54
55MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
56
57static struct sockaddr route_dst = { 2, PF_ROUTE, };
58static struct sockaddr route_src = { 2, PF_ROUTE, };
59static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, };
60static struct sockproto route_proto = { PF_ROUTE, };
61
62struct walkarg {
63 int w_tmemsize;
64 int w_op, w_arg;
65 caddr_t w_tmem;
66 struct sysctl_req *w_req;
67};
68
69static struct mbuf *
70 rt_msg1(int, struct rt_addrinfo *);
71static int rt_msg2(int, struct rt_addrinfo *, caddr_t, struct walkarg *);
72static int rt_xaddrs(caddr_t, caddr_t, struct rt_addrinfo *);
73static int sysctl_dumpentry(struct radix_node *rn, void *vw);
74static int sysctl_iflist(int af, struct walkarg *w);
75static int route_output(struct mbuf *, struct socket *);
76static void rt_setmetrics(u_long, struct rt_metrics *, struct rt_metrics *);
77
78/* Sleazy use of local variables throughout file, warning!!!! */
79#define dst info.rti_info[RTAX_DST]
80#define gate info.rti_info[RTAX_GATEWAY]
81#define netmask info.rti_info[RTAX_NETMASK]
82#define genmask info.rti_info[RTAX_GENMASK]
83#define ifpaddr info.rti_info[RTAX_IFP]
84#define ifaaddr info.rti_info[RTAX_IFA]
85#define brdaddr info.rti_info[RTAX_BRD]
86
87/*
88 * It really doesn't make any sense at all for this code to share much
89 * with raw_usrreq.c, since its functionality is so restricted. XXX
90 */
91static int
92rts_abort(struct socket *so)
93{
94 int s, error;
95 s = splnet();
96 error = raw_usrreqs.pru_abort(so);
97 splx(s);
98 return error;
99}
100
101/* pru_accept is EOPNOTSUPP */
102
103static int
104rts_attach(struct socket *so, int proto, struct thread *td)
105{
106 struct rawcb *rp;
107 int s, error;
108
109 if (sotorawcb(so) != 0)
110 return EISCONN; /* XXX panic? */
111 /* XXX */
112 MALLOC(rp, struct rawcb *, sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
113 if (rp == 0)
114 return ENOBUFS;
115
116 /*
117 * The splnet() is necessary to block protocols from sending
118 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
119 * this PCB is extant but incompletely initialized.
120 * Probably we should try to do more of this work beforehand and
121 * eliminate the spl.
122 */
123 s = splnet();
124 so->so_pcb = (caddr_t)rp;
125 error = raw_attach(so, proto);
126 rp = sotorawcb(so);
127 if (error) {
128 splx(s);
129 so->so_pcb = NULL;
130 free(rp, M_PCB);
131 return error;
132 }
133 switch(rp->rcb_proto.sp_protocol) {
134 case AF_INET:
135 route_cb.ip_count++;
136 break;
137 case AF_INET6:
138 route_cb.ip6_count++;
139 break;
140 case AF_IPX:
141 route_cb.ipx_count++;
142 break;
143 case AF_NS:
144 route_cb.ns_count++;
145 break;
146 }
147 rp->rcb_faddr = &route_src;
148 route_cb.any_count++;
149 SIGIO_SLOCK();
150 soisconnected_locked(so);
151 so->so_options |= SO_USELOOPBACK;
152 SIGIO_SUNLOCK();
153 splx(s);
154 return 0;
155}
156
157static int
158rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
159{
160 int s, error;
161 s = splnet();
162 error = raw_usrreqs.pru_bind(so, nam, td); /* xxx just EINVAL */
163 splx(s);
164 return error;
165}
166
167static int
168rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
169{
170 int s, error;
171 s = splnet();
172 error = raw_usrreqs.pru_connect(so, nam, td); /* XXX just EINVAL */
173 splx(s);
174 return error;
175}
176
177/* pru_connect2 is EOPNOTSUPP */
178/* pru_control is EOPNOTSUPP */
179
180static int
181rts_detach(struct socket *so)
182{
183 struct rawcb *rp = sotorawcb(so);
184 int s, error;
185
186 s = splnet();
187 if (rp != 0) {
188 switch(rp->rcb_proto.sp_protocol) {
189 case AF_INET:
190 route_cb.ip_count--;
191 break;
192 case AF_INET6:
193 route_cb.ip6_count--;
194 break;
195 case AF_IPX:
196 route_cb.ipx_count--;
197 break;
198 case AF_NS:
199 route_cb.ns_count--;
200 break;
201 }
202 route_cb.any_count--;
203 }
204 error = raw_usrreqs.pru_detach(so);
205 splx(s);
206 return error;
207}
208
209static int
210rts_disconnect(struct socket *so)
211{
212 int s, error;
213 s = splnet();
214 error = raw_usrreqs.pru_disconnect(so);
215 splx(s);
216 return error;
217}
218
219/* pru_listen is EOPNOTSUPP */
220
221static int
222rts_peeraddr(struct socket *so, struct sockaddr **nam)
223{
224 int s, error;
225 s = splnet();
226 error = raw_usrreqs.pru_peeraddr(so, nam);
227 splx(s);
228 return error;
229}
230
231/* pru_rcvd is EOPNOTSUPP */
232/* pru_rcvoob is EOPNOTSUPP */
233
234static int
235rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
236 struct mbuf *control, struct thread *td)
237{
238 int s, error;
239 s = splnet();
240 error = raw_usrreqs.pru_send(so, flags, m, nam, control, td);
241 splx(s);
242 return error;
243}
244
245/* pru_sense is null */
246
247static int
248rts_shutdown(struct socket *so)
249{
250 int s, error;
251 s = splnet();
252 error = raw_usrreqs.pru_shutdown(so);
253 splx(s);
254 return error;
255}
256
257static int
258rts_sockaddr(struct socket *so, struct sockaddr **nam)
259{
260 int s, error;
261 s = splnet();
262 error = raw_usrreqs.pru_sockaddr(so, nam);
263 splx(s);
264 return error;
265}
266
267static struct pr_usrreqs route_usrreqs = {
268 rts_abort, pru_accept_notsupp, rts_attach, rts_bind, rts_connect,
269 pru_connect2_notsupp, pru_control_notsupp, rts_detach, rts_disconnect,
270 pru_listen_notsupp, rts_peeraddr, pru_rcvd_notsupp, pru_rcvoob_notsupp,
271 rts_send, pru_sense_null, rts_shutdown, rts_sockaddr,
272 sosend, soreceive, sopoll
273};
274
275/*ARGSUSED*/
276static int
277route_output(m, so)
278 register struct mbuf *m;
279 struct socket *so;
280{
281 register struct rt_msghdr *rtm = 0;
282 register struct rtentry *rt = 0;
283 struct rtentry *saved_nrt = 0;
284 struct radix_node_head *rnh;
285 struct rt_addrinfo info;
286 int len, error = 0;
287 struct ifnet *ifp = 0;
288 struct ifaddr *ifa = 0;
289
290#define senderr(e) { error = e; goto flush;}
291 if (m == 0 || ((m->m_len < sizeof(long)) &&
292 (m = m_pullup(m, sizeof(long))) == 0))
293 return (ENOBUFS);
294 if ((m->m_flags & M_PKTHDR) == 0)
295 panic("route_output");
296 len = m->m_pkthdr.len;
297 if (len < sizeof(*rtm) ||
298 len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
299 dst = 0;
300 senderr(EINVAL);
301 }
302 R_Malloc(rtm, struct rt_msghdr *, len);
303 if (rtm == 0) {
304 dst = 0;
305 senderr(ENOBUFS);
306 }
307 m_copydata(m, 0, len, (caddr_t)rtm);
308 if (rtm->rtm_version != RTM_VERSION) {
309 dst = 0;
310 senderr(EPROTONOSUPPORT);
311 }
312 rtm->rtm_pid = curproc->p_pid;
313 bzero(&info, sizeof(info));
314 info.rti_addrs = rtm->rtm_addrs;
315 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
316 dst = 0;
317 senderr(EINVAL);
318 }
319 info.rti_flags = rtm->rtm_flags;
320 if (dst == 0 || (dst->sa_family >= AF_MAX)
321 || (gate != 0 && (gate->sa_family >= AF_MAX)))
322 senderr(EINVAL);
323 if (genmask) {
324 struct radix_node *t;
325 t = rn_addmask((caddr_t)genmask, 0, 1);
326 if (t && Bcmp((caddr_t *)genmask + 1, (caddr_t *)t->rn_key + 1,
327 *(u_char *)t->rn_key - 1) == 0)
328 genmask = (struct sockaddr *)(t->rn_key);
329 else
330 senderr(ENOBUFS);
331 }
332
333 /*
334 * Verify that the caller has the appropriate privilege; RTM_GET
335 * is the only operation the non-superuser is allowed.
