xref: /netbsd-src/sys/net/rtsock.c (revision c2f76ff004a2cb67efe5b12d97bd3ef7fe89e18d)
1 /*	$NetBSD: rtsock.c,v 1.132 2010/12/25 20:37:44 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 1.132 2010/12/25 20:37:44 christos Exp $");
65 
66 #include "opt_inet.h"
67 #include "opt_mpls.h"
68 #ifdef _KERNEL_OPT
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 #endif
96 
97 #include <machine/stdarg.h>
98 
99 DOMAIN_DEFINE(routedomain);	/* forward declare and add to link set */
100 
101 struct	sockaddr route_dst = { .sa_len = 2, .sa_family = PF_ROUTE, };
102 struct	sockaddr route_src = { .sa_len = 2, .sa_family = PF_ROUTE, };
103 
104 int	route_maxqlen = IFQ_MAXLEN;
105 static struct	ifqueue route_intrq;
106 static void	*route_sih;
107 
108 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
109 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
110 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
111     struct rt_addrinfo *);
112 static void sysctl_net_route_setup(struct sysctllog **);
113 static int sysctl_dumpentry(struct rtentry *, void *);
114 static int sysctl_iflist(int, struct rt_walkarg *, int);
115 static int sysctl_rtable(SYSCTLFN_PROTO);
116 static void rt_adjustcount(int, int);
117 
118 static void
119 rt_adjustcount(int af, int cnt)
120 {
121 	route_cb.any_count += cnt;
122 	switch (af) {
123 	case AF_INET:
124 		route_cb.ip_count += cnt;
125 		return;
126 #ifdef INET6
127 	case AF_INET6:
128 		route_cb.ip6_count += cnt;
129 		return;
130 #endif
131 	case AF_IPX:
132 		route_cb.ipx_count += cnt;
133 		return;
134 	case AF_ISO:
135 		route_cb.iso_count += cnt;
136 		return;
137 	case AF_MPLS:
138 		route_cb.mpls_count += cnt;
139 		return;
140 	case AF_NS:
141 		route_cb.ns_count += cnt;
142 		return;
143 	}
144 }
145 
146 static void
147 cvtmetrics(struct rt_metrics *ortm, const struct nrt_metrics *rtm)
148 {
149 	ortm->rmx_locks = rtm->rmx_locks;
150 	ortm->rmx_mtu = rtm->rmx_mtu;
151 	ortm->rmx_hopcount = rtm->rmx_hopcount;
152 	ortm->rmx_expire = rtm->rmx_expire;
153 	ortm->rmx_recvpipe = rtm->rmx_recvpipe;
154 	ortm->rmx_sendpipe = rtm->rmx_sendpipe;
155 	ortm->rmx_ssthresh = rtm->rmx_ssthresh;
156 	ortm->rmx_rtt = rtm->rmx_rtt;
157 	ortm->rmx_rttvar = rtm->rmx_rttvar;
158 	ortm->rmx_pksent = rtm->rmx_pksent;
159 }
160 
161 /*ARGSUSED*/
162 int
163 route_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
164 	struct mbuf *control, struct lwp *l)
165 {
166 	int error = 0;
167 	struct rawcb *rp = sotorawcb(so);
168 	int s;
169 
170 	if (req == PRU_ATTACH) {
171 		sosetlock(so);
172 		rp = malloc(sizeof(*rp), M_PCB, M_WAITOK|M_ZERO);
173 		so->so_pcb = rp;
174 	}
175 	if (req == PRU_DETACH && rp)
176 		rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
177 	s = splsoftnet();
178 
179 	/*
180 	 * Don't call raw_usrreq() in the attach case, because
181 	 * we want to allow non-privileged processes to listen on
182 	 * and send "safe" commands to the routing socket.
183 	 */
184 	if (req == PRU_ATTACH) {
185 		if (l == NULL)
186 			error = EACCES;
187 		else
188 			error = raw_attach(so, (int)(long)nam);
189 	} else
190 		error = raw_usrreq(so, req, m, nam, control, l);
191 
192 	rp = sotorawcb(so);
193 	if (req == PRU_ATTACH && rp) {
194 		if (error) {
195 			free(rp, M_PCB);
196 			splx(s);
197 			return error;
198 		}
199 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
200 		rp->rcb_laddr = &route_src;
201 		rp->rcb_faddr = &route_dst;
202 		soisconnected(so);
203 		so->so_options |= SO_USELOOPBACK;
204 	}
205 	splx(s);
206 	return error;
207 }
208 
209 static const struct sockaddr *
210 intern_netmask(const struct sockaddr *mask)
211 {
212 	struct radix_node *rn;
213 	extern struct radix_node_head *mask_rnhead;
214 
215 	if (mask != NULL &&
216 	    (rn = rn_search(mask, mask_rnhead->rnh_treetop)))
217 		mask = (const struct sockaddr *)rn->rn_key;
218 
219 	return mask;
220 }
221 
222 /*ARGSUSED*/
223 int
224 route_output(struct mbuf *m, ...)
