rtsock.c revision 1.233 1 /* $NetBSD: rtsock.c,v 1.233 2017/12/14 05:47:45 ozaki-r Exp $ */
2
3 /*
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32 /*
33 * Copyright (c) 1988, 1991, 1993
34 * The Regents of the University of California. All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 3. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)rtsock.c 8.7 (Berkeley) 10/12/95
61 */
62
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: rtsock.c,v 1.233 2017/12/14 05:47:45 ozaki-r Exp $");
65
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_mpls.h"
69 #include "opt_compat_netbsd.h"
70 #include "opt_sctp.h"
71 #include "opt_net_mpsafe.h"
72 #endif
73
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/proc.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/domain.h>
80 #include <sys/protosw.h>
81 #include <sys/sysctl.h>
82 #include <sys/kauth.h>
83 #include <sys/kmem.h>
84 #include <sys/intr.h>
85
86 #include <net/if.h>
87 #include <net/if_llatbl.h>
88 #include <net/if_types.h>
89 #include <net/route.h>
90 #include <net/raw_cb.h>
91
92 #include <netinet/in_var.h>
93 #include <netinet/if_inarp.h>
94
95 #include <netmpls/mpls.h>
96
97 #ifdef SCTP
98 extern void sctp_add_ip_address(struct ifaddr *);
99 extern void sctp_delete_ip_address(struct ifaddr *);
100 #endif
101
102 #if defined(COMPAT_14) || defined(COMPAT_50) || defined(COMPAT_70)
103 #include <compat/net/if.h>
104 #include <compat/net/route.h>
105 #endif
106 #ifdef COMPAT_RTSOCK
107 #define RTM_XVERSION RTM_OVERSION
108 #define RTM_XNEWADDR RTM_ONEWADDR
109 #define RTM_XDELADDR RTM_ODELADDR
110 #define RTM_XCHGADDR RTM_OCHGADDR
111 #define RT_XADVANCE(a,b) RT_OADVANCE(a,b)
112 #define RT_XROUNDUP(n) RT_OROUNDUP(n)
113 #define PF_XROUTE PF_OROUTE
114 #define rt_xmsghdr rt_msghdr50
115 #define if_xmsghdr if_msghdr /* if_msghdr50 is for RTM_OIFINFO */
116 #define ifa_xmsghdr ifa_msghdr50
117 #define if_xannouncemsghdr if_announcemsghdr50
118 #define COMPATNAME(x) compat_50_ ## x
119 #define DOMAINNAME "oroute"
120 CTASSERT(sizeof(struct ifa_xmsghdr) == 20);
121 DOMAIN_DEFINE(compat_50_routedomain); /* forward declare and add to link set */
122 #undef COMPAT_70
123 #else /* COMPAT_RTSOCK */
124 #define RTM_XVERSION RTM_VERSION
125 #define RTM_XNEWADDR RTM_NEWADDR
126 #define RTM_XDELADDR RTM_DELADDR
127 #define RTM_XCHGADDR RTM_CHGADDR
128 #define RT_XADVANCE(a,b) RT_ADVANCE(a,b)
129 #define RT_XROUNDUP(n) RT_ROUNDUP(n)
130 #define PF_XROUTE PF_ROUTE
131 #define rt_xmsghdr rt_msghdr
132 #define if_xmsghdr if_msghdr
133 #define ifa_xmsghdr ifa_msghdr
134 #define if_xannouncemsghdr if_announcemsghdr
135 #define COMPATNAME(x) x
136 #define DOMAINNAME "route"
137 CTASSERT(sizeof(struct ifa_xmsghdr) == 32);
138 #ifdef COMPAT_50
139 #define COMPATCALL(name, args) compat_50_ ## name args
140 #endif
141 DOMAIN_DEFINE(routedomain); /* forward declare and add to link set */
142 #undef COMPAT_50
143 #undef COMPAT_14
144 #endif /* COMPAT_RTSOCK */
145
146 #ifndef COMPATCALL
147 #define COMPATCALL(name, args) do { } while (/*CONSTCOND*/ 0)
148 #endif
149
150 #ifdef RTSOCK_DEBUG
151 #define RT_IN_PRINT(info, b, a) (in_print((b), sizeof(b), \
152 &((const struct sockaddr_in *)(info)->rti_info[(a)])->sin_addr), (b))
153 #endif /* RTSOCK_DEBUG */
154
155 struct route_info COMPATNAME(route_info) = {
156 .ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, },
157 .ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, },
158 .ri_maxqlen = IFQ_MAXLEN,
159 };
160
161 #define PRESERVED_RTF (RTF_UP | RTF_GATEWAY | RTF_HOST | RTF_DONE | RTF_MASK)
162
163 static void COMPATNAME(route_init)(void);
164 static int COMPATNAME(route_output)(struct mbuf *, struct socket *);
165
166 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
167 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
168 struct rt_addrinfo *);
169 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
170 static void rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *);
171 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *);
172 static void sysctl_net_route_setup(struct sysctllog **);
173 static int sysctl_dumpentry(struct rtentry *, void *);
174 static int sysctl_iflist(int, struct rt_walkarg *, int);
175 static int sysctl_rtable(SYSCTLFN_PROTO);
176 static void rt_adjustcount(int, int);
177
178 static const struct protosw COMPATNAME(route_protosw)[];
179
180 struct routecb {
181 struct rawcb rocb_rcb;
182 unsigned int rocb_msgfilter;
183 #define RTMSGFILTER(m) (1U << (m))
184 };
185 #define sotoroutecb(so) ((struct routecb *)(so)->so_pcb)
186
187 static struct rawcbhead rt_rawcb;
188 #ifdef NET_MPSAFE
189 static kmutex_t *rt_so_mtx;
190 #endif
191
192 static void
193 rt_adjustcount(int af, int cnt)
194 {
195 struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
196
197 cb->any_count += cnt;
198
199 switch (af) {
200 case AF_INET:
201 cb->ip_count += cnt;
202 return;
203 #ifdef INET6
204 case AF_INET6:
205 cb->ip6_count += cnt;
206 return;
207 #endif
208 case AF_MPLS:
209 cb->mpls_count += cnt;
210 return;
211 }
212 }
213
214 static int
215 COMPATNAME(route_filter)(struct mbuf *m, struct sockproto *proto,
216 struct rawcb *rp)
217 {
218 struct routecb *rop = (struct routecb *)rp;
219 struct rt_xmsghdr *rtm;
220
221 KASSERT(m != NULL);
222 KASSERT(proto != NULL);
223 KASSERT(rp != NULL);
224
225 /* Wrong family for this socket. */
226 if (proto->sp_family != PF_ROUTE)
227 return ENOPROTOOPT;
228
229 /* If no filter set, just return. */
230 if (rop->rocb_msgfilter == 0)
231 return 0;
232
233 /* Ensure we can access rtm_type */
234 if (m->m_len <
235 offsetof(struct rt_xmsghdr, rtm_type) + sizeof(rtm->rtm_type))
236 return EINVAL;
237
238 rtm = mtod(m, struct rt_xmsghdr *);
239 /* If the rtm type is filtered out, return a positive. */
240 if (!(rop->rocb_msgfilter & RTMSGFILTER(rtm->rtm_type)))
241 return EEXIST;
242
243 /* Passed the filter. */
244 return 0;
245 }
246
247 static void
248 rt_pr_init(void)
249 {
250
251 LIST_INIT(&rt_rawcb);
252 }
253
254 static int
255 COMPATNAME(route_attach)(struct socket *so, int proto)
256 {
257 struct rawcb *rp;
258 struct routecb *rop;
259 int s, error;
260
261 KASSERT(sotorawcb(so) == NULL);
262 rop = kmem_zalloc(sizeof(*rop), KM_SLEEP);
263 rp = &rop->rocb_rcb;
264 rp->rcb_len = sizeof(*rop);
265 so->so_pcb = rp;
266
267 s = splsoftnet();
268
269 #ifdef NET_MPSAFE
270 KASSERT(so->so_lock == NULL);
271 mutex_obj_hold(rt_so_mtx);
272 so->so_lock = rt_so_mtx;
273 solock(so);
274 #endif
275
276 if ((error = raw_attach(so, proto, &rt_rawcb)) == 0) {
277 rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
278 rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
279 rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
280 rp->rcb_filter = COMPATNAME(route_filter);
281 }
282 splx(s);
283
284 if (error) {
285 kmem_free(rop, sizeof(*rop));
286 so->so_pcb = NULL;
287 return error;
288 }
289
290 soisconnected(so);
291 so->so_options |= SO_USELOOPBACK;
292 KASSERT(solocked(so));
293
294 return error;
295 }
296
297 static void
298 COMPATNAME(route_detach)(struct socket *so)
299 {
300 struct rawcb *rp = sotorawcb(so);
301 int s;
302
303 KASSERT(rp != NULL);
304 KASSERT(solocked(so));
305
306 s = splsoftnet();
307 rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
308 raw_detach(so);
309 splx(s);
310 }
311
312 static int
313 COMPATNAME(route_accept)(struct socket *so, struct sockaddr *nam)
314 {
315 KASSERT(solocked(so));
316
317 panic("route_accept");
318
319 return EOPNOTSUPP;
320 }
321
322 static int
323 COMPATNAME(route_bind)(struct socket *so, struct sockaddr *nam, struct lwp *l)
324 {
325 KASSERT(solocked(so));
326
327 return EOPNOTSUPP;
328 }
329
330 static int
331 COMPATNAME(route_listen)(struct socket *so, struct lwp *l)
332 {
333 KASSERT(solocked(so));
334
335 return EOPNOTSUPP;
336 }
337
338 static int
339 COMPATNAME(route_connect)(struct socket *so, struct sockaddr *nam, struct lwp *l)
340 {
341 KASSERT(solocked(so));
342
343 return EOPNOTSUPP;
344 }
345
346 static int
347 COMPATNAME(route_connect2)(struct socket *so, struct socket *so2)
348 {
349 KASSERT(solocked(so));
350
351 return EOPNOTSUPP;
352 }
353
354 static int
355 COMPATNAME(route_disconnect)(struct socket *so)
356 {
357 struct rawcb *rp = sotorawcb(so);
358 int s;
359
360 KASSERT(solocked(so));
361 KASSERT(rp != NULL);
362
363 s = splsoftnet();
364 soisdisconnected(so);
365 raw_disconnect(rp);
366 splx(s);
367
368 return 0;
369 }
370
371 static int
372 COMPATNAME(route_shutdown)(struct socket *so)
373 {
374 int s;
375
376 KASSERT(solocked(so));
377
378 /*
379 * Mark the connection as being incapable of further input.
