xref: /netbsd-src/sys/net/rtsock.c (revision c0179c282a5968435315a82f4128c61372c68fc3)
1 /*	$NetBSD: rtsock.c,v 1.91 2006/11/13 19:16:01 dyoung Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * 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 project 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 PROJECT 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 PROJECT 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 
32 /*
33  * Copyright (c) 1988, 1991, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
61  */
62 
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: rtsock.c,v 1.91 2006/11/13 19:16:01 dyoung Exp $");
65 
66 #include "opt_inet.h"
67 
68 #include <sys/param.h>
69 #include <sys/systm.h>
70 #include <sys/proc.h>
71 #include <sys/mbuf.h>
72 #include <sys/socket.h>
73 #include <sys/socketvar.h>
74 #include <sys/domain.h>
75 #include <sys/protosw.h>
76 #include <sys/sysctl.h>
77 #include <sys/kauth.h>
78 #ifdef RTSOCK_DEBUG
79 #include <netinet/in.h>
80 #endif /* RTSOCK_DEBUG */
81 
82 #include <net/if.h>
83 #include <net/route.h>
84 #include <net/raw_cb.h>
85 
86 #include <machine/stdarg.h>
87 
88 DOMAIN_DEFINE(routedomain);	/* forward declare and add to link set */
89 
90 struct	sockaddr route_dst = { .sa_len = 2, .sa_family = PF_ROUTE, };
91 struct	sockaddr route_src = { .sa_len = 2, .sa_family = PF_ROUTE, };
92 struct	sockproto route_proto = { .sp_family = PF_ROUTE, };
93 
94 struct walkarg {
95 	int	w_op;
96 	int	w_arg;
97 	int	w_given;
98 	int	w_needed;
99 	caddr_t	w_where;
100 	int	w_tmemsize;
101 	int	w_tmemneeded;
102 	caddr_t	w_tmem;
103 };
104 
105 static struct mbuf *rt_msg1(int, struct rt_addrinfo *, caddr_t, int);
106 static int rt_msg2(int, struct rt_addrinfo *, caddr_t, struct walkarg *, int *);
107 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
108 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
109     struct rt_addrinfo *);
110 static int sysctl_dumpentry(struct radix_node *, void *);
111 static int sysctl_iflist(int, struct walkarg *, int);
112 static int sysctl_rtable(SYSCTLFN_PROTO);
113 static inline void rt_adjustcount(int, int);
114 
115 /* Sleazy use of local variables throughout file, warning!!!! */
116 #define dst	info.rti_info[RTAX_DST]
117 #define gate	info.rti_info[RTAX_GATEWAY]
118 #define netmask	info.rti_info[RTAX_NETMASK]
119 #define genmask	info.rti_info[RTAX_GENMASK]
120 #define ifpaddr	info.rti_info[RTAX_IFP]
121 #define ifaaddr	info.rti_info[RTAX_IFA]
122 #define brdaddr	info.rti_info[RTAX_BRD]
123 
124 static inline void
125 rt_adjustcount(int af, int cnt)
126 {
127 	route_cb.any_count += cnt;
128 	switch (af) {
129 	case AF_INET:
130 		route_cb.ip_count += cnt;
131 		return;
132 #ifdef INET6
133 	case AF_INET6:
134 		route_cb.ip6_count += cnt;
135 		return;
136 #endif
137 	case AF_IPX:
138 		route_cb.ipx_count += cnt;
139 		return;
140 	case AF_NS:
141 		route_cb.ns_count += cnt;
142 		return;
143 	case AF_ISO:
144 		route_cb.iso_count += cnt;
145 		return;
146 	}
147 }
148 
149 /*ARGSUSED*/
150 int
151 route_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
152 	struct mbuf *control, struct lwp *l)
153 {
154 	int error = 0;
155 	struct rawcb *rp = sotorawcb(so);
156 	int s;
157 
158 	if (req == PRU_ATTACH) {
159 		MALLOC(rp, struct rawcb *, sizeof(*rp), M_PCB, M_WAITOK);
160 		if ((so->so_pcb = rp) != NULL)
161 			memset(so->so_pcb, 0, sizeof(*rp));
162 
163 	}
164 	if (req == PRU_DETACH && rp)
165 		rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
166 	s = splsoftnet();
167 
168 	/*
169 	 * Don't call raw_usrreq() in the attach case, because
170 	 * we want to allow non-privileged processes to listen on
171 	 * and send "safe" commands to the routing socket.
172 	 */
173 	if (req == PRU_ATTACH) {
174 		if (l == 0)
175 			error = EACCES;
176 		else
177 			error = raw_attach(so, (int)(long)nam);
178 	} else
179 		error = raw_usrreq(so, req, m, nam, control, l);
180 
181 	rp = sotorawcb(so);
182 	if (req == PRU_ATTACH && rp) {
183 		if (error) {
184 			free((caddr_t)rp, M_PCB);
185 			splx(s);
186 			return (error);
187 		}
188 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
189 		rp->rcb_laddr = &route_src;
190 		rp->rcb_faddr = &route_dst;
191 		soisconnected(so);
192 		so->so_options |= SO_USELOOPBACK;
193 	}
194 	splx(s);
195 	return (error);
196 }
197 
198 /*ARGSUSED*/
199 int
200 route_output(struct mbuf *m, ...)
