1 /*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 2021 Ng Peng Nam Sean
5 * Copyright (c) 2022 Alexander V. Chernikov <melifaro@FreeBSD.org>
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 */
28
29 #include <sys/cdefs.h>
30 #include "opt_inet.h"
31 #include "opt_inet6.h"
32 #include <sys/types.h>
33 #include <sys/malloc.h>
34 #include <sys/rmlock.h>
35 #include <sys/socket.h>
36
37 #include <net/if.h>
38 #include <net/route.h>
39 #include <net/route/nhop.h>
40 #include <net/route/route_ctl.h>
41 #include <net/route/route_var.h>
42 #include <netinet6/scope6_var.h>
43 #include <netlink/netlink.h>
44 #include <netlink/netlink_ctl.h>
45 #include <netlink/netlink_route.h>
46 #include <netlink/route/route_var.h>
47
48 #define DEBUG_MOD_NAME nl_route
49 #define DEBUG_MAX_LEVEL LOG_DEBUG3
50 #include <netlink/netlink_debug.h>
51 _DECLARE_DEBUG(LOG_INFO);
52
53 static unsigned char
get_rtm_type(const struct nhop_object * nh)54 get_rtm_type(const struct nhop_object *nh)
55 {
56 int nh_flags = nh->nh_flags;
57
58 /* Use the fact that nhg runtime flags are only NHF_MULTIPATH */
59 if (nh_flags & NHF_BLACKHOLE)
60 return (RTN_BLACKHOLE);
61 else if (nh_flags & NHF_REJECT)
62 return (RTN_PROHIBIT);
63 return (RTN_UNICAST);
64 }
65
66 static uint8_t
nl_get_rtm_protocol(const struct nhop_object * nh)67 nl_get_rtm_protocol(const struct nhop_object *nh)
68 {
69 const struct nhgrp_object *nhg = (const struct nhgrp_object *)nh;
70 int rt_flags;
71 uint8_t origin;
72
73 if (NH_IS_NHGRP(nh)) {
74 origin = nhgrp_get_origin(nhg);
75 if (origin != RTPROT_UNSPEC)
76 return (origin);
77 nh = nhg->nhops[0];
78 }
79 origin = nhop_get_origin(nh);
80 if (origin != RTPROT_UNSPEC)
81 return (origin);
82 /* TODO: remove guesswork once all kernel users fill in origin */
83 rt_flags = nhop_get_rtflags(nh);
84 if (rt_flags & RTF_PROTO1)
85 return (RTPROT_ZEBRA);
86 if (rt_flags & RTF_STATIC)
87 return (RTPROT_STATIC);
88 return (RTPROT_KERNEL);
89 }
90
91 static int
get_rtmsg_type_from_rtsock(int cmd)92 get_rtmsg_type_from_rtsock(int cmd)
93 {
94 switch (cmd) {
95 case RTM_ADD:
96 case RTM_CHANGE:
97 case RTM_GET:
98 return NL_RTM_NEWROUTE;
99 case RTM_DELETE:
100 return NL_RTM_DELROUTE;
101 }
102
103 return (0);
104 }
105
106 /*
107 * fibnum heuristics
108 *
109 * if (dump && rtm_table == 0 && !rta_table) RT_ALL_FIBS
110 * msg rtm_table RTA_TABLE result
111 * RTM_GETROUTE/dump 0 - RT_ALL_FIBS
112 * RTM_GETROUTE/dump 1 - 1
113 * RTM_GETROUTE/get 0 - 0
114 *
115 */
116
117 static struct nhop_object *
rc_get_nhop(const struct rib_cmd_info * rc)118 rc_get_nhop(const struct rib_cmd_info *rc)
119 {
120 return ((rc->rc_cmd == RTM_DELETE) ? rc->rc_nh_old : rc->rc_nh_new);
121 }
122
123 static void
dump_rc_nhop_gw(struct nl_writer * nw,const struct nhop_object * nh)124 dump_rc_nhop_gw(struct nl_writer *nw, const struct nhop_object *nh)
125 {
126 #ifdef INET6
127 int upper_family;
128 #endif
129
130 switch (nhop_get_neigh_family(nh)) {
131 case AF_LINK:
132 /* onlink prefix, skip */
133 break;
134 case AF_INET:
135 nlattr_add(nw, NL_RTA_GATEWAY, 4, &nh->gw4_sa.sin_addr);
136 break;
137 #ifdef INET6
138 case AF_INET6:
139 upper_family = nhop_get_upper_family(nh);
140 if (upper_family == AF_INET6) {
141 struct in6_addr gw6 = nh->gw6_sa.sin6_addr;
142 in6_clearscope(&gw6);
143
144 nlattr_add(nw, NL_RTA_GATEWAY, 16, &gw6);
145 } else if (upper_family == AF_INET) {
146 /* IPv4 over IPv6 */
147 struct in6_addr gw6 = nh->gw6_sa.sin6_addr;
148 in6_clearscope(&gw6);
149
150 char buf[20];
151 struct rtvia *via = (struct rtvia *)&buf[0];
152 via->rtvia_family = AF_INET6;
153 memcpy(via->rtvia_addr, &gw6, 16);
154 nlattr_add(nw, NL_RTA_VIA, 17, via);
155 }
156 break;
157 #endif
158 }
159 }
160
161 static void
dump_rc_nhop_mtu(struct nl_writer * nw,const struct nhop_object * nh)162 dump_rc_nhop_mtu(struct nl_writer *nw, const struct nhop_object *nh)
163 {
164 int nla_len = sizeof(struct nlattr) * 2 + sizeof(uint32_t);
165 struct nlattr *nla = nlmsg_reserve_data(nw, nla_len, struct nlattr);
166
167 if (nla == NULL)
168 return;
169 nla->nla_type = NL_RTA_METRICS;
170 nla->nla_len = nla_len;
171 nla++;
172 nla->nla_type = NL_RTAX_MTU;
173 nla->nla_len = sizeof(struct nlattr) + sizeof(uint32_t);
174 *((uint32_t *)(nla + 1)) = nh->nh_mtu;
175 }
176
177 static void
dump_rc_nhg(struct nl_writer * nw,const struct route_nhop_data * rnd,struct rtmsg * rtm)178 dump_rc_nhg(struct nl_writer *nw, const struct route_nhop_data *rnd, struct rtmsg *rtm)
