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 1.14 2007/05/23 08:57:10 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 int
 326 pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b)
 327 {
 328         int     diff;
 329
 330         if ((diff = a->proto - b->proto) != 0)
 331                 return (diff);
 332         if ((diff = a->af - b->af) != 0)
 333                 return (diff);
 334         switch (a->af) {
 335 #ifdef INET
 336         case AF_INET:
 337                 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
 338                         return (1);
 339                 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
 340                         return (-1);
 341                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 342                         return (1);
 343                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 344                         return (-1);
 345                 break;
 346 #endif /* INET */
 347 #ifdef INET6
 348         case AF_INET6:
 349                 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
 350                         return (1);
 351                 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
 352                         return (-1);
 353                 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
 354                         return (1);
 355                 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
 356                         return (-1);
 357                 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
 358                         return (1);
 359                 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
 360                         return (-1);
 361                 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
 362                         return (1);
 363                 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
 364                         return (-1);
 365                 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
 366                         return (1);
 367                 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
 368                         return (-1);
 369                 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
 370                         return (1);
 371                 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
 372                         return (-1);
 373                 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
 374                         return (1);
 375                 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
 376                         return (-1);
 377                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 378                         return (1);
 379                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 380                         return (-1);
 381                 break;
 382 #endif /* INET6 */
 383         }
 384
 385         if ((diff = a->lan.port - b->lan.port) != 0)
 386                 return (diff);
 387         if ((diff = a->ext.port - b->ext.port) != 0)
 388                 return (diff);
 389
 390         return (0);
 391 }
 392
 393 static int
 394 pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b)
 395 {
 396         int     diff;
 397
 398         if ((diff = a->proto - b->proto) != 0)
 399                 return (diff);
 400         if ((diff = a->af - b->af) != 0)
 401                 return (diff);
 402         switch (a->af) {
 403 #ifdef INET
 404         case AF_INET:
 405                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 406                         return (1);
 407                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 408                         return (-1);
 409                 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
 410                         return (1);
 411                 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
 412                         return (-1);
 413                 break;
 414 #endif /* INET */
 415 #ifdef INET6
 416         case AF_INET6:
 417                 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
 418                         return (1);
 419                 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
 420                         return (-1);
 421                 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
 422                         return (1);
 423                 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
 424                         return (-1);
 425                 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
 426                         return (1);
 427                 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
 428                         return (-1);
 429                 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
 430                         return (1);
 431                 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
 432                         return (-1);
 433                 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
 434                         return (1);
 435                 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
 436                         return (-1);
 437                 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
 438                         return (1);
 439                 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
 440                         return (-1);
 441                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 442                         return (1);
 443                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 444                         return (-1);
 445                 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
 446                         return (1);
 447                 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
 448                         return (-1);
 449                 break;
 450 #endif /* INET6 */
 451         }
 452
 453         if ((diff = a->ext.port - b->ext.port) != 0)
 454                 return (diff);
 455         if ((diff = a->gwy.port - b->gwy.port) != 0)
 456                 return (diff);
 457
 458         return (0);
 459 }
 460
 461 static int
 462 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
 463 {
 464         if (a->id > b->id)
 465                 return (1);
 466         if (a->id < b->id)
 467                 return (-1);
 468         if (a->creatorid > b->creatorid)
 469                 return (1);
 470         if (a->creatorid < b->creatorid)
 471                 return (-1);
 472
 473         return (0);
 474 }
 475
 476 #ifdef INET6
 477 void
 478 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
 479 {
 480         switch (af) {
 481 #ifdef INET
 482         case AF_INET:
 483                 dst->addr32[0] = src->addr32[0];
 484                 break;
 485 #endif /* INET */
 486         case AF_INET6:
 487                 dst->addr32[0] = src->addr32[0];
 488                 dst->addr32[1] = src->addr32[1];
 489                 dst->addr32[2] = src->addr32[2];
 490                 dst->addr32[3] = src->addr32[3];
 491                 break;
 492         }
 493 }
 494 #endif
 495
 496 struct pf_state *
 497 pf_find_state_byid(struct pf_state *key)
 498 {
 499         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 500         return (RB_FIND(pf_state_tree_id, &tree_id, key));
 501 }
 502
 503 struct pf_state *
 504 pf_find_state_recurse(struct pfi_kif *kif, struct pf_state *key, u_int8_t tree)
 505 {
 506         struct pf_state *s;
 507
 508         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 509
 510         switch (tree) {
 511         case PF_LAN_EXT:
 512                 for (; kif != NULL; kif = kif->pfik_parent) {
 513                         s = RB_FIND(pf_state_tree_lan_ext,
 514                             &kif->pfik_lan_ext, key);
 515                         if (s != NULL)
 516                                 return (s);
 517                 }
 518                 return (NULL);
 519         case PF_EXT_GWY:
 520                 for (; kif != NULL; kif = kif->pfik_parent) {
 521                         s = RB_FIND(pf_state_tree_ext_gwy,
 522                             &kif->pfik_ext_gwy, key);
 523                         if (s != NULL)
 524                                 return (s);
 525                 }
 526                 return (NULL);
 527         default:
 528                 panic("pf_find_state_recurse");
 529         }
 530 }
 531
 532 struct pf_state *
 533 pf_find_state_all(struct pf_state *key, u_int8_t tree, int *more)
 534 {
 535         struct pf_state *s, *ss = NULL;
 536         struct pfi_kif  *kif;
 537
 538         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 539
 540         switch (tree) {
 541         case PF_LAN_EXT:
 542                 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
 543                         s = RB_FIND(pf_state_tree_lan_ext,
 544                             &kif->pfik_lan_ext, key);
 545                         if (s == NULL)
 546                                 continue;
 547                         if (more == NULL)
 548                                 return (s);
 549                         ss = s;
 550                         (*more)++;
 551                 }
 552                 return (ss);
 553         case PF_EXT_GWY:
 554                 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
 555                         s = RB_FIND(pf_state_tree_ext_gwy,
 556                             &kif->pfik_ext_gwy, key);
 557                         if (s == NULL)
 558                                 continue;
 559                         if (more == NULL)
 560                                 return (s);
 561                         ss = s;
 562                         (*more)++;
 563                 }
 564                 return (ss);
 565         default:
 566                 panic("pf_find_state_all");
 567         }
 568 }
 569
 570 int
 571 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
 572     struct pf_addr *src, sa_family_t af)
 573 {
 574         struct pf_src_node      k;
 575
 576         if (*sn == NULL) {
 577                 k.af = af;
 578                 PF_ACPY(&k.addr, src, af);
 579                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 580                     rule->rpool.opts & PF_POOL_STICKYADDR)
 581                         k.rule.ptr = rule;
 582                 else
 583                         k.rule.ptr = NULL;
 584                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
 585                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
 586         }
 587         if (*sn == NULL) {
 588                 if (!rule->max_src_nodes ||
 589                     rule->src_nodes < rule->max_src_nodes)
 590                         (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
 591                 if ((*sn) == NULL)
 592                         return (-1);
 593                 bzero(*sn, sizeof(struct pf_src_node));
 594                 (*sn)->af = af;
 595                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 596                     rule->rpool.opts & PF_POOL_STICKYADDR)
 597                         (*sn)->rule.ptr = rule;
 598                 else
 599                         (*sn)->rule.ptr = NULL;
 600                 PF_ACPY(&(*sn)->addr, src, af);
 601                 if (RB_INSERT(pf_src_tree,
 602                     &tree_src_tracking, *sn) != NULL) {
 603                         if (pf_status.debug >= PF_DEBUG_MISC) {
 604                                 kprintf("pf: src_tree insert failed: ");
 605                                 pf_print_host(&(*sn)->addr, 0, af);
 606                                 kprintf("\n");
 607                         }
 608                         pool_put(&pf_src_tree_pl, *sn);
 609                         return (-1);
 610                 }
 611                 (*sn)->creation = time_second;
 612                 (*sn)->ruletype = rule->action;
 613                 if ((*sn)->rule.ptr != NULL)
 614                         (*sn)->rule.ptr->src_nodes++;
 615                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
 616                 pf_status.src_nodes++;
 617         } else {
 618                 if (rule->max_src_states &&
 619                     (*sn)->states >= rule->max_src_states)
 620                         return (-1);
 621         }
 622         return (0);
 623 }
 624
 625 int
 626 pf_insert_state(struct pfi_kif *kif, struct pf_state *state)
 627 {
 628         /* Thou MUST NOT insert multiple duplicate keys */
 629         state->u.s.kif = kif;
 630         if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) {
 631                 if (pf_status.debug >= PF_DEBUG_MISC) {
 632                         kprintf("pf: state insert failed: tree_lan_ext");
 633                         kprintf(" lan: ");
 634                         pf_print_host(&state->lan.addr, state->lan.port,
 635                             state->af);
 636                         kprintf(" gwy: ");
 637                         pf_print_host(&state->gwy.addr, state->gwy.port,
 638                             state->af);
 639                         kprintf(" ext: ");
 640                         pf_print_host(&state->ext.addr, state->ext.port,
 641                             state->af);
 642                         if (state->sync_flags & PFSTATE_FROMSYNC)
 643                                 kprintf(" (from sync)");
 644                         kprintf("\n");
 645                 }
 646                 return (-1);
 647         }
 648
 649         if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) {
 650                 if (pf_status.debug >= PF_DEBUG_MISC) {
 651                         kprintf("pf: state insert failed: tree_ext_gwy");
 652                         kprintf(" lan: ");
 653                         pf_print_host(&state->lan.addr, state->lan.port,
 654                             state->af);
 655                         kprintf(" gwy: ");
 656                         pf_print_host(&state->gwy.addr, state->gwy.port,
 657                             state->af);
 658                         kprintf(" ext: ");
 659                         pf_print_host(&state->ext.addr, state->ext.port,
 660                             state->af);
 661                         if (state->sync_flags & PFSTATE_FROMSYNC)
 662                                 kprintf(" (from sync)");
 663                         kprintf("\n");
 664                 }
 665                 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
 666                 return (-1);
 667         }
 668
 669         if (state->id == 0 && state->creatorid == 0) {
 670                 state->id = htobe64(pf_status.stateid++);
 671                 state->creatorid = pf_status.hostid;
 672         }
 673         if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) {
 674                 if (pf_status.debug >= PF_DEBUG_MISC) {
 675                         kprintf("pf: state insert failed: "
 676                             "id: %016" PRIx64 " creatorid: %08" PRIx32,
 677                             be64toh(state->id), ntohl(state->creatorid));
 678                         if (state->sync_flags & PFSTATE_FROMSYNC)
 679                                 kprintf(" (from sync)");
 680                         kprintf("\n");
 681                 }
 682                 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
 683                 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state);
 684                 return (-1);
 685         }
 686         TAILQ_INSERT_HEAD(&state_updates, state, u.s.entry_updates);
 687
 688         pf_status.fcounters[FCNT_STATE_INSERT]++;
 689         pf_status.states++;
 690         pfi_attach_state(kif);
 691 #if NPFSYNC
 692         pfsync_insert_state(state);
 693 #endif
 694         return (0);
 695 }
 696
 697 void
 698 pf_purge_timeout(void *arg)
 699 {
 700         struct callout  *to = arg;
 701
 702         crit_enter();
 703         pf_purge_expired_states();
 704         pf_purge_expired_fragments();
 705         pf_purge_expired_src_nodes();
 706         crit_exit();
 707
 708         callout_reset(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz,
 709             pf_purge_timeout, to);
 710 }
 711
 712 u_int32_t
 713 pf_state_expires(const struct pf_state *state)
 714 {
 715         u_int32_t       timeout;
 716         u_int32_t       start;
 717         u_int32_t       end;
 718         u_int32_t       states;
 719
 720         /* handle all PFTM_* > PFTM_MAX here */
 721         if (state->timeout == PFTM_PURGE)
 722                 return (time_second);
 723         if (state->timeout == PFTM_UNTIL_PACKET)
 724                 return (0);
 725         KASSERT((state->timeout < PFTM_MAX),
 726             ("pf_state_expires: timeout > PFTM_MAX"));
 727         timeout = state->rule.ptr->timeout[state->timeout];
 728         if (!timeout)
 729                 timeout = pf_default_rule.timeout[state->timeout];
 730         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
 731         if (start) {
 732                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
 733                 states = state->rule.ptr->states;
 734         } else {
 735                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
 736                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
 737                 states = pf_status.states;
 738         }
 739         if (end && states > start && start < end) {
 740                 if (states < end)
 741                         return (state->expire + timeout * (end - states) /
 742                             (end - start));
 743                 else
 744                         return (time_second);
 745         }
 746         return (state->expire + timeout);
 747 }
 748
 749 void
 750 pf_purge_expired_src_nodes(void)
 751 {
 752          struct pf_src_node             *cur, *next;
 753
 754          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
 755                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
 756
 757                  if (cur->states <= 0 && cur->expire <= time_second) {
 758                          if (cur->rule.ptr != NULL) {
 759                                  cur->rule.ptr->src_nodes--;
 760                                  if (cur->rule.ptr->states <= 0 &&
 761                                      cur->rule.ptr->max_src_nodes <= 0)
 762                                          pf_rm_rule(NULL, cur->rule.ptr);
 763                          }
 764                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
 765                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
 766                          pf_status.src_nodes--;
 767                          pool_put(&pf_src_tree_pl, cur);
 768                  }
 769          }
 770 }
 771
 772 void
 773 pf_src_tree_remove_state(struct pf_state *s)
 774 {
 775         u_int32_t timeout;
 776
 777         if (s->src_node != NULL) {
 778                 if (--s->src_node->states <= 0) {
 779                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
 780                         if (!timeout)
 781                                 timeout =
 782                                     pf_default_rule.timeout[PFTM_SRC_NODE];
 783                         s->src_node->expire = time_second + timeout;
 784                 }
 785         }
 786         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
 787                 if (--s->nat_src_node->states <= 0) {
 788                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
 789                         if (!timeout)
 790                                 timeout =
 791                                     pf_default_rule.timeout[PFTM_SRC_NODE];
 792                         s->nat_src_node->expire = time_second + timeout;
 793                 }
 794         }
 795         s->src_node = s->nat_src_node = NULL;
 796 }
 797
 798 static int
 799 pf_purge_expired_states_callback(struct pf_state *cur, void *data __unused)
 800 {
 801         if (pf_state_expires(cur) <= time_second) {
 802                 RB_REMOVE(pf_state_tree_ext_gwy,
 803                     &cur->u.s.kif->pfik_ext_gwy, cur);
 804                 RB_REMOVE(pf_state_tree_lan_ext,
 805                     &cur->u.s.kif->pfik_lan_ext, cur);
 806                 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
 807                 if (cur->src.state == PF_TCPS_PROXY_DST) {
 808                         pf_send_tcp(cur->rule.ptr, cur->af,
 809                             &cur->ext.addr, &cur->lan.addr,
 810                             cur->ext.port, cur->lan.port,
 811                             cur->src.seqhi, cur->src.seqlo + 1, 0,
 812                             TH_RST|TH_ACK, 0, 0);
 813                 }
 814 #if NPFSYNC
 815                 pfsync_delete_state(cur);
 816 #endif
 817                 pf_src_tree_remove_state(cur);
 818                 if (--cur->rule.ptr->states <= 0 &&
 819                     cur->rule.ptr->src_nodes <= 0)
 820                         pf_rm_rule(NULL, cur->rule.ptr);
 821                 if (cur->nat_rule.ptr != NULL)
 822                         if (--cur->nat_rule.ptr->states <= 0 &&
 823                                 cur->nat_rule.ptr->src_nodes <= 0)
 824                                 pf_rm_rule(NULL, cur->nat_rule.ptr);
 825                 if (cur->anchor.ptr != NULL)
 826                         if (--cur->anchor.ptr->states <= 0)
 827                                 pf_rm_rule(NULL, cur->anchor.ptr);
 828                 pf_normalize_tcp_cleanup(cur);
 829                 pfi_detach_state(cur->u.s.kif);
 830                 TAILQ_REMOVE(&state_updates, cur, u.s.entry_updates);
 831                 pool_put(&pf_state_pl, cur);
 832                 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
 833                 pf_status.states--;
 834         }
 835         return(0);
 836 }
 837
 838 void
 839 pf_purge_expired_states(void)
 840 {
 841         RB_SCAN(pf_state_tree_id, &tree_id, NULL,
 842                 pf_purge_expired_states_callback, NULL);
 843 }
 844
 845
 846 int
 847 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
 848 {
 849         if (aw->type != PF_ADDR_TABLE)
 850                 return (0);
 851         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
 852                 return (1);
 853         return (0);
 854 }
 855
 856 void
 857 pf_tbladdr_remove(struct pf_addr_wrap *aw)
 858 {
 859         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
 860                 return;
 861         pfr_detach_table(aw->p.tbl);
 862         aw->p.tbl = NULL;
 863 }
 864
 865 void
 866 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
 867 {
 868         struct pfr_ktable *kt = aw->p.tbl;
 869
 870         if (aw->type != PF_ADDR_TABLE || kt == NULL)
 871                 return;
 872         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
 873                 kt = kt->pfrkt_root;
 874         aw->p.tbl = NULL;
 875         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
 876                 kt->pfrkt_cnt : -1;
 877 }
 878
 879 void
 880 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
 881 {
 882         switch (af) {
 883 #ifdef INET
 884         case AF_INET: {
 885                 u_int32_t a = ntohl(addr->addr32[0]);
 886                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
 887                     (a>>8)&255, a&255);
 888                 if (p) {
 889                         p = ntohs(p);
 890                         kprintf(":%u", p);
 891                 }
 892                 break;
 893         }
 894 #endif /* INET */
 895 #ifdef INET6
 896         case AF_INET6: {
 897                 u_int16_t b;
 898                 u_int8_t i, curstart = 255, curend = 0,
 899                     maxstart = 0, maxend = 0;
 900                 for (i = 0; i < 8; i++) {
 901                         if (!addr->addr16[i]) {
 902                                 if (curstart == 255)
 903                                         curstart = i;
 904                                 else
 905                                         curend = i;
 906                         } else {
 907                                 if (curstart) {
 908                                         if ((curend - curstart) >
 909                                             (maxend - maxstart)) {
 910                                                 maxstart = curstart;
 911                                                 maxend = curend;
 912                                                 curstart = 255;
 913                                         }
 914                                 }
 915                         }
 916                 }
 917                 for (i = 0; i < 8; i++) {
 918                         if (i >= maxstart && i <= maxend) {
 919                                 if (maxend != 7) {
 920                                         if (i == maxstart)
 921                                                 kprintf(":");
 922                                 } else {
 923                                         if (i == maxend)
 924                                                 kprintf(":");
 925                                 }
 926                         } else {
 927                                 b = ntohs(addr->addr16[i]);
 928                                 kprintf("%x", b);
 929                                 if (i < 7)
 930                                         kprintf(":");
 931                         }
 932                 }
 933                 if (p) {
 934                         p = ntohs(p);
 935                         kprintf("[%u]", p);
 936                 }
 937                 break;
 938         }
 939 #endif /* INET6 */
 940         }
 941 }
 942
 943 void
 944 pf_print_state(struct pf_state *s)
 945 {
 946         switch (s->proto) {
 947         case IPPROTO_TCP:
 948                 kprintf("TCP ");
 949                 break;
 950         case IPPROTO_UDP:
 951                 kprintf("UDP ");
 952                 break;
 953         case IPPROTO_ICMP:
 954                 kprintf("ICMP ");
 955                 break;
 956         case IPPROTO_ICMPV6:
 957                 kprintf("ICMPV6 ");
 958                 break;
 959         default:
 960                 kprintf("%u ", s->proto);
 961                 break;
 962         }
 963         pf_print_host(&s->lan.addr, s->lan.port, s->af);
 964         kprintf(" ");
 965         pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
 966         kprintf(" ");
 967         pf_print_host(&s->ext.addr, s->ext.port, s->af);
 968         kprintf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
 969             s->src.seqhi, s->src.max_win, s->src.seqdiff);
 970         if (s->src.wscale && s->dst.wscale)
 971                 kprintf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
 972         kprintf("]");
 973         kprintf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
 974             s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
 975         if (s->src.wscale && s->dst.wscale)
 976                 kprintf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
 977         kprintf("]");
 978         kprintf(" %u:%u", s->src.state, s->dst.state);
 979 }
 980
 981 void
 982 pf_print_flags(u_int8_t f)
 983 {
 984         if (f)
 985                 kprintf(" ");
 986         if (f & TH_FIN)
 987                 kprintf("F");
 988         if (f & TH_SYN)
 989                 kprintf("S");
 990         if (f & TH_RST)
 991                 kprintf("R");
 992         if (f & TH_PUSH)
 993                 kprintf("P");
 994         if (f & TH_ACK)
 995                 kprintf("A");
 996         if (f & TH_URG)
 997                 kprintf("U");
 998         if (f & TH_ECE)
 999                 kprintf("E");
1000         if (f & TH_CWR)
1001                 kprintf("W");
1002 }
1003
1004 #define PF_SET_SKIP_STEPS(i)                                    \
1005         do {                                                    \
1006                 while (head[i] != cur) {                        \
1007                         head[i]->skip[i].ptr = cur;             \
1008                         head[i] = TAILQ_NEXT(head[i], entries); \
1009                 }                                               \
1010         } while (0)
1011
1012 void
1013 pf_calc_skip_steps(struct pf_rulequeue *rules)
1014 {
1015         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1016         int i;
1017
1018         cur = TAILQ_FIRST(rules);
1019         prev = cur;
1020         for (i = 0; i < PF_SKIP_COUNT; ++i)
1021                 head[i] = cur;
1022         while (cur != NULL) {
1023
1024                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1025                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1026                 if (cur->direction != prev->direction)
1027                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1028                 if (cur->af != prev->af)
1029                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1030                 if (cur->proto != prev->proto)
1031                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1032                 if (cur->src.not != prev->src.not ||
1033                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1034                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1035                 if (cur->src.port[0] != prev->src.port[0] ||
1036                     cur->src.port[1] != prev->src.port[1] ||
1037                     cur->src.port_op != prev->src.port_op)
1038                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1039                 if (cur->dst.not != prev->dst.not ||
1040                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1041                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1042                 if (cur->dst.port[0] != prev->dst.port[0] ||
1043                     cur->dst.port[1] != prev->dst.port[1] ||
1044                     cur->dst.port_op != prev->dst.port_op)
1045                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1046
1047                 prev = cur;
1048                 cur = TAILQ_NEXT(cur, entries);
1049         }
1050         for (i = 0; i < PF_SKIP_COUNT; ++i)
1051                 PF_SET_SKIP_STEPS(i);
1052 }
1053
1054 int
1055 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1056 {
1057         if (aw1->type != aw2->type)
1058                 return (1);
1059         switch (aw1->type) {
1060         case PF_ADDR_ADDRMASK:
1061                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1062                         return (1);
1063                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1064                         return (1);
1065                 return (0);
1066         case PF_ADDR_DYNIFTL:
1067                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1068         case PF_ADDR_NOROUTE:
1069                 return (0);
1070         case PF_ADDR_TABLE:
1071                 return (aw1->p.tbl != aw2->p.tbl);
1072         default:
1073                 kprintf("invalid address type: %d\n", aw1->type);
1074                 return (1);
1075         }
1076 }
1077
1078 void
1079 pf_update_anchor_rules(void)
1080 {
1081         struct pf_rule  *rule;
1082         int              i;
1083
1084         for (i = 0; i < PF_RULESET_MAX; ++i)
1085                 TAILQ_FOREACH(rule, pf_main_ruleset.rules[i].active.ptr,
1086                     entries)
