xref: /netbsd-src/sys/net/rtsock.c (revision 3816d47b2c42fcd6e549e3407f842a5b1a1d23ad)
1 /*	$NetBSD: rtsock.c,v 1.127 2009/09/16 15:23:04 pooka 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 1.127 2009/09/16 15:23:04 pooka Exp $");
65 
66 #include "opt_inet.h"
67 #ifdef _KERNEL_OPT
68 #include "opt_compat_netbsd.h"
69 #endif
70 
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/proc.h>
74 #include <sys/mbuf.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/sysctl.h>
80 #include <sys/kauth.h>
81 #include <sys/intr.h>
82 #ifdef RTSOCK_DEBUG
83 #include <netinet/in.h>
84 #endif /* RTSOCK_DEBUG */
85 
86 #include <net/if.h>
87 #include <net/route.h>
88 #include <net/raw_cb.h>
89 
90 #if defined(COMPAT_14) || defined(COMPAT_50)
91 #include <compat/net/if.h>
92 #endif
93 
94 #include <machine/stdarg.h>
95 
96 DOMAIN_DEFINE(routedomain);	/* forward declare and add to link set */
97 
98 struct	sockaddr route_dst = { .sa_len = 2, .sa_family = PF_ROUTE, };
99 struct	sockaddr route_src = { .sa_len = 2, .sa_family = PF_ROUTE, };
100 
101 int	route_maxqlen = IFQ_MAXLEN;
102 static struct	ifqueue route_intrq;
103 static void	*route_sih;
104 
105 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
106 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
107 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
108     struct rt_addrinfo *);
109 static void sysctl_net_route_setup(struct sysctllog **);
110 static int sysctl_dumpentry(struct rtentry *, void *);
111 static int sysctl_iflist(int, struct rt_walkarg *, int);
112 static int sysctl_rtable(SYSCTLFN_PROTO);
113 static void rt_adjustcount(int, int);
114 
115 static void
116 rt_adjustcount(int af, int cnt)
117 {
118 	route_cb.any_count += cnt;
119 	switch (af) {
120 	case AF_INET:
121 		route_cb.ip_count += cnt;
122 		return;
123 #ifdef INET6
124 	case AF_INET6:
125 		route_cb.ip6_count += cnt;
126 		return;
127 #endif
128 	case AF_IPX:
129 		route_cb.ipx_count += cnt;
130 		return;
131 	case AF_NS:
132 		route_cb.ns_count += cnt;
133 		return;
134 	case AF_ISO:
135 		route_cb.iso_count += cnt;
136 		return;
137 	}
138 }
139 
140 static void
141 cvtmetrics(struct rt_metrics *ortm, const struct nrt_metrics *rtm)
142 {
143 	ortm->rmx_locks = rtm->rmx_locks;
144 	ortm->rmx_mtu = rtm->rmx_mtu;
145 	ortm->rmx_hopcount = rtm->rmx_hopcount;
146 	ortm->rmx_expire = rtm->rmx_expire;
147 	ortm->rmx_recvpipe = rtm->rmx_recvpipe;
148 	ortm->rmx_sendpipe = rtm->rmx_sendpipe;
149 	ortm->rmx_ssthresh = rtm->rmx_ssthresh;
150 	ortm->rmx_rtt = rtm->rmx_rtt;
151 	ortm->rmx_rttvar = rtm->rmx_rttvar;
152 	ortm->rmx_pksent = rtm->rmx_pksent;
153 }
154 
155 /*ARGSUSED*/
156 int
157 route_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
158 	struct mbuf *control, struct lwp *l)
159 {
160 	int error = 0;
161 	struct rawcb *rp = sotorawcb(so);
162 	int s;
163 
164 	if (req == PRU_ATTACH) {
165 		sosetlock(so);
166 		rp = malloc(sizeof(*rp), M_PCB, M_WAITOK|M_ZERO);
167 		so->so_pcb = rp;
168 	}
169 	if (req == PRU_DETACH && rp)
170 		rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
171 	s = splsoftnet();
172 
173 	/*
174 	 * Don't call raw_usrreq() in the attach case, because
175 	 * we want to allow non-privileged processes to listen on
176 	 * and send "safe" commands to the routing socket.
177 	 */
178 	if (req == PRU_ATTACH) {
179 		if (l == NULL)
180 			error = EACCES;
181 		else
182 			error = raw_attach(so, (int)(long)nam);
183 	} else
184 		error = raw_usrreq(so, req, m, nam, control, l);
185 
186 	rp = sotorawcb(so);
187 	if (req == PRU_ATTACH && rp) {
188 		if (error) {
189 			free(rp, M_PCB);
190 			splx(s);
191 			return error;
192 		}
193 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
194 		rp->rcb_laddr = &route_src;
195 		rp->rcb_faddr = &route_dst;
196 		soisconnected(so);
197 		so->so_options |= SO_USELOOPBACK;
198 	}
199 	splx(s);
200 	return error;
201 }
202 
203 static const struct sockaddr *
204 intern_netmask(const struct sockaddr *mask)
205 {
206 	struct radix_node *rn;
207 	extern struct radix_node_head *mask_rnhead;
208 
209 	if (mask != NULL &&
210 	    (rn = rn_search(mask, mask_rnhead->rnh_treetop)))
211 		mask = (const struct sockaddr *)rn->rn_key;
212 
213 	return mask;
214 }
215 
216 /*ARGSUSED*/
217 int
218 route_output(struct mbuf *m, ...)
