xref: /openbsd-src/sys/net/rtsock.c (revision d13be5d47e4149db2549a9828e244d59dbc43f15)
1 /*	$OpenBSD: rtsock.c,v 1.118 2011/04/07 15:30:16 miod Exp $	*/
2 /*	$NetBSD: rtsock.c,v 1.18 1996/03/29 00:32:10 cgd Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1988, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)rtsock.c	8.6 (Berkeley) 2/11/95
62  */
63 
64 #include <sys/param.h>
65 #include <sys/systm.h>
66 #include <sys/proc.h>
67 #include <sys/mbuf.h>
68 #include <sys/socket.h>
69 #include <sys/socketvar.h>
70 #include <sys/domain.h>
71 #include <sys/protosw.h>
72 
73 #include <uvm/uvm_extern.h>
74 #include <sys/sysctl.h>
75 
76 #include <net/if.h>
77 #include <net/route.h>
78 #include <net/raw_cb.h>
79 
80 #ifdef MPLS
81 #include <netmpls/mpls.h>
82 #endif
83 
84 #include <sys/stdarg.h>
85 #include <sys/kernel.h>
86 #include <sys/timeout.h>
87 
88 struct sockaddr		route_dst = { 2, PF_ROUTE, };
89 struct sockaddr		route_src = { 2, PF_ROUTE, };
90 struct sockproto	route_proto = { PF_ROUTE, };
91 
92 struct walkarg {
93 	int	w_op, w_arg, w_given, w_needed, w_tmemsize;
94 	caddr_t	w_where, w_tmem;
95 };
96 
97 int	route_ctloutput(int, struct socket *, int, int, struct mbuf **);
98 void	route_input(struct mbuf *m0, ...);
99 
100 struct mbuf	*rt_msg1(int, struct rt_addrinfo *);
101 int		 rt_msg2(int, int, struct rt_addrinfo *, caddr_t,
102 		     struct walkarg *);
103 void		 rt_xaddrs(caddr_t, caddr_t, struct rt_addrinfo *);
104 
105 /* Sleazy use of local variables throughout file, warning!!!! */
106 #define dst	info.rti_info[RTAX_DST]
107 #define gate	info.rti_info[RTAX_GATEWAY]
108 #define netmask	info.rti_info[RTAX_NETMASK]
109 #define genmask	info.rti_info[RTAX_GENMASK]
110 #define ifpaddr	info.rti_info[RTAX_IFP]
111 #define ifaaddr	info.rti_info[RTAX_IFA]
112 #define brdaddr	info.rti_info[RTAX_BRD]
113 
114 struct routecb {
115 	struct rawcb	rcb;
116 	struct timeout	timeout;
117 	unsigned int	msgfilter;
118 	unsigned int	flags;
119 	u_int		rtableid;
120 };
121 #define	sotoroutecb(so)	((struct routecb *)(so)->so_pcb)
122 
123 /*
124  * These flags and timeout are used for indicating to userland (via a
125  * RTM_DESYNC msg) when the route socket has overflowed and messages
126  * have been lost.
127  */
128 #define ROUTECB_FLAG_DESYNC	0x1	/* Route socket out of memory */
129 #define ROUTECB_FLAG_FLUSH	0x2	/* Wait until socket is empty before
130 					   queueing more packets */
131 
132 #define ROUTE_DESYNC_RESEND_TIMEOUT	(hz / 5)	/* In hz */
133 
134 void	rt_senddesync(void *);
135 
136 int
137 route_usrreq(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
138     struct mbuf *control, struct proc *p)
139 {
140 	struct rawcb	*rp;
141 	struct routecb	*rop;
142 	int		 s, af;
143 	int		 error = 0;
144 
145 	s = splsoftnet();
146 	rp = sotorawcb(so);
147 
148 	switch (req) {
149 	case PRU_ATTACH:
150 		/*
151 		 * use the rawcb but allocate a routecb, this
152 		 * code does not care about the additional fields
153 		 * and works directly on the raw socket.
154 		 */
155 		rop = malloc(sizeof(struct routecb), M_PCB, M_WAITOK|M_ZERO);
156 		rp = &rop->rcb;
157 		so->so_pcb = rp;
158 		/* Init the timeout structure */
159 		timeout_set(&((struct routecb *)rp)->timeout, rt_senddesync, rp);
160 		/*
161 		 * Don't call raw_usrreq() in the attach case, because
162 		 * we want to allow non-privileged processes to listen
163 		 * on and send "safe" commands to the routing socket.
164 		 */
165 		if (curproc == 0)
166 			error = EACCES;
167 		else
168 			error = raw_attach(so, (int)(long)nam);
169 		if (error) {
170 			free(rp, M_PCB);
171 			splx(s);
172 			return (error);
173 		}
174 		rop->rtableid = curproc->p_p->ps_rtableid;
175 		af = rp->rcb_proto.sp_protocol;
176 		if (af == AF_INET)
177 			route_cb.ip_count++;
178 		else if (af == AF_INET6)
179 			route_cb.ip6_count++;
180 #ifdef MPLS
181 		else if (af == AF_MPLS)
182 			route_cb.mpls_count++;
183 #endif
184 		rp->rcb_faddr = &route_src;
185 		route_cb.any_count++;
186 		soisconnected(so);
187 		so->so_options |= SO_USELOOPBACK;
188 		break;
189 
190 	case PRU_RCVD:
191 		rop = (struct routecb *)rp;
192 
193 		/*
194 		 * If we are in a FLUSH state, check if the buffer is
195 		 * empty so that we can clear the flag.
