Home | History | Annotate | Line # | Download | only in net
rtsock.c revision 1.200
      1 /*	$NetBSD: rtsock.c,v 1.200 2017/01/19 06:58:55 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 1.200 2017/01/19 06:58:55 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 static void
    181 rt_adjustcount(int af, int cnt)
    182 {
    183 	struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
    184 
    185 	cb->any_count += cnt;
    186 
    187 	switch (af) {
    188 	case AF_INET:
    189 		cb->ip_count += cnt;
    190 		return;
    191 #ifdef INET6
    192 	case AF_INET6:
    193 		cb->ip6_count += cnt;
    194 		return;
    195 #endif
    196 	case AF_MPLS:
    197 		cb->mpls_count += cnt;
    198 		return;
    199 	}
    200 }
    201 
    202 static int
    203 COMPATNAME(route_attach)(struct socket *so, int proto)
    204 {
    205 	struct rawcb *rp;
    206 	int s, error;
    207 
    208 	KASSERT(sotorawcb(so) == NULL);
    209 	rp = kmem_zalloc(sizeof(*rp), KM_SLEEP);
    210 	rp->rcb_len = sizeof(*rp);
    211 	so->so_pcb = rp;
    212 
    213 	s = splsoftnet();
    214 	if ((error = raw_attach(so, proto)) == 0) {
    215 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
    216 		rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
    217 		rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
    218 	}
    219 	splx(s);
    220 
    221 	if (error) {
    222 		kmem_free(rp, sizeof(*rp));
    223 		so->so_pcb = NULL;
    224 		return error;
    225 	}
    226 
    227 	soisconnected(so);
    228 	so->so_options |= SO_USELOOPBACK;
    229 	KASSERT(solocked(so));
    230 
    231 	return error;
    232 }
    233 
    234 static void
    235 COMPATNAME(route_detach)(struct socket *so)
    236 {
    237 	struct rawcb *rp = sotorawcb(so);
    238 	int s;
    239 
    240 	KASSERT(rp != NULL);
    241 	KASSERT(solocked(so));
    242 
    243 	s = splsoftnet();
    244 	rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
    245 	raw_detach(so);
    246 	splx(s);
    247 }
    248 
    249 static int
    250 COMPATNAME(route_accept)(struct socket *so, struct sockaddr *nam)
    251 {
    252 	KASSERT(solocked(so));
    253 
    254 	panic("route_accept");
    255 
    256 	return EOPNOTSUPP;
    257 }
    258 
    259 static int
    260 COMPATNAME(route_bind)(struct socket *so, struct sockaddr *nam, struct lwp *l)
    261 {
    262 	KASSERT(solocked(so));
    263 
    264 	return EOPNOTSUPP;
    265 }
    266 
    267 static int
    268 COMPATNAME(route_listen)(struct socket *so, struct lwp *l)
    269 {
    270 	KASSERT(solocked(so));
    271 
    272 	return EOPNOTSUPP;
    273 }
    274 
    275 static int
    276 COMPATNAME(route_connect)(struct socket *so, struct sockaddr *nam, struct lwp *l)
    277 {
    278 	KASSERT(solocked(so));
    279 
    280 	return EOPNOTSUPP;
    281 }
    282 
    283 static int
    284 COMPATNAME(route_connect2)(struct socket *so, struct socket *so2)
    285 {
    286 	KASSERT(solocked(so));
    287 
    288 	return EOPNOTSUPP;
    289 }
    290 
    291 static int
    292 COMPATNAME(route_disconnect)(struct socket *so)
    293 {
    294 	struct rawcb *rp = sotorawcb(so);
    295 	int s;
    296 
    297 	KASSERT(solocked(so));
    298 	KASSERT(rp != NULL);
    299 
    300 	s = splsoftnet();
    301 	soisdisconnected(so);
    302 	raw_disconnect(rp);
    303 	splx(s);
    304 
    305 	return 0;
    306 }
    307 
    308 static int
    309 COMPATNAME(route_shutdown)(struct socket *so)
    310 {
    311 	int s;
    312 
    313 	KASSERT(solocked(so));
    314 
    315 	/*
    316 	 * Mark the connection as being incapable of further input.
    317 	 */
    318 	s = splsoftnet();
    319 	socantsendmore(so);
    320 	splx(s);
    321 	return 0;
    322 }
    323 
    324 static int
    325 COMPATNAME(route_abort)(struct socket *so)
    326 {
    327 	KASSERT(solocked(so));
    328 
    329 	panic("route_abort");
    330 
    331 	return EOPNOTSUPP;
    332 }
    333 
    334 static int
    335 COMPATNAME(route_ioctl)(struct socket *so, u_long cmd, void *nam,
    336     struct ifnet * ifp)
    337 {
    338 	return EOPNOTSUPP;
    339 }
    340 
    341 static int
    342 COMPATNAME(route_stat)(struct socket *so, struct stat *ub)
    343 {
    344 	KASSERT(solocked(so));
    345 
    346 	return 0;
    347 }
    348 
    349 static int
    350 COMPATNAME(route_peeraddr)(struct socket *so, struct sockaddr *nam)
    351 {
    352 	struct rawcb *rp = sotorawcb(so);
    353 
    354 	KASSERT(solocked(so));
    355 	KASSERT(rp != NULL);
    356 	KASSERT(nam != NULL);
    357 
    358 	if (rp->rcb_faddr == NULL)
    359 		return ENOTCONN;
    360 
    361 	raw_setpeeraddr(rp, nam);
    362 	return 0;
    363 }
    364 
    365 static int
    366 COMPATNAME(route_sockaddr)(struct socket *so, struct sockaddr *nam)
    367 {
    368 	struct rawcb *rp = sotorawcb(so);
    369 
    370 	KASSERT(solocked(so));
    371 	KASSERT(rp != NULL);
    372 	KASSERT(nam != NULL);
    373 
    374 	if (rp->rcb_faddr == NULL)
    375 		return ENOTCONN;
    376 
    377 	raw_setsockaddr(rp, nam);
    378 	return 0;
    379 }
    380 
    381 static int
    382 COMPATNAME(route_rcvd)(struct socket *so, int flags, struct lwp *l)
    383 {
    384 	KASSERT(solocked(so));
    385 
    386 	return EOPNOTSUPP;
    387 }
    388 
    389 static int
    390 COMPATNAME(route_recvoob)(struct socket *so, struct mbuf *m, int flags)
    391 {
    392 	KASSERT(solocked(so));
    393 
    394 	return EOPNOTSUPP;
    395 }
    396 
    397 static int
    398 COMPATNAME(route_send)(struct socket *so, struct mbuf *m,
    399     struct sockaddr *nam, struct mbuf *control, struct lwp *l)
    400 {
    401 	int error = 0;
    402 	int s;
    403 
    404 	KASSERT(solocked(so));
    405 	KASSERT(so->so_proto == &COMPATNAME(route_protosw)[0]);
    406 
    407 	s = splsoftnet();
    408 	error = raw_send(so, m, nam, control, l, &COMPATNAME(route_output));
    409 	splx(s);
    410 
    411 	return error;
    412 }
    413 
    414 static int
    415 COMPATNAME(route_sendoob)(struct socket *so, struct mbuf *m,
    416     struct mbuf *control)
    417 {
    418 	KASSERT(solocked(so));
    419 
    420 	m_freem(m);
    421 	m_freem(control);
    422 
    423 	return EOPNOTSUPP;
    424 }
    425 static int
    426 COMPATNAME(route_purgeif)(struct socket *so, struct ifnet *ifp)
    427 {
    428 
    429 	panic("route_purgeif");
    430 
    431 	return EOPNOTSUPP;
    432 }
    433 
    434 #ifdef INET
    435 static int
    436 route_get_sdl_index(struct rt_addrinfo *info, int *sdl_index)
    437 {
    438 	struct rtentry *nrt;
    439 	int error;
    440 
    441 	error = rtrequest1(RTM_GET, info, &nrt);
    442 	if (error != 0)
    443 		return error;
    444 	/*
    445 	 * nrt->rt_ifp->if_index may not be correct
    446 	 * due to changing to ifplo0.
