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 1.8 2005/12/11 13:00:17 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)) printf 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                                 printf("pf: src_tree insert failed: ");
 603                                 pf_print_host(&(*sn)->addr, 0, af);
 604                                 printf("\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                         printf("pf: state insert failed: tree_lan_ext");
 631                         printf(" lan: ");
 632                         pf_print_host(&state->lan.addr, state->lan.port,
 633                             state->af);
 634                         printf(" gwy: ");
 635                         pf_print_host(&state->gwy.addr, state->gwy.port,
 636                             state->af);
 637                         printf(" ext: ");
 638                         pf_print_host(&state->ext.addr, state->ext.port,
 639                             state->af);
 640                         if (state->sync_flags & PFSTATE_FROMSYNC)
 641                                 printf(" (from sync)");
 642                         printf("\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                         printf("pf: state insert failed: tree_ext_gwy");
 650                         printf(" lan: ");
 651                         pf_print_host(&state->lan.addr, state->lan.port,
 652                             state->af);
 653                         printf(" gwy: ");
 654                         pf_print_host(&state->gwy.addr, state->gwy.port,
 655                             state->af);
 656                         printf(" ext: ");
 657                         pf_print_host(&state->ext.addr, state->ext.port,
 658                             state->af);
 659                         if (state->sync_flags & PFSTATE_FROMSYNC)
 660                                 printf(" (from sync)");
 661                         printf("\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                         printf("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                                 printf(" (from sync)");
 678                         printf("\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 void
 797 pf_purge_expired_states(void)
 798 {
 799         struct pf_state         *cur, *next;
 800
 801         for (cur = RB_MIN(pf_state_tree_id, &tree_id);
 802             cur; cur = next) {
 803                 next = RB_NEXT(pf_state_tree_id, &tree_id, cur);
 804
 805                 if (pf_state_expires(cur) <= time_second) {
 806                         if (cur->src.state == PF_TCPS_PROXY_DST)
 807                                 pf_send_tcp(cur->rule.ptr, cur->af,
 808                                     &cur->ext.addr, &cur->lan.addr,
 809                                     cur->ext.port, cur->lan.port,
 810                                     cur->src.seqhi, cur->src.seqlo + 1, 0,
 811                                     TH_RST|TH_ACK, 0, 0);
 812                         RB_REMOVE(pf_state_tree_ext_gwy,
 813                             &cur->u.s.kif->pfik_ext_gwy, cur);
 814                         RB_REMOVE(pf_state_tree_lan_ext,
 815                             &cur->u.s.kif->pfik_lan_ext, cur);
 816                         RB_REMOVE(pf_state_tree_id, &tree_id, cur);
 817 #if NPFSYNC
 818                         pfsync_delete_state(cur);
 819 #endif
 820                         pf_src_tree_remove_state(cur);
 821                         if (--cur->rule.ptr->states <= 0 &&
 822                             cur->rule.ptr->src_nodes <= 0)
 823                                 pf_rm_rule(NULL, cur->rule.ptr);
 824                         if (cur->nat_rule.ptr != NULL)
 825                                 if (--cur->nat_rule.ptr->states <= 0 &&
 826                                         cur->nat_rule.ptr->src_nodes <= 0)
 827                                         pf_rm_rule(NULL, cur->nat_rule.ptr);
 828                         if (cur->anchor.ptr != NULL)
 829                                 if (--cur->anchor.ptr->states <= 0)
 830                                         pf_rm_rule(NULL, cur->anchor.ptr);
 831                         pf_normalize_tcp_cleanup(cur);
 832                         pfi_detach_state(cur->u.s.kif);
 833                         TAILQ_REMOVE(&state_updates, cur, u.s.entry_updates);
 834                         pool_put(&pf_state_pl, cur);
 835                         pf_status.fcounters[FCNT_STATE_REMOVALS]++;
 836                         pf_status.states--;
 837                 }
 838         }
 839 }
 840
 841 int
 842 pf_tbladdr_setup(struct pf_ruleset *rs, struct pf_addr_wrap *aw)
 843 {
 844         if (aw->type != PF_ADDR_TABLE)
 845                 return (0);
 846         if ((aw->p.tbl = pfr_attach_table(rs, aw->v.tblname)) == NULL)
 847                 return (1);
 848         return (0);
 849 }
 850
 851 void
 852 pf_tbladdr_remove(struct pf_addr_wrap *aw)
 853 {
 854         if (aw->type != PF_ADDR_TABLE || aw->p.tbl == NULL)
 855                 return;
 856         pfr_detach_table(aw->p.tbl);
 857         aw->p.tbl = NULL;
 858 }
 859
 860 void
 861 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
 862 {
 863         struct pfr_ktable *kt = aw->p.tbl;
 864
 865         if (aw->type != PF_ADDR_TABLE || kt == NULL)
 866                 return;
 867         if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
 868                 kt = kt->pfrkt_root;
 869         aw->p.tbl = NULL;
 870         aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
 871                 kt->pfrkt_cnt : -1;
 872 }
 873
 874 void
 875 pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af)
 876 {
 877         switch (af) {
 878 #ifdef INET
 879         case AF_INET: {
 880                 u_int32_t a = ntohl(addr->addr32[0]);
 881                 printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255,
 882                     (a>>8)&255, a&255);
 883                 if (p) {
 884                         p = ntohs(p);
 885                         printf(":%u", p);
 886                 }
 887                 break;
 888         }
 889 #endif /* INET */
 890 #ifdef INET6
 891         case AF_INET6: {
 892                 u_int16_t b;
 893                 u_int8_t i, curstart = 255, curend = 0,
 894                     maxstart = 0, maxend = 0;
 895                 for (i = 0; i < 8; i++) {
 896                         if (!addr->addr16[i]) {
 897                                 if (curstart == 255)
 898                                         curstart = i;
 899                                 else
 900                                         curend = i;
 901                         } else {
 902                                 if (curstart) {
 903                                         if ((curend - curstart) >
 904                                             (maxend - maxstart)) {
 905                                                 maxstart = curstart;
 906                                                 maxend = curend;
 907                                                 curstart = 255;
 908                                         }
 909                                 }
 910                         }
 911                 }
 912                 for (i = 0; i < 8; i++) {
 913                         if (i >= maxstart && i <= maxend) {
 914                                 if (maxend != 7) {
 915                                         if (i == maxstart)
 916                                                 printf(":");
 917                                 } else {
 918                                         if (i == maxend)
 919                                                 printf(":");
 920                                 }
 921                         } else {
 922                                 b = ntohs(addr->addr16[i]);
 923                                 printf("%x", b);
 924                                 if (i < 7)
 925                                         printf(":");
 926                         }
 927                 }
 928                 if (p) {
 929                         p = ntohs(p);
 930                         printf("[%u]", p);
 931                 }
 932                 break;
 933         }
 934 #endif /* INET6 */
 935         }
 936 }
 937
 938 void
 939 pf_print_state(struct pf_state *s)
 940 {
 941         switch (s->proto) {
 942         case IPPROTO_TCP:
 943                 printf("TCP ");
 944                 break;
 945         case IPPROTO_UDP:
 946                 printf("UDP ");
 947                 break;
 948         case IPPROTO_ICMP:
 949                 printf("ICMP ");
 950                 break;
 951         case IPPROTO_ICMPV6:
 952                 printf("ICMPV6 ");
 953                 break;
 954         default:
 955                 printf("%u ", s->proto);
 956                 break;
 957         }
 958         pf_print_host(&s->lan.addr, s->lan.port, s->af);
 959         printf(" ");
 960         pf_print_host(&s->gwy.addr, s->gwy.port, s->af);
 961         printf(" ");
 962         pf_print_host(&s->ext.addr, s->ext.port, s->af);
 963         printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo,
 964             s->src.seqhi, s->src.max_win, s->src.seqdiff);
 965         if (s->src.wscale && s->dst.wscale)
 966                 printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK);
 967         printf("]");
 968         printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo,
 969             s->dst.seqhi, s->dst.max_win, s->dst.seqdiff);
 970         if (s->src.wscale && s->dst.wscale)
 971                 printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK);
 972         printf("]");
 973         printf(" %u:%u", s->src.state, s->dst.state);
 974 }
 975
 976 void
 977 pf_print_flags(u_int8_t f)
 978 {
 979         if (f)
 980                 printf(" ");
 981         if (f & TH_FIN)
 982                 printf("F");
 983         if (f & TH_SYN)
 984                 printf("S");
 985         if (f & TH_RST)
 986                 printf("R");
 987         if (f & TH_PUSH)
 988                 printf("P");
 989         if (f & TH_ACK)
 990                 printf("A");
 991         if (f & TH_URG)
 992                 printf("U");
 993         if (f & TH_ECE)
 994                 printf("E");
 995         if (f & TH_CWR)
 996                 printf("W");
 997 }
 998
 999 #define PF_SET_SKIP_STEPS(i)                                    \
1000         do {                                                    \
1001                 while (head[i] != cur) {                        \
1002                         head[i]->skip[i].ptr = cur;             \
1003                         head[i] = TAILQ_NEXT(head[i], entries); \
1004                 }                                               \
1005         } while (0)
1006
1007 void
1008 pf_calc_skip_steps(struct pf_rulequeue *rules)
1009 {
1010         struct pf_rule *cur, *prev, *head[PF_SKIP_COUNT];
1011         int i;
1012
1013         cur = TAILQ_FIRST(rules);
1014         prev = cur;
1015         for (i = 0; i < PF_SKIP_COUNT; ++i)
1016                 head[i] = cur;
1017         while (cur != NULL) {
1018
1019                 if (cur->kif != prev->kif || cur->ifnot != prev->ifnot)
1020                         PF_SET_SKIP_STEPS(PF_SKIP_IFP);
1021                 if (cur->direction != prev->direction)
1022                         PF_SET_SKIP_STEPS(PF_SKIP_DIR);
1023                 if (cur->af != prev->af)
1024                         PF_SET_SKIP_STEPS(PF_SKIP_AF);
1025                 if (cur->proto != prev->proto)
1026                         PF_SET_SKIP_STEPS(PF_SKIP_PROTO);
1027                 if (cur->src.not != prev->src.not ||
1028                     pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr))
1029                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR);
1030                 if (cur->src.port[0] != prev->src.port[0] ||
1031                     cur->src.port[1] != prev->src.port[1] ||
1032                     cur->src.port_op != prev->src.port_op)
1033                         PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT);
1034                 if (cur->dst.not != prev->dst.not ||
1035                     pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr))
1036                         PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR);
1037                 if (cur->dst.port[0] != prev->dst.port[0] ||
1038                     cur->dst.port[1] != prev->dst.port[1] ||
1039                     cur->dst.port_op != prev->dst.port_op)
1040                         PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT);
1041
1042                 prev = cur;
1043                 cur = TAILQ_NEXT(cur, entries);
1044         }
1045         for (i = 0; i < PF_SKIP_COUNT; ++i)
1046                 PF_SET_SKIP_STEPS(i);
1047 }
1048
1049 int
1050 pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2)
1051 {
1052         if (aw1->type != aw2->type)
1053                 return (1);
1054         switch (aw1->type) {
1055         case PF_ADDR_ADDRMASK:
1056                 if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, 0))
1057                         return (1);
1058                 if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, 0))
1059                         return (1);
1060                 return (0);
1061         case PF_ADDR_DYNIFTL:
1062                 return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt);
1063         case PF_ADDR_NOROUTE:
1064                 return (0);
1065         case PF_ADDR_TABLE:
1066                 return (aw1->p.tbl != aw2->p.tbl);
1067         default:
1068                 printf("invalid address type: %d\n", aw1->type);
1069                 return (1);
1070         }
1071 }
1072
1073 void
1074 pf_update_anchor_rules(void)
1075 {
1076         struct pf_rule  *rule;
1077         int              i;
1078
1079         for (i = 0; i < PF_RULESET_MAX; ++i)
1080                 TAILQ_FOREACH(rule, pf_main_ruleset.rules[i].active.ptr,
1081                     entries)
1082                         if (rule->anchorname[0])
1083                                 rule->anchor = pf_find_anchor(rule->anchorname);
1084                         else
1085                                 rule->anchor = NULL;
1086 }
1087
1088 u_int16_t
1089 pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp)
1090 {
1091         u_int32_t       l;
1092
1093         if (udp && !cksum)
1094                 return (0x0000);
1095         l = cksum + old - new;
1096         l = (l >> 16) + (l & 65535);
1097         l = l & 65535;
1098         if (udp && !l)
1099                 return (0xFFFF);
1100         return (l);
1101 }
1102
1103 void
1104 pf_change_ap(struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc,
1105     struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af)
1106 {
1107         struct pf_addr  ao;
1108         u_int16_t       po = *p;
1109
1110         PF_ACPY(&ao, a, af);
1111         PF_ACPY(a, an, af);
1112
1113         *p = pn;
1114
1115         switch (af) {
1116 #ifdef INET
1117         case AF_INET:
1118                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1119                     ao.addr16[0], an->addr16[0], 0),
1120                     ao.addr16[1], an->addr16[1], 0);
1121                 *p = pn;
1122                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1123                     ao.addr16[0], an->addr16[0], u),
1124                     ao.addr16[1], an->addr16[1], u),
1125                     po, pn, u);
1126                 break;
1127 #endif /* INET */
1128 #ifdef INET6
1129         case AF_INET6:
1130                 *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1131                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1132                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(*pc,
1133                     ao.addr16[0], an->addr16[0], u),
1134                     ao.addr16[1], an->addr16[1], u),
1135                     ao.addr16[2], an->addr16[2], u),
1136                     ao.addr16[3], an->addr16[3], u),
1137                     ao.addr16[4], an->addr16[4], u),
1138                     ao.addr16[5], an->addr16[5], u),
1139                     ao.addr16[6], an->addr16[6], u),
1140                     ao.addr16[7], an->addr16[7], u),
1141                     po, pn, u);
1142                 break;
1143 #endif /* INET6 */
1144         }
1145 }
1146
1147
1148 /* Changes a u_int32_t.  Uses a void * so there are no align restrictions */
1149 void
1150 pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u)
1151 {
1152         u_int32_t       ao;
1153
1154         memcpy(&ao, a, sizeof(ao));
1155         memcpy(a, &an, sizeof(u_int32_t));
1156         *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u),
1157             ao % 65536, an % 65536, u);
1158 }
1159
1160 #ifdef INET6
1161 void
1162 pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u)
1163 {
1164         struct pf_addr  ao;
1165
1166         PF_ACPY(&ao, a, AF_INET6);
1167         PF_ACPY(a, an, AF_INET6);
1168
1169         *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1170             pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1171             pf_cksum_fixup(pf_cksum_fixup(*c,
1172             ao.addr16[0], an->addr16[0], u),
1173             ao.addr16[1], an->addr16[1], u),
1174             ao.addr16[2], an->addr16[2], u),
1175             ao.addr16[3], an->addr16[3], u),
1176             ao.addr16[4], an->addr16[4], u),
1177             ao.addr16[5], an->addr16[5], u),
1178             ao.addr16[6], an->addr16[6], u),
1179             ao.addr16[7], an->addr16[7], u);
1180 }
1181 #endif /* INET6 */
1182
1183 void
1184 pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa,
1185     struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c,
1186     u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af)
1187 {
1188         struct pf_addr  oia, ooa;
1189
1190         PF_ACPY(&oia, ia, af);
1191         PF_ACPY(&ooa, oa, af);
1192
1193         /* Change inner protocol port, fix inner protocol checksum. */
1194         if (ip != NULL) {
1195                 u_int16_t       oip = *ip;
1196                 u_int32_t       opc = 0;
1197
1198                 if (pc != NULL)
1199                         opc = *pc;
1200                 *ip = np;
1201                 if (pc != NULL)
1202                         *pc = pf_cksum_fixup(*pc, oip, *ip, u);
1203                 *ic = pf_cksum_fixup(*ic, oip, *ip, 0);
1204                 if (pc != NULL)
1205                         *ic = pf_cksum_fixup(*ic, opc, *pc, 0);
1206         }
1207         /* Change inner ip address, fix inner ip and icmp checksums. */
1208         PF_ACPY(ia, na, af);
1209         switch (af) {
1210 #ifdef INET
1211         case AF_INET: {
1212                 u_int32_t        oh2c = *h2c;
1213
1214                 *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c,
1215                     oia.addr16[0], ia->addr16[0], 0),
1216                     oia.addr16[1], ia->addr16[1], 0);
1217                 *ic = pf_cksum_fixup(pf_cksum_fixup(*ic,
1218                     oia.addr16[0], ia->addr16[0], 0),
1219                     oia.addr16[1], ia->addr16[1], 0);
1220                 *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0);
1221                 break;
1222         }
1223 #endif /* INET */
1224 #ifdef INET6
1225         case AF_INET6:
1226                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1227                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1228                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1229                     oia.addr16[0], ia->addr16[0], u),
1230                     oia.addr16[1], ia->addr16[1], u),
1231                     oia.addr16[2], ia->addr16[2], u),
1232                     oia.addr16[3], ia->addr16[3], u),
1233                     oia.addr16[4], ia->addr16[4], u),
1234                     oia.addr16[5], ia->addr16[5], u),
1235                     oia.addr16[6], ia->addr16[6], u),
1236                     oia.addr16[7], ia->addr16[7], u);
1237                 break;
1238 #endif /* INET6 */
1239         }
1240         /* Change outer ip address, fix outer ip or icmpv6 checksum. */
1241         PF_ACPY(oa, na, af);
1242         switch (af) {
1243 #ifdef INET
1244         case AF_INET:
1245                 *hc = pf_cksum_fixup(pf_cksum_fixup(*hc,
1246                     ooa.addr16[0], oa->addr16[0], 0),
1247                     ooa.addr16[1], oa->addr16[1], 0);
1248                 break;
1249 #endif /* INET */
1250 #ifdef INET6
1251         case AF_INET6:
1252                 *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1253                     pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup(
1254                     pf_cksum_fixup(pf_cksum_fixup(*ic,
1255                     ooa.addr16[0], oa->addr16[0], u),
1256                     ooa.addr16[1], oa->addr16[1], u),
1257                     ooa.addr16[2], oa->addr16[2], u),
1258                     ooa.addr16[3], oa->addr16[3], u),
1259                     ooa.addr16[4], oa->addr16[4], u),
1260                     ooa.addr16[5], oa->addr16[5], u),
1261                     ooa.addr16[6], oa->addr16[6], u),
1262                     ooa.addr16[7], oa->addr16[7], u);
1263                 break;
1264 #endif /* INET6 */
1265         }
1266 }
1267
1268 void
1269 pf_send_tcp(const struct pf_rule *r, sa_family_t af,
1270     const struct pf_addr *saddr, const struct pf_addr *daddr,
1271     u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack,
1272     u_int8_t flags, u_int16_t win, u_int16_t mss, u_int8_t ttl)
1273 {
1274         struct mbuf     *m;
1275         int              len = 0, tlen;
1276 #ifdef INET
1277         struct ip       *h = NULL;
1278 #endif /* INET */
1279 #ifdef INET6
1280         struct ip6_hdr  *h6 = NULL;
1281 #endif /* INET6 */
1282         struct tcphdr   *th = NULL;
1283         char *opt;
1284
1285         /* maximum segment size tcp option */
1286         tlen = sizeof(struct tcphdr);
1287         if (mss)
1288                 tlen += 4;
1289
1290         switch (af) {
1291 #ifdef INET
1292         case AF_INET:
1293                 len = sizeof(struct ip) + tlen;
1294                 break;
1295 #endif /* INET */
1296 #ifdef INET6
1297         case AF_INET6:
1298                 len = sizeof(struct ip6_hdr) + tlen;
1299                 break;
1300 #endif /* INET6 */
1301         }
1302
1303         /* create outgoing mbuf */
1304         m = m_gethdr(MB_DONTWAIT, MT_HEADER);
1305         if (m == NULL)
1306                 return;
1307         m->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
1308 #ifdef ALTQ
1309         if (r != NULL && r->qid) {
1310                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
1311                 m->m_pkthdr.altq_qid = r->qid;
1312                 m->m_pkthdr.ecn_af = af;
1313                 m->m_pkthdr.header = mtod(m, struct ip *);
1314         }
1315 #endif
1316         m->m_data += max_linkhdr;
1317         m->m_pkthdr.len = m->m_len = len;
1318         m->m_pkthdr.rcvif = NULL;
1319         bzero(m->m_data, len);
1320         switch (af) {
1321 #ifdef INET
1322         case AF_INET:
1323                 h = mtod(m, struct ip *);
1324
1325                 /* IP header fields included in the TCP checksum */
1326                 h->ip_p = IPPROTO_TCP;
1327                 h->ip_len = tlen;
1328                 h->ip_src.s_addr = saddr->v4.s_addr;
1329                 h->ip_dst.s_addr = daddr->v4.s_addr;
1330
1331                 th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip));
1332                 break;
1333 #endif /* INET */
1334 #ifdef INET6
1335         case AF_INET6:
1336                 h6 = mtod(m, struct ip6_hdr *);
1337
1338                 /* IP header fields included in the TCP checksum */
1339                 h6->ip6_nxt = IPPROTO_TCP;
1340                 h6->ip6_plen = htons(tlen);
1341                 memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr));
1342                 memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr));
1343
1344                 th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr));
1345                 break;
1346 #endif /* INET6 */
1347         }
1348
1349         /* TCP header */
1350         th->th_sport = sport;
1351         th->th_dport = dport;
1352         th->th_seq = htonl(seq);
1353         th->th_ack = htonl(ack);
1354         th->th_off = tlen >> 2;
1355         th->th_flags = flags;
1356         th->th_win = htons(win);
1357
1358         if (mss) {
1359                 opt = (char *)(th + 1);
1360                 opt[0] = TCPOPT_MAXSEG;
1361                 opt[1] = 4;
1362                 mss = htons(mss);
1363                 bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2);
1364         }
1365
1366         switch (af) {
1367 #ifdef INET
1368         case AF_INET:
1369                 /* TCP checksum */
1370                 th->th_sum = in_cksum(m, len);
1371
1372                 /* Finish the IP header */
1373                 h->ip_v = 4;
1374                 h->ip_hl = sizeof(*h) >> 2;
1375                 h->ip_tos = IPTOS_LOWDELAY;
1376                 h->ip_len = len;
1377                 h->ip_off = path_mtu_discovery ? IP_DF : 0;
1378                 h->ip_ttl = ttl ? ttl : ip_defttl;
1379                 h->ip_sum = 0;
1380                 ip_output(m, (void *)NULL, (void *)NULL, 0, (void *)NULL,
1381                     (void *)NULL);
1382                 break;
1383 #endif /* INET */
1384 #ifdef INET6
1385         case AF_INET6:
1386                 /* TCP checksum */
1387                 th->th_sum = in6_cksum(m, IPPROTO_TCP,
1388                     sizeof(struct ip6_hdr), tlen);
1389
1390                 h6->ip6_vfc |= IPV6_VERSION;
1391                 h6->ip6_hlim = IPV6_DEFHLIM;
1392
1393                 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
1394                 break;
1395 #endif /* INET6 */
1396         }
1397 }
1398
1399 void
1400 pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af,
1401     struct pf_rule *r)
1402 {
1403         struct mbuf     *m0;
1404
1405         m0 = m_copypacket(m, MB_DONTWAIT);
1406         if (m0 == NULL)
1407                 return;
1408         m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
1409
1410 #ifdef ALTQ
1411         if (r->qid) {
1412                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
1413                 m->m_pkthdr.altq_qid = r->qid;
1414                 m->m_pkthdr.ecn_af = af;
1415                 m->m_pkthdr.header = mtod(m0, struct ip *);
1416         }
1417 #endif
1418
1419         switch (af) {
1420 #ifdef INET
1421         case AF_INET:
1422                 icmp_error(m0, type, code, 0, 0);
1423                 break;
1424 #endif /* INET */
1425 #ifdef INET6
1426         case AF_INET6:
1427                 icmp6_error(m0, type, code, 0);
1428                 break;
1429 #endif /* INET6 */
1430         }
1431 }
1432
1433 /*
1434  * Return 1 if the addresses a and b match (with mask m), otherwise return 0.
1435  * If n is 0, they match if they are equal. If n is != 0, they match if they
1436  * are different.
