xref: /netbsd-src/sys/net/rtsock.c (revision 46f5119e40af2e51998f686b2fdcc76b5488f7f3)
1 /*	$NetBSD: rtsock.c,v 1.135 2011/03/31 19:40:52 dyoung Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1988, 1991, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
61  */
62 
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: rtsock.c,v 1.135 2011/03/31 19:40:52 dyoung Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_mpls.h"
69 #include "opt_compat_netbsd.h"
70 #endif
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/proc.h>
75 #include <sys/mbuf.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/domain.h>
79 #include <sys/protosw.h>
80 #include <sys/sysctl.h>
81 #include <sys/kauth.h>
82 #include <sys/intr.h>
83 #ifdef RTSOCK_DEBUG
84 #include <netinet/in.h>
85 #endif /* RTSOCK_DEBUG */
86 
87 #include <net/if.h>
88 #include <net/route.h>
89 #include <net/raw_cb.h>
90 
91 #include <netmpls/mpls.h>
92 
93 #if defined(COMPAT_14) || defined(COMPAT_50)
94 #include <compat/net/if.h>
95 #include <compat/net/route.h>
96 #endif
97 #ifdef COMPAT_RTSOCK
98 #define	RTM_XVERSION	RTM_OVERSION
99 #define	RT_XADVANCE(a,b) RT_OADVANCE(a,b)
100 #define	RT_XROUNDUP(n)	RT_OROUNDUP(n)
101 #define	PF_XROUTE	PF_OROUTE
102 #define	rt_xmsghdr	rt_msghdr50
103 #define	if_xmsghdr	if_msghdr	/* if_msghdr50 is for RTM_OIFINFO */
104 #define	ifa_xmsghdr	ifa_msghdr50
105 #define	if_xannouncemsghdr	if_announcemsghdr50
106 #define	COMPATNAME(x)	compat_50_ ## x
107 #define	DOMAINNAME	"oroute"
108 CTASSERT(sizeof(struct ifa_xmsghdr) == 20);
109 DOMAIN_DEFINE(compat_50_routedomain); /* forward declare and add to link set */
110 #else
111 #define	RTM_XVERSION	RTM_VERSION
112 #define	RT_XADVANCE(a,b) RT_ADVANCE(a,b)
113 #define	RT_XROUNDUP(n)	RT_ROUNDUP(n)
114 #define	PF_XROUTE	PF_ROUTE
115 #define	rt_xmsghdr	rt_msghdr
116 #define	if_xmsghdr	if_msghdr
117 #define	ifa_xmsghdr	ifa_msghdr
118 #define	if_xannouncemsghdr	if_announcemsghdr
119 #define	COMPATNAME(x)	x
120 #define	DOMAINNAME	"route"
121 CTASSERT(sizeof(struct ifa_xmsghdr) == 24);
122 #ifdef COMPAT_50
123 #define	COMPATCALL(name, args)	compat_50_ ## name args
124 #endif
125 DOMAIN_DEFINE(routedomain); /* forward declare and add to link set */
126 #undef COMPAT_50
127 #undef COMPAT_14
128 #endif
129 
130 #ifndef COMPATCALL
131 #define	COMPATCALL(name, args)	do { } while (/*CONSTCOND*/ 0)
132 #endif
133 
134 #include <machine/stdarg.h>
135 
136 struct route_info COMPATNAME(route_info) = {
137 	.ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, },
138 	.ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, },
139 	.ri_maxqlen = IFQ_MAXLEN,
140 };
141 
142 #define	PRESERVED_RTF	(RTF_UP | RTF_GATEWAY | RTF_HOST | RTF_DONE | RTF_MASK)
143 
144 static void COMPATNAME(route_init)(void);
145 static int COMPATNAME(route_output)(struct mbuf *, ...);
146 static int COMPATNAME(route_usrreq)(struct socket *,
147 	    int, struct mbuf *, struct mbuf *, struct mbuf *, struct lwp *);
148 
149 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
150 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
151 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
152     struct rt_addrinfo *);
153 static void rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *);
154 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *);
155 static void sysctl_net_route_setup(struct sysctllog **);
156 static int sysctl_dumpentry(struct rtentry *, void *);
157 static int sysctl_iflist(int, struct rt_walkarg *, int);
158 static int sysctl_rtable(SYSCTLFN_PROTO);
159 static void rt_adjustcount(int, int);
160 
161 static void
162 rt_adjustcount(int af, int cnt)
163 {
164 	struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
165 
166 	cb->any_count += cnt;
167 
168 	switch (af) {
169 	case AF_INET:
170 		cb->ip_count += cnt;
171 		return;
172 #ifdef INET6
173 	case AF_INET6:
174 		cb->ip6_count += cnt;
175 		return;
176 #endif
177 	case AF_ISO:
178 		cb->iso_count += cnt;
179 		return;
180 	case AF_MPLS:
181 		cb->mpls_count += cnt;
182 		return;
183 	}
184 }
185 
186 /*ARGSUSED*/
187 int
188 COMPATNAME(route_usrreq)(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
189 	struct mbuf *control, struct lwp *l)
190 {
191 	int error = 0;
192 	struct rawcb *rp = sotorawcb(so);
193 	int s;
194 
195 	if (req == PRU_ATTACH) {
196 		sosetlock(so);
197 		rp = malloc(sizeof(*rp), M_PCB, M_WAITOK|M_ZERO);
198 		so->so_pcb = rp;
199 	}
200 	if (req == PRU_DETACH && rp)
201 		rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
202 	s = splsoftnet();
203 
204 	/*
205 	 * Don't call raw_usrreq() in the attach case, because
206 	 * we want to allow non-privileged processes to listen on
207 	 * and send "safe" commands to the routing socket.
