xref: /netbsd-src/sys/net/rtsock.c (revision da9817918ec7e88db2912a2882967c7570a83f47)
1 /*	$NetBSD: rtsock.c,v 1.125 2009/04/02 21:02:06 christos 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 1.125 2009/04/02 21:02:06 christos 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 int sysctl_dumpentry(struct rtentry *, void *);
110 static int sysctl_iflist(int, struct rt_walkarg *, int);
111 static int sysctl_rtable(SYSCTLFN_PROTO);
112 static void rt_adjustcount(int, int);
113 
114 static void
115 rt_adjustcount(int af, int cnt)
116 {
117 	route_cb.any_count += cnt;
118 	switch (af) {
119 	case AF_INET:
120 		route_cb.ip_count += cnt;
121 		return;
122 #ifdef INET6
123 	case AF_INET6:
124 		route_cb.ip6_count += cnt;
125 		return;
126 #endif
127 	case AF_IPX:
128 		route_cb.ipx_count += cnt;
129 		return;
130 	case AF_NS:
131 		route_cb.ns_count += cnt;
132 		return;
133 	case AF_ISO:
134 		route_cb.iso_count += cnt;
135 		return;
136 	}
137 }
138 
139 static void
140 cvtmetrics(struct rt_metrics *ortm, const struct nrt_metrics *rtm)
141 {
142 	ortm->rmx_locks = rtm->rmx_locks;
143 	ortm->rmx_mtu = rtm->rmx_mtu;
144 	ortm->rmx_hopcount = rtm->rmx_hopcount;
145 	ortm->rmx_expire = rtm->rmx_expire;
146 	ortm->rmx_recvpipe = rtm->rmx_recvpipe;
147 	ortm->rmx_sendpipe = rtm->rmx_sendpipe;
148 	ortm->rmx_ssthresh = rtm->rmx_ssthresh;
149 	ortm->rmx_rtt = rtm->rmx_rtt;
150 	ortm->rmx_rttvar = rtm->rmx_rttvar;
151 	ortm->rmx_pksent = rtm->rmx_pksent;
152 }
153 
154 /*ARGSUSED*/
155 int
156 route_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
157 	struct mbuf *control, struct lwp *l)
158 {
159 	int error = 0;
160 	struct rawcb *rp = sotorawcb(so);
161 	int s;
162 
163 	if (req == PRU_ATTACH) {
164 		sosetlock(so);
165 		rp = malloc(sizeof(*rp), M_PCB, M_WAITOK|M_ZERO);
166 		so->so_pcb = rp;
167 	}
168 	if (req == PRU_DETACH && rp)
169 		rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
170 	s = splsoftnet();
171 
172 	/*
173 	 * Don't call raw_usrreq() in the attach case, because
174 	 * we want to allow non-privileged processes to listen on
175 	 * and send "safe" commands to the routing socket.
176 	 */
177 	if (req == PRU_ATTACH) {
178 		if (l == NULL)
179 			error = EACCES;
180 		else
181 			error = raw_attach(so, (int)(long)nam);
182 	} else
183 		error = raw_usrreq(so, req, m, nam, control, l);
184 
185 	rp = sotorawcb(so);
186 	if (req == PRU_ATTACH && rp) {
187 		if (error) {
188 			free(rp, M_PCB);
189 			splx(s);
190 			return error;
191 		}
192 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
193 		rp->rcb_laddr = &route_src;
194 		rp->rcb_faddr = &route_dst;
195 		soisconnected(so);
196 		so->so_options |= SO_USELOOPBACK;
197 	}
198 	splx(s);
199 	return error;
200 }
201 
202 static const struct sockaddr *
203 intern_netmask(const struct sockaddr *mask)
204 {
205 	struct radix_node *rn;
206 	extern struct radix_node_head *mask_rnhead;
207 
208 	if (mask != NULL &&
209 	    (rn = rn_search(mask, mask_rnhead->rnh_treetop)))
210 		mask = (const struct sockaddr *)rn->rn_key;
211 
212 	return mask;
213 }
214 
215 /*ARGSUSED*/
216 int
217 route_output(struct mbuf *m, ...)