201 {
202 	struct rt_msghdr *rtm = 0;
203 	struct radix_node *rn = 0;
204 	struct rtentry *rt = 0;
205 	struct rtentry *saved_nrt = 0;
206 	struct radix_node_head *rnh;
207 	struct rt_addrinfo info;
208 	int len, error = 0;
209 	struct ifnet *ifp = 0;
210 	struct ifaddr *ifa = 0;
211 	struct socket *so;
212 	va_list ap;
213 	sa_family_t family;
214 
215 	va_start(ap, m);
216 	so = va_arg(ap, struct socket *);
217 	va_end(ap);
218 
219 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
220 	if (m == 0 || ((m->m_len < sizeof(int32_t)) &&
221 	   (m = m_pullup(m, sizeof(int32_t))) == 0))
222 		return (ENOBUFS);
223 	if ((m->m_flags & M_PKTHDR) == 0)
224 		panic("route_output");
225 	len = m->m_pkthdr.len;
226 	if (len < sizeof(*rtm) ||
227 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
228 		dst = 0;
229 		senderr(EINVAL);
230 	}
231 	R_Malloc(rtm, struct rt_msghdr *, len);
232 	if (rtm == 0) {
233 		dst = 0;
234 		senderr(ENOBUFS);
235 	}
236 	m_copydata(m, 0, len, (caddr_t)rtm);
237 	if (rtm->rtm_version != RTM_VERSION) {
238 		dst = 0;
239 		senderr(EPROTONOSUPPORT);
240 	}
241 	rtm->rtm_pid = curproc->p_pid;
242 	memset(&info, 0, sizeof(info));
243 	info.rti_addrs = rtm->rtm_addrs;
244 	if (rt_xaddrs(rtm->rtm_type, (caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
245 		senderr(EINVAL);
246 	info.rti_flags = rtm->rtm_flags;
247 #ifdef RTSOCK_DEBUG
248 	if (dst->sa_family == AF_INET) {
249 		printf("%s: extracted dst %s\n", __func__,
250 		    inet_ntoa(((const struct sockaddr_in *)dst)->sin_addr));
251 	}
252 #endif /* RTSOCK_DEBUG */
253 	if (dst == 0 || (dst->sa_family >= AF_MAX))
254 		senderr(EINVAL);
255 	if (gate != 0 && (gate->sa_family >= AF_MAX))
256 		senderr(EINVAL);
257 	if (genmask) {
258 		struct radix_node *t;
259 		t = rn_addmask(genmask, 0, 1);
260 		if (t && genmask->sa_len >= ((const struct sockaddr *)t->rn_key)->sa_len &&
261 		    Bcmp((const char *const *)genmask + 1, (const char *const *)t->rn_key + 1,
262 		    ((const struct sockaddr *)t->rn_key)->sa_len) - 1)
263 			genmask = (const struct sockaddr *)(t->rn_key);
264 		else
265 			senderr(ENOBUFS);
266 	}
267 
268 	/*
269 	 * Verify that the caller has the appropriate privilege; RTM_GET
270 	 * is the only operation the non-superuser is allowed.
271 	 */
272 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
273 	    0, rtm, NULL, NULL) != 0)
274 		senderr(EACCES);
275 
276 	switch (rtm->rtm_type) {
277 
278 	case RTM_ADD:
279 		if (gate == 0)
280 			senderr(EINVAL);
281 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
282 		if (error == 0 && saved_nrt) {
283 			rt_setmetrics(rtm->rtm_inits,
284 			    &rtm->rtm_rmx, &saved_nrt->rt_rmx);
285 			saved_nrt->rt_refcnt--;
286 			saved_nrt->rt_genmask = genmask;
287 		}
288 		break;
289 
290 	case RTM_DELETE:
291 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
292 		if (error == 0) {
293 			(rt = saved_nrt)->rt_refcnt++;
294 			goto report;
295 		}
296 		break;
297 
298 	case RTM_GET:
299 	case RTM_CHANGE:
300 	case RTM_LOCK:
301 		if ((rnh = rt_tables[dst->sa_family]) == 0) {
302 			senderr(EAFNOSUPPORT);
303 		}
304 		rn = rnh->rnh_lookup(dst, netmask, rnh);
305 		if (rn == NULL || (rn->rn_flags & RNF_ROOT) != 0) {
306 			senderr(ESRCH);
307 		}
308 		rt = (struct rtentry *)rn;
309 		rt->rt_refcnt++;
310 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
311 			struct radix_node *rnn;
312 			extern struct radix_node_head *mask_rnhead;
313 
314 			if (Bcmp(dst, rt_key(rt), dst->sa_len) != 0)
315 				senderr(ESRCH);
316 			if (netmask && (rnn = rn_search(netmask,
317 					    mask_rnhead->rnh_treetop)))
318 				netmask = (const struct sockaddr *)rnn->rn_key;
319 			for (rnn = rt->rt_nodes; rnn; rnn = rnn->rn_dupedkey)
320 				if (netmask == (const struct sockaddr *)rnn->rn_mask)
321 					break;
322 			if (rnn == 0)
323 				senderr(ETOOMANYREFS);
324 			rt = (struct rtentry *)rnn;
325 		}
326 
327 		switch (rtm->rtm_type) {
328 		case RTM_GET:
329 		report:
330 			dst = rt_key(rt);
331 			gate = rt->rt_gateway;
332 			netmask = rt_mask(rt);
333 			genmask = rt->rt_genmask;
334 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
335 				;
336 			else if ((ifp = rt->rt_ifp) != NULL) {
337 				const struct ifaddr *rtifa;
338 				ifpaddr = TAILQ_FIRST(&ifp->if_addrlist)->ifa_addr;
339                                 /* rtifa used to be simply rt->rt_ifa.