179 {
180 const struct nhgrp_object *nhg = rnd->rnd_nhgrp;
181 const struct weightened_nhop *wn;
182 struct nhop_object *nh;
183 uint32_t uidx, num_nhops, nh_expire;
184 uint32_t base_rtflags, rtflags, nhop_weight, nhop_metric;
185
186 MPASS((NH_IS_NHGRP(rnd->rnd_nhop)));
187
188 /* select a nhop from nhgrp to not confuse non-mpath consumers */
189 nhop_weight = RT_DEFAULT_WEIGHT;
190 nhop_metric = RT_DEFAULT_METRIC;
191 nh = nhop_select_func(rnd->rnd_nhop, 0);
192 rtflags = nhop_get_rtflags(nh);
193 if (nh->nh_flags & NHF_GATEWAY)
194 dump_rc_nhop_gw(nw, nh);
195
196 wn = nhgrp_get_nhops(nhg, &num_nhops);
197 base_rtflags = nhop_get_rtflags(wn[0].nh);
198 uidx = nhgrp_get_uidx(nhg);
199 if (uidx != 0)
200 nlattr_add_u32(nw, NL_RTA_NH_ID, uidx);
201 nlattr_add_u32(nw, NL_RTA_KNH_ID, nhgrp_get_idx(nhg));
202 nlattr_add_u32(nw, NL_RTA_RTFLAGS, base_rtflags);
203
204 if (rtflags & RTF_FIXEDMTU)
205 dump_rc_nhop_mtu(nw, nh);
206 /* In any case, fill outgoing interface */
207 nlattr_add_u32(nw, NL_RTA_OIF, if_getindex(nh->nh_ifp));
208
209 int off = nlattr_add_nested(nw, NL_RTA_MULTIPATH);
210 if (off == 0)
211 return;
212
213 for (int i = 0; i < num_nhops; i++) {
214 int nh_off = nlattr_save_offset(nw);
215 struct rtnexthop *rtnh = nlmsg_reserve_object(nw, struct rtnexthop);
216 if (rtnh == NULL)
217 return;
218 rtnh->rtnh_flags = 0;
219 rtnh->rtnh_ifindex = if_getindex(wn[i].nh->nh_ifp);
220 rtnh->rtnh_hops = MIN(wn[i].weight, UINT8_MAX);
221 dump_rc_nhop_gw(nw, wn[i].nh);
222 uint32_t rtflags = nhop_get_rtflags(wn[i].nh);
223 if (rtflags != base_rtflags)
224 nlattr_add_u32(nw, NL_RTA_RTFLAGS, rtflags);
225 if (rtflags & RTF_FIXEDMTU)
226 dump_rc_nhop_mtu(nw, wn[i].nh);
227 nlattr_add_u32(nw, NL_RTA_PRIORITY, nhop_get_metric(wn[i].nh));
228 nh_expire = nhop_get_expire(wn[i].nh);
229 if (nh_expire > 0)
230 nlattr_add_u32(nw, NL_RTA_EXPIRES, nh_expire - time_uptime);
231 rtnh = nlattr_restore_offset(nw, nh_off, struct rtnexthop);
232
233 if (nh == wn[i].nh) {
234 nhop_weight = wn[i].weight;
235 nhop_metric = nhop_get_metric(wn[i].nh);
236 }
237 /*
238 * nlattr_add() allocates 4-byte aligned storage, no need to aligh
239 * length here
240 * */
241 rtnh->rtnh_len = nlattr_save_offset(nw) - nh_off;
242 }
243 nlattr_set_len(nw, off);
244 nlattr_add_u32(nw, NL_RTA_PRIORITY, nhop_metric);
245 if (nhop_weight != RT_DEFAULT_WEIGHT)
246 nlattr_add_u32(nw, NL_RTA_WEIGHT, nhop_weight);
247 }
248
249 static void
dump_rc_nhop(struct nl_writer * nw,const struct route_nhop_data * rnd,struct rtmsg * rtm)250 dump_rc_nhop(struct nl_writer *nw, const struct route_nhop_data *rnd, struct rtmsg *rtm)
251 {
252 const struct nhop_object *nh = rnd->rnd_nhop;
253 uint32_t rtflags, uidx, nh_expire;
254
255 if (NH_IS_NHGRP(rnd->rnd_nhop)) {
256 dump_rc_nhg(nw, rnd, rtm);
257 return;
258 }
259
260 rtflags = nhop_get_rtflags(nh);
261 /*
262 * IPv4 over IPv6
263 * ('RTA_VIA', {'family': 10, 'addr': 'fe80::20c:29ff:fe67:2dd'}), ('RTA_OIF', 2),
264 * IPv4 w/ gw
265 * ('RTA_GATEWAY', '172.16.107.131'), ('RTA_OIF', 2)],
266 * Direct route:
267 * ('RTA_OIF', 2)
268 */
269 if (nh->nh_flags & NHF_GATEWAY)
270 dump_rc_nhop_gw(nw, nh);
271
272 uidx = nhop_get_uidx(nh);
273 if (uidx != 0)
274 nlattr_add_u32(nw, NL_RTA_NH_ID, uidx);
275 nlattr_add_u32(nw, NL_RTA_KNH_ID, nhop_get_idx(nh));
276 nlattr_add_u32(nw, NL_RTA_RTFLAGS, rtflags);
277
278 if (rtflags & RTF_FIXEDMTU)
279 dump_rc_nhop_mtu(nw, nh);
280 nh_expire = nhop_get_expire(nh);
281 if (nh_expire > 0)
282 nlattr_add_u32(nw, NL_RTA_EXPIRES, nh_expire - time_uptime);
283
284 /* In any case, fill outgoing interface */
285 nlattr_add_u32(nw, NL_RTA_OIF, if_getindex(nh->nh_ifp));
286
287 nlattr_add_u32(nw, NL_RTA_PRIORITY, nhop_get_metric(nh));
288 if (rnd->rnd_weight != RT_DEFAULT_WEIGHT)
289 nlattr_add_u32(nw, NL_RTA_WEIGHT, rnd->rnd_weight);
290 }
291
292 /*
293 * Dumps output from a rib command into an rtmsg
294 */
295
296 static int
dump_px(uint32_t fibnum,const struct nlmsghdr * hdr,const struct rtentry * rt,struct route_nhop_data * rnd,struct nl_writer * nw)297 dump_px(uint32_t fibnum, const struct nlmsghdr *hdr,
298 const struct rtentry *rt, struct route_nhop_data *rnd,
299 struct nl_writer *nw)
300 {
301 struct rtmsg *rtm;
302 int error = 0;
303
304 NET_EPOCH_ASSERT();
305
306 if (!nlmsg_reply(nw, hdr, sizeof(struct rtmsg)))
307 goto enomem;
308
309 int family = rt_get_family(rt);
310 int rtm_off = nlattr_save_offset(nw);
