xref: /netbsd-src/sys/net/rtsock.c (revision 2c6fc41c810f5088457889d00eba558e8bc74d9e)
1 /*	$NetBSD: rtsock.c,v 1.147 2014/05/21 20:43:56 rmind 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 1.147 2014/05/21 20:43:56 rmind 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/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/kmem.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 struct route_info COMPATNAME(route_info) = {
135 	.ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, },
136 	.ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, },
137 	.ri_maxqlen = IFQ_MAXLEN,
138 };
139 
140 #define	PRESERVED_RTF	(RTF_UP | RTF_GATEWAY | RTF_HOST | RTF_DONE | RTF_MASK)
141 
142 static void COMPATNAME(route_init)(void);
143 static int COMPATNAME(route_output)(struct mbuf *, ...);
144 
145 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
146 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
147 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
148     struct rt_addrinfo *);
149 static void rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *);
150 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *);
151 static void sysctl_net_route_setup(struct sysctllog **);
152 static int sysctl_dumpentry(struct rtentry *, void *);
153 static int sysctl_iflist(int, struct rt_walkarg *, int);
154 static int sysctl_rtable(SYSCTLFN_PROTO);
155 static void rt_adjustcount(int, int);
156 
157 static void
158 rt_adjustcount(int af, int cnt)
159 {
160 	struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
161 
162 	cb->any_count += cnt;
163 
164 	switch (af) {
165 	case AF_INET:
166 		cb->ip_count += cnt;
167 		return;
168 #ifdef INET6
169 	case AF_INET6:
170 		cb->ip6_count += cnt;
171 		return;
172 #endif
173 	case AF_MPLS:
174 		cb->mpls_count += cnt;
175 		return;
176 	}
177 }
178 
179 static int
180 COMPATNAME(route_attach)(struct socket *so, int proto)
181 {
182 	struct rawcb *rp;
183 	int s, error;
184 
185 	KASSERT(sotorawcb(so) == NULL);
186 	rp = kmem_zalloc(sizeof(*rp), KM_SLEEP);
187 	rp->rcb_len = sizeof(*rp);
188 	so->so_pcb = rp;
189 
190 	s = splsoftnet();
191 	if ((error = raw_attach(so, proto)) == 0) {
192 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
193 		rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
194 		rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
195 	}
196 	splx(s);
197 
198 	if (error) {
199 		kmem_free(rp, sizeof(*rp));
200 		so->so_pcb = NULL;
201 		return error;
202 	}
203 
204 	soisconnected(so);
205 	so->so_options |= SO_USELOOPBACK;
206 	KASSERT(solocked(so));
207 
208 	return error;
209 }
210 
211 static void
212 COMPATNAME(route_detach)(struct socket *so)
213 {
214 	struct rawcb *rp = sotorawcb(so);
215 	int s;
216 
217 	KASSERT(rp != NULL);
218 	KASSERT(solocked(so));
219 
220 	s = splsoftnet();
221 	rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
222 	raw_detach(so);
223 	splx(s);
224 }
225 
226 static int
227 COMPATNAME(route_usrreq)(struct socket *so, int req, struct mbuf *m,
228     struct mbuf *nam, struct mbuf *control, struct lwp *l)
229 {
230 	int s, error = 0;
231 
232 	KASSERT(req != PRU_ATTACH);
233 	KASSERT(req != PRU_DETACH);
234 
235 	s = splsoftnet();
236 	error = raw_usrreq(so, req, m, nam, control, l);
237 	splx(s);
238 
239 	return error;
240 }
241 
242 /*ARGSUSED*/
243 int
244 COMPATNAME(route_output)(struct mbuf *m, ...)
245 {
246 	struct sockproto proto = { .sp_family = PF_XROUTE, };
247 	struct rt_xmsghdr *rtm = NULL;
248 	struct rt_xmsghdr *old_rtm = NULL;
249 	struct rtentry *rt = NULL;
250 	struct rtentry *saved_nrt = NULL;
251 	struct rt_addrinfo info;
252 	int len, error = 0;
253 	struct ifnet *ifp = NULL;
254 	struct ifaddr *ifa = NULL;
255 	struct socket *so;
256 	va_list ap;
257 	sa_family_t family;
258 
259 	va_start(ap, m);
260 	so = va_arg(ap, struct socket *);
261 	va_end(ap);
262 
263 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
264 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
265 	   (m = m_pullup(m, sizeof(int32_t))) == NULL))
266 		return ENOBUFS;
267 	if ((m->m_flags & M_PKTHDR) == 0)
268 		panic("%s", __func__);
269 	len = m->m_pkthdr.len;
270 	if (len < sizeof(*rtm) ||
271 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
272 		info.rti_info[RTAX_DST] = NULL;
273 		senderr(EINVAL);
274 	}
275 	R_Malloc(rtm, struct rt_xmsghdr *, len);
276 	if (rtm == NULL) {
277 		info.rti_info[RTAX_DST] = NULL;
278 		senderr(ENOBUFS);
279 	}
280 	m_copydata(m, 0, len, rtm);
281 	if (rtm->rtm_version != RTM_XVERSION) {
282 		info.rti_info[RTAX_DST] = NULL;
283 		senderr(EPROTONOSUPPORT);
284 	}
285 	rtm->rtm_pid = curproc->p_pid;
286 	memset(&info, 0, sizeof(info));
287 	info.rti_addrs = rtm->rtm_addrs;
288 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
289 	    &info)) {
290 		senderr(EINVAL);
291 	}
292 	info.rti_flags = rtm->rtm_flags;
293 #ifdef RTSOCK_DEBUG
294 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
295 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
296 		    inet_ntoa(((const struct sockaddr_in *)
297 		    info.rti_info[RTAX_DST])->sin_addr));
298 	}
299 #endif /* RTSOCK_DEBUG */
300 	if (info.rti_info[RTAX_DST] == NULL ||
301 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
302 		senderr(EINVAL);
303 	}
304 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
305 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
306 		senderr(EINVAL);
307 	}
308 
309 	/*
310 	 * Verify that the caller has the appropriate privilege; RTM_GET
311 	 * is the only operation the non-superuser is allowed.