340                                  * If rt->rt_ifa != NULL, then
341                                  * rt_get_ifa() != NULL.  So this
342                                  * ought to still be safe. --dyoung
343 				 */
344 				rtifa = rt_get_ifa(rt);
345 				ifaaddr = rtifa->ifa_addr;
346 #ifdef RTSOCK_DEBUG
347 				if (ifaaddr->sa_family == AF_INET) {
348 					printf("%s: copying out RTAX_IFA %s ",
349 					    __func__,
350 					    inet_ntoa(((const struct sockaddr_in *)ifaaddr)->sin_addr));
351 					printf("for dst %s ifa_getifa %p ifa_seqno %p\n",
352 					    inet_ntoa(((const struct sockaddr_in *)dst)->sin_addr),
353 					    (void *)rtifa->ifa_getifa, rtifa->ifa_seqno);
354 				}
355 #endif /* RTSOCK_DEBUG */
356 				if (ifp->if_flags & IFF_POINTOPOINT)
357 					brdaddr = rtifa->ifa_dstaddr;
358 				else
359 					brdaddr = 0;
360 				rtm->rtm_index = ifp->if_index;
361 			} else {
362 				ifpaddr = 0;
363 				ifaaddr = 0;
364 			}
365 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)0,
366 			    (struct walkarg *)0, &len);
367 			if (len > rtm->rtm_msglen) {
368 				struct rt_msghdr *new_rtm;
369 				R_Malloc(new_rtm, struct rt_msghdr *, len);
370 				if (new_rtm == 0)
371 					senderr(ENOBUFS);
372 				Bcopy(rtm, new_rtm, rtm->rtm_msglen);
373 				Free(rtm); rtm = new_rtm;
374 			}
375 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm,
376 			    (struct walkarg *)0, 0);
377 			rtm->rtm_flags = rt->rt_flags;
378 			rtm->rtm_rmx = rt->rt_rmx;
379 			rtm->rtm_addrs = info.rti_addrs;
380 			break;
381 
382 		case RTM_CHANGE:
383 			/*
384 			 * new gateway could require new ifaddr, ifp;
385 			 * flags may also be different; ifp may be specified
386 			 * by ll sockaddr when protocol address is ambiguous
387 			 */
388 			if ((error = rt_getifa(&info)) != 0)
389 				senderr(error);
390 			if (gate && rt_setgate(rt, rt_key(rt), gate))
391 				senderr(EDQUOT);
392 			/* new gateway could require new ifaddr, ifp;
393 			   flags may also be different; ifp may be specified
394 			   by ll sockaddr when protocol address is ambiguous */
395 			if (ifpaddr && (ifa = ifa_ifwithnet(ifpaddr)) &&
396 			    (ifp = ifa->ifa_ifp) && (ifaaddr || gate))
397 				ifa = ifaof_ifpforaddr(ifaaddr ? ifaaddr : gate,
398 				    ifp);
399 			else if ((ifaaddr && (ifa = ifa_ifwithaddr(ifaaddr))) ||
400 			    (gate && (ifa = ifa_ifwithroute(rt->rt_flags,
401 			    rt_key(rt), gate))))
402 				ifp = ifa->ifa_ifp;
403 			if (ifa) {
404 				struct ifaddr *oifa = rt->rt_ifa;
405 				if (oifa != ifa) {
406 					if (oifa && oifa->ifa_rtrequest) {
407 						oifa->ifa_rtrequest(RTM_DELETE,
408 						    rt, &info);
409 					}
410 					rt_replace_ifa(rt, ifa);
411 					rt->rt_ifp = ifp;
412 				}
413 			}
414 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
415 			    &rt->rt_rmx);
416 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
417 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
418 			if (genmask)
419 				rt->rt_genmask = genmask;
420 			/*
421 			 * Fall into
422 			 */
423 		case RTM_LOCK:
424 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
425 			rt->rt_rmx.rmx_locks |=
426 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
427 			break;
428 		}
429 		break;
430 
431 	default:
432 		senderr(EOPNOTSUPP);
433 	}
434 
435 flush:
436 	if (rtm) {
437 		if (error)
438 			rtm->rtm_errno = error;
439 		else
440 			rtm->rtm_flags |= RTF_DONE;
441 	}
442 	family = dst ? dst->sa_family : 0;
443 	if (rt)
444 		rtfree(rt);
445     {
446 	struct rawcb *rp = 0;
447 	/*
448 	 * Check to see if we don't want our own messages.
449 	 */
450 	if ((so->so_options & SO_USELOOPBACK) == 0) {
451 		if (route_cb.any_count <= 1) {
452 			if (rtm)
453 				Free(rtm);
454 			m_freem(m);
455 			return (error);
456 		}
457 		/* There is another listener, so construct message */
458 		rp = sotorawcb(so);
459 	}
460 	if (rtm) {
461 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
462 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
463 			m_freem(m);
464 			m = NULL;
465 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
466 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
467 		Free(rtm);
468 	}
469 	if (rp)
470 		rp->rcb_proto.sp_family = 0; /* Avoid us */
471 	if (family)
472 		route_proto.sp_protocol = family;
473 	if (m)
474 		raw_input(m, &route_proto, &route_src, &route_dst);
475 	if (rp)
476 		rp->rcb_proto.sp_family = PF_ROUTE;
477     }
478 	return (error);
479 }
480 
481 void
482 rt_setmetrics(u_long which, const struct rt_metrics *in, struct rt_metrics *out)
483 {
484 #define metric(f, e) if (which & (f)) out->e = in->e;
485 	metric(RTV_RPIPE, rmx_recvpipe);
486 	metric(RTV_SPIPE, rmx_sendpipe);
487 	metric(RTV_SSTHRESH, rmx_ssthresh);
488 	metric(RTV_RTT, rmx_rtt);
489 	metric(RTV_RTTVAR, rmx_rttvar);
490 	metric(RTV_HOPCOUNT, rmx_hopcount);
491 	metric(RTV_MTU, rmx_mtu);
492 	metric(RTV_EXPIRE, rmx_expire);
493 #undef metric
494 }
495 
496 #define ROUNDUP(a) \
497 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
498 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
499 
500 static int
501 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim, struct rt_addrinfo *rtinfo)
502 {
503 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
504 	int i;
505 
506 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
507 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
508 			continue;
509 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
510 		ADVANCE(cp, sa);
511 	}
512 
513 	/* Check for extra addresses specified, except RTM_GET asking for interface info.  */
514 	if (rtmtype == RTM_GET) {
515 		if (((rtinfo->rti_addrs & (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
516 			return (1);
517 	} else {
518 		if ((rtinfo->rti_addrs & (~0 << i)) != 0)
519 			return (1);
520 	}
521 	/* Check for bad data length.  */
522 	if (cp != cplim) {
523 		if (i == RTAX_NETMASK + 1 && sa &&
524 		    cp - ROUNDUP(sa->sa_len) + sa->sa_len == cplim)
525 			/*
526 			 * The last sockaddr was netmask.
