xref: /netbsd-src/sys/net/rtsock.c (revision 2980e352a13e8f0b545a366830c411e7a542ada8)
1 /*	$NetBSD: rtsock.c,v 1.109 2008/06/15 20:36:55 cube 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 1.109 2008/06/15 20:36:55 cube 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 #include <sys/intr.h>
79 #ifdef RTSOCK_DEBUG
80 #include <netinet/in.h>
81 #endif /* RTSOCK_DEBUG */
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 #include <net/raw_cb.h>
86 
87 #include <machine/stdarg.h>
88 
89 DOMAIN_DEFINE(routedomain);	/* forward declare and add to link set */
90 
91 struct	sockaddr route_dst = { .sa_len = 2, .sa_family = PF_ROUTE, };
92 struct	sockaddr route_src = { .sa_len = 2, .sa_family = PF_ROUTE, };
93 
94 int	route_maxqlen = IFQ_MAXLEN;
95 static struct	ifqueue route_intrq;
96 static void	*route_sih;
97 
98 struct walkarg {
99 	int	w_op;
100 	int	w_arg;
101 	int	w_given;
102 	int	w_needed;
103 	void *	w_where;
104 	int	w_tmemsize;
105 	int	w_tmemneeded;
106 	void *	w_tmem;
107 };
108 
109 static struct mbuf *rt_msg1(int, struct rt_addrinfo *, void *, int);
110 static int rt_msg2(int, struct rt_addrinfo *, void *, struct walkarg *, int *);
111 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
112 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
113     struct rt_addrinfo *);
114 static int sysctl_dumpentry(struct rtentry *, void *);
115 static int sysctl_iflist(int, struct walkarg *, int);
116 static int sysctl_rtable(SYSCTLFN_PROTO);
117 static inline void rt_adjustcount(int, int);
118 static void route_enqueue(struct mbuf *, int);
119 
120 /* Sleazy use of local variables throughout file, warning!!!! */
121 #define dst	info.rti_info[RTAX_DST]
122 #define gate	info.rti_info[RTAX_GATEWAY]
123 #define netmask	info.rti_info[RTAX_NETMASK]
124 #define ifpaddr	info.rti_info[RTAX_IFP]
125 #define ifaaddr	info.rti_info[RTAX_IFA]
126 #define brdaddr	info.rti_info[RTAX_BRD]
127 
128 static inline void
129 rt_adjustcount(int af, int cnt)
130 {
131 	route_cb.any_count += cnt;
132 	switch (af) {
133 	case AF_INET:
134 		route_cb.ip_count += cnt;
135 		return;
136 #ifdef INET6
137 	case AF_INET6:
138 		route_cb.ip6_count += cnt;
139 		return;
140 #endif
141 	case AF_IPX:
142 		route_cb.ipx_count += cnt;
143 		return;
144 	case AF_NS:
145 		route_cb.ns_count += cnt;
146 		return;
147 	case AF_ISO:
148 		route_cb.iso_count += cnt;
149 		return;
150 	}
151 }
152 
153 /*ARGSUSED*/
154 int
155 route_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
156 	struct mbuf *control, struct lwp *l)
157 {
158 	int error = 0;
159 	struct rawcb *rp = sotorawcb(so);
160 	int s;
161 
162 	if (req == PRU_ATTACH) {
163 		sosetlock(so);
164 		MALLOC(rp, struct rawcb *, sizeof(*rp), M_PCB, M_WAITOK|M_ZERO);
165 		so->so_pcb = rp;
166 	}
167 	if (req == PRU_DETACH && rp)
168 		rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
169 	s = splsoftnet();
170 
171 	/*
172 	 * Don't call raw_usrreq() in the attach case, because
173 	 * we want to allow non-privileged processes to listen on
174 	 * and send "safe" commands to the routing socket.
175 	 */
176 	if (req == PRU_ATTACH) {
177 		if (l == NULL)
178 			error = EACCES;
179 		else
180 			error = raw_attach(so, (int)(long)nam);
181 	} else
182 		error = raw_usrreq(so, req, m, nam, control, l);
183 
184 	rp = sotorawcb(so);
185 	if (req == PRU_ATTACH && rp) {
186 		if (error) {
187 			free((void *)rp, M_PCB);
188 			splx(s);
189 			return error;
190 		}
191 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
192 		rp->rcb_laddr = &route_src;
193 		rp->rcb_faddr = &route_dst;
194 		soisconnected(so);
195 		so->so_options |= SO_USELOOPBACK;
196 	}
197 	splx(s);
198 	return error;
199 }
200 
201 static const struct sockaddr *
202 intern_netmask(const struct sockaddr *mask)
203 {
204 	struct radix_node *rn;
205 	extern struct radix_node_head *mask_rnhead;
206 
207 	if (mask != NULL &&
208 	    (rn = rn_search(mask, mask_rnhead->rnh_treetop)))
209 		mask = (const struct sockaddr *)rn->rn_key;
210 
211 	return mask;
212 }
213 
214 /*ARGSUSED*/
215 int
216 route_output(struct mbuf *m, ...)
