]> git.proxmox.com Git - mirror_frr.git/blob - zebra/rt_netlink.c
projects / mirror_frr.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge pull request #2764 from opensourcerouting/isis-srcdest
[mirror_frr.git] / zebra / rt_netlink.c
1 /* Kernel routing table updates using netlink over GNU/Linux system.
2 * Copyright (C) 1997, 98, 99 Kunihiro Ishiguro
3 *
4 * This file is part of GNU Zebra.
5 *
6 * GNU Zebra is free software; you can redistribute it and/or modify it
7 * under the terms of the GNU General Public License as published by the
8 * Free Software Foundation; either version 2, or (at your option) any
9 * later version.
10 *
11 * GNU Zebra is distributed in the hope that it will be useful, but
12 * WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License along
17 * with this program; see the file COPYING; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include <zebra.h>
22
23 #ifdef HAVE_NETLINK
24
25 #include <net/if_arp.h>
26 #include <linux/lwtunnel.h>
27 #include <linux/mpls_iptunnel.h>
28 #include <linux/neighbour.h>
29 #include <linux/rtnetlink.h>
30
31 /* Hack for GNU libc version 2. */
32 #ifndef MSG_TRUNC
33 #define MSG_TRUNC 0x20
34 #endif /* MSG_TRUNC */
35
36 #include "linklist.h"
37 #include "if.h"
38 #include "log.h"
39 #include "prefix.h"
40 #include "connected.h"
41 #include "table.h"
42 #include "memory.h"
43 #include "zebra_memory.h"
44 #include "rib.h"
45 #include "thread.h"
46 #include "privs.h"
47 #include "nexthop.h"
48 #include "vrf.h"
49 #include "vty.h"
50 #include "mpls.h"
51 #include "vxlan.h"
52
53 #include "zebra/zapi_msg.h"
54 #include "zebra/zebra_ns.h"
55 #include "zebra/zebra_vrf.h"
56 #include "zebra/rt.h"
57 #include "zebra/redistribute.h"
58 #include "zebra/interface.h"
59 #include "zebra/debug.h"
60 #include "zebra/rtadv.h"
61 #include "zebra/zebra_ptm.h"
62 #include "zebra/zebra_mpls.h"
63 #include "zebra/kernel_netlink.h"
64 #include "zebra/rt_netlink.h"
65 #include "zebra/zebra_mroute.h"
66 #include "zebra/zebra_vxlan.h"
67
68 #ifndef AF_MPLS
69 #define AF_MPLS 28
70 #endif
71
72 static vlanid_t filter_vlan = 0;
73
74 struct gw_family_t {
75 uint16_t filler;
76 uint16_t family;
77 union g_addr gate;
78 };
79
80 char ipv4_ll_buf[16] = "169.254.0.1";
81 struct in_addr ipv4_ll;
82
83 /*
84 * The ipv4_ll data structure is used for all 5549
85 * additions to the kernel. Let's figure out the
86 * correct value one time instead for every
87 * install/remove of a 5549 type route
88 */
89 void rt_netlink_init(void)
90 {
91 inet_pton(AF_INET, ipv4_ll_buf, &ipv4_ll);
92 }
93
94 static inline int is_selfroute(int proto)
95 {
96 if ((proto == RTPROT_BGP) || (proto == RTPROT_OSPF)
97 || (proto == RTPROT_ZSTATIC) || (proto == RTPROT_ZEBRA)
98 || (proto == RTPROT_ISIS) || (proto == RTPROT_RIPNG)
99 || (proto == RTPROT_NHRP) || (proto == RTPROT_EIGRP)
100 || (proto == RTPROT_LDP) || (proto == RTPROT_BABEL)
101 || (proto == RTPROT_RIP) || (proto == RTPROT_SHARP)
102 || (proto == RTPROT_PBR)) {
103 return 1;
104 }
105
106 return 0;
107 }
108
109 static inline int zebra2proto(int proto)
110 {
111 switch (proto) {
112 case ZEBRA_ROUTE_BABEL:
113 proto = RTPROT_BABEL;
114 break;
115 case ZEBRA_ROUTE_BGP:
116 proto = RTPROT_BGP;
117 break;
118 case ZEBRA_ROUTE_OSPF:
119 case ZEBRA_ROUTE_OSPF6:
120 proto = RTPROT_OSPF;
121 break;
122 case ZEBRA_ROUTE_STATIC:
123 proto = RTPROT_ZSTATIC;
124 break;
125 case ZEBRA_ROUTE_ISIS:
126 proto = RTPROT_ISIS;
127 break;
128 case ZEBRA_ROUTE_RIP:
129 proto = RTPROT_RIP;
130 break;
131 case ZEBRA_ROUTE_RIPNG:
132 proto = RTPROT_RIPNG;
133 break;
134 case ZEBRA_ROUTE_NHRP:
135 proto = RTPROT_NHRP;
136 break;
137 case ZEBRA_ROUTE_EIGRP:
138 proto = RTPROT_EIGRP;
139 break;
140 case ZEBRA_ROUTE_LDP:
141 proto = RTPROT_LDP;
142 break;
143 case ZEBRA_ROUTE_SHARP:
144 proto = RTPROT_SHARP;
145 break;
146 case ZEBRA_ROUTE_PBR:
147 proto = RTPROT_PBR;
148 break;
149 default:
150 /*
151 * When a user adds a new protocol this will show up
152 * to let them know to do something about it. This
153 * is intentionally a warn because we should see
154 * this as part of development of a new protocol
155 */
156 zlog_warn("%s: Please add this protocol(%d) to proper rt_netlink.c handling",
157 __PRETTY_FUNCTION__, proto);
158 proto = RTPROT_ZEBRA;
159 break;
160 }
161
162 return proto;
163 }
164
165 static inline int proto2zebra(int proto, int family)
166 {
167 switch (proto) {
168 case RTPROT_BABEL:
169 proto = ZEBRA_ROUTE_BABEL;
170 break;
171 case RTPROT_BGP:
172 proto = ZEBRA_ROUTE_BGP;
173 break;
174 case RTPROT_OSPF:
175 proto = (family == AFI_IP) ? ZEBRA_ROUTE_OSPF
176 : ZEBRA_ROUTE_OSPF6;
177 break;
178 case RTPROT_ISIS:
179 proto = ZEBRA_ROUTE_ISIS;
180 break;
181 case RTPROT_RIP:
182 proto = ZEBRA_ROUTE_RIP;
183 break;
184 case RTPROT_RIPNG:
185 proto = ZEBRA_ROUTE_RIPNG;
186 break;
187 case RTPROT_NHRP:
188 proto = ZEBRA_ROUTE_NHRP;
189 break;
190 case RTPROT_EIGRP:
191 proto = ZEBRA_ROUTE_EIGRP;
192 break;
193 case RTPROT_LDP:
194 proto = ZEBRA_ROUTE_LDP;
195 break;
196 case RTPROT_STATIC:
197 case RTPROT_ZSTATIC:
198 proto = ZEBRA_ROUTE_STATIC;
199 break;
200 case RTPROT_SHARP:
201 proto = ZEBRA_ROUTE_SHARP;
202 break;
203 case RTPROT_PBR:
204 proto = ZEBRA_ROUTE_PBR;
205 break;
206 default:
207 /*
208 * When a user adds a new protocol this will show up
209 * to let them know to do something about it. This
210 * is intentionally a warn because we should see
211 * this as part of development of a new protocol
212 */
213 zlog_warn("%s: Please add this protocol(%d) to proper rt_netlink.c handling",
214 __PRETTY_FUNCTION__,
215 proto);
216 proto = ZEBRA_ROUTE_KERNEL;
217 break;
218 }
219 return proto;
220 }
221
222 /*
223 Pending: create an efficient table_id (in a tree/hash) based lookup)
224 */
225 static vrf_id_t vrf_lookup_by_table(uint32_t table_id, ns_id_t ns_id)
226 {
227 struct vrf *vrf;
228 struct zebra_vrf *zvrf;
229
230 RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) {
231 zvrf = vrf->info;
232 if (zvrf == NULL)
233 continue;
234 /* case vrf with netns : match the netnsid */
235 if (vrf_is_backend_netns()) {
236 if (ns_id == zvrf_id(zvrf))
237 return zvrf_id(zvrf);
238 } else {
239 /* VRF is VRF_BACKEND_VRF_LITE */
240 if (zvrf->table_id != table_id)
241 continue;
242 return zvrf_id(zvrf);
243 }
244 }
245
246 return VRF_DEFAULT;
247 }
248
249 /* Looking up routing table by netlink interface. */
250 static int netlink_route_change_read_unicast(struct nlmsghdr *h, ns_id_t ns_id,
251 int startup)
252 {
253 int len;
254 struct rtmsg *rtm;
255 struct rtattr *tb[RTA_MAX + 1];
256 uint8_t flags = 0;
257 struct prefix p;
258 struct prefix_ipv6 src_p = {};
259 vrf_id_t vrf_id;
260
261 char anyaddr[16] = {0};
262
263 int proto = ZEBRA_ROUTE_KERNEL;
264 int index = 0;
265 int table;
266 int metric = 0;
267 uint32_t mtu = 0;
268 uint8_t distance = 0;
269 route_tag_t tag = 0;
270
271 void *dest = NULL;
272 void *gate = NULL;
273 void *prefsrc = NULL; /* IPv4 preferred source host address */
274 void *src = NULL; /* IPv6 srcdest source prefix */
275 enum blackhole_type bh_type = BLACKHOLE_UNSPEC;
276
277 rtm = NLMSG_DATA(h);
278
279 if (startup && h->nlmsg_type != RTM_NEWROUTE)
280 return 0;
281 switch (rtm->rtm_type) {
282 case RTN_UNICAST:
283 break;
284 case RTN_BLACKHOLE:
285 bh_type = BLACKHOLE_NULL;
286 break;
287 case RTN_UNREACHABLE:
288 bh_type = BLACKHOLE_REJECT;
289 break;
290 case RTN_PROHIBIT:
291 bh_type = BLACKHOLE_ADMINPROHIB;
292 break;
293 default:
294 return 0;
295 }
296
297 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct rtmsg));
298 if (len < 0) {
299 zlog_err("%s: Message received from netlink is of a broken size %d %zu",
300 __PRETTY_FUNCTION__, h->nlmsg_len,
301 (size_t)NLMSG_LENGTH(sizeof(struct rtmsg)));
302 return -1;
303 }
304
305 memset(tb, 0, sizeof tb);
306 netlink_parse_rtattr(tb, RTA_MAX, RTM_RTA(rtm), len);
307
308 if (rtm->rtm_flags & RTM_F_CLONED)
309 return 0;
310 if (rtm->rtm_protocol == RTPROT_REDIRECT)
311 return 0;
312 if (rtm->rtm_protocol == RTPROT_KERNEL)
313 return 0;
314
315 if (!startup && is_selfroute(rtm->rtm_protocol)
316 && h->nlmsg_type == RTM_NEWROUTE) {
317 if (IS_ZEBRA_DEBUG_KERNEL)
318 zlog_debug("Route type: %d Received that we think we have originated, ignoring",
319 rtm->rtm_protocol);
320 return 0;
321 }
322
323 /* We don't care about change notifications for the MPLS table. */
324 /* TODO: Revisit this. */
325 if (rtm->rtm_family == AF_MPLS)
326 return 0;
327
328 /* Table corresponding to route. */
329 if (tb[RTA_TABLE])
330 table = *(int *)RTA_DATA(tb[RTA_TABLE]);
331 else
332 table = rtm->rtm_table;
333
334 /* Map to VRF */
335 vrf_id = vrf_lookup_by_table(table, ns_id);
336 if (vrf_id == VRF_DEFAULT) {
337 if (!is_zebra_valid_kernel_table(table)
338 && !is_zebra_main_routing_table(table))
339 return 0;
340 }
341
342 /* Route which inserted by Zebra. */
343 if (is_selfroute(rtm->rtm_protocol)) {
344 flags |= ZEBRA_FLAG_SELFROUTE;
345 proto = proto2zebra(rtm->rtm_protocol, rtm->rtm_family);
346 }
347 if (tb[RTA_OIF])
348 index = *(int *)RTA_DATA(tb[RTA_OIF]);
349
350 if (tb[RTA_DST])
351 dest = RTA_DATA(tb[RTA_DST]);
352 else
353 dest = anyaddr;
354
355 if (tb[RTA_SRC])
356 src = RTA_DATA(tb[RTA_SRC]);
357 else
358 src = anyaddr;
359
360 if (tb[RTA_PREFSRC])
361 prefsrc = RTA_DATA(tb[RTA_PREFSRC]);
362
363 if (tb[RTA_GATEWAY])
364 gate = RTA_DATA(tb[RTA_GATEWAY]);
365
366 if (tb[RTA_PRIORITY])
367 metric = *(int *)RTA_DATA(tb[RTA_PRIORITY]);
368
369 #if defined(SUPPORT_REALMS)
370 if (tb[RTA_FLOW])
371 tag = *(uint32_t *)RTA_DATA(tb[RTA_FLOW]);
372 #endif
373
374 if (tb[RTA_METRICS]) {
375 struct rtattr *mxrta[RTAX_MAX + 1];
376
377 memset(mxrta, 0, sizeof mxrta);
378 netlink_parse_rtattr(mxrta, RTAX_MAX, RTA_DATA(tb[RTA_METRICS]),
379 RTA_PAYLOAD(tb[RTA_METRICS]));
380
381 if (mxrta[RTAX_MTU])
382 mtu = *(uint32_t *)RTA_DATA(mxrta[RTAX_MTU]);
383 }
384
385 if (rtm->rtm_family == AF_INET) {
386 p.family = AF_INET;
387 if (rtm->rtm_dst_len > IPV4_MAX_BITLEN) {
388 zlog_err(
389 "Invalid destination prefix length: %u received from kernel route change",
390 rtm->rtm_dst_len);
391 return -1;
392 }
393 memcpy(&p.u.prefix4, dest, 4);
394 p.prefixlen = rtm->rtm_dst_len;
395
396 if (rtm->rtm_src_len != 0) {
397 char buf[PREFIX_STRLEN];
398 zlog_warn("unsupported IPv4 sourcedest route (dest %s vrf %u)",
399 prefix2str(&p, buf, sizeof(buf)), vrf_id);
400 return 0;
401 }
402
403 /* Force debug below to not display anything for source */
404 src_p.prefixlen = 0;
405 } else if (rtm->rtm_family == AF_INET6) {
406 p.family = AF_INET6;
407 if (rtm->rtm_dst_len > IPV6_MAX_BITLEN) {
408 zlog_err(
409 "Invalid destination prefix length: %u received from kernel route change",
410 rtm->rtm_dst_len);
411 return -1;
412 }
413 memcpy(&p.u.prefix6, dest, 16);
414 p.prefixlen = rtm->rtm_dst_len;
415
416 src_p.family = AF_INET6;
417 if (rtm->rtm_src_len > IPV6_MAX_BITLEN) {
418 zlog_err(
419 "Invalid source prefix length: %u received from kernel route change",
420 rtm->rtm_src_len);
421 return -1;
422 }
423 memcpy(&src_p.prefix, src, 16);
424 src_p.prefixlen = rtm->rtm_src_len;
425 }
426
427 /*
428 * For ZEBRA_ROUTE_KERNEL types:
429 *
430 * The metric/priority of the route received from the kernel
431 * is a 32 bit number. We are going to interpret the high
432 * order byte as the Admin Distance and the low order 3 bytes
433 * as the metric.