527 			 * We accept this for now for the sake of old
528 			 * binaries or third party softwares.
529 			 */
530 			;
531 		else
532 			return (1);
533 	}
534 	return (0);
535 }
536 
537 static struct mbuf *
538 rt_msg1(int type, struct rt_addrinfo *rtinfo, caddr_t data, int datalen)
539 {
540 	struct rt_msghdr *rtm;
541 	struct mbuf *m;
542 	int i;
543 	const struct sockaddr *sa;
544 	int len, dlen;
545 
546 	m = m_gethdr(M_DONTWAIT, MT_DATA);
547 	if (m == 0)
548 		return (m);
549 	MCLAIM(m, &routedomain.dom_mowner);
550 	switch (type) {
551 
552 	case RTM_DELADDR:
553 	case RTM_NEWADDR:
554 		len = sizeof(struct ifa_msghdr);
555 		break;
556 
557 #ifdef COMPAT_14
558 	case RTM_OIFINFO:
559 		len = sizeof(struct if_msghdr14);
560 		break;
561 #endif
562 
563 	case RTM_IFINFO:
564 		len = sizeof(struct if_msghdr);
565 		break;
566 
567 	case RTM_IFANNOUNCE:
568 	case RTM_IEEE80211:
569 		len = sizeof(struct if_announcemsghdr);
570 		break;
571 
572 	default:
573 		len = sizeof(struct rt_msghdr);
574 	}
575 	if (len > MHLEN + MLEN)
576 		panic("rt_msg1: message too long");
577 	else if (len > MHLEN) {
578 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
579 		if (m->m_next == NULL) {
580 			m_freem(m);
581 			return (NULL);
582 		}
583 		MCLAIM(m->m_next, m->m_owner);
584 		m->m_pkthdr.len = len;
585 		m->m_len = MHLEN;
586 		m->m_next->m_len = len - MHLEN;
587 	} else {
588 		m->m_pkthdr.len = m->m_len = len;
589 	}
590 	m->m_pkthdr.rcvif = 0;
591 	m_copyback(m, 0, datalen, data);
592 	rtm = mtod(m, struct rt_msghdr *);
593 	for (i = 0; i < RTAX_MAX; i++) {
594 		if ((sa = rtinfo->rti_info[i]) == NULL)
595 			continue;
596 		rtinfo->rti_addrs |= (1 << i);
597 		dlen = ROUNDUP(sa->sa_len);
598 		m_copyback(m, len, dlen, sa);
599 		len += dlen;
600 	}
601 	if (m->m_pkthdr.len != len) {
602 		m_freem(m);
603 		return (NULL);
604 	}
605 	rtm->rtm_msglen = len;
606 	rtm->rtm_version = RTM_VERSION;
607 	rtm->rtm_type = type;
608 	return (m);
609 }
610 
611 /*
612  * rt_msg2
613  *
614  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
615  *		returns the length of the message in 'lenp'.
616  *
617  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
618  *	the message
619  * otherwise walkarg's w_needed is updated and if the user buffer is
620  *	specified and w_needed indicates space exists the information is copied
621  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
622  *	if the allocation fails ENOBUFS is returned.
623  */
624 static int
625 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w,
626 	int *lenp)
627 {
628 	int i;
629 	int len, dlen, second_time = 0;
630 	caddr_t cp0;
631 
632 	rtinfo->rti_addrs = 0;
633 again:
634 	switch (type) {
635 
636 	case RTM_DELADDR:
637 	case RTM_NEWADDR:
638 		len = sizeof(struct ifa_msghdr);
639 		break;
640 #ifdef COMPAT_14
641 	case RTM_OIFINFO:
642 		len = sizeof(struct if_msghdr14);
643 		break;
644 #endif
645 
646 	case RTM_IFINFO:
647 		len = sizeof(struct if_msghdr);
648 		break;
649 
650 	default:
651 		len = sizeof(struct rt_msghdr);
652 	}
653 	if ((cp0 = cp) != NULL)
654 		cp += len;
655 	for (i = 0; i < RTAX_MAX; i++) {
656 		const struct sockaddr *sa;
657 
658 		if ((sa = rtinfo->rti_info[i]) == 0)
659 			continue;
660 		rtinfo->rti_addrs |= (1 << i);
661 		dlen = ROUNDUP(sa->sa_len);
662 		if (cp) {
663 			bcopy(sa, cp, (unsigned)dlen);
664 			cp += dlen;
665 		}
666 		len += dlen;
667 	}
668 	if (cp == 0 && w != NULL && !second_time) {
669 		struct walkarg *rw = w;
670 
671 		rw->w_needed += len;
672 		if (rw->w_needed <= 0 && rw->w_where) {
673 			if (rw->w_tmemsize < len) {
674 				if (rw->w_tmem)
675 					free(rw->w_tmem, M_RTABLE);
676 				rw->w_tmem = (caddr_t) malloc(len, M_RTABLE,
677 				    M_NOWAIT);
678 				if (rw->w_tmem)
679 					rw->w_tmemsize = len;
680 			}
681 			if (rw->w_tmem) {
682 				cp = rw->w_tmem;
683 				second_time = 1;
684 				goto again;
685 			} else {
686 				rw->w_tmemneeded = len;
687 				return (ENOBUFS);
688 			}
689 		}
690 	}
691 	if (cp) {
692 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
693 
694 		rtm->rtm_version = RTM_VERSION;
695 		rtm->rtm_type = type;
696 		rtm->rtm_msglen = len;
697 	}
698 	if (lenp)
699 		*lenp = len;
700 	return (0);
701 }
702 
703 /*
704  * This routine is called to generate a message from the routing
705  * socket indicating that a redirect has occurred, a routing lookup
706  * has failed, or that a protocol has detected timeouts to a particular
707  * destination.
