xref: /dflybsd-src/sys/net/rtsock.c (revision e19e5bbc20dd1d64f1833c5d0ac7a605c8e9bfa0)
1 /*
2  * Copyright (c) 2004, 2005 The DragonFly Project.  All rights reserved.
3  *
4  * This code is derived from software contributed to The DragonFly Project
5  * by Jeffrey M. Hsu.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of The DragonFly Project nor the names of its
16  *    contributors may be used to endorse or promote products derived
17  *    from this software without specific, prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
22  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
23  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
25  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
27  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
28  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
29  * 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	8.7 (Berkeley) 10/12/95
62  * $FreeBSD: src/sys/net/rtsock.c,v 1.44.2.11 2002/12/04 14:05:41 ru Exp $
63  */
64 
65 #include "opt_sctp.h"
66 
67 #include <sys/param.h>
68 #include <sys/systm.h>
69 #include <sys/kernel.h>
70 #include <sys/sysctl.h>
71 #include <sys/proc.h>
72 #include <sys/priv.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/domain.h>
79 
80 #include <sys/thread2.h>
81 #include <sys/socketvar2.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 #include <net/raw_cb.h>
86 #include <net/netmsg2.h>
87 #include <net/netisr2.h>
88 
89 #ifdef SCTP
90 extern void sctp_add_ip_address(struct ifaddr *ifa);
91 extern void sctp_delete_ip_address(struct ifaddr *ifa);
92 #endif /* SCTP */
93 
94 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
95 
96 static struct route_cb {
97 	int	ip_count;
98 	int	ip6_count;
99 	int	ns_count;
100 	int	any_count;
101 } route_cb;
102 
103 static const struct sockaddr route_src = { 2, PF_ROUTE, };
104 
105 struct walkarg {
106 	int	w_tmemsize;
107 	int	w_op, w_arg;
108 	void	*w_tmem;
109 	struct sysctl_req *w_req;
110 };
111 
112 static struct mbuf *
113 		rt_msg_mbuf (int, struct rt_addrinfo *);
114 static void	rt_msg_buffer (int, struct rt_addrinfo *, void *buf, int len);
115 static int	rt_msgsize (int type, struct rt_addrinfo *rtinfo);
116 static int	rt_xaddrs (char *, char *, struct rt_addrinfo *);
117 static int	sysctl_dumpentry (struct radix_node *rn, void *vw);
118 static int	sysctl_iflist (int af, struct walkarg *w);
119 static int	route_output(struct mbuf *, struct socket *, ...);
120 static void	rt_setmetrics (u_long, struct rt_metrics *,
121 			       struct rt_metrics *);
122 
123 /*
124  * It really doesn't make any sense at all for this code to share much
125  * with raw_usrreq.c, since its functionality is so restricted.  XXX
126  */
127 static void
128 rts_abort(netmsg_t msg)
129 {
130 	crit_enter();
131 	raw_usrreqs.pru_abort(msg);
132 	/* msg invalid now */
133 	crit_exit();
134 }
135 
136 /* pru_accept is EOPNOTSUPP */
137 
138 static void
139 rts_attach(netmsg_t msg)
140 {
141 	struct socket *so = msg->base.nm_so;
142 	struct pru_attach_info *ai = msg->attach.nm_ai;
143 	struct rawcb *rp;
144 	int proto = msg->attach.nm_proto;
145 	int error;
146 
147 	crit_enter();
148 	if (sotorawcb(so) != NULL) {
149 		error = EISCONN;
150 		goto done;
151 	}
152 
153 	rp = kmalloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
154 
155 	/*
156 	 * The critical section is necessary to block protocols from sending
157 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
158 	 * this PCB is extant but incompletely initialized.
159 	 * Probably we should try to do more of this work beforehand and
160 	 * eliminate the critical section.
161 	 */
162 	so->so_pcb = rp;
163 	soreference(so);	/* so_pcb assignment */
164 	error = raw_attach(so, proto, ai->sb_rlimit);
165 	rp = sotorawcb(so);
166 	if (error) {
167 		kfree(rp, M_PCB);
168 		goto done;
169 	}
170 	switch(rp->rcb_proto.sp_protocol) {
171 	case AF_INET:
172 		route_cb.ip_count++;
173 		break;
174 	case AF_INET6:
175 		route_cb.ip6_count++;
176 		break;
177 	}
178 	rp->rcb_faddr = &route_src;
179 	route_cb.any_count++;
180 	soisconnected(so);
181 	so->so_options |= SO_USELOOPBACK;
182 	error = 0;
183 done:
184 	crit_exit();
185 	lwkt_replymsg(&msg->lmsg, error);
186 }
187 
188 static void
189 rts_bind(netmsg_t msg)
190 {
191 	crit_enter();
192 	raw_usrreqs.pru_bind(msg); /* xxx just EINVAL */
193 	/* msg invalid now */
194 	crit_exit();
195 }
196 
197 static void
198 rts_connect(netmsg_t msg)
199 {
200 	crit_enter();
201 	raw_usrreqs.pru_connect(msg); /* XXX just EINVAL */
202 	/* msg invalid now */
203 	crit_exit();
204 }
205 
206 /* pru_connect2 is EOPNOTSUPP */
207 /* pru_control is EOPNOTSUPP */
208 
209 static void
210 rts_detach(netmsg_t msg)
211 {
212 	struct socket *so = msg->base.nm_so;
213 	struct rawcb *rp = sotorawcb(so);
214 
215 	crit_enter();
216 	if (rp != NULL) {
217 		switch(rp->rcb_proto.sp_protocol) {
218 		case AF_INET:
219 			route_cb.ip_count--;
220 			break;
221 		case AF_INET6:
222 			route_cb.ip6_count--;
223 			break;
224 		}
225 		route_cb.any_count--;
226 	}
227 	raw_usrreqs.pru_detach(msg);
228 	/* msg invalid now */
229 	crit_exit();
230 }
231 
232 static void
233 rts_disconnect(netmsg_t msg)
234 {
235 	crit_enter();
236 	raw_usrreqs.pru_disconnect(msg);
237 	/* msg invalid now */
238 	crit_exit();
239 }
240 
241 /* pru_listen is EOPNOTSUPP */
242 
243 static void
244 rts_peeraddr(netmsg_t msg)
245 {
246 	crit_enter();
247 	raw_usrreqs.pru_peeraddr(msg);
248 	/* msg invalid now */
249 	crit_exit();
250 }
251 
252 /* pru_rcvd is EOPNOTSUPP */
253 /* pru_rcvoob is EOPNOTSUPP */
254 
255 static void
256 rts_send(netmsg_t msg)
257 {
258 	crit_enter();
259 	raw_usrreqs.pru_send(msg);
260 	/* msg invalid now */
261 	crit_exit();
262 }
263 
264 /* pru_sense is null */
265 
266 static void
267 rts_shutdown(netmsg_t msg)
268 {
269 	crit_enter();
270 	raw_usrreqs.pru_shutdown(msg);
271 	/* msg invalid now */
272 	crit_exit();
273 }
274 
275 static void
276 rts_sockaddr(netmsg_t msg)
277 {
278 	crit_enter();
279 	raw_usrreqs.pru_sockaddr(msg);
280 	/* msg invalid now */
281 	crit_exit();
282 }
283 
284 static struct pr_usrreqs route_usrreqs = {
285 	.pru_abort = rts_abort,
286 	.pru_accept = pr_generic_notsupp,
287 	.pru_attach = rts_attach,
288 	.pru_bind = rts_bind,
289 	.pru_connect = rts_connect,
290 	.pru_connect2 = pr_generic_notsupp,
291 	.pru_control = pr_generic_notsupp,
292 	.pru_detach = rts_detach,
293 	.pru_disconnect = rts_disconnect,
294 	.pru_listen = pr_generic_notsupp,
295 	.pru_peeraddr = rts_peeraddr,
296 	.pru_rcvd = pr_generic_notsupp,
297 	.pru_rcvoob = pr_generic_notsupp,
298 	.pru_send = rts_send,
299 	.pru_sense = pru_sense_null,
300 	.pru_shutdown = rts_shutdown,
301 	.pru_sockaddr = rts_sockaddr,
302 	.pru_sosend = sosend,
303 	.pru_soreceive = soreceive
304 };
305 
306 static __inline sa_family_t
307 familyof(struct sockaddr *sa)
308 {
309 	return (sa != NULL ? sa->sa_family : 0);
310 }
311 
312 /*
313  * Routing socket input function.  The packet must be serialized onto cpu 0.
