Home | History | Annotate | Line # | Download | only in net
rtsock.c revision 1.45.2.2
      1 /*	$NetBSD: rtsock.c,v 1.45.2.2 2001/06/21 20:08:20 nathanw 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. All advertising materials mentioning features or use of this software
     45  *    must display the following acknowledgement:
     46  *	This product includes software developed by the University of
     47  *	California, Berkeley and its contributors.
     48  * 4. Neither the name of the University nor the names of its contributors
     49  *    may be used to endorse or promote products derived from this software
     50  *    without specific prior written permission.
     51  *
     52  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     53  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     54  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     55  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     56  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     57  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     58  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     59  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     60  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     61  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     62  * SUCH DAMAGE.
     63  *
     64  *	@(#)rtsock.c	8.7 (Berkeley) 10/12/95
     65  */
     66 
     67 #include "opt_inet.h"
     68 
     69 #include <sys/param.h>
     70 #include <sys/systm.h>
     71 #include <sys/lwp.h>
     72 #include <sys/proc.h>
     73 #include <sys/mbuf.h>
     74 #include <sys/socket.h>
     75 #include <sys/socketvar.h>
     76 #include <sys/domain.h>
     77 #include <sys/protosw.h>
     78 
     79 #include <uvm/uvm_extern.h>
     80 
     81 #include <sys/sysctl.h>
     82 
     83 #include <net/if.h>
     84 #include <net/route.h>
     85 #include <net/raw_cb.h>
     86 
     87 #include <machine/stdarg.h>
     88 
     89 struct	sockaddr route_dst = { 2, PF_ROUTE, };
     90 struct	sockaddr route_src = { 2, PF_ROUTE, };
     91 struct	sockproto route_proto = { PF_ROUTE, };
     92 
     93 struct walkarg {
     94 	int	w_op;
     95 	int	w_arg;
     96 	int	w_given;
     97 	int	w_needed;
     98 	caddr_t	w_where;
     99 	int	w_tmemsize;
    100 	int	w_tmemneeded;
    101 	caddr_t	w_tmem;
    102 };
    103 
    104 static struct mbuf *rt_msg1 __P((int, struct rt_addrinfo *, caddr_t, int));
    105 static int rt_msg2 __P((int, struct rt_addrinfo *, caddr_t, struct walkarg *,
    106     int *));
    107 static int rt_xaddrs __P((caddr_t, caddr_t, struct rt_addrinfo *));
    108 static int sysctl_dumpentry __P((struct radix_node *, void *));
    109 static int sysctl_iflist __P((int, struct walkarg *, int));
    110 static int sysctl_rtable __P((int *, u_int, void *, size_t *, void *, size_t));
    111 static __inline void rt_adjustcount __P((int, int));
    112 
    113 /* Sleazy use of local variables throughout file, warning!!!! */
    114 #define dst	info.rti_info[RTAX_DST]
    115 #define gate	info.rti_info[RTAX_GATEWAY]
    116 #define netmask	info.rti_info[RTAX_NETMASK]
    117 #define genmask	info.rti_info[RTAX_GENMASK]
    118 #define ifpaddr	info.rti_info[RTAX_IFP]
    119 #define ifaaddr	info.rti_info[RTAX_IFA]
    120 #define brdaddr	info.rti_info[RTAX_BRD]
    121 
    122 static __inline void
    123 rt_adjustcount(af, cnt)
    124 	int af, cnt;
    125 {
    126 	route_cb.any_count += cnt;
    127 	switch (af) {
    128 	case AF_INET:
    129 		route_cb.ip_count += cnt;
    130 		return;
    131 #ifdef INET6
    132 	case AF_INET6:
    133 		route_cb.ip6_count += cnt;
    134 		return;
    135 #endif
    136 	case AF_IPX:
    137 		route_cb.ipx_count += cnt;
    138 		return;
    139 	case AF_NS:
    140 		route_cb.ns_count += cnt;
    141 		return;
    142 	case AF_ISO:
    143 		route_cb.iso_count += cnt;
    144 		return;
    145 	}
    146 }
    147 
    148 /*ARGSUSED*/
    149 int
    150 route_usrreq(so, req, m, nam, control, p)
    151 	struct socket *so;
    152 	int req;
    153 	struct mbuf *m, *nam, *control;
    154 	struct proc *p;
    155 {
    156 	int error = 0;
    157 	struct rawcb *rp = sotorawcb(so);
    158 	int s;
    159 
    160 	if (req == PRU_ATTACH) {
    161 		MALLOC(rp, struct rawcb *, sizeof(*rp), M_PCB, M_WAITOK);
    162 		if ((so->so_pcb = rp) != NULL)
    163 			bzero(so->so_pcb, sizeof(*rp));
    164 
    165 	}
    166 	if (req == PRU_DETACH && rp)
    167 		rt_adjustcount(rp->rcb_proto.sp_protocol, -1);
    168 	s = splsoftnet();
    169 
    170 	/*
    171 	 * Don't call raw_usrreq() in the attach case, because
    172 	 * we want to allow non-privileged processes to listen on
    173 	 * and send "safe" commands to the routing socket.
    174 	 */
    175 	if (req == PRU_ATTACH) {
    176 		if (p == 0)
    177 			error = EACCES;
    178 		else
    179 			error = raw_attach(so, (int)(long)nam);
    180 	} else
    181 		error = raw_usrreq(so, req, m, nam, control, p);
    182 
    183 	rp = sotorawcb(so);
    184 	if (req == PRU_ATTACH && rp) {
    185 		if (error) {
    186 			free((caddr_t)rp, M_PCB);
    187 			splx(s);
    188 			return (error);
    189 		}
    190 		rt_adjustcount(rp->rcb_proto.sp_protocol, 1);
    191 		rp->rcb_laddr = &route_src;
    192 		rp->rcb_faddr = &route_dst;
    193 		soisconnected(so);
    194 		so->so_options |= SO_USELOOPBACK;
    195 	}
    196 	splx(s);
    197 	return (error);
    198 }
    199 
    200 /*ARGSUSED*/
    201 int
    202 #if __STDC__
    203 route_output(struct mbuf *m, ...)
