xref: /dragonfly/sys/net/rtsock.c (revision 55f88487)
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_inet6.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 #include <sys/jail.h>
80 
81 #include <sys/thread2.h>
82 #include <sys/socketvar2.h>
83 
84 #include <net/if.h>
85 #include <net/if_var.h>
86 #include <net/route.h>
87 #include <net/raw_cb.h>
88 #include <net/netmsg2.h>
89 #include <net/netisr2.h>
90 
91 #ifdef INET6
92 #include <netinet/in_var.h>
93 #endif
94 
95 /* sa_family is after sa_len, rest is data */
96 #define	_SA_MINSIZE	(offsetof(struct sockaddr, sa_family) + \
97 			 sizeof(((struct sockaddr *)0)->sa_family))
98 
99 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
100 
101 static struct route_cb {
102 	int	ip_count;
103 	int	ip6_count;
104 	int	any_count;
105 } route_cb;
106 
107 static const struct sockaddr route_src = { 2, PF_ROUTE, };
108 
109 struct walkarg {
110 	int	w_tmemsize;
111 	int	w_op, w_arg;
112 	void	*w_tmem;
113 	struct sysctl_req *w_req;
114 };
115 
116 #ifndef RTTABLE_DUMP_MSGCNT_MAX
117 /* Should be large enough for dupkeys */
118 #define RTTABLE_DUMP_MSGCNT_MAX		64
119 #endif
120 
121 struct rttable_walkarg {
122 	int	w_op;
123 	int	w_arg;
124 	int	w_bufsz;
125 	void	*w_buf;
126 
127 	int	w_buflen;
128 
129 	const char *w_key;
130 	const char *w_mask;
131 
132 	struct sockaddr_storage w_key0;
133 	struct sockaddr_storage w_mask0;
134 };
135 
136 struct netmsg_rttable_walk {
137 	struct netmsg_base	base;
138 	int			af;
139 	struct rttable_walkarg	*w;
140 };
141 
142 struct routecb {
143 	struct rawcb	rocb_rcb;
144 	unsigned int	rocb_msgfilter;
145 	char		*rocb_missfilter;
146 	size_t		rocb_missfilterlen;
147 };
148 #define	sotoroutecb(so)	((struct routecb *)(so)->so_pcb)
149 
150 static struct mbuf *
151 		rt_msg_mbuf (int, struct rt_addrinfo *);
152 static void	rt_msg_buffer (int, struct rt_addrinfo *, void *buf, int len);
153 static int	rt_msgsize(int type, const struct rt_addrinfo *rtinfo);
154 static int	rt_xaddrs (char *, char *, struct rt_addrinfo *);
155 static int	sysctl_rttable(int af, struct sysctl_req *req, int op, int arg);
156 static int	if_addrflags(const struct ifaddr *ifa);
157 static int	sysctl_iflist (int af, struct walkarg *w);
158 static int	route_output(struct mbuf *, struct socket *, ...);
159 static void	rt_setmetrics (u_long, struct rt_metrics *,
160 			       struct rt_metrics *);
161 
162 /*
163  * It really doesn't make any sense at all for this code to share much
164  * with raw_usrreq.c, since its functionality is so restricted.  XXX
165  */
166 static void
167 rts_abort(netmsg_t msg)
168 {
169 	crit_enter();
170 	raw_usrreqs.pru_abort(msg);
171 	/* msg invalid now */
172 	crit_exit();
173 }
174 
175 static int
176 rts_filter(struct mbuf *m, const struct sockproto *proto,
177 	const struct rawcb *rp)
178 {
179 	const struct routecb *rop = (const struct routecb *)rp;
180 	const struct rt_msghdr *rtm;
181 
182 	KKASSERT(m != NULL);
183 	KKASSERT(proto != NULL);
184 	KKASSERT(rp != NULL);
185 
186 	/* Wrong family for this socket. */
187 	if (proto->sp_family != PF_ROUTE)
188 		return ENOPROTOOPT;
189 
190 	/* If no filter set, just return. */
191 	if (rop->rocb_msgfilter == 0 && rop->rocb_missfilterlen == 0)
192 		return 0;
193 
194 	/* Ensure we can access rtm_type */
195 	if (m->m_len <
196 	    offsetof(struct rt_msghdr, rtm_type) + sizeof(rtm->rtm_type))
197 		return EINVAL;
198 
199 	rtm = mtod(m, const struct rt_msghdr *);
200 	/* If the rtm type is filtered out, return a positive. */
201 	if (rop->rocb_msgfilter != 0 &&
202 	    !(rop->rocb_msgfilter & ROUTE_FILTER(rtm->rtm_type)))
203 		return EEXIST;
204 
205 	if (rop->rocb_missfilterlen != 0 && rtm->rtm_type == RTM_MISS) {
206 		CTASSERT(RTAX_DST == 0);
207 		struct sockaddr *sa;
208 		struct sockaddr_storage ss;
209 		struct sockaddr *dst = (struct sockaddr *)&ss;
210 		char *cp = rop->rocb_missfilter;
211 		char *ep = cp + rop->rocb_missfilterlen;
212 
213 		/* Ensure we can access sa_len */
214 		if (m->m_pkthdr.len < sizeof(*rtm) + _SA_MINSIZE)
215 			return EINVAL;
216 		m_copydata(m, sizeof(*rtm) + offsetof(struct sockaddr, sa_len),
217 		    sizeof(ss.ss_len), (caddr_t)&ss);
218 		if (ss.ss_len < _SA_MINSIZE ||
219 		    ss.ss_len > sizeof(ss) ||
220 		    m->m_pkthdr.len < sizeof(*rtm) + ss.ss_len)
221 			return EINVAL;
222 		/* Copy out the destination sockaddr */
223 		m_copydata(m, sizeof(*rtm), ss.ss_len, (caddr_t)&ss);
224 
225 		/* Find a matching sockaddr in the filter */
226 		while (cp < ep) {
227 			sa = (struct sockaddr *)cp;
228 			if (sa->sa_len == dst->sa_len &&
229 			    memcmp(sa, dst, sa->sa_len) == 0)
230 				break;
231 			cp += RT_ROUNDUP(sa->sa_len);
232 		}
233 		if (cp == ep)
234 			return EEXIST;
235 	}
236 
237 	/* Passed the filter. */
238 	return 0;
239 }
240 
241 
242 /* pru_accept is EOPNOTSUPP */
243 
244 static void
245 rts_attach(netmsg_t msg)
246 {
247 	struct socket *so = msg->base.nm_so;
248 	struct pru_attach_info *ai = msg->attach.nm_ai;
249 	struct rawcb *rp;
250 	struct routecb *rop;
251 	int proto = msg->attach.nm_proto;
252 	int error;
253 
254 	crit_enter();
255 	if (sotorawcb(so) != NULL) {
256 		error = EISCONN;
257 		goto done;
258 	}
259 
260 	rop = kmalloc(sizeof *rop, M_PCB, M_WAITOK | M_ZERO);
261 	rp = &rop->rocb_rcb;
262 
263 	/*
264 	 * The critical section is necessary to block protocols from sending
265 	 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
266 	 * this PCB is extant but incompletely initialized.
267 	 * Probably we should try to do more of this work beforehand and
268 	 * eliminate the critical section.
