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 1.13 2006/12/22 23:44:57 swildner 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/route.h>
  66
  67 #include <netinet/in.h>
  68 #include <netinet/in_var.h>
  69 #include <netinet/in_systm.h>
  70 #include <netinet/ip.h>
  71 #include <netinet/ip_var.h>
  72 #include <netinet/tcp.h>
  73 #include <netinet/tcp_seq.h>
  74 #include <netinet/udp.h>
  75 #include <netinet/ip_icmp.h>
  76 #include <netinet/in_pcb.h>
  77 #include <netinet/tcp_timer.h>
  78 #include <netinet/tcp_var.h>
  79 #include <netinet/udp_var.h>
  80 #include <netinet/icmp_var.h>
  81
  82 #include <net/pf/pfvar.h>
  83 #include <net/pf/if_pflog.h>
  84
  85 #if NPFSYNC > 0
  86 #include <net/pf/if_pfsync.h>
  87 #endif /* NPFSYNC > 0 */
  88
  89 #ifdef INET6
  90 #include <netinet/ip6.h>
  91 #include <netinet/in_pcb.h>
  92 #include <netinet/icmp6.h>
  93 #include <netinet6/nd6.h>
  94 #include <netinet6/ip6_var.h>
  95 #include <netinet6/in6_pcb.h>
  96 #endif /* INET6 */
  97
  98 #include <sys/in_cksum.h>
  99 #include <machine/limits.h>
 100 #include <sys/msgport2.h>
 101 #include <sys/ucred.h>
 102
 103 extern int ip_optcopy(struct ip *, struct ip *);
 104
 105 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
 106
 107 /*
 108  * Global variables
 109  */
 110
 111 struct pf_anchorqueue    pf_anchors;
 112 struct pf_ruleset        pf_main_ruleset;
 113 struct pf_altqqueue      pf_altqs[2];
 114 struct pf_palist         pf_pabuf;
 115 struct pf_altqqueue     *pf_altqs_active;
 116 struct pf_altqqueue     *pf_altqs_inactive;
 117 struct pf_status         pf_status;
 118
 119 u_int32_t                ticket_altqs_active;
 120 u_int32_t                ticket_altqs_inactive;
 121 int                      altqs_inactive_open;
 122 u_int32_t                ticket_pabuf;
 123
 124 struct callout           pf_expire_to;                  /* expire timeout */
 125
 126 vm_zone_t                pf_src_tree_pl, pf_rule_pl;
 127 vm_zone_t                pf_state_pl, pf_altq_pl, pf_pooladdr_pl;
 128
 129 void                     pf_print_host(struct pf_addr *, u_int16_t, u_int8_t);
 130 void                     pf_print_state(struct pf_state *);
 131 void                     pf_print_flags(u_int8_t);
 132
 133 u_int16_t                pf_cksum_fixup(u_int16_t, u_int16_t, u_int16_t,
 134                             u_int8_t);
 135 void                     pf_change_ap(struct pf_addr *, u_int16_t *,
 136                             u_int16_t *, u_int16_t *, struct pf_addr *,
 137                             u_int16_t, u_int8_t, sa_family_t);
 138 #ifdef INET6
 139 void                     pf_change_a6(struct pf_addr *, u_int16_t *,
 140                             struct pf_addr *, u_int8_t);
 141 #endif /* INET6 */
 142 void                     pf_change_icmp(struct pf_addr *, u_int16_t *,
 143                             struct pf_addr *, struct pf_addr *, u_int16_t,
 144                             u_int16_t *, u_int16_t *, u_int16_t *,
 145                             u_int16_t *, u_int8_t, sa_family_t);
 146 void                     pf_send_tcp(const struct pf_rule *, sa_family_t,
 147                             const struct pf_addr *, const struct pf_addr *,
 148                             u_int16_t, u_int16_t, u_int32_t, u_int32_t,
 149                             u_int8_t, u_int16_t, u_int16_t, u_int8_t);
 150 void                     pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t,
 151                             sa_family_t, struct pf_rule *);
 152 struct pf_rule          *pf_match_translation(struct pf_pdesc *, struct mbuf *,
 153                             int, int, struct pfi_kif *,
 154                             struct pf_addr *, u_int16_t, struct pf_addr *,
 155                             u_int16_t, int);
 156 struct pf_rule          *pf_get_translation(struct pf_pdesc *, struct mbuf *,
 157                             int, int, struct pfi_kif *, struct pf_src_node **,
 158                             struct pf_addr *, u_int16_t,
 159                             struct pf_addr *, u_int16_t,
 160                             struct pf_addr *, u_int16_t *);
 161 int                      pf_test_tcp(struct pf_rule **, struct pf_state **,
 162                             int, struct pfi_kif *, struct mbuf *, int,
 163                             void *, struct pf_pdesc *, struct pf_rule **,
 164                             struct pf_ruleset **);
 165 int                      pf_test_udp(struct pf_rule **, struct pf_state **,
 166                             int, struct pfi_kif *, struct mbuf *, int,
 167                             void *, struct pf_pdesc *, struct pf_rule **,
 168                             struct pf_ruleset **);
 169 int                      pf_test_icmp(struct pf_rule **, struct pf_state **,
 170                             int, struct pfi_kif *, struct mbuf *, int,
 171                             void *, struct pf_pdesc *, struct pf_rule **,
 172                             struct pf_ruleset **);
 173 int                      pf_test_other(struct pf_rule **, struct pf_state **,
 174                             int, struct pfi_kif *, struct mbuf *, int, void *,
 175                             struct pf_pdesc *, struct pf_rule **,
 176                             struct pf_ruleset **);
 177 int                      pf_test_fragment(struct pf_rule **, int,
 178                             struct pfi_kif *, struct mbuf *, void *,
 179                             struct pf_pdesc *, struct pf_rule **,
 180                             struct pf_ruleset **);
 181 int                      pf_test_state_tcp(struct pf_state **, int,
 182                             struct pfi_kif *, struct mbuf *, int,
 183                             void *, struct pf_pdesc *, u_short *);
 184 int                      pf_test_state_udp(struct pf_state **, int,
 185                             struct pfi_kif *, struct mbuf *, int,
 186                             void *, struct pf_pdesc *);
 187 int                      pf_test_state_icmp(struct pf_state **, int,
 188                             struct pfi_kif *, struct mbuf *, int,
 189                             void *, struct pf_pdesc *);
 190 int                      pf_test_state_other(struct pf_state **, int,
 191                             struct pfi_kif *, struct pf_pdesc *);
 192 static int               pf_match_tag(struct mbuf *, struct pf_rule *,
 193                                       struct pf_rule *, int *);
 194 void                     pf_hash(struct pf_addr *, struct pf_addr *,
 195                             struct pf_poolhashkey *, sa_family_t);
 196 int                      pf_map_addr(u_int8_t, struct pf_rule *,
 197                             struct pf_addr *, struct pf_addr *,
 198                             struct pf_addr *, struct pf_src_node **);
 199 int                      pf_get_sport(sa_family_t, u_int8_t, struct pf_rule *,
 200                             struct pf_addr *, struct pf_addr *, u_int16_t,
 201                             struct pf_addr *, u_int16_t*, u_int16_t, u_int16_t,
 202                             struct pf_src_node **);
 203 void                     pf_route(struct mbuf **, struct pf_rule *, int,
 204                             struct ifnet *, struct pf_state *);
 205 void                     pf_route6(struct mbuf **, struct pf_rule *, int,
 206                             struct ifnet *, struct pf_state *);
 207 int                      pf_socket_lookup(uid_t *, gid_t *,
 208                             int, struct pf_pdesc *);
 209 u_int8_t                 pf_get_wscale(struct mbuf *, int, u_int16_t,
 210                             sa_family_t);
 211 u_int16_t                pf_get_mss(struct mbuf *, int, u_int16_t,
 212                             sa_family_t);
 213 u_int16_t                pf_calc_mss(struct pf_addr *, sa_family_t,
 214                                 u_int16_t);
 215 void                     pf_set_rt_ifp(struct pf_state *,
 216                             struct pf_addr *);
 217 int                      pf_check_proto_cksum(struct mbuf *, int, int,
 218                             u_int8_t, sa_family_t);
 219 int                      pf_addr_wrap_neq(struct pf_addr_wrap *,
 220                             struct pf_addr_wrap *);
 221 struct pf_state         *pf_find_state_recurse(struct pfi_kif *,
 222                             struct pf_state *, u_int8_t);
 223
 224 struct pf_pool_limit pf_pool_limits[PF_LIMIT_MAX];
 225
 226 #define STATE_LOOKUP()                                                  \
 227         do {                                                            \
 228                 if (direction == PF_IN)                                 \
 229                         *state = pf_find_state_recurse(         \
 230                             kif, &key, PF_EXT_GWY);                     \
 231                 else                                                    \
 232                         *state = pf_find_state_recurse(         \
 233                             kif, &key, PF_LAN_EXT);                     \
 234                 if (*state == NULL)                                     \
 235                         return (PF_DROP);                               \
 236                 if (direction == PF_OUT &&                              \
 237                     (((*state)->rule.ptr->rt == PF_ROUTETO &&           \
 238                     (*state)->rule.ptr->direction == PF_OUT) ||         \
 239                     ((*state)->rule.ptr->rt == PF_REPLYTO &&            \
 240                     (*state)->rule.ptr->direction == PF_IN)) &&         \
 241                     (*state)->rt_kif != NULL &&                         \
 242                     (*state)->rt_kif != kif)                            \
 243                         return (PF_PASS);                               \
 244         } while (0)
 245
 246 #define STATE_TRANSLATE(s) \
 247         (s)->lan.addr.addr32[0] != (s)->gwy.addr.addr32[0] || \
 248         ((s)->af == AF_INET6 && \
 249         ((s)->lan.addr.addr32[1] != (s)->gwy.addr.addr32[1] || \
 250         (s)->lan.addr.addr32[2] != (s)->gwy.addr.addr32[2] || \
 251         (s)->lan.addr.addr32[3] != (s)->gwy.addr.addr32[3])) || \
 252         (s)->lan.port != (s)->gwy.port
 253
 254 #define BOUND_IFACE(r, k) (((r)->rule_flag & PFRULE_IFBOUND) ? (k) :   \
 255         ((r)->rule_flag & PFRULE_GRBOUND) ? (k)->pfik_parent :         \
 256         (k)->pfik_parent->pfik_parent)
 257
 258 static int pf_src_compare(struct pf_src_node *, struct pf_src_node *);
 259 static int pf_state_compare_lan_ext(struct pf_state *,
 260         struct pf_state *);
 261 static int pf_state_compare_ext_gwy(struct pf_state *,
 262         struct pf_state *);
 263 static int pf_state_compare_id(struct pf_state *,
 264         struct pf_state *);
 265
 266 struct pf_src_tree tree_src_tracking;
 267
 268 struct pf_state_tree_id tree_id;
 269 struct pf_state_queue state_updates;
 270
 271 RB_GENERATE(pf_src_tree, pf_src_node, entry, pf_src_compare);
 272 RB_GENERATE(pf_state_tree_lan_ext, pf_state,
 273     u.s.entry_lan_ext, pf_state_compare_lan_ext);
 274 RB_GENERATE(pf_state_tree_ext_gwy, pf_state,
 275     u.s.entry_ext_gwy, pf_state_compare_ext_gwy);
 276 RB_GENERATE(pf_state_tree_id, pf_state,
 277     u.s.entry_id, pf_state_compare_id);
 278
 279 static int
 280 pf_src_compare(struct pf_src_node *a, struct pf_src_node *b)
 281 {
 282         int     diff;
 283
 284         if (a->rule.ptr > b->rule.ptr)
 285                 return (1);
 286         if (a->rule.ptr < b->rule.ptr)
 287                 return (-1);
 288         if ((diff = a->af - b->af) != 0)
 289                 return (diff);
 290         switch (a->af) {
 291 #ifdef INET
 292         case AF_INET:
 293                 if (a->addr.addr32[0] > b->addr.addr32[0])
 294                         return (1);
 295                 if (a->addr.addr32[0] < b->addr.addr32[0])
 296                         return (-1);
 297                 break;
 298 #endif /* INET */
 299 #ifdef INET6
 300         case AF_INET6:
 301                 if (a->addr.addr32[3] > b->addr.addr32[3])
 302                         return (1);
 303                 if (a->addr.addr32[3] < b->addr.addr32[3])
 304                         return (-1);
 305                 if (a->addr.addr32[2] > b->addr.addr32[2])
 306                         return (1);
 307                 if (a->addr.addr32[2] < b->addr.addr32[2])
 308                         return (-1);
 309                 if (a->addr.addr32[1] > b->addr.addr32[1])
 310                         return (1);
 311                 if (a->addr.addr32[1] < b->addr.addr32[1])
 312                         return (-1);
 313                 if (a->addr.addr32[0] > b->addr.addr32[0])
 314                         return (1);
 315                 if (a->addr.addr32[0] < b->addr.addr32[0])
 316                         return (-1);
 317                 break;
 318 #endif /* INET6 */
 319         }
 320         return (0);
 321 }
 322
 323 static int
 324 pf_state_compare_lan_ext(struct pf_state *a, struct pf_state *b)
 325 {
 326         int     diff;
 327
 328         if ((diff = a->proto - b->proto) != 0)
 329                 return (diff);
 330         if ((diff = a->af - b->af) != 0)
 331                 return (diff);
 332         switch (a->af) {
 333 #ifdef INET
 334         case AF_INET:
 335                 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
 336                         return (1);
 337                 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
 338                         return (-1);
 339                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 340                         return (1);
 341                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 342                         return (-1);
 343                 break;
 344 #endif /* INET */
 345 #ifdef INET6
 346         case AF_INET6:
 347                 if (a->lan.addr.addr32[3] > b->lan.addr.addr32[3])
 348                         return (1);
 349                 if (a->lan.addr.addr32[3] < b->lan.addr.addr32[3])
 350                         return (-1);
 351                 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
 352                         return (1);
 353                 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
 354                         return (-1);
 355                 if (a->lan.addr.addr32[2] > b->lan.addr.addr32[2])
 356                         return (1);
 357                 if (a->lan.addr.addr32[2] < b->lan.addr.addr32[2])
 358                         return (-1);
 359                 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
 360                         return (1);
 361                 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
 362                         return (-1);
 363                 if (a->lan.addr.addr32[1] > b->lan.addr.addr32[1])
 364                         return (1);
 365                 if (a->lan.addr.addr32[1] < b->lan.addr.addr32[1])
 366                         return (-1);
 367                 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
 368                         return (1);
 369                 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
 370                         return (-1);
 371                 if (a->lan.addr.addr32[0] > b->lan.addr.addr32[0])
 372                         return (1);
 373                 if (a->lan.addr.addr32[0] < b->lan.addr.addr32[0])
 374                         return (-1);
 375                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 376                         return (1);
 377                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 378                         return (-1);
 379                 break;
 380 #endif /* INET6 */
 381         }
 382
 383         if ((diff = a->lan.port - b->lan.port) != 0)
 384                 return (diff);
 385         if ((diff = a->ext.port - b->ext.port) != 0)
 386                 return (diff);
 387
 388         return (0);
 389 }
 390
 391 static int
 392 pf_state_compare_ext_gwy(struct pf_state *a, struct pf_state *b)
 393 {
 394         int     diff;
 395
 396         if ((diff = a->proto - b->proto) != 0)
 397                 return (diff);
 398         if ((diff = a->af - b->af) != 0)
 399                 return (diff);
 400         switch (a->af) {
 401 #ifdef INET
 402         case AF_INET:
 403                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 404                         return (1);
 405                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 406                         return (-1);
 407                 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
 408                         return (1);
 409                 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
 410                         return (-1);
 411                 break;
 412 #endif /* INET */
 413 #ifdef INET6
 414         case AF_INET6:
 415                 if (a->ext.addr.addr32[3] > b->ext.addr.addr32[3])
 416                         return (1);
 417                 if (a->ext.addr.addr32[3] < b->ext.addr.addr32[3])
 418                         return (-1);
 419                 if (a->gwy.addr.addr32[3] > b->gwy.addr.addr32[3])
 420                         return (1);
 421                 if (a->gwy.addr.addr32[3] < b->gwy.addr.addr32[3])
 422                         return (-1);
 423                 if (a->ext.addr.addr32[2] > b->ext.addr.addr32[2])
 424                         return (1);
 425                 if (a->ext.addr.addr32[2] < b->ext.addr.addr32[2])
 426                         return (-1);
 427                 if (a->gwy.addr.addr32[2] > b->gwy.addr.addr32[2])
 428                         return (1);
 429                 if (a->gwy.addr.addr32[2] < b->gwy.addr.addr32[2])
 430                         return (-1);
 431                 if (a->ext.addr.addr32[1] > b->ext.addr.addr32[1])
 432                         return (1);
 433                 if (a->ext.addr.addr32[1] < b->ext.addr.addr32[1])
 434                         return (-1);
 435                 if (a->gwy.addr.addr32[1] > b->gwy.addr.addr32[1])
 436                         return (1);
 437                 if (a->gwy.addr.addr32[1] < b->gwy.addr.addr32[1])
 438                         return (-1);
 439                 if (a->ext.addr.addr32[0] > b->ext.addr.addr32[0])
 440                         return (1);
 441                 if (a->ext.addr.addr32[0] < b->ext.addr.addr32[0])
 442                         return (-1);
 443                 if (a->gwy.addr.addr32[0] > b->gwy.addr.addr32[0])
 444                         return (1);
 445                 if (a->gwy.addr.addr32[0] < b->gwy.addr.addr32[0])
 446                         return (-1);
 447                 break;
 448 #endif /* INET6 */
 449         }
 450
 451         if ((diff = a->ext.port - b->ext.port) != 0)
 452                 return (diff);
 453         if ((diff = a->gwy.port - b->gwy.port) != 0)
 454                 return (diff);
 455
 456         return (0);
 457 }
 458
 459 static int
 460 pf_state_compare_id(struct pf_state *a, struct pf_state *b)
 461 {
 462         if (a->id > b->id)
 463                 return (1);
 464         if (a->id < b->id)
 465                 return (-1);
 466         if (a->creatorid > b->creatorid)
 467                 return (1);
 468         if (a->creatorid < b->creatorid)
 469                 return (-1);
 470
 471         return (0);
 472 }
 473
 474 #ifdef INET6
 475 void
 476 pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af)
 477 {
 478         switch (af) {
 479 #ifdef INET
 480         case AF_INET:
 481                 dst->addr32[0] = src->addr32[0];
 482                 break;
 483 #endif /* INET */
 484         case AF_INET6:
 485                 dst->addr32[0] = src->addr32[0];
 486                 dst->addr32[1] = src->addr32[1];
 487                 dst->addr32[2] = src->addr32[2];
 488                 dst->addr32[3] = src->addr32[3];
 489                 break;
 490         }
 491 }
 492 #endif
 493
 494 struct pf_state *
 495 pf_find_state_byid(struct pf_state *key)
 496 {
 497         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 498         return (RB_FIND(pf_state_tree_id, &tree_id, key));
 499 }
 500
 501 struct pf_state *
 502 pf_find_state_recurse(struct pfi_kif *kif, struct pf_state *key, u_int8_t tree)
 503 {
 504         struct pf_state *s;
 505
 506         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 507
 508         switch (tree) {
 509         case PF_LAN_EXT:
 510                 for (; kif != NULL; kif = kif->pfik_parent) {
 511                         s = RB_FIND(pf_state_tree_lan_ext,
 512                             &kif->pfik_lan_ext, key);
 513                         if (s != NULL)
 514                                 return (s);
 515                 }
 516                 return (NULL);
 517         case PF_EXT_GWY:
 518                 for (; kif != NULL; kif = kif->pfik_parent) {
 519                         s = RB_FIND(pf_state_tree_ext_gwy,
 520                             &kif->pfik_ext_gwy, key);
 521                         if (s != NULL)
 522                                 return (s);
 523                 }
 524                 return (NULL);
 525         default:
 526                 panic("pf_find_state_recurse");
 527         }
 528 }
 529
 530 struct pf_state *
 531 pf_find_state_all(struct pf_state *key, u_int8_t tree, int *more)
 532 {
 533         struct pf_state *s, *ss = NULL;
 534         struct pfi_kif  *kif;
 535
 536         pf_status.fcounters[FCNT_STATE_SEARCH]++;
 537
 538         switch (tree) {
 539         case PF_LAN_EXT:
 540                 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
 541                         s = RB_FIND(pf_state_tree_lan_ext,
 542                             &kif->pfik_lan_ext, key);
 543                         if (s == NULL)
 544                                 continue;
 545                         if (more == NULL)
 546                                 return (s);
 547                         ss = s;
 548                         (*more)++;
 549                 }
 550                 return (ss);
 551         case PF_EXT_GWY:
 552                 TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states) {
 553                         s = RB_FIND(pf_state_tree_ext_gwy,
 554                             &kif->pfik_ext_gwy, key);
 555                         if (s == NULL)
 556                                 continue;
 557                         if (more == NULL)
 558                                 return (s);
 559                         ss = s;
 560                         (*more)++;
 561                 }
 562                 return (ss);
 563         default:
 564                 panic("pf_find_state_all");
 565         }
 566 }
 567
 568 int
 569 pf_insert_src_node(struct pf_src_node **sn, struct pf_rule *rule,
 570     struct pf_addr *src, sa_family_t af)
 571 {
 572         struct pf_src_node      k;
 573
 574         if (*sn == NULL) {
 575                 k.af = af;
 576                 PF_ACPY(&k.addr, src, af);
 577                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 578                     rule->rpool.opts & PF_POOL_STICKYADDR)
 579                         k.rule.ptr = rule;
 580                 else
 581                         k.rule.ptr = NULL;
 582                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
 583                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
 584         }
 585         if (*sn == NULL) {
 586                 if (!rule->max_src_nodes ||
 587                     rule->src_nodes < rule->max_src_nodes)
 588                         (*sn) = pool_get(&pf_src_tree_pl, PR_NOWAIT);
 589                 if ((*sn) == NULL)
 590                         return (-1);
 591                 bzero(*sn, sizeof(struct pf_src_node));
 592                 (*sn)->af = af;
 593                 if (rule->rule_flag & PFRULE_RULESRCTRACK ||
 594                     rule->rpool.opts & PF_POOL_STICKYADDR)
 595                         (*sn)->rule.ptr = rule;
 596                 else
 597                         (*sn)->rule.ptr = NULL;
 598                 PF_ACPY(&(*sn)->addr, src, af);
 599                 if (RB_INSERT(pf_src_tree,
 600                     &tree_src_tracking, *sn) != NULL) {
 601                         if (pf_status.debug >= PF_DEBUG_MISC) {
 602                                 kprintf("pf: src_tree insert failed: ");
 603                                 pf_print_host(&(*sn)->addr, 0, af);
 604                                 kprintf("\n");
 605                         }
 606                         pool_put(&pf_src_tree_pl, *sn);
 607                         return (-1);
 608                 }
 609                 (*sn)->creation = time_second;
 610                 (*sn)->ruletype = rule->action;
 611                 if ((*sn)->rule.ptr != NULL)
 612                         (*sn)->rule.ptr->src_nodes++;
 613                 pf_status.scounters[SCNT_SRC_NODE_INSERT]++;
 614                 pf_status.src_nodes++;
 615         } else {
 616                 if (rule->max_src_states &&
 617                     (*sn)->states >= rule->max_src_states)
 618                         return (-1);
 619         }
 620         return (0);
 621 }
 622
 623 int
 624 pf_insert_state(struct pfi_kif *kif, struct pf_state *state)
 625 {
 626         /* Thou MUST NOT insert multiple duplicate keys */
 627         state->u.s.kif = kif;
 628         if (RB_INSERT(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state)) {
 629                 if (pf_status.debug >= PF_DEBUG_MISC) {
 630                         kprintf("pf: state insert failed: tree_lan_ext");
 631                         kprintf(" lan: ");
 632                         pf_print_host(&state->lan.addr, state->lan.port,
 633                             state->af);
 634                         kprintf(" gwy: ");
 635                         pf_print_host(&state->gwy.addr, state->gwy.port,
 636                             state->af);
 637                         kprintf(" ext: ");
 638                         pf_print_host(&state->ext.addr, state->ext.port,
 639                             state->af);
 640                         if (state->sync_flags & PFSTATE_FROMSYNC)
 641                                 kprintf(" (from sync)");
 642                         kprintf("\n");
 643                 }
 644                 return (-1);
 645         }
 646
 647         if (RB_INSERT(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state)) {
 648                 if (pf_status.debug >= PF_DEBUG_MISC) {
 649                         kprintf("pf: state insert failed: tree_ext_gwy");
 650                         kprintf(" lan: ");
 651                         pf_print_host(&state->lan.addr, state->lan.port,
 652                             state->af);
 653                         kprintf(" gwy: ");
 654                         pf_print_host(&state->gwy.addr, state->gwy.port,
 655                             state->af);
 656                         kprintf(" ext: ");
 657                         pf_print_host(&state->ext.addr, state->ext.port,
 658                             state->af);
 659                         if (state->sync_flags & PFSTATE_FROMSYNC)
 660                                 kprintf(" (from sync)");
 661                         kprintf("\n");
 662                 }
 663                 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
 664                 return (-1);
 665         }
 666
 667         if (state->id == 0 && state->creatorid == 0) {
 668                 state->id = htobe64(pf_status.stateid++);
 669                 state->creatorid = pf_status.hostid;
 670         }
 671         if (RB_INSERT(pf_state_tree_id, &tree_id, state) != NULL) {
 672                 if (pf_status.debug >= PF_DEBUG_MISC) {
 673                         kprintf("pf: state insert failed: "
 674                             "id: %016" PRIx64 " creatorid: %08" PRIx32,
 675                             be64toh(state->id), ntohl(state->creatorid));
 676                         if (state->sync_flags & PFSTATE_FROMSYNC)
 677                                 kprintf(" (from sync)");
 678                         kprintf("\n");
 679                 }
 680                 RB_REMOVE(pf_state_tree_lan_ext, &kif->pfik_lan_ext, state);
 681                 RB_REMOVE(pf_state_tree_ext_gwy, &kif->pfik_ext_gwy, state);
 682                 return (-1);
 683         }
 684         TAILQ_INSERT_HEAD(&state_updates, state, u.s.entry_updates);
 685
 686         pf_status.fcounters[FCNT_STATE_INSERT]++;
 687         pf_status.states++;
 688         pfi_attach_state(kif);
 689 #if NPFSYNC
 690         pfsync_insert_state(state);
 691 #endif