208 	 */
209 	if (req == PRU_ATTACH) {
210 		if (l == NULL)
211 			error = EACCES;
212 		else
213 			error = raw_attach(so, (int)(long)nam);
214 	} else
215 		error = raw_usrreq(so, req, m, nam, control, l);
216 
217 	rp = sotorawcb(so);
218 	if (req == PRU_ATTACH && rp) {
219 		if (error) {
220 			free(rp, M_PCB);
221 			splx(s);
222 			return error;
223 		}
224 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
225 		rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
226 		rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
227 		soisconnected(so);
228 		so->so_options |= SO_USELOOPBACK;
229 	}
230 	splx(s);
231 	return error;
232 }
233 
234 /*ARGSUSED*/
235 int
236 COMPATNAME(route_output)(struct mbuf *m, ...)
237 {
238 	struct sockproto proto = { .sp_family = PF_XROUTE, };
239 	struct rt_xmsghdr *rtm = NULL;
240 	struct rt_xmsghdr *old_rtm = NULL;
241 	struct rtentry *rt = NULL;
242 	struct rtentry *saved_nrt = NULL;
243 	struct rt_addrinfo info;
244 	int len, error = 0;
245 	struct ifnet *ifp = NULL;
246 	struct ifaddr *ifa = NULL;
247 	struct socket *so;
248 	va_list ap;
249 	sa_family_t family;
250 
251 	va_start(ap, m);
252 	so = va_arg(ap, struct socket *);
253 	va_end(ap);
254 
255 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
256 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
257 	   (m = m_pullup(m, sizeof(int32_t))) == NULL))
258 		return ENOBUFS;
259 	if ((m->m_flags & M_PKTHDR) == 0)
260 		panic("%s", __func__);
261 	len = m->m_pkthdr.len;
262 	if (len < sizeof(*rtm) ||
263 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
264 		info.rti_info[RTAX_DST] = NULL;
265 		senderr(EINVAL);
266 	}
267 	R_Malloc(rtm, struct rt_xmsghdr *, len);
268 	if (rtm == NULL) {
269 		info.rti_info[RTAX_DST] = NULL;
270 		senderr(ENOBUFS);
271 	}
272 	m_copydata(m, 0, len, rtm);
273 	if (rtm->rtm_version != RTM_XVERSION) {
274 		info.rti_info[RTAX_DST] = NULL;
275 		senderr(EPROTONOSUPPORT);
276 	}
277 	rtm->rtm_pid = curproc->p_pid;
278 	memset(&info, 0, sizeof(info));
279 	info.rti_addrs = rtm->rtm_addrs;
280 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
281 	    &info)) {
282 		senderr(EINVAL);
283 	}
284 	info.rti_flags = rtm->rtm_flags;
285 #ifdef RTSOCK_DEBUG
286 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
287 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
288 		    inet_ntoa(((const struct sockaddr_in *)
289 		    info.rti_info[RTAX_DST])->sin_addr));
290 	}
291 #endif /* RTSOCK_DEBUG */
292 	if (info.rti_info[RTAX_DST] == NULL ||
293 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
294 		senderr(EINVAL);
295 	}
296 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
297 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
298 		senderr(EINVAL);
299 	}
300 
301 	/*
302 	 * Verify that the caller has the appropriate privilege; RTM_GET
303 	 * is the only operation the non-superuser is allowed.
304 	 */
305 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
306 	    0, rtm, NULL, NULL) != 0)
307 		senderr(EACCES);
308 
309 	switch (rtm->rtm_type) {
310 
311 	case RTM_ADD:
312 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
313 			senderr(EINVAL);
314 		}
315 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
316 		if (error == 0 && saved_nrt) {
317 			rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
318 			saved_nrt->rt_refcnt--;
319 		}
320 		break;
321 
322 	case RTM_DELETE:
323 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
324 		if (error == 0) {
325 			(rt = saved_nrt)->rt_refcnt++;
326 			goto report;
327 		}
328 		break;
329 
330 	case RTM_GET:
331 	case RTM_CHANGE:
332 	case RTM_LOCK:
333                 /* XXX This will mask info.rti_info[RTAX_DST] with
334 		 * info.rti_info[RTAX_NETMASK] before
335                  * searching.  It did not used to do that.  --dyoung
336 		 */
337 		error = rtrequest1(RTM_GET, &info, &rt);
338 		if (error != 0)
339 			senderr(error);
340 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
341 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
342 			    info.rti_info[RTAX_DST]->sa_len) != 0)
343 				senderr(ESRCH);
344 			if (info.rti_info[RTAX_NETMASK] == NULL &&
345 			    rt_mask(rt) != NULL)
346 				senderr(ETOOMANYREFS);
347 		}
348 
349 		switch (rtm->rtm_type) {
350 		case RTM_GET:
351 		report:
352 			info.rti_info[RTAX_DST] = rt_getkey(rt);
353 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
354 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
355 			info.rti_info[RTAX_TAG] = (struct sockaddr*)rt_gettag(rt);
356 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
357 				;
358 			else if ((ifp = rt->rt_ifp) != NULL) {
359 				const struct ifaddr *rtifa;
360 				info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
361                                 /* rtifa used to be simply rt->rt_ifa.