269 	 */
270 	so->so_pcb = rp;
271 	soreference(so);	/* so_pcb assignment */
272 	error = raw_attach(so, proto, ai->sb_rlimit);
273 	rp = sotorawcb(so);
274 	if (error) {
275 		kfree(rop, M_PCB);
276 		goto done;
277 	}
278 	switch(rp->rcb_proto.sp_protocol) {
279 	case AF_INET:
280 		route_cb.ip_count++;
281 		break;
282 	case AF_INET6:
283 		route_cb.ip6_count++;
284 		break;
285 	}
286 	rp->rcb_faddr = &route_src;
287 	rp->rcb_filter = rts_filter;
288 	route_cb.any_count++;
289 	soisconnected(so);
290 	so->so_options |= SO_USELOOPBACK;
291 	error = 0;
292 done:
293 	crit_exit();
294 	lwkt_replymsg(&msg->lmsg, error);
295 }
296 
297 static void
298 rts_bind(netmsg_t msg)
299 {
300 	crit_enter();
301 	raw_usrreqs.pru_bind(msg); /* xxx just EINVAL */
302 	/* msg invalid now */
303 	crit_exit();
304 }
305 
306 static void
307 rts_connect(netmsg_t msg)
308 {
309 	crit_enter();
310 	raw_usrreqs.pru_connect(msg); /* XXX just EINVAL */
311 	/* msg invalid now */
312 	crit_exit();
313 }
314 
315 /* pru_connect2 is EOPNOTSUPP */
316 /* pru_control is EOPNOTSUPP */
317 
318 static void
319 rts_detach(netmsg_t msg)
320 {
321 	struct socket *so = msg->base.nm_so;
322 	struct rawcb *rp = sotorawcb(so);
323 	struct routecb *rop = (struct routecb *)rp;
324 
325 	crit_enter();
326 	if (rop->rocb_missfilterlen != 0)
327 		kfree(rop->rocb_missfilter, M_PCB);
328 	if (rp != NULL) {
329 		switch(rp->rcb_proto.sp_protocol) {
330 		case AF_INET:
331 			route_cb.ip_count--;
332 			break;
333 		case AF_INET6:
334 			route_cb.ip6_count--;
335 			break;
336 		}
337 		route_cb.any_count--;
338 	}
339 	raw_usrreqs.pru_detach(msg);
340 	/* msg invalid now */
341 	crit_exit();
342 }
343 
344 static void
345 rts_disconnect(netmsg_t msg)
346 {
347 	crit_enter();
348 	raw_usrreqs.pru_disconnect(msg);
349 	/* msg invalid now */
350 	crit_exit();
351 }
352 
353 /* pru_listen is EOPNOTSUPP */
354 
355 static void
356 rts_peeraddr(netmsg_t msg)
357 {
358 	crit_enter();
359 	raw_usrreqs.pru_peeraddr(msg);
360 	/* msg invalid now */
361 	crit_exit();
362 }
363 
364 /* pru_rcvd is EOPNOTSUPP */
365 /* pru_rcvoob is EOPNOTSUPP */
366 
367 static void
368 rts_send(netmsg_t msg)
369 {
370 	crit_enter();
371 	raw_usrreqs.pru_send(msg);
372 	/* msg invalid now */
373 	crit_exit();
374 }
375 
376 /* pru_sense is null */
377 
378 static void
379 rts_shutdown(netmsg_t msg)
380 {
381 	crit_enter();
382 	raw_usrreqs.pru_shutdown(msg);
383 	/* msg invalid now */
384 	crit_exit();
385 }
386 
387 static void
388 rts_sockaddr(netmsg_t msg)
389 {
390 	crit_enter();
391 	raw_usrreqs.pru_sockaddr(msg);
392 	/* msg invalid now */
393 	crit_exit();
394 }
395 
396 static struct pr_usrreqs route_usrreqs = {
397 	.pru_abort = rts_abort,
398 	.pru_accept = pr_generic_notsupp,
399 	.pru_attach = rts_attach,
400 	.pru_bind = rts_bind,
401 	.pru_connect = rts_connect,
402 	.pru_connect2 = pr_generic_notsupp,
403 	.pru_control = pr_generic_notsupp,
404 	.pru_detach = rts_detach,
405 	.pru_disconnect = rts_disconnect,
406 	.pru_listen = pr_generic_notsupp,
407 	.pru_peeraddr = rts_peeraddr,
408 	.pru_rcvd = pr_generic_notsupp,
409 	.pru_rcvoob = pr_generic_notsupp,
410 	.pru_send = rts_send,
411 	.pru_sense = pru_sense_null,
412 	.pru_shutdown = rts_shutdown,
413 	.pru_sockaddr = rts_sockaddr,
414 	.pru_sosend = sosend,
415 	.pru_soreceive = soreceive
416 };
417 
418 static __inline sa_family_t
419 familyof(struct sockaddr *sa)
420 {
421 	return (sa != NULL ? sa->sa_family : 0);
422 }
423 
424 /*
425  * Routing socket input function.  The packet must be serialized onto cpu 0.
426  * We use the cpu0_soport() netisr processing loop to handle it.
427  *
428  * This looks messy but it means that anyone, including interrupt code,
429  * can send a message to the routing socket.
430  */
431 static void
432 rts_input_handler(netmsg_t msg)
433 {
434 	static const struct sockaddr route_dst = { 2, PF_ROUTE, };
435 	struct sockproto route_proto;
436 	struct netmsg_packet *pmsg = &msg->packet;
437 	struct mbuf *m;
438 	sa_family_t family;
439 	struct rawcb *skip;
440 
441 	family = pmsg->base.lmsg.u.ms_result;
442 	route_proto.sp_family = PF_ROUTE;
443 	route_proto.sp_protocol = family;
444 
445 	m = pmsg->nm_packet;
446 	M_ASSERTPKTHDR(m);
447 
448 	skip = m->m_pkthdr.header;
449 	m->m_pkthdr.header = NULL;
450 
451 	raw_input(m, &route_proto, &route_src, &route_dst, skip);
452 }
453 
454 static void
455 rts_input_skip(struct mbuf *m, sa_family_t family, struct rawcb *skip)
456 {
457 	struct netmsg_packet *pmsg;
458 	lwkt_port_t port;
459 
460 	M_ASSERTPKTHDR(m);
461 
462 	port = netisr_cpuport(0);	/* XXX same as for routing socket */
463 	pmsg = &m->m_hdr.mh_netmsg;
464 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
465 		    0, rts_input_handler);
466 	pmsg->nm_packet = m;
467 	pmsg->base.lmsg.u.ms_result = family;
468 	m->m_pkthdr.header = skip; /* XXX steal field in pkthdr */
469 	lwkt_sendmsg(port, &pmsg->base.lmsg);
470 }
471 
472 static __inline void
473 rts_input(struct mbuf *m, sa_family_t family)
474 {
475 	rts_input_skip(m, family, NULL);
476 }
477 
478 static void
479 route_ctloutput(netmsg_t msg)
480 {
481 	struct socket *so = msg->ctloutput.base.nm_so;
482 	struct sockopt *sopt = msg->ctloutput.nm_sopt;
483 	struct routecb *rop = sotoroutecb(so);
484 	int error;
485 	unsigned int msgfilter;
486 	unsigned char *cp, *ep;
487 	size_t len;
488 	struct sockaddr *sa;
489 
490 	if (sopt->sopt_level != AF_ROUTE) {
491 		error = EINVAL;
492 		goto out;
493 	}
494 
495 	error = 0;
496 
497 	switch (sopt->sopt_dir) {
498 	case SOPT_SET:
499 		switch (sopt->sopt_name) {
500 		case ROUTE_MSGFILTER:
501 			error = soopt_to_kbuf(sopt, &msgfilter,
502 			    sizeof(msgfilter), sizeof(msgfilter));
503 			if (error == 0)
504 				rop->rocb_msgfilter = msgfilter;
505 			break;
506 		case RO_MISSFILTER:
507 			/* Validate the data */
508 			len = 0;
509 			cp = sopt->sopt_val;
510 			ep = cp + sopt->sopt_valsize;
511 			while (cp < ep) {
512 				if (ep - cp <
513 				    offsetof(struct sockaddr, sa_len) +
514 				    sizeof(sa->sa_len))
515 					break;
516 				if (++len > RO_FILTSA_MAX) {
517 					error = ENOBUFS;
518 					break;
519 				}
520 				sa = (struct sockaddr *)cp;
521 				if (sa->sa_len < _SA_MINSIZE ||
522 				    sa->sa_len > sizeof(struct sockaddr_storage))
523 					break;
524 				cp += RT_ROUNDUP(sa->sa_len);
525 			}
526 			if (cp != ep) {
527 				if (error == 0)
528 					error = EINVAL;
529 				break;
530 			}
531 			if (rop->rocb_missfilterlen != 0)
532 				kfree(rop->rocb_missfilter, M_PCB);
533 			if (sopt->sopt_valsize != 0) {
534 				rop->rocb_missfilter =
535 				    kmalloc(sopt->sopt_valsize,
536 				            M_PCB, M_WAITOK | M_NULLOK);
537 				if (rop->rocb_missfilter == NULL) {
538 					rop->rocb_missfilterlen = 0;
539 					error = ENOBUFS;
540 					break;
541 				}
542 			} else
543 				rop->rocb_missfilter = NULL;
544 			rop->rocb_missfilterlen = sopt->sopt_valsize;
545 			if (rop->rocb_missfilterlen != 0)
546 				memcpy(rop->rocb_missfilter, sopt->sopt_val,
547 				    rop->rocb_missfilterlen);
548 			break;
549 		default:
550 			error = ENOPROTOOPT;
551 			break;
552 		}
553 		break;
554 	case SOPT_GET:
555 		switch (sopt->sopt_name) {
556 		case ROUTE_MSGFILTER:
557 			msgfilter = rop->rocb_msgfilter;
558 			soopt_from_kbuf(sopt, &msgfilter, sizeof(msgfilter));
559 			break;
560 		case RO_MISSFILTER:
561 			soopt_from_kbuf(sopt, rop->rocb_missfilter,
562 			    rop->rocb_missfilterlen);
563 			break;
564 		default:
565 			error = ENOPROTOOPT;
566 			break;
567 		}
568 	}
569 out:
570 	lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
571 }
572 
573 
574 
575 static void *
576 reallocbuf_nofree(void *ptr, size_t len, size_t olen)
577 {
578 	void *newptr;
579 
580 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
581 	if (newptr == NULL)
582 		return NULL;
583 	bcopy(ptr, newptr, olen);
584 	if (olen < len)
585 		bzero((char *)newptr + olen, len - olen);
586 
587 	return (newptr);
588 }
589 
590 /*
591  * Internal helper routine for route_output().