314  * We use the cpu0_soport() netisr processing loop to handle it.
315  *
316  * This looks messy but it means that anyone, including interrupt code,
317  * can send a message to the routing socket.
318  */
319 static void
320 rts_input_handler(netmsg_t msg)
321 {
322 	static const struct sockaddr route_dst = { 2, PF_ROUTE, };
323 	struct sockproto route_proto;
324 	struct netmsg_packet *pmsg = &msg->packet;
325 	struct mbuf *m;
326 	sa_family_t family;
327 	struct rawcb *skip;
328 
329 	family = pmsg->base.lmsg.u.ms_result;
330 	route_proto.sp_family = PF_ROUTE;
331 	route_proto.sp_protocol = family;
332 
333 	m = pmsg->nm_packet;
334 	M_ASSERTPKTHDR(m);
335 
336 	skip = m->m_pkthdr.header;
337 	m->m_pkthdr.header = NULL;
338 
339 	raw_input(m, &route_proto, &route_src, &route_dst, skip);
340 }
341 
342 static void
343 rts_input_skip(struct mbuf *m, sa_family_t family, struct rawcb *skip)
344 {
345 	struct netmsg_packet *pmsg;
346 	lwkt_port_t port;
347 
348 	M_ASSERTPKTHDR(m);
349 
350 	port = netisr_cpuport(0);	/* XXX same as for routing socket */
351 	pmsg = &m->m_hdr.mh_netmsg;
352 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
353 		    0, rts_input_handler);
354 	pmsg->nm_packet = m;
355 	pmsg->base.lmsg.u.ms_result = family;
356 	m->m_pkthdr.header = skip; /* XXX steal field in pkthdr */
357 	lwkt_sendmsg(port, &pmsg->base.lmsg);
358 }
359 
360 static __inline void
361 rts_input(struct mbuf *m, sa_family_t family)
362 {
363 	rts_input_skip(m, family, NULL);
364 }
365 
366 static void *
367 reallocbuf_nofree(void *ptr, size_t len, size_t olen)
368 {
369 	void *newptr;
370 
371 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
372 	if (newptr == NULL)
373 		return NULL;
374 	bcopy(ptr, newptr, olen);
375 	return (newptr);
376 }
377 
378 /*
379  * Internal helper routine for route_output().
380  */
381 static int
382 _fillrtmsg(struct rt_msghdr **prtm, struct rtentry *rt,
383 	   struct rt_addrinfo *rtinfo)
384 {
385 	int msglen;
386 	struct rt_msghdr *rtm = *prtm;
387 
388 	/* Fill in rt_addrinfo for call to rt_msg_buffer(). */
389 	rtinfo->rti_dst = rt_key(rt);
390 	rtinfo->rti_gateway = rt->rt_gateway;
391 	rtinfo->rti_netmask = rt_mask(rt);		/* might be NULL */
392 	rtinfo->rti_genmask = rt->rt_genmask;		/* might be NULL */
393 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
394 		if (rt->rt_ifp != NULL) {
395 			rtinfo->rti_ifpaddr =
396 			    TAILQ_FIRST(&rt->rt_ifp->if_addrheads[mycpuid])
397 			    ->ifa->ifa_addr;
398 			rtinfo->rti_ifaaddr = rt->rt_ifa->ifa_addr;
399 			if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
400 				rtinfo->rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
401 			rtm->rtm_index = rt->rt_ifp->if_index;
402 		} else {
403 			rtinfo->rti_ifpaddr = NULL;
404 			rtinfo->rti_ifaaddr = NULL;
405 		}
406 	} else if (rt->rt_ifp != NULL) {
407 		rtm->rtm_index = rt->rt_ifp->if_index;
408 	}
409 
410 	msglen = rt_msgsize(rtm->rtm_type, rtinfo);
411 	if (rtm->rtm_msglen < msglen) {
412 		/* NOTE: Caller will free the old rtm accordingly */
413 		rtm = reallocbuf_nofree(rtm, msglen, rtm->rtm_msglen);
414 		if (rtm == NULL)
415 			return (ENOBUFS);
416 		*prtm = rtm;
417 	}
418 	rt_msg_buffer(rtm->rtm_type, rtinfo, rtm, msglen);
419 
420 	rtm->rtm_flags = rt->rt_flags;
421 	rtm->rtm_rmx = rt->rt_rmx;
422 	rtm->rtm_addrs = rtinfo->rti_addrs;
423 
424 	return (0);
425 }
426 
427 struct rtm_arg {
428 	struct rt_msghdr	*bak_rtm;
429 	struct rt_msghdr	*new_rtm;
430 };
431 
432 static int
433 fillrtmsg(struct rtm_arg *arg, struct rtentry *rt,
434 	  struct rt_addrinfo *rtinfo)
435 {
436 	struct rt_msghdr *rtm = arg->new_rtm;
437 	int error;
438 
439 	error = _fillrtmsg(&rtm, rt, rtinfo);
440 	if (!error) {
441 		if (arg->new_rtm != rtm) {
442 			/*
443 			 * _fillrtmsg() just allocated a new rtm;
444 			 * if the previously allocated rtm is not
445 			 * the backing rtm, it should be freed.
