Home | History | Annotate | Line # | Download | only in net
rtsock.c revision 1.133
      1 /*	$NetBSD: rtsock.c,v 1.133 2011/02/01 01:39:20 matt 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 1.133 2011/02/01 01:39:20 matt 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 static void COMPATNAME(route_init)(void);
    143 static int COMPATNAME(route_output)(struct mbuf *, ...);
    144 static int COMPATNAME(route_usrreq)(struct socket *,
    145 	    int, struct mbuf *, struct mbuf *, struct mbuf *, struct lwp *);
    146 
    147 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *);
    148 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *);
    149 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int,
    150     struct rt_addrinfo *);
    151 static void rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *);
    152 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *);
    153 static void sysctl_net_route_setup(struct sysctllog **);
    154 static int sysctl_dumpentry(struct rtentry *, void *);
    155 static int sysctl_iflist(int, struct rt_walkarg *, int);
    156 static int sysctl_rtable(SYSCTLFN_PROTO);
    157 static void rt_adjustcount(int, int);
    158 
    159 static void
    160 rt_adjustcount(int af, int cnt)
    161 {
    162 	struct route_cb * const cb = &COMPATNAME(route_info).ri_cb;
    163 
    164 	cb->any_count += cnt;
    165 
    166 	switch (af) {
    167 	case AF_INET:
    168 		cb->ip_count += cnt;
    169 		return;
    170 #ifdef INET6
    171 	case AF_INET6:
    172 		cb->ip6_count += cnt;
    173 		return;
    174 #endif
    175 	case AF_ISO:
    176 		cb->iso_count += cnt;
    177 		return;
    178 	case AF_MPLS:
    179 		cb->mpls_count += cnt;
    180 		return;
    181 	}
    182 }
    183 
    184 /*ARGSUSED*/
    185 int
    186 COMPATNAME(route_usrreq)(struct socket *so, int req, struct mbuf *m, struct mbuf *nam,
    187 	struct mbuf *control, struct lwp *l)
    188 {
    189 	int error = 0;
    190 	struct rawcb *rp = sotorawcb(so);
    191 	int s;
    192 
    193 	if (req == PRU_ATTACH) {
    194 		sosetlock(so);
    195 		rp = malloc(sizeof(*rp), M_PCB, M_WAITOK|M_ZERO);
    196 		so->so_pcb = rp;
    197 	}
    198 	if (req == PRU_DETACH && rp)
    199 		rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
    200 	s = splsoftnet();
    201 
    202 	/*
    203 	 * Don't call raw_usrreq() in the attach case, because
    204 	 * we want to allow non-privileged processes to listen on
    205 	 * and send "safe" commands to the routing socket.
    206 	 */
    207 	if (req == PRU_ATTACH) {
    208 		if (l == NULL)
    209 			error = EACCES;
    210 		else
    211 			error = raw_attach(so, (int)(long)nam);
    212 	} else
    213 		error = raw_usrreq(so, req, m, nam, control, l);
    214 
    215 	rp = sotorawcb(so);
    216 	if (req == PRU_ATTACH && rp) {
    217 		if (error) {
    218 			free(rp, M_PCB);
    219 			splx(s);
    220 			return error;
    221 		}
    222 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
    223 		rp->rcb_laddr = &COMPATNAME(route_info).ri_src;
    224 		rp->rcb_faddr = &COMPATNAME(route_info).ri_dst;
    225 		soisconnected(so);
    226 		so->so_options |= SO_USELOOPBACK;
    227 	}
    228 	splx(s);
    229 	return error;
    230 }
    231 
    232 static const struct sockaddr *
    233 intern_netmask(const struct sockaddr *mask)
    234 {
    235 	struct radix_node *rn;
    236 	extern struct radix_node_head *mask_rnhead;
    237 
    238 	if (mask != NULL &&
    239 	    (rn = rn_search(mask, mask_rnhead->rnh_treetop)))
    240 		mask = (const struct sockaddr *)rn->rn_key;
    241 
    242 	return mask;
    243 }
    244 
    245 /*ARGSUSED*/
    246 int
    247 COMPATNAME(route_output)(struct mbuf *m, ...)
    248 {
    249 	struct sockproto proto = { .sp_family = PF_XROUTE, };
    250 	struct rt_xmsghdr *rtm = NULL;
    251 	struct rt_xmsghdr *old_rtm = NULL;
    252 	struct rtentry *rt = NULL;
    253 	struct rtentry *saved_nrt = NULL;
    254 	struct rt_addrinfo info;
    255 	int len, error = 0;
    256 	struct ifnet *ifp = NULL;
    257 	struct ifaddr *ifa = NULL;
    258 	struct socket *so;
    259 	va_list ap;
    260 	sa_family_t family;
    261 
    262 	va_start(ap, m);
    263 	so = va_arg(ap, struct socket *);
    264 	va_end(ap);
    265 
    266 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0)
    267 	if (m == NULL || ((m->m_len < sizeof(int32_t)) &&
    268 	   (m = m_pullup(m, sizeof(int32_t))) == NULL))
    269 		return ENOBUFS;
    270 	if ((m->m_flags & M_PKTHDR) == 0)
    271 		panic("%s", __func__);
    272 	len = m->m_pkthdr.len;
    273 	if (len < sizeof(*rtm) ||
    274 	    len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) {
    275 		info.rti_info[RTAX_DST] = NULL;
    276 		senderr(EINVAL);
    277 	}
    278 	R_Malloc(rtm, struct rt_xmsghdr *, len);
    279 	if (rtm == NULL) {
    280 		info.rti_info[RTAX_DST] = NULL;
    281 		senderr(ENOBUFS);
    282 	}
    283 	m_copydata(m, 0, len, rtm);
    284 	if (rtm->rtm_version != RTM_XVERSION) {
    285 		info.rti_info[RTAX_DST] = NULL;
    286 		senderr(EPROTONOSUPPORT);
    287 	}
    288 	rtm->rtm_pid = curproc->p_pid;
    289 	memset(&info, 0, sizeof(info));
    290 	info.rti_addrs = rtm->rtm_addrs;
    291 	if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm,
    292 	    &info)) {
    293 		senderr(EINVAL);
    294 	}
    295 	info.rti_flags = rtm->rtm_flags;
    296 #ifdef RTSOCK_DEBUG
    297 	if (info.rti_info[RTAX_DST]->sa_family == AF_INET) {
    298 		printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__,
    299 		    inet_ntoa(((const struct sockaddr_in *)
    300 		    info.rti_info[RTAX_DST])->sin_addr));
    301 	}
    302 #endif /* RTSOCK_DEBUG */
    303 	if (info.rti_info[RTAX_DST] == NULL ||
    304 	    (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) {
    305 		senderr(EINVAL);
    306 	}
    307 	if (info.rti_info[RTAX_GATEWAY] != NULL &&
    308 	    (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) {
    309 		senderr(EINVAL);
    310 	}
    311 
    312 	/*
    313 	 * Verify that the caller has the appropriate privilege; RTM_GET
    314 	 * is the only operation the non-superuser is allowed.
