Home | History | Annotate | Line # | Download | only in net
route.c revision 1.9
      1 /*
      2  * Copyright (c) 1980, 1986, 1991 Regents of the University of California.
      3  * All rights reserved.
      4  *
      5  * Redistribution and use in source and binary forms, with or without
      6  * modification, are permitted provided that the following conditions
      7  * are met:
      8  * 1. Redistributions of source code must retain the above copyright
      9  *    notice, this list of conditions and the following disclaimer.
     10  * 2. Redistributions in binary form must reproduce the above copyright
     11  *    notice, this list of conditions and the following disclaimer in the
     12  *    documentation and/or other materials provided with the distribution.
     13  * 3. All advertising materials mentioning features or use of this software
     14  *    must display the following acknowledgement:
     15  *	This product includes software developed by the University of
     16  *	California, Berkeley and its contributors.
     17  * 4. Neither the name of the University nor the names of its contributors
     18  *    may be used to endorse or promote products derived from this software
     19  *    without specific prior written permission.
     20  *
     21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     31  * SUCH DAMAGE.
     32  *
     33  *	from: @(#)route.c	7.22 (Berkeley) 6/27/91
     34  *	$Id: route.c,v 1.9 1994/05/11 09:26:49 mycroft Exp $
     35  */
     36 
     37 #include <sys/param.h>
     38 #include <sys/systm.h>
     39 #include <sys/proc.h>
     40 #include <sys/mbuf.h>
     41 #include <sys/socket.h>
     42 #include <sys/socketvar.h>
     43 #include <sys/domain.h>
     44 #include <sys/protosw.h>
     45 #include <sys/ioctl.h>
     46 
     47 #include <net/if.h>
     48 #include <net/route.h>
     49 #include <net/raw_cb.h>
     50 #include <net/netisr.h>
     51 
     52 #include <netinet/in.h>
     53 #include <netinet/in_var.h>
     54 
     55 #ifdef NS
     56 #include <netns/ns.h>
     57 #endif
     58 
     59 #include <machine/cpu.h>
     60 
     61 #define	SA(p) ((struct sockaddr *)(p))
     62 
     63 int	rttrash;		/* routes not in table but not freed */
     64 struct	sockaddr wildcard;	/* zero valued cookie for wildcard searches */
     65 int	rthashsize = RTHASHSIZ;	/* for netstat, etc. */
     66 
     67 static int rtinits_done = 0;
     68 struct radix_node_head *ns_rnhead, *in_rnhead;
     69 struct radix_node *rn_match(), *rn_delete(), *rn_addroute();
     70 
     71 void
     72 rtinitheads()
     73 {
     74 	if (rtinits_done == 0 &&
     75 #ifdef NS
     76 	    rn_inithead(&ns_rnhead, 16, AF_NS) &&
     77 #endif
     78 	    rn_inithead(&in_rnhead, 32, AF_INET))
     79 		rtinits_done = 1;
     80 }
     81 
     82 /*
     83  * Packet routing routines.