219 {
220 	struct sockproto proto = { .sp_family = PF_ROUTE, };
221 	struct rt_msghdr *rtm = NULL;
222 	struct rt_msghdr *old_rtm = NULL;
223 	struct rtentry *rt = NULL;
224 	struct rtentry *saved_nrt = NULL;
225 	struct rt_addrinfo info;
226 	int len, error = 0;
227 	struct ifnet *ifp = NULL;
228 	struct ifaddr *ifa = NULL;
229 	struct socket *so;
230 	va_list ap;
231 	sa_family_t family;
232 
233 	va_start(ap, m);
234 	so = va_arg(ap, struct socket *);
235 	va_end(ap);
236 
237 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
238 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
239 	   (m = m_pullup(m, sizeof(int32_t))) == NULL))
240 		return ENOBUFS;
241 	if ((m->m_flags & M_PKTHDR) == 0)
242 		panic("route_output");
243 	len = m->m_pkthdr.len;
244 	if (len < sizeof(*rtm) ||
245 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
246 		info.rti_info[RTAX_DST] = NULL;
247 		senderr(EINVAL);
248 	}
249 	R_Malloc(rtm, struct rt_msghdr *, len);
250 	if (rtm == NULL) {
251 		info.rti_info[RTAX_DST] = NULL;
252 		senderr(ENOBUFS);
253 	}
254 	m_copydata(m, 0, len, rtm);
255 	if (rtm->rtm_version != RTM_VERSION) {
256 		info.rti_info[RTAX_DST] = NULL;
257 		senderr(EPROTONOSUPPORT);
258 	}
259 	rtm->rtm_pid = curproc->p_pid;
260 	memset(&info, 0, sizeof(info));
261 	info.rti_addrs = rtm->rtm_addrs;
262 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
263 	    &info))
264 		senderr(EINVAL);
265 	info.rti_flags = rtm->rtm_flags;
266 #ifdef RTSOCK_DEBUG
267 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
268 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
269 		    inet_ntoa(((const struct sockaddr_in *)
270 		    info.rti_info[RTAX_DST])->sin_addr));
271 	}
272 #endif /* RTSOCK_DEBUG */
273 	if (info.rti_info[RTAX_DST] == NULL ||
274 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX))
275 		senderr(EINVAL);
276 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
277 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
278 		senderr(EINVAL);
279 
280 	/*
281 	 * Verify that the caller has the appropriate privilege; RTM_GET
282 	 * is the only operation the non-superuser is allowed.
283 	 */
284 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
285 	    0, rtm, NULL, NULL) != 0)
286 		senderr(EACCES);
287 
288 	switch (rtm->rtm_type) {
289 
290 	case RTM_ADD:
291 		if (info.rti_info[RTAX_GATEWAY] == NULL)
292 			senderr(EINVAL);
293 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
294 		if (error == 0 && saved_nrt) {
295 			rt_setmetrics(rtm->rtm_inits,
296 			    &rtm->rtm_rmx, &saved_nrt->rt_rmx);
297 			saved_nrt->rt_refcnt--;
298 		}
299 		break;
300 
301 	case RTM_DELETE:
302 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
303 		if (error == 0) {
304 			(rt = saved_nrt)->rt_refcnt++;
305 			goto report;
306 		}
307 		break;
308 
309 	case RTM_GET:
310 	case RTM_CHANGE:
311 	case RTM_LOCK:
312                 /* XXX This will mask info.rti_info[RTAX_DST] with
313 		 * info.rti_info[RTAX_NETMASK] before
314                  * searching.  It did not used to do that.  --dyoung
315 		 */
316 		error = rtrequest1(RTM_GET, &info, &rt);
317 		if (error != 0)
318 			senderr(error);
319 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
320 			struct radix_node *rn;
321 
322 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
323 			    info.rti_info[RTAX_DST]->sa_len) != 0)
324 				senderr(ESRCH);
325 			info.rti_info[RTAX_NETMASK] = intern_netmask(
326 			    info.rti_info[RTAX_NETMASK]);
327 			for (rn = rt->rt_nodes; rn; rn = rn->rn_dupedkey)
328 				if (info.rti_info[RTAX_NETMASK] ==
329 				    (const struct sockaddr *)rn->rn_mask)
330 					break;
331 			if (rn == NULL)
332 				senderr(ETOOMANYREFS);
333 			rt = (struct rtentry *)rn;
334 		}
335 
336 		switch (rtm->rtm_type) {
337 		case RTM_GET:
338 		report:
339 			info.rti_info[RTAX_DST] = rt_getkey(rt);
340 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
341 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
342 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
343 				;
344 			else if ((ifp = rt->rt_ifp) != NULL) {
345 				const struct ifaddr *rtifa;
346 				info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
347                                 /* rtifa used to be simply rt->rt_ifa.
