xref: /freebsd-13.1/sys/net/route.c (revision b9772822)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1980, 1986, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
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  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)route.c	8.3.1.1 (Berkeley) 2/23/95
32  * $FreeBSD$
33  */
34 /************************************************************************
35  * Note: In this file a 'fib' is a "forwarding information base"	*
36  * Which is the new name for an in kernel routing (next hop) table.	*
37  ***********************************************************************/
38 
39 #include "opt_inet.h"
40 #include "opt_inet6.h"
41 #include "opt_mrouting.h"
42 #include "opt_route.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/socket.h>
49 #include <sys/sysctl.h>
50 #include <sys/syslog.h>
51 #include <sys/sysproto.h>
52 #include <sys/proc.h>
53 #include <sys/domain.h>
54 #include <sys/eventhandler.h>
55 #include <sys/kernel.h>
56 #include <sys/lock.h>
57 #include <sys/rmlock.h>
58 
59 #include <net/if.h>
60 #include <net/if_var.h>
61 #include <net/if_dl.h>
62 #include <net/route.h>
63 #include <net/route/route_ctl.h>
64 #include <net/route/route_var.h>
65 #include <net/route/nhop.h>
66 #include <net/vnet.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/ip_mroute.h>
70 
71 VNET_PCPUSTAT_DEFINE(struct rtstat, rtstat);
72 
73 VNET_PCPUSTAT_SYSINIT(rtstat);
74 #ifdef VIMAGE
75 VNET_PCPUSTAT_SYSUNINIT(rtstat);
76 #endif
77 
78 EVENTHANDLER_LIST_DEFINE(rt_addrmsg);
79 
80 static int rt_ifdelroute(const struct rtentry *rt, const struct nhop_object *,
81     void *arg);
82 static int rt_exportinfo(struct rtentry *rt, struct nhop_object *nh,
83     struct rt_addrinfo *info, int flags);
84 
85 /*
86  * route initialization must occur before ip6_init2(), which happenas at
87  * SI_ORDER_MIDDLE.
88  */
89 static void
route_init(void)90 route_init(void)
91 {
92 
93 	nhops_init();
94 }
95 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, NULL);
96 
97 struct rib_head *
rt_table_init(int offset,int family,u_int fibnum)98 rt_table_init(int offset, int family, u_int fibnum)
99 {
100 	struct rib_head *rh;
101 
102 	rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO);
103 
104 	/* TODO: These details should be hidded inside radix.c */
105 	/* Init masks tree */
106 	rn_inithead_internal(&rh->head, rh->rnh_nodes, offset);
107 	rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0);
108 	rh->head.rnh_masks = &rh->rmhead;
109 
110 	/* Save metadata associated with this routing table. */
111 	rh->rib_family = family;
112 	rh->rib_fibnum = fibnum;
113 #ifdef VIMAGE
114 	rh->rib_vnet = curvnet;
115 #endif
116 
117 	tmproutes_init(rh);
118 
119 	/* Init locks */
120 	RIB_LOCK_INIT(rh);
121 
122 	nhops_init_rib(rh);
123 
124 	/* Init subscription system */
125 	rib_init_subscriptions(rh);
126 
127 	/* Finally, set base callbacks */
128 	rh->rnh_addaddr = rn_addroute;
129 	rh->rnh_deladdr = rn_delete;
130 	rh->rnh_matchaddr = rn_match;
131 	rh->rnh_lookup = rn_lookup;
132 	rh->rnh_walktree = rn_walktree;
133 	rh->rnh_walktree_from = rn_walktree_from;
134 
135 	return (rh);
136 }
137 
138 static int
rt_freeentry(struct radix_node * rn,void * arg)139 rt_freeentry(struct radix_node *rn, void *arg)
140 {
141 	struct radix_head * const rnh = arg;
142 	struct radix_node *x;
143 
144 	x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
145 	if (x != NULL)
146 		R_Free(x);
147 	return (0);
148 }
149 
150 void
rt_table_destroy(struct rib_head * rh)151 rt_table_destroy(struct rib_head *rh)
152 {
153 
154 	RIB_WLOCK(rh);
155 	rh->rib_dying = true;
156 	RIB_WUNLOCK(rh);
157 
158 #ifdef FIB_ALGO
159 	fib_destroy_rib(rh);
160 #endif
161 
162 	tmproutes_destroy(rh);
163 
164 	rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
165 
166 	nhops_destroy_rib(rh);
167 
168 	rib_destroy_subscriptions(rh);
169 
170 	/* Assume table is already empty */
171 	RIB_LOCK_DESTROY(rh);
172 	free(rh, M_RTABLE);
173 }
174 
175 /*
176  * Adds a temporal redirect entry to the routing table.
