xref: /freebsd/sys/net/rtsock.c (revision 6e8394b8baa7d5d9153ab90de6824bcd19b3b4e1)
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  *	$Id: rtsock.c,v 1.39 1999/04/28 11:37:38 phk Exp $
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_IPX:
137 		route_cb.ipx_count++;
138 		break;
139 	case AF_NS:
140 		route_cb.ns_count++;
141 		break;
142 	case AF_ISO:
143 		route_cb.iso_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_IPX:
190 			route_cb.ipx_count--;
191 			break;
192 		case AF_NS:
193 			route_cb.ns_count--;
194 			break;
195 		case AF_ISO:
196 			route_cb.iso_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(genmask, t->rn_key, *(u_char *)genmask) == 0)
322 			genmask = (struct sockaddr *)(t->rn_key);
323 		else
324 			senderr(ENOBUFS);
325 	}
326 	switch (rtm->rtm_type) {
327 
328 	case RTM_ADD:
329 		if (gate == 0)
330 			senderr(EINVAL);
331 		error = rtrequest(RTM_ADD, dst, gate, netmask,
332 					rtm->rtm_flags, &saved_nrt);
333 		if (error == 0 && saved_nrt) {
334 			rt_setmetrics(rtm->rtm_inits,
335 				&rtm->rtm_rmx, &saved_nrt->rt_rmx);
336 			saved_nrt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
337 			saved_nrt->rt_rmx.rmx_locks |=
338 				(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
339 			saved_nrt->rt_refcnt--;
340 			saved_nrt->rt_genmask = genmask;
341 		}
342 		break;
343 
344 	case RTM_DELETE:
345 		error = rtrequest(RTM_DELETE, dst, gate, netmask,
346 				rtm->rtm_flags, &saved_nrt);
347 		if (error == 0) {
348 			if ((rt = saved_nrt))
349 				rt->rt_refcnt++;
350 			goto report;
351 		}
352 		break;
353 
354 	case RTM_GET:
355 	case RTM_CHANGE:
356 	case RTM_LOCK:
357 		if ((rnh = rt_tables[dst->sa_family]) == 0) {
358 			senderr(EAFNOSUPPORT);
359 		} else if ((rt = (struct rtentry *)
360 				rnh->rnh_lookup(dst, netmask, rnh)) != NULL)
361 			rt->rt_refcnt++;
362 		else
363 			senderr(ESRCH);
364 		switch(rtm->rtm_type) {
365 
366 		case RTM_GET:
367 		report:
368 			dst = rt_key(rt);
369 			gate = rt->rt_gateway;
370 			netmask = rt_mask(rt);
371 			genmask = rt->rt_genmask;
372 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
373 				ifp = rt->rt_ifp;
374 				if (ifp) {
375 					ifpaddr = ifp->if_addrhead.tqh_first->ifa_addr;
376 					ifaaddr = rt->rt_ifa->ifa_addr;
377 					rtm->rtm_index = ifp->if_index;
378 				} else {
379 					ifpaddr = 0;
380 					ifaaddr = 0;
381 			    }
382 			}
383 			len = rt_msg2(rtm->rtm_type, &info, (caddr_t)0,
384 				(struct walkarg *)0);
385 			if (len > rtm->rtm_msglen) {
386 				struct rt_msghdr *new_rtm;
387 				R_Malloc(new_rtm, struct rt_msghdr *, len);
388 				if (new_rtm == 0)
389 					senderr(ENOBUFS);
390 				Bcopy(rtm, new_rtm, rtm->rtm_msglen);
391 				Free(rtm); rtm = new_rtm;
392 			}
393 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm,
394 				(struct walkarg *)0);
395 			rtm->rtm_flags = rt->rt_flags;
396 			rtm->rtm_rmx = rt->rt_rmx;
397 			rtm->rtm_addrs = info.rti_addrs;
398 			break;
399 
400 		case RTM_CHANGE:
401 			if (gate && (error = rt_setgate(rt, rt_key(rt), gate)))
402 				senderr(error);
403 
404 			/*
405 			 * If they tried to change things but didn't specify
406 			 * the required gateway, then just use the old one.
407 			 * This can happen if the user tries to change the
408 			 * flags on the default route without changing the
409 			 * default gateway.  Changing flags still doesn't work.