    204 #else
    205 route_output(m, va_alist)
    206 	struct mbuf *m;
    207 	va_dcl
    208 #endif
    209 {
    210 	struct rt_msghdr *rtm = 0;
    211 	struct radix_node *rn = 0;
    212 	struct rtentry *rt = 0;
    213 	struct rtentry *saved_nrt = 0;
    214 	struct radix_node_head *rnh;
    215 	struct rt_addrinfo info;
    216 	int len, error = 0;
    217 	struct ifnet *ifp = 0;
    218 	struct ifaddr *ifa = 0;
    219 	struct socket *so;
    220 	va_list ap;
    221 
    222 	va_start(ap, m);
    223 	so = va_arg(ap, struct socket *);
    224 	va_end(ap);
    225 
    226 #define senderr(e) do { error = e; goto flush;} while (0)
    227 	if (m == 0 || ((m->m_len < sizeof(int32_t)) &&
    228 	   (m = m_pullup(m, sizeof(int32_t))) == 0))
    229 		return (ENOBUFS);
    230 	if ((m->m_flags & M_PKTHDR) == 0)
    231 		panic("route_output");
    232 	len = m->m_pkthdr.len;
    233 	if (len < sizeof(*rtm) ||
    234 	    len != mtod(m, struct rt_msghdr *)->rtm_msglen) {
    235 		dst = 0;
    236 		senderr(EINVAL);
    237 	}
    238 	R_Malloc(rtm, struct rt_msghdr *, len);
    239 	if (rtm == 0) {
    240 		dst = 0;
    241 		senderr(ENOBUFS);
    242 	}
    243 	m_copydata(m, 0, len, (caddr_t)rtm);
    244 	if (rtm->rtm_version != RTM_VERSION) {
    245 		dst = 0;
    246 		senderr(EPROTONOSUPPORT);
    247 	}
    248 	rtm->rtm_pid = curproc->l_proc->p_pid;
    249 	bzero(&info, sizeof(info));
    250 	info.rti_addrs = rtm->rtm_addrs;
    251 	if (rt_xaddrs((caddr_t)(rtm + 1), len + (caddr_t)rtm, &info))
    252 		senderr(EINVAL);
    253 	info.rti_flags = rtm->rtm_flags;
    254 	if (dst == 0 || (dst->sa_family >= AF_MAX))
    255 		senderr(EINVAL);
    256 	if (gate != 0 && (gate->sa_family >= AF_MAX))
    257 		senderr(EINVAL);
    258 	if (genmask) {
    259 		struct radix_node *t;
    260 		t = rn_addmask((caddr_t)genmask, 0, 1);
    261 		if (t && Bcmp(genmask, t->rn_key, *(u_char *)genmask) == 0)
    262 			genmask = (struct sockaddr *)(t->rn_key);
    263 		else
    264 			senderr(ENOBUFS);
    265 	}
    266 
    267 	/*
    268 	 * Verify that the caller has the appropriate privilege; RTM_GET
    269 	 * is the only operation the non-superuser is allowed.
    270 	 */
    271 	if (rtm->rtm_type != RTM_GET &&
    272 	    suser(curproc->l_proc->p_ucred, &curproc->l_proc->p_acflag) != 0)
    273 		senderr(EACCES);
    274 
    275 	switch (rtm->rtm_type) {
    276 
    277 	case RTM_ADD:
    278 		if (gate == 0)
    279 			senderr(EINVAL);
    280 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
    281 		if (error == 0 && saved_nrt) {
    282 			rt_setmetrics(rtm->rtm_inits,
    283 			    &rtm->rtm_rmx, &saved_nrt->rt_rmx);
    284 			saved_nrt->rt_refcnt--;
    285 			saved_nrt->rt_genmask = genmask;
    286 		}
    287 		break;
    288 
    289 	case RTM_DELETE:
    290 		error = rtrequest1(rtm->rtm_type, &info, &saved_nrt);
    291 		if (error == 0) {
    292 			(rt = saved_nrt)->rt_refcnt++;
    293 			goto report;
    294 		}
    295 		break;
    296 
    297 	case RTM_GET:
    298 	case RTM_CHANGE:
    299 	case RTM_LOCK:
    300 		if ((rnh = rt_tables[dst->sa_family]) == 0) {
    301 			senderr(EAFNOSUPPORT);
    302 		}
    303 		rn = rnh->rnh_lookup(dst, netmask, rnh);
    304 		if (rn == NULL || (rn->rn_flags & RNF_ROOT) != 0) {
    305 			senderr(ESRCH);
    306 		}
    307 		rt = (struct rtentry *)rn;
    308 		rt->rt_refcnt++;
    309 
    310 		switch(rtm->rtm_type) {
    311 
    312 		case RTM_GET:
    313 		report:
    314 			dst = rt_key(rt);
    315 			gate = rt->rt_gateway;
    316 			netmask = rt_mask(rt);
    317 			genmask = rt->rt_genmask;
    318 			if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) {
    319 				if ((ifp = rt->rt_ifp) != NULL) {
    320 					ifpaddr = ifp->if_addrlist.tqh_first->ifa_addr;
    321 					ifaaddr = rt->rt_ifa->ifa_addr;
    322 					if (ifp->if_flags & IFF_POINTOPOINT)
    323 						brdaddr = rt->rt_ifa->ifa_dstaddr;
    324 					else
    325 						brdaddr = 0;
    326 					rtm->rtm_index = ifp->if_index;
    327 				} else {
    328 					ifpaddr = 0;
    329 					ifaaddr = 0;
    330 				}
    331 			}
    332 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)0,
    333 			    (struct walkarg *)0, &len);
    334 			if (len > rtm->rtm_msglen) {
    335 				struct rt_msghdr *new_rtm;
    336 				R_Malloc(new_rtm, struct rt_msghdr *, len);
    337 				if (new_rtm == 0)
    338 					senderr(ENOBUFS);
    339 				Bcopy(rtm, new_rtm, rtm->rtm_msglen);
    340 				Free(rtm); rtm = new_rtm;
    341 			}
    342 			(void)rt_msg2(rtm->rtm_type, &info, (caddr_t)rtm,
    343 			    (struct walkarg *)0, 0);
    344 			rtm->rtm_flags = rt->rt_flags;
    345 			rtm->rtm_rmx = rt->rt_rmx;
    346 			rtm->rtm_addrs = info.rti_addrs;
    347 			break;
    348 
    349 		case RTM_CHANGE:
    350 			/*