1087                         if (rule->anchorname[0])
1088                                 rule->anchor = pf_find_anchor(rule->anchorname);
1089                         else
1090                                 rule->anchor = NULL;
1091 }
1092
1093 u_int16_t
1094 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1095 {
1096         u_int32_t       l;
1097
1098         if (udp && !cksum)
1099                 return (0x0000);
1100         l = cksum + old - new;
1101         l = (l >> 16) + (l & 65535);
1102         l = l & 65535;
1103         if (udp && !l)
1104                 return (0xFFFF);
1105         return (l);
1106 }
1107
1108 void
1109 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1110     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1111 {
1112         struct pf_addr  ao;
1113         u_int16_t       po = *p;
1114
1115         PF_ACPY(&ao, a, af);
1116         PF_ACPY(a, an, af);
1117
1118         *p = pn;
1119
1120         switch (af) {
1121 #ifdef INET
1122         case AF_INET:
1123                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1124                     ao.addr16[0], an->addr16[0], 0),
1125                     ao.addr16[1], an->addr16[1], 0);
1126                 *p = pn;
1127                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1128                     ao.addr16[0], an->addr16[0], u),
1129                     ao.addr16[1], an->addr16[1], u),
1130                     po, pn, u);
1131                 break;
1132 #endif /* INET */
1133 #ifdef INET6
1134         case AF_INET6:
1135                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1136                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1137                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1138                     ao.addr16[0], an->addr16[0], u),
1139                     ao.addr16[1], an->addr16[1], u),
1140                     ao.addr16[2], an->addr16[2], u),
1141                     ao.addr16[3], an->addr16[3], u),
1142                     ao.addr16[4], an->addr16[4], u),
1143                     ao.addr16[5], an->addr16[5], u),
1144                     ao.addr16[6], an->addr16[6], u),
1145                     ao.addr16[7], an->addr16[7], u),
1146                     po, pn, u);
1147                 break;
1148 #endif /* INET6 */
1149         }
1150 }
1151
1152
1153 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1154 void
1155 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1156 {
1157         u_int32_t       ao;
1158
1159         memcpy(&ao, a, sizeof(ao));
1160         memcpy(a, &an, sizeof(u_int32_t));
1161         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1162             ao % 65536, an % 65536, u);
1163 }
1164
1165 #ifdef INET6
1166 void
1167 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1168 {
1169         struct pf_addr  ao;
1170
1171         PF_ACPY(&ao, a, AF_INET6);
1172         PF_ACPY(a, an, AF_INET6);
1173
1174         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1175             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1176             pf_cksum_fixup(pf_cksum_fixup(*c,
1177             ao.addr16[0], an->addr16[0], u),
1178             ao.addr16[1], an->addr16[1], u),
1179             ao.addr16[2], an->addr16[2], u),
1180             ao.addr16[3], an->addr16[3], u),
1181             ao.addr16[4], an->addr16[4], u),
1182             ao.addr16[5], an->addr16[5], u),
1183             ao.addr16[6], an->addr16[6], u),
1184             ao.addr16[7], an->addr16[7], u);
1185 }
1186 #endif /* INET6 */
1187
1188 void
1189 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1190     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1191     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1192 {
1193         struct pf_addr  oia, ooa;
1194
1195         PF_ACPY(&oia, ia, af);
1196         PF_ACPY(&ooa, oa, af);
1197
1198         /* Change inner protocol port, fix inner protocol checksum. */
1199         if (ip != NULL) {
1200                 u_int16_t       oip = *ip;
1201                 u_int32_t       opc = 0;
1202
1203                 if (pc != NULL)
1204                         opc = *pc;
1205                 *ip = np;
1206                 if (pc != NULL)
1207                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1208                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1209                 if (pc != NULL)
1210                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1211         }
1212         /* Change inner ip address, fix inner ip and icmp checksums. */
1213         PF_ACPY(ia, na, af);
1214         switch (af) {
1215 #ifdef INET
1216         case AF_INET: {
1217                 u_int32_t        oh2c = *h2c;
1218
1219                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1220                     oia.addr16[0], ia->addr16[0], 0),
1221                     oia.addr16[1], ia->addr16[1], 0);
1222                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1223                     oia.addr16[0], ia->addr16[0], 0),
1224                     oia.addr16[1], ia->addr16[1], 0);
1225                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1226                 break;
1227         }
1228 #endif /* INET */
1229 #ifdef INET6
1230         case AF_INET6:
1231                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1232                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1233                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1234                     oia.addr16[0], ia->addr16[0], u),
1235                     oia.addr16[1], ia->addr16[1], u),
1236                     oia.addr16[2], ia->addr16[2], u),
1237                     oia.addr16[3], ia->addr16[3], u),
1238                     oia.addr16[4], ia->addr16[4], u),
1239                     oia.addr16[5], ia->addr16[5], u),
1240                     oia.addr16[6], ia->addr16[6], u),
1241                     oia.addr16[7], ia->addr16[7], u);
1242                 break;
1243 #endif /* INET6 */
1244         }
1245         /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1246         PF_ACPY(oa, na, af);
1247         switch (af) {
1248 #ifdef INET
1249         case AF_INET:
1250                 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1251                     ooa.addr16[0], oa->addr16[0], 0),
1252                     ooa.addr16[1], oa->addr16[1], 0);
1253                 break;
1254 #endif /* INET */
1255 #ifdef INET6
1256         case AF_INET6:
1257                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1258                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1259                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1260                     ooa.addr16[0], oa->addr16[0], u),
1261                     ooa.addr16[1], oa->addr16[1], u),
1262                     ooa.addr16[2], oa->addr16[2], u),
1263                     ooa.addr16[3], oa->addr16[3], u),
1264                     ooa.addr16[4], oa->addr16[4], u),
1265                     ooa.addr16[5], oa->addr16[5], u),
1266                     ooa.addr16[6], oa->addr16[6], u),
1267                     ooa.addr16[7], oa->addr16[7], u);
1268                 break;
1269 #endif /* INET6 */
1270         }
1271 }
1272
1273 void
1274 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1275     const struct pf_addr *saddr, const struct pf_addr *daddr,
1276     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1277     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl)
1278 {
1279         struct mbuf     *m;
1280         int              len = 0, tlen;
1281 #ifdef INET
1282         struct ip       *h = NULL;
1283 #endif /* INET */
1284 #ifdef INET6
1285         struct ip6_hdr  *h6 = NULL;
1286 #endif /* INET6 */
1287         struct tcphdr   *th = NULL;
1288         char *opt;
1289
1290         /* maximum segment size tcp option */
1291         tlen = sizeof(struct tcphdr);
1292         if (mss)
1293                 tlen += 4;
1294
1295         switch (af) {
1296 #ifdef INET
1297         case AF_INET:
1298                 len = sizeof(struct ip) + tlen;
1299                 break;
1300 #endif /* INET */
1301 #ifdef INET6
1302         case AF_INET6:
1303                 len = sizeof(struct ip6_hdr) + tlen;
1304                 break;
1305 #endif /* INET6 */
1306         }
1307
1308         /* create outgoing mbuf */
1309         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1310         if (m == NULL)
1311                 return;
1312         m->m_pkthdr.fw_flags = PF_MBUF_GENERATED;
1313 #ifdef ALTQ
1314         if (r != NULL && r->qid) {
1315                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
1316                 m->m_pkthdr.altq_qid = r->qid;
1317                 m->m_pkthdr.ecn_af = af;
1318                 m->m_pkthdr.header = mtod(m, struct ip *);
1319         }
1320 #endif
1321         m->m_data += max_linkhdr;
1322         m->m_pkthdr.len = m->m_len = len;
1323         m->m_pkthdr.rcvif = NULL;
1324         bzero(m->m_data, len);
1325         switch (af) {
1326 #ifdef INET
1327         case AF_INET:
1328                 h = mtod(m, struct ip *);
1329
1330                 /* IP header fields included in the TCP checksum */
1331                 h->ip_p = IPPROTO_TCP;
1332                 h->ip_len = tlen;
1333                 h->ip_src.s_addr = saddr->v4.s_addr;
1334                 h->ip_dst.s_addr = daddr->v4.s_addr;
1335
1336                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1337                 break;
1338 #endif /* INET */
1339 #ifdef INET6
1340         case AF_INET6:
1341                 h6 = mtod(m, struct ip6_hdr *);
1342
1343                 /* IP header fields included in the TCP checksum */
1344                 h6->ip6_nxt = IPPROTO_TCP;
1345                 h6->ip6_plen = htons(tlen);
1346                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1347                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1348
1349                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1350                 break;
1351 #endif /* INET6 */
1352         }
1353
1354         /* TCP header */
1355         th->th_sport = sport;
1356         th->th_dport = dport;
1357         th->th_seq = htonl(seq);
1358         th->th_ack = htonl(ack);
1359         th->th_off = tlen >> 2;
1360         th->th_flags = flags;
1361         th->th_win = htons(win);
1362
1363         if (mss) {
1364                 opt = (char *)(th + 1);
1365                 opt[0] = TCPOPT_MAXSEG;
1366                 opt[1] = 4;
1367                 mss = htons(mss);
1368                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1369         }
1370
1371         switch (af) {
1372 #ifdef INET
1373         case AF_INET:
1374                 /* TCP checksum */
1375                 th->th_sum = in_cksum(m, len);
1376
1377                 /* Finish the IP header */
1378                 h->ip_v = 4;
1379                 h->ip_hl = sizeof(*h) >> 2;
1380                 h->ip_tos = IPTOS_LOWDELAY;
1381                 h->ip_len = len;
1382                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1383                 h->ip_ttl = ttl ? ttl : ip_defttl;
1384                 h->ip_sum = 0;
1385                 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1386                     (void *)NULL);
1387                 break;
1388 #endif /* INET */
1389 #ifdef INET6
1390         case AF_INET6:
1391                 /* TCP checksum */
1392                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1393                     sizeof(struct ip6_hdr), tlen);
1394
1395                 h6->ip6_vfc |= IPV6_VERSION;
1396                 h6->ip6_hlim = IPV6_DEFHLIM;
1397
1398                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1399                 break;
1400 #endif /* INET6 */
1401         }
1402 }
1403
1404 void
1405 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1406     struct pf_rule *r)
1407 {
1408         struct mbuf     *m0;
1409
1410         m0 = m_copypacket(m, MB_DONTWAIT);
1411         if (m0 == NULL)
1412                 return;
1413         m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
1414
1415 #ifdef ALTQ
1416         if (r->qid) {
1417                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
1418                 m->m_pkthdr.altq_qid = r->qid;
1419                 m->m_pkthdr.ecn_af = af;
1420                 m->m_pkthdr.header = mtod(m0, struct ip *);
1421         }
1422 #endif
1423
1424         switch (af) {
1425 #ifdef INET
1426         case AF_INET:
1427                 icmp_error(m0, type, code, 0, 0);
1428                 break;
1429 #endif /* INET */
1430 #ifdef INET6
1431         case AF_INET6:
1432                 icmp6_error(m0, type, code, 0);
1433                 break;
1434 #endif /* INET6 */
1435         }
1436 }
1437
1438 /*
1439  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1440  * If n is 0, they match if they are equal. If n is != 0, they match if they
1441  * are different.
1442  */
1443 int
1444 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1445     struct pf_addr *b, sa_family_t af)
1446 {
1447         int     match = 0;
1448
1449         switch (af) {
1450 #ifdef INET
1451         case AF_INET:
1452                 if ((a->addr32[0] & m->addr32[0]) ==
1453                     (b->addr32[0] & m->addr32[0]))
1454                         match++;
1455                 break;
1456 #endif /* INET */
1457 #ifdef INET6
1458         case AF_INET6:
1459                 if (((a->addr32[0] & m->addr32[0]) ==
1460                      (b->addr32[0] & m->addr32[0])) &&
1461                     ((a->addr32[1] & m->addr32[1]) ==
1462                      (b->addr32[1] & m->addr32[1])) &&
1463                     ((a->addr32[2] & m->addr32[2]) ==
1464                      (b->addr32[2] & m->addr32[2])) &&
1465                     ((a->addr32[3] & m->addr32[3]) ==
1466                      (b->addr32[3] & m->addr32[3])))
1467                         match++;
1468                 break;
1469 #endif /* INET6 */
1470         }
1471         if (match) {
1472                 if (n)
1473                         return (0);
1474                 else
1475                         return (1);
1476         } else {
1477                 if (n)
1478                         return (1);
1479                 else
1480                         return (0);
1481         }
1482 }
1483
1484 int
1485 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1486 {
1487         switch (op) {
1488         case PF_OP_IRG:
1489                 return ((p > a1) && (p < a2));
1490         case PF_OP_XRG:
1491                 return ((p < a1) || (p > a2));
1492         case PF_OP_RRG:
1493                 return ((p >= a1) && (p <= a2));
1494         case PF_OP_EQ:
1495                 return (p == a1);
1496         case PF_OP_NE:
1497                 return (p != a1);
1498         case PF_OP_LT:
1499                 return (p < a1);
1500         case PF_OP_LE:
1501                 return (p <= a1);
1502         case PF_OP_GT:
1503                 return (p > a1);
1504         case PF_OP_GE:
1505                 return (p >= a1);
1506         }
1507         return (0); /* never reached */
1508 }
1509
1510 int
1511 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1512 {
1513         a1 = ntohs(a1);
1514         a2 = ntohs(a2);
1515         p = ntohs(p);
1516         return (pf_match(op, a1, a2, p));
1517 }
1518
1519 int
1520 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1521 {
1522         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1523                 return (0);
1524         return (pf_match(op, a1, a2, u));
1525 }
1526
1527 int
1528 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1529 {
1530         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1531                 return (0);
1532         return (pf_match(op, a1, a2, g));
1533 }
1534
1535 static int
1536 pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_rule *nat_rule,
1537              int *tag)
1538 {
1539         if (*tag == -1) {       /* find mbuf tag */
1540                 if (nat_rule != NULL && nat_rule->tag)
1541                         *tag = nat_rule->tag;
1542                 else if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
1543                         *tag = m->m_pkthdr.pf_tag;
1544                 else
1545                         *tag = 0;
1546         }
1547
1548         return ((!r->match_tag_not && r->match_tag == *tag) ||
1549             (r->match_tag_not && r->match_tag != *tag));
1550 }
1551
1552 void
1553 pf_tag_packet(struct mbuf *m, int tag)
1554 {
1555         if (tag <= 0)
1556                 return;
1557
1558         m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1559         m->m_pkthdr.pf_tag = tag;
1560 }
1561
1562 #define PF_STEP_INTO_ANCHOR(r, a, s, n)                                 \
1563         do {                                                            \
1564                 if ((r) == NULL || (r)->anchor == NULL ||               \
1565                     (s) != NULL || (a) != NULL)                         \
1566                         panic("PF_STEP_INTO_ANCHOR");                   \
1567                 (a) = (r);                                              \
1568                 (s) = TAILQ_FIRST(&(r)->anchor->rulesets);              \
1569                 (r) = NULL;                                             \
1570                 while ((s) != NULL && ((r) =                            \
1571                     TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL)     \
1572                         (s) = TAILQ_NEXT((s), entries);                 \
1573                 if ((r) == NULL) {                                      \
1574                         (r) = TAILQ_NEXT((a), entries);                 \
1575                         (a) = NULL;                                     \
1576                 }                                                       \
1577         } while (0)
1578
1579 #define PF_STEP_OUT_OF_ANCHOR(r, a, s, n)                               \
1580         do {                                                            \
1581                 if ((r) != NULL || (a) == NULL || (s) == NULL)          \
1582                         panic("PF_STEP_OUT_OF_ANCHOR");                 \
1583                 (s) = TAILQ_NEXT((s), entries);                         \
1584                 while ((s) != NULL && ((r) =                            \
1585                     TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL)     \
1586                         (s) = TAILQ_NEXT((s), entries);                 \
1587                 if ((r) == NULL) {                                      \
1588                         (r) = TAILQ_NEXT((a), entries);                 \
1589                         (a) = NULL;                                     \
1590                 }                                                       \
1591         } while (0)
1592
1593 #ifdef INET6
1594 void
1595 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
1596     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
1597 {
1598         switch (af) {
1599 #ifdef INET
1600         case AF_INET:
1601                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1602                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1603                 break;
1604 #endif /* INET */
1605         case AF_INET6:
1606                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1607                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1608                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
1609                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
1610                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
1611                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
1612                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
1613                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
1614                 break;
1615         }
1616 }
1617
1618 void
1619 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
1620 {
1621         switch (af) {
1622 #ifdef INET
1623         case AF_INET:
1624                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
1625                 break;
1626 #endif /* INET */
1627         case AF_INET6:
1628                 if (addr->addr32[3] == 0xffffffff) {
1629                         addr->addr32[3] = 0;
1630                         if (addr->addr32[2] == 0xffffffff) {
1631                                 addr->addr32[2] = 0;
1632                                 if (addr->addr32[1] == 0xffffffff) {
1633                                         addr->addr32[1] = 0;
1634                                         addr->addr32[0] =
1635                                             htonl(ntohl(addr->addr32[0]) + 1);
1636                                 } else
1637                                         addr->addr32[1] =
1638                                             htonl(ntohl(addr->addr32[1]) + 1);
1639                         } else
1640                                 addr->addr32[2] =
1641                                     htonl(ntohl(addr->addr32[2]) + 1);
1642                 } else
1643                         addr->addr32[3] =
1644                             htonl(ntohl(addr->addr32[3]) + 1);
1645                 break;
1646         }
1647 }
1648 #endif /* INET6 */
1649
1650 #define mix(a,b,c) \
1651         do {                                    \
1652                 a -= b; a -= c; a ^= (c >> 13); \
1653                 b -= c; b -= a; b ^= (a << 8);  \
1654                 c -= a; c -= b; c ^= (b >> 13); \
1655                 a -= b; a -= c; a ^= (c >> 12); \
1656                 b -= c; b -= a; b ^= (a << 16); \
1657                 c -= a; c -= b; c ^= (b >> 5);  \
1658                 a -= b; a -= c; a ^= (c >> 3);  \
1659                 b -= c; b -= a; b ^= (a << 10); \
1660                 c -= a; c -= b; c ^= (b >> 15); \
1661         } while (0)
1662
1663 /*
1664  * hash function based on bridge_hash in if_bridge.c
1665  */
1666 void
1667 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
1668     struct pf_poolhashkey *key, sa_family_t af)
1669 {
1670         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
1671
1672         switch (af) {
1673 #ifdef INET
1674         case AF_INET:
1675                 a += inaddr->addr32[0];
1676                 b += key->key32[1];
1677                 mix(a, b, c);
1678                 hash->addr32[0] = c + key->key32[2];
1679                 break;
1680 #endif /* INET */
1681 #ifdef INET6
1682         case AF_INET6:
1683                 a += inaddr->addr32[0];
1684                 b += inaddr->addr32[2];
1685                 mix(a, b, c);
1686                 hash->addr32[0] = c;
1687                 a += inaddr->addr32[1];
1688                 b += inaddr->addr32[3];
1689                 c += key->key32[1];
1690                 mix(a, b, c);
1691                 hash->addr32[1] = c;
1692                 a += inaddr->addr32[2];
1693                 b += inaddr->addr32[1];
1694                 c += key->key32[2];
1695                 mix(a, b, c);
1696                 hash->addr32[2] = c;
1697                 a += inaddr->addr32[3];
1698                 b += inaddr->addr32[0];
1699                 c += key->key32[3];
1700                 mix(a, b, c);
1701                 hash->addr32[3] = c;
1702                 break;
1703 #endif /* INET6 */
1704         }
1705 }
1706
1707 int
1708 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
1709     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
1710 {
1711         unsigned char            hash[16];
1712         struct pf_pool          *rpool = &r->rpool;
1713         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
1714         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
1715         struct pf_pooladdr      *acur = rpool->cur;
1716         struct pf_src_node       k;
1717
1718         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
1719             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1720                 k.af = af;
1721                 PF_ACPY(&k.addr, saddr, af);
1722                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
1723                     r->rpool.opts & PF_POOL_STICKYADDR)
1724                         k.rule.ptr = r;
1725                 else
1726                         k.rule.ptr = NULL;
1727                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
1728                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
1729                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
1730                         PF_ACPY(naddr, &(*sn)->raddr, af);
1731                         if (pf_status.debug >= PF_DEBUG_MISC) {
1732                                 kprintf("pf_map_addr: src tracking maps ");
1733                                 pf_print_host(&k.addr, 0, af);
1734                                 kprintf(" to ");
1735                                 pf_print_host(naddr, 0, af);
1736                                 kprintf("\n");
1737                         }
1738                         return (0);
1739                 }
1740         }
1741
1742         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
1743                 return (1);
1744         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1745                 if (af == AF_INET) {
1746                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
1747                             (rpool->opts & PF_POOL_TYPEMASK) !=
1748                             PF_POOL_ROUNDROBIN)
1749                                 return (1);
1750                          raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
1751                          rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
1752                 } else {
1753                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
1754                             (rpool->opts & PF_POOL_TYPEMASK) !=
1755                             PF_POOL_ROUNDROBIN)
1756                                 return (1);
1757                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
1758                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
1759                 }
1760         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1761                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
1762                         return (1); /* unsupported */
1763         } else {
1764                 raddr = &rpool->cur->addr.v.a.addr;
1765                 rmask = &rpool->cur->addr.v.a.mask;
1766         }
1767
1768         switch (rpool->opts & PF_POOL_TYPEMASK) {
1769         case PF_POOL_NONE:
1770                 PF_ACPY(naddr, raddr, af);
1771                 break;
1772         case PF_POOL_BITMASK:
1773                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
1774                 break;
1775         case PF_POOL_RANDOM:
1776                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
1777                         switch (af) {
1778 #ifdef INET
1779                         case AF_INET:
1780                                 rpool->counter.addr32[0] = karc4random();
1781                                 break;
1782 #endif /* INET */
1783 #ifdef INET6
1784                         case AF_INET6:
1785                                 if (rmask->addr32[3] != 0xffffffff)
1786                                         rpool->counter.addr32[3] = karc4random();
1787                                 else
1788                                         break;
1789                                 if (rmask->addr32[2] != 0xffffffff)
1790                                         rpool->counter.addr32[2] = karc4random();
1791                                 else
1792                                         break;
1793                                 if (rmask->addr32[1] != 0xffffffff)
1794                                         rpool->counter.addr32[1] = karc4random();
1795                                 else
1796                                         break;
1797                                 if (rmask->addr32[0] != 0xffffffff)
1798                                         rpool->counter.addr32[0] = karc4random();
1799                                 break;
1800 #endif /* INET6 */
1801                         }
1802                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1803                         PF_ACPY(init_addr, naddr, af);
1804
1805                 } else {
1806                         PF_AINC(&rpool->counter, af);
1807                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1808                 }
1809                 break;
1810         case PF_POOL_SRCHASH:
1811                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
1812                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
1813                 break;
1814         case PF_POOL_ROUNDROBIN:
1815                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1816                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
1817                             &rpool->tblidx, &rpool->counter,
1818                             &raddr, &rmask, af))
1819                                 goto get_addr;
1820                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1821                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
1822                             &rpool->tblidx, &rpool->counter,
1823                             &raddr, &rmask, af))
1824                                 goto get_addr;
1825                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
1826                         goto get_addr;
1827
1828         try_next:
1829                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
1830                         rpool->cur = TAILQ_FIRST(&rpool->list);
1831                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1832                         rpool->tblidx = -1;
1833                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
1834                             &rpool->tblidx, &rpool->counter,
1835                             &raddr, &rmask, af)) {
1836                                 /* table contains no address of type 'af' */
1837                                 if (rpool->cur != acur)
1838                                         goto try_next;
1839                                 return (1);
1840                         }
1841                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1842                         rpool->tblidx = -1;
1843                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
1844                             &rpool->tblidx, &rpool->counter,
1845                             &raddr, &rmask, af)) {
1846                                 /* table contains no address of type 'af' */
1847                                 if (rpool->cur != acur)
1848                                         goto try_next;
1849                                 return (1);
1850                         }
1851                 } else {
1852                         raddr = &rpool->cur->addr.v.a.addr;
1853                         rmask = &rpool->cur->addr.v.a.mask;
1854                         PF_ACPY(&rpool->counter, raddr, af);
1855                 }
1856
1857         get_addr:
1858                 PF_ACPY(naddr, &rpool->counter, af);
1859                 PF_AINC(&rpool->counter, af);
1860                 break;
1861         }
1862         if (*sn != NULL)
1863                 PF_ACPY(&(*sn)->raddr, naddr, af);
1864
1865         if (pf_status.debug >= PF_DEBUG_MISC &&
