Home | History | Annotate | Line # | Download | only in net
rtsock.c revision 1.154
      1 /*	$NetBSD: rtsock.c,v 1.154 2014/07/09 04:54:03 rtr 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 1.154 2014/07/09 04:54:03 rtr Exp $");
     65 
     66 #ifdef _KERNEL_OPT
     67 #include "opt_inet.h"
     68 #include "opt_mpls.h"
     69 #include "opt_compat_netbsd.h"
     70 #endif
     71 
     72 #include <sys/param.h>
     73 #include <sys/systm.h>
     74 #include <sys/proc.h>
     75 #include <sys/socket.h>
     76 #include <sys/socketvar.h>
     77 #include <sys/domain.h>
     78 #include <sys/protosw.h>
     79 #include <sys/sysctl.h>
     80 #include <sys/kauth.h>
     81 #include <sys/kmem.h>
     82 #include <sys/intr.h>
     83 #ifdef RTSOCK_DEBUG
     84 #include <netinet/in.h>
     85 #endif /* RTSOCK_DEBUG */
     86 
     87 #include <net/if.h>
     88 #include <net/route.h>
     89 #include <net/raw_cb.h>
     90 
     91 #include <netmpls/mpls.h>
     92 
     93 #if defined(COMPAT_14) || defined(COMPAT_50)
     94 #include <compat/net/if.h>
     95 #include <compat/net/route.h>
     96 #endif
     97 #ifdef COMPAT_RTSOCK
     98 #define	RTM_XVERSION	RTM_OVERSION
     99 #define	RT_XADVANCE(a,b) RT_OADVANCE(a,b)
    100 #define	RT_XROUNDUP(n)	RT_OROUNDUP(n)
    101 #define	PF_XROUTE	PF_OROUTE
    102 #define	rt_xmsghdr	rt_msghdr50
    103 #define	if_xmsghdr	if_msghdr	/* if_msghdr50 is for RTM_OIFINFO */
    104 #define	ifa_xmsghdr	ifa_msghdr50
    105 #define	if_xannouncemsghdr	if_announcemsghdr50
    106 #define	COMPATNAME(x)	compat_50_ ## x
    107 #define	DOMAINNAME	"oroute"
    108 CTASSERT(sizeof(struct ifa_xmsghdr) == 20);
    109 DOMAIN_DEFINE(compat_50_routedomain); /* forward declare and add to link set */
    110 #else
    111 #define	RTM_XVERSION	RTM_VERSION
    112 #define	RT_XADVANCE(a,b) RT_ADVANCE(a,b)
    113 #define	RT_XROUNDUP(n)	RT_ROUNDUP(n)
    114 #define	PF_XROUTE	PF_ROUTE
    115 #define	rt_xmsghdr	rt_msghdr
    116 #define	if_xmsghdr	if_msghdr
    117 #define	ifa_xmsghdr	ifa_msghdr
    118 #define	if_xannouncemsghdr	if_announcemsghdr
    119 #define	COMPATNAME(x)	x
    120 #define	DOMAINNAME	"route"
    121 CTASSERT(sizeof(struct ifa_xmsghdr) == 24);
    122 #ifdef COMPAT_50
    123 #define	COMPATCALL(name, args)	compat_50_ ## name args
    124 #endif
    125 DOMAIN_DEFINE(routedomain); /* forward declare and add to link set */
    126 #undef COMPAT_50
    127 #undef COMPAT_14
    128 #endif
    129 
    130 #ifndef COMPATCALL
    131 #define	COMPATCALL(name, args)	do { } while (/*CONSTCOND*/ 0)
    132 #endif
    133 
    134 struct route_info COMPATNAME(route_info) = {
    135 	.ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, },
    136 	.ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, },
    137 	.ri_maxqlen = IFQ_MAXLEN,
    138 };
    139 
    140 #define	PRESERVED_RTF	(RTF_UP | RTF_GATEWAY | RTF_HOST | RTF_DONE | RTF_MASK)
    141 
    142 static void COMPATNAME(route_init)(void);
    143 static int COMPATNAME(route_output)(struct mbuf *, ...);
    144 
    145 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
    146 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
    147 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
    148     struct rt_addrinfo *);
    149 static void rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *);
    150 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *);
    151 static void sysctl_net_route_setup(struct sysctllog **);
    152 static int sysctl_dumpentry(struct rtentry *, void *);
    153 static int sysctl_iflist(int, struct rt_walkarg *, int);
    154 static int sysctl_rtable(SYSCTLFN_PROTO);
    155 static void rt_adjustcount(int, int);
    156 
    157 static void
    158 rt_adjustcount(int af, int cnt)
    159 {
    160 	struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
    161 
    162 	cb->any_count += cnt;
    163 
    164 	switch (af) {
    165 	case AF_INET:
    166 		cb->ip_count += cnt;
    167 		return;
    168 #ifdef INET6
    169 	case AF_INET6:
    170 		cb->ip6_count += cnt;
    171 		return;
    172 #endif
    173 	case AF_MPLS:
    174 		cb->mpls_count += cnt;
    175 		return;
    176 	}
    177 }
    178 
    179 static int
    180 COMPATNAME(route_attach)(struct socket *so, int proto)
    181 {
    182 	struct rawcb *rp;
    183 	int s, error;
    184 
    185 	KASSERT(sotorawcb(so) == NULL);
    186 	rp = kmem_zalloc(sizeof(*rp), KM_SLEEP);
    187 	rp->rcb_len = sizeof(*rp);
    188 	so->so_pcb = rp;
    189 
    190 	s = splsoftnet();
    191 	if ((error = raw_attach(so, proto)) == 0) {
    192 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
    193 		rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
    194 		rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
    195 	}
    196 	splx(s);
    197 
    198 	if (error) {
    199 		kmem_free(rp, sizeof(*rp));
    200 		so->so_pcb = NULL;
    201 		return error;
    202 	}
    203 
    204 	soisconnected(so);
    205 	so->so_options |= SO_USELOOPBACK;
    206 	KASSERT(solocked(so));
    207 
    208 	return error;
    209 }
    210 
    211 static void
    212 COMPATNAME(route_detach)(struct socket *so)
    213 {
    214 	struct rawcb *rp = sotorawcb(so);
    215 	int s;
    216 
    217 	KASSERT(rp != NULL);
    218 	KASSERT(solocked(so));
    219 
    220 	s = splsoftnet();
    221 	rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
    222 	raw_detach(so);
    223 	splx(s);
    224 }
    225 
    226 static int
    227 COMPATNAME(route_ioctl)(struct socket *so, u_long cmd, void *nam,
    228     struct ifnet * ifp)
    229 {
    230 	return EOPNOTSUPP;
    231 }
    232 
    233 static int
    234 COMPATNAME(route_stat)(struct socket *so, struct stat *ub)
    235 {
    236 	KASSERT(solocked(so));
    237 
    238 	return 0;
    239 }
    240 
    241 static int
    242 COMPATNAME(route_peeraddr)(struct socket *so, struct mbuf *nam)
    243 {