410 			 */
411 			if ((rt->rt_flags & RTF_GATEWAY) && !gate)
412 				gate = rt->rt_gateway;
413 
414 			/* new gateway could require new ifaddr, ifp;
415 			   flags may also be different; ifp may be specified
416 			   by ll sockaddr when protocol address is ambiguous */
417 			if (ifpaddr && (ifa = ifa_ifwithnet(ifpaddr)) &&
418 			    (ifp = ifa->ifa_ifp) && (ifaaddr || gate))
419 				ifa = ifaof_ifpforaddr(ifaaddr ? ifaaddr : gate,
420 							ifp);
421 			else if ((ifaaddr && (ifa = ifa_ifwithaddr(ifaaddr))) ||
422 				 (gate && (ifa = ifa_ifwithroute(rt->rt_flags,
423 							rt_key(rt), gate))))
424 				ifp = ifa->ifa_ifp;
425 			if (ifa) {
426 				register struct ifaddr *oifa = rt->rt_ifa;
427 				if (oifa != ifa) {
428 				    if (oifa && oifa->ifa_rtrequest)
429 					oifa->ifa_rtrequest(RTM_DELETE,
430 								rt, gate);
431 				    IFAFREE(rt->rt_ifa);
432 				    rt->rt_ifa = ifa;
433 				    ifa->ifa_refcnt++;
434 				    rt->rt_ifp = ifp;
435 				}
436 			}
437 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
438 					&rt->rt_rmx);
439 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
440 			       rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, gate);
441 			if (genmask)
442 				rt->rt_genmask = genmask;
443 			/*
444 			 * Fall into
445 			 */
446 		case RTM_LOCK:
447 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
448 			rt->rt_rmx.rmx_locks |=
449 				(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
450 			break;
451 		}
452 		break;
453 
454 	default:
455 		senderr(EOPNOTSUPP);
456 	}
457 
458 flush:
459 	if (rtm) {
460 		if (error)
461 			rtm->rtm_errno = error;
462 		else
463 			rtm->rtm_flags |= RTF_DONE;
464 	}
465 	if (rt)
466 		rtfree(rt);
467     {
468 	register struct rawcb *rp = 0;
469 	/*
470 	 * Check to see if we don't want our own messages.
471 	 */
472 	if ((so->so_options & SO_USELOOPBACK) == 0) {
473 		if (route_cb.any_count <= 1) {
474 			if (rtm)
475 				Free(rtm);
476 			m_freem(m);
477 			return (error);
478 		}
479 		/* There is another listener, so construct message */
480 		rp = sotorawcb(so);
481 	}
482 	if (rtm) {
483 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
484 		Free(rtm);
485 	}
486 	if (rp)
487 		rp->rcb_proto.sp_family = 0; /* Avoid us */
488 	if (dst)
489 		route_proto.sp_protocol = dst->sa_family;
490 	raw_input(m, &route_proto, &route_src, &route_dst);
491 	if (rp)
492 		rp->rcb_proto.sp_family = PF_ROUTE;
493     }
494 	return (error);
495 }
496 
497 static void
498 rt_setmetrics(which, in, out)
499 	u_long which;
500 	register struct rt_metrics *in, *out;
501 {
502 #define metric(f, e) if (which & (f)) out->e = in->e;
503 	metric(RTV_RPIPE, rmx_recvpipe);
504 	metric(RTV_SPIPE, rmx_sendpipe);
505 	metric(RTV_SSTHRESH, rmx_ssthresh);
506 	metric(RTV_RTT, rmx_rtt);
507 	metric(RTV_RTTVAR, rmx_rttvar);
508 	metric(RTV_HOPCOUNT, rmx_hopcount);
509 	metric(RTV_MTU, rmx_mtu);
510 	metric(RTV_EXPIRE, rmx_expire);
511 #undef metric
512 }
513 
514 #define ROUNDUP(a) \
515 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
516 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
517 
518 
519 /*
520  * Extract the addresses of the passed sockaddrs.
521  * Do a little sanity checking so as to avoid bad memory references.
522  * This data is derived straight from userland.
