]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - sys/net/rtsock.c
Merge from OpenBSD 5.0 (this is a dummy change, the vendor change does not
[FreeBSD/FreeBSD.git] / sys / net / rtsock.c
1 /*-
2  * Copyright (c) 1988, 1991, 1993
3  *      The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *      @(#)rtsock.c    8.7 (Berkeley) 10/12/95
30  * $FreeBSD$
31  */
32 #include "opt_compat.h"
33 #include "opt_sctp.h"
34 #include "opt_mpath.h"
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37
38 #include <sys/param.h>
39 #include <sys/jail.h>
40 #include <sys/kernel.h>
41 #include <sys/domain.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/priv.h>
46 #include <sys/proc.h>
47 #include <sys/protosw.h>
48 #include <sys/rwlock.h>
49 #include <sys/signalvar.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/sysctl.h>
53 #include <sys/systm.h>
54
55 #include <net/if.h>
56 #include <net/if_dl.h>
57 #include <net/if_llatbl.h>
58 #include <net/if_types.h>
59 #include <net/netisr.h>
60 #include <net/raw_cb.h>
61 #include <net/route.h>
62 #include <net/vnet.h>
63
64 #include <netinet/in.h>
65 #include <netinet/if_ether.h>
66 #include <netinet/ip_carp.h>
67 #ifdef INET6
68 #include <netinet6/scope6_var.h>
69 #endif
70
71 #if defined(INET) || defined(INET6)
72 #ifdef SCTP
73 extern void sctp_addr_change(struct ifaddr *ifa, int cmd);
74 #endif /* SCTP */
75 #endif
76
77 #ifdef COMPAT_FREEBSD32
78 #include <sys/mount.h>
79 #include <compat/freebsd32/freebsd32.h>
80
81 struct if_data32 {
82         uint8_t ifi_type;
83         uint8_t ifi_physical;
84         uint8_t ifi_addrlen;
85         uint8_t ifi_hdrlen;
86         uint8_t ifi_link_state;
87         uint8_t ifi_vhid;
88         uint8_t ifi_spare_char2;
89         uint8_t ifi_datalen;
90         uint32_t ifi_mtu;
91         uint32_t ifi_metric;
92         uint32_t ifi_baudrate;
93         uint32_t ifi_ipackets;
94         uint32_t ifi_ierrors;
95         uint32_t ifi_opackets;
96         uint32_t ifi_oerrors;
97         uint32_t ifi_collisions;
98         uint32_t ifi_ibytes;
99         uint32_t ifi_obytes;
100         uint32_t ifi_imcasts;
101         uint32_t ifi_omcasts;
102         uint32_t ifi_iqdrops;
103         uint32_t ifi_noproto;
104         uint32_t ifi_hwassist;
105         int32_t ifi_epoch;
106         struct  timeval32 ifi_lastchange;
107 };
108
109 struct if_msghdr32 {
110         uint16_t ifm_msglen;
111         uint8_t ifm_version;
112         uint8_t ifm_type;
113         int32_t ifm_addrs;
114         int32_t ifm_flags;
115         uint16_t ifm_index;
116         struct  if_data32 ifm_data;
117 };
118 #endif
119
120 MALLOC_DEFINE(M_RTABLE, "routetbl", "routing tables");
121
122 /* NB: these are not modified */
123 static struct   sockaddr route_src = { 2, PF_ROUTE, };
124 static struct   sockaddr sa_zero   = { sizeof(sa_zero), AF_INET, };
125
126 /* These are external hooks for CARP. */
127 int     (*carp_get_vhid_p)(struct ifaddr *);
128
129 /*
130  * Used by rtsock/raw_input callback code to decide whether to filter the update
131  * notification to a socket bound to a particular FIB.
132  */
133 #define RTS_FILTER_FIB  M_PROTO8
134 #define RTS_ALLFIBS     -1
135
136 static struct {
137         int     ip_count;       /* attached w/ AF_INET */
138         int     ip6_count;      /* attached w/ AF_INET6 */
139         int     ipx_count;      /* attached w/ AF_IPX */
140         int     any_count;      /* total attached */
141 } route_cb;
142
143 struct mtx rtsock_mtx;
144 MTX_SYSINIT(rtsock, &rtsock_mtx, "rtsock route_cb lock", MTX_DEF);
145
146 #define RTSOCK_LOCK()   mtx_lock(&rtsock_mtx)
147 #define RTSOCK_UNLOCK() mtx_unlock(&rtsock_mtx)
148 #define RTSOCK_LOCK_ASSERT()    mtx_assert(&rtsock_mtx, MA_OWNED)
149
150 static SYSCTL_NODE(_net, OID_AUTO, route, CTLFLAG_RD, 0, "");
151
152 struct walkarg {
153         int     w_tmemsize;
154         int     w_op, w_arg;
155         caddr_t w_tmem;
156         struct sysctl_req *w_req;
157 };
158
159 static void     rts_input(struct mbuf *m);
160 static struct mbuf *rt_msg1(int type, struct rt_addrinfo *rtinfo);
161 static int      rt_msg2(int type, struct rt_addrinfo *rtinfo,
162                         caddr_t cp, struct walkarg *w);
163 static int      rt_xaddrs(caddr_t cp, caddr_t cplim,
164                         struct rt_addrinfo *rtinfo);
165 static int      sysctl_dumpentry(struct radix_node *rn, void *vw);
166 static int      sysctl_iflist(int af, struct walkarg *w);
167 static int      sysctl_ifmalist(int af, struct walkarg *w);
168 static int      route_output(struct mbuf *m, struct socket *so);
169 static void     rt_setmetrics(u_long which, const struct rt_metrics *in,
170                         struct rt_metrics_lite *out);
171 static void     rt_getmetrics(const struct rt_metrics_lite *in,
172                         struct rt_metrics *out);
173 static void     rt_dispatch(struct mbuf *, sa_family_t);
174
175 static struct netisr_handler rtsock_nh = {
176         .nh_name = "rtsock",
177         .nh_handler = rts_input,
178         .nh_proto = NETISR_ROUTE,
179         .nh_policy = NETISR_POLICY_SOURCE,
180 };
181
182 static int
183 sysctl_route_netisr_maxqlen(SYSCTL_HANDLER_ARGS)
184 {
185         int error, qlimit;
186
187         netisr_getqlimit(&rtsock_nh, &qlimit);
188         error = sysctl_handle_int(oidp, &qlimit, 0, req);
189         if (error || !req->newptr)
190                 return (error);
191         if (qlimit < 1)
192                 return (EINVAL);
193         return (netisr_setqlimit(&rtsock_nh, qlimit));
194 }
195 SYSCTL_PROC(_net_route, OID_AUTO, netisr_maxqlen, CTLTYPE_INT|CTLFLAG_RW,
196     0, 0, sysctl_route_netisr_maxqlen, "I",
197     "maximum routing socket dispatch queue length");
198
199 static void
200 rts_init(void)
201 {
202         int tmp;
203
204         if (TUNABLE_INT_FETCH("net.route.netisr_maxqlen", &tmp))
205                 rtsock_nh.nh_qlimit = tmp;
206         netisr_register(&rtsock_nh);
207 }
208 SYSINIT(rtsock, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rts_init, 0);
209
210 static int
211 raw_input_rts_cb(struct mbuf *m, struct sockproto *proto, struct sockaddr *src,
212     struct rawcb *rp)
213 {
214         int fibnum;
215
216         KASSERT(m != NULL, ("%s: m is NULL", __func__));
217         KASSERT(proto != NULL, ("%s: proto is NULL", __func__));
218         KASSERT(rp != NULL, ("%s: rp is NULL", __func__));
219
220         /* No filtering requested. */
221         if ((m->m_flags & RTS_FILTER_FIB) == 0)
222                 return (0);
223
224         /* Check if it is a rts and the fib matches the one of the socket. */
225         fibnum = M_GETFIB(m);
226         if (proto->sp_family != PF_ROUTE ||
227             rp->rcb_socket == NULL ||
228             rp->rcb_socket->so_fibnum == fibnum)
229                 return (0);
230
231         /* Filtering requested and no match, the socket shall be skipped. */
232         return (1);
233 }
234
235 static void
236 rts_input(struct mbuf *m)
237 {
238         struct sockproto route_proto;
239         unsigned short *family;
240         struct m_tag *tag;
241
242         route_proto.sp_family = PF_ROUTE;
243         tag = m_tag_find(m, PACKET_TAG_RTSOCKFAM, NULL);
244         if (tag != NULL) {
245                 family = (unsigned short *)(tag + 1);
246                 route_proto.sp_protocol = *family;
247                 m_tag_delete(m, tag);
248         } else
249                 route_proto.sp_protocol = 0;
250
251         raw_input_ext(m, &route_proto, &route_src, raw_input_rts_cb);
252 }
253
254 /*
255  * It really doesn't make any sense at all for this code to share much
256  * with raw_usrreq.c, since its functionality is so restricted.  XXX
257  */
258 static void
259 rts_abort(struct socket *so)
260 {
261
262         raw_usrreqs.pru_abort(so);
263 }
264
265 static void
266 rts_close(struct socket *so)
267 {
268
269         raw_usrreqs.pru_close(so);
270 }
271
272 /* pru_accept is EOPNOTSUPP */
273
274 static int
275 rts_attach(struct socket *so, int proto, struct thread *td)
276 {
277         struct rawcb *rp;
278         int s, error;
279
280         KASSERT(so->so_pcb == NULL, ("rts_attach: so_pcb != NULL"));
281
282         /* XXX */
283         rp = malloc(sizeof *rp, M_PCB, M_WAITOK | M_ZERO);
284         if (rp == NULL)
285                 return ENOBUFS;
286
287         /*
288          * The splnet() is necessary to block protocols from sending
289          * error notifications (like RTM_REDIRECT or RTM_LOSING) while
290          * this PCB is extant but incompletely initialized.