217 {
218 	struct sockproto proto = { .sp_family = PF_ROUTE, };
219 	struct rt_msghdr *rtm = NULL;
220 	struct rtentry *rt = NULL;
221 	struct rtentry *saved_nrt = NULL;
222 	struct rt_addrinfo info;
223 	int len, error = 0;
224 	struct ifnet *ifp = NULL;
225 	struct ifaddr *ifa = NULL;
226 	struct socket *so;
227 	va_list ap;
228 	sa_family_t family;
229 
230 	va_start(ap, m);
231 	so = va_arg(ap, struct socket *);
232 	va_end(ap);
233 
234 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
235 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
236 	   (m = m_pullup(m, sizeof(int32_t))) == NULL))
237 		return ENOBUFS;
238 	if ((m->m_flags & M_PKTHDR) == 0)
239 		panic("route_output");
240 	len = m->m_pkthdr.len;
241 	if (len < sizeof(*rtm) ||
242 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
243 		dst = NULL;
244 		senderr(EINVAL);
245 	}
246 	R_Malloc(rtm, struct rt_msghdr *, len);
247 	if (rtm == NULL) {
248 		dst = NULL;
249 		senderr(ENOBUFS);
250 	}
251 	m_copydata(m, 0, len, (void *)rtm);
252 	if (rtm->rtm_version != RTM_VERSION) {
253 		dst = NULL;
254 		senderr(EPROTONOSUPPORT);
255 	}
256 	rtm->rtm_pid = curproc->p_pid;
257 	memset(&info, 0, sizeof(info));
258 	info.rti_addrs = rtm->rtm_addrs;
259 	if (rt_xaddrs(rtm->rtm_type, (void *)(rtm + 1), len + (char *)rtm, &info))
260 		senderr(EINVAL);
261 	info.rti_flags = rtm->rtm_flags;
262 #ifdef RTSOCK_DEBUG
263 	if (dst->sa_family == AF_INET) {
264 		printf("%s: extracted dst %s\n", __func__,
265 		    inet_ntoa(((const struct sockaddr_in *)dst)->sin_addr));
266 	}
267 #endif /* RTSOCK_DEBUG */
268 	if (dst == NULL || (dst->sa_family >= AF_MAX))
269 		senderr(EINVAL);
270 	if (gate != NULL && (gate->sa_family >= AF_MAX))
271 		senderr(EINVAL);
272 
273 	/*
274 	 * Verify that the caller has the appropriate privilege; RTM_GET
275 	 * is the only operation the non-superuser is allowed.
276 	 */
277 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
278 	    0, rtm, NULL, NULL) != 0)
279 		senderr(EACCES);
280 
281 	switch (rtm->rtm_type) {
282 
283 	case RTM_ADD:
284 		if (gate == NULL)
285 			senderr(EINVAL);
286 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
287 		if (error == 0 && saved_nrt) {
288 			rt_setmetrics(rtm->rtm_inits,
289 			    &rtm->rtm_rmx, &saved_nrt->rt_rmx);
290 			saved_nrt->rt_refcnt--;
291 		}
292 		break;
293 
294 	case RTM_DELETE:
295 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
296 		if (error == 0) {
297 			(rt = saved_nrt)->rt_refcnt++;
298 			goto report;
299 		}
300 		break;
301 
302 	case RTM_GET:
303 	case RTM_CHANGE:
304 	case RTM_LOCK:
305                 /* XXX This will mask dst with netmask before
306                  * searching.  It did not used to do that.  --dyoung
307 		 */
308 		error = rtrequest1(RTM_GET, &info, &rt);
309 		if (error != 0)
310 			senderr(error);
311 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
312 			struct radix_node *rn;
313 
314 			if (memcmp(dst, rt_getkey(rt), dst->sa_len) != 0)
315 				senderr(ESRCH);
316 			netmask = intern_netmask(netmask);
317 			for (rn = rt->rt_nodes; rn; rn = rn->rn_dupedkey)
318 				if (netmask == (const struct sockaddr *)rn->rn_mask)
319 					break;
320 			if (rn == NULL)
321 				senderr(ETOOMANYREFS);
322 			rt = (struct rtentry *)rn;
323 		}
324 
325 		switch (rtm->rtm_type) {
326 		case RTM_GET:
327 		report:
328 			dst = rt_getkey(rt);
329 			gate = rt->rt_gateway;
330 			netmask = rt_mask(rt);
331 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
332 				;
333 			else if ((ifp = rt->rt_ifp) != NULL) {
334 				const struct ifaddr *rtifa;
335 				ifpaddr = ifp->if_dl->ifa_addr;
336                                 /* rtifa used to be simply rt->rt_ifa.