446 			 */
447 			if (arg->new_rtm != arg->bak_rtm)
448 				kfree(arg->new_rtm, M_RTABLE);
449 			arg->new_rtm = rtm;
450 		}
451 	}
452 	return error;
453 }
454 
455 static void route_output_add_callback(int, int, struct rt_addrinfo *,
456 					struct rtentry *, void *);
457 static void route_output_delete_callback(int, int, struct rt_addrinfo *,
458 					struct rtentry *, void *);
459 static int route_output_get_callback(int, struct rt_addrinfo *,
460 				     struct rtentry *, void *, int);
461 static int route_output_change_callback(int, struct rt_addrinfo *,
462 					struct rtentry *, void *, int);
463 static int route_output_lock_callback(int, struct rt_addrinfo *,
464 				      struct rtentry *, void *, int);
465 
466 /*ARGSUSED*/
467 static int
468 route_output(struct mbuf *m, struct socket *so, ...)
469 {
470 	struct rtm_arg arg;
471 	struct rt_msghdr *rtm = NULL;
472 	struct rawcb *rp = NULL;
473 	struct pr_output_info *oi;
474 	struct rt_addrinfo rtinfo;
475 	sa_family_t family;
476 	int len, error = 0;
477 	__va_list ap;
478 
479 	M_ASSERTPKTHDR(m);
480 
481 	__va_start(ap, so);
482 	oi = __va_arg(ap, struct pr_output_info *);
483 	__va_end(ap);
484 
485 	family = familyof(NULL);
486 
487 #define gotoerr(e) { error = e; goto flush;}
488 
489 	if (m == NULL ||
490 	    (m->m_len < sizeof(long) &&
491 	     (m = m_pullup(m, sizeof(long))) == NULL))
492 		return (ENOBUFS);
493 	len = m->m_pkthdr.len;
494 	if (len < sizeof(struct rt_msghdr) ||
495 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
496 		gotoerr(EINVAL);
497 
498 	rtm = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
499 	if (rtm == NULL)
500 		gotoerr(ENOBUFS);
501 
502 	m_copydata(m, 0, len, (caddr_t)rtm);
503 	if (rtm->rtm_version != RTM_VERSION)
504 		gotoerr(EPROTONOSUPPORT);
505 
506 	rtm->rtm_pid = oi->p_pid;
507 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
508 	rtinfo.rti_addrs = rtm->rtm_addrs;
509 	if (rt_xaddrs((char *)(rtm + 1), (char *)rtm + len, &rtinfo) != 0)
510 		gotoerr(EINVAL);
511 
512 	rtinfo.rti_flags = rtm->rtm_flags;
513 	if (rtinfo.rti_dst == NULL || rtinfo.rti_dst->sa_family >= AF_MAX ||
514 	    (rtinfo.rti_gateway && rtinfo.rti_gateway->sa_family >= AF_MAX))
515 		gotoerr(EINVAL);
516 
517 	family = familyof(rtinfo.rti_dst);
518 
519 	/*
520 	 * Verify that the caller has the appropriate privilege; RTM_GET
521 	 * is the only operation the non-superuser is allowed.
522 	 */
523 	if (rtm->rtm_type != RTM_GET &&
524 	    priv_check_cred(so->so_cred, PRIV_ROOT, 0) != 0)
525 		gotoerr(EPERM);
526 
527 	if (rtinfo.rti_genmask != NULL) {
528 		error = rtmask_add_global(rtinfo.rti_genmask,
529 		    rtm->rtm_type != RTM_GET ?
530 		    RTREQ_PRIO_HIGH : RTREQ_PRIO_NORM);
531 		if (error)
532 			goto flush;
533 	}
534 
535 	switch (rtm->rtm_type) {
536 	case RTM_ADD:
537 		if (rtinfo.rti_gateway == NULL) {
538 			error = EINVAL;
539 		} else {
540 			error = rtrequest1_global(RTM_ADD, &rtinfo,
541 			    route_output_add_callback, rtm, RTREQ_PRIO_HIGH);
542 		}
543 		break;
544 	case RTM_DELETE:
545 		/*
546 		 * Backing rtm (bak_rtm) could _not_ be freed during
547 		 * rtrequest1_global or rtsearch_global, even if the
548 		 * callback reallocates the rtm due to its size changes,
549 		 * since rtinfo points to the backing rtm's memory area.
550 		 * After rtrequest1_global or rtsearch_global returns,
551 		 * it is safe to free the backing rtm, since rtinfo will
552 		 * not be used anymore.
553 		 *
554 		 * new_rtm will be used to save the new rtm allocated
555 		 * by rtrequest1_global or rtsearch_global.
556 		 */
557 		arg.bak_rtm = rtm;
558 		arg.new_rtm = rtm;
559 		error = rtrequest1_global(RTM_DELETE, &rtinfo,
560 		    route_output_delete_callback, &arg, RTREQ_PRIO_HIGH);
561 		rtm = arg.new_rtm;
562 		if (rtm != arg.bak_rtm)
563 			kfree(arg.bak_rtm, M_RTABLE);
564 		break;
565 	case RTM_GET:
566 		/* See the comment in RTM_DELETE */
567 		arg.bak_rtm = rtm;
568 		arg.new_rtm = rtm;
569 		error = rtsearch_global(RTM_GET, &rtinfo,
570 		    route_output_get_callback, &arg, RTS_NOEXACTMATCH,
571 		    RTREQ_PRIO_NORM);
572 		rtm = arg.new_rtm;
573 		if (rtm != arg.bak_rtm)
574 			kfree(arg.bak_rtm, M_RTABLE);
575 		break;
576 	case RTM_CHANGE:
577 		error = rtsearch_global(RTM_CHANGE, &rtinfo,
578 		    route_output_change_callback, rtm, RTS_EXACTMATCH,
579 		    RTREQ_PRIO_HIGH);
580 		break;
581 	case RTM_LOCK:
582 		error = rtsearch_global(RTM_LOCK, &rtinfo,
583 		    route_output_lock_callback, rtm, RTS_EXACTMATCH,
584 		    RTREQ_PRIO_HIGH);
585 		break;
586 	default:
587 		error = EOPNOTSUPP;
588 		break;
589 	}
590 flush:
591 	if (rtm != NULL) {
592 		if (error != 0)
593 			rtm->rtm_errno = error;
594 		else
595 			rtm->rtm_flags |= RTF_DONE;
596 	}
597 
598 	/*
599 	 * Check to see if we don't want our own messages.