218 {
219 	struct sockproto proto = { .sp_family = PF_ROUTE, };
220 	struct rt_msghdr *rtm = NULL;
221 	struct rt_msghdr *old_rtm = NULL;
222 	struct rtentry *rt = NULL;
223 	struct rtentry *saved_nrt = NULL;
224 	struct rt_addrinfo info;
225 	int len, error = 0;
226 	struct ifnet *ifp = NULL;
227 	struct ifaddr *ifa = NULL;
228 	struct socket *so;
229 	va_list ap;
230 	sa_family_t family;
231 
232 	va_start(ap, m);
233 	so = va_arg(ap, struct socket *);
234 	va_end(ap);
235 
236 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
237 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
238 	   (m = m_pullup(m, sizeof(int32_t))) == NULL))
239 		return ENOBUFS;
240 	if ((m->m_flags & M_PKTHDR) == 0)
241 		panic("route_output");
242 	len = m->m_pkthdr.len;
243 	if (len < sizeof(*rtm) ||
244 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
245 		info.rti_info[RTAX_DST] = NULL;
246 		senderr(EINVAL);
247 	}
248 	R_Malloc(rtm, struct rt_msghdr *, len);
249 	if (rtm == NULL) {
250 		info.rti_info[RTAX_DST] = NULL;
251 		senderr(ENOBUFS);
252 	}
253 	m_copydata(m, 0, len, rtm);
254 	if (rtm->rtm_version != RTM_VERSION) {
255 		info.rti_info[RTAX_DST] = NULL;
256 		senderr(EPROTONOSUPPORT);
257 	}
258 	rtm->rtm_pid = curproc->p_pid;
259 	memset(&info, 0, sizeof(info));
260 	info.rti_addrs = rtm->rtm_addrs;
261 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
262 	    &info))
263 		senderr(EINVAL);
264 	info.rti_flags = rtm->rtm_flags;
265 #ifdef RTSOCK_DEBUG
266 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
267 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
268 		    inet_ntoa(((const struct sockaddr_in *)
269 		    info.rti_info[RTAX_DST])->sin_addr));
270 	}
271 #endif /* RTSOCK_DEBUG */
272 	if (info.rti_info[RTAX_DST] == NULL ||
273 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX))
274 		senderr(EINVAL);
275 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
276 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
277 		senderr(EINVAL);
278 
279 	/*
280 	 * Verify that the caller has the appropriate privilege; RTM_GET
281 	 * is the only operation the non-superuser is allowed.
282 	 */
283 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
284 	    0, rtm, NULL, NULL) != 0)
285 		senderr(EACCES);
286 
287 	switch (rtm->rtm_type) {
288 
289 	case RTM_ADD:
290 		if (info.rti_info[RTAX_GATEWAY] == NULL)
291 			senderr(EINVAL);
292 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
293 		if (error == 0 && saved_nrt) {
294 			rt_setmetrics(rtm->rtm_inits,
295 			    &rtm->rtm_rmx, &saved_nrt->rt_rmx);
296 			saved_nrt->rt_refcnt--;
297 		}
298 		break;
299 
300 	case RTM_DELETE:
301 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
302 		if (error == 0) {
303 			(rt = saved_nrt)->rt_refcnt++;
304 			goto report;
305 		}
306 		break;
307 
308 	case RTM_GET:
309 	case RTM_CHANGE:
310 	case RTM_LOCK:
311                 /* XXX This will mask info.rti_info[RTAX_DST] with
312 		 * info.rti_info[RTAX_NETMASK] before
313                  * searching.  It did not used to do that.  --dyoung
314 		 */
315 		error = rtrequest1(RTM_GET, &info, &rt);
316 		if (error != 0)
317 			senderr(error);
318 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
319 			struct radix_node *rn;
320 
321 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
322 			    info.rti_info[RTAX_DST]->sa_len) != 0)
323 				senderr(ESRCH);
324 			info.rti_info[RTAX_NETMASK] = intern_netmask(
325 			    info.rti_info[RTAX_NETMASK]);
326 			for (rn = rt->rt_nodes; rn; rn = rn->rn_dupedkey)
327 				if (info.rti_info[RTAX_NETMASK] ==
328 				    (const struct sockaddr *)rn->rn_mask)
329 					break;
330 			if (rn == NULL)
331 				senderr(ETOOMANYREFS);
332 			rt = (struct rtentry *)rn;
333 		}
334 
335 		switch (rtm->rtm_type) {
336 		case RTM_GET:
337 		report:
338 			info.rti_info[RTAX_DST] = rt_getkey(rt);
339 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
340 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
341 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
342 				;
343 			else if ((ifp = rt->rt_ifp) != NULL) {
344 				const struct ifaddr *rtifa;
345 				info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
346                                 /* rtifa used to be simply rt->rt_ifa.
