sys/net/pf/pf.c

   1 /*      $FreeBSD: src/sys/contrib/pf/net/pf.c,v 1.19 2004/09/11 11:18:25 mlaier Exp $   */
   2 /*      $OpenBSD: pf.c,v 1.433.2.2 2004/07/17 03:22:34 brad Exp $ */
   3 /* add  $OpenBSD: pf.c,v 1.448 2004/05/11 07:34:11 dhartmei Exp $ */
   4 /*      $DragonFly: src/sys/net/pf/pf.c,v 1.16 2008/04/06 21:12:41 dillon Exp $ */
   5
   6 /*
   7  * Copyright (c) 2004 The DragonFly Project.  All rights reserved.
   8  *
   9  * Copyright (c) 2001 Daniel Hartmeier
  10  * Copyright (c) 2002,2003 Henning Brauer
  11  * All rights reserved.
  12  *
  13  * Redistribution and use in source and binary forms, with or without
  14  * modification, are permitted provided that the following conditions
  15  * are met:
  16  *
  17  *    - Redistributions of source code must retain the above copyright
  18  *      notice, this list of conditions and the following disclaimer.
  19  *    - Redistributions in binary form must reproduce the above
  20  *      copyright notice, this list of conditions and the following
  21  *      disclaimer in the documentation and/or other materials provided
  22  *      with the distribution.
  23  *
  24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  25  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  26  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
  27  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
  28  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
  29  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
  30  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  31  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  32  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  34  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  35  * POSSIBILITY OF SUCH DAMAGE.
  36  *
  37  * Effort sponsored in part by the Defense Advanced Research Projects
  38  * Agency (DARPA) and Air Force Research Laboratory, Air Force
  39  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
  40  *
  41  */
  42
  43 #include "opt_inet.h"
  44 #include "opt_inet6.h"
  45 #include "use_pfsync.h"
  46
  47 #include <sys/param.h>
  48 #include <sys/systm.h>
  49 #include <sys/malloc.h>
  50 #include <sys/mbuf.h>
  51 #include <sys/filio.h>
  52 #include <sys/socket.h>
  53 #include <sys/socketvar.h>
  54 #include <sys/kernel.h>
  55 #include <sys/time.h>
  56 #include <sys/sysctl.h>
  57 #include <sys/endian.h>
  58 #include <vm/vm_zone.h>
  59
  60 #include <machine/inttypes.h>
  61
  62 #include <net/if.h>
  63 #include <net/if_types.h>
  64 #include <net/bpf.h>
  65 #include <net/netisr.h>
  66 #include <net/route.h>
  67
  68 #include <netinet/in.h>
  69 #include <netinet/in_var.h>
  70 #include <netinet/in_systm.h>
  71 #include <netinet/ip.h>
  72 #include <netinet/ip_var.h>
  73 #include <netinet/tcp.h>
  74 #include <netinet/tcp_seq.h>
  75 #include <netinet/udp.h>
  76 #include <netinet/ip_icmp.h>
  77 #include <netinet/in_pcb.h>
  78 #include <netinet/tcp_timer.h>
  79 #include <netinet/tcp_var.h>
  80 #include <netinet/udp_var.h>
  81 #include <netinet/icmp_var.h>
  82
  83 #include <net/pf/pfvar.h>
  84 #include <net/pf/if_pflog.h>
  85
  86 #if NPFSYNC > 0
  87 #include <net/pf/if_pfsync.h>
  88 #endif /* NPFSYNC > 0 */
  89
  90 #ifdef INET6
  91 #include <netinet/ip6.h>
  92 #include <netinet/in_pcb.h>
  93 #include <netinet/icmp6.h>
  94 #include <netinet6/nd6.h>
  95 #include <netinet6/ip6_var.h>
  96 #include <netinet6/in6_pcb.h>
  97 #endif /* INET6 */
  98
  99 #include <sys/in_cksum.h>
 100 #include <sys/ucred.h>
 101 #include <machine/limits.h>
 102 #include <sys/msgport2.h>
 103 #include <net/netmsg2.h>
 104
 105 extern int ip_optcopy(struct ip *, struct ip *);
 106
 107 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
 108
 109 /*
 110  * Global variables
 111  */
 112
 113 struct pf_anchorqueue    pf_anchors;
 114 struct pf_ruleset        pf_main_ruleset;
 115 struct pf_altqqueue      pf_altqs[2];
 116 struct pf_palist         pf_pabuf;
 117 struct pf_altqqueue     *pf_altqs_active;
 118 struct pf_altqqueue     *pf_altqs_inactive;
 119 struct pf_status         pf_status;
 120
 121 u_int32_t                ticket_altqs_active;
 122 u_int32_t                ticket_altqs_inactive;
 123 int                      altqs_inactive_open;
 124 u_int32_t                ticket_pabuf;
 125
 126 struct callout           pf_expire_to;                  /* expire timeout */
 127
 128 vm_zone_t                pf_src_tree_pl, pf_rule_pl;
 129 vm_zone_t                pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
 130
 131 void                     pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
 132 void                     pf_print_state(struct pf_state *);
 133 void                     pf_print_flags(u_int8_t);
 134
 135 u_int16_t                pf_cksum_fixup(u_int16_t, u_int16_t, u_int16_t,
 136                             u_int8_t);
 137 void                     pf_change_ap(struct pf_addr *, u_int16_t *,
 138                             u_int16_t *, u_int16_t *, struct pf_addr *,
 139                             u_int16_t, u_int8_t, sa_family_t);
 140 #ifdef INET6
 141 void                     pf_change_a6(struct pf_addr *, u_int16_t *,
 142                             struct pf_addr *, u_int8_t);
 143 #endif /* INET6 */
 144 void                     pf_change_icmp(struct pf_addr *, u_int16_t *,
 145                             struct pf_addr *, struct pf_addr *, u_int16_t,
 146                             u_int16_t *, u_int16_t *, u_int16_t *,
 147                             u_int16_t *, u_int8_t, sa_family_t);
 148 void                     pf_send_tcp(const struct pf_rule *, sa_family_t,
 149                             const struct pf_addr *, const struct pf_addr *,
 150                             u_int16_t, u_int16_t, u_int32_t, u_int32_t,
 151                             u_int8_t, u_int16_t, u_int16_t, u_int8_t);
 152 void                     pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
 153                             sa_family_t, struct pf_rule *);
 154 struct pf_rule          *pf_match_translation(struct pf_pdesc *, struct mbuf *,
 155                             int, int, struct pfi_kif *,
 156                             struct pf_addr *, u_int16_t, struct pf_addr *,
 157                             u_int16_t, int);
 158 struct pf_rule          *pf_get_translation(struct pf_pdesc *, struct mbuf *,
 159                             int, int, struct pfi_kif *, struct pf_src_node **,
 160                             struct pf_addr *, u_int16_t,
 161                             struct pf_addr *, u_int16_t,
 162                             struct pf_addr *, u_int16_t *);
 163 int                      pf_test_tcp(struct pf_rule **, struct pf_state **,
 164                             int, struct pfi_kif *, struct mbuf *, int,
 165                             void *, struct pf_pdesc *, struct pf_rule **,
 166                             struct pf_ruleset **);
 167 int                      pf_test_udp(struct pf_rule **, struct pf_state **,
 168                             int, struct pfi_kif *, struct mbuf *, int,
 169                             void *, struct pf_pdesc *, struct pf_rule **,
 170                             struct pf_ruleset **);
 171 int                      pf_test_icmp(struct pf_rule **, struct pf_state **,
 172                             int, struct pfi_kif *, struct mbuf *, int,
 173                             void *, struct pf_pdesc *, struct pf_rule **,
 174                             struct pf_ruleset **);
 175 int                      pf_test_other(struct pf_rule **, struct pf_state **,
 176                             int, struct pfi_kif *, struct mbuf *, int, void *,
 177                             struct pf_pdesc *, struct pf_rule **,
 178                             struct pf_ruleset **);
 179 int                      pf_test_fragment(struct pf_rule **, int,
 180                             struct pfi_kif *, struct mbuf *, void *,
 181                             struct pf_pdesc *, struct pf_rule **,
 182                             struct pf_ruleset **);
 183 int                      pf_test_state_tcp(struct pf_state **, int,
 184                             struct pfi_kif *, struct mbuf *, int,
 185                             void *, struct pf_pdesc *, u_short *);
 186 int                      pf_test_state_udp(struct pf_state **, int,
 187                             struct pfi_kif *, struct mbuf *, int,
 188                             void *, struct pf_pdesc *);
 189 int                      pf_test_state_icmp(struct pf_state **, int,
 190                             struct pfi_kif *, struct mbuf *, int,
 191                             void *, struct pf_pdesc *);
 192 int                      pf_test_state_other(struct pf_state **, int,
 193                             struct pfi_kif *, struct pf_pdesc *);
 194 static int               pf_match_tag(struct mbuf *, struct pf_rule *,
 195                                       struct pf_rule *, int *);
 196 void                     pf_hash(struct pf_addr *, struct pf_addr *,
 197                             struct pf_poolhashkey *, sa_family_t);
 198 int                      pf_map_addr(u_int8_t, struct pf_rule *,
 199                             struct pf_addr *, struct pf_addr *,
 200                             struct pf_addr *, struct pf_src_node **);
 201 int                      pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
 202                             struct pf_addr *, struct pf_addr *, u_int16_t,
 203                             struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
 204                             struct pf_src_node **);
 205 void                     pf_route(struct mbuf **, struct pf_rule *, int,
 206                             struct ifnet *, struct pf_state *);
 207 void                     pf_route6(struct mbuf **, struct pf_rule *, int,
 208                             struct ifnet *, struct pf_state *);
 209 int                      pf_socket_lookup(uid_t *, gid_t *,
 210                             int, struct pf_pdesc *);
 211 u_int8_t                 pf_get_wscale(struct mbuf *, int, u_int16_t,
 212                             sa_family_t);
 213 u_int16_t                pf_get_mss(struct mbuf *, int, u_int16_t,
 214                             sa_family_t);
 215 u_int16_t                pf_calc_mss(struct pf_addr *, sa_family_t,
 216                                 u_int16_t);
 217 void                     pf_set_rt_ifp(struct pf_state *,
 218                             struct pf_addr *);
 219 int                      pf_check_proto_cksum(struct mbuf *, int, int,
 220                             u_int8_t, sa_family_t);
 221 int                      pf_addr_wrap_neq(struct pf_addr_wrap *,
 222                             struct pf_addr_wrap *);
 223 struct pf_state         *pf_find_state_recurse(struct pfi_kif *,
 224                             struct pf_state *, u_int8_t);
 225
 226 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX];
 227
 228 #define STATE_LOOKUP()                                                  \
 229         do {                                                            \
 230                 if (direction == PF_IN)                                 \
 231                         *state = pf_find_state_recurse(         \
 232                             kif, &key, PF_EXT_GWY);                     \
 233                 else                                                    \
 234                         *state = pf_find_state_recurse(         \
 235                             kif, &key, PF_LAN_EXT);                     \
 236                 if (*state == NULL)                                     \
 237                         return (PF_DROP);                               \
 238                 if (direction == PF_OUT &&                              \
 239                     (((*state)->rule.ptr->rt == PF_ROUTETO &&           \
 240                     (*state)->rule.ptr->direction == PF_OUT) ||         \
 241                     ((*state)->rule.ptr->rt == PF_REPLYTO &&            \
 242                     (*state)->rule.ptr->direction == PF_IN)) &&         \
 243                     (*state)->rt_kif != NULL &&                         \
 244                     (*state)->rt_kif != kif)                            \
 245                         return (PF_PASS);                               \
 246         } while (0)
 247
 248 #define STATE_TRANSLATE(s) \
 249         (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \
 250         ((s)->af == AF_INET6 && \
 251         ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \
 252         (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \
 253         (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \
 254         (s)->lan.port != (s)->gwy.port
 255
 256 #define BOUND_IFACE(r, k) (((r)->rule_flag & PFRULE_IFBOUND) ? (k) :   \
 257         ((r)->rule_flag & PFRULE_GRBOUND) ? (k)->pfik_parent :         \
 258         (k)->pfik_parent->pfik_parent)
 259
 260 static int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
 261 static int pf_state_compare_lan_ext(struct pf_state *,
 262         struct pf_state *);
 263 static int pf_state_compare_ext_gwy(struct pf_state *,
 264         struct pf_state *);
 265 static int pf_state_compare_id(struct pf_state *,
 266         struct pf_state *);
 267
 268 struct pf_src_tree tree_src_tracking;
 269
 270 struct pf_state_tree_id tree_id;
 271 struct pf_state_queue state_updates;
 272
 273 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
 274 RB_GENERATE(pf_state_tree_lan_ext, pf_state,
 275     u.s.entry_lan_ext, pf_state_compare_lan_ext);
 276 RB_GENERATE(pf_state_tree_ext_gwy, pf_state,
 277     u.s.entry_ext_gwy, pf_state_compare_ext_gwy);
 278 RB_GENERATE(pf_state_tree_id, pf_state,
 279     u.s.entry_id, pf_state_compare_id);
 280
 281 static int
 282 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
 283 {
 284         int     diff;
 285
 286         if (a->rule.ptr > b->rule.ptr)
 287                 return (1);
 288         if (a->rule.ptr < b->rule.ptr)
 289                 return (-1);
 290         if ((diff = a->af - b->af) != 0)
 291                 return (diff);
 292         switch (a->af) {
 293 #ifdef INET
 294         case AF_INET:
 295                 if (a->addr.addr32[0] > b->addr.addr32[0])
 296                         return (1);
 297                 if (a->addr.addr32[0] < b->addr.addr32[0])
 298                         return (-1);
 299                 break;
 300 #endif /* INET */
 301 #ifdef INET6
 302         case AF_INET6:
 303                 if (a->addr.addr32[3] > b->addr.addr32[3])
 304                         return (1);
 305                 if (a->addr.addr32[3] < b->addr.addr32[3])
 306                         return (-1);
 307                 if (a->addr.addr32[2] > b->addr.addr32[2])
 308                         return (1);
 309                 if (a->addr.addr32[2] < b->addr.addr32[2])
 310                         return (-1);
 311                 if (a->addr.addr32[1] > b->addr.addr32[1])
 312                         return (1);
 313                 if (a->addr.addr32[1] < b->addr.addr32[1])
 314                         return (-1);
 315                 if (a->addr.addr32[0] > b->addr.addr32[0])
 316                         return (1);
 317                 if (a->addr.addr32[0] < b->addr.addr32[0])
 318                         return (-1);
 319                 break;
 320 #endif /* INET6 */
 321         }
 322         return (0);
 323 }
 324
 325 static
 326 int
 327 pf_state_hash(struct pf_state *s)
 328 {
 329         int hv = (intptr_t)s / sizeof(*s);
 330
 331         hv ^= crc32(&s->lan, sizeof(s->lan));
 332         hv ^= crc32(&s->gwy, sizeof(s->gwy));
 333         hv ^= crc32(&s->ext, sizeof(s->ext));
 334         return(hv);
 335 }
 336
 337 static int
 338 pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b)
 339 {
 340         int     diff;
 341
 342         if ((diff = a->proto - b->proto) != 0)
 343                 return (diff);
 344         if ((diff = a->af - b->af) != 0)
 345                 return (diff);
 346         switch (a->af) {
 347 #ifdef INET
 348         case AF_INET:
 349                 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
 350                         return (1);
 351                 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
 352                         return (-1);
 353                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 354                         return (1);
 355                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 356                         return (-1);
 357                 break;
 358 #endif /* INET */
 359 #ifdef INET6
 360         case AF_INET6:
 361                 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
 362                         return (1);
 363                 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
 364                         return (-1);
 365                 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
 366                         return (1);
 367                 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
 368                         return (-1);
 369                 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
 370                         return (1);
 371                 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
 372                         return (-1);
 373                 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
 374                         return (1);
 375                 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
 376                         return (-1);
 377                 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
 378                         return (1);
 379                 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
 380                         return (-1);
 381                 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
 382                         return (1);
 383                 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
 384                         return (-1);
 385                 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
 386                         return (1);
 387                 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
 388                         return (-1);
 389                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 390                         return (1);
 391                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 392                         return (-1);
 393                 break;
 394 #endif /* INET6 */
 395         }
 396
 397         if ((diff = a->lan.port - b->lan.port) != 0)
 398                 return (diff);
 399         if ((diff = a->ext.port - b->ext.port) != 0)
 400                 return (diff);
 401
 402         return (0);
 403 }
 404
 405 static int
 406 pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b)
 407 {
 408         int     diff;
 409
 410         if ((diff = a->proto - b->proto) != 0)
 411                 return (diff);
 412         if ((diff = a->af - b->af) != 0)
 413                 return (diff);
 414         switch (a->af) {
 415 #ifdef INET
 416         case AF_INET:
 417                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 418                         return (1);
 419                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 420                         return (-1);
 421                 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
 422                         return (1);
 423                 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
 424                         return (-1);
 425                 break;
 426 #endif /* INET */
 427 #ifdef INET6
 428         case AF_INET6:
 429                 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
 430                         return (1);
 431                 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
 432                         return (-1);
 433                 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
 434                         return (1);
 435                 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
 436                         return (-1);
 437                 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
 438                         return (1);
 439                 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
 440                         return (-1);
 441                 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
 442                         return (1);
 443                 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
 444                         return (-1);
 445                 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
 446                         return (1);
 447                 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
 448                         return (-1);
 449                 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
 450                         return (1);
 451                 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
 452                         return (-1);
 453                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 454                         return (1);
 455                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 456                         return (-1);
 457                 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
 458                         return (1);
 459                 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
 460                         return (-1);
 461                 break;
 462 #endif /* INET6 */
 463         }
 464
 465         if ((diff = a->ext.port - b->ext.port) != 0)
 466                 return (diff);
 467         if ((diff = a->gwy.port - b->gwy.port) != 0)
 468                 return (diff);
 469
 470         return (0);
 471 }
 472
 473 static int
 474 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
 475 {
 476         if (a->id > b->id)
 477                 return (1);
 478         if (a->id < b->id)
 479                 return (-1);
 480         if (a->creatorid > b->creatorid)
 481                 return (1);
 482         if (a->creatorid < b->creatorid)
 483                 return (-1);
 484
 485         return (0);
 486 }
 487
 488 #ifdef INET6
 489 void
 490 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
 491 {
 492         switch (af) {
 493 #ifdef INET
 494         case AF_INET:
 495                 dst->addr32[0] = src->addr32[0];
 496                 break;
 497 #endif /* INET */
 498         case AF_INET6:
 499                 dst->addr32[0] = src->addr32[0];
 500                 dst->addr32[1] = src->addr32[1];
 501                 dst->addr32[2] = src->addr32[2];
 502                 dst->addr32[3] = src->addr32[3];
 503                 break;
 504         }
 505 }
 506 #endif
 507
 508 struct pf_state *
 509 pf_find_state_byid(struct pf_state *key)
 510 {
 511         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 512         return (RB_FIND(pf_state_tree_id, &tree_id, key));
 513 }
 514
 515 struct pf_state *
 516 pf_find_state_recurse(struct pfi_kif *kif, struct pf_state *key, u_int8_t tree)
 517 {
 518         struct pf_state *s;
 519
 520         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 521
 522         switch (tree) {
 523         case PF_LAN_EXT:
 524                 for (; kif != NULL; kif = kif->pfik_parent) {
 525                         s = RB_FIND(pf_state_tree_lan_ext,
 526                             &kif->pfik_lan_ext, key);
 527                         if (s != NULL)
 528                                 return (s);
 529                 }
 530                 return (NULL);
 531         case PF_EXT_GWY:
 532                 for (; kif != NULL; kif = kif->pfik_parent) {
 533                         s = RB_FIND(pf_state_tree_ext_gwy,
 534                             &kif->pfik_ext_gwy, key);
 535                         if (s != NULL)
 536                                 return (s);
 537                 }
 538                 return (NULL);
 539         default:
 540                 panic("pf_find_state_recurse");
 541         }
 542 }
 543
 544 struct pf_state *
 545 pf_find_state_all(struct pf_state *key, u_int8_t tree, int *more)
 546 {
 547         struct pf_state *s, *ss = NULL;
 548         struct pfi_kif  *kif;
 549
 550         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 551
 552         switch (tree) {
 553         case PF_LAN_EXT:
 554                 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
 555                         s = RB_FIND(pf_state_tree_lan_ext,
 556                             &kif->pfik_lan_ext, key);
 557                         if (s == NULL)
 558                                 continue;
 559                         if (more == NULL)
 560                                 return (s);
 561                         ss = s;
 562                         (*more)++;
 563                 }
 564                 return (ss);
 565         case PF_EXT_GWY:
 566                 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
 567                         s = RB_FIND(pf_state_tree_ext_gwy,
 568                             &kif->pfik_ext_gwy, key);
 569                         if (s == NULL)
 570                                 continue;
 571                         if (more == NULL)
 572                                 return (s);
 573                         ss = s;
 574                         (*more)++;
 575                 }
 576                 return (ss);
 577         default:
 578                 panic("pf_find_state_all");
 579         }
 580 }
 581
 582 int
 583 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
 584     struct pf_addr *src, sa_family_t af)
 585 {
 586         struct pf_src_node      k;
 587
 588         if (*sn == NULL) {
 589                 k.af = af;
 590                 PF_ACPY(&k.addr, src, af);
 591                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 592                     rule->rpool.opts & PF_POOL_STICKYADDR)
 593                         k.rule.ptr = rule;
 594                 else
 595                         k.rule.ptr = NULL;
 596                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
 597                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
 598         }
 599         if (*sn == NULL) {
 600                 if (!rule->max_src_nodes ||
 601                     rule->src_nodes < rule->max_src_nodes)
 602                         (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
 603                 if ((*sn) == NULL)
 604                         return (-1);
 605                 bzero(*sn, sizeof(struct pf_src_node));
 606                 (*sn)->af = af;
 607                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 608                     rule->rpool.opts & PF_POOL_STICKYADDR)
 609                         (*sn)->rule.ptr = rule;
 610                 else
 611                         (*sn)->rule.ptr = NULL;
 612                 PF_ACPY(&(*sn)->addr, src, af);
 613                 if (RB_INSERT(pf_src_tree,
 614                     &tree_src_tracking, *sn) != NULL) {
 615                         if (pf_status.debug >= PF_DEBUG_MISC) {
 616                                 kprintf("pf: src_tree insert failed: ");
 617                                 pf_print_host(&(*sn)->addr, 0, af);
 618                                 kprintf("\n");
 619                         }
 620                         pool_put(&pf_src_tree_pl, *sn);
 621                         return (-1);
 622                 }
 623                 (*sn)->creation = time_second;
 624                 (*sn)->ruletype = rule->action;
 625                 if ((*sn)->rule.ptr != NULL)
 626                         (*sn)->rule.ptr->src_nodes++;
 627                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
 628                 pf_status.src_nodes++;
 629         } else {
 630                 if (rule->max_src_states &&
 631                     (*sn)->states >= rule->max_src_states)
 632                         return (-1);
 633         }
 634         return (0);
 635 }
 636
 637 int
 638 pf_insert_state(struct pfi_kif *kif, struct pf_state *state)
 639 {
 640         /* Thou MUST NOT insert multiple duplicate keys */
 641         state->u.s.kif = kif;
 642         if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) {
 643                 if (pf_status.debug >= PF_DEBUG_MISC) {
 644                         kprintf("pf: state insert failed: tree_lan_ext");
 645                         kprintf(" lan: ");
 646                         pf_print_host(&state->lan.addr, state->lan.port,
 647                             state->af);
 648                         kprintf(" gwy: ");
 649                         pf_print_host(&state->gwy.addr, state->gwy.port,
 650                             state->af);
 651                         kprintf(" ext: ");
 652                         pf_print_host(&state->ext.addr, state->ext.port,
 653                             state->af);
 654                         if (state->sync_flags & PFSTATE_FROMSYNC)
 655                                 kprintf(" (from sync)");
 656                         kprintf("\n");
 657                 }
 658                 return (-1);
 659         }
 660
 661         if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) {
 662                 if (pf_status.debug >= PF_DEBUG_MISC) {
 663                         kprintf("pf: state insert failed: tree_ext_gwy");
 664                         kprintf(" lan: ");
 665                         pf_print_host(&state->lan.addr, state->lan.port,
 666                             state->af);
 667                         kprintf(" gwy: ");
 668                         pf_print_host(&state->gwy.addr, state->gwy.port,
 669                             state->af);
 670                         kprintf(" ext: ");
 671                         pf_print_host(&state->ext.addr, state->ext.port,
 672                             state->af);
 673                         if (state->sync_flags & PFSTATE_FROMSYNC)
 674                                 kprintf(" (from sync)");