434 *
435 * This will allow us to do two things:
436 * 1) Allow the creation of kernel routes that can be
437 * overridden by zebra.
438 * 2) Allow the old behavior for 'most' kernel route types
439 * if a user enters 'ip route ...' v4 routes get a metric
440 * of 0 and v6 routes get a metric of 1024. Both of these
441 * values will end up with a admin distance of 0, which
442 * will cause them to win for the purposes of zebra.
443 */
444 if (proto == ZEBRA_ROUTE_KERNEL) {
445 distance = (metric >> 24) & 0xFF;
446 metric = (metric & 0x00FFFFFF);
447 }
448
449 if (IS_ZEBRA_DEBUG_KERNEL) {
450 char buf[PREFIX_STRLEN];
451 char buf2[PREFIX_STRLEN];
452 zlog_debug("%s %s%s%s vrf %u(%u) metric: %d Admin Distance: %d",
453 nl_msg_type_to_str(h->nlmsg_type),
454 prefix2str(&p, buf, sizeof(buf)),
455 src_p.prefixlen ? " from " : "",
456 src_p.prefixlen
457 ? prefix2str(&src_p, buf2, sizeof(buf2))
458 : "",
459 vrf_id, table, metric, distance);
460 }
461
462 afi_t afi = AFI_IP;
463 if (rtm->rtm_family == AF_INET6)
464 afi = AFI_IP6;
465
466 if (h->nlmsg_type == RTM_NEWROUTE) {
467 struct interface *ifp;
468 vrf_id_t nh_vrf_id = vrf_id;
469
470 if (!tb[RTA_MULTIPATH]) {
471 struct nexthop nh;
472 size_t sz = (afi == AFI_IP) ? 4 : 16;
473
474 memset(&nh, 0, sizeof(nh));
475
476 if (bh_type == BLACKHOLE_UNSPEC) {
477 if (index && !gate)
478 nh.type = NEXTHOP_TYPE_IFINDEX;
479 else if (index && gate)
480 nh.type =
481 (afi == AFI_IP)
482 ? NEXTHOP_TYPE_IPV4_IFINDEX
483 : NEXTHOP_TYPE_IPV6_IFINDEX;
484 else if (!index && gate)
485 nh.type = (afi == AFI_IP)
486 ? NEXTHOP_TYPE_IPV4
487 : NEXTHOP_TYPE_IPV6;
488 else {
489 nh.type = NEXTHOP_TYPE_BLACKHOLE;
490 nh.bh_type = bh_type;
491 }
492 } else {
493 nh.type = NEXTHOP_TYPE_BLACKHOLE;
494 nh.bh_type = bh_type;
495 }
496 nh.ifindex = index;
497 if (prefsrc)
498 memcpy(&nh.src, prefsrc, sz);
499 if (gate)
500 memcpy(&nh.gate, gate, sz);
501
502 if (index) {
503 ifp = if_lookup_by_index_per_ns(
504 zebra_ns_lookup(ns_id),
505 index);
506 if (ifp)
507 nh_vrf_id = ifp->vrf_id;
508 }
509 nh.vrf_id = nh_vrf_id;
510
511 rib_add(afi, SAFI_UNICAST, vrf_id, proto, 0, flags, &p,
512 &src_p, &nh, table, metric, mtu, distance, tag);
513 } else {
514 /* This is a multipath route */
515
516 struct route_entry *re;
517 struct rtnexthop *rtnh =
518 (struct rtnexthop *)RTA_DATA(tb[RTA_MULTIPATH]);
519
520 len = RTA_PAYLOAD(tb[RTA_MULTIPATH]);
521
522 re = XCALLOC(MTYPE_RE, sizeof(struct route_entry));
523 re->type = proto;
524 re->distance = distance;
525 re->flags = flags;
526 re->metric = metric;
527 re->mtu = mtu;
528 re->vrf_id = vrf_id;
529 re->table = table;
530 re->nexthop_num = 0;
531 re->uptime = time(NULL);
532 re->tag = tag;
533
534 for (;;) {
535 vrf_id_t nh_vrf_id;
536 if (len < (int)sizeof(*rtnh)
537 || rtnh->rtnh_len > len)
538 break;
539
540 index = rtnh->rtnh_ifindex;
541 if (index) {
542 /*
543 * Yes we are looking this up
544 * for every nexthop and just
545 * using the last one looked
546 * up right now
547 */
548 ifp = if_lookup_by_index_per_ns(
549 zebra_ns_lookup(ns_id),
550 index);
551 if (ifp)
552 nh_vrf_id = ifp->vrf_id;
553 else {
554 zlog_warn(
555 "%s: Unknown interface %u specified, defaulting to VRF_DEFAULT",
556 __PRETTY_FUNCTION__,
557 index);
558 nh_vrf_id = VRF_DEFAULT;
559 }
560 } else
561 nh_vrf_id = vrf_id;
562
563 gate = 0;
564 if (rtnh->rtnh_len > sizeof(*rtnh)) {
565 memset(tb, 0, sizeof(tb));
566 netlink_parse_rtattr(
567 tb, RTA_MAX, RTNH_DATA(rtnh),
568 rtnh->rtnh_len - sizeof(*rtnh));
569 if (tb[RTA_GATEWAY])
570 gate = RTA_DATA(
571 tb[RTA_GATEWAY]);
572 }
573
574 if (gate) {
575 if (rtm->rtm_family == AF_INET) {
576 if (index)
577 route_entry_nexthop_ipv4_ifindex_add(
578 re, gate,
579 prefsrc, index,
580 nh_vrf_id);
581 else
582 route_entry_nexthop_ipv4_add(
583 re, gate,
584 prefsrc,
585 nh_vrf_id);
586 } else if (rtm->rtm_family
587 == AF_INET6) {
588 if (index)
589 route_entry_nexthop_ipv6_ifindex_add(
590 re, gate, index,
591 nh_vrf_id);
592 else
593 route_entry_nexthop_ipv6_add(
594 re, gate,
595 nh_vrf_id);
596 }
597 } else
598 route_entry_nexthop_ifindex_add(
599 re, index, nh_vrf_id);
600
601 if (rtnh->rtnh_len == 0)
602 break;
603
604 len -= NLMSG_ALIGN(rtnh->rtnh_len);
605 rtnh = RTNH_NEXT(rtnh);
606 }
607
608 zserv_nexthop_num_warn(__func__,
609 (const struct prefix *)&p,
610 re->nexthop_num);
611 if (re->nexthop_num == 0)
612 XFREE(MTYPE_RE, re);
613 else
614 rib_add_multipath(afi, SAFI_UNICAST, &p,
615 &src_p, re);
616 }
617 } else {
618 if (!tb[RTA_MULTIPATH]) {
619 struct nexthop nh;
620 size_t sz = (afi == AFI_IP) ? 4 : 16;
621
622 memset(&nh, 0, sizeof(nh));
623 if (bh_type == BLACKHOLE_UNSPEC) {
624 if (index && !gate)
625 nh.type = NEXTHOP_TYPE_IFINDEX;
626 else if (index && gate)
627 nh.type =
628 (afi == AFI_IP)
629 ? NEXTHOP_TYPE_IPV4_IFINDEX
630 : NEXTHOP_TYPE_IPV6_IFINDEX;
631 else if (!index && gate)
632 nh.type = (afi == AFI_IP)
633 ? NEXTHOP_TYPE_IPV4
634 : NEXTHOP_TYPE_IPV6;
635 else {
636 nh.type = NEXTHOP_TYPE_BLACKHOLE;
637 nh.bh_type = BLACKHOLE_UNSPEC;
638 }
639 } else {
640 nh.type = NEXTHOP_TYPE_BLACKHOLE;
641 nh.bh_type = bh_type;
642 }
643 nh.ifindex = index;
644 if (gate)
645 memcpy(&nh.gate, gate, sz);
646 rib_delete(afi, SAFI_UNICAST, vrf_id, proto, 0, flags,
647 &p, &src_p, &nh, table, metric, true);
648 } else {
649 /* XXX: need to compare the entire list of nexthops
650 * here for NLM_F_APPEND stupidity */
651 rib_delete(afi, SAFI_UNICAST, vrf_id, proto, 0, flags,
652 &p, &src_p, NULL, table, metric, true);
653 }
654 }
655
656 return 0;
657 }
658
659 static struct mcast_route_data *mroute = NULL;
660
661 static int netlink_route_change_read_multicast(struct nlmsghdr *h,
662 ns_id_t ns_id, int startup)
663 {
664 int len;
665 struct rtmsg *rtm;
666 struct rtattr *tb[RTA_MAX + 1];
667 struct mcast_route_data *m;
668 struct mcast_route_data mr;
669 int iif = 0;
670 int count;
671 int oif[256];
672 int oif_count = 0;
673 char sbuf[40];
674 char gbuf[40];
675 char oif_list[256] = "\0";
676 vrf_id_t vrf;
677 int table;
678
679 if (mroute)
680 m = mroute;
681 else {
682 memset(&mr, 0, sizeof(mr));
683 m = &mr;
684 }
685
686 rtm = NLMSG_DATA(h);
687
688 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct rtmsg));
689
690 memset(tb, 0, sizeof tb);
691 netlink_parse_rtattr(tb, RTA_MAX, RTM_RTA(rtm), len);
692
693 if (tb[RTA_TABLE])
694 table = *(int *)RTA_DATA(tb[RTA_TABLE]);
695 else
696 table = rtm->rtm_table;
697
698 vrf = vrf_lookup_by_table(table, ns_id);
699
700 if (tb[RTA_IIF])
701 iif = *(int *)RTA_DATA(tb[RTA_IIF]);
702
703 if (tb[RTA_SRC])
704 m->sg.src = *(struct in_addr *)RTA_DATA(tb[RTA_SRC]);
705
706 if (tb[RTA_DST])
707 m->sg.grp = *(struct in_addr *)RTA_DATA(tb[RTA_DST]);
708
709 if ((RTA_EXPIRES <= RTA_MAX) && tb[RTA_EXPIRES])
710 m->lastused = *(unsigned long long *)RTA_DATA(tb[RTA_EXPIRES]);
711
712 if (tb[RTA_MULTIPATH]) {
713 struct rtnexthop *rtnh =
714 (struct rtnexthop *)RTA_DATA(tb[RTA_MULTIPATH]);
715
716 len = RTA_PAYLOAD(tb[RTA_MULTIPATH]);
717 for (;;) {
718 if (len < (int)sizeof(*rtnh) || rtnh->rtnh_len > len)
719 break;
720
721 oif[oif_count] = rtnh->rtnh_ifindex;
722 oif_count++;
723
724 if (rtnh->rtnh_len == 0)
725 break;
726
727 len -= NLMSG_ALIGN(rtnh->rtnh_len);
728 rtnh = RTNH_NEXT(rtnh);
729 }
730 }
731
732 if (IS_ZEBRA_DEBUG_KERNEL) {
733 struct interface *ifp;
734 strlcpy(sbuf, inet_ntoa(m->sg.src), sizeof(sbuf));
735 strlcpy(gbuf, inet_ntoa(m->sg.grp), sizeof(gbuf));
736 for (count = 0; count < oif_count; count++) {
737 ifp = if_lookup_by_index(oif[count], vrf);
738 char temp[256];
739
740 sprintf(temp, "%s ", ifp->name);
741 strcat(oif_list, temp);
742 }
743 struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(vrf);
744 ifp = if_lookup_by_index(iif, vrf);
745 zlog_debug(
746 "MCAST VRF: %s(%d) %s (%s,%s) IIF: %s OIF: %s jiffies: %lld",
747 zvrf->vrf->name, vrf, nl_msg_type_to_str(h->nlmsg_type),
748 sbuf, gbuf, ifp->name, oif_list, m->lastused);
749 }
750 return 0;
751 }
752
753 int netlink_route_change(struct nlmsghdr *h, ns_id_t ns_id, int startup)
754 {
755 int len;
756 struct rtmsg *rtm;
757
758 rtm = NLMSG_DATA(h);
759
760 if (!(h->nlmsg_type == RTM_NEWROUTE || h->nlmsg_type == RTM_DELROUTE)) {
761 /* If this is not route add/delete message print warning. */
762 zlog_warn("Kernel message: %d NS %u\n", h->nlmsg_type, ns_id);
763 return 0;
764 }
765
766 if (!(rtm->rtm_family == AF_INET || rtm->rtm_family == AF_INET6)) {
767 zlog_warn(
768 "Invalid address family: %u received from kernel route change: %u",
769 rtm->rtm_family, h->nlmsg_type);
770 return 0;
771 }
772
773 /* Connected route. */
774 if (IS_ZEBRA_DEBUG_KERNEL)
