefinedNetworks, /* 0.2.0 */
    .connectListDefinedNetworks = networkConnectListDefinedNetworks, /* 0.2.0 */
    .connectListAllNetworks = networkConnectListAllNetworks, /* 0.10.2 */
4236 4237
    .connectNetworkEventRegisterAny = networkConnectNetworkEventRegisterAny, /* 1.2.1 */
    .connectNetworkEventDeregisterAny = networkConnectNetworkEventDeregisterAny, /* 1.2.1 */
4238 4239
    .networkLookupByUUID = networkLookupByUUID, /* 0.2.0 */
    .networkLookupByName = networkLookupByName, /* 0.2.0 */
4240 4241
    .networkCreateXML = networkCreateXML, /* 0.2.0 */
    .networkDefineXML = networkDefineXML, /* 0.2.0 */
4242
    .networkUndefine = networkUndefine, /* 0.2.0 */
4243
    .networkUpdate = networkUpdate, /* 0.10.2 */
4244
    .networkCreate = networkCreate, /* 0.2.0 */
4245 4246 4247 4248 4249 4250 4251
    .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 */
4252
    .networkGetDHCPLeases = networkGetDHCPLeases, /* 1.2.6 */
4253 4254 4255
};

static virStateDriver networkStateDriver = {
4256
    .name = "bridge",
4257
    .stateInitialize  = networkStateInitialize,
4258
    .stateAutoStart  = networkStateAutoStart,
4259 4260
    .stateCleanup = networkStateCleanup,
    .stateReload = networkStateReload,
4261 4262
};

4263 4264
int
networkRegister(void)
4265
{
4266
    if (virSetSharedNetworkDriver(&networkDriver) < 0)
4267
        return -1;
4268 4269
    if (virRegisterStateDriver(&networkStateDriver) < 0)
        return -1;
4270 4271
    return 0;
}
4272

4273

4274 4275
/********************************************************/

4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310
/* A unified function to log network connections and disconnections */

static void
networkLogAllocation(virNetworkDefPtr netdef,
                     virDomainNetType actualType,
                     virNetworkForwardIfDefPtr dev,
                     virDomainNetDefPtr iface,
                     bool inUse)
{
    char macStr[VIR_MAC_STRING_BUFLEN];
    const char *verb = inUse ? "using" : "releasing";

    if (!dev) {
        VIR_INFO("MAC %s %s network %s (%d connections)",
                 virMacAddrFormat(&iface->mac, macStr), verb,
                 netdef->name, netdef->connections);
    } else {
        if (actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV) {
            VIR_INFO("MAC %s %s network %s (%d connections) "
                     "physical device %04x:%02x:%02x.%x (%d connections)",
                     virMacAddrFormat(&iface->mac, macStr), verb,
                     netdef->name, netdef->connections,
                     dev->device.pci.domain, dev->device.pci.bus,
                     dev->device.pci.slot, dev->device.pci.function,
                     dev->connections);
        } else {
            VIR_INFO("MAC %s %s network %s (%d connections) "
                     "physical device %s (%d connections)",
                     virMacAddrFormat(&iface->mac, macStr), verb,
                     netdef->name, netdef->connections,
                     dev->device.dev, dev->connections);
        }
    }
}

4311

4312 4313 4314 4315 4316 4317 4318 4319 4320
/* Private API to deal with logical switch capabilities.
 * These functions are exported so that other parts of libvirt can
 * call them, but are not part of the public API and not in the
 * driver's function table. If we ever have more than one network
 * driver, we will need to present these functions via a second
 * "backend" function table.
 */

/* networkAllocateActualDevice:
4321
 * @dom: domain definition that @iface belongs to
4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332
 * @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
4333 4334
networkAllocateActualDevice(virDomainDefPtr dom,
                            virDomainNetDefPtr iface)
4335
{
4336
    virNetworkDriverStatePtr driver = networkGetDriver();
4337
    virDomainNetType actualType = iface->type;
4338
    virNetworkObjPtr obj = NULL;
4339
    virNetworkDefPtr netdef = NULL;
4340
    virNetDevBandwidthPtr bandwidth = NULL;
4341 4342 4343
    virPortGroupDefPtr portgroup = NULL;
    virNetDevVPortProfilePtr virtport = iface->virtPortProfile;
    virNetDevVlanPtr vlan = NULL;
4344
    virNetworkForwardIfDefPtr dev = NULL;
4345
    size_t i;
4346 4347 4348
    int ret = -1;

    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
4349
        goto validate;
4350 4351 4352 4353

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

4354 4355
    obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
    if (!obj) {
4356 4357 4358
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
4359
        goto error;
4360
    }
4361
    netdef = obj->def;
4362

4363
    if (!virNetworkObjIsActive(obj)) {
4364 4365 4366 4367 4368 4369
        virReportError(VIR_ERR_OPERATION_INVALID,
                       _("network '%s' is not active"),
                       netdef->name);
        goto error;
    }

4370 4371 4372
    if (VIR_ALLOC(iface->data.network.actual) < 0)
        goto error;

4373 4374 4375
    /* portgroup can be present for any type of network, in particular
     * for bandwidth information, so we need to check for that and
     * fill it in appropriately for all forward types.
J
Ján Tomko 已提交
4376
     */
4377 4378 4379 4380 4381 4382
    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.
     */
4383 4384 4385 4386 4387 4388

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

4389 4390
    if (bandwidth && virNetDevBandwidthCopy(&iface->data.network.actual->bandwidth,
                                            bandwidth) < 0)
4391
        goto error;
4392

4393 4394 4395 4396 4397 4398 4399 4400
    /* copy appropriate vlan info to actualNet */
    if (iface->vlan.nTags > 0)
        vlan = &iface->vlan;
    else if (portgroup && portgroup->vlan.nTags > 0)
        vlan = &portgroup->vlan;
    else if (netdef->vlan.nTags > 0)
        vlan = &netdef->vlan;