    244 	struct rawcb *rp = sotorawcb(so);
    245 
    246 	KASSERT(solocked(so));
    247 	KASSERT(rp != NULL);
    248 	KASSERT(nam != NULL);
    249 
    250 	if (rp->rcb_faddr == NULL)
    251 		return ENOTCONN;
    252 
    253 	raw_setpeeraddr(rp, nam);
    254 	return 0;
    255 }
    256 
    257 static int
    258 COMPATNAME(route_sockaddr)(struct socket *so, struct mbuf *nam)
    259 {
    260 	struct rawcb *rp = sotorawcb(so);
    261 
    262 	KASSERT(solocked(so));
    263 	KASSERT(rp != NULL);
    264 	KASSERT(nam != NULL);
    265 
    266 	if (rp->rcb_faddr == NULL)
    267 		return ENOTCONN;
    268 
    269 	raw_setsockaddr(rp, nam);
    270 	return 0;
    271 }
    272 
    273 static int
    274 COMPATNAME(route_usrreq)(struct socket *so, int req, struct mbuf *m,
    275     struct mbuf *nam, struct mbuf *control, struct lwp *l)
    276 {
    277 	int s, error = 0;
    278 
    279 	KASSERT(req != PRU_ATTACH);
    280 	KASSERT(req != PRU_DETACH);
    281 	KASSERT(req != PRU_CONTROL);
    282 	KASSERT(req != PRU_SENSE);
    283 	KASSERT(req != PRU_PEERADDR);
    284 	KASSERT(req != PRU_SOCKADDR);
    285 
    286 	s = splsoftnet();
    287 	error = raw_usrreq(so, req, m, nam, control, l);
    288 	splx(s);
    289 
    290 	return error;
    291 }
    292 
    293 /*ARGSUSED*/
    294 int
    295 COMPATNAME(route_output)(struct mbuf *m, ...)
    296 {
    297 	struct sockproto proto = { .sp_family = PF_XROUTE, };
    298 	struct rt_xmsghdr *rtm = NULL;
    299 	struct rt_xmsghdr *old_rtm = NULL;
    300 	struct rtentry *rt = NULL;
    301 	struct rtentry *saved_nrt = NULL;
    302 	struct rt_addrinfo info;
    303 	int len, error = 0;
    304 	struct ifnet *ifp = NULL;
    305 	struct ifaddr *ifa = NULL;
    306 	struct socket *so;
    307 	va_list ap;
    308 	sa_family_t family;
    309 
    310 	va_start(ap, m);
    311 	so = va_arg(ap, struct socket *);
    312 	va_end(ap);
    313 
    314 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
    315 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
    316 	   (m = m_pullup(m, sizeof(int32_t))) == NULL))
    317 		return ENOBUFS;
    318 	if ((m->m_flags & M_PKTHDR) == 0)
    319 		panic("%s", __func__);
    320 	len = m->m_pkthdr.len;
    321 	if (len < sizeof(*rtm) ||
    322 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
    323 		info.rti_info[RTAX_DST] = NULL;
    324 		senderr(EINVAL);
    325 	}
    326 	R_Malloc(rtm, struct rt_xmsghdr *, len);
    327 	if (rtm == NULL) {
    328 		info.rti_info[RTAX_DST] = NULL;
    329 		senderr(ENOBUFS);
    330 	}
    331 	m_copydata(m, 0, len, rtm);
    332 	if (rtm->rtm_version != RTM_XVERSION) {
    333 		info.rti_info[RTAX_DST] = NULL;
    334 		senderr(EPROTONOSUPPORT);
    335 	}
    336 	rtm->rtm_pid = curproc->p_pid;
    337 	memset(&info, 0, sizeof(info));
    338 	info.rti_addrs = rtm->rtm_addrs;
    339 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
    340 	    &info)) {
    341 		senderr(EINVAL);
    342 	}
    343 	info.rti_flags = rtm->rtm_flags;
    344 #ifdef RTSOCK_DEBUG
    345 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
    346 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
    347 		    inet_ntoa(((const struct sockaddr_in *)
    348 		    info.rti_info[RTAX_DST])->sin_addr));
    349 	}
    350 #endif /* RTSOCK_DEBUG */
    351 	if (info.rti_info[RTAX_DST] == NULL ||
    352 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
    353 		senderr(EINVAL);
    354 	}
    355 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
    356 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
    357 		senderr(EINVAL);
    358 	}
    359 
    360 	/*
    361 	 * Verify that the caller has the appropriate privilege; RTM_GET
    362 	 * is the only operation the non-superuser is allowed.
    363 	 */
    364 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
    365 	    0, rtm, NULL, NULL) != 0)
    366 		senderr(EACCES);
    367 
    368 	switch (rtm->rtm_type) {
    369 
    370 	case RTM_ADD:
    371 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
    372 			senderr(EINVAL);
    373 		}
    374 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
    375 		if (error == 0 && saved_nrt) {
    376 			rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
    377 			saved_nrt->rt_refcnt--;
    378 		}
    379 		break;
    380 
    381 	case RTM_DELETE:
    382 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
    383 		if (error == 0) {
    384 			(rt = saved_nrt)->rt_refcnt++;
    385 			goto report;
    386 		}
    387 		break;
    388 
    389 	case RTM_GET:
    390 	case RTM_CHANGE:
    391 	case RTM_LOCK:
    392                 /* XXX This will mask info.rti_info[RTAX_DST] with
    393 		 * info.rti_info[RTAX_NETMASK] before
    394                  * searching.  It did not used to do that.  --dyoung
    395 		 */
    396 		error = rtrequest1(RTM_GET, &info, &rt);
    397 		if (error != 0)
    398 			senderr(error);
    399 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
    400 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
    401 			    info.rti_info[RTAX_DST]->sa_len) != 0)
    402 				senderr(ESRCH);
    403 			if (info.rti_info[RTAX_NETMASK] == NULL &&
    404 			    rt_mask(rt) != NULL)
    405 				senderr(ETOOMANYREFS);
    406 		}
    407 
    408 		switch (rtm->rtm_type) {
    409 		case RTM_GET:
    410 		report:
    411 			info.rti_info[RTAX_DST] = rt_getkey(rt);
    412 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    413 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    414 			info.rti_info[RTAX_TAG] = rt_gettag(rt);
    415 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
    416 				;
    417 			else if ((ifp = rt->rt_ifp) != NULL) {
    418 				const struct ifaddr *rtifa;
    419 				info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
    420                                 /* rtifa used to be simply rt->rt_ifa.