380 */
381 s = splsoftnet();
382 socantsendmore(so);
383 splx(s);
384 return 0;
385 }
386
387 static int
388 COMPATNAME(route_abort)(struct socket *so)
389 {
390 KASSERT(solocked(so));
391
392 panic("route_abort");
393
394 return EOPNOTSUPP;
395 }
396
397 static int
398 COMPATNAME(route_ioctl)(struct socket *so, u_long cmd, void *nam,
399 struct ifnet * ifp)
400 {
401 return EOPNOTSUPP;
402 }
403
404 static int
405 COMPATNAME(route_stat)(struct socket *so, struct stat *ub)
406 {
407 KASSERT(solocked(so));
408
409 return 0;
410 }
411
412 static int
413 COMPATNAME(route_peeraddr)(struct socket *so, struct sockaddr *nam)
414 {
415 struct rawcb *rp = sotorawcb(so);
416
417 KASSERT(solocked(so));
418 KASSERT(rp != NULL);
419 KASSERT(nam != NULL);
420
421 if (rp->rcb_faddr == NULL)
422 return ENOTCONN;
423
424 raw_setpeeraddr(rp, nam);
425 return 0;
426 }
427
428 static int
429 COMPATNAME(route_sockaddr)(struct socket *so, struct sockaddr *nam)
430 {
431 struct rawcb *rp = sotorawcb(so);
432
433 KASSERT(solocked(so));
434 KASSERT(rp != NULL);
435 KASSERT(nam != NULL);
436
437 if (rp->rcb_faddr == NULL)
438 return ENOTCONN;
439
440 raw_setsockaddr(rp, nam);
441 return 0;
442 }
443
444 static int
445 COMPATNAME(route_rcvd)(struct socket *so, int flags, struct lwp *l)
446 {
447 KASSERT(solocked(so));
448
449 return EOPNOTSUPP;
450 }
451
452 static int
453 COMPATNAME(route_recvoob)(struct socket *so, struct mbuf *m, int flags)
454 {
455 KASSERT(solocked(so));
456
457 return EOPNOTSUPP;
458 }
459
460 static int
461 COMPATNAME(route_send)(struct socket *so, struct mbuf *m,
462 struct sockaddr *nam, struct mbuf *control, struct lwp *l)
463 {
464 int error = 0;
465 int s;
466
467 KASSERT(solocked(so));
468 KASSERT(so->so_proto == &COMPATNAME(route_protosw)[0]);
469
470 s = splsoftnet();
471 error = raw_send(so, m, nam, control, l, &COMPATNAME(route_output));
472 splx(s);
473
474 return error;
475 }
476
477 static int
478 COMPATNAME(route_sendoob)(struct socket *so, struct mbuf *m,
479 struct mbuf *control)
480 {
481 KASSERT(solocked(so));
482
483 m_freem(m);
484 m_freem(control);
485
486 return EOPNOTSUPP;
487 }
488 static int
489 COMPATNAME(route_purgeif)(struct socket *so, struct ifnet *ifp)
490 {
491
492 panic("route_purgeif");
493
494 return EOPNOTSUPP;
495 }
496
497 #if defined(INET) || defined(INET6)
498 static int __noinline
499 route_get_sdl_index(struct rt_addrinfo *info, int *sdl_index)
500 {
501 struct rtentry *nrt;
502 int error;
503
504 error = rtrequest1(RTM_GET, info, &nrt);
505 if (error != 0)
506 return error;
507 /*
508 * nrt->rt_ifp->if_index may not be correct
509 * due to changing to ifplo0.
510 */
511 *sdl_index = satosdl(nrt->rt_gateway)->sdl_index;
512 rt_unref(nrt);
513
514 return 0;
515 }
516 #endif
517
518 static void __noinline
519 route_get_sdl(const struct ifnet *ifp, const struct sockaddr *dst,
520 struct sockaddr_dl *sdl, int *flags)
521 {
522 struct llentry *la;
523
524 KASSERT(ifp != NULL);
525
526 IF_AFDATA_RLOCK(ifp);
527 switch (dst->sa_family) {
528 case AF_INET:
529 la = lla_lookup(LLTABLE(ifp), 0, dst);
530 break;
531 case AF_INET6:
532 la = lla_lookup(LLTABLE6(ifp), 0, dst);
533 break;
534 default:
535 la = NULL;
536 KASSERTMSG(0, "Invalid AF=%d\n", dst->sa_family);
537 break;
538 }
539 IF_AFDATA_RUNLOCK(ifp);
540
541 void *a = (LLE_IS_VALID(la) && (la->la_flags & LLE_VALID) == LLE_VALID)
542 ? &la->ll_addr : NULL;
543
544 a = sockaddr_dl_init(sdl, sizeof(*sdl), ifp->if_index, ifp->if_type,
545 NULL, 0, a, ifp->if_addrlen);
546 KASSERT(a != NULL);
547
548 if (la != NULL) {
549 *flags = la->la_flags;
550 LLE_RUNLOCK(la);
551 }
552 }
553
554 static int __noinline
555 route_output_report(struct rtentry *rt, struct rt_addrinfo *info,
556 struct rt_xmsghdr *rtm, struct rt_xmsghdr **new_rtm)
557 {
558 int len;
559
560 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
561 const struct ifaddr *rtifa;
562 const struct ifnet *ifp = rt->rt_ifp;
563
564 info->rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
565 /* rtifa used to be simply rt->rt_ifa.
566 * If rt->rt_ifa != NULL, then
567 * rt_get_ifa() != NULL. So this
568 * ought to still be safe. --dyoung
569 */
570 rtifa = rt_get_ifa(rt);
571 info->rti_info[RTAX_IFA] = rtifa->ifa_addr;
572 #ifdef RTSOCK_DEBUG
573 if (info->rti_info[RTAX_IFA]->sa_family == AF_INET) {
574 char ibuf[INET_ADDRSTRLEN];
575 char abuf[INET_ADDRSTRLEN];
576 printf("%s: copying out RTAX_IFA %s "
577 "for info->rti_info[RTAX_DST] %s "
578 "ifa_getifa %p ifa_seqno %p\n",
579 __func__,
580 RT_IN_PRINT(info, ibuf, RTAX_IFA),
581 RT_IN_PRINT(info, abuf, RTAX_DST),
582 (void *)rtifa->ifa_getifa,
583 rtifa->ifa_seqno);
584 }
585 #endif /* RTSOCK_DEBUG */
586 if (ifp->if_flags & IFF_POINTOPOINT)
587 info->rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
588 else
589 info->rti_info[RTAX_BRD] = NULL;
590 rtm->rtm_index = ifp->if_index;
591 }
592 (void)rt_msg2(rtm->rtm_type, info, NULL, NULL, &len);
593 if (len > rtm->rtm_msglen) {
594 struct rt_xmsghdr *old_rtm = rtm;
595 R_Malloc(*new_rtm, struct rt_xmsghdr *, len);
596 if (*new_rtm == NULL)
597 return ENOBUFS;
598 (void)memcpy(*new_rtm, old_rtm, old_rtm->rtm_msglen);
599 rtm = *new_rtm;
600 }
601 (void)rt_msg2(rtm->rtm_type, info, rtm, NULL, 0);
602 rtm->rtm_flags = rt->rt_flags;
603 rtm_setmetrics(rt, rtm);
604 rtm->rtm_addrs = info->rti_addrs;
605
606 return 0;
607 }
608
609 static struct ifaddr * __noinline
610 route_output_get_ifa(const struct rt_addrinfo info, const struct rtentry *rt,
611 struct ifnet **ifp, struct psref *psref_ifp, struct psref *psref)
612 {
613 struct ifaddr *ifa = NULL;
614
615 *ifp = NULL;
616 if (info.rti_info[RTAX_IFP] != NULL) {
617 ifa = ifa_ifwithnet_psref(info.rti_info[RTAX_IFP], psref);
618 if (ifa == NULL)
619 goto next;
620 *ifp = ifa->ifa_ifp;
621 if_acquire(*ifp, psref_ifp);
622 if (info.rti_info[RTAX_IFA] == NULL &&
623 info.rti_info[RTAX_GATEWAY] == NULL)
624 goto next;
625 if (info.rti_info[RTAX_IFA] == NULL) {
626 /* route change <dst> <gw> -ifp <if> */
627 ifa = ifaof_ifpforaddr_psref(info.rti_info[RTAX_GATEWAY],
628 *ifp, psref);
629 } else {
630 /* route change <dst> -ifp <if> -ifa <addr> */
631 ifa = ifa_ifwithaddr_psref(info.rti_info[RTAX_IFA], psref);
632 if (ifa != NULL)
633 goto out;
634 ifa = ifaof_ifpforaddr_psref(info.rti_info[RTAX_IFA],
635 *ifp, psref);
636 }
637 goto out;
638 }
639 next:
640 if (info.rti_info[RTAX_IFA] != NULL) {
641 /* route change <dst> <gw> -ifa <addr> */
642 ifa = ifa_ifwithaddr_psref(info.rti_info[RTAX_IFA], psref);
643 if (ifa != NULL)
644 goto out;
645 }
646 if (info.rti_info[RTAX_GATEWAY] != NULL) {
647 /* route change <dst> <gw> */
648 ifa = ifa_ifwithroute_psref(rt->rt_flags, rt_getkey(rt),
649 info.rti_info[RTAX_GATEWAY], psref);
650 }
651 out:
652 if (ifa != NULL && *ifp == NULL) {