347                                  * If rt->rt_ifa != NULL, then
348                                  * rt_get_ifa() != NULL.  So this
349                                  * ought to still be safe. --dyoung
350 				 */
351 				rtifa = rt_get_ifa(rt);
352 				info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
353 #ifdef RTSOCK_DEBUG
354 				if (info.rti_info[RTAX_IFA]->sa_family ==
355 				    AF_INET) {
356 					printf("%s: copying out RTAX_IFA %s ",
357 					    __func__, inet_ntoa(
358 					    (const struct sockaddr_in *)
359 					    info.rti_info[RTAX_IFA])->sin_addr);
360 					printf("for info.rti_info[RTAX_DST] %s "
361 					    "ifa_getifa %p ifa_seqno %p\n",
362 					    inet_ntoa(
363 					    (const struct sockaddr_in *)
364 					    info.rti_info[RTAX_DST])->sin_addr),
365 					    (void *)rtifa->ifa_getifa,
366 					    rtifa->ifa_seqno);
367 				}
368 #endif /* RTSOCK_DEBUG */
369 				if (ifp->if_flags & IFF_POINTOPOINT) {
370 					info.rti_info[RTAX_BRD] =
371 					    rtifa->ifa_dstaddr;
372 				} else
373 					info.rti_info[RTAX_BRD] = NULL;
374 				rtm->rtm_index = ifp->if_index;
375 			} else {
376 				info.rti_info[RTAX_IFP] = NULL;
377 				info.rti_info[RTAX_IFA] = NULL;
378 			}
379 			(void)rt_msg2(rtm->rtm_type, &info, NULL, NULL, &len);
380 			if (len > rtm->rtm_msglen) {
381 				old_rtm = rtm;
382 				R_Malloc(rtm, struct rt_msghdr *, len);
383 				if (rtm == NULL)
384 					senderr(ENOBUFS);
385 				(void)memcpy(rtm, old_rtm, old_rtm->rtm_msglen);
386 			}
387 			(void)rt_msg2(rtm->rtm_type, &info, rtm, NULL, 0);
388 			rtm->rtm_flags = rt->rt_flags;
389 			cvtmetrics(&rtm->rtm_rmx, &rt->rt_rmx);
390 			rtm->rtm_addrs = info.rti_addrs;
391 			break;
392 
393 		case RTM_CHANGE:
394 			/*
395 			 * new gateway could require new ifaddr, ifp;
396 			 * flags may also be different; ifp may be specified
397 			 * by ll sockaddr when protocol address is ambiguous
398 			 */
399 			if ((error = rt_getifa(&info)) != 0)
400 				senderr(error);
401 			if (info.rti_info[RTAX_GATEWAY] &&
402 			    rt_setgate(rt, info.rti_info[RTAX_GATEWAY]))
403 				senderr(EDQUOT);
404 			/* new gateway could require new ifaddr, ifp;
405 			   flags may also be different; ifp may be specified
406 			   by ll sockaddr when protocol address is ambiguous */
407 			if (info.rti_info[RTAX_IFP] &&
408 			    (ifa = ifa_ifwithnet(info.rti_info[RTAX_IFP])) &&
409 			    (ifp = ifa->ifa_ifp) && (info.rti_info[RTAX_IFA] ||
410 			    info.rti_info[RTAX_GATEWAY])) {
411 				ifa = ifaof_ifpforaddr(info.rti_info[RTAX_IFA] ?
412 				    info.rti_info[RTAX_IFA] :
413 				    info.rti_info[RTAX_GATEWAY], ifp);
414 			} else if ((info.rti_info[RTAX_IFA] &&
415 			    (ifa = ifa_ifwithaddr(info.rti_info[RTAX_IFA]))) ||
416 			    (info.rti_info[RTAX_GATEWAY] &&
417 			    (ifa = ifa_ifwithroute(rt->rt_flags,
418 			    rt_getkey(rt), info.rti_info[RTAX_GATEWAY])))) {
419 				ifp = ifa->ifa_ifp;
420 			}
421 			if (ifa) {
422 				struct ifaddr *oifa = rt->rt_ifa;
423 				if (oifa != ifa) {
424 					if (oifa && oifa->ifa_rtrequest) {
425 						oifa->ifa_rtrequest(RTM_DELETE,
426 						    rt, &info);
427 					}
428 					rt_replace_ifa(rt, ifa);
429 					rt->rt_ifp = ifp;
430 				}
431 			}
432 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
433 			    &rt->rt_rmx);
434 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
435 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
436 			/*FALLTHROUGH*/
437 		case RTM_LOCK:
438 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
439 			rt->rt_rmx.rmx_locks |=
440 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
441 			break;
442 		}
443 		break;
444 
445 	default:
446 		senderr(EOPNOTSUPP);
447 	}
448 
449 flush:
450 	if (rtm) {
451 		if (error)
452 			rtm->rtm_errno = error;
453 		else
454 			rtm->rtm_flags |= RTF_DONE;
455 	}
456 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
457 	    0;
458 	/* We cannot free old_rtm until we have stopped using the
459 	 * pointers in info, some of which may point to sockaddrs
460 	 * in old_rtm.
461 	 */
462 	if (old_rtm != NULL)
463 		Free(old_rtm);
464 	if (rt)
465 		rtfree(rt);
466     {
467 	struct rawcb *rp = NULL;
468 	/*
469 	 * Check to see if we don't want our own messages.
470 	 */
471 	if ((so->so_options & SO_USELOOPBACK) == 0) {
472 		if (route_cb.any_count <= 1) {
473 			if (rtm)
474 				Free(rtm);
475 			m_freem(m);
476 			return error;
477 		}
478 		/* There is another listener, so construct message */
479 		rp = sotorawcb(so);
480 	}
481 	if (rtm) {
482 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
483 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
484 			m_freem(m);
485 			m = NULL;
486 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
487 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
488 		Free(rtm);
489 	}
490 	if (rp)
491 		rp->rcb_proto.sp_family = 0; /* Avoid us */
492 	if (family)
493 		proto.sp_protocol = family;
494 	if (m)
495 		raw_input(m, &proto, &route_src, &route_dst);
496 	if (rp)
497 		rp->rcb_proto.sp_family = PF_ROUTE;
498     }
499 	return error;
500 }
501 
502 void
503 rt_setmetrics(u_long which, const struct rt_metrics *in, struct nrt_metrics *out)
504 {
505 #define metric(f, e) if (which & (f)) out->e = in->e;
506 	metric(RTV_RPIPE, rmx_recvpipe);
507 	metric(RTV_SPIPE, rmx_sendpipe);
508 	metric(RTV_SSTHRESH, rmx_ssthresh);
509 	metric(RTV_RTT, rmx_rtt);
510 	metric(RTV_RTTVAR, rmx_rttvar);
511 	metric(RTV_HOPCOUNT, rmx_hopcount);
512 	metric(RTV_MTU, rmx_mtu);
513 	/* XXX time_t: Will not work after February 2145 (u_long time) */
514 	metric(RTV_EXPIRE, rmx_expire);
515 #undef metric
516 }
517 
518 static int
519 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
520     struct rt_addrinfo *rtinfo)
521 {
522 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
523 	int i;
524 
525 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
526 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
527 			continue;
528 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
529 		RT_ADVANCE(cp, sa);
530 	}
531 
532 	/*
533 	 * Check for extra addresses specified, except RTM_GET asking
534 	 * for interface info.