    421                                  * If rt->rt_ifa != NULL, then
    422                                  * rt_get_ifa() != NULL.  So this
    423                                  * ought to still be safe. --dyoung
    424 				 */
    425 				rtifa = rt_get_ifa(rt);
    426 				info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
    427 #ifdef RTSOCK_DEBUG
    428 				if (info.rti_info[RTAX_IFA]->sa_family ==
    429 				    AF_INET) {
    430 					printf("%s: copying out RTAX_IFA %s ",
    431 					    __func__, inet_ntoa(
    432 					    ((const struct sockaddr_in *)
    433 					    info.rti_info[RTAX_IFA])->sin_addr)
    434 					    );
    435 					printf("for info.rti_info[RTAX_DST] %s "
    436 					    "ifa_getifa %p ifa_seqno %p\n",
    437 					    inet_ntoa(
    438 					    ((const struct sockaddr_in *)
    439 					    info.rti_info[RTAX_DST])->sin_addr),
    440 					    (void *)rtifa->ifa_getifa,
    441 					    rtifa->ifa_seqno);
    442 				}
    443 #endif /* RTSOCK_DEBUG */
    444 				if (ifp->if_flags & IFF_POINTOPOINT) {
    445 					info.rti_info[RTAX_BRD] =
    446 					    rtifa->ifa_dstaddr;
    447 				} else
    448 					info.rti_info[RTAX_BRD] = NULL;
    449 				rtm->rtm_index = ifp->if_index;
    450 			} else {
    451 				info.rti_info[RTAX_IFP] = NULL;
    452 				info.rti_info[RTAX_IFA] = NULL;
    453 			}
    454 			(void)rt_msg2(rtm->rtm_type, &info, NULL, NULL, &len);
    455 			if (len > rtm->rtm_msglen) {
    456 				old_rtm = rtm;
    457 				R_Malloc(rtm, struct rt_xmsghdr *, len);
    458 				if (rtm == NULL)
    459 					senderr(ENOBUFS);
    460 				(void)memcpy(rtm, old_rtm, old_rtm->rtm_msglen);
    461 			}
    462 			(void)rt_msg2(rtm->rtm_type, &info, rtm, NULL, 0);
    463 			rtm->rtm_flags = rt->rt_flags;
    464 			rtm_setmetrics(rt, rtm);
    465 			rtm->rtm_addrs = info.rti_addrs;
    466 			break;
    467 
    468 		case RTM_CHANGE:
    469 			/*
    470 			 * new gateway could require new ifaddr, ifp;
    471 			 * flags may also be different; ifp may be specified
    472 			 * by ll sockaddr when protocol address is ambiguous
    473 			 */
    474 			if ((error = rt_getifa(&info)) != 0)
    475 				senderr(error);
    476 			if (info.rti_info[RTAX_GATEWAY] &&
    477 			    rt_setgate(rt, info.rti_info[RTAX_GATEWAY]))
    478 				senderr(EDQUOT);
    479 			if (info.rti_info[RTAX_TAG])
    480 				rt_settag(rt, info.rti_info[RTAX_TAG]);
    481 			/* new gateway could require new ifaddr, ifp;
    482 			   flags may also be different; ifp may be specified
    483 			   by ll sockaddr when protocol address is ambiguous */
    484 			if (info.rti_info[RTAX_IFP] &&
    485 			    (ifa = ifa_ifwithnet(info.rti_info[RTAX_IFP])) &&
    486 			    (ifp = ifa->ifa_ifp) && (info.rti_info[RTAX_IFA] ||
    487 			    info.rti_info[RTAX_GATEWAY])) {
    488 				if (info.rti_info[RTAX_IFA] == NULL ||
    489 				    (ifa = ifa_ifwithaddr(
    490 				    info.rti_info[RTAX_IFA])) == NULL)
    491 					ifa = ifaof_ifpforaddr(
    492 					    info.rti_info[RTAX_IFA] ?
    493 					    info.rti_info[RTAX_IFA] :
    494 					    info.rti_info[RTAX_GATEWAY], ifp);
    495 			} else if ((info.rti_info[RTAX_IFA] &&
    496 			    (ifa = ifa_ifwithaddr(info.rti_info[RTAX_IFA]))) ||
    497 			    (info.rti_info[RTAX_GATEWAY] &&
    498 			    (ifa = ifa_ifwithroute(rt->rt_flags,
    499 			    rt_getkey(rt), info.rti_info[RTAX_GATEWAY])))) {
    500 				ifp = ifa->ifa_ifp;
    501 			}
    502 			if (ifa) {
    503 				struct ifaddr *oifa = rt->rt_ifa;
    504 				if (oifa != ifa) {
    505 					if (oifa && oifa->ifa_rtrequest) {
    506 						oifa->ifa_rtrequest(RTM_DELETE,
    507 						    rt, &info);
    508 					}
    509 					rt_replace_ifa(rt, ifa);
    510 					rt->rt_ifp = ifp;
    511 				}
    512 			}
    513 			if (ifp && rt->rt_ifp != ifp)
    514 				rt->rt_ifp = ifp;
    515 			rt_setmetrics(rtm->rtm_inits, rtm, rt);
    516 			if (rt->rt_flags != info.rti_flags)
    517 				rt->rt_flags = (info.rti_flags & ~PRESERVED_RTF)
    518 				    | (rt->rt_flags & PRESERVED_RTF);
    519 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
    520 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
    521 			/*FALLTHROUGH*/
    522 		case RTM_LOCK:
    523 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
    524 			rt->rt_rmx.rmx_locks |=
    525 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
    526 			break;
    527 		}
    528 		break;
    529 
    530 	default:
    531 		senderr(EOPNOTSUPP);
    532 	}
    533 
    534 flush:
    535 	if (rtm) {
    536 		if (error)
    537 			rtm->rtm_errno = error;
    538 		else
    539 			rtm->rtm_flags |= RTF_DONE;
    540 	}
    541 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
    542 	    0;
    543 	/* We cannot free old_rtm until we have stopped using the
    544 	 * pointers in info, some of which may point to sockaddrs
    545 	 * in old_rtm.
