bsd/netinet/kpi_ipfilter.c

   1 /*
   2  * Copyright (c) 2008 Apple Computer, 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
  29 #include <sys/param.h>  /* for definition of NULL */
  30 #include <sys/errno.h>
  31 #include <sys/malloc.h>
  32 #include <sys/socket.h>
  33 #include <sys/mbuf.h>
  34 #include <sys/systm.h>
  35 #include <libkern/OSAtomic.h>
  36
  37 #include <machine/endian.h>
  38
  39 #define _IP_VHL
  40 #include <net/if_var.h>
  41 #include <net/route.h>
  42 #include <net/kpi_protocol.h>
  43
  44 #include <netinet/in_systm.h>
  45 #include <netinet/in.h>
  46 #include <netinet/in_var.h>
  47 #include <netinet6/in6_var.h>
  48 #include <netinet/ip.h>
  49 #include <netinet/ip6.h>
  50 #include <netinet/ip_var.h>
  51 #include <netinet6/ip6_var.h>
  52 #include <netinet/kpi_ipfilter_var.h>
  53
  54
  55 /*
  56  * kipf_lock and kipf_ref protect the linkage of the list of IP filters
  57  * An IP filter can be removed only when kipf_ref is zero
  58  * If an IP filter cannot be removed because kipf_ref is not null, then
  59  * the IP filter is marjed and kipf_delayed_remove is set so that when
  60  * kipf_ref eventually goes down to zero, the IP filter is removed
  61  */
  62 static lck_mtx_t *kipf_lock = 0;
  63 static u_int32_t kipf_ref = 0;
  64 static u_int32_t kipf_delayed_remove = 0;
  65 u_int32_t kipf_count = 0;
  66
  67 __private_extern__ struct ipfilter_list ipv4_filters = TAILQ_HEAD_INITIALIZER(ipv4_filters);
  68 __private_extern__ struct ipfilter_list ipv6_filters = TAILQ_HEAD_INITIALIZER(ipv6_filters);
  69 __private_extern__ struct ipfilter_list tbr_filters = TAILQ_HEAD_INITIALIZER(tbr_filters);
  70
  71 __private_extern__ void
  72 ipf_ref(void)
  73 {
  74         lck_mtx_lock(kipf_lock);
  75     kipf_ref++;
  76         lck_mtx_unlock(kipf_lock);
  77 }
  78
  79 __private_extern__ void
  80 ipf_unref(void)
  81 {
  82         lck_mtx_lock(kipf_lock);
  83
  84     if (kipf_ref == 0)
  85         panic("ipf_unref: kipf_ref == 0\n");
  86
  87     kipf_ref--;
  88     if (kipf_ref == 0 && kipf_delayed_remove != 0) {
  89         struct ipfilter *filter;
  90
  91                 while ((filter = TAILQ_FIRST(&tbr_filters))) {
  92                         ipf_detach_func ipf_detach = filter->ipf_filter.ipf_detach;
  93                         void* cookie = filter->ipf_filter.cookie;
  94
  95                         TAILQ_REMOVE(filter->ipf_head, filter, ipf_link);
  96                         TAILQ_REMOVE(&tbr_filters, filter, ipf_tbr);
  97                         kipf_delayed_remove--;
  98
  99                         if (ipf_detach) {
 100                                 lck_mtx_unlock(kipf_lock);
 101                                 ipf_detach(cookie);
 102                                 lck_mtx_lock(kipf_lock);
 103                                 /* In case some filter got to run while we released the lock */
 104                                 if (kipf_ref != 0)
 105                                         break;
 106                         }
 107                 }
 108         }
 109         lck_mtx_unlock(kipf_lock);
 110 }
 111
 112 static errno_t
 113 ipf_add(
 114         const struct ipf_filter* filter,
 115         ipfilter_t *filter_ref,
 116         struct ipfilter_list *head)
 117 {
 118         struct ipfilter *new_filter;
 119         if (filter->name == NULL || (filter->ipf_input == NULL && filter->ipf_output == NULL))
 120                 return EINVAL;
 121
 122         MALLOC(new_filter, struct ipfilter*, sizeof(*new_filter), M_IFADDR, M_WAITOK);
 123         if (new_filter == NULL)
 124                 return ENOMEM;
 125
 126         lck_mtx_lock(kipf_lock);
 127         new_filter->ipf_filter = *filter;
 128         new_filter->ipf_head = head;
 129
 130         TAILQ_INSERT_HEAD(head, new_filter, ipf_link);
 131
 132         lck_mtx_unlock(kipf_lock);
 133
 134         *filter_ref = (ipfilter_t)new_filter;
 135
 136         /* This will force TCP to re-evaluate its use of TSO */
 137         OSAddAtomic(1, &kipf_count);
 138         if (use_routegenid)
 139                 routegenid_update();
 140
 141         return 0;
 142 }
 143
 144 errno_t
 145 ipf_addv4(
 146         const struct ipf_filter* filter,
 147         ipfilter_t *filter_ref)
 148 {
 149         return ipf_add(filter, filter_ref, &ipv4_filters);
 150 }
 151