535 	 */
536 	if (rtmtype == RTM_GET) {
537 		if (((rtinfo->rti_addrs &
538 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
539 			return 1;
540 	} else if ((rtinfo->rti_addrs & (~0 << i)) != 0)
541 		return 1;
542 	/* Check for bad data length.  */
543 	if (cp != cplim) {
544 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
545 		    cp - RT_ROUNDUP(sa->sa_len) + sa->sa_len == cplim)
546 			/*
547 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
548 			 * We accept this for now for the sake of old
549 			 * binaries or third party softwares.
550 			 */
551 			;
552 		else
553 			return 1;
554 	}
555 	return 0;
556 }
557 
558 struct mbuf *
559 rt_msg1(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
560 {
561 	struct rt_msghdr *rtm;
562 	struct mbuf *m;
563 	int i;
564 	const struct sockaddr *sa;
565 	int len, dlen;
566 
567 	m = m_gethdr(M_DONTWAIT, MT_DATA);
568 	if (m == NULL)
569 		return m;
570 	MCLAIM(m, &routedomain.dom_mowner);
571 	switch (type) {
572 
573 	case RTM_DELADDR:
574 	case RTM_NEWADDR:
575 		len = sizeof(struct ifa_msghdr);
576 		break;
577 
578 #ifdef COMPAT_14
579 	case RTM_OOIFINFO:
580 		len = sizeof(struct if_msghdr14);
581 		break;
582 #endif
583 #ifdef COMPAT_50
584 	case RTM_OIFINFO:
585 		len = sizeof(struct if_msghdr50);
586 		break;
587 #endif
588 
589 	case RTM_IFINFO:
590 		len = sizeof(struct if_msghdr);
591 		break;
592 
593 	case RTM_IFANNOUNCE:
594 	case RTM_IEEE80211:
595 		len = sizeof(struct if_announcemsghdr);
596 		break;
597 
598 	default:
599 		len = sizeof(struct rt_msghdr);
600 	}
601 	if (len > MHLEN + MLEN)
602 		panic("rt_msg1: message too long");
603 	else if (len > MHLEN) {
604 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
605 		if (m->m_next == NULL) {
606 			m_freem(m);
607 			return NULL;
608 		}
609 		MCLAIM(m->m_next, m->m_owner);
610 		m->m_pkthdr.len = len;
611 		m->m_len = MHLEN;
612 		m->m_next->m_len = len - MHLEN;
613 	} else {
614 		m->m_pkthdr.len = m->m_len = len;
615 	}
616 	m->m_pkthdr.rcvif = NULL;
617 	m_copyback(m, 0, datalen, data);
618 	if (len > datalen)
619 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
620 	rtm = mtod(m, struct rt_msghdr *);
621 	for (i = 0; i < RTAX_MAX; i++) {
622 		if ((sa = rtinfo->rti_info[i]) == NULL)
623 			continue;
624 		rtinfo->rti_addrs |= (1 << i);
625 		dlen = RT_ROUNDUP(sa->sa_len);
626 		m_copyback(m, len, dlen, sa);
627 		len += dlen;
628 	}
629 	if (m->m_pkthdr.len != len) {
630 		m_freem(m);
631 		return NULL;
632 	}
633 	rtm->rtm_msglen = len;
634 	rtm->rtm_version = RTM_VERSION;
635 	rtm->rtm_type = type;
636 	return m;
637 }
638 
639 /*
640  * rt_msg2
641  *
642  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
643  *		returns the length of the message in 'lenp'.
644  *
645  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
646  *	the message
647  * otherwise walkarg's w_needed is updated and if the user buffer is
648  *	specified and w_needed indicates space exists the information is copied
649  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
650  *	if the allocation fails ENOBUFS is returned.