1437  */
1438 int
1439 pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m,
1440     struct pf_addr *b, sa_family_t af)
1441 {
1442         int     match = 0;
1443
1444         switch (af) {
1445 #ifdef INET
1446         case AF_INET:
1447                 if ((a->addr32[0] & m->addr32[0]) ==
1448                     (b->addr32[0] & m->addr32[0]))
1449                         match++;
1450                 break;
1451 #endif /* INET */
1452 #ifdef INET6
1453         case AF_INET6:
1454                 if (((a->addr32[0] & m->addr32[0]) ==
1455                      (b->addr32[0] & m->addr32[0])) &&
1456                     ((a->addr32[1] & m->addr32[1]) ==
1457                      (b->addr32[1] & m->addr32[1])) &&
1458                     ((a->addr32[2] & m->addr32[2]) ==
1459                      (b->addr32[2] & m->addr32[2])) &&
1460                     ((a->addr32[3] & m->addr32[3]) ==
1461                      (b->addr32[3] & m->addr32[3])))
1462                         match++;
1463                 break;
1464 #endif /* INET6 */
1465         }
1466         if (match) {
1467                 if (n)
1468                         return (0);
1469                 else
1470                         return (1);
1471         } else {
1472                 if (n)
1473                         return (1);
1474                 else
1475                         return (0);
1476         }
1477 }
1478
1479 int
1480 pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p)
1481 {
1482         switch (op) {
1483         case PF_OP_IRG:
1484                 return ((p > a1) && (p < a2));
1485         case PF_OP_XRG:
1486                 return ((p < a1) || (p > a2));
1487         case PF_OP_RRG:
1488                 return ((p >= a1) && (p <= a2));
1489         case PF_OP_EQ:
1490                 return (p == a1);
1491         case PF_OP_NE:
1492                 return (p != a1);
1493         case PF_OP_LT:
1494                 return (p < a1);
1495         case PF_OP_LE:
1496                 return (p <= a1);
1497         case PF_OP_GT:
1498                 return (p > a1);
1499         case PF_OP_GE:
1500                 return (p >= a1);
1501         }
1502         return (0); /* never reached */
1503 }
1504
1505 int
1506 pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p)
1507 {
1508         a1 = ntohs(a1);
1509         a2 = ntohs(a2);
1510         p = ntohs(p);
1511         return (pf_match(op, a1, a2, p));
1512 }
1513
1514 int
1515 pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u)
1516 {
1517         if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1518                 return (0);
1519         return (pf_match(op, a1, a2, u));
1520 }
1521
1522 int
1523 pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g)
1524 {
1525         if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE)
1526                 return (0);
1527         return (pf_match(op, a1, a2, g));
1528 }
1529
1530 static int
1531 pf_match_tag(struct mbuf *m, struct pf_rule *r, struct pf_rule *nat_rule,
1532              int *tag)
1533 {
1534         if (*tag == -1) {       /* find mbuf tag */
1535                 if (nat_rule != NULL && nat_rule->tag)
1536                         *tag = nat_rule->tag;
1537                 else if (m->m_pkthdr.fw_flags & PF_MBUF_TAGGED)
1538                         *tag = m->m_pkthdr.pf_tag;
1539                 else
1540                         *tag = 0;
1541         }
1542
1543         return ((!r->match_tag_not && r->match_tag == *tag) ||
1544             (r->match_tag_not && r->match_tag != *tag));
1545 }
1546
1547 void
1548 pf_tag_packet(struct mbuf *m, int tag)
1549 {
1550         if (tag <= 0)
1551                 return;
1552
1553         m->m_pkthdr.fw_flags |= PF_MBUF_TAGGED;
1554         m->m_pkthdr.pf_tag = tag;
1555 }
1556
1557 #define PF_STEP_INTO_ANCHOR(r, a, s, n)                                 \
1558         do {                                                            \
1559                 if ((r) == NULL || (r)->anchor == NULL ||               \
1560                     (s) != NULL || (a) != NULL)                         \
1561                         panic("PF_STEP_INTO_ANCHOR");                   \
1562                 (a) = (r);                                              \
1563                 (s) = TAILQ_FIRST(&(r)->anchor->rulesets);              \
1564                 (r) = NULL;                                             \
1565                 while ((s) != NULL && ((r) =                            \
1566                     TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL)     \
1567                         (s) = TAILQ_NEXT((s), entries);                 \
1568                 if ((r) == NULL) {                                      \
1569                         (r) = TAILQ_NEXT((a), entries);                 \
1570                         (a) = NULL;                                     \
1571                 }                                                       \
1572         } while (0)
1573
1574 #define PF_STEP_OUT_OF_ANCHOR(r, a, s, n)                               \
1575         do {                                                            \
1576                 if ((r) != NULL || (a) == NULL || (s) == NULL)          \
1577                         panic("PF_STEP_OUT_OF_ANCHOR");                 \
1578                 (s) = TAILQ_NEXT((s), entries);                         \
1579                 while ((s) != NULL && ((r) =                            \
1580                     TAILQ_FIRST((s)->rules[n].active.ptr)) == NULL)     \
1581                         (s) = TAILQ_NEXT((s), entries);                 \
1582                 if ((r) == NULL) {                                      \
1583                         (r) = TAILQ_NEXT((a), entries);                 \
1584                         (a) = NULL;                                     \
1585                 }                                                       \
1586         } while (0)
1587
1588 #ifdef INET6
1589 void
1590 pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr,
1591     struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af)
1592 {
1593         switch (af) {
1594 #ifdef INET
1595         case AF_INET:
1596                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1597                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1598                 break;
1599 #endif /* INET */
1600         case AF_INET6:
1601                 naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) |
1602                 ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]);
1603                 naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) |
1604                 ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]);
1605                 naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) |
1606                 ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]);
1607                 naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) |
1608                 ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]);
1609                 break;
1610         }
1611 }
1612
1613 void
1614 pf_addr_inc(struct pf_addr *addr, sa_family_t af)
1615 {
1616         switch (af) {
1617 #ifdef INET
1618         case AF_INET:
1619                 addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1);
1620                 break;
1621 #endif /* INET */
1622         case AF_INET6:
1623                 if (addr->addr32[3] == 0xffffffff) {
1624                         addr->addr32[3] = 0;
1625                         if (addr->addr32[2] == 0xffffffff) {
1626                                 addr->addr32[2] = 0;
1627                                 if (addr->addr32[1] == 0xffffffff) {
1628                                         addr->addr32[1] = 0;
1629                                         addr->addr32[0] =
1630                                             htonl(ntohl(addr->addr32[0]) + 1);
1631                                 } else
1632                                         addr->addr32[1] =
1633                                             htonl(ntohl(addr->addr32[1]) + 1);
1634                         } else
1635                                 addr->addr32[2] =
1636                                     htonl(ntohl(addr->addr32[2]) + 1);
1637                 } else
1638                         addr->addr32[3] =
1639                             htonl(ntohl(addr->addr32[3]) + 1);
1640                 break;
1641         }
1642 }
1643 #endif /* INET6 */
1644
1645 #define mix(a,b,c) \
1646         do {                                    \
1647                 a -= b; a -= c; a ^= (c >> 13); \
1648                 b -= c; b -= a; b ^= (a << 8);  \
1649                 c -= a; c -= b; c ^= (b >> 13); \
1650                 a -= b; a -= c; a ^= (c >> 12); \
1651                 b -= c; b -= a; b ^= (a << 16); \
1652                 c -= a; c -= b; c ^= (b >> 5);  \
1653                 a -= b; a -= c; a ^= (c >> 3);  \
1654                 b -= c; b -= a; b ^= (a << 10); \
1655                 c -= a; c -= b; c ^= (b >> 15); \
1656         } while (0)
1657
1658 /*
1659  * hash function based on bridge_hash in if_bridge.c
1660  */
1661 void
1662 pf_hash(struct pf_addr *inaddr, struct pf_addr *hash,
1663     struct pf_poolhashkey *key, sa_family_t af)
1664 {
1665         u_int32_t       a = 0x9e3779b9, b = 0x9e3779b9, c = key->key32[0];
1666
1667         switch (af) {
1668 #ifdef INET
1669         case AF_INET:
1670                 a += inaddr->addr32[0];
1671                 b += key->key32[1];
1672                 mix(a, b, c);
1673                 hash->addr32[0] = c + key->key32[2];
1674                 break;
1675 #endif /* INET */
1676 #ifdef INET6
1677         case AF_INET6:
1678                 a += inaddr->addr32[0];
1679                 b += inaddr->addr32[2];
1680                 mix(a, b, c);
1681                 hash->addr32[0] = c;
1682                 a += inaddr->addr32[1];
1683                 b += inaddr->addr32[3];
1684                 c += key->key32[1];
1685                 mix(a, b, c);
1686                 hash->addr32[1] = c;
1687                 a += inaddr->addr32[2];
1688                 b += inaddr->addr32[1];
1689                 c += key->key32[2];
1690                 mix(a, b, c);
1691                 hash->addr32[2] = c;
1692                 a += inaddr->addr32[3];
1693                 b += inaddr->addr32[0];
1694                 c += key->key32[3];
1695                 mix(a, b, c);
1696                 hash->addr32[3] = c;
1697                 break;
1698 #endif /* INET6 */
1699         }
1700 }
1701
1702 int
1703 pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
1704     struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
1705 {
1706         unsigned char            hash[16];
1707         struct pf_pool          *rpool = &r->rpool;
1708         struct pf_addr          *raddr = &rpool->cur->addr.v.a.addr;
1709         struct pf_addr          *rmask = &rpool->cur->addr.v.a.mask;
1710         struct pf_pooladdr      *acur = rpool->cur;
1711         struct pf_src_node       k;
1712
1713         if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
1714             (r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1715                 k.af = af;
1716                 PF_ACPY(&k.addr, saddr, af);
1717                 if (r->rule_flag & PFRULE_RULESRCTRACK ||
1718                     r->rpool.opts & PF_POOL_STICKYADDR)
1719                         k.rule.ptr = r;
1720                 else
1721                         k.rule.ptr = NULL;
1722                 pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
1723                 *sn = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
1724                 if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
1725                         PF_ACPY(naddr, &(*sn)->raddr, af);
1726                         if (pf_status.debug >= PF_DEBUG_MISC) {
1727                                 printf("pf_map_addr: src tracking maps ");
1728                                 pf_print_host(&k.addr, 0, af);
1729                                 printf(" to ");
1730                                 pf_print_host(naddr, 0, af);
1731                                 printf("\n");
1732                         }
1733                         return (0);
1734                 }
1735         }
1736
1737         if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
1738                 return (1);
1739         if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1740                 if (af == AF_INET) {
1741                         if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
1742                             (rpool->opts & PF_POOL_TYPEMASK) !=
1743                             PF_POOL_ROUNDROBIN)
1744                                 return (1);
1745                          raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
1746                          rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
1747                 } else {
1748                         if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
1749                             (rpool->opts & PF_POOL_TYPEMASK) !=
1750                             PF_POOL_ROUNDROBIN)
1751                                 return (1);
1752                         raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
1753                         rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
1754                 }
1755         } else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1756                 if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
1757                         return (1); /* unsupported */
1758         } else {
1759                 raddr = &rpool->cur->addr.v.a.addr;
1760                 rmask = &rpool->cur->addr.v.a.mask;
1761         }
1762
1763         switch (rpool->opts & PF_POOL_TYPEMASK) {
1764         case PF_POOL_NONE:
1765                 PF_ACPY(naddr, raddr, af);
1766                 break;
1767         case PF_POOL_BITMASK:
1768                 PF_POOLMASK(naddr, raddr, rmask, saddr, af);
1769                 break;
1770         case PF_POOL_RANDOM:
1771                 if (init_addr != NULL && PF_AZERO(init_addr, af)) {
1772                         switch (af) {
1773 #ifdef INET
1774                         case AF_INET:
1775                                 rpool->counter.addr32[0] = arc4random();
1776                                 break;
1777 #endif /* INET */
1778 #ifdef INET6
1779                         case AF_INET6:
1780                                 if (rmask->addr32[3] != 0xffffffff)
1781                                         rpool->counter.addr32[3] = arc4random();
1782                                 else
1783                                         break;
1784                                 if (rmask->addr32[2] != 0xffffffff)
1785                                         rpool->counter.addr32[2] = arc4random();
1786                                 else
1787                                         break;
1788                                 if (rmask->addr32[1] != 0xffffffff)
1789                                         rpool->counter.addr32[1] = arc4random();
1790                                 else
1791                                         break;
1792                                 if (rmask->addr32[0] != 0xffffffff)
1793                                         rpool->counter.addr32[0] = arc4random();
1794                                 break;
1795 #endif /* INET6 */
1796                         }
1797                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1798                         PF_ACPY(init_addr, naddr, af);
1799
1800                 } else {
1801                         PF_AINC(&rpool->counter, af);
1802                         PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
1803                 }
1804                 break;
1805         case PF_POOL_SRCHASH:
1806                 pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
1807                 PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
1808                 break;
1809         case PF_POOL_ROUNDROBIN:
1810                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1811                         if (!pfr_pool_get(rpool->cur->addr.p.tbl,
1812                             &rpool->tblidx, &rpool->counter,
1813                             &raddr, &rmask, af))
1814                                 goto get_addr;
1815                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1816                         if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
1817                             &rpool->tblidx, &rpool->counter,
1818                             &raddr, &rmask, af))
1819                                 goto get_addr;
1820                 } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
1821                         goto get_addr;
1822
1823         try_next:
1824                 if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
1825                         rpool->cur = TAILQ_FIRST(&rpool->list);
1826                 if (rpool->cur->addr.type == PF_ADDR_TABLE) {
1827                         rpool->tblidx = -1;
1828                         if (pfr_pool_get(rpool->cur->addr.p.tbl,
1829                             &rpool->tblidx, &rpool->counter,
1830                             &raddr, &rmask, af)) {
1831                                 /* table contains no address of type 'af' */
1832                                 if (rpool->cur != acur)
1833                                         goto try_next;
1834                                 return (1);
1835                         }
1836                 } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
1837                         rpool->tblidx = -1;
1838                         if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
1839                             &rpool->tblidx, &rpool->counter,
1840                             &raddr, &rmask, af)) {
1841                                 /* table contains no address of type 'af' */
1842                                 if (rpool->cur != acur)
1843                                         goto try_next;
1844                                 return (1);
1845                         }
1846                 } else {
1847                         raddr = &rpool->cur->addr.v.a.addr;
1848                         rmask = &rpool->cur->addr.v.a.mask;
1849                         PF_ACPY(&rpool->counter, raddr, af);
1850                 }
1851
1852         get_addr:
1853                 PF_ACPY(naddr, &rpool->counter, af);
1854                 PF_AINC(&rpool->counter, af);
1855                 break;
1856         }
1857         if (*sn != NULL)
1858                 PF_ACPY(&(*sn)->raddr, naddr, af);
1859
1860         if (pf_status.debug >= PF_DEBUG_MISC &&
1861             (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
1862                 printf("pf_map_addr: selected address ");
1863                 pf_print_host(naddr, 0, af);
1864                 printf("\n");
1865         }
1866
1867         return (0);
1868 }
1869
1870 int
1871 pf_get_sport(sa_family_t af, u_int8_t proto, struct pf_rule *r,
1872     struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t dport,
1873     struct pf_addr *naddr, u_int16_t *nport, u_int16_t low, u_int16_t high,
1874     struct pf_src_node **sn)
1875 {
1876         struct pf_state         key;
1877         struct pf_addr          init_addr;
1878         u_int16_t               cut;
1879
1880         bzero(&init_addr, sizeof(init_addr));
1881         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
1882                 return (1);
1883
1884         do {
1885                 key.af = af;
1886                 key.proto = proto;
1887                 PF_ACPY(&key.ext.addr, daddr, key.af);
1888                 PF_ACPY(&key.gwy.addr, naddr, key.af);
1889                 key.ext.port = dport;
1890
1891                 /*
1892                  * port search; start random, step;
1893                  * similar 2 portloop in in_pcbbind
1894                  */
1895                 if (!(proto == IPPROTO_TCP || proto == IPPROTO_UDP)) {
1896                         key.gwy.port = 0;
1897                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
1898                                 return (0);
1899                 } else if (low == 0 && high == 0) {
1900                         key.gwy.port = *nport;
1901                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL)
1902                                 return (0);
1903                 } else if (low == high) {
1904                         key.gwy.port = htons(low);
1905                         if (pf_find_state_all(&key, PF_EXT_GWY, NULL) == NULL) {
1906                                 *nport = htons(low);
1907                                 return (0);
1908                         }
1909                 } else {
1910                         u_int16_t tmp;
1911
1912                         if (low > high) {
1913                                 tmp = low;
1914                                 low = high;
1915                                 high = tmp;
1916                         }
1917                         /* low < high */
1918                         cut = arc4random() % (1 + high - low) + low;
1919                         /* low <= cut <= high */
1920                         for (tmp = cut; tmp <= high; ++(tmp)) {
1921                                 key.gwy.port = htons(tmp);
1922                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
1923                                     NULL) {
1924                                         *nport = htons(tmp);
1925                                         return (0);
1926                                 }
1927                         }
1928                         for (tmp = cut - 1; tmp >= low; --(tmp)) {
1929                                 key.gwy.port = htons(tmp);
1930                                 if (pf_find_state_all(&key, PF_EXT_GWY, NULL) ==
1931                                     NULL) {
1932                                         *nport = htons(tmp);
1933                                         return (0);
1934                                 }
1935                         }
1936                 }
1937
1938                 switch (r->rpool.opts & PF_POOL_TYPEMASK) {
1939                 case PF_POOL_RANDOM:
1940                 case PF_POOL_ROUNDROBIN:
1941                         if (pf_map_addr(af, r, saddr, naddr, &init_addr, sn))
1942                                 return (1);
1943                         break;
1944                 case PF_POOL_NONE:
1945                 case PF_POOL_SRCHASH:
1946                 case PF_POOL_BITMASK:
1947                 default:
1948                         return (1);
1949                 }
1950         } while (! PF_AEQ(&init_addr, naddr, af) );
1951
1952         return (1);                                     /* none available */
1953 }
1954
1955 struct pf_rule *
1956 pf_match_translation(struct pf_pdesc *pd, struct mbuf *m, int off,
1957     int direction, struct pfi_kif *kif, struct pf_addr *saddr, u_int16_t sport,
1958     struct pf_addr *daddr, u_int16_t dport, int rs_num)
1959 {
1960         struct pf_rule          *r, *rm = NULL, *anchorrule = NULL;
1961         struct pf_ruleset       *ruleset = NULL;
1962
1963         r = TAILQ_FIRST(pf_main_ruleset.rules[rs_num].active.ptr);
1964         while (r && rm == NULL) {
1965                 struct pf_rule_addr     *src = NULL, *dst = NULL;
1966                 struct pf_addr_wrap     *xdst = NULL;
1967
1968                 if (r->action == PF_BINAT && direction == PF_IN) {
1969                         src = &r->dst;
1970                         if (r->rpool.cur != NULL)
1971                                 xdst = &r->rpool.cur->addr;
1972                 } else {
1973                         src = &r->src;
1974                         dst = &r->dst;
1975                 }
1976
1977                 r->evaluations++;
1978                 if (r->kif != NULL &&
1979                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
1980                         r = r->skip[PF_SKIP_IFP].ptr;
1981                 else if (r->direction && r->direction != direction)
1982                         r = r->skip[PF_SKIP_DIR].ptr;
1983                 else if (r->af && r->af != pd->af)
1984                         r = r->skip[PF_SKIP_AF].ptr;
1985                 else if (r->proto && r->proto != pd->proto)
1986                         r = r->skip[PF_SKIP_PROTO].ptr;
1987                 else if (PF_MISMATCHAW(&src->addr, saddr, pd->af, src->not))
1988                         r = r->skip[src == &r->src ? PF_SKIP_SRC_ADDR :
1989                             PF_SKIP_DST_ADDR].ptr;
1990                 else if (src->port_op && !pf_match_port(src->port_op,
1991                     src->port[0], src->port[1], sport))
1992                         r = r->skip[src == &r->src ? PF_SKIP_SRC_PORT :
1993                             PF_SKIP_DST_PORT].ptr;
1994                 else if (dst != NULL &&
1995                     PF_MISMATCHAW(&dst->addr, daddr, pd->af, dst->not))
1996                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
1997                 else if (xdst != NULL && PF_MISMATCHAW(xdst, daddr, pd->af, 0))
1998                         r = TAILQ_NEXT(r, entries);
1999                 else if (dst != NULL && dst->port_op &&
2000                     !pf_match_port(dst->port_op, dst->port[0],
2001                     dst->port[1], dport))
2002                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2003                 else if (r->os_fingerprint != PF_OSFP_ANY && (pd->proto !=
2004                     IPPROTO_TCP || !pf_osfp_match(pf_osfp_fingerprint(pd, m,
2005                     off, pd->hdr.tcp), r->os_fingerprint)))
2006                         r = TAILQ_NEXT(r, entries);
2007                 else if (r->anchorname[0] && r->anchor == NULL)
2008                         r = TAILQ_NEXT(r, entries);
2009                 else if (r->anchor == NULL)
2010                                 rm = r;
2011                 else
2012                         PF_STEP_INTO_ANCHOR(r, anchorrule, ruleset, rs_num);
2013                 if (r == NULL && anchorrule != NULL)
2014                         PF_STEP_OUT_OF_ANCHOR(r, anchorrule, ruleset,
2015                             rs_num);
2016         }
2017         if (rm != NULL && (rm->action == PF_NONAT ||
2018             rm->action == PF_NORDR || rm->action == PF_NOBINAT))
2019                 return (NULL);
2020         return (rm);
2021 }
2022
2023 struct pf_rule *
2024 pf_get_translation(struct pf_pdesc *pd, struct mbuf *m, int off, int direction,
2025     struct pfi_kif *kif, struct pf_src_node **sn,
2026     struct pf_addr *saddr, u_int16_t sport,
2027     struct pf_addr *daddr, u_int16_t dport,
2028     struct pf_addr *naddr, u_int16_t *nport)
2029 {
2030         struct pf_rule  *r = NULL;
2031
2032         if (direction == PF_OUT) {
2033                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2034                     sport, daddr, dport, PF_RULESET_BINAT);
2035                 if (r == NULL)
2036                         r = pf_match_translation(pd, m, off, direction, kif,
2037                             saddr, sport, daddr, dport, PF_RULESET_NAT);
2038         } else {
2039                 r = pf_match_translation(pd, m, off, direction, kif, saddr,
2040                     sport, daddr, dport, PF_RULESET_RDR);
2041                 if (r == NULL)
2042                         r = pf_match_translation(pd, m, off, direction, kif,
2043                             saddr, sport, daddr, dport, PF_RULESET_BINAT);
2044         }
2045
2046         if (r != NULL) {
2047                 switch (r->action) {
2048                 case PF_NONAT:
2049                 case PF_NOBINAT:
2050                 case PF_NORDR:
2051                         return (NULL);
2052                 case PF_NAT:
2053                         if (pf_get_sport(pd->af, pd->proto, r, saddr,
2054                             daddr, dport, naddr, nport, r->rpool.proxy_port[0],
2055                             r->rpool.proxy_port[1], sn)) {
2056                                 DPFPRINTF(PF_DEBUG_MISC,
2057                                     ("pf: NAT proxy port allocation "
2058                                     "(%u-%u) failed\n",
2059                                     r->rpool.proxy_port[0],
2060                                     r->rpool.proxy_port[1]));
2061                                 return (NULL);
2062                         }
2063                         break;
2064                 case PF_BINAT:
2065                         switch (direction) {
2066                         case PF_OUT:
2067                                 if (r->rpool.cur->addr.type == PF_ADDR_DYNIFTL){
2068                                         if (pd->af == AF_INET) {
2069                                                 if (r->rpool.cur->addr.p.dyn->
2070                                                     pfid_acnt4 < 1)
2071                                                         return (NULL);
2072                                                 PF_POOLMASK(naddr,
2073                                                     &r->rpool.cur->addr.p.dyn->
2074                                                     pfid_addr4,
2075                                                     &r->rpool.cur->addr.p.dyn->
2076                                                     pfid_mask4,
2077                                                     saddr, AF_INET);
2078                                         } else {
2079                                                 if (r->rpool.cur->addr.p.dyn->
2080                                                     pfid_acnt6 < 1)
2081                                                         return (NULL);
2082                                                 PF_POOLMASK(naddr,
2083                                                     &r->rpool.cur->addr.p.dyn->
2084                                                     pfid_addr6,
2085                                                     &r->rpool.cur->addr.p.dyn->
2086                                                     pfid_mask6,
2087                                                     saddr, AF_INET6);
2088                                         }
2089                                 } else
2090                                         PF_POOLMASK(naddr,
2091                                             &r->rpool.cur->addr.v.a.addr,
2092                                             &r->rpool.cur->addr.v.a.mask,
2093                                             saddr, pd->af);
2094                                 break;
2095                         case PF_IN:
2096                                 if (r->src.addr.type == PF_ADDR_DYNIFTL){
2097                                         if (pd->af == AF_INET) {
2098                                                 if (r->src.addr.p.dyn->
2099                                                     pfid_acnt4 < 1)
2100                                                         return (NULL);
2101                                                 PF_POOLMASK(naddr,
2102                                                     &r->src.addr.p.dyn->
2103                                                     pfid_addr4,
2104                                                     &r->src.addr.p.dyn->
2105                                                     pfid_mask4,
2106                                                     daddr, AF_INET);
2107                                         } else {
2108                                                 if (r->src.addr.p.dyn->
2109                                                     pfid_acnt6 < 1)
2110                                                         return (NULL);
2111                                                 PF_POOLMASK(naddr,
2112                                                     &r->src.addr.p.dyn->
2113                                                     pfid_addr6,
2114                                                     &r->src.addr.p.dyn->
2115                                                     pfid_mask6,
2116                                                     daddr, AF_INET6);
2117                                         }
2118                                 } else
2119                                         PF_POOLMASK(naddr,
2120                                             &r->src.addr.v.a.addr,
2121                                             &r->src.addr.v.a.mask, daddr,
2122                                             pd->af);
2123                                 break;
2124                         }
2125                         break;
2126                 case PF_RDR: {
2127                         if (pf_map_addr(r->af, r, saddr, naddr, NULL, sn))
2128                                 return (NULL);
2129
2130                         if (r->rpool.proxy_port[1]) {
2131                                 u_int32_t       tmp_nport;
2132
2133                                 tmp_nport = ((ntohs(dport) -
2134                                     ntohs(r->dst.port[0])) %
2135                                     (r->rpool.proxy_port[1] -
2136                                     r->rpool.proxy_port[0] + 1)) +
2137                                     r->rpool.proxy_port[0];
2138
2139                                 /* wrap around if necessary */
2140                                 if (tmp_nport > 65535)
2141                                         tmp_nport -= 65535;
2142                                 *nport = htons((u_int16_t)tmp_nport);
2143                         } else if (r->rpool.proxy_port[0])
2144                                 *nport = htons(r->rpool.proxy_port[0]);
2145                         break;
2146                 }
2147                 default:
2148                         return (NULL);
2149                 }
2150         }
2151
2152         return (r);
2153 }
2154
2155 #ifdef SMP
2156 struct netmsg_hashlookup {
2157         struct lwkt_msg         nm_lmsg;
2158         struct inpcb            **nm_pinp;
2159         struct inpcbinfo        *nm_pcbinfo;
2160         struct pf_addr          *nm_saddr;
2161         struct pf_addr          *nm_daddr;
2162         uint16_t                nm_sport;
2163         uint16_t                nm_dport;
2164         sa_family_t             nm_af;
2165 };
2166
2167 static int
2168 in_pcblookup_hash_handler(struct lwkt_msg *msg0)
2169 {
2170         struct netmsg_hashlookup *msg = (struct netmsg_hashlookup *)msg0;
2171
2172         if (msg->nm_af == AF_INET)
2173                 *msg->nm_pinp = in_pcblookup_hash(msg->nm_pcbinfo,
2174                     msg->nm_saddr->v4, msg->nm_sport, msg->nm_daddr->v4,
2175                     msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2176 #ifdef INET6
2177         else
2178                 *msg->nm_pinp = in6_pcblookup_hash(msg->nm_pcbinfo,
2179                     &msg->nm_saddr->v6, msg->nm_sport, &msg->nm_daddr->v6,
2180                     msg->nm_dport, INPLOOKUP_WILDCARD, NULL);
2181 #endif /* INET6 */
2182         lwkt_replymsg(&msg->nm_lmsg, 0);
2183         return (EASYNC);
2184 }
2185 #endif /* SMP */
2186
2187 int
2188 pf_socket_lookup(uid_t *uid, gid_t *gid, int direction, struct pf_pdesc *pd)
2189 {
2190         struct pf_addr          *saddr, *daddr;
2191         u_int16_t                sport, dport;
2192         struct inpcbinfo        *pi;
2193         struct inpcb            *inp;
2194 #ifdef SMP
2195         struct netmsg_hashlookup *msg = NULL;
2196 #endif
2197         int                      pi_cpu = 0;
2198
2199         *uid = UID_MAX;
2200         *gid = GID_MAX;
2201         if (direction == PF_IN) {
2202                 saddr = pd->src;
2203                 daddr = pd->dst;
2204         } else {
2205                 saddr = pd->dst;
2206                 daddr = pd->src;
2207         }
2208         switch (pd->proto) {
2209         case IPPROTO_TCP:
2210                 sport = pd->hdr.tcp->th_sport;
2211                 dport = pd->hdr.tcp->th_dport;
2212
2213                 pi_cpu = tcp_addrcpu(saddr->v4.s_addr, sport, daddr->v4.s_addr, dport);
2214                 pi = &tcbinfo[pi_cpu];
2215 #ifdef SMP
2216                 /*
2217                  * Our netstack runs lockless on MP systems
2218                  * (only for TCP connections at the moment).