291          * Probably we should try to do more of this work beforehand and
292          * eliminate the spl.
293          */
294         s = splnet();
295         so->so_pcb = (caddr_t)rp;
296         so->so_fibnum = td->td_proc->p_fibnum;
297         error = raw_attach(so, proto);
298         rp = sotorawcb(so);
299         if (error) {
300                 splx(s);
301                 so->so_pcb = NULL;
302                 free(rp, M_PCB);
303                 return error;
304         }
305         RTSOCK_LOCK();
306         switch(rp->rcb_proto.sp_protocol) {
307         case AF_INET:
308                 route_cb.ip_count++;
309                 break;
310         case AF_INET6:
311                 route_cb.ip6_count++;
312                 break;
313         case AF_IPX:
314                 route_cb.ipx_count++;
315                 break;
316         }
317         route_cb.any_count++;
318         RTSOCK_UNLOCK();
319         soisconnected(so);
320         so->so_options |= SO_USELOOPBACK;
321         splx(s);
322         return 0;
323 }
324
325 static int
326 rts_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
327 {
328
329         return (raw_usrreqs.pru_bind(so, nam, td)); /* xxx just EINVAL */
330 }
331
332 static int
333 rts_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
334 {
335
336         return (raw_usrreqs.pru_connect(so, nam, td)); /* XXX just EINVAL */
337 }
338
339 /* pru_connect2 is EOPNOTSUPP */
340 /* pru_control is EOPNOTSUPP */
341
342 static void
343 rts_detach(struct socket *so)
344 {
345         struct rawcb *rp = sotorawcb(so);
346
347         KASSERT(rp != NULL, ("rts_detach: rp == NULL"));
348
349         RTSOCK_LOCK();
350         switch(rp->rcb_proto.sp_protocol) {
351         case AF_INET:
352                 route_cb.ip_count--;
353                 break;
354         case AF_INET6:
355                 route_cb.ip6_count--;
356                 break;
357         case AF_IPX:
358                 route_cb.ipx_count--;
359                 break;
360         }
361         route_cb.any_count--;
362         RTSOCK_UNLOCK();
363         raw_usrreqs.pru_detach(so);
364 }
365
366 static int
367 rts_disconnect(struct socket *so)
368 {
369
370         return (raw_usrreqs.pru_disconnect(so));
371 }
372
373 /* pru_listen is EOPNOTSUPP */
374
375 static int
376 rts_peeraddr(struct socket *so, struct sockaddr **nam)
377 {
378
379         return (raw_usrreqs.pru_peeraddr(so, nam));
380 }
381
382 /* pru_rcvd is EOPNOTSUPP */
383 /* pru_rcvoob is EOPNOTSUPP */
384
385 static int
386 rts_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
387          struct mbuf *control, struct thread *td)
388 {
389
390         return (raw_usrreqs.pru_send(so, flags, m, nam, control, td));
391 }
392
393 /* pru_sense is null */
394
395 static int
396 rts_shutdown(struct socket *so)
397 {
398
399         return (raw_usrreqs.pru_shutdown(so));
400 }
401
402 static int
403 rts_sockaddr(struct socket *so, struct sockaddr **nam)
404 {
405
406         return (raw_usrreqs.pru_sockaddr(so, nam));
407 }
408
409 static struct pr_usrreqs route_usrreqs = {
410         .pru_abort =            rts_abort,
411         .pru_attach =           rts_attach,
412         .pru_bind =             rts_bind,
413         .pru_connect =          rts_connect,
414         .pru_detach =           rts_detach,
415         .pru_disconnect =       rts_disconnect,
416         .pru_peeraddr =         rts_peeraddr,
417         .pru_send =             rts_send,
418         .pru_shutdown =         rts_shutdown,
419         .pru_sockaddr =         rts_sockaddr,
420         .pru_close =            rts_close,
421 };
422
423 #ifndef _SOCKADDR_UNION_DEFINED
424 #define _SOCKADDR_UNION_DEFINED
425 /*
426  * The union of all possible address formats we handle.
427  */
428 union sockaddr_union {
429         struct sockaddr         sa;
430         struct sockaddr_in      sin;
431         struct sockaddr_in6     sin6;
432 };
433 #endif /* _SOCKADDR_UNION_DEFINED */
434
435 static int
436 rtm_get_jailed(struct rt_addrinfo *info, struct ifnet *ifp,
437     struct rtentry *rt, union sockaddr_union *saun, struct ucred *cred)
438 {
439
440         /* First, see if the returned address is part of the jail. */
441         if (prison_if(cred, rt->rt_ifa->ifa_addr) == 0) {
442                 info->rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
443                 return (0);
444         }
445
446         switch (info->rti_info[RTAX_DST]->sa_family) {
447 #ifdef INET
448         case AF_INET:
449         {
450                 struct in_addr ia;
451                 struct ifaddr *ifa;
452                 int found;
453
454                 found = 0;
455                 /*
456                  * Try to find an address on the given outgoing interface
457                  * that belongs to the jail.
458                  */
459                 IF_ADDR_LOCK(ifp);
460                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
461                         struct sockaddr *sa;
462                         sa = ifa->ifa_addr;
463                         if (sa->sa_family != AF_INET)
464                                 continue;
465                         ia = ((struct sockaddr_in *)sa)->sin_addr;
466                         if (prison_check_ip4(cred, &ia) == 0) {
467                                 found = 1;
468                                 break;
469                         }
470                 }
471                 IF_ADDR_UNLOCK(ifp);
472                 if (!found) {
473                         /*
474                          * As a last resort return the 'default' jail address.
475                          */
476                         ia = ((struct sockaddr_in *)rt->rt_ifa->ifa_addr)->
477                             sin_addr;
478                         if (prison_get_ip4(cred, &ia) != 0)
479                                 return (ESRCH);
480                 }
481                 bzero(&saun->sin, sizeof(struct sockaddr_in));
482                 saun->sin.sin_len = sizeof(struct sockaddr_in);
483                 saun->sin.sin_family = AF_INET;
484                 saun->sin.sin_addr.s_addr = ia.s_addr;
485                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin;
486                 break;
487         }
488 #endif
489 #ifdef INET6
490         case AF_INET6:
491         {
492                 struct in6_addr ia6;
493                 struct ifaddr *ifa;
494                 int found;
495
496                 found = 0;
497                 /*
498                  * Try to find an address on the given outgoing interface
499                  * that belongs to the jail.