    315 	 */
    316 	if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_ROUTE,
    317 	    0, rtm, NULL, NULL) != 0)
    318 		senderr(EACCES);
    319 
    320 	switch (rtm->rtm_type) {
    321 
    322 	case RTM_ADD:
    323 		if (info.rti_info[RTAX_GATEWAY] == NULL) {
    324 			senderr(EINVAL);
    325 		}
    326 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
    327 		if (error == 0 && saved_nrt) {
    328 			rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt);
    329 			saved_nrt->rt_refcnt--;
    330 		}
    331 		break;
    332 
    333 	case RTM_DELETE:
    334 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
    335 		if (error == 0) {
    336 			(rt = saved_nrt)->rt_refcnt++;
    337 			goto report;
    338 		}
    339 		break;
    340 
    341 	case RTM_GET:
    342 	case RTM_CHANGE:
    343 	case RTM_LOCK:
    344                 /* XXX This will mask info.rti_info[RTAX_DST] with
    345 		 * info.rti_info[RTAX_NETMASK] before
    346                  * searching.  It did not used to do that.  --dyoung
    347 		 */
    348 		error = rtrequest1(RTM_GET, &info, &rt);
    349 		if (error != 0)
    350 			senderr(error);
    351 		if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */
    352 			struct radix_node *rn;
    353 
    354 			if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt),
    355 			    info.rti_info[RTAX_DST]->sa_len) != 0)
    356 				senderr(ESRCH);
    357 			info.rti_info[RTAX_NETMASK] = intern_netmask(
    358 			    info.rti_info[RTAX_NETMASK]);
    359 			for (rn = rt->rt_nodes; rn; rn = rn->rn_dupedkey)
    360 				if (info.rti_info[RTAX_NETMASK] ==
    361 				    (const struct sockaddr *)rn->rn_mask)
    362 					break;
    363 			if (rn == NULL)
    364 				senderr(ETOOMANYREFS);
    365 			rt = (struct rtentry *)rn;
    366 		}
    367 
    368 		switch (rtm->rtm_type) {
    369 		case RTM_GET:
    370 		report:
    371 			info.rti_info[RTAX_DST] = rt_getkey(rt);
    372 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    373 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    374 			info.rti_info[RTAX_TAG] = (struct sockaddr*)rt_gettag(rt);
    375 			if ((rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) == 0)
    376 				;
    377 			else if ((ifp = rt->rt_ifp) != NULL) {
    378 				const struct ifaddr *rtifa;
    379 				info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
    380                                 /* rtifa used to be simply rt->rt_ifa.
    381                                  * If rt->rt_ifa != NULL, then
    382                                  * rt_get_ifa() != NULL.  So this
    383                                  * ought to still be safe. --dyoung
    384 				 */
    385 				rtifa = rt_get_ifa(rt);
    386 				info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
    387 #ifdef RTSOCK_DEBUG
    388 				if (info.rti_info[RTAX_IFA]->sa_family ==
    389 				    AF_INET) {
    390 					printf("%s: copying out RTAX_IFA %s ",
    391 					    __func__, inet_ntoa(
    392 					    ((const struct sockaddr_in *)
    393 					    info.rti_info[RTAX_IFA])->sin_addr)
    394 					    );
    395 					printf("for info.rti_info[RTAX_DST] %s "
    396 					    "ifa_getifa %p ifa_seqno %p\n",
    397 					    inet_ntoa(
    398 					    ((const struct sockaddr_in *)
    399 					    info.rti_info[RTAX_DST])->sin_addr),
    400 					    (void *)rtifa->ifa_getifa,
    401 					    rtifa->ifa_seqno);
    402 				}
    403 #endif /* RTSOCK_DEBUG */
    404 				if (ifp->if_flags & IFF_POINTOPOINT) {
    405 					info.rti_info[RTAX_BRD] =
    406 					    rtifa->ifa_dstaddr;
    407 				} else
    408 					info.rti_info[RTAX_BRD] = NULL;
    409 				rtm->rtm_index = ifp->if_index;
    410 			} else {
    411 				info.rti_info[RTAX_IFP] = NULL;
    412 				info.rti_info[RTAX_IFA] = NULL;
    413 			}
    414 			(void)rt_msg2(rtm->rtm_type, &info, NULL, NULL, &len);
    415 			if (len > rtm->rtm_msglen) {
    416 				old_rtm = rtm;
    417 				R_Malloc(rtm, struct rt_xmsghdr *, len);
    418 				if (rtm == NULL)
    419 					senderr(ENOBUFS);
    420 				(void)memcpy(rtm, old_rtm, old_rtm->rtm_msglen);
    421 			}
    422 			(void)rt_msg2(rtm->rtm_type, &info, rtm, NULL, 0);
    423 			rtm->rtm_flags = rt->rt_flags;
    424 			rtm_setmetrics(rt, rtm);
    425 			rtm->rtm_addrs = info.rti_addrs;
    426 			break;
    427 
    428 		case RTM_CHANGE:
    429 			/*
    430 			 * new gateway could require new ifaddr, ifp;
    431 			 * flags may also be different; ifp may be specified
    432 			 * by ll sockaddr when protocol address is ambiguous
    433 			 */
    434 			if ((error = rt_getifa(&info)) != 0)
    435 				senderr(error);
    436 			if (info.rti_info[RTAX_GATEWAY] &&
    437 			    rt_setgate(rt, info.rti_info[RTAX_GATEWAY]))
    438 				senderr(EDQUOT);
    439 			if (info.rti_info[RTAX_TAG])
    440 				rt_settag(rt, info.rti_info[RTAX_TAG]);
    441 			/* new gateway could require new ifaddr, ifp;
    442 			   flags may also be different; ifp may be specified
    443 			   by ll sockaddr when protocol address is ambiguous */
    444 			if (info.rti_info[RTAX_IFP] &&
    445 			    (ifa = ifa_ifwithnet(info.rti_info[RTAX_IFP])) &&
    446 			    (ifp = ifa->ifa_ifp) && (info.rti_info[RTAX_IFA] ||
    447 			    info.rti_info[RTAX_GATEWAY])) {
    448 				if (info.rti_info[RTAX_IFA] == NULL ||
    449 				    (ifa = ifa_ifwithaddr(
    450 				    info.rti_info[RTAX_IFA])) == NULL)
    451 					ifa = ifaof_ifpforaddr(
    452 					    info.rti_info[RTAX_IFA] ?