775 zlog_debug("%s %s %s proto %s NS %u",
776 nl_msg_type_to_str(h->nlmsg_type),
777 nl_family_to_str(rtm->rtm_family),
778 nl_rttype_to_str(rtm->rtm_type),
779 nl_rtproto_to_str(rtm->rtm_protocol), ns_id);
780
781
782 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct rtmsg));
783 if (len < 0) {
784 zlog_err("%s: Message received from netlink is of a broken size: %d %zu",
785 __PRETTY_FUNCTION__,
786 h->nlmsg_len,
787 (size_t)NLMSG_LENGTH(sizeof(struct rtmsg)));
788 return -1;
789 }
790
791 if (rtm->rtm_type == RTN_MULTICAST)
792 netlink_route_change_read_multicast(h, ns_id, startup);
793 else
794 netlink_route_change_read_unicast(h, ns_id, startup);
795 return 0;
796 }
797
798 /* Request for specific route information from the kernel */
799 static int netlink_request_route(struct zebra_ns *zns, int family, int type)
800 {
801 struct {
802 struct nlmsghdr n;
803 struct rtmsg rtm;
804 } req;
805
806 /* Form the request, specifying filter (rtattr) if needed. */
807 memset(&req, 0, sizeof(req));
808 req.n.nlmsg_type = type;
809 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
810 req.rtm.rtm_family = family;
811
812 return netlink_request(&zns->netlink_cmd, &req.n);
813 }
814
815 /* Routing table read function using netlink interface. Only called
816 bootstrap time. */
817 int netlink_route_read(struct zebra_ns *zns)
818 {
819 int ret;
820
821 /* Get IPv4 routing table. */
822 ret = netlink_request_route(zns, AF_INET, RTM_GETROUTE);
823 if (ret < 0)
824 return ret;
825 ret = netlink_parse_info(netlink_route_change_read_unicast,
826 &zns->netlink_cmd, zns, 0, 1);
827 if (ret < 0)
828 return ret;
829
830 /* Get IPv6 routing table. */
831 ret = netlink_request_route(zns, AF_INET6, RTM_GETROUTE);
832 if (ret < 0)
833 return ret;
834 ret = netlink_parse_info(netlink_route_change_read_unicast,
835 &zns->netlink_cmd, zns, 0, 1);
836 if (ret < 0)
837 return ret;
838
839 return 0;
840 }
841
842 static void _netlink_route_nl_add_gateway_info(uint8_t route_family,
843 uint8_t gw_family,
844 struct nlmsghdr *nlmsg,
845 size_t req_size, int bytelen,
846 struct nexthop *nexthop)
847 {
848 if (route_family == AF_MPLS) {
849 struct gw_family_t gw_fam;
850
851 gw_fam.family = gw_family;
852 if (gw_family == AF_INET)
853 memcpy(&gw_fam.gate.ipv4, &nexthop->gate.ipv4, bytelen);
854 else
855 memcpy(&gw_fam.gate.ipv6, &nexthop->gate.ipv6, bytelen);
856 addattr_l(nlmsg, req_size, RTA_VIA, &gw_fam.family,
857 bytelen + 2);
858 } else {
859 if (gw_family == AF_INET)
860 addattr_l(nlmsg, req_size, RTA_GATEWAY,
861 &nexthop->gate.ipv4, bytelen);
862 else
863 addattr_l(nlmsg, req_size, RTA_GATEWAY,
864 &nexthop->gate.ipv6, bytelen);
865 }
866 }
867
868 static void _netlink_route_rta_add_gateway_info(uint8_t route_family,
869 uint8_t gw_family,
870 struct rtattr *rta,
871 struct rtnexthop *rtnh,
872 size_t req_size, int bytelen,
873 struct nexthop *nexthop)
874 {
875 if (route_family == AF_MPLS) {
876 struct gw_family_t gw_fam;
877
878 gw_fam.family = gw_family;
879 if (gw_family == AF_INET)
880 memcpy(&gw_fam.gate.ipv4, &nexthop->gate.ipv4, bytelen);
881 else
882 memcpy(&gw_fam.gate.ipv6, &nexthop->gate.ipv6, bytelen);
883 rta_addattr_l(rta, req_size, RTA_VIA, &gw_fam.family,
884 bytelen + 2);
885 rtnh->rtnh_len += RTA_LENGTH(bytelen + 2);
886 } else {
887 if (gw_family == AF_INET)
888 rta_addattr_l(rta, req_size, RTA_GATEWAY,
889 &nexthop->gate.ipv4, bytelen);
890 else
891 rta_addattr_l(rta, req_size, RTA_GATEWAY,
892 &nexthop->gate.ipv6, bytelen);
893 rtnh->rtnh_len += sizeof(struct rtattr) + bytelen;
894 }
895 }
896
897 /* This function takes a nexthop as argument and adds
898 * the appropriate netlink attributes to an existing
899 * netlink message.
900 *
901 * @param routedesc: Human readable description of route type
902 * (direct/recursive, single-/multipath)
903 * @param bytelen: Length of addresses in bytes.
904 * @param nexthop: Nexthop information
905 * @param nlmsg: nlmsghdr structure to fill in.
906 * @param req_size: The size allocated for the message.
907 */
908 static void _netlink_route_build_singlepath(const char *routedesc, int bytelen,
909 struct nexthop *nexthop,
910 struct nlmsghdr *nlmsg,
911 struct rtmsg *rtmsg,
912 size_t req_size, int cmd)
913 {
914 struct mpls_label_stack *nh_label;
915 mpls_lse_t out_lse[MPLS_MAX_LABELS];
916 int num_labels = 0;
917 char label_buf[256];
918
919 /*
920 * label_buf is *only* currently used within debugging.
921 * As such when we assign it we are guarding it inside
922 * a debug test. If you want to change this make sure
923 * you fix this assumption
924 */
925 label_buf[0] = '\0';
926
927 assert(nexthop);
928 for (struct nexthop *nh = nexthop; nh; nh = nh->rparent) {
929 char label_buf1[20];
930
931 nh_label = nh->nh_label;
932 if (!nh_label || !nh_label->num_labels)
933 continue;
934
935 for (int i = 0; i < nh_label->num_labels; i++) {
936 if (nh_label->label[i] == MPLS_LABEL_IMPLICIT_NULL)
937 continue;
938
939 if (IS_ZEBRA_DEBUG_KERNEL) {
940 if (!num_labels)
941 sprintf(label_buf, "label %u",
942 nh_label->label[i]);
943 else {
944 sprintf(label_buf1, "/%u",
945 nh_label->label[i]);
946 strlcat(label_buf, label_buf1,
947 sizeof(label_buf));
948 }
949 }
950
951 out_lse[num_labels] =
952 mpls_lse_encode(nh_label->label[i], 0, 0, 0);
953 num_labels++;
954 }
955 }
956
957 if (num_labels) {
958 /* Set the BoS bit */
959 out_lse[num_labels - 1] |= htonl(1 << MPLS_LS_S_SHIFT);
960
961 if (rtmsg->rtm_family == AF_MPLS)
962 addattr_l(nlmsg, req_size, RTA_NEWDST, &out_lse,
963 num_labels * sizeof(mpls_lse_t));
964 else {
965 struct rtattr *nest;
966 uint16_t encap = LWTUNNEL_ENCAP_MPLS;
967
968 addattr_l(nlmsg, req_size, RTA_ENCAP_TYPE, &encap,
969 sizeof(uint16_t));
970 nest = addattr_nest(nlmsg, req_size, RTA_ENCAP);
971 addattr_l(nlmsg, req_size, MPLS_IPTUNNEL_DST, &out_lse,
972 num_labels * sizeof(mpls_lse_t));
973 addattr_nest_end(nlmsg, nest);
974 }
975 }
976
977 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK))
978 rtmsg->rtm_flags |= RTNH_F_ONLINK;
979
980 if (rtmsg->rtm_family == AF_INET
981 && (nexthop->type == NEXTHOP_TYPE_IPV6
982 || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX)) {
983 rtmsg->rtm_flags |= RTNH_F_ONLINK;
984 addattr_l(nlmsg, req_size, RTA_GATEWAY, &ipv4_ll, 4);
985 addattr32(nlmsg, req_size, RTA_OIF, nexthop->ifindex);
986
987 if (nexthop->rmap_src.ipv4.s_addr && (cmd == RTM_NEWROUTE))
988 addattr_l(nlmsg, req_size, RTA_PREFSRC,
989 &nexthop->rmap_src.ipv4, bytelen);
990 else if (nexthop->src.ipv4.s_addr && (cmd == RTM_NEWROUTE))
991 addattr_l(nlmsg, req_size, RTA_PREFSRC,
992 &nexthop->src.ipv4, bytelen);
993
994 if (IS_ZEBRA_DEBUG_KERNEL)
995 zlog_debug(
996 " 5549: _netlink_route_build_singlepath() (%s): "
997 "nexthop via %s %s if %u(%u)",
998 routedesc, ipv4_ll_buf, label_buf,
999 nexthop->ifindex, nexthop->vrf_id);
1000 return;
1001 }
1002
1003 if (nexthop->type == NEXTHOP_TYPE_IPV4
1004 || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX) {
1005 /* Send deletes to the kernel without specifying the next-hop */
1006 if (cmd != RTM_DELROUTE)
1007 _netlink_route_nl_add_gateway_info(
1008 rtmsg->rtm_family, AF_INET, nlmsg, req_size,
1009 bytelen, nexthop);
1010
1011 if (cmd == RTM_NEWROUTE) {
1012 if (nexthop->rmap_src.ipv4.s_addr)
1013 addattr_l(nlmsg, req_size, RTA_PREFSRC,
1014 &nexthop->rmap_src.ipv4, bytelen);
1015 else if (nexthop->src.ipv4.s_addr)
1016 addattr_l(nlmsg, req_size, RTA_PREFSRC,
1017 &nexthop->src.ipv4, bytelen);
1018 }
1019
1020 if (IS_ZEBRA_DEBUG_KERNEL)
1021 zlog_debug(
1022 "netlink_route_multipath() (%s): "
1023 "nexthop via %s %s if %u(%u)",
1024 routedesc, inet_ntoa(nexthop->gate.ipv4),
1025 label_buf, nexthop->ifindex, nexthop->vrf_id);
1026 }
1027
1028 if (nexthop->type == NEXTHOP_TYPE_IPV6
1029 || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX) {
1030 _netlink_route_nl_add_gateway_info(rtmsg->rtm_family, AF_INET6,
1031 nlmsg, req_size, bytelen,
1032 nexthop);
1033
1034 if (cmd == RTM_NEWROUTE) {
1035 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->rmap_src.ipv6))
1036 addattr_l(nlmsg, req_size, RTA_PREFSRC,
1037 &nexthop->rmap_src.ipv6, bytelen);
1038 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->src.ipv6))
1039 addattr_l(nlmsg, req_size, RTA_PREFSRC,
1040 &nexthop->src.ipv6, bytelen);
1041 }
1042
1043 if (IS_ZEBRA_DEBUG_KERNEL)
1044 zlog_debug(
1045 "netlink_route_multipath() (%s): "
1046 "nexthop via %s %s if %u(%u)",
1047 routedesc, inet6_ntoa(nexthop->gate.ipv6),
1048 label_buf, nexthop->ifindex, nexthop->vrf_id);
1049 }
1050
1051 /*
1052 * We have the ifindex so we should always send it
1053 * This is especially useful if we are doing route
1054 * leaking.