311 rtm = nlmsg_reserve_object(nw, struct rtmsg);
312 rtm->rtm_family = family;
313 rtm->rtm_dst_len = 0;
314 rtm->rtm_src_len = 0;
315 rtm->rtm_tos = 0;
316 if (fibnum < 255)
317 rtm->rtm_table = (unsigned char)fibnum;
318 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
319 rtm->rtm_protocol = nl_get_rtm_protocol(rnd->rnd_nhop);
320 rtm->rtm_type = get_rtm_type(rnd->rnd_nhop);
321
322 nlattr_add_u32(nw, NL_RTA_TABLE, fibnum);
323
324 int plen = 0;
325 #if defined(INET) || defined(INET6)
326 uint32_t scopeid;
327 #endif
328 switch (family) {
329 #ifdef INET
330 case AF_INET:
331 {
332 struct in_addr addr;
333 rt_get_inet_prefix_plen(rt, &addr, &plen, &scopeid);
334 nlattr_add(nw, NL_RTA_DST, 4, &addr);
335 break;
336 }
337 #endif
338 #ifdef INET6
339 case AF_INET6:
340 {
341 struct in6_addr addr;
342 rt_get_inet6_prefix_plen(rt, &addr, &plen, &scopeid);
343 nlattr_add(nw, NL_RTA_DST, 16, &addr);
344 break;
345 }
346 #endif
347 default:
348 FIB_LOG(LOG_NOTICE, fibnum, family, "unsupported rt family: %d", family);
349 error = EAFNOSUPPORT;
350 goto flush;
351 }
352
353 rtm = nlattr_restore_offset(nw, rtm_off, struct rtmsg);
354 if (plen > 0)
355 rtm->rtm_dst_len = plen;
356 if (NH_IS_NHGRP(rnd->rnd_nhop))
357 dump_rc_nhg(nw, rnd, rtm);
358 else
359 dump_rc_nhop(nw, rnd, rtm);
360
361 if (nlmsg_end(nw))
362 return (0);
363 enomem:
364 error = ENOMEM;
365 flush:
366 nlmsg_abort(nw);
367 return (error);
368 }
369
370 static int
family_to_group(int family)371 family_to_group(int family)
372 {
373 switch (family) {
374 case AF_INET:
375 return (RTNLGRP_IPV4_ROUTE);
376 case AF_INET6:
377 return (RTNLGRP_IPV6_ROUTE);
378 }
379 return (0);
380 }
381
382 static void
report_operation(uint32_t fibnum,struct rib_cmd_info * rc,struct nlpcb * nlp,struct nlmsghdr * hdr)383 report_operation(uint32_t fibnum, struct rib_cmd_info *rc,
384 struct nlpcb *nlp, struct nlmsghdr *hdr)
385 {
386 struct nl_writer nw;
387 uint32_t group_id = family_to_group(rt_get_family(rc->rc_rt));
388
389 if (nl_writer_group(&nw, NLMSG_SMALL, NETLINK_ROUTE, group_id, 0,
390 false)) {
391 struct route_nhop_data rnd = {
392 .rnd_nhop = rc_get_nhop(rc),
393 .rnd_weight = rc->rc_nh_weight,
394 };
395 hdr->nlmsg_flags &= ~(NLM_F_REPLACE | NLM_F_CREATE);
396 hdr->nlmsg_flags &= ~(NLM_F_EXCL | NLM_F_APPEND);
397 switch (rc->rc_cmd) {
398 case RTM_ADD:
399 hdr->nlmsg_type = NL_RTM_NEWROUTE;
400 hdr->nlmsg_flags |= NLM_F_CREATE | NLM_F_EXCL;
401 break;
402 case RTM_CHANGE:
403 hdr->nlmsg_type = NL_RTM_NEWROUTE;
404 hdr->nlmsg_flags |= NLM_F_REPLACE;
405 break;
406 case RTM_DELETE:
407 hdr->nlmsg_type = NL_RTM_DELROUTE;
408 break;
409 }
410 dump_px(fibnum, hdr, rc->rc_rt, &rnd, &nw);
411 nlmsg_flush(&nw);
412 }
413
414 rtsock_callback_p->route_f(fibnum, rc);
415 }
416
417 static void
set_scope6(struct sockaddr * sa,struct ifnet * ifp)418 set_scope6(struct sockaddr *sa, struct ifnet *ifp)
419 {
420 #ifdef INET6
421 if (sa != NULL && sa->sa_family == AF_INET6 && ifp != NULL) {
422 struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)sa;
423
424 if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr))
425 in6_set_unicast_scopeid(&sa6->sin6_addr, if_getindex(ifp));
426 }
427 #endif
428 }
429
430 struct rta_mpath_nh {
431 struct sockaddr *gw;
432 struct ifnet *ifp;
433 uint8_t rtnh_flags;
434 uint8_t rtnh_weight;
435 };
436
437 #define _IN(_field) offsetof(struct rtnexthop, _field)
438 #define _OUT(_field) offsetof(struct rta_mpath_nh, _field)
439 const static struct nlattr_parser nla_p_rtnh[] = {
440 { .type = NL_RTA_GATEWAY, .off = _OUT(gw), .cb = nlattr_get_ip },
441 { .type = NL_RTA_VIA, .off = _OUT(gw), .cb = nlattr_get_ipvia },
442 };
443 const static struct nlfield_parser nlf_p_rtnh[] = {
444 { .off_in = _IN(rtnh_flags), .off_out = _OUT(rtnh_flags), .cb = nlf_get_u8 },
445 { .off_in = _IN(rtnh_hops), .off_out = _OUT(rtnh_weight), .cb = nlf_get_u8 },
446 { .off_in = _IN(rtnh_ifindex), .off_out = _OUT(ifp), .cb = nlf_get_ifpz },
447 };
448 #undef _IN
449 #undef _OUT
450
451 static bool
post_p_rtnh(void * _attrs,struct nl_pstate * npt __unused)452 post_p_rtnh(void *_attrs, struct nl_pstate *npt __unused)
453 {
454 struct rta_mpath_nh *attrs = (struct rta_mpath_nh *)_attrs;
455
456 set_scope6(attrs->gw, attrs->ifp);
457 return (true);
458 }
459 NL_DECLARE_PARSER_EXT(mpath_parser, struct rtnexthop, NULL, nlf_p_rtnh, nla_p_rtnh, post_p_rtnh);
460
461 struct rta_mpath {
462 u_int num_nhops;
463 struct rta_mpath_nh nhops[0];