    351 			 * new gateway could require new ifaddr, ifp;
    352 			 * flags may also be different; ifp may be specified
    353 			 * by ll sockaddr when protocol address is ambiguous
    354 			 */
    355 			if ((error = rt_getifa(&info)) != 0)
    356 				senderr(error);
    357 			if (gate && rt_setgate(rt, rt_key(rt), gate))
    358 				senderr(EDQUOT);
    359 			/* new gateway could require new ifaddr, ifp;
    360 			   flags may also be different; ifp may be specified
    361 			   by ll sockaddr when protocol address is ambiguous */
    362 			if (ifpaddr && (ifa = ifa_ifwithnet(ifpaddr)) &&
    363 			    (ifp = ifa->ifa_ifp) && (ifaaddr || gate))
    364 				ifa = ifaof_ifpforaddr(ifaaddr ? ifaaddr : gate,
    365 				    ifp);
    366 			else if ((ifaaddr && (ifa = ifa_ifwithaddr(ifaaddr))) ||
    367 			    (gate && (ifa = ifa_ifwithroute(rt->rt_flags,
    368 			    rt_key(rt), gate))))
    369 				ifp = ifa->ifa_ifp;
    370 			if (ifa) {
    371 				struct ifaddr *oifa = rt->rt_ifa;
    372 				if (oifa != ifa) {
    373 				    if (oifa && oifa->ifa_rtrequest)
    374 					oifa->ifa_rtrequest(RTM_DELETE, rt,
    375 					    &info);
    376 				    IFAFREE(rt->rt_ifa);
    377 				    rt->rt_ifa = ifa;
    378 				    IFAREF(rt->rt_ifa);
    379 				    rt->rt_ifp = ifp;
    380 				}
    381 			}
    382 			rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx,
    383 			    &rt->rt_rmx);
    384 			if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest)
    385 				rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info);
    386 			if (genmask)
    387 				rt->rt_genmask = genmask;
    388 			/*
    389 			 * Fall into
    390 			 */
    391 		case RTM_LOCK:
    392 			rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits);
    393 			rt->rt_rmx.rmx_locks |=
    394 			    (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks);
    395 			break;
    396 		}
    397 		break;
    398 
    399 	default:
    400 		senderr(EOPNOTSUPP);
    401 	}
    402 
    403 flush:
    404 	if (rtm) {
    405 		if (error)
    406 			rtm->rtm_errno = error;
    407 		else
    408 			rtm->rtm_flags |= RTF_DONE;
    409 	}
    410 	if (rt)
    411 		rtfree(rt);
    412     {
    413 	struct rawcb *rp = 0;
    414 	/*
    415 	 * Check to see if we don't want our own messages.
    416 	 */
    417 	if ((so->so_options & SO_USELOOPBACK) == 0) {
    418 		if (route_cb.any_count <= 1) {
    419 			if (rtm)
    420 				Free(rtm);
    421 			m_freem(m);
    422 			return (error);
    423 		}
    424 		/* There is another listener, so construct message */
    425 		rp = sotorawcb(so);
    426 	}
    427 	if (rtm) {
    428 		m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm);
    429 		if (m->m_pkthdr.len < rtm->rtm_msglen) {
    430 			m_freem(m);
    431 			m = NULL;
    432 		} else if (m->m_pkthdr.len > rtm->rtm_msglen)
    433 			m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len);
    434 		Free(rtm);
    435 	}
    436 	if (rp)
    437 		rp->rcb_proto.sp_family = 0; /* Avoid us */
    438 	if (dst)
    439 		route_proto.sp_protocol = dst->sa_family;
    440 	if (m)
    441 		raw_input(m, &route_proto, &route_src, &route_dst);
    442 	if (rp)
    443 		rp->rcb_proto.sp_family = PF_ROUTE;
    444     }
    445 	return (error);
    446 }
    447 
    448 void
    449 rt_setmetrics(which, in, out)
    450 	u_long which;
    451 	struct rt_metrics *in, *out;
    452 {
    453 #define metric(f, e) if (which & (f)) out->e = in->e;
    454 	metric(RTV_RPIPE, rmx_recvpipe);
    455 	metric(RTV_SPIPE, rmx_sendpipe);
    456 	metric(RTV_SSTHRESH, rmx_ssthresh);
    457 	metric(RTV_RTT, rmx_rtt);
    458 	metric(RTV_RTTVAR, rmx_rttvar);
    459 	metric(RTV_HOPCOUNT, rmx_hopcount);
    460 	metric(RTV_MTU, rmx_mtu);
    461 	metric(RTV_EXPIRE, rmx_expire);
    462 #undef metric
    463 }
    464 
    465 #define ROUNDUP(a) \
    466 	((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
    467 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len))
    468 
    469 static int
    470 rt_xaddrs(cp, cplim, rtinfo)
    471 	caddr_t cp, cplim;
    472 	struct rt_addrinfo *rtinfo;
    473 {
    474 	struct sockaddr *sa;
    475 	int i;
    476 
    477 	bzero(rtinfo->rti_info, sizeof(rtinfo->rti_info));
    478 	for (i = 0; (i < RTAX_MAX) && (cp < cplim); i++) {
    479 		if ((rtinfo->rti_addrs & (1 << i)) == 0)
    480 			continue;
    481 		rtinfo->rti_info[i] = sa = (struct sockaddr *)cp;
    482 		ADVANCE(cp, sa);
    483 	}
    484 
    485 	/* Check for extra addresses specified.  */
    486 	if ((rtinfo->rti_addrs & (~0 << i)) != 0)
    487 		return (1);
    488 	/* Check for bad data length.  */
    489 	if (cp != cplim) {
    490 		if (i == RTAX_NETMASK + 1 &&
    491 		    cp - ROUNDUP(sa->sa_len) + sa->sa_len == cplim)
    492 			/*
    493 			 * The last sockaddr was netmask.