177  * @fibnum: fib number
178  * @dst: destination to install redirect to
179  * @gateway: gateway to go via
180  * @author: sockaddr of originating router, can be NULL
181  * @ifp: interface to use for the redirected route
182  * @flags: set of flags to add. Allowed: RTF_GATEWAY
183  * @lifetime_sec: time in seconds to expire this redirect.
184  *
185  * Retuns 0 on success, errno otherwise.
186  */
187 int
rib_add_redirect(u_int fibnum,struct sockaddr * dst,struct sockaddr * gateway,struct sockaddr * author,struct ifnet * ifp,int flags,int lifetime_sec)188 rib_add_redirect(u_int fibnum, struct sockaddr *dst, struct sockaddr *gateway,
189     struct sockaddr *author, struct ifnet *ifp, int flags, int lifetime_sec)
190 {
191 	struct rib_cmd_info rc;
192 	int error;
193 	struct rt_addrinfo info;
194 	struct rt_metrics rti_rmx;
195 	struct ifaddr *ifa;
196 
197 	NET_EPOCH_ASSERT();
198 
199 	if (rt_tables_get_rnh(fibnum, dst->sa_family) == NULL)
200 		return (EAFNOSUPPORT);
201 
202 	/* Verify the allowed flag mask. */
203 	KASSERT(((flags & ~(RTF_GATEWAY)) == 0),
204 	    ("invalid redirect flags: %x", flags));
205 	flags |= RTF_HOST | RTF_DYNAMIC;
206 
207 	/* Get the best ifa for the given interface and gateway. */
208 	if ((ifa = ifaof_ifpforaddr(gateway, ifp)) == NULL)
209 		return (ENETUNREACH);
210 
211 	bzero(&info, sizeof(info));
212 	info.rti_info[RTAX_DST] = dst;
213 	info.rti_info[RTAX_GATEWAY] = gateway;
214 	info.rti_ifa = ifa;
215 	info.rti_ifp = ifp;
216 	info.rti_flags = flags;
217 
218 	/* Setup route metrics to define expire time. */
219 	bzero(&rti_rmx, sizeof(rti_rmx));
220 	/* Set expire time as absolute. */
221 	rti_rmx.rmx_expire = lifetime_sec + time_second;
222 	info.rti_mflags |= RTV_EXPIRE;
223 	info.rti_rmx = &rti_rmx;
224 
225 	error = rib_action(fibnum, RTM_ADD, &info, &rc);
226 
227 	if (error != 0) {
228 		/* TODO: add per-fib redirect stats. */
229 		return (error);
230 	}
231 
232 	RTSTAT_INC(rts_dynamic);
233 
234 	/* Send notification of a route addition to userland. */
235 	bzero(&info, sizeof(info));
236 	info.rti_info[RTAX_DST] = dst;
237 	info.rti_info[RTAX_GATEWAY] = gateway;
238 	info.rti_info[RTAX_AUTHOR] = author;
239 	rt_missmsg_fib(RTM_REDIRECT, &info, flags | RTF_UP, error, fibnum);
240 
241 	return (0);
242 }
243 
244 /*
245  * Routing table ioctl interface.
246  */
247 int
rtioctl_fib(u_long req,caddr_t data,u_int fibnum)248 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
249 {
250 
251 	/*
252 	 * If more ioctl commands are added here, make sure the proper
253 	 * super-user checks are being performed because it is possible for
254 	 * prison-root to make it this far if raw sockets have been enabled
255 	 * in jails.
