Home | History | Annotate | Line # | Download | only in net
rtsock.c revision 1.132.4.2
      1 /*	$NetBSD: rtsock.c,v 1.132.4.2 2011/02/17 12:00:49 bouyer 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 1.132.4.2 2011/02/17 12:00:49 bouyer 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/mbuf.h>
     76 #include <sys/socket.h>
     77 #include <sys/socketvar.h>
     78 #include <sys/domain.h>
     79 #include <sys/protosw.h>
     80 #include <sys/sysctl.h>
     81 #include <sys/kauth.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 #include <machine/stdarg.h>
    135 
    136 struct route_info COMPATNAME(route_info) = {
    137 	.ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, },
    138 	.ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, },
    139 	.ri_maxqlen = IFQ_MAXLEN,
    140 };
    141 
    142 #define	PRESERVED_RTF	(RTF_UP | RTF_GATEWAY | RTF_HOST | RTF_DONE | RTF_MASK)
    143 
    144 static void COMPATNAME(route_init)(void);
    145 static int COMPATNAME(route_output)(struct mbuf *, ...);
    146 static int COMPATNAME(route_usrreq)(struct socket *,
    147 	    int, struct mbuf *, struct mbuf *, struct mbuf *, struct lwp *);
    148 
    149 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
    150 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
    151 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
    152     struct rt_addrinfo *);
    153 static void rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *);
    154 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *);
    155 static void sysctl_net_route_setup(struct sysctllog **);
    156 static int sysctl_dumpentry(struct rtentry *, void *);
    157 static int sysctl_iflist(int, struct rt_walkarg *, int);
    158 static int sysctl_rtable(SYSCTLFN_PROTO);
    159 static void rt_adjustcount(int, int);
    160 
    161 static void
    162 rt_adjustcount(int af, int cnt)
    163 {
    164 	struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
    165 
    166 	cb->any_count += cnt;
    167 
    168 	switch (af) {
    169 	case AF_INET:
    170 		cb->ip_count += cnt;
    171 		return;
    172 #ifdef INET6
    173 	case AF_INET6:
    174 		cb->ip6_count += cnt;
    175 		return;
    176 #endif
    177 	case AF_ISO:
    178 		cb->iso_count += cnt;
    179 		return;
    180 	case AF_MPLS:
    181 		cb->mpls_count += cnt;
    182 		return;
    183 	}
    184 }
    185 
    186 /*ARGSUSED*/
    187 int
    188 COMPATNAME(route_usrreq)(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
    189 	struct mbuf *control, struct lwp *l)
    190 {
    191 	int error = 0;
    192 	struct rawcb *rp = sotorawcb(so);
    193 	int s;
    194 
    195 	if (req == PRU_ATTACH) {
    196 		sosetlock(so);
    197 		rp = malloc(sizeof(*rp), M_PCB, M_WAITOK|M_ZERO);
    198 		so->so_pcb = rp;
    199 	}
    200 	if (req == PRU_DETACH && rp)
    201 		rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
    202 	s = splsoftnet();
    203 
    204 	/*
    205 	 * Don't call raw_usrreq() in the attach case, because
    206 	 * we want to allow non-privileged processes to listen on
    207 	 * and send "safe" commands to the routing socket.
    208 	 */
    209 	if (req == PRU_ATTACH) {
    210 		if (l == NULL)
    211 			error = EACCES;
    212 		else
    213 			error = raw_attach(so, (int)(long)nam);
    214 	} else
    215 		error = raw_usrreq(so, req, m, nam, control, l);
    216 
    217 	rp = sotorawcb(so);
    218 	if (req == PRU_ATTACH && rp) {
    219 		if (error) {
    220 			free(rp, M_PCB);
    221 			splx(s);
    222 			return error;
    223 		}
    224 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
    225 		rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
    226 		rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
    227 		soisconnected(so);
    228 		so->so_options |= SO_USELOOPBACK;
    229 	}
    230 	splx(s);
    231 	return error;
    232 }
    233 
    234 static const struct sockaddr *
    235 intern_netmask(const struct sockaddr *mask)
    236 {
    237 	struct radix_node *rn;
    238 	extern struct radix_node_head *mask_rnhead;
    239 
    240 	if (mask != NULL &&
    241 	    (rn = rn_search(mask, mask_rnhead->rnh_treetop)))
    242 		mask = (const struct sockaddr *)rn->rn_key;
    243 
    244 	return mask;
