bsd/netinet/ip_encap.c

   1 /*
   2  * Copyright (c) 2000-2016 Apple Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. The rights granted to you under the License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28 /*      $FreeBSD: src/sys/netinet/ip_encap.c,v 1.1.2.2 2001/07/03 11:01:46 ume Exp $    */
  29 /*      $KAME: ip_encap.c,v 1.41 2001/03/15 08:35:08 itojun Exp $       */
  30
  31 /*
  32  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
  33  * All rights reserved.
  34  *
  35  * Redistribution and use in source and binary forms, with or without
  36  * modification, are permitted provided that the following conditions
  37  * are met:
  38  * 1. Redistributions of source code must retain the above copyright
  39  *    notice, this list of conditions and the following disclaimer.
  40  * 2. Redistributions in binary form must reproduce the above copyright
  41  *    notice, this list of conditions and the following disclaimer in the
  42  *    documentation and/or other materials provided with the distribution.
  43  * 3. Neither the name of the project nor the names of its contributors
  44  *    may be used to endorse or promote products derived from this software
  45  *    without specific prior written permission.
  46  *
  47  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  48  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  49  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  50  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  51  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  52  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  53  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  54  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  55  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  56  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  57  * SUCH DAMAGE.
  58  */
  59 /*
  60  * My grandfather said that there's a devil inside tunnelling technology...
  61  *
  62  * We have surprisingly many protocols that want packets with IP protocol
  63  * #4 or #41.  Here's a list of protocols that want protocol #41:
  64  *      RFC1933 configured tunnel
  65  *      RFC1933 automatic tunnel
  66  *      RFC2401 IPsec tunnel
  67  *      RFC2473 IPv6 generic packet tunnelling
  68  *      RFC2529 6over4 tunnel
  69  *      mobile-ip6 (uses RFC2473)
  70  *      6to4 tunnel
  71  * Here's a list of protocol that want protocol #4:
  72  *      RFC1853 IPv4-in-IPv4 tunnelling
  73  *      RFC2003 IPv4 encapsulation within IPv4
  74  *      RFC2344 reverse tunnelling for mobile-ip4
  75  *      RFC2401 IPsec tunnel
  76  * Well, what can I say.  They impose different en/decapsulation mechanism
  77  * from each other, so they need separate protocol handler.  The only one
  78  * we can easily determine by protocol # is IPsec, which always has
  79  * AH/ESP header right after outer IP header.
  80  *
  81  * So, clearly good old protosw does not work for protocol #4 and #41.
  82  * The code will let you match protocol via src/dst address pair.
  83  */
  84 /* XXX is M_NETADDR correct? */
  85
  86 #include <sys/param.h>
  87 #include <sys/systm.h>
  88 #include <sys/socket.h>
  89 #include <sys/sockio.h>
  90 #include <sys/mbuf.h>
  91 #include <sys/mcache.h>
  92 #include <sys/errno.h>
  93 #include <sys/domain.h>
  94 #include <sys/protosw.h>
  95 #include <sys/queue.h>
  96
  97 #include <net/if.h>
  98 #include <net/route.h>
  99
 100 #include <netinet/in.h>
 101 #include <netinet/in_systm.h>
 102 #include <netinet/ip.h>
 103 #include <netinet/ip_var.h>
 104 #include <netinet/ip_encap.h>
 105
 106 #include <netinet/ip6.h>
 107 #include <netinet6/ip6_var.h>
 108 #include <netinet6/ip6protosw.h>
 109
 110 #include <net/net_osdep.h>
 111
 112 #ifndef __APPLE__
 113 #include <sys/kernel.h>
 114 #include <sys/malloc.h>
 115 MALLOC_DEFINE(M_NETADDR, "Export Host", "Export host address structure");
 116 #endif
 117
 118 static void encap_init(struct protosw *, struct domain *);
 119 static void encap_add_locked(struct encaptab *);
 120 static int mask_match(const struct encaptab *, const struct sockaddr *,
 121     const struct sockaddr *);
 122 static void encap_fillarg(struct mbuf *, void *arg);
 123
 124 #ifndef LIST_HEAD_INITIALIZER