    546 	 */
    547 	if (old_rtm != NULL)
    548 		Free(old_rtm);
    549 	if (rt)
    550 		rtfree(rt);
    551     {
    552 	struct rawcb *rp = NULL;
    553 	/*
    554 	 * Check to see if we don't want our own messages.
    555 	 */
    556 	if ((so->so_options & SO_USELOOPBACK) == 0) {
    557 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
    558 			if (rtm)
    559 				Free(rtm);
    560 			m_freem(m);
    561 			return error;
    562 		}
    563 		/* There is another listener, so construct message */
    564 		rp = sotorawcb(so);
    565 	}
    566 	if (rtm) {
    567 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
    568 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
    569 			m_freem(m);
    570 			m = NULL;
    571 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
    572 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
    573 		Free(rtm);
    574 	}
    575 	if (rp)
    576 		rp->rcb_proto.sp_family = 0; /* Avoid us */
    577 	if (family)
    578 		proto.sp_protocol = family;
    579 	if (m)
    580 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
    581 		    &COMPATNAME(route_info).ri_dst);
    582 	if (rp)
    583 		rp->rcb_proto.sp_family = PF_XROUTE;
    584     }
    585 	return error;
    586 }
    587 
    588 static void
    589 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
    590 {
    591 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
    592 	metric(RTV_RPIPE, rmx_recvpipe);
    593 	metric(RTV_SPIPE, rmx_sendpipe);
    594 	metric(RTV_SSTHRESH, rmx_ssthresh);
    595 	metric(RTV_RTT, rmx_rtt);
    596 	metric(RTV_RTTVAR, rmx_rttvar);
    597 	metric(RTV_HOPCOUNT, rmx_hopcount);
    598 	metric(RTV_MTU, rmx_mtu);
    599 	metric(RTV_EXPIRE, rmx_expire);
    600 #undef metric
    601 }
    602 
    603 static void
    604 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
    605 {
    606 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
    607 	metric(rmx_recvpipe);
    608 	metric(rmx_sendpipe);
    609 	metric(rmx_ssthresh);
    610 	metric(rmx_rtt);
    611 	metric(rmx_rttvar);
    612 	metric(rmx_hopcount);
    613 	metric(rmx_mtu);
    614 	metric(rmx_expire);
    615 #undef metric
    616 }
    617 
    618 static int
    619 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
    620     struct rt_addrinfo *rtinfo)
    621 {
    622 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
    623 	int i;
    624 
    625 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
    626 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
    627 			continue;
    628 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
    629 		RT_XADVANCE(cp, sa);
    630 	}
    631 
    632 	/*
    633 	 * Check for extra addresses specified, except RTM_GET asking
    634 	 * for interface info.
    635 	 */
    636 	if (rtmtype == RTM_GET) {
    637 		if (((rtinfo->rti_addrs &
    638 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
    639 			return 1;
    640 	} else if ((rtinfo->rti_addrs & (~0 << i)) != 0)
    641 		return 1;
    642 	/* Check for bad data length.  */
    643 	if (cp != cplim) {
    644 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
    645 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
    646 			/*
    647 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
    648 			 * We accept this for now for the sake of old
    649 			 * binaries or third party softwares.
    650 			 */
    651 			;
    652 		else
    653 			return 1;
    654 	}
    655 	return 0;
    656 }
    657 
    658 static int
    659 rt_getlen(int type)
    660 {
    661 #ifndef COMPAT_RTSOCK
    662 	CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
    663 	CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
    664 	CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
    665 	CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
    666 #endif
    667 
    668 	switch (type) {
    669 	case RTM_DELADDR:
    670 	case RTM_NEWADDR:
    671 	case RTM_CHGADDR:
    672 		return sizeof(struct ifa_xmsghdr);
    673 
    674 	case RTM_OOIFINFO:
    675 #ifdef COMPAT_14
    676 		return sizeof(struct if_msghdr14);
    677 #else
    678 #ifdef DIAGNOSTIC
    679 		printf("RTM_OOIFINFO\n");
    680 #endif
    681 		return -1;
    682 #endif
    683 	case RTM_OIFINFO:
    684 #ifdef COMPAT_50
    685 		return sizeof(struct if_msghdr50);
    686 #else
    687 #ifdef DIAGNOSTIC
    688 		printf("RTM_OIFINFO\n");
    689 #endif
    690 		return -1;
    691 #endif
    692 
    693 	case RTM_IFINFO:
    694 		return sizeof(struct if_xmsghdr);
    695 
    696 	case RTM_IFANNOUNCE:
    697 	case RTM_IEEE80211:
    698 		return sizeof(struct if_xannouncemsghdr);
    699 
    700 	default:
    701 		return sizeof(struct rt_xmsghdr);
    702 	}
    703 }
    704 
    705 
    706 struct mbuf *
    707 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
    708 {
    709 	struct rt_xmsghdr *rtm;
    710 	struct mbuf *m;
    711 	int i;
    712 	const struct sockaddr *sa;
    713 	int len, dlen;
    714 
    715 	m = m_gethdr(M_DONTWAIT, MT_DATA);
    716 	if (m == NULL)
    717 		return m;
    718 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
    719 
    720 	if ((len = rt_getlen(type)) == -1)
    721 		goto out;
    722 	if (len > MHLEN + MLEN)
    723 		panic("%s: message too long", __func__);
    724 	else if (len > MHLEN) {
    725 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
    726 		if (m->m_next == NULL)
    727 			goto out;
    728 		MCLAIM(m->m_next, m->m_owner);
    729 		m->m_pkthdr.len = len;
    730 		m->m_len = MHLEN;
    731 		m->m_next->m_len = len - MHLEN;
    732 	} else {
    733 		m->m_pkthdr.len = m->m_len = len;
    734 	}
    735 	m->m_pkthdr.rcvif = NULL;
    736 	m_copyback(m, 0, datalen, data);
    737 	if (len > datalen)
    738 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
    739 	rtm = mtod(m, struct rt_xmsghdr *);
    740 	for (i = 0; i < RTAX_MAX; i++) {
    741 		if ((sa = rtinfo->rti_info[i]) == NULL)
    742 			continue;
    743 		rtinfo->rti_addrs |= (1 << i);
    744 		dlen = RT_XROUNDUP(sa->sa_len);
    745 		m_copyback(m, len, sa->sa_len, sa);
    746 		if (dlen != sa->sa_len) {
    747 			/*
    748 			 * Up to 6 + 1 nul's since roundup is to
    749 			 * sizeof(uint64_t) (8 bytes)
    750 			 */
    751 			m_copyback(m, len + sa->sa_len,
    752 			    dlen - sa->sa_len, "\0\0\0\0\0\0");
    753 		}
    754 		len += dlen;
    755 	}
    756 	if (m->m_pkthdr.len != len)
    757 		goto out;
    758 	rtm->rtm_msglen = len;
    759 	rtm->rtm_version = RTM_XVERSION;
    760 	rtm->rtm_type = type;
    761 	return m;
    762 out:
    763 	m_freem(m);
    764 	return NULL;
    765 }
    766 
    767 /*
    768  * rt_msg2
    769  *
    770  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
    771  *		returns the length of the message in 'lenp'.
