xref: /freebsd/sys/net/rtsock.c (revision 817420dc8eac7df799c78f5309b75092b7f7cd40)
1 /*
2  * Copyright (c) 1988, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)rtsock.c	8.5 (Berkeley) 11/2/94
34  * $FreeBSD$
35  */
36 
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/kernel.h>
41 #include <sys/sysctl.h>
42 #include <sys/proc.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <sys/domain.h>
48 #include <sys/protosw.h>
49 
50 #include <net/if.h>
51 #include <net/route.h>
52 #include <net/raw_cb.h>
53 
54 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
55 
56 static struct	sockaddr route_dst = { 2, PF_ROUTE, };
57 static struct	sockaddr route_src = { 2, PF_ROUTE, };
58 static struct	sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
59 static struct	sockproto route_proto = { PF_ROUTE, };
60 
61 struct walkarg {
62 	int	w_tmemsize;
63 	int	w_op, w_arg;
64 	caddr_t	w_tmem;
65 	struct sysctl_req *w_req;
66 };
67 
68 static struct mbuf *
69 		rt_msg1 __P((int, struct rt_addrinfo *));
70 static int	rt_msg2 __P((int,
71 		    struct rt_addrinfo *, caddr_t, struct walkarg *));
72 static int	rt_xaddrs __P((caddr_t, caddr_t, struct rt_addrinfo *));
73 static int	sysctl_dumpentry __P((struct radix_node *rn, void *vw));
74 static int	sysctl_iflist __P((int af, struct walkarg *w));
75 static int	 route_output __P((struct mbuf *, struct socket *));
76 static void	 rt_setmetrics __P((u_long, struct rt_metrics *, struct rt_metrics *));
77 
78 /* Sleazy use of local variables throughout file, warning!!!! */
79 #define dst	info.rti_info[RTAX_DST]
80 #define gate	info.rti_info[RTAX_GATEWAY]
81 #define netmask	info.rti_info[RTAX_NETMASK]
82 #define genmask	info.rti_info[RTAX_GENMASK]
83 #define ifpaddr	info.rti_info[RTAX_IFP]
84 #define ifaaddr	info.rti_info[RTAX_IFA]
85 #define brdaddr	info.rti_info[RTAX_BRD]
86 
87 /*
88  * It really doesn't make any sense at all for this code to share much
89  * with raw_usrreq.c, since its functionality is so restricted.  XXX
90  */
91 static int
92 rts_abort(struct socket *so)
93 {
94 	int s, error;
95 	s = splnet();
96 	error = raw_usrreqs.pru_abort(so);
97 	splx(s);
98 	return error;
99 }
100 
101 /* pru_accept is EOPNOTSUPP */
102 
103 static int
104 rts_attach(struct socket *so, int proto, struct proc *p)
105 {
106 	struct rawcb *rp;
107 	int s, error;
108 
109 	if (sotorawcb(so) != 0)
110 		return EISCONN;	/* XXX panic? */
111 	MALLOC(rp, struct rawcb *, sizeof *rp, M_PCB, M_WAITOK); /* XXX */
112 	if (rp == 0)
113 		return ENOBUFS;
114 	bzero(rp, sizeof *rp);
115 
116 	/*
117 	 * The splnet() is necessary to block protocols from sending
118 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
119 	 * this PCB is extant but incompletely initialized.
120 	 * Probably we should try to do more of this work beforehand and
121 	 * eliminate the spl.