2219                  *
2220                  * As we are not allowed to read another CPU's tcbinfo,
2221                  * we have to ask that CPU via remote call to search the
2222                  * table for us.
2223                  *
2224                  * Prepare a msg iff data belongs to another CPU.
2225                  */
2226                 if (pi_cpu != mycpu->gd_cpuid) {
2227                         msg = malloc(sizeof(*msg), M_LWKTMSG, M_INTWAIT);
2228                         lwkt_initmsg(&msg->nm_lmsg, &netisr_afree_rport, 0,
2229                             lwkt_cmd_func(in_pcblookup_hash_handler),
2230                             lwkt_cmd_op_none);
2231                         msg->nm_pinp = &inp;
2232                         msg->nm_pcbinfo = pi;
2233                         msg->nm_saddr = saddr;
2234                         msg->nm_sport = sport;
2235                         msg->nm_daddr = daddr;
2236                         msg->nm_dport = dport;
2237                         msg->nm_af = pd->af;
2238                 }
2239 #endif /* SMP */
2240                 break;
2241         case IPPROTO_UDP:
2242                 sport = pd->hdr.udp->uh_sport;
2243                 dport = pd->hdr.udp->uh_dport;
2244                 pi = &udbinfo;
2245                 break;
2246         default:
2247                 return (0);
2248         }
2249         if (direction != PF_IN) {
2250                 u_int16_t       p;
2251
2252                 p = sport;
2253                 sport = dport;
2254                 dport = p;
2255         }
2256         switch (pd->af) {
2257 #ifdef INET6
2258         case AF_INET6:
2259 #ifdef SMP
2260                 /*
2261                  * Query other CPU, second part
2262                  *
2263                  * msg only gets initialized when:
2264                  * 1) packet is TCP
2265                  * 2) the info belongs to another CPU
2266                  *
2267                  * Use some switch/case magic to avoid code duplication.
2268                  */
2269                 if (msg == NULL)
2270 #endif /* SMP */
2271                 {
2272                         inp = in6_pcblookup_hash(pi, &saddr->v6, sport,
2273                             &daddr->v6, dport, INPLOOKUP_WILDCARD, NULL);
2274
2275                         if (inp == NULL)
2276                                 return (0);
2277                         break;
2278                 }
2279                 /* FALLTHROUGH if SMP and on other CPU */
2280 #endif /* INET6 */
2281         case AF_INET:
2282 #ifdef SMP
2283                 if (msg != NULL) {
2284                         lwkt_sendmsg(tcp_cport(pi_cpu), &msg->nm_lmsg);
2285                 } else
2286 #endif /* SMP */
2287                 {
2288                         inp = in_pcblookup_hash(pi, saddr->v4, sport, daddr->v4,
2289                             dport, INPLOOKUP_WILDCARD, NULL);
2290                 }
2291                 if (inp == NULL)
2292                         return (0);
2293                 break;
2294
2295         default:
2296                 return (0);
2297         }
2298         *uid = inp->inp_socket->so_cred->cr_uid;
2299         *gid = inp->inp_socket->so_cred->cr_groups[0];
2300         return (1);
2301 }
2302
2303 u_int8_t
2304 pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2305 {
2306         int              hlen;
2307         u_int8_t         hdr[60];
2308         u_int8_t        *opt, optlen;
2309         u_int8_t         wscale = 0;
2310
2311         hlen = th_off << 2;             /* hlen <= sizeof(hdr) */
2312         if (hlen <= sizeof(struct tcphdr))
2313                 return (0);
2314         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2315                 return (0);
2316         opt = hdr + sizeof(struct tcphdr);
2317         hlen -= sizeof(struct tcphdr);
2318         while (hlen >= 3) {
2319                 switch (*opt) {
2320                 case TCPOPT_EOL:
2321                 case TCPOPT_NOP:
2322                         ++opt;
2323                         --hlen;
2324                         break;
2325                 case TCPOPT_WINDOW:
2326                         wscale = opt[2];
2327                         if (wscale > TCP_MAX_WINSHIFT)
2328                                 wscale = TCP_MAX_WINSHIFT;
2329                         wscale |= PF_WSCALE_FLAG;
2330                         /* FALLTHROUGH */
2331                 default:
2332                         optlen = opt[1];
2333                         if (optlen < 2)
2334                                 optlen = 2;
2335                         hlen -= optlen;
2336                         opt += optlen;
2337                         break;
2338                 }
2339         }
2340         return (wscale);
2341 }
2342
2343 u_int16_t
2344 pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af)
2345 {
2346         int              hlen;
2347         u_int8_t         hdr[60];
2348         u_int8_t        *opt, optlen;
2349         u_int16_t        mss = tcp_mssdflt;
2350
2351         hlen = th_off << 2;     /* hlen <= sizeof(hdr) */
2352         if (hlen <= sizeof(struct tcphdr))
2353                 return (0);
2354         if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af))
2355                 return (0);
2356         opt = hdr + sizeof(struct tcphdr);
2357         hlen -= sizeof(struct tcphdr);
2358         while (hlen >= TCPOLEN_MAXSEG) {
2359                 switch (*opt) {
2360                 case TCPOPT_EOL:
2361                 case TCPOPT_NOP:
2362                         ++opt;
2363                         --hlen;
2364                         break;
2365                 case TCPOPT_MAXSEG:
2366                         bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2);
2367                         /* FALLTHROUGH */
2368                 default:
2369                         optlen = opt[1];
2370                         if (optlen < 2)
2371                                 optlen = 2;
2372                         hlen -= optlen;
2373                         opt += optlen;
2374                         break;
2375                 }
2376         }
2377         return (mss);
2378 }
2379
2380 u_int16_t
2381 pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer)
2382 {
2383 #ifdef INET
2384         struct sockaddr_in      *dst;
2385         struct route             ro;
2386 #endif /* INET */
2387 #ifdef INET6
2388         struct sockaddr_in6     *dst6;
2389         struct route_in6         ro6;
2390 #endif /* INET6 */
2391         struct rtentry          *rt = NULL;
2392         int                      hlen = 0;
2393         u_int16_t                mss = tcp_mssdflt;
2394
2395         switch (af) {
2396 #ifdef INET
2397         case AF_INET:
2398                 hlen = sizeof(struct ip);
2399                 bzero(&ro, sizeof(ro));
2400                 dst = (struct sockaddr_in *)&ro.ro_dst;
2401                 dst->sin_family = AF_INET;
2402                 dst->sin_len = sizeof(*dst);
2403                 dst->sin_addr = addr->v4;
2404                 rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
2405                 rt = ro.ro_rt;
2406                 break;
2407 #endif /* INET */
2408 #ifdef INET6
2409         case AF_INET6:
2410                 hlen = sizeof(struct ip6_hdr);
2411                 bzero(&ro6, sizeof(ro6));
2412                 dst6 = (struct sockaddr_in6 *)&ro6.ro_dst;
2413                 dst6->sin6_family = AF_INET6;
2414                 dst6->sin6_len = sizeof(*dst6);
2415                 dst6->sin6_addr = addr->v6;
2416                 rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING));
2417                 rt = ro6.ro_rt;
2418                 break;
2419 #endif /* INET6 */
2420         }
2421
2422         if (rt && rt->rt_ifp) {
2423                 mss = rt->rt_ifp->if_mtu - hlen - sizeof(struct tcphdr);
2424                 mss = max(tcp_mssdflt, mss);
2425                 RTFREE(rt);
2426         }
2427         mss = min(mss, offer);
2428         mss = max(mss, 64);             /* sanity - at least max opt space */
2429         return (mss);
2430 }
2431
2432 void
2433 pf_set_rt_ifp(struct pf_state *s, struct pf_addr *saddr)
2434 {
2435         struct pf_rule *r = s->rule.ptr;
2436
2437         s->rt_kif = NULL;
2438         if (!r->rt || r->rt == PF_FASTROUTE)
2439                 return;
2440         switch (s->af) {
2441 #ifdef INET
2442         case AF_INET:
2443                 pf_map_addr(AF_INET, r, saddr, &s->rt_addr, NULL,
2444                     &s->nat_src_node);
2445                 s->rt_kif = r->rpool.cur->kif;
2446                 break;
2447 #endif /* INET */
2448 #ifdef INET6
2449         case AF_INET6:
2450                 pf_map_addr(AF_INET6, r, saddr, &s->rt_addr, NULL,
2451                     &s->nat_src_node);
2452                 s->rt_kif = r->rpool.cur->kif;
2453                 break;
2454 #endif /* INET6 */
2455         }
2456 }
2457
2458 int
2459 pf_test_tcp(struct pf_rule **rm, struct pf_state **sm, int direction,
2460     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2461     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2462 {
2463         struct pf_rule          *nr = NULL;
2464         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
2465         struct tcphdr           *th = pd->hdr.tcp;
2466         u_int16_t                bport, nport = 0;
2467         sa_family_t              af = pd->af;
2468         int                      lookup = -1;
2469         uid_t                    uid;
2470         gid_t                    gid;
2471         struct pf_rule          *r, *a = NULL;
2472         struct pf_ruleset       *ruleset = NULL;
2473         struct pf_src_node      *nsn = NULL;
2474         u_short                  reason;
2475         int                      rewrite = 0;
2476         int                      tag = -1;
2477         u_int16_t                mss = tcp_mssdflt;
2478
2479         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2480
2481         if (direction == PF_OUT) {
2482                 bport = nport = th->th_sport;
2483                 /* check outgoing packet for BINAT/NAT */
2484                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2485                     saddr, th->th_sport, daddr, th->th_dport,
2486                     &pd->naddr, &nport)) != NULL) {
2487                         PF_ACPY(&pd->baddr, saddr, af);
2488                         pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2489                             &th->th_sum, &pd->naddr, nport, 0, af);
2490                         rewrite++;
2491                         if (nr->natpass)
2492                                 r = NULL;
2493                         pd->nat_rule = nr;
2494                 }
2495         } else {
2496                 bport = nport = th->th_dport;
2497                 /* check incoming packet for BINAT/RDR */
2498                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2499                     saddr, th->th_sport, daddr, th->th_dport,
2500                     &pd->naddr, &nport)) != NULL) {
2501                         PF_ACPY(&pd->baddr, daddr, af);
2502                         pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2503                             &th->th_sum, &pd->naddr, nport, 0, af);
2504                         rewrite++;
2505                         if (nr->natpass)
2506                                 r = NULL;
2507                         pd->nat_rule = nr;
2508                 }
2509         }
2510
2511         while (r != NULL) {
2512                 r->evaluations++;
2513                 if (r->kif != NULL &&
2514                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2515                         r = r->skip[PF_SKIP_IFP].ptr;
2516                 else if (r->direction && r->direction != direction)
2517                         r = r->skip[PF_SKIP_DIR].ptr;
2518                 else if (r->af && r->af != af)
2519                         r = r->skip[PF_SKIP_AF].ptr;
2520                 else if (r->proto && r->proto != IPPROTO_TCP)
2521                         r = r->skip[PF_SKIP_PROTO].ptr;
2522                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2523                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2524                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2525                     r->src.port[0], r->src.port[1], th->th_sport))
2526                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
2527                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2528                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2529                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2530                     r->dst.port[0], r->dst.port[1], th->th_dport))
2531                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2532                 else if (r->tos && !(r->tos & pd->tos))
2533                         r = TAILQ_NEXT(r, entries);
2534                 else if (r->rule_flag & PFRULE_FRAGMENT)
2535                         r = TAILQ_NEXT(r, entries);
2536                 else if ((r->flagset & th->th_flags) != r->flags)
2537                         r = TAILQ_NEXT(r, entries);
2538                 else if (r->uid.op && (lookup != -1 || (lookup =
2539                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2540                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2541                     uid))
2542                         r = TAILQ_NEXT(r, entries);
2543                 else if (r->gid.op && (lookup != -1 || (lookup =
2544                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2545                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2546                     gid))
2547                         r = TAILQ_NEXT(r, entries);
2548                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2549                         r = TAILQ_NEXT(r, entries);
2550                 else if (r->anchorname[0] && r->anchor == NULL)
2551                         r = TAILQ_NEXT(r, entries);
2552                 else if (r->os_fingerprint != PF_OSFP_ANY && !pf_osfp_match(
2553                     pf_osfp_fingerprint(pd, m, off, th), r->os_fingerprint))
2554                         r = TAILQ_NEXT(r, entries);
2555                 else {
2556                         if (r->tag)
2557                                 tag = r->tag;
2558                         if (r->anchor == NULL) {
2559                                 *rm = r;
2560                                 *am = a;
2561                                 *rsm = ruleset;
2562                                 if ((*rm)->quick)
2563                                         break;
2564                                 r = TAILQ_NEXT(r, entries);
2565                         } else
2566                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2567                                     PF_RULESET_FILTER);
2568                 }
2569                 if (r == NULL && a != NULL)
2570                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2571                             PF_RULESET_FILTER);
2572         }
2573         r = *rm;
2574         a = *am;
2575         ruleset = *rsm;
2576
2577         REASON_SET(&reason, PFRES_MATCH);
2578
2579         if (r->log) {
2580                 if (rewrite)
2581                         m_copyback(m, off, sizeof(*th), (caddr_t)th);
2582                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2583         }
2584
2585         if ((r->action == PF_DROP) &&
2586             ((r->rule_flag & PFRULE_RETURNRST) ||
2587             (r->rule_flag & PFRULE_RETURNICMP) ||
2588             (r->rule_flag & PFRULE_RETURN))) {
2589                 /* undo NAT changes, if they have taken place */
2590                 if (nr != NULL) {
2591                         if (direction == PF_OUT) {
2592                                 pf_change_ap(saddr, &th->th_sport, pd->ip_sum,
2593                                     &th->th_sum, &pd->baddr, bport, 0, af);
2594                                 rewrite++;
2595                         } else {
2596                                 pf_change_ap(daddr, &th->th_dport, pd->ip_sum,
2597                                     &th->th_sum, &pd->baddr, bport, 0, af);
2598                                 rewrite++;
2599                         }
2600                 }
2601                 if (((r->rule_flag & PFRULE_RETURNRST) ||
2602                     (r->rule_flag & PFRULE_RETURN)) &&
2603                     !(th->th_flags & TH_RST)) {
2604                         u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
2605
2606                         if (th->th_flags & TH_SYN)
2607                                 ack++;
2608                         if (th->th_flags & TH_FIN)
2609                                 ack++;
2610                         pf_send_tcp(r, af, pd->dst,
2611                             pd->src, th->th_dport, th->th_sport,
2612                             ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0,
2613                             r->return_ttl);
2614                 } else if ((af == AF_INET) && r->return_icmp)
2615                         pf_send_icmp(m, r->return_icmp >> 8,
2616                             r->return_icmp & 255, af, r);
2617                 else if ((af == AF_INET6) && r->return_icmp6)
2618                         pf_send_icmp(m, r->return_icmp6 >> 8,
2619                             r->return_icmp6 & 255, af, r);
2620         }
2621
2622         if (r->action == PF_DROP)
2623                 return (PF_DROP);
2624
2625         pf_tag_packet(m, tag);
2626
2627         if (r->keep_state || nr != NULL ||
2628             (pd->flags & PFDESC_TCP_NORM)) {
2629                 /* create new state */
2630                 u_int16_t        len;
2631                 struct pf_state *s = NULL;
2632                 struct pf_src_node *sn = NULL;
2633
2634                 len = pd->tot_len - off - (th->th_off << 2);
2635
2636                 /* check maximums */
2637                 if (r->max_states && (r->states >= r->max_states))
2638                         goto cleanup;
2639                 /* src node for flter rule */
2640                 if ((r->rule_flag & PFRULE_SRCTRACK ||
2641                     r->rpool.opts & PF_POOL_STICKYADDR) &&
2642                     pf_insert_src_node(&sn, r, saddr, af) != 0)
2643                         goto cleanup;
2644                 /* src node for translation rule */
2645                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2646                     ((direction == PF_OUT &&
2647                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2648                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2649                         goto cleanup;
2650                 s = pool_get(&pf_state_pl, PR_NOWAIT);
2651                 if (s == NULL) {
2652 cleanup:
2653                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2654                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2655                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2656                                 pf_status.src_nodes--;
2657                                 pool_put(&pf_src_tree_pl, sn);
2658                         }
2659                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2660                             nsn->expire == 0) {
2661                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2662                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2663                                 pf_status.src_nodes--;
2664                                 pool_put(&pf_src_tree_pl, nsn);
2665                         }
2666                         REASON_SET(&reason, PFRES_MEMORY);
2667                         return (PF_DROP);
2668                 }
2669                 bzero(s, sizeof(*s));
2670                 r->states++;
2671                 if (a != NULL)
2672                         a->states++;
2673                 s->rule.ptr = r;
2674                 s->nat_rule.ptr = nr;
2675                 if (s->nat_rule.ptr != NULL)
2676                         s->nat_rule.ptr->states++;
2677                 s->anchor.ptr = a;
2678                 s->allow_opts = r->allow_opts;
2679                 s->log = r->log & 2;
2680                 s->proto = IPPROTO_TCP;
2681                 s->direction = direction;
2682                 s->af = af;
2683                 if (direction == PF_OUT) {
2684                         PF_ACPY(&s->gwy.addr, saddr, af);
2685                         s->gwy.port = th->th_sport;             /* sport */
2686                         PF_ACPY(&s->ext.addr, daddr, af);
2687                         s->ext.port = th->th_dport;
2688                         if (nr != NULL) {
2689                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
2690                                 s->lan.port = bport;
2691                         } else {
2692                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
2693                                 s->lan.port = s->gwy.port;
2694                         }
2695                 } else {
2696                         PF_ACPY(&s->lan.addr, daddr, af);
2697                         s->lan.port = th->th_dport;
2698                         PF_ACPY(&s->ext.addr, saddr, af);
2699                         s->ext.port = th->th_sport;
2700                         if (nr != NULL) {
2701                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
2702                                 s->gwy.port = bport;
2703                         } else {
2704                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
2705                                 s->gwy.port = s->lan.port;
2706                         }
2707                 }
2708
2709                 s->src.seqlo = ntohl(th->th_seq);
2710                 s->src.seqhi = s->src.seqlo + len + 1;
2711                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2712                     r->keep_state == PF_STATE_MODULATE) {
2713                         /* Generate sequence number modulator */
2714                         while ((s->src.seqdiff = arc4random()) == 0)
2715                                 ;
2716                         pf_change_a(&th->th_seq, &th->th_sum,
2717                             htonl(s->src.seqlo + s->src.seqdiff), 0);
2718                         rewrite = 1;
2719                 } else
2720                         s->src.seqdiff = 0;
2721                 if (th->th_flags & TH_SYN) {
2722                         s->src.seqhi++;
2723                         s->src.wscale = pf_get_wscale(m, off, th->th_off, af);
2724                 }
2725                 s->src.max_win = MAX(ntohs(th->th_win), 1);
2726                 if (s->src.wscale & PF_WSCALE_MASK) {
2727                         /* Remove scale factor from initial window */
2728                         int win = s->src.max_win;
2729                         win += 1 << (s->src.wscale & PF_WSCALE_MASK);
2730                         s->src.max_win = (win - 1) >>
2731                             (s->src.wscale & PF_WSCALE_MASK);
2732                 }
2733                 if (th->th_flags & TH_FIN)
2734                         s->src.seqhi++;
2735                 s->dst.seqhi = 1;
2736                 s->dst.max_win = 1;
2737                 s->src.state = TCPS_SYN_SENT;
2738                 s->dst.state = TCPS_CLOSED;
2739                 s->creation = time_second;
2740                 s->expire = time_second;
2741                 s->timeout = PFTM_TCP_FIRST_PACKET;
2742                 pf_set_rt_ifp(s, saddr);
2743                 if (sn != NULL) {
2744                         s->src_node = sn;
2745                         s->src_node->states++;
2746                 }
2747                 if (nsn != NULL) {
2748                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
2749                         s->nat_src_node = nsn;
2750                         s->nat_src_node->states++;
2751                 }
2752                 if ((pd->flags & PFDESC_TCP_NORM) && pf_normalize_tcp_init(m,
2753                     off, pd, th, &s->src, &s->dst)) {
2754                         REASON_SET(&reason, PFRES_MEMORY);
2755                         pf_src_tree_remove_state(s);
2756                         pool_put(&pf_state_pl, s);
2757                         return (PF_DROP);
2758                 }
2759                 if ((pd->flags & PFDESC_TCP_NORM) && s->src.scrub &&
2760                     pf_normalize_tcp_stateful(m, off, pd, &reason, th, &s->src,
2761                     &s->dst, &rewrite)) {
2762                         pf_normalize_tcp_cleanup(s);
2763                         pf_src_tree_remove_state(s);
2764                         pool_put(&pf_state_pl, s);
2765                         return (PF_DROP);
2766                 }
2767                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
2768                         pf_normalize_tcp_cleanup(s);
2769                         REASON_SET(&reason, PFRES_MEMORY);
2770                         pf_src_tree_remove_state(s);
2771                         pool_put(&pf_state_pl, s);
2772                         return (PF_DROP);
2773                 } else
2774                         *sm = s;
2775                 if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN &&
2776                     r->keep_state == PF_STATE_SYNPROXY) {
2777                         s->src.state = PF_TCPS_PROXY_SRC;
2778                         if (nr != NULL) {
2779                                 if (direction == PF_OUT) {
2780                                         pf_change_ap(saddr, &th->th_sport,
2781                                             pd->ip_sum, &th->th_sum, &pd->baddr,
2782                                             bport, 0, af);
2783                                 } else {
2784                                         pf_change_ap(daddr, &th->th_dport,
2785                                             pd->ip_sum, &th->th_sum, &pd->baddr,
2786                                             bport, 0, af);
2787                                 }
2788                         }
2789                         s->src.seqhi = arc4random();
2790                         /* Find mss option */
2791                         mss = pf_get_mss(m, off, th->th_off, af);
2792                         mss = pf_calc_mss(saddr, af, mss);
2793                         mss = pf_calc_mss(daddr, af, mss);
2794                         s->src.mss = mss;
2795                         pf_send_tcp(r, af, daddr, saddr, th->th_dport,
2796                             th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1,
2797                             TH_SYN|TH_ACK, 0, s->src.mss, 0);
2798                         return (PF_SYNPROXY_DROP);
2799                 }
2800         }
2801
2802         /* copy back packet headers if we performed NAT operations */
2803         if (rewrite)
2804                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
2805
2806         return (PF_PASS);
2807 }
2808
2809 int
2810 pf_test_udp(struct pf_rule **rm, struct pf_state **sm, int direction,
2811     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
2812     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
2813 {
2814         struct pf_rule          *nr = NULL;
2815         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
