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 #include <linux/nexthop.h>
31
32 /* Hack for GNU libc version 2. */
33 #ifndef MSG_TRUNC
34 #define MSG_TRUNC 0x20
35 #endif /* MSG_TRUNC */
36
37 #include "linklist.h"
38 #include "if.h"
39 #include "log.h"
40 #include "prefix.h"
41 #include "connected.h"
42 #include "table.h"
43 #include "memory.h"
44 #include "zebra_memory.h"
45 #include "rib.h"
46 #include "thread.h"
47 #include "privs.h"
48 #include "nexthop.h"
49 #include "vrf.h"
50 #include "vty.h"
51 #include "mpls.h"
52 #include "vxlan.h"
53 #include "printfrr.h"
54
55 #include "zebra/zapi_msg.h"
56 #include "zebra/zebra_ns.h"
57 #include "zebra/zebra_vrf.h"
58 #include "zebra/rt.h"
59 #include "zebra/redistribute.h"
60 #include "zebra/interface.h"
61 #include "zebra/debug.h"
62 #include "zebra/rtadv.h"
63 #include "zebra/zebra_ptm.h"
64 #include "zebra/zebra_mpls.h"
65 #include "zebra/kernel_netlink.h"
66 #include "zebra/rt_netlink.h"
67 #include "zebra/zebra_nhg.h"
68 #include "zebra/zebra_mroute.h"
69 #include "zebra/zebra_vxlan.h"
70 #include "zebra/zebra_errors.h"
71 #include "zebra/zebra_evpn_mh.h"
72
73 #ifndef AF_MPLS
74 #define AF_MPLS 28
75 #endif
76
77 /* Re-defining as I am unable to include <linux/if_bridge.h> which has the
78 * UAPI for MAC sync. */
79 #ifndef _UAPI_LINUX_IF_BRIDGE_H
80 /* FDB notification bits for NDA_NOTIFY:
81 * - BR_FDB_NFY_STATIC - notify on activity/expire even for a static entry
82 * - BR_FDB_NFY_INACTIVE - mark as inactive to avoid double notification,
83 * used with BR_FDB_NFY_STATIC (kernel controlled)
84 */
85 enum {
86 BR_FDB_NFY_STATIC,
87 BR_FDB_NFY_INACTIVE,
88 BR_FDB_NFY_MAX
89 };
90 #endif
91
92 static vlanid_t filter_vlan = 0;
93
94 /* We capture whether the current kernel supports nexthop ids; by
95 * default, we'll use them if possible. There's also a configuration
96 * available to _disable_ use of kernel nexthops.
97 */
98 static bool supports_nh;
99
100 struct gw_family_t {
101 uint16_t filler;
102 uint16_t family;
103 union g_addr gate;
104 };
105
106 static const char ipv4_ll_buf[16] = "169.254.0.1";
107 static struct in_addr ipv4_ll;
108
109 /* Is this a ipv4 over ipv6 route? */
is_route_v4_over_v6(unsigned char rtm_family,enum nexthop_types_t nexthop_type)110 static bool is_route_v4_over_v6(unsigned char rtm_family,
111 enum nexthop_types_t nexthop_type)
112 {
113 if (rtm_family == AF_INET
114 && (nexthop_type == NEXTHOP_TYPE_IPV6
115 || nexthop_type == NEXTHOP_TYPE_IPV6_IFINDEX))
116 return true;
117
118 return false;
119 }
120
121 /* Helper to control use of kernel-level nexthop ids */
kernel_nexthops_supported(void)122 static bool kernel_nexthops_supported(void)
123 {
124 return (supports_nh && !vrf_is_backend_netns()
125 && zebra_nhg_kernel_nexthops_enabled());
126 }
127
128 /*
129 * The ipv4_ll data structure is used for all 5549
130 * additions to the kernel. Let's figure out the
131 * correct value one time instead for every
132 * install/remove of a 5549 type route
133 */
rt_netlink_init(void)134 void rt_netlink_init(void)
135 {
136 inet_pton(AF_INET, ipv4_ll_buf, &ipv4_ll);
137 }
138
139 /*
140 * Mapping from dataplane neighbor flags to netlink flags
141 */
neigh_flags_to_netlink(uint8_t dplane_flags)142 static uint8_t neigh_flags_to_netlink(uint8_t dplane_flags)
143 {
144 uint8_t flags = 0;
145
146 if (dplane_flags & DPLANE_NTF_EXT_LEARNED)
147 flags |= NTF_EXT_LEARNED;
148 if (dplane_flags & DPLANE_NTF_ROUTER)
149 flags |= NTF_ROUTER;
150 if (dplane_flags & DPLANE_NTF_USE)
151 flags |= NTF_USE;
152
153 return flags;
154 }
155
156 /*
157 * Mapping from dataplane neighbor state to netlink state
158 */
neigh_state_to_netlink(uint16_t dplane_state)159 static uint16_t neigh_state_to_netlink(uint16_t dplane_state)
160 {
161 uint16_t state = 0;
162
163 if (dplane_state & DPLANE_NUD_REACHABLE)
164 state |= NUD_REACHABLE;
165 if (dplane_state & DPLANE_NUD_STALE)
166 state |= NUD_STALE;
167 if (dplane_state & DPLANE_NUD_NOARP)
168 state |= NUD_NOARP;
169 if (dplane_state & DPLANE_NUD_PROBE)
170 state |= NUD_PROBE;
171 if (dplane_state & DPLANE_NUD_INCOMPLETE)
172 state |= NUD_INCOMPLETE;
173
174 return state;
175 }
176
177
is_selfroute(int proto)178 static inline bool is_selfroute(int proto)
179 {
180 if ((proto == RTPROT_BGP) || (proto == RTPROT_OSPF)
181 || (proto == RTPROT_ZSTATIC) || (proto == RTPROT_ZEBRA)
182 || (proto == RTPROT_ISIS) || (proto == RTPROT_RIPNG)
183 || (proto == RTPROT_NHRP) || (proto == RTPROT_EIGRP)
184 || (proto == RTPROT_LDP) || (proto == RTPROT_BABEL)
185 || (proto == RTPROT_RIP) || (proto == RTPROT_SHARP)
186 || (proto == RTPROT_PBR) || (proto == RTPROT_OPENFABRIC)
187 || (proto == RTPROT_SRTE)) {
188 return true;
189 }
190
191 return false;
192 }
193
zebra2proto(int proto)194 static inline int zebra2proto(int proto)
195 {
196 switch (proto) {
197 case ZEBRA_ROUTE_BABEL:
198 proto = RTPROT_BABEL;
199 break;
200 case ZEBRA_ROUTE_BGP:
201 proto = RTPROT_BGP;
202 break;
203 case ZEBRA_ROUTE_OSPF:
204 case ZEBRA_ROUTE_OSPF6:
205 proto = RTPROT_OSPF;
206 break;
207 case ZEBRA_ROUTE_STATIC:
208 proto = RTPROT_ZSTATIC;
209 break;
210 case ZEBRA_ROUTE_ISIS:
211 proto = RTPROT_ISIS;
212 break;
213 case ZEBRA_ROUTE_RIP:
214 proto = RTPROT_RIP;
215 break;
216 case ZEBRA_ROUTE_RIPNG:
217 proto = RTPROT_RIPNG;
218 break;
219 case ZEBRA_ROUTE_NHRP:
220 proto = RTPROT_NHRP;
221 break;
222 case ZEBRA_ROUTE_EIGRP:
223 proto = RTPROT_EIGRP;
224 break;
225 case ZEBRA_ROUTE_LDP:
226 proto = RTPROT_LDP;
227 break;
228 case ZEBRA_ROUTE_SHARP:
229 proto = RTPROT_SHARP;
230 break;
231 case ZEBRA_ROUTE_PBR:
232 proto = RTPROT_PBR;
233 break;
234 case ZEBRA_ROUTE_OPENFABRIC:
235 proto = RTPROT_OPENFABRIC;
236 break;
237 case ZEBRA_ROUTE_SRTE:
238 proto = RTPROT_SRTE;
239 break;
240 case ZEBRA_ROUTE_TABLE:
241 case ZEBRA_ROUTE_NHG:
242 proto = RTPROT_ZEBRA;
243 break;
244 default:
245 /*
246 * When a user adds a new protocol this will show up
247 * to let them know to do something about it. This
248 * is intentionally a warn because we should see
249 * this as part of development of a new protocol
250 */
251 zlog_debug(
252 "%s: Please add this protocol(%d) to proper rt_netlink.c handling",
253 __func__, proto);
254 proto = RTPROT_ZEBRA;
255 break;
256 }
257
258 return proto;
259 }
260
proto2zebra(int proto,int family,bool is_nexthop)261 static inline int proto2zebra(int proto, int family, bool is_nexthop)
262 {
263 switch (proto) {
264 case RTPROT_BABEL:
265 proto = ZEBRA_ROUTE_BABEL;
266 break;
267 case RTPROT_BGP:
268 proto = ZEBRA_ROUTE_BGP;
269 break;
270 case RTPROT_OSPF:
271 proto = (family == AF_INET) ? ZEBRA_ROUTE_OSPF
272 : ZEBRA_ROUTE_OSPF6;
273 break;
274 case RTPROT_ISIS:
275 proto = ZEBRA_ROUTE_ISIS;
276 break;
277 case RTPROT_RIP:
278 proto = ZEBRA_ROUTE_RIP;
279 break;
280 case RTPROT_RIPNG:
281 proto = ZEBRA_ROUTE_RIPNG;
282 break;
283 case RTPROT_NHRP:
284 proto = ZEBRA_ROUTE_NHRP;
285 break;
286 case RTPROT_EIGRP:
287 proto = ZEBRA_ROUTE_EIGRP;
288 break;
289 case RTPROT_LDP:
290 proto = ZEBRA_ROUTE_LDP;
291 break;
292 case RTPROT_STATIC:
293 case RTPROT_ZSTATIC:
294 proto = ZEBRA_ROUTE_STATIC;
295 break;
296 case RTPROT_SHARP:
297 proto = ZEBRA_ROUTE_SHARP;
298 break;
299 case RTPROT_PBR:
300 proto = ZEBRA_ROUTE_PBR;
301 break;
302 case RTPROT_OPENFABRIC:
303 proto = ZEBRA_ROUTE_OPENFABRIC;
304 break;
305 case RTPROT_SRTE:
306 proto = ZEBRA_ROUTE_SRTE;
307 break;
308 case RTPROT_ZEBRA:
309 if (is_nexthop) {
310 proto = ZEBRA_ROUTE_NHG;
311 break;
312 }
313 /* Intentional fall thru */
314 default:
315 /*
316 * When a user adds a new protocol this will show up
317 * to let them know to do something about it. This
318 * is intentionally a warn because we should see
319 * this as part of development of a new protocol
320 */
321 zlog_debug(
322 "%s: Please add this protocol(%d) to proper rt_netlink.c handling",
323 __func__, proto);
324 proto = ZEBRA_ROUTE_KERNEL;
325 break;
326 }
327 return proto;
328 }
329
330 /*
331 Pending: create an efficient table_id (in a tree/hash) based lookup)
332 */
vrf_lookup_by_table(uint32_t table_id,ns_id_t ns_id)333 vrf_id_t vrf_lookup_by_table(uint32_t table_id, ns_id_t ns_id)
334 {
335 struct vrf *vrf;
336 struct zebra_vrf *zvrf;
337
338 RB_FOREACH (vrf, vrf_id_head, &vrfs_by_id) {
339 zvrf = vrf->info;
340 if (zvrf == NULL)
341 continue;
342 /* case vrf with netns : match the netnsid */
343 if (vrf_is_backend_netns()) {
344 if (ns_id == zvrf_id(zvrf))
345 return zvrf_id(zvrf);
346 } else {
347 /* VRF is VRF_BACKEND_VRF_LITE */
348 if (zvrf->table_id != table_id)
349 continue;
350 return zvrf_id(zvrf);
351 }
352 }
353
354 return VRF_DEFAULT;
355 }
356
357 /**
358 * @parse_encap_mpls() - Parses encapsulated mpls attributes
359 * @tb: Pointer to rtattr to look for nested items in.
360 * @labels: Pointer to store labels in.
361 *
362 * Return: Number of mpls labels found.
363 */
parse_encap_mpls(struct rtattr * tb,mpls_label_t * labels)364 static int parse_encap_mpls(struct rtattr *tb, mpls_label_t *labels)
365 {
366 struct rtattr *tb_encap[MPLS_IPTUNNEL_MAX + 1] = {0};
367 mpls_lse_t *lses = NULL;
368 int num_labels = 0;
369 uint32_t ttl = 0;
370 uint32_t bos = 0;
371 uint32_t exp = 0;
372 mpls_label_t label = 0;
373
374 netlink_parse_rtattr_nested(tb_encap, MPLS_IPTUNNEL_MAX, tb);
375 lses = (mpls_lse_t *)RTA_DATA(tb_encap[MPLS_IPTUNNEL_DST]);
376 while (!bos && num_labels < MPLS_MAX_LABELS) {
377 mpls_lse_decode(lses[num_labels], &label, &ttl, &exp, &bos);
378 labels[num_labels++] = label;
379 }
380
381 return num_labels;
382 }
383
384 static struct nexthop
parse_nexthop_unicast(ns_id_t ns_id,struct rtmsg * rtm,struct rtattr ** tb,enum blackhole_type bh_type,int index,void * prefsrc,void * gate,afi_t afi,vrf_id_t vrf_id)385 parse_nexthop_unicast(ns_id_t ns_id, struct rtmsg *rtm, struct rtattr **tb,
386 enum blackhole_type bh_type, int index, void *prefsrc,
387 void *gate, afi_t afi, vrf_id_t vrf_id)
388 {
389 struct interface *ifp = NULL;
390 struct nexthop nh = {0};
391 mpls_label_t labels[MPLS_MAX_LABELS] = {0};
392 int num_labels = 0;
393
394 vrf_id_t nh_vrf_id = vrf_id;
395 size_t sz = (afi == AFI_IP) ? 4 : 16;
396
397 if (bh_type == BLACKHOLE_UNSPEC) {
398 if (index && !gate)
399 nh.type = NEXTHOP_TYPE_IFINDEX;
400 else if (index && gate)
401 nh.type = (afi == AFI_IP) ? NEXTHOP_TYPE_IPV4_IFINDEX
402 : NEXTHOP_TYPE_IPV6_IFINDEX;
403 else if (!index && gate)
404 nh.type = (afi == AFI_IP) ? NEXTHOP_TYPE_IPV4
405 : NEXTHOP_TYPE_IPV6;
406 else {
407 nh.type = NEXTHOP_TYPE_BLACKHOLE;
408 nh.bh_type = bh_type;
409 }
410 } else {
411 nh.type = NEXTHOP_TYPE_BLACKHOLE;
412 nh.bh_type = bh_type;
413 }
414 nh.ifindex = index;
415 if (prefsrc)
416 memcpy(&nh.src, prefsrc, sz);
417 if (gate)
418 memcpy(&nh.gate, gate, sz);
419
420 if (index) {
421 ifp = if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id), index);
422 if (ifp)
423 nh_vrf_id = ifp->vrf_id;
424 }
425 nh.vrf_id = nh_vrf_id;
426
427 if (tb[RTA_ENCAP] && tb[RTA_ENCAP_TYPE]
428 && *(uint16_t *)RTA_DATA(tb[RTA_ENCAP_TYPE])
429 == LWTUNNEL_ENCAP_MPLS) {
430 num_labels = parse_encap_mpls(tb[RTA_ENCAP], labels);
431 }
432
433 if (rtm->rtm_flags & RTNH_F_ONLINK)
434 SET_FLAG(nh.flags, NEXTHOP_FLAG_ONLINK);
435
436 if (num_labels)
437 nexthop_add_labels(&nh, ZEBRA_LSP_STATIC, num_labels, labels);
438
439 return nh;
440 }
441
parse_multipath_nexthops_unicast(ns_id_t ns_id,struct nexthop_group * ng,struct rtmsg * rtm,struct rtnexthop * rtnh,struct rtattr ** tb,void * prefsrc,vrf_id_t vrf_id)442 static uint8_t parse_multipath_nexthops_unicast(ns_id_t ns_id,
443 struct nexthop_group *ng,
444 struct rtmsg *rtm,
445 struct rtnexthop *rtnh,
446 struct rtattr **tb,
447 void *prefsrc, vrf_id_t vrf_id)
448 {
449 void *gate = NULL;
450 struct interface *ifp = NULL;
451 int index = 0;
452 /* MPLS labels */
453 mpls_label_t labels[MPLS_MAX_LABELS] = {0};
454 int num_labels = 0;
455 struct rtattr *rtnh_tb[RTA_MAX + 1] = {};
456
457 int len = RTA_PAYLOAD(tb[RTA_MULTIPATH]);
458 vrf_id_t nh_vrf_id = vrf_id;
459
460 for (;;) {
461 struct nexthop *nh = NULL;
462
463 if (len < (int)sizeof(*rtnh) || rtnh->rtnh_len > len)
464 break;
465
466 index = rtnh->rtnh_ifindex;
467 if (index) {
468 /*
469 * Yes we are looking this up
470 * for every nexthop and just
471 * using the last one looked
472 * up right now
473 */
474 ifp = if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id),
475 index);
476 if (ifp)
477 nh_vrf_id = ifp->vrf_id;
478 else {
479 flog_warn(
480 EC_ZEBRA_UNKNOWN_INTERFACE,
481 "%s: Unknown interface %u specified, defaulting to VRF_DEFAULT",
482 __func__, index);
483 nh_vrf_id = VRF_DEFAULT;
484 }
485 } else
486 nh_vrf_id = vrf_id;
487
488 if (rtnh->rtnh_len > sizeof(*rtnh)) {
489 memset(rtnh_tb, 0, sizeof(rtnh_tb));
490
491 netlink_parse_rtattr(rtnh_tb, RTA_MAX, RTNH_DATA(rtnh),
492 rtnh->rtnh_len - sizeof(*rtnh));
493 if (rtnh_tb[RTA_GATEWAY])
494 gate = RTA_DATA(rtnh_tb[RTA_GATEWAY]);
495 if (rtnh_tb[RTA_ENCAP] && rtnh_tb[RTA_ENCAP_TYPE]
496 && *(uint16_t *)RTA_DATA(rtnh_tb[RTA_ENCAP_TYPE])
497 == LWTUNNEL_ENCAP_MPLS) {
498 num_labels = parse_encap_mpls(
499 rtnh_tb[RTA_ENCAP], labels);
500 }
501 }
502
503 if (gate && rtm->rtm_family == AF_INET) {
504 if (index)
505 nh = nexthop_from_ipv4_ifindex(
506 gate, prefsrc, index, nh_vrf_id);
507 else
508 nh = nexthop_from_ipv4(gate, prefsrc,
509 nh_vrf_id);
510 } else if (gate && rtm->rtm_family == AF_INET6) {
511 if (index)
512 nh = nexthop_from_ipv6_ifindex(
513 gate, index, nh_vrf_id);
514 else
515 nh = nexthop_from_ipv6(gate, nh_vrf_id);
516 } else
517 nh = nexthop_from_ifindex(index, nh_vrf_id);
518
519 if (nh) {
520 nh->weight = rtnh->rtnh_hops + 1;
521
522 if (num_labels)
523 nexthop_add_labels(nh, ZEBRA_LSP_STATIC,
524 num_labels, labels);
525
526 if (rtnh->rtnh_flags & RTNH_F_ONLINK)
527 SET_FLAG(nh->flags, NEXTHOP_FLAG_ONLINK);
528
529 /* Add to temporary list */
530 nexthop_group_add_sorted(ng, nh);
531 }
532
533 if (rtnh->rtnh_len == 0)
534 break;
535
536 len -= NLMSG_ALIGN(rtnh->rtnh_len);
537 rtnh = RTNH_NEXT(rtnh);
538 }
539
540 uint8_t nhop_num = nexthop_group_nexthop_num(ng);
541
542 return nhop_num;
543 }
544
545 /* Looking up routing table by netlink interface. */
netlink_route_change_read_unicast(struct nlmsghdr * h,ns_id_t ns_id,int startup)546 static int netlink_route_change_read_unicast(struct nlmsghdr *h, ns_id_t ns_id,
547 int startup)
548 {
549 int len;
550 struct rtmsg *rtm;
551 struct rtattr *tb[RTA_MAX + 1];
552 uint8_t flags = 0;
553 struct prefix p;
554 struct prefix_ipv6 src_p = {};
555 vrf_id_t vrf_id;
556 bool selfroute;
557
558 char anyaddr[16] = {0};
559
560 int proto = ZEBRA_ROUTE_KERNEL;
561 int index = 0;
562 int table;
563 int metric = 0;
564 uint32_t mtu = 0;
565 uint8_t distance = 0;
566 route_tag_t tag = 0;
567 uint32_t nhe_id = 0;
568
569 void *dest = NULL;
570 void *gate = NULL;
571 void *prefsrc = NULL; /* IPv4 preferred source host address */
572 void *src = NULL; /* IPv6 srcdest source prefix */
573 enum blackhole_type bh_type = BLACKHOLE_UNSPEC;
574
575 rtm = NLMSG_DATA(h);
576
577 if (startup && h->nlmsg_type != RTM_NEWROUTE)
578 return 0;
579 switch (rtm->rtm_type) {
580 case RTN_UNICAST:
581 break;
582 case RTN_BLACKHOLE:
583 bh_type = BLACKHOLE_NULL;
584 break;
585 case RTN_UNREACHABLE:
586 bh_type = BLACKHOLE_REJECT;
587 break;
588 case RTN_PROHIBIT:
589 bh_type = BLACKHOLE_ADMINPROHIB;
590 break;
591 default:
592 if (IS_ZEBRA_DEBUG_KERNEL)
593 zlog_debug("Route rtm_type: %s(%d) intentionally ignoring",
594 nl_rttype_to_str(rtm->rtm_type),
595 rtm->rtm_type);
596 return 0;
597 }
598
599 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct rtmsg));