    772  *
    773  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
    774  *	the message
    775  * otherwise walkarg's w_needed is updated and if the user buffer is
    776  *	specified and w_needed indicates space exists the information is copied
    777  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
    778  *	if the allocation fails ENOBUFS is returned.
    779  */
    780 static int
    781 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
    782 	int *lenp)
    783 {
    784 	int i;
    785 	int len, dlen, second_time = 0;
    786 	char *cp0, *cp = cpv;
    787 
    788 	rtinfo->rti_addrs = 0;
    789 again:
    790 	if ((len = rt_getlen(type)) == -1)
    791 		return EINVAL;
    792 
    793 	if ((cp0 = cp) != NULL)
    794 		cp += len;
    795 	for (i = 0; i < RTAX_MAX; i++) {
    796 		const struct sockaddr *sa;
    797 
    798 		if ((sa = rtinfo->rti_info[i]) == NULL)
    799 			continue;
    800 		rtinfo->rti_addrs |= (1 << i);
    801 		dlen = RT_XROUNDUP(sa->sa_len);
    802 		if (cp) {
    803 			int diff = dlen - sa->sa_len;
    804 			(void)memcpy(cp, sa, (size_t)sa->sa_len);
    805 			cp += sa->sa_len;
    806 			if (diff > 0) {
    807 				(void)memset(cp, 0, (size_t)diff);
    808 				cp += diff;
    809 			}
    810 		}
    811 		len += dlen;
    812 	}
    813 	if (cp == NULL && w != NULL && !second_time) {
    814 		struct rt_walkarg *rw = w;
    815 
    816 		rw->w_needed += len;
    817 		if (rw->w_needed <= 0 && rw->w_where) {
    818 			if (rw->w_tmemsize < len) {
    819 				if (rw->w_tmem)
    820 					free(rw->w_tmem, M_RTABLE);
    821 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
    822 				if (rw->w_tmem)
    823 					rw->w_tmemsize = len;
    824 				else
    825 					rw->w_tmemsize = 0;
    826 			}
    827 			if (rw->w_tmem) {
    828 				cp = rw->w_tmem;
    829 				second_time = 1;
    830 				goto again;
    831 			} else {
    832 				rw->w_tmemneeded = len;
    833 				return ENOBUFS;
    834 			}
    835 		}
    836 	}
    837 	if (cp) {
    838 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
    839 
    840 		rtm->rtm_version = RTM_XVERSION;
    841 		rtm->rtm_type = type;
    842 		rtm->rtm_msglen = len;
    843 	}
    844 	if (lenp)
    845 		*lenp = len;
    846 	return 0;
    847 }
    848 
    849 /*
    850  * This routine is called to generate a message from the routing
    851  * socket indicating that a redirect has occurred, a routing lookup
    852  * has failed, or that a protocol has detected timeouts to a particular
    853  * destination.
    854  */
    855 void
    856 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
    857     int error)
    858 {
    859 	struct rt_xmsghdr rtm;
    860 	struct mbuf *m;
    861 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
    862 	struct rt_addrinfo info = *rtinfo;
    863 
    864 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
    865 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    866 		return;
    867 	memset(&rtm, 0, sizeof(rtm));
    868 	rtm.rtm_flags = RTF_DONE | flags;
    869 	rtm.rtm_errno = error;
    870 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
    871 	if (m == NULL)
    872 		return;
    873 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
    874 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
    875 }
    876 
    877 /*
    878  * This routine is called to generate a message from the routing
    879  * socket indicating that the status of a network interface has changed.
    880  */
    881 void
    882 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
    883 {
    884 	struct if_xmsghdr ifm;
    885 	struct mbuf *m;
    886 	struct rt_addrinfo info;
    887 
    888 	COMPATCALL(rt_ifmsg, (ifp));
    889 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    890 		return;
    891 	(void)memset(&info, 0, sizeof(info));
    892 	(void)memset(&ifm, 0, sizeof(ifm));
    893 	ifm.ifm_index = ifp->if_index;
    894 	ifm.ifm_flags = ifp->if_flags;
    895 	ifm.ifm_data = ifp->if_data;
    896 	ifm.ifm_addrs = 0;
    897 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
    898 	if (m == NULL)
    899 		return;
    900 	COMPATNAME(route_enqueue)(m, 0);
    901 #ifdef COMPAT_14
    902 	compat_14_rt_oifmsg(ifp);
    903 #endif
    904 #ifdef COMPAT_50
    905 	compat_50_rt_oifmsg(ifp);
    906 #endif
    907 }
    908 
    909 
    910 /*
    911  * This is called to generate messages from the routing socket
    912  * indicating a network interface has had addresses associated with it.
    913  * if we ever reverse the logic and replace messages TO the routing
    914  * socket indicate a request to configure interfaces, then it will
    915  * be unnecessary as the routing socket will automatically generate
    916  * copies of it.