4401 4402
    if (vlan && virNetDevVlanCopy(&iface->data.network.actual->vlan, vlan) < 0)
        goto error;
4403

4404 4405 4406 4407 4408 4409 4410 4411 4412 4413
    if (iface->trustGuestRxFilters)
       iface->data.network.actual->trustGuestRxFilters
          = iface->trustGuestRxFilters;
    else if (portgroup && portgroup->trustGuestRxFilters)
       iface->data.network.actual->trustGuestRxFilters
          = portgroup->trustGuestRxFilters;
    else if (netdef->trustGuestRxFilters)
       iface->data.network.actual->trustGuestRxFilters
          = netdef->trustGuestRxFilters;

4414 4415
    if ((netdef->forward.type == VIR_NETWORK_FORWARD_NONE) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_NAT) ||
4416 4417
        (netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_OPEN)) {
4418
        /* for these forward types, the actual net type really *is*
4419
         * NETWORK; we just keep the info from the portgroup in
4420
         * iface->data.network.actual
J
Ján Tomko 已提交
4421
         */
4422
        iface->data.network.actual->type = VIR_DOMAIN_NET_TYPE_NETWORK;
4423

4424
        /* we also store the bridge device and macTableManager settings
4425 4426 4427 4428 4429 4430 4431
         * in iface->data.network.actual->data.bridge for later use
         * after the domain's tap device is created (to attach to the
         * bridge and set flood/learning mode on the tap device)
         */
        if (VIR_STRDUP(iface->data.network.actual->data.bridge.brname,
                       netdef->bridge) < 0)
            goto error;
4432 4433
        iface->data.network.actual->data.bridge.macTableManager
           = netdef->macTableManager;
4434

4435
        if (networkPlugBandwidth(obj, iface) < 0)
4436 4437
            goto error;

4438
    } else if ((netdef->forward.type == VIR_NETWORK_FORWARD_BRIDGE) &&
4439
               netdef->bridge) {
4440 4441 4442 4443 4444

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

4445
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_BRIDGE;
4446 4447
        if (VIR_STRDUP(iface->data.network.actual->data.bridge.brname,
                       netdef->bridge) < 0)
4448
            goto error;
4449 4450
        iface->data.network.actual->data.bridge.macTableManager
           = netdef->macTableManager;
4451

4452 4453 4454 4455 4456 4457 4458 4459
        /* 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) {
4460
            goto error;
4461 4462 4463 4464 4465 4466 4467 4468 4469 4470
        }
        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);
4471
                goto error;
4472 4473 4474
            }
        }

4475
    } else if (netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
4476

4477
        virDomainHostdevSubsysPCIBackendType backend;
4478

4479
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_HOSTDEV;
4480
        if (networkCreateInterfacePool(netdef) < 0)
4481 4482 4483
            goto error;

        /* pick first dev with 0 connections */
4484 4485 4486
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].connections == 0) {
                dev = &netdef->forward.ifs[i];
4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500
                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;
4501
        iface->data.network.actual->data.hostdev.def.managed = netdef->forward.managed ? 1 : 0;
4502
        iface->data.network.actual->data.hostdev.def.source.subsys.type = dev->type;
4503
        iface->data.network.actual->data.hostdev.def.source.subsys.u.pci.addr = dev->device.pci;
4504

E
Eric Blake 已提交
4505
        switch (netdef->forward.driverName) {
4506
        case VIR_NETWORK_FORWARD_DRIVER_NAME_DEFAULT:
4507
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_DEFAULT;
4508 4509
            break;
        case VIR_NETWORK_FORWARD_DRIVER_NAME_KVM:
4510
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_KVM;
4511 4512
            break;
        case VIR_NETWORK_FORWARD_DRIVER_NAME_VFIO:
4513
            backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_VFIO;
4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524
            break;
        default:
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("unrecognized driver name value %d "
                             " in network '%s'"),
                           netdef->forward.driverName, netdef->name);
            goto error;
        }
        iface->data.network.actual->data.hostdev.def.source.subsys.u.pci.backend
            = backend;

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
        /* 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;
            }
        }

4550 4551 4552 4553
    } 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)) {
4554 4555 4556 4557 4558 4559

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

        /* Set type=direct and appropriate <source mode='xxx'/> */
4560
        iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_DIRECT;
4561
        switch (netdef->forward.type) {
4562
        case VIR_NETWORK_FORWARD_BRIDGE:
4563
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_BRIDGE;
4564 4565
            break;
        case VIR_NETWORK_FORWARD_PRIVATE:
4566
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PRIVATE;
4567 4568
            break;
        case VIR_NETWORK_FORWARD_VEPA:
4569
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_VEPA;
4570 4571
            break;
        case VIR_NETWORK_FORWARD_PASSTHROUGH:
4572
            iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PASSTHRU;
4573 4574 4575
            break;
        }

4576 4577 4578 4579 4580 4581 4582 4583
        /* 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) {
4584
            goto error;
4585
        }
4586
        virtport = iface->data.network.actual->virtPortProfile;
4587
        if (virtport) {
4588 4589 4590 4591 4592 4593 4594 4595
            /* 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);
4596
                goto error;
4597 4598
            }
        }
4599

4600 4601 4602
        /* 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).
         */
4603
        if ((netdef->forward.nifs <= 0) && (netdef->forward.npfs <= 0)) {
4604 4605 4606 4607
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' uses a direct mode, but "
                             "has no forward dev and no interface pool"),
                           netdef->name);
4608
            goto error;
4609 4610 4611
        } else {
            /* pick an interface from the pool */

4612
            if (networkCreateInterfacePool(netdef) < 0)
4613 4614
                goto error;