    447 	 */
    448 	*sdl_index = satosdl(nrt->rt_gateway)->sdl_index;
    449 	rt_unref(nrt);
    450 
    451 	return 0;
    452 }
    453 #endif /* INET */
    454 
    455 static void
    456 route_get_sdl(const struct ifnet *ifp, const struct sockaddr *dst,
    457     struct sockaddr_dl *sdl, int *flags)
    458 {
    459 	struct llentry *la;
    460 
    461 	KASSERT(ifp != NULL);
    462 
    463 	IF_AFDATA_RLOCK(ifp);
    464 	switch (dst->sa_family) {
    465 	case AF_INET:
    466 		la = lla_lookup(LLTABLE(ifp), 0, dst);
    467 		break;
    468 	case AF_INET6:
    469 		la = lla_lookup(LLTABLE6(ifp), 0, dst);
    470 		break;
    471 	default:
    472 		la = NULL;
    473 		KASSERTMSG(0, "Invalid AF=%d\n", dst->sa_family);
    474 		break;
    475 	}
    476 	IF_AFDATA_RUNLOCK(ifp);
    477 
    478 	void *a = (LLE_IS_VALID(la) && (la->la_flags & LLE_VALID) == LLE_VALID)
    479 	    ? &la->ll_addr : NULL;
    480 
    481 	a = sockaddr_dl_init(sdl, sizeof(*sdl), ifp->if_index, ifp->if_type,
    482 	    NULL, 0, a, ifp->if_addrlen);
    483 	KASSERT(a != NULL);
    484 
    485 	if (la != NULL) {
    486 		*flags = la->la_flags;
    487 		LLE_RUNLOCK(la);
    488 	}
    489 }
    490 
    491 static int
    492 route_output_report(struct rtentry *rt, struct rt_addrinfo *info,
    493     struct rt_xmsghdr *rtm, struct rt_xmsghdr **new_rtm)
    494 {
    495 	int len;
    496 	struct ifnet *ifp;
    497 
    498 	if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
    499 		;
    500 	else if ((ifp = rt->rt_ifp) != NULL) {
    501 		const struct ifaddr *rtifa;
    502 		info->rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
    503 		/* rtifa used to be simply rt->rt_ifa.
    504 		 * If rt->rt_ifa != NULL, then
    505 		 * rt_get_ifa() != NULL.  So this
    506 		 * ought to still be safe. --dyoung
    507 		 */
    508 		rtifa = rt_get_ifa(rt);
    509 		info->rti_info[RTAX_IFA] = rtifa->ifa_addr;
    510 #ifdef RTSOCK_DEBUG
    511 		if (info->rti_info[RTAX_IFA]->sa_family == AF_INET) {
    512 			char ibuf[INET_ADDRSTRLEN];
    513 			char abuf[INET_ADDRSTRLEN];
    514 			printf("%s: copying out RTAX_IFA %s "
    515 			    "for info->rti_info[RTAX_DST] %s "
    516 			    "ifa_getifa %p ifa_seqno %p\n",
    517 			    __func__,
    518 			    RT_IN_PRINT(info, ibuf, RTAX_IFA),
    519 			    RT_IN_PRINT(info, abuf, RTAX_DST),
    520 			    (void *)rtifa->ifa_getifa,
    521 			    rtifa->ifa_seqno);
    522 		}
    523 #endif /* RTSOCK_DEBUG */
    524 		if (ifp->if_flags & IFF_POINTOPOINT)
    525 			info->rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
    526 		else
    527 			info->rti_info[RTAX_BRD] = NULL;
    528 		rtm->rtm_index = ifp->if_index;
    529 	} else {
    530 		info->rti_info[RTAX_IFP] = NULL;
    531 		info->rti_info[RTAX_IFA] = NULL;
    532 	}
    533 	(void)rt_msg2(rtm->rtm_type, info, NULL, NULL, &len);
    534 	if (len > rtm->rtm_msglen) {
    535 		struct rt_xmsghdr *old_rtm = rtm;
    536 		R_Malloc(*new_rtm, struct rt_xmsghdr *, len);
    537 		if (*new_rtm == NULL)
    538 			return ENOBUFS;
    539 		(void)memcpy(*new_rtm, old_rtm, old_rtm->rtm_msglen);
    540 		rtm = *new_rtm;
    541 	}
    542 	(void)rt_msg2(rtm->rtm_type, info, rtm, NULL, 0);
    543 	rtm->rtm_flags = rt->rt_flags;
    544 	rtm_setmetrics(rt, rtm);
    545 	rtm->rtm_addrs = info->rti_addrs;
    546 
    547 	return 0;
    548 }
    549 
    550 static struct ifaddr *
    551 route_output_get_ifa(const struct rt_addrinfo info, const struct rtentry *rt,
    552     struct ifnet **ifp, struct psref *psref)
    553 {
    554 	struct ifaddr *ifa = NULL;
    555 
    556 	*ifp = NULL;
    557 	if (info.rti_info[RTAX_IFP] != NULL) {
    558 		ifa = ifa_ifwithnet_psref(info.rti_info[RTAX_IFP], psref);
    559 		if (ifa == NULL)
    560 			goto next;
    561 		*ifp = ifa->ifa_ifp;
    562 		if (info.rti_info[RTAX_IFA] == NULL &&
    563 		    info.rti_info[RTAX_GATEWAY] == NULL)
    564 			goto next;
    565 		if (info.rti_info[RTAX_IFA] == NULL) {
    566 			/* route change <dst> <gw> -ifp <if> */
    567 			ifa = ifaof_ifpforaddr_psref(info.rti_info[RTAX_GATEWAY],
    568 			    *ifp, psref);
    569 		} else {
    570 			/* route change <dst> -ifp <if> -ifa <addr> */
    571 			ifa = ifa_ifwithaddr_psref(info.rti_info[RTAX_IFA], psref);
    572 			if (ifa != NULL)
    573 				goto out;
    574 			ifa = ifaof_ifpforaddr_psref(info.rti_info[RTAX_IFA],
    575 			    *ifp, psref);
    576 		}
    577 		goto out;
    578 	}
    579 next:
    580 	if (info.rti_info[RTAX_IFA] != NULL) {
    581 		/* route change <dst> <gw> -ifa <addr> */
    582 		ifa = ifa_ifwithaddr_psref(info.rti_info[RTAX_IFA], psref);
    583 		if (ifa != NULL)
    584 			goto out;
    585 	}
    586 	if (info.rti_info[RTAX_GATEWAY] != NULL) {
    587 		/* route change <dst> <gw> */
    588 		ifa = ifa_ifwithroute_psref(rt->rt_flags, rt_getkey(rt),
    589 		    info.rti_info[RTAX_GATEWAY], psref);
    590 	}
    591 out:
    592 	if (ifa != NULL && *ifp == NULL)
    593 		*ifp = ifa->ifa_ifp;
    594 	return ifa;
    595 }
    596 
    597 static int
    598 route_output_change(struct rtentry *rt, struct rt_addrinfo *info,
    599     struct rt_xmsghdr *rtm)
    600 {
    601 	int error = 0;
    602 	struct ifnet *ifp, *new_ifp;
    603 	struct ifaddr *ifa, *new_ifa;
    604 	struct psref psref_ifa, psref_new_ifa, psref_ifp;
    605 
    606 	/*
    607 	 * new gateway could require new ifaddr, ifp;
    608 	 * flags may also be different; ifp may be specified
    609 	 * by ll sockaddr when protocol address is ambiguous
    610 	 */
    611 	ifp = rt_getifp(info, &psref_ifp);
    612 	ifa = rt_getifa(info, &psref_ifa);
    613 	if (ifa == NULL) {
    614 		error = ENETUNREACH;
    615 		goto out;
    616 	}
    617 	if (info->rti_info[RTAX_GATEWAY]) {
    618 		error = rt_setgate(rt, info->rti_info[RTAX_GATEWAY]);
    619 		if (error != 0)
    620 			goto out;
    621 	}
    622 	if (info->rti_info[RTAX_TAG]) {
    623 		const struct sockaddr *tag;
    624 		tag = rt_settag(rt, info->rti_info[RTAX_TAG]);
    625 		if (tag == NULL) {
    626 			error = ENOBUFS;
    627 			goto out;
    628 		}
    629 	}
    630 	/* new gateway could require new ifaddr, ifp;
    631 	   flags may also be different; ifp may be specified
    632 	   by ll sockaddr when protocol address is ambiguous */
    633 	new_ifa = route_output_get_ifa(*info, rt, &new_ifp, &psref_new_ifa);
    634 	if (new_ifa != NULL) {
    635 		ifa_release(ifa, &psref_ifa);
    636 		ifa = new_ifa;
    637 	}
    638 	if (ifa) {
    639 		struct ifaddr *oifa = rt->rt_ifa;
    640 		if (oifa != ifa &&
    641 		    !ifa_is_destroying(ifa) && !if_is_deactivated(new_ifp)) {
    642 			if (oifa && oifa->ifa_rtrequest)
    643 				oifa->ifa_rtrequest(RTM_DELETE, rt, info);
    644 			rt_replace_ifa(rt, ifa);
    645 			rt->rt_ifp = new_ifp;
    646 		}
    647 		if (new_ifa == NULL)
    648 			ifa_release(ifa, &psref_ifa);
    649 	}
    650 	ifa_release(new_ifa, &psref_new_ifa);
    651 	if (new_ifp && rt->rt_ifp != new_ifp
    652 	    && !if_is_deactivated(new_ifp))
    653 		rt->rt_ifp = new_ifp;