337                                  * If rt->rt_ifa != NULL, then
338                                  * rt_get_ifa() != NULL.  So this
339                                  * ought to still be safe. --dyoung
340 				 */
341 				rtifa = rt_get_ifa(rt);
342 				ifaaddr = rtifa->ifa_addr;
343 #ifdef RTSOCK_DEBUG
344 				if (ifaaddr->sa_family == AF_INET) {
345 					printf("%s: copying out RTAX_IFA %s ",
346 					    __func__,
347 					    inet_ntoa(((const struct sockaddr_in *)ifaaddr)->sin_addr));
348 					printf("for dst %s ifa_getifa %p ifa_seqno %p\n",
349 					    inet_ntoa(((const struct sockaddr_in *)dst)->sin_addr),
350 					    (void *)rtifa->ifa_getifa, rtifa->ifa_seqno);
351 				}
352 #endif /* RTSOCK_DEBUG */
353 				if (ifp->if_flags & IFF_POINTOPOINT)
354 					brdaddr = rtifa->ifa_dstaddr;
355 				else
356 					brdaddr = NULL;
357 				rtm->rtm_index = ifp->if_index;
358 			} else {
359 				ifpaddr = NULL;
360 				ifaaddr = NULL;
361 			}
362 			(void)rt_msg2(rtm->rtm_type, &info, NULL, NULL, &len);
363 			if (len > rtm->rtm_msglen) {
364 				struct rt_msghdr *new_rtm;
365 				R_Malloc(new_rtm, struct rt_msghdr *, len);
366 				if (new_rtm == NULL)
367 					senderr(ENOBUFS);
368 				memmove(new_rtm, rtm, rtm->rtm_msglen);
369 				Free(rtm); rtm = new_rtm;
370 			}
371 			(void)rt_msg2(rtm->rtm_type, &info, (void *)rtm,
372 			    NULL, 0);
373 			rtm->rtm_flags = rt->rt_flags;
374 			rtm->rtm_rmx = rt->rt_rmx;
375 			rtm->rtm_addrs = info.rti_addrs;
376 			break;
377 
378 		case RTM_CHANGE:
379 			/*
380 			 * new gateway could require new ifaddr, ifp;
381 			 * flags may also be different; ifp may be specified
382 			 * by ll sockaddr when protocol address is ambiguous
383 			 */
384 			if ((error = rt_getifa(&info)) != 0)
385 				senderr(error);
386 			if (gate && rt_setgate(rt, gate))
387 				senderr(EDQUOT);
388 			/* new gateway could require new ifaddr, ifp;
389 			   flags may also be different; ifp may be specified
390 			   by ll sockaddr when protocol address is ambiguous */
391 			if (ifpaddr && (ifa = ifa_ifwithnet(ifpaddr)) &&
392 			    (ifp = ifa->ifa_ifp) && (ifaaddr || gate))
393 				ifa = ifaof_ifpforaddr(ifaaddr ? ifaaddr : gate,
394 				    ifp);
395 			else if ((ifaaddr && (ifa = ifa_ifwithaddr(ifaaddr))) ||
396 			    (gate && (ifa = ifa_ifwithroute(rt->rt_flags,
397 			    rt_getkey(rt), gate))))
398 				ifp = ifa->ifa_ifp;
399 			if (ifa) {
400 				struct ifaddr *oifa = rt->rt_ifa;
401 				if (oifa != ifa) {
402 					if (oifa && oifa->ifa_rtrequest) {
403 						oifa->ifa_rtrequest(RTM_DELETE,
404 						    rt, &info);
405 					}
406 					rt_replace_ifa(rt, ifa);
407 					rt->rt_ifp = ifp;
408 				}
409 			}
410 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
411 			    &rt->rt_rmx);
412 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
413 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
414 			/*
415 			 * Fall into
416 			 */
417 		case RTM_LOCK:
418 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
419 			rt->rt_rmx.rmx_locks |=
420 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
421 			break;
422 		}
423 		break;
424 
425 	default:
426 		senderr(EOPNOTSUPP);
427 	}
428 
429 flush:
430 	if (rtm) {
431 		if (error)
432 			rtm->rtm_errno = error;
433 		else
434 			rtm->rtm_flags |= RTF_DONE;
435 	}
436 	family = dst ? dst->sa_family : 0;
437 	if (rt)
438 		rtfree(rt);
439     {
440 	struct rawcb *rp = NULL;
441 	/*
442 	 * Check to see if we don't want our own messages.
443 	 */
444 	if ((so->so_options & SO_USELOOPBACK) == 0) {
445 		if (route_cb.any_count <= 1) {
446 			if (rtm)
447 				Free(rtm);
448 			m_freem(m);
449 			return error;
450 		}
451 		/* There is another listener, so construct message */
452 		rp = sotorawcb(so);
453 	}
454 	if (rtm) {
455 		m_copyback(m, 0, rtm->rtm_msglen, (void *)rtm);
456 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
457 			m_freem(m);
458 			m = NULL;
459 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
460 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
461 		Free(rtm);
462 	}
463 	if (rp)
464 		rp->rcb_proto.sp_family = 0; /* Avoid us */
465 	if (family)
466 		proto.sp_protocol = family;
467 	if (m)
468 		raw_input(m, &proto, &route_src, &route_dst);
469 	if (rp)
470 		rp->rcb_proto.sp_family = PF_ROUTE;
471     }
472 	return error;
473 }
474 
475 void
476 rt_setmetrics(u_long which, const struct rt_metrics *in, struct rt_metrics *out)
477 {
478 #define metric(f, e) if (which & (f)) out->e = in->e;
479 	metric(RTV_RPIPE, rmx_recvpipe);
480 	metric(RTV_SPIPE, rmx_sendpipe);
481 	metric(RTV_SSTHRESH, rmx_ssthresh);
482 	metric(RTV_RTT, rmx_rtt);
483 	metric(RTV_RTTVAR, rmx_rttvar);
484 	metric(RTV_HOPCOUNT, rmx_hopcount);
485 	metric(RTV_MTU, rmx_mtu);
486 	metric(RTV_EXPIRE, rmx_expire);
487 #undef metric
488 }
489 
490 #define ROUNDUP(a) \
491 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
492 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
493 
494 static int
495 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim, struct rt_addrinfo *rtinfo)
496 {
497 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
498 	int i;
499 
500 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
501 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
502 			continue;
503 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
504 		ADVANCE(cp, sa);
505 	}
506 
507 	/* Check for extra addresses specified, except RTM_GET asking for interface info.  */
508 	if (rtmtype == RTM_GET) {
509 		if (((rtinfo->rti_addrs & (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
510 			return 1;
511 	} else {
512 		if ((rtinfo->rti_addrs & (~0 << i)) != 0)
513 			return 1;
514 	}
515 	/* Check for bad data length.  */
516 	if (cp != cplim) {
517 		if (i == RTAX_NETMASK + 1 && sa &&
518 		    cp - ROUNDUP(sa->sa_len) + sa->sa_len == cplim)
519 			/*
520 			 * The last sockaddr was netmask.
521 			 * We accept this for now for the sake of old
522 			 * binaries or third party softwares.