362                                  * If rt->rt_ifa != NULL, then
363                                  * rt_get_ifa() != NULL.  So this
364                                  * ought to still be safe. --dyoung
365 				 */
366 				rtifa = rt_get_ifa(rt);
367 				info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
368 #ifdef RTSOCK_DEBUG
369 				if (info.rti_info[RTAX_IFA]->sa_family ==
370 				    AF_INET) {
371 					printf("%s: copying out RTAX_IFA %s ",
372 					    __func__, inet_ntoa(
373 					    ((const struct sockaddr_in *)
374 					    info.rti_info[RTAX_IFA])->sin_addr)
375 					    );
376 					printf("for info.rti_info[RTAX_DST] %s "
377 					    "ifa_getifa %p ifa_seqno %p\n",
378 					    inet_ntoa(
379 					    ((const struct sockaddr_in *)
380 					    info.rti_info[RTAX_DST])->sin_addr),
381 					    (void *)rtifa->ifa_getifa,
382 					    rtifa->ifa_seqno);
383 				}
384 #endif /* RTSOCK_DEBUG */
385 				if (ifp->if_flags & IFF_POINTOPOINT) {
386 					info.rti_info[RTAX_BRD] =
387 					    rtifa->ifa_dstaddr;
388 				} else
389 					info.rti_info[RTAX_BRD] = NULL;
390 				rtm->rtm_index = ifp->if_index;
391 			} else {
392 				info.rti_info[RTAX_IFP] = NULL;
393 				info.rti_info[RTAX_IFA] = NULL;
394 			}
395 			(void)rt_msg2(rtm->rtm_type, &info, NULL, NULL, &len);
396 			if (len > rtm->rtm_msglen) {
397 				old_rtm = rtm;
398 				R_Malloc(rtm, struct rt_xmsghdr *, len);
399 				if (rtm == NULL)
400 					senderr(ENOBUFS);
401 				(void)memcpy(rtm, old_rtm, old_rtm->rtm_msglen);
402 			}
403 			(void)rt_msg2(rtm->rtm_type, &info, rtm, NULL, 0);
404 			rtm->rtm_flags = rt->rt_flags;
405 			rtm_setmetrics(rt, rtm);
406 			rtm->rtm_addrs = info.rti_addrs;
407 			break;
408 
409 		case RTM_CHANGE:
410 			/*
411 			 * new gateway could require new ifaddr, ifp;
412 			 * flags may also be different; ifp may be specified
413 			 * by ll sockaddr when protocol address is ambiguous
414 			 */
415 			if ((error = rt_getifa(&info)) != 0)
416 				senderr(error);
417 			if (info.rti_info[RTAX_GATEWAY] &&
418 			    rt_setgate(rt, info.rti_info[RTAX_GATEWAY]))
419 				senderr(EDQUOT);
420 			if (info.rti_info[RTAX_TAG])
421 				rt_settag(rt, info.rti_info[RTAX_TAG]);
422 			/* new gateway could require new ifaddr, ifp;
423 			   flags may also be different; ifp may be specified
424 			   by ll sockaddr when protocol address is ambiguous */
425 			if (info.rti_info[RTAX_IFP] &&
426 			    (ifa = ifa_ifwithnet(info.rti_info[RTAX_IFP])) &&
427 			    (ifp = ifa->ifa_ifp) && (info.rti_info[RTAX_IFA] ||
428 			    info.rti_info[RTAX_GATEWAY])) {
429 				if (info.rti_info[RTAX_IFA] == NULL ||
430 				    (ifa = ifa_ifwithaddr(
431 				    info.rti_info[RTAX_IFA])) == NULL)
432 					ifa = ifaof_ifpforaddr(
433 					    info.rti_info[RTAX_IFA] ?
434 					    info.rti_info[RTAX_IFA] :
435 					    info.rti_info[RTAX_GATEWAY], ifp);
436 			} else if ((info.rti_info[RTAX_IFA] &&
437 			    (ifa = ifa_ifwithaddr(info.rti_info[RTAX_IFA]))) ||
438 			    (info.rti_info[RTAX_GATEWAY] &&
439 			    (ifa = ifa_ifwithroute(rt->rt_flags,
440 			    rt_getkey(rt), info.rti_info[RTAX_GATEWAY])))) {
441 				ifp = ifa->ifa_ifp;
442 			}
443 			if (ifa) {
444 				struct ifaddr *oifa = rt->rt_ifa;
445 				if (oifa != ifa) {
446 					if (oifa && oifa->ifa_rtrequest) {
447 						oifa->ifa_rtrequest(RTM_DELETE,
448 						    rt, &info);
449 					}
450 					rt_replace_ifa(rt, ifa);
451 					rt->rt_ifp = ifp;
452 				}
453 			}
454 			if (ifp && rt->rt_ifp != ifp)
455 				rt->rt_ifp = ifp;
456 			rt_setmetrics(rtm->rtm_inits, rtm, rt);
457 			if (rt->rt_flags != info.rti_flags)
458 				rt->rt_flags = (info.rti_flags & ~PRESERVED_RTF)
459 				    | (rt->rt_flags & PRESERVED_RTF);
460 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
461 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
462 			/*FALLTHROUGH*/
463 		case RTM_LOCK:
464 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
465 			rt->rt_rmx.rmx_locks |=
466 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
467 			break;
468 		}
469 		break;
470 
471 	default:
472 		senderr(EOPNOTSUPP);
473 	}
474 
475 flush:
476 	if (rtm) {
477 		if (error)
478 			rtm->rtm_errno = error;
479 		else
480 			rtm->rtm_flags |= RTF_DONE;
481 	}
482 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
483 	    0;
484 	/* We cannot free old_rtm until we have stopped using the
485 	 * pointers in info, some of which may point to sockaddrs
486 	 * in old_rtm.