    654 	rt_setmetrics(rtm->rtm_inits, rtm, rt);
    655 	if (rt->rt_flags != info->rti_flags)
    656 		rt->rt_flags = (info->rti_flags & ~PRESERVED_RTF)
    657 		    | (rt->rt_flags & PRESERVED_RTF);
    658 	if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
    659 		rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
    660 out:
    661 	if_put(ifp, &psref_ifp);
    662 
    663 	return error;
    664 }
    665 
    666 /*ARGSUSED*/
    667 int
    668 COMPATNAME(route_output)(struct mbuf *m, struct socket *so)
    669 {
    670 	struct sockproto proto = { .sp_family = PF_XROUTE, };
    671 	struct rt_xmsghdr *rtm = NULL;
    672 	struct rt_xmsghdr *old_rtm = NULL, *new_rtm = NULL;
    673 	struct rtentry *rt = NULL;
    674 	struct rtentry *saved_nrt = NULL;
    675 	struct rt_addrinfo info;
    676 	int len, error = 0;
    677 	sa_family_t family;
    678 	struct sockaddr_dl sdl;
    679 	int bound = curlwp_bind();
    680 	bool do_rt_free = false;
    681 
    682 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
    683 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
    684 	   (m = m_pullup(m, sizeof(int32_t))) == NULL)) {
    685 		error = ENOBUFS;
    686 		goto out;
    687 	}
    688 	if ((m->m_flags & M_PKTHDR) == 0)
    689 		panic("%s", __func__);
    690 	len = m->m_pkthdr.len;
    691 	if (len < sizeof(*rtm) ||
    692 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
    693 		info.rti_info[RTAX_DST] = NULL;
    694 		senderr(EINVAL);
    695 	}
    696 	R_Malloc(rtm, struct rt_xmsghdr *, len);
    697 	if (rtm == NULL) {
    698 		info.rti_info[RTAX_DST] = NULL;
    699 		senderr(ENOBUFS);
    700 	}
    701 	m_copydata(m, 0, len, rtm);
    702 	if (rtm->rtm_version != RTM_XVERSION) {
    703 		info.rti_info[RTAX_DST] = NULL;
    704 		senderr(EPROTONOSUPPORT);
    705 	}
    706 	rtm->rtm_pid = curproc->p_pid;
    707 	memset(&info, 0, sizeof(info));
    708 	info.rti_addrs = rtm->rtm_addrs;
    709 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
    710 	    &info)) {
    711 		senderr(EINVAL);
    712 	}
    713 	info.rti_flags = rtm->rtm_flags;
    714 #ifdef RTSOCK_DEBUG
    715 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
    716 		char abuf[INET_ADDRSTRLEN];
    717 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
    718 		    RT_IN_PRINT(&info, abuf, RTAX_DST));
    719 	}
    720 #endif /* RTSOCK_DEBUG */
    721 	if (info.rti_info[RTAX_DST] == NULL ||
    722 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
    723 		senderr(EINVAL);
    724 	}
    725 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
    726 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
    727 		senderr(EINVAL);
    728 	}
    729 
    730 	/*
    731 	 * Verify that the caller has the appropriate privilege; RTM_GET
    732 	 * is the only operation the non-superuser is allowed.
    733 	 */
    734 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
    735 	    0, rtm, NULL, NULL) != 0)
    736 		senderr(EACCES);
    737 
    738 	switch (rtm->rtm_type) {
    739 
    740 	case RTM_ADD:
    741 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
    742 			senderr(EINVAL);
    743 		}
    744 #ifdef INET
    745 		/* support for new ARP code with keeping backcompat */
    746 		if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) {
    747 			const struct sockaddr_dl *sdlp =
    748 			    satocsdl(info.rti_info[RTAX_GATEWAY]);
    749 
    750 			/* Allow routing requests by interface index */
    751 			if (sdlp->sdl_nlen == 0 && sdlp->sdl_alen == 0
    752 			    && sdlp->sdl_slen == 0)
    753 				goto fallback;
    754 			/*
    755 			 * Old arp binaries don't set the sdl_index
    756 			 * so we have to complement it.
    757 			 */
    758 			int sdl_index = sdlp->sdl_index;
    759 			if (sdl_index == 0) {
    760 				error = route_get_sdl_index(&info, &sdl_index);
    761 				if (error != 0)
    762 					goto fallback;
    763 			} else if (
    764 			    info.rti_info[RTAX_DST]->sa_family == AF_INET) {
    765 				/*
    766 				 * XXX workaround for SIN_PROXY case; proxy arp
    767 				 * entry should be in an interface that has
    768 				 * a network route including the destination,
    769 				 * not a local (link) route that may not be a
    770 				 * desired place, for example a tap.
    771 				 */
    772 				const struct sockaddr_inarp *sina =
    773 				    (const struct sockaddr_inarp *)
    774 				    info.rti_info[RTAX_DST];
    775 				if (sina->sin_other & SIN_PROXY) {
    776 					error = route_get_sdl_index(&info,
    777 					    &sdl_index);
    778 					if (error != 0)
    779 						goto fallback;
    780 				}
    781 			}
    782 			error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
    783 			    rtm->rtm_rmx.rmx_expire, &info, sdl_index);
    784 			break;
    785 		}
    786 	fallback:
    787 #endif /* INET */
    788 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
    789 		if (error == 0) {
    790 			rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
    791 			rt_unref(saved_nrt);
    792 		}
    793 		break;
    794 
    795 	case RTM_DELETE:
    796 #ifdef INET
    797 		/* support for new ARP code */
    798 		if (info.rti_info[RTAX_GATEWAY] &&
    799 		    (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) &&
    800 		    (rtm->rtm_flags & RTF_LLDATA) != 0) {
    801 			error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags,
    802 			    rtm->rtm_rmx.rmx_expire, &info, 0);
    803 			break;
    804 		}
    805 #endif /* INET */
    806 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
    807 		if (error != 0)
    808 			break;
    809 
    810 		rt = saved_nrt;
    811 		do_rt_free = true;
    812 		info.rti_info[RTAX_DST] = rt_getkey(rt);
    813 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    814 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    815 		info.rti_info[RTAX_TAG] = rt_gettag(rt);
    816 		error = route_output_report(rt, &info, rtm, &new_rtm);
    817 		if (error)
    818 			senderr(error);
    819 		if (new_rtm != NULL) {
    820 			old_rtm = rtm;
    821 			rtm = new_rtm;
    822 		}
    823 		break;
    824 
    825 	case RTM_GET:
    826 	case RTM_CHANGE:
    827 	case RTM_LOCK:
    828                 /* XXX This will mask info.rti_info[RTAX_DST] with
    829 		 * info.rti_info[RTAX_NETMASK] before
    830                  * searching.  It did not used to do that.  --dyoung
    831 		 */
    832 		rt = NULL;
    833 		error = rtrequest1(RTM_GET, &info, &rt);
    834 		if (error != 0)
    835 			senderr(error);
    836 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
    837 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
    838 			    info.rti_info[RTAX_DST]->sa_len) != 0)
    839 				senderr(ESRCH);
    840 			if (info.rti_info[RTAX_NETMASK] == NULL &&
    841 			    rt_mask(rt) != NULL)
    842 				senderr(ETOOMANYREFS);
    843 		}
    844 
    845 		/*
    846 		 * XXX if arp/ndp requests an L2 entry, we have to obtain
    847 		 * it from lltable while for the route command we have to
    848 		 * return a route as it is. How to distinguish them?