523 			 */
524 			;
525 		else
526 			return 1;
527 	}
528 	return 0;
529 }
530 
531 static struct mbuf *
532 rt_msg1(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
533 {
534 	struct rt_msghdr *rtm;
535 	struct mbuf *m;
536 	int i;
537 	const struct sockaddr *sa;
538 	int len, dlen;
539 
540 	m = m_gethdr(M_DONTWAIT, MT_DATA);
541 	if (m == NULL)
542 		return m;
543 	MCLAIM(m, &routedomain.dom_mowner);
544 	switch (type) {
545 
546 	case RTM_DELADDR:
547 	case RTM_NEWADDR:
548 		len = sizeof(struct ifa_msghdr);
549 		break;
550 
551 #ifdef COMPAT_14
552 	case RTM_OIFINFO:
553 		len = sizeof(struct if_msghdr14);
554 		break;
555 #endif
556 
557 	case RTM_IFINFO:
558 		len = sizeof(struct if_msghdr);
559 		break;
560 
561 	case RTM_IFANNOUNCE:
562 	case RTM_IEEE80211:
563 		len = sizeof(struct if_announcemsghdr);
564 		break;
565 
566 	default:
567 		len = sizeof(struct rt_msghdr);
568 	}
569 	if (len > MHLEN + MLEN)
570 		panic("rt_msg1: message too long");
571 	else if (len > MHLEN) {
572 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
573 		if (m->m_next == NULL) {
574 			m_freem(m);
575 			return NULL;
576 		}
577 		MCLAIM(m->m_next, m->m_owner);
578 		m->m_pkthdr.len = len;
579 		m->m_len = MHLEN;
580 		m->m_next->m_len = len - MHLEN;
581 	} else {
582 		m->m_pkthdr.len = m->m_len = len;
583 	}
584 	m->m_pkthdr.rcvif = NULL;
585 	m_copyback(m, 0, datalen, data);
586 	if (len > datalen)
587 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
588 	rtm = mtod(m, struct rt_msghdr *);
589 	for (i = 0; i < RTAX_MAX; i++) {
590 		if ((sa = rtinfo->rti_info[i]) == NULL)
591 			continue;
592 		rtinfo->rti_addrs |= (1 << i);
593 		dlen = ROUNDUP(sa->sa_len);
594 		m_copyback(m, len, dlen, sa);
595 		len += dlen;
596 	}
597 	if (m->m_pkthdr.len != len) {
598 		m_freem(m);
599 		return NULL;
600 	}
601 	rtm->rtm_msglen = len;
602 	rtm->rtm_version = RTM_VERSION;
603 	rtm->rtm_type = type;
604 	return m;
605 }
606 
607 /*
608  * rt_msg2
609  *
610  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
611  *		returns the length of the message in 'lenp'.
612  *
613  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
614  *	the message
615  * otherwise walkarg's w_needed is updated and if the user buffer is
616  *	specified and w_needed indicates space exists the information is copied
617  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
618  *	if the allocation fails ENOBUFS is returned.
619  */
620 static int
621 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct walkarg *w,
622 	int *lenp)
623 {
624 	int i;
625 	int len, dlen, second_time = 0;
626 	char *cp0, *cp = cpv;
627 
628 	rtinfo->rti_addrs = 0;
629 again:
630 	switch (type) {
631 
632 	case RTM_DELADDR:
633 	case RTM_NEWADDR:
634 		len = sizeof(struct ifa_msghdr);
635 		break;
636 #ifdef COMPAT_14
637 	case RTM_OIFINFO:
638 		len = sizeof(struct if_msghdr14);
639 		break;
640 #endif
641 
642 	case RTM_IFINFO:
643 		len = sizeof(struct if_msghdr);
644 		break;
645 
646 	default:
647 		len = sizeof(struct rt_msghdr);
648 	}
649 	if ((cp0 = cp) != NULL)
650 		cp += len;
651 	for (i = 0; i < RTAX_MAX; i++) {
652 		const struct sockaddr *sa;
653 
654 		if ((sa = rtinfo->rti_info[i]) == NULL)
655 			continue;
656 		rtinfo->rti_addrs |= (1 << i);
657 		dlen = ROUNDUP(sa->sa_len);
658 		if (cp) {
659 			bcopy(sa, cp, (unsigned)dlen);
660 			cp += dlen;
661 		}
662 		len += dlen;
663 	}
664 	if (cp == NULL && w != NULL && !second_time) {
665 		struct walkarg *rw = w;
666 
667 		rw->w_needed += len;
668 		if (rw->w_needed <= 0 && rw->w_where) {
669 			if (rw->w_tmemsize < len) {
670 				if (rw->w_tmem)
671 					free(rw->w_tmem, M_RTABLE);
672 				rw->w_tmem = (void *) malloc(len, M_RTABLE,
673 				    M_NOWAIT);
674 				if (rw->w_tmem)
675 					rw->w_tmemsize = len;
676 			}
677 			if (rw->w_tmem) {
678 				cp = rw->w_tmem;
679 				second_time = 1;
680 				goto again;
681 			} else {
682 				rw->w_tmemneeded = len;
683 				return ENOBUFS;
684 			}
685 		}
686 	}
687 	if (cp) {
688 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
689 
690 		rtm->rtm_version = RTM_VERSION;
691 		rtm->rtm_type = type;
692 		rtm->rtm_msglen = len;
693 	}
694 	if (lenp)
695 		*lenp = len;
696 	return 0;
697 }
698 
699 /*
700  * This routine is called to generate a message from the routing
701  * socket indicating that a redirect has occurred, a routing lookup
702  * has failed, or that a protocol has detected timeouts to a particular
703  * destination.