336 */
337 if (rtm->rtm_type != RTM_GET && (error = suser(curthread)) != 0)
338 senderr(error);
339
340 switch (rtm->rtm_type) {
341
342 case RTM_ADD:
343 if (gate == 0)
344 senderr(EINVAL);
345 error = rtrequest1(RTM_ADD, &info, &saved_nrt);
346 if (error == 0 && saved_nrt) {
347 rt_setmetrics(rtm->rtm_inits,
348 &rtm->rtm_rmx, &saved_nrt->rt_rmx);
349 saved_nrt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
350 saved_nrt->rt_rmx.rmx_locks |=
351 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
352 saved_nrt->rt_refcnt--;
353 saved_nrt->rt_genmask = genmask;
354 }
355 break;
356
357 case RTM_DELETE:
358 error = rtrequest1(RTM_DELETE, &info, &saved_nrt);
359 if (error == 0) {
360 if ((rt = saved_nrt))
361 rt->rt_refcnt++;
362 goto report;
363 }
364 break;
365
366 case RTM_GET:
367 case RTM_CHANGE:
368 case RTM_LOCK:
369 if ((rnh = rt_tables[dst->sa_family]) == 0) {
370 senderr(EAFNOSUPPORT);
371 } else if ((rt = (struct rtentry *)
372 rnh->rnh_lookup(dst, netmask, rnh)) != NULL)
373 rt->rt_refcnt++;
374 else
375 senderr(ESRCH);
376 switch(rtm->rtm_type) {
377
378 case RTM_GET:
379 report:
380 dst = rt_key(rt);
381 gate = rt->rt_gateway;
382 netmask = rt_mask(rt);
383 genmask = rt->rt_genmask;
384 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
385 ifp = rt->rt_ifp;
386 if (ifp) {
387 ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
388 ifaaddr = rt->rt_ifa->ifa_addr;
389 if (ifp->if_flags & IFF_POINTOPOINT)
390 brdaddr = rt->rt_ifa->ifa_dstaddr;
391 rtm->rtm_index = ifp->if_index;
392 } else {
393 ifpaddr = 0;
394 ifaaddr = 0;
395 }
396 }
397 len = rt_msg2(rtm->rtm_type, &info, (caddr_t)0,
398 (struct walkarg *)0);
399 if (len > rtm->rtm_msglen) {
400 struct rt_msghdr *new_rtm;
401 R_Malloc(new_rtm, struct rt_msghdr *, len);
402 if (new_rtm == 0)
403 senderr(ENOBUFS);
404 Bcopy(rtm, new_rtm, rtm->rtm_msglen);
405 Free(rtm); rtm = new_rtm;
406 }
407 (void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm,
408 (struct walkarg *)0);
409 rtm->rtm_flags = rt->rt_flags;
410 rtm->rtm_rmx = rt->rt_rmx;
411 rtm->rtm_addrs = info.rti_addrs;
412 break;
413
414 case RTM_CHANGE:
415 /* new gateway could require new ifaddr, ifp;
416 flags may also be different; ifp may be specified
417 by ll sockaddr when protocol address is ambiguous */
418#define equal(a1, a2) (bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0)
419 if ((rt->rt_flags & RTF_GATEWAY && gate != NULL) ||
420 ifpaddr != NULL ||
421 (ifaaddr != NULL &&
422 !equal(ifaaddr, rt->rt_ifa->ifa_addr))) {
423 if ((error = rt_getifa(&info)) != 0)
424 senderr(error);
425 }
426 if (gate != NULL &&
427 (error = rt_setgate(rt, rt_key(rt), gate)) != 0)
428 senderr(error);
429 if ((ifa = info.rti_ifa) != NULL) {
430 register struct ifaddr *oifa = rt->rt_ifa;
431 if (oifa != ifa) {
432 if (oifa && oifa->ifa_rtrequest)
433 oifa->ifa_rtrequest(RTM_DELETE, rt,
434 &info);
435 IFAFREE(rt->rt_ifa);
436 rt->rt_ifa = ifa;
437 ifa->ifa_refcnt++;
438 rt->rt_ifp = info.rti_ifp;
439 }
440 }
441 rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
442 &rt->rt_rmx);
443 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
444 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
445 if (genmask)
446 rt->rt_genmask = genmask;
447 /*
448 * Fall into
449 */
450 case RTM_LOCK:
451 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
452 rt->rt_rmx.rmx_locks |=
453 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
454 break;
455 }
456 break;
457
458 default:
459 senderr(EOPNOTSUPP);
460 }
461
462flush:
463 if (rtm) {
464 if (error)
465 rtm->rtm_errno = error;
466 else
467 rtm->rtm_flags |= RTF_DONE;
468 }
469 if (rt)
470 rtfree(rt);
471 {
472 register struct rawcb *rp = 0;
473 /*
474 * Check to see if we don't want our own messages.
475 */
476 if ((so->so_options & SO_USELOOPBACK) == 0) {
477 if (route_cb.any_count <= 1) {
478 if (rtm)
479 Free(rtm);
480 m_freem(m);
481 return (error);
482 }
483 /* There is another listener, so construct message */
484 rp = sotorawcb(so);
485 }
486 if (rtm) {
487 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
488 if (m->m_pkthdr.len < rtm->rtm_msglen) {
489 m_freem(m);
490 m = NULL;
491 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
492 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
493 Free(rtm);
494 }
495 if (rp)
496 rp->rcb_proto.sp_family = 0; /* Avoid us */
497 if (dst)
498 route_proto.sp_protocol = dst->sa_family;
499 if (m)
500 raw_input(m, &route_proto, &route_src, &route_dst);
501 if (rp)
502 rp->rcb_proto.sp_family = PF_ROUTE;
503 }
504 return (error);
505}
506
507static void
508rt_setmetrics(which, in, out)
509 u_long which;
510 register struct rt_metrics *in, *out;
511{
512#define metric(f, e) if (which & (f)) out->e = in->e;
513 metric(RTV_RPIPE, rmx_recvpipe);
514 metric(RTV_SPIPE, rmx_sendpipe);
515 metric(RTV_SSTHRESH, rmx_ssthresh);
516 metric(RTV_RTT, rmx_rtt);
517 metric(RTV_RTTVAR, rmx_rttvar);
518 metric(RTV_HOPCOUNT, rmx_hopcount);
519 metric(RTV_MTU, rmx_mtu);
520 metric(RTV_EXPIRE, rmx_expire);
521#undef metric
522}
523
524#define ROUNDUP(a) \
525 ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
526#define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
527
528
529/*
530 * Extract the addresses of the passed sockaddrs.
531 * Do a little sanity checking so as to avoid bad memory references.
532 * This data is derived straight from userland.
533 */
534static int
535rt_xaddrs(cp, cplim, rtinfo)
536 register caddr_t cp, cplim;
537 register struct rt_addrinfo *rtinfo;
538{
539 register struct sockaddr *sa;
540 register int i;
541
542 for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
543 if ((rtinfo->rti_addrs & (1 << i)) == 0)
544 continue;
545 sa = (struct sockaddr *)cp;
546 /*
547 * It won't fit.
548 */
549 if ( (cp + sa->sa_len) > cplim ) {
550 return (EINVAL);
551 }
552
553 /*
554 * there are no more.. quit now
555 * If there are more bits, they are in error.
556 * I've seen this. route(1) can evidently generate these.
557 * This causes kernel to core dump.
558 * for compatibility, If we see this, point to a safe address.