592  */
593 static int
594 _fillrtmsg(struct rt_msghdr **prtm, struct rtentry *rt,
595 	   struct rt_addrinfo *rtinfo)
596 {
597 	int msglen;
598 	struct rt_msghdr *rtm = *prtm;
599 
600 	/* Fill in rt_addrinfo for call to rt_msg_buffer(). */
601 	rtinfo->rti_dst = rt_key(rt);
602 	rtinfo->rti_gateway = rt->rt_gateway;
603 	rtinfo->rti_netmask = rt_mask(rt);		/* might be NULL */
604 	rtinfo->rti_genmask = rt->rt_genmask;		/* might be NULL */
605 	if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
606 		if (rt->rt_ifp != NULL) {
607 			rtinfo->rti_ifpaddr =
608 			    TAILQ_FIRST(&rt->rt_ifp->if_addrheads[mycpuid])
609 			    ->ifa->ifa_addr;
610 			rtinfo->rti_ifaaddr = rt->rt_ifa->ifa_addr;
611 			if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
612 				rtinfo->rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
613 			rtm->rtm_index = rt->rt_ifp->if_index;
614 		} else {
615 			rtinfo->rti_ifpaddr = NULL;
616 			rtinfo->rti_ifaaddr = NULL;
617 		}
618 	} else if (rt->rt_ifp != NULL) {
619 		rtm->rtm_index = rt->rt_ifp->if_index;
620 	}
621 
622 	msglen = rt_msgsize(rtm->rtm_type, rtinfo);
623 	if (rtm->rtm_msglen < msglen) {
624 		/* NOTE: Caller will free the old rtm accordingly */
625 		rtm = reallocbuf_nofree(rtm, msglen, rtm->rtm_msglen);
626 		if (rtm == NULL)
627 			return (ENOBUFS);
628 		*prtm = rtm;
629 	}
630 	rt_msg_buffer(rtm->rtm_type, rtinfo, rtm, msglen);
631 
632 	rtm->rtm_flags = rt->rt_flags;
633 	rtm->rtm_rmx = rt->rt_rmx;
634 	rtm->rtm_addrs = rtinfo->rti_addrs;
635 
636 	return (0);
637 }
638 
639 struct rtm_arg {
640 	struct rt_msghdr	*bak_rtm;
641 	struct rt_msghdr	*new_rtm;
642 };
643 
644 static int
645 fillrtmsg(struct rtm_arg *arg, struct rtentry *rt,
646 	  struct rt_addrinfo *rtinfo)
647 {
648 	struct rt_msghdr *rtm = arg->new_rtm;
649 	int error;
650 
651 	error = _fillrtmsg(&rtm, rt, rtinfo);
652 	if (!error) {
653 		if (arg->new_rtm != rtm) {
654 			/*
655 			 * _fillrtmsg() just allocated a new rtm;
656 			 * if the previously allocated rtm is not
657 			 * the backing rtm, it should be freed.
658 			 */
659 			if (arg->new_rtm != arg->bak_rtm)
660 				kfree(arg->new_rtm, M_RTABLE);
661 			arg->new_rtm = rtm;
662 		}
663 	}
664 	return error;
665 }
666 
667 static void route_output_add_callback(int, int, struct rt_addrinfo *,
668 					struct rtentry *, void *);
669 static void route_output_delete_callback(int, int, struct rt_addrinfo *,
670 					struct rtentry *, void *);
671 static int route_output_get_callback(int, struct rt_addrinfo *,
672 				     struct rtentry *, void *, int);
673 static int route_output_change_callback(int, struct rt_addrinfo *,
674 					struct rtentry *, void *, int);
675 static int route_output_lock_callback(int, struct rt_addrinfo *,
676 				      struct rtentry *, void *, int);
677 
678 /*ARGSUSED*/
679 static int
680 route_output(struct mbuf *m, struct socket *so, ...)
681 {
682 	struct rtm_arg arg;
683 	struct rt_msghdr *rtm = NULL;
684 	struct rawcb *rp = NULL;
685 	struct pr_output_info *oi;
686 	struct rt_addrinfo rtinfo;
687 	sa_family_t family;
688 	int len, error = 0;
689 	__va_list ap;
690 
691 	M_ASSERTPKTHDR(m);
692 
693 	__va_start(ap, so);
694 	oi = __va_arg(ap, struct pr_output_info *);
695 	__va_end(ap);
696 
697 	family = familyof(NULL);
698 
699 #define gotoerr(e) { error = e; goto flush;}
700 
701 	if (m == NULL ||
702 	    (m->m_len < sizeof(long) &&
703 	     (m = m_pullup(m, sizeof(long))) == NULL))
704 		return (ENOBUFS);
705 	len = m->m_pkthdr.len;
706 	if (len < sizeof(struct rt_msghdr) ||
707 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen)
708 		gotoerr(EINVAL);
709 
710 	rtm = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
711 	if (rtm == NULL)
712 		gotoerr(ENOBUFS);
713 
714 	m_copydata(m, 0, len, (caddr_t)rtm);
715 	if (rtm->rtm_version != RTM_VERSION)
716 		gotoerr(EPROTONOSUPPORT);
717 
718 	rtm->rtm_pid = oi->p_pid;
719 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
720 	rtinfo.rti_addrs = rtm->rtm_addrs;
721 	if (rt_xaddrs((char *)(rtm + 1), (char *)rtm + len, &rtinfo) != 0)
722 		gotoerr(EINVAL);
723 
724 	rtinfo.rti_flags = rtm->rtm_flags;
725 	if (rtinfo.rti_dst == NULL || rtinfo.rti_dst->sa_family >= AF_MAX ||
726 	    (rtinfo.rti_gateway && rtinfo.rti_gateway->sa_family >= AF_MAX))
727 		gotoerr(EINVAL);
728 
729 	family = familyof(rtinfo.rti_dst);
730 
731 	/*
732 	 * Verify that the caller has the appropriate privilege; RTM_GET
733 	 * is the only operation the non-superuser is allowed.
734 	 */
735 	if (rtm->rtm_type != RTM_GET &&
736 	    priv_check_cred(so->so_cred, PRIV_ROOT, 0) != 0)
737 		gotoerr(EPERM);
738 
739 	if (rtinfo.rti_genmask != NULL) {
740 		error = rtmask_add_global(rtinfo.rti_genmask,
741 		    rtm->rtm_type != RTM_GET ?
742 		    RTREQ_PRIO_HIGH : RTREQ_PRIO_NORM);
743 		if (error)
744 			goto flush;
745 	}
746 
747 	switch (rtm->rtm_type) {
748 	case RTM_ADD:
749 		if (rtinfo.rti_gateway == NULL) {
750 			error = EINVAL;
751 		} else {
752 			error = rtrequest1_global(RTM_ADD, &rtinfo,
753 			    route_output_add_callback, rtm, RTREQ_PRIO_HIGH);
754 		}
755 		break;
756 	case RTM_DELETE:
757 		/*
758 		 * Backing rtm (bak_rtm) could _not_ be freed during
759 		 * rtrequest1_global or rtsearch_global, even if the
760 		 * callback reallocates the rtm due to its size changes,
761 		 * since rtinfo points to the backing rtm's memory area.
762 		 * After rtrequest1_global or rtsearch_global returns,
763 		 * it is safe to free the backing rtm, since rtinfo will
764 		 * not be used anymore.
765 		 *
766 		 * new_rtm will be used to save the new rtm allocated
767 		 * by rtrequest1_global or rtsearch_global.
768 		 */
769 		arg.bak_rtm = rtm;
770 		arg.new_rtm = rtm;
771 		error = rtrequest1_global(RTM_DELETE, &rtinfo,
772 		    route_output_delete_callback, &arg, RTREQ_PRIO_HIGH);
773 		rtm = arg.new_rtm;
774 		if (rtm != arg.bak_rtm)
775 			kfree(arg.bak_rtm, M_RTABLE);
776 		break;
777 	case RTM_GET:
778 		/* See the comment in RTM_DELETE */
779 		arg.bak_rtm = rtm;
780 		arg.new_rtm = rtm;
781 		error = rtsearch_global(RTM_GET, &rtinfo,
782 		    route_output_get_callback, &arg, RTS_NOEXACTMATCH,
783 		    RTREQ_PRIO_NORM);
784 		rtm = arg.new_rtm;
785 		if (rtm != arg.bak_rtm)
786 			kfree(arg.bak_rtm, M_RTABLE);
787 		break;
788 	case RTM_CHANGE:
789 		error = rtsearch_global(RTM_CHANGE, &rtinfo,
790 		    route_output_change_callback, rtm, RTS_EXACTMATCH,
791 		    RTREQ_PRIO_HIGH);
792 		break;
793 	case RTM_LOCK:
794 		error = rtsearch_global(RTM_LOCK, &rtinfo,
795 		    route_output_lock_callback, rtm, RTS_EXACTMATCH,
796 		    RTREQ_PRIO_HIGH);
797 		break;
798 	default:
799 		error = EOPNOTSUPP;
800 		break;
801 	}
802 flush:
803 	if (rtm != NULL) {
804 		if (error != 0)
805 			rtm->rtm_errno = error;
806 		else
807 			rtm->rtm_flags |= RTF_DONE;
808 	}
809 
810 	/*
811 	 * Check to see if we don't want our own messages.