348                                  * If rt->rt_ifa != NULL, then
349                                  * rt_get_ifa() != NULL.  So this
350                                  * ought to still be safe. --dyoung
351 				 */
352 				rtifa = rt_get_ifa(rt);
353 				info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
354 #ifdef RTSOCK_DEBUG
355 				if (info.rti_info[RTAX_IFA]->sa_family ==
356 				    AF_INET) {
357 					printf("%s: copying out RTAX_IFA %s ",
358 					    __func__, inet_ntoa(
359 					    ((const struct sockaddr_in *)
360 					    info.rti_info[RTAX_IFA])->sin_addr)
361 					    );
362 					printf("for info.rti_info[RTAX_DST] %s "
363 					    "ifa_getifa %p ifa_seqno %p\n",
364 					    inet_ntoa(
365 					    ((const struct sockaddr_in *)
366 					    info.rti_info[RTAX_DST])->sin_addr),
367 					    (void *)rtifa->ifa_getifa,
368 					    rtifa->ifa_seqno);
369 				}
370 #endif /* RTSOCK_DEBUG */
371 				if (ifp->if_flags & IFF_POINTOPOINT) {
372 					info.rti_info[RTAX_BRD] =
373 					    rtifa->ifa_dstaddr;
374 				} else
375 					info.rti_info[RTAX_BRD] = NULL;
376 				rtm->rtm_index = ifp->if_index;
377 			} else {
378 				info.rti_info[RTAX_IFP] = NULL;
379 				info.rti_info[RTAX_IFA] = NULL;
380 			}
381 			(void)rt_msg2(rtm->rtm_type, &info, NULL, NULL, &len);
382 			if (len > rtm->rtm_msglen) {
383 				old_rtm = rtm;
384 				R_Malloc(rtm, struct rt_msghdr *, len);
385 				if (rtm == NULL)
386 					senderr(ENOBUFS);
387 				(void)memcpy(rtm, old_rtm, old_rtm->rtm_msglen);
388 			}
389 			(void)rt_msg2(rtm->rtm_type, &info, rtm, NULL, 0);
390 			rtm->rtm_flags = rt->rt_flags;
391 			cvtmetrics(&rtm->rtm_rmx, &rt->rt_rmx);
392 			rtm->rtm_addrs = info.rti_addrs;
393 			break;
394 
395 		case RTM_CHANGE:
396 			/*
397 			 * new gateway could require new ifaddr, ifp;
398 			 * flags may also be different; ifp may be specified
399 			 * by ll sockaddr when protocol address is ambiguous
400 			 */
401 			if ((error = rt_getifa(&info)) != 0)
402 				senderr(error);
403 			if (info.rti_info[RTAX_GATEWAY] &&
404 			    rt_setgate(rt, info.rti_info[RTAX_GATEWAY]))
405 				senderr(EDQUOT);
406 			/* new gateway could require new ifaddr, ifp;
407 			   flags may also be different; ifp may be specified
408 			   by ll sockaddr when protocol address is ambiguous */
409 			if (info.rti_info[RTAX_IFP] &&
410 			    (ifa = ifa_ifwithnet(info.rti_info[RTAX_IFP])) &&
411 			    (ifp = ifa->ifa_ifp) && (info.rti_info[RTAX_IFA] ||
412 			    info.rti_info[RTAX_GATEWAY])) {
413 				ifa = ifaof_ifpforaddr(info.rti_info[RTAX_IFA] ?
414 				    info.rti_info[RTAX_IFA] :
415 				    info.rti_info[RTAX_GATEWAY], ifp);
416 			} else if ((info.rti_info[RTAX_IFA] &&
417 			    (ifa = ifa_ifwithaddr(info.rti_info[RTAX_IFA]))) ||
418 			    (info.rti_info[RTAX_GATEWAY] &&
419 			    (ifa = ifa_ifwithroute(rt->rt_flags,
420 			    rt_getkey(rt), info.rti_info[RTAX_GATEWAY])))) {
421 				ifp = ifa->ifa_ifp;
422 			}
423 			if (ifa) {
424 				struct ifaddr *oifa = rt->rt_ifa;
425 				if (oifa != ifa) {
426 					if (oifa && oifa->ifa_rtrequest) {
427 						oifa->ifa_rtrequest(RTM_DELETE,
428 						    rt, &info);
429 					}
430 					rt_replace_ifa(rt, ifa);
431 					rt->rt_ifp = ifp;
432 				}
433 			}
434 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
435 			    &rt->rt_rmx);
436 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
437 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
438 			/*FALLTHROUGH*/
439 		case RTM_LOCK:
440 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
441 			rt->rt_rmx.rmx_locks |=
442 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
443 			break;
444 		}
445 		break;
446 
447 	default:
448 		senderr(EOPNOTSUPP);
449 	}
450 
451 flush:
452 	if (rtm) {
453 		if (error)
454 			rtm->rtm_errno = error;
455 		else
456 			rtm->rtm_flags |= RTF_DONE;
457 	}
458 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
459 	    0;
460 	/* We cannot free old_rtm until we have stopped using the
461 	 * pointers in info, some of which may point to sockaddrs
462 	 * in old_rtm.