 692         return (0);
 693 }
 694
 695 void
 696 pf_purge_timeout(void *arg)
 697 {
 698         struct callout  *to = arg;
 699
 700         crit_enter();
 701         pf_purge_expired_states();
 702         pf_purge_expired_fragments();
 703         pf_purge_expired_src_nodes();
 704         crit_exit();
 705
 706         callout_reset(to, pf_default_rule.timeout[PFTM_INTERVAL] * hz,
 707             pf_purge_timeout, to);
 708 }
 709
 710 u_int32_t
 711 pf_state_expires(const struct pf_state *state)
 712 {
 713         u_int32_t       timeout;
 714         u_int32_t       start;
 715         u_int32_t       end;
 716         u_int32_t       states;
 717
 718         /* handle all PFTM_* > PFTM_MAX here */
 719         if (state->timeout == PFTM_PURGE)
 720                 return (time_second);
 721         if (state->timeout == PFTM_UNTIL_PACKET)
 722                 return (0);
 723         KASSERT((state->timeout < PFTM_MAX),
 724             ("pf_state_expires: timeout > PFTM_MAX"));
 725         timeout = state->rule.ptr->timeout[state->timeout];
 726         if (!timeout)
 727                 timeout = pf_default_rule.timeout[state->timeout];
 728         start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START];
 729         if (start) {
 730                 end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END];
 731                 states = state->rule.ptr->states;
 732         } else {
 733                 start = pf_default_rule.timeout[PFTM_ADAPTIVE_START];
 734                 end = pf_default_rule.timeout[PFTM_ADAPTIVE_END];
 735                 states = pf_status.states;
 736         }
 737         if (end && states > start && start < end) {
 738                 if (states < end)
 739                         return (state->expire + timeout * (end - states) /
 740                             (end - start));
 741                 else
 742                         return (time_second);
 743         }
 744         return (state->expire + timeout);
 745 }
 746
 747 void
 748 pf_purge_expired_src_nodes(void)
 749 {
 750          struct pf_src_node             *cur, *next;
 751
 752          for (cur = RB_MIN(pf_src_tree, &tree_src_tracking); cur; cur = next) {
 753                  next = RB_NEXT(pf_src_tree, &tree_src_tracking, cur);
 754
 755                  if (cur->states <= 0 && cur->expire <= time_second) {
 756                          if (cur->rule.ptr != NULL) {
 757                                  cur->rule.ptr->src_nodes--;
 758                                  if (cur->rule.ptr->states <= 0 &&
 759                                      cur->rule.ptr->max_src_nodes <= 0)
 760                                          pf_rm_rule(NULL, cur->rule.ptr);
 761                          }
 762                          RB_REMOVE(pf_src_tree, &tree_src_tracking, cur);
 763                          pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
 764                          pf_status.src_nodes--;
 765                          pool_put(&pf_src_tree_pl, cur);
 766                  }
 767          }
 768 }
 769
 770 void
 771 pf_src_tree_remove_state(struct pf_state *s)
 772 {
 773         u_int32_t timeout;
 774
 775         if (s->src_node != NULL) {
 776                 if (--s->src_node->states <= 0) {
 777                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
 778                         if (!timeout)
 779                                 timeout =
 780                                     pf_default_rule.timeout[PFTM_SRC_NODE];
 781                         s->src_node->expire = time_second + timeout;
 782                 }
 783         }
 784         if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) {
 785                 if (--s->nat_src_node->states <= 0) {
 786                         timeout = s->rule.ptr->timeout[PFTM_SRC_NODE];
 787                         if (!timeout)
 788                                 timeout =
 789                                     pf_default_rule.timeout[PFTM_SRC_NODE];
 790                         s->nat_src_node->expire = time_second + timeout;
 791                 }
 792         }
 793         s->src_node = s->nat_src_node = NULL;
 794 }
 795
 796 static int
 797 pf_purge_expired_states_callback(struct pf_state *cur, void *data __unused)
 798 {
 799         if (pf_state_expires(cur) <= time_second) {
 800                 RB_REMOVE(pf_state_tree_ext_gwy,
 801                     &cur->u.s.kif->pfik_ext_gwy, cur);
 802                 RB_REMOVE(pf_state_tree_lan_ext,
 803                     &cur->u.s.kif->pfik_lan_ext, cur);
 804                 RB_REMOVE(pf_state_tree_id, &tree_id, cur);
 805                 if (cur->src.state == PF_TCPS_PROXY_DST) {
 806                         pf_send_tcp(cur->rule.ptr, cur->af,
 807                             &cur->ext.addr, &cur->lan.addr,
 808                             cur->ext.port, cur->lan.port,
 809                             cur->src.seqhi, cur->src.seqlo + 1, 0,
 810                             TH_RST|TH_ACK, 0, 0);
 811                 }
 812 #if NPFSYNC
 813                 pfsync_delete_state(cur);
 814 #endif
 815                 pf_src_tree_remove_state(cur);
 816                 if (--cur->rule.ptr->states <= 0 &&
 817                     cur->rule.ptr->src_nodes <= 0)
 818                         pf_rm_rule(NULL, cur->rule.ptr);
 819                 if (cur->nat_rule.ptr != NULL)
 820                         if (--cur->nat_rule.ptr->states <= 0 &&
 821                                 cur->nat_rule.ptr->src_nodes <= 0)
 822                                 pf_rm_rule(NULL, cur->nat_rule.ptr);
 823                 if (cur->anchor.ptr != NULL)
 824                         if (--cur->anchor.ptr->states <= 0)
 825                                 pf_rm_rule(NULL, cur->anchor.ptr);
 826                 pf_normalize_tcp_cleanup(cur);
 827                 pfi_detach_state(cur->u.s.kif);
 828                 TAILQ_REMOVE(&state_updates, cur, u.s.entry_updates);
 829                 pool_put(&pf_state_pl, cur);
 830                 pf_status.fcounters[FCNT_STATE_REMOVALS]++;
 831                 pf_status.states--;
 832         }
 833         return(0);
 834 }
 835
 836 void
 837 pf_purge_expired_states(void)
 838 {
 839         RB_SCAN(pf_state_tree_id, &tree_id, NULL,
 840                 pf_purge_expired_states_callback, NULL);
 841 }
 842
 843
 844 int
 845 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
 846 {
 847         if (aw->type != PF_ADDR_TABLE)
 848                 return (0);
 849         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
 850                 return (1);
 851         return (0);
 852 }
 853
 854 void
 855 pf_tbladdr_remove(struct pf_addr_wrap *aw)
 856 {
 857         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
 858                 return;
 859         pfr_detach_table(aw->p.tbl);
 860         aw->p.tbl = NULL;
 861 }
 862
 863 void
 864 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
 865 {
 866         struct pfr_ktable *kt = aw->p.tbl;
 867
 868         if (aw->type != PF_ADDR_TABLE || kt == NULL)
 869                 return;
 870         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
 871                 kt = kt->pfrkt_root;
 872         aw->p.tbl = NULL;
 873         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
 874                 kt->pfrkt_cnt : -1;
 875 }
 876
 877 void
 878 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
 879 {
 880         switch (af) {
 881 #ifdef INET
 882         case AF_INET: {
 883                 u_int32_t a = ntohl(addr->addr32[0]);
 884                 kprintf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
 885                     (a>>8)&255, a&255);
 886                 if (p) {
 887                         p = ntohs(p);
 888                         kprintf(":%u", p);
 889                 }
 890                 break;
 891         }
 892 #endif /* INET */
 893 #ifdef INET6
 894         case AF_INET6: {
 895                 u_int16_t b;
 896                 u_int8_t i, curstart = 255, curend = 0,
 897                     maxstart = 0, maxend = 0;
 898                 for (i = 0; i < 8; i++) {
 899                         if (!addr->addr16[i]) {
 900                                 if (curstart == 255)
 901                                         curstart = i;
 902                                 else
 903                                         curend = i;
 904                         } else {
 905                                 if (curstart) {
 906                                         if ((curend - curstart) >
 907                                             (maxend - maxstart)) {
 908                                                 maxstart = curstart;
 909                                                 maxend = curend;
 910                                                 curstart = 255;
 911                                         }
 912                                 }
 913                         }
 914                 }
 915                 for (i = 0; i < 8; i++) {
 916                         if (i >= maxstart && i <= maxend) {
 917                                 if (maxend != 7) {
 918                                         if (i == maxstart)
 919                                                 kprintf(":");
 920                                 } else {
 921                                         if (i == maxend)
 922                                                 kprintf(":");
 923                                 }
 924                         } else {
 925                                 b = ntohs(addr->addr16[i]);
 926                                 kprintf("%x", b);
 927                                 if (i < 7)
 928                                         kprintf(":");
 929                         }
 930                 }
 931                 if (p) {
 932                         p = ntohs(p);
 933                         kprintf("[%u]", p);
 934                 }
 935                 break;
 936         }
 937 #endif /* INET6 */
 938         }
 939 }
 940
 941 void
 942 pf_print_state(struct pf_state *s)
 943 {
 944         switch (s->proto) {
 945         case IPPROTO_TCP:
 946                 kprintf("TCP ");
 947                 break;
 948         case IPPROTO_UDP:
 949                 kprintf("UDP ");
 950                 break;
 951         case IPPROTO_ICMP:
 952                 kprintf("ICMP ");
 953                 break;
 954         case IPPROTO_ICMPV6:
 955                 kprintf("ICMPV6 ");
 956                 break;
 957         default:
 958                 kprintf("%u ", s->proto);
 959                 break;
 960         }
 961         pf_print_host(&s->lan.addr, s->lan.port, s->af);
 962         kprintf(" ");
 963         pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
 964         kprintf(" ");
 965         pf_print_host(&s->ext.addr, s->ext.port, s->af);
 966         kprintf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
 967             s->src.seqhi, s->src.max_win, s->src.seqdiff);
 968         if (s->src.wscale && s->dst.wscale)
 969                 kprintf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
 970         kprintf("]");
 971         kprintf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
 972             s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
 973         if (s->src.wscale && s->dst.wscale)
 974                 kprintf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
 975         kprintf("]");
 976         kprintf(" %u:%u", s->src.state, s->dst.state);
 977 }
 978
 979 void
 980 pf_print_flags(u_int8_t f)
 981 {
 982         if (f)
 983                 kprintf(" ");
 984         if (f & TH_FIN)
 985                 kprintf("F");
 986         if (f & TH_SYN)
 987                 kprintf("S");
 988         if (f & TH_RST)
 989                 kprintf("R");
 990         if (f & TH_PUSH)
 991                 kprintf("P");
 992         if (f & TH_ACK)
 993                 kprintf("A");
 994         if (f & TH_URG)
 995                 kprintf("U");
 996         if (f & TH_ECE)
 997                 kprintf("E");
 998         if (f & TH_CWR)
 999                 kprintf("W");
1000 }
1001
1002 #define PF_SET_SKIP_STEPS(i)                                    \
1003         do {                                                    \
1004                 while (head[i] != cur) {                        \
1005                         head[i]->skip[i].ptr = cur;             \
1006                         head[i] = TAILQ_NEXT(head[i], entries); \
1007                 }                                               \
1008         } while (0)
1009
1010 void
1011 pf_calc_skip_steps(struct pf_rulequeue *rules)
1012 {
1013         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1014         int i;
1015
1016         cur = TAILQ_FIRST(rules);
1017         prev = cur;
1018         for (i = 0; i < PF_SKIP_COUNT; ++i)
1019                 head[i] = cur;
1020         while (cur != NULL) {
1021
1022                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1023                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1024                 if (cur->direction != prev->direction)
1025                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1026                 if (cur->af != prev->af)
1027                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1028                 if (cur->proto != prev->proto)
1029                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1030                 if (cur->src.not != prev->src.not ||
1031                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1032                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1033                 if (cur->src.port[0] != prev->src.port[0] ||
1034                     cur->src.port[1] != prev->src.port[1] ||
1035                     cur->src.port_op != prev->src.port_op)
1036                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1037                 if (cur->dst.not != prev->dst.not ||
1038                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1039                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1040                 if (cur->dst.port[0] != prev->dst.port[0] ||
1041                     cur->dst.port[1] != prev->dst.port[1] ||
1042                     cur->dst.port_op != prev->dst.port_op)
1043                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1044
1045                 prev = cur;
1046                 cur = TAILQ_NEXT(cur, entries);
1047         }
1048         for (i = 0; i < PF_SKIP_COUNT; ++i)
1049                 PF_SET_SKIP_STEPS(i);
1050 }
1051
1052 int
1053 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1054 {
1055         if (aw1->type != aw2->type)
1056                 return (1);
1057         switch (aw1->type) {
1058         case PF_ADDR_ADDRMASK:
1059                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1060                         return (1);
1061                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1062                         return (1);
1063                 return (0);
1064         case PF_ADDR_DYNIFTL:
1065                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1066         case PF_ADDR_NOROUTE:
1067                 return (0);
1068         case PF_ADDR_TABLE:
1069                 return (aw1->p.tbl != aw2->p.tbl);
1070         default:
1071                 kprintf("invalid address type: %d\n", aw1->type);
1072                 return (1);
1073         }
1074 }
1075
1076 void
1077 pf_update_anchor_rules(void)
1078 {
1079         struct pf_rule  *rule;
1080         int              i;
1081
1082         for (i = 0; i < PF_RULESET_MAX; ++i)
1083                 TAILQ_FOREACH(rule, pf_main_ruleset.rules[i].active.ptr,
1084                     entries)
1085                         if (rule->anchorname[0])
1086                                 rule->anchor = pf_find_anchor(rule->anchorname);
1087                         else
1088                                 rule->anchor = NULL;
1089 }
1090
1091 u_int16_t
1092 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1093 {
1094         u_int32_t       l;
1095
1096         if (udp && !cksum)
1097                 return (0x0000);
1098         l = cksum + old - new;
1099         l = (l >> 16) + (l & 65535);
1100         l = l & 65535;
1101         if (udp && !l)
1102                 return (0xFFFF);
1103         return (l);
1104 }
1105
1106 void
1107 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1108     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1109 {
1110         struct pf_addr  ao;
1111         u_int16_t       po = *p;
1112
1113         PF_ACPY(&ao, a, af);
1114         PF_ACPY(a, an, af);
1115
1116         *p = pn;
1117
1118         switch (af) {
1119 #ifdef INET
1120         case AF_INET:
1121                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1122                     ao.addr16[0], an->addr16[0], 0),
1123                     ao.addr16[1], an->addr16[1], 0);
1124                 *p = pn;
1125                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1126                     ao.addr16[0], an->addr16[0], u),
1127                     ao.addr16[1], an->addr16[1], u),
1128                     po, pn, u);
1129                 break;
1130 #endif /* INET */
1131 #ifdef INET6
1132         case AF_INET6:
1133                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1134                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1135                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1136                     ao.addr16[0], an->addr16[0], u),
1137                     ao.addr16[1], an->addr16[1], u),
1138                     ao.addr16[2], an->addr16[2], u),
1139                     ao.addr16[3], an->addr16[3], u),
1140                     ao.addr16[4], an->addr16[4], u),
1141                     ao.addr16[5], an->addr16[5], u),
1142                     ao.addr16[6], an->addr16[6], u),
1143                     ao.addr16[7], an->addr16[7], u),
1144                     po, pn, u);
1145                 break;
1146 #endif /* INET6 */
1147         }
1148 }
1149
1150
1151 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1152 void
1153 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1154 {
1155         u_int32_t       ao;
1156
1157         memcpy(&ao, a, sizeof(ao));
1158         memcpy(a, &an, sizeof(u_int32_t));
1159         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1160             ao % 65536, an % 65536, u);
1161 }
1162
1163 #ifdef INET6
1164 void
1165 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1166 {
1167         struct pf_addr  ao;
1168
1169         PF_ACPY(&ao, a, AF_INET6);
1170         PF_ACPY(a, an, AF_INET6);
1171
1172         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1173             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1174             pf_cksum_fixup(pf_cksum_fixup(*c,
1175             ao.addr16[0], an->addr16[0], u),
1176             ao.addr16[1], an->addr16[1], u),
1177             ao.addr16[2], an->addr16[2], u),
1178             ao.addr16[3], an->addr16[3], u),
1179             ao.addr16[4], an->addr16[4], u),
1180             ao.addr16[5], an->addr16[5], u),
1181             ao.addr16[6], an->addr16[6], u),
1182             ao.addr16[7], an->addr16[7], u);
1183 }
1184 #endif /* INET6 */
1185
1186 void
1187 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1188     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1189     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1190 {
1191         struct pf_addr  oia, ooa;
1192
1193         PF_ACPY(&oia, ia, af);
1194         PF_ACPY(&ooa, oa, af);
1195
1196         /* Change inner protocol port, fix inner protocol checksum. */
1197         if (ip != NULL) {
1198                 u_int16_t       oip = *ip;
1199                 u_int32_t       opc = 0;
1200
1201                 if (pc != NULL)
1202                         opc = *pc;
1203                 *ip = np;
1204                 if (pc != NULL)
1205                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1206                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1207                 if (pc != NULL)
1208                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1209         }
1210         /* Change inner ip address, fix inner ip and icmp checksums. */
1211         PF_ACPY(ia, na, af);
1212         switch (af) {
1213 #ifdef INET
1214         case AF_INET: {
1215                 u_int32_t        oh2c = *h2c;
1216
1217                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1218                     oia.addr16[0], ia->addr16[0], 0),
1219                     oia.addr16[1], ia->addr16[1], 0);
1220                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1221                     oia.addr16[0], ia->addr16[0], 0),
1222                     oia.addr16[1], ia->addr16[1], 0);
1223                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1224                 break;
1225         }
1226 #endif /* INET */
1227 #ifdef INET6
1228         case AF_INET6:
1229                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1230                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1231                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1232                     oia.addr16[0], ia->addr16[0], u),
1233                     oia.addr16[1], ia->addr16[1], u),
1234                     oia.addr16[2], ia->addr16[2], u),
1235                     oia.addr16[3], ia->addr16[3], u),
1236                     oia.addr16[4], ia->addr16[4], u),
1237                     oia.addr16[5], ia->addr16[5], u),
1238                     oia.addr16[6], ia->addr16[6], u),
1239                     oia.addr16[7], ia->addr16[7], u);
1240                 break;
1241 #endif /* INET6 */
1242         }
1243         /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1244         PF_ACPY(oa, na, af);
1245         switch (af) {
1246 #ifdef INET
1247         case AF_INET:
1248                 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1249                     ooa.addr16[0], oa->addr16[0], 0),
1250                     ooa.addr16[1], oa->addr16[1], 0);
1251                 break;
1252 #endif /* INET */
1253 #ifdef INET6
1254         case AF_INET6:
1255                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1256                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1257                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1258                     ooa.addr16[0], oa->addr16[0], u),
1259                     ooa.addr16[1], oa->addr16[1], u),
1260                     ooa.addr16[2], oa->addr16[2], u),
1261                     ooa.addr16[3], oa->addr16[3], u),
1262                     ooa.addr16[4], oa->addr16[4], u),
1263                     ooa.addr16[5], oa->addr16[5], u),
1264                     ooa.addr16[6], oa->addr16[6], u),
1265                     ooa.addr16[7], oa->addr16[7], u);
1266                 break;
1267 #endif /* INET6 */
1268         }
1269 }
1270
1271 void
1272 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1273     const struct pf_addr *saddr, const struct pf_addr *daddr,
1274     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1275     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl)
1276 {
1277         struct mbuf     *m;
1278         int              len = 0, tlen;
1279 #ifdef INET
1280         struct ip       *h = NULL;
1281 #endif /* INET */
1282 #ifdef INET6
1283         struct ip6_hdr  *h6 = NULL;
1284 #endif /* INET6 */
1285         struct tcphdr   *th = NULL;
1286         char *opt;
1287
1288         /* maximum segment size tcp option */
1289         tlen = sizeof(struct tcphdr);
1290         if (mss)
1291                 tlen += 4;
1292
1293         switch (af) {
1294 #ifdef INET
1295         case AF_INET:
1296                 len = sizeof(struct ip) + tlen;
1297                 break;
1298 #endif /* INET */
1299 #ifdef INET6
1300         case AF_INET6:
1301                 len = sizeof(struct ip6_hdr) + tlen;
1302                 break;
1303 #endif /* INET6 */
1304         }
1305
1306         /* create outgoing mbuf */
1307         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1308         if (m == NULL)
1309                 return;
1310         m->m_pkthdr.fw_flags = PF_MBUF_GENERATED;
1311 #ifdef ALTQ
1312         if (r != NULL && r->qid) {
1313                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
1314                 m->m_pkthdr.altq_qid = r->qid;
1315                 m->m_pkthdr.ecn_af = af;
1316                 m->m_pkthdr.header = mtod(m, struct ip *);
1317         }
1318 #endif
1319         m->m_data += max_linkhdr;
1320         m->m_pkthdr.len = m->m_len = len;
1321         m->m_pkthdr.rcvif = NULL;
1322         bzero(m->m_data, len);
1323         switch (af) {
1324 #ifdef INET
1325         case AF_INET:
1326                 h = mtod(m, struct ip *);
1327
1328                 /* IP header fields included in the TCP checksum */
1329                 h->ip_p = IPPROTO_TCP;
1330                 h->ip_len = tlen;
1331                 h->ip_src.s_addr = saddr->v4.s_addr;
1332                 h->ip_dst.s_addr = daddr->v4.s_addr;
1333
1334                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1335                 break;
1336 #endif /* INET */
1337 #ifdef INET6
1338         case AF_INET6:
1339                 h6 = mtod(m, struct ip6_hdr *);
1340
1341                 /* IP header fields included in the TCP checksum */
1342                 h6->ip6_nxt = IPPROTO_TCP;
1343                 h6->ip6_plen = htons(tlen);
1344                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1345                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1346
1347                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1348                 break;
1349 #endif /* INET6 */
1350         }
1351
1352         /* TCP header */
1353         th->th_sport = sport;
1354         th->th_dport = dport;
1355         th->th_seq = htonl(seq);
1356         th->th_ack = htonl(ack);
1357         th->th_off = tlen >> 2;
1358         th->th_flags = flags;
1359         th->th_win = htons(win);
1360
1361         if (mss) {
1362                 opt = (char *)(th + 1);
1363                 opt[0] = TCPOPT_MAXSEG;
1364                 opt[1] = 4;
1365                 mss = htons(mss);
1366                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1367         }
1368
1369         switch (af) {
1370 #ifdef INET
1371         case AF_INET:
1372                 /* TCP checksum */
1373                 th->th_sum = in_cksum(m, len);
1374
1375                 /* Finish the IP header */
1376                 h->ip_v = 4;
1377                 h->ip_hl = sizeof(*h) >> 2;
1378                 h->ip_tos = IPTOS_LOWDELAY;
1379                 h->ip_len = len;
1380                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1381                 h->ip_ttl = ttl ? ttl : ip_defttl;
1382                 h->ip_sum = 0;
1383                 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1384                     (void *)NULL);
1385                 break;
1386 #endif /* INET */
1387 #ifdef INET6
1388         case AF_INET6:
1389                 /* TCP checksum */
1390                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1391                     sizeof(struct ip6_hdr), tlen);
1392
1393                 h6->ip6_vfc |= IPV6_VERSION;
1394                 h6->ip6_hlim = IPV6_DEFHLIM;
1395
1396                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1397                 break;
1398 #endif /* INET6 */
1399         }
1400 }
1401
1402 void
1403 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1404     struct pf_rule *r)
1405 {
1406         struct mbuf     *m0;
1407
1408         m0 = m_copypacket(m, MB_DONTWAIT);
1409         if (m0 == NULL)
1410                 return;
1411         m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
1412
1413 #ifdef ALTQ
1414         if (r->qid) {
1415                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
1416                 m->m_pkthdr.altq_qid = r->qid;
1417                 m->m_pkthdr.ecn_af = af;
1418                 m->m_pkthdr.header = mtod(m0, struct ip *);
1419         }
1420 #endif
1421
1422         switch (af) {
1423 #ifdef INET
1424         case AF_INET:
1425                 icmp_error(m0, type, code, 0, 0);
1426                 break;
1427 #endif /* INET */
1428 #ifdef INET6
1429         case AF_INET6:
1430                 icmp6_error(m0, type, code, 0);
1431                 break;
1432 #endif /* INET6 */
1433         }
1434 }
1435
1436 /*
1437  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1438  * If n is 0, they match if they are equal. If n is != 0, they match if they
1439  * are different.