653 *ifp = ifa->ifa_ifp;
654 if_acquire(*ifp, psref_ifp);
655 }
656 if (ifa == NULL && *ifp != NULL) {
657 if_put(*ifp, psref_ifp);
658 *ifp = NULL;
659 }
660 return ifa;
661 }
662
663 static int __noinline
664 route_output_change(struct rtentry *rt, struct rt_addrinfo *info,
665 struct rt_xmsghdr *rtm)
666 {
667 int error = 0;
668 struct ifnet *ifp = NULL, *new_ifp;
669 struct ifaddr *ifa = NULL, *new_ifa;
670 struct psref psref_ifa, psref_new_ifa, psref_ifp, psref_new_ifp;
671 bool newgw, ifp_changed = false;
672
673 /*
674 * New gateway could require new ifaddr, ifp;
675 * flags may also be different; ifp may be specified
676 * by ll sockaddr when protocol address is ambiguous
677 */
678 newgw = info->rti_info[RTAX_GATEWAY] != NULL &&
679 sockaddr_cmp(info->rti_info[RTAX_GATEWAY], rt->rt_gateway) != 0;
680
681 if (newgw || info->rti_info[RTAX_IFP] != NULL ||
682 info->rti_info[RTAX_IFA] != NULL) {
683 ifp = rt_getifp(info, &psref_ifp);
684 /* info refers ifp so we need to keep a reference */
685 ifa = rt_getifa(info, &psref_ifa);
686 if (ifa == NULL) {
687 error = ENETUNREACH;
688 goto out;
689 }
690 }
691 if (newgw) {
692 error = rt_setgate(rt, info->rti_info[RTAX_GATEWAY]);
693 if (error != 0)
694 goto out;
695 }
696 if (info->rti_info[RTAX_TAG]) {
697 const struct sockaddr *tag;
698 tag = rt_settag(rt, info->rti_info[RTAX_TAG]);
699 if (tag == NULL) {
700 error = ENOBUFS;
701 goto out;
702 }
703 }
704 /*
705 * New gateway could require new ifaddr, ifp;
706 * flags may also be different; ifp may be specified
707 * by ll sockaddr when protocol address is ambiguous
708 */
709 new_ifa = route_output_get_ifa(*info, rt, &new_ifp, &psref_new_ifp,
710 &psref_new_ifa);
711 if (new_ifa != NULL) {
712 ifa_release(ifa, &psref_ifa);
713 ifa = new_ifa;
714 }
715 if (ifa) {
716 struct ifaddr *oifa = rt->rt_ifa;
717 if (oifa != ifa && !ifa_is_destroying(ifa) &&
718 new_ifp != NULL && !if_is_deactivated(new_ifp)) {
719 if (oifa && oifa->ifa_rtrequest)
720 oifa->ifa_rtrequest(RTM_DELETE, rt, info);
721 rt_replace_ifa(rt, ifa);
722 rt->rt_ifp = new_ifp;
723 ifp_changed = true;
724 }
725 if (new_ifa == NULL)
726 ifa_release(ifa, &psref_ifa);
727 }
728 ifa_release(new_ifa, &psref_new_ifa);
729 if (new_ifp && rt->rt_ifp != new_ifp && !if_is_deactivated(new_ifp)) {
730 rt->rt_ifp = new_ifp;
731 ifp_changed = true;
732 }
733 rt_setmetrics(rtm->rtm_inits, rtm, rt);
734 if (rt->rt_flags != info->rti_flags) {
735 rt->rt_flags = (info->rti_flags & ~PRESERVED_RTF) |
736 (rt->rt_flags & PRESERVED_RTF);
737 }
738 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
739 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
740 #if defined(INET) || defined(INET6)
741 if (ifp_changed && rt_mask(rt) != NULL)
742 lltable_prefix_free(rt_getkey(rt)->sa_family, rt_getkey(rt),
743 rt_mask(rt), 0);
744 #else
745 (void)ifp_changed; /* XXX gcc */
746 #endif
747 out:
748 if_put(new_ifp, &psref_new_ifp);
749 if_put(ifp, &psref_ifp);
750
751 return error;
752 }
753
754 /*ARGSUSED*/
755 int
756 COMPATNAME(route_output)(struct mbuf *m, struct socket *so)
757 {
758 struct sockproto proto = { .sp_family = PF_XROUTE, };
759 struct rt_xmsghdr *rtm = NULL;
760 struct rt_xmsghdr *old_rtm = NULL, *new_rtm = NULL;
761 struct rtentry *rt = NULL;
762 struct rtentry *saved_nrt = NULL;
763 struct rt_addrinfo info;
764 int len, error = 0;
765 sa_family_t family;
766 struct sockaddr_dl sdl;
767 int bound = curlwp_bind();
768 bool do_rt_free = false;
769 struct sockaddr_storage netmask;
770
771 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
772 if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
773 (m = m_pullup(m, sizeof(int32_t))) == NULL)) {
774 error = ENOBUFS;
775 goto out;
776 }
777 if ((m->m_flags & M_PKTHDR) == 0)
778 panic("%s", __func__);
779 len = m->m_pkthdr.len;
780 if (len < sizeof(*rtm) ||
781 len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
782 info.rti_info[RTAX_DST] = NULL;
783 senderr(EINVAL);
784 }
785 R_Malloc(rtm, struct rt_xmsghdr *, len);
786 if (rtm == NULL) {
787 info.rti_info[RTAX_DST] = NULL;
788 senderr(ENOBUFS);
789 }
790 m_copydata(m, 0, len, rtm);
791 if (rtm->rtm_version != RTM_XVERSION) {
792 info.rti_info[RTAX_DST] = NULL;
793 senderr(EPROTONOSUPPORT);
794 }
795 rtm->rtm_pid = curproc->p_pid;
796 memset(&info, 0, sizeof(info));
797 info.rti_addrs = rtm->rtm_addrs;
798 if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
799 &info)) {
800 senderr(EINVAL);
801 }
802 info.rti_flags = rtm->rtm_flags;
803 #ifdef RTSOCK_DEBUG
804 if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
805 char abuf[INET_ADDRSTRLEN];
806 printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
807 RT_IN_PRINT(&info, abuf, RTAX_DST));
808 }
809 #endif /* RTSOCK_DEBUG */
810 if (info.rti_info[RTAX_DST] == NULL ||
811 (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
812 senderr(EINVAL);
813 }
814 if (info.rti_info[RTAX_GATEWAY] != NULL &&
815 (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
816 senderr(EINVAL);
817 }
818
819 /*
820 * Verify that the caller has the appropriate privilege; RTM_GET
821 * is the only operation the non-superuser is allowed.
822 */
823 if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
824 0, rtm, NULL, NULL) != 0)
825 senderr(EACCES);
826
827 /*
828 * route(8) passes a sockaddr truncated with prefixlen.
829 * The kernel doesn't expect such sockaddr and need to
830 * use a buffer that is big enough for the sockaddr expected
831 * (padded with 0's). We keep the original length of the sockaddr.
832 */
833 if (info.rti_info[RTAX_NETMASK]) {
834 /*
835 * Use the family of RTAX_DST, because RTAX_NETMASK
836 * can have a zero family if it comes from the radix
837 * tree via rt_mask().
838 */
839 socklen_t sa_len = sockaddr_getsize_by_family(
840 info.rti_info[RTAX_DST]->sa_family);
841 socklen_t masklen = sockaddr_getlen(
842 info.rti_info[RTAX_NETMASK]);
843 if (sa_len != 0 && sa_len > masklen) {
844 KASSERT(sa_len <= sizeof(netmask));
845 memcpy(&netmask, info.rti_info[RTAX_NETMASK], masklen);
846 memset((char *)&netmask + masklen, 0, sa_len - masklen);
847 info.rti_info[RTAX_NETMASK] = sstocsa(&netmask);
848 }
849 }
850
851 switch (rtm->rtm_type) {
852
853 case RTM_ADD:
854 if (info.rti_info[RTAX_GATEWAY] == NULL) {
855 senderr(EINVAL);
856 }
857 #if defined(INET) || defined(INET6)
858 /* support for new ARP/NDP code with keeping backcompat */
859 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) {
860 const struct sockaddr_dl *sdlp =
861 satocsdl(info.rti_info[RTAX_GATEWAY]);
862
863 /* Allow routing requests by interface index */
864 if (sdlp->sdl_nlen == 0 && sdlp->sdl_alen == 0
865 && sdlp->sdl_slen == 0)
866 goto fallback;
867 /*
868 * Old arp binaries don't set the sdl_index
869 * so we have to complement it.