2816         struct udphdr           *uh = pd->hdr.udp;
2817         u_int16_t                bport, nport = 0;
2818         sa_family_t              af = pd->af;
2819         int                      lookup = -1;
2820         uid_t                    uid;
2821         gid_t                    gid;
2822         struct pf_rule          *r, *a = NULL;
2823         struct pf_ruleset       *ruleset = NULL;
2824         struct pf_src_node      *nsn = NULL;
2825         u_short                  reason;
2826         int                      rewrite = 0;
2827         int                      tag = -1;
2828
2829         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
2830
2831         if (direction == PF_OUT) {
2832                 bport = nport = uh->uh_sport;
2833                 /* check outgoing packet for BINAT/NAT */
2834                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
2835                     saddr, uh->uh_sport, daddr, uh->uh_dport,
2836                     &pd->naddr, &nport)) != NULL) {
2837                         PF_ACPY(&pd->baddr, saddr, af);
2838                         pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2839                             &uh->uh_sum, &pd->naddr, nport, 1, af);
2840                         rewrite++;
2841                         if (nr->natpass)
2842                                 r = NULL;
2843                         pd->nat_rule = nr;
2844                 }
2845         } else {
2846                 bport = nport = uh->uh_dport;
2847                 /* check incoming packet for BINAT/RDR */
2848                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
2849                     saddr, uh->uh_sport, daddr, uh->uh_dport, &pd->naddr,
2850                     &nport)) != NULL) {
2851                         PF_ACPY(&pd->baddr, daddr, af);
2852                         pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2853                             &uh->uh_sum, &pd->naddr, nport, 1, af);
2854                         rewrite++;
2855                         if (nr->natpass)
2856                                 r = NULL;
2857                         pd->nat_rule = nr;
2858                 }
2859         }
2860
2861         while (r != NULL) {
2862                 r->evaluations++;
2863                 if (r->kif != NULL &&
2864                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
2865                         r = r->skip[PF_SKIP_IFP].ptr;
2866                 else if (r->direction && r->direction != direction)
2867                         r = r->skip[PF_SKIP_DIR].ptr;
2868                 else if (r->af && r->af != af)
2869                         r = r->skip[PF_SKIP_AF].ptr;
2870                 else if (r->proto && r->proto != IPPROTO_UDP)
2871                         r = r->skip[PF_SKIP_PROTO].ptr;
2872                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
2873                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
2874                 else if (r->src.port_op && !pf_match_port(r->src.port_op,
2875                     r->src.port[0], r->src.port[1], uh->uh_sport))
2876                         r = r->skip[PF_SKIP_SRC_PORT].ptr;
2877                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
2878                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
2879                 else if (r->dst.port_op && !pf_match_port(r->dst.port_op,
2880                     r->dst.port[0], r->dst.port[1], uh->uh_dport))
2881                         r = r->skip[PF_SKIP_DST_PORT].ptr;
2882                 else if (r->tos && !(r->tos & pd->tos))
2883                         r = TAILQ_NEXT(r, entries);
2884                 else if (r->rule_flag & PFRULE_FRAGMENT)
2885                         r = TAILQ_NEXT(r, entries);
2886                 else if (r->uid.op && (lookup != -1 || (lookup =
2887                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2888                     !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1],
2889                     uid))
2890                         r = TAILQ_NEXT(r, entries);
2891                 else if (r->gid.op && (lookup != -1 || (lookup =
2892                     pf_socket_lookup(&uid, &gid, direction, pd), 1)) &&
2893                     !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1],
2894                     gid))
2895                         r = TAILQ_NEXT(r, entries);
2896                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
2897                         r = TAILQ_NEXT(r, entries);
2898                 else if (r->anchorname[0] && r->anchor == NULL)
2899                         r = TAILQ_NEXT(r, entries);
2900                 else if (r->os_fingerprint != PF_OSFP_ANY)
2901                         r = TAILQ_NEXT(r, entries);
2902                 else {
2903                         if (r->tag)
2904                                 tag = r->tag;
2905                         if (r->anchor == NULL) {
2906                                 *rm = r;
2907                                 *am = a;
2908                                 *rsm = ruleset;
2909                                 if ((*rm)->quick)
2910                                         break;
2911                                 r = TAILQ_NEXT(r, entries);
2912                         } else
2913                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
2914                                     PF_RULESET_FILTER);
2915                 }
2916                 if (r == NULL && a != NULL)
2917                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
2918                             PF_RULESET_FILTER);
2919         }
2920         r = *rm;
2921         a = *am;
2922         ruleset = *rsm;
2923
2924         REASON_SET(&reason, PFRES_MATCH);
2925
2926         if (r->log) {
2927                 if (rewrite)
2928                         m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
2929                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
2930         }
2931
2932         if ((r->action == PF_DROP) &&
2933             ((r->rule_flag & PFRULE_RETURNICMP) ||
2934             (r->rule_flag & PFRULE_RETURN))) {
2935                 /* undo NAT changes, if they have taken place */
2936                 if (nr != NULL) {
2937                         if (direction == PF_OUT) {
2938                                 pf_change_ap(saddr, &uh->uh_sport, pd->ip_sum,
2939                                     &uh->uh_sum, &pd->baddr, bport, 1, af);
2940                                 rewrite++;
2941                         } else {
2942                                 pf_change_ap(daddr, &uh->uh_dport, pd->ip_sum,
2943                                     &uh->uh_sum, &pd->baddr, bport, 1, af);
2944                                 rewrite++;
2945                         }
2946                 }
2947                 if ((af == AF_INET) && r->return_icmp)
2948                         pf_send_icmp(m, r->return_icmp >> 8,
2949                             r->return_icmp & 255, af, r);
2950                 else if ((af == AF_INET6) && r->return_icmp6)
2951                         pf_send_icmp(m, r->return_icmp6 >> 8,
2952                             r->return_icmp6 & 255, af, r);
2953         }
2954
2955         if (r->action == PF_DROP)
2956                 return (PF_DROP);
2957
2958         pf_tag_packet(m, tag);
2959
2960         if (r->keep_state || nr != NULL) {
2961                 /* create new state */
2962                 struct pf_state *s = NULL;
2963                 struct pf_src_node *sn = NULL;
2964
2965                 /* check maximums */
2966                 if (r->max_states && (r->states >= r->max_states))
2967                         goto cleanup;
2968                 /* src node for flter rule */
2969                 if ((r->rule_flag & PFRULE_SRCTRACK ||
2970                     r->rpool.opts & PF_POOL_STICKYADDR) &&
2971                     pf_insert_src_node(&sn, r, saddr, af) != 0)
2972                         goto cleanup;
2973                 /* src node for translation rule */
2974                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
2975                     ((direction == PF_OUT &&
2976                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
2977                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
2978                         goto cleanup;
2979                 s = pool_get(&pf_state_pl, PR_NOWAIT);
2980                 if (s == NULL) {
2981 cleanup:
2982                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
2983                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
2984                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2985                                 pf_status.src_nodes--;
2986                                 pool_put(&pf_src_tree_pl, sn);
2987                         }
2988                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
2989                             nsn->expire == 0) {
2990                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
2991                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
2992                                 pf_status.src_nodes--;
2993                                 pool_put(&pf_src_tree_pl, nsn);
2994                         }
2995                         REASON_SET(&reason, PFRES_MEMORY);
2996                         return (PF_DROP);
2997                 }
2998                 bzero(s, sizeof(*s));
2999                 r->states++;
3000                 if (a != NULL)
3001                         a->states++;
3002                 s->rule.ptr = r;
3003                 s->nat_rule.ptr = nr;
3004                 if (s->nat_rule.ptr != NULL)
3005                         s->nat_rule.ptr->states++;
3006                 s->anchor.ptr = a;
3007                 s->allow_opts = r->allow_opts;
3008                 s->log = r->log & 2;
3009                 s->proto = IPPROTO_UDP;
3010                 s->direction = direction;
3011                 s->af = af;
3012                 if (direction == PF_OUT) {
3013                         PF_ACPY(&s->gwy.addr, saddr, af);
3014                         s->gwy.port = uh->uh_sport;
3015                         PF_ACPY(&s->ext.addr, daddr, af);
3016                         s->ext.port = uh->uh_dport;
3017                         if (nr != NULL) {
3018                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3019                                 s->lan.port = bport;
3020                         } else {
3021                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3022                                 s->lan.port = s->gwy.port;
3023                         }
3024                 } else {
3025                         PF_ACPY(&s->lan.addr, daddr, af);
3026                         s->lan.port = uh->uh_dport;
3027                         PF_ACPY(&s->ext.addr, saddr, af);
3028                         s->ext.port = uh->uh_sport;
3029                         if (nr != NULL) {
3030                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3031                                 s->gwy.port = bport;
3032                         } else {
3033                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3034                                 s->gwy.port = s->lan.port;
3035                         }
3036                 }
3037                 s->src.state = PFUDPS_SINGLE;
3038                 s->dst.state = PFUDPS_NO_TRAFFIC;
3039                 s->creation = time_second;
3040                 s->expire = time_second;
3041                 s->timeout = PFTM_UDP_FIRST_PACKET;
3042                 pf_set_rt_ifp(s, saddr);
3043                 if (sn != NULL) {
3044                         s->src_node = sn;
3045                         s->src_node->states++;
3046                 }
3047                 if (nsn != NULL) {
3048                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3049                         s->nat_src_node = nsn;
3050                         s->nat_src_node->states++;
3051                 }
3052                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3053                         REASON_SET(&reason, PFRES_MEMORY);
3054                         pf_src_tree_remove_state(s);
3055                         pool_put(&pf_state_pl, s);
3056                         return (PF_DROP);
3057                 } else
3058                         *sm = s;
3059         }
3060
3061         /* copy back packet headers if we performed NAT operations */
3062         if (rewrite)
3063                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
3064
3065         return (PF_PASS);
3066 }
3067
3068 int
3069 pf_test_icmp(struct pf_rule **rm, struct pf_state **sm, int direction,
3070     struct pfi_kif *kif, struct mbuf *m, int off, void *h,
3071     struct pf_pdesc *pd, struct pf_rule **am, struct pf_ruleset **rsm)
3072 {
3073         struct pf_rule          *nr = NULL;
3074         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3075         struct pf_rule          *r, *a = NULL;
3076         struct pf_ruleset       *ruleset = NULL;
3077         struct pf_src_node      *nsn = NULL;
3078         u_short                  reason;
3079         u_int16_t                icmpid = 0;
3080         sa_family_t              af = pd->af;
3081         u_int8_t                 icmptype = 0, icmpcode = 0;
3082         int                      state_icmp = 0;
3083         int                      tag = -1;
3084 #ifdef INET6
3085         int                      rewrite = 0;
3086 #endif /* INET6 */
3087
3088         switch (pd->proto) {
3089 #ifdef INET
3090         case IPPROTO_ICMP:
3091                 icmptype = pd->hdr.icmp->icmp_type;
3092                 icmpcode = pd->hdr.icmp->icmp_code;
3093                 icmpid = pd->hdr.icmp->icmp_id;
3094
3095                 if (icmptype == ICMP_UNREACH ||
3096                     icmptype == ICMP_SOURCEQUENCH ||
3097                     icmptype == ICMP_REDIRECT ||
3098                     icmptype == ICMP_TIMXCEED ||
3099                     icmptype == ICMP_PARAMPROB)
3100                         state_icmp++;
3101                 break;
3102 #endif /* INET */
3103 #ifdef INET6
3104         case IPPROTO_ICMPV6:
3105                 icmptype = pd->hdr.icmp6->icmp6_type;
3106                 icmpcode = pd->hdr.icmp6->icmp6_code;
3107                 icmpid = pd->hdr.icmp6->icmp6_id;
3108
3109                 if (icmptype == ICMP6_DST_UNREACH ||
3110                     icmptype == ICMP6_PACKET_TOO_BIG ||
3111                     icmptype == ICMP6_TIME_EXCEEDED ||
3112                     icmptype == ICMP6_PARAM_PROB)
3113                         state_icmp++;
3114                 break;
3115 #endif /* INET6 */
3116         }
3117
3118         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3119
3120         if (direction == PF_OUT) {
3121                 /* check outgoing packet for BINAT/NAT */
3122                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3123                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3124                         PF_ACPY(&pd->baddr, saddr, af);
3125                         switch (af) {
3126 #ifdef INET
3127                         case AF_INET:
3128                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3129                                     pd->naddr.v4.s_addr, 0);
3130                                 break;
3131 #endif /* INET */
3132 #ifdef INET6
3133                         case AF_INET6:
3134                                 pf_change_a6(saddr, &pd->hdr.icmp6->icmp6_cksum,
3135                                     &pd->naddr, 0);
3136                                 rewrite++;
3137                                 break;
3138 #endif /* INET6 */
3139                         }
3140                         if (nr->natpass)
3141                                 r = NULL;
3142                         pd->nat_rule = nr;
3143                 }
3144         } else {
3145                 /* check incoming packet for BINAT/RDR */
3146                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3147                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3148                         PF_ACPY(&pd->baddr, daddr, af);
3149                         switch (af) {
3150 #ifdef INET
3151                         case AF_INET:
3152                                 pf_change_a(&daddr->v4.s_addr,
3153                                     pd->ip_sum, pd->naddr.v4.s_addr, 0);
3154                                 break;
3155 #endif /* INET */
3156 #ifdef INET6
3157                         case AF_INET6:
3158                                 pf_change_a6(daddr, &pd->hdr.icmp6->icmp6_cksum,
3159                                     &pd->naddr, 0);
3160                                 rewrite++;
3161                                 break;
3162 #endif /* INET6 */
3163                         }
3164                         if (nr->natpass)
3165                                 r = NULL;
3166                         pd->nat_rule = nr;
3167                 }
3168         }
3169
3170         while (r != NULL) {
3171                 r->evaluations++;
3172                 if (r->kif != NULL &&
3173                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3174                         r = r->skip[PF_SKIP_IFP].ptr;
3175                 else if (r->direction && r->direction != direction)
3176                         r = r->skip[PF_SKIP_DIR].ptr;
3177                 else if (r->af && r->af != af)
3178                         r = r->skip[PF_SKIP_AF].ptr;
3179                 else if (r->proto && r->proto != pd->proto)
3180                         r = r->skip[PF_SKIP_PROTO].ptr;
3181                 else if (PF_MISMATCHAW(&r->src.addr, saddr, af, r->src.not))
3182                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3183                 else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, r->dst.not))
3184                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3185                 else if (r->type && r->type != icmptype + 1)
3186                         r = TAILQ_NEXT(r, entries);
3187                 else if (r->code && r->code != icmpcode + 1)
3188                         r = TAILQ_NEXT(r, entries);
3189                 else if (r->tos && !(r->tos & pd->tos))
3190                         r = TAILQ_NEXT(r, entries);
3191                 else if (r->rule_flag & PFRULE_FRAGMENT)
3192                         r = TAILQ_NEXT(r, entries);
3193                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3194                         r = TAILQ_NEXT(r, entries);
3195                 else if (r->anchorname[0] && r->anchor == NULL)
3196                         r = TAILQ_NEXT(r, entries);
3197                 else if (r->os_fingerprint != PF_OSFP_ANY)
3198                         r = TAILQ_NEXT(r, entries);
3199                 else {
3200                         if (r->tag)
3201                                 tag = r->tag;
3202                         if (r->anchor == NULL) {
3203                                 *rm = r;
3204                                 *am = a;
3205                                 *rsm = ruleset;
3206                                 if ((*rm)->quick)
3207                                         break;
3208                                 r = TAILQ_NEXT(r, entries);
3209                         } else
3210                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3211                                     PF_RULESET_FILTER);
3212                 }
3213                 if (r == NULL && a != NULL)
3214                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3215                             PF_RULESET_FILTER);
3216         }
3217         r = *rm;
3218         a = *am;
3219         ruleset = *rsm;
3220
3221         REASON_SET(&reason, PFRES_MATCH);
3222
3223         if (r->log) {
3224 #ifdef INET6
3225                 if (rewrite)
3226                         m_copyback(m, off, sizeof(struct icmp6_hdr),
3227                             (caddr_t)pd->hdr.icmp6);
3228 #endif /* INET6 */
3229                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3230         }
3231
3232         if (r->action != PF_PASS)
3233                 return (PF_DROP);
3234
3235         pf_tag_packet(m, tag);
3236
3237         if (!state_icmp && (r->keep_state || nr != NULL)) {
3238                 /* create new state */
3239                 struct pf_state *s = NULL;
3240                 struct pf_src_node *sn = NULL;
3241
3242                 /* check maximums */
3243                 if (r->max_states && (r->states >= r->max_states))
3244                         goto cleanup;
3245                 /* src node for flter rule */
3246                 if ((r->rule_flag & PFRULE_SRCTRACK ||
3247                     r->rpool.opts & PF_POOL_STICKYADDR) &&
3248                     pf_insert_src_node(&sn, r, saddr, af) != 0)
3249                         goto cleanup;
3250                 /* src node for translation rule */
3251                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3252                     ((direction == PF_OUT &&
3253                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3254                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3255                         goto cleanup;
3256                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3257                 if (s == NULL) {
3258 cleanup:
3259                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3260                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3261                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3262                                 pf_status.src_nodes--;
3263                                 pool_put(&pf_src_tree_pl, sn);
3264                         }
3265                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3266                             nsn->expire == 0) {
3267                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3268                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3269                                 pf_status.src_nodes--;
3270                                 pool_put(&pf_src_tree_pl, nsn);
3271                         }
3272                         REASON_SET(&reason, PFRES_MEMORY);
3273                         return (PF_DROP);
3274                 }
3275                 bzero(s, sizeof(*s));
3276                 r->states++;
3277                 if (a != NULL)
3278                         a->states++;
3279                 s->rule.ptr = r;
3280                 s->nat_rule.ptr = nr;
3281                 if (s->nat_rule.ptr != NULL)
3282                         s->nat_rule.ptr->states++;
3283                 s->anchor.ptr = a;
3284                 s->allow_opts = r->allow_opts;
3285                 s->log = r->log & 2;
3286                 s->proto = pd->proto;
3287                 s->direction = direction;
3288                 s->af = af;
3289                 if (direction == PF_OUT) {
3290                         PF_ACPY(&s->gwy.addr, saddr, af);
3291                         s->gwy.port = icmpid;
3292                         PF_ACPY(&s->ext.addr, daddr, af);
3293                         s->ext.port = icmpid;
3294                         if (nr != NULL)
3295                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3296                         else
3297                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3298                         s->lan.port = icmpid;
3299                 } else {
3300                         PF_ACPY(&s->lan.addr, daddr, af);
3301                         s->lan.port = icmpid;
3302                         PF_ACPY(&s->ext.addr, saddr, af);
3303                         s->ext.port = icmpid;
3304                         if (nr != NULL)
3305                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3306                         else
3307                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3308                         s->gwy.port = icmpid;
3309                 }
3310                 s->creation = time_second;
3311                 s->expire = time_second;
3312                 s->timeout = PFTM_ICMP_FIRST_PACKET;
3313                 pf_set_rt_ifp(s, saddr);
3314                 if (sn != NULL) {
3315                         s->src_node = sn;
3316                         s->src_node->states++;
3317                 }
3318                 if (nsn != NULL) {
3319                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3320                         s->nat_src_node = nsn;
3321                         s->nat_src_node->states++;
3322                 }
3323                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3324                         REASON_SET(&reason, PFRES_MEMORY);
3325                         pf_src_tree_remove_state(s);
3326                         pool_put(&pf_state_pl, s);
3327                         return (PF_DROP);
3328                 } else
3329                         *sm = s;
3330         }
3331
3332 #ifdef INET6
3333         /* copy back packet headers if we performed IPv6 NAT operations */
3334         if (rewrite)
3335                 m_copyback(m, off, sizeof(struct icmp6_hdr),
3336                     (caddr_t)pd->hdr.icmp6);
3337 #endif /* INET6 */
3338
3339         return (PF_PASS);
3340 }
3341
3342 int
3343 pf_test_other(struct pf_rule **rm, struct pf_state **sm, int direction,
3344     struct pfi_kif *kif, struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3345     struct pf_rule **am, struct pf_ruleset **rsm)
3346 {
3347         struct pf_rule          *nr = NULL;
3348         struct pf_rule          *r, *a = NULL;
3349         struct pf_ruleset       *ruleset = NULL;
3350         struct pf_src_node      *nsn = NULL;
3351         struct pf_addr          *saddr = pd->src, *daddr = pd->dst;
3352         sa_family_t              af = pd->af;
3353         u_short                  reason;
3354         int                      tag = -1;
3355
3356         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3357
3358         if (direction == PF_OUT) {
3359                 /* check outgoing packet for BINAT/NAT */
3360                 if ((nr = pf_get_translation(pd, m, off, PF_OUT, kif, &nsn,
3361                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3362                         PF_ACPY(&pd->baddr, saddr, af);
3363                         switch (af) {
3364 #ifdef INET
3365                         case AF_INET:
3366                                 pf_change_a(&saddr->v4.s_addr, pd->ip_sum,
3367                                     pd->naddr.v4.s_addr, 0);
3368                                 break;
3369 #endif /* INET */
3370 #ifdef INET6
3371                         case AF_INET6:
3372                                 PF_ACPY(saddr, &pd->naddr, af);
3373                                 break;
3374 #endif /* INET6 */
3375                         }
3376                         if (nr->natpass)
3377                                 r = NULL;
3378                         pd->nat_rule = nr;
3379                 }
3380         } else {
3381                 /* check incoming packet for BINAT/RDR */
3382                 if ((nr = pf_get_translation(pd, m, off, PF_IN, kif, &nsn,
3383                     saddr, 0, daddr, 0, &pd->naddr, NULL)) != NULL) {
3384                         PF_ACPY(&pd->baddr, daddr, af);