312 	 */
313 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
314 	    0, rtm, NULL, NULL) != 0)
315 		senderr(EACCES);
316 
317 	switch (rtm->rtm_type) {
318 
319 	case RTM_ADD:
320 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
321 			senderr(EINVAL);
322 		}
323 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
324 		if (error == 0 && saved_nrt) {
325 			rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
326 			saved_nrt->rt_refcnt--;
327 		}
328 		break;
329 
330 	case RTM_DELETE:
331 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
332 		if (error == 0) {
333 			(rt = saved_nrt)->rt_refcnt++;
334 			goto report;
335 		}
336 		break;
337 
338 	case RTM_GET:
339 	case RTM_CHANGE:
340 	case RTM_LOCK:
341                 /* XXX This will mask info.rti_info[RTAX_DST] with
342 		 * info.rti_info[RTAX_NETMASK] before
343                  * searching.  It did not used to do that.  --dyoung
344 		 */
345 		error = rtrequest1(RTM_GET, &info, &rt);
346 		if (error != 0)
347 			senderr(error);
348 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
349 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
350 			    info.rti_info[RTAX_DST]->sa_len) != 0)
351 				senderr(ESRCH);
352 			if (info.rti_info[RTAX_NETMASK] == NULL &&
353 			    rt_mask(rt) != NULL)
354 				senderr(ETOOMANYREFS);
355 		}
356 
357 		switch (rtm->rtm_type) {
358 		case RTM_GET:
359 		report:
360 			info.rti_info[RTAX_DST] = rt_getkey(rt);
361 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
362 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
363 			info.rti_info[RTAX_TAG] = rt_gettag(rt);
364 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
365 				;
366 			else if ((ifp = rt->rt_ifp) != NULL) {
367 				const struct ifaddr *rtifa;
368 				info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
369                                 /* rtifa used to be simply rt->rt_ifa.
370                                  * If rt->rt_ifa != NULL, then
371                                  * rt_get_ifa() != NULL.  So this
372                                  * ought to still be safe. --dyoung
373 				 */
374 				rtifa = rt_get_ifa(rt);
375 				info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
376 #ifdef RTSOCK_DEBUG
377 				if (info.rti_info[RTAX_IFA]->sa_family ==
378 				    AF_INET) {
379 					printf("%s: copying out RTAX_IFA %s ",
380 					    __func__, inet_ntoa(
381 					    ((const struct sockaddr_in *)
382 					    info.rti_info[RTAX_IFA])->sin_addr)
383 					    );
384 					printf("for info.rti_info[RTAX_DST] %s "
385 					    "ifa_getifa %p ifa_seqno %p\n",
386 					    inet_ntoa(
387 					    ((const struct sockaddr_in *)
388 					    info.rti_info[RTAX_DST])->sin_addr),
389 					    (void *)rtifa->ifa_getifa,
390 					    rtifa->ifa_seqno);
391 				}
392 #endif /* RTSOCK_DEBUG */
393 				if (ifp->if_flags & IFF_POINTOPOINT) {
394 					info.rti_info[RTAX_BRD] =
395 					    rtifa->ifa_dstaddr;
396 				} else
397 					info.rti_info[RTAX_BRD] = NULL;
398 				rtm->rtm_index = ifp->if_index;
399 			} else {
400 				info.rti_info[RTAX_IFP] = NULL;
401 				info.rti_info[RTAX_IFA] = NULL;
402 			}
403 			(void)rt_msg2(rtm->rtm_type, &info, NULL, NULL, &len);
404 			if (len > rtm->rtm_msglen) {
405 				old_rtm = rtm;
406 				R_Malloc(rtm, struct rt_xmsghdr *, len);
407 				if (rtm == NULL)
408 					senderr(ENOBUFS);
409 				(void)memcpy(rtm, old_rtm, old_rtm->rtm_msglen);
410 			}
411 			(void)rt_msg2(rtm->rtm_type, &info, rtm, NULL, 0);
412 			rtm->rtm_flags = rt->rt_flags;
413 			rtm_setmetrics(rt, rtm);
414 			rtm->rtm_addrs = info.rti_addrs;
415 			break;
416 
417 		case RTM_CHANGE:
418 			/*
419 			 * new gateway could require new ifaddr, ifp;
420 			 * flags may also be different; ifp may be specified
421 			 * by ll sockaddr when protocol address is ambiguous
422 			 */
423 			if ((error = rt_getifa(&info)) != 0)
424 				senderr(error);
425 			if (info.rti_info[RTAX_GATEWAY] &&
426 			    rt_setgate(rt, info.rti_info[RTAX_GATEWAY]))
427 				senderr(EDQUOT);
428 			if (info.rti_info[RTAX_TAG])
429 				rt_settag(rt, info.rti_info[RTAX_TAG]);
430 			/* new gateway could require new ifaddr, ifp;
431 			   flags may also be different; ifp may be specified
432 			   by ll sockaddr when protocol address is ambiguous */
433 			if (info.rti_info[RTAX_IFP] &&
434 			    (ifa = ifa_ifwithnet(info.rti_info[RTAX_IFP])) &&
435 			    (ifp = ifa->ifa_ifp) && (info.rti_info[RTAX_IFA] ||
436 			    info.rti_info[RTAX_GATEWAY])) {
437 				if (info.rti_info[RTAX_IFA] == NULL ||
438 				    (ifa = ifa_ifwithaddr(
439 				    info.rti_info[RTAX_IFA])) == NULL)
440 					ifa = ifaof_ifpforaddr(
441 					    info.rti_info[RTAX_IFA] ?