225 {
226 	struct sockproto proto = { .sp_family = PF_ROUTE, };
227 	struct rt_msghdr *rtm = NULL;
228 	struct rt_msghdr *old_rtm = NULL;
229 	struct rtentry *rt = NULL;
230 	struct rtentry *saved_nrt = NULL;
231 	struct rt_addrinfo info;
232 	int len, error = 0;
233 	struct ifnet *ifp = NULL;
234 	struct ifaddr *ifa = NULL;
235 	struct socket *so;
236 	va_list ap;
237 	sa_family_t family;
238 
239 	va_start(ap, m);
240 	so = va_arg(ap, struct socket *);
241 	va_end(ap);
242 
243 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
244 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
245 	   (m = m_pullup(m, sizeof(int32_t))) == NULL))
246 		return ENOBUFS;
247 	if ((m->m_flags & M_PKTHDR) == 0)
248 		panic("route_output");
249 	len = m->m_pkthdr.len;
250 	if (len < sizeof(*rtm) ||
251 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
252 		info.rti_info[RTAX_DST] = NULL;
253 		senderr(EINVAL);
254 	}
255 	R_Malloc(rtm, struct rt_msghdr *, len);
256 	if (rtm == NULL) {
257 		info.rti_info[RTAX_DST] = NULL;
258 		senderr(ENOBUFS);
259 	}
260 	m_copydata(m, 0, len, rtm);
261 	if (rtm->rtm_version != RTM_VERSION) {
262 		info.rti_info[RTAX_DST] = NULL;
263 		senderr(EPROTONOSUPPORT);
264 	}
265 	rtm->rtm_pid = curproc->p_pid;
266 	memset(&info, 0, sizeof(info));
267 	info.rti_addrs = rtm->rtm_addrs;
268 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
269 	    &info))
270 		senderr(EINVAL);
271 	info.rti_flags = rtm->rtm_flags;
272 #ifdef RTSOCK_DEBUG
273 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
274 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
275 		    inet_ntoa(((const struct sockaddr_in *)
276 		    info.rti_info[RTAX_DST])->sin_addr));
277 	}
278 #endif /* RTSOCK_DEBUG */
279 	if (info.rti_info[RTAX_DST] == NULL ||
280 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX))
281 		senderr(EINVAL);
282 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
283 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
284 		senderr(EINVAL);
285 
286 	/*
287 	 * Verify that the caller has the appropriate privilege; RTM_GET
288 	 * is the only operation the non-superuser is allowed.
289 	 */
290 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
291 	    0, rtm, NULL, NULL) != 0)
292 		senderr(EACCES);
293 
294 	switch (rtm->rtm_type) {
295 
296 	case RTM_ADD:
297 		if (info.rti_info[RTAX_GATEWAY] == NULL)
298 			senderr(EINVAL);
299 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
300 		if (error == 0 && saved_nrt) {
301 			rt_setmetrics(rtm->rtm_inits,
302 			    &rtm->rtm_rmx, &saved_nrt->rt_rmx);
303 			saved_nrt->rt_refcnt--;
304 		}
305 		break;
306 
307 	case RTM_DELETE:
308 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
309 		if (error == 0) {
310 			(rt = saved_nrt)->rt_refcnt++;
311 			goto report;
312 		}
313 		break;
314 
315 	case RTM_GET:
316 	case RTM_CHANGE:
317 	case RTM_LOCK:
318                 /* XXX This will mask info.rti_info[RTAX_DST] with
319 		 * info.rti_info[RTAX_NETMASK] before
320                  * searching.  It did not used to do that.  --dyoung
321 		 */
322 		error = rtrequest1(RTM_GET, &info, &rt);
323 		if (error != 0)
324 			senderr(error);
325 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
326 			struct radix_node *rn;
327 
328 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
329 			    info.rti_info[RTAX_DST]->sa_len) != 0)
330 				senderr(ESRCH);
331 			info.rti_info[RTAX_NETMASK] = intern_netmask(
332 			    info.rti_info[RTAX_NETMASK]);
333 			for (rn = rt->rt_nodes; rn; rn = rn->rn_dupedkey)
334 				if (info.rti_info[RTAX_NETMASK] ==
335 				    (const struct sockaddr *)rn->rn_mask)
336 					break;
337 			if (rn == NULL)
338 				senderr(ETOOMANYREFS);
339 			rt = (struct rtentry *)rn;
340 		}
341 
342 		switch (rtm->rtm_type) {
343 		case RTM_GET:
344 		report:
345 			info.rti_info[RTAX_DST] = rt_getkey(rt);
346 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
347 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
348 			info.rti_info[RTAX_TAG] = (struct sockaddr*)rt_gettag(rt);
349 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
350 				;
351 			else if ((ifp = rt->rt_ifp) != NULL) {
352 				const struct ifaddr *rtifa;
353 				info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
354                                 /* rtifa used to be simply rt->rt_ifa.