704  */
705 void
706 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
707 {
708 	struct rt_msghdr rtm;
709 	struct mbuf *m;
710 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
711 
712 	if (route_cb.any_count == 0)
713 		return;
714 	memset(&rtm, 0, sizeof(rtm));
715 	rtm.rtm_flags = RTF_DONE | flags;
716 	rtm.rtm_errno = error;
717 	m = rt_msg1(type, rtinfo, (void *)&rtm, sizeof(rtm));
718 	if (m == NULL)
719 		return;
720 	mtod(m, struct rt_msghdr *)->rtm_addrs = rtinfo->rti_addrs;
721 	route_enqueue(m, sa ? sa->sa_family : 0);
722 }
723 
724 /*
725  * This routine is called to generate a message from the routing
726  * socket indicating that the status of a network interface has changed.
727  */
728 void
729 rt_ifmsg(struct ifnet *ifp)
730 {
731 	struct if_msghdr ifm;
732 #ifdef COMPAT_14
733 	struct if_msghdr14 oifm;
734 #endif
735 	struct mbuf *m;
736 	struct rt_addrinfo info;
737 
738 	if (route_cb.any_count == 0)
739 		return;
740 	memset(&info, 0, sizeof(info));
741 	memset(&ifm, 0, sizeof(ifm));
742 	ifm.ifm_index = ifp->if_index;
743 	ifm.ifm_flags = ifp->if_flags;
744 	ifm.ifm_data = ifp->if_data;
745 	ifm.ifm_addrs = 0;
746 	m = rt_msg1(RTM_IFINFO, &info, (void *)&ifm, sizeof(ifm));
747 	if (m == NULL)
748 		return;
749 	route_enqueue(m, 0);
750 #ifdef COMPAT_14
751 	memset(&info, 0, sizeof(info));
752 	memset(&oifm, 0, sizeof(oifm));
753 	oifm.ifm_index = ifp->if_index;
754 	oifm.ifm_flags = ifp->if_flags;
755 	oifm.ifm_data.ifi_type = ifp->if_data.ifi_type;
756 	oifm.ifm_data.ifi_addrlen = ifp->if_data.ifi_addrlen;
757 	oifm.ifm_data.ifi_hdrlen = ifp->if_data.ifi_hdrlen;
758 	oifm.ifm_data.ifi_mtu = ifp->if_data.ifi_mtu;
759 	oifm.ifm_data.ifi_metric = ifp->if_data.ifi_metric;
760 	oifm.ifm_data.ifi_baudrate = ifp->if_data.ifi_baudrate;
761 	oifm.ifm_data.ifi_ipackets = ifp->if_data.ifi_ipackets;
762 	oifm.ifm_data.ifi_ierrors = ifp->if_data.ifi_ierrors;
763 	oifm.ifm_data.ifi_opackets = ifp->if_data.ifi_opackets;
764 	oifm.ifm_data.ifi_oerrors = ifp->if_data.ifi_oerrors;
765 	oifm.ifm_data.ifi_collisions = ifp->if_data.ifi_collisions;
766 	oifm.ifm_data.ifi_ibytes = ifp->if_data.ifi_ibytes;
767 	oifm.ifm_data.ifi_obytes = ifp->if_data.ifi_obytes;
768 	oifm.ifm_data.ifi_imcasts = ifp->if_data.ifi_imcasts;
769 	oifm.ifm_data.ifi_omcasts = ifp->if_data.ifi_omcasts;
770 	oifm.ifm_data.ifi_iqdrops = ifp->if_data.ifi_iqdrops;
771 	oifm.ifm_data.ifi_noproto = ifp->if_data.ifi_noproto;
772 	oifm.ifm_data.ifi_lastchange = ifp->if_data.ifi_lastchange;
773 	oifm.ifm_addrs = 0;
774 	m = rt_msg1(RTM_OIFINFO, &info, (void *)&oifm, sizeof(oifm));
775 	if (m == NULL)
776 		return;
777 	route_enqueue(m, 0);
778 #endif
779 }
780 
781 /*
782  * This is called to generate messages from the routing socket
783  * indicating a network interface has had addresses associated with it.
784  * if we ever reverse the logic and replace messages TO the routing
785  * socket indicate a request to configure interfaces, then it will
786  * be unnecessary as the routing socket will automatically generate
787  * copies of it.
788  */
789 void
790 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
791 {
792 	struct rt_addrinfo info;
793 	const struct sockaddr *sa = NULL;
794 	int pass;
795 	struct mbuf *m = NULL;
796 	struct ifnet *ifp = ifa->ifa_ifp;
797 
798 	if (route_cb.any_count == 0)
799 		return;
800 	for (pass = 1; pass < 3; pass++) {
801 		memset(&info, 0, sizeof(info));
802 		if ((cmd == RTM_ADD && pass == 1) ||
803 		    (cmd == RTM_DELETE && pass == 2)) {
804 			struct ifa_msghdr ifam;
805 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
806 
807 			ifaaddr = sa = ifa->ifa_addr;
808 			ifpaddr = ifp->if_dl->ifa_addr;
809 			netmask = ifa->ifa_netmask;
810 			brdaddr = ifa->ifa_dstaddr;
811 			memset(&ifam, 0, sizeof(ifam));
812 			ifam.ifam_index = ifp->if_index;
813 			ifam.ifam_metric = ifa->ifa_metric;
814 			ifam.ifam_flags = ifa->ifa_flags;
815 			m = rt_msg1(ncmd, &info, (void *)&ifam, sizeof(ifam));
816 			if (m == NULL)
817 				continue;
818 			mtod(m, struct ifa_msghdr *)->ifam_addrs =
819 			    info.rti_addrs;
820 		}
821 		if ((cmd == RTM_ADD && pass == 2) ||
822 		    (cmd == RTM_DELETE && pass == 1)) {
823 			struct rt_msghdr rtm;
824 
825 			if (rt == NULL)
826 				continue;
827 			netmask = rt_mask(rt);
828 			dst = sa = rt_getkey(rt);
829 			gate = rt->rt_gateway;
830 			memset(&rtm, 0, sizeof(rtm));
831 			rtm.rtm_index = ifp->if_index;
832 			rtm.rtm_flags |= rt->rt_flags;
833 			rtm.rtm_errno = error;
834 			m = rt_msg1(cmd, &info, (void *)&rtm, sizeof(rtm));
835 			if (m == NULL)
836 				continue;
837 			mtod(m, struct rt_msghdr *)->rtm_addrs = info.rti_addrs;
838 		}
839 #ifdef DIAGNOSTIC
840 		if (m == NULL)
841 			panic("%s: called with wrong command", __func__);
842 #endif
843 		route_enqueue(m, sa ? sa->sa_family : 0);
844 	}
845 }
846 
847 static struct mbuf *
848 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
849     struct rt_addrinfo *info)
850 {
851 	struct if_announcemsghdr ifan;
852 
853 	memset(info, 0, sizeof(*info));
854 	memset(&ifan, 0, sizeof(ifan));
855 	ifan.ifan_index = ifp->if_index;
856 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
857 	ifan.ifan_what = what;
858 	return rt_msg1(type, info, (void *)&ifan, sizeof(ifan));
859 }
860 
861 /*
862  * This is called to generate routing socket messages indicating
863  * network interface arrival and departure.