1055 */
1056 if (nexthop->type != NEXTHOP_TYPE_BLACKHOLE)
1057 addattr32(nlmsg, req_size, RTA_OIF, nexthop->ifindex);
1058
1059 if (nexthop->type == NEXTHOP_TYPE_IFINDEX
1060 || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX) {
1061 if (cmd == RTM_NEWROUTE) {
1062 if (nexthop->rmap_src.ipv4.s_addr)
1063 addattr_l(nlmsg, req_size, RTA_PREFSRC,
1064 &nexthop->rmap_src.ipv4, bytelen);
1065 else if (nexthop->src.ipv4.s_addr)
1066 addattr_l(nlmsg, req_size, RTA_PREFSRC,
1067 &nexthop->src.ipv4, bytelen);
1068 }
1069
1070 if (IS_ZEBRA_DEBUG_KERNEL)
1071 zlog_debug(
1072 "netlink_route_multipath() (%s): "
1073 "nexthop via if %u(%u)",
1074 routedesc, nexthop->ifindex, nexthop->vrf_id);
1075 }
1076
1077 if (nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX) {
1078 if (cmd == RTM_NEWROUTE) {
1079 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->rmap_src.ipv6))
1080 addattr_l(nlmsg, req_size, RTA_PREFSRC,
1081 &nexthop->rmap_src.ipv6, bytelen);
1082 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->src.ipv6))
1083 addattr_l(nlmsg, req_size, RTA_PREFSRC,
1084 &nexthop->src.ipv6, bytelen);
1085 }
1086
1087 if (IS_ZEBRA_DEBUG_KERNEL)
1088 zlog_debug(
1089 "netlink_route_multipath() (%s): "
1090 "nexthop via if %u(%u)",
1091 routedesc, nexthop->ifindex, nexthop->vrf_id);
1092 }
1093 }
1094
1095 /* This function takes a nexthop as argument and
1096 * appends to the given rtattr/rtnexthop pair the
1097 * representation of the nexthop. If the nexthop
1098 * defines a preferred source, the src parameter
1099 * will be modified to point to that src, otherwise
1100 * it will be kept unmodified.
1101 *
1102 * @param routedesc: Human readable description of route type
1103 * (direct/recursive, single-/multipath)
1104 * @param bytelen: Length of addresses in bytes.
1105 * @param nexthop: Nexthop information
1106 * @param rta: rtnetlink attribute structure
1107 * @param rtnh: pointer to an rtnetlink nexthop structure
1108 * @param src: pointer pointing to a location where
1109 * the prefsrc should be stored.
1110 */
1111 static void _netlink_route_build_multipath(const char *routedesc, int bytelen,
1112 struct nexthop *nexthop,
1113 struct rtattr *rta,
1114 struct rtnexthop *rtnh,
1115 struct rtmsg *rtmsg,
1116 union g_addr **src)
1117 {
1118 struct mpls_label_stack *nh_label;
1119 mpls_lse_t out_lse[MPLS_MAX_LABELS];
1120 int num_labels = 0;
1121 char label_buf[256];
1122
1123 rtnh->rtnh_len = sizeof(*rtnh);
1124 rtnh->rtnh_flags = 0;
1125 rtnh->rtnh_hops = 0;
1126 rta->rta_len += rtnh->rtnh_len;
1127
1128 /*
1129 * label_buf is *only* currently used within debugging.
1130 * As such when we assign it we are guarding it inside
1131 * a debug test. If you want to change this make sure
1132 * you fix this assumption
1133 */
1134 label_buf[0] = '\0';
1135
1136 assert(nexthop);
1137 for (struct nexthop *nh = nexthop; nh; nh = nh->rparent) {
1138 char label_buf1[20];
1139
1140 nh_label = nh->nh_label;
1141 if (!nh_label || !nh_label->num_labels)
1142 continue;
1143
1144 for (int i = 0; i < nh_label->num_labels; i++) {
1145 if (nh_label->label[i] == MPLS_LABEL_IMPLICIT_NULL)
1146 continue;
1147
1148 if (IS_ZEBRA_DEBUG_KERNEL) {
1149 if (!num_labels)
1150 sprintf(label_buf, "label %u",
1151 nh_label->label[i]);
1152 else {
1153 sprintf(label_buf1, "/%u",
1154 nh_label->label[i]);
1155 strlcat(label_buf, label_buf1,
1156 sizeof(label_buf));
1157 }
1158 }
1159
1160 out_lse[num_labels] =
1161 mpls_lse_encode(nh_label->label[i], 0, 0, 0);
1162 num_labels++;
1163 }
1164 }
1165
1166 if (num_labels) {
1167 /* Set the BoS bit */
1168 out_lse[num_labels - 1] |= htonl(1 << MPLS_LS_S_SHIFT);
1169
1170 if (rtmsg->rtm_family == AF_MPLS) {
1171 rta_addattr_l(rta, NL_PKT_BUF_SIZE, RTA_NEWDST,
1172 &out_lse,
1173 num_labels * sizeof(mpls_lse_t));
1174 rtnh->rtnh_len +=
1175 RTA_LENGTH(num_labels * sizeof(mpls_lse_t));
1176 } else {
1177 struct rtattr *nest;
1178 uint16_t encap = LWTUNNEL_ENCAP_MPLS;
1179 int len = rta->rta_len;
1180
1181 rta_addattr_l(rta, NL_PKT_BUF_SIZE, RTA_ENCAP_TYPE,
1182 &encap, sizeof(uint16_t));
1183 nest = rta_nest(rta, NL_PKT_BUF_SIZE, RTA_ENCAP);
1184 rta_addattr_l(rta, NL_PKT_BUF_SIZE, MPLS_IPTUNNEL_DST,
1185 &out_lse,
1186 num_labels * sizeof(mpls_lse_t));
1187 rta_nest_end(rta, nest);
1188 rtnh->rtnh_len += rta->rta_len - len;
1189 }
1190 }
1191
1192 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK))
1193 rtnh->rtnh_flags |= RTNH_F_ONLINK;
1194
1195 if (rtmsg->rtm_family == AF_INET
1196 && (nexthop->type == NEXTHOP_TYPE_IPV6
1197 || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX)) {
1198 bytelen = 4;
1199 rtnh->rtnh_flags |= RTNH_F_ONLINK;
1200 rta_addattr_l(rta, NL_PKT_BUF_SIZE, RTA_GATEWAY, &ipv4_ll,
1201 bytelen);
1202 rtnh->rtnh_len += sizeof(struct rtattr) + bytelen;
1203 rtnh->rtnh_ifindex = nexthop->ifindex;
1204
1205 if (nexthop->rmap_src.ipv4.s_addr)
1206 *src = &nexthop->rmap_src;
1207 else if (nexthop->src.ipv4.s_addr)
1208 *src = &nexthop->src;
1209
1210 if (IS_ZEBRA_DEBUG_KERNEL)
1211 zlog_debug(
1212 " 5549: netlink_route_build_multipath() (%s): "
1213 "nexthop via %s %s if %u",
1214 routedesc, ipv4_ll_buf, label_buf,
1215 nexthop->ifindex);
1216 return;
1217 }
1218
1219 if (nexthop->type == NEXTHOP_TYPE_IPV4
1220 || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX) {
1221 _netlink_route_rta_add_gateway_info(rtmsg->rtm_family, AF_INET,
1222 rta, rtnh, NL_PKT_BUF_SIZE,
1223 bytelen, nexthop);
1224 if (nexthop->rmap_src.ipv4.s_addr)
1225 *src = &nexthop->rmap_src;
1226 else if (nexthop->src.ipv4.s_addr)
1227 *src = &nexthop->src;
1228
1229 if (IS_ZEBRA_DEBUG_KERNEL)
1230 zlog_debug(
1231 "netlink_route_multipath() (%s): "
1232 "nexthop via %s %s if %u",
1233 routedesc, inet_ntoa(nexthop->gate.ipv4),
1234 label_buf, nexthop->ifindex);
1235 }
1236 if (nexthop->type == NEXTHOP_TYPE_IPV6
1237 || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX) {
1238 _netlink_route_rta_add_gateway_info(rtmsg->rtm_family, AF_INET6,
1239 rta, rtnh, NL_PKT_BUF_SIZE,
1240 bytelen, nexthop);
1241
1242 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->rmap_src.ipv6))
1243 *src = &nexthop->rmap_src;
1244 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->src.ipv6))
1245 *src = &nexthop->src;
1246
1247 if (IS_ZEBRA_DEBUG_KERNEL)
1248 zlog_debug(
1249 "netlink_route_multipath() (%s): "
1250 "nexthop via %s %s if %u",
1251 routedesc, inet6_ntoa(nexthop->gate.ipv6),
1252 label_buf, nexthop->ifindex);
1253 }
1254
1255 /*
1256 * We have figured out the ifindex so we should always send it
1257 * This is especially useful if we are doing route
1258 * leaking.
1259 */
1260 if (nexthop->type != NEXTHOP_TYPE_BLACKHOLE)
1261 rtnh->rtnh_ifindex = nexthop->ifindex;
1262
1263 /* ifindex */
1264 if (nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX
1265 || nexthop->type == NEXTHOP_TYPE_IFINDEX) {
1266 if (nexthop->rmap_src.ipv4.s_addr)
1267 *src = &nexthop->rmap_src;
1268 else if (nexthop->src.ipv4.s_addr)
1269 *src = &nexthop->src;
1270
1271 if (IS_ZEBRA_DEBUG_KERNEL)
1272 zlog_debug(
1273 "netlink_route_multipath() (%s): "
1274 "nexthop via if %u",
1275 routedesc, nexthop->ifindex);
1276 }
1277 }
1278
1279 static inline void _netlink_mpls_build_singlepath(const char *routedesc,
1280 zebra_nhlfe_t *nhlfe,
1281 struct nlmsghdr *nlmsg,
1282 struct rtmsg *rtmsg,
1283 size_t req_size, int cmd)
1284 {
1285 int bytelen;
1286 uint8_t family;
1287
1288 family = NHLFE_FAMILY(nhlfe);
1289 bytelen = (family == AF_INET ? 4 : 16);
1290 _netlink_route_build_singlepath(routedesc, bytelen, nhlfe->nexthop,
1291 nlmsg, rtmsg, req_size, cmd);
1292 }
1293
1294
1295 static inline void
1296 _netlink_mpls_build_multipath(const char *routedesc, zebra_nhlfe_t *nhlfe,
1297 struct rtattr *rta, struct rtnexthop *rtnh,
1298 struct rtmsg *rtmsg, union g_addr **src)
1299 {
1300 int bytelen;
1301 uint8_t family;
1302
1303 family = NHLFE_FAMILY(nhlfe);
1304 bytelen = (family == AF_INET ? 4 : 16);
1305 _netlink_route_build_multipath(routedesc, bytelen, nhlfe->nexthop, rta,
1306 rtnh, rtmsg, src);
1307 }
1308
1309
1310 /* Log debug information for netlink_route_multipath
1311 * if debug logging is enabled.