463 	 */
464 	if (old_rtm != NULL)
465 		Free(old_rtm);
466 	if (rt)
467 		rtfree(rt);
468     {
469 	struct rawcb *rp = NULL;
470 	/*
471 	 * Check to see if we don't want our own messages.
472 	 */
473 	if ((so->so_options & SO_USELOOPBACK) == 0) {
474 		if (route_cb.any_count <= 1) {
475 			if (rtm)
476 				Free(rtm);
477 			m_freem(m);
478 			return error;
479 		}
480 		/* There is another listener, so construct message */
481 		rp = sotorawcb(so);
482 	}
483 	if (rtm) {
484 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
485 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
486 			m_freem(m);
487 			m = NULL;
488 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
489 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
490 		Free(rtm);
491 	}
492 	if (rp)
493 		rp->rcb_proto.sp_family = 0; /* Avoid us */
494 	if (family)
495 		proto.sp_protocol = family;
496 	if (m)
497 		raw_input(m, &proto, &route_src, &route_dst);
498 	if (rp)
499 		rp->rcb_proto.sp_family = PF_ROUTE;
500     }
501 	return error;
502 }
503 
504 void
505 rt_setmetrics(u_long which, const struct rt_metrics *in, struct nrt_metrics *out)
506 {
507 #define metric(f, e) if (which & (f)) out->e = in->e;
508 	metric(RTV_RPIPE, rmx_recvpipe);
509 	metric(RTV_SPIPE, rmx_sendpipe);
510 	metric(RTV_SSTHRESH, rmx_ssthresh);
511 	metric(RTV_RTT, rmx_rtt);
512 	metric(RTV_RTTVAR, rmx_rttvar);
513 	metric(RTV_HOPCOUNT, rmx_hopcount);
514 	metric(RTV_MTU, rmx_mtu);
515 	/* XXX time_t: Will not work after February 2145 (u_long time) */
516 	metric(RTV_EXPIRE, rmx_expire);
517 #undef metric
518 }
519 
520 static int
521 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
522     struct rt_addrinfo *rtinfo)
523 {
524 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
525 	int i;
526 
527 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
528 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
529 			continue;
530 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
531 		RT_ADVANCE(cp, sa);
532 	}
533 
534 	/*
535 	 * Check for extra addresses specified, except RTM_GET asking
536 	 * for interface info.
537 	 */
538 	if (rtmtype == RTM_GET) {
539 		if (((rtinfo->rti_addrs &
540 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
541 			return 1;
542 	} else if ((rtinfo->rti_addrs & (~0 << i)) != 0)
543 		return 1;
544 	/* Check for bad data length.  */
545 	if (cp != cplim) {
546 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
547 		    cp - RT_ROUNDUP(sa->sa_len) + sa->sa_len == cplim)
548 			/*
549 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
550 			 * We accept this for now for the sake of old
551 			 * binaries or third party softwares.
552 			 */
553 			;
554 		else
555 			return 1;
556 	}
557 	return 0;
558 }
559 
560 struct mbuf *
561 rt_msg1(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
562 {
563 	struct rt_msghdr *rtm;
564 	struct mbuf *m;
565 	int i;
566 	const struct sockaddr *sa;
567 	int len, dlen;
568 
569 	m = m_gethdr(M_DONTWAIT, MT_DATA);
570 	if (m == NULL)
571 		return m;
572 	MCLAIM(m, &routedomain.dom_mowner);
573 	switch (type) {
574 
575 	case RTM_DELADDR:
576 	case RTM_NEWADDR:
577 		len = sizeof(struct ifa_msghdr);
578 		break;
579 
580 #ifdef COMPAT_14
581 	case RTM_OOIFINFO:
582 		len = sizeof(struct if_msghdr14);
583 		break;
584 #endif
585 #ifdef COMPAT_50
586 	case RTM_OIFINFO:
587 		len = sizeof(struct if_msghdr50);
588 		break;
589 #endif
590 
591 	case RTM_IFINFO:
592 		len = sizeof(struct if_msghdr);
593 		break;
594 
595 	case RTM_IFANNOUNCE:
596 	case RTM_IEEE80211:
597 		len = sizeof(struct if_announcemsghdr);
598 		break;
599 
600 	default:
601 		len = sizeof(struct rt_msghdr);
602 	}
603 	if (len > MHLEN + MLEN)
604 		panic("rt_msg1: message too long");
605 	else if (len > MHLEN) {
606 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
607 		if (m->m_next == NULL) {
608 			m_freem(m);
609 			return NULL;
610 		}
611 		MCLAIM(m->m_next, m->m_owner);
612 		m->m_pkthdr.len = len;
613 		m->m_len = MHLEN;
614 		m->m_next->m_len = len - MHLEN;
615 	} else {
616 		m->m_pkthdr.len = m->m_len = len;
617 	}
618 	m->m_pkthdr.rcvif = NULL;
619 	m_copyback(m, 0, datalen, data);
620 	if (len > datalen)
621 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
622 	rtm = mtod(m, struct rt_msghdr *);
623 	for (i = 0; i < RTAX_MAX; i++) {
624 		if ((sa = rtinfo->rti_info[i]) == NULL)
625 			continue;
626 		rtinfo->rti_addrs |= (1 << i);
627 		dlen = RT_ROUNDUP(sa->sa_len);
628 		m_copyback(m, len, dlen, sa);
629 		len += dlen;
630 	}
631 	if (m->m_pkthdr.len != len) {
632 		m_freem(m);
633 		return NULL;
634 	}
635 	rtm->rtm_msglen = len;
636 	rtm->rtm_version = RTM_VERSION;
637 	rtm->rtm_type = type;
638 	return m;
639 }
640 
641 /*
642  * rt_msg2
643  *
644  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
645  *		returns the length of the message in 'lenp'.
