sys/netinet/ip_encap.c

   1 /*      $FreeBSD: src/sys/netinet/ip_encap.c,v 1.1.2.5 2003/01/23 21:06:45 sam Exp $    */
   2 /*      $KAME: ip_encap.c,v 1.41 2001/03/15 08:35:08 itojun Exp $       */
   3
   4 /*
   5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
   6  * All rights reserved.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the name of the project nor the names of its contributors
  17  *    may be used to endorse or promote products derived from this software
  18  *    without specific prior written permission.
  19  *
  20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  30  * SUCH DAMAGE.
  31  */
  32 /*
  33  * My grandfather said that there's a devil inside tunnelling technology...
  34  *
  35  * We have surprisingly many protocols that want packets with IP protocol
  36  * #4 or #41.  Here's a list of protocols that want protocol #41:
  37  *      RFC1933 configured tunnel
  38  *      RFC1933 automatic tunnel
  39  *      RFC2401 IPsec tunnel
  40  *      RFC2473 IPv6 generic packet tunnelling
  41  *      RFC2529 6over4 tunnel
  42  *      mobile-ip6 (uses RFC2473)
  43  *      RFC3056 6to4 tunnel
  44  *      isatap tunnel
  45  * Here's a list of protocol that want protocol #4:
  46  *      RFC1853 IPv4-in-IPv4 tunnelling
  47  *      RFC2003 IPv4 encapsulation within IPv4
  48  *      RFC2344 reverse tunnelling for mobile-ip4
  49  *      RFC2401 IPsec tunnel
  50  * Well, what can I say.  They impose different en/decapsulation mechanism
  51  * from each other, so they need separate protocol handler.  The only one
  52  * we can easily determine by protocol # is IPsec, which always has
  53  * AH/ESP/IPComp header right after outer IP header.
  54  *
  55  * So, clearly good old protosw does not work for protocol #4 and #41.
  56  * The code will let you match protocol via src/dst address pair.
  57  */
  58
  59 #include "opt_inet.h"
  60 #include "opt_inet6.h"
  61
  62 #include <sys/param.h>
  63 #include <sys/systm.h>
  64 #include <sys/socket.h>
  65 #include <sys/sockio.h>
  66 #include <sys/mbuf.h>
  67 #include <sys/errno.h>
  68 #include <sys/protosw.h>
  69 #include <sys/queue.h>
  70
  71 #include <net/if.h>
  72 #include <net/route.h>
  73
  74 #include <netinet/in.h>
  75 #include <netinet/in_systm.h>
  76 #include <netinet/ip.h>
  77 #include <netinet/ip_var.h>
  78 #include <netinet/ip_encap.h>
  79
  80 #ifdef INET6
  81 #include <netinet/ip6.h>
  82 #include <netinet6/ip6_var.h>
  83 #include <netinet6/ip6protosw.h>
  84 #endif
  85
  86 #include <machine/stdarg.h>
  87
  88 #include <net/net_osdep.h>
  89
  90 #include <sys/kernel.h>
  91 #include <sys/malloc.h>
  92 #include <sys/thread2.h>
  93 MALLOC_DEFINE(M_IPENCAP, "IP Encapsulation", "IP Encapsulation");
  94
  95 static void encap_add (struct encaptab *);
  96 static int mask_match (const struct encaptab *, const struct sockaddr *,
  97                 const struct sockaddr *);
  98 static void encap_fillarg (struct mbuf *, const struct encaptab *);
  99
 100 #ifndef LIST_HEAD_INITIALIZER
 101 /* rely upon BSS initialization */
 102 LIST_HEAD(, encaptab) encaptab;
 103 #else
 104 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
 105 #endif
 106
 107 int     (*ipip_input)(struct mbuf **, int *, int); /* hook for mrouting */
 108
 109 void
 110 encap_init(void)
 111 {
 112         static int initialized = 0;
 113
 114         if (initialized)
 115                 return;
 116         initialized++;
 117 #if 0
 118         /*
 119          * we cannot use LIST_INIT() here, since drivers may want to call
 120          * encap_attach(), on driver attach.  encap_init() will be called
 121          * on AF_INET{,6} initialization, which happens after driver
 122          * initialization - using LIST_INIT() here can nuke encap_attach()
 123          * from drivers.