196 		 */
197 		if (((rop->flags & ROUTECB_FLAG_FLUSH) != 0) &&
198 		    ((sbspace(&rp->rcb_socket->so_rcv) ==
199 		    rp->rcb_socket->so_rcv.sb_hiwat)))
200 			rop->flags &= ~ROUTECB_FLAG_FLUSH;
201 		break;
202 
203 	case PRU_DETACH:
204 		if (rp) {
205 			timeout_del(&((struct routecb *)rp)->timeout);
206 			af = rp->rcb_proto.sp_protocol;
207 			if (af == AF_INET)
208 				route_cb.ip_count--;
209 			else if (af == AF_INET6)
210 				route_cb.ip6_count--;
211 #ifdef MPLS
212 			else if (af == AF_MPLS)
213 				route_cb.mpls_count--;
214 #endif
215 			route_cb.any_count--;
216 		}
217 		/* FALLTHROUGH */
218 	default:
219 		error = raw_usrreq(so, req, m, nam, control, p);
220 	}
221 
222 	splx(s);
223 	return (error);
224 }
225 
226 int
227 route_ctloutput(int op, struct socket *so, int level, int optname,
228     struct mbuf **mp)
229 {
230 	struct routecb *rop = sotoroutecb(so);
231 	struct mbuf *m = *mp;
232 	int error = 0;
233 	unsigned int tid;
234 
235 	if (level != AF_ROUTE) {
236 		error = EINVAL;
237 		if (op == PRCO_SETOPT && *mp)
238 			m_free(*mp);
239 		return (error);
240 	}
241 
242 	switch (op) {
243 	case PRCO_SETOPT:
244 		switch (optname) {
245 		case ROUTE_MSGFILTER:
246 			if (m == NULL || m->m_len != sizeof(unsigned int))
247 				error = EINVAL;
248 			else
249 				rop->msgfilter = *mtod(m, unsigned int *);
250 			break;
251 		case ROUTE_TABLEFILTER:
252 			if (m == NULL || m->m_len != sizeof(unsigned int)) {
253 				error = EINVAL;
254 				break;
255 			}
256 			tid = *mtod(m, unsigned int *);
257 			if (tid != RTABLE_ANY && !rtable_exists(tid))
258 				error = ENOENT;
259 			else
260 				rop->rtableid = tid;
261 			break;
262 		default:
263 			error = ENOPROTOOPT;
264 			break;
265 		}
266 		if (m)
267 			m_free(m);
268 		break;
269 	case PRCO_GETOPT:
270 		switch (optname) {
271 		case ROUTE_MSGFILTER:
272 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
273 			m->m_len = sizeof(unsigned int);
274 			*mtod(m, unsigned int *) = rop->msgfilter;
275 			break;
276 		case ROUTE_TABLEFILTER:
277 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
278 			m->m_len = sizeof(unsigned int);
279 			*mtod(m, unsigned int *) = rop->rtableid;
280 			break;
281 		default:
282 			error = ENOPROTOOPT;
283 			break;
284 		}
285 	}
286 	return (error);
287 }
288 
289 void
290 rt_senddesync(void *data)
291 {
292 	struct rawcb	*rp;
293 	struct routecb	*rop;
294 	struct mbuf	*desync_mbuf;
295 
296 	rp = (struct rawcb *)data;
297 	rop = (struct routecb *)rp;
298 
299 	/* If we are in a DESYNC state, try to send a RTM_DESYNC packet */
300 	if ((rop->flags & ROUTECB_FLAG_DESYNC) != 0) {
301 		/*
302 		 * If we fail to alloc memory or if sbappendaddr()
303 		 * fails, re-add timeout and try again.
304 		 */
305 		desync_mbuf = rt_msg1(RTM_DESYNC, NULL);
306 		if ((desync_mbuf != NULL) &&
307 		    (sbappendaddr(&rp->rcb_socket->so_rcv, &route_src,
308 		    desync_mbuf, (struct mbuf *)0) != 0)) {
309 			rop->flags &= ~ROUTECB_FLAG_DESYNC;
310 			sorwakeup(rp->rcb_socket);
311 		} else {
312 			if (desync_mbuf)
313 				m_freem(desync_mbuf);
314 			/* Re-add timeout to try sending msg again */
315 			timeout_add(&rop->timeout, ROUTE_DESYNC_RESEND_TIMEOUT);
316 		}
317 	}
318 }
319 
320 void
321 route_input(struct mbuf *m0, ...)
322 {
323 	struct rawcb *rp;
324 	struct routecb *rop;
325 	struct rt_msghdr *rtm;
326 	struct mbuf *m = m0;
327 	int sockets = 0;
328 	struct socket *last = NULL;
329 	va_list ap;
330 	struct sockproto *proto;
331 	struct sockaddr *sosrc, *sodst;
332 
333 	va_start(ap, m0);
334 	proto = va_arg(ap, struct sockproto *);
335 	sosrc = va_arg(ap, struct sockaddr *);
336 	sodst = va_arg(ap, struct sockaddr *);
337 	va_end(ap);
338 
339 	/* ensure that we can access the rtm_type via mtod() */
340 	if (m->m_len < offsetof(struct rt_msghdr, rtm_type) + 1) {
341 		m_freem(m);
342 		return;
343 	}
344 
345 	LIST_FOREACH(rp, &rawcb, rcb_list) {
346 		if (rp->rcb_socket->so_state & SS_CANTRCVMORE)
347 			continue;
348 		if (rp->rcb_proto.sp_family != proto->sp_family)
349 			continue;
350 		if (rp->rcb_proto.sp_protocol && proto->sp_protocol &&
351 		    rp->rcb_proto.sp_protocol != proto->sp_protocol)
352 			continue;
353 		/*
354 		 * We assume the lower level routines have
355 		 * placed the address in a canonical format
356 		 * suitable for a structure comparison.
357 		 *
358 		 * Note that if the lengths are not the same
359 		 * the comparison will fail at the first byte.
360 		 */
361 #define	equal(a1, a2) \
362   (bcmp((caddr_t)(a1), (caddr_t)(a2), a1->sa_len) == 0)
363 		if (rp->rcb_laddr && !equal(rp->rcb_laddr, sodst))
364 			continue;
365 		if (rp->rcb_faddr && !equal(rp->rcb_faddr, sosrc))
366 			continue;
367 
368 		/* filter messages that the process does not want */
369 		rop = (struct routecb *)rp;
370 		rtm = mtod(m, struct rt_msghdr *);
371 		if (rop->msgfilter != 0 && !(rop->msgfilter & (1 <<
372 		    rtm->rtm_type)))
373 			continue;
374 		switch (rtm->rtm_type) {
375 		case RTM_IFANNOUNCE:
376 		case RTM_DESYNC:
377 			/* no tableid */
378 			break;
379 		case RTM_RESOLVE:
380 		case RTM_NEWADDR:
381 		case RTM_DELADDR:
382 		case RTM_IFINFO:
383 			/* check against rdomain id */
384 			if (rop->rtableid != RTABLE_ANY &&
385 			    rtable_l2(rop->rtableid) != rtm->rtm_tableid)
386 				continue;
387 			break;
388 		default:
389 			/* check against rtable id */
390 			if (rop->rtableid != RTABLE_ANY &&
391 			    rop->rtableid != rtm->rtm_tableid)
392 				continue;
393 			break;
394 		}
395 
396 		/*
397 		 * Check to see if the flush flag is set. If so, don't queue
398 		 * any more messages until the flag is cleared.
399 		 */
400 		if ((rop->flags & ROUTECB_FLAG_FLUSH) != 0)
401 			continue;
402 
403 		if (last) {
404 			struct mbuf *n;
405 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
406 				if (sbspace(&last->so_rcv) < (2 * MSIZE) ||
407 				    sbappendaddr(&last->so_rcv, sosrc,
408 				    n, (struct mbuf *)0) == 0) {
409 					/*
410 					 * Flag socket as desync'ed and
411 					 * flush required
412 					 */
413 					sotoroutecb(last)->flags |=
414 					    ROUTECB_FLAG_DESYNC |
415 					    ROUTECB_FLAG_FLUSH;
416 					rt_senddesync((void *) sotorawcb(last));
417 					m_freem(n);
418 				} else {
419 					sorwakeup(last);
420 					sockets++;
421 				}
422 			}
423 		}
424 		last = rp->rcb_socket;
425 	}
426 	if (last) {
427 		if (sbspace(&last->so_rcv) < (2 * MSIZE) ||
428 		    sbappendaddr(&last->so_rcv, sosrc,
429 		    m, (struct mbuf *)0) == 0) {
430 			/* Flag socket as desync'ed and flush required */
431 			sotoroutecb(last)->flags |=
432 			    ROUTECB_FLAG_DESYNC | ROUTECB_FLAG_FLUSH;
433 			rt_senddesync((void *) sotorawcb(last));
434 			m_freem(m);
435 		} else {
436 			sorwakeup(last);
437 			sockets++;
438 		}
439 	} else
440 		m_freem(m);
441 }
442 
443 int
444 route_output(struct mbuf *m, ...)