523  */
524 static int
525 rt_xaddrs(cp, cplim, rtinfo)
526 	register caddr_t cp, cplim;
527 	register struct rt_addrinfo *rtinfo;
528 {
529 	register struct sockaddr *sa;
530 	register int i;
531 
532 	bzero(rtinfo->rti_info, sizeof(rtinfo->rti_info));
533 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
534 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
535 			continue;
536 		sa = (struct sockaddr *)cp;
537 		/*
538 		 * It won't fit.
539 		 */
540 		if ( (cp + sa->sa_len) > cplim ) {
541 			return (EINVAL);
542 		}
543 
544 		/*
545 		 * there are no more.. quit now
546 		 * If there are more bits, they are in error.
547 		 * I've seen this. route(1) can evidently generate these.
548 		 * This causes kernel to core dump.
549 		 * for compatibility, If we see this, point to a safe address.
550 		 */
551 		if (sa->sa_len == 0) {
552 			rtinfo->rti_info[i] = &sa_zero;
553 			return (0); /* should be EINVAL but for compat */
554 		}
555 
556 		/* accept it */
557 		rtinfo->rti_info[i] = sa;
558 		ADVANCE(cp, sa);
559 	}
560 	return (0);
561 }
562 
563 static struct mbuf *
564 rt_msg1(type, rtinfo)
565 	int type;
566 	register struct rt_addrinfo *rtinfo;
567 {
568 	register struct rt_msghdr *rtm;
569 	register struct mbuf *m;
570 	register int i;
571 	register struct sockaddr *sa;
572 	int len, dlen;
573 
574 	m = m_gethdr(M_DONTWAIT, MT_DATA);
575 	if (m == 0)
576 		return (m);
577 	switch (type) {
578 
579 	case RTM_DELADDR:
580 	case RTM_NEWADDR:
581 		len = sizeof(struct ifa_msghdr);
582 		break;
583 
584 	case RTM_DELMADDR:
585 	case RTM_NEWMADDR:
586 		len = sizeof(struct ifma_msghdr);
587 		break;
588 
589 	case RTM_IFINFO:
590 		len = sizeof(struct if_msghdr);
591 		break;
592 
593 	default:
594 		len = sizeof(struct rt_msghdr);
595 	}
596 	if (len > MHLEN)
597 		panic("rt_msg1");
598 	m->m_pkthdr.len = m->m_len = len;
599 	m->m_pkthdr.rcvif = 0;
600 	rtm = mtod(m, struct rt_msghdr *);
601 	bzero((caddr_t)rtm, len);
602 	for (i = 0; i < RTAX_MAX; i++) {
603 		if ((sa = rtinfo->rti_info[i]) == NULL)
604 			continue;
605 		rtinfo->rti_addrs |= (1 << i);
606 		dlen = ROUNDUP(sa->sa_len);
607 		m_copyback(m, len, dlen, (caddr_t)sa);
608 		len += dlen;
609 	}
610 	if (m->m_pkthdr.len != len) {
611 		m_freem(m);
612 		return (NULL);
613 	}
614 	rtm->rtm_msglen = len;
615 	rtm->rtm_version = RTM_VERSION;
616 	rtm->rtm_type = type;
617 	return (m);
618 }
619 
620 static int
621 rt_msg2(type, rtinfo, cp, w)
622 	int type;
623 	register struct rt_addrinfo *rtinfo;
624 	caddr_t cp;
625 	struct walkarg *w;
626 {
627 	register int i;
628 	int len, dlen, second_time = 0;
629 	caddr_t cp0;
630 
631 	rtinfo->rti_addrs = 0;
632 again:
633 	switch (type) {
634 
635 	case RTM_DELADDR:
636 	case RTM_NEWADDR:
637 		len = sizeof(struct ifa_msghdr);
638 		break;
639 
640 	case RTM_IFINFO:
641 		len = sizeof(struct if_msghdr);
642 		break;
643 
644 	default:
645 		len = sizeof(struct rt_msghdr);
646 	}
647 	cp0 = cp;
648 	if (cp0)
649 		cp += len;
650 	for (i = 0; i < RTAX_MAX; i++) {
651 		register struct sockaddr *sa;
652 
653 		if ((sa = rtinfo->rti_info[i]) == 0)
654 			continue;
655 		rtinfo->rti_addrs |= (1 << i);
656 		dlen = ROUNDUP(sa->sa_len);
657 		if (cp) {
658 			bcopy((caddr_t)sa, cp, (unsigned)dlen);
659 			cp += dlen;
660 		}
661 		len += dlen;
662 	}
663 	if (cp == 0 && w != NULL && !second_time) {
664 		register struct walkarg *rw = w;
665 
666 		if (rw->w_req) {
667 			if (rw->w_tmemsize < len) {
668 				if (rw->w_tmem)
669 					free(rw->w_tmem, M_RTABLE);
670 				rw->w_tmem = (caddr_t)
671 					malloc(len, M_RTABLE, M_NOWAIT);
672 				if (rw->w_tmem)
673 					rw->w_tmemsize = len;
674 			}
675 			if (rw->w_tmem) {
676 				cp = rw->w_tmem;
677 				second_time = 1;
678 				goto again;
679 			}
680 		}
681 	}
682 	if (cp) {
683 		register struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
684 
685 		rtm->rtm_version = RTM_VERSION;
686 		rtm->rtm_type = type;
687 		rtm->rtm_msglen = len;
688 	}
689 	return (len);
690 }
691 
692 /*
693  * This routine is called to generate a message from the routing
694  * socket indicating that a redirect has occured, a routing lookup
695  * has failed, or that a protocol has detected timeouts to a particular
696  * destination.