3385                         switch (af) {
3386 #ifdef INET
3387                         case AF_INET:
3388                                 pf_change_a(&daddr->v4.s_addr,
3389                                     pd->ip_sum, pd->naddr.v4.s_addr, 0);
3390                                 break;
3391 #endif /* INET */
3392 #ifdef INET6
3393                         case AF_INET6:
3394                                 PF_ACPY(daddr, &pd->naddr, af);
3395                                 break;
3396 #endif /* INET6 */
3397                         }
3398                         if (nr->natpass)
3399                                 r = NULL;
3400                         pd->nat_rule = nr;
3401                 }
3402         }
3403
3404         while (r != NULL) {
3405                 r->evaluations++;
3406                 if (r->kif != NULL &&
3407                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3408                         r = r->skip[PF_SKIP_IFP].ptr;
3409                 else if (r->direction && r->direction != direction)
3410                         r = r->skip[PF_SKIP_DIR].ptr;
3411                 else if (r->af && r->af != af)
3412                         r = r->skip[PF_SKIP_AF].ptr;
3413                 else if (r->proto && r->proto != pd->proto)
3414                         r = r->skip[PF_SKIP_PROTO].ptr;
3415                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3416                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3417                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3418                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3419                 else if (r->tos && !(r->tos & pd->tos))
3420                         r = TAILQ_NEXT(r, entries);
3421                 else if (r->rule_flag & PFRULE_FRAGMENT)
3422                         r = TAILQ_NEXT(r, entries);
3423                 else if (r->match_tag && !pf_match_tag(m, r, nr, &tag))
3424                         r = TAILQ_NEXT(r, entries);
3425                 else if (r->anchorname[0] && r->anchor == NULL)
3426                         r = TAILQ_NEXT(r, entries);
3427                 else if (r->os_fingerprint != PF_OSFP_ANY)
3428                         r = TAILQ_NEXT(r, entries);
3429                 else {
3430                         if (r->tag)
3431                                 tag = r->tag;
3432                         if (r->anchor == NULL) {
3433                                 *rm = r;
3434                                 *am = a;
3435                                 *rsm = ruleset;
3436                                 if ((*rm)->quick)
3437                                         break;
3438                                 r = TAILQ_NEXT(r, entries);
3439                         } else
3440                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3441                                     PF_RULESET_FILTER);
3442                 }
3443                 if (r == NULL && a != NULL)
3444                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3445                             PF_RULESET_FILTER);
3446         }
3447         r = *rm;
3448         a = *am;
3449         ruleset = *rsm;
3450
3451         REASON_SET(&reason, PFRES_MATCH);
3452
3453         if (r->log)
3454                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3455
3456         if ((r->action == PF_DROP) &&
3457             ((r->rule_flag & PFRULE_RETURNICMP) ||
3458             (r->rule_flag & PFRULE_RETURN))) {
3459                 struct pf_addr *a = NULL;
3460
3461                 if (nr != NULL) {
3462                         if (direction == PF_OUT)
3463                                 a = saddr;
3464                         else
3465                                 a = daddr;
3466                 }
3467                 if (a != NULL) {
3468                         switch (af) {
3469 #ifdef INET
3470                         case AF_INET:
3471                                 pf_change_a(&a->v4.s_addr, pd->ip_sum,
3472                                     pd->baddr.v4.s_addr, 0);
3473                                 break;
3474 #endif /* INET */
3475 #ifdef INET6
3476                         case AF_INET6:
3477                                 PF_ACPY(a, &pd->baddr, af);
3478                                 break;
3479 #endif /* INET6 */
3480                         }
3481                 }
3482                 if ((af == AF_INET) && r->return_icmp)
3483                         pf_send_icmp(m, r->return_icmp >> 8,
3484                             r->return_icmp & 255, af, r);
3485                 else if ((af == AF_INET6) && r->return_icmp6)
3486                         pf_send_icmp(m, r->return_icmp6 >> 8,
3487                             r->return_icmp6 & 255, af, r);
3488         }
3489
3490         if (r->action != PF_PASS)
3491                 return (PF_DROP);
3492
3493         pf_tag_packet(m, tag);
3494
3495         if (r->keep_state || nr != NULL) {
3496                 /* create new state */
3497                 struct pf_state *s = NULL;
3498                 struct pf_src_node *sn = NULL;
3499
3500                 /* check maximums */
3501                 if (r->max_states && (r->states >= r->max_states))
3502                         goto cleanup;
3503                 /* src node for flter rule */
3504                 if ((r->rule_flag & PFRULE_SRCTRACK ||
3505                     r->rpool.opts & PF_POOL_STICKYADDR) &&
3506                     pf_insert_src_node(&sn, r, saddr, af) != 0)
3507                         goto cleanup;
3508                 /* src node for translation rule */
3509                 if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) &&
3510                     ((direction == PF_OUT &&
3511                     pf_insert_src_node(&nsn, nr, &pd->baddr, af) != 0) ||
3512                     (pf_insert_src_node(&nsn, nr, saddr, af) != 0)))
3513                         goto cleanup;
3514                 s = pool_get(&pf_state_pl, PR_NOWAIT);
3515                 if (s == NULL) {
3516 cleanup:
3517                         if (sn != NULL && sn->states == 0 && sn->expire == 0) {
3518                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, sn);
3519                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3520                                 pf_status.src_nodes--;
3521                                 pool_put(&pf_src_tree_pl, sn);
3522                         }
3523                         if (nsn != sn && nsn != NULL && nsn->states == 0 &&
3524                             nsn->expire == 0) {
3525                                 RB_REMOVE(pf_src_tree, &tree_src_tracking, nsn);
3526                                 pf_status.scounters[SCNT_SRC_NODE_REMOVALS]++;
3527                                 pf_status.src_nodes--;
3528                                 pool_put(&pf_src_tree_pl, nsn);
3529                         }
3530                         REASON_SET(&reason, PFRES_MEMORY);
3531                         return (PF_DROP);
3532                 }
3533                 bzero(s, sizeof(*s));
3534                 r->states++;
3535                 if (a != NULL)
3536                         a->states++;
3537                 s->rule.ptr = r;
3538                 s->nat_rule.ptr = nr;
3539                 if (s->nat_rule.ptr != NULL)
3540                         s->nat_rule.ptr->states++;
3541                 s->anchor.ptr = a;
3542                 s->allow_opts = r->allow_opts;
3543                 s->log = r->log & 2;
3544                 s->proto = pd->proto;
3545                 s->direction = direction;
3546                 s->af = af;
3547                 if (direction == PF_OUT) {
3548                         PF_ACPY(&s->gwy.addr, saddr, af);
3549                         PF_ACPY(&s->ext.addr, daddr, af);
3550                         if (nr != NULL)
3551                                 PF_ACPY(&s->lan.addr, &pd->baddr, af);
3552                         else
3553                                 PF_ACPY(&s->lan.addr, &s->gwy.addr, af);
3554                 } else {
3555                         PF_ACPY(&s->lan.addr, daddr, af);
3556                         PF_ACPY(&s->ext.addr, saddr, af);
3557                         if (nr != NULL)
3558                                 PF_ACPY(&s->gwy.addr, &pd->baddr, af);
3559                         else
3560                                 PF_ACPY(&s->gwy.addr, &s->lan.addr, af);
3561                 }
3562                 s->src.state = PFOTHERS_SINGLE;
3563                 s->dst.state = PFOTHERS_NO_TRAFFIC;
3564                 s->creation = time_second;
3565                 s->expire = time_second;
3566                 s->timeout = PFTM_OTHER_FIRST_PACKET;
3567                 pf_set_rt_ifp(s, saddr);
3568                 if (sn != NULL) {
3569                         s->src_node = sn;
3570                         s->src_node->states++;
3571                 }
3572                 if (nsn != NULL) {
3573                         PF_ACPY(&nsn->raddr, &pd->naddr, af);
3574                         s->nat_src_node = nsn;
3575                         s->nat_src_node->states++;
3576                 }
3577                 if (pf_insert_state(BOUND_IFACE(r, kif), s)) {
3578                         REASON_SET(&reason, PFRES_MEMORY);
3579                         pf_src_tree_remove_state(s);
3580                         pool_put(&pf_state_pl, s);
3581                         return (PF_DROP);
3582                 } else
3583                         *sm = s;
3584         }
3585
3586         return (PF_PASS);
3587 }
3588
3589 int
3590 pf_test_fragment(struct pf_rule **rm, int direction, struct pfi_kif *kif,
3591     struct mbuf *m, void *h, struct pf_pdesc *pd, struct pf_rule **am,
3592     struct pf_ruleset **rsm)
3593 {
3594         struct pf_rule          *r, *a = NULL;
3595         struct pf_ruleset       *ruleset = NULL;
3596         sa_family_t              af = pd->af;
3597         u_short                  reason;
3598         int                      tag = -1;
3599
3600         r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr);
3601         while (r != NULL) {
3602                 r->evaluations++;
3603                 if (r->kif != NULL &&
3604                     (r->kif != kif && r->kif != kif->pfik_parent) == !r->ifnot)
3605                         r = r->skip[PF_SKIP_IFP].ptr;
3606                 else if (r->direction && r->direction != direction)
3607                         r = r->skip[PF_SKIP_DIR].ptr;
3608                 else if (r->af && r->af != af)
3609                         r = r->skip[PF_SKIP_AF].ptr;
3610                 else if (r->proto && r->proto != pd->proto)
3611                         r = r->skip[PF_SKIP_PROTO].ptr;
3612                 else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, r->src.not))
3613                         r = r->skip[PF_SKIP_SRC_ADDR].ptr;
3614                 else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, r->dst.not))
3615                         r = r->skip[PF_SKIP_DST_ADDR].ptr;
3616                 else if (r->tos && !(r->tos & pd->tos))
3617                         r = TAILQ_NEXT(r, entries);
3618                 else if (r->src.port_op || r->dst.port_op ||
3619                     r->flagset || r->type || r->code ||
3620                     r->os_fingerprint != PF_OSFP_ANY)
3621                         r = TAILQ_NEXT(r, entries);
3622                 else if (r->match_tag && !pf_match_tag(m, r, NULL, &tag))
3623                         r = TAILQ_NEXT(r, entries);
3624                 else if (r->anchorname[0] && r->anchor == NULL)
3625                         r = TAILQ_NEXT(r, entries);
3626                 else {
3627                         if (r->anchor == NULL) {
3628                                 *rm = r;
3629                                 *am = a;
3630                                 *rsm = ruleset;
3631                                 if ((*rm)->quick)
3632                                         break;
3633                                 r = TAILQ_NEXT(r, entries);
3634                         } else
3635                                 PF_STEP_INTO_ANCHOR(r, a, ruleset,
3636                                     PF_RULESET_FILTER);
3637                 }
3638                 if (r == NULL && a != NULL)
3639                         PF_STEP_OUT_OF_ANCHOR(r, a, ruleset,
3640                             PF_RULESET_FILTER);
3641         }
3642         r = *rm;
3643         a = *am;
3644         ruleset = *rsm;
3645
3646         REASON_SET(&reason, PFRES_MATCH);
3647
3648         if (r->log)
3649                 PFLOG_PACKET(kif, h, m, af, direction, reason, r, a, ruleset);
3650
3651         if (r->action != PF_PASS)
3652                 return (PF_DROP);
3653
3654         pf_tag_packet(m, tag);
3655
3656         return (PF_PASS);
3657 }
3658
3659 int
3660 pf_test_state_tcp(struct pf_state **state, int direction, struct pfi_kif *kif,
3661     struct mbuf *m, int off, void *h, struct pf_pdesc *pd,
3662     u_short *reason)
3663 {
3664         struct pf_state          key;
3665         struct tcphdr           *th = pd->hdr.tcp;
3666         u_int16_t                win = ntohs(th->th_win);
3667         u_int32_t                ack, end, seq;
3668         u_int8_t                 sws, dws;
3669         int                      ackskew;
3670         int                      copyback = 0;
3671         struct pf_state_peer    *src, *dst;
3672
3673         key.af = pd->af;
3674         key.proto = IPPROTO_TCP;
3675         if (direction == PF_IN) {
3676                 PF_ACPY(&key.ext.addr, pd->src, key.af);
3677                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
3678                 key.ext.port = th->th_sport;
3679                 key.gwy.port = th->th_dport;
3680         } else {
3681                 PF_ACPY(&key.lan.addr, pd->src, key.af);
3682                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
3683                 key.lan.port = th->th_sport;
3684                 key.ext.port = th->th_dport;
3685         }
3686
3687         STATE_LOOKUP();
3688
3689         if (direction == (*state)->direction) {
3690                 src = &(*state)->src;
3691                 dst = &(*state)->dst;
3692         } else {
3693                 src = &(*state)->dst;
3694                 dst = &(*state)->src;
3695         }
3696
3697         if ((*state)->src.state == PF_TCPS_PROXY_SRC) {
3698                 if (direction != (*state)->direction)
3699                         return (PF_SYNPROXY_DROP);
3700                 if (th->th_flags & TH_SYN) {
3701                         if (ntohl(th->th_seq) != (*state)->src.seqlo)
3702                                 return (PF_DROP);
3703                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3704                             pd->src, th->th_dport, th->th_sport,
3705                             (*state)->src.seqhi, ntohl(th->th_seq) + 1,
3706                             TH_SYN|TH_ACK, 0, (*state)->src.mss, 0);
3707                         return (PF_SYNPROXY_DROP);
3708                 } else if (!(th->th_flags & TH_ACK) ||
3709                     (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3710                     (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3711                         return (PF_DROP);
3712                 else
3713                         (*state)->src.state = PF_TCPS_PROXY_DST;
3714         }
3715         if ((*state)->src.state == PF_TCPS_PROXY_DST) {
3716                 struct pf_state_host *src, *dst;
3717
3718                 if (direction == PF_OUT) {
3719                         src = &(*state)->gwy;
3720                         dst = &(*state)->ext;
3721                 } else {
3722                         src = &(*state)->ext;
3723                         dst = &(*state)->lan;
3724                 }
3725                 if (direction == (*state)->direction) {
3726                         if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) ||
3727                             (ntohl(th->th_ack) != (*state)->src.seqhi + 1) ||
3728                             (ntohl(th->th_seq) != (*state)->src.seqlo + 1))
3729                                 return (PF_DROP);
3730                         (*state)->src.max_win = MAX(ntohs(th->th_win), 1);
3731                         if ((*state)->dst.seqhi == 1)
3732                                 (*state)->dst.seqhi = arc4random();
3733                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3734                             &dst->addr, src->port, dst->port,
3735                             (*state)->dst.seqhi, 0, TH_SYN, 0,
3736                             (*state)->src.mss, 0);
3737                         return (PF_SYNPROXY_DROP);
3738                 } else if (((th->th_flags & (TH_SYN|TH_ACK)) !=
3739                     (TH_SYN|TH_ACK)) ||
3740                     (ntohl(th->th_ack) != (*state)->dst.seqhi + 1))
3741                         return (PF_DROP);
3742                 else {
3743                         (*state)->dst.max_win = MAX(ntohs(th->th_win), 1);
3744                         (*state)->dst.seqlo = ntohl(th->th_seq);
3745                         pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst,
3746                             pd->src, th->th_dport, th->th_sport,
3747                             ntohl(th->th_ack), ntohl(th->th_seq) + 1,
3748                             TH_ACK, (*state)->src.max_win, 0, 0);
3749                         pf_send_tcp((*state)->rule.ptr, pd->af, &src->addr,
3750                             &dst->addr, src->port, dst->port,
3751                             (*state)->src.seqhi + 1, (*state)->src.seqlo + 1,
3752                             TH_ACK, (*state)->dst.max_win, 0, 0);
3753                         (*state)->src.seqdiff = (*state)->dst.seqhi -
3754                             (*state)->src.seqlo;
3755                         (*state)->dst.seqdiff = (*state)->src.seqhi -
3756                             (*state)->dst.seqlo;
3757                         (*state)->src.seqhi = (*state)->src.seqlo +
3758                             (*state)->src.max_win;
3759                         (*state)->dst.seqhi = (*state)->dst.seqlo +
3760                             (*state)->dst.max_win;
3761                         (*state)->src.wscale = (*state)->dst.wscale = 0;
3762                         (*state)->src.state = (*state)->dst.state =
3763                             TCPS_ESTABLISHED;
3764                         return (PF_SYNPROXY_DROP);
3765                 }
3766         }
3767
3768         if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) {
3769                 sws = src->wscale & PF_WSCALE_MASK;
3770                 dws = dst->wscale & PF_WSCALE_MASK;
3771         } else
3772                 sws = dws = 0;
3773
3774         /*
3775          * Sequence tracking algorithm from Guido van Rooij's paper:
3776          *   http://www.madison-gurkha.com/publications/tcp_filtering/
3777          *      tcp_filtering.ps
3778          */
3779
3780         seq = ntohl(th->th_seq);
3781         if (src->seqlo == 0) {
3782                 /* First packet from this end. Set its state */
3783
3784                 if ((pd->flags & PFDESC_TCP_NORM || dst->scrub) &&
3785                     src->scrub == NULL) {
3786                         if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) {
3787                                 REASON_SET(reason, PFRES_MEMORY);
3788                                 return (PF_DROP);
3789                         }
3790                 }
3791
3792                 /* Deferred generation of sequence number modulator */
3793                 if (dst->seqdiff && !src->seqdiff) {
3794                         while ((src->seqdiff = arc4random()) == 0)
3795                                 ;
3796                         ack = ntohl(th->th_ack) - dst->seqdiff;
3797                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3798                             src->seqdiff), 0);
3799                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3800                         copyback = 1;
3801                 } else {
3802                         ack = ntohl(th->th_ack);
3803                 }
3804
3805                 end = seq + pd->p_len;
3806                 if (th->th_flags & TH_SYN) {
3807                         end++;
3808                         if (dst->wscale & PF_WSCALE_FLAG) {
3809                                 src->wscale = pf_get_wscale(m, off, th->th_off,
3810                                     pd->af);
3811                                 if (src->wscale & PF_WSCALE_FLAG) {
3812                                         /* Remove scale factor from initial
3813                                          * window */
3814                                         sws = src->wscale & PF_WSCALE_MASK;
3815                                         win = ((u_int32_t)win + (1 << sws) - 1)
3816                                             >> sws;
3817                                         dws = dst->wscale & PF_WSCALE_MASK;
3818                                 } else {
3819                                         /* fixup other window */
3820                                         dst->max_win <<= dst->wscale &
3821                                             PF_WSCALE_MASK;
3822                                         /* in case of a retrans SYN|ACK */
3823                                         dst->wscale = 0;
3824                                 }
3825                         }
3826                 }
3827                 if (th->th_flags & TH_FIN)
3828                         end++;
3829
3830                 src->seqlo = seq;
3831                 if (src->state < TCPS_SYN_SENT)
3832                         src->state = TCPS_SYN_SENT;
3833
3834                 /*
3835                  * May need to slide the window (seqhi may have been set by
3836                  * the crappy stack check or if we picked up the connection
3837                  * after establishment)
3838                  */
3839                 if (src->seqhi == 1 ||
3840                     SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi))
3841                         src->seqhi = end + MAX(1, dst->max_win << dws);
3842                 if (win > src->max_win)
3843                         src->max_win = win;
3844
3845         } else {
3846                 ack = ntohl(th->th_ack) - dst->seqdiff;
3847                 if (src->seqdiff) {
3848                         /* Modulate sequence numbers */
3849                         pf_change_a(&th->th_seq, &th->th_sum, htonl(seq +
3850                             src->seqdiff), 0);
3851                         pf_change_a(&th->th_ack, &th->th_sum, htonl(ack), 0);
3852                         copyback = 1;
3853                 }
3854                 end = seq + pd->p_len;
3855                 if (th->th_flags & TH_SYN)
3856                         end++;
3857                 if (th->th_flags & TH_FIN)
3858                         end++;
3859         }
3860
3861         if ((th->th_flags & TH_ACK) == 0) {
3862                 /* Let it pass through the ack skew check */
3863                 ack = dst->seqlo;
3864         } else if ((ack == 0 &&
3865             (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) ||
3866             /* broken tcp stacks do not set ack */
3867             (dst->state < TCPS_SYN_SENT)) {
3868                 /*
3869                  * Many stacks (ours included) will set the ACK number in an
3870                  * FIN|ACK if the SYN times out -- no sequence to ACK.
3871                  */
3872                 ack = dst->seqlo;
3873         }
3874
3875         if (seq == end) {
3876                 /* Ease sequencing restrictions on no data packets */
3877                 seq = src->seqlo;
3878                 end = seq;
3879         }
3880
3881         ackskew = dst->seqlo - ack;
3882
3883 #define MAXACKWINDOW (0xffff + 1500)    /* 1500 is an arbitrary fudge factor */
3884         if (SEQ_GEQ(src->seqhi, end) &&
3885             /* Last octet inside other's window space */
3886             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) &&
3887             /* Retrans: not more than one window back */
3888             (ackskew >= -MAXACKWINDOW) &&
3889             /* Acking not more than one reassembled fragment backwards */
3890             (ackskew <= (MAXACKWINDOW << sws))) {
3891             /* Acking not more than one window forward */
3892
3893                 /* update max window */
3894                 if (src->max_win < win)
3895                         src->max_win = win;
3896                 /* synchronize sequencing */
3897                 if (SEQ_GT(end, src->seqlo))
3898                         src->seqlo = end;
3899                 /* slide the window of what the other end can send */
3900                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3901                         dst->seqhi = ack + MAX((win << sws), 1);
3902
3903
3904                 /* update states */
3905                 if (th->th_flags & TH_SYN)
3906                         if (src->state < TCPS_SYN_SENT)
3907                                 src->state = TCPS_SYN_SENT;
3908                 if (th->th_flags & TH_FIN)
3909                         if (src->state < TCPS_CLOSING)
3910                                 src->state = TCPS_CLOSING;
3911                 if (th->th_flags & TH_ACK) {
3912                         if (dst->state == TCPS_SYN_SENT)
3913                                 dst->state = TCPS_ESTABLISHED;
3914                         else if (dst->state == TCPS_CLOSING)
3915                                 dst->state = TCPS_FIN_WAIT_2;
3916                 }
3917                 if (th->th_flags & TH_RST)
3918                         src->state = dst->state = TCPS_TIME_WAIT;
3919
3920                 /* update expire time */
3921                 (*state)->expire = time_second;
3922                 if (src->state >= TCPS_FIN_WAIT_2 &&
3923                     dst->state >= TCPS_FIN_WAIT_2)
3924                         (*state)->timeout = PFTM_TCP_CLOSED;
3925                 else if (src->state >= TCPS_FIN_WAIT_2 ||
3926                     dst->state >= TCPS_FIN_WAIT_2)
3927                         (*state)->timeout = PFTM_TCP_FIN_WAIT;
3928                 else if (src->state < TCPS_ESTABLISHED ||
3929                     dst->state < TCPS_ESTABLISHED)
3930                         (*state)->timeout = PFTM_TCP_OPENING;
3931                 else if (src->state >= TCPS_CLOSING ||
3932                     dst->state >= TCPS_CLOSING)
3933                         (*state)->timeout = PFTM_TCP_CLOSING;
3934                 else
3935                         (*state)->timeout = PFTM_TCP_ESTABLISHED;
3936
3937                 /* Fall through to PASS packet */
3938
3939         } else if ((dst->state < TCPS_SYN_SENT ||
3940                 dst->state >= TCPS_FIN_WAIT_2 ||
3941                 src->state >= TCPS_FIN_WAIT_2) &&
3942             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) &&
3943             /* Within a window forward of the originating packet */
3944             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
3945             /* Within a window backward of the originating packet */
3946
3947                 /*
3948                  * This currently handles three situations:
3949                  *  1) Stupid stacks will shotgun SYNs before their peer
3950                  *     replies.