    453 					    info.rti_info[RTAX_IFA] :
    454 					    info.rti_info[RTAX_GATEWAY], ifp);
    455 			} else if ((info.rti_info[RTAX_IFA] &&
    456 			    (ifa = ifa_ifwithaddr(info.rti_info[RTAX_IFA]))) ||
    457 			    (info.rti_info[RTAX_GATEWAY] &&
    458 			    (ifa = ifa_ifwithroute(rt->rt_flags,
    459 			    rt_getkey(rt), info.rti_info[RTAX_GATEWAY])))) {
    460 				ifp = ifa->ifa_ifp;
    461 			}
    462 			if (ifa) {
    463 				struct ifaddr *oifa = rt->rt_ifa;
    464 				if (oifa != ifa) {
    465 					if (oifa && oifa->ifa_rtrequest) {
    466 						oifa->ifa_rtrequest(RTM_DELETE,
    467 						    rt, &info);
    468 					}
    469 					rt_replace_ifa(rt, ifa);
    470 					rt->rt_ifp = ifp;
    471 				}
    472 			}
    473 			if (ifp && rt->rt_ifp != ifp)
    474 				rt->rt_ifp = ifp;
    475 			rt_setmetrics(rtm->rtm_inits, rtm, rt);
    476 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
    477 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
    478 			/*FALLTHROUGH*/
    479 		case RTM_LOCK:
    480 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
    481 			rt->rt_rmx.rmx_locks |=
    482 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
    483 			break;
    484 		}
    485 		break;
    486 
    487 	default:
    488 		senderr(EOPNOTSUPP);
    489 	}
    490 
    491 flush:
    492 	if (rtm) {
    493 		if (error)
    494 			rtm->rtm_errno = error;
    495 		else
    496 			rtm->rtm_flags |= RTF_DONE;
    497 	}
    498 	family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family :
    499 	    0;
    500 	/* We cannot free old_rtm until we have stopped using the
    501 	 * pointers in info, some of which may point to sockaddrs
    502 	 * in old_rtm.
    503 	 */
    504 	if (old_rtm != NULL)
    505 		Free(old_rtm);
    506 	if (rt)
    507 		rtfree(rt);
    508     {
    509 	struct rawcb *rp = NULL;
    510 	/*
    511 	 * Check to see if we don't want our own messages.
    512 	 */
    513 	if ((so->so_options & SO_USELOOPBACK) == 0) {
    514 		if (COMPATNAME(route_info).ri_cb.any_count <= 1) {
    515 			if (rtm)
    516 				Free(rtm);
    517 			m_freem(m);
    518 			return error;
    519 		}
    520 		/* There is another listener, so construct message */
    521 		rp = sotorawcb(so);
    522 	}
    523 	if (rtm) {
    524 		m_copyback(m, 0, rtm->rtm_msglen, rtm);
    525 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
    526 			m_freem(m);
    527 			m = NULL;
    528 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
    529 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
    530 		Free(rtm);
    531 	}
    532 	if (rp)
    533 		rp->rcb_proto.sp_family = 0; /* Avoid us */
    534 	if (family)
    535 		proto.sp_protocol = family;
    536 	if (m)
    537 		raw_input(m, &proto, &COMPATNAME(route_info).ri_src,
    538 		    &COMPATNAME(route_info).ri_dst);
    539 	if (rp)
    540 		rp->rcb_proto.sp_family = PF_XROUTE;
    541     }
    542 	return error;
    543 }
    544 
    545 static void
    546 rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out)
    547 {
    548 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e;
    549 	metric(RTV_RPIPE, rmx_recvpipe);
    550 	metric(RTV_SPIPE, rmx_sendpipe);
    551 	metric(RTV_SSTHRESH, rmx_ssthresh);
    552 	metric(RTV_RTT, rmx_rtt);
    553 	metric(RTV_RTTVAR, rmx_rttvar);
    554 	metric(RTV_HOPCOUNT, rmx_hopcount);
    555 	metric(RTV_MTU, rmx_mtu);
    556 	metric(RTV_EXPIRE, rmx_expire);
    557 #undef metric
    558 }
    559 
    560 static void
    561 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out)
    562 {
    563 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e;
    564 	metric(rmx_recvpipe);
    565 	metric(rmx_sendpipe);
    566 	metric(rmx_ssthresh);
    567 	metric(rmx_rtt);
    568 	metric(rmx_rttvar);
    569 	metric(rmx_hopcount);
    570 	metric(rmx_mtu);
    571 	metric(rmx_expire);
    572 #undef metric
    573 }
    574 
    575 static int
    576 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim,
    577     struct rt_addrinfo *rtinfo)
    578 {
    579 	const struct sockaddr *sa = NULL;	/* Quell compiler warning */
    580 	int i;
    581 
    582 	for (i = 0; i < RTAX_MAX && cp < cplim; i++) {
    583 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
    584 			continue;
    585 		rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp;
    586 		RT_XADVANCE(cp, sa);
    587 	}
    588 
    589 	/*
    590 	 * Check for extra addresses specified, except RTM_GET asking
    591 	 * for interface info.