 675                         kprintf("\n");
 676                 }
 677                 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
 678                 return (-1);
 679         }
 680
 681         if (state->id == 0 && state->creatorid == 0) {
 682                 state->id = htobe64(pf_status.stateid++);
 683                 state->creatorid = pf_status.hostid;
 684         }
 685         if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) {
 686                 if (pf_status.debug >= PF_DEBUG_MISC) {
 687                         kprintf("pf: state insert failed: "
 688                             "id: %016" PRIx64 " creatorid: %08" PRIx32,
 689                             be64toh(state->id), ntohl(state->creatorid));
 690                         if (state->sync_flags & PFSTATE_FROMSYNC)
 691                                 kprintf(" (from sync)");
 692                         kprintf("\n");
 693                 }
 694                 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
 695                 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state);
 696                 return (-1);
 697         }
 698         TAILQ_INSERT_HEAD(&state_updates, state, u.s.entry_updates);
 699
 700         pf_status.fcounters[FCNT_STATE_INSERT]++;
 701         pf_status.states++;
 702         pfi_attach_state(kif);
 703 #if NPFSYNC
 704         pfsync_insert_state(state);
 705 #endif
 706         return (0);
 707 }
 708
 709 void
 710 pf_purge_timeout(void *arg)
 711 {
 712         struct callout  *to = arg;
 713
 714         crit_enter();
 715         pf_purge_expired_states();
 716         pf_purge_expired_fragments();
 717         pf_purge_expired_src_nodes();
 718         crit_exit();
 719
 720         callout_reset(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz,
 721             pf_purge_timeout, to);
 722 }
 723
 724 u_int32_t
 725 pf_state_expires(const struct pf_state *state)
 726 {
 727         u_int32_t       timeout;
 728         u_int32_t       start;
 729         u_int32_t       end;
 730         u_int32_t       states;
 731
 732         /* handle all PFTM_* > PFTM_MAX here */
 733         if (state->timeout == PFTM_PURGE)
 734                 return (time_second);
 735         if (state->timeout == PFTM_UNTIL_PACKET)
 736                 return (0);
 737         KASSERT((state->timeout < PFTM_MAX),
 738             ("pf_state_expires: timeout > PFTM_MAX"));
 739         timeout = state->rule.ptr->timeout[state->timeout];
 740         if (!timeout)
 741                 timeout = pf_default_rule.timeout[state->timeout];
 742         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
 743         if (start) {
 744                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
 745                 states = state->rule.ptr->states;
 746         } else {
 747                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
 748                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
 749                 states = pf_status.states;
 750         }
 751         if (end && states > start && start < end) {
 752                 if (states < end)
 753                         return (state->expire + timeout * (end - states) /
 754                             (end - start));
 755                 else
 756                         return (time_second);
 757         }
 758         return (state->expire + timeout);
 759 }
 760
 761 void
 762 pf_purge_expired_src_nodes(void)
 763 {
 764          struct pf_src_node             *cur, *next;
 765
 766          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
 767                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
 768
 769                  if (cur->states <= 0 && cur->expire <= time_second) {
 770                          if (cur->rule.ptr != NULL) {
 771                                  cur->rule.ptr->src_nodes--;
 772                                  if (cur->rule.ptr->states <= 0 &&
 773                                      cur->rule.ptr->max_src_nodes <= 0)
 774                                          pf_rm_rule(NULL, cur->rule.ptr);
 775                          }
 776                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
 777                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
 778                          pf_status.src_nodes--;
 779                          pool_put(&pf_src_tree_pl, cur);
 780                  }
 781          }
 782 }
 783
 784 void
 785 pf_src_tree_remove_state(struct pf_state *s)
 786 {
 787         u_int32_t timeout;
 788
 789         if (s->src_node != NULL) {
 790                 if (--s->src_node->states <= 0) {
 791                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
 792                         if (!timeout)
 793                                 timeout =
 794                                     pf_default_rule.timeout[PFTM_SRC_NODE];
 795                         s->src_node->expire = time_second + timeout;
 796                 }
 797         }
 798         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
 799                 if (--s->nat_src_node->states <= 0) {
 800                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
 801                         if (!timeout)
 802                                 timeout =
 803                                     pf_default_rule.timeout[PFTM_SRC_NODE];
 804                         s->nat_src_node->expire = time_second + timeout;
 805                 }
 806         }
 807         s->src_node = s->nat_src_node = NULL;
 808 }
 809
 810 static int
 811 pf_purge_expired_states_callback(struct pf_state *cur, void *data __unused)
 812 {
 813         if (pf_state_expires(cur) <= time_second) {
 814                 RB_REMOVE(pf_state_tree_ext_gwy,
 815                     &cur->u.s.kif->pfik_ext_gwy, cur);
 816                 RB_REMOVE(pf_state_tree_lan_ext,
 817                     &cur->u.s.kif->pfik_lan_ext, cur);
 818                 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
 819                 if (cur->src.state == PF_TCPS_PROXY_DST) {
 820                         pf_send_tcp(cur->rule.ptr, cur->af,
 821                             &cur->ext.addr, &cur->lan.addr,
 822                             cur->ext.port, cur->lan.port,
 823                             cur->src.seqhi, cur->src.seqlo + 1, 0,
 824                             TH_RST|TH_ACK, 0, 0);
 825                 }
 826 #if NPFSYNC
 827                 pfsync_delete_state(cur);
 828 #endif
 829                 pf_src_tree_remove_state(cur);
 830                 if (--cur->rule.ptr->states <= 0 &&
 831                     cur->rule.ptr->src_nodes <= 0)
 832                         pf_rm_rule(NULL, cur->rule.ptr);
 833                 if (cur->nat_rule.ptr != NULL)
 834                         if (--cur->nat_rule.ptr->states <= 0 &&
 835                                 cur->nat_rule.ptr->src_nodes <= 0)
 836                                 pf_rm_rule(NULL, cur->nat_rule.ptr);
 837                 if (cur->anchor.ptr != NULL)
 838                         if (--cur->anchor.ptr->states <= 0)
 839                                 pf_rm_rule(NULL, cur->anchor.ptr);
 840                 pf_normalize_tcp_cleanup(cur);
 841                 pfi_detach_state(cur->u.s.kif);
 842                 TAILQ_REMOVE(&state_updates, cur, u.s.entry_updates);
 843                 pool_put(&pf_state_pl, cur);
 844                 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
 845                 pf_status.states--;
 846         }
 847         return(0);
 848 }
 849
 850 void
 851 pf_purge_expired_states(void)
 852 {
 853         RB_SCAN(pf_state_tree_id, &tree_id, NULL,
 854                 pf_purge_expired_states_callback, NULL);
 855 }
 856
 857
 858 int
 859 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
 860 {
 861         if (aw->type != PF_ADDR_TABLE)
 862                 return (0);
 863         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
 864                 return (1);
 865         return (0);
 866 }
 867
 868 void
 869 pf_tbladdr_remove(struct pf_addr_wrap *aw)
 870 {
 871         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
 872                 return;
 873         pfr_detach_table(aw->p.tbl);
 874         aw->p.tbl = NULL;
 875 }
 876
 877 void
 878 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
 879 {
 880         struct pfr_ktable *kt = aw->p.tbl;
 881
 882         if (aw->type != PF_ADDR_TABLE || kt == NULL)
 883                 return;
 884         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
 885                 kt = kt->pfrkt_root;
 886         aw->p.tbl = NULL;
 887         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
 888                 kt->pfrkt_cnt : -1;
 889 }
 890
 891 void
 892 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
 893 {
 894         switch (af) {
 895 #ifdef INET
 896         case AF_INET: {
 897                 u_int32_t a = ntohl(addr->addr32[0]);
 898                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
 899                     (a>>8)&255, a&255);
 900                 if (p) {
 901                         p = ntohs(p);
 902                         kprintf(":%u", p);
 903                 }
 904                 break;
 905         }
 906 #endif /* INET */
 907 #ifdef INET6
 908         case AF_INET6: {
 909                 u_int16_t b;
 910                 u_int8_t i, curstart = 255, curend = 0,
 911                     maxstart = 0, maxend = 0;
 912                 for (i = 0; i < 8; i++) {
 913                         if (!addr->addr16[i]) {
 914                                 if (curstart == 255)
 915                                         curstart = i;
 916                                 else
 917                                         curend = i;
 918                         } else {
 919                                 if (curstart) {
 920                                         if ((curend - curstart) >
 921                                             (maxend - maxstart)) {
 922                                                 maxstart = curstart;
 923                                                 maxend = curend;
 924                                                 curstart = 255;
 925                                         }
 926                                 }
 927                         }
 928                 }
 929                 for (i = 0; i < 8; i++) {
 930                         if (i >= maxstart && i <= maxend) {
 931                                 if (maxend != 7) {
 932                                         if (i == maxstart)
 933                                                 kprintf(":");
 934                                 } else {
 935                                         if (i == maxend)
 936                                                 kprintf(":");
 937                                 }
 938                         } else {
 939                                 b = ntohs(addr->addr16[i]);
 940                                 kprintf("%x", b);
 941                                 if (i < 7)
 942                                         kprintf(":");
 943                         }
 944                 }
 945                 if (p) {
 946                         p = ntohs(p);
 947                         kprintf("[%u]", p);
 948                 }
 949                 break;
 950         }
 951 #endif /* INET6 */
 952         }
 953 }
 954
 955 void
 956 pf_print_state(struct pf_state *s)
 957 {
 958         switch (s->proto) {
 959         case IPPROTO_TCP:
 960                 kprintf("TCP ");
 961                 break;
 962         case IPPROTO_UDP:
 963                 kprintf("UDP ");
 964                 break;
 965         case IPPROTO_ICMP:
 966                 kprintf("ICMP ");
 967                 break;
 968         case IPPROTO_ICMPV6:
 969                 kprintf("ICMPV6 ");
 970                 break;
 971         default:
 972                 kprintf("%u ", s->proto);
 973                 break;
 974         }
 975         pf_print_host(&s->lan.addr, s->lan.port, s->af);
 976         kprintf(" ");
 977         pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
 978         kprintf(" ");
 979         pf_print_host(&s->ext.addr, s->ext.port, s->af);
 980         kprintf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
 981             s->src.seqhi, s->src.max_win, s->src.seqdiff);
 982         if (s->src.wscale && s->dst.wscale)
 983                 kprintf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
 984         kprintf("]");
 985         kprintf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
 986             s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
 987         if (s->src.wscale && s->dst.wscale)
 988                 kprintf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
 989         kprintf("]");
 990         kprintf(" %u:%u", s->src.state, s->dst.state);
 991 }
 992
 993 void
 994 pf_print_flags(u_int8_t f)
 995 {
 996         if (f)
 997                 kprintf(" ");
 998         if (f & TH_FIN)
 999                 kprintf("F");
1000         if (f & TH_SYN)
1001                 kprintf("S");
1002         if (f & TH_RST)
1003                 kprintf("R");
1004         if (f & TH_PUSH)
1005                 kprintf("P");
1006         if (f & TH_ACK)
1007                 kprintf("A");
1008         if (f & TH_URG)
1009                 kprintf("U");
1010         if (f & TH_ECE)
1011                 kprintf("E");
1012         if (f & TH_CWR)
1013                 kprintf("W");
1014 }
1015
1016 #define PF_SET_SKIP_STEPS(i)                                    \
1017         do {                                                    \
1018                 while (head[i] != cur) {                        \
1019                         head[i]->skip[i].ptr = cur;             \
1020                         head[i] = TAILQ_NEXT(head[i], entries); \
1021                 }                                               \
1022         } while (0)
1023
1024 void
1025 pf_calc_skip_steps(struct pf_rulequeue *rules)
1026 {
1027         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1028         int i;
1029
1030         cur = TAILQ_FIRST(rules);
1031         prev = cur;
1032         for (i = 0; i < PF_SKIP_COUNT; ++i)
1033                 head[i] = cur;
1034         while (cur != NULL) {
1035
1036                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1037                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1038                 if (cur->direction != prev->direction)
1039                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1040                 if (cur->af != prev->af)
1041                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1042                 if (cur->proto != prev->proto)
1043                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1044                 if (cur->src.not != prev->src.not ||
1045                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1046                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1047                 if (cur->src.port[0] != prev->src.port[0] ||
1048                     cur->src.port[1] != prev->src.port[1] ||
1049                     cur->src.port_op != prev->src.port_op)
1050                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1051                 if (cur->dst.not != prev->dst.not ||
1052                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1053                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1054                 if (cur->dst.port[0] != prev->dst.port[0] ||
1055                     cur->dst.port[1] != prev->dst.port[1] ||
1056                     cur->dst.port_op != prev->dst.port_op)
1057                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1058
1059                 prev = cur;
1060                 cur = TAILQ_NEXT(cur, entries);
1061         }
1062         for (i = 0; i < PF_SKIP_COUNT; ++i)
1063                 PF_SET_SKIP_STEPS(i);
1064 }
1065
1066 int
1067 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1068 {
1069         if (aw1->type != aw2->type)
1070                 return (1);
1071         switch (aw1->type) {
1072         case PF_ADDR_ADDRMASK:
1073                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1074                         return (1);
1075                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1076                         return (1);
1077                 return (0);
1078         case PF_ADDR_DYNIFTL:
1079                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1080         case PF_ADDR_NOROUTE:
1081                 return (0);
1082         case PF_ADDR_TABLE:
1083                 return (aw1->p.tbl != aw2->p.tbl);
1084         default:
1085                 kprintf("invalid address type: %d\n", aw1->type);
1086                 return (1);
1087         }
1088 }
1089
1090 void
1091 pf_update_anchor_rules(void)
1092 {
1093         struct pf_rule  *rule;
1094         int              i;
1095
1096         for (i = 0; i < PF_RULESET_MAX; ++i)
1097                 TAILQ_FOREACH(rule, pf_main_ruleset.rules[i].active.ptr,
1098                     entries)
1099                         if (rule->anchorname[0])
1100                                 rule->anchor = pf_find_anchor(rule->anchorname);
1101                         else
1102                                 rule->anchor = NULL;
1103 }
1104
1105 u_int16_t
1106 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1107 {
1108         u_int32_t       l;
1109
1110         if (udp && !cksum)
1111                 return (0x0000);
1112         l = cksum + old - new;
1113         l = (l >> 16) + (l & 65535);
1114         l = l & 65535;
1115         if (udp && !l)
1116                 return (0xFFFF);
1117         return (l);
1118 }
1119
1120 void
1121 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1122     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1123 {
1124         struct pf_addr  ao;
1125         u_int16_t       po = *p;
1126
1127         PF_ACPY(&ao, a, af);
1128         PF_ACPY(a, an, af);
1129
1130         *p = pn;
1131
1132         switch (af) {
1133 #ifdef INET
1134         case AF_INET:
1135                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1136                     ao.addr16[0], an->addr16[0], 0),
1137                     ao.addr16[1], an->addr16[1], 0);
1138                 *p = pn;
1139                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1140                     ao.addr16[0], an->addr16[0], u),
1141                     ao.addr16[1], an->addr16[1], u),
1142                     po, pn, u);
1143                 break;
1144 #endif /* INET */
1145 #ifdef INET6
1146         case AF_INET6:
1147                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1148                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1149                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1150                     ao.addr16[0], an->addr16[0], u),
1151                     ao.addr16[1], an->addr16[1], u),
1152                     ao.addr16[2], an->addr16[2], u),
1153                     ao.addr16[3], an->addr16[3], u),
1154                     ao.addr16[4], an->addr16[4], u),
1155                     ao.addr16[5], an->addr16[5], u),
1156                     ao.addr16[6], an->addr16[6], u),
1157                     ao.addr16[7], an->addr16[7], u),
1158                     po, pn, u);
1159                 break;
1160 #endif /* INET6 */
1161         }
1162 }
1163
1164
1165 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1166 void
1167 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1168 {
1169         u_int32_t       ao;
1170
1171         memcpy(&ao, a, sizeof(ao));
1172         memcpy(a, &an, sizeof(u_int32_t));
1173         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1174             ao % 65536, an % 65536, u);
1175 }
1176
1177 #ifdef INET6
1178 void
1179 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1180 {
1181         struct pf_addr  ao;
1182
1183         PF_ACPY(&ao, a, AF_INET6);
1184         PF_ACPY(a, an, AF_INET6);
1185
1186         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1187             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1188             pf_cksum_fixup(pf_cksum_fixup(*c,
1189             ao.addr16[0], an->addr16[0], u),
1190             ao.addr16[1], an->addr16[1], u),
1191             ao.addr16[2], an->addr16[2], u),
1192             ao.addr16[3], an->addr16[3], u),
1193             ao.addr16[4], an->addr16[4], u),
1194             ao.addr16[5], an->addr16[5], u),
1195             ao.addr16[6], an->addr16[6], u),
1196             ao.addr16[7], an->addr16[7], u);
1197 }
1198 #endif /* INET6 */
1199
1200 void
1201 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1202     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1203     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1204 {
1205         struct pf_addr  oia, ooa;
1206
1207         PF_ACPY(&oia, ia, af);
1208         PF_ACPY(&ooa, oa, af);
1209
1210         /* Change inner protocol port, fix inner protocol checksum. */
1211         if (ip != NULL) {
1212                 u_int16_t       oip = *ip;
1213                 u_int32_t       opc = 0;
1214
1215                 if (pc != NULL)
1216                         opc = *pc;
1217                 *ip = np;
1218                 if (pc != NULL)
1219                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1220                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1221                 if (pc != NULL)
1222                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1223         }
1224         /* Change inner ip address, fix inner ip and icmp checksums. */
1225         PF_ACPY(ia, na, af);
1226         switch (af) {
1227 #ifdef INET
1228         case AF_INET: {
1229                 u_int32_t        oh2c = *h2c;
1230
1231                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1232                     oia.addr16[0], ia->addr16[0], 0),
1233                     oia.addr16[1], ia->addr16[1], 0);
1234                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1235                     oia.addr16[0], ia->addr16[0], 0),
1236                     oia.addr16[1], ia->addr16[1], 0);
1237                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1238                 break;
1239         }
1240 #endif /* INET */
1241 #ifdef INET6
1242         case AF_INET6:
1243                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1244                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1245                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1246                     oia.addr16[0], ia->addr16[0], u),
1247                     oia.addr16[1], ia->addr16[1], u),
1248                     oia.addr16[2], ia->addr16[2], u),
1249                     oia.addr16[3], ia->addr16[3], u),
1250                     oia.addr16[4], ia->addr16[4], u),
1251                     oia.addr16[5], ia->addr16[5], u),
1252                     oia.addr16[6], ia->addr16[6], u),
1253                     oia.addr16[7], ia->addr16[7], u);
1254                 break;
1255 #endif /* INET6 */
1256         }
1257         /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1258         PF_ACPY(oa, na, af);
1259         switch (af) {
1260 #ifdef INET
1261         case AF_INET:
1262                 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1263                     ooa.addr16[0], oa->addr16[0], 0),
1264                     ooa.addr16[1], oa->addr16[1], 0);
1265                 break;
1266 #endif /* INET */
1267 #ifdef INET6
1268         case AF_INET6:
1269                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1270                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1271                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1272                     ooa.addr16[0], oa->addr16[0], u),
1273                     ooa.addr16[1], oa->addr16[1], u),
1274                     ooa.addr16[2], oa->addr16[2], u),
1275                     ooa.addr16[3], oa->addr16[3], u),
1276                     ooa.addr16[4], oa->addr16[4], u),
1277                     ooa.addr16[5], oa->addr16[5], u),
1278                     ooa.addr16[6], oa->addr16[6], u),
1279                     ooa.addr16[7], oa->addr16[7], u);
1280                 break;
1281 #endif /* INET6 */
1282         }
1283 }
1284
1285 void
1286 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1287     const struct pf_addr *saddr, const struct pf_addr *daddr,
1288     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1289     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl)
1290 {
1291         struct mbuf     *m;
1292         int              len = 0, tlen;
1293 #ifdef INET
1294         struct ip       *h = NULL;
1295 #endif /* INET */
1296 #ifdef INET6
1297         struct ip6_hdr  *h6 = NULL;
1298 #endif /* INET6 */
1299         struct tcphdr   *th = NULL;
1300         char *opt;
1301
1302         /* maximum segment size tcp option */
1303         tlen = sizeof(struct tcphdr);
1304         if (mss)
1305                 tlen += 4;
1306
1307         switch (af) {
1308 #ifdef INET
1309         case AF_INET:
1310                 len = sizeof(struct ip) + tlen;
1311                 break;
1312 #endif /* INET */
1313 #ifdef INET6
1314         case AF_INET6:
1315                 len = sizeof(struct ip6_hdr) + tlen;
1316                 break;
1317 #endif /* INET6 */
1318         }
1319
1320         /* create outgoing mbuf */
1321         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1322         if (m == NULL)
1323                 return;
1324         m->m_pkthdr.fw_flags = PF_MBUF_GENERATED;
1325 #ifdef ALTQ
1326         if (r != NULL && r->qid) {
1327                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
1328                 m->m_pkthdr.altq_qid = r->qid;
1329                 m->m_pkthdr.ecn_af = af;
1330                 m->m_pkthdr.header = mtod(m, struct ip *);
1331         }
1332 #endif
1333         m->m_data += max_linkhdr;
1334         m->m_pkthdr.len = m->m_len = len;
1335         m->m_pkthdr.rcvif = NULL;
1336         bzero(m->m_data, len);
1337         switch (af) {
1338 #ifdef INET
1339         case AF_INET:
1340                 h = mtod(m, struct ip *);
1341
1342                 /* IP header fields included in the TCP checksum */
1343                 h->ip_p = IPPROTO_TCP;
1344                 h->ip_len = tlen;
1345                 h->ip_src.s_addr = saddr->v4.s_addr;
1346                 h->ip_dst.s_addr = daddr->v4.s_addr;
1347
1348                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1349                 break;
1350 #endif /* INET */
1351 #ifdef INET6
1352         case AF_INET6:
1353                 h6 = mtod(m, struct ip6_hdr *);
1354
1355                 /* IP header fields included in the TCP checksum */
1356                 h6->ip6_nxt = IPPROTO_TCP;
1357                 h6->ip6_plen = htons(tlen);
1358                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1359                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1360
1361                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1362                 break;
1363 #endif /* INET6 */
1364         }
1365
1366         /* TCP header */
1367         th->th_sport = sport;
1368         th->th_dport = dport;
1369         th->th_seq = htonl(seq);
1370         th->th_ack = htonl(ack);
1371         th->th_off = tlen >> 2;
1372         th->th_flags = flags;
1373         th->th_win = htons(win);
1374
1375         if (mss) {
1376                 opt = (char *)(th + 1);
1377                 opt[0] = TCPOPT_MAXSEG;
1378                 opt[1] = 4;
1379                 mss = htons(mss);
1380                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1381         }
1382
1383         switch (af) {
1384 #ifdef INET
1385         case AF_INET:
1386                 /* TCP checksum */
1387                 th->th_sum = in_cksum(m, len);
1388
1389                 /* Finish the IP header */
1390                 h->ip_v = 4;
1391                 h->ip_hl = sizeof(*h) >> 2;
1392                 h->ip_tos = IPTOS_LOWDELAY;
1393                 h->ip_len = len;
1394                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1395                 h->ip_ttl = ttl ? ttl : ip_defttl;
1396                 h->ip_sum = 0;
1397                 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1398                     (void *)NULL);
1399                 break;
1400 #endif /* INET */
1401 #ifdef INET6
1402         case AF_INET6:
1403                 /* TCP checksum */
1404                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1405                     sizeof(struct ip6_hdr), tlen);
1406
1407                 h6->ip6_vfc |= IPV6_VERSION;
1408                 h6->ip6_hlim = IPV6_DEFHLIM;
1409
1410                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1411                 break;
1412 #endif /* INET6 */
1413         }
1414 }
1415
1416 void
1417 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1418     struct pf_rule *r)
1419 {
1420         struct mbuf     *m0;
1421
1422         m0 = m_copypacket(m, MB_DONTWAIT);
1423         if (m0 == NULL)
1424                 return;
1425         m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
1426
1427 #ifdef ALTQ
1428         if (r->qid) {
1429                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
1430                 m->m_pkthdr.altq_qid = r->qid;
1431                 m->m_pkthdr.ecn_af = af;
1432                 m->m_pkthdr.header = mtod(m0, struct ip *);
1433         }
1434 #endif
1435
1436         switch (af) {
1437 #ifdef INET
1438         case AF_INET:
1439                 icmp_error(m0, type, code, 0, 0);
1440                 break;
1441 #endif /* INET */
1442 #ifdef INET6
1443         case AF_INET6:
1444                 icmp6_error(m0, type, code, 0);
1445                 break;
1446 #endif /* INET6 */
1447         }
1448 }
1449
1450 /*
1451  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1452  * If n is 0, they match if they are equal. If n is != 0, they match if they
1453  * are different.