4615 4616 4617 4618 4619
            /* 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.
4620
             */
4621 4622
            if ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
                ((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
4623 4624 4625
                 iface->data.network.actual->virtPortProfile &&
                 (iface->data.network.actual->virtPortProfile->virtPortType
                  == VIR_NETDEV_VPORT_PROFILE_8021QBH))) {
4626

4627
                /* pick first dev with 0 connections */
4628 4629 4630
                for (i = 0; i < netdef->forward.nifs; i++) {
                    if (netdef->forward.ifs[i].connections == 0) {
                        dev = &netdef->forward.ifs[i];
4631 4632 4633 4634 4635
                        break;
                    }
                }
            } else {
                /* pick least used dev */
4636
                dev = &netdef->forward.ifs[0];
4637 4638 4639
                for (i = 1; i < netdef->forward.nifs; i++) {
                    if (netdef->forward.ifs[i].connections < dev->connections)
                        dev = &netdef->forward.ifs[i];
4640 4641 4642 4643
                }
            }
            /* dev points at the physical device we want to use */
            if (!dev) {
4644 4645 4646 4647
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               _("network '%s' requires exclusive access "
                                 "to interfaces, but none are available"),
                               netdef->name);
4648
                goto error;
4649
            }
4650 4651
            if (VIR_STRDUP(iface->data.network.actual->data.direct.linkdev,
                           dev->device.dev) < 0)
4652
                goto error;
4653 4654 4655
        }
    }

4656
    if (networkMacMgrAdd(driver, obj, dom->name, &iface->mac) < 0)
M
Michal Privoznik 已提交
4657 4658
        goto error;

4659
    if (virNetDevVPortProfileCheckComplete(virtport, true) < 0)
4660
        goto error;
4661

4662
 validate:
4663 4664 4665 4666 4667
    /* 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.
     */

4668
    if (virDomainNetGetActualVlan(iface)) {
4669 4670 4671 4672
        /* vlan configuration via libvirt is only supported for PCI
         * Passthrough SR-IOV devices (hostdev or macvtap passthru
         * mode) and openvswitch bridges. Otherwise log an error and
         * fail
4673 4674
         */
        if (!(actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV ||
4675 4676 4677
              (actualType == VIR_DOMAIN_NET_TYPE_DIRECT &&
               virDomainNetGetActualDirectMode(iface)
               == VIR_NETDEV_MACVLAN_MODE_PASSTHRU) ||
4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696
              (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;
        }
    }
4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707
    if (virDomainNetGetActualBandwidth(iface)) {
        /* bandwidth configuration via libvirt is not supported for
         * hostdev network devices
         */
        if (actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV) {
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("bandwidth settings are not supported "
                             "for hostdev interfaces"));
            goto error;
        }
    }
4708 4709 4710

    if (netdef) {
        netdef->connections++;
4711
        if (dev)
4712 4713
            dev->connections++;
        /* finally we can call the 'plugged' hook script if any */
4714
        if (networkRunHook(obj, dom, iface,
4715 4716 4717
                           VIR_HOOK_NETWORK_OP_IFACE_PLUGGED,
                           VIR_HOOK_SUBOP_BEGIN) < 0) {
            /* adjust for failure */
4718
            netdef->connections--;
4719 4720 4721 4722
            if (dev)
                dev->connections--;
            goto error;
        }
4723
        networkLogAllocation(netdef, actualType, dev, iface, true);
4724 4725
    }

4726
    ret = 0;
4727

4728
 cleanup:
4729
    virNetworkObjEndAPI(&obj);
4730 4731
    return ret;

4732
 error:
4733
    if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
4734 4735 4736
        virDomainActualNetDefFree(iface->data.network.actual);
        iface->data.network.actual = NULL;
    }
4737
    goto cleanup;
4738 4739
}

4740

4741
/* networkNotifyActualDevice:
4742
 * @dom: domain definition that @iface belongs to
4743 4744 4745 4746 4747
 * @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
4748 4749
 * order, or re-attach the interface's tap device to the network's
 * bridge.
4750
 *
4751
 * No return value (but does log any failures)
4752
 */
4753
void
4754 4755
networkNotifyActualDevice(virDomainDefPtr dom,
                          virDomainNetDefPtr iface)
4756
{
4757
    virNetworkDriverStatePtr driver = networkGetDriver();
4758
    virDomainNetType actualType = virDomainNetGetActualType(iface);
4759
    virNetworkObjPtr obj;
4760
    virNetworkDefPtr netdef;
4761
    virNetworkForwardIfDefPtr dev = NULL;
4762
    size_t i;
4763
    char *master = NULL;
4764 4765

    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
4766
        return;
4767

4768 4769
    obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
    if (!obj) {
4770 4771 4772
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
4773 4774
        goto error;
    }
4775
    netdef = obj->def;
4776

4777
    if (!virNetworkObjIsActive(obj)) {
4778 4779 4780 4781 4782 4783
        virReportError(VIR_ERR_OPERATION_INVALID,
                       _("network '%s' is not active"),
                       netdef->name);
        goto error;
    }

4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794
    /* if we're restarting libvirtd after an upgrade from a version
     * that didn't save bridge name in actualNetDef for
     * actualType==network, we need to copy it in so that it will be
     * available in all cases
     */
    if (actualType == VIR_DOMAIN_NET_TYPE_NETWORK &&
        !iface->data.network.actual->data.bridge.brname &&
        (VIR_STRDUP(iface->data.network.actual->data.bridge.brname,
                    netdef->bridge) < 0))
            goto error;