    849 		 * For newer arp/ndp, RTF_LLDATA flag set by arp/ndp
    850 		 * indicates an L2 entry is requested. For old arp/ndp
    851 		 * binaries, we check RTF_UP flag is NOT set; it works
    852 		 * by the fact that arp/ndp don't set it while the route
    853 		 * command sets it.
    854 		 */
    855 		if (((rtm->rtm_flags & RTF_LLDATA) != 0 ||
    856 		     (rtm->rtm_flags & RTF_UP) == 0) &&
    857 		    rtm->rtm_type == RTM_GET &&
    858 		    sockaddr_cmp(rt_getkey(rt), info.rti_info[RTAX_DST]) != 0) {
    859 			int ll_flags = 0;
    860 			route_get_sdl(rt->rt_ifp, info.rti_info[RTAX_DST], &sdl,
    861 			    &ll_flags);
    862 			info.rti_info[RTAX_GATEWAY] = sstocsa(&sdl);
    863 			error = route_output_report(rt, &info, rtm, &new_rtm);
    864 			if (error)
    865 				senderr(error);
    866 			if (new_rtm != NULL) {
    867 				old_rtm = rtm;
    868 				rtm = new_rtm;
    869 			}
    870 			rtm->rtm_flags |= RTF_LLDATA;
    871 			rtm->rtm_flags |= (ll_flags & LLE_STATIC) ? RTF_STATIC : 0;
    872 			break;
    873 		}
    874 
    875 		switch (rtm->rtm_type) {
    876 		case RTM_GET:
    877 			info.rti_info[RTAX_DST] = rt_getkey(rt);
    878 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    879 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    880 			info.rti_info[RTAX_TAG] = rt_gettag(rt);
    881 			error = route_output_report(rt, &info, rtm, &new_rtm);
    882 			if (error)
    883 				senderr(error);
    884 			if (new_rtm != NULL) {
    885 				old_rtm = rtm;
    886 				rtm = new_rtm;
    887 			}
    888 			break;
    889 
    890 		case RTM_CHANGE:
    891 #ifdef NET_MPSAFE
    892 			error = rt_update_prepare(rt);
    893 			if (error == 0) {
    894 				error = route_output_change(rt, &info, rtm);
    895 				rt_update_finish(rt);
    896 			}
    897 #else
    898 			error = route_output_change(rt, &info, rtm);
    899 #endif
    900 			if (error != 0)
    901 				goto flush;
    902 			/*FALLTHROUGH*/
    903 		case RTM_LOCK:
    904 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
    905 			rt->rt_rmx.rmx_locks |=
    906 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
    907 			break;
    908 		}
    909 		break;
    910 
    911 	default:
    912 		senderr(EOPNOTSUPP);
    913 	}
    914 
    915 flush:
    916 	if (rtm) {
    917 		if (error)
    918 			rtm->rtm_errno = error;
    919 		else
    920 			rtm->rtm_flags |= RTF_DONE;
    921 	}
    922 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
    923 	    0;
    924 	/* We cannot free old_rtm until we have stopped using the
    925 	 * pointers in info, some of which may point to sockaddrs
    926 	 * in old_rtm.
    927 	 */
    928 	if (old_rtm != NULL)
    929 		Free(old_rtm);
    930 	if (rt) {
    931 		if (do_rt_free)
    932 			rt_free(rt);
    933 		else
    934 			rt_unref(rt);
    935 	}
    936     {
    937 	struct rawcb *rp = NULL;
    938 	/*
    939 	 * Check to see if we don't want our own messages.
    940 	 */
    941 	if ((so->so_options & SO_USELOOPBACK) == 0) {
    942 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
    943 			if (rtm)
    944 				Free(rtm);
    945 			m_freem(m);
    946 			goto out;
    947 		}
    948 		/* There is another listener, so construct message */
    949 		rp = sotorawcb(so);
    950 	}
    951 	if (rtm) {
    952 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
    953 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
    954 			m_freem(m);
    955 			m = NULL;
    956 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
    957 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
    958 		Free(rtm);
    959 	}
    960 	if (rp)
    961 		rp->rcb_proto.sp_family = 0; /* Avoid us */
    962 	if (family)
    963 		proto.sp_protocol = family;
    964 	if (m)
    965 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
    966 		    &COMPATNAME(route_info).ri_dst);
    967 	if (rp)
    968 		rp->rcb_proto.sp_family = PF_XROUTE;
    969     }
    970 out:
    971 	curlwp_bindx(bound);
    972 	return error;
    973 }
    974 
    975 static void
    976 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
    977 {
    978 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
    979 	metric(RTV_RPIPE, rmx_recvpipe);
    980 	metric(RTV_SPIPE, rmx_sendpipe);
    981 	metric(RTV_SSTHRESH, rmx_ssthresh);
    982 	metric(RTV_RTT, rmx_rtt);
    983 	metric(RTV_RTTVAR, rmx_rttvar);
    984 	metric(RTV_HOPCOUNT, rmx_hopcount);
    985 	metric(RTV_MTU, rmx_mtu);
    986 #undef metric
    987 	if (which & RTV_EXPIRE) {
    988 		out->rt_rmx.rmx_expire = in->rtm_rmx.rmx_expire ?
    989 		    time_wall_to_mono(in->rtm_rmx.rmx_expire) : 0;
    990 	}
    991 }
    992 
    993 static void
    994 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
    995 {
    996 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
    997 	metric(rmx_recvpipe);
    998 	metric(rmx_sendpipe);
    999 	metric(rmx_ssthresh);
   1000 	metric(rmx_rtt);
   1001 	metric(rmx_rttvar);
   1002 	metric(rmx_hopcount);
   1003 	metric(rmx_mtu);
   1004 #undef metric
   1005 	out->rtm_rmx.rmx_expire = in->rt_rmx.rmx_expire ?
   1006 	    time_mono_to_wall(in->rt_rmx.rmx_expire) : 0;
   1007 }
   1008 
   1009 static int
   1010 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
   1011     struct rt_addrinfo *rtinfo)
   1012 {
   1013 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
   1014 	int i;
   1015 
   1016 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
   1017 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
   1018 			continue;
   1019 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
   1020 		RT_XADVANCE(cp, sa);
   1021 	}
   1022 
   1023 	/*
   1024 	 * Check for extra addresses specified, except RTM_GET asking
   1025 	 * for interface info.
   1026 	 */
   1027 	if (rtmtype == RTM_GET) {
   1028 		if (((rtinfo->rti_addrs &
   1029 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0U << i)) != 0)
   1030 			return 1;
   1031 	} else if ((rtinfo->rti_addrs & (~0U << i)) != 0)
   1032 		return 1;
   1033 	/* Check for bad data length.  */
   1034 	if (cp != cplim) {
   1035 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
   1036 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
   1037 			/*
   1038 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
   1039 			 * We accept this for now for the sake of old
   1040 			 * binaries or third party softwares.