    494 			 * We accept this for now for the sake of old
    495 			 * binaries or third party softwares.
    496 			 */
    497 			;
    498 		else
    499 			return (1);
    500 	}
    501 	return (0);
    502 }
    503 
    504 static struct mbuf *
    505 rt_msg1(type, rtinfo, data, datalen)
    506 	int type;
    507 	struct rt_addrinfo *rtinfo;
    508 	caddr_t data;
    509 	int datalen;
    510 {
    511 	struct rt_msghdr *rtm;
    512 	struct mbuf *m;
    513 	int i;
    514 	struct sockaddr *sa;
    515 	int len, dlen;
    516 
    517 	m = m_gethdr(M_DONTWAIT, MT_DATA);
    518 	if (m == 0)
    519 		return (m);
    520 	switch (type) {
    521 
    522 	case RTM_DELADDR:
    523 	case RTM_NEWADDR:
    524 		len = sizeof(struct ifa_msghdr);
    525 		break;
    526 
    527 #ifdef COMPAT_14
    528 	case RTM_OIFINFO:
    529 		len = sizeof(struct if_msghdr14);
    530 		break;
    531 #endif
    532 
    533 	case RTM_IFINFO:
    534 		len = sizeof(struct if_msghdr);
    535 		break;
    536 
    537 	case RTM_IFANNOUNCE:
    538 		len = sizeof(struct if_announcemsghdr);
    539 		break;
    540 
    541 	default:
    542 		len = sizeof(struct rt_msghdr);
    543 	}
    544 	if (len > MHLEN + MLEN)
    545 		panic("rt_msg1: message too long");
    546 	else if (len > MHLEN) {
    547 		m->m_next = m_get(M_DONTWAIT, MT_DATA);
    548 		if (m->m_next == NULL) {
    549 			m_freem(m);
    550 			return (NULL);
    551 		}
    552 		m->m_pkthdr.len = len;
    553 		m->m_len = MHLEN;
    554 		m->m_next->m_len = len - MHLEN;
    555 	} else {
    556 		m->m_pkthdr.len = m->m_len = len;
    557 	}
    558 	m->m_pkthdr.rcvif = 0;
    559 	m_copyback(m, 0, datalen, data);
    560 	rtm = mtod(m, struct rt_msghdr *);
    561 	for (i = 0; i < RTAX_MAX; i++) {
    562 		if ((sa = rtinfo->rti_info[i]) == NULL)
    563 			continue;
    564 		rtinfo->rti_addrs |= (1 << i);
    565 		dlen = ROUNDUP(sa->sa_len);
    566 		m_copyback(m, len, dlen, (caddr_t)sa);
    567 		len += dlen;
    568 	}
    569 	if (m->m_pkthdr.len != len) {
    570 		m_freem(m);
    571 		return (NULL);
    572 	}
    573 	rtm->rtm_msglen = len;
    574 	rtm->rtm_version = RTM_VERSION;
    575 	rtm->rtm_type = type;
    576 	return (m);
    577 }
    578 
    579 /*
    580  * rt_msg2
    581  *
    582  *	 fills 'cp' or 'w'.w_tmem with the routing socket message and
    583  *		returns the length of the message in 'lenp'.
    584  *
    585  * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold
    586  *	the message
    587  * otherwise walkarg's w_needed is updated and if the user buffer is
    588  *	specified and w_needed indicates space exists the information is copied
    589  *	into the temp space (w_tmem). w_tmem is [re]allocated if necessary,
    590  *	if the allocation fails ENOBUFS is returned.