559 */
560 if (sa->sa_len == 0) {
561 rtinfo->rti_info[i] = &sa_zero;
562 return (0); /* should be EINVAL but for compat */
563 }
564
565 /* accept it */
566 rtinfo->rti_info[i] = sa;
567 ADVANCE(cp, sa);
568 }
569 return (0);
570}
571
572static struct mbuf *
573rt_msg1(type, rtinfo)
574 int type;
575 register struct rt_addrinfo *rtinfo;
576{
577 register struct rt_msghdr *rtm;
578 register struct mbuf *m;
579 register int i;
580 register struct sockaddr *sa;
581 int len, dlen;
582
583 switch (type) {
584
585 case RTM_DELADDR:
586 case RTM_NEWADDR:
587 len = sizeof(struct ifa_msghdr);
588 break;
589
590 case RTM_DELMADDR:
591 case RTM_NEWMADDR:
592 len = sizeof(struct ifma_msghdr);
593 break;
594
595 case RTM_IFINFO:
596 len = sizeof(struct if_msghdr);
597 break;
598
599 case RTM_IFANNOUNCE:
600 len = sizeof(struct if_announcemsghdr);
601 break;
602
603 default:
604 len = sizeof(struct rt_msghdr);
605 }
606 if (len > MCLBYTES)
607 panic("rt_msg1");
608 m = m_gethdr(M_DONTWAIT, MT_DATA);
609 if (m && len > MHLEN) {
610 MCLGET(m, M_DONTWAIT);
611 if ((m->m_flags & M_EXT) == 0) {
612 m_free(m);
613 m = NULL;
614 }
615 }
616 if (m == 0)
617 return (m);
618 m->m_pkthdr.len = m->m_len = len;
619 m->m_pkthdr.rcvif = 0;
620 rtm = mtod(m, struct rt_msghdr *);
621 bzero((caddr_t)rtm, len);
622 for (i = 0; i < RTAX_MAX; i++) {
623 if ((sa = rtinfo->rti_info[i]) == NULL)
624 continue;
625 rtinfo->rti_addrs |= (1 << i);
626 dlen = ROUNDUP(sa->sa_len);
627 m_copyback(m, len, dlen, (caddr_t)sa);
628 len += dlen;
629 }
630 if (m->m_pkthdr.len != len) {
631 m_freem(m);
632 return (NULL);
633 }
634 rtm->rtm_msglen = len;
635 rtm->rtm_version = RTM_VERSION;
636 rtm->rtm_type = type;
637 return (m);
638}
639
640static int
641rt_msg2(type, rtinfo, cp, w)
642 int type;
643 register struct rt_addrinfo *rtinfo;
644 caddr_t cp;
645 struct walkarg *w;
646{
647 register int i;
648 int len, dlen, second_time = 0;
649 caddr_t cp0;
650
651 rtinfo->rti_addrs = 0;
652again:
653 switch (type) {
654
655 case RTM_DELADDR:
656 case RTM_NEWADDR:
657 len = sizeof(struct ifa_msghdr);
658 break;
659
660 case RTM_IFINFO:
661 len = sizeof(struct if_msghdr);
662 break;
663
664 default:
665 len = sizeof(struct rt_msghdr);
666 }
667 cp0 = cp;
668 if (cp0)
669 cp += len;
670 for (i = 0; i < RTAX_MAX; i++) {
671 register struct sockaddr *sa;
672
673 if ((sa = rtinfo->rti_info[i]) == 0)
674 continue;
675 rtinfo->rti_addrs |= (1 << i);
676 dlen = ROUNDUP(sa->sa_len);
677 if (cp) {
678 bcopy((caddr_t)sa, cp, (unsigned)dlen);
679 cp += dlen;
680 }
681 len += dlen;
682 }
683 len = ALIGN(len);
684 if (cp == 0 && w != NULL && !second_time) {
685 register struct walkarg *rw = w;
686
687 if (rw->w_req) {
688 if (rw->w_tmemsize < len) {
689 if (rw->w_tmem)
690 free(rw->w_tmem, M_RTABLE);
691 rw->w_tmem = (caddr_t)
692 malloc(len, M_RTABLE, M_NOWAIT);
693 if (rw->w_tmem)
694 rw->w_tmemsize = len;
695 }
696 if (rw->w_tmem) {
697 cp = rw->w_tmem;
698 second_time = 1;
699 goto again;
700 }
701 }
702 }
703 if (cp) {
704 register struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
705
706 rtm->rtm_version = RTM_VERSION;
707 rtm->rtm_type = type;
708 rtm->rtm_msglen = len;
709 }
710 return (len);
711}
712
713/*
714 * This routine is called to generate a message from the routing
715 * socket indicating that a redirect has occured, a routing lookup
716 * has failed, or that a protocol has detected timeouts to a particular
717 * destination.
718 */
719void
720rt_missmsg(type, rtinfo, flags, error)
721 int type, flags, error;
722 register struct rt_addrinfo *rtinfo;
723{
724 register struct rt_msghdr *rtm;
725 register struct mbuf *m;
726 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
727
728 if (route_cb.any_count == 0)
729 return;
730 m = rt_msg1(type, rtinfo);
731 if (m == 0)
732 return;
733 rtm = mtod(m, struct rt_msghdr *);
734 rtm->rtm_flags = RTF_DONE | flags;
735 rtm->rtm_errno = error;
736 rtm->rtm_addrs = rtinfo->rti_addrs;
737 route_proto.sp_protocol = sa ? sa->sa_family : 0;
738 raw_input(m, &route_proto, &route_src, &route_dst);
739}
740
741/*
742 * This routine is called to generate a message from the routing
743 * socket indicating that the status of a network interface has changed.
744 */
745void
746rt_ifmsg(ifp)
747 register struct ifnet *ifp;
748{
749 register struct if_msghdr *ifm;
750 struct mbuf *m;
751 struct rt_addrinfo info;
752
753 if (route_cb.any_count == 0)
754 return;
755 bzero((caddr_t)&info, sizeof(info));
756 m = rt_msg1(RTM_IFINFO, &info);
757 if (m == 0)
758 return;
759 ifm = mtod(m, struct if_msghdr *);
760 ifm->ifm_index = ifp->if_index;
761 ifm->ifm_flags = (u_short)ifp->if_flags;
762 ifm->ifm_data = ifp->if_data;
763 ifm->ifm_addrs = 0;
764 route_proto.sp_protocol = 0;
765 raw_input(m, &route_proto, &route_src, &route_dst);
766}
767
768/*
769 * This is called to generate messages from the routing socket
770 * indicating a network interface has had addresses associated with it.
771 * if we ever reverse the logic and replace messages TO the routing
772 * socket indicate a request to configure interfaces, then it will
773 * be unnecessary as the routing socket will automatically generate
774 * copies of it.
775 */
776void
777rt_newaddrmsg(cmd, ifa, error, rt)
778 int cmd, error;
779 register struct ifaddr *ifa;
780 register struct rtentry *rt;
781{
782 struct rt_addrinfo info;
783 struct sockaddr *sa = 0;
784 int pass;
785 struct mbuf *m = 0;
786 struct ifnet *ifp = ifa->ifa_ifp;
787
788 if (route_cb.any_count == 0)
789 return;
790 for (pass = 1; pass < 3; pass++) {
791 bzero((caddr_t)&info, sizeof(info));
792 if ((cmd == RTM_ADD && pass == 1) ||
793 (cmd == RTM_DELETE && pass == 2)) {
794 register struct ifa_msghdr *ifam;
795 int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
796
797 ifaaddr = sa = ifa->ifa_addr;
798 ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
799 netmask = ifa->ifa_netmask;
800 brdaddr = ifa->ifa_dstaddr;
801 if ((m = rt_msg1(ncmd, &info)) == NULL)
802 continue;
803 ifam = mtod(m, struct ifa_msghdr *);
804 ifam->ifam_index = ifp->if_index;
805 ifam->ifam_metric = ifa->ifa_metric;
806 ifam->ifam_flags = ifa->ifa_flags;
807 ifam->ifam_addrs = info.rti_addrs;
808 }
809 if ((cmd == RTM_ADD && pass == 2) ||
810 (cmd == RTM_DELETE && pass == 1)) {
811 register struct rt_msghdr *rtm;
812
813 if (rt == 0)
814 continue;
815 netmask = rt_mask(rt);
816 dst = sa = rt_key(rt);
817 gate = rt->rt_gateway;
818 if ((m = rt_msg1(cmd, &info)) == NULL)
819 continue;
820 rtm = mtod(m, struct rt_msghdr *);
821 rtm->rtm_index = ifp->if_index;
822 rtm->rtm_flags |= rt->rt_flags;
823 rtm->rtm_errno = error;
824 rtm->rtm_addrs = info.rti_addrs;
825 }
826 route_proto.sp_protocol = sa ? sa->sa_family : 0;
827 raw_input(m, &route_proto, &route_src, &route_dst);
828 }
829}
830
831/*
832 * This is the analogue to the rt_newaddrmsg which performs the same
833 * function but for multicast group memberhips. This is easier since
834 * there is no route state to worry about.