487 	 */
488 	if (old_rtm != NULL)
489 		Free(old_rtm);
490 	if (rt)
491 		rtfree(rt);
492     {
493 	struct rawcb *rp = NULL;
494 	/*
495 	 * Check to see if we don't want our own messages.
496 	 */
497 	if ((so->so_options & SO_USELOOPBACK) == 0) {
498 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
499 			if (rtm)
500 				Free(rtm);
501 			m_freem(m);
502 			return error;
503 		}
504 		/* There is another listener, so construct message */
505 		rp = sotorawcb(so);
506 	}
507 	if (rtm) {
508 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
509 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
510 			m_freem(m);
511 			m = NULL;
512 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
513 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
514 		Free(rtm);
515 	}
516 	if (rp)
517 		rp->rcb_proto.sp_family = 0; /* Avoid us */
518 	if (family)
519 		proto.sp_protocol = family;
520 	if (m)
521 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
522 		    &COMPATNAME(route_info).ri_dst);
523 	if (rp)
524 		rp->rcb_proto.sp_family = PF_XROUTE;
525     }
526 	return error;
527 }
528 
529 static void
530 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
531 {
532 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
533 	metric(RTV_RPIPE, rmx_recvpipe);
534 	metric(RTV_SPIPE, rmx_sendpipe);
535 	metric(RTV_SSTHRESH, rmx_ssthresh);
536 	metric(RTV_RTT, rmx_rtt);
537 	metric(RTV_RTTVAR, rmx_rttvar);
538 	metric(RTV_HOPCOUNT, rmx_hopcount);
539 	metric(RTV_MTU, rmx_mtu);
540 	metric(RTV_EXPIRE, rmx_expire);
541 #undef metric
542 }
543 
544 static void
545 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
546 {
547 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
548 	metric(rmx_recvpipe);
549 	metric(rmx_sendpipe);
550 	metric(rmx_ssthresh);
551 	metric(rmx_rtt);
552 	metric(rmx_rttvar);
553 	metric(rmx_hopcount);
554 	metric(rmx_mtu);
555 	metric(rmx_expire);
556 #undef metric
557 }
558 
559 static int
560 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
561     struct rt_addrinfo *rtinfo)
562 {
563 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
564 	int i;
565 
566 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
567 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
568 			continue;
569 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
570 		RT_XADVANCE(cp, sa);
571 	}
572 
573 	/*
574 	 * Check for extra addresses specified, except RTM_GET asking
575 	 * for interface info.
576 	 */
577 	if (rtmtype == RTM_GET) {
578 		if (((rtinfo->rti_addrs &
579 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
580 			return 1;
581 	} else if ((rtinfo->rti_addrs & (~0 << i)) != 0)
582 		return 1;
583 	/* Check for bad data length.  */
584 	if (cp != cplim) {
585 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
586 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
587 			/*
588 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
589 			 * We accept this for now for the sake of old
590 			 * binaries or third party softwares.
591 			 */
592 			;
593 		else
594 			return 1;
595 	}
596 	return 0;
597 }
598 
599 static int
600 rt_getlen(int type)
601 {
602 #ifndef COMPAT_RTSOCK
603 	CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
604 	CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
605 	CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
606 	CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
607 #endif
608 
609 	switch (type) {
610 	case RTM_DELADDR:
611 	case RTM_NEWADDR:
612 	case RTM_CHGADDR:
613 		return sizeof(struct ifa_xmsghdr);
614 
615 	case RTM_OOIFINFO:
616 #ifdef COMPAT_14
617 		return sizeof(struct if_msghdr14);
618 #else
619 #ifdef DIAGNOSTIC
620 		printf("RTM_OOIFINFO\n");
621 #endif
622 		return -1;
623 #endif
624 	case RTM_OIFINFO:
625 #ifdef COMPAT_50
626 		return sizeof(struct if_msghdr50);
627 #else
628 #ifdef DIAGNOSTIC
629 		printf("RTM_OIFINFO\n");
630 #endif
631 		return -1;
632 #endif
633 
634 	case RTM_IFINFO:
635 		return sizeof(struct if_xmsghdr);
636 
637 	case RTM_IFANNOUNCE:
638 	case RTM_IEEE80211:
639 		return sizeof(struct if_xannouncemsghdr);
640 
641 	default:
642 		return sizeof(struct rt_xmsghdr);
643 	}
644 }
645 
646 
647 struct mbuf *
648 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
649 {
650 	struct rt_xmsghdr *rtm;
651 	struct mbuf *m;
652 	int i;
653 	const struct sockaddr *sa;
654 	int len, dlen;
655 
656 	m = m_gethdr(M_DONTWAIT, MT_DATA);
657 	if (m == NULL)
658 		return m;
659 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
660 
661 	if ((len = rt_getlen(type)) == -1)
662 		goto out;
663 	if (len > MHLEN + MLEN)
664 		panic("%s: message too long", __func__);
665 	else if (len > MHLEN) {
666 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
667 		if (m->m_next == NULL)
668 			goto out;
669 		MCLAIM(m->m_next, m->m_owner);
670 		m->m_pkthdr.len = len;
671 		m->m_len = MHLEN;
672 		m->m_next->m_len = len - MHLEN;
673 	} else {
674 		m->m_pkthdr.len = m->m_len = len;
675 	}
676 	m->m_pkthdr.rcvif = NULL;
677 	m_copyback(m, 0, datalen, data);
678 	if (len > datalen)
679 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
680 	rtm = mtod(m, struct rt_xmsghdr *);
681 	for (i = 0; i < RTAX_MAX; i++) {
682 		if ((sa = rtinfo->rti_info[i]) == NULL)
683 			continue;
684 		rtinfo->rti_addrs |= (1 << i);
685 		dlen = RT_XROUNDUP(sa->sa_len);
686 		m_copyback(m, len, sa->sa_len, sa);
687 		if (dlen != sa->sa_len) {
688 			m_copyback(m, len + sa->sa_len,
689 			    dlen - sa->sa_len, "\0\0\0\0\0\0");
690 		}
691 		len += dlen;
692 	}
693 	if (m->m_pkthdr.len != len)
694 		goto out;
695 	rtm->rtm_msglen = len;
696 	rtm->rtm_version = RTM_XVERSION;
697 	rtm->rtm_type = type;
698 	return m;
699 out:
700 	m_freem(m);
701 	return NULL;
702 }
703 
704 /*
705  * rt_msg2
706  *
707  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
708  *		returns the length of the message in 'lenp'.