442 					    info.rti_info[RTAX_IFA] :
443 					    info.rti_info[RTAX_GATEWAY], ifp);
444 			} else if ((info.rti_info[RTAX_IFA] &&
445 			    (ifa = ifa_ifwithaddr(info.rti_info[RTAX_IFA]))) ||
446 			    (info.rti_info[RTAX_GATEWAY] &&
447 			    (ifa = ifa_ifwithroute(rt->rt_flags,
448 			    rt_getkey(rt), info.rti_info[RTAX_GATEWAY])))) {
449 				ifp = ifa->ifa_ifp;
450 			}
451 			if (ifa) {
452 				struct ifaddr *oifa = rt->rt_ifa;
453 				if (oifa != ifa) {
454 					if (oifa && oifa->ifa_rtrequest) {
455 						oifa->ifa_rtrequest(RTM_DELETE,
456 						    rt, &info);
457 					}
458 					rt_replace_ifa(rt, ifa);
459 					rt->rt_ifp = ifp;
460 				}
461 			}
462 			if (ifp && rt->rt_ifp != ifp)
463 				rt->rt_ifp = ifp;
464 			rt_setmetrics(rtm->rtm_inits, rtm, rt);
465 			if (rt->rt_flags != info.rti_flags)
466 				rt->rt_flags = (info.rti_flags & ~PRESERVED_RTF)
467 				    | (rt->rt_flags & PRESERVED_RTF);
468 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
469 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
470 			/*FALLTHROUGH*/
471 		case RTM_LOCK:
472 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
473 			rt->rt_rmx.rmx_locks |=
474 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
475 			break;
476 		}
477 		break;
478 
479 	default:
480 		senderr(EOPNOTSUPP);
481 	}
482 
483 flush:
484 	if (rtm) {
485 		if (error)
486 			rtm->rtm_errno = error;
487 		else
488 			rtm->rtm_flags |= RTF_DONE;
489 	}
490 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
491 	    0;
492 	/* We cannot free old_rtm until we have stopped using the
493 	 * pointers in info, some of which may point to sockaddrs
494 	 * in old_rtm.
495 	 */
496 	if (old_rtm != NULL)
497 		Free(old_rtm);
498 	if (rt)
499 		rtfree(rt);
500     {
501 	struct rawcb *rp = NULL;
502 	/*
503 	 * Check to see if we don't want our own messages.
504 	 */
505 	if ((so->so_options & SO_USELOOPBACK) == 0) {
506 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
507 			if (rtm)
508 				Free(rtm);
509 			m_freem(m);
510 			return error;
511 		}
512 		/* There is another listener, so construct message */
513 		rp = sotorawcb(so);
514 	}
515 	if (rtm) {
516 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
517 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
518 			m_freem(m);
519 			m = NULL;
520 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
521 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
522 		Free(rtm);
523 	}
524 	if (rp)
525 		rp->rcb_proto.sp_family = 0; /* Avoid us */
526 	if (family)
527 		proto.sp_protocol = family;
528 	if (m)
529 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
530 		    &COMPATNAME(route_info).ri_dst);
531 	if (rp)
532 		rp->rcb_proto.sp_family = PF_XROUTE;
533     }
534 	return error;
535 }
536 
537 static void
538 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
539 {
540 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
541 	metric(RTV_RPIPE, rmx_recvpipe);
542 	metric(RTV_SPIPE, rmx_sendpipe);
543 	metric(RTV_SSTHRESH, rmx_ssthresh);
544 	metric(RTV_RTT, rmx_rtt);
545 	metric(RTV_RTTVAR, rmx_rttvar);
546 	metric(RTV_HOPCOUNT, rmx_hopcount);
547 	metric(RTV_MTU, rmx_mtu);
548 	metric(RTV_EXPIRE, rmx_expire);
549 #undef metric
550 }
551 
552 static void
553 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
554 {
555 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
556 	metric(rmx_recvpipe);
557 	metric(rmx_sendpipe);
558 	metric(rmx_ssthresh);
559 	metric(rmx_rtt);
560 	metric(rmx_rttvar);
561 	metric(rmx_hopcount);
562 	metric(rmx_mtu);
563 	metric(rmx_expire);
564 #undef metric
565 }
566 
567 static int
568 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
569     struct rt_addrinfo *rtinfo)
570 {
571 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
572 	int i;
573 
574 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
575 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
576 			continue;
577 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
578 		RT_XADVANCE(cp, sa);
579 	}
580 
581 	/*
582 	 * Check for extra addresses specified, except RTM_GET asking
583 	 * for interface info.