835 */
836void
837rt_newmaddrmsg(cmd, ifma)
838 int cmd;
839 struct ifmultiaddr *ifma;
840{
841 struct rt_addrinfo info;
842 struct mbuf *m = 0;
843 struct ifnet *ifp = ifma->ifma_ifp;
844 struct ifma_msghdr *ifmam;
845
846 if (route_cb.any_count == 0)
847 return;
848
849 bzero((caddr_t)&info, sizeof(info));
850 ifaaddr = ifma->ifma_addr;
851 if (ifp && TAILQ_FIRST(&ifp->if_addrhead))
852 ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
853 else
854 ifpaddr = NULL;
855 /*
856 * If a link-layer address is present, present it as a ``gateway''
857 * (similarly to how ARP entries, e.g., are presented).
858 */
859 gate = ifma->ifma_lladdr;
860 if ((m = rt_msg1(cmd, &info)) == NULL)
861 return;
862 ifmam = mtod(m, struct ifma_msghdr *);
863 ifmam->ifmam_index = ifp->if_index;
864 ifmam->ifmam_addrs = info.rti_addrs;
865 route_proto.sp_protocol = ifma->ifma_addr->sa_family;
866 raw_input(m, &route_proto, &route_src, &route_dst);
867}
868
869/*
870 * This is called to generate routing socket messages indicating
871 * network interface arrival and departure.
872 */
873void
874rt_ifannouncemsg(ifp, what)
875 struct ifnet *ifp;
876 int what;
877{
878 struct if_announcemsghdr *ifan;
879 struct mbuf *m;
880 struct rt_addrinfo info;
881
882 if (route_cb.any_count == 0)
883 return;
884 bzero((caddr_t)&info, sizeof(info));
885 m = rt_msg1(RTM_IFANNOUNCE, &info);
886 if (m == NULL)
887 return;
888 ifan = mtod(m, struct if_announcemsghdr *);
889 ifan->ifan_index = ifp->if_index;
890 snprintf(ifan->ifan_name, sizeof(ifan->ifan_name),
891 "%s%d", ifp->if_name, ifp->if_unit);
892 ifan->ifan_what = what;
893 route_proto.sp_protocol = 0;
894 raw_input(m, &route_proto, &route_src, &route_dst);
895 }
896
897/*
898 * This is used in dumping the kernel table via sysctl().
899 */
900int
901sysctl_dumpentry(rn, vw)
902 struct radix_node *rn;
903 void *vw;
904{
905 register struct walkarg *w = vw;
906 register struct rtentry *rt = (struct rtentry *)rn;
907 int error = 0, size;
908 struct rt_addrinfo info;
909
910 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
911 return 0;
912 bzero((caddr_t)&info, sizeof(info));
913 dst = rt_key(rt);
914 gate = rt->rt_gateway;
915 netmask = rt_mask(rt);
916 genmask = rt->rt_genmask;
917 if (rt->rt_ifp) {
918 ifpaddr = TAILQ_FIRST(&rt->rt_ifp->if_addrhead)->ifa_addr;
919 ifaaddr = rt->rt_ifa->ifa_addr;
920 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
921 brdaddr = rt->rt_ifa->ifa_dstaddr;
922 }
923 size = rt_msg2(RTM_GET, &info, 0, w);
924 if (w->w_req && w->w_tmem) {
925 register struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
926
927 rtm->rtm_flags = rt->rt_flags;
928 rtm->rtm_use = rt->rt_use;
929 rtm->rtm_rmx = rt->rt_rmx;
930 rtm->rtm_index = rt->rt_ifp->if_index;
931 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
932 rtm->rtm_addrs = info.rti_addrs;
933 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
934 return (error);
935 }
936 return (error);
937}
938
939int
940sysctl_iflist(af, w)
941 int af;
942 register struct walkarg *w;
943{
944 register struct ifnet *ifp;
945 register struct ifaddr *ifa;
946 struct rt_addrinfo info;
947 int len, error = 0;
948
949 bzero((caddr_t)&info, sizeof(info));
950 TAILQ_FOREACH(ifp, &ifnet, if_link) {
951 if (w->w_arg && w->w_arg != ifp->if_index)
952 continue;
953 ifa = TAILQ_FIRST(&ifp->if_addrhead);
954 ifpaddr = ifa->ifa_addr;
955 len = rt_msg2(RTM_IFINFO, &info, (caddr_t)0, w);
956 ifpaddr = 0;
957 if (w->w_req && w->w_tmem) {
958 register struct if_msghdr *ifm;
959
960 ifm = (struct if_msghdr *)w->w_tmem;
961 ifm->ifm_index = ifp->if_index;
962 ifm->ifm_flags = (u_short)ifp->if_flags;
963 ifm->ifm_data = ifp->if_data;
964 ifm->ifm_addrs = info.rti_addrs;
965 error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
966 if (error)
967 goto done;
968 }
969 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != 0) {
970 if (af && af != ifa->ifa_addr->sa_family)
971 continue;
972 if (jailed(curthread->td_ucred) &&
973 prison_if(curthread->td_ucred, ifa->ifa_addr))
974 continue;
975 ifaaddr = ifa->ifa_addr;
976 netmask = ifa->ifa_netmask;
977 brdaddr = ifa->ifa_dstaddr;
978 len = rt_msg2(RTM_NEWADDR, &info, 0, w);
979 if (w->w_req && w->w_tmem) {
980 register struct ifa_msghdr *ifam;
981
982 ifam = (struct ifa_msghdr *)w->w_tmem;
983 ifam->ifam_index = ifa->ifa_ifp->if_index;
984 ifam->ifam_flags = ifa->ifa_flags;
985 ifam->ifam_metric = ifa->ifa_metric;
986 ifam->ifam_addrs = info.rti_addrs;
987 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
988 if (error)
989 goto done;
990 }
991 }
992 ifaaddr = netmask = brdaddr = 0;
993 }
994done:
995 return (error);
996}
997
998static int
999sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1000{
1001 int *name = (int *)arg1;
1002 u_int namelen = arg2;
1003 register struct radix_node_head *rnh;
1004 int i, s, error = EINVAL;
1005 u_char af;
1006 struct walkarg w;
1007
1008 name ++;
1009 namelen--;
1010 if (req->newptr)
1011 return (EPERM);
1012 if (namelen != 3)
1013 return ((namelen < 3) ? EISDIR : ENOTDIR);
1014 af = name[0];
1015 Bzero(&w, sizeof(w));
1016 w.w_op = name[1];
1017 w.w_arg = name[2];
1018 w.w_req = req;
1019
1020 s = splnet();
1021 switch (w.w_op) {
1022
1023 case NET_RT_DUMP:
1024 case NET_RT_FLAGS:
1025 for (i = 1; i <= AF_MAX; i++)
1026 if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
1027 (error = rnh->rnh_walktree(rnh,
1028 sysctl_dumpentry, &w)))
1029 break;
1030 break;
1031
1032 case NET_RT_IFLIST:
1033 error = sysctl_iflist(af, &w);
1034 }
1035 splx(s);
1036 if (w.w_tmem)
1037 free(w.w_tmem, M_RTABLE);
1038 return (error);
1039}
1040
1041SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1042
1043/*
1044 * Definitions of protocols supported in the ROUTE domain.