    592 	 */
    593 	if (rtmtype == RTM_GET) {
    594 		if (((rtinfo->rti_addrs &
    595 		    (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0 << i)) != 0)
    596 			return 1;
    597 	} else if ((rtinfo->rti_addrs & (~0 << i)) != 0)
    598 		return 1;
    599 	/* Check for bad data length.  */
    600 	if (cp != cplim) {
    601 		if (i == RTAX_NETMASK + 1 && sa != NULL &&
    602 		    cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim)
    603 			/*
    604 			 * The last sockaddr was info.rti_info[RTAX_NETMASK].
    605 			 * We accept this for now for the sake of old
    606 			 * binaries or third party softwares.
    607 			 */
    608 			;
    609 		else
    610 			return 1;
    611 	}
    612 	return 0;
    613 }
    614 
    615 static int
    616 rt_getlen(int type)
    617 {
    618 #ifndef COMPAT_RTSOCK
    619 	CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t));
    620 	CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t));
    621 	CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t));
    622 	CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t));
    623 #endif
    624 
    625 	switch (type) {
    626 	case RTM_DELADDR:
    627 	case RTM_NEWADDR:
    628 	case RTM_CHGADDR:
    629 		return sizeof(struct ifa_xmsghdr);
    630 
    631 	case RTM_OOIFINFO:
    632 #ifdef COMPAT_14
    633 		return sizeof(struct if_msghdr14);
    634 #else
    635 #ifdef DIAGNOSTIC
    636 		printf("RTM_OOIFINFO\n");
    637 #endif
    638 		return -1;
    639 #endif
    640 	case RTM_OIFINFO:
    641 #ifdef COMPAT_50
    642 		return sizeof(struct if_msghdr50);
    643 #else
    644 #ifdef DIAGNOSTIC
    645 		printf("RTM_OIFINFO\n");
    646 #endif
    647 		return -1;
    648 #endif
    649 
    650 	case RTM_IFINFO:
    651 		return sizeof(struct if_xmsghdr);
    652 
    653 	case RTM_IFANNOUNCE:
    654 	case RTM_IEEE80211:
    655 		return sizeof(struct if_xannouncemsghdr);
    656 
    657 	default:
    658 		return sizeof(struct rt_xmsghdr);
    659 	}
    660 }
    661 
    662 
    663 struct mbuf *
    664 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen)
    665 {
    666 	struct rt_xmsghdr *rtm;
    667 	struct mbuf *m;
    668 	int i;
    669 	const struct sockaddr *sa;
    670 	int len, dlen;
    671 
    672 	m = m_gethdr(M_DONTWAIT, MT_DATA);
    673 	if (m == NULL)
    674 		return m;
    675 	MCLAIM(m, &COMPATNAME(routedomain).dom_mowner);
    676 
    677 	if ((len = rt_getlen(type)) == -1)
    678 		goto out;
    679 	if (len > MHLEN + MLEN)
    680 		panic("%s: message too long", __func__);
    681 	else if (len > MHLEN) {
    682 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
    683 		if (m->m_next == NULL)
    684 			goto out;
    685 		MCLAIM(m->m_next, m->m_owner);
    686 		m->m_pkthdr.len = len;
    687 		m->m_len = MHLEN;
    688 		m->m_next->m_len = len - MHLEN;
    689 	} else {
    690 		m->m_pkthdr.len = m->m_len = len;
    691 	}
    692 	m->m_pkthdr.rcvif = NULL;
    693 	m_copyback(m, 0, datalen, data);
    694 	if (len > datalen)
    695 		(void)memset(mtod(m, char *) + datalen, 0, len - datalen);
    696 	rtm = mtod(m, struct rt_xmsghdr *);
    697 	for (i = 0; i < RTAX_MAX; i++) {
    698 		if ((sa = rtinfo->rti_info[i]) == NULL)
    699 			continue;
    700 		rtinfo->rti_addrs |= (1 << i);
    701 		dlen = RT_XROUNDUP(sa->sa_len);
    702 		m_copyback(m, len, sa->sa_len, sa);
    703 		if (dlen != sa->sa_len) {
    704 			m_copyback(m, len + sa->sa_len,
    705 			    dlen - sa->sa_len, "\0\0\0\0\0\0");
    706 		}
    707 		len += dlen;
    708 	}
    709 	if (m->m_pkthdr.len != len)
    710 		goto out;
    711 	rtm->rtm_msglen = len;
    712 	rtm->rtm_version = RTM_XVERSION;
    713 	rtm->rtm_type = type;
    714 	return m;
    715 out:
    716 	m_freem(m);
    717 	return NULL;
    718 }
    719 
    720 /*
    721  * rt_msg2
    722  *
    723  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
    724  *		returns the length of the message in 'lenp'.
    725  *
    726  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
    727  *	the message
    728  * otherwise walkarg's w_needed is updated and if the user buffer is
    729  *	specified and w_needed indicates space exists the information is copied
    730  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
    731  *	if the allocation fails ENOBUFS is returned.