3951                  *  2) When PF catches an already established stream (the
3952                  *     firewall rebooted, the state table was flushed, routes
3953                  *     changed...)
3954                  *  3) Packets get funky immediately after the connection
3955                  *     closes (this should catch Solaris spurious ACK|FINs
3956                  *     that web servers like to spew after a close)
3957                  *
3958                  * This must be a little more careful than the above code
3959                  * since packet floods will also be caught here. We don't
3960                  * update the TTL here to mitigate the damage of a packet
3961                  * flood and so the same code can handle awkward establishment
3962                  * and a loosened connection close.
3963                  * In the establishment case, a correct peer response will
3964                  * validate the connection, go through the normal state code
3965                  * and keep updating the state TTL.
3966                  */
3967
3968                 if (pf_status.debug >= PF_DEBUG_MISC) {
3969                         printf("pf: loose state match: ");
3970                         pf_print_state(*state);
3971                         pf_print_flags(th->th_flags);
3972                         printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d\n",
3973                             seq, ack, pd->p_len, ackskew,
3974                             (*state)->packets[0], (*state)->packets[1]);
3975                 }
3976
3977                 /* update max window */
3978                 if (src->max_win < win)
3979                         src->max_win = win;
3980                 /* synchronize sequencing */
3981                 if (SEQ_GT(end, src->seqlo))
3982                         src->seqlo = end;
3983                 /* slide the window of what the other end can send */
3984                 if (SEQ_GEQ(ack + (win << sws), dst->seqhi))
3985                         dst->seqhi = ack + MAX((win << sws), 1);
3986
3987                 /*
3988                  * Cannot set dst->seqhi here since this could be a shotgunned
3989                  * SYN and not an already established connection.
3990                  */
3991
3992                 if (th->th_flags & TH_FIN)
3993                         if (src->state < TCPS_CLOSING)
3994                                 src->state = TCPS_CLOSING;
3995                 if (th->th_flags & TH_RST)
3996                         src->state = dst->state = TCPS_TIME_WAIT;
3997
3998                 /* Fall through to PASS packet */
3999
4000         } else {
4001                 if ((*state)->dst.state == TCPS_SYN_SENT &&
4002                     (*state)->src.state == TCPS_SYN_SENT) {
4003                         /* Send RST for state mismatches during handshake */
4004                         if (!(th->th_flags & TH_RST)) {
4005                                 u_int32_t ack = ntohl(th->th_seq) + pd->p_len;
4006
4007                                 if (th->th_flags & TH_SYN)
4008                                         ack++;
4009                                 if (th->th_flags & TH_FIN)
4010                                         ack++;
4011                                 pf_send_tcp((*state)->rule.ptr, pd->af,
4012                                     pd->dst, pd->src, th->th_dport,
4013                                     th->th_sport, ntohl(th->th_ack), ack,
4014                                     TH_RST|TH_ACK, 0, 0,
4015                                     (*state)->rule.ptr->return_ttl);
4016                         }
4017                         src->seqlo = 0;
4018                         src->seqhi = 1;
4019                         src->max_win = 1;
4020                 } else if (pf_status.debug >= PF_DEBUG_MISC) {
4021                         printf("pf: BAD state: ");
4022                         pf_print_state(*state);
4023                         pf_print_flags(th->th_flags);
4024                         printf(" seq=%u ack=%u len=%u ackskew=%d pkts=%d:%d "
4025                             "dir=%s,%s\n", seq, ack, pd->p_len, ackskew,
4026                             (*state)->packets[0], (*state)->packets[1],
4027                             direction == PF_IN ? "in" : "out",
4028                             direction == (*state)->direction ? "fwd" : "rev");
4029                         printf("pf: State failure on: %c %c %c %c | %c %c\n",
4030                             SEQ_GEQ(src->seqhi, end) ? ' ' : '1',
4031                             SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ?
4032                             ' ': '2',
4033                             (ackskew >= -MAXACKWINDOW) ? ' ' : '3',
4034                             (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4',
4035                             SEQ_GEQ(src->seqhi + MAXACKWINDOW, end) ?' ' :'5',
4036                             SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6');
4037                 }
4038                 return (PF_DROP);
4039         }
4040
4041         if (dst->scrub || src->scrub) {
4042                 if (pf_normalize_tcp_stateful(m, off, pd, reason, th,
4043                     src, dst, &copyback))
4044                         return (PF_DROP);
4045         }
4046
4047         /* Any packets which have gotten here are to be passed */
4048
4049         /* translate source/destination address, if necessary */
4050         if (STATE_TRANSLATE(*state)) {
4051                 if (direction == PF_OUT)
4052                         pf_change_ap(pd->src, &th->th_sport, pd->ip_sum,
4053                             &th->th_sum, &(*state)->gwy.addr,
4054                             (*state)->gwy.port, 0, pd->af);
4055                 else
4056                         pf_change_ap(pd->dst, &th->th_dport, pd->ip_sum,
4057                             &th->th_sum, &(*state)->lan.addr,
4058                             (*state)->lan.port, 0, pd->af);
4059                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4060         } else if (copyback) {
4061                 /* Copyback sequence modulation or stateful scrub changes */
4062                 m_copyback(m, off, sizeof(*th), (caddr_t)th);
4063         }
4064
4065         return (PF_PASS);
4066 }
4067
4068 int
4069 pf_test_state_udp(struct pf_state **state, int direction, struct pfi_kif *kif,
4070     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4071 {
4072         struct pf_state_peer    *src, *dst;
4073         struct pf_state          key;
4074         struct udphdr           *uh = pd->hdr.udp;
4075
4076         key.af = pd->af;
4077         key.proto = IPPROTO_UDP;
4078         if (direction == PF_IN) {
4079                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4080                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4081                 key.ext.port = uh->uh_sport;
4082                 key.gwy.port = uh->uh_dport;
4083         } else {
4084                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4085                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4086                 key.lan.port = uh->uh_sport;
4087                 key.ext.port = uh->uh_dport;
4088         }
4089
4090         STATE_LOOKUP();
4091
4092         if (direction == (*state)->direction) {
4093                 src = &(*state)->src;
4094                 dst = &(*state)->dst;
4095         } else {
4096                 src = &(*state)->dst;
4097                 dst = &(*state)->src;
4098         }
4099
4100         /* update states */
4101         if (src->state < PFUDPS_SINGLE)
4102                 src->state = PFUDPS_SINGLE;
4103         if (dst->state == PFUDPS_SINGLE)
4104                 dst->state = PFUDPS_MULTIPLE;
4105
4106         /* update expire time */
4107         (*state)->expire = time_second;
4108         if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE)
4109                 (*state)->timeout = PFTM_UDP_MULTIPLE;
4110         else
4111                 (*state)->timeout = PFTM_UDP_SINGLE;
4112
4113         /* translate source/destination address, if necessary */
4114         if (STATE_TRANSLATE(*state)) {
4115                 if (direction == PF_OUT)
4116                         pf_change_ap(pd->src, &uh->uh_sport, pd->ip_sum,
4117                             &uh->uh_sum, &(*state)->gwy.addr,
4118                             (*state)->gwy.port, 1, pd->af);
4119                 else
4120                         pf_change_ap(pd->dst, &uh->uh_dport, pd->ip_sum,
4121                             &uh->uh_sum, &(*state)->lan.addr,
4122                             (*state)->lan.port, 1, pd->af);
4123                 m_copyback(m, off, sizeof(*uh), (caddr_t)uh);
4124         }
4125
4126         return (PF_PASS);
4127 }
4128
4129 int
4130 pf_test_state_icmp(struct pf_state **state, int direction, struct pfi_kif *kif,
4131     struct mbuf *m, int off, void *h, struct pf_pdesc *pd)
4132 {
4133         struct pf_addr  *saddr = pd->src, *daddr = pd->dst;
4134         u_int16_t        icmpid = 0;
4135         u_int16_t       *icmpsum = NULL;
4136         u_int8_t         icmptype = 0;
4137         int              state_icmp = 0;
4138
4139         switch (pd->proto) {
4140 #ifdef INET
4141         case IPPROTO_ICMP:
4142                 icmptype = pd->hdr.icmp->icmp_type;
4143                 icmpid = pd->hdr.icmp->icmp_id;
4144                 icmpsum = &pd->hdr.icmp->icmp_cksum;
4145
4146                 if (icmptype == ICMP_UNREACH ||
4147                     icmptype == ICMP_SOURCEQUENCH ||
4148                     icmptype == ICMP_REDIRECT ||
4149                     icmptype == ICMP_TIMXCEED ||
4150                     icmptype == ICMP_PARAMPROB)
4151                         state_icmp++;
4152                 break;
4153 #endif /* INET */
4154 #ifdef INET6
4155         case IPPROTO_ICMPV6:
4156                 icmptype = pd->hdr.icmp6->icmp6_type;
4157                 icmpid = pd->hdr.icmp6->icmp6_id;
4158                 icmpsum = &pd->hdr.icmp6->icmp6_cksum;
4159
4160                 if (icmptype == ICMP6_DST_UNREACH ||
4161                     icmptype == ICMP6_PACKET_TOO_BIG ||
4162                     icmptype == ICMP6_TIME_EXCEEDED ||
4163                     icmptype == ICMP6_PARAM_PROB)
4164                         state_icmp++;
4165                 break;
4166 #endif /* INET6 */
4167         }
4168
4169         if (!state_icmp) {
4170
4171                 /*
4172                  * ICMP query/reply message not related to a TCP/UDP packet.
4173                  * Search for an ICMP state.
4174                  */
4175                 struct pf_state         key;
4176
4177                 key.af = pd->af;
4178                 key.proto = pd->proto;
4179                 if (direction == PF_IN) {
4180                         PF_ACPY(&key.ext.addr, pd->src, key.af);
4181                         PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4182                         key.ext.port = icmpid;
4183                         key.gwy.port = icmpid;
4184                 } else {
4185                         PF_ACPY(&key.lan.addr, pd->src, key.af);
4186                         PF_ACPY(&key.ext.addr, pd->dst, key.af);
4187                         key.lan.port = icmpid;
4188                         key.ext.port = icmpid;
4189                 }
4190
4191                 STATE_LOOKUP();
4192
4193                 (*state)->expire = time_second;
4194                 (*state)->timeout = PFTM_ICMP_ERROR_REPLY;
4195
4196                 /* translate source/destination address, if necessary */
4197                 if (PF_ANEQ(&(*state)->lan.addr, &(*state)->gwy.addr, pd->af)) {
4198                         if (direction == PF_OUT) {
4199                                 switch (pd->af) {
4200 #ifdef INET
4201                                 case AF_INET:
4202                                         pf_change_a(&saddr->v4.s_addr,
4203                                             pd->ip_sum,
4204                                             (*state)->gwy.addr.v4.s_addr, 0);
4205                                         break;
4206 #endif /* INET */
4207 #ifdef INET6
4208                                 case AF_INET6:
4209                                         pf_change_a6(saddr,
4210                                             &pd->hdr.icmp6->icmp6_cksum,
4211                                             &(*state)->gwy.addr, 0);
4212                                         m_copyback(m, off,
4213                                             sizeof(struct icmp6_hdr),
4214                                             (caddr_t)pd->hdr.icmp6);
4215                                         break;
4216 #endif /* INET6 */
4217                                 }
4218                         } else {
4219                                 switch (pd->af) {
4220 #ifdef INET
4221                                 case AF_INET:
4222                                         pf_change_a(&daddr->v4.s_addr,
4223                                             pd->ip_sum,
4224                                             (*state)->lan.addr.v4.s_addr, 0);
4225                                         break;
4226 #endif /* INET */
4227 #ifdef INET6
4228                                 case AF_INET6:
4229                                         pf_change_a6(daddr,
4230                                             &pd->hdr.icmp6->icmp6_cksum,
4231                                             &(*state)->lan.addr, 0);
4232                                         m_copyback(m, off,
4233                                             sizeof(struct icmp6_hdr),
4234                                             (caddr_t)pd->hdr.icmp6);
4235                                         break;
4236 #endif /* INET6 */
4237                                 }
4238                         }
4239                 }
4240
4241                 return (PF_PASS);
4242
4243         } else {
4244                 /*
4245                  * ICMP error message in response to a TCP/UDP packet.
4246                  * Extract the inner TCP/UDP header and search for that state.
4247                  */
4248
4249                 struct pf_pdesc pd2;
4250 #ifdef INET
4251                 struct ip       h2;
4252 #endif /* INET */
4253 #ifdef INET6
4254                 struct ip6_hdr  h2_6;
4255                 int             terminal = 0;
4256 #endif /* INET6 */
4257                 int             ipoff2 = 0;
4258                 int             off2 = 0;
4259
4260                 pd2.af = pd->af;
4261                 switch (pd->af) {
4262 #ifdef INET
4263                 case AF_INET:
4264                         /* offset of h2 in mbuf chain */
4265                         ipoff2 = off + ICMP_MINLEN;
4266
4267                         if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2),
4268                             NULL, NULL, pd2.af)) {
4269                                 DPFPRINTF(PF_DEBUG_MISC,
4270                                     ("pf: ICMP error message too short "
4271                                     "(ip)\n"));
4272                                 return (PF_DROP);
4273                         }
4274                         /*
4275                          * ICMP error messages don't refer to non-first
4276                          * fragments
4277                          */
4278                         /*
4279                          * Note: We are dealing with an encapsulated
4280                          * header. This means ip_off/ip_len are not
4281                          * in host byte order!
4282                          */
4283                         if (h2.ip_off & htons(IP_OFFMASK))
4284                                 return (PF_DROP);
4285
4286                         /* offset of protocol header that follows h2 */
4287                         off2 = ipoff2 + (h2.ip_hl << 2);
4288
4289                         pd2.proto = h2.ip_p;
4290                         pd2.src = (struct pf_addr *)&h2.ip_src;
4291                         pd2.dst = (struct pf_addr *)&h2.ip_dst;
4292                         pd2.ip_sum = &h2.ip_sum;
4293                         break;
4294 #endif /* INET */
4295 #ifdef INET6
4296                 case AF_INET6:
4297                         ipoff2 = off + sizeof(struct icmp6_hdr);
4298
4299                         if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6),
4300                             NULL, NULL, pd2.af)) {
4301                                 DPFPRINTF(PF_DEBUG_MISC,
4302                                     ("pf: ICMP error message too short "
4303                                     "(ip6)\n"));
4304                                 return (PF_DROP);
4305                         }
4306                         pd2.proto = h2_6.ip6_nxt;
4307                         pd2.src = (struct pf_addr *)&h2_6.ip6_src;
4308                         pd2.dst = (struct pf_addr *)&h2_6.ip6_dst;
4309                         pd2.ip_sum = NULL;
4310                         off2 = ipoff2 + sizeof(h2_6);
4311                         do {
4312                                 switch (pd2.proto) {
4313                                 case IPPROTO_FRAGMENT:
4314                                         /*
4315                                          * ICMPv6 error messages for
4316                                          * non-first fragments
4317                                          */
4318                                         return (PF_DROP);
4319                                 case IPPROTO_AH:
4320                                 case IPPROTO_HOPOPTS:
4321                                 case IPPROTO_ROUTING:
4322                                 case IPPROTO_DSTOPTS: {
4323                                         /* get next header and header length */
4324                                         struct ip6_ext opt6;
4325
4326                                         if (!pf_pull_hdr(m, off2, &opt6,
4327                                             sizeof(opt6), NULL, NULL, pd2.af)) {
4328                                                 DPFPRINTF(PF_DEBUG_MISC,
4329                                                     ("pf: ICMPv6 short opt\n"));
4330                                                 return (PF_DROP);
4331                                         }
4332                                         if (pd2.proto == IPPROTO_AH)
4333                                                 off2 += (opt6.ip6e_len + 2) * 4;
4334                                         else
4335                                                 off2 += (opt6.ip6e_len + 1) * 8;
4336                                         pd2.proto = opt6.ip6e_nxt;
4337                                         /* goto the next header */
4338                                         break;
4339                                 }
4340                                 default:
4341                                         terminal++;
4342                                         break;
4343                                 }
4344                         } while (!terminal);
4345                         break;
4346 #endif /* INET6 */
4347                 }
4348
4349                 switch (pd2.proto) {
4350                 case IPPROTO_TCP: {
4351                         struct tcphdr            th;
4352                         u_int32_t                seq;
4353                         struct pf_state          key;
4354                         struct pf_state_peer    *src, *dst;
4355                         u_int8_t                 dws;
4356                         int                      copyback = 0;
4357
4358                         /*
4359                          * Only the first 8 bytes of the TCP header can be
4360                          * expected. Don't access any TCP header fields after
4361                          * th_seq, an ackskew test is not possible.
4362                          */
4363                         if (!pf_pull_hdr(m, off2, &th, 8, NULL, NULL, pd2.af)) {
4364                                 DPFPRINTF(PF_DEBUG_MISC,
4365                                     ("pf: ICMP error message too short "
4366                                     "(tcp)\n"));
4367                                 return (PF_DROP);
4368                         }
4369
4370                         key.af = pd2.af;
4371                         key.proto = IPPROTO_TCP;
4372                         if (direction == PF_IN) {
4373                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4374                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4375                                 key.ext.port = th.th_dport;
4376                                 key.gwy.port = th.th_sport;
4377                         } else {
4378                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4379                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4380                                 key.lan.port = th.th_dport;
4381                                 key.ext.port = th.th_sport;
4382                         }
4383
4384                         STATE_LOOKUP();
4385
4386                         if (direction == (*state)->direction) {
4387                                 src = &(*state)->dst;
4388                                 dst = &(*state)->src;
4389                         } else {
4390                                 src = &(*state)->src;
4391                                 dst = &(*state)->dst;
4392                         }
4393
4394                         if (src->wscale && dst->wscale &&
4395                             !(th.th_flags & TH_SYN))
4396                                 dws = dst->wscale & PF_WSCALE_MASK;
4397                         else
4398                                 dws = 0;
4399
4400                         /* Demodulate sequence number */
4401                         seq = ntohl(th.th_seq) - src->seqdiff;
4402                         if (src->seqdiff) {
4403                                 pf_change_a(&th.th_seq, icmpsum,
4404                                     htonl(seq), 0);
4405                                 copyback = 1;
4406                         }
4407
4408                         if (!SEQ_GEQ(src->seqhi, seq) ||
4409                             !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))) {
4410                                 if (pf_status.debug >= PF_DEBUG_MISC) {
4411                                         printf("pf: BAD ICMP %d:%d ",
4412                                             icmptype, pd->hdr.icmp->icmp_code);
4413                                         pf_print_host(pd->src, 0, pd->af);
4414                                         printf(" -> ");
4415                                         pf_print_host(pd->dst, 0, pd->af);
4416                                         printf(" state: ");
4417                                         pf_print_state(*state);
4418                                         printf(" seq=%u\n", seq);
4419                                 }
4420                                 return (PF_DROP);
4421                         }
4422
4423                         if (STATE_TRANSLATE(*state)) {
4424                                 if (direction == PF_IN) {
4425                                         pf_change_icmp(pd2.src, &th.th_sport,
4426                                             daddr, &(*state)->lan.addr,
4427                                             (*state)->lan.port, NULL,
4428                                             pd2.ip_sum, icmpsum,
4429                                             pd->ip_sum, 0, pd2.af);
4430                                 } else {
4431                                         pf_change_icmp(pd2.dst, &th.th_dport,
4432                                             saddr, &(*state)->gwy.addr,
4433                                             (*state)->gwy.port, NULL,
4434                                             pd2.ip_sum, icmpsum,
4435                                             pd->ip_sum, 0, pd2.af);
4436                                 }
4437                                 copyback = 1;
4438                         }
4439
4440                         if (copyback) {
4441                                 switch (pd2.af) {
4442 #ifdef INET
4443                                 case AF_INET:
4444                                         m_copyback(m, off, ICMP_MINLEN,
4445                                             (caddr_t)pd->hdr.icmp);
4446                                         m_copyback(m, ipoff2, sizeof(h2),
4447                                             (caddr_t)&h2);
4448                                         break;
4449 #endif /* INET */
4450 #ifdef INET6
4451                                 case AF_INET6:
4452                                         m_copyback(m, off,
4453                                             sizeof(struct icmp6_hdr),
4454                                             (caddr_t)pd->hdr.icmp6);
4455                                         m_copyback(m, ipoff2, sizeof(h2_6),
4456                                             (caddr_t)&h2_6);
4457                                         break;
4458 #endif /* INET6 */
4459                                 }
4460                                 m_copyback(m, off2, 8, (caddr_t)&th);
4461                         }
4462
4463                         return (PF_PASS);
4464                         break;
4465                 }
4466                 case IPPROTO_UDP: {
4467                         struct udphdr           uh;
4468                         struct pf_state         key;
4469
4470                         if (!pf_pull_hdr(m, off2, &uh, sizeof(uh),
4471                             NULL, NULL, pd2.af)) {
4472                                 DPFPRINTF(PF_DEBUG_MISC,
4473                                     ("pf: ICMP error message too short "
4474                                     "(udp)\n"));
4475                                 return (PF_DROP);
4476                         }
4477
4478                         key.af = pd2.af;
4479                         key.proto = IPPROTO_UDP;
4480                         if (direction == PF_IN) {
4481                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4482                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4483                                 key.ext.port = uh.uh_dport;
4484                                 key.gwy.port = uh.uh_sport;
4485                         } else {
4486                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4487                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4488                                 key.lan.port = uh.uh_dport;
4489                                 key.ext.port = uh.uh_sport;
4490                         }
4491
4492                         STATE_LOOKUP();
4493
4494                         if (STATE_TRANSLATE(*state)) {
4495                                 if (direction == PF_IN) {
4496                                         pf_change_icmp(pd2.src, &uh.uh_sport,
4497                                             daddr, &(*state)->lan.addr,
4498                                             (*state)->lan.port, &uh.uh_sum,
4499                                             pd2.ip_sum, icmpsum,
4500                                             pd->ip_sum, 1, pd2.af);
4501                                 } else {
4502                                         pf_change_icmp(pd2.dst, &uh.uh_dport,
4503                                             saddr, &(*state)->gwy.addr,
4504                                             (*state)->gwy.port, &uh.uh_sum,
4505                                             pd2.ip_sum, icmpsum,
4506                                             pd->ip_sum, 1, pd2.af);
4507                                 }
4508                                 switch (pd2.af) {
4509 #ifdef INET
4510                                 case AF_INET:
4511                                         m_copyback(m, off, ICMP_MINLEN,
4512                                             (caddr_t)pd->hdr.icmp);
4513                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4514                                         break;
4515 #endif /* INET */
4516 #ifdef INET6
4517                                 case AF_INET6:
4518                                         m_copyback(m, off,
4519                                             sizeof(struct icmp6_hdr),
4520                                             (caddr_t)pd->hdr.icmp6);
4521                                         m_copyback(m, ipoff2, sizeof(h2_6),
4522                                             (caddr_t)&h2_6);
4523                                         break;
4524 #endif /* INET6 */
4525                                 }
4526                                 m_copyback(m, off2, sizeof(uh), (caddr_t)&uh);
4527                         }
4528
4529                         return (PF_PASS);
4530                         break;
4531                 }
4532 #ifdef INET
4533                 case IPPROTO_ICMP: {
4534                         struct icmp             iih;
4535                         struct pf_state         key;
4536
4537                         if (!pf_pull_hdr(m, off2, &iih, ICMP_MINLEN,
4538                             NULL, NULL, pd2.af)) {