4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819
    /* see if we're connected to the correct bridge */
    if (netdef->bridge) {
        if (virNetDevGetMaster(iface->ifname, &master) < 0)
            goto error;

        if (STRNEQ_NULLABLE(netdef->bridge, master)) {
            /* disconnect from current (incorrect) bridge */
            if (master)
                ignore_value(virNetDevBridgeRemovePort(master, iface->ifname));

            /* attach/reattach to correct bridge.
             * NB: we can't notify the guest of any MTU change anyway,
             * so there is no point in trying to learn the actualMTU
             * (final arg to virNetDevTapAttachBridge())
             */
            if (virNetDevTapAttachBridge(iface->ifname, netdef->bridge,
                                         &iface->mac, dom->uuid,
                                         virDomainNetGetActualVirtPortProfile(iface),
                                         virDomainNetGetActualVlan(iface),
                                         iface->mtu, NULL) < 0) {
                goto error;
            }
        }
    }

4820
    if (!iface->data.network.actual ||
4821 4822
        (actualType != VIR_DOMAIN_NET_TYPE_DIRECT &&
         actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV)) {
4823 4824
        VIR_DEBUG("Nothing to claim from network %s", iface->data.network.name);
        goto success;
4825 4826
    }

4827
    if (networkCreateInterfacePool(netdef) < 0)
4828
        goto error;
4829

4830
    if (netdef->forward.nifs == 0) {
4831
        virReportError(VIR_ERR_INTERNAL_ERROR,
4832 4833
                       _("network '%s' uses a direct or hostdev mode, "
                         "but has no forward dev and no interface pool"),
4834
                       netdef->name);
4835
        goto error;
4836
    }
4837

4838 4839
    if (actualType == VIR_DOMAIN_NET_TYPE_DIRECT) {
        const char *actualDev;
4840

4841 4842 4843 4844 4845 4846 4847 4848 4849
        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 */
4850 4851
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
4852
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
4853 4854
                STREQ(actualDev, netdef->forward.ifs[i].device.dev)) {
                dev = &netdef->forward.ifs[i];
4855 4856 4857 4858 4859
                break;
            }
        }
        /* dev points at the physical device we want to use */
        if (!dev) {
4860
            virReportError(VIR_ERR_INTERNAL_ERROR,
4861 4862
                           _("network '%s' doesn't have dev='%s' "
                             "in use by domain"),
4863
                           netdef->name, actualDev);
4864
            goto error;
4865 4866
        }

4867
        /* PASSTHROUGH mode and PRIVATE Mode + 802.1Qbh both require
4868 4869
         * exclusive access to a device, so current connections count
         * must be 0 in those cases.
4870
         */
4871
        if ((dev->connections > 0) &&
4872 4873
            ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
             ((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
4874 4875
              iface->data.network.actual->virtPortProfile &&
              (iface->data.network.actual->virtPortProfile->virtPortType
4876
               == VIR_NETDEV_VPORT_PROFILE_8021QBH)))) {
4877
            virReportError(VIR_ERR_INTERNAL_ERROR,
4878 4879
                           _("network '%s' claims dev='%s' is already in "
                             "use by a different domain"),
4880
                           netdef->name, actualDev);
4881
            goto error;
4882
        }
4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894
    }  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 */
4895 4896
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
4897
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
4898
                virPCIDeviceAddressEqual(&hostdev->source.subsys.u.pci.addr,
4899 4900
                                         &netdef->forward.ifs[i].device.pci)) {
                dev = &netdef->forward.ifs[i];
4901 4902 4903 4904 4905 4906 4907 4908 4909
                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,
4910 4911 4912 4913
                           hostdev->source.subsys.u.pci.addr.domain,
                           hostdev->source.subsys.u.pci.addr.bus,
                           hostdev->source.subsys.u.pci.addr.slot,
                           hostdev->source.subsys.u.pci.addr.function);
J
Ján Tomko 已提交
4914
            goto error;
4915 4916 4917 4918 4919 4920 4921
        }

        /* 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) &&
4922
            netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
4923 4924 4925 4926 4927 4928 4929 4930 4931
            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;
        }
4932 4933
    }

4934
 success:
4935
    netdef->connections++;
4936 4937
    if (dev)
        dev->connections++;
4938
    /* finally we can call the 'plugged' hook script if any */
4939
    if (networkRunHook(obj, dom, iface, VIR_HOOK_NETWORK_OP_IFACE_PLUGGED,
4940 4941 4942 4943 4944 4945 4946
                       VIR_HOOK_SUBOP_BEGIN) < 0) {
        /* adjust for failure */
        if (dev)
            dev->connections--;
        netdef->connections--;
        goto error;
    }
4947
    networkLogAllocation(netdef, actualType, dev, iface, true);
4948

4949
 cleanup:
4950
    virNetworkObjEndAPI(&obj);
4951
    VIR_FREE(master);
4952
    return;
4953

4954
 error:
4955
    goto cleanup;
4956 4957 4958 4959
}


/* networkReleaseActualDevice:
4960
 * @dom: domain definition that @iface belongs to
4961 4962 4963 4964 4965 4966 4967 4968 4969 4970
 * @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
4971 4972
networkReleaseActualDevice(virDomainDefPtr dom,
                           virDomainNetDefPtr iface)
4973
{
4974
    virNetworkDriverStatePtr driver = networkGetDriver();
4975
    virDomainNetType actualType = virDomainNetGetActualType(iface);
4976
    virNetworkObjPtr obj;
4977
    virNetworkDefPtr netdef;
4978
    virNetworkForwardIfDefPtr dev = NULL;
4979 4980
    size_t i;
    int ret = -1;
4981 4982

    if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
J
Ján Tomko 已提交
4983
        return 0;
4984

4985 4986
    obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
    if (!obj) {
4987 4988 4989
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
4990 4991
        goto error;
    }
4992
    netdef = obj->def;
4993