122 	 */
123 	s = splnet();
124 	so->so_pcb = (caddr_t)rp;
125 	error = raw_usrreqs.pru_attach(so, proto, p);
126 	rp = sotorawcb(so);
127 	if (error) {
128 		splx(s);
129 		free(rp, M_PCB);
130 		return error;
131 	}
132 	switch(rp->rcb_proto.sp_protocol) {
133 	case AF_INET:
134 		route_cb.ip_count++;
135 		break;
136 	case AF_INET6:
137 		route_cb.ip6_count++;
138 		break;
139 	case AF_IPX:
140 		route_cb.ipx_count++;
141 		break;
142 	case AF_NS:
143 		route_cb.ns_count++;
144 		break;
145 	}
146 	rp->rcb_faddr = &route_src;
147 	route_cb.any_count++;
148 	soisconnected(so);
149 	so->so_options |= SO_USELOOPBACK;
150 	splx(s);
151 	return 0;
152 }
153 
154 static int
155 rts_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
156 {
157 	int s, error;
158 	s = splnet();
159 	error = raw_usrreqs.pru_bind(so, nam, p); /* xxx just EINVAL */
160 	splx(s);
161 	return error;
162 }
163 
164 static int
165 rts_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
166 {
167 	int s, error;
168 	s = splnet();
169 	error = raw_usrreqs.pru_connect(so, nam, p); /* XXX just EINVAL */
170 	splx(s);
171 	return error;
172 }
173 
174 /* pru_connect2 is EOPNOTSUPP */
175 /* pru_control is EOPNOTSUPP */
176 
177 static int
178 rts_detach(struct socket *so)
179 {
180 	struct rawcb *rp = sotorawcb(so);
181 	int s, error;
182 
183 	s = splnet();
184 	if (rp != 0) {
185 		switch(rp->rcb_proto.sp_protocol) {
186 		case AF_INET:
187 			route_cb.ip_count--;
188 			break;
189 		case AF_INET6:
190 			route_cb.ip6_count--;
191 			break;
192 		case AF_IPX:
193 			route_cb.ipx_count--;
194 			break;
195 		case AF_NS:
196 			route_cb.ns_count--;
197 			break;
198 		}
199 		route_cb.any_count--;
200 	}
201 	error = raw_usrreqs.pru_detach(so);
202 	splx(s);
203 	return error;
204 }
205 
206 static int
207 rts_disconnect(struct socket *so)
208 {
209 	int s, error;
210 	s = splnet();
211 	error = raw_usrreqs.pru_disconnect(so);
212 	splx(s);
213 	return error;
214 }
215 
216 /* pru_listen is EOPNOTSUPP */
217 
218 static int
219 rts_peeraddr(struct socket *so, struct sockaddr **nam)
220 {
221 	int s, error;
222 	s = splnet();
223 	error = raw_usrreqs.pru_peeraddr(so, nam);
224 	splx(s);
225 	return error;
226 }
227 
228 /* pru_rcvd is EOPNOTSUPP */
229 /* pru_rcvoob is EOPNOTSUPP */
230 
231 static int
232 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
233 	 struct mbuf *control, struct proc *p)
234 {
235 	int s, error;
236 	s = splnet();
237 	error = raw_usrreqs.pru_send(so, flags, m, nam, control, p);
238 	splx(s);
239 	return error;
240 }
241 
242 /* pru_sense is null */
243 
244 static int
245 rts_shutdown(struct socket *so)
246 {
247 	int s, error;
248 	s = splnet();
249 	error = raw_usrreqs.pru_shutdown(so);
250 	splx(s);
251 	return error;
252 }
253 
254 static int
255 rts_sockaddr(struct socket *so, struct sockaddr **nam)
256 {
257 	int s, error;
258 	s = splnet();
259 	error = raw_usrreqs.pru_sockaddr(so, nam);
260 	splx(s);
261 	return error;
262 }
263 
264 static struct pr_usrreqs route_usrreqs = {
265 	rts_abort, pru_accept_notsupp, rts_attach, rts_bind, rts_connect,
266 	pru_connect2_notsupp, pru_control_notsupp, rts_detach, rts_disconnect,
267 	pru_listen_notsupp, rts_peeraddr, pru_rcvd_notsupp, pru_rcvoob_notsupp,
268 	rts_send, pru_sense_null, rts_shutdown, rts_sockaddr,
269 	sosend, soreceive, sopoll
270 };
271 
272 /*ARGSUSED*/
273 static int
274 route_output(m, so)
275 	register struct mbuf *m;
276 	struct socket *so;
277 {
278 	register struct rt_msghdr *rtm = 0;
279 	register struct rtentry *rt = 0;
280 	struct rtentry *saved_nrt = 0;
281 	struct radix_node_head *rnh;
282 	struct rt_addrinfo info;
283 	int len, error = 0;
284 	struct ifnet *ifp = 0;
285 	struct ifaddr *ifa = 0;