1454  */
1455 int
1456 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1457     struct pf_addr *b, sa_family_t af)
1458 {
1459         int     match = 0;
1460
1461         switch (af) {
1462 #ifdef INET
1463         case AF_INET:
1464                 if ((a->addr32[0] & m->addr32[0]) ==
1465                     (b->addr32[0] & m->addr32[0]))
1466                         match++;
1467                 break;
1468 #endif /* INET */
1469 #ifdef INET6
1470         case AF_INET6:
1471                 if (((a->addr32[0] & m->addr32[0]) ==
1472                      (b->addr32[0] & m->addr32[0])) &&
1473                     ((a->addr32[1] & m->addr32[1]) ==
1474                      (b->addr32[1] & m->addr32[1])) &&
1475                     ((a->addr32[2] & m->addr32[2]) ==
1476                      (b->addr32[2] & m->addr32[2])) &&
1477                     ((a->addr32[3] & m->addr32[3]) ==
1478                      (b->addr32[3] & m->addr32[3])))
1479                         match++;
1480                 break;
1481 #endif /* INET6 */
1482         }
1483         if (match) {
1484                 if (n)
1485                         return (0);
1486                 else
1487                         return (1);
1488         } else {
1489                 if (n)
1490                         return (1);
1491                 else
1492                         return (0);
1493         }
1494 }
1495
1496 int
1497 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1498 {
1499         switch (op) {
1500         case PF_OP_IRG:
1501                 return ((p > a1) && (p < a2));
1502         case PF_OP_XRG:
1503                 return ((p < a1) || (p > a2));
1504         case PF_OP_RRG:
1505                 return ((p >= a1) && (p <= a2));
1506         case PF_OP_EQ:
1507                 return (p == a1);
1508         case PF_OP_NE:
1509                 return (p != a1);
1510         case PF_OP_LT:
1511                 return (p < a1);
1512         case PF_OP_LE:
1513                 return (p <= a1);
1514         case PF_OP_GT:
1515                 return (p > a1);
1516         case PF_OP_GE:
1517                 return (p >= a1);
1518         }
1519         return (0); /* never reached */
1520 }
1521
1522 int
1523 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1524 {
1525         a1 = ntohs(a1);
1526         a2 = ntohs(a2);
1527         p = ntohs(p);
1528         return (pf_match(op, a1, a2, p));
1529 }
1530
1531 int
1532 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1533 {
1534         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1535                 return (0);
1536         return (pf_match(op, a1, a2, u));
1537 }
1538
1539 int
1540 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1541 {
1542         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1543                 return (0);
1544         return (pf_match(op, a1, a2, g));
1545 }
1546
1547 static int
1548 pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_rule *nat_rule,
1549              int *tag)
1550 {
1551         if (*tag == -1) {       /* find mbuf tag */
1552                 if (nat_rule != NULL && nat_rule->tag)
1553                         *tag = nat_rule->tag;
1554                 else if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
1555                         *tag = m->m_pkthdr.pf_tag;
1556                 else
1557                         *tag = 0;
1558         }
1559
1560         return ((!r->match_tag_not && r->match_tag == *tag) ||
1561             (r->match_tag_not && r->match_tag != *tag));
1562 }
1563
1564 void
1565 pf_tag_packet(struct mbuf *m, int tag)
1566 {
1567         if (tag <= 0)
1568                 return;
1569
1570         m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1571         m->m_pkthdr.pf_tag = tag;
1572 }
1573
1574 #define PF_STEP_INTO_ANCHOR(r, a, s, n)                                 \
1575         do {                                                            \
1576                 if ((r) == NULL || (r)->anchor == NULL ||               \
1577                     (s) != NULL || (a) != NULL)                         \
1578                         panic("PF_STEP_INTO_ANCHOR");                   \
1579                 (a) = (r);                                              \
1580                 (s) = TAILQ_FIRST(&(r)->anchor->rulesets);              \
1581                 (r) = NULL;                                             \
1582                 while ((s) != NULL && ((r) =                            \
1583                     TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL)     \
1584                         (s) = TAILQ_NEXT((s), entries);                 \
1585                 if ((r) == NULL) {                                      \
1586                         (r) = TAILQ_NEXT((a), entries);                 \
1587                         (a) = NULL;                                     \
1588                 }                                                       \
1589         } while (0)
1590
1591 #define PF_STEP_OUT_OF_ANCHOR(r, a, s, n)                               \
1592         do {                                                            \
1593                 if ((r) != NULL || (a) == NULL || (s) == NULL)          \
1594                         panic("PF_STEP_OUT_OF_ANCHOR");                 \
1595                 (s) = TAILQ_NEXT((s), entries);                         \
1596                 while ((s) != NULL && ((r) =                            \
1597                     TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL)     \
1598                         (s) = TAILQ_NEXT((s), entries);                 \
1599                 if ((r) == NULL) {                                      \
1600                         (r) = TAILQ_NEXT((a), entries);                 \
1601                         (a) = NULL;                                     \
1602                 }                                                       \
1603         } while (0)
1604
1605 #ifdef INET6
1606 void
1607 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
1608     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
1609 {
1610         switch (af) {
1611 #ifdef INET
1612         case AF_INET:
1613                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1614                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1615                 break;
1616 #endif /* INET */
1617         case AF_INET6:
1618                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1619                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1620                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
1621                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
1622                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
1623                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
1624                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
1625                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
1626                 break;
1627         }
1628 }
1629
1630 void
1631 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
1632 {
1633         switch (af) {
1634 #ifdef INET
1635         case AF_INET:
1636                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
1637                 break;
1638 #endif /* INET */
1639         case AF_INET6:
1640                 if (addr->addr32[3] == 0xffffffff) {
1641                         addr->addr32[3] = 0;
1642                         if (addr->addr32[2] == 0xffffffff) {
1643                                 addr->addr32[2] = 0;
1644                                 if (addr->addr32[1] == 0xffffffff) {
1645                                         addr->addr32[1] = 0;
1646                                         addr->addr32[0] =
1647                                             htonl(ntohl(addr->addr32[0]) + 1);
1648                                 } else
1649                                         addr->addr32[1] =
1650                                             htonl(ntohl(addr->addr32[1]) + 1);
1651                         } else
1652                                 addr->addr32[2] =
1653                                     htonl(ntohl(addr->addr32[2]) + 1);
1654                 } else
1655                         addr->addr32[3] =
1656                             htonl(ntohl(addr->addr32[3]) + 1);
1657                 break;
1658         }
1659 }
1660 #endif /* INET6 */
1661
1662 #define mix(a,b,c) \
1663         do {                                    \
1664                 a -= b; a -= c; a ^= (c >> 13); \
1665                 b -= c; b -= a; b ^= (a << 8);  \
1666                 c -= a; c -= b; c ^= (b >> 13); \
1667                 a -= b; a -= c; a ^= (c >> 12); \
1668                 b -= c; b -= a; b ^= (a << 16); \
1669                 c -= a; c -= b; c ^= (b >> 5);  \
1670                 a -= b; a -= c; a ^= (c >> 3);  \
1671                 b -= c; b -= a; b ^= (a << 10); \
1672                 c -= a; c -= b; c ^= (b >> 15); \
1673         } while (0)
1674
1675 /*
1676  * hash function based on bridge_hash in if_bridge.c
1677  */
1678 void
1679 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
1680     struct pf_poolhashkey *key, sa_family_t af)
1681 {
1682         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
1683
1684         switch (af) {
1685 #ifdef INET
1686         case AF_INET:
1687                 a += inaddr->addr32[0];
1688                 b += key->key32[1];
1689                 mix(a, b, c);
1690                 hash->addr32[0] = c + key->key32[2];
1691                 break;
1692 #endif /* INET */
1693 #ifdef INET6
1694         case AF_INET6:
1695                 a += inaddr->addr32[0];
1696                 b += inaddr->addr32[2];
1697                 mix(a, b, c);
1698                 hash->addr32[0] = c;
1699                 a += inaddr->addr32[1];
1700                 b += inaddr->addr32[3];
1701                 c += key->key32[1];
1702                 mix(a, b, c);
1703                 hash->addr32[1] = c;
1704                 a += inaddr->addr32[2];
1705                 b += inaddr->addr32[1];
1706                 c += key->key32[2];
1707                 mix(a, b, c);
1708                 hash->addr32[2] = c;
1709                 a += inaddr->addr32[3];
1710                 b += inaddr->addr32[0];
1711                 c += key->key32[3];
1712                 mix(a, b, c);
1713                 hash->addr32[3] = c;
1714                 break;
1715 #endif /* INET6 */
1716         }
1717 }
1718
1719 int
1720 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
1721     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
1722 {
1723         unsigned char            hash[16];
1724         struct pf_pool          *rpool = &r->rpool;
1725         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
1726         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
1727         struct pf_pooladdr      *acur = rpool->cur;
1728         struct pf_src_node       k;
1729
1730         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
1731             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1732                 k.af = af;
1733                 PF_ACPY(&k.addr, saddr, af);
1734                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
1735                     r->rpool.opts & PF_POOL_STICKYADDR)
1736                         k.rule.ptr = r;
1737                 else
1738                         k.rule.ptr = NULL;
1739                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
1740                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
1741                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
1742                         PF_ACPY(naddr, &(*sn)->raddr, af);
1743                         if (pf_status.debug >= PF_DEBUG_MISC) {
1744                                 kprintf("pf_map_addr: src tracking maps ");
1745                                 pf_print_host(&k.addr, 0, af);
1746                                 kprintf(" to ");
1747                                 pf_print_host(naddr, 0, af);
1748                                 kprintf("\n");
1749                         }
1750                         return (0);
1751                 }
1752         }
1753
1754         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
1755                 return (1);
1756         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1757                 if (af == AF_INET) {
1758                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
1759                             (rpool->opts & PF_POOL_TYPEMASK) !=
1760                             PF_POOL_ROUNDROBIN)
1761                                 return (1);
1762                          raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
1763                          rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
1764                 } else {
1765                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
1766                             (rpool->opts & PF_POOL_TYPEMASK) !=
1767                             PF_POOL_ROUNDROBIN)
1768                                 return (1);
1769                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
1770                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
1771                 }
1772         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1773                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
1774                         return (1); /* unsupported */
1775         } else {
1776                 raddr = &rpool->cur->addr.v.a.addr;
1777                 rmask = &rpool->cur->addr.v.a.mask;
1778         }
1779
1780         switch (rpool->opts & PF_POOL_TYPEMASK) {
1781         case PF_POOL_NONE:
1782                 PF_ACPY(naddr, raddr, af);
1783                 break;
1784         case PF_POOL_BITMASK:
1785                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
1786                 break;
1787         case PF_POOL_RANDOM:
1788                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
1789                         switch (af) {
1790 #ifdef INET
1791                         case AF_INET:
1792                                 rpool->counter.addr32[0] = karc4random();
1793                                 break;
1794 #endif /* INET */
1795 #ifdef INET6
1796                         case AF_INET6:
1797                                 if (rmask->addr32[3] != 0xffffffff)
1798                                         rpool->counter.addr32[3] = karc4random();
1799                                 else
1800                                         break;
1801                                 if (rmask->addr32[2] != 0xffffffff)
1802                                         rpool->counter.addr32[2] = karc4random();
1803                                 else
1804                                         break;
1805                                 if (rmask->addr32[1] != 0xffffffff)
1806                                         rpool->counter.addr32[1] = karc4random();
1807                                 else
1808                                         break;
1809                                 if (rmask->addr32[0] != 0xffffffff)
1810                                         rpool->counter.addr32[0] = karc4random();
1811                                 break;
1812 #endif /* INET6 */
1813                         }
1814                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1815                         PF_ACPY(init_addr, naddr, af);
1816
1817                 } else {
1818                         PF_AINC(&rpool->counter, af);
1819                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1820                 }
1821                 break;
1822         case PF_POOL_SRCHASH:
1823                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
1824                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
1825                 break;
1826         case PF_POOL_ROUNDROBIN:
1827                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1828                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
1829                             &rpool->tblidx, &rpool->counter,
1830                             &raddr, &rmask, af))
1831                                 goto get_addr;
1832                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1833                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
1834                             &rpool->tblidx, &rpool->counter,
1835                             &raddr, &rmask, af))
1836                                 goto get_addr;
1837                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
1838                         goto get_addr;
1839
1840         try_next:
1841                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
1842                         rpool->cur = TAILQ_FIRST(&rpool->list);
1843                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1844                         rpool->tblidx = -1;
1845                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
1846                             &rpool->tblidx, &rpool->counter,
1847                             &raddr, &rmask, af)) {
1848                                 /* table contains no address of type 'af' */
1849                                 if (rpool->cur != acur)
1850                                         goto try_next;
1851                                 return (1);
1852                         }
1853                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1854                         rpool->tblidx = -1;
1855                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
1856                             &rpool->tblidx, &rpool->counter,
1857                             &raddr, &rmask, af)) {
1858                                 /* table contains no address of type 'af' */
1859                                 if (rpool->cur != acur)
1860                                         goto try_next;
1861                                 return (1);
1862                         }
1863                 } else {
1864                         raddr = &rpool->cur->addr.v.a.addr;
1865                         rmask = &rpool->cur->addr.v.a.mask;
1866                         PF_ACPY(&rpool->counter, raddr, af);
1867                 }
1868
1869         get_addr:
1870                 PF_ACPY(naddr, &rpool->counter, af);
1871                 PF_AINC(&rpool->counter, af);
1872                 break;
1873         }
1874         if (*sn != NULL)
1875                 PF_ACPY(&(*sn)->raddr, naddr, af);
1876
1877         if (pf_status.debug >= PF_DEBUG_MISC &&
1878             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1879                 kprintf("pf_map_addr: selected address ");
1880                 pf_print_host(naddr, 0, af);
1881                 kprintf("\n");
1882         }
1883
1884         return (0);
1885 }
1886
1887 int
1888 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
1889     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
1890     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
1891     struct pf_src_node **sn)
1892 {
1893         struct pf_state         key;
1894         struct pf_addr          init_addr;
1895         u_int16_t               cut;
1896
1897         bzero(&init_addr, sizeof(init_addr));
1898         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
1899                 return (1);
1900
1901         do {
1902                 key.af = af;
1903                 key.proto = proto;
1904                 PF_ACPY(&key.ext.addr, daddr, key.af);
1905                 PF_ACPY(&key.gwy.addr, naddr, key.af);
1906                 key.ext.port = dport;
1907
1908                 /*
1909                  * port search; start random, step;
1910                  * similar 2 portloop in in_pcbbind
1911                  */
1912                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP)) {
1913                         key.gwy.port = 0;
1914                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
1915                                 return (0);
1916                 } else if (low == 0 && high == 0) {
1917                         key.gwy.port = *nport;
1918                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
1919                                 return (0);
1920                 } else if (low == high) {
1921                         key.gwy.port = htons(low);
1922                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
1923                                 *nport = htons(low);
1924                                 return (0);
1925                         }
1926                 } else {
1927                         u_int16_t tmp;
1928
1929                         if (low > high) {
1930                                 tmp = low;
1931                                 low = high;
1932                                 high = tmp;
1933                         }
1934                         /* low < high */
1935                         cut = karc4random() % (1 + high - low) + low;
1936                         /* low <= cut <= high */
1937                         for (tmp = cut; tmp <= high; ++(tmp)) {
1938                                 key.gwy.port = htons(tmp);
1939                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
1940                                     NULL) {
1941                                         *nport = htons(tmp);
1942                                         return (0);
1943                                 }
1944                         }
1945                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
1946                                 key.gwy.port = htons(tmp);
1947                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
1948                                     NULL) {
1949                                         *nport = htons(tmp);
1950                                         return (0);
1951                                 }
1952                         }
1953                 }
1954
1955                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
1956                 case PF_POOL_RANDOM:
1957                 case PF_POOL_ROUNDROBIN:
1958                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
1959                                 return (1);
1960                         break;
1961                 case PF_POOL_NONE:
1962                 case PF_POOL_SRCHASH:
1963                 case PF_POOL_BITMASK:
1964                 default:
1965                         return (1);
1966                 }
1967         } while (! PF_AEQ(&init_addr, naddr, af) );
1968
1969         return (1);                                     /* none available */
1970 }
1971
1972 struct pf_rule *
1973 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
1974     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
1975     struct pf_addr *daddr, u_int16_t dport, int rs_num)
1976 {
1977         struct pf_rule          *r, *rm = NULL, *anchorrule = NULL;
1978         struct pf_ruleset       *ruleset = NULL;
1979
1980         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
1981         while (r && rm == NULL) {
1982                 struct pf_rule_addr     *src = NULL, *dst = NULL;
1983                 struct pf_addr_wrap     *xdst = NULL;
1984
1985                 if (r->action == PF_BINAT && direction == PF_IN) {
1986                         src = &r->dst;
1987                         if (r->rpool.cur != NULL)
1988                                 xdst = &r->rpool.cur->addr;
1989                 } else {
1990                         src = &r->src;
1991                         dst = &r->dst;
1992                 }
1993
1994                 r->evaluations++;
1995                 if (r->kif != NULL &&
1996                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
1997                         r = r->skip[PF_SKIP_IFP].ptr;
1998                 else if (r->direction && r->direction != direction)
1999                         r = r->skip[PF_SKIP_DIR].ptr;
2000                 else if (r->af && r->af != pd->af)
2001                         r = r->skip[PF_SKIP_AF].ptr;
2002                 else if (r->proto && r->proto != pd->proto)
2003                         r = r->skip[PF_SKIP_PROTO].ptr;
2004                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->not))
2005                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
2006                             PF_SKIP_DST_ADDR].ptr;
2007                 else if (src->port_op && !pf_match_port(src->port_op,
2008                     src->port[0], src->port[1], sport))
2009                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
2010                             PF_SKIP_DST_PORT].ptr;
2011                 else if (dst != NULL &&
2012                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->not))
2013                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2014                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0))
2015                         r = TAILQ_NEXT(r, entries);
2016                 else if (dst != NULL && dst->port_op &&
2017                     !pf_match_port(dst->port_op, dst->port[0],
2018                     dst->port[1], dport))
2019                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2020                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2021                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2022                     off, pd->hdr.tcp), r->os_fingerprint)))
2023                         r = TAILQ_NEXT(r, entries);
2024                 else if (r->anchorname[0] && r->anchor == NULL)
2025                         r = TAILQ_NEXT(r, entries);
2026                 else if (r->anchor == NULL)
2027                                 rm = r;
2028                 else
2029                         PF_STEP_INTO_ANCHOR(r, anchorrule, ruleset, rs_num);
2030                 if (r == NULL && anchorrule != NULL)
2031                         PF_STEP_OUT_OF_ANCHOR(r, anchorrule, ruleset,
2032                             rs_num);
2033         }
2034         if (rm != NULL && (rm->action == PF_NONAT ||
2035             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2036                 return (NULL);
2037         return (rm);
2038 }
2039
2040 struct pf_rule *
2041 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2042     struct pfi_kif *kif, struct pf_src_node **sn,
2043     struct pf_addr *saddr, u_int16_t sport,
2044     struct pf_addr *daddr, u_int16_t dport,
2045     struct pf_addr *naddr, u_int16_t *nport)
2046 {
2047         struct pf_rule  *r = NULL;
2048
2049         if (direction == PF_OUT) {
2050                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2051                     sport, daddr, dport, PF_RULESET_BINAT);
2052                 if (r == NULL)
2053                         r = pf_match_translation(pd, m, off, direction, kif,
2054                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2055         } else {
2056                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2057                     sport, daddr, dport, PF_RULESET_RDR);
2058                 if (r == NULL)
2059                         r = pf_match_translation(pd, m, off, direction, kif,
2060                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2061         }
2062
2063         if (r != NULL) {
2064                 switch (r->action) {
2065                 case PF_NONAT:
2066                 case PF_NOBINAT:
2067                 case PF_NORDR:
2068                         return (NULL);
2069                 case PF_NAT:
2070                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2071                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2072                             r->rpool.proxy_port[1], sn)) {
2073                                 DPFPRINTF(PF_DEBUG_MISC,
2074                                     ("pf: NAT proxy port allocation "
2075                                     "(%u-%u) failed\n",
2076                                     r->rpool.proxy_port[0],
2077                                     r->rpool.proxy_port[1]));
2078                                 return (NULL);
2079                         }
2080                         break;
2081                 case PF_BINAT:
2082                         switch (direction) {
2083                         case PF_OUT:
2084                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2085                                         if (pd->af == AF_INET) {
2086                                                 if (r->rpool.cur->addr.p.dyn->
2087                                                     pfid_acnt4 < 1)
2088                                                         return (NULL);
2089                                                 PF_POOLMASK(naddr,
2090                                                     &r->rpool.cur->addr.p.dyn->
2091                                                     pfid_addr4,
2092                                                     &r->rpool.cur->addr.p.dyn->
2093                                                     pfid_mask4,
2094                                                     saddr, AF_INET);
2095                                         } else {
2096                                                 if (r->rpool.cur->addr.p.dyn->
2097                                                     pfid_acnt6 < 1)
2098                                                         return (NULL);
2099                                                 PF_POOLMASK(naddr,
2100                                                     &r->rpool.cur->addr.p.dyn->
2101                                                     pfid_addr6,
2102                                                     &r->rpool.cur->addr.p.dyn->
2103                                                     pfid_mask6,
2104                                                     saddr, AF_INET6);
2105                                         }
2106                                 } else
2107                                         PF_POOLMASK(naddr,
2108                                             &r->rpool.cur->addr.v.a.addr,
2109                                             &r->rpool.cur->addr.v.a.mask,
2110                                             saddr, pd->af);
2111                                 break;
2112                         case PF_IN:
2113                                 if (r->src.addr.type == PF_ADDR_DYNIFTL){
2114                                         if (pd->af == AF_INET) {
2115                                                 if (r->src.addr.p.dyn->
2116                                                     pfid_acnt4 < 1)
2117                                                         return (NULL);
2118                                                 PF_POOLMASK(naddr,
2119                                                     &r->src.addr.p.dyn->
2120                                                     pfid_addr4,
2121                                                     &r->src.addr.p.dyn->
2122                                                     pfid_mask4,
2123                                                     daddr, AF_INET);
2124                                         } else {
2125                                                 if (r->src.addr.p.dyn->
2126                                                     pfid_acnt6 < 1)
2127                                                         return (NULL);
2128                                                 PF_POOLMASK(naddr,
2129                                                     &r->src.addr.p.dyn->
2130                                                     pfid_addr6,
2131                                                     &r->src.addr.p.dyn->
2132                                                     pfid_mask6,
2133                                                     daddr, AF_INET6);
2134                                         }
2135                                 } else
2136                                         PF_POOLMASK(naddr,
2137                                             &r->src.addr.v.a.addr,
2138                                             &r->src.addr.v.a.mask, daddr,
2139                                             pd->af);
2140                                 break;
2141                         }
2142                         break;
2143                 case PF_RDR: {
2144                         if (pf_map_addr(r->af, r, saddr, naddr, NULL, sn))
2145                                 return (NULL);
2146
2147                         if (r->rpool.proxy_port[1]) {
2148                                 u_int32_t       tmp_nport;
2149
2150                                 tmp_nport = ((ntohs(dport) -
2151                                     ntohs(r->dst.port[0])) %
2152                                     (r->rpool.proxy_port[1] -
2153                                     r->rpool.proxy_port[0] + 1)) +
2154                                     r->rpool.proxy_port[0];
2155
2156                                 /* wrap around if necessary */
2157                                 if (tmp_nport > 65535)
2158                                         tmp_nport -= 65535;
2159                                 *nport = htons((u_int16_t)tmp_nport);
2160                         } else if (r->rpool.proxy_port[0])
2161                                 *nport = htons(r->rpool.proxy_port[0]);
2162                         break;
2163                 }
2164                 default:
2165                         return (NULL);
2166                 }
2167         }
2168
2169         return (r);
2170 }
2171
2172 #ifdef SMP
2173 struct netmsg_hashlookup {
2174         struct netmsg           nm_netmsg;
2175         struct inpcb            **nm_pinp;
2176         struct inpcbinfo        *nm_pcbinfo;
2177         struct pf_addr          *nm_saddr;
2178         struct pf_addr          *nm_daddr;
2179         uint16_t                nm_sport;
2180         uint16_t                nm_dport;
2181         sa_family_t             nm_af;
2182 };
2183
2184 static void
2185 in_pcblookup_hash_handler(struct netmsg *msg0)
2186 {
2187         struct netmsg_hashlookup *msg = (struct netmsg_hashlookup *)msg0;
2188
2189         if (msg->nm_af == AF_INET)
2190                 *msg->nm_pinp = in_pcblookup_hash(msg->nm_pcbinfo,
2191                     msg->nm_saddr->v4, msg->nm_sport, msg->nm_daddr->v4,
2192                     msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2193 #ifdef INET6
2194         else
2195                 *msg->nm_pinp = in6_pcblookup_hash(msg->nm_pcbinfo,
2196                     &msg->nm_saddr->v6, msg->nm_sport, &msg->nm_daddr->v6,
2197                     msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2198 #endif /* INET6 */
2199         lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
2200 }
2201 #endif /* SMP */
2202
2203 int
2204 pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd)
2205 {
2206         struct pf_addr          *saddr, *daddr;
2207         u_int16_t                sport, dport;
2208         struct inpcbinfo        *pi;
2209         struct inpcb            *inp;
2210 #ifdef SMP
2211         struct netmsg_hashlookup *msg = NULL;
2212 #endif
2213         int                      pi_cpu = 0;
2214
2215         *uid = UID_MAX;
2216         *gid = GID_MAX;
2217         if (direction == PF_IN) {
2218                 saddr = pd->src;
2219                 daddr = pd->dst;
2220         } else {
2221                 saddr = pd->dst;
2222                 daddr = pd->src;
2223         }
2224         switch (pd->proto) {
2225         case IPPROTO_TCP:
2226                 sport = pd->hdr.tcp->th_sport;
2227                 dport = pd->hdr.tcp->th_dport;
2228
2229                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2230                 pi = &tcbinfo[pi_cpu];
2231 #ifdef SMP
2232                 /*
2233                  * Our netstack runs lockless on MP systems
2234                  * (only for TCP connections at the moment).
2235                  *
2236                  * As we are not allowed to read another CPU's tcbinfo,
2237                  * we have to ask that CPU via remote call to search the
2238                  * table for us.
2239                  *
2240                  * Prepare a msg iff data belongs to another CPU.