584 	 */
585 	if (rtmtype == RTM_GET) {
586 		if (((rtinfo->rti_addrs &
587 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
588 			return 1;
589 	} else if ((rtinfo->rti_addrs & (~0 << i)) != 0)
590 		return 1;
591 	/* Check for bad data length.  */
592 	if (cp != cplim) {
593 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
594 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
595 			/*
596 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
597 			 * We accept this for now for the sake of old
598 			 * binaries or third party softwares.
599 			 */
600 			;
601 		else
602 			return 1;
603 	}
604 	return 0;
605 }
606 
607 static int
608 rt_getlen(int type)
609 {
610 #ifndef COMPAT_RTSOCK
611 	CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
612 	CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
613 	CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
614 	CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
615 #endif
616 
617 	switch (type) {
618 	case RTM_DELADDR:
619 	case RTM_NEWADDR:
620 	case RTM_CHGADDR:
621 		return sizeof(struct ifa_xmsghdr);
622 
623 	case RTM_OOIFINFO:
624 #ifdef COMPAT_14
625 		return sizeof(struct if_msghdr14);
626 #else
627 #ifdef DIAGNOSTIC
628 		printf("RTM_OOIFINFO\n");
629 #endif
630 		return -1;
631 #endif
632 	case RTM_OIFINFO:
633 #ifdef COMPAT_50
634 		return sizeof(struct if_msghdr50);
635 #else
636 #ifdef DIAGNOSTIC
637 		printf("RTM_OIFINFO\n");
638 #endif
639 		return -1;
640 #endif
641 
642 	case RTM_IFINFO:
643 		return sizeof(struct if_xmsghdr);
644 
645 	case RTM_IFANNOUNCE:
646 	case RTM_IEEE80211:
647 		return sizeof(struct if_xannouncemsghdr);
648 
649 	default:
650 		return sizeof(struct rt_xmsghdr);
651 	}
652 }
653 
654 
655 struct mbuf *
656 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
657 {
658 	struct rt_xmsghdr *rtm;
659 	struct mbuf *m;
660 	int i;
661 	const struct sockaddr *sa;
662 	int len, dlen;
663 
664 	m = m_gethdr(M_DONTWAIT, MT_DATA);
665 	if (m == NULL)
666 		return m;
667 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
668 
669 	if ((len = rt_getlen(type)) == -1)
670 		goto out;
671 	if (len > MHLEN + MLEN)
672 		panic("%s: message too long", __func__);
673 	else if (len > MHLEN) {
674 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
675 		if (m->m_next == NULL)
676 			goto out;
677 		MCLAIM(m->m_next, m->m_owner);
678 		m->m_pkthdr.len = len;
679 		m->m_len = MHLEN;
680 		m->m_next->m_len = len - MHLEN;
681 	} else {
682 		m->m_pkthdr.len = m->m_len = len;
683 	}
684 	m->m_pkthdr.rcvif = NULL;
685 	m_copyback(m, 0, datalen, data);
686 	if (len > datalen)
687 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
688 	rtm = mtod(m, struct rt_xmsghdr *);
689 	for (i = 0; i < RTAX_MAX; i++) {
690 		if ((sa = rtinfo->rti_info[i]) == NULL)
691 			continue;
692 		rtinfo->rti_addrs |= (1 << i);
693 		dlen = RT_XROUNDUP(sa->sa_len);
694 		m_copyback(m, len, sa->sa_len, sa);
695 		if (dlen != sa->sa_len) {
696 			/*
697 			 * Up to 6 + 1 nul's since roundup is to
698 			 * sizeof(uint64_t) (8 bytes)
699 			 */
700 			m_copyback(m, len + sa->sa_len,
701 			    dlen - sa->sa_len, "\0\0\0\0\0\0");
702 		}
703 		len += dlen;
704 	}
705 	if (m->m_pkthdr.len != len)
706 		goto out;
707 	rtm->rtm_msglen = len;
708 	rtm->rtm_version = RTM_XVERSION;
709 	rtm->rtm_type = type;
710 	return m;
711 out:
712 	m_freem(m);
713 	return NULL;
714 }
715 
716 /*
717  * rt_msg2
718  *
719  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
720  *		returns the length of the message in 'lenp'.
721  *
722  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
723  *	the message
724  * otherwise walkarg's w_needed is updated and if the user buffer is
725  *	specified and w_needed indicates space exists the information is copied
726  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
727  *	if the allocation fails ENOBUFS is returned.