600 if (len < 0) {
601 zlog_err(
602 "%s: Message received from netlink is of a broken size %d %zu",
603 __func__, h->nlmsg_len,
604 (size_t)NLMSG_LENGTH(sizeof(struct rtmsg)));
605 return -1;
606 }
607
608 memset(tb, 0, sizeof(tb));
609 netlink_parse_rtattr(tb, RTA_MAX, RTM_RTA(rtm), len);
610
611 if (rtm->rtm_flags & RTM_F_CLONED)
612 return 0;
613 if (rtm->rtm_protocol == RTPROT_REDIRECT)
614 return 0;
615 if (rtm->rtm_protocol == RTPROT_KERNEL)
616 return 0;
617
618 selfroute = is_selfroute(rtm->rtm_protocol);
619
620 if (!startup && selfroute && h->nlmsg_type == RTM_NEWROUTE) {
621 if (IS_ZEBRA_DEBUG_KERNEL)
622 zlog_debug("Route type: %d Received that we think we have originated, ignoring",
623 rtm->rtm_protocol);
624 return 0;
625 }
626
627 /* We don't care about change notifications for the MPLS table. */
628 /* TODO: Revisit this. */
629 if (rtm->rtm_family == AF_MPLS)
630 return 0;
631
632 /* Table corresponding to route. */
633 if (tb[RTA_TABLE])
634 table = *(int *)RTA_DATA(tb[RTA_TABLE]);
635 else
636 table = rtm->rtm_table;
637
638 /* Map to VRF */
639 vrf_id = vrf_lookup_by_table(table, ns_id);
640 if (vrf_id == VRF_DEFAULT) {
641 if (!is_zebra_valid_kernel_table(table)
642 && !is_zebra_main_routing_table(table))
643 return 0;
644 }
645
646 /* Route which inserted by Zebra. */
647 if (selfroute) {
648 flags |= ZEBRA_FLAG_SELFROUTE;
649 proto = proto2zebra(rtm->rtm_protocol, rtm->rtm_family, false);
650 }
651 if (tb[RTA_OIF])
652 index = *(int *)RTA_DATA(tb[RTA_OIF]);
653
654 if (tb[RTA_DST])
655 dest = RTA_DATA(tb[RTA_DST]);
656 else
657 dest = anyaddr;
658
659 if (tb[RTA_SRC])
660 src = RTA_DATA(tb[RTA_SRC]);
661 else
662 src = anyaddr;
663
664 if (tb[RTA_PREFSRC])
665 prefsrc = RTA_DATA(tb[RTA_PREFSRC]);
666
667 if (tb[RTA_GATEWAY])
668 gate = RTA_DATA(tb[RTA_GATEWAY]);
669
670 if (tb[RTA_NH_ID])
671 nhe_id = *(uint32_t *)RTA_DATA(tb[RTA_NH_ID]);
672
673 if (tb[RTA_PRIORITY])
674 metric = *(int *)RTA_DATA(tb[RTA_PRIORITY]);
675
676 #if defined(SUPPORT_REALMS)
677 if (tb[RTA_FLOW])
678 tag = *(uint32_t *)RTA_DATA(tb[RTA_FLOW]);
679 #endif
680
681 if (tb[RTA_METRICS]) {
682 struct rtattr *mxrta[RTAX_MAX + 1];
683
684 memset(mxrta, 0, sizeof(mxrta));
685 netlink_parse_rtattr(mxrta, RTAX_MAX, RTA_DATA(tb[RTA_METRICS]),
686 RTA_PAYLOAD(tb[RTA_METRICS]));
687
688 if (mxrta[RTAX_MTU])
689 mtu = *(uint32_t *)RTA_DATA(mxrta[RTAX_MTU]);
690 }
691
692 if (rtm->rtm_family == AF_INET) {
693 p.family = AF_INET;
694 if (rtm->rtm_dst_len > IPV4_MAX_BITLEN) {
695 zlog_err(
696 "Invalid destination prefix length: %u received from kernel route change",
697 rtm->rtm_dst_len);
698 return -1;
699 }
700 memcpy(&p.u.prefix4, dest, 4);
701 p.prefixlen = rtm->rtm_dst_len;
702
703 if (rtm->rtm_src_len != 0) {
704 char buf[PREFIX_STRLEN];
705 flog_warn(
706 EC_ZEBRA_UNSUPPORTED_V4_SRCDEST,
707 "unsupported IPv4 sourcedest route (dest %s vrf %u)",
708 prefix2str(&p, buf, sizeof(buf)), vrf_id);
709 return 0;
710 }
711
712 /* Force debug below to not display anything for source */
713 src_p.prefixlen = 0;
714 } else if (rtm->rtm_family == AF_INET6) {
715 p.family = AF_INET6;
716 if (rtm->rtm_dst_len > IPV6_MAX_BITLEN) {
717 zlog_err(
718 "Invalid destination prefix length: %u received from kernel route change",
719 rtm->rtm_dst_len);
720 return -1;
721 }
722 memcpy(&p.u.prefix6, dest, 16);
723 p.prefixlen = rtm->rtm_dst_len;
724
725 src_p.family = AF_INET6;
726 if (rtm->rtm_src_len > IPV6_MAX_BITLEN) {
727 zlog_err(
728 "Invalid source prefix length: %u received from kernel route change",
729 rtm->rtm_src_len);
730 return -1;
731 }
732 memcpy(&src_p.prefix, src, 16);
733 src_p.prefixlen = rtm->rtm_src_len;
734 }
735
736 /*
737 * For ZEBRA_ROUTE_KERNEL types:
738 *
739 * The metric/priority of the route received from the kernel
740 * is a 32 bit number. We are going to interpret the high
741 * order byte as the Admin Distance and the low order 3 bytes
742 * as the metric.
743 *
744 * This will allow us to do two things:
745 * 1) Allow the creation of kernel routes that can be
746 * overridden by zebra.
747 * 2) Allow the old behavior for 'most' kernel route types
748 * if a user enters 'ip route ...' v4 routes get a metric
749 * of 0 and v6 routes get a metric of 1024. Both of these
750 * values will end up with a admin distance of 0, which
751 * will cause them to win for the purposes of zebra.
752 */
753 if (proto == ZEBRA_ROUTE_KERNEL) {
754 distance = (metric >> 24) & 0xFF;
755 metric = (metric & 0x00FFFFFF);
756 }
757
758 if (IS_ZEBRA_DEBUG_KERNEL) {
759 char buf[PREFIX_STRLEN];
760 char buf2[PREFIX_STRLEN];
761 zlog_debug(
762 "%s %s%s%s vrf %s(%u) table_id: %u metric: %d Admin Distance: %d",
763 nl_msg_type_to_str(h->nlmsg_type),
764 prefix2str(&p, buf, sizeof(buf)),
765 src_p.prefixlen ? " from " : "",
766 src_p.prefixlen ? prefix2str(&src_p, buf2, sizeof(buf2))
767 : "",
768 vrf_id_to_name(vrf_id), vrf_id, table, metric,
769 distance);
770 }
771
772 afi_t afi = AFI_IP;
773 if (rtm->rtm_family == AF_INET6)
774 afi = AFI_IP6;
775
776 if (h->nlmsg_type == RTM_NEWROUTE) {
777
778 if (!tb[RTA_MULTIPATH]) {
779 struct nexthop nh = {0};
780
781 if (!nhe_id) {
782 nh = parse_nexthop_unicast(
783 ns_id, rtm, tb, bh_type, index, prefsrc,
784 gate, afi, vrf_id);
785 }
786 rib_add(afi, SAFI_UNICAST, vrf_id, proto, 0, flags, &p,
787 &src_p, &nh, nhe_id, table, metric, mtu,
788 distance, tag);
789 } else {
790 /* This is a multipath route */
791 struct route_entry *re;
792 struct nexthop_group *ng = NULL;
793 struct rtnexthop *rtnh =
794 (struct rtnexthop *)RTA_DATA(tb[RTA_MULTIPATH]);
795
796 re = XCALLOC(MTYPE_RE, sizeof(struct route_entry));
797 re->type = proto;
798 re->distance = distance;
799 re->flags = flags;
800 re->metric = metric;
801 re->mtu = mtu;
802 re->vrf_id = vrf_id;
803 re->table = table;
804 re->uptime = monotime(NULL);
805 re->tag = tag;
806 re->nhe_id = nhe_id;
807
808 if (!nhe_id) {
809 uint8_t nhop_num;
810
811 /* Use temporary list of nexthops; parse
812 * message payload's nexthops.
813 */
814 ng = nexthop_group_new();
815 nhop_num =
816 parse_multipath_nexthops_unicast(
817 ns_id, ng, rtm, rtnh, tb,
818 prefsrc, vrf_id);
819
820 zserv_nexthop_num_warn(
821 __func__, (const struct prefix *)&p,
822 nhop_num);
823
824 if (nhop_num == 0) {
825 nexthop_group_delete(&ng);
826 ng = NULL;
827 }
828 }
829
830 if (nhe_id || ng)
831 rib_add_multipath(afi, SAFI_UNICAST, &p,
832 &src_p, re, ng);
833 else
834 XFREE(MTYPE_RE, re);
835 }
836 } else {
837 if (nhe_id) {
838 rib_delete(afi, SAFI_UNICAST, vrf_id, proto, 0, flags,
839 &p, &src_p, NULL, nhe_id, table, metric,
840 distance, true, false);
841 } else {
842 if (!tb[RTA_MULTIPATH]) {
843 struct nexthop nh;
844
845 nh = parse_nexthop_unicast(
846 ns_id, rtm, tb, bh_type, index, prefsrc,
847 gate, afi, vrf_id);
848 rib_delete(afi, SAFI_UNICAST, vrf_id, proto, 0,
849 flags, &p, &src_p, &nh, 0, table,
850 metric, distance, true, false);
851 } else {
852 /* XXX: need to compare the entire list of
853 * nexthops here for NLM_F_APPEND stupidity */
854 rib_delete(afi, SAFI_UNICAST, vrf_id, proto, 0,
855 flags, &p, &src_p, NULL, 0, table,
856 metric, distance, true, false);
857 }
858 }
859 }
860
861 return 0;
862 }
863
864 static struct mcast_route_data *mroute = NULL;
865
netlink_route_change_read_multicast(struct nlmsghdr * h,ns_id_t ns_id,int startup)866 static int netlink_route_change_read_multicast(struct nlmsghdr *h,
867 ns_id_t ns_id, int startup)
868 {
869 int len;
870 struct rtmsg *rtm;
871 struct rtattr *tb[RTA_MAX + 1];
872 struct mcast_route_data *m;
873 struct mcast_route_data mr;
874 int iif = 0;
875 int count;
876 int oif[256];
877 int oif_count = 0;
878 char sbuf[40];
879 char gbuf[40];
880 char oif_list[256] = "\0";
881 vrf_id_t vrf;
882 int table;
883
884 if (mroute)
885 m = mroute;
886 else {
887 memset(&mr, 0, sizeof(mr));
888 m = &mr;
889 }
890
891 rtm = NLMSG_DATA(h);
892
893 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct rtmsg));
894
895 memset(tb, 0, sizeof(tb));
896 netlink_parse_rtattr(tb, RTA_MAX, RTM_RTA(rtm), len);
897
898 if (tb[RTA_TABLE])
899 table = *(int *)RTA_DATA(tb[RTA_TABLE]);
900 else
901 table = rtm->rtm_table;
902
903 vrf = vrf_lookup_by_table(table, ns_id);
904
905 if (tb[RTA_IIF])
906 iif = *(int *)RTA_DATA(tb[RTA_IIF]);
907
908 if (tb[RTA_SRC])
909 m->sg.src = *(struct in_addr *)RTA_DATA(tb[RTA_SRC]);
910
911 if (tb[RTA_DST])
912 m->sg.grp = *(struct in_addr *)RTA_DATA(tb[RTA_DST]);
913
914 if (tb[RTA_EXPIRES])
915 m->lastused = *(unsigned long long *)RTA_DATA(tb[RTA_EXPIRES]);
916
917 if (tb[RTA_MULTIPATH]) {
918 struct rtnexthop *rtnh =
919 (struct rtnexthop *)RTA_DATA(tb[RTA_MULTIPATH]);
920
921 len = RTA_PAYLOAD(tb[RTA_MULTIPATH]);
922 for (;;) {
923 if (len < (int)sizeof(*rtnh) || rtnh->rtnh_len > len)
924 break;
925
926 oif[oif_count] = rtnh->rtnh_ifindex;
927 oif_count++;
928
929 if (rtnh->rtnh_len == 0)
930 break;
931
932 len -= NLMSG_ALIGN(rtnh->rtnh_len);
933 rtnh = RTNH_NEXT(rtnh);
934 }
935 }
936
937 if (IS_ZEBRA_DEBUG_KERNEL) {
938 struct interface *ifp = NULL;
939 struct zebra_vrf *zvrf = NULL;
940
941 strlcpy(sbuf, inet_ntoa(m->sg.src), sizeof(sbuf));
942 strlcpy(gbuf, inet_ntoa(m->sg.grp), sizeof(gbuf));
943 for (count = 0; count < oif_count; count++) {
944 ifp = if_lookup_by_index(oif[count], vrf);
945 char temp[256];
946
947 snprintf(temp, sizeof(temp), "%s(%d) ",
948 ifp ? ifp->name : "Unknown", oif[count]);
949 strlcat(oif_list, temp, sizeof(oif_list));
950 }
951 zvrf = zebra_vrf_lookup_by_id(vrf);
952 ifp = if_lookup_by_index(iif, vrf);
953 zlog_debug(
954 "MCAST VRF: %s(%d) %s (%s,%s) IIF: %s(%d) OIF: %s jiffies: %lld",
955 zvrf_name(zvrf), vrf, nl_msg_type_to_str(h->nlmsg_type),
956 sbuf, gbuf, ifp ? ifp->name : "Unknown", iif, oif_list,
957 m->lastused);
958 }
959 return 0;
960 }
961
netlink_route_change(struct nlmsghdr * h,ns_id_t ns_id,int startup)962 int netlink_route_change(struct nlmsghdr *h, ns_id_t ns_id, int startup)
963 {
964 int len;
965 struct rtmsg *rtm;
966
967 rtm = NLMSG_DATA(h);
968
969 if (!(h->nlmsg_type == RTM_NEWROUTE || h->nlmsg_type == RTM_DELROUTE)) {
970 /* If this is not route add/delete message print warning. */
971 zlog_debug("Kernel message: %s NS %u",
972 nl_msg_type_to_str(h->nlmsg_type), ns_id);
973 return 0;
974 }
975
976 if (!(rtm->rtm_family == AF_INET ||
977 rtm->rtm_family == AF_INET6 ||
978 rtm->rtm_family == RTNL_FAMILY_IPMR )) {
979 flog_warn(
980 EC_ZEBRA_UNKNOWN_FAMILY,
981 "Invalid address family: %u received from kernel route change: %s",
982 rtm->rtm_family, nl_msg_type_to_str(h->nlmsg_type));
983 return 0;
984 }
985
986 /* Connected route. */
987 if (IS_ZEBRA_DEBUG_KERNEL)
988 zlog_debug("%s %s %s proto %s NS %u",
989 nl_msg_type_to_str(h->nlmsg_type),
990 nl_family_to_str(rtm->rtm_family),
991 nl_rttype_to_str(rtm->rtm_type),
992 nl_rtproto_to_str(rtm->rtm_protocol), ns_id);
993
994
995 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct rtmsg));
996 if (len < 0) {
997 zlog_err(
998 "%s: Message received from netlink is of a broken size: %d %zu",
999 __func__, h->nlmsg_len,
1000 (size_t)NLMSG_LENGTH(sizeof(struct rtmsg)));
1001 return -1;
1002 }
1003
1004 if (rtm->rtm_type == RTN_MULTICAST)
1005 netlink_route_change_read_multicast(h, ns_id, startup);
1006 else
1007 netlink_route_change_read_unicast(h, ns_id, startup);
1008 return 0;
1009 }
1010
1011 /* Request for specific route information from the kernel */
netlink_request_route(struct zebra_ns * zns,int family,int type)1012 static int netlink_request_route(struct zebra_ns *zns, int family, int type)
1013 {
1014 struct {
1015 struct nlmsghdr n;
1016 struct rtmsg rtm;
1017 } req;
1018
1019 /* Form the request, specifying filter (rtattr) if needed. */
1020 memset(&req, 0, sizeof(req));
1021 req.n.nlmsg_type = type;
1022 req.n.nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST;
1023 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
1024 req.rtm.rtm_family = family;
1025
1026 return netlink_request(&zns->netlink_cmd, &req);
1027 }
1028
1029 /* Routing table read function using netlink interface. Only called
1030 bootstrap time. */
netlink_route_read(struct zebra_ns * zns)1031 int netlink_route_read(struct zebra_ns *zns)
1032 {
1033 int ret;
1034 struct zebra_dplane_info dp_info;
1035
1036 zebra_dplane_info_from_zns(&dp_info, zns, true /*is_cmd*/);
1037
1038 /* Get IPv4 routing table. */
1039 ret = netlink_request_route(zns, AF_INET, RTM_GETROUTE);
1040 if (ret < 0)
1041 return ret;
1042 ret = netlink_parse_info(netlink_route_change_read_unicast,
1043 &zns->netlink_cmd, &dp_info, 0, 1);
1044 if (ret < 0)
1045 return ret;
1046
1047 /* Get IPv6 routing table. */
1048 ret = netlink_request_route(zns, AF_INET6, RTM_GETROUTE);
1049 if (ret < 0)
1050 return ret;
1051 ret = netlink_parse_info(netlink_route_change_read_unicast,
1052 &zns->netlink_cmd, &dp_info, 0, 1);
1053 if (ret < 0)
1054 return ret;
1055
1056 return 0;
1057 }
1058
1059 /*
1060 * The function returns true if the gateway info could be added
1061 * to the message, otherwise false is returned.
1062 */
_netlink_route_add_gateway_info(uint8_t route_family,uint8_t gw_family,struct nlmsghdr * nlmsg,size_t req_size,int bytelen,const struct nexthop * nexthop)1063 static bool _netlink_route_add_gateway_info(uint8_t route_family,
1064 uint8_t gw_family,
1065 struct nlmsghdr *nlmsg,
1066 size_t req_size, int bytelen,
1067 const struct nexthop *nexthop)
1068 {
1069 if (route_family == AF_MPLS) {
1070 struct gw_family_t gw_fam;
1071
1072 gw_fam.family = gw_family;
1073 if (gw_family == AF_INET)
1074 memcpy(&gw_fam.gate.ipv4, &nexthop->gate.ipv4, bytelen);
1075 else
1076 memcpy(&gw_fam.gate.ipv6, &nexthop->gate.ipv6, bytelen);
1077 if (!nl_attr_put(nlmsg, req_size, RTA_VIA, &gw_fam.family,
1078 bytelen + 2))
1079 return false;
1080 } else {
1081 if (!(nexthop->rparent
1082 && IS_MAPPED_IPV6(&nexthop->rparent->gate.ipv6))) {
1083 if (gw_family == AF_INET) {
1084 if (!nl_attr_put(nlmsg, req_size, RTA_GATEWAY,
1085 &nexthop->gate.ipv4, bytelen))
1086 return false;
1087 } else {
1088 if (!nl_attr_put(nlmsg, req_size, RTA_GATEWAY,
1089 &nexthop->gate.ipv6, bytelen))
1090 return false;
1091 }
1092 }
1093 }
1094
1095 return true;
1096 }
1097
build_label_stack(struct mpls_label_stack * nh_label,mpls_lse_t * out_lse,char * label_buf,size_t label_buf_size)1098 static int build_label_stack(struct mpls_label_stack *nh_label,
1099 mpls_lse_t *out_lse, char *label_buf,
1100 size_t label_buf_size)
1101 {
1102 char label_buf1[20];
1103 int num_labels = 0;
1104
1105 for (int i = 0; nh_label && i < nh_label->num_labels; i++) {
1106 if (nh_label->label[i] == MPLS_LABEL_IMPLICIT_NULL)
1107 continue;
1108
1109 if (IS_ZEBRA_DEBUG_KERNEL) {
1110 if (!num_labels)
1111 sprintf(label_buf, "label %u",
1112 nh_label->label[i]);
1113 else {
1114 snprintf(label_buf1, sizeof(label_buf1), "/%u",
1115 nh_label->label[i]);
1116 strlcat(label_buf, label_buf1, label_buf_size);
1117 }
1118 }
1119
1120 out_lse[num_labels] =
1121 mpls_lse_encode(nh_label->label[i], 0, 0, 0);
1122 num_labels++;
1123 }
1124
1125 return num_labels;
1126 }
1127
_netlink_route_encode_label_info(struct mpls_label_stack * nh_label,struct nlmsghdr * nlmsg,size_t buflen,struct rtmsg * rtmsg,char * label_buf,size_t label_buf_size)1128 static bool _netlink_route_encode_label_info(struct mpls_label_stack *nh_label,
1129 struct nlmsghdr *nlmsg,
1130 size_t buflen, struct rtmsg *rtmsg,
1131 char *label_buf,
1132 size_t label_buf_size)
1133 {
1134 mpls_lse_t out_lse[MPLS_MAX_LABELS];
1135 int num_labels;
1136
1137 /*
1138 * label_buf is *only* currently used within debugging.