812 	 */
813 	if (!(so->so_options & SO_USELOOPBACK)) {
814 		if (route_cb.any_count <= 1) {
815 			if (rtm != NULL)
816 				kfree(rtm, M_RTABLE);
817 			m_freem(m);
818 			return (error);
819 		}
820 		/* There is another listener, so construct message */
821 		rp = sotorawcb(so);
822 	}
823 	if (rtm != NULL) {
824 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
825 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
826 			m_freem(m);
827 			m = NULL;
828 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
829 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
830 		kfree(rtm, M_RTABLE);
831 	}
832 	if (m != NULL)
833 		rts_input_skip(m, family, rp);
834 	return (error);
835 }
836 
837 static void
838 route_output_add_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
839 			  struct rtentry *rt, void *arg)
840 {
841 	struct rt_msghdr *rtm = arg;
842 
843 	if (error == 0 && rt != NULL) {
844 		rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
845 		    &rt->rt_rmx);
846 		rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
847 		rt->rt_rmx.rmx_locks |=
848 		    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
849 		if (rtinfo->rti_genmask != NULL) {
850 			rt->rt_genmask = rtmask_purelookup(rtinfo->rti_genmask);
851 			if (rt->rt_genmask == NULL) {
852 				/*
853 				 * This should not happen, since we
854 				 * have already installed genmask
855 				 * on each CPU before we reach here.
856 				 */
857 				panic("genmask is gone!?");
858 			}
859 		} else {
860 			rt->rt_genmask = NULL;
861 		}
862 		rtm->rtm_index = rt->rt_ifp->if_index;
863 	}
864 }
865 
866 static void
867 route_output_delete_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
868 			  struct rtentry *rt, void *arg)
869 {
870 	if (error == 0 && rt) {
871 		++rt->rt_refcnt;
872 		if (fillrtmsg(arg, rt, rtinfo) != 0) {
873 			error = ENOBUFS;
874 			/* XXX no way to return the error */
875 		}
876 		--rt->rt_refcnt;
877 	}
878 	if (rt && rt->rt_refcnt == 0) {
879 		++rt->rt_refcnt;
880 		rtfree(rt);
881 	}
882 }
883 
884 static int
885 route_output_get_callback(int cmd, struct rt_addrinfo *rtinfo,
886 			  struct rtentry *rt, void *arg, int found_cnt)
887 {
888 	int error, found = 0;
889 
890 	if (((rtinfo->rti_flags ^ rt->rt_flags) & RTF_HOST) == 0)
891 		found = 1;
892 
893 	error = fillrtmsg(arg, rt, rtinfo);
894 	if (!error && found) {
895 		/* Got the exact match, we could return now! */
896 		error = EJUSTRETURN;
897 	}
898 	return error;
899 }
900 
901 static int
902 route_output_change_callback(int cmd, struct rt_addrinfo *rtinfo,
903 			     struct rtentry *rt, void *arg, int found_cnt)
904 {
905 	struct rt_msghdr *rtm = arg;
906 	struct ifaddr *ifa;
907 	int error = 0;
908 
909 	/*
910 	 * new gateway could require new ifaddr, ifp;
911 	 * flags may also be different; ifp may be specified
912 	 * by ll sockaddr when protocol address is ambiguous
913 	 */
914 	if (((rt->rt_flags & RTF_GATEWAY) && rtinfo->rti_gateway != NULL) ||
915 	    rtinfo->rti_ifpaddr != NULL ||
916 	    (rtinfo->rti_ifaaddr != NULL &&
917 	     !sa_equal(rtinfo->rti_ifaaddr, rt->rt_ifa->ifa_addr))) {
918 		error = rt_getifa(rtinfo);
919 		if (error != 0)
920 			goto done;
921 	}
922 	if (rtinfo->rti_gateway != NULL) {
923 		/*
924 		 * We only need to generate rtmsg upon the
925 		 * first route to be changed.
926 		 */
927 		error = rt_setgate(rt, rt_key(rt), rtinfo->rti_gateway);
928 		if (error != 0)
929 			goto done;
930 	}
931 	if ((ifa = rtinfo->rti_ifa) != NULL) {
932 		struct ifaddr *oifa = rt->rt_ifa;
933 
934 		if (oifa != ifa) {
935 			if (oifa && oifa->ifa_rtrequest)
936 				oifa->ifa_rtrequest(RTM_DELETE, rt);
937 			IFAFREE(rt->rt_ifa);
938 			IFAREF(ifa);
939 			rt->rt_ifa = ifa;
940 			rt->rt_ifp = rtinfo->rti_ifp;
941 		}
942 	}
943 	rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, &rt->rt_rmx);
944 	if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
945 		rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt);
946 	if (rtinfo->rti_genmask != NULL) {
947 		rt->rt_genmask = rtmask_purelookup(rtinfo->rti_genmask);
948 		if (rt->rt_genmask == NULL) {
949 			/*
950 			 * This should not happen, since we
951 			 * have already installed genmask
952 			 * on each CPU before we reach here.
953 			 */
954 			panic("genmask is gone!?");
955 		}
956 	}
957 	rtm->rtm_index = rt->rt_ifp->if_index;
958 	if (found_cnt == 1)
959 		rt_rtmsg(RTM_CHANGE, rt, rt->rt_ifp, 0);
960 done:
961 	return error;
962 }
963 
964 static int
965 route_output_lock_callback(int cmd, struct rt_addrinfo *rtinfo,
966 			   struct rtentry *rt, void *arg,
967 			   int found_cnt __unused)
968 {
969 	struct rt_msghdr *rtm = arg;
970 
971 	rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
972 	rt->rt_rmx.rmx_locks |=
973 		(rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
974 	return 0;
975 }
976 
977 static void
978 rt_setmetrics(u_long which, struct rt_metrics *in, struct rt_metrics *out)
979 {
980 #define setmetric(flag, elt) if (which & (flag)) out->elt = in->elt;
981 	setmetric(RTV_RPIPE, rmx_recvpipe);
982 	setmetric(RTV_SPIPE, rmx_sendpipe);
983 	setmetric(RTV_SSTHRESH, rmx_ssthresh);
984 	setmetric(RTV_RTT, rmx_rtt);
985 	setmetric(RTV_RTTVAR, rmx_rttvar);
986 	setmetric(RTV_HOPCOUNT, rmx_hopcount);
987 	setmetric(RTV_MTU, rmx_mtu);
988 	setmetric(RTV_EXPIRE, rmx_expire);
989 	setmetric(RTV_MSL, rmx_msl);
990 	setmetric(RTV_IWMAXSEGS, rmx_iwmaxsegs);
991 	setmetric(RTV_IWCAPSEGS, rmx_iwcapsegs);
992 #undef setmetric
993 }
994 
995 /*
996  * Extract the addresses of the passed sockaddrs.
997  * Do a little sanity checking so as to avoid bad memory references.
998  * This data is derived straight from userland.
999  */
1000 static int
1001 rt_xaddrs(char *cp, char *cplim, struct rt_addrinfo *rtinfo)
1002 {
1003 	struct sockaddr *sa;
1004 	int i;
1005 
1006 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
1007 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
1008 			continue;
1009 		sa = (struct sockaddr *)cp;
1010 		/*
1011 		 * It won't fit.
1012 		 */
1013 		if ((cp + sa->sa_len) > cplim) {
1014 			return (EINVAL);
1015 		}
1016 
1017 		/*
1018 		 * There are no more...  Quit now.
1019 		 * If there are more bits, they are in error.
1020 		 * I've seen this.  route(1) can evidently generate these.
1021 		 * This causes kernel to core dump.
1022 		 * For compatibility, if we see this, point to a safe address.