1312 *
1313 * @param cmd: Netlink command which is to be processed
1314 * @param p: Prefix for which the change is due
1315 * @param family: Address family which the change concerns
1316 * @param zvrf: The vrf we are in
1317 * @param tableid: The table we are working on
1318 */
1319 static void _netlink_route_debug(int cmd, const struct prefix *p,
1320 int family, vrf_id_t vrfid,
1321 uint32_t tableid)
1322 {
1323 if (IS_ZEBRA_DEBUG_KERNEL) {
1324 char buf[PREFIX_STRLEN];
1325 zlog_debug(
1326 "netlink_route_multipath(): %s %s vrf %u(%u)",
1327 nl_msg_type_to_str(cmd),
1328 prefix2str(p, buf, sizeof(buf)),
1329 vrfid, tableid);
1330 }
1331 }
1332
1333 static void _netlink_mpls_debug(int cmd, uint32_t label, const char *routedesc)
1334 {
1335 if (IS_ZEBRA_DEBUG_KERNEL)
1336 zlog_debug("netlink_mpls_multipath() (%s): %s %u/20", routedesc,
1337 nl_msg_type_to_str(cmd), label);
1338 }
1339
1340 static int netlink_neigh_update(int cmd, int ifindex, uint32_t addr, char *lla,
1341 int llalen, ns_id_t ns_id)
1342 {
1343 struct {
1344 struct nlmsghdr n;
1345 struct ndmsg ndm;
1346 char buf[256];
1347 } req;
1348
1349 struct zebra_ns *zns = zebra_ns_lookup(ns_id);
1350
1351 memset(&req, 0, sizeof(req));
1352
1353 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
1354 req.n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
1355 req.n.nlmsg_type = cmd; // RTM_NEWNEIGH or RTM_DELNEIGH
1356 req.n.nlmsg_pid = zns->netlink_cmd.snl.nl_pid;
1357
1358 req.ndm.ndm_family = AF_INET;
1359 req.ndm.ndm_state = NUD_PERMANENT;
1360 req.ndm.ndm_ifindex = ifindex;
1361 req.ndm.ndm_type = RTN_UNICAST;
1362
1363 addattr_l(&req.n, sizeof(req), NDA_DST, &addr, 4);
1364 addattr_l(&req.n, sizeof(req), NDA_LLADDR, lla, llalen);
1365
1366 return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
1367 0);
1368 }
1369
1370 /* Routing table change via netlink interface. */
1371 /* Update flag indicates whether this is a "replace" or not. */
1372 static int netlink_route_multipath(int cmd, const struct prefix *p,
1373 const struct prefix *src_p,
1374 struct route_entry *re,
1375 int update)
1376 {
1377 int bytelen;
1378 struct sockaddr_nl snl;
1379 struct nexthop *nexthop = NULL;
1380 unsigned int nexthop_num;
1381 int family = PREFIX_FAMILY(p);
1382 const char *routedesc;
1383 int setsrc = 0;
1384 union g_addr src;
1385
1386 struct {
1387 struct nlmsghdr n;
1388 struct rtmsg r;
1389 char buf[NL_PKT_BUF_SIZE];
1390 } req;
1391
1392 struct zebra_ns *zns;
1393 struct zebra_vrf *zvrf = vrf_info_lookup(re->vrf_id);
1394
1395 zns = zvrf->zns;
1396 memset(&req, 0, sizeof req - NL_PKT_BUF_SIZE);
1397
1398 bytelen = (family == AF_INET ? 4 : 16);
1399
1400 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
1401 req.n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
1402 if ((cmd == RTM_NEWROUTE) && update)
1403 req.n.nlmsg_flags |= NLM_F_REPLACE;
1404 req.n.nlmsg_type = cmd;
1405 req.n.nlmsg_pid = zns->netlink_cmd.snl.nl_pid;
1406
1407 req.r.rtm_family = family;
1408 req.r.rtm_dst_len = p->prefixlen;
1409 req.r.rtm_src_len = src_p ? src_p->prefixlen : 0;
1410 req.r.rtm_protocol = zebra2proto(re->type);
1411 req.r.rtm_scope = RT_SCOPE_UNIVERSE;
1412
1413 /*
1414 * blackhole routes are not RTN_UNICAST, they are
1415 * RTN_ BLACKHOLE|UNREACHABLE|PROHIBIT
1416 * so setting this value as a RTN_UNICAST would
1417 * cause the route lookup of just the prefix
1418 * to fail. So no need to specify this for
1419 * the RTM_DELROUTE case
1420 */
1421 if (cmd != RTM_DELROUTE)
1422 req.r.rtm_type = RTN_UNICAST;
1423
1424 addattr_l(&req.n, sizeof req, RTA_DST, &p->u.prefix, bytelen);
1425 if (src_p)
1426 addattr_l(&req.n, sizeof req, RTA_SRC, &src_p->u.prefix,
1427 bytelen);
1428
1429 /* Metric. */
1430 /* Hardcode the metric for all routes coming from zebra. Metric isn't
1431 * used
1432 * either by the kernel or by zebra. Its purely for calculating best
1433 * path(s)
1434 * by the routing protocol and for communicating with protocol peers.
1435 */
1436 addattr32(&req.n, sizeof req, RTA_PRIORITY, NL_DEFAULT_ROUTE_METRIC);
1437 #if defined(SUPPORT_REALMS)
1438 if (re->tag > 0 && re->tag <= 255)
1439 addattr32(&req.n, sizeof req, RTA_FLOW, re->tag);
1440 #endif
1441 /* Table corresponding to this route. */
1442 if (re->table < 256)
1443 req.r.rtm_table = re->table;
1444 else {
1445 req.r.rtm_table = RT_TABLE_UNSPEC;
1446 addattr32(&req.n, sizeof req, RTA_TABLE, re->table);
1447 }
1448
1449 _netlink_route_debug(cmd, p, family, zvrf_id(zvrf), re->table);
1450
1451 /*
1452 * If we are not updating the route and we have received
1453 * a route delete, then all we need to fill in is the
1454 * prefix information to tell the kernel to schwack
1455 * it.
1456 */
1457 if (!update && cmd == RTM_DELROUTE)
1458 goto skip;
1459
1460 if (re->mtu || re->nexthop_mtu) {
1461 char buf[NL_PKT_BUF_SIZE];
1462 struct rtattr *rta = (void *)buf;
1463 uint32_t mtu = re->mtu;
1464 if (!mtu || (re->nexthop_mtu && re->nexthop_mtu < mtu))
1465 mtu = re->nexthop_mtu;
1466 rta->rta_type = RTA_METRICS;
1467 rta->rta_len = RTA_LENGTH(0);
1468 rta_addattr_l(rta, NL_PKT_BUF_SIZE, RTAX_MTU, &mtu, sizeof mtu);
1469 addattr_l(&req.n, NL_PKT_BUF_SIZE, RTA_METRICS, RTA_DATA(rta),
1470 RTA_PAYLOAD(rta));
1471 }
1472
1473 /* Count overall nexthops so we can decide whether to use singlepath
1474 * or multipath case. */
1475 nexthop_num = 0;
1476 for (ALL_NEXTHOPS(re->ng, nexthop)) {
1477 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
1478 continue;
1479 if (cmd == RTM_NEWROUTE && !NEXTHOP_IS_ACTIVE(nexthop->flags))
1480 continue;
1481 if (cmd == RTM_DELROUTE
1482 && !CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB))
1483 continue;
1484
1485 nexthop_num++;
1486 }
1487
1488 /* Singlepath case. */
1489 if (nexthop_num == 1 || multipath_num == 1) {
1490 nexthop_num = 0;
1491 for (ALL_NEXTHOPS(re->ng, nexthop)) {
1492 /*
1493 * So we want to cover 2 types of blackhole
1494 * routes here:
1495 * 1) A normal blackhole route( ala from a static
1496 * install.
1497 * 2) A recursively resolved blackhole route
1498 */
1499 if (nexthop->type == NEXTHOP_TYPE_BLACKHOLE) {
1500 switch (nexthop->bh_type) {
1501 case BLACKHOLE_ADMINPROHIB:
1502 req.r.rtm_type = RTN_PROHIBIT;
1503 break;
1504 case BLACKHOLE_REJECT:
1505 req.r.rtm_type = RTN_UNREACHABLE;
1506 break;
1507 default:
1508 req.r.rtm_type = RTN_BLACKHOLE;
1509 break;
1510 }
1511 goto skip;
1512 }
1513 if (CHECK_FLAG(nexthop->flags,
1514 NEXTHOP_FLAG_RECURSIVE)) {
1515 if (!setsrc) {
1516 if (family == AF_INET) {
1517 if (nexthop->rmap_src.ipv4
1518 .s_addr
1519 != 0) {
1520 src.ipv4 =
1521 nexthop->rmap_src
1522 .ipv4;
1523 setsrc = 1;
1524 } else if (nexthop->src.ipv4
1525 .s_addr
1526 != 0) {
1527 src.ipv4 =
1528 nexthop->src
1529 .ipv4;
1530 setsrc = 1;
1531 }
1532 } else if (family == AF_INET6) {
1533 if (!IN6_IS_ADDR_UNSPECIFIED(
1534 &nexthop->rmap_src
1535 .ipv6)) {
1536 src.ipv6 =
1537 nexthop->rmap_src
1538 .ipv6;
1539 setsrc = 1;
1540 } else if (
1541 !IN6_IS_ADDR_UNSPECIFIED(
1542 &nexthop->src
1543 .ipv6)) {
1544 src.ipv6 =
1545 nexthop->src
1546 .ipv6;
1547 setsrc = 1;
1548 }
1549 }
1550 }
1551 continue;
1552 }
1553
1554 if ((cmd == RTM_NEWROUTE
1555 && NEXTHOP_IS_ACTIVE(nexthop->flags))
1556 || (cmd == RTM_DELROUTE
1557 && CHECK_FLAG(nexthop->flags,
1558 NEXTHOP_FLAG_FIB))) {
1559 routedesc = nexthop->rparent
1560 ? "recursive, single-path"
1561 : "single-path";
1562
1563 _netlink_route_build_singlepath(
1564 routedesc, bytelen, nexthop, &req.n,
1565 &req.r, sizeof req, cmd);
1566 nexthop_num++;
1567 break;
1568 }
1569 }
1570 if (setsrc && (cmd == RTM_NEWROUTE)) {
1571 if (family == AF_INET)
1572 addattr_l(&req.n, sizeof req, RTA_PREFSRC,
1573 &src.ipv4, bytelen);
1574 else if (family == AF_INET6)
1575 addattr_l(&req.n, sizeof req, RTA_PREFSRC,
1576 &src.ipv6, bytelen);
1577 }
1578 } else {
1579 char buf[NL_PKT_BUF_SIZE];
1580 struct rtattr *rta = (void *)buf;
1581 struct rtnexthop *rtnh;
1582 union g_addr *src1 = NULL;
1583
1584 rta->rta_type = RTA_MULTIPATH;
1585 rta->rta_len = RTA_LENGTH(0);
1586 rtnh = RTA_DATA(rta);
1587
1588 nexthop_num = 0;
1589 for (ALL_NEXTHOPS(re->ng, nexthop)) {
1590 if (nexthop_num >= multipath_num)
1591 break;
1592
1593 if (CHECK_FLAG(nexthop->flags,
1594 NEXTHOP_FLAG_RECURSIVE)) {
1595 /* This only works for IPv4 now */
1596 if (!setsrc) {
1597 if (family == AF_INET) {
1598 if (nexthop->rmap_src.ipv4
1599 .s_addr
1600 != 0) {
1601 src.ipv4 =
1602 nexthop->rmap_src
1603 .ipv4;
1604 setsrc = 1;
1605 } else if (nexthop->src.ipv4
1606 .s_addr
1607 != 0) {
1608 src.ipv4 =
1609 nexthop->src
1610 .ipv4;
1611 setsrc = 1;
1612 }
1613 } else if (family == AF_INET6) {
1614 if (!IN6_IS_ADDR_UNSPECIFIED(
1615 &nexthop->rmap_src
1616 .ipv6)) {
1617 src.ipv6 =
1618 nexthop->rmap_src
1619 .ipv6;
1620 setsrc = 1;
1621 } else if (
1622 !IN6_IS_ADDR_UNSPECIFIED(
1623 &nexthop->src
1624 .ipv6)) {
1625 src.ipv6 =
1626 nexthop->src
1627 .ipv6;
1628 setsrc = 1;
1629 }
1630 }
1631 }
1632 continue;
1633 }
1634
1635 if ((cmd == RTM_NEWROUTE
1636 && NEXTHOP_IS_ACTIVE(nexthop->flags))
1637 || (cmd == RTM_DELROUTE
1638 && CHECK_FLAG(nexthop->flags,
1639 NEXTHOP_FLAG_FIB))) {
1640 routedesc = nexthop->rparent
1641 ? "recursive, multipath"
1642 : "multipath";
1643 nexthop_num++;
1644
1645 _netlink_route_build_multipath(
1646 routedesc, bytelen, nexthop, rta, rtnh,
1647 &req.r, &src1);
1648 rtnh = RTNH_NEXT(rtnh);
1649
1650 if (!setsrc && src1) {
1651 if (family == AF_INET)
1652 src.ipv4 = src1->ipv4;
1653 else if (family == AF_INET6)
1654 src.ipv6 = src1->ipv6;
1655
1656 setsrc = 1;
1657 }
1658 }
1659 }
1660 if (setsrc && (cmd == RTM_NEWROUTE)) {
1661 if (family == AF_INET)
1662 addattr_l(&req.n, sizeof req, RTA_PREFSRC,
1663 &src.ipv4, bytelen);
1664 else if (family == AF_INET6)
1665 addattr_l(&req.n, sizeof req, RTA_PREFSRC,
1666 &src.ipv6, bytelen);
1667 if (IS_ZEBRA_DEBUG_KERNEL)
1668 zlog_debug("Setting source");
1669 }
1670
1671 if (rta->rta_len > RTA_LENGTH(0))
1672 addattr_l(&req.n, NL_PKT_BUF_SIZE, RTA_MULTIPATH,
1673 RTA_DATA(rta), RTA_PAYLOAD(rta));
1674 }
1675
1676 /* If there is no useful nexthop then return. */
1677 if (nexthop_num == 0) {
1678 if (IS_ZEBRA_DEBUG_KERNEL)
1679 zlog_debug(
1680 "netlink_route_multipath(): No useful nexthop.");
1681 return 0;
1682 }
1683
1684 skip:
1685
1686 /* Destination netlink address. */
1687 memset(&snl, 0, sizeof snl);
1688 snl.nl_family = AF_NETLINK;
1689
1690 /* Talk to netlink socket. */
1691 return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
1692 0);
1693 }
1694
1695 int kernel_get_ipmr_sg_stats(struct zebra_vrf *zvrf, void *in)
1696 {
1697 int suc = 0;
1698 struct mcast_route_data *mr = (struct mcast_route_data *)in;
1699 struct {
1700 struct nlmsghdr n;
1701 struct ndmsg ndm;
1702 char buf[256];
1703 } req;
1704
1705 mroute = mr;
1706 struct zebra_ns *zns;
1707
1708 zns = zvrf->zns;
1709 memset(&req, 0, sizeof(req));
1710
1711 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
1712 req.n.nlmsg_flags = NLM_F_REQUEST;
1713 req.n.nlmsg_pid = zns->netlink_cmd.snl.nl_pid;
1714
1715 req.ndm.ndm_family = RTNL_FAMILY_IPMR;
1716 req.n.nlmsg_type = RTM_GETROUTE;
1717
1718 addattr_l(&req.n, sizeof(req), RTA_IIF, &mroute->ifindex, 4);
1719 addattr_l(&req.n, sizeof(req), RTA_OIF, &mroute->ifindex, 4);
1720 addattr_l(&req.n, sizeof(req), RTA_SRC, &mroute->sg.src.s_addr, 4);
1721 addattr_l(&req.n, sizeof(req), RTA_DST, &mroute->sg.grp.s_addr, 4);
1722 addattr_l(&req.n, sizeof(req), RTA_TABLE, &zvrf->table_id, 4);
1723
1724 suc = netlink_talk(netlink_route_change_read_multicast, &req.n,
1725 &zns->netlink_cmd, zns, 0);
1726
1727 mroute = NULL;
1728 return suc;
1729 }
1730
1731 enum dp_req_result kernel_route_rib(struct route_node *rn,
1732 const struct prefix *p,
1733 const struct prefix *src_p,
1734 struct route_entry *old,
1735 struct route_entry *new)
1736 {
1737 int ret = 0;
1738
1739 assert(old || new);
1740
1741 if (new) {
1742 if (p->family == AF_INET || v6_rr_semantics)
1743 ret = netlink_route_multipath(RTM_NEWROUTE, p, src_p,
1744 new, (old) ? 1 : 0);
1745 else {
1746 /*
1747 * So v6 route replace semantics are not in
1748 * the kernel at this point as I understand it.