1440  */
1441 int
1442 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1443     struct pf_addr *b, sa_family_t af)
1444 {
1445         int     match = 0;
1446
1447         switch (af) {
1448 #ifdef INET
1449         case AF_INET:
1450                 if ((a->addr32[0] & m->addr32[0]) ==
1451                     (b->addr32[0] & m->addr32[0]))
1452                         match++;
1453                 break;
1454 #endif /* INET */
1455 #ifdef INET6
1456         case AF_INET6:
1457                 if (((a->addr32[0] & m->addr32[0]) ==
1458                      (b->addr32[0] & m->addr32[0])) &&
1459                     ((a->addr32[1] & m->addr32[1]) ==
1460                      (b->addr32[1] & m->addr32[1])) &&
1461                     ((a->addr32[2] & m->addr32[2]) ==
1462                      (b->addr32[2] & m->addr32[2])) &&
1463                     ((a->addr32[3] & m->addr32[3]) ==
1464                      (b->addr32[3] & m->addr32[3])))
1465                         match++;
1466                 break;
1467 #endif /* INET6 */
1468         }
1469         if (match) {
1470                 if (n)
1471                         return (0);
1472                 else
1473                         return (1);
1474         } else {
1475                 if (n)
1476                         return (1);
1477                 else
1478                         return (0);
1479         }
1480 }
1481
1482 int
1483 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1484 {
1485         switch (op) {
1486         case PF_OP_IRG:
1487                 return ((p > a1) && (p < a2));
1488         case PF_OP_XRG:
1489                 return ((p < a1) || (p > a2));
1490         case PF_OP_RRG:
1491                 return ((p >= a1) && (p <= a2));
1492         case PF_OP_EQ:
1493                 return (p == a1);
1494         case PF_OP_NE:
1495                 return (p != a1);
1496         case PF_OP_LT:
1497                 return (p < a1);
1498         case PF_OP_LE:
1499                 return (p <= a1);
1500         case PF_OP_GT:
1501                 return (p > a1);
1502         case PF_OP_GE:
1503                 return (p >= a1);
1504         }
1505         return (0); /* never reached */
1506 }
1507
1508 int
1509 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1510 {
1511         a1 = ntohs(a1);
1512         a2 = ntohs(a2);
1513         p = ntohs(p);
1514         return (pf_match(op, a1, a2, p));
1515 }
1516
1517 int
1518 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1519 {
1520         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1521                 return (0);
1522         return (pf_match(op, a1, a2, u));
1523 }
1524
1525 int
1526 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1527 {
1528         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1529                 return (0);
1530         return (pf_match(op, a1, a2, g));
1531 }
1532
1533 static int
1534 pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_rule *nat_rule,
1535              int *tag)
1536 {
1537         if (*tag == -1) {       /* find mbuf tag */
1538                 if (nat_rule != NULL && nat_rule->tag)
1539                         *tag = nat_rule->tag;
1540                 else if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
1541                         *tag = m->m_pkthdr.pf_tag;
1542                 else
1543                         *tag = 0;
1544         }
1545
1546         return ((!r->match_tag_not && r->match_tag == *tag) ||
1547             (r->match_tag_not && r->match_tag != *tag));
1548 }
1549
1550 void
1551 pf_tag_packet(struct mbuf *m, int tag)
1552 {
1553         if (tag <= 0)
1554                 return;
1555
1556         m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1557         m->m_pkthdr.pf_tag = tag;
1558 }
1559
1560 #define PF_STEP_INTO_ANCHOR(r, a, s, n)                                 \
1561         do {                                                            \
1562                 if ((r) == NULL || (r)->anchor == NULL ||               \
1563                     (s) != NULL || (a) != NULL)                         \
1564                         panic("PF_STEP_INTO_ANCHOR");                   \
1565                 (a) = (r);                                              \
1566                 (s) = TAILQ_FIRST(&(r)->anchor->rulesets);              \
1567                 (r) = NULL;                                             \
1568                 while ((s) != NULL && ((r) =                            \
1569                     TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL)     \
1570                         (s) = TAILQ_NEXT((s), entries);                 \
1571                 if ((r) == NULL) {                                      \
1572                         (r) = TAILQ_NEXT((a), entries);                 \
1573                         (a) = NULL;                                     \
1574                 }                                                       \
1575         } while (0)
1576
1577 #define PF_STEP_OUT_OF_ANCHOR(r, a, s, n)                               \
1578         do {                                                            \
1579                 if ((r) != NULL || (a) == NULL || (s) == NULL)          \
1580                         panic("PF_STEP_OUT_OF_ANCHOR");                 \
1581                 (s) = TAILQ_NEXT((s), entries);                         \
1582                 while ((s) != NULL && ((r) =                            \
1583                     TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL)     \
1584                         (s) = TAILQ_NEXT((s), entries);                 \
1585                 if ((r) == NULL) {                                      \
1586                         (r) = TAILQ_NEXT((a), entries);                 \
1587                         (a) = NULL;                                     \
1588                 }                                                       \
1589         } while (0)
1590
1591 #ifdef INET6
1592 void
1593 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
1594     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
1595 {
1596         switch (af) {
1597 #ifdef INET
1598         case AF_INET:
1599                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1600                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1601                 break;
1602 #endif /* INET */
1603         case AF_INET6:
1604                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1605                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1606                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
1607                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
1608                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
1609                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
1610                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
1611                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
1612                 break;
1613         }
1614 }
1615
1616 void
1617 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
1618 {
1619         switch (af) {
1620 #ifdef INET
1621         case AF_INET:
1622                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
1623                 break;
1624 #endif /* INET */
1625         case AF_INET6:
1626                 if (addr->addr32[3] == 0xffffffff) {
1627                         addr->addr32[3] = 0;
1628                         if (addr->addr32[2] == 0xffffffff) {
1629                                 addr->addr32[2] = 0;
1630                                 if (addr->addr32[1] == 0xffffffff) {
1631                                         addr->addr32[1] = 0;
1632                                         addr->addr32[0] =
1633                                             htonl(ntohl(addr->addr32[0]) + 1);
1634                                 } else
1635                                         addr->addr32[1] =
1636                                             htonl(ntohl(addr->addr32[1]) + 1);
1637                         } else
1638                                 addr->addr32[2] =
1639                                     htonl(ntohl(addr->addr32[2]) + 1);
1640                 } else
1641                         addr->addr32[3] =
1642                             htonl(ntohl(addr->addr32[3]) + 1);
1643                 break;
1644         }
1645 }
1646 #endif /* INET6 */
1647
1648 #define mix(a,b,c) \
1649         do {                                    \
1650                 a -= b; a -= c; a ^= (c >> 13); \
1651                 b -= c; b -= a; b ^= (a << 8);  \
1652                 c -= a; c -= b; c ^= (b >> 13); \
1653                 a -= b; a -= c; a ^= (c >> 12); \
1654                 b -= c; b -= a; b ^= (a << 16); \
1655                 c -= a; c -= b; c ^= (b >> 5);  \
1656                 a -= b; a -= c; a ^= (c >> 3);  \
1657                 b -= c; b -= a; b ^= (a << 10); \
1658                 c -= a; c -= b; c ^= (b >> 15); \
1659         } while (0)
1660
1661 /*
1662  * hash function based on bridge_hash in if_bridge.c
1663  */
1664 void
1665 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
1666     struct pf_poolhashkey *key, sa_family_t af)
1667 {
1668         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
1669
1670         switch (af) {
1671 #ifdef INET
1672         case AF_INET:
1673                 a += inaddr->addr32[0];
1674                 b += key->key32[1];
1675                 mix(a, b, c);
1676                 hash->addr32[0] = c + key->key32[2];
1677                 break;
1678 #endif /* INET */
1679 #ifdef INET6
1680         case AF_INET6:
1681                 a += inaddr->addr32[0];
1682                 b += inaddr->addr32[2];
1683                 mix(a, b, c);
1684                 hash->addr32[0] = c;
1685                 a += inaddr->addr32[1];
1686                 b += inaddr->addr32[3];
1687                 c += key->key32[1];
1688                 mix(a, b, c);
1689                 hash->addr32[1] = c;
1690                 a += inaddr->addr32[2];
1691                 b += inaddr->addr32[1];
1692                 c += key->key32[2];
1693                 mix(a, b, c);
1694                 hash->addr32[2] = c;
1695                 a += inaddr->addr32[3];
1696                 b += inaddr->addr32[0];
1697                 c += key->key32[3];
1698                 mix(a, b, c);
1699                 hash->addr32[3] = c;
1700                 break;
1701 #endif /* INET6 */
1702         }
1703 }
1704
1705 int
1706 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
1707     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
1708 {
1709         unsigned char            hash[16];
1710         struct pf_pool          *rpool = &r->rpool;
1711         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
1712         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
1713         struct pf_pooladdr      *acur = rpool->cur;
1714         struct pf_src_node       k;
1715
1716         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
1717             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1718                 k.af = af;
1719                 PF_ACPY(&k.addr, saddr, af);
1720                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
1721                     r->rpool.opts & PF_POOL_STICKYADDR)
1722                         k.rule.ptr = r;
1723                 else
1724                         k.rule.ptr = NULL;
1725                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
1726                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
1727                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
1728                         PF_ACPY(naddr, &(*sn)->raddr, af);
1729                         if (pf_status.debug >= PF_DEBUG_MISC) {
1730                                 kprintf("pf_map_addr: src tracking maps ");
1731                                 pf_print_host(&k.addr, 0, af);
1732                                 kprintf(" to ");
1733                                 pf_print_host(naddr, 0, af);
1734                                 kprintf("\n");
1735                         }
1736                         return (0);
1737                 }
1738         }
1739
1740         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
1741                 return (1);
1742         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1743                 if (af == AF_INET) {
1744                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
1745                             (rpool->opts & PF_POOL_TYPEMASK) !=
1746                             PF_POOL_ROUNDROBIN)
1747                                 return (1);
1748                          raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
1749                          rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
1750                 } else {
1751                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
1752                             (rpool->opts & PF_POOL_TYPEMASK) !=
1753                             PF_POOL_ROUNDROBIN)
1754                                 return (1);
1755                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
1756                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
1757                 }
1758         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1759                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
1760                         return (1); /* unsupported */
1761         } else {
1762                 raddr = &rpool->cur->addr.v.a.addr;
1763                 rmask = &rpool->cur->addr.v.a.mask;
1764         }
1765
1766         switch (rpool->opts & PF_POOL_TYPEMASK) {
1767         case PF_POOL_NONE:
1768                 PF_ACPY(naddr, raddr, af);
1769                 break;
1770         case PF_POOL_BITMASK:
1771                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
1772                 break;
1773         case PF_POOL_RANDOM:
1774                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
1775                         switch (af) {
1776 #ifdef INET
1777                         case AF_INET:
1778                                 rpool->counter.addr32[0] = karc4random();
1779                                 break;
1780 #endif /* INET */
1781 #ifdef INET6
1782                         case AF_INET6:
1783                                 if (rmask->addr32[3] != 0xffffffff)
1784                                         rpool->counter.addr32[3] = karc4random();
1785                                 else
1786                                         break;
1787                                 if (rmask->addr32[2] != 0xffffffff)
1788                                         rpool->counter.addr32[2] = karc4random();
1789                                 else
1790                                         break;
1791                                 if (rmask->addr32[1] != 0xffffffff)
1792                                         rpool->counter.addr32[1] = karc4random();
1793                                 else
1794                                         break;
1795                                 if (rmask->addr32[0] != 0xffffffff)
1796                                         rpool->counter.addr32[0] = karc4random();
1797                                 break;
1798 #endif /* INET6 */
1799                         }
1800                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1801                         PF_ACPY(init_addr, naddr, af);
1802
1803                 } else {
1804                         PF_AINC(&rpool->counter, af);
1805                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1806                 }
1807                 break;
1808         case PF_POOL_SRCHASH:
1809                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
1810                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
1811                 break;
1812         case PF_POOL_ROUNDROBIN:
1813                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1814                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
1815                             &rpool->tblidx, &rpool->counter,
1816                             &raddr, &rmask, af))
1817                                 goto get_addr;
1818                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1819                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
1820                             &rpool->tblidx, &rpool->counter,
1821                             &raddr, &rmask, af))
1822                                 goto get_addr;
1823                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
1824                         goto get_addr;
1825
1826         try_next:
1827                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
1828                         rpool->cur = TAILQ_FIRST(&rpool->list);
1829                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1830                         rpool->tblidx = -1;
1831                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
1832                             &rpool->tblidx, &rpool->counter,
1833                             &raddr, &rmask, af)) {
1834                                 /* table contains no address of type 'af' */
1835                                 if (rpool->cur != acur)
1836                                         goto try_next;
1837                                 return (1);
1838                         }
1839                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1840                         rpool->tblidx = -1;
1841                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
1842                             &rpool->tblidx, &rpool->counter,
1843                             &raddr, &rmask, af)) {
1844                                 /* table contains no address of type 'af' */
1845                                 if (rpool->cur != acur)
1846                                         goto try_next;
1847                                 return (1);
1848                         }
1849                 } else {
1850                         raddr = &rpool->cur->addr.v.a.addr;
1851                         rmask = &rpool->cur->addr.v.a.mask;
1852                         PF_ACPY(&rpool->counter, raddr, af);
1853                 }
1854
1855         get_addr:
1856                 PF_ACPY(naddr, &rpool->counter, af);
1857                 PF_AINC(&rpool->counter, af);
1858                 break;
1859         }
1860         if (*sn != NULL)
1861                 PF_ACPY(&(*sn)->raddr, naddr, af);
1862
1863         if (pf_status.debug >= PF_DEBUG_MISC &&
1864             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1865                 kprintf("pf_map_addr: selected address ");
1866                 pf_print_host(naddr, 0, af);
1867                 kprintf("\n");
1868         }
1869
1870         return (0);
1871 }
1872
1873 int
1874 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
1875     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
1876     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
1877     struct pf_src_node **sn)
1878 {
1879         struct pf_state         key;
1880         struct pf_addr          init_addr;
1881         u_int16_t               cut;
1882
1883         bzero(&init_addr, sizeof(init_addr));
1884         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
1885                 return (1);
1886
1887         do {
1888                 key.af = af;
1889                 key.proto = proto;
1890                 PF_ACPY(&key.ext.addr, daddr, key.af);
1891                 PF_ACPY(&key.gwy.addr, naddr, key.af);
1892                 key.ext.port = dport;
1893
1894                 /*
1895                  * port search; start random, step;
1896                  * similar 2 portloop in in_pcbbind
1897                  */
1898                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP)) {
1899                         key.gwy.port = 0;
1900                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
1901                                 return (0);
1902                 } else if (low == 0 && high == 0) {
1903                         key.gwy.port = *nport;
1904                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
1905                                 return (0);
1906                 } else if (low == high) {
1907                         key.gwy.port = htons(low);
1908                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
1909                                 *nport = htons(low);
1910                                 return (0);
1911                         }
1912                 } else {
1913                         u_int16_t tmp;
1914
1915                         if (low > high) {
1916                                 tmp = low;
1917                                 low = high;
1918                                 high = tmp;
1919                         }
1920                         /* low < high */
1921                         cut = karc4random() % (1 + high - low) + low;
1922                         /* low <= cut <= high */
1923                         for (tmp = cut; tmp <= high; ++(tmp)) {
1924                                 key.gwy.port = htons(tmp);
1925                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
1926                                     NULL) {
1927                                         *nport = htons(tmp);
1928                                         return (0);
1929                                 }
1930                         }
1931                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
1932                                 key.gwy.port = htons(tmp);
1933                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
1934                                     NULL) {
1935                                         *nport = htons(tmp);
1936                                         return (0);
1937                                 }
1938                         }
1939                 }
1940
1941                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
1942                 case PF_POOL_RANDOM:
1943                 case PF_POOL_ROUNDROBIN:
1944                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
1945                                 return (1);
1946                         break;
1947                 case PF_POOL_NONE:
1948                 case PF_POOL_SRCHASH:
1949                 case PF_POOL_BITMASK:
1950                 default:
1951                         return (1);
1952                 }
1953         } while (! PF_AEQ(&init_addr, naddr, af) );
1954
1955         return (1);                                     /* none available */
1956 }
1957
1958 struct pf_rule *
1959 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
1960     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
1961     struct pf_addr *daddr, u_int16_t dport, int rs_num)
1962 {
1963         struct pf_rule          *r, *rm = NULL, *anchorrule = NULL;
1964         struct pf_ruleset       *ruleset = NULL;
1965
1966         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
1967         while (r && rm == NULL) {
1968                 struct pf_rule_addr     *src = NULL, *dst = NULL;
1969                 struct pf_addr_wrap     *xdst = NULL;
1970
1971                 if (r->action == PF_BINAT && direction == PF_IN) {
1972                         src = &r->dst;
1973                         if (r->rpool.cur != NULL)
1974                                 xdst = &r->rpool.cur->addr;
1975                 } else {
1976                         src = &r->src;
1977                         dst = &r->dst;
1978                 }
1979
1980                 r->evaluations++;
1981                 if (r->kif != NULL &&
1982                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
1983                         r = r->skip[PF_SKIP_IFP].ptr;
1984                 else if (r->direction && r->direction != direction)
1985                         r = r->skip[PF_SKIP_DIR].ptr;
1986                 else if (r->af && r->af != pd->af)
1987                         r = r->skip[PF_SKIP_AF].ptr;
1988                 else if (r->proto && r->proto != pd->proto)
1989                         r = r->skip[PF_SKIP_PROTO].ptr;
1990                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->not))
1991                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
1992                             PF_SKIP_DST_ADDR].ptr;
1993                 else if (src->port_op && !pf_match_port(src->port_op,
1994                     src->port[0], src->port[1], sport))
1995                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
1996                             PF_SKIP_DST_PORT].ptr;
1997                 else if (dst != NULL &&
1998                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->not))
1999                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2000                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0))
2001                         r = TAILQ_NEXT(r, entries);
2002                 else if (dst != NULL && dst->port_op &&
2003                     !pf_match_port(dst->port_op, dst->port[0],
2004                     dst->port[1], dport))
2005                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2006                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2007                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2008                     off, pd->hdr.tcp), r->os_fingerprint)))
2009                         r = TAILQ_NEXT(r, entries);
2010                 else if (r->anchorname[0] && r->anchor == NULL)
2011                         r = TAILQ_NEXT(r, entries);
2012                 else if (r->anchor == NULL)
2013                                 rm = r;
2014                 else
2015                         PF_STEP_INTO_ANCHOR(r, anchorrule, ruleset, rs_num);
2016                 if (r == NULL && anchorrule != NULL)
2017                         PF_STEP_OUT_OF_ANCHOR(r, anchorrule, ruleset,
2018                             rs_num);
2019         }
2020         if (rm != NULL && (rm->action == PF_NONAT ||
2021             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2022                 return (NULL);
2023         return (rm);
2024 }
2025
2026 struct pf_rule *
2027 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2028     struct pfi_kif *kif, struct pf_src_node **sn,
2029     struct pf_addr *saddr, u_int16_t sport,
2030     struct pf_addr *daddr, u_int16_t dport,
2031     struct pf_addr *naddr, u_int16_t *nport)
2032 {
2033         struct pf_rule  *r = NULL;
2034
2035         if (direction == PF_OUT) {
2036                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2037                     sport, daddr, dport, PF_RULESET_BINAT);
2038                 if (r == NULL)
2039                         r = pf_match_translation(pd, m, off, direction, kif,
2040                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2041         } else {
2042                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2043                     sport, daddr, dport, PF_RULESET_RDR);
2044                 if (r == NULL)
2045                         r = pf_match_translation(pd, m, off, direction, kif,
2046                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2047         }
2048
2049         if (r != NULL) {
2050                 switch (r->action) {
2051                 case PF_NONAT:
2052                 case PF_NOBINAT:
2053                 case PF_NORDR:
2054                         return (NULL);
2055                 case PF_NAT:
2056                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2057                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2058                             r->rpool.proxy_port[1], sn)) {
2059                                 DPFPRINTF(PF_DEBUG_MISC,
2060                                     ("pf: NAT proxy port allocation "
2061                                     "(%u-%u) failed\n",
2062                                     r->rpool.proxy_port[0],
2063                                     r->rpool.proxy_port[1]));
2064                                 return (NULL);
2065                         }
2066                         break;
2067                 case PF_BINAT:
2068                         switch (direction) {
2069                         case PF_OUT:
2070                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2071                                         if (pd->af == AF_INET) {
2072                                                 if (r->rpool.cur->addr.p.dyn->
2073                                                     pfid_acnt4 < 1)
2074                                                         return (NULL);
2075                                                 PF_POOLMASK(naddr,
2076                                                     &r->rpool.cur->addr.p.dyn->
2077                                                     pfid_addr4,
2078                                                     &r->rpool.cur->addr.p.dyn->
2079                                                     pfid_mask4,
2080                                                     saddr, AF_INET);
2081                                         } else {
2082                                                 if (r->rpool.cur->addr.p.dyn->
2083                                                     pfid_acnt6 < 1)
2084                                                         return (NULL);
2085                                                 PF_POOLMASK(naddr,
2086                                                     &r->rpool.cur->addr.p.dyn->
2087                                                     pfid_addr6,
2088                                                     &r->rpool.cur->addr.p.dyn->
2089                                                     pfid_mask6,
2090                                                     saddr, AF_INET6);
2091                                         }
2092                                 } else
2093                                         PF_POOLMASK(naddr,
2094                                             &r->rpool.cur->addr.v.a.addr,
2095                                             &r->rpool.cur->addr.v.a.mask,
2096                                             saddr, pd->af);
2097                                 break;
2098                         case PF_IN:
2099                                 if (r->src.addr.type == PF_ADDR_DYNIFTL){
2100                                         if (pd->af == AF_INET) {
2101                                                 if (r->src.addr.p.dyn->
2102                                                     pfid_acnt4 < 1)
2103                                                         return (NULL);
2104                                                 PF_POOLMASK(naddr,
2105                                                     &r->src.addr.p.dyn->
2106                                                     pfid_addr4,
2107                                                     &r->src.addr.p.dyn->
2108                                                     pfid_mask4,
2109                                                     daddr, AF_INET);
2110                                         } else {
2111                                                 if (r->src.addr.p.dyn->
2112                                                     pfid_acnt6 < 1)
2113                                                         return (NULL);
2114                                                 PF_POOLMASK(naddr,
2115                                                     &r->src.addr.p.dyn->
2116                                                     pfid_addr6,
2117                                                     &r->src.addr.p.dyn->
2118                                                     pfid_mask6,
2119                                                     daddr, AF_INET6);
2120                                         }
2121                                 } else
2122                                         PF_POOLMASK(naddr,
2123                                             &r->src.addr.v.a.addr,
2124                                             &r->src.addr.v.a.mask, daddr,
2125                                             pd->af);
2126                                 break;
2127                         }
2128                         break;
2129                 case PF_RDR: {
2130                         if (pf_map_addr(r->af, r, saddr, naddr, NULL, sn))
2131                                 return (NULL);
2132
2133                         if (r->rpool.proxy_port[1]) {
2134                                 u_int32_t       tmp_nport;
2135
2136                                 tmp_nport = ((ntohs(dport) -
2137                                     ntohs(r->dst.port[0])) %
2138                                     (r->rpool.proxy_port[1] -
2139                                     r->rpool.proxy_port[0] + 1)) +
2140                                     r->rpool.proxy_port[0];
2141
2142                                 /* wrap around if necessary */
2143                                 if (tmp_nport > 65535)
2144                                         tmp_nport -= 65535;
2145                                 *nport = htons((u_int16_t)tmp_nport);
2146                         } else if (r->rpool.proxy_port[0])
2147                                 *nport = htons(r->rpool.proxy_port[0]);
2148                         break;
2149                 }
2150                 default:
2151                         return (NULL);
2152                 }
2153         }
2154
2155         return (r);
2156 }
2157
2158 #ifdef SMP
2159 struct netmsg_hashlookup {
2160         struct lwkt_msg         nm_lmsg;
2161         struct inpcb            **nm_pinp;
2162         struct inpcbinfo        *nm_pcbinfo;
2163         struct pf_addr          *nm_saddr;
2164         struct pf_addr          *nm_daddr;
2165         uint16_t                nm_sport;
2166         uint16_t                nm_dport;
2167         sa_family_t             nm_af;
2168 };
2169
2170 static int
2171 in_pcblookup_hash_handler(struct lwkt_msg *msg0)
2172 {
2173         struct netmsg_hashlookup *msg = (struct netmsg_hashlookup *)msg0;
2174
2175         if (msg->nm_af == AF_INET)
2176                 *msg->nm_pinp = in_pcblookup_hash(msg->nm_pcbinfo,
2177                     msg->nm_saddr->v4, msg->nm_sport, msg->nm_daddr->v4,
2178                     msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2179 #ifdef INET6
2180         else
2181                 *msg->nm_pinp = in6_pcblookup_hash(msg->nm_pcbinfo,
2182                     &msg->nm_saddr->v6, msg->nm_sport, &msg->nm_daddr->v6,
2183                     msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2184 #endif /* INET6 */
2185         lwkt_replymsg(&msg->nm_lmsg, 0);
2186         return (EASYNC);
2187 }
2188 #endif /* SMP */
2189
2190 int
2191 pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd)
2192 {
2193         struct pf_addr          *saddr, *daddr;
2194         u_int16_t                sport, dport;
2195         struct inpcbinfo        *pi;
2196         struct inpcb            *inp;
2197 #ifdef SMP
2198         struct netmsg_hashlookup *msg = NULL;
2199 #endif
2200         int                      pi_cpu = 0;
2201
2202         *uid = UID_MAX;
2203         *gid = GID_MAX;
2204         if (direction == PF_IN) {
2205                 saddr = pd->src;
2206                 daddr = pd->dst;
2207         } else {
2208                 saddr = pd->dst;
2209                 daddr = pd->src;
2210         }
2211         switch (pd->proto) {
2212         case IPPROTO_TCP:
2213                 sport = pd->hdr.tcp->th_sport;
2214                 dport = pd->hdr.tcp->th_dport;
2215
2216                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2217                 pi = &tcbinfo[pi_cpu];
2218 #ifdef SMP
2219                 /*
2220                  * Our netstack runs lockless on MP systems
2221                  * (only for TCP connections at the moment).