4994 4995
    if (iface->data.network.actual &&
        (netdef->forward.type == VIR_NETWORK_FORWARD_NONE ||
4996
         netdef->forward.type == VIR_NETWORK_FORWARD_NAT ||
4997 4998
         netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE ||
         netdef->forward.type == VIR_NETWORK_FORWARD_OPEN) &&
4999
        networkUnplugBandwidth(obj, iface) < 0)
5000 5001
        goto error;

5002 5003 5004
    if ((!iface->data.network.actual) ||
        ((actualType != VIR_DOMAIN_NET_TYPE_DIRECT) &&
         (actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV))) {
5005 5006
        VIR_DEBUG("Nothing to release to network %s", iface->data.network.name);
        goto success;
5007 5008
    }

5009
    if (netdef->forward.nifs == 0) {
5010
        virReportError(VIR_ERR_INTERNAL_ERROR,
5011
                       _("network '%s' uses a direct/hostdev mode, but "
5012 5013
                         "has no forward dev and no interface pool"),
                       netdef->name);
5014
        goto error;
5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026
    }

    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;
        }
5027

5028 5029
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
5030
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
5031 5032
                STREQ(actualDev, netdef->forward.ifs[i].device.dev)) {
                dev = &netdef->forward.ifs[i];
5033 5034 5035
                break;
            }
        }
5036

5037
        if (!dev) {
5038
            virReportError(VIR_ERR_INTERNAL_ERROR,
5039 5040
                           _("network '%s' doesn't have dev='%s' "
                             "in use by domain"),
5041
                           netdef->name, actualDev);
5042
            goto error;
5043
        }
5044 5045 5046 5047 5048 5049 5050 5051 5052 5053
    } 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;
        }

5054 5055
        for (i = 0; i < netdef->forward.nifs; i++) {
            if (netdef->forward.ifs[i].type
5056
                == VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
5057
                virPCIDeviceAddressEqual(&hostdev->source.subsys.u.pci.addr,
5058 5059
                                         &netdef->forward.ifs[i].device.pci)) {
                dev = &netdef->forward.ifs[i];
5060 5061 5062 5063 5064 5065 5066 5067 5068
                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,
5069 5070 5071 5072
                           hostdev->source.subsys.u.pci.addr.domain,
                           hostdev->source.subsys.u.pci.addr.bus,
                           hostdev->source.subsys.u.pci.addr.slot,
                           hostdev->source.subsys.u.pci.addr.function);
J
Ján Tomko 已提交
5073
            goto error;
5074
        }
J
Ján Tomko 已提交
5075
    }
5076

5077
 success:
5078
    networkMacMgrDel(driver, obj, dom->name, &iface->mac);
M
Michal Privoznik 已提交
5079

5080
    if (iface->data.network.actual) {
5081
        netdef->connections--;
5082 5083
        if (dev)
            dev->connections--;
5084
        /* finally we can call the 'unplugged' hook script if any */
5085
        networkRunHook(obj, dom, iface, VIR_HOOK_NETWORK_OP_IFACE_UNPLUGGED,
5086
                       VIR_HOOK_SUBOP_BEGIN);
5087
        networkLogAllocation(netdef, actualType, dev, iface, false);
5088
    }
5089
    ret = 0;
5090
 cleanup:
5091
    virNetworkObjEndAPI(&obj);
5092 5093 5094 5095
    if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
        virDomainActualNetDefFree(iface->data.network.actual);
        iface->data.network.actual = NULL;
    }
5096
    return ret;
5097

5098
 error:
5099
    goto cleanup;
5100
}
5101

5102

5103 5104 5105 5106 5107
/*
 * networkGetNetworkAddress:
 * @netname: the name of a network
 * @netaddr: string representation of IP address for that network.
 *
5108
 * Attempt to return an IP address associated with the named
5109 5110 5111 5112
 * 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.
 *
5113
 * Note: This function returns the first IP address it finds. It might
5114 5115 5116 5117 5118 5119 5120 5121 5122
 * 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
5123 5124
networkGetNetworkAddress(const char *netname,
                         char **netaddr)
5125
{
5126
    virNetworkDriverStatePtr driver = networkGetDriver();
5127
    int ret = -1;
5128
    virNetworkObjPtr obj;
5129
    virNetworkDefPtr netdef;
5130
    virNetworkIPDefPtr ipdef;
5131 5132
    virSocketAddr addr;
    virSocketAddrPtr addrptr = NULL;
5133
    char *dev_name = NULL;
5134 5135

    *netaddr = NULL;
5136 5137
    obj = virNetworkObjFindByName(driver->networks, netname);
    if (!obj) {
5138 5139 5140
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       netname);
5141
        goto cleanup;
5142
    }
5143
    netdef = obj->def;
5144

5145
    switch (netdef->forward.type) {
5146 5147 5148
    case VIR_NETWORK_FORWARD_NONE:
    case VIR_NETWORK_FORWARD_NAT:
    case VIR_NETWORK_FORWARD_ROUTE:
5149
    case VIR_NETWORK_FORWARD_OPEN:
5150
        ipdef = virNetworkDefGetIPByIndex(netdef, AF_UNSPEC, 0);
5151
        if (!ipdef) {
5152
            virReportError(VIR_ERR_INTERNAL_ERROR,
5153
                           _("network '%s' doesn't have an IP address"),
5154
                           netdef->name);
5155
            goto cleanup;
5156 5157 5158 5159 5160
        }
        addrptr = &ipdef->address;
        break;

    case VIR_NETWORK_FORWARD_BRIDGE:
5161
        if ((dev_name = netdef->bridge))
5162 5163
            break;
        /*
5164 5165
         * fall through if netdef->bridge wasn't set, since that is
         * macvtap bridge mode network.
5166
         */
5167 5168
        ATTRIBUTE_FALLTHROUGH;