    245 }
    246 
    247 /*ARGSUSED*/
    248 int
    249 COMPATNAME(route_output)(struct mbuf *m, ...)
    250 {
    251 	struct sockproto proto = { .sp_family = PF_XROUTE, };
    252 	struct rt_xmsghdr *rtm = NULL;
    253 	struct rt_xmsghdr *old_rtm = NULL;
    254 	struct rtentry *rt = NULL;
    255 	struct rtentry *saved_nrt = NULL;
    256 	struct rt_addrinfo info;
    257 	int len, error = 0;
    258 	struct ifnet *ifp = NULL;
    259 	struct ifaddr *ifa = NULL;
    260 	struct socket *so;
    261 	va_list ap;
    262 	sa_family_t family;
    263 
    264 	va_start(ap, m);
    265 	so = va_arg(ap, struct socket *);
    266 	va_end(ap);
    267 
    268 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
    269 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
    270 	   (m = m_pullup(m, sizeof(int32_t))) == NULL))
    271 		return ENOBUFS;
    272 	if ((m->m_flags & M_PKTHDR) == 0)
    273 		panic("%s", __func__);
    274 	len = m->m_pkthdr.len;
    275 	if (len < sizeof(*rtm) ||
    276 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
    277 		info.rti_info[RTAX_DST] = NULL;
    278 		senderr(EINVAL);
    279 	}
    280 	R_Malloc(rtm, struct rt_xmsghdr *, len);
    281 	if (rtm == NULL) {
    282 		info.rti_info[RTAX_DST] = NULL;
    283 		senderr(ENOBUFS);
    284 	}
    285 	m_copydata(m, 0, len, rtm);
    286 	if (rtm->rtm_version != RTM_XVERSION) {
    287 		info.rti_info[RTAX_DST] = NULL;
    288 		senderr(EPROTONOSUPPORT);
    289 	}
    290 	rtm->rtm_pid = curproc->p_pid;
    291 	memset(&info, 0, sizeof(info));
    292 	info.rti_addrs = rtm->rtm_addrs;
    293 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
    294 	    &info)) {
    295 		senderr(EINVAL);
    296 	}
    297 	info.rti_flags = rtm->rtm_flags;
    298 #ifdef RTSOCK_DEBUG
    299 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
    300 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
    301 		    inet_ntoa(((const struct sockaddr_in *)
    302 		    info.rti_info[RTAX_DST])->sin_addr));
    303 	}
    304 #endif /* RTSOCK_DEBUG */
    305 	if (info.rti_info[RTAX_DST] == NULL ||
    306 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
    307 		senderr(EINVAL);
    308 	}
    309 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
    310 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
    311 		senderr(EINVAL);
    312 	}
    313 
    314 	/*
    315 	 * Verify that the caller has the appropriate privilege; RTM_GET
    316 	 * is the only operation the non-superuser is allowed.
    317 	 */
    318 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
    319 	    0, rtm, NULL, NULL) != 0)
    320 		senderr(EACCES);
    321 
    322 	switch (rtm->rtm_type) {
    323 
    324 	case RTM_ADD:
    325 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
    326 			senderr(EINVAL);
    327 		}
    328 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
    329 		if (error == 0 && saved_nrt) {
    330 			rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
    331 			saved_nrt->rt_refcnt--;
    332 		}
    333 		break;
    334 
    335 	case RTM_DELETE:
    336 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
    337 		if (error == 0) {
    338 			(rt = saved_nrt)->rt_refcnt++;
    339 			goto report;
    340 		}
    341 		break;
    342 
    343 	case RTM_GET:
    344 	case RTM_CHANGE:
    345 	case RTM_LOCK:
    346                 /* XXX This will mask info.rti_info[RTAX_DST] with
    347 		 * info.rti_info[RTAX_NETMASK] before
    348                  * searching.  It did not used to do that.  --dyoung
    349 		 */
    350 		error = rtrequest1(RTM_GET, &info, &rt);
    351 		if (error != 0)
    352 			senderr(error);
    353 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
    354 			struct radix_node *rn;
    355 
    356 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
    357 			    info.rti_info[RTAX_DST]->sa_len) != 0)
    358 				senderr(ESRCH);
    359 			info.rti_info[RTAX_NETMASK] = intern_netmask(
    360 			    info.rti_info[RTAX_NETMASK]);
    361 			for (rn = rt->rt_nodes; rn; rn = rn->rn_dupedkey)
    362 				if (info.rti_info[RTAX_NETMASK] ==
    363 				    (const struct sockaddr *)rn->rn_mask)
    364 					break;
    365 			if (rn == NULL)
    366 				senderr(ETOOMANYREFS);
    367 			rt = (struct rtentry *)rn;
    368 		}
    369 
    370 		switch (rtm->rtm_type) {
    371 		case RTM_GET:
    372 		report:
    373 			info.rti_info[RTAX_DST] = rt_getkey(rt);
    374 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    375 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    376 			info.rti_info[RTAX_TAG] = (struct sockaddr*)rt_gettag(rt);
    377 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
    378 				;
    379 			else if ((ifp = rt->rt_ifp) != NULL) {
    380 				const struct ifaddr *rtifa;
    381 				info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
    382                                 /* rtifa used to be simply rt->rt_ifa.
    383                                  * If rt->rt_ifa != NULL, then