4539                                 DPFPRINTF(PF_DEBUG_MISC,
4540                                     ("pf: ICMP error message too short i"
4541                                     "(icmp)\n"));
4542                                 return (PF_DROP);
4543                         }
4544
4545                         key.af = pd2.af;
4546                         key.proto = IPPROTO_ICMP;
4547                         if (direction == PF_IN) {
4548                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4549                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4550                                 key.ext.port = iih.icmp_id;
4551                                 key.gwy.port = iih.icmp_id;
4552                         } else {
4553                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4554                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4555                                 key.lan.port = iih.icmp_id;
4556                                 key.ext.port = iih.icmp_id;
4557                         }
4558
4559                         STATE_LOOKUP();
4560
4561                         if (STATE_TRANSLATE(*state)) {
4562                                 if (direction == PF_IN) {
4563                                         pf_change_icmp(pd2.src, &iih.icmp_id,
4564                                             daddr, &(*state)->lan.addr,
4565                                             (*state)->lan.port, NULL,
4566                                             pd2.ip_sum, icmpsum,
4567                                             pd->ip_sum, 0, AF_INET);
4568                                 } else {
4569                                         pf_change_icmp(pd2.dst, &iih.icmp_id,
4570                                             saddr, &(*state)->gwy.addr,
4571                                             (*state)->gwy.port, NULL,
4572                                             pd2.ip_sum, icmpsum,
4573                                             pd->ip_sum, 0, AF_INET);
4574                                 }
4575                                 m_copyback(m, off, ICMP_MINLEN, (caddr_t)pd->hdr.icmp);
4576                                 m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4577                                 m_copyback(m, off2, ICMP_MINLEN, (caddr_t)&iih);
4578                         }
4579
4580                         return (PF_PASS);
4581                         break;
4582                 }
4583 #endif /* INET */
4584 #ifdef INET6
4585                 case IPPROTO_ICMPV6: {
4586                         struct icmp6_hdr        iih;
4587                         struct pf_state         key;
4588
4589                         if (!pf_pull_hdr(m, off2, &iih,
4590                             sizeof(struct icmp6_hdr), NULL, NULL, pd2.af)) {
4591                                 DPFPRINTF(PF_DEBUG_MISC,
4592                                     ("pf: ICMP error message too short "
4593                                     "(icmp6)\n"));
4594                                 return (PF_DROP);
4595                         }
4596
4597                         key.af = pd2.af;
4598                         key.proto = IPPROTO_ICMPV6;
4599                         if (direction == PF_IN) {
4600                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4601                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4602                                 key.ext.port = iih.icmp6_id;
4603                                 key.gwy.port = iih.icmp6_id;
4604                         } else {
4605                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4606                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4607                                 key.lan.port = iih.icmp6_id;
4608                                 key.ext.port = iih.icmp6_id;
4609                         }
4610
4611                         STATE_LOOKUP();
4612
4613                         if (STATE_TRANSLATE(*state)) {
4614                                 if (direction == PF_IN) {
4615                                         pf_change_icmp(pd2.src, &iih.icmp6_id,
4616                                             daddr, &(*state)->lan.addr,
4617                                             (*state)->lan.port, NULL,
4618                                             pd2.ip_sum, icmpsum,
4619                                             pd->ip_sum, 0, AF_INET6);
4620                                 } else {
4621                                         pf_change_icmp(pd2.dst, &iih.icmp6_id,
4622                                             saddr, &(*state)->gwy.addr,
4623                                             (*state)->gwy.port, NULL,
4624                                             pd2.ip_sum, icmpsum,
4625                                             pd->ip_sum, 0, AF_INET6);
4626                                 }
4627                                 m_copyback(m, off, sizeof(struct icmp6_hdr),
4628                                     (caddr_t)pd->hdr.icmp6);
4629                                 m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6);
4630                                 m_copyback(m, off2, sizeof(struct icmp6_hdr),
4631                                     (caddr_t)&iih);
4632                         }
4633
4634                         return (PF_PASS);
4635                         break;
4636                 }
4637 #endif /* INET6 */
4638                 default: {
4639                         struct pf_state         key;
4640
4641                         key.af = pd2.af;
4642                         key.proto = pd2.proto;
4643                         if (direction == PF_IN) {
4644                                 PF_ACPY(&key.ext.addr, pd2.dst, key.af);
4645                                 PF_ACPY(&key.gwy.addr, pd2.src, key.af);
4646                                 key.ext.port = 0;
4647                                 key.gwy.port = 0;
4648                         } else {
4649                                 PF_ACPY(&key.lan.addr, pd2.dst, key.af);
4650                                 PF_ACPY(&key.ext.addr, pd2.src, key.af);
4651                                 key.lan.port = 0;
4652                                 key.ext.port = 0;
4653                         }
4654
4655                         STATE_LOOKUP();
4656
4657                         if (STATE_TRANSLATE(*state)) {
4658                                 if (direction == PF_IN) {
4659                                         pf_change_icmp(pd2.src, NULL,
4660                                             daddr, &(*state)->lan.addr,
4661                                             0, NULL,
4662                                             pd2.ip_sum, icmpsum,
4663                                             pd->ip_sum, 0, pd2.af);
4664                                 } else {
4665                                         pf_change_icmp(pd2.dst, NULL,
4666                                             saddr, &(*state)->gwy.addr,
4667                                             0, NULL,
4668                                             pd2.ip_sum, icmpsum,
4669                                             pd->ip_sum, 0, pd2.af);
4670                                 }
4671                                 switch (pd2.af) {
4672 #ifdef INET
4673                                 case AF_INET:
4674                                         m_copyback(m, off, ICMP_MINLEN,
4675                                             (caddr_t)pd->hdr.icmp);
4676                                         m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2);
4677                                         break;
4678 #endif /* INET */
4679 #ifdef INET6
4680                                 case AF_INET6:
4681                                         m_copyback(m, off,
4682                                             sizeof(struct icmp6_hdr),
4683                                             (caddr_t)pd->hdr.icmp6);
4684                                         m_copyback(m, ipoff2, sizeof(h2_6),
4685                                             (caddr_t)&h2_6);
4686                                         break;
4687 #endif /* INET6 */
4688                                 }
4689                         }
4690
4691                         return (PF_PASS);
4692                         break;
4693                 }
4694                 }
4695         }
4696 }
4697
4698 int
4699 pf_test_state_other(struct pf_state **state, int direction, struct pfi_kif *kif,
4700     struct pf_pdesc *pd)
4701 {
4702         struct pf_state_peer    *src, *dst;
4703         struct pf_state          key;
4704
4705         key.af = pd->af;
4706         key.proto = pd->proto;
4707         if (direction == PF_IN) {
4708                 PF_ACPY(&key.ext.addr, pd->src, key.af);
4709                 PF_ACPY(&key.gwy.addr, pd->dst, key.af);
4710                 key.ext.port = 0;
4711                 key.gwy.port = 0;
4712         } else {
4713                 PF_ACPY(&key.lan.addr, pd->src, key.af);
4714                 PF_ACPY(&key.ext.addr, pd->dst, key.af);
4715                 key.lan.port = 0;
4716                 key.ext.port = 0;
4717         }
4718
4719         STATE_LOOKUP();
4720
4721         if (direction == (*state)->direction) {
4722                 src = &(*state)->src;
4723                 dst = &(*state)->dst;
4724         } else {
4725                 src = &(*state)->dst;
4726                 dst = &(*state)->src;
4727         }
4728
4729         /* update states */
4730         if (src->state < PFOTHERS_SINGLE)
4731                 src->state = PFOTHERS_SINGLE;
4732         if (dst->state == PFOTHERS_SINGLE)
4733                 dst->state = PFOTHERS_MULTIPLE;
4734
4735         /* update expire time */
4736         (*state)->expire = time_second;
4737         if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE)
4738                 (*state)->timeout = PFTM_OTHER_MULTIPLE;
4739         else
4740                 (*state)->timeout = PFTM_OTHER_SINGLE;
4741
4742         /* translate source/destination address, if necessary */
4743         if (STATE_TRANSLATE(*state)) {
4744                 if (direction == PF_OUT)
4745                         switch (pd->af) {
4746 #ifdef INET
4747                         case AF_INET:
4748                                 pf_change_a(&pd->src->v4.s_addr,
4749                                     pd->ip_sum, (*state)->gwy.addr.v4.s_addr,
4750                                     0);
4751                                 break;
4752 #endif /* INET */
4753 #ifdef INET6
4754                         case AF_INET6:
4755                                 PF_ACPY(pd->src, &(*state)->gwy.addr, pd->af);
4756                                 break;
4757 #endif /* INET6 */
4758                         }
4759                 else
4760                         switch (pd->af) {
4761 #ifdef INET
4762                         case AF_INET:
4763                                 pf_change_a(&pd->dst->v4.s_addr,
4764                                     pd->ip_sum, (*state)->lan.addr.v4.s_addr,
4765                                     0);
4766                                 break;
4767 #endif /* INET */
4768 #ifdef INET6
4769                         case AF_INET6:
4770                                 PF_ACPY(pd->dst, &(*state)->lan.addr, pd->af);
4771                                 break;
4772 #endif /* INET6 */
4773                         }
4774         }
4775
4776         return (PF_PASS);
4777 }
4778
4779 /*
4780  * ipoff and off are measured from the start of the mbuf chain.
4781  * h must be at "ipoff" on the mbuf chain.
4782  */
4783 void *
4784 pf_pull_hdr(struct mbuf *m, int off, void *p, int len,
4785     u_short *actionp, u_short *reasonp, sa_family_t af)
4786 {
4787         switch (af) {
4788 #ifdef INET
4789         case AF_INET: {
4790                 struct ip       *h = mtod(m, struct ip *);
4791                 u_int16_t        fragoff = (h->ip_off & IP_OFFMASK) << 3;
4792
4793                 if (fragoff) {
4794                         if (fragoff >= len)
4795                                 ACTION_SET(actionp, PF_PASS);
4796                         else {
4797                                 ACTION_SET(actionp, PF_DROP);
4798                                 REASON_SET(reasonp, PFRES_FRAG);
4799                         }
4800                         return (NULL);
4801                 }
4802                 if (m->m_pkthdr.len < off + len ||
4803                     h->ip_len < off + len) {
4804                         ACTION_SET(actionp, PF_DROP);
4805                         REASON_SET(reasonp, PFRES_SHORT);
4806                         return (NULL);
4807                 }
4808                 break;
4809         }
4810 #endif /* INET */
4811 #ifdef INET6
4812         case AF_INET6: {
4813                 struct ip6_hdr  *h = mtod(m, struct ip6_hdr *);
4814
4815                 if (m->m_pkthdr.len < off + len ||
4816                     (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) <
4817                     (unsigned)(off + len)) {
4818                         ACTION_SET(actionp, PF_DROP);
4819                         REASON_SET(reasonp, PFRES_SHORT);
4820                         return (NULL);
4821                 }
4822                 break;
4823         }
4824 #endif /* INET6 */
4825         }
4826         m_copydata(m, off, len, p);
4827         return (p);
4828 }
4829
4830 int
4831 pf_routable(struct pf_addr *addr, sa_family_t af)
4832 {
4833         struct sockaddr_in      *dst;
4834         struct route             ro;
4835         int                      ret = 0;
4836
4837         bzero(&ro, sizeof(ro));
4838         dst = satosin(&ro.ro_dst);
4839         dst->sin_family = af;
4840         dst->sin_len = sizeof(*dst);
4841         dst->sin_addr = addr->v4;
4842         rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING));
4843
4844         if (ro.ro_rt != NULL) {
4845                 ret = 1;
4846                 RTFREE(ro.ro_rt);
4847         }
4848
4849         return (ret);
4850 }
4851
4852 #ifdef INET
4853 void
4854 pf_route(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
4855     struct pf_state *s)
4856 {
4857         struct mbuf             *m0, *m1;
4858         struct route             iproute;
4859         struct route            *ro = NULL;
4860         struct sockaddr_in      *dst;
4861         struct ip               *ip;
4862         struct ifnet            *ifp = NULL;
4863         struct pf_addr           naddr;
4864         struct pf_src_node      *sn = NULL;
4865         int                      error = 0;
4866         int sw_csum;
4867
4868         if (m == NULL || *m == NULL || r == NULL ||
4869             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
4870                 panic("pf_route: invalid parameters");
4871
4872         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
4873                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
4874                 (*m)->m_pkthdr.pf_routed = 1;
4875         } else {
4876                 if ((*m)->m_pkthdr.pf_routed > 3) {
4877                         m0 = *m;
4878                         *m = NULL;
4879                         goto bad;
4880                 }
4881                 (*m)->m_pkthdr.pf_routed++;
4882         }
4883
4884         if (r->rt == PF_DUPTO) {
4885                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
4886                         return;
4887         } else {
4888                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
4889                         return;
4890                 m0 = *m;
4891         }
4892
4893         if (m0->m_len < sizeof(struct ip))
4894                 panic("pf_route: m0->m_len < sizeof(struct ip)");
4895         ip = mtod(m0, struct ip *);
4896
4897         ro = &iproute;
4898         bzero((caddr_t)ro, sizeof(*ro));
4899         dst = satosin(&ro->ro_dst);
4900         dst->sin_family = AF_INET;
4901         dst->sin_len = sizeof(*dst);
4902         dst->sin_addr = ip->ip_dst;
4903
4904         if (r->rt == PF_FASTROUTE) {
4905                 rtalloc(ro);
4906                 if (ro->ro_rt == 0) {
4907                         ipstat.ips_noroute++;
4908                         goto bad;
4909                 }
4910
4911                 ifp = ro->ro_rt->rt_ifp;
4912                 ro->ro_rt->rt_use++;
4913
4914                 if (ro->ro_rt->rt_flags & RTF_GATEWAY)
4915                         dst = satosin(ro->ro_rt->rt_gateway);
4916         } else {
4917                 if (TAILQ_EMPTY(&r->rpool.list))
4918                         panic("pf_route: TAILQ_EMPTY(&r->rpool.list)");
4919                 if (s == NULL) {
4920                         pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src,
4921                             &naddr, NULL, &sn);
4922                         if (!PF_AZERO(&naddr, AF_INET))
4923                                 dst->sin_addr.s_addr = naddr.v4.s_addr;
4924                         ifp = r->rpool.cur->kif ?
4925                             r->rpool.cur->kif->pfik_ifp : NULL;
4926                 } else {
4927                         if (!PF_AZERO(&s->rt_addr, AF_INET))
4928                                 dst->sin_addr.s_addr =
4929                                     s->rt_addr.v4.s_addr;
4930                         ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
4931                 }
4932         }
4933         if (ifp == NULL)
4934                 goto bad;
4935
4936         if (oifp != ifp) {
4937                 if (pf_test(PF_OUT, ifp, &m0) != PF_PASS)
4938                         goto bad;
4939                 else if (m0 == NULL)
4940                         goto done;
4941                 if (m0->m_len < sizeof(struct ip))
4942                         panic("pf_route: m0->m_len < sizeof(struct ip)");
4943                 ip = mtod(m0, struct ip *);
4944         }
4945
4946         /* Copied from ip_output. */
4947         m0->m_pkthdr.csum_flags |= CSUM_IP;
4948         sw_csum = m0->m_pkthdr.csum_flags & ~ifp->if_hwassist;
4949         if (sw_csum & CSUM_DELAY_DATA) {
4950                 in_delayed_cksum(m0);
4951                 sw_csum &= ~CSUM_DELAY_DATA;
4952         }
4953         m0->m_pkthdr.csum_flags &= ifp->if_hwassist;
4954
4955         /*
4956          * If small enough for interface, or the interface will take
4957          * care of the fragmentation for us, can just send directly.
4958          */
4959         if (ip->ip_len <= ifp->if_mtu || ((ifp->if_hwassist & CSUM_FRAGMENT) &&
4960                 (ip->ip_off & IP_DF) == 0)) {
4961                 ip->ip_len = htons(ip->ip_len);
4962                 ip->ip_off = htons(ip->ip_off);
4963                 ip->ip_sum = 0;
4964                 if (sw_csum & CSUM_DELAY_IP) {
4965                         /* From KAME */
4966                         if (ip->ip_v == IPVERSION &&
4967                             (ip->ip_hl << 2) == sizeof(*ip)) {
4968                                 ip->ip_sum = in_cksum_hdr(ip);
4969                         } else {
4970                                 ip->ip_sum = in_cksum(m0, ip->ip_hl << 2);
4971                         }
4972                 }
4973
4974                 lwkt_serialize_enter(ifp->if_serializer);
4975                 error = (*ifp->if_output)(ifp, m0, sintosa(dst), ro->ro_rt);
4976                 lwkt_serialize_exit(ifp->if_serializer);
4977                 goto done;
4978         }
4979
4980         /*
4981          * Too large for interface; fragment if possible.
4982          * Must be able to put at least 8 bytes per fragment.
4983          */
4984         if (ip->ip_off & IP_DF) {
4985                 ipstat.ips_cantfrag++;
4986                 if (r->rt != PF_DUPTO) {
4987                         icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0,
4988                             ifp->if_mtu);
4989                         goto done;
4990                 } else
4991                         goto bad;
4992         }
4993
4994         m1 = m0;
4995         error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist, sw_csum);
4996         if (error)
4997                 goto bad;
4998
4999         for (m0 = m1; m0; m0 = m1) {
5000                 m1 = m0->m_nextpkt;
5001                 m0->m_nextpkt = 0;
5002                 if (error == 0) {
5003                         lwkt_serialize_enter(ifp->if_serializer);
5004                         error = (*ifp->if_output)(ifp, m0, sintosa(dst),
5005                                     NULL);
5006                         lwkt_serialize_exit(ifp->if_serializer);
5007                 } else {
5008                         m_freem(m0);
5009                 }
5010         }
5011
5012         if (error == 0)
5013                 ipstat.ips_fragmented++;
5014
5015 done:
5016         if (r->rt != PF_DUPTO)
5017                 *m = NULL;
5018         if (ro == &iproute && ro->ro_rt)
5019                 RTFREE(ro->ro_rt);
5020         return;
5021
5022 bad:
5023         m_freem(m0);
5024         goto done;
5025 }
5026 #endif /* INET */
5027
5028 #ifdef INET6
5029 void
5030 pf_route6(struct mbuf **m, struct pf_rule *r, int dir, struct ifnet *oifp,
5031     struct pf_state *s)
5032 {
5033         struct mbuf             *m0;
5034         struct route_in6         ip6route;
5035         struct route_in6        *ro;
5036         struct sockaddr_in6     *dst;
5037         struct ip6_hdr          *ip6;
5038         struct ifnet            *ifp = NULL;
5039         struct pf_addr           naddr;
5040         struct pf_src_node      *sn = NULL;
5041         int                      error = 0;
5042
5043         if (m == NULL || *m == NULL || r == NULL ||
5044             (dir != PF_IN && dir != PF_OUT) || oifp == NULL)
5045                 panic("pf_route6: invalid parameters");
5046
5047         if (((*m)->m_pkthdr.fw_flags & PF_MBUF_ROUTED) == 0) {
5048                 (*m)->m_pkthdr.fw_flags |= PF_MBUF_ROUTED;
5049                 (*m)->m_pkthdr.pf_routed = 1;
5050         } else {
5051                 if ((*m)->m_pkthdr.pf_routed > 3) {
5052                         m0 = *m;
5053                         *m = NULL;
5054                         goto bad;
5055                 }
5056                 (*m)->m_pkthdr.pf_routed++;
5057         }
5058
5059         if (r->rt == PF_DUPTO) {
5060                 if ((m0 = m_dup(*m, MB_DONTWAIT)) == NULL)
5061                         return;
5062         } else {
5063                 if ((r->rt == PF_REPLYTO) == (r->direction == dir))
5064                         return;
5065                 m0 = *m;
5066         }
5067
5068         if (m0->m_len < sizeof(struct ip6_hdr))
5069                 panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5070         ip6 = mtod(m0, struct ip6_hdr *);
5071
5072         ro = &ip6route;
5073         bzero((caddr_t)ro, sizeof(*ro));
5074         dst = (struct sockaddr_in6 *)&ro->ro_dst;
5075         dst->sin6_family = AF_INET6;
5076         dst->sin6_len = sizeof(*dst);
5077         dst->sin6_addr = ip6->ip6_dst;
5078
5079         /* Cheat. */
5080         if (r->rt == PF_FASTROUTE) {
5081                 m0->m_pkthdr.fw_flags |= PF_MBUF_GENERATED;
5082                 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
5083                 return;
5084         }
5085
5086         if (TAILQ_EMPTY(&r->rpool.list))
5087                 panic("pf_route6: TAILQ_EMPTY(&r->rpool.list)");
5088         if (s == NULL) {
5089                 pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src,
5090                     &naddr, NULL, &sn);
5091                 if (!PF_AZERO(&naddr, AF_INET6))
5092                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5093                             &naddr, AF_INET6);
5094                 ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL;
5095         } else {
5096                 if (!PF_AZERO(&s->rt_addr, AF_INET6))
5097                         PF_ACPY((struct pf_addr *)&dst->sin6_addr,
5098                             &s->rt_addr, AF_INET6);
5099                 ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL;
5100         }
5101         if (ifp == NULL)
5102                 goto bad;
5103
5104         if (oifp != ifp) {
5105                 if (pf_test6(PF_OUT, ifp, &m0) != PF_PASS)
5106                         goto bad;
5107                 else if (m0 == NULL)
5108                         goto done;
5109                 if (m0->m_len < sizeof(struct ip6_hdr))
5110                         panic("pf_route6: m0->m_len < sizeof(struct ip6_hdr)");
5111                 ip6 = mtod(m0, struct ip6_hdr *);
5112         }
5113
5114         /*
5115          * If the packet is too large for the outgoing interface,
5116          * send back an icmp6 error.
5117          */
5118         if (IN6_IS_ADDR_LINKLOCAL(&dst->sin6_addr))
5119                 dst->sin6_addr.s6_addr16[1] = htons(ifp->if_index);
5120         if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) {
5121                 error = nd6_output(ifp, ifp, m0, dst, NULL);
5122         } else {
5123                 in6_ifstat_inc(ifp, ifs6_in_toobig);
5124                 if (r->rt != PF_DUPTO)
5125                         icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
5126                 else
5127                         goto bad;
5128         }
5129
5130 done:
5131         if (r->rt != PF_DUPTO)
5132                 *m = NULL;
5133         return;
5134
5135 bad:
5136         m_freem(m0);
5137         goto done;
5138 }
5139 #endif /* INET6 */
5140
5141
5142 /*
5143  * check protocol (tcp/udp/icmp/icmp6) checksum and set mbuf flag
5144  *   off is the offset where the protocol header starts
5145  *   len is the total length of protocol header plus payload
5146  * returns 0 when the checksum is valid, otherwise returns 1.
5147  */
5148 /*
5149  * XXX
5150  * FreeBSD supports cksum offload for the following drivers.
5151  * em(4), gx(4), lge(4), nge(4), ti(4), xl(4)
5152  * If we can make full use of it we would outperform ipfw/ipfilter in
5153  * very heavy traffic.
5154  * I have not tested 'cause I don't have NICs that supports cksum offload.
5155  * (There might be problems. Typical phenomena would be
5156  *   1. No route message for UDP packet.
5157  *   2. No connection acceptance from external hosts regardless of rule set.)
5158  */
5159 int
5160 pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p,
5161     sa_family_t af)
5162 {
5163         u_int16_t sum = 0;
5164         int hw_assist = 0;
5165         struct ip *ip;
5166
5167         if (off < sizeof(struct ip) || len < sizeof(struct udphdr))
5168                 return (1);
5169         if (m->m_pkthdr.len < off + len)
5170                 return (1);
5171
5172         switch (p) {
5173         case IPPROTO_TCP:
5174         case IPPROTO_UDP:
5175                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
5176                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) {
5177                                 sum = m->m_pkthdr.csum_data;
5178                         } else {
5179                                 ip = mtod(m, struct ip *);
5180                                 sum = in_pseudo(ip->ip_src.s_addr,
5181                                         ip->ip_dst.s_addr, htonl((u_short)len +
5182                                         m->m_pkthdr.csum_data + p));
5183                         }
5184                         sum ^= 0xffff;
5185                         ++hw_assist;
5186                 }
5187                 break;
5188         case IPPROTO_ICMP:
5189 #ifdef INET6
5190         case IPPROTO_ICMPV6:
5191 #endif /* INET6 */
5192                 break;
5193         default:
5194                 return (1);
5195         }
5196
5197         if (!hw_assist) {
5198                 switch (af) {
5199                 case AF_INET:
5200                         if (p == IPPROTO_ICMP) {
5201                                 if (m->m_len < off)
5202                                         return (1);
5203                                 m->m_data += off;
5204                                 m->m_len -= off;
5205                                 sum = in_cksum(m, len);
5206                                 m->m_data -= off;
5207                                 m->m_len += off;
5208                         } else {
5209                                 if (m->m_len < sizeof(struct ip))
5210                                         return (1);
5211                                 sum = in_cksum_range(m, p, off, len);
5212                                 if (sum == 0) {
5213                                         m->m_pkthdr.csum_flags |=
5214                                             (CSUM_DATA_VALID |
5215                                              CSUM_PSEUDO_HDR);
5216                                         m->m_pkthdr.csum_data = 0xffff;
5217                                 }
5218                         }
5219                         break;
5220 #ifdef INET6
5221                 case AF_INET6:
5222                         if (m->m_len < sizeof(struct ip6_hdr))
5223                                 return (1);
5224                         sum = in6_cksum(m, p, off, len);
5225                         /*
5226                          * XXX
5227                          * IPv6 H/W cksum off-load not supported yet!