 125 /* rely upon BSS initialization */
 126 LIST_HEAD(, encaptab) encaptab;
 127 #else
 128 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
 129 #endif
 130
 131 decl_lck_rw_data(static, encaptab_lock);
 132
 133 static void
 134 encap_init(struct protosw *pp, struct domain *dp)
 135 {
 136 #pragma unused(dp)
 137         static int encap_initialized = 0;
 138         lck_grp_attr_t *encaptab_grp_attrib = NULL;
 139         lck_attr_t *encaptab_lck_attrib = NULL;
 140         lck_grp_t *encaptab_lck_group = NULL;
 141
 142         VERIFY((pp->pr_flags & (PR_INITIALIZED | PR_ATTACHED)) == PR_ATTACHED);
 143
 144         /* This gets called by more than one protocols, so initialize once */
 145         if (encap_initialized) {
 146                 return;
 147         }
 148
 149         encaptab_grp_attrib = lck_grp_attr_alloc_init();
 150         encaptab_lck_group = lck_grp_alloc_init("encaptab lock", encaptab_grp_attrib);
 151         lck_grp_attr_free(encaptab_grp_attrib);
 152
 153         encaptab_lck_attrib = lck_attr_alloc_init();
 154         lck_rw_init(&encaptab_lock, encaptab_lck_group, encaptab_lck_attrib);
 155
 156         lck_grp_free(encaptab_lck_group);
 157         lck_attr_free(encaptab_lck_attrib);
 158
 159         encap_initialized = 1;
 160 #if 0
 161         /*
 162          * we cannot use LIST_INIT() here, since drivers may want to call
 163          * encap_attach(), on driver attach.  encap_init() will be called
 164          * on AF_INET{,6} initialization, which happens after driver
 165          * initialization - using LIST_INIT() here can nuke encap_attach()
 166          * from drivers.
 167          */
 168         LIST_INIT(&encaptab);
 169 #endif
 170 }
 171
 172 void
 173 encap4_init(struct protosw *pp, struct domain *dp)
 174 {
 175         encap_init(pp, dp);
 176 }
 177
 178 void
 179 encap6_init(struct ip6protosw *pp, struct domain *dp)
 180 {
 181         encap_init((struct protosw *)pp, dp);
 182 }
 183
 184 #if INET
 185 void
 186 encap4_input(struct mbuf *m, int off)
 187 {
 188         int proto;
 189         struct ip *ip;
 190         struct sockaddr_in s, d;
 191         const struct protosw *psw;
 192         struct encaptab *ep, *match;
 193         int prio, matchprio;
 194         void *match_arg = NULL;
 195
 196 #ifndef __APPLE__
 197         va_start(ap, m);
 198         off = va_arg(ap, int);
 199         proto = va_arg(ap, int);
 200         va_end(ap);
 201 #endif
 202
 203         /* Expect 32-bit aligned data pointer on strict-align platforms */
 204         MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
 205
 206         ip = mtod(m, struct ip *);
 207 #ifdef __APPLE__
 208         proto = ip->ip_p;
 209 #endif
 210
 211         bzero(&s, sizeof(s));
 212         s.sin_family = AF_INET;
 213         s.sin_len = sizeof(struct sockaddr_in);
 214         s.sin_addr = ip->ip_src;
 215         bzero(&d, sizeof(d));
 216         d.sin_family = AF_INET;
 217         d.sin_len = sizeof(struct sockaddr_in);
 218         d.sin_addr = ip->ip_dst;
 219
 220         match = NULL;
 221         matchprio = 0;
 222
 223         lck_rw_lock_shared(&encaptab_lock);
 224         for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
 225                 if (ep->af != AF_INET) {
 226                         continue;
 227                 }
 228                 if (ep->proto >= 0 && ep->proto != proto) {
 229                         continue;
 230                 }
 231                 if (ep->func) {
 232                         prio = (*ep->func)(m, off, proto, ep->arg);
 233                 } else {
 234                         /*
 235                          * it's inbound traffic, we need to match in reverse
 236                          * order
 237                          */
 238                         prio = mask_match(ep, (struct sockaddr *)&d,
 239                             (struct sockaddr *)&s);
 240                 }
 241
 242                 /*
 243                  * We prioritize the matches by using bit length of the
 244                  * matches.  mask_match() and user-supplied matching function
 245                  * should return the bit length of the matches (for example,
 246                  * if both src/dst are matched for IPv4, 64 should be returned).