709  *
710  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
711  *	the message
712  * otherwise walkarg's w_needed is updated and if the user buffer is
713  *	specified and w_needed indicates space exists the information is copied
714  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
715  *	if the allocation fails ENOBUFS is returned.
716  */
717 static int
718 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
719 	int *lenp)
720 {
721 	int i;
722 	int len, dlen, second_time = 0;
723 	char *cp0, *cp = cpv;
724 
725 	rtinfo->rti_addrs = 0;
726 again:
727 	if ((len = rt_getlen(type)) == -1)
728 		return EINVAL;
729 
730 	if ((cp0 = cp) != NULL)
731 		cp += len;
732 	for (i = 0; i < RTAX_MAX; i++) {
733 		const struct sockaddr *sa;
734 
735 		if ((sa = rtinfo->rti_info[i]) == NULL)
736 			continue;
737 		rtinfo->rti_addrs |= (1 << i);
738 		dlen = RT_XROUNDUP(sa->sa_len);
739 		if (cp) {
740 			(void)memcpy(cp, sa, (size_t)dlen);
741 			cp += dlen;
742 		}
743 		len += dlen;
744 	}
745 	if (cp == NULL && w != NULL && !second_time) {
746 		struct rt_walkarg *rw = w;
747 
748 		rw->w_needed += len;
749 		if (rw->w_needed <= 0 && rw->w_where) {
750 			if (rw->w_tmemsize < len) {
751 				if (rw->w_tmem)
752 					free(rw->w_tmem, M_RTABLE);
753 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
754 				if (rw->w_tmem)
755 					rw->w_tmemsize = len;
756 				else
757 					rw->w_tmemsize = 0;
758 			}
759 			if (rw->w_tmem) {
760 				cp = rw->w_tmem;
761 				second_time = 1;
762 				goto again;
763 			} else {
764 				rw->w_tmemneeded = len;
765 				return ENOBUFS;
766 			}
767 		}
768 	}
769 	if (cp) {
770 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
771 
772 		rtm->rtm_version = RTM_XVERSION;
773 		rtm->rtm_type = type;
774 		rtm->rtm_msglen = len;
775 	}
776 	if (lenp)
777 		*lenp = len;
778 	return 0;
779 }
780 
781 /*
782  * This routine is called to generate a message from the routing
783  * socket indicating that a redirect has occurred, a routing lookup
784  * has failed, or that a protocol has detected timeouts to a particular
785  * destination.
786  */
787 void
788 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
789     int error)
790 {
791 	struct rt_xmsghdr rtm;
792 	struct mbuf *m;
793 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
794 	struct rt_addrinfo info = *rtinfo;
795 
796 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
797 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
798 		return;
799 	memset(&rtm, 0, sizeof(rtm));
800 	rtm.rtm_flags = RTF_DONE | flags;
801 	rtm.rtm_errno = error;
802 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
803 	if (m == NULL)
804 		return;
805 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
806 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
807 }
808 
809 /*
810  * This routine is called to generate a message from the routing
811  * socket indicating that the status of a network interface has changed.
812  */
813 void
814 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
815 {
816 	struct if_xmsghdr ifm;
817 	struct mbuf *m;
818 	struct rt_addrinfo info;
819 
820 	COMPATCALL(rt_ifmsg, (ifp));
821 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
822 		return;
823 	(void)memset(&info, 0, sizeof(info));
824 	(void)memset(&ifm, 0, sizeof(ifm));
825 	ifm.ifm_index = ifp->if_index;
826 	ifm.ifm_flags = ifp->if_flags;
827 	ifm.ifm_data = ifp->if_data;
828 	ifm.ifm_addrs = 0;
829 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
830 	if (m == NULL)
831 		return;
832 	COMPATNAME(route_enqueue)(m, 0);
833 #ifdef COMPAT_14
834 	compat_14_rt_oifmsg(ifp);
835 #endif
836 #ifdef COMPAT_50
837 	compat_50_rt_oifmsg(ifp);
838 #endif
839 }
840 
841 
842 /*
843  * This is called to generate messages from the routing socket
844  * indicating a network interface has had addresses associated with it.
845  * if we ever reverse the logic and replace messages TO the routing
846  * socket indicate a request to configure interfaces, then it will
847  * be unnecessary as the routing socket will automatically generate
848  * copies of it.