286 
287 #define senderr(e) { error = e; goto flush;}
288 	if (m == 0 || ((m->m_len < sizeof(long)) &&
289 		       (m = m_pullup(m, sizeof(long))) == 0))
290 		return (ENOBUFS);
291 	if ((m->m_flags & M_PKTHDR) == 0)
292 		panic("route_output");
293 	len = m->m_pkthdr.len;
294 	if (len < sizeof(*rtm) ||
295 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
296 		dst = 0;
297 		senderr(EINVAL);
298 	}
299 	R_Malloc(rtm, struct rt_msghdr *, len);
300 	if (rtm == 0) {
301 		dst = 0;
302 		senderr(ENOBUFS);
303 	}
304 	m_copydata(m, 0, len, (caddr_t)rtm);
305 	if (rtm->rtm_version != RTM_VERSION) {
306 		dst = 0;
307 		senderr(EPROTONOSUPPORT);
308 	}
309 	rtm->rtm_pid = curproc->p_pid;
310 	info.rti_addrs = rtm->rtm_addrs;
311 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
312 		dst = 0;
313 		senderr(EINVAL);
314 	}
315 	if (dst == 0 || (dst->sa_family >= AF_MAX)
316 	    || (gate != 0 && (gate->sa_family >= AF_MAX)))
317 		senderr(EINVAL);
318 	if (genmask) {
319 		struct radix_node *t;
320 		t = rn_addmask((caddr_t)genmask, 0, 1);
321 		if (t && Bcmp((caddr_t *)genmask + 1, (caddr_t *)t->rn_key + 1,
322 			      *(u_char *)t->rn_key - 1) == 0)
323 			genmask = (struct sockaddr *)(t->rn_key);
324 		else
325 			senderr(ENOBUFS);
326 	}
327 	switch (rtm->rtm_type) {
328 
329 	case RTM_ADD:
330 		if (gate == 0)
331 			senderr(EINVAL);
332 		error = rtrequest(RTM_ADD, dst, gate, netmask,
333 					rtm->rtm_flags, &saved_nrt);
334 		if (error == 0 && saved_nrt) {
335 			rt_setmetrics(rtm->rtm_inits,
336 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
337 			saved_nrt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
338 			saved_nrt->rt_rmx.rmx_locks |=
339 				(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
340 			saved_nrt->rt_refcnt--;
341 			saved_nrt->rt_genmask = genmask;
342 		}
343 		break;
344 
345 	case RTM_DELETE:
346 		error = rtrequest(RTM_DELETE, dst, gate, netmask,
347 				rtm->rtm_flags, &saved_nrt);
348 		if (error == 0) {
349 			if ((rt = saved_nrt))
350 				rt->rt_refcnt++;
351 			goto report;
352 		}
353 		break;
354 
355 	case RTM_GET:
356 	case RTM_CHANGE:
357 	case RTM_LOCK:
358 		if ((rnh = rt_tables[dst->sa_family]) == 0) {
359 			senderr(EAFNOSUPPORT);
360 		} else if ((rt = (struct rtentry *)
361 				rnh->rnh_lookup(dst, netmask, rnh)) != NULL)
362 			rt->rt_refcnt++;
363 		else
364 			senderr(ESRCH);
365 		switch(rtm->rtm_type) {
366 
367 		case RTM_GET:
368 		report:
369 			dst = rt_key(rt);
370 			gate = rt->rt_gateway;
371 			netmask = rt_mask(rt);
372 			genmask = rt->rt_genmask;
373 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
374 				ifp = rt->rt_ifp;
375 				if (ifp) {
376 					ifpaddr = ifp->if_addrhead.tqh_first->ifa_addr;
377 					ifaaddr = rt->rt_ifa->ifa_addr;
378 					rtm->rtm_index = ifp->if_index;
379 				} else {
380 					ifpaddr = 0;
381 					ifaaddr = 0;
382 			    }
383 			}
384 			len = rt_msg2(rtm->rtm_type, &info, (caddr_t)0,
385 				(struct walkarg *)0);
386 			if (len > rtm->rtm_msglen) {
387 				struct rt_msghdr *new_rtm;
388 				R_Malloc(new_rtm, struct rt_msghdr *, len);
389 				if (new_rtm == 0)
390 					senderr(ENOBUFS);
391 				Bcopy(rtm, new_rtm, rtm->rtm_msglen);
392 				Free(rtm); rtm = new_rtm;
393 			}
394 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm,
395 				(struct walkarg *)0);
396 			rtm->rtm_flags = rt->rt_flags;
397 			rtm->rtm_rmx = rt->rt_rmx;
398 			rtm->rtm_addrs = info.rti_addrs;
399 			break;
400 
401 		case RTM_CHANGE:
402 			if (gate && (error = rt_setgate(rt, rt_key(rt), gate)))
403 				senderr(error);
404 
405 			/*
406 			 * If they tried to change things but didn't specify
407 			 * the required gateway, then just use the old one.