445 {
446 	struct rt_msghdr	*rtm = NULL;
447 	struct radix_node	*rn = NULL;
448 	struct rtentry		*rt = NULL;
449 	struct rtentry		*saved_nrt = NULL;
450 	struct radix_node_head	*rnh;
451 	struct rt_addrinfo	 info;
452 	int			 len, newgate, error = 0;
453 	struct ifnet		*ifp = NULL;
454 	struct ifaddr		*ifa = NULL;
455 	struct socket		*so;
456 	struct rawcb		*rp = NULL;
457 	struct sockaddr_rtlabel	 sa_rt;
458 #ifdef MPLS
459 	struct sockaddr_mpls	 sa_mpls, *psa_mpls;
460 #endif
461 	const char		*label;
462 	va_list			 ap;
463 	u_int			 tableid;
464 	u_int8_t		 prio;
465 
466 	va_start(ap, m);
467 	so = va_arg(ap, struct socket *);
468 	va_end(ap);
469 
470 	dst = NULL;	/* for error handling (goto flush) */
471 	if (m == 0 || ((m->m_len < sizeof(int32_t)) &&
472 	    (m = m_pullup(m, sizeof(int32_t))) == 0))
473 		return (ENOBUFS);
474 	if ((m->m_flags & M_PKTHDR) == 0)
475 		panic("route_output");
476 	len = m->m_pkthdr.len;
477 	if (len < offsetof(struct rt_msghdr, rtm_type) + 1 ||
478 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
479 		error = EINVAL;
480 		goto fail;
481 	}
482 	switch (mtod(m, struct rt_msghdr *)->rtm_version) {
483 	case RTM_VERSION:
484 		if (len < sizeof(struct rt_msghdr)) {
485 			error = EINVAL;
486 			goto fail;
487 		}
488 		if (len > RTM_MAXSIZE) {
489 			error = EMSGSIZE;
490 			goto fail;
491 		}
492 		R_Malloc(rtm, struct rt_msghdr *, len);
493 		if (rtm == 0) {
494 			error = ENOBUFS;
495 			goto fail;
496 		}
497 		m_copydata(m, 0, len, (caddr_t)rtm);
498 		break;
499 	default:
500 		error = EPROTONOSUPPORT;
501 		goto fail;
502 	}
503 	rtm->rtm_pid = curproc->p_pid;
504 	if (rtm->rtm_hdrlen == 0)	/* old client */
505 		rtm->rtm_hdrlen = sizeof(struct rt_msghdr);
506 	if (len < rtm->rtm_hdrlen) {
507 		error = EINVAL;
508 		goto fail;
509 	}
510 
511 	/* Verify that the caller is sending an appropriate message early */
512 	switch (rtm->rtm_type) {
513 	case RTM_ADD:
514 	case RTM_DELETE:
515 	case RTM_GET:
516 	case RTM_CHANGE:
517 	case RTM_LOCK:
518 		break;
519 	default:
520 		error = EOPNOTSUPP;
521 		goto fail;
522 	}
523 
524 	/*
525 	 * Verify that the caller has the appropriate privilege; RTM_GET
526 	 * is the only operation the non-superuser is allowed.
527 	 */
528 	if (rtm->rtm_type != RTM_GET && suser(curproc, 0) != 0) {
529 		error = EACCES;
530 		goto fail;
531 	}
532 
533 	tableid = rtm->rtm_tableid;
534 	if (!rtable_exists(tableid)) {
535 		if (rtm->rtm_type == RTM_ADD) {
536 			if ((error = rtable_add(tableid)) != 0)
537 				goto flush;
538 		} else {
539 			error = EINVAL;
540 			goto flush;
541 		}
542 	}
543 
544 	/* make sure that kernel-only bits are not set */
545 	rtm->rtm_priority &= RTP_MASK;
546 
547 	if (rtm->rtm_priority != 0) {
548 		if (rtm->rtm_priority > RTP_MAX) {
549 			error = EINVAL;
550 			goto fail;
551 		}
552 		prio = rtm->rtm_priority;
553 	} else if (rtm->rtm_type != RTM_ADD)
554 		prio = RTP_ANY;
555 	else if (rtm->rtm_flags & RTF_STATIC)
556 		prio = 0;
557 	else
558 		prio = RTP_DEFAULT;
559 
560 	bzero(&info, sizeof(info));
561 	info.rti_addrs = rtm->rtm_addrs;
562 	rt_xaddrs(rtm->rtm_hdrlen + (caddr_t)rtm, len + (caddr_t)rtm, &info);
563 	info.rti_flags = rtm->rtm_flags;
564 	if (dst == 0 || dst->sa_family >= AF_MAX ||
565 	    (gate != 0 && gate->sa_family >= AF_MAX)) {
566 		error = EINVAL;
567 		goto flush;
568 	}
569 	if (genmask) {
570 		struct radix_node	*t;
571 		t = rn_addmask(genmask, 0, 1);
572 		if (t && genmask->sa_len >=
573 		    ((struct sockaddr *)t->rn_key)->sa_len &&
574 		    Bcmp((caddr_t *)genmask + 1, (caddr_t *)t->rn_key + 1,
575 		    ((struct sockaddr *)t->rn_key)->sa_len) - 1)
576 			genmask = (struct sockaddr *)(t->rn_key);
577 		else {
578 			error = ENOBUFS;
579 			goto flush;
580 		}
581 	}
582 #ifdef MPLS
583 	info.rti_mpls = rtm->rtm_mpls;
584 #endif
585 
586 	switch (rtm->rtm_type) {
587 	case RTM_ADD:
588 		if (gate == 0) {
589 			error = EINVAL;
590 			goto flush;
591 		}
592 		error = rtrequest1(rtm->rtm_type, &info, prio, &saved_nrt,
593 		    tableid);
594 		if (error == 0 && saved_nrt) {
595 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
596 			    &saved_nrt->rt_rmx);
597 			saved_nrt->rt_refcnt--;
598 			saved_nrt->rt_genmask = genmask;
599 			/* write back the priority the kernel used */
600 			rtm->rtm_priority = saved_nrt->rt_priority & RTP_MASK;
601 			rtm->rtm_index = saved_nrt->rt_ifp->if_index;
602 			rtm->rtm_flags = saved_nrt->rt_flags;
603 		}
604 		break;
605 	case RTM_DELETE:
606 		error = rtrequest1(rtm->rtm_type, &info, prio, &saved_nrt,
607 		    tableid);
608 		if (error == 0) {
609 			(rt = saved_nrt)->rt_refcnt++;
610 			goto report;
611 		}
612 		break;
613 	case RTM_GET:
614 	case RTM_CHANGE:
615 	case RTM_LOCK:
616 		if ((rnh = rt_gettable(dst->sa_family, tableid)) == NULL) {
617 			error = EAFNOSUPPORT;
618 			goto flush;
619 		}
620 		rn = rt_lookup(dst, netmask, tableid);
621 		if (rn == NULL || (rn->rn_flags & RNF_ROOT) != 0) {
622 			error = ESRCH;
623 			goto flush;
624 		}
625 		rt = (struct rtentry *)rn;
626 #ifndef SMALL_KERNEL
627 		/*
628 		 * for RTM_CHANGE/LOCK, if we got multipath routes,
629 		 * we require users to specify a matching RTAX_GATEWAY.