2222                  *
2223                  * As we are not allowed to read another CPU's tcbinfo,
2224                  * we have to ask that CPU via remote call to search the
2225                  * table for us.
2226                  *
2227                  * Prepare a msg iff data belongs to another CPU.
2228                  */
2229                 if (pi_cpu != mycpu->gd_cpuid) {
2230                         msg = kmalloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2231                         lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0,
2232                             lwkt_cmd_func(in_pcblookup_hash_handler),
2233                             lwkt_cmd_op_none);
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), &msg->nm_lmsg);
2288                 } else
2289 #endif /* SMP */
2290                 {
2291                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2292                             dport, INPLOOKUP_WILDCARD, NULL);
2293                 }
2294                 if (inp == NULL)
2295                         return (0);
2296                 break;
2297
2298         default:
2299                 return (0);
2300         }
2301         *uid = inp->inp_socket->so_cred->cr_uid;
2302         *gid = inp->inp_socket->so_cred->cr_groups[0];
2303         return (1);
2304 }
2305
2306 u_int8_t
2307 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2308 {
2309         int              hlen;
2310         u_int8_t         hdr[60];
2311         u_int8_t        *opt, optlen;
2312         u_int8_t         wscale = 0;
2313
2314         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
2315         if (hlen <= sizeof(struct tcphdr))
2316                 return (0);
2317         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2318                 return (0);
2319         opt = hdr + sizeof(struct tcphdr);
2320         hlen -= sizeof(struct tcphdr);
2321         while (hlen >= 3) {
2322                 switch (*opt) {
2323                 case TCPOPT_EOL:
2324                 case TCPOPT_NOP:
2325                         ++opt;
2326                         --hlen;
2327                         break;
2328                 case TCPOPT_WINDOW:
2329                         wscale = opt[2];
2330                         if (wscale > TCP_MAX_WINSHIFT)
2331                                 wscale = TCP_MAX_WINSHIFT;
2332                         wscale |= PF_WSCALE_FLAG;
2333                         /* FALLTHROUGH */
2334                 default:
2335                         optlen = opt[1];
2336                         if (optlen < 2)
2337                                 optlen = 2;
2338                         hlen -= optlen;
2339                         opt += optlen;
2340                         break;
2341                 }
2342         }
2343         return (wscale);
2344 }
2345
2346 u_int16_t
2347 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2348 {
2349         int              hlen;
2350         u_int8_t         hdr[60];
2351         u_int8_t        *opt, optlen;
2352         u_int16_t        mss = tcp_mssdflt;
2353
2354         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
2355         if (hlen <= sizeof(struct tcphdr))
2356                 return (0);
2357         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2358                 return (0);
2359         opt = hdr + sizeof(struct tcphdr);
2360         hlen -= sizeof(struct tcphdr);
2361         while (hlen >= TCPOLEN_MAXSEG) {
2362                 switch (*opt) {
2363                 case TCPOPT_EOL:
2364                 case TCPOPT_NOP:
2365                         ++opt;
2366                         --hlen;
2367                         break;
2368                 case TCPOPT_MAXSEG:
2369                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2370                         /* FALLTHROUGH */
2371                 default:
2372                         optlen = opt[1];
2373                         if (optlen < 2)
2374                                 optlen = 2;
2375                         hlen -= optlen;
2376                         opt += optlen;
2377                         break;
2378                 }
2379         }
2380         return (mss);
2381 }
2382
2383 u_int16_t
2384 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2385 {
2386 #ifdef INET
2387         struct sockaddr_in      *dst;
2388         struct route             ro;
2389 #endif /* INET */
2390 #ifdef INET6
2391         struct sockaddr_in6     *dst6;
2392         struct route_in6         ro6;
2393 #endif /* INET6 */
2394         struct rtentry          *rt = NULL;
2395         int                      hlen = 0;
2396         u_int16_t                mss = tcp_mssdflt;
2397
2398         switch (af) {
2399 #ifdef INET
2400         case AF_INET:
2401                 hlen = sizeof(struct ip);
2402                 bzero(&ro, sizeof(ro));
2403                 dst = (struct sockaddr_in *)&ro.ro_dst;
2404                 dst->sin_family = AF_INET;
2405                 dst->sin_len = sizeof(*dst);
2406                 dst->sin_addr = addr->v4;
2407                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2408                 rt = ro.ro_rt;
2409                 break;
2410 #endif /* INET */
2411 #ifdef INET6
2412         case AF_INET6:
2413                 hlen = sizeof(struct ip6_hdr);
2414                 bzero(&ro6, sizeof(ro6));
2415                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2416                 dst6->sin6_family = AF_INET6;
2417                 dst6->sin6_len = sizeof(*dst6);
2418                 dst6->sin6_addr = addr->v6;
2419                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
2420                 rt = ro6.ro_rt;
2421                 break;
2422 #endif /* INET6 */
2423         }
2424
2425         if (rt && rt->rt_ifp) {
2426                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2427                 mss = max(tcp_mssdflt, mss);
2428                 RTFREE(rt);
2429         }
2430         mss = min(mss, offer);
2431         mss = max(mss, 64);             /* sanity - at least max opt space */
2432         return (mss);
2433 }
2434
2435 void
2436 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2437 {
2438         struct pf_rule *r = s->rule.ptr;
2439
2440         s->rt_kif = NULL;
2441         if (!r->rt || r->rt == PF_FASTROUTE)
2442                 return;
2443         switch (s->af) {
2444 #ifdef INET
2445         case AF_INET:
2446                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2447                     &s->nat_src_node);
2448                 s->rt_kif = r->rpool.cur->kif;
2449                 break;
2450 #endif /* INET */
2451 #ifdef INET6
2452         case AF_INET6:
2453                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2454                     &s->nat_src_node);
2455                 s->rt_kif = r->rpool.cur->kif;
2456                 break;
2457 #endif /* INET6 */
2458         }
2459 }
2460
2461 int
2462 pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
2463     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2464     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2465 {
2466         struct pf_rule          *nr = NULL;
2467         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
2468         struct tcphdr           *th = pd->hdr.tcp;
2469         u_int16_t                bport, nport = 0;
2470         sa_family_t              af = pd->af;
2471         int                      lookup = -1;
2472         uid_t                    uid;
2473         gid_t                    gid;
2474         struct pf_rule          *r, *a = NULL;
2475         struct pf_ruleset       *ruleset = NULL;
2476         struct pf_src_node      *nsn = NULL;
2477         u_short                  reason;
2478         int                      rewrite = 0;
2479         int                      tag = -1;
2480         u_int16_t                mss = tcp_mssdflt;
2481
2482         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2483
2484         if (direction == PF_OUT) {
2485                 bport = nport = th->th_sport;
2486                 /* check outgoing packet for BINAT/NAT */
2487                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2488                     saddr, th->th_sport, daddr, th->th_dport,
2489                     &pd->naddr, &nport)) != NULL) {
2490                         PF_ACPY(&pd->baddr, saddr, af);
2491                         pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2492                             &th->th_sum, &pd->naddr, nport, 0, af);
2493                         rewrite++;
2494                         if (nr->natpass)
2495                                 r = NULL;
2496                         pd->nat_rule = nr;
2497                 }
2498         } else {
2499                 bport = nport = th->th_dport;
2500                 /* check incoming packet for BINAT/RDR */
2501                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2502                     saddr, th->th_sport, daddr, th->th_dport,
2503                     &pd->naddr, &nport)) != NULL) {
2504                         PF_ACPY(&pd->baddr, daddr, af);
2505                         pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2506                             &th->th_sum, &pd->naddr, nport, 0, af);
2507                         rewrite++;
2508                         if (nr->natpass)
2509                                 r = NULL;
2510                         pd->nat_rule = nr;
2511                 }
2512         }
2513
2514         while (r != NULL) {
2515                 r->evaluations++;
2516                 if (r->kif != NULL &&
2517                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2518                         r = r->skip[PF_SKIP_IFP].ptr;
2519                 else if (r->direction && r->direction != direction)
2520                         r = r->skip[PF_SKIP_DIR].ptr;
2521                 else if (r->af && r->af != af)
2522                         r = r->skip[PF_SKIP_AF].ptr;
2523                 else if (r->proto && r->proto != IPPROTO_TCP)
2524                         r = r->skip[PF_SKIP_PROTO].ptr;
2525                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2526                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2527                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2528                     r->src.port[0], r->src.port[1], th->th_sport))
2529                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
2530                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2531                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2532                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2533                     r->dst.port[0], r->dst.port[1], th->th_dport))
2534                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2535                 else if (r->tos && !(r->tos & pd->tos))
2536                         r = TAILQ_NEXT(r, entries);
2537                 else if (r->rule_flag & PFRULE_FRAGMENT)
2538                         r = TAILQ_NEXT(r, entries);
2539                 else if ((r->flagset & th->th_flags) != r->flags)
2540                         r = TAILQ_NEXT(r, entries);
2541                 else if (r->uid.op && (lookup != -1 || (lookup =
2542                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2543                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2544                     uid))
2545                         r = TAILQ_NEXT(r, entries);
2546                 else if (r->gid.op && (lookup != -1 || (lookup =
2547                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2548                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2549                     gid))
2550                         r = TAILQ_NEXT(r, entries);
2551                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2552                         r = TAILQ_NEXT(r, entries);
2553                 else if (r->anchorname[0] && r->anchor == NULL)
2554                         r = TAILQ_NEXT(r, entries);
2555                 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
2556                     pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
2557                         r = TAILQ_NEXT(r, entries);
2558                 else {
2559                         if (r->tag)
2560                                 tag = r->tag;
2561                         if (r->anchor == NULL) {
2562                                 *rm = r;
2563                                 *am = a;
2564                                 *rsm = ruleset;
2565                                 if ((*rm)->quick)
2566                                         break;
2567                                 r = TAILQ_NEXT(r, entries);
2568                         } else
2569                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2570                                     PF_RULESET_FILTER);
2571                 }
2572                 if (r == NULL && a != NULL)
2573                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2574                             PF_RULESET_FILTER);
2575         }
2576         r = *rm;
2577         a = *am;
2578         ruleset = *rsm;
2579
2580         REASON_SET(&reason, PFRES_MATCH);
2581
2582         if (r->log) {
2583                 if (rewrite)
2584                         m_copyback(m, off, sizeof(*th), (caddr_t)th);
2585                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2586         }
2587
2588         if ((r->action == PF_DROP) &&
2589             ((r->rule_flag & PFRULE_RETURNRST) ||
2590             (r->rule_flag & PFRULE_RETURNICMP) ||
2591             (r->rule_flag & PFRULE_RETURN))) {
2592                 /* undo NAT changes, if they have taken place */
2593                 if (nr != NULL) {
2594                         if (direction == PF_OUT) {
2595                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2596                                     &th->th_sum, &pd->baddr, bport, 0, af);
2597                                 rewrite++;
2598                         } else {
2599                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2600                                     &th->th_sum, &pd->baddr, bport, 0, af);
2601                                 rewrite++;
2602                         }
2603                 }
2604                 if (((r->rule_flag & PFRULE_RETURNRST) ||
2605                     (r->rule_flag & PFRULE_RETURN)) &&
2606                     !(th->th_flags & TH_RST)) {
2607                         u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
2608
2609                         if (th->th_flags & TH_SYN)
2610                                 ack++;
2611                         if (th->th_flags & TH_FIN)
2612                                 ack++;
2613                         pf_send_tcp(r, af, pd->dst,
2614                             pd->src, th->th_dport, th->th_sport,
2615                             ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
2616                             r->return_ttl);
2617                 } else if ((af == AF_INET) && r->return_icmp)
2618                         pf_send_icmp(m, r->return_icmp >> 8,
2619                             r->return_icmp & 255, af, r);
2620                 else if ((af == AF_INET6) && r->return_icmp6)
2621                         pf_send_icmp(m, r->return_icmp6 >> 8,
2622                             r->return_icmp6 & 255, af, r);
2623         }
2624
2625         if (r->action == PF_DROP)
2626                 return (PF_DROP);
2627
2628         pf_tag_packet(m, tag);
2629
2630         if (r->keep_state || nr != NULL ||
2631             (pd->flags & PFDESC_TCP_NORM)) {
2632                 /* create new state */
2633                 u_int16_t        len;
2634                 struct pf_state *s = NULL;
2635                 struct pf_src_node *sn = NULL;
2636
2637                 len = pd->tot_len - off - (th->th_off << 2);
2638
2639                 /* check maximums */
2640                 if (r->max_states && (r->states >= r->max_states))
2641                         goto cleanup;
2642                 /* src node for flter rule */
2643                 if ((r->rule_flag & PFRULE_SRCTRACK ||
2644                     r->rpool.opts & PF_POOL_STICKYADDR) &&
2645                     pf_insert_src_node(&sn, r, saddr, af) != 0)
2646                         goto cleanup;
2647                 /* src node for translation rule */
2648                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2649                     ((direction == PF_OUT &&
2650                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2651                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2652                         goto cleanup;
2653                 s = pool_get(&pf_state_pl, PR_NOWAIT);
2654                 if (s == NULL) {
2655 cleanup:
2656                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2657                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2658                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2659                                 pf_status.src_nodes--;
2660                                 pool_put(&pf_src_tree_pl, sn);
2661                         }
2662                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2663                             nsn->expire == 0) {
2664                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2665                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2666                                 pf_status.src_nodes--;
2667                                 pool_put(&pf_src_tree_pl, nsn);
2668                         }
2669                         REASON_SET(&reason, PFRES_MEMORY);
2670                         return (PF_DROP);
2671                 }
2672                 bzero(s, sizeof(*s));
2673                 r->states++;
2674                 if (a != NULL)
2675                         a->states++;
2676                 s->rule.ptr = r;
2677                 s->nat_rule.ptr = nr;
2678                 if (s->nat_rule.ptr != NULL)
2679                         s->nat_rule.ptr->states++;
2680                 s->anchor.ptr = a;
2681                 s->allow_opts = r->allow_opts;
2682                 s->log = r->log & 2;
2683                 s->proto = IPPROTO_TCP;
2684                 s->direction = direction;
2685                 s->af = af;
2686                 if (direction == PF_OUT) {
2687                         PF_ACPY(&s->gwy.addr, saddr, af);
2688                         s->gwy.port = th->th_sport;             /* sport */
2689                         PF_ACPY(&s->ext.addr, daddr, af);
2690                         s->ext.port = th->th_dport;
2691                         if (nr != NULL) {
2692                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
2693                                 s->lan.port = bport;
2694                         } else {
2695                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2696                                 s->lan.port = s->gwy.port;
2697                         }
2698                 } else {
2699                         PF_ACPY(&s->lan.addr, daddr, af);
2700                         s->lan.port = th->th_dport;
2701                         PF_ACPY(&s->ext.addr, saddr, af);
2702                         s->ext.port = th->th_sport;
2703                         if (nr != NULL) {
2704                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
2705                                 s->gwy.port = bport;
2706                         } else {
2707                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2708                                 s->gwy.port = s->lan.port;
2709                         }
2710                 }
2711
2712                 s->src.seqlo = ntohl(th->th_seq);
2713                 s->src.seqhi = s->src.seqlo + len + 1;
2714                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2715                     r->keep_state == PF_STATE_MODULATE) {
2716                         /* Generate sequence number modulator */
2717                         while ((s->src.seqdiff = karc4random()) == 0)
2718                                 ;
2719                         pf_change_a(&th->th_seq, &th->th_sum,
2720                             htonl(s->src.seqlo + s->src.seqdiff), 0);
2721                         rewrite = 1;
2722                 } else
2723                         s->src.seqdiff = 0;
2724                 if (th->th_flags & TH_SYN) {
2725                         s->src.seqhi++;
2726                         s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
2727                 }
2728                 s->src.max_win = MAX(ntohs(th->th_win), 1);
2729                 if (s->src.wscale & PF_WSCALE_MASK) {
2730                         /* Remove scale factor from initial window */
2731                         int win = s->src.max_win;
2732                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
2733                         s->src.max_win = (win - 1) >>
2734                             (s->src.wscale & PF_WSCALE_MASK);
2735                 }
2736                 if (th->th_flags & TH_FIN)
2737                         s->src.seqhi++;
2738                 s->dst.seqhi = 1;
2739                 s->dst.max_win = 1;
2740                 s->src.state = TCPS_SYN_SENT;
2741                 s->dst.state = TCPS_CLOSED;
2742                 s->creation = time_second;
2743                 s->expire = time_second;
2744                 s->timeout = PFTM_TCP_FIRST_PACKET;
2745                 pf_set_rt_ifp(s, saddr);
2746                 if (sn != NULL) {
2747                         s->src_node = sn;
2748                         s->src_node->states++;
2749                 }
2750                 if (nsn != NULL) {
2751                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
2752                         s->nat_src_node = nsn;
2753                         s->nat_src_node->states++;
2754                 }
2755                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
2756                     off, pd, th, &s->src, &s->dst)) {
2757                         REASON_SET(&reason, PFRES_MEMORY);
2758                         pf_src_tree_remove_state(s);
2759                         pool_put(&pf_state_pl, s);
2760                         return (PF_DROP);
2761                 }
2762                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
2763                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, &s->src,
2764                     &s->dst, &rewrite)) {
2765                         pf_normalize_tcp_cleanup(s);
2766                         pf_src_tree_remove_state(s);
2767                         pool_put(&pf_state_pl, s);
2768                         return (PF_DROP);
2769                 }
2770                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
2771                         pf_normalize_tcp_cleanup(s);
2772                         REASON_SET(&reason, PFRES_MEMORY);
2773                         pf_src_tree_remove_state(s);
2774                         pool_put(&pf_state_pl, s);
2775                         return (PF_DROP);
2776                 } else
2777                         *sm = s;
2778                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2779                     r->keep_state == PF_STATE_SYNPROXY) {
2780                         s->src.state = PF_TCPS_PROXY_SRC;
2781                         if (nr != NULL) {
2782                                 if (direction == PF_OUT) {
2783                                         pf_change_ap(saddr, &th->th_sport,
2784                                             pd->ip_sum, &th->th_sum, &pd->baddr,
2785                                             bport, 0, af);
2786                                 } else {
2787                                         pf_change_ap(daddr, &th->th_dport,
2788                                             pd->ip_sum, &th->th_sum, &pd->baddr,
2789                                             bport, 0, af);
2790                                 }
2791                         }
2792                         s->src.seqhi = karc4random();
2793                         /* Find mss option */
2794                         mss = pf_get_mss(m, off, th->th_off, af);
2795                         mss = pf_calc_mss(saddr, af, mss);
2796                         mss = pf_calc_mss(daddr, af, mss);
2797                         s->src.mss = mss;
2798                         pf_send_tcp(r, af, daddr, saddr, th->th_dport,
2799                             th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
2800                             TH_SYN|TH_ACK, 0, s->src.mss, 0);
2801                         return (PF_SYNPROXY_DROP);
2802                 }
2803         }
2804
2805         /* copy back packet headers if we performed NAT operations */
2806         if (rewrite)
2807                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2808
2809         return (PF_PASS);
2810 }
2811
2812 int
2813 pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
2814     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2815     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2816 {
2817         struct pf_rule          *nr = NULL;
2818         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
2819         struct udphdr           *uh = pd->hdr.udp;
2820         u_int16_t                bport, nport = 0;
2821         sa_family_t              af = pd->af;
2822         int                      lookup = -1;
2823         uid_t                    uid;
2824         gid_t                    gid;
2825         struct pf_rule          *r, *a = NULL;
2826         struct pf_ruleset       *ruleset = NULL;
2827         struct pf_src_node      *nsn = NULL;
2828         u_short                  reason;
2829         int                      rewrite = 0;
2830         int                      tag = -1;
2831
2832         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2833
2834         if (direction == PF_OUT) {
2835                 bport = nport = uh->uh_sport;
2836                 /* check outgoing packet for BINAT/NAT */
2837                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2838                     saddr, uh->uh_sport, daddr, uh->uh_dport,
2839                     &pd->naddr, &nport)) != NULL) {
2840                         PF_ACPY(&pd->baddr, saddr, af);
2841                         pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2842                             &uh->uh_sum, &pd->naddr, nport, 1, af);
2843                         rewrite++;
2844                         if (nr->natpass)
2845                                 r = NULL;
2846                         pd->nat_rule = nr;
2847                 }
2848         } else {
2849                 bport = nport = uh->uh_dport;
2850                 /* check incoming packet for BINAT/RDR */
2851                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2852                     saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
2853                     &nport)) != NULL) {
2854                         PF_ACPY(&pd->baddr, daddr, af);
2855                         pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2856                             &uh->uh_sum, &pd->naddr, nport, 1, af);
2857                         rewrite++;
2858                         if (nr->natpass)
2859                                 r = NULL;
2860                         pd->nat_rule = nr;
2861                 }
2862         }
2863
2864         while (r != NULL) {
2865                 r->evaluations++;
2866                 if (r->kif != NULL &&
2867                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2868                         r = r->skip[PF_SKIP_IFP].ptr;
2869                 else if (r->direction && r->direction != direction)
2870                         r = r->skip[PF_SKIP_DIR].ptr;
2871                 else if (r->af && r->af != af)
2872                         r = r->skip[PF_SKIP_AF].ptr;
2873                 else if (r->proto && r->proto != IPPROTO_UDP)
2874                         r = r->skip[PF_SKIP_PROTO].ptr;
2875                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2876                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2877                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2878                     r->src.port[0], r->src.port[1], uh->uh_sport))
2879                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
2880                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2881                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2882                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2883                     r->dst.port[0], r->dst.port[1], uh->uh_dport))
2884                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2885                 else if (r->tos && !(r->tos & pd->tos))
2886                         r = TAILQ_NEXT(r, entries);
2887                 else if (r->rule_flag & PFRULE_FRAGMENT)
2888                         r = TAILQ_NEXT(r, entries);
2889                 else if (r->uid.op && (lookup != -1 || (lookup =
2890                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2891                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2892                     uid))
2893                         r = TAILQ_NEXT(r, entries);
2894                 else if (r->gid.op && (lookup != -1 || (lookup =
2895                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2896                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2897                     gid))
2898                         r = TAILQ_NEXT(r, entries);
2899                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2900                         r = TAILQ_NEXT(r, entries);
2901                 else if (r->anchorname[0] && r->anchor == NULL)
2902                         r = TAILQ_NEXT(r, entries);
2903                 else if (r->os_fingerprint != PF_OSFP_ANY)
2904                         r = TAILQ_NEXT(r, entries);
2905                 else {
2906                         if (r->tag)
2907                                 tag = r->tag;
2908                         if (r->anchor == NULL) {
2909                                 *rm = r;
2910                                 *am = a;
2911                                 *rsm = ruleset;
2912                                 if ((*rm)->quick)
2913                                         break;
2914                                 r = TAILQ_NEXT(r, entries);
2915                         } else
2916                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2917                                     PF_RULESET_FILTER);
2918                 }
2919                 if (r == NULL && a != NULL)
2920                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2921                             PF_RULESET_FILTER);
2922         }
2923         r = *rm;
2924         a = *am;
2925         ruleset = *rsm;
2926
2927         REASON_SET(&reason, PFRES_MATCH);
2928
2929         if (r->log) {
2930                 if (rewrite)
2931                         m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
2932                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2933         }
2934
2935         if ((r->action == PF_DROP) &&
2936             ((r->rule_flag & PFRULE_RETURNICMP) ||
2937             (r->rule_flag & PFRULE_RETURN))) {
2938                 /* undo NAT changes, if they have taken place */
2939                 if (nr != NULL) {
2940                         if (direction == PF_OUT) {
2941                                 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2942                                     &uh->uh_sum, &pd->baddr, bport, 1, af);
2943                                 rewrite++;
2944                         } else {
2945                                 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2946                                     &uh->uh_sum, &pd->baddr, bport, 1, af);
2947                                 rewrite++;
2948                         }
2949                 }
2950                 if ((af == AF_INET) && r->return_icmp)
2951                         pf_send_icmp(m, r->return_icmp >> 8,
2952                             r->return_icmp & 255, af, r);
2953                 else if ((af == AF_INET6) && r->return_icmp6)
2954                         pf_send_icmp(m, r->return_icmp6 >> 8,
2955                             r->return_icmp6 & 255, af, r);
2956         }
2957
2958         if (r->action == PF_DROP)
2959                 return (PF_DROP);
2960
2961         pf_tag_packet(m, tag);
2962
2963         if (r->keep_state || nr != NULL) {
2964                 /* create new state */
2965                 struct pf_state *s = NULL;
2966                 struct pf_src_node *sn = NULL;
2967
2968                 /* check maximums */
2969                 if (r->max_states && (r->states >= r->max_states))
2970                         goto cleanup;
2971                 /* src node for flter rule */
2972                 if ((r->rule_flag & PFRULE_SRCTRACK ||
2973                     r->rpool.opts & PF_POOL_STICKYADDR) &&
2974                     pf_insert_src_node(&sn, r, saddr, af) != 0)
2975                         goto cleanup;
2976                 /* src node for translation rule */
2977                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2978                     ((direction == PF_OUT &&
2979                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2980                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2981                         goto cleanup;
2982                 s = pool_get(&pf_state_pl, PR_NOWAIT);
2983                 if (s == NULL) {
2984 cleanup:
2985                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2986                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2987                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2988                                 pf_status.src_nodes--;
2989                                 pool_put(&pf_src_tree_pl, sn);
2990                         }
2991                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2992                             nsn->expire == 0) {
2993                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2994                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2995                                 pf_status.src_nodes--;
2996                                 pool_put(&pf_src_tree_pl, nsn);
2997                         }
2998                         REASON_SET(&reason, PFRES_MEMORY);
2999                         return (PF_DROP);
3000                 }
3001                 bzero(s, sizeof(*s));
3002                 r->states++;
3003                 if (a != NULL)
3004                         a->states++;
3005                 s->rule.ptr = r;
3006                 s->nat_rule.ptr = nr;
3007                 if (s->nat_rule.ptr != NULL)
3008                         s->nat_rule.ptr->states++;
3009                 s->anchor.ptr = a;
3010                 s->allow_opts = r->allow_opts;
3011                 s->log = r->log & 2;
3012                 s->proto = IPPROTO_UDP;
3013                 s->direction = direction;
3014                 s->af = af;
3015                 if (direction == PF_OUT) {
3016                         PF_ACPY(&s->gwy.addr, saddr, af);
3017                         s->gwy.port = uh->uh_sport;
3018                         PF_ACPY(&s->ext.addr, daddr, af);
3019                         s->ext.port = uh->uh_dport;
3020                         if (nr != NULL) {
3021                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3022                                 s->lan.port = bport;
3023                         } else {
3024                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3025                                 s->lan.port = s->gwy.port;
3026                         }
3027                 } else {
3028                         PF_ACPY(&s->lan.addr, daddr, af);
3029                         s->lan.port = uh->uh_dport;
3030                         PF_ACPY(&s->ext.addr, saddr, af);
3031                         s->ext.port = uh->uh_sport;
3032                         if (nr != NULL) {
3033                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3034                                 s->gwy.port = bport;
3035                         } else {
3036                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3037                                 s->gwy.port = s->lan.port;
3038                         }
3039                 }
3040                 s->src.state = PFUDPS_SINGLE;
3041                 s->dst.state = PFUDPS_NO_TRAFFIC;
3042                 s->creation = time_second;
3043                 s->expire = time_second;
3044                 s->timeout = PFTM_UDP_FIRST_PACKET;
3045                 pf_set_rt_ifp(s, saddr);
3046                 if (sn != NULL) {
3047                         s->src_node = sn;
3048                         s->src_node->states++;
3049                 }
3050                 if (nsn != NULL) {
3051                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3052                         s->nat_src_node = nsn;
3053                         s->nat_src_node->states++;
3054                 }
3055                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3056                         REASON_SET(&reason, PFRES_MEMORY);
3057                         pf_src_tree_remove_state(s);
3058                         pool_put(&pf_state_pl, s);
3059                         return (PF_DROP);
3060                 } else
3061                         *sm = s;
3062         }
3063
3064         /* copy back packet headers if we performed NAT operations */
3065         if (rewrite)
3066                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3067
3068         return (PF_PASS);
3069 }
3070
3071 int