464 };
465
466 static int
nlattr_get_multipath(struct nlattr * nla,struct nl_pstate * npt,const void * arg,void * target)467 nlattr_get_multipath(struct nlattr *nla, struct nl_pstate *npt,
468 const void *arg, void *target)
469 {
470 struct rta_mpath *mp;
471 struct rtnexthop *rtnh;
472 uint16_t data_len, len;
473 u_int max_nhops;
474 int error;
475
476 data_len = nla->nla_len - sizeof(struct nlattr);
477 max_nhops = data_len / sizeof(struct rtnexthop);
478
479 mp = npt_alloc(npt, (max_nhops + 2) * sizeof(struct rta_mpath_nh));
480 if (mp == NULL) {
481 NLMSG_REPORT_ERR_MSG(npt, "%s: too many RTA_MULTIPATH", __func__);
482 return (ENOMEM);
483 }
484 mp->num_nhops = 0;
485
486 for (rtnh = (struct rtnexthop *)(nla + 1); data_len > 0; ) {
487 struct rta_mpath_nh *mpnh;
488
489 len = NL_ITEM_ALIGN(rtnh->rtnh_len);
490 if (__predict_false(rtnh->rtnh_len <= sizeof(*rtnh) ||
491 len < rtnh->rtnh_len || len > data_len)) {
492 NLMSG_REPORT_ERR_MSG(npt, "%s: bad length %u",
493 __func__, rtnh->rtnh_len);
494 return (EINVAL);
495 }
496 mpnh = &mp->nhops[mp->num_nhops++];
497 error = nl_parse_header(rtnh, rtnh->rtnh_len, &mpath_parser,
498 npt, mpnh);
499 if (error != 0) {
500 NLMSG_REPORT_ERR_MSG(npt,
501 "RTA_MULTIPATH: nexthop %u: parse failed",
502 mp->num_nhops - 1);
503 return (error);
504 }
505 data_len -= len;
506 rtnh = (struct rtnexthop *)((char *)rtnh + len);
507 }
508 if (data_len != 0 || mp->num_nhops == 0) {
509 NLMSG_REPORT_ERR_MSG(npt, "invalid RTA_MULTIPATH attr");
510 return (EINVAL);
511 }
512
513 *((struct rta_mpath **)target) = mp;
514 return (0);
515 }
516
517
518 struct nl_parsed_route {
519 struct sockaddr *rta_dst;
520 struct sockaddr *rta_gw;
521 struct ifnet *rta_oif;
522 struct rta_mpath *rta_multipath;
523 uint32_t rta_table;
524 uint32_t rta_rtflags;
525 uint32_t rta_nh_id;
526 uint32_t rta_metric;
527 uint32_t rta_weight;
528 uint32_t rta_expire;
529 uint32_t rtax_mtu;
530 uint8_t rtm_table;
531 uint8_t rtm_family;
532 uint8_t rtm_dst_len;
533 uint8_t rtm_protocol;
534 uint8_t rtm_type;
535 uint32_t rtm_flags;
536 };
537
538 #define _IN(_field) offsetof(struct rtmsg, _field)
539 #define _OUT(_field) offsetof(struct nl_parsed_route, _field)
540 static struct nlattr_parser nla_p_rtmetrics[] = {
541 { .type = NL_RTAX_MTU, .off = _OUT(rtax_mtu), .cb = nlattr_get_uint32 },
542 };
543 NL_DECLARE_ATTR_PARSER(metrics_parser, nla_p_rtmetrics);
544
545 static const struct nlattr_parser nla_p_rtmsg[] = {
546 { .type = NL_RTA_DST, .off = _OUT(rta_dst), .cb = nlattr_get_ip },
547 { .type = NL_RTA_OIF, .off = _OUT(rta_oif), .cb = nlattr_get_ifp },
548 { .type = NL_RTA_GATEWAY, .off = _OUT(rta_gw), .cb = nlattr_get_ip },
549 { .type = NL_RTA_PRIORITY, .off = _OUT(rta_metric), .cb = nlattr_get_uint32 },
550 { .type = NL_RTA_METRICS, .arg = &metrics_parser, .cb = nlattr_get_nested },
551 { .type = NL_RTA_MULTIPATH, .off = _OUT(rta_multipath), .cb = nlattr_get_multipath },
552 { .type = NL_RTA_WEIGHT, .off = _OUT(rta_weight), .cb = nlattr_get_uint32 },
553 { .type = NL_RTA_RTFLAGS, .off = _OUT(rta_rtflags), .cb = nlattr_get_uint32 },
554 { .type = NL_RTA_TABLE, .off = _OUT(rta_table), .cb = nlattr_get_uint32 },
555 { .type = NL_RTA_VIA, .off = _OUT(rta_gw), .cb = nlattr_get_ipvia },
556 { .type = NL_RTA_EXPIRES, .off = _OUT(rta_expire), .cb = nlattr_get_uint32 },
557 { .type = NL_RTA_NH_ID, .off = _OUT(rta_nh_id), .cb = nlattr_get_uint32 },
558 };
559
560 static const struct nlfield_parser nlf_p_rtmsg[] = {
561 { .off_in = _IN(rtm_family), .off_out = _OUT(rtm_family), .cb = nlf_get_u8 },
562 { .off_in = _IN(rtm_dst_len), .off_out = _OUT(rtm_dst_len), .cb = nlf_get_u8 },
563 { .off_in = _IN(rtm_protocol), .off_out = _OUT(rtm_protocol), .cb = nlf_get_u8 },
564 { .off_in = _IN(rtm_type), .off_out = _OUT(rtm_type), .cb = nlf_get_u8 },
565 { .off_in = _IN(rtm_table), .off_out = _OUT(rtm_table), .cb = nlf_get_u8 },
566 { .off_in = _IN(rtm_flags), .off_out = _OUT(rtm_flags), .cb = nlf_get_u32 },
567 };
568 #undef _IN
569 #undef _OUT
570
571 static bool
post_p_rtmsg(void * _attrs,struct nl_pstate * npt __unused)572 post_p_rtmsg(void *_attrs, struct nl_pstate *npt __unused)
573 {
574 struct nl_parsed_route *attrs = (struct nl_parsed_route *)_attrs;
575
576 set_scope6(attrs->rta_dst, attrs->rta_oif);
577 set_scope6(attrs->rta_gw, attrs->rta_oif);
578 return (true);
579 }
580 NL_DECLARE_PARSER_EXT(rtm_parser, struct rtmsg, NULL, nlf_p_rtmsg, nla_p_rtmsg, post_p_rtmsg);
581