    384                                  * rt_get_ifa() != NULL.  So this
    385                                  * ought to still be safe. --dyoung
    386 				 */
    387 				rtifa = rt_get_ifa(rt);
    388 				info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
    389 #ifdef RTSOCK_DEBUG
    390 				if (info.rti_info[RTAX_IFA]->sa_family ==
    391 				    AF_INET) {
    392 					printf("%s: copying out RTAX_IFA %s ",
    393 					    __func__, inet_ntoa(
    394 					    ((const struct sockaddr_in *)
    395 					    info.rti_info[RTAX_IFA])->sin_addr)
    396 					    );
    397 					printf("for info.rti_info[RTAX_DST] %s "
    398 					    "ifa_getifa %p ifa_seqno %p\n",
    399 					    inet_ntoa(
    400 					    ((const struct sockaddr_in *)
    401 					    info.rti_info[RTAX_DST])->sin_addr),
    402 					    (void *)rtifa->ifa_getifa,
    403 					    rtifa->ifa_seqno);
    404 				}
    405 #endif /* RTSOCK_DEBUG */
    406 				if (ifp->if_flags & IFF_POINTOPOINT) {
    407 					info.rti_info[RTAX_BRD] =
    408 					    rtifa->ifa_dstaddr;
    409 				} else
    410 					info.rti_info[RTAX_BRD] = NULL;
    411 				rtm->rtm_index = ifp->if_index;
    412 			} else {
    413 				info.rti_info[RTAX_IFP] = NULL;
    414 				info.rti_info[RTAX_IFA] = NULL;
    415 			}
    416 			(void)rt_msg2(rtm->rtm_type, &info, NULL, NULL, &len);
    417 			if (len > rtm->rtm_msglen) {
    418 				old_rtm = rtm;
    419 				R_Malloc(rtm, struct rt_xmsghdr *, len);
    420 				if (rtm == NULL)
    421 					senderr(ENOBUFS);
    422 				(void)memcpy(rtm, old_rtm, old_rtm->rtm_msglen);
    423 			}
    424 			(void)rt_msg2(rtm->rtm_type, &info, rtm, NULL, 0);
    425 			rtm->rtm_flags = rt->rt_flags;
    426 			rtm_setmetrics(rt, rtm);
    427 			rtm->rtm_addrs = info.rti_addrs;
    428 			break;
    429 
    430 		case RTM_CHANGE:
    431 			/*
    432 			 * new gateway could require new ifaddr, ifp;
    433 			 * flags may also be different; ifp may be specified
    434 			 * by ll sockaddr when protocol address is ambiguous
    435 			 */
    436 			if ((error = rt_getifa(&info)) != 0)
    437 				senderr(error);
    438 			if (info.rti_info[RTAX_GATEWAY] &&
    439 			    rt_setgate(rt, info.rti_info[RTAX_GATEWAY]))
    440 				senderr(EDQUOT);
    441 			if (info.rti_info[RTAX_TAG])
    442 				rt_settag(rt, info.rti_info[RTAX_TAG]);
    443 			/* new gateway could require new ifaddr, ifp;
    444 			   flags may also be different; ifp may be specified
    445 			   by ll sockaddr when protocol address is ambiguous */
    446 			if (info.rti_info[RTAX_IFP] &&
    447 			    (ifa = ifa_ifwithnet(info.rti_info[RTAX_IFP])) &&
    448 			    (ifp = ifa->ifa_ifp) && (info.rti_info[RTAX_IFA] ||
    449 			    info.rti_info[RTAX_GATEWAY])) {
    450 				if (info.rti_info[RTAX_IFA] == NULL ||
    451 				    (ifa = ifa_ifwithaddr(
    452 				    info.rti_info[RTAX_IFA])) == NULL)
    453 					ifa = ifaof_ifpforaddr(
    454 					    info.rti_info[RTAX_IFA] ?
    455 					    info.rti_info[RTAX_IFA] :
    456 					    info.rti_info[RTAX_GATEWAY], ifp);
    457 			} else if ((info.rti_info[RTAX_IFA] &&
    458 			    (ifa = ifa_ifwithaddr(info.rti_info[RTAX_IFA]))) ||
    459 			    (info.rti_info[RTAX_GATEWAY] &&
    460 			    (ifa = ifa_ifwithroute(rt->rt_flags,
    461 			    rt_getkey(rt), info.rti_info[RTAX_GATEWAY])))) {
    462 				ifp = ifa->ifa_ifp;
    463 			}
    464 			if (ifa) {
    465 				struct ifaddr *oifa = rt->rt_ifa;
    466 				if (oifa != ifa) {
    467 					if (oifa && oifa->ifa_rtrequest) {
    468 						oifa->ifa_rtrequest(RTM_DELETE,
    469 						    rt, &info);
    470 					}
    471 					rt_replace_ifa(rt, ifa);
    472 					rt->rt_ifp = ifp;
    473 				}
    474 			}
    475 			if (ifp && rt->rt_ifp != ifp)
    476 				rt->rt_ifp = ifp;
    477 			rt_setmetrics(rtm->rtm_inits, rtm, rt);
    478 			if (rt->rt_flags != info.rti_flags)
    479 				rt->rt_flags = (info.rti_flags & ~PRESERVED_RTF)
    480 				    | (rt->rt_flags & PRESERVED_RTF);
    481 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
    482 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
    483 			/*FALLTHROUGH*/
    484 		case RTM_LOCK:
    485 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
    486 			rt->rt_rmx.rmx_locks |=
    487 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
    488 			break;
    489 		}
    490 		break;
    491 
    492 	default:
    493 		senderr(EOPNOTSUPP);
    494 	}
    495 
    496 flush:
    497 	if (rtm) {
    498 		if (error)
    499 			rtm->rtm_errno = error;
    500 		else
    501 			rtm->rtm_flags |= RTF_DONE;
    502 	}
    503 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
    504 	    0;
    505 	/* We cannot free old_rtm until we have stopped using the
    506 	 * pointers in info, some of which may point to sockaddrs
    507 	 * in old_rtm.
    508 	 */
    509 	if (old_rtm != NULL)
    510 		Free(old_rtm);
    511 	if (rt)
    512 		rtfree(rt);
    513     {
    514 	struct rawcb *rp = NULL;
    515 	/*
    516 	 * Check to see if we don't want our own messages.