355                                  * If rt->rt_ifa != NULL, then
356                                  * rt_get_ifa() != NULL.  So this
357                                  * ought to still be safe. --dyoung
358 				 */
359 				rtifa = rt_get_ifa(rt);
360 				info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
361 #ifdef RTSOCK_DEBUG
362 				if (info.rti_info[RTAX_IFA]->sa_family ==
363 				    AF_INET) {
364 					printf("%s: copying out RTAX_IFA %s ",
365 					    __func__, inet_ntoa(
366 					    ((const struct sockaddr_in *)
367 					    info.rti_info[RTAX_IFA])->sin_addr)
368 					    );
369 					printf("for info.rti_info[RTAX_DST] %s "
370 					    "ifa_getifa %p ifa_seqno %p\n",
371 					    inet_ntoa(
372 					    ((const struct sockaddr_in *)
373 					    info.rti_info[RTAX_DST])->sin_addr),
374 					    (void *)rtifa->ifa_getifa,
375 					    rtifa->ifa_seqno);
376 				}
377 #endif /* RTSOCK_DEBUG */
378 				if (ifp->if_flags & IFF_POINTOPOINT) {
379 					info.rti_info[RTAX_BRD] =
380 					    rtifa->ifa_dstaddr;
381 				} else
382 					info.rti_info[RTAX_BRD] = NULL;
383 				rtm->rtm_index = ifp->if_index;
384 			} else {
385 				info.rti_info[RTAX_IFP] = NULL;
386 				info.rti_info[RTAX_IFA] = NULL;
387 			}
388 			(void)rt_msg2(rtm->rtm_type, &info, NULL, NULL, &len);
389 			if (len > rtm->rtm_msglen) {
390 				old_rtm = rtm;
391 				R_Malloc(rtm, struct rt_msghdr *, len);
392 				if (rtm == NULL)
393 					senderr(ENOBUFS);
394 				(void)memcpy(rtm, old_rtm, old_rtm->rtm_msglen);
395 			}
396 			(void)rt_msg2(rtm->rtm_type, &info, rtm, NULL, 0);
397 			rtm->rtm_flags = rt->rt_flags;
398 			cvtmetrics(&rtm->rtm_rmx, &rt->rt_rmx);
399 			rtm->rtm_addrs = info.rti_addrs;
400 			break;
401 
402 		case RTM_CHANGE:
403 			/*
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 			 */
408 			if ((error = rt_getifa(&info)) != 0)
409 				senderr(error);
410 			if (info.rti_info[RTAX_GATEWAY] &&
411 			    rt_setgate(rt, info.rti_info[RTAX_GATEWAY]))
412 				senderr(EDQUOT);
413 			if (info.rti_info[RTAX_TAG])
414 				rt_settag(rt, info.rti_info[RTAX_TAG]);
415 			/* new gateway could require new ifaddr, ifp;
416 			   flags may also be different; ifp may be specified
417 			   by ll sockaddr when protocol address is ambiguous */
418 			if (info.rti_info[RTAX_IFP] &&
419 			    (ifa = ifa_ifwithnet(info.rti_info[RTAX_IFP])) &&
420 			    (ifp = ifa->ifa_ifp) && (info.rti_info[RTAX_IFA] ||
421 			    info.rti_info[RTAX_GATEWAY])) {
422 				if (info.rti_info[RTAX_IFA] == NULL ||
423 				    (ifa = ifa_ifwithaddr(
424 				    info.rti_info[RTAX_IFA])) == NULL)
425 					ifa = ifaof_ifpforaddr(
426 					    info.rti_info[RTAX_IFA] ?
427 					    info.rti_info[RTAX_IFA] :
428 					    info.rti_info[RTAX_GATEWAY], ifp);
429 			} else if ((info.rti_info[RTAX_IFA] &&
430 			    (ifa = ifa_ifwithaddr(info.rti_info[RTAX_IFA]))) ||
431 			    (info.rti_info[RTAX_GATEWAY] &&
432 			    (ifa = ifa_ifwithroute(rt->rt_flags,
433 			    rt_getkey(rt), info.rti_info[RTAX_GATEWAY])))) {
434 				ifp = ifa->ifa_ifp;
435 			}
436 			if (ifa) {
437 				struct ifaddr *oifa = rt->rt_ifa;
438 				if (oifa != ifa) {
439 					if (oifa && oifa->ifa_rtrequest) {
440 						oifa->ifa_rtrequest(RTM_DELETE,
441 						    rt, &info);
442 					}
443 					rt_replace_ifa(rt, ifa);
444 					rt->rt_ifp = ifp;
445 				}
446 			}
447 			if (ifp && rt->rt_ifp != ifp)
448 				rt->rt_ifp = ifp;
449 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
450 			    &rt->rt_rmx);
451 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
452 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
453 			/*FALLTHROUGH*/
454 		case RTM_LOCK:
455 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
456 			rt->rt_rmx.rmx_locks |=
457 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
458 			break;
459 		}
460 		break;
461 
462 	default:
463 		senderr(EOPNOTSUPP);
464 	}
465 
466 flush:
467 	if (rtm) {
468 		if (error)
469 			rtm->rtm_errno = error;
470 		else
471 			rtm->rtm_flags |= RTF_DONE;
472 	}
473 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
474 	    0;
475 	/* We cannot free old_rtm until we have stopped using the
476 	 * pointers in info, some of which may point to sockaddrs
477 	 * in old_rtm.
478 	 */
479 	if (old_rtm != NULL)
480 		Free(old_rtm);
481 	if (rt)
482 		rtfree(rt);
483     {
484 	struct rawcb *rp = NULL;
485 	/*
486 	 * Check to see if we don't want our own messages.