728  */
729 static int
730 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
731 	int *lenp)
732 {
733 	int i;
734 	int len, dlen, second_time = 0;
735 	char *cp0, *cp = cpv;
736 
737 	rtinfo->rti_addrs = 0;
738 again:
739 	if ((len = rt_getlen(type)) == -1)
740 		return EINVAL;
741 
742 	if ((cp0 = cp) != NULL)
743 		cp += len;
744 	for (i = 0; i < RTAX_MAX; i++) {
745 		const struct sockaddr *sa;
746 
747 		if ((sa = rtinfo->rti_info[i]) == NULL)
748 			continue;
749 		rtinfo->rti_addrs |= (1 << i);
750 		dlen = RT_XROUNDUP(sa->sa_len);
751 		if (cp) {
752 			int diff = dlen - sa->sa_len;
753 			(void)memcpy(cp, sa, (size_t)sa->sa_len);
754 			cp += sa->sa_len;
755 			if (diff > 0) {
756 				(void)memset(cp, 0, (size_t)diff);
757 				cp += diff;
758 			}
759 		}
760 		len += dlen;
761 	}
762 	if (cp == NULL && w != NULL && !second_time) {
763 		struct rt_walkarg *rw = w;
764 
765 		rw->w_needed += len;
766 		if (rw->w_needed <= 0 && rw->w_where) {
767 			if (rw->w_tmemsize < len) {
768 				if (rw->w_tmem)
769 					free(rw->w_tmem, M_RTABLE);
770 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
771 				if (rw->w_tmem)
772 					rw->w_tmemsize = len;
773 				else
774 					rw->w_tmemsize = 0;
775 			}
776 			if (rw->w_tmem) {
777 				cp = rw->w_tmem;
778 				second_time = 1;
779 				goto again;
780 			} else {
781 				rw->w_tmemneeded = len;
782 				return ENOBUFS;
783 			}
784 		}
785 	}
786 	if (cp) {
787 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
788 
789 		rtm->rtm_version = RTM_XVERSION;
790 		rtm->rtm_type = type;
791 		rtm->rtm_msglen = len;
792 	}
793 	if (lenp)
794 		*lenp = len;
795 	return 0;
796 }
797 
798 /*
799  * This routine is called to generate a message from the routing
800  * socket indicating that a redirect has occurred, a routing lookup
801  * has failed, or that a protocol has detected timeouts to a particular
802  * destination.
803  */
804 void
805 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
806     int error)
807 {
808 	struct rt_xmsghdr rtm;
809 	struct mbuf *m;
810 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
811 	struct rt_addrinfo info = *rtinfo;
812 
813 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
814 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
815 		return;
816 	memset(&rtm, 0, sizeof(rtm));
817 	rtm.rtm_flags = RTF_DONE | flags;
818 	rtm.rtm_errno = error;
819 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
820 	if (m == NULL)
821 		return;
822 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
823 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
824 }
825 
826 /*
827  * This routine is called to generate a message from the routing
828  * socket indicating that the status of a network interface has changed.
829  */
830 void
831 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
832 {
833 	struct if_xmsghdr ifm;
834 	struct mbuf *m;
835 	struct rt_addrinfo info;
836 
837 	COMPATCALL(rt_ifmsg, (ifp));
838 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
839 		return;
840 	(void)memset(&info, 0, sizeof(info));
841 	(void)memset(&ifm, 0, sizeof(ifm));
842 	ifm.ifm_index = ifp->if_index;
843 	ifm.ifm_flags = ifp->if_flags;
844 	ifm.ifm_data = ifp->if_data;
845 	ifm.ifm_addrs = 0;
846 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
847 	if (m == NULL)
848 		return;
849 	COMPATNAME(route_enqueue)(m, 0);
850 #ifdef COMPAT_14
851 	compat_14_rt_oifmsg(ifp);
852 #endif
853 #ifdef COMPAT_50
854 	compat_50_rt_oifmsg(ifp);
855 #endif
856 }
857 
858 
859 /*
860  * This is called to generate messages from the routing socket
861  * indicating a network interface has had addresses associated with it.
862  * if we ever reverse the logic and replace messages TO the routing
863  * socket indicate a request to configure interfaces, then it will
864  * be unnecessary as the routing socket will automatically generate
865  * copies of it.