2241                  */
2242                 if (pi_cpu != mycpu->gd_cpuid) {
2243                         msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2244                         netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0,
2245                                     in_pcblookup_hash_handler);
2246                         msg->nm_pinp = &inp;
2247                         msg->nm_pcbinfo = pi;
2248                         msg->nm_saddr = saddr;
2249                         msg->nm_sport = sport;
2250                         msg->nm_daddr = daddr;
2251                         msg->nm_dport = dport;
2252                         msg->nm_af = pd->af;
2253                 }
2254 #endif /* SMP */
2255                 break;
2256         case IPPROTO_UDP:
2257                 sport = pd->hdr.udp->uh_sport;
2258                 dport = pd->hdr.udp->uh_dport;
2259                 pi = &udbinfo;
2260                 break;
2261         default:
2262                 return (0);
2263         }
2264         if (direction != PF_IN) {
2265                 u_int16_t       p;
2266
2267                 p = sport;
2268                 sport = dport;
2269                 dport = p;
2270         }
2271         switch (pd->af) {
2272 #ifdef INET6
2273         case AF_INET6:
2274 #ifdef SMP
2275                 /*
2276                  * Query other CPU, second part
2277                  *
2278                  * msg only gets initialized when:
2279                  * 1) packet is TCP
2280                  * 2) the info belongs to another CPU
2281                  *
2282                  * Use some switch/case magic to avoid code duplication.
2283                  */
2284                 if (msg == NULL)
2285 #endif /* SMP */
2286                 {
2287                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2288                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2289
2290                         if (inp == NULL)
2291                                 return (0);
2292                         break;
2293                 }
2294                 /* FALLTHROUGH if SMP and on other CPU */
2295 #endif /* INET6 */
2296         case AF_INET:
2297 #ifdef SMP
2298                 if (msg != NULL) {
2299                         lwkt_sendmsg(tcp_cport(pi_cpu),
2300                                      &msg->nm_netmsg.nm_lmsg);
2301                 } else
2302 #endif /* SMP */
2303                 {
2304                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2305                             dport, INPLOOKUP_WILDCARD, NULL);
2306                 }
2307                 if (inp == NULL)
2308                         return (0);
2309                 break;
2310
2311         default:
2312                 return (0);
2313         }
2314         *uid = inp->inp_socket->so_cred->cr_uid;
2315         *gid = inp->inp_socket->so_cred->cr_groups[0];
2316         return (1);
2317 }
2318
2319 u_int8_t
2320 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2321 {
2322         int              hlen;
2323         u_int8_t         hdr[60];
2324         u_int8_t        *opt, optlen;
2325         u_int8_t         wscale = 0;
2326
2327         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
2328         if (hlen <= sizeof(struct tcphdr))
2329                 return (0);
2330         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2331                 return (0);
2332         opt = hdr + sizeof(struct tcphdr);
2333         hlen -= sizeof(struct tcphdr);
2334         while (hlen >= 3) {
2335                 switch (*opt) {
2336                 case TCPOPT_EOL:
2337                 case TCPOPT_NOP:
2338                         ++opt;
2339                         --hlen;
2340                         break;
2341                 case TCPOPT_WINDOW:
2342                         wscale = opt[2];
2343                         if (wscale > TCP_MAX_WINSHIFT)
2344                                 wscale = TCP_MAX_WINSHIFT;
2345                         wscale |= PF_WSCALE_FLAG;
2346                         /* FALLTHROUGH */
2347                 default:
2348                         optlen = opt[1];
2349                         if (optlen < 2)
2350                                 optlen = 2;
2351                         hlen -= optlen;
2352                         opt += optlen;
2353                         break;
2354                 }
2355         }
2356         return (wscale);
2357 }
2358
2359 u_int16_t
2360 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2361 {
2362         int              hlen;
2363         u_int8_t         hdr[60];
2364         u_int8_t        *opt, optlen;
2365         u_int16_t        mss = tcp_mssdflt;
2366
2367         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
2368         if (hlen <= sizeof(struct tcphdr))
2369                 return (0);
2370         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2371                 return (0);
2372         opt = hdr + sizeof(struct tcphdr);
2373         hlen -= sizeof(struct tcphdr);
2374         while (hlen >= TCPOLEN_MAXSEG) {
2375                 switch (*opt) {
2376                 case TCPOPT_EOL:
2377                 case TCPOPT_NOP:
2378                         ++opt;
2379                         --hlen;
2380                         break;
2381                 case TCPOPT_MAXSEG:
2382                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2383                         /* FALLTHROUGH */
2384                 default:
2385                         optlen = opt[1];
2386                         if (optlen < 2)
2387                                 optlen = 2;
2388                         hlen -= optlen;
2389                         opt += optlen;
2390                         break;
2391                 }
2392         }
2393         return (mss);
2394 }
2395
2396 u_int16_t
2397 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2398 {
2399 #ifdef INET
2400         struct sockaddr_in      *dst;
2401         struct route             ro;
2402 #endif /* INET */
2403 #ifdef INET6
2404         struct sockaddr_in6     *dst6;
2405         struct route_in6         ro6;
2406 #endif /* INET6 */
2407         struct rtentry          *rt = NULL;
2408         int                      hlen = 0;
2409         u_int16_t                mss = tcp_mssdflt;
2410
2411         switch (af) {
2412 #ifdef INET
2413         case AF_INET:
2414                 hlen = sizeof(struct ip);
2415                 bzero(&ro, sizeof(ro));
2416                 dst = (struct sockaddr_in *)&ro.ro_dst;
2417                 dst->sin_family = AF_INET;
2418                 dst->sin_len = sizeof(*dst);
2419                 dst->sin_addr = addr->v4;
2420                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2421                 rt = ro.ro_rt;
2422                 break;
2423 #endif /* INET */
2424 #ifdef INET6
2425         case AF_INET6:
2426                 hlen = sizeof(struct ip6_hdr);
2427                 bzero(&ro6, sizeof(ro6));
2428                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2429                 dst6->sin6_family = AF_INET6;
2430                 dst6->sin6_len = sizeof(*dst6);
2431                 dst6->sin6_addr = addr->v6;
2432                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
2433                 rt = ro6.ro_rt;
2434                 break;
2435 #endif /* INET6 */
2436         }
2437
2438         if (rt && rt->rt_ifp) {
2439                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2440                 mss = max(tcp_mssdflt, mss);
2441                 RTFREE(rt);
2442         }
2443         mss = min(mss, offer);
2444         mss = max(mss, 64);             /* sanity - at least max opt space */
2445         return (mss);
2446 }
2447
2448 void
2449 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2450 {
2451         struct pf_rule *r = s->rule.ptr;
2452
2453         s->rt_kif = NULL;
2454         if (!r->rt || r->rt == PF_FASTROUTE)
2455                 return;
2456         switch (s->af) {
2457 #ifdef INET
2458         case AF_INET:
2459                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2460                     &s->nat_src_node);
2461                 s->rt_kif = r->rpool.cur->kif;
2462                 break;
2463 #endif /* INET */
2464 #ifdef INET6
2465         case AF_INET6:
2466                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2467                     &s->nat_src_node);
2468                 s->rt_kif = r->rpool.cur->kif;
2469                 break;
2470 #endif /* INET6 */
2471         }
2472 }
2473
2474 int
2475 pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
2476     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2477     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2478 {
2479         struct pf_rule          *nr = NULL;
2480         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
2481         struct tcphdr           *th = pd->hdr.tcp;
2482         u_int16_t                bport, nport = 0;
2483         sa_family_t              af = pd->af;
2484         int                      lookup = -1;
2485         uid_t                    uid;
2486         gid_t                    gid;
2487         struct pf_rule          *r, *a = NULL;
2488         struct pf_ruleset       *ruleset = NULL;
2489         struct pf_src_node      *nsn = NULL;
2490         u_short                  reason;
2491         int                      rewrite = 0;
2492         int                      tag = -1;
2493         u_int16_t                mss = tcp_mssdflt;
2494
2495         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2496
2497         if (direction == PF_OUT) {
2498                 bport = nport = th->th_sport;
2499                 /* check outgoing packet for BINAT/NAT */
2500                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2501                     saddr, th->th_sport, daddr, th->th_dport,
2502                     &pd->naddr, &nport)) != NULL) {
2503                         PF_ACPY(&pd->baddr, saddr, af);
2504                         pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2505                             &th->th_sum, &pd->naddr, nport, 0, af);
2506                         rewrite++;
2507                         if (nr->natpass)
2508                                 r = NULL;
2509                         pd->nat_rule = nr;
2510                 }
2511         } else {
2512                 bport = nport = th->th_dport;
2513                 /* check incoming packet for BINAT/RDR */
2514                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2515                     saddr, th->th_sport, daddr, th->th_dport,
2516                     &pd->naddr, &nport)) != NULL) {
2517                         PF_ACPY(&pd->baddr, daddr, af);
2518                         pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2519                             &th->th_sum, &pd->naddr, nport, 0, af);
2520                         rewrite++;
2521                         if (nr->natpass)
2522                                 r = NULL;
2523                         pd->nat_rule = nr;
2524                 }
2525         }
2526
2527         while (r != NULL) {
2528                 r->evaluations++;
2529                 if (r->kif != NULL &&
2530                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2531                         r = r->skip[PF_SKIP_IFP].ptr;
2532                 else if (r->direction && r->direction != direction)
2533                         r = r->skip[PF_SKIP_DIR].ptr;
2534                 else if (r->af && r->af != af)
2535                         r = r->skip[PF_SKIP_AF].ptr;
2536                 else if (r->proto && r->proto != IPPROTO_TCP)
2537                         r = r->skip[PF_SKIP_PROTO].ptr;
2538                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2539                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2540                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2541                     r->src.port[0], r->src.port[1], th->th_sport))
2542                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
2543                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2544                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2545                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2546                     r->dst.port[0], r->dst.port[1], th->th_dport))
2547                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2548                 else if (r->tos && !(r->tos & pd->tos))
2549                         r = TAILQ_NEXT(r, entries);
2550                 else if (r->rule_flag & PFRULE_FRAGMENT)
2551                         r = TAILQ_NEXT(r, entries);
2552                 else if ((r->flagset & th->th_flags) != r->flags)
2553                         r = TAILQ_NEXT(r, entries);
2554                 else if (r->uid.op && (lookup != -1 || (lookup =
2555                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2556                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2557                     uid))
2558                         r = TAILQ_NEXT(r, entries);
2559                 else if (r->gid.op && (lookup != -1 || (lookup =
2560                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2561                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2562                     gid))
2563                         r = TAILQ_NEXT(r, entries);
2564                 else if (r->prob && r->prob <= karc4random())
2565                         r = TAILQ_NEXT(r, entries);
2566                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2567                         r = TAILQ_NEXT(r, entries);
2568                 else if (r->anchorname[0] && r->anchor == NULL)
2569                         r = TAILQ_NEXT(r, entries);
2570                 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
2571                     pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
2572                         r = TAILQ_NEXT(r, entries);
2573                 else {
2574                         if (r->tag)
2575                                 tag = r->tag;
2576                         if (r->anchor == NULL) {
2577                                 *rm = r;
2578                                 *am = a;
2579                                 *rsm = ruleset;
2580                                 if ((*rm)->quick)
2581                                         break;
2582                                 r = TAILQ_NEXT(r, entries);
2583                         } else
2584                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2585                                     PF_RULESET_FILTER);
2586                 }
2587                 if (r == NULL && a != NULL)
2588                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2589                             PF_RULESET_FILTER);
2590         }
2591         r = *rm;
2592         a = *am;
2593         ruleset = *rsm;
2594
2595         REASON_SET(&reason, PFRES_MATCH);
2596
2597         if (r->log) {
2598                 if (rewrite)
2599                         m_copyback(m, off, sizeof(*th), (caddr_t)th);
2600                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2601         }
2602
2603         if ((r->action == PF_DROP) &&
2604             ((r->rule_flag & PFRULE_RETURNRST) ||
2605             (r->rule_flag & PFRULE_RETURNICMP) ||
2606             (r->rule_flag & PFRULE_RETURN))) {
2607                 /* undo NAT changes, if they have taken place */
2608                 if (nr != NULL) {
2609                         if (direction == PF_OUT) {
2610                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2611                                     &th->th_sum, &pd->baddr, bport, 0, af);
2612                                 rewrite++;
2613                         } else {
2614                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2615                                     &th->th_sum, &pd->baddr, bport, 0, af);
2616                                 rewrite++;
2617                         }
2618                 }
2619                 if (((r->rule_flag & PFRULE_RETURNRST) ||
2620                     (r->rule_flag & PFRULE_RETURN)) &&
2621                     !(th->th_flags & TH_RST)) {
2622                         u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
2623
2624                         if (th->th_flags & TH_SYN)
2625                                 ack++;
2626                         if (th->th_flags & TH_FIN)
2627                                 ack++;
2628                         pf_send_tcp(r, af, pd->dst,
2629                             pd->src, th->th_dport, th->th_sport,
2630                             ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
2631                             r->return_ttl);
2632                 } else if ((af == AF_INET) && r->return_icmp)
2633                         pf_send_icmp(m, r->return_icmp >> 8,
2634                             r->return_icmp & 255, af, r);
2635                 else if ((af == AF_INET6) && r->return_icmp6)
2636                         pf_send_icmp(m, r->return_icmp6 >> 8,
2637                             r->return_icmp6 & 255, af, r);
2638         }
2639
2640         if (r->action == PF_DROP)
2641                 return (PF_DROP);
2642
2643         pf_tag_packet(m, tag);
2644
2645         if (r->keep_state || nr != NULL ||
2646             (pd->flags & PFDESC_TCP_NORM)) {
2647                 /* create new state */
2648                 u_int16_t        len;
2649                 struct pf_state *s = NULL;
2650                 struct pf_src_node *sn = NULL;
2651
2652                 len = pd->tot_len - off - (th->th_off << 2);
2653
2654                 /* check maximums */
2655                 if (r->max_states && (r->states >= r->max_states))
2656                         goto cleanup;
2657                 /* src node for flter rule */
2658                 if ((r->rule_flag & PFRULE_SRCTRACK ||
2659                     r->rpool.opts & PF_POOL_STICKYADDR) &&
2660                     pf_insert_src_node(&sn, r, saddr, af) != 0)
2661                         goto cleanup;
2662                 /* src node for translation rule */
2663                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2664                     ((direction == PF_OUT &&
2665                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2666                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2667                         goto cleanup;
2668                 s = pool_get(&pf_state_pl, PR_NOWAIT);
2669                 if (s == NULL) {
2670 cleanup:
2671                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2672                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2673                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2674                                 pf_status.src_nodes--;
2675                                 pool_put(&pf_src_tree_pl, sn);
2676                         }
2677                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2678                             nsn->expire == 0) {
2679                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2680                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2681                                 pf_status.src_nodes--;
2682                                 pool_put(&pf_src_tree_pl, nsn);
2683                         }
2684                         REASON_SET(&reason, PFRES_MEMORY);
2685                         return (PF_DROP);
2686                 }
2687                 bzero(s, sizeof(*s));
2688                 r->states++;
2689                 if (a != NULL)
2690                         a->states++;
2691                 s->rule.ptr = r;
2692                 s->nat_rule.ptr = nr;
2693                 if (s->nat_rule.ptr != NULL)
2694                         s->nat_rule.ptr->states++;
2695                 s->anchor.ptr = a;
2696                 s->allow_opts = r->allow_opts;
2697                 s->log = r->log & 2;
2698                 s->proto = IPPROTO_TCP;
2699                 s->direction = direction;
2700                 s->af = af;
2701                 if (direction == PF_OUT) {
2702                         PF_ACPY(&s->gwy.addr, saddr, af);
2703                         s->gwy.port = th->th_sport;             /* sport */
2704                         PF_ACPY(&s->ext.addr, daddr, af);
2705                         s->ext.port = th->th_dport;
2706                         if (nr != NULL) {
2707                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
2708                                 s->lan.port = bport;
2709                         } else {
2710                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2711                                 s->lan.port = s->gwy.port;
2712                         }
2713                 } else {
2714                         PF_ACPY(&s->lan.addr, daddr, af);
2715                         s->lan.port = th->th_dport;
2716                         PF_ACPY(&s->ext.addr, saddr, af);
2717                         s->ext.port = th->th_sport;
2718                         if (nr != NULL) {
2719                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
2720                                 s->gwy.port = bport;
2721                         } else {
2722                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2723                                 s->gwy.port = s->lan.port;
2724                         }
2725                 }
2726
2727                 s->src.seqlo = ntohl(th->th_seq);
2728                 s->src.seqhi = s->src.seqlo + len + 1;
2729                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2730                     r->keep_state == PF_STATE_MODULATE) {
2731                         /* Generate sequence number modulator */
2732                         while ((s->src.seqdiff = karc4random()) == 0)
2733                                 ;
2734                         pf_change_a(&th->th_seq, &th->th_sum,
2735                             htonl(s->src.seqlo + s->src.seqdiff), 0);
2736                         rewrite = 1;
2737                 } else
2738                         s->src.seqdiff = 0;
2739                 if (th->th_flags & TH_SYN) {
2740                         s->src.seqhi++;
2741                         s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
2742                 }
2743                 s->src.max_win = MAX(ntohs(th->th_win), 1);
2744                 if (s->src.wscale & PF_WSCALE_MASK) {
2745                         /* Remove scale factor from initial window */
2746                         int win = s->src.max_win;
2747                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
2748                         s->src.max_win = (win - 1) >>
2749                             (s->src.wscale & PF_WSCALE_MASK);
2750                 }
2751                 if (th->th_flags & TH_FIN)
2752                         s->src.seqhi++;
2753                 s->dst.seqhi = 1;
2754                 s->dst.max_win = 1;
2755                 s->src.state = TCPS_SYN_SENT;
2756                 s->dst.state = TCPS_CLOSED;
2757                 s->creation = time_second;
2758                 s->expire = time_second;
2759                 s->timeout = PFTM_TCP_FIRST_PACKET;
2760                 pf_set_rt_ifp(s, saddr);
2761                 if (sn != NULL) {
2762                         s->src_node = sn;
2763                         s->src_node->states++;
2764                 }
2765                 if (nsn != NULL) {
2766                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
2767                         s->nat_src_node = nsn;
2768                         s->nat_src_node->states++;
2769                 }
2770                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
2771                     off, pd, th, &s->src, &s->dst)) {
2772                         REASON_SET(&reason, PFRES_MEMORY);
2773                         pf_src_tree_remove_state(s);
2774                         pool_put(&pf_state_pl, s);
2775                         return (PF_DROP);
2776                 }
2777                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
2778                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, &s->src,
2779                     &s->dst, &rewrite)) {
2780                         pf_normalize_tcp_cleanup(s);
2781                         pf_src_tree_remove_state(s);
2782                         pool_put(&pf_state_pl, s);
2783                         return (PF_DROP);
2784                 }
2785                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
2786                         pf_normalize_tcp_cleanup(s);
2787                         REASON_SET(&reason, PFRES_MEMORY);
2788                         pf_src_tree_remove_state(s);
2789                         pool_put(&pf_state_pl, s);
2790                         return (PF_DROP);
2791                 } else
2792                         *sm = s;
2793                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2794                     r->keep_state == PF_STATE_SYNPROXY) {
2795                         s->src.state = PF_TCPS_PROXY_SRC;
2796                         if (nr != NULL) {
2797                                 if (direction == PF_OUT) {
2798                                         pf_change_ap(saddr, &th->th_sport,
2799                                             pd->ip_sum, &th->th_sum, &pd->baddr,
2800                                             bport, 0, af);
2801                                 } else {
2802                                         pf_change_ap(daddr, &th->th_dport,
2803                                             pd->ip_sum, &th->th_sum, &pd->baddr,
2804                                             bport, 0, af);
2805                                 }
2806                         }
2807                         s->src.seqhi = karc4random();
2808                         /* Find mss option */
2809                         mss = pf_get_mss(m, off, th->th_off, af);
2810                         mss = pf_calc_mss(saddr, af, mss);
2811                         mss = pf_calc_mss(daddr, af, mss);
2812                         s->src.mss = mss;
2813                         pf_send_tcp(r, af, daddr, saddr, th->th_dport,
2814                             th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
2815                             TH_SYN|TH_ACK, 0, s->src.mss, 0);
2816                         return (PF_SYNPROXY_DROP);
2817                 }
2818         }
2819
2820         /* copy back packet headers if we performed NAT operations */
2821         if (rewrite)
2822                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2823
2824         return (PF_PASS);
2825 }
2826
2827 int
2828 pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
2829     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2830     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2831 {
2832         struct pf_rule          *nr = NULL;
2833         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
2834         struct udphdr           *uh = pd->hdr.udp;
2835         u_int16_t                bport, nport = 0;
2836         sa_family_t              af = pd->af;
2837         int                      lookup = -1;
2838         uid_t                    uid;
2839         gid_t                    gid;
2840         struct pf_rule          *r, *a = NULL;
2841         struct pf_ruleset       *ruleset = NULL;
2842         struct pf_src_node      *nsn = NULL;
2843         u_short                  reason;
2844         int                      rewrite = 0;
2845         int                      tag = -1;
2846
2847         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2848
2849         if (direction == PF_OUT) {
2850                 bport = nport = uh->uh_sport;
2851                 /* check outgoing packet for BINAT/NAT */
2852                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2853                     saddr, uh->uh_sport, daddr, uh->uh_dport,
2854                     &pd->naddr, &nport)) != NULL) {
2855                         PF_ACPY(&pd->baddr, saddr, af);
2856                         pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2857                             &uh->uh_sum, &pd->naddr, nport, 1, af);
2858                         rewrite++;
2859                         if (nr->natpass)
2860                                 r = NULL;
2861                         pd->nat_rule = nr;
2862                 }
2863         } else {
2864                 bport = nport = uh->uh_dport;
2865                 /* check incoming packet for BINAT/RDR */
2866                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2867                     saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
2868                     &nport)) != NULL) {
2869                         PF_ACPY(&pd->baddr, daddr, af);
2870                         pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2871                             &uh->uh_sum, &pd->naddr, nport, 1, af);
2872                         rewrite++;
2873                         if (nr->natpass)
2874                                 r = NULL;
2875                         pd->nat_rule = nr;
2876                 }
2877         }
2878
2879         while (r != NULL) {
2880                 r->evaluations++;
2881                 if (r->kif != NULL &&
2882                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2883                         r = r->skip[PF_SKIP_IFP].ptr;
2884                 else if (r->direction && r->direction != direction)
2885                         r = r->skip[PF_SKIP_DIR].ptr;
2886                 else if (r->af && r->af != af)
2887                         r = r->skip[PF_SKIP_AF].ptr;
2888                 else if (r->proto && r->proto != IPPROTO_UDP)
2889                         r = r->skip[PF_SKIP_PROTO].ptr;
2890                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2891                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2892                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2893                     r->src.port[0], r->src.port[1], uh->uh_sport))
2894                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
2895                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2896                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2897                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2898                     r->dst.port[0], r->dst.port[1], uh->uh_dport))
2899                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2900                 else if (r->tos && !(r->tos & pd->tos))
2901                         r = TAILQ_NEXT(r, entries);
2902                 else if (r->rule_flag & PFRULE_FRAGMENT)
2903                         r = TAILQ_NEXT(r, entries);
2904                 else if (r->uid.op && (lookup != -1 || (lookup =
2905                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2906                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2907                     uid))
2908                         r = TAILQ_NEXT(r, entries);
2909                 else if (r->gid.op && (lookup != -1 || (lookup =
2910                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2911                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2912                     gid))
2913                         r = TAILQ_NEXT(r, entries);
2914                 else if (r->prob && r->prob <= karc4random())
2915                         r = TAILQ_NEXT(r, entries);
2916                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2917                         r = TAILQ_NEXT(r, entries);
2918                 else if (r->anchorname[0] && r->anchor == NULL)
2919                         r = TAILQ_NEXT(r, entries);
2920                 else if (r->os_fingerprint != PF_OSFP_ANY)
2921                         r = TAILQ_NEXT(r, entries);
2922                 else {
2923                         if (r->tag)
2924                                 tag = r->tag;
2925                         if (r->anchor == NULL) {
2926                                 *rm = r;
2927                                 *am = a;
2928                                 *rsm = ruleset;
2929                                 if ((*rm)->quick)
2930                                         break;
2931                                 r = TAILQ_NEXT(r, entries);
2932                         } else
2933                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2934                                     PF_RULESET_FILTER);
2935                 }
2936                 if (r == NULL && a != NULL)
2937                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2938                             PF_RULESET_FILTER);
2939         }
2940         r = *rm;
2941         a = *am;
2942         ruleset = *rsm;
2943
2944         REASON_SET(&reason, PFRES_MATCH);
2945
2946         if (r->log) {
2947                 if (rewrite)
2948                         m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
2949                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2950         }
2951
2952         if ((r->action == PF_DROP) &&
2953             ((r->rule_flag & PFRULE_RETURNICMP) ||
2954             (r->rule_flag & PFRULE_RETURN))) {
2955                 /* undo NAT changes, if they have taken place */
2956                 if (nr != NULL) {
2957                         if (direction == PF_OUT) {
2958                                 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2959                                     &uh->uh_sum, &pd->baddr, bport, 1, af);
2960                                 rewrite++;
2961                         } else {
2962                                 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2963                                     &uh->uh_sum, &pd->baddr, bport, 1, af);
2964                                 rewrite++;
2965                         }
2966                 }
2967                 if ((af == AF_INET) && r->return_icmp)
2968                         pf_send_icmp(m, r->return_icmp >> 8,
2969                             r->return_icmp & 255, af, r);
2970                 else if ((af == AF_INET6) && r->return_icmp6)
2971                         pf_send_icmp(m, r->return_icmp6 >> 8,
2972                             r->return_icmp6 & 255, af, r);
2973         }
2974
2975         if (r->action == PF_DROP)
2976                 return (PF_DROP);
2977
2978         pf_tag_packet(m, tag);
2979
2980         if (r->keep_state || nr != NULL) {
2981                 /* create new state */
2982                 struct pf_state *s = NULL;
2983                 struct pf_src_node *sn = NULL;
2984
2985                 /* check maximums */
2986                 if (r->max_states && (r->states >= r->max_states))
2987                         goto cleanup;
2988                 /* src node for flter rule */
2989                 if ((r->rule_flag & PFRULE_SRCTRACK ||
2990                     r->rpool.opts & PF_POOL_STICKYADDR) &&
2991                     pf_insert_src_node(&sn, r, saddr, af) != 0)
2992                         goto cleanup;
2993                 /* src node for translation rule */
2994                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2995                     ((direction == PF_OUT &&
2996                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2997                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2998                         goto cleanup;
2999                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3000                 if (s == NULL) {
3001 cleanup:
3002                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3003                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3004                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3005                                 pf_status.src_nodes--;
3006                                 pool_put(&pf_src_tree_pl, sn);
3007                         }
3008                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3009                             nsn->expire == 0) {
3010                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3011                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3012                                 pf_status.src_nodes--;
3013                                 pool_put(&pf_src_tree_pl, nsn);
3014                         }
3015                         REASON_SET(&reason, PFRES_MEMORY);
3016                         return (PF_DROP);
3017                 }
3018                 bzero(s, sizeof(*s));
3019                 r->states++;
3020                 if (a != NULL)
3021                         a->states++;
3022                 s->rule.ptr = r;
3023                 s->nat_rule.ptr = nr;
3024                 if (s->nat_rule.ptr != NULL)
3025                         s->nat_rule.ptr->states++;
3026                 s->anchor.ptr = a;
3027                 s->allow_opts = r->allow_opts;
3028                 s->log = r->log & 2;
3029                 s->proto = IPPROTO_UDP;
3030                 s->direction = direction;
3031                 s->af = af;
3032                 if (direction == PF_OUT) {
3033                         PF_ACPY(&s->gwy.addr, saddr, af);
3034                         s->gwy.port = uh->uh_sport;
3035                         PF_ACPY(&s->ext.addr, daddr, af);
3036                         s->ext.port = uh->uh_dport;
3037                         if (nr != NULL) {
3038                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3039                                 s->lan.port = bport;
3040                         } else {
3041                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3042                                 s->lan.port = s->gwy.port;
3043                         }
3044                 } else {
3045                         PF_ACPY(&s->lan.addr, daddr, af);
3046                         s->lan.port = uh->uh_dport;
3047                         PF_ACPY(&s->ext.addr, saddr, af);
3048                         s->ext.port = uh->uh_sport;
3049                         if (nr != NULL) {
3050                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3051                                 s->gwy.port = bport;
3052                         } else {
3053                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3054                                 s->gwy.port = s->lan.port;
3055                         }
3056                 }
3057                 s->src.state = PFUDPS_SINGLE;
3058                 s->dst.state = PFUDPS_NO_TRAFFIC;
3059                 s->creation = time_second;
3060                 s->expire = time_second;
3061                 s->timeout = PFTM_UDP_FIRST_PACKET;
3062                 pf_set_rt_ifp(s, saddr);
3063                 if (sn != NULL) {
3064                         s->src_node = sn;
3065                         s->src_node->states++;
3066                 }
3067                 if (nsn != NULL) {
3068                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3069                         s->nat_src_node = nsn;