    517 	 */
    518 	if ((so->so_options & SO_USELOOPBACK) == 0) {
    519 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
    520 			if (rtm)
    521 				Free(rtm);
    522 			m_freem(m);
    523 			return error;
    524 		}
    525 		/* There is another listener, so construct message */
    526 		rp = sotorawcb(so);
    527 	}
    528 	if (rtm) {
    529 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
    530 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
    531 			m_freem(m);
    532 			m = NULL;
    533 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
    534 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
    535 		Free(rtm);
    536 	}
    537 	if (rp)
    538 		rp->rcb_proto.sp_family = 0; /* Avoid us */
    539 	if (family)
    540 		proto.sp_protocol = family;
    541 	if (m)
    542 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
    543 		    &COMPATNAME(route_info).ri_dst);
    544 	if (rp)
    545 		rp->rcb_proto.sp_family = PF_XROUTE;
    546     }
    547 	return error;
    548 }
    549 
    550 static void
    551 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
    552 {
    553 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
    554 	metric(RTV_RPIPE, rmx_recvpipe);
    555 	metric(RTV_SPIPE, rmx_sendpipe);
    556 	metric(RTV_SSTHRESH, rmx_ssthresh);
    557 	metric(RTV_RTT, rmx_rtt);
    558 	metric(RTV_RTTVAR, rmx_rttvar);
    559 	metric(RTV_HOPCOUNT, rmx_hopcount);
    560 	metric(RTV_MTU, rmx_mtu);
    561 	metric(RTV_EXPIRE, rmx_expire);
    562 #undef metric
    563 }
    564 
    565 static void
    566 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
    567 {
    568 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
    569 	metric(rmx_recvpipe);
    570 	metric(rmx_sendpipe);
    571 	metric(rmx_ssthresh);
    572 	metric(rmx_rtt);
    573 	metric(rmx_rttvar);
    574 	metric(rmx_hopcount);
    575 	metric(rmx_mtu);
    576 	metric(rmx_expire);
    577 #undef metric
    578 }
    579 
    580 static int
    581 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
    582     struct rt_addrinfo *rtinfo)
    583 {
    584 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
    585 	int i;
    586 
    587 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
    588 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
    589 			continue;
    590 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
    591 		RT_XADVANCE(cp, sa);
    592 	}
    593 
    594 	/*
    595 	 * Check for extra addresses specified, except RTM_GET asking
    596 	 * for interface info.
    597 	 */
    598 	if (rtmtype == RTM_GET) {
    599 		if (((rtinfo->rti_addrs &
    600 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
    601 			return 1;
    602 	} else if ((rtinfo->rti_addrs & (~0 << i)) != 0)
    603 		return 1;
    604 	/* Check for bad data length.  */
    605 	if (cp != cplim) {
    606 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
    607 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
    608 			/*
    609 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
    610 			 * We accept this for now for the sake of old
    611 			 * binaries or third party softwares.
    612 			 */
    613 			;
    614 		else
    615 			return 1;
    616 	}
    617 	return 0;
    618 }
    619 
    620 static int
    621 rt_getlen(int type)
    622 {
    623 #ifndef COMPAT_RTSOCK
    624 	CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
    625 	CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
    626 	CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
    627 	CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
    628 #endif
    629 
    630 	switch (type) {
    631 	case RTM_DELADDR:
    632 	case RTM_NEWADDR:
    633 	case RTM_CHGADDR:
    634 		return sizeof(struct ifa_xmsghdr);
    635 
    636 	case RTM_OOIFINFO:
    637 #ifdef COMPAT_14
    638 		return sizeof(struct if_msghdr14);
    639 #else
    640 #ifdef DIAGNOSTIC
    641 		printf("RTM_OOIFINFO\n");
    642 #endif
    643 		return -1;
    644 #endif
    645 	case RTM_OIFINFO:
    646 #ifdef COMPAT_50
    647 		return sizeof(struct if_msghdr50);
    648 #else
    649 #ifdef DIAGNOSTIC
    650 		printf("RTM_OIFINFO\n");
    651 #endif
    652 		return -1;
    653 #endif
    654 
    655 	case RTM_IFINFO:
    656 		return sizeof(struct if_xmsghdr);
    657 
    658 	case RTM_IFANNOUNCE:
    659 	case RTM_IEEE80211:
    660 		return sizeof(struct if_xannouncemsghdr);
    661 
    662 	default:
    663 		return sizeof(struct rt_xmsghdr);
    664 	}
    665 }
    666 
    667 
    668 struct mbuf *
    669 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
    670 {
    671 	struct rt_xmsghdr *rtm;
    672 	struct mbuf *m;
    673 	int i;
    674 	const struct sockaddr *sa;
    675 	int len, dlen;
    676 
    677 	m = m_gethdr(M_DONTWAIT, MT_DATA);
    678 	if (m == NULL)
    679 		return m;
    680 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
    681 
    682 	if ((len = rt_getlen(type)) == -1)
    683 		goto out;
    684 	if (len > MHLEN + MLEN)
    685 		panic("%s: message too long", __func__);
    686 	else if (len > MHLEN) {
    687 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
    688 		if (m->m_next == NULL)
    689 			goto out;
    690 		MCLAIM(m->m_next, m->m_owner);
    691 		m->m_pkthdr.len = len;
    692 		m->m_len = MHLEN;
    693 		m->m_next->m_len = len - MHLEN;
    694 	} else {
    695 		m->m_pkthdr.len = m->m_len = len;
    696 	}
    697 	m->m_pkthdr.rcvif = NULL;
    698 	m_copyback(m, 0, datalen, data);
    699 	if (len > datalen)
    700 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
    701 	rtm = mtod(m, struct rt_xmsghdr *);
    702 	for (i = 0; i < RTAX_MAX; i++) {
    703 		if ((sa = rtinfo->rti_info[i]) == NULL)
    704 			continue;
    705 		rtinfo->rti_addrs |= (1 << i);
    706 		dlen = RT_XROUNDUP(sa->sa_len);
    707 		m_copyback(m, len, sa->sa_len, sa);
    708 		if (dlen != sa->sa_len) {
    709 			m_copyback(m, len + sa->sa_len,
    710 			    dlen - sa->sa_len, "\0\0\0\0\0\0");
    711 		}
    712 		len += dlen;
    713 	}
    714 	if (m->m_pkthdr.len != len)
    715 		goto out;
    716 	rtm->rtm_msglen = len;
    717 	rtm->rtm_version = RTM_XVERSION;
    718 	rtm->rtm_type = type;
    719 	return m;
    720 out:
    721 	m_freem(m);
    722 	return NULL;
    723 }
    724 
    725 /*
    726  * rt_msg2
    727  *
    728  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
    729  *		returns the length of the message in 'lenp'.