1023 		 */
1024 		if (sa->sa_len == 0) {
1025 			static struct sockaddr sa_zero = {
1026 				sizeof sa_zero, AF_INET,
1027 			};
1028 
1029 			rtinfo->rti_info[i] = &sa_zero;
1030 			kprintf("rtsock: received more addr bits than sockaddrs.\n");
1031 			return (0); /* should be EINVAL but for compat */
1032 		}
1033 
1034 		/* Accept the sockaddr. */
1035 		rtinfo->rti_info[i] = sa;
1036 		cp += RT_ROUNDUP(sa->sa_len);
1037 	}
1038 	return (0);
1039 }
1040 
1041 static int
1042 rt_msghdrsize(int type)
1043 {
1044 	switch (type) {
1045 	case RTM_DELADDR:
1046 	case RTM_NEWADDR:
1047 		return sizeof(struct ifa_msghdr);
1048 	case RTM_DELMADDR:
1049 	case RTM_NEWMADDR:
1050 		return sizeof(struct ifma_msghdr);
1051 	case RTM_IFINFO:
1052 		return sizeof(struct if_msghdr);
1053 	case RTM_IFANNOUNCE:
1054 	case RTM_IEEE80211:
1055 		return sizeof(struct if_announcemsghdr);
1056 	default:
1057 		return sizeof(struct rt_msghdr);
1058 	}
1059 }
1060 
1061 static int
1062 rt_msgsize(int type, const struct rt_addrinfo *rtinfo)
1063 {
1064 	int len, i;
1065 
1066 	len = rt_msghdrsize(type);
1067 	for (i = 0; i < RTAX_MAX; i++) {
1068 		if (rtinfo->rti_info[i] != NULL)
1069 			len += RT_ROUNDUP(rtinfo->rti_info[i]->sa_len);
1070 	}
1071 	len = ALIGN(len);
1072 	return len;
1073 }
1074 
1075 /*
1076  * Build a routing message in a buffer.
1077  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
1078  * to the end of the buffer after the message header.
1079  *
1080  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
1081  * This side-effect can be avoided if we reorder the addrs bitmask field in all
1082  * the route messages to line up so we can set it here instead of back in the
1083  * calling routine.
1084  *
1085  * NOTE! The buffer may already contain a partially filled-out rtm via
1086  *	 _fillrtmsg().
1087  */
1088 static void
1089 rt_msg_buffer(int type, struct rt_addrinfo *rtinfo, void *buf, int msglen)
1090 {
1091 	struct rt_msghdr *rtm;
1092 	char *cp;
1093 	int dlen, i;
1094 
1095 	rtm = (struct rt_msghdr *) buf;
1096 	rtm->rtm_version = RTM_VERSION;
1097 	rtm->rtm_type = type;
1098 	rtm->rtm_msglen = msglen;
1099 
1100 	cp = (char *)buf + rt_msghdrsize(type);
1101 	rtinfo->rti_addrs = 0;
1102 	for (i = 0; i < RTAX_MAX; i++) {
1103 		struct sockaddr *sa;
1104 
1105 		if ((sa = rtinfo->rti_info[i]) == NULL)
1106 			continue;
1107 		rtinfo->rti_addrs |= (1 << i);
1108 		dlen = RT_ROUNDUP(sa->sa_len);
1109 		bcopy(sa, cp, dlen);
1110 		cp += dlen;
1111 	}
1112 }
1113 
1114 /*
1115  * Build a routing message in a mbuf chain.
1116  * Copy the addresses in the rtinfo->rti_info[] sockaddr array
1117  * to the end of the mbuf after the message header.
1118  *
1119  * Set the rtinfo->rti_addrs bitmask of addresses present in rtinfo->rti_info[].
1120  * This side-effect can be avoided if we reorder the addrs bitmask field in all
1121  * the route messages to line up so we can set it here instead of back in the
1122  * calling routine.
1123  */
1124 static struct mbuf *
1125 rt_msg_mbuf(int type, struct rt_addrinfo *rtinfo)
1126 {
1127 	struct mbuf *m;
1128 	struct rt_msghdr *rtm;
1129 	int hlen, len;
1130 	int i;
1131 
1132 	hlen = rt_msghdrsize(type);
1133 	KASSERT(hlen <= MCLBYTES, ("rt_msg_mbuf: hlen %d doesn't fit", hlen));
1134 
1135 	m = m_getl(hlen, M_NOWAIT, MT_DATA, M_PKTHDR, NULL);
1136 	if (m == NULL)
1137 		return (NULL);
1138 	mbuftrackid(m, 32);
1139 	m->m_pkthdr.len = m->m_len = hlen;
1140 	m->m_pkthdr.rcvif = NULL;
1141 	rtinfo->rti_addrs = 0;
1142 	len = hlen;
1143 	for (i = 0; i < RTAX_MAX; i++) {
1144 		struct sockaddr *sa;
1145 		int dlen;
1146 
1147 		if ((sa = rtinfo->rti_info[i]) == NULL)
1148 			continue;
1149 		rtinfo->rti_addrs |= (1 << i);
1150 		dlen = RT_ROUNDUP(sa->sa_len);
1151 		m_copyback(m, len, dlen, (caddr_t)sa); /* can grow mbuf chain */
1152 		len += dlen;
1153 	}
1154 	if (m->m_pkthdr.len != len) { /* one of the m_copyback() calls failed */
1155 		m_freem(m);
1156 		return (NULL);
1157 	}
1158 	rtm = mtod(m, struct rt_msghdr *);
1159 	bzero(rtm, hlen);
1160 	rtm->rtm_msglen = len;
1161 	rtm->rtm_version = RTM_VERSION;
1162 	rtm->rtm_type = type;
1163 	return (m);
1164 }
1165 
1166 /*
1167  * This routine is called to generate a message from the routing
1168  * socket indicating that a redirect has occurred, a routing lookup
1169  * has failed, or that a protocol has detected timeouts to a particular
1170  * destination.
1171  */
1172 void
1173 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1174 {
1175 	struct sockaddr *dst = rtinfo->rti_info[RTAX_DST];
1176 	struct rt_msghdr *rtm;
1177 	struct mbuf *m;
1178 
1179 	if (route_cb.any_count == 0)
1180 		return;
1181 	m = rt_msg_mbuf(type, rtinfo);
1182 	if (m == NULL)
1183 		return;
1184 	rtm = mtod(m, struct rt_msghdr *);
1185 	rtm->rtm_flags = RTF_DONE | flags;
1186 	rtm->rtm_errno = error;
1187 	rtm->rtm_addrs = rtinfo->rti_addrs;
1188 	rts_input(m, familyof(dst));
1189 }
1190 
1191 void
1192 rt_dstmsg(int type, struct sockaddr *dst, int error)
1193 {
1194 	struct rt_msghdr *rtm;
1195 	struct rt_addrinfo addrs;
1196 	struct mbuf *m;
1197 
1198 	if (route_cb.any_count == 0)
1199 		return;
1200 	bzero(&addrs, sizeof(struct rt_addrinfo));
1201 	addrs.rti_info[RTAX_DST] = dst;
1202 	m = rt_msg_mbuf(type, &addrs);
1203 	if (m == NULL)
1204 		return;
1205 	rtm = mtod(m, struct rt_msghdr *);
1206 	rtm->rtm_flags = RTF_DONE;
1207 	rtm->rtm_errno = error;
1208 	rtm->rtm_addrs = addrs.rti_addrs;
1209 	rts_input(m, familyof(dst));
1210 }
1211 
1212 /*
1213  * This routine is called to generate a message from the routing
1214  * socket indicating that the status of a network interface has changed.
1215  */
1216 void
1217 rt_ifmsg(struct ifnet *ifp)
1218 {
1219 	struct if_msghdr *ifm;
1220 	struct mbuf *m;
1221 	struct rt_addrinfo rtinfo;
1222 
1223 	if (route_cb.any_count == 0)
1224 		return;
1225 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1226 	m = rt_msg_mbuf(RTM_IFINFO, &rtinfo);
1227 	if (m == NULL)
1228 		return;
1229 	ifm = mtod(m, struct if_msghdr *);
1230 	ifm->ifm_index = ifp->if_index;
1231 	ifm->ifm_flags = ifp->if_flags;
1232 	ifm->ifm_data = ifp->if_data;
1233 	ifm->ifm_addrs = 0;
1234 	rts_input(m, 0);
1235 }
1236 
1237 static void
1238 rt_ifamsg(int cmd, struct ifaddr *ifa)
1239 {
1240 	struct ifa_msghdr *ifam;
1241 	struct rt_addrinfo rtinfo;
1242 	struct mbuf *m;
1243 	struct ifnet *ifp = ifa->ifa_ifp;
1244 
1245 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1246 	rtinfo.rti_ifaaddr = ifa->ifa_addr;
1247 	rtinfo.rti_ifpaddr =
1248 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1249 	rtinfo.rti_netmask = ifa->ifa_netmask;
1250 	rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
1251 
1252 	m = rt_msg_mbuf(cmd, &rtinfo);
1253 	if (m == NULL)
1254 		return;
1255 
1256 	ifam = mtod(m, struct ifa_msghdr *);
1257 	ifam->ifam_index = ifp->if_index;
1258 	ifam->ifam_flags = ifa->ifa_flags;
1259 	ifam->ifam_addrs = rtinfo.rti_addrs;
1260 	ifam->ifam_addrflags = if_addrflags(ifa);
1261 	ifam->ifam_metric = ifa->ifa_metric;
1262 
1263 	rts_input(m, familyof(ifa->ifa_addr));
1264 }
1265 
1266 void
1267 rt_rtmsg(int cmd, struct rtentry *rt, struct ifnet *ifp, int error)
1268 {
1269 	struct rt_msghdr *rtm;
1270 	struct rt_addrinfo rtinfo;
1271 	struct mbuf *m;
1272 	struct sockaddr *dst;
1273 
1274 	if (rt == NULL)
1275 		return;
1276 
1277 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1278 	rtinfo.rti_dst = dst = rt_key(rt);
1279 	rtinfo.rti_gateway = rt->rt_gateway;
1280 	rtinfo.rti_netmask = rt_mask(rt);
1281 	if (ifp != NULL) {
1282 		rtinfo.rti_ifpaddr =
1283 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1284 	}
1285 	if (rt->rt_ifa != NULL)
1286 		rtinfo.rti_ifaaddr = rt->rt_ifa->ifa_addr;
1287 
1288 	m = rt_msg_mbuf(cmd, &rtinfo);
1289 	if (m == NULL)
1290 		return;
1291 
1292 	rtm = mtod(m, struct rt_msghdr *);
1293 	if (ifp != NULL)
1294 		rtm->rtm_index = ifp->if_index;
1295 	rtm->rtm_flags |= rt->rt_flags;
1296 	rtm->rtm_errno = error;
1297 	rtm->rtm_addrs = rtinfo.rti_addrs;
1298 
1299 	rts_input(m, familyof(dst));
1300 }
1301 
1302 /*
1303  * This is called to generate messages from the routing socket
1304  * indicating a network interface has had addresses associated with it.