1139 * As such when we assign it we are guarding it inside
1140 * a debug test. If you want to change this make sure
1141 * you fix this assumption
1142 */
1143 label_buf[0] = '\0';
1144
1145 num_labels =
1146 build_label_stack(nh_label, out_lse, label_buf, label_buf_size);
1147
1148 if (num_labels) {
1149 /* Set the BoS bit */
1150 out_lse[num_labels - 1] |= htonl(1 << MPLS_LS_S_SHIFT);
1151
1152 if (rtmsg->rtm_family == AF_MPLS) {
1153 if (!nl_attr_put(nlmsg, buflen, RTA_NEWDST, &out_lse,
1154 num_labels * sizeof(mpls_lse_t)))
1155 return false;
1156 } else {
1157 struct rtattr *nest;
1158
1159 if (!nl_attr_put16(nlmsg, buflen, RTA_ENCAP_TYPE,
1160 LWTUNNEL_ENCAP_MPLS))
1161 return false;
1162
1163 nest = nl_attr_nest(nlmsg, buflen, RTA_ENCAP);
1164 if (!nest)
1165 return false;
1166
1167 if (!nl_attr_put(nlmsg, buflen, MPLS_IPTUNNEL_DST,
1168 &out_lse,
1169 num_labels * sizeof(mpls_lse_t)))
1170 return false;
1171 nl_attr_nest_end(nlmsg, nest);
1172 }
1173 }
1174
1175 return true;
1176 }
1177
_netlink_route_encode_nexthop_src(const struct nexthop * nexthop,int family,struct nlmsghdr * nlmsg,size_t buflen,int bytelen)1178 static bool _netlink_route_encode_nexthop_src(const struct nexthop *nexthop,
1179 int family,
1180 struct nlmsghdr *nlmsg,
1181 size_t buflen, int bytelen)
1182 {
1183 if (family == AF_INET) {
1184 if (nexthop->rmap_src.ipv4.s_addr != INADDR_ANY) {
1185 if (!nl_attr_put(nlmsg, buflen, RTA_PREFSRC,
1186 &nexthop->rmap_src.ipv4, bytelen))
1187 return false;
1188 } else if (nexthop->src.ipv4.s_addr != INADDR_ANY) {
1189 if (!nl_attr_put(nlmsg, buflen, RTA_PREFSRC,
1190 &nexthop->src.ipv4, bytelen))
1191 return false;
1192 }
1193 } else if (family == AF_INET6) {
1194 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->rmap_src.ipv6)) {
1195 if (!nl_attr_put(nlmsg, buflen, RTA_PREFSRC,
1196 &nexthop->rmap_src.ipv6, bytelen))
1197 return false;
1198 } else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->src.ipv6)) {
1199 if (!nl_attr_put(nlmsg, buflen, RTA_PREFSRC,
1200 &nexthop->src.ipv6, bytelen))
1201 return false;
1202 }
1203 }
1204
1205 return true;
1206 }
1207
1208 /* This function takes a nexthop as argument and adds
1209 * the appropriate netlink attributes to an existing
1210 * netlink message.
1211 *
1212 * @param routedesc: Human readable description of route type
1213 * (direct/recursive, single-/multipath)
1214 * @param bytelen: Length of addresses in bytes.
1215 * @param nexthop: Nexthop information
1216 * @param nlmsg: nlmsghdr structure to fill in.
1217 * @param req_size: The size allocated for the message.
1218 *
1219 * The function returns true if the nexthop could be added
1220 * to the message, otherwise false is returned.
1221 */
_netlink_route_build_singlepath(const struct prefix * p,const char * routedesc,int bytelen,const struct nexthop * nexthop,struct nlmsghdr * nlmsg,struct rtmsg * rtmsg,size_t req_size,int cmd)1222 static bool _netlink_route_build_singlepath(const struct prefix *p,
1223 const char *routedesc, int bytelen,
1224 const struct nexthop *nexthop,
1225 struct nlmsghdr *nlmsg,
1226 struct rtmsg *rtmsg,
1227 size_t req_size, int cmd)
1228 {
1229
1230 char label_buf[256];
1231 struct vrf *vrf;
1232 char addrstr[INET6_ADDRSTRLEN];
1233
1234 assert(nexthop);
1235
1236 vrf = vrf_lookup_by_id(nexthop->vrf_id);
1237
1238 if (!_netlink_route_encode_label_info(nexthop->nh_label, nlmsg,
1239 req_size, rtmsg, label_buf,
1240 sizeof(label_buf)))
1241 return false;
1242
1243 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK))
1244 rtmsg->rtm_flags |= RTNH_F_ONLINK;
1245
1246 if (is_route_v4_over_v6(rtmsg->rtm_family, nexthop->type)) {
1247 rtmsg->rtm_flags |= RTNH_F_ONLINK;
1248 if (!nl_attr_put(nlmsg, req_size, RTA_GATEWAY, &ipv4_ll, 4))
1249 return false;
1250 if (!nl_attr_put32(nlmsg, req_size, RTA_OIF, nexthop->ifindex))
1251 return false;
1252
1253 if (cmd == RTM_NEWROUTE) {
1254 if (!_netlink_route_encode_nexthop_src(
1255 nexthop, AF_INET, nlmsg, req_size, bytelen))
1256 return false;
1257 }
1258
1259 if (IS_ZEBRA_DEBUG_KERNEL)
1260 zlog_debug("%s: 5549 (%s): %pFX nexthop via %s %s if %u vrf %s(%u)",
1261 __func__, routedesc, p, ipv4_ll_buf,
1262 label_buf, nexthop->ifindex,
1263 VRF_LOGNAME(vrf), nexthop->vrf_id);
1264 return true;
1265 }
1266
1267 if (nexthop->type == NEXTHOP_TYPE_IPV4
1268 || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX) {
1269 /* Send deletes to the kernel without specifying the next-hop */
1270 if (cmd != RTM_DELROUTE) {
1271 if (!_netlink_route_add_gateway_info(
1272 rtmsg->rtm_family, AF_INET, nlmsg, req_size,
1273 bytelen, nexthop))
1274 return false;
1275 }
1276
1277 if (cmd == RTM_NEWROUTE) {
1278 if (!_netlink_route_encode_nexthop_src(
1279 nexthop, AF_INET, nlmsg, req_size, bytelen))
1280 return false;
1281 }
1282
1283 if (IS_ZEBRA_DEBUG_KERNEL) {
1284 inet_ntop(AF_INET, &nexthop->gate.ipv4, addrstr,
1285 sizeof(addrstr));
1286 zlog_debug("%s: (%s): %pFX nexthop via %s %s if %u vrf %s(%u)",
1287 __func__, routedesc, p, addrstr, label_buf,
1288 nexthop->ifindex, VRF_LOGNAME(vrf),
1289 nexthop->vrf_id);
1290 }
1291 }
1292
1293 if (nexthop->type == NEXTHOP_TYPE_IPV6
1294 || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX) {
1295 if (!_netlink_route_add_gateway_info(rtmsg->rtm_family,
1296 AF_INET6, nlmsg, req_size,
1297 bytelen, nexthop))
1298 return false;
1299
1300 if (cmd == RTM_NEWROUTE) {
1301 if (!_netlink_route_encode_nexthop_src(
1302 nexthop, AF_INET6, nlmsg, req_size,
1303 bytelen))
1304 return false;
1305 }
1306
1307 if (IS_ZEBRA_DEBUG_KERNEL) {
1308 inet_ntop(AF_INET6, &nexthop->gate.ipv6, addrstr,
1309 sizeof(addrstr));
1310 zlog_debug("%s: (%s): %pFX nexthop via %s %s if %u vrf %s(%u)",
1311 __func__, routedesc, p, addrstr, label_buf,
1312 nexthop->ifindex, VRF_LOGNAME(vrf),
1313 nexthop->vrf_id);
1314 }
1315 }
1316
1317 /*
1318 * We have the ifindex so we should always send it
1319 * This is especially useful if we are doing route
1320 * leaking.
1321 */
1322 if (nexthop->type != NEXTHOP_TYPE_BLACKHOLE) {
1323 if (!nl_attr_put32(nlmsg, req_size, RTA_OIF, nexthop->ifindex))
1324 return false;
1325 }
1326
1327 if (nexthop->type == NEXTHOP_TYPE_IFINDEX) {
1328 if (cmd == RTM_NEWROUTE) {
1329 if (!_netlink_route_encode_nexthop_src(
1330 nexthop, AF_INET, nlmsg, req_size, bytelen))
1331 return false;
1332 }
1333
1334 if (IS_ZEBRA_DEBUG_KERNEL)
1335 zlog_debug("%s: (%s): %pFX nexthop via if %u vrf %s(%u)",
1336 __func__, routedesc, p, nexthop->ifindex,
1337 VRF_LOGNAME(vrf), nexthop->vrf_id);
1338 }
1339
1340 return true;
1341 }
1342
1343 /* This function takes a nexthop as argument and
1344 * appends to the given netlink msg. If the nexthop
1345 * defines a preferred source, the src parameter
1346 * will be modified to point to that src, otherwise
1347 * it will be kept unmodified.
1348 *
1349 * @param routedesc: Human readable description of route type
1350 * (direct/recursive, single-/multipath)
1351 * @param bytelen: Length of addresses in bytes.
1352 * @param nexthop: Nexthop information
1353 * @param nlmsg: nlmsghdr structure to fill in.
1354 * @param req_size: The size allocated for the message.
1355 * @param src: pointer pointing to a location where
1356 * the prefsrc should be stored.
1357 *
1358 * The function returns true if the nexthop could be added
1359 * to the message, otherwise false is returned.
1360 */
_netlink_route_build_multipath(const struct prefix * p,const char * routedesc,int bytelen,const struct nexthop * nexthop,struct nlmsghdr * nlmsg,size_t req_size,struct rtmsg * rtmsg,const union g_addr ** src)1361 static bool _netlink_route_build_multipath(const struct prefix *p,
1362 const char *routedesc, int bytelen,
1363 const struct nexthop *nexthop,
1364 struct nlmsghdr *nlmsg,
1365 size_t req_size, struct rtmsg *rtmsg,
1366 const union g_addr **src)
1367 {
1368 char label_buf[256];
1369 struct vrf *vrf;
1370 struct rtnexthop *rtnh;
1371
1372 rtnh = nl_attr_rtnh(nlmsg, req_size);
1373 if (rtnh == NULL)
1374 return false;
1375
1376 assert(nexthop);
1377
1378 vrf = vrf_lookup_by_id(nexthop->vrf_id);
1379
1380 if (!_netlink_route_encode_label_info(nexthop->nh_label, nlmsg,
1381 req_size, rtmsg, label_buf,
1382 sizeof(label_buf)))
1383 return false;
1384
1385 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ONLINK))
1386 rtnh->rtnh_flags |= RTNH_F_ONLINK;
1387
1388 if (is_route_v4_over_v6(rtmsg->rtm_family, nexthop->type)) {
1389 rtnh->rtnh_flags |= RTNH_F_ONLINK;
1390 if (!nl_attr_put(nlmsg, req_size, RTA_GATEWAY, &ipv4_ll, 4))
1391 return false;
1392 rtnh->rtnh_ifindex = nexthop->ifindex;
1393 if (nexthop->weight)
1394 rtnh->rtnh_hops = nexthop->weight - 1;
1395
1396 if (nexthop->rmap_src.ipv4.s_addr != INADDR_ANY)
1397 *src = &nexthop->rmap_src;
1398 else if (nexthop->src.ipv4.s_addr != INADDR_ANY)
1399 *src = &nexthop->src;
1400
1401 if (IS_ZEBRA_DEBUG_KERNEL)
1402 zlog_debug(
1403 "%s: 5549 (%s): %pFX nexthop via %s %s if %u vrf %s(%u)",
1404 __func__, routedesc, p, ipv4_ll_buf, label_buf,
1405 nexthop->ifindex, VRF_LOGNAME(vrf),
1406 nexthop->vrf_id);
1407 nl_attr_rtnh_end(nlmsg, rtnh);
1408 return true;
1409 }
1410
1411 if (nexthop->type == NEXTHOP_TYPE_IPV4
1412 || nexthop->type == NEXTHOP_TYPE_IPV4_IFINDEX) {
1413 if (!_netlink_route_add_gateway_info(rtmsg->rtm_family, AF_INET,
1414 nlmsg, req_size, bytelen,
1415 nexthop))
1416 return false;
1417
1418 if (nexthop->rmap_src.ipv4.s_addr != INADDR_ANY)
1419 *src = &nexthop->rmap_src;
1420 else if (nexthop->src.ipv4.s_addr != INADDR_ANY)
1421 *src = &nexthop->src;
1422
1423 if (IS_ZEBRA_DEBUG_KERNEL)
1424 zlog_debug("%s: (%s): %pFX nexthop via %pI4 %s if %u vrf %s(%u)",
1425 __func__, routedesc, p, &nexthop->gate.ipv4,
1426 label_buf, nexthop->ifindex,
1427 VRF_LOGNAME(vrf), nexthop->vrf_id);
1428 }
1429 if (nexthop->type == NEXTHOP_TYPE_IPV6
1430 || nexthop->type == NEXTHOP_TYPE_IPV6_IFINDEX) {
1431 if (!_netlink_route_add_gateway_info(rtmsg->rtm_family,
1432 AF_INET6, nlmsg, req_size,
1433 bytelen, nexthop))
1434 return false;
1435
1436 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->rmap_src.ipv6))
1437 *src = &nexthop->rmap_src;
1438 else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->src.ipv6))
1439 *src = &nexthop->src;
1440
1441 if (IS_ZEBRA_DEBUG_KERNEL)
1442 zlog_debug("%s: (%s): %pFX nexthop via %pI6 %s if %u vrf %s(%u)",
1443 __func__, routedesc, p, &nexthop->gate.ipv6,
1444 label_buf, nexthop->ifindex,
1445 VRF_LOGNAME(vrf), nexthop->vrf_id);
1446 }
1447
1448 /*
1449 * We have figured out the ifindex so we should always send it
1450 * This is especially useful if we are doing route
1451 * leaking.
1452 */
1453 if (nexthop->type != NEXTHOP_TYPE_BLACKHOLE)
1454 rtnh->rtnh_ifindex = nexthop->ifindex;
1455
1456 /* ifindex */
1457 if (nexthop->type == NEXTHOP_TYPE_IFINDEX) {
1458 if (nexthop->rmap_src.ipv4.s_addr != INADDR_ANY)
1459 *src = &nexthop->rmap_src;
1460 else if (nexthop->src.ipv4.s_addr != INADDR_ANY)
1461 *src = &nexthop->src;
1462
1463 if (IS_ZEBRA_DEBUG_KERNEL)
1464 zlog_debug("%s: (%s): %pFX nexthop via if %u vrf %s(%u)",
1465 __func__, routedesc, p, nexthop->ifindex,
1466 VRF_LOGNAME(vrf), nexthop->vrf_id);
1467 }
1468
1469 if (nexthop->weight)
1470 rtnh->rtnh_hops = nexthop->weight - 1;
1471
1472 nl_attr_rtnh_end(nlmsg, rtnh);
1473 return true;
1474 }
1475
_netlink_mpls_build_singlepath(const struct prefix * p,const char * routedesc,const zebra_nhlfe_t * nhlfe,struct nlmsghdr * nlmsg,struct rtmsg * rtmsg,size_t req_size,int cmd)1476 static inline bool _netlink_mpls_build_singlepath(const struct prefix *p,
1477 const char *routedesc,
1478 const zebra_nhlfe_t *nhlfe,
1479 struct nlmsghdr *nlmsg,
1480 struct rtmsg *rtmsg,
1481 size_t req_size, int cmd)
1482 {
1483 int bytelen;
1484 uint8_t family;
1485
1486 family = NHLFE_FAMILY(nhlfe);
1487 bytelen = (family == AF_INET ? 4 : 16);
1488 return _netlink_route_build_singlepath(p, routedesc, bytelen,
1489 nhlfe->nexthop, nlmsg, rtmsg,
1490 req_size, cmd);
1491 }
1492
1493
1494 static inline bool
_netlink_mpls_build_multipath(const struct prefix * p,const char * routedesc,const zebra_nhlfe_t * nhlfe,struct nlmsghdr * nlmsg,size_t req_size,struct rtmsg * rtmsg,const union g_addr ** src)1495 _netlink_mpls_build_multipath(const struct prefix *p, const char *routedesc,
1496 const zebra_nhlfe_t *nhlfe,
1497 struct nlmsghdr *nlmsg, size_t req_size,
1498 struct rtmsg *rtmsg, const union g_addr **src)
1499 {
1500 int bytelen;
1501 uint8_t family;
1502
1503 family = NHLFE_FAMILY(nhlfe);
1504 bytelen = (family == AF_INET ? 4 : 16);
1505 return _netlink_route_build_multipath(p, routedesc, bytelen,
1506 nhlfe->nexthop, nlmsg, req_size,
1507 rtmsg, src);
1508 }
1509
_netlink_mpls_debug(int cmd,uint32_t label,const char * routedesc)1510 static void _netlink_mpls_debug(int cmd, uint32_t label, const char *routedesc)
1511 {
1512 if (IS_ZEBRA_DEBUG_KERNEL)
1513 zlog_debug("netlink_mpls_multipath_msg_encode() (%s): %s %u/20",
1514 routedesc, nl_msg_type_to_str(cmd), label);
1515 }
1516
netlink_neigh_update(int cmd,int ifindex,uint32_t addr,char * lla,int llalen,ns_id_t ns_id)1517 static int netlink_neigh_update(int cmd, int ifindex, uint32_t addr, char *lla,
1518 int llalen, ns_id_t ns_id)
1519 {
1520 uint8_t protocol = RTPROT_ZEBRA;
1521 struct {
1522 struct nlmsghdr n;
1523 struct ndmsg ndm;
1524 char buf[256];
1525 } req;
1526
1527 struct zebra_ns *zns = zebra_ns_lookup(ns_id);
1528
1529 memset(&req, 0, sizeof(req));
1530
1531 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
1532 req.n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
1533 req.n.nlmsg_type = cmd; // RTM_NEWNEIGH or RTM_DELNEIGH
1534 req.n.nlmsg_pid = zns->netlink_cmd.snl.nl_pid;
1535
1536 req.ndm.ndm_family = AF_INET;
1537 req.ndm.ndm_state = NUD_PERMANENT;
1538 req.ndm.ndm_ifindex = ifindex;
1539 req.ndm.ndm_type = RTN_UNICAST;
1540
1541 nl_attr_put(&req.n, sizeof(req), NDA_PROTOCOL, &protocol,
1542 sizeof(protocol));
1543 nl_attr_put32(&req.n, sizeof(req), NDA_DST, addr);
1544 nl_attr_put(&req.n, sizeof(req), NDA_LLADDR, lla, llalen);
1545
1546 return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
1547 0);
1548 }
1549
nexthop_set_src(const struct nexthop * nexthop,int family,union g_addr * src)1550 static bool nexthop_set_src(const struct nexthop *nexthop, int family,
1551 union g_addr *src)
1552 {
1553 if (family == AF_INET) {
1554 if (nexthop->rmap_src.ipv4.s_addr != INADDR_ANY) {
1555 src->ipv4 = nexthop->rmap_src.ipv4;
1556 return true;
1557 } else if (nexthop->src.ipv4.s_addr != INADDR_ANY) {
1558 src->ipv4 = nexthop->src.ipv4;
1559 return true;
1560 }
1561 } else if (family == AF_INET6) {
1562 if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->rmap_src.ipv6)) {
1563 src->ipv6 = nexthop->rmap_src.ipv6;
1564 return true;
1565 } else if (!IN6_IS_ADDR_UNSPECIFIED(&nexthop->src.ipv6)) {
1566 src->ipv6 = nexthop->src.ipv6;
1567 return true;
1568 }
1569 }
1570
1571 return false;
1572 }
1573
1574 /*
1575 * The function returns true if the attribute could be added
1576 * to the message, otherwise false is returned.