708  */
709 void
710 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
711 {
712 	struct rt_msghdr rtm;
713 	struct mbuf *m;
714 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
715 
716 	if (route_cb.any_count == 0)
717 		return;
718 	memset(&rtm, 0, sizeof(rtm));
719 	rtm.rtm_flags = RTF_DONE | flags;
720 	rtm.rtm_errno = error;
721 	m = rt_msg1(type, rtinfo, (caddr_t)&rtm, sizeof(rtm));
722 	if (m == 0)
723 		return;
724 	mtod(m, struct rt_msghdr *)->rtm_addrs = rtinfo->rti_addrs;
725 	route_proto.sp_protocol = sa ? sa->sa_family : 0;
726 	raw_input(m, &route_proto, &route_src, &route_dst);
727 }
728 
729 /*
730  * This routine is called to generate a message from the routing
731  * socket indicating that the status of a network interface has changed.
732  */
733 void
734 rt_ifmsg(struct ifnet *ifp)
735 {
736 	struct if_msghdr ifm;
737 #ifdef COMPAT_14
738 	struct if_msghdr14 oifm;
739 #endif
740 	struct mbuf *m;
741 	struct rt_addrinfo info;
742 
743 	if (route_cb.any_count == 0)
744 		return;
745 	memset(&info, 0, sizeof(info));
746 	memset(&ifm, 0, sizeof(ifm));
747 	ifm.ifm_index = ifp->if_index;
748 	ifm.ifm_flags = ifp->if_flags;
749 	ifm.ifm_data = ifp->if_data;
750 	ifm.ifm_addrs = 0;
751 	m = rt_msg1(RTM_IFINFO, &info, (caddr_t)&ifm, sizeof(ifm));
752 	if (m == 0)
753 		return;
754 	route_proto.sp_protocol = 0;
755 	raw_input(m, &route_proto, &route_src, &route_dst);
756 #ifdef COMPAT_14
757 	memset(&info, 0, sizeof(info));
758 	memset(&oifm, 0, sizeof(oifm));
759 	oifm.ifm_index = ifp->if_index;
760 	oifm.ifm_flags = ifp->if_flags;
761 	oifm.ifm_data.ifi_type = ifp->if_data.ifi_type;
762 	oifm.ifm_data.ifi_addrlen = ifp->if_data.ifi_addrlen;
763 	oifm.ifm_data.ifi_hdrlen = ifp->if_data.ifi_hdrlen;
764 	oifm.ifm_data.ifi_mtu = ifp->if_data.ifi_mtu;
765 	oifm.ifm_data.ifi_metric = ifp->if_data.ifi_metric;
766 	oifm.ifm_data.ifi_baudrate = ifp->if_data.ifi_baudrate;
767 	oifm.ifm_data.ifi_ipackets = ifp->if_data.ifi_ipackets;
768 	oifm.ifm_data.ifi_ierrors = ifp->if_data.ifi_ierrors;
769 	oifm.ifm_data.ifi_opackets = ifp->if_data.ifi_opackets;
770 	oifm.ifm_data.ifi_oerrors = ifp->if_data.ifi_oerrors;
771 	oifm.ifm_data.ifi_collisions = ifp->if_data.ifi_collisions;
772 	oifm.ifm_data.ifi_ibytes = ifp->if_data.ifi_ibytes;
773 	oifm.ifm_data.ifi_obytes = ifp->if_data.ifi_obytes;
774 	oifm.ifm_data.ifi_imcasts = ifp->if_data.ifi_imcasts;
775 	oifm.ifm_data.ifi_omcasts = ifp->if_data.ifi_omcasts;
776 	oifm.ifm_data.ifi_iqdrops = ifp->if_data.ifi_iqdrops;
777 	oifm.ifm_data.ifi_noproto = ifp->if_data.ifi_noproto;
778 	oifm.ifm_data.ifi_lastchange = ifp->if_data.ifi_lastchange;
779 	oifm.ifm_addrs = 0;
780 	m = rt_msg1(RTM_OIFINFO, &info, (caddr_t)&oifm, sizeof(oifm));
781 	if (m == 0)
782 		return;
783 	route_proto.sp_protocol = 0;
784 	raw_input(m, &route_proto, &route_src, &route_dst);
785 #endif
786 }
787 
788 /*
789  * This is called to generate messages from the routing socket
790  * indicating a network interface has had addresses associated with it.
791  * if we ever reverse the logic and replace messages TO the routing
792  * socket indicate a request to configure interfaces, then it will
793  * be unnecessary as the routing socket will automatically generate
794  * copies of it.