1305  * if we ever reverse the logic and replace messages TO the routing
1306  * socket indicate a request to configure interfaces, then it will
1307  * be unnecessary as the routing socket will automatically generate
1308  * copies of it.
1309  */
1310 void
1311 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1312 {
1313 	if (route_cb.any_count == 0)
1314 		return;
1315 
1316 	if (cmd == RTM_ADD) {
1317 		rt_ifamsg(RTM_NEWADDR, ifa);
1318 		rt_rtmsg(RTM_ADD, rt, ifa->ifa_ifp, error);
1319 	} else {
1320 		KASSERT((cmd == RTM_DELETE), ("unknown cmd %d", cmd));
1321 		rt_rtmsg(RTM_DELETE, rt, ifa->ifa_ifp, error);
1322 		rt_ifamsg(RTM_DELADDR, ifa);
1323 	}
1324 }
1325 
1326 /*
1327  * This is the analogue to the rt_newaddrmsg which performs the same
1328  * function but for multicast group memberhips.  This is easier since
1329  * there is no route state to worry about.
1330  */
1331 void
1332 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1333 {
1334 	struct rt_addrinfo rtinfo;
1335 	struct mbuf *m = NULL;
1336 	struct ifnet *ifp = ifma->ifma_ifp;
1337 	struct ifma_msghdr *ifmam;
1338 
1339 	if (route_cb.any_count == 0)
1340 		return;
1341 
1342 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1343 	rtinfo.rti_ifaaddr = ifma->ifma_addr;
1344 	if (ifp != NULL && !TAILQ_EMPTY(&ifp->if_addrheads[mycpuid])) {
1345 		rtinfo.rti_ifpaddr =
1346 		TAILQ_FIRST(&ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1347 	}
1348 	/*
1349 	 * If a link-layer address is present, present it as a ``gateway''
1350 	 * (similarly to how ARP entries, e.g., are presented).
1351 	 */
1352 	rtinfo.rti_gateway = ifma->ifma_lladdr;
1353 
1354 	m = rt_msg_mbuf(cmd, &rtinfo);
1355 	if (m == NULL)
1356 		return;
1357 
1358 	ifmam = mtod(m, struct ifma_msghdr *);
1359 	ifmam->ifmam_index = ifp->if_index;
1360 	ifmam->ifmam_addrs = rtinfo.rti_addrs;
1361 
1362 	rts_input(m, familyof(ifma->ifma_addr));
1363 }
1364 
1365 static struct mbuf *
1366 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1367 		     struct rt_addrinfo *info)
1368 {
1369 	struct if_announcemsghdr *ifan;
1370 	struct mbuf *m;
1371 
1372 	if (route_cb.any_count == 0)
1373 		return NULL;
1374 
1375 	bzero(info, sizeof(*info));
1376 	m = rt_msg_mbuf(type, info);
1377 	if (m == NULL)
1378 		return NULL;
1379 
1380 	ifan = mtod(m, struct if_announcemsghdr *);
1381 	ifan->ifan_index = ifp->if_index;
1382 	strlcpy(ifan->ifan_name, ifp->if_xname, sizeof ifan->ifan_name);
1383 	ifan->ifan_what = what;
1384 	return m;
1385 }
1386 
1387 /*
1388  * This is called to generate routing socket messages indicating
1389  * IEEE80211 wireless events.
1390  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1391  */
1392 void
1393 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1394 {
1395 	struct rt_addrinfo info;
1396 	struct mbuf *m;
1397 
1398 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1399 	if (m == NULL)
1400 		return;
1401 
1402 	/*
1403 	 * Append the ieee80211 data.  Try to stick it in the
1404 	 * mbuf containing the ifannounce msg; otherwise allocate
1405 	 * a new mbuf and append.
1406 	 *
1407 	 * NB: we assume m is a single mbuf.
1408 	 */
1409 	if (data_len > M_TRAILINGSPACE(m)) {
1410 		/* XXX use m_getb(data_len, M_NOWAIT, MT_DATA, 0); */
1411 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1412 		if (n == NULL) {
1413 			m_freem(m);
1414 			return;
1415 		}
1416 		KKASSERT(data_len <= M_TRAILINGSPACE(n));
1417 		bcopy(data, mtod(n, void *), data_len);
1418 		n->m_len = data_len;
1419 		m->m_next = n;
1420 	} else if (data_len > 0) {
1421 		bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1422 		m->m_len += data_len;
1423 	}
1424 	mbuftrackid(m, 33);
1425 	if (m->m_flags & M_PKTHDR)
1426 		m->m_pkthdr.len += data_len;
1427 	mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1428 	rts_input(m, 0);
1429 }
1430 
1431 /*
1432  * This is called to generate routing socket messages indicating
1433  * network interface arrival and departure.
1434  */
1435 void
1436 rt_ifannouncemsg(struct ifnet *ifp, int what)
1437 {
1438 	struct rt_addrinfo addrinfo;
1439 	struct mbuf *m;
1440 
1441 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &addrinfo);
1442 	if (m != NULL)
1443 		rts_input(m, 0);
1444 }
1445 
1446 static int
1447 resizewalkarg(struct walkarg *w, int len)
1448 {
1449 	void *newptr;
1450 
1451 	newptr = kmalloc(len, M_RTABLE, M_INTWAIT | M_NULLOK);
1452 	if (newptr == NULL)
1453 		return (ENOMEM);
1454 	if (w->w_tmem != NULL)
1455 		kfree(w->w_tmem, M_RTABLE);
1456 	w->w_tmem = newptr;
1457 	w->w_tmemsize = len;
1458 	bzero(newptr, len);
1459 
1460 	return (0);
1461 }
1462 
1463 static void
1464 ifnet_compute_stats(struct ifnet *ifp)
1465 {
1466 	IFNET_STAT_GET(ifp, ipackets, ifp->if_ipackets);
1467 	IFNET_STAT_GET(ifp, ierrors, ifp->if_ierrors);
1468 	IFNET_STAT_GET(ifp, opackets, ifp->if_opackets);
1469 	IFNET_STAT_GET(ifp, collisions, ifp->if_collisions);
1470 	IFNET_STAT_GET(ifp, ibytes, ifp->if_ibytes);
1471 	IFNET_STAT_GET(ifp, obytes, ifp->if_obytes);
1472 	IFNET_STAT_GET(ifp, imcasts, ifp->if_imcasts);
1473 	IFNET_STAT_GET(ifp, omcasts, ifp->if_omcasts);
1474 	IFNET_STAT_GET(ifp, iqdrops, ifp->if_iqdrops);
1475 	IFNET_STAT_GET(ifp, noproto, ifp->if_noproto);
1476 	IFNET_STAT_GET(ifp, oqdrops, ifp->if_oqdrops);
1477 }
1478 
1479 static int
1480 if_addrflags(const struct ifaddr *ifa)
1481 {
1482 	switch (ifa->ifa_addr->sa_family) {
1483 #ifdef INET6
1484 	case AF_INET6:
1485 		return ((const struct in6_ifaddr *)ifa)->ia6_flags;
1486 #endif
1487 	default:
1488 		return 0;
1489 	}
1490 }
1491 
1492 static int
1493 sysctl_iflist(int af, struct walkarg *w)
1494 {
1495 	struct ifnet *ifp;
1496 	struct rt_addrinfo rtinfo;
1497 	int msglen, error;
1498 
1499 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
1500 
1501 	ifnet_lock();
1502 	TAILQ_FOREACH(ifp, &ifnetlist, if_link) {
1503 		struct ifaddr_container *ifac, *ifac_mark;
1504 		struct ifaddr_marker mark;
1505 		struct ifaddrhead *head;
1506 		struct ifaddr *ifa;
1507 
1508 		if (w->w_arg && w->w_arg != ifp->if_index)
1509 			continue;
1510 		head = &ifp->if_addrheads[mycpuid];
1511 		/*
1512 		 * There is no need to reference the first ifaddr
1513 		 * even if the following resizewalkarg() blocks,
1514 		 * since the first ifaddr will not be destroyed
1515 		 * when the ifnet lock is held.