256 	 */
257 #ifdef INET
258 	/* Multicast goop, grrr... */
259 	return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
260 #else /* INET */
261 	return ENXIO;
262 #endif /* INET */
263 }
264 
265 struct ifaddr *
ifa_ifwithroute(int flags,const struct sockaddr * dst,const struct sockaddr * gateway,u_int fibnum)266 ifa_ifwithroute(int flags, const struct sockaddr *dst,
267     const struct sockaddr *gateway, u_int fibnum)
268 {
269 	struct ifaddr *ifa;
270 
271 	NET_EPOCH_ASSERT();
272 	if ((flags & RTF_GATEWAY) == 0) {
273 		/*
274 		 * If we are adding a route to an interface,
275 		 * and the interface is a pt to pt link
276 		 * we should search for the destination
277 		 * as our clue to the interface.  Otherwise
278 		 * we can use the local address.
279 		 */
280 		ifa = NULL;
281 		if (flags & RTF_HOST)
282 			ifa = ifa_ifwithdstaddr(dst, fibnum);
283 		if (ifa == NULL)
284 			ifa = ifa_ifwithaddr(gateway);
285 	} else {
286 		/*
287 		 * If we are adding a route to a remote net
288 		 * or host, the gateway may still be on the
289 		 * other end of a pt to pt link.
290 		 */
291 		ifa = ifa_ifwithdstaddr(gateway, fibnum);
292 	}
293 	if (ifa == NULL)
294 		ifa = ifa_ifwithnet(gateway, 0, fibnum);
295 	if (ifa == NULL) {
296 		struct nhop_object *nh;
297 
298 		nh = rib_lookup(fibnum, gateway, NHR_NONE, 0);
299 
300 		/*
301 		 * dismiss a gateway that is reachable only
302 		 * through the default router
303 		 */
304 		if ((nh == NULL) || (nh->nh_flags & NHF_DEFAULT))
305 			return (NULL);
306 		ifa = nh->nh_ifa;
307 	}
308 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
309 		struct ifaddr *oifa = ifa;
310 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
311 		if (ifa == NULL)
312 			ifa = oifa;
313 	}
314 
315 	return (ifa);
316 }
317 
318 /*
319  * Copy most of @rt data into @info.
320  *
321  * If @flags contains NHR_COPY, copies dst,netmask and gw to the
322  * pointers specified by @info structure. Assume such pointers
323  * are zeroed sockaddr-like structures with sa_len field initialized
324  * to reflect size of the provided buffer. if no NHR_COPY is specified,
325  * point dst,netmask and gw @info fields to appropriate @rt values.
326  *
327  * if @flags contains NHR_REF, do refcouting on rt_ifp and rt_ifa.
328  *
329  * Returns 0 on success.
330  */
331 static int
rt_exportinfo(struct rtentry * rt,struct nhop_object * nh,struct rt_addrinfo * info,int flags)332 rt_exportinfo(struct rtentry *rt, struct nhop_object *nh,
333     struct rt_addrinfo *info, int flags)
334 {
335 	struct rt_metrics *rmx;
336 	struct sockaddr *src, *dst;
337 	int sa_len;
338 
339 	if (flags & NHR_COPY) {
340 		/* Copy destination if dst is non-zero */
341 		src = rt_key(rt);
342 		dst = info->rti_info[RTAX_DST];
343 		sa_len = src->sa_len;
344 		if (dst != NULL) {
345 			if (src->sa_len > dst->sa_len)
346 				return (ENOMEM);
347 			memcpy(dst, src, src->sa_len);
348 			info->rti_addrs |= RTA_DST;
349 		}
350 
351 		/* Copy mask if set && dst is non-zero */
352 		src = rt_mask(rt);
353 		dst = info->rti_info[RTAX_NETMASK];
354 		if (src != NULL && dst != NULL) {
355 			/*
356 			 * Radix stores different value in sa_len,
357 			 * assume rt_mask() to have the same length
358 			 * as rt_key()