500                  */
501                 IF_ADDR_LOCK(ifp);
502                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
503                         struct sockaddr *sa;
504                         sa = ifa->ifa_addr;
505                         if (sa->sa_family != AF_INET6)
506                                 continue;
507                         bcopy(&((struct sockaddr_in6 *)sa)->sin6_addr,
508                             &ia6, sizeof(struct in6_addr));
509                         if (prison_check_ip6(cred, &ia6) == 0) {
510                                 found = 1;
511                                 break;
512                         }
513                 }
514                 IF_ADDR_UNLOCK(ifp);
515                 if (!found) {
516                         /*
517                          * As a last resort return the 'default' jail address.
518                          */
519                         ia6 = ((struct sockaddr_in6 *)rt->rt_ifa->ifa_addr)->
520                             sin6_addr;
521                         if (prison_get_ip6(cred, &ia6) != 0)
522                                 return (ESRCH);
523                 }
524                 bzero(&saun->sin6, sizeof(struct sockaddr_in6));
525                 saun->sin6.sin6_len = sizeof(struct sockaddr_in6);
526                 saun->sin6.sin6_family = AF_INET6;
527                 bcopy(&ia6, &saun->sin6.sin6_addr, sizeof(struct in6_addr));
528                 if (sa6_recoverscope(&saun->sin6) != 0)
529                         return (ESRCH);
530                 info->rti_info[RTAX_IFA] = (struct sockaddr *)&saun->sin6;
531                 break;
532         }
533 #endif
534         default:
535                 return (ESRCH);
536         }
537         return (0);
538 }
539
540 /*ARGSUSED*/
541 static int
542 route_output(struct mbuf *m, struct socket *so)
543 {
544 #define sa_equal(a1, a2) (bcmp((a1), (a2), (a1)->sa_len) == 0)
545         struct rt_msghdr *rtm = NULL;
546         struct rtentry *rt = NULL;
547         struct radix_node_head *rnh;
548         struct rt_addrinfo info;
549         int len, error = 0;
550         struct ifnet *ifp = NULL;
551         union sockaddr_union saun;
552         sa_family_t saf = AF_UNSPEC;
553
554 #define senderr(e) { error = e; goto flush;}
555         if (m == NULL || ((m->m_len < sizeof(long)) &&
556                        (m = m_pullup(m, sizeof(long))) == NULL))
557                 return (ENOBUFS);
558         if ((m->m_flags & M_PKTHDR) == 0)
559                 panic("route_output");
560         len = m->m_pkthdr.len;
561         if (len < sizeof(*rtm) ||
562             len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
563                 info.rti_info[RTAX_DST] = NULL;
564                 senderr(EINVAL);
565         }
566         R_Malloc(rtm, struct rt_msghdr *, len);
567         if (rtm == NULL) {
568                 info.rti_info[RTAX_DST] = NULL;
569                 senderr(ENOBUFS);
570         }
571         m_copydata(m, 0, len, (caddr_t)rtm);
572         if (rtm->rtm_version != RTM_VERSION) {
573                 info.rti_info[RTAX_DST] = NULL;
574                 senderr(EPROTONOSUPPORT);
575         }
576         rtm->rtm_pid = curproc->p_pid;
577         bzero(&info, sizeof(info));
578         info.rti_addrs = rtm->rtm_addrs;
579         if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info)) {
580                 info.rti_info[RTAX_DST] = NULL;
581                 senderr(EINVAL);
582         }
583         info.rti_flags = rtm->rtm_flags;
584         if (info.rti_info[RTAX_DST] == NULL ||
585             info.rti_info[RTAX_DST]->sa_family >= AF_MAX ||
586             (info.rti_info[RTAX_GATEWAY] != NULL &&
587              info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX))
588                 senderr(EINVAL);
589         saf = info.rti_info[RTAX_DST]->sa_family;
590         /*
591          * Verify that the caller has the appropriate privilege; RTM_GET
592          * is the only operation the non-superuser is allowed.
593          */
594         if (rtm->rtm_type != RTM_GET) {
595                 error = priv_check(curthread, PRIV_NET_ROUTE);
596                 if (error)
597                         senderr(error);
598         }
599
600         /*
601          * The given gateway address may be an interface address.
602          * For example, issuing a "route change" command on a route
603          * entry that was created from a tunnel, and the gateway
604          * address given is the local end point. In this case the 
605          * RTF_GATEWAY flag must be cleared or the destination will
606          * not be reachable even though there is no error message.
607          */
608         if (info.rti_info[RTAX_GATEWAY] != NULL &&
609             info.rti_info[RTAX_GATEWAY]->sa_family != AF_LINK) {
610                 struct route gw_ro;
611
612                 bzero(&gw_ro, sizeof(gw_ro));
613                 gw_ro.ro_dst = *info.rti_info[RTAX_GATEWAY];
614                 rtalloc_ign_fib(&gw_ro, 0, so->so_fibnum);
615                 /* 
616                  * A host route through the loopback interface is 
617                  * installed for each interface adddress. In pre 8.0
618                  * releases the interface address of a PPP link type
619                  * is not reachable locally. This behavior is fixed as 
620                  * part of the new L2/L3 redesign and rewrite work. The
621                  * signature of this interface address route is the
622                  * AF_LINK sa_family type of the rt_gateway, and the
623                  * rt_ifp has the IFF_LOOPBACK flag set.
624                  */
625                 if (gw_ro.ro_rt != NULL &&
626                     gw_ro.ro_rt->rt_gateway->sa_family == AF_LINK &&
627                     gw_ro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)
628                         info.rti_flags &= ~RTF_GATEWAY;
629                 if (gw_ro.ro_rt != NULL)
630                         RTFREE(gw_ro.ro_rt);
631         }
632
633         switch (rtm->rtm_type) {
634                 struct rtentry *saved_nrt;
635
636         case RTM_ADD:
637                 if (info.rti_info[RTAX_GATEWAY] == NULL)
638                         senderr(EINVAL);
639                 saved_nrt = NULL;
640
641                 /* support for new ARP code */
642                 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK &&
643                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
644                         error = lla_rt_output(rtm, &info);
645                         break;
646                 }
647                 error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt,
648                     so->so_fibnum);
649                 if (error == 0 && saved_nrt) {
650                         RT_LOCK(saved_nrt);
651                         rt_setmetrics(rtm->rtm_inits,
652                                 &rtm->rtm_rmx, &saved_nrt->rt_rmx);
653                         rtm->rtm_index = saved_nrt->rt_ifp->if_index;
654                         RT_REMREF(saved_nrt);
655                         RT_UNLOCK(saved_nrt);
656                 }
657                 break;
658
659         case RTM_DELETE:
660                 saved_nrt = NULL;
661                 /* support for new ARP code */
662                 if (info.rti_info[RTAX_GATEWAY] && 
663                     (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
664                     (rtm->rtm_flags & RTF_LLDATA) != 0) {
665                         error = lla_rt_output(rtm, &info);
666                         break;
667                 }
668                 error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt,
669                     so->so_fibnum);
670                 if (error == 0) {
671                         RT_LOCK(saved_nrt);
672                         rt = saved_nrt;
673                         goto report;
674                 }
675                 break;
676
677         case RTM_GET:
678         case RTM_CHANGE:
679         case RTM_LOCK:
680                 rnh = rt_tables_get_rnh(so->so_fibnum,
681                     info.rti_info[RTAX_DST]->sa_family);
682                 if (rnh == NULL)
683                         senderr(EAFNOSUPPORT);
684                 RADIX_NODE_HEAD_RLOCK(rnh);
685                 rt = (struct rtentry *) rnh->rnh_lookup(info.rti_info[RTAX_DST],
686                         info.rti_info[RTAX_NETMASK], rnh);
687                 if (rt == NULL) {       /* XXX looks bogus */
688                         RADIX_NODE_HEAD_RUNLOCK(rnh);
689                         senderr(ESRCH);
690                 }
691 #ifdef RADIX_MPATH
692                 /*
693                  * for RTM_CHANGE/LOCK, if we got multipath routes,
694                  * we require users to specify a matching RTAX_GATEWAY.
695                  *
696                  * for RTM_GET, gate is optional even with multipath.
697                  * if gate == NULL the first match is returned.
698                  * (no need to call rt_mpath_matchgate if gate == NULL)
699                  */
700                 if (rn_mpath_capable(rnh) &&
701                     (rtm->rtm_type != RTM_GET || info.rti_info[RTAX_GATEWAY])) {
702                         rt = rt_mpath_matchgate(rt, info.rti_info[RTAX_GATEWAY]);
703                         if (!rt) {
704                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
705                                 senderr(ESRCH);
706                         }
707                 }
708 #endif
709                 /*
710                  * If performing proxied L2 entry insertion, and
711                  * the actual PPP host entry is found, perform
712                  * another search to retrieve the prefix route of
713                  * the local end point of the PPP link.