1749 * So let's do a delete than an add.
1750 * In the future once v6 route replace semantics
1751 * are in we can figure out what to do here to
1752 * allow working with old and new kernels.
1753 *
1754 * I'm also intentionally ignoring the failure case
1755 * of the route delete. If that happens yeah we're
1756 * screwed.
1757 */
1758 if (old)
1759 netlink_route_multipath(RTM_DELROUTE, p, src_p,
1760 old, 0);
1761 ret = netlink_route_multipath(RTM_NEWROUTE, p, src_p,
1762 new, 0);
1763 }
1764 kernel_route_rib_pass_fail(rn, p, new,
1765 (!ret) ? DP_INSTALL_SUCCESS
1766 : DP_INSTALL_FAILURE);
1767 return DP_REQUEST_SUCCESS;
1768 }
1769
1770 if (old) {
1771 ret = netlink_route_multipath(RTM_DELROUTE, p, src_p, old, 0);
1772
1773 kernel_route_rib_pass_fail(rn, p, old,
1774 (!ret) ? DP_DELETE_SUCCESS
1775 : DP_DELETE_FAILURE);
1776 }
1777
1778 return DP_REQUEST_SUCCESS;
1779 }
1780
1781 int kernel_neigh_update(int add, int ifindex, uint32_t addr, char *lla,
1782 int llalen, ns_id_t ns_id)
1783 {
1784 return netlink_neigh_update(add ? RTM_NEWNEIGH : RTM_DELNEIGH, ifindex,
1785 addr, lla, llalen, ns_id);
1786 }
1787
1788 /*
1789 * Add remote VTEP to the flood list for this VxLAN interface (VNI). This
1790 * is done by adding an FDB entry with a MAC of 00:00:00:00:00:00.
1791 */
1792 static int netlink_vxlan_flood_list_update(struct interface *ifp,
1793 struct in_addr *vtep_ip, int cmd)
1794 {
1795 struct zebra_ns *zns;
1796 struct {
1797 struct nlmsghdr n;
1798 struct ndmsg ndm;
1799 char buf[256];
1800 } req;
1801 uint8_t dst_mac[6] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0};
1802 struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(ifp->vrf_id);
1803
1804 zns = zvrf->zns;
1805 memset(&req, 0, sizeof(req));
1806
1807 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
1808 req.n.nlmsg_flags = NLM_F_REQUEST;
1809 if (cmd == RTM_NEWNEIGH)
1810 req.n.nlmsg_flags |= (NLM_F_CREATE | NLM_F_APPEND);
1811 req.n.nlmsg_type = cmd;
1812 req.ndm.ndm_family = PF_BRIDGE;
1813 req.ndm.ndm_state = NUD_NOARP | NUD_PERMANENT;
1814 req.ndm.ndm_flags |= NTF_SELF; // Handle by "self", not "master"
1815
1816
1817 addattr_l(&req.n, sizeof(req), NDA_LLADDR, &dst_mac, 6);
1818 req.ndm.ndm_ifindex = ifp->ifindex;
1819 addattr_l(&req.n, sizeof(req), NDA_DST, &vtep_ip->s_addr, 4);
1820
1821 return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
1822 0);
1823 }
1824
1825 /*
1826 * Add remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1827 * adding
1828 * a "flood" MAC FDB entry.
1829 */
1830 int kernel_add_vtep(vni_t vni, struct interface *ifp, struct in_addr *vtep_ip)
1831 {
1832 if (IS_ZEBRA_DEBUG_VXLAN)
1833 zlog_debug("Install %s into flood list for VNI %u intf %s(%u)",
1834 inet_ntoa(*vtep_ip), vni, ifp->name, ifp->ifindex);
1835
1836 return netlink_vxlan_flood_list_update(ifp, vtep_ip, RTM_NEWNEIGH);
1837 }
1838
1839 /*
1840 * Remove remote VTEP for this VxLAN interface (VNI). In Linux, this involves
1841 * deleting the "flood" MAC FDB entry.
1842 */
1843 int kernel_del_vtep(vni_t vni, struct interface *ifp, struct in_addr *vtep_ip)
1844 {
1845 if (IS_ZEBRA_DEBUG_VXLAN)
1846 zlog_debug(
1847 "Uninstall %s from flood list for VNI %u intf %s(%u)",
1848 inet_ntoa(*vtep_ip), vni, ifp->name, ifp->ifindex);
1849
1850 return netlink_vxlan_flood_list_update(ifp, vtep_ip, RTM_DELNEIGH);
1851 }
1852
1853 #ifndef NDA_RTA
1854 #define NDA_RTA(r) \
1855 ((struct rtattr *)(((char *)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
1856 #endif
1857
1858 static int netlink_macfdb_change(struct nlmsghdr *h, int len, ns_id_t ns_id)
1859 {
1860 struct ndmsg *ndm;
1861 struct interface *ifp;
1862 struct zebra_if *zif;
1863 struct rtattr *tb[NDA_MAX + 1];
1864 struct interface *br_if;
1865 struct ethaddr mac;
1866 vlanid_t vid = 0;
1867 struct prefix vtep_ip;
1868 int vid_present = 0, dst_present = 0;
1869 char buf[ETHER_ADDR_STRLEN];
1870 char vid_buf[20];
1871 char dst_buf[30];
1872 uint8_t sticky = 0;
1873
1874 ndm = NLMSG_DATA(h);
1875
1876 /* We only process macfdb notifications if EVPN is enabled */
1877 if (!is_evpn_enabled())
1878 return 0;
1879
1880 /* The interface should exist. */
1881 ifp = if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id),
1882 ndm->ndm_ifindex);
1883 if (!ifp || !ifp->info)
1884 return 0;
1885
1886 /* The interface should be something we're interested in. */
1887 if (!IS_ZEBRA_IF_BRIDGE_SLAVE(ifp))
1888 return 0;
1889
1890 /* Drop "permanent" entries. */
1891 if (ndm->ndm_state & NUD_PERMANENT)
1892 return 0;
1893
1894 zif = (struct zebra_if *)ifp->info;
1895 if ((br_if = zif->brslave_info.br_if) == NULL) {
1896 zlog_warn("%s family %s IF %s(%u) brIF %u - no bridge master",
1897 nl_msg_type_to_str(h->nlmsg_type),
1898 nl_family_to_str(ndm->ndm_family), ifp->name,
1899 ndm->ndm_ifindex, zif->brslave_info.bridge_ifindex);
1900 return 0;
1901 }
1902
1903 /* Parse attributes and extract fields of interest. */
1904 memset(tb, 0, sizeof tb);
1905 netlink_parse_rtattr(tb, NDA_MAX, NDA_RTA(ndm), len);
1906
1907 if (!tb[NDA_LLADDR]) {
1908 zlog_warn("%s family %s IF %s(%u) brIF %u - no LLADDR",
1909 nl_msg_type_to_str(h->nlmsg_type),
1910 nl_family_to_str(ndm->ndm_family), ifp->name,
1911 ndm->ndm_ifindex, zif->brslave_info.bridge_ifindex);
1912 return 0;
1913 }
1914
1915 if (RTA_PAYLOAD(tb[NDA_LLADDR]) != ETH_ALEN) {
1916 zlog_warn(
1917 "%s family %s IF %s(%u) brIF %u - LLADDR is not MAC, len %lu",
1918 nl_msg_type_to_str(h->nlmsg_type),
1919 nl_family_to_str(ndm->ndm_family), ifp->name,
1920 ndm->ndm_ifindex, zif->brslave_info.bridge_ifindex,
1921 (unsigned long)RTA_PAYLOAD(tb[NDA_LLADDR]));
1922 return 0;
1923 }
1924
1925 memcpy(&mac, RTA_DATA(tb[NDA_LLADDR]), ETH_ALEN);
1926
1927 if ((NDA_VLAN <= NDA_MAX) && tb[NDA_VLAN]) {
1928 vid_present = 1;
1929 vid = *(uint16_t *)RTA_DATA(tb[NDA_VLAN]);
1930 sprintf(vid_buf, " VLAN %u", vid);
1931 }
1932
1933 if (tb[NDA_DST]) {
1934 /* TODO: Only IPv4 supported now. */
1935 dst_present = 1;
1936 vtep_ip.family = AF_INET;
1937 vtep_ip.prefixlen = IPV4_MAX_BITLEN;
1938 memcpy(&(vtep_ip.u.prefix4.s_addr), RTA_DATA(tb[NDA_DST]),
1939 IPV4_MAX_BYTELEN);
1940 sprintf(dst_buf, " dst %s", inet_ntoa(vtep_ip.u.prefix4));
1941 }
1942
1943 sticky = (ndm->ndm_state & NUD_NOARP) ? 1 : 0;
1944
1945 if (IS_ZEBRA_DEBUG_KERNEL)
1946 zlog_debug("Rx %s family %s IF %s(%u)%s %sMAC %s%s",
1947 nl_msg_type_to_str(h->nlmsg_type),
1948 nl_family_to_str(ndm->ndm_family), ifp->name,
1949 ndm->ndm_ifindex, vid_present ? vid_buf : "",
1950 sticky ? "sticky " : "",
1951 prefix_mac2str(&mac, buf, sizeof(buf)),
1952 dst_present ? dst_buf : "");
1953
1954 if (filter_vlan && vid != filter_vlan)
1955 return 0;
1956
1957 /* If add or update, do accordingly if learnt on a "local" interface; if
1958 * the notification is over VxLAN, this has to be related to
1959 * multi-homing,
1960 * so perform an implicit delete of any local entry (if it exists).
1961 */
1962 if (h->nlmsg_type == RTM_NEWNEIGH) {
1963 /* Drop "permanent" entries. */
1964 if (ndm->ndm_state & NUD_PERMANENT)
1965 return 0;
1966
1967 if (IS_ZEBRA_IF_VXLAN(ifp))
1968 return zebra_vxlan_check_del_local_mac(ifp, br_if, &mac,
1969 vid);
1970
1971 return zebra_vxlan_local_mac_add_update(ifp, br_if, &mac, vid,
1972 sticky);
1973 }
1974
1975 /* This is a delete notification.
1976 * 1. For a MAC over VxLan, check if it needs to be refreshed(readded)
1977 * 2. For a MAC over "local" interface, delete the mac
1978 * Note: We will get notifications from both bridge driver and VxLAN
1979 * driver.
1980 * Ignore the notification from VxLan driver as it is also generated
1981 * when mac moves from remote to local.
1982 */
1983 if (dst_present)
1984 return 0;
1985
1986 if (IS_ZEBRA_IF_VXLAN(ifp))
1987 return zebra_vxlan_check_readd_remote_mac(ifp, br_if, &mac,
1988 vid);
1989
1990 return zebra_vxlan_local_mac_del(ifp, br_if, &mac, vid);
1991 }
1992
1993 static int netlink_macfdb_table(struct nlmsghdr *h, ns_id_t ns_id, int startup)
1994 {
1995 int len;
1996 struct ndmsg *ndm;
1997
1998 if (h->nlmsg_type != RTM_NEWNEIGH)
1999 return 0;
2000
2001 /* Length validity. */
2002 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct ndmsg));
2003 if (len < 0)
2004 return -1;
2005
2006 /* We are interested only in AF_BRIDGE notifications. */
2007 ndm = NLMSG_DATA(h);
2008 if (ndm->ndm_family != AF_BRIDGE)
2009 return 0;
2010
2011 return netlink_macfdb_change(h, len, ns_id);
2012 }
2013
2014 /* Request for MAC FDB information from the kernel */
2015 static int netlink_request_macs(struct zebra_ns *zns, int family, int type,
2016 ifindex_t master_ifindex)
2017 {
2018 struct {
2019 struct nlmsghdr n;
2020 struct ifinfomsg ifm;
2021 char buf[256];
2022 } req;
2023
2024 /* Form the request, specifying filter (rtattr) if needed. */
2025 memset(&req, 0, sizeof(req));
2026 req.n.nlmsg_type = type;
2027 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg));
2028 req.ifm.ifi_family = family;
2029 if (master_ifindex)
2030 addattr32(&req.n, sizeof(req), IFLA_MASTER, master_ifindex);
2031
2032 return netlink_request(&zns->netlink_cmd, &req.n);
2033 }
2034
2035 /*
2036 * MAC forwarding database read using netlink interface. This is invoked
2037 * at startup.