1577 */
netlink_route_nexthop_encap(struct nlmsghdr * n,size_t nlen,struct nexthop * nh)1578 static int netlink_route_nexthop_encap(struct nlmsghdr *n, size_t nlen,
1579 struct nexthop *nh)
1580 {
1581 struct rtattr *nest;
1582
1583 switch (nh->nh_encap_type) {
1584 case NET_VXLAN:
1585 if (!nl_attr_put16(n, nlen, RTA_ENCAP_TYPE, nh->nh_encap_type))
1586 return false;
1587
1588 nest = nl_attr_nest(n, nlen, RTA_ENCAP);
1589 if (!nest)
1590 return false;
1591
1592 if (!nl_attr_put32(n, nlen, 0 /* VXLAN_VNI */,
1593 nh->nh_encap.vni))
1594 return false;
1595 nl_attr_nest_end(n, nest);
1596 break;
1597 }
1598
1599 return true;
1600 }
1601
1602 /*
1603 * Routing table change via netlink interface, using a dataplane context object
1604 *
1605 * Returns -1 on failure, 0 when the msg doesn't fit entirely in the buffer
1606 * otherwise the number of bytes written to buf.
1607 */
netlink_route_multipath_msg_encode(int cmd,struct zebra_dplane_ctx * ctx,uint8_t * data,size_t datalen,bool fpm,bool force_nhg)1608 ssize_t netlink_route_multipath_msg_encode(int cmd,
1609 struct zebra_dplane_ctx *ctx,
1610 uint8_t *data, size_t datalen,
1611 bool fpm, bool force_nhg)
1612 {
1613 int bytelen;
1614 struct nexthop *nexthop = NULL;
1615 unsigned int nexthop_num;
1616 const char *routedesc;
1617 bool setsrc = false;
1618 union g_addr src;
1619 const struct prefix *p, *src_p;
1620 uint32_t table_id;
1621
1622 struct {
1623 struct nlmsghdr n;
1624 struct rtmsg r;
1625 char buf[];
1626 } *req = (void *)data;
1627
1628 p = dplane_ctx_get_dest(ctx);
1629 src_p = dplane_ctx_get_src(ctx);
1630
1631 if (datalen < sizeof(*req))
1632 return 0;
1633
1634 memset(req, 0, sizeof(*req));
1635
1636 bytelen = (p->family == AF_INET ? 4 : 16);
1637
1638 req->n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
1639 req->n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
1640
1641 if ((cmd == RTM_NEWROUTE) &&
1642 ((p->family == AF_INET) || v6_rr_semantics))
1643 req->n.nlmsg_flags |= NLM_F_REPLACE;
1644
1645 req->n.nlmsg_type = cmd;
1646
1647 req->n.nlmsg_pid = dplane_ctx_get_ns(ctx)->nls.snl.nl_pid;
1648
1649 req->r.rtm_family = p->family;
1650 req->r.rtm_dst_len = p->prefixlen;
1651 req->r.rtm_src_len = src_p ? src_p->prefixlen : 0;
1652 req->r.rtm_scope = RT_SCOPE_UNIVERSE;
1653
1654 if (cmd == RTM_DELROUTE)
1655 req->r.rtm_protocol = zebra2proto(dplane_ctx_get_old_type(ctx));
1656 else
1657 req->r.rtm_protocol = zebra2proto(dplane_ctx_get_type(ctx));
1658
1659 /*
1660 * blackhole routes are not RTN_UNICAST, they are
1661 * RTN_ BLACKHOLE|UNREACHABLE|PROHIBIT
1662 * so setting this value as a RTN_UNICAST would
1663 * cause the route lookup of just the prefix
1664 * to fail. So no need to specify this for
1665 * the RTM_DELROUTE case
1666 */
1667 if (cmd != RTM_DELROUTE)
1668 req->r.rtm_type = RTN_UNICAST;
1669
1670 if (!nl_attr_put(&req->n, datalen, RTA_DST, &p->u.prefix, bytelen))
1671 return 0;
1672 if (src_p) {
1673 if (!nl_attr_put(&req->n, datalen, RTA_SRC, &src_p->u.prefix,
1674 bytelen))
1675 return 0;
1676 }
1677
1678 /* Metric. */
1679 /* Hardcode the metric for all routes coming from zebra. Metric isn't
1680 * used
1681 * either by the kernel or by zebra. Its purely for calculating best
1682 * path(s)
1683 * by the routing protocol and for communicating with protocol peers.
1684 */
1685 if (!nl_attr_put32(&req->n, datalen, RTA_PRIORITY,
1686 NL_DEFAULT_ROUTE_METRIC))
1687 return 0;
1688
1689 #if defined(SUPPORT_REALMS)
1690 {
1691 route_tag_t tag;
1692
1693 if (cmd == RTM_DELROUTE)
1694 tag = dplane_ctx_get_old_tag(ctx);
1695 else
1696 tag = dplane_ctx_get_tag(ctx);
1697
1698 if (tag > 0 && tag <= 255) {
1699 if (!nl_attr_put32(&req->n, datalen, RTA_FLOW, tag))
1700 return 0;
1701 }
1702 }
1703 #endif
1704 /* Table corresponding to this route. */
1705 table_id = dplane_ctx_get_table(ctx);
1706 if (table_id < 256)
1707 req->r.rtm_table = table_id;
1708 else {
1709 req->r.rtm_table = RT_TABLE_UNSPEC;
1710 if (!nl_attr_put32(&req->n, datalen, RTA_TABLE, table_id))
1711 return 0;
1712 }
1713
1714 if (IS_ZEBRA_DEBUG_KERNEL)
1715 zlog_debug(
1716 "%s: %s %pFX vrf %u(%u)", __func__,
1717 nl_msg_type_to_str(cmd), p, dplane_ctx_get_vrf(ctx),
1718 table_id);
1719
1720 /*
1721 * If we are not updating the route and we have received
1722 * a route delete, then all we need to fill in is the
1723 * prefix information to tell the kernel to schwack
1724 * it.
1725 */
1726 if (cmd == RTM_DELROUTE)
1727 return NLMSG_ALIGN(req->n.nlmsg_len);
1728
1729 if (dplane_ctx_get_mtu(ctx) || dplane_ctx_get_nh_mtu(ctx)) {
1730 struct rtattr *nest;
1731 uint32_t mtu = dplane_ctx_get_mtu(ctx);
1732 uint32_t nexthop_mtu = dplane_ctx_get_nh_mtu(ctx);
1733
1734 if (!mtu || (nexthop_mtu && nexthop_mtu < mtu))
1735 mtu = nexthop_mtu;
1736
1737 nest = nl_attr_nest(&req->n, datalen, RTA_METRICS);
1738 if (nest == NULL)
1739 return 0;
1740
1741 if (!nl_attr_put(&req->n, datalen, RTAX_MTU, &mtu, sizeof(mtu)))
1742 return 0;
1743 nl_attr_nest_end(&req->n, nest);
1744 }
1745
1746 /*
1747 * Always install blackhole routes without using nexthops, because of
1748 * the following kernel problems:
1749 * 1. Kernel nexthops don't suport unreachable/prohibit route types.
1750 * 2. Blackhole kernel nexthops are deleted when loopback is down.
1751 */
1752 nexthop = dplane_ctx_get_ng(ctx)->nexthop;
1753 if (nexthop) {
1754 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
1755 nexthop = nexthop->resolved;
1756
1757 if (nexthop->type == NEXTHOP_TYPE_BLACKHOLE) {
1758 switch (nexthop->bh_type) {
1759 case BLACKHOLE_ADMINPROHIB:
1760 req->r.rtm_type = RTN_PROHIBIT;
1761 break;
1762 case BLACKHOLE_REJECT:
1763 req->r.rtm_type = RTN_UNREACHABLE;
1764 break;
1765 default:
1766 req->r.rtm_type = RTN_BLACKHOLE;
1767 break;
1768 }
1769 return NLMSG_ALIGN(req->n.nlmsg_len);
1770 }
1771 }
1772
1773 if ((!fpm && kernel_nexthops_supported()) || (fpm && force_nhg)) {
1774 /* Kernel supports nexthop objects */
1775 if (IS_ZEBRA_DEBUG_KERNEL)
1776 zlog_debug("%s: %pFX nhg_id is %u", __func__, p,
1777 dplane_ctx_get_nhe_id(ctx));
1778
1779 if (!nl_attr_put32(&req->n, datalen, RTA_NH_ID,
1780 dplane_ctx_get_nhe_id(ctx)))
1781 return 0;
1782
1783 /* Have to determine src still */
1784 for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), nexthop)) {
1785 if (setsrc)
1786 break;
1787
1788 setsrc = nexthop_set_src(nexthop, p->family, &src);
1789 }
1790
1791 if (setsrc) {
1792 if (p->family == AF_INET) {
1793 if (!nl_attr_put(&req->n, datalen, RTA_PREFSRC,
1794 &src.ipv4, bytelen))
1795 return 0;
1796 } else if (p->family == AF_INET6) {
1797 if (!nl_attr_put(&req->n, datalen, RTA_PREFSRC,
1798 &src.ipv6, bytelen))
1799 return 0;
1800 }
1801 }
1802
1803 return NLMSG_ALIGN(req->n.nlmsg_len);
1804 }
1805
1806 /* Count overall nexthops so we can decide whether to use singlepath
1807 * or multipath case.
1808 */
1809 nexthop_num = 0;
1810 for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), nexthop)) {
1811 if (CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_RECURSIVE))
1812 continue;
1813 if (!NEXTHOP_IS_ACTIVE(nexthop->flags))
1814 continue;
1815
1816 nexthop_num++;
1817 }
1818
1819 /* Singlepath case. */
1820 if (nexthop_num == 1) {
1821 nexthop_num = 0;
1822 for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), nexthop)) {
1823 if (CHECK_FLAG(nexthop->flags,
1824 NEXTHOP_FLAG_RECURSIVE)) {
1825
1826 if (setsrc)
1827 continue;
1828
1829 setsrc = nexthop_set_src(nexthop, p->family,
1830 &src);
1831 continue;
1832 }
1833
1834 if (NEXTHOP_IS_ACTIVE(nexthop->flags)) {
1835 routedesc = nexthop->rparent
1836 ? "recursive, single-path"
1837 : "single-path";
1838
1839 if (!_netlink_route_build_singlepath(
1840 p, routedesc, bytelen, nexthop,
1841 &req->n, &req->r, datalen, cmd))
1842 return 0;
1843 nexthop_num++;
1844 break;
1845 }
1846
1847 /*
1848 * Add encapsulation information when installing via
1849 * FPM.
1850 */
1851 if (fpm) {
1852 if (!netlink_route_nexthop_encap(
1853 &req->n, datalen, nexthop))
1854 return 0;
1855 }
1856 }
1857
1858 if (setsrc) {
1859 if (p->family == AF_INET) {
1860 if (!nl_attr_put(&req->n, datalen, RTA_PREFSRC,
1861 &src.ipv4, bytelen))
1862 return 0;
1863 } else if (p->family == AF_INET6) {
1864 if (!nl_attr_put(&req->n, datalen, RTA_PREFSRC,
1865 &src.ipv6, bytelen))
1866 return 0;
1867 }
1868 }
1869 } else { /* Multipath case */
1870 struct rtattr *nest;
1871 const union g_addr *src1 = NULL;
1872
1873 nest = nl_attr_nest(&req->n, datalen, RTA_MULTIPATH);
1874 if (nest == NULL)
1875 return 0;
1876
1877 nexthop_num = 0;
1878 for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx), nexthop)) {
1879 if (CHECK_FLAG(nexthop->flags,
1880 NEXTHOP_FLAG_RECURSIVE)) {
1881 /* This only works for IPv4 now */
1882 if (setsrc)
1883 continue;
1884
1885 setsrc = nexthop_set_src(nexthop, p->family,
1886 &src);
1887 continue;
1888 }
1889
1890 if (NEXTHOP_IS_ACTIVE(nexthop->flags)) {
1891 routedesc = nexthop->rparent
1892 ? "recursive, multipath"
1893 : "multipath";
1894 nexthop_num++;
1895
1896 if (!_netlink_route_build_multipath(
1897 p, routedesc, bytelen, nexthop,
1898 &req->n, datalen, &req->r, &src1))
1899 return 0;
1900
1901 if (!setsrc && src1) {
1902 if (p->family == AF_INET)
1903 src.ipv4 = src1->ipv4;
1904 else if (p->family == AF_INET6)
1905 src.ipv6 = src1->ipv6;
1906
1907 setsrc = 1;
1908 }
1909 }
1910 }
1911
1912 nl_attr_nest_end(&req->n, nest);
1913
1914 /*
1915 * Add encapsulation information when installing via
1916 * FPM.
1917 */
1918 if (fpm) {
1919 for (ALL_NEXTHOPS_PTR(dplane_ctx_get_ng(ctx),
1920 nexthop)) {
1921 if (CHECK_FLAG(nexthop->flags,
1922 NEXTHOP_FLAG_RECURSIVE))
1923 continue;
1924 if (!netlink_route_nexthop_encap(
1925 &req->n, datalen, nexthop))
1926 return 0;
1927 }
1928 }
1929
1930
1931 if (setsrc) {
1932 if (p->family == AF_INET) {
1933 if (!nl_attr_put(&req->n, datalen, RTA_PREFSRC,
1934 &src.ipv4, bytelen))
1935 return 0;
1936 } else if (p->family == AF_INET6) {
1937 if (!nl_attr_put(&req->n, datalen, RTA_PREFSRC,
1938 &src.ipv6, bytelen))
1939 return 0;
1940 }
1941 if (IS_ZEBRA_DEBUG_KERNEL)
1942 zlog_debug("Setting source");
1943 }
1944 }
1945
1946 /* If there is no useful nexthop then return. */
1947 if (nexthop_num == 0) {
1948 if (IS_ZEBRA_DEBUG_KERNEL)
1949 zlog_debug("%s: No useful nexthop.", __func__);
1950 }
1951
1952 return NLMSG_ALIGN(req->n.nlmsg_len);
1953 }
1954
kernel_get_ipmr_sg_stats(struct zebra_vrf * zvrf,void * in)1955 int kernel_get_ipmr_sg_stats(struct zebra_vrf *zvrf, void *in)
1956 {
1957 uint32_t actual_table;
1958 int suc = 0;
1959 struct mcast_route_data *mr = (struct mcast_route_data *)in;
1960 struct {
1961 struct nlmsghdr n;
1962 struct ndmsg ndm;
1963 char buf[256];
1964 } req;
1965
1966 mroute = mr;
1967 struct zebra_ns *zns;
1968
1969 zns = zvrf->zns;
1970 memset(&req, 0, sizeof(req));
1971
1972 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
1973 req.n.nlmsg_flags = NLM_F_REQUEST;
1974 req.n.nlmsg_pid = zns->netlink_cmd.snl.nl_pid;
1975
1976 req.ndm.ndm_family = RTNL_FAMILY_IPMR;
1977 req.n.nlmsg_type = RTM_GETROUTE;
1978
1979 nl_attr_put32(&req.n, sizeof(req), RTA_IIF, mroute->ifindex);
1980 nl_attr_put32(&req.n, sizeof(req), RTA_OIF, mroute->ifindex);
1981 nl_attr_put32(&req.n, sizeof(req), RTA_SRC, mroute->sg.src.s_addr);
1982 nl_attr_put32(&req.n, sizeof(req), RTA_DST, mroute->sg.grp.s_addr);
1983 /*
1984 * What?
1985 *
1986 * So during the namespace cleanup we started storing
1987 * the zvrf table_id for the default table as RT_TABLE_MAIN
1988 * which is what the normal routing table for ip routing is.
1989 * This change caused this to break our lookups of sg data
1990 * because prior to this change the zvrf->table_id was 0
1991 * and when the pim multicast kernel code saw a 0,
1992 * it was auto-translated to RT_TABLE_DEFAULT. But since
1993 * we are now passing in RT_TABLE_MAIN there is no auto-translation
1994 * and the kernel goes screw you and the delicious cookies you
1995 * are trying to give me. So now we have this little hack.
1996 */
1997 actual_table = (zvrf->table_id == RT_TABLE_MAIN) ? RT_TABLE_DEFAULT :
1998 zvrf->table_id;
1999 nl_attr_put32(&req.n, sizeof(req), RTA_TABLE, actual_table);
2000
2001 suc = netlink_talk(netlink_route_change_read_multicast, &req.n,
2002 &zns->netlink_cmd, zns, 0);
2003
2004 mroute = NULL;
2005 return suc;
2006 }
2007
2008 /* Char length to debug ID with */
2009 #define ID_LENGTH 10
2010
_netlink_nexthop_build_group(struct nlmsghdr * n,size_t req_size,uint32_t id,const struct nh_grp * z_grp,const uint8_t count)2011 static bool _netlink_nexthop_build_group(struct nlmsghdr *n, size_t req_size,
2012 uint32_t id,
2013 const struct nh_grp *z_grp,
2014 const uint8_t count)
2015 {
2016 struct nexthop_grp grp[count];
2017 /* Need space for max group size, "/", and null term */
2018 char buf[(MULTIPATH_NUM * (ID_LENGTH + 1)) + 1];
2019 char buf1[ID_LENGTH + 2];
2020
2021 buf[0] = '\0';
2022
2023 memset(grp, 0, sizeof(grp));
2024
2025 if (count) {
2026 for (int i = 0; i < count; i++) {
2027 grp[i].id = z_grp[i].id;
2028 grp[i].weight = z_grp[i].weight - 1;
2029
2030 if (IS_ZEBRA_DEBUG_KERNEL) {
2031 if (i == 0)
2032 snprintf(buf, sizeof(buf1), "group %u",
2033 grp[i].id);
2034 else {
2035 snprintf(buf1, sizeof(buf1), "/%u",
2036 grp[i].id);
2037 strlcat(buf, buf1, sizeof(buf));
2038 }
2039 }
2040 }
2041 if (!nl_attr_put(n, req_size, NHA_GROUP, grp,
2042 count * sizeof(*grp)))
2043 return false;
2044 }
2045
2046 if (IS_ZEBRA_DEBUG_KERNEL)
2047 zlog_debug("%s: ID (%u): %s", __func__, id, buf);
2048
2049 return true;
2050 }
2051
2052 /**
2053 * Next hop packet encoding helper function.
2054 *
2055 * \param[in] cmd netlink command.
2056 * \param[in] ctx dataplane context (information snapshot).
2057 * \param[out] buf buffer to hold the packet.
2058 * \param[in] buflen amount of buffer bytes.
2059 *
2060 * \returns -1 on failure, 0 when the msg doesn't fit entirely in the buffer
2061 * otherwise the number of bytes written to buf.
2062 */
netlink_nexthop_msg_encode(uint16_t cmd,const struct zebra_dplane_ctx * ctx,void * buf,size_t buflen)2063 ssize_t netlink_nexthop_msg_encode(uint16_t cmd,
2064 const struct zebra_dplane_ctx *ctx,
2065 void *buf, size_t buflen)
2066 {
2067 struct {
2068 struct nlmsghdr n;
2069 struct nhmsg nhm;
2070 char buf[];
2071 } *req = buf;
2072
2073 mpls_lse_t out_lse[MPLS_MAX_LABELS];
2074 char label_buf[256];
2075 int num_labels = 0;
2076
2077 label_buf[0] = '\0';
2078
2079 if (buflen < sizeof(*req))
2080 return 0;
2081
2082 memset(req, 0, sizeof(*req));
2083
2084 req->n.nlmsg_len = NLMSG_LENGTH(sizeof(struct nhmsg));
2085 req->n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
2086
2087 if (cmd == RTM_NEWNEXTHOP)
2088 req->n.nlmsg_flags |= NLM_F_REPLACE;
2089
2090 req->n.nlmsg_type = cmd;
2091 req->n.nlmsg_pid = dplane_ctx_get_ns(ctx)->nls.snl.nl_pid;
2092
2093 req->nhm.nh_family = AF_UNSPEC;
2094 /* TODO: Scope? */
2095
2096 uint32_t id = dplane_ctx_get_nhe_id(ctx);
2097
2098 if (!id) {
2099 flog_err(
2100 EC_ZEBRA_NHG_FIB_UPDATE,
2101 "Failed trying to update a nexthop group in the kernel that does not have an ID");
2102 return -1;
2103 }
2104
2105 if (!nl_attr_put32(&req->n, buflen, NHA_ID, id))
2106 return 0;
2107
2108 if (cmd == RTM_NEWNEXTHOP) {
2109 /*
2110 * We distinguish between a "group", which is a collection
2111 * of ids, and a singleton nexthop with an id. The
2112 * group is installed as an id that just refers to a list of
2113 * other ids.