849  */
850 void
851 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
852     struct rtentry *rt)
853 {
854 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
855 	struct rt_addrinfo info;
856 	const struct sockaddr *sa;
857 	int pass;
858 	struct mbuf *m;
859 	struct ifnet *ifp = ifa->ifa_ifp;
860 	struct rt_xmsghdr rtm;
861 	struct ifa_xmsghdr ifam;
862 	int ncmd;
863 
864 	COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
865 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
866 		return;
867 	for (pass = 1; pass < 3; pass++) {
868 		memset(&info, 0, sizeof(info));
869 		switch (cmdpass(cmd, pass)) {
870 		case cmdpass(RTM_ADD, 1):
871 		case cmdpass(RTM_CHANGE, 1):
872 		case cmdpass(RTM_DELETE, 2):
873 			switch (cmd) {
874 			case RTM_DELETE:
875 				ncmd = RTM_DELADDR;
876 				break;
877 			case RTM_CHANGE:
878 				ncmd = RTM_CHGADDR;
879 				break;
880 			default:
881 				ncmd = RTM_NEWADDR;
882 			}
883 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
884 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
885 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
886 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
887 			memset(&ifam, 0, sizeof(ifam));
888 			ifam.ifam_index = ifp->if_index;
889 			ifam.ifam_metric = ifa->ifa_metric;
890 			ifam.ifam_flags = ifa->ifa_flags;
891 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
892 			if (m == NULL)
893 				continue;
894 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
895 			    info.rti_addrs;
896 			break;
897 		case cmdpass(RTM_ADD, 2):
898 		case cmdpass(RTM_CHANGE, 2):
899 		case cmdpass(RTM_DELETE, 1):
900 			if (rt == NULL)
901 				continue;
902 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
903 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
904 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
905 			memset(&rtm, 0, sizeof(rtm));
906 			rtm.rtm_index = ifp->if_index;
907 			rtm.rtm_flags |= rt->rt_flags;
908 			rtm.rtm_errno = error;
909 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
910 			if (m == NULL)
911 				continue;
912 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
913 			break;
914 		default:
915 			continue;
916 		}
917 #ifdef DIAGNOSTIC
918 		if (m == NULL)
919 			panic("%s: called with wrong command", __func__);
920 #endif
921 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
922 	}
923 #undef cmdpass
924 }
925 
926 static struct mbuf *
927 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
928     struct rt_addrinfo *info)
929 {
930 	struct if_xannouncemsghdr ifan;
931 
932 	memset(info, 0, sizeof(*info));
933 	memset(&ifan, 0, sizeof(ifan));
934 	ifan.ifan_index = ifp->if_index;
935 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
936 	ifan.ifan_what = what;
937 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
938 }
939 
940 /*
941  * This is called to generate routing socket messages indicating
942  * network interface arrival and departure.
943  */
944 void
945 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
946 {
947 	struct mbuf *m;
948 	struct rt_addrinfo info;
949 
950 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
951 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
952 		return;
953 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
954 	if (m == NULL)
955 		return;
956 	COMPATNAME(route_enqueue)(m, 0);
957 }
958 
959 /*
960  * This is called to generate routing socket messages indicating
961  * IEEE80211 wireless events.
962  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
963  */
964 void
965 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
966 	size_t data_len)
967 {
968 	struct mbuf *m;
969 	struct rt_addrinfo info;
970 
971 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
972 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
973 		return;
974 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
975 	if (m == NULL)
976 		return;
977 	/*
978 	 * Append the ieee80211 data.  Try to stick it in the
979 	 * mbuf containing the ifannounce msg; otherwise allocate
980 	 * a new mbuf and append.
981 	 *
982 	 * NB: we assume m is a single mbuf.
983 	 */
984 	if (data_len > M_TRAILINGSPACE(m)) {
985 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
986 		if (n == NULL) {
987 			m_freem(m);
988 			return;
989 		}
990 		(void)memcpy(mtod(n, void *), data, data_len);
991 		n->m_len = data_len;
992 		m->m_next = n;
993 	} else if (data_len > 0) {
994 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
995 		m->m_len += data_len;
996 	}
997 	if (m->m_flags & M_PKTHDR)
998 		m->m_pkthdr.len += data_len;
999 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
1000 	COMPATNAME(route_enqueue)(m, 0);
1001 }
1002 
1003 /*
1004  * This is used in dumping the kernel table via sysctl().