   1041 			 */
   1042 			;
   1043 		else
   1044 			return 1;
   1045 	}
   1046 	return 0;
   1047 }
   1048 
   1049 static int
   1050 rt_getlen(int type)
   1051 {
   1052 #ifndef COMPAT_RTSOCK
   1053 	CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
   1054 	CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
   1055 	CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
   1056 	CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
   1057 #endif
   1058 
   1059 	switch (type) {
   1060 	case RTM_ODELADDR:
   1061 	case RTM_ONEWADDR:
   1062 	case RTM_OCHGADDR:
   1063 #ifdef COMPAT_70
   1064 		return sizeof(struct ifa_msghdr70);
   1065 #else
   1066 #ifdef DIAGNOSTIC
   1067 		printf("RTM_ONEWADDR\n");
   1068 #endif
   1069 		return -1;
   1070 #endif
   1071 	case RTM_DELADDR:
   1072 	case RTM_NEWADDR:
   1073 	case RTM_CHGADDR:
   1074 		return sizeof(struct ifa_xmsghdr);
   1075 
   1076 	case RTM_OOIFINFO:
   1077 #ifdef COMPAT_14
   1078 		return sizeof(struct if_msghdr14);
   1079 #else
   1080 #ifdef DIAGNOSTIC
   1081 		printf("RTM_OOIFINFO\n");
   1082 #endif
   1083 		return -1;
   1084 #endif
   1085 	case RTM_OIFINFO:
   1086 #ifdef COMPAT_50
   1087 		return sizeof(struct if_msghdr50);
   1088 #else
   1089 #ifdef DIAGNOSTIC
   1090 		printf("RTM_OIFINFO\n");
   1091 #endif
   1092 		return -1;
   1093 #endif
   1094 
   1095 	case RTM_IFINFO:
   1096 		return sizeof(struct if_xmsghdr);
   1097 
   1098 	case RTM_IFANNOUNCE:
   1099 	case RTM_IEEE80211:
   1100 		return sizeof(struct if_xannouncemsghdr);
   1101 
   1102 	default:
   1103 		return sizeof(struct rt_xmsghdr);
   1104 	}
   1105 }
   1106 
   1107 
   1108 struct mbuf *
   1109 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
   1110 {
   1111 	struct rt_xmsghdr *rtm;
   1112 	struct mbuf *m;
   1113 	int i;
   1114 	const struct sockaddr *sa;
   1115 	int len, dlen;
   1116 
   1117 	m = m_gethdr(M_DONTWAIT, MT_DATA);
   1118 	if (m == NULL)
   1119 		return m;
   1120 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
   1121 
   1122 	if ((len = rt_getlen(type)) == -1)
   1123 		goto out;
   1124 	if (len > MHLEN + MLEN)
   1125 		panic("%s: message too long", __func__);
   1126 	else if (len > MHLEN) {
   1127 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
   1128 		if (m->m_next == NULL)
   1129 			goto out;
   1130 		MCLAIM(m->m_next, m->m_owner);
   1131 		m->m_pkthdr.len = len;
   1132 		m->m_len = MHLEN;
   1133 		m->m_next->m_len = len - MHLEN;
   1134 	} else {
   1135 		m->m_pkthdr.len = m->m_len = len;
   1136 	}
   1137 	m_reset_rcvif(m);
   1138 	m_copyback(m, 0, datalen, data);
   1139 	if (len > datalen)
   1140 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
   1141 	rtm = mtod(m, struct rt_xmsghdr *);
   1142 	for (i = 0; i < RTAX_MAX; i++) {
   1143 		if ((sa = rtinfo->rti_info[i]) == NULL)
   1144 			continue;
   1145 		rtinfo->rti_addrs |= (1 << i);
   1146 		dlen = RT_XROUNDUP(sa->sa_len);
   1147 		m_copyback(m, len, sa->sa_len, sa);
   1148 		if (dlen != sa->sa_len) {
   1149 			/*
   1150 			 * Up to 6 + 1 nul's since roundup is to
   1151 			 * sizeof(uint64_t) (8 bytes)
   1152 			 */
   1153 			m_copyback(m, len + sa->sa_len,
   1154 			    dlen - sa->sa_len, "\0\0\0\0\0\0");
   1155 		}
   1156 		len += dlen;
   1157 	}
   1158 	if (m->m_pkthdr.len != len)
   1159 		goto out;
   1160 	rtm->rtm_msglen = len;
   1161 	rtm->rtm_version = RTM_XVERSION;
   1162 	rtm->rtm_type = type;
   1163 	return m;
   1164 out:
   1165 	m_freem(m);
   1166 	return NULL;
   1167 }
   1168 
   1169 /*
   1170  * rt_msg2
   1171  *
   1172  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
   1173  *		returns the length of the message in 'lenp'.
   1174  *
   1175  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
   1176  *	the message
   1177  * otherwise walkarg's w_needed is updated and if the user buffer is
   1178  *	specified and w_needed indicates space exists the information is copied
   1179  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
   1180  *	if the allocation fails ENOBUFS is returned.
   1181  */
   1182 static int
   1183 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
   1184 	int *lenp)
   1185 {
   1186 	int i;
   1187 	int len, dlen, second_time = 0;
   1188 	char *cp0, *cp = cpv;
   1189 
   1190 	rtinfo->rti_addrs = 0;
   1191 again:
   1192 	if ((len = rt_getlen(type)) == -1)
   1193 		return EINVAL;
   1194 
   1195 	if ((cp0 = cp) != NULL)
   1196 		cp += len;
   1197 	for (i = 0; i < RTAX_MAX; i++) {
   1198 		const struct sockaddr *sa;
   1199 
   1200 		if ((sa = rtinfo->rti_info[i]) == NULL)
   1201 			continue;
   1202 		rtinfo->rti_addrs |= (1 << i);
   1203 		dlen = RT_XROUNDUP(sa->sa_len);
   1204 		if (cp) {
   1205 			int diff = dlen - sa->sa_len;
   1206 			(void)memcpy(cp, sa, (size_t)sa->sa_len);
   1207 			cp += sa->sa_len;
   1208 			if (diff > 0) {
   1209 				(void)memset(cp, 0, (size_t)diff);
   1210 				cp += diff;
   1211 			}
   1212 		}
   1213 		len += dlen;
   1214 	}
   1215 	if (cp == NULL && w != NULL && !second_time) {
   1216 		struct rt_walkarg *rw = w;
   1217 
   1218 		rw->w_needed += len;
   1219 		if (rw->w_needed <= 0 && rw->w_where) {
   1220 			if (rw->w_tmemsize < len) {
   1221 				if (rw->w_tmem)
   1222 					free(rw->w_tmem, M_RTABLE);
   1223 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
   1224 				if (rw->w_tmem)
   1225 					rw->w_tmemsize = len;
   1226 				else
   1227 					rw->w_tmemsize = 0;
   1228 			}
   1229 			if (rw->w_tmem) {
   1230 				cp = rw->w_tmem;
   1231 				second_time = 1;
   1232 				goto again;
   1233 			} else {
   1234 				rw->w_tmemneeded = len;
   1235 				return ENOBUFS;
   1236 			}
   1237 		}
   1238 	}
   1239 	if (cp) {
   1240 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
   1241 
   1242 		rtm->rtm_version = RTM_XVERSION;
   1243 		rtm->rtm_type = type;
   1244 		rtm->rtm_msglen = len;
   1245 	}
   1246 	if (lenp)
   1247 		*lenp = len;
   1248 	return 0;
   1249 }
   1250 
   1251 #ifndef COMPAT_RTSOCK
   1252 int
   1253 rt_msg3(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
   1254 	int *lenp)
   1255 {
   1256 	return rt_msg2(type, rtinfo, cpv, w, lenp);
   1257 }
   1258 #endif
   1259 
   1260 /*
   1261  * This routine is called to generate a message from the routing
   1262  * socket indicating that a redirect has occurred, a routing lookup
   1263  * has failed, or that a protocol has detected timeouts to a particular
   1264  * destination.
   1265  */
   1266 void
   1267 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
   1268     int error)
   1269 {
   1270 	struct rt_xmsghdr rtm;
   1271 	struct mbuf *m;
   1272 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
   1273 	struct rt_addrinfo info = *rtinfo;
   1274 
   1275 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
   1276 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
   1277 		return;
   1278 	memset(&rtm, 0, sizeof(rtm));
   1279 	rtm.rtm_pid = curproc->p_pid;
   1280 	rtm.rtm_flags = RTF_DONE | flags;
   1281 	rtm.rtm_errno = error;
   1282 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
   1283 	if (m == NULL)
   1284 		return;
   1285 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
   1286 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
   1287 }
   1288 
   1289 /*
   1290  * This routine is called to generate a message from the routing
   1291  * socket indicating that the status of a network interface has changed.
   1292  */
   1293 void
   1294 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
   1295 {
   1296 	struct if_xmsghdr ifm;
   1297 	struct mbuf *m;
   1298 	struct rt_addrinfo info;
   1299 
   1300 	COMPATCALL(rt_ifmsg, (ifp));
   1301 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
   1302 		return;
   1303 	(void)memset(&info, 0, sizeof(info));
   1304 	(void)memset(&ifm, 0, sizeof(ifm));
   1305 	ifm.ifm_index = ifp->if_index;
   1306 	ifm.ifm_flags = ifp->if_flags;
   1307 	ifm.ifm_data = ifp->if_data;
   1308 	ifm.ifm_addrs = 0;
   1309 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
   1310 	if (m == NULL)
   1311 		return;
   1312 	COMPATNAME(route_enqueue)(m, 0);
   1313 #ifdef COMPAT_14
   1314 	compat_14_rt_oifmsg(ifp);
   1315 #endif
   1316 #ifdef COMPAT_50
   1317 	compat_50_rt_oifmsg(ifp);
   1318 #endif
   1319 }
   1320 
   1321 #ifndef COMPAT_RTSOCK
   1322 static int
   1323 if_addrflags(struct ifaddr *ifa)
   1324 {
   1325 
   1326 	switch (ifa->ifa_addr->sa_family) {
   1327 #ifdef INET
   1328 	case AF_INET:
   1329 		return ((struct in_ifaddr *)ifa)->ia4_flags;
   1330 #endif
   1331 #ifdef INET6
   1332 	case AF_INET6:
   1333 		return ((struct in6_ifaddr *)ifa)->ia6_flags;
   1334 #endif
   1335 	default:
   1336 		return 0;
   1337 	}
   1338 }
   1339 #endif
   1340 
   1341 /*
   1342  * This is called to generate messages from the routing socket
   1343  * indicating a network interface has had addresses associated with it.