697  */
698 void
699 rt_missmsg(type, rtinfo, flags, error)
700 	int type, flags, error;
701 	register struct rt_addrinfo *rtinfo;
702 {
703 	register struct rt_msghdr *rtm;
704 	register struct mbuf *m;
705 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
706 
707 	if (route_cb.any_count == 0)
708 		return;
709 	m = rt_msg1(type, rtinfo);
710 	if (m == 0)
711 		return;
712 	rtm = mtod(m, struct rt_msghdr *);
713 	rtm->rtm_flags = RTF_DONE | flags;
714 	rtm->rtm_errno = error;
715 	rtm->rtm_addrs = rtinfo->rti_addrs;
716 	route_proto.sp_protocol = sa ? sa->sa_family : 0;
717 	raw_input(m, &route_proto, &route_src, &route_dst);
718 }
719 
720 /*
721  * This routine is called to generate a message from the routing
722  * socket indicating that the status of a network interface has changed.
723  */
724 void
725 rt_ifmsg(ifp)
726 	register struct ifnet *ifp;
727 {
728 	register struct if_msghdr *ifm;
729 	struct mbuf *m;
730 	struct rt_addrinfo info;
731 
732 	if (route_cb.any_count == 0)
733 		return;
734 	bzero((caddr_t)&info, sizeof(info));
735 	m = rt_msg1(RTM_IFINFO, &info);
736 	if (m == 0)
737 		return;
738 	ifm = mtod(m, struct if_msghdr *);
739 	ifm->ifm_index = ifp->if_index;
740 	ifm->ifm_flags = (u_short)ifp->if_flags;
741 	ifm->ifm_data = ifp->if_data;
742 	ifm->ifm_addrs = 0;
743 	route_proto.sp_protocol = 0;
744 	raw_input(m, &route_proto, &route_src, &route_dst);
745 }
746 
747 /*
748  * This is called to generate messages from the routing socket
749  * indicating a network interface has had addresses associated with it.
750  * if we ever reverse the logic and replace messages TO the routing
751  * socket indicate a request to configure interfaces, then it will
752  * be unnecessary as the routing socket will automatically generate
753  * copies of it.