5169 5170 5171
    case VIR_NETWORK_FORWARD_PRIVATE:
    case VIR_NETWORK_FORWARD_VEPA:
    case VIR_NETWORK_FORWARD_PASSTHROUGH:
5172 5173
        if ((netdef->forward.nifs > 0) && netdef->forward.ifs)
            dev_name = netdef->forward.ifs[0].device.dev;
5174

5175
        if (!dev_name) {
5176 5177 5178
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("network '%s' has no associated interface or bridge"),
                           netdef->name);
5179
            goto cleanup;
5180 5181 5182 5183
        }
        break;
    }

5184
    if (dev_name) {
5185
        if (virNetDevIPAddrGet(dev_name, &addr) < 0)
5186
            goto cleanup;
5187
        addrptr = &addr;
5188 5189
    }

5190 5191
    if (!(addrptr &&
          (*netaddr = virSocketAddrFormat(addrptr)))) {
5192
        goto cleanup;
5193 5194
    }

5195
    ret = 0;
5196
 cleanup:
5197
    virNetworkObjEndAPI(&obj);
5198 5199
    return ret;
}
5200

5201

5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214
/* networkGetActualType:
 * @dom: domain definition that @iface belongs to
 * @iface: the original NetDef from the domain
 *
 * Looks up the network reference by iface, and returns the actual
 * type of the connection without allocating any resources.
 *
 * Returns 0 on success, -1 on failure.
 */
int
networkGetActualType(virDomainNetDefPtr iface)
{
    virNetworkDriverStatePtr driver = networkGetDriver();
5215
    virNetworkObjPtr obj = NULL;
5216 5217 5218 5219 5220 5221 5222 5223 5224
    virNetworkDefPtr netdef = NULL;
    int ret = -1;

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

    if (iface->data.network.actual)
        return iface->data.network.actual->type;

5225 5226
    obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
    if (!obj) {
5227 5228 5229 5230 5231
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
        return -1;
    }
5232
    netdef = obj->def;
5233 5234 5235

    if ((netdef->forward.type == VIR_NETWORK_FORWARD_NONE) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_NAT) ||
5236 5237
        (netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE) ||
        (netdef->forward.type == VIR_NETWORK_FORWARD_OPEN)) {
5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269
        /* for these forward types, the actual net type really *is*
         * NETWORK; we just keep the info from the portgroup in
         * iface->data.network.actual
         */
        ret = VIR_DOMAIN_NET_TYPE_NETWORK;

    } else if ((netdef->forward.type == VIR_NETWORK_FORWARD_BRIDGE) &&
               netdef->bridge) {

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

        ret = VIR_DOMAIN_NET_TYPE_BRIDGE;

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

        ret = VIR_DOMAIN_NET_TYPE_HOSTDEV;

    } 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)) {

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

        ret = VIR_DOMAIN_NET_TYPE_DIRECT;

    }

5270
    virNetworkObjEndAPI(&obj);
5271 5272 5273 5274
    return ret;
}


5275 5276 5277
/**
 * networkCheckBandwidth:
 * @net: network QoS
5278
 * @ifaceBand: interface QoS (may be NULL if no QoS)
5279
 * @oldBandwidth: new interface QoS (may be NULL if no QoS)
5280
 * @ifaceMac: interface MAC (used in error messages for identification)
5281 5282
 * @new_rate: new rate for non guaranteed class
 *
5283 5284 5285 5286 5287 5288 5289 5290
 * Function checks if @ifaceBand can be satisfied on @net. However, sometimes it
 * may happen that the interface that @ifaceBand corresponds to is already
 * plugged into the @net and the bandwidth is to be updated. In that case we
 * need to check if new bandwidth can be satisfied. If that's the case
 * @ifaceBand should point to new bandwidth settings and @oldBandwidth to
 * current ones. If you want to suppress this functionality just pass
 * @oldBandwidth == NULL.
 *
5291 5292 5293 5294 5295
 * 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
5296
networkCheckBandwidth(virNetworkObjPtr obj,
5297
                      virNetDevBandwidthPtr ifaceBand,
5298
                      virNetDevBandwidthPtr oldBandwidth,
5299
                      virMacAddr ifaceMac,
5300 5301 5302
                      unsigned long long *new_rate)
{
    int ret = -1;
5303 5304
    virNetDevBandwidthPtr netBand = obj->def->bandwidth;
    unsigned long long tmp_floor_sum = obj->floor_sum;
5305 5306 5307
    unsigned long long tmp_new_rate = 0;
    char ifmac[VIR_MAC_STRING_BUFLEN];

5308
    virMacAddrFormat(&ifaceMac, ifmac);
5309 5310 5311 5312 5313 5314

    if (ifaceBand && ifaceBand->in && ifaceBand->in->floor &&
        !(netBand && netBand->in)) {
        virReportError(VIR_ERR_OPERATION_UNSUPPORTED,
                       _("Invalid use of 'floor' on interface with MAC "
                         "address %s - network '%s' has no inbound QoS set"),
5315
                       ifmac, obj->def->name);
5316 5317 5318
        return -1;
    }

5319 5320
    if (((!ifaceBand || !ifaceBand->in || !ifaceBand->in->floor) &&
         (!oldBandwidth || !oldBandwidth->in || !oldBandwidth->in->floor)) ||
5321 5322
        !netBand || !netBand->in) {
        /* no QoS required, claim success */
5323
        return 1;
5324
    }
5325 5326

    tmp_new_rate = netBand->in->average;
5327 5328 5329 5330
    if (oldBandwidth && oldBandwidth->in)
        tmp_floor_sum -= oldBandwidth->in->floor;
    if (ifaceBand && ifaceBand->in)
        tmp_floor_sum += ifaceBand->in->floor;
5331 5332 5333 5334 5335 5336 5337 5338 5339