   1344  * if we ever reverse the logic and replace messages TO the routing
   1345  * socket indicate a request to configure interfaces, then it will
   1346  * be unnecessary as the routing socket will automatically generate
   1347  * copies of it.
   1348  */
   1349 void
   1350 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
   1351     struct rtentry *rt)
   1352 {
   1353 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
   1354 	struct rt_addrinfo info;
   1355 	const struct sockaddr *sa;
   1356 	int pass;
   1357 	struct mbuf *m;
   1358 	struct ifnet *ifp;
   1359 	struct rt_xmsghdr rtm;
   1360 	struct ifa_xmsghdr ifam;
   1361 	int ncmd;
   1362 
   1363 	KASSERT(ifa != NULL);
   1364 	KASSERT(ifa->ifa_addr != NULL);
   1365 	ifp = ifa->ifa_ifp;
   1366 #ifdef SCTP
   1367 	if (cmd == RTM_ADD) {
   1368 		sctp_add_ip_address(ifa);
   1369 	} else if (cmd == RTM_DELETE) {
   1370 		sctp_delete_ip_address(ifa);
   1371 	}
   1372 #endif
   1373 
   1374 	COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
   1375 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
   1376 		return;
   1377 	for (pass = 1; pass < 3; pass++) {
   1378 		memset(&info, 0, sizeof(info));
   1379 		switch (cmdpass(cmd, pass)) {
   1380 		case cmdpass(RTM_ADD, 1):
   1381 		case cmdpass(RTM_CHANGE, 1):
   1382 		case cmdpass(RTM_DELETE, 2):
   1383 		case cmdpass(RTM_NEWADDR, 1):
   1384 		case cmdpass(RTM_DELADDR, 1):
   1385 		case cmdpass(RTM_CHGADDR, 1):
   1386 			switch (cmd) {
   1387 			case RTM_ADD:
   1388 				ncmd = RTM_XNEWADDR;
   1389 				break;
   1390 			case RTM_DELETE:
   1391 				ncmd = RTM_XDELADDR;
   1392 				break;
   1393 			case RTM_CHANGE:
   1394 				ncmd = RTM_XCHGADDR;
   1395 				break;
   1396 			case RTM_NEWADDR:
   1397 				ncmd = RTM_XNEWADDR;
   1398 				break;
   1399 			case RTM_DELADDR:
   1400 				ncmd = RTM_XDELADDR;
   1401 				break;
   1402 			case RTM_CHGADDR:
   1403 				ncmd = RTM_XCHGADDR;
   1404 				break;
   1405 			default:
   1406 				panic("%s: unknown command %d", __func__, cmd);
   1407 			}
   1408 #ifdef COMPAT_70
   1409 			compat_70_rt_newaddrmsg1(ncmd, ifa);
   1410 #endif
   1411 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
   1412 			KASSERT(ifp->if_dl != NULL);
   1413 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
   1414 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
   1415 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
   1416 			memset(&ifam, 0, sizeof(ifam));
   1417 			ifam.ifam_index = ifp->if_index;
   1418 			ifam.ifam_metric = ifa->ifa_metric;
   1419 			ifam.ifam_flags = ifa->ifa_flags;
   1420 #ifndef COMPAT_RTSOCK
   1421 			ifam.ifam_pid = curproc->p_pid;
   1422 			ifam.ifam_addrflags = if_addrflags(ifa);
   1423 #endif
   1424 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
   1425 			if (m == NULL)
   1426 				continue;
   1427 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
   1428 			    info.rti_addrs;
   1429 			break;
   1430 		case cmdpass(RTM_ADD, 2):
   1431 		case cmdpass(RTM_CHANGE, 2):
   1432 		case cmdpass(RTM_DELETE, 1):
   1433 			if (rt == NULL)
   1434 				continue;
   1435 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
   1436 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
   1437 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
   1438 			memset(&rtm, 0, sizeof(rtm));
   1439 			rtm.rtm_pid = curproc->p_pid;
   1440 			rtm.rtm_index = ifp->if_index;
   1441 			rtm.rtm_flags |= rt->rt_flags;
   1442 			rtm.rtm_errno = error;
   1443 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
   1444 			if (m == NULL)
   1445 				continue;
   1446 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
   1447 			break;
   1448 		default:
   1449 			continue;
   1450 		}
   1451 #ifdef DIAGNOSTIC
   1452 		if (m == NULL)
   1453 			panic("%s: called with wrong command", __func__);
   1454 #endif
   1455 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
   1456 	}
   1457 #undef cmdpass
   1458 
   1459 }
   1460 
   1461 static struct mbuf *
   1462 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
   1463     struct rt_addrinfo *info)
   1464 {
   1465 	struct if_xannouncemsghdr ifan;
   1466 
   1467 	memset(info, 0, sizeof(*info));
   1468 	memset(&ifan, 0, sizeof(ifan));
   1469 	ifan.ifan_index = ifp->if_index;
   1470 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
   1471 	ifan.ifan_what = what;
   1472 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
   1473 }
   1474 
   1475 /*
   1476  * This is called to generate routing socket messages indicating
   1477  * network interface arrival and departure.
   1478  */
   1479 void
   1480 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
   1481 {
   1482 	struct mbuf *m;
   1483 	struct rt_addrinfo info;
   1484 
   1485 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
   1486 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
   1487 		return;
   1488 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
   1489 	if (m == NULL)
   1490 		return;
   1491 	COMPATNAME(route_enqueue)(m, 0);
   1492 }
   1493 
   1494 /*
   1495  * This is called to generate routing socket messages indicating
   1496  * IEEE80211 wireless events.
   1497  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
   1498  */
   1499 void
   1500 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
   1501 	size_t data_len)
   1502 {
   1503 	struct mbuf *m;
   1504 	struct rt_addrinfo info;
   1505 
   1506 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
   1507 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
   1508 		return;
   1509 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
   1510 	if (m == NULL)
   1511 		return;
   1512 	/*
   1513 	 * Append the ieee80211 data.  Try to stick it in the
   1514 	 * mbuf containing the ifannounce msg; otherwise allocate
   1515 	 * a new mbuf and append.
   1516 	 *
   1517 	 * NB: we assume m is a single mbuf.
   1518 	 */
   1519 	if (data_len > M_TRAILINGSPACE(m)) {
   1520 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
   1521 		if (n == NULL) {
   1522 			m_freem(m);
   1523 			return;
   1524 		}
   1525 		(void)memcpy(mtod(n, void *), data, data_len);
   1526 		n->m_len = data_len;
   1527 		m->m_next = n;
   1528 	} else if (data_len > 0) {
   1529 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
   1530 		m->m_len += data_len;
   1531 	}
   1532 	if (m->m_flags & M_PKTHDR)
   1533 		m->m_pkthdr.len += data_len;
   1534 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
   1535 	COMPATNAME(route_enqueue)(m, 0);
   1536 }
   1537 
   1538 /*
   1539  * This is used in dumping the kernel table via sysctl().