487 	 */
488 	if ((so->so_options & SO_USELOOPBACK) == 0) {
489 		if (route_cb.any_count <= 1) {
490 			if (rtm)
491 				Free(rtm);
492 			m_freem(m);
493 			return error;
494 		}
495 		/* There is another listener, so construct message */
496 		rp = sotorawcb(so);
497 	}
498 	if (rtm) {
499 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
500 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
501 			m_freem(m);
502 			m = NULL;
503 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
504 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
505 		Free(rtm);
506 	}
507 	if (rp)
508 		rp->rcb_proto.sp_family = 0; /* Avoid us */
509 	if (family)
510 		proto.sp_protocol = family;
511 	if (m)
512 		raw_input(m, &proto, &route_src, &route_dst);
513 	if (rp)
514 		rp->rcb_proto.sp_family = PF_ROUTE;
515     }
516 	return error;
517 }
518 
519 void
520 rt_setmetrics(u_long which, const struct rt_metrics *in, struct nrt_metrics *out)
521 {
522 #define metric(f, e) if (which & (f)) out->e = in->e;
523 	metric(RTV_RPIPE, rmx_recvpipe);
524 	metric(RTV_SPIPE, rmx_sendpipe);
525 	metric(RTV_SSTHRESH, rmx_ssthresh);
526 	metric(RTV_RTT, rmx_rtt);
527 	metric(RTV_RTTVAR, rmx_rttvar);
528 	metric(RTV_HOPCOUNT, rmx_hopcount);
529 	metric(RTV_MTU, rmx_mtu);
530 	/* XXX time_t: Will not work after February 2145 (u_long time) */
531 	metric(RTV_EXPIRE, rmx_expire);
532 #undef metric
533 }
534 
535 static int
536 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
537     struct rt_addrinfo *rtinfo)
538 {
539 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
540 	int i;
541 
542 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
543 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
544 			continue;
545 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
546 		RT_ADVANCE(cp, sa);
547 	}
548 
549 	/*
550 	 * Check for extra addresses specified, except RTM_GET asking
551 	 * for interface info.
552 	 */
553 	if (rtmtype == RTM_GET) {
554 		if (((rtinfo->rti_addrs &
555 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
556 			return 1;
557 	} else if ((rtinfo->rti_addrs & (~0 << i)) != 0)
558 		return 1;
559 	/* Check for bad data length.  */
560 	if (cp != cplim) {
561 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
562 		    cp - RT_ROUNDUP(sa->sa_len) + sa->sa_len == cplim)
563 			/*
564 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
565 			 * We accept this for now for the sake of old
566 			 * binaries or third party softwares.
567 			 */
568 			;
569 		else
570 			return 1;
571 	}
572 	return 0;
573 }
574 
575 static int
576 rt_getlen(int type)
577 {
578 	switch (type) {
579 	case RTM_DELADDR:
580 	case RTM_NEWADDR:
581 	case RTM_CHGADDR:
582 		return sizeof(struct ifa_msghdr);
583 
584 	case RTM_OOIFINFO:
585 #ifdef COMPAT_14
586 		return sizeof(struct if_msghdr14);
587 #else
588 #ifdef DIAGNOSTIC
589 		printf("RTM_OOIFINFO\n");
590 #endif
591 		return -1;
592 #endif
593 	case RTM_OIFINFO:
594 #ifdef COMPAT_50
595 		return sizeof(struct if_msghdr50);
596 #else
597 #ifdef DIAGNOSTIC
598 		printf("RTM_OIFINFO\n");
599 #endif
600 		return -1;
601 #endif
602 
603 	case RTM_IFINFO:
604 		return sizeof(struct if_msghdr);
605 
606 	case RTM_IFANNOUNCE:
607 	case RTM_IEEE80211:
608 		return sizeof(struct if_announcemsghdr);
609 
610 	default:
611 		return sizeof(struct rt_msghdr);
612 	}
613 }
614 
615 
616 struct mbuf *
617 rt_msg1(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
618 {
619 	struct rt_msghdr *rtm;
620 	struct mbuf *m;
621 	int i;
622 	const struct sockaddr *sa;
623 	int len, dlen;
624 
625 	m = m_gethdr(M_DONTWAIT, MT_DATA);
626 	if (m == NULL)
627 		return m;
628 	MCLAIM(m, &routedomain.dom_mowner);
629 
630 	if ((len = rt_getlen(type)) == -1)
631 		goto out;
632 	if (len > MHLEN + MLEN)
633 		panic("rt_msg1: message too long");
634 	else if (len > MHLEN) {
635 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
636 		if (m->m_next == NULL)
637 			goto out;
638 		MCLAIM(m->m_next, m->m_owner);
639 		m->m_pkthdr.len = len;
640 		m->m_len = MHLEN;
641 		m->m_next->m_len = len - MHLEN;
642 	} else {
643 		m->m_pkthdr.len = m->m_len = len;
644 	}
645 	m->m_pkthdr.rcvif = NULL;
646 	m_copyback(m, 0, datalen, data);
647 	if (len > datalen)
648 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
649 	rtm = mtod(m, struct rt_msghdr *);
650 	for (i = 0; i < RTAX_MAX; i++) {
651 		if ((sa = rtinfo->rti_info[i]) == NULL)
652 			continue;
653 		rtinfo->rti_addrs |= (1 << i);
654 		dlen = RT_ROUNDUP(sa->sa_len);
655 		m_copyback(m, len, dlen, sa);
656 		len += dlen;
657 	}
658 	if (m->m_pkthdr.len != len)
659 		goto out;
660 	rtm->rtm_msglen = len;
661 	rtm->rtm_version = RTM_VERSION;
662 	rtm->rtm_type = type;
663 	return m;
664 out:
665 	m_freem(m);
666 	return NULL;
667 }
668 
669 /*
670  * rt_msg2
671  *
672  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
673  *		returns the length of the message in 'lenp'.