630 		 *
631 		 * for RTM_GET, gate is optional even with multipath.
632 		 * if gate == NULL the first match is returned.
633 		 * (no need to call rt_mpath_matchgate if gate == NULL)
634 		 */
635 		if (rn_mpath_capable(rnh)) {
636 			/* first find correct priority bucket */
637 			rn = rn_mpath_prio(rn, prio);
638 			rt = (struct rtentry *)rn;
639 			if (prio != RTP_ANY &&
640 			    (rt->rt_priority & RTP_MASK) != prio) {
641 				error = ESRCH;
642 				rt->rt_refcnt++;
643 				goto flush;
644 			}
645 
646 			/* if multipath routes */
647 			if (rn_mpath_next(rn, 0)) {
648 				if (gate)
649 					rt = rt_mpath_matchgate(rt, gate, prio);
650 				else if (rtm->rtm_type != RTM_GET)
651 					/*
652 					 * only RTM_GET may use an empty gate
653 					 * on multipath ...
654 					 */
655 					rt = NULL;
656 			} else if (gate && (rtm->rtm_type == RTM_GET ||
657 			    rtm->rtm_type == RTM_LOCK))
658 				/*
659 				 * ... but if a gate is specified RTM_GET
660 				 * and RTM_LOCK must match the gate no matter
661 				 * what.
662 				 */
663 				rt = rt_mpath_matchgate(rt, gate, prio);
664 
665 			if (!rt) {
666 				error = ESRCH;
667 				goto flush;
668 			}
669 			rn = (struct radix_node *)rt;
670 		}
671 #endif
672 		rt->rt_refcnt++;
673 
674 		/*
675 		 * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
676 		 * returns a perfect match in case a netmask is specified.
677 		 * For host routes only a longest prefix match is returned
678 		 * so it is necessary to compare the existence of the netmaks.
679 		 * If both have a netmask rn_lookup() did a perfect match and
680 		 * if none of them have a netmask both are host routes which is
681 		 * also a perfect match.
682 		 */
683 		if (rtm->rtm_type != RTM_GET && !rt_mask(rt) != !netmask) {
684 				error = ESRCH;
685 				goto flush;
686 		}
687 
688 		switch (rtm->rtm_type) {
689 		case RTM_GET:
690 report:
691 			dst = rt_key(rt);
692 			gate = rt->rt_gateway;
693 			netmask = rt_mask(rt);
694 			genmask = rt->rt_genmask;
695 
696 			if (rt->rt_labelid) {
697 				bzero(&sa_rt, sizeof(sa_rt));
698 				sa_rt.sr_len = sizeof(sa_rt);
699 				label = rtlabel_id2name(rt->rt_labelid);
700 				if (label != NULL)
701 					strlcpy(sa_rt.sr_label, label,
702 					    sizeof(sa_rt.sr_label));
703 				info.rti_info[RTAX_LABEL] =
704 				    (struct sockaddr *)&sa_rt;
705 			}
706 #ifdef MPLS
707 			if (rt->rt_flags & RTF_MPLS) {
708 				bzero(&sa_mpls, sizeof(sa_mpls));
709 				sa_mpls.smpls_family = AF_MPLS;
710 				sa_mpls.smpls_len = sizeof(sa_mpls);
711 				sa_mpls.smpls_label = ((struct rt_mpls *)
712 				    rt->rt_llinfo)->mpls_label;
713 				info.rti_info[RTAX_SRC] =
714 				    (struct sockaddr *)&sa_mpls;
715 				info.rti_mpls = ((struct rt_mpls *)
716 				    rt->rt_llinfo)->mpls_operation;
717 				rtm->rtm_mpls = info.rti_mpls;
718 			}
719 #endif
720 			ifpaddr = 0;
721 			ifaaddr = 0;
722 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA) &&
723 			    (ifp = rt->rt_ifp) != NULL) {
724 				ifpaddr =
725 				    TAILQ_FIRST(&ifp->if_addrlist)->ifa_addr;
726 				ifaaddr = rt->rt_ifa->ifa_addr;
727 				if (ifp->if_flags & IFF_POINTOPOINT)
728 					brdaddr = rt->rt_ifa->ifa_dstaddr;
729 				else
730 					brdaddr = 0;
731 				rtm->rtm_index = ifp->if_index;
732 			}
733 			len = rt_msg2(rtm->rtm_type, RTM_VERSION, &info, NULL,
734 			    NULL);
735 			if (len > rtm->rtm_msglen) {
736 				struct rt_msghdr	*new_rtm;
737 				R_Malloc(new_rtm, struct rt_msghdr *, len);
738 				if (new_rtm == 0) {
739 					error = ENOBUFS;
740 					goto flush;
741 				}
742 				Bcopy(rtm, new_rtm, rtm->rtm_msglen);
743 				Free(rtm); rtm = new_rtm;
744 			}
745 			rt_msg2(rtm->rtm_type, RTM_VERSION, &info, (caddr_t)rtm,
746 			    NULL);
747 			rtm->rtm_flags = rt->rt_flags;
748 			rtm->rtm_use = 0;
749 			rtm->rtm_priority = rt->rt_priority & RTP_MASK;
750 			rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
751 			rtm->rtm_addrs = info.rti_addrs;
752 			break;
753 
754 		case RTM_CHANGE:
755 			/*
756 			 * new gateway could require new ifaddr, ifp;