 152 errno_t
 153 ipf_addv6(
 154         const struct ipf_filter* filter,
 155         ipfilter_t *filter_ref)
 156 {
 157         return ipf_add(filter, filter_ref, &ipv6_filters);
 158 }
 159
 160 errno_t
 161 ipf_remove(
 162         ipfilter_t filter_ref)
 163 {
 164         struct ipfilter *match = (struct ipfilter*)filter_ref;
 165         struct ipfilter_list *head;
 166
 167         if (match == 0 || (match->ipf_head != &ipv4_filters && match->ipf_head != &ipv6_filters))
 168                 return EINVAL;
 169
 170         head = match->ipf_head;
 171
 172         lck_mtx_lock(kipf_lock);
 173         TAILQ_FOREACH(match, head, ipf_link) {
 174                 if (match == (struct ipfilter*)filter_ref) {
 175                         ipf_detach_func ipf_detach = match->ipf_filter.ipf_detach;
 176                         void* cookie = match->ipf_filter.cookie;
 177
 178                         /*
 179                          * Cannot detach when they are filters running
 180                          */
 181                         if (kipf_ref) {
 182                                 kipf_delayed_remove++;
 183                                 TAILQ_INSERT_TAIL(&tbr_filters, match, ipf_tbr);
 184                                 match->ipf_filter.ipf_input = 0;
 185                                 match->ipf_filter.ipf_output = 0;
 186                                 lck_mtx_unlock(kipf_lock);
 187                         } else {
 188                                 TAILQ_REMOVE(head, match, ipf_link);
 189                                 lck_mtx_unlock(kipf_lock);
 190                                 if (ipf_detach)
 191                                         ipf_detach(cookie);
 192                                 FREE(match, M_IFADDR);
 193
 194                                 /* This will force TCP to re-evaluate its use of TSO */
 195                                 OSAddAtomic(-1, &kipf_count);
 196                                 if (use_routegenid)
 197                                         routegenid_update();
 198
 199                         }
 200                         return 0;
 201                 }
 202         }
 203         lck_mtx_unlock(kipf_lock);
 204
 205         return ENOENT;
 206 }
 207
 208 int log_for_en1 = 0;
 209
 210 errno_t
 211 ipf_inject_input(
 212         mbuf_t data,
 213         ipfilter_t filter_ref)
 214 {
 215         struct mbuf     *m = (struct mbuf*)data;
 216         struct m_tag *mtag = 0;
 217         struct ip *ip = mtod(m, struct ip *);
 218         u_int8_t        vers;
 219         int hlen;
 220         errno_t error = 0;
 221         protocol_family_t proto;
 222
 223         vers = IP_VHL_V(ip->ip_vhl);
 224
 225         switch (vers) {
 226                 case 4:
 227                         proto = PF_INET;
 228                         break;
 229                 case 6:
 230                         proto = PF_INET6;
 231                         break;
 232                 default:
 233                         error = ENOTSUP;
 234                         goto done;
 235         }
 236
 237         if (filter_ref == 0 && m->m_pkthdr.rcvif == 0) {
 238                 m->m_pkthdr.rcvif = ifunit("lo0");
 239                 m->m_pkthdr.csum_data = 0;
 240                 m->m_pkthdr.csum_flags = 0;
 241                 if (vers == 4) {
 242                         hlen = IP_VHL_HL(ip->ip_vhl) << 2;
 243                         ip->ip_sum = 0;
 244                         ip->ip_sum = in_cksum(m, hlen);
 245                 }
 246         }
 247         if (filter_ref != 0) {
 248                 mtag = m_tag_alloc(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_IPFILT,
 249                                                    sizeof (ipfilter_t), M_NOWAIT);
 250                 if (mtag == NULL) {
 251                         error = ENOMEM;
 252                         goto done;
 253                 }
 254                 *(ipfilter_t*)(mtag+1) = filter_ref;
 255                 m_tag_prepend(m, mtag);
 256         }
 257
 258         error = proto_inject(proto, data);
 259
 260 done:
 261         return error;
 262 }
 263
 264 static errno_t
 265 ipf_injectv4_out(
 266         mbuf_t data,
 267         ipfilter_t filter_ref,
 268         ipf_pktopts_t options)
 269 {
 270         struct route ro;
 271         struct sockaddr_in      *sin = (struct sockaddr_in*)&ro.ro_dst;