1866             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1867                 kprintf("pf_map_addr: selected address ");
1868                 pf_print_host(naddr, 0, af);
1869                 kprintf("\n");
1870         }
1871
1872         return (0);
1873 }
1874
1875 int
1876 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
1877     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
1878     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
1879     struct pf_src_node **sn)
1880 {
1881         struct pf_state         key;
1882         struct pf_addr          init_addr;
1883         u_int16_t               cut;
1884
1885         bzero(&init_addr, sizeof(init_addr));
1886         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
1887                 return (1);
1888
1889         do {
1890                 key.af = af;
1891                 key.proto = proto;
1892                 PF_ACPY(&key.ext.addr, daddr, key.af);
1893                 PF_ACPY(&key.gwy.addr, naddr, key.af);
1894                 key.ext.port = dport;
1895
1896                 /*
1897                  * port search; start random, step;
1898                  * similar 2 portloop in in_pcbbind
1899                  */
1900                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP)) {
1901                         key.gwy.port = 0;
1902                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
1903                                 return (0);
1904                 } else if (low == 0 && high == 0) {
1905                         key.gwy.port = *nport;
1906                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
1907                                 return (0);
1908                 } else if (low == high) {
1909                         key.gwy.port = htons(low);
1910                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
1911                                 *nport = htons(low);
1912                                 return (0);
1913                         }
1914                 } else {
1915                         u_int16_t tmp;
1916
1917                         if (low > high) {
1918                                 tmp = low;
1919                                 low = high;
1920                                 high = tmp;
1921                         }
1922                         /* low < high */
1923                         cut = karc4random() % (1 + high - low) + low;
1924                         /* low <= cut <= high */
1925                         for (tmp = cut; tmp <= high; ++(tmp)) {
1926                                 key.gwy.port = htons(tmp);
1927                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
1928                                     NULL) {
1929                                         *nport = htons(tmp);
1930                                         return (0);
1931                                 }
1932                         }
1933                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
1934                                 key.gwy.port = htons(tmp);
1935                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
1936                                     NULL) {
1937                                         *nport = htons(tmp);
1938                                         return (0);
1939                                 }
1940                         }
1941                 }
1942
1943                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
1944                 case PF_POOL_RANDOM:
1945                 case PF_POOL_ROUNDROBIN:
1946                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
1947                                 return (1);
1948                         break;
1949                 case PF_POOL_NONE:
1950                 case PF_POOL_SRCHASH:
1951                 case PF_POOL_BITMASK:
1952                 default:
1953                         return (1);
1954                 }
1955         } while (! PF_AEQ(&init_addr, naddr, af) );
1956
1957         return (1);                                     /* none available */
1958 }
1959
1960 struct pf_rule *
1961 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
1962     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
1963     struct pf_addr *daddr, u_int16_t dport, int rs_num)
1964 {
1965         struct pf_rule          *r, *rm = NULL, *anchorrule = NULL;
1966         struct pf_ruleset       *ruleset = NULL;
1967
1968         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
1969         while (r && rm == NULL) {
1970                 struct pf_rule_addr     *src = NULL, *dst = NULL;
1971                 struct pf_addr_wrap     *xdst = NULL;
1972
1973                 if (r->action == PF_BINAT && direction == PF_IN) {
1974                         src = &r->dst;
1975                         if (r->rpool.cur != NULL)
1976                                 xdst = &r->rpool.cur->addr;
1977                 } else {
1978                         src = &r->src;
1979                         dst = &r->dst;
1980                 }
1981
1982                 r->evaluations++;
1983                 if (r->kif != NULL &&
1984                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
1985                         r = r->skip[PF_SKIP_IFP].ptr;
1986                 else if (r->direction && r->direction != direction)
1987                         r = r->skip[PF_SKIP_DIR].ptr;
1988                 else if (r->af && r->af != pd->af)
1989                         r = r->skip[PF_SKIP_AF].ptr;
1990                 else if (r->proto && r->proto != pd->proto)
1991                         r = r->skip[PF_SKIP_PROTO].ptr;
1992                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->not))
1993                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
1994                             PF_SKIP_DST_ADDR].ptr;
1995                 else if (src->port_op && !pf_match_port(src->port_op,
1996                     src->port[0], src->port[1], sport))
1997                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
1998                             PF_SKIP_DST_PORT].ptr;
1999                 else if (dst != NULL &&
2000                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->not))
2001                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2002                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0))
2003                         r = TAILQ_NEXT(r, entries);
2004                 else if (dst != NULL && dst->port_op &&
2005                     !pf_match_port(dst->port_op, dst->port[0],
2006                     dst->port[1], dport))
2007                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2008                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2009                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2010                     off, pd->hdr.tcp), r->os_fingerprint)))
2011                         r = TAILQ_NEXT(r, entries);
2012                 else if (r->anchorname[0] && r->anchor == NULL)
2013                         r = TAILQ_NEXT(r, entries);
2014                 else if (r->anchor == NULL)
2015                                 rm = r;
2016                 else
2017                         PF_STEP_INTO_ANCHOR(r, anchorrule, ruleset, rs_num);
2018                 if (r == NULL && anchorrule != NULL)
2019                         PF_STEP_OUT_OF_ANCHOR(r, anchorrule, ruleset,
2020                             rs_num);
2021         }
2022         if (rm != NULL && (rm->action == PF_NONAT ||
2023             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2024                 return (NULL);
2025         return (rm);
2026 }
2027
2028 struct pf_rule *
2029 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2030     struct pfi_kif *kif, struct pf_src_node **sn,
2031     struct pf_addr *saddr, u_int16_t sport,
2032     struct pf_addr *daddr, u_int16_t dport,
2033     struct pf_addr *naddr, u_int16_t *nport)
2034 {
2035         struct pf_rule  *r = NULL;
2036
2037         if (direction == PF_OUT) {
2038                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2039                     sport, daddr, dport, PF_RULESET_BINAT);
2040                 if (r == NULL)
2041                         r = pf_match_translation(pd, m, off, direction, kif,
2042                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2043         } else {
2044                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2045                     sport, daddr, dport, PF_RULESET_RDR);
2046                 if (r == NULL)
2047                         r = pf_match_translation(pd, m, off, direction, kif,
2048                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2049         }
2050
2051         if (r != NULL) {
2052                 switch (r->action) {
2053                 case PF_NONAT:
2054                 case PF_NOBINAT:
2055                 case PF_NORDR:
2056                         return (NULL);
2057                 case PF_NAT:
2058                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2059                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2060                             r->rpool.proxy_port[1], sn)) {
2061                                 DPFPRINTF(PF_DEBUG_MISC,
2062                                     ("pf: NAT proxy port allocation "
2063                                     "(%u-%u) failed\n",
2064                                     r->rpool.proxy_port[0],
2065                                     r->rpool.proxy_port[1]));
2066                                 return (NULL);
2067                         }
2068                         break;
2069                 case PF_BINAT:
2070                         switch (direction) {
2071                         case PF_OUT:
2072                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2073                                         if (pd->af == AF_INET) {
2074                                                 if (r->rpool.cur->addr.p.dyn->
2075                                                     pfid_acnt4 < 1)
2076                                                         return (NULL);
2077                                                 PF_POOLMASK(naddr,
2078                                                     &r->rpool.cur->addr.p.dyn->
2079                                                     pfid_addr4,
2080                                                     &r->rpool.cur->addr.p.dyn->
2081                                                     pfid_mask4,
2082                                                     saddr, AF_INET);
2083                                         } else {
2084                                                 if (r->rpool.cur->addr.p.dyn->
2085                                                     pfid_acnt6 < 1)
2086                                                         return (NULL);
2087                                                 PF_POOLMASK(naddr,
2088                                                     &r->rpool.cur->addr.p.dyn->
2089                                                     pfid_addr6,
2090                                                     &r->rpool.cur->addr.p.dyn->
2091                                                     pfid_mask6,
2092                                                     saddr, AF_INET6);
2093                                         }
2094                                 } else
2095                                         PF_POOLMASK(naddr,
2096                                             &r->rpool.cur->addr.v.a.addr,
2097                                             &r->rpool.cur->addr.v.a.mask,
2098                                             saddr, pd->af);
2099                                 break;
2100                         case PF_IN:
2101                                 if (r->src.addr.type == PF_ADDR_DYNIFTL){
2102                                         if (pd->af == AF_INET) {
2103                                                 if (r->src.addr.p.dyn->
2104                                                     pfid_acnt4 < 1)
2105                                                         return (NULL);
2106                                                 PF_POOLMASK(naddr,
2107                                                     &r->src.addr.p.dyn->
2108                                                     pfid_addr4,
2109                                                     &r->src.addr.p.dyn->
2110                                                     pfid_mask4,
2111                                                     daddr, AF_INET);
2112                                         } else {
2113                                                 if (r->src.addr.p.dyn->
2114                                                     pfid_acnt6 < 1)
2115                                                         return (NULL);
2116                                                 PF_POOLMASK(naddr,
2117                                                     &r->src.addr.p.dyn->
2118                                                     pfid_addr6,
2119                                                     &r->src.addr.p.dyn->
2120                                                     pfid_mask6,
2121                                                     daddr, AF_INET6);
2122                                         }
2123                                 } else
2124                                         PF_POOLMASK(naddr,
2125                                             &r->src.addr.v.a.addr,
2126                                             &r->src.addr.v.a.mask, daddr,
2127                                             pd->af);
2128                                 break;
2129                         }
2130                         break;
2131                 case PF_RDR: {
2132                         if (pf_map_addr(r->af, r, saddr, naddr, NULL, sn))
2133                                 return (NULL);
2134
2135                         if (r->rpool.proxy_port[1]) {
2136                                 u_int32_t       tmp_nport;
2137
2138                                 tmp_nport = ((ntohs(dport) -
2139                                     ntohs(r->dst.port[0])) %
2140                                     (r->rpool.proxy_port[1] -
2141                                     r->rpool.proxy_port[0] + 1)) +
2142                                     r->rpool.proxy_port[0];
2143
2144                                 /* wrap around if necessary */
2145                                 if (tmp_nport > 65535)
2146                                         tmp_nport -= 65535;
2147                                 *nport = htons((u_int16_t)tmp_nport);
2148                         } else if (r->rpool.proxy_port[0])
2149                                 *nport = htons(r->rpool.proxy_port[0]);
2150                         break;
2151                 }
2152                 default:
2153                         return (NULL);
2154                 }
2155         }
2156
2157         return (r);
2158 }
2159
2160 #ifdef SMP
2161 struct netmsg_hashlookup {
2162         struct netmsg           nm_netmsg;
2163         struct inpcb            **nm_pinp;
2164         struct inpcbinfo        *nm_pcbinfo;
2165         struct pf_addr          *nm_saddr;
2166         struct pf_addr          *nm_daddr;
2167         uint16_t                nm_sport;
2168         uint16_t                nm_dport;
2169         sa_family_t             nm_af;
2170 };
2171
2172 static void
2173 in_pcblookup_hash_handler(struct netmsg *msg0)
2174 {
2175         struct netmsg_hashlookup *msg = (struct netmsg_hashlookup *)msg0;
2176
2177         if (msg->nm_af == AF_INET)
2178                 *msg->nm_pinp = in_pcblookup_hash(msg->nm_pcbinfo,
2179                     msg->nm_saddr->v4, msg->nm_sport, msg->nm_daddr->v4,
2180                     msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2181 #ifdef INET6
2182         else
2183                 *msg->nm_pinp = in6_pcblookup_hash(msg->nm_pcbinfo,
2184                     &msg->nm_saddr->v6, msg->nm_sport, &msg->nm_daddr->v6,
2185                     msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2186 #endif /* INET6 */
2187         lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, 0);
2188 }
2189 #endif /* SMP */
2190
2191 int
2192 pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd)
2193 {
2194         struct pf_addr          *saddr, *daddr;
2195         u_int16_t                sport, dport;
2196         struct inpcbinfo        *pi;
2197         struct inpcb            *inp;
2198 #ifdef SMP
2199         struct netmsg_hashlookup *msg = NULL;
2200 #endif
2201         int                      pi_cpu = 0;
2202
2203         *uid = UID_MAX;
2204         *gid = GID_MAX;
2205         if (direction == PF_IN) {
2206                 saddr = pd->src;
2207                 daddr = pd->dst;
2208         } else {
2209                 saddr = pd->dst;
2210                 daddr = pd->src;
2211         }
2212         switch (pd->proto) {
2213         case IPPROTO_TCP:
2214                 sport = pd->hdr.tcp->th_sport;
2215                 dport = pd->hdr.tcp->th_dport;
2216
2217                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2218                 pi = &tcbinfo[pi_cpu];
2219 #ifdef SMP
2220                 /*
2221                  * Our netstack runs lockless on MP systems
2222                  * (only for TCP connections at the moment).
2223                  *
2224                  * As we are not allowed to read another CPU's tcbinfo,
2225                  * we have to ask that CPU via remote call to search the
2226                  * table for us.
2227                  *
2228                  * Prepare a msg iff data belongs to another CPU.
2229                  */
2230                 if (pi_cpu != mycpu->gd_cpuid) {
2231                         msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2232                         netmsg_init(&msg->nm_netmsg, &netisr_afree_rport, 0,
2233                                     in_pcblookup_hash_handler);
2234                         msg->nm_pinp = &inp;
2235                         msg->nm_pcbinfo = pi;
2236                         msg->nm_saddr = saddr;
2237                         msg->nm_sport = sport;
2238                         msg->nm_daddr = daddr;
2239                         msg->nm_dport = dport;
2240                         msg->nm_af = pd->af;
2241                 }
2242 #endif /* SMP */
2243                 break;
2244         case IPPROTO_UDP:
2245                 sport = pd->hdr.udp->uh_sport;
2246                 dport = pd->hdr.udp->uh_dport;
2247                 pi = &udbinfo;
2248                 break;
2249         default:
2250                 return (0);
2251         }
2252         if (direction != PF_IN) {
2253                 u_int16_t       p;
2254
2255                 p = sport;
2256                 sport = dport;
2257                 dport = p;
2258         }
2259         switch (pd->af) {
2260 #ifdef INET6
2261         case AF_INET6:
2262 #ifdef SMP
2263                 /*
2264                  * Query other CPU, second part
2265                  *
2266                  * msg only gets initialized when:
2267                  * 1) packet is TCP
2268                  * 2) the info belongs to another CPU
2269                  *
2270                  * Use some switch/case magic to avoid code duplication.
2271                  */
2272                 if (msg == NULL)
2273 #endif /* SMP */
2274                 {
2275                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2276                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2277
2278                         if (inp == NULL)
2279                                 return (0);
2280                         break;
2281                 }
2282                 /* FALLTHROUGH if SMP and on other CPU */
2283 #endif /* INET6 */
2284         case AF_INET:
2285 #ifdef SMP
2286                 if (msg != NULL) {
2287                         lwkt_sendmsg(tcp_cport(pi_cpu),
2288                                      &msg->nm_netmsg.nm_lmsg);
2289                 } else
2290 #endif /* SMP */
2291                 {
2292                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2293                             dport, INPLOOKUP_WILDCARD, NULL);
2294                 }
2295                 if (inp == NULL)
2296                         return (0);
2297                 break;
2298
2299         default:
2300                 return (0);
2301         }
2302         *uid = inp->inp_socket->so_cred->cr_uid;
2303         *gid = inp->inp_socket->so_cred->cr_groups[0];
2304         return (1);
2305 }
2306
2307 u_int8_t
2308 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2309 {
2310         int              hlen;
2311         u_int8_t         hdr[60];
2312         u_int8_t        *opt, optlen;
2313         u_int8_t         wscale = 0;
2314
2315         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
2316         if (hlen <= sizeof(struct tcphdr))
2317                 return (0);
2318         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2319                 return (0);
2320         opt = hdr + sizeof(struct tcphdr);
2321         hlen -= sizeof(struct tcphdr);
2322         while (hlen >= 3) {
2323                 switch (*opt) {
2324                 case TCPOPT_EOL:
2325                 case TCPOPT_NOP:
2326                         ++opt;
2327                         --hlen;
2328                         break;
2329                 case TCPOPT_WINDOW:
2330                         wscale = opt[2];
2331                         if (wscale > TCP_MAX_WINSHIFT)
2332                                 wscale = TCP_MAX_WINSHIFT;
2333                         wscale |= PF_WSCALE_FLAG;
2334                         /* FALLTHROUGH */
2335                 default:
2336                         optlen = opt[1];
2337                         if (optlen < 2)
2338                                 optlen = 2;
2339                         hlen -= optlen;
2340                         opt += optlen;
2341                         break;
2342                 }
2343         }
2344         return (wscale);
2345 }
2346
2347 u_int16_t
2348 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2349 {
2350         int              hlen;
2351         u_int8_t         hdr[60];
2352         u_int8_t        *opt, optlen;
2353         u_int16_t        mss = tcp_mssdflt;
2354
2355         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
2356         if (hlen <= sizeof(struct tcphdr))
2357                 return (0);
2358         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2359                 return (0);
2360         opt = hdr + sizeof(struct tcphdr);
2361         hlen -= sizeof(struct tcphdr);
2362         while (hlen >= TCPOLEN_MAXSEG) {
2363                 switch (*opt) {
2364                 case TCPOPT_EOL:
2365                 case TCPOPT_NOP:
2366                         ++opt;
2367                         --hlen;
2368                         break;
2369                 case TCPOPT_MAXSEG:
2370                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2371                         /* FALLTHROUGH */
2372                 default:
2373                         optlen = opt[1];
2374                         if (optlen < 2)
2375                                 optlen = 2;
2376                         hlen -= optlen;
2377                         opt += optlen;
2378                         break;
2379                 }
2380         }
2381         return (mss);
2382 }
2383
2384 u_int16_t
2385 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2386 {
2387 #ifdef INET
2388         struct sockaddr_in      *dst;
2389         struct route             ro;
2390 #endif /* INET */
2391 #ifdef INET6
2392         struct sockaddr_in6     *dst6;
2393         struct route_in6         ro6;
2394 #endif /* INET6 */
2395         struct rtentry          *rt = NULL;
2396         int                      hlen = 0;
2397         u_int16_t                mss = tcp_mssdflt;
2398
2399         switch (af) {
2400 #ifdef INET
2401         case AF_INET:
2402                 hlen = sizeof(struct ip);
2403                 bzero(&ro, sizeof(ro));
2404                 dst = (struct sockaddr_in *)&ro.ro_dst;
2405                 dst->sin_family = AF_INET;
2406                 dst->sin_len = sizeof(*dst);
2407                 dst->sin_addr = addr->v4;
2408                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2409                 rt = ro.ro_rt;
2410                 break;
2411 #endif /* INET */
2412 #ifdef INET6
2413         case AF_INET6:
2414                 hlen = sizeof(struct ip6_hdr);
2415                 bzero(&ro6, sizeof(ro6));
2416                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2417                 dst6->sin6_family = AF_INET6;
2418                 dst6->sin6_len = sizeof(*dst6);
2419                 dst6->sin6_addr = addr->v6;
2420                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
2421                 rt = ro6.ro_rt;
2422                 break;
2423 #endif /* INET6 */
2424         }
2425
2426         if (rt && rt->rt_ifp) {
2427                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2428                 mss = max(tcp_mssdflt, mss);
2429                 RTFREE(rt);
2430         }
2431         mss = min(mss, offer);
2432         mss = max(mss, 64);             /* sanity - at least max opt space */
2433         return (mss);
2434 }
2435
2436 void
2437 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2438 {
2439         struct pf_rule *r = s->rule.ptr;
2440
2441         s->rt_kif = NULL;
2442         if (!r->rt || r->rt == PF_FASTROUTE)
2443                 return;
2444         switch (s->af) {
2445 #ifdef INET
2446         case AF_INET:
2447                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2448                     &s->nat_src_node);
2449                 s->rt_kif = r->rpool.cur->kif;
2450                 break;
2451 #endif /* INET */
2452 #ifdef INET6
2453         case AF_INET6:
2454                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2455                     &s->nat_src_node);
2456                 s->rt_kif = r->rpool.cur->kif;
2457                 break;
2458 #endif /* INET6 */
2459         }
2460 }
2461
2462 int
2463 pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
2464     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2465     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2466 {
2467         struct pf_rule          *nr = NULL;
2468         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
2469         struct tcphdr           *th = pd->hdr.tcp;
2470         u_int16_t                bport, nport = 0;
2471         sa_family_t              af = pd->af;
2472         int                      lookup = -1;
2473         uid_t                    uid;
2474         gid_t                    gid;
2475         struct pf_rule          *r, *a = NULL;
2476         struct pf_ruleset       *ruleset = NULL;
2477         struct pf_src_node      *nsn = NULL;
2478         u_short                  reason;
2479         int                      rewrite = 0;
2480         int                      tag = -1;
2481         u_int16_t                mss = tcp_mssdflt;
2482
2483         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2484
2485         if (direction == PF_OUT) {
2486                 bport = nport = th->th_sport;
2487                 /* check outgoing packet for BINAT/NAT */
2488                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2489                     saddr, th->th_sport, daddr, th->th_dport,
2490                     &pd->naddr, &nport)) != NULL) {
2491                         PF_ACPY(&pd->baddr, saddr, af);
2492                         pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2493                             &th->th_sum, &pd->naddr, nport, 0, af);
2494                         rewrite++;
2495                         if (nr->natpass)
2496                                 r = NULL;
2497                         pd->nat_rule = nr;
2498                 }
2499         } else {
2500                 bport = nport = th->th_dport;
2501                 /* check incoming packet for BINAT/RDR */
2502                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2503                     saddr, th->th_sport, daddr, th->th_dport,
2504                     &pd->naddr, &nport)) != NULL) {
2505                         PF_ACPY(&pd->baddr, daddr, af);
2506                         pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2507                             &th->th_sum, &pd->naddr, nport, 0, af);
2508                         rewrite++;
2509                         if (nr->natpass)
2510                                 r = NULL;
2511                         pd->nat_rule = nr;
2512                 }
2513         }
2514
2515         while (r != NULL) {
2516                 r->evaluations++;
2517                 if (r->kif != NULL &&
2518                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2519                         r = r->skip[PF_SKIP_IFP].ptr;
2520                 else if (r->direction && r->direction != direction)
2521                         r = r->skip[PF_SKIP_DIR].ptr;
2522                 else if (r->af && r->af != af)
2523                         r = r->skip[PF_SKIP_AF].ptr;
2524                 else if (r->proto && r->proto != IPPROTO_TCP)
2525                         r = r->skip[PF_SKIP_PROTO].ptr;
2526                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2527                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2528                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2529                     r->src.port[0], r->src.port[1], th->th_sport))
2530                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
2531                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2532                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2533                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2534                     r->dst.port[0], r->dst.port[1], th->th_dport))
2535                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2536                 else if (r->tos && !(r->tos & pd->tos))
2537                         r = TAILQ_NEXT(r, entries);
2538                 else if (r->rule_flag & PFRULE_FRAGMENT)
2539                         r = TAILQ_NEXT(r, entries);
2540                 else if ((r->flagset & th->th_flags) != r->flags)
2541                         r = TAILQ_NEXT(r, entries);
2542                 else if (r->uid.op && (lookup != -1 || (lookup =
2543                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2544                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2545                     uid))
2546                         r = TAILQ_NEXT(r, entries);
2547                 else if (r->gid.op && (lookup != -1 || (lookup =
2548                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2549                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2550                     gid))
2551                         r = TAILQ_NEXT(r, entries);
2552                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2553                         r = TAILQ_NEXT(r, entries);
2554                 else if (r->anchorname[0] && r->anchor == NULL)
2555                         r = TAILQ_NEXT(r, entries);
2556                 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
2557                     pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
2558                         r = TAILQ_NEXT(r, entries);
2559                 else {
2560                         if (r->tag)
2561                                 tag = r->tag;