866  */
867 void
868 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
869     struct rtentry *rt)
870 {
871 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
872 	struct rt_addrinfo info;
873 	const struct sockaddr *sa;
874 	int pass;
875 	struct mbuf *m;
876 	struct ifnet *ifp;
877 	struct rt_xmsghdr rtm;
878 	struct ifa_xmsghdr ifam;
879 	int ncmd;
880 
881 	KASSERT(ifa != NULL);
882 	ifp = ifa->ifa_ifp;
883 	COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
884 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
885 		return;
886 	for (pass = 1; pass < 3; pass++) {
887 		memset(&info, 0, sizeof(info));
888 		switch (cmdpass(cmd, pass)) {
889 		case cmdpass(RTM_ADD, 1):
890 		case cmdpass(RTM_CHANGE, 1):
891 		case cmdpass(RTM_DELETE, 2):
892 		case cmdpass(RTM_NEWADDR, 1):
893 		case cmdpass(RTM_DELADDR, 1):
894 		case cmdpass(RTM_CHGADDR, 1):
895 			switch (cmd) {
896 			case RTM_ADD:
897 				ncmd = RTM_NEWADDR;
898 				break;
899 			case RTM_DELETE:
900 				ncmd = RTM_DELADDR;
901 				break;
902 			case RTM_CHANGE:
903 				ncmd = RTM_CHGADDR;
904 				break;
905 			default:
906 				ncmd = cmd;
907 			}
908 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
909 			KASSERT(ifp->if_dl != NULL);
910 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
911 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
912 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
913 			memset(&ifam, 0, sizeof(ifam));
914 			ifam.ifam_index = ifp->if_index;
915 			ifam.ifam_metric = ifa->ifa_metric;
916 			ifam.ifam_flags = ifa->ifa_flags;
917 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
918 			if (m == NULL)
919 				continue;
920 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
921 			    info.rti_addrs;
922 			break;
923 		case cmdpass(RTM_ADD, 2):
924 		case cmdpass(RTM_CHANGE, 2):
925 		case cmdpass(RTM_DELETE, 1):
926 			if (rt == NULL)
927 				continue;
928 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
929 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
930 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
931 			memset(&rtm, 0, sizeof(rtm));
932 			rtm.rtm_index = ifp->if_index;
933 			rtm.rtm_flags |= rt->rt_flags;
934 			rtm.rtm_errno = error;
935 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
936 			if (m == NULL)
937 				continue;
938 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
939 			break;
940 		default:
941 			continue;
942 		}
943 #ifdef DIAGNOSTIC
944 		if (m == NULL)
945 			panic("%s: called with wrong command", __func__);
946 #endif
947 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
948 	}
949 #undef cmdpass
950 }
951 
952 static struct mbuf *
953 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
954     struct rt_addrinfo *info)
955 {
956 	struct if_xannouncemsghdr ifan;
957 
958 	memset(info, 0, sizeof(*info));
959 	memset(&ifan, 0, sizeof(ifan));
960 	ifan.ifan_index = ifp->if_index;
961 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
962 	ifan.ifan_what = what;
963 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
964 }
965 
966 /*
967  * This is called to generate routing socket messages indicating
968  * network interface arrival and departure.
969  */
970 void
971 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
972 {
973 	struct mbuf *m;
974 	struct rt_addrinfo info;
975 
976 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
977 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
978 		return;
979 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
980 	if (m == NULL)
981 		return;
982 	COMPATNAME(route_enqueue)(m, 0);
983 }
984 
985 /*
986  * This is called to generate routing socket messages indicating
987  * IEEE80211 wireless events.
988  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
989  */
990 void
991 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
992 	size_t data_len)
993 {
994 	struct mbuf *m;
995 	struct rt_addrinfo info;
996 
997 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
998 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
999 		return;
1000 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1001 	if (m == NULL)
1002 		return;
1003 	/*
1004 	 * Append the ieee80211 data.  Try to stick it in the
1005 	 * mbuf containing the ifannounce msg; otherwise allocate
1006 	 * a new mbuf and append.
1007 	 *
1008 	 * NB: we assume m is a single mbuf.
1009 	 */
1010 	if (data_len > M_TRAILINGSPACE(m)) {
1011 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1012 		if (n == NULL) {
1013 			m_freem(m);
1014 			return;
1015 		}
1016 		(void)memcpy(mtod(n, void *), data, data_len);
1017 		n->m_len = data_len;
1018 		m->m_next = n;
1019 	} else if (data_len > 0) {
1020 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
1021 		m->m_len += data_len;
1022 	}
1023 	if (m->m_flags & M_PKTHDR)
1024 		m->m_pkthdr.len += data_len;
1025 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
1026 	COMPATNAME(route_enqueue)(m, 0);
1027 }
1028 
1029 /*
1030  * This is used in dumping the kernel table via sysctl().
1031  */
1032 static int
1033 sysctl_dumpentry(struct rtentry *rt, void *v)
1034 {
1035 	struct rt_walkarg *w = v;
1036 	int error = 0, size;
1037 	struct rt_addrinfo info;
1038 
1039 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1040 		return 0;
1041 	memset(&info, 0, sizeof(info));
1042 	info.rti_info[RTAX_DST] = rt_getkey(rt);
1043 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1044 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1045 	info.rti_info[RTAX_TAG] = rt_gettag(rt);
1046 	if (rt->rt_ifp) {
1047 		const struct ifaddr *rtifa;
1048 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
1049 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
1050 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
1051 		 * --dyoung
1052 		 */
1053 		rtifa = rt_get_ifa(rt);
1054 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
1055 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1056 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
1057 	}
1058 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
1059 		return error;
1060 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1061 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