 272         struct ip       *ip;
 273         struct mbuf     *m = (struct mbuf*)data;
 274         errno_t error = 0;
 275         struct m_tag *mtag = 0;
 276         struct ip_moptions *imo = 0, ip_moptions;
 277
 278         /* Make the IP header contiguous in the mbuf */
 279         if ((size_t)m->m_len < sizeof(struct ip)) {
 280                 m = m_pullup(m, sizeof(struct ip));
 281                 if (m == NULL) return ENOMEM;
 282         }
 283         ip = (struct ip*)m_mtod(m);
 284
 285         if (filter_ref != 0) {
 286                 mtag = m_tag_alloc(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_IPFILT,
 287                                                    sizeof (ipfilter_t), M_NOWAIT);
 288                 if (mtag == NULL) {
 289                         m_freem(m);
 290                         return ENOMEM;
 291                 }
 292                 *(ipfilter_t*)(mtag+1) = filter_ref;
 293                 m_tag_prepend(m, mtag);
 294         }
 295
 296         if (options && (options->ippo_flags & IPPOF_MCAST_OPTS)) {
 297                 imo = &ip_moptions;
 298
 299                 bzero(imo, sizeof(struct ip6_moptions));
 300                 imo->imo_multicast_ifp = options->ippo_mcast_ifnet;
 301                 imo->imo_multicast_ttl = options->ippo_mcast_ttl;
 302                 imo->imo_multicast_loop = options->ippo_mcast_loop;
 303         }
 304
 305         /* Fill out a route structure and get a route */
 306         bzero(&ro, sizeof(struct route));
 307         sin->sin_len = sizeof(struct sockaddr_in);
 308         sin->sin_family = AF_INET;
 309         sin->sin_port = 0;
 310         sin->sin_addr = ip->ip_dst;
 311         rtalloc(&ro);
 312         if (ro.ro_rt == NULL) {
 313                 m_freem(m);
 314                 return ENETUNREACH;
 315         }
 316
 317         /* Put ip_len and ip_off in host byte order, ip_output expects that */
 318
 319 #if BYTE_ORDER != BIG_ENDIAN
 320         NTOHS(ip->ip_len);
 321         NTOHS(ip->ip_off);
 322 #endif
 323
 324         /* Send  */
 325         error = ip_output(m, NULL, &ro, IP_ALLOWBROADCAST | IP_RAWOUTPUT, imo, NULL);
 326
 327         /* Release the route */
 328         if (ro.ro_rt)
 329                 rtfree(ro.ro_rt);
 330
 331         return error;
 332 }
 333
 334 #if INET6
 335 static errno_t
 336 ipf_injectv6_out(
 337         mbuf_t data,
 338         ipfilter_t filter_ref,
 339         ipf_pktopts_t options)
 340 {
 341         struct route_in6 ro;
 342         struct sockaddr_in6     *sin6 = &ro.ro_dst;
 343         struct ip6_hdr  *ip6;
 344         struct mbuf     *m = (struct mbuf*)data;
 345         errno_t error = 0;
 346         struct m_tag *mtag = 0;
 347         struct ip6_moptions *im6o = 0, ip6_moptions;
 348
 349         /* Make the IP header contiguous in the mbuf */
 350         if ((size_t)m->m_len < sizeof(struct ip6_hdr)) {
 351                 m = m_pullup(m, sizeof(struct ip6_hdr));
 352                 if (m == NULL) return ENOMEM;
 353         }
 354         ip6 = (struct ip6_hdr*)m_mtod(m);
 355
 356         if (filter_ref != 0) {
 357                 mtag = m_tag_alloc(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_IPFILT,
 358                                                    sizeof (ipfilter_t), M_NOWAIT);
 359                 if (mtag == NULL) {
 360                         m_freem(m);
 361                         return ENOMEM;
 362                 }
 363                 *(ipfilter_t*)(mtag+1) = filter_ref;
 364                 m_tag_prepend(m, mtag);
 365         }
 366
 367         if (options && (options->ippo_flags & IPPOF_MCAST_OPTS)) {
 368                 im6o = &ip6_moptions;
 369
 370                 bzero(im6o, sizeof(struct ip6_moptions));
 371                 im6o->im6o_multicast_ifp = options->ippo_mcast_ifnet;
 372                 im6o->im6o_multicast_hlim = options->ippo_mcast_ttl;
 373                 im6o->im6o_multicast_loop = options->ippo_mcast_loop;
 374         }
 375
 376
 377         /* Fill out a route structure and get a route */
 378         bzero(&ro, sizeof(struct route_in6));
 379         sin6->sin6_len = sizeof(struct sockaddr_in6);
 380         sin6->sin6_family = AF_INET6;
 381         sin6->sin6_addr = ip6->ip6_dst;
 382 #if 0
 383         /* This is breaks loopback multicast! */