2562                         if (r->anchor == NULL) {
2563                                 *rm = r;
2564                                 *am = a;
2565                                 *rsm = ruleset;
2566                                 if ((*rm)->quick)
2567                                         break;
2568                                 r = TAILQ_NEXT(r, entries);
2569                         } else
2570                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2571                                     PF_RULESET_FILTER);
2572                 }
2573                 if (r == NULL && a != NULL)
2574                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2575                             PF_RULESET_FILTER);
2576         }
2577         r = *rm;
2578         a = *am;
2579         ruleset = *rsm;
2580
2581         REASON_SET(&reason, PFRES_MATCH);
2582
2583         if (r->log) {
2584                 if (rewrite)
2585                         m_copyback(m, off, sizeof(*th), (caddr_t)th);
2586                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2587         }
2588
2589         if ((r->action == PF_DROP) &&
2590             ((r->rule_flag & PFRULE_RETURNRST) ||
2591             (r->rule_flag & PFRULE_RETURNICMP) ||
2592             (r->rule_flag & PFRULE_RETURN))) {
2593                 /* undo NAT changes, if they have taken place */
2594                 if (nr != NULL) {
2595                         if (direction == PF_OUT) {
2596                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2597                                     &th->th_sum, &pd->baddr, bport, 0, af);
2598                                 rewrite++;
2599                         } else {
2600                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2601                                     &th->th_sum, &pd->baddr, bport, 0, af);
2602                                 rewrite++;
2603                         }
2604                 }
2605                 if (((r->rule_flag & PFRULE_RETURNRST) ||
2606                     (r->rule_flag & PFRULE_RETURN)) &&
2607                     !(th->th_flags & TH_RST)) {
2608                         u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
2609
2610                         if (th->th_flags & TH_SYN)
2611                                 ack++;
2612                         if (th->th_flags & TH_FIN)
2613                                 ack++;
2614                         pf_send_tcp(r, af, pd->dst,
2615                             pd->src, th->th_dport, th->th_sport,
2616                             ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
2617                             r->return_ttl);
2618                 } else if ((af == AF_INET) && r->return_icmp)
2619                         pf_send_icmp(m, r->return_icmp >> 8,
2620                             r->return_icmp & 255, af, r);
2621                 else if ((af == AF_INET6) && r->return_icmp6)
2622                         pf_send_icmp(m, r->return_icmp6 >> 8,
2623                             r->return_icmp6 & 255, af, r);
2624         }
2625
2626         if (r->action == PF_DROP)
2627                 return (PF_DROP);
2628
2629         pf_tag_packet(m, tag);
2630
2631         if (r->keep_state || nr != NULL ||
2632             (pd->flags & PFDESC_TCP_NORM)) {
2633                 /* create new state */
2634                 u_int16_t        len;
2635                 struct pf_state *s = NULL;
2636                 struct pf_src_node *sn = NULL;
2637
2638                 len = pd->tot_len - off - (th->th_off << 2);
2639
2640                 /* check maximums */
2641                 if (r->max_states && (r->states >= r->max_states))
2642                         goto cleanup;
2643                 /* src node for flter rule */
2644                 if ((r->rule_flag & PFRULE_SRCTRACK ||
2645                     r->rpool.opts & PF_POOL_STICKYADDR) &&
2646                     pf_insert_src_node(&sn, r, saddr, af) != 0)
2647                         goto cleanup;
2648                 /* src node for translation rule */
2649                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2650                     ((direction == PF_OUT &&
2651                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2652                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2653                         goto cleanup;
2654                 s = pool_get(&pf_state_pl, PR_NOWAIT);
2655                 if (s == NULL) {
2656 cleanup:
2657                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2658                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2659                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2660                                 pf_status.src_nodes--;
2661                                 pool_put(&pf_src_tree_pl, sn);
2662                         }
2663                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2664                             nsn->expire == 0) {
2665                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2666                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2667                                 pf_status.src_nodes--;
2668                                 pool_put(&pf_src_tree_pl, nsn);
2669                         }
2670                         REASON_SET(&reason, PFRES_MEMORY);
2671                         return (PF_DROP);
2672                 }
2673                 bzero(s, sizeof(*s));
2674                 r->states++;
2675                 if (a != NULL)
2676                         a->states++;
2677                 s->rule.ptr = r;
2678                 s->nat_rule.ptr = nr;
2679                 if (s->nat_rule.ptr != NULL)
2680                         s->nat_rule.ptr->states++;
2681                 s->anchor.ptr = a;
2682                 s->allow_opts = r->allow_opts;
2683                 s->log = r->log & 2;
2684                 s->proto = IPPROTO_TCP;
2685                 s->direction = direction;
2686                 s->af = af;
2687                 if (direction == PF_OUT) {
2688                         PF_ACPY(&s->gwy.addr, saddr, af);
2689                         s->gwy.port = th->th_sport;             /* sport */
2690                         PF_ACPY(&s->ext.addr, daddr, af);
2691                         s->ext.port = th->th_dport;
2692                         if (nr != NULL) {
2693                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
2694                                 s->lan.port = bport;
2695                         } else {
2696                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2697                                 s->lan.port = s->gwy.port;
2698                         }
2699                 } else {
2700                         PF_ACPY(&s->lan.addr, daddr, af);
2701                         s->lan.port = th->th_dport;
2702                         PF_ACPY(&s->ext.addr, saddr, af);
2703                         s->ext.port = th->th_sport;
2704                         if (nr != NULL) {
2705                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
2706                                 s->gwy.port = bport;
2707                         } else {
2708                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2709                                 s->gwy.port = s->lan.port;
2710                         }
2711                 }
2712
2713                 s->src.seqlo = ntohl(th->th_seq);
2714                 s->src.seqhi = s->src.seqlo + len + 1;
2715                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2716                     r->keep_state == PF_STATE_MODULATE) {
2717                         /* Generate sequence number modulator */
2718                         while ((s->src.seqdiff = karc4random()) == 0)
2719                                 ;
2720                         pf_change_a(&th->th_seq, &th->th_sum,
2721                             htonl(s->src.seqlo + s->src.seqdiff), 0);
2722                         rewrite = 1;
2723                 } else
2724                         s->src.seqdiff = 0;
2725                 if (th->th_flags & TH_SYN) {
2726                         s->src.seqhi++;
2727                         s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
2728                 }
2729                 s->src.max_win = MAX(ntohs(th->th_win), 1);
2730                 if (s->src.wscale & PF_WSCALE_MASK) {
2731                         /* Remove scale factor from initial window */
2732                         int win = s->src.max_win;
2733                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
2734                         s->src.max_win = (win - 1) >>
2735                             (s->src.wscale & PF_WSCALE_MASK);
2736                 }
2737                 if (th->th_flags & TH_FIN)
2738                         s->src.seqhi++;
2739                 s->dst.seqhi = 1;
2740                 s->dst.max_win = 1;
2741                 s->src.state = TCPS_SYN_SENT;
2742                 s->dst.state = TCPS_CLOSED;
2743                 s->creation = time_second;
2744                 s->expire = time_second;
2745                 s->timeout = PFTM_TCP_FIRST_PACKET;
2746                 pf_set_rt_ifp(s, saddr);
2747                 if (sn != NULL) {
2748                         s->src_node = sn;
2749                         s->src_node->states++;
2750                 }
2751                 if (nsn != NULL) {
2752                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
2753                         s->nat_src_node = nsn;
2754                         s->nat_src_node->states++;
2755                 }
2756                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
2757                     off, pd, th, &s->src, &s->dst)) {
2758                         REASON_SET(&reason, PFRES_MEMORY);
2759                         pf_src_tree_remove_state(s);
2760                         pool_put(&pf_state_pl, s);
2761                         return (PF_DROP);
2762                 }
2763                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
2764                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, &s->src,
2765                     &s->dst, &rewrite)) {
2766                         pf_normalize_tcp_cleanup(s);
2767                         pf_src_tree_remove_state(s);
2768                         pool_put(&pf_state_pl, s);
2769                         return (PF_DROP);
2770                 }
2771                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
2772                         pf_normalize_tcp_cleanup(s);
2773                         REASON_SET(&reason, PFRES_MEMORY);
2774                         pf_src_tree_remove_state(s);
2775                         pool_put(&pf_state_pl, s);
2776                         return (PF_DROP);
2777                 } else
2778                         *sm = s;
2779                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2780                     r->keep_state == PF_STATE_SYNPROXY) {
2781                         s->src.state = PF_TCPS_PROXY_SRC;
2782                         if (nr != NULL) {
2783                                 if (direction == PF_OUT) {
2784                                         pf_change_ap(saddr, &th->th_sport,
2785                                             pd->ip_sum, &th->th_sum, &pd->baddr,
2786                                             bport, 0, af);
2787                                 } else {
2788                                         pf_change_ap(daddr, &th->th_dport,
2789                                             pd->ip_sum, &th->th_sum, &pd->baddr,
2790                                             bport, 0, af);
2791                                 }
2792                         }
2793                         s->src.seqhi = karc4random();
2794                         /* Find mss option */
2795                         mss = pf_get_mss(m, off, th->th_off, af);
2796                         mss = pf_calc_mss(saddr, af, mss);
2797                         mss = pf_calc_mss(daddr, af, mss);
2798                         s->src.mss = mss;
2799                         pf_send_tcp(r, af, daddr, saddr, th->th_dport,
2800                             th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
2801                             TH_SYN|TH_ACK, 0, s->src.mss, 0);
2802                         return (PF_SYNPROXY_DROP);
2803                 }
2804         }
2805
2806         /* copy back packet headers if we performed NAT operations */
2807         if (rewrite)
2808                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2809
2810         return (PF_PASS);
2811 }
2812
2813 int
2814 pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
2815     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2816     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2817 {
2818         struct pf_rule          *nr = NULL;
2819         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
2820         struct udphdr           *uh = pd->hdr.udp;
2821         u_int16_t                bport, nport = 0;
2822         sa_family_t              af = pd->af;
2823         int                      lookup = -1;
2824         uid_t                    uid;
2825         gid_t                    gid;
2826         struct pf_rule          *r, *a = NULL;
2827         struct pf_ruleset       *ruleset = NULL;
2828         struct pf_src_node      *nsn = NULL;
2829         u_short                  reason;
2830         int                      rewrite = 0;
2831         int                      tag = -1;
2832
2833         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2834
2835         if (direction == PF_OUT) {
2836                 bport = nport = uh->uh_sport;
2837                 /* check outgoing packet for BINAT/NAT */
2838                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2839                     saddr, uh->uh_sport, daddr, uh->uh_dport,
2840                     &pd->naddr, &nport)) != NULL) {
2841                         PF_ACPY(&pd->baddr, saddr, af);
2842                         pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2843                             &uh->uh_sum, &pd->naddr, nport, 1, af);
2844                         rewrite++;
2845                         if (nr->natpass)
2846                                 r = NULL;
2847                         pd->nat_rule = nr;
2848                 }
2849         } else {
2850                 bport = nport = uh->uh_dport;
2851                 /* check incoming packet for BINAT/RDR */
2852                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2853                     saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
2854                     &nport)) != NULL) {
2855                         PF_ACPY(&pd->baddr, daddr, af);
2856                         pf_change_ap(daddr, &uh->uh_dport, 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         }
2864
2865         while (r != NULL) {
2866                 r->evaluations++;
2867                 if (r->kif != NULL &&
2868                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2869                         r = r->skip[PF_SKIP_IFP].ptr;
2870                 else if (r->direction && r->direction != direction)
2871                         r = r->skip[PF_SKIP_DIR].ptr;
2872                 else if (r->af && r->af != af)
2873                         r = r->skip[PF_SKIP_AF].ptr;
2874                 else if (r->proto && r->proto != IPPROTO_UDP)
2875                         r = r->skip[PF_SKIP_PROTO].ptr;
2876                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2877                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2878                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2879                     r->src.port[0], r->src.port[1], uh->uh_sport))
2880                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
2881                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2882                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2883                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2884                     r->dst.port[0], r->dst.port[1], uh->uh_dport))
2885                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2886                 else if (r->tos && !(r->tos & pd->tos))
2887                         r = TAILQ_NEXT(r, entries);
2888                 else if (r->rule_flag & PFRULE_FRAGMENT)
2889                         r = TAILQ_NEXT(r, entries);
2890                 else if (r->uid.op && (lookup != -1 || (lookup =
2891                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2892                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2893                     uid))
2894                         r = TAILQ_NEXT(r, entries);
2895                 else if (r->gid.op && (lookup != -1 || (lookup =
2896                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2897                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2898                     gid))
2899                         r = TAILQ_NEXT(r, entries);
2900                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2901                         r = TAILQ_NEXT(r, entries);
2902                 else if (r->anchorname[0] && r->anchor == NULL)
2903                         r = TAILQ_NEXT(r, entries);
2904                 else if (r->os_fingerprint != PF_OSFP_ANY)
2905                         r = TAILQ_NEXT(r, entries);
2906                 else {
2907                         if (r->tag)
2908                                 tag = r->tag;
2909                         if (r->anchor == NULL) {
2910                                 *rm = r;
2911                                 *am = a;
2912                                 *rsm = ruleset;
2913                                 if ((*rm)->quick)
2914                                         break;
2915                                 r = TAILQ_NEXT(r, entries);
2916                         } else
2917                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2918                                     PF_RULESET_FILTER);
2919                 }
2920                 if (r == NULL && a != NULL)
2921                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2922                             PF_RULESET_FILTER);
2923         }
2924         r = *rm;
2925         a = *am;
2926         ruleset = *rsm;
2927
2928         REASON_SET(&reason, PFRES_MATCH);
2929
2930         if (r->log) {
2931                 if (rewrite)
2932                         m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
2933                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2934         }
2935
2936         if ((r->action == PF_DROP) &&
2937             ((r->rule_flag & PFRULE_RETURNICMP) ||
2938             (r->rule_flag & PFRULE_RETURN))) {
2939                 /* undo NAT changes, if they have taken place */
2940                 if (nr != NULL) {
2941                         if (direction == PF_OUT) {
2942                                 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2943                                     &uh->uh_sum, &pd->baddr, bport, 1, af);
2944                                 rewrite++;
2945                         } else {
2946                                 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2947                                     &uh->uh_sum, &pd->baddr, bport, 1, af);
2948                                 rewrite++;
2949                         }
2950                 }
2951                 if ((af == AF_INET) && r->return_icmp)
2952                         pf_send_icmp(m, r->return_icmp >> 8,
2953                             r->return_icmp & 255, af, r);
2954                 else if ((af == AF_INET6) && r->return_icmp6)
2955                         pf_send_icmp(m, r->return_icmp6 >> 8,
2956                             r->return_icmp6 & 255, af, r);
2957         }
2958
2959         if (r->action == PF_DROP)
2960                 return (PF_DROP);
2961
2962         pf_tag_packet(m, tag);
2963
2964         if (r->keep_state || nr != NULL) {
2965                 /* create new state */
2966                 struct pf_state *s = NULL;
2967                 struct pf_src_node *sn = NULL;
2968
2969                 /* check maximums */
2970                 if (r->max_states && (r->states >= r->max_states))
2971                         goto cleanup;
2972                 /* src node for flter rule */
2973                 if ((r->rule_flag & PFRULE_SRCTRACK ||
2974                     r->rpool.opts & PF_POOL_STICKYADDR) &&
2975                     pf_insert_src_node(&sn, r, saddr, af) != 0)
2976                         goto cleanup;
2977                 /* src node for translation rule */
2978                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2979                     ((direction == PF_OUT &&
2980                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2981                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2982                         goto cleanup;
2983                 s = pool_get(&pf_state_pl, PR_NOWAIT);
2984                 if (s == NULL) {
2985 cleanup:
2986                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2987                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2988                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2989                                 pf_status.src_nodes--;
2990                                 pool_put(&pf_src_tree_pl, sn);
2991                         }
2992                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2993                             nsn->expire == 0) {
2994                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2995                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2996                                 pf_status.src_nodes--;
2997                                 pool_put(&pf_src_tree_pl, nsn);
2998                         }
2999                         REASON_SET(&reason, PFRES_MEMORY);
3000                         return (PF_DROP);
3001                 }
3002                 bzero(s, sizeof(*s));
3003                 r->states++;
3004                 if (a != NULL)
3005                         a->states++;
3006                 s->rule.ptr = r;
3007                 s->nat_rule.ptr = nr;
3008                 if (s->nat_rule.ptr != NULL)
3009                         s->nat_rule.ptr->states++;
3010                 s->anchor.ptr = a;
3011                 s->allow_opts = r->allow_opts;
3012                 s->log = r->log & 2;
3013                 s->proto = IPPROTO_UDP;
3014                 s->direction = direction;
3015                 s->af = af;
3016                 if (direction == PF_OUT) {
3017                         PF_ACPY(&s->gwy.addr, saddr, af);
3018                         s->gwy.port = uh->uh_sport;
3019                         PF_ACPY(&s->ext.addr, daddr, af);
3020                         s->ext.port = uh->uh_dport;
3021                         if (nr != NULL) {
3022                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3023                                 s->lan.port = bport;
3024                         } else {
3025                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3026                                 s->lan.port = s->gwy.port;
3027                         }
3028                 } else {
3029                         PF_ACPY(&s->lan.addr, daddr, af);
3030                         s->lan.port = uh->uh_dport;
3031                         PF_ACPY(&s->ext.addr, saddr, af);
3032                         s->ext.port = uh->uh_sport;
3033                         if (nr != NULL) {
3034                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3035                                 s->gwy.port = bport;
3036                         } else {
3037                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3038                                 s->gwy.port = s->lan.port;
3039                         }
3040                 }
3041                 s->src.state = PFUDPS_SINGLE;
3042                 s->dst.state = PFUDPS_NO_TRAFFIC;
3043                 s->creation = time_second;
3044                 s->expire = time_second;
3045                 s->timeout = PFTM_UDP_FIRST_PACKET;
3046                 pf_set_rt_ifp(s, saddr);
3047                 if (sn != NULL) {
3048                         s->src_node = sn;
3049                         s->src_node->states++;
3050                 }
3051                 if (nsn != NULL) {
3052                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3053                         s->nat_src_node = nsn;
3054                         s->nat_src_node->states++;
3055                 }
3056                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3057                         REASON_SET(&reason, PFRES_MEMORY);
3058                         pf_src_tree_remove_state(s);
3059                         pool_put(&pf_state_pl, s);
3060                         return (PF_DROP);
3061                 } else
3062                         *sm = s;
3063         }
3064
3065         /* copy back packet headers if we performed NAT operations */
3066         if (rewrite)
3067                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3068
3069         return (PF_PASS);
3070 }
3071
3072 int
3073 pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3074     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3075     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3076 {
3077         struct pf_rule          *nr = NULL;
3078         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3079         struct pf_rule          *r, *a = NULL;
3080         struct pf_ruleset       *ruleset = NULL;
3081         struct pf_src_node      *nsn = NULL;
3082         u_short                  reason;
3083         u_int16_t                icmpid = 0;
3084         sa_family_t              af = pd->af;
3085         u_int8_t                 icmptype = 0, icmpcode = 0;
3086         int                      state_icmp = 0;
3087         int                      tag = -1;
3088 #ifdef INET6
3089         int                      rewrite = 0;
3090 #endif /* INET6 */
3091
3092         switch (pd->proto) {
3093 #ifdef INET
3094         case IPPROTO_ICMP:
3095                 icmptype = pd->hdr.icmp->icmp_type;
3096                 icmpcode = pd->hdr.icmp->icmp_code;
3097                 icmpid = pd->hdr.icmp->icmp_id;
3098
3099                 if (icmptype == ICMP_UNREACH ||
3100                     icmptype == ICMP_SOURCEQUENCH ||
3101                     icmptype == ICMP_REDIRECT ||
3102                     icmptype == ICMP_TIMXCEED ||
3103                     icmptype == ICMP_PARAMPROB)
3104                         state_icmp++;
3105                 break;
3106 #endif /* INET */
3107 #ifdef INET6
3108         case IPPROTO_ICMPV6:
3109                 icmptype = pd->hdr.icmp6->icmp6_type;
3110                 icmpcode = pd->hdr.icmp6->icmp6_code;
3111                 icmpid = pd->hdr.icmp6->icmp6_id;
3112
3113                 if (icmptype == ICMP6_DST_UNREACH ||
3114                     icmptype == ICMP6_PACKET_TOO_BIG ||
3115                     icmptype == ICMP6_TIME_EXCEEDED ||
3116                     icmptype == ICMP6_PARAM_PROB)
3117                         state_icmp++;
3118                 break;
3119 #endif /* INET6 */
3120         }
3121
3122         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3123
3124         if (direction == PF_OUT) {
3125                 /* check outgoing packet for BINAT/NAT */
3126                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3127                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3128                         PF_ACPY(&pd->baddr, saddr, af);
3129                         switch (af) {
3130 #ifdef INET
3131                         case AF_INET:
3132                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3133                                     pd->naddr.v4.s_addr, 0);
3134                                 break;
3135 #endif /* INET */
3136 #ifdef INET6
3137                         case AF_INET6:
3138                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3139                                     &pd->naddr, 0);
3140                                 rewrite++;
3141                                 break;
3142 #endif /* INET6 */
3143                         }
3144                         if (nr->natpass)
3145                                 r = NULL;
3146                         pd->nat_rule = nr;
3147                 }
3148         } else {
3149                 /* check incoming packet for BINAT/RDR */
3150                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3151                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3152                         PF_ACPY(&pd->baddr, daddr, af);
3153                         switch (af) {
3154 #ifdef INET
3155                         case AF_INET:
3156                                 pf_change_a(&daddr->v4.s_addr,
3157                                     pd->ip_sum, pd->naddr.v4.s_addr, 0);
3158                                 break;
3159 #endif /* INET */
3160 #ifdef INET6
3161                         case AF_INET6:
3162                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3163                                     &pd->naddr, 0);
3164                                 rewrite++;
3165                                 break;
3166 #endif /* INET6 */
3167                         }
3168                         if (nr->natpass)
3169                                 r = NULL;
3170                         pd->nat_rule = nr;
3171                 }
3172         }
3173
3174         while (r != NULL) {
3175                 r->evaluations++;
3176                 if (r->kif != NULL &&
3177                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3178                         r = r->skip[PF_SKIP_IFP].ptr;
3179                 else if (r->direction && r->direction != direction)
3180                         r = r->skip[PF_SKIP_DIR].ptr;
3181                 else if (r->af && r->af != af)
3182                         r = r->skip[PF_SKIP_AF].ptr;
3183                 else if (r->proto && r->proto != pd->proto)
3184                         r = r->skip[PF_SKIP_PROTO].ptr;
3185                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3186                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3187                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3188                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3189                 else if (r->type && r->type != icmptype + 1)
3190                         r = TAILQ_NEXT(r, entries);
3191                 else if (r->code && r->code != icmpcode + 1)
3192                         r = TAILQ_NEXT(r, entries);
3193                 else if (r->tos && !(r->tos & pd->tos))
3194                         r = TAILQ_NEXT(r, entries);
3195                 else if (r->rule_flag & PFRULE_FRAGMENT)
3196                         r = TAILQ_NEXT(r, entries);
3197                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3198                         r = TAILQ_NEXT(r, entries);
3199                 else if (r->anchorname[0] && r->anchor == NULL)
3200                         r = TAILQ_NEXT(r, entries);
3201                 else if (r->os_fingerprint != PF_OSFP_ANY)
3202                         r = TAILQ_NEXT(r, entries);
3203                 else {
3204                         if (r->tag)
3205                                 tag = r->tag;
3206                         if (r->anchor == NULL) {
3207                                 *rm = r;