600 	 */
601 	if (!(so->so_options & SO_USELOOPBACK)) {
602 		if (route_cb.any_count <= 1) {
603 			if (rtm != NULL)
604 				kfree(rtm, M_RTABLE);
605 			m_freem(m);
606 			return (error);
607 		}
608 		/* There is another listener, so construct message */
609 		rp = sotorawcb(so);
610 	}
611 	if (rtm != NULL) {
612 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
613 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
614 			m_freem(m);
615 			m = NULL;
616 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
617 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
618 		kfree(rtm, M_RTABLE);
619 	}
620 	if (m != NULL)
621 		rts_input_skip(m, family, rp);
622 	return (error);
623 }
624 
625 static void
626 route_output_add_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
627 			  struct rtentry *rt, void *arg)
628 {
629 	struct rt_msghdr *rtm = arg;
630 
631 	if (error == 0 && rt != NULL) {
632 		rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
633 		    &rt->rt_rmx);
634 		rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
635 		rt->rt_rmx.rmx_locks |=
636 		    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
637 		if (rtinfo->rti_genmask != NULL) {
638 			rt->rt_genmask = rtmask_purelookup(rtinfo->rti_genmask);
639 			if (rt->rt_genmask == NULL) {
640 				/*
641 				 * This should not happen, since we
642 				 * have already installed genmask
643 				 * on each CPU before we reach here.
644 				 */
645 				panic("genmask is gone!?");
646 			}
647 		} else {
648 			rt->rt_genmask = NULL;
649 		}
650 		rtm->rtm_index = rt->rt_ifp->if_index;
651 	}
652 }
653 
654 static void
655 route_output_delete_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
656 			  struct rtentry *rt, void *arg)
657 {
658 	if (error == 0 && rt) {
659 		++rt->rt_refcnt;
660 		if (fillrtmsg(arg, rt, rtinfo) != 0) {
661 			error = ENOBUFS;
662 			/* XXX no way to return the error */
663 		}
664 		--rt->rt_refcnt;
665 	}
666 	if (rt && rt->rt_refcnt == 0) {
667 		++rt->rt_refcnt;
668 		rtfree(rt);
669 	}
670 }
671 
672 static int
673 route_output_get_callback(int cmd, struct rt_addrinfo *rtinfo,
674 			  struct rtentry *rt, void *arg, int found_cnt)
675 {
676 	int error, found = 0;
677 
678 	if (((rtinfo->rti_flags ^ rt->rt_flags) & RTF_HOST) == 0)
679 		found = 1;
680 
681 	error = fillrtmsg(arg, rt, rtinfo);
682 	if (!error && found) {
683 		/* Got the exact match, we could return now! */
684 		error = EJUSTRETURN;
685 	}
686 	return error;
687 }
688 
689 static int
690 route_output_change_callback(int cmd, struct rt_addrinfo *rtinfo,
691 			     struct rtentry *rt, void *arg, int found_cnt)
692 {
693 	struct rt_msghdr *rtm = arg;
694 	struct ifaddr *ifa;
695 	int error = 0;
696 
697 	/*
698 	 * new gateway could require new ifaddr, ifp;
699 	 * flags may also be different; ifp may be specified
700 	 * by ll sockaddr when protocol address is ambiguous
701 	 */
702 	if (((rt->rt_flags & RTF_GATEWAY) && rtinfo->rti_gateway != NULL) ||
703 	    rtinfo->rti_ifpaddr != NULL ||
704 	    (rtinfo->rti_ifaaddr != NULL &&
705 	     !sa_equal(rtinfo->rti_ifaaddr, rt->rt_ifa->ifa_addr))) {
706 		error = rt_getifa(rtinfo);
707 		if (error != 0)
708 			goto done;
709 	}
710 	if (rtinfo->rti_gateway != NULL) {
711 		/*
712 		 * We only need to generate rtmsg upon the
713 		 * first route to be changed.
714 		 */
715 		error = rt_setgate(rt, rt_key(rt), rtinfo->rti_gateway,
716 			found_cnt == 1 ? RTL_REPORTMSG : RTL_DONTREPORT);
717 		if (error != 0)
718 			goto done;
719 	}
720 	if ((ifa = rtinfo->rti_ifa) != NULL) {
721 		struct ifaddr *oifa = rt->rt_ifa;
722 
723 		if (oifa != ifa) {
724 			if (oifa && oifa->ifa_rtrequest)
725 				oifa->ifa_rtrequest(RTM_DELETE, rt);
726 			IFAFREE(rt->rt_ifa);
727 			IFAREF(ifa);
728 			rt->rt_ifa = ifa;
729 			rt->rt_ifp = rtinfo->rti_ifp;
730 		}
731 	}
732 	rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, &rt->rt_rmx);
733 	if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
734 		rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt);
735 	if (rtinfo->rti_genmask != NULL) {
736 		rt->rt_genmask = rtmask_purelookup(rtinfo->rti_genmask);
737 		if (rt->rt_genmask == NULL) {
738 			/*
739 			 * This should not happen, since we
740 			 * have already installed genmask
741 			 * on each CPU before we reach here.
742 			 */
743 			panic("genmask is gone!?");
744 		}
745 	}
746 	rtm->rtm_index = rt->rt_ifp->if_index;
747 done:
748 	return error;
749 }
750 
751 static int
752 route_output_lock_callback(int cmd, struct rt_addrinfo *rtinfo,
753 			   struct rtentry *rt, void *arg,
754 			   int found_cnt __unused)
755 {
756 	struct rt_msghdr *rtm = arg;
757 
758 	rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
759 	rt->rt_rmx.rmx_locks |=
760 		(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
761 	return 0;
762 }
763 
764 static void
765 rt_setmetrics(u_long which, struct rt_metrics *in, struct rt_metrics *out)
766 {
767 #define setmetric(flag, elt) if (which & (flag)) out->elt = in->elt;
768 	setmetric(RTV_RPIPE, rmx_recvpipe);
769 	setmetric(RTV_SPIPE, rmx_sendpipe);
770 	setmetric(RTV_SSTHRESH, rmx_ssthresh);
771 	setmetric(RTV_RTT, rmx_rtt);
772 	setmetric(RTV_RTTVAR, rmx_rttvar);
773 	setmetric(RTV_HOPCOUNT, rmx_hopcount);
774 	setmetric(RTV_MTU, rmx_mtu);
775 	setmetric(RTV_EXPIRE, rmx_expire);
776 	setmetric(RTV_MSL, rmx_msl);
777 	setmetric(RTV_IWMAXSEGS, rmx_iwmaxsegs);
778 	setmetric(RTV_IWCAPSEGS, rmx_iwcapsegs);
779 #undef setmetric
780 }
781 
782 #define ROUNDUP(a) \
783 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
784 
785 /*
786  * Extract the addresses of the passed sockaddrs.