757 			 * flags may also be different; ifp may be specified
758 			 * by ll sockaddr when protocol address is ambiguous
759 			 */
760 			if ((error = rt_getifa(&info, tableid)) != 0)
761 				goto flush;
762 			newgate = 0;
763 			if (gate)
764 				if (rt->rt_gateway == NULL ||
765 				    bcmp(rt->rt_gateway, gate, gate->sa_len))
766 					newgate = 1;
767 			if (gate && rt_setgate(rt, rt_key(rt), gate, tableid)) {
768 				error = EDQUOT;
769 				goto flush;
770 			}
771 			if (ifpaddr &&
772 			    (ifa = ifa_ifwithnet(ifpaddr, tableid)) &&
773 			    (ifp = ifa->ifa_ifp) && (ifaaddr || gate))
774 				ifa = ifaof_ifpforaddr(ifaaddr ? ifaaddr : gate,
775 				    ifp);
776 			else if ((ifaaddr &&
777 			    (ifa = ifa_ifwithaddr(ifaaddr, tableid))) ||
778 			    (gate && (ifa = ifa_ifwithroute(rt->rt_flags,
779 			    rt_key(rt), gate, tableid))))
780 				ifp = ifa->ifa_ifp;
781 			if (ifa) {
782 				struct ifaddr *oifa = rt->rt_ifa;
783 				if (oifa != ifa) {
784 				    if (oifa && oifa->ifa_rtrequest)
785 					oifa->ifa_rtrequest(RTM_DELETE, rt,
786 					    &info);
787 				    IFAFREE(rt->rt_ifa);
788 				    rt->rt_ifa = ifa;
789 				    ifa->ifa_refcnt++;
790 				    rt->rt_ifp = ifp;
791 #ifndef SMALL_KERNEL
792 				    /* recheck link state after ifp change */
793 				    rt_if_linkstate_change(
794 					(struct radix_node *)rt, ifp, tableid);
795 #endif
796 				}
797 			}
798 #ifdef MPLS
799 			if ((rtm->rtm_flags & RTF_MPLS) &&
800 			    info.rti_info[RTAX_SRC] != NULL) {
801 				struct rt_mpls *rt_mpls;
802 
803 				psa_mpls = (struct sockaddr_mpls *)
804 				    info.rti_info[RTAX_SRC];
805 
806 				if (rt->rt_llinfo == NULL) {
807 					rt->rt_llinfo = (caddr_t)
808 					    malloc(sizeof(struct rt_mpls),
809 					    M_TEMP, M_NOWAIT|M_ZERO);
810 				}
811 				if (rt->rt_llinfo == NULL) {
812 					error = ENOMEM;
813 					goto flush;
814 				}
815 
816 				rt_mpls = (struct rt_mpls *)rt->rt_llinfo;
817 
818 				if (psa_mpls != NULL) {
819 					rt_mpls->mpls_label =
820 					    psa_mpls->smpls_label;
821 				}
822 
823 				rt_mpls->mpls_operation = info.rti_mpls;
824 
825 				/* XXX: set experimental bits */
826 
827 				rt->rt_flags |= RTF_MPLS;
828 			} else if (newgate || ((rtm->rtm_fmask & RTF_MPLS) &&
829 			    !(rtm->rtm_flags & RTF_MPLS))) {
830 				/* if gateway changed remove MPLS information */
831 				if (rt->rt_llinfo != NULL &&
832 				    rt->rt_flags & RTF_MPLS) {
833 					free(rt->rt_llinfo, M_TEMP);
834 					rt->rt_llinfo = NULL;
835 					rt->rt_flags &= ~RTF_MPLS;
836 				}
837 			}
838 #endif
839 			/* Hack to allow some flags to be toggled */
840 			if (rtm->rtm_fmask & RTF_FMASK)
841 				rt->rt_flags = (rt->rt_flags &
842 				    ~rtm->rtm_fmask) |
843 				    (rtm->rtm_flags & rtm->rtm_fmask);
844 
845 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
846 			    &rt->rt_rmx);
847 			rtm->rtm_index = rt->rt_ifp->if_index;
848 			rtm->rtm_priority = rt->rt_priority & RTP_MASK;
849 			rtm->rtm_flags = rt->rt_flags;
850 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
851 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
852 			if (genmask)
853 				rt->rt_genmask = genmask;
854 			if (info.rti_info[RTAX_LABEL] != NULL) {
855 				char *rtlabel = ((struct sockaddr_rtlabel *)
856 				    info.rti_info[RTAX_LABEL])->sr_label;
857 				rtlabel_unref(rt->rt_labelid);
858 				rt->rt_labelid =
859 				    rtlabel_name2id(rtlabel);
860 			}
861 			if_group_routechange(dst, netmask);
862 			/* FALLTHROUGH */
863 		case RTM_LOCK:
864 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
865 			rt->rt_rmx.rmx_locks |=
866 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
867 			rtm->rtm_priority = rt->rt_priority & RTP_MASK;
868 			break;
869 		}
870 		break;
871 	}
872 
873 flush:
874 	if (rtm) {
875 		if (error)
876 			rtm->rtm_errno = error;
877 		else {
878 			rtm->rtm_flags |= RTF_DONE;
879 		}
880 	}
881 	if (dst)
882 		route_proto.sp_protocol = dst->sa_family;
883 	if (rt)
884 		rtfree(rt);
885 
886 	/*
887 	 * Check to see if we don't want our own messages.