651  */
652 static int
653 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
654 	int *lenp)
655 {
656 	int i;
657 	int len, dlen, second_time = 0;
658 	char *cp0, *cp = cpv;
659 
660 	rtinfo->rti_addrs = 0;
661 again:
662 	switch (type) {
663 
664 	case RTM_DELADDR:
665 	case RTM_NEWADDR:
666 		len = sizeof(struct ifa_msghdr);
667 		break;
668 #ifdef COMPAT_14
669 	case RTM_OOIFINFO:
670 		len = sizeof(struct if_msghdr14);
671 		break;
672 #endif
673 #ifdef COMPAT_50
674 	case RTM_OIFINFO:
675 		len = sizeof(struct if_msghdr50);
676 		break;
677 #endif
678 
679 	case RTM_IFINFO:
680 		len = sizeof(struct if_msghdr);
681 		break;
682 
683 	default:
684 		len = sizeof(struct rt_msghdr);
685 	}
686 	if ((cp0 = cp) != NULL)
687 		cp += len;
688 	for (i = 0; i < RTAX_MAX; i++) {
689 		const struct sockaddr *sa;
690 
691 		if ((sa = rtinfo->rti_info[i]) == NULL)
692 			continue;
693 		rtinfo->rti_addrs |= (1 << i);
694 		dlen = RT_ROUNDUP(sa->sa_len);
695 		if (cp) {
696 			(void)memcpy(cp, sa, (size_t)dlen);
697 			cp += dlen;
698 		}
699 		len += dlen;
700 	}
701 	if (cp == NULL && w != NULL && !second_time) {
702 		struct rt_walkarg *rw = w;
703 
704 		rw->w_needed += len;
705 		if (rw->w_needed <= 0 && rw->w_where) {
706 			if (rw->w_tmemsize < len) {
707 				if (rw->w_tmem)
708 					free(rw->w_tmem, M_RTABLE);
709 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
710 				if (rw->w_tmem)
711 					rw->w_tmemsize = len;
712 				else
713 					rw->w_tmemsize = 0;
714 			}
715 			if (rw->w_tmem) {
716 				cp = rw->w_tmem;
717 				second_time = 1;
718 				goto again;
719 			} else {
720 				rw->w_tmemneeded = len;
721 				return ENOBUFS;
722 			}
723 		}
724 	}
725 	if (cp) {
726 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
727 
728 		rtm->rtm_version = RTM_VERSION;
729 		rtm->rtm_type = type;
730 		rtm->rtm_msglen = len;
731 	}
732 	if (lenp)
733 		*lenp = len;
734 	return 0;
735 }
736 
737 /*
738  * This routine is called to generate a message from the routing
739  * socket indicating that a redirect has occurred, a routing lookup
740  * has failed, or that a protocol has detected timeouts to a particular
741  * destination.
742  */
743 void
744 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
745 {
746 	struct rt_msghdr rtm;
747 	struct mbuf *m;
748 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
749 
750 	if (route_cb.any_count == 0)
751 		return;
752 	memset(&rtm, 0, sizeof(rtm));
753 	rtm.rtm_flags = RTF_DONE | flags;
754 	rtm.rtm_errno = error;
755 	m = rt_msg1(type, rtinfo, &rtm, sizeof(rtm));
756 	if (m == NULL)
757 		return;
758 	mtod(m, struct rt_msghdr *)->rtm_addrs = rtinfo->rti_addrs;
759 	route_enqueue(m, sa ? sa->sa_family : 0);
760 }
761 
762 /*
763  * This routine is called to generate a message from the routing
764  * socket indicating that the status of a network interface has changed.
765  */
766 void
767 rt_ifmsg(struct ifnet *ifp)
768 {
769 	struct if_msghdr ifm;
770 	struct mbuf *m;
771 	struct rt_addrinfo info;
772 
773 	if (route_cb.any_count == 0)
774 		return;
775 	(void)memset(&info, 0, sizeof(info));
776 	(void)memset(&ifm, 0, sizeof(ifm));
777 	ifm.ifm_index = ifp->if_index;
778 	ifm.ifm_flags = ifp->if_flags;
779 	ifm.ifm_data = ifp->if_data;
780 	ifm.ifm_addrs = 0;
781 	m = rt_msg1(RTM_IFINFO, &info, &ifm, sizeof(ifm));
782 	if (m == NULL)
783 		return;
784 	route_enqueue(m, 0);
785 #ifdef COMPAT_14
786 	compat_14_rt_ifmsg(ifp, &ifm);
787 #endif
788 #ifdef COMPAT_50
789 	compat_50_rt_ifmsg(ifp, &ifm);
790 #endif
791 }
792 
793 
794 /*
795  * This is called to generate messages from the routing socket
796  * indicating a network interface has had addresses associated with it.
797  * if we ever reverse the logic and replace messages TO the routing
798  * socket indicate a request to configure interfaces, then it will
799  * be unnecessary as the routing socket will automatically generate
800  * copies of it.