    591  */
    592 static int
    593 rt_msg2(type, rtinfo, cp, w, lenp)
    594 	int type;
    595 	struct rt_addrinfo *rtinfo;
    596 	caddr_t cp;
    597 	struct walkarg *w;
    598 	int *lenp;
    599 {
    600 	int i;
    601 	int len, dlen, second_time = 0;
    602 	caddr_t cp0;
    603 
    604 	rtinfo->rti_addrs = 0;
    605 again:
    606 	switch (type) {
    607 
    608 	case RTM_DELADDR:
    609 	case RTM_NEWADDR:
    610 		len = sizeof(struct ifa_msghdr);
    611 		break;
    612 #ifdef COMPAT_14
    613 	case RTM_OIFINFO:
    614 		len = sizeof(struct if_msghdr14);
    615 		break;
    616 #endif
    617 
    618 	case RTM_IFINFO:
    619 		len = sizeof(struct if_msghdr);
    620 		break;
    621 
    622 	default:
    623 		len = sizeof(struct rt_msghdr);
    624 	}
    625 	if ((cp0 = cp) != NULL)
    626 		cp += len;
    627 	for (i = 0; i < RTAX_MAX; i++) {
    628 		struct sockaddr *sa;
    629 
    630 		if ((sa = rtinfo->rti_info[i]) == 0)
    631 			continue;
    632 		rtinfo->rti_addrs |= (1 << i);
    633 		dlen = ROUNDUP(sa->sa_len);
    634 		if (cp) {
    635 			bcopy(sa, cp, (unsigned)dlen);
    636 			cp += dlen;
    637 		}
    638 		len += dlen;
    639 	}
    640 	if (cp == 0 && w != NULL && !second_time) {
    641 		struct walkarg *rw = w;
    642 
    643 		rw->w_needed += len;
    644 		if (rw->w_needed <= 0 && rw->w_where) {
    645 			if (rw->w_tmemsize < len) {
    646 				if (rw->w_tmem)
    647 					free(rw->w_tmem, M_RTABLE);
    648 				rw->w_tmem = (caddr_t) malloc(len, M_RTABLE,
    649 				    M_NOWAIT);
    650 				if (rw->w_tmem)
    651 					rw->w_tmemsize = len;
    652 			}
    653 			if (rw->w_tmem) {
    654 				cp = rw->w_tmem;
    655 				second_time = 1;
    656 				goto again;
    657 			} else {
    658 				rw->w_tmemneeded = len;
    659 				return (ENOBUFS);
    660 			}
    661 		}
    662 	}
    663 	if (cp) {
    664 		struct rt_msghdr *rtm = (struct rt_msghdr *)cp0;
    665 
    666 		rtm->rtm_version = RTM_VERSION;
    667 		rtm->rtm_type = type;
    668 		rtm->rtm_msglen = len;
    669 	}
    670 	if (lenp)
    671 		*lenp = len;
    672 	return (0);
    673 }
    674 
    675 /*
    676  * This routine is called to generate a message from the routing
    677  * socket indicating that a redirect has occured, a routing lookup
    678  * has failed, or that a protocol has detected timeouts to a particular
    679  * destination.
    680  */
    681 void
    682 rt_missmsg(type, rtinfo, flags, error)
    683 	int type, flags, error;
    684 	struct rt_addrinfo *rtinfo;
    685 {
    686 	struct rt_msghdr rtm;
    687 	struct mbuf *m;
    688 	struct sockaddr *sa = rtinfo->rti_info[RTAX_DST];
    689 
    690 	if (route_cb.any_count == 0)
    691 		return;
    692 	bzero(&rtm, sizeof(rtm));
    693 	rtm.rtm_flags = RTF_DONE | flags;
    694 	rtm.rtm_errno = error;
    695 	m = rt_msg1(type, rtinfo, (caddr_t)&rtm, sizeof(rtm));
    696 	if (m == 0)
    697 		return;
    698 	mtod(m, struct rt_msghdr *)->rtm_addrs = rtinfo->rti_addrs;
    699 	route_proto.sp_protocol = sa ? sa->sa_family : 0;
    700 	raw_input(m, &route_proto, &route_src, &route_dst);
    701 }
    702 
    703 /*
    704  * This routine is called to generate a message from the routing
    705  * socket indicating that the status of a network interface has changed.
    706  */
    707 void
    708 rt_ifmsg(ifp)
    709 	struct ifnet *ifp;
    710 {
    711 	struct if_msghdr ifm;
    712 #ifdef COMPAT_14
    713 	struct if_msghdr14 oifm;
    714 #endif
    715 	struct mbuf *m;
    716 	struct rt_addrinfo info;
    717 
    718 	if (route_cb.any_count == 0)
    719 		return;
    720 	bzero(&info, sizeof(info));
    721 	bzero(&ifm, sizeof(ifm));
    722 	ifm.ifm_index = ifp->if_index;
    723 	ifm.ifm_flags = ifp->if_flags;
    724 	ifm.ifm_data = ifp->if_data;
    725 	ifm.ifm_addrs = 0;
    726 	m = rt_msg1(RTM_IFINFO, &info, (caddr_t)&ifm, sizeof(ifm));
    727 	if (m == 0)
    728 		return;
    729 	route_proto.sp_protocol = 0;
    730 	raw_input(m, &route_proto, &route_src, &route_dst);
    731 #ifdef COMPAT_14
    732 	bzero(&info, sizeof(info));
    733 	bzero(&oifm, sizeof(oifm));
    734 	oifm.ifm_index = ifp->if_index;
    735 	oifm.ifm_flags = ifp->if_flags;
    736 	oifm.ifm_data.ifi_type = ifp->if_data.ifi_type;
    737 	oifm.ifm_data.ifi_addrlen = ifp->if_data.ifi_addrlen;
    738 	oifm.ifm_data.ifi_hdrlen = ifp->if_data.ifi_hdrlen;
    739 	oifm.ifm_data.ifi_mtu = ifp->if_data.ifi_mtu;
    740 	oifm.ifm_data.ifi_metric = ifp->if_data.ifi_metric;
    741 	oifm.ifm_data.ifi_baudrate = ifp->if_data.ifi_baudrate;
    742 	oifm.ifm_data.ifi_ipackets = ifp->if_data.ifi_ipackets;
    743 	oifm.ifm_data.ifi_ierrors = ifp->if_data.ifi_ierrors;
    744 	oifm.ifm_data.ifi_opackets = ifp->if_data.ifi_opackets;
    745 	oifm.ifm_data.ifi_oerrors = ifp->if_data.ifi_oerrors;
    746 	oifm.ifm_data.ifi_collisions = ifp->if_data.ifi_collisions;
    747 	oifm.ifm_data.ifi_ibytes = ifp->if_data.ifi_ibytes;
    748 	oifm.ifm_data.ifi_obytes = ifp->if_data.ifi_obytes;
    749 	oifm.ifm_data.ifi_imcasts = ifp->if_data.ifi_imcasts;
    750 	oifm.ifm_data.ifi_omcasts = ifp->if_data.ifi_omcasts;
    751 	oifm.ifm_data.ifi_iqdrops = ifp->if_data.ifi_iqdrops;
    752 	oifm.ifm_data.ifi_noproto = ifp->if_data.ifi_noproto;
    753 	oifm.ifm_data.ifi_lastchange = ifp->if_data.ifi_lastchange;
    754 	oifm.ifm_addrs = 0;
    755 	m = rt_msg1(RTM_OIFINFO, &info, (caddr_t)&oifm, sizeof(oifm));
    756 	if (m == 0)
    757 		return;
    758 	route_proto.sp_protocol = 0;
    759 	raw_input(m, &route_proto, &route_src, &route_dst);
    760 #endif
    761 }
    762 
    763 /*
    764  * This is called to generate messages from the routing socket
    765  * indicating a network interface has had addresses associated with it.