    730  *
    731  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
    732  *	the message
    733  * otherwise walkarg's w_needed is updated and if the user buffer is
    734  *	specified and w_needed indicates space exists the information is copied
    735  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
    736  *	if the allocation fails ENOBUFS is returned.
    737  */
    738 static int
    739 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
    740 	int *lenp)
    741 {
    742 	int i;
    743 	int len, dlen, second_time = 0;
    744 	char *cp0, *cp = cpv;
    745 
    746 	rtinfo->rti_addrs = 0;
    747 again:
    748 	if ((len = rt_getlen(type)) == -1)
    749 		return EINVAL;
    750 
    751 	if ((cp0 = cp) != NULL)
    752 		cp += len;
    753 	for (i = 0; i < RTAX_MAX; i++) {
    754 		const struct sockaddr *sa;
    755 
    756 		if ((sa = rtinfo->rti_info[i]) == NULL)
    757 			continue;
    758 		rtinfo->rti_addrs |= (1 << i);
    759 		dlen = RT_XROUNDUP(sa->sa_len);
    760 		if (cp) {
    761 			(void)memcpy(cp, sa, (size_t)dlen);
    762 			cp += dlen;
    763 		}
    764 		len += dlen;
    765 	}
    766 	if (cp == NULL && w != NULL && !second_time) {
    767 		struct rt_walkarg *rw = w;
    768 
    769 		rw->w_needed += len;
    770 		if (rw->w_needed <= 0 && rw->w_where) {
    771 			if (rw->w_tmemsize < len) {
    772 				if (rw->w_tmem)
    773 					free(rw->w_tmem, M_RTABLE);
    774 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
    775 				if (rw->w_tmem)
    776 					rw->w_tmemsize = len;
    777 				else
    778 					rw->w_tmemsize = 0;
    779 			}
    780 			if (rw->w_tmem) {
    781 				cp = rw->w_tmem;
    782 				second_time = 1;
    783 				goto again;
    784 			} else {
    785 				rw->w_tmemneeded = len;
    786 				return ENOBUFS;
    787 			}
    788 		}
    789 	}
    790 	if (cp) {
    791 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
    792 
    793 		rtm->rtm_version = RTM_XVERSION;
    794 		rtm->rtm_type = type;
    795 		rtm->rtm_msglen = len;
    796 	}
    797 	if (lenp)
    798 		*lenp = len;
    799 	return 0;
    800 }
    801 
    802 /*
    803  * This routine is called to generate a message from the routing
    804  * socket indicating that a redirect has occurred, a routing lookup
    805  * has failed, or that a protocol has detected timeouts to a particular
    806  * destination.
    807  */
    808 void
    809 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
    810     int error)
    811 {
    812 	struct rt_xmsghdr rtm;
    813 	struct mbuf *m;
    814 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
    815 	struct rt_addrinfo info = *rtinfo;
    816 
    817 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
    818 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    819 		return;
    820 	memset(&rtm, 0, sizeof(rtm));
    821 	rtm.rtm_flags = RTF_DONE | flags;
    822 	rtm.rtm_errno = error;
    823 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
    824 	if (m == NULL)
    825 		return;
    826 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
    827 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
    828 }
    829 
    830 /*
    831  * This routine is called to generate a message from the routing
    832  * socket indicating that the status of a network interface has changed.
    833  */
    834 void
    835 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
    836 {
    837 	struct if_xmsghdr ifm;
    838 	struct mbuf *m;
    839 	struct rt_addrinfo info;
    840 
    841 	COMPATCALL(rt_ifmsg, (ifp));
    842 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    843 		return;
    844 	(void)memset(&info, 0, sizeof(info));
    845 	(void)memset(&ifm, 0, sizeof(ifm));
    846 	ifm.ifm_index = ifp->if_index;
    847 	ifm.ifm_flags = ifp->if_flags;
    848 	ifm.ifm_data = ifp->if_data;
    849 	ifm.ifm_addrs = 0;
    850 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
    851 	if (m == NULL)
    852 		return;
    853 	COMPATNAME(route_enqueue)(m, 0);
    854 #ifdef COMPAT_14
    855 	compat_14_rt_oifmsg(ifp);
    856 #endif
    857 #ifdef COMPAT_50
    858 	compat_50_rt_oifmsg(ifp);
    859 #endif
    860 }
    861 
    862 
    863 /*
    864  * This is called to generate messages from the routing socket
    865  * indicating a network interface has had addresses associated with it.
    866  * if we ever reverse the logic and replace messages TO the routing
    867  * socket indicate a request to configure interfaces, then it will
    868  * be unnecessary as the routing socket will automatically generate
    869  * copies of it.