 124          */
 125         LIST_INIT(&encaptab);
 126 #endif
 127 }
 128
 129 #ifdef INET
 130 int
 131 encap4_input(struct mbuf **mp, int *offp, int proto)
 132 {
 133         struct mbuf *m = *mp;
 134         int off = *offp;
 135         struct ip *ip;
 136         struct sockaddr_in s, d;
 137         const struct protosw *psw;
 138         struct encaptab *ep, *match;
 139         int prio, matchprio;
 140
 141         if (!IN_NETISR(0)) {
 142                 /*
 143                  * NOTE:
 144                  * Some NICs, noticeably igb(4) and ix(4), use inner IP
 145                  * datagram to calculate the packet hash, which leads us
 146                  * here.
 147                  */
 148                 m->m_flags &= ~M_HASH;
 149                 m = ip_rehashm(m);
 150                 if (m != NULL) {
 151                         lwkt_port_t port = netisr_hashport(m->m_pkthdr.hash);
 152
 153                         KASSERT(port != &curthread->td_msgport,
 154                             ("mbuf hash recursion"));
 155                         ip_transport_redispatch(port, m, off);
 156                 }
 157                 return (IPPROTO_DONE);
 158         }
 159
 160         ip = mtod(m, struct ip *);
 161         *mp = NULL;
 162
 163         bzero(&s, sizeof s);
 164         s.sin_family = AF_INET;
 165         s.sin_len = sizeof(struct sockaddr_in);
 166         s.sin_addr = ip->ip_src;
 167         bzero(&d, sizeof d);
 168         d.sin_family = AF_INET;
 169         d.sin_len = sizeof(struct sockaddr_in);
 170         d.sin_addr = ip->ip_dst;
 171
 172         match = NULL;
 173         matchprio = 0;
 174         for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
 175                 if (ep->af != AF_INET)
 176                         continue;
 177                 if (ep->proto >= 0 && ep->proto != proto)
 178                         continue;
 179                 if (ep->func)
 180                         prio = (*ep->func)(m, off, proto, ep->arg);
 181                 else {
 182                         /*
 183                          * it's inbound traffic, we need to match in reverse
 184                          * order
 185                          */
 186                         prio = mask_match(ep,
 187                                           (struct sockaddr *)&d,
 188                                           (struct sockaddr *)&s);
 189                 }
 190
 191                 /*
 192                  * We prioritize the matches by using bit length of the
 193                  * matches.  mask_match() and user-supplied matching function
 194                  * should return the bit length of the matches (for example,
 195                  * if both src/dst are matched for IPv4, 64 should be returned).
 196                  * 0 or negative return value means "it did not match".
 197                  *
 198                  * The question is, since we have two "mask" portion, we
 199                  * cannot really define total order between entries.
 200                  * For example, which of these should be preferred?
 201                  * mask_match() returns 48 (32 + 16) for both of them.
 202                  *      src=3ffe::/16, dst=3ffe:501::/32
 203                  *      src=3ffe:501::/32, dst=3ffe::/16
 204                  *
 205                  * We need to loop through all the possible candidates
 206                  * to get the best match - the search takes O(n) for
 207                  * n attachments (i.e. interfaces).
 208                  */
 209                 if (prio <= 0)
 210                         continue;
 211                 if (prio > matchprio) {
 212                         matchprio = prio;
 213                         match = ep;
 214                 }
 215         }
 216
 217         if (match) {
 218                 /* found a match, "match" has the best one */
 219                 psw = match->psw;
 220                 if (psw && psw->pr_input) {
 221                         encap_fillarg(m, match);
 222                         *mp = m;
 223                         (*psw->pr_input)(mp, offp, proto);
 224                 } else {
 225                         m_freem(m);
 226                 }
 227                 return(IPPROTO_DONE);
 228         }
 229
 230         /* for backward compatibility */
 231         if (proto == IPPROTO_IPV4 && ipip_input) {
 232                 *mp = m;