1045 */
1046
1047extern struct domain routedomain; /* or at least forward */
1048
1049static struct protosw routesw[] = {
1050{ SOCK_RAW, &routedomain, 0, PR_ATOMIC|PR_ADDR,
1051 0, route_output, raw_ctlinput, 0,
1052 0,
1053 raw_init, 0, 0, 0,
1054 &route_usrreqs
1055}
1056};
1057
1058static struct domain routedomain =
1059 { PF_ROUTE, "route", 0, 0, 0,
1060 routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] };
1061
1062DOMAIN_SET(route);
| 57
58MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
59
60static struct sockaddr route_dst = { 2, PF_ROUTE, };
61static struct sockaddr route_src = { 2, PF_ROUTE, };
62static struct sockaddr sa_zero = { sizeof(sa_zero), AF_INET, };
63static struct sockproto route_proto = { PF_ROUTE, };
64
65struct walkarg {
66 int w_tmemsize;
67 int w_op, w_arg;
68 caddr_t w_tmem;
69 struct sysctl_req *w_req;
70};
71
72static struct mbuf *
73 rt_msg1(int, struct rt_addrinfo *);
74static int rt_msg2(int, struct rt_addrinfo *, caddr_t, struct walkarg *);
75static int rt_xaddrs(caddr_t, caddr_t, struct rt_addrinfo *);
76static int sysctl_dumpentry(struct radix_node *rn, void *vw);
77static int sysctl_iflist(int af, struct walkarg *w);
78static int route_output(struct mbuf *, struct socket *);
79static void rt_setmetrics(u_long, struct rt_metrics *, struct rt_metrics *);
80
81/* Sleazy use of local variables throughout file, warning!!!! */
82#define dst info.rti_info[RTAX_DST]
83#define gate info.rti_info[RTAX_GATEWAY]
84#define netmask info.rti_info[RTAX_NETMASK]
85#define genmask info.rti_info[RTAX_GENMASK]
86#define ifpaddr info.rti_info[RTAX_IFP]
87#define ifaaddr info.rti_info[RTAX_IFA]
88#define brdaddr info.rti_info[RTAX_BRD]
89
90/*
91 * It really doesn't make any sense at all for this code to share much
92 * with raw_usrreq.c, since its functionality is so restricted. XXX
93 */
94static int
95rts_abort(struct socket *so)
96{
97 int s, error;
98 s = splnet();
99 error = raw_usrreqs.pru_abort(so);
100 splx(s);
101 return error;
102}
103
104/* pru_accept is EOPNOTSUPP */
105
106static int
107rts_attach(struct socket *so, int proto, struct thread *td)
108{
109 struct rawcb *rp;
110 int s, error;
111
112 if (sotorawcb(so) != 0)
113 return EISCONN; /* XXX panic? */
114 /* XXX */
115 MALLOC(rp, struct rawcb *, sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
116 if (rp == 0)
117 return ENOBUFS;
118
119 /*
120 * The splnet() is necessary to block protocols from sending
121 * error notifications (like RTM_REDIRECT or RTM_LOSING) while
122 * this PCB is extant but incompletely initialized.
123 * Probably we should try to do more of this work beforehand and
124 * eliminate the spl.
125 */
126 s = splnet();
127 so->so_pcb = (caddr_t)rp;
128 error = raw_attach(so, proto);
129 rp = sotorawcb(so);
130 if (error) {
131 splx(s);
132 so->so_pcb = NULL;
133 free(rp, M_PCB);
134 return error;
135 }
136 switch(rp->rcb_proto.sp_protocol) {
137 case AF_INET:
138 route_cb.ip_count++;
139 break;
140 case AF_INET6:
141 route_cb.ip6_count++;
142 break;
143 case AF_IPX:
144 route_cb.ipx_count++;
145 break;
146 case AF_NS:
147 route_cb.ns_count++;
148 break;
149 }
150 rp->rcb_faddr = &route_src;
151 route_cb.any_count++;
152 SIGIO_SLOCK();
153 soisconnected_locked(so);
154 so->so_options |= SO_USELOOPBACK;
155 SIGIO_SUNLOCK();
156 splx(s);
157 return 0;
158}
159
160static int
161rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
162{
163 int s, error;
164 s = splnet();
165 error = raw_usrreqs.pru_bind(so, nam, td); /* xxx just EINVAL */
166 splx(s);
167 return error;
168}
169
170static int
171rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
172{
173 int s, error;
174 s = splnet();
175 error = raw_usrreqs.pru_connect(so, nam, td); /* XXX just EINVAL */
176 splx(s);
177 return error;
178}
179
180/* pru_connect2 is EOPNOTSUPP */
181/* pru_control is EOPNOTSUPP */
182
183static int
184rts_detach(struct socket *so)
185{
186 struct rawcb *rp = sotorawcb(so);
187 int s, error;
188
189 s = splnet();
190 if (rp != 0) {
191 switch(rp->rcb_proto.sp_protocol) {
192 case AF_INET:
193 route_cb.ip_count--;
194 break;
195 case AF_INET6:
196 route_cb.ip6_count--;
197 break;
198 case AF_IPX:
199 route_cb.ipx_count--;
200 break;
201 case AF_NS:
202 route_cb.ns_count--;
203 break;
204 }
205 route_cb.any_count--;
206 }
207 error = raw_usrreqs.pru_detach(so);
208 splx(s);
209 return error;
210}
211
212static int
213rts_disconnect(struct socket *so)
214{
215 int s, error;
216 s = splnet();
217 error = raw_usrreqs.pru_disconnect(so);
218 splx(s);
219 return error;
220}
221
222/* pru_listen is EOPNOTSUPP */
223
224static int
225rts_peeraddr(struct socket *so, struct sockaddr **nam)
226{
227 int s, error;
228 s = splnet();
229 error = raw_usrreqs.pru_peeraddr(so, nam);
230 splx(s);
231 return error;
232}
233
234/* pru_rcvd is EOPNOTSUPP */
235/* pru_rcvoob is EOPNOTSUPP */
236
237static int
238rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
239 struct mbuf *control, struct thread *td)
240{
241 int s, error;
242 s = splnet();
243 error = raw_usrreqs.pru_send(so, flags, m, nam, control, td);
244 splx(s);
245 return error;
246}
247
248/* pru_sense is null */
249
250static int
251rts_shutdown(struct socket *so)
252{
253 int s, error;
254 s = splnet();
255 error = raw_usrreqs.pru_shutdown(so);
256 splx(s);
257 return error;
258}
259
260static int
261rts_sockaddr(struct socket *so, struct sockaddr **nam)
262{
263 int s, error;
264 s = splnet();
265 error = raw_usrreqs.pru_sockaddr(so, nam);
266 splx(s);
267 return error;
268}
269
270static struct pr_usrreqs route_usrreqs = {
271 rts_abort, pru_accept_notsupp, rts_attach, rts_bind, rts_connect,
272 pru_connect2_notsupp, pru_control_notsupp, rts_detach, rts_disconnect,
273 pru_listen_notsupp, rts_peeraddr, pru_rcvd_notsupp, pru_rcvoob_notsupp,
274 rts_send, pru_sense_null, rts_shutdown, rts_sockaddr,
275 sosend, soreceive, sopoll
276};
277
278/*ARGSUSED*/
279static int
280route_output(m, so)
281 register struct mbuf *m;
282 struct socket *so;
283{
284 register struct rt_msghdr *rtm = 0;
285 register struct rtentry *rt = 0;
286 struct rtentry *saved_nrt = 0;
287 struct radix_node_head *rnh;
288 struct rt_addrinfo info;
289 int len, error = 0;
290 struct ifnet *ifp = 0;
291 struct ifaddr *ifa = 0;
292
293#define senderr(e) { error = e; goto flush;}
294 if (m == 0 || ((m->m_len < sizeof(long)) &&
295 (m = m_pullup(m, sizeof(long))) == 0))
296 return (ENOBUFS);
297 if ((m->m_flags & M_PKTHDR) == 0)
298 panic("route_output");
299 len = m->m_pkthdr.len;
300 if (len < sizeof(*rtm) ||
301 len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
302 dst = 0;
303 senderr(EINVAL);
304 }
305 R_Malloc(rtm, struct rt_msghdr *, len);
306 if (rtm == 0) {
307 dst = 0;
308 senderr(ENOBUFS);
309 }
310 m_copydata(m, 0, len, (caddr_t)rtm);
311 if (rtm->rtm_version != RTM_VERSION) {
312 dst = 0;
313 senderr(EPROTONOSUPPORT);
314 }
315 rtm->rtm_pid = curproc->p_pid;
316 bzero(&info, sizeof(info));
317 info.rti_addrs = rtm->rtm_addrs;
318 if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
319 dst = 0;
320 senderr(EINVAL);
321 }
322 info.rti_flags = rtm->rtm_flags;
323 if (dst == 0 || (dst->sa_family >= AF_MAX)
324 || (gate != 0 && (gate->sa_family >= AF_MAX)))
325 senderr(EINVAL);
326 if (genmask) {
327 struct radix_node *t;
328 t = rn_addmask((caddr_t)genmask, 0, 1);
329 if (t && Bcmp((caddr_t *)genmask + 1, (caddr_t *)t->rn_key + 1,
330 *(u_char *)t->rn_key - 1) == 0)
331 genmask = (struct sockaddr *)(t->rn_key);
332 else
333 senderr(ENOBUFS);
334 }
335
336 /*
337 * Verify that the caller has the appropriate privilege; RTM_GET
338 * is the only operation the non-superuser is allowed.