408 			 * This can happen if the user tries to change the
409 			 * flags on the default route without changing the
410 			 * default gateway.  Changing flags still doesn't work.
411 			 */
412 			if ((rt->rt_flags & RTF_GATEWAY) && !gate)
413 				gate = rt->rt_gateway;
414 
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 (ifpaddr && (ifa = ifa_ifwithnet(ifpaddr)) &&
419 			    (ifp = ifa->ifa_ifp) && (ifaaddr || gate))
420 				ifa = ifaof_ifpforaddr(ifaaddr ? ifaaddr : gate,
421 							ifp);
422 			else if ((ifaaddr && (ifa = ifa_ifwithaddr(ifaaddr))) ||
423 				 (gate && (ifa = ifa_ifwithroute(rt->rt_flags,
424 							rt_key(rt), gate))))
425 				ifp = ifa->ifa_ifp;
426 			if (ifa) {
427 				register struct ifaddr *oifa = rt->rt_ifa;
428 				if (oifa != ifa) {
429 				    if (oifa && oifa->ifa_rtrequest)
430 					oifa->ifa_rtrequest(RTM_DELETE,
431 								rt, gate);
432 				    IFAFREE(rt->rt_ifa);
433 				    rt->rt_ifa = ifa;
434 				    ifa->ifa_refcnt++;
435 				    rt->rt_ifp = ifp;
436 				}
437 			}
438 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
439 					&rt->rt_rmx);
440 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
441 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, gate);
442 			if (genmask)
443 				rt->rt_genmask = genmask;
444 			/*
445 			 * Fall into
446 			 */
447 		case RTM_LOCK:
448 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
449 			rt->rt_rmx.rmx_locks |=
450 				(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
451 			break;
452 		}
453 		break;
454 
455 	default:
456 		senderr(EOPNOTSUPP);
457 	}
458 
459 flush:
460 	if (rtm) {
461 		if (error)
462 			rtm->rtm_errno = error;
463 		else
464 			rtm->rtm_flags |= RTF_DONE;
465 	}
466 	if (rt)
467 		rtfree(rt);
468     {
469 	register struct rawcb *rp = 0;
470 	/*
471 	 * Check to see if we don't want our own messages.
472 	 */
473 	if ((so->so_options & SO_USELOOPBACK) == 0) {
474 		if (route_cb.any_count <= 1) {
475 			if (rtm)
476 				Free(rtm);
477 			m_freem(m);
478 			return (error);
479 		}
480 		/* There is another listener, so construct message */
481 		rp = sotorawcb(so);
482 	}
483 	if (rtm) {
484 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
485 		Free(rtm);
486 	}
487 	if (rp)
488 		rp->rcb_proto.sp_family = 0; /* Avoid us */
489 	if (dst)
490 		route_proto.sp_protocol = dst->sa_family;
491 	raw_input(m, &route_proto, &route_src, &route_dst);
492 	if (rp)
493 		rp->rcb_proto.sp_family = PF_ROUTE;
494     }
495 	return (error);
496 }
497 
498 static void
499 rt_setmetrics(which, in, out)
500 	u_long which;
501 	register struct rt_metrics *in, *out;
502 {
503 #define metric(f, e) if (which & (f)) out->e = in->e;
504 	metric(RTV_RPIPE, rmx_recvpipe);
505 	metric(RTV_SPIPE, rmx_sendpipe);
506 	metric(RTV_SSTHRESH, rmx_ssthresh);
507 	metric(RTV_RTT, rmx_rtt);
508 	metric(RTV_RTTVAR, rmx_rttvar);
509 	metric(RTV_HOPCOUNT, rmx_hopcount);
510 	metric(RTV_MTU, rmx_mtu);
511 	metric(RTV_EXPIRE, rmx_expire);
512 #undef metric
513 }
514 
515 #define ROUNDUP(a) \
516 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
517 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
518 
519 
520 /*
521  * Extract the addresses of the passed sockaddrs.
522  * Do a little sanity checking so as to avoid bad memory references.