2038 */
2039 int netlink_macfdb_read(struct zebra_ns *zns)
2040 {
2041 int ret;
2042
2043 /* Get bridge FDB table. */
2044 ret = netlink_request_macs(zns, AF_BRIDGE, RTM_GETNEIGH, 0);
2045 if (ret < 0)
2046 return ret;
2047 /* We are reading entire table. */
2048 filter_vlan = 0;
2049 ret = netlink_parse_info(netlink_macfdb_table, &zns->netlink_cmd, zns,
2050 0, 1);
2051
2052 return ret;
2053 }
2054
2055 /*
2056 * MAC forwarding database read using netlink interface. This is for a
2057 * specific bridge and matching specific access VLAN (if VLAN-aware bridge).
2058 */
2059 int netlink_macfdb_read_for_bridge(struct zebra_ns *zns, struct interface *ifp,
2060 struct interface *br_if)
2061 {
2062 struct zebra_if *br_zif;
2063 struct zebra_if *zif;
2064 struct zebra_l2info_vxlan *vxl;
2065 int ret = 0;
2066
2067
2068 /* Save VLAN we're filtering on, if needed. */
2069 br_zif = (struct zebra_if *)br_if->info;
2070 zif = (struct zebra_if *)ifp->info;
2071 vxl = &zif->l2info.vxl;
2072 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif))
2073 filter_vlan = vxl->access_vlan;
2074
2075 /* Get bridge FDB table for specific bridge - we do the VLAN filtering.
2076 */
2077 ret = netlink_request_macs(zns, AF_BRIDGE, RTM_GETNEIGH,
2078 br_if->ifindex);
2079 if (ret < 0)
2080 return ret;
2081 ret = netlink_parse_info(netlink_macfdb_table, &zns->netlink_cmd, zns,
2082 0, 0);
2083
2084 /* Reset VLAN filter. */
2085 filter_vlan = 0;
2086 return ret;
2087 }
2088
2089 static int netlink_macfdb_update(struct interface *ifp, vlanid_t vid,
2090 struct ethaddr *mac, struct in_addr vtep_ip,
2091 int local, int cmd, uint8_t sticky)
2092 {
2093 struct zebra_ns *zns;
2094 struct {
2095 struct nlmsghdr n;
2096 struct ndmsg ndm;
2097 char buf[256];
2098 } req;
2099 int dst_alen;
2100 struct zebra_if *zif;
2101 struct interface *br_if;
2102 struct zebra_if *br_zif;
2103 char buf[ETHER_ADDR_STRLEN];
2104 int vid_present = 0, dst_present = 0;
2105 char vid_buf[20];
2106 char dst_buf[30];
2107 struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(ifp->vrf_id);
2108
2109 zns = zvrf->zns;
2110 zif = ifp->info;
2111 if ((br_if = zif->brslave_info.br_if) == NULL) {
2112 zlog_warn("MAC %s on IF %s(%u) - no mapping to bridge",
2113 (cmd == RTM_NEWNEIGH) ? "add" : "del", ifp->name,
2114 ifp->ifindex);
2115 return -1;
2116 }
2117
2118 memset(&req, 0, sizeof(req));
2119
2120 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
2121 req.n.nlmsg_flags = NLM_F_REQUEST;
2122 if (cmd == RTM_NEWNEIGH)
2123 req.n.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
2124 req.n.nlmsg_type = cmd;
2125 req.ndm.ndm_family = AF_BRIDGE;
2126 req.ndm.ndm_flags |= NTF_SELF | NTF_MASTER;
2127 req.ndm.ndm_state = NUD_REACHABLE;
2128
2129 if (sticky)
2130 req.ndm.ndm_state |= NUD_NOARP;
2131 else
2132 req.ndm.ndm_flags |= NTF_EXT_LEARNED;
2133
2134 addattr_l(&req.n, sizeof(req), NDA_LLADDR, mac, 6);
2135 req.ndm.ndm_ifindex = ifp->ifindex;
2136 if (!local) {
2137 dst_alen = 4; // TODO: hardcoded
2138 addattr_l(&req.n, sizeof(req), NDA_DST, &vtep_ip, dst_alen);
2139 dst_present = 1;
2140 sprintf(dst_buf, " dst %s", inet_ntoa(vtep_ip));
2141 }
2142 br_zif = (struct zebra_if *)br_if->info;
2143 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif) && vid > 0) {
2144 addattr16(&req.n, sizeof(req), NDA_VLAN, vid);
2145 vid_present = 1;
2146 sprintf(vid_buf, " VLAN %u", vid);
2147 }
2148 addattr32(&req.n, sizeof(req), NDA_MASTER, br_if->ifindex);
2149
2150 if (IS_ZEBRA_DEBUG_KERNEL)
2151 zlog_debug("Tx %s family %s IF %s(%u)%s %sMAC %s%s",
2152 nl_msg_type_to_str(cmd),
2153 nl_family_to_str(req.ndm.ndm_family), ifp->name,
2154 ifp->ifindex, vid_present ? vid_buf : "",
2155 sticky ? "sticky " : "",
2156 prefix_mac2str(mac, buf, sizeof(buf)),
2157 dst_present ? dst_buf : "");
2158
2159 return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
2160 0);
2161 }
2162
2163 #define NUD_VALID \
2164 (NUD_PERMANENT | NUD_NOARP | NUD_REACHABLE | NUD_PROBE | NUD_STALE \
2165 | NUD_DELAY)
2166
2167 static int netlink_ipneigh_change(struct nlmsghdr *h, int len, ns_id_t ns_id)
2168 {
2169 struct ndmsg *ndm;
2170 struct interface *ifp;
2171 struct zebra_if *zif;
2172 struct rtattr *tb[NDA_MAX + 1];
2173 struct interface *link_if;
2174 struct ethaddr mac;
2175 struct ipaddr ip;
2176 char buf[ETHER_ADDR_STRLEN];
2177 char buf2[INET6_ADDRSTRLEN];
2178 int mac_present = 0;
2179 uint8_t ext_learned;
2180 uint8_t router_flag;
2181
2182 ndm = NLMSG_DATA(h);
2183
2184 /* The interface should exist. */
2185 ifp = if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id),
2186 ndm->ndm_ifindex);
2187 if (!ifp || !ifp->info)
2188 return 0;
2189
2190 zif = (struct zebra_if *)ifp->info;
2191
2192 /* Parse attributes and extract fields of interest. */
2193 memset(tb, 0, sizeof tb);
2194 netlink_parse_rtattr(tb, NDA_MAX, NDA_RTA(ndm), len);
2195
2196 if (!tb[NDA_DST]) {
2197 zlog_warn("%s family %s IF %s(%u) - no DST",
2198 nl_msg_type_to_str(h->nlmsg_type),
2199 nl_family_to_str(ndm->ndm_family), ifp->name,
2200 ndm->ndm_ifindex);
2201 return 0;
2202 }
2203
2204 memset(&ip, 0, sizeof(struct ipaddr));
2205 ip.ipa_type = (ndm->ndm_family == AF_INET) ? IPADDR_V4 : IPADDR_V6;
2206 memcpy(&ip.ip.addr, RTA_DATA(tb[NDA_DST]), RTA_PAYLOAD(tb[NDA_DST]));
2207
2208 /* Drop some "permanent" entries. */
2209 if (ndm->ndm_state & NUD_PERMANENT) {
2210 char buf[16] = "169.254.0.1";
2211 struct in_addr ipv4_ll;
2212
2213 if (ndm->ndm_family != AF_INET)
2214 return 0;
2215
2216 if (!zif->v6_2_v4_ll_neigh_entry)
2217 return 0;
2218
2219 if (h->nlmsg_type != RTM_DELNEIGH)
2220 return 0;
2221
2222 inet_pton(AF_INET, buf, &ipv4_ll);
2223 if (ipv4_ll.s_addr != ip.ip._v4_addr.s_addr)
2224 return 0;
2225
2226 if_nbr_ipv6ll_to_ipv4ll_neigh_update(
2227 ifp, &zif->v6_2_v4_ll_addr6, true);
2228 return 0;
2229 }
2230
2231 /* The neighbor is present on an SVI. From this, we locate the
2232 * underlying
2233 * bridge because we're only interested in neighbors on a VxLAN bridge.
2234 * The bridge is located based on the nature of the SVI:
2235 * (a) In the case of a VLAN-aware bridge, the SVI is a L3 VLAN
2236 * interface
2237 * and is linked to the bridge
2238 * (b) In the case of a VLAN-unaware bridge, the SVI is the bridge
2239 * inteface
2240 * itself
2241 */
2242 if (IS_ZEBRA_IF_VLAN(ifp)) {
2243 link_if = if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id),
2244 zif->link_ifindex);
2245 if (!link_if)
2246 return 0;
2247 } else if (IS_ZEBRA_IF_BRIDGE(ifp))
2248 link_if = ifp;
2249 else
2250 return 0;
2251
2252 memset(&mac, 0, sizeof(struct ethaddr));
2253 if (h->nlmsg_type == RTM_NEWNEIGH) {
2254 if (tb[NDA_LLADDR]) {
2255 if (RTA_PAYLOAD(tb[NDA_LLADDR]) != ETH_ALEN) {
2256 zlog_warn(
2257 "%s family %s IF %s(%u) - LLADDR is not MAC, len %lu",
2258 nl_msg_type_to_str(h->nlmsg_type),
2259 nl_family_to_str(ndm->ndm_family),
2260 ifp->name, ndm->ndm_ifindex,
2261 (unsigned long)RTA_PAYLOAD(
2262 tb[NDA_LLADDR]));
2263 return 0;
2264 }
2265
2266 mac_present = 1;
2267 memcpy(&mac, RTA_DATA(tb[NDA_LLADDR]), ETH_ALEN);
2268 }
2269
2270 ext_learned = (ndm->ndm_flags & NTF_EXT_LEARNED) ? 1 : 0;
2271 router_flag = (ndm->ndm_flags & NTF_ROUTER) ? 1 : 0;
2272
2273 if (IS_ZEBRA_DEBUG_KERNEL)
2274 zlog_debug(
2275 "Rx %s family %s IF %s(%u) IP %s MAC %s state 0x%x flags 0x%x",
2276 nl_msg_type_to_str(h->nlmsg_type),
2277 nl_family_to_str(ndm->ndm_family), ifp->name,
2278 ndm->ndm_ifindex,
2279 ipaddr2str(&ip, buf2, sizeof(buf2)),
2280 mac_present
2281 ? prefix_mac2str(&mac, buf, sizeof(buf))
2282 : "",
2283 ndm->ndm_state, ndm->ndm_flags);
2284
2285 /* If the neighbor state is valid for use, process as an add or
2286 * update
2287 * else process as a delete. Note that the delete handling may
2288 * result
2289 * in re-adding the neighbor if it is a valid "remote" neighbor.
2290 */
2291 if (ndm->ndm_state & NUD_VALID)
2292 return zebra_vxlan_handle_kernel_neigh_update(
2293 ifp, link_if, &ip, &mac, ndm->ndm_state,
2294 ext_learned, router_flag);
2295
2296 return zebra_vxlan_handle_kernel_neigh_del(ifp, link_if, &ip);
2297 }
2298
2299 if (IS_ZEBRA_DEBUG_KERNEL)
2300 zlog_debug("Rx %s family %s IF %s(%u) IP %s",
2301 nl_msg_type_to_str(h->nlmsg_type),
2302 nl_family_to_str(ndm->ndm_family), ifp->name,
2303 ndm->ndm_ifindex,
2304 ipaddr2str(&ip, buf2, sizeof(buf2)));
2305
2306 /* Process the delete - it may result in re-adding the neighbor if it is
2307 * a valid "remote" neighbor.