    732  */
    733 static int
    734 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w,
    735 	int *lenp)
    736 {
    737 	int i;
    738 	int len, dlen, second_time = 0;
    739 	char *cp0, *cp = cpv;
    740 
    741 	rtinfo->rti_addrs = 0;
    742 again:
    743 	if ((len = rt_getlen(type)) == -1)
    744 		return EINVAL;
    745 
    746 	if ((cp0 = cp) != NULL)
    747 		cp += len;
    748 	for (i = 0; i < RTAX_MAX; i++) {
    749 		const struct sockaddr *sa;
    750 
    751 		if ((sa = rtinfo->rti_info[i]) == NULL)
    752 			continue;
    753 		rtinfo->rti_addrs |= (1 << i);
    754 		dlen = RT_XROUNDUP(sa->sa_len);
    755 		if (cp) {
    756 			(void)memcpy(cp, sa, (size_t)dlen);
    757 			cp += dlen;
    758 		}
    759 		len += dlen;
    760 	}
    761 	if (cp == NULL && w != NULL && !second_time) {
    762 		struct rt_walkarg *rw = w;
    763 
    764 		rw->w_needed += len;
    765 		if (rw->w_needed <= 0 && rw->w_where) {
    766 			if (rw->w_tmemsize < len) {
    767 				if (rw->w_tmem)
    768 					free(rw->w_tmem, M_RTABLE);
    769 				rw->w_tmem = malloc(len, M_RTABLE, M_NOWAIT);
    770 				if (rw->w_tmem)
    771 					rw->w_tmemsize = len;
    772 				else
    773 					rw->w_tmemsize = 0;
    774 			}
    775 			if (rw->w_tmem) {
    776 				cp = rw->w_tmem;
    777 				second_time = 1;
    778 				goto again;
    779 			} else {
    780 				rw->w_tmemneeded = len;
    781 				return ENOBUFS;
    782 			}
    783 		}
    784 	}
    785 	if (cp) {
    786 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0;
    787 
    788 		rtm->rtm_version = RTM_XVERSION;
    789 		rtm->rtm_type = type;
    790 		rtm->rtm_msglen = len;
    791 	}
    792 	if (lenp)
    793 		*lenp = len;
    794 	return 0;
    795 }
    796 
    797 /*
    798  * This routine is called to generate a message from the routing
    799  * socket indicating that a redirect has occurred, a routing lookup
    800  * has failed, or that a protocol has detected timeouts to a particular
    801  * destination.
    802  */
    803 void
    804 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags,
    805     int error)
    806 {
    807 	struct rt_xmsghdr rtm;
    808 	struct mbuf *m;
    809 	const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
    810 	struct rt_addrinfo info = *rtinfo;
    811 
    812 	COMPATCALL(rt_missmsg, (type, rtinfo, flags, error));
    813 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    814 		return;
    815 	memset(&rtm, 0, sizeof(rtm));
    816 	rtm.rtm_flags = RTF_DONE | flags;
    817 	rtm.rtm_errno = error;
    818 	m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm));
    819 	if (m == NULL)
    820 		return;
    821 	mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
    822 	COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
    823 }
    824 
    825 /*
    826  * This routine is called to generate a message from the routing
    827  * socket indicating that the status of a network interface has changed.
    828  */
    829 void
    830 COMPATNAME(rt_ifmsg)(struct ifnet *ifp)
    831 {
    832 	struct if_xmsghdr ifm;
    833 	struct mbuf *m;
    834 	struct rt_addrinfo info;
    835 
    836 	COMPATCALL(rt_ifmsg, (ifp));
    837 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    838 		return;
    839 	(void)memset(&info, 0, sizeof(info));
    840 	(void)memset(&ifm, 0, sizeof(ifm));
    841 	ifm.ifm_index = ifp->if_index;
    842 	ifm.ifm_flags = ifp->if_flags;
    843 	ifm.ifm_data = ifp->if_data;
    844 	ifm.ifm_addrs = 0;
    845 	m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm));
    846 	if (m == NULL)
    847 		return;
    848 	COMPATNAME(route_enqueue)(m, 0);
    849 #ifdef COMPAT_14
    850 	compat_14_rt_oifmsg(ifp);
    851 #endif
    852 #ifdef COMPAT_50
    853 	compat_50_rt_oifmsg(ifp);
    854 #endif
    855 }
    856 
    857 
    858 /*
    859  * This is called to generate messages from the routing socket
    860  * indicating a network interface has had addresses associated with it.
    861  * if we ever reverse the logic and replace messages TO the routing
    862  * socket indicate a request to configure interfaces, then it will
    863  * be unnecessary as the routing socket will automatically generate
    864  * copies of it.
    865  */
    866 void
    867 COMPATNAME(rt_newaddrmsg)(int cmd, struct ifaddr *ifa, int error,
    868     struct rtentry *rt)
    869 {
    870 #define	cmdpass(__cmd, __pass)	(((__cmd) << 2) | (__pass))
    871 	struct rt_addrinfo info;
    872 	const struct sockaddr *sa;
    873 	int pass;
    874 	struct mbuf *m;
    875 	struct ifnet *ifp = ifa->ifa_ifp;
    876 	struct rt_xmsghdr rtm;
    877 	struct ifa_xmsghdr ifam;
    878 	int ncmd;
    879 
    880 	COMPATCALL(rt_newaddrmsg, (cmd, ifa, error, rt));
    881 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    882 		return;
    883 	for (pass = 1; pass < 3; pass++) {
    884 		memset(&info, 0, sizeof(info));
    885 		switch (cmdpass(cmd, pass)) {
    886 		case cmdpass(RTM_ADD, 1):
    887 		case cmdpass(RTM_CHANGE, 1):
    888 		case cmdpass(RTM_DELETE, 2):
    889 			switch (cmd) {
    890 			case RTM_DELETE:
    891 				ncmd = RTM_DELADDR;
    892 				break;
    893 			case RTM_CHANGE:
    894 				ncmd = RTM_CHGADDR;
    895 				break;
    896 			default:
    897 				ncmd = RTM_NEWADDR;
    898 			}
    899 			info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr;
    900 			info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
    901 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
    902 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
    903 			memset(&ifam, 0, sizeof(ifam));
    904 			ifam.ifam_index = ifp->if_index;
    905 			ifam.ifam_metric = ifa->ifa_metric;
    906 			ifam.ifam_flags = ifa->ifa_flags;
    907 			m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam));
    908 			if (m == NULL)
    909 				continue;
    910 			mtod(m, struct ifa_xmsghdr *)->ifam_addrs =
    911 			    info.rti_addrs;
    912 			break;
    913 		case cmdpass(RTM_ADD, 2):
    914 		case cmdpass(RTM_CHANGE, 2):
    915 		case cmdpass(RTM_DELETE, 1):
    916 			if (rt == NULL)
    917 				continue;
    918 			info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    919 			info.rti_info[RTAX_DST] = sa = rt_getkey(rt);
    920 			info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    921 			memset(&rtm, 0, sizeof(rtm));
    922 			rtm.rtm_index = ifp->if_index;
    923 			rtm.rtm_flags |= rt->rt_flags;
    924 			rtm.rtm_errno = error;
    925 			m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm));
    926 			if (m == NULL)
    927 				continue;
    928 			mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs;
    929 			break;
    930 		default:
    931 			continue;
    932 		}
    933 #ifdef DIAGNOSTIC
    934 		if (m == NULL)
    935 			panic("%s: called with wrong command", __func__);
    936 #endif
    937 		COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0);
    938 	}
    939 #undef cmdpass
    940 }
    941 
    942 static struct mbuf *
    943 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what,
    944     struct rt_addrinfo *info)
    945 {
    946 	struct if_xannouncemsghdr ifan;
    947 
    948 	memset(info, 0, sizeof(*info));
    949 	memset(&ifan, 0, sizeof(ifan));
    950 	ifan.ifan_index = ifp->if_index;
    951 	strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name));
    952 	ifan.ifan_what = what;
    953 	return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan));
    954 }
    955 
    956 /*
    957  * This is called to generate routing socket messages indicating
    958  * network interface arrival and departure.