801  */
802 void
803 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
804 {
805 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
806 	struct rt_addrinfo info;
807 	const struct sockaddr *sa;
808 	int pass;
809 	struct mbuf *m;
810 	struct ifnet *ifp = ifa->ifa_ifp;
811 	struct rt_msghdr rtm;
812 	struct ifa_msghdr ifam;
813 	int ncmd;
814 
815 	if (route_cb.any_count == 0)
816 		return;
817 	for (pass = 1; pass < 3; pass++) {
818 		memset(&info, 0, sizeof(info));
819 		switch (cmdpass(cmd, pass)) {
820 		case cmdpass(RTM_ADD, 1):
821 		case cmdpass(RTM_CHANGE, 1):
822 		case cmdpass(RTM_DELETE, 2):
823 			if (cmd == RTM_ADD)
824 				ncmd = RTM_NEWADDR;
825 			else
826 				ncmd = RTM_DELADDR;
827 
828 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
829 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
830 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
831 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
832 			memset(&ifam, 0, sizeof(ifam));
833 			ifam.ifam_index = ifp->if_index;
834 			ifam.ifam_metric = ifa->ifa_metric;
835 			ifam.ifam_flags = ifa->ifa_flags;
836 			m = rt_msg1(ncmd, &info, &ifam, sizeof(ifam));
837 			if (m == NULL)
838 				continue;
839 			mtod(m, struct ifa_msghdr *)->ifam_addrs =
840 			    info.rti_addrs;
841 			break;
842 		case cmdpass(RTM_ADD, 2):
843 		case cmdpass(RTM_CHANGE, 2):
844 		case cmdpass(RTM_DELETE, 1):
845 			if (rt == NULL)
846 				continue;
847 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
848 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
849 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
850 			memset(&rtm, 0, sizeof(rtm));
851 			rtm.rtm_index = ifp->if_index;
852 			rtm.rtm_flags |= rt->rt_flags;
853 			rtm.rtm_errno = error;
854 			m = rt_msg1(cmd, &info, &rtm, sizeof(rtm));
855 			if (m == NULL)
856 				continue;
857 			mtod(m, struct rt_msghdr *)->rtm_addrs = info.rti_addrs;
858 			break;
859 		default:
860 			continue;
861 		}
862 #ifdef DIAGNOSTIC
863 		if (m == NULL)
864 			panic("%s: called with wrong command", __func__);
865 #endif
866 		route_enqueue(m, sa ? sa->sa_family : 0);
867 	}
868 #undef cmdpass
869 }
870 
871 static struct mbuf *
872 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
873     struct rt_addrinfo *info)
874 {
875 	struct if_announcemsghdr ifan;
876 
877 	memset(info, 0, sizeof(*info));
878 	memset(&ifan, 0, sizeof(ifan));
879 	ifan.ifan_index = ifp->if_index;
880 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
881 	ifan.ifan_what = what;
882 	return rt_msg1(type, info, &ifan, sizeof(ifan));
883 }
884 
885 /*
886  * This is called to generate routing socket messages indicating
887  * network interface arrival and departure.
888  */
889 void
890 rt_ifannouncemsg(struct ifnet *ifp, int what)
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_IFANNOUNCE, what, &info);
898 	if (m == NULL)
899 		return;
900 	route_enqueue(m, 0);
901 }
902 
903 /*
904  * This is called to generate routing socket messages indicating
905  * IEEE80211 wireless events.
906  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
907  */
908 void
909 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
910 {
911 	struct mbuf *m;
912 	struct rt_addrinfo info;
913 
914 	if (route_cb.any_count == 0)
915 		return;
916 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
917 	if (m == NULL)
918 		return;
919 	/*
920 	 * Append the ieee80211 data.  Try to stick it in the
921 	 * mbuf containing the ifannounce msg; otherwise allocate
922 	 * a new mbuf and append.
923 	 *
924 	 * NB: we assume m is a single mbuf.
925 	 */
926 	if (data_len > M_TRAILINGSPACE(m)) {
927 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
928 		if (n == NULL) {
929 			m_freem(m);
930 			return;
931 		}
932 		(void)memcpy(mtod(n, void *), data, data_len);
933 		n->m_len = data_len;
934 		m->m_next = n;
935 	} else if (data_len > 0) {
936 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
937 		m->m_len += data_len;
938 	}
939 	if (m->m_flags & M_PKTHDR)
940 		m->m_pkthdr.len += data_len;
941 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
942 	route_enqueue(m, 0);
943 }
944 
945 /*
946  * This is used in dumping the kernel table via sysctl().
947  */
948 static int
949 sysctl_dumpentry(struct rtentry *rt, void *v)
950 {
951 	struct rt_walkarg *w = v;
952 	int error = 0, size;
953 	struct rt_addrinfo info;
954 
955 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
956 		return 0;
957 	memset(&info, 0, sizeof(info));
958 	info.rti_info[RTAX_DST] = rt_getkey(rt);
959 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