523  * This data is derived straight from userland.
524  */
525 static int
526 rt_xaddrs(cp, cplim, rtinfo)
527 	register caddr_t cp, cplim;
528 	register struct rt_addrinfo *rtinfo;
529 {
530 	register struct sockaddr *sa;
531 	register int i;
532 
533 	bzero(rtinfo->rti_info, sizeof(rtinfo->rti_info));
534 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
535 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
536 			continue;
537 		sa = (struct sockaddr *)cp;
538 		/*
539 		 * It won't fit.
540 		 */
541 		if ( (cp + sa->sa_len) > cplim ) {
542 			return (EINVAL);
543 		}
544 
545 		/*
546 		 * there are no more.. quit now
547 		 * If there are more bits, they are in error.
548 		 * I've seen this. route(1) can evidently generate these.
549 		 * This causes kernel to core dump.
550 		 * for compatibility, If we see this, point to a safe address.
551 		 */
552 		if (sa->sa_len == 0) {
553 			rtinfo->rti_info[i] = &sa_zero;
554 			return (0); /* should be EINVAL but for compat */
555 		}
556 
557 		/* accept it */
558 		rtinfo->rti_info[i] = sa;
559 		ADVANCE(cp, sa);
560 	}
561 	return (0);
562 }
563 
564 static struct mbuf *
565 rt_msg1(type, rtinfo)
566 	int type;
567 	register struct rt_addrinfo *rtinfo;
568 {
569 	register struct rt_msghdr *rtm;
570 	register struct mbuf *m;
571 	register int i;
572 	register struct sockaddr *sa;
573 	int len, dlen;
574 
575 	m = m_gethdr(M_DONTWAIT, MT_DATA);
576 	if (m == 0)
577 		return (m);
578 	switch (type) {
579 
580 	case RTM_DELADDR:
581 	case RTM_NEWADDR:
582 		len = sizeof(struct ifa_msghdr);
583 		break;
584 
585 	case RTM_DELMADDR:
586 	case RTM_NEWMADDR:
587 		len = sizeof(struct ifma_msghdr);
588 		break;
589 
590 	case RTM_IFINFO:
591 		len = sizeof(struct if_msghdr);
592 		break;
593 
594 	default:
595 		len = sizeof(struct rt_msghdr);
596 	}
597 	if (len > MHLEN)
598 		panic("rt_msg1");
599 	m->m_pkthdr.len = m->m_len = len;
600 	m->m_pkthdr.rcvif = 0;
601 	rtm = mtod(m, struct rt_msghdr *);
602 	bzero((caddr_t)rtm, len);
603 	for (i = 0; i < RTAX_MAX; i++) {
604 		if ((sa = rtinfo->rti_info[i]) == NULL)
605 			continue;
606 		rtinfo->rti_addrs |= (1 << i);
607 		dlen = ROUNDUP(sa->sa_len);
608 		m_copyback(m, len, dlen, (caddr_t)sa);
609 		len += dlen;
610 	}
611 	if (m->m_pkthdr.len != len) {
612 		m_freem(m);
613 		return (NULL);
614 	}
615 	rtm->rtm_msglen = len;
616 	rtm->rtm_version = RTM_VERSION;
617 	rtm->rtm_type = type;
618 	return (m);
619 }
620 
621 static int
622 rt_msg2(type, rtinfo, cp, w)
623 	int type;
624 	register struct rt_addrinfo *rtinfo;
625 	caddr_t cp;
626 	struct walkarg *w;
627 {
628 	register int i;
629 	int len, dlen, second_time = 0;
630 	caddr_t cp0;
631 
632 	rtinfo->rti_addrs = 0;
633 again:
634 	switch (type) {
635 
636 	case RTM_DELADDR:
637 	case RTM_NEWADDR:
638 		len = sizeof(struct ifa_msghdr);
639 		break;
640 
641 	case RTM_IFINFO:
642 		len = sizeof(struct if_msghdr);
643 		break;
644 
645 	default:
646 		len = sizeof(struct rt_msghdr);
647 	}
648 	cp0 = cp;
649 	if (cp0)
650 		cp += len;