714                  */
715                 if (rtm->rtm_flags & RTF_ANNOUNCE) {
716                         struct sockaddr laddr;
717
718                         if (rt->rt_ifp != NULL && 
719                             rt->rt_ifp->if_type == IFT_PROPVIRTUAL) {
720                                 struct ifaddr *ifa;
721
722                                 ifa = ifa_ifwithnet(info.rti_info[RTAX_DST], 1);
723                                 if (ifa != NULL)
724                                         rt_maskedcopy(ifa->ifa_addr,
725                                                       &laddr,
726                                                       ifa->ifa_netmask);
727                         } else
728                                 rt_maskedcopy(rt->rt_ifa->ifa_addr,
729                                               &laddr,
730                                               rt->rt_ifa->ifa_netmask);
731                         /* 
732                          * refactor rt and no lock operation necessary
733                          */
734                         rt = (struct rtentry *)rnh->rnh_matchaddr(&laddr, rnh);
735                         if (rt == NULL) {
736                                 RADIX_NODE_HEAD_RUNLOCK(rnh);
737                                 senderr(ESRCH);
738                         }
739                 } 
740                 RT_LOCK(rt);
741                 RT_ADDREF(rt);
742                 RADIX_NODE_HEAD_RUNLOCK(rnh);
743
744                 /* 
745                  * Fix for PR: 82974
746                  *
747                  * RTM_CHANGE/LOCK need a perfect match, rn_lookup()
748                  * returns a perfect match in case a netmask is
749                  * specified.  For host routes only a longest prefix
750                  * match is returned so it is necessary to compare the
751                  * existence of the netmask.  If both have a netmask
752                  * rnh_lookup() did a perfect match and if none of them
753                  * have a netmask both are host routes which is also a
754                  * perfect match.
755                  */
756
757                 if (rtm->rtm_type != RTM_GET && 
758                     (!rt_mask(rt) != !info.rti_info[RTAX_NETMASK])) {
759                         RT_UNLOCK(rt);
760                         senderr(ESRCH);
761                 }
762
763                 switch(rtm->rtm_type) {
764
765                 case RTM_GET:
766                 report:
767                         RT_LOCK_ASSERT(rt);
768                         if ((rt->rt_flags & RTF_HOST) == 0
769                             ? jailed_without_vnet(curthread->td_ucred)
770                             : prison_if(curthread->td_ucred,
771                             rt_key(rt)) != 0) {
772                                 RT_UNLOCK(rt);
773                                 senderr(ESRCH);
774                         }
775                         info.rti_info[RTAX_DST] = rt_key(rt);
776                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
777                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
778                         info.rti_info[RTAX_GENMASK] = 0;
779                         if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
780                                 ifp = rt->rt_ifp;
781                                 if (ifp) {
782                                         info.rti_info[RTAX_IFP] =
783                                             ifp->if_addr->ifa_addr;
784                                         error = rtm_get_jailed(&info, ifp, rt,
785                                             &saun, curthread->td_ucred);
786                                         if (error != 0) {
787                                                 RT_UNLOCK(rt);
788                                                 senderr(error);
789                                         }
790                                         if (ifp->if_flags & IFF_POINTOPOINT)
791                                                 info.rti_info[RTAX_BRD] =
792                                                     rt->rt_ifa->ifa_dstaddr;
793                                         rtm->rtm_index = ifp->if_index;
794                                 } else {
795                                         info.rti_info[RTAX_IFP] = NULL;
796                                         info.rti_info[RTAX_IFA] = NULL;
797                                 }
798                         } else if ((ifp = rt->rt_ifp) != NULL) {
799                                 rtm->rtm_index = ifp->if_index;
800                         }
801                         len = rt_msg2(rtm->rtm_type, &info, NULL, NULL);
802                         if (len > rtm->rtm_msglen) {
803                                 struct rt_msghdr *new_rtm;
804                                 R_Malloc(new_rtm, struct rt_msghdr *, len);
805                                 if (new_rtm == NULL) {
806                                         RT_UNLOCK(rt);
807                                         senderr(ENOBUFS);
808                                 }
809                                 bcopy(rtm, new_rtm, rtm->rtm_msglen);
810                                 Free(rtm); rtm = new_rtm;
811                         }
812                         (void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm, NULL);
813                         rtm->rtm_flags = rt->rt_flags;
814                         rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
815                         rtm->rtm_addrs = info.rti_addrs;
816                         break;
817
818                 case RTM_CHANGE:
819                         /*
820                          * New gateway could require new ifaddr, ifp;
821                          * flags may also be different; ifp may be specified
822                          * by ll sockaddr when protocol address is ambiguous
823                          */
824                         if (((rt->rt_flags & RTF_GATEWAY) &&
825                              info.rti_info[RTAX_GATEWAY] != NULL) ||
826                             info.rti_info[RTAX_IFP] != NULL ||
827                             (info.rti_info[RTAX_IFA] != NULL &&
828                              !sa_equal(info.rti_info[RTAX_IFA],
829                                        rt->rt_ifa->ifa_addr))) {
830                                 RT_UNLOCK(rt);
831                                 RADIX_NODE_HEAD_LOCK(rnh);
832                                 error = rt_getifa_fib(&info, rt->rt_fibnum);
833                                 /*
834                                  * XXXRW: Really we should release this
835                                  * reference later, but this maintains
836                                  * historical behavior.
837                                  */
838                                 if (info.rti_ifa != NULL)
839                                         ifa_free(info.rti_ifa);
840                                 RADIX_NODE_HEAD_UNLOCK(rnh);
841                                 if (error != 0)
842                                         senderr(error);
843                                 RT_LOCK(rt);
844                         }
845                         if (info.rti_ifa != NULL &&
846                             info.rti_ifa != rt->rt_ifa &&
847                             rt->rt_ifa != NULL &&
848                             rt->rt_ifa->ifa_rtrequest != NULL) {
849                                 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt,
850                                     &info);
851                                 ifa_free(rt->rt_ifa);
852                         }
853                         if (info.rti_info[RTAX_GATEWAY] != NULL) {
854                                 RT_UNLOCK(rt);
855                                 RADIX_NODE_HEAD_LOCK(rnh);
856                                 RT_LOCK(rt);
857                                 
858                                 error = rt_setgate(rt, rt_key(rt),
859                                     info.rti_info[RTAX_GATEWAY]);
860                                 RADIX_NODE_HEAD_UNLOCK(rnh);
861                                 if (error != 0) {
862                                         RT_UNLOCK(rt);
863                                         senderr(error);
864                                 }
865                                 rt->rt_flags |= (RTF_GATEWAY & info.rti_flags);
866                         }
867                         if (info.rti_ifa != NULL &&
868                             info.rti_ifa != rt->rt_ifa) {
869                                 ifa_ref(info.rti_ifa);
870                                 rt->rt_ifa = info.rti_ifa;
871                                 rt->rt_ifp = info.rti_ifp;
872                         }
873                         /* Allow some flags to be toggled on change. */
874                         rt->rt_flags = (rt->rt_flags & ~RTF_FMASK) |
875                                     (rtm->rtm_flags & RTF_FMASK);
876                         rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
877                                         &rt->rt_rmx);
878                         rtm->rtm_index = rt->rt_ifp->if_index;
879                         if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
880                                rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
881                         /* FALLTHROUGH */
882                 case RTM_LOCK:
883                         /* We don't support locks anymore */
884                         break;
885                 }
886                 RT_UNLOCK(rt);
887                 break;
888
889         default:
890                 senderr(EOPNOTSUPP);
891         }
892
893 flush:
894         if (rtm) {
895                 if (error)
896                         rtm->rtm_errno = error;
897                 else
898                         rtm->rtm_flags |= RTF_DONE;
899         }
900         if (rt)         /* XXX can this be true? */
901                 RTFREE(rt);
902     {
903         struct rawcb *rp = NULL;
904         /*
905          * Check to see if we don't want our own messages.