    870  */
    871 void
    872 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
    873     struct rtentry *rt)
    874 {
    875 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
    876 	struct rt_addrinfo info;
    877 	const struct sockaddr *sa;
    878 	int pass;
    879 	struct mbuf *m;
    880 	struct ifnet *ifp = ifa->ifa_ifp;
    881 	struct rt_xmsghdr rtm;
    882 	struct ifa_xmsghdr ifam;
    883 	int ncmd;
    884 
    885 	COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
    886 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    887 		return;
    888 	for (pass = 1; pass < 3; pass++) {
    889 		memset(&info, 0, sizeof(info));
    890 		switch (cmdpass(cmd, pass)) {
    891 		case cmdpass(RTM_ADD, 1):
    892 		case cmdpass(RTM_CHANGE, 1):
    893 		case cmdpass(RTM_DELETE, 2):
    894 			switch (cmd) {
    895 			case RTM_DELETE:
    896 				ncmd = RTM_DELADDR;
    897 				break;
    898 			case RTM_CHANGE:
    899 				ncmd = RTM_CHGADDR;
    900 				break;
    901 			default:
    902 				ncmd = RTM_NEWADDR;
    903 			}
    904 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
    905 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
    906 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
    907 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
    908 			memset(&ifam, 0, sizeof(ifam));
    909 			ifam.ifam_index = ifp->if_index;
    910 			ifam.ifam_metric = ifa->ifa_metric;
    911 			ifam.ifam_flags = ifa->ifa_flags;
    912 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
    913 			if (m == NULL)
    914 				continue;
    915 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
    916 			    info.rti_addrs;
    917 			break;
    918 		case cmdpass(RTM_ADD, 2):
    919 		case cmdpass(RTM_CHANGE, 2):
    920 		case cmdpass(RTM_DELETE, 1):
    921 			if (rt == NULL)
    922 				continue;
    923 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    924 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
    925 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    926 			memset(&rtm, 0, sizeof(rtm));
    927 			rtm.rtm_index = ifp->if_index;
    928 			rtm.rtm_flags |= rt->rt_flags;
    929 			rtm.rtm_errno = error;
    930 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
    931 			if (m == NULL)
    932 				continue;
    933 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
    934 			break;
    935 		default:
    936 			continue;
    937 		}
    938 #ifdef DIAGNOSTIC
    939 		if (m == NULL)
    940 			panic("%s: called with wrong command", __func__);
    941 #endif
    942 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
    943 	}
    944 #undef cmdpass
    945 }
    946 
    947 static struct mbuf *
    948 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
    949     struct rt_addrinfo *info)
    950 {
    951 	struct if_xannouncemsghdr ifan;
    952 
    953 	memset(info, 0, sizeof(*info));
    954 	memset(&ifan, 0, sizeof(ifan));
    955 	ifan.ifan_index = ifp->if_index;
    956 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
    957 	ifan.ifan_what = what;
    958 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
    959 }
    960 
    961 /*
    962  * This is called to generate routing socket messages indicating
    963  * network interface arrival and departure.
    964  */
    965 void
    966 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
    967 {
    968 	struct mbuf *m;
    969 	struct rt_addrinfo info;
    970 
    971 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
    972 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    973 		return;
    974 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
    975 	if (m == NULL)
    976 		return;
    977 	COMPATNAME(route_enqueue)(m, 0);
    978 }
    979 
    980 /*
    981  * This is called to generate routing socket messages indicating
    982  * IEEE80211 wireless events.
    983  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
    984  */
    985 void
    986 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
    987 	size_t data_len)
    988 {
    989 	struct mbuf *m;
    990 	struct rt_addrinfo info;
    991 
    992 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
    993 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    994 		return;
    995 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
    996 	if (m == NULL)
    997 		return;
    998 	/*
    999 	 * Append the ieee80211 data.  Try to stick it in the
   1000 	 * mbuf containing the ifannounce msg; otherwise allocate
   1001 	 * a new mbuf and append.
   1002 	 *
   1003 	 * NB: we assume m is a single mbuf.
   1004 	 */
   1005 	if (data_len > M_TRAILINGSPACE(m)) {
   1006 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
   1007 		if (n == NULL) {
   1008 			m_freem(m);
   1009 			return;
   1010 		}
   1011 		(void)memcpy(mtod(n, void *), data, data_len);
   1012 		n->m_len = data_len;
   1013 		m->m_next = n;
   1014 	} else if (data_len > 0) {
   1015 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
   1016 		m->m_len += data_len;
   1017 	}
   1018 	if (m->m_flags & M_PKTHDR)
   1019 		m->m_pkthdr.len += data_len;
   1020 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
   1021 	COMPATNAME(route_enqueue)(m, 0);
   1022 }
   1023 
   1024 /*
   1025  * This is used in dumping the kernel table via sysctl().
   1026  */
   1027 static int
   1028 sysctl_dumpentry(struct rtentry *rt, void *v)
   1029 {
   1030 	struct rt_walkarg *w = v;
   1031 	int error = 0, size;
   1032 	struct rt_addrinfo info;
   1033 
   1034 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
   1035 		return 0;
   1036 	memset(&info, 0, sizeof(info));
   1037 	info.rti_info[RTAX_DST] = rt_getkey(rt);
   1038 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
   1039 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
   1040 	if (rt->rt_ifp) {
   1041 		const struct ifaddr *rtifa;
   1042 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
   1043 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
   1044 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
   1045 		 * --dyoung
   1046 		 */
   1047 		rtifa = rt_get_ifa(rt);
   1048 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
   1049 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
   1050 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
   1051 	}
   1052 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
   1053 		return error;
   1054 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1055 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
   1056 
   1057 		rtm->rtm_flags = rt->rt_flags;
   1058 		rtm->rtm_use = rt->rt_use;