651 	for (i = 0; i < RTAX_MAX; i++) {
652 		register struct sockaddr *sa;
653 
654 		if ((sa = rtinfo->rti_info[i]) == 0)
655 			continue;
656 		rtinfo->rti_addrs |= (1 << i);
657 		dlen = ROUNDUP(sa->sa_len);
658 		if (cp) {
659 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
660 			cp += dlen;
661 		}
662 		len += dlen;
663 	}
664 	if (cp == 0 && w != NULL && !second_time) {
665 		register struct walkarg *rw = w;
666 
667 		if (rw->w_req) {
668 			if (rw->w_tmemsize < len) {
669 				if (rw->w_tmem)
670 					free(rw->w_tmem, M_RTABLE);
671 				rw->w_tmem = (caddr_t)
672 					malloc(len, M_RTABLE, M_NOWAIT);
673 				if (rw->w_tmem)
674 					rw->w_tmemsize = len;
675 			}
676 			if (rw->w_tmem) {
677 				cp = rw->w_tmem;
678 				second_time = 1;
679 				goto again;
680 			}
681 		}
682 	}
683 	if (cp) {
684 		register struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
685 
686 		rtm->rtm_version = RTM_VERSION;
687 		rtm->rtm_type = type;
688 		rtm->rtm_msglen = len;
689 	}
690 	return (len);
691 }
692 
693 /*
694  * This routine is called to generate a message from the routing
695  * socket indicating that a redirect has occured, a routing lookup
696  * has failed, or that a protocol has detected timeouts to a particular
697  * destination.
698  */
699 void
700 rt_missmsg(type, rtinfo, flags, error)
701 	int type, flags, error;
702 	register struct rt_addrinfo *rtinfo;
703 {
704 	register struct rt_msghdr *rtm;
705 	register struct mbuf *m;
706 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
707 
708 	if (route_cb.any_count == 0)
709 		return;
710 	m = rt_msg1(type, rtinfo);
711 	if (m == 0)
712 		return;
713 	rtm = mtod(m, struct rt_msghdr *);
714 	rtm->rtm_flags = RTF_DONE | flags;
715 	rtm->rtm_errno = error;
716 	rtm->rtm_addrs = rtinfo->rti_addrs;
717 	route_proto.sp_protocol = sa ? sa->sa_family : 0;
718 	raw_input(m, &route_proto, &route_src, &route_dst);
719 }
720 
721 /*
722  * This routine is called to generate a message from the routing
723  * socket indicating that the status of a network interface has changed.
724  */
725 void
726 rt_ifmsg(ifp)
727 	register struct ifnet *ifp;
728 {
729 	register struct if_msghdr *ifm;
730 	struct mbuf *m;
731 	struct rt_addrinfo info;
732 
733 	if (route_cb.any_count == 0)
734 		return;
735 	bzero((caddr_t)&info, sizeof(info));
736 	m = rt_msg1(RTM_IFINFO, &info);
737 	if (m == 0)
738 		return;
739 	ifm = mtod(m, struct if_msghdr *);
740 	ifm->ifm_index = ifp->if_index;
741 	ifm->ifm_flags = (u_short)ifp->if_flags;
742 	ifm->ifm_data = ifp->if_data;
743 	ifm->ifm_addrs = 0;
744 	route_proto.sp_protocol = 0;
745 	raw_input(m, &route_proto, &route_src, &route_dst);
746 }
747 
748 /*
749  * This is called to generate messages from the routing socket
750  * indicating a network interface has had addresses associated with it.
751  * if we ever reverse the logic and replace messages TO the routing
752  * socket indicate a request to configure interfaces, then it will
753  * be unnecessary as the routing socket will automatically generate
754  * copies of it.