906          */
907         if ((so->so_options & SO_USELOOPBACK) == 0) {
908                 if (route_cb.any_count <= 1) {
909                         if (rtm)
910                                 Free(rtm);
911                         m_freem(m);
912                         return (error);
913                 }
914                 /* There is another listener, so construct message */
915                 rp = sotorawcb(so);
916         }
917         if (rtm) {
918                 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
919                 if (m->m_pkthdr.len < rtm->rtm_msglen) {
920                         m_freem(m);
921                         m = NULL;
922                 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
923                         m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
924         }
925         if (m) {
926                 M_SETFIB(m, so->so_fibnum);
927                 m->m_flags |= RTS_FILTER_FIB;
928                 if (rp) {
929                         /*
930                          * XXX insure we don't get a copy by
931                          * invalidating our protocol
932                          */
933                         unsigned short family = rp->rcb_proto.sp_family;
934                         rp->rcb_proto.sp_family = 0;
935                         rt_dispatch(m, saf);
936                         rp->rcb_proto.sp_family = family;
937                 } else
938                         rt_dispatch(m, saf);
939         }
940         /* info.rti_info[RTAX_DST] (used above) can point inside of rtm */
941         if (rtm)
942                 Free(rtm);
943     }
944         return (error);
945 #undef  sa_equal
946 }
947
948 static void
949 rt_setmetrics(u_long which, const struct rt_metrics *in,
950         struct rt_metrics_lite *out)
951 {
952 #define metric(f, e) if (which & (f)) out->e = in->e;
953         /*
954          * Only these are stored in the routing entry since introduction
955          * of tcp hostcache. The rest is ignored.
956          */
957         metric(RTV_MTU, rmx_mtu);
958         metric(RTV_WEIGHT, rmx_weight);
959         /* Userland -> kernel timebase conversion. */
960         if (which & RTV_EXPIRE)
961                 out->rmx_expire = in->rmx_expire ?
962                     in->rmx_expire - time_second + time_uptime : 0;
963 #undef metric
964 }
965
966 static void
967 rt_getmetrics(const struct rt_metrics_lite *in, struct rt_metrics *out)
968 {
969 #define metric(e) out->e = in->e;
970         bzero(out, sizeof(*out));
971         metric(rmx_mtu);
972         metric(rmx_weight);
973         /* Kernel -> userland timebase conversion. */
974         out->rmx_expire = in->rmx_expire ?
975             in->rmx_expire - time_uptime + time_second : 0;
976 #undef metric
977 }
978
979 /*
980  * Extract the addresses of the passed sockaddrs.
981  * Do a little sanity checking so as to avoid bad memory references.
982  * This data is derived straight from userland.
983  */
984 static int
985 rt_xaddrs(caddr_t cp, caddr_t cplim, struct rt_addrinfo *rtinfo)
986 {
987         struct sockaddr *sa;
988         int i;
989
990         for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
991                 if ((rtinfo->rti_addrs & (1 << i)) == 0)
992                         continue;
993                 sa = (struct sockaddr *)cp;
994                 /*
995                  * It won't fit.
996                  */
997                 if (cp + sa->sa_len > cplim)
998                         return (EINVAL);
999                 /*
1000                  * there are no more.. quit now
1001                  * If there are more bits, they are in error.
1002                  * I've seen this. route(1) can evidently generate these. 
1003                  * This causes kernel to core dump.
1004                  * for compatibility, If we see this, point to a safe address.
1005                  */
1006                 if (sa->sa_len == 0) {
1007                         rtinfo->rti_info[i] = &sa_zero;
1008                         return (0); /* should be EINVAL but for compat */
1009                 }
1010                 /* accept it */
1011                 rtinfo->rti_info[i] = sa;
1012                 cp += SA_SIZE(sa);
1013         }
1014         return (0);
1015 }
1016
1017 static struct mbuf *
1018 rt_msg1(int type, struct rt_addrinfo *rtinfo)
1019 {
1020         struct rt_msghdr *rtm;
1021         struct mbuf *m;
1022         int i;
1023         struct sockaddr *sa;
1024         int len, dlen;
1025
1026         switch (type) {
1027
1028         case RTM_DELADDR:
1029         case RTM_NEWADDR:
1030                 len = sizeof(struct ifa_msghdr);
1031                 break;
1032
1033         case RTM_DELMADDR:
1034         case RTM_NEWMADDR:
1035                 len = sizeof(struct ifma_msghdr);
1036                 break;
1037
1038         case RTM_IFINFO:
1039                 len = sizeof(struct if_msghdr);
1040                 break;
1041
1042         case RTM_IFANNOUNCE:
1043         case RTM_IEEE80211:
1044                 len = sizeof(struct if_announcemsghdr);
1045                 break;
1046
1047         default:
1048                 len = sizeof(struct rt_msghdr);
1049         }
1050         if (len > MCLBYTES)
1051                 panic("rt_msg1");
1052         m = m_gethdr(M_DONTWAIT, MT_DATA);
1053         if (m && len > MHLEN) {
1054                 MCLGET(m, M_DONTWAIT);
1055                 if ((m->m_flags & M_EXT) == 0) {
1056                         m_free(m);
1057                         m = NULL;
1058                 }
1059         }
1060         if (m == NULL)
1061                 return (m);
1062         m->m_pkthdr.len = m->m_len = len;
1063         m->m_pkthdr.rcvif = NULL;
1064         rtm = mtod(m, struct rt_msghdr *);
1065         bzero((caddr_t)rtm, len);
1066         for (i = 0; i < RTAX_MAX; i++) {
1067                 if ((sa = rtinfo->rti_info[i]) == NULL)
1068                         continue;
1069                 rtinfo->rti_addrs |= (1 << i);
1070                 dlen = SA_SIZE(sa);
1071                 m_copyback(m, len, dlen, (caddr_t)sa);
1072                 len += dlen;
1073         }
1074         if (m->m_pkthdr.len != len) {
1075                 m_freem(m);
1076                 return (NULL);
1077         }
1078         rtm->rtm_msglen = len;
1079         rtm->rtm_version = RTM_VERSION;
1080         rtm->rtm_type = type;
1081         return (m);
1082 }
1083
1084 static int
1085 rt_msg2(int type, struct rt_addrinfo *rtinfo, caddr_t cp, struct walkarg *w)
1086 {
1087         int i;
1088         int len, dlen, second_time = 0;
1089         caddr_t cp0;
1090
1091         rtinfo->rti_addrs = 0;
1092 again:
1093         switch (type) {
1094
1095         case RTM_DELADDR:
1096         case RTM_NEWADDR:
1097                 len = sizeof(struct ifa_msghdr);
1098                 break;
1099
1100         case RTM_IFINFO:
1101 #ifdef COMPAT_FREEBSD32
1102                 if (w != NULL && w->w_req->flags & SCTL_MASK32) {
1103                         len = sizeof(struct if_msghdr32);
1104                         break;
1105                 }
1106 #endif
1107                 len = sizeof(struct if_msghdr);
1108                 break;
1109
1110         case RTM_NEWMADDR:
1111                 len = sizeof(struct ifma_msghdr);
1112                 break;
1113
1114         default:
1115                 len = sizeof(struct rt_msghdr);
1116         }
1117         cp0 = cp;
1118         if (cp0)
1119                 cp += len;
1120         for (i = 0; i < RTAX_MAX; i++) {
1121                 struct sockaddr *sa;
1122
1123                 if ((sa = rtinfo->rti_info[i]) == NULL)
1124                         continue;
1125                 rtinfo->rti_addrs |= (1 << i);
1126                 dlen = SA_SIZE(sa);
1127                 if (cp) {
1128                         bcopy((caddr_t)sa, cp, (unsigned)dlen);
1129                         cp += dlen;
1130                 }
1131                 len += dlen;
1132         }
1133         len = ALIGN(len);
1134         if (cp == NULL && w != NULL && !second_time) {
1135                 struct walkarg *rw = w;
1136
1137                 if (rw->w_req) {
1138                         if (rw->w_tmemsize < len) {
1139                                 if (rw->w_tmem)
1140                                         free(rw->w_tmem, M_RTABLE);
1141                                 rw->w_tmem = (caddr_t)
1142                                         malloc(len, M_RTABLE, M_NOWAIT);
1143                                 if (rw->w_tmem)
1144                                         rw->w_tmemsize = len;
1145                         }
1146                         if (rw->w_tmem) {
1147                                 cp = rw->w_tmem;
1148                                 second_time = 1;
1149                                 goto again;
1150                         }
1151                 }
1152         }
1153         if (cp) {
1154                 struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
1155
1156                 rtm->rtm_version = RTM_VERSION;
1157                 rtm->rtm_type = type;
1158                 rtm->rtm_msglen = len;
1159         }
1160         return (len);
1161 }
1162
1163 /*
1164  * This routine is called to generate a message from the routing
1165  * socket indicating that a redirect has occured, a routing lookup
1166  * has failed, or that a protocol has detected timeouts to a particular
1167  * destination.