359 			 */
360 			if (sa_len > dst->sa_len)
361 				return (ENOMEM);
362 			memcpy(dst, src, src->sa_len);
363 			info->rti_addrs |= RTA_NETMASK;
364 		}
365 
366 		/* Copy gateway is set && dst is non-zero */
367 		src = &nh->gw_sa;
368 		dst = info->rti_info[RTAX_GATEWAY];
369 		if ((nhop_get_rtflags(nh) & RTF_GATEWAY) &&
370 		    src != NULL && dst != NULL) {
371 			if (src->sa_len > dst->sa_len)
372 				return (ENOMEM);
373 			memcpy(dst, src, src->sa_len);
374 			info->rti_addrs |= RTA_GATEWAY;
375 		}
376 	} else {
377 		info->rti_info[RTAX_DST] = rt_key(rt);
378 		info->rti_addrs |= RTA_DST;
379 		if (rt_mask(rt) != NULL) {
380 			info->rti_info[RTAX_NETMASK] = rt_mask(rt);
381 			info->rti_addrs |= RTA_NETMASK;
382 		}
383 		if (nhop_get_rtflags(nh) & RTF_GATEWAY) {
384 			info->rti_info[RTAX_GATEWAY] = &nh->gw_sa;
385 			info->rti_addrs |= RTA_GATEWAY;
386 		}
387 	}
388 
389 	rmx = info->rti_rmx;
390 	if (rmx != NULL) {
391 		info->rti_mflags |= RTV_MTU;
392 		rmx->rmx_mtu = nh->nh_mtu;
393 	}
394 
395 	info->rti_flags = rt->rte_flags | nhop_get_rtflags(nh);
396 	info->rti_ifp = nh->nh_ifp;
397 	info->rti_ifa = nh->nh_ifa;
398 	if (flags & NHR_REF) {
399 		if_ref(info->rti_ifp);
400 		ifa_ref(info->rti_ifa);
401 	}
402 
403 	return (0);
404 }
405 
406 /*
407  * Lookups up route entry for @dst in RIB database for fib @fibnum.
408  * Exports entry data to @info using rt_exportinfo().
409  *
410  * If @flags contains NHR_REF, refcouting is performed on rt_ifp and rt_ifa.
411  * All references can be released later by calling rib_free_info().
412  *
413  * Returns 0 on success.
414  * Returns ENOENT for lookup failure, ENOMEM for export failure.
415  */
416 int
rib_lookup_info(uint32_t fibnum,const struct sockaddr * dst,uint32_t flags,uint32_t flowid,struct rt_addrinfo * info)417 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
418     uint32_t flowid, struct rt_addrinfo *info)
419 {
420 	RIB_RLOCK_TRACKER;
421 	struct rib_head *rh;
422 	struct radix_node *rn;
423 	struct rtentry *rt;
424 	struct nhop_object *nh;
425 	int error;
426 
427 	KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
428 	rh = rt_tables_get_rnh(fibnum, dst->sa_family);
429 	if (rh == NULL)
430 		return (ENOENT);
431 
432 	RIB_RLOCK(rh);
433 	rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
434 	if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
435 		rt = RNTORT(rn);
436 		nh = nhop_select(rt->rt_nhop, flowid);
437 		/* Ensure route & ifp is UP */
438 		if (RT_LINK_IS_UP(nh->nh_ifp)) {
439 			flags = (flags & NHR_REF) | NHR_COPY;
440 			error = rt_exportinfo(rt, nh, info, flags);
441 			RIB_RUNLOCK(rh);
442 
443 			return (error);
444 		}
445 	}
446 	RIB_RUNLOCK(rh);
447 
448 	return (ENOENT);
449 }
450 
451 /*
452  * Releases all references acquired by rib_lookup_info() when
453  * called with NHR_REF flags.
454  */
455 void
rib_free_info(struct rt_addrinfo * info)456 rib_free_info(struct rt_addrinfo *info)
457 {
458 
459 	ifa_free(info->rti_ifa);
460 	if_rele(info->rti_ifp);
461 }
462 
463 /*
464  * Delete Routes for a Network Interface
465  *
466  * Called for each routing entry via the rnh->rnh_walktree() call above
467  * to delete all route entries referencing a detaching network interface.