888 	 */
889 	if (!(so->so_options & SO_USELOOPBACK)) {
890 		if (route_cb.any_count <= 1) {
891 fail:
892 			if (rtm)
893 				Free(rtm);
894 			m_freem(m);
895 			return (error);
896 		}
897 		/* There is another listener, so construct message */
898 		rp = sotorawcb(so);
899 	}
900 	if (rp)
901 		rp->rcb_proto.sp_family = 0; /* Avoid us */
902 	if (rtm) {
903 		if (m_copyback(m, 0, rtm->rtm_msglen, rtm, M_NOWAIT)) {
904 			m_freem(m);
905 			m = NULL;
906 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
907 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
908 		Free(rtm);
909 	}
910 	if (m)
911 		route_input(m, &route_proto, &route_src, &route_dst);
912 	if (rp)
913 		rp->rcb_proto.sp_family = PF_ROUTE;
914 
915 	return (error);
916 }
917 
918 void
919 rt_setmetrics(u_long which, struct rt_metrics *in, struct rt_kmetrics *out)
920 {
921 	if (which & RTV_MTU)
922 		out->rmx_mtu = in->rmx_mtu;
923 	if (which & RTV_EXPIRE)
924 		out->rmx_expire = in->rmx_expire;
925 	/* RTV_PRIORITY handled befor */
926 }
927 
928 void
929 rt_getmetrics(struct rt_kmetrics *in, struct rt_metrics *out)
930 {
931 	bzero(out, sizeof(*out));
932 	out->rmx_locks = in->rmx_locks;
933 	out->rmx_mtu = in->rmx_mtu;
934 	out->rmx_expire = in->rmx_expire;
935 	out->rmx_pksent = in->rmx_pksent;
936 }
937 
938 #define ROUNDUP(a) \
939 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
940 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
941 
942 void
943 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
944 {
945 	struct sockaddr	*sa;
946 	int		 i;
947 
948 	bzero(rtinfo->rti_info, sizeof(rtinfo->rti_info));
949 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
950 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
951 			continue;
952 		rtinfo->rti_info[i] = sa = (struct sockaddr *)cp;
953 		ADVANCE(cp, sa);
954 	}
955 }
956 
957 struct mbuf *
958 rt_msg1(int type, struct rt_addrinfo *rtinfo)
959 {
960 	struct rt_msghdr	*rtm;
961 	struct mbuf		*m;
962 	int			 i;
963 	struct sockaddr		*sa;
964 	int			 len, dlen, hlen;
965 
966 	switch (type) {
967 	case RTM_DELADDR:
968 	case RTM_NEWADDR:
969 		len = sizeof(struct ifa_msghdr);
970 		break;
971 	case RTM_IFINFO:
972 		len = sizeof(struct if_msghdr);
973 		break;
974 	case RTM_IFANNOUNCE:
975 		len = sizeof(struct if_announcemsghdr);
976 		break;
977 	default:
978 		len = sizeof(struct rt_msghdr);
979 		break;
980 	}
981 	if (len > MCLBYTES)
982 		panic("rt_msg1");
983 	m = m_gethdr(M_DONTWAIT, MT_DATA);
984 	if (m && len > MHLEN) {
985 		MCLGET(m, M_DONTWAIT);
986 		if ((m->m_flags & M_EXT) == 0) {
987 			m_free(m);
988 			m = NULL;
989 		}
990 	}
991 	if (m == 0)
992 		return (m);
993 	m->m_pkthdr.len = m->m_len = hlen = len;
994 	m->m_pkthdr.rcvif = NULL;
995 	rtm = mtod(m, struct rt_msghdr *);
996 	bzero(rtm, len);
997 	for (i = 0; i < RTAX_MAX; i++) {
998 		if (rtinfo == NULL || (sa = rtinfo->rti_info[i]) == NULL)
999 			continue;
1000 		rtinfo->rti_addrs |= (1 << i);
1001 		dlen = ROUNDUP(sa->sa_len);
1002 		if (m_copyback(m, len, dlen, sa, M_NOWAIT)) {
1003 			m_freem(m);
1004 			return (NULL);
1005 		}
1006 		len += dlen;
1007 	}
1008 	rtm->rtm_msglen = len;
1009 	rtm->rtm_hdrlen = hlen;
1010 	rtm->rtm_version = RTM_VERSION;
1011 	rtm->rtm_type = type;
1012 	return (m);
1013 }
1014 
1015 int
1016 rt_msg2(int type, int vers, struct rt_addrinfo *rtinfo, caddr_t cp,
1017     struct walkarg *w)
1018 {
1019 	int		i;
1020 	int		len, dlen, hlen, second_time = 0;
1021 	caddr_t		cp0;
1022 
1023 	rtinfo->rti_addrs = 0;
1024 again:
1025 	switch (type) {
1026 	case RTM_DELADDR:
1027 	case RTM_NEWADDR:
1028 		len = sizeof(struct ifa_msghdr);
1029 		break;
1030 	case RTM_IFINFO:
1031 		len = sizeof(struct if_msghdr);
1032 		break;
1033 	default:
1034 		len = sizeof(struct rt_msghdr);
1035 		break;
1036 	}
1037 	hlen = len;
1038 	if ((cp0 = cp) != NULL)
1039 		cp += len;
1040 	for (i = 0; i < RTAX_MAX; i++) {
1041 		struct sockaddr *sa;
1042 
1043 		if ((sa = rtinfo->rti_info[i]) == 0)
1044 			continue;
1045 		rtinfo->rti_addrs |= (1 << i);
1046 		dlen = ROUNDUP(sa->sa_len);
1047 		if (cp) {
1048 			bcopy(sa, cp, (size_t)dlen);
1049 			cp += dlen;
1050 		}
1051 		len += dlen;
1052 	}
1053 	/* align message length to the next natural boundary */
1054 	len = ALIGN(len);
1055 	if (cp == 0 && w != NULL && !second_time) {
1056 		struct walkarg *rw = w;
1057 
1058 		rw->w_needed += len;
1059 		if (rw->w_needed <= 0 && rw->w_where) {
1060 			if (rw->w_tmemsize < len) {
1061 				if (rw->w_tmem)
1062 					free(rw->w_tmem, M_RTABLE);
1063 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
1064 				if (rw->w_tmem)
1065 					rw->w_tmemsize = len;
1066 			}
1067 			if (rw->w_tmem) {
1068 				cp = rw->w_tmem;
1069 				second_time = 1;
1070 				goto again;
1071 			} else
1072 				rw->w_where = 0;
1073 		}
1074 	}
1075 	if (cp && w)		/* clear the message header */
1076 		bzero(cp0, hlen);
1077 
1078 	if (cp) {
1079 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1080 
1081 		rtm->rtm_version = RTM_VERSION;
1082 		rtm->rtm_type = type;
1083 		rtm->rtm_msglen = len;
1084 		rtm->rtm_hdrlen = hlen;
1085 	}
1086 	return (len);
1087 }
1088 
1089 /*
1090  * This routine is called to generate a message from the routing
1091  * socket indicating that a redirect has occurred, a routing lookup
1092  * has failed, or that a protocol has detected timeouts to a particular
1093  * destination.
1094  */
1095 void
1096 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags,
1097     struct ifnet *ifp, int error, u_int tableid)
1098 {
1099 	struct rt_msghdr	*rtm;
1100 	struct mbuf		*m;
1101 	struct sockaddr		*sa = rtinfo->rti_info[RTAX_DST];
1102 
1103 	if (route_cb.any_count == 0)
1104 		return;
1105 	m = rt_msg1(type, rtinfo);
1106 	if (m == 0)
1107 		return;
1108 	rtm = mtod(m, struct rt_msghdr *);
1109 	rtm->rtm_flags = RTF_DONE | flags;
1110 	rtm->rtm_errno = error;
1111 	rtm->rtm_tableid = tableid;
1112 	rtm->rtm_addrs = rtinfo->rti_addrs;
1113 	if (ifp != NULL)
1114 		rtm->rtm_index = ifp->if_index;
1115 	if (sa == NULL)
1116 		route_proto.sp_protocol = 0;
1117 	else
1118 		route_proto.sp_protocol = sa->sa_family;
1119 	route_input(m, &route_proto, &route_src, &route_dst);
1120 }
1121 
1122 /*
1123  * This routine is called to generate a message from the routing
1124  * socket indicating that the status of a network interface has changed.