3072 pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3073     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3074     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3075 {
3076         struct pf_rule          *nr = NULL;
3077         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3078         struct pf_rule          *r, *a = NULL;
3079         struct pf_ruleset       *ruleset = NULL;
3080         struct pf_src_node      *nsn = NULL;
3081         u_short                  reason;
3082         u_int16_t                icmpid = 0;
3083         sa_family_t              af = pd->af;
3084         u_int8_t                 icmptype = 0, icmpcode = 0;
3085         int                      state_icmp = 0;
3086         int                      tag = -1;
3087 #ifdef INET6
3088         int                      rewrite = 0;
3089 #endif /* INET6 */
3090
3091         switch (pd->proto) {
3092 #ifdef INET
3093         case IPPROTO_ICMP:
3094                 icmptype = pd->hdr.icmp->icmp_type;
3095                 icmpcode = pd->hdr.icmp->icmp_code;
3096                 icmpid = pd->hdr.icmp->icmp_id;
3097
3098                 if (icmptype == ICMP_UNREACH ||
3099                     icmptype == ICMP_SOURCEQUENCH ||
3100                     icmptype == ICMP_REDIRECT ||
3101                     icmptype == ICMP_TIMXCEED ||
3102                     icmptype == ICMP_PARAMPROB)
3103                         state_icmp++;
3104                 break;
3105 #endif /* INET */
3106 #ifdef INET6
3107         case IPPROTO_ICMPV6:
3108                 icmptype = pd->hdr.icmp6->icmp6_type;
3109                 icmpcode = pd->hdr.icmp6->icmp6_code;
3110                 icmpid = pd->hdr.icmp6->icmp6_id;
3111
3112                 if (icmptype == ICMP6_DST_UNREACH ||
3113                     icmptype == ICMP6_PACKET_TOO_BIG ||
3114                     icmptype == ICMP6_TIME_EXCEEDED ||
3115                     icmptype == ICMP6_PARAM_PROB)
3116                         state_icmp++;
3117                 break;
3118 #endif /* INET6 */
3119         }
3120
3121         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3122
3123         if (direction == PF_OUT) {
3124                 /* check outgoing packet for BINAT/NAT */
3125                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3126                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3127                         PF_ACPY(&pd->baddr, saddr, af);
3128                         switch (af) {
3129 #ifdef INET
3130                         case AF_INET:
3131                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3132                                     pd->naddr.v4.s_addr, 0);
3133                                 break;
3134 #endif /* INET */
3135 #ifdef INET6
3136                         case AF_INET6:
3137                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3138                                     &pd->naddr, 0);
3139                                 rewrite++;
3140                                 break;
3141 #endif /* INET6 */
3142                         }
3143                         if (nr->natpass)
3144                                 r = NULL;
3145                         pd->nat_rule = nr;
3146                 }
3147         } else {
3148                 /* check incoming packet for BINAT/RDR */
3149                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3150                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3151                         PF_ACPY(&pd->baddr, daddr, af);
3152                         switch (af) {
3153 #ifdef INET
3154                         case AF_INET:
3155                                 pf_change_a(&daddr->v4.s_addr,
3156                                     pd->ip_sum, pd->naddr.v4.s_addr, 0);
3157                                 break;
3158 #endif /* INET */
3159 #ifdef INET6
3160                         case AF_INET6:
3161                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3162                                     &pd->naddr, 0);
3163                                 rewrite++;
3164                                 break;
3165 #endif /* INET6 */
3166                         }
3167                         if (nr->natpass)
3168                                 r = NULL;
3169                         pd->nat_rule = nr;
3170                 }
3171         }
3172
3173         while (r != NULL) {
3174                 r->evaluations++;
3175                 if (r->kif != NULL &&
3176                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3177                         r = r->skip[PF_SKIP_IFP].ptr;
3178                 else if (r->direction && r->direction != direction)
3179                         r = r->skip[PF_SKIP_DIR].ptr;
3180                 else if (r->af && r->af != af)
3181                         r = r->skip[PF_SKIP_AF].ptr;
3182                 else if (r->proto && r->proto != pd->proto)
3183                         r = r->skip[PF_SKIP_PROTO].ptr;
3184                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3185                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3186                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3187                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3188                 else if (r->type && r->type != icmptype + 1)
3189                         r = TAILQ_NEXT(r, entries);
3190                 else if (r->code && r->code != icmpcode + 1)
3191                         r = TAILQ_NEXT(r, entries);
3192                 else if (r->tos && !(r->tos & pd->tos))
3193                         r = TAILQ_NEXT(r, entries);
3194                 else if (r->rule_flag & PFRULE_FRAGMENT)
3195                         r = TAILQ_NEXT(r, entries);
3196                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3197                         r = TAILQ_NEXT(r, entries);
3198                 else if (r->anchorname[0] && r->anchor == NULL)
3199                         r = TAILQ_NEXT(r, entries);
3200                 else if (r->os_fingerprint != PF_OSFP_ANY)
3201                         r = TAILQ_NEXT(r, entries);
3202                 else {
3203                         if (r->tag)
3204                                 tag = r->tag;
3205                         if (r->anchor == NULL) {
3206                                 *rm = r;
3207                                 *am = a;
3208                                 *rsm = ruleset;
3209                                 if ((*rm)->quick)
3210                                         break;
3211                                 r = TAILQ_NEXT(r, entries);
3212                         } else
3213                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3214                                     PF_RULESET_FILTER);
3215                 }
3216                 if (r == NULL && a != NULL)
3217                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3218                             PF_RULESET_FILTER);
3219         }
3220         r = *rm;
3221         a = *am;
3222         ruleset = *rsm;
3223
3224         REASON_SET(&reason, PFRES_MATCH);
3225
3226         if (r->log) {
3227 #ifdef INET6
3228                 if (rewrite)
3229                         m_copyback(m, off, sizeof(struct icmp6_hdr),
3230                             (caddr_t)pd->hdr.icmp6);
3231 #endif /* INET6 */
3232                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3233         }
3234
3235         if (r->action != PF_PASS)
3236                 return (PF_DROP);
3237
3238         pf_tag_packet(m, tag);
3239
3240         if (!state_icmp && (r->keep_state || nr != NULL)) {
3241                 /* create new state */
3242                 struct pf_state *s = NULL;
3243                 struct pf_src_node *sn = NULL;
3244
3245                 /* check maximums */
3246                 if (r->max_states && (r->states >= r->max_states))
3247                         goto cleanup;
3248                 /* src node for flter rule */
3249                 if ((r->rule_flag & PFRULE_SRCTRACK ||
3250                     r->rpool.opts & PF_POOL_STICKYADDR) &&
3251                     pf_insert_src_node(&sn, r, saddr, af) != 0)
3252                         goto cleanup;
3253                 /* src node for translation rule */
3254                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3255                     ((direction == PF_OUT &&
3256                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3257                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3258                         goto cleanup;
3259                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3260                 if (s == NULL) {
3261 cleanup:
3262                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3263                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3264                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3265                                 pf_status.src_nodes--;
3266                                 pool_put(&pf_src_tree_pl, sn);
3267                         }
3268                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3269                             nsn->expire == 0) {
3270                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3271                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3272                                 pf_status.src_nodes--;
3273                                 pool_put(&pf_src_tree_pl, nsn);
3274                         }
3275                         REASON_SET(&reason, PFRES_MEMORY);
3276                         return (PF_DROP);
3277                 }
3278                 bzero(s, sizeof(*s));
3279                 r->states++;
3280                 if (a != NULL)
3281                         a->states++;
3282                 s->rule.ptr = r;
3283                 s->nat_rule.ptr = nr;
3284                 if (s->nat_rule.ptr != NULL)
3285                         s->nat_rule.ptr->states++;
3286                 s->anchor.ptr = a;
3287                 s->allow_opts = r->allow_opts;
3288                 s->log = r->log & 2;
3289                 s->proto = pd->proto;
3290                 s->direction = direction;
3291                 s->af = af;
3292                 if (direction == PF_OUT) {
3293                         PF_ACPY(&s->gwy.addr, saddr, af);
3294                         s->gwy.port = icmpid;
3295                         PF_ACPY(&s->ext.addr, daddr, af);
3296                         s->ext.port = icmpid;
3297                         if (nr != NULL)
3298                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3299                         else
3300                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3301                         s->lan.port = icmpid;
3302                 } else {
3303                         PF_ACPY(&s->lan.addr, daddr, af);
3304                         s->lan.port = icmpid;
3305                         PF_ACPY(&s->ext.addr, saddr, af);
3306                         s->ext.port = icmpid;
3307                         if (nr != NULL)
3308                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3309                         else
3310                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3311                         s->gwy.port = icmpid;
3312                 }
3313                 s->creation = time_second;
3314                 s->expire = time_second;
3315                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3316                 pf_set_rt_ifp(s, saddr);
3317                 if (sn != NULL) {
3318                         s->src_node = sn;
3319                         s->src_node->states++;
3320                 }
3321                 if (nsn != NULL) {
3322                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3323                         s->nat_src_node = nsn;
3324                         s->nat_src_node->states++;
3325                 }
3326                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3327                         REASON_SET(&reason, PFRES_MEMORY);
3328                         pf_src_tree_remove_state(s);
3329                         pool_put(&pf_state_pl, s);
3330                         return (PF_DROP);
3331                 } else
3332                         *sm = s;
3333         }
3334
3335 #ifdef INET6
3336         /* copy back packet headers if we performed IPv6 NAT operations */
3337         if (rewrite)
3338                 m_copyback(m, off, sizeof(struct icmp6_hdr),
3339                     (caddr_t)pd->hdr.icmp6);
3340 #endif /* INET6 */
3341
3342         return (PF_PASS);
3343 }
3344
3345 int
3346 pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3347     struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3348     struct pf_rule **am, struct pf_ruleset **rsm)
3349 {
3350         struct pf_rule          *nr = NULL;
3351         struct pf_rule          *r, *a = NULL;
3352         struct pf_ruleset       *ruleset = NULL;
3353         struct pf_src_node      *nsn = NULL;
3354         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3355         sa_family_t              af = pd->af;
3356         u_short                  reason;
3357         int                      tag = -1;
3358
3359         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3360
3361         if (direction == PF_OUT) {
3362                 /* check outgoing packet for BINAT/NAT */
3363                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3364                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3365                         PF_ACPY(&pd->baddr, saddr, af);
3366                         switch (af) {
3367 #ifdef INET
3368                         case AF_INET:
3369                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3370                                     pd->naddr.v4.s_addr, 0);
3371                                 break;
3372 #endif /* INET */
3373 #ifdef INET6
3374                         case AF_INET6:
3375                                 PF_ACPY(saddr, &pd->naddr, af);
3376                                 break;
3377 #endif /* INET6 */
3378                         }
3379                         if (nr->natpass)
3380                                 r = NULL;
3381                         pd->nat_rule = nr;
3382                 }
3383         } else {
3384                 /* check incoming packet for BINAT/RDR */
3385                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3386                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3387                         PF_ACPY(&pd->baddr, daddr, af);
3388                         switch (af) {
3389 #ifdef INET
3390                         case AF_INET:
3391                                 pf_change_a(&daddr->v4.s_addr,
3392                                     pd->ip_sum, pd->naddr.v4.s_addr, 0);
3393                                 break;
3394 #endif /* INET */
3395 #ifdef INET6
3396                         case AF_INET6:
3397                                 PF_ACPY(daddr, &pd->naddr, af);
3398                                 break;
3399 #endif /* INET6 */
3400                         }
3401                         if (nr->natpass)
3402                                 r = NULL;
3403                         pd->nat_rule = nr;
3404                 }
3405         }
3406
3407         while (r != NULL) {
3408                 r->evaluations++;
3409                 if (r->kif != NULL &&
3410                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3411                         r = r->skip[PF_SKIP_IFP].ptr;
3412                 else if (r->direction && r->direction != direction)
3413                         r = r->skip[PF_SKIP_DIR].ptr;
3414                 else if (r->af && r->af != af)
3415                         r = r->skip[PF_SKIP_AF].ptr;
3416                 else if (r->proto && r->proto != pd->proto)
3417                         r = r->skip[PF_SKIP_PROTO].ptr;
3418                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3419                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3420                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3421                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3422                 else if (r->tos && !(r->tos & pd->tos))
3423                         r = TAILQ_NEXT(r, entries);
3424                 else if (r->rule_flag & PFRULE_FRAGMENT)
3425                         r = TAILQ_NEXT(r, entries);
3426                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3427                         r = TAILQ_NEXT(r, entries);
3428                 else if (r->anchorname[0] && r->anchor == NULL)
3429                         r = TAILQ_NEXT(r, entries);
3430                 else if (r->os_fingerprint != PF_OSFP_ANY)
3431                         r = TAILQ_NEXT(r, entries);
3432                 else {
3433                         if (r->tag)
3434                                 tag = r->tag;
3435                         if (r->anchor == NULL) {
3436                                 *rm = r;
3437                                 *am = a;
3438                                 *rsm = ruleset;
3439                                 if ((*rm)->quick)
3440                                         break;
3441                                 r = TAILQ_NEXT(r, entries);
3442                         } else
3443                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3444                                     PF_RULESET_FILTER);
3445                 }
3446                 if (r == NULL && a != NULL)
3447                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3448                             PF_RULESET_FILTER);
3449         }
3450         r = *rm;
3451         a = *am;
3452         ruleset = *rsm;
3453
3454         REASON_SET(&reason, PFRES_MATCH);
3455
3456         if (r->log)
3457                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3458
3459         if ((r->action == PF_DROP) &&
3460             ((r->rule_flag & PFRULE_RETURNICMP) ||
3461             (r->rule_flag & PFRULE_RETURN))) {
3462                 struct pf_addr *a = NULL;
3463
3464                 if (nr != NULL) {
3465                         if (direction == PF_OUT)
3466                                 a = saddr;
3467                         else
3468                                 a = daddr;
3469                 }
3470                 if (a != NULL) {
3471                         switch (af) {
3472 #ifdef INET
3473                         case AF_INET:
3474                                 pf_change_a(&a->v4.s_addr, pd->ip_sum,
3475                                     pd->baddr.v4.s_addr, 0);
3476                                 break;
3477 #endif /* INET */
3478 #ifdef INET6
3479                         case AF_INET6:
3480                                 PF_ACPY(a, &pd->baddr, af);
3481                                 break;
3482 #endif /* INET6 */
3483                         }
3484                 }
3485                 if ((af == AF_INET) && r->return_icmp)
3486                         pf_send_icmp(m, r->return_icmp >> 8,
3487                             r->return_icmp & 255, af, r);
3488                 else if ((af == AF_INET6) && r->return_icmp6)
3489                         pf_send_icmp(m, r->return_icmp6 >> 8,
3490                             r->return_icmp6 & 255, af, r);
3491         }
3492
3493         if (r->action != PF_PASS)
3494                 return (PF_DROP);
3495
3496         pf_tag_packet(m, tag);
3497
3498         if (r->keep_state || nr != NULL) {
3499                 /* create new state */
3500                 struct pf_state *s = NULL;
3501                 struct pf_src_node *sn = NULL;
3502
3503                 /* check maximums */
3504                 if (r->max_states && (r->states >= r->max_states))
3505                         goto cleanup;
3506                 /* src node for flter rule */
3507                 if ((r->rule_flag & PFRULE_SRCTRACK ||
3508                     r->rpool.opts & PF_POOL_STICKYADDR) &&
3509                     pf_insert_src_node(&sn, r, saddr, af) != 0)
3510                         goto cleanup;
3511                 /* src node for translation rule */
3512                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3513                     ((direction == PF_OUT &&
3514                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3515                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3516                         goto cleanup;
3517                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3518                 if (s == NULL) {
3519 cleanup:
3520                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3521                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3522                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3523                                 pf_status.src_nodes--;
3524                                 pool_put(&pf_src_tree_pl, sn);
3525                         }
3526                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3527                             nsn->expire == 0) {
3528                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3529                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3530                                 pf_status.src_nodes--;
3531                                 pool_put(&pf_src_tree_pl, nsn);
3532                         }
3533                         REASON_SET(&reason, PFRES_MEMORY);
3534                         return (PF_DROP);
3535                 }
3536                 bzero(s, sizeof(*s));
3537                 r->states++;
3538                 if (a != NULL)
3539                         a->states++;
3540                 s->rule.ptr = r;
3541                 s->nat_rule.ptr = nr;
3542                 if (s->nat_rule.ptr != NULL)
3543                         s->nat_rule.ptr->states++;
3544                 s->anchor.ptr = a;
3545                 s->allow_opts = r->allow_opts;
3546                 s->log = r->log & 2;
3547                 s->proto = pd->proto;
3548                 s->direction = direction;
3549                 s->af = af;
3550                 if (direction == PF_OUT) {
3551                         PF_ACPY(&s->gwy.addr, saddr, af);
3552                         PF_ACPY(&s->ext.addr, daddr, af);
3553                         if (nr != NULL)
3554                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3555                         else
3556                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3557                 } else {
3558                         PF_ACPY(&s->lan.addr, daddr, af);
3559                         PF_ACPY(&s->ext.addr, saddr, af);
3560                         if (nr != NULL)
3561                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3562                         else
3563                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3564                 }
3565                 s->src.state = PFOTHERS_SINGLE;
3566                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3567                 s->creation = time_second;
3568                 s->expire = time_second;
3569                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3570                 pf_set_rt_ifp(s, saddr);
3571                 if (sn != NULL) {
3572                         s->src_node = sn;
3573                         s->src_node->states++;
3574                 }
3575                 if (nsn != NULL) {
3576                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3577                         s->nat_src_node = nsn;
3578                         s->nat_src_node->states++;
3579                 }
3580                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3581                         REASON_SET(&reason, PFRES_MEMORY);
3582                         pf_src_tree_remove_state(s);
3583                         pool_put(&pf_state_pl, s);
3584                         return (PF_DROP);
3585                 } else
3586                         *sm = s;
3587         }
3588
3589         return (PF_PASS);
3590 }
3591
3592 int
3593 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3594     struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3595     struct pf_ruleset **rsm)
3596 {
3597         struct pf_rule          *r, *a = NULL;
3598         struct pf_ruleset       *ruleset = NULL;
3599         sa_family_t              af = pd->af;
3600         u_short                  reason;
3601         int                      tag = -1;
3602
3603         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3604         while (r != NULL) {
3605                 r->evaluations++;
3606                 if (r->kif != NULL &&
3607                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3608                         r = r->skip[PF_SKIP_IFP].ptr;
3609                 else if (r->direction && r->direction != direction)
3610                         r = r->skip[PF_SKIP_DIR].ptr;
3611                 else if (r->af && r->af != af)
3612                         r = r->skip[PF_SKIP_AF].ptr;
3613                 else if (r->proto && r->proto != pd->proto)
3614                         r = r->skip[PF_SKIP_PROTO].ptr;
3615                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3616                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3617                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3618                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3619                 else if (r->tos && !(r->tos & pd->tos))
3620                         r = TAILQ_NEXT(r, entries);
3621                 else if (r->src.port_op || r->dst.port_op ||
3622                     r->flagset || r->type || r->code ||
3623                     r->os_fingerprint != PF_OSFP_ANY)
3624                         r = TAILQ_NEXT(r, entries);
3625                 else if (r->match_tag && !pf_match_tag(m, r, NULL, &tag))
3626                         r = TAILQ_NEXT(r, entries);
3627                 else if (r->anchorname[0] && r->anchor == NULL)
3628                         r = TAILQ_NEXT(r, entries);
3629                 else {
3630                         if (r->anchor == NULL) {
3631                                 *rm = r;
3632                                 *am = a;
3633                                 *rsm = ruleset;
3634                                 if ((*rm)->quick)
3635                                         break;
3636                                 r = TAILQ_NEXT(r, entries);
3637                         } else
3638                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3639                                     PF_RULESET_FILTER);
3640                 }
3641                 if (r == NULL && a != NULL)
3642                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3643                             PF_RULESET_FILTER);
3644         }
3645         r = *rm;
3646         a = *am;
3647         ruleset = *rsm;
3648
3649         REASON_SET(&reason, PFRES_MATCH);
3650
3651         if (r->log)
3652                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3653
3654         if (r->action != PF_PASS)
3655                 return (PF_DROP);
3656
3657         pf_tag_packet(m, tag);
3658
3659         return (PF_PASS);
3660 }
3661
3662 int
3663 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3664     struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3665     u_short *reason)
3666 {
3667         struct pf_state          key;
3668         struct tcphdr           *th = pd->hdr.tcp;
3669         u_int16_t                win = ntohs(th->th_win);
3670         u_int32_t                ack, end, seq;
3671         u_int8_t                 sws, dws;
3672         int                      ackskew;
3673         int                      copyback = 0;
3674         struct pf_state_peer    *src, *dst;
3675
3676         key.af = pd->af;
3677         key.proto = IPPROTO_TCP;
3678         if (direction == PF_IN) {
3679                 PF_ACPY(&key.ext.addr, pd->src, key.af);
3680                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3681                 key.ext.port = th->th_sport;
3682                 key.gwy.port = th->th_dport;
3683         } else {
3684                 PF_ACPY(&key.lan.addr, pd->src, key.af);
3685                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3686                 key.lan.port = th->th_sport;
3687                 key.ext.port = th->th_dport;
3688         }
3689
3690         STATE_LOOKUP();
3691
3692         if (direction == (*state)->direction) {
3693                 src = &(*state)->src;
3694                 dst = &(*state)->dst;
3695         } else {
3696                 src = &(*state)->dst;
3697                 dst = &(*state)->src;
3698         }
3699
3700         if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3701                 if (direction != (*state)->direction)
3702                         return (PF_SYNPROXY_DROP);
3703                 if (th->th_flags & TH_SYN) {
3704                         if (ntohl(th->th_seq) != (*state)->src.seqlo)
3705                                 return (PF_DROP);
3706                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3707                             pd->src, th->th_dport, th->th_sport,
3708                             (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3709                             TH_SYN|TH_ACK, 0, (*state)->src.mss, 0);
3710                         return (PF_SYNPROXY_DROP);
3711                 } else if (!(th->th_flags & TH_ACK) ||
3712                     (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3713                     (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3714                         return (PF_DROP);
3715                 else
3716                         (*state)->src.state = PF_TCPS_PROXY_DST;
3717         }
3718         if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3719                 struct pf_state_host *src, *dst;
3720
3721                 if (direction == PF_OUT) {
3722                         src = &(*state)->gwy;
3723                         dst = &(*state)->ext;
3724                 } else {
3725                         src = &(*state)->ext;
3726                         dst = &(*state)->lan;
3727                 }
3728                 if (direction == (*state)->direction) {
3729                         if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3730                             (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3731                             (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3732                                 return (PF_DROP);
3733                         (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3734                         if ((*state)->dst.seqhi == 1)
3735                                 (*state)->dst.seqhi = karc4random();
3736                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3737                             &dst->addr, src->port, dst->port,
3738                             (*state)->dst.seqhi, 0, TH_SYN, 0,
3739                             (*state)->src.mss, 0);
3740                         return (PF_SYNPROXY_DROP);
3741                 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3742                     (TH_SYN|TH_ACK)) ||
3743                     (ntohl(th->th_ack) != (*state)->dst.seqhi + 1))
3744                         return (PF_DROP);
3745                 else {
3746                         (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3747                         (*state)->dst.seqlo = ntohl(th->th_seq);
3748                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3749                             pd->src, th->th_dport, th->th_sport,
3750                             ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3751                             TH_ACK, (*state)->src.max_win, 0, 0);
3752                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3753                             &dst->addr, src->port, dst->port,
3754                             (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3755                             TH_ACK, (*state)->dst.max_win, 0, 0);
3756                         (*state)->src.seqdiff = (*state)->dst.seqhi -
3757                             (*state)->src.seqlo;
3758                         (*state)->dst.seqdiff = (*state)->src.seqhi -
3759                             (*state)->dst.seqlo;
3760                         (*state)->src.seqhi = (*state)->src.seqlo +
3761                             (*state)->src.max_win;
3762                         (*state)->dst.seqhi = (*state)->dst.seqlo +
3763                             (*state)->dst.max_win;
3764                         (*state)->src.wscale = (*state)->dst.wscale = 0;
3765                         (*state)->src.state = (*state)->dst.state =
3766                             TCPS_ESTABLISHED;
3767                         return (PF_SYNPROXY_DROP);
3768                 }
3769         }
3770
3771         if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3772                 sws = src->wscale & PF_WSCALE_MASK;
3773                 dws = dst->wscale & PF_WSCALE_MASK;
3774         } else
3775                 sws = dws = 0;
3776
3777         /*
3778          * Sequence tracking algorithm from Guido van Rooij's paper:
3779          *   http://www.madison-gurkha.com/publications/tcp_filtering/
3780          *      tcp_filtering.ps
3781          */
3782
3783         seq = ntohl(th->th_seq);
3784         if (src->seqlo == 0) {
3785                 /* First packet from this end. Set its state */
3786
3787                 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3788                     src->scrub == NULL) {
3789                         if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3790                                 REASON_SET(reason, PFRES_MEMORY);
3791                                 return (PF_DROP);
3792                         }
3793                 }
3794
3795                 /* Deferred generation of sequence number modulator */
3796                 if (dst->seqdiff && !src->seqdiff) {
3797                         while ((src->seqdiff = karc4random()) == 0)
3798                                 ;
3799                         ack = ntohl(th->th_ack) - dst->seqdiff;
3800                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3801                             src->seqdiff), 0);
3802                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3803                         copyback = 1;
3804                 } else {
3805                         ack = ntohl(th->th_ack);
3806                 }
3807
3808                 end = seq + pd->p_len;
3809                 if (th->th_flags & TH_SYN) {
3810                         end++;
3811                         if (dst->wscale & PF_WSCALE_FLAG) {
3812                                 src->wscale = pf_get_wscale(m, off, th->th_off,
3813                                     pd->af);
3814                                 if (src->wscale & PF_WSCALE_FLAG) {
3815                                         /* Remove scale factor from initial
3816                                          * window */
3817                                         sws = src->wscale & PF_WSCALE_MASK;
3818                                         win = ((u_int32_t)win + (1 << sws) - 1)
3819                                             >> sws;
3820                                         dws = dst->wscale & PF_WSCALE_MASK;
3821                                 } else {
3822                                         /* fixup other window */
3823                                         dst->max_win <<= dst->wscale &
3824                                             PF_WSCALE_MASK;
3825                                         /* in case of a retrans SYN|ACK */
3826                                         dst->wscale = 0;
3827                                 }
3828                         }
3829                 }
3830                 if (th->th_flags & TH_FIN)
3831                         end++;
3832
3833                 src->seqlo = seq;
3834                 if (src->state < TCPS_SYN_SENT)
3835                         src->state = TCPS_SYN_SENT;
3836
3837                 /*
3838                  * May need to slide the window (seqhi may have been set by
3839                  * the crappy stack check or if we picked up the connection
3840                  * after establishment)
3841                  */
3842                 if (src->seqhi == 1 ||
3843                     SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3844                         src->seqhi = end + MAX(1, dst->max_win << dws);
3845                 if (win > src->max_win)
3846                         src->max_win = win;
3847
3848         } else {
3849                 ack = ntohl(th->th_ack) - dst->seqdiff;
3850                 if (src->seqdiff) {
3851                         /* Modulate sequence numbers */
3852                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3853                             src->seqdiff), 0);
3854                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3855                         copyback = 1;
3856                 }
3857                 end = seq + pd->p_len;
3858                 if (th->th_flags & TH_SYN)
3859                         end++;
3860                 if (th->th_flags & TH_FIN)
3861                         end++;
3862         }
3863
3864         if ((th->th_flags & TH_ACK) == 0) {
3865                 /* Let it pass through the ack skew check */
3866                 ack = dst->seqlo;
3867         } else if ((ack == 0 &&
3868             (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3869             /* broken tcp stacks do not set ack */
3870             (dst->state < TCPS_SYN_SENT)) {
3871                 /*
3872                  * Many stacks (ours included) will set the ACK number in an
3873                  * FIN|ACK if the SYN times out -- no sequence to ACK.