2114 */
2115 if (dplane_ctx_get_nhe_nh_grp_count(ctx)) {
2116 if (!_netlink_nexthop_build_group(
2117 &req->n, buflen, id,
2118 dplane_ctx_get_nhe_nh_grp(ctx),
2119 dplane_ctx_get_nhe_nh_grp_count(ctx)))
2120 return 0;
2121 } else {
2122 const struct nexthop *nh =
2123 dplane_ctx_get_nhe_ng(ctx)->nexthop;
2124 afi_t afi = dplane_ctx_get_nhe_afi(ctx);
2125
2126 if (afi == AFI_IP)
2127 req->nhm.nh_family = AF_INET;
2128 else if (afi == AFI_IP6)
2129 req->nhm.nh_family = AF_INET6;
2130
2131 switch (nh->type) {
2132 case NEXTHOP_TYPE_IPV4:
2133 case NEXTHOP_TYPE_IPV4_IFINDEX:
2134 if (!nl_attr_put(&req->n, buflen, NHA_GATEWAY,
2135 &nh->gate.ipv4,
2136 IPV4_MAX_BYTELEN))
2137 return 0;
2138 break;
2139 case NEXTHOP_TYPE_IPV6:
2140 case NEXTHOP_TYPE_IPV6_IFINDEX:
2141 if (!nl_attr_put(&req->n, buflen, NHA_GATEWAY,
2142 &nh->gate.ipv6,
2143 IPV6_MAX_BYTELEN))
2144 return 0;
2145 break;
2146 case NEXTHOP_TYPE_BLACKHOLE:
2147 if (!nl_attr_put(&req->n, buflen, NHA_BLACKHOLE,
2148 NULL, 0))
2149 return 0;
2150 /* Blackhole shouldn't have anymore attributes
2151 */
2152 goto nexthop_done;
2153 case NEXTHOP_TYPE_IFINDEX:
2154 /* Don't need anymore info for this */
2155 break;
2156 }
2157
2158 if (!nh->ifindex) {
2159 flog_err(
2160 EC_ZEBRA_NHG_FIB_UPDATE,
2161 "Context received for kernel nexthop update without an interface");
2162 return -1;
2163 }
2164
2165 if (!nl_attr_put32(&req->n, buflen, NHA_OIF,
2166 nh->ifindex))
2167 return 0;
2168
2169 if (CHECK_FLAG(nh->flags, NEXTHOP_FLAG_ONLINK))
2170 req->nhm.nh_flags |= RTNH_F_ONLINK;
2171
2172 num_labels =
2173 build_label_stack(nh->nh_label, out_lse,
2174 label_buf, sizeof(label_buf));
2175
2176 if (num_labels) {
2177 /* Set the BoS bit */
2178 out_lse[num_labels - 1] |=
2179 htonl(1 << MPLS_LS_S_SHIFT);
2180
2181 /*
2182 * TODO: MPLS unsupported for now in kernel.
2183 */
2184 if (req->nhm.nh_family == AF_MPLS)
2185 goto nexthop_done;
2186 #if 0
2187 if (!nl_attr_put(&req->n, buflen, NHA_NEWDST,
2188 &out_lse,
2189 num_labels
2190 * sizeof(mpls_lse_t)))
2191 return 0;
2192 #endif
2193 else {
2194 struct rtattr *nest;
2195 uint16_t encap = LWTUNNEL_ENCAP_MPLS;
2196
2197 if (!nl_attr_put16(&req->n, buflen,
2198 NHA_ENCAP_TYPE,
2199 encap))
2200 return 0;
2201 nest = nl_attr_nest(&req->n, buflen,
2202 NHA_ENCAP);
2203 if (!nest)
2204 return 0;
2205 if (!nl_attr_put(
2206 &req->n, buflen,
2207 MPLS_IPTUNNEL_DST, &out_lse,
2208 num_labels
2209 * sizeof(
2210 mpls_lse_t)))
2211 return 0;
2212 nl_attr_nest_end(&req->n, nest);
2213 }
2214 }
2215
2216 nexthop_done:
2217
2218 if (IS_ZEBRA_DEBUG_KERNEL)
2219 zlog_debug("%s: ID (%u): %pNHv(%d) vrf %s(%u) %s ",
2220 __func__, id, nh, nh->ifindex,
2221 vrf_id_to_name(nh->vrf_id),
2222 nh->vrf_id, label_buf);
2223 }
2224
2225 req->nhm.nh_protocol =
2226 zebra2proto(dplane_ctx_get_nhe_type(ctx));
2227
2228 } else if (cmd != RTM_DELNEXTHOP) {
2229 flog_err(
2230 EC_ZEBRA_NHG_FIB_UPDATE,
2231 "Nexthop group kernel update command (%d) does not exist",
2232 cmd);
2233 return -1;
2234 }
2235
2236 if (IS_ZEBRA_DEBUG_KERNEL)
2237 zlog_debug("%s: %s, id=%u", __func__, nl_msg_type_to_str(cmd),
2238 id);
2239
2240 return NLMSG_ALIGN(req->n.nlmsg_len);
2241 }
2242
netlink_nexthop_msg_encoder(struct zebra_dplane_ctx * ctx,void * buf,size_t buflen)2243 static ssize_t netlink_nexthop_msg_encoder(struct zebra_dplane_ctx *ctx,
2244 void *buf, size_t buflen)
2245 {
2246 enum dplane_op_e op;
2247 int cmd = 0;
2248
2249 op = dplane_ctx_get_op(ctx);
2250 if (op == DPLANE_OP_NH_INSTALL || op == DPLANE_OP_NH_UPDATE)
2251 cmd = RTM_NEWNEXTHOP;
2252 else if (op == DPLANE_OP_NH_DELETE)
2253 cmd = RTM_DELNEXTHOP;
2254 else {
2255 flog_err(EC_ZEBRA_NHG_FIB_UPDATE,
2256 "Context received for kernel nexthop update with incorrect OP code (%u)",
2257 op);
2258 return -1;
2259 }
2260
2261 return netlink_nexthop_msg_encode(cmd, ctx, buf, buflen);
2262 }
2263
2264 enum netlink_msg_status
netlink_put_nexthop_update_msg(struct nl_batch * bth,struct zebra_dplane_ctx * ctx)2265 netlink_put_nexthop_update_msg(struct nl_batch *bth,
2266 struct zebra_dplane_ctx *ctx)
2267 {
2268 /* Nothing to do if the kernel doesn't support nexthop objects */
2269 if (!kernel_nexthops_supported())
2270 return FRR_NETLINK_SUCCESS;
2271
2272 return netlink_batch_add_msg(bth, ctx, netlink_nexthop_msg_encoder,
2273 false);
2274 }
2275
netlink_newroute_msg_encoder(struct zebra_dplane_ctx * ctx,void * buf,size_t buflen)2276 static ssize_t netlink_newroute_msg_encoder(struct zebra_dplane_ctx *ctx,
2277 void *buf, size_t buflen)
2278 {
2279 return netlink_route_multipath_msg_encode(RTM_NEWROUTE, ctx, buf,
2280 buflen, false, false);
2281 }
2282
netlink_delroute_msg_encoder(struct zebra_dplane_ctx * ctx,void * buf,size_t buflen)2283 static ssize_t netlink_delroute_msg_encoder(struct zebra_dplane_ctx *ctx,
2284 void *buf, size_t buflen)
2285 {
2286 return netlink_route_multipath_msg_encode(RTM_DELROUTE, ctx, buf,
2287 buflen, false, false);
2288 }
2289
2290 enum netlink_msg_status
netlink_put_route_update_msg(struct nl_batch * bth,struct zebra_dplane_ctx * ctx)2291 netlink_put_route_update_msg(struct nl_batch *bth, struct zebra_dplane_ctx *ctx)
2292 {
2293 int cmd;
2294 const struct prefix *p = dplane_ctx_get_dest(ctx);
2295
2296 if (dplane_ctx_get_op(ctx) == DPLANE_OP_ROUTE_DELETE) {
2297 cmd = RTM_DELROUTE;
2298 } else if (dplane_ctx_get_op(ctx) == DPLANE_OP_ROUTE_INSTALL) {
2299 cmd = RTM_NEWROUTE;
2300 } else if (dplane_ctx_get_op(ctx) == DPLANE_OP_ROUTE_UPDATE) {
2301
2302 if (p->family == AF_INET || v6_rr_semantics) {
2303 /* Single 'replace' operation */
2304
2305 /*
2306 * With route replace semantics in place
2307 * for v4 routes and the new route is a system
2308 * route we do not install anything.
2309 * The problem here is that the new system
2310 * route should cause us to withdraw from
2311 * the kernel the old non-system route
2312 */
2313 if (RSYSTEM_ROUTE(dplane_ctx_get_type(ctx))
2314 && !RSYSTEM_ROUTE(dplane_ctx_get_old_type(ctx)))
2315 netlink_batch_add_msg(
2316 bth, ctx, netlink_delroute_msg_encoder,
2317 true);
2318 } else {
2319 /*
2320 * So v6 route replace semantics are not in
2321 * the kernel at this point as I understand it.
2322 * so let's do a delete then an add.
2323 * In the future once v6 route replace semantics
2324 * are in we can figure out what to do here to
2325 * allow working with old and new kernels.
2326 *
2327 * I'm also intentionally ignoring the failure case
2328 * of the route delete. If that happens yeah we're
2329 * screwed.
2330 */
2331 if (!RSYSTEM_ROUTE(dplane_ctx_get_old_type(ctx)))
2332 netlink_batch_add_msg(
2333 bth, ctx, netlink_delroute_msg_encoder,
2334 true);
2335 }
2336
2337 cmd = RTM_NEWROUTE;
2338 } else
2339 return FRR_NETLINK_ERROR;
2340
2341 if (RSYSTEM_ROUTE(dplane_ctx_get_type(ctx)))
2342 return FRR_NETLINK_SUCCESS;
2343
2344 return netlink_batch_add_msg(bth, ctx,
2345 cmd == RTM_NEWROUTE
2346 ? netlink_newroute_msg_encoder
2347 : netlink_delroute_msg_encoder,
2348 false);
2349 }
2350
2351 /**
2352 * netlink_nexthop_process_nh() - Parse the gatway/if info from a new nexthop
2353 *
2354 * @tb: Netlink RTA data
2355 * @family: Address family in the nhmsg
2356 * @ifp: Interface connected - this should be NULL, we fill it in
2357 * @ns_id: Namspace id
2358 *
2359 * Return: New nexthop
2360 */
netlink_nexthop_process_nh(struct rtattr ** tb,unsigned char family,struct interface ** ifp,ns_id_t ns_id)2361 static struct nexthop netlink_nexthop_process_nh(struct rtattr **tb,
2362 unsigned char family,
2363 struct interface **ifp,
2364 ns_id_t ns_id)
2365 {
2366 struct nexthop nh = {};
2367 void *gate = NULL;
2368 enum nexthop_types_t type = 0;
2369 int if_index = 0;
2370 size_t sz = 0;
2371 struct interface *ifp_lookup;
2372
2373 if_index = *(int *)RTA_DATA(tb[NHA_OIF]);
2374
2375
2376 if (tb[NHA_GATEWAY]) {
2377 switch (family) {
2378 case AF_INET:
2379 type = NEXTHOP_TYPE_IPV4_IFINDEX;
2380 sz = 4;
2381 break;
2382 case AF_INET6:
2383 type = NEXTHOP_TYPE_IPV6_IFINDEX;
2384 sz = 16;
2385 break;
2386 default:
2387 flog_warn(
2388 EC_ZEBRA_BAD_NHG_MESSAGE,
2389 "Nexthop gateway with bad address family (%d) received from kernel",
2390 family);
2391 return nh;
2392 }
2393 gate = RTA_DATA(tb[NHA_GATEWAY]);
2394 } else
2395 type = NEXTHOP_TYPE_IFINDEX;
2396
2397 if (type)
2398 nh.type = type;
2399
2400 if (gate)
2401 memcpy(&(nh.gate), gate, sz);
2402
2403 if (if_index)
2404 nh.ifindex = if_index;
2405
2406 ifp_lookup =
2407 if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id), nh.ifindex);
2408
2409 if (ifp)
2410 *ifp = ifp_lookup;
2411 if (ifp_lookup)
2412 nh.vrf_id = ifp_lookup->vrf_id;
2413 else {
2414 flog_warn(
2415 EC_ZEBRA_UNKNOWN_INTERFACE,
2416 "%s: Unknown nexthop interface %u received, defaulting to VRF_DEFAULT",
2417 __func__, nh.ifindex);
2418
2419 nh.vrf_id = VRF_DEFAULT;
2420 }
2421
2422 if (tb[NHA_ENCAP] && tb[NHA_ENCAP_TYPE]) {
2423 uint16_t encap_type = *(uint16_t *)RTA_DATA(tb[NHA_ENCAP_TYPE]);
2424 int num_labels = 0;
2425
2426 mpls_label_t labels[MPLS_MAX_LABELS] = {0};
2427
2428 if (encap_type == LWTUNNEL_ENCAP_MPLS)
2429 num_labels = parse_encap_mpls(tb[NHA_ENCAP], labels);
2430
2431 if (num_labels)
2432 nexthop_add_labels(&nh, ZEBRA_LSP_STATIC, num_labels,
2433 labels);
2434 }
2435
2436 return nh;
2437 }
2438
netlink_nexthop_process_group(struct rtattr ** tb,struct nh_grp * z_grp,int z_grp_size)2439 static int netlink_nexthop_process_group(struct rtattr **tb,
2440 struct nh_grp *z_grp, int z_grp_size)
2441 {
2442 uint8_t count = 0;
2443 /* linux/nexthop.h group struct */
2444 struct nexthop_grp *n_grp = NULL;
2445
2446 n_grp = (struct nexthop_grp *)RTA_DATA(tb[NHA_GROUP]);
2447 count = (RTA_PAYLOAD(tb[NHA_GROUP]) / sizeof(*n_grp));
2448
2449 if (!count || (count * sizeof(*n_grp)) != RTA_PAYLOAD(tb[NHA_GROUP])) {
2450 flog_warn(EC_ZEBRA_BAD_NHG_MESSAGE,
2451 "Invalid nexthop group received from the kernel");
2452 return count;
2453 }
2454
2455 #if 0
2456 // TODO: Need type for something?
2457 zlog_debug("Nexthop group type: %d",
2458 *((uint16_t *)RTA_DATA(tb[NHA_GROUP_TYPE])));
2459
2460 #endif
2461
2462 for (int i = 0; ((i < count) && (i < z_grp_size)); i++) {
2463 z_grp[i].id = n_grp[i].id;
2464 z_grp[i].weight = n_grp[i].weight + 1;
2465 }
2466 return count;
2467 }
2468
2469 /**
2470 * netlink_nexthop_change() - Read in change about nexthops from the kernel
2471 *
2472 * @h: Netlink message header
2473 * @ns_id: Namspace id
2474 * @startup: Are we reading under startup conditions?
2475 *
2476 * Return: Result status
2477 */
netlink_nexthop_change(struct nlmsghdr * h,ns_id_t ns_id,int startup)2478 int netlink_nexthop_change(struct nlmsghdr *h, ns_id_t ns_id, int startup)
2479 {
2480 int len;
2481 /* nexthop group id */
2482 uint32_t id;
2483 unsigned char family;
2484 int type;
2485 afi_t afi = AFI_UNSPEC;
2486 vrf_id_t vrf_id = VRF_DEFAULT;
2487 struct interface *ifp = NULL;
2488 struct nhmsg *nhm = NULL;
2489 struct nexthop nh = {};
2490 struct nh_grp grp[MULTIPATH_NUM] = {};
2491 /* Count of nexthops in group array */
2492 uint8_t grp_count = 0;
2493 struct rtattr *tb[NHA_MAX + 1] = {};
2494
2495 nhm = NLMSG_DATA(h);
2496
2497 if (ns_id)
2498 vrf_id = ns_id;
2499
2500 if (startup && h->nlmsg_type != RTM_NEWNEXTHOP)
2501 return 0;
2502
2503 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct nhmsg));
2504 if (len < 0) {
2505 zlog_warn(
2506 "%s: Message received from netlink is of a broken size %d %zu",
2507 __func__, h->nlmsg_len,
2508 (size_t)NLMSG_LENGTH(sizeof(struct nhmsg)));
2509 return -1;
2510 }
2511
2512 netlink_parse_rtattr(tb, NHA_MAX, RTM_NHA(nhm), len);
2513
2514
2515 if (!tb[NHA_ID]) {
2516 flog_warn(
2517 EC_ZEBRA_BAD_NHG_MESSAGE,
2518 "Nexthop group without an ID received from the kernel");
2519 return -1;
2520 }
2521
2522 /* We use the ID key'd nhg table for kernel updates */
2523 id = *((uint32_t *)RTA_DATA(tb[NHA_ID]));
2524
2525 if (zebra_evpn_mh_is_fdb_nh(id)) {
2526 /* If this is a L2 NH just ignore it */
2527 if (IS_ZEBRA_DEBUG_KERNEL || IS_ZEBRA_DEBUG_EVPN_MH_NH) {
2528 zlog_debug("Ignore kernel update (%u) for fdb-nh 0x%x",
2529 h->nlmsg_type, id);
2530 }
2531 return 0;
2532 }
2533
2534 family = nhm->nh_family;
2535 afi = family2afi(family);
2536
2537 type = proto2zebra(nhm->nh_protocol, 0, true);
2538
2539 if (IS_ZEBRA_DEBUG_KERNEL)
2540 zlog_debug("%s ID (%u) %s NS %u",
2541 nl_msg_type_to_str(h->nlmsg_type), id,
2542 nl_family_to_str(family), ns_id);
2543
2544
2545 if (h->nlmsg_type == RTM_NEWNEXTHOP) {
2546 if (tb[NHA_GROUP]) {
2547 /**
2548 * If this is a group message its only going to have
2549 * an array of nexthop IDs associated with it
2550 */
2551 grp_count = netlink_nexthop_process_group(
2552 tb, grp, array_size(grp));
2553 } else {
2554 if (tb[NHA_BLACKHOLE]) {
2555 /**
2556 * This nexthop is just for blackhole-ing
2557 * traffic, it should not have an OIF, GATEWAY,
2558 * or ENCAP
2559 */
2560 nh.type = NEXTHOP_TYPE_BLACKHOLE;
2561 nh.bh_type = BLACKHOLE_UNSPEC;
2562 } else if (tb[NHA_OIF])
2563 /**
2564 * This is a true new nexthop, so we need
2565 * to parse the gateway and device info
2566 */
2567 nh = netlink_nexthop_process_nh(tb, family,
2568 &ifp, ns_id);
2569 else {
2570
2571 flog_warn(
2572 EC_ZEBRA_BAD_NHG_MESSAGE,
2573 "Invalid Nexthop message received from the kernel with ID (%u)",
2574 id);
2575 return -1;
2576 }
2577 SET_FLAG(nh.flags, NEXTHOP_FLAG_ACTIVE);
2578 if (nhm->nh_flags & RTNH_F_ONLINK)
2579 SET_FLAG(nh.flags, NEXTHOP_FLAG_ONLINK);
2580 vrf_id = nh.vrf_id;
2581 }
2582
2583 if (zebra_nhg_kernel_find(id, &nh, grp, grp_count, vrf_id, afi,
2584 type, startup))
2585 return -1;
2586
2587 } else if (h->nlmsg_type == RTM_DELNEXTHOP)
2588 zebra_nhg_kernel_del(id, vrf_id);
2589
2590 return 0;
2591 }
2592
2593 /**
2594 * netlink_request_nexthop() - Request nextop information from the kernel
2595 * @zns: Zebra namespace
2596 * @family: AF_* netlink family
2597 * @type: RTM_* route type
2598 *
2599 * Return: Result status
2600 */
netlink_request_nexthop(struct zebra_ns * zns,int family,int type)2601 static int netlink_request_nexthop(struct zebra_ns *zns, int family, int type)
2602 {
2603 struct {
2604 struct nlmsghdr n;
2605 struct nhmsg nhm;
2606 } req;
2607
2608 /* Form the request, specifying filter (rtattr) if needed. */
2609 memset(&req, 0, sizeof(req));
2610 req.n.nlmsg_type = type;
2611 req.n.nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST;
2612 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct nhmsg));
2613 req.nhm.nh_family = family;
2614
2615 return netlink_request(&zns->netlink_cmd, &req);
2616 }
2617
2618
2619 /**
2620 * netlink_nexthop_read() - Nexthop read function using netlink interface
2621 *
2622 * @zns: Zebra name space
2623 *
2624 * Return: Result status
2625 * Only called at bootstrap time.