1516 		 */
1517 		ifac = TAILQ_FIRST(head);
1518 		ifa = ifac->ifa;
1519 		rtinfo.rti_ifpaddr = ifa->ifa_addr;
1520 		msglen = rt_msgsize(RTM_IFINFO, &rtinfo);
1521 		if (w->w_tmemsize < msglen && resizewalkarg(w, msglen) != 0) {
1522 			ifnet_unlock();
1523 			return (ENOMEM);
1524 		}
1525 		rt_msg_buffer(RTM_IFINFO, &rtinfo, w->w_tmem, msglen);
1526 		rtinfo.rti_ifpaddr = NULL;
1527 		if (w->w_req != NULL && w->w_tmem != NULL) {
1528 			struct if_msghdr *ifm = w->w_tmem;
1529 
1530 			ifm->ifm_index = ifp->if_index;
1531 			ifm->ifm_flags = ifp->if_flags;
1532 			ifnet_compute_stats(ifp);
1533 			ifm->ifm_data = ifp->if_data;
1534 			ifm->ifm_addrs = rtinfo.rti_addrs;
1535 			error = SYSCTL_OUT(w->w_req, ifm, msglen);
1536 			if (error) {
1537 				ifnet_unlock();
1538 				return (error);
1539 			}
1540 		}
1541 		/*
1542 		 * Add a marker, since SYSCTL_OUT() could block and during
1543 		 * that period the list could be changed.
1544 		 */
1545 		ifa_marker_init(&mark, ifp);
1546 		ifac_mark = &mark.ifac;
1547 		TAILQ_INSERT_AFTER(head, ifac, ifac_mark, ifa_link);
1548 		while ((ifac = TAILQ_NEXT(ifac_mark, ifa_link)) != NULL) {
1549 			TAILQ_REMOVE(head, ifac_mark, ifa_link);
1550 			TAILQ_INSERT_AFTER(head, ifac, ifac_mark, ifa_link);
1551 
1552 			ifa = ifac->ifa;
1553 
1554 			/* Ignore marker */
1555 			if (ifa->ifa_addr->sa_family == AF_UNSPEC)
1556 				continue;
1557 
1558 			if (af && af != ifa->ifa_addr->sa_family)
1559 				continue;
1560 			if (curproc->p_ucred->cr_prison &&
1561 			    prison_if(curproc->p_ucred, ifa->ifa_addr))
1562 				continue;
1563 			rtinfo.rti_ifaaddr = ifa->ifa_addr;
1564 			rtinfo.rti_netmask = ifa->ifa_netmask;
1565 			rtinfo.rti_bcastaddr = ifa->ifa_dstaddr;
1566 			msglen = rt_msgsize(RTM_NEWADDR, &rtinfo);
1567 			/*
1568 			 * Keep a reference on this ifaddr, so that it will
1569 			 * not be destroyed if the following resizewalkarg()
1570 			 * blocks.
1571 			 */
1572 			IFAREF(ifa);
1573 			if (w->w_tmemsize < msglen &&
1574 			    resizewalkarg(w, msglen) != 0) {
1575 				IFAFREE(ifa);
1576 				TAILQ_REMOVE(head, ifac_mark, ifa_link);
1577 				ifnet_unlock();
1578 				return (ENOMEM);
1579 			}
1580 			rt_msg_buffer(RTM_NEWADDR, &rtinfo, w->w_tmem, msglen);
1581 			if (w->w_req != NULL) {
1582 				struct ifa_msghdr *ifam = w->w_tmem;
1583 
1584 				ifam->ifam_index = ifa->ifa_ifp->if_index;
1585 				ifam->ifam_flags = ifa->ifa_flags;
1586 				ifam->ifam_addrs = rtinfo.rti_addrs;
1587 				ifam->ifam_addrflags = if_addrflags(ifa);
1588 				ifam->ifam_metric = ifa->ifa_metric;
1589 				error = SYSCTL_OUT(w->w_req, w->w_tmem, msglen);
1590 				if (error) {
1591 					IFAFREE(ifa);
1592 					TAILQ_REMOVE(head, ifac_mark, ifa_link);
1593 					ifnet_unlock();
1594 					return (error);
1595 				}
1596 			}
1597 			IFAFREE(ifa);
1598 		}
1599 		TAILQ_REMOVE(head, ifac_mark, ifa_link);
1600 		rtinfo.rti_netmask = NULL;
1601 		rtinfo.rti_ifaaddr = NULL;
1602 		rtinfo.rti_bcastaddr = NULL;
1603 	}
1604 	ifnet_unlock();
1605 	return (0);
1606 }
1607 
1608 static int
1609 rttable_walkarg_create(struct rttable_walkarg *w, int op, int arg)
1610 {
1611 	struct rt_addrinfo rtinfo;
1612 	struct sockaddr_storage ss;
1613 	int i, msglen;
1614 
1615 	memset(w, 0, sizeof(*w));
1616 	w->w_op = op;
1617 	w->w_arg = arg;
1618 
1619 	memset(&ss, 0, sizeof(ss));
1620 	ss.ss_len = sizeof(ss);
1621 
1622 	memset(&rtinfo, 0, sizeof(rtinfo));
1623 	for (i = 0; i < RTAX_MAX; ++i)
1624 		rtinfo.rti_info[i] = (struct sockaddr *)&ss;
1625 	msglen = rt_msgsize(RTM_GET, &rtinfo);
1626 
1627 	w->w_bufsz = msglen * RTTABLE_DUMP_MSGCNT_MAX;
1628 	w->w_buf = kmalloc(w->w_bufsz, M_TEMP, M_WAITOK | M_NULLOK);
1629 	if (w->w_buf == NULL)
1630 		return ENOMEM;
1631 	return 0;
1632 }
1633 
1634 static void
1635 rttable_walkarg_destroy(struct rttable_walkarg *w)
1636 {
1637 	kfree(w->w_buf, M_TEMP);
1638 }
1639 
1640 static void
1641 rttable_entry_rtinfo(struct rt_addrinfo *rtinfo, struct radix_node *rn)
1642 {
1643 	struct rtentry *rt = (struct rtentry *)rn;
1644 
1645 	bzero(rtinfo, sizeof(*rtinfo));
1646 	rtinfo->rti_dst = rt_key(rt);
1647 	rtinfo->rti_gateway = rt->rt_gateway;
1648 	rtinfo->rti_netmask = rt_mask(rt);
1649 	rtinfo->rti_genmask = rt->rt_genmask;
1650 	if (rt->rt_ifp != NULL) {
1651 		rtinfo->rti_ifpaddr =
1652 		TAILQ_FIRST(&rt->rt_ifp->if_addrheads[mycpuid])->ifa->ifa_addr;
1653 		rtinfo->rti_ifaaddr = rt->rt_ifa->ifa_addr;
1654 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1655 			rtinfo->rti_bcastaddr = rt->rt_ifa->ifa_dstaddr;
1656 	}
1657 }
1658 
1659 static int
1660 rttable_walk_entry(struct radix_node *rn, void *xw)
1661 {
1662 	struct rttable_walkarg *w = xw;
1663 	struct rtentry *rt = (struct rtentry *)rn;
1664 	struct rt_addrinfo rtinfo;
1665 	struct rt_msghdr *rtm;
1666 	boolean_t save = FALSE;
1667 	int msglen, w_bufleft;
1668 	void *ptr;
1669 
1670 	rttable_entry_rtinfo(&rtinfo, rn);
1671 	msglen = rt_msgsize(RTM_GET, &rtinfo);
1672 
1673 	w_bufleft = w->w_bufsz - w->w_buflen;
1674 
1675 	if (rn->rn_dupedkey != NULL) {
1676 		struct radix_node *rn1 = rn;
1677 		int total_msglen = msglen;
1678 
1679 		/*
1680 		 * Make sure that we have enough space left for all
1681 		 * dupedkeys, since rn_walktree_at always starts
1682 		 * from the first dupedkey.