870 */
871 int sdl_index = sdlp->sdl_index;
872 if (sdl_index == 0) {
873 error = route_get_sdl_index(&info, &sdl_index);
874 if (error != 0)
875 goto fallback;
876 } else if (
877 info.rti_info[RTAX_DST]->sa_family == AF_INET) {
878 /*
879 * XXX workaround for SIN_PROXY case; proxy arp
880 * entry should be in an interface that has
881 * a network route including the destination,
882 * not a local (link) route that may not be a
883 * desired place, for example a tap.
884 */
885 const struct sockaddr_inarp *sina =
886 (const struct sockaddr_inarp *)
887 info.rti_info[RTAX_DST];
888 if (sina->sin_other & SIN_PROXY) {
889 error = route_get_sdl_index(&info,
890 &sdl_index);
891 if (error != 0)
892 goto fallback;
893 }
894 }
895 error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
896 rtm->rtm_rmx.rmx_expire, &info, sdl_index);
897 break;
898 }
899 fallback:
900 #endif /* defined(INET) || defined(INET6) */
901 error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
902 if (error == 0) {
903 rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
904 rt_unref(saved_nrt);
905 }
906 break;
907
908 case RTM_DELETE:
909 #if defined(INET) || defined(INET6)
910 /* support for new ARP/NDP code */
911 if (info.rti_info[RTAX_GATEWAY] &&
912 (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
913 (rtm->rtm_flags & RTF_LLDATA) != 0) {
914 const struct sockaddr_dl *sdlp =
915 satocsdl(info.rti_info[RTAX_GATEWAY]);
916 error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
917 rtm->rtm_rmx.rmx_expire, &info, sdlp->sdl_index);
918 rtm->rtm_flags &= ~RTF_UP;
919 break;
920 }
921 #endif
922 error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
923 if (error != 0)
924 break;
925
926 rt = saved_nrt;
927 do_rt_free = true;
928 info.rti_info[RTAX_DST] = rt_getkey(rt);
929 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
930 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
931 info.rti_info[RTAX_TAG] = rt_gettag(rt);
932 error = route_output_report(rt, &info, rtm, &new_rtm);
933 if (error)
934 senderr(error);
935 if (new_rtm != NULL) {
936 old_rtm = rtm;
937 rtm = new_rtm;
938 }
939 break;
940
941 case RTM_GET:
942 case RTM_CHANGE:
943 case RTM_LOCK:
944 /* XXX This will mask info.rti_info[RTAX_DST] with
945 * info.rti_info[RTAX_NETMASK] before
946 * searching. It did not used to do that. --dyoung
947 */
948 rt = NULL;
949 error = rtrequest1(RTM_GET, &info, &rt);
950 if (error != 0)
951 senderr(error);
952 if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
953 if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
954 info.rti_info[RTAX_DST]->sa_len) != 0)
955 senderr(ESRCH);
956 if (info.rti_info[RTAX_NETMASK] == NULL &&
957 rt_mask(rt) != NULL)
958 senderr(ETOOMANYREFS);
959 }
960
961 /*
962 * XXX if arp/ndp requests an L2 entry, we have to obtain
963 * it from lltable while for the route command we have to
964 * return a route as it is. How to distinguish them?
965 * For newer arp/ndp, RTF_LLDATA flag set by arp/ndp
966 * indicates an L2 entry is requested. For old arp/ndp
967 * binaries, we check RTF_UP flag is NOT set; it works
968 * by the fact that arp/ndp don't set it while the route
969 * command sets it.
970 */
971 if (((rtm->rtm_flags & RTF_LLDATA) != 0 ||
972 (rtm->rtm_flags & RTF_UP) == 0) &&
973 rtm->rtm_type == RTM_GET &&
974 sockaddr_cmp(rt_getkey(rt), info.rti_info[RTAX_DST]) != 0) {
975 int ll_flags = 0;
976 route_get_sdl(rt->rt_ifp, info.rti_info[RTAX_DST], &sdl,
977 &ll_flags);
978 info.rti_info[RTAX_GATEWAY] = sstocsa(&sdl);
979 error = route_output_report(rt, &info, rtm, &new_rtm);
980 if (error)
981 senderr(error);
982 if (new_rtm != NULL) {
983 old_rtm = rtm;
984 rtm = new_rtm;
985 }
986 rtm->rtm_flags |= RTF_LLDATA;
987 rtm->rtm_flags &= ~RTF_CONNECTED;
988 rtm->rtm_flags |= (ll_flags & LLE_STATIC) ? RTF_STATIC : 0;
989 break;
990 }
991
992 switch (rtm->rtm_type) {
993 case RTM_GET:
994 info.rti_info[RTAX_DST] = rt_getkey(rt);
995 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
996 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
997 info.rti_info[RTAX_TAG] = rt_gettag(rt);
998 error = route_output_report(rt, &info, rtm, &new_rtm);
999 if (error)
1000 senderr(error);
1001 if (new_rtm != NULL) {
1002 old_rtm = rtm;
1003 rtm = new_rtm;
1004 }
1005 break;
1006
1007 case RTM_CHANGE:
1008 #ifdef NET_MPSAFE
1009 error = rt_update_prepare(rt);
1010 if (error == 0) {
1011 error = route_output_change(rt, &info, rtm);
1012 rt_update_finish(rt);
1013 }
1014 #else
1015 error = route_output_change(rt, &info, rtm);
1016 #endif
1017 if (error != 0)
1018 goto flush;
1019 /*FALLTHROUGH*/
1020 case RTM_LOCK:
1021 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
1022 rt->rt_rmx.rmx_locks |=
1023 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
1024 break;
1025 }
1026 break;
1027
1028 default:
1029 senderr(EOPNOTSUPP);
1030 }
1031
1032 flush:
1033 if (rtm) {
1034 if (error)
1035 rtm->rtm_errno = error;
1036 else
1037 rtm->rtm_flags |= RTF_DONE;
1038 }
1039 family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
1040 0;
1041 /* We cannot free old_rtm until we have stopped using the
1042 * pointers in info, some of which may point to sockaddrs
1043 * in old_rtm.
1044 */
1045 if (old_rtm != NULL)
1046 Free(old_rtm);
1047 if (rt) {
1048 if (do_rt_free)
1049 rt_free(rt);
1050 else
1051 rt_unref(rt);
1052 }
1053 {
1054 struct rawcb *rp = NULL;
1055 /*
1056 * Check to see if we don't want our own messages.
1057 */
1058 if ((so->so_options & SO_USELOOPBACK) == 0) {
1059 if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
1060 if (rtm)
1061 Free(rtm);
1062 m_freem(m);
1063 goto out;
1064 }
1065 /* There is another listener, so construct message */
1066 rp = sotorawcb(so);
1067 }
1068 if (rtm) {
1069 m_copyback(m, 0, rtm->rtm_msglen, rtm);
1070 if (m->m_pkthdr.len < rtm->rtm_msglen) {
1071 m_freem(m);
1072 m = NULL;
1073 } else if (m->m_pkthdr.len > rtm->rtm_msglen)
1074 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
1075 Free(rtm);
1076 }
1077 if (rp)
1078 rp->rcb_proto.sp_family = 0; /* Avoid us */
1079 if (family)
1080 proto.sp_protocol = family;
1081 if (m)
1082 raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
1083 &COMPATNAME(route_info).ri_dst, &rt_rawcb);
1084 if (rp)
1085 rp->rcb_proto.sp_family = PF_XROUTE;
1086 }
1087 out:
1088 curlwp_bindx(bound);
1089 return error;
1090 }
1091
1092 static int
1093 route_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1094 {
1095 struct routecb *rop = sotoroutecb(so);
1096 int error = 0;
1097 unsigned char *rtm_type;
1098 size_t len;
1099 unsigned int msgfilter;
1100
1101 KASSERT(solocked(so));
1102
1103 if (sopt->sopt_level != AF_ROUTE) {
1104 error = ENOPROTOOPT;
1105 } else switch (op) {
1106 case PRCO_SETOPT:
1107 switch (sopt->sopt_name) {
1108 case RO_MSGFILTER:
1109 msgfilter = 0;
1110 for (rtm_type = sopt->sopt_data, len = sopt->sopt_size;
1111 len != 0;
1112 rtm_type++, len -= sizeof(*rtm_type))
1113 {
1114 /* Guard against overflowing our storage. */
1115 if (*rtm_type >= sizeof(msgfilter) * CHAR_BIT) {
1116 error = EOVERFLOW;
1117 break;
1118 }
1119 msgfilter |= RTMSGFILTER(*rtm_type);
1120 }
1121 if (error == 0)
1122 rop->rocb_msgfilter = msgfilter;
1123 break;
1124 default:
1125 error = ENOPROTOOPT;
1126 break;
1127 }
1128 break;
1129 case PRCO_GETOPT:
1130 switch (sopt->sopt_name) {
1131 case RO_MSGFILTER:
1132 error = ENOTSUP;
1133 break;
1134 default:
1135 error = ENOPROTOOPT;
1136 break;
1137 }
1138 }
1139 return error;
1140 }
1141
1142 static void
1143 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
1144 {
1145 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
1146 metric(RTV_RPIPE, rmx_recvpipe);
1147 metric(RTV_SPIPE, rmx_sendpipe);
1148 metric(RTV_SSTHRESH, rmx_ssthresh);
1149 metric(RTV_RTT, rmx_rtt);
1150 metric(RTV_RTTVAR, rmx_rttvar);
1151 metric(RTV_HOPCOUNT, rmx_hopcount);
1152 metric(RTV_MTU, rmx_mtu);
1153 #undef metric
1154 if (which & RTV_EXPIRE) {
1155 out->rt_rmx.rmx_expire = in->rtm_rmx.rmx_expire ?