    959  */
    960 void
    961 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what)
    962 {
    963 	struct mbuf *m;
    964 	struct rt_addrinfo info;
    965 
    966 	COMPATCALL(rt_ifannouncemsg, (ifp, what));
    967 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    968 		return;
    969 	m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info);
    970 	if (m == NULL)
    971 		return;
    972 	COMPATNAME(route_enqueue)(m, 0);
    973 }
    974 
    975 /*
    976  * This is called to generate routing socket messages indicating
    977  * IEEE80211 wireless events.
    978  * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way.
    979  */
    980 void
    981 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data,
    982 	size_t data_len)
    983 {
    984 	struct mbuf *m;
    985 	struct rt_addrinfo info;
    986 
    987 	COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len));
    988 	if (COMPATNAME(route_info).ri_cb.any_count == 0)
    989 		return;
    990 	m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info);
    991 	if (m == NULL)
    992 		return;
    993 	/*
    994 	 * Append the ieee80211 data.  Try to stick it in the
    995 	 * mbuf containing the ifannounce msg; otherwise allocate
    996 	 * a new mbuf and append.
    997 	 *
    998 	 * NB: we assume m is a single mbuf.
    999 	 */
   1000 	if (data_len > M_TRAILINGSPACE(m)) {
   1001 		struct mbuf *n = m_get(M_NOWAIT, MT_DATA);
   1002 		if (n == NULL) {
   1003 			m_freem(m);
   1004 			return;
   1005 		}
   1006 		(void)memcpy(mtod(n, void *), data, data_len);
   1007 		n->m_len = data_len;
   1008 		m->m_next = n;
   1009 	} else if (data_len > 0) {
   1010 		(void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len);
   1011 		m->m_len += data_len;
   1012 	}
   1013 	if (m->m_flags & M_PKTHDR)
   1014 		m->m_pkthdr.len += data_len;
   1015 	mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len;
   1016 	COMPATNAME(route_enqueue)(m, 0);
   1017 }
   1018 
   1019 /*
   1020  * This is used in dumping the kernel table via sysctl().
   1021  */
   1022 static int
   1023 sysctl_dumpentry(struct rtentry *rt, void *v)
   1024 {
   1025 	struct rt_walkarg *w = v;
   1026 	int error = 0, size;
   1027 	struct rt_addrinfo info;
   1028 
   1029 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
   1030 		return 0;
   1031 	memset(&info, 0, sizeof(info));
   1032 	info.rti_info[RTAX_DST] = rt_getkey(rt);
   1033 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
   1034 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
   1035 	if (rt->rt_ifp) {
   1036 		const struct ifaddr *rtifa;
   1037 		info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
   1038 		/* rtifa used to be simply rt->rt_ifa.  If rt->rt_ifa != NULL,
   1039 		 * then rt_get_ifa() != NULL.  So this ought to still be safe.