787  * Do a little sanity checking so as to avoid bad memory references.
788  * This data is derived straight from userland.
789  */
790 static int
791 rt_xaddrs(char *cp, char *cplim, struct rt_addrinfo *rtinfo)
792 {
793 	struct sockaddr *sa;
794 	int i;
795 
796 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
797 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
798 			continue;
799 		sa = (struct sockaddr *)cp;
800 		/*
801 		 * It won't fit.
802 		 */
803 		if ((cp + sa->sa_len) > cplim) {
804 			return (EINVAL);
805 		}
806 
807 		/*
808 		 * There are no more...  Quit now.
809 		 * If there are more bits, they are in error.
810 		 * I've seen this.  route(1) can evidently generate these.
811 		 * This causes kernel to core dump.
812 		 * For compatibility, if we see this, point to a safe address.
813 		 */
814 		if (sa->sa_len == 0) {
815 			static struct sockaddr sa_zero = {
816 				sizeof sa_zero, AF_INET,
817 			};
818 
819 			rtinfo->rti_info[i] = &sa_zero;
820 			kprintf("rtsock: received more addr bits than sockaddrs.\n");
821 			return (0); /* should be EINVAL but for compat */
822 		}
823 
824 		/* Accept the sockaddr. */
825 		rtinfo->rti_info[i] = sa;
826 		cp += ROUNDUP(sa->sa_len);
827 	}
828 	return (0);
829 }
830 
831 static int
832 rt_msghdrsize(int type)
833 {
834 	switch (type) {
835 	case RTM_DELADDR:
836 	case RTM_NEWADDR:
837 		return sizeof(struct ifa_msghdr);
838 	case RTM_DELMADDR:
839 	case RTM_NEWMADDR:
840 		return sizeof(struct ifma_msghdr);
841 	case RTM_IFINFO:
842 		return sizeof(struct if_msghdr);
843 	case RTM_IFANNOUNCE:
844 	case RTM_IEEE80211:
845 		return sizeof(struct if_announcemsghdr);
846 	default:
847 		return sizeof(struct rt_msghdr);
848 	}
849 }
850 
851 static int
852 rt_msgsize(int type, struct rt_addrinfo *rtinfo)
853 {
854 	int len, i;
855 
856 	len = rt_msghdrsize(type);
857 	for (i = 0; i < RTAX_MAX; i++) {
858 		if (rtinfo->rti_info[i] != NULL)
859 			len += ROUNDUP(rtinfo->rti_info[i]->sa_len);
860 	}
861 	len = ALIGN(len);
862 	return len;
863 }
864 
865 /*
866  * Build a routing message in a buffer.
867  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
868  * to the end of the buffer after the message header.
869  *
870  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
871  * This side-effect can be avoided if we reorder the addrs bitmask field in all
872  * the route messages to line up so we can set it here instead of back in the
873  * calling routine.
874  */
875 static void
876 rt_msg_buffer(int type, struct rt_addrinfo *rtinfo, void *buf, int msglen)
877 {
878 	struct rt_msghdr *rtm;
879 	char *cp;
880 	int dlen, i;
881 
882 	rtm = (struct rt_msghdr *) buf;
883 	rtm->rtm_version = RTM_VERSION;
884 	rtm->rtm_type = type;
885 	rtm->rtm_msglen = msglen;
886 
887 	cp = (char *)buf + rt_msghdrsize(type);
888 	rtinfo->rti_addrs = 0;
889 	for (i = 0; i < RTAX_MAX; i++) {
890 		struct sockaddr *sa;
891 
892 		if ((sa = rtinfo->rti_info[i]) == NULL)
893 			continue;
894 		rtinfo->rti_addrs |= (1 << i);
895 		dlen = ROUNDUP(sa->sa_len);
896 		bcopy(sa, cp, dlen);
897 		cp += dlen;
898 	}
899 }
900 
901 /*
902  * Build a routing message in a mbuf chain.
903  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
904  * to the end of the mbuf after the message header.
905  *
906  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
907  * This side-effect can be avoided if we reorder the addrs bitmask field in all
908  * the route messages to line up so we can set it here instead of back in the
909  * calling routine.
910  */
911 static struct mbuf *
912 rt_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
913 {
914 	struct mbuf *m;
915 	struct rt_msghdr *rtm;
916 	int hlen, len;
917 	int i;
918 
919 	hlen = rt_msghdrsize(type);
920 	KASSERT(hlen <= MCLBYTES, ("rt_msg_mbuf: hlen %d doesn't fit", hlen));
921 
922 	m = m_getl(hlen, MB_DONTWAIT, MT_DATA, M_PKTHDR, NULL);
923 	if (m == NULL)
924 		return (NULL);
925 	mbuftrackid(m, 32);
926 	m->m_pkthdr.len = m->m_len = hlen;
927 	m->m_pkthdr.rcvif = NULL;
928 	rtinfo->rti_addrs = 0;
929 	len = hlen;
930 	for (i = 0; i < RTAX_MAX; i++) {
931 		struct sockaddr *sa;
932 		int dlen;
933 
934 		if ((sa = rtinfo->rti_info[i]) == NULL)
935 			continue;
936 		rtinfo->rti_addrs |= (1 << i);
937 		dlen = ROUNDUP(sa->sa_len);
938 		m_copyback(m, len, dlen, (caddr_t)sa); /* can grow mbuf chain */
939 		len += dlen;
940 	}
941 	if (m->m_pkthdr.len != len) { /* one of the m_copyback() calls failed */
942 		m_freem(m);
943 		return (NULL);
944 	}
945 	rtm = mtod(m, struct rt_msghdr *);
946 	bzero(rtm, hlen);
947 	rtm->rtm_msglen = len;
948 	rtm->rtm_version = RTM_VERSION;
949 	rtm->rtm_type = type;
950 	return (m);
951 }
952 
953 /*
954  * This routine is called to generate a message from the routing
955  * socket indicating that a redirect has occurred, a routing lookup
956  * has failed, or that a protocol has detected timeouts to a particular
957  * destination.