3070                         s->nat_src_node->states++;
3071                 }
3072                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3073                         REASON_SET(&reason, PFRES_MEMORY);
3074                         pf_src_tree_remove_state(s);
3075                         pool_put(&pf_state_pl, s);
3076                         return (PF_DROP);
3077                 } else
3078                         *sm = s;
3079         }
3080
3081         /* copy back packet headers if we performed NAT operations */
3082         if (rewrite)
3083                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3084
3085         return (PF_PASS);
3086 }
3087
3088 int
3089 pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3090     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3091     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3092 {
3093         struct pf_rule          *nr = NULL;
3094         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3095         struct pf_rule          *r, *a = NULL;
3096         struct pf_ruleset       *ruleset = NULL;
3097         struct pf_src_node      *nsn = NULL;
3098         u_short                  reason;
3099         u_int16_t                icmpid = 0;
3100         sa_family_t              af = pd->af;
3101         u_int8_t                 icmptype = 0, icmpcode = 0;
3102         int                      state_icmp = 0;
3103         int                      tag = -1;
3104 #ifdef INET6
3105         int                      rewrite = 0;
3106 #endif /* INET6 */
3107
3108         switch (pd->proto) {
3109 #ifdef INET
3110         case IPPROTO_ICMP:
3111                 icmptype = pd->hdr.icmp->icmp_type;
3112                 icmpcode = pd->hdr.icmp->icmp_code;
3113                 icmpid = pd->hdr.icmp->icmp_id;
3114
3115                 if (icmptype == ICMP_UNREACH ||
3116                     icmptype == ICMP_SOURCEQUENCH ||
3117                     icmptype == ICMP_REDIRECT ||
3118                     icmptype == ICMP_TIMXCEED ||
3119                     icmptype == ICMP_PARAMPROB)
3120                         state_icmp++;
3121                 break;
3122 #endif /* INET */
3123 #ifdef INET6
3124         case IPPROTO_ICMPV6:
3125                 icmptype = pd->hdr.icmp6->icmp6_type;
3126                 icmpcode = pd->hdr.icmp6->icmp6_code;
3127                 icmpid = pd->hdr.icmp6->icmp6_id;
3128
3129                 if (icmptype == ICMP6_DST_UNREACH ||
3130                     icmptype == ICMP6_PACKET_TOO_BIG ||
3131                     icmptype == ICMP6_TIME_EXCEEDED ||
3132                     icmptype == ICMP6_PARAM_PROB)
3133                         state_icmp++;
3134                 break;
3135 #endif /* INET6 */
3136         }
3137
3138         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3139
3140         if (direction == PF_OUT) {
3141                 /* check outgoing packet for BINAT/NAT */
3142                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3143                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3144                         PF_ACPY(&pd->baddr, saddr, af);
3145                         switch (af) {
3146 #ifdef INET
3147                         case AF_INET:
3148                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3149                                     pd->naddr.v4.s_addr, 0);
3150                                 break;
3151 #endif /* INET */
3152 #ifdef INET6
3153                         case AF_INET6:
3154                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3155                                     &pd->naddr, 0);
3156                                 rewrite++;
3157                                 break;
3158 #endif /* INET6 */
3159                         }
3160                         if (nr->natpass)
3161                                 r = NULL;
3162                         pd->nat_rule = nr;
3163                 }
3164         } else {
3165                 /* check incoming packet for BINAT/RDR */
3166                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3167                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3168                         PF_ACPY(&pd->baddr, daddr, af);
3169                         switch (af) {
3170 #ifdef INET
3171                         case AF_INET:
3172                                 pf_change_a(&daddr->v4.s_addr,
3173                                     pd->ip_sum, pd->naddr.v4.s_addr, 0);
3174                                 break;
3175 #endif /* INET */
3176 #ifdef INET6
3177                         case AF_INET6:
3178                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3179                                     &pd->naddr, 0);
3180                                 rewrite++;
3181                                 break;
3182 #endif /* INET6 */
3183                         }
3184                         if (nr->natpass)
3185                                 r = NULL;
3186                         pd->nat_rule = nr;
3187                 }
3188         }
3189
3190         while (r != NULL) {
3191                 r->evaluations++;
3192                 if (r->kif != NULL &&
3193                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3194                         r = r->skip[PF_SKIP_IFP].ptr;
3195                 else if (r->direction && r->direction != direction)
3196                         r = r->skip[PF_SKIP_DIR].ptr;
3197                 else if (r->af && r->af != af)
3198                         r = r->skip[PF_SKIP_AF].ptr;
3199                 else if (r->proto && r->proto != pd->proto)
3200                         r = r->skip[PF_SKIP_PROTO].ptr;
3201                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3202                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3203                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3204                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3205                 else if (r->type && r->type != icmptype + 1)
3206                         r = TAILQ_NEXT(r, entries);
3207                 else if (r->code && r->code != icmpcode + 1)
3208                         r = TAILQ_NEXT(r, entries);
3209                 else if (r->tos && !(r->tos & pd->tos))
3210                         r = TAILQ_NEXT(r, entries);
3211                 else if (r->rule_flag & PFRULE_FRAGMENT)
3212                         r = TAILQ_NEXT(r, entries);
3213                 else if (r->prob && r->prob <= karc4random())
3214                         r = TAILQ_NEXT(r, entries);
3215                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3216                         r = TAILQ_NEXT(r, entries);
3217                 else if (r->anchorname[0] && r->anchor == NULL)
3218                         r = TAILQ_NEXT(r, entries);
3219                 else if (r->os_fingerprint != PF_OSFP_ANY)
3220                         r = TAILQ_NEXT(r, entries);
3221                 else {
3222                         if (r->tag)
3223                                 tag = r->tag;
3224                         if (r->anchor == NULL) {
3225                                 *rm = r;
3226                                 *am = a;
3227                                 *rsm = ruleset;
3228                                 if ((*rm)->quick)
3229                                         break;
3230                                 r = TAILQ_NEXT(r, entries);
3231                         } else
3232                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3233                                     PF_RULESET_FILTER);
3234                 }
3235                 if (r == NULL && a != NULL)
3236                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3237                             PF_RULESET_FILTER);
3238         }
3239         r = *rm;
3240         a = *am;
3241         ruleset = *rsm;
3242
3243         REASON_SET(&reason, PFRES_MATCH);
3244
3245         if (r->log) {
3246 #ifdef INET6
3247                 if (rewrite)
3248                         m_copyback(m, off, sizeof(struct icmp6_hdr),
3249                             (caddr_t)pd->hdr.icmp6);
3250 #endif /* INET6 */
3251                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3252         }
3253
3254         if (r->action != PF_PASS)
3255                 return (PF_DROP);
3256
3257         pf_tag_packet(m, tag);
3258
3259         if (!state_icmp && (r->keep_state || nr != NULL)) {
3260                 /* create new state */
3261                 struct pf_state *s = NULL;
3262                 struct pf_src_node *sn = NULL;
3263
3264                 /* check maximums */
3265                 if (r->max_states && (r->states >= r->max_states))
3266                         goto cleanup;
3267                 /* src node for flter rule */
3268                 if ((r->rule_flag & PFRULE_SRCTRACK ||
3269                     r->rpool.opts & PF_POOL_STICKYADDR) &&
3270                     pf_insert_src_node(&sn, r, saddr, af) != 0)
3271                         goto cleanup;
3272                 /* src node for translation rule */
3273                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3274                     ((direction == PF_OUT &&
3275                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3276                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3277                         goto cleanup;
3278                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3279                 if (s == NULL) {
3280 cleanup:
3281                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3282                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3283                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3284                                 pf_status.src_nodes--;
3285                                 pool_put(&pf_src_tree_pl, sn);
3286                         }
3287                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3288                             nsn->expire == 0) {
3289                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3290                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3291                                 pf_status.src_nodes--;
3292                                 pool_put(&pf_src_tree_pl, nsn);
3293                         }
3294                         REASON_SET(&reason, PFRES_MEMORY);
3295                         return (PF_DROP);
3296                 }
3297                 bzero(s, sizeof(*s));
3298                 r->states++;
3299                 if (a != NULL)
3300                         a->states++;
3301                 s->rule.ptr = r;
3302                 s->nat_rule.ptr = nr;
3303                 if (s->nat_rule.ptr != NULL)
3304                         s->nat_rule.ptr->states++;
3305                 s->anchor.ptr = a;
3306                 s->allow_opts = r->allow_opts;
3307                 s->log = r->log & 2;
3308                 s->proto = pd->proto;
3309                 s->direction = direction;
3310                 s->af = af;
3311                 if (direction == PF_OUT) {
3312                         PF_ACPY(&s->gwy.addr, saddr, af);
3313                         s->gwy.port = icmpid;
3314                         PF_ACPY(&s->ext.addr, daddr, af);
3315                         s->ext.port = icmpid;
3316                         if (nr != NULL)
3317                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3318                         else
3319                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3320                         s->lan.port = icmpid;
3321                 } else {
3322                         PF_ACPY(&s->lan.addr, daddr, af);
3323                         s->lan.port = icmpid;
3324                         PF_ACPY(&s->ext.addr, saddr, af);
3325                         s->ext.port = icmpid;
3326                         if (nr != NULL)
3327                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3328                         else
3329                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3330                         s->gwy.port = icmpid;
3331                 }
3332                 s->creation = time_second;
3333                 s->expire = time_second;
3334                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3335                 pf_set_rt_ifp(s, saddr);
3336                 if (sn != NULL) {
3337                         s->src_node = sn;
3338                         s->src_node->states++;
3339                 }
3340                 if (nsn != NULL) {
3341                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3342                         s->nat_src_node = nsn;
3343                         s->nat_src_node->states++;
3344                 }
3345                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3346                         REASON_SET(&reason, PFRES_MEMORY);
3347                         pf_src_tree_remove_state(s);
3348                         pool_put(&pf_state_pl, s);
3349                         return (PF_DROP);
3350                 } else
3351                         *sm = s;
3352         }
3353
3354 #ifdef INET6
3355         /* copy back packet headers if we performed IPv6 NAT operations */
3356         if (rewrite)
3357                 m_copyback(m, off, sizeof(struct icmp6_hdr),
3358                     (caddr_t)pd->hdr.icmp6);
3359 #endif /* INET6 */
3360
3361         return (PF_PASS);
3362 }
3363
3364 int
3365 pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3366     struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3367     struct pf_rule **am, struct pf_ruleset **rsm)
3368 {
3369         struct pf_rule          *nr = NULL;
3370         struct pf_rule          *r, *a = NULL;
3371         struct pf_ruleset       *ruleset = NULL;
3372         struct pf_src_node      *nsn = NULL;
3373         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3374         sa_family_t              af = pd->af;
3375         u_short                  reason;
3376         int                      tag = -1;
3377
3378         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3379
3380         if (direction == PF_OUT) {
3381                 /* check outgoing packet for BINAT/NAT */
3382                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3383                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3384                         PF_ACPY(&pd->baddr, saddr, af);
3385                         switch (af) {
3386 #ifdef INET
3387                         case AF_INET:
3388                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3389                                     pd->naddr.v4.s_addr, 0);
3390                                 break;
3391 #endif /* INET */
3392 #ifdef INET6
3393                         case AF_INET6:
3394                                 PF_ACPY(saddr, &pd->naddr, af);
3395                                 break;
3396 #endif /* INET6 */
3397                         }
3398                         if (nr->natpass)
3399                                 r = NULL;
3400                         pd->nat_rule = nr;
3401                 }
3402         } else {
3403                 /* check incoming packet for BINAT/RDR */
3404                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3405                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3406                         PF_ACPY(&pd->baddr, daddr, af);
3407                         switch (af) {
3408 #ifdef INET
3409                         case AF_INET:
3410                                 pf_change_a(&daddr->v4.s_addr,
3411                                     pd->ip_sum, pd->naddr.v4.s_addr, 0);
3412                                 break;
3413 #endif /* INET */
3414 #ifdef INET6
3415                         case AF_INET6:
3416                                 PF_ACPY(daddr, &pd->naddr, af);
3417                                 break;
3418 #endif /* INET6 */
3419                         }
3420                         if (nr->natpass)
3421                                 r = NULL;
3422                         pd->nat_rule = nr;
3423                 }
3424         }
3425
3426         while (r != NULL) {
3427                 r->evaluations++;
3428                 if (r->kif != NULL &&
3429                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3430                         r = r->skip[PF_SKIP_IFP].ptr;
3431                 else if (r->direction && r->direction != direction)
3432                         r = r->skip[PF_SKIP_DIR].ptr;
3433                 else if (r->af && r->af != af)
3434                         r = r->skip[PF_SKIP_AF].ptr;
3435                 else if (r->proto && r->proto != pd->proto)
3436                         r = r->skip[PF_SKIP_PROTO].ptr;
3437                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3438                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3439                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3440                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3441                 else if (r->tos && !(r->tos & pd->tos))
3442                         r = TAILQ_NEXT(r, entries);
3443                 else if (r->rule_flag & PFRULE_FRAGMENT)
3444                         r = TAILQ_NEXT(r, entries);
3445                 else if (r->prob && r->prob <= karc4random())
3446                         r = TAILQ_NEXT(r, entries);
3447                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3448                         r = TAILQ_NEXT(r, entries);
3449                 else if (r->anchorname[0] && r->anchor == NULL)
3450                         r = TAILQ_NEXT(r, entries);
3451                 else if (r->os_fingerprint != PF_OSFP_ANY)
3452                         r = TAILQ_NEXT(r, entries);
3453                 else {
3454                         if (r->tag)
3455                                 tag = r->tag;
3456                         if (r->anchor == NULL) {
3457                                 *rm = r;
3458                                 *am = a;
3459                                 *rsm = ruleset;
3460                                 if ((*rm)->quick)
3461                                         break;
3462                                 r = TAILQ_NEXT(r, entries);
3463                         } else
3464                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3465                                     PF_RULESET_FILTER);
3466                 }
3467                 if (r == NULL && a != NULL)
3468                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3469                             PF_RULESET_FILTER);
3470         }
3471         r = *rm;
3472         a = *am;
3473         ruleset = *rsm;
3474
3475         REASON_SET(&reason, PFRES_MATCH);
3476
3477         if (r->log)
3478                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3479
3480         if ((r->action == PF_DROP) &&
3481             ((r->rule_flag & PFRULE_RETURNICMP) ||
3482             (r->rule_flag & PFRULE_RETURN))) {
3483                 struct pf_addr *a = NULL;
3484
3485                 if (nr != NULL) {
3486                         if (direction == PF_OUT)
3487                                 a = saddr;
3488                         else
3489                                 a = daddr;
3490                 }
3491                 if (a != NULL) {
3492                         switch (af) {
3493 #ifdef INET
3494                         case AF_INET:
3495                                 pf_change_a(&a->v4.s_addr, pd->ip_sum,
3496                                     pd->baddr.v4.s_addr, 0);
3497                                 break;
3498 #endif /* INET */
3499 #ifdef INET6
3500                         case AF_INET6:
3501                                 PF_ACPY(a, &pd->baddr, af);
3502                                 break;
3503 #endif /* INET6 */
3504                         }
3505                 }
3506                 if ((af == AF_INET) && r->return_icmp)
3507                         pf_send_icmp(m, r->return_icmp >> 8,
3508                             r->return_icmp & 255, af, r);
3509                 else if ((af == AF_INET6) && r->return_icmp6)
3510                         pf_send_icmp(m, r->return_icmp6 >> 8,
3511                             r->return_icmp6 & 255, af, r);
3512         }
3513
3514         if (r->action != PF_PASS)
3515                 return (PF_DROP);
3516
3517         pf_tag_packet(m, tag);
3518
3519         if (r->keep_state || nr != NULL) {
3520                 /* create new state */
3521                 struct pf_state *s = NULL;
3522                 struct pf_src_node *sn = NULL;
3523
3524                 /* check maximums */
3525                 if (r->max_states && (r->states >= r->max_states))
3526                         goto cleanup;
3527                 /* src node for flter rule */
3528                 if ((r->rule_flag & PFRULE_SRCTRACK ||
3529                     r->rpool.opts & PF_POOL_STICKYADDR) &&
3530                     pf_insert_src_node(&sn, r, saddr, af) != 0)
3531                         goto cleanup;
3532                 /* src node for translation rule */
3533                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3534                     ((direction == PF_OUT &&
3535                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3536                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3537                         goto cleanup;
3538                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3539                 if (s == NULL) {
3540 cleanup:
3541                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3542                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3543                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3544                                 pf_status.src_nodes--;
3545                                 pool_put(&pf_src_tree_pl, sn);
3546                         }
3547                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3548                             nsn->expire == 0) {
3549                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3550                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3551                                 pf_status.src_nodes--;
3552                                 pool_put(&pf_src_tree_pl, nsn);
3553                         }
3554                         REASON_SET(&reason, PFRES_MEMORY);
3555                         return (PF_DROP);
3556                 }
3557                 bzero(s, sizeof(*s));
3558                 r->states++;
3559                 if (a != NULL)
3560                         a->states++;
3561                 s->rule.ptr = r;
3562                 s->nat_rule.ptr = nr;
3563                 if (s->nat_rule.ptr != NULL)
3564                         s->nat_rule.ptr->states++;
3565                 s->anchor.ptr = a;
3566                 s->allow_opts = r->allow_opts;
3567                 s->log = r->log & 2;
3568                 s->proto = pd->proto;
3569                 s->direction = direction;
3570                 s->af = af;
3571                 if (direction == PF_OUT) {
3572                         PF_ACPY(&s->gwy.addr, saddr, af);
3573                         PF_ACPY(&s->ext.addr, daddr, af);
3574                         if (nr != NULL)
3575                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3576                         else
3577                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3578                 } else {
3579                         PF_ACPY(&s->lan.addr, daddr, af);
3580                         PF_ACPY(&s->ext.addr, saddr, af);
3581                         if (nr != NULL)
3582                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3583                         else
3584                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3585                 }
3586                 s->src.state = PFOTHERS_SINGLE;
3587                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3588                 s->creation = time_second;
3589                 s->expire = time_second;
3590                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3591                 pf_set_rt_ifp(s, saddr);
3592                 if (sn != NULL) {
3593                         s->src_node = sn;
3594                         s->src_node->states++;
3595                 }
3596                 if (nsn != NULL) {
3597                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3598                         s->nat_src_node = nsn;
3599                         s->nat_src_node->states++;
3600                 }
3601                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3602                         REASON_SET(&reason, PFRES_MEMORY);
3603                         pf_src_tree_remove_state(s);
3604                         pool_put(&pf_state_pl, s);
3605                         return (PF_DROP);
3606                 } else
3607                         *sm = s;
3608         }
3609
3610         return (PF_PASS);
3611 }
3612
3613 int
3614 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3615     struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3616     struct pf_ruleset **rsm)
3617 {
3618         struct pf_rule          *r, *a = NULL;
3619         struct pf_ruleset       *ruleset = NULL;
3620         sa_family_t              af = pd->af;
3621         u_short                  reason;
3622         int                      tag = -1;
3623
3624         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3625         while (r != NULL) {
3626                 r->evaluations++;
3627                 if (r->kif != NULL &&
3628                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3629                         r = r->skip[PF_SKIP_IFP].ptr;
3630                 else if (r->direction && r->direction != direction)
3631                         r = r->skip[PF_SKIP_DIR].ptr;
3632                 else if (r->af && r->af != af)
3633                         r = r->skip[PF_SKIP_AF].ptr;
3634                 else if (r->proto && r->proto != pd->proto)
3635                         r = r->skip[PF_SKIP_PROTO].ptr;
3636                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3637                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3638                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3639                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3640                 else if (r->tos && !(r->tos & pd->tos))
3641                         r = TAILQ_NEXT(r, entries);
3642                 else if (r->src.port_op || r->dst.port_op ||
3643                     r->flagset || r->type || r->code ||
3644                     r->os_fingerprint != PF_OSFP_ANY)
3645                         r = TAILQ_NEXT(r, entries);
3646                 else if (r->prob && r->prob <= karc4random())
3647                         r = TAILQ_NEXT(r, entries);
3648                 else if (r->match_tag && !pf_match_tag(m, r, NULL, &tag))
3649                         r = TAILQ_NEXT(r, entries);
3650                 else if (r->anchorname[0] && r->anchor == NULL)
3651                         r = TAILQ_NEXT(r, entries);
3652                 else {
3653                         if (r->anchor == NULL) {
3654                                 *rm = r;
3655                                 *am = a;
3656                                 *rsm = ruleset;
3657                                 if ((*rm)->quick)
3658                                         break;
3659                                 r = TAILQ_NEXT(r, entries);
3660                         } else
3661                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3662                                     PF_RULESET_FILTER);
3663                 }
3664                 if (r == NULL && a != NULL)
3665                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3666                             PF_RULESET_FILTER);
3667         }
3668         r = *rm;
3669         a = *am;
3670         ruleset = *rsm;
3671
3672         REASON_SET(&reason, PFRES_MATCH);
3673
3674         if (r->log)
3675                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3676
3677         if (r->action != PF_PASS)
3678                 return (PF_DROP);
3679
3680         pf_tag_packet(m, tag);
3681
3682         return (PF_PASS);
3683 }
3684
3685 int
3686 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3687     struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3688     u_short *reason)
3689 {
3690         struct pf_state          key;
3691         struct tcphdr           *th = pd->hdr.tcp;
3692         u_int16_t                win = ntohs(th->th_win);
3693         u_int32_t                ack, end, seq;
3694         u_int8_t                 sws, dws;
3695         int                      ackskew;
3696         int                      copyback = 0;
3697         struct pf_state_peer    *src, *dst;
3698
3699         key.af = pd->af;
3700         key.proto = IPPROTO_TCP;
3701         if (direction == PF_IN) {
3702                 PF_ACPY(&key.ext.addr, pd->src, key.af);
3703                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3704                 key.ext.port = th->th_sport;
3705                 key.gwy.port = th->th_dport;
3706         } else {
3707                 PF_ACPY(&key.lan.addr, pd->src, key.af);
3708                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3709                 key.lan.port = th->th_sport;
3710                 key.ext.port = th->th_dport;
3711         }
3712
3713         STATE_LOOKUP();
3714
3715         if (direction == (*state)->direction) {
3716                 src = &(*state)->src;
3717                 dst = &(*state)->dst;
3718         } else {
3719                 src = &(*state)->dst;
3720                 dst = &(*state)->src;
3721         }
3722
3723         if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3724                 if (direction != (*state)->direction)
3725                         return (PF_SYNPROXY_DROP);
3726                 if (th->th_flags & TH_SYN) {
3727                         if (ntohl(th->th_seq) != (*state)->src.seqlo)
3728                                 return (PF_DROP);
3729                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3730                             pd->src, th->th_dport, th->th_sport,
3731                             (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3732                             TH_SYN|TH_ACK, 0, (*state)->src.mss, 0);
3733                         return (PF_SYNPROXY_DROP);
3734                 } else if (!(th->th_flags & TH_ACK) ||
3735                     (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3736                     (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3737                         return (PF_DROP);
3738                 else
3739                         (*state)->src.state = PF_TCPS_PROXY_DST;
3740         }
3741         if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3742                 struct pf_state_host *src, *dst;
3743
3744                 if (direction == PF_OUT) {
3745                         src = &(*state)->gwy;
3746                         dst = &(*state)->ext;
3747                 } else {
3748                         src = &(*state)->ext;
3749                         dst = &(*state)->lan;
3750                 }
3751                 if (direction == (*state)->direction) {
3752                         if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3753                             (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3754                             (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3755                                 return (PF_DROP);
3756                         (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3757                         if ((*state)->dst.seqhi == 1)
3758                                 (*state)->dst.seqhi = karc4random();
3759                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3760                             &dst->addr, src->port, dst->port,
3761                             (*state)->dst.seqhi, 0, TH_SYN, 0,
3762                             (*state)->src.mss, 0);
3763                         return (PF_SYNPROXY_DROP);
3764                 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3765                     (TH_SYN|TH_ACK)) ||
3766                     (ntohl(th->th_ack) != (*state)->dst.seqhi + 1))
3767                         return (PF_DROP);
3768                 else {
3769                         (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3770                         (*state)->dst.seqlo = ntohl(th->th_seq);
3771                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3772                             pd->src, th->th_dport, th->th_sport,
3773                             ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3774                             TH_ACK, (*state)->src.max_win, 0, 0);
3775                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3776                             &dst->addr, src->port, dst->port,
3777                             (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3778                             TH_ACK, (*state)->dst.max_win, 0, 0);
3779                         (*state)->src.seqdiff = (*state)->dst.seqhi -
3780                             (*state)->src.seqlo;
3781                         (*state)->dst.seqdiff = (*state)->src.seqhi -
3782                             (*state)->dst.seqlo;
3783                         (*state)->src.seqhi = (*state)->src.seqlo +
3784                             (*state)->src.max_win;
3785                         (*state)->dst.seqhi = (*state)->dst.seqlo +
3786                             (*state)->dst.max_win;
3787                         (*state)->src.wscale = (*state)->dst.wscale = 0;
3788                         (*state)->src.state = (*state)->dst.state =
3789                             TCPS_ESTABLISHED;
3790                         return (PF_SYNPROXY_DROP);
3791                 }
3792         }
3793
3794         if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3795                 sws = src->wscale & PF_WSCALE_MASK;
3796                 dws = dst->wscale & PF_WSCALE_MASK;
3797         } else
3798                 sws = dws = 0;
3799
3800         /*
3801          * Sequence tracking algorithm from Guido van Rooij's paper:
3802          *   http://www.madison-gurkha.com/publications/tcp_filtering/
3803          *      tcp_filtering.ps
3804          */
3805
3806         seq = ntohl(th->th_seq);
3807         if (src->seqlo == 0) {
3808                 /* First packet from this end. Set its state */
3809
3810                 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3811                     src->scrub == NULL) {
3812                         if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3813                                 REASON_SET(reason, PFRES_MEMORY);
3814                                 return (PF_DROP);
3815                         }
3816                 }
3817
3818                 /* Deferred generation of sequence number modulator */
3819                 if (dst->seqdiff && !src->seqdiff) {
3820                         while ((src->seqdiff = karc4random()) == 0)
3821                                 ;
3822                         ack = ntohl(th->th_ack) - dst->seqdiff;
3823                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3824                             src->seqdiff), 0);
3825                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3826                         copyback = 1;
3827                 } else {
3828                         ack = ntohl(th->th_ack);
3829                 }
3830
3831                 end = seq + pd->p_len;
3832                 if (th->th_flags & TH_SYN) {
3833                         end++;
3834                         if (dst->wscale & PF_WSCALE_FLAG) {
3835                                 src->wscale = pf_get_wscale(m, off, th->th_off,
3836                                     pd->af);
3837                                 if (src->wscale & PF_WSCALE_FLAG) {
3838                                         /* Remove scale factor from initial
3839                                          * window */
3840                                         sws = src->wscale & PF_WSCALE_MASK;
3841                                         win = ((u_int32_t)win + (1 << sws) - 1)
3842                                             >> sws;
3843                                         dws = dst->wscale & PF_WSCALE_MASK;
3844                                 } else {
3845                                         /* fixup other window */
3846                                         dst->max_win <<= dst->wscale &
3847                                             PF_WSCALE_MASK;
3848                                         /* in case of a retrans SYN|ACK */
3849                                         dst->wscale = 0;
3850                                 }
3851                         }
3852                 }
3853                 if (th->th_flags & TH_FIN)
3854                         end++;
3855
3856                 src->seqlo = seq;
3857                 if (src->state < TCPS_SYN_SENT)
3858                         src->state = TCPS_SYN_SENT;
3859
3860                 /*
3861                  * May need to slide the window (seqhi may have been set by
3862                  * the crappy stack check or if we picked up the connection
3863                  * after establishment)
3864                  */
3865                 if (src->seqhi == 1 ||
3866                     SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3867                         src->seqhi = end + MAX(1, dst->max_win << dws);
3868                 if (win > src->max_win)
3869                         src->max_win = win;
3870
3871         } else {
3872                 ack = ntohl(th->th_ack) - dst->seqdiff;
3873                 if (src->seqdiff) {
3874                         /* Modulate sequence numbers */
3875                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3876                             src->seqdiff), 0);
3877                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3878                         copyback = 1;
3879                 }
3880                 end = seq + pd->p_len;
3881                 if (th->th_flags & TH_SYN)
3882                         end++;
3883                 if (th->th_flags & TH_FIN)
3884                         end++;
3885         }
3886
3887         if ((th->th_flags & TH_ACK) == 0) {
3888                 /* Let it pass through the ack skew check */
3889                 ack = dst->seqlo;
3890         } else if ((ack == 0 &&
3891             (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3892             /* broken tcp stacks do not set ack */
3893             (dst->state < TCPS_SYN_SENT)) {
3894                 /*
3895                  * Many stacks (ours included) will set the ACK number in an
3896                  * FIN|ACK if the SYN times out -- no sequence to ACK.