339 */
340 if (rtm->rtm_type != RTM_GET && (error = suser(curthread)) != 0)
341 senderr(error);
342
343 switch (rtm->rtm_type) {
344
345 case RTM_ADD:
346 if (gate == 0)
347 senderr(EINVAL);
348 error = rtrequest1(RTM_ADD, &info, &saved_nrt);
349 if (error == 0 && saved_nrt) {
350 rt_setmetrics(rtm->rtm_inits,
351 &rtm->rtm_rmx, &saved_nrt->rt_rmx);
352 saved_nrt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
353 saved_nrt->rt_rmx.rmx_locks |=
354 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
355 saved_nrt->rt_refcnt--;
356 saved_nrt->rt_genmask = genmask;
357 }
358 break;
359
360 case RTM_DELETE:
361 error = rtrequest1(RTM_DELETE, &info, &saved_nrt);
362 if (error == 0) {
363 if ((rt = saved_nrt))
364 rt->rt_refcnt++;
365 goto report;
366 }
367 break;
368
369 case RTM_GET:
370 case RTM_CHANGE:
371 case RTM_LOCK:
372 if ((rnh = rt_tables[dst->sa_family]) == 0) {
373 senderr(EAFNOSUPPORT);
374 } else if ((rt = (struct rtentry *)
375 rnh->rnh_lookup(dst, netmask, rnh)) != NULL)
376 rt->rt_refcnt++;
377 else
378 senderr(ESRCH);
379 switch(rtm->rtm_type) {
380
381 case RTM_GET:
382 report:
383 dst = rt_key(rt);
384 gate = rt->rt_gateway;
385 netmask = rt_mask(rt);
386 genmask = rt->rt_genmask;
387 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
388 ifp = rt->rt_ifp;
389 if (ifp) {
390 ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
391 ifaaddr = rt->rt_ifa->ifa_addr;
392 if (ifp->if_flags & IFF_POINTOPOINT)
393 brdaddr = rt->rt_ifa->ifa_dstaddr;
394 rtm->rtm_index = ifp->if_index;
395 } else {
396 ifpaddr = 0;
397 ifaaddr = 0;
398 }
399 }
400 len = rt_msg2(rtm->rtm_type, &info, (caddr_t)0,
401 (struct walkarg *)0);
402 if (len > rtm->rtm_msglen) {
403 struct rt_msghdr *new_rtm;
404 R_Malloc(new_rtm, struct rt_msghdr *, len);
405 if (new_rtm == 0)
406 senderr(ENOBUFS);
407 Bcopy(rtm, new_rtm, rtm->rtm_msglen);
408 Free(rtm); rtm = new_rtm;
409 }
410 (void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm,
411 (struct walkarg *)0);
412 rtm->rtm_flags = rt->rt_flags;
413 rtm->rtm_rmx = rt->rt_rmx;
414 rtm->rtm_addrs = info.rti_addrs;
415 break;
416
417 case RTM_CHANGE:
418 /* new gateway could require new ifaddr, ifp;
419 flags may also be different; ifp may be specified
420 by ll sockaddr when protocol address is ambiguous */
421#define equal(a1, a2) (bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0)
422 if ((rt->rt_flags & RTF_GATEWAY && gate != NULL) ||
423 ifpaddr != NULL ||
424 (ifaaddr != NULL &&
425 !equal(ifaaddr, rt->rt_ifa->ifa_addr))) {
426 if ((error = rt_getifa(&info)) != 0)
427 senderr(error);
428 }
429 if (gate != NULL &&
430 (error = rt_setgate(rt, rt_key(rt), gate)) != 0)
431 senderr(error);
432 if ((ifa = info.rti_ifa) != NULL) {
433 register struct ifaddr *oifa = rt->rt_ifa;
434 if (oifa != ifa) {
435 if (oifa && oifa->ifa_rtrequest)
436 oifa->ifa_rtrequest(RTM_DELETE, rt,
437 &info);
438 IFAFREE(rt->rt_ifa);
439 rt->rt_ifa = ifa;
440 ifa->ifa_refcnt++;
441 rt->rt_ifp = info.rti_ifp;
442 }
443 }
444 rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
445 &rt->rt_rmx);
446 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
447 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
448 if (genmask)
449 rt->rt_genmask = genmask;
450 /*
451 * Fall into
452 */
453 case RTM_LOCK:
454 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
455 rt->rt_rmx.rmx_locks |=
456 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
457 break;
458 }
459 break;
460
461 default:
462 senderr(EOPNOTSUPP);
463 }
464
465flush:
466 if (rtm) {
467 if (error)
468 rtm->rtm_errno = error;
469 else
470 rtm->rtm_flags |= RTF_DONE;
471 }
472 if (rt)
473 rtfree(rt);
474 {
475 register struct rawcb *rp = 0;
476 /*
477 * Check to see if we don't want our own messages.
478 */
479 if ((so->so_options & SO_USELOOPBACK) == 0) {
480 if (route_cb.any_count <= 1) {
481 if (rtm)
482 Free(rtm);
483 m_freem(m);
484 return (error);
485 }
486 /* There is another listener, so construct message */
487 rp = sotorawcb(so);
488 }
489 if (rtm) {
490 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
491 if (m->m_pkthdr.len < rtm->rtm_msglen) {
492 m_freem(m);
493 m = NULL;
494 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
495 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
496 Free(rtm);
497 }
498 if (rp)
499 rp->rcb_proto.sp_family = 0; /* Avoid us */
500 if (dst)
501 route_proto.sp_protocol = dst->sa_family;
502 if (m)
503 raw_input(m, &route_proto, &route_src, &route_dst);
504 if (rp)
505 rp->rcb_proto.sp_family = PF_ROUTE;
506 }
507 return (error);
508}
509
510static void
511rt_setmetrics(which, in, out)
512 u_long which;
513 register struct rt_metrics *in, *out;
514{
515#define metric(f, e) if (which & (f)) out->e = in->e;
516 metric(RTV_RPIPE, rmx_recvpipe);
517 metric(RTV_SPIPE, rmx_sendpipe);
518 metric(RTV_SSTHRESH, rmx_ssthresh);
519 metric(RTV_RTT, rmx_rtt);
520 metric(RTV_RTTVAR, rmx_rttvar);
521 metric(RTV_HOPCOUNT, rmx_hopcount);
522 metric(RTV_MTU, rmx_mtu);
523 metric(RTV_EXPIRE, rmx_expire);
524#undef metric
525}
526
527#define ROUNDUP(a) \
528 ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
529#define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
530
531
532/*
533 * Extract the addresses of the passed sockaddrs.
534 * Do a little sanity checking so as to avoid bad memory references.
535 * This data is derived straight from userland.
536 */
537static int
538rt_xaddrs(cp, cplim, rtinfo)
539 register caddr_t cp, cplim;
540 register struct rt_addrinfo *rtinfo;
541{
542 register struct sockaddr *sa;
543 register int i;
544
545 for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
546 if ((rtinfo->rti_addrs & (1 << i)) == 0)
547 continue;
548 sa = (struct sockaddr *)cp;
549 /*
550 * It won't fit.
551 */
552 if ( (cp + sa->sa_len) > cplim ) {
553 return (EINVAL);
554 }
555
556 /*
557 * there are no more.. quit now
558 * If there are more bits, they are in error.
559 * I've seen this. route(1) can evidently generate these.
560 * This causes kernel to core dump.
561 * for compatibility, If we see this, point to a safe address.