582 struct netlink_walkargs {
583 struct nl_writer *nw;
584 struct route_nhop_data rnd;
585 struct nlmsghdr hdr;
586 struct nlpcb *nlp;
587 uint32_t fibnum;
588 int family;
589 int error;
590 int count;
591 int dumped;
592 int dumped_tables;
593 };
594
595 static int
dump_rtentry(struct rtentry * rt,void * _arg)596 dump_rtentry(struct rtentry *rt, void *_arg)
597 {
598 struct netlink_walkargs *wa = (struct netlink_walkargs *)_arg;
599 int error;
600
601 wa->count++;
602 if (wa->error != 0)
603 return (0);
604 if (!rt_is_exportable(rt, nlp_get_cred(wa->nlp)))
605 return (0);
606 wa->dumped++;
607
608 rt_get_rnd(rt, &wa->rnd);
609
610 error = dump_px(wa->fibnum, &wa->hdr, rt, &wa->rnd, wa->nw);
611
612 IF_DEBUG_LEVEL(LOG_DEBUG3) {
613 char rtbuf[INET6_ADDRSTRLEN + 5];
614 FIB_LOG(LOG_DEBUG3, wa->fibnum, wa->family,
615 "Dump %s, error %d",
616 rt_print_buf(rt, rtbuf, sizeof(rtbuf)), error);
617 }
618 wa->error = error;
619
620 return (0);
621 }
622
623 static void
dump_rtable_one(struct netlink_walkargs * wa,uint32_t fibnum,int family)624 dump_rtable_one(struct netlink_walkargs *wa, uint32_t fibnum, int family)
625 {
626 FIB_LOG(LOG_DEBUG2, fibnum, family, "Start dump");
627 wa->count = 0;
628 wa->dumped = 0;
629
630 rib_walk(fibnum, family, false, dump_rtentry, wa);
631
632 wa->dumped_tables++;
633
634 FIB_LOG(LOG_DEBUG2, fibnum, family, "End dump, iterated %d dumped %d",
635 wa->count, wa->dumped);
636 }
637
638 static int
dump_rtable_fib(struct netlink_walkargs * wa,uint32_t fibnum,int family)639 dump_rtable_fib(struct netlink_walkargs *wa, uint32_t fibnum, int family)
640 {
641 wa->fibnum = fibnum;
642
643 if (family == AF_UNSPEC) {
644 for (int i = 0; i < AF_MAX; i++) {
645 if (rt_tables_get_rnh(fibnum, i) != 0) {
646 wa->family = i;
647 dump_rtable_one(wa, fibnum, i);
648 if (wa->error != 0)
649 break;
650 }
651 }
652 } else {
653 if (rt_tables_get_rnh(fibnum, family) != 0) {
654 wa->family = family;
655 dump_rtable_one(wa, fibnum, family);
656 }
657 }
658
659 return (wa->error);
660 }
661
662 static int
handle_rtm_getroute(struct nlpcb * nlp,struct nl_parsed_route * attrs,struct nlmsghdr * hdr,struct nl_pstate * npt)663 handle_rtm_getroute(struct nlpcb *nlp, struct nl_parsed_route *attrs,
664 struct nlmsghdr *hdr, struct nl_pstate *npt)
665 {
666 RIB_RLOCK_TRACKER;
667 struct rib_head *rnh;
668 const struct rtentry *rt;
669 struct route_nhop_data rnd;
670 uint32_t fibnum = attrs->rta_table;
671 sa_family_t family = attrs->rtm_family;
672
673 if (attrs->rta_dst == NULL) {
674 NLMSG_REPORT_ERR_MSG(npt, "No RTA_DST supplied");
675 return (EINVAL);
676 }
677
678 rnh = rt_tables_get_rnh(fibnum, family);
679 if (rnh == NULL)
680 return (EAFNOSUPPORT);
681
682 RIB_RLOCK(rnh);
683
684 struct sockaddr *dst = attrs->rta_dst;
685
686 if (attrs->rtm_flags & RTM_F_PREFIX)
687 rt = rib_lookup_prefix_plen(rnh, dst, attrs->rtm_dst_len, &rnd);
688 else
689 rt = (const struct rtentry *)rnh->rnh_matchaddr(dst, &rnh->head);
690 if (rt == NULL) {
691 RIB_RUNLOCK(rnh);
692 return (ESRCH);
693 }
694
695 rt_get_rnd(rt, &rnd);
696
697 RIB_RUNLOCK(rnh);
698
699 if (!rt_is_exportable(rt, nlp_get_cred(nlp)))
700 return (ESRCH);
701
702 IF_DEBUG_LEVEL(LOG_DEBUG2) {
703 char rtbuf[NHOP_PRINT_BUFSIZE] __unused, nhbuf[NHOP_PRINT_BUFSIZE] __unused;
704 FIB_LOG(LOG_DEBUG2, fibnum, family, "getroute completed: got %s for %s",
705 nhop_print_buf_any(rnd.rnd_nhop, nhbuf, sizeof(nhbuf)),
706 rt_print_buf(rt, rtbuf, sizeof(rtbuf)));
707 }
708
709 hdr->nlmsg_type = NL_RTM_NEWROUTE;
710 dump_px(fibnum, hdr, rt, &rnd, npt->nw);
711
712 return (0);
713 }
714
715 static int
handle_rtm_dump(struct nlpcb * nlp,uint32_t fibnum,int family,struct nlmsghdr * hdr,struct nl_writer * nw)716 handle_rtm_dump(struct nlpcb *nlp, uint32_t fibnum, int family,
717 struct nlmsghdr *hdr, struct nl_writer *nw)
718 {
719 struct netlink_walkargs wa = {
720 .nlp = nlp,
721 .nw = nw,
722 .hdr.nlmsg_pid = hdr->nlmsg_pid,
723 .hdr.nlmsg_seq = hdr->nlmsg_seq,
724 .hdr.nlmsg_type = NL_RTM_NEWROUTE,
725 .hdr.nlmsg_flags = hdr->nlmsg_flags | NLM_F_MULTI,
726 };
727
728 if (fibnum == RT_TABLE_UNSPEC) {
729 for (int i = 0; i < V_rt_numfibs; i++) {
730 dump_rtable_fib(&wa, i, family);
731 if (wa.error != 0)
732 break;
733 }
734 } else
735 dump_rtable_fib(&wa, fibnum, family);
736
737 if (wa.error == 0 && wa.dumped_tables == 0) {
738 FIB_LOG(LOG_DEBUG, fibnum, family, "incorrect fibnum/family");
739 wa.error = ESRCH;
740 // How do we propagate it?