2626 */
netlink_nexthop_read(struct zebra_ns * zns)2627 int netlink_nexthop_read(struct zebra_ns *zns)
2628 {
2629 int ret;
2630 struct zebra_dplane_info dp_info;
2631
2632 zebra_dplane_info_from_zns(&dp_info, zns, true /*is_cmd*/);
2633
2634 /* Get nexthop objects */
2635 ret = netlink_request_nexthop(zns, AF_UNSPEC, RTM_GETNEXTHOP);
2636 if (ret < 0)
2637 return ret;
2638 ret = netlink_parse_info(netlink_nexthop_change, &zns->netlink_cmd,
2639 &dp_info, 0, 1);
2640
2641 if (!ret)
2642 /* If we succesfully read in nexthop objects,
2643 * this kernel must support them.
2644 */
2645 supports_nh = true;
2646
2647 if (IS_ZEBRA_DEBUG_KERNEL || IS_ZEBRA_DEBUG_NHG)
2648 zlog_debug("Nexthop objects %ssupported on this kernel",
2649 supports_nh ? "" : "not ");
2650
2651 return ret;
2652 }
2653
2654
kernel_neigh_update(int add,int ifindex,uint32_t addr,char * lla,int llalen,ns_id_t ns_id)2655 int kernel_neigh_update(int add, int ifindex, uint32_t addr, char *lla,
2656 int llalen, ns_id_t ns_id)
2657 {
2658 return netlink_neigh_update(add ? RTM_NEWNEIGH : RTM_DELNEIGH, ifindex,
2659 addr, lla, llalen, ns_id);
2660 }
2661
2662 /**
2663 * netlink_neigh_update_msg_encode() - Common helper api for encoding
2664 * evpn neighbor update as netlink messages using dataplane context object.
2665 * Here, a neighbor refers to a bridge forwarding database entry for
2666 * either unicast forwarding or head-end replication or an IP neighbor
2667 * entry.
2668 * @ctx: Dataplane context
2669 * @cmd: Netlink command (RTM_NEWNEIGH or RTM_DELNEIGH)
2670 * @mac: A neighbor cache link layer address
2671 * @ip: A neighbor cache n/w layer destination address
2672 * In the case of bridge FDB, this represnts the remote
2673 * VTEP IP.
2674 * @replace_obj: Whether NEW request should replace existing object or
2675 * add to the end of the list
2676 * @family: AF_* netlink family
2677 * @type: RTN_* route type
2678 * @flags: NTF_* flags
2679 * @state: NUD_* states
2680 * @data: data buffer pointer
2681 * @datalen: total amount of data buffer space
2682 *
2683 * Return: 0 when the msg doesn't fit entirely in the buffer
2684 * otherwise the number of bytes written to buf.
2685 */
netlink_neigh_update_msg_encode(const struct zebra_dplane_ctx * ctx,int cmd,const struct ethaddr * mac,const struct ipaddr * ip,bool replace_obj,uint8_t family,uint8_t type,uint8_t flags,uint16_t state,uint32_t nhg_id,bool nfy,uint8_t nfy_flags,void * data,size_t datalen)2686 static ssize_t netlink_neigh_update_msg_encode(
2687 const struct zebra_dplane_ctx *ctx, int cmd, const struct ethaddr *mac,
2688 const struct ipaddr *ip, bool replace_obj, uint8_t family, uint8_t type,
2689 uint8_t flags, uint16_t state, uint32_t nhg_id,
2690 bool nfy, uint8_t nfy_flags,
2691 void *data, size_t datalen)
2692 {
2693 uint8_t protocol = RTPROT_ZEBRA;
2694 struct {
2695 struct nlmsghdr n;
2696 struct ndmsg ndm;
2697 char buf[];
2698 } *req = data;
2699 int ipa_len;
2700 enum dplane_op_e op;
2701
2702 if (datalen < sizeof(*req))
2703 return 0;
2704 memset(req, 0, sizeof(*req));
2705
2706 op = dplane_ctx_get_op(ctx);
2707
2708 req->n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
2709 req->n.nlmsg_flags = NLM_F_REQUEST;
2710 if (cmd == RTM_NEWNEIGH)
2711 req->n.nlmsg_flags |=
2712 NLM_F_CREATE
2713 | (replace_obj ? NLM_F_REPLACE : NLM_F_APPEND);
2714 req->n.nlmsg_type = cmd;
2715 req->ndm.ndm_family = family;
2716 req->ndm.ndm_type = type;
2717 req->ndm.ndm_state = state;
2718 req->ndm.ndm_flags = flags;
2719 req->ndm.ndm_ifindex = dplane_ctx_get_ifindex(ctx);
2720
2721 if (!nl_attr_put(&req->n, datalen, NDA_PROTOCOL, &protocol,
2722 sizeof(protocol)))
2723 return 0;
2724
2725 if (mac) {
2726 if (!nl_attr_put(&req->n, datalen, NDA_LLADDR, mac, 6))
2727 return 0;
2728 }
2729
2730 if (nhg_id) {
2731 if (!nl_attr_put32(&req->n, datalen, NDA_NH_ID, nhg_id))
2732 return 0;
2733 }
2734 if (nfy) {
2735 if (!nl_attr_put(&req->n, datalen, NDA_NOTIFY,
2736 &nfy_flags, sizeof(nfy_flags)))
2737 return 0;
2738 }
2739
2740 ipa_len = IS_IPADDR_V4(ip) ? IPV4_MAX_BYTELEN : IPV6_MAX_BYTELEN;
2741 if (!nl_attr_put(&req->n, datalen, NDA_DST, &ip->ip.addr, ipa_len))
2742 return 0;
2743
2744 if (op == DPLANE_OP_MAC_INSTALL || op == DPLANE_OP_MAC_DELETE) {
2745 vlanid_t vid = dplane_ctx_mac_get_vlan(ctx);
2746
2747 if (vid > 0) {
2748 if (!nl_attr_put16(&req->n, datalen, NDA_VLAN, vid))
2749 return 0;
2750 }
2751
2752 if (!nl_attr_put32(&req->n, datalen, NDA_MASTER,
2753 dplane_ctx_mac_get_br_ifindex(ctx)))
2754 return 0;
2755 }
2756
2757 return NLMSG_ALIGN(req->n.nlmsg_len);
2758 }
2759
2760 /*
2761 * Add remote VTEP to the flood list for this VxLAN interface (VNI). This
2762 * is done by adding an FDB entry with a MAC of 00:00:00:00:00:00.
2763 */
2764 static ssize_t
netlink_vxlan_flood_update_ctx(const struct zebra_dplane_ctx * ctx,int cmd,void * buf,size_t buflen)2765 netlink_vxlan_flood_update_ctx(const struct zebra_dplane_ctx *ctx, int cmd,
2766 void *buf, size_t buflen)
2767 {
2768 struct ethaddr dst_mac = {.octet = {0}};
2769
2770 return netlink_neigh_update_msg_encode(
2771 ctx, cmd, &dst_mac, dplane_ctx_neigh_get_ipaddr(ctx), false,
2772 PF_BRIDGE, 0, NTF_SELF, (NUD_NOARP | NUD_PERMANENT), 0 /*nhg*/,
2773 false /*nfy*/, 0 /*nfy_flags*/, buf, buflen);
2774 }
2775
2776 #ifndef NDA_RTA
2777 #define NDA_RTA(r) \
2778 ((struct rtattr *)(((char *)(r)) + NLMSG_ALIGN(sizeof(struct ndmsg))))
2779 #endif
2780
netlink_macfdb_change(struct nlmsghdr * h,int len,ns_id_t ns_id)2781 static int netlink_macfdb_change(struct nlmsghdr *h, int len, ns_id_t ns_id)
2782 {
2783 struct ndmsg *ndm;
2784 struct interface *ifp;
2785 struct zebra_if *zif;
2786 struct rtattr *tb[NDA_MAX + 1];
2787 struct interface *br_if;
2788 struct ethaddr mac;
2789 vlanid_t vid = 0;
2790 struct in_addr vtep_ip;
2791 int vid_present = 0, dst_present = 0;
2792 char buf[ETHER_ADDR_STRLEN];
2793 char vid_buf[20];
2794 char dst_buf[30];
2795 bool sticky;
2796 bool local_inactive = false;
2797 bool dp_static = false;
2798 uint32_t nhg_id = 0;
2799
2800 ndm = NLMSG_DATA(h);
2801
2802 /* We only process macfdb notifications if EVPN is enabled */
2803 if (!is_evpn_enabled())
2804 return 0;
2805
2806 /* Parse attributes and extract fields of interest. Do basic
2807 * validation of the fields.
2808 */
2809 memset(tb, 0, sizeof tb);
2810 netlink_parse_rtattr(tb, NDA_MAX, NDA_RTA(ndm), len);
2811
2812 if (!tb[NDA_LLADDR]) {
2813 if (IS_ZEBRA_DEBUG_KERNEL)
2814 zlog_debug("%s AF_BRIDGE IF %u - no LLADDR",
2815 nl_msg_type_to_str(h->nlmsg_type),
2816 ndm->ndm_ifindex);
2817 return 0;
2818 }
2819
2820 if (RTA_PAYLOAD(tb[NDA_LLADDR]) != ETH_ALEN) {
2821 if (IS_ZEBRA_DEBUG_KERNEL)
2822 zlog_debug(
2823 "%s AF_BRIDGE IF %u - LLADDR is not MAC, len %lu",
2824 nl_msg_type_to_str(h->nlmsg_type), ndm->ndm_ifindex,
2825 (unsigned long)RTA_PAYLOAD(tb[NDA_LLADDR]));
2826 return 0;
2827 }
2828
2829 memcpy(&mac, RTA_DATA(tb[NDA_LLADDR]), ETH_ALEN);
2830
2831 if ((NDA_VLAN <= NDA_MAX) && tb[NDA_VLAN]) {
2832 vid_present = 1;
2833 vid = *(uint16_t *)RTA_DATA(tb[NDA_VLAN]);
2834 snprintf(vid_buf, sizeof(vid_buf), " VLAN %u", vid);
2835 }
2836
2837 if (tb[NDA_DST]) {
2838 /* TODO: Only IPv4 supported now. */
2839 dst_present = 1;
2840 memcpy(&vtep_ip.s_addr, RTA_DATA(tb[NDA_DST]),
2841 IPV4_MAX_BYTELEN);
2842 snprintf(dst_buf, sizeof(dst_buf), " dst %s",
2843 inet_ntoa(vtep_ip));
2844 }
2845
2846 if (tb[NDA_NH_ID])
2847 nhg_id = *(uint32_t *)RTA_DATA(tb[NDA_NH_ID]);
2848
2849 if (ndm->ndm_state & NUD_STALE)
2850 local_inactive = true;
2851
2852 if (tb[NDA_NOTIFY]) {
2853 uint8_t nfy_flags;
2854
2855 dp_static = true;
2856 nfy_flags = *(uint8_t *)RTA_DATA(tb[NDA_NOTIFY]);
2857 /* local activity has not been detected on the entry */
2858 if (nfy_flags & (1 << BR_FDB_NFY_INACTIVE))
2859 local_inactive = true;
2860 }
2861
2862 if (IS_ZEBRA_DEBUG_KERNEL)
2863 zlog_debug("Rx %s AF_BRIDGE IF %u%s st 0x%x fl 0x%x MAC %s%s nhg %d",
2864 nl_msg_type_to_str(h->nlmsg_type),
2865 ndm->ndm_ifindex, vid_present ? vid_buf : "",
2866 ndm->ndm_state, ndm->ndm_flags,
2867 prefix_mac2str(&mac, buf, sizeof(buf)),
2868 dst_present ? dst_buf : "", nhg_id);
2869
2870 /* The interface should exist. */
2871 ifp = if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id),
2872 ndm->ndm_ifindex);
2873 if (!ifp || !ifp->info)
2874 return 0;
2875
2876 /* The interface should be something we're interested in. */
2877 if (!IS_ZEBRA_IF_BRIDGE_SLAVE(ifp))
2878 return 0;
2879
2880 zif = (struct zebra_if *)ifp->info;
2881 if ((br_if = zif->brslave_info.br_if) == NULL) {
2882 if (IS_ZEBRA_DEBUG_KERNEL)
2883 zlog_debug(
2884 "%s AF_BRIDGE IF %s(%u) brIF %u - no bridge master",
2885 nl_msg_type_to_str(h->nlmsg_type), ifp->name,
2886 ndm->ndm_ifindex,
2887 zif->brslave_info.bridge_ifindex);
2888 return 0;
2889 }
2890
2891 sticky = !!(ndm->ndm_flags & NTF_STICKY);
2892
2893 if (filter_vlan && vid != filter_vlan) {
2894 if (IS_ZEBRA_DEBUG_KERNEL)
2895 zlog_debug(" Filtered due to filter vlan: %d",
2896 filter_vlan);
2897 return 0;
2898 }
2899
2900 /* If add or update, do accordingly if learnt on a "local" interface; if
2901 * the notification is over VxLAN, this has to be related to
2902 * multi-homing,
2903 * so perform an implicit delete of any local entry (if it exists).
2904 */
2905 if (h->nlmsg_type == RTM_NEWNEIGH) {
2906 /* Drop "permanent" entries. */
2907 if (ndm->ndm_state & NUD_PERMANENT) {
2908 if (IS_ZEBRA_DEBUG_KERNEL)
2909 zlog_debug(
2910 " Dropping entry because of NUD_PERMANENT");
2911 return 0;
2912 }
2913
2914 if (IS_ZEBRA_IF_VXLAN(ifp))
2915 return zebra_vxlan_check_del_local_mac(ifp, br_if, &mac,
2916 vid);
2917
2918 return zebra_vxlan_local_mac_add_update(ifp, br_if, &mac, vid,
2919 sticky, local_inactive, dp_static);
2920 }
2921
2922 /* This is a delete notification.
2923 * Ignore the notification with IP dest as it may just signify that the
2924 * MAC has moved from remote to local. The exception is the special
2925 * all-zeros MAC that represents the BUM flooding entry; we may have
2926 * to readd it. Otherwise,
2927 * 1. For a MAC over VxLan, check if it needs to be refreshed(readded)
2928 * 2. For a MAC over "local" interface, delete the mac
2929 * Note: We will get notifications from both bridge driver and VxLAN
2930 * driver.
2931 */
2932 if (nhg_id)
2933 return 0;
2934
2935 if (dst_present) {
2936 u_char zero_mac[6] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0};
2937
2938 if (!memcmp(zero_mac, mac.octet, ETH_ALEN))
2939 return zebra_vxlan_check_readd_vtep(ifp, vtep_ip);
2940 return 0;
2941 }
2942
2943 if (IS_ZEBRA_IF_VXLAN(ifp))
2944 return zebra_vxlan_check_readd_remote_mac(ifp, br_if, &mac,
2945 vid);
2946
2947 return zebra_vxlan_local_mac_del(ifp, br_if, &mac, vid);
2948 }
2949
netlink_macfdb_table(struct nlmsghdr * h,ns_id_t ns_id,int startup)2950 static int netlink_macfdb_table(struct nlmsghdr *h, ns_id_t ns_id, int startup)
2951 {
2952 int len;
2953 struct ndmsg *ndm;
2954
2955 if (h->nlmsg_type != RTM_NEWNEIGH)
2956 return 0;
2957
2958 /* Length validity. */
2959 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct ndmsg));
2960 if (len < 0)
2961 return -1;
2962
2963 /* We are interested only in AF_BRIDGE notifications. */
2964 ndm = NLMSG_DATA(h);
2965 if (ndm->ndm_family != AF_BRIDGE)
2966 return 0;
2967
2968 return netlink_macfdb_change(h, len, ns_id);
2969 }
2970
2971 /* Request for MAC FDB information from the kernel */
netlink_request_macs(struct nlsock * netlink_cmd,int family,int type,ifindex_t master_ifindex)2972 static int netlink_request_macs(struct nlsock *netlink_cmd, int family,
2973 int type, ifindex_t master_ifindex)
2974 {
2975 struct {
2976 struct nlmsghdr n;
2977 struct ifinfomsg ifm;
2978 char buf[256];
2979 } req;
2980
2981 /* Form the request, specifying filter (rtattr) if needed. */
2982 memset(&req, 0, sizeof(req));
2983 req.n.nlmsg_type = type;
2984 req.n.nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST;
2985 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ifinfomsg));
2986 req.ifm.ifi_family = family;
2987 if (master_ifindex)
2988 nl_attr_put32(&req.n, sizeof(req), IFLA_MASTER, master_ifindex);
2989
2990 return netlink_request(netlink_cmd, &req);
2991 }
2992
2993 /*
2994 * MAC forwarding database read using netlink interface. This is invoked
2995 * at startup.
2996 */
netlink_macfdb_read(struct zebra_ns * zns)2997 int netlink_macfdb_read(struct zebra_ns *zns)
2998 {
2999 int ret;
3000 struct zebra_dplane_info dp_info;
3001
3002 zebra_dplane_info_from_zns(&dp_info, zns, true /*is_cmd*/);
3003
3004 /* Get bridge FDB table. */
3005 ret = netlink_request_macs(&zns->netlink_cmd, AF_BRIDGE, RTM_GETNEIGH,
3006 0);
3007 if (ret < 0)
3008 return ret;
3009 /* We are reading entire table. */
3010 filter_vlan = 0;
3011 ret = netlink_parse_info(netlink_macfdb_table, &zns->netlink_cmd,
3012 &dp_info, 0, 1);
3013
3014 return ret;
3015 }
3016
3017 /*
3018 * MAC forwarding database read using netlink interface. This is for a
3019 * specific bridge and matching specific access VLAN (if VLAN-aware bridge).
3020 */
netlink_macfdb_read_for_bridge(struct zebra_ns * zns,struct interface * ifp,struct interface * br_if)3021 int netlink_macfdb_read_for_bridge(struct zebra_ns *zns, struct interface *ifp,
3022 struct interface *br_if)
3023 {
3024 struct zebra_if *br_zif;
3025 struct zebra_if *zif;
3026 struct zebra_l2info_vxlan *vxl;
3027 struct zebra_dplane_info dp_info;
3028 int ret = 0;
3029
3030 zebra_dplane_info_from_zns(&dp_info, zns, true /*is_cmd*/);
3031
3032 /* Save VLAN we're filtering on, if needed. */
3033 br_zif = (struct zebra_if *)br_if->info;
3034 zif = (struct zebra_if *)ifp->info;
3035 vxl = &zif->l2info.vxl;
3036 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif))
3037 filter_vlan = vxl->access_vlan;
3038
3039 /* Get bridge FDB table for specific bridge - we do the VLAN filtering.
3040 */
3041 ret = netlink_request_macs(&zns->netlink_cmd, AF_BRIDGE, RTM_GETNEIGH,
3042 br_if->ifindex);
3043 if (ret < 0)
3044 return ret;
3045 ret = netlink_parse_info(netlink_macfdb_table, &zns->netlink_cmd,
3046 &dp_info, 0, 0);
3047
3048 /* Reset VLAN filter. */
3049 filter_vlan = 0;
3050 return ret;
3051 }
3052
3053
3054 /* Request for MAC FDB for a specific MAC address in VLAN from the kernel */
netlink_request_specific_mac_in_bridge(struct zebra_ns * zns,int family,int type,struct interface * br_if,struct ethaddr * mac,vlanid_t vid)3055 static int netlink_request_specific_mac_in_bridge(struct zebra_ns *zns,
3056 int family,
3057 int type,
3058 struct interface *br_if,
3059 struct ethaddr *mac,
3060 vlanid_t vid)
3061 {
3062 struct {
3063 struct nlmsghdr n;
3064 struct ndmsg ndm;
3065 char buf[256];
3066 } req;
3067 struct zebra_if *br_zif;
3068 char buf[ETHER_ADDR_STRLEN];
3069
3070 memset(&req, 0, sizeof(req));
3071 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
3072 req.n.nlmsg_type = type; /* RTM_GETNEIGH */
3073 req.n.nlmsg_flags = NLM_F_REQUEST;
3074 req.ndm.ndm_family = family; /* AF_BRIDGE */
3075 /* req.ndm.ndm_state = NUD_REACHABLE; */
3076
3077 nl_attr_put(&req.n, sizeof(req), NDA_LLADDR, mac, 6);
3078
3079 br_zif = (struct zebra_if *)br_if->info;
3080 if (IS_ZEBRA_IF_BRIDGE_VLAN_AWARE(br_zif) && vid > 0)
3081 nl_attr_put16(&req.n, sizeof(req), NDA_VLAN, vid);
3082
3083 nl_attr_put32(&req.n, sizeof(req), NDA_MASTER, br_if->ifindex);
3084
3085 if (IS_ZEBRA_DEBUG_KERNEL)
3086 zlog_debug(
3087 "%s: Tx family %s IF %s(%u) vrf %s(%u) MAC %s vid %u",
3088 __func__, nl_family_to_str(req.ndm.ndm_family),
3089 br_if->name, br_if->ifindex,
3090 vrf_id_to_name(br_if->vrf_id), br_if->vrf_id,
3091 prefix_mac2str(mac, buf, sizeof(buf)), vid);
3092
3093 return netlink_request(&zns->netlink_cmd, &req);
3094 }
3095
netlink_macfdb_read_specific_mac(struct zebra_ns * zns,struct interface * br_if,struct ethaddr * mac,vlanid_t vid)3096 int netlink_macfdb_read_specific_mac(struct zebra_ns *zns,
3097 struct interface *br_if,
3098 struct ethaddr *mac, vlanid_t vid)
3099 {
3100 int ret = 0;
3101 struct zebra_dplane_info dp_info;
3102
3103 zebra_dplane_info_from_zns(&dp_info, zns, true /*is_cmd*/);
3104
3105 /* Get bridge FDB table for specific bridge - we do the VLAN filtering.