1005  */
1006 static int
1007 sysctl_dumpentry(struct rtentry *rt, void *v)
1008 {
1009 	struct rt_walkarg *w = v;
1010 	int error = 0, size;
1011 	struct rt_addrinfo info;
1012 
1013 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1014 		return 0;
1015 	memset(&info, 0, sizeof(info));
1016 	info.rti_info[RTAX_DST] = rt_getkey(rt);
1017 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1018 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1019 	if (rt->rt_ifp) {
1020 		const struct ifaddr *rtifa;
1021 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
1022 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
1023 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
1024 		 * --dyoung
1025 		 */
1026 		rtifa = rt_get_ifa(rt);
1027 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
1028 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1029 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
1030 	}
1031 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
1032 		return error;
1033 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1034 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
1035 
1036 		rtm->rtm_flags = rt->rt_flags;
1037 		rtm->rtm_use = rt->rt_use;
1038 		rtm_setmetrics(rt, rtm);
1039 		KASSERT(rt->rt_ifp != NULL);
1040 		rtm->rtm_index = rt->rt_ifp->if_index;
1041 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1042 		rtm->rtm_addrs = info.rti_addrs;
1043 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1044 			w->w_where = NULL;
1045 		else
1046 			w->w_where = (char *)w->w_where + size;
1047 	}
1048 	return error;
1049 }
1050 
1051 static int
1052 sysctl_iflist(int af, struct rt_walkarg *w, int type)
1053 {
1054 	struct ifnet *ifp;
1055 	struct ifaddr *ifa;
1056 	struct	rt_addrinfo info;
1057 	int	len, error = 0;
1058 
1059 	memset(&info, 0, sizeof(info));
1060 	IFNET_FOREACH(ifp) {
1061 		if (w->w_arg && w->w_arg != ifp->if_index)
1062 			continue;
1063 		if (IFADDR_EMPTY(ifp))
1064 			continue;
1065 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1066 		switch (type) {
1067 		case NET_RT_IFLIST:
1068 			error = rt_msg2(RTM_IFINFO, &info, NULL, w, &len);
1069 			break;
1070 #ifdef COMPAT_14
1071 		case NET_RT_OOIFLIST:
1072 			error = rt_msg2(RTM_OOIFINFO, &info, NULL, w, &len);
1073 			break;
1074 #endif
1075 #ifdef COMPAT_50
1076 		case NET_RT_OIFLIST:
1077 			error = rt_msg2(RTM_OIFINFO, &info, NULL, w, &len);
1078 			break;
1079 #endif
1080 		default:
1081 			panic("sysctl_iflist(1)");
1082 		}
1083 		if (error)
1084 			return error;
1085 		info.rti_info[RTAX_IFP] = NULL;
1086 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1087 			switch (type) {
1088 			case NET_RT_IFLIST: {
1089 				struct if_xmsghdr *ifm;
1090 
1091 				ifm = (struct if_xmsghdr *)w->w_tmem;
1092 				ifm->ifm_index = ifp->if_index;
1093 				ifm->ifm_flags = ifp->if_flags;
1094 				ifm->ifm_data = ifp->if_data;
1095 				ifm->ifm_addrs = info.rti_addrs;
1096 				error = copyout(ifm, w->w_where, len);
1097 				if (error)
1098 					return error;
1099 				w->w_where = (char *)w->w_where + len;
1100 				break;
1101 			}
1102 
1103 #ifdef COMPAT_14
1104 			case NET_RT_OOIFLIST:
1105 				error = compat_14_iflist(ifp, w, &info, len);
1106 				if (error)
1107 					return error;
1108 				break;
1109 #endif
1110 #ifdef COMPAT_50
1111 			case NET_RT_OIFLIST:
1112 				error = compat_50_iflist(ifp, w, &info, len);
1113 				if (error)
1114 					return error;
1115 				break;
1116 #endif
1117 			default:
1118 				panic("sysctl_iflist(2)");
1119 			}
1120 		}
1121 		IFADDR_FOREACH(ifa, ifp) {
1122 			if (af && af != ifa->ifa_addr->sa_family)
1123 				continue;
1124 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1125 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1126 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1127 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
1128 				return error;
1129 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1130 				struct ifa_xmsghdr *ifam;
1131 
1132 				ifam = (struct ifa_xmsghdr *)w->w_tmem;
1133 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1134 				ifam->ifam_flags = ifa->ifa_flags;
1135 				ifam->ifam_metric = ifa->ifa_metric;
1136 				ifam->ifam_addrs = info.rti_addrs;
1137 				error = copyout(w->w_tmem, w->w_where, len);
1138 				if (error)
1139 					return error;
1140 				w->w_where = (char *)w->w_where + len;
1141 			}
1142 		}
1143 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1144 		    info.rti_info[RTAX_BRD] = NULL;
1145 	}
1146 	return 0;
1147 }
1148 
1149 static int
1150 sysctl_rtable(SYSCTLFN_ARGS)
1151 {
1152 	void 	*where = oldp;
1153 	size_t	*given = oldlenp;
1154 	const void *new = newp;
1155 	int	i, s, error = EINVAL;
1156 	u_char  af;
1157 	struct	rt_walkarg w;
1158 
1159 	if (namelen == 1 && name[0] == CTL_QUERY)
1160 		return sysctl_query(SYSCTLFN_CALL(rnode));
1161 
1162 	if (new)
1163 		return EPERM;
1164 	if (namelen != 3)
1165 		return EINVAL;
1166 	af = name[0];
1167 	w.w_tmemneeded = 0;
1168 	w.w_tmemsize = 0;
1169 	w.w_tmem = NULL;
1170 again:
1171 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1172 	if (w.w_tmemneeded) {
1173 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
1174 		w.w_tmemsize = w.w_tmemneeded;
1175 		w.w_tmemneeded = 0;
1176 	}
1177 	w.w_op = name[1];
1178 	w.w_arg = name[2];
1179 	w.w_given = *given;
1180 	w.w_needed = 0 - w.w_given;
1181 	w.w_where = where;
1182 
1183 	s = splsoftnet();
1184 	switch (w.w_op) {
1185 
1186 	case NET_RT_DUMP:
1187 	case NET_RT_FLAGS:
1188 		for (i = 1; i <= AF_MAX; i++)
1189 			if ((af == 0 || af == i) &&
1190 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
1191 				break;
1192 		break;
1193 
1194 #ifdef COMPAT_14
1195 	case NET_RT_OOIFLIST:
1196 		error = sysctl_iflist(af, &w, w.w_op);
1197 		break;
1198 #endif
1199 #ifdef COMPAT_50
1200 	case NET_RT_OIFLIST:
1201 		error = sysctl_iflist(af, &w, w.w_op);
1202 		break;
1203 #endif
1204 	case NET_RT_IFLIST:
1205 		error = sysctl_iflist(af, &w, w.w_op);
1206 		break;
1207 	}
1208 	splx(s);
1209 
1210 	/* check to see if we couldn't allocate memory with NOWAIT */
1211 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1212 		goto again;
1213 
1214 	if (w.w_tmem)
1215 		free(w.w_tmem, M_RTABLE);
1216 	w.w_needed += w.w_given;
1217 	if (where) {
1218 		*given = (char *)w.w_where - (char *)where;
1219 		if (*given < w.w_needed)
1220 			return ENOMEM;
1221 	} else {
1222 		*given = (11 * w.w_needed) / 10;
1223 	}
1224 	return error;
1225 }
1226 
1227 /*
1228  * Routing message software interrupt routine
1229  */
1230 static void
1231 COMPATNAME(route_intr)(void *cookie)
1232 {
1233 	struct sockproto proto = { .sp_family = PF_XROUTE, };
1234 	struct route_info * const ri = &COMPATNAME(route_info);
1235 	struct mbuf *m;
1236 	int s;
1237 
1238 	mutex_enter(softnet_lock);
1239 	KERNEL_LOCK(1, NULL);
1240 	while (!IF_IS_EMPTY(&ri->ri_intrq)) {
1241 		s = splnet();
1242 		IF_DEQUEUE(&ri->ri_intrq, m);
1243 		splx(s);
1244 		if (m == NULL)
1245 			break;
1246 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
1247 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst);
1248 	}
1249 	KERNEL_UNLOCK_ONE(NULL);
1250 	mutex_exit(softnet_lock);
1251 }
1252 
1253 /*
1254  * Enqueue a message to the software interrupt routine.