     84  */
     85 void
     86 rtalloc(ro)
     87 	register struct route *ro;
     88 {
     89 	if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP))
     90 		return;				 /* XXX */
     91 	ro->ro_rt = rtalloc1(&ro->ro_dst, 1);
     92 }
     93 
     94 struct rtentry *
     95 rtalloc1(dst, report)
     96 	register struct sockaddr *dst;
     97 	int  report;
     98 {
     99 	register struct radix_node_head *rnh;
    100 	register struct rtentry *rt;
    101 	register struct radix_node *rn;
    102 	struct rtentry *newrt = 0;
    103 	int  s = splnet(), err = 0, msgtype = RTM_MISS;
    104 
    105 	for (rnh = radix_node_head; rnh && (dst->sa_family != rnh->rnh_af); )
    106 		rnh = rnh->rnh_next;
    107 	if (rnh && rnh->rnh_treetop &&
    108 	    (rn = rn_match((caddr_t)dst, rnh->rnh_treetop)) &&
    109 	    ((rn->rn_flags & RNF_ROOT) == 0)) {
    110 		newrt = rt = (struct rtentry *)rn;
    111 		if (report && (rt->rt_flags & RTF_CLONING)) {
    112 			err = rtrequest(RTM_RESOLVE, dst, SA(0),
    113 			    SA(0), 0, &newrt);
    114 			if (err) {
    115 				newrt = rt;
    116 				rt->rt_refcnt++;
    117 				goto miss;
    118 			}
    119 			if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) {
    120 				msgtype = RTM_RESOLVE;
    121 				goto miss;
    122 			}
    123 		} else
    124 			rt->rt_refcnt++;
    125 	} else {
    126 		rtstat.rts_unreach++;
    127 	miss:	if (report)
    128 			rt_missmsg(msgtype, dst, SA(0), SA(0), SA(0), 0, err);
    129 	}
    130 	splx(s);
    131 	return (newrt);
    132 }
    133 
    134 void
    135 rtfree(rt)
    136 	register struct rtentry *rt;
    137 {
    138 	register struct ifaddr *ifa;
    139 	if (rt == 0)
    140 		panic("rtfree");
    141 	rt->rt_refcnt--;
    142 	if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) {
    143 		rttrash--;
    144 		if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
    145 			panic ("rtfree 2");
    146 		free((caddr_t)rt, M_RTABLE);
    147 	}
    148 }
    149 
    150 /*
    151  * Force a routing table entry to the specified
    152  * destination to go through the given gateway.
    153  * Normally called as a result of a routing redirect
    154  * message from the network layer.
    155  *
    156  * N.B.: must be called at splnet
    157  *
    158  */
    159 int
    160 rtredirect(dst, gateway, netmask, flags, src, rtp)
    161 	struct sockaddr *dst, *gateway, *netmask, *src;
    162 	int flags;
    163 	struct rtentry **rtp;
    164 {
    165 	register struct rtentry *rt = 0;
    166 	int error = 0;
    167 	short *stat = 0;
    168 
    169 	/* verify the gateway is directly reachable */
    170 	if (ifa_ifwithnet(gateway) == 0) {
    171 		error = ENETUNREACH;
    172 		goto out;
    173 	}
    174 	rt = rtalloc1(dst, 0);
    175 	/*
    176 	 * If the redirect isn't from our current router for this dst,
    177 	 * it's either old or wrong.  If it redirects us to ourselves,
    178 	 * we have a routing loop, perhaps as a result of an interface
    179 	 * going down recently.
    180 	 */
    181 #define	equal(a1, a2) (bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0)
    182 	if (!(flags & RTF_DONE) && rt && !equal(src, rt->rt_gateway))
    183 		error = EINVAL;
    184 	else if (ifa_ifwithaddr(gateway))
    185 		error = EHOSTUNREACH;
    186 	if (error)
    187 		goto done;
    188 	/*
    189 	 * Create a new entry if we just got back a wildcard entry
    190 	 * or the the lookup failed.  This is necessary for hosts
    191 	 * which use routing redirects generated by smart gateways
    192 	 * to dynamically build the routing tables.
    193 	 */
    194 	if ((rt == 0) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
    195 		goto create;
    196 	/*
    197 	 * Don't listen to the redirect if it's
    198 	 * for a route to an interface.
    199 	 */
    200 	if (rt->rt_flags & RTF_GATEWAY) {
    201 		if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
    202 			/*
    203 			 * Changing from route to net => route to host.
    204 			 * Create new route, rather than smashing route to net.
    205 			 */
    206 		create:
    207 			flags |=  RTF_GATEWAY | RTF_DYNAMIC;
    208 			error = rtrequest((int)RTM_ADD, dst, gateway,
    209 				    SA(0), flags,
    210 				    (struct rtentry **)0);
    211 			stat = &rtstat.rts_dynamic;
    212 		} else {
    213 			/*
    214 			 * Smash the current notion of the gateway to
    215 			 * this destination.  Should check about netmask!!!