646  *
647  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
648  *	the message
649  * otherwise walkarg's w_needed is updated and if the user buffer is
650  *	specified and w_needed indicates space exists the information is copied
651  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
652  *	if the allocation fails ENOBUFS is returned.
653  */
654 static int
655 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
656 	int *lenp)
657 {
658 	int i;
659 	int len, dlen, second_time = 0;
660 	char *cp0, *cp = cpv;
661 
662 	rtinfo->rti_addrs = 0;
663 again:
664 	switch (type) {
665 
666 	case RTM_DELADDR:
667 	case RTM_NEWADDR:
668 		len = sizeof(struct ifa_msghdr);
669 		break;
670 #ifdef COMPAT_14
671 	case RTM_OOIFINFO:
672 		len = sizeof(struct if_msghdr14);
673 		break;
674 #endif
675 #ifdef COMPAT_50
676 	case RTM_OIFINFO:
677 		len = sizeof(struct if_msghdr50);
678 		break;
679 #endif
680 
681 	case RTM_IFINFO:
682 		len = sizeof(struct if_msghdr);
683 		break;
684 
685 	default:
686 		len = sizeof(struct rt_msghdr);
687 	}
688 	if ((cp0 = cp) != NULL)
689 		cp += len;
690 	for (i = 0; i < RTAX_MAX; i++) {
691 		const struct sockaddr *sa;
692 
693 		if ((sa = rtinfo->rti_info[i]) == NULL)
694 			continue;
695 		rtinfo->rti_addrs |= (1 << i);
696 		dlen = RT_ROUNDUP(sa->sa_len);
697 		if (cp) {
698 			(void)memcpy(cp, sa, (size_t)dlen);
699 			cp += dlen;
700 		}
701 		len += dlen;
702 	}
703 	if (cp == NULL && w != NULL && !second_time) {
704 		struct rt_walkarg *rw = w;
705 
706 		rw->w_needed += len;
707 		if (rw->w_needed <= 0 && rw->w_where) {
708 			if (rw->w_tmemsize < len) {
709 				if (rw->w_tmem)
710 					free(rw->w_tmem, M_RTABLE);
711 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
712 				if (rw->w_tmem)
713 					rw->w_tmemsize = len;
714 				else
715 					rw->w_tmemsize = 0;
716 			}
717 			if (rw->w_tmem) {
718 				cp = rw->w_tmem;
719 				second_time = 1;
720 				goto again;
721 			} else {
722 				rw->w_tmemneeded = len;
723 				return ENOBUFS;
724 			}
725 		}
726 	}
727 	if (cp) {
728 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
729 
730 		rtm->rtm_version = RTM_VERSION;
731 		rtm->rtm_type = type;
732 		rtm->rtm_msglen = len;
733 	}
734 	if (lenp)
735 		*lenp = len;
736 	return 0;
737 }
738 
739 /*
740  * This routine is called to generate a message from the routing
741  * socket indicating that a redirect has occurred, a routing lookup
742  * has failed, or that a protocol has detected timeouts to a particular
743  * destination.
744  */
745 void
746 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
747 {
748 	struct rt_msghdr rtm;
749 	struct mbuf *m;
750 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
751 
752 	if (route_cb.any_count == 0)
753 		return;
754 	memset(&rtm, 0, sizeof(rtm));
755 	rtm.rtm_flags = RTF_DONE | flags;
756 	rtm.rtm_errno = error;
757 	m = rt_msg1(type, rtinfo, &rtm, sizeof(rtm));
758 	if (m == NULL)
759 		return;
760 	mtod(m, struct rt_msghdr *)->rtm_addrs = rtinfo->rti_addrs;
761 	route_enqueue(m, sa ? sa->sa_family : 0);
762 }
763 
764 /*
765  * This routine is called to generate a message from the routing
766  * socket indicating that the status of a network interface has changed.