562 */
563 if (sa->sa_len == 0) {
564 rtinfo->rti_info[i] = &sa_zero;
565 return (0); /* should be EINVAL but for compat */
566 }
567
568 /* accept it */
569 rtinfo->rti_info[i] = sa;
570 ADVANCE(cp, sa);
571 }
572 return (0);
573}
574
575static struct mbuf *
576rt_msg1(type, rtinfo)
577 int type;
578 register struct rt_addrinfo *rtinfo;
579{
580 register struct rt_msghdr *rtm;
581 register struct mbuf *m;
582 register int i;
583 register struct sockaddr *sa;
584 int len, dlen;
585
586 switch (type) {
587
588 case RTM_DELADDR:
589 case RTM_NEWADDR:
590 len = sizeof(struct ifa_msghdr);
591 break;
592
593 case RTM_DELMADDR:
594 case RTM_NEWMADDR:
595 len = sizeof(struct ifma_msghdr);
596 break;
597
598 case RTM_IFINFO:
599 len = sizeof(struct if_msghdr);
600 break;
601
602 case RTM_IFANNOUNCE:
603 len = sizeof(struct if_announcemsghdr);
604 break;
605
606 default:
607 len = sizeof(struct rt_msghdr);
608 }
609 if (len > MCLBYTES)
610 panic("rt_msg1");
611 m = m_gethdr(M_DONTWAIT, MT_DATA);
612 if (m && len > MHLEN) {
613 MCLGET(m, M_DONTWAIT);
614 if ((m->m_flags & M_EXT) == 0) {
615 m_free(m);
616 m = NULL;
617 }
618 }
619 if (m == 0)
620 return (m);
621 m->m_pkthdr.len = m->m_len = len;
622 m->m_pkthdr.rcvif = 0;
623 rtm = mtod(m, struct rt_msghdr *);
624 bzero((caddr_t)rtm, len);
625 for (i = 0; i < RTAX_MAX; i++) {
626 if ((sa = rtinfo->rti_info[i]) == NULL)
627 continue;
628 rtinfo->rti_addrs |= (1 << i);
629 dlen = ROUNDUP(sa->sa_len);
630 m_copyback(m, len, dlen, (caddr_t)sa);
631 len += dlen;
632 }
633 if (m->m_pkthdr.len != len) {
634 m_freem(m);
635 return (NULL);
636 }
637 rtm->rtm_msglen = len;
638 rtm->rtm_version = RTM_VERSION;
639 rtm->rtm_type = type;
640 return (m);
641}
642
643static int
644rt_msg2(type, rtinfo, cp, w)
645 int type;
646 register struct rt_addrinfo *rtinfo;
647 caddr_t cp;
648 struct walkarg *w;
649{
650 register int i;
651 int len, dlen, second_time = 0;
652 caddr_t cp0;
653
654 rtinfo->rti_addrs = 0;
655again:
656 switch (type) {
657
658 case RTM_DELADDR:
659 case RTM_NEWADDR:
660 len = sizeof(struct ifa_msghdr);
661 break;
662
663 case RTM_IFINFO:
664 len = sizeof(struct if_msghdr);
665 break;
666
667 default:
668 len = sizeof(struct rt_msghdr);
669 }
670 cp0 = cp;
671 if (cp0)
672 cp += len;
673 for (i = 0; i < RTAX_MAX; i++) {
674 register struct sockaddr *sa;
675
676 if ((sa = rtinfo->rti_info[i]) == 0)
677 continue;
678 rtinfo->rti_addrs |= (1 << i);
679 dlen = ROUNDUP(sa->sa_len);
680 if (cp) {
681 bcopy((caddr_t)sa, cp, (unsigned)dlen);
682 cp += dlen;
683 }
684 len += dlen;
685 }
686 len = ALIGN(len);
687 if (cp == 0 && w != NULL && !second_time) {
688 register struct walkarg *rw = w;
689
690 if (rw->w_req) {
691 if (rw->w_tmemsize < len) {
692 if (rw->w_tmem)
693 free(rw->w_tmem, M_RTABLE);
694 rw->w_tmem = (caddr_t)
695 malloc(len, M_RTABLE, M_NOWAIT);
696 if (rw->w_tmem)
697 rw->w_tmemsize = len;
698 }
699 if (rw->w_tmem) {
700 cp = rw->w_tmem;
701 second_time = 1;
702 goto again;
703 }
704 }
705 }
706 if (cp) {
707 register struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
708
709 rtm->rtm_version = RTM_VERSION;
710 rtm->rtm_type = type;
711 rtm->rtm_msglen = len;
712 }
713 return (len);
714}
715
716/*
717 * This routine is called to generate a message from the routing
718 * socket indicating that a redirect has occured, a routing lookup
719 * has failed, or that a protocol has detected timeouts to a particular
720 * destination.
721 */
722void
723rt_missmsg(type, rtinfo, flags, error)
724 int type, flags, error;
725 register struct rt_addrinfo *rtinfo;
726{
727 register struct rt_msghdr *rtm;
728 register struct mbuf *m;
729 struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
730
731 if (route_cb.any_count == 0)
732 return;
733 m = rt_msg1(type, rtinfo);
734 if (m == 0)
735 return;
736 rtm = mtod(m, struct rt_msghdr *);
737 rtm->rtm_flags = RTF_DONE | flags;
738 rtm->rtm_errno = error;
739 rtm->rtm_addrs = rtinfo->rti_addrs;
740 route_proto.sp_protocol = sa ? sa->sa_family : 0;
741 raw_input(m, &route_proto, &route_src, &route_dst);
742}
743
744/*
745 * This routine is called to generate a message from the routing
746 * socket indicating that the status of a network interface has changed.
747 */
748void
749rt_ifmsg(ifp)
750 register struct ifnet *ifp;
751{
752 register struct if_msghdr *ifm;
753 struct mbuf *m;
754 struct rt_addrinfo info;
755
756 if (route_cb.any_count == 0)
757 return;
758 bzero((caddr_t)&info, sizeof(info));
759 m = rt_msg1(RTM_IFINFO, &info);
760 if (m == 0)
761 return;
762 ifm = mtod(m, struct if_msghdr *);
763 ifm->ifm_index = ifp->if_index;
764 ifm->ifm_flags = (u_short)ifp->if_flags;
765 ifm->ifm_data = ifp->if_data;
766 ifm->ifm_addrs = 0;
767 route_proto.sp_protocol = 0;
768 raw_input(m, &route_proto, &route_src, &route_dst);
769}
770
771/*
772 * This is called to generate messages from the routing socket
773 * indicating a network interface has had addresses associated with it.
774 * if we ever reverse the logic and replace messages TO the routing
775 * socket indicate a request to configure interfaces, then it will
776 * be unnecessary as the routing socket will automatically generate
777 * copies of it.
778 */
779void
780rt_newaddrmsg(cmd, ifa, error, rt)
781 int cmd, error;
782 register struct ifaddr *ifa;
783 register struct rtentry *rt;
784{
785 struct rt_addrinfo info;
786 struct sockaddr *sa = 0;
787 int pass;
788 struct mbuf *m = 0;
789 struct ifnet *ifp = ifa->ifa_ifp;
790
791 if (route_cb.any_count == 0)
792 return;
793 for (pass = 1; pass < 3; pass++) {
794 bzero((caddr_t)&info, sizeof(info));
795 if ((cmd == RTM_ADD && pass == 1) ||
796 (cmd == RTM_DELETE && pass == 2)) {
797 register struct ifa_msghdr *ifam;
798 int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
799
800 ifaaddr = sa = ifa->ifa_addr;
801 ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
802 netmask = ifa->ifa_netmask;
803 brdaddr = ifa->ifa_dstaddr;
804 if ((m = rt_msg1(ncmd, &info)) == NULL)
805 continue;
806 ifam = mtod(m, struct ifa_msghdr *);
807 ifam->ifam_index = ifp->if_index;
808 ifam->ifam_metric = ifa->ifa_metric;
809 ifam->ifam_flags = ifa->ifa_flags;
810 ifam->ifam_addrs = info.rti_addrs;
811 }
812 if ((cmd == RTM_ADD && pass == 2) ||
813 (cmd == RTM_DELETE && pass == 1)) {
814 register struct rt_msghdr *rtm;
815
816 if (rt == 0)
817 continue;
818 netmask = rt_mask(rt);
819 dst = sa = rt_key(rt);
820 gate = rt->rt_gateway;
821 if ((m = rt_msg1(cmd, &info)) == NULL)
822 continue;
823 rtm = mtod(m, struct rt_msghdr *);
824 rtm->rtm_index = ifp->if_index;
825 rtm->rtm_flags |= rt->rt_flags;
826 rtm->rtm_errno = error;
827 rtm->rtm_addrs = info.rti_addrs;
828 }
829 route_proto.sp_protocol = sa ? sa->sa_family : 0;
830 raw_input(m, &route_proto, &route_src, &route_dst);
831 }
832}
833
834/*
835 * This is the analogue to the rt_newaddrmsg which performs the same
836 * function but for multicast group memberhips. This is easier since
837 * there is no route state to worry about.