674  *
675  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
676  *	the message
677  * otherwise walkarg's w_needed is updated and if the user buffer is
678  *	specified and w_needed indicates space exists the information is copied
679  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
680  *	if the allocation fails ENOBUFS is returned.
681  */
682 static int
683 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
684 	int *lenp)
685 {
686 	int i;
687 	int len, dlen, second_time = 0;
688 	char *cp0, *cp = cpv;
689 
690 	rtinfo->rti_addrs = 0;
691 again:
692 	if ((len = rt_getlen(type)) == -1)
693 		return EINVAL;
694 
695 	if ((cp0 = cp) != NULL)
696 		cp += len;
697 	for (i = 0; i < RTAX_MAX; i++) {
698 		const struct sockaddr *sa;
699 
700 		if ((sa = rtinfo->rti_info[i]) == NULL)
701 			continue;
702 		rtinfo->rti_addrs |= (1 << i);
703 		dlen = RT_ROUNDUP(sa->sa_len);
704 		if (cp) {
705 			(void)memcpy(cp, sa, (size_t)dlen);
706 			cp += dlen;
707 		}
708 		len += dlen;
709 	}
710 	if (cp == NULL && w != NULL && !second_time) {
711 		struct rt_walkarg *rw = w;
712 
713 		rw->w_needed += len;
714 		if (rw->w_needed <= 0 && rw->w_where) {
715 			if (rw->w_tmemsize < len) {
716 				if (rw->w_tmem)
717 					free(rw->w_tmem, M_RTABLE);
718 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
719 				if (rw->w_tmem)
720 					rw->w_tmemsize = len;
721 				else
722 					rw->w_tmemsize = 0;
723 			}
724 			if (rw->w_tmem) {
725 				cp = rw->w_tmem;
726 				second_time = 1;
727 				goto again;
728 			} else {
729 				rw->w_tmemneeded = len;
730 				return ENOBUFS;
731 			}
732 		}
733 	}
734 	if (cp) {
735 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
736 
737 		rtm->rtm_version = RTM_VERSION;
738 		rtm->rtm_type = type;
739 		rtm->rtm_msglen = len;
740 	}
741 	if (lenp)
742 		*lenp = len;
743 	return 0;
744 }
745 
746 /*
747  * This routine is called to generate a message from the routing
748  * socket indicating that a redirect has occurred, a routing lookup
749  * has failed, or that a protocol has detected timeouts to a particular
750  * destination.
751  */
752 void
753 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
754 {
755 	struct rt_msghdr rtm;
756 	struct mbuf *m;
757 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
758 
759 	if (route_cb.any_count == 0)
760 		return;
761 	memset(&rtm, 0, sizeof(rtm));
762 	rtm.rtm_flags = RTF_DONE | flags;
763 	rtm.rtm_errno = error;
764 	m = rt_msg1(type, rtinfo, &rtm, sizeof(rtm));
765 	if (m == NULL)
766 		return;
767 	mtod(m, struct rt_msghdr *)->rtm_addrs = rtinfo->rti_addrs;
768 	route_enqueue(m, sa ? sa->sa_family : 0);
769 }
770 
771 /*
772  * This routine is called to generate a message from the routing
773  * socket indicating that the status of a network interface has changed.
774  */
775 void
776 rt_ifmsg(struct ifnet *ifp)
777 {
778 	struct if_msghdr ifm;
779 	struct mbuf *m;
780 	struct rt_addrinfo info;
781 
782 	if (route_cb.any_count == 0)
783 		return;
784 	(void)memset(&info, 0, sizeof(info));
785 	(void)memset(&ifm, 0, sizeof(ifm));
786 	ifm.ifm_index = ifp->if_index;
787 	ifm.ifm_flags = ifp->if_flags;
788 	ifm.ifm_data = ifp->if_data;
789 	ifm.ifm_addrs = 0;
790 	m = rt_msg1(RTM_IFINFO, &info, &ifm, sizeof(ifm));
791 	if (m == NULL)
792 		return;
793 	route_enqueue(m, 0);
794 #ifdef COMPAT_14
795 	compat_14_rt_ifmsg(ifp, &ifm);
796 #endif
797 #ifdef COMPAT_50
798 	compat_50_rt_ifmsg(ifp, &ifm);
799 #endif
800 }
801 
802 
803 /*
804  * This is called to generate messages from the routing socket
805  * indicating a network interface has had addresses associated with it.
806  * if we ever reverse the logic and replace messages TO the routing
807  * socket indicate a request to configure interfaces, then it will
808  * be unnecessary as the routing socket will automatically generate
809  * copies of it.