 384         /* The scope ID should already at s6_addr16[1] */
 385         if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
 386                 /* Hack, pull the scope_id out of the dest addr */
 387                 sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
 388                 ip6->ip6_dst.s6_addr16[1] = 0;
 389         } else
 390                 sin6->sin6_scope_id = 0;
 391 #endif
 392         rtalloc((struct route*)&ro);
 393         if (ro.ro_rt == NULL) {
 394                 m_freem(m);
 395                 return ENETUNREACH;
 396         }
 397
 398         /* Send  */
 399         error = ip6_output(m, NULL, &ro, 0, im6o, NULL, 0);
 400
 401         /* Release the route */
 402         if (ro.ro_rt)
 403                 rtfree(ro.ro_rt);
 404
 405         return error;
 406 }
 407 #endif /* INET6 */
 408
 409 errno_t
 410 ipf_inject_output(
 411         mbuf_t data,
 412         ipfilter_t filter_ref,
 413         ipf_pktopts_t options)
 414 {
 415         struct mbuf     *m = (struct mbuf*)data;
 416         u_int8_t        vers;
 417         errno_t         error = 0;
 418
 419         /* Make one byte of the header contiguous in the mbuf */
 420         if (m->m_len < 1) {
 421                 m = m_pullup(m, 1);
 422                 if (m == NULL)
 423                         goto done;
 424         }
 425
 426         vers = (*(u_int8_t*)m_mtod(m)) >> 4;
 427         switch (vers)
 428         {
 429                 case 4:
 430                         error = ipf_injectv4_out(data, filter_ref, options);
 431                         break;
 432 #if INET6
 433                 case 6:
 434                         error = ipf_injectv6_out(data, filter_ref, options);
 435                         break;
 436 #endif
 437                 default:
 438                         m_freem(m);
 439                         error = ENOTSUP;
 440                         break;
 441         }
 442
 443 done:
 444         return error;
 445 }
 446
 447 __private_extern__ ipfilter_t
 448 ipf_get_inject_filter(struct mbuf *m)
 449 {
 450         ipfilter_t filter_ref = 0;
 451         struct m_tag *mtag;
 452
 453         mtag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_IPFILT, NULL);
 454         if (mtag) {
 455                 filter_ref = *(ipfilter_t *)(mtag+1);
 456
 457                 m_tag_delete(m, mtag);
 458         }
 459         return filter_ref;
 460 }
 461
 462 __private_extern__ int
 463 ipf_init(void)
 464 {
 465         int error = 0;
 466         lck_grp_attr_t *grp_attributes = 0;
 467         lck_attr_t *lck_attributes = 0;
 468         lck_grp_t *lck_grp = 0;
 469
 470         grp_attributes = lck_grp_attr_alloc_init();
 471         if (grp_attributes == 0) {
 472                 printf("ipf_init: lck_grp_attr_alloc_init failed\n");
 473                 error = ENOMEM;
 474                 goto done;
 475         }
 476
 477         lck_grp = lck_grp_alloc_init("IP Filter", grp_attributes);
 478         if (lck_grp == 0) {
 479                 printf("ipf_init: lck_grp_alloc_init failed\n");
 480                 error = ENOMEM;
 481                 goto done;
 482         }
 483
 484         lck_attributes = lck_attr_alloc_init();
 485         if (lck_attributes == 0) {
 486                 printf("ipf_init: lck_attr_alloc_init failed\n");
 487                 error = ENOMEM;
 488                 goto done;
 489         }
 490
 491         kipf_lock = lck_mtx_alloc_init(lck_grp, lck_attributes);
 492         if (kipf_lock == 0) {
 493                 printf("ipf_init: lck_mtx_alloc_init failed\n");
 494                 error = ENOMEM;
 495                 goto done;
 496         }
 497         done:
 498         if (error != 0) {
 499                 if (kipf_lock) {
 500                         lck_mtx_free(kipf_lock, lck_grp);
 501                         kipf_lock = 0;
 502                 }
 503         }
 504         if (lck_grp) {
 505                 lck_grp_free(lck_grp);
 506                 lck_grp = 0;
 507         }
 508         if (grp_attributes) {
 509                 lck_grp_attr_free(grp_attributes);
 510                 grp_attributes = 0;
 511         }
 512         if (lck_attributes) {
 513                 lck_attr_free(lck_attributes);
 514                 lck_attributes = 0;
 515         }
 516
 517         return error;
 518 }