3106 */
3107 ret = netlink_request_specific_mac_in_bridge(zns, AF_BRIDGE,
3108 RTM_GETNEIGH,
3109 br_if, mac, vid);
3110 if (ret < 0)
3111 return ret;
3112
3113 ret = netlink_parse_info(netlink_macfdb_table, &zns->netlink_cmd,
3114 &dp_info, 1, 0);
3115
3116 return ret;
3117 }
3118
3119 /*
3120 * Netlink-specific handler for MAC updates using dataplane context object.
3121 */
netlink_macfdb_update_ctx(struct zebra_dplane_ctx * ctx,void * data,size_t datalen)3122 ssize_t netlink_macfdb_update_ctx(struct zebra_dplane_ctx *ctx, void *data,
3123 size_t datalen)
3124 {
3125 struct ipaddr vtep_ip;
3126 vlanid_t vid;
3127 ssize_t total;
3128 int cmd;
3129 uint8_t flags;
3130 uint16_t state;
3131 uint32_t nhg_id;
3132 uint32_t update_flags;
3133 bool nfy = false;
3134 uint8_t nfy_flags = 0;
3135
3136 cmd = dplane_ctx_get_op(ctx) == DPLANE_OP_MAC_INSTALL
3137 ? RTM_NEWNEIGH : RTM_DELNEIGH;
3138
3139 flags = NTF_MASTER;
3140 state = NUD_REACHABLE;
3141
3142 update_flags = dplane_ctx_mac_get_update_flags(ctx);
3143 if (update_flags & DPLANE_MAC_REMOTE) {
3144 flags |= NTF_SELF;
3145 if (dplane_ctx_mac_is_sticky(ctx)) {
3146 /* NUD_NOARP prevents the entry from expiring */
3147 state |= NUD_NOARP;
3148 /* sticky the entry from moving */
3149 flags |= NTF_STICKY;
3150 } else {
3151 flags |= NTF_EXT_LEARNED;
3152 }
3153 /* if it was static-local previously we need to clear the
3154 * notify flags on replace with remote
3155 */
3156 if (update_flags & DPLANE_MAC_WAS_STATIC)
3157 nfy = true;
3158 } else {
3159 /* local mac */
3160 if (update_flags & DPLANE_MAC_SET_STATIC) {
3161 nfy_flags |= (1 << BR_FDB_NFY_STATIC);
3162 state |= NUD_NOARP;
3163 }
3164
3165 if (update_flags & DPLANE_MAC_SET_INACTIVE)
3166 nfy_flags |= (1 << BR_FDB_NFY_INACTIVE);
3167
3168 nfy = true;
3169 }
3170
3171 nhg_id = dplane_ctx_mac_get_nhg_id(ctx);
3172 vtep_ip.ipaddr_v4 = *(dplane_ctx_mac_get_vtep_ip(ctx));
3173 SET_IPADDR_V4(&vtep_ip);
3174
3175 if (IS_ZEBRA_DEBUG_KERNEL) {
3176 char ipbuf[PREFIX_STRLEN];
3177 char buf[ETHER_ADDR_STRLEN];
3178 char vid_buf[20];
3179 const struct ethaddr *mac = dplane_ctx_mac_get_addr(ctx);
3180
3181 vid = dplane_ctx_mac_get_vlan(ctx);
3182 if (vid > 0)
3183 snprintf(vid_buf, sizeof(vid_buf), " VLAN %u", vid);
3184 else
3185 vid_buf[0] = '\0';
3186
3187 zlog_debug("Tx %s family %s IF %s(%u)%s %sMAC %s dst %s nhg %u%s%s%s%s%s",
3188 nl_msg_type_to_str(cmd), nl_family_to_str(AF_BRIDGE),
3189 dplane_ctx_get_ifname(ctx),
3190 dplane_ctx_get_ifindex(ctx), vid_buf,
3191 dplane_ctx_mac_is_sticky(ctx) ? "sticky " : "",
3192 prefix_mac2str(mac, buf, sizeof(buf)),
3193 ipaddr2str(&vtep_ip, ipbuf, sizeof(ipbuf)),
3194 nhg_id,
3195 (update_flags &
3196 DPLANE_MAC_REMOTE) ? " rem" : "",
3197 (update_flags &
3198 DPLANE_MAC_WAS_STATIC) ? " clr_sync" : "",
3199 (update_flags &
3200 DPLANE_MAC_SET_STATIC) ? " static" : "",
3201 (update_flags &
3202 DPLANE_MAC_SET_INACTIVE) ? " inactive" : "",
3203 (nfy &
3204 DPLANE_MAC_SET_INACTIVE) ? " nfy" : "");
3205 }
3206
3207 total = netlink_neigh_update_msg_encode(
3208 ctx, cmd, dplane_ctx_mac_get_addr(ctx), &vtep_ip, true,
3209 AF_BRIDGE, 0, flags, state, nhg_id, nfy, nfy_flags,
3210 data, datalen);
3211
3212 return total;
3213 }
3214
3215 /*
3216 * In the event the kernel deletes ipv4 link-local neighbor entries created for
3217 * 5549 support, re-install them.
3218 */
netlink_handle_5549(struct ndmsg * ndm,struct zebra_if * zif,struct interface * ifp,struct ipaddr * ip,bool handle_failed)3219 static void netlink_handle_5549(struct ndmsg *ndm, struct zebra_if *zif,
3220 struct interface *ifp, struct ipaddr *ip,
3221 bool handle_failed)
3222 {
3223 if (ndm->ndm_family != AF_INET)
3224 return;
3225
3226 if (!zif->v6_2_v4_ll_neigh_entry)
3227 return;
3228
3229 if (ipv4_ll.s_addr != ip->ip._v4_addr.s_addr)
3230 return;
3231
3232 if (handle_failed && ndm->ndm_state & NUD_FAILED) {
3233 zlog_info("Neighbor Entry for %s has entered a failed state, not reinstalling",
3234 ifp->name);
3235 return;
3236 }
3237
3238 if_nbr_ipv6ll_to_ipv4ll_neigh_update(ifp, &zif->v6_2_v4_ll_addr6, true);
3239 }
3240
3241 #define NUD_VALID \
3242 (NUD_PERMANENT | NUD_NOARP | NUD_REACHABLE | NUD_PROBE | NUD_STALE \
3243 | NUD_DELAY)
3244 #define NUD_LOCAL_ACTIVE \
3245 (NUD_PERMANENT | NUD_NOARP | NUD_REACHABLE)
3246
netlink_ipneigh_change(struct nlmsghdr * h,int len,ns_id_t ns_id)3247 static int netlink_ipneigh_change(struct nlmsghdr *h, int len, ns_id_t ns_id)
3248 {
3249 struct ndmsg *ndm;
3250 struct interface *ifp;
3251 struct zebra_if *zif;
3252 struct rtattr *tb[NDA_MAX + 1];
3253 struct interface *link_if;
3254 struct ethaddr mac;
3255 struct ipaddr ip;
3256 struct vrf *vrf;
3257 char buf[ETHER_ADDR_STRLEN];
3258 char buf2[INET6_ADDRSTRLEN];
3259 int mac_present = 0;
3260 bool is_ext;
3261 bool is_router;
3262 bool local_inactive;
3263
3264 ndm = NLMSG_DATA(h);
3265
3266 /* The interface should exist. */
3267 ifp = if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id),
3268 ndm->ndm_ifindex);
3269 if (!ifp || !ifp->info)
3270 return 0;
3271
3272 vrf = vrf_lookup_by_id(ifp->vrf_id);
3273 zif = (struct zebra_if *)ifp->info;
3274
3275 /* Parse attributes and extract fields of interest. */
3276 memset(tb, 0, sizeof(tb));
3277 netlink_parse_rtattr(tb, NDA_MAX, NDA_RTA(ndm), len);
3278
3279 if (!tb[NDA_DST]) {
3280 zlog_debug("%s family %s IF %s(%u) vrf %s(%u) - no DST",
3281 nl_msg_type_to_str(h->nlmsg_type),
3282 nl_family_to_str(ndm->ndm_family), ifp->name,
3283 ndm->ndm_ifindex, VRF_LOGNAME(vrf), ifp->vrf_id);
3284 return 0;
3285 }
3286
3287 memset(&ip, 0, sizeof(struct ipaddr));
3288 ip.ipa_type = (ndm->ndm_family == AF_INET) ? IPADDR_V4 : IPADDR_V6;
3289 memcpy(&ip.ip.addr, RTA_DATA(tb[NDA_DST]), RTA_PAYLOAD(tb[NDA_DST]));
3290
3291 /* if kernel deletes our rfc5549 neighbor entry, re-install it */
3292 if (h->nlmsg_type == RTM_DELNEIGH && (ndm->ndm_state & NUD_PERMANENT)) {
3293 netlink_handle_5549(ndm, zif, ifp, &ip, false);
3294 if (IS_ZEBRA_DEBUG_KERNEL)
3295 zlog_debug(
3296 "\tNeighbor Entry Received is a 5549 entry, finished");
3297 return 0;
3298 }
3299
3300 /* if kernel marks our rfc5549 neighbor entry invalid, re-install it */
3301 if (h->nlmsg_type == RTM_NEWNEIGH && !(ndm->ndm_state & NUD_VALID))
3302 netlink_handle_5549(ndm, zif, ifp, &ip, true);
3303
3304 /* The neighbor is present on an SVI. From this, we locate the
3305 * underlying
3306 * bridge because we're only interested in neighbors on a VxLAN bridge.
3307 * The bridge is located based on the nature of the SVI:
3308 * (a) In the case of a VLAN-aware bridge, the SVI is a L3 VLAN
3309 * interface
3310 * and is linked to the bridge
3311 * (b) In the case of a VLAN-unaware bridge, the SVI is the bridge
3312 * inteface
3313 * itself
3314 */
3315 if (IS_ZEBRA_IF_VLAN(ifp)) {
3316 link_if = if_lookup_by_index_per_ns(zebra_ns_lookup(ns_id),
3317 zif->link_ifindex);
3318 if (!link_if)
3319 return 0;
3320 } else if (IS_ZEBRA_IF_BRIDGE(ifp))
3321 link_if = ifp;
3322 else {
3323 if (IS_ZEBRA_DEBUG_KERNEL)
3324 zlog_debug(
3325 "\tNeighbor Entry received is not on a VLAN or a BRIDGE, ignoring");
3326 return 0;
3327 }
3328
3329 memset(&mac, 0, sizeof(struct ethaddr));
3330 if (h->nlmsg_type == RTM_NEWNEIGH) {
3331 if (tb[NDA_LLADDR]) {
3332 if (RTA_PAYLOAD(tb[NDA_LLADDR]) != ETH_ALEN) {
3333 if (IS_ZEBRA_DEBUG_KERNEL)
3334 zlog_debug(
3335 "%s family %s IF %s(%u) vrf %s(%u) - LLADDR is not MAC, len %lu",
3336 nl_msg_type_to_str(
3337 h->nlmsg_type),
3338 nl_family_to_str(
3339 ndm->ndm_family),
3340 ifp->name, ndm->ndm_ifindex,
3341 VRF_LOGNAME(vrf), ifp->vrf_id,
3342 (unsigned long)RTA_PAYLOAD(
3343 tb[NDA_LLADDR]));
3344 return 0;
3345 }
3346
3347 mac_present = 1;
3348 memcpy(&mac, RTA_DATA(tb[NDA_LLADDR]), ETH_ALEN);
3349 }
3350
3351 is_ext = !!(ndm->ndm_flags & NTF_EXT_LEARNED);
3352 is_router = !!(ndm->ndm_flags & NTF_ROUTER);
3353
3354 if (IS_ZEBRA_DEBUG_KERNEL)
3355 zlog_debug(
3356 "Rx %s family %s IF %s(%u) vrf %s(%u) IP %s MAC %s state 0x%x flags 0x%x",
3357 nl_msg_type_to_str(h->nlmsg_type),
3358 nl_family_to_str(ndm->ndm_family), ifp->name,
3359 ndm->ndm_ifindex, VRF_LOGNAME(vrf), ifp->vrf_id,
3360 ipaddr2str(&ip, buf2, sizeof(buf2)),
3361 mac_present
3362 ? prefix_mac2str(&mac, buf, sizeof(buf))
3363 : "",
3364 ndm->ndm_state, ndm->ndm_flags);
3365
3366 /* If the neighbor state is valid for use, process as an add or
3367 * update
3368 * else process as a delete. Note that the delete handling may
3369 * result
3370 * in re-adding the neighbor if it is a valid "remote" neighbor.
3371 */
3372 if (ndm->ndm_state & NUD_VALID) {
3373 local_inactive = !(ndm->ndm_state & NUD_LOCAL_ACTIVE);
3374
3375 /* XXX - populate dp-static based on the sync flags
3376 * in the kernel
3377 */
3378 return zebra_vxlan_handle_kernel_neigh_update(
3379 ifp, link_if, &ip, &mac, ndm->ndm_state,
3380 is_ext, is_router, local_inactive,
3381 false /* dp_static */);
3382 }
3383
3384 return zebra_vxlan_handle_kernel_neigh_del(ifp, link_if, &ip);
3385 }
3386
3387 if (IS_ZEBRA_DEBUG_KERNEL)
3388 zlog_debug("Rx %s family %s IF %s(%u) vrf %s(%u) IP %s",
3389 nl_msg_type_to_str(h->nlmsg_type),
3390 nl_family_to_str(ndm->ndm_family), ifp->name,
3391 ndm->ndm_ifindex, VRF_LOGNAME(vrf), ifp->vrf_id,
3392 ipaddr2str(&ip, buf2, sizeof(buf2)));
3393
3394 /* Process the delete - it may result in re-adding the neighbor if it is
3395 * a valid "remote" neighbor.
3396 */
3397 return zebra_vxlan_handle_kernel_neigh_del(ifp, link_if, &ip);
3398 }
3399
netlink_neigh_table(struct nlmsghdr * h,ns_id_t ns_id,int startup)3400 static int netlink_neigh_table(struct nlmsghdr *h, ns_id_t ns_id, int startup)
3401 {
3402 int len;
3403 struct ndmsg *ndm;
3404
3405 if (h->nlmsg_type != RTM_NEWNEIGH)
3406 return 0;
3407
3408 /* Length validity. */
3409 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct ndmsg));
3410 if (len < 0)
3411 return -1;
3412
3413 /* We are interested only in AF_INET or AF_INET6 notifications. */
3414 ndm = NLMSG_DATA(h);
3415 if (ndm->ndm_family != AF_INET && ndm->ndm_family != AF_INET6)
3416 return 0;
3417
3418 return netlink_neigh_change(h, len);
3419 }
3420
3421 /* Request for IP neighbor information from the kernel */
netlink_request_neigh(struct nlsock * netlink_cmd,int family,int type,ifindex_t ifindex)3422 static int netlink_request_neigh(struct nlsock *netlink_cmd, int family,
3423 int type, ifindex_t ifindex)
3424 {
3425 struct {
3426 struct nlmsghdr n;
3427 struct ndmsg ndm;
3428 char buf[256];
3429 } req;
3430
3431 /* Form the request, specifying filter (rtattr) if needed. */
3432 memset(&req, 0, sizeof(req));
3433 req.n.nlmsg_type = type;
3434 req.n.nlmsg_flags = NLM_F_ROOT | NLM_F_MATCH | NLM_F_REQUEST;
3435 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
3436 req.ndm.ndm_family = family;
3437 if (ifindex)
3438 nl_attr_put32(&req.n, sizeof(req), NDA_IFINDEX, ifindex);
3439
3440 return netlink_request(netlink_cmd, &req);
3441 }
3442
3443 /*
3444 * IP Neighbor table read using netlink interface. This is invoked
3445 * at startup.
3446 */
netlink_neigh_read(struct zebra_ns * zns)3447 int netlink_neigh_read(struct zebra_ns *zns)
3448 {
3449 int ret;
3450 struct zebra_dplane_info dp_info;
3451
3452 zebra_dplane_info_from_zns(&dp_info, zns, true /*is_cmd*/);
3453
3454 /* Get IP neighbor table. */
3455 ret = netlink_request_neigh(&zns->netlink_cmd, AF_UNSPEC, RTM_GETNEIGH,
3456 0);
3457 if (ret < 0)
3458 return ret;
3459 ret = netlink_parse_info(netlink_neigh_table, &zns->netlink_cmd,
3460 &dp_info, 0, 1);
3461
3462 return ret;
3463 }
3464
3465 /*
3466 * IP Neighbor table read using netlink interface. This is for a specific
3467 * VLAN device.
3468 */
netlink_neigh_read_for_vlan(struct zebra_ns * zns,struct interface * vlan_if)3469 int netlink_neigh_read_for_vlan(struct zebra_ns *zns, struct interface *vlan_if)
3470 {
3471 int ret = 0;
3472 struct zebra_dplane_info dp_info;
3473
3474 zebra_dplane_info_from_zns(&dp_info, zns, true /*is_cmd*/);
3475
3476 ret = netlink_request_neigh(&zns->netlink_cmd, AF_UNSPEC, RTM_GETNEIGH,
3477 vlan_if->ifindex);
3478 if (ret < 0)
3479 return ret;
3480 ret = netlink_parse_info(netlink_neigh_table, &zns->netlink_cmd,
3481 &dp_info, 0, 0);
3482
3483 return ret;
3484 }
3485
3486 /*
3487 * Request for a specific IP in VLAN (SVI) device from IP Neighbor table,
3488 * read using netlink interface.