3897                  */
3898                 ack = dst->seqlo;
3899         }
3900
3901         if (seq == end) {
3902                 /* Ease sequencing restrictions on no data packets */
3903                 seq = src->seqlo;
3904                 end = seq;
3905         }
3906
3907         ackskew = dst->seqlo - ack;
3908
3909 #define MAXACKWINDOW (0xffff + 1500)    /* 1500 is an arbitrary fudge factor */
3910         if (SEQ_GEQ(src->seqhi, end) &&
3911             /* Last octet inside other's window space */
3912             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3913             /* Retrans: not more than one window back */
3914             (ackskew >= -MAXACKWINDOW) &&
3915             /* Acking not more than one reassembled fragment backwards */
3916             (ackskew <= (MAXACKWINDOW << sws))) {
3917             /* Acking not more than one window forward */
3918
3919                 /* update max window */
3920                 if (src->max_win < win)
3921                         src->max_win = win;
3922                 /* synchronize sequencing */
3923                 if (SEQ_GT(end, src->seqlo))
3924                         src->seqlo = end;
3925                 /* slide the window of what the other end can send */
3926                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3927                         dst->seqhi = ack + MAX((win << sws), 1);
3928
3929
3930                 /* update states */
3931                 if (th->th_flags & TH_SYN)
3932                         if (src->state < TCPS_SYN_SENT)
3933                                 src->state = TCPS_SYN_SENT;
3934                 if (th->th_flags & TH_FIN)
3935                         if (src->state < TCPS_CLOSING)
3936                                 src->state = TCPS_CLOSING;
3937                 if (th->th_flags & TH_ACK) {
3938                         if (dst->state == TCPS_SYN_SENT)
3939                                 dst->state = TCPS_ESTABLISHED;
3940                         else if (dst->state == TCPS_CLOSING)
3941                                 dst->state = TCPS_FIN_WAIT_2;
3942                 }
3943                 if (th->th_flags & TH_RST)
3944                         src->state = dst->state = TCPS_TIME_WAIT;
3945
3946                 /* update expire time */
3947                 (*state)->expire = time_second;
3948                 if (src->state >= TCPS_FIN_WAIT_2 &&
3949                     dst->state >= TCPS_FIN_WAIT_2)
3950                         (*state)->timeout = PFTM_TCP_CLOSED;
3951                 else if (src->state >= TCPS_FIN_WAIT_2 ||
3952                     dst->state >= TCPS_FIN_WAIT_2)
3953                         (*state)->timeout = PFTM_TCP_FIN_WAIT;
3954                 else if (src->state < TCPS_ESTABLISHED ||
3955                     dst->state < TCPS_ESTABLISHED)
3956                         (*state)->timeout = PFTM_TCP_OPENING;
3957                 else if (src->state >= TCPS_CLOSING ||
3958                     dst->state >= TCPS_CLOSING)
3959                         (*state)->timeout = PFTM_TCP_CLOSING;
3960                 else
3961                         (*state)->timeout = PFTM_TCP_ESTABLISHED;
3962
3963                 /* Fall through to PASS packet */
3964
3965         } else if ((dst->state < TCPS_SYN_SENT ||
3966                 dst->state >= TCPS_FIN_WAIT_2 ||
3967                 src->state >= TCPS_FIN_WAIT_2) &&
3968             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
3969             /* Within a window forward of the originating packet */
3970             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
3971             /* Within a window backward of the originating packet */
3972
3973                 /*
3974                  * This currently handles three situations:
3975                  *  1) Stupid stacks will shotgun SYNs before their peer
3976                  *     replies.
3977                  *  2) When PF catches an already established stream (the
3978                  *     firewall rebooted, the state table was flushed, routes
3979                  *     changed...)
3980                  *  3) Packets get funky immediately after the connection
3981                  *     closes (this should catch Solaris spurious ACK|FINs
3982                  *     that web servers like to spew after a close)
3983                  *
3984                  * This must be a little more careful than the above code
3985                  * since packet floods will also be caught here. We don't
3986                  * update the TTL here to mitigate the damage of a packet
3987                  * flood and so the same code can handle awkward establishment
3988                  * and a loosened connection close.
3989                  * In the establishment case, a correct peer response will
3990                  * validate the connection, go through the normal state code
3991                  * and keep updating the state TTL.
3992                  */
3993
3994                 if (pf_status.debug >= PF_DEBUG_MISC) {
3995                         kprintf("pf: loose state match: ");
3996                         pf_print_state(*state);
3997                         pf_print_flags(th->th_flags);
3998                         kprintf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n",
3999                             seq, ack, pd->p_len, ackskew,
4000                             (*state)->packets[0], (*state)->packets[1]);
4001                 }
4002
4003                 /* update max window */
4004                 if (src->max_win < win)
4005                         src->max_win = win;
4006                 /* synchronize sequencing */
4007                 if (SEQ_GT(end, src->seqlo))
4008                         src->seqlo = end;
4009                 /* slide the window of what the other end can send */
4010                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
4011                         dst->seqhi = ack + MAX((win << sws), 1);
4012
4013                 /*
4014                  * Cannot set dst->seqhi here since this could be a shotgunned
4015                  * SYN and not an already established connection.
4016                  */
4017
4018                 if (th->th_flags & TH_FIN)
4019                         if (src->state < TCPS_CLOSING)
4020                                 src->state = TCPS_CLOSING;
4021                 if (th->th_flags & TH_RST)
4022                         src->state = dst->state = TCPS_TIME_WAIT;
4023
4024                 /* Fall through to PASS packet */
4025
4026         } else {
4027                 if ((*state)->dst.state == TCPS_SYN_SENT &&
4028                     (*state)->src.state == TCPS_SYN_SENT) {
4029                         /* Send RST for state mismatches during handshake */
4030                         if (!(th->th_flags & TH_RST)) {
4031                                 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
4032
4033                                 if (th->th_flags & TH_SYN)
4034                                         ack++;
4035                                 if (th->th_flags & TH_FIN)
4036                                         ack++;
4037                                 pf_send_tcp((*state)->rule.ptr, pd->af,
4038                                     pd->dst, pd->src, th->th_dport,
4039                                     th->th_sport, ntohl(th->th_ack), ack,
4040                                     TH_RST|TH_ACK, 0, 0,
4041                                     (*state)->rule.ptr->return_ttl);
4042                         }
4043                         src->seqlo = 0;
4044                         src->seqhi = 1;
4045                         src->max_win = 1;
4046                 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4047                         kprintf("pf: BAD state: ");
4048                         pf_print_state(*state);
4049                         pf_print_flags(th->th_flags);
4050                         kprintf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d "
4051                             "dir=%s,%s\n", seq, ack, pd->p_len, ackskew,
4052                             (*state)->packets[0], (*state)->packets[1],
4053                             direction == PF_IN ? "in" : "out",
4054                             direction == (*state)->direction ? "fwd" : "rev");
4055                         kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4056                             SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4057                             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4058                             ' ': '2',
4059                             (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4060                             (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4061                             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4062                             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4063                 }
4064                 return (PF_DROP);
4065         }
4066
4067         if (dst->scrub || src->scrub) {
4068                 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4069                     src, dst, &copyback))
4070                         return (PF_DROP);
4071         }
4072
4073         /* Any packets which have gotten here are to be passed */
4074
4075         /* translate source/destination address, if necessary */
4076         if (STATE_TRANSLATE(*state)) {
4077                 if (direction == PF_OUT)
4078                         pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4079                             &th->th_sum, &(*state)->gwy.addr,
4080                             (*state)->gwy.port, 0, pd->af);
4081                 else
4082                         pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4083                             &th->th_sum, &(*state)->lan.addr,
4084                             (*state)->lan.port, 0, pd->af);
4085                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4086         } else if (copyback) {
4087                 /* Copyback sequence modulation or stateful scrub changes */
4088                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4089         }
4090
4091         return (PF_PASS);
4092 }
4093
4094 int
4095 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4096     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4097 {
4098         struct pf_state_peer    *src, *dst;
4099         struct pf_state          key;
4100         struct udphdr           *uh = pd->hdr.udp;
4101
4102         key.af = pd->af;
4103         key.proto = IPPROTO_UDP;
4104         if (direction == PF_IN) {
4105                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4106                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4107                 key.ext.port = uh->uh_sport;
4108                 key.gwy.port = uh->uh_dport;
4109         } else {
4110                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4111                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4112                 key.lan.port = uh->uh_sport;
4113                 key.ext.port = uh->uh_dport;
4114         }
4115
4116         STATE_LOOKUP();
4117
4118         if (direction == (*state)->direction) {
4119                 src = &(*state)->src;
4120                 dst = &(*state)->dst;
4121         } else {
4122                 src = &(*state)->dst;
4123                 dst = &(*state)->src;
4124         }
4125
4126         /* update states */
4127         if (src->state < PFUDPS_SINGLE)
4128                 src->state = PFUDPS_SINGLE;
4129         if (dst->state == PFUDPS_SINGLE)
4130                 dst->state = PFUDPS_MULTIPLE;
4131
4132         /* update expire time */
4133         (*state)->expire = time_second;
4134         if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4135                 (*state)->timeout = PFTM_UDP_MULTIPLE;
4136         else
4137                 (*state)->timeout = PFTM_UDP_SINGLE;
4138
4139         /* translate source/destination address, if necessary */
4140         if (STATE_TRANSLATE(*state)) {
4141                 if (direction == PF_OUT)
4142                         pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4143                             &uh->uh_sum, &(*state)->gwy.addr,
4144                             (*state)->gwy.port, 1, pd->af);
4145                 else
4146                         pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4147                             &uh->uh_sum, &(*state)->lan.addr,
4148                             (*state)->lan.port, 1, pd->af);
4149                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4150         }
4151
4152         return (PF_PASS);
4153 }
4154
4155 int
4156 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4157     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4158 {
4159         struct pf_addr  *saddr = pd->src, *daddr = pd->dst;
4160         u_int16_t        icmpid = 0;
4161         u_int16_t       *icmpsum = NULL;
4162         u_int8_t         icmptype = 0;
4163         int              state_icmp = 0;
4164
4165         switch (pd->proto) {
4166 #ifdef INET
4167         case IPPROTO_ICMP:
4168                 icmptype = pd->hdr.icmp->icmp_type;
4169                 icmpid = pd->hdr.icmp->icmp_id;
4170                 icmpsum = &pd->hdr.icmp->icmp_cksum;
4171
4172                 if (icmptype == ICMP_UNREACH ||
4173                     icmptype == ICMP_SOURCEQUENCH ||
4174                     icmptype == ICMP_REDIRECT ||
4175                     icmptype == ICMP_TIMXCEED ||
4176                     icmptype == ICMP_PARAMPROB)
4177                         state_icmp++;
4178                 break;
4179 #endif /* INET */
4180 #ifdef INET6
4181         case IPPROTO_ICMPV6:
4182                 icmptype = pd->hdr.icmp6->icmp6_type;
4183                 icmpid = pd->hdr.icmp6->icmp6_id;
4184                 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4185
4186                 if (icmptype == ICMP6_DST_UNREACH ||
4187                     icmptype == ICMP6_PACKET_TOO_BIG ||
4188                     icmptype == ICMP6_TIME_EXCEEDED ||
4189                     icmptype == ICMP6_PARAM_PROB)
4190                         state_icmp++;
4191                 break;
4192 #endif /* INET6 */
4193         }
4194
4195         if (!state_icmp) {
4196
4197                 /*
4198                  * ICMP query/reply message not related to a TCP/UDP packet.
4199                  * Search for an ICMP state.
4200                  */
4201                 struct pf_state         key;
4202
4203                 key.af = pd->af;
4204                 key.proto = pd->proto;
4205                 if (direction == PF_IN) {
4206                         PF_ACPY(&key.ext.addr, pd->src, key.af);
4207                         PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4208                         key.ext.port = icmpid;
4209                         key.gwy.port = icmpid;
4210                 } else {
4211                         PF_ACPY(&key.lan.addr, pd->src, key.af);
4212                         PF_ACPY(&key.ext.addr, pd->dst, key.af);
4213                         key.lan.port = icmpid;
4214                         key.ext.port = icmpid;
4215                 }
4216
4217                 STATE_LOOKUP();
4218
4219                 (*state)->expire = time_second;
4220                 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4221
4222                 /* translate source/destination address, if necessary */
4223                 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) {
4224                         if (direction == PF_OUT) {
4225                                 switch (pd->af) {
4226 #ifdef INET
4227                                 case AF_INET:
4228                                         pf_change_a(&saddr->v4.s_addr,
4229                                             pd->ip_sum,
4230                                             (*state)->gwy.addr.v4.s_addr, 0);
4231                                         break;
4232 #endif /* INET */
4233 #ifdef INET6
4234                                 case AF_INET6:
4235                                         pf_change_a6(saddr,
4236                                             &pd->hdr.icmp6->icmp6_cksum,
4237                                             &(*state)->gwy.addr, 0);
4238                                         m_copyback(m, off,
4239                                             sizeof(struct icmp6_hdr),
4240                                             (caddr_t)pd->hdr.icmp6);
4241                                         break;
4242 #endif /* INET6 */
4243                                 }
4244                         } else {
4245                                 switch (pd->af) {
4246 #ifdef INET
4247                                 case AF_INET:
4248                                         pf_change_a(&daddr->v4.s_addr,
4249                                             pd->ip_sum,
4250                                             (*state)->lan.addr.v4.s_addr, 0);
4251                                         break;
4252 #endif /* INET */
4253 #ifdef INET6
4254                                 case AF_INET6:
4255                                         pf_change_a6(daddr,
4256                                             &pd->hdr.icmp6->icmp6_cksum,
4257                                             &(*state)->lan.addr, 0);
4258                                         m_copyback(m, off,
4259                                             sizeof(struct icmp6_hdr),
4260                                             (caddr_t)pd->hdr.icmp6);
4261                                         break;
4262 #endif /* INET6 */
4263                                 }
4264                         }
4265                 }
4266
4267                 return (PF_PASS);
4268
4269         } else {
4270                 /*
4271                  * ICMP error message in response to a TCP/UDP packet.
4272                  * Extract the inner TCP/UDP header and search for that state.
4273                  */
4274
4275                 struct pf_pdesc pd2;
4276 #ifdef INET
4277                 struct ip       h2;
4278 #endif /* INET */
4279 #ifdef INET6
4280                 struct ip6_hdr  h2_6;
4281                 int             terminal = 0;
4282 #endif /* INET6 */
4283                 int             ipoff2 = 0;
4284                 int             off2 = 0;
4285
4286                 pd2.af = pd->af;
4287                 switch (pd->af) {
4288 #ifdef INET
4289                 case AF_INET:
4290                         /* offset of h2 in mbuf chain */
4291                         ipoff2 = off + ICMP_MINLEN;
4292
4293                         if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4294                             NULL, NULL, pd2.af)) {
4295                                 DPFPRINTF(PF_DEBUG_MISC,
4296                                     ("pf: ICMP error message too short "
4297                                     "(ip)\n"));
4298                                 return (PF_DROP);
4299                         }
4300                         /*
4301                          * ICMP error messages don't refer to non-first
4302                          * fragments
4303                          */
4304                         /*
4305                          * Note: We are dealing with an encapsulated
4306                          * header. This means ip_off/ip_len are not
4307                          * in host byte order!
4308                          */
4309                         if (h2.ip_off & htons(IP_OFFMASK))
4310                                 return (PF_DROP);
4311
4312                         /* offset of protocol header that follows h2 */
4313                         off2 = ipoff2 + (h2.ip_hl << 2);
4314
4315                         pd2.proto = h2.ip_p;
4316                         pd2.src = (struct pf_addr *)&h2.ip_src;
4317                         pd2.dst = (struct pf_addr *)&h2.ip_dst;
4318                         pd2.ip_sum = &h2.ip_sum;
4319                         break;
4320 #endif /* INET */
4321 #ifdef INET6
4322                 case AF_INET6:
4323                         ipoff2 = off + sizeof(struct icmp6_hdr);
4324
4325                         if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4326                             NULL, NULL, pd2.af)) {
4327                                 DPFPRINTF(PF_DEBUG_MISC,
4328                                     ("pf: ICMP error message too short "
4329                                     "(ip6)\n"));
4330                                 return (PF_DROP);
4331                         }
4332                         pd2.proto = h2_6.ip6_nxt;
4333                         pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4334                         pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4335                         pd2.ip_sum = NULL;
4336                         off2 = ipoff2 + sizeof(h2_6);
4337                         do {
4338                                 switch (pd2.proto) {
4339                                 case IPPROTO_FRAGMENT:
4340                                         /*
4341                                          * ICMPv6 error messages for
4342                                          * non-first fragments
4343                                          */
4344                                         return (PF_DROP);
4345                                 case IPPROTO_AH:
4346                                 case IPPROTO_HOPOPTS:
4347                                 case IPPROTO_ROUTING:
4348                                 case IPPROTO_DSTOPTS: {
4349                                         /* get next header and header length */
4350                                         struct ip6_ext opt6;
4351
4352                                         if (!pf_pull_hdr(m, off2, &opt6,
4353                                             sizeof(opt6), NULL, NULL, pd2.af)) {
4354                                                 DPFPRINTF(PF_DEBUG_MISC,
4355                                                     ("pf: ICMPv6 short opt\n"));
4356                                                 return (PF_DROP);
4357                                         }
4358                                         if (pd2.proto == IPPROTO_AH)
4359                                                 off2 += (opt6.ip6e_len + 2) * 4;
4360                                         else
4361                                                 off2 += (opt6.ip6e_len + 1) * 8;
4362                                         pd2.proto = opt6.ip6e_nxt;
4363                                         /* goto the next header */
4364                                         break;
4365                                 }
4366                                 default:
4367                                         terminal++;
4368                                         break;
4369                                 }
4370                         } while (!terminal);
4371                         break;
4372 #endif /* INET6 */
4373                 }
4374
4375                 switch (pd2.proto) {
4376                 case IPPROTO_TCP: {
4377                         struct tcphdr            th;
4378                         u_int32_t                seq;
4379                         struct pf_state          key;
4380                         struct pf_state_peer    *src, *dst;
4381                         u_int8_t                 dws;
4382                         int                      copyback = 0;
4383
4384                         /*
4385                          * Only the first 8 bytes of the TCP header can be
4386                          * expected. Don't access any TCP header fields after
4387                          * th_seq, an ackskew test is not possible.