795  */
796 void
797 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
798 {
799 	struct rt_addrinfo info;
800 	struct sockaddr *sa = NULL;
801 	int pass;
802 	struct mbuf *m = NULL;
803 	struct ifnet *ifp = ifa->ifa_ifp;
804 
805 	if (route_cb.any_count == 0)
806 		return;
807 	for (pass = 1; pass < 3; pass++) {
808 		memset(&info, 0, sizeof(info));
809 		if ((cmd == RTM_ADD && pass == 1) ||
810 		    (cmd == RTM_DELETE && pass == 2)) {
811 			struct ifa_msghdr ifam;
812 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
813 
814 			ifaaddr = sa = ifa->ifa_addr;
815 			ifpaddr = TAILQ_FIRST(&ifp->if_addrlist)->ifa_addr;
816 			netmask = ifa->ifa_netmask;
817 			brdaddr = ifa->ifa_dstaddr;
818 			memset(&ifam, 0, sizeof(ifam));
819 			ifam.ifam_index = ifp->if_index;
820 			ifam.ifam_metric = ifa->ifa_metric;
821 			ifam.ifam_flags = ifa->ifa_flags;
822 			m = rt_msg1(ncmd, &info, (caddr_t)&ifam, sizeof(ifam));
823 			if (m == NULL)
824 				continue;
825 			mtod(m, struct ifa_msghdr *)->ifam_addrs =
826 			    info.rti_addrs;
827 		}
828 		if ((cmd == RTM_ADD && pass == 2) ||
829 		    (cmd == RTM_DELETE && pass == 1)) {
830 			struct rt_msghdr rtm;
831 
832 			if (rt == 0)
833 				continue;
834 			netmask = rt_mask(rt);
835 			dst = sa = rt_key(rt);
836 			gate = rt->rt_gateway;
837 			memset(&rtm, 0, sizeof(rtm));
838 			rtm.rtm_index = ifp->if_index;
839 			rtm.rtm_flags |= rt->rt_flags;
840 			rtm.rtm_errno = error;
841 			m = rt_msg1(cmd, &info, (caddr_t)&rtm, sizeof(rtm));
842 			if (m == NULL)
843 				continue;
844 			mtod(m, struct rt_msghdr *)->rtm_addrs = info.rti_addrs;
845 		}
846 		route_proto.sp_protocol = sa ? sa->sa_family : 0;
847 		raw_input(m, &route_proto, &route_src, &route_dst);
848 	}
849 }
850 
851 static struct mbuf *
852 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
853     struct rt_addrinfo *info)
854 {
855 	struct if_announcemsghdr ifan;
856 
857 	memset(info, 0, sizeof(*info));
858 	memset(&ifan, 0, sizeof(ifan));
859 	ifan.ifan_index = ifp->if_index;
860 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
861 	ifan.ifan_what = what;
862 	return rt_msg1(type, info, (caddr_t)&ifan, sizeof(ifan));
863 }
864 
865 /*
866  * This is called to generate routing socket messages indicating
867  * network interface arrival and departure.
868  */
869 void
870 rt_ifannouncemsg(struct ifnet *ifp, int what)
871 {
872 	struct mbuf *m;
873 	struct rt_addrinfo info;
874 
875 	if (route_cb.any_count == 0)
876 		return;
877 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
878 	if (m == NULL)
879 		return;
880 	route_proto.sp_protocol = 0;
881 	raw_input(m, &route_proto, &route_src, &route_dst);
882 }
883 
884 /*
885  * This is called to generate routing socket messages indicating
886  * IEEE80211 wireless events.
887  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
888  */
889 void
890 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
891 {
892 	struct mbuf *m;
893 	struct rt_addrinfo info;
894 
895 	if (route_cb.any_count == 0)
896 		return;
897 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
898 	if (m == NULL)
899 		return;
900 	/*
901 	 * Append the ieee80211 data.  Try to stick it in the
902 	 * mbuf containing the ifannounce msg; otherwise allocate
903 	 * a new mbuf and append.
904 	 *
905 	 * NB: we assume m is a single mbuf.
906 	 */
907 	if (data_len > M_TRAILINGSPACE(m)) {
908 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
909 		if (n == NULL) {
910 			m_freem(m);
911 			return;
912 		}
913 		(void)memcpy(mtod(n, void *), data, data_len);
914 		n->m_len = data_len;
915 		m->m_next = n;
916 	} else if (data_len > 0) {
917 		(void)memcpy(mtod(m, u_int8_t *) + m->m_len, data, data_len);
918 		m->m_len += data_len;
919 	}
920 	if (m->m_flags & M_PKTHDR)
921 		m->m_pkthdr.len += data_len;
922 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
923 	route_proto.sp_protocol = 0;
924 	raw_input(m, &route_proto, &route_src, &route_dst);
925 }
926 
927 /*
928  * This is used in dumping the kernel table via sysctl().