3874                  */
3875                 ack = dst->seqlo;
3876         }
3877
3878         if (seq == end) {
3879                 /* Ease sequencing restrictions on no data packets */
3880                 seq = src->seqlo;
3881                 end = seq;
3882         }
3883
3884         ackskew = dst->seqlo - ack;
3885
3886 #define MAXACKWINDOW (0xffff + 1500)    /* 1500 is an arbitrary fudge factor */
3887         if (SEQ_GEQ(src->seqhi, end) &&
3888             /* Last octet inside other's window space */
3889             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3890             /* Retrans: not more than one window back */
3891             (ackskew >= -MAXACKWINDOW) &&
3892             /* Acking not more than one reassembled fragment backwards */
3893             (ackskew <= (MAXACKWINDOW << sws))) {
3894             /* Acking not more than one window forward */
3895
3896                 /* update max window */
3897                 if (src->max_win < win)
3898                         src->max_win = win;
3899                 /* synchronize sequencing */
3900                 if (SEQ_GT(end, src->seqlo))
3901                         src->seqlo = end;
3902                 /* slide the window of what the other end can send */
3903                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3904                         dst->seqhi = ack + MAX((win << sws), 1);
3905
3906
3907                 /* update states */
3908                 if (th->th_flags & TH_SYN)
3909                         if (src->state < TCPS_SYN_SENT)
3910                                 src->state = TCPS_SYN_SENT;
3911                 if (th->th_flags & TH_FIN)
3912                         if (src->state < TCPS_CLOSING)
3913                                 src->state = TCPS_CLOSING;
3914                 if (th->th_flags & TH_ACK) {
3915                         if (dst->state == TCPS_SYN_SENT)
3916                                 dst->state = TCPS_ESTABLISHED;
3917                         else if (dst->state == TCPS_CLOSING)
3918                                 dst->state = TCPS_FIN_WAIT_2;
3919                 }
3920                 if (th->th_flags & TH_RST)
3921                         src->state = dst->state = TCPS_TIME_WAIT;
3922
3923                 /* update expire time */
3924                 (*state)->expire = time_second;
3925                 if (src->state >= TCPS_FIN_WAIT_2 &&
3926                     dst->state >= TCPS_FIN_WAIT_2)
3927                         (*state)->timeout = PFTM_TCP_CLOSED;
3928                 else if (src->state >= TCPS_FIN_WAIT_2 ||
3929                     dst->state >= TCPS_FIN_WAIT_2)
3930                         (*state)->timeout = PFTM_TCP_FIN_WAIT;
3931                 else if (src->state < TCPS_ESTABLISHED ||
3932                     dst->state < TCPS_ESTABLISHED)
3933                         (*state)->timeout = PFTM_TCP_OPENING;
3934                 else if (src->state >= TCPS_CLOSING ||
3935                     dst->state >= TCPS_CLOSING)
3936                         (*state)->timeout = PFTM_TCP_CLOSING;
3937                 else
3938                         (*state)->timeout = PFTM_TCP_ESTABLISHED;
3939
3940                 /* Fall through to PASS packet */
3941
3942         } else if ((dst->state < TCPS_SYN_SENT ||
3943                 dst->state >= TCPS_FIN_WAIT_2 ||
3944                 src->state >= TCPS_FIN_WAIT_2) &&
3945             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
3946             /* Within a window forward of the originating packet */
3947             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
3948             /* Within a window backward of the originating packet */
3949
3950                 /*
3951                  * This currently handles three situations:
3952                  *  1) Stupid stacks will shotgun SYNs before their peer
3953                  *     replies.
3954                  *  2) When PF catches an already established stream (the
3955                  *     firewall rebooted, the state table was flushed, routes
3956                  *     changed...)
3957                  *  3) Packets get funky immediately after the connection
3958                  *     closes (this should catch Solaris spurious ACK|FINs
3959                  *     that web servers like to spew after a close)
3960                  *
3961                  * This must be a little more careful than the above code
3962                  * since packet floods will also be caught here. We don't
3963                  * update the TTL here to mitigate the damage of a packet
3964                  * flood and so the same code can handle awkward establishment
3965                  * and a loosened connection close.
3966                  * In the establishment case, a correct peer response will
3967                  * validate the connection, go through the normal state code
3968                  * and keep updating the state TTL.
3969                  */
3970
3971                 if (pf_status.debug >= PF_DEBUG_MISC) {
3972                         kprintf("pf: loose state match: ");
3973                         pf_print_state(*state);
3974                         pf_print_flags(th->th_flags);
3975                         kprintf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n",
3976                             seq, ack, pd->p_len, ackskew,
3977                             (*state)->packets[0], (*state)->packets[1]);
3978                 }
3979
3980                 /* update max window */
3981                 if (src->max_win < win)
3982                         src->max_win = win;
3983                 /* synchronize sequencing */
3984                 if (SEQ_GT(end, src->seqlo))
3985                         src->seqlo = end;
3986                 /* slide the window of what the other end can send */
3987                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3988                         dst->seqhi = ack + MAX((win << sws), 1);
3989
3990                 /*
3991                  * Cannot set dst->seqhi here since this could be a shotgunned
3992                  * SYN and not an already established connection.
3993                  */
3994
3995                 if (th->th_flags & TH_FIN)
3996                         if (src->state < TCPS_CLOSING)
3997                                 src->state = TCPS_CLOSING;
3998                 if (th->th_flags & TH_RST)
3999                         src->state = dst->state = TCPS_TIME_WAIT;
4000
4001                 /* Fall through to PASS packet */
4002
4003         } else {
4004                 if ((*state)->dst.state == TCPS_SYN_SENT &&
4005                     (*state)->src.state == TCPS_SYN_SENT) {
4006                         /* Send RST for state mismatches during handshake */
4007                         if (!(th->th_flags & TH_RST)) {
4008                                 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
4009
4010                                 if (th->th_flags & TH_SYN)
4011                                         ack++;
4012                                 if (th->th_flags & TH_FIN)
4013                                         ack++;
4014                                 pf_send_tcp((*state)->rule.ptr, pd->af,
4015                                     pd->dst, pd->src, th->th_dport,
4016                                     th->th_sport, ntohl(th->th_ack), ack,
4017                                     TH_RST|TH_ACK, 0, 0,
4018                                     (*state)->rule.ptr->return_ttl);
4019                         }
4020                         src->seqlo = 0;
4021                         src->seqhi = 1;
4022                         src->max_win = 1;
4023                 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4024                         kprintf("pf: BAD state: ");
4025                         pf_print_state(*state);
4026                         pf_print_flags(th->th_flags);
4027                         kprintf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d "
4028                             "dir=%s,%s\n", seq, ack, pd->p_len, ackskew,
4029                             (*state)->packets[0], (*state)->packets[1],
4030                             direction == PF_IN ? "in" : "out",
4031                             direction == (*state)->direction ? "fwd" : "rev");
4032                         kprintf("pf: State failure on: %c %c %c %c | %c %c\n",
4033                             SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4034                             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4035                             ' ': '2',
4036                             (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4037                             (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4038                             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4039                             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4040                 }
4041                 return (PF_DROP);
4042         }
4043
4044         if (dst->scrub || src->scrub) {
4045                 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4046                     src, dst, &copyback))
4047                         return (PF_DROP);
4048         }
4049
4050         /* Any packets which have gotten here are to be passed */
4051
4052         /* translate source/destination address, if necessary */
4053         if (STATE_TRANSLATE(*state)) {
4054                 if (direction == PF_OUT)
4055                         pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4056                             &th->th_sum, &(*state)->gwy.addr,
4057                             (*state)->gwy.port, 0, pd->af);
4058                 else
4059                         pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4060                             &th->th_sum, &(*state)->lan.addr,
4061                             (*state)->lan.port, 0, pd->af);
4062                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4063         } else if (copyback) {
4064                 /* Copyback sequence modulation or stateful scrub changes */
4065                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4066         }
4067
4068         return (PF_PASS);
4069 }
4070
4071 int
4072 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4073     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4074 {
4075         struct pf_state_peer    *src, *dst;
4076         struct pf_state          key;
4077         struct udphdr           *uh = pd->hdr.udp;
4078
4079         key.af = pd->af;
4080         key.proto = IPPROTO_UDP;
4081         if (direction == PF_IN) {
4082                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4083                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4084                 key.ext.port = uh->uh_sport;
4085                 key.gwy.port = uh->uh_dport;
4086         } else {
4087                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4088                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4089                 key.lan.port = uh->uh_sport;
4090                 key.ext.port = uh->uh_dport;
4091         }
4092
4093         STATE_LOOKUP();
4094
4095         if (direction == (*state)->direction) {
4096                 src = &(*state)->src;
4097                 dst = &(*state)->dst;
4098         } else {
4099                 src = &(*state)->dst;
4100                 dst = &(*state)->src;
4101         }
4102
4103         /* update states */
4104         if (src->state < PFUDPS_SINGLE)
4105                 src->state = PFUDPS_SINGLE;
4106         if (dst->state == PFUDPS_SINGLE)
4107                 dst->state = PFUDPS_MULTIPLE;
4108
4109         /* update expire time */
4110         (*state)->expire = time_second;
4111         if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4112                 (*state)->timeout = PFTM_UDP_MULTIPLE;
4113         else
4114                 (*state)->timeout = PFTM_UDP_SINGLE;
4115
4116         /* translate source/destination address, if necessary */
4117         if (STATE_TRANSLATE(*state)) {
4118                 if (direction == PF_OUT)
4119                         pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4120                             &uh->uh_sum, &(*state)->gwy.addr,
4121                             (*state)->gwy.port, 1, pd->af);
4122                 else
4123                         pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4124                             &uh->uh_sum, &(*state)->lan.addr,
4125                             (*state)->lan.port, 1, pd->af);
4126                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4127         }
4128
4129         return (PF_PASS);
4130 }
4131
4132 int
4133 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4134     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4135 {
4136         struct pf_addr  *saddr = pd->src, *daddr = pd->dst;
4137         u_int16_t        icmpid = 0;
4138         u_int16_t       *icmpsum = NULL;
4139         u_int8_t         icmptype = 0;
4140         int              state_icmp = 0;
4141
4142         switch (pd->proto) {
4143 #ifdef INET
4144         case IPPROTO_ICMP:
4145                 icmptype = pd->hdr.icmp->icmp_type;
4146                 icmpid = pd->hdr.icmp->icmp_id;
4147                 icmpsum = &pd->hdr.icmp->icmp_cksum;
4148
4149                 if (icmptype == ICMP_UNREACH ||
4150                     icmptype == ICMP_SOURCEQUENCH ||
4151                     icmptype == ICMP_REDIRECT ||
4152                     icmptype == ICMP_TIMXCEED ||
4153                     icmptype == ICMP_PARAMPROB)
4154                         state_icmp++;
4155                 break;
4156 #endif /* INET */
4157 #ifdef INET6
4158         case IPPROTO_ICMPV6:
4159                 icmptype = pd->hdr.icmp6->icmp6_type;
4160                 icmpid = pd->hdr.icmp6->icmp6_id;
4161                 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4162
4163                 if (icmptype == ICMP6_DST_UNREACH ||
4164                     icmptype == ICMP6_PACKET_TOO_BIG ||
4165                     icmptype == ICMP6_TIME_EXCEEDED ||
4166                     icmptype == ICMP6_PARAM_PROB)
4167                         state_icmp++;
4168                 break;
4169 #endif /* INET6 */
4170         }
4171
4172         if (!state_icmp) {
4173
4174                 /*
4175                  * ICMP query/reply message not related to a TCP/UDP packet.
4176                  * Search for an ICMP state.
4177                  */
4178                 struct pf_state         key;
4179
4180                 key.af = pd->af;
4181                 key.proto = pd->proto;
4182                 if (direction == PF_IN) {
4183                         PF_ACPY(&key.ext.addr, pd->src, key.af);
4184                         PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4185                         key.ext.port = icmpid;
4186                         key.gwy.port = icmpid;
4187                 } else {
4188                         PF_ACPY(&key.lan.addr, pd->src, key.af);
4189                         PF_ACPY(&key.ext.addr, pd->dst, key.af);
4190                         key.lan.port = icmpid;
4191                         key.ext.port = icmpid;
4192                 }
4193
4194                 STATE_LOOKUP();
4195
4196                 (*state)->expire = time_second;
4197                 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4198
4199                 /* translate source/destination address, if necessary */
4200                 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) {
4201                         if (direction == PF_OUT) {
4202                                 switch (pd->af) {
4203 #ifdef INET
4204                                 case AF_INET:
4205                                         pf_change_a(&saddr->v4.s_addr,
4206                                             pd->ip_sum,
4207                                             (*state)->gwy.addr.v4.s_addr, 0);
4208                                         break;
4209 #endif /* INET */
4210 #ifdef INET6
4211                                 case AF_INET6:
4212                                         pf_change_a6(saddr,
4213                                             &pd->hdr.icmp6->icmp6_cksum,
4214                                             &(*state)->gwy.addr, 0);
4215                                         m_copyback(m, off,
4216                                             sizeof(struct icmp6_hdr),
4217                                             (caddr_t)pd->hdr.icmp6);
4218                                         break;
4219 #endif /* INET6 */
4220                                 }
4221                         } else {
4222                                 switch (pd->af) {
4223 #ifdef INET
4224                                 case AF_INET:
4225                                         pf_change_a(&daddr->v4.s_addr,
4226                                             pd->ip_sum,
4227                                             (*state)->lan.addr.v4.s_addr, 0);
4228                                         break;
4229 #endif /* INET */
4230 #ifdef INET6
4231                                 case AF_INET6:
4232                                         pf_change_a6(daddr,
4233                                             &pd->hdr.icmp6->icmp6_cksum,
4234                                             &(*state)->lan.addr, 0);
4235                                         m_copyback(m, off,
4236                                             sizeof(struct icmp6_hdr),
4237                                             (caddr_t)pd->hdr.icmp6);
4238                                         break;
4239 #endif /* INET6 */
4240                                 }
4241                         }
4242                 }
4243
4244                 return (PF_PASS);
4245
4246         } else {
4247                 /*
4248                  * ICMP error message in response to a TCP/UDP packet.
4249                  * Extract the inner TCP/UDP header and search for that state.
4250                  */
4251
4252                 struct pf_pdesc pd2;
4253 #ifdef INET
4254                 struct ip       h2;
4255 #endif /* INET */
4256 #ifdef INET6
4257                 struct ip6_hdr  h2_6;
4258                 int             terminal = 0;
4259 #endif /* INET6 */
4260                 int             ipoff2 = 0;
4261                 int             off2 = 0;
4262
4263                 pd2.af = pd->af;
4264                 switch (pd->af) {
4265 #ifdef INET
4266                 case AF_INET:
4267                         /* offset of h2 in mbuf chain */
4268                         ipoff2 = off + ICMP_MINLEN;
4269
4270                         if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4271                             NULL, NULL, pd2.af)) {
4272                                 DPFPRINTF(PF_DEBUG_MISC,
4273                                     ("pf: ICMP error message too short "
4274                                     "(ip)\n"));
4275                                 return (PF_DROP);
4276                         }
4277                         /*
4278                          * ICMP error messages don't refer to non-first
4279                          * fragments
4280                          */
4281                         /*
4282                          * Note: We are dealing with an encapsulated
4283                          * header. This means ip_off/ip_len are not
4284                          * in host byte order!
4285                          */
4286                         if (h2.ip_off & htons(IP_OFFMASK))
4287                                 return (PF_DROP);
4288
4289                         /* offset of protocol header that follows h2 */
4290                         off2 = ipoff2 + (h2.ip_hl << 2);
4291
4292                         pd2.proto = h2.ip_p;
4293                         pd2.src = (struct pf_addr *)&h2.ip_src;
4294                         pd2.dst = (struct pf_addr *)&h2.ip_dst;
4295                         pd2.ip_sum = &h2.ip_sum;
4296                         break;
4297 #endif /* INET */
4298 #ifdef INET6
4299                 case AF_INET6:
4300                         ipoff2 = off + sizeof(struct icmp6_hdr);
4301
4302                         if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4303                             NULL, NULL, pd2.af)) {
4304                                 DPFPRINTF(PF_DEBUG_MISC,
4305                                     ("pf: ICMP error message too short "
4306                                     "(ip6)\n"));
4307                                 return (PF_DROP);
4308                         }
4309                         pd2.proto = h2_6.ip6_nxt;
4310                         pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4311                         pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4312                         pd2.ip_sum = NULL;
4313                         off2 = ipoff2 + sizeof(h2_6);
4314                         do {
4315                                 switch (pd2.proto) {
4316                                 case IPPROTO_FRAGMENT:
4317                                         /*
4318                                          * ICMPv6 error messages for
4319                                          * non-first fragments
4320                                          */
4321                                         return (PF_DROP);
4322                                 case IPPROTO_AH:
4323                                 case IPPROTO_HOPOPTS:
4324                                 case IPPROTO_ROUTING:
4325                                 case IPPROTO_DSTOPTS: {
4326                                         /* get next header and header length */
4327                                         struct ip6_ext opt6;
4328
4329                                         if (!pf_pull_hdr(m, off2, &opt6,
4330                                             sizeof(opt6), NULL, NULL, pd2.af)) {
4331                                                 DPFPRINTF(PF_DEBUG_MISC,
4332                                                     ("pf: ICMPv6 short opt\n"));
4333                                                 return (PF_DROP);
4334                                         }
4335                                         if (pd2.proto == IPPROTO_AH)
4336                                                 off2 += (opt6.ip6e_len + 2) * 4;
4337                                         else
4338                                                 off2 += (opt6.ip6e_len + 1) * 8;
4339                                         pd2.proto = opt6.ip6e_nxt;
4340                                         /* goto the next header */
4341                                         break;
4342                                 }
4343                                 default:
4344                                         terminal++;
4345                                         break;
4346                                 }
4347                         } while (!terminal);
4348                         break;
4349 #endif /* INET6 */
4350                 }
4351
4352                 switch (pd2.proto) {
4353                 case IPPROTO_TCP: {
4354                         struct tcphdr            th;
4355                         u_int32_t                seq;
4356                         struct pf_state          key;
4357                         struct pf_state_peer    *src, *dst;
4358                         u_int8_t                 dws;
4359                         int                      copyback = 0;
4360
4361                         /*
4362                          * Only the first 8 bytes of the TCP header can be
4363                          * expected. Don't access any TCP header fields after
4364                          * th_seq, an ackskew test is not possible.