864  */
865 void
866 rt_ifannouncemsg(struct ifnet *ifp, int what)
867 {
868 	struct mbuf *m;
869 	struct rt_addrinfo info;
870 
871 	if (route_cb.any_count == 0)
872 		return;
873 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
874 	if (m == NULL)
875 		return;
876 	route_enqueue(m, 0);
877 }
878 
879 /*
880  * This is called to generate routing socket messages indicating
881  * IEEE80211 wireless events.
882  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
883  */
884 void
885 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
886 {
887 	struct mbuf *m;
888 	struct rt_addrinfo info;
889 
890 	if (route_cb.any_count == 0)
891 		return;
892 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
893 	if (m == NULL)
894 		return;
895 	/*
896 	 * Append the ieee80211 data.  Try to stick it in the
897 	 * mbuf containing the ifannounce msg; otherwise allocate
898 	 * a new mbuf and append.
899 	 *
900 	 * NB: we assume m is a single mbuf.
901 	 */
902 	if (data_len > M_TRAILINGSPACE(m)) {
903 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
904 		if (n == NULL) {
905 			m_freem(m);
906 			return;
907 		}
908 		(void)memcpy(mtod(n, void *), data, data_len);
909 		n->m_len = data_len;
910 		m->m_next = n;
911 	} else if (data_len > 0) {
912 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
913 		m->m_len += data_len;
914 	}
915 	if (m->m_flags & M_PKTHDR)
916 		m->m_pkthdr.len += data_len;
917 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
918 	route_enqueue(m, 0);
919 }
920 
921 /*
922  * This is used in dumping the kernel table via sysctl().
923  */
924 static int
925 sysctl_dumpentry(struct rtentry *rt, void *v)
926 {
927 	struct walkarg *w = v;
928 	int error = 0, size;
929 	struct rt_addrinfo info;
930 
931 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
932 		return 0;
933 	memset(&info, 0, sizeof(info));
934 	dst = rt_getkey(rt);
935 	gate = rt->rt_gateway;
936 	netmask = rt_mask(rt);
937 	if (rt->rt_ifp) {
938 		const struct ifaddr *rtifa;
939 		ifpaddr = rt->rt_ifp->if_dl->ifa_addr;
940 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
941 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
942 		 * --dyoung
943 		 */
944 		rtifa = rt_get_ifa(rt);
945 		ifaaddr = rtifa->ifa_addr;
946 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
947 			brdaddr = rtifa->ifa_dstaddr;
948 	}
949 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
950 		return error;
951 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
952 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
953 
954 		rtm->rtm_flags = rt->rt_flags;
955 		rtm->rtm_use = rt->rt_use;
956 		rtm->rtm_rmx = rt->rt_rmx;
957 		KASSERT(rt->rt_ifp != NULL);
958 		rtm->rtm_index = rt->rt_ifp->if_index;
959 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
960 		rtm->rtm_addrs = info.rti_addrs;
961 		if ((error = copyout(rtm, w->w_where, size)) != 0)
962 			w->w_where = NULL;
963 		else
964 			w->w_where = (char *)w->w_where + size;
965 	}
966 	return error;
967 }
968 
969 static int
970 sysctl_iflist(int af, struct walkarg *w, int type)
971 {
972 	struct ifnet *ifp;
973 	struct ifaddr *ifa;
974 	struct	rt_addrinfo info;
975 	int	len, error = 0;
976 
977 	memset(&info, 0, sizeof(info));
978 	IFNET_FOREACH(ifp) {
979 		if (w->w_arg && w->w_arg != ifp->if_index)
980 			continue;
981 		if (IFADDR_EMPTY(ifp))
982 			continue;
983 		ifpaddr = ifp->if_dl->ifa_addr;
984 		switch (type) {
985 		case NET_RT_IFLIST:
986 			error =
987 			    rt_msg2(RTM_IFINFO, &info, NULL, w, &len);
988 			break;
989 #ifdef COMPAT_14
990 		case NET_RT_OIFLIST:
991 			error =
992 			    rt_msg2(RTM_OIFINFO, &info, NULL, w, &len);
993 			break;
994 #endif
995 		default:
996 			panic("sysctl_iflist(1)");
997 		}
998 		if (error)
999 			return error;
1000 		ifpaddr = NULL;
1001 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1002 			switch (type) {
1003 			case NET_RT_IFLIST: {
1004 				struct if_msghdr *ifm;
1005 
1006 				ifm = (struct if_msghdr *)w->w_tmem;
1007 				ifm->ifm_index = ifp->if_index;