   1540  */
   1541 static int
   1542 sysctl_dumpentry(struct rtentry *rt, void *v)
   1543 {
   1544 	struct rt_walkarg *w = v;
   1545 	int error = 0, size;
   1546 	struct rt_addrinfo info;
   1547 
   1548 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
   1549 		return 0;
   1550 	memset(&info, 0, sizeof(info));
   1551 	info.rti_info[RTAX_DST] = rt_getkey(rt);
   1552 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
   1553 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
   1554 	info.rti_info[RTAX_TAG] = rt_gettag(rt);
   1555 	if (rt->rt_ifp) {
   1556 		const struct ifaddr *rtifa;
   1557 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
   1558 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
   1559 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
   1560 		 * --dyoung
   1561 		 */
   1562 		rtifa = rt_get_ifa(rt);
   1563 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
   1564 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
   1565 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
   1566 	}
   1567 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
   1568 		return error;
   1569 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1570 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
   1571 
   1572 		rtm->rtm_flags = rt->rt_flags;
   1573 		rtm->rtm_use = rt->rt_use;
   1574 		rtm_setmetrics(rt, rtm);
   1575 		KASSERT(rt->rt_ifp != NULL);
   1576 		rtm->rtm_index = rt->rt_ifp->if_index;
   1577 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
   1578 		rtm->rtm_addrs = info.rti_addrs;
   1579 		if ((error = copyout(rtm, w->w_where, size)) != 0)
   1580 			w->w_where = NULL;
   1581 		else
   1582 			w->w_where = (char *)w->w_where + size;
   1583 	}
   1584 	return error;
   1585 }
   1586 
   1587 static int
   1588 sysctl_iflist_if(struct ifnet *ifp, struct rt_walkarg *w,
   1589     struct rt_addrinfo *info, size_t len)
   1590 {
   1591 	struct if_xmsghdr *ifm;
   1592 	int error;
   1593 
   1594 	ifm = (struct if_xmsghdr *)w->w_tmem;
   1595 	ifm->ifm_index = ifp->if_index;
   1596 	ifm->ifm_flags = ifp->if_flags;
   1597 	ifm->ifm_data = ifp->if_data;
   1598 	ifm->ifm_addrs = info->rti_addrs;
   1599 	if ((error = copyout(ifm, w->w_where, len)) == 0)
   1600 		w->w_where = (char *)w->w_where + len;
   1601 	return error;
   1602 }
   1603 
   1604 static int
   1605 sysctl_iflist_addr(struct rt_walkarg *w, struct ifaddr *ifa,
   1606      struct rt_addrinfo *info)
   1607 {
   1608 	int len, error;
   1609 
   1610 	if ((error = rt_msg2(RTM_XNEWADDR, info, 0, w, &len)))
   1611 		return error;
   1612 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1613 		struct ifa_xmsghdr *ifam;
   1614 
   1615 		ifam = (struct ifa_xmsghdr *)w->w_tmem;
   1616 		ifam->ifam_index = ifa->ifa_ifp->if_index;
   1617 		ifam->ifam_flags = ifa->ifa_flags;
   1618 		ifam->ifam_metric = ifa->ifa_metric;
   1619 		ifam->ifam_addrs = info->rti_addrs;
   1620 #ifndef COMPAT_RTSOCK
   1621 		ifam->ifam_pid = 0;
   1622 		ifam->ifam_addrflags = if_addrflags(ifa);
   1623 #endif
   1624 		if ((error = copyout(w->w_tmem, w->w_where, len)) == 0)
   1625 			w->w_where = (char *)w->w_where + len;
   1626 	}
   1627 	return error;
   1628 }
   1629 
   1630 static int
   1631 sysctl_iflist(int af, struct rt_walkarg *w, int type)
   1632 {
   1633 	struct ifnet *ifp;
   1634 	struct ifaddr *ifa;
   1635 	struct	rt_addrinfo info;
   1636 	int	cmd, len, error = 0;
   1637 	int	(*iflist_if)(struct ifnet *, struct rt_walkarg *,
   1638 			     struct rt_addrinfo *, size_t);
   1639 	int	(*iflist_addr)(struct rt_walkarg *, struct ifaddr *,
   1640 			       struct rt_addrinfo *);
   1641 	int s;
   1642 	struct psref psref;
   1643 	int bound = curlwp_bind();
   1644 
   1645 	switch (type) {
   1646 	case NET_RT_IFLIST:
   1647 		cmd = RTM_IFINFO;
   1648 		iflist_if = sysctl_iflist_if;
   1649 		iflist_addr = sysctl_iflist_addr;
   1650 		break;
   1651 #ifdef COMPAT_14
   1652 	case NET_RT_OOOIFLIST:
   1653 		cmd = RTM_OOIFINFO;
   1654 		iflist_if = compat_14_iflist;
   1655 		iflist_addr = compat_70_iflist_addr;
   1656 		break;
   1657 #endif
   1658 #ifdef COMPAT_50
   1659 	case NET_RT_OOIFLIST:
   1660 		cmd = RTM_OIFINFO;
   1661 		iflist_if = compat_50_iflist;
   1662 		iflist_addr = compat_70_iflist_addr;
   1663 		break;
   1664 #endif
   1665 #ifdef COMPAT_70
   1666 	case NET_RT_OIFLIST:
   1667 		cmd = RTM_IFINFO;
   1668 		iflist_if = sysctl_iflist_if;
   1669 		iflist_addr = compat_70_iflist_addr;
   1670 		break;
   1671 #endif
   1672 	default:
   1673 #ifdef DIAGNOSTIC
   1674 		printf("sysctl_iflist\n");
   1675 #endif
   1676 		return EINVAL;
   1677 	}
   1678 
   1679 	memset(&info, 0, sizeof(info));
   1680 
   1681 	s = pserialize_read_enter();
   1682 	IFNET_READER_FOREACH(ifp) {
   1683 		if (w->w_arg && w->w_arg != ifp->if_index)
   1684 			continue;
   1685 		if (IFADDR_READER_EMPTY(ifp))
   1686 			continue;
   1687 
   1688 		psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class);
   1689 		pserialize_read_exit(s);
   1690 
   1691 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
   1692 		if ((error = rt_msg2(cmd, &info, NULL, w, &len)) != 0)
   1693 			goto release_exit;
   1694 		info.rti_info[RTAX_IFP] = NULL;
   1695 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1696 			if ((error = iflist_if(ifp, w, &info, len)) != 0)
   1697 				goto release_exit;
   1698 		}
   1699 		IFADDR_READER_FOREACH(ifa, ifp) {
   1700 			if (af && af != ifa->ifa_addr->sa_family)
   1701 				continue;
   1702 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
   1703 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
   1704 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
   1705 			if ((error = iflist_addr(w, ifa, &info)) != 0)
   1706 				goto release_exit;
   1707 		}
   1708 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
   1709 		    info.rti_info[RTAX_BRD] = NULL;
   1710 
   1711 		s = pserialize_read_enter();
   1712 		psref_release(&psref, &ifp->if_psref, ifnet_psref_class);
   1713 	}
   1714 	pserialize_read_exit(s);
   1715 	curlwp_bindx(bound);
   1716 
   1717 	return 0;
   1718 
   1719 release_exit:
   1720 	psref_release(&psref, &ifp->if_psref, ifnet_psref_class);
   1721 	curlwp_bindx(bound);
   1722 	return error;
   1723 }
   1724 
   1725 static int
   1726 sysctl_rtable(SYSCTLFN_ARGS)
   1727 {
   1728 	void 	*where = oldp;
   1729 	size_t	*given = oldlenp;
   1730 	int	i, s, error = EINVAL;
   1731 	u_char  af;
   1732 	struct	rt_walkarg w;
   1733 
   1734 	if (namelen == 1 && name[0] == CTL_QUERY)
   1735 		return sysctl_query(SYSCTLFN_CALL(rnode));
   1736 
   1737 	if (newp)
   1738 		return EPERM;
   1739 	if (namelen != 3)
   1740 		return EINVAL;
   1741 	af = name[0];
   1742 	w.w_tmemneeded = 0;
   1743 	w.w_tmemsize = 0;
   1744 	w.w_tmem = NULL;
   1745 again:
   1746 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
   1747 	if (w.w_tmemneeded) {
   1748 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
   1749 		w.w_tmemsize = w.w_tmemneeded;