    917  */
    918 void
    919 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
    920     struct rtentry *rt)
    921 {
    922 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
    923 	struct rt_addrinfo info;
    924 	const struct sockaddr *sa;
    925 	int pass;
    926 	struct mbuf *m;
    927 	struct ifnet *ifp;
    928 	struct rt_xmsghdr rtm;
    929 	struct ifa_xmsghdr ifam;
    930 	int ncmd;
    931 
    932 	KASSERT(ifa != NULL);
    933 	ifp = ifa->ifa_ifp;
    934 	COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
    935 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    936 		return;
    937 	for (pass = 1; pass < 3; pass++) {
    938 		memset(&info, 0, sizeof(info));
    939 		switch (cmdpass(cmd, pass)) {
    940 		case cmdpass(RTM_ADD, 1):
    941 		case cmdpass(RTM_CHANGE, 1):
    942 		case cmdpass(RTM_DELETE, 2):
    943 		case cmdpass(RTM_NEWADDR, 1):
    944 		case cmdpass(RTM_DELADDR, 1):
    945 		case cmdpass(RTM_CHGADDR, 1):
    946 			switch (cmd) {
    947 			case RTM_ADD:
    948 				ncmd = RTM_NEWADDR;
    949 				break;
    950 			case RTM_DELETE:
    951 				ncmd = RTM_DELADDR;
    952 				break;
    953 			case RTM_CHANGE:
    954 				ncmd = RTM_CHGADDR;
    955 				break;
    956 			default:
    957 				ncmd = cmd;
    958 			}
    959 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
    960 			KASSERT(ifp->if_dl != NULL);
    961 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
    962 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
    963 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
    964 			memset(&ifam, 0, sizeof(ifam));
    965 			ifam.ifam_index = ifp->if_index;
    966 			ifam.ifam_metric = ifa->ifa_metric;
    967 			ifam.ifam_flags = ifa->ifa_flags;
    968 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
    969 			if (m == NULL)
    970 				continue;
    971 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
    972 			    info.rti_addrs;
    973 			break;
    974 		case cmdpass(RTM_ADD, 2):
    975 		case cmdpass(RTM_CHANGE, 2):
    976 		case cmdpass(RTM_DELETE, 1):
    977 			if (rt == NULL)
    978 				continue;
    979 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    980 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
    981 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    982 			memset(&rtm, 0, sizeof(rtm));
    983 			rtm.rtm_index = ifp->if_index;
    984 			rtm.rtm_flags |= rt->rt_flags;
    985 			rtm.rtm_errno = error;
    986 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
    987 			if (m == NULL)
    988 				continue;
    989 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
    990 			break;
    991 		default:
    992 			continue;
    993 		}
    994 #ifdef DIAGNOSTIC
    995 		if (m == NULL)
    996 			panic("%s: called with wrong command", __func__);
    997 #endif
    998 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
    999 	}
   1000 #undef cmdpass
   1001 }
   1002 
   1003 static struct mbuf *
   1004 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
   1005     struct rt_addrinfo *info)
   1006 {
   1007 	struct if_xannouncemsghdr ifan;
   1008 
   1009 	memset(info, 0, sizeof(*info));
   1010 	memset(&ifan, 0, sizeof(ifan));
   1011 	ifan.ifan_index = ifp->if_index;
   1012 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
   1013 	ifan.ifan_what = what;
   1014 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
   1015 }
   1016 
   1017 /*
   1018  * This is called to generate routing socket messages indicating
   1019  * network interface arrival and departure.
   1020  */
   1021 void
   1022 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
   1023 {
   1024 	struct mbuf *m;
   1025 	struct rt_addrinfo info;
   1026 
   1027 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
   1028 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
   1029 		return;
   1030 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
   1031 	if (m == NULL)
   1032 		return;
   1033 	COMPATNAME(route_enqueue)(m, 0);
   1034 }
   1035 
   1036 /*
   1037  * This is called to generate routing socket messages indicating
   1038  * IEEE80211 wireless events.
   1039  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
   1040  */
   1041 void
   1042 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
   1043 	size_t data_len)
   1044 {
   1045 	struct mbuf *m;
   1046 	struct rt_addrinfo info;
   1047 
   1048 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
   1049 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
   1050 		return;
   1051 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
   1052 	if (m == NULL)
   1053 		return;
   1054 	/*
   1055 	 * Append the ieee80211 data.  Try to stick it in the
   1056 	 * mbuf containing the ifannounce msg; otherwise allocate
   1057 	 * a new mbuf and append.
   1058 	 *
   1059 	 * NB: we assume m is a single mbuf.
   1060 	 */
   1061 	if (data_len > M_TRAILINGSPACE(m)) {
   1062 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
   1063 		if (n == NULL) {
   1064 			m_freem(m);
   1065 			return;
   1066 		}
   1067 		(void)memcpy(mtod(n, void *), data, data_len);
   1068 		n->m_len = data_len;
   1069 		m->m_next = n;
   1070 	} else if (data_len > 0) {
   1071 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
   1072 		m->m_len += data_len;
   1073 	}
   1074 	if (m->m_flags & M_PKTHDR)
   1075 		m->m_pkthdr.len += data_len;
   1076 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
   1077 	COMPATNAME(route_enqueue)(m, 0);
   1078 }
   1079 
   1080 /*
   1081  * This is used in dumping the kernel table via sysctl().