810  */
811 void
812 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
813 {
814 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
815 	struct rt_addrinfo info;
816 	const struct sockaddr *sa;
817 	int pass;
818 	struct mbuf *m;
819 	struct ifnet *ifp = ifa->ifa_ifp;
820 	struct rt_msghdr rtm;
821 	struct ifa_msghdr ifam;
822 	int ncmd;
823 
824 	if (route_cb.any_count == 0)
825 		return;
826 	for (pass = 1; pass < 3; pass++) {
827 		memset(&info, 0, sizeof(info));
828 		switch (cmdpass(cmd, pass)) {
829 		case cmdpass(RTM_ADD, 1):
830 		case cmdpass(RTM_CHANGE, 1):
831 		case cmdpass(RTM_DELETE, 2):
832 			switch (cmd) {
833 			case RTM_DELETE:
834 				ncmd = RTM_DELADDR;
835 				break;
836 			case RTM_CHANGE:
837 				ncmd = RTM_CHGADDR;
838 				break;
839 			default:
840 				ncmd = RTM_NEWADDR;
841 			}
842 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
843 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
844 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
845 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
846 			memset(&ifam, 0, sizeof(ifam));
847 			ifam.ifam_index = ifp->if_index;
848 			ifam.ifam_metric = ifa->ifa_metric;
849 			ifam.ifam_flags = ifa->ifa_flags;
850 			m = rt_msg1(ncmd, &info, &ifam, sizeof(ifam));
851 			if (m == NULL)
852 				continue;
853 			mtod(m, struct ifa_msghdr *)->ifam_addrs =
854 			    info.rti_addrs;
855 			break;
856 		case cmdpass(RTM_ADD, 2):
857 		case cmdpass(RTM_CHANGE, 2):
858 		case cmdpass(RTM_DELETE, 1):
859 			if (rt == NULL)
860 				continue;
861 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
862 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
863 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
864 			memset(&rtm, 0, sizeof(rtm));
865 			rtm.rtm_index = ifp->if_index;
866 			rtm.rtm_flags |= rt->rt_flags;
867 			rtm.rtm_errno = error;
868 			m = rt_msg1(cmd, &info, &rtm, sizeof(rtm));
869 			if (m == NULL)
870 				continue;
871 			mtod(m, struct rt_msghdr *)->rtm_addrs = info.rti_addrs;
872 			break;
873 		default:
874 			continue;
875 		}
876 #ifdef DIAGNOSTIC
877 		if (m == NULL)
878 			panic("%s: called with wrong command", __func__);
879 #endif
880 		route_enqueue(m, sa ? sa->sa_family : 0);
881 	}
882 #undef cmdpass
883 }
884 
885 static struct mbuf *
886 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
887     struct rt_addrinfo *info)
888 {
889 	struct if_announcemsghdr ifan;
890 
891 	memset(info, 0, sizeof(*info));
892 	memset(&ifan, 0, sizeof(ifan));
893 	ifan.ifan_index = ifp->if_index;
894 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
895 	ifan.ifan_what = what;
896 	return rt_msg1(type, info, &ifan, sizeof(ifan));
897 }
898 
899 /*
900  * This is called to generate routing socket messages indicating
901  * network interface arrival and departure.
902  */
903 void
904 rt_ifannouncemsg(struct ifnet *ifp, int what)
905 {
906 	struct mbuf *m;
907 	struct rt_addrinfo info;
908 
909 	if (route_cb.any_count == 0)
910 		return;
911 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
912 	if (m == NULL)
913 		return;
914 	route_enqueue(m, 0);
915 }
916 
917 /*
918  * This is called to generate routing socket messages indicating
919  * IEEE80211 wireless events.
920  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
921  */
922 void
923 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
924 {
925 	struct mbuf *m;
926 	struct rt_addrinfo info;
927 
928 	if (route_cb.any_count == 0)
929 		return;
930 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
931 	if (m == NULL)
932 		return;
933 	/*
934 	 * Append the ieee80211 data.  Try to stick it in the
935 	 * mbuf containing the ifannounce msg; otherwise allocate
936 	 * a new mbuf and append.
937 	 *
938 	 * NB: we assume m is a single mbuf.
939 	 */
940 	if (data_len > M_TRAILINGSPACE(m)) {
941 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
942 		if (n == NULL) {
943 			m_freem(m);
944 			return;
945 		}
946 		(void)memcpy(mtod(n, void *), data, data_len);
947 		n->m_len = data_len;
948 		m->m_next = n;
949 	} else if (data_len > 0) {
950 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
951 		m->m_len += data_len;
952 	}
953 	if (m->m_flags & M_PKTHDR)
954 		m->m_pkthdr.len += data_len;
955 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
956 	route_enqueue(m, 0);
957 }
958 
959 /*
960  * This is used in dumping the kernel table via sysctl().