4388                          */
4389                         if (!pf_pull_hdr(m, off2, &th, 8, NULL, NULL, pd2.af)) {
4390                                 DPFPRINTF(PF_DEBUG_MISC,
4391                                     ("pf: ICMP error message too short "
4392                                     "(tcp)\n"));
4393                                 return (PF_DROP);
4394                         }
4395
4396                         key.af = pd2.af;
4397                         key.proto = IPPROTO_TCP;
4398                         if (direction == PF_IN) {
4399                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4400                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4401                                 key.ext.port = th.th_dport;
4402                                 key.gwy.port = th.th_sport;
4403                         } else {
4404                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4405                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4406                                 key.lan.port = th.th_dport;
4407                                 key.ext.port = th.th_sport;
4408                         }
4409
4410                         STATE_LOOKUP();
4411
4412                         if (direction == (*state)->direction) {
4413                                 src = &(*state)->dst;
4414                                 dst = &(*state)->src;
4415                         } else {
4416                                 src = &(*state)->src;
4417                                 dst = &(*state)->dst;
4418                         }
4419
4420                         if (src->wscale && dst->wscale &&
4421                             !(th.th_flags & TH_SYN))
4422                                 dws = dst->wscale & PF_WSCALE_MASK;
4423                         else
4424                                 dws = 0;
4425
4426                         /* Demodulate sequence number */
4427                         seq = ntohl(th.th_seq) - src->seqdiff;
4428                         if (src->seqdiff) {
4429                                 pf_change_a(&th.th_seq, icmpsum,
4430                                     htonl(seq), 0);
4431                                 copyback = 1;
4432                         }
4433
4434                         if (!SEQ_GEQ(src->seqhi, seq) ||
4435                             !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4436                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4437                                         kprintf("pf: BAD ICMP %d:%d ",
4438                                             icmptype, pd->hdr.icmp->icmp_code);
4439                                         pf_print_host(pd->src, 0, pd->af);
4440                                         kprintf(" -> ");
4441                                         pf_print_host(pd->dst, 0, pd->af);
4442                                         kprintf(" state: ");
4443                                         pf_print_state(*state);
4444                                         kprintf(" seq=%u\n", seq);
4445                                 }
4446                                 return (PF_DROP);
4447                         }
4448
4449                         if (STATE_TRANSLATE(*state)) {
4450                                 if (direction == PF_IN) {
4451                                         pf_change_icmp(pd2.src, &th.th_sport,
4452                                             daddr, &(*state)->lan.addr,
4453                                             (*state)->lan.port, NULL,
4454                                             pd2.ip_sum, icmpsum,
4455                                             pd->ip_sum, 0, pd2.af);
4456                                 } else {
4457                                         pf_change_icmp(pd2.dst, &th.th_dport,
4458                                             saddr, &(*state)->gwy.addr,
4459                                             (*state)->gwy.port, NULL,
4460                                             pd2.ip_sum, icmpsum,
4461                                             pd->ip_sum, 0, pd2.af);
4462                                 }
4463                                 copyback = 1;
4464                         }
4465
4466                         if (copyback) {
4467                                 switch (pd2.af) {
4468 #ifdef INET
4469                                 case AF_INET:
4470                                         m_copyback(m, off, ICMP_MINLEN,
4471                                             (caddr_t)pd->hdr.icmp);
4472                                         m_copyback(m, ipoff2, sizeof(h2),
4473                                             (caddr_t)&h2);
4474                                         break;
4475 #endif /* INET */
4476 #ifdef INET6
4477                                 case AF_INET6:
4478                                         m_copyback(m, off,
4479                                             sizeof(struct icmp6_hdr),
4480                                             (caddr_t)pd->hdr.icmp6);
4481                                         m_copyback(m, ipoff2, sizeof(h2_6),
4482                                             (caddr_t)&h2_6);
4483                                         break;
4484 #endif /* INET6 */
4485                                 }
4486                                 m_copyback(m, off2, 8, (caddr_t)&th);
4487                         }
4488
4489                         return (PF_PASS);
4490                         break;
4491                 }
4492                 case IPPROTO_UDP: {
4493                         struct udphdr           uh;
4494                         struct pf_state         key;
4495
4496                         if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4497                             NULL, NULL, pd2.af)) {
4498                                 DPFPRINTF(PF_DEBUG_MISC,
4499                                     ("pf: ICMP error message too short "
4500                                     "(udp)\n"));
4501                                 return (PF_DROP);
4502                         }
4503
4504                         key.af = pd2.af;
4505                         key.proto = IPPROTO_UDP;
4506                         if (direction == PF_IN) {
4507                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4508                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4509                                 key.ext.port = uh.uh_dport;
4510                                 key.gwy.port = uh.uh_sport;
4511                         } else {
4512                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4513                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4514                                 key.lan.port = uh.uh_dport;
4515                                 key.ext.port = uh.uh_sport;
4516                         }
4517
4518                         STATE_LOOKUP();
4519
4520                         if (STATE_TRANSLATE(*state)) {
4521                                 if (direction == PF_IN) {
4522                                         pf_change_icmp(pd2.src, &uh.uh_sport,
4523                                             daddr, &(*state)->lan.addr,
4524                                             (*state)->lan.port, &uh.uh_sum,
4525                                             pd2.ip_sum, icmpsum,
4526                                             pd->ip_sum, 1, pd2.af);
4527                                 } else {
4528                                         pf_change_icmp(pd2.dst, &uh.uh_dport,
4529                                             saddr, &(*state)->gwy.addr,
4530                                             (*state)->gwy.port, &uh.uh_sum,
4531                                             pd2.ip_sum, icmpsum,
4532                                             pd->ip_sum, 1, pd2.af);
4533                                 }
4534                                 switch (pd2.af) {
4535 #ifdef INET
4536                                 case AF_INET:
4537                                         m_copyback(m, off, ICMP_MINLEN,
4538                                             (caddr_t)pd->hdr.icmp);
4539                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4540                                         break;
4541 #endif /* INET */
4542 #ifdef INET6
4543                                 case AF_INET6:
4544                                         m_copyback(m, off,
4545                                             sizeof(struct icmp6_hdr),
4546                                             (caddr_t)pd->hdr.icmp6);
4547                                         m_copyback(m, ipoff2, sizeof(h2_6),
4548                                             (caddr_t)&h2_6);
4549                                         break;
4550 #endif /* INET6 */
4551                                 }
4552                                 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4553                         }
4554
4555                         return (PF_PASS);
4556                         break;
4557                 }
4558 #ifdef INET
4559                 case IPPROTO_ICMP: {
4560                         struct icmp             iih;
4561                         struct pf_state         key;
4562
4563                         if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4564                             NULL, NULL, pd2.af)) {
4565                                 DPFPRINTF(PF_DEBUG_MISC,
4566                                     ("pf: ICMP error message too short i"
4567                                     "(icmp)\n"));
4568                                 return (PF_DROP);
4569                         }
4570
4571                         key.af = pd2.af;
4572                         key.proto = IPPROTO_ICMP;
4573                         if (direction == PF_IN) {
4574                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4575                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4576                                 key.ext.port = iih.icmp_id;
4577                                 key.gwy.port = iih.icmp_id;
4578                         } else {
4579                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4580                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4581                                 key.lan.port = iih.icmp_id;
4582                                 key.ext.port = iih.icmp_id;
4583                         }
4584
4585                         STATE_LOOKUP();
4586
4587                         if (STATE_TRANSLATE(*state)) {
4588                                 if (direction == PF_IN) {
4589                                         pf_change_icmp(pd2.src, &iih.icmp_id,
4590                                             daddr, &(*state)->lan.addr,
4591                                             (*state)->lan.port, NULL,
4592                                             pd2.ip_sum, icmpsum,
4593                                             pd->ip_sum, 0, AF_INET);
4594                                 } else {
4595                                         pf_change_icmp(pd2.dst, &iih.icmp_id,
4596                                             saddr, &(*state)->gwy.addr,
4597                                             (*state)->gwy.port, NULL,
4598                                             pd2.ip_sum, icmpsum,
4599                                             pd->ip_sum, 0, AF_INET);
4600                                 }
4601                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4602                                 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4603                                 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4604                         }
4605
4606                         return (PF_PASS);
4607                         break;
4608                 }
4609 #endif /* INET */
4610 #ifdef INET6
4611                 case IPPROTO_ICMPV6: {
4612                         struct icmp6_hdr        iih;
4613                         struct pf_state         key;
4614
4615                         if (!pf_pull_hdr(m, off2, &iih,
4616                             sizeof(struct icmp6_hdr), NULL, NULL, pd2.af)) {
4617                                 DPFPRINTF(PF_DEBUG_MISC,
4618                                     ("pf: ICMP error message too short "
4619                                     "(icmp6)\n"));
4620                                 return (PF_DROP);
4621                         }
4622
4623                         key.af = pd2.af;
4624                         key.proto = IPPROTO_ICMPV6;
4625                         if (direction == PF_IN) {
4626                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4627                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4628                                 key.ext.port = iih.icmp6_id;
4629                                 key.gwy.port = iih.icmp6_id;
4630                         } else {
4631                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4632                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4633                                 key.lan.port = iih.icmp6_id;
4634                                 key.ext.port = iih.icmp6_id;
4635                         }
4636
4637                         STATE_LOOKUP();
4638
4639                         if (STATE_TRANSLATE(*state)) {
4640                                 if (direction == PF_IN) {
4641                                         pf_change_icmp(pd2.src, &iih.icmp6_id,
4642                                             daddr, &(*state)->lan.addr,
4643                                             (*state)->lan.port, NULL,
4644                                             pd2.ip_sum, icmpsum,
4645                                             pd->ip_sum, 0, AF_INET6);
4646                                 } else {
4647                                         pf_change_icmp(pd2.dst, &iih.icmp6_id,
4648                                             saddr, &(*state)->gwy.addr,
4649                                             (*state)->gwy.port, NULL,
4650                                             pd2.ip_sum, icmpsum,
4651                                             pd->ip_sum, 0, AF_INET6);
4652                                 }
4653                                 m_copyback(m, off, sizeof(struct icmp6_hdr),
4654                                     (caddr_t)pd->hdr.icmp6);
4655                                 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
4656                                 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4657                                     (caddr_t)&iih);
4658                         }
4659
4660                         return (PF_PASS);
4661                         break;
4662                 }
4663 #endif /* INET6 */
4664                 default: {
4665                         struct pf_state         key;
4666
4667                         key.af = pd2.af;
4668                         key.proto = pd2.proto;
4669                         if (direction == PF_IN) {
4670                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4671                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4672                                 key.ext.port = 0;
4673                                 key.gwy.port = 0;
4674                         } else {
4675                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4676                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4677                                 key.lan.port = 0;
4678                                 key.ext.port = 0;
4679                         }
4680
4681                         STATE_LOOKUP();
4682
4683                         if (STATE_TRANSLATE(*state)) {
4684                                 if (direction == PF_IN) {
4685                                         pf_change_icmp(pd2.src, NULL,
4686                                             daddr, &(*state)->lan.addr,
4687                                             0, NULL,
4688                                             pd2.ip_sum, icmpsum,
4689                                             pd->ip_sum, 0, pd2.af);
4690                                 } else {
4691                                         pf_change_icmp(pd2.dst, NULL,
4692                                             saddr, &(*state)->gwy.addr,
4693                                             0, NULL,
4694                                             pd2.ip_sum, icmpsum,
4695                                             pd->ip_sum, 0, pd2.af);
4696                                 }
4697                                 switch (pd2.af) {
4698 #ifdef INET
4699                                 case AF_INET:
4700                                         m_copyback(m, off, ICMP_MINLEN,
4701                                             (caddr_t)pd->hdr.icmp);
4702                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4703                                         break;
4704 #endif /* INET */
4705 #ifdef INET6
4706                                 case AF_INET6:
4707                                         m_copyback(m, off,
4708                                             sizeof(struct icmp6_hdr),
4709                                             (caddr_t)pd->hdr.icmp6);
4710                                         m_copyback(m, ipoff2, sizeof(h2_6),
4711                                             (caddr_t)&h2_6);
4712                                         break;
4713 #endif /* INET6 */
4714                                 }
4715                         }
4716
4717                         return (PF_PASS);
4718                         break;
4719                 }
4720                 }
4721         }
4722 }
4723
4724 int
4725 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4726     struct pf_pdesc *pd)
4727 {
4728         struct pf_state_peer    *src, *dst;
4729         struct pf_state          key;
4730
4731         key.af = pd->af;
4732         key.proto = pd->proto;
4733         if (direction == PF_IN) {
4734                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4735                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4736                 key.ext.port = 0;
4737                 key.gwy.port = 0;
4738         } else {
4739                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4740                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4741                 key.lan.port = 0;
4742                 key.ext.port = 0;
4743         }
4744
4745         STATE_LOOKUP();
4746
4747         if (direction == (*state)->direction) {
4748                 src = &(*state)->src;
4749                 dst = &(*state)->dst;
4750         } else {
4751                 src = &(*state)->dst;
4752                 dst = &(*state)->src;
4753         }
4754
4755         /* update states */
4756         if (src->state < PFOTHERS_SINGLE)
4757                 src->state = PFOTHERS_SINGLE;
4758         if (dst->state == PFOTHERS_SINGLE)
4759                 dst->state = PFOTHERS_MULTIPLE;
4760
4761         /* update expire time */
4762         (*state)->expire = time_second;
4763         if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4764                 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4765         else
4766                 (*state)->timeout = PFTM_OTHER_SINGLE;
4767
4768         /* translate source/destination address, if necessary */
4769         if (STATE_TRANSLATE(*state)) {
4770                 if (direction == PF_OUT)
4771                         switch (pd->af) {
4772 #ifdef INET
4773                         case AF_INET:
4774                                 pf_change_a(&pd->src->v4.s_addr,
4775                                     pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
4776                                     0);
4777                                 break;
4778 #endif /* INET */
4779 #ifdef INET6
4780                         case AF_INET6:
4781                                 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
4782                                 break;
4783 #endif /* INET6 */
4784                         }
4785                 else
4786                         switch (pd->af) {
4787 #ifdef INET
4788                         case AF_INET:
4789                                 pf_change_a(&pd->dst->v4.s_addr,
4790                                     pd->ip_sum, (*state)->lan.addr.v4.s_addr,
4791                                     0);
4792                                 break;
4793 #endif /* INET */
4794 #ifdef INET6
4795                         case AF_INET6:
4796                                 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
4797                                 break;
4798 #endif /* INET6 */
4799                         }
4800         }
4801
4802         return (PF_PASS);
4803 }
4804
4805 /*
4806  * ipoff and off are measured from the start of the mbuf chain.
4807  * h must be at "ipoff" on the mbuf chain.
4808  */
4809 void *
4810 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4811     u_short *actionp, u_short *reasonp, sa_family_t af)
4812 {
4813         switch (af) {
4814 #ifdef INET
4815         case AF_INET: {
4816                 struct ip       *h = mtod(m, struct ip *);
4817                 u_int16_t        fragoff = (h->ip_off & IP_OFFMASK) << 3;
4818
4819                 if (fragoff) {
4820                         if (fragoff >= len)
4821                                 ACTION_SET(actionp, PF_PASS);
4822                         else {
4823                                 ACTION_SET(actionp, PF_DROP);
4824                                 REASON_SET(reasonp, PFRES_FRAG);
4825                         }
4826                         return (NULL);
4827                 }
4828                 if (m->m_pkthdr.len < off + len ||
4829                     h->ip_len < off + len) {
4830                         ACTION_SET(actionp, PF_DROP);
4831                         REASON_SET(reasonp, PFRES_SHORT);
4832                         return (NULL);
4833                 }
4834                 break;
4835         }
4836 #endif /* INET */
4837 #ifdef INET6
4838         case AF_INET6: {
4839                 struct ip6_hdr  *h = mtod(m, struct ip6_hdr *);
4840
4841                 if (m->m_pkthdr.len < off + len ||
4842                     (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
4843                     (unsigned)(off + len)) {
4844                         ACTION_SET(actionp, PF_DROP);
4845                         REASON_SET(reasonp, PFRES_SHORT);
4846                         return (NULL);
4847                 }
4848                 break;
4849         }
4850 #endif /* INET6 */
4851         }
4852         m_copydata(m, off, len, p);
4853         return (p);
4854 }
4855
4856 int
4857 pf_routable(struct pf_addr *addr, sa_family_t af)
4858 {
4859         struct sockaddr_in      *dst;
4860         struct route             ro;
4861         int                      ret = 0;
4862
4863         bzero(&ro, sizeof(ro));
4864         dst = satosin(&ro.ro_dst);
4865         dst->sin_family = af;
4866         dst->sin_len = sizeof(*dst);
4867         dst->sin_addr = addr->v4;
4868         rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
4869
4870         if (ro.ro_rt != NULL) {
4871                 ret = 1;
4872                 RTFREE(ro.ro_rt);
4873         }
4874
4875         return (ret);
4876 }
4877
4878 #ifdef INET
4879 void
4880 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
4881     struct pf_state *s)
4882 {
4883         struct mbuf             *m0, *m1;
4884         struct route             iproute;
4885         struct route            *ro = NULL;
4886         struct sockaddr_in      *dst;
4887         struct ip               *ip;
4888         struct ifnet            *ifp = NULL;
4889         struct pf_addr           naddr;
4890         struct pf_src_node      *sn = NULL;
4891         int                      error = 0;
4892         int sw_csum;
4893
4894         if (m == NULL || *m == NULL || r == NULL ||
4895             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
4896                 panic("pf_route: invalid parameters");
4897
4898         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
4899                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
4900                 (*m)->m_pkthdr.pf_routed = 1;
4901         } else {
4902                 if ((*m)->m_pkthdr.pf_routed > 3) {
4903                         m0 = *m;
4904                         *m = NULL;
4905                         goto bad;
4906                 }
4907                 (*m)->m_pkthdr.pf_routed++;
4908         }
4909
4910         if (r->rt == PF_DUPTO) {
4911                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
4912                         return;
4913         } else {
4914                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
4915                         return;
4916                 m0 = *m;
4917         }
4918
4919         if (m0->m_len < sizeof(struct ip))
4920                 panic("pf_route: m0->m_len < sizeof(struct ip)");
4921         ip = mtod(m0, struct ip *);
4922
4923         ro = &iproute;
4924         bzero((caddr_t)ro, sizeof(*ro));
4925         dst = satosin(&ro->ro_dst);
4926         dst->sin_family = AF_INET;
4927         dst->sin_len = sizeof(*dst);
4928         dst->sin_addr = ip->ip_dst;
4929
4930         if (r->rt == PF_FASTROUTE) {
4931                 rtalloc(ro);
4932                 if (ro->ro_rt == 0) {
4933                         ipstat.ips_noroute++;
4934                         goto bad;
4935                 }
4936
4937                 ifp = ro->ro_rt->rt_ifp;
4938                 ro->ro_rt->rt_use++;
4939
4940                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
4941                         dst = satosin(ro->ro_rt->rt_gateway);
4942         } else {
4943                 if (TAILQ_EMPTY(&r->rpool.list))
4944                         panic("pf_route: TAILQ_EMPTY(&r->rpool.list)");
4945                 if (s == NULL) {
4946                         pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
4947                             &naddr, NULL, &sn);
4948                         if (!PF_AZERO(&naddr, AF_INET))
4949                                 dst->sin_addr.s_addr = naddr.v4.s_addr;
4950                         ifp = r->rpool.cur->kif ?
4951                             r->rpool.cur->kif->pfik_ifp : NULL;
4952                 } else {
4953                         if (!PF_AZERO(&s->rt_addr, AF_INET))
4954                                 dst->sin_addr.s_addr =
4955                                     s->rt_addr.v4.s_addr;
4956                         ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
4957                 }
4958         }
4959         if (ifp == NULL)
4960                 goto bad;
4961
4962         if (oifp != ifp) {
4963                 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS)
4964                         goto bad;
4965                 else if (m0 == NULL)
4966                         goto done;
4967                 if (m0->m_len < sizeof(struct ip))
4968                         panic("pf_route: m0->m_len < sizeof(struct ip)");
4969                 ip = mtod(m0, struct ip *);
4970         }
4971
4972         /* Copied from ip_output. */
4973         m0->m_pkthdr.csum_flags |= CSUM_IP;
4974         sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
4975         if (sw_csum & CSUM_DELAY_DATA) {
4976                 in_delayed_cksum(m0);
4977                 sw_csum &= ~CSUM_DELAY_DATA;
4978         }
4979         m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
4980
4981         /*
4982          * If small enough for interface, or the interface will take
4983          * care of the fragmentation for us, can just send directly.
4984          */
4985         if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
4986                 (ip->ip_off & IP_DF) == 0)) {
4987                 ip->ip_len = htons(ip->ip_len);
4988                 ip->ip_off = htons(ip->ip_off);
4989                 ip->ip_sum = 0;
4990                 if (sw_csum & CSUM_DELAY_IP) {
4991                         /* From KAME */
4992                         if (ip->ip_v == IPVERSION &&
4993                             (ip->ip_hl << 2) == sizeof(*ip)) {
4994                                 ip->ip_sum = in_cksum_hdr(ip);
4995                         } else {
4996                                 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
4997                         }
4998                 }
4999
5000                 lwkt_serialize_enter(ifp->if_serializer);
5001                 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt);
5002                 lwkt_serialize_exit(ifp->if_serializer);
5003                 goto done;
5004         }
5005
5006         /*
5007          * Too large for interface; fragment if possible.
5008          * Must be able to put at least 8 bytes per fragment.
5009          */
5010         if (ip->ip_off & IP_DF) {
5011                 ipstat.ips_cantfrag++;
5012                 if (r->rt != PF_DUPTO) {
5013                         icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
5014                             ifp->if_mtu);
5015                         goto done;
5016                 } else
5017                         goto bad;
5018         }
5019
5020         m1 = m0;
5021         error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5022         if (error)
5023                 goto bad;
5024
5025         for (m0 = m1; m0; m0 = m1) {
5026                 m1 = m0->m_nextpkt;
5027                 m0->m_nextpkt = 0;
5028                 if (error == 0) {
5029                         lwkt_serialize_enter(ifp->if_serializer);
5030                         error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5031                                     NULL);
5032                         lwkt_serialize_exit(ifp->if_serializer);
5033                 } else {
5034                         m_freem(m0);
5035                 }
5036         }
5037
5038         if (error == 0)
5039                 ipstat.ips_fragmented++;
5040
5041 done:
5042         if (r->rt != PF_DUPTO)
5043                 *m = NULL;
5044         if (ro == &iproute && ro->ro_rt)
5045                 RTFREE(ro->ro_rt);
5046         return;
5047
5048 bad:
5049         m_freem(m0);
5050         goto done;
5051 }
5052 #endif /* INET */
5053
5054 #ifdef INET6
5055 void
5056 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5057     struct pf_state *s)
5058 {
5059         struct mbuf             *m0;
5060         struct route_in6         ip6route;
5061         struct route_in6        *ro;
5062         struct sockaddr_in6     *dst;
5063         struct ip6_hdr          *ip6;
5064         struct ifnet            *ifp = NULL;
5065         struct pf_addr           naddr;
5066         struct pf_src_node      *sn = NULL;
5067         int                      error = 0;
5068
5069         if (m == NULL || *m == NULL || r == NULL ||
5070             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5071                 panic("pf_route6: invalid parameters");
5072
5073         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5074                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5075                 (*m)->m_pkthdr.pf_routed = 1;
5076         } else {
5077                 if ((*m)->m_pkthdr.pf_routed > 3) {
5078                         m0 = *m;
5079                         *m = NULL;
5080                         goto bad;
5081                 }
5082                 (*m)->m_pkthdr.pf_routed++;
5083         }
5084
5085         if (r->rt == PF_DUPTO) {
5086                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5087                         return;
5088         } else {
5089                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5090                         return;
5091                 m0 = *m;
5092         }
5093
5094         if (m0->m_len < sizeof(struct ip6_hdr))
5095                 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5096         ip6 = mtod(m0, struct ip6_hdr *);
5097
5098         ro = &ip6route;
5099         bzero((caddr_t)ro, sizeof(*ro));
5100         dst = (struct sockaddr_in6 *)&ro->ro_dst;
5101         dst->sin6_family = AF_INET6;
5102         dst->sin6_len = sizeof(*dst);
5103         dst->sin6_addr = ip6->ip6_dst;
5104
5105         /* Cheat. */
5106         if (r->rt == PF_FASTROUTE) {
5107                 m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
5108                 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5109                 return;
5110         }
5111
5112         if (TAILQ_EMPTY(&r->rpool.list))
5113                 panic("pf_route6: TAILQ_EMPTY(&r->rpool.list)");
5114         if (s == NULL) {
5115                 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5116                     &naddr, NULL, &sn);
5117                 if (!PF_AZERO(&naddr, AF_INET6))
5118                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5119                             &naddr, AF_INET6);
5120                 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5121         } else {
5122                 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5123                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5124                             &s->rt_addr, AF_INET6);
5125                 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5126         }
5127         if (ifp == NULL)
5128                 goto bad;
5129
5130         if (oifp != ifp) {
5131                 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS)
5132                         goto bad;
5133                 else if (m0 == NULL)
5134                         goto done;
5135                 if (m0->m_len < sizeof(struct ip6_hdr))
5136                         panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5137                 ip6 = mtod(m0, struct ip6_hdr *);
5138         }
5139
5140         /*
5141          * If the packet is too large for the outgoing interface,
5142          * send back an icmp6 error.