   1082  */
   1083 static int
   1084 sysctl_dumpentry(struct rtentry *rt, void *v)
   1085 {
   1086 	struct rt_walkarg *w = v;
   1087 	int error = 0, size;
   1088 	struct rt_addrinfo info;
   1089 
   1090 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
   1091 		return 0;
   1092 	memset(&info, 0, sizeof(info));
   1093 	info.rti_info[RTAX_DST] = rt_getkey(rt);
   1094 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
   1095 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
   1096 	info.rti_info[RTAX_TAG] = rt_gettag(rt);
   1097 	if (rt->rt_ifp) {
   1098 		const struct ifaddr *rtifa;
   1099 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
   1100 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
   1101 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
   1102 		 * --dyoung
   1103 		 */
   1104 		rtifa = rt_get_ifa(rt);
   1105 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
   1106 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
   1107 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
   1108 	}
   1109 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
   1110 		return error;
   1111 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1112 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
   1113 
   1114 		rtm->rtm_flags = rt->rt_flags;
   1115 		rtm->rtm_use = rt->rt_use;
   1116 		rtm_setmetrics(rt, rtm);
   1117 		KASSERT(rt->rt_ifp != NULL);
   1118 		rtm->rtm_index = rt->rt_ifp->if_index;
   1119 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
   1120 		rtm->rtm_addrs = info.rti_addrs;
   1121 		if ((error = copyout(rtm, w->w_where, size)) != 0)
   1122 			w->w_where = NULL;
   1123 		else
   1124 			w->w_where = (char *)w->w_where + size;
   1125 	}
   1126 	return error;
   1127 }
   1128 
   1129 static int
   1130 sysctl_iflist(int af, struct rt_walkarg *w, int type)
   1131 {
   1132 	struct ifnet *ifp;
   1133 	struct ifaddr *ifa;
   1134 	struct	rt_addrinfo info;
   1135 	int	len, error = 0;
   1136 
   1137 	memset(&info, 0, sizeof(info));
   1138 	IFNET_FOREACH(ifp) {
   1139 		if (w->w_arg && w->w_arg != ifp->if_index)
   1140 			continue;
   1141 		if (IFADDR_EMPTY(ifp))
   1142 			continue;
   1143 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
   1144 		switch (type) {
   1145 		case NET_RT_IFLIST:
   1146 			error = rt_msg2(RTM_IFINFO, &info, NULL, w, &len);
   1147 			break;
   1148 #ifdef COMPAT_14
   1149 		case NET_RT_OOIFLIST:
   1150 			error = rt_msg2(RTM_OOIFINFO, &info, NULL, w, &len);
   1151 			break;
   1152 #endif
   1153 #ifdef COMPAT_50
   1154 		case NET_RT_OIFLIST:
   1155 			error = rt_msg2(RTM_OIFINFO, &info, NULL, w, &len);
   1156 			break;
   1157 #endif
   1158 		default:
   1159 			panic("sysctl_iflist(1)");
   1160 		}
   1161 		if (error)
   1162 			return error;
   1163 		info.rti_info[RTAX_IFP] = NULL;
   1164 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1165 			switch (type) {
   1166 			case NET_RT_IFLIST: {
   1167 				struct if_xmsghdr *ifm;
   1168 
   1169 				ifm = (struct if_xmsghdr *)w->w_tmem;
   1170 				ifm->ifm_index = ifp->if_index;
   1171 				ifm->ifm_flags = ifp->if_flags;
   1172 				ifm->ifm_data = ifp->if_data;
   1173 				ifm->ifm_addrs = info.rti_addrs;
   1174 				error = copyout(ifm, w->w_where, len);
   1175 				if (error)
   1176 					return error;
   1177 				w->w_where = (char *)w->w_where + len;
   1178 				break;
   1179 			}
   1180 
   1181 #ifdef COMPAT_14
   1182 			case NET_RT_OOIFLIST:
   1183 				error = compat_14_iflist(ifp, w, &info, len);
   1184 				if (error)
   1185 					return error;
   1186 				break;
   1187 #endif
   1188 #ifdef COMPAT_50
   1189 			case NET_RT_OIFLIST:
   1190 				error = compat_50_iflist(ifp, w, &info, len);
   1191 				if (error)
   1192 					return error;
   1193 				break;
   1194 #endif
   1195 			default:
   1196 				panic("sysctl_iflist(2)");
   1197 			}
   1198 		}
   1199 		IFADDR_FOREACH(ifa, ifp) {
   1200 			if (af && af != ifa->ifa_addr->sa_family)
   1201 				continue;
   1202 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
   1203 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
   1204 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
   1205 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
   1206 				return error;
   1207 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1208 				struct ifa_xmsghdr *ifam;
   1209 
   1210 				ifam = (struct ifa_xmsghdr *)w->w_tmem;
   1211 				ifam->ifam_index = ifa->ifa_ifp->if_index;
   1212 				ifam->ifam_flags = ifa->ifa_flags;
   1213 				ifam->ifam_metric = ifa->ifa_metric;
   1214 				ifam->ifam_addrs = info.rti_addrs;
   1215 				error = copyout(w->w_tmem, w->w_where, len);
   1216 				if (error)
   1217 					return error;
   1218 				w->w_where = (char *)w->w_where + len;
   1219 			}
   1220 		}
   1221 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
   1222 		    info.rti_info[RTAX_BRD] = NULL;
   1223 	}
   1224 	return 0;
   1225 }
   1226 
   1227 static int
   1228 sysctl_rtable(SYSCTLFN_ARGS)
   1229 {
   1230 	void 	*where = oldp;
   1231 	size_t	*given = oldlenp;
   1232 	const void *new = newp;
   1233 	int	i, s, error = EINVAL;
   1234 	u_char  af;
   1235 	struct	rt_walkarg w;
   1236 
   1237 	if (namelen == 1 && name[0] == CTL_QUERY)
   1238 		return sysctl_query(SYSCTLFN_CALL(rnode));
   1239 
   1240 	if (new)
   1241 		return EPERM;
   1242 	if (namelen != 3)
   1243 		return EINVAL;
   1244 	af = name[0];
   1245 	w.w_tmemneeded = 0;
   1246 	w.w_tmemsize = 0;
   1247 	w.w_tmem = NULL;
   1248 again:
   1249 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
   1250 	if (w.w_tmemneeded) {
   1251 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
   1252 		w.w_tmemsize = w.w_tmemneeded;
   1253 		w.w_tmemneeded = 0;
   1254 	}
   1255 	w.w_op = name[1];
   1256 	w.w_arg = name[2];
   1257 	w.w_given = *given;
   1258 	w.w_needed = 0 - w.w_given;
   1259 	w.w_where = where;
   1260 
   1261 	s = splsoftnet();
   1262 	switch (w.w_op) {
   1263 
   1264 	case NET_RT_DUMP:
   1265 	case NET_RT_FLAGS:
   1266 		for (i = 1; i <= AF_MAX; i++)
   1267 			if ((af == 0 || af == i) &&
   1268 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
   1269 				break;
   1270 		break;
   1271 
   1272 #ifdef COMPAT_14
   1273 	case NET_RT_OOIFLIST:
   1274 		error = sysctl_iflist(af, &w, w.w_op);
   1275 		break;
   1276 #endif
   1277 #ifdef COMPAT_50
   1278 	case NET_RT_OIFLIST:
   1279 		error = sysctl_iflist(af, &w, w.w_op);
   1280 		break;
   1281 #endif
   1282 	case NET_RT_IFLIST:
   1283 		error = sysctl_iflist(af, &w, w.w_op);
   1284 		break;
   1285 	}
   1286 	splx(s);
   1287 
   1288 	/* check to see if we couldn't allocate memory with NOWAIT */
   1289 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
   1290 		goto again;
   1291 
   1292 	if (w.w_tmem)
   1293 		free(w.w_tmem, M_RTABLE);
   1294 	w.w_needed += w.w_given;
   1295 	if (where) {
   1296 		*given = (char *)w.w_where - (char *)where;
   1297 		if (*given < w.w_needed)
   1298 			return ENOMEM;
   1299 	} else {
   1300 		*given = (11 * w.w_needed) / 10;
   1301 	}
   1302 	return error;
   1303 }
   1304 
   1305 /*
   1306  * Routing message software interrupt routine
   1307  */
   1308 static void
   1309 COMPATNAME(route_intr)(void *cookie)
   1310 {
   1311 	struct sockproto proto = { .sp_family = PF_XROUTE, };
   1312 	struct route_info * const ri = &COMPATNAME(route_info);
   1313 	struct mbuf *m;
   1314 	int s;
   1315 
   1316 	mutex_enter(softnet_lock);
   1317 	KERNEL_LOCK(1, NULL);
   1318 	while (!IF_IS_EMPTY(&ri->ri_intrq)) {
   1319 		s = splnet();
   1320 		IF_DEQUEUE(&ri->ri_intrq, m);
   1321 		splx(s);
   1322 		if (m == NULL)
   1323 			break;
   1324 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
   1325 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst);
   1326 	}
   1327 	KERNEL_UNLOCK_ONE(NULL);
   1328 	mutex_exit(softnet_lock);
   1329 }
   1330 
   1331 /*
   1332  * Enqueue a message to the software interrupt routine.