   1059 		rtm_setmetrics(rt, rtm);
   1060 		KASSERT(rt->rt_ifp != NULL);
   1061 		rtm->rtm_index = rt->rt_ifp->if_index;
   1062 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
   1063 		rtm->rtm_addrs = info.rti_addrs;
   1064 		if ((error = copyout(rtm, w->w_where, size)) != 0)
   1065 			w->w_where = NULL;
   1066 		else
   1067 			w->w_where = (char *)w->w_where + size;
   1068 	}
   1069 	return error;
   1070 }
   1071 
   1072 static int
   1073 sysctl_iflist(int af, struct rt_walkarg *w, int type)
   1074 {
   1075 	struct ifnet *ifp;
   1076 	struct ifaddr *ifa;
   1077 	struct	rt_addrinfo info;
   1078 	int	len, error = 0;
   1079 
   1080 	memset(&info, 0, sizeof(info));
   1081 	IFNET_FOREACH(ifp) {
   1082 		if (w->w_arg && w->w_arg != ifp->if_index)
   1083 			continue;
   1084 		if (IFADDR_EMPTY(ifp))
   1085 			continue;
   1086 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
   1087 		switch (type) {
   1088 		case NET_RT_IFLIST:
   1089 			error = rt_msg2(RTM_IFINFO, &info, NULL, w, &len);
   1090 			break;
   1091 #ifdef COMPAT_14
   1092 		case NET_RT_OOIFLIST:
   1093 			error = rt_msg2(RTM_OOIFINFO, &info, NULL, w, &len);
   1094 			break;
   1095 #endif
   1096 #ifdef COMPAT_50
   1097 		case NET_RT_OIFLIST:
   1098 			error = rt_msg2(RTM_OIFINFO, &info, NULL, w, &len);
   1099 			break;
   1100 #endif
   1101 		default:
   1102 			panic("sysctl_iflist(1)");
   1103 		}
   1104 		if (error)
   1105 			return error;
   1106 		info.rti_info[RTAX_IFP] = NULL;
   1107 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1108 			switch (type) {
   1109 			case NET_RT_IFLIST: {
   1110 				struct if_xmsghdr *ifm;
   1111 
   1112 				ifm = (struct if_xmsghdr *)w->w_tmem;
   1113 				ifm->ifm_index = ifp->if_index;
   1114 				ifm->ifm_flags = ifp->if_flags;
   1115 				ifm->ifm_data = ifp->if_data;
   1116 				ifm->ifm_addrs = info.rti_addrs;
   1117 				error = copyout(ifm, w->w_where, len);
   1118 				if (error)
   1119 					return error;
   1120 				w->w_where = (char *)w->w_where + len;
   1121 				break;
   1122 			}
   1123 
   1124 #ifdef COMPAT_14
   1125 			case NET_RT_OOIFLIST:
   1126 				error = compat_14_iflist(ifp, w, &info, len);
   1127 				if (error)
   1128 					return error;
   1129 				break;
   1130 #endif
   1131 #ifdef COMPAT_50
   1132 			case NET_RT_OIFLIST:
   1133 				error = compat_50_iflist(ifp, w, &info, len);
   1134 				if (error)
   1135 					return error;
   1136 				break;
   1137 #endif
   1138 			default:
   1139 				panic("sysctl_iflist(2)");
   1140 			}
   1141 		}
   1142 		IFADDR_FOREACH(ifa, ifp) {
   1143 			if (af && af != ifa->ifa_addr->sa_family)
   1144 				continue;
   1145 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
   1146 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
   1147 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
   1148 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
   1149 				return error;
   1150 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1151 				struct ifa_xmsghdr *ifam;
   1152 
   1153 				ifam = (struct ifa_xmsghdr *)w->w_tmem;
   1154 				ifam->ifam_index = ifa->ifa_ifp->if_index;
   1155 				ifam->ifam_flags = ifa->ifa_flags;
   1156 				ifam->ifam_metric = ifa->ifa_metric;
   1157 				ifam->ifam_addrs = info.rti_addrs;
   1158 				error = copyout(w->w_tmem, w->w_where, len);
   1159 				if (error)
   1160 					return error;
   1161 				w->w_where = (char *)w->w_where + len;
   1162 			}
   1163 		}
   1164 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
   1165 		    info.rti_info[RTAX_BRD] = NULL;
   1166 	}
   1167 	return 0;
   1168 }
   1169 
   1170 static int
   1171 sysctl_rtable(SYSCTLFN_ARGS)
   1172 {
   1173 	void 	*where = oldp;
   1174 	size_t	*given = oldlenp;
   1175 	const void *new = newp;
   1176 	int	i, s, error = EINVAL;
   1177 	u_char  af;
   1178 	struct	rt_walkarg w;
   1179 
   1180 	if (namelen == 1 && name[0] == CTL_QUERY)
   1181 		return sysctl_query(SYSCTLFN_CALL(rnode));
   1182 
   1183 	if (new)
   1184 		return EPERM;
   1185 	if (namelen != 3)
   1186 		return EINVAL;
   1187 	af = name[0];
   1188 	w.w_tmemneeded = 0;
   1189 	w.w_tmemsize = 0;
   1190 	w.w_tmem = NULL;
   1191 again:
   1192 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
   1193 	if (w.w_tmemneeded) {
   1194 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
   1195 		w.w_tmemsize = w.w_tmemneeded;
   1196 		w.w_tmemneeded = 0;
   1197 	}
   1198 	w.w_op = name[1];
   1199 	w.w_arg = name[2];
   1200 	w.w_given = *given;
   1201 	w.w_needed = 0 - w.w_given;
   1202 	w.w_where = where;
   1203 
   1204 	s = splsoftnet();
   1205 	switch (w.w_op) {
   1206 
   1207 	case NET_RT_DUMP:
   1208 	case NET_RT_FLAGS:
   1209 		for (i = 1; i <= AF_MAX; i++)
   1210 			if ((af == 0 || af == i) &&
   1211 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
   1212 				break;
   1213 		break;
   1214 
   1215 #ifdef COMPAT_14
   1216 	case NET_RT_OOIFLIST:
   1217 		error = sysctl_iflist(af, &w, w.w_op);
   1218 		break;
   1219 #endif
   1220 #ifdef COMPAT_50
   1221 	case NET_RT_OIFLIST:
   1222 		error = sysctl_iflist(af, &w, w.w_op);
   1223 		break;
   1224 #endif
   1225 	case NET_RT_IFLIST:
   1226 		error = sysctl_iflist(af, &w, w.w_op);
   1227 		break;
   1228 	}
   1229 	splx(s);
   1230 
   1231 	/* check to see if we couldn't allocate memory with NOWAIT */
   1232 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
   1233 		goto again;
   1234 
   1235 	if (w.w_tmem)
   1236 		free(w.w_tmem, M_RTABLE);
   1237 	w.w_needed += w.w_given;
   1238 	if (where) {
   1239 		*given = (char *)w.w_where - (char *)where;
   1240 		if (*given < w.w_needed)
   1241 			return ENOMEM;
   1242 	} else {
   1243 		*given = (11 * w.w_needed) / 10;
   1244 	}
   1245 	return error;
   1246 }
   1247 
   1248 /*
   1249  * Routing message software interrupt routine
   1250  */
   1251 static void
   1252 COMPATNAME(route_intr)(void *cookie)
   1253 {
   1254 	struct sockproto proto = { .sp_family = PF_XROUTE, };
   1255 	struct route_info * const ri = &COMPATNAME(route_info);
   1256 	struct mbuf *m;
   1257 	int s;
   1258 
   1259 	mutex_enter(softnet_lock);
   1260 	KERNEL_LOCK(1, NULL);
   1261 	while (!IF_IS_EMPTY(&ri->ri_intrq)) {
   1262 		s = splnet();
   1263 		IF_DEQUEUE(&ri->ri_intrq, m);
   1264 		splx(s);
   1265 		if (m == NULL)
   1266 			break;
   1267 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
   1268 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst);
   1269 	}
   1270 	KERNEL_UNLOCK_ONE(NULL);
   1271 	mutex_exit(softnet_lock);
   1272 }
   1273 
   1274 /*
   1275  * Enqueue a message to the software interrupt routine.