958  */
959 void
960 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
961 {
962 	struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
963 	struct rt_msghdr *rtm;
964 	struct mbuf *m;
965 
966 	if (route_cb.any_count == 0)
967 		return;
968 	m = rt_msg_mbuf(type, rtinfo);
969 	if (m == NULL)
970 		return;
971 	rtm = mtod(m, struct rt_msghdr *);
972 	rtm->rtm_flags = RTF_DONE | flags;
973 	rtm->rtm_errno = error;
974 	rtm->rtm_addrs = rtinfo->rti_addrs;
975 	rts_input(m, familyof(dst));
976 }
977 
978 void
979 rt_dstmsg(int type, struct sockaddr *dst, int error)
980 {
981 	struct rt_msghdr *rtm;
982 	struct rt_addrinfo addrs;
983 	struct mbuf *m;
984 
985 	if (route_cb.any_count == 0)
986 		return;
987 	bzero(&addrs, sizeof(struct rt_addrinfo));
988 	addrs.rti_info[RTAX_DST] = dst;
989 	m = rt_msg_mbuf(type, &addrs);
990 	if (m == NULL)
991 		return;
992 	rtm = mtod(m, struct rt_msghdr *);
993 	rtm->rtm_flags = RTF_DONE;
994 	rtm->rtm_errno = error;
995 	rtm->rtm_addrs = addrs.rti_addrs;
996 	rts_input(m, familyof(dst));
997 }
998 
999 /*
1000  * This routine is called to generate a message from the routing
1001  * socket indicating that the status of a network interface has changed.
1002  */
1003 void
1004 rt_ifmsg(struct ifnet *ifp)
1005 {
1006 	struct if_msghdr *ifm;
1007 	struct mbuf *m;
1008 	struct rt_addrinfo rtinfo;
1009 
1010 	if (route_cb.any_count == 0)
1011 		return;
1012 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1013 	m = rt_msg_mbuf(RTM_IFINFO, &rtinfo);
1014 	if (m == NULL)
1015 		return;
1016 	ifm = mtod(m, struct if_msghdr *);
1017 	ifm->ifm_index = ifp->if_index;
1018 	ifm->ifm_flags = ifp->if_flags;
1019 	ifm->ifm_data = ifp->if_data;
1020 	ifm->ifm_addrs = 0;
1021 	rts_input(m, 0);
1022 }
1023 
1024 static void
1025 rt_ifamsg(int cmd, struct ifaddr *ifa)
1026 {
1027 	struct ifa_msghdr *ifam;
1028 	struct rt_addrinfo rtinfo;
1029 	struct mbuf *m;
1030 	struct ifnet *ifp = ifa->ifa_ifp;
1031 
1032 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1033 	rtinfo.rti_ifaaddr = ifa->ifa_addr;
1034 	rtinfo.rti_ifpaddr =
1035 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1036 	rtinfo.rti_netmask = ifa->ifa_netmask;
1037 	rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
1038 
1039 	m = rt_msg_mbuf(cmd, &rtinfo);
1040 	if (m == NULL)
1041 		return;
1042 
1043 	ifam = mtod(m, struct ifa_msghdr *);
1044 	ifam->ifam_index = ifp->if_index;
1045 	ifam->ifam_metric = ifa->ifa_metric;
1046 	ifam->ifam_flags = ifa->ifa_flags;
1047 	ifam->ifam_addrs = rtinfo.rti_addrs;
1048 
1049 	rts_input(m, familyof(ifa->ifa_addr));
1050 }
1051 
1052 void
1053 rt_rtmsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int error)
1054 {
1055 	struct rt_msghdr *rtm;
1056 	struct rt_addrinfo rtinfo;
1057 	struct mbuf *m;
1058 	struct sockaddr *dst;
1059 
1060 	if (rt == NULL)
1061 		return;
1062 
1063 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1064 	rtinfo.rti_dst = dst = rt_key(rt);
1065 	rtinfo.rti_gateway = rt->rt_gateway;
1066 	rtinfo.rti_netmask = rt_mask(rt);
1067 	if (ifp != NULL) {
1068 		rtinfo.rti_ifpaddr =
1069 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1070 	}
1071 	rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
1072 
1073 	m = rt_msg_mbuf(cmd, &rtinfo);
1074 	if (m == NULL)
1075 		return;
1076 
1077 	rtm = mtod(m, struct rt_msghdr *);
1078 	if (ifp != NULL)
1079 		rtm->rtm_index = ifp->if_index;
1080 	rtm->rtm_flags |= rt->rt_flags;
1081 	rtm->rtm_errno = error;
1082 	rtm->rtm_addrs = rtinfo.rti_addrs;
1083 
1084 	rts_input(m, familyof(dst));
1085 }
1086 
1087 /*
1088  * This is called to generate messages from the routing socket
1089  * indicating a network interface has had addresses associated with it.
1090  * if we ever reverse the logic and replace messages TO the routing
1091  * socket indicate a request to configure interfaces, then it will
1092  * be unnecessary as the routing socket will automatically generate
1093  * copies of it.
1094  */
1095 void
1096 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1097 {
1098 #ifdef SCTP
1099 	/*
1100 	 * notify the SCTP stack
1101 	 * this will only get called when an address is added/deleted
1102 	 * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1103 	 */
1104 	if (cmd == RTM_ADD)
1105 		sctp_add_ip_address(ifa);
1106 	else if (cmd == RTM_DELETE)
1107 		sctp_delete_ip_address(ifa);
1108 #endif /* SCTP */
1109 
1110 	if (route_cb.any_count == 0)
1111 		return;
1112 
1113 	if (cmd == RTM_ADD) {
1114 		rt_ifamsg(RTM_NEWADDR, ifa);
1115 		rt_rtmsg(RTM_ADD, rt, ifa->ifa_ifp, error);
1116 	} else {
1117 		KASSERT((cmd == RTM_DELETE), ("unknown cmd %d", cmd));
1118 		rt_rtmsg(RTM_DELETE, rt, ifa->ifa_ifp, error);
1119 		rt_ifamsg(RTM_DELADDR, ifa);
1120 	}
1121 }
1122 
1123 /*
1124  * This is the analogue to the rt_newaddrmsg which performs the same
1125  * function but for multicast group memberhips.  This is easier since
1126  * there is no route state to worry about.
1127  */
1128 void
1129 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1130 {
1131 	struct rt_addrinfo rtinfo;
1132 	struct mbuf *m = NULL;
1133 	struct ifnet *ifp = ifma->ifma_ifp;
1134 	struct ifma_msghdr *ifmam;
1135 
1136 	if (route_cb.any_count == 0)
1137 		return;
1138 
1139 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1140 	rtinfo.rti_ifaaddr = ifma->ifma_addr;
1141 	if (ifp != NULL && !TAILQ_EMPTY(&ifp->if_addrheads[mycpuid])) {
1142 		rtinfo.rti_ifpaddr =
1143 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1144 	}
1145 	/*
1146 	 * If a link-layer address is present, present it as a ``gateway''
1147 	 * (similarly to how ARP entries, e.g., are presented).