1062 
1063 		rtm->rtm_flags = rt->rt_flags;
1064 		rtm->rtm_use = rt->rt_use;
1065 		rtm_setmetrics(rt, rtm);
1066 		KASSERT(rt->rt_ifp != NULL);
1067 		rtm->rtm_index = rt->rt_ifp->if_index;
1068 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1069 		rtm->rtm_addrs = info.rti_addrs;
1070 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1071 			w->w_where = NULL;
1072 		else
1073 			w->w_where = (char *)w->w_where + size;
1074 	}
1075 	return error;
1076 }
1077 
1078 static int
1079 sysctl_iflist(int af, struct rt_walkarg *w, int type)
1080 {
1081 	struct ifnet *ifp;
1082 	struct ifaddr *ifa;
1083 	struct	rt_addrinfo info;
1084 	int	len, error = 0;
1085 
1086 	memset(&info, 0, sizeof(info));
1087 	IFNET_FOREACH(ifp) {
1088 		if (w->w_arg && w->w_arg != ifp->if_index)
1089 			continue;
1090 		if (IFADDR_EMPTY(ifp))
1091 			continue;
1092 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1093 		switch (type) {
1094 		case NET_RT_IFLIST:
1095 			error = rt_msg2(RTM_IFINFO, &info, NULL, w, &len);
1096 			break;
1097 #ifdef COMPAT_14
1098 		case NET_RT_OOIFLIST:
1099 			error = rt_msg2(RTM_OOIFINFO, &info, NULL, w, &len);
1100 			break;
1101 #endif
1102 #ifdef COMPAT_50
1103 		case NET_RT_OIFLIST:
1104 			error = rt_msg2(RTM_OIFINFO, &info, NULL, w, &len);
1105 			break;
1106 #endif
1107 		default:
1108 			panic("sysctl_iflist(1)");
1109 		}
1110 		if (error)
1111 			return error;
1112 		info.rti_info[RTAX_IFP] = NULL;
1113 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1114 			switch (type) {
1115 			case NET_RT_IFLIST: {
1116 				struct if_xmsghdr *ifm;
1117 
1118 				ifm = (struct if_xmsghdr *)w->w_tmem;
1119 				ifm->ifm_index = ifp->if_index;
1120 				ifm->ifm_flags = ifp->if_flags;
1121 				ifm->ifm_data = ifp->if_data;
1122 				ifm->ifm_addrs = info.rti_addrs;
1123 				error = copyout(ifm, w->w_where, len);
1124 				if (error)
1125 					return error;
1126 				w->w_where = (char *)w->w_where + len;
1127 				break;
1128 			}
1129 
1130 #ifdef COMPAT_14
1131 			case NET_RT_OOIFLIST:
1132 				error = compat_14_iflist(ifp, w, &info, len);
1133 				if (error)
1134 					return error;
1135 				break;
1136 #endif
1137 #ifdef COMPAT_50
1138 			case NET_RT_OIFLIST:
1139 				error = compat_50_iflist(ifp, w, &info, len);
1140 				if (error)
1141 					return error;
1142 				break;
1143 #endif
1144 			default:
1145 				panic("sysctl_iflist(2)");
1146 			}
1147 		}
1148 		IFADDR_FOREACH(ifa, ifp) {
1149 			if (af && af != ifa->ifa_addr->sa_family)
1150 				continue;
1151 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1152 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1153 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1154 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
1155 				return error;
1156 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1157 				struct ifa_xmsghdr *ifam;
1158 
1159 				ifam = (struct ifa_xmsghdr *)w->w_tmem;
1160 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1161 				ifam->ifam_flags = ifa->ifa_flags;
1162 				ifam->ifam_metric = ifa->ifa_metric;
1163 				ifam->ifam_addrs = info.rti_addrs;
1164 				error = copyout(w->w_tmem, w->w_where, len);
1165 				if (error)
1166 					return error;
1167 				w->w_where = (char *)w->w_where + len;
1168 			}
1169 		}
1170 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1171 		    info.rti_info[RTAX_BRD] = NULL;
1172 	}
1173 	return 0;
1174 }
1175 
1176 static int
1177 sysctl_rtable(SYSCTLFN_ARGS)
1178 {
1179 	void 	*where = oldp;
1180 	size_t	*given = oldlenp;
1181 	const void *new = newp;
1182 	int	i, s, error = EINVAL;
1183 	u_char  af;
1184 	struct	rt_walkarg w;
1185 
1186 	if (namelen == 1 && name[0] == CTL_QUERY)
1187 		return sysctl_query(SYSCTLFN_CALL(rnode));
1188 
1189 	if (new)
1190 		return EPERM;
1191 	if (namelen != 3)
1192 		return EINVAL;
1193 	af = name[0];
1194 	w.w_tmemneeded = 0;
1195 	w.w_tmemsize = 0;
1196 	w.w_tmem = NULL;
1197 again:
1198 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1199 	if (w.w_tmemneeded) {
1200 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
1201 		w.w_tmemsize = w.w_tmemneeded;
1202 		w.w_tmemneeded = 0;
1203 	}
1204 	w.w_op = name[1];
1205 	w.w_arg = name[2];
1206 	w.w_given = *given;
1207 	w.w_needed = 0 - w.w_given;
1208 	w.w_where = where;
1209 
1210 	s = splsoftnet();
1211 	switch (w.w_op) {
1212 
1213 	case NET_RT_DUMP:
1214 	case NET_RT_FLAGS:
1215 		for (i = 1; i <= AF_MAX; i++)
1216 			if ((af == 0 || af == i) &&
1217 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
1218 				break;
1219 		break;
1220 
1221 #ifdef COMPAT_14
1222 	case NET_RT_OOIFLIST:
1223 		error = sysctl_iflist(af, &w, w.w_op);
1224 		break;
1225 #endif
1226 #ifdef COMPAT_50
1227 	case NET_RT_OIFLIST:
1228 		error = sysctl_iflist(af, &w, w.w_op);
1229 		break;
1230 #endif
1231 	case NET_RT_IFLIST:
1232 		error = sysctl_iflist(af, &w, w.w_op);
1233 		break;
1234 	}
1235 	splx(s);
1236 
1237 	/* check to see if we couldn't allocate memory with NOWAIT */
1238 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1239 		goto again;
1240 
1241 	if (w.w_tmem)
1242 		free(w.w_tmem, M_RTABLE);
1243 	w.w_needed += w.w_given;
1244 	if (where) {
1245 		*given = (char *)w.w_where - (char *)where;
1246 		if (*given < w.w_needed)
1247 			return ENOMEM;
1248 	} else {
1249 		*given = (11 * w.w_needed) / 10;
1250 	}
1251 	return error;
1252 }
1253 
1254 /*
1255  * Routing message software interrupt routine
1256  */
1257 static void
1258 COMPATNAME(route_intr)(void *cookie)
1259 {
1260 	struct sockproto proto = { .sp_family = PF_XROUTE, };
1261 	struct route_info * const ri = &COMPATNAME(route_info);
1262 	struct mbuf *m;
1263 	int s;
1264 
1265 	mutex_enter(softnet_lock);
1266 	KERNEL_LOCK(1, NULL);
1267 	while (!IF_IS_EMPTY(&ri->ri_intrq)) {
1268 		s = splnet();
1269 		IF_DEQUEUE(&ri->ri_intrq, m);
1270 		splx(s);
1271 		if (m == NULL)
1272 			break;
1273 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
1274 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst);
1275 	}
1276 	KERNEL_UNLOCK_ONE(NULL);
1277 	mutex_exit(softnet_lock);
1278 }
1279 
1280 /*
1281  * Enqueue a message to the software interrupt routine.