   1040 		 * --dyoung
   1041 		 */
   1042 		rtifa = rt_get_ifa(rt);
   1043 		info.rti_info[RTAX_IFA] = rtifa->ifa_addr;
   1044 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
   1045 			info.rti_info[RTAX_BRD] = rtifa->ifa_dstaddr;
   1046 	}
   1047 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
   1048 		return error;
   1049 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1050 		struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)w->w_tmem;
   1051 
   1052 		rtm->rtm_flags = rt->rt_flags;
   1053 		rtm->rtm_use = rt->rt_use;
   1054 		rtm_setmetrics(rt, rtm);
   1055 		KASSERT(rt->rt_ifp != NULL);
   1056 		rtm->rtm_index = rt->rt_ifp->if_index;
   1057 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
   1058 		rtm->rtm_addrs = info.rti_addrs;
   1059 		if ((error = copyout(rtm, w->w_where, size)) != 0)
   1060 			w->w_where = NULL;
   1061 		else
   1062 			w->w_where = (char *)w->w_where + size;
   1063 	}
   1064 	return error;
   1065 }
   1066 
   1067 static int
   1068 sysctl_iflist(int af, struct rt_walkarg *w, int type)
   1069 {
   1070 	struct ifnet *ifp;
   1071 	struct ifaddr *ifa;
   1072 	struct	rt_addrinfo info;
   1073 	int	len, error = 0;
   1074 
   1075 	memset(&info, 0, sizeof(info));
   1076 	IFNET_FOREACH(ifp) {
   1077 		if (w->w_arg && w->w_arg != ifp->if_index)
   1078 			continue;
   1079 		if (IFADDR_EMPTY(ifp))
   1080 			continue;
   1081 		info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr;
   1082 		switch (type) {
   1083 		case NET_RT_IFLIST:
   1084 			error = rt_msg2(RTM_IFINFO, &info, NULL, w, &len);
   1085 			break;
   1086 #ifdef COMPAT_14
   1087 		case NET_RT_OOIFLIST:
   1088 			error = rt_msg2(RTM_OOIFINFO, &info, NULL, w, &len);
   1089 			break;
   1090 #endif
   1091 #ifdef COMPAT_50
   1092 		case NET_RT_OIFLIST:
   1093 			error = rt_msg2(RTM_OIFINFO, &info, NULL, w, &len);
   1094 			break;
   1095 #endif
   1096 		default:
   1097 			panic("sysctl_iflist(1)");
   1098 		}
   1099 		if (error)
   1100 			return error;
   1101 		info.rti_info[RTAX_IFP] = NULL;
   1102 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1103 			switch (type) {
   1104 			case NET_RT_IFLIST: {
   1105 				struct if_xmsghdr *ifm;
   1106 
   1107 				ifm = (struct if_xmsghdr *)w->w_tmem;
   1108 				ifm->ifm_index = ifp->if_index;
   1109 				ifm->ifm_flags = ifp->if_flags;
   1110 				ifm->ifm_data = ifp->if_data;
   1111 				ifm->ifm_addrs = info.rti_addrs;
   1112 				error = copyout(ifm, w->w_where, len);
   1113 				if (error)
   1114 					return error;
   1115 				w->w_where = (char *)w->w_where + len;
   1116 				break;
   1117 			}
   1118 
   1119 #ifdef COMPAT_14
   1120 			case NET_RT_OOIFLIST:
   1121 				error = compat_14_iflist(ifp, w, &info, len);
   1122 				if (error)
   1123 					return error;
   1124 				break;
   1125 #endif
   1126 #ifdef COMPAT_50
   1127 			case NET_RT_OIFLIST:
   1128 				error = compat_50_iflist(ifp, w, &info, len);
   1129 				if (error)
   1130 					return error;
   1131 				break;
   1132 #endif
   1133 			default:
   1134 				panic("sysctl_iflist(2)");
   1135 			}
   1136 		}
   1137 		IFADDR_FOREACH(ifa, ifp) {
   1138 			if (af && af != ifa->ifa_addr->sa_family)
   1139 				continue;
   1140 			info.rti_info[RTAX_IFA] = ifa->ifa_addr;
   1141 			info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
   1142 			info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr;
   1143 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
   1144 				return error;
   1145 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1146 				struct ifa_xmsghdr *ifam;
   1147 
   1148 				ifam = (struct ifa_xmsghdr *)w->w_tmem;
   1149 				ifam->ifam_index = ifa->ifa_ifp->if_index;
   1150 				ifam->ifam_flags = ifa->ifa_flags;
   1151 				ifam->ifam_metric = ifa->ifa_metric;
   1152 				ifam->ifam_addrs = info.rti_addrs;
   1153 				error = copyout(w->w_tmem, w->w_where, len);
   1154 				if (error)
   1155 					return error;
   1156 				w->w_where = (char *)w->w_where + len;
   1157 			}
   1158 		}
   1159 		info.rti_info[RTAX_IFA] = info.rti_info[RTAX_NETMASK] =
   1160 		    info.rti_info[RTAX_BRD] = NULL;
   1161 	}
   1162 	return 0;
   1163 }
   1164 
   1165 static int
   1166 sysctl_rtable(SYSCTLFN_ARGS)
   1167 {
   1168 	void 	*where = oldp;
   1169 	size_t	*given = oldlenp;
   1170 	const void *new = newp;
   1171 	int	i, s, error = EINVAL;
   1172 	u_char  af;
   1173 	struct	rt_walkarg w;
   1174 
   1175 	if (namelen == 1 && name[0] == CTL_QUERY)
   1176 		return sysctl_query(SYSCTLFN_CALL(rnode));
   1177 
   1178 	if (new)
   1179 		return EPERM;
   1180 	if (namelen != 3)
   1181 		return EINVAL;
   1182 	af = name[0];
   1183 	w.w_tmemneeded = 0;
   1184 	w.w_tmemsize = 0;
   1185 	w.w_tmem = NULL;
   1186 again:
   1187 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
   1188 	if (w.w_tmemneeded) {
   1189 		w.w_tmem = malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
   1190 		w.w_tmemsize = w.w_tmemneeded;
   1191 		w.w_tmemneeded = 0;
   1192 	}
   1193 	w.w_op = name[1];
   1194 	w.w_arg = name[2];
   1195 	w.w_given = *given;
   1196 	w.w_needed = 0 - w.w_given;
   1197 	w.w_where = where;
   1198 
   1199 	s = splsoftnet();
   1200 	switch (w.w_op) {
   1201 
   1202 	case NET_RT_DUMP:
   1203 	case NET_RT_FLAGS:
   1204 		for (i = 1; i <= AF_MAX; i++)
   1205 			if ((af == 0 || af == i) &&
   1206 			    (error = rt_walktree(i, sysctl_dumpentry, &w)))
   1207 				break;
   1208 		break;
   1209 
   1210 #ifdef COMPAT_14
   1211 	case NET_RT_OOIFLIST:
   1212 		error = sysctl_iflist(af, &w, w.w_op);
   1213 		break;
   1214 #endif
   1215 #ifdef COMPAT_50
   1216 	case NET_RT_OIFLIST:
   1217 		error = sysctl_iflist(af, &w, w.w_op);
   1218 		break;
   1219 #endif
   1220 	case NET_RT_IFLIST:
   1221 		error = sysctl_iflist(af, &w, w.w_op);
   1222 		break;
   1223 	}
   1224 	splx(s);
   1225 
   1226 	/* check to see if we couldn't allocate memory with NOWAIT */
   1227 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
   1228 		goto again;
   1229 
   1230 	if (w.w_tmem)
   1231 		free(w.w_tmem, M_RTABLE);
   1232 	w.w_needed += w.w_given;
   1233 	if (where) {
   1234 		*given = (char *)w.w_where - (char *)where;
   1235 		if (*given < w.w_needed)
   1236 			return ENOMEM;
   1237 	} else {
   1238 		*given = (11 * w.w_needed) / 10;
   1239 	}
   1240 	return error;
   1241 }
   1242 
   1243 /*
   1244  * Routing message software interrupt routine
   1245  */
   1246 static void
   1247 COMPATNAME(route_intr)(void *cookie)
   1248 {
   1249 	struct sockproto proto = { .sp_family = PF_XROUTE, };
   1250 	struct route_info * const ri = &COMPATNAME(route_info);
   1251 	struct mbuf *m;
   1252 	int s;
   1253 
   1254 	mutex_enter(softnet_lock);
   1255 	KERNEL_LOCK(1, NULL);
   1256 	while (!IF_IS_EMPTY(&ri->ri_intrq)) {
   1257 		s = splnet();
   1258 		IF_DEQUEUE(&ri->ri_intrq, m);
   1259 		splx(s);
   1260 		if (m == NULL)
   1261 			break;
   1262 		proto.sp_protocol = M_GETCTX(m, uintptr_t);
   1263 		raw_input(m, &proto, &ri->ri_src, &ri->ri_dst);
   1264 	}
   1265 	KERNEL_UNLOCK_ONE(NULL);
   1266 	mutex_exit(softnet_lock);
   1267 }
   1268 
   1269 /*
   1270  * Enqueue a message to the software interrupt routine.