1156 time_wall_to_mono(in->rtm_rmx.rmx_expire) : 0;
1157 }
1158 }
1159
1160 static void
1161 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
1162 {
1163 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
1164 metric(rmx_recvpipe);
1165 metric(rmx_sendpipe);
1166 metric(rmx_ssthresh);
1167 metric(rmx_rtt);
1168 metric(rmx_rttvar);
1169 metric(rmx_hopcount);
1170 metric(rmx_mtu);
1171 metric(rmx_locks);
1172 #undef metric
1173 out->rtm_rmx.rmx_expire = in->rt_rmx.rmx_expire ?
1174 time_mono_to_wall(in->rt_rmx.rmx_expire) : 0;
1175 }
1176
1177 static int
1178 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
1179 struct rt_addrinfo *rtinfo)
1180 {
1181 const struct sockaddr *sa = NULL; /* Quell compiler warning */
1182 int i;
1183
1184 for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
1185 if ((rtinfo->rti_addrs & (1 << i)) == 0)
1186 continue;
1187 rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
1188 RT_XADVANCE(cp, sa);
1189 }
1190
1191 /*
1192 * Check for extra addresses specified, except RTM_GET asking
1193 * for interface info.
1194 */
1195 if (rtmtype == RTM_GET) {
1196 if (((rtinfo->rti_addrs &
1197 (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0U << i)) != 0)
1198 return 1;
1199 } else if ((rtinfo->rti_addrs & (~0U << i)) != 0)
1200 return 1;
1201 /* Check for bad data length. */
1202 if (cp != cplim) {
1203 if (i == RTAX_NETMASK + 1 && sa != NULL &&
1204 cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
1205 /*
1206 * The last sockaddr was info.rti_info[RTAX_NETMASK].
1207 * We accept this for now for the sake of old
1208 * binaries or third party softwares.
1209 */
1210 ;
1211 else
1212 return 1;
1213 }
1214 return 0;
1215 }
1216
1217 static int
1218 rt_getlen(int type)
1219 {
1220 #ifndef COMPAT_RTSOCK
1221 CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
1222 CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
1223 CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
1224 CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
1225 #endif
1226
1227 switch (type) {
1228 case RTM_ODELADDR:
1229 case RTM_ONEWADDR:
1230 case RTM_OCHGADDR:
1231 #ifdef COMPAT_70
1232 return sizeof(struct ifa_msghdr70);
1233 #else
1234 #ifdef RTSOCK_DEBUG
1235 printf("%s: unsupported RTM type %d\n", __func__, type);
1236 #endif
1237 return -1;
1238 #endif
1239 case RTM_DELADDR:
1240 case RTM_NEWADDR:
1241 case RTM_CHGADDR:
1242 return sizeof(struct ifa_xmsghdr);
1243
1244 case RTM_OOIFINFO:
1245 #ifdef COMPAT_14
1246 return sizeof(struct if_msghdr14);
1247 #else
1248 #ifdef RTSOCK_DEBUG
1249 printf("%s: unsupported RTM type RTM_OOIFINFO\n", __func__);
1250 #endif
1251 return -1;
1252 #endif
1253 case RTM_OIFINFO:
1254 #ifdef COMPAT_50
1255 return sizeof(struct if_msghdr50);
1256 #else
1257 #ifdef RTSOCK_DEBUG
1258 printf("%s: unsupported RTM type RTM_OIFINFO\n", __func__);
1259 #endif
1260 return -1;
1261 #endif
1262
1263 case RTM_IFINFO:
1264 return sizeof(struct if_xmsghdr);
1265
1266 case RTM_IFANNOUNCE:
1267 case RTM_IEEE80211:
1268 return sizeof(struct if_xannouncemsghdr);
1269
1270 default:
1271 return sizeof(struct rt_xmsghdr);
1272 }
1273 }
1274
1275
1276 struct mbuf *
1277 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
1278 {
1279 struct rt_xmsghdr *rtm;
1280 struct mbuf *m;
1281 int i;
1282 const struct sockaddr *sa;
1283 int len, dlen;
1284
1285 m = m_gethdr(M_DONTWAIT, MT_DATA);
1286 if (m == NULL)
1287 return m;
1288 MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
1289
1290 if ((len = rt_getlen(type)) == -1)
1291 goto out;
1292 if (len > MHLEN + MLEN)
1293 panic("%s: message too long", __func__);
1294 else if (len > MHLEN) {
1295 m->m_next = m_get(M_DONTWAIT, MT_DATA);
1296 if (m->m_next == NULL)
1297 goto out;
1298 MCLAIM(m->m_next, m->m_owner);
1299 m->m_pkthdr.len = len;
1300 m->m_len = MHLEN;
1301 m->m_next->m_len = len - MHLEN;
1302 } else {
1303 m->m_pkthdr.len = m->m_len = len;
1304 }
1305 m_reset_rcvif(m);
1306 m_copyback(m, 0, datalen, data);
1307 if (len > datalen)
1308 (void)memset(mtod(m, char *) + datalen, 0, len - datalen);
1309 rtm = mtod(m, struct rt_xmsghdr *);
1310 for (i = 0; i < RTAX_MAX; i++) {
1311 if ((sa = rtinfo->rti_info[i]) == NULL)
1312 continue;
1313 rtinfo->rti_addrs |= (1 << i);
1314 dlen = RT_XROUNDUP(sa->sa_len);
1315 m_copyback(m, len, sa->sa_len, sa);
1316 if (dlen != sa->sa_len) {
1317 /*
1318 * Up to 6 + 1 nul's since roundup is to
1319 * sizeof(uint64_t) (8 bytes)
1320 */
1321 m_copyback(m, len + sa->sa_len,
1322 dlen - sa->sa_len, "\0\0\0\0\0\0");
1323 }
1324 len += dlen;
1325 }
1326 if (m->m_pkthdr.len != len)
1327 goto out;
1328 rtm->rtm_msglen = len;
1329 rtm->rtm_version = RTM_XVERSION;
1330 rtm->rtm_type = type;
1331 return m;
1332 out:
1333 m_freem(m);
1334 return NULL;
1335 }
1336
1337 /*
1338 * rt_msg2
1339 *
1340 * fills 'cp' or 'w'.w_tmem with the routing socket message and
1341 * returns the length of the message in 'lenp'.
1342 *
1343 * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
1344 * the message
1345 * otherwise walkarg's w_needed is updated and if the user buffer is
1346 * specified and w_needed indicates space exists the information is copied
1347 * into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
1348 * if the allocation fails ENOBUFS is returned.
1349 */
1350 static int
1351 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1352 int *lenp)
1353 {
1354 int i;
1355 int len, dlen, second_time = 0;
1356 char *cp0, *cp = cpv;
1357
1358 rtinfo->rti_addrs = 0;
1359 again:
1360 if ((len = rt_getlen(type)) == -1)
1361 return EINVAL;
1362
1363 if ((cp0 = cp) != NULL)
1364 cp += len;
1365 for (i = 0; i < RTAX_MAX; i++) {
1366 const struct sockaddr *sa;
1367
1368 if ((sa = rtinfo->rti_info[i]) == NULL)
1369 continue;
1370 rtinfo->rti_addrs |= (1 << i);
1371 dlen = RT_XROUNDUP(sa->sa_len);
1372 if (cp) {
1373 int diff = dlen - sa->sa_len;
1374 (void)memcpy(cp, sa, (size_t)sa->sa_len);
1375 cp += sa->sa_len;
1376 if (diff > 0) {
1377 (void)memset(cp, 0, (size_t)diff);
1378 cp += diff;
1379 }
1380 }
1381 len += dlen;
1382 }
1383 if (cp == NULL && w != NULL && !second_time) {
1384 struct rt_walkarg *rw = w;
1385
1386 rw->w_needed += len;
1387 if (rw->w_needed <= 0 && rw->w_where) {
1388 if (rw->w_tmemsize < len) {
1389 if (rw->w_tmem)
1390 kmem_free(rw->w_tmem, rw->w_tmemsize);
1391 rw->w_tmem = kmem_alloc(len, KM_SLEEP);
1392 rw->w_tmemsize = len;
1393 }
1394 if (rw->w_tmem) {
1395 cp = rw->w_tmem;
1396 second_time = 1;
1397 goto again;
1398 } else {
1399 rw->w_tmemneeded = len;
1400 return ENOBUFS;
1401 }
1402 }
1403 }
1404 if (cp) {
1405 struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
1406
1407 rtm->rtm_version = RTM_XVERSION;
1408 rtm->rtm_type = type;
1409 rtm->rtm_msglen = len;
1410 }
1411 if (lenp)
1412 *lenp = len;
1413 return 0;
1414 }
1415
1416 #ifndef COMPAT_RTSOCK
1417 int
1418 rt_msg3(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
1419 int *lenp)
1420 {
1421 return rt_msg2(type, rtinfo, cpv, w, lenp);
1422 }
1423 #endif
1424
1425 /*
1426 * This routine is called to generate a message from the routing
1427 * socket indicating that a redirect has occurred, a routing lookup
1428 * has failed, or that a protocol has detected timeouts to a particular
1429 * destination.