741 }
742
743 if (!nlmsg_end_dump(wa.nw, wa.error, &wa.hdr)) {
744 NL_LOG(LOG_DEBUG, "Unable to finalize the dump");
745 return (ENOMEM);
746 }
747
748 return (wa.error);
749 }
750
751 static struct nhop_object *
finalize_nhop(struct nhop_object * nh,const struct sockaddr * dst,int * perror)752 finalize_nhop(struct nhop_object *nh, const struct sockaddr *dst, int *perror)
753 {
754 /*
755 * The following MUST be filled:
756 * nh_ifp, nh_ifa, nh_gw
757 */
758 if (nh->gw_sa.sa_family == 0) {
759 /*
760 * Empty gateway. Can be direct route with RTA_OIF set.
761 */
762 if (nh->nh_ifp != NULL)
763 nhop_set_direct_gw(nh, nh->nh_ifp);
764 else {
765 NL_LOG(LOG_DEBUG, "empty gateway and interface, skipping");
766 *perror = EINVAL;
767 return (NULL);
768 }
769 /* Both nh_ifp and gateway are set */
770 } else {
771 /* Gateway is set up, we can derive ifp if not set */
772 if (nh->nh_ifp == NULL) {
773 uint32_t fibnum = nhop_get_fibnum(nh);
774 uint32_t flags = 0;
775
776 if (nh->nh_flags & NHF_GATEWAY)
777 flags = RTF_GATEWAY;
778 else if (nh->nh_flags & NHF_HOST)
779 flags = RTF_HOST;
780
781 struct ifaddr *ifa = ifa_ifwithroute(flags, dst, &nh->gw_sa, fibnum);
782 if (ifa == NULL) {
783 NL_LOG(LOG_DEBUG, "Unable to determine ifp, skipping");
784 *perror = EINVAL;
785 return (NULL);
786 }
787 nhop_set_transmit_ifp(nh, ifa->ifa_ifp);
788 }
789 }
790 /* Both nh_ifp and gateway are set */
791 if (nh->nh_ifa == NULL) {
792 const struct sockaddr *gw_sa = &nh->gw_sa;
793
794 if (gw_sa->sa_family != dst->sa_family) {
795 /*
796 * Use dst as the target for determining the default
797 * preferred ifa IF
798 * 1) the gateway is link-level (e.g. direct route)
799 * 2) the gateway family is different (e.g. IPv4 over IPv6).
800 */
801 gw_sa = dst;
802 }
803
804 struct ifaddr *ifa = ifaof_ifpforaddr(gw_sa, nh->nh_ifp);
805 if (ifa == NULL) {
806 /* Try link-level ifa. */
807 gw_sa = &nh->gw_sa;
808 ifa = ifaof_ifpforaddr(gw_sa, nh->nh_ifp);
809 if (ifa == NULL) {
810 NL_LOG(LOG_DEBUG, "Unable to determine ifa, skipping");
811 *perror = EINVAL;
812 return (NULL);
813 }
814 }
815 nhop_set_src(nh, ifa);
816 }
817
818 return (nhop_get_nhop(nh, perror));
819 }
820
821 static int
get_pxflag(const struct nl_parsed_route * attrs)822 get_pxflag(const struct nl_parsed_route *attrs)
823 {
824 int pxflag = 0;
825 switch (attrs->rtm_family) {
826 case AF_INET:
827 if (attrs->rtm_dst_len == 32)
828 pxflag = NHF_HOST;
829 else if (attrs->rtm_dst_len == 0)
830 pxflag = NHF_DEFAULT;
831 break;
832 case AF_INET6:
833 if (attrs->rtm_dst_len == 128)
834 pxflag = NHF_HOST;
835 else if (attrs->rtm_dst_len == 0)
836 pxflag = NHF_DEFAULT;
837 break;
838 }
839
840 return (pxflag);
841 }
842
843 static int
get_op_flags(int nlm_flags)844 get_op_flags(int nlm_flags)
845 {
846 int op_flags = 0;
847
848 op_flags |= (nlm_flags & NLM_F_REPLACE) ? RTM_F_REPLACE : 0;
849 op_flags |= (nlm_flags & NLM_F_EXCL) ? RTM_F_EXCL : 0;
850 op_flags |= (nlm_flags & NLM_F_CREATE) ? RTM_F_CREATE : 0;
851 op_flags |= (nlm_flags & NLM_F_APPEND) ? RTM_F_APPEND : 0;
852
853 return (op_flags);
854 }
855
856 static int
create_nexthop_one(struct nl_parsed_route * attrs,struct rta_mpath_nh * mpnh,struct nl_pstate * npt,struct nhop_object ** pnh)857 create_nexthop_one(struct nl_parsed_route *attrs, struct rta_mpath_nh *mpnh,
858 struct nl_pstate *npt, struct nhop_object **pnh)
859 {
860 int error;
861
862 if (mpnh->gw == NULL)
863 return (EINVAL);
864
865 struct nhop_object *nh = nhop_alloc(attrs->rta_table, attrs->rtm_family);
866 if (nh == NULL)
867 return (ENOMEM);
868
869 error = nl_set_nexthop_gw(nh, mpnh->gw, mpnh->ifp, npt);
870 if (error != 0) {
871 nhop_free(nh);
872 return (error);
873 }
874 if (mpnh->ifp != NULL)
875 nhop_set_transmit_ifp(nh, mpnh->ifp);
876 nhop_set_pxtype_flag(nh, get_pxflag(attrs));
877 nhop_set_rtflags(nh, attrs->rta_rtflags);
878 nhop_set_metric(nh, attrs->rta_metric);
879 if (attrs->rtm_protocol > RTPROT_STATIC)
880 nhop_set_origin(nh, attrs->rtm_protocol);
881
882 *pnh = finalize_nhop(nh, attrs->rta_dst, &error);
883
884 return (error);
885 }
886
887 static struct nhop_object *
create_nexthop_from_attrs(struct nl_parsed_route * attrs,struct nl_pstate * npt,int * perror)888 create_nexthop_from_attrs(struct nl_parsed_route *attrs,
889 struct nl_pstate *npt, int *perror)
890 {
891 struct nhop_object *nh = NULL;
892 int error = 0;
893 uint32_t nh_expire = 0;
894
895 if (attrs->rta_multipath != NULL) {
896 /* Multipath w/o explicit nexthops */