3208                                 *am = a;
3209                                 *rsm = ruleset;
3210                                 if ((*rm)->quick)
3211                                         break;
3212                                 r = TAILQ_NEXT(r, entries);
3213                         } else
3214                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3215                                     PF_RULESET_FILTER);
3216                 }
3217                 if (r == NULL && a != NULL)
3218                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3219                             PF_RULESET_FILTER);
3220         }
3221         r = *rm;
3222         a = *am;
3223         ruleset = *rsm;
3224
3225         REASON_SET(&reason, PFRES_MATCH);
3226
3227         if (r->log) {
3228 #ifdef INET6
3229                 if (rewrite)
3230                         m_copyback(m, off, sizeof(struct icmp6_hdr),
3231                             (caddr_t)pd->hdr.icmp6);
3232 #endif /* INET6 */
3233                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3234         }
3235
3236         if (r->action != PF_PASS)
3237                 return (PF_DROP);
3238
3239         pf_tag_packet(m, tag);
3240
3241         if (!state_icmp && (r->keep_state || nr != NULL)) {
3242                 /* create new state */
3243                 struct pf_state *s = NULL;
3244                 struct pf_src_node *sn = NULL;
3245
3246                 /* check maximums */
3247                 if (r->max_states && (r->states >= r->max_states))
3248                         goto cleanup;
3249                 /* src node for flter rule */
3250                 if ((r->rule_flag & PFRULE_SRCTRACK ||
3251                     r->rpool.opts & PF_POOL_STICKYADDR) &&
3252                     pf_insert_src_node(&sn, r, saddr, af) != 0)
3253                         goto cleanup;
3254                 /* src node for translation rule */
3255                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3256                     ((direction == PF_OUT &&
3257                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3258                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3259                         goto cleanup;
3260                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3261                 if (s == NULL) {
3262 cleanup:
3263                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3264                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3265                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3266                                 pf_status.src_nodes--;
3267                                 pool_put(&pf_src_tree_pl, sn);
3268                         }
3269                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3270                             nsn->expire == 0) {
3271                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3272                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3273                                 pf_status.src_nodes--;
3274                                 pool_put(&pf_src_tree_pl, nsn);
3275                         }
3276                         REASON_SET(&reason, PFRES_MEMORY);
3277                         return (PF_DROP);
3278                 }
3279                 bzero(s, sizeof(*s));
3280                 r->states++;
3281                 if (a != NULL)
3282                         a->states++;
3283                 s->rule.ptr = r;
3284                 s->nat_rule.ptr = nr;
3285                 if (s->nat_rule.ptr != NULL)
3286                         s->nat_rule.ptr->states++;
3287                 s->anchor.ptr = a;
3288                 s->allow_opts = r->allow_opts;
3289                 s->log = r->log & 2;
3290                 s->proto = pd->proto;
3291                 s->direction = direction;
3292                 s->af = af;
3293                 if (direction == PF_OUT) {
3294                         PF_ACPY(&s->gwy.addr, saddr, af);
3295                         s->gwy.port = icmpid;
3296                         PF_ACPY(&s->ext.addr, daddr, af);
3297                         s->ext.port = icmpid;
3298                         if (nr != NULL)
3299                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3300                         else
3301                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3302                         s->lan.port = icmpid;
3303                 } else {
3304                         PF_ACPY(&s->lan.addr, daddr, af);
3305                         s->lan.port = icmpid;
3306                         PF_ACPY(&s->ext.addr, saddr, af);
3307                         s->ext.port = icmpid;
3308                         if (nr != NULL)
3309                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3310                         else
3311                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3312                         s->gwy.port = icmpid;
3313                 }
3314                 s->creation = time_second;
3315                 s->expire = time_second;
3316                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3317                 pf_set_rt_ifp(s, saddr);
3318                 if (sn != NULL) {
3319                         s->src_node = sn;
3320                         s->src_node->states++;
3321                 }
3322                 if (nsn != NULL) {
3323                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3324                         s->nat_src_node = nsn;
3325                         s->nat_src_node->states++;
3326                 }
3327                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3328                         REASON_SET(&reason, PFRES_MEMORY);
3329                         pf_src_tree_remove_state(s);
3330                         pool_put(&pf_state_pl, s);
3331                         return (PF_DROP);
3332                 } else
3333                         *sm = s;
3334         }
3335
3336 #ifdef INET6
3337         /* copy back packet headers if we performed IPv6 NAT operations */
3338         if (rewrite)
3339                 m_copyback(m, off, sizeof(struct icmp6_hdr),
3340                     (caddr_t)pd->hdr.icmp6);
3341 #endif /* INET6 */
3342
3343         return (PF_PASS);
3344 }
3345
3346 int
3347 pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3348     struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3349     struct pf_rule **am, struct pf_ruleset **rsm)
3350 {
3351         struct pf_rule          *nr = NULL;
3352         struct pf_rule          *r, *a = NULL;
3353         struct pf_ruleset       *ruleset = NULL;
3354         struct pf_src_node      *nsn = NULL;
3355         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3356         sa_family_t              af = pd->af;
3357         u_short                  reason;
3358         int                      tag = -1;
3359
3360         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3361
3362         if (direction == PF_OUT) {
3363                 /* check outgoing packet for BINAT/NAT */
3364                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3365                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3366                         PF_ACPY(&pd->baddr, saddr, af);
3367                         switch (af) {
3368 #ifdef INET
3369                         case AF_INET:
3370                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3371                                     pd->naddr.v4.s_addr, 0);
3372                                 break;
3373 #endif /* INET */
3374 #ifdef INET6
3375                         case AF_INET6:
3376                                 PF_ACPY(saddr, &pd->naddr, af);
3377                                 break;
3378 #endif /* INET6 */
3379                         }
3380                         if (nr->natpass)
3381                                 r = NULL;
3382                         pd->nat_rule = nr;
3383                 }
3384         } else {
3385                 /* check incoming packet for BINAT/RDR */
3386                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3387                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3388                         PF_ACPY(&pd->baddr, daddr, af);
3389                         switch (af) {
3390 #ifdef INET
3391                         case AF_INET:
3392                                 pf_change_a(&daddr->v4.s_addr,
3393                                     pd->ip_sum, pd->naddr.v4.s_addr, 0);
3394                                 break;
3395 #endif /* INET */
3396 #ifdef INET6
3397                         case AF_INET6:
3398                                 PF_ACPY(daddr, &pd->naddr, af);
3399                                 break;
3400 #endif /* INET6 */
3401                         }
3402                         if (nr->natpass)
3403                                 r = NULL;
3404                         pd->nat_rule = nr;
3405                 }
3406         }
3407
3408         while (r != NULL) {
3409                 r->evaluations++;
3410                 if (r->kif != NULL &&
3411                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3412                         r = r->skip[PF_SKIP_IFP].ptr;
3413                 else if (r->direction && r->direction != direction)
3414                         r = r->skip[PF_SKIP_DIR].ptr;
3415                 else if (r->af && r->af != af)
3416                         r = r->skip[PF_SKIP_AF].ptr;
3417                 else if (r->proto && r->proto != pd->proto)
3418                         r = r->skip[PF_SKIP_PROTO].ptr;
3419                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3420                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3421                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3422                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3423                 else if (r->tos && !(r->tos & pd->tos))
3424                         r = TAILQ_NEXT(r, entries);
3425                 else if (r->rule_flag & PFRULE_FRAGMENT)
3426                         r = TAILQ_NEXT(r, entries);
3427                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3428                         r = TAILQ_NEXT(r, entries);
3429                 else if (r->anchorname[0] && r->anchor == NULL)
3430                         r = TAILQ_NEXT(r, entries);
3431                 else if (r->os_fingerprint != PF_OSFP_ANY)
3432                         r = TAILQ_NEXT(r, entries);
3433                 else {
3434                         if (r->tag)
3435                                 tag = r->tag;
3436                         if (r->anchor == NULL) {
3437                                 *rm = r;
3438                                 *am = a;
3439                                 *rsm = ruleset;
3440                                 if ((*rm)->quick)
3441                                         break;
3442                                 r = TAILQ_NEXT(r, entries);
3443                         } else
3444                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3445                                     PF_RULESET_FILTER);
3446                 }
3447                 if (r == NULL && a != NULL)
3448                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3449                             PF_RULESET_FILTER);
3450         }
3451         r = *rm;
3452         a = *am;
3453         ruleset = *rsm;
3454
3455         REASON_SET(&reason, PFRES_MATCH);
3456
3457         if (r->log)
3458                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3459
3460         if ((r->action == PF_DROP) &&
3461             ((r->rule_flag & PFRULE_RETURNICMP) ||
3462             (r->rule_flag & PFRULE_RETURN))) {
3463                 struct pf_addr *a = NULL;
3464
3465                 if (nr != NULL) {
3466                         if (direction == PF_OUT)
3467                                 a = saddr;
3468                         else
3469                                 a = daddr;
3470                 }
3471                 if (a != NULL) {
3472                         switch (af) {
3473 #ifdef INET
3474                         case AF_INET:
3475                                 pf_change_a(&a->v4.s_addr, pd->ip_sum,
3476                                     pd->baddr.v4.s_addr, 0);
3477                                 break;
3478 #endif /* INET */
3479 #ifdef INET6
3480                         case AF_INET6:
3481                                 PF_ACPY(a, &pd->baddr, af);
3482                                 break;
3483 #endif /* INET6 */
3484                         }
3485                 }
3486                 if ((af == AF_INET) && r->return_icmp)
3487                         pf_send_icmp(m, r->return_icmp >> 8,
3488                             r->return_icmp & 255, af, r);
3489                 else if ((af == AF_INET6) && r->return_icmp6)
3490                         pf_send_icmp(m, r->return_icmp6 >> 8,
3491                             r->return_icmp6 & 255, af, r);
3492         }
3493
3494         if (r->action != PF_PASS)
3495                 return (PF_DROP);
3496
3497         pf_tag_packet(m, tag);
3498
3499         if (r->keep_state || nr != NULL) {
3500                 /* create new state */
3501                 struct pf_state *s = NULL;
3502                 struct pf_src_node *sn = NULL;
3503
3504                 /* check maximums */
3505                 if (r->max_states && (r->states >= r->max_states))
3506                         goto cleanup;
3507                 /* src node for flter rule */
3508                 if ((r->rule_flag & PFRULE_SRCTRACK ||
3509                     r->rpool.opts & PF_POOL_STICKYADDR) &&
3510                     pf_insert_src_node(&sn, r, saddr, af) != 0)
3511                         goto cleanup;
3512                 /* src node for translation rule */
3513                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3514                     ((direction == PF_OUT &&
3515                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3516                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3517                         goto cleanup;
3518                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3519                 if (s == NULL) {
3520 cleanup:
3521                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3522                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3523                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3524                                 pf_status.src_nodes--;
3525                                 pool_put(&pf_src_tree_pl, sn);
3526                         }
3527                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3528                             nsn->expire == 0) {
3529                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3530                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3531                                 pf_status.src_nodes--;
3532                                 pool_put(&pf_src_tree_pl, nsn);
3533                         }
3534                         REASON_SET(&reason, PFRES_MEMORY);
3535                         return (PF_DROP);
3536                 }
3537                 bzero(s, sizeof(*s));
3538                 r->states++;
3539                 if (a != NULL)
3540                         a->states++;
3541                 s->rule.ptr = r;
3542                 s->nat_rule.ptr = nr;
3543                 if (s->nat_rule.ptr != NULL)
3544                         s->nat_rule.ptr->states++;
3545                 s->anchor.ptr = a;
3546                 s->allow_opts = r->allow_opts;
3547                 s->log = r->log & 2;
3548                 s->proto = pd->proto;
3549                 s->direction = direction;
3550                 s->af = af;
3551                 if (direction == PF_OUT) {
3552                         PF_ACPY(&s->gwy.addr, saddr, af);
3553                         PF_ACPY(&s->ext.addr, daddr, af);
3554                         if (nr != NULL)
3555                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3556                         else
3557                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3558                 } else {
3559                         PF_ACPY(&s->lan.addr, daddr, af);
3560                         PF_ACPY(&s->ext.addr, saddr, af);
3561                         if (nr != NULL)
3562                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3563                         else
3564                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3565                 }
3566                 s->src.state = PFOTHERS_SINGLE;
3567                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3568                 s->creation = time_second;
3569                 s->expire = time_second;
3570                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3571                 pf_set_rt_ifp(s, saddr);
3572                 if (sn != NULL) {
3573                         s->src_node = sn;
3574                         s->src_node->states++;
3575                 }
3576                 if (nsn != NULL) {
3577                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3578                         s->nat_src_node = nsn;
3579                         s->nat_src_node->states++;
3580                 }
3581                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3582                         REASON_SET(&reason, PFRES_MEMORY);
3583                         pf_src_tree_remove_state(s);
3584                         pool_put(&pf_state_pl, s);
3585                         return (PF_DROP);
3586                 } else
3587                         *sm = s;
3588         }
3589
3590         return (PF_PASS);
3591 }
3592
3593 int
3594 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3595     struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3596     struct pf_ruleset **rsm)
3597 {
3598         struct pf_rule          *r, *a = NULL;
3599         struct pf_ruleset       *ruleset = NULL;
3600         sa_family_t              af = pd->af;
3601         u_short                  reason;
3602         int                      tag = -1;
3603
3604         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3605         while (r != NULL) {
3606                 r->evaluations++;
3607                 if (r->kif != NULL &&
3608                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3609                         r = r->skip[PF_SKIP_IFP].ptr;
3610                 else if (r->direction && r->direction != direction)
3611                         r = r->skip[PF_SKIP_DIR].ptr;
3612                 else if (r->af && r->af != af)
3613                         r = r->skip[PF_SKIP_AF].ptr;
3614                 else if (r->proto && r->proto != pd->proto)
3615                         r = r->skip[PF_SKIP_PROTO].ptr;
3616                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3617                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3618                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3619                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3620                 else if (r->tos && !(r->tos & pd->tos))
3621                         r = TAILQ_NEXT(r, entries);
3622                 else if (r->src.port_op || r->dst.port_op ||
3623                     r->flagset || r->type || r->code ||
3624                     r->os_fingerprint != PF_OSFP_ANY)
3625                         r = TAILQ_NEXT(r, entries);
3626                 else if (r->match_tag && !pf_match_tag(m, r, NULL, &tag))
3627                         r = TAILQ_NEXT(r, entries);
3628                 else if (r->anchorname[0] && r->anchor == NULL)
3629                         r = TAILQ_NEXT(r, entries);
3630                 else {
3631                         if (r->anchor == NULL) {
3632                                 *rm = r;
3633                                 *am = a;
3634                                 *rsm = ruleset;
3635                                 if ((*rm)->quick)
3636                                         break;
3637                                 r = TAILQ_NEXT(r, entries);
3638                         } else
3639                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3640                                     PF_RULESET_FILTER);
3641                 }
3642                 if (r == NULL && a != NULL)
3643                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3644                             PF_RULESET_FILTER);
3645         }
3646         r = *rm;
3647         a = *am;
3648         ruleset = *rsm;
3649
3650         REASON_SET(&reason, PFRES_MATCH);
3651
3652         if (r->log)
3653                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3654
3655         if (r->action != PF_PASS)
3656                 return (PF_DROP);
3657
3658         pf_tag_packet(m, tag);
3659
3660         return (PF_PASS);
3661 }
3662
3663 int
3664 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3665     struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3666     u_short *reason)
3667 {
3668         struct pf_state          key;
3669         struct tcphdr           *th = pd->hdr.tcp;
3670         u_int16_t                win = ntohs(th->th_win);
3671         u_int32_t                ack, end, seq;
3672         u_int8_t                 sws, dws;
3673         int                      ackskew;
3674         int                      copyback = 0;
3675         struct pf_state_peer    *src, *dst;
3676
3677         key.af = pd->af;
3678         key.proto = IPPROTO_TCP;
3679         if (direction == PF_IN) {
3680                 PF_ACPY(&key.ext.addr, pd->src, key.af);
3681                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3682                 key.ext.port = th->th_sport;
3683                 key.gwy.port = th->th_dport;
3684         } else {
3685                 PF_ACPY(&key.lan.addr, pd->src, key.af);
3686                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3687                 key.lan.port = th->th_sport;
3688                 key.ext.port = th->th_dport;
3689         }
3690
3691         STATE_LOOKUP();
3692
3693         if (direction == (*state)->direction) {
3694                 src = &(*state)->src;
3695                 dst = &(*state)->dst;
3696         } else {
3697                 src = &(*state)->dst;
3698                 dst = &(*state)->src;
3699         }
3700
3701         if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3702                 if (direction != (*state)->direction)
3703                         return (PF_SYNPROXY_DROP);
3704                 if (th->th_flags & TH_SYN) {
3705                         if (ntohl(th->th_seq) != (*state)->src.seqlo)
3706                                 return (PF_DROP);
3707                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3708                             pd->src, th->th_dport, th->th_sport,
3709                             (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3710                             TH_SYN|TH_ACK, 0, (*state)->src.mss, 0);
3711                         return (PF_SYNPROXY_DROP);
3712                 } else if (!(th->th_flags & TH_ACK) ||
3713                     (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3714                     (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3715                         return (PF_DROP);
3716                 else
3717                         (*state)->src.state = PF_TCPS_PROXY_DST;
3718         }
3719         if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3720                 struct pf_state_host *src, *dst;
3721
3722                 if (direction == PF_OUT) {
3723                         src = &(*state)->gwy;
3724                         dst = &(*state)->ext;
3725                 } else {
3726                         src = &(*state)->ext;
3727                         dst = &(*state)->lan;
3728                 }
3729                 if (direction == (*state)->direction) {
3730                         if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3731                             (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3732                             (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3733                                 return (PF_DROP);
3734                         (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3735                         if ((*state)->dst.seqhi == 1)
3736                                 (*state)->dst.seqhi = karc4random();
3737                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3738                             &dst->addr, src->port, dst->port,
3739                             (*state)->dst.seqhi, 0, TH_SYN, 0,
3740                             (*state)->src.mss, 0);
3741                         return (PF_SYNPROXY_DROP);
3742                 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3743                     (TH_SYN|TH_ACK)) ||
3744                     (ntohl(th->th_ack) != (*state)->dst.seqhi + 1))
3745                         return (PF_DROP);
3746                 else {
3747                         (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3748                         (*state)->dst.seqlo = ntohl(th->th_seq);
3749                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3750                             pd->src, th->th_dport, th->th_sport,
3751                             ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3752                             TH_ACK, (*state)->src.max_win, 0, 0);
3753                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3754                             &dst->addr, src->port, dst->port,
3755                             (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3756                             TH_ACK, (*state)->dst.max_win, 0, 0);
3757                         (*state)->src.seqdiff = (*state)->dst.seqhi -
3758                             (*state)->src.seqlo;
3759                         (*state)->dst.seqdiff = (*state)->src.seqhi -
3760                             (*state)->dst.seqlo;
3761                         (*state)->src.seqhi = (*state)->src.seqlo +
3762                             (*state)->src.max_win;
3763                         (*state)->dst.seqhi = (*state)->dst.seqlo +
3764                             (*state)->dst.max_win;
3765                         (*state)->src.wscale = (*state)->dst.wscale = 0;
3766                         (*state)->src.state = (*state)->dst.state =
3767                             TCPS_ESTABLISHED;
3768                         return (PF_SYNPROXY_DROP);
3769                 }
3770         }
3771
3772         if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3773                 sws = src->wscale & PF_WSCALE_MASK;
3774                 dws = dst->wscale & PF_WSCALE_MASK;
3775         } else
3776                 sws = dws = 0;
3777
3778         /*
3779          * Sequence tracking algorithm from Guido van Rooij's paper:
3780          *   http://www.madison-gurkha.com/publications/tcp_filtering/
3781          *      tcp_filtering.ps
3782          */
3783
3784         seq = ntohl(th->th_seq);
3785         if (src->seqlo == 0) {
3786                 /* First packet from this end. Set its state */
3787
3788                 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3789                     src->scrub == NULL) {
3790                         if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3791                                 REASON_SET(reason, PFRES_MEMORY);
3792                                 return (PF_DROP);
3793                         }
3794                 }
3795
3796                 /* Deferred generation of sequence number modulator */
3797                 if (dst->seqdiff && !src->seqdiff) {
3798                         while ((src->seqdiff = karc4random()) == 0)
3799                                 ;
3800                         ack = ntohl(th->th_ack) - dst->seqdiff;
3801                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3802                             src->seqdiff), 0);
3803                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3804                         copyback = 1;
3805                 } else {
3806                         ack = ntohl(th->th_ack);
3807                 }
3808
3809                 end = seq + pd->p_len;
3810                 if (th->th_flags & TH_SYN) {
3811                         end++;
3812                         if (dst->wscale & PF_WSCALE_FLAG) {
3813                                 src->wscale = pf_get_wscale(m, off, th->th_off,
3814                                     pd->af);
3815                                 if (src->wscale & PF_WSCALE_FLAG) {
3816                                         /* Remove scale factor from initial
3817                                          * window */
3818                                         sws = src->wscale & PF_WSCALE_MASK;
3819                                         win = ((u_int32_t)win + (1 << sws) - 1)
3820                                             >> sws;
3821                                         dws = dst->wscale & PF_WSCALE_MASK;
3822                                 } else {
3823                                         /* fixup other window */
3824                                         dst->max_win <<= dst->wscale &
3825                                             PF_WSCALE_MASK;
3826                                         /* in case of a retrans SYN|ACK */
3827                                         dst->wscale = 0;
3828                                 }
3829                         }
3830                 }
3831                 if (th->th_flags & TH_FIN)
3832                         end++;
3833
3834                 src->seqlo = seq;
3835                 if (src->state < TCPS_SYN_SENT)
3836                         src->state = TCPS_SYN_SENT;
3837
3838                 /*
3839                  * May need to slide the window (seqhi may have been set by
3840                  * the crappy stack check or if we picked up the connection
3841                  * after establishment)
3842                  */
3843                 if (src->seqhi == 1 ||
3844                     SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3845                         src->seqhi = end + MAX(1, dst->max_win << dws);
3846                 if (win > src->max_win)
3847                         src->max_win = win;
3848
3849         } else {
3850                 ack = ntohl(th->th_ack) - dst->seqdiff;
3851                 if (src->seqdiff) {
3852                         /* Modulate sequence numbers */
3853                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3854                             src->seqdiff), 0);
3855                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3856                         copyback = 1;
3857                 }
3858                 end = seq + pd->p_len;
3859                 if (th->th_flags & TH_SYN)
3860                         end++;
3861                 if (th->th_flags & TH_FIN)
3862                         end++;
3863         }
3864
3865         if ((th->th_flags & TH_ACK) == 0) {
3866                 /* Let it pass through the ack skew check */
3867                 ack = dst->seqlo;
3868         } else if ((ack == 0 &&
3869             (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3870             /* broken tcp stacks do not set ack */
3871             (dst->state < TCPS_SYN_SENT)) {
3872                 /*
3873                  * Many stacks (ours included) will set the ACK number in an
3874                  * FIN|ACK if the SYN times out -- no sequence to ACK.