1430 */
1431 void
1432 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
1433 int error)
1434 {
1435 struct rt_xmsghdr rtm;
1436 struct mbuf *m;
1437 const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
1438 struct rt_addrinfo info = *rtinfo;
1439
1440 COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
1441 if (COMPATNAME(route_info).ri_cb.any_count == 0)
1442 return;
1443 memset(&rtm, 0, sizeof(rtm));
1444 rtm.rtm_pid = curproc->p_pid;
1445 rtm.rtm_flags = RTF_DONE | flags;
1446 rtm.rtm_errno = error;
1447 m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
1448 if (m == NULL)
1449 return;
1450 mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1451 COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1452 }
1453
1454 /*
1455 * This routine is called to generate a message from the routing
1456 * socket indicating that the status of a network interface has changed.
1457 */
1458 void
1459 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
1460 {
1461 struct if_xmsghdr ifm;
1462 struct mbuf *m;
1463 struct rt_addrinfo info;
1464
1465 COMPATCALL(rt_ifmsg, (ifp));
1466 if (COMPATNAME(route_info).ri_cb.any_count == 0)
1467 return;
1468 (void)memset(&info, 0, sizeof(info));
1469 (void)memset(&ifm, 0, sizeof(ifm));
1470 ifm.ifm_index = ifp->if_index;
1471 ifm.ifm_flags = ifp->if_flags;
1472 ifm.ifm_data = ifp->if_data;
1473 ifm.ifm_addrs = 0;
1474 m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
1475 if (m == NULL)
1476 return;
1477 COMPATNAME(route_enqueue)(m, 0);
1478 #ifdef COMPAT_14
1479 compat_14_rt_oifmsg(ifp);
1480 #endif
1481 #ifdef COMPAT_50
1482 compat_50_rt_oifmsg(ifp);
1483 #endif
1484 }
1485
1486 #ifndef COMPAT_RTSOCK
1487 static int
1488 if_addrflags(struct ifaddr *ifa)
1489 {
1490
1491 switch (ifa->ifa_addr->sa_family) {
1492 #ifdef INET
1493 case AF_INET:
1494 return ((struct in_ifaddr *)ifa)->ia4_flags;
1495 #endif
1496 #ifdef INET6
1497 case AF_INET6:
1498 return ((struct in6_ifaddr *)ifa)->ia6_flags;
1499 #endif
1500 default:
1501 return 0;
1502 }
1503 }
1504 #endif
1505
1506 /*
1507 * This is called to generate messages from the routing socket
1508 * indicating a network interface has had addresses associated with it.
1509 * if we ever reverse the logic and replace messages TO the routing
1510 * socket indicate a request to configure interfaces, then it will
1511 * be unnecessary as the routing socket will automatically generate
1512 * copies of it.
1513 */
1514 void
1515 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
1516 struct rtentry *rt)
1517 {
1518 #define cmdpass(__cmd, __pass) (((__cmd) << 2) | (__pass))
1519 struct rt_addrinfo info;
1520 const struct sockaddr *sa;
1521 int pass;
1522 struct mbuf *m;
1523 struct ifnet *ifp;
1524 struct rt_xmsghdr rtm;
1525 struct ifa_xmsghdr ifam;
1526 int ncmd;
1527
1528 KASSERT(ifa != NULL);
1529 KASSERT(ifa->ifa_addr != NULL);
1530 ifp = ifa->ifa_ifp;
1531 #ifdef SCTP
1532 if (cmd == RTM_ADD) {
1533 sctp_add_ip_address(ifa);
1534 } else if (cmd == RTM_DELETE) {
1535 sctp_delete_ip_address(ifa);
1536 }
1537 #endif
1538
1539 COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
1540 if (COMPATNAME(route_info).ri_cb.any_count == 0)
1541 return;
1542 for (pass = 1; pass < 3; pass++) {
1543 memset(&info, 0, sizeof(info));
1544 switch (cmdpass(cmd, pass)) {
1545 case cmdpass(RTM_ADD, 1):
1546 case cmdpass(RTM_CHANGE, 1):
1547 case cmdpass(RTM_DELETE, 2):
1548 case cmdpass(RTM_NEWADDR, 1):
1549 case cmdpass(RTM_DELADDR, 1):
1550 case cmdpass(RTM_CHGADDR, 1):
1551 switch (cmd) {
1552 case RTM_ADD:
1553 ncmd = RTM_XNEWADDR;
1554 break;
1555 case RTM_DELETE:
1556 ncmd = RTM_XDELADDR;
1557 break;
1558 case RTM_CHANGE:
1559 ncmd = RTM_XCHGADDR;
1560 break;
1561 case RTM_NEWADDR:
1562 ncmd = RTM_XNEWADDR;
1563 break;
1564 case RTM_DELADDR:
1565 ncmd = RTM_XDELADDR;
1566 break;
1567 case RTM_CHGADDR:
1568 ncmd = RTM_XCHGADDR;
1569 break;
1570 default:
1571 panic("%s: unknown command %d", __func__, cmd);
1572 }
1573 #ifdef COMPAT_70
1574 compat_70_rt_newaddrmsg1(ncmd, ifa);
1575 #endif
1576 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
1577 KASSERT(ifp->if_dl != NULL);
1578 info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1579 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1580 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1581 memset(&ifam, 0, sizeof(ifam));
1582 ifam.ifam_index = ifp->if_index;
1583 ifam.ifam_metric = ifa->ifa_metric;
1584 ifam.ifam_flags = ifa->ifa_flags;
1585 #ifndef COMPAT_RTSOCK
1586 ifam.ifam_pid = curproc->p_pid;
1587 ifam.ifam_addrflags = if_addrflags(ifa);
1588 #endif
1589 m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
1590 if (m == NULL)
1591 continue;
1592 mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
1593 info.rti_addrs;
1594 break;
1595 case cmdpass(RTM_ADD, 2):
1596 case cmdpass(RTM_CHANGE, 2):
1597 case cmdpass(RTM_DELETE, 1):
1598 if (rt == NULL)
1599 continue;
1600 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1601 info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
1602 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1603 memset(&rtm, 0, sizeof(rtm));
1604 rtm.rtm_pid = curproc->p_pid;
1605 rtm.rtm_index = ifp->if_index;
1606 rtm.rtm_flags |= rt->rt_flags;
1607 rtm.rtm_errno = error;
1608 m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
1609 if (m == NULL)
1610 continue;
1611 mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
1612 break;
1613 default:
1614 continue;
1615 }
1616 KASSERTMSG(m != NULL, "called with wrong command");
1617 COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
1618 }
1619 #undef cmdpass
1620
1621 }
1622
1623 static struct mbuf *
1624 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
1625 struct rt_addrinfo *info)
1626 {
1627 struct if_xannouncemsghdr ifan;
1628
1629 memset(info, 0, sizeof(*info));
1630 memset(&ifan, 0, sizeof(ifan));
1631 ifan.ifan_index = ifp->if_index;
1632 strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
1633 ifan.ifan_what = what;
1634 return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
1635 }
1636
1637 /*
1638 * This is called to generate routing socket messages indicating
1639 * network interface arrival and departure.
1640 */
1641 void
1642 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
1643 {
1644 struct mbuf *m;
1645 struct rt_addrinfo info;
1646
1647 COMPATCALL(rt_ifannouncemsg, (ifp, what));
1648 if (COMPATNAME(route_info).ri_cb.any_count == 0)
1649 return;
1650 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
1651 if (m == NULL)
1652 return;
1653 COMPATNAME(route_enqueue)(m, 0);
1654 }
1655
1656 /*
1657 * This is called to generate routing socket messages indicating
1658 * IEEE80211 wireless events.
1659 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
1660 */
1661 void
1662 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
1663 size_t data_len)
1664 {
1665 struct mbuf *m;
1666 struct rt_addrinfo info;
1667
1668 COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
1669 if (COMPATNAME(route_info).ri_cb.any_count == 0)
1670 return;
1671 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
1672 if (m == NULL)
1673 return;
1674 /*
1675 * Append the ieee80211 data. Try to stick it in the
1676 * mbuf containing the ifannounce msg; otherwise allocate
1677 * a new mbuf and append.
1678 *
1679 * NB: we assume m is a single mbuf.
1680 */
1681 if (data_len > M_TRAILINGSPACE(m)) {
1682 struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
1683 if (n == NULL) {
1684 m_freem(m);
1685 return;
1686 }
1687 (void)memcpy(mtod(n, void *), data, data_len);
1688 n->m_len = data_len;
1689 m->m_next = n;
1690 } else if (data_len > 0) {
1691 (void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
1692 m->m_len += data_len;
1693 }
1694 if (m->m_flags & M_PKTHDR)
1695 m->m_pkthdr.len += data_len;
1696 mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
1697 COMPATNAME(route_enqueue)(m, 0);
1698 }
1699
1700 #ifndef COMPAT_RTSOCK
1701 /*
1702 * Send a routing message as mimicing that a cloned route is added.