897 int num_nhops = attrs->rta_multipath->num_nhops;
898 struct weightened_nhop *wn = npt_alloc(npt, sizeof(*wn) * num_nhops);
899
900 if (wn == NULL) {
901 *perror = ENOMEM;
902 return (NULL);
903 }
904 for (int i = 0; i < num_nhops; i++) {
905 struct rta_mpath_nh *mpnh = &attrs->rta_multipath->nhops[i];
906
907 error = create_nexthop_one(attrs, mpnh, npt, &wn[i].nh);
908 if (error != 0) {
909 for (int j = 0; j < i; j++)
910 nhop_free(wn[j].nh);
911 break;
912 }
913 wn[i].weight = mpnh->rtnh_weight > 0 ? mpnh->rtnh_weight : 1;
914 }
915 if (error == 0) {
916 struct rib_head *rh = nhop_get_rh(wn[0].nh);
917 struct nhgrp_object *nhg;
918
919 nhg = nhgrp_alloc(rh->rib_fibnum, rh->rib_family,
920 wn, num_nhops, perror);
921 if (nhg != NULL) {
922 if (attrs->rtm_protocol > RTPROT_STATIC)
923 nhgrp_set_origin(nhg, attrs->rtm_protocol);
924 nhg = nhgrp_get_nhgrp(nhg, perror);
925 }
926 for (int i = 0; i < num_nhops; i++)
927 nhop_free(wn[i].nh);
928 if (nhg != NULL)
929 return ((struct nhop_object *)nhg);
930 error = *perror;
931 }
932 *perror = error;
933 } else {
934 nh = nhop_alloc(attrs->rta_table, attrs->rtm_family);
935 if (nh == NULL) {
936 *perror = ENOMEM;
937 return (NULL);
938 }
939 if (attrs->rta_gw != NULL) {
940 *perror = nl_set_nexthop_gw(nh, attrs->rta_gw, attrs->rta_oif, npt);
941 if (*perror != 0) {
942 nhop_free(nh);
943 return (NULL);
944 }
945 }
946 if (attrs->rta_oif != NULL)
947 nhop_set_transmit_ifp(nh, attrs->rta_oif);
948 if (attrs->rtax_mtu != 0)
949 nhop_set_mtu(nh, attrs->rtax_mtu, true);
950 if (attrs->rta_expire > 0) {
951 nh_expire = attrs->rta_expire - time_second + time_uptime;
952 nhop_set_expire(nh, nh_expire);
953 }
954 if (attrs->rta_rtflags & RTF_BROADCAST)
955 nhop_set_broadcast(nh, true);
956 if (attrs->rtm_protocol > RTPROT_STATIC)
957 nhop_set_origin(nh, attrs->rtm_protocol);
958 nhop_set_metric(nh, attrs->rta_metric);
959 nhop_set_pxtype_flag(nh, get_pxflag(attrs));
960 nhop_set_rtflags(nh, attrs->rta_rtflags);
961
962 switch (attrs->rtm_type) {
963 case RTN_UNICAST:
964 break;
965 case RTN_BLACKHOLE:
966 nhop_set_blackhole(nh, RTF_BLACKHOLE);
967 break;
968 case RTN_PROHIBIT:
969 case RTN_UNREACHABLE:
970 nhop_set_blackhole(nh, RTF_REJECT);
971 break;
972 /* TODO: return ENOTSUP for other types if strict option is set */
973 }
974
975 nh = finalize_nhop(nh, attrs->rta_dst, perror);
976 }
977
978 return (nh);
979 }
980
981 /* pre-2.6.19 Linux API compatibility: prefer RTA_TABLE, fall back to rtm_table */
982 static inline void
old_linux_compat(struct nl_parsed_route * attrs)983 old_linux_compat(struct nl_parsed_route *attrs)
984 {
985 if (attrs->rtm_table > 0 && attrs->rta_table == 0)
986 attrs->rta_table = attrs->rtm_table;
987 }
988
989 static int
rtnl_handle_newroute(struct nlmsghdr * hdr,struct nlpcb * nlp,struct nl_pstate * npt)990 rtnl_handle_newroute(struct nlmsghdr *hdr, struct nlpcb *nlp,
991 struct nl_pstate *npt)
992 {
993 struct rib_cmd_info rc = {};
994 struct nhop_object *nh = NULL;
995 int error;
996
997 struct nl_parsed_route attrs = {};
998 error = nl_parse_nlmsg(hdr, &rtm_parser, npt, &attrs);
999 if (error != 0)
1000 return (error);
1001
1002 /* Check if we have enough data */
1003 if (attrs.rta_dst == NULL) {
1004 NL_LOG(LOG_DEBUG, "missing RTA_DST");
1005 return (EINVAL);
1006 }
1007
1008 old_linux_compat(&attrs);
1009 if (attrs.rta_table >= V_rt_numfibs || attrs.rtm_family > AF_MAX) {
1010 NLMSG_REPORT_ERR_MSG(npt, "invalid fib");
1011 return (EINVAL);
1012 }
1013
1014 if (attrs.rta_nh_id != 0) {
1015 /* Referenced uindex */
1016 int pxflag = get_pxflag(&attrs);
1017 nh = nl_find_nhop(attrs.rta_table, attrs.rtm_family, attrs.rta_nh_id,
1018 pxflag, &error);
1019 if (error != 0)
1020 return (error);
1021 } else {
1022 nh = create_nexthop_from_attrs(&attrs, npt, &error);
1023 if (error != 0) {
1024 NL_LOG(LOG_DEBUG, "Error creating nexthop");
1025 return (error);
1026 }
1027 }
1028
1029 if (!NH_IS_NHGRP(nh) && attrs.rta_weight == 0)
1030 attrs.rta_weight = RT_DEFAULT_WEIGHT;
1031 struct route_nhop_data rnd = { .rnd_nhop = nh, .rnd_weight = attrs.rta_weight };
1032 int op_flags = get_op_flags(hdr->nlmsg_flags);
1033
1034 error = rib_add_route_px(attrs.rta_table, attrs.rta_dst, attrs.rtm_dst_len,
1035 &rnd, op_flags, &rc);
1036 if (error == 0)
1037 report_operation(attrs.rta_table, &rc, nlp, hdr);
1038 return (error);
1039 }
1040
1041 static int
path_match_func(const struct rtentry * rt,const struct nhop_object * nh,void * _data)1042 path_match_func(const struct rtentry *rt, const struct nhop_object *nh, void *_data)