1125  */
1126 void
1127 rt_ifmsg(struct ifnet *ifp)
1128 {
1129 	struct if_msghdr	*ifm;
1130 	struct mbuf		*m;
1131 
1132 	if (route_cb.any_count == 0)
1133 		return;
1134 	m = rt_msg1(RTM_IFINFO, NULL);
1135 	if (m == 0)
1136 		return;
1137 	ifm = mtod(m, struct if_msghdr *);
1138 	ifm->ifm_index = ifp->if_index;
1139 	ifm->ifm_tableid = ifp->if_rdomain;
1140 	ifm->ifm_flags = ifp->if_flags;
1141 	ifm->ifm_xflags = ifp->if_xflags;
1142 	ifm->ifm_data = ifp->if_data;
1143 	ifm->ifm_addrs = 0;
1144 	route_proto.sp_protocol = 0;
1145 	route_input(m, &route_proto, &route_src, &route_dst);
1146 }
1147 
1148 /*
1149  * This is called to generate messages from the routing socket
1150  * indicating a network interface has had addresses associated with it.
1151  * if we ever reverse the logic and replace messages TO the routing
1152  * socket indicate a request to configure interfaces, then it will
1153  * be unnecessary as the routing socket will automatically generate
1154  * copies of it.
1155  */
1156 void
1157 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1158 {
1159 	struct rt_addrinfo	 info;
1160 	struct sockaddr		*sa = NULL;
1161 	int			 pass;
1162 	struct mbuf		*m = NULL;
1163 	struct ifnet		*ifp = ifa->ifa_ifp;
1164 
1165 	if (route_cb.any_count == 0)
1166 		return;
1167 	for (pass = 1; pass < 3; pass++) {
1168 		bzero(&info, sizeof(info));
1169 		if ((cmd == RTM_ADD && pass == 1) ||
1170 		    (cmd == RTM_DELETE && pass == 2)) {
1171 			struct ifa_msghdr	*ifam;
1172 			int			 ncmd;
1173 
1174 			if (cmd == RTM_ADD)
1175 				ncmd = RTM_NEWADDR;
1176 			else
1177 				ncmd = RTM_DELADDR;
1178 
1179 			ifaaddr = sa = ifa->ifa_addr;
1180 			ifpaddr = TAILQ_FIRST(&ifp->if_addrlist)->ifa_addr;
1181 			netmask = ifa->ifa_netmask;
1182 			brdaddr = ifa->ifa_dstaddr;
1183 			if ((m = rt_msg1(ncmd, &info)) == NULL)
1184 				continue;
1185 			ifam = mtod(m, struct ifa_msghdr *);
1186 			ifam->ifam_index = ifp->if_index;
1187 			ifam->ifam_metric = ifa->ifa_metric;
1188 			ifam->ifam_flags = ifa->ifa_flags;
1189 			ifam->ifam_addrs = info.rti_addrs;
1190 			ifam->ifam_tableid = ifp->if_rdomain;
1191 		}
1192 		if ((cmd == RTM_ADD && pass == 2) ||
1193 		    (cmd == RTM_DELETE && pass == 1)) {
1194 			struct rt_msghdr *rtm;
1195 
1196 			if (rt == 0)
1197 				continue;
1198 			netmask = rt_mask(rt);
1199 			dst = sa = rt_key(rt);
1200 			gate = rt->rt_gateway;
1201 			if ((m = rt_msg1(cmd, &info)) == NULL)
1202 				continue;
1203 			rtm = mtod(m, struct rt_msghdr *);
1204 			rtm->rtm_index = ifp->if_index;
1205 			rtm->rtm_flags |= rt->rt_flags;
1206 			rtm->rtm_priority = rt->rt_priority & RTP_MASK;
1207 			rtm->rtm_errno = error;
1208 			rtm->rtm_addrs = info.rti_addrs;
1209 			rtm->rtm_tableid = ifp->if_rdomain;
1210 		}
1211 		if (sa == NULL)
1212 			route_proto.sp_protocol = 0;
1213 		else
1214 			route_proto.sp_protocol = sa->sa_family;
1215 		route_input(m, &route_proto, &route_src, &route_dst);
1216 	}
1217 }
1218 
1219 /*
1220  * This is called to generate routing socket messages indicating
1221  * network interface arrival and departure.
1222  */
1223 void
1224 rt_ifannouncemsg(struct ifnet *ifp, int what)
1225 {
1226 	struct if_announcemsghdr	*ifan;
1227 	struct mbuf			*m;
1228 
1229 	if (route_cb.any_count == 0)
1230 		return;
1231 	m = rt_msg1(RTM_IFANNOUNCE, NULL);
1232 	if (m == 0)
1233 		return;
1234 	ifan = mtod(m, struct if_announcemsghdr *);
1235 	ifan->ifan_index = ifp->if_index;
1236 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof(ifan->ifan_name));
1237 	ifan->ifan_what = what;
1238 	route_proto.sp_protocol = 0;
1239 	route_input(m, &route_proto, &route_src, &route_dst);
1240 }
1241 
1242 /*
1243  * This is used in dumping the kernel table via sysctl().
1244  */
1245 int
1246 sysctl_dumpentry(struct radix_node *rn, void *v, u_int id)
1247 {
1248 	struct walkarg		*w = v;
1249 	struct rtentry		*rt = (struct rtentry *)rn;
1250 	int			 error = 0, size;
1251 	struct rt_addrinfo	 info;
1252 #ifdef MPLS
1253 	struct sockaddr_mpls	 sa_mpls;
1254 #endif
1255 	struct sockaddr_rtlabel	 sa_rt;
1256 	const char		*label;
1257 
1258 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1259 		return 0;
1260 	bzero(&info, sizeof(info));
1261 	dst = rt_key(rt);
1262 	gate = rt->rt_gateway;
1263 	netmask = rt_mask(rt);
1264 	genmask = rt->rt_genmask;
1265 	if (rt->rt_ifp) {
1266 		ifpaddr = TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
1267 		ifaaddr = rt->rt_ifa->ifa_addr;
1268 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1269 			brdaddr = rt->rt_ifa->ifa_dstaddr;
1270 	}
1271 	if (rt->rt_labelid) {
1272 		bzero(&sa_rt, sizeof(sa_rt));
1273 		sa_rt.sr_len = sizeof(sa_rt);
1274 		label = rtlabel_id2name(rt->rt_labelid);
1275 		if (label != NULL) {
1276 			strlcpy(sa_rt.sr_label, label,
1277 			    sizeof(sa_rt.sr_label));
1278 			info.rti_info[RTAX_LABEL] =
1279 			    (struct sockaddr *)&sa_rt;
1280 		}
1281 	}
1282 #ifdef MPLS
1283 	if (rt->rt_flags & RTF_MPLS) {
1284 		bzero(&sa_mpls, sizeof(sa_mpls));
1285 		sa_mpls.smpls_family = AF_MPLS;
1286 		sa_mpls.smpls_len = sizeof(sa_mpls);
1287 		sa_mpls.smpls_label = ((struct rt_mpls *)
1288 		    rt->rt_llinfo)->mpls_label;
1289 		info.rti_info[RTAX_SRC] = (struct sockaddr *)&sa_mpls;
1290 		info.rti_mpls = ((struct rt_mpls *)