   1271  */
   1272 void
   1273 COMPATNAME(route_enqueue)(struct mbuf *m, int family)
   1274 {
   1275 	struct route_info * const ri = &COMPATNAME(route_info);
   1276 	int s, wasempty;
   1277 
   1278 	s = splnet();
   1279 	if (IF_QFULL(&ri->ri_intrq)) {
   1280 		IF_DROP(&ri->ri_intrq);
   1281 		m_freem(m);
   1282 	} else {
   1283 		wasempty = IF_IS_EMPTY(&ri->ri_intrq);
   1284 		M_SETCTX(m, (uintptr_t)family);
   1285 		IF_ENQUEUE(&ri->ri_intrq, m);
   1286 		if (wasempty)
   1287 			softint_schedule(ri->ri_sih);
   1288 	}
   1289 	splx(s);
   1290 }
   1291 
   1292 static void
   1293 COMPATNAME(route_init)(void)
   1294 {
   1295 	struct route_info * const ri = &COMPATNAME(route_info);
   1296 
   1297 #ifndef COMPAT_RTSOCK
   1298 	rt_init();
   1299 #endif
   1300 
   1301 	sysctl_net_route_setup(NULL);
   1302 	ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen;
   1303 	ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE,
   1304 	    COMPATNAME(route_intr), NULL);
   1305 }
   1306 
   1307 /*
   1308  * Definitions of protocols supported in the ROUTE domain.
   1309  */
   1310 #ifndef COMPAT_RTSOCK
   1311 PR_WRAP_USRREQ(route_usrreq);
   1312 #else
   1313 PR_WRAP_USRREQ(compat_50_route_usrreq);
   1314 #endif
   1315 
   1316 static const struct protosw COMPATNAME(route_protosw)[] = {
   1317 	{
   1318 		.pr_type = SOCK_RAW,
   1319 		.pr_domain = &COMPATNAME(routedomain),
   1320 		.pr_flags = PR_ATOMIC|PR_ADDR,
   1321 		.pr_input = raw_input,
   1322 		.pr_output = COMPATNAME(route_output),
   1323 		.pr_ctlinput = raw_ctlinput,
   1324 		.pr_usrreq = COMPATNAME(route_usrreq_wrapper),
   1325 		.pr_init = raw_init,
   1326 	},
   1327 };
   1328 
   1329 struct domain COMPATNAME(routedomain) = {
   1330 	.dom_family = PF_XROUTE,
   1331 	.dom_name = DOMAINNAME,
   1332 	.dom_init = COMPATNAME(route_init),
   1333 	.dom_protosw = COMPATNAME(route_protosw),
   1334 	.dom_protoswNPROTOSW =
   1335 	    &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))],
   1336 };
   1337 
   1338 static void
   1339 sysctl_net_route_setup(struct sysctllog **clog)
   1340 {
   1341 	const struct sysctlnode *rnode = NULL;
   1342 
   1343 	sysctl_createv(clog, 0, NULL, NULL,
   1344 		       CTLFLAG_PERMANENT,
   1345 		       CTLTYPE_NODE, "net", NULL,
   1346 		       NULL, 0, NULL, 0,
   1347 		       CTL_NET, CTL_EOL);
   1348 
   1349 	sysctl_createv(clog, 0, NULL, &rnode,
   1350 		       CTLFLAG_PERMANENT,
   1351 		       CTLTYPE_NODE, DOMAINNAME,
   1352 		       SYSCTL_DESCR("PF_ROUTE information"),
   1353 		       NULL, 0, NULL, 0,
   1354 		       CTL_NET, PF_XROUTE, CTL_EOL);
   1355 
   1356 	sysctl_createv(clog, 0, NULL, NULL,
   1357 		       CTLFLAG_PERMANENT,
   1358 		       CTLTYPE_NODE, "rtable",
   1359 		       SYSCTL_DESCR("Routing table information"),
   1360 		       sysctl_rtable, 0, NULL, 0,
   1361 		       CTL_NET, PF_XROUTE, 0 /* any protocol */, CTL_EOL);
   1362 
   1363 	sysctl_createv(clog, 0, &rnode, NULL,
   1364 		       CTLFLAG_PERMANENT,
   1365 		       CTLTYPE_STRUCT, "stats",
   1366 		       SYSCTL_DESCR("Routing statistics"),
   1367 		       NULL, 0, &rtstat, sizeof(rtstat),
   1368 		       CTL_CREATE, CTL_EOL);
   1369 }
   1370