1703 */
1704 void
1705 rt_clonedmsg(const struct sockaddr *dst, const struct ifnet *ifp,
1706 const struct rtentry *rt)
1707 {
1708 struct rt_addrinfo info;
1709 /* Mimic flags exactly */
1710 #define RTF_LLINFO 0x400
1711 #define RTF_CLONED 0x2000
1712 int flags = RTF_UP | RTF_HOST | RTF_DONE | RTF_LLINFO | RTF_CLONED;
1713 union {
1714 struct sockaddr sa;
1715 struct sockaddr_storage ss;
1716 struct sockaddr_dl sdl;
1717 } u;
1718 uint8_t namelen = strlen(ifp->if_xname);
1719 uint8_t addrlen = ifp->if_addrlen;
1720
1721 if (rt == NULL)
1722 return; /* XXX */
1723
1724 memset(&info, 0, sizeof(info));
1725 info.rti_info[RTAX_DST] = dst;
1726 sockaddr_dl_init(&u.sdl, sizeof(u.ss), ifp->if_index, ifp->if_type,
1727 NULL, namelen, NULL, addrlen);
1728 info.rti_info[RTAX_GATEWAY] = &u.sa;
1729
1730 rt_missmsg(RTM_ADD, &info, flags, 0);
1731 #undef RTF_LLINFO
1732 #undef RTF_CLONED
1733 }
1734 #endif /* COMPAT_RTSOCK */
1735
1736 /*
1737 * This is used in dumping the kernel table via sysctl().
1738 */
1739 static int
1740 sysctl_dumpentry(struct rtentry *rt, void *v)
1741 {
1742 struct rt_walkarg *w = v;
1743 int error = 0, size;
1744 struct rt_addrinfo info;
1745
1746 if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
1747 return 0;
1748 memset(&info, 0, sizeof(info));
1749 info.rti_info[RTAX_DST] = rt_getkey(rt);
1750 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1751 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1752 info.rti_info[RTAX_TAG] = rt_gettag(rt);
1753 if (rt->rt_ifp) {
1754 const struct ifaddr *rtifa;
1755 info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
1756 /* rtifa used to be simply rt->rt_ifa. If rt->rt_ifa != NULL,
1757 * then rt_get_ifa() != NULL. So this ought to still be safe.
1758 * --dyoung
1759 */
1760 rtifa = rt_get_ifa(rt);
1761 info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
1762 if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
1763 info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
1764 }
1765 if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
1766 return error;
1767 if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1768 struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
1769
1770 rtm->rtm_flags = rt->rt_flags;
1771 rtm->rtm_use = rt->rt_use;
1772 rtm_setmetrics(rt, rtm);
1773 KASSERT(rt->rt_ifp != NULL);
1774 rtm->rtm_index = rt->rt_ifp->if_index;
1775 rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
1776 rtm->rtm_addrs = info.rti_addrs;
1777 if ((error = copyout(rtm, w->w_where, size)) != 0)
1778 w->w_where = NULL;
1779 else
1780 w->w_where = (char *)w->w_where + size;
1781 }
1782 return error;
1783 }
1784
1785 static int
1786 sysctl_iflist_if(struct ifnet *ifp, struct rt_walkarg *w,
1787 struct rt_addrinfo *info, size_t len)
1788 {
1789 struct if_xmsghdr *ifm;
1790 int error;
1791
1792 ifm = (struct if_xmsghdr *)w->w_tmem;
1793 ifm->ifm_index = ifp->if_index;
1794 ifm->ifm_flags = ifp->if_flags;
1795 ifm->ifm_data = ifp->if_data;
1796 ifm->ifm_addrs = info->rti_addrs;
1797 if ((error = copyout(ifm, w->w_where, len)) == 0)
1798 w->w_where = (char *)w->w_where + len;
1799 return error;
1800 }
1801
1802 static int
1803 sysctl_iflist_addr(struct rt_walkarg *w, struct ifaddr *ifa,
1804 struct rt_addrinfo *info)
1805 {
1806 int len, error;
1807
1808 if ((error = rt_msg2(RTM_XNEWADDR, info, 0, w, &len)))
1809 return error;
1810 if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1811 struct ifa_xmsghdr *ifam;
1812
1813 ifam = (struct ifa_xmsghdr *)w->w_tmem;
1814 ifam->ifam_index = ifa->ifa_ifp->if_index;
1815 ifam->ifam_flags = ifa->ifa_flags;
1816 ifam->ifam_metric = ifa->ifa_metric;
1817 ifam->ifam_addrs = info->rti_addrs;
1818 #ifndef COMPAT_RTSOCK
1819 ifam->ifam_pid = 0;
1820 ifam->ifam_addrflags = if_addrflags(ifa);
1821 #endif
1822 if ((error = copyout(w->w_tmem, w->w_where, len)) == 0)
1823 w->w_where = (char *)w->w_where + len;
1824 }
1825 return error;
1826 }
1827
1828 static int
1829 sysctl_iflist(int af, struct rt_walkarg *w, int type)
1830 {
1831 struct ifnet *ifp;
1832 struct ifaddr *ifa;
1833 struct rt_addrinfo info;
1834 int cmd, len, error = 0;
1835 int (*iflist_if)(struct ifnet *, struct rt_walkarg *,
1836 struct rt_addrinfo *, size_t);
1837 int (*iflist_addr)(struct rt_walkarg *, struct ifaddr *,
1838 struct rt_addrinfo *);
1839 int s;
1840 struct psref psref;
1841 int bound;
1842
1843 switch (type) {
1844 case NET_RT_IFLIST:
1845 cmd = RTM_IFINFO;
1846 iflist_if = sysctl_iflist_if;
1847 iflist_addr = sysctl_iflist_addr;
1848 break;
1849 #ifdef COMPAT_14
1850 case NET_RT_OOOIFLIST:
1851 cmd = RTM_OOIFINFO;
1852 iflist_if = compat_14_iflist;
1853 iflist_addr = compat_70_iflist_addr;
1854 break;
1855 #endif
1856 #ifdef COMPAT_50
1857 case NET_RT_OOIFLIST:
1858 cmd = RTM_OIFINFO;
1859 iflist_if = compat_50_iflist;
1860 iflist_addr = compat_70_iflist_addr;
1861 break;
1862 #endif
1863 #ifdef COMPAT_70
1864 case NET_RT_OIFLIST:
1865 cmd = RTM_IFINFO;
1866 iflist_if = sysctl_iflist_if;
1867 iflist_addr = compat_70_iflist_addr;
1868 break;
1869 #endif
1870 default:
1871 #ifdef RTSOCK_DEBUG
1872 printf("%s: unsupported IFLIST type %d\n", __func__, type);
1873 #endif
1874 return EINVAL;
1875 }
1876
1877 memset(&info, 0, sizeof(info));
1878
1879 bound = curlwp_bind();
1880 s = pserialize_read_enter();
1881 IFNET_READER_FOREACH(ifp) {
1882 int _s;
1883 if (w->w_arg && w->w_arg != ifp->if_index)
1884 continue;
1885 if (IFADDR_READER_EMPTY(ifp))
1886 continue;
1887
1888 if_acquire(ifp, &psref);
1889 pserialize_read_exit(s);
1890
1891 info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
1892 if ((error = rt_msg2(cmd, &info, NULL, w, &len)) != 0)
1893 goto release_exit;
1894 info.rti_info[RTAX_IFP] = NULL;
1895 if (w->w_where && w->w_tmem && w->w_needed <= 0) {
1896 if ((error = iflist_if(ifp, w, &info, len)) != 0)
1897 goto release_exit;
1898 }
1899 _s = pserialize_read_enter();
1900 IFADDR_READER_FOREACH(ifa, ifp) {
1901 struct psref _psref;
1902 if (af && af != ifa->ifa_addr->sa_family)
1903 continue;
1904 ifa_acquire(ifa, &_psref);
1905 pserialize_read_exit(_s);
1906
1907 info.rti_info[RTAX_IFA] = ifa->ifa_addr;
1908 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
1909 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
1910 error = iflist_addr(w, ifa, &info);
1911
1912 _s = pserialize_read_enter();
1913 ifa_release(ifa, &_psref);
1914 if (error != 0) {
1915 pserialize_read_exit(_s);
1916 goto release_exit;
1917 }
1918 }
1919 pserialize_read_exit(_s);
1920 info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
1921 info.rti_info[RTAX_BRD] = NULL;
1922
1923 s = pserialize_read_enter();
1924 if_release(ifp, &psref);
1925 }
1926 pserialize_read_exit(s);
1927 curlwp_bindx(bound);
1928
1929 return 0;
1930
1931 release_exit:
1932 if_release(ifp, &psref);
1933 curlwp_bindx(bound);
1934 return error;
1935 }
1936
1937 static int
1938 sysctl_rtable(SYSCTLFN_ARGS)
1939 {
1940 void *where = oldp;
1941 size_t *given = oldlenp;
1942 int i, s, error = EINVAL;
1943 u_char af;
1944 struct rt_walkarg w;
1945
1946 if (namelen == 1 && name[0] == CTL_QUERY)
1947 return sysctl_query(SYSCTLFN_CALL(rnode));
1948
1949 if (newp)
1950 return EPERM;
1951 if (namelen != 3)
1952 return EINVAL;
1953 af = name[0];
1954 w.w_tmemneeded = 0;
1955 w.w_tmemsize = 0;
1956 w.w_tmem = NULL;
1957 again:
1958 /* we may return here if a later [re]alloc of the t_mem buffer fails */
1959 if (w.w_tmemneeded) {
1960 w.w_tmem = kmem_alloc(w.w_tmemneeded, KM_SLEEP);
1961 w.w_tmemsize = w.w_tmemneeded;
1962 w.w_tmemneeded = 0;
1963 }
1964 w.w_op = name[1];
1965 w.w_arg = name[2];
1966 w.w_given = *given;
1967 w.w_needed = 0 - w.w_given;
1968 w.w_where = where;
1969
1970 s = splsoftnet();
1971 switch (w.w_op) {
1972
1973 case NET_RT_DUMP:
1974 case NET_RT_FLAGS:
1975 #if defined(INET) || defined(INET6)
1976 /*
1977 * take care of llinfo entries, the caller must
1978 * specify an AF
1979 */
1980 if (w.w_op == NET_RT_FLAGS &&
1981 (w.w_arg == 0 || w.w_arg & RTF_LLDATA)) {
1982 if (af != 0)
1983 error = lltable_sysctl_dump(af, &w);
1984 else
1985 error = EINVAL;
1986 break;
1987 }
1988 #endif
1989
1990 for (i = 1; i <= AF_MAX; i++) {
1991 if (af == 0 || af == i) {
1992 error = rt_walktree(i, sysctl_dumpentry, &w);
1993 if (error != 0)
1994 break;
1995 #if defined(INET) || defined(INET6)
1996 /*
1997 * Return ARP/NDP entries too for
1998 * backward compatibility.