1291 		    rt->rt_llinfo)->mpls_operation;
1292 	}
1293 #endif
1294 
1295 	size = rt_msg2(RTM_GET, RTM_VERSION, &info, NULL, w);
1296 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1297 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1298 
1299 		rtm->rtm_flags = rt->rt_flags;
1300 		rtm->rtm_priority = rt->rt_priority & RTP_MASK;
1301 		rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1302 		rtm->rtm_rmx.rmx_refcnt = rt->rt_refcnt;
1303 		rtm->rtm_index = rt->rt_ifp->if_index;
1304 		rtm->rtm_addrs = info.rti_addrs;
1305 		rtm->rtm_tableid = id;
1306 #ifdef MPLS
1307 		rtm->rtm_mpls = info.rti_mpls;
1308 #endif
1309 		if ((error = copyout(rtm, w->w_where, size)) != 0)
1310 			w->w_where = NULL;
1311 		else
1312 			w->w_where += size;
1313 	}
1314 	return (error);
1315 }
1316 
1317 int
1318 sysctl_iflist(int af, struct walkarg *w)
1319 {
1320 	struct ifnet		*ifp;
1321 	struct ifaddr		*ifa;
1322 	struct rt_addrinfo	 info;
1323 	int			 len, error = 0;
1324 
1325 	bzero(&info, sizeof(info));
1326 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1327 		if (w->w_arg && w->w_arg != ifp->if_index)
1328 			continue;
1329 		ifa = TAILQ_FIRST(&ifp->if_addrlist);
1330 		if (!ifa)
1331 			continue;
1332 		ifpaddr = ifa->ifa_addr;
1333 		len = rt_msg2(RTM_IFINFO, RTM_VERSION, &info, 0, w);
1334 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1335 			struct if_msghdr *ifm;
1336 
1337 			ifm = (struct if_msghdr *)w->w_tmem;
1338 			ifm->ifm_index = ifp->if_index;
1339 			ifm->ifm_tableid = ifp->if_rdomain;
1340 			ifm->ifm_flags = ifp->if_flags;
1341 			ifm->ifm_data = ifp->if_data;
1342 			ifm->ifm_addrs = info.rti_addrs;
1343 			error = copyout(ifm, w->w_where, len);
1344 			if (error)
1345 				return (error);
1346 			w->w_where += len;
1347 		}
1348 		ifpaddr = 0;
1349 		while ((ifa = TAILQ_NEXT(ifa, ifa_list)) !=
1350 		    TAILQ_END(&ifp->if_addrlist)) {
1351 			if (af && af != ifa->ifa_addr->sa_family)
1352 				continue;
1353 			ifaaddr = ifa->ifa_addr;
1354 			netmask = ifa->ifa_netmask;
1355 			brdaddr = ifa->ifa_dstaddr;
1356 			len = rt_msg2(RTM_NEWADDR, RTM_VERSION, &info, 0, w);
1357 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1358 				struct ifa_msghdr *ifam;
1359 
1360 				ifam = (struct ifa_msghdr *)w->w_tmem;
1361 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1362 				ifam->ifam_flags = ifa->ifa_flags;
1363 				ifam->ifam_metric = ifa->ifa_metric;
1364 				ifam->ifam_addrs = info.rti_addrs;
1365 				error = copyout(w->w_tmem, w->w_where, len);
1366 				if (error)
1367 					return (error);
1368 				w->w_where += len;
1369 			}
1370 		}
1371 		ifaaddr = netmask = brdaddr = 0;
1372 	}
1373 	return (0);
1374 }
1375 
1376 int
1377 sysctl_rtable(int *name, u_int namelen, void *where, size_t *given, void *new,
1378     size_t newlen)
1379 {
1380 	struct radix_node_head	*rnh;
1381 	int			 i, s, error = EINVAL;
1382 	u_char  		 af;
1383 	struct walkarg		 w;
1384 	struct rt_tableinfo	 tableinfo;
1385 	u_int			 tableid = 0;
1386 
1387 	if (new)
1388 		return (EPERM);
1389 	if (namelen < 3 || namelen > 4)
1390 		return (EINVAL);
1391 	af = name[0];
1392 	bzero(&w, sizeof(w));
1393 	w.w_where = where;
1394 	w.w_given = *given;
1395 	w.w_needed = 0 - w.w_given;
1396 	w.w_op = name[1];
1397 	w.w_arg = name[2];
1398 
1399 	if (namelen == 4) {
1400 		tableid = name[3];
1401 		if (!rtable_exists(tableid))
1402 			return (ENOENT);
1403 	} else
1404 		tableid = curproc->p_p->ps_rtableid;
1405 
1406 	s = splsoftnet();
1407 	switch (w.w_op) {
1408 
1409 	case NET_RT_DUMP:
1410 	case NET_RT_FLAGS:
1411 		for (i = 1; i <= AF_MAX; i++)
1412 			if ((rnh = rt_gettable(i, tableid)) != NULL &&
1413 			    (af == 0 || af == i) &&
1414 			    (error = (*rnh->rnh_walktree)(rnh,
1415 			    sysctl_dumpentry, &w)))
1416 				break;
1417 		break;
1418 
1419 	case NET_RT_IFLIST:
1420 		error = sysctl_iflist(af, &w);
1421 		break;
1422 
1423 	case NET_RT_STATS:
1424 		error = sysctl_rdstruct(where, given, new,
1425 		    &rtstat, sizeof(rtstat));
1426 		splx(s);
1427 		return (error);
1428 	case NET_RT_TABLE:
1429 		tableid = w.w_arg;
1430 		if (!rtable_exists(tableid)) {
1431 			splx(s);
1432 			return (ENOENT);
1433 		}
1434 		tableinfo.rti_tableid = tableid;
1435 		tableinfo.rti_domainid = rtable_l2(tableid);
1436 		error = sysctl_rdstruct(where, given, new,
1437 		    &tableinfo, sizeof(tableinfo));
1438 		splx(s);
1439 		return (error);
1440 	}
1441 	splx(s);
1442 	if (w.w_tmem)
1443 		free(w.w_tmem, M_RTABLE);
1444 	w.w_needed += w.w_given;
1445 	if (where) {
1446 		*given = w.w_where - (caddr_t)where;
1447 		if (*given < w.w_needed)
1448 			return (ENOMEM);
1449 	} else
1450 		*given = (11 * w.w_needed) / 10;
1451 
1452 	return (error);
1453 }
1454 
1455 /*
1456  * Definitions of protocols supported in the ROUTE domain.
1457  */
1458 
1459 extern	struct domain routedomain;		/* or at least forward */
1460 
1461 struct protosw routesw[] = {
1462 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR|PR_WANTRCVD,
1463   route_input,	route_output,	raw_ctlinput,	route_ctloutput,
1464   route_usrreq,
1465   raw_init,	0,		0,		0,
1466   sysctl_rtable,
1467 }
1468 };
1469 
1470 struct domain routedomain =
1471     { PF_ROUTE, "route", route_init, 0, 0,
1472       routesw, &routesw[nitems(routesw)] };
1473