3489 */
netlink_request_specific_neigh_in_vlan(struct zebra_ns * zns,int type,struct ipaddr * ip,ifindex_t ifindex)3490 static int netlink_request_specific_neigh_in_vlan(struct zebra_ns *zns,
3491 int type, struct ipaddr *ip,
3492 ifindex_t ifindex)
3493 {
3494 struct {
3495 struct nlmsghdr n;
3496 struct ndmsg ndm;
3497 char buf[256];
3498 } req;
3499 int ipa_len;
3500
3501 /* Form the request, specifying filter (rtattr) if needed. */
3502 memset(&req, 0, sizeof(req));
3503 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct ndmsg));
3504 req.n.nlmsg_flags = NLM_F_REQUEST;
3505 req.n.nlmsg_type = type; /* RTM_GETNEIGH */
3506 req.ndm.ndm_ifindex = ifindex;
3507
3508 if (IS_IPADDR_V4(ip)) {
3509 ipa_len = IPV4_MAX_BYTELEN;
3510 req.ndm.ndm_family = AF_INET;
3511
3512 } else {
3513 ipa_len = IPV6_MAX_BYTELEN;
3514 req.ndm.ndm_family = AF_INET6;
3515 }
3516
3517 nl_attr_put(&req.n, sizeof(req), NDA_DST, &ip->ip.addr, ipa_len);
3518
3519 if (IS_ZEBRA_DEBUG_KERNEL) {
3520 char buf[INET6_ADDRSTRLEN];
3521
3522 zlog_debug("%s: Tx %s family %s IF %u IP %s flags 0x%x",
3523 __func__, nl_msg_type_to_str(type),
3524 nl_family_to_str(req.ndm.ndm_family), ifindex,
3525 ipaddr2str(ip, buf, sizeof(buf)), req.n.nlmsg_flags);
3526 }
3527
3528 return netlink_request(&zns->netlink_cmd, &req);
3529 }
3530
netlink_neigh_read_specific_ip(struct ipaddr * ip,struct interface * vlan_if)3531 int netlink_neigh_read_specific_ip(struct ipaddr *ip,
3532 struct interface *vlan_if)
3533 {
3534 int ret = 0;
3535 struct zebra_ns *zns;
3536 struct zebra_vrf *zvrf = zebra_vrf_lookup_by_id(vlan_if->vrf_id);
3537 char buf[INET6_ADDRSTRLEN];
3538 struct zebra_dplane_info dp_info;
3539
3540 zns = zvrf->zns;
3541
3542 zebra_dplane_info_from_zns(&dp_info, zns, true /*is_cmd*/);
3543
3544 if (IS_ZEBRA_DEBUG_KERNEL)
3545 zlog_debug("%s: neigh request IF %s(%u) IP %s vrf %s(%u)",
3546 __func__, vlan_if->name, vlan_if->ifindex,
3547 ipaddr2str(ip, buf, sizeof(buf)),
3548 vrf_id_to_name(vlan_if->vrf_id), vlan_if->vrf_id);
3549
3550 ret = netlink_request_specific_neigh_in_vlan(zns, RTM_GETNEIGH, ip,
3551 vlan_if->ifindex);
3552 if (ret < 0)
3553 return ret;
3554
3555 ret = netlink_parse_info(netlink_neigh_table, &zns->netlink_cmd,
3556 &dp_info, 1, 0);
3557
3558 return ret;
3559 }
3560
netlink_neigh_change(struct nlmsghdr * h,ns_id_t ns_id)3561 int netlink_neigh_change(struct nlmsghdr *h, ns_id_t ns_id)
3562 {
3563 int len;
3564 struct ndmsg *ndm;
3565
3566 if (!(h->nlmsg_type == RTM_NEWNEIGH || h->nlmsg_type == RTM_DELNEIGH))
3567 return 0;
3568
3569 /* Length validity. */
3570 len = h->nlmsg_len - NLMSG_LENGTH(sizeof(struct ndmsg));
3571 if (len < 0) {
3572 zlog_err(
3573 "%s: Message received from netlink is of a broken size %d %zu",
3574 __func__, h->nlmsg_len,
3575 (size_t)NLMSG_LENGTH(sizeof(struct ndmsg)));
3576 return -1;
3577 }
3578
3579 /* Is this a notification for the MAC FDB or IP neighbor table? */
3580 ndm = NLMSG_DATA(h);
3581 if (ndm->ndm_family == AF_BRIDGE)
3582 return netlink_macfdb_change(h, len, ns_id);
3583
3584 if (ndm->ndm_type != RTN_UNICAST)
3585 return 0;
3586
3587 if (ndm->ndm_family == AF_INET || ndm->ndm_family == AF_INET6)
3588 return netlink_ipneigh_change(h, len, ns_id);
3589 else {
3590 flog_warn(
3591 EC_ZEBRA_UNKNOWN_FAMILY,
3592 "Invalid address family: %u received from kernel neighbor change: %s",
3593 ndm->ndm_family, nl_msg_type_to_str(h->nlmsg_type));
3594 return 0;
3595 }
3596
3597 return 0;
3598 }
3599
3600 /*
3601 * Utility neighbor-update function, using info from dplane context.
3602 */
netlink_neigh_update_ctx(const struct zebra_dplane_ctx * ctx,int cmd,void * buf,size_t buflen)3603 static ssize_t netlink_neigh_update_ctx(const struct zebra_dplane_ctx *ctx,
3604 int cmd, void *buf, size_t buflen)
3605 {
3606 const struct ipaddr *ip;
3607 const struct ethaddr *mac;
3608 uint8_t flags;
3609 uint16_t state;
3610 uint8_t family;
3611
3612 ip = dplane_ctx_neigh_get_ipaddr(ctx);
3613 mac = dplane_ctx_neigh_get_mac(ctx);
3614 if (is_zero_mac(mac))
3615 mac = NULL;
3616
3617 flags = neigh_flags_to_netlink(dplane_ctx_neigh_get_flags(ctx));
3618 state = neigh_state_to_netlink(dplane_ctx_neigh_get_state(ctx));
3619
3620 family = IS_IPADDR_V4(ip) ? AF_INET : AF_INET6;
3621
3622 if (IS_ZEBRA_DEBUG_KERNEL) {
3623 char buf[INET6_ADDRSTRLEN];
3624 char buf2[ETHER_ADDR_STRLEN];
3625
3626 zlog_debug(
3627 "Tx %s family %s IF %s(%u) Neigh %s MAC %s flags 0x%x state 0x%x",
3628 nl_msg_type_to_str(cmd), nl_family_to_str(family),
3629 dplane_ctx_get_ifname(ctx), dplane_ctx_get_ifindex(ctx),
3630 ipaddr2str(ip, buf, sizeof(buf)),
3631 mac ? prefix_mac2str(mac, buf2, sizeof(buf2)) : "null",
3632 flags, state);
3633 }
3634
3635 return netlink_neigh_update_msg_encode(
3636 ctx, cmd, mac, ip, true, family, RTN_UNICAST, flags, state,
3637 0 /*nhg*/, false /*nfy*/, 0 /*nfy_flags*/, buf, buflen);
3638 }
3639
netlink_neigh_msg_encoder(struct zebra_dplane_ctx * ctx,void * buf,size_t buflen)3640 static ssize_t netlink_neigh_msg_encoder(struct zebra_dplane_ctx *ctx,
3641 void *buf, size_t buflen)
3642 {
3643 ssize_t ret;
3644
3645 switch (dplane_ctx_get_op(ctx)) {
3646 case DPLANE_OP_NEIGH_INSTALL:
3647 case DPLANE_OP_NEIGH_UPDATE:
3648 case DPLANE_OP_NEIGH_DISCOVER:
3649 ret = netlink_neigh_update_ctx(ctx, RTM_NEWNEIGH, buf, buflen);
3650 break;
3651 case DPLANE_OP_NEIGH_DELETE:
3652 ret = netlink_neigh_update_ctx(ctx, RTM_DELNEIGH, buf, buflen);
3653 break;
3654 case DPLANE_OP_VTEP_ADD:
3655 ret = netlink_vxlan_flood_update_ctx(ctx, RTM_NEWNEIGH, buf,
3656 buflen);
3657 break;
3658 case DPLANE_OP_VTEP_DELETE:
3659 ret = netlink_vxlan_flood_update_ctx(ctx, RTM_DELNEIGH, buf,
3660 buflen);
3661 break;
3662 default:
3663 ret = -1;
3664 }
3665
3666 return ret;
3667 }
3668
3669 /*
3670 * Update MAC, using dataplane context object.
3671 */
3672
netlink_put_mac_update_msg(struct nl_batch * bth,struct zebra_dplane_ctx * ctx)3673 enum netlink_msg_status netlink_put_mac_update_msg(struct nl_batch *bth,
3674 struct zebra_dplane_ctx *ctx)
3675 {
3676 return netlink_batch_add_msg(bth, ctx, netlink_macfdb_update_ctx,
3677 false);
3678 }
3679
3680 enum netlink_msg_status
netlink_put_neigh_update_msg(struct nl_batch * bth,struct zebra_dplane_ctx * ctx)3681 netlink_put_neigh_update_msg(struct nl_batch *bth, struct zebra_dplane_ctx *ctx)
3682 {
3683 return netlink_batch_add_msg(bth, ctx, netlink_neigh_msg_encoder,
3684 false);
3685 }
3686
3687 /*
3688 * MPLS label forwarding table change via netlink interface, using dataplane
3689 * context information.
3690 */
netlink_mpls_multipath_msg_encode(int cmd,struct zebra_dplane_ctx * ctx,void * buf,size_t buflen)3691 ssize_t netlink_mpls_multipath_msg_encode(int cmd, struct zebra_dplane_ctx *ctx,
3692 void *buf, size_t buflen)
3693 {
3694 mpls_lse_t lse;
3695 const struct nhlfe_list_head *head;
3696 const zebra_nhlfe_t *nhlfe;
3697 struct nexthop *nexthop = NULL;
3698 unsigned int nexthop_num;
3699 const char *routedesc;
3700 int route_type;
3701 struct prefix p = {0};
3702
3703 struct {
3704 struct nlmsghdr n;
3705 struct rtmsg r;
3706 char buf[0];
3707 } *req = buf;
3708
3709 if (buflen < sizeof(*req))
3710 return 0;
3711
3712 memset(req, 0, sizeof(*req));
3713
3714 /*
3715 * Count # nexthops so we can decide whether to use singlepath
3716 * or multipath case.
3717 */
3718 nexthop_num = 0;
3719 head = dplane_ctx_get_nhlfe_list(ctx);
3720 frr_each(nhlfe_list_const, head, nhlfe) {
3721 nexthop = nhlfe->nexthop;
3722 if (!nexthop)
3723 continue;
3724 if (cmd == RTM_NEWROUTE) {
3725 /* Count all selected NHLFEs */
3726 if (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_SELECTED)
3727 && CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_ACTIVE))
3728 nexthop_num++;
3729 } else { /* DEL */
3730 /* Count all installed NHLFEs */
3731 if (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_INSTALLED)
3732 && CHECK_FLAG(nexthop->flags, NEXTHOP_FLAG_FIB))
3733 nexthop_num++;
3734 }
3735 }
3736
3737 if ((nexthop_num == 0) ||
3738 (!dplane_ctx_get_best_nhlfe(ctx) && (cmd != RTM_DELROUTE)))
3739 return 0;
3740
3741 req->n.nlmsg_len = NLMSG_LENGTH(sizeof(struct rtmsg));
3742 req->n.nlmsg_flags = NLM_F_CREATE | NLM_F_REQUEST;
3743 req->n.nlmsg_type = cmd;
3744 req->n.nlmsg_pid = dplane_ctx_get_ns(ctx)->nls.snl.nl_pid;
3745
3746 req->r.rtm_family = AF_MPLS;
3747 req->r.rtm_table = RT_TABLE_MAIN;
3748 req->r.rtm_dst_len = MPLS_LABEL_LEN_BITS;
3749 req->r.rtm_scope = RT_SCOPE_UNIVERSE;
3750 req->r.rtm_type = RTN_UNICAST;
3751
3752 if (cmd == RTM_NEWROUTE) {
3753 /* We do a replace to handle update. */
3754 req->n.nlmsg_flags |= NLM_F_REPLACE;
3755
3756 /* set the protocol value if installing */
3757 route_type = re_type_from_lsp_type(
3758 dplane_ctx_get_best_nhlfe(ctx)->type);
3759 req->r.rtm_protocol = zebra2proto(route_type);
3760 }
3761
3762 /* Fill destination */
3763 lse = mpls_lse_encode(dplane_ctx_get_in_label(ctx), 0, 0, 1);
3764 if (!nl_attr_put(&req->n, buflen, RTA_DST, &lse, sizeof(mpls_lse_t)))
3765 return 0;
3766
3767 /* Fill nexthops (paths) based on single-path or multipath. The paths
3768 * chosen depend on the operation.
3769 */
3770 if (nexthop_num == 1) {
3771 routedesc = "single-path";
3772 _netlink_mpls_debug(cmd, dplane_ctx_get_in_label(ctx),
3773 routedesc);
3774
3775 nexthop_num = 0;
3776 frr_each(nhlfe_list_const, head, nhlfe) {
3777 nexthop = nhlfe->nexthop;
3778 if (!nexthop)
3779 continue;
3780
3781 if ((cmd == RTM_NEWROUTE
3782 && (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_SELECTED)
3783 && CHECK_FLAG(nexthop->flags,
3784 NEXTHOP_FLAG_ACTIVE)))
3785 || (cmd == RTM_DELROUTE
3786 && (CHECK_FLAG(nhlfe->flags,
3787 NHLFE_FLAG_INSTALLED)
3788 && CHECK_FLAG(nexthop->flags,
3789 NEXTHOP_FLAG_FIB)))) {
3790 /* Add the gateway */
3791 if (!_netlink_mpls_build_singlepath(
3792 &p, routedesc, nhlfe, &req->n,
3793 &req->r, buflen, cmd))
3794 return false;
3795
3796 nexthop_num++;
3797 break;
3798 }
3799 }
3800 } else { /* Multipath case */
3801 struct rtattr *nest;
3802 const union g_addr *src1 = NULL;
3803
3804 nest = nl_attr_nest(&req->n, buflen, RTA_MULTIPATH);
3805 if (!nest)
3806 return 0;
3807
3808 routedesc = "multipath";
3809 _netlink_mpls_debug(cmd, dplane_ctx_get_in_label(ctx),
3810 routedesc);
3811
3812 nexthop_num = 0;
3813 frr_each(nhlfe_list_const, head, nhlfe) {
3814 nexthop = nhlfe->nexthop;
3815 if (!nexthop)
3816 continue;
3817
3818 if ((cmd == RTM_NEWROUTE
3819 && (CHECK_FLAG(nhlfe->flags, NHLFE_FLAG_SELECTED)
3820 && CHECK_FLAG(nexthop->flags,
3821 NEXTHOP_FLAG_ACTIVE)))
3822 || (cmd == RTM_DELROUTE
3823 && (CHECK_FLAG(nhlfe->flags,
3824 NHLFE_FLAG_INSTALLED)
3825 && CHECK_FLAG(nexthop->flags,
3826 NEXTHOP_FLAG_FIB)))) {
3827 nexthop_num++;
3828
3829 /* Build the multipath */
3830 if (!_netlink_mpls_build_multipath(
3831 &p, routedesc, nhlfe, &req->n,
3832 buflen, &req->r, &src1))
3833 return 0;
3834 }
3835 }
3836
3837 /* Add the multipath */
3838 nl_attr_nest_end(&req->n, nest);
3839 }
3840
3841 return NLMSG_ALIGN(req->n.nlmsg_len);
3842 }
3843
3844 /****************************************************************************
3845 * This code was developed in a branch that didn't have dplane APIs for
3846 * MAC updates. Hence the use of the legacy style. It will be moved to
3847 * the new dplane style pre-merge to master. XXX
3848 */
netlink_fdb_nh_update(uint32_t nh_id,struct in_addr vtep_ip)3849 static int netlink_fdb_nh_update(uint32_t nh_id, struct in_addr vtep_ip)
3850 {
3851 struct {
3852 struct nlmsghdr n;
3853 struct nhmsg nhm;
3854 char buf[256];
3855 } req;
3856 int cmd = RTM_NEWNEXTHOP;
3857 struct zebra_vrf *zvrf;
3858 struct zebra_ns *zns;
3859
3860 zvrf = zebra_vrf_get_evpn();
3861 if (!zvrf)
3862 return -1;
3863 zns = zvrf->zns;
3864
3865 memset(&req, 0, sizeof(req));
3866
3867 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct nhmsg));
3868 req.n.nlmsg_flags = NLM_F_REQUEST;
3869 req.n.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
3870 req.n.nlmsg_type = cmd;
3871 req.nhm.nh_family = AF_INET;
3872
3873 if (!nl_attr_put32(&req.n, sizeof(req), NHA_ID, nh_id))
3874 return -1;
3875 if (!nl_attr_put(&req.n, sizeof(req), NHA_FDB, NULL, 0))
3876 return -1;
3877 if (!nl_attr_put(&req.n, sizeof(req), NHA_GATEWAY,
3878 &vtep_ip, IPV4_MAX_BYTELEN))
3879 return -1;
3880
3881 if (IS_ZEBRA_DEBUG_KERNEL || IS_ZEBRA_DEBUG_EVPN_MH_NH) {
3882 zlog_debug("Tx %s fdb-nh 0x%x %s",
3883 nl_msg_type_to_str(cmd), nh_id, inet_ntoa(vtep_ip));
3884 }
3885
3886 return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
3887 0);
3888 }
3889
netlink_fdb_nh_del(uint32_t nh_id)3890 static int netlink_fdb_nh_del(uint32_t nh_id)
3891 {
3892 struct {
3893 struct nlmsghdr n;
3894 struct nhmsg nhm;
3895 char buf[256];
3896 } req;
3897 int cmd = RTM_DELNEXTHOP;
3898 struct zebra_vrf *zvrf;
3899 struct zebra_ns *zns;
3900
3901 zvrf = zebra_vrf_get_evpn();
3902 if (!zvrf)
3903 return -1;
3904 zns = zvrf->zns;
3905
3906 memset(&req, 0, sizeof(req));
3907
3908 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct nhmsg));
3909 req.n.nlmsg_flags = NLM_F_REQUEST;
3910 req.n.nlmsg_type = cmd;
3911 req.nhm.nh_family = AF_UNSPEC;
3912
3913 if (!nl_attr_put32(&req.n, sizeof(req), NHA_ID, nh_id))
3914 return -1;
3915
3916 if (IS_ZEBRA_DEBUG_KERNEL || IS_ZEBRA_DEBUG_EVPN_MH_NH) {
3917 zlog_debug("Tx %s fdb-nh 0x%x",
3918 nl_msg_type_to_str(cmd), nh_id);
3919 }
3920
3921 return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
3922 0);
3923 }
3924
netlink_fdb_nhg_update(uint32_t nhg_id,uint32_t nh_cnt,struct nh_grp * nh_ids)3925 static int netlink_fdb_nhg_update(uint32_t nhg_id, uint32_t nh_cnt,
3926 struct nh_grp *nh_ids)
3927 {
3928 struct {
3929 struct nlmsghdr n;
3930 struct nhmsg nhm;
3931 char buf[256];
3932 } req;
3933 int cmd = RTM_NEWNEXTHOP;
3934 struct zebra_vrf *zvrf;
3935 struct zebra_ns *zns;
3936 struct nexthop_grp grp[nh_cnt];
3937 uint32_t i;
3938
3939 zvrf = zebra_vrf_get_evpn();
3940 if (!zvrf)
3941 return -1;
3942 zns = zvrf->zns;
3943
3944 memset(&req, 0, sizeof(req));
3945
3946 req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct nhmsg));
3947 req.n.nlmsg_flags = NLM_F_REQUEST;
3948 req.n.nlmsg_flags |= (NLM_F_CREATE | NLM_F_REPLACE);
3949 req.n.nlmsg_type = cmd;
3950 req.nhm.nh_family = AF_UNSPEC;
3951
3952 if (!nl_attr_put32(&req.n, sizeof(req), NHA_ID, nhg_id))
3953 return -1;
3954 if (!nl_attr_put(&req.n, sizeof(req), NHA_FDB, NULL, 0))
3955 return -1;
3956 memset(&grp, 0, sizeof(grp));
3957 for (i = 0; i < nh_cnt; ++i) {
3958 grp[i].id = nh_ids[i].id;
3959 grp[i].weight = nh_ids[i].weight;
3960 }
3961 if (!nl_attr_put(&req.n, sizeof(req), NHA_GROUP,
3962 grp, nh_cnt * sizeof(struct nexthop_grp)))
3963 return -1;
3964
3965
3966 if (IS_ZEBRA_DEBUG_KERNEL || IS_ZEBRA_DEBUG_EVPN_MH_NH) {
3967 char vtep_str[ES_VTEP_LIST_STR_SZ];
3968 char nh_buf[16];
3969
3970 vtep_str[0] = '\0';
3971 for (i = 0; i < nh_cnt; ++i) {
3972 snprintf(nh_buf, sizeof(nh_buf), "%u ",
3973 grp[i].id);
3974 strlcat(vtep_str, nh_buf, sizeof(vtep_str));
3975 }
3976
3977 zlog_debug("Tx %s fdb-nhg 0x%x %s",
3978 nl_msg_type_to_str(cmd), nhg_id, vtep_str);
3979 }
3980
3981 return netlink_talk(netlink_talk_filter, &req.n, &zns->netlink_cmd, zns,
3982 0);
3983 }
3984
netlink_fdb_nhg_del(uint32_t nhg_id)3985 static int netlink_fdb_nhg_del(uint32_t nhg_id)
3986 {
3987 return netlink_fdb_nh_del(nhg_id);
3988 }
3989
kernel_upd_mac_nh(uint32_t nh_id,struct in_addr vtep_ip)3990 int kernel_upd_mac_nh(uint32_t nh_id, struct in_addr vtep_ip)
3991 {
3992 return netlink_fdb_nh_update(nh_id, vtep_ip);
3993 }
3994
kernel_del_mac_nh(uint32_t nh_id)3995 int kernel_del_mac_nh(uint32_t nh_id)
3996 {
3997 return netlink_fdb_nh_del(nh_id);
3998 }
3999
kernel_upd_mac_nhg(uint32_t nhg_id,uint32_t nh_cnt,struct nh_grp * nh_ids)4000 int kernel_upd_mac_nhg(uint32_t nhg_id, uint32_t nh_cnt,
4001 struct nh_grp *nh_ids)
4002 {
4003 return netlink_fdb_nhg_update(nhg_id, nh_cnt, nh_ids);
4004 }
4005
kernel_del_mac_nhg(uint32_t nhg_id)4006 int kernel_del_mac_nhg(uint32_t nhg_id)
4007 {
4008 return netlink_fdb_nhg_del(nhg_id);
4009 }
4010
4011 #endif /* HAVE_NETLINK */
4012