 247                  * 0 or negative return value means "it did not match".
 248                  *
 249                  * The question is, since we have two "mask" portion, we
 250                  * cannot really define total order between entries.
 251                  * For example, which of these should be preferred?
 252                  * mask_match() returns 48 (32 + 16) for both of them.
 253                  *      src=3ffe::/16, dst=3ffe:501::/32
 254                  *      src=3ffe:501::/32, dst=3ffe::/16
 255                  *
 256                  * We need to loop through all the possible candidates
 257                  * to get the best match - the search takes O(n) for
 258                  * n attachments (i.e. interfaces).
 259                  */
 260                 if (prio <= 0) {
 261                         continue;
 262                 }
 263                 if (prio > matchprio) {
 264                         matchprio = prio;
 265                         match = ep;
 266                         psw = (const struct protosw *)match->psw;
 267                         match_arg = ep->arg;
 268                 }
 269         }
 270         lck_rw_unlock_shared(&encaptab_lock);
 271
 272         if (match) {
 273                 /* found a match, "match" has the best one */
 274                 if (psw && psw->pr_input) {
 275                         encap_fillarg(m, match_arg);
 276                         (*psw->pr_input)(m, off);
 277                 } else {
 278                         m_freem(m);
 279                 }
 280                 return;
 281         }
 282
 283         /* last resort: inject to raw socket */
 284         rip_input(m, off);
 285 }
 286 #endif
 287
 288 int
 289 encap6_input(struct mbuf **mp, int *offp, int proto)
 290 {
 291         struct mbuf *m = *mp;
 292         struct ip6_hdr *ip6;
 293         struct sockaddr_in6 s, d;
 294         const struct ip6protosw *psw;
 295         struct encaptab *ep, *match;
 296         int prio, matchprio;
 297         void *match_arg = NULL;
 298
 299         /* Expect 32-bit aligned data pointer on strict-align platforms */
 300         MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m);
 301
 302         ip6 = mtod(m, struct ip6_hdr *);
 303         bzero(&s, sizeof(s));
 304         s.sin6_family = AF_INET6;
 305         s.sin6_len = sizeof(struct sockaddr_in6);
 306         s.sin6_addr = ip6->ip6_src;
 307         bzero(&d, sizeof(d));
 308         d.sin6_family = AF_INET6;
 309         d.sin6_len = sizeof(struct sockaddr_in6);
 310         d.sin6_addr = ip6->ip6_dst;
 311
 312         match = NULL;
 313         matchprio = 0;
 314
 315         lck_rw_lock_shared(&encaptab_lock);
 316         for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
 317                 if (ep->af != AF_INET6) {
 318                         continue;
 319                 }
 320                 if (ep->proto >= 0 && ep->proto != proto) {
 321                         continue;
 322                 }
 323                 if (ep->func) {
 324                         prio = (*ep->func)(m, *offp, proto, ep->arg);
 325                 } else {
 326                         /*
 327                          * it's inbound traffic, we need to match in reverse
 328                          * order
 329                          */
 330                         prio = mask_match(ep, (struct sockaddr *)&d,
 331                             (struct sockaddr *)&s);
 332                 }
 333
 334                 /* see encap4_input() for issues here */
 335                 if (prio <= 0) {
 336                         continue;
 337                 }
 338                 if (prio > matchprio) {
 339                         matchprio = prio;
 340                         match = ep;
 341                         psw = (const struct ip6protosw *)match->psw;
 342                         match_arg = ep->arg;
 343                 }
 344         }
 345         lck_rw_unlock_shared(&encaptab_lock);
 346
 347         if (match) {
 348                 /* found a match */
 349                 if (psw && psw->pr_input) {
 350                         encap_fillarg(m, match_arg);
 351                         return (*psw->pr_input)(mp, offp, proto);
 352                 } else {
 353                         m_freem(m);
 354                         return IPPROTO_DONE;
 355                 }
 356         }
 357
 358         /* last resort: inject to raw socket */
 359         return rip6_input(mp, offp, proto);
 360 }
 361
 362 static void
 363 encap_add_locked(struct encaptab *ep)
 364 {
 365         LCK_RW_ASSERT(&encaptab_lock, LCK_RW_ASSERT_EXCLUSIVE);
 366         LIST_INSERT_HEAD(&encaptab, ep, chain);
 367 }
 368
 369 /*
 370  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
 371  * length of mask (sm and dm) is assumed to be same as sp/dp.