    216 			 */
    217 			if (gateway->sa_len <= rt->rt_gateway->sa_len) {
    218 				Bcopy(gateway, rt->rt_gateway, gateway->sa_len);
    219 				rt->rt_flags |= RTF_MODIFIED;
    220 				flags |= RTF_MODIFIED;
    221 				stat = &rtstat.rts_newgateway;
    222 			} else
    223 				error = ENOSPC;
    224 		}
    225 	} else
    226 		error = EHOSTUNREACH;
    227 done:
    228 	if (rt) {
    229 		if (rtp && !error)
    230 			*rtp = rt;
    231 		else
    232 			rtfree(rt);
    233 	}
    234 out:
    235 	if (error)
    236 		rtstat.rts_badredirect++;
    237 	else if (stat != NULL)
    238 		(*stat)++;
    239 	rt_missmsg(RTM_REDIRECT, dst, gateway, netmask, src, flags, error);
    240 }
    241 
    242 /*
    243  * Routing table ioctl interface.
    244  */
    245 int
    246 rtioctl(req, data, p)
    247 	int req;
    248 	caddr_t data;
    249 	struct proc *p;
    250 {
    251 #ifndef COMPAT_43
    252 	return (EOPNOTSUPP);
    253 #else
    254 	register struct ortentry *entry = (struct ortentry *)data;
    255 	int error;
    256 	struct sockaddr *netmask = 0;
    257 
    258 	if (req == SIOCADDRT)
    259 		req = RTM_ADD;
    260 	else if (req == SIOCDELRT)
    261 		req = RTM_DELETE;
    262 	else
    263 		return (EINVAL);
    264 
    265 	if (error = suser(p->p_ucred, &p->p_acflag))
    266 		return (error);
    267 #if BYTE_ORDER != BIG_ENDIAN
    268 	if (entry->rt_dst.sa_family == 0 && entry->rt_dst.sa_len < 16) {
    269 		entry->rt_dst.sa_family = entry->rt_dst.sa_len;
    270 		entry->rt_dst.sa_len = 16;
    271 	}
    272 	if (entry->rt_gateway.sa_family == 0 && entry->rt_gateway.sa_len < 16) {
    273 		entry->rt_gateway.sa_family = entry->rt_gateway.sa_len;
    274 		entry->rt_gateway.sa_len = 16;
    275 	}
    276 #else
    277 	if (entry->rt_dst.sa_len == 0)
    278 		entry->rt_dst.sa_len = 16;
    279 	if (entry->rt_gateway.sa_len == 0)
    280 		entry->rt_gateway.sa_len = 16;
    281 #endif
    282 	if ((entry->rt_flags & RTF_HOST) == 0)
    283 		switch (entry->rt_dst.sa_family) {
    284 #ifdef INET
    285 		case AF_INET:
    286 			{
    287 				extern struct sockaddr_in icmpmask;
    288 				struct sockaddr_in *dst_in =
    289 					(struct sockaddr_in *)&entry->rt_dst;
    290 
    291 				in_sockmaskof(dst_in->sin_addr, &icmpmask);
    292 				netmask = (struct sockaddr *)&icmpmask;
    293 			}
    294 			break;
    295 #endif
    296 #ifdef NS
    297 		case AF_NS:
    298 			{
    299 				extern struct sockaddr_ns ns_netmask;
    300 				netmask = (struct sockaddr *)&ns_netmask;
    301 			}
    302 #endif
    303 		}
    304 	error =  rtrequest(req, &(entry->rt_dst), &(entry->rt_gateway), netmask,
    305 				entry->rt_flags, (struct rtentry **)0);
    306 	rt_missmsg((req == RTM_ADD ? RTM_OLDADD : RTM_OLDDEL),
    307 		   &(entry->rt_dst), &(entry->rt_gateway),
    308 		   netmask, SA(0), entry->rt_flags, error);
    309 	return (error);
    310 #endif
    311 }
    312 
    313 struct ifaddr *
    314 ifa_ifwithroute(flags, dst, gateway)
    315 int	flags;
    316 struct sockaddr	*dst, *gateway;
    317 {
    318 	register struct ifaddr *ifa;
    319 	if ((flags & RTF_GATEWAY) == 0) {
    320 		/*
    321 		 * If we are adding a route to an interface,
    322 		 * and the interface is a pt to pt link
    323 		 * we should search for the destination
    324 		 * as our clue to the interface.  Otherwise
    325 		 * we can use the local address.