754  */
755 void
756 rt_newaddrmsg(cmd, ifa, error, rt)
757 	int cmd, error;
758 	register struct ifaddr *ifa;
759 	register struct rtentry *rt;
760 {
761 	struct rt_addrinfo info;
762 	struct sockaddr *sa = 0;
763 	int pass;
764 	struct mbuf *m = 0;
765 	struct ifnet *ifp = ifa->ifa_ifp;
766 
767 	if (route_cb.any_count == 0)
768 		return;
769 	for (pass = 1; pass < 3; pass++) {
770 		bzero((caddr_t)&info, sizeof(info));
771 		if ((cmd == RTM_ADD && pass == 1) ||
772 		    (cmd == RTM_DELETE && pass == 2)) {
773 			register struct ifa_msghdr *ifam;
774 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
775 
776 			ifaaddr = sa = ifa->ifa_addr;
777 			ifpaddr = ifp->if_addrhead.tqh_first->ifa_addr;
778 			netmask = ifa->ifa_netmask;
779 			brdaddr = ifa->ifa_dstaddr;
780 			if ((m = rt_msg1(ncmd, &info)) == NULL)
781 				continue;
782 			ifam = mtod(m, struct ifa_msghdr *);
783 			ifam->ifam_index = ifp->if_index;
784 			ifam->ifam_metric = ifa->ifa_metric;
785 			ifam->ifam_flags = ifa->ifa_flags;
786 			ifam->ifam_addrs = info.rti_addrs;
787 		}
788 		if ((cmd == RTM_ADD && pass == 2) ||
789 		    (cmd == RTM_DELETE && pass == 1)) {
790 			register struct rt_msghdr *rtm;
791 
792 			if (rt == 0)
793 				continue;
794 			netmask = rt_mask(rt);
795 			dst = sa = rt_key(rt);
796 			gate = rt->rt_gateway;
797 			if ((m = rt_msg1(cmd, &info)) == NULL)
798 				continue;
799 			rtm = mtod(m, struct rt_msghdr *);
800 			rtm->rtm_index = ifp->if_index;
801 			rtm->rtm_flags |= rt->rt_flags;
802 			rtm->rtm_errno = error;
803 			rtm->rtm_addrs = info.rti_addrs;
804 		}
805 		route_proto.sp_protocol = sa ? sa->sa_family : 0;
806 		raw_input(m, &route_proto, &route_src, &route_dst);
807 	}
808 }
809 
810 /*
811  * This is the analogue to the rt_newaddrmsg which performs the same
812  * function but for multicast group memberhips.  This is easier since
813  * there is no route state to worry about.
814  */
815 void
816 rt_newmaddrmsg(cmd, ifma)
817 	int cmd;
818 	struct ifmultiaddr *ifma;
819 {
820 	struct rt_addrinfo info;
821 	struct mbuf *m = 0;
822 	struct ifnet *ifp = ifma->ifma_ifp;
823 	struct ifma_msghdr *ifmam;
824 
825 	if (route_cb.any_count == 0)
826 		return;
827 
828 	bzero((caddr_t)&info, sizeof(info));
829 	ifaaddr = ifma->ifma_addr;
830 	ifpaddr = ifp->if_addrhead.tqh_first->ifa_addr;
831 	/*
832 	 * If a link-layer address is present, present it as a ``gateway''
833 	 * (similarly to how ARP entries, e.g., are presented).
834 	 */
835 	gate = ifma->ifma_lladdr;
836 	if ((m = rt_msg1(cmd, &info)) == NULL)
837 		return;
838 	ifmam = mtod(m, struct ifma_msghdr *);
839 	ifmam->ifmam_index = ifp->if_index;
840 	ifmam->ifmam_addrs = info.rti_addrs;
841 	route_proto.sp_protocol = ifma->ifma_addr->sa_family;
842 	raw_input(m, &route_proto, &route_src, &route_dst);
843 }
844 
845 /*
846  * This is used in dumping the kernel table via sysctl().