    766  * if we ever reverse the logic and replace messages TO the routing
    767  * socket indicate a request to configure interfaces, then it will
    768  * be unnecessary as the routing socket will automatically generate
    769  * copies of it.
    770  */
    771 void
    772 rt_newaddrmsg(cmd, ifa, error, rt)
    773 	int cmd, error;
    774 	struct ifaddr *ifa;
    775 	struct rtentry *rt;
    776 {
    777 	struct rt_addrinfo info;
    778 	struct sockaddr *sa = NULL;
    779 	int pass;
    780 	struct mbuf *m = NULL;
    781 	struct ifnet *ifp = ifa->ifa_ifp;
    782 
    783 	if (route_cb.any_count == 0)
    784 		return;
    785 	for (pass = 1; pass < 3; pass++) {
    786 		bzero(&info, sizeof(info));
    787 		if ((cmd == RTM_ADD && pass == 1) ||
    788 		    (cmd == RTM_DELETE && pass == 2)) {
    789 			struct ifa_msghdr ifam;
    790 			int ncmd = cmd == RTM_ADD ? RTM_NEWADDR : RTM_DELADDR;
    791 
    792 			ifaaddr = sa = ifa->ifa_addr;
    793 			ifpaddr = ifp->if_addrlist.tqh_first->ifa_addr;
    794 			netmask = ifa->ifa_netmask;
    795 			brdaddr = ifa->ifa_dstaddr;
    796 			bzero(&ifam, sizeof(ifam));
    797 			ifam.ifam_index = ifp->if_index;
    798 			ifam.ifam_metric = ifa->ifa_metric;
    799 			ifam.ifam_flags = ifa->ifa_flags;
    800 			m = rt_msg1(ncmd, &info, (caddr_t)&ifam, sizeof(ifam));
    801 			if (m == NULL)
    802 				continue;
    803 			mtod(m, struct ifa_msghdr *)->ifam_addrs =
    804 			    info.rti_addrs;
    805 		}
    806 		if ((cmd == RTM_ADD && pass == 2) ||
    807 		    (cmd == RTM_DELETE && pass == 1)) {
    808 			struct rt_msghdr rtm;
    809 
    810 			if (rt == 0)
    811 				continue;
    812 			netmask = rt_mask(rt);
    813 			dst = sa = rt_key(rt);
    814 			gate = rt->rt_gateway;
    815 			bzero(&rtm, sizeof(rtm));
    816 			rtm.rtm_index = ifp->if_index;
    817 			rtm.rtm_flags |= rt->rt_flags;
    818 			rtm.rtm_errno = error;
    819 			m = rt_msg1(cmd, &info, (caddr_t)&rtm, sizeof(rtm));
    820 			if (m == NULL)
    821 				continue;
    822 			mtod(m, struct rt_msghdr *)->rtm_addrs = info.rti_addrs;
    823 		}
    824 		route_proto.sp_protocol = sa ? sa->sa_family : 0;
    825 		raw_input(m, &route_proto, &route_src, &route_dst);
    826 	}
    827 }
    828 
    829 /*
    830  * This is called to generate routing socket messages indicating
    831  * network interface arrival and departure.
    832  */
    833 void
    834 rt_ifannouncemsg(ifp, what)
    835 	struct ifnet *ifp;
    836 	int what;
    837 {
    838 	struct if_announcemsghdr ifan;
    839 	struct mbuf *m;
    840 	struct rt_addrinfo info;
    841 
    842 	if (route_cb.any_count == 0)
    843 		return;
    844 	bzero(&info, sizeof(info));
    845 	bzero(&ifan, sizeof(ifan));
    846 	ifan.ifan_index = ifp->if_index;
    847 	strcpy(ifan.ifan_name, ifp->if_xname);
    848 	ifan.ifan_what = what;
    849 	m = rt_msg1(RTM_IFANNOUNCE, &info, (caddr_t)&ifan, sizeof(ifan));
    850 	if (m == 0)
    851 		return;
    852 	route_proto.sp_protocol = 0;
    853 	raw_input(m, &route_proto, &route_src, &route_dst);
    854 }
    855 
    856 /*
    857  * This is used in dumping the kernel table via sysctl().