1255  */
1256 void
1257 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
1258 {
1259 	struct route_info * const ri = &COMPATNAME(route_info);
1260 	int s, wasempty;
1261 
1262 	s = splnet();
1263 	if (IF_QFULL(&ri->ri_intrq)) {
1264 		IF_DROP(&ri->ri_intrq);
1265 		m_freem(m);
1266 	} else {
1267 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
1268 		M_SETCTX(m, (uintptr_t)family);
1269 		IF_ENQUEUE(&ri->ri_intrq, m);
1270 		if (wasempty)
1271 			softint_schedule(ri->ri_sih);
1272 	}
1273 	splx(s);
1274 }
1275 
1276 static void
1277 COMPATNAME(route_init)(void)
1278 {
1279 	struct route_info * const ri = &COMPATNAME(route_info);
1280 
1281 #ifndef COMPAT_RTSOCK
1282 	rt_init();
1283 #endif
1284 
1285 	sysctl_net_route_setup(NULL);
1286 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
1287 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
1288 	    COMPATNAME(route_intr), NULL);
1289 }
1290 
1291 /*
1292  * Definitions of protocols supported in the ROUTE domain.
1293  */
1294 #ifndef COMPAT_RTSOCK
1295 PR_WRAP_USRREQ(route_usrreq);
1296 #else
1297 PR_WRAP_USRREQ(compat_50_route_usrreq);
1298 #endif
1299 
1300 static const struct protosw COMPATNAME(route_protosw)[] = {
1301 	{
1302 		.pr_type = SOCK_RAW,
1303 		.pr_domain = &COMPATNAME(routedomain),
1304 		.pr_flags = PR_ATOMIC|PR_ADDR,
1305 		.pr_input = raw_input,
1306 		.pr_output = COMPATNAME(route_output),
1307 		.pr_ctlinput = raw_ctlinput,
1308 		.pr_usrreq = COMPATNAME(route_usrreq_wrapper),
1309 		.pr_init = raw_init,
1310 	},
1311 };
1312 
1313 struct domain COMPATNAME(routedomain) = {
1314 	.dom_family = PF_XROUTE,
1315 	.dom_name = DOMAINNAME,
1316 	.dom_init = COMPATNAME(route_init),
1317 	.dom_protosw = COMPATNAME(route_protosw),
1318 	.dom_protoswNPROTOSW =
1319 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
1320 };
1321 
1322 static void
1323 sysctl_net_route_setup(struct sysctllog **clog)
1324 {
1325 	const struct sysctlnode *rnode = NULL;
1326 
1327 	sysctl_createv(clog, 0, NULL, NULL,
1328 		       CTLFLAG_PERMANENT,
1329 		       CTLTYPE_NODE, "net", NULL,
1330 		       NULL, 0, NULL, 0,
1331 		       CTL_NET, CTL_EOL);
1332 
1333 	sysctl_createv(clog, 0, NULL, &rnode,
1334 		       CTLFLAG_PERMANENT,
1335 		       CTLTYPE_NODE, DOMAINNAME,
1336 		       SYSCTL_DESCR("PF_ROUTE information"),
1337 		       NULL, 0, NULL, 0,
1338 		       CTL_NET, PF_XROUTE, CTL_EOL);
1339 
1340 	sysctl_createv(clog, 0, NULL, NULL,
1341 		       CTLFLAG_PERMANENT,
1342 		       CTLTYPE_NODE, "rtable",
1343 		       SYSCTL_DESCR("Routing table information"),
1344 		       sysctl_rtable, 0, NULL, 0,
1345 		       CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
1346 
1347 	sysctl_createv(clog, 0, &rnode, NULL,
1348 		       CTLFLAG_PERMANENT,
1349 		       CTLTYPE_STRUCT, "stats",
1350 		       SYSCTL_DESCR("Routing statistics"),
1351 		       NULL, 0, &rtstat, sizeof(rtstat),
1352 		       CTL_CREATE, CTL_EOL);
1353 }
1354