1043 {
1044 struct nl_parsed_route *attrs = (struct nl_parsed_route *)_data;
1045
1046 if (attrs->rta_metric != 0 && attrs->rta_metric != nhop_get_metric(nh))
1047 return (0);
1048
1049 if ((attrs->rta_gw != NULL) && !rib_match_gw(rt, nh, attrs->rta_gw))
1050 return (0);
1051
1052 if ((attrs->rta_oif != NULL) && (attrs->rta_oif != nh->nh_ifp))
1053 return (0);
1054
1055 return (1);
1056 }
1057
1058 static int
rtnl_handle_delroute(struct nlmsghdr * hdr,struct nlpcb * nlp,struct nl_pstate * npt)1059 rtnl_handle_delroute(struct nlmsghdr *hdr, struct nlpcb *nlp,
1060 struct nl_pstate *npt)
1061 {
1062 struct rib_cmd_info rc;
1063 int error;
1064
1065 struct nl_parsed_route attrs = {};
1066 error = nl_parse_nlmsg(hdr, &rtm_parser, npt, &attrs);
1067 if (error != 0)
1068 return (error);
1069
1070 if (attrs.rta_dst == NULL) {
1071 NLMSG_REPORT_ERR_MSG(npt, "RTA_DST is not set");
1072 return (ESRCH);
1073 }
1074
1075 old_linux_compat(&attrs);
1076 if (attrs.rta_table >= V_rt_numfibs || attrs.rtm_family > AF_MAX) {
1077 NLMSG_REPORT_ERR_MSG(npt, "invalid fib");
1078 return (EINVAL);
1079 }
1080
1081 error = rib_del_route_px(attrs.rta_table, attrs.rta_dst,
1082 attrs.rtm_dst_len, path_match_func, &attrs,
1083 (attrs.rta_rtflags & RTF_PINNED) ? RTM_F_FORCE : 0, &rc);
1084 if (error == 0)
1085 report_operation(attrs.rta_table, &rc, nlp, hdr);
1086 return (error);
1087 }
1088
1089 static int
rtnl_handle_getroute(struct nlmsghdr * hdr,struct nlpcb * nlp,struct nl_pstate * npt)1090 rtnl_handle_getroute(struct nlmsghdr *hdr, struct nlpcb *nlp, struct nl_pstate *npt)
1091 {
1092 int error;
1093
1094 struct nl_parsed_route attrs = {};
1095 error = nl_parse_nlmsg(hdr, &rtm_parser, npt, &attrs);
1096 if (error != 0)
1097 return (error);
1098
1099 old_linux_compat(&attrs);
1100 if (attrs.rta_table >= V_rt_numfibs || attrs.rtm_family > AF_MAX) {
1101 NLMSG_REPORT_ERR_MSG(npt, "invalid fib");
1102 return (EINVAL);
1103 }
1104
1105 if (hdr->nlmsg_flags & NLM_F_DUMP)
1106 error = handle_rtm_dump(nlp, attrs.rta_table, attrs.rtm_family, hdr, npt->nw);
1107 else
1108 error = handle_rtm_getroute(nlp, &attrs, hdr, npt);
1109
1110 return (error);
1111 }
1112
1113 void
rtnl_handle_route_event(uint32_t fibnum,const struct rib_cmd_info * rc)1114 rtnl_handle_route_event(uint32_t fibnum, const struct rib_cmd_info *rc)
1115 {
1116 struct nl_writer nw;
1117 int family, nlm_flags = 0;
1118
1119 family = rt_get_family(rc->rc_rt);
1120
1121 /* XXX: check if there are active listeners first */
1122
1123 /* TODO: consider passing PID/type/seq */
1124 switch (rc->rc_cmd) {
1125 case RTM_ADD:
1126 nlm_flags = NLM_F_EXCL | NLM_F_CREATE;
1127 break;
1128 case RTM_CHANGE:
1129 nlm_flags = NLM_F_REPLACE;
1130 break;
1131 case RTM_DELETE:
1132 nlm_flags = 0;
1133 break;
1134 }
1135 IF_DEBUG_LEVEL(LOG_DEBUG2) {
1136 char rtbuf[NHOP_PRINT_BUFSIZE] __unused;
1137 FIB_LOG(LOG_DEBUG2, fibnum, family,
1138 "received event %s for %s / nlm_flags=%X",
1139 rib_print_cmd(rc->rc_cmd),
1140 rt_print_buf(rc->rc_rt, rtbuf, sizeof(rtbuf)),
1141 nlm_flags);
1142 }
1143
1144 struct nlmsghdr hdr = {
1145 .nlmsg_flags = nlm_flags,
1146 .nlmsg_type = get_rtmsg_type_from_rtsock(rc->rc_cmd),
1147 };
1148
1149 struct route_nhop_data rnd = {
1150 .rnd_nhop = rc_get_nhop(rc),
1151 .rnd_weight = rc->rc_nh_weight,
1152 };
1153
1154 uint32_t group_id = family_to_group(family);
1155 if (!nl_writer_group(&nw, NLMSG_SMALL, NETLINK_ROUTE, group_id, 0,
1156 false)) {
1157 NL_LOG(LOG_DEBUG, "error allocating event buffer");
1158 return;
1159 }
1160
1161 dump_px(fibnum, &hdr, rc->rc_rt, &rnd, &nw);
1162 nlmsg_flush(&nw);
1163 }
1164
1165 static const struct rtnl_cmd_handler cmd_handlers[] = {
1166 {
1167 .cmd = NL_RTM_GETROUTE,
1168 .name = "RTM_GETROUTE",
1169 .cb = &rtnl_handle_getroute,
1170 .flags = RTNL_F_ALLOW_NONVNET_JAIL,
1171 },
1172 {
1173 .cmd = NL_RTM_DELROUTE,
1174 .name = "RTM_DELROUTE",
1175 .cb = &rtnl_handle_delroute,
1176 .priv = PRIV_NET_ROUTE,
1177 .flags = RTNL_F_ALLOW_NONVNET_JAIL,
1178 },
1179 {
1180 .cmd = NL_RTM_NEWROUTE,
1181 .name = "RTM_NEWROUTE",
1182 .cb = &rtnl_handle_newroute,
1183 .priv = PRIV_NET_ROUTE,
1184 .flags = RTNL_F_ALLOW_NONVNET_JAIL,
1185 }
1186 };
1187
1188 static const struct nlhdr_parser *all_parsers[] = {&mpath_parser, &metrics_parser, &rtm_parser};
1189
1190 void
rtnl_routes_init(void)1191 rtnl_routes_init(void)
1192 {
1193 NL_VERIFY_PARSERS(all_parsers);
1194 rtnl_register_messages(cmd_handlers, nitems(cmd_handlers));
1195 }
1196