 233                 ipip_input(mp, offp, proto);
 234                 return(IPPROTO_DONE);
 235         }
 236
 237         /* last resort: inject to raw socket */
 238         *mp = m;
 239         rip_input(mp, offp, proto);
 240         return(IPPROTO_DONE);
 241 }
 242 #endif
 243
 244 #ifdef INET6
 245 int
 246 encap6_input(struct mbuf **mp, int *offp, int proto)
 247 {
 248         struct mbuf *m = *mp;
 249         struct ip6_hdr *ip6;
 250         struct sockaddr_in6 s, d;
 251         const struct protosw *psw;
 252         struct encaptab *ep, *match;
 253         int prio, matchprio;
 254
 255         ip6 = mtod(m, struct ip6_hdr *);
 256
 257         bzero(&s, sizeof s);
 258         s.sin6_family = AF_INET6;
 259         s.sin6_len = sizeof(struct sockaddr_in6);
 260         s.sin6_addr = ip6->ip6_src;
 261         bzero(&d, sizeof d);
 262         d.sin6_family = AF_INET6;
 263         d.sin6_len = sizeof(struct sockaddr_in6);
 264         d.sin6_addr = ip6->ip6_dst;
 265
 266         match = NULL;
 267         matchprio = 0;
 268         for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
 269                 if (ep->af != AF_INET6)
 270                         continue;
 271                 if (ep->proto >= 0 && ep->proto != proto)
 272                         continue;
 273                 if (ep->func)
 274                         prio = (*ep->func)(m, *offp, proto, ep->arg);
 275                 else {
 276                         /*
 277                          * it's inbound traffic, we need to match in reverse
 278                          * order
 279                          */
 280                         prio = mask_match(ep, (struct sockaddr *)&d,
 281                             (struct sockaddr *)&s);
 282                 }
 283
 284                 /* see encap4_input() for issues here */
 285                 if (prio <= 0)
 286                         continue;
 287                 if (prio > matchprio) {
 288                         matchprio = prio;
 289                         match = ep;
 290                 }
 291         }
 292
 293         if (match) {
 294                 /* found a match */
 295                 psw = match->psw;
 296                 if (psw && psw->pr_input) {
 297                         encap_fillarg(m, match);
 298                         return (*psw->pr_input)(mp, offp, proto);
 299                 } else {
 300                         m_freem(m);
 301                         return IPPROTO_DONE;
 302                 }
 303         }
 304
 305         /* last resort: inject to raw socket */
 306         return rip6_input(mp, offp, proto);
 307 }
 308 #endif
 309
 310 static void
 311 encap_add(struct encaptab *ep)
 312 {
 313
 314         LIST_INSERT_HEAD(&encaptab, ep, chain);
 315 }
 316
 317 /*
 318  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
 319  * length of mask (sm and dm) is assumed to be same as sp/dp.
 320  * Return value will be necessary as input (cookie) for encap_detach().
 321  */
 322 const struct encaptab *
 323 encap_attach(int af, int proto, const struct sockaddr *sp,
 324              const struct sockaddr *sm, const struct sockaddr *dp,
 325              const struct sockaddr *dm, const struct protosw *psw, void *arg)
 326 {
 327         struct encaptab *ep;
 328
 329         crit_enter();
 330         /* sanity check on args */
 331         if (sp->sa_len > sizeof ep->src || dp->sa_len > sizeof ep->dst)
 332                 goto fail;
 333         if (sp->sa_len != dp->sa_len)
 334                 goto fail;
 335         if (af != sp->sa_family || af != dp->sa_family)
 336                 goto fail;
 337
 338         /* check if anyone have already attached with exactly same config */
 339         for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
 340                 if (ep->af != af)
 341                         continue;
 342                 if (ep->proto != proto)
 343                         continue;
 344                 if (ep->src.ss_len != sp->sa_len ||
 345                     bcmp(&ep->src, sp, sp->sa_len) != 0 ||
 346                     bcmp(&ep->srcmask, sm, sp->sa_len) != 0)
 347                         continue;
 348                 if (ep->dst.ss_len != dp->sa_len ||
 349                     bcmp(&ep->dst, dp, dp->sa_len) != 0 ||
 350                     bcmp(&ep->dstmask, dm, dp->sa_len) != 0)
 351                         continue;
 352
 353                 goto fail;