1282  */
1283 void
1284 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
1285 {
1286 	struct route_info * const ri = &COMPATNAME(route_info);
1287 	int s, wasempty;
1288 
1289 	s = splnet();
1290 	if (IF_QFULL(&ri->ri_intrq)) {
1291 		IF_DROP(&ri->ri_intrq);
1292 		m_freem(m);
1293 	} else {
1294 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
1295 		M_SETCTX(m, (uintptr_t)family);
1296 		IF_ENQUEUE(&ri->ri_intrq, m);
1297 		if (wasempty)
1298 			softint_schedule(ri->ri_sih);
1299 	}
1300 	splx(s);
1301 }
1302 
1303 static void
1304 COMPATNAME(route_init)(void)
1305 {
1306 	struct route_info * const ri = &COMPATNAME(route_info);
1307 
1308 #ifndef COMPAT_RTSOCK
1309 	rt_init();
1310 #endif
1311 
1312 	sysctl_net_route_setup(NULL);
1313 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
1314 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
1315 	    COMPATNAME(route_intr), NULL);
1316 }
1317 
1318 /*
1319  * Definitions of protocols supported in the ROUTE domain.
1320  */
1321 #ifndef COMPAT_RTSOCK
1322 PR_WRAP_USRREQS(route);
1323 #else
1324 PR_WRAP_USRREQS(compat_50_route);
1325 #endif
1326 
1327 static const struct pr_usrreqs route_usrreqs = {
1328 	.pr_attach	= COMPATNAME(route_attach_wrapper),
1329 	.pr_detach	= COMPATNAME(route_detach_wrapper),
1330 	.pr_generic	= COMPATNAME(route_usrreq_wrapper),
1331 };
1332 
1333 static const struct protosw COMPATNAME(route_protosw)[] = {
1334 	{
1335 		.pr_type = SOCK_RAW,
1336 		.pr_domain = &COMPATNAME(routedomain),
1337 		.pr_flags = PR_ATOMIC|PR_ADDR,
1338 		.pr_input = raw_input,
1339 		.pr_output = COMPATNAME(route_output),
1340 		.pr_ctlinput = raw_ctlinput,
1341 		.pr_usrreqs = &route_usrreqs,
1342 		.pr_init = raw_init,
1343 	},
1344 };
1345 
1346 struct domain COMPATNAME(routedomain) = {
1347 	.dom_family = PF_XROUTE,
1348 	.dom_name = DOMAINNAME,
1349 	.dom_init = COMPATNAME(route_init),
1350 	.dom_protosw = COMPATNAME(route_protosw),
1351 	.dom_protoswNPROTOSW =
1352 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
1353 };
1354 
1355 static void
1356 sysctl_net_route_setup(struct sysctllog **clog)
1357 {
1358 	const struct sysctlnode *rnode = NULL;
1359 
1360 	sysctl_createv(clog, 0, NULL, &rnode,
1361 		       CTLFLAG_PERMANENT,
1362 		       CTLTYPE_NODE, DOMAINNAME,
1363 		       SYSCTL_DESCR("PF_ROUTE information"),
1364 		       NULL, 0, NULL, 0,
1365 		       CTL_NET, PF_XROUTE, CTL_EOL);
1366 
1367 	sysctl_createv(clog, 0, NULL, NULL,
1368 		       CTLFLAG_PERMANENT,
1369 		       CTLTYPE_NODE, "rtable",
1370 		       SYSCTL_DESCR("Routing table information"),
1371 		       sysctl_rtable, 0, NULL, 0,
1372 		       CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
1373 
1374 	sysctl_createv(clog, 0, &rnode, NULL,
1375 		       CTLFLAG_PERMANENT,
1376 		       CTLTYPE_STRUCT, "stats",
1377 		       SYSCTL_DESCR("Routing statistics"),
1378 		       NULL, 0, &rtstat, sizeof(rtstat),
1379 		       CTL_CREATE, CTL_EOL);
1380 }
1381