755  */
756 void
757 rt_newaddrmsg(cmd, ifa, error, rt)
758 	int cmd, error;
759 	register struct ifaddr *ifa;
760 	register struct rtentry *rt;
761 {
762 	struct rt_addrinfo info;
763 	struct sockaddr *sa = 0;
764 	int pass;
765 	struct mbuf *m = 0;
766 	struct ifnet *ifp = ifa->ifa_ifp;
767 
768 	if (route_cb.any_count == 0)
769 		return;
770 	for (pass = 1; pass < 3; pass++) {
771 		bzero((caddr_t)&info, sizeof(info));
772 		if ((cmd == RTM_ADD && pass == 1) ||
773 		    (cmd == RTM_DELETE && pass == 2)) {
774 			register struct ifa_msghdr *ifam;
775 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
776 
777 			ifaaddr = sa = ifa->ifa_addr;
778 			ifpaddr = ifp->if_addrhead.tqh_first->ifa_addr;
779 			netmask = ifa->ifa_netmask;
780 			brdaddr = ifa->ifa_dstaddr;
781 			if ((m = rt_msg1(ncmd, &info)) == NULL)
782 				continue;
783 			ifam = mtod(m, struct ifa_msghdr *);
784 			ifam->ifam_index = ifp->if_index;
785 			ifam->ifam_metric = ifa->ifa_metric;
786 			ifam->ifam_flags = ifa->ifa_flags;
787 			ifam->ifam_addrs = info.rti_addrs;
788 		}
789 		if ((cmd == RTM_ADD && pass == 2) ||
790 		    (cmd == RTM_DELETE && pass == 1)) {
791 			register struct rt_msghdr *rtm;
792 
793 			if (rt == 0)
794 				continue;
795 			netmask = rt_mask(rt);
796 			dst = sa = rt_key(rt);
797 			gate = rt->rt_gateway;
798 			if ((m = rt_msg1(cmd, &info)) == NULL)
799 				continue;
800 			rtm = mtod(m, struct rt_msghdr *);
801 			rtm->rtm_index = ifp->if_index;
802 			rtm->rtm_flags |= rt->rt_flags;
803 			rtm->rtm_errno = error;
804 			rtm->rtm_addrs = info.rti_addrs;
805 		}
806 		route_proto.sp_protocol = sa ? sa->sa_family : 0;
807 		raw_input(m, &route_proto, &route_src, &route_dst);
808 	}
809 }
810 
811 /*
812  * This is the analogue to the rt_newaddrmsg which performs the same
813  * function but for multicast group memberhips.  This is easier since
814  * there is no route state to worry about.
815  */
816 void
817 rt_newmaddrmsg(cmd, ifma)
818 	int cmd;
819 	struct ifmultiaddr *ifma;
820 {
821 	struct rt_addrinfo info;
822 	struct mbuf *m = 0;
823 	struct ifnet *ifp = ifma->ifma_ifp;
824 	struct ifma_msghdr *ifmam;
825 
826 	if (route_cb.any_count == 0)
827 		return;
828 
829 	bzero((caddr_t)&info, sizeof(info));
830 	ifaaddr = ifma->ifma_addr;
831 	if (ifp && ifp->if_addrhead.tqh_first)
832 		ifpaddr = ifp->if_addrhead.tqh_first->ifa_addr;
833 	else
834 		ifpaddr = NULL;
835 	/*
836 	 * If a link-layer address is present, present it as a ``gateway''
837 	 * (similarly to how ARP entries, e.g., are presented).
838 	 */
839 	gate = ifma->ifma_lladdr;
840 	if ((m = rt_msg1(cmd, &info)) == NULL)
841 		return;
842 	ifmam = mtod(m, struct ifma_msghdr *);
843 	ifmam->ifmam_index = ifp->if_index;
844 	ifmam->ifmam_addrs = info.rti_addrs;
845 	route_proto.sp_protocol = ifma->ifma_addr->sa_family;
846 	raw_input(m, &route_proto, &route_src, &route_dst);
847 }
848 
849 /*
850  * This is used in dumping the kernel table via sysctl().