 372  * Return value will be necessary as input (cookie) for encap_detach().
 373  */
 374 const struct encaptab *
 375 encap_attach(int af, int proto, const struct sockaddr *sp,
 376     const struct sockaddr *sm, const struct sockaddr *dp,
 377     const struct sockaddr *dm, const struct protosw *psw, void *arg)
 378 {
 379         struct encaptab *ep = NULL;
 380         struct encaptab *new_ep = NULL;
 381         int error;
 382
 383         /* sanity check on args */
 384         if (sp->sa_len > sizeof(new_ep->src) || dp->sa_len > sizeof(new_ep->dst)) {
 385                 error = EINVAL;
 386                 goto fail;
 387         }
 388         if (sp->sa_len != dp->sa_len) {
 389                 error = EINVAL;
 390                 goto fail;
 391         }
 392         if (af != sp->sa_family || af != dp->sa_family) {
 393                 error = EINVAL;
 394                 goto fail;
 395         }
 396
 397         new_ep = _MALLOC(sizeof(*new_ep), M_NETADDR, M_WAITOK | M_ZERO);
 398         if (new_ep == NULL) {
 399                 error = ENOBUFS;
 400                 goto fail;
 401         }
 402
 403         /* check if anyone have already attached with exactly same config */
 404         lck_rw_lock_exclusive(&encaptab_lock);
 405         for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
 406                 if (ep->af != af) {
 407                         continue;
 408                 }
 409                 if (ep->proto != proto) {
 410                         continue;
 411                 }
 412                 if (ep->src.ss_len != sp->sa_len ||
 413                     bcmp(&ep->src, sp, sp->sa_len) != 0 ||
 414                     bcmp(&ep->srcmask, sm, sp->sa_len) != 0) {
 415                         continue;
 416                 }
 417                 if (ep->dst.ss_len != dp->sa_len ||
 418                     bcmp(&ep->dst, dp, dp->sa_len) != 0 ||
 419                     bcmp(&ep->dstmask, dm, dp->sa_len) != 0) {
 420                         continue;
 421                 }
 422
 423                 error = EEXIST;
 424                 goto fail_locked;
 425         }
 426
 427         new_ep->af = af;
 428         new_ep->proto = proto;
 429         bcopy(sp, &new_ep->src, sp->sa_len);
 430         bcopy(sm, &new_ep->srcmask, sp->sa_len);
 431         bcopy(dp, &new_ep->dst, dp->sa_len);
 432         bcopy(dm, &new_ep->dstmask, dp->sa_len);
 433         new_ep->psw = psw;
 434         new_ep->arg = arg;
 435
 436         encap_add_locked(new_ep);
 437         lck_rw_unlock_exclusive(&encaptab_lock);
 438
 439         error = 0;
 440         return new_ep;
 441
 442 fail_locked:
 443         lck_rw_unlock_exclusive(&encaptab_lock);
 444         if (new_ep != NULL) {
 445                 _FREE(new_ep, M_NETADDR);
 446         }
 447 fail:
 448         return NULL;
 449 }
 450
 451 const struct encaptab *
 452 encap_attach_func( int af, int proto,
 453     int (*func)(const struct mbuf *, int, int, void *),
 454     const struct protosw *psw, void *arg)
 455 {
 456         struct encaptab *ep;
 457         int error;
 458
 459         /* sanity check on args */
 460         if (!func) {
 461                 error = EINVAL;
 462                 goto fail;
 463         }
 464
 465         ep = _MALLOC(sizeof(*ep), M_NETADDR, M_WAITOK | M_ZERO); /* XXX */
 466         if (ep == NULL) {
 467                 error = ENOBUFS;
 468                 goto fail;
 469         }
 470
 471         ep->af = af;
 472         ep->proto = proto;
 473         ep->func = func;
 474         ep->psw = psw;
 475         ep->arg = arg;
 476
 477         lck_rw_lock_exclusive(&encaptab_lock);
 478         encap_add_locked(ep);