3875                  */
3876                 ack = dst->seqlo;
3877         }
3878
3879         if (seq == end) {
3880                 /* Ease sequencing restrictions on no data packets */
3881                 seq = src->seqlo;
3882                 end = seq;
3883         }
3884
3885         ackskew = dst->seqlo - ack;
3886
3887 #define MAXACKWINDOW (0xffff + 1500)    /* 1500 is an arbitrary fudge factor */
3888         if (SEQ_GEQ(src->seqhi, end) &&
3889             /* Last octet inside other's window space */
3890             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3891             /* Retrans: not more than one window back */
3892             (ackskew >= -MAXACKWINDOW) &&
3893             /* Acking not more than one reassembled fragment backwards */
3894             (ackskew <= (MAXACKWINDOW << sws))) {
3895             /* Acking not more than one window forward */
3896
3897                 /* update max window */
3898                 if (src->max_win < win)
3899                         src->max_win = win;
3900                 /* synchronize sequencing */
3901                 if (SEQ_GT(end, src->seqlo))
3902                         src->seqlo = end;
3903                 /* slide the window of what the other end can send */
3904                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3905                         dst->seqhi = ack + MAX((win << sws), 1);
3906
3907
3908                 /* update states */
3909                 if (th->th_flags & TH_SYN)
3910                         if (src->state < TCPS_SYN_SENT)
3911                                 src->state = TCPS_SYN_SENT;
3912                 if (th->th_flags & TH_FIN)
3913                         if (src->state < TCPS_CLOSING)
3914                                 src->state = TCPS_CLOSING;
3915                 if (th->th_flags & TH_ACK) {
3916                         if (dst->state == TCPS_SYN_SENT)
3917                                 dst->state = TCPS_ESTABLISHED;
3918                         else if (dst->state == TCPS_CLOSING)
3919                                 dst->state = TCPS_FIN_WAIT_2;
3920                 }
3921                 if (th->th_flags & TH_RST)
3922                         src->state = dst->state = TCPS_TIME_WAIT;
3923
3924                 /* update expire time */
3925                 (*state)->expire = time_second;
3926                 if (src->state >= TCPS_FIN_WAIT_2 &&
3927                     dst->state >= TCPS_FIN_WAIT_2)
3928                         (*state)->timeout = PFTM_TCP_CLOSED;
3929                 else if (src->state >= TCPS_FIN_WAIT_2 ||
3930                     dst->state >= TCPS_FIN_WAIT_2)
3931                         (*state)->timeout = PFTM_TCP_FIN_WAIT;
3932                 else if (src->state < TCPS_ESTABLISHED ||
3933                     dst->state < TCPS_ESTABLISHED)
3934                         (*state)->timeout = PFTM_TCP_OPENING;
3935                 else if (src->state >= TCPS_CLOSING ||
3936                     dst->state >= TCPS_CLOSING)
3937                         (*state)->timeout = PFTM_TCP_CLOSING;
3938                 else
3939                         (*state)->timeout = PFTM_TCP_ESTABLISHED;
3940
3941                 /* Fall through to PASS packet */
3942
3943         } else if ((dst->state < TCPS_SYN_SENT ||
3944                 dst->state >= TCPS_FIN_WAIT_2 ||
3945                 src->state >= TCPS_FIN_WAIT_2) &&
3946             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
3947             /* Within a window forward of the originating packet */
3948             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
3949             /* Within a window backward of the originating packet */
3950
3951                 /*
3952                  * This currently handles three situations:
3953                  *  1) Stupid stacks will shotgun SYNs before their peer
3954                  *     replies.
3955                  *  2) When PF catches an already established stream (the
3956                  *     firewall rebooted, the state table was flushed, routes
3957                  *     changed...)
3958                  *  3) Packets get funky immediately after the connection
3959                  *     closes (this should catch Solaris spurious ACK|FINs
3960                  *     that web servers like to spew after a close)
3961                  *
3962                  * This must be a little more careful than the above code
3963                  * since packet floods will also be caught here. We don't
3964                  * update the TTL here to mitigate the damage of a packet
3965                  * flood and so the same code can handle awkward establishment
3966                  * and a loosened connection close.
3967                  * In the establishment case, a correct peer response will
3968                  * validate the connection, go through the normal state code
3969                  * and keep updating the state TTL.
3970                  */
3971
3972                 if (pf_status.debug >= PF_DEBUG_MISC) {
3973                         kprintf("pf: loose state match: ");
3974                         pf_print_state(*state);
3975                         pf_print_flags(th->th_flags);
3976                         kprintf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n",
3977                             seq, ack, pd->p_len, ackskew,
3978                             (*state)->packets[0], (*state)->packets[1]);
3979                 }
3980
3981                 /* update max window */
3982                 if (src->max_win < win)
3983                         src->max_win = win;
3984                 /* synchronize sequencing */
3985                 if (SEQ_GT(end, src->seqlo))
3986                         src->seqlo = end;
3987                 /* slide the window of what the other end can send */
3988                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3989                         dst->seqhi = ack + MAX((win << sws), 1);
3990
3991                 /*
3992                  * Cannot set dst->seqhi here since this could be a shotgunned
3993                  * SYN and not an already established connection.
3994                  */
3995
3996                 if (th->th_flags & TH_FIN)
3997                         if (src->state < TCPS_CLOSING)
3998                                 src->state = TCPS_CLOSING;
3999                 if (th->th_flags & TH_RST)
4000                         src->state = dst->state = TCPS_TIME_WAIT;
4001
4002                 /* Fall through to PASS packet */
4003
4004         } else {
4005                 if ((*state)->dst.state == TCPS_SYN_SENT &&
4006                     (*state)->src.state == TCPS_SYN_SENT) {
4007                         /* Send RST for state mismatches during handshake */
4008                         if (!(th->th_flags & TH_RST)) {
4009                                 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
4010
4011                                 if (th->th_flags & TH_SYN)
4012                                         ack++;
4013                                 if (th->th_flags & TH_FIN)
4014                                         ack++;
4015                                 pf_send_tcp((*state)->rule.ptr, pd->af,
4016                                     pd->dst, pd->src, th->th_dport,
4017                                     th->th_sport, ntohl(th->th_ack), ack,
4018                                     TH_RST|TH_ACK, 0, 0,
4019                                     (*state)->rule.ptr->return_ttl);
4020                         }
4021                         src->seqlo = 0;
4022                         src->seqhi = 1;
4023                         src->max_win = 1;
4024                 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4025                         kprintf("pf: BAD state: ");
4026                         pf_print_state(*state);
4027                         pf_print_flags(th->th_flags);
4028                         kprintf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d "
4029                             "dir=%s,%s\n", seq, ack, pd->p_len, ackskew,
4030                             (*state)->packets[0], (*state)->packets[1],
4031                             direction == PF_IN ? "in" : "out",
4032                             direction == (*state)->direction ? "fwd" : "rev");
4033                         kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4034                             SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4035                             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4036                             ' ': '2',
4037                             (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4038                             (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4039                             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4040                             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4041                 }
4042                 return (PF_DROP);
4043         }
4044
4045         if (dst->scrub || src->scrub) {
4046                 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4047                     src, dst, &copyback))
4048                         return (PF_DROP);
4049         }
4050
4051         /* Any packets which have gotten here are to be passed */
4052
4053         /* translate source/destination address, if necessary */
4054         if (STATE_TRANSLATE(*state)) {
4055                 if (direction == PF_OUT)
4056                         pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4057                             &th->th_sum, &(*state)->gwy.addr,
4058                             (*state)->gwy.port, 0, pd->af);
4059                 else
4060                         pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4061                             &th->th_sum, &(*state)->lan.addr,
4062                             (*state)->lan.port, 0, pd->af);
4063                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4064         } else if (copyback) {
4065                 /* Copyback sequence modulation or stateful scrub changes */
4066                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4067         }
4068
4069         return (PF_PASS);
4070 }
4071
4072 int
4073 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4074     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4075 {
4076         struct pf_state_peer    *src, *dst;
4077         struct pf_state          key;
4078         struct udphdr           *uh = pd->hdr.udp;
4079
4080         key.af = pd->af;
4081         key.proto = IPPROTO_UDP;
4082         if (direction == PF_IN) {
4083                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4084                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4085                 key.ext.port = uh->uh_sport;
4086                 key.gwy.port = uh->uh_dport;
4087         } else {
4088                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4089                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4090                 key.lan.port = uh->uh_sport;
4091                 key.ext.port = uh->uh_dport;
4092         }
4093
4094         STATE_LOOKUP();
4095
4096         if (direction == (*state)->direction) {
4097                 src = &(*state)->src;
4098                 dst = &(*state)->dst;
4099         } else {
4100                 src = &(*state)->dst;
4101                 dst = &(*state)->src;
4102         }
4103
4104         /* update states */
4105         if (src->state < PFUDPS_SINGLE)
4106                 src->state = PFUDPS_SINGLE;
4107         if (dst->state == PFUDPS_SINGLE)
4108                 dst->state = PFUDPS_MULTIPLE;
4109
4110         /* update expire time */
4111         (*state)->expire = time_second;
4112         if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4113                 (*state)->timeout = PFTM_UDP_MULTIPLE;
4114         else
4115                 (*state)->timeout = PFTM_UDP_SINGLE;
4116
4117         /* translate source/destination address, if necessary */
4118         if (STATE_TRANSLATE(*state)) {
4119                 if (direction == PF_OUT)
4120                         pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4121                             &uh->uh_sum, &(*state)->gwy.addr,
4122                             (*state)->gwy.port, 1, pd->af);
4123                 else
4124                         pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4125                             &uh->uh_sum, &(*state)->lan.addr,
4126                             (*state)->lan.port, 1, pd->af);
4127                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4128         }
4129
4130         return (PF_PASS);
4131 }
4132
4133 int
4134 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4135     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4136 {
4137         struct pf_addr  *saddr = pd->src, *daddr = pd->dst;
4138         u_int16_t        icmpid = 0;
4139         u_int16_t       *icmpsum = NULL;
4140         u_int8_t         icmptype = 0;
4141         int              state_icmp = 0;
4142
4143         switch (pd->proto) {
4144 #ifdef INET
4145         case IPPROTO_ICMP:
4146                 icmptype = pd->hdr.icmp->icmp_type;
4147                 icmpid = pd->hdr.icmp->icmp_id;
4148                 icmpsum = &pd->hdr.icmp->icmp_cksum;
4149
4150                 if (icmptype == ICMP_UNREACH ||
4151                     icmptype == ICMP_SOURCEQUENCH ||
4152                     icmptype == ICMP_REDIRECT ||
4153                     icmptype == ICMP_TIMXCEED ||
4154                     icmptype == ICMP_PARAMPROB)
4155                         state_icmp++;
4156                 break;
4157 #endif /* INET */
4158 #ifdef INET6
4159         case IPPROTO_ICMPV6:
4160                 icmptype = pd->hdr.icmp6->icmp6_type;
4161                 icmpid = pd->hdr.icmp6->icmp6_id;
4162                 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4163
4164                 if (icmptype == ICMP6_DST_UNREACH ||
4165                     icmptype == ICMP6_PACKET_TOO_BIG ||
4166                     icmptype == ICMP6_TIME_EXCEEDED ||
4167                     icmptype == ICMP6_PARAM_PROB)
4168                         state_icmp++;
4169                 break;
4170 #endif /* INET6 */
4171         }
4172
4173         if (!state_icmp) {
4174
4175                 /*
4176                  * ICMP query/reply message not related to a TCP/UDP packet.
4177                  * Search for an ICMP state.
4178                  */
4179                 struct pf_state         key;
4180
4181                 key.af = pd->af;
4182                 key.proto = pd->proto;
4183                 if (direction == PF_IN) {
4184                         PF_ACPY(&key.ext.addr, pd->src, key.af);
4185                         PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4186                         key.ext.port = icmpid;
4187                         key.gwy.port = icmpid;
4188                 } else {
4189                         PF_ACPY(&key.lan.addr, pd->src, key.af);
4190                         PF_ACPY(&key.ext.addr, pd->dst, key.af);
4191                         key.lan.port = icmpid;
4192                         key.ext.port = icmpid;
4193                 }
4194
4195                 STATE_LOOKUP();
4196
4197                 (*state)->expire = time_second;
4198                 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4199
4200                 /* translate source/destination address, if necessary */
4201                 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) {
4202                         if (direction == PF_OUT) {
4203                                 switch (pd->af) {
4204 #ifdef INET
4205                                 case AF_INET:
4206                                         pf_change_a(&saddr->v4.s_addr,
4207                                             pd->ip_sum,
4208                                             (*state)->gwy.addr.v4.s_addr, 0);
4209                                         break;
4210 #endif /* INET */
4211 #ifdef INET6
4212                                 case AF_INET6:
4213                                         pf_change_a6(saddr,
4214                                             &pd->hdr.icmp6->icmp6_cksum,
4215                                             &(*state)->gwy.addr, 0);
4216                                         m_copyback(m, off,
4217                                             sizeof(struct icmp6_hdr),
4218                                             (caddr_t)pd->hdr.icmp6);
4219                                         break;
4220 #endif /* INET6 */
4221                                 }
4222                         } else {
4223                                 switch (pd->af) {
4224 #ifdef INET
4225                                 case AF_INET:
4226                                         pf_change_a(&daddr->v4.s_addr,
4227                                             pd->ip_sum,
4228                                             (*state)->lan.addr.v4.s_addr, 0);
4229                                         break;
4230 #endif /* INET */
4231 #ifdef INET6
4232                                 case AF_INET6:
4233                                         pf_change_a6(daddr,
4234                                             &pd->hdr.icmp6->icmp6_cksum,
4235                                             &(*state)->lan.addr, 0);
4236                                         m_copyback(m, off,
4237                                             sizeof(struct icmp6_hdr),
4238                                             (caddr_t)pd->hdr.icmp6);
4239                                         break;
4240 #endif /* INET6 */
4241                                 }
4242                         }
4243                 }
4244
4245                 return (PF_PASS);
4246
4247         } else {
4248                 /*
4249                  * ICMP error message in response to a TCP/UDP packet.
4250                  * Extract the inner TCP/UDP header and search for that state.
4251                  */
4252
4253                 struct pf_pdesc pd2;
4254 #ifdef INET
4255                 struct ip       h2;
4256 #endif /* INET */
4257 #ifdef INET6
4258                 struct ip6_hdr  h2_6;
4259                 int             terminal = 0;
4260 #endif /* INET6 */
4261                 int             ipoff2 = 0;
4262                 int             off2 = 0;
4263
4264                 pd2.af = pd->af;
4265                 switch (pd->af) {
4266 #ifdef INET
4267                 case AF_INET:
4268                         /* offset of h2 in mbuf chain */
4269                         ipoff2 = off + ICMP_MINLEN;
4270
4271                         if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4272                             NULL, NULL, pd2.af)) {
4273                                 DPFPRINTF(PF_DEBUG_MISC,
4274                                     ("pf: ICMP error message too short "
4275                                     "(ip)\n"));
4276                                 return (PF_DROP);
4277                         }
4278                         /*
4279                          * ICMP error messages don't refer to non-first
4280                          * fragments
4281                          */
4282                         /*
4283                          * Note: We are dealing with an encapsulated
4284                          * header. This means ip_off/ip_len are not
4285                          * in host byte order!
4286                          */
4287                         if (h2.ip_off & htons(IP_OFFMASK))
4288                                 return (PF_DROP);
4289
4290                         /* offset of protocol header that follows h2 */
4291                         off2 = ipoff2 + (h2.ip_hl << 2);
4292
4293                         pd2.proto = h2.ip_p;
4294                         pd2.src = (struct pf_addr *)&h2.ip_src;
4295                         pd2.dst = (struct pf_addr *)&h2.ip_dst;
4296                         pd2.ip_sum = &h2.ip_sum;
4297                         break;
4298 #endif /* INET */
4299 #ifdef INET6
4300                 case AF_INET6:
4301                         ipoff2 = off + sizeof(struct icmp6_hdr);
4302
4303                         if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4304                             NULL, NULL, pd2.af)) {
4305                                 DPFPRINTF(PF_DEBUG_MISC,
4306                                     ("pf: ICMP error message too short "
4307                                     "(ip6)\n"));
4308                                 return (PF_DROP);
4309                         }
4310                         pd2.proto = h2_6.ip6_nxt;
4311                         pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4312                         pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4313                         pd2.ip_sum = NULL;
4314                         off2 = ipoff2 + sizeof(h2_6);
4315                         do {
4316                                 switch (pd2.proto) {
4317                                 case IPPROTO_FRAGMENT:
4318                                         /*
4319                                          * ICMPv6 error messages for
4320                                          * non-first fragments
4321                                          */
4322                                         return (PF_DROP);
4323                                 case IPPROTO_AH:
4324                                 case IPPROTO_HOPOPTS:
4325                                 case IPPROTO_ROUTING:
4326                                 case IPPROTO_DSTOPTS: {
4327                                         /* get next header and header length */
4328                                         struct ip6_ext opt6;
4329
4330                                         if (!pf_pull_hdr(m, off2, &opt6,
4331                                             sizeof(opt6), NULL, NULL, pd2.af)) {
4332                                                 DPFPRINTF(PF_DEBUG_MISC,
4333                                                     ("pf: ICMPv6 short opt\n"));
4334                                                 return (PF_DROP);
4335                                         }
4336                                         if (pd2.proto == IPPROTO_AH)
4337                                                 off2 += (opt6.ip6e_len + 2) * 4;
4338                                         else
4339                                                 off2 += (opt6.ip6e_len + 1) * 8;
4340                                         pd2.proto = opt6.ip6e_nxt;
4341                                         /* goto the next header */
4342                                         break;
4343                                 }
4344                                 default:
4345                                         terminal++;
4346                                         break;
4347                                 }
4348                         } while (!terminal);
4349                         break;
4350 #endif /* INET6 */
4351                 }
4352
4353                 switch (pd2.proto) {
4354                 case IPPROTO_TCP: {
4355                         struct tcphdr            th;
4356                         u_int32_t                seq;
4357                         struct pf_state          key;
4358                         struct pf_state_peer    *src, *dst;
4359                         u_int8_t                 dws;
4360                         int                      copyback = 0;
4361
4362                         /*
4363                          * Only the first 8 bytes of the TCP header can be
4364                          * expected. Don't access any TCP header fields after
4365                          * th_seq, an ackskew test is not possible.