    858  */
    859 static int
    860 sysctl_dumpentry(rn, v)
    861 	struct radix_node *rn;
    862 	void *v;
    863 {
    864 	struct walkarg *w = v;
    865 	struct rtentry *rt = (struct rtentry *)rn;
    866 	int error = 0, size;
    867 	struct rt_addrinfo info;
    868 
    869 	if (w->w_op == NET_RT_FLAGS && !(rt->rt_flags & w->w_arg))
    870 		return 0;
    871 	bzero(&info, sizeof(info));
    872 	dst = rt_key(rt);
    873 	gate = rt->rt_gateway;
    874 	netmask = rt_mask(rt);
    875 	genmask = rt->rt_genmask;
    876 	if (rt->rt_ifp) {
    877 		ifpaddr = rt->rt_ifp->if_addrlist.tqh_first->ifa_addr;
    878 		ifaaddr = rt->rt_ifa->ifa_addr;
    879 		if (rt->rt_ifp->if_flags & IFF_POINTOPOINT)
    880 			brdaddr = rt->rt_ifa->ifa_dstaddr;
    881 	}
    882 	if ((error = rt_msg2(RTM_GET, &info, 0, w, &size)))
    883 		return (error);
    884 	if (w->w_where && w->w_tmem && w->w_needed <= 0) {
    885 		struct rt_msghdr *rtm = (struct rt_msghdr *)w->w_tmem;
    886 
    887 		rtm->rtm_flags = rt->rt_flags;
    888 		rtm->rtm_use = rt->rt_use;
    889 		rtm->rtm_rmx = rt->rt_rmx;
    890 		rtm->rtm_index = rt->rt_ifp->if_index;
    891 		rtm->rtm_errno = rtm->rtm_pid = rtm->rtm_seq = 0;
    892 		rtm->rtm_addrs = info.rti_addrs;
    893 		if ((error = copyout(rtm, w->w_where, size)) != 0)
    894 			w->w_where = NULL;
    895 		else
    896 			w->w_where += size;
    897 	}
    898 	return (error);
    899 }
    900 
    901 static int
    902 sysctl_iflist(af, w, type)
    903 	int	af;
    904 	struct	walkarg *w;
    905 	int type;
    906 {
    907 	struct ifnet *ifp;
    908 	struct ifaddr *ifa;
    909 	struct	rt_addrinfo info;
    910 	int	len, error = 0;
    911 
    912 	bzero(&info, sizeof(info));
    913 	for (ifp = ifnet.tqh_first; ifp != 0; ifp = ifp->if_list.tqe_next) {
    914 		if (w->w_arg && w->w_arg != ifp->if_index)
    915 			continue;
    916 		ifa = ifp->if_addrlist.tqh_first;
    917 		ifpaddr = ifa->ifa_addr;
    918 		switch(type) {
    919 		case NET_RT_IFLIST:
    920 			error =
    921 			    rt_msg2(RTM_IFINFO, &info, (caddr_t)0, w, &len);
    922 			break;
    923 #ifdef COMPAT_14
    924 		case NET_RT_OIFLIST:
    925 			error =
    926 			    rt_msg2(RTM_OIFINFO, &info, (caddr_t)0, w, &len);
    927 			break;
    928 #endif
    929 		default:
    930 			panic("sysctl_iflist(1)");
    931 		}
    932 		if (error)
    933 			return (error);
    934 		ifpaddr = 0;
    935 		if (w->w_where && w->w_tmem && w->w_needed <= 0) {
    936 			switch(type) {
    937 			case NET_RT_IFLIST: {
    938 				struct if_msghdr *ifm;
    939 
    940 				ifm = (struct if_msghdr *)w->w_tmem;
    941 				ifm->ifm_index = ifp->if_index;
    942 				ifm->ifm_flags = ifp->if_flags;
    943 				ifm->ifm_data = ifp->if_data;
    944 				ifm->ifm_addrs = info.rti_addrs;
    945 				error = copyout(ifm, w->w_where, len);
    946 				if (error)
    947 					return (error);
    948 				w->w_where += len;
    949 				break;
    950 			}
    951 
    952 #ifdef COMPAT_14
    953 			case NET_RT_OIFLIST: {
    954 				struct if_msghdr14 *ifm;
    955 
    956 				ifm = (struct if_msghdr14 *)w->w_tmem;
    957 				ifm->ifm_index = ifp->if_index;
    958 				ifm->ifm_flags = ifp->if_flags;
    959 				ifm->ifm_data.ifi_type = ifp->if_data.ifi_type;
    960 				ifm->ifm_data.ifi_addrlen =
    961 				    ifp->if_data.ifi_addrlen;
    962 				ifm->ifm_data.ifi_hdrlen =
    963 				    ifp->if_data.ifi_hdrlen;
    964 				ifm->ifm_data.ifi_mtu = ifp->if_data.ifi_mtu;
    965 				ifm->ifm_data.ifi_metric =
    966 				    ifp->if_data.ifi_metric;
    967 				ifm->ifm_data.ifi_baudrate =
    968 				    ifp->if_data.ifi_baudrate;
    969 				ifm->ifm_data.ifi_ipackets =
    970 				    ifp->if_data.ifi_ipackets;
    971 				ifm->ifm_data.ifi_ierrors =
    972 				    ifp->if_data.ifi_ierrors;
    973 				ifm->ifm_data.ifi_opackets =
    974 				    ifp->if_data.ifi_opackets;
    975 				ifm->ifm_data.ifi_oerrors =
    976 				    ifp->if_data.ifi_oerrors;
    977 				ifm->ifm_data.ifi_collisions =
    978 				    ifp->if_data.ifi_collisions;
    979 				ifm->ifm_data.ifi_ibytes =
    980 				    ifp->if_data.ifi_ibytes;
    981 				ifm->ifm_data.ifi_obytes =
    982 				    ifp->if_data.ifi_obytes;
    983 				ifm->ifm_data.ifi_imcasts =
    984 				    ifp->if_data.ifi_imcasts;