960 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
961 	if (rt->rt_ifp) {
962 		const struct ifaddr *rtifa;
963 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
964 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
965 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
966 		 * --dyoung
967 		 */
968 		rtifa = rt_get_ifa(rt);
969 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
970 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
971 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
972 	}
973 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
974 		return error;
975 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
976 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
977 
978 		rtm->rtm_flags = rt->rt_flags;
979 		rtm->rtm_use = rt->rt_use;
980 		cvtmetrics(&rtm->rtm_rmx, &rt->rt_rmx);
981 		KASSERT(rt->rt_ifp != NULL);
982 		rtm->rtm_index = rt->rt_ifp->if_index;
983 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
984 		rtm->rtm_addrs = info.rti_addrs;
985 		if ((error = copyout(rtm, w->w_where, size)) != 0)
986 			w->w_where = NULL;
987 		else
988 			w->w_where = (char *)w->w_where + size;
989 	}
990 	return error;
991 }
992 
993 static int
994 sysctl_iflist(int af, struct rt_walkarg *w, int type)
995 {
996 	struct ifnet *ifp;
997 	struct ifaddr *ifa;
998 	struct	rt_addrinfo info;
999 	int	len, error = 0;
1000 
1001 	memset(&info, 0, sizeof(info));
1002 	IFNET_FOREACH(ifp) {
1003 		if (w->w_arg && w->w_arg != ifp->if_index)
1004 			continue;
1005 		if (IFADDR_EMPTY(ifp))
1006 			continue;
1007 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1008 		switch (type) {
1009 		case NET_RT_IFLIST:
1010 			error = rt_msg2(RTM_IFINFO, &info, NULL, w, &len);
1011 			break;
1012 #ifdef COMPAT_14
1013 		case NET_RT_OOIFLIST:
1014 			error = rt_msg2(RTM_OOIFINFO, &info, NULL, w, &len);
1015 			break;
1016 #endif
1017 #ifdef COMPAT_50
1018 		case NET_RT_OIFLIST:
1019 			error = rt_msg2(RTM_OIFINFO, &info, NULL, w, &len);
1020 			break;
1021 #endif
1022 		default:
1023 			panic("sysctl_iflist(1)");
1024 		}
1025 		if (error)
1026 			return error;
1027 		info.rti_info[RTAX_IFP] = NULL;
1028 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1029 			switch (type) {
1030 			case NET_RT_IFLIST: {
1031 				struct if_msghdr *ifm;
1032 
1033 				ifm = (struct if_msghdr *)w->w_tmem;
1034 				ifm->ifm_index = ifp->if_index;
1035 				ifm->ifm_flags = ifp->if_flags;
1036 				ifm->ifm_data = ifp->if_data;
1037 				ifm->ifm_addrs = info.rti_addrs;
1038 				error = copyout(ifm, w->w_where, len);
1039 				if (error)
1040 					return error;
1041 				w->w_where = (char *)w->w_where + len;
1042 				break;
1043 			}
1044 
1045 #ifdef COMPAT_14
1046 			case NET_RT_OOIFLIST:
1047 				error = compat_14_iflist(ifp, w, &info, len);
1048 				if (error)
1049 					return error;
1050 				break;
1051 #endif
1052 #ifdef COMPAT_50
1053 			case NET_RT_OIFLIST:
1054 				error = compat_50_iflist(ifp, w, &info, len);
1055 				if (error)
1056 					return error;
1057 				break;
1058 #endif
1059 			default:
1060 				panic("sysctl_iflist(2)");
1061 			}
1062 		}
1063 		IFADDR_FOREACH(ifa, ifp) {
1064 			if (af && af != ifa->ifa_addr->sa_family)
1065 				continue;
1066 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1067 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1068 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1069 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
1070 				return error;
1071 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1072 				struct ifa_msghdr *ifam;
1073 
1074 				ifam = (struct ifa_msghdr *)w->w_tmem;
1075 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1076 				ifam->ifam_flags = ifa->ifa_flags;
1077 				ifam->ifam_metric = ifa->ifa_metric;
1078 				ifam->ifam_addrs = info.rti_addrs;
1079 				error = copyout(w->w_tmem, w->w_where, len);
1080 				if (error)
1081 					return error;
1082 				w->w_where = (char *)w->w_where + len;
1083 			}
1084 		}
1085 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1086 		    info.rti_info[RTAX_BRD] = NULL;
1087 	}
1088 	return 0;
1089 }
1090 
1091 static int
1092 sysctl_rtable(SYSCTLFN_ARGS)
1093 {
1094 	void 	*where = oldp;
1095 	size_t	*given = oldlenp;
1096 	const void *new = newp;
1097 	int	i, s, error = EINVAL;
1098 	u_char  af;
1099 	struct	rt_walkarg w;
1100 
1101 	if (namelen == 1 && name[0] == CTL_QUERY)
1102 		return sysctl_query(SYSCTLFN_CALL(rnode));
1103 
1104 	if (new)
1105 		return EPERM;
1106 	if (namelen != 3)
1107 		return EINVAL;
1108 	af = name[0];
1109 	w.w_tmemneeded = 0;
1110 	w.w_tmemsize = 0;
1111 	w.w_tmem = NULL;