838 */
839void
840rt_newmaddrmsg(cmd, ifma)
841 int cmd;
842 struct ifmultiaddr *ifma;
843{
844 struct rt_addrinfo info;
845 struct mbuf *m = 0;
846 struct ifnet *ifp = ifma->ifma_ifp;
847 struct ifma_msghdr *ifmam;
848
849 if (route_cb.any_count == 0)
850 return;
851
852 bzero((caddr_t)&info, sizeof(info));
853 ifaaddr = ifma->ifma_addr;
854 if (ifp && TAILQ_FIRST(&ifp->if_addrhead))
855 ifpaddr = TAILQ_FIRST(&ifp->if_addrhead)->ifa_addr;
856 else
857 ifpaddr = NULL;
858 /*
859 * If a link-layer address is present, present it as a ``gateway''
860 * (similarly to how ARP entries, e.g., are presented).
861 */
862 gate = ifma->ifma_lladdr;
863 if ((m = rt_msg1(cmd, &info)) == NULL)
864 return;
865 ifmam = mtod(m, struct ifma_msghdr *);
866 ifmam->ifmam_index = ifp->if_index;
867 ifmam->ifmam_addrs = info.rti_addrs;
868 route_proto.sp_protocol = ifma->ifma_addr->sa_family;
869 raw_input(m, &route_proto, &route_src, &route_dst);
870}
871
872/*
873 * This is called to generate routing socket messages indicating
874 * network interface arrival and departure.
875 */
876void
877rt_ifannouncemsg(ifp, what)
878 struct ifnet *ifp;
879 int what;
880{
881 struct if_announcemsghdr *ifan;
882 struct mbuf *m;
883 struct rt_addrinfo info;
884
885 if (route_cb.any_count == 0)
886 return;
887 bzero((caddr_t)&info, sizeof(info));
888 m = rt_msg1(RTM_IFANNOUNCE, &info);
889 if (m == NULL)
890 return;
891 ifan = mtod(m, struct if_announcemsghdr *);
892 ifan->ifan_index = ifp->if_index;
893 snprintf(ifan->ifan_name, sizeof(ifan->ifan_name),
894 "%s%d", ifp->if_name, ifp->if_unit);
895 ifan->ifan_what = what;
896 route_proto.sp_protocol = 0;
897 raw_input(m, &route_proto, &route_src, &route_dst);
898 }
899
900/*
901 * This is used in dumping the kernel table via sysctl().
902 */
903int
904sysctl_dumpentry(rn, vw)
905 struct radix_node *rn;
906 void *vw;
907{
908 register struct walkarg *w = vw;
909 register struct rtentry *rt = (struct rtentry *)rn;
910 int error = 0, size;
911 struct rt_addrinfo info;
912
913 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
914 return 0;
915 bzero((caddr_t)&info, sizeof(info));
916 dst = rt_key(rt);
917 gate = rt->rt_gateway;
918 netmask = rt_mask(rt);
919 genmask = rt->rt_genmask;
920 if (rt->rt_ifp) {
921 ifpaddr = TAILQ_FIRST(&rt->rt_ifp->if_addrhead)->ifa_addr;
922 ifaaddr = rt->rt_ifa->ifa_addr;
923 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
924 brdaddr = rt->rt_ifa->ifa_dstaddr;
925 }
926 size = rt_msg2(RTM_GET, &info, 0, w);
927 if (w->w_req && w->w_tmem) {
928 register struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
929
930 rtm->rtm_flags = rt->rt_flags;
931 rtm->rtm_use = rt->rt_use;
932 rtm->rtm_rmx = rt->rt_rmx;
933 rtm->rtm_index = rt->rt_ifp->if_index;
934 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
935 rtm->rtm_addrs = info.rti_addrs;
936 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
937 return (error);
938 }
939 return (error);
940}
941
942int
943sysctl_iflist(af, w)
944 int af;
945 register struct walkarg *w;
946{
947 register struct ifnet *ifp;
948 register struct ifaddr *ifa;
949 struct rt_addrinfo info;
950 int len, error = 0;
951
952 bzero((caddr_t)&info, sizeof(info));
953 TAILQ_FOREACH(ifp, &ifnet, if_link) {
954 if (w->w_arg && w->w_arg != ifp->if_index)
955 continue;
956 ifa = TAILQ_FIRST(&ifp->if_addrhead);
957 ifpaddr = ifa->ifa_addr;
958 len = rt_msg2(RTM_IFINFO, &info, (caddr_t)0, w);
959 ifpaddr = 0;
960 if (w->w_req && w->w_tmem) {
961 register struct if_msghdr *ifm;
962
963 ifm = (struct if_msghdr *)w->w_tmem;
964 ifm->ifm_index = ifp->if_index;
965 ifm->ifm_flags = (u_short)ifp->if_flags;
966 ifm->ifm_data = ifp->if_data;
967 ifm->ifm_addrs = info.rti_addrs;
968 error = SYSCTL_OUT(w->w_req,(caddr_t)ifm, len);
969 if (error)
970 goto done;
971 }
972 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != 0) {
973 if (af && af != ifa->ifa_addr->sa_family)
974 continue;
975 if (jailed(curthread->td_ucred) &&
976 prison_if(curthread->td_ucred, ifa->ifa_addr))
977 continue;
978 ifaaddr = ifa->ifa_addr;
979 netmask = ifa->ifa_netmask;
980 brdaddr = ifa->ifa_dstaddr;
981 len = rt_msg2(RTM_NEWADDR, &info, 0, w);
982 if (w->w_req && w->w_tmem) {
983 register struct ifa_msghdr *ifam;
984
985 ifam = (struct ifa_msghdr *)w->w_tmem;
986 ifam->ifam_index = ifa->ifa_ifp->if_index;
987 ifam->ifam_flags = ifa->ifa_flags;
988 ifam->ifam_metric = ifa->ifa_metric;
989 ifam->ifam_addrs = info.rti_addrs;
990 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
991 if (error)
992 goto done;
993 }
994 }
995 ifaaddr = netmask = brdaddr = 0;
996 }
997done:
998 return (error);
999}
1000
1001static int
1002sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1003{
1004 int *name = (int *)arg1;
1005 u_int namelen = arg2;
1006 register struct radix_node_head *rnh;
1007 int i, s, error = EINVAL;
1008 u_char af;
1009 struct walkarg w;
1010
1011 name ++;
1012 namelen--;
1013 if (req->newptr)
1014 return (EPERM);
1015 if (namelen != 3)
1016 return ((namelen < 3) ? EISDIR : ENOTDIR);
1017 af = name[0];
1018 Bzero(&w, sizeof(w));
1019 w.w_op = name[1];
1020 w.w_arg = name[2];
1021 w.w_req = req;
1022
1023 s = splnet();
1024 switch (w.w_op) {
1025
1026 case NET_RT_DUMP:
1027 case NET_RT_FLAGS:
1028 for (i = 1; i <= AF_MAX; i++)
1029 if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
1030 (error = rnh->rnh_walktree(rnh,
1031 sysctl_dumpentry, &w)))
1032 break;
1033 break;
1034
1035 case NET_RT_IFLIST:
1036 error = sysctl_iflist(af, &w);
1037 }
1038 splx(s);
1039 if (w.w_tmem)
1040 free(w.w_tmem, M_RTABLE);
1041 return (error);
1042}
1043
1044SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1045
1046/*
1047 * Definitions of protocols supported in the ROUTE domain.
1048 */
1049
1050extern struct domain routedomain; /* or at least forward */
1051
1052static struct protosw routesw[] = {
1053{ SOCK_RAW, &routedomain, 0, PR_ATOMIC|PR_ADDR,
1054 0, route_output, raw_ctlinput, 0,
1055 0,
1056 raw_init, 0, 0, 0,
1057 &route_usrreqs
1058}
1059};
1060
1061static struct domain routedomain =
1062 { PF_ROUTE, "route", 0, 0, 0,
1063 routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] };
1064
1065DOMAIN_SET(route);
|