468  *
469  * Arguments:
470  *	rt	pointer to rtentry
471  *	nh	pointer to nhop
472  *	arg	argument passed to rnh->rnh_walktree() - detaching interface
473  *
474  * Returns:
475  *	0	successful
476  *	errno	failed - reason indicated
477  */
478 static int
rt_ifdelroute(const struct rtentry * rt,const struct nhop_object * nh,void * arg)479 rt_ifdelroute(const struct rtentry *rt, const struct nhop_object *nh, void *arg)
480 {
481 	struct ifnet	*ifp = arg;
482 
483 	if (nh->nh_ifp != ifp)
484 		return (0);
485 
486 	/*
487 	 * Protect (sorta) against walktree recursion problems
488 	 * with cloned routes
489 	 */
490 	if ((rt->rte_flags & RTF_UP) == 0)
491 		return (0);
492 
493 	return (1);
494 }
495 
496 void
rt_flushifroutes(struct ifnet * ifp)497 rt_flushifroutes(struct ifnet *ifp)
498 {
499 
500 	rib_foreach_table_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
501 }
502 
503 /*
504  * Tries to extract interface from RTAX_IFP passed in rt_addrinfo.
505  * Interface can be specified ether as interface index (sdl_index) or
506  * the interface name (sdl_data).
507  *
508  * Returns found ifp or NULL
509  */
510 static struct ifnet *
info_get_ifp(struct rt_addrinfo * info)511 info_get_ifp(struct rt_addrinfo *info)
512 {
513 	const struct sockaddr_dl *sdl;
514 
515 	sdl = (const struct sockaddr_dl *)info->rti_info[RTAX_IFP];
516 	if (sdl->sdl_family != AF_LINK)
517 		return (NULL);
518 
519 	if (sdl->sdl_index != 0)
520 		return (ifnet_byindex(sdl->sdl_index));
521 	if (sdl->sdl_nlen > 0) {
522 		char if_name[IF_NAMESIZE];
523 		if (sdl->sdl_nlen + offsetof(struct sockaddr_dl, sdl_data) > sdl->sdl_len)
524 			return (NULL);
525 		if (sdl->sdl_nlen >= IF_NAMESIZE)
526 			return (NULL);
527 		bzero(if_name, sizeof(if_name));
528 		memcpy(if_name, sdl->sdl_data, sdl->sdl_nlen);
529 		return (ifunit(if_name));
530 	}
531 
532 	return (NULL);
533 }
534 
535 /*
536  * Calculates proper ifa/ifp for the cases when gateway AF is different
537  * from dst AF.
538  *
539  * Returns 0 on success.
540  */
541 __noinline static int
rt_getifa_family(struct rt_addrinfo * info,uint32_t fibnum)542 rt_getifa_family(struct rt_addrinfo *info, uint32_t fibnum)
543 {
544 	if (info->rti_ifp == NULL) {
545 		struct ifaddr *ifa = NULL;
546 		/*
547 		 * No transmit interface specified. Guess it by checking gw sa.
548 		 */
549 		const struct sockaddr *gw = info->rti_info[RTAX_GATEWAY];
550 		ifa = ifa_ifwithroute(RTF_GATEWAY, gw, gw, fibnum);
551 		if (ifa == NULL)
552 			return (ENETUNREACH);
553 		info->rti_ifp = ifa->ifa_ifp;
554 	}
555 
556 	/* Prefer address from outgoing interface */
557 	info->rti_ifa = ifaof_ifpforaddr(info->rti_info[RTAX_DST], info->rti_ifp);
558 #ifdef INET
559 	if (info->rti_ifa == NULL) {
560 		/* Use first found IPv4 address */
561 		bool loopback_ok = info->rti_ifp->if_flags & IFF_LOOPBACK;
562 		info->rti_ifa = (struct ifaddr *)in_findlocal(fibnum, loopback_ok);
563 	}
564 #endif
565 	if (info->rti_ifa == NULL)
566 		return (ENETUNREACH);
567 	return (0);
568 }
569 
570 /*
571  * Look up rt_addrinfo for a specific fib.
572  *
573  * Assume basic consistency checks are executed by callers:
574  * RTAX_DST exists, if RTF_GATEWAY is set, RTAX_GATEWAY exists as well.