961  */
962 static int
963 sysctl_dumpentry(struct rtentry *rt, void *v)
964 {
965 	struct rt_walkarg *w = v;
966 	int error = 0, size;
967 	struct rt_addrinfo info;
968 
969 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
970 		return 0;
971 	memset(&info, 0, sizeof(info));
972 	info.rti_info[RTAX_DST] = rt_getkey(rt);
973 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
974 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
975 	if (rt->rt_ifp) {
976 		const struct ifaddr *rtifa;
977 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
978 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
979 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
980 		 * --dyoung
981 		 */
982 		rtifa = rt_get_ifa(rt);
983 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
984 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
985 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
986 	}
987 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
988 		return error;
989 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
990 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
991 
992 		rtm->rtm_flags = rt->rt_flags;
993 		rtm->rtm_use = rt->rt_use;
994 		cvtmetrics(&rtm->rtm_rmx, &rt->rt_rmx);
995 		KASSERT(rt->rt_ifp != NULL);
996 		rtm->rtm_index = rt->rt_ifp->if_index;
997 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
998 		rtm->rtm_addrs = info.rti_addrs;
999 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1000 			w->w_where = NULL;
1001 		else
1002 			w->w_where = (char *)w->w_where + size;
1003 	}
1004 	return error;
1005 }
1006 
1007 static int
1008 sysctl_iflist(int af, struct rt_walkarg *w, int type)
1009 {
1010 	struct ifnet *ifp;
1011 	struct ifaddr *ifa;
1012 	struct	rt_addrinfo info;
1013 	int	len, error = 0;
1014 
1015 	memset(&info, 0, sizeof(info));
1016 	IFNET_FOREACH(ifp) {
1017 		if (w->w_arg && w->w_arg != ifp->if_index)
1018 			continue;
1019 		if (IFADDR_EMPTY(ifp))
1020 			continue;
1021 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1022 		switch (type) {
1023 		case NET_RT_IFLIST:
1024 			error = rt_msg2(RTM_IFINFO, &info, NULL, w, &len);
1025 			break;
1026 #ifdef COMPAT_14
1027 		case NET_RT_OOIFLIST:
1028 			error = rt_msg2(RTM_OOIFINFO, &info, NULL, w, &len);
1029 			break;
1030 #endif
1031 #ifdef COMPAT_50
1032 		case NET_RT_OIFLIST:
1033 			error = rt_msg2(RTM_OIFINFO, &info, NULL, w, &len);
1034 			break;
1035 #endif
1036 		default:
1037 			panic("sysctl_iflist(1)");
1038 		}
1039 		if (error)
1040 			return error;
1041 		info.rti_info[RTAX_IFP] = NULL;
1042 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1043 			switch (type) {
1044 			case NET_RT_IFLIST: {
1045 				struct if_msghdr *ifm;
1046 
1047 				ifm = (struct if_msghdr *)w->w_tmem;
1048 				ifm->ifm_index = ifp->if_index;
1049 				ifm->ifm_flags = ifp->if_flags;
1050 				ifm->ifm_data = ifp->if_data;
1051 				ifm->ifm_addrs = info.rti_addrs;
1052 				error = copyout(ifm, w->w_where, len);
1053 				if (error)
1054 					return error;
1055 				w->w_where = (char *)w->w_where + len;
1056 				break;
1057 			}
1058 
1059 #ifdef COMPAT_14
1060 			case NET_RT_OOIFLIST:
1061 				error = compat_14_iflist(ifp, w, &info, len);
1062 				if (error)
1063 					return error;
1064 				break;
1065 #endif
1066 #ifdef COMPAT_50
1067 			case NET_RT_OIFLIST:
1068 				error = compat_50_iflist(ifp, w, &info, len);
1069 				if (error)
1070 					return error;
1071 				break;
1072 #endif
1073 			default:
1074 				panic("sysctl_iflist(2)");
1075 			}
1076 		}
1077 		IFADDR_FOREACH(ifa, ifp) {
1078 			if (af && af != ifa->ifa_addr->sa_family)
1079 				continue;
1080 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1081 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1082 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1083 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
1084 				return error;
1085 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1086 				struct ifa_msghdr *ifam;
1087 
1088 				ifam = (struct ifa_msghdr *)w->w_tmem;
1089 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1090 				ifam->ifam_flags = ifa->ifa_flags;
1091 				ifam->ifam_metric = ifa->ifa_metric;
1092 				ifam->ifam_addrs = info.rti_addrs;
1093 				error = copyout(w->w_tmem, w->w_where, len);
1094 				if (error)
1095 					return error;
1096 				w->w_where = (char *)w->w_where + len;
1097 			}
1098 		}
1099 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1100 		    info.rti_info[RTAX_BRD] = NULL;
1101 	}
1102 	return 0;
1103 }
1104 
1105 static int
1106 sysctl_rtable(SYSCTLFN_ARGS)
1107 {
1108 	void 	*where = oldp;
1109 	size_t	*given = oldlenp;
1110 	const void *new = newp;
1111 	int	i, s, error = EINVAL;
1112 	u_char  af;
1113 	struct	rt_walkarg w;
1114 
1115 	if (namelen == 1 && name[0] == CTL_QUERY)
1116 		return sysctl_query(SYSCTLFN_CALL(rnode));