847  */
848 int
849 sysctl_dumpentry(rn, vw)
850 	struct radix_node *rn;
851 	void *vw;
852 {
853 	register struct walkarg *w = vw;
854 	register struct rtentry *rt = (struct rtentry *)rn;
855 	int error = 0, size;
856 	struct rt_addrinfo info;
857 
858 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
859 		return 0;
860 	bzero((caddr_t)&info, sizeof(info));
861 	dst = rt_key(rt);
862 	gate = rt->rt_gateway;
863 	netmask = rt_mask(rt);
864 	genmask = rt->rt_genmask;
865 	size = rt_msg2(RTM_GET, &info, 0, w);
866 	if (w->w_req && w->w_tmem) {
867 		register struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
868 
869 		rtm->rtm_flags = rt->rt_flags;
870 		rtm->rtm_use = rt->rt_use;
871 		rtm->rtm_rmx = rt->rt_rmx;
872 		rtm->rtm_index = rt->rt_ifp->if_index;
873 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
874 		rtm->rtm_addrs = info.rti_addrs;
875 		error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
876 		return (error);
877 	}
878 	return (error);
879 }
880 
881 int
882 sysctl_iflist(af, w)
883 	int	af;
884 	register struct	walkarg *w;
885 {
886 	register struct ifnet *ifp;
887 	register struct ifaddr *ifa;
888 	struct	rt_addrinfo info;
889 	int	len, error = 0;
890 
891 	bzero((caddr_t)&info, sizeof(info));
892 	for (ifp = ifnet.tqh_first; ifp; ifp = ifp->if_link.tqe_next) {
893 		if (w->w_arg && w->w_arg != ifp->if_index)
894 			continue;
895 		ifa = ifp->if_addrhead.tqh_first;
896 		ifpaddr = ifa->ifa_addr;
897 		len = rt_msg2(RTM_IFINFO, &info, (caddr_t)0, w);
898 		ifpaddr = 0;
899 		if (w->w_req && w->w_tmem) {
900 			register struct if_msghdr *ifm;
901 
902 			ifm = (struct if_msghdr *)w->w_tmem;
903 			ifm->ifm_index = ifp->if_index;
904 			ifm->ifm_flags = (u_short)ifp->if_flags;
905 			ifm->ifm_data = ifp->if_data;
906 			ifm->ifm_addrs = info.rti_addrs;
907 			error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
908 			if (error)
909 				return (error);
910 		}
911 		while ((ifa = ifa->ifa_link.tqe_next) != 0) {
912 			if (af && af != ifa->ifa_addr->sa_family)
913 				continue;
914 			if (curproc->p_prison && prison_if(curproc, ifa->ifa_addr))
915 				continue;
916 			ifaaddr = ifa->ifa_addr;
917 			netmask = ifa->ifa_netmask;
918 			brdaddr = ifa->ifa_dstaddr;
919 			len = rt_msg2(RTM_NEWADDR, &info, 0, w);
920 			if (w->w_req && w->w_tmem) {
921 				register struct ifa_msghdr *ifam;
922 
923 				ifam = (struct ifa_msghdr *)w->w_tmem;
924 				ifam->ifam_index = ifa->ifa_ifp->if_index;
925 				ifam->ifam_flags = ifa->ifa_flags;
926 				ifam->ifam_metric = ifa->ifa_metric;
927 				ifam->ifam_addrs = info.rti_addrs;
928 				error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
929 				if (error)
930 					return (error);
931 			}
932 		}
933 		ifaaddr = netmask = brdaddr = 0;
934 	}
935 	return (0);
936 }
937 
938 static int
939 sysctl_rtsock SYSCTL_HANDLER_ARGS
940 {
941 	int	*name = (int *)arg1;
942 	u_int	namelen = arg2;
943 	register struct radix_node_head *rnh;
944 	int	i, s, error = EINVAL;
945 	u_char  af;
946 	struct	walkarg w;
947 
948 	name ++;
949 	namelen--;
950 	if (req->newptr)
951 		return (EPERM);
952 	if (namelen != 3)
953 		return (EINVAL);
954 	af = name[0];
955 	Bzero(&w, sizeof(w));
956 	w.w_op = name[1];
957 	w.w_arg = name[2];
958 	w.w_req = req;
959 
960 	s = splnet();
961 	switch (w.w_op) {
962 
963 	case NET_RT_DUMP:
964 	case NET_RT_FLAGS:
965 		for (i = 1; i <= AF_MAX; i++)
966 			if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
967 			    (error = rnh->rnh_walktree(rnh,
968 							sysctl_dumpentry, &w)))
969 				break;
970 		break;
971 
972 	case NET_RT_IFLIST:
973 		error = sysctl_iflist(af, &w);
974 	}
975 	splx(s);
976 	if (w.w_tmem)
977 		free(w.w_tmem, M_RTABLE);
978 	return (error);
979 }
980 
981 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
982 
983 /*
984  * Definitions of protocols supported in the ROUTE domain.
985  */
986 
987 extern struct domain routedomain;		/* or at least forward */
988 
989 static struct protosw routesw[] = {
990 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
991   0,		route_output,	raw_ctlinput,	0,
992   0,
993   raw_init,	0,		0,		0,
994   &route_usrreqs
995 }
996 };
997 
998 static struct domain routedomain =
999     { PF_ROUTE, "route", 0, 0, 0,
1000       routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] };
1001 
1002 DOMAIN_SET(route);
1003