   1750 		w.w_tmemneeded = 0;
   1751 	}
   1752 	w.w_op = name[1];
   1753 	w.w_arg = name[2];
   1754 	w.w_given = *given;
   1755 	w.w_needed = 0 - w.w_given;
   1756 	w.w_where = where;
   1757 
   1758 	s = splsoftnet();
   1759 	switch (w.w_op) {
   1760 
   1761 	case NET_RT_DUMP:
   1762 	case NET_RT_FLAGS:
   1763 #ifdef INET
   1764 		/*
   1765 		 * take care of llinfo entries, the caller must
   1766 		 * specify an AF
   1767 		 */
   1768 		if (w.w_op == NET_RT_FLAGS &&
   1769 		    (w.w_arg == 0 || w.w_arg & RTF_LLDATA)) {
   1770 			if (af != 0)
   1771 				error = lltable_sysctl_dumparp(af, &w);
   1772 			else
   1773 				error = EINVAL;
   1774 			break;
   1775 		}
   1776 #endif /* INET */
   1777 
   1778 		for (i = 1; i <= AF_MAX; i++)
   1779 			if ((af == 0 || af == i) &&
   1780 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
   1781 				break;
   1782 		break;
   1783 
   1784 #ifdef COMPAT_14
   1785 	case NET_RT_OOOIFLIST:
   1786 		error = sysctl_iflist(af, &w, w.w_op);
   1787 		break;
   1788 #endif
   1789 #ifdef COMPAT_50
   1790 	case NET_RT_OOIFLIST:
   1791 		error = sysctl_iflist(af, &w, w.w_op);
   1792 		break;
   1793 #endif
   1794 #ifdef COMPAT_70
   1795 	case NET_RT_OIFLIST:
   1796 		error = sysctl_iflist(af, &w, w.w_op);
   1797 		break;
   1798 #endif
   1799 	case NET_RT_IFLIST:
   1800 		error = sysctl_iflist(af, &w, w.w_op);
   1801 		break;
   1802 	}
   1803 	splx(s);
   1804 
   1805 	/* check to see if we couldn't allocate memory with NOWAIT */
   1806 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
   1807 		goto again;
   1808 
   1809 	if (w.w_tmem)
   1810 		free(w.w_tmem, M_RTABLE);
   1811 	w.w_needed += w.w_given;
   1812 	if (where) {
   1813 		*given = (char *)w.w_where - (char *)where;
   1814 		if (*given < w.w_needed)
   1815 			return ENOMEM;
   1816 	} else {
   1817 		*given = (11 * w.w_needed) / 10;
   1818 	}
   1819 	return error;
   1820 }
   1821 
   1822 /*
   1823  * Routing message software interrupt routine
   1824  */
   1825 static void
   1826 COMPATNAME(route_intr)(void *cookie)
   1827 {
   1828 	struct sockproto proto = { .sp_family = PF_XROUTE, };
   1829 	struct route_info * const ri = &COMPATNAME(route_info);
   1830 	struct mbuf *m;
   1831 
   1832 	mutex_enter(softnet_lock);
   1833 	KERNEL_LOCK(1, NULL);
   1834 	for (;;) {
   1835 		IFQ_LOCK(&ri->ri_intrq);
   1836 		IF_DEQUEUE(&ri->ri_intrq, m);
   1837 		IFQ_UNLOCK(&ri->ri_intrq);
   1838 		if (m == NULL)
   1839 			break;
   1840 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
   1841 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst);
   1842 	}
   1843 	KERNEL_UNLOCK_ONE(NULL);
   1844 	mutex_exit(softnet_lock);
   1845 }
   1846 
   1847 /*
   1848  * Enqueue a message to the software interrupt routine.
   1849  */
   1850 void
   1851 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
   1852 {
   1853 	struct route_info * const ri = &COMPATNAME(route_info);
   1854 	int wasempty;
   1855 
   1856 	IFQ_LOCK(&ri->ri_intrq);
   1857 	if (IF_QFULL(&ri->ri_intrq)) {
   1858 		IF_DROP(&ri->ri_intrq);
   1859 		IFQ_UNLOCK(&ri->ri_intrq);
   1860 		m_freem(m);
   1861 	} else {
   1862 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
   1863 		M_SETCTX(m, (uintptr_t)family);
   1864 		IF_ENQUEUE(&ri->ri_intrq, m);
   1865 		IFQ_UNLOCK(&ri->ri_intrq);
   1866 		if (wasempty) {
   1867 			kpreempt_disable();
   1868 			softint_schedule(ri->ri_sih);
   1869 			kpreempt_enable();
   1870 		}
   1871 	}
   1872 }
   1873 
   1874 static void
   1875 COMPATNAME(route_init)(void)
   1876 {
   1877 	struct route_info * const ri = &COMPATNAME(route_info);
   1878 
   1879 #ifndef COMPAT_RTSOCK
   1880 	rt_init();
   1881 #endif
   1882 
   1883 	sysctl_net_route_setup(NULL);
   1884 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
   1885 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
   1886 	    COMPATNAME(route_intr), NULL);
   1887 	IFQ_LOCK_INIT(&ri->ri_intrq);
   1888 }
   1889 
   1890 /*
   1891  * Definitions of protocols supported in the ROUTE domain.
   1892  */
   1893 #ifndef COMPAT_RTSOCK
   1894 PR_WRAP_USRREQS(route);
   1895 #else
   1896 PR_WRAP_USRREQS(compat_50_route);
   1897 #endif
   1898 
   1899 static const struct pr_usrreqs route_usrreqs = {
   1900 	.pr_attach	= COMPATNAME(route_attach_wrapper),
   1901 	.pr_detach	= COMPATNAME(route_detach_wrapper),
   1902 	.pr_accept	= COMPATNAME(route_accept_wrapper),
   1903 	.pr_bind	= COMPATNAME(route_bind_wrapper),
   1904 	.pr_listen	= COMPATNAME(route_listen_wrapper),
   1905 	.pr_connect	= COMPATNAME(route_connect_wrapper),
   1906 	.pr_connect2	= COMPATNAME(route_connect2_wrapper),
   1907 	.pr_disconnect	= COMPATNAME(route_disconnect_wrapper),
   1908 	.pr_shutdown	= COMPATNAME(route_shutdown_wrapper),
   1909 	.pr_abort	= COMPATNAME(route_abort_wrapper),
   1910 	.pr_ioctl	= COMPATNAME(route_ioctl_wrapper),
   1911 	.pr_stat	= COMPATNAME(route_stat_wrapper),
   1912 	.pr_peeraddr	= COMPATNAME(route_peeraddr_wrapper),
   1913 	.pr_sockaddr	= COMPATNAME(route_sockaddr_wrapper),
   1914 	.pr_rcvd	= COMPATNAME(route_rcvd_wrapper),
   1915 	.pr_recvoob	= COMPATNAME(route_recvoob_wrapper),
   1916 	.pr_send	= COMPATNAME(route_send_wrapper),
   1917 	.pr_sendoob	= COMPATNAME(route_sendoob_wrapper),
   1918 	.pr_purgeif	= COMPATNAME(route_purgeif_wrapper),
   1919 };
   1920 
   1921 static const struct protosw COMPATNAME(route_protosw)[] = {
   1922 	{
   1923 		.pr_type = SOCK_RAW,
   1924 		.pr_domain = &COMPATNAME(routedomain),
   1925 		.pr_flags = PR_ATOMIC|PR_ADDR,
   1926 		.pr_input = raw_input,
   1927 		.pr_ctlinput = raw_ctlinput,
   1928 		.pr_usrreqs = &route_usrreqs,
   1929 		.pr_init = raw_init,
   1930 	},
   1931 };
   1932 
   1933 struct domain COMPATNAME(routedomain) = {
   1934 	.dom_family = PF_XROUTE,
   1935 	.dom_name = DOMAINNAME,
   1936 	.dom_init = COMPATNAME(route_init),
   1937 	.dom_protosw = COMPATNAME(route_protosw),
   1938 	.dom_protoswNPROTOSW =
   1939 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
   1940 };
   1941 
   1942 static void
   1943 sysctl_net_route_setup(struct sysctllog **clog)
   1944 {
   1945 	const struct sysctlnode *rnode = NULL;
   1946 
   1947 	sysctl_createv(clog, 0, NULL, &rnode,
   1948 		       CTLFLAG_PERMANENT,
   1949 		       CTLTYPE_NODE, DOMAINNAME,
   1950 		       SYSCTL_DESCR("PF_ROUTE information"),
   1951 		       NULL, 0, NULL, 0,
   1952 		       CTL_NET, PF_XROUTE, CTL_EOL);
   1953 
   1954 	sysctl_createv(clog, 0, NULL, NULL,
   1955 		       CTLFLAG_PERMANENT,
   1956 		       CTLTYPE_NODE, "rtable",
   1957 		       SYSCTL_DESCR("Routing table information"),
   1958 		       sysctl_rtable, 0, NULL, 0,
   1959 		       CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
   1960 
   1961 	sysctl_createv(clog, 0, &rnode, NULL,
   1962 		       CTLFLAG_PERMANENT,
   1963 		       CTLTYPE_STRUCT, "stats",
   1964 		       SYSCTL_DESCR("Routing statistics"),
   1965 		       NULL, 0, &rtstat, sizeof(rtstat),
   1966 		       CTL_CREATE, CTL_EOL);
   1967 }
   1968