1168  */
1169 void
1170 rt_missmsg_fib(int type, struct rt_addrinfo *rtinfo, int flags, int error,
1171     int fibnum)
1172 {
1173         struct rt_msghdr *rtm;
1174         struct mbuf *m;
1175         struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1176
1177         if (route_cb.any_count == 0)
1178                 return;
1179         m = rt_msg1(type, rtinfo);
1180         if (m == NULL)
1181                 return;
1182
1183         if (fibnum != RTS_ALLFIBS) {
1184                 KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: fibnum out "
1185                     "of range 0 <= %d < %d", __func__, fibnum, rt_numfibs));
1186                 M_SETFIB(m, fibnum);
1187                 m->m_flags |= RTS_FILTER_FIB;
1188         }
1189
1190         rtm = mtod(m, struct rt_msghdr *);
1191         rtm->rtm_flags = RTF_DONE | flags;
1192         rtm->rtm_errno = error;
1193         rtm->rtm_addrs = rtinfo->rti_addrs;
1194         rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1195 }
1196
1197 void
1198 rt_missmsg(int type, struct rt_addrinfo *rtinfo, int flags, int error)
1199 {
1200
1201         rt_missmsg_fib(type, rtinfo, flags, error, RTS_ALLFIBS);
1202 }
1203
1204 /*
1205  * This routine is called to generate a message from the routing
1206  * socket indicating that the status of a network interface has changed.
1207  */
1208 void
1209 rt_ifmsg(struct ifnet *ifp)
1210 {
1211         struct if_msghdr *ifm;
1212         struct mbuf *m;
1213         struct rt_addrinfo info;
1214
1215         if (route_cb.any_count == 0)
1216                 return;
1217         bzero((caddr_t)&info, sizeof(info));
1218         m = rt_msg1(RTM_IFINFO, &info);
1219         if (m == NULL)
1220                 return;
1221         ifm = mtod(m, struct if_msghdr *);
1222         ifm->ifm_index = ifp->if_index;
1223         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1224         ifm->ifm_data = ifp->if_data;
1225         ifm->ifm_addrs = 0;
1226         rt_dispatch(m, AF_UNSPEC);
1227 }
1228
1229 /*
1230  * This is called to generate messages from the routing socket
1231  * indicating a network interface has had addresses associated with it.
1232  * if we ever reverse the logic and replace messages TO the routing
1233  * socket indicate a request to configure interfaces, then it will
1234  * be unnecessary as the routing socket will automatically generate
1235  * copies of it.
1236  */
1237 void
1238 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
1239     int fibnum)
1240 {
1241         struct rt_addrinfo info;
1242         struct sockaddr *sa = NULL;
1243         int pass;
1244         struct mbuf *m = NULL;
1245         struct ifnet *ifp = ifa->ifa_ifp;
1246
1247         KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
1248                 ("unexpected cmd %u", cmd));
1249 #if defined(INET) || defined(INET6)
1250 #ifdef SCTP
1251         /*
1252          * notify the SCTP stack
1253          * this will only get called when an address is added/deleted
1254          * XXX pass the ifaddr struct instead if ifa->ifa_addr...
1255          */
1256         sctp_addr_change(ifa, cmd);
1257 #endif /* SCTP */
1258 #endif
1259         if (route_cb.any_count == 0)
1260                 return;
1261         for (pass = 1; pass < 3; pass++) {
1262                 bzero((caddr_t)&info, sizeof(info));
1263                 if ((cmd == RTM_ADD && pass == 1) ||
1264                     (cmd == RTM_DELETE && pass == 2)) {
1265                         struct ifa_msghdr *ifam;
1266                         int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
1267
1268                         info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1269                         info.rti_info[RTAX_IFP] = ifp->if_addr->ifa_addr;
1270                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1271                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1272                         if ((m = rt_msg1(ncmd, &info)) == NULL)
1273                                 continue;
1274                         ifam = mtod(m, struct ifa_msghdr *);
1275                         ifam->ifam_index = ifp->if_index;
1276                         ifam->ifam_metric = ifa->ifa_metric;
1277                         ifam->ifam_flags = ifa->ifa_flags;
1278                         ifam->ifam_addrs = info.rti_addrs;
1279                 }
1280                 if ((cmd == RTM_ADD && pass == 2) ||
1281                     (cmd == RTM_DELETE && pass == 1)) {
1282                         struct rt_msghdr *rtm;
1283
1284                         if (rt == NULL)
1285                                 continue;
1286                         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1287                         info.rti_info[RTAX_DST] = sa = rt_key(rt);
1288                         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1289                         if ((m = rt_msg1(cmd, &info)) == NULL)
1290                                 continue;
1291                         rtm = mtod(m, struct rt_msghdr *);
1292                         rtm->rtm_index = ifp->if_index;
1293                         rtm->rtm_flags |= rt->rt_flags;
1294                         rtm->rtm_errno = error;
1295                         rtm->rtm_addrs = info.rti_addrs;
1296                 }
1297                 if (fibnum != RTS_ALLFIBS) {
1298                         KASSERT(fibnum >= 0 && fibnum < rt_numfibs, ("%s: "
1299                             "fibnum out of range 0 <= %d < %d", __func__,
1300                              fibnum, rt_numfibs));
1301                         M_SETFIB(m, fibnum);
1302                         m->m_flags |= RTS_FILTER_FIB;
1303                 }
1304                 rt_dispatch(m, sa ? sa->sa_family : AF_UNSPEC);
1305         }
1306 }
1307
1308 void
1309 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
1310 {
1311
1312         rt_newaddrmsg_fib(cmd, ifa, error, rt, RTS_ALLFIBS);
1313 }
1314
1315 /*
1316  * This is the analogue to the rt_newaddrmsg which performs the same
1317  * function but for multicast group memberhips.  This is easier since
1318  * there is no route state to worry about.
1319  */
1320 void
1321 rt_newmaddrmsg(int cmd, struct ifmultiaddr *ifma)
1322 {
1323         struct rt_addrinfo info;
1324         struct mbuf *m = NULL;
1325         struct ifnet *ifp = ifma->ifma_ifp;
1326         struct ifma_msghdr *ifmam;
1327
1328         if (route_cb.any_count == 0)
1329                 return;
1330
1331         bzero((caddr_t)&info, sizeof(info));
1332         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1333         info.rti_info[RTAX_IFP] = ifp ? ifp->if_addr->ifa_addr : NULL;
1334         /*
1335          * If a link-layer address is present, present it as a ``gateway''
1336          * (similarly to how ARP entries, e.g., are presented).
1337          */
1338         info.rti_info[RTAX_GATEWAY] = ifma->ifma_lladdr;
1339         m = rt_msg1(cmd, &info);
1340         if (m == NULL)
1341                 return;
1342         ifmam = mtod(m, struct ifma_msghdr *);
1343         KASSERT(ifp != NULL, ("%s: link-layer multicast address w/o ifp\n",
1344             __func__));
1345         ifmam->ifmam_index = ifp->if_index;
1346         ifmam->ifmam_addrs = info.rti_addrs;
1347         rt_dispatch(m, ifma->ifma_addr ? ifma->ifma_addr->sa_family : AF_UNSPEC);
1348 }
1349
1350 static struct mbuf *
1351 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1352         struct rt_addrinfo *info)
1353 {
1354         struct if_announcemsghdr *ifan;
1355         struct mbuf *m;
1356
1357         if (route_cb.any_count == 0)
1358                 return NULL;
1359         bzero((caddr_t)info, sizeof(*info));
1360         m = rt_msg1(type, info);
1361         if (m != NULL) {
1362                 ifan = mtod(m, struct if_announcemsghdr *);
1363                 ifan->ifan_index = ifp->if_index;
1364                 strlcpy(ifan->ifan_name, ifp->if_xname,
1365                         sizeof(ifan->ifan_name));
1366                 ifan->ifan_what = what;
1367         }
1368         return m;
1369 }
1370
1371 /*
1372  * This is called to generate routing socket messages indicating
1373  * IEEE80211 wireless events.