767  */
768 void
769 rt_ifmsg(struct ifnet *ifp)
770 {
771 	struct if_msghdr ifm;
772 	struct mbuf *m;
773 	struct rt_addrinfo info;
774 
775 	if (route_cb.any_count == 0)
776 		return;
777 	(void)memset(&info, 0, sizeof(info));
778 	(void)memset(&ifm, 0, sizeof(ifm));
779 	ifm.ifm_index = ifp->if_index;
780 	ifm.ifm_flags = ifp->if_flags;
781 	ifm.ifm_data = ifp->if_data;
782 	ifm.ifm_addrs = 0;
783 	m = rt_msg1(RTM_IFINFO, &info, &ifm, sizeof(ifm));
784 	if (m == NULL)
785 		return;
786 	route_enqueue(m, 0);
787 #ifdef COMPAT_14
788 	compat_14_rt_ifmsg(ifp, &ifm);
789 #endif
790 #ifdef COMPAT_50
791 	compat_50_rt_ifmsg(ifp, &ifm);
792 #endif
793 }
794 
795 
796 /*
797  * This is called to generate messages from the routing socket
798  * indicating a network interface has had addresses associated with it.
799  * if we ever reverse the logic and replace messages TO the routing
800  * socket indicate a request to configure interfaces, then it will
801  * be unnecessary as the routing socket will automatically generate
802  * copies of it.
803  */
804 void
805 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
806 {
807 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
808 	struct rt_addrinfo info;
809 	const struct sockaddr *sa;
810 	int pass;
811 	struct mbuf *m;
812 	struct ifnet *ifp = ifa->ifa_ifp;
813 	struct rt_msghdr rtm;
814 	struct ifa_msghdr ifam;
815 	int ncmd;
816 
817 	if (route_cb.any_count == 0)
818 		return;
819 	for (pass = 1; pass < 3; pass++) {
820 		memset(&info, 0, sizeof(info));
821 		switch (cmdpass(cmd, pass)) {
822 		case cmdpass(RTM_ADD, 1):
823 		case cmdpass(RTM_CHANGE, 1):
824 		case cmdpass(RTM_DELETE, 2):
825 			if (cmd == RTM_ADD)
826 				ncmd = RTM_NEWADDR;
827 			else
828 				ncmd = RTM_DELADDR;
829 
830 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
831 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
832 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
833 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
834 			memset(&ifam, 0, sizeof(ifam));
835 			ifam.ifam_index = ifp->if_index;
836 			ifam.ifam_metric = ifa->ifa_metric;
837 			ifam.ifam_flags = ifa->ifa_flags;
838 			m = rt_msg1(ncmd, &info, &ifam, sizeof(ifam));
839 			if (m == NULL)
840 				continue;
841 			mtod(m, struct ifa_msghdr *)->ifam_addrs =
842 			    info.rti_addrs;
843 			break;
844 		case cmdpass(RTM_ADD, 2):
845 		case cmdpass(RTM_CHANGE, 2):
846 		case cmdpass(RTM_DELETE, 1):
847 			if (rt == NULL)
848 				continue;
849 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
850 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
851 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
852 			memset(&rtm, 0, sizeof(rtm));
853 			rtm.rtm_index = ifp->if_index;
854 			rtm.rtm_flags |= rt->rt_flags;
855 			rtm.rtm_errno = error;
856 			m = rt_msg1(cmd, &info, &rtm, sizeof(rtm));
857 			if (m == NULL)
858 				continue;
859 			mtod(m, struct rt_msghdr *)->rtm_addrs = info.rti_addrs;
860 			break;
861 		default:
862 			continue;
863 		}
864 #ifdef DIAGNOSTIC
865 		if (m == NULL)
866 			panic("%s: called with wrong command", __func__);
867 #endif
868 		route_enqueue(m, sa ? sa->sa_family : 0);
869 	}
870 #undef cmdpass
871 }
872 
873 static struct mbuf *
874 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
875     struct rt_addrinfo *info)
876 {
877 	struct if_announcemsghdr ifan;
878 
879 	memset(info, 0, sizeof(*info));
880 	memset(&ifan, 0, sizeof(ifan));
881 	ifan.ifan_index = ifp->if_index;
882 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
883 	ifan.ifan_what = what;
884 	return rt_msg1(type, info, &ifan, sizeof(ifan));
885 }
886 
887 /*
888  * This is called to generate routing socket messages indicating
889  * network interface arrival and departure.
890  */
891 void
892 rt_ifannouncemsg(struct ifnet *ifp, int what)
893 {
894 	struct mbuf *m;
895 	struct rt_addrinfo info;
896 
897 	if (route_cb.any_count == 0)
898 		return;
899 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
900 	if (m == NULL)
901 		return;
902 	route_enqueue(m, 0);
903 }
904 
905 /*
906  * This is called to generate routing socket messages indicating
907  * IEEE80211 wireless events.
908  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
909  */
910 void
911 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
912 {
913 	struct mbuf *m;
914 	struct rt_addrinfo info;
915 
916 	if (route_cb.any_count == 0)
917 		return;
918 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
919 	if (m == NULL)
920 		return;
921 	/*
922 	 * Append the ieee80211 data.  Try to stick it in the
923 	 * mbuf containing the ifannounce msg; otherwise allocate
924 	 * a new mbuf and append.
925 	 *
926 	 * NB: we assume m is a single mbuf.