575  */
576 int
rt_getifa_fib(struct rt_addrinfo * info,u_int fibnum)577 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
578 {
579 	const struct sockaddr *dst, *gateway, *ifaaddr;
580 	int error, flags;
581 
582 	dst = info->rti_info[RTAX_DST];
583 	gateway = info->rti_info[RTAX_GATEWAY];
584 	ifaaddr = info->rti_info[RTAX_IFA];
585 	flags = info->rti_flags;
586 
587 	/*
588 	 * ifp may be specified by sockaddr_dl
589 	 * when protocol address is ambiguous.
590 	 */
591 	error = 0;
592 
593 	/* If we have interface specified by RTAX_IFP address, try to use it */
594 	if ((info->rti_ifp == NULL) && (info->rti_info[RTAX_IFP] != NULL))
595 		info->rti_ifp = info_get_ifp(info);
596 	/*
597 	 * If we have source address specified, try to find it
598 	 * TODO: avoid enumerating all ifas on all interfaces.
599 	 */
600 	if (info->rti_ifa == NULL && ifaaddr != NULL)
601 		info->rti_ifa = ifa_ifwithaddr(ifaaddr);
602 	if ((info->rti_ifa == NULL) && ((info->rti_flags & RTF_GATEWAY) != 0) &&
603 	    (gateway->sa_family != dst->sa_family))
604 		return (rt_getifa_family(info, fibnum));
605 	if (info->rti_ifa == NULL) {
606 		const struct sockaddr *sa;
607 
608 		/*
609 		 * Most common use case for the userland-supplied routes.
610 		 *
611 		 * Choose sockaddr to select ifa.
612 		 * -- if ifp is set --
613 		 * Order of preference:
614 		 * 1) IFA address
615 		 * 2) gateway address
616 		 *   Note: for interface routes link-level gateway address
617 		 *     is specified to indicate the interface index without
618 		 *     specifying RTF_GATEWAY. In this case, ignore gateway
619 		 *   Note: gateway AF may be different from dst AF. In this case,
620 		 *   ignore gateway
621 		 * 3) final destination.
622 		 * 4) if all of these fails, try to get at least link-level ifa.
623 		 * -- else --
624 		 * try to lookup gateway or dst in the routing table to get ifa
625 		 */
626 		if (info->rti_info[RTAX_IFA] != NULL)
627 			sa = info->rti_info[RTAX_IFA];
628 		else if ((info->rti_flags & RTF_GATEWAY) != 0 &&
629 		    gateway->sa_family == dst->sa_family)
630 			sa = gateway;
631 		else
632 			sa = dst;
633 		if (info->rti_ifp != NULL) {
634 			info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
635 			/* Case 4 */
636 			if (info->rti_ifa == NULL && gateway != NULL)
637 				info->rti_ifa = ifaof_ifpforaddr(gateway, info->rti_ifp);
638 		} else if (dst != NULL && gateway != NULL)
639 			info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
640 							fibnum);
641 		else if (sa != NULL)
642 			info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
643 							fibnum);
644 	}
645 	if (info->rti_ifa != NULL) {
646 		if (info->rti_ifp == NULL)
647 			info->rti_ifp = info->rti_ifa->ifa_ifp;
648 	} else
649 		error = ENETUNREACH;
650 	return (error);
651 }
652 
653 void
rt_updatemtu(struct ifnet * ifp)654 rt_updatemtu(struct ifnet *ifp)
655 {
656 	struct rib_head *rnh;
657 	int mtu;
658 	int i, j;
659 
660 	/*
661 	 * Try to update rt_mtu for all routes using this interface
662 	 * Unfortunately the only way to do this is to traverse all
663 	 * routing tables in all fibs/domains.