1148 	 */
1149 	rtinfo.rti_gateway = ifma->ifma_lladdr;
1150 
1151 	m = rt_msg_mbuf(cmd, &rtinfo);
1152 	if (m == NULL)
1153 		return;
1154 
1155 	ifmam = mtod(m, struct ifma_msghdr *);
1156 	ifmam->ifmam_index = ifp->if_index;
1157 	ifmam->ifmam_addrs = rtinfo.rti_addrs;
1158 
1159 	rts_input(m, familyof(ifma->ifma_addr));
1160 }
1161 
1162 static struct mbuf *
1163 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1164 		     struct rt_addrinfo *info)
1165 {
1166 	struct if_announcemsghdr *ifan;
1167 	struct mbuf *m;
1168 
1169 	if (route_cb.any_count == 0)
1170 		return NULL;
1171 
1172 	bzero(info, sizeof(*info));
1173 	m = rt_msg_mbuf(type, info);
1174 	if (m == NULL)
1175 		return NULL;
1176 
1177 	ifan = mtod(m, struct if_announcemsghdr *);
1178 	ifan->ifan_index = ifp->if_index;
1179 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof ifan->ifan_name);
1180 	ifan->ifan_what = what;
1181 	return m;
1182 }
1183 
1184 /*
1185  * This is called to generate routing socket messages indicating
1186  * IEEE80211 wireless events.
1187  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1188  */
1189 void
1190 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1191 {
1192 	struct rt_addrinfo info;
1193 	struct mbuf *m;
1194 
1195 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1196 	if (m == NULL)
1197 		return;
1198 
1199 	/*
1200 	 * Append the ieee80211 data.  Try to stick it in the
1201 	 * mbuf containing the ifannounce msg; otherwise allocate
1202 	 * a new mbuf and append.
1203 	 *
1204 	 * NB: we assume m is a single mbuf.
1205 	 */
1206 	if (data_len > M_TRAILINGSPACE(m)) {
1207 		/* XXX use m_getb(data_len, MB_DONTWAIT, MT_DATA, 0); */
1208 		struct mbuf *n = m_get(MB_DONTWAIT, MT_DATA);
1209 		if (n == NULL) {
1210 			m_freem(m);
1211 			return;
1212 		}
1213 		KKASSERT(data_len <= M_TRAILINGSPACE(n));
1214 		bcopy(data, mtod(n, void *), data_len);
1215 		n->m_len = data_len;
1216 		m->m_next = n;
1217 	} else if (data_len > 0) {
1218 		bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1219 		m->m_len += data_len;
1220 	}
1221 	mbuftrackid(m, 33);
1222 	if (m->m_flags & M_PKTHDR)
1223 		m->m_pkthdr.len += data_len;
1224 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1225 	rts_input(m, 0);
1226 }
1227 
1228 /*
1229  * This is called to generate routing socket messages indicating
1230  * network interface arrival and departure.
1231  */
1232 void
1233 rt_ifannouncemsg(struct ifnet *ifp, int what)
1234 {
1235 	struct rt_addrinfo addrinfo;
1236 	struct mbuf *m;
1237 
1238 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &addrinfo);
1239 	if (m != NULL)
1240 		rts_input(m, 0);
1241 }
1242 
1243 static int
1244 resizewalkarg(struct walkarg *w, int len)
1245 {
1246 	void *newptr;
1247 
1248 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
1249 	if (newptr == NULL)
1250 		return (ENOMEM);
1251 	if (w->w_tmem != NULL)
1252 		kfree(w->w_tmem, M_RTABLE);
1253 	w->w_tmem = newptr;
1254 	w->w_tmemsize = len;
1255 	return (0);
1256 }
1257 
1258 /*
1259  * This is used in dumping the kernel table via sysctl().
1260  */
1261 int
1262 sysctl_dumpentry(struct radix_node *rn, void *vw)
1263 {
1264 	struct walkarg *w = vw;
1265 	struct rtentry *rt = (struct rtentry *)rn;
1266 	struct rt_addrinfo rtinfo;
1267 	int error, msglen;
1268 
1269 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1270 		return 0;
1271 
1272 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1273 	rtinfo.rti_dst = rt_key(rt);
1274 	rtinfo.rti_gateway = rt->rt_gateway;
1275 	rtinfo.rti_netmask = rt_mask(rt);
1276 	rtinfo.rti_genmask = rt->rt_genmask;
1277 	if (rt->rt_ifp != NULL) {
1278 		rtinfo.rti_ifpaddr =
1279 		TAILQ_FIRST(&rt->rt_ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1280 		rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
1281 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1282 			rtinfo.rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
1283 	}
1284 	msglen = rt_msgsize(RTM_GET, &rtinfo);
1285 	if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1286 		return (ENOMEM);
1287 	rt_msg_buffer(RTM_GET, &rtinfo, w->w_tmem, msglen);
1288 	if (w->w_req != NULL) {
1289 		struct rt_msghdr *rtm = w->w_tmem;
1290 
1291 		rtm->rtm_flags = rt->rt_flags;
1292 		rtm->rtm_use = rt->rt_use;
1293 		rtm->rtm_rmx = rt->rt_rmx;
1294 		rtm->rtm_index = rt->rt_ifp->if_index;
1295 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1296 		rtm->rtm_addrs = rtinfo.rti_addrs;
1297 		error = SYSCTL_OUT(w->w_req, rtm, msglen);
1298 		return (error);
1299 	}
1300 	return (0);
1301 }
1302 
1303 static void
1304 ifnet_compute_stats(struct ifnet *ifp)
1305 {
1306 	IFNET_STAT_GET(ifp, ipackets, ifp->if_ipackets);
1307 	IFNET_STAT_GET(ifp, ierrors, ifp->if_ierrors);
1308 	IFNET_STAT_GET(ifp, opackets, ifp->if_opackets);
1309 	IFNET_STAT_GET(ifp, collisions, ifp->if_collisions);
1310 	IFNET_STAT_GET(ifp, ibytes, ifp->if_ibytes);