1999 */
2000 error = lltable_sysctl_dump(i, &w);
2001 if (error != 0)
2002 break;
2003 #endif
2004 }
2005 }
2006 break;
2007
2008 #ifdef COMPAT_14
2009 case NET_RT_OOOIFLIST:
2010 error = sysctl_iflist(af, &w, w.w_op);
2011 break;
2012 #endif
2013 #ifdef COMPAT_50
2014 case NET_RT_OOIFLIST:
2015 error = sysctl_iflist(af, &w, w.w_op);
2016 break;
2017 #endif
2018 #ifdef COMPAT_70
2019 case NET_RT_OIFLIST:
2020 error = sysctl_iflist(af, &w, w.w_op);
2021 break;
2022 #endif
2023 case NET_RT_IFLIST:
2024 error = sysctl_iflist(af, &w, w.w_op);
2025 break;
2026 }
2027 splx(s);
2028
2029 /* check to see if we couldn't allocate memory with NOWAIT */
2030 if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
2031 goto again;
2032
2033 if (w.w_tmem)
2034 kmem_free(w.w_tmem, w.w_tmemsize);
2035 w.w_needed += w.w_given;
2036 if (where) {
2037 *given = (char *)w.w_where - (char *)where;
2038 if (*given < w.w_needed)
2039 return ENOMEM;
2040 } else {
2041 *given = (11 * w.w_needed) / 10;
2042 }
2043 return error;
2044 }
2045
2046 /*
2047 * Routing message software interrupt routine
2048 */
2049 static void
2050 COMPATNAME(route_intr)(void *cookie)
2051 {
2052 struct sockproto proto = { .sp_family = PF_XROUTE, };
2053 struct route_info * const ri = &COMPATNAME(route_info);
2054 struct mbuf *m;
2055
2056 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
2057 for (;;) {
2058 IFQ_LOCK(&ri->ri_intrq);
2059 IF_DEQUEUE(&ri->ri_intrq, m);
2060 IFQ_UNLOCK(&ri->ri_intrq);
2061 if (m == NULL)
2062 break;
2063 proto.sp_protocol = M_GETCTX(m, uintptr_t);
2064 #ifdef NET_MPSAFE
2065 mutex_enter(rt_so_mtx);
2066 #endif
2067 raw_input(m, &proto, &ri->ri_src, &ri->ri_dst, &rt_rawcb);
2068 #ifdef NET_MPSAFE
2069 mutex_exit(rt_so_mtx);
2070 #endif
2071 }
2072 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
2073 }
2074
2075 /*
2076 * Enqueue a message to the software interrupt routine.
2077 */
2078 void
2079 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
2080 {
2081 struct route_info * const ri = &COMPATNAME(route_info);
2082 int wasempty;
2083
2084 IFQ_LOCK(&ri->ri_intrq);
2085 if (IF_QFULL(&ri->ri_intrq)) {
2086 IF_DROP(&ri->ri_intrq);
2087 IFQ_UNLOCK(&ri->ri_intrq);
2088 m_freem(m);
2089 } else {
2090 wasempty = IF_IS_EMPTY(&ri->ri_intrq);
2091 M_SETCTX(m, (uintptr_t)family);
2092 IF_ENQUEUE(&ri->ri_intrq, m);
2093 IFQ_UNLOCK(&ri->ri_intrq);
2094 if (wasempty) {
2095 kpreempt_disable();
2096 softint_schedule(ri->ri_sih);
2097 kpreempt_enable();
2098 }
2099 }
2100 }
2101
2102 static void
2103 COMPATNAME(route_init)(void)
2104 {
2105 struct route_info * const ri = &COMPATNAME(route_info);
2106
2107 #ifndef COMPAT_RTSOCK
2108 rt_init();
2109 #endif
2110 #ifdef NET_MPSAFE
2111 rt_so_mtx = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
2112 #endif
2113
2114 sysctl_net_route_setup(NULL);
2115 ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
2116 ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
2117 COMPATNAME(route_intr), NULL);
2118 IFQ_LOCK_INIT(&ri->ri_intrq);
2119 }
2120
2121 /*
2122 * Definitions of protocols supported in the ROUTE domain.
2123 */
2124 #ifndef COMPAT_RTSOCK
2125 PR_WRAP_USRREQS(route);
2126 #else
2127 PR_WRAP_USRREQS(compat_50_route);
2128 #endif
2129
2130 static const struct pr_usrreqs route_usrreqs = {
2131 .pr_attach = COMPATNAME(route_attach_wrapper),
2132 .pr_detach = COMPATNAME(route_detach_wrapper),
2133 .pr_accept = COMPATNAME(route_accept_wrapper),
2134 .pr_bind = COMPATNAME(route_bind_wrapper),
2135 .pr_listen = COMPATNAME(route_listen_wrapper),
2136 .pr_connect = COMPATNAME(route_connect_wrapper),
2137 .pr_connect2 = COMPATNAME(route_connect2_wrapper),
2138 .pr_disconnect = COMPATNAME(route_disconnect_wrapper),
2139 .pr_shutdown = COMPATNAME(route_shutdown_wrapper),
2140 .pr_abort = COMPATNAME(route_abort_wrapper),
2141 .pr_ioctl = COMPATNAME(route_ioctl_wrapper),
2142 .pr_stat = COMPATNAME(route_stat_wrapper),
2143 .pr_peeraddr = COMPATNAME(route_peeraddr_wrapper),
2144 .pr_sockaddr = COMPATNAME(route_sockaddr_wrapper),
2145 .pr_rcvd = COMPATNAME(route_rcvd_wrapper),
2146 .pr_recvoob = COMPATNAME(route_recvoob_wrapper),
2147 .pr_send = COMPATNAME(route_send_wrapper),
2148 .pr_sendoob = COMPATNAME(route_sendoob_wrapper),
2149 .pr_purgeif = COMPATNAME(route_purgeif_wrapper),
2150 };
2151
2152 static const struct protosw COMPATNAME(route_protosw)[] = {
2153 {
2154 .pr_type = SOCK_RAW,
2155 .pr_domain = &COMPATNAME(routedomain),
2156 .pr_flags = PR_ATOMIC|PR_ADDR,
2157 .pr_input = raw_input,
2158 .pr_ctlinput = raw_ctlinput,
2159 .pr_ctloutput = route_ctloutput,
2160 .pr_usrreqs = &route_usrreqs,
2161 .pr_init = rt_pr_init,
2162 },
2163 };
2164
2165 struct domain COMPATNAME(routedomain) = {
2166 .dom_family = PF_XROUTE,
2167 .dom_name = DOMAINNAME,
2168 .dom_init = COMPATNAME(route_init),
2169 .dom_protosw = COMPATNAME(route_protosw),
2170 .dom_protoswNPROTOSW =
2171 &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
2172 };
2173
2174 static void
2175 sysctl_net_route_setup(struct sysctllog **clog)
2176 {
2177 const struct sysctlnode *rnode = NULL;
2178
2179 sysctl_createv(clog, 0, NULL, &rnode,
2180 CTLFLAG_PERMANENT,
2181 CTLTYPE_NODE, DOMAINNAME,
2182 SYSCTL_DESCR("PF_ROUTE information"),
2183 NULL, 0, NULL, 0,
2184 CTL_NET, PF_XROUTE, CTL_EOL);
2185
2186 sysctl_createv(clog, 0, NULL, NULL,
2187 CTLFLAG_PERMANENT,
2188 CTLTYPE_NODE, "rtable",
2189 SYSCTL_DESCR("Routing table information"),
2190 sysctl_rtable, 0, NULL, 0,
2191 CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
2192
2193 sysctl_createv(clog, 0, &rnode, NULL,
2194 CTLFLAG_PERMANENT,
2195 CTLTYPE_STRUCT, "stats",
2196 SYSCTL_DESCR("Routing statistics"),
2197 NULL, 0, &rtstat, sizeof(rtstat),
2198 CTL_CREATE, CTL_EOL);
2199 }
2200