    /* 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,
5340 5341
                           obj->def->bridge,
                           obj->def->name);
5342 5343 5344 5345 5346 5347 5348 5349 5350
            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,
5351 5352
                       obj->def->bridge,
                       obj->def->name);
5353 5354 5355
        goto cleanup;
    }

5356 5357
    if (new_rate)
        *new_rate = tmp_new_rate;
5358 5359
    ret = 0;

5360
 cleanup:
5361 5362 5363
    return ret;
}

5364

5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375
/**
 * 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
5376
networkNextClassID(virNetworkObjPtr obj)
5377
{
5378
    ssize_t ret = 0;
5379

5380
    ret = virBitmapNextClearBit(obj->class_id, -1);
5381

5382
    if (ret < 0 || virBitmapSetBit(obj->class_id, ret) < 0)
5383 5384 5385 5386 5387
        return -1;

    return ret;
}

5388

5389
static int
5390
networkPlugBandwidthImpl(virNetworkObjPtr obj,
5391 5392 5393
                         virDomainNetDefPtr iface,
                         virNetDevBandwidthPtr ifaceBand,
                         unsigned long long new_rate)
5394
{
5395
    virNetworkDriverStatePtr driver = networkGetDriver();
5396
    ssize_t class_id = 0;
5397 5398
    int plug_ret;
    int ret = -1;
5399 5400

    /* generate new class_id */
5401
    if ((class_id = networkNextClassID(obj)) < 0) {
5402 5403 5404 5405 5406
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not generate next class ID"));
        goto cleanup;
    }

5407
    plug_ret = virNetDevBandwidthPlug(obj->def->bridge, obj->def->bandwidth,
5408
                                      &iface->mac, ifaceBand, class_id);
5409
    if (plug_ret < 0) {
5410
        ignore_value(virNetDevBandwidthUnplug(obj->def->bridge, class_id));
5411 5412 5413 5414 5415 5416
        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 */
5417
    obj->floor_sum += ifaceBand->in->floor;
5418
    /* update status file */
5419 5420 5421
    if (virNetworkObjSaveStatus(driver->stateDir, obj) < 0) {
        ignore_value(virBitmapClearBit(obj->class_id, class_id));
        obj->floor_sum -= ifaceBand->in->floor;
5422
        iface->data.network.actual->class_id = 0;
5423
        ignore_value(virNetDevBandwidthUnplug(obj->def->bridge, class_id));
5424 5425
        goto cleanup;
    }
5426
    /* update rate for non guaranteed NICs */
5427 5428 5429
    new_rate -= obj->floor_sum;
    if (virNetDevBandwidthUpdateRate(obj->def->bridge, 2,
                                     obj->def->bandwidth, new_rate) < 0)
5430
        VIR_WARN("Unable to update rate for 1:2 class on %s bridge",
5431
                 obj->def->bridge);
5432 5433

    ret = 0;
5434 5435 5436 5437 5438 5439
 cleanup:
    return ret;
}


static int
5440
networkPlugBandwidth(virNetworkObjPtr obj,
5441 5442 5443 5444 5445 5446 5447 5448
                     virDomainNetDefPtr iface)
{
    int ret = -1;
    int plug_ret;
    unsigned long long new_rate = 0;
    char ifmac[VIR_MAC_STRING_BUFLEN];
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);

5449
    if ((plug_ret = networkCheckBandwidth(obj, ifaceBand, NULL,
5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469
                                          iface->mac, &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;
    }

5470
    if (networkPlugBandwidthImpl(obj, iface, ifaceBand, new_rate) < 0)
5471 5472 5473
        goto cleanup;

    ret = 0;
5474

5475
 cleanup:
5476 5477 5478
    return ret;
}

5479

5480
static int
5481
networkUnplugBandwidth(virNetworkObjPtr obj,
5482 5483
                       virDomainNetDefPtr iface)
{
5484
    virNetworkDriverStatePtr driver = networkGetDriver();
5485 5486
    int ret = 0;
    unsigned long long new_rate;
5487
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
5488 5489 5490

    if (iface->data.network.actual &&
        iface->data.network.actual->class_id) {
5491
        if (!obj->def->bandwidth || !obj->def->bandwidth->in) {
5492
            VIR_WARN("Network %s has no bandwidth but unplug requested",
5493
                     obj->def->name);
5494 5495
            goto cleanup;
        }
5496
        /* we must remove class from bridge */
5497
        new_rate = obj->def->bandwidth->in->average;
5498

5499 5500
        if (obj->def->bandwidth->in->peak > 0)
            new_rate = obj->def->bandwidth->in->peak;
5501

5502
        ret = virNetDevBandwidthUnplug(obj->def->bridge,
5503 5504 5505 5506
                                       iface->data.network.actual->class_id);
        if (ret < 0)
            goto cleanup;
        /* update sum of 'floor'-s of attached NICs */
5507
        obj->floor_sum -= ifaceBand->in->floor;
5508
        /* return class ID */
5509
        ignore_value(virBitmapClearBit(obj->class_id,
5510 5511
                                       iface->data.network.actual->class_id));
        /* update status file */
5512 5513 5514
        if (virNetworkObjSaveStatus(driver->stateDir, obj) < 0) {
            obj->floor_sum += ifaceBand->in->floor;
            ignore_value(virBitmapSetBit(obj->class_id,
5515 5516 5517
                                         iface->data.network.actual->class_id));
            goto cleanup;
        }
5518
        /* update rate for non guaranteed NICs */
5519 5520 5521
        new_rate -= obj->floor_sum;
        if (virNetDevBandwidthUpdateRate(obj->def->bridge, 2,
                                         obj->def->bandwidth, new_rate) < 0)
5522
            VIR_WARN("Unable to update rate for 1:2 class on %s bridge",
5523
                     obj->def->bridge);
5524 5525 5526 5527
        /* no class is associated any longer */
        iface->data.network.actual->class_id = 0;
    }