1311 	IFNET_STAT_GET(ifp, obytes, ifp->if_obytes);
1312 	IFNET_STAT_GET(ifp, imcasts, ifp->if_imcasts);
1313 	IFNET_STAT_GET(ifp, omcasts, ifp->if_omcasts);
1314 	IFNET_STAT_GET(ifp, iqdrops, ifp->if_iqdrops);
1315 	IFNET_STAT_GET(ifp, noproto, ifp->if_noproto);
1316 }
1317 
1318 static int
1319 sysctl_iflist(int af, struct walkarg *w)
1320 {
1321 	struct ifnet *ifp;
1322 	struct rt_addrinfo rtinfo;
1323 	int msglen, error;
1324 
1325 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1326 	TAILQ_FOREACH(ifp, &ifnet, if_link) {
1327 		struct ifaddr_container *ifac;
1328 		struct ifaddr *ifa;
1329 
1330 		if (w->w_arg && w->w_arg != ifp->if_index)
1331 			continue;
1332 		ifac = TAILQ_FIRST(&ifp->if_addrheads[mycpuid]);
1333 		ifa = ifac->ifa;
1334 		rtinfo.rti_ifpaddr = ifa->ifa_addr;
1335 		msglen = rt_msgsize(RTM_IFINFO, &rtinfo);
1336 		if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0)
1337 			return (ENOMEM);
1338 		rt_msg_buffer(RTM_IFINFO, &rtinfo, w->w_tmem, msglen);
1339 		rtinfo.rti_ifpaddr = NULL;
1340 		if (w->w_req != NULL && w->w_tmem != NULL) {
1341 			struct if_msghdr *ifm = w->w_tmem;
1342 
1343 			ifm->ifm_index = ifp->if_index;
1344 			ifm->ifm_flags = ifp->if_flags;
1345 			ifnet_compute_stats(ifp);
1346 			ifm->ifm_data = ifp->if_data;
1347 			ifm->ifm_addrs = rtinfo.rti_addrs;
1348 			error = SYSCTL_OUT(w->w_req, ifm, msglen);
1349 			if (error)
1350 				return (error);
1351 		}
1352 		while ((ifac = TAILQ_NEXT(ifac, ifa_link)) != NULL) {
1353 			ifa = ifac->ifa;
1354 
1355 			if (af && af != ifa->ifa_addr->sa_family)
1356 				continue;
1357 			if (curproc->p_ucred->cr_prison &&
1358 			    prison_if(curproc->p_ucred, ifa->ifa_addr))
1359 				continue;
1360 			rtinfo.rti_ifaaddr = ifa->ifa_addr;
1361 			rtinfo.rti_netmask = ifa->ifa_netmask;
1362 			rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
1363 			msglen = rt_msgsize(RTM_NEWADDR, &rtinfo);
1364 			if (w->w_tmemsize < msglen &&
1365 			    resizewalkarg(w, msglen) != 0)
1366 				return (ENOMEM);
1367 			rt_msg_buffer(RTM_NEWADDR, &rtinfo, w->w_tmem, msglen);
1368 			if (w->w_req != NULL) {
1369 				struct ifa_msghdr *ifam = w->w_tmem;
1370 
1371 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1372 				ifam->ifam_flags = ifa->ifa_flags;
1373 				ifam->ifam_metric = ifa->ifa_metric;
1374 				ifam->ifam_addrs = rtinfo.rti_addrs;
1375 				error = SYSCTL_OUT(w->w_req, w->w_tmem, msglen);
1376 				if (error)
1377 					return (error);
1378 			}
1379 		}
1380 		rtinfo.rti_netmask = NULL;
1381 		rtinfo.rti_ifaaddr = NULL;
1382 		rtinfo.rti_bcastaddr = NULL;
1383 	}
1384 	return (0);
1385 }
1386 
1387 static int
1388 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1389 {
1390 	int	*name = (int *)arg1;
1391 	u_int	namelen = arg2;
1392 	struct radix_node_head *rnh;
1393 	int	i, error = EINVAL;
1394 	int	origcpu;
1395 	u_char  af;
1396 	struct	walkarg w;
1397 
1398 	name ++;
1399 	namelen--;
1400 	if (req->newptr)
1401 		return (EPERM);
1402 	if (namelen != 3 && namelen != 4)
1403 		return (EINVAL);
1404 	af = name[0];
1405 	bzero(&w, sizeof w);
1406 	w.w_op = name[1];
1407 	w.w_arg = name[2];
1408 	w.w_req = req;
1409 
1410 	/*
1411 	 * Optional third argument specifies cpu, used primarily for
1412 	 * debugging the route table.
1413 	 */
1414 	if (namelen == 4) {
1415 		if (name[3] < 0 || name[3] >= ncpus)
1416 			return (EINVAL);
1417 		origcpu = mycpuid;
1418 		lwkt_migratecpu(name[3]);
1419 	} else {
1420 		origcpu = -1;
1421 	}
1422 	crit_enter();
1423 	switch (w.w_op) {
1424 	case NET_RT_DUMP:
1425 	case NET_RT_FLAGS:
1426 		for (i = 1; i <= AF_MAX; i++)
1427 			if ((rnh = rt_tables[mycpuid][i]) &&
1428 			    (af == 0 || af == i) &&
1429 			    (error = rnh->rnh_walktree(rnh,
1430 						       sysctl_dumpentry, &w)))
1431 				break;
1432 		break;
1433 
1434 	case NET_RT_IFLIST:
1435 		error = sysctl_iflist(af, &w);
1436 	}
1437 	crit_exit();
1438 	if (w.w_tmem != NULL)
1439 		kfree(w.w_tmem, M_RTABLE);
1440 	if (origcpu >= 0)
1441 		lwkt_migratecpu(origcpu);
1442 	return (error);
1443 }
1444 
1445 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1446 
1447 /*
1448  * Definitions of protocols supported in the ROUTE domain.
1449  */
1450 
1451 static struct domain routedomain;		/* or at least forward */
1452 
1453 static struct protosw routesw[] = {
1454     {
1455 	.pr_type = SOCK_RAW,
1456 	.pr_domain = &routedomain,
1457 	.pr_protocol = 0,
1458 	.pr_flags = PR_ATOMIC|PR_ADDR,
1459 	.pr_input = NULL,
1460 	.pr_output = route_output,
1461 	.pr_ctlinput = raw_ctlinput,
1462 	.pr_ctloutput = NULL,
1463 	.pr_ctlport = cpu0_ctlport,
1464 
1465 	.pr_init = raw_init,
1466 	.pr_usrreqs = &route_usrreqs
1467     }
1468 };
1469 
1470 static struct domain routedomain = {
1471 	PF_ROUTE, "route", NULL, NULL, NULL,
1472 	routesw, &routesw[(sizeof routesw)/(sizeof routesw[0])],
1473 };
1474 
1475 DOMAIN_SET(route);
1476 
1477