1112 again:
1113 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1114 	if (w.w_tmemneeded) {
1115 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
1116 		w.w_tmemsize = w.w_tmemneeded;
1117 		w.w_tmemneeded = 0;
1118 	}
1119 	w.w_op = name[1];
1120 	w.w_arg = name[2];
1121 	w.w_given = *given;
1122 	w.w_needed = 0 - w.w_given;
1123 	w.w_where = where;
1124 
1125 	s = splsoftnet();
1126 	switch (w.w_op) {
1127 
1128 	case NET_RT_DUMP:
1129 	case NET_RT_FLAGS:
1130 		for (i = 1; i <= AF_MAX; i++)
1131 			if ((af == 0 || af == i) &&
1132 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
1133 				break;
1134 		break;
1135 
1136 #ifdef COMPAT_14
1137 	case NET_RT_OOIFLIST:
1138 		error = sysctl_iflist(af, &w, w.w_op);
1139 		break;
1140 #endif
1141 #ifdef COMPAT_50
1142 	case NET_RT_OIFLIST:
1143 		error = sysctl_iflist(af, &w, w.w_op);
1144 		break;
1145 #endif
1146 
1147 	case NET_RT_IFLIST:
1148 		error = sysctl_iflist(af, &w, w.w_op);
1149 	}
1150 	splx(s);
1151 
1152 	/* check to see if we couldn't allocate memory with NOWAIT */
1153 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1154 		goto again;
1155 
1156 	if (w.w_tmem)
1157 		free(w.w_tmem, M_RTABLE);
1158 	w.w_needed += w.w_given;
1159 	if (where) {
1160 		*given = (char *)w.w_where - (char *)where;
1161 		if (*given < w.w_needed)
1162 			return ENOMEM;
1163 	} else {
1164 		*given = (11 * w.w_needed) / 10;
1165 	}
1166 	return error;
1167 }
1168 
1169 /*
1170  * Routing message software interrupt routine
1171  */
1172 static void
1173 route_intr(void *cookie)
1174 {
1175 	struct sockproto proto = { .sp_family = PF_ROUTE, };
1176 	struct mbuf *m;
1177 	int s;
1178 
1179 	mutex_enter(softnet_lock);
1180 	KERNEL_LOCK(1, NULL);
1181 	while (!IF_IS_EMPTY(&route_intrq)) {
1182 		s = splnet();
1183 		IF_DEQUEUE(&route_intrq, m);
1184 		splx(s);
1185 		if (m == NULL)
1186 			break;
1187 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
1188 		raw_input(m, &proto, &route_src, &route_dst);
1189 	}
1190 	KERNEL_UNLOCK_ONE(NULL);
1191 	mutex_exit(softnet_lock);
1192 }
1193 
1194 /*
1195  * Enqueue a message to the software interrupt routine.
1196  */
1197 void
1198 route_enqueue(struct mbuf *m, int family)
1199 {
1200 	int s, wasempty;
1201 
1202 	s = splnet();
1203 	if (IF_QFULL(&route_intrq)) {
1204 		IF_DROP(&route_intrq);
1205 		m_freem(m);
1206 	} else {
1207 		wasempty = IF_IS_EMPTY(&route_intrq);
1208 		M_SETCTX(m, (uintptr_t)family);
1209 		IF_ENQUEUE(&route_intrq, m);
1210 		if (wasempty)
1211 			softint_schedule(route_sih);
1212 	}
1213 	splx(s);
1214 }
1215 
1216 void
1217 rt_init(void)
1218 {
1219 
1220 	route_intrq.ifq_maxlen = route_maxqlen;
1221 	route_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
1222 	    route_intr, NULL);
1223 }
1224 
1225 /*
1226  * Definitions of protocols supported in the ROUTE domain.
1227  */
1228 PR_WRAP_USRREQ(route_usrreq)
1229 #define	route_usrreq	route_usrreq_wrapper
1230 
1231 const struct protosw routesw[] = {
1232 	{
1233 		.pr_type = SOCK_RAW,
1234 		.pr_domain = &routedomain,
1235 		.pr_flags = PR_ATOMIC|PR_ADDR,
1236 		.pr_input = raw_input,
1237 		.pr_output = route_output,
1238 		.pr_ctlinput = raw_ctlinput,
1239 		.pr_usrreq = route_usrreq,
1240 		.pr_init = raw_init,
1241 	},
1242 };
1243 
1244 struct domain routedomain = {
1245 	.dom_family = PF_ROUTE,
1246 	.dom_name = "route",
1247 	.dom_init = route_init,
1248 	.dom_protosw = routesw,
1249 	.dom_protoswNPROTOSW = &routesw[__arraycount(routesw)],
1250 };
1251 
1252 SYSCTL_SETUP(sysctl_net_route_setup, "sysctl net.route subtree setup")
1253 {
1254 	const struct sysctlnode *rnode = NULL;
1255 
1256 	sysctl_createv(clog, 0, NULL, NULL,
1257 		       CTLFLAG_PERMANENT,
1258 		       CTLTYPE_NODE, "net", NULL,
1259 		       NULL, 0, NULL, 0,
1260 		       CTL_NET, CTL_EOL);
1261 
1262 	sysctl_createv(clog, 0, NULL, &rnode,
1263 		       CTLFLAG_PERMANENT,
1264 		       CTLTYPE_NODE, "route",
1265 		       SYSCTL_DESCR("PF_ROUTE information"),
1266 		       NULL, 0, NULL, 0,
1267 		       CTL_NET, PF_ROUTE, CTL_EOL);
1268 	sysctl_createv(clog, 0, NULL, NULL,
1269 		       CTLFLAG_PERMANENT,
1270 		       CTLTYPE_NODE, "rtable",
1271 		       SYSCTL_DESCR("Routing table information"),
1272 		       sysctl_rtable, 0, NULL, 0,
1273 		       CTL_NET, PF_ROUTE, 0 /* any protocol */, CTL_EOL);
1274 	sysctl_createv(clog, 0, &rnode, NULL,
1275 		       CTLFLAG_PERMANENT,
1276 		       CTLTYPE_STRUCT, "stats",
1277 		       SYSCTL_DESCR("Routing statistics"),
1278 		       NULL, 0, &rtstat, sizeof(rtstat),
1279 		       CTL_CREATE, CTL_EOL);
1280 }
1281