929  */
930 static int
931 sysctl_dumpentry(struct radix_node *rn, void *v)
932 {
933 	struct walkarg *w = v;
934 	struct rtentry *rt = (struct rtentry *)rn;
935 	int error = 0, size;
936 	struct rt_addrinfo info;
937 
938 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
939 		return 0;
940 	memset(&info, 0, sizeof(info));
941 	dst = rt_key(rt);
942 	gate = rt->rt_gateway;
943 	netmask = rt_mask(rt);
944 	genmask = rt->rt_genmask;
945 	if (rt->rt_ifp) {
946 		const struct ifaddr *rtifa;
947 		ifpaddr = TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
948 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
949 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
950 		 * --dyoung
951 		 */
952 		rtifa = rt_get_ifa(rt);
953 		ifaaddr = rtifa->ifa_addr;
954 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
955 			brdaddr = rtifa->ifa_dstaddr;
956 	}
957 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
958 		return (error);
959 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
960 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
961 
962 		rtm->rtm_flags = rt->rt_flags;
963 		rtm->rtm_use = rt->rt_use;
964 		rtm->rtm_rmx = rt->rt_rmx;
965 		KASSERT(rt->rt_ifp != NULL);
966 		rtm->rtm_index = rt->rt_ifp->if_index;
967 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
968 		rtm->rtm_addrs = info.rti_addrs;
969 		if ((error = copyout(rtm, w->w_where, size)) != 0)
970 			w->w_where = NULL;
971 		else
972 			w->w_where += size;
973 	}
974 	return (error);
975 }
976 
977 static int
978 sysctl_iflist(int af, struct walkarg *w, int type)
979 {
980 	struct ifnet *ifp;
981 	struct ifaddr *ifa;
982 	struct	rt_addrinfo info;
983 	int	len, error = 0;
984 
985 	memset(&info, 0, sizeof(info));
986 	IFNET_FOREACH(ifp) {
987 		if (w->w_arg && w->w_arg != ifp->if_index)
988 			continue;
989 		ifa = TAILQ_FIRST(&ifp->if_addrlist);
990 		if (ifa == NULL)
991 			continue;
992 		ifpaddr = ifa->ifa_addr;
993 		switch (type) {
994 		case NET_RT_IFLIST:
995 			error =
996 			    rt_msg2(RTM_IFINFO, &info, (caddr_t)0, w, &len);
997 			break;
998 #ifdef COMPAT_14
999 		case NET_RT_OIFLIST:
1000 			error =
1001 			    rt_msg2(RTM_OIFINFO, &info, (caddr_t)0, w, &len);
1002 			break;
1003 #endif
1004 		default:
1005 			panic("sysctl_iflist(1)");
1006 		}
1007 		if (error)
1008 			return (error);
1009 		ifpaddr = 0;
1010 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1011 			switch (type) {
1012 			case NET_RT_IFLIST: {
1013 				struct if_msghdr *ifm;
1014 
1015 				ifm = (struct if_msghdr *)w->w_tmem;
1016 				ifm->ifm_index = ifp->if_index;
1017 				ifm->ifm_flags = ifp->if_flags;
1018 				ifm->ifm_data = ifp->if_data;
1019 				ifm->ifm_addrs = info.rti_addrs;
1020 				error = copyout(ifm, w->w_where, len);
1021 				if (error)
1022 					return (error);
1023 				w->w_where += len;
1024 				break;
1025 			}
1026 
1027 #ifdef COMPAT_14
1028 			case NET_RT_OIFLIST: {
1029 				struct if_msghdr14 *ifm;
1030 
1031 				ifm = (struct if_msghdr14 *)w->w_tmem;
1032 				ifm->ifm_index = ifp->if_index;
1033 				ifm->ifm_flags = ifp->if_flags;
1034 				ifm->ifm_data.ifi_type = ifp->if_data.ifi_type;
1035 				ifm->ifm_data.ifi_addrlen =
1036 				    ifp->if_data.ifi_addrlen;
1037 				ifm->ifm_data.ifi_hdrlen =
1038 				    ifp->if_data.ifi_hdrlen;
1039 				ifm->ifm_data.ifi_mtu = ifp->if_data.ifi_mtu;
1040 				ifm->ifm_data.ifi_metric =
1041 				    ifp->if_data.ifi_metric;
1042 				ifm->ifm_data.ifi_baudrate =
1043 				    ifp->if_data.ifi_baudrate;
1044 				ifm->ifm_data.ifi_ipackets =
1045 				    ifp->if_data.ifi_ipackets;
1046 				ifm->ifm_data.ifi_ierrors =
1047 				    ifp->if_data.ifi_ierrors;
1048 				ifm->ifm_data.ifi_opackets =
1049 				    ifp->if_data.ifi_opackets;
1050 				ifm->ifm_data.ifi_oerrors =
1051 				    ifp->if_data.ifi_oerrors;
1052 				ifm->ifm_data.ifi_collisions =
1053 				    ifp->if_data.ifi_collisions;
1054 				ifm->ifm_data.ifi_ibytes =
1055 				    ifp->if_data.ifi_ibytes;
1056 				ifm->ifm_data.ifi_obytes =
1057 				    ifp->if_data.ifi_obytes;
1058 				ifm->ifm_data.ifi_imcasts =
1059 				    ifp->if_data.ifi_imcasts;
1060 				ifm->ifm_data.ifi_omcasts =
1061 				    ifp->if_data.ifi_omcasts;
1062 				ifm->ifm_data.ifi_iqdrops =
1063 				    ifp->if_data.ifi_iqdrops;
1064 				ifm->ifm_data.ifi_noproto =
1065 				    ifp->if_data.ifi_noproto;
1066 				ifm->ifm_data.ifi_lastchange =
1067 				    ifp->if_data.ifi_lastchange;
1068 				ifm->ifm_addrs = info.rti_addrs;
1069 				error = copyout(ifm, w->w_where, len);
1070 				if (error)
1071 					return (error);
1072 				w->w_where += len;