    326 		 */
    327 		ifa = 0;
    328 		if (flags & RTF_HOST)
    329 			ifa = ifa_ifwithdstaddr(dst);
    330 		if (ifa == 0)
    331 			ifa = ifa_ifwithaddr(gateway);
    332 	} else {
    333 		/*
    334 		 * If we are adding a route to a remote net
    335 		 * or host, the gateway may still be on the
    336 		 * other end of a pt to pt link.
    337 		 */
    338 		ifa = ifa_ifwithdstaddr(gateway);
    339 	}
    340 	if (ifa == 0)
    341 		ifa = ifa_ifwithnet(gateway);
    342 	if (ifa == 0) {
    343 		struct rtentry *rt = rtalloc1(dst, 0);
    344 		if (rt == 0)
    345 			return (0);
    346 		rt->rt_refcnt--;
    347 		if ((ifa = rt->rt_ifa) == 0)
    348 			return (0);
    349 	}
    350 	if (ifa->ifa_addr->sa_family != dst->sa_family) {
    351 		struct ifaddr *oifa = ifa, *ifaof_ifpforaddr();
    352 		ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
    353 		if (ifa == 0)
    354 			ifa = oifa;
    355 	}
    356 	return (ifa);
    357 }
    358 
    359 #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
    360 
    361 int
    362 rtrequest(req, dst, gateway, netmask, flags, ret_nrt)
    363 	int req, flags;
    364 	struct sockaddr *dst, *gateway, *netmask;
    365 	struct rtentry **ret_nrt;
    366 {
    367 	int s = splnet(), len, error = 0;
    368 	register struct rtentry *rt;
    369 	register struct radix_node *rn;
    370 	register struct radix_node_head *rnh;
    371 	struct ifaddr *ifa, *ifa_ifwithdstaddr();
    372 	struct sockaddr *ndst;
    373 	u_char af = dst->sa_family;
    374 #define senderr(x) { error = x ; goto bad; }
    375 
    376 	if (rtinits_done == 0)
    377 		rtinitheads();
    378 	for (rnh = radix_node_head; rnh && (af != rnh->rnh_af); )
    379 		rnh = rnh->rnh_next;
    380 	if (rnh == 0)
    381 		senderr(ESRCH);
    382 	if (flags & RTF_HOST)
    383 		netmask = 0;
    384 	switch (req) {
    385 	case RTM_DELETE:
    386 		if (ret_nrt && (rt = *ret_nrt)) {
    387 			RTFREE(rt);
    388 			*ret_nrt = 0;
    389 		}
    390 		if ((rn = rn_delete((caddr_t)dst, (caddr_t)netmask,
    391 					rnh->rnh_treetop)) == 0)
    392 			senderr(ESRCH);
    393 		if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
    394 			panic ("rtrequest delete");
    395 		rt = (struct rtentry *)rn;
    396 		rt->rt_flags &= ~RTF_UP;
    397 		if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest)
    398 			ifa->ifa_rtrequest(RTM_DELETE, rt, SA(0));
    399 		rttrash++;
    400 		if (rt->rt_refcnt <= 0)
    401 			rtfree(rt);
    402 		break;
    403 
    404 	case RTM_RESOLVE:
    405 		if (ret_nrt == 0 || (rt = *ret_nrt) == 0)
    406 			senderr(EINVAL);
    407 		ifa = rt->rt_ifa;
    408 		flags = rt->rt_flags & ~RTF_CLONING;
    409 		gateway = rt->rt_gateway;
    410 		if ((netmask = rt->rt_genmask) == 0)
    411 			flags |= RTF_HOST;
    412 		goto makeroute;
    413 
    414 	case RTM_ADD:
    415 		if ((ifa = ifa_ifwithroute(flags, dst, gateway)) == 0)
    416 			senderr(ENETUNREACH);
    417 	makeroute:
    418 		len = sizeof (*rt) + ROUNDUP(gateway->sa_len)
    419 		    + ROUNDUP(dst->sa_len);
    420 		R_Malloc(rt, struct rtentry *, len);
    421 		if (rt == 0)
    422 			senderr(ENOBUFS);
    423 		Bzero(rt, len);