1374  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1375  */
1376 void
1377 rt_ieee80211msg(struct ifnet *ifp, int what, void *data, size_t data_len)
1378 {
1379         struct mbuf *m;
1380         struct rt_addrinfo info;
1381
1382         m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1383         if (m != NULL) {
1384                 /*
1385                  * Append the ieee80211 data.  Try to stick it in the
1386                  * mbuf containing the ifannounce msg; otherwise allocate
1387                  * a new mbuf and append.
1388                  *
1389                  * NB: we assume m is a single mbuf.
1390                  */
1391                 if (data_len > M_TRAILINGSPACE(m)) {
1392                         struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1393                         if (n == NULL) {
1394                                 m_freem(m);
1395                                 return;
1396                         }
1397                         bcopy(data, mtod(n, void *), data_len);
1398                         n->m_len = data_len;
1399                         m->m_next = n;
1400                 } else if (data_len > 0) {
1401                         bcopy(data, mtod(m, u_int8_t *) + m->m_len, data_len);
1402                         m->m_len += data_len;
1403                 }
1404                 if (m->m_flags & M_PKTHDR)
1405                         m->m_pkthdr.len += data_len;
1406                 mtod(m, struct if_announcemsghdr *)->ifan_msglen += data_len;
1407                 rt_dispatch(m, AF_UNSPEC);
1408         }
1409 }
1410
1411 /*
1412  * This is called to generate routing socket messages indicating
1413  * network interface arrival and departure.
1414  */
1415 void
1416 rt_ifannouncemsg(struct ifnet *ifp, int what)
1417 {
1418         struct mbuf *m;
1419         struct rt_addrinfo info;
1420
1421         m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1422         if (m != NULL)
1423                 rt_dispatch(m, AF_UNSPEC);
1424 }
1425
1426 static void
1427 rt_dispatch(struct mbuf *m, sa_family_t saf)
1428 {
1429         struct m_tag *tag;
1430
1431         /*
1432          * Preserve the family from the sockaddr, if any, in an m_tag for
1433          * use when injecting the mbuf into the routing socket buffer from
1434          * the netisr.
1435          */
1436         if (saf != AF_UNSPEC) {
1437                 tag = m_tag_get(PACKET_TAG_RTSOCKFAM, sizeof(unsigned short),
1438                     M_NOWAIT);
1439                 if (tag == NULL) {
1440                         m_freem(m);
1441                         return;
1442                 }
1443                 *(unsigned short *)(tag + 1) = saf;
1444                 m_tag_prepend(m, tag);
1445         }
1446 #ifdef VIMAGE
1447         if (V_loif)
1448                 m->m_pkthdr.rcvif = V_loif;
1449         else {
1450                 m_freem(m);
1451                 return;
1452         }
1453 #endif
1454         netisr_queue(NETISR_ROUTE, m);  /* mbuf is free'd on failure. */
1455 }
1456
1457 /*
1458  * This is used in dumping the kernel table via sysctl().
1459  */
1460 static int
1461 sysctl_dumpentry(struct radix_node *rn, void *vw)
1462 {
1463         struct walkarg *w = vw;
1464         struct rtentry *rt = (struct rtentry *)rn;
1465         int error = 0, size;
1466         struct rt_addrinfo info;
1467
1468         if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1469                 return 0;
1470         if ((rt->rt_flags & RTF_HOST) == 0
1471             ? jailed_without_vnet(w->w_req->td->td_ucred)
1472             : prison_if(w->w_req->td->td_ucred, rt_key(rt)) != 0)
1473                 return (0);
1474         bzero((caddr_t)&info, sizeof(info));
1475         info.rti_info[RTAX_DST] = rt_key(rt);
1476         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1477         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1478         info.rti_info[RTAX_GENMASK] = 0;
1479         if (rt->rt_ifp) {
1480                 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr;
1481                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
1482                 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1483                         info.rti_info[RTAX_BRD] = rt->rt_ifa->ifa_dstaddr;
1484         }
1485         size = rt_msg2(RTM_GET, &info, NULL, w);
1486         if (w->w_req && w->w_tmem) {
1487                 struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
1488
1489                 rtm->rtm_flags = rt->rt_flags;
1490                 /*
1491                  * let's be honest about this being a retarded hack
1492                  */
1493                 rtm->rtm_fmask = rt->rt_rmx.rmx_pksent;
1494                 rt_getmetrics(&rt->rt_rmx, &rtm->rtm_rmx);
1495                 rtm->rtm_index = rt->rt_ifp->if_index;
1496                 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1497                 rtm->rtm_addrs = info.rti_addrs;
1498                 error = SYSCTL_OUT(w->w_req, (caddr_t)rtm, size);
1499                 return (error);
1500         }
1501         return (error);
1502 }
1503
1504 #ifdef COMPAT_FREEBSD32
1505 static void
1506 copy_ifdata32(struct if_data *src, struct if_data32 *dst)
1507 {
1508
1509         bzero(dst, sizeof(*dst));
1510         CP(*src, *dst, ifi_type);
1511         CP(*src, *dst, ifi_physical);
1512         CP(*src, *dst, ifi_addrlen);
1513         CP(*src, *dst, ifi_hdrlen);
1514         CP(*src, *dst, ifi_link_state);
1515         CP(*src, *dst, ifi_vhid);
1516         dst->ifi_datalen = sizeof(struct if_data32);
1517         CP(*src, *dst, ifi_mtu);
1518         CP(*src, *dst, ifi_metric);
1519         CP(*src, *dst, ifi_baudrate);
1520         CP(*src, *dst, ifi_ipackets);
1521         CP(*src, *dst, ifi_ierrors);
1522         CP(*src, *dst, ifi_opackets);
1523         CP(*src, *dst, ifi_oerrors);
1524         CP(*src, *dst, ifi_collisions);
1525         CP(*src, *dst, ifi_ibytes);
1526         CP(*src, *dst, ifi_obytes);
1527         CP(*src, *dst, ifi_imcasts);
1528         CP(*src, *dst, ifi_omcasts);
1529         CP(*src, *dst, ifi_iqdrops);
1530         CP(*src, *dst, ifi_noproto);
1531         CP(*src, *dst, ifi_hwassist);
1532         CP(*src, *dst, ifi_epoch);
1533         TV_CP(*src, *dst, ifi_lastchange);
1534 }
1535 #endif
1536
1537 static int
1538 sysctl_iflist(int af, struct walkarg *w)
1539 {
1540         struct ifnet *ifp;
1541         struct ifaddr *ifa;
1542         struct rt_addrinfo info;
1543         int len, error = 0;
1544
1545         bzero((caddr_t)&info, sizeof(info));
1546         IFNET_RLOCK();
1547         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1548                 if (w->w_arg && w->w_arg != ifp->if_index)
1549                         continue;
1550                 IF_ADDR_LOCK(ifp);
1551                 ifa = ifp->if_addr;
1552                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
1553                 len = rt_msg2(RTM_IFINFO, &info, NULL, w);
1554                 info.rti_info[RTAX_IFP] = NULL;
1555                 if (w->w_req && w->w_tmem) {
1556                         struct if_msghdr *ifm;
1557
1558 #ifdef COMPAT_FREEBSD32
1559                         if (w->w_req->flags & SCTL_MASK32) {
1560                                 struct if_msghdr32 *ifm32;
1561
1562                                 ifm32 = (struct if_msghdr32 *)w->w_tmem;
1563                                 ifm32->ifm_index = ifp->if_index;
1564                                 ifm32->ifm_flags = ifp->if_flags |
1565                                     ifp->if_drv_flags;
1566                                 copy_ifdata32(&ifp->if_data, &ifm32->ifm_data);
1567                                 if (carp_get_vhid_p != NULL)
1568                                         ifm32->ifm_data.ifi_vhid =
1569                                             (*carp_get_vhid_p)(ifa);
1570                                 ifm32->ifm_addrs = info.rti_addrs;
1571                                 error = SYSCTL_OUT(w->w_req, (caddr_t)ifm32,
1572                                     len);
1573                                 goto sysctl_out;
1574                         }
1575 #endif
1576                         ifm = (struct if_msghdr *)w->w_tmem;
1577                         ifm->ifm_index = ifp->if_index;
1578                         ifm->ifm_flags = ifp->if_flags | ifp->if_drv_flags;
1579                         ifm->ifm_data = ifp->if_data;
1580                         if (carp_get_vhid_p != NULL)