1117 
1118 	if (new)
1119 		return EPERM;
1120 	if (namelen != 3)
1121 		return EINVAL;
1122 	af = name[0];
1123 	w.w_tmemneeded = 0;
1124 	w.w_tmemsize = 0;
1125 	w.w_tmem = NULL;
1126 again:
1127 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
1128 	if (w.w_tmemneeded) {
1129 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
1130 		w.w_tmemsize = w.w_tmemneeded;
1131 		w.w_tmemneeded = 0;
1132 	}
1133 	w.w_op = name[1];
1134 	w.w_arg = name[2];
1135 	w.w_given = *given;
1136 	w.w_needed = 0 - w.w_given;
1137 	w.w_where = where;
1138 
1139 	s = splsoftnet();
1140 	switch (w.w_op) {
1141 
1142 	case NET_RT_DUMP:
1143 	case NET_RT_FLAGS:
1144 		for (i = 1; i <= AF_MAX; i++)
1145 			if ((af == 0 || af == i) &&
1146 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
1147 				break;
1148 		break;
1149 
1150 #ifdef COMPAT_14
1151 	case NET_RT_OOIFLIST:
1152 		error = sysctl_iflist(af, &w, w.w_op);
1153 		break;
1154 #endif
1155 #ifdef COMPAT_50
1156 	case NET_RT_OIFLIST:
1157 		error = sysctl_iflist(af, &w, w.w_op);
1158 		break;
1159 #endif
1160 
1161 	case NET_RT_IFLIST:
1162 		error = sysctl_iflist(af, &w, w.w_op);
1163 	}
1164 	splx(s);
1165 
1166 	/* check to see if we couldn't allocate memory with NOWAIT */
1167 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
1168 		goto again;
1169 
1170 	if (w.w_tmem)
1171 		free(w.w_tmem, M_RTABLE);
1172 	w.w_needed += w.w_given;
1173 	if (where) {
1174 		*given = (char *)w.w_where - (char *)where;
1175 		if (*given < w.w_needed)
1176 			return ENOMEM;
1177 	} else {
1178 		*given = (11 * w.w_needed) / 10;
1179 	}
1180 	return error;
1181 }
1182 
1183 /*
1184  * Routing message software interrupt routine
1185  */
1186 static void
1187 route_intr(void *cookie)
1188 {
1189 	struct sockproto proto = { .sp_family = PF_ROUTE, };
1190 	struct mbuf *m;
1191 	int s;
1192 
1193 	mutex_enter(softnet_lock);
1194 	KERNEL_LOCK(1, NULL);
1195 	while (!IF_IS_EMPTY(&route_intrq)) {
1196 		s = splnet();
1197 		IF_DEQUEUE(&route_intrq, m);
1198 		splx(s);
1199 		if (m == NULL)
1200 			break;
1201 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
1202 		raw_input(m, &proto, &route_src, &route_dst);
1203 	}
1204 	KERNEL_UNLOCK_ONE(NULL);
1205 	mutex_exit(softnet_lock);
1206 }
1207 
1208 /*
1209  * Enqueue a message to the software interrupt routine.
1210  */
1211 void
1212 route_enqueue(struct mbuf *m, int family)
1213 {
1214 	int s, wasempty;
1215 
1216 	s = splnet();
1217 	if (IF_QFULL(&route_intrq)) {
1218 		IF_DROP(&route_intrq);
1219 		m_freem(m);
1220 	} else {
1221 		wasempty = IF_IS_EMPTY(&route_intrq);
1222 		M_SETCTX(m, (uintptr_t)family);
1223 		IF_ENQUEUE(&route_intrq, m);
1224 		if (wasempty)
1225 			softint_schedule(route_sih);
1226 	}
1227 	splx(s);
1228 }
1229 
1230 void
1231 rt_init(void)
1232 {
1233 
1234 	sysctl_net_route_setup(NULL);
1235 	route_intrq.ifq_maxlen = route_maxqlen;
1236 	route_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
1237 	    route_intr, NULL);
1238 }
1239 
1240 /*
1241  * Definitions of protocols supported in the ROUTE domain.
1242  */
1243 PR_WRAP_USRREQ(route_usrreq)
1244 #define	route_usrreq	route_usrreq_wrapper
1245 
1246 const struct protosw routesw[] = {
1247 	{
1248 		.pr_type = SOCK_RAW,
1249 		.pr_domain = &routedomain,
1250 		.pr_flags = PR_ATOMIC|PR_ADDR,
1251 		.pr_input = raw_input,
1252 		.pr_output = route_output,
1253 		.pr_ctlinput = raw_ctlinput,
1254 		.pr_usrreq = route_usrreq,
1255 		.pr_init = raw_init,
1256 	},
1257 };
1258 
1259 struct domain routedomain = {
1260 	.dom_family = PF_ROUTE,
1261 	.dom_name = "route",
1262 	.dom_init = route_init,
1263 	.dom_protosw = routesw,
1264 	.dom_protoswNPROTOSW = &routesw[__arraycount(routesw)],
1265 };
1266 
1267 static void
1268 sysctl_net_route_setup(struct sysctllog **clog)
1269 {
1270 	const struct sysctlnode *rnode = NULL;
1271 
1272 	sysctl_createv(clog, 0, NULL, NULL,
1273 		       CTLFLAG_PERMANENT,
1274 		       CTLTYPE_NODE, "net", NULL,
1275 		       NULL, 0, NULL, 0,
1276 		       CTL_NET, CTL_EOL);
1277 
1278 	sysctl_createv(clog, 0, NULL, &rnode,
1279 		       CTLFLAG_PERMANENT,
1280 		       CTLTYPE_NODE, "route",
1281 		       SYSCTL_DESCR("PF_ROUTE information"),
1282 		       NULL, 0, NULL, 0,
1283 		       CTL_NET, PF_ROUTE, CTL_EOL);
1284 	sysctl_createv(clog, 0, NULL, NULL,
1285 		       CTLFLAG_PERMANENT,
1286 		       CTLTYPE_NODE, "rtable",
1287 		       SYSCTL_DESCR("Routing table information"),
1288 		       sysctl_rtable, 0, NULL, 0,
1289 		       CTL_NET, PF_ROUTE, 0 /* any protocol */, CTL_EOL);
1290 	sysctl_createv(clog, 0, &rnode, NULL,
1291 		       CTLFLAG_PERMANENT,
1292 		       CTLTYPE_STRUCT, "stats",
1293 		       SYSCTL_DESCR("Routing statistics"),
1294 		       NULL, 0, &rtstat, sizeof(rtstat),
1295 		       CTL_CREATE, CTL_EOL);
1296 }
1297