    424 		ndst = (struct sockaddr *)(rt + 1);
    425 		if (netmask) {
    426 			rt_maskedcopy(dst, ndst, netmask);
    427 		} else
    428 			Bcopy(dst, ndst, dst->sa_len);
    429 		rn = rn_addroute((caddr_t)ndst, (caddr_t)netmask,
    430 					rnh->rnh_treetop, rt->rt_nodes);
    431 		if (rn == 0) {
    432 			free((caddr_t)rt, M_RTABLE);
    433 			senderr(EEXIST);
    434 		}
    435 		rt->rt_ifa = ifa;
    436 		rt->rt_ifp = ifa->ifa_ifp;
    437 		rt->rt_flags = RTF_UP | flags;
    438 		rt->rt_gateway = (struct sockaddr *)
    439 					(rn->rn_key + ROUNDUP(dst->sa_len));
    440 		Bcopy(gateway, rt->rt_gateway, gateway->sa_len);
    441 		if (req == RTM_RESOLVE)
    442 			rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */
    443 		if (ifa->ifa_rtrequest)
    444 			ifa->ifa_rtrequest(req, rt, SA(ret_nrt ? *ret_nrt : 0));
    445 		if (ret_nrt) {
    446 			*ret_nrt = rt;
    447 			rt->rt_refcnt++;
    448 		}
    449 		break;
    450 	}
    451 bad:
    452 	splx(s);
    453 	return (error);
    454 }
    455 
    456 void
    457 rt_maskedcopy(src, dst, netmask)
    458 	struct sockaddr *src, *dst, *netmask;
    459 {
    460 	register u_char *cp1 = (u_char *)src;
    461 	register u_char *cp2 = (u_char *)dst;
    462 	register u_char *cp3 = (u_char *)netmask;
    463 	u_char *cplim = cp2 + *cp3;
    464 	u_char *cplim2 = cp2 + *cp1;
    465 
    466 	*cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
    467 	cp3 += 2;
    468 	if (cplim > cplim2)
    469 		cplim = cplim2;
    470 	while (cp2 < cplim)
    471 		*cp2++ = *cp1++ & *cp3++;
    472 	if (cp2 < cplim2)
    473 		bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
    474 }
    475 /*
    476  * Set up a routing table entry, normally
    477  * for an interface.
    478  */
    479 int
    480 rtinit(ifa, cmd, flags)
    481 	register struct ifaddr *ifa;
    482 	int cmd, flags;
    483 {
    484 	register struct rtentry *rt;
    485 	register struct sockaddr *dst;
    486 	register struct sockaddr *deldst;
    487 	struct mbuf *m = 0;
    488 	int error;
    489 
    490 	dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
    491 	if (ifa->ifa_flags & IFA_ROUTE) {
    492 		if ((rt = ifa->ifa_rt) && (rt->rt_flags & RTF_UP) == 0) {
    493 			RTFREE(rt);
    494 			ifa->ifa_rt = 0;
    495 		}
    496 	}
    497 	if (cmd == RTM_DELETE) {
    498 		if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
    499 			m = m_get(M_WAIT, MT_SONAME);
    500 			deldst = mtod(m, struct sockaddr *);
    501 			rt_maskedcopy(dst, deldst, ifa->ifa_netmask);
    502 			dst = deldst;
    503 		}
    504 		if (rt = rtalloc1(dst, 0)) {
    505 			rt->rt_refcnt--;
    506 			if (rt->rt_ifa != ifa) {
    507 				if (m)
    508 					(void) m_free(m);
    509 				return (flags & RTF_HOST ? EHOSTUNREACH
    510 							: ENETUNREACH);
    511 			}
    512 		}
    513 	}
    514 	error = rtrequest(cmd, dst, ifa->ifa_addr, ifa->ifa_netmask,
    515 			flags | ifa->ifa_flags, &ifa->ifa_rt);
    516 	if (m)
    517 		(void) m_free(m);
    518 	if (cmd == RTM_ADD && error == 0 && (rt = ifa->ifa_rt)
    519 						&& rt->rt_ifa != ifa) {
    520 		rt->rt_ifa = ifa;
    521 		rt->rt_ifp = ifa->ifa_ifp;
    522 	}
    523 	return (error);
    524 }
    525