1581                                 ifm->ifm_data.ifi_vhid =
1582                                     (*carp_get_vhid_p)(ifa);
1583                         ifm->ifm_addrs = info.rti_addrs;
1584                         error = SYSCTL_OUT(w->w_req, (caddr_t)ifm, len);
1585 #ifdef COMPAT_FREEBSD32
1586                 sysctl_out:
1587 #endif
1588                         if (error)
1589                                 goto done;
1590                 }
1591                 while ((ifa = TAILQ_NEXT(ifa, ifa_link)) != NULL) {
1592                         if (af && af != ifa->ifa_addr->sa_family)
1593                                 continue;
1594                         if (prison_if(w->w_req->td->td_ucred,
1595                             ifa->ifa_addr) != 0)
1596                                 continue;
1597                         info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1598                         info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1599                         info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1600                         len = rt_msg2(RTM_NEWADDR, &info, NULL, w);
1601                         if (w->w_req && w->w_tmem) {
1602                                 struct ifa_msghdr *ifam;
1603
1604                                 ifam = (struct ifa_msghdr *)w->w_tmem;
1605                                 ifam->ifam_index = ifa->ifa_ifp->if_index;
1606                                 ifam->ifam_flags = ifa->ifa_flags;
1607                                 ifam->ifam_metric = ifa->ifa_metric;
1608                                 ifam->ifam_addrs = info.rti_addrs;
1609                                 if (carp_get_vhid_p != NULL)
1610                                         ifam->ifam_data.ifi_vhid =
1611                                             (*carp_get_vhid_p)(ifa);
1612                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1613                                 if (error)
1614                                         goto done;
1615                         }
1616                 }
1617                 IF_ADDR_UNLOCK(ifp);
1618                 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1619                         info.rti_info[RTAX_BRD] = NULL;
1620         }
1621 done:
1622         if (ifp != NULL)
1623                 IF_ADDR_UNLOCK(ifp);
1624         IFNET_RUNLOCK();
1625         return (error);
1626 }
1627
1628 static int
1629 sysctl_ifmalist(int af, struct walkarg *w)
1630 {
1631         struct ifnet *ifp;
1632         struct ifmultiaddr *ifma;
1633         struct  rt_addrinfo info;
1634         int     len, error = 0;
1635         struct ifaddr *ifa;
1636
1637         bzero((caddr_t)&info, sizeof(info));
1638         IFNET_RLOCK();
1639         TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1640                 if (w->w_arg && w->w_arg != ifp->if_index)
1641                         continue;
1642                 ifa = ifp->if_addr;
1643                 info.rti_info[RTAX_IFP] = ifa ? ifa->ifa_addr : NULL;
1644                 IF_ADDR_LOCK(ifp);
1645                 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1646                         if (af && af != ifma->ifma_addr->sa_family)
1647                                 continue;
1648                         if (prison_if(w->w_req->td->td_ucred,
1649                             ifma->ifma_addr) != 0)
1650                                 continue;
1651                         info.rti_info[RTAX_IFA] = ifma->ifma_addr;
1652                         info.rti_info[RTAX_GATEWAY] =
1653                             (ifma->ifma_addr->sa_family != AF_LINK) ?
1654                             ifma->ifma_lladdr : NULL;
1655                         len = rt_msg2(RTM_NEWMADDR, &info, NULL, w);
1656                         if (w->w_req && w->w_tmem) {
1657                                 struct ifma_msghdr *ifmam;
1658
1659                                 ifmam = (struct ifma_msghdr *)w->w_tmem;
1660                                 ifmam->ifmam_index = ifma->ifma_ifp->if_index;
1661                                 ifmam->ifmam_flags = 0;
1662                                 ifmam->ifmam_addrs = info.rti_addrs;
1663                                 error = SYSCTL_OUT(w->w_req, w->w_tmem, len);
1664                                 if (error) {
1665                                         IF_ADDR_UNLOCK(ifp);
1666                                         goto done;
1667                                 }
1668                         }
1669                 }
1670                 IF_ADDR_UNLOCK(ifp);
1671         }
1672 done:
1673         IFNET_RUNLOCK();
1674         return (error);
1675 }
1676
1677 static int
1678 sysctl_rtsock(SYSCTL_HANDLER_ARGS)
1679 {
1680         int     *name = (int *)arg1;
1681         u_int   namelen = arg2;
1682         struct radix_node_head *rnh = NULL; /* silence compiler. */
1683         int     i, lim, error = EINVAL;
1684         u_char  af;
1685         struct  walkarg w;
1686
1687         name ++;
1688         namelen--;
1689         if (req->newptr)
1690                 return (EPERM);
1691         if (namelen != 3)
1692                 return ((namelen < 3) ? EISDIR : ENOTDIR);
1693         af = name[0];
1694         if (af > AF_MAX)
1695                 return (EINVAL);
1696         bzero(&w, sizeof(w));
1697         w.w_op = name[1];
1698         w.w_arg = name[2];
1699         w.w_req = req;
1700
1701         error = sysctl_wire_old_buffer(req, 0);
1702         if (error)
1703                 return (error);
1704         switch (w.w_op) {
1705
1706         case NET_RT_DUMP:
1707         case NET_RT_FLAGS:
1708                 if (af == 0) {                  /* dump all tables */
1709                         i = 1;
1710                         lim = AF_MAX;
1711                 } else                          /* dump only one table */
1712                         i = lim = af;
1713
1714                 /*
1715                  * take care of llinfo entries, the caller must
1716                  * specify an AF
1717                  */
1718                 if (w.w_op == NET_RT_FLAGS &&
1719                     (w.w_arg == 0 || w.w_arg & RTF_LLINFO)) {
1720                         if (af != 0)
1721                                 error = lltable_sysctl_dumparp(af, w.w_req);
1722                         else
1723                                 error = EINVAL;
1724                         break;
1725                 }
1726                 /*
1727                  * take care of routing entries
1728                  */
1729                 for (error = 0; error == 0 && i <= lim; i++) {
1730                         rnh = rt_tables_get_rnh(req->td->td_proc->p_fibnum, i);
1731                         if (rnh != NULL) {
1732                                 RADIX_NODE_HEAD_LOCK(rnh); 
1733                                 error = rnh->rnh_walktree(rnh,
1734                                     sysctl_dumpentry, &w);
1735                                 RADIX_NODE_HEAD_UNLOCK(rnh);
1736                         } else if (af != 0)
1737                                 error = EAFNOSUPPORT;
1738                 }
1739                 break;
1740
1741         case NET_RT_IFLIST:
1742                 error = sysctl_iflist(af, &w);
1743                 break;
1744
1745         case NET_RT_IFMALIST:
1746                 error = sysctl_ifmalist(af, &w);
1747                 break;
1748         }
1749         if (w.w_tmem)
1750                 free(w.w_tmem, M_RTABLE);
1751         return (error);
1752 }
1753
1754 static SYSCTL_NODE(_net, PF_ROUTE, routetable, CTLFLAG_RD, sysctl_rtsock, "");
1755
1756 /*
1757  * Definitions of protocols supported in the ROUTE domain.
1758  */
1759
1760 static struct domain routedomain;               /* or at least forward */
1761
1762 static struct protosw routesw[] = {
1763 {
1764         .pr_type =              SOCK_RAW,
1765         .pr_domain =            &routedomain,
1766         .pr_flags =             PR_ATOMIC|PR_ADDR,
1767         .pr_output =            route_output,
1768         .pr_ctlinput =          raw_ctlinput,
1769         .pr_init =              raw_init,
1770         .pr_usrreqs =           &route_usrreqs
1771 }
1772 };
1773
1774 static struct domain routedomain = {
1775         .dom_family =           PF_ROUTE,
1776         .dom_name =              "route",
1777         .dom_protosw =          routesw,
1778         .dom_protoswNPROTOSW =  &routesw[sizeof(routesw)/sizeof(routesw[0])]
1779 };
1780
1781 VNET_DOMAIN_SET(route);