5143          */
5144         if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5145                 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5146         if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5147                 error = nd6_output(ifp, ifp, m0, dst, NULL);
5148         } else {
5149                 in6_ifstat_inc(ifp, ifs6_in_toobig);
5150                 if (r->rt != PF_DUPTO)
5151                         icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5152                 else
5153                         goto bad;
5154         }
5155
5156 done:
5157         if (r->rt != PF_DUPTO)
5158                 *m = NULL;
5159         return;
5160
5161 bad:
5162         m_freem(m0);
5163         goto done;
5164 }
5165 #endif /* INET6 */
5166
5167
5168 /*
5169  * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5170  *   off is the offset where the protocol header starts
5171  *   len is the total length of protocol header plus payload
5172  * returns 0 when the checksum is valid, otherwise returns 1.
5173  */
5174 /*
5175  * XXX
5176  * FreeBSD supports cksum offload for the following drivers.
5177  * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5178  * If we can make full use of it we would outperform ipfw/ipfilter in
5179  * very heavy traffic.
5180  * I have not tested 'cause I don't have NICs that supports cksum offload.
5181  * (There might be problems. Typical phenomena would be
5182  *   1. No route message for UDP packet.
5183  *   2. No connection acceptance from external hosts regardless of rule set.)
5184  */
5185 int
5186 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5187     sa_family_t af)
5188 {
5189         u_int16_t sum = 0;
5190         int hw_assist = 0;
5191         struct ip *ip;
5192
5193         if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5194                 return (1);
5195         if (m->m_pkthdr.len < off + len)
5196                 return (1);
5197
5198         switch (p) {
5199         case IPPROTO_TCP:
5200         case IPPROTO_UDP:
5201                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5202                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5203                                 sum = m->m_pkthdr.csum_data;
5204                         } else {
5205                                 ip = mtod(m, struct ip *);
5206                                 sum = in_pseudo(ip->ip_src.s_addr,
5207                                         ip->ip_dst.s_addr, htonl((u_short)len +
5208                                         m->m_pkthdr.csum_data + p));
5209                         }
5210                         sum ^= 0xffff;
5211                         ++hw_assist;
5212                 }
5213                 break;
5214         case IPPROTO_ICMP:
5215 #ifdef INET6
5216         case IPPROTO_ICMPV6:
5217 #endif /* INET6 */
5218                 break;
5219         default:
5220                 return (1);
5221         }
5222
5223         if (!hw_assist) {
5224                 switch (af) {
5225                 case AF_INET:
5226                         if (p == IPPROTO_ICMP) {
5227                                 if (m->m_len < off)
5228                                         return (1);
5229                                 m->m_data += off;
5230                                 m->m_len -= off;
5231                                 sum = in_cksum(m, len);
5232                                 m->m_data -= off;
5233                                 m->m_len += off;
5234                         } else {
5235                                 if (m->m_len < sizeof(struct ip))
5236                                         return (1);
5237                                 sum = in_cksum_range(m, p, off, len);
5238                                 if (sum == 0) {
5239                                         m->m_pkthdr.csum_flags |=
5240                                             (CSUM_DATA_VALID |
5241                                              CSUM_PSEUDO_HDR);
5242                                         m->m_pkthdr.csum_data = 0xffff;
5243                                 }
5244                         }
5245                         break;
5246 #ifdef INET6
5247                 case AF_INET6:
5248                         if (m->m_len < sizeof(struct ip6_hdr))
5249                                 return (1);
5250                         sum = in6_cksum(m, p, off, len);
5251                         /*
5252                          * XXX
5253                          * IPv6 H/W cksum off-load not supported yet!
5254                          *
5255                          * if (sum == 0) {
5256                          *      m->m_pkthdr.csum_flags |=
5257                          *          (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5258                          *      m->m_pkthdr.csum_data = 0xffff;
5259                          *}
5260                          */
5261                         break;
5262 #endif /* INET6 */
5263                 default:
5264                         return (1);
5265                 }
5266         }
5267         if (sum) {
5268                 switch (p) {
5269                 case IPPROTO_TCP:
5270                         tcpstat.tcps_rcvbadsum++;
5271                         break;
5272                 case IPPROTO_UDP:
5273                         udpstat.udps_badsum++;
5274                         break;
5275                 case IPPROTO_ICMP:
5276                         icmpstat.icps_checksum++;
5277                         break;
5278 #ifdef INET6
5279                 case IPPROTO_ICMPV6:
5280                         icmp6stat.icp6s_checksum++;
5281                         break;
5282 #endif /* INET6 */
5283                 }
5284                 return (1);
5285         }
5286         return (0);
5287 }
5288
5289 #ifdef INET
5290 int
5291 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0)
5292 {
5293         struct pfi_kif          *kif;
5294         u_short                  action, reason = 0, log = 0;
5295         struct mbuf             *m = *m0;
5296         struct ip               *h = NULL;
5297         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5298         struct pf_state         *s = NULL;
5299         struct pf_ruleset       *ruleset = NULL;
5300         struct pf_pdesc          pd;
5301         int                      off, dirndx, pqid = 0;
5302
5303         if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5304                 return (PF_PASS);
5305
5306         kif = pfi_index2kif[ifp->if_index];
5307         if (kif == NULL)
5308                 return (PF_DROP);
5309
5310 #ifdef DIAGNOSTIC
5311         if ((m->m_flags & M_PKTHDR) == 0)
5312                 panic("non-M_PKTHDR is passed to pf_test");
5313 #endif
5314
5315         memset(&pd, 0, sizeof(pd));
5316         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5317                 action = PF_DROP;
5318                 REASON_SET(&reason, PFRES_SHORT);
5319                 log = 1;
5320                 goto done;
5321         }
5322
5323         /* We do IP header normalization and packet reassembly here */
5324         if (pf_normalize_ip(m0, dir, kif, &reason) != PF_PASS) {
5325                 action = PF_DROP;
5326                 goto done;
5327         }
5328         m = *m0;
5329         h = mtod(m, struct ip *);
5330
5331         off = h->ip_hl << 2;
5332         if (off < (int)sizeof(*h)) {
5333                 action = PF_DROP;
5334                 REASON_SET(&reason, PFRES_SHORT);
5335                 log = 1;
5336                 goto done;
5337         }
5338
5339         pd.src = (struct pf_addr *)&h->ip_src;
5340         pd.dst = (struct pf_addr *)&h->ip_dst;
5341         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5342         pd.ip_sum = &h->ip_sum;
5343         pd.proto = h->ip_p;
5344         pd.af = AF_INET;
5345         pd.tos = h->ip_tos;
5346         pd.tot_len = h->ip_len;
5347
5348         /* handle fragments that didn't get reassembled by normalization */
5349         if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5350                 action = pf_test_fragment(&r, dir, kif, m, h,
5351                     &pd, &a, &ruleset);
5352                 goto done;
5353         }
5354
5355         switch (h->ip_p) {
5356
5357         case IPPROTO_TCP: {
5358                 struct tcphdr   th;
5359
5360                 pd.hdr.tcp = &th;
5361                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5362                     &action, &reason, AF_INET)) {
5363                         log = action != PF_PASS;
5364                         goto done;
5365                 }
5366                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5367                     h->ip_len - off, IPPROTO_TCP, AF_INET)) {
5368                         action = PF_DROP;
5369                         goto done;
5370                 }
5371                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5372                 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5373                         pqid = 1;
5374                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5375                 if (action == PF_DROP)
5376                         goto done;
5377                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5378                     &reason);
5379                 if (action == PF_PASS) {
5380 #if NPFSYNC
5381                         pfsync_update_state(s);
5382 #endif
5383                         r = s->rule.ptr;
5384                         a = s->anchor.ptr;
5385                         log = s->log;
5386                 } else if (s == NULL)
5387                         action = pf_test_tcp(&r, &s, dir, kif,
5388                             m, off, h, &pd, &a, &ruleset);
5389                 break;
5390         }
5391
5392         case IPPROTO_UDP: {
5393                 struct udphdr   uh;
5394
5395                 pd.hdr.udp = &uh;
5396                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5397                     &action, &reason, AF_INET)) {
5398                         log = action != PF_PASS;
5399                         goto done;
5400                 }
5401                 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5402                     off, h->ip_len - off, IPPROTO_UDP, AF_INET)) {
5403                         action = PF_DROP;
5404                         goto done;
5405                 }
5406                 if (uh.uh_dport == 0 ||
5407                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5408                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5409                         action = PF_DROP;
5410                         goto done;
5411                 }
5412                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5413                 if (action == PF_PASS) {
5414 #if NPFSYNC
5415                         pfsync_update_state(s);
5416 #endif
5417                         r = s->rule.ptr;
5418                         a = s->anchor.ptr;
5419                         log = s->log;
5420                 } else if (s == NULL)
5421                         action = pf_test_udp(&r, &s, dir, kif,
5422                             m, off, h, &pd, &a, &ruleset);
5423                 break;
5424         }
5425
5426         case IPPROTO_ICMP: {
5427                 struct icmp     ih;
5428
5429                 pd.hdr.icmp = &ih;
5430                 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5431                     &action, &reason, AF_INET)) {
5432                         log = action != PF_PASS;
5433                         goto done;
5434                 }
5435                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5436                     h->ip_len - off, IPPROTO_ICMP, AF_INET)) {
5437                         action = PF_DROP;
5438                         goto done;
5439                 }
5440                 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd);
5441                 if (action == PF_PASS) {
5442 #if NPFSYNC
5443                         pfsync_update_state(s);
5444 #endif
5445                         r = s->rule.ptr;
5446                         a = s->anchor.ptr;
5447                         log = s->log;
5448                 } else if (s == NULL)
5449                         action = pf_test_icmp(&r, &s, dir, kif,
5450                             m, off, h, &pd, &a, &ruleset);
5451                 break;
5452         }
5453
5454         default:
5455                 action = pf_test_state_other(&s, dir, kif, &pd);
5456                 if (action == PF_PASS) {
5457 #if NPFSYNC
5458                         pfsync_update_state(s);
5459 #endif
5460                         r = s->rule.ptr;
5461                         a = s->anchor.ptr;
5462                         log = s->log;
5463                 } else if (s == NULL)
5464                         action = pf_test_other(&r, &s, dir, kif, m, off, h,
5465                             &pd, &a, &ruleset);
5466                 break;
5467         }
5468
5469 done:
5470         if (action == PF_PASS && h->ip_hl > 5 &&
5471             !((s && s->allow_opts) || r->allow_opts)) {
5472                 action = PF_DROP;
5473                 REASON_SET(&reason, PFRES_SHORT);
5474                 log = 1;
5475                 DPFPRINTF(PF_DEBUG_MISC,
5476                     ("pf: dropping packet with ip options\n"));
5477         }
5478
5479 #ifdef ALTQ
5480         if (action == PF_PASS && r->qid) {
5481                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5482                 if (pd.tos == IPTOS_LOWDELAY)
5483                         m->m_pkthdr.altq_qid = r->pqid;
5484                 else
5485                         m->m_pkthdr.altq_qid = r->qid;
5486                 if (s) {
5487                         m->m_pkthdr.fw_flags |= ALTQ_MBUF_STATE_HASHED;
5488                         m->m_pkthdr.altq_state_hash = pf_state_hash(s);
5489                 }
5490                 m->m_pkthdr.ecn_af = AF_INET;
5491                 m->m_pkthdr.header = h;
5492         }
5493 #endif
5494
5495         /*
5496          * connections redirected to loopback should not match sockets
5497          * bound specifically to loopback due to security implications,
5498          * see tcp_input() and in_pcblookup_listen().
5499          */
5500         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5501             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5502             (s->nat_rule.ptr->action == PF_RDR ||
5503             s->nat_rule.ptr->action == PF_BINAT) &&
5504             (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
5505                 action = PF_DROP;
5506                 REASON_SET(&reason, PFRES_MEMORY);
5507         }
5508
5509         m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5510
5511         if (log)
5512                 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, r, a, ruleset);
5513
5514         kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5515         kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5516
5517         if (action == PF_PASS || r->action == PF_DROP) {
5518                 r->packets++;
5519                 r->bytes += pd.tot_len;
5520                 if (a != NULL) {
5521                         a->packets++;
5522                         a->bytes += pd.tot_len;
5523                 }
5524                 if (s != NULL) {
5525                         dirndx = (dir == s->direction) ? 0 : 1;
5526                         s->packets[dirndx]++;
5527                         s->bytes[dirndx] += pd.tot_len;
5528                         if (s->nat_rule.ptr != NULL) {
5529                                 s->nat_rule.ptr->packets++;
5530                                 s->nat_rule.ptr->bytes += pd.tot_len;
5531                         }
5532                         if (s->src_node != NULL) {
5533                                 s->src_node->packets++;
5534                                 s->src_node->bytes += pd.tot_len;
5535                         }
5536                         if (s->nat_src_node != NULL) {
5537                                 s->nat_src_node->packets++;
5538                                 s->nat_src_node->bytes += pd.tot_len;
5539                         }
5540                 }
5541                 tr = r;
5542                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5543                 if (nr != NULL) {
5544                         struct pf_addr *x;
5545                         /*
5546                          * XXX: we need to make sure that the addresses
5547                          * passed to pfr_update_stats() are the same than
5548                          * the addresses used during matching (pfr_match)
5549                          */
5550                         if (r == &pf_default_rule) {
5551                                 tr = nr;
5552                                 x = (s == NULL || s->direction == dir) ?
5553                                     &pd.baddr : &pd.naddr;
5554                         } else
5555                                 x = (s == NULL || s->direction == dir) ?
5556                                     &pd.naddr : &pd.baddr;
5557                         if (x == &pd.baddr || s == NULL) {
5558                                 /* we need to change the address */
5559                                 if (dir == PF_OUT)
5560                                         pd.src = x;
5561                                 else
5562                                         pd.dst = x;
5563                         }
5564                 }
5565                 if (tr->src.addr.type == PF_ADDR_TABLE)
5566                         pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5567                             s->direction == dir) ? pd.src : pd.dst, pd.af,
5568                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5569                             tr->src.not);
5570                 if (tr->dst.addr.type == PF_ADDR_TABLE)
5571                         pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5572                             s->direction == dir) ? pd.dst : pd.src, pd.af,
5573                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5574                             tr->dst.not);
5575         }
5576
5577
5578         if (action == PF_SYNPROXY_DROP) {
5579                 m_freem(*m0);
5580                 *m0 = NULL;
5581                 action = PF_PASS;
5582         } else if (r->rt)
5583                 /* pf_route can free the mbuf causing *m0 to become NULL */
5584                 pf_route(m0, r, dir, ifp, s);
5585
5586         return (action);
5587 }
5588 #endif /* INET */
5589
5590 #ifdef INET6
5591 int
5592 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0)
5593 {
5594         struct pfi_kif          *kif;
5595         u_short                  action, reason = 0, log = 0;
5596         struct mbuf             *m = *m0;
5597         struct ip6_hdr          *h = NULL;
5598         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5599         struct pf_state         *s = NULL;
5600         struct pf_ruleset       *ruleset = NULL;
5601         struct pf_pdesc          pd;
5602         int                      off, terminal = 0, dirndx;
5603
5604         if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5605                 return (PF_PASS);
5606
5607         kif = pfi_index2kif[ifp->if_index];
5608         if (kif == NULL)
5609                 return (PF_DROP);
5610
5611 #ifdef DIAGNOSTIC
5612         if ((m->m_flags & M_PKTHDR) == 0)
5613                 panic("non-M_PKTHDR is passed to pf_test");
5614 #endif
5615
5616         memset(&pd, 0, sizeof(pd));
5617         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5618                 action = PF_DROP;
5619                 REASON_SET(&reason, PFRES_SHORT);
5620                 log = 1;
5621                 goto done;
5622         }
5623
5624         /* We do IP header normalization and packet reassembly here */
5625         if (pf_normalize_ip6(m0, dir, kif, &reason) != PF_PASS) {
5626                 action = PF_DROP;
5627                 goto done;
5628         }
5629         m = *m0;
5630         h = mtod(m, struct ip6_hdr *);
5631
5632         pd.src = (struct pf_addr *)&h->ip6_src;
5633         pd.dst = (struct pf_addr *)&h->ip6_dst;
5634         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
5635         pd.ip_sum = NULL;
5636         pd.af = AF_INET6;
5637         pd.tos = 0;
5638         pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5639
5640         off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5641         pd.proto = h->ip6_nxt;
5642         do {
5643                 switch (pd.proto) {
5644                 case IPPROTO_FRAGMENT:
5645                         action = pf_test_fragment(&r, dir, kif, m, h,
5646                             &pd, &a, &ruleset);
5647                         if (action == PF_DROP)
5648                                 REASON_SET(&reason, PFRES_FRAG);
5649                         goto done;
5650                 case IPPROTO_AH:
5651                 case IPPROTO_HOPOPTS:
5652                 case IPPROTO_ROUTING:
5653                 case IPPROTO_DSTOPTS: {
5654                         /* get next header and header length */
5655                         struct ip6_ext  opt6;
5656
5657                         if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
5658                             NULL, NULL, pd.af)) {
5659                                 DPFPRINTF(PF_DEBUG_MISC,
5660                                     ("pf: IPv6 short opt\n"));
5661                                 action = PF_DROP;
5662                                 REASON_SET(&reason, PFRES_SHORT);
5663                                 log = 1;
5664                                 goto done;
5665                         }
5666                         if (pd.proto == IPPROTO_AH)
5667                                 off += (opt6.ip6e_len + 2) * 4;
5668                         else
5669                                 off += (opt6.ip6e_len + 1) * 8;
5670                         pd.proto = opt6.ip6e_nxt;
5671                         /* goto the next header */
5672                         break;
5673                 }
5674                 default:
5675                         terminal++;
5676                         break;
5677                 }
5678         } while (!terminal);
5679
5680         switch (pd.proto) {
5681
5682         case IPPROTO_TCP: {
5683                 struct tcphdr   th;
5684
5685                 pd.hdr.tcp = &th;
5686                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5687                     &action, &reason, AF_INET6)) {
5688                         log = action != PF_PASS;
5689                         goto done;
5690                 }
5691                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5692                     ntohs(h->ip6_plen), IPPROTO_TCP, AF_INET6)) {
5693                         action = PF_DROP;
5694                         goto done;
5695                 }
5696                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5697                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5698                 if (action == PF_DROP)
5699                         goto done;
5700                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5701                     &reason);
5702                 if (action == PF_PASS) {
5703 #if NPFSYNC
5704                         pfsync_update_state(s);
5705 #endif
5706                         r = s->rule.ptr;
5707                         a = s->anchor.ptr;
5708                         log = s->log;
5709                 } else if (s == NULL)
5710                         action = pf_test_tcp(&r, &s, dir, kif,
5711                             m, off, h, &pd, &a, &ruleset);
5712                 break;
5713         }
5714
5715         case IPPROTO_UDP: {
5716                 struct udphdr   uh;
5717
5718                 pd.hdr.udp = &uh;
5719                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5720                     &action, &reason, AF_INET6)) {
5721                         log = action != PF_PASS;
5722                         goto done;
5723                 }
5724                 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5725                     off, ntohs(h->ip6_plen), IPPROTO_UDP, AF_INET6)) {
5726                         action = PF_DROP;
5727                         goto done;
5728                 }
5729                 if (uh.uh_dport == 0 ||
5730                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5731                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5732                         action = PF_DROP;
5733                         goto done;
5734                 }
5735                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5736                 if (action == PF_PASS) {
5737 #if NPFSYNC
5738                         pfsync_update_state(s);
5739 #endif
5740                         r = s->rule.ptr;
5741                         a = s->anchor.ptr;
5742                         log = s->log;
5743                 } else if (s == NULL)
5744                         action = pf_test_udp(&r, &s, dir, kif,
5745                             m, off, h, &pd, &a, &ruleset);
5746                 break;
5747         }
5748
5749         case IPPROTO_ICMPV6: {
5750                 struct icmp6_hdr        ih;
5751
5752                 pd.hdr.icmp6 = &ih;
5753                 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
5754                     &action, &reason, AF_INET6)) {
5755                         log = action != PF_PASS;
5756                         goto done;
5757                 }
5758                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5759                     ntohs(h->ip6_plen), IPPROTO_ICMPV6, AF_INET6)) {
5760                         action = PF_DROP;
5761                         goto done;
5762                 }
5763                 action = pf_test_state_icmp(&s, dir, kif,
5764                     m, off, h, &pd);
5765                 if (action == PF_PASS) {
5766 #if NPFSYNC
5767                         pfsync_update_state(s);
5768 #endif
5769                         r = s->rule.ptr;
5770                         a = s->anchor.ptr;
5771                         log = s->log;
5772                 } else if (s == NULL)
5773                         action = pf_test_icmp(&r, &s, dir, kif,
5774                             m, off, h, &pd, &a, &ruleset);
5775                 break;
5776         }
5777
5778         default:
5779                 action = pf_test_state_other(&s, dir, kif, &pd);
5780                 if (action == PF_PASS) {
5781                         r = s->rule.ptr;
5782                         a = s->anchor.ptr;
5783                         log = s->log;
5784                 } else if (s == NULL)
5785                         action = pf_test_other(&r, &s, dir, kif, m, off, h,
5786                             &pd, &a, &ruleset);
5787                 break;
5788         }
5789
5790 done:
5791         /* XXX handle IPv6 options, if not allowed. not implemented. */
5792
5793 #ifdef ALTQ
5794         if (action == PF_PASS && r->qid) {
5795                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5796                 if (pd.tos == IPTOS_LOWDELAY)
5797                         m->m_pkthdr.altq_qid = r->pqid;
5798                 else
5799                         m->m_pkthdr.altq_qid = r->qid;
5800                 if (s) {
5801                         m->m_pkthdr.fw_flags |= ALTQ_MBUF_STATE_HASHED;
5802                         m->m_pkthdr.altq_state_hash = pf_state_hash(s);
5803                 }
5804                 m->m_pkthdr.ecn_af = AF_INET6;
5805                 m->m_pkthdr.header = h;
5806         }
5807 #endif
5808
5809         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5810             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5811             (s->nat_rule.ptr->action == PF_RDR ||
5812             s->nat_rule.ptr->action == PF_BINAT) &&
5813             IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) {
5814                 action = PF_DROP;
5815                 REASON_SET(&reason, PFRES_MEMORY);
5816         }
5817
5818         m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5819
5820         if (log)
5821                 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, r, a, ruleset);
5822
5823         kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5824         kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
5825
5826         if (action == PF_PASS || r->action == PF_DROP) {
5827                 r->packets++;
5828                 r->bytes += pd.tot_len;
5829                 if (a != NULL) {
5830                         a->packets++;
5831                         a->bytes += pd.tot_len;
5832                 }
5833                 if (s != NULL) {
5834                         dirndx = (dir == s->direction) ? 0 : 1;
5835                         s->packets[dirndx]++;
5836                         s->bytes[dirndx] += pd.tot_len;
5837                         if (s->nat_rule.ptr != NULL) {
5838                                 s->nat_rule.ptr->packets++;
5839                                 s->nat_rule.ptr->bytes += pd.tot_len;
5840                         }
5841                         if (s->src_node != NULL) {
5842                                 s->src_node->packets++;
5843                                 s->src_node->bytes += pd.tot_len;
5844                         }
5845                         if (s->nat_src_node != NULL) {
5846                                 s->nat_src_node->packets++;
5847                                 s->nat_src_node->bytes += pd.tot_len;
5848                         }
5849                 }
5850                 tr = r;
5851                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5852                 if (nr != NULL) {
5853                         struct pf_addr *x;
5854                         /*
5855                          * XXX: we need to make sure that the addresses
5856                          * passed to pfr_update_stats() are the same than
5857                          * the addresses used during matching (pfr_match)
5858                          */
5859                         if (r == &pf_default_rule) {
5860                                 tr = nr;
5861                                 x = (s == NULL || s->direction == dir) ?
5862                                     &pd.baddr : &pd.naddr;
5863                         } else {
5864                                 x = (s == NULL || s->direction == dir) ?
5865                                     &pd.naddr : &pd.baddr;
5866                         }
5867                         if (x == &pd.baddr || s == NULL) {
5868                                 if (dir == PF_OUT)
5869                                         pd.src = x;
5870                                 else
5871                                         pd.dst = x;
5872                         }
5873                 }
5874                 if (tr->src.addr.type == PF_ADDR_TABLE)
5875                         pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5876                             s->direction == dir) ? pd.src : pd.dst, pd.af,
5877                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5878                             tr->src.not);
5879                 if (tr->dst.addr.type == PF_ADDR_TABLE)
5880                         pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5881                             s->direction == dir) ? pd.dst : pd.src, pd.af,
5882                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5883                             tr->dst.not);
5884         }
5885
5886
5887         if (action == PF_SYNPROXY_DROP) {
5888                 m_freem(*m0);
5889                 *m0 = NULL;
5890                 action = PF_PASS;
5891         } else if (r->rt)
5892                 /* pf_route6 can free the mbuf causing *m0 to become NULL */
5893                 pf_route6(m0, r, dir, ifp, s);
5894
5895         return (action);
5896 }
5897 #endif /* INET6 */