664 	 */
665 	for (i = 1; i <= AF_MAX; i++) {
666 		mtu = if_getmtu_family(ifp, i);
667 		for (j = 0; j < rt_numfibs; j++) {
668 			rnh = rt_tables_get_rnh(j, i);
669 			if (rnh == NULL)
670 				continue;
671 			nhops_update_ifmtu(rnh, ifp, mtu);
672 		}
673 	}
674 }
675 
676 #if 0
677 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
678 int rt_print(char *buf, int buflen, struct rtentry *rt);
679 
680 int
681 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
682 {
683 	void *paddr = NULL;
684 
685 	switch (s->sa_family) {
686 	case AF_INET:
687 		paddr = &((struct sockaddr_in *)s)->sin_addr;
688 		break;
689 	case AF_INET6:
690 		paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
691 		break;
692 	}
693 
694 	if (paddr == NULL)
695 		return (0);
696 
697 	if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
698 		return (0);
699 
700 	return (strlen(buf));
701 }
702 
703 int
704 rt_print(char *buf, int buflen, struct rtentry *rt)
705 {
706 	struct sockaddr *addr, *mask;
707 	int i = 0;
708 
709 	addr = rt_key(rt);
710 	mask = rt_mask(rt);
711 
712 	i = p_sockaddr(buf, buflen, addr);
713 	if (!(rt->rt_flags & RTF_HOST)) {
714 		buf[i++] = '/';
715 		i += p_sockaddr(buf + i, buflen - i, mask);
716 	}
717 
718 	if (rt->rt_flags & RTF_GATEWAY) {
719 		buf[i++] = '>';
720 		i += p_sockaddr(buf + i, buflen - i, &rt->rt_nhop->gw_sa);
721 	}
722 
723 	return (i);
724 }
725 #endif
726 
727 void
rt_maskedcopy(struct sockaddr * src,struct sockaddr * dst,struct sockaddr * netmask)728 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
729 {
730 	u_char *cp1 = (u_char *)src;
731 	u_char *cp2 = (u_char *)dst;
732 	u_char *cp3 = (u_char *)netmask;
733 	u_char *cplim = cp2 + *cp3;
734 	u_char *cplim2 = cp2 + *cp1;
735 
736 	*cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
737 	cp3 += 2;
738 	if (cplim > cplim2)
739 		cplim = cplim2;
740 	while (cp2 < cplim)
741 		*cp2++ = *cp1++ & *cp3++;
742 	if (cp2 < cplim2)
743 		bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
744 }
745 
746 /*
747  * Announce interface address arrival/withdraw
748  * Returns 0 on success.
749  */
750 int
rt_addrmsg(int cmd,struct ifaddr * ifa,int fibnum)751 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
752 {
753 
754 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
755 	    ("unexpected cmd %d", cmd));
756 	KASSERT((fibnum >= 0 && fibnum < rt_numfibs),
757 	    ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
758 
759 	EVENTHANDLER_DIRECT_INVOKE(rt_addrmsg, ifa, cmd);
760 
761 	if (V_rt_add_addr_allfibs)
762 		fibnum = RT_ALL_FIBS;
763 	return (rtsock_addrmsg(cmd, ifa, fibnum));
764 }
765 
766 /*
767  * Announce kernel-originated route addition/removal to rtsock based on @rt data.
768  * cmd: RTM_ cmd
769  * @rt: valid rtentry
770  * @nh: nhop object to announce
771  * @fibnum: fib id or RT_ALL_FIBS
772  *
773  * Returns 0 on success.
774  */
775 int
rt_routemsg(int cmd,struct rtentry * rt,struct nhop_object * nh,int fibnum)776 rt_routemsg(int cmd, struct rtentry *rt, struct nhop_object *nh,
777     int fibnum)
778 {
779 
780 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
781 	    ("unexpected cmd %d", cmd));
782 
783 	KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
784 	    ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
785 
786 	KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
787 
788 	return (rtsock_routemsg(cmd, rt, nh, fibnum));
789 }
790 
791 /*
792  * Announce kernel-originated route addition/removal to rtsock based on @rt data.
793  * cmd: RTM_ cmd
794  * @info: addrinfo structure with valid data.
795  * @fibnum: fib id or RT_ALL_FIBS
796  *
797  * Returns 0 on success.
798  */
799 int
rt_routemsg_info(int cmd,struct rt_addrinfo * info,int fibnum)800 rt_routemsg_info(int cmd, struct rt_addrinfo *info, int fibnum)
801 {
802 
803 	KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE || cmd == RTM_CHANGE,
804 	    ("unexpected cmd %d", cmd));
805 
806 	KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
807 	    ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
808 
809 	KASSERT(info->rti_info[RTAX_DST] != NULL, (":%s: RTAX_DST must be supplied", __func__));
810 
811 	return (rtsock_routemsg_info(cmd, info, fibnum));
812 }
813