2308 */
2309 return zebra_vxlan_handle_kernel_neigh_del(ifp, link_if, &ip);
2310 }
2311
2312 static int netlink_neigh_table(struct nlmsghdr *h, ns_id_t ns_id, int startup)
2313 {
2314 int len;
2315 struct ndmsg *ndm;
2316
2317 if (h->nlmsg_type != RTM_NEWNEIGH)
2318 return 0;
2319
2320 /* Length validity. */
2321 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct ndmsg));
2322 if (len < 0)
2323 return -1;
2324
2325 /* We are interested only in AF_INET or AF_INET6 notifications. */
2326 ndm = NLMSG_DATA(h);
2327 if (ndm->ndm_family != AF_INET && ndm->ndm_family != AF_INET6)
2328 return 0;
2329
2330 return netlink_neigh_change(h, len);
2331 }
2332
2333 /* Request for IP neighbor information from the kernel */
2334 static int netlink_request_neigh(struct zebra_ns *zns, int family, int type,
2335 ifindex_t ifindex)
2336 {
2337 struct {
2338 struct nlmsghdr n;
2339 struct ndmsg ndm;
2340 char buf[256];
2341 } req;
2342
2343 /* Form the request, specifying filter (rtattr) if needed. */
2344 memset(&req, 0, sizeof(req));
2345 req.n.nlmsg_type = type;
2346 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
2347 req.ndm.ndm_family = family;
2348 if (ifindex)
2349 addattr32(&req.n, sizeof(req), NDA_IFINDEX, ifindex);
2350
2351 return netlink_request(&zns->netlink_cmd, &req.n);
2352 }
2353
2354 /*
2355 * IP Neighbor table read using netlink interface. This is invoked
2356 * at startup.
2357 */
2358 int netlink_neigh_read(struct zebra_ns *zns)
2359 {
2360 int ret;
2361
2362 /* Get IP neighbor table. */
2363 ret = netlink_request_neigh(zns, AF_UNSPEC, RTM_GETNEIGH, 0);
2364 if (ret < 0)
2365 return ret;
2366 ret = netlink_parse_info(netlink_neigh_table, &zns->netlink_cmd, zns, 0,
2367 1);
2368
2369 return ret;
2370 }
2371
2372 /*
2373 * IP Neighbor table read using netlink interface. This is for a specific
2374 * VLAN device.
2375 */
2376 int netlink_neigh_read_for_vlan(struct zebra_ns *zns, struct interface *vlan_if)
2377 {
2378 int ret = 0;
2379
2380 ret = netlink_request_neigh(zns, AF_UNSPEC, RTM_GETNEIGH,
2381 vlan_if->ifindex);
2382 if (ret < 0)
2383 return ret;
2384 ret = netlink_parse_info(netlink_neigh_table, &zns->netlink_cmd, zns, 0,
2385 0);
2386
2387 return ret;
2388 }
2389
2390 int netlink_neigh_change(struct nlmsghdr *h, ns_id_t ns_id)
2391 {
2392 int len;
2393 struct ndmsg *ndm;
2394
2395 if (!(h->nlmsg_type == RTM_NEWNEIGH || h->nlmsg_type == RTM_DELNEIGH))
2396 return 0;
2397
2398 /* Length validity. */
2399 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct ndmsg));
2400 if (len < 0) {
2401 zlog_err("%s: Message received from netlink is of a broken size %d %zu",
2402 __PRETTY_FUNCTION__, h->nlmsg_len,
2403 (size_t)NLMSG_LENGTH(sizeof(struct ndmsg)));
2404 return -1;
2405 }
2406
2407 /* Is this a notification for the MAC FDB or IP neighbor table? */
2408 ndm = NLMSG_DATA(h);
2409 if (ndm->ndm_family == AF_BRIDGE)
2410 return netlink_macfdb_change(h, len, ns_id);
2411
2412 if (ndm->ndm_type != RTN_UNICAST)
2413 return 0;
2414
2415 if (ndm->ndm_family == AF_INET || ndm->ndm_family == AF_INET6)
2416 return netlink_ipneigh_change(h, len, ns_id);
2417 else {
2418 zlog_warn(
2419 "Invalid address family: %u received from kernel neighbor change: %u",
2420 ndm->ndm_family, h->nlmsg_type);
2421 return 0;
2422 }
2423
2424 return 0;
2425 }
2426
2427 static int netlink_neigh_update2(struct interface *ifp, struct ipaddr *ip,
2428 struct ethaddr *mac, uint8_t flags,
2429 uint16_t state, int cmd)
2430 {
2431 struct {
2432 struct nlmsghdr n;
2433 struct ndmsg ndm;
2434 char buf[256];
2435 } req;
2436 int ipa_len;
2437
2438 struct zebra_ns *zns;
2439 char buf[INET6_ADDRSTRLEN];
2440 char buf2[ETHER_ADDR_STRLEN];
2441 struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(ifp->vrf_id);
2442
2443 zns = zvrf->zns;
2444 memset(&req, 0, sizeof(req));
2445
2446 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
2447 req.n.nlmsg_flags = NLM_F_REQUEST;
2448 if (cmd == RTM_NEWNEIGH)
2449 req.n.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
2450 req.n.nlmsg_type = cmd; // RTM_NEWNEIGH or RTM_DELNEIGH
2451 req.ndm.ndm_family = IS_IPADDR_V4(ip) ? AF_INET : AF_INET6;
2452 req.ndm.ndm_state = state;
2453 req.ndm.ndm_ifindex = ifp->ifindex;
2454 req.ndm.ndm_type = RTN_UNICAST;
2455 req.ndm.ndm_flags = flags;
2456
2457 ipa_len = IS_IPADDR_V4(ip) ? IPV4_MAX_BYTELEN : IPV6_MAX_BYTELEN;
2458 addattr_l(&req.n, sizeof(req), NDA_DST, &ip->ip.addr, ipa_len);
2459 if (mac)
2460 addattr_l(&req.n, sizeof(req), NDA_LLADDR, mac, 6);
2461
2462 if (IS_ZEBRA_DEBUG_KERNEL)
2463 zlog_debug("Tx %s family %s IF %s(%u) Neigh %s MAC %s flags 0x%x",
2464 nl_msg_type_to_str(cmd),
2465 nl_family_to_str(req.ndm.ndm_family), ifp->name,
2466 ifp->ifindex, ipaddr2str(ip, buf, sizeof(buf)),
2467 mac ? prefix_mac2str(mac, buf2, sizeof(buf2))
2468 : "null", flags);
2469
2470 return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
2471 0);
2472 }
2473
2474 int kernel_add_mac(struct interface *ifp, vlanid_t vid, struct ethaddr *mac,
2475 struct in_addr vtep_ip, uint8_t sticky)
2476 {
2477 return netlink_macfdb_update(ifp, vid, mac, vtep_ip, 0, RTM_NEWNEIGH,
2478 sticky);
2479 }
2480
2481 int kernel_del_mac(struct interface *ifp, vlanid_t vid, struct ethaddr *mac,
2482 struct in_addr vtep_ip, int local)
2483 {
2484 return netlink_macfdb_update(ifp, vid, mac, vtep_ip, local,
2485 RTM_DELNEIGH, 0);
2486 }
2487
2488 int kernel_add_neigh(struct interface *ifp, struct ipaddr *ip,
2489 struct ethaddr *mac, uint8_t flags)
2490 {
2491 return netlink_neigh_update2(ifp, ip, mac, flags,
2492 NUD_NOARP, RTM_NEWNEIGH);
2493 }
2494
2495 int kernel_del_neigh(struct interface *ifp, struct ipaddr *ip)
2496 {
2497 return netlink_neigh_update2(ifp, ip, NULL, 0, 0, RTM_DELNEIGH);
2498 }
2499
2500 /*
2501 * MPLS label forwarding table change via netlink interface.
2502 */
2503 int netlink_mpls_multipath(int cmd, zebra_lsp_t *lsp)
2504 {
2505 mpls_lse_t lse;
2506 zebra_nhlfe_t *nhlfe;
2507 struct nexthop *nexthop = NULL;
2508 unsigned int nexthop_num;
2509 const char *routedesc;
2510 struct zebra_ns *zns = zebra_ns_lookup(NS_DEFAULT);
2511 int route_type;
2512
2513 struct {
2514 struct nlmsghdr n;
2515 struct rtmsg r;
2516 char buf[NL_PKT_BUF_SIZE];
2517 } req;
2518
2519 memset(&req, 0, sizeof req - NL_PKT_BUF_SIZE);
2520
2521 /*
2522 * Count # nexthops so we can decide whether to use singlepath
2523 * or multipath case.
2524 */
2525 nexthop_num = 0;
2526 for (nhlfe = lsp->nhlfe_list; nhlfe; nhlfe = nhlfe->next) {
2527 nexthop = nhlfe->nexthop;
2528 if (!nexthop)
2529 continue;
2530 if (cmd == RTM_NEWROUTE) {
2531 /* Count all selected NHLFEs */
2532 if (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_SELECTED)
2533 && CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE))
2534 nexthop_num++;
2535 } else /* DEL */
2536 {
2537 /* Count all installed NHLFEs */
2538 if (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_INSTALLED)
2539 && CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB))
2540 nexthop_num++;
2541 }
2542 }
2543
2544 if ((nexthop_num == 0) || (!lsp->best_nhlfe && (cmd != RTM_DELROUTE)))
2545 return 0;
2546
2547 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
2548 req.n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
2549 req.n.nlmsg_type = cmd;
2550 req.n.nlmsg_pid = zns->netlink_cmd.snl.nl_pid;
2551
2552 req.r.rtm_family = AF_MPLS;
2553 req.r.rtm_table = RT_TABLE_MAIN;
2554 req.r.rtm_dst_len = MPLS_LABEL_LEN_BITS;
2555 req.r.rtm_scope = RT_SCOPE_UNIVERSE;
2556 req.r.rtm_type = RTN_UNICAST;
2557
2558 if (cmd == RTM_NEWROUTE) {
2559 /* We do a replace to handle update. */
2560 req.n.nlmsg_flags |= NLM_F_REPLACE;
2561
2562 /* set the protocol value if installing */
2563 route_type = re_type_from_lsp_type(lsp->best_nhlfe->type);
2564 req.r.rtm_protocol = zebra2proto(route_type);
2565 }
2566
2567 /* Fill destination */
2568 lse = mpls_lse_encode(lsp->ile.in_label, 0, 0, 1);
2569 addattr_l(&req.n, sizeof req, RTA_DST, &lse, sizeof(mpls_lse_t));
2570
2571 /* Fill nexthops (paths) based on single-path or multipath. The paths
2572 * chosen depend on the operation.
2573 */
2574 if (nexthop_num == 1 || multipath_num == 1) {
2575 routedesc = "single-path";
2576 _netlink_mpls_debug(cmd, lsp->ile.in_label, routedesc);
2577
2578 nexthop_num = 0;
2579 for (nhlfe = lsp->nhlfe_list; nhlfe; nhlfe = nhlfe->next) {
2580 nexthop = nhlfe->nexthop;
2581 if (!nexthop)
2582 continue;
2583
2584 if ((cmd == RTM_NEWROUTE
2585 && (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_SELECTED)
2586 && CHECK_FLAG(nexthop->flags,
2587 NEXTHOP_FLAG_ACTIVE)))
2588 || (cmd == RTM_DELROUTE
2589 && (CHECK_FLAG(nhlfe->flags,
2590 NHLFE_FLAG_INSTALLED)
2591 && CHECK_FLAG(nexthop->flags,
2592 NEXTHOP_FLAG_FIB)))) {
2593 /* Add the gateway */
2594 _netlink_mpls_build_singlepath(routedesc, nhlfe,
2595 &req.n, &req.r,
2596 sizeof req, cmd);
2597 nexthop_num++;
2598 break;
2599 }
2600 }
2601 } else /* Multipath case */
2602 {
2603 char buf[NL_PKT_BUF_SIZE];
2604 struct rtattr *rta = (void *)buf;
2605 struct rtnexthop *rtnh;
2606 union g_addr *src1 = NULL;
2607
2608 rta->rta_type = RTA_MULTIPATH;
2609 rta->rta_len = RTA_LENGTH(0);
2610 rtnh = RTA_DATA(rta);
2611
2612 routedesc = "multipath";
2613 _netlink_mpls_debug(cmd, lsp->ile.in_label, routedesc);
2614
2615 nexthop_num = 0;
2616 for (nhlfe = lsp->nhlfe_list; nhlfe; nhlfe = nhlfe->next) {
2617 nexthop = nhlfe->nexthop;
2618 if (!nexthop)
2619 continue;
2620
2621 if (nexthop_num >= multipath_num)
2622 break;
2623
2624 if ((cmd == RTM_NEWROUTE
2625 && (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_SELECTED)
2626 && CHECK_FLAG(nexthop->flags,
2627 NEXTHOP_FLAG_ACTIVE)))
2628 || (cmd == RTM_DELROUTE
2629 && (CHECK_FLAG(nhlfe->flags,
2630 NHLFE_FLAG_INSTALLED)
2631 && CHECK_FLAG(nexthop->flags,
2632 NEXTHOP_FLAG_FIB)))) {
2633 nexthop_num++;
2634
2635 /* Build the multipath */
2636 _netlink_mpls_build_multipath(routedesc, nhlfe,
2637 rta, rtnh, &req.r,
2638 &src1);
2639 rtnh = RTNH_NEXT(rtnh);
2640 }
2641 }
2642
2643 /* Add the multipath */
2644 if (rta->rta_len > RTA_LENGTH(0))
2645 addattr_l(&req.n, NL_PKT_BUF_SIZE, RTA_MULTIPATH,
2646 RTA_DATA(rta), RTA_PAYLOAD(rta));
2647 }
2648
2649 /* Talk to netlink socket. */
2650 return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
2651 0);
2652 }
2653 #endif /* HAVE_NETLINK */
FRRouting mirror