1683 		 */
1684 		while ((rn1 = rn1->rn_dupedkey) != NULL) {
1685 			struct rt_addrinfo rtinfo1;
1686 			int msglen1;
1687 
1688 			if (rn1->rn_flags & RNF_ROOT)
1689 				continue;
1690 
1691 			rttable_entry_rtinfo(&rtinfo1, rn1);
1692 			msglen1 = rt_msgsize(RTM_GET, &rtinfo1);
1693 			total_msglen += msglen1;
1694 		}
1695 
1696 		if (total_msglen > w_bufleft) {
1697 			if (total_msglen > w->w_bufsz) {
1698 				static int logged = 0;
1699 
1700 				if (!logged) {
1701 					kprintf("buffer is too small for "
1702 					    "all dupedkeys, increase "
1703 					    "RTTABLE_DUMP_MSGCNT_MAX\n");
1704 					logged = 1;
1705 				}
1706 				return ENOMEM;
1707 			}
1708 			save = TRUE;
1709 		}
1710 	} else if (msglen > w_bufleft) {
1711 		save = TRUE;
1712 	}
1713 
1714 	if (save) {
1715 		/*
1716 		 * Not enough buffer left; remember the position
1717 		 * to start from upon next round.
1718 		 */
1719 		KASSERT(msglen <= w->w_bufsz, ("msg too long %d", msglen));
1720 
1721 		KASSERT(rtinfo.rti_dst->sa_len <= sizeof(w->w_key0),
1722 		    ("key too long %d", rtinfo.rti_dst->sa_len));
1723 		memset(&w->w_key0, 0, sizeof(w->w_key0));
1724 		memcpy(&w->w_key0, rtinfo.rti_dst, rtinfo.rti_dst->sa_len);
1725 		w->w_key = (const char *)&w->w_key0;
1726 
1727 		if (rtinfo.rti_netmask != NULL) {
1728 			KASSERT(
1729 			    rtinfo.rti_netmask->sa_len <= sizeof(w->w_mask0),
1730 			    ("mask too long %d", rtinfo.rti_netmask->sa_len));
1731 			memset(&w->w_mask0, 0, sizeof(w->w_mask0));
1732 			memcpy(&w->w_mask0, rtinfo.rti_netmask,
1733 			    rtinfo.rti_netmask->sa_len);
1734 			w->w_mask = (const char *)&w->w_mask0;
1735 		} else {
1736 			w->w_mask = NULL;
1737 		}
1738 		return EJUSTRETURN;
1739 	}
1740 
1741 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1742 		return 0;
1743 
1744 	ptr = ((uint8_t *)w->w_buf) + w->w_buflen;
1745 	rt_msg_buffer(RTM_GET, &rtinfo, ptr, msglen);
1746 
1747 	rtm = (struct rt_msghdr *)ptr;
1748 	rtm->rtm_flags = rt->rt_flags;
1749 	rtm->rtm_use = rt->rt_use;
1750 	rtm->rtm_rmx = rt->rt_rmx;
1751 	rtm->rtm_index = rt->rt_ifp->if_index;
1752 	rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1753 	rtm->rtm_addrs = rtinfo.rti_addrs;
1754 
1755 	w->w_buflen += msglen;
1756 
1757 	return 0;
1758 }
1759 
1760 static void
1761 rttable_walk_dispatch(netmsg_t msg)
1762 {
1763 	struct netmsg_rttable_walk *nmsg = (struct netmsg_rttable_walk *)msg;
1764 	struct radix_node_head *rnh = rt_tables[mycpuid][nmsg->af];
1765 	struct rttable_walkarg *w = nmsg->w;
1766 	int error;
1767 
1768 	error = rnh->rnh_walktree_at(rnh, w->w_key, w->w_mask,
1769 	    rttable_walk_entry, w);
1770 	lwkt_replymsg(&nmsg->base.lmsg, error);
1771 }
1772 
1773 static int
1774 sysctl_rttable(int af, struct sysctl_req *req, int op, int arg)
1775 {
1776 	struct rttable_walkarg w;
1777 	int error, i;
1778 
1779 	error = rttable_walkarg_create(&w, op, arg);
1780 	if (error)
1781 		return error;
1782 
1783 	error = EINVAL;
1784 	for (i = 1; i <= AF_MAX; i++) {
1785 		if (rt_tables[mycpuid][i] != NULL && (af == 0 || af == i)) {
1786 			w.w_key = NULL;
1787 			w.w_mask = NULL;
1788 			for (;;) {
1789 				struct netmsg_rttable_walk nmsg;
1790 
1791 				netmsg_init(&nmsg.base, NULL,
1792 				    &curthread->td_msgport, 0,
1793 				    rttable_walk_dispatch);
1794 				nmsg.af = i;
1795 				nmsg.w = &w;
1796 
1797 				w.w_buflen = 0;
1798 
1799 				error = lwkt_domsg(netisr_cpuport(mycpuid),
1800 				    &nmsg.base.lmsg, 0);
1801 				if (error && error != EJUSTRETURN)
1802 					goto done;
1803 
1804 				if (req != NULL && w.w_buflen > 0) {
1805 					int error1;
1806 
1807 					error1 = SYSCTL_OUT(req, w.w_buf,
1808 					    w.w_buflen);
1809 					if (error1) {
1810 						error = error1;
1811 						goto done;
1812 					}
1813 				}
1814 				if (error == 0) /* done */
1815 					break;
1816 			}
1817 		}
1818 	}
1819 done:
1820 	rttable_walkarg_destroy(&w);
1821 	return error;
1822 }
1823 
1824 static int
1825 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1826 {
1827 	int	*name = (int *)arg1;
1828 	u_int	namelen = arg2;
1829 	int	error = EINVAL;
1830 	int	origcpu, cpu;
1831 	u_char  af;
1832 	struct	walkarg w;
1833 
1834 	name ++;
1835 	namelen--;
1836 	if (req->newptr)
1837 		return (EPERM);
1838 	if (namelen != 3 && namelen != 4)
1839 		return (EINVAL);
1840 	af = name[0];
1841 	bzero(&w, sizeof w);
1842 	w.w_op = name[1];
1843 	w.w_arg = name[2];
1844 	w.w_req = req;
1845 
1846 	/*
1847 	 * Optional third argument specifies cpu, used primarily for
1848 	 * debugging the route table.
1849 	 */
1850 	if (namelen == 4) {
1851 		if (name[3] < 0 || name[3] >= netisr_ncpus)
1852 			return (EINVAL);
1853 		cpu = name[3];
1854 	} else {
1855 		/*
1856 		 * Target cpu is not specified, use cpu0 then, so that
1857 		 * the result set will be relatively stable.
1858 		 */
1859 		cpu = 0;
1860 	}
1861 	origcpu = mycpuid;
1862 	lwkt_migratecpu(cpu);
1863 
1864 	switch (w.w_op) {
1865 	case NET_RT_DUMP:
1866 	case NET_RT_FLAGS:
1867 		error = sysctl_rttable(af, w.w_req, w.w_op, w.w_arg);
1868 		break;
1869 
1870 	case NET_RT_IFLIST:
1871 		error = sysctl_iflist(af, &w);
1872 		break;
1873 	}
1874 	if (w.w_tmem != NULL)
1875 		kfree(w.w_tmem, M_RTABLE);
1876 
1877 	lwkt_migratecpu(origcpu);
1878 	return (error);
1879 }
1880 
1881 SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1882 
1883 /*
1884  * Definitions of protocols supported in the ROUTE domain.
1885  */
1886 
1887 static struct domain routedomain;		/* or at least forward */
1888 
1889 static struct protosw routesw[] = {
1890     {
1891 	.pr_type = SOCK_RAW,
1892 	.pr_domain = &routedomain,
1893 	.pr_protocol = 0,
1894 	.pr_flags = PR_ATOMIC|PR_ADDR,
1895 	.pr_input = NULL,
1896 	.pr_output = route_output,
1897 	.pr_ctlinput = raw_ctlinput,
1898 	.pr_ctloutput = route_ctloutput,
1899 	.pr_ctlport = cpu0_ctlport,
1900 
1901 	.pr_init = raw_init,
1902 	.pr_usrreqs = &route_usrreqs
1903     }
1904 };
1905 
1906 static struct domain routedomain = {
1907 	.dom_family		= AF_ROUTE,
1908 	.dom_name		= "route",
1909 	.dom_init		= NULL,
1910 	.dom_externalize	= NULL,
1911 	.dom_dispose		= NULL,
1912 	.dom_protosw		= routesw,
1913 	.dom_protoswNPROTOSW	= &routesw[(sizeof routesw)/(sizeof routesw[0])],
1914 	.dom_next		= SLIST_ENTRY_INITIALIZER,
1915 	.dom_rtattach		= NULL,
1916 	.dom_rtoffset		= 0,
1917 	.dom_maxrtkey		= 0,
1918 	.dom_ifattach		= NULL,
1919 	.dom_ifdetach		= NULL
1920 };
1921 
1922 DOMAIN_SET(route);
1923 
1924