 479         lck_rw_unlock_exclusive(&encaptab_lock);
 480
 481         error = 0;
 482         return ep;
 483
 484 fail:
 485         return NULL;
 486 }
 487
 488 int
 489 encap_detach(const struct encaptab *cookie)
 490 {
 491         const struct encaptab *ep = cookie;
 492         struct encaptab *p;
 493
 494         lck_rw_lock_exclusive(&encaptab_lock);
 495         for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
 496                 if (p == ep) {
 497                         LIST_REMOVE(p, chain);
 498                         lck_rw_unlock_exclusive(&encaptab_lock);
 499                         _FREE(p, M_NETADDR);    /*XXX*/
 500                         return 0;
 501                 }
 502         }
 503         lck_rw_unlock_exclusive(&encaptab_lock);
 504
 505         return EINVAL;
 506 }
 507
 508 static int
 509 mask_match(const struct encaptab *ep, const struct sockaddr *sp,
 510     const struct sockaddr *dp)
 511 {
 512         struct sockaddr_storage s;
 513         struct sockaddr_storage d;
 514         int i;
 515         const u_int8_t *p, *q;
 516         u_int8_t *r;
 517         int matchlen;
 518
 519         if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d)) {
 520                 return 0;
 521         }
 522         if (sp->sa_family != ep->af || dp->sa_family != ep->af) {
 523                 return 0;
 524         }
 525         if (sp->sa_len != ep->src.ss_len || dp->sa_len != ep->dst.ss_len) {
 526                 return 0;
 527         }
 528
 529         matchlen = 0;
 530
 531         p = (const u_int8_t *)sp;
 532         q = (const u_int8_t *)&ep->srcmask;
 533         r = (u_int8_t *)&s;
 534         for (i = 0; i < sp->sa_len; i++) {
 535                 r[i] = p[i] & q[i];
 536                 /* XXX estimate */
 537                 matchlen += (q[i] ? 8 : 0);
 538         }
 539
 540         p = (const u_int8_t *)dp;
 541         q = (const u_int8_t *)&ep->dstmask;
 542         r = (u_int8_t *)&d;
 543         for (i = 0; i < dp->sa_len; i++) {
 544                 r[i] = p[i] & q[i];
 545                 /* XXX rough estimate */
 546                 matchlen += (q[i] ? 8 : 0);
 547         }
 548
 549         /* need to overwrite len/family portion as we don't compare them */
 550         s.ss_len = sp->sa_len;
 551         s.ss_family = sp->sa_family;
 552         d.ss_len = dp->sa_len;
 553         d.ss_family = dp->sa_family;
 554
 555         if (bcmp(&s, &ep->src, ep->src.ss_len) == 0 &&
 556             bcmp(&d, &ep->dst, ep->dst.ss_len) == 0) {
 557                 return matchlen;
 558         } else {
 559                 return 0;
 560         }
 561 }
 562
 563 struct encaptabtag {
 564         void*                   *arg;
 565 };
 566
 567 static void
 568 encap_fillarg(
 569         struct mbuf *m,
 570         void *arg)
 571 {
 572         struct m_tag    *tag;
 573         struct encaptabtag *et;
 574
 575         tag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_ENCAP,
 576             sizeof(struct encaptabtag), M_WAITOK, m);
 577
 578         if (tag != NULL) {
 579                 et = (struct encaptabtag*)(tag + 1);
 580                 et->arg = arg;
 581                 m_tag_prepend(m, tag);
 582         }
 583 }
 584
 585 void *
 586 encap_getarg(struct mbuf *m)
 587 {
 588         struct m_tag    *tag;
 589         struct encaptabtag *et;
 590         void *p = NULL;
 591
 592         tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_ENCAP, NULL);
 593         if (tag) {
 594                 et = (struct encaptabtag*)(tag + 1);
 595                 p = et->arg;
 596                 m_tag_delete(m, tag);
 597         }
 598
 599         return p;
 600 }