927 	 */
928 	if (data_len > M_TRAILINGSPACE(m)) {
929 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
930 		if (n == NULL) {
931 			m_freem(m);
932 			return;
933 		}
934 		(void)memcpy(mtod(n, void *), data, data_len);
935 		n->m_len = data_len;
936 		m->m_next = n;
937 	} else if (data_len > 0) {
938 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
939 		m->m_len += data_len;
940 	}
941 	if (m->m_flags & M_PKTHDR)
942 		m->m_pkthdr.len += data_len;
943 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
944 	route_enqueue(m, 0);
945 }
946 
947 /*
948  * This is used in dumping the kernel table via sysctl().
949  */
950 static int
951 sysctl_dumpentry(struct rtentry *rt, void *v)
952 {
953 	struct rt_walkarg *w = v;
954 	int error = 0, size;
955 	struct rt_addrinfo info;
956 
957 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
958 		return 0;
959 	memset(&info, 0, sizeof(info));
960 	info.rti_info[RTAX_DST] = rt_getkey(rt);
961 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
962 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
963 	if (rt->rt_ifp) {
964 		const struct ifaddr *rtifa;
965 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
966 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
967 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
968 		 * --dyoung
969 		 */
970 		rtifa = rt_get_ifa(rt);
971 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
972 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
973 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
974 	}
975 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
976 		return error;
977 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
978 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
979 
980 		rtm->rtm_flags = rt->rt_flags;
981 		rtm->rtm_use = rt->rt_use;
982 		cvtmetrics(&rtm->rtm_rmx, &rt->rt_rmx);
983 		KASSERT(rt->rt_ifp != NULL);
984 		rtm->rtm_index = rt->rt_ifp->if_index;
985 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
986 		rtm->rtm_addrs = info.rti_addrs;
987 		if ((error = copyout(rtm, w->w_where, size)) != 0)
988 			w->w_where = NULL;
989 		else
990 			w->w_where = (char *)w->w_where + size;
991 	}
992 	return error;
993 }
994 
995 static int
996 sysctl_iflist(int af, struct rt_walkarg *w, int type)
997 {
998 	struct ifnet *ifp;
999 	struct ifaddr *ifa;
1000 	struct	rt_addrinfo info;
1001 	int	len, error = 0;
1002 
1003 	memset(&info, 0, sizeof(info));
1004 	IFNET_FOREACH(ifp) {
1005 		if (w->w_arg && w->w_arg != ifp->if_index)
1006 			continue;
1007 		if (IFADDR_EMPTY(ifp))
1008 			continue;
1009 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1010 		switch (type) {
1011 		case NET_RT_IFLIST:
1012 			error = rt_msg2(RTM_IFINFO, &info, NULL, w, &len);
1013 			break;
1014 #ifdef COMPAT_14
1015 		case NET_RT_OOIFLIST:
1016 			error = rt_msg2(RTM_OOIFINFO, &info, NULL, w, &len);
1017 			break;
1018 #endif
1019 #ifdef COMPAT_50
1020 		case NET_RT_OIFLIST:
1021 			error = rt_msg2(RTM_OIFINFO, &info, NULL, w, &len);
1022 			break;
1023 #endif
1024 		default:
1025 			panic("sysctl_iflist(1)");
1026 		}
1027 		if (error)
1028 			return error;
1029 		info.rti_info[RTAX_IFP] = NULL;
1030 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1031 			switch (type) {
1032 			case NET_RT_IFLIST: {
1033 				struct if_msghdr *ifm;
1034 
1035 				ifm = (struct if_msghdr *)w->w_tmem;
1036 				ifm->ifm_index = ifp->if_index;
1037 				ifm->ifm_flags = ifp->if_flags;
1038 				ifm->ifm_data = ifp->if_data;
1039 				ifm->ifm_addrs = info.rti_addrs;
1040 				error = copyout(ifm, w->w_where, len);
1041 				if (error)
1042 					return error;
1043 				w->w_where = (char *)w->w_where + len;
1044 				break;
1045 			}
1046 
1047 #ifdef COMPAT_14
1048 			case NET_RT_OOIFLIST:
1049 				error = compat_14_iflist(ifp, w, &info, len);
1050 				if (error)
1051 					return error;
1052 				break;
1053 #endif
1054 #ifdef COMPAT_50
1055 			case NET_RT_OIFLIST:
1056 				error = compat_50_iflist(ifp, w, &info, len);
1057 				if (error)
1058 					return error;
1059 				break;
1060 #endif
1061 			default:
1062 				panic("sysctl_iflist(2)");
1063 			}
1064 		}
1065 		IFADDR_FOREACH(ifa, ifp) {
1066 			if (af && af != ifa->ifa_addr->sa_family)
1067 				continue;