   1276  */
   1277 void
   1278 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
   1279 {
   1280 	struct route_info * const ri = &COMPATNAME(route_info);
   1281 	int s, wasempty;
   1282 
   1283 	s = splnet();
   1284 	if (IF_QFULL(&ri->ri_intrq)) {
   1285 		IF_DROP(&ri->ri_intrq);
   1286 		m_freem(m);
   1287 	} else {
   1288 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
   1289 		M_SETCTX(m, (uintptr_t)family);
   1290 		IF_ENQUEUE(&ri->ri_intrq, m);
   1291 		if (wasempty)
   1292 			softint_schedule(ri->ri_sih);
   1293 	}
   1294 	splx(s);
   1295 }
   1296 
   1297 static void
   1298 COMPATNAME(route_init)(void)
   1299 {
   1300 	struct route_info * const ri = &COMPATNAME(route_info);
   1301 
   1302 #ifndef COMPAT_RTSOCK
   1303 	rt_init();
   1304 #endif
   1305 
   1306 	sysctl_net_route_setup(NULL);
   1307 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
   1308 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
   1309 	    COMPATNAME(route_intr), NULL);
   1310 }
   1311 
   1312 /*
   1313  * Definitions of protocols supported in the ROUTE domain.
   1314  */
   1315 #ifndef COMPAT_RTSOCK
   1316 PR_WRAP_USRREQ(route_usrreq);
   1317 #else
   1318 PR_WRAP_USRREQ(compat_50_route_usrreq);
   1319 #endif
   1320 
   1321 static const struct protosw COMPATNAME(route_protosw)[] = {
   1322 	{
   1323 		.pr_type = SOCK_RAW,
   1324 		.pr_domain = &COMPATNAME(routedomain),
   1325 		.pr_flags = PR_ATOMIC|PR_ADDR,
   1326 		.pr_input = raw_input,
   1327 		.pr_output = COMPATNAME(route_output),
   1328 		.pr_ctlinput = raw_ctlinput,
   1329 		.pr_usrreq = COMPATNAME(route_usrreq_wrapper),
   1330 		.pr_init = raw_init,
   1331 	},
   1332 };
   1333 
   1334 struct domain COMPATNAME(routedomain) = {
   1335 	.dom_family = PF_XROUTE,
   1336 	.dom_name = DOMAINNAME,
   1337 	.dom_init = COMPATNAME(route_init),
   1338 	.dom_protosw = COMPATNAME(route_protosw),
   1339 	.dom_protoswNPROTOSW =
   1340 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
   1341 };
   1342 
   1343 static void
   1344 sysctl_net_route_setup(struct sysctllog **clog)
   1345 {
   1346 	const struct sysctlnode *rnode = NULL;
   1347 
   1348 	sysctl_createv(clog, 0, NULL, NULL,
   1349 		       CTLFLAG_PERMANENT,
   1350 		       CTLTYPE_NODE, "net", NULL,
   1351 		       NULL, 0, NULL, 0,
   1352 		       CTL_NET, CTL_EOL);
   1353 
   1354 	sysctl_createv(clog, 0, NULL, &rnode,
   1355 		       CTLFLAG_PERMANENT,
   1356 		       CTLTYPE_NODE, DOMAINNAME,
   1357 		       SYSCTL_DESCR("PF_ROUTE information"),
   1358 		       NULL, 0, NULL, 0,
   1359 		       CTL_NET, PF_XROUTE, CTL_EOL);
   1360 
   1361 	sysctl_createv(clog, 0, NULL, NULL,
   1362 		       CTLFLAG_PERMANENT,
   1363 		       CTLTYPE_NODE, "rtable",
   1364 		       SYSCTL_DESCR("Routing table information"),
   1365 		       sysctl_rtable, 0, NULL, 0,
   1366 		       CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
   1367 
   1368 	sysctl_createv(clog, 0, &rnode, NULL,
   1369 		       CTLFLAG_PERMANENT,
   1370 		       CTLTYPE_STRUCT, "stats",
   1371 		       SYSCTL_DESCR("Routing statistics"),
   1372 		       NULL, 0, &rtstat, sizeof(rtstat),
   1373 		       CTL_CREATE, CTL_EOL);
   1374 }
   1375