5528
 cleanup:
5529 5530
    return ret;
}
5531

5532

5533
static void
5534
networkNetworkObjTaint(virNetworkObjPtr obj,
5535
                       virNetworkTaintFlags taint)
5536
{
5537
    if (virNetworkObjTaint(obj, taint)) {
5538
        char uuidstr[VIR_UUID_STRING_BUFLEN];
5539
        virUUIDFormat(obj->def->uuid, uuidstr);
5540 5541

        VIR_WARN("Network name='%s' uuid=%s is tainted: %s",
5542
                 obj->def->name,
5543 5544 5545 5546
                 uuidstr,
                 virNetworkTaintTypeToString(taint));
    }
}
5547 5548 5549 5550 5551 5552


static bool
networkBandwidthGenericChecks(virDomainNetDefPtr iface,
                              virNetDevBandwidthPtr newBandwidth)
{
5553
    virNetDevBandwidthPtr ifaceBand;
5554 5555 5556 5557 5558 5559 5560 5561
    unsigned long long old_floor, new_floor;

    if (virDomainNetGetActualType(iface) != VIR_DOMAIN_NET_TYPE_NETWORK) {
        /* This is not an interface that's plugged into a network.
         * We don't care. Thus from our POV bandwidth change is allowed. */
        return false;
    }

5562
    ifaceBand = virDomainNetGetActualBandwidth(iface);
5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578
    old_floor = new_floor = 0;

    if (ifaceBand && ifaceBand->in)
        old_floor = ifaceBand->in->floor;
    if (newBandwidth && newBandwidth->in)
        new_floor = newBandwidth->in->floor;

    return new_floor != old_floor;
}


bool
networkBandwidthChangeAllowed(virDomainNetDefPtr iface,
                              virNetDevBandwidthPtr newBandwidth)
{
    virNetworkDriverStatePtr driver = networkGetDriver();
5579
    virNetworkObjPtr obj = NULL;
5580 5581 5582 5583 5584 5585
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
    bool ret = false;

    if (!networkBandwidthGenericChecks(iface, newBandwidth))
        return true;

5586 5587
    obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
    if (!obj) {
5588 5589 5590 5591 5592 5593
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
        return false;
    }

5594
    if (networkCheckBandwidth(obj, newBandwidth, ifaceBand, iface->mac, NULL) < 0)
5595 5596 5597 5598 5599
        goto cleanup;

    ret = true;

 cleanup:
5600
    virNetworkObjEndAPI(&obj);
5601 5602
    return ret;
}
5603 5604 5605 5606 5607 5608 5609


int
networkBandwidthUpdate(virDomainNetDefPtr iface,
                       virNetDevBandwidthPtr newBandwidth)
{
    virNetworkDriverStatePtr driver = networkGetDriver();
5610
    virNetworkObjPtr obj = NULL;
5611 5612 5613 5614 5615 5616 5617 5618
    virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
    unsigned long long new_rate = 0;
    int plug_ret;
    int ret = -1;

    if (!networkBandwidthGenericChecks(iface, newBandwidth))
        return 0;

5619 5620
    obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
    if (!obj) {
5621 5622 5623 5624 5625 5626
        virReportError(VIR_ERR_NO_NETWORK,
                       _("no network with matching name '%s'"),
                       iface->data.network.name);
        return ret;
    }

5627
    if ((plug_ret = networkCheckBandwidth(obj, newBandwidth, ifaceBand,
5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640
                                          iface->mac, &new_rate)) < 0) {
        /* helper reported error */
        goto cleanup;
    }

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

    /* Okay, there are three possible scenarios: */

5641
    if (ifaceBand && ifaceBand->in && ifaceBand->in->floor &&
5642
        newBandwidth->in && newBandwidth->in->floor) {
5643 5644
        /* Either we just need to update @floor .. */

5645
        if (virNetDevBandwidthUpdateRate(obj->def->bridge,
5646
                                         iface->data.network.actual->class_id,
5647
                                         obj->def->bandwidth,
5648 5649 5650
                                         newBandwidth->in->floor) < 0)
            goto cleanup;

5651 5652 5653
        obj->floor_sum -= ifaceBand->in->floor;
        obj->floor_sum += newBandwidth->in->floor;
        new_rate -= obj->floor_sum;
5654

5655 5656 5657
        if (virNetDevBandwidthUpdateRate(obj->def->bridge, 2,
                                         obj->def->bandwidth, new_rate) < 0 ||
            virNetworkObjSaveStatus(driver->stateDir, obj) < 0) {
5658
            /* Ouch, rollback */
5659 5660
            obj->floor_sum -= newBandwidth->in->floor;
            obj->floor_sum += ifaceBand->in->floor;
5661

5662
            ignore_value(virNetDevBandwidthUpdateRate(obj->def->bridge,
5663
                                                      iface->data.network.actual->class_id,
5664
                                                      obj->def->bandwidth,
5665 5666 5667 5668 5669 5670
                                                      ifaceBand->in->floor));
            goto cleanup;
        }
    } else if (newBandwidth->in && newBandwidth->in->floor) {
        /* .. or we need to plug in new .. */

5671
        if (networkPlugBandwidthImpl(obj, iface, newBandwidth, new_rate) < 0)
5672 5673 5674 5675
            goto cleanup;
    } else {
        /* .. or unplug old. */

5676
        if (networkUnplugBandwidth(obj, iface) < 0)
5677 5678 5679 5680 5681
            goto cleanup;
    }

    ret = 0;
 cleanup:
5682
    virNetworkObjEndAPI(&obj);
5683 5684
    return ret;
}