4366                          */
4367                         if (!pf_pull_hdr(m, off2, &th, 8, NULL, NULL, pd2.af)) {
4368                                 DPFPRINTF(PF_DEBUG_MISC,
4369                                     ("pf: ICMP error message too short "
4370                                     "(tcp)\n"));
4371                                 return (PF_DROP);
4372                         }
4373
4374                         key.af = pd2.af;
4375                         key.proto = IPPROTO_TCP;
4376                         if (direction == PF_IN) {
4377                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4378                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4379                                 key.ext.port = th.th_dport;
4380                                 key.gwy.port = th.th_sport;
4381                         } else {
4382                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4383                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4384                                 key.lan.port = th.th_dport;
4385                                 key.ext.port = th.th_sport;
4386                         }
4387
4388                         STATE_LOOKUP();
4389
4390                         if (direction == (*state)->direction) {
4391                                 src = &(*state)->dst;
4392                                 dst = &(*state)->src;
4393                         } else {
4394                                 src = &(*state)->src;
4395                                 dst = &(*state)->dst;
4396                         }
4397
4398                         if (src->wscale && dst->wscale &&
4399                             !(th.th_flags & TH_SYN))
4400                                 dws = dst->wscale & PF_WSCALE_MASK;
4401                         else
4402                                 dws = 0;
4403
4404                         /* Demodulate sequence number */
4405                         seq = ntohl(th.th_seq) - src->seqdiff;
4406                         if (src->seqdiff) {
4407                                 pf_change_a(&th.th_seq, icmpsum,
4408                                     htonl(seq), 0);
4409                                 copyback = 1;
4410                         }
4411
4412                         if (!SEQ_GEQ(src->seqhi, seq) ||
4413                             !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4414                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4415                                         kprintf("pf: BAD ICMP %d:%d ",
4416                                             icmptype, pd->hdr.icmp->icmp_code);
4417                                         pf_print_host(pd->src, 0, pd->af);
4418                                         kprintf(" -> ");
4419                                         pf_print_host(pd->dst, 0, pd->af);
4420                                         kprintf(" state: ");
4421                                         pf_print_state(*state);
4422                                         kprintf(" seq=%u\n", seq);
4423                                 }
4424                                 return (PF_DROP);
4425                         }
4426
4427                         if (STATE_TRANSLATE(*state)) {
4428                                 if (direction == PF_IN) {
4429                                         pf_change_icmp(pd2.src, &th.th_sport,
4430                                             daddr, &(*state)->lan.addr,
4431                                             (*state)->lan.port, NULL,
4432                                             pd2.ip_sum, icmpsum,
4433                                             pd->ip_sum, 0, pd2.af);
4434                                 } else {
4435                                         pf_change_icmp(pd2.dst, &th.th_dport,
4436                                             saddr, &(*state)->gwy.addr,
4437                                             (*state)->gwy.port, NULL,
4438                                             pd2.ip_sum, icmpsum,
4439                                             pd->ip_sum, 0, pd2.af);
4440                                 }
4441                                 copyback = 1;
4442                         }
4443
4444                         if (copyback) {
4445                                 switch (pd2.af) {
4446 #ifdef INET
4447                                 case AF_INET:
4448                                         m_copyback(m, off, ICMP_MINLEN,
4449                                             (caddr_t)pd->hdr.icmp);
4450                                         m_copyback(m, ipoff2, sizeof(h2),
4451                                             (caddr_t)&h2);
4452                                         break;
4453 #endif /* INET */
4454 #ifdef INET6
4455                                 case AF_INET6:
4456                                         m_copyback(m, off,
4457                                             sizeof(struct icmp6_hdr),
4458                                             (caddr_t)pd->hdr.icmp6);
4459                                         m_copyback(m, ipoff2, sizeof(h2_6),
4460                                             (caddr_t)&h2_6);
4461                                         break;
4462 #endif /* INET6 */
4463                                 }
4464                                 m_copyback(m, off2, 8, (caddr_t)&th);
4465                         }
4466
4467                         return (PF_PASS);
4468                         break;
4469                 }
4470                 case IPPROTO_UDP: {
4471                         struct udphdr           uh;
4472                         struct pf_state         key;
4473
4474                         if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4475                             NULL, NULL, pd2.af)) {
4476                                 DPFPRINTF(PF_DEBUG_MISC,
4477                                     ("pf: ICMP error message too short "
4478                                     "(udp)\n"));
4479                                 return (PF_DROP);
4480                         }
4481
4482                         key.af = pd2.af;
4483                         key.proto = IPPROTO_UDP;
4484                         if (direction == PF_IN) {
4485                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4486                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4487                                 key.ext.port = uh.uh_dport;
4488                                 key.gwy.port = uh.uh_sport;
4489                         } else {
4490                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4491                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4492                                 key.lan.port = uh.uh_dport;
4493                                 key.ext.port = uh.uh_sport;
4494                         }
4495
4496                         STATE_LOOKUP();
4497
4498                         if (STATE_TRANSLATE(*state)) {
4499                                 if (direction == PF_IN) {
4500                                         pf_change_icmp(pd2.src, &uh.uh_sport,
4501                                             daddr, &(*state)->lan.addr,
4502                                             (*state)->lan.port, &uh.uh_sum,
4503                                             pd2.ip_sum, icmpsum,
4504                                             pd->ip_sum, 1, pd2.af);
4505                                 } else {
4506                                         pf_change_icmp(pd2.dst, &uh.uh_dport,
4507                                             saddr, &(*state)->gwy.addr,
4508                                             (*state)->gwy.port, &uh.uh_sum,
4509                                             pd2.ip_sum, icmpsum,
4510                                             pd->ip_sum, 1, pd2.af);
4511                                 }
4512                                 switch (pd2.af) {
4513 #ifdef INET
4514                                 case AF_INET:
4515                                         m_copyback(m, off, ICMP_MINLEN,
4516                                             (caddr_t)pd->hdr.icmp);
4517                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4518                                         break;
4519 #endif /* INET */
4520 #ifdef INET6
4521                                 case AF_INET6:
4522                                         m_copyback(m, off,
4523                                             sizeof(struct icmp6_hdr),
4524                                             (caddr_t)pd->hdr.icmp6);
4525                                         m_copyback(m, ipoff2, sizeof(h2_6),
4526                                             (caddr_t)&h2_6);
4527                                         break;
4528 #endif /* INET6 */
4529                                 }
4530                                 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4531                         }
4532
4533                         return (PF_PASS);
4534                         break;
4535                 }
4536 #ifdef INET
4537                 case IPPROTO_ICMP: {
4538                         struct icmp             iih;
4539                         struct pf_state         key;
4540
4541                         if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4542                             NULL, NULL, pd2.af)) {
4543                                 DPFPRINTF(PF_DEBUG_MISC,
4544                                     ("pf: ICMP error message too short i"
4545                                     "(icmp)\n"));
4546                                 return (PF_DROP);
4547                         }
4548
4549                         key.af = pd2.af;
4550                         key.proto = IPPROTO_ICMP;
4551                         if (direction == PF_IN) {
4552                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4553                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4554                                 key.ext.port = iih.icmp_id;
4555                                 key.gwy.port = iih.icmp_id;
4556                         } else {
4557                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4558                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4559                                 key.lan.port = iih.icmp_id;
4560                                 key.ext.port = iih.icmp_id;
4561                         }
4562
4563                         STATE_LOOKUP();
4564
4565                         if (STATE_TRANSLATE(*state)) {
4566                                 if (direction == PF_IN) {
4567                                         pf_change_icmp(pd2.src, &iih.icmp_id,
4568                                             daddr, &(*state)->lan.addr,
4569                                             (*state)->lan.port, NULL,
4570                                             pd2.ip_sum, icmpsum,
4571                                             pd->ip_sum, 0, AF_INET);
4572                                 } else {
4573                                         pf_change_icmp(pd2.dst, &iih.icmp_id,
4574                                             saddr, &(*state)->gwy.addr,
4575                                             (*state)->gwy.port, NULL,
4576                                             pd2.ip_sum, icmpsum,
4577                                             pd->ip_sum, 0, AF_INET);
4578                                 }
4579                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4580                                 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4581                                 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4582                         }
4583
4584                         return (PF_PASS);
4585                         break;
4586                 }
4587 #endif /* INET */
4588 #ifdef INET6
4589                 case IPPROTO_ICMPV6: {
4590                         struct icmp6_hdr        iih;
4591                         struct pf_state         key;
4592
4593                         if (!pf_pull_hdr(m, off2, &iih,
4594                             sizeof(struct icmp6_hdr), NULL, NULL, pd2.af)) {
4595                                 DPFPRINTF(PF_DEBUG_MISC,
4596                                     ("pf: ICMP error message too short "
4597                                     "(icmp6)\n"));
4598                                 return (PF_DROP);
4599                         }
4600
4601                         key.af = pd2.af;
4602                         key.proto = IPPROTO_ICMPV6;
4603                         if (direction == PF_IN) {
4604                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4605                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4606                                 key.ext.port = iih.icmp6_id;
4607                                 key.gwy.port = iih.icmp6_id;
4608                         } else {
4609                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4610                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4611                                 key.lan.port = iih.icmp6_id;
4612                                 key.ext.port = iih.icmp6_id;
4613                         }
4614
4615                         STATE_LOOKUP();
4616
4617                         if (STATE_TRANSLATE(*state)) {
4618                                 if (direction == PF_IN) {
4619                                         pf_change_icmp(pd2.src, &iih.icmp6_id,
4620                                             daddr, &(*state)->lan.addr,
4621                                             (*state)->lan.port, NULL,
4622                                             pd2.ip_sum, icmpsum,
4623                                             pd->ip_sum, 0, AF_INET6);
4624                                 } else {
4625                                         pf_change_icmp(pd2.dst, &iih.icmp6_id,
4626                                             saddr, &(*state)->gwy.addr,
4627                                             (*state)->gwy.port, NULL,
4628                                             pd2.ip_sum, icmpsum,
4629                                             pd->ip_sum, 0, AF_INET6);
4630                                 }
4631                                 m_copyback(m, off, sizeof(struct icmp6_hdr),
4632                                     (caddr_t)pd->hdr.icmp6);
4633                                 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
4634                                 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4635                                     (caddr_t)&iih);
4636                         }
4637
4638                         return (PF_PASS);
4639                         break;
4640                 }
4641 #endif /* INET6 */
4642                 default: {
4643                         struct pf_state         key;
4644
4645                         key.af = pd2.af;
4646                         key.proto = pd2.proto;
4647                         if (direction == PF_IN) {
4648                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4649                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4650                                 key.ext.port = 0;
4651                                 key.gwy.port = 0;
4652                         } else {
4653                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4654                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4655                                 key.lan.port = 0;
4656                                 key.ext.port = 0;
4657                         }
4658
4659                         STATE_LOOKUP();
4660
4661                         if (STATE_TRANSLATE(*state)) {
4662                                 if (direction == PF_IN) {
4663                                         pf_change_icmp(pd2.src, NULL,
4664                                             daddr, &(*state)->lan.addr,
4665                                             0, NULL,
4666                                             pd2.ip_sum, icmpsum,
4667                                             pd->ip_sum, 0, pd2.af);
4668                                 } else {
4669                                         pf_change_icmp(pd2.dst, NULL,
4670                                             saddr, &(*state)->gwy.addr,
4671                                             0, NULL,
4672                                             pd2.ip_sum, icmpsum,
4673                                             pd->ip_sum, 0, pd2.af);
4674                                 }
4675                                 switch (pd2.af) {
4676 #ifdef INET
4677                                 case AF_INET:
4678                                         m_copyback(m, off, ICMP_MINLEN,
4679                                             (caddr_t)pd->hdr.icmp);
4680                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4681                                         break;
4682 #endif /* INET */
4683 #ifdef INET6
4684                                 case AF_INET6:
4685                                         m_copyback(m, off,
4686                                             sizeof(struct icmp6_hdr),
4687                                             (caddr_t)pd->hdr.icmp6);
4688                                         m_copyback(m, ipoff2, sizeof(h2_6),
4689                                             (caddr_t)&h2_6);
4690                                         break;
4691 #endif /* INET6 */
4692                                 }
4693                         }
4694
4695                         return (PF_PASS);
4696                         break;
4697                 }
4698                 }
4699         }
4700 }
4701
4702 int
4703 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4704     struct pf_pdesc *pd)
4705 {
4706         struct pf_state_peer    *src, *dst;
4707         struct pf_state          key;
4708
4709         key.af = pd->af;
4710         key.proto = pd->proto;
4711         if (direction == PF_IN) {
4712                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4713                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4714                 key.ext.port = 0;
4715                 key.gwy.port = 0;
4716         } else {
4717                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4718                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4719                 key.lan.port = 0;
4720                 key.ext.port = 0;
4721         }
4722
4723         STATE_LOOKUP();
4724
4725         if (direction == (*state)->direction) {
4726                 src = &(*state)->src;
4727                 dst = &(*state)->dst;
4728         } else {
4729                 src = &(*state)->dst;
4730                 dst = &(*state)->src;
4731         }
4732
4733         /* update states */
4734         if (src->state < PFOTHERS_SINGLE)
4735                 src->state = PFOTHERS_SINGLE;
4736         if (dst->state == PFOTHERS_SINGLE)
4737                 dst->state = PFOTHERS_MULTIPLE;
4738
4739         /* update expire time */
4740         (*state)->expire = time_second;
4741         if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4742                 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4743         else
4744                 (*state)->timeout = PFTM_OTHER_SINGLE;
4745
4746         /* translate source/destination address, if necessary */
4747         if (STATE_TRANSLATE(*state)) {
4748                 if (direction == PF_OUT)
4749                         switch (pd->af) {
4750 #ifdef INET
4751                         case AF_INET:
4752                                 pf_change_a(&pd->src->v4.s_addr,
4753                                     pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
4754                                     0);
4755                                 break;
4756 #endif /* INET */
4757 #ifdef INET6
4758                         case AF_INET6:
4759                                 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
4760                                 break;
4761 #endif /* INET6 */
4762                         }
4763                 else
4764                         switch (pd->af) {
4765 #ifdef INET
4766                         case AF_INET:
4767                                 pf_change_a(&pd->dst->v4.s_addr,
4768                                     pd->ip_sum, (*state)->lan.addr.v4.s_addr,
4769                                     0);
4770                                 break;
4771 #endif /* INET */
4772 #ifdef INET6
4773                         case AF_INET6:
4774                                 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
4775                                 break;
4776 #endif /* INET6 */
4777                         }
4778         }
4779
4780         return (PF_PASS);
4781 }
4782
4783 /*
4784  * ipoff and off are measured from the start of the mbuf chain.
4785  * h must be at "ipoff" on the mbuf chain.
4786  */
4787 void *
4788 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4789     u_short *actionp, u_short *reasonp, sa_family_t af)
4790 {
4791         switch (af) {
4792 #ifdef INET
4793         case AF_INET: {
4794                 struct ip       *h = mtod(m, struct ip *);
4795                 u_int16_t        fragoff = (h->ip_off & IP_OFFMASK) << 3;
4796
4797                 if (fragoff) {
4798                         if (fragoff >= len)
4799                                 ACTION_SET(actionp, PF_PASS);
4800                         else {
4801                                 ACTION_SET(actionp, PF_DROP);
4802                                 REASON_SET(reasonp, PFRES_FRAG);
4803                         }
4804                         return (NULL);
4805                 }
4806                 if (m->m_pkthdr.len < off + len ||
4807                     h->ip_len < off + len) {
4808                         ACTION_SET(actionp, PF_DROP);
4809                         REASON_SET(reasonp, PFRES_SHORT);
4810                         return (NULL);
4811                 }
4812                 break;
4813         }
4814 #endif /* INET */
4815 #ifdef INET6
4816         case AF_INET6: {
4817                 struct ip6_hdr  *h = mtod(m, struct ip6_hdr *);
4818
4819                 if (m->m_pkthdr.len < off + len ||
4820                     (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
4821                     (unsigned)(off + len)) {
4822                         ACTION_SET(actionp, PF_DROP);
4823                         REASON_SET(reasonp, PFRES_SHORT);
4824                         return (NULL);
4825                 }
4826                 break;
4827         }
4828 #endif /* INET6 */
4829         }
4830         m_copydata(m, off, len, p);
4831         return (p);
4832 }
4833
4834 int
4835 pf_routable(struct pf_addr *addr, sa_family_t af)
4836 {
4837         struct sockaddr_in      *dst;
4838         struct route             ro;
4839         int                      ret = 0;
4840
4841         bzero(&ro, sizeof(ro));
4842         dst = satosin(&ro.ro_dst);
4843         dst->sin_family = af;
4844         dst->sin_len = sizeof(*dst);
4845         dst->sin_addr = addr->v4;
4846         rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
4847
4848         if (ro.ro_rt != NULL) {
4849                 ret = 1;
4850                 RTFREE(ro.ro_rt);
4851         }
4852
4853         return (ret);
4854 }
4855
4856 #ifdef INET
4857 void
4858 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
4859     struct pf_state *s)
4860 {
4861         struct mbuf             *m0, *m1;
4862         struct route             iproute;
4863         struct route            *ro = NULL;
4864         struct sockaddr_in      *dst;
4865         struct ip               *ip;
4866         struct ifnet            *ifp = NULL;
4867         struct pf_addr           naddr;
4868         struct pf_src_node      *sn = NULL;
4869         int                      error = 0;
4870         int sw_csum;
4871
4872         if (m == NULL || *m == NULL || r == NULL ||
4873             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
4874                 panic("pf_route: invalid parameters");
4875
4876         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
4877                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
4878                 (*m)->m_pkthdr.pf_routed = 1;
4879         } else {
4880                 if ((*m)->m_pkthdr.pf_routed > 3) {
4881                         m0 = *m;
4882                         *m = NULL;
4883                         goto bad;
4884                 }
4885                 (*m)->m_pkthdr.pf_routed++;
4886         }
4887
4888         if (r->rt == PF_DUPTO) {
4889                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
4890                         return;
4891         } else {
4892                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
4893                         return;
4894                 m0 = *m;
4895         }
4896
4897         if (m0->m_len < sizeof(struct ip))
4898                 panic("pf_route: m0->m_len < sizeof(struct ip)");
4899         ip = mtod(m0, struct ip *);
4900
4901         ro = &iproute;
4902         bzero((caddr_t)ro, sizeof(*ro));
4903         dst = satosin(&ro->ro_dst);
4904         dst->sin_family = AF_INET;
4905         dst->sin_len = sizeof(*dst);
4906         dst->sin_addr = ip->ip_dst;
4907
4908         if (r->rt == PF_FASTROUTE) {
4909                 rtalloc(ro);
4910                 if (ro->ro_rt == 0) {
4911                         ipstat.ips_noroute++;
4912                         goto bad;
4913                 }
4914
4915                 ifp = ro->ro_rt->rt_ifp;
4916                 ro->ro_rt->rt_use++;
4917
4918                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
4919                         dst = satosin(ro->ro_rt->rt_gateway);
4920         } else {
4921                 if (TAILQ_EMPTY(&r->rpool.list))
4922                         panic("pf_route: TAILQ_EMPTY(&r->rpool.list)");
4923                 if (s == NULL) {
4924                         pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
4925                             &naddr, NULL, &sn);
4926                         if (!PF_AZERO(&naddr, AF_INET))
4927                                 dst->sin_addr.s_addr = naddr.v4.s_addr;
4928                         ifp = r->rpool.cur->kif ?
4929                             r->rpool.cur->kif->pfik_ifp : NULL;
4930                 } else {
4931                         if (!PF_AZERO(&s->rt_addr, AF_INET))
4932                                 dst->sin_addr.s_addr =
4933                                     s->rt_addr.v4.s_addr;
4934                         ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
4935                 }
4936         }
4937         if (ifp == NULL)
4938                 goto bad;
4939
4940         if (oifp != ifp) {
4941                 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS)
4942                         goto bad;
4943                 else if (m0 == NULL)
4944                         goto done;
4945                 if (m0->m_len < sizeof(struct ip))
4946                         panic("pf_route: m0->m_len < sizeof(struct ip)");
4947                 ip = mtod(m0, struct ip *);
4948         }
4949
4950         /* Copied from ip_output. */
4951         m0->m_pkthdr.csum_flags |= CSUM_IP;
4952         sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
4953         if (sw_csum & CSUM_DELAY_DATA) {
4954                 in_delayed_cksum(m0);
4955                 sw_csum &= ~CSUM_DELAY_DATA;
4956         }
4957         m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
4958
4959         /*
4960          * If small enough for interface, or the interface will take
4961          * care of the fragmentation for us, can just send directly.
4962          */
4963         if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
4964                 (ip->ip_off & IP_DF) == 0)) {
4965                 ip->ip_len = htons(ip->ip_len);
4966                 ip->ip_off = htons(ip->ip_off);
4967                 ip->ip_sum = 0;
4968                 if (sw_csum & CSUM_DELAY_IP) {
4969                         /* From KAME */
4970                         if (ip->ip_v == IPVERSION &&
4971                             (ip->ip_hl << 2) == sizeof(*ip)) {
4972                                 ip->ip_sum = in_cksum_hdr(ip);
4973                         } else {
4974                                 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
4975                         }
4976                 }
4977
4978                 lwkt_serialize_enter(ifp->if_serializer);
4979                 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt);
4980                 lwkt_serialize_exit(ifp->if_serializer);
4981                 goto done;
4982         }
4983
4984         /*
4985          * Too large for interface; fragment if possible.
4986          * Must be able to put at least 8 bytes per fragment.
4987          */
4988         if (ip->ip_off & IP_DF) {
4989                 ipstat.ips_cantfrag++;
4990                 if (r->rt != PF_DUPTO) {
4991                         icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
4992                             ifp->if_mtu);
4993                         goto done;
4994                 } else
4995                         goto bad;
4996         }
4997
4998         m1 = m0;
4999         error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
5000         if (error)
5001                 goto bad;
5002
5003         for (m0 = m1; m0; m0 = m1) {
5004                 m1 = m0->m_nextpkt;
5005                 m0->m_nextpkt = 0;
5006                 if (error == 0) {
5007                         lwkt_serialize_enter(ifp->if_serializer);
5008                         error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5009                                     NULL);
5010                         lwkt_serialize_exit(ifp->if_serializer);
5011                 } else {
5012                         m_freem(m0);
5013                 }
5014         }
5015
5016         if (error == 0)
5017                 ipstat.ips_fragmented++;
5018
5019 done:
5020         if (r->rt != PF_DUPTO)
5021                 *m = NULL;
5022         if (ro == &iproute && ro->ro_rt)
5023                 RTFREE(ro->ro_rt);
5024         return;
5025
5026 bad:
5027         m_freem(m0);
5028         goto done;
5029 }
5030 #endif /* INET */
5031
5032 #ifdef INET6
5033 void
5034 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5035     struct pf_state *s)
5036 {
5037         struct mbuf             *m0;
5038         struct route_in6         ip6route;
5039         struct route_in6        *ro;
5040         struct sockaddr_in6     *dst;
5041         struct ip6_hdr          *ip6;
5042         struct ifnet            *ifp = NULL;
5043         struct pf_addr           naddr;
5044         struct pf_src_node      *sn = NULL;
5045         int                      error = 0;
5046
5047         if (m == NULL || *m == NULL || r == NULL ||
5048             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5049                 panic("pf_route6: invalid parameters");
5050
5051         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5052                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5053                 (*m)->m_pkthdr.pf_routed = 1;
5054         } else {
5055                 if ((*m)->m_pkthdr.pf_routed > 3) {
5056                         m0 = *m;
5057                         *m = NULL;
5058                         goto bad;
5059                 }
5060                 (*m)->m_pkthdr.pf_routed++;
5061         }
5062
5063         if (r->rt == PF_DUPTO) {
5064                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5065                         return;
5066         } else {
5067                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5068                         return;
5069                 m0 = *m;
5070         }
5071
5072         if (m0->m_len < sizeof(struct ip6_hdr))
5073                 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5074         ip6 = mtod(m0, struct ip6_hdr *);
5075
5076         ro = &ip6route;
5077         bzero((caddr_t)ro, sizeof(*ro));
5078         dst = (struct sockaddr_in6 *)&ro->ro_dst;
5079         dst->sin6_family = AF_INET6;
5080         dst->sin6_len = sizeof(*dst);
5081         dst->sin6_addr = ip6->ip6_dst;
5082
5083         /* Cheat. */
5084         if (r->rt == PF_FASTROUTE) {
5085                 m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
5086                 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5087                 return;
5088         }
5089
5090         if (TAILQ_EMPTY(&r->rpool.list))
5091                 panic("pf_route6: TAILQ_EMPTY(&r->rpool.list)");
5092         if (s == NULL) {
5093                 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5094                     &naddr, NULL, &sn);
5095                 if (!PF_AZERO(&naddr, AF_INET6))
5096                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5097                             &naddr, AF_INET6);
5098                 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5099         } else {
5100                 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5101                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5102                             &s->rt_addr, AF_INET6);
5103                 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5104         }
5105         if (ifp == NULL)
5106                 goto bad;
5107
5108         if (oifp != ifp) {
5109                 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS)
5110                         goto bad;
5111                 else if (m0 == NULL)
5112                         goto done;
5113                 if (m0->m_len < sizeof(struct ip6_hdr))
5114                         panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5115                 ip6 = mtod(m0, struct ip6_hdr *);
5116         }
5117
5118         /*
5119          * If the packet is too large for the outgoing interface,
5120          * send back an icmp6 error.
5121          */
5122         if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5123                 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5124         if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5125                 error = nd6_output(ifp, ifp, m0, dst, NULL);
5126         } else {
5127                 in6_ifstat_inc(ifp, ifs6_in_toobig);
5128                 if (r->rt != PF_DUPTO)
5129                         icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5130                 else
5131                         goto bad;
5132         }
5133
5134 done:
5135         if (r->rt != PF_DUPTO)
5136                 *m = NULL;
5137         return;
5138
5139 bad:
5140         m_freem(m0);
5141         goto done;
5142 }
5143 #endif /* INET6 */
5144
5145
5146 /*
5147  * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5148  *   off is the offset where the protocol header starts
5149  *   len is the total length of protocol header plus payload
5150  * returns 0 when the checksum is valid, otherwise returns 1.
5151  */
5152 /*
5153  * XXX
5154  * FreeBSD supports cksum offload for the following drivers.
5155  * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5156  * If we can make full use of it we would outperform ipfw/ipfilter in
5157  * very heavy traffic.
5158  * I have not tested 'cause I don't have NICs that supports cksum offload.
5159  * (There might be problems. Typical phenomena would be
5160  *   1. No route message for UDP packet.
5161  *   2. No connection acceptance from external hosts regardless of rule set.)