5228                          *
5229                          * if (sum == 0) {
5230                          *      m->m_pkthdr.csum_flags |=
5231                          *          (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
5232                          *      m->m_pkthdr.csum_data = 0xffff;
5233                          *}
5234                          */
5235                         break;
5236 #endif /* INET6 */
5237                 default:
5238                         return (1);
5239                 }
5240         }
5241         if (sum) {
5242                 switch (p) {
5243                 case IPPROTO_TCP:
5244                         tcpstat.tcps_rcvbadsum++;
5245                         break;
5246                 case IPPROTO_UDP:
5247                         udpstat.udps_badsum++;
5248                         break;
5249                 case IPPROTO_ICMP:
5250                         icmpstat.icps_checksum++;
5251                         break;
5252 #ifdef INET6
5253                 case IPPROTO_ICMPV6:
5254                         icmp6stat.icp6s_checksum++;
5255                         break;
5256 #endif /* INET6 */
5257                 }
5258                 return (1);
5259         }
5260         return (0);
5261 }
5262
5263 #ifdef INET
5264 int
5265 pf_test(int dir, struct ifnet *ifp, struct mbuf **m0)
5266 {
5267         struct pfi_kif          *kif;
5268         u_short                  action, reason = 0, log = 0;
5269         struct mbuf             *m = *m0;
5270         struct ip               *h = NULL;
5271         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5272         struct pf_state         *s = NULL;
5273         struct pf_ruleset       *ruleset = NULL;
5274         struct pf_pdesc          pd;
5275         int                      off, dirndx, pqid = 0;
5276
5277         if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5278                 return (PF_PASS);
5279
5280         kif = pfi_index2kif[ifp->if_index];
5281         if (kif == NULL)
5282                 return (PF_DROP);
5283
5284 #ifdef DIAGNOSTIC
5285         if ((m->m_flags & M_PKTHDR) == 0)
5286                 panic("non-M_PKTHDR is passed to pf_test");
5287 #endif
5288
5289         memset(&pd, 0, sizeof(pd));
5290         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5291                 action = PF_DROP;
5292                 REASON_SET(&reason, PFRES_SHORT);
5293                 log = 1;
5294                 goto done;
5295         }
5296
5297         /* We do IP header normalization and packet reassembly here */
5298         if (pf_normalize_ip(m0, dir, kif, &reason) != PF_PASS) {
5299                 action = PF_DROP;
5300                 goto done;
5301         }
5302         m = *m0;
5303         h = mtod(m, struct ip *);
5304
5305         off = h->ip_hl << 2;
5306         if (off < (int)sizeof(*h)) {
5307                 action = PF_DROP;
5308                 REASON_SET(&reason, PFRES_SHORT);
5309                 log = 1;
5310                 goto done;
5311         }
5312
5313         pd.src = (struct pf_addr *)&h->ip_src;
5314         pd.dst = (struct pf_addr *)&h->ip_dst;
5315         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET);
5316         pd.ip_sum = &h->ip_sum;
5317         pd.proto = h->ip_p;
5318         pd.af = AF_INET;
5319         pd.tos = h->ip_tos;
5320         pd.tot_len = h->ip_len;
5321
5322         /* handle fragments that didn't get reassembled by normalization */
5323         if (h->ip_off & (IP_MF | IP_OFFMASK)) {
5324                 action = pf_test_fragment(&r, dir, kif, m, h,
5325                     &pd, &a, &ruleset);
5326                 goto done;
5327         }
5328
5329         switch (h->ip_p) {
5330
5331         case IPPROTO_TCP: {
5332                 struct tcphdr   th;
5333
5334                 pd.hdr.tcp = &th;
5335                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5336                     &action, &reason, AF_INET)) {
5337                         log = action != PF_PASS;
5338                         goto done;
5339                 }
5340                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5341                     h->ip_len - off, IPPROTO_TCP, AF_INET)) {
5342                         action = PF_DROP;
5343                         goto done;
5344                 }
5345                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5346                 if ((th.th_flags & TH_ACK) && pd.p_len == 0)
5347                         pqid = 1;
5348                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5349                 if (action == PF_DROP)
5350                         goto done;
5351                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5352                     &reason);
5353                 if (action == PF_PASS) {
5354 #if NPFSYNC
5355                         pfsync_update_state(s);
5356 #endif
5357                         r = s->rule.ptr;
5358                         a = s->anchor.ptr;
5359                         log = s->log;
5360                 } else if (s == NULL)
5361                         action = pf_test_tcp(&r, &s, dir, kif,
5362                             m, off, h, &pd, &a, &ruleset);
5363                 break;
5364         }
5365
5366         case IPPROTO_UDP: {
5367                 struct udphdr   uh;
5368
5369                 pd.hdr.udp = &uh;
5370                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5371                     &action, &reason, AF_INET)) {
5372                         log = action != PF_PASS;
5373                         goto done;
5374                 }
5375                 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5376                     off, h->ip_len - off, IPPROTO_UDP, AF_INET)) {
5377                         action = PF_DROP;
5378                         goto done;
5379                 }
5380                 if (uh.uh_dport == 0 ||
5381                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5382                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5383                         action = PF_DROP;
5384                         goto done;
5385                 }
5386                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5387                 if (action == PF_PASS) {
5388 #if NPFSYNC
5389                         pfsync_update_state(s);
5390 #endif
5391                         r = s->rule.ptr;
5392                         a = s->anchor.ptr;
5393                         log = s->log;
5394                 } else if (s == NULL)
5395                         action = pf_test_udp(&r, &s, dir, kif,
5396                             m, off, h, &pd, &a, &ruleset);
5397                 break;
5398         }
5399
5400         case IPPROTO_ICMP: {
5401                 struct icmp     ih;
5402
5403                 pd.hdr.icmp = &ih;
5404                 if (!pf_pull_hdr(m, off, &ih, ICMP_MINLEN,
5405                     &action, &reason, AF_INET)) {
5406                         log = action != PF_PASS;
5407                         goto done;
5408                 }
5409                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5410                     h->ip_len - off, IPPROTO_ICMP, AF_INET)) {
5411                         action = PF_DROP;
5412                         goto done;
5413                 }
5414                 action = pf_test_state_icmp(&s, dir, kif, m, off, h, &pd);
5415                 if (action == PF_PASS) {
5416 #if NPFSYNC
5417                         pfsync_update_state(s);
5418 #endif
5419                         r = s->rule.ptr;
5420                         a = s->anchor.ptr;
5421                         log = s->log;
5422                 } else if (s == NULL)
5423                         action = pf_test_icmp(&r, &s, dir, kif,
5424                             m, off, h, &pd, &a, &ruleset);
5425                 break;
5426         }
5427
5428         default:
5429                 action = pf_test_state_other(&s, dir, kif, &pd);
5430                 if (action == PF_PASS) {
5431 #if NPFSYNC
5432                         pfsync_update_state(s);
5433 #endif
5434                         r = s->rule.ptr;
5435                         a = s->anchor.ptr;
5436                         log = s->log;
5437                 } else if (s == NULL)
5438                         action = pf_test_other(&r, &s, dir, kif, m, off, h,
5439                             &pd, &a, &ruleset);
5440                 break;
5441         }
5442
5443 done:
5444         if (action == PF_PASS && h->ip_hl > 5 &&
5445             !((s && s->allow_opts) || r->allow_opts)) {
5446                 action = PF_DROP;
5447                 REASON_SET(&reason, PFRES_SHORT);
5448                 log = 1;
5449                 DPFPRINTF(PF_DEBUG_MISC,
5450                     ("pf: dropping packet with ip options\n"));
5451         }
5452
5453 #ifdef ALTQ
5454         if (action == PF_PASS && r->qid) {
5455                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5456                 if (pd.tos == IPTOS_LOWDELAY)
5457                         m->m_pkthdr.altq_qid = r->pqid;
5458                 else
5459                         m->m_pkthdr.altq_qid = r->qid;
5460                 m->m_pkthdr.ecn_af = AF_INET;
5461                 m->m_pkthdr.header = h;
5462         }
5463 #endif
5464
5465         /*
5466          * connections redirected to loopback should not match sockets
5467          * bound specifically to loopback due to security implications,
5468          * see tcp_input() and in_pcblookup_listen().
5469          */
5470         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5471             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5472             (s->nat_rule.ptr->action == PF_RDR ||
5473             s->nat_rule.ptr->action == PF_BINAT) &&
5474             (ntohl(pd.dst->v4.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) {
5475                 action = PF_DROP;
5476                 REASON_SET(&reason, PFRES_MEMORY);
5477         }
5478
5479         m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5480
5481         if (log)
5482                 PFLOG_PACKET(kif, h, m, AF_INET, dir, reason, r, a, ruleset);
5483
5484         kif->pfik_bytes[0][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5485         kif->pfik_packets[0][dir == PF_OUT][action != PF_PASS]++;
5486
5487         if (action == PF_PASS || r->action == PF_DROP) {
5488                 r->packets++;
5489                 r->bytes += pd.tot_len;
5490                 if (a != NULL) {
5491                         a->packets++;
5492                         a->bytes += pd.tot_len;
5493                 }
5494                 if (s != NULL) {
5495                         dirndx = (dir == s->direction) ? 0 : 1;
5496                         s->packets[dirndx]++;
5497                         s->bytes[dirndx] += pd.tot_len;
5498                         if (s->nat_rule.ptr != NULL) {
5499                                 s->nat_rule.ptr->packets++;
5500                                 s->nat_rule.ptr->bytes += pd.tot_len;
5501                         }
5502                         if (s->src_node != NULL) {
5503                                 s->src_node->packets++;
5504                                 s->src_node->bytes += pd.tot_len;
5505                         }
5506                         if (s->nat_src_node != NULL) {
5507                                 s->nat_src_node->packets++;
5508                                 s->nat_src_node->bytes += pd.tot_len;
5509                         }
5510                 }
5511                 tr = r;
5512                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5513                 if (nr != NULL) {
5514                         struct pf_addr *x;
5515                         /*
5516                          * XXX: we need to make sure that the addresses
5517                          * passed to pfr_update_stats() are the same than
5518                          * the addresses used during matching (pfr_match)
5519                          */
5520                         if (r == &pf_default_rule) {
5521                                 tr = nr;
5522                                 x = (s == NULL || s->direction == dir) ?
5523                                     &pd.baddr : &pd.naddr;
5524                         } else
5525                                 x = (s == NULL || s->direction == dir) ?
5526                                     &pd.naddr : &pd.baddr;
5527                         if (x == &pd.baddr || s == NULL) {
5528                                 /* we need to change the address */
5529                                 if (dir == PF_OUT)
5530                                         pd.src = x;
5531                                 else
5532                                         pd.dst = x;
5533                         }
5534                 }
5535                 if (tr->src.addr.type == PF_ADDR_TABLE)
5536                         pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5537                             s->direction == dir) ? pd.src : pd.dst, pd.af,
5538                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5539                             tr->src.not);
5540                 if (tr->dst.addr.type == PF_ADDR_TABLE)
5541                         pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5542                             s->direction == dir) ? pd.dst : pd.src, pd.af,
5543                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5544                             tr->dst.not);
5545         }
5546
5547
5548         if (action == PF_SYNPROXY_DROP) {
5549                 m_freem(*m0);
5550                 *m0 = NULL;
5551                 action = PF_PASS;
5552         } else if (r->rt)
5553                 /* pf_route can free the mbuf causing *m0 to become NULL */
5554                 pf_route(m0, r, dir, ifp, s);
5555
5556         return (action);
5557 }
5558 #endif /* INET */
5559
5560 #ifdef INET6
5561 int
5562 pf_test6(int dir, struct ifnet *ifp, struct mbuf **m0)
5563 {
5564         struct pfi_kif          *kif;
5565         u_short                  action, reason = 0, log = 0;
5566         struct mbuf             *m = *m0;
5567         struct ip6_hdr          *h = NULL;
5568         struct pf_rule          *a = NULL, *r = &pf_default_rule, *tr, *nr;
5569         struct pf_state         *s = NULL;
5570         struct pf_ruleset       *ruleset = NULL;
5571         struct pf_pdesc          pd;
5572         int                      off, terminal = 0, dirndx;
5573
5574         if (!pf_status.running || (m->m_pkthdr.fw_flags & PF_MBUF_GENERATED))
5575                 return (PF_PASS);
5576
5577         kif = pfi_index2kif[ifp->if_index];
5578         if (kif == NULL)
5579                 return (PF_DROP);
5580
5581 #ifdef DIAGNOSTIC
5582         if ((m->m_flags & M_PKTHDR) == 0)
5583                 panic("non-M_PKTHDR is passed to pf_test");
5584 #endif
5585
5586         memset(&pd, 0, sizeof(pd));
5587         if (m->m_pkthdr.len < (int)sizeof(*h)) {
5588                 action = PF_DROP;
5589                 REASON_SET(&reason, PFRES_SHORT);
5590                 log = 1;
5591                 goto done;
5592         }
5593
5594         /* We do IP header normalization and packet reassembly here */
5595         if (pf_normalize_ip6(m0, dir, kif, &reason) != PF_PASS) {
5596                 action = PF_DROP;
5597                 goto done;
5598         }
5599         m = *m0;
5600         h = mtod(m, struct ip6_hdr *);
5601
5602         pd.src = (struct pf_addr *)&h->ip6_src;
5603         pd.dst = (struct pf_addr *)&h->ip6_dst;
5604         PF_ACPY(&pd.baddr, dir == PF_OUT ? pd.src : pd.dst, AF_INET6);
5605         pd.ip_sum = NULL;
5606         pd.af = AF_INET6;
5607         pd.tos = 0;
5608         pd.tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr);
5609
5610         off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr);
5611         pd.proto = h->ip6_nxt;
5612         do {
5613                 switch (pd.proto) {
5614                 case IPPROTO_FRAGMENT:
5615                         action = pf_test_fragment(&r, dir, kif, m, h,
5616                             &pd, &a, &ruleset);
5617                         if (action == PF_DROP)
5618                                 REASON_SET(&reason, PFRES_FRAG);
5619                         goto done;
5620                 case IPPROTO_AH:
5621                 case IPPROTO_HOPOPTS:
5622                 case IPPROTO_ROUTING:
5623                 case IPPROTO_DSTOPTS: {
5624                         /* get next header and header length */
5625                         struct ip6_ext  opt6;
5626
5627                         if (!pf_pull_hdr(m, off, &opt6, sizeof(opt6),
5628                             NULL, NULL, pd.af)) {
5629                                 DPFPRINTF(PF_DEBUG_MISC,
5630                                     ("pf: IPv6 short opt\n"));
5631                                 action = PF_DROP;
5632                                 REASON_SET(&reason, PFRES_SHORT);
5633                                 log = 1;
5634                                 goto done;
5635                         }
5636                         if (pd.proto == IPPROTO_AH)
5637                                 off += (opt6.ip6e_len + 2) * 4;
5638                         else
5639                                 off += (opt6.ip6e_len + 1) * 8;
5640                         pd.proto = opt6.ip6e_nxt;
5641                         /* goto the next header */
5642                         break;
5643                 }
5644                 default:
5645                         terminal++;
5646                         break;
5647                 }
5648         } while (!terminal);
5649
5650         switch (pd.proto) {
5651
5652         case IPPROTO_TCP: {
5653                 struct tcphdr   th;
5654
5655                 pd.hdr.tcp = &th;
5656                 if (!pf_pull_hdr(m, off, &th, sizeof(th),
5657                     &action, &reason, AF_INET6)) {
5658                         log = action != PF_PASS;
5659                         goto done;
5660                 }
5661                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5662                     ntohs(h->ip6_plen), IPPROTO_TCP, AF_INET6)) {
5663                         action = PF_DROP;
5664                         goto done;
5665                 }
5666                 pd.p_len = pd.tot_len - off - (th.th_off << 2);
5667                 action = pf_normalize_tcp(dir, kif, m, 0, off, h, &pd);
5668                 if (action == PF_DROP)
5669                         goto done;
5670                 action = pf_test_state_tcp(&s, dir, kif, m, off, h, &pd,
5671                     &reason);
5672                 if (action == PF_PASS) {
5673 #if NPFSYNC
5674                         pfsync_update_state(s);
5675 #endif
5676                         r = s->rule.ptr;
5677                         a = s->anchor.ptr;
5678                         log = s->log;
5679                 } else if (s == NULL)
5680                         action = pf_test_tcp(&r, &s, dir, kif,
5681                             m, off, h, &pd, &a, &ruleset);
5682                 break;
5683         }
5684
5685         case IPPROTO_UDP: {
5686                 struct udphdr   uh;
5687
5688                 pd.hdr.udp = &uh;
5689                 if (!pf_pull_hdr(m, off, &uh, sizeof(uh),
5690                     &action, &reason, AF_INET6)) {
5691                         log = action != PF_PASS;
5692                         goto done;
5693                 }
5694                 if (dir == PF_IN && uh.uh_sum && pf_check_proto_cksum(m,
5695                     off, ntohs(h->ip6_plen), IPPROTO_UDP, AF_INET6)) {
5696                         action = PF_DROP;
5697                         goto done;
5698                 }
5699                 if (uh.uh_dport == 0 ||
5700                     ntohs(uh.uh_ulen) > m->m_pkthdr.len - off ||
5701                     ntohs(uh.uh_ulen) < sizeof(struct udphdr)) {
5702                         action = PF_DROP;
5703                         goto done;
5704                 }
5705                 action = pf_test_state_udp(&s, dir, kif, m, off, h, &pd);
5706                 if (action == PF_PASS) {
5707 #if NPFSYNC
5708                         pfsync_update_state(s);
5709 #endif
5710                         r = s->rule.ptr;
5711                         a = s->anchor.ptr;
5712                         log = s->log;
5713                 } else if (s == NULL)
5714                         action = pf_test_udp(&r, &s, dir, kif,
5715                             m, off, h, &pd, &a, &ruleset);
5716                 break;
5717         }
5718
5719         case IPPROTO_ICMPV6: {
5720                 struct icmp6_hdr        ih;
5721
5722                 pd.hdr.icmp6 = &ih;
5723                 if (!pf_pull_hdr(m, off, &ih, sizeof(ih),
5724                     &action, &reason, AF_INET6)) {
5725                         log = action != PF_PASS;
5726                         goto done;
5727                 }
5728                 if (dir == PF_IN && pf_check_proto_cksum(m, off,
5729                     ntohs(h->ip6_plen), IPPROTO_ICMPV6, AF_INET6)) {
5730                         action = PF_DROP;
5731                         goto done;
5732                 }
5733                 action = pf_test_state_icmp(&s, dir, kif,
5734                     m, off, h, &pd);
5735                 if (action == PF_PASS) {
5736 #if NPFSYNC
5737                         pfsync_update_state(s);
5738 #endif
5739                         r = s->rule.ptr;
5740                         a = s->anchor.ptr;
5741                         log = s->log;
5742                 } else if (s == NULL)
5743                         action = pf_test_icmp(&r, &s, dir, kif,
5744                             m, off, h, &pd, &a, &ruleset);
5745                 break;
5746         }
5747
5748         default:
5749                 action = pf_test_state_other(&s, dir, kif, &pd);
5750                 if (action == PF_PASS) {
5751                         r = s->rule.ptr;
5752                         a = s->anchor.ptr;
5753                         log = s->log;
5754                 } else if (s == NULL)
5755                         action = pf_test_other(&r, &s, dir, kif, m, off, h,
5756                             &pd, &a, &ruleset);
5757                 break;
5758         }
5759
5760 done:
5761         /* XXX handle IPv6 options, if not allowed. not implemented. */
5762
5763 #ifdef ALTQ
5764         if (action == PF_PASS && r->qid) {
5765                 m->m_pkthdr.fw_flags |= ALTQ_MBUF_TAGGED;
5766                 if (pd.tos == IPTOS_LOWDELAY)
5767                         m->m_pkthdr.altq_qid = r->pqid;
5768                 else
5769                         m->m_pkthdr.altq_qid = r->qid;
5770                 m->m_pkthdr.ecn_af = AF_INET6;
5771                 m->m_pkthdr.header = h;
5772         }
5773 #endif
5774
5775         if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP ||
5776             pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL &&
5777             (s->nat_rule.ptr->action == PF_RDR ||
5778             s->nat_rule.ptr->action == PF_BINAT) &&
5779             IN6_IS_ADDR_LOOPBACK(&pd.dst->v6)) {
5780                 action = PF_DROP;
5781                 REASON_SET(&reason, PFRES_MEMORY);
5782         }
5783
5784         m->m_pkthdr.fw_flags |= PF_MBUF_TRANSLATE_LOCALHOST;
5785
5786         if (log)
5787                 PFLOG_PACKET(kif, h, m, AF_INET6, dir, reason, r, a, ruleset);
5788
5789         kif->pfik_bytes[1][dir == PF_OUT][action != PF_PASS] += pd.tot_len;
5790         kif->pfik_packets[1][dir == PF_OUT][action != PF_PASS]++;
5791
5792         if (action == PF_PASS || r->action == PF_DROP) {
5793                 r->packets++;
5794                 r->bytes += pd.tot_len;
5795                 if (a != NULL) {
5796                         a->packets++;
5797                         a->bytes += pd.tot_len;
5798                 }
5799                 if (s != NULL) {
5800                         dirndx = (dir == s->direction) ? 0 : 1;
5801                         s->packets[dirndx]++;
5802                         s->bytes[dirndx] += pd.tot_len;
5803                         if (s->nat_rule.ptr != NULL) {
5804                                 s->nat_rule.ptr->packets++;
5805                                 s->nat_rule.ptr->bytes += pd.tot_len;
5806                         }
5807                         if (s->src_node != NULL) {
5808                                 s->src_node->packets++;
5809                                 s->src_node->bytes += pd.tot_len;
5810                         }
5811                         if (s->nat_src_node != NULL) {
5812                                 s->nat_src_node->packets++;
5813                                 s->nat_src_node->bytes += pd.tot_len;
5814                         }
5815                 }
5816                 tr = r;
5817                 nr = (s != NULL) ? s->nat_rule.ptr : pd.nat_rule;
5818                 if (nr != NULL) {
5819                         struct pf_addr *x;
5820                         /*
5821                          * XXX: we need to make sure that the addresses
5822                          * passed to pfr_update_stats() are the same than
5823                          * the addresses used during matching (pfr_match)
5824                          */
5825                         if (r == &pf_default_rule) {
5826                                 tr = nr;
5827                                 x = (s == NULL || s->direction == dir) ?
5828                                     &pd.baddr : &pd.naddr;
5829                         } else {
5830                                 x = (s == NULL || s->direction == dir) ?
5831                                     &pd.naddr : &pd.baddr;
5832                         }
5833                         if (x == &pd.baddr || s == NULL) {
5834                                 if (dir == PF_OUT)
5835                                         pd.src = x;
5836                                 else
5837                                         pd.dst = x;
5838                         }
5839                 }
5840                 if (tr->src.addr.type == PF_ADDR_TABLE)
5841                         pfr_update_stats(tr->src.addr.p.tbl, (s == NULL ||
5842                             s->direction == dir) ? pd.src : pd.dst, pd.af,
5843                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5844                             tr->src.not);
5845                 if (tr->dst.addr.type == PF_ADDR_TABLE)
5846                         pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL ||
5847                             s->direction == dir) ? pd.dst : pd.src, pd.af,
5848                             pd.tot_len, dir == PF_OUT, r->action == PF_PASS,
5849                             tr->dst.not);
5850         }
5851
5852
5853         if (action == PF_SYNPROXY_DROP) {
5854                 m_freem(*m0);
5855                 *m0 = NULL;
5856                 action = PF_PASS;
5857         } else if (r->rt)
5858                 /* pf_route6 can free the mbuf causing *m0 to become NULL */
5859                 pf_route6(m0, r, dir, ifp, s);
5860
5861         return (action);
5862 }
5863 #endif /* INET6 */