    985 				ifm->ifm_data.ifi_omcasts =
    986 				    ifp->if_data.ifi_omcasts;
    987 				ifm->ifm_data.ifi_iqdrops =
    988 				    ifp->if_data.ifi_iqdrops;
    989 				ifm->ifm_data.ifi_noproto =
    990 				    ifp->if_data.ifi_noproto;
    991 				ifm->ifm_data.ifi_lastchange =
    992 				    ifp->if_data.ifi_lastchange;
    993 				ifm->ifm_addrs = info.rti_addrs;
    994 				error = copyout(ifm, w->w_where, len);
    995 				if (error)
    996 					return (error);
    997 				w->w_where += len;
    998 				break;
    999 			}
   1000 #endif
   1001 			default:
   1002 				panic("sysctl_iflist(2)");
   1003 			}
   1004 		}
   1005 		while ((ifa = ifa->ifa_list.tqe_next) != NULL) {
   1006 			if (af && af != ifa->ifa_addr->sa_family)
   1007 				continue;
   1008 			ifaaddr = ifa->ifa_addr;
   1009 			netmask = ifa->ifa_netmask;
   1010 			brdaddr = ifa->ifa_dstaddr;
   1011 			if ((error = rt_msg2(RTM_NEWADDR, &info, 0, w, &len)))
   1012 				return (error);
   1013 			if (w->w_where && w->w_tmem && w->w_needed <= 0) {
   1014 				struct ifa_msghdr *ifam;
   1015 
   1016 				ifam = (struct ifa_msghdr *)w->w_tmem;
   1017 				ifam->ifam_index = ifa->ifa_ifp->if_index;
   1018 				ifam->ifam_flags = ifa->ifa_flags;
   1019 				ifam->ifam_metric = ifa->ifa_metric;
   1020 				ifam->ifam_addrs = info.rti_addrs;
   1021 				error = copyout(w->w_tmem, w->w_where, len);
   1022 				if (error)
   1023 					return (error);
   1024 				w->w_where += len;
   1025 			}
   1026 		}
   1027 		ifaaddr = netmask = brdaddr = 0;
   1028 	}
   1029 	return (0);
   1030 }
   1031 
   1032 static int
   1033 sysctl_rtable(name, namelen, where, given, new, newlen)
   1034 	int	*name;
   1035 	u_int	namelen;
   1036 	void 	*where;
   1037 	size_t	*given;
   1038 	void	*new;
   1039 	size_t	newlen;
   1040 {
   1041 	struct radix_node_head *rnh;
   1042 	int	i, s, error = EINVAL;
   1043 	u_char  af;
   1044 	struct	walkarg w;
   1045 
   1046 	if (new)
   1047 		return (EPERM);
   1048 	if (namelen != 3)
   1049 		return (EINVAL);
   1050 	af = name[0];
   1051 	w.w_tmemneeded = 0;
   1052 	w.w_tmemsize = 0;
   1053 	w.w_tmem = NULL;
   1054 again:
   1055 	/* we may return here if a later [re]alloc of the t_mem buffer fails */
   1056 	if (w.w_tmemneeded) {
   1057 		w.w_tmem = (caddr_t) malloc(w.w_tmemneeded, M_RTABLE, M_WAITOK);
   1058 		w.w_tmemsize = w.w_tmemneeded;
   1059 		w.w_tmemneeded = 0;
   1060 	}
   1061 	w.w_op = name[1];
   1062 	w.w_arg = name[2];
   1063 	w.w_given = *given;
   1064 	w.w_needed = 0 - w.w_given;
   1065 	w.w_where = where;
   1066 
   1067 	s = splsoftnet();
   1068 	switch (w.w_op) {
   1069 
   1070 	case NET_RT_DUMP:
   1071 	case NET_RT_FLAGS:
   1072 		for (i = 1; i <= AF_MAX; i++)
   1073 			if ((rnh = rt_tables[i]) && (af == 0 || af == i) &&
   1074 			    (error = (*rnh->rnh_walktree)(rnh,
   1075 			    sysctl_dumpentry, &w)))
   1076 				break;
   1077 		break;
   1078 
   1079 #ifdef COMPAT_14
   1080 	case NET_RT_OIFLIST:
   1081 		error = sysctl_iflist(af, &w, w.w_op);
   1082 		break;
   1083 #endif
   1084 
   1085 	case NET_RT_IFLIST:
   1086 		error = sysctl_iflist(af, &w, w.w_op);
   1087 	}
   1088 	splx(s);
   1089 
   1090 	/* check to see if we couldn't allocate memory with NOWAIT */
   1091 	if (error == ENOBUFS && w.w_tmem == 0 && w.w_tmemneeded)
   1092 		goto again;
   1093 
   1094 	if (w.w_tmem)
   1095 		free(w.w_tmem, M_RTABLE);
   1096 	w.w_needed += w.w_given;
   1097 	if (where) {
   1098 		*given = w.w_where - (caddr_t) where;
   1099 		if (*given < w.w_needed)
   1100 			return (ENOMEM);
   1101 	} else {
   1102 		*given = (11 * w.w_needed) / 10;
   1103 	}
   1104 	return (error);
   1105 }
   1106 
   1107 /*
   1108  * Definitions of protocols supported in the ROUTE domain.
   1109  */
   1110 
   1111 extern	struct domain routedomain;		/* or at least forward */
   1112 
   1113 struct protosw routesw[] = {
   1114 { SOCK_RAW,	&routedomain,	0,		PR_ATOMIC|PR_ADDR,
   1115   raw_input,	route_output,	raw_ctlinput,	0,
   1116   route_usrreq,
   1117   raw_init,	0,		0,		0,
   1118   sysctl_rtable,
   1119 }
   1120 };
   1121 
   1122 struct domain routedomain =
   1123     { PF_ROUTE, "route", route_init, 0, 0,
   1124       routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] };
   1125