851  */
852 int
853 sysctl_dumpentry(rn, vw)
854 	struct radix_node *rn;
855 	void *vw;
856 {
857 	register struct walkarg *w = vw;
858 	register struct rtentry *rt = (struct rtentry *)rn;
859 	int error = 0, size;
860 	struct rt_addrinfo info;
861 
862 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
863 		return 0;
864 	bzero((caddr_t)&info, sizeof(info));
865 	dst = rt_key(rt);
866 	gate = rt->rt_gateway;
867 	netmask = rt_mask(rt);
868 	genmask = rt->rt_genmask;
869 	size = rt_msg2(RTM_GET, &info, 0, w);
870 	if (w->w_req && w->w_tmem) {
871 		register struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
872 
873 		rtm->rtm_flags = rt->rt_flags;
874 		rtm->rtm_use = rt->rt_use;
875 		rtm->rtm_rmx = rt->rt_rmx;
876 		rtm->rtm_index = rt->rt_ifp->if_index;
877 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
878 		rtm->rtm_addrs = info.rti_addrs;
879 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
880 		return (error);
881 	}
882 	return (error);
883 }
884 
885 int
886 sysctl_iflist(af, w)
887 	int	af;
888 	register struct	walkarg *w;
889 {
890 	register struct ifnet *ifp;
891 	register struct ifaddr *ifa;
892 	struct	rt_addrinfo info;
893 	int	len, error = 0;
894 
895 	bzero((caddr_t)&info, sizeof(info));
896 	for (ifp = ifnet.tqh_first; ifp; ifp = ifp->if_link.tqe_next) {
897 		if (w->w_arg && w->w_arg != ifp->if_index)
898 			continue;
899 		ifa = ifp->if_addrhead.tqh_first;
900 		ifpaddr = ifa->ifa_addr;
901 		len = rt_msg2(RTM_IFINFO, &info, (caddr_t)0, w);
902 		ifpaddr = 0;
903 		if (w->w_req && w->w_tmem) {
904 			register struct if_msghdr *ifm;
905 
906 			ifm = (struct if_msghdr *)w->w_tmem;
907 			ifm->ifm_index = ifp->if_index;
908 			ifm->ifm_flags = (u_short)ifp->if_flags;
909 			ifm->ifm_data = ifp->if_data;
910 			ifm->ifm_addrs = info.rti_addrs;
911 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
912 			if (error)
913 				return (error);
914 		}
915 		while ((ifa = ifa->ifa_link.tqe_next) != 0) {
916 			if (af && af != ifa->ifa_addr->sa_family)
917 				continue;
918 			if (curproc->p_prison && prison_if(curproc, ifa->ifa_addr))
919 				continue;
920 			ifaaddr = ifa->ifa_addr;
921 			netmask = ifa->ifa_netmask;
922 			brdaddr = ifa->ifa_dstaddr;
923 			len = rt_msg2(RTM_NEWADDR, &info, 0, w);
924 			if (w->w_req && w->w_tmem) {
925 				register struct ifa_msghdr *ifam;
926 
927 				ifam = (struct ifa_msghdr *)w->w_tmem;
928 				ifam->ifam_index = ifa->ifa_ifp->if_index;
929 				ifam->ifam_flags = ifa->ifa_flags;
930 				ifam->ifam_metric = ifa->ifa_metric;
931 				ifam->ifam_addrs = info.rti_addrs;
932 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
933 				if (error)
934 					return (error);
935 			}
936 		}
937 		ifaaddr = netmask = brdaddr = 0;
938 	}
939 	return (0);
940 }
941 
942 static int
943 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
944 {
945 	int	*name = (int *)arg1;
946 	u_int	namelen = arg2;
947 	register struct radix_node_head *rnh;
948 	int	i, s, error = EINVAL;
949 	u_char  af;
950 	struct	walkarg w;
951 
952 	name ++;
953 	namelen--;
954 	if (req->newptr)
955 		return (EPERM);
956 	if (namelen != 3)
957 		return (EINVAL);
958 	af = name[0];
959 	Bzero(&w, sizeof(w));
960 	w.w_op = name[1];
961 	w.w_arg = name[2];
962 	w.w_req = req;
963 
964 	s = splnet();
965 	switch (w.w_op) {
966 
967 	case NET_RT_DUMP:
968 	case NET_RT_FLAGS:
969 		for (i = 1; i <= AF_MAX; i++)
970 			if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
971 			    (error = rnh->rnh_walktree(rnh,
972 							sysctl_dumpentry, &w)))
973 				break;
974 		break;
975 
976 	case NET_RT_IFLIST:
977 		error = sysctl_iflist(af, &w);
978 	}
979 	splx(s);
980 	if (w.w_tmem)
981 		free(w.w_tmem, M_RTABLE);
982 	return (error);
983 }
984 
985 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
986 
987 /*
988  * Definitions of protocols supported in the ROUTE domain.
989  */
990 
991 extern struct domain routedomain;		/* or at least forward */
992 
993 static struct protosw routesw[] = {
994 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
995   0,		route_output,	raw_ctlinput,	0,
996   0,
997   raw_init,	0,		0,		0,
998   &route_usrreqs
999 }
1000 };
1001 
1002 static struct domain routedomain =
1003     { PF_ROUTE, "route", 0, 0, 0,
1004       routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] };
1005 
1006 DOMAIN_SET(route);
1007