1073 				break;
1074 			}
1075 #endif
1076 			default:
1077 				panic("sysctl_iflist(2)");
1078 			}
1079 		}
1080 		while ((ifa = TAILQ_NEXT(ifa, ifa_list)) != NULL) {
1081 			if (af && af != ifa->ifa_addr->sa_family)
1082 				continue;
1083 			ifaaddr = ifa->ifa_addr;
1084 			netmask = ifa->ifa_netmask;
1085 			brdaddr = ifa->ifa_dstaddr;
1086 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
1087 				return (error);
1088 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1089 				struct ifa_msghdr *ifam;
1090 
1091 				ifam = (struct ifa_msghdr *)w->w_tmem;
1092 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1093 				ifam->ifam_flags = ifa->ifa_flags;
1094 				ifam->ifam_metric = ifa->ifa_metric;
1095 				ifam->ifam_addrs = info.rti_addrs;
1096 				error = copyout(w->w_tmem, w->w_where, len);
1097 				if (error)
1098 					return (error);
1099 				w->w_where += len;
1100 			}
1101 		}
1102 		ifaaddr = netmask = brdaddr = 0;
1103 	}
1104 	return (0);
1105 }
1106 
1107 static int
1108 sysctl_rtable(SYSCTLFN_ARGS)
1109 {
1110 	void 	*where = oldp;
1111 	size_t	*given = oldlenp;
1112 	const void *new = newp;
1113 	struct radix_node_head *rnh;
1114 	int	i, s, error = EINVAL;
1115 	u_char  af;
1116 	struct	walkarg w;
1117 
1118 	if (namelen == 1 && name[0] == CTL_QUERY)
1119 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1120 
1121 	if (new)
1122 		return (EPERM);
1123 	if (namelen != 3)
1124 		return (EINVAL);
1125 	af = name[0];
1126 	w.w_tmemneeded = 0;
1127 	w.w_tmemsize = 0;
1128 	w.w_tmem = NULL;
1129 again:
1130 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1131 	if (w.w_tmemneeded) {
1132 		w.w_tmem = (caddr_t) malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
1133 		w.w_tmemsize = w.w_tmemneeded;
1134 		w.w_tmemneeded = 0;
1135 	}
1136 	w.w_op = name[1];
1137 	w.w_arg = name[2];
1138 	w.w_given = *given;
1139 	w.w_needed = 0 - w.w_given;
1140 	w.w_where = where;
1141 
1142 	s = splsoftnet();
1143 	switch (w.w_op) {
1144 
1145 	case NET_RT_DUMP:
1146 	case NET_RT_FLAGS:
1147 		for (i = 1; i <= AF_MAX; i++)
1148 			if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
1149 			    (error = (*rnh->rnh_walktree)(rnh,
1150 			    sysctl_dumpentry, &w)))
1151 				break;
1152 		break;
1153 
1154 #ifdef COMPAT_14
1155 	case NET_RT_OIFLIST:
1156 		error = sysctl_iflist(af, &w, w.w_op);
1157 		break;
1158 #endif
1159 
1160 	case NET_RT_IFLIST:
1161 		error = sysctl_iflist(af, &w, w.w_op);
1162 	}
1163 	splx(s);
1164 
1165 	/* check to see if we couldn't allocate memory with NOWAIT */
1166 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1167 		goto again;
1168 
1169 	if (w.w_tmem)
1170 		free(w.w_tmem, M_RTABLE);
1171 	w.w_needed += w.w_given;
1172 	if (where) {
1173 		*given = w.w_where - (caddr_t) where;
1174 		if (*given < w.w_needed)
1175 			return (ENOMEM);
1176 	} else {
1177 		*given = (11 * w.w_needed) / 10;
1178 	}
1179 	return (error);
1180 }
1181 
1182 /*
1183  * Definitions of protocols supported in the ROUTE domain.
1184  */
1185 
1186 const struct protosw routesw[] = {
1187 {
1188 	SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
1189 	raw_input,	route_output,	raw_ctlinput,	0,
1190 	route_usrreq,
1191 	raw_init,	0,		0,		0,
1192 } };
1193 
1194 struct domain routedomain = {
1195 	.dom_family = PF_ROUTE,
1196 	.dom_name = "route",
1197 	.dom_init = route_init,
1198 	.dom_protosw = routesw,
1199 	.dom_protoswNPROTOSW = &routesw[sizeof(routesw)/sizeof(routesw[0])],
1200 };
1201 
1202 SYSCTL_SETUP(sysctl_net_route_setup, "sysctl net.route subtree setup")
1203 {
1204 	const struct sysctlnode *rnode = NULL;
1205 
1206 	sysctl_createv(clog, 0, NULL, NULL,
1207 		       CTLFLAG_PERMANENT,
1208 		       CTLTYPE_NODE, "net", NULL,
1209 		       NULL, 0, NULL, 0,
1210 		       CTL_NET, CTL_EOL);
1211 
1212 	sysctl_createv(clog, 0, NULL, &rnode,
1213 		       CTLFLAG_PERMANENT,
1214 		       CTLTYPE_NODE, "route",
1215 		       SYSCTL_DESCR("PF_ROUTE information"),
1216 		       NULL, 0, NULL, 0,
1217 		       CTL_NET, PF_ROUTE, CTL_EOL);
1218 	sysctl_createv(clog, 0, NULL, NULL,
1219 		       CTLFLAG_PERMANENT,
1220 		       CTLTYPE_NODE, "rtable",
1221 		       SYSCTL_DESCR("Routing table information"),
1222 		       sysctl_rtable, 0, NULL, 0,
1223 		       CTL_NET, PF_ROUTE, 0 /* any protocol */, CTL_EOL);
1224 	sysctl_createv(clog, 0, &rnode, NULL,
1225 		       CTLFLAG_PERMANENT,
1226 		       CTLTYPE_STRUCT, "stats",
1227 		       SYSCTL_DESCR("Routing statistics"),
1228 		       NULL, 0, &rtstat, sizeof(rtstat),
1229 		       CTL_CREATE, CTL_EOL);
1230 }
1231