1068 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1069 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1070 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1071 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
1072 				return error;
1073 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1074 				struct ifa_msghdr *ifam;
1075 
1076 				ifam = (struct ifa_msghdr *)w->w_tmem;
1077 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1078 				ifam->ifam_flags = ifa->ifa_flags;
1079 				ifam->ifam_metric = ifa->ifa_metric;
1080 				ifam->ifam_addrs = info.rti_addrs;
1081 				error = copyout(w->w_tmem, w->w_where, len);
1082 				if (error)
1083 					return error;
1084 				w->w_where = (char *)w->w_where + len;
1085 			}
1086 		}
1087 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1088 		    info.rti_info[RTAX_BRD] = NULL;
1089 	}
1090 	return 0;
1091 }
1092 
1093 static int
1094 sysctl_rtable(SYSCTLFN_ARGS)
1095 {
1096 	void 	*where = oldp;
1097 	size_t	*given = oldlenp;
1098 	const void *new = newp;
1099 	int	i, s, error = EINVAL;
1100 	u_char  af;
1101 	struct	rt_walkarg w;
1102 
1103 	if (namelen == 1 && name[0] == CTL_QUERY)
1104 		return sysctl_query(SYSCTLFN_CALL(rnode));
1105 
1106 	if (new)
1107 		return EPERM;
1108 	if (namelen != 3)
1109 		return EINVAL;
1110 	af = name[0];
1111 	w.w_tmemneeded = 0;
1112 	w.w_tmemsize = 0;
1113 	w.w_tmem = NULL;
1114 again:
1115 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1116 	if (w.w_tmemneeded) {
1117 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
1118 		w.w_tmemsize = w.w_tmemneeded;
1119 		w.w_tmemneeded = 0;
1120 	}
1121 	w.w_op = name[1];
1122 	w.w_arg = name[2];
1123 	w.w_given = *given;
1124 	w.w_needed = 0 - w.w_given;
1125 	w.w_where = where;
1126 
1127 	s = splsoftnet();
1128 	switch (w.w_op) {
1129 
1130 	case NET_RT_DUMP:
1131 	case NET_RT_FLAGS:
1132 		for (i = 1; i <= AF_MAX; i++)
1133 			if ((af == 0 || af == i) &&
1134 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
1135 				break;
1136 		break;
1137 
1138 #ifdef COMPAT_14
1139 	case NET_RT_OOIFLIST:
1140 		error = sysctl_iflist(af, &w, w.w_op);
1141 		break;
1142 #endif
1143 #ifdef COMPAT_50
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 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 	sysctl_net_route_setup(NULL);
1223 	route_intrq.ifq_maxlen = route_maxqlen;
1224 	route_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
1225 	    route_intr, NULL);
1226 }
1227 
1228 /*
1229  * Definitions of protocols supported in the ROUTE domain.
1230  */
1231 PR_WRAP_USRREQ(route_usrreq)
1232 #define	route_usrreq	route_usrreq_wrapper
1233 
1234 const struct protosw routesw[] = {
1235 	{
1236 		.pr_type = SOCK_RAW,
1237 		.pr_domain = &routedomain,
1238 		.pr_flags = PR_ATOMIC|PR_ADDR,
1239 		.pr_input = raw_input,
1240 		.pr_output = route_output,
1241 		.pr_ctlinput = raw_ctlinput,
1242 		.pr_usrreq = route_usrreq,
1243 		.pr_init = raw_init,
1244 	},
1245 };
1246 
1247 struct domain routedomain = {
1248 	.dom_family = PF_ROUTE,
1249 	.dom_name = "route",
1250 	.dom_init = route_init,
1251 	.dom_protosw = routesw,
1252 	.dom_protoswNPROTOSW = &routesw[__arraycount(routesw)],
1253 };
1254 
1255 static void
1256 sysctl_net_route_setup(struct sysctllog **clog)
1257 {
1258 	const struct sysctlnode *rnode = NULL;
1259 
1260 	sysctl_createv(clog, 0, NULL, NULL,
1261 		       CTLFLAG_PERMANENT,
1262 		       CTLTYPE_NODE, "net", NULL,
1263 		       NULL, 0, NULL, 0,
1264 		       CTL_NET, CTL_EOL);
1265 
1266 	sysctl_createv(clog, 0, NULL, &rnode,
1267 		       CTLFLAG_PERMANENT,
1268 		       CTLTYPE_NODE, "route",
1269 		       SYSCTL_DESCR("PF_ROUTE information"),
1270 		       NULL, 0, NULL, 0,
1271 		       CTL_NET, PF_ROUTE, CTL_EOL);
1272 	sysctl_createv(clog, 0, NULL, NULL,
1273 		       CTLFLAG_PERMANENT,
1274 		       CTLTYPE_NODE, "rtable",
1275 		       SYSCTL_DESCR("Routing table information"),
1276 		       sysctl_rtable, 0, NULL, 0,
1277 		       CTL_NET, PF_ROUTE, 0 /* any protocol */, CTL_EOL);
1278 	sysctl_createv(clog, 0, &rnode, NULL,
1279 		       CTLFLAG_PERMANENT,
1280 		       CTLTYPE_STRUCT, "stats",
1281 		       SYSCTL_DESCR("Routing statistics"),
1282 		       NULL, 0, &rtstat, sizeof(rtstat),
1283 		       CTL_CREATE, CTL_EOL);
1284 }
1285