4365                          */
4366                         if (!pf_pull_hdr(m, off2, &th, 8, NULL, NULL, pd2.af)) {
4367                                 DPFPRINTF(PF_DEBUG_MISC,
4368                                     ("pf: ICMP error message too short "
4369                                     "(tcp)\n"));
4370                                 return (PF_DROP);
4371                         }
4372
4373                         key.af = pd2.af;
4374                         key.proto = IPPROTO_TCP;
4375                         if (direction == PF_IN) {
4376                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4377                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4378                                 key.ext.port = th.th_dport;
4379                                 key.gwy.port = th.th_sport;
4380                         } else {
4381                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4382                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4383                                 key.lan.port = th.th_dport;
4384                                 key.ext.port = th.th_sport;
4385                         }
4386
4387                         STATE_LOOKUP();
4388
4389                         if (direction == (*state)->direction) {
4390                                 src = &(*state)->dst;
4391                                 dst = &(*state)->src;
4392                         } else {
4393                                 src = &(*state)->src;
4394                                 dst = &(*state)->dst;
4395                         }
4396
4397                         if (src->wscale && dst->wscale &&
4398                             !(th.th_flags & TH_SYN))
4399                                 dws = dst->wscale & PF_WSCALE_MASK;
4400                         else
4401                                 dws = 0;
4402
4403                         /* Demodulate sequence number */
4404                         seq = ntohl(th.th_seq) - src->seqdiff;
4405                         if (src->seqdiff) {
4406                                 pf_change_a(&th.th_seq, icmpsum,
4407                                     htonl(seq), 0);
4408                                 copyback = 1;
4409                         }
4410
4411                         if (!SEQ_GEQ(src->seqhi, seq) ||
4412                             !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4413                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4414                                         kprintf("pf: BAD ICMP %d:%d ",
4415                                             icmptype, pd->hdr.icmp->icmp_code);
4416                                         pf_print_host(pd->src, 0, pd->af);
4417                                         kprintf(" -> ");
4418                                         pf_print_host(pd->dst, 0, pd->af);
4419                                         kprintf(" state: ");
4420                                         pf_print_state(*state);
4421                                         kprintf(" seq=%u\n", seq);
4422                                 }
4423                                 return (PF_DROP);
4424                         }
4425
4426                         if (STATE_TRANSLATE(*state)) {
4427                                 if (direction == PF_IN) {
4428                                         pf_change_icmp(pd2.src, &th.th_sport,
4429                                             daddr, &(*state)->lan.addr,
4430                                             (*state)->lan.port, NULL,
4431                                             pd2.ip_sum, icmpsum,
4432                                             pd->ip_sum, 0, pd2.af);
4433                                 } else {
4434                                         pf_change_icmp(pd2.dst, &th.th_dport,
4435                                             saddr, &(*state)->gwy.addr,
4436                                             (*state)->gwy.port, NULL,
4437                                             pd2.ip_sum, icmpsum,
4438                                             pd->ip_sum, 0, pd2.af);
4439                                 }
4440                                 copyback = 1;
4441                         }
4442
4443                         if (copyback) {
4444                                 switch (pd2.af) {
4445 #ifdef INET
4446                                 case AF_INET:
4447                                         m_copyback(m, off, ICMP_MINLEN,
4448                                             (caddr_t)pd->hdr.icmp);
4449                                         m_copyback(m, ipoff2, sizeof(h2),
4450                                             (caddr_t)&h2);
4451                                         break;
4452 #endif /* INET */
4453 #ifdef INET6
4454                                 case AF_INET6:
4455                                         m_copyback(m, off,
4456                                             sizeof(struct icmp6_hdr),
4457                                             (caddr_t)pd->hdr.icmp6);
4458                                         m_copyback(m, ipoff2, sizeof(h2_6),
4459                                             (caddr_t)&h2_6);
4460                                         break;
4461 #endif /* INET6 */
4462                                 }
4463                                 m_copyback(m, off2, 8, (caddr_t)&th);
4464                         }
4465
4466                         return (PF_PASS);
4467                         break;
4468                 }
4469                 case IPPROTO_UDP: {
4470                         struct udphdr           uh;
4471                         struct pf_state         key;
4472
4473                         if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4474                             NULL, NULL, pd2.af)) {
4475                                 DPFPRINTF(PF_DEBUG_MISC,
4476                                     ("pf: ICMP error message too short "
4477                                     "(udp)\n"));
4478                                 return (PF_DROP);
4479                         }
4480
4481                         key.af = pd2.af;
4482                         key.proto = IPPROTO_UDP;
4483                         if (direction == PF_IN) {
4484                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4485                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4486                                 key.ext.port = uh.uh_dport;
4487                                 key.gwy.port = uh.uh_sport;
4488                         } else {
4489                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4490                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4491                                 key.lan.port = uh.uh_dport;
4492                                 key.ext.port = uh.uh_sport;
4493                         }
4494
4495                         STATE_LOOKUP();
4496
4497                         if (STATE_TRANSLATE(*state)) {
4498                                 if (direction == PF_IN) {
4499                                         pf_change_icmp(pd2.src, &uh.uh_sport,
4500                                             daddr, &(*state)->lan.addr,
4501                                             (*state)->lan.port, &uh.uh_sum,
4502                                             pd2.ip_sum, icmpsum,
4503                                             pd->ip_sum, 1, pd2.af);
4504                                 } else {
4505                                         pf_change_icmp(pd2.dst, &uh.uh_dport,
4506                                             saddr, &(*state)->gwy.addr,
4507                                             (*state)->gwy.port, &uh.uh_sum,
4508                                             pd2.ip_sum, icmpsum,
4509                                             pd->ip_sum, 1, pd2.af);
4510                                 }
4511                                 switch (pd2.af) {
4512 #ifdef INET
4513                                 case AF_INET:
4514                                         m_copyback(m, off, ICMP_MINLEN,
4515                                             (caddr_t)pd->hdr.icmp);
4516                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4517                                         break;
4518 #endif /* INET */
4519 #ifdef INET6
4520                                 case AF_INET6:
4521                                         m_copyback(m, off,
4522                                             sizeof(struct icmp6_hdr),
4523                                             (caddr_t)pd->hdr.icmp6);
4524                                         m_copyback(m, ipoff2, sizeof(h2_6),
4525                                             (caddr_t)&h2_6);
4526                                         break;
4527 #endif /* INET6 */
4528                                 }
4529                                 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4530                         }
4531
4532                         return (PF_PASS);
4533                         break;
4534                 }
4535 #ifdef INET
4536                 case IPPROTO_ICMP: {
4537                         struct icmp             iih;
4538                         struct pf_state         key;
4539
4540                         if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4541                             NULL, NULL, pd2.af)) {
4542                                 DPFPRINTF(PF_DEBUG_MISC,
4543                                     ("pf: ICMP error message too short i"
4544                                     "(icmp)\n"));
4545                                 return (PF_DROP);
4546                         }
4547
4548                         key.af = pd2.af;
4549                         key.proto = IPPROTO_ICMP;
4550                         if (direction == PF_IN) {
4551                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4552                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4553                                 key.ext.port = iih.icmp_id;
4554                                 key.gwy.port = iih.icmp_id;
4555                         } else {
4556                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4557                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4558                                 key.lan.port = iih.icmp_id;
4559                                 key.ext.port = iih.icmp_id;
4560                         }
4561
4562                         STATE_LOOKUP();
4563
4564                         if (STATE_TRANSLATE(*state)) {
4565                                 if (direction == PF_IN) {
4566                                         pf_change_icmp(pd2.src, &iih.icmp_id,
4567                                             daddr, &(*state)->lan.addr,
4568                                             (*state)->lan.port, NULL,
4569                                             pd2.ip_sum, icmpsum,
4570                                             pd->ip_sum, 0, AF_INET);
4571                                 } else {
4572                                         pf_change_icmp(pd2.dst, &iih.icmp_id,
4573                                             saddr, &(*state)->gwy.addr,
4574                                             (*state)->gwy.port, NULL,
4575                                             pd2.ip_sum, icmpsum,
4576                                             pd->ip_sum, 0, AF_INET);
4577                                 }
4578                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4579                                 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4580                                 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4581                         }
4582
4583                         return (PF_PASS);
4584                         break;
4585                 }
4586 #endif /* INET */
4587 #ifdef INET6
4588                 case IPPROTO_ICMPV6: {
4589                         struct icmp6_hdr        iih;
4590                         struct pf_state         key;
4591
4592                         if (!pf_pull_hdr(m, off2, &iih,
4593                             sizeof(struct icmp6_hdr), NULL, NULL, pd2.af)) {
4594                                 DPFPRINTF(PF_DEBUG_MISC,
4595                                     ("pf: ICMP error message too short "
4596                                     "(icmp6)\n"));
4597                                 return (PF_DROP);
4598                         }
4599
4600                         key.af = pd2.af;
4601                         key.proto = IPPROTO_ICMPV6;
4602                         if (direction == PF_IN) {
4603                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4604                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4605                                 key.ext.port = iih.icmp6_id;
4606                                 key.gwy.port = iih.icmp6_id;
4607                         } else {
4608                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4609                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4610                                 key.lan.port = iih.icmp6_id;
4611                                 key.ext.port = iih.icmp6_id;
4612                         }
4613
4614                         STATE_LOOKUP();
4615
4616                         if (STATE_TRANSLATE(*state)) {
4617                                 if (direction == PF_IN) {
4618                                         pf_change_icmp(pd2.src, &iih.icmp6_id,
4619                                             daddr, &(*state)->lan.addr,
4620                                             (*state)->lan.port, NULL,
4621                                             pd2.ip_sum, icmpsum,
4622                                             pd->ip_sum, 0, AF_INET6);
4623                                 } else {
4624                                         pf_change_icmp(pd2.dst, &iih.icmp6_id,
4625                                             saddr, &(*state)->gwy.addr,
4626                                             (*state)->gwy.port, NULL,
4627                                             pd2.ip_sum, icmpsum,
4628                                             pd->ip_sum, 0, AF_INET6);
4629                                 }
4630                                 m_copyback(m, off, sizeof(struct icmp6_hdr),
4631                                     (caddr_t)pd->hdr.icmp6);
4632                                 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
4633                                 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4634                                     (caddr_t)&iih);
4635                         }
4636
4637                         return (PF_PASS);
4638                         break;
4639                 }
4640 #endif /* INET6 */
4641                 default: {
4642                         struct pf_state         key;
4643
4644                         key.af = pd2.af;
4645                         key.proto = pd2.proto;
4646                         if (direction == PF_IN) {
4647                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4648                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4649                                 key.ext.port = 0;
4650                                 key.gwy.port = 0;
4651                         } else {
4652                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4653                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4654                                 key.lan.port = 0;
4655                                 key.ext.port = 0;
4656                         }
4657
4658                         STATE_LOOKUP();
4659
4660                         if (STATE_TRANSLATE(*state)) {
4661                                 if (direction == PF_IN) {
4662                                         pf_change_icmp(pd2.src, NULL,
4663                                             daddr, &(*state)->lan.addr,
4664                                             0, NULL,
4665                                             pd2.ip_sum, icmpsum,
4666                                             pd->ip_sum, 0, pd2.af);
4667                                 } else {
4668                                         pf_change_icmp(pd2.dst, NULL,
4669                                             saddr, &(*state)->gwy.addr,
4670                                             0, NULL,
4671                                             pd2.ip_sum, icmpsum,
4672                                             pd->ip_sum, 0, pd2.af);
4673                                 }
4674                                 switch (pd2.af) {
4675 #ifdef INET
4676                                 case AF_INET:
4677                                         m_copyback(m, off, ICMP_MINLEN,
4678                                             (caddr_t)pd->hdr.icmp);
4679                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4680                                         break;
4681 #endif /* INET */
4682 #ifdef INET6
4683                                 case AF_INET6:
4684                                         m_copyback(m, off,
4685                                             sizeof(struct icmp6_hdr),
4686                                             (caddr_t)pd->hdr.icmp6);
4687                                         m_copyback(m, ipoff2, sizeof(h2_6),
4688                                             (caddr_t)&h2_6);
4689                                         break;
4690 #endif /* INET6 */
4691                                 }
4692                         }
4693
4694                         return (PF_PASS);
4695                         break;
4696                 }
4697                 }
4698         }
4699 }
4700
4701 int
4702 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4703     struct pf_pdesc *pd)
4704 {
4705         struct pf_state_peer    *src, *dst;
4706         struct pf_state          key;
4707
4708         key.af = pd->af;
4709         key.proto = pd->proto;
4710         if (direction == PF_IN) {
4711                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4712                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4713                 key.ext.port = 0;
4714                 key.gwy.port = 0;
4715         } else {
4716                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4717                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4718                 key.lan.port = 0;
4719                 key.ext.port = 0;
4720         }
4721
4722         STATE_LOOKUP();
4723
4724         if (direction == (*state)->direction) {
4725                 src = &(*state)->src;
4726                 dst = &(*state)->dst;
4727         } else {
4728                 src = &(*state)->dst;
4729                 dst = &(*state)->src;
4730         }
4731
4732         /* update states */
4733         if (src->state < PFOTHERS_SINGLE)
4734                 src->state = PFOTHERS_SINGLE;
4735         if (dst->state == PFOTHERS_SINGLE)
4736                 dst->state = PFOTHERS_MULTIPLE;
4737
4738         /* update expire time */
4739         (*state)->expire = time_second;
4740         if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4741                 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4742         else
4743                 (*state)->timeout = PFTM_OTHER_SINGLE;
4744
4745         /* translate source/destination address, if necessary */
4746         if (STATE_TRANSLATE(*state)) {
4747                 if (direction == PF_OUT)
4748                         switch (pd->af) {
4749 #ifdef INET
4750                         case AF_INET:
4751                                 pf_change_a(&pd->src->v4.s_addr,
4752                                     pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
4753                                     0);
4754                                 break;
4755 #endif /* INET */
4756 #ifdef INET6
4757                         case AF_INET6:
4758                                 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
4759                                 break;
4760 #endif /* INET6 */
4761                         }
4762                 else
4763                         switch (pd->af) {
4764 #ifdef INET
4765                         case AF_INET:
4766                                 pf_change_a(&pd->dst->v4.s_addr,
4767                                     pd->ip_sum, (*state)->lan.addr.v4.s_addr,
4768                                     0);
4769                                 break;
4770 #endif /* INET */
4771 #ifdef INET6
4772                         case AF_INET6:
4773                                 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
4774                                 break;
4775 #endif /* INET6 */
4776                         }
4777         }
4778
4779         return (PF_PASS);
4780 }
4781
4782 /*
4783  * ipoff and off are measured from the start of the mbuf chain.
4784  * h must be at "ipoff" on the mbuf chain.
4785  */
4786 void *
4787 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4788     u_short *actionp, u_short *reasonp, sa_family_t af)
4789 {
4790         switch (af) {
4791 #ifdef INET
4792         case AF_INET: {
4793                 struct ip       *h = mtod(m, struct ip *);
4794                 u_int16_t        fragoff = (h->ip_off & IP_OFFMASK) << 3;
4795
4796                 if (fragoff) {
4797                         if (fragoff >= len)
4798                                 ACTION_SET(actionp, PF_PASS);
4799                         else {
4800                                 ACTION_SET(actionp, PF_DROP);
4801                                 REASON_SET(reasonp, PFRES_FRAG);
4802                         }
4803                         return (NULL);
4804                 }
4805                 if (m->m_pkthdr.len < off + len ||
4806                     h->ip_len < off + len) {
4807                         ACTION_SET(actionp, PF_DROP);
4808                         REASON_SET(reasonp, PFRES_SHORT);
4809                         return (NULL);
4810                 }
4811                 break;
4812         }
4813 #endif /* INET */
4814 #ifdef INET6
4815         case AF_INET6: {
4816                 struct ip6_hdr  *h = mtod(m, struct ip6_hdr *);
4817
4818                 if (m->m_pkthdr.len < off + len ||
4819                     (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
4820                     (unsigned)(off + len)) {
4821                         ACTION_SET(actionp, PF_DROP);
4822                         REASON_SET(reasonp, PFRES_SHORT);
4823                         return (NULL);
4824                 }
4825                 break;
4826         }
4827 #endif /* INET6 */
4828         }
4829         m_copydata(m, off, len, p);
4830         return (p);
4831 }
4832
4833 int
4834 pf_routable(struct pf_addr *addr, sa_family_t af)
4835 {
4836         struct sockaddr_in      *dst;
4837         struct route             ro;
4838         int                      ret = 0;
4839
4840         bzero(&ro, sizeof(ro));
4841         dst = satosin(&ro.ro_dst);
4842         dst->sin_family = af;
4843         dst->sin_len = sizeof(*dst);
4844         dst->sin_addr = addr->v4;
4845         rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
4846
4847         if (ro.ro_rt != NULL) {
4848                 ret = 1;
4849                 RTFREE(ro.ro_rt);
4850         }
4851
4852         return (ret);
4853 }
4854
4855 #ifdef INET
4856 void
4857 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
4858     struct pf_state *s)
4859 {
4860         struct mbuf             *m0, *m1;
4861         struct route             iproute;
4862         struct route            *ro = NULL;
4863         struct sockaddr_in      *dst;
4864         struct ip               *ip;
4865         struct ifnet            *ifp = NULL;
4866         struct pf_addr           naddr;
4867         struct pf_src_node      *sn = NULL;
4868         int                      error = 0;
4869         int sw_csum;
4870
4871         if (m == NULL || *m == NULL || r == NULL ||
4872             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
4873                 panic("pf_route: invalid parameters");
4874
4875         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
4876                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
4877                 (*m)->m_pkthdr.pf_routed = 1;
4878         } else {
4879                 if ((*m)->m_pkthdr.pf_routed > 3) {
4880                         m0 = *m;
4881                         *m = NULL;
4882                         goto bad;
4883                 }
4884                 (*m)->m_pkthdr.pf_routed++;
4885         }
4886
4887         if (r->rt == PF_DUPTO) {
4888                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
4889                         return;
4890         } else {
4891                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
4892                         return;
4893                 m0 = *m;
4894         }
4895
4896         if (m0->m_len < sizeof(struct ip))
4897                 panic("pf_route: m0->m_len < sizeof(struct ip)");
4898         ip = mtod(m0, struct ip *);
4899
4900         ro = &iproute;
4901         bzero((caddr_t)ro, sizeof(*ro));
4902         dst = satosin(&ro->ro_dst);
4903         dst->sin_family = AF_INET;
4904         dst->sin_len = sizeof(*dst);
4905         dst->sin_addr = ip->ip_dst;
4906
4907         if (r->rt == PF_FASTROUTE) {
4908                 rtalloc(ro);
4909                 if (ro->ro_rt == 0) {
4910                         ipstat.ips_noroute++;
4911                         goto bad;
4912                 }
4913
4914                 ifp = ro->ro_rt->rt_ifp;
4915                 ro->ro_rt->rt_use++;
4916
4917                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
4918                         dst = satosin(ro->ro_rt->rt_gateway);
4919         } else {
4920                 if (TAILQ_EMPTY(&r->rpool.list))
4921                         panic("pf_route: TAILQ_EMPTY(&r->rpool.list)");
4922                 if (s == NULL) {
4923                         pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
4924                             &naddr, NULL, &sn);
4925                         if (!PF_AZERO(&naddr, AF_INET))
4926                                 dst->sin_addr.s_addr = naddr.v4.s_addr;
4927                         ifp = r->rpool.cur->kif ?
4928                             r->rpool.cur->kif->pfik_ifp : NULL;
4929                 } else {
4930                         if (!PF_AZERO(&s->rt_addr, AF_INET))
4931                                 dst->sin_addr.s_addr =
4932                                     s->rt_addr.v4.s_addr;
4933                         ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
4934                 }
4935         }
4936         if (ifp == NULL)
4937                 goto bad;
4938
4939         if (oifp != ifp) {
4940                 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS)
4941                         goto bad;
4942                 else if (m0 == NULL)
4943                         goto done;
4944                 if (m0->m_len < sizeof(struct ip))
4945                         panic("pf_route: m0->m_len < sizeof(struct ip)");
4946                 ip = mtod(m0, struct ip *);
4947         }
4948
4949         /* Copied from ip_output. */
4950         m0->m_pkthdr.csum_flags |= CSUM_IP;
4951         sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
4952         if (sw_csum & CSUM_DELAY_DATA) {
4953                 in_delayed_cksum(m0);
4954                 sw_csum &= ~CSUM_DELAY_DATA;
4955         }
4956         m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
4957
4958         /*
4959          * If small enough for interface, or the interface will take
4960          * care of the fragmentation for us, can just send directly.
4961          */
4962         if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
4963                 (ip->ip_off & IP_DF) == 0)) {
4964                 ip->ip_len = htons(ip->ip_len);
4965                 ip->ip_off = htons(ip->ip_off);
4966                 ip->ip_sum = 0;
4967                 if (sw_csum & CSUM_DELAY_IP) {
4968                         /* From KAME */
4969                         if (ip->ip_v == IPVERSION &&
4970                             (ip->ip_hl << 2) == sizeof(*ip)) {
4971                                 ip->ip_sum = in_cksum_hdr(ip);
4972                         } else {
4973                                 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
4974                         }
4975                 }
4976
4977                 lwkt_serialize_enter(ifp->if_serializer);
4978                 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt);
4979                 lwkt_serialize_exit(ifp->if_serializer);
4980                 goto done;
4981         }
4982
4983         /*
4984          * Too large for interface; fragment if possible.
4985          * Must be able to put at least 8 bytes per fragment.
4986          */
4987         if (ip->ip_off & IP_DF) {
4988                 ipstat.ips_cantfrag++;
4989                 if (r->rt != PF_DUPTO) {
4990                         icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
4991                             ifp->if_mtu);
4992                         goto done;
4993                 } else
4994                         goto bad;
4995         }
4996
4997         m1 = m0;
4998         error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
4999         if (error)
5000                 goto bad;
5001
5002         for (m0 = m1; m0; m0 = m1) {
5003                 m1 = m0->m_nextpkt;
5004                 m0->m_nextpkt = 0;
5005                 if (error == 0) {
5006                         lwkt_serialize_enter(ifp->if_serializer);
5007                         error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5008                                     NULL);
5009                         lwkt_serialize_exit(ifp->if_serializer);
5010                 } else {
5011                         m_freem(m0);
5012                 }
5013         }
5014
5015         if (error == 0)
5016                 ipstat.ips_fragmented++;
5017
5018 done:
5019         if (r->rt != PF_DUPTO)
5020                 *m = NULL;
5021         if (ro == &iproute && ro->ro_rt)
5022                 RTFREE(ro->ro_rt);
5023         return;
5024
5025 bad:
5026         m_freem(m0);
5027         goto done;
5028 }
5029 #endif /* INET */
5030
5031 #ifdef INET6
5032 void
5033 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5034     struct pf_state *s)
5035 {
5036         struct mbuf             *m0;
5037         struct route_in6         ip6route;
5038         struct route_in6        *ro;
5039         struct sockaddr_in6     *dst;
5040         struct ip6_hdr          *ip6;
5041         struct ifnet            *ifp = NULL;
5042         struct pf_addr           naddr;
5043         struct pf_src_node      *sn = NULL;
5044         int                      error = 0;
5045
5046         if (m == NULL || *m == NULL || r == NULL ||
5047             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5048                 panic("pf_route6: invalid parameters");
5049
5050         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5051                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5052                 (*m)->m_pkthdr.pf_routed = 1;
5053         } else {
5054                 if ((*m)->m_pkthdr.pf_routed > 3) {
5055                         m0 = *m;
5056                         *m = NULL;
5057                         goto bad;
5058                 }
5059                 (*m)->m_pkthdr.pf_routed++;
5060         }
5061
5062         if (r->rt == PF_DUPTO) {
5063                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5064                         return;
5065         } else {
5066                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5067                         return;
5068                 m0 = *m;
5069         }
5070
5071         if (m0->m_len < sizeof(struct ip6_hdr))
5072                 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5073         ip6 = mtod(m0, struct ip6_hdr *);
5074
5075         ro = &ip6route;
5076         bzero((caddr_t)ro, sizeof(*ro));
5077         dst = (struct sockaddr_in6 *)&ro->ro_dst;
5078         dst->sin6_family = AF_INET6;
5079         dst->sin6_len = sizeof(*dst);
5080         dst->sin6_addr = ip6->ip6_dst;
5081
5082         /* Cheat. */
5083         if (r->rt == PF_FASTROUTE) {
5084                 m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
5085                 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5086                 return;
5087         }
5088
5089         if (TAILQ_EMPTY(&r->rpool.list))
5090                 panic("pf_route6: TAILQ_EMPTY(&r->rpool.list)");
5091         if (s == NULL) {
5092                 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5093                     &naddr, NULL, &sn);
5094                 if (!PF_AZERO(&naddr, AF_INET6))
5095                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5096                             &naddr, AF_INET6);
5097                 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5098         } else {
5099                 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5100                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5101                             &s->rt_addr, AF_INET6);
5102                 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5103         }
5104         if (ifp == NULL)
5105                 goto bad;
5106
5107         if (oifp != ifp) {
5108                 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS)
5109                         goto bad;
5110                 else if (m0 == NULL)
5111                         goto done;
5112                 if (m0->m_len < sizeof(struct ip6_hdr))
5113                         panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5114                 ip6 = mtod(m0, struct ip6_hdr *);
5115         }
5116
5117         /*
5118          * If the packet is too large for the outgoing interface,
5119          * send back an icmp6 error.