 354         }
 355
 356         ep = kmalloc(sizeof *ep, M_IPENCAP, M_INTWAIT | M_ZERO | M_NULLOK);
 357         if (ep == NULL)
 358                 goto fail;
 359
 360         ep->af = af;
 361         ep->proto = proto;
 362         bcopy(sp, &ep->src, sp->sa_len);
 363         bcopy(sm, &ep->srcmask, sp->sa_len);
 364         bcopy(dp, &ep->dst, dp->sa_len);
 365         bcopy(dm, &ep->dstmask, dp->sa_len);
 366         ep->psw = psw;
 367         ep->arg = arg;
 368
 369         encap_add(ep);
 370
 371         crit_exit();
 372         return ep;
 373
 374 fail:
 375         crit_exit();
 376         return NULL;
 377 }
 378
 379 const struct encaptab *
 380 encap_attach_func(int af, int proto,
 381                   int (*func)(const struct mbuf *, int, int, void *),
 382                   const struct protosw *psw, void *arg)
 383 {
 384         struct encaptab *ep;
 385
 386         crit_enter();
 387         /* sanity check on args */
 388         if (!func)
 389                 goto fail;
 390
 391         ep = kmalloc(sizeof *ep, M_IPENCAP, M_INTWAIT | M_ZERO | M_NULLOK);
 392         if (ep == NULL)
 393                 goto fail;
 394
 395         ep->af = af;
 396         ep->proto = proto;
 397         ep->func = func;
 398         ep->psw = psw;
 399         ep->arg = arg;
 400
 401         encap_add(ep);
 402
 403         crit_exit();
 404         return ep;
 405
 406 fail:
 407         crit_exit();
 408         return NULL;
 409 }
 410
 411 int
 412 encap_detach(const struct encaptab *cookie)
 413 {
 414         const struct encaptab *ep = cookie;
 415         struct encaptab *p;
 416
 417         for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
 418                 if (p == ep) {
 419                         LIST_REMOVE(p, chain);
 420                         kfree(p, M_IPENCAP);    /*XXX*/
 421                         return 0;
 422                 }
 423         }
 424
 425         return EINVAL;
 426 }
 427
 428 static int
 429 mask_match(const struct encaptab *ep, const struct sockaddr *sp,
 430            const struct sockaddr *dp)
 431 {
 432         struct sockaddr_storage s;
 433         struct sockaddr_storage d;
 434         int i;
 435         const u_int8_t *p, *q;
 436         u_int8_t *r;
 437         int matchlen;
 438
 439         if (sp->sa_len > sizeof s || dp->sa_len > sizeof d)
 440                 return 0;
 441         if (sp->sa_family != ep->af || dp->sa_family != ep->af)
 442                 return 0;
 443         if (sp->sa_len != ep->src.ss_len || dp->sa_len != ep->dst.ss_len)
 444                 return 0;
 445
 446         matchlen = 0;
 447
 448         p = (const u_int8_t *)sp;
 449         q = (const u_int8_t *)&ep->srcmask;
 450         r = (u_int8_t *)&s;
 451         for (i = 0 ; i < sp->sa_len; i++) {
 452                 r[i] = p[i] & q[i];
 453                 /* XXX estimate */
 454                 matchlen += (q[i] ? 8 : 0);
 455         }
 456
 457         p = (const u_int8_t *)dp;
 458         q = (const u_int8_t *)&ep->dstmask;
 459         r = (u_int8_t *)&d;
 460         for (i = 0 ; i < dp->sa_len; i++) {
 461                 r[i] = p[i] & q[i];
 462                 /* XXX rough estimate */
 463                 matchlen += (q[i] ? 8 : 0);
 464         }
 465
 466         /* need to overwrite len/family portion as we don't compare them */
 467         s.ss_len = sp->sa_len;
 468         s.ss_family = sp->sa_family;
 469         d.ss_len = dp->sa_len;
 470         d.ss_family = dp->sa_family;
 471
 472         if (bcmp(&s, &ep->src, ep->src.ss_len) == 0 &&
 473             bcmp(&d, &ep->dst, ep->dst.ss_len) == 0) {
 474                 return matchlen;
 475         } else
 476                 return 0;
 477 }
 478
 479 static void
 480 encap_fillarg(struct mbuf *m, const struct encaptab *ep)
 481 {
 482         struct m_tag *tag;
 483
 484         tag = m_tag_get(PACKET_TAG_ENCAP, sizeof(void *), M_NOWAIT);
 485         if (tag != NULL) {
 486                 *(void **)m_tag_data(tag) = ep->arg;
 487                 m_tag_prepend(m, tag);
 488         }
 489 }
 490
 491 void *
 492 encap_getarg(struct mbuf *m)
 493 {
 494         void *p = NULL;
 495         struct m_tag *tag;
 496
 497         tag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
 498         if (tag != NULL) {
 499                 p = *(void **)m_tag_data(tag);
 500                 m_tag_delete(m, tag);
 501         }
 502         return p;
 503 }