5162  */
5163 int
5164 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5165     sa_family_t af)
5166 {
5167         u_int16_t sum = 0;
5168         int hw_assist = 0;
5169         struct ip *ip;
5170
5171         if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5172                 return (1);
5173         if (m->m_pkthdr.len < off + len)
5174                 return (1);
5175
5176         switch (p) {
5177         case IPPROTO_TCP:
5178         case IPPROTO_UDP:
5179                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5180                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5181                                 sum = m->m_pkthdr.csum_data;
5182                         } else {
5183                                 ip = mtod(m, struct ip *);
5184                                 sum = in_pseudo(ip->ip_src.s_addr,
5185                                         ip->ip_dst.s_addr, htonl((u_short)len +
5186                                         m->m_pkthdr.csum_data + p));
5187                         }
5188                         sum ^= 0xffff;
5189                         ++hw_assist;
5190                 }
5191                 break;
5192         case IPPROTO_ICMP:
5193 #ifdef INET6
5194         case IPPROTO_ICMPV6:
5195 #endif /* INET6 */
5196                 break;
5197         default:
5198                 return (1);
5199         }
5200
5201         if (!hw_assist) {
5202                 switch (af) {
5203                 case AF_INET:
5204                         if (p == IPPROTO_ICMP) {
5205                                 if (m->m_len < off)
5206                                         return (1);
5207                                 m->m_data += off;
5208                                 m->m_len -= off;
5209                                 sum = in_cksum(m, len);
5210                                 m->m_data -= off;
5211                                 m->m_len += off;
5212                         } else {
5213                                 if (m->m_len < sizeof(struct ip))
5214                                         return (1);
5215                                 sum = in_cksum_range(m, p, off, len);
5216                                 if (sum == 0) {
5217                                         m->m_pkthdr.csum_flags |=
5218                                             (CSUM_DATA_VALID |
5219                                              CSUM_PSEUDO_HDR);
5220                                         m->m_pkthdr.csum_data = 0xffff;
5221                                 }
5222                         }
5223                         break;
5224 #ifdef INET6
5225                 case AF_INET6:
5226                         if (m->m_len < sizeof(struct ip6_hdr))
5227                                 return (1);
5228                         sum = in6_cksum(m, p, off, len);
5229                         /*
5230                          * XXX
5231                          * IPv6 H/W cksum off-load not supported yet!
5232                          *
5233                          * if (sum == 0) {
5234                          *      m->m_pkthdr.csum_flags |=
5235                          *          (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5236                          *      m->m_pkthdr.csum_data = 0xffff;
5237                          *}
5238                          */
5239                         break;
5240 #endif /* INET6 */
5241                 default:
5242                         return (1);
5243                 }
5244         }
5245         if (sum) {
5246                 switch (p) {
5247                 case IPPROTO_TCP:
5248                         tcpstat.tcps_rcvbadsum++;
5249                         break;
5250                 case IPPROTO_UDP:
5251                         udpstat.udps_badsum++;
5252                         break;
5253                 case IPPROTO_ICMP:
5254                         icmpstat.icps_checksum++;
5255                         break;
5256 #ifdef INET6
5257                 case IPPROTO_ICMPV6:
5258                         icmp6stat.icp6s_checksum++;
5259                         break;
5260 #endif /* INET6 */
5261                 }
5262                 return (1);
5263         }
5264         return (0);
5265 }
5266
5267 #ifdef INET
5268 int
5269 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0)
5270 {
5271         struct pfi_kif          *kif;
5272         u_short                  action, reason = 0, log = 0;
5273         struct mbuf             *m = *m0;
5274         struct ip               *h = NULL;
5275         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5276         struct pf_state         *s = NULL;
5277         struct pf_ruleset       *ruleset = NULL;
5278         struct pf_pdesc          pd;
5279         int                      off, dirndx, pqid = 0;
5280
5281         if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5282                 return (PF_PASS);
5283
5284         kif = pfi_index2kif[ifp->if_index];
5285         if (kif == NULL)
5286                 return (PF_DROP);
5287
5288 #ifdef DIAGNOSTIC
5289         if ((m->m_flags & M_PKTHDR) == 0)
5290                 panic("non-M_PKTHDR is passed to pf_test");
5291 #endif
5292
5293         memset(&pd, 0, sizeof(pd));
5294         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5295                 action = PF_DROP;
5296                 REASON_SET(&reason, PFRES_SHORT);
5297                 log = 1;
5298                 goto done;
5299         }
5300
5301         /* We do IP header normalization and packet reassembly here */
5302         if (pf_normalize_ip(m0, dir, kif, &reason) != PF_PASS) {
5303                 action = PF_DROP;
5304                 goto done;
5305         }
5306         m = *m0;
5307         h = mtod(m, struct ip *);
5308
5309         off = h->ip_hl << 2;
5310         if (off < (int)sizeof(*h)) {
5311                 action = PF_DROP;
5312                 REASON_SET(&reason, PFRES_SHORT);
5313                 log = 1;
5314                 goto done;
5315         }
5316
5317         pd.src = (struct pf_addr *)&h->ip_src;
5318         pd.dst = (struct pf_addr *)&h->ip_dst;
5319         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5320         pd.ip_sum = &h->ip_sum;
5321         pd.proto = h->ip_p;
5322         pd.af = AF_INET;
5323         pd.tos = h->ip_tos;
5324         pd.tot_len = h->ip_len;
5325
5326         /* handle fragments that didn't get reassembled by normalization */
5327         if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5328                 action = pf_test_fragment(&r, dir, kif, m, h,
5329                     &pd, &a, &ruleset);
5330                 goto done;
5331         }
5332
5333         switch (h->ip_p) {
5334
5335         case IPPROTO_TCP: {
5336                 struct tcphdr   th;
5337
5338                 pd.hdr.tcp = &th;
5339                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5340                     &action, &reason, AF_INET)) {
5341                         log = action != PF_PASS;
5342                         goto done;
5343                 }
5344                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5345                     h->ip_len - off, IPPROTO_TCP, AF_INET)) {
5346                         action = PF_DROP;
5347                         goto done;
5348                 }
5349                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5350                 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5351                         pqid = 1;
5352                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5353                 if (action == PF_DROP)
5354                         goto done;
5355                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5356                     &reason);
5357                 if (action == PF_PASS) {
5358 #if NPFSYNC
5359                         pfsync_update_state(s);
5360 #endif
5361                         r = s->rule.ptr;
5362                         a = s->anchor.ptr;
5363                         log = s->log;
5364                 } else if (s == NULL)
5365                         action = pf_test_tcp(&r, &s, dir, kif,
5366                             m, off, h, &pd, &a, &ruleset);
5367                 break;
5368         }
5369
5370         case IPPROTO_UDP: {
5371                 struct udphdr   uh;
5372
5373                 pd.hdr.udp = &uh;
5374                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5375                     &action, &reason, AF_INET)) {
5376                         log = action != PF_PASS;
5377                         goto done;
5378                 }
5379                 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5380                     off, h->ip_len - off, IPPROTO_UDP, AF_INET)) {
5381                         action = PF_DROP;
5382                         goto done;
5383                 }
5384                 if (uh.uh_dport == 0 ||
5385                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5386                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5387                         action = PF_DROP;
5388                         goto done;
5389                 }
5390                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5391                 if (action == PF_PASS) {
5392 #if NPFSYNC
5393                         pfsync_update_state(s);
5394 #endif
5395                         r = s->rule.ptr;
5396                         a = s->anchor.ptr;
5397                         log = s->log;
5398                 } else if (s == NULL)
5399                         action = pf_test_udp(&r, &s, dir, kif,
5400                             m, off, h, &pd, &a, &ruleset);
5401                 break;
5402         }
5403
5404         case IPPROTO_ICMP: {
5405                 struct icmp     ih;
5406
5407                 pd.hdr.icmp = &ih;
5408                 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5409                     &action, &reason, AF_INET)) {
5410                         log = action != PF_PASS;
5411                         goto done;
5412                 }
5413                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5414                     h->ip_len - off, IPPROTO_ICMP, AF_INET)) {
5415                         action = PF_DROP;
5416                         goto done;
5417                 }
5418                 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd);
5419                 if (action == PF_PASS) {
5420 #if NPFSYNC
5421                         pfsync_update_state(s);
5422 #endif
5423                         r = s->rule.ptr;
5424                         a = s->anchor.ptr;
5425                         log = s->log;
5426                 } else if (s == NULL)
5427                         action = pf_test_icmp(&r, &s, dir, kif,
5428                             m, off, h, &pd, &a, &ruleset);
5429                 break;
5430         }
5431
5432         default:
5433                 action = pf_test_state_other(&s, dir, kif, &pd);
5434                 if (action == PF_PASS) {
5435 #if NPFSYNC
5436                         pfsync_update_state(s);
5437 #endif
5438                         r = s->rule.ptr;
5439                         a = s->anchor.ptr;
5440                         log = s->log;
5441                 } else if (s == NULL)
5442                         action = pf_test_other(&r, &s, dir, kif, m, off, h,
5443                             &pd, &a, &ruleset);
5444                 break;
5445         }
5446
5447 done:
5448         if (action == PF_PASS && h->ip_hl > 5 &&
5449             !((s && s->allow_opts) || r->allow_opts)) {
5450                 action = PF_DROP;
5451                 REASON_SET(&reason, PFRES_SHORT);
5452                 log = 1;
5453                 DPFPRINTF(PF_DEBUG_MISC,
5454                     ("pf: dropping packet with ip options\n"));
5455         }
5456
5457 #ifdef ALTQ
5458         if (action == PF_PASS && r->qid) {
5459                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5460                 if (pd.tos == IPTOS_LOWDELAY)
5461                         m->m_pkthdr.altq_qid = r->pqid;
5462                 else
5463                         m->m_pkthdr.altq_qid = r->qid;
5464                 m->m_pkthdr.ecn_af = AF_INET;
5465                 m->m_pkthdr.header = h;
5466         }
5467 #endif
5468
5469         /*
5470          * connections redirected to loopback should not match sockets
5471          * bound specifically to loopback due to security implications,
5472          * see tcp_input() and in_pcblookup_listen().
5473          */
5474         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5475             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5476             (s->nat_rule.ptr->action == PF_RDR ||
5477             s->nat_rule.ptr->action == PF_BINAT) &&
5478             (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
5479                 action = PF_DROP;
5480                 REASON_SET(&reason, PFRES_MEMORY);
5481         }
5482
5483         m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5484
5485         if (log)
5486                 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, r, a, ruleset);
5487
5488         kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5489         kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5490
5491         if (action == PF_PASS || r->action == PF_DROP) {
5492                 r->packets++;
5493                 r->bytes += pd.tot_len;
5494                 if (a != NULL) {
5495                         a->packets++;
5496                         a->bytes += pd.tot_len;
5497                 }
5498                 if (s != NULL) {
5499                         dirndx = (dir == s->direction) ? 0 : 1;
5500                         s->packets[dirndx]++;
5501                         s->bytes[dirndx] += pd.tot_len;
5502                         if (s->nat_rule.ptr != NULL) {
5503                                 s->nat_rule.ptr->packets++;
5504                                 s->nat_rule.ptr->bytes += pd.tot_len;
5505                         }
5506                         if (s->src_node != NULL) {
5507                                 s->src_node->packets++;
5508                                 s->src_node->bytes += pd.tot_len;
5509                         }
5510                         if (s->nat_src_node != NULL) {
5511                                 s->nat_src_node->packets++;
5512                                 s->nat_src_node->bytes += pd.tot_len;
5513                         }
5514                 }
5515                 tr = r;
5516                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5517                 if (nr != NULL) {
5518                         struct pf_addr *x;
5519                         /*
5520                          * XXX: we need to make sure that the addresses
5521                          * passed to pfr_update_stats() are the same than
5522                          * the addresses used during matching (pfr_match)
5523                          */
5524                         if (r == &pf_default_rule) {
5525                                 tr = nr;
5526                                 x = (s == NULL || s->direction == dir) ?
5527                                     &pd.baddr : &pd.naddr;
5528                         } else
5529                                 x = (s == NULL || s->direction == dir) ?
5530                                     &pd.naddr : &pd.baddr;
5531                         if (x == &pd.baddr || s == NULL) {
5532                                 /* we need to change the address */
5533                                 if (dir == PF_OUT)
5534                                         pd.src = x;
5535                                 else
5536                                         pd.dst = x;
5537                         }
5538                 }
5539                 if (tr->src.addr.type == PF_ADDR_TABLE)
5540                         pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5541                             s->direction == dir) ? pd.src : pd.dst, pd.af,
5542                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5543                             tr->src.not);
5544                 if (tr->dst.addr.type == PF_ADDR_TABLE)
5545                         pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5546                             s->direction == dir) ? pd.dst : pd.src, pd.af,
5547                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5548                             tr->dst.not);
5549         }
5550
5551
5552         if (action == PF_SYNPROXY_DROP) {
5553                 m_freem(*m0);
5554                 *m0 = NULL;
5555                 action = PF_PASS;
5556         } else if (r->rt)
5557                 /* pf_route can free the mbuf causing *m0 to become NULL */
5558                 pf_route(m0, r, dir, ifp, s);
5559
5560         return (action);
5561 }
5562 #endif /* INET */
5563
5564 #ifdef INET6
5565 int
5566 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0)
5567 {
5568         struct pfi_kif          *kif;
5569         u_short                  action, reason = 0, log = 0;
5570         struct mbuf             *m = *m0;
5571         struct ip6_hdr          *h = NULL;
5572         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5573         struct pf_state         *s = NULL;
5574         struct pf_ruleset       *ruleset = NULL;
5575         struct pf_pdesc          pd;
5576         int                      off, terminal = 0, dirndx;
5577
5578         if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5579                 return (PF_PASS);
5580
5581         kif = pfi_index2kif[ifp->if_index];
5582         if (kif == NULL)
5583                 return (PF_DROP);
5584
5585 #ifdef DIAGNOSTIC
5586         if ((m->m_flags & M_PKTHDR) == 0)
5587                 panic("non-M_PKTHDR is passed to pf_test");
5588 #endif
5589
5590         memset(&pd, 0, sizeof(pd));
5591         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5592                 action = PF_DROP;
5593                 REASON_SET(&reason, PFRES_SHORT);
5594                 log = 1;
5595                 goto done;
5596         }
5597
5598         /* We do IP header normalization and packet reassembly here */
5599         if (pf_normalize_ip6(m0, dir, kif, &reason) != PF_PASS) {
5600                 action = PF_DROP;
5601                 goto done;
5602         }
5603         m = *m0;
5604         h = mtod(m, struct ip6_hdr *);
5605
5606         pd.src = (struct pf_addr *)&h->ip6_src;
5607         pd.dst = (struct pf_addr *)&h->ip6_dst;
5608         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
5609         pd.ip_sum = NULL;
5610         pd.af = AF_INET6;
5611         pd.tos = 0;
5612         pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5613
5614         off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5615         pd.proto = h->ip6_nxt;
5616         do {
5617                 switch (pd.proto) {
5618                 case IPPROTO_FRAGMENT:
5619                         action = pf_test_fragment(&r, dir, kif, m, h,
5620                             &pd, &a, &ruleset);
5621                         if (action == PF_DROP)
5622                                 REASON_SET(&reason, PFRES_FRAG);
5623                         goto done;
5624                 case IPPROTO_AH:
5625                 case IPPROTO_HOPOPTS:
5626                 case IPPROTO_ROUTING:
5627                 case IPPROTO_DSTOPTS: {
5628                         /* get next header and header length */
5629                         struct ip6_ext  opt6;
5630
5631                         if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
5632                             NULL, NULL, pd.af)) {
5633                                 DPFPRINTF(PF_DEBUG_MISC,
5634                                     ("pf: IPv6 short opt\n"));
5635                                 action = PF_DROP;
5636                                 REASON_SET(&reason, PFRES_SHORT);
5637                                 log = 1;
5638                                 goto done;
5639                         }
5640                         if (pd.proto == IPPROTO_AH)
5641                                 off += (opt6.ip6e_len + 2) * 4;
5642                         else
5643                                 off += (opt6.ip6e_len + 1) * 8;
5644                         pd.proto = opt6.ip6e_nxt;
5645                         /* goto the next header */
5646                         break;
5647                 }
5648                 default:
5649                         terminal++;
5650                         break;
5651                 }
5652         } while (!terminal);
5653
5654         switch (pd.proto) {
5655
5656         case IPPROTO_TCP: {
5657                 struct tcphdr   th;
5658
5659                 pd.hdr.tcp = &th;
5660                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5661                     &action, &reason, AF_INET6)) {
5662                         log = action != PF_PASS;
5663                         goto done;
5664                 }
5665                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5666                     ntohs(h->ip6_plen), IPPROTO_TCP, AF_INET6)) {
5667                         action = PF_DROP;
5668                         goto done;
5669                 }
5670                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5671                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5672                 if (action == PF_DROP)
5673                         goto done;
5674                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5675                     &reason);
5676                 if (action == PF_PASS) {
5677 #if NPFSYNC
5678                         pfsync_update_state(s);
5679 #endif
5680                         r = s->rule.ptr;
5681                         a = s->anchor.ptr;
5682                         log = s->log;
5683                 } else if (s == NULL)
5684                         action = pf_test_tcp(&r, &s, dir, kif,
5685                             m, off, h, &pd, &a, &ruleset);
5686                 break;
5687         }
5688
5689         case IPPROTO_UDP: {
5690                 struct udphdr   uh;
5691
5692                 pd.hdr.udp = &uh;
5693                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5694                     &action, &reason, AF_INET6)) {
5695                         log = action != PF_PASS;
5696                         goto done;
5697                 }
5698                 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5699                     off, ntohs(h->ip6_plen), IPPROTO_UDP, AF_INET6)) {
5700                         action = PF_DROP;
5701                         goto done;
5702                 }
5703                 if (uh.uh_dport == 0 ||
5704                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5705                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5706                         action = PF_DROP;
5707                         goto done;
5708                 }
5709                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5710                 if (action == PF_PASS) {
5711 #if NPFSYNC
5712                         pfsync_update_state(s);
5713 #endif
5714                         r = s->rule.ptr;
5715                         a = s->anchor.ptr;
5716                         log = s->log;
5717                 } else if (s == NULL)
5718                         action = pf_test_udp(&r, &s, dir, kif,
5719                             m, off, h, &pd, &a, &ruleset);
5720                 break;
5721         }
5722
5723         case IPPROTO_ICMPV6: {
5724                 struct icmp6_hdr        ih;
5725
5726                 pd.hdr.icmp6 = &ih;
5727                 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
5728                     &action, &reason, AF_INET6)) {
5729                         log = action != PF_PASS;
5730                         goto done;
5731                 }
5732                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5733                     ntohs(h->ip6_plen), IPPROTO_ICMPV6, AF_INET6)) {
5734                         action = PF_DROP;
5735                         goto done;
5736                 }
5737                 action = pf_test_state_icmp(&s, dir, kif,
5738                     m, off, h, &pd);
5739                 if (action == PF_PASS) {
5740 #if NPFSYNC
5741                         pfsync_update_state(s);
5742 #endif
5743                         r = s->rule.ptr;
5744                         a = s->anchor.ptr;
5745                         log = s->log;
5746                 } else if (s == NULL)
5747                         action = pf_test_icmp(&r, &s, dir, kif,
5748                             m, off, h, &pd, &a, &ruleset);
5749                 break;
5750         }
5751
5752         default:
5753                 action = pf_test_state_other(&s, dir, kif, &pd);
5754                 if (action == PF_PASS) {
5755                         r = s->rule.ptr;
5756                         a = s->anchor.ptr;
5757                         log = s->log;
5758                 } else if (s == NULL)
5759                         action = pf_test_other(&r, &s, dir, kif, m, off, h,
5760                             &pd, &a, &ruleset);
5761                 break;
5762         }
5763
5764 done:
5765         /* XXX handle IPv6 options, if not allowed. not implemented. */
5766
5767 #ifdef ALTQ
5768         if (action == PF_PASS && r->qid) {
5769                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5770                 if (pd.tos == IPTOS_LOWDELAY)
5771                         m->m_pkthdr.altq_qid = r->pqid;
5772                 else
5773                         m->m_pkthdr.altq_qid = r->qid;
5774                 m->m_pkthdr.ecn_af = AF_INET6;
5775                 m->m_pkthdr.header = h;
5776         }
5777 #endif
5778
5779         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5780             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5781             (s->nat_rule.ptr->action == PF_RDR ||
5782             s->nat_rule.ptr->action == PF_BINAT) &&
5783             IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) {
5784                 action = PF_DROP;
5785                 REASON_SET(&reason, PFRES_MEMORY);
5786         }
5787
5788         m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5789
5790         if (log)
5791                 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, r, a, ruleset);
5792
5793         kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5794         kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
5795
5796         if (action == PF_PASS || r->action == PF_DROP) {
5797                 r->packets++;
5798                 r->bytes += pd.tot_len;
5799                 if (a != NULL) {
5800                         a->packets++;
5801                         a->bytes += pd.tot_len;
5802                 }
5803                 if (s != NULL) {
5804                         dirndx = (dir == s->direction) ? 0 : 1;
5805                         s->packets[dirndx]++;
5806                         s->bytes[dirndx] += pd.tot_len;
5807                         if (s->nat_rule.ptr != NULL) {
5808                                 s->nat_rule.ptr->packets++;
5809                                 s->nat_rule.ptr->bytes += pd.tot_len;
5810                         }
5811                         if (s->src_node != NULL) {
5812                                 s->src_node->packets++;
5813                                 s->src_node->bytes += pd.tot_len;
5814                         }
5815                         if (s->nat_src_node != NULL) {
5816                                 s->nat_src_node->packets++;
5817                                 s->nat_src_node->bytes += pd.tot_len;
5818                         }
5819                 }
5820                 tr = r;
5821                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5822                 if (nr != NULL) {
5823                         struct pf_addr *x;
5824                         /*
5825                          * XXX: we need to make sure that the addresses
5826                          * passed to pfr_update_stats() are the same than
5827                          * the addresses used during matching (pfr_match)
5828                          */
5829                         if (r == &pf_default_rule) {
5830                                 tr = nr;
5831                                 x = (s == NULL || s->direction == dir) ?
5832                                     &pd.baddr : &pd.naddr;
5833                         } else {
5834                                 x = (s == NULL || s->direction == dir) ?
5835                                     &pd.naddr : &pd.baddr;
5836                         }
5837                         if (x == &pd.baddr || s == NULL) {
5838                                 if (dir == PF_OUT)
5839                                         pd.src = x;
5840                                 else
5841                                         pd.dst = x;
5842                         }
5843                 }
5844                 if (tr->src.addr.type == PF_ADDR_TABLE)
5845                         pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5846                             s->direction == dir) ? pd.src : pd.dst, pd.af,
5847                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5848                             tr->src.not);
5849                 if (tr->dst.addr.type == PF_ADDR_TABLE)
5850                         pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5851                             s->direction == dir) ? pd.dst : pd.src, pd.af,
5852                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5853                             tr->dst.not);
5854         }
5855
5856
5857         if (action == PF_SYNPROXY_DROP) {
5858                 m_freem(*m0);
5859                 *m0 = NULL;
5860                 action = PF_PASS;
5861         } else if (r->rt)
5862                 /* pf_route6 can free the mbuf causing *m0 to become NULL */
5863                 pf_route6(m0, r, dir, ifp, s);
5864
5865         return (action);
5866 }
5867 #endif /* INET6 */