   1333  */
   1334 void
   1335 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
   1336 {
   1337 	struct route_info * const ri = &COMPATNAME(route_info);
   1338 	int s, wasempty;
   1339 
   1340 	s = splnet();
   1341 	if (IF_QFULL(&ri->ri_intrq)) {
   1342 		IF_DROP(&ri->ri_intrq);
   1343 		m_freem(m);
   1344 	} else {
   1345 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
   1346 		M_SETCTX(m, (uintptr_t)family);
   1347 		IF_ENQUEUE(&ri->ri_intrq, m);
   1348 		if (wasempty)
   1349 			softint_schedule(ri->ri_sih);
   1350 	}
   1351 	splx(s);
   1352 }
   1353 
   1354 static void
   1355 COMPATNAME(route_init)(void)
   1356 {
   1357 	struct route_info * const ri = &COMPATNAME(route_info);
   1358 
   1359 #ifndef COMPAT_RTSOCK
   1360 	rt_init();
   1361 #endif
   1362 
   1363 	sysctl_net_route_setup(NULL);
   1364 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
   1365 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
   1366 	    COMPATNAME(route_intr), NULL);
   1367 }
   1368 
   1369 /*
   1370  * Definitions of protocols supported in the ROUTE domain.
   1371  */
   1372 #ifndef COMPAT_RTSOCK
   1373 PR_WRAP_USRREQS(route);
   1374 #else
   1375 PR_WRAP_USRREQS(compat_50_route);
   1376 #endif
   1377 
   1378 static const struct pr_usrreqs route_usrreqs = {
   1379 	.pr_attach	= COMPATNAME(route_attach_wrapper),
   1380 	.pr_detach	= COMPATNAME(route_detach_wrapper),
   1381 	.pr_ioctl	= COMPATNAME(route_ioctl_wrapper),
   1382 	.pr_stat	= COMPATNAME(route_stat_wrapper),
   1383 	.pr_peeraddr	= COMPATNAME(route_peeraddr_wrapper),
   1384 	.pr_sockaddr	= COMPATNAME(route_sockaddr_wrapper),
   1385 	.pr_generic	= COMPATNAME(route_usrreq_wrapper),
   1386 };
   1387 
   1388 static const struct protosw COMPATNAME(route_protosw)[] = {
   1389 	{
   1390 		.pr_type = SOCK_RAW,
   1391 		.pr_domain = &COMPATNAME(routedomain),
   1392 		.pr_flags = PR_ATOMIC|PR_ADDR,
   1393 		.pr_input = raw_input,
   1394 		.pr_output = COMPATNAME(route_output),
   1395 		.pr_ctlinput = raw_ctlinput,
   1396 		.pr_usrreqs = &route_usrreqs,
   1397 		.pr_init = raw_init,
   1398 	},
   1399 };
   1400 
   1401 struct domain COMPATNAME(routedomain) = {
   1402 	.dom_family = PF_XROUTE,
   1403 	.dom_name = DOMAINNAME,
   1404 	.dom_init = COMPATNAME(route_init),
   1405 	.dom_protosw = COMPATNAME(route_protosw),
   1406 	.dom_protoswNPROTOSW =
   1407 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
   1408 };
   1409 
   1410 static void
   1411 sysctl_net_route_setup(struct sysctllog **clog)
   1412 {
   1413 	const struct sysctlnode *rnode = NULL;
   1414 
   1415 	sysctl_createv(clog, 0, NULL, &rnode,
   1416 		       CTLFLAG_PERMANENT,
   1417 		       CTLTYPE_NODE, DOMAINNAME,
   1418 		       SYSCTL_DESCR("PF_ROUTE information"),
   1419 		       NULL, 0, NULL, 0,
   1420 		       CTL_NET, PF_XROUTE, CTL_EOL);
   1421 
   1422 	sysctl_createv(clog, 0, NULL, NULL,
   1423 		       CTLFLAG_PERMANENT,
   1424 		       CTLTYPE_NODE, "rtable",
   1425 		       SYSCTL_DESCR("Routing table information"),
   1426 		       sysctl_rtable, 0, NULL, 0,
   1427 		       CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
   1428 
   1429 	sysctl_createv(clog, 0, &rnode, NULL,
   1430 		       CTLFLAG_PERMANENT,
   1431 		       CTLTYPE_STRUCT, "stats",
   1432 		       SYSCTL_DESCR("Routing statistics"),
   1433 		       NULL, 0, &rtstat, sizeof(rtstat),
   1434 		       CTL_CREATE, CTL_EOL);
   1435 }
   1436