1008 				ifm->ifm_flags = ifp->if_flags;
1009 				ifm->ifm_data = ifp->if_data;
1010 				ifm->ifm_addrs = info.rti_addrs;
1011 				error = copyout(ifm, w->w_where, len);
1012 				if (error)
1013 					return error;
1014 				w->w_where = (char *)w->w_where + len;
1015 				break;
1016 			}
1017 
1018 #ifdef COMPAT_14
1019 			case NET_RT_OIFLIST: {
1020 				struct if_msghdr14 *ifm;
1021 
1022 				ifm = (struct if_msghdr14 *)w->w_tmem;
1023 				ifm->ifm_index = ifp->if_index;
1024 				ifm->ifm_flags = ifp->if_flags;
1025 				ifm->ifm_data.ifi_type = ifp->if_data.ifi_type;
1026 				ifm->ifm_data.ifi_addrlen =
1027 				    ifp->if_data.ifi_addrlen;
1028 				ifm->ifm_data.ifi_hdrlen =
1029 				    ifp->if_data.ifi_hdrlen;
1030 				ifm->ifm_data.ifi_mtu = ifp->if_data.ifi_mtu;
1031 				ifm->ifm_data.ifi_metric =
1032 				    ifp->if_data.ifi_metric;
1033 				ifm->ifm_data.ifi_baudrate =
1034 				    ifp->if_data.ifi_baudrate;
1035 				ifm->ifm_data.ifi_ipackets =
1036 				    ifp->if_data.ifi_ipackets;
1037 				ifm->ifm_data.ifi_ierrors =
1038 				    ifp->if_data.ifi_ierrors;
1039 				ifm->ifm_data.ifi_opackets =
1040 				    ifp->if_data.ifi_opackets;
1041 				ifm->ifm_data.ifi_oerrors =
1042 				    ifp->if_data.ifi_oerrors;
1043 				ifm->ifm_data.ifi_collisions =
1044 				    ifp->if_data.ifi_collisions;
1045 				ifm->ifm_data.ifi_ibytes =
1046 				    ifp->if_data.ifi_ibytes;
1047 				ifm->ifm_data.ifi_obytes =
1048 				    ifp->if_data.ifi_obytes;
1049 				ifm->ifm_data.ifi_imcasts =
1050 				    ifp->if_data.ifi_imcasts;
1051 				ifm->ifm_data.ifi_omcasts =
1052 				    ifp->if_data.ifi_omcasts;
1053 				ifm->ifm_data.ifi_iqdrops =
1054 				    ifp->if_data.ifi_iqdrops;
1055 				ifm->ifm_data.ifi_noproto =
1056 				    ifp->if_data.ifi_noproto;
1057 				ifm->ifm_data.ifi_lastchange =
1058 				    ifp->if_data.ifi_lastchange;
1059 				ifm->ifm_addrs = info.rti_addrs;
1060 				error = copyout(ifm, w->w_where, len);
1061 				if (error)
1062 					return error;
1063 				w->w_where = (char *)w->w_where + len;
1064 				break;
1065 			}
1066 #endif
1067 			default:
1068 				panic("sysctl_iflist(2)");
1069 			}
1070 		}
1071 		IFADDR_FOREACH(ifa, ifp) {
1072 			if (af && af != ifa->ifa_addr->sa_family)
1073 				continue;
1074 			ifaaddr = ifa->ifa_addr;
1075 			netmask = ifa->ifa_netmask;
1076 			brdaddr = ifa->ifa_dstaddr;
1077 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
1078 				return error;
1079 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1080 				struct ifa_msghdr *ifam;
1081 
1082 				ifam = (struct ifa_msghdr *)w->w_tmem;
1083 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1084 				ifam->ifam_flags = ifa->ifa_flags;
1085 				ifam->ifam_metric = ifa->ifa_metric;
1086 				ifam->ifam_addrs = info.rti_addrs;
1087 				error = copyout(w->w_tmem, w->w_where, len);
1088 				if (error)
1089 					return error;
1090 				w->w_where = (char *)w->w_where + len;
1091 			}
1092 		}
1093 		ifaaddr = netmask = brdaddr = NULL;
1094 	}
1095 	return 0;
1096 }
1097 
1098 static int
1099 sysctl_rtable(SYSCTLFN_ARGS)
1100 {
1101 	void 	*where = oldp;
1102 	size_t	*given = oldlenp;
1103 	const void *new = newp;
1104 	int	i, s, error = EINVAL;
1105 	u_char  af;
1106 	struct	walkarg w;
1107 
1108 	if (namelen == 1 && name[0] == CTL_QUERY)
1109 		return sysctl_query(SYSCTLFN_CALL(rnode));
1110 
1111 	if (new)
1112 		return EPERM;
1113 	if (namelen != 3)
1114 		return EINVAL;
1115 	af = name[0];
1116 	w.w_tmemneeded = 0;
1117 	w.w_tmemsize = 0;
1118 	w.w_tmem = NULL;
1119 again:
1120 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1121 	if (w.w_tmemneeded) {
1122 		w.w_tmem = (void *) malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
1123 		w.w_tmemsize = w.w_tmemneeded;
1124 		w.w_tmemneeded = 0;
1125 	}
1126 	w.w_op = name[1];
1127 	w.w_arg = name[2];
1128 	w.w_given = *given;
1129 	w.w_needed = 0 - w.w_given;
1130 	w.w_where = where;
1131 
1132 	s = splsoftnet();
1133 	switch (w.w_op) {
1134 
1135 	case NET_RT_DUMP:
1136 	case NET_RT_FLAGS:
1137 		for (i = 1; i <= AF_MAX; i++)
1138 			if ((af == 0 || af == i) &&
1139 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
1140 				break;
1141 		break;
1142 
1143 #ifdef COMPAT_14
1144 	case NET_RT_OIFLIST:
1145 		error = sysctl_iflist(af, &w, w.w_op);
1146 		break;
1147 #endif
1148 
1149 	case NET_RT_IFLIST:
1150 		error = sysctl_iflist(af, &w, w.w_op);
1151 	}
1152 	splx(s);
1153 
1154 	/* check to see if we couldn't allocate memory with NOWAIT */
1155 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1156 		goto again;
1157 
1158 	if (w.w_tmem)
1159 		free(w.w_tmem, M_RTABLE);
1160 	w.w_needed += w.w_given;
1161 	if (where) {
1162 		*given = (char *)w.w_where - (char *)where;
1163 		if (*given < w.w_needed)
1164 			return ENOMEM;
1165 	} else {
1166 		*given = (11 * w.w_needed) / 10;
1167 	}
1168 	return error;
1169 }
1170 
1171 /*
1172  * Routing message software interrupt routine
1173  */
1174 static void
1175 route_intr(void *cookie)
1176 {
1177 	struct sockproto proto = { .sp_family = PF_ROUTE, };
1178 	struct mbuf *m;
1179 	int s;
1180 
1181 	mutex_enter(softnet_lock);
1182 	KERNEL_LOCK(1, NULL);
1183 	while (!IF_IS_EMPTY(&route_intrq)) {
1184 		s = splnet();
1185 		IF_DEQUEUE(&route_intrq, m);
1186 		splx(s);
1187 		if (m == NULL)
1188 			break;
1189 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
1190 		raw_input(m, &proto, &route_src, &route_dst);
1191 	}
1192 	KERNEL_UNLOCK_ONE(NULL);
1193 	mutex_exit(softnet_lock);
1194 }
1195 
1196 /*
1197  * Enqueue a message to the software interrupt routine.
1198  */
1199 static void
1200 route_enqueue(struct mbuf *m, int family)
1201 {
1202 	int s, wasempty;
1203 
1204 	s = splnet();
1205 	if (IF_QFULL(&route_intrq)) {
1206 		IF_DROP(&route_intrq);
1207 		m_freem(m);
1208 	} else {
1209 		wasempty = IF_IS_EMPTY(&route_intrq);
1210 		M_SETCTX(m, (uintptr_t)family);
1211 		IF_ENQUEUE(&route_intrq, m);
1212 		if (wasempty)
1213 			softint_schedule(route_sih);
1214 	}
1215 	splx(s);
1216 }
1217 
1218 void
1219 rt_init(void)
1220 {
1221 
1222 	route_intrq.ifq_maxlen = route_maxqlen;
1223 	route_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
1224 	    route_intr, NULL);
1225 }
1226 
1227 /*
1228  * Definitions of protocols supported in the ROUTE domain.
1229  */
1230 PR_WRAP_USRREQ(route_usrreq)
1231 #define	route_usrreq	route_usrreq_wrapper
1232 
1233 const struct protosw routesw[] = {
1234 	{
1235 		.pr_type = SOCK_RAW,
1236 		.pr_domain = &routedomain,
1237 		.pr_flags = PR_ATOMIC|PR_ADDR,
1238 		.pr_input = raw_input,
1239 		.pr_output = route_output,
1240 		.pr_ctlinput = raw_ctlinput,
1241 		.pr_usrreq = route_usrreq,
1242 		.pr_init = raw_init,
1243 	},
1244 };
1245 
1246 struct domain routedomain = {
1247 	.dom_family = PF_ROUTE,
1248 	.dom_name = "route",
1249 	.dom_init = route_init,
1250 	.dom_protosw = routesw,
1251 	.dom_protoswNPROTOSW = &routesw[__arraycount(routesw)],
1252 };
1253 
1254 SYSCTL_SETUP(sysctl_net_route_setup, "sysctl net.route subtree setup")
1255 {
1256 	const struct sysctlnode *rnode = NULL;
1257 
1258 	sysctl_createv(clog, 0, NULL, NULL,
1259 		       CTLFLAG_PERMANENT,
1260 		       CTLTYPE_NODE, "net", NULL,
1261 		       NULL, 0, NULL, 0,
1262 		       CTL_NET, CTL_EOL);
1263 
1264 	sysctl_createv(clog, 0, NULL, &rnode,
1265 		       CTLFLAG_PERMANENT,
1266 		       CTLTYPE_NODE, "route",
1267 		       SYSCTL_DESCR("PF_ROUTE information"),
1268 		       NULL, 0, NULL, 0,
1269 		       CTL_NET, PF_ROUTE, CTL_EOL);
1270 	sysctl_createv(clog, 0, NULL, NULL,
1271 		       CTLFLAG_PERMANENT,
1272 		       CTLTYPE_NODE, "rtable",
1273 		       SYSCTL_DESCR("Routing table information"),
1274 		       sysctl_rtable, 0, NULL, 0,
1275 		       CTL_NET, PF_ROUTE, 0 /* any protocol */, CTL_EOL);
1276 	sysctl_createv(clog, 0, &rnode, NULL,
1277 		       CTLFLAG_PERMANENT,
1278 		       CTLTYPE_STRUCT, "stats",
1279 		       SYSCTL_DESCR("Routing statistics"),
1280 		       NULL, 0, &rtstat, sizeof(rtstat),
1281 		       CTL_CREATE, CTL_EOL);
1282 }
1283