5120          */
5121         if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5122                 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5123         if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5124                 error = nd6_output(ifp, ifp, m0, dst, NULL);
5125         } else {
5126                 in6_ifstat_inc(ifp, ifs6_in_toobig);
5127                 if (r->rt != PF_DUPTO)
5128                         icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5129                 else
5130                         goto bad;
5131         }
5132
5133 done:
5134         if (r->rt != PF_DUPTO)
5135                 *m = NULL;
5136         return;
5137
5138 bad:
5139         m_freem(m0);
5140         goto done;
5141 }
5142 #endif /* INET6 */
5143
5144
5145 /*
5146  * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5147  *   off is the offset where the protocol header starts
5148  *   len is the total length of protocol header plus payload
5149  * returns 0 when the checksum is valid, otherwise returns 1.
5150  */
5151 /*
5152  * XXX
5153  * FreeBSD supports cksum offload for the following drivers.
5154  * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5155  * If we can make full use of it we would outperform ipfw/ipfilter in
5156  * very heavy traffic.
5157  * I have not tested 'cause I don't have NICs that supports cksum offload.
5158  * (There might be problems. Typical phenomena would be
5159  *   1. No route message for UDP packet.
5160  *   2. No connection acceptance from external hosts regardless of rule set.)
5161  */
5162 int
5163 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5164     sa_family_t af)
5165 {
5166         u_int16_t sum = 0;
5167         int hw_assist = 0;
5168         struct ip *ip;
5169
5170         if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5171                 return (1);
5172         if (m->m_pkthdr.len < off + len)
5173                 return (1);
5174
5175         switch (p) {
5176         case IPPROTO_TCP:
5177         case IPPROTO_UDP:
5178                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5179                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5180                                 sum = m->m_pkthdr.csum_data;
5181                         } else {
5182                                 ip = mtod(m, struct ip *);
5183                                 sum = in_pseudo(ip->ip_src.s_addr,
5184                                         ip->ip_dst.s_addr, htonl((u_short)len +
5185                                         m->m_pkthdr.csum_data + p));
5186                         }
5187                         sum ^= 0xffff;
5188                         ++hw_assist;
5189                 }
5190                 break;
5191         case IPPROTO_ICMP:
5192 #ifdef INET6
5193         case IPPROTO_ICMPV6:
5194 #endif /* INET6 */
5195                 break;
5196         default:
5197                 return (1);
5198         }
5199
5200         if (!hw_assist) {
5201                 switch (af) {
5202                 case AF_INET:
5203                         if (p == IPPROTO_ICMP) {
5204                                 if (m->m_len < off)
5205                                         return (1);
5206                                 m->m_data += off;
5207                                 m->m_len -= off;
5208                                 sum = in_cksum(m, len);
5209                                 m->m_data -= off;
5210                                 m->m_len += off;
5211                         } else {
5212                                 if (m->m_len < sizeof(struct ip))
5213                                         return (1);
5214                                 sum = in_cksum_range(m, p, off, len);
5215                                 if (sum == 0) {
5216                                         m->m_pkthdr.csum_flags |=
5217                                             (CSUM_DATA_VALID |
5218                                              CSUM_PSEUDO_HDR);
5219                                         m->m_pkthdr.csum_data = 0xffff;
5220                                 }
5221                         }
5222                         break;
5223 #ifdef INET6
5224                 case AF_INET6:
5225                         if (m->m_len < sizeof(struct ip6_hdr))
5226                                 return (1);
5227                         sum = in6_cksum(m, p, off, len);
5228                         /*
5229                          * XXX
5230                          * IPv6 H/W cksum off-load not supported yet!
5231                          *
5232                          * if (sum == 0) {
5233                          *      m->m_pkthdr.csum_flags |=
5234                          *          (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5235                          *      m->m_pkthdr.csum_data = 0xffff;
5236                          *}
5237                          */
5238                         break;
5239 #endif /* INET6 */
5240                 default:
5241                         return (1);
5242                 }
5243         }
5244         if (sum) {
5245                 switch (p) {
5246                 case IPPROTO_TCP:
5247                         tcpstat.tcps_rcvbadsum++;
5248                         break;
5249                 case IPPROTO_UDP:
5250                         udpstat.udps_badsum++;
5251                         break;
5252                 case IPPROTO_ICMP:
5253                         icmpstat.icps_checksum++;
5254                         break;
5255 #ifdef INET6
5256                 case IPPROTO_ICMPV6:
5257                         icmp6stat.icp6s_checksum++;
5258                         break;
5259 #endif /* INET6 */
5260                 }
5261                 return (1);
5262         }
5263         return (0);
5264 }
5265
5266 #ifdef INET
5267 int
5268 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0)
5269 {
5270         struct pfi_kif          *kif;
5271         u_short                  action, reason = 0, log = 0;
5272         struct mbuf             *m = *m0;
5273         struct ip               *h = NULL;
5274         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5275         struct pf_state         *s = NULL;
5276         struct pf_ruleset       *ruleset = NULL;
5277         struct pf_pdesc          pd;
5278         int                      off, dirndx, pqid = 0;
5279
5280         if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5281                 return (PF_PASS);
5282
5283         kif = pfi_index2kif[ifp->if_index];
5284         if (kif == NULL)
5285                 return (PF_DROP);
5286
5287 #ifdef DIAGNOSTIC
5288         if ((m->m_flags & M_PKTHDR) == 0)
5289                 panic("non-M_PKTHDR is passed to pf_test");
5290 #endif
5291
5292         memset(&pd, 0, sizeof(pd));
5293         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5294                 action = PF_DROP;
5295                 REASON_SET(&reason, PFRES_SHORT);
5296                 log = 1;
5297                 goto done;
5298         }
5299
5300         /* We do IP header normalization and packet reassembly here */
5301         if (pf_normalize_ip(m0, dir, kif, &reason) != PF_PASS) {
5302                 action = PF_DROP;
5303                 goto done;
5304         }
5305         m = *m0;
5306         h = mtod(m, struct ip *);
5307
5308         off = h->ip_hl << 2;
5309         if (off < (int)sizeof(*h)) {
5310                 action = PF_DROP;
5311                 REASON_SET(&reason, PFRES_SHORT);
5312                 log = 1;
5313                 goto done;
5314         }
5315
5316         pd.src = (struct pf_addr *)&h->ip_src;
5317         pd.dst = (struct pf_addr *)&h->ip_dst;
5318         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5319         pd.ip_sum = &h->ip_sum;
5320         pd.proto = h->ip_p;
5321         pd.af = AF_INET;
5322         pd.tos = h->ip_tos;
5323         pd.tot_len = h->ip_len;
5324
5325         /* handle fragments that didn't get reassembled by normalization */
5326         if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5327                 action = pf_test_fragment(&r, dir, kif, m, h,
5328                     &pd, &a, &ruleset);
5329                 goto done;
5330         }
5331
5332         switch (h->ip_p) {
5333
5334         case IPPROTO_TCP: {
5335                 struct tcphdr   th;
5336
5337                 pd.hdr.tcp = &th;
5338                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5339                     &action, &reason, AF_INET)) {
5340                         log = action != PF_PASS;
5341                         goto done;
5342                 }
5343                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5344                     h->ip_len - off, IPPROTO_TCP, AF_INET)) {
5345                         action = PF_DROP;
5346                         goto done;
5347                 }
5348                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5349                 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5350                         pqid = 1;
5351                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5352                 if (action == PF_DROP)
5353                         goto done;
5354                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5355                     &reason);
5356                 if (action == PF_PASS) {
5357 #if NPFSYNC
5358                         pfsync_update_state(s);
5359 #endif
5360                         r = s->rule.ptr;
5361                         a = s->anchor.ptr;
5362                         log = s->log;
5363                 } else if (s == NULL)
5364                         action = pf_test_tcp(&r, &s, dir, kif,
5365                             m, off, h, &pd, &a, &ruleset);
5366                 break;
5367         }
5368
5369         case IPPROTO_UDP: {
5370                 struct udphdr   uh;
5371
5372                 pd.hdr.udp = &uh;
5373                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5374                     &action, &reason, AF_INET)) {
5375                         log = action != PF_PASS;
5376                         goto done;
5377                 }
5378                 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5379                     off, h->ip_len - off, IPPROTO_UDP, AF_INET)) {
5380                         action = PF_DROP;
5381                         goto done;
5382                 }
5383                 if (uh.uh_dport == 0 ||
5384                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5385                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5386                         action = PF_DROP;
5387                         goto done;
5388                 }
5389                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5390                 if (action == PF_PASS) {
5391 #if NPFSYNC
5392                         pfsync_update_state(s);
5393 #endif
5394                         r = s->rule.ptr;
5395                         a = s->anchor.ptr;
5396                         log = s->log;
5397                 } else if (s == NULL)
5398                         action = pf_test_udp(&r, &s, dir, kif,
5399                             m, off, h, &pd, &a, &ruleset);
5400                 break;
5401         }
5402
5403         case IPPROTO_ICMP: {
5404                 struct icmp     ih;
5405
5406                 pd.hdr.icmp = &ih;
5407                 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5408                     &action, &reason, AF_INET)) {
5409                         log = action != PF_PASS;
5410                         goto done;
5411                 }
5412                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5413                     h->ip_len - off, IPPROTO_ICMP, AF_INET)) {
5414                         action = PF_DROP;
5415                         goto done;
5416                 }
5417                 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd);
5418                 if (action == PF_PASS) {
5419 #if NPFSYNC
5420                         pfsync_update_state(s);
5421 #endif
5422                         r = s->rule.ptr;
5423                         a = s->anchor.ptr;
5424                         log = s->log;
5425                 } else if (s == NULL)
5426                         action = pf_test_icmp(&r, &s, dir, kif,
5427                             m, off, h, &pd, &a, &ruleset);
5428                 break;
5429         }
5430
5431         default:
5432                 action = pf_test_state_other(&s, dir, kif, &pd);
5433                 if (action == PF_PASS) {
5434 #if NPFSYNC
5435                         pfsync_update_state(s);
5436 #endif
5437                         r = s->rule.ptr;
5438                         a = s->anchor.ptr;
5439                         log = s->log;
5440                 } else if (s == NULL)
5441                         action = pf_test_other(&r, &s, dir, kif, m, off, h,
5442                             &pd, &a, &ruleset);
5443                 break;
5444         }
5445
5446 done:
5447         if (action == PF_PASS && h->ip_hl > 5 &&
5448             !((s && s->allow_opts) || r->allow_opts)) {
5449                 action = PF_DROP;
5450                 REASON_SET(&reason, PFRES_SHORT);
5451                 log = 1;
5452                 DPFPRINTF(PF_DEBUG_MISC,
5453                     ("pf: dropping packet with ip options\n"));
5454         }
5455
5456 #ifdef ALTQ
5457         if (action == PF_PASS && r->qid) {
5458                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5459                 if (pd.tos == IPTOS_LOWDELAY)
5460                         m->m_pkthdr.altq_qid = r->pqid;
5461                 else
5462                         m->m_pkthdr.altq_qid = r->qid;
5463                 m->m_pkthdr.ecn_af = AF_INET;
5464                 m->m_pkthdr.header = h;
5465         }
5466 #endif
5467
5468         /*
5469          * connections redirected to loopback should not match sockets
5470          * bound specifically to loopback due to security implications,
5471          * see tcp_input() and in_pcblookup_listen().
5472          */
5473         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5474             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5475             (s->nat_rule.ptr->action == PF_RDR ||
5476             s->nat_rule.ptr->action == PF_BINAT) &&
5477             (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
5478                 action = PF_DROP;
5479                 REASON_SET(&reason, PFRES_MEMORY);
5480         }
5481
5482         m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5483
5484         if (log)
5485                 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, r, a, ruleset);
5486
5487         kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5488         kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5489
5490         if (action == PF_PASS || r->action == PF_DROP) {
5491                 r->packets++;
5492                 r->bytes += pd.tot_len;
5493                 if (a != NULL) {
5494                         a->packets++;
5495                         a->bytes += pd.tot_len;
5496                 }
5497                 if (s != NULL) {
5498                         dirndx = (dir == s->direction) ? 0 : 1;
5499                         s->packets[dirndx]++;
5500                         s->bytes[dirndx] += pd.tot_len;
5501                         if (s->nat_rule.ptr != NULL) {
5502                                 s->nat_rule.ptr->packets++;
5503                                 s->nat_rule.ptr->bytes += pd.tot_len;
5504                         }
5505                         if (s->src_node != NULL) {
5506                                 s->src_node->packets++;
5507                                 s->src_node->bytes += pd.tot_len;
5508                         }
5509                         if (s->nat_src_node != NULL) {
5510                                 s->nat_src_node->packets++;
5511                                 s->nat_src_node->bytes += pd.tot_len;
5512                         }
5513                 }
5514                 tr = r;
5515                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5516                 if (nr != NULL) {
5517                         struct pf_addr *x;
5518                         /*
5519                          * XXX: we need to make sure that the addresses
5520                          * passed to pfr_update_stats() are the same than
5521                          * the addresses used during matching (pfr_match)
5522                          */
5523                         if (r == &pf_default_rule) {
5524                                 tr = nr;
5525                                 x = (s == NULL || s->direction == dir) ?
5526                                     &pd.baddr : &pd.naddr;
5527                         } else
5528                                 x = (s == NULL || s->direction == dir) ?
5529                                     &pd.naddr : &pd.baddr;
5530                         if (x == &pd.baddr || s == NULL) {
5531                                 /* we need to change the address */
5532                                 if (dir == PF_OUT)
5533                                         pd.src = x;
5534                                 else
5535                                         pd.dst = x;
5536                         }
5537                 }
5538                 if (tr->src.addr.type == PF_ADDR_TABLE)
5539                         pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5540                             s->direction == dir) ? pd.src : pd.dst, pd.af,
5541                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5542                             tr->src.not);
5543                 if (tr->dst.addr.type == PF_ADDR_TABLE)
5544                         pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5545                             s->direction == dir) ? pd.dst : pd.src, pd.af,
5546                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5547                             tr->dst.not);
5548         }
5549
5550
5551         if (action == PF_SYNPROXY_DROP) {
5552                 m_freem(*m0);
5553                 *m0 = NULL;
5554                 action = PF_PASS;
5555         } else if (r->rt)
5556                 /* pf_route can free the mbuf causing *m0 to become NULL */
5557                 pf_route(m0, r, dir, ifp, s);
5558
5559         return (action);
5560 }
5561 #endif /* INET */
5562
5563 #ifdef INET6
5564 int
5565 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0)
5566 {
5567         struct pfi_kif          *kif;
5568         u_short                  action, reason = 0, log = 0;
5569         struct mbuf             *m = *m0;
5570         struct ip6_hdr          *h = NULL;
5571         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5572         struct pf_state         *s = NULL;
5573         struct pf_ruleset       *ruleset = NULL;
5574         struct pf_pdesc          pd;
5575         int                      off, terminal = 0, dirndx;
5576
5577         if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5578                 return (PF_PASS);
5579
5580         kif = pfi_index2kif[ifp->if_index];
5581         if (kif == NULL)
5582                 return (PF_DROP);
5583
5584 #ifdef DIAGNOSTIC
5585         if ((m->m_flags & M_PKTHDR) == 0)
5586                 panic("non-M_PKTHDR is passed to pf_test");
5587 #endif
5588
5589         memset(&pd, 0, sizeof(pd));
5590         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5591                 action = PF_DROP;
5592                 REASON_SET(&reason, PFRES_SHORT);
5593                 log = 1;
5594                 goto done;
5595         }
5596
5597         /* We do IP header normalization and packet reassembly here */
5598         if (pf_normalize_ip6(m0, dir, kif, &reason) != PF_PASS) {
5599                 action = PF_DROP;
5600                 goto done;
5601         }
5602         m = *m0;
5603         h = mtod(m, struct ip6_hdr *);
5604
5605         pd.src = (struct pf_addr *)&h->ip6_src;
5606         pd.dst = (struct pf_addr *)&h->ip6_dst;
5607         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
5608         pd.ip_sum = NULL;
5609         pd.af = AF_INET6;
5610         pd.tos = 0;
5611         pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5612
5613         off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5614         pd.proto = h->ip6_nxt;
5615         do {
5616                 switch (pd.proto) {
5617                 case IPPROTO_FRAGMENT:
5618                         action = pf_test_fragment(&r, dir, kif, m, h,
5619                             &pd, &a, &ruleset);
5620                         if (action == PF_DROP)
5621                                 REASON_SET(&reason, PFRES_FRAG);
5622                         goto done;
5623                 case IPPROTO_AH:
5624                 case IPPROTO_HOPOPTS:
5625                 case IPPROTO_ROUTING:
5626                 case IPPROTO_DSTOPTS: {
5627                         /* get next header and header length */
5628                         struct ip6_ext  opt6;
5629
5630                         if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
5631                             NULL, NULL, pd.af)) {
5632                                 DPFPRINTF(PF_DEBUG_MISC,
5633                                     ("pf: IPv6 short opt\n"));
5634                                 action = PF_DROP;
5635                                 REASON_SET(&reason, PFRES_SHORT);
5636                                 log = 1;
5637                                 goto done;
5638                         }
5639                         if (pd.proto == IPPROTO_AH)
5640                                 off += (opt6.ip6e_len + 2) * 4;
5641                         else
5642                                 off += (opt6.ip6e_len + 1) * 8;
5643                         pd.proto = opt6.ip6e_nxt;
5644                         /* goto the next header */
5645                         break;
5646                 }
5647                 default:
5648                         terminal++;
5649                         break;
5650                 }
5651         } while (!terminal);
5652
5653         switch (pd.proto) {
5654
5655         case IPPROTO_TCP: {
5656                 struct tcphdr   th;
5657
5658                 pd.hdr.tcp = &th;
5659                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5660                     &action, &reason, AF_INET6)) {
5661                         log = action != PF_PASS;
5662                         goto done;
5663                 }
5664                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5665                     ntohs(h->ip6_plen), IPPROTO_TCP, AF_INET6)) {
5666                         action = PF_DROP;
5667                         goto done;
5668                 }
5669                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5670                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5671                 if (action == PF_DROP)
5672                         goto done;
5673                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5674                     &reason);
5675                 if (action == PF_PASS) {
5676 #if NPFSYNC
5677                         pfsync_update_state(s);
5678 #endif
5679                         r = s->rule.ptr;
5680                         a = s->anchor.ptr;
5681                         log = s->log;
5682                 } else if (s == NULL)
5683                         action = pf_test_tcp(&r, &s, dir, kif,
5684                             m, off, h, &pd, &a, &ruleset);
5685                 break;
5686         }
5687
5688         case IPPROTO_UDP: {
5689                 struct udphdr   uh;
5690
5691                 pd.hdr.udp = &uh;
5692                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5693                     &action, &reason, AF_INET6)) {
5694                         log = action != PF_PASS;
5695                         goto done;
5696                 }
5697                 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5698                     off, ntohs(h->ip6_plen), IPPROTO_UDP, AF_INET6)) {
5699                         action = PF_DROP;
5700                         goto done;
5701                 }
5702                 if (uh.uh_dport == 0 ||
5703                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5704                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5705                         action = PF_DROP;
5706                         goto done;
5707                 }
5708                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5709                 if (action == PF_PASS) {
5710 #if NPFSYNC
5711                         pfsync_update_state(s);
5712 #endif
5713                         r = s->rule.ptr;
5714                         a = s->anchor.ptr;
5715                         log = s->log;
5716                 } else if (s == NULL)
5717                         action = pf_test_udp(&r, &s, dir, kif,
5718                             m, off, h, &pd, &a, &ruleset);
5719                 break;
5720         }
5721
5722         case IPPROTO_ICMPV6: {
5723                 struct icmp6_hdr        ih;
5724
5725                 pd.hdr.icmp6 = &ih;
5726                 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
5727                     &action, &reason, AF_INET6)) {
5728                         log = action != PF_PASS;
5729                         goto done;
5730                 }
5731                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5732                     ntohs(h->ip6_plen), IPPROTO_ICMPV6, AF_INET6)) {
5733                         action = PF_DROP;
5734                         goto done;
5735                 }
5736                 action = pf_test_state_icmp(&s, dir, kif,
5737                     m, off, h, &pd);
5738                 if (action == PF_PASS) {
5739 #if NPFSYNC
5740                         pfsync_update_state(s);
5741 #endif
5742                         r = s->rule.ptr;
5743                         a = s->anchor.ptr;
5744                         log = s->log;
5745                 } else if (s == NULL)
5746                         action = pf_test_icmp(&r, &s, dir, kif,
5747                             m, off, h, &pd, &a, &ruleset);
5748                 break;
5749         }
5750
5751         default:
5752                 action = pf_test_state_other(&s, dir, kif, &pd);
5753                 if (action == PF_PASS) {
5754                         r = s->rule.ptr;
5755                         a = s->anchor.ptr;
5756                         log = s->log;
5757                 } else if (s == NULL)
5758                         action = pf_test_other(&r, &s, dir, kif, m, off, h,
5759                             &pd, &a, &ruleset);
5760                 break;
5761         }
5762
5763 done:
5764         /* XXX handle IPv6 options, if not allowed. not implemented. */
5765
5766 #ifdef ALTQ
5767         if (action == PF_PASS && r->qid) {
5768                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5769                 if (pd.tos == IPTOS_LOWDELAY)
5770                         m->m_pkthdr.altq_qid = r->pqid;
5771                 else
5772                         m->m_pkthdr.altq_qid = r->qid;
5773                 m->m_pkthdr.ecn_af = AF_INET6;
5774                 m->m_pkthdr.header = h;
5775         }
5776 #endif
5777
5778         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5779             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5780             (s->nat_rule.ptr->action == PF_RDR ||
5781             s->nat_rule.ptr->action == PF_BINAT) &&
5782             IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) {
5783                 action = PF_DROP;
5784                 REASON_SET(&reason, PFRES_MEMORY);
5785         }
5786
5787         m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5788
5789         if (log)
5790                 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, r, a, ruleset);
5791
5792         kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5793         kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
5794
5795         if (action == PF_PASS || r->action == PF_DROP) {
5796                 r->packets++;
5797                 r->bytes += pd.tot_len;
5798                 if (a != NULL) {
5799                         a->packets++;
5800                         a->bytes += pd.tot_len;
5801                 }
5802                 if (s != NULL) {
5803                         dirndx = (dir == s->direction) ? 0 : 1;
5804                         s->packets[dirndx]++;
5805                         s->bytes[dirndx] += pd.tot_len;
5806                         if (s->nat_rule.ptr != NULL) {
5807                                 s->nat_rule.ptr->packets++;
5808                                 s->nat_rule.ptr->bytes += pd.tot_len;
5809                         }
5810                         if (s->src_node != NULL) {
5811                                 s->src_node->packets++;
5812                                 s->src_node->bytes += pd.tot_len;
5813                         }
5814                         if (s->nat_src_node != NULL) {
5815                                 s->nat_src_node->packets++;
5816                                 s->nat_src_node->bytes += pd.tot_len;
5817                         }
5818                 }
5819                 tr = r;
5820                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5821                 if (nr != NULL) {
5822                         struct pf_addr *x;
5823                         /*
5824                          * XXX: we need to make sure that the addresses
5825                          * passed to pfr_update_stats() are the same than
5826                          * the addresses used during matching (pfr_match)
5827                          */
5828                         if (r == &pf_default_rule) {
5829                                 tr = nr;
5830                                 x = (s == NULL || s->direction == dir) ?
5831                                     &pd.baddr : &pd.naddr;
5832                         } else {
5833                                 x = (s == NULL || s->direction == dir) ?
5834                                     &pd.naddr : &pd.baddr;
5835                         }
5836                         if (x == &pd.baddr || s == NULL) {
5837                                 if (dir == PF_OUT)
5838                                         pd.src = x;
5839                                 else
5840                                         pd.dst = x;
5841                         }
5842                 }
5843                 if (tr->src.addr.type == PF_ADDR_TABLE)
5844                         pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5845                             s->direction == dir) ? pd.src : pd.dst, pd.af,
5846                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5847                             tr->src.not);
5848                 if (tr->dst.addr.type == PF_ADDR_TABLE)
5849                         pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5850                             s->direction == dir) ? pd.dst : pd.src, pd.af,
5851                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5852                             tr->dst.not);
5853         }
5854
5855
5856         if (action == PF_SYNPROXY_DROP) {
5857                 m_freem(*m0);
5858                 *m0 = NULL;
5859                 action = PF_PASS;
5860         } else if (r->rt)
5861                 /* pf_route6 can free the mbuf causing *m0 to become NULL */
5862                 pf_route6(m0, r, dir, ifp, s);
5863
5864         return (action);
5865 }
5866 #endif /* INET6 */